A characterization of the chemical vapor deposition of gallium arsenide and indium phosphide in the hydride and chloride...


Material Information

A characterization of the chemical vapor deposition of gallium arsenide and indium phosphide in the hydride and chloride systems
Physical Description:
vi, 205 leaves : ill. ; 28 cm.
Meyer, Douglas John, 1953-
Publication Date:


Subjects / Keywords:
Vapor-plating   ( lcsh )
Indium phosphide   ( lcsh )
Gallium arsenide semiconductors   ( lcsh )
Compound semiconductors   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis (Ph. D.)--University of Florida, 1984.
Includes bibliographical references (leaves 195-204).
Statement of Responsibility:
by Douglas John Meyer.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000491164
notis - ACQ9666
oclc - 11959318
System ID:

Full Text








The author would like to thank Dr. Paul Holloway and Dr. Gar

Hoflund for their advice during the design of the mass spectrometer

vacuum system. The assistance provided by Dave Cox during construc-

tion of the vacuum system was greatly appreciated.

It is a pleasure to thank Ron Baxley and Tracy Lambert for their

instrumental help in constructing the laboratory.

The author wishes to express a special appreciation to his

committee chairman, Dr. Tim Anderson. His guidance and encourage-

ment throughout this project was greatly appreciated.

It is also a pleasure to express appreciation to Melissa Maher

for typing this rather lengthy manuscript.

Finally, the author wishes to express his sincere gratitude to

G. Yard, whose undying efforts to watch over the laboratory during

the author's presence and absence were greatly appreciated.


ACKNOWLEDGEMENTS ..................... ................. .11

ABSTRACT ........................................................


Importance of III-V Semiconducting Materials ..... 1
Epitaxy of III-V Semiconducting Materials ...............5

TWO REVIEW OF THE LITERATURE ..............................14
Impurities in GaAs Epitaxial Films Grown by the
Chloride Process ........ .. ... .................... 14
Impurities in InP Epitaxial Films Grown by the
Chloride Process ..................................... 17
Impurities in GaAs Epitaxial Films Grown by the
Hydride Process .......... ......... ............. 20
Impurities in InP Epitaxial Films Grown by the
Hydride Process ...................................... 22
The Thermal Decomposition of NH3 .. 24
The Thermal Decomposition of PH ......................27
The Thermal Decomposition of As .................28

AND CHLORIDE PROCESSES ..............................30
Introduction .. .......................................30
Method of Calculation for Complex Chemical
Equilibrium ........................................ 31
Thermodynamic Models of CVD ............................35
Models for the CVD Source and Pre-Source
Zones ........................................ 35
Models for the CVD Mixing and Deposition
Zones ....................................... 38
Solid State Defect Chemistry ..........................44

FOUR THERMOCHEMICAL PROPERTIES .............................49
Introduction .......... .......... ..................... 49
Pseudo-Steady State Constraint for the Liquid
Source Boat ........................................50
The Ga-As-C1-H System ..................................54
The In-P-C1-H System ................................. 57
The Si-C1-H System .................................... 61


Introduction ........................................ 73
Experimental Apparatus and Method .....................76

SIX RESULTS AND DISCUSSION ................................86
Chemical Equilibrium Investigation ....................86
Introduction ................................... 86
The GaAs Chloride System ..........................87
The GaAs Hydride System .......................... 112
The InP Chloride System .........................121
The InP Hydride System ..........................130
Thermal Decomposition of NH3, PHI, and AsH3 ...........133
Results of the Thermal Decomposition
Measurements ................................. 133
Analysis of Uncertainties in the Rate Data .......143
Determination of Activation Energies .............150
Implications of the Measurements on Other
Reactions in the Decomposition Chain .............154
Comparison Between Equilibrium and Kinetic Models
for VH3 Decomposition .............................. 160

Conclusions .........................................168
Recommendations ...................................... 170


CHEMICAL EQUILIBRIA ................................173


REFERENCES ...... ............................................... 195

BIOGRAPHICAL SKETCH ............................................205

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




December 1984

Chairman: Timothy James Anderson
Major Department: Chemical Engineering

The fundamental chemistry surrounding the chemical vapor deposi-

tion of GaAs and InP in the hydride and chloride processes was investi-

gated. Chemical equilibrium calculations showed that IIIC1, V4 and V2

were the dominant group III and V species in the vapor phase. These

calculations also demonstrated that vapor phase silicon species, formed

by the interaction of H2 and HC1 with the reactor wall, may be present

at compositions up to 1 ppm under typical operation conditions. It was

shown that the formation of vapor phase silicon species can be sup-

pressed by the addition of small amounts of H20 or by replacing the

H2 carrier gas with an inert. The unintentional incorporation of

silicon into III-V epitaxial layers may be decreased by reducing the

amount of silicon species in the vapor phase or by shifting these

silicon species from hydrogen rich to chlorine rich species through

the addition of HC1 or VC13.

The use of solid and liquid group III sources in the chloride

process was compared. In the GaAs system, the liquid source yielded

a much greater degree of supersaturation than did the solid source.

This difference was much less pronounced for the InP system.

The equilibrium chemistry of the hydride process was found to

behave similarly to that of the chloride process. The degree of

supersaturation present in the hydride process was found to be lower

than that in the chloride process.

The thermal decompositions of NH3, PH3 and AsH3 were studied in

a constant volume reactor using a mass spectrometer. It was found

that PH3 and AsH3 could be adequately represented by first order de-

composition reactions for temperatures greater than 850 K and 780 K,

respectively. A reverse reaction between VH and H2 was found to be

present in all of the VH3 decompositions. The activation energies for

the effective first order rate constants were 60.2, 36.5 and 29.2 kcal/

mole for NH3, PH3 and AsH3, respectively.


Importance of III-V Semiconducting Materials

The development of the solid state electronics industry has princi-

pally centered around the semiconductor material Si due, primarily, to

its good electrical properties, available high purity and the relatively

simple chemistry surrounding the fabrication of silicon semiconductor

devices. The rapid growth experienced by the semiconductor industry has

imposed further demands on device characteristics, such as extremely

high speed and optical emission properties, which cannot be met by silicon

as a result of its relatively low electron mobility and fixed indirect

bandgap. The desire to satisfy these new demands has spurred the

development of compound semiconductors composed of the elements Al, Ga

and In from column IIIa and P, As and Sb from column Vb of the periodic

table. Semiconductors formed from combinations of these elements

provide a large selection of electrical and physical properties such

as electron mobility, bandgap energy, possible direct bandgap and crystal

lattice parameter. Table 1-1 compares these parameters for all of the

III-V binary semiconductors and silicon. Electron mobilities ranging

from 80 to 100000 cm2/V-s and bandgap energies ranging from 0.18 to

2.45 eV are available using the binary III-V semiconductors. This

tremendous range of electrical properties contrasts sharply with the

single available set of properties provided by silicon.

Table 1-1

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

Bandgap Electron Mobility
Type Energy (eV) (cm /V-s)

indirect 1.12 1350

direct 0.18 100000

direct 0.36 22600

direct 0.70 5000

direct 1.28 4000

direct 1.43 8500

indirect 1.60 200

indirect 2.16 180

indirect 2.26 300

indirect 2.45 80

Source: Streetman [1].

Lattice Constant






















Semiconductors consisting of elements from the group III and V

columns of the periodic table need not be limited to simple binary

compounds. Indeed, the use of ternary and quaternary solid solutions

provides a means of varying the physical and electrical properties of

III-V materials continuously between the limits in Table 1-1. The

vastly increased electron mobility, which is characteristic of many

of the III-V semiconductors over that of silicon, makes these materials

ideally suited to advancing the current capabilities of high speed de-

vices (e.g. oscillators and mixers in the 100 GHZ range, central pro-

cessors and computer memories).

Currently, the most important application for semiconductors made

from III-V materials is in the production of optoelectronic devices.

The energy band structure for many III-V materials is of a direct

nature; that is, a third particle (e.g. phonon) is unnecessary for the

generation and recombination of free carriers. This allows for the

efficient conversion of electromagnetic radiation into electrical power

(photovoltaic cells) or, in the opposite situation, emission of radia-

tion from the semiconductor device (light emitting diodes, lasers).

The available bandgap energies for binary III-V devices having a

direct band structure result in device emission characteristics which

range from the infrared into the visible region of the electromagnetic

radiation spectrum.

Figure 1-1 plots the lattice parameters of many possible III-V

compounds against the bandgap energies of these compounds at 300K.

The solid dots represent binary compounds and the lines connecting each

dot represent ternary solid solutions of intermediate composition be-

tween the binaries. Solid lines signify direct bandgap materials


In b


6.2 -
o GaSb A1Sb
L InAs 0

S6.0 -


"- 5.8

5.6-- \

GaP --oA1P

0 0.5 1.0 1.5 2.0 2.5

Bandgap (eV)

Figure 1-1
Lattice Parameter and Bandgap Energy of Various III-V Semiconductors


while broken lines indicate indirect materials. Essentially, the entire

area enclosed in Figure 1-1 is accessible to the designer when employing

ternary and quaternary III-V solid solutions. This flexibility is ex-

tremely useful for the optimal design of new solid-state electronic


Currently, the quaternary system InxGalxAs yP1- is receiving much

attention. The available ranges for the lattice parameter and bandgap

energy in this system are given by the cross hatched surface shown in

Figure 1-1. Thus, the lattice parameter and bandgap energy may be

specified independently with the composition of the quaternary solution

chosen to meet these specifications. The availability of a degree of

freedom in the lattice parameter is extremely important since currently

only GaAs, GaSb, GaP, InP and InSb are available in bulk crystal form

for use as substrate materials. A disparity of greater than 0.1% be-

tween the lattice parameter of the substrate and epitaxial layer induces

the formation of interface defects in the crystal structure which can

degrade the device performance. One important application of the qua-

ternary InxGalxAsy P1y is in the development of heterojunction lasers

for use as transmitters in optical fiber communication systems [2].

Currently, available optical fibers exhibit minima in attenuation and

dispersion characteristics for radiation of approximately 1 eV [2].

Choosing the values of x=0.8 and y=0.35 yields an emission energy of

-1.11 eV [3] with a lattice parameter which closely matches that of

the InP substrate.

Epitaxy of III-V Semiconducting Materials

There currently exists three primary methods for growing epitaxial

III-V films: liquid phase epitaxy (LPE), molecular beam epitaxy (MBE)

and chemical vapor deposition (CVD). LPE is the growth of thin single

crystal layers from a liquid solution. The driving force for film

deposition is most commonly provided by decreasing the temperature of

the substrate relative to that of the liquid. The advantages of LPE


1) The method is capable of growing multicomponent layers with
a high reactivity disparity among the elements.

2) The equipment is relatively simple and inexpensive.

3) A large selection of dopants is available.

4) The process is near equilibrium at the surface thus allowing

5) The growth occurs below the film melting temperature.

6) The growth rate can be high.

7) The impurity distribution coefficients are generally favorable.

There are, however, several drawbacks with LPE. Frequently, the pres-

ence of surface defects such as incomplete melt removal, terraces,

pinholes and miniscus lines degrade the material. The thickness

uniformity can be poor and, for solid solution films, inherent composi-

tion gradients are present. Furthermore, LPE is a small scale batch

operation and heteroepitaxy can be difficult.

