Citation
Growth and modeling of III-V compound semiconductor optoelectronic materials with device applications

Material Information

Title:
Growth and modeling of III-V compound semiconductor optoelectronic materials with device applications
Series Title:
Growth and modeling of III-V compound semiconductor optoelectronic materials with device applications
Creator:
Howard, Arnold John,
Place of Publication:
Gainesville FL
Publisher:
University of Florida
Publication Date:

Subjects

Subjects / Keywords:
Crystals ( jstor )
Doping ( jstor )
Flow velocity ( jstor )
Hydrides ( jstor )
Hydrogen ( jstor )
Inlets ( jstor )
Room temperature ( jstor )
Semiconductors ( jstor )
Wavelengths ( jstor )
Zinc ( jstor )
City of Gainesville ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Arnold John Howard. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
025577103 ( alephbibnum )
24529515 ( oclc )

Downloads

This item has the following downloads:

oai.xml

UF00090193_00001.pdf

00006.txt

00064a.txt

UF00090193_00001_0140.txt

UF00090193_00001_0096.txt

00199.txt

UF00090193_00001_0141.txt

UF00090193_00001_0196.txt

UF00090193_00001_0239.txt

00206.txt

UF00090193_00001_0201.txt

00026.txt

00047.txt

00080.txt

UF00090193_00001_0013.txt

UF00090193_00001_0262.txt

UF00090193_00001_0011.txt

UF00090193_00001_0063.txt

00058.txt

UF00090193_00001_0202.txt

UF00090193_00001_0164.txt

UF00090193_00001_0195.txt

00105.txt

00060.txt

00054.txt

00092.txt

UF00090193_00001_0108.txt

UF00090193_00001_0074.txt

00205a.txt

UF00090193_00001_0211.txt

UF00090193_00001_0083.txt

00233.txt

UF00090193_00001_0040.txt

UF00090193_00001_0071.txt

UF00090193_00001_0233.txt

UF00090193_00001_0157.txt

00051.txt

UF00090193_00001_0118.txt

UF00090193_00001_0248.txt

UF00090193_00001_0210.txt

UF00090193_00001_0134.txt

UF00090193_00001_0150.txt

UF00090193_00001_0227.txt

UF00090193_00001_0128.txt

UF00090193_00001_0183.txt

00177.txt

UF00090193_00001_0156.txt

00231.txt

00252.txt

00055.txt

00061.txt

00153.txt

00162.txt

UF00090193_00001_0084.txt

00137.txt

00205.txt

00253.txt

00183.txt

UF00090193_00001_0213.txt

00067.txt

00142.txt

00181.txt

00237.txt

00037.txt

UF00090193_00001_0039.txt

00033.txt

00215.txt

00100.txt

00224.txt

00096.txt

UF00090193_00001_0064.txt

00145.txt

UF00090193_00001_0152.txt

UF00090193_00001_0113.txt

UF00090193_00001_0127.txt

UF00090193_00001_0222.txt

UF00090193_00001_0076.txt

UF00090193_00001_0166.txt

00108.txt

UF00090193_00001_0027.txt

00235a.txt

UF00090193_00001_0041.txt

00174.txt

UF00090193_00001_0145.txt

UF00090193_00001_0250.txt

UF00090193_00001_0116.txt

UF00090193_00001_0221.txt

00062.txt

UF00090193_00001_0090.txt

00105a.txt

00002.txt

UF00090193_00001_0269.txt

UF00090193_00001_0034.txt

UF00090193_00001_0271.txt

00112.txt

00146.txt

UF00090193_00001_0101.txt

00243.txt

UF00090193_00001_0160.txt

00076.txt

00057.txt

UF00090193_00001_0225.txt

UF00090193_00001_0189.txt

00148.txt

UF00090193_00001_0206.txt

00182.txt

UF00090193_00001_0138.txt

UF00090193_00001_0079.txt

UF00090193_00001_0162.txt

00158.txt

00087.txt

00066.txt

UF00090193_00001_0067.txt

UF00090193_00001_0060.txt

00109a.txt

00186.txt

00073.txt

00075.txt

UF00090193_00001_0257.txt

UF00090193_00001_0209.txt

UF00090193_00001_0178.txt

00194.txt

UF00090193_00001_0058.txt

UF00090193_00001_0030.txt

00007.txt

UF00090193_00001_0054.txt

UF00090193_00001_0175.txt

00127.txt

UF00090193_00001_0266.txt

00235.txt

00027.txt

00063.txt

UF00090193_00001_0208.txt

UF00090193_00001_0153.txt

UF00090193_00001_0144.txt

UF00090193_00001_0172.txt

UF00090193_00001_0115.txt

00114.txt

UF00090193_00001_0173.txt

00221.txt

00091.txt

00071.txt

00120.txt

UF00090193_00001_0053.txt

00059.txt

UF00090193_00001_0236.txt

00223.txt

UF00090193_00001_0146.txt

UF00090193_00001_0057.txt

UF00090193_00001_0018.txt

00136.txt

UF00090193_00001_0088.txt

00259.txt

UF00090193_00001_0169.txt

00150.txt

UF00090193_00001_0256.txt

UF00090193_00001_0133.txt

UF00090193_00001_0080.txt

UF00090193_00001_0028.txt

UF00090193_00001_0218.txt

00042.txt

UF00090193_00001_0261.txt

UF00090193_00001_0235.txt

00012.txt

UF00090193_00001_0253.txt

00201.txt

00156.txt

00125.txt

00023.txt

UF00090193_00001_0212.txt

UF00090193_00001_0252.txt

00167.txt

00039.txt

00218.txt

UF00090193_00001_0014.txt

00122.txt

UF00090193_00001_0165.txt

UF00090193_00001_0075.txt

00258.txt

00163.txt

00255.txt

UF00090193_00001_0207.txt

00256.txt

UF00090193_00001_0097.txt

00133.txt

00210.txt

UF00090193_00001_0098.txt

UF00090193_00001_0179.txt

00072.txt

UF00090193_00001_0155.txt

00081.txt

00020.txt

UF00090193_00001_0191.txt

UF00090193_00001_0023.txt

UF00090193_00001_0234.txt

UF00090193_00001_0119.txt

00038.txt

00213.txt

UF00090193_00001_0106.txt

UF00090193_00001_0120.txt

00250.txt

UF00090193_00001_0020.txt

00188.txt

UF00090193_00001_0019.txt

00179.txt

UF00090193_00001_0094.txt

00193.txt

00151.txt

00101.txt

00011.txt

00238.txt

UF00090193_00001_0085.txt

00190.txt

00047a.txt

00160.txt

00034.txt

00010.txt

UF00090193_00001_0033.txt

00083.txt

00157.txt

00143.txt

00024.txt

UF00090193_00001_0258.txt

00110.txt

UF00090193_00001_0042.txt

00093.txt

UF00090193_00001_0190.txt

00117.txt

00247.txt

00234.txt

UF00090193_00001_0251.txt

00152.txt

00184.txt

UF00090193_00001_0089.txt

00022.txt

00204.txt

00119.txt

00189.txt

00168.txt

00111.txt

00154.txt

UF00090193_00001_0046.txt

UF00090193_00001_0184.txt

UF00090193_00001_0228.txt

00248.txt

00207.txt

UF00090193_00001_0037.txt

UF00090193_00001_0091.txt

00019.txt

UF00090193_00001_0095.txt

00203.txt

UF00090193_00001_0255.txt

00251.txt

UF00090193_00001_0036.txt

00126.txt

00135.txt

UF00090193_00001_0232.txt

UF00090193_00001_0021.txt

UF00090193_00001_0237.txt

UF00090193_00001_0002.txt

00172.txt

00191.txt

UF00090193_00001_pdf.txt

00170.txt

00220.txt

00246.txt

UF00090193_00001_0010.txt

00169.txt

00070.txt

00032.txt

UF00090193_00001_0197.txt

UF00090193_00001_0187.txt

UF00090193_00001_0203.txt

UF00090193_00001_0254.txt

UF00090193_00001_0032.txt

UF00090193_00001_0230.txt

00138.txt

00068.txt

UF00090193_00001_0045.txt

UF00090193_00001_0176.txt

UF00090193_00001_0192.txt

00241.txt

00107.txt

UF00090193_00001_0226.txt

UF00090193_00001_0242.txt

00217.txt

UF00090193_00001_0214.txt

UF00090193_00001_0171.txt

00128.txt

UF00090193_00001_0026.txt

UF00090193_00001_0220.txt

00140.txt

UF00090193_00001_0114.txt

00212.txt

00064.txt

00008.txt

UF00090193_00001_0048.txt

UF00090193_00001_0167.txt

UF00090193_00001_0159.txt

00035.txt

UF00090193_00001_0259.txt

UF00090193_00001_0092.txt

00095.txt

00200.txt

UF00090193_00001_0047.txt

UF00090193_00001_0185.txt

00090.txt

00196.txt

00016.txt

UF00090193_00001_0245.txt

UF00090193_00001_0081.txt

UF00090193_00001_0038.txt

00222.txt

UF00090193_00001_0056.txt

00116.txt

UF00090193_00001_0112.txt

UF00090193_00001_0043.txt

00118.txt

00005.txt

00103.txt

00208.txt

UF00090193_00001_0050.txt

UF00090193_00001_0051.txt

00166.txt

UF00090193_00001_0136.txt

00197.txt

00063a.txt

00017.txt

UF00090193_00001_0008.txt

00139.txt

oai_xml.txt

UF00090193_00001_0124.txt

00178.txt

00097.txt

UF00090193_00001_0003.txt

UF00090193_00001_0149.txt

00050.txt

UF00090193_00001_0182.txt

00121.txt

UF00090193_00001_0130.txt

UF00090193_00001_0163.txt

UF00090193_00001_0199.txt

00085.txt

00195.txt

00018.txt

UF00090193_00001_0087.txt

00227.txt

UF00090193_00001_0109.txt

UF00090193_00001_0121.txt

00098.txt

UF00090193_00001_0131.txt

Copyright.txt

UF00090193_00001_0017.txt

00209.txt

00113.txt

00052.txt

00144.txt

00084.txt

00069.txt

00245.txt

00134.txt

00239.txt

00069a.txt

00004.txt

UF00090193_00001_0001.txt

UF00090193_00001_0198.txt

UF00090193_00001_0078.txt

UF00090193_00001_0205.txt

00088.txt

00187.txt

UF00090193_00001_0093.txt

UF00090193_00001_0137.txt

00240.txt

UF00090193_00001_0139.txt

UF00090193_00001_0082.txt

UF00090193_00001_0012.txt

00029.txt

00257.txt

00175.txt

00226.txt

UF00090193_00001_0186.txt

00074.txt

UF00090193_00001_0126.txt

00254.txt

UF00090193_00001_0135.txt

UF00090193_00001_0068.txt

00249.txt

UF00090193_00001_0059.txt

00132.txt

UF00090193_00001_0142.txt

UF00090193_00001_0241.txt

00077.txt

UF00090193_00001_0107.txt

UF00090193_00001_0125.txt

UF00090193_00001_0086.txt

00219.txt

00041.txt

UF00090193_00001_0007.txt

UF00090193_00001_0217.txt

00236.txt

00053.txt

UF00090193_00001_0219.txt

00164.txt

00198.txt

UF00090193_00001_0177.txt

00229.txt

UF00090193_00001_0272.txt

00104.txt

00185.txt

UF00090193_00001_0147.txt

UF00090193_00001_0015.txt

00115.txt

00078.txt

UF00090193_00001_0102.txt

00149.txt

00141.txt

UF00090193_00001_0052.txt

UF00090193_00001_0244.txt

00131.txt

00021.txt

00103a.txt

00028.txt

00216.txt

UF00090193_00001_0270.txt

00031.txt

00009.txt

00230.txt

UF00090193_00001_0070.txt

UF00090193_00001_0188.txt

00046.txt

UF00090193_00001_0103.txt

UF00090193_00001_0143.txt

UF00090193_00001_0154.txt

UF00090193_00001_0104.txt

UF00090193_00001_0151.txt

UF00090193_00001_0009.txt

00147.txt

UF00090193_00001_0168.txt

UF00090193_00001_0243.txt

00044.txt

00013.txt

UF00090193_00001_0065.txt

00228.txt

UF00090193_00001_0029.txt

UF00090193_00001_0267.txt

UF00090193_00001_0024.txt

UF00090193_00001_0246.txt

00001.txt

00109.txt

00225.txt

UF00090193_00001_0073.txt

00099.txt

UF00090193_00001_0049.txt

UF00090193_00001_0031.txt

00102.txt

UF00090193_00001_0240.txt

UF00090193_00001_0129.txt

00180.txt

UF00090193_00001_0066.txt

UF00090193_00001_0265.txt

00040.txt

00129.txt

UF00090193_00001_0216.txt

00094.txt

00159.txt

UF00090193_00001_0229.txt

00014.txt

UF00090193_00001_0194.txt

UF00090193_00001_0100.txt

00086.txt

UF00090193_00001_0006.txt

UF00090193_00001_0231.txt

UF00090193_00001_0077.txt

00242.txt

00232.txt

UF00090193_00001_0238.txt

00130.txt

UF00090193_00001_0170.txt

00049.txt

00079.txt

UF00090193_00001_0062.txt

UF00090193_00001_0161.txt

UF00090193_00001_0223.txt

UF00090193_00001_0110.txt

UF00090193_00001_0132.txt

UF00090193_00001_0016.txt

00048.txt

00165.txt

UF00090193_00001_0181.txt

00211.txt

00123.txt

00187a.txt

UF00090193_00001_0117.txt

00065.txt

00261.txt

00106.txt

UF00090193_00001_0200.txt

00214.txt

UF00090193_00001_0035.txt

00015.txt

UF00090193_00001_0260.txt

UF00090193_00001_0263.txt

00056.txt

UF00090193_00001_0247.txt

00192.txt

00045.txt

UF00090193_00001_0022.txt

UF00090193_00001_0180.txt

UF00090193_00001_0224.txt

00161.txt

00171.txt

UF00090193_00001_0005.txt

00176.txt

00173.txt

UF00090193_00001_0072.txt

00202.txt

UF00090193_00001_0105.txt

UF00090193_00001_0122.txt

00030.txt

UF00090193_00001_0158.txt

UF00090193_00001_0193.txt

00244.txt

UF00090193_00001_0148.txt

00089.txt

UF00090193_00001_0174.txt

UF00090193_00001_0061.txt

UF00090193_00001_0055.txt

UF00090193_00001_0204.txt

00082.txt

UF00090193_00001_0025.txt

UF00090193_00001_0268.txt

UF00090193_00001_0069.txt

UF00090193_00001_0099.txt

UF00090193_00001_0123.txt

00155.txt

00036.txt

UF00090193_00001_0004.txt

UF00090193_00001_0111.txt

00124.txt

UF00090193_00001_0249.txt

00260.txt

00043.txt

UF00090193_00001_0215.txt

UF00090193_00001_0044.txt

00025.txt

UF00090193_00001_0264.txt

00003.txt


Full Text











GROWTH AND MODELING OF III-V COMPOUND
SEMICONDUCTOR OPTOELECTRONIC MATERIALS
WITH DEVICE APPLICATIONS







By

ARNOLD JOHN HOWARD


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA


1990














ACKNOWLEDGEMENTS

The author would like to thank his advisor and committee

chairman, Dr. T. J. Anderson, for his guidance and

encouragement throughout this educational program. Also, many

thanks go to the other members of his graduate committee, Dr.

M. E. Orazem, Dr. G. Bosman, Dr. S. S. Li and Dr. A. J.

SpringThorpe, for their advice and useful discussions.

The author wishes to extend his deepest thanks to Dr.

Balu Pathangey, Dr. Yasuhiro Hayakawa, Mrs. Vesna Jovic and

Mr. Pete Axson for their friendship and assistance in the

construction and operation of the MOCVD system. Special

thanks are in order for Dr. Pathangey for his assistance with

the dopant modeling portion of this work. The author is also

indebted to Mr. Keith Rambo, Mr. Dean Heinz and Mrs. Cheryl

Heinz, Dr. Shiro Sakai and Mr. Whitey Herrlinger for their

assistance at the University of Florida Surge Area.

The author had the good fortune to have been a visitor at

Bell Northern Research (BNR) in Ottawa, Canada, and is

grateful for the hospitality and advice of Dr. A. J.

SpringThorpe, Dr. C. Blaauw, Dr. M. N. Svilans, Dr. C. J.

Miner and Dr. I. C. Bassignana and several other members of

the scientific staff at BNR. The author also gratefully








acknowledges the financial support of BNR, DARPA and

Microfabritech for portions of this study.

Warm, personal thanks go to the author's family,

especially his wife Robin, and his mother Mary, for their

moral support and patience throughout the course of his

graduate studies. This work is dedicated to them and also to

his father, the late Peter V. Howard.


iii













TABLE OF CONTENTS


ACKNOWLEDGEMENTS ................................. ...... ii

ABSTRACT................................................ vi

CHAPTERS

I INTRODUCTION....................................... 1

1.1 III-V Semiconductors........................... 1
1.2 Epitaxial Growth Techniques................... 6

II METAL ORGANIC CHEMICAL VAPOR DEPOSITION ........... 13

2.1 A Brief History of MOCVD..................... 13
2.2 MOCVD Systems................................ 15
2.3 A Review of the Literature on InP Based MOCVD 22
2.3.1 InP Homoepitaxy....................... 22
2.3.2 GaInAs/InP ...... ..................... 30
2.3.3 GaInAsP/InP.......................... 33
2.3.4 InP Based Devices...................... 37
2.4 A Description of the MOCVD System. ........... 40
2.4.1 Introduction......................... 40
2.4.2 Gas Delivery System................... 43
2.4.3 Reactor and Heating System............ 47
2.4.4 Exhaust/Scrubbing System.............. 50
2.4.5 Safety ................................ 51
2.5 Determination of Optimum Growth Conditions
Based on Thin Film Characterization .......... 52
2.5.1 Experimental Approach................. 52
2.5.2 InP................................... 56
2.5.3 Growth of GaInAs Lattice-Matched to
InP .............................. .. 90
2.5.4 Growth of GaInAsP Lattice-Matched to
InP ................................... 108

III P-TYPE DOPING OF MOCVD INP: EXPERIMENTS AND
MODELING ......................................... 118

3.1 A Review of the Literature on p-Type Doping
of InP........................................ 118
3.1.1 Introduction.......................... 118
3.1.2 Bulk Crystal Growth................... 119
iv








3.1.3 Liquid Phase Epitaxy.................. 121
3.1.4 Molecular Beam Epitaxy................ 122
3.1.5 Chemical Vapor Deposition ............. 123
3.2 MOCVD Growth and Characterization of Mg-Doped
InP Using bis-(Methylcyclopentadienyl)
Magnesium as a Dopant Source................ 130
3.2.1 Introduction.......................... 130
3.2.2 MOCVD Growth.......................... 130
3.2.3 Results and Discussion ................ 131
3.2.4 Conclusions........................... 145
3.3 Experimental DMCd and DEZn for p-Type Doping
of InP by MOCVD.............................. 145
3.3.1 Introduction.......................... 145
3.3.2 DMCd Results.......................... 146
3.3.3 DEZn Results.......................... 148
3.4 Modeling of p-Type Doping of InP using DEZn.. 156
3.4.1 Introduction.......................... 156
3.4.2 Point Defect Structure................ 164
3.4.3 Discussion of Results................. 173

IV EPITAXIALLY GROWN INTERFERENCE FILTERS............ 177

4.1 Theory of Interference Filters............... 177
4.2 MBE Grown AlGaAs/GaAs Devices ................ 182
4.2.1 Introduction.......................... 182
4.2.2 Electrical Theory..................... 184
4.2.3 Experimental.......................... 185
4.2.4 Electrical Testing.................... 193
4.2.5 Conclusions........................... 204
4.2.6 Addendum.............................. 205
4.3 MOCVD Grown GaInAsP Devices ................... 209

V CONCLUSIONS AND RECOMMENDATIONS ................... 216

APPENDICES

A THEORY AND OPERATION OF THE LOW TEMPERATURE HALL
EFFECT SYSTEM .................................... 220

B OPERATION OF THE QUATERNARY MOCVD SYSTEM.......... 234

C DOPANT MODEL COMPUTER PROGRAM..................... 248

REFERENCES ............................................. 252

BIOGRAPHICALSKETCH....................................262














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy

GROWTH AND MODELING OF III-V COMPOUND
SEMICONDUCTOR OPTOELECTRONIC MATERIALS
WITH DEVICE APPLICATIONS

By

ARNOLD JOHN HOWARD

December 1990

Chairperson: Timothy James Anderson
Major Department: Chemical Engineering

Several topics have been undertaken during the course of

this degree which are associated with understanding and

improving semiconductor processing. The growth, modeling and

characterization of III-V compound semiconductor materials and

optoelectronic devices has been emphasized. Epitaxial layers

of GaxIn1.xAsyP1.y with lattice-matched alloy compositions over

the range from x=0, y=0 (InP) to x=0.47, y=1 (Ga47In53As)

have been grown by metal organic chemical vapor deposition

(MOCVD) on InP substrates. Both the MOCVD system, used to

grow these layers, and a low temperature Hall effect system,

used to characterize these layers, were designed and

installed. The results from several other analytical

techniques were used to determine the optimal growth

conditions for high quality epitaxial layers.








The use of diethylzinc (DEZn), bis-(methylcyclo-

pentadienyl) magnesium (MCp2Mg) and dimethylcadmium (DMCd) as

p-type dopant sources for MOCVD InP was investigated at BNR in

Ottawa, Canada. It has been experimentally observed that the

carrier concentration dependence on dopant partial pressure in

the MOCVD reactor is different for each of these three

dopants. A novel model of the p-doping process of MOCVD InP

using DEZn has been developed that incorporates an equilibrium

boundary condition between the gas phase and solid phase point

defects. The results of this model indicate that at high DEZn

gas phase mole fractions, which results in low solid-phase

electrical activity, the dominant electrically inactive point

defects are intersticial zinc and zinc completed with a

phosphorous divacancy.

A novel optoelectronic device has been fabricated and

modeled which contains p-n heterojunctions in an optical

interference filter. Structures were grown by molecular beam

epitaxy at BNR using the GaAs/AlGaAs material system and by

MOCVD at the University of Florida using the InP/GaInAsP

material system. Structures with peak reflectivities at 1.3

and 1.40 microns were grown and good crystalline quality were

confirmed. Electrical bistability was observed in a forty-

layer device which has never been reported before in a

structure of this size.


vii













CHAPTER I
INTRODUCTION


1.1 III-V Semiconductors


Since the invention of the transistor in 1948 by

Shockley, Brattain, and Bardeen, there has been a revolution

in the electronics industry. Up to that time the vacuum tube

diode and triode were the most used electronic devices, but

then the transistor device using a semiconductor crystal as

its starting material was fabricated. The microchip, which is

the fundamental building block of present day computers,

contains a large number of tiny semiconductor transistors

using typically single crystals of silicon as a starting

material. Silicon has been the "workhorse" for the

electronics industry primarily due to its availability in high

single crystalline purity, ease of use in device fabrication,

and of course its good electrical properties. But, the

relatively low electron mobility and fixed indirect bandgap of

silicon makes it not suitable for present-day optoelectronic

device applications. As a consequence of these new demands,

research into the development of semiconductors with variable

electrical and optical properties has flourished.

Compound semiconductors such as GaAs, InP and others

composed of elements from group IIIA (Al, Ga, In) and group VA

1








2

(P, As, Sb) columns of the periodic table have electrical and

optical properties superior to those of silicon for certain

modern-day device applications. III-V materials have a wide

range of bandgap energies (0.18 to 2.4 eV), where the bandgap

energy is defined as the energy difference between the lowest

electron state in the conduction band and the highest hole

state in the valence band allowed in the semiconductor. Some

compound semiconductors have direct bandgaps, meaning that the

conversion of photons (light) to electrons (energy) or vice

versa, does not involve a third particle, such as a phonon.

The direct bandgap III-V compounds also have large electron

mobilities where mobility is defined (at low electric fields)

as the ratio of absolute electron velocity to the magnitude of

the electric field. A listing of these parameters and the

lattice constants of silicon and binary III-V semiconductors

is shown in Table 1[1]. As shown in this table, as much as a

two order of magnitude increase in electron mobility is

possible by using III-V compound semiconductors instead of

silicon for electronic devices. It is also significant to

note that a wide range of compound semiconductors can be

formed by creating solid solutions of the individual

semiconductors; hence, a wide selection of compound

semiconductors exists with a wide range of electrical and

optical properties.










Table 1

Properties of Silicon and III-V Binary
Semiconductors at 300 K


Bandgap Bandgap Electron Lattice
Type Energy(eV) Mobility Constant
(cm2/V-s) (Angstroms)
Si indirect 1.12 1350 5.43
InSb direct 0.18 100000 6.48
InAs direct 0.36 22600 6.06
GaSb direct 0.70 5000 6.09
InP direct 1.28 4000 5.87
GaAs direct 1.43 8500 5.65
AlSb indirect 1.60 200 6.14
AlAs indirect 2.16 180 5.66
GaP indirect 2.26 300 5.45
AlP indirect 2.45 80 5.46


Source: Streetman[1].








4

Since III-V compound semiconductors present a wide range

in values of direct bandgap energy, mobility, and lattice

constant, semiconductor devices have wider ranges of

application. The bandgap energy (Eg) of a semiconductor is

related to the cut-off wavelength (Ag) by the following

equation: Eg(eV) = 1.24/Ag (gm). The cut-off wavelength of a

semiconductor is the longest wavelength to which a detector

fabricated from this same semiconductor will respond. Another

degree of freedom available is the ability to form completely

miscible substitutional solid solutions independently on both

the group III and group V sublattices. In other words, not

only simple binary III-V compounds, but also III-III'-V or

III-V-V' ternary and III-III'-V-V' quaternary single

crystalline semiconductors such as AlxGa1. As, GaxInl-xAs,

GaASyP,1. and GaxIn_1.As P1. can be created. By using ternary

and quaternary semiconductors, it is possible to vary the

physical and electrical properties of these materials

continuously between the property limits of the constituent

binary compounds listed in Table 1. A plot of the lattice

constant versus the bandgap energy (at 300 K) for III-V

compound semiconductors is shown in Figure 1. Solid dots

indicate binary compounds, solid lines connecting dots

represent direct bandgap ternary solid solutions, and dashed

lines connecting dots represent indirect bandgap ternary solid

solutions.








5



6.6

In b


6.4 i




6.2 i
SA1 Sb
S\ GaSb
S \InAs

6.0

I-n


S5.8-\-


SA1As
GaAs
5.6 --


GaPS- -Al P

5.4 I

0 0.5 1.0 1.5 2.0 2.5

Bandgap (eV)




Figure 1: Lattice parameter and bandgap energy of various
III-V semiconductors








6

Basically, the entire area bounded by the solid and dashed

lines is available for use in the design of new III-V compound

semiconductor devices.

The cross-hatched area shown in Figure 1 is the lattice

parameter-bandgap energy space of the quaternary material

GaxInl.xAsyP1.y. This material has many optoelectronic device

applications due to its wide range of bandgap energy (0.36 to

2.26 eV) and possible lattice constants. Most semiconducting

single crystalline ternary and quaternary materials are

epitaxially grown on a substrate of nearly the same lattice

constant. Hence, the two compositional degrees of freedom

available with the GaInAsP system are important because

presently only GaAs, GaSb, GaP, InP, InAs and InSb are

available for use as substrate materials. For GaInAsP on InP,

or any other heteroepitaxial materials system, a difference of

lattice constant (lattice-mismatch) of greater than 0.1%

between the grown film and substrate leads to, for film

thicknesses greater than the critical thickness, the formation

of structural defects which can degrade device performance.

This problem puts strict demands on the epitaxial growth

technique employed.

1.2 Epitaxial Growth Techniques

The word "epitaxy" is derived from Greek and means

"arranged upon." Epitaxial films of III-V materials are

usually grown or arranged upon substrates with equivalent

crystalline structure and lattice constant. The two most








7

common substrate materials used today for III-V homoepitaxy,

(i.e., growth on a crystal of the same composition), or

heteroepitaxy, (i.e., growth on a crystal of different

composition or crystal structure), are GaAs and InP. The

substrates are cut from bulk crystals along a particular

crystal orientation from two to four inch diameter boules

which are sometimes created by withdrawing a seed crystal from

a heated liquid melt. Epitaxial growth on these substrates is

accomplished by exposing the heated surface to a flux of group

IIIA metals and group VA non-metals. The flux can be supplied

from a liquid, vapor, or molecular beam source. This

distinction defines the three primary methods for growing

epitaxial III-V films: liquid phase epitaxy (LPE), vapor

phase epitaxy (VPE), (also known as chemical vapor deposition

(CVD)), and molecular beam epitaxy (MBE). Each technique has

its own advantages and drawbacks, which will be discussed in

the following paragraphs.

LPE is a growth technique which can be used to deposit

thin single crystal layers of III-V compound semiconductors

from a heated liquid solution by decreasing the temperature of

the substrate relative to the solution. It is a relatively

simple, inexpensive, near equilibrium (reproducible) growth

technique that is well understood. The growth rate can be

high and a wide range of both p- and n-type dopants are

available and their incorporation is controllable. LPE has

been used to grow InP[2] and high quality GaInAs[3] on InP for








8

laser applications. However, problems such as surface

defects, poor thickness and compositional uniformity, and

difficulty in growing abrupt heterojunctions have made the LPE

technique unsuitable for present-day device fabrication

demands. For simple layer structures such as the AlGaAs solid

state laser used in compact disc players, LPE is perfectly

adequate.

There are three distinct VPE or CVD chemistries: chloride

CVD, hydride CVD and metal organic CVD. The chloride (or

sometimes referred to as halide) CVD process for GaAs growth,

as an example, uses AsC13 and metallic Gallium as sources in

an open tube system with H2 (as a carrier gas) to transport

reactants from the source zone, through a temperature gradient

zone to the deposition zone. The chloride CVD process is a

surface-kinetically limited process requiring careful source

composition control and accurate temperature control

throughout the system for reproducibility. Also, it is

difficult to vary the V/III ratio and transients are long so

abruptness is bad in chloride CVD. GaInAsP has been grown by

the chloride CVD method[4] but other CVD techniques are more

convenient and flexible for growing ternary and quaternary

III-V compounds. Hence, the chloride process is usually only

used to grow high purity epitaxial GaAs.

The hydride CVD process for growth of III-V compound

semiconductors differs from the chloride process by replacing

column V chlorides such AsC13 or PCl3 with column V hydrides







9

like AsH3 or PH3. For InP growth, HC1 gas is first reacted

with liquid indium metal in the source zone. The gaseous

product InCl is then carried by H2 to mix and react with PH3

to deposit InP in the growth zone. Similar to the chloride

process, accurate temperature control is required for this

three-zone process which is also surface-kinetically limited

in the low-temperature growth regime. The hydride process is

currently widely used for light-emitting diode (LED)

applications using GaAs1l.Px. It has also found application

in the growth of III-V GaInAsP and GaInAs for LED's, lasers

and detectors[5]. One major advantage the hydride system

provides over the chloride system is the ability to vary the

vapor phase V/III molar ratio by adjusting the inlet flow

rates of hydrides and HC1. One drawback of both the hydride

and chloride systems is that they are hot-wall systems;

interaction between the gas and heated SiO2 reactor wall

occurs which results in unintentional silicon incorporation

into grown layers. Due to its successful use, especially in

LED fabrication, hydride CVD will continue to have a

significant role in the growth of III-V materials.

The third type of CVD or VPE process is metal organic

chemical vapor deposition (MOCVD). The MOCVD process involves

an irreversible pyrolysis reaction of vapor-phase mixtures

usually of group IIIA metal organic sources and group VA

hydride sources. For InP, as an example, trimethylindium

(TMIn) and PH3 diluted in H2 would flow into an open cold-wall







10

quartz tube, decompose in the presence of a heated substrate,

and then deposit an epitaxial layer. Under normal deposition

conditions, the MOCVD process is kinetically limited by mass

transport of the column III source through a stagnant layer

near the growing surface. The MOCVD process is capable of

growing a wide variety of films with excellent abruptness

uniformity over large substrate areas. The principal device

area where MOCVD has made an impact is optoelectronics. A

thorough review of the MOCVD literature has been written by

Ludowise[6] and a brief history of MOCVD with emphasis on InP

based materials and device applications is presented in

section 2.1.

Molecular beam epitaxy (MBE) is a technique capable of

growing epitaxial films one atomic layer at a time. MBE makes

use of controlled evaporation from one or more thermal sources

to direct beams of atoms or molecules onto a heated substrate

under ultra-high vacuum conditions. During a MBE growth the

substrate temperature is generally kept relatively low (500-

6000C for GaAs). MBE growth rates are typically slow (0.1 -

2iLm/hr) which in combination with low growth temperatures

permits precise layer thickness, doping and compositional

control[7]. For GaAs, the As4 beam flux is much greater than

the Ga beam flux, and both fluxes are dependent upon the

temperature of the effusion oven, molecular weight of the

emitted atom, orifice area, and source cell to wafer distance.

With a properly placed two-inch rotating wafer, nonuniformity








11

of the growth film may be reduced to less than ten percent.

Numerous devices such as LEDs, lasers, FETs and HBTs have been

fabricated with MBE. One unique bonus with MBE is that in-

situ monitoring devices such as RHEED, mass spectrometers,

Auger spectrometers and ion gauges are feasible and

commonplace. However, large expense and limited throughput

restrict MBE usage.

During the last decade, several novel deposition

techniques for III-V compound semiconductors have emerged.

Each one is a spin-off of either MOCVD, MBE, or a combination

of the two and has some relative advantages and disadvantages;

none of these new techniques are widely used. One such

technique is called atomic layer epitaxy or ALE. For III-V

compound semiconductors, ALE proceeds by depositing a

monolayer of a group III metal followed by depositing a

monolayer of group V atoms, in a layer-by-layer sequence[8].

In ALE, grown layer thickness is determined by the number of

cycles rather than the time of growth. Another relatively new

technique is called chemical beam epitaxy (CBE) or metal

organic molecular beam epitaxy (MOMBE). In MOMBE, all of the

group IIIA and VA sources are metal organic; TMAs replaces As4

and AsH3, TEP replaces P2 and PH3[9]. The remaining aspects

of MOMBE are essentially the same as conventional MBE. A

final technique which is similar to ALE but is performed in a

conventional low pressure MOCVD system is called flow

modulation epitaxy or FME. FME has been used to grow InP at








12

temperatures as low as 3300C by alternately pulsing PH3 and

TMIn with intervening H2 purge steps into the reactor with

each step lasting on the order of several seconds[10]. So, it

is evident that there are a wide variety of epitaxial growth

techniques for III-V compound semiconductors. Most of the

experimental work presented in this dissertation made use of

a low pressure MOCVD system. One study made use of a MBE

system for a GaAs/AlGaAs device, but the rest of the

literature review will focus on the use of MOCVD for growth of

GaInAsP/InP materials for optoelectronic device applications.













CHAPTER II
METAL ORGANIC CHEMICAL VAPOR DEPOSITION


2.1 A Brief History of MOCVD


MOCVD refers to the deposition of multiconstituent films

using one or more metal organic compounds as sources, and the

term was originated by Manasevit[11]. He demonstrated that

single crystalline GaAs could be deposited using TEGa and AsH3

in an open tube reactor. Shortly after this report, it was

discovered that by mixing metal organic and hydrides of

different elements, binary and ternary III-V compounds such as

GaP, GaAsP, GaAsSb, A1N, GaN, InAs, GaInAs, InAsP, and InP

could be formed in a manner similar to GaAs[12-16]. Practical

information was also reported in these early efforts such as

the observation that GaAs film growth rate was mass transport

limited by the metal organic group III source and independent

of temperature below 8000C. Also, n-type doping using H2S,

H2Se and p-type doping using DEZn and DMCd were achieved.

Soon after the demonstration of high quality MOCVD grown

material, all MOCVD grown devices such as FETs,

photocathodes[17,18] and GaAs/AlGaAs DH laser diodes[19] were

reported. These milestones resulted in a rapid increase in

MOCVD research and development, and as a result, it has been

demonstrated that MOCVD has the capability of growing a wide

13








14

variety of device quality III-V compound semiconducting

materials.

The MOCVD process has been used for the epitaxy of most

III-V compound semiconductors. The basic overall reaction is

M11,(Alk)3(g) + XVH3(g) ----> M,11X(s) + 2 Alkane(g). In this

reaction, the organometallic and the hydride typically are

irreversibly pyrolyzed by the heat of the susceptor and

substrate to form molecular (or atomic) intermediates which

may react in the gas phase or on the substrate surface[20].

Some researchers say that the breaking of metal-carbon bonds

occur on the semiconductor surface[21]. Contradicting reports

such as these may be related to differences in test conditions

or configurations from team-to-team, but the only conclusion

that can be made is that the complete MOCVD process is not

well understood at this time.

Each step of the MOCVD process is not known, but there

are several practical trends that are generally agreed upon.

Some trends for GaAs and InP are: (1) III-V MOCVD growth

between 550 and 7500C is mass transport limited in the column

III source (below 5500C growth is reaction limited); hence

growth rates are determined by metal organic fluxes and are

temperature independent; (2) for ternary and quaternary films,

group III solid phase compositions are linearly related to gas

phase group III compositions, however, solid phase group V

compositions have a non-linear dependence on gas phase group

V compositions due to large differences in cracking








15

temperatures (PH3 is more difficult to decompose than AsH3);

(3) gas phase V/III ratios have a strong effect on background

carrier concentrations and p- and n-type doping levels; (4) in

some cases, group III metal organic react with group V

hydrides to form adducts and polymers that may be involatile

liquids or solids; and (5) group II alkyls act as p-type

dopant sources, group VI hydrides act as n-type dopant

sources, but depending on growth conditions, group IV hydrides

can act as donor or acceptor sources in III-V's. The extent

to which these trends apply to all III-V materials does vary,

but they are definitely useful in designing an MOCVD system or

in optimizing material specific growth conditions.

2.2 MOCVD Systems

Most MOCVD systems use quartz reactors oriented either

vertically[ll] where gas flow is usually down, or horizontally

[22] where gas flow is usually over a wedge shaped susceptor.

Other less common systems incorporate barrel reactors[23]

where growth can occur on multiple wafers with reactants

flowing from top to bottom, or "chimney" reactors[24] where

gas flow is up and wafers and susceptor are held vertically.

In a vertical reactor, uniform growth rates are more difficult

to achieve than in a horizontal reactor because with the

geometry of a vertical reactor, nonlaminar, turbulent gas flow

can more easily occur. In a horizontal reactor, a boundary

layer zone forms above the susceptor where the flow rate is

lower than that of the bulk gas above. The stagnant layer








16

thickness can increase along a flat susceptor in the gas flow

direction which in combination with the depletion of reactants

due to deposition, can result in nonuniform grown layer

thicknesses. This variation can be nullified by either

tilting the susceptor at an angle of 5 to 100 or inserting a

baffle into the reactor at an angle positioned above the

horizontal susceptor. This gradually reduces the cross-

sectional area in the reactor above the wafer resulting in a

gradual increase in linear gas velocity causing the boundary

layer to have a uniform thickness profile. With a constant

boundary layer thickness and constant diffusion coefficient of

group III source molecule, the flux of the mass transfer

limited reactant to the growing surface will be constant. As

a result of this, uniform semiconductor films can be grown

over large area substrates.

When MOCVD growth of mixed crystals involves the use of

more than one group III compound, GaInAs for example, solid

phase compositional non-uniformity can result. This can

result from concentration gradients in the gas flow direction

due to slight differences in the magnitude of the diffusion

coefficients. One way to avoid this problem is by selecting

group III metal organic sources with similar molecular weights

and correspondingly similar diffusion coefficients. Another

way of avoiding this problem is by using low pressure (= 0.1

atm) operation instead of atmospheric pressure. At reduced

pressures, the linear gas velocity increases and accordingly








17

the stagnant layer thickness decreases. Also, at lower

pressures, diffusion coefficients increase, making more abrupt

heterojunctions possible. Low pressure operation also reduces

the occurrence probability of unwanted parasitic reactions.

Because of all of these advantages, low pressure MOCVD growth

has become widely used especially for multi-wafer scale-up

production applications.

Another problem with the traditional horizontal "Bass

type"[17] reactor is due to the fact that having a cold dense

gas above a hot, less dense gas is unstable because of

gravity. This can cause natural convection which results in

closed stream-line gas flow patterns. This problem can

usually be minimized by operating at reduced pressures[25].

The best solution, however, appears to be the use of an

inverted reactor geometry[26] which completely eliminates

thermal buoyancy effects. In this geometry, the susceptor is

located at the highest and hottest point of a horizontal

reactor with the wafer mounted upon it facing downwards.

Another obvious benefit with this design is the elimination of

the problem of particles falling on the substrate before and

during epitaxial growth which can lead to structural defects.

With this design, improved GaInAs compositional uniformity and

a complete elimination of parasitic deposition on the quartz

wall opposite the susceptor have also been reported[26].

Another final technique that has been used to improve both

thickness and compositional uniformity is the use of moving








18

substrate holders in both circular and even planetary motion

configurations[27]. These techniques can greatly improve

uniformity, but also, unfortunately, greatly add to machine

complexity and expense.

Aside from the reactor, the other major parts of an MOCVD

system are the gas delivery, heating exhaust/scrubbing and

safety systems. Most gas handling systems are constructed

from high purity stainless steel tubing, valves (air-operated

and manual), regulators, electronic mass flow controllers and

filters or purifiers. The system typically delivers metal

organic (from bubblers), hydrides (from high pressure gas

cylinders) and most often hydrogen (from a palladium-alloy

purifier) to a fast switching manifold which directs gases to

the reactor or to a vent line. Gas manifolds should be

situated as close to the reactor inlet as possible to minimize

tube length and improve interface abruptness capabilities.

Most systems use manifolds with a linear valve arrangement,

but only in a radial manifold arrangement is the length from

each valve to the reactor potentially the same for all of the

gases[28]. In the "vent/run" type system discussed above, the

vent line and reactor line are "pressure balanced" so that

transient times associated with gases adjusting to and flowing

from a high pressure to a low pressure line, or vice versa,

can be eliminated.

For heating systems, most MOCVD reactors are heated by

inductively coupling RF power to a graphite susceptor. This







19

is ideal because it is a non-contact method, it selectively

heats only the graphite, and it is easy to configure by

arranging a copper coil around the susceptor portion of the

reactor. The RF generator size required depends on susceptor

size, gas velocity, coupling efficiency, and reactor wall

cooling mechanism. Infrared heating from quartz-halogen lamps

has also been used but, as wall deposition increases, non-

uniform heating may occur. A resistance heater embedded into

the graphite is another option but deposition on electrical

feedthroughs complicates reactor cleaning. Most heating

systems use an embedded thermocouple feedback system to

control temperature. Optical pyrometers have also been used

but wall deposition can result in false readings.

After passing through the heated zone of the MOCVD

reactor, some of the toxic gas sources still remain uncracked

and undeposited. These gases have to be neutralized before

being discharged into the atmosphere. Hazardous gases such as

AsH3, PH3 and SiH4 are commonly used in the MOCVD of III-V

semiconductors and are difficult to neutralize or "scrub"

especially when they are used in combination. There are four

different types of scrubbing systems commercially available;

depending on the application one alone may be inadequate. The

four types are liquid scrubbers, thermal crackers, dry powder

scrubbers, and incinerators. Liquid based scrubbers are most

commonly used and work by bubbling the toxic gas through a

basic (pH > 10) solution of sodium hypochlorite and sodium








20

hydroxide diluted in water where neutralized salts and acids

are products. Some gases are nearly insoluble in water,

though. Thermal crackers basically operate by heating the

exhaust stream to approximately 9500C to thermally decompose

toxic gases into less toxic compounds. Clogging and also

insufficient heat transfer at high flow rates are problems

with this technique. Dry scrubbers use powders such as

activated carbon or diatomaceous earth mixed with iron

chlorides to react with the toxic gases. This technique also

has problems associated with efficient gas solid contacting

and disposal of toxic corrosive powders. A final technique,

the incinerator or "burn box," operates in such a way that

gases are mixed with a fuel gas and oxygen and then ignited by

a pilot flame or electric igniter[29]. All of the four

scrubbing techniques have their individual problems. The

scrubbing system should ideally be a combination of two or

more of the individual systems in case one system fails.

As discussed, MOCVD of III-V compound semiconductors

presently involves the use of highly toxic and explosive

source gases. There has been some work on the use of less

toxic sources for MOCVD such as tertiarybutylphosphine (TBP)

[30] instead of phosphine (PH3), (the threshold limit value

(TLV) of PH3 is 0.3ppm while that of TBP is greater than

1000ppm), but the material grown with these new sources is

generally inferior. In any event, there must always be an

integrated safety component to all MOCVD systems.








21

Several papers have been published on the important topic

of MOCVD safety[31,32]. These papers are very useful when

designing the layout of an MOCVD machine and laboratory.

First, all MOCVD systems must have both toxic gas and hydrogen

sensors in and around them connected to an alarm. These

sensors must be capable of shutting down the machine in the

event of a detected leak. In order to shut down quickly, all

gas lines must be equipped with normally closed air-operated

valves. Toxic gas lines should also be double contained

(which is extremely expensive) and equipped with pressure

sensors that sound an alarm for abnormally high pressures.

Compressed air, hydrogen and nitrogen lines should have

sensors for abnormally low pressure which can, if activated,

shut the machine down. The reactor and pump exhaust line

should also have similar pressure sensors. Of course, smoke,

fire, and cooling water flow detectors and sensors are

necessary. Other things such as micro-switches on panel doors

and available supplied air masks are also required. Finally,

SCBAs and trained users should always be available outside of

the facility for emergencies. All of the above mentioned

safety issues/design features are important for the design and

operation of a modern MOCVD system. Of course, the most

important thing for lab safety is to provide adequate operator

training focusing on the nature of the toxic sources and how

to treat them.








22

2.3 A Review of the Literature on InP Based MOCVD

2.3.1 InP Homoepitaxv

The first reported growth of indium phosphide (InP) by

MOCVD was in 1969 by Manasevit and Simpson[12]. This work as

well as other early efforts[33-35] used triethyl indium (TEI)

and phosphine (PH3) as indium and phosphorus sources. For

several years, problems such as low, nonuniform growth rates

and high impurity levels were encountered. One source of

these problems was found by relating the observation of a

white smoke at certain growth conditions to the uncontrolled

gas phase reaction between metal organic indium sources and

phosphine. This reaction occurs at low temperatures <1000C,

is parasitic in nature, and produces a non-volatile liquid

polymer.

One method used to minimize this problem was the use of

low pressure reactor systems[36-38] to decrease the residence

time of unreacted species upstream of the growth region.

Another method involves the use of adducts such as TMI-TMP

[39] or TMIn-TEP[40] as indium sources which will not complex

with PH3. Another technique is to keep the reactants apart

and only let them mix just prior to the growth region. This

method can, however, lead to uniformity problems. The most

recent improvement is the use of TMIn (a solid powder at room

temperature which melts at 880C) instead of TEIn (a liquid at

room temperature), as the indium source[41-43]. TMIn also

decomposes at a much higher temperature (>3000C) than TEIn







23

(<1000C) and therefore is less likely to react upstream of the

heated growth zone. Also, TMIn has a much higher vapor

pressure than TEIn which is experimentally convenient because

heated gas lines would no longer be necessary.

Other techniques such as the use of hydrogen-nitrogen

mixtures as the carrier gas[41] and phosphine pre-crackers[44]

have been tried with varying degrees of success and merit. A

final conclusion is that proper reactor geometry, system

design, and growth conditions are very important for avoiding

parasitic gas phase reactions and obtaining superior InP thin

film quality. Currently, the reaction at certain growth

conditions of TMIn and PH3 in a properly designed MOCVD system

can yield uniform, high quality epitaxial InP with no evidence

of indium prereaction problems.

The deposition of high quality layers of InP for device

applications requires precise control of their unintentionally

introduced (undoped) and intentionally introduced (both p- and

n-type doped) impurity concentrations. An important condition

for obtaining reproducible p- and n-type doping levels is the

ability to grow undoped material with a reproducibly low

background carrier concentration. To obtain low background

levels, one needs a contamination free MOCVD system equipped

with high purity sources and optimized growth conditions.

Most conventional MOCVD systems are constructed from

ultra-high purity components such as electropolished welded

316 stainless steel and semiconductor-grade low-sodium content







24

quartzware. Coupled with the use of palladium-alloy diffused

hydrogen as a carrier gas, these precautions usually eliminate

the system as a source of high background impurity level

problems. In addition to high purity equipment, ultra high

purity sources contained in stainless steel bubblers and

corrosion resistant coated cylinders are required. For InP,

phosphine with five nines purity (99.999%) and diphos purified

(doubly sublimed) trimethylindium are both commercially

available. As purification technologies advance, then

progressively lower background doping levels surely will

follow.

The most important material's issues in the InP growth

area are the effect of growth conditions, substrate quality,

substrate orientation and substrate wafer cleaning techniques

on material quality. Several papers have been published on

each topic and the basis of comparison presented usually

involves characterization results of thin films such as room

temperature (300 K) and/or liquid nitrogen temperature (77 K)

mobilities and undoped carrier concentrations (ND-NA) cm3, etch

pit densities (EPD), photoluminescence (PL) intensities and

occasionally device performances.

The effect of growth conditions on properties of InP

grown by MOCVD has been studied by several research teams[45-

48]. Razeghi and Duchemin[45] showed that the growth rate of

InP is linearly dependent on indium metal organic reactor

partial pressure and independent of the phosphine partial







25

pressure. They also compared the growth rate of undoped InP

on (100), (111), and (115) InP oriented substrates and they

reported excellent film quality on (100) 20 towards (110) and

(115) 20 towards (111). Eguchi et al.[46] studied the effect

of V/III (phosphine to metal organic indium) ratio on EPD and

electrical properties and reported superior material at high

V/III ratios (>300). This result was in agreement with

Kasemset's[47] earlier work on both V/III ratio and growth

temperature effects. However, a survey of the effect of

growth temperature on layer quality is less conclusive.

Kasemset[47] indicates that a decrease in background carrier

concentration results upon increasing growth temperature,

while Scott et al.[48] report the opposite trend. This

discrepancy is probably due to different dominant impurities

in each group's TMIn source with correspondingly different

incorporation mechanisms. Most teams report high quality

MOCVD InP grown at temperatures between 5500C and 6750C.

Below 5500C, growth rates drop and material quality degrades.

Above 7000C, background carrier concentrations increase.

Presently, high quality two inch diameter wafers of InP

are commercially available as both doped (p and n-type) and

semi-insulating. Variation of results from team to team in

early research efforts and even today may be due in part to

the lack of reproducibility of substrate properties from batch

to batch and vendor to vendor. A recent paper from Knight et

al.[49] reports this problem. They observed a correlation







26

between leakage current of p-i-n InP based photodiodes and

substrate quality. Consequently, they set up a nondestructive

PL wafer-mapping system to evaluate grown film quality before

investing further processing time. Non-destructive techniques

such as PL mapping will remain essential unless wafer quality

control improves.

Proper wafer cleaning is also very important for the

growth of high-quality InP. Tuck and Baker[50] in 1973

published work on the chemical etching of (111) and (100) InP.

They compared the merits and disadvantages of using the

following four etching solutions: (1) 1HCL:1HNO3; (2)concen-

trated HC1; (3) 0.4N Fe3+; and (4) 1% bromine in methanol,

based on etching rate and hillock delineation. Nishitani and

Kotani[51] presented the use of H202-H2SO4-H20 solutions for

etching (100) and (111) oriented InP. Recently, studies have

been reported using sulphur to chemically passivate the

surfaces of InP and GaAs[52-53]. The goal of this work is to

reduce the substrate surface recombination velocity in order

to improve device performance. Another interesting study

compared several wafer cleaning methods using the surface

science techniques ISS, ESCA and AES[54]. This report states

that using a 5:1:1 mixture of H2SO4:H202:H20 in combination

with solvent degreasing step yields an InP surface with the

least amount of absorbed carbon and oxygen relative to the

other methods tested. Most crystal growth teams develop their







27

own technique of wafer preparation using a combination of the

methods reviewed above.

Once an MOCVD system has been optimized for growing high

purity InP and a proper substrate vendor, orientation and

cleaning procedure have all been selected, most authors report

that the source purity of both the metal organic indium and

phosphine have the strongest influence on background carrier

concentrations and mobilities. The initial work on InP growth

[12,33-34] reported room temperature carrier concentrations of

n=0.17 to 1.4 x 106cm'3 and 300 K and 77 K electron mobilities

of 3500-4200 and 16,000-36,000cm2/volt-sec, respectively.

After two decades of technological advancement in purification

techniques and machine design, the highest reported 77 K

mobility for undoped InP is now 305,000cm2/volt-sec with a

corresponding carrier concentration of n= 5 x 1013cm'3[55].

Based on low temperature PL it appears the dominant residual

acceptor in MOCVD InP is zinc[55]. For many years both carbon

and manganese(47] have also been reported as compensating

acceptors and silicon has been reported as the dominant

donor[56]. The recent work of Bose et al.[55] caution against

identifying PL peaks as carbon since the transverse optical

phonon replicas of the free-exciton recombination occur at the

same energy as carbon. For most teams, however, 300 K and

77 K mobilities of 4,700 and 80,000cm2/volt-sec, respectively,

and a carrier concentration of n= 1 x 1014cm"3 are typical for







28

undoped MOCVD InP. So, it is evident that device quality

unintentionally doped InP can be grown by MOCVD.

Most semiconductor devices require a junction of some

type in the host material where two materials with either

different electrical or optical properties meet. An

electrical junction can be created by post growth processing

techniques such as ion-implantation or diffusion of a donor or

acceptor into the host crystal. Another way is to just create

the junction in-situ during the MOCVD growth by adding a small

quantity of a donor or acceptor source into the inlet gas

stream. High quality undoped InP is usually n-type with a

background carrier concentration of n = 1 x 1014- 1015cm3.

The carrier concentration n or (ND-NA) can be increased by

adding an InP donor species to the inlet gas stream of the

MOCVD reactor. InP can be doped n-type by using H2S[45],

H2Se[43] and SiH4[43] or Si2H6[57] as sources. For each

source, the free carrier concentration is essentially

proportional to the dopant source mole fraction in the reactor

inlet stream. Controllable n-type doping from 015cm'3 to

1020cm'3 can be achieved without a significant decrease in

material quality by using a combination of these sources for

different parts of this wide incorporation range. Doping

levels and diffusion rates of these dopants are affected to

varying degrees by changes in growth conditions such as

temperature, V/III ratio, and indium mole fraction. For H2S,

the free carrier concentration in deposited InP layers







29

decreases when the growth temperature increases. For SiH4,

the opposite trend is observed, because the incorporation of

silane is reaction limited whereas H2S is adsorption limited.

Si is amphoteric in InP, acting as an acceptor or donor,

depending on site selection (adjusted by changes in the V/III

ratio used during the growth). H2S does not compensate itself

in InP but the diffusion coefficient of S is greater than that

of Si in InP. Depending on the device application, a suitable

n-type dopant source for InP is apparently available.

InP can also be doped p-type by adding an InP acceptor

species to the inlet gas stream of the reactor. The gas

stream must contain enough of an acceptor species to increase

the electrically active extrinsic acceptor level above the

electrically active intrinsic donor level. The metal organic

(MO) compounds DESn[45], DMZn[58], DMCd[59], Cp2Mg[60] and

MCp2Mg[61] act as sources for acceptors in InP with varying

degrees of success. There is not one single MO acceptor

source which dopes InP p-type over a wide doping range (1015-

1019cm3) without extended diffusion or surface morphology

degradation. This is the reason that so many different

sources have been investigated for InP as suitable p-doping

sources. This dilemma, in combination with the observation

that the electrical activation of some p-type dopants in InP

is much less than unity, was a driving force for the extensive

literature review and model development for p-doping of MOCVD

InP in Chapter III of this dissertation. The most commonly








30

used p-dopant source is DEZn which can be used to dope InP

over the range p = 1015-1018cm3. DEZn is sufficient for most

device applications, but its relatively high diffusion

coefficient at typical growth temperatures, D = 3 x 10'13

cm2/sec[58], does make it unsuitable for some device

applications. For a more extensive discussion on p-doping of

InP, the reader is referred to Chapter III of this

dissertation.

2.3.2. GaInAs/InP

The ternary compound GaxIn1.xAs can be grown lattice-

matched to InP by MOCVD. Unlike the AlGaAs/GaAs material

system, the GaInAs/InP system is not lattice-matched for all

compositions. Ga47In53As has an energy gap of 0.75 eV (Ag =

1.67 pm) and is the only composition which is lattice-matched

to InP. This ternary film can be grown by carefully

controlling the gallium to indium metal organic source

composition of the gas inlet to an MOCVD reactor. This is one

of the most severe heteroepitaxial growth scenarios possible

as the group V sublattice must be changed from pure phosphorus

to pure arsenic.

The first reported MOCVD growth of GaxIn1lxAs was on a

GaAs substrate[61] and hence was lattice-mismatched. The

early efforts provided useful information such as

compositional uniformity, merits of methyl versus ethyl MO

sources, gas phase reactions and purity for later GaInAs/InP

work. One conclusion that was useful for GaInAs/InP work was







31

the improvement in material quality observed upon using TEGa

instead of TMGa as the gallium source. Another useful

observation was that the solid phase composition is controlled

by and almost equal to the gas phase ratio [TMIn]/([TMIn] +

[TEGa]). Finally, the GaInAs growth rate is proportional to

the sum of the metal organic gas phase concentrations.

Lattice-matched Ga47In53As grown on InP by low pressure

MOCVD using TEGa and TEIn was first reported by Hirtz et al.

[62] in 1980. As stated previously, the choice of group III

alkyl sources used for GaInAs growth is critical. Using TMGa

and TEIn results in poor surface morphology, whereas using

TEGa and TEIn results in nearly featureless material over the

composition range 0.4 < x < 0.6 (Gax). This phenomena has

been attributed to a TMGa-InP substrate steric hindrance to

the heterogeneous decomposition of TEIn[63]. When growing on

InP, the initial stage of growth of GaInAs is also complicated

by the incongruent evaporation of phosphorus from the InP

substrates upon heating. It has been shown that the

morphological, optical and electrical properties of the GaInAs

epitaxial layer depend heavily on minimizing InP substrate

damage during the transition from PH3 to AsH3[64]. The best

approach is using an InP buffer layer and then allowing the

indium flow to continue while rapidly switching phosphine to

the vent and TEGa and AsH3 to a low pressure reactor.

As is the case for growth on GaAs substrates, the

composition of GaInAs is linearly dependent on the flow rate







32

of TEGa for a fixed TEIn or TMIn flow. Both the composition

and growth rate are independent of AsH3 flow for fixed metal

organic flow. The growth rate is independent of growth

temperature (500-6500C), but the composition can be slightly

affected due to slight differences in gallium and indium

source cracking efficiencies. If the composition of a layer

is different from the lattice-matched value, this layer is

mismatched. The lattice-mismatch between layer and substrate

is defined as Aa/a = (aL a)/ a when aL is the measured

room temperature (strained) lattice parameter and a is the

lattice parameter of the substrate. Razeghi et al.[38] have

reported that the mobility of a semiconducting layer is

dependent on the amount of mismatch in the layer relative to

the substrate. At optimum growth conditions, a lattice-

matching of Aa/a < 0.04% has been achieved resulting in Hall

mobilities of 12,000 (300 K), 100,000 (77 K) and 260,000cm2/

volt-sec (2 K) and background carrier concentrations of 0.7 -

1.0 x 1015cm'3[65]. Such high mobilities at 2 K are explained

by the existence of a two-dimensional electron gas formed at

the interface between undoped InP buffer and GaInAs layers and

are indicative of superior material quality.

It is evident that MOCVD can be used to produce undoped

Ga 47In.53As lattice-matched to InP, with very high quality

electronic properties. Intentionally doped both p- and n-type

Ga47In.53As on InP is also producible. Razeghi[65] presented

data on p-type doping using DEZn and n-type doping using H2S







33

of GaInAs by MOCVD. Zinc doped GaInAs carrier concentrations

are reported to decrease with increasing growth temperatures

over the range p = 1017 018cm-3. The opposite behavior is

observed for sulphur doped GaInAs and this trend is confirmed

by Logan et al.[66]; a carrier concentration of 1020cm'3 is

reported for a growth temperature of 5250C and at nearly

identical conditions except Tg = 6250C, n = 8 x 017cm'3. Wide

ranges of both p- and n-type doping are attainable for GaInAs

on InP by MOCVD which is useful for device applications.

2.3.3 GaInAsP/InP

The quaternary solid solution GaxIn.xAAs P1., is an alloy

semiconductor which can be lattice-matched to InP and GaAs

substrates. GaInAsP is a direct bandgap semiconductor (when

lattice-matched to GaAs or InP) which can be a very efficient

light emitter over the wavelength range 0.65 0.87gm

(lattice-matched to GaAs) and 0.92 1.65gm (lattice-matched

to InP). Very little work has been reported on quaternary

growth on GaAs substrates[67,68] due to greater interest in

GaInAsP alloys lattice-matched to InP substrates for optical

fiber device applications. Optoelectronic devices operating

at 1.3/m or 1.55Am wavelength regions have immediate

commercial applications because light transmission through

silica fibers exhibits low loss at 1.3/m and low dispersion at

1.55gm.

The MOCVD growth of GaInAsP alloys on InP substrates was

first reported using ethyl alkyls in a low pressure







34

system[63]. This same team later reported growth of nearly

the entire quaternary composition range lattice-matched to InP

and presented device test results of broad area and stripe

lasers fabricated using six different quaternary

compositions[69]. After these initial reports, numerous

quaternary related papers have been published on novel growth

techniques, relationship between gas phase growth conditions

to solid phase compositions, electrical and optical material

properties, and device applications. Each of these topics

will be reviewed in the following paragraphs.

Growth techniques used for depositing quaternary alloys

are direct extensions of techniques used for growing InP and

GaInAs. The growth rate is similarly proportional to the sum

of the partial pressure of TEGa and TMIn, and is independent

of the phosphorus and/or arsenic partial pressure. Similar to

the growth of GaInAs, the solid phase group III composition is

essentially equal to the gas phase metal organic composition

introduced to the MOCVD reactor. However, the behavior of

incorporation of group V elements is different and much more

difficult to control since both arsine and phosphine are

required and they do not incorporate with the same

probability. It is much more difficult to incorporate

phosphorus than arsenic at a fixed growth temperature since

the cracking temperature of PH3 is higher than AsH3. To

alleviate this problem, some workers have used precracked

PH3[38,70] or in-situ adduct formation techniques[71] (which







35

is a more easily cracked species) both with varying degrees of

success. Most teams use the low pressure MOCVD technique

which makes PH3/AsH3 ratios as high as 200 (which is required

for 1.0pm wavelength quaternaries) more safely attainable.

The operating parameters inlet partial pressures, total

pressure, deposition temperature, and V/III ratio have an

effect on the growth rate and composition of deposited

quaternary materials. Because of this, numerous papers have

been written relating the gas phase growth conditions to solid

phase material quality and composition. Razeghi[65] has

published graphs on which are plotted the relationship between

growth conditions ratios and bandgap wavelengths for the full

range of quaternary materials which are lattice-matched to

InP. The three ratios are: (1) Rg = PH3 / (PH3+AsH3) ; (2) R3

= TEGa/(TEGa + TEIn); and (3) R5/R3. Using these graphs,

which are only valid for a growth temperature of 6500C and

total flow rate of 7 liters/min, one can estimate growth

conditions for any lattice-matched quaternary composition.

Similarly, Fujii et al.[72] and Sugou et al.[70] present

quaternary compositions as a function of (In/Ga) and (P/As)

ratios in the gas phase for fixed V/III ratios. Koukitu and

Seki[73] use a thermodynamic approach to compute the solid

composition as a function of input mole ratio for several

quaternary III-V alloy systems. They also compute the

equilibrium partial pressures of gaseous species over

GaInAsP/InP as a function of temperature, V/III ratio and







36

Ag(Am). Using one of the above mentioned techniques, good

estimates of optimum gas phase growth conditions for the full

range of solid phase quaternary compositions can be predicted.

Several papers have been written presenting optical and

electrical property data as a function of composition for

GaxInl.AsyP1. lattice-matched to InP. One important

realization is that for lattice-matching, y is related to x by

the following simple relation: y = 2.16 x. This greatly

facilitates presenting data as it can be plotted as a function

of y or x under the assumption that the lattice-matching

condition is realized. A paper by Nahory et al.[74] presents

useful experimental lattice constant and bandgap values as a

function of composition relative to lattice constant values

predicted using Vegard's Law. Vegard's Law states that for a

lattice-matched system, the lattice parameter of the

quaternary can be deduced from those of the constituent

binaries. This team also presents the empirical relation for

bandgap variation (Eg(eV)) with composition.

Eg(y) = 1.35 0.72y + 0.12y2 (1)

Another group presented undoped electron and hole mobilities

as a function of composition, y. For undoped GaInAsP, room

temperature electron mobilities range from 4000 to

11,000cm2/volt-sec (y = 0 to y = 1) and room temperature hole

mobilities range from 130 to 200cm2/volt-sec (y=0 to y=l) [75].

Both p- and n-type doping of GaInAsP/InP grown by MOCVD

have been reported. Extensive doping studies as a function of








37

quaternary composition has not, however, been reported.

Saxena et al.[76] present data on 1.3Am p-GaInAsP doped by

DEZn over the range 1018-1019cm',and n-GaInAsP doped by DETe

and H2S. Using tellurium, n-type doping from 3x1017 to

5x109cm'3 is reported and, with sulphur a lower range 5x1016

to 3x1018cm'3 is reported. Meyer et al.[77] report n-doping

of 1.3 and 1.55Am GaInAsP using H2S over the range 1016 to

1019cm3. It is evident that p- and n-doping of GaInAsP is

possible over a wide range of doping levels which is

significant for device applications.

2.3.4 InP Based Devices

The wide range of compounds that can be grown with large

area uniformity by MOCVD make it suitable for fabrication of

long wavelength opto-electronic device structures. Several

different types of electronic, optical and opto-electronic

devices have been grown by MOCVD in the GaxInlxASyP1.y on InP

material systems. Manasevit et al.[35] in 1978 showed that

solar cells in which an InP active region grown by MOCVD can

perform as well as ones fabricated by other techniques. Other

devices such as lasers, field effect transistors, photo

detectors and waveguides have been grown by MOCVD and will be

discussed in the following paragraphs.

Some of the first devices grown using MOCVD GaInAsP were

broad area and stripe double heterostructure (DH) lasers which

lased at wavelengths between 1.15 and 1.54jm[69]. Distributed

feedback (DFB) lasers have also been successfully grown by








38

MOCVD. With MOCVD, it is possible to overgrow onto sub-micron

diffraction gratings which is vital to the operation of these

devices[79]. A typical laser structure composed of these

materials would start with a nr-InP substrate, followed by a

21m thick n* InP layer, then a 0.2Am active layer of

lattice-matched GaInAsP (undoped), then a 2Mm thick p-InP

layer ending with a p* InP contact layer of 0.2Am thickness.

The lasing wavelength is determined by the composition of the

active layer. Surface emitting semiconductor diode lasers,

which emit light perpendicular to the grown layer surface have

also been fabricated in an array form using MOCVD grown

GaInAsP[80].

Microwave devices such as Gunn diodes and metal to

semiconductor field-effect transistors (MESFET's) have been

grown using MOCVD InP. For Gunn effect devices, even though

the mobility of InP is lower than GaAs, other characteristics

such as cut-off frequency, acceleration-deceleration time,

relaxation time and peak-to-valley ratio are better in InP.

These devices require a three-layer structure of n+-n-n+-InP

grown on n+-InP and have been successfully grown by MOCVD for

60 GHz[36], and 94 GHz[81] operation. MESFET's have also

successfully been fabricated by the use of undoped InP grown

by MOCVD on Fe-doped substrates. The electrical properties of

Au-InP Schottky diodes are reasonable and comparable to other

crystal growth techniques[82].







39

The use of the ternary material GaInAs lattice-matched to

InP for long-wavelength photodetectors is well established.

Traditionally, LPE and hydride VPE are used but, MOCVD grown

p-i-n photodiodes have also been prepared[65]. The most

important device characteristics required of detectors are low

capacitance, low dark field leakage current and high quantum

efficiency. To attain these goals, low background doping

levels, accurate lattice-matching and an abrupt p-n junction

are required and all of these are possible with MOCVD.

Actually, with MOCVD the need for a post-growth zinc diffusion

processing step can usually be eliminated since in-situ p-

doping is possible. Several teams have reported improvements

in p-i-n photodiode performance by adjusting MOCVD growth

conditions[76], layer structure[65], and Schottky barrier

height enhancement[83]. MOCVD grown structures with leakage

currents as low as 3 pA at -10V using a 100/m device diameter

have been fabricated on two inch diameter InP substrates[84].

In addition to the above structures, a number of other

optoelectronic devices have been fabricated using MOCVD grown

InP based materials. Two dimensional electron gas (2DEG) and

multiple quantum well structures have been grown making use of

the extremely high mobilities, (in excess of 180,000cm2/volt-

sec at 9.2 K) possible with these materials[85]. Guided wave

devices such as optical waveguides and phase modulators have

also been grown[86]. Finally, GaInAsP/InP interference

filters have recently been grown by MOCVD[87]. The theory








40

and results of low pressure MOCVD grown GaInAsP/InP and MBE

grown AlGaAs/GaAs electrically tuneable interference filters

are presented in Chapter IV of this dissertation. More

extensive reviews of the wide range of MOCVD grown opto-

electronic devices using InP based materials are available in

the literature[65,88].

2.4. A Description of the MOCVD System

2.4.1 Introduction

The experimental apparatus used for the growth of

epitaxial layers of GaxIn1.AsyP1.y on InP substrates is a

commercial MOCVD system custom built for the University of

Florida by Nippon Sanso K.K. (Japan Oxygen Inc.). A

photograph of the front and a simplified schematic of the

Japan Oxygen MOCVD System are shown in Figures 2 and 3. The

complete operating procedures for performing epitaxial growths

and maintenance (e.g., such as reactor cleaning), are

presented in Appendix B of this text. The four basic parts of

the MOCVD system which are described in the following

paragraphs are: (1) the gas delivery system; (2) the reactor

and heating system; (3) the exhaust/scrubbing system; and (4)

the safety system. The gas delivery system, reactor, exhaust

and safety system are all integrated inside the MOCVD system

which is shown in the photograph in Figure 2. The heating

system is a separate unit (20 kW RF generator) as is the

scrubbing system which is located outside the building for

ease of maintenance reasons.
























































Figure 2: Photograph of the quaternary MOCVD system.














































p-dopant Vacuum pump


Figure 3: Quaternary MOCVD simplified flow diagram.









2.4.2.Gas Delivery System

The gas delivery system connects the sources to the

reactor and provides a method of transporting them in a

controlled fashion. Since impurity levels must be kept to a

minimum, all components of the gas delivery system are

constructed from electropolished 316L stainless steel and

connected with metal-gasket leak-tight couplings. Also for

improved purity, 0.2Am particle filters are installed at all

gas inlet points. All lines were wrapped with electrothermal

heating tape and aluminum foil and are heated during standby

mode to 500C to help desorb any of the sources or impurities

adsorbed on the inner walls of the stainless steel tubing.

The flow of gases is controlled by a combination of

manual valves, needle valves, pneumatic valves, check valves,

electronic mass flow controllers and regulators. The range of

possible flow rates for each source and the carrier gases

(hydrogen and nitrogen) are given in Table 2. The house

nitrogen gas which is mainly used for purging the MOCVD system

before a reactor or source change, passes through a molecular

sieve cartridge (Matheson Model 451) before entering the

machine. The house hydrogen, which is the carrier gas in the

system, is purified by diffusing it through a heated (4000C)

palladium-alloy membrane which is part of a 0-20 liter/min

hydrogen purifier system (Matheson, Series 8370V) that was

installed inside the Japan Oxygen machine.








44

Table 2

Flow Rate Ranges of Sources, Vendors and Purity


Source Flow Rate Range Vendor(Purity)

H2 0-20 SLM Gator Oxygen

(Alloy Diffused)

N2 0-10 SLM Linde(LN2 Boil-off)

AsH3 0-50 sccm Matheson (ULSI Grade)

PH3 0-200 sccm Solkatronic

(Micropure Grade)

1000ppm H2S 0-50 sccm Matheson

(in H2) (ULSI Grade)

TMIn 0-300 sccm Air Products

(Diphos Grade)

TEGa 0-100 sccm Akzo

(Electronic Grade)

DEZn 0-50 sccm Morton Thiokol

(Electronic Grade)







45

The gas delivery system for the metal organic sources

trimethylindium, triethylgallium and p-dopant diethylzinc

(which are held in stainless steel bubblers in temperature

controlled baths) consists of pneumatic, needle and manual

valves which are attached to the inlet and outlet ports of the

bubblers. Each metal organic line also has a pressure sensor

attached to it just before the inlet of the bubbler. The

metal organic sources are solids or liquids at room

temperature with fairly low, temperature dependent equilibrium

vapor pressures. By varying the temperature of the bubbler

bath, the hydrogen flow rate through the bubbler, and the

pressure of the bubbler region of the gas delivery system (by

opening or closing the needle valve), a controllable range of

metal organic source flow rates can be attained.

Since the hydrides (arsine, phosphine) and n-type dopant

(1000 ppm H2S diluted in H2) are highly toxic, combustible and

at high pressure, the gas delivery system for these sources is

slightly more complicated than for the metal organic. Each

line has a regulator, air operated valve and a manifold

attached to it. The manifold contains a high-pressure, high-

purity nitrogen purge line, a hydrogen purge line, and a third

line which can be used to evacuate the hydride line or vent

the hydride source directly to the scrubber. During normal

"standby" operation, palladium-alloy diffused hydrogen is

purging the hydride line to the vent. Only during toxic gas

flow is the hydrogen purge interrupted. Check valves on the








46

hydrogen and nitrogen lines prevent back flow of toxic gas

through the manifold to the rest of the system.

The "heart" of the gas delivery system is the fast-

switching "vent/run" manifold mounted just prior to the inlet

of the reactor. The status of the "vent/run" valves (opening

and closing) and the timing of this sequence can be manually

or automatically controlled. Automatic control is made

possible by using a process sequencer which is capable of

storing 150 valve patterns and the corresponding times for

each pattern or layer. The MOCVD manifold is also "pressure

balanced" which means that it is possible to adjust the

pressure difference between the two vent lines and the two

reactor lines to almost zero. This is accomplished by

carefully adjusting tube lengths and by installing a dead

volume in the vent line to counter balance the volume of the

reactor. Pressure balancing is especially important for MOCVD

growth of superlattice or multiple quantum well structures.

To keep the pressure difference between the reactor and vent

lines equal to zero (which is measured using the two

differential pressure indicators) each main source line has a

hydrogen compensation line. It is necessary for the MOCVD

growth of InP and related materials to keep the metal organic

separated from the hydrides to prevent parasitic gas phase

prereactions from occurring, hence the reason for the two

vent/two reactor line design.








47

2.4.3 Reactor and Heating System

The Japan Oxygen MOCVD system has a horizontal 4 inch

I.D. reactor which is equipped with a water cooling jacket

(see Figure 4) to minimize side wall deposition. At the inlet

of the high purity quartz reactor, two inlet lines exist to

keep the group III metal organic separated from the group V

hydrides at low temperatures. The dilute group III sources

enter the reactor through a 6 mm O.D. high purity quartz tube

10cm downstream from the inlet of the group V sources. These

two gas streams should ideally be well mixed and flowing under

fully developed laminar flow conditions before the mixture

encounters the high purity quartz deflector/silicon carbide

coated graphite susceptor heat source. The deflector -

susceptor unit is tapered at an angle of 170 with respect to

horizontal to improve growth rate uniformity.

The graphite susceptor is heated by radio frequency

inductive heating. A Lepel series T-15-3-KC-TL 15 kilowatt RF

generator is used to generate radio waves over the frequency

range 80-900 KHz. Some of these radio waves are picked up by

a 3/8" copper coil that is wrapped around the reactor. A

platinum/rhodium "R-type" thermocouple sealed in a high purity

quartz tube is embedded inside the graphite wedge. This

thermocouple is connected to a West series 2070 microprocessor

based temperature controller which was installed on the MOCVD

system and sends DC current to the RF generator to control















H2 and Hydride
Gas Inlet


Sample
Position


Quartz
Wafer Tray


Fork
Mechanism


/


Quartz Graphite Tnermocouple
Deflector Susceptor Tube


Figure 4: MOCVD reactor and loading mechanism.


H2 and MO
Inlet


,, __________,








49

its power output. Automatic temperature control from room

temperature to growth temperatures (550-7000C) to bakeout

temperatures (900-9500C) is possible with this elaborate

heating system.

As shown in Figure 4, a high-purity quartz sample tray

sits on top of part of the graphite susceptor. This tray

holds substrates as large as 1/4 of a two inch diameter wafer.

The tray and substrate are placed onto and removed from the

susceptor by an electro-mechanical fork which is capable of

precise x-y-z motion. The fork can move 90 cm horizontally

from the load lock area (where the sample and tray are loaded

and unloaded) through a gate valve and shutter valve, by the

gas exhaust port to directly above the susceptor. Precise

mechanical sample loading makes it possible to use the reactor

even after side wall deposition has obstructed view of the

susceptor.