Molecular beam epitaxy is a method for growing epitaxial thin

films of semiconductors by impinging one or more thermal energy beams

of atoms or molecules onto a heated substrate under ultra-high vacuum

conditions. The distinguishing characteristic of MBE is the slow

growth rate (0.1-2 um/hr) that permits precise control of layer

thickness, composition and doping profiles. Furthermore, it is possi-

ble to achieve spatial resolution not offered by other techniques. As

with LPE, the growth temperatures are lower than those encountered in


CVD systems. Additionally, in situ analysis of the surface structure

and reaction conditions during growth are possible. However, the equip-

ment is very expensive and the throughput is low.

Commercially, the most successful technique for the growth of

epitaxial films has been chemical vapor deposition. The major advan-

tages given by CVD over MBE and LPE are relatively fast growth rates

(0.1-1 pm/min.), the ability to scale up research equipment to accomo-

date many large substrates, the availability of a variety of source

gases and the ease in which dopants may be changed during the deposition


Three source chemistries dominate the CVD process for III-V

materials: group III metalorganic (MOCVD), group V hydride and group

V halide sources. The MOCVD technique involves an irreversible pyroly-

sis reaction in which a group III metalorganic gaseous species is fed

to a cold-wall reactor along with a group V species (usually a hydride).

These species then contact a heated substrate, decompose and deposit

an epitaxial layer onto the substrate. Due to the nature of the source

gas, carbon, which is liberated during the decomposition of the group

III metalorganic species, is incorporated as an unintentional acceptor

in MOCVD grown epitaxial layers. Other unintentional dopants include

Mg, Zn and Si which are usually present as impurities in the metalor-

ganic source material. The cold-wall design employed in the MOCVD

system results in large amounts of deposition on the reactor wall

and dopants which have been deposited on the wall often desorb during

later stages of the deposition process. This makes it difficult to

grow epitaxial layers with abrupt doping profiles using MOCVD. A major

advantage associated with the MOCVD process is its ability to success-

fully grow epitaxial layers which contain aluminum.

The most successful CVD techniques for the growth of III-V

epitaxial layers, which do not contain aluminum, have been the hydride

and chloride systems. A schematic representation of the chloride CVD

system is shown in Figure 1-2. The reactor consists of source, mixing

and deposition zones which are usually operated at 100 kPa pressure.

Due to the exothermic nature of the overall deposition reaction, the

reactor is hot-wall design, the temperature of the mixing zone is

greater than or equal to that of the source zone and the deposition

zone temperature is normally less than that of the source zone. Typi-

cally, hydrogen is used as the carrier gas with the concentration of

the group V chloride in the inlet vapor being on the order of 1% by

volume. The group III source is either the III-V stoichiometric com-

pound (in order to avoid an initial source transient) or the group III

liquid metal saturated with the group V element (generally available in

higher purity). Upon entering the reactor, the group V chloride decom-

poses to form primarily V2, V4 and HC1 vapor species. The HC1 then

reacts with the group III source to form III-C1 and other high chlorides.

The mixing zone allows the species in the vapor to equilibrate while

being transported to the lower temperature deposition zone where the

group III and V vapor species react at the substrate surface to deposit

an epitaxial layer. When a liquid group III source is used, the ratio

of group III to group V atoms in the vapor is fixed at approximately 3

since essentially all of the chlorine atoms on the group V chloride

react to predominantly form III-C1. The use of a III-V stoichiometric

compound as the group III source limits the III/V ratio to a maximum

value of 1 since one group V atom is released from the solid for each

group III atom that reacts to form III-C1.




0 B0
O- --


U0 U


o- S.

NO -
0 1

0 0 4-
- C (- a
4 > -1 1


.- F- 4-',




The hydride CVD process is shown schematically in Figure 1-3.

The source zone of the hydride system consists of two mass transfer

isolated inlets, one for the group V species and one for the group III

species. The group V hydride, at a typical inlet composition of 1%,

is introduced into the source zone where it decomposes to form primarily

V2, V4 and H2. As in the chloride process, the group III element is

transported principally as the mono-chloride by the reaction of HC1

(typically the HC1 inlet concentration is 1%) with the liquid group III

metal. One major advantage the hydride system provides over the chlo-

ride is the ability to vary the vapor III/V ratio by adjusting the in-

let compositions or flowrates of VH3 and HC1. Typically, the source

and mixing zones in the hydride system are operated at higher temper-

atures than those of the chloride system in order to increase the rate

of VH3 decomposition (the decomposition kinetics of VH3 are much slower

than VC13). Again, hydrogen is usually used as the carrier gas and the

mixing and deposition zones provide functions equivalent to those in

the chloride system. Indeed, the equilibrium chemistry of the two

systems are identical after the source zone.

Both the chloride and hydride systems are hot-wall designs (heated

from the outside of the reactor tube by conduction). Therefore, depo-

sition does not occur on the reactor wall and sharp dopant profiles and

heterostructures may be grown. Unfortunately, the hot-wall design

allows interactions between the vapor and the reactor wall (usually

quartz) which results in the introduction of Si and 0 into the vapor

phase. The unintentional incorporation of Si into hydride and chloride

grown epitaxial layers is a major problem in these processes.



> -= 3



> ()






I L.

-3 *r-
0- -


The source materials used in the chloride and hydride processes

are the purest of all the III-V CVD systems. All of the systems use

H2 as a carrier gas which is usually diffused through a palladium

barrier in order to obtain extremely pure H2. The VC13 used in the

chloride system is a liquid which is introduced to the reactor by

bubbling H2 through it at a controlled temperature. The impurities

present in VC13 liquid are typically very small and the bubbling pro-

cess tends to further purify the inlet gas over that of the liquid.

The HC1 used in the hydride system is a source of many impurities if

it is taken from a high pressure gas cylinder. This problem can be

circumvented, however, by using AsC13 as a source of chlorine atoms,

cracking the AsC13 at high temperature in the presence of H2, and

depositing the As as a solid in a trap before introduction to the

hydride CVD reactor.

An understanding of the chemistry involved in the chloride and

hydride CVD processes is essential in order to advance these technolo-

gies. In this study, a complex chemical equilibrium analysis is pre-

sented which identifies the principal vapor phase species which must

be accounted for in order to understand these CVD processes. The

effects of reactor temperature, pressure and inlet concentration on

the equilibrium chemistry of each process are evaluated. Further,

species which may be added to or removed from the system in order to

suppress the amount of silicon incorporated into the epitaxial layer

are identified. The influence exerted by the vapor phase on the

point defect structure of the epitaxial layer is discussed relative

to the unintentional incorporation of silicon.

Ban [4] has suggested that the sluggish decomposition kinetics

of PH3 and AsH3 may prohibit the use of a thermodynamic analysis of

the hydride process. Therefore, the decomposition rates of NH3, PH3

and AsH3 were studied in order to ascertain the degree to which an

analysis based on the assumption of chemical equilibrium can be applied

to the hydride process.

Both the chloride and hydride processes were investigated for the

deposition of homoepitaxial GaAs and InP. Due to the application of

a consistent basis set of operating parameters, direct comparisons

between the systems are made.


Impurities in GaAs Epitaxial Films Grown by the Chloride Process

The feasibility of applying the chloride system CVD technique to

the epitaxial growth of high purity GaAs was first demonstrated by

Knight et al. [5] and Effer [6]. Initially, the commercially avail-

able AsC13 contained sufficient impurities to cause significant con-

tamination of the epitaxial layers and therefore the purity of the

feed materials was believed to be the controlling factor for this

system [7]. As better quality AsC13 became available, Cairns and

Fairman [8,9] and DiLorenzo et al. [10] found that an increase in the

AsC13 mole fraction in the inlet gas stream resulted in a decrease in

unintentional impurity incorporation in the epitaxial layer.

For materials grown in their laboratory, DiLorenzo and Moore [11]

identified the primary unintentional dopant as being silicon, through

the use of photoluminescence spectra. Further, they proposed a thermo-

dynamic model for the generation of vapor phase chlorosilanes as a

result of the interaction of HC1 with the quartz (Si02) reactor wall

and presented an expression for the activity of solid silicon (i.e. as

an impurity) as a function of the partial pressures of the chloro-

silanes. Their model showed that increasing the vapor HC1 concentra-

tion (e.g. as a result of AsC13 decomposition) decreased the condensed

phase silicon activity by further stabilizing the silicon species in

the vapor phase in the form of chlorosilanes. Additionally, their


model predicted that the generation of vapor phase silicon species

could be supressed by the introduction of H20 vapor into the system.

Rai-Choudhury [12] performed a thermodynamic analysis on the in-

corporation of silicon into GaAs epitaxial layers. His results reflected

those of DiLorenzo and Moore [11] when considering the effects of H20

and HC1, but he also showed that higher temperatures increased the

amount of vapor phase silicon species.

The work of Ashen et al. [13] further supported the conclusion that

silicon was an impurity in GaAs. A BN lined reactor was used to grow

epitaxial layers from liquid Ga sources which were doped with Si. Com-

paring the electrical characteristics of these epitaxial layers to layers

grown from pure Ga sources provided qualitative evidence for the presence

of Si in GaAs. The effect of AsC13 concentration on the amount of Si in-

corporated into the epitaxial layer was also verified by their experiments.

Additionally, these studies provided evidence which indicated that Si was

much more likely to reside on a Ga site than an As site and therefore,

behaves as a donor. This conclusion was also supported by Beiden et al.


Wolfe, Stillman and Korn [15] have identified, through intentional

doping and determination of ionization energies, three unintentional

impurities, Si, C and one unknown (possibly Te), in GaAs grown by the

chloride CVD system. Also, due to the results of Solomon [16] which

showed that oxygen may be a shallow donor in GaAs, they attempted to dope

the epitaxial layer with oxygen by adding Ga203 to the liquid gallium

source. The oxygen, however, was not incorporated into the epitaxial

layer, but did reduce the amount of silicon which was incorporated into

the layer. This reduction in background doping due to the presence


of oxygen was also investigated by Palm et al. [17] by injecting a

hydrogen-oxygen mixture into the mixing zone of a chloride system

CVD reactor. Using silane as an intentional dopant, the presence of

oxygen was found to reduce the incorporation of silicon in the epitaxial

layers by as much as four orders of magnitude.

Seki et al. [18] performed a thermodynamic analysis of the GaAs

chloride process in order to identify the effects of inerts, HC1 and

substrate temperature on the activity of silicon in the epitaxial

layers. The analysis predicted that increasing the HC1 content or de-

creasing the substrate temperature lowered the silicon activity. In

addition, replacing the hydrogen carrier gas with an inert gas caused

a very large reduction in the silicon activity.

The effect of replacing the hydrogen carrier gas with an inert was

investigated by Ozeki et al. [19]. Through far infrared photocon-

ductivity measurements, it was determined that sulfur was the dominant

residual donor present in epitaxial GaAs when N2 was used in place of

H2 as the carrier gas. It was also found that the dominant residual

donor when H2 was used as the carrier gas was sometimes Si and some-

times S. A possible source of S in the system was not discussed

(although it was presumably in the feed gases) and elaboration on the

growth conditions which caused Si or S to be dominant was not provided.

A thermodynamic analysis of the chloride CVD system performed by

Boucher and Hollan [20] assumed solid GaAs as the group III source

material. By comparison with experiment, it was found that the

dominant group III and group V species present in the vapor were GaCI

and As4. Under the experimental conditions investigated, the deposition

process was kinetically controlled with an activation energy of -40

kcal/mole, and reproducible growth conditions could be attained only

if 10% or less of the thermodynamically available GaAs was deposited

from the vapor phase.

Gentner et al. [21] also studied the chloride process experimentally

and presented a thermodynamic analysis over a greater range of tempera-

ture, pressure and inlet AsC13 composition than did previous investiga-

tors. They found that As2 became the dominant group V species below

10 kPa pressure and that GaCl was always the dominant group III species.