To prevent the introduction of "dirty" room air to the

reactor during each loading of a substrate, the air in the

load-lock is evacuated by a rotary pump (Edwards model E2M2)

to a roughing pressure of 10-2 torr. Then a turbomolecular

pump (Balzers model TPH050) is turned on and evacuates the

load-lock to a pressure of 10"7 torr. At this point the load

lock is isolated and backfilled with the ultra high purity

hydrogen. The sample is now ready for loading and the ultra

high vacuum gate valve (VAT Ltd. model MSS4) is opened.







50

2.4.4 Exhaust/Scrubbing System

The waste products from the reactor flow through an

exhaust port in the water cooled flange which is bolted to the

exit of the reactor to a particle filter (Fuji Ltd.) which is

made out of glass fibers. The gas can then either flow at

atmospheric pressure directly to the scrubber or it can be

evacuated from the system by a rotary pump (Edwards model E2M

18) and then continue onto the scrubber. The rotary pump

makes it possible to grow films at low pressures (0.05 to 0.2

atm) which generally improves thickness uniformity.

The scrubber which was added to the Japan Oxygen MOCVD

system was built by Advanced Concepts (model 9625). It is a

liquid based scrubber which is designed to scrub toxic gases

and exhaust clean gas to atmosphere. For safety, the exhaust

was connected by a fireproof duct to the building's room air

scrubber. The scrubbing solution consists of a 80:5:2 by

volume mixture of water, 15% sodium hypochlorite, and 50%

sodium hydroxide. The pH and oxidation reduction potential

(ORP) of the scrubbing solution are constantly monitored to

evaluate the solution's scrubbing potential (pH > 10.0 and ORP

> 200 mV). The scrubber has an efficient gas-liquid venturi

contactor and a packed bed which can handle higher than

required toxic gas flow rates. For arsine and phosphine the

overall chemical neutralization reactions are:

AsH3(g)+3NaOC1(1)+H20(l) = H3AsO4(l)+3NaCl(s)+H2(g) (2)

PH3(g)+3NaOC1(1)+H20(l) = H3PO4(1)+3NaCl(s)+H2(g) (3)








51

The products of these reactions are less hazardous and for

added safety, when the system requires draining, the solution

flows to the building's neutralization pit.

2.4.5 Safety

Arsine, phosphine and hydrogen sulfide are highly toxic

combustible gases, therefore, several safety features are

installed in the MOCVD system, and several safety practices

must be followed in the facility. First of all, the facility

has a rule that at least two competent people must be present

in order to do any work in the building. The facility also

has an alarm system and card reader to deny access to

unauthorized personnel. The clean room has an eight point

toxic gas monitor (MDA, Inc.) which has a resolution of 1 ppb

for all hydrides. In the unlikely event of a toxic gas

detection anywhere in the facility, the MOCVD system will

completely shut down (all air-operated valves are normally

closed). The MOCVD system also has a four point hydrogen gas

detector (Matheson) connected to it so that if hydrogen levels

exceed 50 ppm, the machine will shut down. Also the facility

has a helium leak detector (Varian) which is used to find

actual leak points before the MOCVD system is used and to

check connections after valve replacements or reactor changes.

Not only is the facility well equipped with safety

features and practices, but the Japan Oxygen MOCVD System

itself has several integrated safety systems. There are two

types of alarms, facility failure and machine failure. If







52

either power fails, compressed air pressure drops, cooling

water pressure drops or temperature increases, the machine

will automatically alarm and all air operated valves will

close. There are smoke detectors in the machine and fire

detectors in the room. There are pressure sensors on the

hydride lines, in the reactor, and on the exhaust line and for

each, if a certain pressure value is exceeded, an alarm will

sound and the machine will shut down. There is also a pH, ORP

and temperature sensor on the scrubber which triggers an alarm

in the clean room if any of these values are out of the safety

range. Finally, there is a compressed breathable air supply

always on hand for reactor or source changes and two SCBAs

available for the emergency response team. It is evident that

safety is a big concern and since the machine was constructed

in Japan, there is even an earthquake sensor attached to it.



2.5 Determination of Optimum Growth Conditions Based on Thin
Film Characterization

2.5.1 Experimental Method

There are several experimental parameters that must be

determined before performing a MOCVD growth. Using the

simplest case as an example, undoped InP on InP, the first

thing that must be decided is what type of substrate is

required. For all of the GaxInl.xASyP1.y on InP experiments

performed in the Japan Oxygen MOCVD system, InP oriented (100)

20 towards (110) purchased from Sumitomo Inc. were used. Both

semi-insulating (iron doped) and n-type (n 8x1018cm'3, sulfur







53

doped) InP were used together or by themselves depending on

the purpose of the experiment. Next, the proper substrate

cleaning procedure must be decided. Also, experimental growth

conditions such as: growth temperature, growth pressure,

total hydrogen flow rate, trimethylindium mole fraction,

phosphine mole fraction and finally the length of time of the

planned deposition, must be determined. From the literature,

estimates of optimum growth conditions and proper wafer

cleaning procedures can be obtained, but these values or

procedures are system specific and hence had to be

experimentally optimized for the Japan Oxygen MOCVD system at

the University of Florida.

The procedure for calculating growth conditions for

undoped InP on InP by MOCVD is fairly straight forward. The

growth pressure, temperature and time must be chosen based on

knowledge of the material and capabilities of the system. It

should be noted that for convenience, some simplifications

were made in the derivation of the equations used to calculate

the growth conditions. Specifically, the ideal gas law was

used and it is assumed that the total flow rate to the reactor

is equal to the hydrogen flow rate and the bubbler pressure

was much greater than any metal organic vapor pressure. The

TMIn and PH3 flow rates are commonly expressed as TMIn mole

fraction (MFTmI) and PH3 mole fraction (MFPH3) or their ratio,

the V/III ratio, which in this case is defined as:










V FPH3 (4)
III MFTHIn

where:


F2
MFpH3 (5)





FH2,THIn VPTIn (T)
MFTMIn = (6)
FH2 Pb,TMIn



and FpH3 is the total pure phosphine flow rate (cm3/min), FH2

is the total hydrogen flow rate (cm3/min), FH2,TMIn is the

hydrogen flow rate through the TMIn bubbler (cm3/min),

VPTMIn(T) (mm Hg) is the bubbler temperature dependent TMIn

vapor pressure (torr), and Pb,TMIn is total pressure at the

TMIn bubbler (torr). All volumetric flow rates are measured

at standard conditions (300 K and 760 torr). The temperature

dependent vapor pressure equations for the metal organic

installed in the MOCVD system which were supplied by their

vendors (see Table 2) are shown in Table 3. It is therefore

possible with the use of the above equations to determine

conditions for a MOCVD experiment. Of course some initial

values must be known and others can be based on literature

values.










Table 3

Metal Organic Vapor Pressure Equations


Source Vapor Pressure Equation



log1oP(mm Hg) = 10.52 3014
TMIn T(K)





log,0P(mm Hg) = 8.224 2222
TEGa T(K)




2190
logo1P(mm Hg) = 8.28 2190
DEZn T(K)









2.5.2 InP

The method for determining the optimum growth conditions

for undoped InP on InP, n-type InP on InP and p-type InP on

InP will be discussed in the following sections. Before

presenting the results of this operating parameter sensitivity

and optimization study, the criteria for good epitaxial layers

are: mirror-like low defect density surfaces, uniform growth

rates over a large area substrate, less than n = 1x1015cm'3

background (ND-NA) doping levels for undoped InP with

corresponding high room temperature electron mobilities (j 2

3000cm2/volt-sec), experimentally convenient growth rates of

l-2jm/hour, and for doped layers and heterostructures,

atomically abrupt junctions and interfaces, respectively. All

of the above criteria have been determined by using various

thin film characterization techniques available on the campus

of the University of Florida.

2.5.2a Undoped InP

An extensive optimization study has been performed on the

MOCVD growth of undoped InP on InP. The results of this study

have been applied to aid in the determination of optimum

growth conditions for n- and p-type InP, undoped, n- and p-

type GaInAs, and undoped, n- and p-type GaInAsP on InP; these

optima will be discussed later. The timing and sequence of

events for a typical MOCVD growth are presented in the

Appendix B of this text.







57

After a growth has been performed, the first property

studied is the surface morphology of the thin film viewed

under an optical microscope. The thickness of the film can

also be determined by cleaving a sample and etching it for two

minutes in a 6:4:50 solution (measured by weight) of potassium

hydroxide, potassium ferricyanide, and deionized water to

delineate the grown layer from the substrate and then viewing

it under a microscope. A Nikon Optiphot microscope equipped

with Nomarski phase contrasting was used to view the surface

and delineated layer of the third growth in the machine, Q003.

Polaroid photos of the images observed at 2000X magnification

are shown in Figure 5. As shown, the surface morphology has

a ripple in it and this is sometimes referred to as an "orange

peel" surface. The side view photo of the sample was taken at

the leading edge of the sample with respect to the gas flow

direction. As shown, the layer thickness is clearly greater

(1.5Am) near the corner compared to the center (1.Om)

probably due to the lower growth rate right on the corner

which is at a different crystalline orientation, the (111)

orientation.

Several experimental parameters have been varied in an

attempt to improve the surface morphology of undoped InP grown

on InP. The growth temperature was varied from 450 to 7500C

and it was found that layers grown at temperatures between 575

and 7000C had better morphologies than those grown at higher

or lower temperatures. At lower temperatures lack of












































'1


Figure 5: Nomarski photographs of an InP plane view surface
and stained edge cross section taken at 2000X
magnification.







59

adequate phosphine decomposition and at higher temperatures,

slightly higher growth rates were the causes for inferior

material quality. The total hydrogen flow rate and growth

pressure were also varied but had little to no effect on

surface morphology. The TMIn mole fraction was varied and at

values greater than 1xl0"4 resulting material quality degraded

probably due to increased growth rates. The V/III ratio had

the most pronounced effect on layer surface morphology. This

effect can be clearly seen in Figure 6 which shows two InP

surfaces grown at identical conditions, except that the V/III

ratio was almost tripled by increasing the phosphine flow rate

to the reactor from 35cm3/min to 100cm3/min. This is in

agreement with the growth temperature study which concluded

that poor material results from insufficient phosphine

decomposition.

Another study performed with the goal of improving the

surface morphology of undoped InP on InP was varying the wafer

preparation procedure employed. Substrates were prepared

using five different techniques and then viewed under the

microscope. The procedures for each are as follows: (1)

filtered nitrogen blow off, five minutes each of warm acetone,

warm propanol, warm methanol, and then filtered nitrogen

blowoff; (2) procedure (1) followed by a 1 minute etch in 20%

nitric acid in methanol, methanol rinse, DI rinse, five minute

etch in room temperature 5:1:1 (sulfuric acid: hydrogen

peroxide: DI water), DI rinse, methanol rinse, filtered








60



I. .- i.





















































ratio on InP surface morphology (top: V/III=50,
bottom: V/III= 41).







61

nitrogen blowoff; (3) procedure (2) without the nitric acid in

methanol etch; (4) procedure (2) followed by a 1% bromine in

methanol etch for 6 minutes, methanol rinse, filtered nitrogen

blow dry; and (5) 2 minutes of surface treatment with the

UV/ozone cleaning system (UVOCS Inc).

Under a microscope at 200X magnification, particles were

observed on the surfaces prepared using procedures (1) and

(2). Procedure (4) resulted in a wavy surface probably due to

uneven bromine etching. Samples prepared using procedures (3)

and (5) had the best surfaces. Fresh substrates were cleaned

using procedures (3) and (5), loaded into the reactor, and

undoped InP was grown on both of them at the same time. This

experiment was performed several times and layers grown using

procedure (3) were consistently equal to or better than layers

grown using procedure (5). Procedure (3) was chosen as the

optimum and the exact details of this procedure are given in

Table 4.

The effect of growth conditions on the uniformity of

undoped epitaxial InP on InP was also studied. The total

volumetric flow rate was the only parameter that had any

significant effect on layer uniformity. Layers grown using

total hydrogen flow rates of 3, 5, 7 and 9 standard liters per

minute (SLM), had corresponding thickness variations of +/-10,

8, 6 and 6%. The increased variation at the lower flow rates

probably is associated with an increased boundary layer








62

Table 4

Optimum InP(100) Wafer Preparation Procedure





1. Cleave wafer and blow off with filtered N2.

2. Degrease teflon beakers and tweezers with warm methanol.

3. Place substrates into beaker with warm acetone for 5

minutes.

4. Place in warm propanol for 5 minutes.

5. Place in warm methanol for 5 minutes.

6. Rinse in running DI water for 1 minute.

7. Etch in a 5:1:1 solution of H2SO4:H202:DI at RT for 5 min.

8. Rinse in running DI for 1 minute.

9. Place in room temperature methanol for 1 minute.

10. Blow off with filtered nitrogen.

11. Load into reactor on wafer tray from oven.

12. Anneal wafer at growth temperature for ten minutes with H2

and PH3 flowing.







63

thicknesses resulting in depletion of TMIn near the growing

surface. This theory is confirmed by the observation that the

growth rate was lower, 0.95 and 1.2gm/hour for 3SLM and 5SLM,

than it was at higher flow rates (1.4 im/hour for 7SLM and

9SLM). The uniformity of undoped InP grown using a total

hydrogen flow rate of 7SLM is presented in Figure 7. A total

flow rate of 7SLM for hydrogen was chosen as optimum over 9SLM

because with 9SLM at a fixed TMIn mole fraction almost 30%

more TMIn material would be required.

The effect of several parameters on InP growth rate and

interface quality was also studied. As already stated the

total hydrogen flow rate has an effect on the growth rate.

The growth pressure and V/III ratio appear, however, to have

little to no effect. The growth temperature has a nonlinear

effect on the growth rate; at 450C the growth rate was

0.8Mm/hour, but going from 550 to 7500C the growth rate only

changed from 1.4 to 1.5Am/hour. This can be explained as

follows: at low temperatures (below 5500C), InP growth is

kinetically limited and consequently temperature dependent.

At substrate temperatures above 5500C, InP growth is TmIn

transport limited. This is confirmed by Figure 8 which shows

the linear relationship between InP growth rate and TMIn mole

fraction grown at 6000C. Some of the data for Figure 8 was

taken from the SEM micrograph shown in Figure 9. Figure 9 is

a scanning electron micrograph of a cross section from growth















m N 0



5
DISTANCE


10
FROM


15
LEADING


20
EDGE (mm)


n


-I


-20


S S


-10 0
DISTANCE FROM


10
CENTER (mm)


Figure 7: MOCVD thickness uniformity study.
variation, bottom: radial variation
the leading edge.


Top: axial
at 12 mm from


25


U U
Em..





















































0.0 f
0.00000


0.00005 0.00010 0.00015



TMIn Hole Fraction


0.00020


Figure 8: The effect of TMIn mole fraction on InP growth
rate.










* -


'Ul


Figure 9: SEM micrograph of growth Q054 at 20,000X.
layers are InP, dark are GaInAs.


Light


IEH








67

Q054 which shows InP layers grown with different MFTMIn for

different lengths of time separated by thinner dark GaInAs

marker layers. Figure 9 also shows the excellent abruptness

of interfaces between two different materials grown at 6000C

and 7SLM.

Another analytical technique used to evaluate material

quality is an electrochemical C-V profiler. The C-V profiler

(BioRad, Model PN4200) is thoroughly explained by Blood[89].

It uses a Schottky and ohmic contacts in an electrochemical

cell filled with .0.1M HC1 (for InP) to slowly etch away

material while it measures the change in capacitance as a

function of changing applied voltage. From this C-V data, the

net carrier concentration can be calculated (n = ND-NA or p =

NA-N,, where NA and ND are the number of acceptors and donors

per cm3) and plotted as a function of depth. A C-V profile of

growth Q056 which is undoped InP on n-type (8x1018cm'3)InP is

shown in Figure 10. From this figure one can get a rough

estimate of 1.2gm for the layer thickness. Also, the

epilayer-substrate interface abruptness can be assessed. Good

uniformity of the carrier concentration throughout the layer

can also be observed. Finally an average value of the

background carrier concentration for undoped InP of

n=3x1015cm'3 can be obtained. It is evident that a lot of

useful information can be obtained from an electrochemical C-V

profile, the only drawback to this technique is that it is

destructive.







































.5 1.0


X (um)


Figure 10: C-V profile of growth Q056 (undoped InP on n+InP).


15 L
0.


0


i i I r I I I


L


i








69

The effect of growth conditions on the background carrier

concentration of undoped InP has also been studied. The

growth pressure, total hydrogen flow rate, and TMIn mole

fraction apparently have little effect on the background

carrier concentration. The growth temperature and V/III

ratio, however, both have a strong effect. The background

carrier concentration measured by the C-V profiler with

respect to growth temperature is plotted in Figure 11. At a

low growth temperature, high carrier concentrations may be

associated with phosphorus vacancies in the InP crystal

(impurities are completing with these vacancies). At higher

temperatures, electrically active source impurities are

incorporating more efficiently. It appears that based on

background concentrations, the optimum growth temperature

should be between 5750C and 6250C for undoped InP. The

undoped InP carrier concentration is also strongly affected by

the V/III ratio as shown in Figure 12. Based on this figure,

high purity material can be grown with high PH3 flow rates.

Another analytical technique which is commonly used to

evaluate material quality is PL. The technique is explained

well by Dean[90]. Briefly, it typically uses laser light

energy incident upon a sample to stimulate the emission of

photons at discreet energy levels. These discreet energy

levels can be detected and related to impurities, defects, or

host crystalline energy level transitions. The resolution of

closely spaced transitions can be improved by reducing the










































300 400 500 600
TC


700 800 900


Figure 11: The effect of growth temperature, Tg, on undoped
(n-type) InP background carrier concentration.


10"


E
0




010

a
0



0
o
L.


(10


0


10


Ha=7.0 sl, P =80 torr
xi=1x 10-, V/f1=50


SI I I


I


















o 10
rj )
I H2=7.0 sin, P,=80 torr
E X,,=lx10-, T,= 600C



c--
I
C
0
r
10
O.
C \

i
0)

0 \







0.00 40.00 80.00 120.00 160.00
V/Ill Ratio








Figure 12: The effect of V/III ratio on C-V background carrier
concentration in undoped (n-type) InP.







72

temperature of the sample. The PL spectra of two different

InP samples grown at different V/III ratios and measured at

4.2 K are shown in Figures 13 and 14. The peaks labelled

(D,x) and (FE) are host crystalline donor to bound exciton

and free exciton transitions. The other peaks maybe due to

impurities such as carbon or crystalline defects like

phosphorus vacancies (Vp). Based on the reduction in area

under the peaks labelled (Vp), it is evident that increasing

the V/III ratio or phosphine flow rate definitely has a

positive effect of the material quality of undoped InP.

In addition to the C-V profiler, the Hall effect can also

be used to determine electrical properties of semiconductors.

A low temperature (room temperature to 7 K) Hall effect system

was set up at the University of Florida by this researcher.

The theory, and a manual explaining the use of the Hall effect

system is given in the Appendix A of this text. Using data

from the Hall effect system, an average value of the mobility,

resistivity and carrier concentration can be calculated if the

thickness of the measured film is known. These values are

usually measured at 300 and 77 K (liquid nitrogen temperature)

and sometimes plotted as a function of reciprocal temperature

to 4.2 K. The liquid nitrogen temperature mobility (p.K) is

widely used as a measure of purity of undoped semiconducting

films. In Figure 15, p A is plotted as a function of growth

temperatures for undoped InP. The highest mobility for InP























(Dox)

0.8752 eV
0.8750 eV

(vp) ??


d -
cu
ts ^
^ r


0.8746 eV

0.8744 eV


(FE)


10600 100oo 11600 112oo 1 o00 11600 1100oo


12000


Wavenumbers (cm-1)


Figure 13:


PL spectrum of an InP sample grown with a V/III
ratio of 140 measured at 4.2 K.






















(D, x)


(Vp)??

>o

I S gS

M o a

(, R), Carbon -
,m- I


0.8751 eV


10600 10800 11000 00 i 0 11400 11600 11800 12000
Wavenumbers (cm-1)















Figure 14: PL spectrum of an InP sample grown with a V/III
ratio of 219 measured at 4.2 K.























70000












50000
03
N









30000
1 40000-
-I








30000-





20000
500 550 600 650 700



Tg (c)













Figure 15: The effect of growth temperature on 77 K mobility
of undoped InP (bars indicate the range of data) .







76

material from in the Japan Oxygen MOCVD was measured to be

61,800cm2/volt-sec for a sample grown at a temperature of

6000C. There is a lot of scatter in the mobility data. The

scatter may be due to the fact that different material sources

with different impurity levels were used to grow these layers.

Also, the reactor was not changed/baked-out after each growth.

However, the average mobility values indicate that higher

quality material can be grown at 6000C and this is in

agreement with previously discussed characterization results.

Based on all of the observed characterization trends,

apparent optimum MOCVD growth conditions for undoped InP on

InP have been determined. These growth conditions are

presented in Table 5 along with the room temperature and 77 K

electrical properties of undoped InP grown at these

conditions. These optimum conditions are significant because

they are the basis for estimates of growth conditions for

intentionally doped InP, unintentionally and intentionally

doped GaInAs and GaInAsP, both lattice-matched to InP.

2.5.2b Growth of n-type InP Using H2S

The Japan Oxygen MOCVD system was designed with the plan

of using dilute H2S as an n-type doping source. A cylinder

containing a mixture of 1000ppm H2S and the balance ultra high

purity hydrogen (Matheson, ULSI grade) is connected to the

MOCVD system. The flow rate range measured at room

temperature and 760 torr for this mixture, as shown in Table

2, is 0-50cm3/min (sccm). Epitaxial layers of InP were grown








77

Table 5

Undoped InP Growth Conditions and Electrical Properties


Total H2 Flow Rate

Growth Temperature

TMIn Mole Fraction

V/III Ratio

Growth Pressure


= 7 SLM

= 6000C

= 0.7-10'4

= 419

= 80 Torr


at 300 K :


ND

ND

11300K

11300K


= 2.1014 cm'3 (lowest)

= 2*1015 cm'3 (average)

= 3461 cm2/volt-sec(highest)

= 2800 cm2/volt-sec(average)


at 77 K :


ND NA

ND N

I77K

977K


= 11014 cm'3 (lowest)

= 2*1015 cm'3 (average)

= 61800 cm2/volt-sec (highest)

= 40500 cm2/volt-sec (average)







78

with H2S mixture flow rates of 2 to 50 sccm, which corresponds

to gas phase mole fractions of H2S of 2.8x10"7 and 7.16x106,

respectively. A C-V profile of a S-doped InP grown layer is

shown in Figure 16. As one can see, a roughly l.lgm thick

layer of 2.5x1018cm"3 n-type material was deposited on a n*-InP

substrate, and this was achieved using a H2S mole fraction of

2.43x10'6. The "hump" in the profile in Figure 16 is a C-V

profiler error which occurs at interfaces. The relationship

between the measured C-V carrier concentration and H2S mixture

flow rate for several n-type samples is presented in Figure

17. As shown, a wide linear incorporation rate of sulfur in

InP is possible resulting in doping levels from 5x1017 to

2.5x1019cm'3. The surface morphology of the grown material

when viewed under the Nikon microscope at 2000x appeared

unaffected by the presence of the sulfur atoms even at the

highest n-type doping level. Room temperature Hall effect

measurements were also performed on the sulfur doped samples.

Hall carrier concentrations agreed with C-V measurements and

Hall mobilities (A300K) ranged from 498cm2/volt-sec at the

lowest doping level to 1064cm2/volt-sec at the highest level.

2.5.2c Growth of p-type InP Using DEZn

P-type conversion of MOCVD InP was achieved by mixing the

metal organic source diethylzinc (DEZn) with the standard gas

mixture used to grow undoped InP. In order to get a wide

range of p-type doping, the temperature of the DEZn bubbler

was varied from -20 to 200C resulting in a change in the




































E



0













iB
0.0


X (uml


Figure 16: C-V profile of a H2S doped InP film.









80









1020-





0








0
U El
o X
0






u l
H 7















10 "5 10 4 0
iol7---------------- .I




Partial Pressure (Torr)













Figure 17: The effect of H2S partial pressure on InP carrier
concentration.







81

vapor pressure of DEZn in the bubbler from 1.39 torr to 14.00

torr. The hydrogen flow rate through the bubbler ranged from

5 to 50 sccm and by carefully adjusting the opening of the

DEZn needle valve it was possible to keep the bubbler at a

pressure of 500 torr.

Characterization of the p-type InP material included

thickness and surface morphology measurements using the

optical microscope, C-V profiles, Hall measurements and

secondary in mass spectroscopy (SIMS). The SIMS technique,

which is explained elsewhere[91], was used to determine the

total atomic zinc concentration incorporated into several p-

type InP layers as a function of depth. This total zinc level

can be compared to the carrier concentration to determine the

percentage of electrically active zinc atoms. Also, based on

the depth of the atomic zinc profile, the extent of zinc

diffusion can be assessed relative to the epitaxial layer

thickness measurements. A C-V profile of growth Q080 which

was deposited at Tg=6000C, Pg=80 torr and V/III=50.0 is shown

in Figure 18. During the growth, a DEZn mole fraction of

4.06x10'5 was used which resulted in a C-V measured hole

concentration of 1.6x108cm"3, for the layer grown on the n+-

InP substrate. A room temperature Hall effect measured hole

concentration of 2.5x1018cm'3 was calculated for the layer

grown on the semi-insulating substrate. The room temperature

Hall mobility and resistivity for this p-type sample were

61.3cm2/volt-sec and 0.0402 ohm-cm, respectively.









































X (um)


Figure 18: A C-V profile of a DEZn doped InP film.







83

The relationship between the C-V measured hole concentration

and the corresponding DEZn partial pressure used for the

growth is plotted in Figure 19. As indicated, controllable p-

type doping from 2x1017cm3 to 3x1018cm3 was realized. Hall

measurements were attempted on all of the p-type layers grown

on semi-insulating InP substrates, but the alloyed indium

contacts were generally not ohmic in nature. Hence, only the

C-V carrier concentration results from films grown on n-type

substrates were used. The room temperature Hall carrier

concentrations of semi-insulating samples with ohmic contacts

agreed quite well with the C-V carrier concentrations on n-

type substrates.

The effect of several growth conditions on the p-type

doping of MOCVD InP was investigated. The following growth

parameters: DEZn mole fraction, growth pressure, V/III ratio,

growth temperature, and total hydrogen flow rate, were varied

during individual experiments. It was later discovered that

the effect of the extended diffusion of atomic zinc in InP

during growth influenced the spatial variation of zinc's

incorporation. Basically, any variation in incorporation that

might have existed due to variations in growth conditions

during the deposition, were negated by the rapid diffusion

rate of zinc. For example, Figure 20 shows a C-V profile of

growth Q105 during which the DEZn mole fraction was varied

from 0.2 to 1.0x10-5 and the C-V hole concentration of















19


10


z3 -
I-

T 0 "-
z


T%= 6000C
P p3=0.4 Torr
P9.n=.008 Torr


10 "17-
0.0001


I I I I I I I I I
0.001

PODEZn(Torr)


I I


0.01


Figure 19: The effect of DEZn partial pressure on InP hole
concentration.


- - --- - -


II


















20




n-Type


19


E DEZn Flowrate
2 Adjusted


\ I,/

p-Type





i7
0 1 2 3 1
X (um)












Figure 20: C-V profile of InP:Zn grown with different DEZn
partial pressures.







86

9xl017cm-3 as shown, is essentially constant throughout the

grown film. The large "dip" in the profile shown in Figure 20

is commonly observed at p-n electrical interfaces. A SIMS

measurement was also performed on this sample (see Figure 21)

and the atomic zinc concentration profile yielded essentially

the same result. Similar results were observed when comparing

the SIMS and C-V profiles of samples grown to study the effect

of the other system parameters on zinc incorporation.

One interesting and useful conclusion that can be derived

from the SIMS and C-V profiles of growth Q095, during which

the growth pressure was varied from 38 to 760 torr, is the

extent of zinc diffusion into the InP substrate. As shown in

Figure 22, the grown layer thickness of this sample is about

1.4Am and the zinc doping level is unaffected by the change in

growth pressure. In Figure 23, the atomic zinc level (from

SIMS) is essentially constant throughout the profile, but a

"spike" occurs in the zinc profile right at a depth of 1.4Mm.

The SIMS operators at BNR, where the data was taken, say that

this "spike" is due to silicon-zinc complex which results from

Si on the surface of the InP substrate wafer after cleaning

and it reproducibly indicates the position of the grown layer-

substrate interface. These "spikes" were observed in several

of the SIMS profiles done at BNR. Based on the depth of the

zinc profile in Figure 23, zinc has diffused approximately 0.5

microns into the InP substrate. Using this diffusion length,
























PROCEsBOE DATA
5 Aug ea Ca


0.0 0.5


BNR
PILE: 0105


1.5
DEPTH (micronr)


Figure 21: Atomic zinc profile of sample Q105 measured by
SIMS.








88












20




n-Type



19




z
C /
o p-Type












0 2 3
X (um)














Figure 22: C-V profile of growth Q095 during which the growth
pressure was varied.
























PROCESSED DATA
5 Aug a8 CE


4n-24--


o 1 "spike"



101






10 1



-1> Cc+Zn


0.0 1.5 2.0


1.5 2.0
DEPTM (microns)


Figure 23: Atomic zinc profile of sample Q095 measured by
SIMS.


BNR
FILE: 009o


0.0







90

L = 0.5 microns, and the length of time that the sample was at

growth temperature, t = 90 minutes, a rough estimate of the

diffusion coefficient, D, of zinc in InP can be calculated

using this equation:


D(cm2/sec) = L2(cm)/t(sec) (7)

A value of D = 4.6x1013cm2/sec is calculated which agrees very

well with the range of values that were reported by Nelson and

Westbrook[58], D = (1-6)x 10"3cm2/sec, for zinc in InP.

A review of the literature on the topic of p-type doping

of InP by several growth techniques, extensive data on zinc,

magnesium and cadmium p-type doping of InP by MOCVD, (which

was acquired when this investigator was a visitor at BNR) and

a theoretical model of the p-type doping process of MOCVD InP

is all presented in Chapter III of this text. The reader is

therefore referred to Chapter III for a more detailed and in-

depth discussion on p-type doping of InP.

2.5.3 Growth of GaInAs Lattice-Matched to InP

The mixed crystal Ga47In.53As which is lattice-matched to

InP, has been grown using the Japan Oxygen MOCVD system. The

growth conditions used were similar to the optimum conditions

for InP growth. Based on the results of several different

characterization techniques, optimum growth conditions for

undoped Ga47In.53As were determined. Whenever p- and n-doped

GaInAs was required for device applications, test layers were

grown to calibrate for the required DEZn and H2S gas phase







91

mole fractions. The timing and sequence for a typical MOCVD

growth of GaInAs on InP is presented in Appendix B of this

text.

Since GaxInl.xAs has only one composition (x=.47) which

is lattice-matched to InP, precise control of both the TMIn

and TEGa flow rates to the reactor is crucial. A small change

in the TMIn to TEGa gas phase molar ratio is approximately

equivalent to the change in deposited solid phase molar ratio.

Unfortunately, small changes in the solid phase composition

dramatically affect the material's quality and both electrical

and optical properties. When the lattice-mismatch (Aa/a),

where a is the lattice constant, is greater than approximately

0.5% and layer thickness is greater than the critical

thickness of the material (= 1000A for GaInAs), the strain in

the epitaxial layer is enough to form cracks or dislocations

which can propagate throughout the grown film. Dislocations

appear as a "cross-hatched" pattern and are clearly visible in

the surface of a grown layer as viewed under an optical

microscope. The surface morphology of GaInAs deposited on InP

is directly related to the degree of mismatch in the thin film

relative to the substrate.

Using X-ray diffraction (XRD), it is possible to

determine the lattice constant of a deposited thin film

relative to that of the substrate. With the lattice constant,

one can determine the lattice-mismatch and composition, x, of

a mixed crystal such as GaxInl-xAs. The XRD technique is







92

explained by Cullity[92]. It relies on the periodic structure

of a crystal to scatter incident X-rays in such a way that

some of the scattered beams will be in phase and reinforce

each other to form diffracted beams. Scattered rays will be

in phase if Bragg's law is satisfied:

nl = 2dsin0 (8)



where:


d= (9)
(h2 +k2 +12 )1/2



and I = 1.54051A (for copper Kal), (hkl) are miller indices

(usually (004)), n is an integer (equal to 1 for first order

reflection) and 9 is the angle of incidence of X-rays. Figure

24 shows the XRD spectra from two different GaxIni-xAs/InP

growths. The plot on the left is the spectrum for growth Q015

which is lattice matched. The reason there are two peaks is

because two different X-ray wavelengths close to each other

(Kal and Ka2 from copper) were incident upon the sample. The

spectrum on the right is from growth Q007 and it has a third

broad peak at a lower 28 angle relative to the InP substrate

peaks which when inserted into Bragg's law yields a lattice

constant of a = 5.897A. This is the lattice constant of

Ga 40In 6As (see Table 1 for the lattice constants of InAs and

GaAs), which has a lattice-mismatch of Aa/a = 4.6x10-3 or

0.46% on InP. If a GaInAs/InP sample had an X-ray spectrum













Ga47 In.As / InP


InP


I Q007
y 1
0015 Ga. In, As

C

L cL
4 2 0 Angle (deg)








Figure 24: X-ray diffraction patterns of GaInAs samples grown
on InP (left: lattice-matched, right: lattice-
mismatched) .




Full Text
133
Figure 39; The effect of H2 flow rate through the bubbler on
[Mg] as determined by SIMS.


CONCENTRATION (ntomo/cc)
89
PROCESSED DATA SNR
S Aug SB Ca FILE: aose
Figure 23: Atomic zinc profile of sample Q095 measured by
SIMS.


134
plot is the rapid increase of the Mg concentration with a
small change in the H2 flow rate. This unfortunate phenomenon
has also been reported from Mg doping of GaAs, using the
MCpjjMg source [114].
At low Mg concentrations smooth layers were grown, but at
a concentration of I019cm'3, the surface morphology started to
deteriorate as witnessed by the appearance of stacking faults.
The morphology at a particular Mg concentration was similar
for layers grown on Fe-doped and on S-doped InP substrates.
Several samples were investigated by TEM, using both plan view
and cross-sectional samples. No features indicative of
increased defect densities or precipitates could be detected
at any Mg doping level below I019cm'3, but at higher doping
levels stacking faults were found up to a density greater than
109cm'2.
3.2.3b Mo Diffusion in S-doped InP Substrates
Figure 40 shows SIMS profiles of Mg for layers grown at
different Mg doping levels on S-doped substrates. In addition
to the Mg in the epilayer, Mg spikes, like the one clearly
seen in Figure 38, were observed at the epilayer/substrate
interface, with a subsequent rapid decay of the Mg signal to
the instrumental detection limit, at I015cm'3. For layers
grown with a high Mg doping concentration, ([Mg] > 3xl019cm'3) ,
a Mg concentration of I019cm'3 was attained in the doping
spikes, independent of the Mg level in the as-grown epilayer.


189
The layer structure for the interference filter as shown
in Figure 60 has a very large (200nm) period so that the
satellite peaks are very closely spaced (100 arc-sec) and can
only be resolved by double crystal XRD. The experimental
rocking curve is shown in Figure 62. The epitaxial super
lattice structure is of high quality as demonstrated by the
very large number of diffraction peaks which can be observed
on either side of the main substrate peak: up to five orders
of diffraction on the low angle side and up to 1'2 on the high
angle side. T, the average period thickness calculated from
the fringe spacing is 200nm. The average AlAs concentration
in the superlattice, Xy, as calculated from the angular
spacing between the substrate and zero-order peak is 17%,
which corresponds an aluminum concentration of x=0.34 in
AlxGa,j.xAs.
Figure 62 also shows a simulated rocking curve which is
based on a dynamical scattering theory that is presented in
Hill[152]. This curve was generated using the TEM determined
value for It is encouraging to note that the
simulation does show the correct general trends; the satellite
peaks are decreasing in intensity with both increasing and
decreasing angle. It even predicts the much lower intensity
satellites for the low angle side and the alternating high and
low intensity for the even and odd numbered satellites on the
high angle side.


232
A.4 Calculations
Now that the twenty-two voltage measurements have been
recorded, for a sample of known thickness (the thickness of an
epitaxial layer when growth is on a SI substrate) and with a
value of resistance for the standard resistor used the
equations of reference [157] can be used to calculated the
resistivity of the semiconductor (p in ohm-cm) and the Hall
coefficient (RH in cm3/coulomb), which is negative for n-type
and positive for p-type samples. Then with a known Hall
proportionality factor (r = 1 for GaAs) the charged carrier
concentration (n or p) can be calculated (/i in cm2/volt-sec)
can be calculated as /i=RH/r. These three values (p, /i and n
or p) are all temperature dependant and provide a good basis
of comparison in determining the quality of a grown epitaxial
film.
For investigators who are interested in extremely
accurate values of p, ju and n, several references[159,161-163]
present correlations for samples of non-ideal geometries and
finite contact sizes. It is advised that square samples have
a perimeter length (Lp) less than 1.5 cm and with thicknesses
less than 0.1 cm. Also, the contacts should be as close as
possible to the corners (or edges) of the sample and of size
less than 0.01*Lp. This author advises that if errors of 10%
are tolerable then using the previously mentioned correlations
are not necessary.


Reflectivity
194
Figure 64: Theoretical (solid) and experimental (dashed)
reflectivity spectra taken from wafer center.


6
Basically, the entire area bounded by the solid and dashed
lines is available for use in the design of new III-V compound
semiconductor devices.
The cross-hatched area shown in Figure 1 is the lattice
parameter-bandgap energy space of the quaternary material
Ga^n^ASyP^y. This material has many optoelectronic device
applications due to its wide range of bandgap energy (0.36 to
2.26 eV) and possible lattice constants. Most semiconducting
single crystalline ternary and quaternary materials are
epitaxially grown on a substrate of nearly the same lattice
constant. Hence, the two compositional degrees of freedom
available with the GalnAsP system are important because
presently only GaAs, GaSb, GaP, InP, InAs and InSb are
available for use as substrate materials. For GalnAsP on InP,
or any other heteroepitaxial materials system, a difference of
lattice constant (lattice-mismatch) of greater than 0.1%
between the grown film and substrate leads to, for film
thicknesses greater than the critical thickness, the formation
of structural defects which can degrade device performance.
This problem puts strict demands on the epitaxial growth
technique employed.
1.2 Epitaxial Growth Techniques
The word "epitaxy" is derived from Greek and means
"arranged upon." Epitaxial films of III-V materials are
usually grown or arranged upon substrates with equivalent
crystalline structure and lattice constant. The two most


255
[52] Shin, J., Geib, K. M., Wilmensen, C. W. and Lillental-
Weber, Z. 1989 Amer. Vac. Soc. Symp., Boston, MA, Paper
EMl-ThMl.
[53] Lee, H. H., Racicot, R. J., and Lee, S. H., Appl. Phys.
Lett. 54 (1989) 724.
[54] Hoekje, S. J., Ph.D. Dissertation, Univ. of Florida,
(1990).
[55] Bose, S. S., Szafranek, I., Kim, M. H. and Stillman, G.
E., Appl. Phys. Lett. 56 (1990) 752.
[56] Jones, A. C., Jacobs, P. R., Cafferty, R. Scott, M. D.
Moore, A. H. and Wright, P. J., J. Cryst. Growth 77
(1986) 47.
[57] Blaauw, C., Shepherd, F. R., Miner, C. J. and
SpringThorpe, A. J., J. Electron. Mater. 19 (1990) 1.
[58] Nelson, A. W. and Westbrook, L. D., J. Cryst. Growth 68
(1984) 102.
[59] Blaauw, C., Emmerstorfer, B. and SpringThorpe, A. J., J.
Cryst. Growth 84. (1987) 431.
[60] Blaauw, C., Bruce, R. A., Miner, C. J., Howard, A. J.,
Emmerstorfer, B. and SpringThorpe, A. J., J. Electron.
Mater. 18 (1989) 567.
[61] Manasevit, H. M. and Simpson, W. I., J. Electrochem.
Soc. 118 (1971) 291.
[62] Hirtz, J. P., Larivan, J. P., Duchemin, J. P. and
Pearsall, T. P., Electron. Lett. 16 (1980) 415.
[63] Duchemin, J. P., Hirtz, J. P., Razeghi, M., Bonnet, M.
and Hersee, S. D., J. Cryst. Growth 55 (1981) 64.
[64] Hockly, M. and White, E. A. D., J. Cryst. Growth 68
(1984) 334.
[65] Razeghi. M, The MOCVD Challenge. Vol. 1. Adam Hilger,
Bristol, England, 1989.
[66] Logan, R. A., Tanbun-Ek, T. and Sergent, A. M., J. Appl.
Phys. 65 (1989) 3723.
Ludowise, M. J., Dietze, W. T. and Lewis, C. R., Inst.
Phys. Conf. Ser. 65 (1983) 93.
[67]


4
Since III-V compound semiconductors present a wide range
in values of direct bandgap energy, mobility, and lattice
constant, semiconductor devices have wider ranges of
application. The bandgap energy (Eg) of a semiconductor is
related to the cut-off wavelength (Ag) by the following
equation: Eg(eV) = 1.24/Ag (un). The cut-off wavelength of a
semiconductor is the longest wavelength to which a detector
fabricated from this same semiconductor will respond. Another
degree of freedom available is the ability to form completely
miscible substitutional solid solutions independently on both
the group III and group V sublattices. In other words, not
only simple binary III-V compounds, but also III-III'-V or
III-V-V ternary and III-III'-V-V quaternary single
crystalline semiconductors such as Al^a^^s, GaxIn.,.xAs,
GaASyP^y and Ga^n^^SyP^y can be created. By using ternary
and quaternary semiconductors, it is possible to vary the
physical and electrical properties of these materials
continuously between the property limits of the constituent
binary compounds listed in Table 1. A plot of the lattice
constant versus the bandgap energy (at 300 K) for III-V
compound semiconductors is shown in Figure 1. Solid dots
indicate binary compounds, solid lines connecting dots
represent direct bandgap ternary solid solutions, and dashed
lines connecting dots represent indirect bandgap ternary solid
solutions.


56
2.5.2 InP
The method for determining the optimum growth conditions
for undoped InP on InP, n-type InP on InP and p-type InP on
InP will be discussed in the following sections. Before
presenting the results of this operating parameter sensitivity
and optimization study, the criteria for good epitaxial layers
are: mirror-like low defect density surfaces, uniform growth
rates over a large area substrate, less than n = lxl015cm'3
background (ND-NA) doping levels for undoped InP with
corresponding high room temperature electron mobilities (ju >
3000cm2/volt-sec), experimentally convenient growth rates of
1-2/xm/hour, and for doped layers and heterostructures,
atomically abrupt junctions and interfaces, respectively. All
of the above criteria have been determined by using various
thin film characterization techniques available on the campus
of the University of Florida.
2.5.2a Undooed InP
An extensive optimization study has been performed on the
MOCVD growth of undoped InP on InP. The results of this study
have been applied to aid in the determination of optimum
growth conditions for n- and p-type InP, undoped, n- and p-
type GalnAs, and undoped, n- and p-type GalnAsP on InP; these
optima will be discussed later. The timing and sequence of
events for a typical MOCVD growth are presented in the
Appendix B of this text.


Reflectivity or Transmission
195
Figure 65: Reflectivity and transmission (dashed) spectra
taken near the edge of the wafer.


157
Figure 53: The effect of growth temperature on Zn (from DEZn)
incorporation in InP.


187
Figure 61:
TEM cross-section photo of a
etched edge of stack MBE464.
cleaved (110) and


101
Figure 28;
TEM micrograph of alternating layers of InP (light)
and GalnAs (dark) at 247,500X.


67
Q054 which shows InP layers grown with different MFTHIn for
different lengths of time separated by thinner dark GalnAs
marker layers. Figure 9 also shows the excellent abruptness
of interfaces between two different materials grown at 600C
and 7SLM.
Another analytical technique used to evaluate material
quality is an electrochemical C-V profiler. The C-V profiler
(BioRad, Model PN4200) is thoroughly explained by Blood[89].
It uses a Schottky and ohmic contacts in an electrochemical
cell filled with 0.1M HC1 (for InP) to slowly etch away
material while it measures the change in capacitance as a
function of changing applied voltage. From this C-V data, the
net carrier concentration can be calculated (n = ND-NA or p =
Na-Nd, where NA and N0 are the number of acceptors and donors
per cm3) and plotted as a function of depth. A C-V profile of
growth Q056 which is undoped InP on n-type (8xl018cm'3) InP is
shown in Figure 10. From this figure one can get a rough
estimate of 1.2/im for the layer thickness. Also, the
epilayer-substrate interface abruptness can be assessed. Good
uniformity of the carrier concentration throughout the layer
can also be observed. Finally an average value of the
background carrier concentration for undoped InP of
n=3xl015cm'3 can be obtained. It is evident that a lot of
useful information can be obtained from an electrochemical C-V
profile, the only drawback to this technique is that it is
destructive.


204
together respectively to allow parallel operation with RI
modulation capability, but no switching function.
4.2.5 Conclusions
A functional interference filter operating at a central
wavelength of A.=1300nm has been fabricated from the GaAs-
AlGaAs material system by MBE. The stack was doped alter
nately p- and n-type hence creating a novel optoelectronic
switch element. The I versus V characteristics were measured
and as predicted, only the PNPN device demonstrated electrical
bistability. As a detector, photocurrent measured for the
PNPN device peaked near the GaAs bandgap and changed direction
as a function of wavelength for the NPN device. As expected,
neither device showed any photoresponse for wavelengths
greater than lOOOnm.
The expected spectral shift of the reflectivity in
response to current modulation could not be verified because
of local thermal shifting of the characteristics resulting
from the relatively high holding current measured in these
unoptimized PNPN devices. Multiple quantum wells are also
being considered to take advantage of electric field effects.
The p-n junctions can also be operated in a parallel mode,
(with side contacts) circumventing the necessity to switch the
thyristor structure (inherent in the series mode) to the "ON
state in order to achieve RI modulation by carrier injection.


76
material from in the Japan Oxygen MOCVD was measured to be
61,800cm2/volt-sec for a sample grown at a temperature of
600C. There is a lot of scatter in the mobility data. The
scatter may be due to the fact that different material sources
with different impurity levels were used to grow these layers.
Also, the reactor was not changed/baked-out after each growth.
However, the average mobility values indicate that higher
quality material can be grown at 600C and this is in
agreement with previously discussed characterization results.
Based on all of the observed characterization trends,
apparent optimum MOCVD growth conditions for undoped InP on
InP have been determined. These growth conditions are
presented in Table 5 along with the room temperature and 77 K
electrical properties of undoped InP grown at these
conditions. These optimum conditions are significant because
they are the basis for estimates of growth conditions for
intentionally doped InP, unintentionally and intentionally
doped GalnAs and GalnAsP, both lattice-matched to InP.
2.5.2b Growth of n-tvne InP Using H:S
The Japan Oxygen MOCVD system was designed with the plan
of using dilute H2S as an n-type doping source. A cylinder
containing a mixture of lOOOppm H2S and the balance ultra high
purity hydrogen (Matheson, ULSI grade) is connected to the
MOCVD system. The flow rate range measured at room
temperature and 760 torr for this mixture, as shown in Table
2, is 0-50cm3/min (seem). Epitaxial layers of InP were grown


33
of GalnAs by MOCVD. Zinc doped GalnAs carrier concentrations
are reported to decrease with increasing growth temperatures
over the range p = 1017 I018cm*3. The opposite behavior is
observed for sulphur doped GalnAs and this trend is confirmed
by Logan et al.[66]; a carrier concentration of 102cm'3 is
reported for a growth temperature of 525C and at nearly
identical conditions except Tg = 625C, n = 8 x I017cm'3. Wide
ranges of both p- and n-type doping are attainable for GalnAs
on InP by MOCVD which is useful for device applications.
2.3.3 GalnAsP/InP
The quaternary solid solution GaxIn1.xAsyP1.y is an alloy
semiconductor which can be lattice-matched to InP and GaAs
substrates. GalnAsP is a direct bandgap semiconductor (when
lattice-matched to GaAs or InP) which can be a very efficient
light emitter over the wavelength range 0.65 0.87/xm
(lattice-matched to GaAs) and 0.92 1.65jttm (lattice-matched
to InP) Very little work has been reported on quaternary
growth on GaAs substrates[67,68] due to greater interest in
GalnAsP alloys lattice-matched to InP substrates for optical
fiber device applications. Optoelectronic devices operating
at 1.3 Mm or 1.55/im wavelength regions have immediate
commercial applications because light transmission through
silica fibers exhibits low loss at 1.3/xm and low dispersion at
1.55jum.
The MOCVD growth of GalnAsP alloys on InP substrates was
first reported using ethyl alkyls in a low pressure


64
E
3.
W
(O
Ui
z
*
o
X
I
O 5 10 15 20 25
DISTANCE FROM LEADING EDGE (mm)
E
(0
(0
UI
z
*

X
H
-20 -10 0 10 20
DISTANCE FROM CENTER (mm)
Figure 7: MOCVD thickness uniformity study. Top: axial
variation, bottom: radial variation at 12 mm from
the leading edge.


226
contacts but also be isolated from each other. A one inch
square by .025" thick alumina plate with a 2E-06 cm thick
chromium adhesion layer and a 2E-05 cm thick gold conductive
layer was obtained from Materials Research Corporation. This
plate was cut into 3/8" by 1" rectangular pieces by using a
Micromatic precision watering machine equipped with a diamond
impregnated blade. Onto these plates a pattern of four pads
was drawn with a permanent ink marker. The plates were etched
in a gold etching solution of KI:I2:H20 in the respective
volumetric ratios 4:1:40 for ten minutes or until the gold was
removed. Then the plates were etched in a one part potassium
to 3 part ferrocyanide by volume solution until the gray
chromium layer was removed exposing the underlying non-
conductive alumina. The permanent ink was finally removed
with acetone revealing four isolated gold pads. To each of
these four pads were soldered a 4 cm piece of non-magnetic
wire with an Amphenol Incorporated female type plug connector.
To the center of the alumina mounting plate the small
semiconductor sample can be attached with a portion of Crycon
grease (which does not out-gas at low temperatures). The
grease keeps the wafer stuck in place to the alumina plate.
A very fine 1 mil thick gold wire must be ultrasonically bound
from each ohmic contact to the edge of the adjacent gold pad.
Request from the operator in the microelectronics lab that the
wires are as short as possible so that they can withstand
higher currents before burning up. This last step completes


150
-6
DEZn flow rate (x10 Moles/minute)
Figure 49; Total zinc incorporation in MOCVD InP from both
bubbler and cylinder sources as determined by SIMS.


Ill
Angle (deg)
Figure 34; X-ray diffraction patterns of quaternary films
grown on InP substrates (left: lattice-matched,
right: lattice-mismatched).


241
reads 490 torr (indium bubbler), open AV207, adjust NV203 so
the reactor is at 80 torr. Flow the MO to the vent (TMIn
only). Make sure AV63 is open and AV64 is closed. Adjust the
pressure of the bubblers (PI60 = 490 torr) Open the MO
bubbler (at first open the outlet valve and then open the
inlet valve). Close AV62 and then check the pressure of the
bubbler, set it to be 500 torr and stable.
Heat the substrates by setting the controller temperature
to 650C (172R) Turn on the RF plate and power switches
(auto mode) Flow the toxic gas PH3 to the reactor for 10
minutes. Check H2 detector point 8 and constantly monitor it.
Open the PH3 cylinder slightly and adjust the outlet pressure
to 0.5 kgf/cm2 (use the black gloves and keep all machine
doors completely closed until all PH3 is gone) Open AV22 and
then open AV21. Set MFC20 to maximum (300 seem). Check H2
detector point 6 then return it to auto. Start the program in
10 minutes if the temperature is stable. Close the PH3
cylinder at an appropriate time (this depends on flow rate and
cylinder pressure), possibly before the end of the growth.
Close the MO bubbler after the growth ends by first opening
AV62 and then close the inlet valve and last, close the outlet
valve. Reduce the flow rate of the MO to 30 seem. End the
program always 3 minutes after the MO goes to the vent. Turn
off the RF power, plate switches and filament of the RF
generator. When the temperature of the substrate is less than
400C (100R), turn off the RF generator power switch. When


Response [V]
203
Wavelength[nm]
Figure 70; Photoresponse spectrum for a NPN device


127
DMCd desorption from the growing layer at high temperatures.
They also measured the diffusion coefficient of Cd from SIMS
profiles and report a value (at 600C) of Dcd = 0.4-l.4xl015
cm2/sec. Yang et al.[112] presented data for one DMCd
simultaneous InP deposition on three different orientations of
InP. They reported carrier concentrations of p = 1. Ixl017cm'3,
p = 8.7xl017cm"3 and n= lxl016cm'3 for the (100), (111A) and
(111B) orientations, respectively. The Cd incorporation is
most efficient on the (111A) orientation possibly due to the
increased number of Indium vacancies on this type of surface.
Overall Cd seems to be a fairly good p-type dopant due to its
wide incorporation range and low bulk diffusivity, but it is
not a suitable dopant for low pressure growth applications due
to the deterioration in surface morphology.
The use of Cp;>Mg[58] and MCp2Mg[60] as precursors for Mg
p-type doping of InP has been reported. Nelson and Westbrook
[58] performed atmospheric pressure adduct growths of InP and
obtained hole concentrations from p = 3xl016 to 2xl018cm'3 at
CpjMg vapor flows of lxlO'8 to lxlO'7 moles/min. The doping
level varied as the square of the dopant vapor flow, therefore
incorporation is mass transfer controlled. At lower dopant
flow rates, no Mg was detected by SIMS. They also report a
diffusion coefficient (at 600C) of DMg = 2-4xl0'15cm2/sec
which is based on SIMS profiles. In Blaauw et al.[60] similar
"threshold-like dopant incorporation using MCpjMg during low
pressure growths (76 torr) was reported. We also observed


143
10
20
10
16
epilayer/substrate
interface
C-V Profiling
SIMS
Depth (m)
Figure 45: Mg and hole concentration profiles of InP:Mg on
InP:Fe.


APPENDIX A
THEORY AND OPERATION OF THE LOW TEMPERATURE HALL EFFECT SYSTEM
A.l Theory
If current is flowing in the x-direction in a probed
semiconducting sample and a magnetic field is being applied in
the z-direction, the flowing charged particles will be
deflected in the mutually perpendicular y-direction. To
maintain the steady-state flow of particles through the
device, an electric field will be induced in the y-direction.
The establishment of the electric field Ey, (as shown in
Figure 78) is known as the Hall effect[156].
By performing an experiment, like the one mentioned
above, with a known applied magnetic field (Bz) and current
density (Jx) one can measure the induced voltage (VH) From
these values, the proportionality constant:
Rh = Ey/(JX*BZ) (37)
can be calculated. With RH, the Hall coefficient, and a
previously measured value of resistivity, p, for the sample,
it is now straight forward to calculate the majority carrier
concentration and carrier mobility of the semiconductor.
Knowledge of these two temperature dependent values is a way
of characterizing the type (n or p) purity and functionality
of a semiconductor.
220


121
material and reducing the net hole concentration. Theories
such as this one are proposed and considered in the model
development portion (section 3.4) of this chapter.
3.1.3 Liquid Phase Epitaxy
The LPE technique has been successfully used to fabricate
many optoelectronic devices which use p-InP layers as part of
the device structure. Zinc[2], magnesium[2], cadmium[2] and
manganese[102] have been successfully used to achieve p-type
conversion in LPE grown InP and related materials. The
dopants are added to the In melt in pure metallic form, or as
alloys. The distribution coefficients for Zn, Cd and Mg in
InP by LPE are 0.7, 0.001 and 0.05-0.5, respectively[2]. Zinc
is the most commonly used dopant for several reasons. It
controllably incorporates to form p-type material with hole
concentrations from p = lxlO16 to 2xl018cm'3, whereas with Cd,
p-doping levels from p = 5xl016 to 2xl018cm3 exist[2]. With
Mg, hole data were uncontrollably scattered from p = 2xl017 to
6xl018cm3[2]. Mn is only capable of doping material at low
p-levels from 3xl016 to Ixl017cm'3[l02]. Cd diffuses less than
Zn and much less than Mg. But, the equilibrium vapor pressure
of Cd (at 625C) is eight times greater than that of zinc (200
vs. 25 torr). Hence, Cd rapidly evaporates from the heated
melt, which results in varying doping level profiles and
contamination of neighboring melts. Consequently, Zn is the
preferred p-type dopant for LPE growth of InP.


176
electrically inactive interstitial zinc, Zn{/ and the complex
VpZnInVp gradually increase in relative concentration as ZnIn'
decreases (p < [Zn]). The model also shows that the complex
Zn,nVp is essentially non-existent at all zinc partial
pressures.
The trends shown in Figure 57 agree very well with the
theories upon which the model was developed. However, other
theories such as compensation due to ionized interstitial zinc
and zinc-hydrogen passivation must also be evaluated by
developing new model equations. Also, this work would be more
convincing if less constants had to be fitted to experimental
data. Unfortunately, InP has not received as much attention
in the literature as GaAs and, consequently, little or no
information on lattice dilation or high temperature Hall
effect measurements are available. Perhaps future researchers
can fill these important gaps. This method could also be
applied to understanding the effects of varying the growth
temperature and the V/III ratio on zinc incorporation in InP
and extend the model to analyzing cadmium and magnesium doping
of InP. Hopefully by completing this puzzling topic with the
aid of modelling, an "ideal" p-type dopant for InP will
eventually be found.


149
CO
£
c;
c
O'
I HM
4-*
CO
CD
O
c
o
CJ
c
N1
Depth (ym)
Figure 48; SIMS profile of atomic zinc in InP grown on InP:S
(sample B316).


77
Table 5
Undoped InP Growth Conditions and Electrical Properties
Total H2 Flow Rate
= 7 SLM
Growth Temperature
= 600C
TMIn Mole Fraction
-41

o
H


O
II
V/III Ratio
= 419
Growth Pressure
=80 Torr
at 300 K :
nd na
= 2*1014 cm'3 (lowest)
nd na
= 2 1015 cm'3 (average)
^300K
= 3461 cm2/volt-sec(highest)
^300K
= 2800 cm2/volt-sec(average)
at 77 K :
nd na
= iio14 cm'3 (lowest)
nd na
= 2 1015 cm'3 (average)
^77K
= 61800 cm2/volt-sec (highest)
^77K
= 40500 cm2/volt-sec (average)


206
9 X
1025 A
1050 A
GaAs:Si(n = 5xl017 cm'3 )
A13Ga 7 As:Si(n = 2xl017 cm 3 )
*
GaAs:Be(p = 2xl017 cm3 )
Al 3Ga-7As:Be(p = 2xl017 cm3 )
GaAs buffer
n+ GaAs
SUBSTRATE
Figure 71; MBE572 targeted structure.


32
of TEGa for a fixed TEIn or TMIn flow. Both the composition
and growth rate are independent of AsH3 flow for fixed metal
organic flow. The growth rate is independent of growth
temperature (500-650C), but the composition can be slightly
affected due to slight differences in gallium and indium
source cracking efficiencies. If the composition of a layer
is different from the lattice-matched value, this layer is
mismatched. The lattice-mismatch between layer and substrate
is defined as Aa/a = (aL a)/ a when aL is the measured
room temperature (strained) lattice parameter and a is the
lattice parameter of the substrate. Razeghi et al.[38] have
reported that the mobility of a semiconducting layer is
dependent on the amount of mismatch in the layer relative to
the substrate. At optimum growth conditions, a lattice
matching of Aa/a < 0.04% has been achieved resulting in Hall
mobilities of 12,000 (300 K), 100,000 (77 K) and 260,000cm2/
volt-sec (2 K) and background carrier concentrations of 0.7 -
1.0 x I015cm'3[65]. Such high mobilities at 2 K are explained
by the existence of a two-dimensional electron gas formed at
the interface between undoped InP buffer and GalnAs layers and
are indicative of superior material quality.
It is evident that MOCVD can be used to produce undoped
Ga 47In 53As lattice-matched to InP, with very high quality
electronic properties. Intentionally doped both p- and n-type
Ga 47In 53As on InP is also producible. Razeghi[65] presented
data on p-type doping using DEZn and n-type doping using H2S


57
After a growth has been performed, the first property
studied is the surface morphology of the thin film viewed
under an optical microscope. The thickness of the film can
also be determined by cleaving a sample and etching it for two
minutes in a 6:4:50 solution (measured by weight) of potassium
hydroxide, potassium ferricyanide, and deionized water to
delineate the grown layer from the substrate and then viewing
it under a microscope. A Nikon Optiphot microscope equipped
with Nomarski phase contrasting was used to view the surface
and delineated layer of the third growth in the machine, Q003.
Polaroid photos of the images observed at 2000X magnification
are shown in Figure 5. As shown, the surface morphology has
a ripple in it and this is sometimes referred to as an "orange
peel" surface. The side view photo of the sample was taken at
the leading edge of the sample with respect to the gas flow
direction. As shown, the layer thickness is clearly greater
(1.5/Ltm) near the corner compared to the center (1.0/xm)
probably due to the lower growth rate right on the corner
which is at a different crystalline orientation, the (111)
orientation.
Several experimental parameters have been varied in an
attempt to improve the surface morphology of undoped InP grown
on InP. The growth temperature was varied from 450 to 750C
and it was found that layers grown at temperatures between 575
and 700C had better morphologies than those grown at higher
or lower temperatures. At lower temperatures lack of


145
3.2.4 Conclusions
The use of MCpjMg as a p-dopant source for MOCVD InP has
been investigated and the Mg incorporation was found to be
non-linear. For Mg concentrations above I019cm3 the layer
morphology deteriorated (stacking faults were observed by
TEM). Extended diffusion of the Mg in the grown layers into
both InP:S and InP:Fe substrates was observed. It appears
based on the shapes of C-V and SIMS profiles, that immobile
Mg-S and Mg-Fe complexes form in InP. At doping levels
exceeding I019cm'3, significant electrical compensation takes
place; the net hole concentration decreases with increasing
[Mg] in the layers. Because of these abnormal incorporation
and doping characteristics, Mg may be considered an unsuitable
dopant for InP in some device applications.
3.3 Experimental DMCd and DEZn p-Tvpe Doping of MOCVD InP
3.3.1 Introduction
In addition to Mg which was discussed above, the use of
Cd and Zn as p-type dopants for MOCVD InP has been studied
while this investigator was a visitor at BNR. During these
experiments, the metal organic (MO) compounds dimethylcadmium
(DMCd) and diethylzinc (DEZn) have been used as Cd and Zn
sources, respectively. The effect of MO flow rate on layer
morphology, atomic incorporation and electrical activation in
InP has been determined by Nomarski optical microscopy, SIMS,
Hall and C-V measurements. For DEZn, the effect of varying


106
shown in Figure 30. Some of the peaks are not as sharp or
well defined as the others, but clearly quantum energy level
transitions have been realized by using PL on this sample. By
using the Kronig-Penney theory[93], which is a simplified
mathematical representation of the periodic potential function
for electron motion, the following equation can be derived:
E
n
n2 h2
8m *L2
(10)
where En is energy in eV, n = 1,2,3... quantum levels, h =
6.624xl0'34J-sec (Planck's Constant), m* is the effective mass
(.041 m0, for electrons and 0.50 m0, for heavy holes, m0 =
9.109xl031Kg) and L is the well thickness (A). Using
equation (10) one can find the theoretical confinement energy,
AE = E^-Eg = Ec1+Ev1 = 6.1911x10*19/L2 eV (where EC1 and EV1
are the first energy levels of the conduction and valence
bands, respectively). Of course the measured AE is the
difference between the position of each individual peak, Epealc,
and the bandgap, Eg, of Ga 47In 53As at 4.2 K (0.812eV). A
plot comparing the theoretical and experimentally measured
confinement energies versus GalnAs well thickness for growth
Q136 is plotted in Figure 32. The agreement is good for
thinner wells but deviates from theory for thicker wells. In
conclusion, the growth of low dimensional structures using the
Japan Oxygen MOCVD system is possible and has been confirmed
by TEM, SIMS and low temperature PL results.


120
convenient near unity effective distribution coefficient;
K(Zn) = 0.6[99], K(Zn) = 1.4[100]. Zn is a fast diffuser in
InP (D=(l-6)xl0*13cm2/sec[58]) so in situations where abrupt
doping profiles are required, Cd doping is sometimes used.
The distribution coefficient of Cd is low though, K(Cd) = 0.1-
0.2[99]. This means that as the crystal is pulled, the Cd
concentration in the melt increases and as a result, the
atomic concentration in the solid also increases. Zn doped
InP bulk crystals have been fabricated by LEC and the
characteristics of the material have been reported by Roksnoer
and Van Rijbroek-Van Den Boom[101]. They report a Zn
distribution coefficient of K(Zn) = 0.90 which is constant
over the atomic concentration range of [Zn] = 7xl017 to
2xl019cm'3. They also report that their material was
dislocation free for concentrations greater than [Zn] =
lxl018cm'3. The most interesting data, however, were their
Hall carrier concentration results. Their observation was
that for crystals containing less than 5x1018 zinc atoms/cm3,
p [Zn] but, at [Zn] > 5xl018cm3, p < [Zn]. At the highest
atomic level [Zn] = 2.4xl019cm'3, p = 1.5xl01scm'3. Thus, a
significant part of the zinc atoms were electrically inactive.
They propose that precipitates had beh formed in the heavily
doped crystals. But in only the most heavily doped one ([Zn]
= 2.4xl019cm3) were small pits detected by preferential
etching. It is possible that Zn at high doping levels also
incorporates in some unknown form as a donor, compensating the


215
estimates or a combination of all three are different from the
expected values used to determine the layer structure and
growth conditions. To determine the possible sources of
error, separate parts of the MOCVD sample Q240 were given for
both TEM and High Resolution XRD analysis. Until these
characterization results are available, no further MOCVD
growths or device processing and testing are planned for the
interference filter project.


69
The effect of growth conditions on the background carrier
concentration of undoped InP has also been studied. The
growth pressure, total hydrogen flow rate, and TMIn mole
fraction apparently have little effect on the background
carrier concentration. The growth temperature and V/III
ratio, however, both have a strong effect. The background
carrier concentration measured by the C-V profiler with
respect to growth temperature is plotted in Figure 11. At a
low growth temperature, high carrier concentrations may be
associated with phosphorus vacancies in the InP crystal
(impurities are complexing with these vacancies). At higher
temperatures, electrically active source impurities are
incorporating more efficiently. It appears that based on
background concentrations, the optimum growth temperature
should be between 575C and 625C for undoped InP. The
undoped InP carrier concentration is also strongly affected by
the V/III ratio as shown in Figure 12. Based on this figure,
high purity material can be grown with high PH3 flow rates.
Another analytical technique which is commonly used to
evaluate material quality is PL. The technique is explained
well by Dean[90]. Briefly, it typically uses laser light
energy incident upon a sample to stimulate the emission of
photons at discreet energy levels. These discreet energy
levels can be detected and related to impurities, defects, or
host crystalline energy level transitions. The resolution of
closely spaced transitions can be improved by reducing the


159
p (cm
Dopant Partial Pressure (atm)
Figure 54: The effect of dopant partial pressure on hole
concentration for InP:Mg, InP:Cd and InP:Zn.


20
hydroxide diluted in water where neutralized salts and acids
are products. Some gases are nearly insoluble in water,
though. Thermal crackers basically operate by heating the
exhaust stream to approximately 950C to thermally decompose
toxic gases into less toxic compounds. Clogging and also
insufficient heat transfer at high flow rates are problems
with this technique. Dry scrubbers use powders such as
activated carbon or diatomaceous earth mixed with iron
chlorides to react with the toxic gases. This technique also
has problems associated with efficient gas solid contacting
and disposal of toxic corrosive powders. A final technique,
the incinerator or "burn box," operates in such a way that
gases are mixed with a fuel gas and oxygen and then ignited by
a pilot flame or electric igniter[29]. All of the four
scrubbing techniques have their individual problems. The
scrubbing system should ideally be a combination of two or
more of the individual systems in case one system fails.
As discussed, MOCVD of III-V compound semiconductors
presently involves the use of highly toxic and explosive
source gases. There has been some work on the use of less
toxic sources for MOCVD such as tertiarybutylphosphine (TBP)
[30] instead of phosphine (PH3), (the threshold limit value
(TLV) of PH3 is 0.3ppm while that of TBP is greater than
lOOOppm), but the material grown with these new sources is
generally inferior. In any event, there must always be an
integrated safety component to all MOCVD systems.


34
system[63]. This same team later reported growth of nearly
the entire quaternary composition range lattice-matched to InP
and presented device test results of broad area and stripe
lasers fabricated using six different quaternary
compositions[69]. After these initial reports, numerous
quaternary related papers have been published on novel growth
techniques, relationship between gas phase growth conditions
to solid phase compositions, electrical and optical material
properties, and device applications. Each of these topics
will be reviewed in the following paragraphs.
Growth techniques used for depositing quaternary alloys
are direct extensions of techniques used for growing InP and
GalnAs. The growth rate is similarly proportional to the sum
of the partial pressure of TEGa and TMIn, and is independent
of the phosphorus and/or arsenic partial pressure. Similar to
the growth of GalnAs, the solid phase group III composition is
essentially equal to the gas phase metal organic composition
introduced to the MOCVD reactor. However, the behavior of
incorporation of group V elements is different and much more
difficult to control since both arsine and phosphine are
required and they do not incorporate with the same
probability. It is much more difficult to incorporate
phosphorus than arsenic at a fixed growth temperature since
the cracking temperature of PH3 is higher than AsH3. To
alleviate this problem, some workers have used precracked
PH3 [38,70] or in-situ adduct formation techniques[71] (which


138
related to the substrate iron doping level (0.3 10xl017cm'3)
and suggests the existence of an immobile Fe-Mg complex.
3.2.3d Electrical Characteristics
Electrochemical C-V profiling measurements were carried
out for layers grown on S-doped substrates, and flat profiles
were typically obtained. A typical C-V profile is shown in
Figure 42. As shown, the hole concentration for this sample,
growth B229, is p=2xl018cm'3 and the layer thickness is roughly
one micron. The hole carrier concentration determined by C-V
profiling is shown in Figure 43 as a function of the Mg (SIMS)
concentration in the layers. Up to a concentration level of
[Mg] 2xl018cm'3, an approximately linear relationship between
C-V profile hole concentrations and SIMS measured atomic Mg
concentrations (on both S- and Fe-InP) was observed. As [Mg]
was increased further, the net hole concentration gradually
decreased from a maximum value of p 2xl018cm'3 to a value of
p 8xl016cm'3 [Mg] 3xl019cm3. This suggests that at high
levels, Mg is acting as a donor, compensating InP.
Hall measurements were performed on layers grown on Fe-
doped substrates but the results were not considered reliable
due to the uncertainty in depth from the extended diffusion of
electrically active Mg atoms. Electrochemical C-V profiling,
using the front contact method, was also performed on the
InP:Fe layers. The SIMS and C-V results for a sample doped at
[Mg] = 2xl018cm'3 are shown in Figure 44. The agreement
between SIMS and C-V profiles confirms that all the Mg is


243
heat tapes and exit the clean room. Turn off the scrubber,
check H2 supply pressures (500 psi minimum).
B.5.4 Reactor Cleaning
Purge the reactor overnight with hydrogen after a four
hour bake at 900C and 35 torr (4 slm Hydrogen flowing to the
reactor) Disconnect the water line (inlet and outlet) close
all water valves. Divert gas flow by closing AV119, 125, 200
then opening AV118, 124 and turn off the heat tapes.
Disconnect the thermocouple cable connector. Evacuate the
reactor using RP1 through the load-lock to 3 torr. Close the
doors and open slowly MV207 to let air into the reactor to a
pressure of 760 torr (there is a chance that a fire may occur
in the reactor tube). Repeat evacuation and fill four more
times.
Using the particle masks, disconnect the reactor and
thermocouple. Cover all with aluminum foil. Clean the MO
inlet tube, susceptor bed and reactor immediately, using the
wafer tray procedure (reactor requires 2 hours of DI rinsing
though). Clean the thermocouple tube using methanol, clean
room paper and a clean razor blade. Clean the exhaust area
(wearing particle mask) with methanol and clean room paper.
Dispose of waste in toxic waste garbage can (fill out proper
forms).
Install the clean thermocouple tube, and reactor with
clean graphite susceptor and susceptor bed. Before connecting


109
Figure 33: Surfaces of quaternary films grown on InP (top:
lattice-mismatched, bottom: lattice-matched).


REFLECTIVITY (A. U.)
214
WAVELENGTH (microns)
Figure 77; Experimental reflectivity spectrum for sample Q240.


83
The relationship between the C-V measured hole concentration
and the corresponding DEZn partial pressure used for the
growth is plotted in Figure 19. As indicated, controllable p-
type doping from 2xl017cm*3 to 3xl018cm*3 was realized. Hall
measurements were attempted on all of the p-type layers grown
on semi-insulating InP substrates, but the alloyed indium
contacts were generally not ohmic in nature. Hence, only the
C-V carrier concentration results from films grown on n-type
substrates were used. The room temperature Hall carrier
concentrations of semi-insulating samples with ohmic contacts
agreed quite well with the C-V carrier concentrations on n-
type substrates.
The effect of several growth conditions on the p-type
doping of MOCVD InP was investigated. The following growth
parameters: DEZn mole fraction, growth pressure, V/III ratio,
growth temperature, and total hydrogen flow rate, were varied
during individual experiments. It was later discovered that
the effect of the extended diffusion of atomic zinc in InP
during growth influenced the spatial variation of zinc's
incorporation. Basically, any variation in incorporation that
might have existed due to variations in growth conditions
during the deposition, were negated by the rapid diffusion
rate of zinc. For example, Figure 20 shows a C-V profile of
growth Q105 during which the DEZn mole fraction was varied
from 0.2 to l.OxlO'5 and the C-V hole concentration of


BIOGRAPHICAL SKETCH
Arnold John Howard was born on January 6, 1963, in
Huntington, NY, which is located on the north shore of Long
Island. He grew up there and graduated from Huntington High
School in 1981. He then enrolled at the University of
Pennsylvania in Philadelphia, Pennsylvania, and graduated in
1985 with a B.S. degree in chemical engineering. While at
Penn he met Robin Lustig, another Penn Quaker, at a party that
he and his friends had stumbled upon; Robin later became his
wife on April 7, 1990. After graduating, the author spent six
weeks in nine different countries of Europe. After all that
traveling, he enrolled at the University of Virginia in
Charlottesville. As part of his master's program, the author
spent a summer at the Oak Ridge National Laboratory performing
research on the separation of sugars. He received his M.S. in
chemical engineering from UVa in January of 1987 and then went
further south to begin the Ph.D. program at the University of
Florida in Gainesville. The author had the good fortune of
choosing a research project which involved being a visitor at
Bell Northern Research in Ottawa, Canada. (Gainesville was
too hot!) The author plans to work at Sandia National
Laboratories in Albuquerque, New Mexico, upon completion of
his Doctor of Philosophy degree in chemical engineering.
262


27
own technique of wafer preparation using a combination of the
methods reviewed above.
Once an MOCVD system has been optimized for growing high
purity InP and a proper substrate vendor, orientation and
cleaning procedure have all been selected, most authors report
that the source purity of both the metal organic indium and
phosphine have the strongest influence on background carrier
concentrations and mobilities. The initial work on InP growth
[12,33-34] reported room temperature carrier concentrations of
n=0.17 to 1.4 x I016cm*3 and 300 K and 77 K electron mobilities
of 3500-4200 and 16,000-36,000cm2/volt-sec, respectively.
After two decades of technological advancement in purification
techniques and machine design, the highest reported 77 K
mobility for undoped InP is now 305,000cm2/volt-sec with a
corresponding carrier concentration of n= 5 x I013cm3[55].
Based on low temperature PL it appears the dominant residual
acceptor in MOCVD InP is zinc[55]. For many years both carbon
and manganese[47] have also been reported as compensating
acceptors and silicon has been reported as the dominant
donor[56]. The recent work of Bose et al.[55] caution against
identifying PL peaks as carbon since the transverse optical
phonon replicas of the free-exciton recombination occur at the
same energy as carbon. For most teams, however, 300 K and
77 K mobilities of 4,700 and 80,000cm2/volt-sec, respectively,
and a carrier concentration of n= 1 x 10ucm'3 are typical for


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
,\.
Anthony J. SpringThorpe
Manager of Epitaxy
Bell Northern Research, Canada
This dissertation was submitted to the Graduate Faculty
of the College of Engineering and to the Graduate School and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
December 1990
/ Loius" Cl
A^Winfred M. Phillips
v Dean, College of Engineering
Madelyn M. Lockhart
Dean, Graduate School


94
with more than two large peaks, like sample Q007, then no
further characterization was performed on the sample.
For lattice-matched Ga 47In 53As on InP, the C-V profiler
is a characterization technique which was routinely used. A
C-V profile of growth Q005 is shown in Figure 25, which is an
undoped GalnAs layer deposited on n+-InP. As shown, the grown
layer thickness is approximately 0.7/xm and the substrate-
epilayer electrical interface is extremely abrupt. The
background doping level of this sample varies slightly from
2xl015 to 8xl014cm'3, but the average is very low which is
indicative of high quality material. The C-V profiler is less
accurate for profiling lattice-mismatched GalnAs/InP samples
and consequently was not used for some films.
Another technique that was used to investigate the purity
and also optical properties of GalnAs/InP samples is low
temperature (4.2 K) PL. Figure 26 shows the PL spectra for a
precisely lattice-matched Ga 467In 533As on InP grown by MOCVD.
Before PL was performed the degree of lattice-matching was
determined by XRD. This was necessary as the PL technique is
rather time consuming and it also requires a great deal of
operator expertise. Also, the PL set-up was not always
equipped to test the GaxIn1.xAs on InP samples as they require
a detector which responds to light over the wavelength range
0.9 to 1.7 microns. The spectra are quite valuable, though, as
a lot of information can be derived from them. Based on the
peak position in wavenumbers (inverse wavelength)


153
Figure 51; The relationship between SIMS zinc concentration
and hole concentration in InP:Zn.


218
measurements and lattice parameter measurements. With this
added information, it would be possible to estimate more of
the unknown equilibrium constants which are required in the
model. This would improve the accuracy of the unknown
constants that are estimated by a non-linear regression
analysis program which uses the experimental SIMS and hole
concentration data. It is also recommended that other point
defect reactions be considered to validate or invalidate the
model which was presented in this dissertation. Mechanisms
such as the formation of a hydride complex as proposed by Cole
et al.[128] should be considered. In addition, the charged
state of all the point defects at room temperature is another
uncertainty that must be considered. The study of all the
possible doping mechanisms is hence proposed. Finally, it is
recommended that future researchers apply the point defect
analysis technique to understanding the cadmium and magnesium
doping data sets for MOCVD InP which are presented in this
dissertation. Cadmium appears to have an understandable
linear incorporation dependence on its precursor partial
pressure, but, magnesium has a dependency that suggests strong
compensation and possibly n-type conversion at high source
pressures.
A novel optoelectronic device incorporating multiple p-n
heterojunctions in an optical interference filter has been
presented. Potential applications for this device include an
electrically tuneable optical filter for optoelectronic


182
theoretically possible to get a higher reflectivity for the
same number of layers using GalnAsP/InP material system.
4.2 MBE Grown AlGaAs/GaAs Devices
4.2.1 Introduction
A novel optoelectronic structure is presented, based on
an optical interference filter with a stack of quarter-wave p-
n heterojunctions in the GaAs/AlGaAs III-V semiconductor
system. Combining these passive filters with multiple p-n
junctions to introduce electrical functions, to the best of
our knowledge, has never been reported before. Such devices
can be used as a passive optical filter as well as active
optoelectronic switch elements.
Since A = 1300nm is a wavelength of interest in fibre-
optic communications, it was chosen as the central peak
reflectivity wavelength. For quarter-wave films, the number
of periods of high and low RI layers required to attain a
desired reflectivity can be calculated using equation (35).
Useful estimates of the room temperature refractive indices
(RI) of AlGaAs at photon energies below the direct band
edge can be calculated from a semi-empirical formula of
Afromowitz [146]. The pertinent values used at a wavelength of
1300nm are: n(GaAs, Eg=1.42eV) = 3.408 and n(Al 3Ga 7As, Eg
= 1.827eV) =3.251. A typical structure of 20 periods of
97.5nm GaAs/ 102.5nm Al0 3Ga0 yAs (see Figure 60) with an
expected peak reflectivity of 84% was chosen (see equations
(31) (35)).