At large AsC13 inlet compositions, the higher gallium chlorides became

more pronounced but never competed with the monochloride as the dominant

species. They concluded, based on a kinetic model [22], a mass transfer

model and experimental results, that the deposition of GaAs was kinet-

ically rather than mass transfer controlled.

Shaw [23] studied the transport kinetics of the GaAs chloride

system in the source and deposition zones. He found an activation

energy of 49.1 kcal/mole in reasonable agreement with that of Boucher

and Hollan [20], for a surface reaction associated with the deposition


Impurities in InP Epitaxial Films Grown by the Chloride Process

The epitaxial growth of InP using a chloride CVD system was first

demonstrated by Clark et al. [24] and later by Hales et al. [25]. Both

groups of investigators reported limitations on the purity of their

epitaxial layers due to unintentional dopants. Joyce and Williams [26]

tentatively identified the impurities as being Si and Zn acceptors.

They also found evidence of a donor which was thought to be amphoteric


The dependence of unintentional doping on PC13 mole fraction in

the InP chloride system was first reported by Clark [27]. The simi-

larity between the GaAs and InP chloride system reactors combined with

the analogous dependencies on the group V hydride mole fraction supported

the belief that Si was an impurity in InP epitaxial layers. Clarke [28]

later studied the effect of III/V ratio in the vapor phase on the un-

intentional doping of InP epi-layers and found p-type conductivity for

III/V < 3 and n-type for III/V > 3, with a minimum in the free carrier

concentration at III/V z 3. No explanation was offered for these


Easton [29] investigated the unintentional doping of InP epitaxial

layers grown by the chloride system and concluded that S (acting as a

donor) was the major impurity and that the origin of the S was the PC13

liquid source. Using mass spectrometric analysis, Easton found Si, S

and Zn present in the PC13 source at levels between 1 ppm and 10 ppm

(atomic) and Fe, Cu, Cd and Sn at -0.7 ppm. These same elements were

found in the unused bulk In liquid at levels below 0.1 ppm. Analysis

of the used In source liquid revealed impurity levels approximately 10

times larger than those in the unused liquid.

These results support the work of Weiner [30] who proposed models

for the contamination of a Ga liquid source by Si in the GaAs and GaP

systems. Weiner's results showed that the liquid group III metal im-

purity level increased as the exposure to the CVD environment increased.

He also found the level of Si contamination to be inversely related to

the partial pressure of H20 in the system.

Fairhurst et al. [31] studied the InP halide system using both

PC13 and PBr3. They found that oxyhalide salts were present in both


phosphorous liquids at approximately the 100 ppm level. The presence

of oxygen was expected to decrease the level of impurity incorporation

in the epitaxial layers. This effect was not observed however, pre-

sumably due to this level of oxygen contamination being too low to be

significant. Equilibrium calculations were performed which showed

InCl and P4 to be the dominant group III and V species in the vapor

over a temperature range of 850 K to 1150 K and an inlet PC13 mole

fraction range of 0.1% to 6%. These results agreed with those of

Boucher and Hollan [20] for the analogous GaAs system.

Hales and Knight [32] investigated the effect of introducing

oxygen into the system in order to reduce the level of impurities in

InP. They observed a monotonic decrease in free electron density for

additions of 02 up to 24 ppm. The electron mobility (measured at

77 K) however reached a very broad maximum at approximately 16 ppm

of added 02, which suggests that oxygen was becoming incorporated

into the epitaxial layer and that there is a limit to the degree of

benefit which may be obtained through oxygen addition. They also

observed POC13 to be an impurity in the liquid PC13 used in the

chloride system.

Investigations of the dependence of impurity incorporation on

PC13 inlet composition, total flowrate and deposition zone temperature

were carried out by Chevrier et al. [33]. They observed a decrease

in free carrier concentration with increasing PC13 concentration,

as did other investigators, but also found that the impurity concen-

tration increased with increasing total flowrate. This velocity effect

had not been reported before and suggests the presence of a mass

transfer resistance at the group III source (if impurities are picked

up from the liquid metal) or at the substrate in the deposition zone.

They also studied the intentional doping of InP as a function of

deposition zone temperature using SnC14. Lower free electron con-

centrations and higher electron mobilities were observed when the

deposition zone temperature was decreased from 950 K to 910 K. Thus,

the uptake of group IV impurities (Sn, Si, etc.) was apparently re-

duced by lowering the deposition zone temperature.

Cardwell et al. [34] found kinetic limitations in both the source

and deposition zones. The previously reported effect of PC13 mole

fraction on impurity levels in the epitaxial layers was observed.

Intentional doping of InP using Sn followed the same behavior as that

of unintentional dopants and therefore supports the use of Sn for

studies regarding the reduction of unintentional impurities. In con-

trast to Chevrier et al. [33], no dependence of impurity uptake on

total flowrate was found.

A thermodynamic analysis comparing the GaAs and InP chloride

systems using the stoichiometric III-V solid as the group III source

material was reported by Shaw [35]. His results also confirmed GaC1,

As4, InCl and P4 to be the dominant group III and V vapor species in

these systems. Further, the degree of supersaturation in the deposi-

tion zone was calculated to be less for InP than for GaAs under

analogous conditions. Since solid III-V source materials were em-

ployed etching conditions were predicted whenever the deposition zone

temperature was greater than that of the source zone.

Impurities in GaAs Epitaxial Films Grown by the Hydride Process

The feasibility of applying a hydride CVD system for the growth of

high purity GaAs was demonstrated by Enstrom and Peterson [36]. Since


the hydride system consists of a hot-wall quartz reactor with the

elements H, C1, Ga and As present in the vapor, one would expect it

to show an unintentional impurity incorporation problem similar to that

of the chloride system. Pogge and Kemlage [37] investigated the

effects of HC1, AsH3 and PH3 on the unintentional doping of GaAs

and GaP grown by the hydride system. They found that the free carrier

concentration decreased with increasing HC1, AsH3 or PH3 composition.

The effect of HC1 was less than that of the group V hydrides and

changes in PH3 showed larger effects than did AsH3. They concluded

that the HC1 effect on the vapor phase composition was similar to that

of the chloride system. Further, they concluded that As4 and P4 caused

blockage of the available surface sites on the substrate due to the

large size of these molecules. The unintentional dopant was assumed

to be Si generated from reactions with the quartz wall.

Kennedy et al. [38] investigated the effect of HC1 inlet com-

position and additions of HC1 downstream of the source zone on the un-

intentional doping of GaAs grown in a hydride CVD reactor. Increasing

the HC1 inlet composition greatly reduced the free carrier density in

the epitaxial layer. In contrast to this result, however, when HC1

was added downstream of the source zone the free carrier density was

found to increase. This led to the conclusion that the equilibrium

model as proposed by DiLorenzo and Moore [11] for the chloride system

was not applicable to the hydride system. However, the HC1 which was

injected may not have been as pure as that which was generated from

the decomposition of AsC13 in the chloride system and therefore may

have introduced additional impurities into the epi-layer. These effects

were also observed by Enstrom and Appert [39].


The work of Skromme et al. [40] identified some of the unintentional

donors and acceptors present in GaAs and InP prepared by the hydride

CVD system. They found C, Zn, Cu and Mn as acceptors and Si, S and Ge

as donors in GaAs. Epitaxial InP was found to contain Zn, C or Mg and

an unidentifiable acceptor along with Si and S as donors. Additionally,

in one of the laboratories (Honeywell) where the epi-layers were grown,

an increase in the impurity concentration in epitaxial GaAs was noted as

the HC1 gas cylinders "aged". This effect, however, was not observed

at the other laboratory (Hanscom AFB).

The effect of pressure was studied experimentally by Putz et al.

[41] from 1 kPa to 100 kPa. They found that the unintentional doping

of GaAs was reduced at pressures below 100 kPa.

Impurities in InP Epitaxial Films Grown by the Hydride Process

Growth of InP epitaxial layers using the hydride system has been

demonstrated by Olsen [42] and Hyder [43] among others. Both of these

investigators observed unintentional impurity incorporation similar to

that occurring in the GaAs system. Hyder also found that for the ter-

nary InxGa -xAs (x=0.53), a maximum in electron mobility occurred when

the III/V ratio in the vapor phase was approximately 2, but the effect

of III/V ratio on free carrier concentration was not discussed.

Zinkiewicz et al. [44] also studied the growth of InP and the ternary

InxGa -xAs in the hydride system. They found Zn, Cu and Hg to be pre-

sent as unintentional donors.

Anderson [45] studied the hydride system for InP growth in order

to determine the effect of HC1 mole fraction, H2 flowrate and mixing

zone temperature on unintentional impurity incorporation. He found that

these parameters caused only minor changes in the electrical behavior

of the InP epitaxial layers. This suggests that the InP hydride sys-

tem may perform somewhat differently than the GaAs hydride system.

Jones [46] performed a thermodynamic analysis of the InP hydride

system in order to understand the effect of process parameters on unin-

tentional Si incorporation. The calculations predicted that decreasing

temperatures lowered the silicon activity in the epitaxial layer. Addi-

tionally, the silicon activity was decreased by increasing the PH3 and/or

HC1 inlet composition. Very little effect was noted when HC1 was added

downstream of the source zone. His analysis used liquid In as the group

III source material and showed InCl and P4 to be the dominant group III

and V vapor species.

Ban and Ettenberg [47] coupled a mass spectrometer to a hydride

system reactor used for the growth of InxGal-xP. They compared measured

vapor species to those predicted by a thermodynamic model and concluded

that the model yielded an acceptable representation of the system. The

major shortcomings of the model were an overprediction of the amount of

InCl generated from the heterogeneous reaction of HC1 and In liquid, and

a predicted greater degree of dissociation for PH3 than was measured.

Due to the slow decomposition kinetics of PH3 and the potential mass

transfer and kinetic limitations associated with heterogeneous reactions,

these discrepancies were not surprising. Their mass spectrometric

measurements identified the major vapor phase species as being InCl,

GaC1, HC1, PH3, P2, P4 and H2.

Usui and Watanabe [48] investigated the effects of temperature and

oxygen additions on the unintentional doping of hydride grown InP. Ad-

ditions of 02 in the 1 ppm to 10 ppm range decreased the free carrier

concentration about one order of magnitude, but further additions caused


the free carrier concentration to increase, presumably due to uptake

of oxygen by the epitaxial layer. The liquid In source that was used

in these experiments was found to have a gettering effect on impuri-

ties in the inlet gases. Unused In showed less than 1 ppm levels of

Si, S, Sn, Te, Zn, Fe and Cu, while In exposed to the CVD environment

contained increased levels (-2 ppm) of Fe, Cu and Sn. Increasing the

source zone temperature caused a decrease in the free carrier concen-

trations in InP epi-layers due to an increased ability of the In

liquid to getter impurities at high temperature. Thus, the purity of

source gases still appears to be a major problem in the hydride system.

The Thermal Decomposition of NH

The thermal decomposition of the trihydrides of N, P and As have

been studied by many investigators and, for temperatures below 1500 K,

a general consensus exists that these reactions are almost entirely

heterogeneous in nature. Bamford and Tipper [49] have reviewed the

literature relevant to the homogeneous pyrolysis of ammonia at temper-

ature above 2000 K and found the reaction to be characterized by an

activation energy of approximately 100 kcal/mole. Based on the ob-

served activation energy and the results of experiments with deuterated

ammonia, the initiating step in the pyrolysis reaction sequence was

proposed to be:

NH3 + M ---> NH + H2 + M 2-1

where M represents any gas molecule. They also found evidence that a

reaction which forms NH3 is likely to be present in the decomposition

chain reaction sequence, but were unable to identify the nature of

this reaction.