199
As shown the current remains fairly low up to 12 volts, but
then begins to increase. At a current of 27mA (Vth= 12 volts)
the device turns "ON", negative resistance results when the
neighboring junctions interact and the voltage drops to 3
volts. If the current is held above ihold, equal to 30mA for
this device, the device will stay "ON", as shown. This
represents a relatively high power dissipation, which can
degrade the device performance by thermal drift and non-
uniform current injection.
A similar device was fabricated without the added zinc
diffusion step, giving a NPN (odd number of junctions) type
structure. As predicted by Katz[149] this structure does not
show bistability. A typical I-V and R-V of this device is
shown in Figure 68. The current is low, substantially less
than in the PNPN case. For this reason the NPN structure does
not appear to be suitable for RI modulation by the carrier
injection method.
Photoresponse measurements were performed on the PNPN
device by focusing the monochromator exit light onto some
electrically probed samples. The photoresponse measurement
was done at normal incidence (9=0) and the spectra were not
corrected with respect to the monochromator output. The
response was measured directly at the lock-in amplifier (see
Figure 63). The spectral response of the PNPN device used as
a detector has a large peak at 850nm due to absorption
occurring primarily in GaAs, as shown in Figure 69.


131
temperature of 625C. A deflector and baffle were used in the
reactor to improve both the gas mixing and the gas flow
pattern, respectively. A TMIn mole fraction of 0.7 x 104 and
a V/III ratio of 140 were used for the growths. Both S-doped
(n8xl018cm*3) and Fe-doped (semi-insulating) (100) oriented
InP substrates were used. Prior to loading into the reactor,
the substrates were mechanically polished using a bromine-
methanol solution, rinsed in methanol, and blown dry with
filtered high purity nitrogen. Clean quartzware (the liner
tube and gas deflector) was used for each individual MOCVD
growth and the graphite susceptor was baked-out before each
series of experiments to aid in experimental accuracy and
reproducibi1ity.
3.2.3 Results and Discussion
3.2.3a. Mo Incorporation and Laver Morphology
The atomic Mg concentration profile in the InP:Mg layers
was determined by SIMS using 157CsMg+ molecular ions with a
24Mg implant as a standard. A representative SIMS profile of
growth B232 is shown in Figure 38. During this growth, the
flow rate of hydrogen through the MCpgMg bubbler was 12.5 seem
and this resulted in a [Mg] = 3xl018cm3, atomic magnesium
concentration. All of the magnesium doped layers had a growth
rate of approximately 1.0/Lim/hour; independent of magnesium
concentration. Figure 39 shows the effect of the H2 flow rate
through the MCp^g bubbler on the atomic Mg concentration in
the layers determined by SIMS. An interesting feature of this


72
temperature of the sample. The PL spectra of two different
InP samples grown at different V/III ratios and measured at
4.2 K are shown in Figures 13 and 14. The peaks labelled
(D,x) and (FE) are host crystalline donor to bound exciton
and free exciton transitions. The other peaks maybe due to
impurities such as carbon or crystalline defects like
phosphorus vacancies (Vp) Based on the reduction in area
under the peaks labelled (Vp) it is evident that increasing
the V/III ratio or phosphine flow rate definitely has a
positive effect of the material quality of undoped InP.
In addition to the C-V profiler, the Hall effect can also
be used to determine electrical properties of semiconductors.
A low temperature (room temperature to 7 K) Hall effect system
was set up at the University of Florida by this researcher.
The theory, and a manual explaining the use of the Hall effect
system is given in the Appendix A of this text. Using data
from the Hall effect system, an average value of the mobility,
resistivity and carrier concentration can be calculated if the
thickness of the measured film is known. These values are
usually measured at 300 and 77 K (liquid nitrogen temperature)
and sometimes plotted as a function of reciprocal temperature
to 4.2 K. The liquid nitrogen temperature mobility (ii-m) is
widely used as a measure of purity of undoped semiconducting
films. In Figure 15, is plotted as a function of growth
temperatures for undoped InP. The highest mobility for InP


Zn or hole concentration (cm
155
CO
o
o
o
o


SIMS, low Zn
CV low Zn
SIMS,high Zn
o CV ,high Zn
50 100 150 200
V/lll ratio in gas flow
250
Figure 52: The effect of V/III ratio on Zn incorporation and
hole concentration.


60
Figure 6; Nomarski photographs showing the effect
ratio on InP surface morphology (top:
bottom: V/III=141).
of V/III
V/III=50,


53
doped) InP were used together or by themselves depending on
the purpose of the experiment. Next, the proper substrate
cleaning procedure must be decided. Also, experimental growth
conditions such as: growth temperature, growth pressure,
total hydrogen flow rate, trimechylindium mole fraction,
phosphine mole fraction and finally the length of time of the
planned deposition, must be determined. From the literature,
estimates of optimum growth conditions and proper wafer
cleaning procedures can be obtained, but these values or
procedures are system specific and hence had to be
experimentally optimized for the Japan Oxygen MOCVD system at
the University of Florida.
The procedure for calculating growth conditions for
undoped InP on InP by MOCVD is fairly straight forward. The
growth pressure, temperature and time must be chosen based on
knowledge of the material and capabilities of the system. It
should be noted that for convenience, some simplifications
were made in the derivation of the equations used to calculate
the growth conditions. Specifically, the ideal gas law was
used and it is assumed that the total flow rate to the reactor
is equal to the hydrogen flow rate and the bubbler pressure
was much greater than any metal organic vapor pressure. The
TMIn and PH3 flow rates are commonly expressed as TMIn mole
fraction (MFTHIn) and PH3 mole fraction (MFPH3) or their ratio,
the V/III ratio, which in this case is defined as:


250
where:
Y(l) = [p]RT
(43)
Y (2)
'[P]*T
-I3!T¡
(44)
Y(3)
*5[P]rt
*c5B2
(45)
Y (4)
5CP]rt
dBj
(46)
Y (5)
[PRT
5B4
(47)
The resulting values of the best fit are: 6^375.3 mole
fractionVatm, B2=2.5 mole fraction'1 atm'1/4, and Bj^l.23 10'18
mole fraction.
The non-linear regression program was executed a second
time to solve for the constants B5 B7 by fitting the SIMS
concentration model equation (28) to the SIMS concentration
data. The model and variational equations used for the second
program execution are:
Y(1)=B(1)*B(2) 2.36E-06*X ((1+(B(3)/l.0E+31 /
B(2) / 0.00236))A0.5) / ((1.0E+31 + B(1)*B(2) *
2.36E06*X)A0.5) (48)


202
The spectral photoresponse of the NPN device used as a
detector shows two peaks at 650 and 850nm (due to Al0 3Ga0 7As
and GaAs absorption, respectively) of opposite polarity (see
Figure 70) The relative intensity of the peaks varied
according to the position probed on the wafer. The absolute
intensity was also dependent on the device selected and on the
alignment of the optical system, but it was much lower than in
the PNPN device. It resembled the appearances of two opposing
detectors almost canceling each other's response. Sinc no
response was observed at longer wavelengths (1000-6000nm) as
expected and shown in Figure 69, the PNPN device has great
potential as a tuneable electro-optic device with either
electrical or optical control.
Using the RI results of Sell et al.[153] and Cross and
Adams[154], at typical electrical testing conditions (20 volts
through a 50 micron square device) a change in carrier
concentration from 1016 to I018cm'3 would result in a 0.8%
decrease in the refractive index of the narrow band GaAs.
This would result in a lOnm peak shift and a 30% drop in
intensity for reflected light at 1300nm. As indicated above,
the presently achieved high switching currents precluded
optical measurement of the expected spectral shift due to non-
uniform current distribution in the sample and heating
effects. One approach to the high switching current problem
is to redesign the PNPN junction stack to lower ihold.
Alternatively, all the p layers and n layers can be connected


171
Using the equilibrium constant relationships given in Table 8,
with yn = Yp = i# equations 23-25 can be solved analytically.
The solutions are:
[B4 + B,B2 pZn (Pp4) 1/4]1/2
[1 + V (B2B4(Pp4)1/4)]1/2
(26)
Prt
BiB2pZn(Pp4)V4
P
(27)
and
[Zn] =pZn
BiB2(Pp4)
1/4
Pgt
hBs
f W*
B.
(1 + -
B,
(Pp4) V4
(28)
where B,,- B7 are constants defined as follows: B, = K^K^, B2
= Khl/Kh2' B3= *^3' B4 = *^4' 5 = Kd8' B6 ~ Kd10 an{* B7 =
Kdn/K,,,. It is significant to note that when equation 26 is
substituted into equation 27, the expected (often reported in
the literature) p versus p2n square root dependency exists.
For InP, the electron-hole pair equilibrium constant,
Kh4(B4) can be approximated using the following equation[ 137]:
Kh4 = 4(2tt(memh) 1/2kT/h 2]3 exp(-||)
(29)


51
The products of these reactions are less hazardous and for
added safety, when the system requires draining, the solution
flows to the building's neutralization pit.
2.4.5 Safety
Arsine, phosphine and hydrogen sulfide are highly toxic
combustible gases, therefore, several safety features are
installed in the MOCVD system, and several safety practices
must be followed in the facility. First of all, the facility
has a rule that at least two competent people must be present
in order to do any work in the building. The facility also
has an alarm system and card reader to deny access to
unauthorized personnel. The clean room has an eight point
toxic gas monitor (MDA, Inc.) which has a resolution of 1 ppb
for all hydrides. In the unlikely event of a toxic gas
detection anywhere in the facility, the MOCVD system will
completely shut down (all air-operated valves are normally
closed). The MOCVD system also has a four point hydrogen gas
detector (Matheson) connected to it so that if hydrogen levels
exceed 50 ppm, the machine will shut down. Also the facility
has a helium leak detector (Varian) which is used to find
actual leak points before the MOCVD system is used and to
check connections after valve replacements or reactor changes.
Not only is the facility well equipped with safety
features and practices, but the Japan Oxygen MOCVD System
itself has several integrated safety systems. There are two
types of alarms, facility failure and machine failure. If


23
(<100C) and therefore is less likely to react upstream of the
heated growth zone. Also, TMIn has a much higher vapor
pressure than TEIn which is experimentally convenient because
heated gas lines would no longer be necessary.
Other techniques such as the use of hydrogen-nitrogen
mixtures as the carrier gas[41] and phosphine pre-crackers [44]
have been tried with varying degrees of success and merit. A
final conclusion is that proper reactor geometry, system
design, and growth conditions are very important for avoiding
parasitic gas phase reactions and obtaining superior InP thin
film quality. Currently, the reaction at certain growth
conditions of TMIn and PH3 in a properly designed MOCVD system
can yield uniform, high quality epitaxial InP with no evidence
of indium prereaction problems.
The deposition of high quality layers of InP for device
applications requires precise control of their unintentionally
introduced (undoped) and intentionally introduced (both p- and
n-type doped) impurity concentrations. An important condition
for obtaining reproducible p- and n-type doping levels is the
ability to grow undoped material with a reproducibly low
background carrier concentration. To obtain low background
levels, one needs a contamination free MOCVD system equipped
with high purity sources and optimized growth conditions.
Most conventional MOCVD systems are constructed from
ultra-high purity components such as electropolished welded
316 stainless steel and semiconductor-grade low-sodium content


47
2.4.3 Reactor and Heating System
The Japan Oxygen MOCVD system has a horizontal 4 inch
I.D. reactor which is equipped with a water cooling jacket
(see Figure 4) to minimize side wall deposition. At the inlet
of the high purity quartz reactor, two inlet lines exist to
keep the group III metal organics separated from the group V
hydrides at low temperatures. The dilute group III sources
enter the reactor through a 6 mm O.D. high purity quartz tube
10cm downstream from the inlet of the group V sources. These
two gas streams should ideally be well mixed and flowing under
fully developed laminar flow conditions before the mixture
encounters the high purity quartz deflector/silicon carbide
coated graphite susceptor heat source. The deflector -
susceptor unit is tapered at an angle of 17 with respect to
horizontal to improve growth rate uniformity.
The graphite susceptor is heated by radio frequency
inductive heating. A Lepel series T-15-3-KC-TL 15 kilowatt RF
generator is used to generate radio waves over the frequency
range 80-900 KHz. Some of these radio waves are picked up by
a 3/8" copper coil that is wrapped around the reactor. A
platinum/rhodium "R-type" thermocouple sealed in a high purity
quartz tube is embedded inside the graphite wedge. This
thermocouple is connected to a West series 2070 microprocessor
based temperature controller which was installed on the MOCVD
system and sends DC current to the RF generator to control


168
Table 8
InP Point Defect Constants and Electroneutrality Relation
Reactions
Equilibrium Relations
\PA{g) + Vi = Pi
Pp + Vi = Pi + VP
Vp = V<+ e~
0 = e~ + k+
In in + Vi = Irii + VIn
0 = Vin + VP
Vln = Vfn + h+
Zn(g) + Vi = Zrii
Zn(g) + Vin = ZnJn + h+
Znin + Vp = ZnjnVp
Pp + ZninVp = VpZninVp + Pi
sb
ii
(11)
pp<
= [P¡] [VP]
(12)
jj- T'nnJVjJ]
Khz [Vp]
(13)
I

(14)
Km = [In¡\ [V,\
(15)
I (16)
Kh7
A/l7 ~ [Vm]
(17)
(18)
(19)
= 1*
(20)
rs \VpZninVp]\Pi\ icyi \
KdU ~ [ZnInVP} (21)
Tl +
in
+
Z"Fn] = P + [Vp+]
(22)


259
[117] Skromme, B. J., Stillman, G. E., Oberstar, J. D. and
Chan, S. S., Appl. Phys. Lett. (1984) 319.
[118] Williams, R. S., Barnes, P. A. and Feldman, L. C., Appl.
Phys. Lett. 36 (1980) 760.
[119] Lennard, W. N., Swanson, M. L., Eger, D., SpringThorpe,
A. J. and Shepherd, F. R., J. Electron. Mater. 17
(1988) 1.
[120] Yamada, M., Tien, P. K., Martin, R. J., Nahory, R. E.
and Ballman, A. A., Appl. Phys. Lett. 43. (1983) 594.
[121] Wong, C. C. D. and Bube, R. H., J. Appl. Phys. 55
(1984) 3804.
[122] van Urp, G. J., van Dongen, T., Fontijn, G. M., Jacobs,
J. M. and Tjaden, D. L. A., J. Appl. Phys. 65 (1989)
553.
[123] Marek, H. S. and Serreze, H. B., Appl. Phys. Lett. 51
(1987) 2031.
[124] Kazmierski, C., J. Appl. Phys. 64 (1988) 6573.
[125] Dlubek, G. and Brummer, O., Appl. Phys. Lett. 46 (1985)
1136.
[126] Li, S. S., Wang, W. L. and Shaban, E. H., Solid State
Comm. 51 (1984) 15.
[127] Kami j oh, T., Takano, H. and Sakuta, M., J. Cryst. Growth
67 (1984) 144.
[128] Cole, S., Evans, J. S., Harlow, M. J., Nelson, A. W. and
Wong, S., Electron. Lett. 24 (1988) 813.
[129] Glade, M., Grutzmacher, D., Meyer, R., Woelk, E. G. and
Balk, P., Appl. Phys. Lett. 54 (1989) 2411.
[130] Omeljanovsky, E. M., Pakhomov, A. V. and Polyakov, Y.,
Semicond. Sci. Technol. 4 (1989) 947.
[131] Jackson, D. A., J. Cryst. Growth. 87 (1988) 205.
[132] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 613.
[133] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 627.
[134] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 639.
[135] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 647.


Response [V]
201
Figure 69: Photoresponse spectrum for a PNPN device.


146
the V/III ratio and growth temperature on material properties
has also been studied. The ultimate goal of this work was to
find the overall "ideal p-type dopant for InP based devices
grown in the BNR low pressure MOCVD system.
3.3.2 DMCd Results
The DMCd doping experiments were carried out in the CVT
Ltd. MOCVD system which was discussed in section 3.2.2. The
InPtCd growth conditions were: Pg = 76 torr; Tg = 625C; FH2
= 7 SLM; V/III = 140; DMCd total bubbler pressure = 600 torr;
and, DMCd bubbler temperature = 15.9C (vapor pressure = 3.898
torr). The hydrogen flow rate through the DMCd bubbler was
varied from 90 to 700 seem which resulted in a DMCd molar flow
rate range of 2.94xl0'5 to 2.284xl0'4 moles/min. At the higher
DMCd flow rates, the material quality rapidly degrades due to
the formation of large hillocks possibly from the formation of
Cd-P precipitates. The hole concentration of these grown
layers is plotted as a function of DMCd molar flow rate in
Figure 47. Also plotted in Figure 47 is Cd doping data from
the same reactor of a previous study by Blaauw et al.[59]. A
linear relationship between DMCd flow rate and hole carrier
concentration for both data sets is evident. The main
conclusion is that at high DMCd molar flow rates grown at a
low pressure of 76 torr, InP:Cd material quality is poor.
This is unfortunate because DMCd has a wide incorporation
range and Cd is a relatively slow diffuser in InP[58].


Mg Concentration (cm
137
CO
i
10
20
epilayer/substrate
interface
7
2 3 4 5
Depth (m)
Figure 41: SIMS Mg profiles of InP on Fe-InP (symbols
represent H2 flow rates of 5, 12.5, 22 and 27.
seem to the Mg bubbler, respectively).


165
of the Zn dopant is incorporated into the growing layer and
the rest evaporates and is transported away from the growing
crystal. If the growth temperature is raised, the dopant
evaporation rate increases and consequently, the fraction
incorporated decreases. Also, the doping concentration would
be independent of the growth rate. The behavior described in
case (2) is typical of DEZn doping of InP as observed for
MOCVD growth at BNR, the University of Florida and also
reported in the literature (see sections 2.5.2c, 3.1.5c, and
3.3 of this text).
Several different point defects and defect complexes have
been proposed to explain the apparent electrical inactivity of
incorporated zinc in InP:Zn. These point defects and other
host crystal point defects such as a phosphorus vacancy, Vp,
and an indium vacancy, VIn, are represented in a fictitious
two-dimensional InP lattice which is shown in Figure 56. The
relative concentrations of these defects in a real crystal
depends upon the conditions for which it was grown and the
environment it is presently in. Chemical reactions can be
written for each point defect, and traditional methods of
chemical engineering thermodynamics can also be used to derive
expressions for the concentrations of neutral and charged
point defects and complexes. The assumption will be made that
the concentration of these point defects is small relative to
the host crystal atomic concentration. This will simplify the
equations required. Temperature-dependent equilibrium


108
2.5.4 Growth of GalnAsP Lattice-Matched to InP
Unlike Ga^n^jjAs, a continuous range of compositions
exist for the mixed crystal GaxIn1.xAsyP1.y which are lattice-
matched to InP. The Asy composition can be varied from y=0
(InP) to y=1.0 (Ga 47In 53As), but since the relation y=2.16x
exists for lattice-matched quaternary, Gax is constrained to
only vary from x=0 (InP) to x=.47 (Ga 47In 53As) This wide
range of variable composition corresponds to material which
can emit light at wavelengths from 0.92/im (y=0) to 1.65jum
(y=l), (see Figure 1) and several layers of lattice-matched
quaternary over this wavelength range have been successfully
grown in the Japan Oxygen MOCVD system.
The MOCVD growth conditions for Ga^n^As P,, on InP
were based upon the optimum growth conditions for Ga 47In 53As
on InP. It has been experimentally determined that the
following gas phase growth conditions and the relationship
between them all have a strong influence on solid phase
material quality: MFn,In/MFTEGa, FPH3/FASH3, and the V/III
ratio. The exact values of these three ratios are strongly
composition dependent and initial estimates of these values
for a specific solid phase composition were taken from the
literature[72]. In Figure 33, the effect of growth conditions
on material quality (lattice-matching) is clearly shown. The
upper photo is of a sample from growth Q036 and it has the
characteristics of a lattice-mismatched sample, point and line


166
I I
In
I
P
In
In
p Cfcp-
In
In
Zn
In
In Q
In Q- in Q ,n
P
In
I
P
I
Zn
I ln
I I
Figure 56: Schematic representation of point defects in zinc
doped InP.


46
hydrogen and nitrogen lines prevent back flow of toxic gas
through the manifold to the rest of the system.
The "heart" of the gas delivery system is the fast
switching "vent/run" manifold mounted just prior to the inlet
of the reactor. The status of the "vent/run" valves (opening
and closing) and the timing of this sequence can be manually
or automatically controlled. Automatic control is made
possible by using a process sequencer which is capable of
storing 150 valve patterns and the corresponding times for
each pattern or layer. The MOCVD manifold is also "pressure
balanced" which means that it is possible to adjust the
pressure difference between the two vent lines and the two
reactor lines to almost zero. This is accomplished by
carefully adjusting tube lengths and by installing a dead
volume in the vent line to counter balance the volume of the
reactor. Pressure balancing is especially important for MOCVD
growth of superlattice or multiple quantum well structures.
To keep the pressure difference between the reactor and vent
lines equal to zero (which is measured using the two
differential pressure indicators) each main source line has a
hydrogen compensation line. It is necessary for the MOCVD
growth of InP and related materials to keep the metal organics
separated from the hydrides to prevent parasitic gas phase
prereactions from occurring, hence the reason for the two
vent/two reactor line design.


229
constant signal values). The voltmeter used was actually a
high-quality high-input impedance electrometer (Keithley
Instruments, model 614) to avoid any leakage currents and
consequently increased error.
Now that all electrical connections are made, the
temperature desired for measurement must be reached. This can
be room temperature (300 K) or lower, but the system can not
be operated above room temperature as the cryostage contains
indium seals which would melt. For temperatures below
ambient, the temperature controller and compressor must be
used. Basically the rough pump is first plugged in and run
until the pressure gauge reads 1-2 torr. At this point close
the brass valve, unplug the rough pump and slowly vent the
pump and hose to 760 torr. Now turn on first the compressor,
then the cold head (on the compressor) and then, last, the
temperature controller. The controller must be programmed
with its optimum process parameters as found and explained in
its manual. It was advised that the heater plugs always be
disconnected whenever the temperature readout exceeded 275 K.
When the combination of helium cooling and resistive heating
is operating correctly in a properly evacuated environment,
temperatures below 100 K should be attained and stable in
about 45 minutes. If it is not thermally stable or the
outside of the cryoshield is cold with condensation occurring
then either the chamber was not properly evacuated or there is
a leak in the cryostage.


log N(cm~3)
139
Figure 42: C-V profile of growth B229 (InP:Mg on InP:S).


50
2.4.4 Exhaust/Scrubbina System
The waste products from the reactor flow through an
exhaust port in the water cooled flange which is bolted to the
exit of the reactor to a particle filter (Fuji Ltd.) which is
made out of glass fibers. The gas can then either flow at
atmospheric pressure directly to the scrubber or it can be
evacuated from the system by a rotary pump (Edwards model E2M
18) and then continue onto the scrubber. The rotary pump
makes it possible to grow films at low pressures (0.05 to 0.2
atm) which generally improves thickness uniformity.
The scrubber which was added to the Japan Oxygen MOCVD
system was built by Advanced Concepts (model 9625). It is a
liquid based scrubber which is designed to scrub toxic gases
and exhaust clean gas to atmosphere. For safety, the exhaust
was connected by a fireproof duct to the building's room air
scrubber. The scrubbing solution consists of a 80:5:2 by
volume mixture of water, 15% sodium hypochlorite, and 50%
sodium hydroxide. The pH and oxidation reduction potential
(ORP) of the scrubbing solution are constantly monitored to
evaluate the solution's scrubbing potential (pH >10.0 and ORP
> 200 mV). The scrubber has an efficient gas-liquid venturi
contactor and a packed bed which can handle higher than
required toxic gas flow rates. For arsine and phosphine the
overall chemical neutralization reactions are:
AsH3(g)+3Na0Cl(l)+H20(l) = H3As04(l)+3NaCl(s)+H2(g) (2)
PH3 (g) +3NaOCl (1) +H20 (1) = H3P04(l)+3NaCl(s)+H2(g) (3)


11
of the growth film may be reduced to less than ten percent.
Numerous devices such as LEDs, lasers, FETs and HBTs have been
fabricated with MBE. One unique bonus with MBE is that in-
situ monitoring devices such as RHEED, mass spectrometers,
Auger spectrometers and ion gauges are feasible and
commonplace. However, large expense and limited throughput
restrict MBE usage.
During the last decade, several novel deposition
techniques for III-V compound semiconductors have emerged.
Each one is a spin-off of either MOCVD, MBE, or a combination
of the two and has some relative advantages and disadvantages;
none of these new techniques are widely used. One such
technique is called atomic layer epitaxy or ALE. For III-V
compound semiconductors, ALE proceeds by depositing a
monolayer of a group III metal followed by depositing a
monolayer of group V atoms, in a layer-by-layer sequence[8].
In ALE, grown layer thickness is determined by the number of
cycles rather than the time of growth. Another relatively new
technique is called chemical beam epitaxy (CBE) or metal
organic molecular beam epitaxy (MOMBE). In MOMBE, all of the
group IIIA and VA sources are metal organic; TMAs replaces As4
and AsH3, TEP replaces P2 and PH3[9]. The remaining aspects
of MOMBE are essentially the same as conventional MBE. A
final technique which is similar to ALE but is performed in a
conventional low pressure MOCVD system is called flow
modulation epitaxy or FME. FME has been used to grow InP at


81
vapor pressure of DEZn in the bubbler from 1.39 torr to 14.00
torr. The hydrogen flow rate through the bubbler ranged from
5 to 50 seem and by carefully adjusting the opening of the
DEZn needle valve it was possible to keep the bubbler at a
pressure of 500 torr.
Characterization of the p-type InP material included
thickness and surface morphology measurements using the
optical microscope, C-V profiles, Hall measurements and
secondary in mass spectroscopy (SIMS). The SIMS technique,
which is explained elsewhere[91], was used to determine the
total atomic zinc concentration incorporated into several p-
type InP layers as a function of depth. This total zinc level
can be compared to the carrier concentration to determine the
percentage of electrically active zinc atoms. Also, based on
the depth of the atomic zinc profile, the extent of zinc
diffusion can be assessed relative to the epitaxial layer
thickness measurements. A C-V profile of growth Q080 which
was deposited at Tg=600C, Pg=80 torr and V/III=50.0 is shown
in Figure 18. During the growth, a DEZn mole fraction of
4.06xl0'5 was used which resulted in a C-V measured hole
concentration of 1.6xl018cm'3, for the layer grown on the n+-
InP substrate. A room temperature Hall effect measured hole
concentration of 2.5xl018cm*3 was calculated for the layer
grown on the semi-insulating substrate. The room temperature
Hall mobility and resistivity for this p-type sample were
61.3cm2/volt-sec and 0.0402 ohm-cm, respectively.


162
The degree of electrical activation in zinc doped InP
does vary with the amount of atomically incorporated zinc and
this phenomenon has been investigated by several research
teams. Williams et al.[118] have used proton-induced X-ray
emission (PIXE) combined with channeling techniques to analyze
Zn-doped InP. They concluded that neutral complexes such as
VpZnInVp, (which was first proposed by Hooper and Tuck[104]
based on external diffusion experiments), do not exist but
nonsubstitutional Zn is in the form of randomly distributed
precipitates. More recently, Lennard et al.[119] performed
PIXE experiments on Zn-doped InP and found no evidence for
precipitates, but state that VpZnInVp complexes could explain
electrical inactivation. Yamada et al.[120] theorized that
the neutral complexes VpZnInVp and ZnInVp exist in InP:Zn and
that their presence explains the double diffusion fronts
observed by electron beam induced current (EBIC) and SIMS
analysis of their externally zinc diffused samples.
Interstitial zinc, Znif both positively charged as a
compensating donor, Zn,+ [121,122] and as an intermediate
reacting with substrate donor atoms to form a neutral
complex[123], have also been proposed to explain Zn diffusion
profiles in InP. Another report states that interstitial zinc
must be electrically neutral[124] based on the shape of
simulated diffusion profiles taking the charge of m, of Zn,1",
to be 0, +1, +2. Kazmierski[124] also suggests that other


249
model equation with respect to each unknown model parameter.
The program was first run to solve for the constants Bv B2,
Bj by fitting the hole concentration model equation (27) to
the hole concentration data. The model and variational
equations used for this are:
Y(1)=B(1)*(2.36E-06)*X*(((B(2)A2)+B(3)*B(2) /
B(4)/0.00236)A0.5) / ((B(4) + B(1)*B(2)*
(2.36E-06) *X)A0.5) (38)
Y(2)=(2.36E06)*X (((B(2)A2)+B(3)*B(2)/B(4) /
0.00236)A0.5) (B(4)+0.5 B(l)*
B(2) (2.36E-06)*X) / ((B(4)+B(1)*B(2) *
(2.36E-06)*X)Al.5) (39)
Y(3)=B(1)*(2.36E-06) X*(B(2)*B(4) + 0.5*B(1)*
(B(2)A2) (2.36E-06)*X + 0.5*B(3)/0.00236) /
((((B(2)A2)+B(3)*B(2) / B(4)/0.00236)A0.5) *
((B(4)+B(1)*B(2) (2.36E-06) *X)A(1.5))) (40)
Y(4)=(0.5*B(1)*B(2) (2.36E-06) X/B(4)/0.00236) /
(((B(4)+B(l) B(2)*(2.36E-06)*X)A0.5) *
(((B(2)A2) + B(3)*B(2) / B(4)/0.00236)A0.5)) (41)
Y(5)=(-B(1)*B(2) (2.36E-06)*X) ((B(3)/B(4) /
0.00236) + (0.5*B(1)*B(2) B(3)*(2.36E-06)
*X/0.00236/(B(4)A2)) + 0.5*B(2)) / (((B(4)+B(l)
* B(2)*(2.36E-06) X)A(1.5)) (((B(2)A2)+B(2)
* B(3)/B(4) / 0.00236)A0.5)) (42)


126
date, Cd, Mg and Zn have been used as dopant species for p-
type doping of MOCVD InP. Several different organometallic
compounds have been used as Cd, Mg and Zn sources such as:
dimethylcadmium (DMCd)[58,59,112], cyclopentadienyl magnesium
(Cp^Mg)[58], bis-(methylcyclopentadienyl) magnesium (MCpgMg)
[113], diethylzinc (DEZn)[45,58,112] and dimethylzinc (DMZn)
[58]. The relative merits and disadvantages of the use of
each dopant will be discussed pertinent to specific desired
device characteristics.
The use of Cd as a p-type dopant for MOCVD InP has been
reported by several researchers[58,59,112]. Blaauw et al.[59]
report atmospheric and low pressures MOCVD growth using TMIn,
PHj and DMCd. Nelson and Westbrook[58] report atmospheric
pressure MOCVD growth using adducts. The same doping trend
was reported by Blaauw et al.[59] and Nelson and Westbrook
[58]; a linear incorporation rate of Cd from p = 5xl015cm3 to
2xl018cm3 (measured by Hall) due to a change in DMCd partial
pressure from lxlO'6 to 3xl04 atm. Blaauw et al.[59] reported
this trend for both atmosphere pressure and low pressure (76
torr) growths, but observed extended layer morphology
deterioration at high DMCd partial pressures for only low
pressure growths. Nelson and Westbrook[27] also investigated
the effect of growth temperature (Tg = 550-650C) on DMCd
incorporation. They observed that as the growth temperature
is decreased, the hole concentration at a fixed DMCd flow
rate, increased. They attribute this trend to increased Cd or


RELATNE INTENSITY
73
Figure 13: PL spectrum of an InP sample grown with a V/III
ratio of 140 measured at 4.2 K.


21
Several papers have been published on the important topic
of MOCVD safety[31,32]. These papers are very useful when
designing the layout of an MOCVD machine and laboratory.
First, all MOCVD systems must have both toxic gas and hydrogen
sensors in and around them connected to an alarm. These
sensors must be capable of shutting down the machine in the
event of a detected leak. In order to shut down quickly, all
gas lines must be equipped with normally closed air-operated
valves. Toxic gas lines should also be double contained
(which is extremely expensive) and equipped with pressure
sensors that sound an alarm for abnormally high pressures.
Compressed air, hydrogen and nitrogen lines should have
sensors for abnormally low pressure which can, if activated,
shut the machine down. The reactor and pump exhaust line
should also have similar pressure sensors. Of course, smoke,
fire, and cooling water flow detectors and sensors are
necessary. Other things such as micro-switches on panel doors
and available supplied air masks are also required. Finally,
SCBAs and trained users should always be available outside of
the facility for emergencies. All of the above mentioned
safety issues/design features are important for the design and
operation of a modern MOCVD system. Of course, the most
important thing for lab safety is to provide adequate operator
training focusing on the nature of the toxic sources and how
to treat them.


211
nquat= 3.47 and nInP= 3.17, and using equations (31), (33) and
(36) a peak reflectivity of 0.9197 occurring at X = 1550nm is
expected. If the RI of the quaternary layer is reduced by An
= -.015, the stack will no longer be exactly a quarter-wave
(optically) thick. Hence, the new reflectivity peak will
occur at X = 1543nm and have a maximum value of 0.9090. A
theoretically predicted spectrum (neglecting GalnAsP and InP
absorption) for a 30 layer stack structure of alternating
1550nm quarter-wave thick layers of Ga 375In 625As 81P 19 and
InP is shown in Figure 75. If the central peak in Figure 75
was shifted by 7nm due to a change in the RI of the quaternary
layer, a maximum change of intensity of 30% for reflected
light at 1525nm (the steepest point of the peak) is expected.
This optical change can be electrically controlled hence
potentially creating a switching element.
A p-GalnAsP/n-InP interference filter structure was grown
by MOCVD. The layer structure and growth conditions of MOCVD
sample Q240 are shown in Figure 76. This sample was sent to
BNR for optical testing as a passive interference filter. The
reflectivity versus wavelength spectrum for sample Q240 is
plotted in Figure 77. As shown, the central peak occurs at a
wavelength of A0 = 1400nm. The detected reflectivity signal
was not normalized with respect to the detector's response for
this spectrum. The central wavelength is considerably
different from the expected value of 1550nm. This indicates
that either the growth rates, quaternary composition, RI


254
[35] Manasevit, H. M., Dapkus, P. D., Ruth, R. P., Yank, J.
J., Campbell, A. G., Johnson, R. E. and Moudy, L. A.,
1978 Photovoltaic Spec. Conf., Washington, DC (New York,
NY, IEEE).
[36] Duchemin, J. P., Bonnet, M., Beuchet, G. and Koelsch,
F., Inst. Phys. Conf. Ser. 45 (1979) 10.
[37] Yoshino, J., Iwamoto, T. and Kuhimoto, H., J. Crystal
Growth 55 (1981) 74.
[38] Razeghi, M., Poisson, M. A., Larivain, J. P. and
Duchemin, J. P., J. Electron. Mater. 12. (1983) 371.
[39] Renz, H., Weidlein, J., Benz, K. W. and Pilkuhn, M. H.,
Electron. Lett. 16 (1980) 228.
[40] Moss, R. H. and Evans, J. S., J. Cryst. Growth 55
(1981) 129.
[41] Sacilotti, M., Mircea, A. and Azoulay, R., J. Cryst.
Growth 68 (1983) 111.
[42] Hsu, C. C., Cohen, R. M. and Stringfellow, G. B., J.
Cryst. Growth 63. (1983) 8.
[43] Bass, S. J., Pickering, C. and Young, M. L., J. Cryst.
Growth M (1983) 68.
[44] Duchemin, J. P., J. Vac. Sci. Technol. 18. (1981) 753.
[45] Razeghi, M. and Duchemin, J. P., J. Cryst. Growth 64
(1983) 76.
[46] Eguchi, K., Ohba, Y., Kushibe, M., Funamizu, M. and
Nakanisi, T., J. Cryst. Growth 93 (1988) 88.
[47] Kasemset, D., Hess, K. L., Mohammed, K. and Merz, J. L.,
J. Electron. Mater. 13 (1984) 655.
[48] Scott, M. D., Norman, A. G. and Bradley, R. R., J.
Cryst. Growth 68 (1984) 319.
[49] Knight, D. G., Miner, C. J. and SpringThorpe, A. J.,
1990 Materials Res. Soc. Spring Meeting, San Francisco,
CA, Paper F4.5.
[50] Tuck, B. and Baker, A. J., Jour. Mater. Sci. 8 (1973)
1559.
[51] Nishitani, Y. and Kotani, T., J. Electrochem. Soc. 126
(1979) 2269.


228
should be taken to avoid unnecessarily touching its internal
parts and of course disposable gloves must be worn.
Once the connections have been made to the resistivity
holder, the inner heat shield should be gently slipped over
the cold finger and attached to the cryostage. Next the outer
shield should be placed over the heat shield while making sure
that its "O" ring is in place. Now the rough pump should be
attached to the closed brass valve. If the helium supply and
return lines, the cryostat power line, and the ten pin mating
plug are attached then the cryostage is ready for operation.
The previously discussed four sample lines should be
connected to the binding posts labelled 1, 2, 3 and 4 of the
aluminum shielded switching box. The switching box contains
three double-pole double-throw switches, a four-pole six-throw
switch, a standard 100 ohm resistor and four other binding
posts for current and voltages leads. This box was built by
Mrs. Grazyna Palczewska, another member of our research group,
as per the design in reference[157]. It efficiently permits
the acquisition of Hall voltage data by a simple routine of
switching current flow through the sample and the standard
resistor. The standard resistor used should be of the same
magnitude as the resistance of the sample[157]. The constant
current supply (Lambda, model LQ531) must be attached to the
+1 and -I labelled binding posts. The magnitude of the
current used was always less than 5 mA which was low enough to
avoid resistive heating in the sample (as indicated by


246
pressure drop (note room temperature). Check for leaks with
the Matheson detector by pressurizing to 200 psi with H2 and
open/close all MV's near the cylinder. Allow H2 to flow at
normal conditions.
Wear air masks and evacuate others from clean room during
the first use of new cylinder (before a growth), flow gas to
vent for one hour to remove volatile impurities (of course
turn scrubber on) The cylinder is now ready for routine use,
keep cylinder cabinet doors closed at all times.
B.6 Return from an Emergency Shutdown
After problem has been solved by the team wearing air
packs, if necessary, hit "alarm reset/"BZ reset" to stop
buzzer and alarm light. Evacuate and fill the system with H2
if necessary, five times. Do a complete "preparation" check
(see section B.4). Follow the turn-on procedure at section
B.5.2.
B.7 SHUTDOWN
Return to the "OFF" state is defined in the previous
section. For a complete shutdown, the system should be
completely purged with N2 for 1 hour (all lines up to a closed
source manual valves closed, that is) Now, all three
temperature baths should be shut off. Close the UHP H2 outlet
valve and evacuate the whole system to 1 torr. Then close
MV100, AVI10 and hit the red emergency stop button. Turn off
the heat tapes and skinner valve. Set the Hz purifier
temperature to 0C and when T < 200C, shut off the purifier,


205
4.2.6 Addendum
Another structure, similar to the one of MBE 464, was
grown with the goal of reducing the number of electrical p-n
junctions in the interference stack. The actual layer
structure requested for growth MBE 572 is shown in Figure 71.
With this doping profile, the number of grown p-n junctions is
ten instead of twenty-one, as was the case for the as grown
MBE 464 sample. With less p-n junctions, the holding current
and threshold voltage for this thyristor could be much less
which could improve its switching characteristics.
A TEM photograph of a cleaved (110) and etched edge of
sample MBE 572 is shown in Figure 72. Similar to MBE 464, the
vertical period uniformity and interface abruptness appears to
be quite good. However, at 50,000 times magnification, the
AlGaAs/GaAs layer pair thickness was measured to be 172nm
which is considerably lower than the design value of 200nm.
This was later shown to not significantly affect the passive
optical device performance of this interference filter.
Reflectivity values were also measured on MBE 572 over the
wavelength range of 850-1600nm. The normalized reflectivity
data is plotted in Figure 73. Interestingly enough however,
even though the layer thicknesses according to TEM are off for
this structure, the central peak reflectivity is higher
(0.825) than it was for sample MBE 464. Perhaps the optical
alignment was better for this sample. The location of the
central peak maxima, 1250nm, is also extremely good.


245
off RP2. Use the Matheson H2 detector around the bubbler with
the bubbler at 900 torr and hydrogen flowing. Turn on the
heat tapes. Flow MO to the vent for one hour before use in a
growth to remove volatile impurities.
Bi5.6 Hydride Source Change (PH3 as an example)
Close the gas cylinders tightly (using the black gloves) .
Purge the gas lines with H2 (five times evacuation and fill
up) then let H2 flow overnight. Close AV21, AV22, AV23, AV20
and pressurize the gas lines with cylinder N2 gas and then
release slowly through MV41 and RV40 to the exhaust (turn on
RP1 and RP2 and close exhaust AVs) Repeat this step four
more times then leave the line at 760 torr and close MV23.
Prepare the new cylinder (and gasket which is necessary on H2S
cylinder only) and clean room gloves and wrenches. Put on the
gas mask in a buddy system, set the MDA to sense at QMOCVD
only, and seal off the clean room from normal access. Take
out the old gas cylinder. Put in the new gas cylinder (use
the gasket).
Pressurize the gas lines with N2 gas up to AV20. Check
for leaks with soapy water. Release the N2 pressure through
MV41 and RV40 (prevent air from entering the rest of the
system) If there are no leaks then remove air masks, put MDA
on auto and allow others to enter the clean room. Pressurize
the gas line and then release ten times with N2. Evacuate the
gas lines, and then check for leaks with helium detector.
Pressurize the line overnight with N2 and check for any


7
common substrate materials used today for III-V homoepitaxy,
(i.e., growth on a crystal of the same composition), or
heteroepitaxy, (i.e., growth on a crystal of different
composition or crystal structure), are GaAs and InP. The
substrates are cut from bulk crystals along a particular
crystal orientation from two to four inch diameter boules
which are sometimes created by withdrawing a seed crystal from
a heated liquid melt. Epitaxial growth on these substrates is
accomplished by exposing the heated surface to a flux of group
IIIA metals and group VA non-metals. The flux can be supplied
from a liquid, vapor, or molecular beam source. This
distinction defines the three primary methods for growing
epitaxial III-V films: liquid phase epitaxy (LPE), vapor
phase epitaxy (VPE), (also known as chemical vapor deposition
(CVD)), and molecular beam epitaxy (MBE). Each technique has
its own advantages and drawbacks, which will be discussed in
the following paragraphs.
LPE is a growth technique which can be used to deposit
thin single crystal layers of III-V compound semiconductors
from a heated liquid solution by decreasing the temperature of
the substrate relative to the solution. It is a relatively
simple, inexpensive, near equilibrium (reproducible) growth
technique that is well understood. The growth rate can be
high and a wide range of both p- and n-type dopants are
available and their incorporation is controllable. LPE has
been used to grow InP[2] and high quality GaInAs[3] on InP for


158
bis-(methylcyclopentadienyl) magnesium (MCpgMg) as sources for
the p-type doping of MOCVD InP. A summary of the measured
hole concentration versus dopant partial pressure relationship
for each of these three dopants is presented in Figure 54.
Using the definition of an "ideal" p-type dopant given in
section 3.3.1, the information given in sections 3.3.2 and
3.3.3, and the data shown in this figure, it is clear that
DEZn is the best p-type dopant for MOCVD InP grown at a low
pressure. Consequently, the experimental data and other
information gathered from the literature review on DEZn doping
of InP will be used to formulate a model of the p-type doping
process of MOCVD InP.
Bulk crystal zinc doped InP has been grown by the LEC
method[101]. Epitaxial layers have been grown from the liquid
phase by LPE[103] and the gas phase by hydride[109], chloride
[110] and metal organic[45,58] CVD. Zinc doped crystals or
layers grown using these different techniques have been
reported to have approximately the same following dopant
incorporation and electrical activation relationship relative
to the dopant source concentration: (1) zinc has a solubility
limit in InP of 2-4xl018cm'3, and then forms ZnP; and (2) once
this solubility is attained, at higher source concentrations
measured hole concentration levels sometimes remain constant
and sometimes decrease. For LEC grown bulk crystals[ 101 ], the
decrease in 300 K hole concentration has been attributed to
precipitates without substantial evidence. Wada et al.[103]


68
Figure 10: C-V profile of growth Q056 (undoped InP on n+InP).


Current (jjA)
200
Voltage (V)
Figure 68: I-V and R-V characteristics of a NPN device.
Resistance xIO (&)


90
L = 0.5 microns, and the length of time that the sample was at
growth temperature, t = 90 minutes, a rough estimate of the
diffusion coefficient, D, of zinc in InP can be calculated
using this equation:
D(cm2/sec) = L2(cm)/t (sec) (7)
A value of D = 4.6xl0'13cm2/sec is calculated which agrees very
well with the range of values that were reported by Nelson and
Westbrook[58], D = (l-6)x I0'13cm2/sec, for zinc in InP.
A review of the literature on the topic of p-type doping
of InP by several growth techniques, extensive data on zinc,
magnesium and cadmium p-type doping of InP by MOCVD, (which
was acquired when this investigator was a visitor at BNR) and
a theoretical model of the p-type doping process of MOCVD InP
is all presented in Chapter III of this text. The reader is
therefore referred to Chapter III for a more detailed and in-
depth discussion on p-type doping of InP.
2.5.3 Growth of GalnAs Lattice-Matched to InP
The mixed crystal Ga 47In 53As which is lattice-matched to
InP, has been grown using the Japan Oxygen MOCVD system. The
growth conditions used were similar to the optimum conditions
for InP growth. Based on the results of several different
characterization techniques, optimum growth conditions for
undoped Ga 47In 53As were determined. Whenever p- and n-doped
GalnAs was required for device applications, test layers were
grown to calibrate for the required DEZn and H2S gas phase


91
mole fractions. The timing and sequence for a typical MOCVD
growth of GalnAs on InP is presented in Appendix B of this
text.
Since GaxIn.,.xAs has only one composition (x=.47) which
is lattice-matched to InP, precise control of both the TMIn
and TEGa flow rates to the reactor is crucial. A small change
in the TMIn to TEGa gas phase molar ratio is approximately
equivalent to the change in deposited solid phase molar ratio.
Unfortunately, small changes in the solid phase composition
dramatically affect the material's quality and both electrical
and optical properties. When the lattice-mismatch (Aa/a),
where a is the lattice constant, is greater than approximately
0.5% and layer thickness is greater than the critical
thickness of the material ( 1000 for GalnAs), the strain in
the epitaxial layer is enough to form cracks or dislocations
which can propagate throughout the grown film. Dislocations
appear as a "cross-hatched" pattern and are clearly visible in
the surface of a grown layer as viewed under an optical
microscope. The surface morphology of GalnAs deposited on InP
is directly related to the degree of mismatch in the thin film
relative to the substrate.
Using X-ray diffraction (XRD), it is possible to
determine the lattice constant of a deposited thin film
relative to that of the substrate. With the lattice constant,
one can determine the lattice-mismatch and composition, x, of
a mixed crystal such as GaxIn1.xAs. The XRD technique is


172
where mg/m,, = .077, ntj/m,, = .56, k = 1.38xlO*aJ/K, Tg = 898 K
(growth temperature), h = 6.624xl0'34J-sec, Eg = 1.0536eV (InP
bandgap at 625C), and m0 = 9. llxlO*31Kg. This value turns out
to be Kb4(B4) = 4.888xl029cm*3. Due to the large size of this
constant, all constants, p, and [Zn] data were converted to
mole fractions by a simple conversion factor. The factor
(which is presented in Hurle[132]) is based on the molecular
density, volume and weight of InP, and Avogadro's number. The
factor is: one mole fraction = 2.557xl021cm'3. Using this
factor, K,,4(B4) = 3.196X10'12 mole fraction2.
No direct estimate, either theoretical or experimental is
available in the literature for constants B,- Bj, and B5- B7.
So, estimates were taken from the values reported for GaAs by
Hurle[132]. These values were used as initial guesses in a
non-linear regression analysis program based on the Marquardt
method[138], which is described in the Appendix C. The
experimental hole concentration data (converted to mole
fractions) and their corresponding zinc partial pressures (in
atmospheres) were entered into the program as data set #1.
The atomic zinc data (SIMS) and their corresponding zinc
partial pressures were also entered as data set #2. Both data
sets were taken from the BNR data shown in Figure 55. The
phosphorus partial pressure, pp4 = 2.357xl0'3atm (V/III = 140) ,
was also entered as part of the equations which are shown in
Appendix C. The program was run twice, once for each data
set, and it basically fits the hole equation to data set #1


148
3.3.3 DEZn Results
The DEZn doping experiments carried out in the BNR
reactor made use of two different physical sources in order to
get a wide range of source flow rates. One source was a DEZn
bubbler which was kept at -15C (vapor pressure = 0.619 torr) ,
the other was a high pressure gas cylinder containing a
mixture of 92 ppm DEZn by volume with the balance being UHP
H2. With these concentrations, and a flow rate as low as 5
seem from the cylinder and as high as 200 seem of H2 through
the bubbler, DEZn molar flow rates from l.lxlO'7 to 4.92X10*4
moles/min were attained. Unless noted otherwise, all the
experiments were performed with the same basic conditions as
listed for the DMCd experiments.
After the growths were completed, several methods were
employed to characterize the InP:Zn thin films. The surfaces
were observed under an optical microscope and for all doping
levels, the grown layer morphology was smooth and essentially
featureless. A SIMS profile of the atomic zinc concentration
in InP:Zn sample B316 is presented in Figure 48. As shown,
the [Zn] is 7xl016cm'3 for this 1.4/xm thick layer which was
grown using 175 seem of DEZn from the gas cylinder. The
"spike" in the profile is probably due to a donor-acceptor
complex as was observed in the InP:Mg samples. The [Zn]
(SIMS) is plotted in Figure 49 versus the DEZn molar flow rate
for experiments using both the bubbler and cylinder sources.
It is evident that up to approximately 3xl0*6 moles/minute,


184
4.2.2 Electrical Theory
The RI of a semiconductor can be modulated by changing
any one of several physical parameters[148]. Of primary
interest here are the electrically controllable ones, i.e.,
the applied electric field (Franz-Keldysh effect) and the free
carrier density. The former has been reported[141], but
direct application of the high voltages necessary for
obtaining the electric field gave rise to excessive leakage
currents, which introduced thermal drift of the optical
properties.
The object of this work is to apply the high electric
field found in p-n heterojunctions to achieve modulation of
the RI without having to resort to high applied voltages.
Introducing a p-n junction into every period of the stack,
however, implies that every other junction will be reverse-
biased upon application of an external voltage. Thus, the
possibility of excessive power dissipation and attendant
thermal effects may be reintroduced if attempts are made to
pass current through the device, e.g., to inject carriers.
However, the thickness of layers determined by the
optical requirements (only lOOnm) allows neighboring junctions
to interact, as in the bipolar transistor. The stack can then
be considered as a multi-layered thyristor with at least two
stable switching states achievable. The forward voltage in
the "ON" state can be substantially lower than the breakdown
voltage of any single reverse-biased junction and can also be


16
thickness can increase along a flat susceptor in the gas flow
direction which in combination with the depletion of reactants
due to deposition, can result in nonuniform grown layer
thicknesses. This variation can be nullified by either
tilting the susceptor at an angle of 5 to 10 or inserting a
baffle into the reactor at an angle positioned above the
horizontal susceptor. This gradually reduces the cross-
sectional area in the reactor above the wafer resulting in a
gradual increase in linear gas velocity causing the boundary
layer to have a uniform thickness profile. With a constant
boundary layer thickness and constant diffusion coefficient of
group III source molecule, the flux of the mass transfer
limited reactant to the growing surface will be constant. As
a result of this, uniform semiconductor films can be grown
over large area substrates.
When MOCVD growth of mixed crystals involves the use of
more than one group III compound, GalnAs for example, solid
phase compositional non-uniformity can result. This can
result from concentration gradients in the gas flow direction
due to slight differences in the magnitude of the diffusion
coefficients. One way to avoid this problem is by selecting
group III metal organic sources with similar molecular weights
and correspondingly similar diffusion coefficients. Another
way of avoiding this problem is by using low pressure ( 0.1
atm) operation instead of atmospheric pressure. At reduced
pressures, the linear gas velocity increases and accordingly


227
the sample mounting and preparation stage of the Hall effect
experiment.
A.3 Experimental Hall Effect Measurements
After the sample has been mounted to the resistivity
holder and the holder attached to the cold finger of the
cryostat (CTI-Cryogenics, model 22), make sure that the
cryostat is perpendicular to the magnetic field. The wafer
should also face the right magnet pole labeled "North". Now
the wire from the bottom left gold pad should be connected to
the wire labelled C through a pin of the resistivity holder.
Similarly, the upper left to D, the upper right to E and the
lower right pad should be connected to wire F. These four
wires are wrapped around the cold finger and exit through the
cable connector port at the base of the cryostat. The letters
C, D, E, F are the pin labels on the male mating plug. These
four pins are connected to wires 4,3,2 and 1, respectively.
The cryostage has two lines filled with a pressurized
liquid helium attached to it which should not be bent to a
radius less than two feet (to avoid leaks). These lines with
the aid of the compressor (CTI-Cryogenics, model SC),
temperature controller (Palm Beach Cryophysics, Inc., model
4025), resistive heater (which is wrapped around the top of
the cold finger) and cryostat permit the measurement of Hall
voltages at temperatures as low as 4.2 K. An important note
about the cryostage is that whenever it is being handled, care


112
is PL. The PL spectrum for growth Q120 is shown in Figure 35.
Based on the location of the peak position, and the
temperature dependence of the quaternary bandgap, the room
temperature emission wavelength of sample Q120 is X = 1.554/zm.
This wavelength for lattice-matched quaternary corresponds to
a composition of Ga-4In_6As_913P>0a7 according to Table I of
Nahory et al.[74]. A comparison of the PL and EPMA results
for all lattice-matched quaternary samples are shown in Table
7. As shown, the agreement for some samples is quite good,
but for others, Q242 for example, the agreement is poor. This
probably is due to the fact that only InP, GaAs, and a
lattice-matched GalnAs (on InP) sample were used to calibrate
the EPMA detected atomic analysis results. It would have been
ideal to have also used an InAs and GaP sample, but these
crystals were not available for calibration. Consequently,
whenever samples were found to be lattice-matched by XRD, PL
was performed and solid phase compositions were based on PL
peak positions. The EPMA results were, however, generally
close, and they were also used as feedback for lattice
matching studies when quaternary samples were too mismatched
for XRD and PL analysis.
PL can also be used to assess grown quaternary film
quality based on the shape of peaks, the number of peaks and
also their location in the spectrum. The FWHM of sample Q120
was measured from the spectrum to be 4.46 meV at 4.2 K.


99
The Hall effect was also used to investigate the quality
of undoped GalnAs/InP samples. In this case, the room
temperature mobility (^-¡00K) and carrier concentration (N0-NA)
were used as a basis of comparison between samples grown at
different growth temperatures (600 to 700C) and different
V/III ratios (12.5 to 50) The Hall data for both studies are
presented in Table 6. As shown it appears that a growth
temperature of 600C and V/III ratios between 25 and 37.5
yield better material. It should be noted, though, that at a
V/III ratio of 25, the extended defect density was much higher
than in layers grown using V/III ratios of 37.5 and 50.0.
Since it was evident that high quality GaxIn.,.xAs can be
grown lattice-matched to InP, one important application of
this material system was investigated: low dimensional
structures involving heteroepitaxial GalnAs/InP superlattices
and multiple quantum wells. A TEM (transmission electron
micrograph) photo of a cross-section of alternating 1,150
thick layers of GalnAs and InP is shown in Figure 28. This
layer was grown as a test structure for an optical
interference filter device requiring a layer pair thickness of
2000 and also to test the interface abruptness capability of
the MOCVD system. As shown, it appears that the abruptness is
excellent and the layer pair thickness is very reproducible
throughout the stack structure; both are equally important for
good interference filter response. A SIMS profile (which was
measured at VG Inc.) of another low-dimensional structure,


RELATIVE INTENSTY
113
Figure 35: PL spectrum of sample Q120 measured at 4.2 K.


CHAPTER II
METAL ORGANIC CHEMICAL VAPOR DEPOSITION
2.1 A Brief History of MOCVD
MOCVD refers to the deposition of multiconstituent films
using one or more metal organic compounds as sources, and the
term was originated by Manasevit[ll]. He demonstrated that
single crystalline GaAs could be deposited using TEGa and AsH3
in an open tube reactor. Shortly after this report, it was
discovered that by mixing metal organics and hydrides of
different elements, binary and ternary III-V compounds such as
GaP, GaAsP, GaAsSb, AIN, GaN, InAs, GalnAs, InAsP, and InP
could be formed in a manner similar to GaAs[12-16]. Practical
information was also reported in these early efforts such as
the observation that GaAs film growth rate was mass transport
limited by the metal organic group III source and independent
of temperature below 800C. Also, n-type doping using H2S,
H2Se and p-type doping using DEZn and DMCd were achieved.
Soon after the demonstration of high quality MOCVD grown
material, all MOCVD grown devices such as FETs,
photocathodes[17,18] and GaAs/AlGaAs DH laser diodes[19] were
reported. These milestones resulted in a rapid increase in
MOCVD research and development, and as a result, it has been
demonstrated that MOCVD has the capability of growing a wide
13


12
temperatures as low as 330C by alternately pulsing PH3 and
TMIn with intervening H2 purge steps into the reactor with
each step lasting on the order of several seconds[10]. So, it
is evident that there are a wide variety of epitaxial growth
techniques for III-V compound semiconductors. Most of the
experimental work presented in this dissertation made use of
a low pressure MOCVD system. One study made use of a MBE
system for a GaAs/AlGaAs device, but the rest of the
literature review will focus on the use of MOCVD for growth of
GalnAsP/InP materials for optoelectronic device applications.


CONCENTRATION (atoma/cc)
87
PROCESSED DATA SNR
5 Aug SB Ca FILE: 0108
DEPTH (nlcponi)
Figure 21: Atomic zinc profile of sample Q105 measured by
SIMS.


78
with H2S mixture flow rates of 2 to 50 seem, which corresponds
to gas phase mole fractions of H2S of 2.8xl0'7 and 7.16xl0'6,
respectively. A C-V profile of a S-doped InP grown layer is
shown in Figure 16. As one can see, a roughly 1.1/xm thick
layer of 2.5xl018cm*3 n-type material was deposited on a n+-InP
substrate, and this was achieved using a H2S mole fraction of
2.43xl0'6. The "hump" in the profile in Figure 16 is a C-V
profiler error which occurs at interfaces. The relationship
between the measured C-V carrier concentration and H2S mixture
flow rate for several n-type samples is presented in Figure
17. As shown, a wide linear incorporation rate of sulfur in
InP is possible resulting in doping levels from 5xl017 to
2.5xl019cm'3. The surface morphology of the grown material
when viewed under the Nikon microscope at 2000x appeared
unaffected by the presence of the sulfur atoms even at the
highest n-type doping level. Room temperature Hall effect
measurements were also performed on the sulfur doped samples.
Hall carrier concentrations agreed with C-V measurements and
Hall mobilities (/i300K) ranged from 498cm2/volt-sec at the
lowest doping level to 1064cm2/volt-sec at the highest level.
2.5.2c Growth of p-tvoe InP Using DEZn
P-type conversion of MOCVD InP was achieved by mixing the
metal organic source diethylzinc (DEZn) with the standard gas
mixture used to grow undoped InP. In order to get a wide
range of p-type doping, the temperature of the DEZn bubbler
was varied from -20 to 20C resulting in a change in the


26
between leakage current of p-i-n InP based photodiodes and
substrate quality. Consequently, they set up a nondestructive
PL wafer-mapping system to evaluate grown film quality before
investing further processing time. Non-destructive techniques
such as PL mapping will remain essential unless wafer quality
control improves.
Proper wafer cleaning is also very important for the
growth of high-quality InP. Tuck and Baker[50] in 1973
published work on the chemical etching of (111) and (100) InP.
They compared the merits and disadvantages of using the
following four etching solutions: (1) 1HCL:1HN03; (2)concen
trated HC1; (3) 0.4N Fe3+; and (4) 1% bromine in methanol,
based on etching rate and hillock delineation. Nishitani and
Kotani[51] presented the use of H202-H2S04-H20 solutions for
etching (100) and (111) oriented InP. Recently, studies have
been reported using sulphur to chemically passivate the
surfaces of InP and GaAs[52-53]. The goal of this work is to
reduce the substrate surface recombination velocity in order
to improve device performance. Another interesting study
compared several wafer cleaning methods using the surface
science techniques ISS, ESCA and AES[54]. This report states
that using a 5:1:1 mixture of H2S04:H202:H20 in combination
with solvent degreasing step yields an InP surface with the
least amount of absorbed carbon and oxygen relative to the
other methods tested. Most crystal growth teams develop their


251
Y(2)=(1.OE+31 + 0.5 B(l)*B(2) 2.36E-06*X) *
((1.0+(B(3)/B(2) / 1.0E+31/0.00236))A0.5) *
B(2)*2.36E-06*X / ((1.OE+31 + B(1)*B(2) *
2.36E-06*X)Al.5)
Y(3)=0.5*B(1) 2.36E-06*X ((2*B(2)+(B(3) /
1.0E+31/0.00236)) / ((B(2)A2 + (B(3)*B(2) /
1.0E+31/0.00236))A0.5) B(l) 2.36E-06 X
((B(2)A2 + (B(3)*B(2) / 1.0E+31/0.00236))A0.5
/ (1.OE+31 + B(l)*B(2) 2.36E-06*X))
Y(4)=B(1)*0.001 X*0.5 / 1.OE+31 / ((1.0+(B(3) /
1.0E+31/B(2) / 0.00236))A0.5) / ((1.0E+31
+ B(1) B(2)*2.36E-06 X)A0.5)
where:
Y (1) =-£[Zn]
Y(2) 'aE^Zn]
v -B5
Y(3) = ,^[BZ6n]-
Y (4) =
The resulting values of the best fit are: B5=2708
fraction/atm, B6=1003 mole fraction'1, and B7=11.6 atm1/4
(49)
*
)
(50)
(51)
(52)
(53)
(54)
(55)
mole


35
is a more easily cracked species) both with varying degrees of
success. Most teams use the low pressure MOCVD technique
which makes PHj/AsHj ratios as high as 200 (which is required
for 1.0/xm wavelength quaternaries) more safely attainable.
The operating parameters inlet partial pressures, total
pressure, deposition temperature, and V/III ratio have an
effect on the growth rate and composition of deposited
quaternary materials. Because of this, numerous papers have
been written relating the gas phase growth conditions to solid
phase material quality and composition. Razeghi[65] has
published graphs on which are plotted the relationship between
growth conditions ratios and bandgap wavelengths for the full
range of quaternary materials which are lattice-matched to
InP. The three ratios are: (1) Rj = PH3 / (PH3+AsH3) ; (2) R3
= TEGa/(TEGa + TEIn) ? and (3) Rj/R3. Using these graphs,
which are only valid for a growth temperature of 650C and
total flow rate of 7 liters/min, one can estimate growth
conditions for any lattice-matched quaternary composition.
Similarly, Fujii et al.[72] and Sugou et al.[70] present
quaternary compositions as a function of (In/Ga) and (P/As)
ratios in the gas phase for fixed V/III ratios. Koukitu and
Seki[73] use a thermodynamic approach to compute the solid
composition as a function of input mole ratio for several
quaternary III-V alloy systems. They also compute the
equilibrium partial pressures of gaseous species over
GalnAsP/InP as a function of temperature, V/III ratio and


183
20x
975 n-GaAs: Si 5x17cm 3
1025 p-AI Ga As: Be 2x1017cm'3
r 03 07
1000 n+-GaAs (buffer)
Figure 60: A typical layer structure grown by MBE.


244
the inlet flange to the reactor, check the position of the
susceptor using the fork procedure and a clean wafer tray.
Adjust the position if necessary. Connect the inlet flange to
the reactor using fresh VCR gaskets. Evacuate the reactor to
3 torr as an initial leak check using RP1. Helium leak check
the reactor.
Flow H2 to the reactor and use the Matheson H2 detector
at all connections. Connect the water lines using tie-wraps.
Bake the reactor at 35 torr, 900C, 4 SLM H2 flow for four
hours. Turn on the heat tapes, too. Pressurize the reactor
to 760 torr, set all flows to 0.2 SLM or 30 seem.
B.5.5 Metal Organic Source Change (TMIn as an example)
Purge the MO line overnight with 300 seem H2. Close all
air valves on the system by hitting the red emergency stop
button, and turn off the heat tapes. Evacuate the MO line
using RP2 (open AV209, 202, 65, 63, 62, 61, 60 and NV60
completely in this order). Open AV111 to let N2 flow to the
MO. Close NV60 until PI60 reads 900 torr then open until PI60
reads 70 torr. Repeat this ten times, then put the bubbler at
760 torr by adjusting NV60. Close all air valves. Disconnect
the bubbler using particle masks and clean room gloves (make
sure bubbler valves are real tight) Install the new bubbler
using fresh VCR gaskets. Repeat purge and evacuation steps
five times. Evacuate the whole system using RP2 and helium
leak check the bubbler connections through MV207. Pressurize
the system and return all flows to 0.2 slm or 30 seem. Turn


82
co
i
O)
o
20
19
ie
17
n-Type
0 12 3
X (um)
Figure 18: A C-V profile of a DEZn doped InP film.