The decomposition of NH3 in a quartz vessel was first studied by

Bodenstein and Kranendieck [50] using a manometric method. The amount

of surface area present in the reactor was varied by the addition of

quartz fibers. They concluded that within the temperature range of

their study (1063 K to 1153 K), the reaction appeared to be first order

and was entirely heterogeneous in nature. Further, they found no

change in the reaction rate when H2 or N2 additions were made to the


Hinshelwood and Burke [51] investigated NH3 decomposition in a

quartz vessel for temperatures as high as 1323 K. They also concluded

that the reaction was dominated by the heterogeneous component and

demonstrated a first order dependence on NH3. Additions of H2 to the

reactor decreased the reaction rate while N2 additions were ineffective.

Christiansen and Knuth [52] suggested the following mechanism

for the heterogeneous pyrolysis of NH3 in a quartz vessel:

NH + S NH + H2 + S 2-2
3 2<--
NH NH* 2-3
NH* + NH3< N2 + 2H2 2-4

where S represents a surface site. Their experiments were carried out

in a reactor vessel having a surface to volume ratio (S/V) of 1 and a

surface area of 0.02 M2. They concluded that the forward component of

reaction 2-2 was the rate limiting step and over the temperature range

of 1062 K to 1132 K, this reaction was characterized by an activation

energy of 43 + 5 kcal/mole with an Arrhenius type frequency factor of
5 -1
4.5 x 10 s-

The investigation of Russow and Pewsner [53] into the decomposi-

tion of NH3 in a quartz reactor demonstrated that the reaction followed

a first order dependence with respect to NH3 partial pressure. They

reported an activation energy of 38.2 kcal/mole for the pyrolysis


The decomposition of NH3 on quartz was reported by Szabo and

Ordogh [54] to follow a 1/2 order dependence with NH3 pressure at 913 K

and a first order dependence at 1013 K. The reduced order of reaction

at 913 K was reported to be a result of H2 competing with NH3 for ad-

sorption sites on the quartz reactor wall. The presence of H2 or 02

in the system was found to decrease the reaction rate while no reaction

rate changes were observed when N2 was added. Initial partial pressures

of NH3 ranging from 6.6 x 103 Pa to 2.6 x 104 Pa were tried, but no

changes were observed in the order of reaction over this range.

An activation energy of 34 + 2 kcal/mole for the decomposition of

NH3 on quartz was reported by Voelter and Schoen [55]. The frequency

factor associated with an Arrhenius type temperature dependence was

560 s but they did not report the surface area of the reactor used.

Over the temperature range of 1023 K to 1173 K, the reaction was found

to be first order with respect to NH3 pressure.

Kelvin [56] utilized an infrared spectrometer to measure the out-

let NH3 composition from a plug flow quartz reactor over the tempera-

ture range 833 K to 1373 K. He found that the reaction rate varied

with the reactor S/V to the 0.75 power, confirming the heterogeneous

nature of the reaction. Hydrogen was found to exhibit a strong in-

hibitory effect on the decomposition reaction nature due, primarily,

to a reaction between NH radicals and H2, which forms NH3. Additions

of N2, Ar or He to the reactor resulted only in a dilutent effect.


The Thermal Decomposition of PH3

The decomposition of PH3 in a quartz reactor was first studied by

van't Hoff and Kooj [57]. They found the reaction to be first order

in PH3 pressure over the temperature range 310 K to 512 K. The reac-

tion was believed to be heterogeneous as a result of the increase in

reaction rate, which occurred upon the addition of quartz fibers to the

system. An activation energy of 46.4 kcal/mole was reported.

Trautz and Bhandarkar [58] reported a transition from hetero-

geneous to homogeneous reaction kinetics at 940 K for the decomposition

of PH3 in a porcelain bulb. They reported activation energies of 59

kcal/mole for temperatures below 940 K and 116 kcal/mole for tempera-

tures above 940 K. Based on the large degree of scatter in their re-

sults and the fact that other investigators have not seen this transi-

tion, it is doubtful that a homogeneous decomposition reaction was

actually observed.

Hinshelwood and Topley [59] studied the decomposition of PH3 in a

155 cm3 quartz bulb with surface areas of 210 cm2 to 1600 cm2. They

concluded that in the temperature range 848 K to 1042 K, the reaction

was first order and behaved in a heterogeneous manner with an acti-

vation energy of 46 + 4 kcal/mole. The reaction rate was found to in-

crease with increasing S/V to the 0.8 power.

Devyatykh et al. [60] studied the decomposition of PH3 on glass

and Si over the temperature range 740 K < T < 822 K. They found PH3

decomposition to be first order with activation energies of 44.2 kcal/

mole and 55.3 kcal/mole on the glass and Si surfaces, respectively.

The decomposition of SbH3 was studied on an antimony surface and was

found to have an activation energy of 7.7 kcal/mole. The SbH3 decom-

position reaction was investigated over the temperature range 298 K <

T < 364 K and the reaction order was observed to change from half or-

der at 298 K to first order at 364 K.

The Thermal Decomposition of AsH3

The decomposition of AsH3 on glass, As and Sb has been studied by

Tamaru [61]. The reaction was found to proceed most rapidly on the Sb

surface and slowest on glass. Adding H2 to the system had no effect

on reaction rate and no isotopic exchange was observed when D2 was

added. Tamaru proposed that the reaction mechanism consisted of AsH3

adsorbing on the surface followed by sequential stripping of the hydro-

gen atoms off of the As atom. He believed the rate determining step

to be the removal of the first H atom and assigned an activation energy

of 23.2 kcal/mole to this reaction. Tamaru [62] later attempted the

calculation of the reaction rate constants for AsH3 and SbH3 decomposi-

tion on As and Sb surfaces using a model based on activated complex

theory. His predicted rate constant for AsH3 decomposition was six

orders of magnitude below the observed value while the predicted rate

constant for SbH3 was two orders of magnitude low. The restrictive

assumptions, which required all of the hydrogen bond energies to be the

same and all partition functions to have the value of one, were proba-

bly the reasons for the poor results.

Devyatykh et al. [63] found that the decomposition of AsH3 on a

Si surface obeyed first order kinetics and was characterized by an

activation energy of 50.9 kcal/mole. Their experiments were conducted

from a temperature of 659 K to 707 K.

Kedyarkin and Zorin [64] reported an activation energy of 25.5

kcal/mole, in good agreement with that of Tamaru [61], for the decom-

position of AsH3 on As. They found the reaction to be first order with

respect to AsH3 pressure and heterogeneous in nature. The tempera-

ture range of their study, however, was somewhat restricted (543 < T <

583 K).

The only reported study of AsH3 decomposition on a quartz sur-

face was undertaken by Frolov et al. [65]. Their investigations were

carried out on quartz, Ge, Ga and GaAs surfaces over a temperature

range of 699 K to 909 K. The activation energy for AsH3 decomposition

on quartz was found to be 32.6 kcal/mole with an Arrhenius frequency

factor of 513 s- The activation energies for AsH3 decomposition on

Ge, Ga, Te doped GaAs and Cr doped GaAs were reported as 54, 30, 45

and 27 kcal/mole, respectively, in the presence of H2. When He was

used as the carrier gas in place of H2, these activation energies in-

creased slightly. No explanation was provided for this observation.

Their experimental apparatus consisted of an open tube coupled to an

infrared spectrometer. The surface area or surface to volume ratio of

the reactor was not reported.



Performing a detailed analysis on CVD processes as complex as the

hydride and chloride processes presents a formidable task if results

which allow direct comparison of the systems are desired. A rigorous

treatment would require the solution of the mass, energy and momentum

equations with variable properties and kinetic expressions for many of

the homogeneous and heterogeneous reactions which are present. The

current lack of knowledge surrounding the fundamental chemistry which

underlies these processes precludes this type of analysis. Many of

the essential differences between these processes may, however, be

elucidated by the application of a model based on the assumption of

chemical equilibrium. Due to the high temperatures employed in these

processes, it is expected that many of the chemical reactions proceed

at a very high rate. Therefore, homogeneous reactions are expected to

be near equilibrium. Heterogeneous reactions may be impeded by the

existence of mass transfer barriers. These barriers may be approxi-

mated by either allowing the reaction to reach equilibrium or by ne-

glecting the reaction. The results related to heterogeneous reactions

are therefore expected to be semiquantitative in nature in that they

provide limits to the system composition. These results may then be

compared to experiments in order to determine the degree of influence

exerted by the heterogeneous reactions on the systems.

The resources employed in order to affect the analysis per-

formed in this study include a computer code for the calculation

of multiphase equilibria in systems having many chemical species

(typically more than 20), models for each zone of these CVD processes

which are based on chemical equilibrium constrained to account for

actual mass transfer or kinetic limitations, a model for the point

defect chemistry in the solid epitaxial layer and, finally, a set

of consistent thermodynamic properties for each chemical species in

the system. The computer code, reactor zone models and solid state

defect chemistry are discussed in this Chapter. The choice and

analysis of a set of thermodynamic properties are described in Chap-

ter Four.

Method of Calculation for Complex Chemical Equilibrium

The calculation of complex chemical equilibrium in multicomponent,

multiphase systems has been reviewed most completely by Smith [66].

Essentially, there are two statements of the solution to this problem.

Nonstoichiometric methods, such as the popular Rand algorithm [67],

directly minimize the Gibbs energy of the total system in order to ob-

tain a solution without recourse to a specific set of formation re-

action equations. Stoichiometric methods [68] require that an inde-

pendent set of chemical reactions be in equilibrium. Generally, a

formation reaction is written for each species present in the system

and the corresponding equilibrium constant for each reaction is gen-

erated from the Gibbs energy change of the reaction.

An extension of the Rand algorithm to include not only a gas

phase with an inert species, but also a multicomponent solution and

pure condensed phases, was developed by Anderson [69]. This algorithm

was initially applied to the hydride and chloride CVD systems, but

was susceptible to finding local minima. In particular, some com-

ponent mole fractions sought were as low as 0.1 ppb. The contribu-

tion to the system Gibbs energy for these components was quite small

and the resulting component mole changes were not capable of releasing

the Gibbs energy of the system from the local minima.

A stoichiometric algorithm, presented in Appendix A, was there-

fore developed which was superior to the extended Rand algorithm since

a linearized Gibbs energy function was not required. The stoichio-

metric algorithm performed well for all of the systems studied (the

results were compared to other investigator's calculations and were

found to be independent with respect to initial guesses) and yielded

results which were in agreement with those of the extended Rand al-

gorithm, when it could be successfully applied. The amount of com-

puter memory required for the stoichiometric algorithm was found to

be much less than that required by the extended Rand algorithm in

order to solve identical systems.

The input data which was required in order to perform the cal-

culations consisted of the standard enthalpy and entropy of formation

and heat capacity for each species along with the system temperature,

pressure and inlet composition.

Aside from numerical difficulties, the two major factors, which

must be considered in determining the overall accuracy of the calcu-

lated equilibrium compositions, are the choice of species postulated

to be present in the system and the accuracy of the thermodynamic

data chosen to represent each species. Choosing an appropriate set

of species, which accurately represent the system at equilibrium, is

an inherent difficulty in the calculation of multicomponent equilib-

rium. A true calculation of equilibrium in a given system must in-

clude any chemical species which is formed from any combination of

the elements present in the system. The compilation of such a com-

plete thermodynamic data base can represent a nearly impossible task,

even for systems consisting of only a few elements.

It is important to realize that anytime a possible species is

not included in the data set, a constrained equilibrium calculation

will result. This is most easily understood if the calculation of

chemical equilibrium is considered from the viewpoint of the Rand

algorithm. In the Rand algorithm, multicomponent equilibrium repre-

sents the optimal distribution of a given quantity of elements among

a set of chemical species. The optimizing function for a constant

pressure system is the minimization of the total Gibbs energy. There-

fore, as the number of available chemical species is decreased, the

elements are constrained to reside in a smaller selection of mole-

cules. This causes a shift in the calculated compositions in order

to satisfy the atom balance while keeping the Gibbs energy of the

system as low as possible. In general, the exclusion of a species

will impact the equilibrium composition of the remaining species con-

taining similar atoms that are in the vicinity or below the equilib-

rium composition of the excluded species.

The procedure for developing a species list first excludes those

species not expected to be present because of severe kinetic limita-

tions. In practice, this species set is developed by including only

those species which have been experimentally observed in the system or

in appropriate subsystems. As an example, mass spectroscopic work in


the CVD of GaAs has indicated approximately 10 species, but observa-

tions in the subsystem Si-H-Cl indicates approximately 15 additional

species. The next step consists of generating an initial thermodynamic

data base for all species. This includes thermodynamic compilations

(such as the JANAF tables), data bases of other investigators for sim-

ilar systems and crude estimates for the remaining species. Initial

equilibrium calculations are then performed at the extreme limits of

the study and those species whose compositions are more than three or-

ders of magnitude in mole fraction below the range of interest are ex-

cluded. Finally, the initial thermodynamic data base is completed by

referring to the literature and the results of internal consistency


The sensitivity of the results to errors in the thermodynamic data

was investigated by Smith [70] in terms of a Jacobian, which relates

the changes in the calculated results to changes in the input parameters.

The first order approximation to the result was

N 6n.
16nil < I | 16 1 3-1
j=1 j

Here, n. is the number of moles of specie i present, pv is the standard
1 J
chemical potential of specie j and N is the total of components. This

expression, while simple in form, is extremely difficult to evaluate

due to the complicated and implicit nature of the function n.( ) for

all values of i. If problems seem apparent for some species, this

function can be numerically evaluated. The work of Sirtl and Hunt [71]

and similar calculations performed in Chapter Four showed by means of

a calculated example the effects of changes in the enthalpy of formation

of SiHC13 on the predicted equilibrium ratio of SiC14/SiHC13. This

ratio was found to change by approximately two orders of magnitude

for a 10% change in the standard enthalpy of formation. The shape

of the curve relating this ratio to temperature was also found to

change markedly. Therefore, it is extremely important to critically

review the thermodynamic data set in order to perform meaningful

equilibrium calculations. The absolute composition of the calculated

solution can be no better than the data set employed. Extreme care

must also be used when comparing calculated equilibrium compositions

to experimental process compositions as the latter include possible

kinetic limitations. However, these calculations are of great value

in predicting the directions of compositional changes, particularly

at the high temperatures and low pressures encountered in this study.

Thermodynamic Models of CVD

Models for the CVD Source and Pre-Source Zones

Each of the CVD systems under investigation may be separated,

based on temperature or composition, into the pre-source, source, mixing

and deposition zones. The pre-source zone was investigated as a source

of Si by considering the equilibrium gas phase Si-content in the sys-

tem: Si02(c) in excess, carrier gas (H2 or inert) and vapor reactant

(VH3, VC13 or HC1).

Historically, the chloride process has employed two different

group III source materials (III-V(c) and III(1)) and thus required

two separate model formulations. The III-V(c) chloride source zone

model considered the system: Si02(c) in excess, III-V(c) in excess,

carrier gas (H2 or inert) and VC13(g).

The chloride process which used a group III(1) as the group III

source material was a more complicated situation. Shaw [23] has

studied this source zone and found that, following an initial transient,

a constant rate of mass loss of material occurred. An overall mass
balance on the source boat yields the following expression:

d [n(MI + MV) + nl(xMi + (l-x)Mv)]

nV(yMII + (l-y)Mv) = constant 3-2

while a group III component mass balance on the source boat produces the

d [ ncM + n1xM = nVyMIII 3
dt 2 CMIII l1] =VYII 3-3

In these expressions: nc and n1 are the moles of atoms in the solid
GaAs crust and the liquid III-V mixture, MIII and MV are the molecular

weights of the group III and V elements, x and y are the mole fractions
of the group III element in the liquid and vapor phase, and nV is the
molar rate at which vapor species are formed. If it is assumed that the

solid and liquid phases are in equilibrium, the liquid phase mole frac-

tion is a function of temperature only and therefore constant for a
given process condition. Furthermore, if the actual kinetic processes

produce a steady state value of the crust thickness, the first term on
the left side of both equations is zero. With these assumptions, equa-
*V dn1
tions 3-2 and 3-3 can be solved to show that n = dt and x=y. That

is, the rate at which group III and V atoms are introduced into the
vapor phase is equal to the rate of loss for the melt and the III/V
vapor atom ratio is the same as that in the melt. Another way of view-
ing the situation is to consider the three phase equilibrium problem.

The activity of the group III and V elements in the solid compound can

vary greatly even though the stoichiometry (~1:1) can be very small

and therefore the sum of these two activities is nearly constant. The

large amount of melt in equilibrium with the solid will, however, fix

the activity of each element in the solid, with the group III activity

being considerably higher than that of the group V, which in turn fixes

the vapor phase fugacity. In the event that the assumption of constant

crust thickness is not valid, the dnc/dt terms in equations 3-2 and 3-3

must be retained and the result is

SV cons t (2x-1) dnc (2x-1) dnI 3-4
n = constant = -- dt 2y-I dt
2(x-y) dt 2y-1 dt
The limit that dt = 0 implies that y = 1/2 and the source can be con-
sidered to be pure solid compound. This limit is simply the first case

examined (solid compound source). Thus, an investigation of the two

source zones described here should establish the limits of operation for

the liquid source in the chloride process. In practice, the conditions

of operation may lie somewhere in between with the observed III/V ratio

providing an indicator of the relative rates. However, if x is a con-

stant as determined by the condition of solid-liquid equilibrium and y

is also a constant as witnessed by a constant growth rate, it follows
dnI dnc 1
that both d- and dc are constant. If dn/dt is dependent upon the
dt dt
crust thickness, nc, (i.e. a diffusion limited process), then it is

impossible for dnl/dt to be constant for a finite value of dnc/dt,

which implies operation at one of the limits.

The above considerations motivated a model for the liquid group

III source zone to consist of an ideal vapor phase in equilibrium with

excess Si02(c) and IIIxV-lx(1). The gas inlet stream contained VC13

and carrier gas (H2 or inert). The development of a thermodynamic

data set for the hypothetical specie IIIxV1_x(1) is presented in Chapter


Four. Thus, with this relation, the compound crust does not contri-

bute elements to the system.

Two source zones, one for the thermal decomposition of the group

V hydride and one for volatilization of the group III liquid, are found

in the hydride CVD process. The group V source zone was modeled as an

ideal vapor phase in equilibrium with excess Si02(c). The inlet gas

reactants were the hydride (VH3) and carrier gas (H2 or inert) at con-

stant temperature and pressure. The group III source zone included

excess pure group III liquid in equilibrium with excess Si02(c) and an

ideal vapor phase (HC1 plus carrier gas).

Models for the CVD Mixing and Deposition Zones

Since the only differences between the chloride and hydride sys-

tems exist in the source regions, the mixing and deposition zone models

were identical in both systems. An ideal vapor phase in equilibrium

with excess Si02(c) was used for the mixing zone model. Formation of

solid III-V compound was postulated to be kinetically hindered and thus

assumed not to exist in the mixing zone. As a result, it was possible

for this region to be supersaturated. The model also allowed the addi-

tion of various species (i.e. HC1, H20, VC13, VH3) in order to study

their effects on silicon activity. The gas reactant input for the mix-

ing zone was identical to that calculated from the equilibrium source

zone(s). Consistent with the source and mixing zone models, the vapor

phase of the deposition zone was assumed to be an ideal mixture. Due

to the large volumetric flowrate of gases and the relatively small

deposition rates in these CVD processes, the depletion of group III,

group V and silicon species in the vapor phase as a result of film

deposition or wall interaction was neglected. The equilibrium mixing

zone gas mixture served as the input to the deposition zone. Essen-

tially, the above assumption fixed the moles of each element in the

vapor and the new equilibrium composition was calculated as a result

of temperature change only. This model provided an upper bound for

the computed value of the Si activity since the lower temperatures

found in the deposition zone shifted the wall interaction towards

Si02(c) formation and including the III-V compound deposition with

Si incorporation would remove Si from the gas phase. This model,

therefore, assumes that the epi-film grows from a supersaturated vapor

mixture of pseudo-steady state composition. Furthermore, this pro-

cedure avoided having knowledge of the solid silicon activity co-

efficient. In order to implement this model, the III-V solid phase

was not included in the deposition zone, thus allowing calculation of

the degree of supersaturation in this zone.

The effect of not accounting for depletion of the group III and V

atoms from the vapor phase can be tested by the following simple anal-

ysis. The molar growth rate of an epitaxial layer is

gm = 1PA 3-5

gm = molar growth rate (moles/time)

g1 = linear growth rate (length/time)
p = compound molar density (moles/length3)

A = substrate area lengthh)

A typical set of operating parameters for a hydride CVD process

would specify a total volumetric flowrate of 500 SCCM through each

source zone having an inlet composition of 1% HC1 to the group III

source zone and 1% VH3 composition to the group V source zone. Assuming

that all of the HC1 reacts to form III-C1 results in 3.7 i-moles/s

of group III atoms transported. The molar flowrate of group V atoms

would also be 3.7 j-moles/s. Choosing as typical deposition parameters

a 2.54 cm diameter circular substrate, a linear growth rate of 1 im/min

and given the molar density of GaAs as 0.0367 moles/cm3 [72], the re-

sulting molar growth rate is 0.31 u-moles/s. Thus, in the worst case,

less than 10% of the III and V atoms are depleted. The smaller the

growth rate and substrate surface area or the larger the volumetric

flowrate, the better the approximation becomes. If reaction depletion

were indeed important, highly non-uniform film thickness would occur.

However, this is not experimentally observed. Similar analyses applied

to the GaAs chloride system and the analogous InP systems yield equiv-

alent results.

The activity of silicon in the epi-layer was further studied in

the presence and absence of the Si02 reactor wall. Since the deposition

zone is typically operated at a lower temperature than the source and

mixing zones, inclusion of the reactor wall would be expected to de-

crease the silicon activity as some of the silicon in the vapor phase

is redeposited on the reactor wall in the deposition zone, therefore,

providing an additional method of bounding the maximum value of the

silicon activity in the epitaxial layer. Justification for neglecting

the reactor wall lies in the heterogeneous nature of the gas-wall re-

action. Due to the lower temperature of the deposition zone (-873 K),

it is expected that this heterogeneous reaction does not equilibrate

as rapidly as it should in the source and mixing zones (-973 K). This

expectation arises from the fact that adsorption reaction rates de-

crease strongly, and to a lesser extent molecular diffusivities, with

decreasing temperature. Additionally, the mean residence time is

typically much smaller in the deposition zone. Thus, the reactor wall

in the deposition zone should not interact with the vapor phase as

strongly as it does in the source and mixing zones.

In order to carry out parametric analyses of the two processes,

"base cases" were chosen for each system around which each parameter

could be varied. The base cases were chosen from commonly used oper-

ating parameters reported in the literature shown in Tables 3-1 and

3-2, thus providing results which may be compared to experimental

results. The chloride system base case parameters were as follows.