39
The use of the ternary material GalnAs lattice-matched to
InP for long-wavelength photodetectors is well established.
Traditionally, LPE and hydride VPE are used but, MOCVD grown
p-i-n photodiodes have also been prepared [ 65 ]. The most
important device characteristics required of detectors are low
capacitance, low dark field leakage current and high quantum
efficiency. To attain these goals, low background doping
levels, accurate lattice-matching and an abrupt p-n junction
are required and all of these are possible with MOCVD.
Actually, with MOCVD the need for a post-growth zinc diffusion
processing step can usually be eliminated since in-situ p-
doping is possible. Several teams have reported improvements
in p-i-n photodiode performance by adjusting MOCVD growth
conditions[76], layer structure[65], and Schottky barrier
height enhancement[83]. MOCVD grown structures with leakage
currents as low as 3 pA at -10V using a 100/im device diameter
have been fabricated on two inch diameter InP substrates[84].
In addition to the above structures, a number of other
optoelectronic devices have been fabricated using MOCVD grown
InP based materials. Two dimensional electron gas (2DEG) and
multiple quantum well structures have been grown making use of
the extremely high mobilities, (in excess of 180,000cm2/volt-
sec at 9.2 K) possible with these materials[85]. Guided wave
devices such as optical waveguides and phase modulators have
also been grown[86]. Finally, GalnAsP/InP interference
filters have recently been grown by MOCVD[87]. The theory


CHAPTER IV
EPITAXIALLY GROWN INTERFERENCE FILTERS
4.1 Theory of Interference Filters
If a beam of electromagnetic radiation is incident upon
a structure consisting of films of different materials,
multiple reflections will occur within the structure[139].
Should the distances between the various boundaries be on the
order of the wavelength of the incident light, the reflected
beams will interfere. An optical interference stack makes use
of this effect. These stacks or periodic multilayer di
electric films find application as bandpass filters and
mirrors. Usually, the multilayer is a guarter-wave stack
consisting of alternating layers of high and low refractive
index material with the optical thickness of each layer a
quarter of the particular required wavelength. Passive
interference filters have been previously grown by MBE
[140,141] and MOCVD[142] in the GaAs/AlGaAs material system.
They have also been grown by CBE[143] and MOCVD [87,144] in
the GalnAsP/InP material system.
The most general method of calculating the transmittance
and the reflectance of a multilayer is based on a matrix
formulation of the boundary conditions of the film surfaces
derived from Maxwell's equations[145]. Basically, a two by
177


221
Figure 78; The Hall effect in a semiconductor.


30
used p-dopant source is DEZn which can be used to dope InP
over the range p = I015-I018cm3. DEZn is sufficient for most
device applications, but its relatively high diffusion
coefficient at typical growth temperatures, D = 3 x 10'13
cm2/sec[58], does make it unsuitable for some device
applications. For a more extensive discussion on p-doping of
InP, the reader is referred to Chapter III of this
dissertation.
2.3.2. GalnAs/InP
The ternary compound Ga^n^As can be grown lattice-
matched to InP by MOCVD. Unlike the AlGaAs/GaAs material
system, the GalnAs/InP system is not lattice-matched for all
compositions. Ga 47In 53As has an energy gap of 0.75 eV (Xg =
1.67 tm) and is the only composition which is lattice-matched
to InP. This ternary film can be grown by carefully
controlling the gallium to indium metal organic source
composition of the gas inlet to an MOCVD reactor. This is one
of the most severe heteroepitaxial growth scenarios possible
as the group V sublattice must be changed from pure phosphorus
to pure arsenic.
The first reported MOCVD growth of GaxIn1.xAs was on a
GaAs substrate[61] and hence was lattice-mismatched. The
early efforts provided useful information such as
compositional uniformity, merits of methyl versus ethyl MO
sources, gas phase reactions and purity for later GalnAs/InP
work. One conclusion that was useful for GalnAs/InP work was


84
Figure 19: The effect of DEZn partial pressure on InP hole
concentration.


88
Figure 22: C-V profile of growth Q095 during which the growth
pressure was varied.


APPENDIX C
DOPANT MODEL COMPUTER PROGRAM
As stated in section 3.3.4, a non-linear regression
analysis program based on the Marquardt method was used to
determine values for the equilibrium constants B^Bj, Bs-B7.
The program is not listed here due to its length (which is
approximately 40 pages) The program was written by Dr. Alkis
Constantinides and is available on a diskette which comes with
his book, Applied Numerical Methods with Personal Computers
[138]. A description of how to use the program is given in
section 7-5 of the book. Briefly, regression analysis is the
application of mathematical and statistical methods for the
analysis of experimental data, and the fitting of mathematical
models to these data by the estimation of the unknown
parameters of the models. By performing statistical tests,
the model can be identified or verified. The Marquardt method
uses an interpolation technique which is a combination of the
Gauss-Newton and the steepest-descent methods to obtain values
of the parameters in the model which minimize the overall
(weighted) sum of the squared residuals.
The non-linear regression program is menu-driven for data
input and adjustment. To use the program, one must first
derive a model and variational equations. The variational
equations are obtained by taking partial derivatives of the
248


29
decreases when the growth temperature increases. For SiH4,
the opposite trend is observed, because the incorporation of
silane is reaction limited whereas H2S is adsorption limited.
Si is amphoteric in InP, acting as an acceptor or donor,
depending on site selection (adjusted by changes in the V/III
ratio used during the growth) H2S does not compensate itself
in InP but the diffusion coefficient of S is greater than that
of Si in InP. Depending on the device application, a suitable
n-type dopant source for InP is apparently available.
InP can also be doped p-type by adding an InP acceptor
species to the inlet gas stream of the reactor. The gas
stream must contain enough of an acceptor species to increase
the electrically active extrinsic acceptor level above the
electrically active intrinsic donor level. The metal organic
(MO) compounds DESn[45], DMZn[58], DMCd[59], Cp^gteO] and
MCp2Mg[61] act as sources for acceptors in InP with varying
degrees of success. There is not one single MO acceptor
source which dopes InP p-type over a wide doping range (1015-
I019cm'3) without extended diffusion or surface morphology
degradation. This is the reason that so many different
sources have been investigated for InP as suitable p-doping
sources. This dilemma, in combination with the observation
that the electrical activation of some p-type dopants in InP
is much less than unity, was a driving force for the extensive
literature review and model development for p-doping of MOCVD
InP in Chapter III of this dissertation. The most commonly


225
microelectronics lab at the University of Florida). If all
possible combinations of contacts taken two at a time yield
linear current-voltage plots over a wide range of applied
voltage values, then, the contacts are ohmic. If they are not
all ohmic, then the alloying process must be improved by trial
and error.
Having prepared a square sample with ohmic contacts and
no visible wafer damage (cracks, oxide haze) then this sample
must be attached to an adequate sample holder. For routine
measurements (300 and 77 K) a patterned printed circuit board
with metal clips is available. Samples can be mounted on the
board and inserted into a dewer which rests on a wooden stand
in between the magnets two poles. The dewer could be filled
with liquid nitrogen or left empty depending upon the
temperature required. A sample holder was also developed for
temperature dependent experiments (down to 4.2 K) which
consists of three interconnecting parts: the CTI-Cryogenics
model SH-14R resistivity holder, an oxygen-free copper
conductive plate, and a gold coated alumina mounting plate.
The resistivity holder can be rotated so the semiconductor is
perpendicular to the magnetic field and is also made of oxygen
free copper. The copper plate was made to fit on one face of
the resistivity holder with a channel in it to hold the
alumina plate.
The gold covered alumina plate is the medium by which the
current and voltage leads could be attached to the four ohmic


14
variety of device quality III-V compound semiconducting
materials.
The MOCVD process has been used for the epitaxy of most
III-V compound semiconductors. The basic overall reaction is
Mjjj (Alk)3(g) + XvH3(g) > MjjjX^s) + 2 Alkane(g) In this
reaction, the organometallic and the hydride typically are
irreversibly pyrolyzed by the heat of the susceptor and
substrate to form molecular (or atomic) intermediates which
may react in the gas phase or on the substrate surface[20].
Some researchers say that the breaking of metal-carbon bonds
occur on the semiconductor surface [21]. Contradicting reports
such as these may be related to differences in test conditions
or configurations from team-to-team, but the only conclusion
that can be made is that the complete MOCVD process is not
well understood at this time.
Each step of the MOCVD process is not known, but there
are several practical trends that are generally agreed upon.
Some trends for GaAs and InP are: (1) III-V MOCVD growth
between 550 and 750C is mass transport limited in the column
III source (below 550C growth is reaction limited); hence
growth rates are determined by metal organic fluxes and are
temperature independent; (2) for ternary and quaternary films,
group III solid phase compositions are linearly related to gas
phase group III compositions, however, solid phase group V
compositions have a non-linear dependence on gas phase group
V compositions due to large differences in cracking


230
It was suggested that a test be performed to determine if
the readout temperature is a good representation of the actual
temperature of the sample. This was performed by placing a
second silicon diode sensor, with the same I/V characteristics
as the cold finger sensor, at the exact sample position. Its
four plugs were connected temporarily to four leads from the
37 pin mating plug of the temperature controller. Since the
controller is capable of displaying the temperature at both
sensor locations, this experiment was fairly straight forward.
The results of the test are that it takes approximately
fifteen minutes for the sample location to stably reach the
same temperature as the permanent sensor location.
Once the desired temperature for measurement has been
reached and is stable, the magnetic field should be applied.
To do this, the constant current supply (Walker Scientific,
model HS525) for the magnet (Walker Scientific, model HV4H)
should be turned on by pressing the power button. Next the
cooling water must be turned on to a flow rate great enough to
illuminate the "DC-OFF switch (this indicates adequate flow
through the current supply and magnet). Of course make sure
that the exit line for the water is in the drain and there are
no leaks. Also, make sure that the spacing between the magnet
poles is adequate and that the poles are locked in place.
Now, with the percentage of maximum current dial reading zero,
press the "DC-ON" switch which will activate the current
generator. At this point it is safe to slowly dial in a


ACKNOWLEDGEMENTS
The author would like to thank his advisor and committee
chairman, Dr. T. J. Anderson, for his guidance and
encouragement throughout this educational program. Also, many
thanks go to the other members of his graduate committee, Dr.
M. E. Orazem, Dr. G. Bosman, Dr. S. S. Li and Dr. A. J.
SpringThorpe, for their advice and useful discussions.
The author wishes to extend his deepest thanks to Dr.
Balu Pathangey, Dr. Yasuhiro Hayakawa, Mrs. Vesna Jovic and
Mr. Pete Axson for their friendship and assistance in the
construction and operation of the MOCVD system. Special
thanks are in order for Dr. Pathangey for his assistance with
the dopant modeling portion of this work. The author is also
indebted to Mr. Keith Rambo, Mr. Dean Heinz and Mrs. Cheryl
Heinz, Dr. Shiro Sakai and Mr. Whitey Herrlinger for their
assistance at the University of Florida Surge Area.
The author had the good fortune to have been a visitor at
Bell Northern Research (BNR) in Ottawa, Canada, and is
grateful for the hospitality and advice of Dr. A. J.
Spr ingThorpe, Dr. C. Blaauw, Dr. M. N. Svilans, Dr. C. J.
Miner and Dr. I. C. Bassignana and several other members of
the scientific staff at BNR. The author also gratefully
ii


GROWTH AND MODELING OF III-V COMPOUND
SEMICONDUCTOR OPTOELECTRONIC MATERIALS
WITH DEVICE APPLICATIONS
By
ARNOLD JOHN HOWARD
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1990

ACKNOWLEDGEMENTS
The author would like to thank his advisor and committee
chairman, Dr. T. J. Anderson, for his guidance and
encouragement throughout this educational program. Also, many
thanks go to the other members of his graduate committee, Dr.
M. E. Orazem, Dr. G. Bosman, Dr. S. S. Li and Dr. A. J.
SpringThorpe, for their advice and useful discussions.
The author wishes to extend his deepest thanks to Dr.
Balu Pathangey, Dr. Yasuhiro Hayakawa, Mrs. Vesna Jovic and
Mr. Pete Axson for their friendship and assistance in the
construction and operation of the MOCVD system. Special
thanks are in order for Dr. Pathangey for his assistance with
the dopant modeling portion of this work. The author is also
indebted to Mr. Keith Rambo, Mr. Dean Heinz and Mrs. Cheryl
Heinz, Dr. Shiro Sakai and Mr. Whitey Herrlinger for their
assistance at the University of Florida Surge Area.
The author had the good fortune to have been a visitor at
Bell Northern Research (BNR) in Ottawa, Canada, and is
grateful for the hospitality and advice of Dr. A. J.
Spr ingThorpe, Dr. C. Blaauw, Dr. M. N. Svilans, Dr. C. J.
Miner and Dr. I. C. Bassignana and several other members of
the scientific staff at BNR. The author also gratefully
ii

acknowledges the financial support of BNR, DARPA and
Microfabritech for portions of this study.
Warm, personal thanks go to the author's family,
especially his wife Robin, and his mother Mary, for their
moral support and patience throughout the course of his
graduate studies. This work is dedicated to them and also to
his father, the late Peter V. Howard.
iii

TABLE OF CONTENTS
page
ACKNOWLEDGEMENTS
ABSTRACT vi
CHAPTERS
I INTRODUCTION 1
1.1 III-V Semiconductors 1
1.2 Epitaxial Growth Techniques 6
II METAL ORGANIC CHEMICAL VAPOR DEPOSITION 13
2.1 A Brief History of MOCVD 13
2.2 MOCVD Systems 15
2.3 A Review of the Literature on InP Based MOCVD 22
2.3.1 InP Homoepitaxy 22
2.3.2 GalnAs/InP 30
2.3.3 GalnAsP/InP 33
2.3.4 InP Based Devices 37
2.4 A Description of the MOCVD System 40
2.4.1 Introduction 40
2.4.2 Gas Delivery System 43
2.4.3 Reactor and Heating System 47
2.4.4 Exhaust/Scrubbing System 50
2.4.5 Safety 51
2.5 Determination of Optimum Growth Conditions
Based on Thin Film Characterization 52
2.5.1 Experimental Approach 52
2.5.2 InP 56
2.5.3 Growth of GalziAs Lattice-Matched to
InP 90
2.5.4 Growth of GalnAsP Lattice-Matched to
InP 108
IIIP-TYPE DOPING OF MOCVD INP: EXPERIMENTS AND
MODELING 118
3.1 A Review of the Literature on p-Type Doping
of InP 118
3.1.1 Introduction 118
3.1.2 Bulk Crystal Growth 119
iv

3.1.3 Liquid Phase Epitaxy 121
3.1.4 Molecular Beam Epitaxy 122
3.1.5 Chemical Vapor Deposition 123
3.2 MOCVD Growth and Characterization of Mg-Doped
InP Using bis-(Methylcyclopentadienyl)
Magnesium as a Dopant Source 130
3.2.1 Introduction 130
3.2.2 MOCVD Growth 130
3.2.3 Results and Discussion 131
3.2.4 Conclusions 145
3.3 Experimental DMCd and DEZn for p-Type Doping
of InP by MOCVD 145
3.3.1 Introduction 145
3.3.2 DMCd Results 146
3.3.3 DEZn Results 148
3.4 Modeling of p-Type Doping of InP using DEZn.. 156
3.4.1 Introduction 156
3.4.2 Point Defect Structure 164
3.4.3 Discussion of Results 173
IV EPITAXIALLY GROWN INTERFERENCE FILTERS 177
4.1 Theory of Interference Filters 177
4.2 MBE Grown AlGaAs/GaAs Devices 182
4.2.1 Introduction 182
4.2.2 Electrical Theory 184
4.2.3 Experimental 185
4.2.4 Electrical Testing 193
4.2.5 Conclusions 204
4.2.6 Addendum 205
4.3MOCVD Grown GalnAsP Devices 209
V CONCLUSIONS AND RECOMMENDATIONS 216
APPENDICES
A THEORY AND OPERATION OF THE LOW TEMPERATURE HALL
EFFECT SYSTEM 220
B OPERATION OF THE QUATERNARY MOCVD SYSTEM 234
C DOPANT MODEL COMPUTER PROGRAM 248
REFERENCES 252
BIOGRAPHICAL SKETCH 262
v

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy
GROWTH AND MODELING OF III-V COMPOUND
SEMICONDUCTOR OPTOELECTRONIC MATERIALS
WITH DEVICE APPLICATIONS
By
ARNOLD JOHN HOWARD
December 1990
Chairperson: Timothy James Anderson
Major Department: Chemical Engineering
Several topics have been undertaken during the course of
this degree which are associated with understanding and
improving semiconductor processing. The growth, modeling and
characterization of III-V compound semiconductor materials and
optoelectronic devices has been emphasized. Epitaxial layers
of GaxIn^xASyP^y with lattice-matched alloy compositions over
the range from x=0, y=0 (InP) to x=0.47, y=l (Ga 47In 53As)
have been grown by metal organic chemical vapor deposition
(MOCVD) on InP substrates. Both the MOCVD system, used to
grow these layers, and a low temperature Hall effect system,
used to characterize these layers, were designed and
installed. The results from several other analytical
techniques were used to determine the optimal growth
conditions for high quality epitaxial layers.
vi

The use of diethylzinc (DEZn), bis-(methylcyclo-
pentadienyl) magnesium (MCpjMg) and dimethylcadmium (DMCd) as
p-type dopant sources for MOCVD InP was investigated at BNR in
Ottawa, Canada. It has been experimentally observed that the
carrier concentration dependence on dopant partial pressure in
the MOCVD reactor is different for each of these three
dopants. A novel model of the p-doping process of MOCVD InP
using DEZn has been developed that incorporates an equilibrium
boundary condition between the gas phase and solid phase point
defects. The results of this model indicate that at high DEZn
gas phase mole fractions, which results in low solid-phase
electrical activity, the dominant electrically inactive point
defects are intersticial zinc and zinc complexed with a
phosphorous divacancy.
A novel optoelectronic device has been fabricated and
modeled which contains p-n heterojunctions in an optical
interference filter. Structures were grown by molecular beam
epitaxy at BNR using the GaAs/AlGaAs material system and by
MOCVD at the University of Florida using the InP/GalnAsP
material system. Structures with peak reflectivities at 1.3
and 1.40 microns were grown and good crystalline quality were
confirmed. Electrical bistability was observed in a forty-
layer device which has never been reported before in a
structure of this size.
vii

CHAPTER I
INTRODUCTION
1.1 III-V Semiconductors
Since the invention of the transistor in 1948 by
Shockley, Brattain, and Bardeen, there has been a revolution
in the electronics industry. Up to that time the vacuum tube
diode and triode were the most used electronic devices, but
then the transistor device using a semiconductor crystal as
its starting material was fabricated. The microchip, which is
the fundamental building block of present day computers,
contains a large number of tiny semiconductor transistors
using typically single crystals of silicon as a starting
material. Silicon has been the "workhorse" for the
electronics industry primarily due to its availability in high
single crystalline purity, ease of use in device fabrication,
and of course its good electrical properties. But, the
relatively low electron mobility and fixed indirect bandgap of
silicon makes it not suitable for present-day optoelectronic
device applications. As a consequence of these new demands,
research into the development of semiconductors with variable
electrical and optical properties has flourished.
Compound semiconductors such as GaAs, InP and others
composed of elements from group IIIA (Al, Ga, In) and group VA
1

2
(P, As, Sb) columns of the periodic table have electrical and
optical properties superior to those of silicon for certain
modern-day device applications. III-V materials have a wide
range of bandgap energies (0.18 to 2.4 eV), where the bandgap
energy is defined as the energy difference between the lowest
electron state in the conduction band and the highest hole
state in the valence band allowed in the semiconductor. Some
compound semiconductors have direct bandgaps, meaning that the
conversion of photons (light) to electrons (energy) or vice
versa, does not involve a third particle, such as a phonon.
The direct bandgap III-V compounds also have large electron
mobilities where mobility is defined (at low electric fields)
as the ratio of absolute electron velocity to the magnitude of
the electric field. A listing of these parameters and the
lattice constants of silicon and binary III-V semiconductors
is shown in Table 1(1]. As shown in this table, as much as a
two order of magnitude increase in electron mobility is
possible by using III-V compound semiconductors instead of
silicon for electronic devices. It is also significant to
note that a wide range of compound semiconductors can be
formed by creating solid solutions of the individual
semiconductors? hence, a wide selection of compound
semiconductors exists with a wide range of electrical and
optical properties.

3
Table 1
Properties of Silicon and III-V Binary
Semiconductors at 300 K
Bandgap
Type
Bandgap
Energy(eV)
Electron
Mobility
(cm2/V-s)
Lattice
Constant
(Angstroms)
Si
indirect
1.12
1350
5.43
InSb
direct
0.18
100000
6.48
InAs
direct
0.36
22600
6.06
GaSb
direct
0.70
5000
6.09
InP
direct
1.28
4000
5.87
GaAs
direct
1.43
8500
5.65
AlSb
indirect
1.60
200
6.14
AlAs
indirect
2.16
180
5.66
GaP
indirect
2.26
300
5.45
A1P
indirect
2.45
80
5.46
Source: Streetman[1].

4
Since III-V compound semiconductors present a wide range
in values of direct bandgap energy, mobility, and lattice
constant, semiconductor devices have wider ranges of
application. The bandgap energy (Eg) of a semiconductor is
related to the cut-off wavelength (Ag) by the following
equation: Eg(eV) = 1.24/Ag (un). The cut-off wavelength of a
semiconductor is the longest wavelength to which a detector
fabricated from this same semiconductor will respond. Another
degree of freedom available is the ability to form completely
miscible substitutional solid solutions independently on both
the group III and group V sublattices. In other words, not
only simple binary III-V compounds, but also III-III'-V or
III-V-V ternary and III-III'-V-V quaternary single
crystalline semiconductors such as Al^a^^s, GaxIn.,.xAs,
GaASyP^y and Ga^n^^SyP^y can be created. By using ternary
and quaternary semiconductors, it is possible to vary the
physical and electrical properties of these materials
continuously between the property limits of the constituent
binary compounds listed in Table 1. A plot of the lattice
constant versus the bandgap energy (at 300 K) for III-V
compound semiconductors is shown in Figure 1. Solid dots
indicate binary compounds, solid lines connecting dots
represent direct bandgap ternary solid solutions, and dashed
lines connecting dots represent indirect bandgap ternary solid
solutions.

Lattice Parameter (ft)
5
Figure 1; Lattice parameter and bandgap energy of various
III-V semiconductors

6
Basically, the entire area bounded by the solid and dashed
lines is available for use in the design of new III-V compound
semiconductor devices.
The cross-hatched area shown in Figure 1 is the lattice
parameter-bandgap energy space of the quaternary material
Ga^n^ASyP^y. This material has many optoelectronic device
applications due to its wide range of bandgap energy (0.36 to
2.26 eV) and possible lattice constants. Most semiconducting
single crystalline ternary and quaternary materials are
epitaxially grown on a substrate of nearly the same lattice
constant. Hence, the two compositional degrees of freedom
available with the GalnAsP system are important because
presently only GaAs, GaSb, GaP, InP, InAs and InSb are
available for use as substrate materials. For GalnAsP on InP,
or any other heteroepitaxial materials system, a difference of
lattice constant (lattice-mismatch) of greater than 0.1%
between the grown film and substrate leads to, for film
thicknesses greater than the critical thickness, the formation
of structural defects which can degrade device performance.
This problem puts strict demands on the epitaxial growth
technique employed.
1.2 Epitaxial Growth Techniques
The word "epitaxy" is derived from Greek and means
"arranged upon." Epitaxial films of III-V materials are
usually grown or arranged upon substrates with equivalent
crystalline structure and lattice constant. The two most

7
common substrate materials used today for III-V homoepitaxy,
(i.e., growth on a crystal of the same composition), or
heteroepitaxy, (i.e., growth on a crystal of different
composition or crystal structure), are GaAs and InP. The
substrates are cut from bulk crystals along a particular
crystal orientation from two to four inch diameter boules
which are sometimes created by withdrawing a seed crystal from
a heated liquid melt. Epitaxial growth on these substrates is
accomplished by exposing the heated surface to a flux of group
IIIA metals and group VA non-metals. The flux can be supplied
from a liquid, vapor, or molecular beam source. This
distinction defines the three primary methods for growing
epitaxial III-V films: liquid phase epitaxy (LPE), vapor
phase epitaxy (VPE), (also known as chemical vapor deposition
(CVD)), and molecular beam epitaxy (MBE). Each technique has
its own advantages and drawbacks, which will be discussed in
the following paragraphs.
LPE is a growth technique which can be used to deposit
thin single crystal layers of III-V compound semiconductors
from a heated liquid solution by decreasing the temperature of
the substrate relative to the solution. It is a relatively
simple, inexpensive, near equilibrium (reproducible) growth
technique that is well understood. The growth rate can be
high and a wide range of both p- and n-type dopants are
available and their incorporation is controllable. LPE has
been used to grow InP[2] and high quality GaInAs[3] on InP for

8
laser applications. However, problems such as surface
defects, poor thickness and compositional uniformity, and
difficulty in growing abrupt heterojunctions have made the LPE
technique unsuitable for present-day device fabrication
demands. For simple layer structures such as the AlGaAs solid
state laser used in compact disc players, LPE is perfectly
adequate.
There are three distinct VPE or CVD chemistries: chloride
CVD, hydride CVD and metal organic CVD. The chloride (or
sometimes referred to as halide) CVD process for GaAs growth,
as an example, uses AsCl3 and metallic Gallium as sources in
an open tube system with H2 (as a carrier gas) to transport
reactants from the source zone, through a temperature gradient
zone to the deposition zone. The chloride CVD process is a
surface-kinetically limited process requiring careful source
composition control and accurate temperature control
throughout the system for reproducibility. Also, it is
difficult to vary the V/III ratio and transients are long so
abruptness is bad in chloride CVD. GalnAsP has been grown by
the chloride CVD method[4] but other CVD techniques are more
convenient and flexible for growing ternary and quaternary
III-V compounds. Hence, the chloride process is usually only
used to grow high purity epitaxial GaAs.
The hydride CVD process for growth of III-V compound
semiconductors differs from the chloride process by replacing
column V chlorides such AsC13 or PC13 with column V hydrides

9
like AsH3 or PH3. For InP growth, HC1 gas is first reacted
with liquid indium metal in the source zone. The gaseous
product InCl is then carried by H2 to mix and react with PH3
to deposit InP in the growth zone. Similar to the chloride
process, accurate temperature control is required for this
three-zone process which is also surface-kinetically limited
in the low-temperature growth regime. The hydride process is
currently widely used for light-emitting diode (LED)
applications using GaAs.,_xPx. It has also found application
in the growth of III-V GalnAsP and GalnAs for LED's, lasers
and detectors[5]. One major advantage the hydride system
provides over the chloride system is the ability to vary the
vapor phase V/III molar ratio by adjusting the inlet flow
rates of hydrides and HC1. One drawback of both the hydride
and chloride systems is that they are hot-wall systems;
interaction between the gas and heated Si02 reactor wall
occurs which results in unintentional silicon incorporation
into grown layers. Due to its successful use, especially in
LED fabrication, hydride CVD will continue to have a
significant role in the growth of III-V materials.
The third type of CVD or VPE process is metal organic
chemical vapor deposition (MOCVD) The MOCVD process involves
an irreversible pyrolysis reaction of vapor-phase mixtures
usually of group IIIA metal organic sources and group VA
hydride sources. For InP, as an example, trimethylindium
(TMIn) and PH3 diluted in H2 would flow into an open cold-wall

10
quartz tube, decompose in the presence of a heated substrate,
and then deposit an epitaxial layer. Under normal deposition
conditions, the MOCVD process is kinetically limited by mass
transport of the column III source through a stagnant layer
near the growing surface. The MOCVD process is capable of
growing a wide variety of films with excellent abruptness
uniformity over large substrate areas. The principal device
area where MOCVD has made an impact is optoelectronics. A
thorough review of the MOCVD literature has been written by
Ludowise[6] and a brief history of MOCVD with emphasis on InP
based materials and device applications is presented in
section 2.1.
Molecular beam epitaxy (MBE) is a technique capable of
growing epitaxial films one atomic layer at a time. MBE makes
use of controlled evaporation from one or more thermal sources
to direct beams of atoms or molecules onto a heated substrate
under ultra-high vacuum conditions. During a MBE growth the
substrate temperature is generally kept relatively low (500-
600C for GaAs) MBE growth rates are typically slow (0.1 -
2/xm/hr) which in combination with low growth temperatures
permits precise layer thickness, doping and compositional
control[7]. For GaAs, the As4 beam flux is much greater than
the Ga beam flux, and both fluxes are dependent upon the
temperature of the effusion oven, molecular weight of the
emitted atom, orifice area, and source cell to wafer distance.
With a properly placed two-inch rotating wafer, nonuniformity

11
of the growth film may be reduced to less than ten percent.
Numerous devices such as LEDs, lasers, FETs and HBTs have been
fabricated with MBE. One unique bonus with MBE is that in-
situ monitoring devices such as RHEED, mass spectrometers,
Auger spectrometers and ion gauges are feasible and
commonplace. However, large expense and limited throughput
restrict MBE usage.
During the last decade, several novel deposition
techniques for III-V compound semiconductors have emerged.
Each one is a spin-off of either MOCVD, MBE, or a combination
of the two and has some relative advantages and disadvantages;
none of these new techniques are widely used. One such
technique is called atomic layer epitaxy or ALE. For III-V
compound semiconductors, ALE proceeds by depositing a
monolayer of a group III metal followed by depositing a
monolayer of group V atoms, in a layer-by-layer sequence[8].
In ALE, grown layer thickness is determined by the number of
cycles rather than the time of growth. Another relatively new
technique is called chemical beam epitaxy (CBE) or metal
organic molecular beam epitaxy (MOMBE). In MOMBE, all of the
group IIIA and VA sources are metal organic; TMAs replaces As4
and AsH3, TEP replaces P2 and PH3[9]. The remaining aspects
of MOMBE are essentially the same as conventional MBE. A
final technique which is similar to ALE but is performed in a
conventional low pressure MOCVD system is called flow
modulation epitaxy or FME. FME has been used to grow InP at

12
temperatures as low as 330C by alternately pulsing PH3 and
TMIn with intervening H2 purge steps into the reactor with
each step lasting on the order of several seconds[10]. So, it
is evident that there are a wide variety of epitaxial growth
techniques for III-V compound semiconductors. Most of the
experimental work presented in this dissertation made use of
a low pressure MOCVD system. One study made use of a MBE
system for a GaAs/AlGaAs device, but the rest of the
literature review will focus on the use of MOCVD for growth of
GalnAsP/InP materials for optoelectronic device applications.

CHAPTER II
METAL ORGANIC CHEMICAL VAPOR DEPOSITION
2.1 A Brief History of MOCVD
MOCVD refers to the deposition of multiconstituent films
using one or more metal organic compounds as sources, and the
term was originated by Manasevit[ll]. He demonstrated that
single crystalline GaAs could be deposited using TEGa and AsH3
in an open tube reactor. Shortly after this report, it was
discovered that by mixing metal organics and hydrides of
different elements, binary and ternary III-V compounds such as
GaP, GaAsP, GaAsSb, AIN, GaN, InAs, GalnAs, InAsP, and InP
could be formed in a manner similar to GaAs[12-16]. Practical
information was also reported in these early efforts such as
the observation that GaAs film growth rate was mass transport
limited by the metal organic group III source and independent
of temperature below 800C. Also, n-type doping using H2S,
H2Se and p-type doping using DEZn and DMCd were achieved.
Soon after the demonstration of high quality MOCVD grown
material, all MOCVD grown devices such as FETs,
photocathodes[17,18] and GaAs/AlGaAs DH laser diodes[19] were
reported. These milestones resulted in a rapid increase in
MOCVD research and development, and as a result, it has been
demonstrated that MOCVD has the capability of growing a wide
13

14
variety of device quality III-V compound semiconducting
materials.
The MOCVD process has been used for the epitaxy of most
III-V compound semiconductors. The basic overall reaction is
Mjjj (Alk)3(g) + XvH3(g) > MjjjX^s) + 2 Alkane(g) In this
reaction, the organometallic and the hydride typically are
irreversibly pyrolyzed by the heat of the susceptor and
substrate to form molecular (or atomic) intermediates which
may react in the gas phase or on the substrate surface[20].
Some researchers say that the breaking of metal-carbon bonds
occur on the semiconductor surface [21]. Contradicting reports
such as these may be related to differences in test conditions
or configurations from team-to-team, but the only conclusion
that can be made is that the complete MOCVD process is not
well understood at this time.
Each step of the MOCVD process is not known, but there
are several practical trends that are generally agreed upon.
Some trends for GaAs and InP are: (1) III-V MOCVD growth
between 550 and 750C is mass transport limited in the column
III source (below 550C growth is reaction limited); hence
growth rates are determined by metal organic fluxes and are
temperature independent; (2) for ternary and quaternary films,
group III solid phase compositions are linearly related to gas
phase group III compositions, however, solid phase group V
compositions have a non-linear dependence on gas phase group
V compositions due to large differences in cracking

15
temperatures (PH3 is more difficult to decompose than AsH3) ;
(3) gas phase V/III ratios have a strong effect on background
carrier concentrations and p- and n-type doping levels; (4) in
some cases, group III metal organics react with group V
hydrides to form adducts and polymers that may be involatile
liquids or solids; and (5) group II alkyls act as p-type
dopant sources, group VI hydrides act as n-type dopant
sources, but depending on growth conditions, group IV hydrides
can act as donor or acceptor sources in III-V's. The extent
to which these trends apply to all III-V materials does vary,
but they are definitely useful in designing an MOCVD system or
in optimizing material specific growth conditions.
2.2 MOCVD Systems
Most MOCVD systems use quartz reactors oriented either
vertically[11] where gas flow is usually down, or horizontally
[22] where gas flow is usually over a wedge shaped susceptor.
Other less common systems incorporate barrel reactors[23]
where growth can occur on multiple wafers with reactants
flowing from top to bottom, or "chimney" reactors[24] where
gas flow is up and wafers and susceptor are held vertically.
In a vertical reactor, uniform growth rates are more difficult
to achieve than in a horizontal reactor because with the
geometry of a vertical reactor, nonlaminar, turbulent gas flow
can more easily occur. In a horizontal reactor, a boundary
layer zone forms above the susceptor where the flow rate is
lower than that of the bulk gas above. The stagnant layer

16
thickness can increase along a flat susceptor in the gas flow
direction which in combination with the depletion of reactants
due to deposition, can result in nonuniform grown layer
thicknesses. This variation can be nullified by either
tilting the susceptor at an angle of 5 to 10 or inserting a
baffle into the reactor at an angle positioned above the
horizontal susceptor. This gradually reduces the cross-
sectional area in the reactor above the wafer resulting in a
gradual increase in linear gas velocity causing the boundary
layer to have a uniform thickness profile. With a constant
boundary layer thickness and constant diffusion coefficient of
group III source molecule, the flux of the mass transfer
limited reactant to the growing surface will be constant. As
a result of this, uniform semiconductor films can be grown
over large area substrates.
When MOCVD growth of mixed crystals involves the use of
more than one group III compound, GalnAs for example, solid
phase compositional non-uniformity can result. This can
result from concentration gradients in the gas flow direction
due to slight differences in the magnitude of the diffusion
coefficients. One way to avoid this problem is by selecting
group III metal organic sources with similar molecular weights
and correspondingly similar diffusion coefficients. Another
way of avoiding this problem is by using low pressure ( 0.1
atm) operation instead of atmospheric pressure. At reduced
pressures, the linear gas velocity increases and accordingly

17
the stagnant layer thickness decreases. Also, at lower
pressures, diffusion coefficients increase, making more abrupt
heterojunctions possible. Low pressure operation also reduces
the occurrence probability of unwanted parasitic reactions.
Because of all of these advantages, low pressure MOCVD growth
has become widely used especially for multi-wafer scale-up
production applications.
Another problem with the traditional horizontal "Bass
type" [17] reactor is due to the fact that having a cold dense
gas above a hot, less dense gas is unstable because of
gravity. This can cause natural convection which results in
closed stream-line gas flow patterns. This problem can
usually be minimized by operating at reduced pressures[25].
The best solution, however, appears to be the use of an
inverted reactor geometry[26] which completely eliminates
thermal buoyancy effects. In this geometry, the susceptor is
located at the highest and hottest point of a horizontal
reactor with the wafer mounted upon it facing downwards.
Another obvious benefit with this design is the elimination of
the problem of particles falling on the substrate before and
during epitaxial growth which can lead to structural defects.
With this design, improved GalnAs compositional uniformity and
a complete elimination of parasitic deposition on the quartz
wall opposite the susceptor have also been reported[26].
Another final technique that has been used to improve both
thickness and compositional uniformity is the use of moving

18
substrate holders in both circular and even planetary motion
configurations[27]. These techniques can greatly improve
uniformity, but also, unfortunately, greatly add to machine
complexity and expense.
Aside frcm the reactor, the other major parts of an MOCVD
system are the gas delivery, heating exhaust/scrubbing and
safety systems. Most gas handling systems are constructed
from high purity stainless steel tubing, valves (air-operated
and manual), regulators, electronic mass flow controllers and
filters or purifiers. The system typically delivers metal
organics (from bubblers), hydrides (from high pressure gas
cylinders) and most often hydrogen (from a palladium-alloy
purifier) to a fast switching manifold which directs gases to
the reactor or to a vent line. Gas manifolds should be
situated as close to the reactor inlet as possible to minimize
tube length and improve interface abruptness capabilities.
Most systems use manifolds with a linear valve arrangement,
but only in a radial manifold arrangement is the length from
each valve to the reactor potentially the same for all of the
gases[28]. In the "vent/run" type system discussed above, the
vent line and reactor line are "pressure balanced" so that
transient times associated with gases adjusting to and flowing
from a high pressure to a low pressure line, or vice versa,
can be eliminated.
For heating systems, most MOCVD reactors are heated by
inductively coupling RF power to a graphite susceptor. This

19
is ideal because it is a non-contact method, it selectively
heats only the graphite, and it is easy to configure by
arranging a copper coil around the susceptor portion of the
reactor. The RF generator size required depends on susceptor
size, gas velocity, coupling efficiency, and reactor wall
cooling mechanism. Infrared heating from quartz-halogen lamps
has also been used but, as wall deposition increases, non-
uniform heating may occur. A resistance heater embedded into
the graphite is another option but deposition on electrical
feedthroughs complicates reactor cleaning. Most heating
systems use an embedded thermocouple feedback system to
control temperature. Optical pyrometers have also been used
but wall deposition can result in false readings.
After passing through the heated zone of the MOCVD
reactor, some of the toxic gas sources still remain uncracked
and undeposited. These gases have to be neutralized before
being discharged into the atmosphere. Hazardous gases such as
AsH3, PH3 and SiH4 are commonly used in the MOCVD of III-V
semiconductors and are difficult to neutralize or "scrub"
especially when they are used in combination. There are four
different types of scrubbing systems commercially available;
depending on the application one alone may be inadequate. The
four types are liquid scrubbers, thermal crackers, dry powder
scrubbers, and incinerators. Liquid based scrubbers are most
commonly used and work by bubbling the toxic gas through a
basic (pH > 10) solution of sodium hypochlorite and sodium

20
hydroxide diluted in water where neutralized salts and acids
are products. Some gases are nearly insoluble in water,
though. Thermal crackers basically operate by heating the
exhaust stream to approximately 950C to thermally decompose
toxic gases into less toxic compounds. Clogging and also
insufficient heat transfer at high flow rates are problems
with this technique. Dry scrubbers use powders such as
activated carbon or diatomaceous earth mixed with iron
chlorides to react with the toxic gases. This technique also
has problems associated with efficient gas solid contacting
and disposal of toxic corrosive powders. A final technique,
the incinerator or "burn box," operates in such a way that
gases are mixed with a fuel gas and oxygen and then ignited by
a pilot flame or electric igniter[29]. All of the four
scrubbing techniques have their individual problems. The
scrubbing system should ideally be a combination of two or
more of the individual systems in case one system fails.
As discussed, MOCVD of III-V compound semiconductors
presently involves the use of highly toxic and explosive
source gases. There has been some work on the use of less
toxic sources for MOCVD such as tertiarybutylphosphine (TBP)
[30] instead of phosphine (PH3), (the threshold limit value
(TLV) of PH3 is 0.3ppm while that of TBP is greater than
lOOOppm), but the material grown with these new sources is
generally inferior. In any event, there must always be an
integrated safety component to all MOCVD systems.

21
Several papers have been published on the important topic
of MOCVD safety[31,32]. These papers are very useful when
designing the layout of an MOCVD machine and laboratory.
First, all MOCVD systems must have both toxic gas and hydrogen
sensors in and around them connected to an alarm. These
sensors must be capable of shutting down the machine in the
event of a detected leak. In order to shut down quickly, all
gas lines must be equipped with normally closed air-operated
valves. Toxic gas lines should also be double contained
(which is extremely expensive) and equipped with pressure
sensors that sound an alarm for abnormally high pressures.
Compressed air, hydrogen and nitrogen lines should have
sensors for abnormally low pressure which can, if activated,
shut the machine down. The reactor and pump exhaust line
should also have similar pressure sensors. Of course, smoke,
fire, and cooling water flow detectors and sensors are
necessary. Other things such as micro-switches on panel doors
and available supplied air masks are also required. Finally,
SCBAs and trained users should always be available outside of
the facility for emergencies. All of the above mentioned
safety issues/design features are important for the design and
operation of a modern MOCVD system. Of course, the most
important thing for lab safety is to provide adequate operator
training focusing on the nature of the toxic sources and how
to treat them.

22
2.3 A Review of the Literature on InP Based MOCVD
2.3.1 InP Homoepitaxv
The first reported growth of indium phosphide (InP) by
MOCVD was in 1969 by Manasevit and Simpson[12]. This work as
well as other early efforts[33-35] used triethyl indium (TEI)
and phosphine (PH3) as indium and phosphorus sources. For
several years, problems such as low, nonuniform growth rates
and high impurity levels were encountered. One source of
these problems was found by relating the observation of a
white smoke at certain growth conditions to the uncontrolled
gas phase reaction between metal organic indium sources and
phosphine. This reaction occurs at low temperatures <100C,
is parasitic in nature, and produces a non-volatile liquid
polymer.
One method used to minimize this problem was the use of
low pressure reactor systems[36-38] to decrease the residence
time of unreacted species upstream of the growth region.
Another method involves the use of adducts such as TMI-TMP
[39] or TMIn-TEP[40] as indium sources which will not complex
with PH3. Another technique is to keep the reactants apart
and only let them mix just prior to the growth region. This
method can, however, lead to uniformity problems. The most
recent improvement is the use of TMIn (a solid powder at room
temperature which melts at 88C) instead of TEIn (a liquid at
room temperature), as the indium source[41-43]. TMIn also
decomposes at a much higher temperature (>300C) than TEIn

23
(<100C) and therefore is less likely to react upstream of the
heated growth zone. Also, TMIn has a much higher vapor
pressure than TEIn which is experimentally convenient because
heated gas lines would no longer be necessary.
Other techniques such as the use of hydrogen-nitrogen
mixtures as the carrier gas[41] and phosphine pre-crackers [44]
have been tried with varying degrees of success and merit. A
final conclusion is that proper reactor geometry, system
design, and growth conditions are very important for avoiding
parasitic gas phase reactions and obtaining superior InP thin
film quality. Currently, the reaction at certain growth
conditions of TMIn and PH3 in a properly designed MOCVD system
can yield uniform, high quality epitaxial InP with no evidence
of indium prereaction problems.
The deposition of high quality layers of InP for device
applications requires precise control of their unintentionally
introduced (undoped) and intentionally introduced (both p- and
n-type doped) impurity concentrations. An important condition
for obtaining reproducible p- and n-type doping levels is the
ability to grow undoped material with a reproducibly low
background carrier concentration. To obtain low background
levels, one needs a contamination free MOCVD system equipped
with high purity sources and optimized growth conditions.
Most conventional MOCVD systems are constructed from
ultra-high purity components such as electropolished welded
316 stainless steel and semiconductor-grade low-sodium content

24
quartzware. Coupled with the use of palladium-alloy diffused
hydrogen as a carrier gas, these precautions usually eliminate
the system as a source of high background impurity level
problems. In addition to high purity equipment, ultra high
purity sources contained in stainless steel bubblers and
corrosion resistant coated cylinders are required. For inP,
phosphine with five nines purity (99.999%) and diphos purified
(doubly sublimed) trimethylindium are both commercially
available. As purification technologies advance, then
progressively lower background doping levels surely will
follow.
The most important materials issues in the InP growth
area are the effect of growth conditions, substrate quality,
substrate orientation and substrate wafer cleaning techniques
on material quality. Several papers have been published on
each topic and the basis of comparison presented usually
involves characterization results of thin films such as room
temperature (300 K) and/or liquid nitrogen temperature (77 K)
mobilities and undoped carrier concentrations (NDNA) cm3, etch
pit densities (EPD), photoluminescence (PL) intensities and
occasionally device performances.
The effect of growth conditions on properties of InP
grown by MOCVD has been studied by several research teams[45-
48]. Razeghi and Duchemin[45] showed that the growth rate of
InP is linearly dependent on indium metal organic reactor
partial pressure and independent of the phosphine partial

25
pressure. They also compared the growth rate of undoped InP
on (100), (111), and (115) InP oriented substrates and they
reported excellent film quality on (100) 2 towards (110) and
(115) 2 towards (111). Eguchi et al.[46] studied the effect
of V/III (phosphine to metal organic indium) ratio on EPD and
electrical properties and reported superior material at high
V/III ratios (>300) This result was in agreement with
Kasemset's[47] earlier work on both V/III ratio and growth
temperature effects. However, a survey of the effect of
growth temperature on layer quality is less conclusive.
Kasemset[47] indicates that a decrease in background carrier
concentration results upon increasing growth temperature,
while Scott et al.[48] report the opposite trend. This
discrepancy is probably due to different dominant impurities
in each group's TMIn source with correspondingly different
incorporation mechanisms. Most teams report high quality
MOCVD InP grown at temperatures between 550c and 675C.
Below 550C, growth rates drop and material quality degrades.
Above 700C, background carrier concentrations increase.
Presently, high quality two inch diameter wafers of InP
are commercially available as both doped (p and n-type) and
semi-insulating. Variation of results from team to team in
early research efforts and even today may be due in part to
the lack of reproducibility of substrate properties from batch
to batch and vendor to vendor. A recent paper from Knight et
al.[49] reports this problem. They observed a correlation

26
between leakage current of p-i-n InP based photodiodes and
substrate quality. Consequently, they set up a nondestructive
PL wafer-mapping system to evaluate grown film quality before
investing further processing time. Non-destructive techniques
such as PL mapping will remain essential unless wafer quality
control improves.
Proper wafer cleaning is also very important for the
growth of high-quality InP. Tuck and Baker[50] in 1973
published work on the chemical etching of (111) and (100) InP.
They compared the merits and disadvantages of using the
following four etching solutions: (1) 1HCL:1HN03; (2)concen
trated HC1; (3) 0.4N Fe3+; and (4) 1% bromine in methanol,
based on etching rate and hillock delineation. Nishitani and
Kotani[51] presented the use of H202-H2S04-H20 solutions for
etching (100) and (111) oriented InP. Recently, studies have
been reported using sulphur to chemically passivate the
surfaces of InP and GaAs[52-53]. The goal of this work is to
reduce the substrate surface recombination velocity in order
to improve device performance. Another interesting study
compared several wafer cleaning methods using the surface
science techniques ISS, ESCA and AES[54]. This report states
that using a 5:1:1 mixture of H2S04:H202:H20 in combination
with solvent degreasing step yields an InP surface with the
least amount of absorbed carbon and oxygen relative to the
other methods tested. Most crystal growth teams develop their

27
own technique of wafer preparation using a combination of the
methods reviewed above.
Once an MOCVD system has been optimized for growing high
purity InP and a proper substrate vendor, orientation and
cleaning procedure have all been selected, most authors report
that the source purity of both the metal organic indium and
phosphine have the strongest influence on background carrier
concentrations and mobilities. The initial work on InP growth
[12,33-34] reported room temperature carrier concentrations of
n=0.17 to 1.4 x I016cm*3 and 300 K and 77 K electron mobilities
of 3500-4200 and 16,000-36,000cm2/volt-sec, respectively.
After two decades of technological advancement in purification
techniques and machine design, the highest reported 77 K
mobility for undoped InP is now 305,000cm2/volt-sec with a
corresponding carrier concentration of n= 5 x I013cm3[55].
Based on low temperature PL it appears the dominant residual
acceptor in MOCVD InP is zinc[55]. For many years both carbon
and manganese[47] have also been reported as compensating
acceptors and silicon has been reported as the dominant
donor[56]. The recent work of Bose et al.[55] caution against
identifying PL peaks as carbon since the transverse optical
phonon replicas of the free-exciton recombination occur at the
same energy as carbon. For most teams, however, 300 K and
77 K mobilities of 4,700 and 80,000cm2/volt-sec, respectively,
and a carrier concentration of n= 1 x 10ucm'3 are typical for

28
undoped MOCVD InP. So, it is evident that device quality
unintentionally doped InP can be grown by MOCVD.
Most semiconductor devices require a junction of some
type in the host material where two materials with either
different electrical or optical properties meet. An
electrical junction can be created by post growth processing
techniques such as ion-implantation or diffusion of a donor or
acceptor into the host crystal. Another way is to just create
the junction in-situ during the MOCVD growth by adding a small
quantity of a donor or acceptor source into the inlet gas
stream. High quality undoped InP is usually n-type with a
background carrier concentration of n = 1 x 10u- I015cm*3.
The carrier concentration n or (ND-NA) can be increased by
adding an InP donor species to the inlet gas stream of the
MOCVD reactor. InP can be doped n-type by using H2S[45],
H2Se[43] and SiH4[43] or Si2H6[57] as sources. For each
source, the free carrier concentration is essentially
proportional to the dopant source mole fraction in the reactor
inlet stream. Controllable n-type doping from I015cm'3 to
1020cm3 can be achieved without a significant decrease in
material quality by using a combination of these sources for
different parts of this wide incorporation range. Doping
levels and diffusion rates of these dopants are affected to
varying degrees by changes in growth conditions such as
temperature, V/III ratio, and indium mole fraction. For H2S,
the free carrier concentration in deposited InP layers

29
decreases when the growth temperature increases. For SiH4,
the opposite trend is observed, because the incorporation of
silane is reaction limited whereas H2S is adsorption limited.
Si is amphoteric in InP, acting as an acceptor or donor,
depending on site selection (adjusted by changes in the V/III
ratio used during the growth) H2S does not compensate itself
in InP but the diffusion coefficient of S is greater than that
of Si in InP. Depending on the device application, a suitable
n-type dopant source for InP is apparently available.
InP can also be doped p-type by adding an InP acceptor
species to the inlet gas stream of the reactor. The gas
stream must contain enough of an acceptor species to increase
the electrically active extrinsic acceptor level above the
electrically active intrinsic donor level. The metal organic
(MO) compounds DESn[45], DMZn[58], DMCd[59], Cp^gteO] and
MCp2Mg[61] act as sources for acceptors in InP with varying
degrees of success. There is not one single MO acceptor
source which dopes InP p-type over a wide doping range (1015-
I019cm'3) without extended diffusion or surface morphology
degradation. This is the reason that so many different
sources have been investigated for InP as suitable p-doping
sources. This dilemma, in combination with the observation
that the electrical activation of some p-type dopants in InP
is much less than unity, was a driving force for the extensive
literature review and model development for p-doping of MOCVD
InP in Chapter III of this dissertation. The most commonly

30
used p-dopant source is DEZn which can be used to dope InP
over the range p = I015-I018cm3. DEZn is sufficient for most
device applications, but its relatively high diffusion
coefficient at typical growth temperatures, D = 3 x 10'13
cm2/sec[58], does make it unsuitable for some device
applications. For a more extensive discussion on p-doping of
InP, the reader is referred to Chapter III of this
dissertation.
2.3.2. GalnAs/InP
The ternary compound Ga^n^As can be grown lattice-
matched to InP by MOCVD. Unlike the AlGaAs/GaAs material
system, the GalnAs/InP system is not lattice-matched for all
compositions. Ga 47In 53As has an energy gap of 0.75 eV (Xg =
1.67 tm) and is the only composition which is lattice-matched
to InP. This ternary film can be grown by carefully
controlling the gallium to indium metal organic source
composition of the gas inlet to an MOCVD reactor. This is one
of the most severe heteroepitaxial growth scenarios possible
as the group V sublattice must be changed from pure phosphorus
to pure arsenic.
The first reported MOCVD growth of GaxIn1.xAs was on a
GaAs substrate[61] and hence was lattice-mismatched. The
early efforts provided useful information such as
compositional uniformity, merits of methyl versus ethyl MO
sources, gas phase reactions and purity for later GalnAs/InP
work. One conclusion that was useful for GalnAs/InP work was

31
the improvement in material quality observed upon using TEGa
instead of TMGa as the gallium source. Another useful
observation was that the solid phase composition is controlled
by and almost equal to the gas phase ratio [TMIn]/([TMIn] +
[TEGa]). Finally, the GalnAs growth rate is proportional to
the sum of the metal organic gas phase concentrations.
Lattice-matched Ga 47In>53As grown on InP by low pressure
MOCVD using TEGa and TEIn was first reported by Hirtz et al.
[62] in 1980. As stated previously, the choice of group III
alkyl sources used for GalnAs growth is critical. Using TMGa
and TEIn results in poor surface morphology, whereas using
TEGa and TEIn results in nearly featureless material over the
composition range 0.4 < x < 0.6 (Gax) This phenomena has
been attributed to a TMGa-InP substrate steric hindrance to
the heterogeneous decomposition of TEIn[63]. When growing on
InP, the initial stage of growth of GalnAs is also complicated
by the incongruent evaporation of phosphorus from the InP
substrates upon heating. It has been shown that the
morphological, optical and electrical properties of the GalnAs
epitaxial layer depend heavily on minimizing InP substrate
damage during the transition from PH3 to AsH3[64]. The best
approach is using an InP buffer layer and then allowing the
indium flow to continue while rapidly switching phosphine to
the vent and TEGa and AsH3 to a low pressure reactor.
As is the case for growth on GaAs substrates, the
composition of GalnAs is linearly dependent on the flow rate

32
of TEGa for a fixed TEIn or TMIn flow. Both the composition
and growth rate are independent of AsH3 flow for fixed metal
organic flow. The growth rate is independent of growth
temperature (500-650C), but the composition can be slightly
affected due to slight differences in gallium and indium
source cracking efficiencies. If the composition of a layer
is different from the lattice-matched value, this layer is
mismatched. The lattice-mismatch between layer and substrate
is defined as Aa/a = (aL a)/ a when aL is the measured
room temperature (strained) lattice parameter and a is the
lattice parameter of the substrate. Razeghi et al.[38] have
reported that the mobility of a semiconducting layer is
dependent on the amount of mismatch in the layer relative to
the substrate. At optimum growth conditions, a lattice
matching of Aa/a < 0.04% has been achieved resulting in Hall
mobilities of 12,000 (300 K), 100,000 (77 K) and 260,000cm2/
volt-sec (2 K) and background carrier concentrations of 0.7 -
1.0 x I015cm'3[65]. Such high mobilities at 2 K are explained
by the existence of a two-dimensional electron gas formed at
the interface between undoped InP buffer and GalnAs layers and
are indicative of superior material quality.
It is evident that MOCVD can be used to produce undoped
Ga 47In 53As lattice-matched to InP, with very high quality
electronic properties. Intentionally doped both p- and n-type
Ga 47In 53As on InP is also producible. Razeghi[65] presented
data on p-type doping using DEZn and n-type doping using H2S

33
of GalnAs by MOCVD. Zinc doped GalnAs carrier concentrations
are reported to decrease with increasing growth temperatures
over the range p = 1017 I018cm*3. The opposite behavior is
observed for sulphur doped GalnAs and this trend is confirmed
by Logan et al.[66]; a carrier concentration of 102cm'3 is
reported for a growth temperature of 525C and at nearly
identical conditions except Tg = 625C, n = 8 x I017cm'3. Wide
ranges of both p- and n-type doping are attainable for GalnAs
on InP by MOCVD which is useful for device applications.
2.3.3 GalnAsP/InP
The quaternary solid solution GaxIn1.xAsyP1.y is an alloy
semiconductor which can be lattice-matched to InP and GaAs
substrates. GalnAsP is a direct bandgap semiconductor (when
lattice-matched to GaAs or InP) which can be a very efficient
light emitter over the wavelength range 0.65 0.87/xm
(lattice-matched to GaAs) and 0.92 1.65jttm (lattice-matched
to InP) Very little work has been reported on quaternary
growth on GaAs substrates[67,68] due to greater interest in
GalnAsP alloys lattice-matched to InP substrates for optical
fiber device applications. Optoelectronic devices operating
at 1.3 Mm or 1.55/im wavelength regions have immediate
commercial applications because light transmission through
silica fibers exhibits low loss at 1.3/xm and low dispersion at
1.55jum.
The MOCVD growth of GalnAsP alloys on InP substrates was
first reported using ethyl alkyls in a low pressure

34
system[63]. This same team later reported growth of nearly
the entire quaternary composition range lattice-matched to InP
and presented device test results of broad area and stripe
lasers fabricated using six different quaternary
compositions[69]. After these initial reports, numerous
quaternary related papers have been published on novel growth
techniques, relationship between gas phase growth conditions
to solid phase compositions, electrical and optical material
properties, and device applications. Each of these topics
will be reviewed in the following paragraphs.
Growth techniques used for depositing quaternary alloys
are direct extensions of techniques used for growing InP and
GalnAs. The growth rate is similarly proportional to the sum
of the partial pressure of TEGa and TMIn, and is independent
of the phosphorus and/or arsenic partial pressure. Similar to
the growth of GalnAs, the solid phase group III composition is
essentially equal to the gas phase metal organic composition
introduced to the MOCVD reactor. However, the behavior of
incorporation of group V elements is different and much more
difficult to control since both arsine and phosphine are
required and they do not incorporate with the same
probability. It is much more difficult to incorporate
phosphorus than arsenic at a fixed growth temperature since
the cracking temperature of PH3 is higher than AsH3. To
alleviate this problem, some workers have used precracked
PH3 [38,70] or in-situ adduct formation techniques[71] (which

35
is a more easily cracked species) both with varying degrees of
success. Most teams use the low pressure MOCVD technique
which makes PHj/AsHj ratios as high as 200 (which is required
for 1.0/xm wavelength quaternaries) more safely attainable.
The operating parameters inlet partial pressures, total
pressure, deposition temperature, and V/III ratio have an
effect on the growth rate and composition of deposited
quaternary materials. Because of this, numerous papers have
been written relating the gas phase growth conditions to solid
phase material quality and composition. Razeghi[65] has
published graphs on which are plotted the relationship between
growth conditions ratios and bandgap wavelengths for the full
range of quaternary materials which are lattice-matched to
InP. The three ratios are: (1) Rj = PH3 / (PH3+AsH3) ; (2) R3
= TEGa/(TEGa + TEIn) ? and (3) Rj/R3. Using these graphs,
which are only valid for a growth temperature of 650C and
total flow rate of 7 liters/min, one can estimate growth
conditions for any lattice-matched quaternary composition.
Similarly, Fujii et al.[72] and Sugou et al.[70] present
quaternary compositions as a function of (In/Ga) and (P/As)
ratios in the gas phase for fixed V/III ratios. Koukitu and
Seki[73] use a thermodynamic approach to compute the solid
composition as a function of input mole ratio for several
quaternary III-V alloy systems. They also compute the
equilibrium partial pressures of gaseous species over
GalnAsP/InP as a function of temperature, V/III ratio and

36
Ag(/m) Using one of the above mentioned techniques, good
estimates of optimum gas phase growth conditions for the full
range of solid phase quaternary compositions can be predicted.
Several papers have been written presenting optical and
electrical property data as a function of composition for
GajjIn^j^ASyP^y lattice-matched to InP. One important
realization is that for lattice-matching, y is related to x by
the following simple relation: y = 2.16 x. This greatly
facilitates presenting data as it can be plotted as a function
of y or x under the assumption that the lattice-matching
condition is realized. A paper by Nahory et al.[74] presents
useful experimental lattice constant and bandgap values as a
function of composition relative to lattice constant values
predicted using Vegard's Law. Vegard's Law states that for a
lattice-matched system, the lattice parameter of the
quaternary can be deduced from those of the constituent
binaries. This team also presents the empirical relation for
bandgap variation (Eg(eV)) with composition.
Eg(y) = 1.35 0.72y + 0.12^ (1)
Another group presented undoped electron and hole mobilities
as a function of composition, y. For undoped GalnAsP, room
temperature electron mobilities range from 4000 to
11,000cm2/volt-sec (y = 0 to y = 1) and room temperature hole
mobilities range from 130 to 200cm2/volt-sec (y=0 to y=l) [75].
Both p- and n-type doping of GalnAsP/InP grown by MOCVD
have been reported. Extensive doping studies as a function of

37
quaternary composition has not, however, been reported.
Saxena et al.[76] present data on 1.3jum p-GalnAsP doped by
DEZn over the range I018-I019cm*3,and n-GalnAsP doped by DETe
and H2S. Using tellurium, n-type doping from 3xl017 to
5xl019cm*3 is reported and, with sulphur a lower range 5xl016
to 3xl018cm*3 is reported. Meyer et al.[77] report n-doping
of 1.3 and 1.55im GalnAsP using H2S over the range 1016 to
I019cm*3. It is evident that p- and n-doping of GalnAsP is
possible over a wide range of doping levels which is
significant for device applications.
2.3.4 InP Based Devices
The wide range of compounds that can be grown with large
area uniformity by MOCVD make it suitable for fabrication of
long wavelength opto-electronic device structures. Several
different types of electronic, optical and opto-electronic
devices have been grown by MOCVD in the GaxIn1.xAsyP1.y on InP
material systems. Manasevit et al.[35] in 1978 showed that
solar cells in which an InP active region grown by MOCVD can
perform as well as ones fabricated by other techniques. Other
devices such as lasers, field effect transistors, photo
detectors and waveguides have been grown by MOCVD and will be
discussed in the following paragraphs.
Some of the first devices grown using MOCVD GalnAsP were
broad area and stripe double heterostructure (DH) lasers which
lased at wavelengths between 1.15 and 1.54/xm[69]. Distributed
feedback (DFB) lasers have also been successfully grown by

38
MOCVD. With MOCVD, it is possible to overgrow onto sub-micron
diffraction gratings which is vital to the operation of these
devices[79]. A typical laser structure composed of these
materials would start with a n+-InP substrate, followed by a
2/im thick n+ InP layer, then a 0.2frn active layer of
lattice-matched GalnAsP (undoped), then a 2/xm thick p-InP
layer ending with a p+ InP contact layer of 0.2/im thickness.
The lasing wavelength is determined by the composition of the
active layer. Surface emitting semiconductor diode lasers,
which emit light perpendicular to the grown layer surface have
also been fabricated in an array form using MOCVD grown
GalnAsP[80].
Microwave devices such as Gunn diodes and metal to
semiconductor field-effect transistors (MESFET's) have been
grown using MOCVD InP. For Gunn effect devices, even though
the mobility of InP is lower than GaAs, other characteristics
such as cut-off frequency, acceleration-deceleration time,
relaxation time and peak-to-valley ratio are better in InP.
These devices require a three-layer structure of n+-n-n+-InP
grown on n+-InP and have been successfully grown by MOCVD for
60 GHz[36], and 94 GHz[81] operation. MESFET's have also
successfully been fabricated by the use of undoped InP grown
by MOCVD on Fe-doped substrates. The electrical properties of
Au-InP Schottky diodes are reasonable and comparable to other
crystal growth techniques[82].

39
The use of the ternary material GalnAs lattice-matched to
InP for long-wavelength photodetectors is well established.
Traditionally, LPE and hydride VPE are used but, MOCVD grown
p-i-n photodiodes have also been prepared [ 65 ]. The most
important device characteristics required of detectors are low
capacitance, low dark field leakage current and high quantum
efficiency. To attain these goals, low background doping
levels, accurate lattice-matching and an abrupt p-n junction
are required and all of these are possible with MOCVD.
Actually, with MOCVD the need for a post-growth zinc diffusion
processing step can usually be eliminated since in-situ p-
doping is possible. Several teams have reported improvements
in p-i-n photodiode performance by adjusting MOCVD growth
conditions[76], layer structure[65], and Schottky barrier
height enhancement[83]. MOCVD grown structures with leakage
currents as low as 3 pA at -10V using a 100/im device diameter
have been fabricated on two inch diameter InP substrates[84].
In addition to the above structures, a number of other
optoelectronic devices have been fabricated using MOCVD grown
InP based materials. Two dimensional electron gas (2DEG) and
multiple quantum well structures have been grown making use of
the extremely high mobilities, (in excess of 180,000cm2/volt-
sec at 9.2 K) possible with these materials[85]. Guided wave
devices such as optical waveguides and phase modulators have
also been grown[86]. Finally, GalnAsP/InP interference
filters have recently been grown by MOCVD[87]. The theory

40
and results of low pressure MOCVD grown GalnAsP/InP and MBE
grown AlGaAs/GaAs electrically tuneable interference filters
are presented in Chapter IV of this dissertation. More
extensive reviews of the wide range of MOCVD grown opto
electronic devices using InP based materials are available in
the 1iterature[65,88].
2.4. A Description of the MOCVD System
2.4.1 Introduction
The experimental apparatus used for the growth of
epitaxial layers of Ga^n^jjASyP^y on InP substrates is a
commercial MOCVD system custom built for the University of
Florida by Nippon Sanso K.K. (Japan Oxygen Inc.). A
photograph of the front and a simplified schematic of the
Japan Oxygen MOCVD System are shown in Figures 2 and 3. The
complete operating procedures for performing epitaxial growths
and maintenance (e.g., such as reactor cleaning), are
presented in Appendix B of this text. The four basic parts of
the MOCVD system which are described in the following
paragraphs are: (1) the gas delivery system; (2) the reactor
and heating system; (3) the exhaust/scrubbing system; and (4)
the safety system. The gas delivery system, reactor, exhaust
and safety system are all integrated inside the MOCVD system
which is shown in the photograph in Figure 2. The heating
system is a separate unit (20 kW RF generator) as is the
scrubbing system which is located outside the building for
ease of maintenance reasons.

41
Figure 2: Photograph of the quaternary MOCVD system.

42
Figure 3: Quaternary MOCVD simplified flow diagram.

43
2.4.2.Gas Delivery System
The gas delivery system connects the sources to the
reactor and provides a method of transporting them in a
controlled fashion. Since impurity levels must be kept to a
minimum, all components of the gas delivery system are
constructed from electropolished 316L stainless steel and
connected with metal-gasket leak-tight couplings. Also for
improved purity, 0.2/xm particle filters are installed at all
gas inlet points. All lines were wrapped with electrothermal
heating tape and aluminum foil and are heated during standby
mode to 50C to help desorb any of the sources or impurities
adsorbed on the inner walls of the stainless steel tubing.
The flow of gases is controlled by a combination of
manual valves, needle valves, pneumatic valves, check valves,
electronic mass flow controllers and regulators. The range of
possible flow rates for each source and the carrier gases
(hydrogen and nitrogen) are given in Table 2. The house
nitrogen gas which is mainly used for purging the MOCVD system
before a reactor or source change, passes through a molecular
sieve cartridge (Matheson Model 451) before entering the
machine. The house hydrogen, which is the carrier gas in the
system, is purified by diffusing it through a heated (400C)
palladium-alloy membrane which is part of a 0-20 liter/min
hydrogen purifier system (Matheson, Series 8370V) that was
installed inside the Japan Oxygen machine.

44
Table 2
Flow Rate Ranges of Sources, Vendors and Purity
Source
Flow Rate Range
Vendor(Purity)
h2
0-20 SLM
Gator Oxygen
(Alloy Diffused)
N2
0-10 SLM
Linde(LN2 Boil-off)
AsH3
0-50 seem
Matheson (ULSI Grade)
ph3
0-200 seem
Solkatronic
(Micropure Grade)
lOOOppm H2S
(in H2)
0-50 seem
Matheson
(ULSI Grade)
TMIn
0-300 seem
Air Products
(Diphos Grade)
TEGa
0-100 seem
Akzo
(Electronic Grade)
DEZn
0-50 seem
Morton Thiokol
(Electronic Grade)

45
The gas delivery system for the metal organic sources
trimethylindium, triethylgallium and p-dopant diethylzinc
(which are held in stainless steel bubblers in temperature
controlled baths) consists of pneumatic, needle and manual
valves which are attached to the inlet and outlet ports of the
bubblers. Each metal organic line also has a pressure sensor
attached to it just before the inlet of the bubbler. The
metal organic sources are solids or liquids at room
temperature with fairly low, temperature dependent equilibrium
vapor pressures. By varying the temperature of the bubbler
bath, the hydrogen flow rate through the bubbler, and the
pressure of the bubbler region of the gas delivery system (by
opening or closing the needle valve), a controllable range of
metal organic source flow rates can be attained.
Since the hydrides (arsine, phosphine) and n-type dopant
(1000 ppm H2S diluted in H2) are highly toxic, combustible and
at high pressure, the gas delivery system for these sources is
slightly more complicated than for the metal organics. Each
line has a regulator, air operated valve and a manifold
attached to it. The manifold contains a high-pressure, high-
purity nitrogen purge line, a hydrogen purge line, and a third
line which can be used to evacuate the hydride line or vent
the hydride source directly to the scrubber. During normal
"standby" operation, palladium-alloy diffused hydrogen is
purging the hydride line to the vent. Only during toxic gas
flow is the hydrogen purge interrupted. Check valves on the

46
hydrogen and nitrogen lines prevent back flow of toxic gas
through the manifold to the rest of the system.
The "heart" of the gas delivery system is the fast
switching "vent/run" manifold mounted just prior to the inlet
of the reactor. The status of the "vent/run" valves (opening
and closing) and the timing of this sequence can be manually
or automatically controlled. Automatic control is made
possible by using a process sequencer which is capable of
storing 150 valve patterns and the corresponding times for
each pattern or layer. The MOCVD manifold is also "pressure
balanced" which means that it is possible to adjust the
pressure difference between the two vent lines and the two
reactor lines to almost zero. This is accomplished by
carefully adjusting tube lengths and by installing a dead
volume in the vent line to counter balance the volume of the
reactor. Pressure balancing is especially important for MOCVD
growth of superlattice or multiple quantum well structures.
To keep the pressure difference between the reactor and vent
lines equal to zero (which is measured using the two
differential pressure indicators) each main source line has a
hydrogen compensation line. It is necessary for the MOCVD
growth of InP and related materials to keep the metal organics
separated from the hydrides to prevent parasitic gas phase
prereactions from occurring, hence the reason for the two
vent/two reactor line design.

47
2.4.3 Reactor and Heating System
The Japan Oxygen MOCVD system has a horizontal 4 inch
I.D. reactor which is equipped with a water cooling jacket
(see Figure 4) to minimize side wall deposition. At the inlet
of the high purity quartz reactor, two inlet lines exist to
keep the group III metal organics separated from the group V
hydrides at low temperatures. The dilute group III sources
enter the reactor through a 6 mm O.D. high purity quartz tube
10cm downstream from the inlet of the group V sources. These
two gas streams should ideally be well mixed and flowing under
fully developed laminar flow conditions before the mixture
encounters the high purity quartz deflector/silicon carbide
coated graphite susceptor heat source. The deflector -
susceptor unit is tapered at an angle of 17 with respect to
horizontal to improve growth rate uniformity.
The graphite susceptor is heated by radio frequency
inductive heating. A Lepel series T-15-3-KC-TL 15 kilowatt RF
generator is used to generate radio waves over the frequency
range 80-900 KHz. Some of these radio waves are picked up by
a 3/8" copper coil that is wrapped around the reactor. A
platinum/rhodium "R-type" thermocouple sealed in a high purity
quartz tube is embedded inside the graphite wedge. This
thermocouple is connected to a West series 2070 microprocessor
based temperature controller which was installed on the MOCVD
system and sends DC current to the RF generator to control

Fork
Mechanism
H2 and Hydride
Gas Inlet
Sample Quartz
Position Wafer Tray
Figure 4; MOCVD reactor and loading mechanism.

49
its power output. Automatic temperature control from room
temperature to growth temperatures (550-700C) to bakeout
temperatures (900-950C) is possible with this elaborate
heating system.
As shown in Figure 4, a high-purity quartz sample tray
sits on top of part of the graphite susceptor. This tray
holds substrates as large as 1/4 of a two inch diameter wafer.
The tray and substrate are placed onto and removed from the
susceptor by an electro-mechanical fork which is capable of
precise x-y-z motion. The fork can move 90 cm horizontally
from the load lock area (where the sample and tray are loaded
and unloaded) through a gate valve and shutter valve, by the
gas exhaust port to directly above the susceptor. Precise
mechanical sample loading makes it possible to use the reactor
even after side wall deposition has obstructed view of the
susceptor.
To prevent the introduction of "dirty" room air to the
reactor during each loading of a substrate, the air in the
load-lock is evacuated by a rotary pump (Edwards model E2M2)
to a roughing pressure of 10'2 torr. Then a turbomolecular
pump (Balzers model TPH050) is turned on and evacuates the
load-lock to a pressure of 10'7 torr. At this point the load
lock is isolated and backfilled with the ultra high purity
hydrogen. The sample is now ready for loading and the ultra
high vacuum gate valve (VAT Ltd. model MSS4) is opened.

50
2.4.4 Exhaust/Scrubbina System
The waste products from the reactor flow through an
exhaust port in the water cooled flange which is bolted to the
exit of the reactor to a particle filter (Fuji Ltd.) which is
made out of glass fibers. The gas can then either flow at
atmospheric pressure directly to the scrubber or it can be
evacuated from the system by a rotary pump (Edwards model E2M
18) and then continue onto the scrubber. The rotary pump
makes it possible to grow films at low pressures (0.05 to 0.2
atm) which generally improves thickness uniformity.
The scrubber which was added to the Japan Oxygen MOCVD
system was built by Advanced Concepts (model 9625). It is a
liquid based scrubber which is designed to scrub toxic gases
and exhaust clean gas to atmosphere. For safety, the exhaust
was connected by a fireproof duct to the building's room air
scrubber. The scrubbing solution consists of a 80:5:2 by
volume mixture of water, 15% sodium hypochlorite, and 50%
sodium hydroxide. The pH and oxidation reduction potential
(ORP) of the scrubbing solution are constantly monitored to
evaluate the solution's scrubbing potential (pH >10.0 and ORP
> 200 mV). The scrubber has an efficient gas-liquid venturi
contactor and a packed bed which can handle higher than
required toxic gas flow rates. For arsine and phosphine the
overall chemical neutralization reactions are:
AsH3(g)+3Na0Cl(l)+H20(l) = H3As04(l)+3NaCl(s)+H2(g) (2)
PH3 (g) +3NaOCl (1) +H20 (1) = H3P04(l)+3NaCl(s)+H2(g) (3)

51
The products of these reactions are less hazardous and for
added safety, when the system requires draining, the solution
flows to the building's neutralization pit.
2.4.5 Safety
Arsine, phosphine and hydrogen sulfide are highly toxic
combustible gases, therefore, several safety features are
installed in the MOCVD system, and several safety practices
must be followed in the facility. First of all, the facility
has a rule that at least two competent people must be present
in order to do any work in the building. The facility also
has an alarm system and card reader to deny access to
unauthorized personnel. The clean room has an eight point
toxic gas monitor (MDA, Inc.) which has a resolution of 1 ppb
for all hydrides. In the unlikely event of a toxic gas
detection anywhere in the facility, the MOCVD system will
completely shut down (all air-operated valves are normally
closed). The MOCVD system also has a four point hydrogen gas
detector (Matheson) connected to it so that if hydrogen levels
exceed 50 ppm, the machine will shut down. Also the facility
has a helium leak detector (Varian) which is used to find
actual leak points before the MOCVD system is used and to
check connections after valve replacements or reactor changes.
Not only is the facility well equipped with safety
features and practices, but the Japan Oxygen MOCVD System
itself has several integrated safety systems. There are two
types of alarms, facility failure and machine failure. If

52
either power fails, compressed air pressure drops, cooling
water pressure drops or temperature increases, the machine
will automatically alarm and all air operated valves will
close. There are smoke detectors in the machine and fire
detectors in the room. There are pressure sensors on the
hydride lines, in the reactor, and on the exhaust line and for
each, if a certain pressure value is exceeded, an alarm will
sound and the machine will shut down. There is also a pH, ORP
and temperature sensor on the scrubber which triggers an alarm
in the clean room if any of these values are out of the safety
range. Finally, there is a compressed breathable air supply
always on hand for reactor or source changes and two SCBAs
available for the emergency response team. It is evident that
safety is a big concern and since the machine was constructed
in Japan, there is even an earthquake sensor attached to it.
2.5 Determination of Optimum Growth Conditions Based on Thin
Film Characterization
2.5.1 Experimental Method
There are several experimental parameters that must be
determined before performing a MOCVD growth. Using the
simplest case as an example, undoped InP on InP, the first
thing that must be decided is what type of substrate is
required. For all of the GaxIn1_xAsyP1_y on InP experiments
performed in the Japan Oxygen MOCVD system, InP oriented (100)
2 towards (110) purchased from Sumitomo Inc. were used. Both
semi-insulating (iron doped) and n-type (n 8xl018cm'3,
sulfur

53
doped) InP were used together or by themselves depending on
the purpose of the experiment. Next, the proper substrate
cleaning procedure must be decided. Also, experimental growth
conditions such as: growth temperature, growth pressure,
total hydrogen flow rate, trimechylindium mole fraction,
phosphine mole fraction and finally the length of time of the
planned deposition, must be determined. From the literature,
estimates of optimum growth conditions and proper wafer
cleaning procedures can be obtained, but these values or
procedures are system specific and hence had to be
experimentally optimized for the Japan Oxygen MOCVD system at
the University of Florida.
The procedure for calculating growth conditions for
undoped InP on InP by MOCVD is fairly straight forward. The
growth pressure, temperature and time must be chosen based on
knowledge of the material and capabilities of the system. It
should be noted that for convenience, some simplifications
were made in the derivation of the equations used to calculate
the growth conditions. Specifically, the ideal gas law was
used and it is assumed that the total flow rate to the reactor
is equal to the hydrogen flow rate and the bubbler pressure
was much greater than any metal organic vapor pressure. The
TMIn and PH3 flow rates are commonly expressed as TMIn mole
fraction (MFTHIn) and PH3 mole fraction (MFPH3) or their ratio,
the V/III ratio, which in this case is defined as:

54
where:
V
III
MF,
ph3
MF
THIn
MFPh3
(4)
(5)
MF
THIn
FH2,TMIn
VP-
THIn
(T)
FH2 Pb,TMIn
(6)
and Fph3 is the total pure phosphine flow rate (cm3/min) t fh2
is the total hydrogen flow rate (cm3/min), FH2 THIn is the
hydrogen flow rate through the TMIn bubbler (cm3/min) ,
VPTMIn(T) (nun Hg) is the bubbler temperature dependent TMIn
vapor pressure (torr), and PbTMIn is total pressure at the
TMIn bubbler (torr). All volumetric flow rates are measured
at standard conditions (300 K and 760 torr). The temperature
dependent vapor pressure equations for the metal organics
installed in the MOCVD system which were supplied by their
vendors (see Table 2) are shown in Table 3. It is therefore
possible with the use of the above equations to determine
conditions for a MOCVD experiment. Of course some initial
values must be known and others can be based on literature
values.

55
Table 3
Metal Organic Vapor Pressure Equations
Source
Vapor Pressure Equation
TMIn
log10P(mm Hg) =10.52 *014
TEGa
log10P(mm Hg) = 8.224 2222
DEZn
log10P(mm Hg) = 8.28 2190

56
2.5.2 InP
The method for determining the optimum growth conditions
for undoped InP on InP, n-type InP on InP and p-type InP on
InP will be discussed in the following sections. Before
presenting the results of this operating parameter sensitivity
and optimization study, the criteria for good epitaxial layers
are: mirror-like low defect density surfaces, uniform growth
rates over a large area substrate, less than n = lxl015cm'3
background (ND-NA) doping levels for undoped InP with
corresponding high room temperature electron mobilities (ju >
3000cm2/volt-sec), experimentally convenient growth rates of
1-2/xm/hour, and for doped layers and heterostructures,
atomically abrupt junctions and interfaces, respectively. All
of the above criteria have been determined by using various
thin film characterization techniques available on the campus
of the University of Florida.
2.5.2a Undooed InP
An extensive optimization study has been performed on the
MOCVD growth of undoped InP on InP. The results of this study
have been applied to aid in the determination of optimum
growth conditions for n- and p-type InP, undoped, n- and p-
type GalnAs, and undoped, n- and p-type GalnAsP on InP; these
optima will be discussed later. The timing and sequence of
events for a typical MOCVD growth are presented in the
Appendix B of this text.

57
After a growth has been performed, the first property
studied is the surface morphology of the thin film viewed
under an optical microscope. The thickness of the film can
also be determined by cleaving a sample and etching it for two
minutes in a 6:4:50 solution (measured by weight) of potassium
hydroxide, potassium ferricyanide, and deionized water to
delineate the grown layer from the substrate and then viewing
it under a microscope. A Nikon Optiphot microscope equipped
with Nomarski phase contrasting was used to view the surface
and delineated layer of the third growth in the machine, Q003.
Polaroid photos of the images observed at 2000X magnification
are shown in Figure 5. As shown, the surface morphology has
a ripple in it and this is sometimes referred to as an "orange
peel" surface. The side view photo of the sample was taken at
the leading edge of the sample with respect to the gas flow
direction. As shown, the layer thickness is clearly greater
(1.5/Ltm) near the corner compared to the center (1.0/xm)
probably due to the lower growth rate right on the corner
which is at a different crystalline orientation, the (111)
orientation.
Several experimental parameters have been varied in an
attempt to improve the surface morphology of undoped InP grown
on InP. The growth temperature was varied from 450 to 750C
and it was found that layers grown at temperatures between 575
and 700C had better morphologies than those grown at higher
or lower temperatures. At lower temperatures lack of

58
Figure 5: Nomarski photographs of an InP plane view surface
and stained edge cross section taken at 2000X
magnification.

59
adequate phosphine decomposition and at higher temperatures,
slightly higher growth rates were the causes for inferior
material quality. The total hydrogen flow rate and growth
pressure were also varied but had little to no effect on
surface morphology. The TMIn mole fraction was varied and at
values greater than lxlO'4 resulting material quality degraded
probably due to increased growth rates. The V/III ratio had
the most pronounced effect on layer surface morphology. This
effect can be clearly seen in Figure 6 which shows two inP
surfaces grown at identical conditions, except that the V/III
ratio was almost tripled by increasing the phosphine flow rate
to the reactor from 35cm3/min to 100cm3/min. This is in
agreement with the growth temperature study which concluded
that poor material results from insufficient phosphine
decomposition.
Another study performed with the goal of improving the
surface morphology of undoped InP on InP was varying the wafer
preparation procedure employed. Substrates were prepared
using five different techniques and then viewed under the
microscope. The procedures for each are as follows: (1)
filtered nitrogen blow off, five minutes each of warm acetone,
warm propanol, warm methanol, and then filtered nitrogen
blowoff; (2) procedure (1) followed by a 1 minute etch in 20%
nitric acid in methanol, methanol rinse, DI rinse, five minute
etch in room temperature 5:1:1 (sulfuric acid: hydrogen
peroxide: DI water), DI rinse, methanol rinse, filtered

60
Figure 6; Nomarski photographs showing the effect
ratio on InP surface morphology (top:
bottom: V/III=141).
of V/III
V/III=50,

61
nitrogen blowoff; (3) procedure (2) without the nitric acid in
methanol etch; (4) procedure (2) followed by a 1% bromine in
methanol etch for 6 minutes, methanol rinse, filtered nitrogen
blow dry; and (5) 2 minutes of surface treatment with the
UV/ozone cleaning system (UVOCS Inc).
Under a microscope at 200X magnification, particles were
observed on the surfaces prepared using procedures (1) and
(2) Procedure (4) resulted in a wavy surface probably due to
uneven bromine etching. Samples prepared using procedures (3)
and (5) had the best surfaces. Fresh substrates were cleaned
using procedures (3) and (5), loaded into the reactor, and
undoped InP was grown on both of them at the same time. This
experiment was performed several times and layers grown using
procedure (3) were consistently equal to or better than layers
grown using procedure (5). Procedure (3) was chosen as the
optimum and the exact details of this procedure are given in
Table 4.
The effect of growth conditions on the uniformity of
undoped epitaxial InP on InP was also studied. The total
volumetric flow rate was the only parameter that had any
significant effect on layer uniformity. Layers grown using
total hydrogen flow rates of 3, 5, 7 and 9 standard liters per
minute (SLM) had corresponding thickness variations of +/-10,
8, 6 and 6%. The increased variation at the lower flow rates
probably is associated with an increased boundary layer

62
Table 4
Optimum InP(lOO) Wafer Preparation Procedure
1. Cleave wafer and blow off with filtered N2.
2. Degrease teflon beakers and tweezers with warm methanol.
3. Place substrates into beaker with warm acetone for 5
minutes.
4. Place in warm propanol for 5 minutes.
5. Place in warm methanol for 5 minutes.
6. Rinse in running DI water for 1 minute.
7. Etch in a 5:1:1 solution of H2S0A:H202:DI at RT for 5 min.
8. Rinse in running DI for 1 minute.
9. Place in room temperature methanol for 1 minute.
10. Blow off with filtered nitrogen.
11. Load into reactor on wafer tray from oven.
12. Anneal wafer at growth temperature for ten minutes with H2
and PH3 flowing.

63
thicknesses resulting in depletion of TMIn near the growing
surface. This theory is confirmed by the observation that the
growth rate was lower, 0.95 and 1.2/m/hour for 3SLM and 5SLM,
than it was at higher flow rates (1.4/m/hour for 7SLM and
9SLM). The uniformity of undoped InP grown using a total
hydrogen flow rate of 7SLM is presented in Figure 7. A total
flow rate of 7SLM for hydrogen was chosen as optimum over 9SLM
because with 9SLM at a fixed TMIn mole fraction almost 3 0%
more TMIn material would be required.
The effect of several parameters on InP growth rate and
interface quality was also studied. As already stated the
total hydrogen flow rate has an effect on the growth rate.
The growth pressure and V/III ratio appear, however, to have
little to no effect. The growth temperature has a nonlinear
effect on the growth rate? at 450C the growth rate was
0.8/m/hour, but going from 550 to 750C the growth rate only
changed from 1.4 to 1.5/m/hour. This can be explained as
follows: at low temperatures (below 550C) InP growth is
kinetically limited and consequently temperature dependent.
At substrate temperatures above 550C, InP growth is Tmln
transport limited. This is confirmed by Figure 8 which shows
the linear relationship between InP growth rate and TMIn mole
fraction grown at 600C. Some of the data for Figure 8 was
taken from the SEM micrograph shown in Figure 9. Figure 9 is
a scanning electron micrograph of a cross section from growth

64
E
3.
W
(O
Ui
z
*
o
X
I
O 5 10 15 20 25
DISTANCE FROM LEADING EDGE (mm)
E
(0
(0
UI
z
*

X
H
-20 -10 0 10 20
DISTANCE FROM CENTER (mm)
Figure 7: MOCVD thickness uniformity study. Top: axial
variation, bottom: radial variation at 12 mm from
the leading edge.