Source Zone Temperature = Mixing Zone Temperature : 973 K

Deposition Zone Temperature : 873 K

Inlet VC13 Composition : 1%

Carrier Gas : H2

Pressure : 100 kPa

For the hydride system the following base case was chosen.

Source Zone Temperature = Mixing Zone Temperature : 973 K

Deposition Zone Temperature : 873 K

Inlet HC1 Concentration = Inlet VH3 Concentration : 1%

Carrier Gas : H2

Pressure : 100 kPa

Typically, the source zone of the hydride system is operated at

a higher temperature than that of the chloride system in order to aug-

ment the decomposition of VH3. Due to the strong influence of temper-

ature on the species present, the same temperatures were used in both

systems in order to provide direct comparison between the chloride and

hydride CVD systems.



~--~-- F-F- F

O)C)mm) mmm
i- F- CC\J C\ C\J C\ CJ
I iC) C) C) C) C) C)

0 0 0n o
F- co C)


F- F- F- F-D F

0 000
0 000
0 O AAI o

Cn o O

N. *. CM CMC 0C
.0 ......
C) C r- F- F- C
I I I I I I IC'-*C'-
O- Ln ko 0o 'n Li

C) *Co)C oC)
C)IC) C:)

3 033 3333
r- Ln*-*r- -r ) -r r --*r


- 000000000000




rNN r


C\ C\J




00 00
o o



0 0




CO w- M CO
m m m m m m M Ln

CTjm 00 C0t C 0 1 1

co co 0n

I CM CM r.

cO '

- CM CM rN
I C) C )



C-*C-* ILi)


CD*C) m

c'. I I I

C c;c)

0 CO) C

C) CO coCM (mC)

NC I Ic1


C0 ) 1 01 CM

i- 0C C0-

O C)0N




0-0 o 0 0 0 0 o 0U0
*r- i*r- r- r- *r- ( r- C rL- l-

CL-W-HW aaa Qc
-4- --4- 4-4- --4-


o o LO c0 Co\
, b- I- dZ : -- zj-

0 i- 0 0 0

co co co ,- Co
r- C COj I- C

-- r- r-

coo m o co

co rm-- c
0- 0
C4 N" C\j c'* C\J
JC)o Co Co

Lf O
,- LOc O

I cc

--O r--- O N,
0 00

LO I-r-

i- CD t0-:- C)

Co Co CM rN- LA

. L *
,- Co CM CM CM


oo ( ro ro


co o m co
Cr CMj C- r- C-
mT 01 mT a

CD- i

C0- I C-* I


r- a C\-


I *


*C LO. r- C
NM3 r- O

CM 0

0. Q. Q.. 0.. Q

Solid State Defect Chemistry

The model developed for the deposition zone specified that reactant

depletion was not a significant phenomenon. During steady state growth,

the compound film is exposed to a vapor phase that is invariant with

respect to composition. It follows directly that the vapor phase fu-

gacity of silicon must also be constant and therefore, at equilibrium,

the activity of solid silicon in the epitaxial layer must be constant.

Thus, prediction of the amount of silicon deposited in an epitaxial

layer may be accomplished by determining the solid state silicon con-

centration as functions of the fixed vapor composition and temperature.

A model which relates the point defect structure of the solid to the

solid state silicon concentration must therefore be developed and cou-

pled to the calculated vapor phase composition.

Hurle [73-76] has proposed models for native defects in GaAs and

for Te, Sn and Ge doped GaAs. These models are extended here to in-

clude the formation of antisite defects and to account for the doping

of GaAs by Si. The native point defect model allows for the formation

of Frenkel and Schottky disorder on both the group III and group V sub-

lattices. Silicon is allowed to reside either on group III or group V

lattice sites or as an interstitial. Furthermore, Si substituted on a

group III site is allowed to form a complex with a group III vacancy

or a Si atom on a group V site.

A set of independent formation reactions for neutral species is

shown below.

1/2 As2(g) + Vi = Asi 3-6

1/2 As2(g) + VGa = AGa 3-7

AsAs + Vi = As + VAs 3-8

Ga Ga+ V = Gai + VGa 3-9

GaGa + VAs = GaAs + VGa 3-10

0 = VGa + VAs 3-11

Si(g) + VGa = SiGa 3-12

Si(g) + VAs = SiAs 3-13

Si(g) + Vi = Si. 3-14

The notation used is consistent with that of Hurle [76]. As an

example of this notation, equation 3-6 combines arsenic dimer in the

vapor with a vacancy on an interstitial site to yield an arsenic atom

on an interstitial site. Thus, in the solid state, the subscripts de-

note whether an atom (or vacancy) is residing on the Ga or As sublattice

or in an interstitial location. Equation 3-6 couples the point defect

structure to the vapor phase via the formation of As interstitials.

Frenkel defects are accounted for through equations 3-8 and 3-9.

Schottky defect formation is given by equation 3-11 and the formation

of antisite defects is represented by equations 3-7 and 3-10. Silicon

incorporation is shown in equations 3-12, 3-13 and 3-14.

Consistent with Hurle's model [76], the interstitial species are

assumed to remain electrically neutral while the other defects may

ionize. The following set of formation reactions may be written to re-

present the ionized species.

As = Asa + e- 3-15
Ga Ga
V = V + e- 3-16
As As
Ga = Ga + h+ 3-17
As As
V = Va + h+ 3-18
Ga Ga
Si = Si + e- 3-19
Ga Ga
Si = Si- + h+ 3-20
As As

Si + V =Si V + 3-21
Ga Ga GaGa+ h+ 3-21
Sia+ SiA = SiGSi 3-22
Ga As GaSiAs
0 = e- + h+ 3-23

The ionization of native point defects is illustrated by equations

3-15 through 3-18 while the amphoteric behavior of silicon is represented

by equations 3-19 and 3-20. Equations 3-21 and 3-22 represent silicon

complex formation and equation 3-23 accounts for the generation and re-

combination of electrons and holes.

All that is needed to complete the model is to couple the arsenic

partial pressure to that of gallium through the sublimation reaction

GaAs(c) = Ga(g) + 1/2 As2 3-24

and to write the electroneutrality condition
[h1 + [Si+] + [VA] + [AsG] = [e-] + [Sis] + [Va +
Ga As Ga As Ga

[GaAs] + [SiGaVGa] 3-25
where [ ] denotes concentration.

Expressions for the equilibrium constants for equations 3-6 through

3-24 may be written in the usual manner. As examples, the equilibrium

constants for equations 3-6 and 3-7 are

K = Y [Asil 1/2 3-26
6 Ai As2

K7= As [AGa] [VGa -1 -1/2 3-27
Ga a Ga As2

where yj represents the activity coefficient for species j in the solid


The model represented by equations 3-6 through 3-25 may be simpli-

fied by considering the electroneutrality condition. Hurle [76] has

shown that, at the temperature of growth, the Frenkel defect on the ar-

senic sublattice dominates the electron concentration. Thus, the

electroneutrality condition becomes

[e-] = [V ] 3-28

and the electron concentration at the growth temperature is given by the

relation 1/2
K4KI2 -1/4
[e] = 2 As 3-29
[] 2YAs+Ye- As2

Assuming that the defect structure of the epitaxial layer is

"frozen-in" at the growth temperature, the electron concentration given

by equation 3-29 is used to determine the concentration of each defect

present. At room temperature, the compensation ratio, which is defined

as the ratio of donors to acceptors (ND/NA), is dominated by the ionized

silicon impurities. Using the equilibrium constant expressions for

equations 3-6 through 3-24 along with equation 3-29 yields

N .+
D [SGa] = 1 3-30
A [SiA] + [SiGVG aP-3/2 b
As Ga Ga aPAS +b
where a = K13K16Ye- YSi
K 6K11K12K19K23YSi YV+
As As

b = K11K16K18K21Ye-YSiLa
K23YSi V +
i Ga Ga As

The dependence of the compensation ratio upon As2 partial pressure

for VPE grown GaAs has been found to be very small (ND/NA=3). Thus,

consistent with Hurle's results for Sn and Ge doped material, equation

3-30 indicates that the dominant acceptor in VPE GaAs is the SiGaVGa


An expression for the total silicon present in the epitaxial layer

is given by


[SiTotal [SiGa] + [Sia] + [Si.] + [SiAs] + [Si5s]

+ [SiGaV] + 2[Si Si] 3-31

= aP-3/2 + 5/4 -1/4 3/4
[Si]Tota [(aP~+ dPs + eP + b + 1)P +
TAs2 As2 As2 As2
2 2 2 2

f + gai] aSi 3-32

where: d = (K /2KI3K 1/2K + Ye)/(/2KK3K9
8 1316 K24YSi aYe-)/(K/26KIK13KI9VAs SAs
Ga As As

e = ( 2K16Si Y/2 )/(K/2K Si
Ga As Ga

f = (K3/2K 4K 1 S/2 1)/(K 3/2K1 K2K KK YV
S 14 16 (aYe- 6 11 12K19YSiiYV )
Ga i As
g = (K3/2KK 1/2 K I/2)/( 3/2 K Y +
g = (K 32K16K20K222YSi e- 2)/(K 2K24Yi Si As 'VA
Ga Ga As As

aSi = activity of silicon in the solid phase.

Currently, the thermodynamic data necessary for the evaluation of
the equilibrium constants and activity coefficients is not available.

Therefore, a quantitative application of equation 3-32 is not possible.
It can be seen, however, that the incorporation of Si into an epitaxial

layer increases with increasing aSi and depends, in a complicated way,
upon As2 partial pressure.
This model was developed for GaAs in order to keep the notation
manageable. An analogous model can, of course, be constructed for InP,
which would yield identical dependencies upon aSi and the group V dimer
partial pressure.



Summarized in this chapter are the thermochemical properties

used for the complex chemical equilibrium analysis. The proper se-

lection of a consistent data set is of extreme importance as a small

error in a property value can greatly influence the eventual calcu-

lated equilibrium composition. That this sensitivity can be important

is nicely illustrated in the Si-C1-H subsystem as discussed later.

Essentially, what is required for these calculations is a means of

specifying the partial molar Gibbs energy of each species believed to

be present as a function of temperature, pressure and composition.

Approximately 150 species were initially examined while only those

species that would be present at a mole fraction > 1014 were included

in the analysis. The scheme of representing the data was to fix the

zero enthalpy scale at 298 K and 1 atm with the pure components (stan-

dard states) Ga(c), As(c), In(c), white P(c), H2(g), Si(c), C12(g) and

02(g). The enthalpy of formation of the remaining components at the

standard conditions from the reference components was then determined.

In addition, the absolute entropy at the standard conditions for each

species was selected which allowed a calculation of the standard Gibbs

energy change for all possible reactions at 298 K and 1 atm. Finally,

knowledge of the constant pressure heat capacity and assuming ideal gas

behavior allows the Gibbs energy to be determined at any temperature and

pressure. The gas phase was assumed to be a solution of ideal gases due

to the low pressure and high temperatures investigated. For the con-

densed solutions, the pressure dependence of the thermochemical pro-

perties was neglected. However, the moderate negative deviations from

ideal behavior in composition dependence for the liquid source mixtures

was accounted for and represents one of the significant refinements

contained in these calculations.

Three pieces of information were required for each species with

the standard enthalpy of formation being the most critical, particularly

at the lower temperatures. It is for this quantity that the most un-

certainty exists in the reported value. On the other hand, the standard

entropy is generally quite accurately known, either from low temperature

heat capacity measurements or spectroscopic studies. The high tempera-

ture heat capacities were sometimes estimated, but there exists a par-

tial cancellation of its effect when calculating Gibbs energy changes.