Growth Rate (pin/hr)
65
TMIn Mole Fraction
Figure 8: The effect of TMIn mole fraction on InP
rate.
growth

66
Figure 9: SEM micrograph of growth Q054 at 20,000X.
layers are InP, dark are GalnAs.
Light

67
Q054 which shows InP layers grown with different MFTHIn for
different lengths of time separated by thinner dark GalnAs
marker layers. Figure 9 also shows the excellent abruptness
of interfaces between two different materials grown at 600C
and 7SLM.
Another analytical technique used to evaluate material
quality is an electrochemical C-V profiler. The C-V profiler
(BioRad, Model PN4200) is thoroughly explained by Blood[89].
It uses a Schottky and ohmic contacts in an electrochemical
cell filled with 0.1M HC1 (for InP) to slowly etch away
material while it measures the change in capacitance as a
function of changing applied voltage. From this C-V data, the
net carrier concentration can be calculated (n = ND-NA or p =
Na-Nd, where NA and N0 are the number of acceptors and donors
per cm3) and plotted as a function of depth. A C-V profile of
growth Q056 which is undoped InP on n-type (8xl018cm'3) InP is
shown in Figure 10. From this figure one can get a rough
estimate of 1.2/im for the layer thickness. Also, the
epilayer-substrate interface abruptness can be assessed. Good
uniformity of the carrier concentration throughout the layer
can also be observed. Finally an average value of the
background carrier concentration for undoped InP of
n=3xl015cm'3 can be obtained. It is evident that a lot of
useful information can be obtained from an electrochemical C-V
profile, the only drawback to this technique is that it is
destructive.

68
Figure 10: C-V profile of growth Q056 (undoped InP on n+InP).

69
The effect of growth conditions on the background carrier
concentration of undoped InP has also been studied. The
growth pressure, total hydrogen flow rate, and TMIn mole
fraction apparently have little effect on the background
carrier concentration. The growth temperature and V/III
ratio, however, both have a strong effect. The background
carrier concentration measured by the C-V profiler with
respect to growth temperature is plotted in Figure 11. At a
low growth temperature, high carrier concentrations may be
associated with phosphorus vacancies in the InP crystal
(impurities are complexing with these vacancies). At higher
temperatures, electrically active source impurities are
incorporating more efficiently. It appears that based on
background concentrations, the optimum growth temperature
should be between 575C and 625C for undoped InP. The
undoped InP carrier concentration is also strongly affected by
the V/III ratio as shown in Figure 12. Based on this figure,
high purity material can be grown with high PH3 flow rates.
Another analytical technique which is commonly used to
evaluate material quality is PL. The technique is explained
well by Dean[90]. Briefly, it typically uses laser light
energy incident upon a sample to stimulate the emission of
photons at discreet energy levels. These discreet energy
levels can be detected and related to impurities, defects, or
host crystalline energy level transitions. The resolution of
closely spaced transitions can be improved by reducing the

70
Figure 11: The effect of growth temperature, Tg, on undoped
(n-type) InP background carrier concentration.

71
Figure 12: The effect of V/III ratio on C-V background carrier
concentration in undoped (n-type) InP.

72
temperature of the sample. The PL spectra of two different
InP samples grown at different V/III ratios and measured at
4.2 K are shown in Figures 13 and 14. The peaks labelled
(D,x) and (FE) are host crystalline donor to bound exciton
and free exciton transitions. The other peaks maybe due to
impurities such as carbon or crystalline defects like
phosphorus vacancies (Vp) Based on the reduction in area
under the peaks labelled (Vp) it is evident that increasing
the V/III ratio or phosphine flow rate definitely has a
positive effect of the material quality of undoped InP.
In addition to the C-V profiler, the Hall effect can also
be used to determine electrical properties of semiconductors.
A low temperature (room temperature to 7 K) Hall effect system
was set up at the University of Florida by this researcher.
The theory, and a manual explaining the use of the Hall effect
system is given in the Appendix A of this text. Using data
from the Hall effect system, an average value of the mobility,
resistivity and carrier concentration can be calculated if the
thickness of the measured film is known. These values are
usually measured at 300 and 77 K (liquid nitrogen temperature)
and sometimes plotted as a function of reciprocal temperature
to 4.2 K. The liquid nitrogen temperature mobility (ii-m) is
widely used as a measure of purity of undoped semiconducting
films. In Figure 15, is plotted as a function of growth
temperatures for undoped InP. The highest mobility for InP

RELATNE INTENSITY
73
Figure 13: PL spectrum of an InP sample grown with a V/III
ratio of 140 measured at 4.2 K.

REIATNE INTENSITY
74
Figure 14: PL spectrum of an InP sample grown with a V/III
ratio of 219 measured at 4.2 K.

77K Mobility (ca2/Volt Sec.
75
70000
60000 -
X
50000 -
X
40000 -
30000 -
20000 +"
500
4 '
X

X
X
X
I 1 r-
550 600 650
Tg (c)
700
Figure 15: The effect of growth temperature on 77 K mobility
of undoped InP (bars indicate the range of data).

76
material from in the Japan Oxygen MOCVD was measured to be
61,800cm2/volt-sec for a sample grown at a temperature of
600C. There is a lot of scatter in the mobility data. The
scatter may be due to the fact that different material sources
with different impurity levels were used to grow these layers.
Also, the reactor was not changed/baked-out after each growth.
However, the average mobility values indicate that higher
quality material can be grown at 600C and this is in
agreement with previously discussed characterization results.
Based on all of the observed characterization trends,
apparent optimum MOCVD growth conditions for undoped InP on
InP have been determined. These growth conditions are
presented in Table 5 along with the room temperature and 77 K
electrical properties of undoped InP grown at these
conditions. These optimum conditions are significant because
they are the basis for estimates of growth conditions for
intentionally doped InP, unintentionally and intentionally
doped GalnAs and GalnAsP, both lattice-matched to InP.
2.5.2b Growth of n-tvne InP Using H:S
The Japan Oxygen MOCVD system was designed with the plan
of using dilute H2S as an n-type doping source. A cylinder
containing a mixture of lOOOppm H2S and the balance ultra high
purity hydrogen (Matheson, ULSI grade) is connected to the
MOCVD system. The flow rate range measured at room
temperature and 760 torr for this mixture, as shown in Table
2, is 0-50cm3/min (seem). Epitaxial layers of InP were grown

77
Table 5
Undoped InP Growth Conditions and Electrical Properties
Total H2 Flow Rate
= 7 SLM
Growth Temperature
= 600C
TMIn Mole Fraction
-41

o
H


O
II
V/III Ratio
= 419
Growth Pressure
=80 Torr
at 300 K :
nd na
= 2*1014 cm'3 (lowest)
nd na
= 2 1015 cm'3 (average)
^300K
= 3461 cm2/volt-sec(highest)
^300K
= 2800 cm2/volt-sec(average)
at 77 K :
nd na
= iio14 cm'3 (lowest)
nd na
= 2 1015 cm'3 (average)
^77K
= 61800 cm2/volt-sec (highest)
^77K
= 40500 cm2/volt-sec (average)

78
with H2S mixture flow rates of 2 to 50 seem, which corresponds
to gas phase mole fractions of H2S of 2.8xl0'7 and 7.16xl0'6,
respectively. A C-V profile of a S-doped InP grown layer is
shown in Figure 16. As one can see, a roughly 1.1/xm thick
layer of 2.5xl018cm*3 n-type material was deposited on a n+-InP
substrate, and this was achieved using a H2S mole fraction of
2.43xl0'6. The "hump" in the profile in Figure 16 is a C-V
profiler error which occurs at interfaces. The relationship
between the measured C-V carrier concentration and H2S mixture
flow rate for several n-type samples is presented in Figure
17. As shown, a wide linear incorporation rate of sulfur in
InP is possible resulting in doping levels from 5xl017 to
2.5xl019cm'3. The surface morphology of the grown material
when viewed under the Nikon microscope at 2000x appeared
unaffected by the presence of the sulfur atoms even at the
highest n-type doping level. Room temperature Hall effect
measurements were also performed on the sulfur doped samples.
Hall carrier concentrations agreed with C-V measurements and
Hall mobilities (/i300K) ranged from 498cm2/volt-sec at the
lowest doping level to 1064cm2/volt-sec at the highest level.
2.5.2c Growth of p-tvoe InP Using DEZn
P-type conversion of MOCVD InP was achieved by mixing the
metal organic source diethylzinc (DEZn) with the standard gas
mixture used to grow undoped InP. In order to get a wide
range of p-type doping, the temperature of the DEZn bubbler
was varied from -20 to 20C resulting in a change in the

79
X (um)
Figure 16: C-V profile of a H2S doped InP film.

Electron Cone, (cm-3)
80
Partial Pressure (Torr)
Figure 17; The effect of H2S partial pressure on InP carrier
concentration.

81
vapor pressure of DEZn in the bubbler from 1.39 torr to 14.00
torr. The hydrogen flow rate through the bubbler ranged from
5 to 50 seem and by carefully adjusting the opening of the
DEZn needle valve it was possible to keep the bubbler at a
pressure of 500 torr.
Characterization of the p-type InP material included
thickness and surface morphology measurements using the
optical microscope, C-V profiles, Hall measurements and
secondary in mass spectroscopy (SIMS). The SIMS technique,
which is explained elsewhere[91], was used to determine the
total atomic zinc concentration incorporated into several p-
type InP layers as a function of depth. This total zinc level
can be compared to the carrier concentration to determine the
percentage of electrically active zinc atoms. Also, based on
the depth of the atomic zinc profile, the extent of zinc
diffusion can be assessed relative to the epitaxial layer
thickness measurements. A C-V profile of growth Q080 which
was deposited at Tg=600C, Pg=80 torr and V/III=50.0 is shown
in Figure 18. During the growth, a DEZn mole fraction of
4.06xl0'5 was used which resulted in a C-V measured hole
concentration of 1.6xl018cm'3, for the layer grown on the n+-
InP substrate. A room temperature Hall effect measured hole
concentration of 2.5xl018cm*3 was calculated for the layer
grown on the semi-insulating substrate. The room temperature
Hall mobility and resistivity for this p-type sample were
61.3cm2/volt-sec and 0.0402 ohm-cm, respectively.

82
co
i
O)
o
20
19
ie
17
n-Type
0 12 3
X (um)
Figure 18: A C-V profile of a DEZn doped InP film.

83
The relationship between the C-V measured hole concentration
and the corresponding DEZn partial pressure used for the
growth is plotted in Figure 19. As indicated, controllable p-
type doping from 2xl017cm*3 to 3xl018cm*3 was realized. Hall
measurements were attempted on all of the p-type layers grown
on semi-insulating InP substrates, but the alloyed indium
contacts were generally not ohmic in nature. Hence, only the
C-V carrier concentration results from films grown on n-type
substrates were used. The room temperature Hall carrier
concentrations of semi-insulating samples with ohmic contacts
agreed quite well with the C-V carrier concentrations on n-
type substrates.
The effect of several growth conditions on the p-type
doping of MOCVD InP was investigated. The following growth
parameters: DEZn mole fraction, growth pressure, V/III ratio,
growth temperature, and total hydrogen flow rate, were varied
during individual experiments. It was later discovered that
the effect of the extended diffusion of atomic zinc in InP
during growth influenced the spatial variation of zinc's
incorporation. Basically, any variation in incorporation that
might have existed due to variations in growth conditions
during the deposition, were negated by the rapid diffusion
rate of zinc. For example, Figure 20 shows a C-V profile of
growth Q105 during which the DEZn mole fraction was varied
from 0.2 to l.OxlO'5 and the C-V hole concentration of

84
Figure 19: The effect of DEZn partial pressure on InP hole
concentration.

85
X (um)
Figure 20: C-V profile of InP:Zn grown with different DEZn
partial pressures.

86
9xl017cm'3 as shown, is essentially constant throughout the
grown film. The large "dip" in the profile shown in Figure 20
is commonly observed at p-n electrical interfaces. A SIMS
measurement was also performed on this sample (see Figure 21)
and the atomic zinc concentration profile yielded essentially
the same result. Similar results were observed when comparing
the SIMS and C-V profiles of samples grown to study the effect
of the other system parameters on zinc incorporation.
One interesting and useful conclusion that can be derived
from the SIMS and C-V profiles of growth Q095, during which
the growth pressure was varied from 38 to 760 torr, is the
extent of zinc diffusion into the InP substrate. As shown in
Figure 22, the grown layer thickness of this sample is about
1.4/Ln and the zinc doping level is unaffected by the change in
growth pressure. In Figure 23, the atomic zinc level (from
SIMS) is essentially constant throughout the profile, but a
"spike" occurs in the zinc profile right at a depth of 1.4/xm.
The SIMS operators at BNR, where the data was taken, say that
this "spike" is due to silicon-zinc complex which results from
Si on the surface of the InP substrate wafer after cleaning
and it reproducibly indicates the position of the grown layer-
substrate interface. These "spikes" were observed in several
of the SIMS profiles done at BNR. Based on the depth of the
zinc profile in Figure 23, zinc has diffused approximately 0.5
microns into the InP substrate. Using this diffusion length,

CONCENTRATION (atoma/cc)
87
PROCESSED DATA SNR
5 Aug SB Ca FILE: 0108
DEPTH (nlcponi)
Figure 21: Atomic zinc profile of sample Q105 measured by
SIMS.

88
Figure 22: C-V profile of growth Q095 during which the growth
pressure was varied.

CONCENTRATION (ntomo/cc)
89
PROCESSED DATA SNR
S Aug SB Ca FILE: aose
Figure 23: Atomic zinc profile of sample Q095 measured by
SIMS.

90
L = 0.5 microns, and the length of time that the sample was at
growth temperature, t = 90 minutes, a rough estimate of the
diffusion coefficient, D, of zinc in InP can be calculated
using this equation:
D(cm2/sec) = L2(cm)/t (sec) (7)
A value of D = 4.6xl0'13cm2/sec is calculated which agrees very
well with the range of values that were reported by Nelson and
Westbrook[58], D = (l-6)x I0'13cm2/sec, for zinc in InP.
A review of the literature on the topic of p-type doping
of InP by several growth techniques, extensive data on zinc,
magnesium and cadmium p-type doping of InP by MOCVD, (which
was acquired when this investigator was a visitor at BNR) and
a theoretical model of the p-type doping process of MOCVD InP
is all presented in Chapter III of this text. The reader is
therefore referred to Chapter III for a more detailed and in-
depth discussion on p-type doping of InP.
2.5.3 Growth of GalnAs Lattice-Matched to InP
The mixed crystal Ga 47In 53As which is lattice-matched to
InP, has been grown using the Japan Oxygen MOCVD system. The
growth conditions used were similar to the optimum conditions
for InP growth. Based on the results of several different
characterization techniques, optimum growth conditions for
undoped Ga 47In 53As were determined. Whenever p- and n-doped
GalnAs was required for device applications, test layers were
grown to calibrate for the required DEZn and H2S gas phase

91
mole fractions. The timing and sequence for a typical MOCVD
growth of GalnAs on InP is presented in Appendix B of this
text.
Since GaxIn.,.xAs has only one composition (x=.47) which
is lattice-matched to InP, precise control of both the TMIn
and TEGa flow rates to the reactor is crucial. A small change
in the TMIn to TEGa gas phase molar ratio is approximately
equivalent to the change in deposited solid phase molar ratio.
Unfortunately, small changes in the solid phase composition
dramatically affect the material's quality and both electrical
and optical properties. When the lattice-mismatch (Aa/a),
where a is the lattice constant, is greater than approximately
0.5% and layer thickness is greater than the critical
thickness of the material ( 1000 for GalnAs), the strain in
the epitaxial layer is enough to form cracks or dislocations
which can propagate throughout the grown film. Dislocations
appear as a "cross-hatched" pattern and are clearly visible in
the surface of a grown layer as viewed under an optical
microscope. The surface morphology of GalnAs deposited on InP
is directly related to the degree of mismatch in the thin film
relative to the substrate.
Using X-ray diffraction (XRD), it is possible to
determine the lattice constant of a deposited thin film
relative to that of the substrate. With the lattice constant,
one can determine the lattice-mismatch and composition, x, of
a mixed crystal such as GaxIn1.xAs. The XRD technique is

92
explained by Cullity[92]. It relies on the periodic structure
of a crystal to scatter incident X-rays in such a way that
some of the scattered beams will be in phase and reinforce
each other to form diffracted beams. Scattered rays will be
in phase if Bragg's law is satisfied:
nA. = 2d sin0 (8)
where:
(h2 +k2 +12 )1/2
and X = 1.54051 (for copper Kal) (hkl) are miller indices
(usually (004)), n is an integer (equal to 1 for first order
reflection) and 0 is the angle of incidence of X-rays. Figure
24 shows the XRD spectra from two different GaxIn1.xAs/InP
growths. The plot on the left is the spectrum for growth Q015
which is lattice matched. The reason there are two peaks is
because two different X-ray wavelengths close to each other
(Kal and Ka2 from copper) were incident upon the sample. The
spectrum on the right is from growth Q007 and it has a third
broad peak at a lower 29 angle relative to the InP substrate
peaks which when inserted into Bragg's law yields a lattice
constant of a = 5.897. This is the lattice constant of
Ga 40In_60As (see Table 1 for the lattice constants of InAs and
GaAs) which has a lattice-mismatch of Aa/a = 4.6xl0'3 or
0.46% on InP. If a GalnAs/InP sample had an X-ray spectrum

93
Figure 24: X-ray diffraction patterns of GalnAs samples grown
on InP (left: lattice-matched, right: lattice-
mismatched) .

94
with more than two large peaks, like sample Q007, then no
further characterization was performed on the sample.
For lattice-matched Ga 47In 53As on InP, the C-V profiler
is a characterization technique which was routinely used. A
C-V profile of growth Q005 is shown in Figure 25, which is an
undoped GalnAs layer deposited on n+-InP. As shown, the grown
layer thickness is approximately 0.7/xm and the substrate-
epilayer electrical interface is extremely abrupt. The
background doping level of this sample varies slightly from
2xl015 to 8xl014cm'3, but the average is very low which is
indicative of high quality material. The C-V profiler is less
accurate for profiling lattice-mismatched GalnAs/InP samples
and consequently was not used for some films.
Another technique that was used to investigate the purity
and also optical properties of GalnAs/InP samples is low
temperature (4.2 K) PL. Figure 26 shows the PL spectra for a
precisely lattice-matched Ga 467In 533As on InP grown by MOCVD.
Before PL was performed the degree of lattice-matching was
determined by XRD. This was necessary as the PL technique is
rather time consuming and it also requires a great deal of
operator expertise. Also, the PL set-up was not always
equipped to test the GaxIn1.xAs on InP samples as they require
a detector which responds to light over the wavelength range
0.9 to 1.7 microns. The spectra are quite valuable, though, as
a lot of information can be derived from them. Based on the
peak position in wavenumbers (inverse wavelength)

95
Figure 25: C-V profile of growth Q005 (lattice-matched n-
type GalnAs on InP).

RELATIVE IM1ENSITY
96
12 j
ioi X= 1.67 nm @300 K
8-
(0
kO
lO
r
m
to
<0
0.81016 eV
h 3.47mv fwhm
4 -
0 1-
J
5600 5800
6000
6200
6400
6600
6800
7000
Wavenumbers (cm 1)
Figure 26
PL spectrum of GalnAs lattice-matched to InP
measured at 4.2 K.

97
6533.556cm"1, which corresponds to an energy of 0.81016eV, and
the changes in bandgap with respect to temperature (-2.285xl04
eV/K), the room temperature emission wavelength of this
material is 1.67/m; this is the emission wavelength of lattice
matched Ga ^In 533As on InP. Hence, PL can be used to
determine ternary material composition. Also, the full peak
width at half of the peak's maximum (FWHM) of 3.47meV is
another indication that this layer is of excellent quality.
Finally, the fact that no other peaks exist in the spectrum
indicates that this sample has a very low impurity level.
The PL technique can also be used on lattice-mismatched
samples. Figure 27 shows a PL spectrum of growth Q148 which
was determined by the XRD technique to have the lattice-
mismatched composition Ga 435In 565As. The PL spectrum agrees
with XRD as the room temperature wavelength equivalent of
6380.88cm'1 wavenumbers is 1.714/im which would be indium
arsenide-rich material. Unfortunately, the temperature
dependence of mismatched GaxIn1.xAs is unknown so an exact
composition cannot be calculated. However, when comparing
this spectrum with the one in Figure 26, it is clear based on
the FWHM = 7.94meV being larger and that two peaks exist
instead of one perfect crystal transition, that the material
from growth Q148 is certainly inferior to that of growth Q144
(Figure 26). When comparing PL and XRD, the PL technique is
more accurate as it can correctly indicate that an epitaxial
layer is lattice- mismatched when XRD incorrectly does not.

REIATNE INTENSITY
98
Figure 27: PL spectrum of a lattice mismatched GalnAs film on
InP measured at 4.2 K.

99
The Hall effect was also used to investigate the quality
of undoped GalnAs/InP samples. In this case, the room
temperature mobility (^-¡00K) and carrier concentration (N0-NA)
were used as a basis of comparison between samples grown at
different growth temperatures (600 to 700C) and different
V/III ratios (12.5 to 50) The Hall data for both studies are
presented in Table 6. As shown it appears that a growth
temperature of 600C and V/III ratios between 25 and 37.5
yield better material. It should be noted, though, that at a
V/III ratio of 25, the extended defect density was much higher
than in layers grown using V/III ratios of 37.5 and 50.0.
Since it was evident that high quality GaxIn.,.xAs can be
grown lattice-matched to InP, one important application of
this material system was investigated: low dimensional
structures involving heteroepitaxial GalnAs/InP superlattices
and multiple quantum wells. A TEM (transmission electron
micrograph) photo of a cross-section of alternating 1,150
thick layers of GalnAs and InP is shown in Figure 28. This
layer was grown as a test structure for an optical
interference filter device requiring a layer pair thickness of
2000 and also to test the interface abruptness capability of
the MOCVD system. As shown, it appears that the abruptness is
excellent and the layer pair thickness is very reproducible
throughout the stack structure; both are equally important for
good interference filter response. A SIMS profile (which was
measured at VG Inc.) of another low-dimensional structure,

100
Table 6
Growth Temperature and V/III Ratio Studies on
GalnAs/InP Material (H2 = 5SLM, Pg = 80 Torr)
V/III = 50
Tg=
600C
650C
700C
/iRT (cm2/volt-sec)
6067
4083
3411
Nd Na (cm'3)
3 1015
8 1015
11016
Tg = 600C
V/III =
12.5
25
37.5
50
/iRT (cm2/volt-sec)
*
4763
5141
4083
Nd Na (cm-3)
*
4.5* 1015
3.6* 1015
8 1015
Note: *
Material was very poor and ohmic contacts were
not possible.

101
Figure 28;
TEM micrograph of alternating layers of InP (light)
and GalnAs (dark) at 247,500X.

102
growth Q033, is shown in Figure 29. In this growth a multiple
quantum well (MQW) structure was attempted using GalnAs wells
(60 to 200 thick) which were grown between InP barriers (50G
thick) From the SIMS profile, it is evident that the
position where the phosphorus intensity drops does not exactly
match the position where the gallium and arsenic intensities
increase; the well barrier interface is not well defined.
This was later corrected by growing a similar structure which
made use of the pressure balancing feature of the MOCVD system
(vent and reactor line pressures equal). A TEM cross section
micrograph of growth Q136, which is almost the exact same
structure as Q033, is shown in Figure 30. However, the well
layer thicknesses attempted were much smaller and consequently
a GalnAs quantum well of 13 was grown. This layer took 5
seconds to grow in the MOCVD system using the pressure
balancing mode. As shown in Figure 30, heterostructures with
interface abruptness down to what appears to be the atomic
level, can be grown in the Japan Oxygen MOCVD system.
The PL technique can also be used to characterize the
optical properties of low-dimensional layer structures such as
Q136. When layer thicknesses as small as 13 exist, quantum
size effects can be demonstrated. In Figure 31, a PL spectrum
of sample Q136 is shown. In the figure several quantized
peaks or energy level transitions exist. These peaks are
labelled according to the individual GalnAs quantum wells

Counts
103
SIMS
Profile of
a MQW sample
CQ033).

104
Figure 30:
TEM micrograph
of MQW sample
13 3 A InGaAs
19 + 3 A InGaAs
23 3 A InGaAs
29 3 A InGaAs
33 3 A InGaAs
45 3 A InGaAs
InP SUBSTRATE
Q13 6 .

RELATIVE INTENSIFY
105
Figure 31; PL spectrum of MQW sample Q136 measured at 4.2 K

106
shown in Figure 30. Some of the peaks are not as sharp or
well defined as the others, but clearly quantum energy level
transitions have been realized by using PL on this sample. By
using the Kronig-Penney theory[93], which is a simplified
mathematical representation of the periodic potential function
for electron motion, the following equation can be derived:
E
n
n2 h2
8m *L2
(10)
where En is energy in eV, n = 1,2,3... quantum levels, h =
6.624xl0'34J-sec (Planck's Constant), m* is the effective mass
(.041 m0, for electrons and 0.50 m0, for heavy holes, m0 =
9.109xl031Kg) and L is the well thickness (A). Using
equation (10) one can find the theoretical confinement energy,
AE = E^-Eg = Ec1+Ev1 = 6.1911x10*19/L2 eV (where EC1 and EV1
are the first energy levels of the conduction and valence
bands, respectively). Of course the measured AE is the
difference between the position of each individual peak, Epealc,
and the bandgap, Eg, of Ga 47In 53As at 4.2 K (0.812eV). A
plot comparing the theoretical and experimentally measured
confinement energies versus GalnAs well thickness for growth
Q136 is plotted in Figure 32. The agreement is good for
thinner wells but deviates from theory for thicker wells. In
conclusion, the growth of low dimensional structures using the
Japan Oxygen MOCVD system is possible and has been confirmed
by TEM, SIMS and low temperature PL results.

AB (meV)
107
Figure 32: The effect of well thickness on both experimental
and theoretical confinement energy.

108
2.5.4 Growth of GalnAsP Lattice-Matched to InP
Unlike Ga^n^jjAs, a continuous range of compositions
exist for the mixed crystal GaxIn1.xAsyP1.y which are lattice-
matched to InP. The Asy composition can be varied from y=0
(InP) to y=1.0 (Ga 47In 53As), but since the relation y=2.16x
exists for lattice-matched quaternary, Gax is constrained to
only vary from x=0 (InP) to x=.47 (Ga 47In 53As) This wide
range of variable composition corresponds to material which
can emit light at wavelengths from 0.92/im (y=0) to 1.65jum
(y=l), (see Figure 1) and several layers of lattice-matched
quaternary over this wavelength range have been successfully
grown in the Japan Oxygen MOCVD system.
The MOCVD growth conditions for Ga^n^As P,, on InP
were based upon the optimum growth conditions for Ga 47In 53As
on InP. It has been experimentally determined that the
following gas phase growth conditions and the relationship
between them all have a strong influence on solid phase
material quality: MFn,In/MFTEGa, FPH3/FASH3, and the V/III
ratio. The exact values of these three ratios are strongly
composition dependent and initial estimates of these values
for a specific solid phase composition were taken from the
literature[72]. In Figure 33, the effect of growth conditions
on material quality (lattice-matching) is clearly shown. The
upper photo is of a sample from growth Q036 and it has the
characteristics of a lattice-mismatched sample, point and line

109
Figure 33: Surfaces of quaternary films grown on InP (top:
lattice-mismatched, bottom: lattice-matched).

110
dislocations. The lower photo is of a sample (Q120) that
appears to be lattice-matched. The degree of mismatch can be
determined more accurately by XRD, but since there are two
unknowns, x and y, the exact composition of a lattice-
mismatched quaternary film cannot easily be determined with
this technique. The percent of lattice-mismatch (100 x Aa/a)
for samples Q036 and Q120 are -0.824% and 0.00%, respectively.
The degree of lattice-mismatch for both of these samples was
determined from their XRD spectra which are shown in Figure
34. All that one can tell from the higher angle peak position
of the quaternary material in spectrum Q036, other than the
lattice constant, is that it has too much GaAsyP.,_y to be
matched.
One way of getting a "rough idea of the composition of
GaxIn.,.xAs Pj /InP films is by using the electron microprobe
analysis (EPMA) technique. This analytical tool is available
in the Materials Science and Engineering Department on campus,
but it is not extremely accurate. It was useful, however, for
calibrating the growth conditions as the technique does not
require that the deposited film is lattice-matched or nearly
lattice-matched to the substrate used. It gives the detected
weight fraction or atomic fraction of elements which are being
emitted from the electron bombarded surface. The EPMA
analysis of sample Q120 yielded a quaternary composition of
Ga.353In.647As.80P.20* Another technique which determines more
precisely the composition of a lattice-matched quaternary film

Ill
Angle (deg)
Figure 34; X-ray diffraction patterns of quaternary films
grown on InP substrates (left: lattice-matched,
right: lattice-mismatched).

112
is PL. The PL spectrum for growth Q120 is shown in Figure 35.
Based on the location of the peak position, and the
temperature dependence of the quaternary bandgap, the room
temperature emission wavelength of sample Q120 is X = 1.554/zm.
This wavelength for lattice-matched quaternary corresponds to
a composition of Ga-4In_6As_913P>0a7 according to Table I of
Nahory et al.[74]. A comparison of the PL and EPMA results
for all lattice-matched quaternary samples are shown in Table
7. As shown, the agreement for some samples is quite good,
but for others, Q242 for example, the agreement is poor. This
probably is due to the fact that only InP, GaAs, and a
lattice-matched GalnAs (on InP) sample were used to calibrate
the EPMA detected atomic analysis results. It would have been
ideal to have also used an InAs and GaP sample, but these
crystals were not available for calibration. Consequently,
whenever samples were found to be lattice-matched by XRD, PL
was performed and solid phase compositions were based on PL
peak positions. The EPMA results were, however, generally
close, and they were also used as feedback for lattice
matching studies when quaternary samples were too mismatched
for XRD and PL analysis.
PL can also be used to assess grown quaternary film
quality based on the shape of peaks, the number of peaks and
also their location in the spectrum. The FWHM of sample Q120
was measured from the spectrum to be 4.46 meV at 4.2 K.

RELATIVE INTENSTY
113
Figure 35: PL spectrum of sample Q120 measured at 4.2 K.

114
Table 7
Photoluminescence and Electron Microprobe Analysis of Nearly
Lattice-Matched GaxIn1.xAsyP1.y Films Grown on InP by MOCVD
Run
PL
PL(Xrt)
EPMA
Q035
Ga.437In.563As.960P.040
1.605/xm
Ga .4201n. 580a3.994P. 006
Q119
Ga.330In.670As.761P.239
1.427/xm
Ga.170In.830As.540P.460
Q120
Ga.400In.600As.761P.239
1.554/im
Ga.353In.647As.800P.200
Q164
Ga.390In.610As.895P.105
1.532/xm
Ga.410In.590As.900P.100
Q165
Ga.320In.680As.745P.255
1.400/xm
Ga. 300In. 70(AS. 720P. 280
Q238
Ga.346In.654As.787P.213
1.469Atm
Ga.314In.686As.722P.278
Q239
Ga.301In.699As.706P.294
1.371/xm
Ga.242In.758As.567P.433
Q242
Ga.283In.717As.629P.371
1.324jum
Ga. 0301 n.970a3.042P. 957
Q251
Ga. 159In.841As .350P.650
1.115/xm
Ga.121In.879As.333P.667

115
This, in addition to the observation that the spectrum of Q120
has no other peaks, indicates that the film's compositional
uniformity throughout the film and purity of the film are both
quite good. The FWHM of the PL peak for sample Q239 (1RT=
1.371/xm) is 15.5 meV. This spectrum is shown in Figure 36 and
was taken at 13 K which partially accounts for the larger peak
width due to increased lattice vibration at the higher
measurement temperature. It may also be that the control of
the composition during the MOCVD growth of sample Q239 was
inferior to that of sample Q120.
Another way of determining the purity of quaternary films
is by measuring the background carrier concentration of
undoped layers with the C-V profiler. In addition, the Hall
effect can be used to measure the average background carrier
concentration and the mobility of the layer. A C-V profile of
sample Q239 is shown in Figure 37 and from this profile an
average background carrier concentration of 2xl016cm'3 was
measured. The room temperature Hall carrier concentration and
mobility of this same sample were measured to be 1.8xl016cm'3
and 4630cm2/volt-sec, respectively. The lowest Hall carrier
concentration of an undoped quaternary film was 1.10xl015cm'3
for sample Q166. The highest room temperature Hall mobility
measured was 4810cm2/volt-sec for sample Q037. Of course
these values are composition dependent; higher mobilities are
expected from compositions closer to y=l (GalnAs).

Intensity (A.
116
Wavelength (pm)
Figure 36: PL spectrum of sample Q239 measured at 13 K.

117
Figure 37; OV profile of a n-type quaternary film on InP.

CHAPTER III
P-TYPE DOPING OF MOCVD INP: EXPERIMENTS AND MODELING
3.1 A Review of the Literature on p-Tvpe Doping of InP
3.1.1 Introduction
Epitaxial layers of the III-V compound semiconductors
InP, GalnAs and GalnAsP with compositions lattice-matched to
InP are widely used for the fabrication of optoelectronic
devices. Many optoelectronic devices rely on the creation of
an electrical p-n junction using p-type InP. InP doped p-type
by MOCVD and other crystal growth techniques can be made, but
the process is not understood in detail. At this time, an
"ideal p-type dopant with the following characteristics does
not exist: (1) a wide controllable doping range (p = 1015-
I019cm'3) ; (2) a low bulk diffusivity (D < I0'15cm2/sec) ; (3)
a controllable incorporation rate for a wide range of growth
conditions (growth temperatures, V/III ratios, group III
concentrations); (4) full electrical activity (p = atomic
concentration); (5) no memory effects in the system; and (6)
one that does not degrade layer morphology by the formation of
extended defects.
Several material sources have been used to incorporate
Hg, Cd, Mg, Zn, Mn, Be, Cu, Ca and C atoms as p-type dopants
in InP, with varying degrees of success[94-96]. Because of
118

119
this dilemma, an attempt has been made to understand the
process of p-type doping of InP with the ultimate goal of
identifying an "ideal" p-type dopant source. A review of the
literature on bulk crystal growth, LPE, MBE and CVD (metal
organic, chloride, and hydride) of p-type doping of InP has
been performed. From this search, data on the relationship
between the dopant distribution in the solid phase and the
growth conditions from either the liquid or gas phase have
been used to determine if the process is at equilibrium,
reaction limited, or transport limited. For gas phase growth,
it is evident that dopant incorporation occurs by a reaction
or transport limited process. For the liquid phase growth,
incorporation occurs from a near equilibrium process at the
solid-liquid interface which is also transport limited in the
liquid bulk.
3.2 Bulk Crystal Growth
P-type InP substrates have been used for radiation
resistant solar cells[97], buried heterostructure lasers[98],
and other modern optoelectronic device applications. Liquid
encapsulated Czochralski (LEC) pulling is the most widely used
technique for growing bulk p-InP substrates. Usually a layer
of liquid B203 seals the heated InP melt and a seed crystal is
dipped into the melt and slowly pulled out at a rate which
controls the crystal diameter. Cd, Zn, Be, Mg and Mn have
been used as p-type dopants for InP bulk crystals[99]. Zn is
the most commonly used dopant due to its experimentally

120
convenient near unity effective distribution coefficient;
K(Zn) = 0.6[99], K(Zn) = 1.4[100]. Zn is a fast diffuser in
InP (D=(l-6)xl0*13cm2/sec[58]) so in situations where abrupt
doping profiles are required, Cd doping is sometimes used.
The distribution coefficient of Cd is low though, K(Cd) = 0.1-
0.2[99]. This means that as the crystal is pulled, the Cd
concentration in the melt increases and as a result, the
atomic concentration in the solid also increases. Zn doped
InP bulk crystals have been fabricated by LEC and the
characteristics of the material have been reported by Roksnoer
and Van Rijbroek-Van Den Boom[101]. They report a Zn
distribution coefficient of K(Zn) = 0.90 which is constant
over the atomic concentration range of [Zn] = 7xl017 to
2xl019cm'3. They also report that their material was
dislocation free for concentrations greater than [Zn] =
lxl018cm'3. The most interesting data, however, were their
Hall carrier concentration results. Their observation was
that for crystals containing less than 5x1018 zinc atoms/cm3,
p [Zn] but, at [Zn] > 5xl018cm3, p < [Zn]. At the highest
atomic level [Zn] = 2.4xl019cm'3, p = 1.5xl01scm'3. Thus, a
significant part of the zinc atoms were electrically inactive.
They propose that precipitates had beh formed in the heavily
doped crystals. But in only the most heavily doped one ([Zn]
= 2.4xl019cm3) were small pits detected by preferential
etching. It is possible that Zn at high doping levels also
incorporates in some unknown form as a donor, compensating the

121
material and reducing the net hole concentration. Theories
such as this one are proposed and considered in the model
development portion (section 3.4) of this chapter.
3.1.3 Liquid Phase Epitaxy
The LPE technique has been successfully used to fabricate
many optoelectronic devices which use p-InP layers as part of
the device structure. Zinc[2], magnesium[2], cadmium[2] and
manganese[102] have been successfully used to achieve p-type
conversion in LPE grown InP and related materials. The
dopants are added to the In melt in pure metallic form, or as
alloys. The distribution coefficients for Zn, Cd and Mg in
InP by LPE are 0.7, 0.001 and 0.05-0.5, respectively[2]. Zinc
is the most commonly used dopant for several reasons. It
controllably incorporates to form p-type material with hole
concentrations from p = lxlO16 to 2xl018cm'3, whereas with Cd,
p-doping levels from p = 5xl016 to 2xl018cm3 exist[2]. With
Mg, hole data were uncontrollably scattered from p = 2xl017 to
6xl018cm3[2]. Mn is only capable of doping material at low
p-levels from 3xl016 to Ixl017cm'3[l02]. Cd diffuses less than
Zn and much less than Mg. But, the equilibrium vapor pressure
of Cd (at 625C) is eight times greater than that of zinc (200
vs. 25 torr). Hence, Cd rapidly evaporates from the heated
melt, which results in varying doping level profiles and
contamination of neighboring melts. Consequently, Zn is the
preferred p-type dopant for LPE growth of InP.

122
Wada et al.[103] also presented Zn doping results for LPE
InP. They report a distribution coefficient of K(Zn) = 0.84
for a Zn fraction in the melt of 102 atomic percent which
corresponds to a hole concentration of 2xl018cm'3. Above this
Zn fraction, they report constant then decreasing Hall hole
concentrations and attribute this trend to strong compensation
effects. Interestingly, this is the same trend that was
observed for bulk crystal growth of Zn doped InP. Wada et al.
[103] suggest (without evidence) that the decrease in carrier
concentration may be due in part to precipitate formation, but
is more likely due to compensation by an interstitial zinc
donor complex as proposed by Hooper and Tuck[104].
3.1.4 Molecular Beam Epitaxy
P-type doping of InP grown by MBE has been attempted
using several different group II atoms. Zn and Cd were used
as dopant atoms by Park et al.[105] emanating from a low
energy ion cell during InP MBE growth. Doped films remained
n-type due to the near zero sticking coefficient of these
atoms on InP at the growth temperature ~ 350C. Be doping of
MBE InP grown at 525C has been reported by Panish et al.
[106]. They report successful p-type conversion using Be with
hole concentrations from p = 6xl016cm3 to p = lxl019cm3 with
corresponding Hall mobilities of /ih = 100 to 30 cm2/volt-sec.
Recently, Mg has been used as a potential p-type dopant for
MBE InP grown at 500C. Cheng et al.[107] reported for all
layers up to the maximum beam equivalent pressure of Mg =

123
lxlO'9 mbar, p-type conversion was not obtained. Apparently,
the Mg sticking coefficient is very low for MBE-InP unlike
MBE-GaAs where Mg has been successfully applied[107]. To
date, Be is the most successful p-type dopant for MBE InP.
3.1.5 Chemical Vapor Deposition
3.1.5a Hydride
The hydride CVD technique has also been used to grow p-
type InP layers as part of optoelectronic device structures.
In the review by 01sen[108], two different methods are
mentioned for p-doping InP by the hydride technique and both
involve zinc. One method uses flowing hydrogen to carry
elemental zinc vapor from a heated "zinc bucket" into the
mixing zone of the reaction tube. A second method mentioned
uses diethylzinc as a p-type dopant source for InP, but there
are no electrical results presented for either technique. A
third technique is presented by Jurgensen et al.[109] which
made use of a Zn doped indium source. The weight fraction of
Zn in the indium was varied from 0.6 to 4.5xl05. The hole
concentrations of several subsequently grown layers from each
In/Zn mixture are presented. Surprisingly, independent of the
amount of Zn added to the indium source, the experimental hole
concentrations scatter around the same curve and appear to
only depend on the length of time that the source was heated.
Hole concentrations start at p = lxl018cm'3 (run #1) and
increase to 2.5xl018cm'3 (run #4). They then decrease to
9xl017cm'3 (run #9) Perhaps as time goes on, from run to run,

124
zinc is being gradually depleted from the In/Zn melt due to
incongruent evaporation. The first three samples may actually
have a higher atomic zinc concentration than sample #4, but
they are compensated due to the incorporation of interstitial
donor complexes. This theory agrees with the observed trend
and also with the trends seen for both bulk-grown and LPE-
grown Zn-doped InP.
3.1.5b Chloride
P-type InP has also been achieved by the chloride CVD
technique. Both Zn and Cd doped InP have been reported by
Chevrier and co-workers[110,111]. In the Zn doped InP paper
[110], 1 gram of Zn was added to the In melt which was heated
to 750C. Hole concentrations are given for consecutive
samples which were grown from the same In/Zn melt. The hole
concentration decreases as the number of runs increases and
this trend is attributed to Zn depletion in the source[110].
SIMS measurement results of the atomic zinc concentration [Zn]
are also given for a few samples, and the ratio [Zn]/p varies
from 2.35 for the low doped sample ([Zn] = 2xl018cm'3) to 9.25
for one of the highest doped samples ([Zn] = 3.7xl018cm'3) .
The low doped (p = 8.5xl017cm'3) sample was annealed for two
hours at 300C and the hole concentration was again measured
by the Hall effect method. It was reported to increase to p
= 2xl018cm'3, which is the same value as the measured SIMS
concentration, [Zn]. The authors propose that the annealing
process activates neutral Zn atoms[110]. It is this writer's

125
opinion that the grown-in compensating donor zinc complexes
disassemble upon annealing to create a defect structure which
eventually contains only electrically active zinc acceptors.
Similar to the zinc doping paper, Chevrier et al.[lll]
have published a paper on Cd doping of chloride CVD InP. They
report a linear incorporation of Cd with hole concentrations
(deduced from Hall effect and C-V measurements) in the range
p = lxlO15 to 3xl018cm'3. For cadmium partial pressures below
1.5xl0'4 atm, a distribution coefficient K(Cd) = 0.2, is
reported. Above pcd = 1.5xl0*4 atm, they report that strong
compensation limits the hole concentration to a constant value
of 3xl018cm'3 (for partial pressures up to Pcd = 8.0xl02 atm) .
Chevrier et al.[lll] also performed annealing experiments on
the Cd doped samples and found the opposite trend as reported
for the Zn annealing experiments[110]; hole concentrations
decrease with longer annealing times at 300C. The cadmium
trend, they propose, may be due to increased neutral complex
formation, increased clustering of Cd, activation of a donor
compensating complex, or deep diffusion of Cd. They also
reported gradually decreasing growth rates at high (above
1.5xl04 atm) Cd partial pressures. Because of this fact, and
the relatively high pcd used, this writer has concluded that
the clustering explanation is the most logical explanation.
3.1.4c Metal Organic
MOCVD has been applied to the growth of p-type InP
epitaxial layers for optoelectronic device structures. To

126
date, Cd, Mg and Zn have been used as dopant species for p-
type doping of MOCVD InP. Several different organometallic
compounds have been used as Cd, Mg and Zn sources such as:
dimethylcadmium (DMCd)[58,59,112], cyclopentadienyl magnesium
(Cp^Mg)[58], bis-(methylcyclopentadienyl) magnesium (MCpgMg)
[113], diethylzinc (DEZn)[45,58,112] and dimethylzinc (DMZn)
[58]. The relative merits and disadvantages of the use of
each dopant will be discussed pertinent to specific desired
device characteristics.
The use of Cd as a p-type dopant for MOCVD InP has been
reported by several researchers[58,59,112]. Blaauw et al.[59]
report atmospheric and low pressures MOCVD growth using TMIn,
PHj and DMCd. Nelson and Westbrook[58] report atmospheric
pressure MOCVD growth using adducts. The same doping trend
was reported by Blaauw et al.[59] and Nelson and Westbrook
[58]; a linear incorporation rate of Cd from p = 5xl015cm3 to
2xl018cm3 (measured by Hall) due to a change in DMCd partial
pressure from lxlO'6 to 3xl04 atm. Blaauw et al.[59] reported
this trend for both atmosphere pressure and low pressure (76
torr) growths, but observed extended layer morphology
deterioration at high DMCd partial pressures for only low
pressure growths. Nelson and Westbrook[27] also investigated
the effect of growth temperature (Tg = 550-650C) on DMCd
incorporation. They observed that as the growth temperature
is decreased, the hole concentration at a fixed DMCd flow
rate, increased. They attribute this trend to increased Cd or

127
DMCd desorption from the growing layer at high temperatures.
They also measured the diffusion coefficient of Cd from SIMS
profiles and report a value (at 600C) of Dcd = 0.4-l.4xl015
cm2/sec. Yang et al.[112] presented data for one DMCd
simultaneous InP deposition on three different orientations of
InP. They reported carrier concentrations of p = 1. Ixl017cm'3,
p = 8.7xl017cm"3 and n= lxl016cm'3 for the (100), (111A) and
(111B) orientations, respectively. The Cd incorporation is
most efficient on the (111A) orientation possibly due to the
increased number of Indium vacancies on this type of surface.
Overall Cd seems to be a fairly good p-type dopant due to its
wide incorporation range and low bulk diffusivity, but it is
not a suitable dopant for low pressure growth applications due
to the deterioration in surface morphology.
The use of Cp;>Mg[58] and MCp2Mg[60] as precursors for Mg
p-type doping of InP has been reported. Nelson and Westbrook
[58] performed atmospheric pressure adduct growths of InP and
obtained hole concentrations from p = 3xl016 to 2xl018cm'3 at
CpjMg vapor flows of lxlO'8 to lxlO'7 moles/min. The doping
level varied as the square of the dopant vapor flow, therefore
incorporation is mass transfer controlled. At lower dopant
flow rates, no Mg was detected by SIMS. They also report a
diffusion coefficient (at 600C) of DMg = 2-4xl0'15cm2/sec
which is based on SIMS profiles. In Blaauw et al.[60] similar
"threshold-like dopant incorporation using MCpjMg during low
pressure growths (76 torr) was reported. We also observed

128
substrate doping-level dependent deep diffusion of Mg, a
decrease in hole concentration and stacking fault formation at
high Mg concentrations indicating strong compensation possibly
from a Mg donor-like complex. More details of the MCpgMg work
are presented in the following section of this text. Due to
the uncontrollable super-linear incorporation rate of Mg
during both atmosphere and low pressure growths, and the
compensation and layer morphology degradation at high [Mg] for
low pressure growths, Mg may be unsuitable for certain device
applications.
Zinc doping of MOCVD InP is the most often reported p-
doping method. Two sources, DEZn[45,58,112] and DMZn[58],
have been used in either the bubbler or diluted in a high-
pressure cylinder configuration. Nelson and Westbrook[58]
used both configurations (DEZn in a bubbler at 0C and 750 ppm
DMZn in H2 in a high pressure cylinder) and reported little
difference between the use of either Zn precursor. They
obtained hole concentrations of p = 4xl017 to 2xl018cm3
(calculated from Hall effect data) using Zn, with a square
root dependence on DEZn or DMZn vapor flow from 2xl0'8 to
lxlO6 moles/min. Above lxlO'6 moles/min, the Hall hole
concentration became saturated at 2xl018cm'3 which is similar
to the saturation level (2-3xl018cm'3) reported by the other
crystal growth techniques already discussed. Razeghi and
Duchemin[45] observed a wider range of incorporation p = 0.2-
4.0xl018cm3 from their experiments using organometallic DEZn.

129
They also reported that by increasing the growth temperature
from Tg = 530 to 650C at a fixed DEZn flow, the 300 K hole
cpncentration decreased from 4xl018 to 4xl017cm'3 due to, as
they explain, increased dopant evaporation from the growing
surface.
Nelson and Westbrook[58] also calculated the diffusion
coefficient of Zn in InP based on SIMS depth profiles of
atomic zinc into InP substrates coming from epitaxially grown
InP:Zn layers. They report a diffusion coefficient (at 600C)
of DZn = l-6xl0'13cm2/sec, which is approximately two orders of
magnitude larger than the values they reported for Cd and Mg.
Yang et al. [112] grew a Zn-doped InP layer by MOCVD using DEZn
simultaneously on a (100) and (111B) InP:Fe substrate. The
reported hole concentrations calculated from room temperature
Hall data were p = 6.5xl017 and l.9xl017cm3 on (100) and
(111B) orientations, respectively. Perhaps Zn adsorbs more
strongly on the (100) orientation resulting in a higher hole
concentration. No one reports layer morphology degradation at
high doping levels of Zn. This reason, in combination with
its reasonably wide incorporation range, may explain why zinc
is the most often used p-type dopant. This conclusion is
valid for both low-pressure and atmospheric pressure MOCVD
growth of InP, even though its bulk diffusivity is quite
large.

130
3.2 MOCVD Growth and Characterization of Mg-Doped InP
Using bis-(MethvlcvclopentadienvH Magnesium as a
Dopant Source
3.2.1 Introduction
Epitaxial layers of Mg doped (p-type) InP have been grown
by the low pressure MOCVD method using MCpjMg as a MO source.
Previously, CpgMg doped InP has been grown by the atmospheric
pressure MOCVD method[58,113]. The atomic incorporation and
subsequent diffusion of Mg in InP has been determined by SIMS.
The surface morphology of grown Mg:InP layers has been
investigated using an optical microscope equipped with
Nomarski phase contrasting and also by TEM. The electrical
characteristics of the layers were measured by the Van der
Pauw Hall effect technique and by an electrochemical C-V
profiler. The optical characteristics of the layers were also
measured by PL at 7 K.
3.2.2 MOCVD Growth
The Mg-doped InP crystal growth was performed by this
investigator in a commercial (CVT, Ltd.), custom designed,
MOCVD reactor at BNR in Ottawa, Canada. The organometallics
TMIn and MCp2Mg were used and kept at 17C, 800 torr and 22c,
700 torr, respectively. Fifteen percent phosphine diluted in
UHP hydrogen was also used as a source gas. The carrier gas
was hydrogen and the total H2 flow rate was electronically
controlled to be 7 SLM. A radial manifold existed at the
inlet to the horizontal quartz reactor. The reactor was held
at a pressure of 75 torr and RF inductively heated to a growth

131
temperature of 625C. A deflector and baffle were used in the
reactor to improve both the gas mixing and the gas flow
pattern, respectively. A TMIn mole fraction of 0.7 x 104 and
a V/III ratio of 140 were used for the growths. Both S-doped
(n8xl018cm*3) and Fe-doped (semi-insulating) (100) oriented
InP substrates were used. Prior to loading into the reactor,
the substrates were mechanically polished using a bromine-
methanol solution, rinsed in methanol, and blown dry with
filtered high purity nitrogen. Clean quartzware (the liner
tube and gas deflector) was used for each individual MOCVD
growth and the graphite susceptor was baked-out before each
series of experiments to aid in experimental accuracy and
reproducibi1ity.
3.2.3 Results and Discussion
3.2.3a. Mo Incorporation and Laver Morphology
The atomic Mg concentration profile in the InP:Mg layers
was determined by SIMS using 157CsMg+ molecular ions with a
24Mg implant as a standard. A representative SIMS profile of
growth B232 is shown in Figure 38. During this growth, the
flow rate of hydrogen through the MCpgMg bubbler was 12.5 seem
and this resulted in a [Mg] = 3xl018cm3, atomic magnesium
concentration. All of the magnesium doped layers had a growth
rate of approximately 1.0/Lim/hour; independent of magnesium
concentration. Figure 39 shows the effect of the H2 flow rate
through the MCp^g bubbler on the atomic Mg concentration in
the layers determined by SIMS. An interesting feature of this

Mg Concentration (cm
132
CO
I
Figure 38: Atomic magnesium SIMS profile of sample B232 which
was grown on a S-doped InP substrate.

133
Figure 39; The effect of H2 flow rate through the bubbler on
[Mg] as determined by SIMS.

134
plot is the rapid increase of the Mg concentration with a
small change in the H2 flow rate. This unfortunate phenomenon
has also been reported from Mg doping of GaAs, using the
MCpjjMg source [114].
At low Mg concentrations smooth layers were grown, but at
a concentration of I019cm'3, the surface morphology started to
deteriorate as witnessed by the appearance of stacking faults.
The morphology at a particular Mg concentration was similar
for layers grown on Fe-doped and on S-doped InP substrates.
Several samples were investigated by TEM, using both plan view
and cross-sectional samples. No features indicative of
increased defect densities or precipitates could be detected
at any Mg doping level below I019cm'3, but at higher doping
levels stacking faults were found up to a density greater than
109cm'2.
3.2.3b Mo Diffusion in S-doped InP Substrates
Figure 40 shows SIMS profiles of Mg for layers grown at
different Mg doping levels on S-doped substrates. In addition
to the Mg in the epilayer, Mg spikes, like the one clearly
seen in Figure 38, were observed at the epilayer/substrate
interface, with a subsequent rapid decay of the Mg signal to
the instrumental detection limit, at I015cm'3. For layers
grown with a high Mg doping concentration, ([Mg] > 3xl019cm'3) ,
a Mg concentration of I019cm'3 was attained in the doping
spikes, independent of the Mg level in the as-grown epilayer.

Mg Concentration (cm
135
CO
i
1020
1019
1018
1017
1016
1015
0 12 3
Depth (/urn)
Figure 40: SIMS Mg profiles of InP layers on S-InP (symbols
represent H2 flow rates of 5, 12.5, 22 and 27.5
seem to the Mg bubbler, respectively).

136
In Blaauw et al.[115], results were presented on the diffusion
of Zn in InP. It is proposed that the presence of sulfur and
silicon donor atoms in InP can act as traps and immobilize
zinc. As shown in Figure 40, apparently magnesium can also be
trapped up to a level corresponding to the substrate donor
concentration (n=lxl019cm*3) Also the depth of diffusion (1
to 2 microns) appears to be determined by the total amount of
Mg diffusing across the substrate/epilayer interface and
depends on the diffusion time (one hour) and the epilayer
doping level.
The increase in Mg concentration with epilayer depth for
the Mg profile at the lowest doping level in Figure 40 was
observed in all low concentration Mg-doped layers grown on
both S-doped and Fe-doped substrates. This phenomenon may be
related to a Mg gas phase depletion reaction which occurs with
an increasing rate as the reactor wall gradually gets coated.
3.2.3c Ma Diffusion in Fe-dooed InP Substrates
The atomic Mg concentrations of InP:Mg layers grown on
InP:Fe substrates were the same as the layers simultaneously
grown on InP:S substrates. The Mg diffusion depths however
were much greater, up to 32/xm deep for the layers grown on
InP:Fe substrates. The SIMS profiles for four different
samples are shown in Figure 41. As shown, similar to the
diffusion in InP:S, the Mg is immobilized but at a lower level
of I017cm'3 and then drops off abruptly to the instrumental
detection limit ( I015cm'3) The I017cm'3 plateau must be

Mg Concentration (cm
137
CO
i
10
20
epilayer/substrate
interface
7
2 3 4 5
Depth (m)
Figure 41: SIMS Mg profiles of InP on Fe-InP (symbols
represent H2 flow rates of 5, 12.5, 22 and 27.
seem to the Mg bubbler, respectively).

138
related to the substrate iron doping level (0.3 10xl017cm'3)
and suggests the existence of an immobile Fe-Mg complex.
3.2.3d Electrical Characteristics
Electrochemical C-V profiling measurements were carried
out for layers grown on S-doped substrates, and flat profiles
were typically obtained. A typical C-V profile is shown in
Figure 42. As shown, the hole concentration for this sample,
growth B229, is p=2xl018cm'3 and the layer thickness is roughly
one micron. The hole carrier concentration determined by C-V
profiling is shown in Figure 43 as a function of the Mg (SIMS)
concentration in the layers. Up to a concentration level of
[Mg] 2xl018cm'3, an approximately linear relationship between
C-V profile hole concentrations and SIMS measured atomic Mg
concentrations (on both S- and Fe-InP) was observed. As [Mg]
was increased further, the net hole concentration gradually
decreased from a maximum value of p 2xl018cm'3 to a value of
p 8xl016cm'3 [Mg] 3xl019cm3. This suggests that at high
levels, Mg is acting as a donor, compensating InP.
Hall measurements were performed on layers grown on Fe-
doped substrates but the results were not considered reliable
due to the uncertainty in depth from the extended diffusion of
electrically active Mg atoms. Electrochemical C-V profiling,
using the front contact method, was also performed on the
InP:Fe layers. The SIMS and C-V results for a sample doped at
[Mg] = 2xl018cm'3 are shown in Figure 44. The agreement
between SIMS and C-V profiles confirms that all the Mg is

log N(cm~3)
139
Figure 42: C-V profile of growth B229 (InP:Mg on InP:S).

140
Figure 43: The relationship between atomic [Mg] and hole
concentration for InP:Mg layers.

141
10
19
10
18
CO
i
!*
o
CO
c
0)
o
c
o
o
10
16
10
15
10
14
epilayer/substrate
interface
C-V Profiling
SIMS


CD

2 4 6
depth (/i,m)
8
Figure 44; SIMS and hole concentration profiles of InP:Mg
grown on InP:Fe.

142
electrically active, including that of the diffused Mg at the
[Mg] I017cm'3 plateau. Similar profiles for a highly doped
sample are presented in Figure 45, and show that in the
surface region of the epilayer, where [Mg] is high, a much
lower hole concentration is measured. This result is in
agreement with the relationship between [Mg] and the hole
concentration as observed for layers on S-doped substrates at
high [Mg] (Figure 43).
3.2.3e Photoluminescence Results
The low temperature PL of the Mg doped InP is shown for
several dopant concentrations in Figure 46. The nominally
undoped (n-type) material shows strong donor and free exciton
(D-X) recombination at 874 nm; in addition we see weak band to
acceptor (Zn, e-A) transitions, at 900 nm, related to residual
Zn contaminants[116]. As small guantities of Mg are added,
acceptor-bound excitons (A-X) at 877 nm become apparent[116]
and the dominant e-A transition shifts from 900 nm (e-Zn) to
896nm, corresponding to e-Mg acceptor transitions[117]. In
the middle Mg concentration range, the e-Mg transition
broadens and becomes dominant. At [Mg] > 2xl018cm'3 the
spectra consist of a single broad peak. At the same nominal
excitation conditions, the peak occurs at higher wavelengths
as [Mg] increases. This trend may be related to the existence
of compensating donors at high Mg concentrations as suggested
by C-V measurements.

143
10
20
10
16
epilayer/substrate
interface
C-V Profiling
SIMS
Depth (m)
Figure 45: Mg and hole concentration profiles of InP:Mg on
InP:Fe.

PL Intensity (arbitrary units)
144
Figure 46; PL spectra of InP:Mg layers measured at 7K.

145
3.2.4 Conclusions
The use of MCpjMg as a p-dopant source for MOCVD InP has
been investigated and the Mg incorporation was found to be
non-linear. For Mg concentrations above I019cm3 the layer
morphology deteriorated (stacking faults were observed by
TEM). Extended diffusion of the Mg in the grown layers into
both InP:S and InP:Fe substrates was observed. It appears
based on the shapes of C-V and SIMS profiles, that immobile
Mg-S and Mg-Fe complexes form in InP. At doping levels
exceeding I019cm'3, significant electrical compensation takes
place; the net hole concentration decreases with increasing
[Mg] in the layers. Because of these abnormal incorporation
and doping characteristics, Mg may be considered an unsuitable
dopant for InP in some device applications.
3.3 Experimental DMCd and DEZn p-Tvpe Doping of MOCVD InP
3.3.1 Introduction
In addition to Mg which was discussed above, the use of
Cd and Zn as p-type dopants for MOCVD InP has been studied
while this investigator was a visitor at BNR. During these
experiments, the metal organic (MO) compounds dimethylcadmium
(DMCd) and diethylzinc (DEZn) have been used as Cd and Zn
sources, respectively. The effect of MO flow rate on layer
morphology, atomic incorporation and electrical activation in
InP has been determined by Nomarski optical microscopy, SIMS,
Hall and C-V measurements. For DEZn, the effect of varying

146
the V/III ratio and growth temperature on material properties
has also been studied. The ultimate goal of this work was to
find the overall "ideal p-type dopant for InP based devices
grown in the BNR low pressure MOCVD system.
3.3.2 DMCd Results
The DMCd doping experiments were carried out in the CVT
Ltd. MOCVD system which was discussed in section 3.2.2. The
InPtCd growth conditions were: Pg = 76 torr; Tg = 625C; FH2
= 7 SLM; V/III = 140; DMCd total bubbler pressure = 600 torr;
and, DMCd bubbler temperature = 15.9C (vapor pressure = 3.898
torr). The hydrogen flow rate through the DMCd bubbler was
varied from 90 to 700 seem which resulted in a DMCd molar flow
rate range of 2.94xl0'5 to 2.284xl0'4 moles/min. At the higher
DMCd flow rates, the material quality rapidly degrades due to
the formation of large hillocks possibly from the formation of
Cd-P precipitates. The hole concentration of these grown
layers is plotted as a function of DMCd molar flow rate in
Figure 47. Also plotted in Figure 47 is Cd doping data from
the same reactor of a previous study by Blaauw et al.[59]. A
linear relationship between DMCd flow rate and hole carrier
concentration for both data sets is evident. The main
conclusion is that at high DMCd molar flow rates grown at a
low pressure of 76 torr, InP:Cd material quality is poor.
This is unfortunate because DMCd has a wide incorporation
range and Cd is a relatively slow diffuser in InP[58].

-Nn (cm
147
Figure 47: The effect of DMCd molar flow rate on InP hole
concentration. The circles are data taken from
Blaauw et al.[59]

148
3.3.3 DEZn Results
The DEZn doping experiments carried out in the BNR
reactor made use of two different physical sources in order to
get a wide range of source flow rates. One source was a DEZn
bubbler which was kept at -15C (vapor pressure = 0.619 torr) ,
the other was a high pressure gas cylinder containing a
mixture of 92 ppm DEZn by volume with the balance being UHP
H2. With these concentrations, and a flow rate as low as 5
seem from the cylinder and as high as 200 seem of H2 through
the bubbler, DEZn molar flow rates from l.lxlO'7 to 4.92X10*4
moles/min were attained. Unless noted otherwise, all the
experiments were performed with the same basic conditions as
listed for the DMCd experiments.
After the growths were completed, several methods were
employed to characterize the InP:Zn thin films. The surfaces
were observed under an optical microscope and for all doping
levels, the grown layer morphology was smooth and essentially
featureless. A SIMS profile of the atomic zinc concentration
in InP:Zn sample B316 is presented in Figure 48. As shown,
the [Zn] is 7xl016cm'3 for this 1.4/xm thick layer which was
grown using 175 seem of DEZn from the gas cylinder. The
"spike" in the profile is probably due to a donor-acceptor
complex as was observed in the InP:Mg samples. The [Zn]
(SIMS) is plotted in Figure 49 versus the DEZn molar flow rate
for experiments using both the bubbler and cylinder sources.
It is evident that up to approximately 3xl0*6 moles/minute,

149
CO
£
c;
c
O'
I HM
4-*
CO
CD
O
c
o
CJ
c
N1
Depth (ym)
Figure 48; SIMS profile of atomic zinc in InP grown on InP:S
(sample B316).

150
-6
DEZn flow rate (x10 Moles/minute)
Figure 49; Total zinc incorporation in MOCVD InP from both
bubbler and cylinder sources as determined by SIMS.

151
the [Zn] is linearly proportional to DEZn flow rate, then, a
saturation level of [Zn] = 2-4xl018cm3 exists. Also, the
measured InP growth rate is unaffected by the DEZn flow rate
for all of the experiments performed and, conversely, the
doping level is unaffected by a change in the growth rate.
Electrical characterization of grown layers was also
performed by using an electrochemical C-V profiler and by
performing Hall effect measurements. A typical C-V profile of
an InP:Zn growth with DEZn from the bubbler (sample B318) is
shown in Figure 50. This 2.4/un thick InPtZn film was also
characterized by room temperature Hall effect measurements and
the C-V carrier concentration, p = lxl018cm*3, was confirmed.
Some InP:Zn samples were grown with thin lattice-matched
GaInAs:Zn contact layers to improve the likelihood of making
ohmic contacts. After the contacts were alloyed, the thin
GalnAsrZn layer was removed by etching and Hall measurements
were performed. The Hall and C-V measured 300 K carrier
concentrations are plotted as a function of atomic zinc
concentrations (determined by SIMS) in Figure 51. Less Hall
data is shown as the results from several samples were deemed
unreliable due to rectifying contacts. The hole concentration
(electrically active Zn) for the highly doped samples is less
than the atomic (total) Zn concentration. In other words, at
high DEZn molar flow rates a fraction of the zinc atoms are
incorporated into InP as either electrically neutral or
possibly compensating donor defects.

log N(cm-3)
152
Figure 50: C-V profile of sample B318 InP:Zn (bubbler) grown
on InP:S.

153
Figure 51; The relationship between SIMS zinc concentration
and hole concentration in InP:Zn.

154
In an attempt to obtain more information about the nature
of the DEZn p-type doping process of MOCVD InP, a series of
experiments were performed where the V/III ratio and growth
temperature were independently varied. The V/III ratio for
all of the other DEZn doping experiments was held at 140.
Using a constant DEZn molar flow rate of 5.0xl0*7 moles/min,
V/III ratios of 10, 50, 140, and 200 were used for separate p-
doping experiments. In another series of growths, a higher
DEZn flow rate, 2.0xl0'5 moles/min, was used with the same
four V/III ratios. These eight samples were characterized by
SIMS and C-V profiling and the resulting hole and atomic zinc
concentrations are plotted in Figure 52 as a function of V/III
ratio. It appears that by increasing the V/III ratios, both
the hole and zinc concentrations decrease for both sets of
experiments. This trend may be misleading since the phosphine
source is only 15% concentrated which means that the total
flow rate to the reactor increased by 200 seem when the V/III
ratio was increased to 200, and decreased by 420 seem when the
V/III ratio was reduced to 10, both relative to the base
condition, V/III = 140 and phosphine mixture flow of 457 seem.
An increase or decrease in the total flow rate does reduce or
increase the relative DEZn partial pressure in the reactor for
a fixed DEZn molar flow rate. It's also interesting to note
that, as plotted, the ratio of hole to zinc concentration is
less for the high [Zn] runs than it is for the low [Zn] runs,
indicating less electrical activation.

Zn or hole concentration (cm
155
CO
o
o
o
o


SIMS, low Zn
CV low Zn
SIMS,high Zn
o CV ,high Zn
50 100 150 200
V/lll ratio in gas flow
250
Figure 52: The effect of V/III ratio on Zn incorporation and
hole concentration.

156
The effect of MOCVD growth temperature on atomic zinc
incorporation was also studied. Layers of InP p-doped using
DEZn at a fixed molar flow rate of 2.0xl06 moles/min were
grown at temperatures from 550 to 680C. The effect of growth
temperature on atomic incorporation as determined from SIMS
profiles is plotted in Figure 53. As shown, the atomic Zn
incorporation drops from 4xl018cm'3 to 5xl017cm'3 as the growth
temperature is increased. This is the same trend that was
previously discussed and reported for DEZn doping[45] and DMCd
doping[58]. This trend indicates a build up of DEZn, or some
species containing zinc, near the InP:Zn growing surface which
has a temperature dependent equilibrium vapor pressure.
Hence, at higher temperatures, more zinc evaporates from the
surface, or stays in the gas phase, and less gets incorporated
into the solid.
3.4 Modeling of p-tvoe Doping of InP using DEZn
3.4.1 Introduction
Effective control of p-type doping in InP grown by MOCVD
is important in many solid state devices such as lasers,
photodetectors, and heterojunction bipolar transistors. The
process of p-type doping is not fully understood at this time.
A model explaining this process could be useful in selecting
a suitable dopant and/or proper doping conditions for a wide
range of optoelectronic device applications. Experiments have
been performed at BNR and at the University of Florida on the
use of the MOs dimethylcadmium (DMCd) diethylzinc (DEZn) and

157
Figure 53: The effect of growth temperature on Zn (from DEZn)
incorporation in InP.

158
bis-(methylcyclopentadienyl) magnesium (MCpgMg) as sources for
the p-type doping of MOCVD InP. A summary of the measured
hole concentration versus dopant partial pressure relationship
for each of these three dopants is presented in Figure 54.
Using the definition of an "ideal" p-type dopant given in
section 3.3.1, the information given in sections 3.3.2 and
3.3.3, and the data shown in this figure, it is clear that
DEZn is the best p-type dopant for MOCVD InP grown at a low
pressure. Consequently, the experimental data and other
information gathered from the literature review on DEZn doping
of InP will be used to formulate a model of the p-type doping
process of MOCVD InP.
Bulk crystal zinc doped InP has been grown by the LEC
method[101]. Epitaxial layers have been grown from the liquid
phase by LPE[103] and the gas phase by hydride[109], chloride
[110] and metal organic[45,58] CVD. Zinc doped crystals or
layers grown using these different techniques have been
reported to have approximately the same following dopant
incorporation and electrical activation relationship relative
to the dopant source concentration: (1) zinc has a solubility
limit in InP of 2-4xl018cm'3, and then forms ZnP; and (2) once
this solubility is attained, at higher source concentrations
measured hole concentration levels sometimes remain constant
and sometimes decrease. For LEC grown bulk crystals[ 101 ], the
decrease in 300 K hole concentration has been attributed to
precipitates without substantial evidence. Wada et al.[103]

159
p (cm
Dopant Partial Pressure (atm)
Figure 54: The effect of dopant partial pressure on hole
concentration for InP:Mg, InP:Cd and InP:Zn.

160
state that a decrease in hole concentration at higher zinc
melt concentrations is due to compensation in the form of an
interstitial zinc donor complex originally proposed by Hooper
and Tuck[104]. Hydride CVD grown material has been reported
to increase and then decrease in hole concentration with the
number of experiments performed using the same indium-zinc
heated source[109]. This writer proposed that this trend is
due to a coupling of zinc compensation and evaporation.
Chloride CVD grown samples with initially low zinc electrical
activation, became, upon annealing, fully active[110]. This
can be explained by the disassociation of compensating defect
complexes due to prolonged heating. Finally, MOCVD grown
samples saturate at p = 2xl018cm'3 and the hole concentration
has a square root dependence on DEZn partial pressure[58].
As the MOCVD growth temperature is increased, hole and
atomic zinc concentrations decrease due to increased zinc
evaporation from the growing surface at higher temperatures
[45]. Most of the above trends for MOCVD grown InP:Zn have
been observed in the BNR DEZn doped InP data which is shown in
Figure 55. Note that at low [Zn], p [Zn], but above [Zn] =
2xl017cm'3, [Zn] eventually becomes much greater than p. In
addition, for the BNR data, the low growth rate of undoped
InP, one micron per hour, was unaffected by the DEZn molar
flow rate even at its maximum, 4.92xl0'4 moles/min, and the
hole concentration was unaffected by a change in growth rate
at a fixed DEZn flow rate.

161
101,1
CO
*E
o,
c
o
cS
o
c
o
o
o
c

o
Q.
10
18.
10
17.
10
16.
SIMS
a

C-V or Hall
i r 1111 n 1 i i" i 111 m" 1 " i11 ir
10'* 10'8 10'7
10'6 105
DEZn Partial Pressure (atm)
Figure 55; SIMS and hole concentration data for InP:Zn which
is used for the model evaluation.

162
The degree of electrical activation in zinc doped InP
does vary with the amount of atomically incorporated zinc and
this phenomenon has been investigated by several research
teams. Williams et al.[118] have used proton-induced X-ray
emission (PIXE) combined with channeling techniques to analyze
Zn-doped InP. They concluded that neutral complexes such as
VpZnInVp, (which was first proposed by Hooper and Tuck[104]
based on external diffusion experiments), do not exist but
nonsubstitutional Zn is in the form of randomly distributed
precipitates. More recently, Lennard et al.[119] performed
PIXE experiments on Zn-doped InP and found no evidence for
precipitates, but state that VpZnInVp complexes could explain
electrical inactivation. Yamada et al.[120] theorized that
the neutral complexes VpZnInVp and ZnInVp exist in InP:Zn and
that their presence explains the double diffusion fronts
observed by electron beam induced current (EBIC) and SIMS
analysis of their externally zinc diffused samples.
Interstitial zinc, Znif both positively charged as a
compensating donor, Zn,+ [121,122] and as an intermediate
reacting with substrate donor atoms to form a neutral
complex[123], have also been proposed to explain Zn diffusion
profiles in InP. Another report states that interstitial zinc
must be electrically neutral[124] based on the shape of
simulated diffusion profiles taking the charge of m, of Zn,1",
to be 0, +1, +2. Kazmierski[124] also suggests that other

163
theories which set m = 1, or 2 do not take into account the
effect of substrate doping on zinc diffusion.
In addition to PIXE, EBIC, SIMS and diffusion theory,
other techniques have been employed to investigate the InP:Zn
activation phenomenon. Positron lifetime measurements have
been performed on InP:Zn crystals by Dlubek et al. [125]. They
report that the crystal doped with Zn to a concentration of
4.5xl018cm'3 shows strong positron trapping by vacancy defects,
but no vacancies are found in crystals doped with Zn to a
lower concentration of 2xl018cm3 or doped with Sn, S or Fe.
They also state that the existence of the complex VpZnInVp,
which is a deep unionized acceptor, is in perfect agreement
with their positron results. DLTS has also been performed on
InP:Zn and a hole trap with an energy of Evt = 0.52eV was
observed[126]. The trap's origin is attributed to either a
phosphorus vacancy or a phosphorus interstitial related
defect. Kamijoh et al.[127] performed 4.2 K PL on undoped InP
grown by MOCVD at different V/III ratios. They conclude that
Zn and C-acceptor impurity incorporation is controlled by the
V/III ratio. The effect of post growth cooling ambient on the
electrical activation of Zn and Cd dopants in MOCVD InP has
also been studied[128,129]. Cole et al.[128] reported nearly
complete activation in samples cooled in H2 only, intermediate
activation ( 50%) in samples cooled in PH3 and H2, and low
activation ( 10%) in samples which cooled to room temperature
with AsH3 and H2 flowing. They rule out the existence of the

164
VpZnInVp complex based on their results and propose that atomic
hydrogen which comes from the pyrolysis of the hydrides, has
an influence on the doping level of p-InP. They support this
theory by showing that SIMS H-profiles are related to doping
profiles. Glade et al.[129] performed annealing experiments
and they proposed that acceptor-hydrogen complexes exist
interstitially and their activation is limited by indium
vacancy diffusion. Also, hydrogen passivation of a p-type
InP:Mn sample showed a drastic decrease in hole concentration
from p = 7.4xl016cm'3 down to p s I013cm*3[l39]. So, it is
clearly evident that there is still much uncertainty about the
electrical activation process of p-dopants in InP.
3.4.2 Point Defect Structure
The incorporation of DEZn during the growth of InP by
MOCVD can be explained qualitatively by the model of Razeghi
and Duchemin[45]. The first assumption is that all the DEZn
arriving at the growing surface is decomposed. This is a good
assumption since the growth temperature is usually 600-650C
and the onset of pyrolysis of DEZn occurs at 332.3C[131].
Hence, the Zn concentration is limited by the rate of arrival
of DEZn to the hot surface. After decomposition there are two
possible limiting cases proposed by Razeghi and Duchemin[45]:
(1) all of the available decomposed dopant source material is
incorporated into the growing layer and the resulting impurity
concentration is independent of temperature and inversely
proportional to the growth rate; and (2) only a small fraction

165
of the Zn dopant is incorporated into the growing layer and
the rest evaporates and is transported away from the growing
crystal. If the growth temperature is raised, the dopant
evaporation rate increases and consequently, the fraction
incorporated decreases. Also, the doping concentration would
be independent of the growth rate. The behavior described in
case (2) is typical of DEZn doping of InP as observed for
MOCVD growth at BNR, the University of Florida and also
reported in the literature (see sections 2.5.2c, 3.1.5c, and
3.3 of this text).
Several different point defects and defect complexes have
been proposed to explain the apparent electrical inactivity of
incorporated zinc in InP:Zn. These point defects and other
host crystal point defects such as a phosphorus vacancy, Vp,
and an indium vacancy, VIn, are represented in a fictitious
two-dimensional InP lattice which is shown in Figure 56. The
relative concentrations of these defects in a real crystal
depends upon the conditions for which it was grown and the
environment it is presently in. Chemical reactions can be
written for each point defect, and traditional methods of
chemical engineering thermodynamics can also be used to derive
expressions for the concentrations of neutral and charged
point defects and complexes. The assumption will be made that
the concentration of these point defects is small relative to
the host crystal atomic concentration. This will simplify the
equations required. Temperature-dependent equilibrium

166
I I
In
I
P
In
In
p Cfcp-
In
In
Zn
In
In Q
In Q- in Q ,n
P
In
I
P
I
Zn
I ln
I I
Figure 56: Schematic representation of point defects in zinc
doped InP.

167
constants can also be obtained from experimental data or from
estimates based on values of similar materials. Hurle has
previously applied the point defect analysis in a series of
papers on undoped[132], tellurium doped[133], tin doped[134],
and germanium doped[135] GaAs. To this writer's knowledge,
this technique, applied to InP, has never been reported
before. This method has been used to understand the relative
concentrations of the point defects of InP:Zn which are shown
in Figure 56. Anti-site defects (Inp and PIn) may also exist,
but it is assumed that their concentrations are small relative
to the other defects already being considered.
As stated previously, formation and ionization equations
and equilibrium constants can be written for the point defects
in InP:Zn. If the concentrations of the defects considered
are small, then it is safe to assume that their activities are
equal to their concentrations (activity coefficients are
unity), except for electrons and holes. Also, if defect
concentrations are small, then the activities of Pp, InIn, V¡
would be unity. Using these conditions, equilibrium constant
expressions for each reaction can also be derived. The
reactions to form the host crystal and zinc related point
defects and their corresponding equilibrium relations are
given in Table 8 (Equations 11-21). Also given in Table 8 is
the electroneutrality relation (22) which is written under the
premise that all stable solids are electrically neutral.

168
Table 8
InP Point Defect Constants and Electroneutrality Relation
Reactions
Equilibrium Relations
\PA{g) + Vi = Pi
Pp + Vi = Pi + VP
Vp = V<+ e~
0 = e~ + k+
In in + Vi = Irii + VIn
0 = Vin + VP
Vln = Vfn + h+
Zn(g) + Vi = Zrii
Zn(g) + Vin = ZnJn + h+
Znin + Vp = ZnjnVp
Pp + ZninVp = VpZninVp + Pi
sb
ii
(11)
pp<
= [P¡] [VP]
(12)
jj- T'nnJVjJ]
Khz [Vp]
(13)
I

(14)
Km = [In¡\ [V,\
(15)
I (16)
Kh7
A/l7 ~ [Vm]
(17)
(18)
(19)
= 1*
(20)
rs \VpZninVp]\Pi\ icyi \
KdU ~ [ZnInVP} (21)
Tl +
in
+
Z"Fn] = P + [Vp+]
(22)

169
The partial pressures of phosphorous and zinc, pp4 and p2n, in
the gas phase near the heated InP surface are known. As
already discussed, the majority of zinc in the gas phase above
the growing InP:Zn crystal is not incorporated into the solid.
Also, InP is typically grown with a large overpressure of PH3
(V/III = 140) which cracks to form P4. Consequently, both pZn
and pP4 are set equal to the gas phase partial pressures at
the inlet to the reactor of DEZn and PH3, respectively.
It has been reported that yn and Yp/ the activity
coefficients of the electrons and holes, are only important
for closely compensated material[132], hence their magnitude
is set to unity. If the eleven constants 1^.,- K^, K^- Kd11,
are known and electroneutrality is satisfied, then the twelve
unknown defect concentrations: Pif Inf, Vp, Vp+, n, p, VIn,
VIn-, Zn}, Zn,n-, Zn,nVp, VpZnInVp, can be calculated. The
charged state of zinc related defects have been chosen based
on values reported in the literature and discussed in the
previous section, and based on the observed incorporation and
electrical characteristics of the InP:Zn data from BNR.
Undoped InP grown by MOCVD in the BNR reactor had a room
temperature carrier concentration of n = l-2xl015cm"3. The
lowest hole concentration achieved from the DEZn doping of InP
experiments was p = 2xl016cm*3 and most layers had much higher
hole concentrations. Hence, it is safe to approximate the
electroneutrality relation by its two dominant members at room

170
temperature (Brouwer's approximation[136]) p and [ZnIn']. So,
a new simpler relation is realized:
P = [ZniJ (23)
which only applies at room temperature. Useful SIMS data on
InP:Zn samples taken at room temperature was also available.
It must be assumed that zinc related defect concentrations at
the growth temperature will be approximately the same at room
temperature. The room temperature total zinc concentration,
[Zn] (by SIMS), is then given by the following equation:
[Zn] = [Zn"n] + [Znf] + [ZnInVp] + [VpZnInVp] (24)
The room temperature electroneutrality equation (equation 23)
does not apply at the growth temperature. The full form of
equation 22 in Table 8 must be used to calculated the hole
concentration at the growth temperature, but it can be
simplified based on an order of magnitude analysis. Hurle
[132] states that group III vacancy concentration [Vln] can be
neglected relative to the group V vacancy concentration. So,
the growth temperature (gt) hole equation now is:
+ [Znfn] [Vp+]
Pgt = ngt
(25)

171
Using the equilibrium constant relationships given in Table 8,
with yn = Yp = i# equations 23-25 can be solved analytically.
The solutions are:
[B4 + B,B2 pZn (Pp4) 1/4]1/2
[1 + V (B2B4(Pp4)1/4)]1/2
(26)
Prt
BiB2pZn(Pp4)V4
P
(27)
and
[Zn] =pZn
BiB2(Pp4)
1/4
Pgt
hBs
f W*
B.
(1 + -
B,
(Pp4) V4
(28)
where B,,- B7 are constants defined as follows: B, = K^K^, B2
= Khl/Kh2' B3= *^3' B4 = *^4' 5 = Kd8' B6 ~ Kd10 an{* B7 =
Kdn/K,,,. It is significant to note that when equation 26 is
substituted into equation 27, the expected (often reported in
the literature) p versus p2n square root dependency exists.
For InP, the electron-hole pair equilibrium constant,
Kh4(B4) can be approximated using the following equation[ 137]:
Kh4 = 4(2tt(memh) 1/2kT/h 2]3 exp(-||)
(29)

172
where mg/m,, = .077, ntj/m,, = .56, k = 1.38xlO*aJ/K, Tg = 898 K
(growth temperature), h = 6.624xl0'34J-sec, Eg = 1.0536eV (InP
bandgap at 625C), and m0 = 9. llxlO*31Kg. This value turns out
to be Kb4(B4) = 4.888xl029cm*3. Due to the large size of this
constant, all constants, p, and [Zn] data were converted to
mole fractions by a simple conversion factor. The factor
(which is presented in Hurle[132]) is based on the molecular
density, volume and weight of InP, and Avogadro's number. The
factor is: one mole fraction = 2.557xl021cm'3. Using this
factor, K,,4(B4) = 3.196X10'12 mole fraction2.
No direct estimate, either theoretical or experimental is
available in the literature for constants B,- Bj, and B5- B7.
So, estimates were taken from the values reported for GaAs by
Hurle[132]. These values were used as initial guesses in a
non-linear regression analysis program based on the Marquardt
method[138], which is described in the Appendix C. The
experimental hole concentration data (converted to mole
fractions) and their corresponding zinc partial pressures (in
atmospheres) were entered into the program as data set #1.
The atomic zinc data (SIMS) and their corresponding zinc
partial pressures were also entered as data set #2. Both data
sets were taken from the BNR data shown in Figure 55. The
phosphorus partial pressure, pp4 = 2.357xl0'3atm (V/III = 140) ,
was also entered as part of the equations which are shown in
Appendix C. The program was run twice, once for each data
set, and it basically fits the hole equation to data set #1

173
and the total zinc equation to the data set #2 by adjusting
the values of the unknown constants. Once convergence has
been achieved, (a 95% confidence limit is met) the program
stops and it prints out the fitted constants. Three constants
Bu B2 and Bj were fitted using data set #1. Three more, Bs,
B6 and Bj were fitted using data set #2. The resulting values
of the best fit are: B1 = 375.3 mole fraction3/atm, B2 = 2.5
mole fraction*1atm*1/4, Bj = l.23xl0'18 mole fraction, Bs = 2708
mole fraction/atm, B6 = 1003 mol fraction'1, Bj = 11.6 atm1/4.
3.4.3 Discussion of Results
A comparison between the model's best fit line and the
data for both data sets (p and [Zn]) is shown in Figure 57.
It is encouraging to see that the hole p2n square root
dependence of the model does agree quite well with the C-V and
Hall data. Also, the total zinc data, which has less scatter,
is fit even better by the model line. It is difficult to
assess the real meaning of some of the evaluated constants
which are actually combinations of two unknowns. But, using
the total zinc data (data set #2) and the fitted constants,
one can get an idea of the relative concentrations of the
proposed point defects ZnIn-, Zn{, VpZn,nVp, and ZnInVp as a
function of zinc partial pressure. These relationships are
plotted in Figure 58. As shown, the point defect model which
was proposed, indicates that at low zinc partial pressures,
ZnIn' makes up the majority of the total incorporated atomic
zinc (pw[Zn]). As the zinc partial pressure is increased,

174
DEZn Partial Pressure (atm)
Figure 57: A comparison of the model (solid lines) and the
experimental data for InPsZn.

% of Total Atomic Zinc
175
DEZn Partial Pressure (atm)
Figure 58: InP point defect distribution based on the model
results.

176
electrically inactive interstitial zinc, Zn{/ and the complex
VpZnInVp gradually increase in relative concentration as ZnIn'
decreases (p < [Zn]). The model also shows that the complex
Zn,nVp is essentially non-existent at all zinc partial
pressures.
The trends shown in Figure 57 agree very well with the
theories upon which the model was developed. However, other
theories such as compensation due to ionized interstitial zinc
and zinc-hydrogen passivation must also be evaluated by
developing new model equations. Also, this work would be more
convincing if less constants had to be fitted to experimental
data. Unfortunately, InP has not received as much attention
in the literature as GaAs and, consequently, little or no
information on lattice dilation or high temperature Hall
effect measurements are available. Perhaps future researchers
can fill these important gaps. This method could also be
applied to understanding the effects of varying the growth
temperature and the V/III ratio on zinc incorporation in InP
and extend the model to analyzing cadmium and magnesium doping
of InP. Hopefully by completing this puzzling topic with the
aid of modelling, an "ideal" p-type dopant for InP will
eventually be found.

CHAPTER IV
EPITAXIALLY GROWN INTERFERENCE FILTERS
4.1 Theory of Interference Filters
If a beam of electromagnetic radiation is incident upon
a structure consisting of films of different materials,
multiple reflections will occur within the structure[139].
Should the distances between the various boundaries be on the
order of the wavelength of the incident light, the reflected
beams will interfere. An optical interference stack makes use
of this effect. These stacks or periodic multilayer di
electric films find application as bandpass filters and
mirrors. Usually, the multilayer is a guarter-wave stack
consisting of alternating layers of high and low refractive
index material with the optical thickness of each layer a
quarter of the particular required wavelength. Passive
interference filters have been previously grown by MBE
[140,141] and MOCVD[142] in the GaAs/AlGaAs material system.
They have also been grown by CBE[143] and MOCVD [87,144] in
the GalnAsP/InP material system.
The most general method of calculating the transmittance
and the reflectance of a multilayer is based on a matrix
formulation of the boundary conditions of the film surfaces
derived from Maxwell's equations[145]. Basically, a two by
177

178
two matrix is formed from the refractive index, n? thickness,
t; angle of incidence of light of wavelength X, 0; and, angle
of propagation of light, 0, for each of the layers (see Figure
59) in the stack. The resulting matrix for layer j is as
follows:
cos <5¡ -Lsin J UJ 1
iUjSinj cos5j
(30)
where the phase is:
=
2 7T
(njt j COS0J-)
(31)
and Uj, the effective refractive index is:
ui =
n¡
parallel
COS0j
nj cos0j perpendicular
(32)
depending on whether the incident light is polarized parallel
or perpendicular to the plane of incidence [145]. The angle is calculated using Snell's Law:
nm sin0 = nj sin0j (33)
where the refractive index of the medium (air), n,,,, is unity.

179
Figure 59; Optical theory of an interference filter.

180
The complete multilayer of Figure 59 is represented by
their product matrix, M:
m,i
im21
imi2
m22
M1M2* Mt
(34)
From the individual elements, m{j, of the matrix M, the
reflectivity (neglecting absorption and scattering) can be
calculated as follows:
(""Hi ~ nsm22)2 + (njigm^ m21)2
(njn +nsm22)2 + (nji^ + m21)2
where ns is the refractive index of the substrate material.
A computer program based on equations (30) to (35) which was
written to calculate the reflectivity as a function of
wavelength for both GaAs and InP based stack structures.
Within the program, estimates of the refractive indices (RI)
of A^Ga^As and Ga^n^ASyP^y materials were calculated at
photon energies below the direct band edge. The RI values are
based on semi-empirical formulas of Afromowitz[146] for
AlxGa.,_xAs and of Broberg and Lindgren[147] for Ga^n^ASyP^y.
For quarter-wave films the number of periods of high and
low RI layers required to attain a desired reflectivity can be
calculated using the following equation from Born and
Wolf(145]:

(36)
R =
nsl n1
2N
1 + !,
n
n2
n
1
2N
where N is the number of layer pairs and n,, n2 are the layers
as shown in Figure 59. It is important to note that if n2/n.,
decreases and/or N is increased, R increases. Also, a larger
value of R will be realized if the stack is ordered such that
n2/n1 is less than unity. It is significant to note that lower
values of the n2/n., ratio will also result in narrower central
bandwidths. The GaAs/AlGaAs material system was used because
of its compatibility with optoelectronic devices. Results of
GaAs/AlGaAs interference filters are presented in the next
section of this text. All GaAs/AlGaAs layers were grown by
MBE at BNR.
MOCVD grown GalnAsP/InP material was also investigated
and the results of filters grown from this material system are
presented in section 4.4.3 of this document. The quaternary
MOCVD system at the University of Florida is equipped with
sources making it possible to grow thin alternating layers of
GaxIn1.xAsyP1.y and InP for interference filter structures.
This is significant because with the lattice-matched GalnAsP/
InP material system, a lower refractive index ratio (n2/n, =
0.9042) is possible (with light at 1300nm) than with the
GaAs/AlGaAs system (minimum ratio of 0.9240). This makes it

182
theoretically possible to get a higher reflectivity for the
same number of layers using GalnAsP/InP material system.
4.2 MBE Grown AlGaAs/GaAs Devices
4.2.1 Introduction
A novel optoelectronic structure is presented, based on
an optical interference filter with a stack of quarter-wave p-
n heterojunctions in the GaAs/AlGaAs III-V semiconductor
system. Combining these passive filters with multiple p-n
junctions to introduce electrical functions, to the best of
our knowledge, has never been reported before. Such devices
can be used as a passive optical filter as well as active
optoelectronic switch elements.
Since A = 1300nm is a wavelength of interest in fibre-
optic communications, it was chosen as the central peak
reflectivity wavelength. For quarter-wave films, the number
of periods of high and low RI layers required to attain a
desired reflectivity can be calculated using equation (35).
Useful estimates of the room temperature refractive indices
(RI) of AlGaAs at photon energies below the direct band
edge can be calculated from a semi-empirical formula of
Afromowitz [146]. The pertinent values used at a wavelength of
1300nm are: n(GaAs, Eg=1.42eV) = 3.408 and n(Al 3Ga 7As, Eg
= 1.827eV) =3.251. A typical structure of 20 periods of
97.5nm GaAs/ 102.5nm Al0 3Ga0 yAs (see Figure 60) with an
expected peak reflectivity of 84% was chosen (see equations
(31) (35)).

183
20x
975 n-GaAs: Si 5x17cm 3
1025 p-AI Ga As: Be 2x1017cm'3
r 03 07
1000 n+-GaAs (buffer)
Figure 60: A typical layer structure grown by MBE.

184
4.2.2 Electrical Theory
The RI of a semiconductor can be modulated by changing
any one of several physical parameters[148]. Of primary
interest here are the electrically controllable ones, i.e.,
the applied electric field (Franz-Keldysh effect) and the free
carrier density. The former has been reported[141], but
direct application of the high voltages necessary for
obtaining the electric field gave rise to excessive leakage
currents, which introduced thermal drift of the optical
properties.
The object of this work is to apply the high electric
field found in p-n heterojunctions to achieve modulation of
the RI without having to resort to high applied voltages.
Introducing a p-n junction into every period of the stack,
however, implies that every other junction will be reverse-
biased upon application of an external voltage. Thus, the
possibility of excessive power dissipation and attendant
thermal effects may be reintroduced if attempts are made to
pass current through the device, e.g., to inject carriers.
However, the thickness of layers determined by the
optical requirements (only lOOnm) allows neighboring junctions
to interact, as in the bipolar transistor. The stack can then
be considered as a multi-layered thyristor with at least two
stable switching states achievable. The forward voltage in
the "ON" state can be substantially lower than the breakdown
voltage of any single reverse-biased junction and can also be

185
accompanied by light emission if there is sufficient carrier
injection.
The novel active optoelectronic switch element thus
created has the useful feature of indicating its state
optically at a wavelength (e.g. 870nm) remote from the
wavelength the element is switching by means of its altered
reflectivity response (e.g. 1300nm). As with thyristors in
general, switching between the "ON and "OFF" states can be
controlled either electrically or optically.
The electronic structure of such a multi-heterojunction
stack is very complex. Simple analysis indicates, that to
achieve bistability, the number of junctions must be even
[149]. More detailed analysis is required to optimize the
modulator function, as well as to design the stack for a given
forward breakdown voltage and holding current. With an
appropriate design, the emitted light could also be involved
in the switching function by way of optical feedback.
4.2.3 Experimental
4.2.3a Crystal Growth
The superlattice was grown by periodic variation of the
A1 content in epitaxially grown AlGaAs crystals on a GaAs
(001) substrate using a VG Semicon V80-H molecular beam
epitaxy (MBE) system. The 2" diameter substrate was rotated
during growth and was prepared for indium-free mounting using
previously described methods[150].

186
Layers were grown at the substrate temperature of 670C
using the oxide removal temperature as a reference[151]. The
GaAs growth rate was ~ 1/Ltm/hr and the A1 flux was adjusted to
achieve the 30% AlAs composition in AlGaAs. As2 was the
arsenic source and the V/III flux ratio was maintained between
3 and 5.
4.2.3b Transmission Electron Microscopy
Shown in Figure 61 is a transmission electron microscopy
(TEM) photograph of a cleaved (110) and etched edge of the
stack. The alternating dark and light delineated layers
represent GaAs and AlGaAs, respectively. As shown, the
interfaces are abrupt and from this photo it is evident that
the structure has good vertical period uniformity. At 210,000
times magnification, the average GaAs/AsxGa,,_xAs layer pair
thickness was measured to be 192nm which is very close to the
design value of 200nm.
4.2.3c Double Crystal Diffractometrv
Three parameters completely determine an AlxGa.,_xAs GaAs
superlattice: t1# the thickness of the GaAs layer, t2, the
thickness of the AlGaAs layer and x, its composition. Double
crystal rocking curves give information on xv, the layer
composition of a virtual crystal having this as an average
composition and T the total period thickness. The averaged
composition, xv, is easily calculated from the angular spacing
between the substrate peak and the main superlattice peak
reflection (or zero order peak) XRD from a material in which

188
the lattice parameter is subject to a one-dimensional
modulation is characterized by satellites around each Bragg
peak of the average lattice and the total period thickness can
be calculated from the angular spacing of these satellite
peaks A6.
A superlattice cannot be completely characterized by
experimental rocking curve data alone since this yields only
two (x and T) of the three necessary parameters. Additional
information is needed, which can be obtained in one of several
ways. The complementary data can be obtained experimentally
by making additional measurements of the optical reflectance,
or PL, otherwise the individual layer thicknesses can be
measured directly by cross-sectional TEM. Alternatively,
since the intensities of the rocking curve satellite peaks are
proportional to the layer thicknesses, they can be found
interactively by fitting intensities of calculated rocking
curves to the experimental ones.
The rocking curves were recorded for the (004) reflection
on a modified Bede 6" double crystal diffractometer in the
parallel (+,-) setting using CuKa(A=1.5418) radiation. A
GaAs (001) single crystal was used as the first crystal: this
crystal when rocked with an identical one gave a rocking curve
FWHM of 10.3 arc-sec. The rocking curves were simulated
using a dynamical scattering theory based solution of the
Takagi-Taupin equations, using the method discussed in detail
by Hill[152].

187
Figure 61:
TEM cross-section photo of a
etched edge of stack MBE464.
cleaved (110) and

189
The layer structure for the interference filter as shown
in Figure 60 has a very large (200nm) period so that the
satellite peaks are very closely spaced (100 arc-sec) and can
only be resolved by double crystal XRD. The experimental
rocking curve is shown in Figure 62. The epitaxial super
lattice structure is of high quality as demonstrated by the
very large number of diffraction peaks which can be observed
on either side of the main substrate peak: up to five orders
of diffraction on the low angle side and up to 1'2 on the high
angle side. T, the average period thickness calculated from
the fringe spacing is 200nm. The average AlAs concentration
in the superlattice, Xy, as calculated from the angular
spacing between the substrate and zero-order peak is 17%,
which corresponds an aluminum concentration of x=0.34 in
AlxGa,j.xAs.
Figure 62 also shows a simulated rocking curve which is
based on a dynamical scattering theory that is presented in
Hill[152]. This curve was generated using the TEM determined
value for It is encouraging to note that the
simulation does show the correct general trends; the satellite
peaks are decreasing in intensity with both increasing and
decreasing angle. It even predicts the much lower intensity
satellites for the low angle side and the alternating high and
low intensity for the even and odd numbered satellites on the
high angle side.

REFLECTIVITY {%)
190
Figure 62: Experimental and theoretical DCD rocking curves.

191
4.2.3d Spectral Scanning
The experimental set-up used to measure the reflectivity
of the device (see Figure 63) consisted of a tungsten halogen
lamp, a chopper (f=400Hz), a 200mm focal length monochromator
with a 600 lines/mm grating, some focusing elements, a lock-in
amplifier, and a filter to separate the first and second order
of the monochromator exit light. The detector used was a
calibrated BNR 250FE GalnAs p-i-n. A computer operated the
monochromator and lock-in via a GPIB interface, and collected
and stored the data for subsequent analysis. Reflectivity
values were measured over the wavelength range of 850 to
1600nm. This was the useable range for the PIN reference
detector operating with a +5 volt reverse bias. The raw data
was stored and later normalized by division with calibration
data for the detector.
Spectral scans were performed at 3 radial positions of
the two inch wafer (2, 6, and 11mm from the center). The
wavelength position of the central peak maximum changed as a
function of radial position due to the well known drop-off of
growth rate towards the edge of a wafer caused by slight
collimation of the gallium and aluminum beam fluxes incident
upon the substrate. Also, the aluminum and gallium flux
distributions have different drop-off rates resulting in a
variation of the ternary composition away from 30% AlAs.
A change in composition means that the refractive index
and optical thickness will change accordingly. This was

192
Figure 63: Spectral scanning system for reflection,
transmission and photoresponse measurements.

193
observed experimentally by the increasing shift of the
position of the central peak away from l=1300nm as the
distance from the wafer center increases (1=1260, 1225, and
1140nm, respectively) with a corresponding drop in the peak
reflectivity (0.75, 0.70 and 0.65, respectively).
The theoretical reflectivity spectrum (calculated using
a program based on equations (30) (35)) and the experimental
one for A=1300nm at 45 incidence are plotted in Figure 64.
The dashed curve is the experimental data with a baseline
which lies approximately 0.05 below the theoretical R=0.30.
As shown, the experimental and theoretical spectra agree quite
well for the forty-layer AlGaAs/GaAs stack.
A small piece of the wafer was cleaved and its back side
was polished to a near defect-free (mirror-like) surface. For
this sample, after a reflectivity scan was performed, the
detector was moved (see Figure 63) to respond to light
transmitted through the sample. As shown in Figure 59, theory
predicts that the position of the peak of the reflectivity and
valley of the transmission scans should occur at the same
wavelength. These two scans are plotted in Figure 65 and as
shown, the agreement with theory is quite good.
4.4 Electrical Testing
Part of the MBE 464 wafer was processed for electrical
testing purposes. A blanket zinc diffusion from the top to
the first p-layer was performed. Part of this zinc diffused
sample was characterized by the SIMS technique. Atomic zinc

Reflectivity
194
Figure 64: Theoretical (solid) and experimental (dashed)
reflectivity spectra taken from wafer center.

Reflectivity or Transmission
195
Figure 65: Reflectivity and transmission (dashed) spectra
taken near the edge of the wafer.

196
and aluminum profiles were measured to assess the abruptness
of the AlGaAs/GaAs heterostructure interface and to determine
the depth of the zinc diffusion. The profiles for the full
stack structure are shown in Figure 66. As shown, both the
heterojunction interface is abrupt and the A1 composition is
constant throughout the stack. The extent of zinc diffusion
can be more clearly seen in Figure 66. It appears that the
zinc diffusion ends in the second GaAs layer, hence creating
the desired PNPN (even number of junctions) type structure.
After the zinc diffusion was completed and confirmed by
SIMS, a Si3N4 dielectric coating was evaporated to a thickness
of 200nm. Fifty micron square devices were delineated by
etching mesas and ohmic contacts were made to the top layer.
The back side of the substrate was lapped down and polished
and a gold ohmic contact layer was evaporated onto this
surface. The whole PNPN structure was then tested.
Individual devices were electrically probed using a HP
4145B semiconductor analyzer. Current was measured as a
function of forward and reverse bias. Some devices exhibited
a shift from an "OFF" to "ON" state under forward bias
conditions when a voltage of 15 to 30 volts was applied. This
is the first reported switch for an electrical structure of
this size (forty layers) An I-V plot of a typical PNPN
device exhibiting bistability is shown in Figure 67. Also
plotted on the same scale is the resistance versus voltage.

CONCENTRATION (atoma/cc)
197
DEPTH (microns)
Figure 66: SIMS profiles of aluminum and zinc in MBE 464.

Current xIO3 (A)
198
Figure 67: I-V and R-V characteristics of a PNPN device
demonstrating bistability.
Resistance xIO5 (2)

199
As shown the current remains fairly low up to 12 volts, but
then begins to increase. At a current of 27mA (Vth= 12 volts)
the device turns "ON", negative resistance results when the
neighboring junctions interact and the voltage drops to 3
volts. If the current is held above ihold, equal to 30mA for
this device, the device will stay "ON", as shown. This
represents a relatively high power dissipation, which can
degrade the device performance by thermal drift and non-
uniform current injection.
A similar device was fabricated without the added zinc
diffusion step, giving a NPN (odd number of junctions) type
structure. As predicted by Katz[149] this structure does not
show bistability. A typical I-V and R-V of this device is
shown in Figure 68. The current is low, substantially less
than in the PNPN case. For this reason the NPN structure does
not appear to be suitable for RI modulation by the carrier
injection method.
Photoresponse measurements were performed on the PNPN
device by focusing the monochromator exit light onto some
electrically probed samples. The photoresponse measurement
was done at normal incidence (9=0) and the spectra were not
corrected with respect to the monochromator output. The
response was measured directly at the lock-in amplifier (see
Figure 63). The spectral response of the PNPN device used as
a detector has a large peak at 850nm due to absorption
occurring primarily in GaAs, as shown in Figure 69.

Current (jjA)
200
Voltage (V)
Figure 68: I-V and R-V characteristics of a NPN device.
Resistance xIO (&)

Response [V]
201
Figure 69: Photoresponse spectrum for a PNPN device.

202
The spectral photoresponse of the NPN device used as a
detector shows two peaks at 650 and 850nm (due to Al0 3Ga0 7As
and GaAs absorption, respectively) of opposite polarity (see
Figure 70) The relative intensity of the peaks varied
according to the position probed on the wafer. The absolute
intensity was also dependent on the device selected and on the
alignment of the optical system, but it was much lower than in
the PNPN device. It resembled the appearances of two opposing
detectors almost canceling each other's response. Sinc no
response was observed at longer wavelengths (1000-6000nm) as
expected and shown in Figure 69, the PNPN device has great
potential as a tuneable electro-optic device with either
electrical or optical control.
Using the RI results of Sell et al.[153] and Cross and
Adams[154], at typical electrical testing conditions (20 volts
through a 50 micron square device) a change in carrier
concentration from 1016 to I018cm'3 would result in a 0.8%
decrease in the refractive index of the narrow band GaAs.
This would result in a lOnm peak shift and a 30% drop in
intensity for reflected light at 1300nm. As indicated above,
the presently achieved high switching currents precluded
optical measurement of the expected spectral shift due to non-
uniform current distribution in the sample and heating
effects. One approach to the high switching current problem
is to redesign the PNPN junction stack to lower ihold.
Alternatively, all the p layers and n layers can be connected

Response [V]
203
Wavelength[nm]
Figure 70; Photoresponse spectrum for a NPN device

204
together respectively to allow parallel operation with RI
modulation capability, but no switching function.
4.2.5 Conclusions
A functional interference filter operating at a central
wavelength of A.=1300nm has been fabricated from the GaAs-
AlGaAs material system by MBE. The stack was doped alter
nately p- and n-type hence creating a novel optoelectronic
switch element. The I versus V characteristics were measured
and as predicted, only the PNPN device demonstrated electrical
bistability. As a detector, photocurrent measured for the
PNPN device peaked near the GaAs bandgap and changed direction
as a function of wavelength for the NPN device. As expected,
neither device showed any photoresponse for wavelengths
greater than lOOOnm.
The expected spectral shift of the reflectivity in
response to current modulation could not be verified because
of local thermal shifting of the characteristics resulting
from the relatively high holding current measured in these
unoptimized PNPN devices. Multiple quantum wells are also
being considered to take advantage of electric field effects.
The p-n junctions can also be operated in a parallel mode,
(with side contacts) circumventing the necessity to switch the
thyristor structure (inherent in the series mode) to the "ON
state in order to achieve RI modulation by carrier injection.

205
4.2.6 Addendum
Another structure, similar to the one of MBE 464, was
grown with the goal of reducing the number of electrical p-n
junctions in the interference stack. The actual layer
structure requested for growth MBE 572 is shown in Figure 71.
With this doping profile, the number of grown p-n junctions is
ten instead of twenty-one, as was the case for the as grown
MBE 464 sample. With less p-n junctions, the holding current
and threshold voltage for this thyristor could be much less
which could improve its switching characteristics.
A TEM photograph of a cleaved (110) and etched edge of
sample MBE 572 is shown in Figure 72. Similar to MBE 464, the
vertical period uniformity and interface abruptness appears to
be quite good. However, at 50,000 times magnification, the
AlGaAs/GaAs layer pair thickness was measured to be 172nm
which is considerably lower than the design value of 200nm.
This was later shown to not significantly affect the passive
optical device performance of this interference filter.
Reflectivity values were also measured on MBE 572 over the
wavelength range of 850-1600nm. The normalized reflectivity
data is plotted in Figure 73. Interestingly enough however,
even though the layer thicknesses according to TEM are off for
this structure, the central peak reflectivity is higher
(0.825) than it was for sample MBE 464. Perhaps the optical
alignment was better for this sample. The location of the
central peak maxima, 1250nm, is also extremely good.

206
9 X
1025 A
1050 A
GaAs:Si(n = 5xl017 cm'3 )
A13Ga 7 As:Si(n = 2xl017 cm 3 )
*
GaAs:Be(p = 2xl017 cm3 )
Al 3Ga-7As:Be(p = 2xl017 cm3 )
GaAs buffer
n+ GaAs
SUBSTRATE
Figure 71; MBE572 targeted structure.

207
Figure 72:
TEM micrograph of a cross-section of the stack
structure MBE572.

>REFLECTIVITY< L .1 /div]
208
Figure 73: Reflectivity spectrum for sample MBE572.

209
Part of sample MBE 572 was processed for electrical
testing purposes following the procedure previously outlined
for MBE 464. A blanket zinc diffusion from the top to the
first p-AlGaAs layer was performed and a SIMS profile of the
atomic zinc and aluminum concentrations is shown in Figure 74.
As shown, the zinc diffusion ends in the fourth layer which
should be p-type AlGaAs, and the aluminum concentration in the
AlGaAs layers is essentially constant. Further processing and
device testing of MBE 572 is planned. Unfortunately, this
investigator had to leave BNR to pursue other interests such
as MOCVD growth of III-V semiconductors.
4.3 MOCVD Grown GalnAsP Devices
Optical interference filters using the GalnAsP/InP
material system have been previously grown by the MOCVD
technigue[87,144] and also by CBE[143], but all devices were
passive. As of the time of this writing, GalnAsP/InP
interference filters with electrical properties, which could
be used as active optoelectronic switch elements, have never
been reported. Similar to GaAs, the RI of GaxIn1.xAsyP1.y is
affected by a change in carrier concentration. A change in
carrier concentration from I016cm'3 to I018cm'3 for GalnAsP (Xg
= 1400nm) using sub-bandgap light at a wavelength of 1550nm,
would result in a change of RI of Anquat = -0.43%[155]. For
a 30 layer quarter-wave thick structure of Ga 375In 625As 81P 19
and InP with RI estimates from Broberg and Lindgren[147] of

CONCEmTHATION (atomo/cc)
210
PROCESSED DATA bnw
9 Mov 88 Ca FILE: MSE572EX
Figure 74; SIMS profiles of zinc and aluminum from a portion
of the MBE572 stack structure.

211
nquat= 3.47 and nInP= 3.17, and using equations (31), (33) and
(36) a peak reflectivity of 0.9197 occurring at X = 1550nm is
expected. If the RI of the quaternary layer is reduced by An
= -.015, the stack will no longer be exactly a quarter-wave
(optically) thick. Hence, the new reflectivity peak will
occur at X = 1543nm and have a maximum value of 0.9090. A
theoretically predicted spectrum (neglecting GalnAsP and InP
absorption) for a 30 layer stack structure of alternating
1550nm quarter-wave thick layers of Ga 375In 625As 81P 19 and
InP is shown in Figure 75. If the central peak in Figure 75
was shifted by 7nm due to a change in the RI of the quaternary
layer, a maximum change of intensity of 30% for reflected
light at 1525nm (the steepest point of the peak) is expected.
This optical change can be electrically controlled hence
potentially creating a switching element.
A p-GalnAsP/n-InP interference filter structure was grown
by MOCVD. The layer structure and growth conditions of MOCVD
sample Q240 are shown in Figure 76. This sample was sent to
BNR for optical testing as a passive interference filter. The
reflectivity versus wavelength spectrum for sample Q240 is
plotted in Figure 77. As shown, the central peak occurs at a
wavelength of A0 = 1400nm. The detected reflectivity signal
was not normalized with respect to the detector's response for
this spectrum. The central wavelength is considerably
different from the expected value of 1550nm. This indicates
that either the growth rates, quaternary composition, RI

1.0
>1
p
H
>
H
P
0
Q)
rH
o.o
Wavelength (ran)
Figure 75: Theoretical reflectivity spectrum for sample Q240
(neglecting GalnAsP absorption).

213
15 X
Growth Conditions
Growth Temperature: 620C
Growth Pressure: 80 Ton-
Hydrogen Flowrate: 7 SLM
Metal Organics
Hydrides
MFTMm 0.737xl0'4
MFPH3 =211xl0'4
MFTEGa = 0.3 3 5x1 O'4
MFashb = 25x10 "
MFDE2n =0.621xl0'4
MFH2S = 0.07 lxl O'4
Figure 76; Layer structure and growth conditions for MOCVD
growth Q240.

REFLECTIVITY (A. U.)
214
WAVELENGTH (microns)
Figure 77; Experimental reflectivity spectrum for sample Q240.

215
estimates or a combination of all three are different from the
expected values used to determine the layer structure and
growth conditions. To determine the possible sources of
error, separate parts of the MOCVD sample Q240 were given for
both TEM and High Resolution XRD analysis. Until these
characterization results are available, no further MOCVD
growths or device processing and testing are planned for the
interference filter project.

CHAPTER V
CONCLUSIONS AND RECOMMENDATIONS
Epitaxial layers of InP, Ga^n^^s, and Ga^n^As P.,
have been grown lattice-matched to InP by MOCVD. However, the
electrical properties of undoped layers of these materials
grown in the quaternary MOCVD machine were not "state of the
art" or, in other words, as good as is possible considering
the quality of the sources presently available. The room
temperature Hall mobilities of undoped InP and Ga 47In 53As,
(on the order of 3000 and 6000 cm2/volt-sec, respectively) are
lower than the highest values reported in the literature
("4500 cm2/volt-sec for InP[55] and "10,000 cm2/volt-sec for
GaInAs[65]). The room temperature carrier concentrations for
undoped films are low though, n ~ 2*1014 cm'3, which suggests
that compensation by an incorporated acceptor impurity is a
cause for the inferior mobilities. The effect of growth
conditions, wafer preparation and sources used on material
quality have already been investigated as possible causes for
compensation. It is recommended that a phosphine purifier be
purchased and installed in the MOCVD system if high purity
InP, GalnAs and GalnAsP layers are required for future device
applications.
216

217
Another potential solution to the compensation problem is
associated with the substrate holding mechanism. Presently,
the substrate is placed on top of a quartz wafer tray during
the MOCVD growth. It is conceivable that this tray is the
source of compensating acceptors. The tray is cleaned and
dried using the same procedure as the reactor, but, unlike the
rest of the reactor, it is not baked for four hours at 950C
and kept in an air-free environment before a deposition. A
clean tray is used for each MOCVD growth, but, the sample is
heated by conduction through the quartz tray and part of the
exposed tray is directly upstream of the sample. It is
recommended that a silicon carbide coated graphite tray be
purchased and used to hold the sample during the MOCVD growths
of structures which require high purity undoped epitaxial
layers.
A model of the p-type doping process of MOCVD InP using
DEZn has been developed, yet never before reported. The
results of this study are that the dominant electrically
inactive point defects in zinc doped InP grown using high DEZn
reactor partial pressures are interstitial zinc, Zn,., and zinc
complexed with a phosphorous divacancy, VpZnInVp. Due to the
fact that very little information is available on the InP:Zn
material system, several assumptions were made during the
course of the development of this model. It is recommended
that future researchers perform extensive characterization on
undoped InP and InP:Zn, such as high temperature Hall effect

218
measurements and lattice parameter measurements. With this
added information, it would be possible to estimate more of
the unknown equilibrium constants which are required in the
model. This would improve the accuracy of the unknown
constants that are estimated by a non-linear regression
analysis program which uses the experimental SIMS and hole
concentration data. It is also recommended that other point
defect reactions be considered to validate or invalidate the
model which was presented in this dissertation. Mechanisms
such as the formation of a hydride complex as proposed by Cole
et al.[128] should be considered. In addition, the charged
state of all the point defects at room temperature is another
uncertainty that must be considered. The study of all the
possible doping mechanisms is hence proposed. Finally, it is
recommended that future researchers apply the point defect
analysis technique to understanding the cadmium and magnesium
doping data sets for MOCVD InP which are presented in this
dissertation. Cadmium appears to have an understandable
linear incorporation dependence on its precursor partial
pressure, but, magnesium has a dependency that suggests strong
compensation and possibly n-type conversion at high source
pressures.
A novel optoelectronic device incorporating multiple p-n
heterojunctions in an optical interference filter has been
presented. Potential applications for this device include an
electrically tuneable optical filter for optoelectronic

219
switching and a selective wavelength detector. Electrical
bistability was observed and passive optical interference was
demonstrated for this device. The combination of the two
device characteristics: a peak shift in the reflectivity
spectrum due to the application of an external bias, was not
observed. The expected spectral shift may have been observed
if the holding current was lower and/or if the current
distribution was more uniform. Consequently, optimized p-n-p-
n structures with fewer layers using multiple quantum wells
are suggested as improvements for future researchers. Also,
broad-area transparent contacts to the epilayer could improve
the current channeling problem.

APPENDIX A
THEORY AND OPERATION OF THE LOW TEMPERATURE HALL EFFECT SYSTEM
A.l Theory
If current is flowing in the x-direction in a probed
semiconducting sample and a magnetic field is being applied in
the z-direction, the flowing charged particles will be
deflected in the mutually perpendicular y-direction. To
maintain the steady-state flow of particles through the
device, an electric field will be induced in the y-direction.
The establishment of the electric field Ey, (as shown in
Figure 78) is known as the Hall effect[156].
By performing an experiment, like the one mentioned
above, with a known applied magnetic field (Bz) and current
density (Jx) one can measure the induced voltage (VH) From
these values, the proportionality constant:
Rh = Ey/(JX*BZ) (37)
can be calculated. With RH, the Hall coefficient, and a
previously measured value of resistivity, p, for the sample,
it is now straight forward to calculate the majority carrier
concentration and carrier mobility of the semiconductor.
Knowledge of these two temperature dependent values is a way
of characterizing the type (n or p) purity and functionality
of a semiconductor.
220

221
Figure 78; The Hall effect in a semiconductor.

222
A.2 Sample Preparation
In order to perform Hall effect measurements, a suitable
sample is needed. This semiconducting sample should be
expendable as the technique used is destructive. If a 2"
diameter wafer exists which needs characterization, a small .5
cm to .75 cm square piece will suffice as the working sample.
This sample can be obtained by first determining the crystal
growth plane which is usually indicated by a flat edge on the
circumference of the wafer. Next a line must be carefully
scribed in the wafer parallel or perpendicular to the flat
edge. The wafer can be broken or cleaved along the scribed
line. With one of the two pieces, this process should be
repeated until an approximately square working sample is
obtained. Since the sample can have maximum dimensions of
only 1 cm2, sufficient material should remain for additional
characterization. It has been reported[157-162] that crystal
orientation affects the Hall measurement results for all
anisotropic materials. Therefore, care should be taken to
establish a standard procedure to prevent misleading and
irreproduceable results.
After the test sample has been cut, make sure the
surfaces are clean. Some samples may require cleaning with a
degreasing solvent such as methanol and rinsing with distilled
deionized water, others may require chemical etching and
material specific techniques are described elsewhere[157].
Next, identify which surface has the deposited epitaxial layer

223
of interest. Depending on the type of sample, different
solders must be attached as close as possible to the four
corners of the wafer. For undoped GaAs and InP, pure indium
works well when cut and pressed onto the "epi" layer with a
clean surgical knife. The indium dots should also be as small
as possible to avoid introducing error. For other materials
such as doped III-V's, Si and Ge different solders are
suggested[157].
Now that the solder is in place, the sample should be
placed in the alloying station on the center of the graphite
heating strip. Check to see that there is a small air gap
between the thermocouple and the strip. Also be careful when
handling the strip as it is quite fragile. Put the "o"-ring
and plexiglass dome in place and tighten the mounting bolts in
a diagonal fashion. Do not tighten the bolts too much as the
dome is fragile.
The next step is probably the most important in the
alloying process: chamber evacuation and heating in a forming
gas (10% H2 in N2) environment. First open the regulator
valve on the forming gas cylinder so that there is about 10
psi of pressure on the flow regulator gauge. Close the line
to the rough pump and open the exit line of the chamber.
Place the exit line in a beaker two-thirds full of water for
a visual indication of the gas flow. Open the inlet valve
line to chamber so that the water in the beaker is bubbling to
the point of almost overflowing (increase cylinder flow if

224
necessary). Next close off the inlet and exit lines and open
the rough pump line. Plug in the rough pump and let it run
until the vacuum gauge reads -30 psi. Open the inlet line and
fill the chamber with forming gas. Then simultaneously close
the rough pump line and open the exit line. Let the gas flow
at the previous bubbling rate for about five minutes with the
hood closed. A flowmeter was attached to the system but it
added too much resistance to gas flow so it was removed.
Now that the chamber is nearly free of air, plug in the
digital thermometer (Omega, model 115KC) which should read
about 27C. At this point plug in the transformer (Signal,
model 10.5) turn it on, and set the dial to 55 (this system is
for GaAs with indium contacts). Close the hood and while
heating for about twenty minutes, make sure the gas flow rate
stays approximately constant. At this point the temperature
should be reading *230C (of course the sample is much hotter
than the thermocouple tip) and the transformer can be turned
off. Keep the forming gas flowing until the temperature reads
less than 50C consequently reducing the chance of forming an
oxide layer. Let the chamber cool down close to ambient
temperature before opening it and removing the sample.
The final step in the sample preparation process is the
testing of the nature of the contacts. The contacts have to
be of the Ohmic type to be used in the Hall effect experiment.
A simple check of contact integrity can be performed on a
conventional transistor curve tracer (one is available in the

225
microelectronics lab at the University of Florida). If all
possible combinations of contacts taken two at a time yield
linear current-voltage plots over a wide range of applied
voltage values, then, the contacts are ohmic. If they are not
all ohmic, then the alloying process must be improved by trial
and error.
Having prepared a square sample with ohmic contacts and
no visible wafer damage (cracks, oxide haze) then this sample
must be attached to an adequate sample holder. For routine
measurements (300 and 77 K) a patterned printed circuit board
with metal clips is available. Samples can be mounted on the
board and inserted into a dewer which rests on a wooden stand
in between the magnets two poles. The dewer could be filled
with liquid nitrogen or left empty depending upon the
temperature required. A sample holder was also developed for
temperature dependent experiments (down to 4.2 K) which
consists of three interconnecting parts: the CTI-Cryogenics
model SH-14R resistivity holder, an oxygen-free copper
conductive plate, and a gold coated alumina mounting plate.
The resistivity holder can be rotated so the semiconductor is
perpendicular to the magnetic field and is also made of oxygen
free copper. The copper plate was made to fit on one face of
the resistivity holder with a channel in it to hold the
alumina plate.
The gold covered alumina plate is the medium by which the
current and voltage leads could be attached to the four ohmic

226
contacts but also be isolated from each other. A one inch
square by .025" thick alumina plate with a 2E-06 cm thick
chromium adhesion layer and a 2E-05 cm thick gold conductive
layer was obtained from Materials Research Corporation. This
plate was cut into 3/8" by 1" rectangular pieces by using a
Micromatic precision watering machine equipped with a diamond
impregnated blade. Onto these plates a pattern of four pads
was drawn with a permanent ink marker. The plates were etched
in a gold etching solution of KI:I2:H20 in the respective
volumetric ratios 4:1:40 for ten minutes or until the gold was
removed. Then the plates were etched in a one part potassium
to 3 part ferrocyanide by volume solution until the gray
chromium layer was removed exposing the underlying non-
conductive alumina. The permanent ink was finally removed
with acetone revealing four isolated gold pads. To each of
these four pads were soldered a 4 cm piece of non-magnetic
wire with an Amphenol Incorporated female type plug connector.
To the center of the alumina mounting plate the small
semiconductor sample can be attached with a portion of Crycon
grease (which does not out-gas at low temperatures). The
grease keeps the wafer stuck in place to the alumina plate.
A very fine 1 mil thick gold wire must be ultrasonically bound
from each ohmic contact to the edge of the adjacent gold pad.
Request from the operator in the microelectronics lab that the
wires are as short as possible so that they can withstand
higher currents before burning up. This last step completes

227
the sample mounting and preparation stage of the Hall effect
experiment.
A.3 Experimental Hall Effect Measurements
After the sample has been mounted to the resistivity
holder and the holder attached to the cold finger of the
cryostat (CTI-Cryogenics, model 22), make sure that the
cryostat is perpendicular to the magnetic field. The wafer
should also face the right magnet pole labeled "North". Now
the wire from the bottom left gold pad should be connected to
the wire labelled C through a pin of the resistivity holder.
Similarly, the upper left to D, the upper right to E and the
lower right pad should be connected to wire F. These four
wires are wrapped around the cold finger and exit through the
cable connector port at the base of the cryostat. The letters
C, D, E, F are the pin labels on the male mating plug. These
four pins are connected to wires 4,3,2 and 1, respectively.
The cryostage has two lines filled with a pressurized
liquid helium attached to it which should not be bent to a
radius less than two feet (to avoid leaks). These lines with
the aid of the compressor (CTI-Cryogenics, model SC),
temperature controller (Palm Beach Cryophysics, Inc., model
4025), resistive heater (which is wrapped around the top of
the cold finger) and cryostat permit the measurement of Hall
voltages at temperatures as low as 4.2 K. An important note
about the cryostage is that whenever it is being handled, care

228
should be taken to avoid unnecessarily touching its internal
parts and of course disposable gloves must be worn.
Once the connections have been made to the resistivity
holder, the inner heat shield should be gently slipped over
the cold finger and attached to the cryostage. Next the outer
shield should be placed over the heat shield while making sure
that its "O" ring is in place. Now the rough pump should be
attached to the closed brass valve. If the helium supply and
return lines, the cryostat power line, and the ten pin mating
plug are attached then the cryostage is ready for operation.
The previously discussed four sample lines should be
connected to the binding posts labelled 1, 2, 3 and 4 of the
aluminum shielded switching box. The switching box contains
three double-pole double-throw switches, a four-pole six-throw
switch, a standard 100 ohm resistor and four other binding
posts for current and voltages leads. This box was built by
Mrs. Grazyna Palczewska, another member of our research group,
as per the design in reference[157]. It efficiently permits
the acquisition of Hall voltage data by a simple routine of
switching current flow through the sample and the standard
resistor. The standard resistor used should be of the same
magnitude as the resistance of the sample[157]. The constant
current supply (Lambda, model LQ531) must be attached to the
+1 and -I labelled binding posts. The magnitude of the
current used was always less than 5 mA which was low enough to
avoid resistive heating in the sample (as indicated by

229
constant signal values). The voltmeter used was actually a
high-quality high-input impedance electrometer (Keithley
Instruments, model 614) to avoid any leakage currents and
consequently increased error.
Now that all electrical connections are made, the
temperature desired for measurement must be reached. This can
be room temperature (300 K) or lower, but the system can not
be operated above room temperature as the cryostage contains
indium seals which would melt. For temperatures below
ambient, the temperature controller and compressor must be
used. Basically the rough pump is first plugged in and run
until the pressure gauge reads 1-2 torr. At this point close
the brass valve, unplug the rough pump and slowly vent the
pump and hose to 760 torr. Now turn on first the compressor,
then the cold head (on the compressor) and then, last, the
temperature controller. The controller must be programmed
with its optimum process parameters as found and explained in
its manual. It was advised that the heater plugs always be
disconnected whenever the temperature readout exceeded 275 K.
When the combination of helium cooling and resistive heating
is operating correctly in a properly evacuated environment,
temperatures below 100 K should be attained and stable in
about 45 minutes. If it is not thermally stable or the
outside of the cryoshield is cold with condensation occurring
then either the chamber was not properly evacuated or there is
a leak in the cryostage.

230
It was suggested that a test be performed to determine if
the readout temperature is a good representation of the actual
temperature of the sample. This was performed by placing a
second silicon diode sensor, with the same I/V characteristics
as the cold finger sensor, at the exact sample position. Its
four plugs were connected temporarily to four leads from the
37 pin mating plug of the temperature controller. Since the
controller is capable of displaying the temperature at both
sensor locations, this experiment was fairly straight forward.
The results of the test are that it takes approximately
fifteen minutes for the sample location to stably reach the
same temperature as the permanent sensor location.
Once the desired temperature for measurement has been
reached and is stable, the magnetic field should be applied.
To do this, the constant current supply (Walker Scientific,
model HS525) for the magnet (Walker Scientific, model HV4H)
should be turned on by pressing the power button. Next the
cooling water must be turned on to a flow rate great enough to
illuminate the "DC-OFF switch (this indicates adequate flow
through the current supply and magnet). Of course make sure
that the exit line for the water is in the drain and there are
no leaks. Also, make sure that the spacing between the magnet
poles is adequate and that the poles are locked in place.
Now, with the percentage of maximum current dial reading zero,
press the "DC-ON" switch which will activate the current
generator. At this point it is safe to slowly dial in a

231
percentage of maximum current. For example, with a 2" pole
spacing at 50% of current and with the reversal switch
illuminated, a magnetic field of about 2600 gauss should be
measurable with the previously zeroed gaussmeter (Jobmaster
Magnets, model 6MIA). The magnetic field will be positive in
the sample when it is facing the pole llxelled "North and the
"DC-ON" switch is illuminated (the "Reversal" switch is not) .
A calibration curve for the particular pole spacing used
should be obtained on a regular basis. Operating at higher
magnetic fields induces higher Hall voltages and is therefore
desired. The sample should always be placed as close to the
center of the magnetic field and as perpendicular to its field
lines as possible.
With the sample in the proper place (shielded from
light), at the right temperature, with a stable low current
passing through it, with the voltmeter and magnet on, record
the Hall voltage value V5(+I,+B) ,f or example, with the switch
in position 5. Then by varying the current (+1,-1), magnetic
field (+B,-B) and position switch (5,6,S) record twelve
voltage values (S stands for the standard resistor reference
voltage). Positions 5 and 6 pass current diagonally through
the samples and consequently measure voltage across the other
respective diagonal. Now with the magnetic field off, put the
switch in positions 1, 2, 3, 4 and S and vary the current (+1,
-I) and record the ten resistivity voltage values Vi.

232
A.4 Calculations
Now that the twenty-two voltage measurements have been
recorded, for a sample of known thickness (the thickness of an
epitaxial layer when growth is on a SI substrate) and with a
value of resistance for the standard resistor used the
equations of reference [157] can be used to calculated the
resistivity of the semiconductor (p in ohm-cm) and the Hall
coefficient (RH in cm3/coulomb), which is negative for n-type
and positive for p-type samples. Then with a known Hall
proportionality factor (r = 1 for GaAs) the charged carrier
concentration (n or p) can be calculated (/i in cm2/volt-sec)
can be calculated as /i=RH/r. These three values (p, /i and n
or p) are all temperature dependant and provide a good basis
of comparison in determining the quality of a grown epitaxial
film.
For investigators who are interested in extremely
accurate values of p, ju and n, several references[159,161-163]
present correlations for samples of non-ideal geometries and
finite contact sizes. It is advised that square samples have
a perimeter length (Lp) less than 1.5 cm and with thicknesses
less than 0.1 cm. Also, the contacts should be as close as
possible to the corners (or edges) of the sample and of size
less than 0.01*Lp. This author advises that if errors of 10%
are tolerable then using the previously mentioned correlations
are not necessary.

233
A.5 Conclusions
The Hall effect has been used for many years as a way of
characterizing semiconductor devices. It is a reliable,
reproducible method with the only real drawback being its
difficult, destructive contacting procedure. For both an InP
and GaAs samples which were grown and characterized by an
industrial lab, the measured mobility, carrier concentration
and resistivity values differed by at most 10% from the
company's measured results. This shows that the described
contact procedure, measurement routine and calculation method
should be adequate for most applications.

APPENDIX B
OPERATION OF THE QUATERNARY MOCVD SYSTEM
B.l SCOPE
This document pertains to the quaternary MOCVD
housed in building 771 on the campus of the University of
Florida.
B.2 PURPOSE
The purpose of this document is to outline the correct/
safe operating procedures for the MOCVD system which is used
for research on optoelectronic materials.
B.3 SAFETY
No one should operate any part of this system without
prior approval of, and training by Dr. Tim Anderson. The
system has its own built-in safety system which will shut down
upon detecting any potentially dangerous situation. If
something related to the system seems to be malfunctioning,
and you are not trained to operate the system, immediately
(hit the red emergency stop, advise others to leave the room
with you), and contact the present operator, C. Heinz, Dr. T.
Anderson (392-2591) or K. Rambo).
If a problem exists in the room such as a fire, also hit
the red emergency stop button. Of course, if the reactor
234

235
explodes or you suspect a hazardous gas leak, do not wait for
the MDA to turn on the whooper leave the Surge building
immediately. Do not return to the room until the emergency
response team has entered the room with their Scott Air packs
and solved the problem. Bottle changes, reactor tube removal
and cleaning and pump oil changes should be done using the
buddy system; with the dual-cylinder air system. No growth
shall be performed unless at least 2 people are in the
building. Always wear safety glasses while in the clean room.
All operators should take part in the arsenic in urine
monitoring. If you ever smell a garlic-like or rotten fish
odor or sense your heart rate increase above normal (for no
apparent reason), leave the room. Never start a deposition or
use the bubbler baths during stormy weather. Always inform
the building director that you are going into the lab and if
you are planning to do a growth. The MOCVD has several
sensors which light-up, flash or cause a buzzer to sound on
the display panel. It has three levels of warning and
corresponding actions which result depending on the level of
danger to the system or operator. (see pages 25-29 of
Hayakawa[164]) .
B.4 PREPARATION
Check with the building operator to see if anything is
abnormal today with regards to the nitrogen, hydrogen, chilled
water, dry air, building power supply, and air scrubbing
monitoring. Check the status of the Advanced Concepts wet

236
scrubber on the back pad by turning it on. Is, the pH > 10,
ORP > 100, pumping pressure > 28 psi, liquid level sufficient?
If not, see scrubber manual for maintenance. Check the
pressure on the inlet H2 and N2 lines, the hydride cylinder
sources. Do we need a bottle changes or six-pack hydrogen
changes? Check the water filter on the chilled water line,
clean if necessary. Check the temperature of the metal
organic source baths. Does growth rate or doping level
changes indicate the need for a source change? Is the
hydrogen purifier operating properly? Is the reactor tube
clean and dry and in place? Leak tight seal? Purged with
H2. Has all source changes or maintenance been followed up by
a helium leak check? Did the whole system pass the leak
checking procedure? If any seal has been broken on a toxic
gas line, wear air masks during first operation of that line
after leak checking. How is the oil level in the roughing
pumps? Does the oil need changing? Change it before and
after growths. Is there a clog? Have you programmed the
process controller with your desired deposition scheme? Test
it. (Programming tips are available in Hayakawa[164]).
B.5 PROCEDURE
B.5.1 Introduction
Now that all the facilities are in place, all safety
precautions have been followed/checked, and the materials are
installed, the normal operation procedure will be outlined.
It is assumed that the system has already gone through its

237
initial start up (leak check, bake out of graphite susceptor,
scrubber N2 purge, individual components testing). It is also
assumed that the reader is familiar with the physical layout
of the system (see Figure 79) and what the individual system
components do. Manual valves (MV), air valves (AV), needle
valves (NV) and regulator valves (RV) are all mentioned in the
procedure (also see Figure 79).
B.5.1 Power Supply and Programmable Controller Explanation
Turn on the 115V and 208V circuit breakers. Turn on
breakers ELB 100 and 200. After the 115 and 208V lights are
illuminated, turn on fuse breakers NFB102, 112 and circuit
protector CP132 which are all located inside the lower right
electrical panel (when facing the front of the machine). Push
the "stand by", "PC ON" then "ON" switches and then all
switches, readouts and emergency sensors will begin to
function. If the "ON" switch is not lit, the system will
function as it stands, but all air operated valves will remain
open (red light) or closed (no light) even if they are pushed
accidentally. The system should be in manual mode now. For
a description of edit, auto modes or how to operate the
programmable controller, see Hayakawa[164]. Check reactor
pressure, if not slightly greater than or equal to 760 torr,
turn on RP2 and open valves from RP2 inlet towards the reactor
to prevent back flow to the reactor. Completely evacuate the
system, start hydrogen flow and then pressurize the system by
closing AV208,9 and then opening AV204 when p = 780 torr.

238
Figure 79: Quaternary MOCVD flow diagram (complete)

239
B.5.2 Definition of the "OFF" or "OVERNIGHT State
The complete system should be purging through valves
AV200, 204 at 760 torr with hydrogen. All source MVs should
be closed (metal organics, hydrides) and AV111 (nitrogen
inlet) should also be closed. Roughing pumps and turbo
molecular pumps should be off. The scrubber should be off,
Lepel RF generator also off. Water flow to Lepel, reactor and
load lock and to bubblers also all off. Hydrogen purifier
should always be on. Close MV14, 24, 34, as they should
always be during toxic gas flow (double check).
B.5.3 How to Grow InP on InP in the MOCVD System
Turn on the scrubber, and check the hydrogen supply on
back pad. Enter the clean room and check the MDA and hydrogen
detector. Flow water to the system, reset water alarm, and
make the deposition program now. Turn on the MO bubbler
controllers and set the temperatures. Turn off the variacs
for the heat tapes. Prepare the substrates. Now put the
substrates into the load-lock room. Evacuate the load-lock
room and close MV102, AV109, then turn RPI and RP2 on, also
open MV201, MV202, NV202 and evacuate to minimum pressure.
Flow N2 gas into the load-lock room and close MV201,
MV202, NV202, then open MV102, MV202, NV202. When the
pressure of the load-lock room becomes 760 torr, close NV202,
MV202, MV102. Load the substrates into the load-lock room
using clean room paper and gloves, close the door tightly.
Evacuate the load-lock room by opening MV201, MV202, NV202

240
slowly. Flow H2 gas into the load-lock room by closing MV201,
MV202, NV202 and opening AV109, MV202, NV202 and when the
pressure of the load-lock room becomes 760 torr, close NV202,
MV202, AV109. Evacuate the load-lock two more times.
Insert the substrates to the reactor from the load-lock
room (fork operation). The pressure of the reactor should be
760 torr. Open MV200 and turn the fork manipulator on. Open
the fork gate valve. Open the shutter valve. Move the fork
forward and put the substrates on the graphite susceptor and
then remove the fork. Close the shutter valve. Close the
gate valve. Close MV200. Turn off the fork manipulator and
double check that MV200 is closed!
Prepare for flowing toxic gases (PH3 Only for InP on InP)
by first closing MV20, MV23, AV204 and then quickly evacuate
the system through AV209 with RP2 on. Increase the set point
of MFC20 to maximum. When MFC20 indicates almost zero seem,
close AV21, AV22 and AV23 (leave AV20 open). Adjust MFC20 to
about 10 seem (set point). Close the regulator (RV20) (turn
it fully counter-clockwise).
Heat the RF filament (power on, solenoid on, RF filament
on) set the pressure of the reactor to 80 torr. Increase the
main hydrogen flow rates to the reactor to 7 slm (total) and
to each vent line 0.5 slm also increase the H2 regulator to 90
psi. Set all other flow rates to overnight flows except the
PH3 compensation line (set to 100 seem) TMIn, and its
compensation line (both set to 143 seem) Adjust NV60 so PI60

241
reads 490 torr (indium bubbler), open AV207, adjust NV203 so
the reactor is at 80 torr. Flow the MO to the vent (TMIn
only). Make sure AV63 is open and AV64 is closed. Adjust the
pressure of the bubblers (PI60 = 490 torr) Open the MO
bubbler (at first open the outlet valve and then open the
inlet valve). Close AV62 and then check the pressure of the
bubbler, set it to be 500 torr and stable.
Heat the substrates by setting the controller temperature
to 650C (172R) Turn on the RF plate and power switches
(auto mode) Flow the toxic gas PH3 to the reactor for 10
minutes. Check H2 detector point 8 and constantly monitor it.
Open the PH3 cylinder slightly and adjust the outlet pressure
to 0.5 kgf/cm2 (use the black gloves and keep all machine
doors completely closed until all PH3 is gone) Open AV22 and
then open AV21. Set MFC20 to maximum (300 seem). Check H2
detector point 6 then return it to auto. Start the program in
10 minutes if the temperature is stable. Close the PH3
cylinder at an appropriate time (this depends on flow rate and
cylinder pressure), possibly before the end of the growth.
Close the MO bubbler after the growth ends by first opening
AV62 and then close the inlet valve and last, close the outlet
valve. Reduce the flow rate of the MO to 30 seem. End the
program always 3 minutes after the MO goes to the vent. Turn
off the RF power, plate switches and filament of the RF
generator. When the temperature of the substrate is less than
400C (100R), turn off the RF generator power switch. When

242
the temperature of the substrate is less than 200C (50R) ,
flow PH3 gas to the vent. When the flow rate of the PH3
becomes 0 seem, purge the PH3 line five times by opening MV20
and MV23 and then close MV23, and wait until MFC20 becomes 0
seem. Repeat this four more times and reduce MFC20 to 30
seem, also make a note of the time.
Pressurize the reactor to 760 torr. Close AV207 and
AV209, and turn off RP2. When the reactor pressure becomes
780 torr, open AV204. Remove the sample from the reactor (30
minutes after the time noted previously) by fork operation.
Remove the samples from the load-lock room by first evacuating
it. Close MV102, AV109, and open MV201, MV202, NV202. Flow
Nz gas into the load-lock room. Close MV201, MV202, NV202 and
open MV102, MV202, NV202. When the pressure of the load-lock
room becomes 760 torr, close NV202, MV202, MV102. Repeat the
evacuation and fill steps three more times then remove the
substrates from the load-lock room. Repeat evacuation and
fill steps one more time, then turn off RP1. Reduce the H2
regulator to 40 psi, bleed line to 100, make all flows 0.2 slm
or 30 seem.
Wash the substrate supporter using aqua regia (3 HC1:1
HN03) for 10 minutes then rinse with DI for one minute then
dip in the HF mix (5 DI:1 HF) for 5 minutes. Follow with a DI
rinse for 15 minutes. Turn off the water flow (to the MOCVD
system), silence the water alarm, open the water pressure
relief into the plastic tray. Turn on the variacs for the

243
heat tapes and exit the clean room. Turn off the scrubber,
check H2 supply pressures (500 psi minimum).
B.5.4 Reactor Cleaning
Purge the reactor overnight with hydrogen after a four
hour bake at 900C and 35 torr (4 slm Hydrogen flowing to the
reactor) Disconnect the water line (inlet and outlet) close
all water valves. Divert gas flow by closing AV119, 125, 200
then opening AV118, 124 and turn off the heat tapes.
Disconnect the thermocouple cable connector. Evacuate the
reactor using RP1 through the load-lock to 3 torr. Close the
doors and open slowly MV207 to let air into the reactor to a
pressure of 760 torr (there is a chance that a fire may occur
in the reactor tube). Repeat evacuation and fill four more
times.
Using the particle masks, disconnect the reactor and
thermocouple. Cover all with aluminum foil. Clean the MO
inlet tube, susceptor bed and reactor immediately, using the
wafer tray procedure (reactor requires 2 hours of DI rinsing
though). Clean the thermocouple tube using methanol, clean
room paper and a clean razor blade. Clean the exhaust area
(wearing particle mask) with methanol and clean room paper.
Dispose of waste in toxic waste garbage can (fill out proper
forms).
Install the clean thermocouple tube, and reactor with
clean graphite susceptor and susceptor bed. Before connecting

244
the inlet flange to the reactor, check the position of the
susceptor using the fork procedure and a clean wafer tray.
Adjust the position if necessary. Connect the inlet flange to
the reactor using fresh VCR gaskets. Evacuate the reactor to
3 torr as an initial leak check using RP1. Helium leak check
the reactor.
Flow H2 to the reactor and use the Matheson H2 detector
at all connections. Connect the water lines using tie-wraps.
Bake the reactor at 35 torr, 900C, 4 SLM H2 flow for four
hours. Turn on the heat tapes, too. Pressurize the reactor
to 760 torr, set all flows to 0.2 SLM or 30 seem.
B.5.5 Metal Organic Source Change (TMIn as an example)
Purge the MO line overnight with 300 seem H2. Close all
air valves on the system by hitting the red emergency stop
button, and turn off the heat tapes. Evacuate the MO line
using RP2 (open AV209, 202, 65, 63, 62, 61, 60 and NV60
completely in this order). Open AV111 to let N2 flow to the
MO. Close NV60 until PI60 reads 900 torr then open until PI60
reads 70 torr. Repeat this ten times, then put the bubbler at
760 torr by adjusting NV60. Close all air valves. Disconnect
the bubbler using particle masks and clean room gloves (make
sure bubbler valves are real tight) Install the new bubbler
using fresh VCR gaskets. Repeat purge and evacuation steps
five times. Evacuate the whole system using RP2 and helium
leak check the bubbler connections through MV207. Pressurize
the system and return all flows to 0.2 slm or 30 seem. Turn

245
off RP2. Use the Matheson H2 detector around the bubbler with
the bubbler at 900 torr and hydrogen flowing. Turn on the
heat tapes. Flow MO to the vent for one hour before use in a
growth to remove volatile impurities.
Bi5.6 Hydride Source Change (PH3 as an example)
Close the gas cylinders tightly (using the black gloves) .
Purge the gas lines with H2 (five times evacuation and fill
up) then let H2 flow overnight. Close AV21, AV22, AV23, AV20
and pressurize the gas lines with cylinder N2 gas and then
release slowly through MV41 and RV40 to the exhaust (turn on
RP1 and RP2 and close exhaust AVs) Repeat this step four
more times then leave the line at 760 torr and close MV23.
Prepare the new cylinder (and gasket which is necessary on H2S
cylinder only) and clean room gloves and wrenches. Put on the
gas mask in a buddy system, set the MDA to sense at QMOCVD
only, and seal off the clean room from normal access. Take
out the old gas cylinder. Put in the new gas cylinder (use
the gasket).
Pressurize the gas lines with N2 gas up to AV20. Check
for leaks with soapy water. Release the N2 pressure through
MV41 and RV40 (prevent air from entering the rest of the
system) If there are no leaks then remove air masks, put MDA
on auto and allow others to enter the clean room. Pressurize
the gas line and then release ten times with N2. Evacuate the
gas lines, and then check for leaks with helium detector.
Pressurize the line overnight with N2 and check for any

246
pressure drop (note room temperature). Check for leaks with
the Matheson detector by pressurizing to 200 psi with H2 and
open/close all MV's near the cylinder. Allow H2 to flow at
normal conditions.
Wear air masks and evacuate others from clean room during
the first use of new cylinder (before a growth), flow gas to
vent for one hour to remove volatile impurities (of course
turn scrubber on) The cylinder is now ready for routine use,
keep cylinder cabinet doors closed at all times.
B.6 Return from an Emergency Shutdown
After problem has been solved by the team wearing air
packs, if necessary, hit "alarm reset/"BZ reset" to stop
buzzer and alarm light. Evacuate and fill the system with H2
if necessary, five times. Do a complete "preparation" check
(see section B.4). Follow the turn-on procedure at section
B.5.2.
B.7 SHUTDOWN
Return to the "OFF" state is defined in the previous
section. For a complete shutdown, the system should be
completely purged with N2 for 1 hour (all lines up to a closed
source manual valves closed, that is) Now, all three
temperature baths should be shut off. Close the UHP H2 outlet
valve and evacuate the whole system to 1 torr. Then close
MV100, AVI10 and hit the red emergency stop button. Turn off
the heat tapes and skinner valve. Set the Hz purifier
temperature to 0C and when T < 200C, shut off the purifier,

247
pressurize it with N2 and close all four MVs. Turn off the
Lepel RF generator's circuit breaker. Purge the rough pumps
with N2 also. Close the nitrogen inlet, dry air and cooling
water inlet and outlet valves. Turn off electric breakers
ELB100, 200, and then the MOCVD breaker. The system should
now be completely off and ready for a long vacation or major
renovation.
B.8 MAINTENANCE
General system maintenance is done as needed. Of course,
occasionally vacuum out the glove box for dust removal and
wipe down the system with a "clean room" detergent. All other
maintenance steps are mentioned as either part of preparation
or procedure steps.

APPENDIX C
DOPANT MODEL COMPUTER PROGRAM
As stated in section 3.3.4, a non-linear regression
analysis program based on the Marquardt method was used to
determine values for the equilibrium constants B^Bj, Bs-B7.
The program is not listed here due to its length (which is
approximately 40 pages) The program was written by Dr. Alkis
Constantinides and is available on a diskette which comes with
his book, Applied Numerical Methods with Personal Computers
[138]. A description of how to use the program is given in
section 7-5 of the book. Briefly, regression analysis is the
application of mathematical and statistical methods for the
analysis of experimental data, and the fitting of mathematical
models to these data by the estimation of the unknown
parameters of the models. By performing statistical tests,
the model can be identified or verified. The Marquardt method
uses an interpolation technique which is a combination of the
Gauss-Newton and the steepest-descent methods to obtain values
of the parameters in the model which minimize the overall
(weighted) sum of the squared residuals.
The non-linear regression program is menu-driven for data
input and adjustment. To use the program, one must first
derive a model and variational equations. The variational
equations are obtained by taking partial derivatives of the
248

249
model equation with respect to each unknown model parameter.
The program was first run to solve for the constants Bv B2,
Bj by fitting the hole concentration model equation (27) to
the hole concentration data. The model and variational
equations used for this are:
Y(1)=B(1)*(2.36E-06)*X*(((B(2)A2)+B(3)*B(2) /
B(4)/0.00236)A0.5) / ((B(4) + B(1)*B(2)*
(2.36E-06) *X)A0.5) (38)
Y(2)=(2.36E06)*X (((B(2)A2)+B(3)*B(2)/B(4) /
0.00236)A0.5) (B(4)+0.5 B(l)*
B(2) (2.36E-06)*X) / ((B(4)+B(1)*B(2) *
(2.36E-06)*X)Al.5) (39)
Y(3)=B(1)*(2.36E-06) X*(B(2)*B(4) + 0.5*B(1)*
(B(2)A2) (2.36E-06)*X + 0.5*B(3)/0.00236) /
((((B(2)A2)+B(3)*B(2) / B(4)/0.00236)A0.5) *
((B(4)+B(1)*B(2) (2.36E-06) *X)A(1.5))) (40)
Y(4)=(0.5*B(1)*B(2) (2.36E-06) X/B(4)/0.00236) /
(((B(4)+B(l) B(2)*(2.36E-06)*X)A0.5) *
(((B(2)A2) + B(3)*B(2) / B(4)/0.00236)A0.5)) (41)
Y(5)=(-B(1)*B(2) (2.36E-06)*X) ((B(3)/B(4) /
0.00236) + (0.5*B(1)*B(2) B(3)*(2.36E-06)
*X/0.00236/(B(4)A2)) + 0.5*B(2)) / (((B(4)+B(l)
* B(2)*(2.36E-06) X)A(1.5)) (((B(2)A2)+B(2)
* B(3)/B(4) / 0.00236)A0.5)) (42)

250
where:
Y(l) = [p]RT
(43)
Y (2)
'[P]*T
-I3!T¡
(44)
Y(3)
*5[P]rt
*c5B2
(45)
Y (4)
5CP]rt
dBj
(46)
Y (5)
[PRT
5B4
(47)
The resulting values of the best fit are: 6^375.3 mole
fractionVatm, B2=2.5 mole fraction'1 atm'1/4, and Bj^l.23 10'18
mole fraction.
The non-linear regression program was executed a second
time to solve for the constants B5 B7 by fitting the SIMS
concentration model equation (28) to the SIMS concentration
data. The model and variational equations used for the second
program execution are:
Y(1)=B(1)*B(2) 2.36E-06*X ((1+(B(3)/l.0E+31 /
B(2) / 0.00236))A0.5) / ((1.0E+31 + B(1)*B(2) *
2.36E06*X)A0.5) (48)

251
Y(2)=(1.OE+31 + 0.5 B(l)*B(2) 2.36E-06*X) *
((1.0+(B(3)/B(2) / 1.0E+31/0.00236))A0.5) *
B(2)*2.36E-06*X / ((1.OE+31 + B(1)*B(2) *
2.36E-06*X)Al.5)
Y(3)=0.5*B(1) 2.36E-06*X ((2*B(2)+(B(3) /
1.0E+31/0.00236)) / ((B(2)A2 + (B(3)*B(2) /
1.0E+31/0.00236))A0.5) B(l) 2.36E-06 X
((B(2)A2 + (B(3)*B(2) / 1.0E+31/0.00236))A0.5
/ (1.OE+31 + B(l)*B(2) 2.36E-06*X))
Y(4)=B(1)*0.001 X*0.5 / 1.OE+31 / ((1.0+(B(3) /
1.0E+31/B(2) / 0.00236))A0.5) / ((1.0E+31
+ B(1) B(2)*2.36E-06 X)A0.5)
where:
Y (1) =-£[Zn]
Y(2) 'aE^Zn]
v -B5
Y(3) = ,^[BZ6n]-
Y (4) =
The resulting values of the best fit are: B5=2708
fraction/atm, B6=1003 mole fraction'1, and B7=11.6 atm1/4
(49)
*
)
(50)
(51)
(52)
(53)
(54)
(55)
mole

REFERENCES
[1] Streetman, B. G., Solid State Electronic Devices. 2nd.
Ed., Prentice-Hall, Englewood Cliffs, NJ, 1980.
[2] Kuphal, E., J. Cryst. Growth 54 (1981) 117.
[3] Bocchi, C., Ferrari, C., Franzosi, P., Fornuto, G.,
Pellegino, S., and Taiarol, F., J. Electron. Mater. 16
(1987) 245.
[4] Lyons, M. H., Faktor, M. M., and Moss, R. H., J. Cryst.
Growth 66 (1984) 269.
[5] Olsen, G. H. and Zamerowski, T. J., IEEE J. of Quantum
Electron. OE-17 (1981) 128.
[6] Ludowise. M. J., J. Appl. Phys. 58 (1985) R31.
[7] Cho, A. Y., J. Vac. Sci. Technol. 16 (1979) 275.
[8] Bedair, S. M., Tischler, M. A., and El-Masry, N., Appl.
Phys. Lett. 47 (1985) 51.
[9] Tsang, W. T., Appl. Phys. Lett. 45 (1984) 1234.
[10] Chen, W. K., Chen, J. C., Anthony, L. and Liu, P. L.,
Appl. Phys. Lett. 55 (1989) 987.
[11] Manasevit, H. M., Appl. Phys. Lett. 12 (1968) 156.
[12] Manasevit, H. M. and Simpson, W. I., J. Electrochem.
Soc. 116 (1969) 1725.
[13] Manasevit. H. M., J. Electrochem. Soc. 118 (1971) 647.
[14] Manasevit, H. M., Erdmann, F. M. and Simpson, W. I., J.
Electrochem. Soc. 118 (1971) 1864.
[15] Manasevit, H. M. and Thorsen, A. C., J. Electrochem.
Soc. 119 (1972) 99.
[16] Manasevit, H. M. and Simpson, W. I., J. Electrochem.
Soc. 120. (1973) 135.
[17] Bass, S. J., J. Cryst. Growth 11 (1975) 172.
252

253
[18] Bass, S. J., J. Cryst. Growth M (1978) 29.
[19] Dupuis, R. D. and Dapkus, P. D., Appl. Phys. Lett. 31
(1977) 839.
[20] Jacko, M. G. and Price, S. J. W., Can. J. Chem. 41
(1964) 1198.
[21] Haigh, J. and O'Brien, S., J. Cryst. Growth 67 (1984)
75.
[22] Ban, V. S., J. Electrochem. Soc. 125 (1978) 317.
[23] Hayafuji, N., Mizuguchi, K., Ochi, S. and Murotani, T.,
J. Cryst. Growth 77 (1986) 281.
[24] Leys, M. R., Van Opdorp, C., Viegers, M. P. A. and
Talen-van der Mheen, H. J., J. Cryst. Growth 68 (1984)
431.
[25] Wang, C. A., Groves, S. H., Palmateer, S. C., Weyburne,
D. W. and Brown, R. A., J. Cryst. Growth 77 (1986) 136.
[26] Puetz, N., Hillier, G. and SpringThorpe, A. J., J.
Electron. Mater. 17 (1988) 136.
[27] Komeno, J., Tanaka, H., Itoh, H., Ohori, T., Takikawa,
M. and Kasai, K. 1987 Electronic Materials Conf., Santa
Barbara, CA, June 24-26, Paper E-8.
[28] Blaauw, C. and Miner, C. J., J. Cryst. Growth 84 (1987)
191.
[29] Eliot, B., Balma, F. and Johnson, F., Solid State Tech.
33 (1990) 89.
[30] Chen, C. H., Cao, D. S. and Stringfellow, G. B., J.
Electron. Mater. 12 (1988) 67.
[31] Johnson, E., Tsui, R., Convey, D., Mellen, N. and
Curless, J., J. Cryst. Growth 68. (1984) 497.
[32] Hess, K. L. and Riccio, R. J., J. Cryst. Growth, 77
(1986) 95.
[33] Rai-Choudhury, P., J. Electrochem. Soc. 116 (1969)
1745.
[34] Baliga, B. J. and Ghandi, S. K., J. Electrochem. Soc.
122 (1975) 683.