Presented below is a discussion of the properties selected. It is noted

that in many instances the thermochemical data presented in the JANAF

tables [77-79] were used and are discussed in these tables, therefore

precluding a discussion here.

Psuedo-Steady State Constraint for Liquid Source Boat

It has been observed that during VPE of GaAs and InP using a liquid

source boat of pure group III metal in the chloride process, an initial

transient period exists in which the composition of the gas stream

leaving the source region is a function of time. Initially, the pure

metal boat is dissolving group V atoms thus producing an excess of group

III chloride. As the metal becomes saturated with the group V element,

a thin crust of the compound is formed at the top surface since the

density of the compound is less than that of the saturated liquid. It

is observed that the crust thickness reaches a steady state value and

therefore, from a simple mass balance, the vapor phase will contain

all of the group V atoms in the input stream plus the amount of group

V atoms generated from the saturated liquid (due to reaction of chlorine

with the group III atom). The exact amount of group V element pro-

duced from the source boat is therefore a function of the temperature

(i.e. the equilibrium group V mole fraction in the liquid is a function

of temperature and the amount of group III atoms leaving depends on the

form in which they leave (e.g. IIIC1, IIIC13, III, etc.)) and the flow

rate (i.e. mass transfer efficiency). The mechanism by which the group

III and V atoms reach the gas/solid interface is not known but is not

required for the thermodynamic model presented here as mass transfer

barriers (e.g. the crust) are assumed not to be present. All that is

required is to assume a new species exists having a stoichiometry

equivalent to the saturated liquid composition.

The thermodynamic properties of the hypothetical liquid species,

AlxBx(1), can be estimated in the following manner. Letting A repre-

sent the group III atom and B the group V atom, consider the reaction


(l-x) A(c) = (l-x) A(1) 4-1

x B(c) = x B(1) 4-2

(l-x) A(1) + xB(1) = AlxBx(1) 4-3

all occurring at the source temperature, T. Since A(c) and B(c) are in

the pure state, the Gibbs energy changes for reactions 4-1 and 4-2 are

the Gibbs energies of formation for the liquid species and can be cal-

culated from the thermodynamic sequence: solid element A or B is taken
from T to its melting temperature T B
from T to its melting temperature Tm or TM, the solid element is melted,

the liquid element is taken from the melting temperature to the ori-

ginal temperature of interest. Approximating the heat capacity

difference between the pure liquid and pure solid, AC as a constant,

the Gibbs energy change for reactions 4-1 and 4-2 are

AG1 = (1-x){AHA (1 T ) + ACA (T TA T In [E]T } 4-4
m m


AG2 =x AH (1 ) + ACB (T TB T n [ 4-5
m m

The Gibbs energy change of the third reaction is simply the Gibbs energy

of mixing. Assuming that a simple solution model describes the liquid

behavior results in

AG3 = (a + bT) x (l-x) + RT[xlnx + (l-x) In(l-x)] 4-6

where a and b are adjustable parameters determined in conjunction with

solid-liquid equilibrium data.

The sum of reactions 4-1 to 4-3 is the desired formation reaction

(l-x)A(c) + x B(c) = A1-x x(l) 4-7

while the corresponding Gibbs energy of formation of A lxBx(1) is the

sum of AG1 to AG3. Given the source temperature, T, the procedure is

to first calculate the liquidus composition, x, from the implicit


T HAB R In[4x(1-x)] + b[ x2 (1-x)2] HAB
TAB m m

a[ x2 (-x)2] 4-8

where AHAB and TAB are the enthalpy of fusion and melting temperature of
m m
the solid compound AB and R is the gas constant. Once the equilibrium

group V composition is determined the standard Gibbs energy of formation

Table 4-1

Thermochemical Properties of GaAs and InP Required

for Calculating AG[A xB(1), T]

Property GaAs Reference InP Reference

TAB (K) 1511 72 1332 71,72,74
AHA (kcal/mole) 25.14 72 14.4 71,72,73
a (cal/mole) 4666 see text 5055 75

b (cal/mole K) -8.74 see text -5.0 75

Table 4-2

Thermochemical Properties of the Elements Ga, In, As and P

Required for Calculating AG [A _B x(1), T]

Property Ga Ref. In Ref. As Ref. P Ref.

T (K) 302.9 84 429.76 84 1090 17 313.3 29

AHm(kcal/mole) 1.336 84 0.78 84 5.123 17 0.157 29

AC (cal/mole K) 0.27 84 -0.2 84 1.0 estimated 0.472 84

of Al-xBx(1) can be calculated from equations 4-4 to 4-6 given T, x

and the required thermochemical properties. Tables 4-1 and 4-2 summa-

rize the thermochemical properties used in these calculations. The

adjustable parameters a and b for GaAs were determined by reduction of

the liquidus measurements of Clariou et al. [80], Hall [81], Muszynski

and Riabeev [82] and Osamura et al. [83] using a maximum likelihood


The Ga-As-Cl-H System

The enthalpy of vaporization of As(c) has been investigated quite

extensively with a reported range [84-99] at 298 K of 34.4 to 38.54

kcal/mole corresponding to the standard enthalpy of formation for

As4(g). The literature has been summarized by Hultgren et al. [84]

to 1973, while a more recent measurement of Rusin et al. [85] by a

static method produced a value of H(AAs4, g, 298 K) = 38.14 kcal/mole.

In addition, Rau [100] has measured the total vapor pressure over solid

and liquid arsenic from 850 to 1400 K with a Bourdon gauge. Analysis

of the low temperature results indicated H (As4, g, 298 K) = 37.34 +

0.2 kcal/mole. The value selected was 37.5 kcal/mole on the basis

that the static methods are believed to be more reliable.

The dissociation enthalpy, H (As4 = 2As2, 298 K), has received

considerable investigation with early mass spectrometric studies pro-

ducing values in the range of 61.5 to 73.5 kcal/mole [101-105]. These

measurements are suspected of overestimating the As4 partial pressure

as a result of the use of a low condensation coefficient for As4. More

recent determinations using MoAs2, Mo2As3, GaAs, InAs and InAs + InSb

sources with improved Knudsen-cell mass-spectrometer designs [106-108]

and reduction of PVT measurements [100] have indicated much lower values


for the dissociation enthalpy (54.21 1.5, 54.26 + 1.1, 54.2 + 1.4 and

54.8 + 1.0 kcal/mole, respectively). The value selected here is the

average of these four measurements, 54.4 + 1.5 kcal/mole. Using the

selected values of the standard enthalpies, AHo(GaAs, c, 298 K) = -19.52

kcal/mole, AHo(As4, g, 298 K) = 37.5 kcal/mole and the above dissociation

enthalpy yields AHo(GaAs(c) = Ga(c) + 1As2(g), 298 K) = 42.5 kcal/mole.

This can be compared to the experimentally determined values of 44.85

[106], 45.06 [107] and 45.4 [109] kcal/mole.

The value adopted for AHo(GaAs, c, 298 K) is -19.52 kcal/mole as

determined by Martosudirdjo and Pratt [110] with a precipitation calori-

metric technique. This value can be compared to the emf work of Abbasov

[111] and Sirota and Yushevich [112] in which values of -19.4 and -20.96

kcal/mole were reported, respectively. These latter two values are expected

to include uncertainties due to the assumed valency of Ga in the gal-

vanic cell and the inability to accurately determine the full tempera-

ture dependence over the relatively small temperature range of measure-

ment. In addition, a considerable number of dissociation pressure

stuides [102-107, 113-118] and flow equilibration investigations with

reactive gases [109, 119] have been performed which contain information

about solid GaAs. However, knowledge of the thermodynamic properites of

other species is required (i.e. As2(g), As4(q), GaCl(g), etc.) to speci-

fy the properites of GaAs(c). Thus, the uncertainty in these'proper-

ties must be considered in addition to those associated with the measure-

ments. This work was, however, used as a test of internal consistency

in the total data set. The standard entropy of GaAs, So(GaAs,c,298 K)

was taken from the low temperature heat capacity measurements of Pies-

bergen [120] while the high temperature heat capacity was taken from the


measurements of Lichter and Sommelet [121] and are in good agreement

with the work of Cox and Pool [122] and the estimates of Marina et al.

[123] and Maslov and Maslov [124].

Very little experimental information is available for the arsenic

chlorides. The reported range for the enthalpy of formation of AsC13(g)

is -52 to -72 kcal/mole [97, 99, 125-130]. The value adopted was

AH (AsC13, g, 298 K) = -62.7 kcal/mole, taken from the enthalpy of

formation of the liquid and the enthalpy of vaporization. The enthalpy

of formation of the mono and dichlorides were taken from the estimates

of Shaulov and Mosin [127] as was the standard entropy and heat capacity.

The enthalpy of formation of arsine was taken as AH (AsH3, g, 298 K) =

16.0 + 1.5 kcal/mole based on the work of Gunn [131] and reported tabu-

lations [97-99]. Finally, the thermochemical properties of the remain-

ing arsenic hydrides were estimated by comparison with the hydrides of

N, P and Sb [77-79].

The thermodynamic information available for the chlorides of gallium

is somewhat scarce and inconsistent. The enthalpy of formation for

GaCl3(g) was determined from AH'(GaC13, c, 298 K) = -125.0 kcal/mole

[132] and the enthalpy of sublimation taken from the vapor pressure

measurements of Kuniya et al. [133], AH (GaC13, g, 298 K) = -107.3 + 3

kcal/mole. The enthalpy of formation of gallium monochloride has the

reported value AH0(GaC1, g, 298 K) = -19.5 kcal/mole and is taken from

the dissociation energy of Barrow [134]. However, a value of AH (GaC1,

g, 298 K) = -12.0 kcal/mole is obtained using AH](GaAs(c) + HC1(g) =

GaC1(g) + As(g) + H2(), 950 K) = 37.52 + 8 kcal/mole determined

by Battat et al. [119] and the thermochemical data selected for the other

species. These results are in sharp contrast to the vapor pressure

measurements of Kuniya et al. [133] who report a second law AH (GaC13(g)

= GaC1(g) + Cl2(g), 1083 K) = 45.912 kcal/mole. The value selected was

AH (GaC1, g, 298 K) = -17.0 + 5 kcal/mole based on the first two re-

ports, considering the value of Barrow [134] slightly more due to the

uncertainties found in the enthalpy of formation for the other species

in the reaction studied by Battat et al. [119]. The enthalpy of forma-

tion for gallium dichloride was taken from the measurements of Battat

et al. [119] using the thermochemical data selected here and correcting

the second law entropy to that calculated by Shaulov and Mosin [135].

The enthalpy of dimerization has been investigated by several authors

[133, 136-139] with the reported enthalpy and entropy of dimerization

in the range, AH0(2GaCl3(g) = Ga2C16(g), 298 K) = -22.6 to -20.9 kcal/

mole and So(2GaCl3(g) = Ga2Clg(g), 298 K) = -31.66 to -36.0 cal/mole-K.