254
[35] Manasevit, H. M., Dapkus, P. D., Ruth, R. P., Yank, J.
J., Campbell, A. G., Johnson, R. E. and Moudy, L. A.,
1978 Photovoltaic Spec. Conf., Washington, DC (New York,
NY, IEEE).
[36] Duchemin, J. P., Bonnet, M., Beuchet, G. and Koelsch,
F., Inst. Phys. Conf. Ser. 45 (1979) 10.
[37] Yoshino, J., Iwamoto, T. and Kuhimoto, H., J. Crystal
Growth 55 (1981) 74.
[38] Razeghi, M., Poisson, M. A., Larivain, J. P. and
Duchemin, J. P., J. Electron. Mater. 12. (1983) 371.
[39] Renz, H., Weidlein, J., Benz, K. W. and Pilkuhn, M. H.,
Electron. Lett. 16 (1980) 228.
[40] Moss, R. H. and Evans, J. S., J. Cryst. Growth 55
(1981) 129.
[41] Sacilotti, M., Mircea, A. and Azoulay, R., J. Cryst.
Growth 68 (1983) 111.
[42] Hsu, C. C., Cohen, R. M. and Stringfellow, G. B., J.
Cryst. Growth 63. (1983) 8.
[43] Bass, S. J., Pickering, C. and Young, M. L., J. Cryst.
Growth M (1983) 68.
[44] Duchemin, J. P., J. Vac. Sci. Technol. 18. (1981) 753.
[45] Razeghi, M. and Duchemin, J. P., J. Cryst. Growth 64
(1983) 76.
[46] Eguchi, K., Ohba, Y., Kushibe, M., Funamizu, M. and
Nakanisi, T., J. Cryst. Growth 93 (1988) 88.
[47] Kasemset, D., Hess, K. L., Mohammed, K. and Merz, J. L.,
J. Electron. Mater. 13 (1984) 655.
[48] Scott, M. D., Norman, A. G. and Bradley, R. R., J.
Cryst. Growth 68 (1984) 319.
[49] Knight, D. G., Miner, C. J. and SpringThorpe, A. J.,
1990 Materials Res. Soc. Spring Meeting, San Francisco,
CA, Paper F4.5.
[50] Tuck, B. and Baker, A. J., Jour. Mater. Sci. 8 (1973)
1559.
[51] Nishitani, Y. and Kotani, T., J. Electrochem. Soc. 126
(1979) 2269.

255
[52] Shin, J., Geib, K. M., Wilmensen, C. W. and Lillental-
Weber, Z. 1989 Amer. Vac. Soc. Symp., Boston, MA, Paper
EMl-ThMl.
[53] Lee, H. H., Racicot, R. J., and Lee, S. H., Appl. Phys.
Lett. 54 (1989) 724.
[54] Hoekje, S. J., Ph.D. Dissertation, Univ. of Florida,
(1990).
[55] Bose, S. S., Szafranek, I., Kim, M. H. and Stillman, G.
E., Appl. Phys. Lett. 56 (1990) 752.
[56] Jones, A. C., Jacobs, P. R., Cafferty, R. Scott, M. D.
Moore, A. H. and Wright, P. J., J. Cryst. Growth 77
(1986) 47.
[57] Blaauw, C., Shepherd, F. R., Miner, C. J. and
SpringThorpe, A. J., J. Electron. Mater. 19 (1990) 1.
[58] Nelson, A. W. and Westbrook, L. D., J. Cryst. Growth 68
(1984) 102.
[59] Blaauw, C., Emmerstorfer, B. and SpringThorpe, A. J., J.
Cryst. Growth 84. (1987) 431.
[60] Blaauw, C., Bruce, R. A., Miner, C. J., Howard, A. J.,
Emmerstorfer, B. and SpringThorpe, A. J., J. Electron.
Mater. 18 (1989) 567.
[61] Manasevit, H. M. and Simpson, W. I., J. Electrochem.
Soc. 118 (1971) 291.
[62] Hirtz, J. P., Larivan, J. P., Duchemin, J. P. and
Pearsall, T. P., Electron. Lett. 16 (1980) 415.
[63] Duchemin, J. P., Hirtz, J. P., Razeghi, M., Bonnet, M.
and Hersee, S. D., J. Cryst. Growth 55 (1981) 64.
[64] Hockly, M. and White, E. A. D., J. Cryst. Growth 68
(1984) 334.
[65] Razeghi. M, The MOCVD Challenge. Vol. 1. Adam Hilger,
Bristol, England, 1989.
[66] Logan, R. A., Tanbun-Ek, T. and Sergent, A. M., J. Appl.
Phys. 65 (1989) 3723.
Ludowise, M. J., Dietze, W. T. and Lewis, C. R., Inst.
Phys. Conf. Ser. 65 (1983) 93.
[67]

256
[68] Iwamoto, T., Mori, K., Mizuta, M. and Kukimoto, H., Jpn.
J. Appl. Phys. 22 (1983) L191.
[69] Duchemin, J. P., Razeghi, M., Hirtz, J. P. and Bonnet,
M., Inst. Phys. Conf. Ser. 63 (1981) 89.
[70] Sogou, S., Kameyana, A., Miyamoto, Y., Furuya, K. and
Suematsu, Y., Jpn. J. Appl. Phys. 23. (1984) 1182.
[71] Moss, R. H. and Spurdens, P. C., J. Cryst. Growth 68
(1984) 96.
[72] Fujii, T., Yamazaki, S. and Nakajima, K. 1988 Intern.
Conf. on Solid State Dev. and Mater., Tokyo, Japan,
Paper D-9-2, 395.
[73] Koukitu, A. and Seki, H., J. Cryst. Growth, 76 (1986)
233.
[74] Nahory, R. E., Pollack, M. A., Johnston, W. D. and
Barns, R. L., Appl. Phys. Lett., 33 (1978) 659.
[75] Hayes, J. R., Patel, D., Adams, A. R. and Greene, P. D.,
J. Electron. Mater. 11 (1982) 155.
[76] Saxena, R., Sardi, V., Oberstar, J., Hodge, L., Keever,
M., Trott, G., Chen, K. L. and Moon, R., J. Cryst.
Growth 77. (1986) 591.
[77] Mayer, R., Grutzmacher, D., Jurgensen, H. and Balk, P.,
J. Cryst. Growth 93. (1988) 285.
[78] Manasevit, H. M., Hess, K. L., Dapkus, P. D., Ruth, R.
P., Yang, J. J., Campbell, A. G., Johnson, R. E., Moudy,
L. A., Bube, R. H., Tabick, L. B., Fahrenbruch, A. L.
and Tsai, M. J., 1978 Conf. Rec., 13th IEEE Photovoltaic
Spec. Conf., 165.
[79] Razeghi, M. and Duchemin, J. P., J. Cryst. Growth 70
(1984) 145.
[80] Walpole, J. N., in Compound Semiconductors: Growth.
Processing, and Devices. P. H. Holloway and T. J.
Anderson (Editors), CRC Press, Boca Raton, FL, 1989.
[81] di Forte-Poisson, M. A., Brylinski, C., Colomer, G.,
Osselin, D., Hersee, S., Duchemin, J. P., Azan, F.,
Lechevallier, D. and Lacombe, J., Electron. Lett. .20
(1984) 1061.
Ogura, M., Inoue, K., Ban, Y, Uno, T., Morisaki, M. and
Hase, N., Jap. J. Appl Phys. 21 (1982) L548.
[82]

257
[83] Kim, J., Li, S. S., Figueroa, L., Carruthers, T. F. and
Wagner, R. S., J. Appl. Phys. 64 (1988) 6536.
[84] Gallant, M., Puetz, N., Zemel, A. and Shepherd, F. R.,
Appl. Phys. Lett. 52 (1988) 733.
[85] Grutzmacher, D., Meyer, R., Balk, P., Berg, C.,
Schapers, T., Luth, H., Zachau, M. and Koch, F., J.
Appl. Phys. 66 (1989) 697
[86] Delacourt, D., Papuchon, M., Poisson, M. A. and Razeghi,
M., Electron. Lett. 23 (1987) 451.
[87] Deppe, D. G., Gerrard, N. D., Pinzone, C. J., Dupuis, R.
D. and Schubert, E. F., Appl. Phys. Lett. 56 (1990) 315.
[88] Nelson, A. W., Cole, S., Wong, S., Harlow, M. J.,
Devlin, W. J., Wake, D., Rodgers, P. M. and Robertson,
M. J., J. Cryst. Growth 22 (1986) 579.
[89] Blood, P., Semicond. Sci. Tech. 1 (1986) 7.
[90] Dean, P. J., Prog. Crystal Growth Charact. 5 (1982) 89.
[91] Holland, R. and Blackmore, G. W., J. Crystal Growth. 68
(1984) 271.
[92] Cullity, B. D., Elements of X-Rav Diffraction. 2nd. Ed.,
Addison-Wellsley, Reading, MA, 1980.
[93] Kronig, R. de L. and Penney, W. G., Proc. R. Soc.
London, A130 (1930) 499.
[94] Skromme, B. J., Stillman, G. E., Oberstar, J. D. and
Chan, S. S., J. Electron. Mater. 13 (1984) 463.
[95] Williams, E. W., Elder, W., Astles, M. G., Webb, M.,
Mullin, J. B., Straughan, B. and Tufton, P. J., J.
Electrochem. Soc. 120 (1973) 1741.
[96] Kubota, E. and Katsui, A., Jap. J. Appl. Phys. 24.
(1985) 1344.
[97] Yamamoto, A., Yamaguchi, M. and Uemura, C., Appl. Phys.
Lett. 47 (1985) 975.
[98] Nakano, Y., Takahei, K., Noguchi, Y., Suzuki, Y., Nagai,
H. and Nawata, K., Electron. Lett. 12 (1981) 782.
[99] Bachman, K. J., Buehler, E., Miller, B. I., McFee, J. H.
and Thiel, F. A., J. Cryst. Growth ,39 (1977) 137.

258
[100] Fang, D., Wang, X., Xu, Y. and Tan, L., J. Cryst. Growth
66 (1984) 317.
[101] Roksnoer, P. J. and Van Rijbroek-VanDenBoom, M. M. B.,
J. Cryst. Growth 66 (1984) 317.
[102] Ishikawa, J., Takahashi, N., Ito, T., Sube, M. and
Kurita, S., J. Electrochem. Soc. 137 (1990) 343.
[103] Wada, O., Majerfeld, A. and Robson, P. N., J.
Electrochem. Soc., 127 (1980) 2278.
[104] Hooper, A. and Tuck, B., Solid State Electron. 19
(1976) 513.
[105] Park, R. M., Stanley, C. R. and Clampitt, R., Inst.
Phys. Conf. Ser. 54 (1980) 235.
[106] Panish, M. B., Temkin, H. and Sumski., S., J. Vac. Sci.
Technol. B3 (1985) 657.
[107] Cheng, T. S., Airaksinen, V. M. and Stanley, C. R., J.
Appl. Phys. 64 (1988) 6662.
[108] Olsen, G. H., Interqrated Circuits: Chemical and
Physical Processing. ACS Symp. Ser. #290 (1985) 221.
[109] Jurgensen, H., Heyen, M. and Balk, P., Inst. Phys. Conf.
Ser. 56 (1981) 55.
[110] Chevrier, J., Huber, A. and Linh, N. T., J. Appl. Phys.
51 (1980) 815.
[111] Chevrier, J., Horache, E., Goldstein, L. and Linh, N.
T., J. Appl. Phys. 53 (1982) 3247.
[112] Yang, J. J., Ruth, R. P. and Manasevit, H. M., J. Appl.
Phys. 52 (1981) 6729.
[113] Bacher, F. R. and Leigh, W. B., J. Cryst. Growth 68
(1984) 102.
[114] Timmons, M. L., Chiang, P. K. and Hattangady, S. V., J.
Cryst. Growth 77 (1986) 37.
[115] Blaauw, C. B., Shepherd, F. R. and Eger, D., J. Appl.
Phys. 66 (1989) 605.
[116] White, A. M., Dean, P. J., Fairhurst, K. M., Bardsley,
W., Williams, E. W. and Day, B., Solid State Comm. 11
(1972) 1099.

259
[117] Skromme, B. J., Stillman, G. E., Oberstar, J. D. and
Chan, S. S., Appl. Phys. Lett. (1984) 319.
[118] Williams, R. S., Barnes, P. A. and Feldman, L. C., Appl.
Phys. Lett. 36 (1980) 760.
[119] Lennard, W. N., Swanson, M. L., Eger, D., SpringThorpe,
A. J. and Shepherd, F. R., J. Electron. Mater. 17
(1988) 1.
[120] Yamada, M., Tien, P. K., Martin, R. J., Nahory, R. E.
and Ballman, A. A., Appl. Phys. Lett. 43. (1983) 594.
[121] Wong, C. C. D. and Bube, R. H., J. Appl. Phys. 55
(1984) 3804.
[122] van Urp, G. J., van Dongen, T., Fontijn, G. M., Jacobs,
J. M. and Tjaden, D. L. A., J. Appl. Phys. 65 (1989)
553.
[123] Marek, H. S. and Serreze, H. B., Appl. Phys. Lett. 51
(1987) 2031.
[124] Kazmierski, C., J. Appl. Phys. 64 (1988) 6573.
[125] Dlubek, G. and Brummer, O., Appl. Phys. Lett. 46 (1985)
1136.
[126] Li, S. S., Wang, W. L. and Shaban, E. H., Solid State
Comm. 51 (1984) 15.
[127] Kami j oh, T., Takano, H. and Sakuta, M., J. Cryst. Growth
67 (1984) 144.
[128] Cole, S., Evans, J. S., Harlow, M. J., Nelson, A. W. and
Wong, S., Electron. Lett. 24 (1988) 813.
[129] Glade, M., Grutzmacher, D., Meyer, R., Woelk, E. G. and
Balk, P., Appl. Phys. Lett. 54 (1989) 2411.
[130] Omeljanovsky, E. M., Pakhomov, A. V. and Polyakov, Y.,
Semicond. Sci. Technol. 4 (1989) 947.
[131] Jackson, D. A., J. Cryst. Growth. 87 (1988) 205.
[132] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 613.
[133] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 627.
[134] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 639.
[135] Hurle, D. T. J., J. Phys. Chem. Solids. 40 (1979) 647.

260
[136] Brouwer, G. Philips Res. Rep. 9 (1954) 366.
[137] Kroger, F. A., Chemistry of Imperfect Crystals. North
Holland, Amsterdam, Holland, 1964.
[138] Constantinides, A., Applied Numerical Methods with
Personal Computers. McGraw-Hill, New York, NY, 1987.
[139] Dobrowolski, J. and Driscoll, W.(editor), Handbook of
Optics. McGraw-Hill, New York, NY, 1977, Section 8.
[140] van der Ziel, J. P. and Ilegems, M., Appl. Optics 14
(1975) 2627.
[141] Yoffe, G. W., Schlom, D. G. and Harris, J. S., Jr., 1987
Ann. Dev. Res. Conf., Santa Barbara, CA., paper 3A-4.
[142] Thorton, R. L., Burnham, R. D. and Streifer, W., Appl.
Phys. Lett., 45 (1984) 1028.
[143] Tai, K., McCall, S. L., Chu, N. G. and Tsang, W. T.,
Appl. Phys. Lett. 51 (1987) 826.
[144] Ritchie, S., Spurdens, P. C., Hewitt, N. P. and Aylett,
M. R., Appl. Phys. Lett. 55 (1989) 1713.
[145] Born, M. and Wolf, E., Principles of Optics. Pergammon,
New York, NY, 1964.
[146] Afromitz, M. A., Solid State Comm. 15 (1979) 59.
[147] Broberg, B. and Lindgren, S., J. Appl. Phys. 55 (1984)
3376.
[148] Cardona, M., Modulation Spectroscopy. Academic Press,
New York, NY, 1969.
[149] Katz, J., Margalit, S. and Yariv, A., IEEE Trans.
Electron Dev. ED-29 (1982) 977.
[150] SpringThorpe, A. J. and Mandeville, P., J. Vac. Sci.
Tech., B4 (1986) 853.
[151] SpringThorpe, A. J., Ingrey, S. J., Emmerstorfer, B. and
Mandeville, P., Appl. Phys. Lett., .50 (1987) 77.
[152] Hill, M. J., Ph.D. Thesis, Durham Univ. England (1985).
[153] Sell, D. D., Casey, H. C. and Wecht, K., J. Appl. Phys.
45 (1974) 2650.
[154] Cross, M. and Adams, M. J., Opto-Electron 6 (1974) 199.

261
[155] Botteldooren, D. and Baets, R., Appl. Phys. Lett. 54
(1989) 1989.
[156] Chien, C. L. and Westgate, R., The Hall Effect and
its Applications. Plenum, New York, NY, 1979.
[157] ASTM Standard Method for "Method and Hall Coefficient in
Extrinsic Semiconductors," F76-73 (1978).
[158] Putley, E. H., Hall Effect and Related Phenomena.
Butterworth & Company, Ltd., London, England, 1960.
[159] Beer, A. C., Galvanomaanetic Effects in Semiconductors.
Academic Press, New York, NY, 1963.
[160] Stillman, G. E., Wolfe, C. M. and Dimmock, J. O., Jour.
Phys. Chem. Solids. 31 (1970) 1199.
[161] Wieder, H. H., Laboratory Notes on Electrical and
Galvanometric Measurements. Elsevier, New York, NY,
1979.
[162] van der Pauw, L. J., Philips Res. Rep. 13 (1958) 1.
[163] Benzaquen, M., J. Amer. Phys. Rev. B., 34 (1986) 8947.
[164]Hayakawa, Y., MOCVD Operating Procedures A Translation.
Nippon Sanso Inc., Hammamatsu, Japan, 1988.

BIOGRAPHICAL SKETCH
Arnold John Howard was born on January 6, 1963, in
Huntington, NY, which is located on the north shore of Long
Island. He grew up there and graduated from Huntington High
School in 1981. He then enrolled at the University of
Pennsylvania in Philadelphia, Pennsylvania, and graduated in
1985 with a B.S. degree in chemical engineering. While at
Penn he met Robin Lustig, another Penn Quaker, at a party that
he and his friends had stumbled upon; Robin later became his
wife on April 7, 1990. After graduating, the author spent six
weeks in nine different countries of Europe. After all that
traveling, he enrolled at the University of Virginia in
Charlottesville. As part of his master's program, the author
spent a summer at the Oak Ridge National Laboratory performing
research on the separation of sugars. He received his M.S. in
chemical engineering from UVa in January of 1987 and then went
further south to begin the Ph.D. program at the University of
Florida in Gainesville. The author had the good fortune of
choosing a research project which involved being a visitor at
Bell Northern Research in Ottawa, Canada. (Gainesville was
too hot!) The author plans to work at Sandia National
Laboratories in Albuquerque, New Mexico, upon completion of
his Doctor of Philosophy degree in chemical engineering.
262

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Timothy J. Anderson, Chairmen
Professor of Chemical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Chemical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Sheng S. Li
Professor of
Electrical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctqrof Philosophy.
Gys Bosman
Associate Professor of
Electrical Engineering

I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
,\.
Anthony J. SpringThorpe
Manager of Epitaxy
Bell Northern Research, Canada
This dissertation was submitted to the Graduate Faculty
of the College of Engineering and to the Graduate School and
was accepted as partial fulfillment of the requirements for
the degree of Doctor of Philosophy.
December 1990
/ Loius" Cl
A^Winfred M. Phillips
v Dean, College of Engineering
Madelyn M. Lockhart
Dean, Graduate School

UF Libraries: Digital Dissertation Project
Page 2 of 2
Internet Distribution Consent Agreement
In reference to the following dissertation:
AUTHOR: Howard, Arnold
TITLE: Growth and modeling of III-V compound semiconductor optoelectronic
materials with device applications / (record number: 1659152)
PUBLICATION DATE: 1990
I,
/IzMu (2.
as copyright holder for the aforementioned
dissertation, hereby grant specific and limited archive and distribution rights to the Board of Trustees of
the University of Florida and its agents. I authorize the University of Florida to digitize and distribute
the dissertation described above for nonprofit, educational purposes via the Internet or successive
technologies.
This is a non-exclusive grant of permissions for specific off-line and on-line uses for an indefinite term.
Off-line uses shall be limited to those specifically allowed by "Eair Useu as prescribed by the terms of
United States copyright legislation (cf, Title 17, U.S. Code) as well as to the maintenance and
preservation of a digital archive, copy. Digitization allows ihe University of Florida to generate image-
and text-based versions as appropriate and to provide and enhance access using search software.
This grant of permissions prohibits use of the digitized versions for commercial use or profit.
Signature of Copyright Holder
/~}/Zmq u> sj.
Printed or Typed Name of Copyright-Holder/Licehsee
Personal information blurred
7h /a gr
Date of Signature
Please print, sign and return to:
Cathleen Martyniak
UF Dissertation Project
Preservation Department
University of Florida Libraries
P.O. Box 117007
Gainesville, FL 32611-7007
7/3/2008



3
Table 1
Properties of Silicon and III-V Binary
Semiconductors at 300 K
Bandgap
Type
Bandgap
Energy(eV)
Electron
Mobility
(cm2/V-s)
Lattice
Constant
(Angstroms)
Si
indirect
1.12
1350
5.43
InSb
direct
0.18
100000
6.48
InAs
direct
0.36
22600
6.06
GaSb
direct
0.70
5000
6.09
InP
direct
1.28
4000
5.87
GaAs
direct
1.43
8500
5.65
AlSb
indirect
1.60
200
6.14
AlAs
indirect
2.16
180
5.66
GaP
indirect
2.26
300
5.45
A1P
indirect
2.45
80
5.46
Source: Streetman[1].


REFLECTIVITY {%)
190
Figure 62: Experimental and theoretical DCD rocking curves.


180
The complete multilayer of Figure 59 is represented by
their product matrix, M:
m,i
im21
imi2
m22
M1M2* Mt
(34)
From the individual elements, m{j, of the matrix M, the
reflectivity (neglecting absorption and scattering) can be
calculated as follows:
(""Hi ~ nsm22)2 + (njigm^ m21)2
(njn +nsm22)2 + (nji^ + m21)2
where ns is the refractive index of the substrate material.
A computer program based on equations (30) to (35) which was
written to calculate the reflectivity as a function of
wavelength for both GaAs and InP based stack structures.
Within the program, estimates of the refractive indices (RI)
of A^Ga^As and Ga^n^ASyP^y materials were calculated at
photon energies below the direct band edge. The RI values are
based on semi-empirical formulas of Afromowitz[146] for
AlxGa.,_xAs and of Broberg and Lindgren[147] for Ga^n^ASyP^y.
For quarter-wave films the number of periods of high and
low RI layers required to attain a desired reflectivity can be
calculated using the following equation from Born and
Wolf(145]:


196
and aluminum profiles were measured to assess the abruptness
of the AlGaAs/GaAs heterostructure interface and to determine
the depth of the zinc diffusion. The profiles for the full
stack structure are shown in Figure 66. As shown, both the
heterojunction interface is abrupt and the A1 composition is
constant throughout the stack. The extent of zinc diffusion
can be more clearly seen in Figure 66. It appears that the
zinc diffusion ends in the second GaAs layer, hence creating
the desired PNPN (even number of junctions) type structure.
After the zinc diffusion was completed and confirmed by
SIMS, a Si3N4 dielectric coating was evaporated to a thickness
of 200nm. Fifty micron square devices were delineated by
etching mesas and ohmic contacts were made to the top layer.
The back side of the substrate was lapped down and polished
and a gold ohmic contact layer was evaporated onto this
surface. The whole PNPN structure was then tested.
Individual devices were electrically probed using a HP
4145B semiconductor analyzer. Current was measured as a
function of forward and reverse bias. Some devices exhibited
a shift from an "OFF" to "ON" state under forward bias
conditions when a voltage of 15 to 30 volts was applied. This
is the first reported switch for an electrical structure of
this size (forty layers) An I-V plot of a typical PNPN
device exhibiting bistability is shown in Figure 67. Also
plotted on the same scale is the resistance versus voltage.


247
pressurize it with N2 and close all four MVs. Turn off the
Lepel RF generator's circuit breaker. Purge the rough pumps
with N2 also. Close the nitrogen inlet, dry air and cooling
water inlet and outlet valves. Turn off electric breakers
ELB100, 200, and then the MOCVD breaker. The system should
now be completely off and ready for a long vacation or major
renovation.
B.8 MAINTENANCE
General system maintenance is done as needed. Of course,
occasionally vacuum out the glove box for dust removal and
wipe down the system with a "clean room" detergent. All other
maintenance steps are mentioned as either part of preparation
or procedure steps.


25
pressure. They also compared the growth rate of undoped InP
on (100), (111), and (115) InP oriented substrates and they
reported excellent film quality on (100) 2 towards (110) and
(115) 2 towards (111). Eguchi et al.[46] studied the effect
of V/III (phosphine to metal organic indium) ratio on EPD and
electrical properties and reported superior material at high
V/III ratios (>300) This result was in agreement with
Kasemset's[47] earlier work on both V/III ratio and growth
temperature effects. However, a survey of the effect of
growth temperature on layer quality is less conclusive.
Kasemset[47] indicates that a decrease in background carrier
concentration results upon increasing growth temperature,
while Scott et al.[48] report the opposite trend. This
discrepancy is probably due to different dominant impurities
in each group's TMIn source with correspondingly different
incorporation mechanisms. Most teams report high quality
MOCVD InP grown at temperatures between 550c and 675C.
Below 550C, growth rates drop and material quality degrades.
Above 700C, background carrier concentrations increase.
Presently, high quality two inch diameter wafers of InP
are commercially available as both doped (p and n-type) and
semi-insulating. Variation of results from team to team in
early research efforts and even today may be due in part to
the lack of reproducibility of substrate properties from batch
to batch and vendor to vendor. A recent paper from Knight et
al.[49] reports this problem. They observed a correlation


223
of interest. Depending on the type of sample, different
solders must be attached as close as possible to the four
corners of the wafer. For undoped GaAs and InP, pure indium
works well when cut and pressed onto the "epi" layer with a
clean surgical knife. The indium dots should also be as small
as possible to avoid introducing error. For other materials
such as doped III-V's, Si and Ge different solders are
suggested[157].
Now that the solder is in place, the sample should be
placed in the alloying station on the center of the graphite
heating strip. Check to see that there is a small air gap
between the thermocouple and the strip. Also be careful when
handling the strip as it is quite fragile. Put the "o"-ring
and plexiglass dome in place and tighten the mounting bolts in
a diagonal fashion. Do not tighten the bolts too much as the
dome is fragile.
The next step is probably the most important in the
alloying process: chamber evacuation and heating in a forming
gas (10% H2 in N2) environment. First open the regulator
valve on the forming gas cylinder so that there is about 10
psi of pressure on the flow regulator gauge. Close the line
to the rough pump and open the exit line of the chamber.
Place the exit line in a beaker two-thirds full of water for
a visual indication of the gas flow. Open the inlet valve
line to chamber so that the water in the beaker is bubbling to
the point of almost overflowing (increase cylinder flow if


38
MOCVD. With MOCVD, it is possible to overgrow onto sub-micron
diffraction gratings which is vital to the operation of these
devices[79]. A typical laser structure composed of these
materials would start with a n+-InP substrate, followed by a
2/im thick n+ InP layer, then a 0.2frn active layer of
lattice-matched GalnAsP (undoped), then a 2/xm thick p-InP
layer ending with a p+ InP contact layer of 0.2/im thickness.
The lasing wavelength is determined by the composition of the
active layer. Surface emitting semiconductor diode lasers,
which emit light perpendicular to the grown layer surface have
also been fabricated in an array form using MOCVD grown
GalnAsP[80].
Microwave devices such as Gunn diodes and metal to
semiconductor field-effect transistors (MESFET's) have been
grown using MOCVD InP. For Gunn effect devices, even though
the mobility of InP is lower than GaAs, other characteristics
such as cut-off frequency, acceleration-deceleration time,
relaxation time and peak-to-valley ratio are better in InP.
These devices require a three-layer structure of n+-n-n+-InP
grown on n+-InP and have been successfully grown by MOCVD for
60 GHz[36], and 94 GHz[81] operation. MESFET's have also
successfully been fabricated by the use of undoped InP grown
by MOCVD on Fe-doped substrates. The electrical properties of
Au-InP Schottky diodes are reasonable and comparable to other
crystal growth techniques[82].


169
The partial pressures of phosphorous and zinc, pp4 and p2n, in
the gas phase near the heated InP surface are known. As
already discussed, the majority of zinc in the gas phase above
the growing InP:Zn crystal is not incorporated into the solid.
Also, InP is typically grown with a large overpressure of PH3
(V/III = 140) which cracks to form P4. Consequently, both pZn
and pP4 are set equal to the gas phase partial pressures at
the inlet to the reactor of DEZn and PH3, respectively.
It has been reported that yn and Yp/ the activity
coefficients of the electrons and holes, are only important
for closely compensated material[132], hence their magnitude
is set to unity. If the eleven constants 1^.,- K^, K^- Kd11,
are known and electroneutrality is satisfied, then the twelve
unknown defect concentrations: Pif Inf, Vp, Vp+, n, p, VIn,
VIn-, Zn}, Zn,n-, Zn,nVp, VpZnInVp, can be calculated. The
charged state of zinc related defects have been chosen based
on values reported in the literature and discussed in the
previous section, and based on the observed incorporation and
electrical characteristics of the InP:Zn data from BNR.
Undoped InP grown by MOCVD in the BNR reactor had a room
temperature carrier concentration of n = l-2xl015cm"3. The
lowest hole concentration achieved from the DEZn doping of InP
experiments was p = 2xl016cm*3 and most layers had much higher
hole concentrations. Hence, it is safe to approximate the
electroneutrality relation by its two dominant members at room


185
accompanied by light emission if there is sufficient carrier
injection.
The novel active optoelectronic switch element thus
created has the useful feature of indicating its state
optically at a wavelength (e.g. 870nm) remote from the
wavelength the element is switching by means of its altered
reflectivity response (e.g. 1300nm). As with thyristors in
general, switching between the "ON and "OFF" states can be
controlled either electrically or optically.
The electronic structure of such a multi-heterojunction
stack is very complex. Simple analysis indicates, that to
achieve bistability, the number of junctions must be even
[149]. More detailed analysis is required to optimize the
modulator function, as well as to design the stack for a given
forward breakdown voltage and holding current. With an
appropriate design, the emitted light could also be involved
in the switching function by way of optical feedback.
4.2.3 Experimental
4.2.3a Crystal Growth
The superlattice was grown by periodic variation of the
A1 content in epitaxially grown AlGaAs crystals on a GaAs
(001) substrate using a VG Semicon V80-H molecular beam
epitaxy (MBE) system. The 2" diameter substrate was rotated
during growth and was prepared for indium-free mounting using
previously described methods[150].


219
switching and a selective wavelength detector. Electrical
bistability was observed and passive optical interference was
demonstrated for this device. The combination of the two
device characteristics: a peak shift in the reflectivity
spectrum due to the application of an external bias, was not
observed. The expected spectral shift may have been observed
if the holding current was lower and/or if the current
distribution was more uniform. Consequently, optimized p-n-p-
n structures with fewer layers using multiple quantum wells
are suggested as improvements for future researchers. Also,
broad-area transparent contacts to the epilayer could improve
the current channeling problem.


235
explodes or you suspect a hazardous gas leak, do not wait for
the MDA to turn on the whooper leave the Surge building
immediately. Do not return to the room until the emergency
response team has entered the room with their Scott Air packs
and solved the problem. Bottle changes, reactor tube removal
and cleaning and pump oil changes should be done using the
buddy system; with the dual-cylinder air system. No growth
shall be performed unless at least 2 people are in the
building. Always wear safety glasses while in the clean room.
All operators should take part in the arsenic in urine
monitoring. If you ever smell a garlic-like or rotten fish
odor or sense your heart rate increase above normal (for no
apparent reason), leave the room. Never start a deposition or
use the bubbler baths during stormy weather. Always inform
the building director that you are going into the lab and if
you are planning to do a growth. The MOCVD has several
sensors which light-up, flash or cause a buzzer to sound on
the display panel. It has three levels of warning and
corresponding actions which result depending on the level of
danger to the system or operator. (see pages 25-29 of
Hayakawa[164]) .
B.4 PREPARATION
Check with the building operator to see if anything is
abnormal today with regards to the nitrogen, hydrogen, chilled
water, dry air, building power supply, and air scrubbing
monitoring. Check the status of the Advanced Concepts wet


207
Figure 72:
TEM micrograph of a cross-section of the stack
structure MBE572.


164
VpZnInVp complex based on their results and propose that atomic
hydrogen which comes from the pyrolysis of the hydrides, has
an influence on the doping level of p-InP. They support this
theory by showing that SIMS H-profiles are related to doping
profiles. Glade et al.[129] performed annealing experiments
and they proposed that acceptor-hydrogen complexes exist
interstitially and their activation is limited by indium
vacancy diffusion. Also, hydrogen passivation of a p-type
InP:Mn sample showed a drastic decrease in hole concentration
from p = 7.4xl016cm'3 down to p s I013cm*3[l39]. So, it is
clearly evident that there is still much uncertainty about the
electrical activation process of p-dopants in InP.
3.4.2 Point Defect Structure
The incorporation of DEZn during the growth of InP by
MOCVD can be explained qualitatively by the model of Razeghi
and Duchemin[45]. The first assumption is that all the DEZn
arriving at the growing surface is decomposed. This is a good
assumption since the growth temperature is usually 600-650C
and the onset of pyrolysis of DEZn occurs at 332.3C[131].
Hence, the Zn concentration is limited by the rate of arrival
of DEZn to the hot surface. After decomposition there are two
possible limiting cases proposed by Razeghi and Duchemin[45]:
(1) all of the available decomposed dopant source material is
incorporated into the growing layer and the resulting impurity
concentration is independent of temperature and inversely
proportional to the growth rate; and (2) only a small fraction


19
is ideal because it is a non-contact method, it selectively
heats only the graphite, and it is easy to configure by
arranging a copper coil around the susceptor portion of the
reactor. The RF generator size required depends on susceptor
size, gas velocity, coupling efficiency, and reactor wall
cooling mechanism. Infrared heating from quartz-halogen lamps
has also been used but, as wall deposition increases, non-
uniform heating may occur. A resistance heater embedded into
the graphite is another option but deposition on electrical
feedthroughs complicates reactor cleaning. Most heating
systems use an embedded thermocouple feedback system to
control temperature. Optical pyrometers have also been used
but wall deposition can result in false readings.
After passing through the heated zone of the MOCVD
reactor, some of the toxic gas sources still remain uncracked
and undeposited. These gases have to be neutralized before
being discharged into the atmosphere. Hazardous gases such as
AsH3, PH3 and SiH4 are commonly used in the MOCVD of III-V
semiconductors and are difficult to neutralize or "scrub"
especially when they are used in combination. There are four
different types of scrubbing systems commercially available;
depending on the application one alone may be inadequate. The
four types are liquid scrubbers, thermal crackers, dry powder
scrubbers, and incinerators. Liquid based scrubbers are most
commonly used and work by bubbling the toxic gas through a
basic (pH > 10) solution of sodium hypochlorite and sodium


213
15 X
Growth Conditions
Growth Temperature: 620C
Growth Pressure: 80 Ton-
Hydrogen Flowrate: 7 SLM
Metal Organics
Hydrides
MFTMm 0.737xl0'4
MFPH3 =211xl0'4
MFTEGa = 0.3 3 5x1 O'4
MFashb = 25x10 "
MFDE2n =0.621xl0'4
MFH2S = 0.07 lxl O'4
Figure 76; Layer structure and growth conditions for MOCVD
growth Q240.


AB (meV)
107
Figure 32: The effect of well thickness on both experimental
and theoretical confinement energy.


41
Figure 2: Photograph of the quaternary MOCVD system.


160
state that a decrease in hole concentration at higher zinc
melt concentrations is due to compensation in the form of an
interstitial zinc donor complex originally proposed by Hooper
and Tuck[104]. Hydride CVD grown material has been reported
to increase and then decrease in hole concentration with the
number of experiments performed using the same indium-zinc
heated source[109]. This writer proposed that this trend is
due to a coupling of zinc compensation and evaporation.
Chloride CVD grown samples with initially low zinc electrical
activation, became, upon annealing, fully active[110]. This
can be explained by the disassociation of compensating defect
complexes due to prolonged heating. Finally, MOCVD grown
samples saturate at p = 2xl018cm'3 and the hole concentration
has a square root dependence on DEZn partial pressure[58].
As the MOCVD growth temperature is increased, hole and
atomic zinc concentrations decrease due to increased zinc
evaporation from the growing surface at higher temperatures
[45]. Most of the above trends for MOCVD grown InP:Zn have
been observed in the BNR DEZn doped InP data which is shown in
Figure 55. Note that at low [Zn], p [Zn], but above [Zn] =
2xl017cm'3, [Zn] eventually becomes much greater than p. In
addition, for the BNR data, the low growth rate of undoped
InP, one micron per hour, was unaffected by the DEZn molar
flow rate even at its maximum, 4.92xl0'4 moles/min, and the
hole concentration was unaffected by a change in growth rate
at a fixed DEZn flow rate.


log N(cm-3)
152
Figure 50: C-V profile of sample B318 InP:Zn (bubbler) grown
on InP:S.


REFERENCES
[1] Streetman, B. G., Solid State Electronic Devices. 2nd.
Ed., Prentice-Hall, Englewood Cliffs, NJ, 1980.
[2] Kuphal, E., J. Cryst. Growth 54 (1981) 117.
[3] Bocchi, C., Ferrari, C., Franzosi, P., Fornuto, G.,
Pellegino, S., and Taiarol, F., J. Electron. Mater. 16
(1987) 245.
[4] Lyons, M. H., Faktor, M. M., and Moss, R. H., J. Cryst.
Growth 66 (1984) 269.
[5] Olsen, G. H. and Zamerowski, T. J., IEEE J. of Quantum
Electron. OE-17 (1981) 128.
[6] Ludowise. M. J., J. Appl. Phys. 58 (1985) R31.
[7] Cho, A. Y., J. Vac. Sci. Technol. 16 (1979) 275.
[8] Bedair, S. M., Tischler, M. A., and El-Masry, N., Appl.
Phys. Lett. 47 (1985) 51.
[9] Tsang, W. T., Appl. Phys. Lett. 45 (1984) 1234.
[10] Chen, W. K., Chen, J. C., Anthony, L. and Liu, P. L.,
Appl. Phys. Lett. 55 (1989) 987.
[11] Manasevit, H. M., Appl. Phys. Lett. 12 (1968) 156.
[12] Manasevit, H. M. and Simpson, W. I., J. Electrochem.
Soc. 116 (1969) 1725.
[13] Manasevit. H. M., J. Electrochem. Soc. 118 (1971) 647.
[14] Manasevit, H. M., Erdmann, F. M. and Simpson, W. I., J.
Electrochem. Soc. 118 (1971) 1864.
[15] Manasevit, H. M. and Thorsen, A. C., J. Electrochem.
Soc. 119 (1972) 99.
[16] Manasevit, H. M. and Simpson, W. I., J. Electrochem.
Soc. 120. (1973) 135.
[17] Bass, S. J., J. Cryst. Growth 11 (1975) 172.
252


85
X (um)
Figure 20: C-V profile of InP:Zn grown with different DEZn
partial pressures.


40
and results of low pressure MOCVD grown GalnAsP/InP and MBE
grown AlGaAs/GaAs electrically tuneable interference filters
are presented in Chapter IV of this dissertation. More
extensive reviews of the wide range of MOCVD grown opto
electronic devices using InP based materials are available in
the 1iterature[65,88].
2.4. A Description of the MOCVD System
2.4.1 Introduction
The experimental apparatus used for the growth of
epitaxial layers of Ga^n^jjASyP^y on InP substrates is a
commercial MOCVD system custom built for the University of
Florida by Nippon Sanso K.K. (Japan Oxygen Inc.). A
photograph of the front and a simplified schematic of the
Japan Oxygen MOCVD System are shown in Figures 2 and 3. The
complete operating procedures for performing epitaxial growths
and maintenance (e.g., such as reactor cleaning), are
presented in Appendix B of this text. The four basic parts of
the MOCVD system which are described in the following
paragraphs are: (1) the gas delivery system; (2) the reactor
and heating system; (3) the exhaust/scrubbing system; and (4)
the safety system. The gas delivery system, reactor, exhaust
and safety system are all integrated inside the MOCVD system
which is shown in the photograph in Figure 2. The heating
system is a separate unit (20 kW RF generator) as is the
scrubbing system which is located outside the building for
ease of maintenance reasons.


178
two matrix is formed from the refractive index, n? thickness,
t; angle of incidence of light of wavelength X, 0; and, angle
of propagation of light, 0, for each of the layers (see Figure
59) in the stack. The resulting matrix for layer j is as
follows:
cos <5¡ -Lsin J UJ 1
iUjSinj cos5j
(30)
where the phase is:
=
2 7T
(njt j COS0J-)
(31)
and Uj, the effective refractive index is:
ui =
n¡
parallel
COS0j
nj cos0j perpendicular
(32)
depending on whether the incident light is polarized parallel
or perpendicular to the plane of incidence [145]. The angle is calculated using Snell's Law:
nm sin0 = nj sin0j (33)
where the refractive index of the medium (air), n,,,, is unity.


238
Figure 79: Quaternary MOCVD flow diagram (complete)


REIATNE INTENSITY
74
Figure 14: PL spectrum of an InP sample grown with a V/III
ratio of 219 measured at 4.2 K.


31
the improvement in material quality observed upon using TEGa
instead of TMGa as the gallium source. Another useful
observation was that the solid phase composition is controlled
by and almost equal to the gas phase ratio [TMIn]/([TMIn] +
[TEGa]). Finally, the GalnAs growth rate is proportional to
the sum of the metal organic gas phase concentrations.
Lattice-matched Ga 47In>53As grown on InP by low pressure
MOCVD using TEGa and TEIn was first reported by Hirtz et al.
[62] in 1980. As stated previously, the choice of group III
alkyl sources used for GalnAs growth is critical. Using TMGa
and TEIn results in poor surface morphology, whereas using
TEGa and TEIn results in nearly featureless material over the
composition range 0.4 < x < 0.6 (Gax) This phenomena has
been attributed to a TMGa-InP substrate steric hindrance to
the heterogeneous decomposition of TEIn[63]. When growing on
InP, the initial stage of growth of GalnAs is also complicated
by the incongruent evaporation of phosphorus from the InP
substrates upon heating. It has been shown that the
morphological, optical and electrical properties of the GalnAs
epitaxial layer depend heavily on minimizing InP substrate
damage during the transition from PH3 to AsH3[64]. The best
approach is using an InP buffer layer and then allowing the
indium flow to continue while rapidly switching phosphine to
the vent and TEGa and AsH3 to a low pressure reactor.
As is the case for growth on GaAs substrates, the
composition of GalnAs is linearly dependent on the flow rate


49
its power output. Automatic temperature control from room
temperature to growth temperatures (550-700C) to bakeout
temperatures (900-950C) is possible with this elaborate
heating system.
As shown in Figure 4, a high-purity quartz sample tray
sits on top of part of the graphite susceptor. This tray
holds substrates as large as 1/4 of a two inch diameter wafer.
The tray and substrate are placed onto and removed from the
susceptor by an electro-mechanical fork which is capable of
precise x-y-z motion. The fork can move 90 cm horizontally
from the load lock area (where the sample and tray are loaded
and unloaded) through a gate valve and shutter valve, by the
gas exhaust port to directly above the susceptor. Precise
mechanical sample loading makes it possible to use the reactor
even after side wall deposition has obstructed view of the
susceptor.
To prevent the introduction of "dirty" room air to the
reactor during each loading of a substrate, the air in the
load-lock is evacuated by a rotary pump (Edwards model E2M2)
to a roughing pressure of 10'2 torr. Then a turbomolecular
pump (Balzers model TPH050) is turned on and evacuates the
load-lock to a pressure of 10'7 torr. At this point the load
lock is isolated and backfilled with the ultra high purity
hydrogen. The sample is now ready for loading and the ultra
high vacuum gate valve (VAT Ltd. model MSS4) is opened.


RELATIVE INTENSIFY
105
Figure 31; PL spectrum of MQW sample Q136 measured at 4.2 K


36
Ag(/m) Using one of the above mentioned techniques, good
estimates of optimum gas phase growth conditions for the full
range of solid phase quaternary compositions can be predicted.
Several papers have been written presenting optical and
electrical property data as a function of composition for
GajjIn^j^ASyP^y lattice-matched to InP. One important
realization is that for lattice-matching, y is related to x by
the following simple relation: y = 2.16 x. This greatly
facilitates presenting data as it can be plotted as a function
of y or x under the assumption that the lattice-matching
condition is realized. A paper by Nahory et al.[74] presents
useful experimental lattice constant and bandgap values as a
function of composition relative to lattice constant values
predicted using Vegard's Law. Vegard's Law states that for a
lattice-matched system, the lattice parameter of the
quaternary can be deduced from those of the constituent
binaries. This team also presents the empirical relation for
bandgap variation (Eg(eV)) with composition.
Eg(y) = 1.35 0.72y + 0.12^ (1)
Another group presented undoped electron and hole mobilities
as a function of composition, y. For undoped GalnAsP, room
temperature electron mobilities range from 4000 to
11,000cm2/volt-sec (y = 0 to y = 1) and room temperature hole
mobilities range from 130 to 200cm2/volt-sec (y=0 to y=l) [75].
Both p- and n-type doping of GalnAsP/InP grown by MOCVD
have been reported. Extensive doping studies as a function of


43
2.4.2.Gas Delivery System
The gas delivery system connects the sources to the
reactor and provides a method of transporting them in a
controlled fashion. Since impurity levels must be kept to a
minimum, all components of the gas delivery system are
constructed from electropolished 316L stainless steel and
connected with metal-gasket leak-tight couplings. Also for
improved purity, 0.2/xm particle filters are installed at all
gas inlet points. All lines were wrapped with electrothermal
heating tape and aluminum foil and are heated during standby
mode to 50C to help desorb any of the sources or impurities
adsorbed on the inner walls of the stainless steel tubing.
The flow of gases is controlled by a combination of
manual valves, needle valves, pneumatic valves, check valves,
electronic mass flow controllers and regulators. The range of
possible flow rates for each source and the carrier gases
(hydrogen and nitrogen) are given in Table 2. The house
nitrogen gas which is mainly used for purging the MOCVD system
before a reactor or source change, passes through a molecular
sieve cartridge (Matheson Model 451) before entering the
machine. The house hydrogen, which is the carrier gas in the
system, is purified by diffusing it through a heated (400C)
palladium-alloy membrane which is part of a 0-20 liter/min
hydrogen purifier system (Matheson, Series 8370V) that was
installed inside the Japan Oxygen machine.


44
Table 2
Flow Rate Ranges of Sources, Vendors and Purity
Source
Flow Rate Range
Vendor(Purity)
h2
0-20 SLM
Gator Oxygen
(Alloy Diffused)
N2
0-10 SLM
Linde(LN2 Boil-off)
AsH3
0-50 seem
Matheson (ULSI Grade)
ph3
0-200 seem
Solkatronic
(Micropure Grade)
lOOOppm H2S
(in H2)
0-50 seem
Matheson
(ULSI Grade)
TMIn
0-300 seem
Air Products
(Diphos Grade)
TEGa
0-100 seem
Akzo
(Electronic Grade)
DEZn
0-50 seem
Morton Thiokol
(Electronic Grade)


129
They also reported that by increasing the growth temperature
from Tg = 530 to 650C at a fixed DEZn flow, the 300 K hole
cpncentration decreased from 4xl018 to 4xl017cm'3 due to, as
they explain, increased dopant evaporation from the growing
surface.
Nelson and Westbrook[58] also calculated the diffusion
coefficient of Zn in InP based on SIMS depth profiles of
atomic zinc into InP substrates coming from epitaxially grown
InP:Zn layers. They report a diffusion coefficient (at 600C)
of DZn = l-6xl0'13cm2/sec, which is approximately two orders of
magnitude larger than the values they reported for Cd and Mg.
Yang et al. [112] grew a Zn-doped InP layer by MOCVD using DEZn
simultaneously on a (100) and (111B) InP:Fe substrate. The
reported hole concentrations calculated from room temperature
Hall data were p = 6.5xl017 and l.9xl017cm3 on (100) and
(111B) orientations, respectively. Perhaps Zn adsorbs more
strongly on the (100) orientation resulting in a higher hole
concentration. No one reports layer morphology degradation at
high doping levels of Zn. This reason, in combination with
its reasonably wide incorporation range, may explain why zinc
is the most often used p-type dopant. This conclusion is
valid for both low-pressure and atmospheric pressure MOCVD
growth of InP, even though its bulk diffusivity is quite
large.


CHAPTER I
INTRODUCTION
1.1 III-V Semiconductors
Since the invention of the transistor in 1948 by
Shockley, Brattain, and Bardeen, there has been a revolution
in the electronics industry. Up to that time the vacuum tube
diode and triode were the most used electronic devices, but
then the transistor device using a semiconductor crystal as
its starting material was fabricated. The microchip, which is
the fundamental building block of present day computers,
contains a large number of tiny semiconductor transistors
using typically single crystals of silicon as a starting
material. Silicon has been the "workhorse" for the
electronics industry primarily due to its availability in high
single crystalline purity, ease of use in device fabrication,
and of course its good electrical properties. But, the
relatively low electron mobility and fixed indirect bandgap of
silicon makes it not suitable for present-day optoelectronic
device applications. As a consequence of these new demands,
research into the development of semiconductors with variable
electrical and optical properties has flourished.
Compound semiconductors such as GaAs, InP and others
composed of elements from group IIIA (Al, Ga, In) and group VA
1


xml record header identifier oai:www.uflib.ufl.edu.ufdc:UF0009019300001datestamp 2009-03-17setSpec [UFDC_OAI_SET]metadata oai_dc:dc xmlns:oai_dc http:www.openarchives.orgOAI2.0oai_dc xmlns:dc http:purl.orgdcelements1.1 xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.openarchives.orgOAI2.0oai_dc.xsd dc:title Growth and modeling of III-V compound semiconductor optoelectronic materials with device applications dc:creator Howard, Arnold John,dc:type Bookdc:identifier http://www.uflib.ufl.edu/ufdc/?b=UF00090193&v=00001001659152 (alephbibnum)24529515 (oclc)dc:source University of Florida


117
Figure 37; OV profile of a n-type quaternary film on InP.


acknowledges the financial support of BNR, DARPA and
Microfabritech for portions of this study.
Warm, personal thanks go to the author's family,
especially his wife Robin, and his mother Mary, for their
moral support and patience throughout the course of his
graduate studies. This work is dedicated to them and also to
his father, the late Peter V. Howard.
iii


142
electrically active, including that of the diffused Mg at the
[Mg] I017cm'3 plateau. Similar profiles for a highly doped
sample are presented in Figure 45, and show that in the
surface region of the epilayer, where [Mg] is high, a much
lower hole concentration is measured. This result is in
agreement with the relationship between [Mg] and the hole
concentration as observed for layers on S-doped substrates at
high [Mg] (Figure 43).
3.2.3e Photoluminescence Results
The low temperature PL of the Mg doped InP is shown for
several dopant concentrations in Figure 46. The nominally
undoped (n-type) material shows strong donor and free exciton
(D-X) recombination at 874 nm; in addition we see weak band to
acceptor (Zn, e-A) transitions, at 900 nm, related to residual
Zn contaminants[116]. As small guantities of Mg are added,
acceptor-bound excitons (A-X) at 877 nm become apparent[116]
and the dominant e-A transition shifts from 900 nm (e-Zn) to
896nm, corresponding to e-Mg acceptor transitions[117]. In
the middle Mg concentration range, the e-Mg transition
broadens and becomes dominant. At [Mg] > 2xl018cm'3 the
spectra consist of a single broad peak. At the same nominal
excitation conditions, the peak occurs at higher wavelengths
as [Mg] increases. This trend may be related to the existence
of compensating donors at high Mg concentrations as suggested
by C-V measurements.


% of Total Atomic Zinc
175
DEZn Partial Pressure (atm)
Figure 58: InP point defect distribution based on the model
results.


123
lxlO'9 mbar, p-type conversion was not obtained. Apparently,
the Mg sticking coefficient is very low for MBE-InP unlike
MBE-GaAs where Mg has been successfully applied[107]. To
date, Be is the most successful p-type dopant for MBE InP.
3.1.5 Chemical Vapor Deposition
3.1.5a Hydride
The hydride CVD technique has also been used to grow p-
type InP layers as part of optoelectronic device structures.
In the review by 01sen[108], two different methods are
mentioned for p-doping InP by the hydride technique and both
involve zinc. One method uses flowing hydrogen to carry
elemental zinc vapor from a heated "zinc bucket" into the
mixing zone of the reaction tube. A second method mentioned
uses diethylzinc as a p-type dopant source for InP, but there
are no electrical results presented for either technique. A
third technique is presented by Jurgensen et al.[109] which
made use of a Zn doped indium source. The weight fraction of
Zn in the indium was varied from 0.6 to 4.5xl05. The hole
concentrations of several subsequently grown layers from each
In/Zn mixture are presented. Surprisingly, independent of the
amount of Zn added to the indium source, the experimental hole
concentrations scatter around the same curve and appear to
only depend on the length of time that the source was heated.
Hole concentrations start at p = lxl018cm'3 (run #1) and
increase to 2.5xl018cm'3 (run #4). They then decrease to
9xl017cm'3 (run #9) Perhaps as time goes on, from run to run,


156
The effect of MOCVD growth temperature on atomic zinc
incorporation was also studied. Layers of InP p-doped using
DEZn at a fixed molar flow rate of 2.0xl06 moles/min were
grown at temperatures from 550 to 680C. The effect of growth
temperature on atomic incorporation as determined from SIMS
profiles is plotted in Figure 53. As shown, the atomic Zn
incorporation drops from 4xl018cm'3 to 5xl017cm'3 as the growth
temperature is increased. This is the same trend that was
previously discussed and reported for DEZn doping[45] and DMCd
doping[58]. This trend indicates a build up of DEZn, or some
species containing zinc, near the InP:Zn growing surface which
has a temperature dependent equilibrium vapor pressure.
Hence, at higher temperatures, more zinc evaporates from the
surface, or stays in the gas phase, and less gets incorporated
into the solid.
3.4 Modeling of p-tvoe Doping of InP using DEZn
3.4.1 Introduction
Effective control of p-type doping in InP grown by MOCVD
is important in many solid state devices such as lasers,
photodetectors, and heterojunction bipolar transistors. The
process of p-type doping is not fully understood at this time.
A model explaining this process could be useful in selecting
a suitable dopant and/or proper doping conditions for a wide
range of optoelectronic device applications. Experiments have
been performed at BNR and at the University of Florida on the
use of the MOs dimethylcadmium (DMCd) diethylzinc (DEZn) and


192
Figure 63: Spectral scanning system for reflection,
transmission and photoresponse measurements.


Electron Cone, (cm-3)
80
Partial Pressure (Torr)
Figure 17; The effect of H2S partial pressure on InP carrier
concentration.


102
growth Q033, is shown in Figure 29. In this growth a multiple
quantum well (MQW) structure was attempted using GalnAs wells
(60 to 200 thick) which were grown between InP barriers (50G
thick) From the SIMS profile, it is evident that the
position where the phosphorus intensity drops does not exactly
match the position where the gallium and arsenic intensities
increase; the well barrier interface is not well defined.
This was later corrected by growing a similar structure which
made use of the pressure balancing feature of the MOCVD system
(vent and reactor line pressures equal). A TEM cross section
micrograph of growth Q136, which is almost the exact same
structure as Q033, is shown in Figure 30. However, the well
layer thicknesses attempted were much smaller and consequently
a GalnAs quantum well of 13 was grown. This layer took 5
seconds to grow in the MOCVD system using the pressure
balancing mode. As shown in Figure 30, heterostructures with
interface abruptness down to what appears to be the atomic
level, can be grown in the Japan Oxygen MOCVD system.
The PL technique can also be used to characterize the
optical properties of low-dimensional layer structures such as
Q136. When layer thicknesses as small as 13 exist, quantum
size effects can be demonstrated. In Figure 31, a PL spectrum
of sample Q136 is shown. In the figure several quantized
peaks or energy level transitions exist. These peaks are
labelled according to the individual GalnAs quantum wells


114
Table 7
Photoluminescence and Electron Microprobe Analysis of Nearly
Lattice-Matched GaxIn1.xAsyP1.y Films Grown on InP by MOCVD
Run
PL
PL(Xrt)
EPMA
Q035
Ga.437In.563As.960P.040
1.605/xm
Ga .4201n. 580a3.994P. 006
Q119
Ga.330In.670As.761P.239
1.427/xm
Ga.170In.830As.540P.460
Q120
Ga.400In.600As.761P.239
1.554/im
Ga.353In.647As.800P.200
Q164
Ga.390In.610As.895P.105
1.532/xm
Ga.410In.590As.900P.100
Q165
Ga.320In.680As.745P.255
1.400/xm
Ga. 300In. 70(AS. 720P. 280
Q238
Ga.346In.654As.787P.213
1.469Atm
Ga.314In.686As.722P.278
Q239
Ga.301In.699As.706P.294
1.371/xm
Ga.242In.758As.567P.433
Q242
Ga.283In.717As.629P.371
1.324jum
Ga. 0301 n.970a3.042P. 957
Q251
Ga. 159In.841As .350P.650
1.115/xm
Ga.121In.879As.333P.667


124
zinc is being gradually depleted from the In/Zn melt due to
incongruent evaporation. The first three samples may actually
have a higher atomic zinc concentration than sample #4, but
they are compensated due to the incorporation of interstitial
donor complexes. This theory agrees with the observed trend
and also with the trends seen for both bulk-grown and LPE-
grown Zn-doped InP.
3.1.5b Chloride
P-type InP has also been achieved by the chloride CVD
technique. Both Zn and Cd doped InP have been reported by
Chevrier and co-workers[110,111]. In the Zn doped InP paper
[110], 1 gram of Zn was added to the In melt which was heated
to 750C. Hole concentrations are given for consecutive
samples which were grown from the same In/Zn melt. The hole
concentration decreases as the number of runs increases and
this trend is attributed to Zn depletion in the source[110].
SIMS measurement results of the atomic zinc concentration [Zn]
are also given for a few samples, and the ratio [Zn]/p varies
from 2.35 for the low doped sample ([Zn] = 2xl018cm'3) to 9.25
for one of the highest doped samples ([Zn] = 3.7xl018cm'3) .
The low doped (p = 8.5xl017cm'3) sample was annealed for two
hours at 300C and the hole concentration was again measured
by the Hall effect method. It was reported to increase to p
= 2xl018cm'3, which is the same value as the measured SIMS
concentration, [Zn]. The authors propose that the annealing
process activates neutral Zn atoms[110]. It is this writer's


10
quartz tube, decompose in the presence of a heated substrate,
and then deposit an epitaxial layer. Under normal deposition
conditions, the MOCVD process is kinetically limited by mass
transport of the column III source through a stagnant layer
near the growing surface. The MOCVD process is capable of
growing a wide variety of films with excellent abruptness
uniformity over large substrate areas. The principal device
area where MOCVD has made an impact is optoelectronics. A
thorough review of the MOCVD literature has been written by
Ludowise[6] and a brief history of MOCVD with emphasis on InP
based materials and device applications is presented in
section 2.1.
Molecular beam epitaxy (MBE) is a technique capable of
growing epitaxial films one atomic layer at a time. MBE makes
use of controlled evaporation from one or more thermal sources
to direct beams of atoms or molecules onto a heated substrate
under ultra-high vacuum conditions. During a MBE growth the
substrate temperature is generally kept relatively low (500-
600C for GaAs) MBE growth rates are typically slow (0.1 -
2/xm/hr) which in combination with low growth temperatures
permits precise layer thickness, doping and compositional
control[7]. For GaAs, the As4 beam flux is much greater than
the Ga beam flux, and both fluxes are dependent upon the
temperature of the effusion oven, molecular weight of the
emitted atom, orifice area, and source cell to wafer distance.
With a properly placed two-inch rotating wafer, nonuniformity


GROWTH AND MODELING OF III-V COMPOUND
SEMICONDUCTOR OPTOELECTRONIC MATERIALS
WITH DEVICE APPLICATIONS
By
ARNOLD JOHN HOWARD
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1990


191
4.2.3d Spectral Scanning
The experimental set-up used to measure the reflectivity
of the device (see Figure 63) consisted of a tungsten halogen
lamp, a chopper (f=400Hz), a 200mm focal length monochromator
with a 600 lines/mm grating, some focusing elements, a lock-in
amplifier, and a filter to separate the first and second order
of the monochromator exit light. The detector used was a
calibrated BNR 250FE GalnAs p-i-n. A computer operated the
monochromator and lock-in via a GPIB interface, and collected
and stored the data for subsequent analysis. Reflectivity
values were measured over the wavelength range of 850 to
1600nm. This was the useable range for the PIN reference
detector operating with a +5 volt reverse bias. The raw data
was stored and later normalized by division with calibration
data for the detector.
Spectral scans were performed at 3 radial positions of
the two inch wafer (2, 6, and 11mm from the center). The
wavelength position of the central peak maximum changed as a
function of radial position due to the well known drop-off of
growth rate towards the edge of a wafer caused by slight
collimation of the gallium and aluminum beam fluxes incident
upon the substrate. Also, the aluminum and gallium flux
distributions have different drop-off rates resulting in a
variation of the ternary composition away from 30% AlAs.
A change in composition means that the refractive index
and optical thickness will change accordingly. This was


71
Figure 12: The effect of V/III ratio on C-V background carrier
concentration in undoped (n-type) InP.


Current xIO3 (A)
198
Figure 67: I-V and R-V characteristics of a PNPN device
demonstrating bistability.
Resistance xIO5 (2)


86
9xl017cm'3 as shown, is essentially constant throughout the
grown film. The large "dip" in the profile shown in Figure 20
is commonly observed at p-n electrical interfaces. A SIMS
measurement was also performed on this sample (see Figure 21)
and the atomic zinc concentration profile yielded essentially
the same result. Similar results were observed when comparing
the SIMS and C-V profiles of samples grown to study the effect
of the other system parameters on zinc incorporation.
One interesting and useful conclusion that can be derived
from the SIMS and C-V profiles of growth Q095, during which
the growth pressure was varied from 38 to 760 torr, is the
extent of zinc diffusion into the InP substrate. As shown in
Figure 22, the grown layer thickness of this sample is about
1.4/Ln and the zinc doping level is unaffected by the change in
growth pressure. In Figure 23, the atomic zinc level (from
SIMS) is essentially constant throughout the profile, but a
"spike" occurs in the zinc profile right at a depth of 1.4/xm.
The SIMS operators at BNR, where the data was taken, say that
this "spike" is due to silicon-zinc complex which results from
Si on the surface of the InP substrate wafer after cleaning
and it reproducibly indicates the position of the grown layer-
substrate interface. These "spikes" were observed in several
of the SIMS profiles done at BNR. Based on the depth of the
zinc profile in Figure 23, zinc has diffused approximately 0.5
microns into the InP substrate. Using this diffusion length,


130
3.2 MOCVD Growth and Characterization of Mg-Doped InP
Using bis-(MethvlcvclopentadienvH Magnesium as a
Dopant Source
3.2.1 Introduction
Epitaxial layers of Mg doped (p-type) InP have been grown
by the low pressure MOCVD method using MCpjMg as a MO source.
Previously, CpgMg doped InP has been grown by the atmospheric
pressure MOCVD method[58,113]. The atomic incorporation and
subsequent diffusion of Mg in InP has been determined by SIMS.
The surface morphology of grown Mg:InP layers has been
investigated using an optical microscope equipped with
Nomarski phase contrasting and also by TEM. The electrical
characteristics of the layers were measured by the Van der
Pauw Hall effect technique and by an electrochemical C-V
profiler. The optical characteristics of the layers were also
measured by PL at 7 K.
3.2.2 MOCVD Growth
The Mg-doped InP crystal growth was performed by this
investigator in a commercial (CVT, Ltd.), custom designed,
MOCVD reactor at BNR in Ottawa, Canada. The organometallics
TMIn and MCp2Mg were used and kept at 17C, 800 torr and 22c,
700 torr, respectively. Fifteen percent phosphine diluted in
UHP hydrogen was also used as a source gas. The carrier gas
was hydrogen and the total H2 flow rate was electronically
controlled to be 7 SLM. A radial manifold existed at the
inlet to the horizontal quartz reactor. The reactor was held
at a pressure of 75 torr and RF inductively heated to a growth


Mg Concentration (cm
132
CO
I
Figure 38: Atomic magnesium SIMS profile of sample B232 which
was grown on a S-doped InP substrate.


77K Mobility (ca2/Volt Sec.
75
70000
60000 -
X
50000 -
X
40000 -
30000 -
20000 +"
500
4 '
X

X
X
X
I 1 r-
550 600 650
Tg (c)
700
Figure 15: The effect of growth temperature on 77 K mobility
of undoped InP (bars indicate the range of data).


Lattice Parameter (ft)
5
Figure 1; Lattice parameter and bandgap energy of various
III-V semiconductors


179
Figure 59; Optical theory of an interference filter.


119
this dilemma, an attempt has been made to understand the
process of p-type doping of InP with the ultimate goal of
identifying an "ideal" p-type dopant source. A review of the
literature on bulk crystal growth, LPE, MBE and CVD (metal
organic, chloride, and hydride) of p-type doping of InP has
been performed. From this search, data on the relationship
between the dopant distribution in the solid phase and the
growth conditions from either the liquid or gas phase have
been used to determine if the process is at equilibrium,
reaction limited, or transport limited. For gas phase growth,
it is evident that dopant incorporation occurs by a reaction
or transport limited process. For the liquid phase growth,
incorporation occurs from a near equilibrium process at the
solid-liquid interface which is also transport limited in the
liquid bulk.
3.2 Bulk Crystal Growth
P-type InP substrates have been used for radiation
resistant solar cells[97], buried heterostructure lasers[98],
and other modern optoelectronic device applications. Liquid
encapsulated Czochralski (LEC) pulling is the most widely used
technique for growing bulk p-InP substrates. Usually a layer
of liquid B203 seals the heated InP melt and a seed crystal is
dipped into the melt and slowly pulled out at a rate which
controls the crystal diameter. Cd, Zn, Be, Mg and Mn have
been used as p-type dopants for InP bulk crystals[99]. Zn is
the most commonly used dopant due to its experimentally


128
substrate doping-level dependent deep diffusion of Mg, a
decrease in hole concentration and stacking fault formation at
high Mg concentrations indicating strong compensation possibly
from a Mg donor-like complex. More details of the MCpgMg work
are presented in the following section of this text. Due to
the uncontrollable super-linear incorporation rate of Mg
during both atmosphere and low pressure growths, and the
compensation and layer morphology degradation at high [Mg] for
low pressure growths, Mg may be unsuitable for certain device
applications.
Zinc doping of MOCVD InP is the most often reported p-
doping method. Two sources, DEZn[45,58,112] and DMZn[58],
have been used in either the bubbler or diluted in a high-
pressure cylinder configuration. Nelson and Westbrook[58]
used both configurations (DEZn in a bubbler at 0C and 750 ppm
DMZn in H2 in a high pressure cylinder) and reported little
difference between the use of either Zn precursor. They
obtained hole concentrations of p = 4xl017 to 2xl018cm3
(calculated from Hall effect data) using Zn, with a square
root dependence on DEZn or DMZn vapor flow from 2xl0'8 to
lxlO6 moles/min. Above lxlO'6 moles/min, the Hall hole
concentration became saturated at 2xl018cm'3 which is similar
to the saturation level (2-3xl018cm'3) reported by the other
crystal growth techniques already discussed. Razeghi and
Duchemin[45] observed a wider range of incorporation p = 0.2-
4.0xl018cm3 from their experiments using organometallic DEZn.


61
nitrogen blowoff; (3) procedure (2) without the nitric acid in
methanol etch; (4) procedure (2) followed by a 1% bromine in
methanol etch for 6 minutes, methanol rinse, filtered nitrogen
blow dry; and (5) 2 minutes of surface treatment with the
UV/ozone cleaning system (UVOCS Inc).
Under a microscope at 200X magnification, particles were
observed on the surfaces prepared using procedures (1) and
(2) Procedure (4) resulted in a wavy surface probably due to
uneven bromine etching. Samples prepared using procedures (3)
and (5) had the best surfaces. Fresh substrates were cleaned
using procedures (3) and (5), loaded into the reactor, and
undoped InP was grown on both of them at the same time. This
experiment was performed several times and layers grown using
procedure (3) were consistently equal to or better than layers
grown using procedure (5). Procedure (3) was chosen as the
optimum and the exact details of this procedure are given in
Table 4.
The effect of growth conditions on the uniformity of
undoped epitaxial InP on InP was also studied. The total
volumetric flow rate was the only parameter that had any
significant effect on layer uniformity. Layers grown using
total hydrogen flow rates of 3, 5, 7 and 9 standard liters per
minute (SLM) had corresponding thickness variations of +/-10,
8, 6 and 6%. The increased variation at the lower flow rates
probably is associated with an increased boundary layer


52
either power fails, compressed air pressure drops, cooling
water pressure drops or temperature increases, the machine
will automatically alarm and all air operated valves will
close. There are smoke detectors in the machine and fire
detectors in the room. There are pressure sensors on the
hydride lines, in the reactor, and on the exhaust line and for
each, if a certain pressure value is exceeded, an alarm will
sound and the machine will shut down. There is also a pH, ORP
and temperature sensor on the scrubber which triggers an alarm
in the clean room if any of these values are out of the safety
range. Finally, there is a compressed breathable air supply
always on hand for reactor or source changes and two SCBAs
available for the emergency response team. It is evident that
safety is a big concern and since the machine was constructed
in Japan, there is even an earthquake sensor attached to it.
2.5 Determination of Optimum Growth Conditions Based on Thin
Film Characterization
2.5.1 Experimental Method
There are several experimental parameters that must be
determined before performing a MOCVD growth. Using the
simplest case as an example, undoped InP on InP, the first
thing that must be decided is what type of substrate is
required. For all of the GaxIn1_xAsyP1_y on InP experiments
performed in the Japan Oxygen MOCVD system, InP oriented (100)
2 towards (110) purchased from Sumitomo Inc. were used. Both
semi-insulating (iron doped) and n-type (n 8xl018cm'3,
sulfur


Mg Concentration (cm
135
CO
i
1020
1019
1018
1017
1016
1015
0 12 3
Depth (/urn)
Figure 40: SIMS Mg profiles of InP layers on S-InP (symbols
represent H2 flow rates of 5, 12.5, 22 and 27.5
seem to the Mg bubbler, respectively).


APPENDIX B
OPERATION OF THE QUATERNARY MOCVD SYSTEM
B.l SCOPE
This document pertains to the quaternary MOCVD
housed in building 771 on the campus of the University of
Florida.
B.2 PURPOSE
The purpose of this document is to outline the correct/
safe operating procedures for the MOCVD system which is used
for research on optoelectronic materials.
B.3 SAFETY
No one should operate any part of this system without
prior approval of, and training by Dr. Tim Anderson. The
system has its own built-in safety system which will shut down
upon detecting any potentially dangerous situation. If
something related to the system seems to be malfunctioning,
and you are not trained to operate the system, immediately
(hit the red emergency stop, advise others to leave the room
with you), and contact the present operator, C. Heinz, Dr. T.
Anderson (392-2591) or K. Rambo).
If a problem exists in the room such as a fire, also hit
the red emergency stop button. Of course, if the reactor
234


100
Table 6
Growth Temperature and V/III Ratio Studies on
GalnAs/InP Material (H2 = 5SLM, Pg = 80 Torr)
V/III = 50
Tg=
600C
650C
700C
/iRT (cm2/volt-sec)
6067
4083
3411
Nd Na (cm'3)
3 1015
8 1015
11016
Tg = 600C
V/III =
12.5
25
37.5
50
/iRT (cm2/volt-sec)
*
4763
5141
4083
Nd Na (cm-3)
*
4.5* 1015
3.6* 1015
8 1015
Note: *
Material was very poor and ohmic contacts were
not possible.


CHAPTER III
P-TYPE DOPING OF MOCVD INP: EXPERIMENTS AND MODELING
3.1 A Review of the Literature on p-Tvpe Doping of InP
3.1.1 Introduction
Epitaxial layers of the III-V compound semiconductors
InP, GalnAs and GalnAsP with compositions lattice-matched to
InP are widely used for the fabrication of optoelectronic
devices. Many optoelectronic devices rely on the creation of
an electrical p-n junction using p-type InP. InP doped p-type
by MOCVD and other crystal growth techniques can be made, but
the process is not understood in detail. At this time, an
"ideal p-type dopant with the following characteristics does
not exist: (1) a wide controllable doping range (p = 1015-
I019cm'3) ; (2) a low bulk diffusivity (D < I0'15cm2/sec) ; (3)
a controllable incorporation rate for a wide range of growth
conditions (growth temperatures, V/III ratios, group III
concentrations); (4) full electrical activity (p = atomic
concentration); (5) no memory effects in the system; and (6)
one that does not degrade layer morphology by the formation of
extended defects.
Several material sources have been used to incorporate
Hg, Cd, Mg, Zn, Mn, Be, Cu, Ca and C atoms as p-type dopants
in InP, with varying degrees of success[94-96]. Because of
118


79
X (um)
Figure 16: C-V profile of a H2S doped InP film.


The use of diethylzinc (DEZn), bis-(methylcyclo-
pentadienyl) magnesium (MCpjMg) and dimethylcadmium (DMCd) as
p-type dopant sources for MOCVD InP was investigated at BNR in
Ottawa, Canada. It has been experimentally observed that the
carrier concentration dependence on dopant partial pressure in
the MOCVD reactor is different for each of these three
dopants. A novel model of the p-doping process of MOCVD InP
using DEZn has been developed that incorporates an equilibrium
boundary condition between the gas phase and solid phase point
defects. The results of this model indicate that at high DEZn
gas phase mole fractions, which results in low solid-phase
electrical activity, the dominant electrically inactive point
defects are intersticial zinc and zinc complexed with a
phosphorous divacancy.
A novel optoelectronic device has been fabricated and
modeled which contains p-n heterojunctions in an optical
interference filter. Structures were grown by molecular beam
epitaxy at BNR using the GaAs/AlGaAs material system and by
MOCVD at the University of Florida using the InP/GalnAsP
material system. Structures with peak reflectivities at 1.3
and 1.40 microns were grown and good crystalline quality were
confirmed. Electrical bistability was observed in a forty-
layer device which has never been reported before in a
structure of this size.
vii


CONCEmTHATION (atomo/cc)
210
PROCESSED DATA bnw
9 Mov 88 Ca FILE: MSE572EX
Figure 74; SIMS profiles of zinc and aluminum from a portion
of the MBE572 stack structure.


1.0
>1
p
H
>
H
P
0
Q)
rH
o.o
Wavelength (ran)
Figure 75: Theoretical reflectivity spectrum for sample Q240
(neglecting GalnAsP absorption).


170
temperature (Brouwer's approximation[136]) p and [ZnIn']. So,
a new simpler relation is realized:
P = [ZniJ (23)
which only applies at room temperature. Useful SIMS data on
InP:Zn samples taken at room temperature was also available.
It must be assumed that zinc related defect concentrations at
the growth temperature will be approximately the same at room
temperature. The room temperature total zinc concentration,
[Zn] (by SIMS), is then given by the following equation:
[Zn] = [Zn"n] + [Znf] + [ZnInVp] + [VpZnInVp] (24)
The room temperature electroneutrality equation (equation 23)
does not apply at the growth temperature. The full form of
equation 22 in Table 8 must be used to calculated the hole
concentration at the growth temperature, but it can be
simplified based on an order of magnitude analysis. Hurle
[132] states that group III vacancy concentration [Vln] can be
neglected relative to the group V vacancy concentration. So,
the growth temperature (gt) hole equation now is:
+ [Znfn] [Vp+]
Pgt = ngt
(25)


UF Libraries: Digital Dissertation Project
Page 2 of 2
Internet Distribution Consent Agreement
In reference to the following dissertation:
AUTHOR: Howard, Arnold
TITLE: Growth and modeling of III-V compound semiconductor optoelectronic
materials with device applications / (record number: 1659152)
PUBLICATION DATE: 1990
I,
/IzMu (2.
as copyright holder for the aforementioned
dissertation, hereby grant specific and limited archive and distribution rights to the Board of Trustees of
the University of Florida and its agents. I authorize the University of Florida to digitize and distribute
the dissertation described above for nonprofit, educational purposes via the Internet or successive
technologies.
This is a non-exclusive grant of permissions for specific off-line and on-line uses for an indefinite term.
Off-line uses shall be limited to those specifically allowed by "Eair Useu as prescribed by the terms of
United States copyright legislation (cf, Title 17, U.S. Code) as well as to the maintenance and
preservation of a digital archive, copy. Digitization allows ihe University of Florida to generate image-
and text-based versions as appropriate and to provide and enhance access using search software.
This grant of permissions prohibits use of the digitized versions for commercial use or profit.
Signature of Copyright Holder
/~}/Zmq u> sj.
Printed or Typed Name of Copyright-Holder/Licehsee
Personal information blurred
7h /a gr
Date of Signature
Please print, sign and return to:
Cathleen Martyniak
UF Dissertation Project
Preservation Department
University of Florida Libraries
P.O. Box 117007
Gainesville, FL 32611-7007
7/3/2008


140
Figure 43: The relationship between atomic [Mg] and hole
concentration for InP:Mg layers.


8
laser applications. However, problems such as surface
defects, poor thickness and compositional uniformity, and
difficulty in growing abrupt heterojunctions have made the LPE
technique unsuitable for present-day device fabrication
demands. For simple layer structures such as the AlGaAs solid
state laser used in compact disc players, LPE is perfectly
adequate.
There are three distinct VPE or CVD chemistries: chloride
CVD, hydride CVD and metal organic CVD. The chloride (or
sometimes referred to as halide) CVD process for GaAs growth,
as an example, uses AsCl3 and metallic Gallium as sources in
an open tube system with H2 (as a carrier gas) to transport
reactants from the source zone, through a temperature gradient
zone to the deposition zone. The chloride CVD process is a
surface-kinetically limited process requiring careful source
composition control and accurate temperature control
throughout the system for reproducibility. Also, it is
difficult to vary the V/III ratio and transients are long so
abruptness is bad in chloride CVD. GalnAsP has been grown by
the chloride CVD method[4] but other CVD techniques are more
convenient and flexible for growing ternary and quaternary
III-V compounds. Hence, the chloride process is usually only
used to grow high purity epitaxial GaAs.
The hydride CVD process for growth of III-V compound
semiconductors differs from the chloride process by replacing
column V chlorides such AsC13 or PC13 with column V hydrides


95
Figure 25: C-V profile of growth Q005 (lattice-matched n-
type GalnAs on InP).


45
The gas delivery system for the metal organic sources
trimethylindium, triethylgallium and p-dopant diethylzinc
(which are held in stainless steel bubblers in temperature
controlled baths) consists of pneumatic, needle and manual
valves which are attached to the inlet and outlet ports of the
bubblers. Each metal organic line also has a pressure sensor
attached to it just before the inlet of the bubbler. The
metal organic sources are solids or liquids at room
temperature with fairly low, temperature dependent equilibrium
vapor pressures. By varying the temperature of the bubbler
bath, the hydrogen flow rate through the bubbler, and the
pressure of the bubbler region of the gas delivery system (by
opening or closing the needle valve), a controllable range of
metal organic source flow rates can be attained.
Since the hydrides (arsine, phosphine) and n-type dopant
(1000 ppm H2S diluted in H2) are highly toxic, combustible and
at high pressure, the gas delivery system for these sources is
slightly more complicated than for the metal organics. Each
line has a regulator, air operated valve and a manifold
attached to it. The manifold contains a high-pressure, high-
purity nitrogen purge line, a hydrogen purge line, and a third
line which can be used to evacuate the hydride line or vent
the hydride source directly to the scrubber. During normal
"standby" operation, palladium-alloy diffused hydrogen is
purging the hydride line to the vent. Only during toxic gas
flow is the hydrogen purge interrupted. Check valves on the


237
initial start up (leak check, bake out of graphite susceptor,
scrubber N2 purge, individual components testing). It is also
assumed that the reader is familiar with the physical layout
of the system (see Figure 79) and what the individual system
components do. Manual valves (MV), air valves (AV), needle
valves (NV) and regulator valves (RV) are all mentioned in the
procedure (also see Figure 79).
B.5.1 Power Supply and Programmable Controller Explanation
Turn on the 115V and 208V circuit breakers. Turn on
breakers ELB 100 and 200. After the 115 and 208V lights are
illuminated, turn on fuse breakers NFB102, 112 and circuit
protector CP132 which are all located inside the lower right
electrical panel (when facing the front of the machine). Push
the "stand by", "PC ON" then "ON" switches and then all
switches, readouts and emergency sensors will begin to
function. If the "ON" switch is not lit, the system will
function as it stands, but all air operated valves will remain
open (red light) or closed (no light) even if they are pushed
accidentally. The system should be in manual mode now. For
a description of edit, auto modes or how to operate the
programmable controller, see Hayakawa[164]. Check reactor
pressure, if not slightly greater than or equal to 760 torr,
turn on RP2 and open valves from RP2 inlet towards the reactor
to prevent back flow to the reactor. Completely evacuate the
system, start hydrogen flow and then pressurize the system by
closing AV208,9 and then opening AV204 when p = 780 torr.


I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Timothy J. Anderson, Chairmen
Professor of Chemical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Chemical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctor of Philosophy.
Sheng S. Li
Professor of
Electrical Engineering
I certify that I have read this study and that in my
opinion it conforms to acceptable standards of scholarly
presentation and is fully adequate, in scope and quality, as
a dissertation for the degree of Doctqrof Philosophy.
Gys Bosman
Associate Professor of
Electrical Engineering


63
thicknesses resulting in depletion of TMIn near the growing
surface. This theory is confirmed by the observation that the
growth rate was lower, 0.95 and 1.2/m/hour for 3SLM and 5SLM,
than it was at higher flow rates (1.4/m/hour for 7SLM and
9SLM). The uniformity of undoped InP grown using a total
hydrogen flow rate of 7SLM is presented in Figure 7. A total
flow rate of 7SLM for hydrogen was chosen as optimum over 9SLM
because with 9SLM at a fixed TMIn mole fraction almost 3 0%
more TMIn material would be required.
The effect of several parameters on InP growth rate and
interface quality was also studied. As already stated the
total hydrogen flow rate has an effect on the growth rate.
The growth pressure and V/III ratio appear, however, to have
little to no effect. The growth temperature has a nonlinear
effect on the growth rate? at 450C the growth rate was
0.8/m/hour, but going from 550 to 750C the growth rate only
changed from 1.4 to 1.5/m/hour. This can be explained as
follows: at low temperatures (below 550C) InP growth is
kinetically limited and consequently temperature dependent.
At substrate temperatures above 550C, InP growth is Tmln
transport limited. This is confirmed by Figure 8 which shows
the linear relationship between InP growth rate and TMIn mole
fraction grown at 600C. Some of the data for Figure 8 was
taken from the SEM micrograph shown in Figure 9. Figure 9 is
a scanning electron micrograph of a cross section from growth


(36)
R =
nsl n1
2N
1 + !,
n
n2
n
1
2N
where N is the number of layer pairs and n,, n2 are the layers
as shown in Figure 59. It is important to note that if n2/n.,
decreases and/or N is increased, R increases. Also, a larger
value of R will be realized if the stack is ordered such that
n2/n1 is less than unity. It is significant to note that lower
values of the n2/n., ratio will also result in narrower central
bandwidths. The GaAs/AlGaAs material system was used because
of its compatibility with optoelectronic devices. Results of
GaAs/AlGaAs interference filters are presented in the next
section of this text. All GaAs/AlGaAs layers were grown by
MBE at BNR.
MOCVD grown GalnAsP/InP material was also investigated
and the results of filters grown from this material system are
presented in section 4.4.3 of this document. The quaternary
MOCVD system at the University of Florida is equipped with
sources making it possible to grow thin alternating layers of
GaxIn1.xAsyP1.y and InP for interference filter structures.
This is significant because with the lattice-matched GalnAsP/
InP material system, a lower refractive index ratio (n2/n, =
0.9042) is possible (with light at 1300nm) than with the
GaAs/AlGaAs system (minimum ratio of 0.9240). This makes it


RELATIVE IM1ENSITY
96
12 j
ioi X= 1.67 nm @300 K
8-
(0
kO
lO
r
m
to
<0
0.81016 eV
h 3.47mv fwhm
4 -
0 1-
J
5600 5800
6000
6200
6400
6600
6800
7000
Wavenumbers (cm 1)
Figure 26
PL spectrum of GalnAs lattice-matched to InP
measured at 4.2 K.


136
In Blaauw et al.[115], results were presented on the diffusion
of Zn in InP. It is proposed that the presence of sulfur and
silicon donor atoms in InP can act as traps and immobilize
zinc. As shown in Figure 40, apparently magnesium can also be
trapped up to a level corresponding to the substrate donor
concentration (n=lxl019cm*3) Also the depth of diffusion (1
to 2 microns) appears to be determined by the total amount of
Mg diffusing across the substrate/epilayer interface and
depends on the diffusion time (one hour) and the epilayer
doping level.
The increase in Mg concentration with epilayer depth for
the Mg profile at the lowest doping level in Figure 40 was
observed in all low concentration Mg-doped layers grown on
both S-doped and Fe-doped substrates. This phenomenon may be
related to a Mg gas phase depletion reaction which occurs with
an increasing rate as the reactor wall gradually gets coated.
3.2.3c Ma Diffusion in Fe-dooed InP Substrates
The atomic Mg concentrations of InP:Mg layers grown on
InP:Fe substrates were the same as the layers simultaneously
grown on InP:S substrates. The Mg diffusion depths however
were much greater, up to 32/xm deep for the layers grown on
InP:Fe substrates. The SIMS profiles for four different
samples are shown in Figure 41. As shown, similar to the
diffusion in InP:S, the Mg is immobilized but at a lower level
of I017cm'3 and then drops off abruptly to the instrumental
detection limit ( I015cm'3) The I017cm'3 plateau must be


-Nn (cm
147
Figure 47: The effect of DMCd molar flow rate on InP hole
concentration. The circles are data taken from
Blaauw et al.[59]


97
6533.556cm"1, which corresponds to an energy of 0.81016eV, and
the changes in bandgap with respect to temperature (-2.285xl04
eV/K), the room temperature emission wavelength of this
material is 1.67/m; this is the emission wavelength of lattice
matched Ga ^In 533As on InP. Hence, PL can be used to
determine ternary material composition. Also, the full peak
width at half of the peak's maximum (FWHM) of 3.47meV is
another indication that this layer is of excellent quality.
Finally, the fact that no other peaks exist in the spectrum
indicates that this sample has a very low impurity level.
The PL technique can also be used on lattice-mismatched
samples. Figure 27 shows a PL spectrum of growth Q148 which
was determined by the XRD technique to have the lattice-
mismatched composition Ga 435In 565As. The PL spectrum agrees
with XRD as the room temperature wavelength equivalent of
6380.88cm'1 wavenumbers is 1.714/im which would be indium
arsenide-rich material. Unfortunately, the temperature
dependence of mismatched GaxIn1.xAs is unknown so an exact
composition cannot be calculated. However, when comparing
this spectrum with the one in Figure 26, it is clear based on
the FWHM = 7.94meV being larger and that two peaks exist
instead of one perfect crystal transition, that the material
from growth Q148 is certainly inferior to that of growth Q144
(Figure 26). When comparing PL and XRD, the PL technique is
more accurate as it can correctly indicate that an epitaxial
layer is lattice- mismatched when XRD incorrectly does not.


PL Intensity (arbitrary units)
144
Figure 46; PL spectra of InP:Mg layers measured at 7K.


2
(P, As, Sb) columns of the periodic table have electrical and
optical properties superior to those of silicon for certain
modern-day device applications. III-V materials have a wide
range of bandgap energies (0.18 to 2.4 eV), where the bandgap
energy is defined as the energy difference between the lowest
electron state in the conduction band and the highest hole
state in the valence band allowed in the semiconductor. Some
compound semiconductors have direct bandgaps, meaning that the
conversion of photons (light) to electrons (energy) or vice
versa, does not involve a third particle, such as a phonon.
The direct bandgap III-V compounds also have large electron
mobilities where mobility is defined (at low electric fields)
as the ratio of absolute electron velocity to the magnitude of
the electric field. A listing of these parameters and the
lattice constants of silicon and binary III-V semiconductors
is shown in Table 1(1]. As shown in this table, as much as a
two order of magnitude increase in electron mobility is
possible by using III-V compound semiconductors instead of
silicon for electronic devices. It is also significant to
note that a wide range of compound semiconductors can be
formed by creating solid solutions of the individual
semiconductors? hence, a wide selection of compound
semiconductors exists with a wide range of electrical and
optical properties.


161
101,1
CO
*E
o,
c
o
cS
o
c
o
o
o
c

o
Q.
10
18.
10
17.
10
16.
SIMS
a

C-V or Hall
i r 1111 n 1 i i" i 111 m" 1 " i11 ir
10'* 10'8 10'7
10'6 105
DEZn Partial Pressure (atm)
Figure 55; SIMS and hole concentration data for InP:Zn which
is used for the model evaluation.


236
scrubber on the back pad by turning it on. Is, the pH > 10,
ORP > 100, pumping pressure > 28 psi, liquid level sufficient?
If not, see scrubber manual for maintenance. Check the
pressure on the inlet H2 and N2 lines, the hydride cylinder
sources. Do we need a bottle changes or six-pack hydrogen
changes? Check the water filter on the chilled water line,
clean if necessary. Check the temperature of the metal
organic source baths. Does growth rate or doping level
changes indicate the need for a source change? Is the
hydrogen purifier operating properly? Is the reactor tube
clean and dry and in place? Leak tight seal? Purged with
H2. Has all source changes or maintenance been followed up by
a helium leak check? Did the whole system pass the leak
checking procedure? If any seal has been broken on a toxic
gas line, wear air masks during first operation of that line
after leak checking. How is the oil level in the roughing
pumps? Does the oil need changing? Change it before and
after growths. Is there a clog? Have you programmed the
process controller with your desired deposition scheme? Test
it. (Programming tips are available in Hayakawa[164]).
B.5 PROCEDURE
B.5.1 Introduction
Now that all the facilities are in place, all safety
precautions have been followed/checked, and the materials are
installed, the normal operation procedure will be outlined.
It is assumed that the system has already gone through its


58
Figure 5: Nomarski photographs of an InP plane view surface
and stained edge cross section taken at 2000X
magnification.


22
2.3 A Review of the Literature on InP Based MOCVD
2.3.1 InP Homoepitaxv
The first reported growth of indium phosphide (InP) by
MOCVD was in 1969 by Manasevit and Simpson[12]. This work as
well as other early efforts[33-35] used triethyl indium (TEI)
and phosphine (PH3) as indium and phosphorus sources. For
several years, problems such as low, nonuniform growth rates
and high impurity levels were encountered. One source of
these problems was found by relating the observation of a
white smoke at certain growth conditions to the uncontrolled
gas phase reaction between metal organic indium sources and
phosphine. This reaction occurs at low temperatures <100C,
is parasitic in nature, and produces a non-volatile liquid
polymer.
One method used to minimize this problem was the use of
low pressure reactor systems[36-38] to decrease the residence
time of unreacted species upstream of the growth region.
Another method involves the use of adducts such as TMI-TMP
[39] or TMIn-TEP[40] as indium sources which will not complex
with PH3. Another technique is to keep the reactants apart
and only let them mix just prior to the growth region. This
method can, however, lead to uniformity problems. The most
recent improvement is the use of TMIn (a solid powder at room
temperature which melts at 88C) instead of TEIn (a liquid at
room temperature), as the indium source[41-43]. TMIn also
decomposes at a much higher temperature (>300C) than TEIn


260
[136] Brouwer, G. Philips Res. Rep. 9 (1954) 366.
[137] Kroger, F. A., Chemistry of Imperfect Crystals. North
Holland, Amsterdam, Holland, 1964.
[138] Constantinides, A., Applied Numerical Methods with
Personal Computers. McGraw-Hill, New York, NY, 1987.
[139] Dobrowolski, J. and Driscoll, W.(editor), Handbook of
Optics. McGraw-Hill, New York, NY, 1977, Section 8.
[140] van der Ziel, J. P. and Ilegems, M., Appl. Optics 14
(1975) 2627.
[141] Yoffe, G. W., Schlom, D. G. and Harris, J. S., Jr., 1987
Ann. Dev. Res. Conf., Santa Barbara, CA., paper 3A-4.
[142] Thorton, R. L., Burnham, R. D. and Streifer, W., Appl.
Phys. Lett., 45 (1984) 1028.
[143] Tai, K., McCall, S. L., Chu, N. G. and Tsang, W. T.,
Appl. Phys. Lett. 51 (1987) 826.
[144] Ritchie, S., Spurdens, P. C., Hewitt, N. P. and Aylett,
M. R., Appl. Phys. Lett. 55 (1989) 1713.
[145] Born, M. and Wolf, E., Principles of Optics. Pergammon,
New York, NY, 1964.
[146] Afromitz, M. A., Solid State Comm. 15 (1979) 59.
[147] Broberg, B. and Lindgren, S., J. Appl. Phys. 55 (1984)
3376.
[148] Cardona, M., Modulation Spectroscopy. Academic Press,
New York, NY, 1969.
[149] Katz, J., Margalit, S. and Yariv, A., IEEE Trans.
Electron Dev. ED-29 (1982) 977.
[150] SpringThorpe, A. J. and Mandeville, P., J. Vac. Sci.
Tech., B4 (1986) 853.
[151] SpringThorpe, A. J., Ingrey, S. J., Emmerstorfer, B. and
Mandeville, P., Appl. Phys. Lett., .50 (1987) 77.
[152] Hill, M. J., Ph.D. Thesis, Durham Univ. England (1985).
[153] Sell, D. D., Casey, H. C. and Wecht, K., J. Appl. Phys.
45 (1974) 2650.
[154] Cross, M. and Adams, M. J., Opto-Electron 6 (1974) 199.


17
the stagnant layer thickness decreases. Also, at lower
pressures, diffusion coefficients increase, making more abrupt
heterojunctions possible. Low pressure operation also reduces
the occurrence probability of unwanted parasitic reactions.
Because of all of these advantages, low pressure MOCVD growth
has become widely used especially for multi-wafer scale-up
production applications.
Another problem with the traditional horizontal "Bass
type" [17] reactor is due to the fact that having a cold dense
gas above a hot, less dense gas is unstable because of
gravity. This can cause natural convection which results in
closed stream-line gas flow patterns. This problem can
usually be minimized by operating at reduced pressures[25].
The best solution, however, appears to be the use of an
inverted reactor geometry[26] which completely eliminates
thermal buoyancy effects. In this geometry, the susceptor is
located at the highest and hottest point of a horizontal
reactor with the wafer mounted upon it facing downwards.
Another obvious benefit with this design is the elimination of
the problem of particles falling on the substrate before and
during epitaxial growth which can lead to structural defects.
With this design, improved GalnAs compositional uniformity and
a complete elimination of parasitic deposition on the quartz
wall opposite the susceptor have also been reported[26].
Another final technique that has been used to improve both
thickness and compositional uniformity is the use of moving


239
B.5.2 Definition of the "OFF" or "OVERNIGHT State
The complete system should be purging through valves
AV200, 204 at 760 torr with hydrogen. All source MVs should
be closed (metal organics, hydrides) and AV111 (nitrogen
inlet) should also be closed. Roughing pumps and turbo
molecular pumps should be off. The scrubber should be off,
Lepel RF generator also off. Water flow to Lepel, reactor and
load lock and to bubblers also all off. Hydrogen purifier
should always be on. Close MV14, 24, 34, as they should
always be during toxic gas flow (double check).
B.5.3 How to Grow InP on InP in the MOCVD System
Turn on the scrubber, and check the hydrogen supply on
back pad. Enter the clean room and check the MDA and hydrogen
detector. Flow water to the system, reset water alarm, and
make the deposition program now. Turn on the MO bubbler
controllers and set the temperatures. Turn off the variacs
for the heat tapes. Prepare the substrates. Now put the
substrates into the load-lock room. Evacuate the load-lock
room and close MV102, AV109, then turn RPI and RP2 on, also
open MV201, MV202, NV202 and evacuate to minimum pressure.
Flow N2 gas into the load-lock room and close MV201,
MV202, NV202, then open MV102, MV202, NV202. When the
pressure of the load-lock room becomes 760 torr, close NV202,
MV202, MV102. Load the substrates into the load-lock room
using clean room paper and gloves, close the door tightly.
Evacuate the load-lock room by opening MV201, MV202, NV202


66
Figure 9: SEM micrograph of growth Q054 at 20,000X.
layers are InP, dark are GalnAs.
Light


42
Figure 3: Quaternary MOCVD simplified flow diagram.


24
quartzware. Coupled with the use of palladium-alloy diffused
hydrogen as a carrier gas, these precautions usually eliminate
the system as a source of high background impurity level
problems. In addition to high purity equipment, ultra high
purity sources contained in stainless steel bubblers and
corrosion resistant coated cylinders are required. For inP,
phosphine with five nines purity (99.999%) and diphos purified
(doubly sublimed) trimethylindium are both commercially
available. As purification technologies advance, then
progressively lower background doping levels surely will
follow.
The most important materials issues in the InP growth
area are the effect of growth conditions, substrate quality,
substrate orientation and substrate wafer cleaning techniques
on material quality. Several papers have been published on
each topic and the basis of comparison presented usually
involves characterization results of thin films such as room
temperature (300 K) and/or liquid nitrogen temperature (77 K)
mobilities and undoped carrier concentrations (NDNA) cm3, etch
pit densities (EPD), photoluminescence (PL) intensities and
occasionally device performances.
The effect of growth conditions on properties of InP
grown by MOCVD has been studied by several research teams[45-
48]. Razeghi and Duchemin[45] showed that the growth rate of
InP is linearly dependent on indium metal organic reactor
partial pressure and independent of the phosphine partial


233
A.5 Conclusions
The Hall effect has been used for many years as a way of
characterizing semiconductor devices. It is a reliable,
reproducible method with the only real drawback being its
difficult, destructive contacting procedure. For both an InP
and GaAs samples which were grown and characterized by an
industrial lab, the measured mobility, carrier concentration
and resistivity values differed by at most 10% from the
company's measured results. This shows that the described
contact procedure, measurement routine and calculation method
should be adequate for most applications.


122
Wada et al.[103] also presented Zn doping results for LPE
InP. They report a distribution coefficient of K(Zn) = 0.84
for a Zn fraction in the melt of 102 atomic percent which
corresponds to a hole concentration of 2xl018cm'3. Above this
Zn fraction, they report constant then decreasing Hall hole
concentrations and attribute this trend to strong compensation
effects. Interestingly, this is the same trend that was
observed for bulk crystal growth of Zn doped InP. Wada et al.
[103] suggest (without evidence) that the decrease in carrier
concentration may be due in part to precipitate formation, but
is more likely due to compensation by an interstitial zinc
donor complex as proposed by Hooper and Tuck[104].
3.1.4 Molecular Beam Epitaxy
P-type doping of InP grown by MBE has been attempted
using several different group II atoms. Zn and Cd were used
as dopant atoms by Park et al.[105] emanating from a low
energy ion cell during InP MBE growth. Doped films remained
n-type due to the near zero sticking coefficient of these
atoms on InP at the growth temperature ~ 350C. Be doping of
MBE InP grown at 525C has been reported by Panish et al.
[106]. They report successful p-type conversion using Be with
hole concentrations from p = 6xl016cm3 to p = lxl019cm3 with
corresponding Hall mobilities of /ih = 100 to 30 cm2/volt-sec.
Recently, Mg has been used as a potential p-type dopant for
MBE InP grown at 500C. Cheng et al.[107] reported for all
layers up to the maximum beam equivalent pressure of Mg =


59
adequate phosphine decomposition and at higher temperatures,
slightly higher growth rates were the causes for inferior
material quality. The total hydrogen flow rate and growth
pressure were also varied but had little to no effect on
surface morphology. The TMIn mole fraction was varied and at
values greater than lxlO'4 resulting material quality degraded
probably due to increased growth rates. The V/III ratio had
the most pronounced effect on layer surface morphology. This
effect can be clearly seen in Figure 6 which shows two inP
surfaces grown at identical conditions, except that the V/III
ratio was almost tripled by increasing the phosphine flow rate
to the reactor from 35cm3/min to 100cm3/min. This is in
agreement with the growth temperature study which concluded
that poor material results from insufficient phosphine
decomposition.
Another study performed with the goal of improving the
surface morphology of undoped InP on InP was varying the wafer
preparation procedure employed. Substrates were prepared
using five different techniques and then viewed under the
microscope. The procedures for each are as follows: (1)
filtered nitrogen blow off, five minutes each of warm acetone,
warm propanol, warm methanol, and then filtered nitrogen
blowoff; (2) procedure (1) followed by a 1 minute etch in 20%
nitric acid in methanol, methanol rinse, DI rinse, five minute
etch in room temperature 5:1:1 (sulfuric acid: hydrogen
peroxide: DI water), DI rinse, methanol rinse, filtered


258
[100] Fang, D., Wang, X., Xu, Y. and Tan, L., J. Cryst. Growth
66 (1984) 317.
[101] Roksnoer, P. J. and Van Rijbroek-VanDenBoom, M. M. B.,
J. Cryst. Growth 66 (1984) 317.
[102] Ishikawa, J., Takahashi, N., Ito, T., Sube, M. and
Kurita, S., J. Electrochem. Soc. 137 (1990) 343.
[103] Wada, O., Majerfeld, A. and Robson, P. N., J.
Electrochem. Soc., 127 (1980) 2278.
[104] Hooper, A. and Tuck, B., Solid State Electron. 19
(1976) 513.
[105] Park, R. M., Stanley, C. R. and Clampitt, R., Inst.
Phys. Conf. Ser. 54 (1980) 235.
[106] Panish, M. B., Temkin, H. and Sumski., S., J. Vac. Sci.
Technol. B3 (1985) 657.
[107] Cheng, T. S., Airaksinen, V. M. and Stanley, C. R., J.
Appl. Phys. 64 (1988) 6662.
[108] Olsen, G. H., Interqrated Circuits: Chemical and
Physical Processing. ACS Symp. Ser. #290 (1985) 221.
[109] Jurgensen, H., Heyen, M. and Balk, P., Inst. Phys. Conf.
Ser. 56 (1981) 55.
[110] Chevrier, J., Huber, A. and Linh, N. T., J. Appl. Phys.
51 (1980) 815.
[111] Chevrier, J., Horache, E., Goldstein, L. and Linh, N.
T., J. Appl. Phys. 53 (1982) 3247.
[112] Yang, J. J., Ruth, R. P. and Manasevit, H. M., J. Appl.
Phys. 52 (1981) 6729.
[113] Bacher, F. R. and Leigh, W. B., J. Cryst. Growth 68
(1984) 102.
[114] Timmons, M. L., Chiang, P. K. and Hattangady, S. V., J.
Cryst. Growth 77 (1986) 37.
[115] Blaauw, C. B., Shepherd, F. R. and Eger, D., J. Appl.
Phys. 66 (1989) 605.
[116] White, A. M., Dean, P. J., Fairhurst, K. M., Bardsley,
W., Williams, E. W. and Day, B., Solid State Comm. 11
(1972) 1099.


209
Part of sample MBE 572 was processed for electrical
testing purposes following the procedure previously outlined
for MBE 464. A blanket zinc diffusion from the top to the
first p-AlGaAs layer was performed and a SIMS profile of the
atomic zinc and aluminum concentrations is shown in Figure 74.
As shown, the zinc diffusion ends in the fourth layer which
should be p-type AlGaAs, and the aluminum concentration in the
AlGaAs layers is essentially constant. Further processing and
device testing of MBE 572 is planned. Unfortunately, this
investigator had to leave BNR to pursue other interests such
as MOCVD growth of III-V semiconductors.
4.3 MOCVD Grown GalnAsP Devices
Optical interference filters using the GalnAsP/InP
material system have been previously grown by the MOCVD
technigue[87,144] and also by CBE[143], but all devices were
passive. As of the time of this writing, GalnAsP/InP
interference filters with electrical properties, which could
be used as active optoelectronic switch elements, have never
been reported. Similar to GaAs, the RI of GaxIn1.xAsyP1.y is
affected by a change in carrier concentration. A change in
carrier concentration from I016cm'3 to I018cm'3 for GalnAsP (Xg
= 1400nm) using sub-bandgap light at a wavelength of 1550nm,
would result in a change of RI of Anquat = -0.43%[155]. For
a 30 layer quarter-wave thick structure of Ga 375In 625As 81P 19
and InP with RI estimates from Broberg and Lindgren[147] of


222
A.2 Sample Preparation
In order to perform Hall effect measurements, a suitable
sample is needed. This semiconducting sample should be
expendable as the technique used is destructive. If a 2"
diameter wafer exists which needs characterization, a small .5
cm to .75 cm square piece will suffice as the working sample.
This sample can be obtained by first determining the crystal
growth plane which is usually indicated by a flat edge on the
circumference of the wafer. Next a line must be carefully
scribed in the wafer parallel or perpendicular to the flat
edge. The wafer can be broken or cleaved along the scribed
line. With one of the two pieces, this process should be
repeated until an approximately square working sample is
obtained. Since the sample can have maximum dimensions of
only 1 cm2, sufficient material should remain for additional
characterization. It has been reported[157-162] that crystal
orientation affects the Hall measurement results for all
anisotropic materials. Therefore, care should be taken to
establish a standard procedure to prevent misleading and
irreproduceable results.
After the test sample has been cut, make sure the
surfaces are clean. Some samples may require cleaning with a
degreasing solvent such as methanol and rinsing with distilled
deionized water, others may require chemical etching and
material specific techniques are described elsewhere[157].
Next, identify which surface has the deposited epitaxial layer


188
the lattice parameter is subject to a one-dimensional
modulation is characterized by satellites around each Bragg
peak of the average lattice and the total period thickness can
be calculated from the angular spacing of these satellite
peaks A6.
A superlattice cannot be completely characterized by
experimental rocking curve data alone since this yields only
two (x and T) of the three necessary parameters. Additional
information is needed, which can be obtained in one of several
ways. The complementary data can be obtained experimentally
by making additional measurements of the optical reflectance,
or PL, otherwise the individual layer thicknesses can be
measured directly by cross-sectional TEM. Alternatively,
since the intensities of the rocking curve satellite peaks are
proportional to the layer thicknesses, they can be found
interactively by fitting intensities of calculated rocking
curves to the experimental ones.
The rocking curves were recorded for the (004) reflection
on a modified Bede 6" double crystal diffractometer in the
parallel (+,-) setting using CuKa(A=1.5418) radiation. A
GaAs (001) single crystal was used as the first crystal: this
crystal when rocked with an identical one gave a rocking curve
FWHM of 10.3 arc-sec. The rocking curves were simulated
using a dynamical scattering theory based solution of the
Takagi-Taupin equations, using the method discussed in detail
by Hill[152].


93
Figure 24: X-ray diffraction patterns of GalnAs samples grown
on InP (left: lattice-matched, right: lattice-
mismatched) .


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy
GROWTH AND MODELING OF III-V COMPOUND
SEMICONDUCTOR OPTOELECTRONIC MATERIALS
WITH DEVICE APPLICATIONS
By
ARNOLD JOHN HOWARD
December 1990
Chairperson: Timothy James Anderson
Major Department: Chemical Engineering
Several topics have been undertaken during the course of
this degree which are associated with understanding and
improving semiconductor processing. The growth, modeling and
characterization of III-V compound semiconductor materials and
optoelectronic devices has been emphasized. Epitaxial layers
of GaxIn^xASyP^y with lattice-matched alloy compositions over
the range from x=0, y=0 (InP) to x=0.47, y=l (Ga 47In 53As)
have been grown by metal organic chemical vapor deposition
(MOCVD) on InP substrates. Both the MOCVD system, used to
grow these layers, and a low temperature Hall effect system,
used to characterize these layers, were designed and
installed. The results from several other analytical
techniques were used to determine the optimal growth
conditions for high quality epitaxial layers.
vi


224
necessary). Next close off the inlet and exit lines and open
the rough pump line. Plug in the rough pump and let it run
until the vacuum gauge reads -30 psi. Open the inlet line and
fill the chamber with forming gas. Then simultaneously close
the rough pump line and open the exit line. Let the gas flow
at the previous bubbling rate for about five minutes with the
hood closed. A flowmeter was attached to the system but it
added too much resistance to gas flow so it was removed.
Now that the chamber is nearly free of air, plug in the
digital thermometer (Omega, model 115KC) which should read
about 27C. At this point plug in the transformer (Signal,
model 10.5) turn it on, and set the dial to 55 (this system is
for GaAs with indium contacts). Close the hood and while
heating for about twenty minutes, make sure the gas flow rate
stays approximately constant. At this point the temperature
should be reading *230C (of course the sample is much hotter
than the thermocouple tip) and the transformer can be turned
off. Keep the forming gas flowing until the temperature reads
less than 50C consequently reducing the chance of forming an
oxide layer. Let the chamber cool down close to ambient
temperature before opening it and removing the sample.
The final step in the sample preparation process is the
testing of the nature of the contacts. The contacts have to
be of the Ohmic type to be used in the Hall effect experiment.
A simple check of contact integrity can be performed on a
conventional transistor curve tracer (one is available in the


70
Figure 11: The effect of growth temperature, Tg, on undoped
(n-type) InP background carrier concentration.


231
percentage of maximum current. For example, with a 2" pole
spacing at 50% of current and with the reversal switch
illuminated, a magnetic field of about 2600 gauss should be
measurable with the previously zeroed gaussmeter (Jobmaster
Magnets, model 6MIA). The magnetic field will be positive in
the sample when it is facing the pole llxelled "North and the
"DC-ON" switch is illuminated (the "Reversal" switch is not) .
A calibration curve for the particular pole spacing used
should be obtained on a regular basis. Operating at higher
magnetic fields induces higher Hall voltages and is therefore
desired. The sample should always be placed as close to the
center of the magnetic field and as perpendicular to its field
lines as possible.
With the sample in the proper place (shielded from
light), at the right temperature, with a stable low current
passing through it, with the voltmeter and magnet on, record
the Hall voltage value V5(+I,+B) ,f or example, with the switch
in position 5. Then by varying the current (+1,-1), magnetic
field (+B,-B) and position switch (5,6,S) record twelve
voltage values (S stands for the standard resistor reference
voltage). Positions 5 and 6 pass current diagonally through
the samples and consequently measure voltage across the other
respective diagonal. Now with the magnetic field off, put the
switch in positions 1, 2, 3, 4 and S and vary the current (+1,
-I) and record the ten resistivity voltage values Vi.


163
theories which set m = 1, or 2 do not take into account the
effect of substrate doping on zinc diffusion.
In addition to PIXE, EBIC, SIMS and diffusion theory,
other techniques have been employed to investigate the InP:Zn
activation phenomenon. Positron lifetime measurements have
been performed on InP:Zn crystals by Dlubek et al. [125]. They
report that the crystal doped with Zn to a concentration of
4.5xl018cm'3 shows strong positron trapping by vacancy defects,
but no vacancies are found in crystals doped with Zn to a
lower concentration of 2xl018cm3 or doped with Sn, S or Fe.
They also state that the existence of the complex VpZnInVp,
which is a deep unionized acceptor, is in perfect agreement
with their positron results. DLTS has also been performed on
InP:Zn and a hole trap with an energy of Evt = 0.52eV was
observed[126]. The trap's origin is attributed to either a
phosphorus vacancy or a phosphorus interstitial related
defect. Kamijoh et al.[127] performed 4.2 K PL on undoped InP
grown by MOCVD at different V/III ratios. They conclude that
Zn and C-acceptor impurity incorporation is controlled by the
V/III ratio. The effect of post growth cooling ambient on the
electrical activation of Zn and Cd dopants in MOCVD InP has
also been studied[128,129]. Cole et al.[128] reported nearly
complete activation in samples cooled in H2 only, intermediate
activation ( 50%) in samples cooled in PH3 and H2, and low
activation ( 10%) in samples which cooled to room temperature
with AsH3 and H2 flowing. They rule out the existence of the


55
Table 3
Metal Organic Vapor Pressure Equations
Source
Vapor Pressure Equation
TMIn
log10P(mm Hg) =10.52 *014
TEGa
log10P(mm Hg) = 8.224 2222
DEZn
log10P(mm Hg) = 8.28 2190


154
In an attempt to obtain more information about the nature
of the DEZn p-type doping process of MOCVD InP, a series of
experiments were performed where the V/III ratio and growth
temperature were independently varied. The V/III ratio for
all of the other DEZn doping experiments was held at 140.
Using a constant DEZn molar flow rate of 5.0xl0*7 moles/min,
V/III ratios of 10, 50, 140, and 200 were used for separate p-
doping experiments. In another series of growths, a higher
DEZn flow rate, 2.0xl0'5 moles/min, was used with the same
four V/III ratios. These eight samples were characterized by
SIMS and C-V profiling and the resulting hole and atomic zinc
concentrations are plotted in Figure 52 as a function of V/III
ratio. It appears that by increasing the V/III ratios, both
the hole and zinc concentrations decrease for both sets of
experiments. This trend may be misleading since the phosphine
source is only 15% concentrated which means that the total
flow rate to the reactor increased by 200 seem when the V/III
ratio was increased to 200, and decreased by 420 seem when the
V/III ratio was reduced to 10, both relative to the base
condition, V/III = 140 and phosphine mixture flow of 457 seem.
An increase or decrease in the total flow rate does reduce or
increase the relative DEZn partial pressure in the reactor for
a fixed DEZn molar flow rate. It's also interesting to note
that, as plotted, the ratio of hole to zinc concentration is
less for the high [Zn] runs than it is for the low [Zn] runs,
indicating less electrical activation.


CHAPTER V
CONCLUSIONS AND RECOMMENDATIONS
Epitaxial layers of InP, Ga^n^^s, and Ga^n^As P.,
have been grown lattice-matched to InP by MOCVD. However, the
electrical properties of undoped layers of these materials
grown in the quaternary MOCVD machine were not "state of the
art" or, in other words, as good as is possible considering
the quality of the sources presently available. The room
temperature Hall mobilities of undoped InP and Ga 47In 53As,
(on the order of 3000 and 6000 cm2/volt-sec, respectively) are
lower than the highest values reported in the literature
("4500 cm2/volt-sec for InP[55] and "10,000 cm2/volt-sec for
GaInAs[65]). The room temperature carrier concentrations for
undoped films are low though, n ~ 2*1014 cm'3, which suggests
that compensation by an incorporated acceptor impurity is a
cause for the inferior mobilities. The effect of growth
conditions, wafer preparation and sources used on material
quality have already been investigated as possible causes for
compensation. It is recommended that a phosphine purifier be
purchased and installed in the MOCVD system if high purity
InP, GalnAs and GalnAsP layers are required for future device
applications.
216


Counts
103
SIMS
Profile of
a MQW sample
CQ033).


125
opinion that the grown-in compensating donor zinc complexes
disassemble upon annealing to create a defect structure which
eventually contains only electrically active zinc acceptors.
Similar to the zinc doping paper, Chevrier et al.[lll]
have published a paper on Cd doping of chloride CVD InP. They
report a linear incorporation of Cd with hole concentrations
(deduced from Hall effect and C-V measurements) in the range
p = lxlO15 to 3xl018cm'3. For cadmium partial pressures below
1.5xl0'4 atm, a distribution coefficient K(Cd) = 0.2, is
reported. Above pcd = 1.5xl0*4 atm, they report that strong
compensation limits the hole concentration to a constant value
of 3xl018cm'3 (for partial pressures up to Pcd = 8.0xl02 atm) .
Chevrier et al.[lll] also performed annealing experiments on
the Cd doped samples and found the opposite trend as reported
for the Zn annealing experiments[110]; hole concentrations
decrease with longer annealing times at 300C. The cadmium
trend, they propose, may be due to increased neutral complex
formation, increased clustering of Cd, activation of a donor
compensating complex, or deep diffusion of Cd. They also
reported gradually decreasing growth rates at high (above
1.5xl04 atm) Cd partial pressures. Because of this fact, and
the relatively high pcd used, this writer has concluded that
the clustering explanation is the most logical explanation.
3.1.4c Metal Organic
MOCVD has been applied to the growth of p-type InP
epitaxial layers for optoelectronic device structures. To


9
like AsH3 or PH3. For InP growth, HC1 gas is first reacted
with liquid indium metal in the source zone. The gaseous
product InCl is then carried by H2 to mix and react with PH3
to deposit InP in the growth zone. Similar to the chloride
process, accurate temperature control is required for this
three-zone process which is also surface-kinetically limited
in the low-temperature growth regime. The hydride process is
currently widely used for light-emitting diode (LED)
applications using GaAs.,_xPx. It has also found application
in the growth of III-V GalnAsP and GalnAs for LED's, lasers
and detectors[5]. One major advantage the hydride system
provides over the chloride system is the ability to vary the
vapor phase V/III molar ratio by adjusting the inlet flow
rates of hydrides and HC1. One drawback of both the hydride
and chloride systems is that they are hot-wall systems;
interaction between the gas and heated Si02 reactor wall
occurs which results in unintentional silicon incorporation
into grown layers. Due to its successful use, especially in
LED fabrication, hydride CVD will continue to have a
significant role in the growth of III-V materials.
The third type of CVD or VPE process is metal organic
chemical vapor deposition (MOCVD) The MOCVD process involves
an irreversible pyrolysis reaction of vapor-phase mixtures
usually of group IIIA metal organic sources and group VA
hydride sources. For InP, as an example, trimethylindium
(TMIn) and PH3 diluted in H2 would flow into an open cold-wall


174
DEZn Partial Pressure (atm)
Figure 57: A comparison of the model (solid lines) and the
experimental data for InPsZn.


110
dislocations. The lower photo is of a sample (Q120) that
appears to be lattice-matched. The degree of mismatch can be
determined more accurately by XRD, but since there are two
unknowns, x and y, the exact composition of a lattice-
mismatched quaternary film cannot easily be determined with
this technique. The percent of lattice-mismatch (100 x Aa/a)
for samples Q036 and Q120 are -0.824% and 0.00%, respectively.
The degree of lattice-mismatch for both of these samples was
determined from their XRD spectra which are shown in Figure
34. All that one can tell from the higher angle peak position
of the quaternary material in spectrum Q036, other than the
lattice constant, is that it has too much GaAsyP.,_y to be
matched.
One way of getting a "rough idea of the composition of
GaxIn.,.xAs Pj /InP films is by using the electron microprobe
analysis (EPMA) technique. This analytical tool is available
in the Materials Science and Engineering Department on campus,
but it is not extremely accurate. It was useful, however, for
calibrating the growth conditions as the technique does not
require that the deposited film is lattice-matched or nearly
lattice-matched to the substrate used. It gives the detected
weight fraction or atomic fraction of elements which are being
emitted from the electron bombarded surface. The EPMA
analysis of sample Q120 yielded a quaternary composition of
Ga.353In.647As.80P.20* Another technique which determines more
precisely the composition of a lattice-matched quaternary film


193
observed experimentally by the increasing shift of the
position of the central peak away from l=1300nm as the
distance from the wafer center increases (1=1260, 1225, and
1140nm, respectively) with a corresponding drop in the peak
reflectivity (0.75, 0.70 and 0.65, respectively).
The theoretical reflectivity spectrum (calculated using
a program based on equations (30) (35)) and the experimental
one for A=1300nm at 45 incidence are plotted in Figure 64.
The dashed curve is the experimental data with a baseline
which lies approximately 0.05 below the theoretical R=0.30.
As shown, the experimental and theoretical spectra agree quite
well for the forty-layer AlGaAs/GaAs stack.
A small piece of the wafer was cleaved and its back side
was polished to a near defect-free (mirror-like) surface. For
this sample, after a reflectivity scan was performed, the
detector was moved (see Figure 63) to respond to light
transmitted through the sample. As shown in Figure 59, theory
predicts that the position of the peak of the reflectivity and
valley of the transmission scans should occur at the same
wavelength. These two scans are plotted in Figure 65 and as
shown, the agreement with theory is quite good.
4.4 Electrical Testing
Part of the MBE 464 wafer was processed for electrical
testing purposes. A blanket zinc diffusion from the top to
the first p-layer was performed. Part of this zinc diffused
sample was characterized by the SIMS technique. Atomic zinc


28
undoped MOCVD InP. So, it is evident that device quality
unintentionally doped InP can be grown by MOCVD.
Most semiconductor devices require a junction of some
type in the host material where two materials with either
different electrical or optical properties meet. An
electrical junction can be created by post growth processing
techniques such as ion-implantation or diffusion of a donor or
acceptor into the host crystal. Another way is to just create
the junction in-situ during the MOCVD growth by adding a small
quantity of a donor or acceptor source into the inlet gas
stream. High quality undoped InP is usually n-type with a
background carrier concentration of n = 1 x 10u- I015cm*3.
The carrier concentration n or (ND-NA) can be increased by
adding an InP donor species to the inlet gas stream of the
MOCVD reactor. InP can be doped n-type by using H2S[45],
H2Se[43] and SiH4[43] or Si2H6[57] as sources. For each
source, the free carrier concentration is essentially
proportional to the dopant source mole fraction in the reactor
inlet stream. Controllable n-type doping from I015cm'3 to
1020cm3 can be achieved without a significant decrease in
material quality by using a combination of these sources for
different parts of this wide incorporation range. Doping
levels and diffusion rates of these dopants are affected to
varying degrees by changes in growth conditions such as
temperature, V/III ratio, and indium mole fraction. For H2S,
the free carrier concentration in deposited InP layers


253
[18] Bass, S. J., J. Cryst. Growth M (1978) 29.
[19] Dupuis, R. D. and Dapkus, P. D., Appl. Phys. Lett. 31
(1977) 839.
[20] Jacko, M. G. and Price, S. J. W., Can. J. Chem. 41
(1964) 1198.
[21] Haigh, J. and O'Brien, S., J. Cryst. Growth 67 (1984)
75.
[22] Ban, V. S., J. Electrochem. Soc. 125 (1978) 317.
[23] Hayafuji, N., Mizuguchi, K., Ochi, S. and Murotani, T.,
J. Cryst. Growth 77 (1986) 281.
[24] Leys, M. R., Van Opdorp, C., Viegers, M. P. A. and
Talen-van der Mheen, H. J., J. Cryst. Growth 68 (1984)
431.
[25] Wang, C. A., Groves, S. H., Palmateer, S. C., Weyburne,
D. W. and Brown, R. A., J. Cryst. Growth 77 (1986) 136.
[26] Puetz, N., Hillier, G. and SpringThorpe, A. J., J.
Electron. Mater. 17 (1988) 136.
[27] Komeno, J., Tanaka, H., Itoh, H., Ohori, T., Takikawa,
M. and Kasai, K. 1987 Electronic Materials Conf., Santa
Barbara, CA, June 24-26, Paper E-8.
[28] Blaauw, C. and Miner, C. J., J. Cryst. Growth 84 (1987)
191.
[29] Eliot, B., Balma, F. and Johnson, F., Solid State Tech.
33 (1990) 89.
[30] Chen, C. H., Cao, D. S. and Stringfellow, G. B., J.
Electron. Mater. 12 (1988) 67.
[31] Johnson, E., Tsui, R., Convey, D., Mellen, N. and
Curless, J., J. Cryst. Growth 68. (1984) 497.
[32] Hess, K. L. and Riccio, R. J., J. Cryst. Growth, 77
(1986) 95.
[33] Rai-Choudhury, P., J. Electrochem. Soc. 116 (1969)
1745.
[34] Baliga, B. J. and Ghandi, S. K., J. Electrochem. Soc.
122 (1975) 683.


256
[68] Iwamoto, T., Mori, K., Mizuta, M. and Kukimoto, H., Jpn.
J. Appl. Phys. 22 (1983) L191.
[69] Duchemin, J. P., Razeghi, M., Hirtz, J. P. and Bonnet,
M., Inst. Phys. Conf. Ser. 63 (1981) 89.
[70] Sogou, S., Kameyana, A., Miyamoto, Y., Furuya, K. and
Suematsu, Y., Jpn. J. Appl. Phys. 23. (1984) 1182.
[71] Moss, R. H. and Spurdens, P. C., J. Cryst. Growth 68
(1984) 96.
[72] Fujii, T., Yamazaki, S. and Nakajima, K. 1988 Intern.
Conf. on Solid State Dev. and Mater., Tokyo, Japan,
Paper D-9-2, 395.
[73] Koukitu, A. and Seki, H., J. Cryst. Growth, 76 (1986)
233.
[74] Nahory, R. E., Pollack, M. A., Johnston, W. D. and
Barns, R. L., Appl. Phys. Lett., 33 (1978) 659.
[75] Hayes, J. R., Patel, D., Adams, A. R. and Greene, P. D.,
J. Electron. Mater. 11 (1982) 155.
[76] Saxena, R., Sardi, V., Oberstar, J., Hodge, L., Keever,
M., Trott, G., Chen, K. L. and Moon, R., J. Cryst.
Growth 77. (1986) 591.
[77] Mayer, R., Grutzmacher, D., Jurgensen, H. and Balk, P.,
J. Cryst. Growth 93. (1988) 285.
[78] Manasevit, H. M., Hess, K. L., Dapkus, P. D., Ruth, R.
P., Yang, J. J., Campbell, A. G., Johnson, R. E., Moudy,
L. A., Bube, R. H., Tabick, L. B., Fahrenbruch, A. L.
and Tsai, M. J., 1978 Conf. Rec., 13th IEEE Photovoltaic
Spec. Conf., 165.
[79] Razeghi, M. and Duchemin, J. P., J. Cryst. Growth 70
(1984) 145.
[80] Walpole, J. N., in Compound Semiconductors: Growth.
Processing, and Devices. P. H. Holloway and T. J.
Anderson (Editors), CRC Press, Boca Raton, FL, 1989.
[81] di Forte-Poisson, M. A., Brylinski, C., Colomer, G.,
Osselin, D., Hersee, S., Duchemin, J. P., Azan, F.,
Lechevallier, D. and Lacombe, J., Electron. Lett. .20
(1984) 1061.
Ogura, M., Inoue, K., Ban, Y, Uno, T., Morisaki, M. and
Hase, N., Jap. J. Appl Phys. 21 (1982) L548.
[82]


240
slowly. Flow H2 gas into the load-lock room by closing MV201,
MV202, NV202 and opening AV109, MV202, NV202 and when the
pressure of the load-lock room becomes 760 torr, close NV202,
MV202, AV109. Evacuate the load-lock two more times.
Insert the substrates to the reactor from the load-lock
room (fork operation). The pressure of the reactor should be
760 torr. Open MV200 and turn the fork manipulator on. Open
the fork gate valve. Open the shutter valve. Move the fork
forward and put the substrates on the graphite susceptor and
then remove the fork. Close the shutter valve. Close the
gate valve. Close MV200. Turn off the fork manipulator and
double check that MV200 is closed!
Prepare for flowing toxic gases (PH3 Only for InP on InP)
by first closing MV20, MV23, AV204 and then quickly evacuate
the system through AV209 with RP2 on. Increase the set point
of MFC20 to maximum. When MFC20 indicates almost zero seem,
close AV21, AV22 and AV23 (leave AV20 open). Adjust MFC20 to
about 10 seem (set point). Close the regulator (RV20) (turn
it fully counter-clockwise).
Heat the RF filament (power on, solenoid on, RF filament
on) set the pressure of the reactor to 80 torr. Increase the
main hydrogen flow rates to the reactor to 7 slm (total) and
to each vent line 0.5 slm also increase the H2 regulator to 90
psi. Set all other flow rates to overnight flows except the
PH3 compensation line (set to 100 seem) TMIn, and its
compensation line (both set to 143 seem) Adjust NV60 so PI60


15
temperatures (PH3 is more difficult to decompose than AsH3) ;
(3) gas phase V/III ratios have a strong effect on background
carrier concentrations and p- and n-type doping levels; (4) in
some cases, group III metal organics react with group V
hydrides to form adducts and polymers that may be involatile
liquids or solids; and (5) group II alkyls act as p-type
dopant sources, group VI hydrides act as n-type dopant
sources, but depending on growth conditions, group IV hydrides
can act as donor or acceptor sources in III-V's. The extent
to which these trends apply to all III-V materials does vary,
but they are definitely useful in designing an MOCVD system or
in optimizing material specific growth conditions.
2.2 MOCVD Systems
Most MOCVD systems use quartz reactors oriented either
vertically[11] where gas flow is usually down, or horizontally
[22] where gas flow is usually over a wedge shaped susceptor.
Other less common systems incorporate barrel reactors[23]
where growth can occur on multiple wafers with reactants
flowing from top to bottom, or "chimney" reactors[24] where
gas flow is up and wafers and susceptor are held vertically.
In a vertical reactor, uniform growth rates are more difficult
to achieve than in a horizontal reactor because with the
geometry of a vertical reactor, nonlaminar, turbulent gas flow
can more easily occur. In a horizontal reactor, a boundary
layer zone forms above the susceptor where the flow rate is
lower than that of the bulk gas above. The stagnant layer


173
and the total zinc equation to the data set #2 by adjusting
the values of the unknown constants. Once convergence has
been achieved, (a 95% confidence limit is met) the program
stops and it prints out the fitted constants. Three constants
Bu B2 and Bj were fitted using data set #1. Three more, Bs,
B6 and Bj were fitted using data set #2. The resulting values
of the best fit are: B1 = 375.3 mole fraction3/atm, B2 = 2.5
mole fraction*1atm*1/4, Bj = l.23xl0'18 mole fraction, Bs = 2708
mole fraction/atm, B6 = 1003 mol fraction'1, Bj = 11.6 atm1/4.
3.4.3 Discussion of Results
A comparison between the model's best fit line and the
data for both data sets (p and [Zn]) is shown in Figure 57.
It is encouraging to see that the hole p2n square root
dependence of the model does agree quite well with the C-V and
Hall data. Also, the total zinc data, which has less scatter,
is fit even better by the model line. It is difficult to
assess the real meaning of some of the evaluated constants
which are actually combinations of two unknowns. But, using
the total zinc data (data set #2) and the fitted constants,
one can get an idea of the relative concentrations of the
proposed point defects ZnIn-, Zn{, VpZn,nVp, and ZnInVp as a
function of zinc partial pressure. These relationships are
plotted in Figure 58. As shown, the point defect model which
was proposed, indicates that at low zinc partial pressures,
ZnIn' makes up the majority of the total incorporated atomic
zinc (pw[Zn]). As the zinc partial pressure is increased,


217
Another potential solution to the compensation problem is
associated with the substrate holding mechanism. Presently,
the substrate is placed on top of a quartz wafer tray during
the MOCVD growth. It is conceivable that this tray is the
source of compensating acceptors. The tray is cleaned and
dried using the same procedure as the reactor, but, unlike the
rest of the reactor, it is not baked for four hours at 950C
and kept in an air-free environment before a deposition. A
clean tray is used for each MOCVD growth, but, the sample is
heated by conduction through the quartz tray and part of the
exposed tray is directly upstream of the sample. It is
recommended that a silicon carbide coated graphite tray be
purchased and used to hold the sample during the MOCVD growths
of structures which require high purity undoped epitaxial
layers.
A model of the p-type doping process of MOCVD InP using
DEZn has been developed, yet never before reported. The
results of this study are that the dominant electrically
inactive point defects in zinc doped InP grown using high DEZn
reactor partial pressures are interstitial zinc, Zn,., and zinc
complexed with a phosphorous divacancy, VpZnInVp. Due to the
fact that very little information is available on the InP:Zn
material system, several assumptions were made during the
course of the development of this model. It is recommended
that future researchers perform extensive characterization on
undoped InP and InP:Zn, such as high temperature Hall effect


3.1.3 Liquid Phase Epitaxy 121
3.1.4 Molecular Beam Epitaxy 122
3.1.5 Chemical Vapor Deposition 123
3.2 MOCVD Growth and Characterization of Mg-Doped
InP Using bis-(Methylcyclopentadienyl)
Magnesium as a Dopant Source 130
3.2.1 Introduction 130
3.2.2 MOCVD Growth 130
3.2.3 Results and Discussion 131
3.2.4 Conclusions 145
3.3 Experimental DMCd and DEZn for p-Type Doping
of InP by MOCVD 145
3.3.1 Introduction 145
3.3.2 DMCd Results 146
3.3.3 DEZn Results 148
3.4 Modeling of p-Type Doping of InP using DEZn.. 156
3.4.1 Introduction 156
3.4.2 Point Defect Structure 164
3.4.3 Discussion of Results 173
IV EPITAXIALLY GROWN INTERFERENCE FILTERS 177
4.1 Theory of Interference Filters 177
4.2 MBE Grown AlGaAs/GaAs Devices 182
4.2.1 Introduction 182
4.2.2 Electrical Theory 184
4.2.3 Experimental 185
4.2.4 Electrical Testing 193
4.2.5 Conclusions 204
4.2.6 Addendum 205
4.3MOCVD Grown GalnAsP Devices 209
V CONCLUSIONS AND RECOMMENDATIONS 216
APPENDICES
A THEORY AND OPERATION OF THE LOW TEMPERATURE HALL
EFFECT SYSTEM 220
B OPERATION OF THE QUATERNARY MOCVD SYSTEM 234
C DOPANT MODEL COMPUTER PROGRAM 248
REFERENCES 252
BIOGRAPHICAL SKETCH 262
v


Growth Rate (pin/hr)
65
TMIn Mole Fraction
Figure 8: The effect of TMIn mole fraction on InP
rate.
growth


REIATNE INTENSITY
98
Figure 27: PL spectrum of a lattice mismatched GalnAs film on
InP measured at 4.2 K.


115
This, in addition to the observation that the spectrum of Q120
has no other peaks, indicates that the film's compositional
uniformity throughout the film and purity of the film are both
quite good. The FWHM of the PL peak for sample Q239 (1RT=
1.371/xm) is 15.5 meV. This spectrum is shown in Figure 36 and
was taken at 13 K which partially accounts for the larger peak
width due to increased lattice vibration at the higher
measurement temperature. It may also be that the control of
the composition during the MOCVD growth of sample Q239 was
inferior to that of sample Q120.
Another way of determining the purity of quaternary films
is by measuring the background carrier concentration of
undoped layers with the C-V profiler. In addition, the Hall
effect can be used to measure the average background carrier
concentration and the mobility of the layer. A C-V profile of
sample Q239 is shown in Figure 37 and from this profile an
average background carrier concentration of 2xl016cm'3 was
measured. The room temperature Hall carrier concentration and
mobility of this same sample were measured to be 1.8xl016cm'3
and 4630cm2/volt-sec, respectively. The lowest Hall carrier
concentration of an undoped quaternary film was 1.10xl015cm'3
for sample Q166. The highest room temperature Hall mobility
measured was 4810cm2/volt-sec for sample Q037. Of course
these values are composition dependent; higher mobilities are
expected from compositions closer to y=l (GalnAs).


151
the [Zn] is linearly proportional to DEZn flow rate, then, a
saturation level of [Zn] = 2-4xl018cm3 exists. Also, the
measured InP growth rate is unaffected by the DEZn flow rate
for all of the experiments performed and, conversely, the
doping level is unaffected by a change in the growth rate.
Electrical characterization of grown layers was also
performed by using an electrochemical C-V profiler and by
performing Hall effect measurements. A typical C-V profile of
an InP:Zn growth with DEZn from the bubbler (sample B318) is
shown in Figure 50. This 2.4/un thick InPtZn film was also
characterized by room temperature Hall effect measurements and
the C-V carrier concentration, p = lxl018cm*3, was confirmed.
Some InP:Zn samples were grown with thin lattice-matched
GaInAs:Zn contact layers to improve the likelihood of making
ohmic contacts. After the contacts were alloyed, the thin
GalnAsrZn layer was removed by etching and Hall measurements
were performed. The Hall and C-V measured 300 K carrier
concentrations are plotted as a function of atomic zinc
concentrations (determined by SIMS) in Figure 51. Less Hall
data is shown as the results from several samples were deemed
unreliable due to rectifying contacts. The hole concentration
(electrically active Zn) for the highly doped samples is less
than the atomic (total) Zn concentration. In other words, at
high DEZn molar flow rates a fraction of the zinc atoms are
incorporated into InP as either electrically neutral or
possibly compensating donor defects.


186
Layers were grown at the substrate temperature of 670C
using the oxide removal temperature as a reference[151]. The
GaAs growth rate was ~ 1/Ltm/hr and the A1 flux was adjusted to
achieve the 30% AlAs composition in AlGaAs. As2 was the
arsenic source and the V/III flux ratio was maintained between
3 and 5.
4.2.3b Transmission Electron Microscopy
Shown in Figure 61 is a transmission electron microscopy
(TEM) photograph of a cleaved (110) and etched edge of the
stack. The alternating dark and light delineated layers
represent GaAs and AlGaAs, respectively. As shown, the
interfaces are abrupt and from this photo it is evident that
the structure has good vertical period uniformity. At 210,000
times magnification, the average GaAs/AsxGa,,_xAs layer pair
thickness was measured to be 192nm which is very close to the
design value of 200nm.
4.2.3c Double Crystal Diffractometrv
Three parameters completely determine an AlxGa.,_xAs GaAs
superlattice: t1# the thickness of the GaAs layer, t2, the
thickness of the AlGaAs layer and x, its composition. Double
crystal rocking curves give information on xv, the layer
composition of a virtual crystal having this as an average
composition and T the total period thickness. The averaged
composition, xv, is easily calculated from the angular spacing
between the substrate peak and the main superlattice peak
reflection (or zero order peak) XRD from a material in which


141
10
19
10
18
CO
i
!*
o
CO
c
0)
o
c
o
o
10
16
10
15
10
14
epilayer/substrate
interface
C-V Profiling
SIMS


CD

2 4 6
depth (/i,m)
8
Figure 44; SIMS and hole concentration profiles of InP:Mg
grown on InP:Fe.


167
constants can also be obtained from experimental data or from
estimates based on values of similar materials. Hurle has
previously applied the point defect analysis in a series of
papers on undoped[132], tellurium doped[133], tin doped[134],
and germanium doped[135] GaAs. To this writer's knowledge,
this technique, applied to InP, has never been reported
before. This method has been used to understand the relative
concentrations of the point defects of InP:Zn which are shown
in Figure 56. Anti-site defects (Inp and PIn) may also exist,
but it is assumed that their concentrations are small relative
to the other defects already being considered.
As stated previously, formation and ionization equations
and equilibrium constants can be written for the point defects
in InP:Zn. If the concentrations of the defects considered
are small, then it is safe to assume that their activities are
equal to their concentrations (activity coefficients are
unity), except for electrons and holes. Also, if defect
concentrations are small, then the activities of Pp, InIn, V¡
would be unity. Using these conditions, equilibrium constant
expressions for each reaction can also be derived. The
reactions to form the host crystal and zinc related point
defects and their corresponding equilibrium relations are
given in Table 8 (Equations 11-21). Also given in Table 8 is
the electroneutrality relation (22) which is written under the
premise that all stable solids are electrically neutral.


54
where:
V
III
MF,
ph3
MF
THIn
MFPh3
(4)
(5)
MF
THIn
FH2,TMIn
VP-
THIn
(T)
FH2 Pb,TMIn
(6)
and Fph3 is the total pure phosphine flow rate (cm3/min) t fh2
is the total hydrogen flow rate (cm3/min), FH2 THIn is the
hydrogen flow rate through the TMIn bubbler (cm3/min) ,
VPTMIn(T) (nun Hg) is the bubbler temperature dependent TMIn
vapor pressure (torr), and PbTMIn is total pressure at the
TMIn bubbler (torr). All volumetric flow rates are measured
at standard conditions (300 K and 760 torr). The temperature
dependent vapor pressure equations for the metal organics
installed in the MOCVD system which were supplied by their
vendors (see Table 2) are shown in Table 3. It is therefore
possible with the use of the above equations to determine
conditions for a MOCVD experiment. Of course some initial
values must be known and others can be based on literature
values.


Fork
Mechanism
H2 and Hydride
Gas Inlet
Sample Quartz
Position Wafer Tray
Figure 4; MOCVD reactor and loading mechanism.


CONCENTRATION (atoma/cc)
197
DEPTH (microns)
Figure 66: SIMS profiles of aluminum and zinc in MBE 464.


18
substrate holders in both circular and even planetary motion
configurations[27]. These techniques can greatly improve
uniformity, but also, unfortunately, greatly add to machine
complexity and expense.
Aside frcm the reactor, the other major parts of an MOCVD
system are the gas delivery, heating exhaust/scrubbing and
safety systems. Most gas handling systems are constructed
from high purity stainless steel tubing, valves (air-operated
and manual), regulators, electronic mass flow controllers and
filters or purifiers. The system typically delivers metal
organics (from bubblers), hydrides (from high pressure gas
cylinders) and most often hydrogen (from a palladium-alloy
purifier) to a fast switching manifold which directs gases to
the reactor or to a vent line. Gas manifolds should be
situated as close to the reactor inlet as possible to minimize
tube length and improve interface abruptness capabilities.
Most systems use manifolds with a linear valve arrangement,
but only in a radial manifold arrangement is the length from
each valve to the reactor potentially the same for all of the
gases[28]. In the "vent/run" type system discussed above, the
vent line and reactor line are "pressure balanced" so that
transient times associated with gases adjusting to and flowing
from a high pressure to a low pressure line, or vice versa,
can be eliminated.
For heating systems, most MOCVD reactors are heated by
inductively coupling RF power to a graphite susceptor. This


261
[155] Botteldooren, D. and Baets, R., Appl. Phys. Lett. 54
(1989) 1989.
[156] Chien, C. L. and Westgate, R., The Hall Effect and
its Applications. Plenum, New York, NY, 1979.
[157] ASTM Standard Method for "Method and Hall Coefficient in
Extrinsic Semiconductors," F76-73 (1978).
[158] Putley, E. H., Hall Effect and Related Phenomena.
Butterworth & Company, Ltd., London, England, 1960.
[159] Beer, A. C., Galvanomaanetic Effects in Semiconductors.
Academic Press, New York, NY, 1963.
[160] Stillman, G. E., Wolfe, C. M. and Dimmock, J. O., Jour.
Phys. Chem. Solids. 31 (1970) 1199.
[161] Wieder, H. H., Laboratory Notes on Electrical and
Galvanometric Measurements. Elsevier, New York, NY,
1979.
[162] van der Pauw, L. J., Philips Res. Rep. 13 (1958) 1.
[163] Benzaquen, M., J. Amer. Phys. Rev. B., 34 (1986) 8947.
[164]Hayakawa, Y., MOCVD Operating Procedures A Translation.
Nippon Sanso Inc., Hammamatsu, Japan, 1988.


62
Table 4
Optimum InP(lOO) Wafer Preparation Procedure
1. Cleave wafer and blow off with filtered N2.
2. Degrease teflon beakers and tweezers with warm methanol.
3. Place substrates into beaker with warm acetone for 5
minutes.
4. Place in warm propanol for 5 minutes.
5. Place in warm methanol for 5 minutes.
6. Rinse in running DI water for 1 minute.
7. Etch in a 5:1:1 solution of H2S0A:H202:DI at RT for 5 min.
8. Rinse in running DI for 1 minute.
9. Place in room temperature methanol for 1 minute.
10. Blow off with filtered nitrogen.
11. Load into reactor on wafer tray from oven.
12. Anneal wafer at growth temperature for ten minutes with H2
and PH3 flowing.


92
explained by Cullity[92]. It relies on the periodic structure
of a crystal to scatter incident X-rays in such a way that
some of the scattered beams will be in phase and reinforce
each other to form diffracted beams. Scattered rays will be
in phase if Bragg's law is satisfied:
nA. = 2d sin0 (8)
where:
(h2 +k2 +12 )1/2
and X = 1.54051 (for copper Kal) (hkl) are miller indices
(usually (004)), n is an integer (equal to 1 for first order
reflection) and 0 is the angle of incidence of X-rays. Figure
24 shows the XRD spectra from two different GaxIn1.xAs/InP
growths. The plot on the left is the spectrum for growth Q015
which is lattice matched. The reason there are two peaks is
because two different X-ray wavelengths close to each other
(Kal and Ka2 from copper) were incident upon the sample. The
spectrum on the right is from growth Q007 and it has a third
broad peak at a lower 29 angle relative to the InP substrate
peaks which when inserted into Bragg's law yields a lattice
constant of a = 5.897. This is the lattice constant of
Ga 40In_60As (see Table 1 for the lattice constants of InAs and
GaAs) which has a lattice-mismatch of Aa/a = 4.6xl0'3 or
0.46% on InP. If a GalnAs/InP sample had an X-ray spectrum


Intensity (A.
116
Wavelength (pm)
Figure 36: PL spectrum of sample Q239 measured at 13 K.


TABLE OF CONTENTS
page
ACKNOWLEDGEMENTS
ABSTRACT vi
CHAPTERS
I INTRODUCTION 1
1.1 III-V Semiconductors 1
1.2 Epitaxial Growth Techniques 6
II METAL ORGANIC CHEMICAL VAPOR DEPOSITION 13
2.1 A Brief History of MOCVD 13
2.2 MOCVD Systems 15
2.3 A Review of the Literature on InP Based MOCVD 22
2.3.1 InP Homoepitaxy 22
2.3.2 GalnAs/InP 30
2.3.3 GalnAsP/InP 33
2.3.4 InP Based Devices 37
2.4 A Description of the MOCVD System 40
2.4.1 Introduction 40
2.4.2 Gas Delivery System 43
2.4.3 Reactor and Heating System 47
2.4.4 Exhaust/Scrubbing System 50
2.4.5 Safety 51
2.5 Determination of Optimum Growth Conditions
Based on Thin Film Characterization 52
2.5.1 Experimental Approach 52
2.5.2 InP 56
2.5.3 Growth of GalziAs Lattice-Matched to
InP 90
2.5.4 Growth of GalnAsP Lattice-Matched to
InP 108
IIIP-TYPE DOPING OF MOCVD INP: EXPERIMENTS AND
MODELING 118
3.1 A Review of the Literature on p-Type Doping
of InP 118
3.1.1 Introduction 118
3.1.2 Bulk Crystal Growth 119
iv


104
Figure 30:
TEM micrograph
of MQW sample
13 3 A InGaAs
19 + 3 A InGaAs
23 3 A InGaAs
29 3 A InGaAs
33 3 A InGaAs
45 3 A InGaAs
InP SUBSTRATE
Q13 6 .


242
the temperature of the substrate is less than 200C (50R) ,
flow PH3 gas to the vent. When the flow rate of the PH3
becomes 0 seem, purge the PH3 line five times by opening MV20
and MV23 and then close MV23, and wait until MFC20 becomes 0
seem. Repeat this four more times and reduce MFC20 to 30
seem, also make a note of the time.
Pressurize the reactor to 760 torr. Close AV207 and
AV209, and turn off RP2. When the reactor pressure becomes
780 torr, open AV204. Remove the sample from the reactor (30
minutes after the time noted previously) by fork operation.
Remove the samples from the load-lock room by first evacuating
it. Close MV102, AV109, and open MV201, MV202, NV202. Flow
Nz gas into the load-lock room. Close MV201, MV202, NV202 and
open MV102, MV202, NV202. When the pressure of the load-lock
room becomes 760 torr, close NV202, MV202, MV102. Repeat the
evacuation and fill steps three more times then remove the
substrates from the load-lock room. Repeat evacuation and
fill steps one more time, then turn off RP1. Reduce the H2
regulator to 40 psi, bleed line to 100, make all flows 0.2 slm
or 30 seem.
Wash the substrate supporter using aqua regia (3 HC1:1
HN03) for 10 minutes then rinse with DI for one minute then
dip in the HF mix (5 DI:1 HF) for 5 minutes. Follow with a DI
rinse for 15 minutes. Turn off the water flow (to the MOCVD
system), silence the water alarm, open the water pressure
relief into the plastic tray. Turn on the variacs for the


>REFLECTIVITY< L .1 /div]
208
Figure 73: Reflectivity spectrum for sample MBE572.


37
quaternary composition has not, however, been reported.
Saxena et al.[76] present data on 1.3jum p-GalnAsP doped by
DEZn over the range I018-I019cm*3,and n-GalnAsP doped by DETe
and H2S. Using tellurium, n-type doping from 3xl017 to
5xl019cm*3 is reported and, with sulphur a lower range 5xl016
to 3xl018cm*3 is reported. Meyer et al.[77] report n-doping
of 1.3 and 1.55im GalnAsP using H2S over the range 1016 to
I019cm*3. It is evident that p- and n-doping of GalnAsP is
possible over a wide range of doping levels which is
significant for device applications.
2.3.4 InP Based Devices
The wide range of compounds that can be grown with large
area uniformity by MOCVD make it suitable for fabrication of
long wavelength opto-electronic device structures. Several
different types of electronic, optical and opto-electronic
devices have been grown by MOCVD in the GaxIn1.xAsyP1.y on InP
material systems. Manasevit et al.[35] in 1978 showed that
solar cells in which an InP active region grown by MOCVD can
perform as well as ones fabricated by other techniques. Other
devices such as lasers, field effect transistors, photo
detectors and waveguides have been grown by MOCVD and will be
discussed in the following paragraphs.
Some of the first devices grown using MOCVD GalnAsP were
broad area and stripe double heterostructure (DH) lasers which
lased at wavelengths between 1.15 and 1.54/xm[69]. Distributed
feedback (DFB) lasers have also been successfully grown by


257
[83] Kim, J., Li, S. S., Figueroa, L., Carruthers, T. F. and
Wagner, R. S., J. Appl. Phys. 64 (1988) 6536.
[84] Gallant, M., Puetz, N., Zemel, A. and Shepherd, F. R.,
Appl. Phys. Lett. 52 (1988) 733.
[85] Grutzmacher, D., Meyer, R., Balk, P., Berg, C.,
Schapers, T., Luth, H., Zachau, M. and Koch, F., J.
Appl. Phys. 66 (1989) 697
[86] Delacourt, D., Papuchon, M., Poisson, M. A. and Razeghi,
M., Electron. Lett. 23 (1987) 451.
[87] Deppe, D. G., Gerrard, N. D., Pinzone, C. J., Dupuis, R.
D. and Schubert, E. F., Appl. Phys. Lett. 56 (1990) 315.
[88] Nelson, A. W., Cole, S., Wong, S., Harlow, M. J.,
Devlin, W. J., Wake, D., Rodgers, P. M. and Robertson,
M. J., J. Cryst. Growth 22 (1986) 579.
[89] Blood, P., Semicond. Sci. Tech. 1 (1986) 7.
[90] Dean, P. J., Prog. Crystal Growth Charact. 5 (1982) 89.
[91] Holland, R. and Blackmore, G. W., J. Crystal Growth. 68
(1984) 271.
[92] Cullity, B. D., Elements of X-Rav Diffraction. 2nd. Ed.,
Addison-Wellsley, Reading, MA, 1980.
[93] Kronig, R. de L. and Penney, W. G., Proc. R. Soc.
London, A130 (1930) 499.
[94] Skromme, B. J., Stillman, G. E., Oberstar, J. D. and
Chan, S. S., J. Electron. Mater. 13 (1984) 463.
[95] Williams, E. W., Elder, W., Astles, M. G., Webb, M.,
Mullin, J. B., Straughan, B. and Tufton, P. J., J.
Electrochem. Soc. 120 (1973) 1741.
[96] Kubota, E. and Katsui, A., Jap. J. Appl. Phys. 24.
(1985) 1344.
[97] Yamamoto, A., Yamaguchi, M. and Uemura, C., Appl. Phys.
Lett. 47 (1985) 975.
[98] Nakano, Y., Takahei, K., Noguchi, Y., Suzuki, Y., Nagai,
H. and Nawata, K., Electron. Lett. 12 (1981) 782.
[99] Bachman, K. J., Buehler, E., Miller, B. I., McFee, J. H.
and Thiel, F. A., J. Cryst. Growth ,39 (1977) 137.