Accepting the enthalpy and entropy of dimerization as -21.0 kcal/mole

and -33.0 cal/mole-K, respectively, and combining these results with

the selected thermochemical data for GaC13 produces AH (Ga2C16, g, 298

K) = -235.6 + 10 kcal/mole and S(Ga2C16, g, 298 K) = 127 + 6.0 cal/

mole-K. The standard entropy and heat capacity for GaC1, GaC12 and

GaC13 were taken from Shaulov and Mosin [135] while the heat capacity

of Ga2C16 was approximated by the value for A12C16 [77]. In addition,

other species are expected to exist (i.e. Ga2C14, Ga2C12) [140, 141],

but no thermochemical data is available.
The In-P-C1-H System

The standard entropy at 298 K and the constant pressure heat capac-

ity of solid and vapor In were taken from Hultgren et al. [84]. As sum-

marized by Hultgren et al. [84], the standard enthalpy of vaporization

of solid In at 298 K that results from application of the third law


to the vapor pressure measurements produces the range of 49.8 to 58.1

kcal/mole for AH (In, g, 298 K). More recent mass spectrometric re-

sults of Panish and Arthur [132] and Farrow [142] suggest the values

of 56.6 and 58.03 kcal/mole, respectively, with the average of these

two values adopted here. In a similar fashion, the thermochemical

properties of phosphorous selected by Hultgren et al. [84] or the

JANAF tables [77] are in agreement with the more recent results [132]

and were adopted for this study. However, there exists a small dif-

ference in the reported AH0(298 K) of the reaction: P4(g) = 2P2(g).

Foxon et al. [143] report a value of AH(298 K) = 57.9 + 1 kcal/mole

while Panish and Arthur [132] reported AH0(298 K) = 53.8 + 0.8 kcal/

mole from third law calculations of their mass spectrometric results.

The third law reduction of the mass spectrometric results of Farrow

[142] produces a value of AH(298 K) = 58.04 + 0.3 kcal/mole, while

the JANAF tables [77] suggest AH0(298 K) = 54.6 + 1.1 kcal/mole. An

average value was selected of AH(298 K) = 56.1 + 2.0 kcal/mole.

A rather wide range in the reported values for the standard en-

thalpy of formation of solid InP exists (-13.52 to -22.3 kcal/mole).

As shown in Table 4-3, the value selected was the average of the two

solution calorimetric determinations as this is a direct determination

of the property. The results from the vapor pressure measurements are

subject to uncertainties in the properties of the vapor phase species

and also the heat capacity of solid InP (e.g. Panish and Arthur [132]

used C for A1Sb which produces a decrease in AH (InP, c, 298 K) of

0.5 kcal/mole, when compared to C (InP, c, T) of Pankratz [144]). The

standard entropy of InP(c) was taken from the low temperature heat

capacity measurements of Piesbergen [120] while the heat capacity

Table 4-3

The Reported Standard Enthalpy of Formation of

InP(c), AH (InP, c, 298 K)




-18.83 + 0.7(a)
-18.58 + 0.7
-22.3 + 1.5 (
-19.33 + 0.1
-17.83 + 1.4(d)
-13.52 + 0.26
-21.0 + 2
-21.5 + 1.5
-14.5 + 0.44

(a) InP(c) = In(c)

(b) InP(c) = In(c)

(c) InP(c) = In(c)

(d) InP(c) = In(c)

(e) InP(c) = In(c)


flow equilibration
mass spectrometry
mass spectrometry
vapor pressure
mass spectrometry
solution calorimetry
bomb calorimetry
bomb calorimetry
solution calorimetry

109, 145
144, 148
referenced in

p (g)

+ P2 (), Hg98 = 360 kcal/mole
taken as 34.34 kcal/mole

+ P4(g), Hg98 = 22.1 kcal/mole
taken as 14.08 kcal/mole

+ P2g Hg98 = 36.5 kcal/mole
taken as 34.34 kcal/mole

+ 2 P2(g), H98 = 35.0 kcal/mole
taken as 34.34 kcal/mole
+ P2(g), H298 = 37.50 + 0.1 kcal/mole
taken as 34.34 kcal/mole

adopted was that measured by Pankratz [144]. The result is in good

agreement with the 298 K value of Piesbergen [120] and in fair agree-

ment with the high temperature measurements ov Cox and Pool [122]

and the suggested value of Maslov and Maslov [124].

Barrow [134] reports the dissociation energy of InCl to be 102.5

kcal/mole and combining this with the value of the enthalpy of forma-

tion of In(g) and Cl(g) yields AH (InC1, g, 298 K) = -16.21 kcal/mole.

However, the atomic fluorescence value for the dissociation energy

also reported by Barrow [134] (D = 104.6 kcal/mole) yields AH (InC1,

g, 298 K) = -18.31 kcal/mole. Klemm and Brautigan [149] reported

AH (InC1, c, 273 K) = -44.6 kcal/mole and when combined with the

enthalpy of sublimation, AHs(InC1, c, 298 K) = 27.8 kcal/mole [150]

gives AH(InC1, g, 298 K) = -16.8 kcal/mole and is the value adopted

here. The standard entropy of InCl was taken from the calculations

of Malakova and Pashinkin [151] while the heat capacity is that

recommended by Kelly [152]. The standard enthalpy of formation and

entropy of InC12 was taken from the estimate of Glassner [153] and the

heat capacity is the same as that listed for GaCl2. The standard en-

thalpy of formation and entropy of GaC1, was taken to be the values

suggested by Mullin and Hurle [97] and the constant pressure heat ca-

pacity estimated by Shaw [98]. The thermochemical properties of the

dimer, In2Cl6, were taken from the values suggested by Schafer and

Binnewies [139].

The standard enthalpy of formation of phosphine was taken from

the decomposition studies of Gunn and Green [154] and the remaining

properties from the JANAF tables [77, 79]. The JANAF tables were also

used for the other phosphorous hydrides, chlorides and oxide vapor

phase species.


The Si-C1-H System

The thermochemical properties of Si have been reviewed by Hultgren

et al. [84] and the JANAF tables [77]. In particular, there exists a

rather large range in the reported third law values of the standard en-

thalpy of vaporization, AH (Si, g, 298 K) = 86.75 to 109.06 kcal/mole.

The value selected was in between the Knudsen studies of Davis et al.

[155] and Grieveson and Alcock [156].

The Si-C1-H system has received considerable attention due to its

importance in the semiconductor industry. An excellent review of the

literature for this system with equilibrium calculations presented is

given by Hunt and Sirtl [157] and Sirtl et al. [158]. The posture taken

here is to assume that SiC14 has the most reliable thermodynamic data

with these values being fixed by the JANAF tables [77]. The reaction

Si(g) + SiCl4(g) = 2SiC12(g) 4-9

has been investigated extensively [159-164]. Employing the thermo-

chemical data for the three species in reaction 4-1 from the JANAF

tables [77], third law values of AH (SiC12, g, 298 K) were calculated

from the experimental data. The effusion-mass spectrometric determin-

ation of Farber and Srivastava [164] over the temperature range of

1593 K to 1792 K produced the value, AH (SiC12, g, 298 K) = -40.39 +

0.3 kcal/mole and showed no temperature dependence. This result is

in good agreement with flow equilibration data of Schafer et al. [159]

(1273 K to 1473 K), Teichmann and Wolf [160] (1223 K to 1575 K) and the

static measurements of Schafer and Nicki [163] (1398 K to 1573 K), with

third law values of -40.62, -40.54 and -40.44 kcal/mole, respectively.

The flow equilibrium values of Antipin and Sergeev [161] (1273 K to

1673 K) and the static values of Ishino et al. [162] (1448 K to 1573 K)

were more negative and exhibited a marked temperature dependence. On

the basis of these calculations, the value AHO(SiC12, g, 298 K) =

-40.4 kcal/mole was selected. The values for the standard enthalpies

of formation of the less stable chlorine SiC1 and SiC13 were computed

from the high temperature flow equilibrium studies of Farber and Srivas-

tava [164]. In the third law analysis, the data previously discussed

was used in conjunction with the heat capacity for SiC1 and SiC13,

suggested by the JANAF tables [77] and produced the value of 47.4 + 0.6

and -93.3 + 0.5 kcal/mole, respectively. These results are in agreement

with the analysis of Rusin et al. [165-167] on the total pressure meas-

urements of Schafer and Nicki [163]. No additional thermodynamic

studies of Si2C16 are known to exist and thus the properties suggested

by Hunt and Sirtl [157] were adopted.

It was pointed out by Hunt and Sirtl [157] that the mole ratio of

SiC14 to SiHC13 is very sensitive to the assumed value of the standard

enthalpy of formation of SiHC13. Indeed, this ratio is seen to vary

by nearly four orders of magnitude at 1000 K when bounded by the ex-

perimental determinations of AH (SiHC13, g, 298 K). Since the work of

Hunt and Sirtl [157] was published, two additional experimental in-

vestigations of the thermodynamic properties of SiHCl3 have been per-

formed. Farber and Srivastava [168], from effusion-mass spectrometric

measurements, determined the reaction enthalpy for

SiCl4(g) + H2(g) = SiHCl3(g) + HC1(g) 4-10

over the temperature range 1155 K to 1500 K. Employing the thermo-

dynamic data listed in Table 4-4 and these results, a relative temper-

ature insensitive third law value for AH (SiHCl3, g, 298 K) = -119.30

+ 1.0 kcal/mole is obtained. Using both static and dynamic methods,


Wolf and Teichmann [169] investigated reaction 4-10 and the reaction
4SiHC13(g) = 3SiCl4(g) + Si(c) + 2H2(g) 4-11

Third law values for AH (SiHC13, g, 298 K) were calculated from the

original results of these authors. The values obtained for reaction

4-11 and for the three data sets with reaction 4-10 were -119.47 +

0.4, -119.83 + 0.9, -119.58 + 0.2 and -119.50 + 0.6 kcal/mole and the

results are seen to be in good agreement with the measurements of

Farber and Srivastava [168]. Since these values were nearly 3 kcal/mole

more negative than those developed by Hunt and Sirtl [157], values of

AH (SiHC13, g, 298 K) were calculated for various experimental SiCl4/

SiHC13 ratios in a fashion similar to Hunt and Sirtl. The experimental

data consisted of a variety of feed mixtures (e.g. SiC14/H2, H2/HC1,

SiHC13/H2) which were contacted with Si(c) at different temperatures

during a deposition/etching process. The results of these calculations

for 14 data sets suggested AHo(SiHC13, g, 298 K) = -118.16 + 1.70 kcal/

mole. Based on these results and the new experimental determinations,

the value adopted was AHo(SiHC13, g, 298 K) = -119.5 + 1.5 kcal/mole.

The standard enthalpy of formation of the di- and mono-chloro-

silanes was taken from the recent measurements of Farber and Srivastava

[168]. In order to obtain a consistent data set, third law values of

these quantities were calculated from the original experimental data

while using the data base adopted here. The adopted values were

MH (SiH2CI2, g, 298 K) = -75.5 + 2 kcal/mole and AHo(SiH3C1, g, 298 K)

= -32.7 + 2.5 kcal/mole.

No additional experimental information on the thermochemistry of

SiH4 and SiH exists and thus the JANAF tables recommendation was adopted.

The standard enthalpy of formation of disilane was taken from the


calculations of Potzinger et al. [170], AH (Si2H6, g, 298 K) = 17.1 +

3 kcal/mole and is compared with the calculations of O'Neal and Ring

[171], AH (Si2H6, g, 298 K) = 19.1 kcal/mole and the value of 16.0 kcal/

mole obtained from the estimated enthalpy of formation for SiH3 (35

kcal/mole) and Si-Si (-54 kcal/mole). The standard entropy and heat

capacity of Si2H6 were obtained by comparison with C2H6.

A summary of the selected thermochemical properties (with two

additional references [172, 173]) is presented in Table 4-4. In or-

der that the stability of the various species might easily be compared,

the standard molar Gibbs energies of formation are plotted in Figures

4-1 through 4-5.

.,- -- K- r- r- r- r- Chj coi r <- '- Lfio Lo rh- a) r-, r-
cOc 00 CO CO \ cj r- i r---r cocormm 0 N NrN N-

~--- -- f ~-- r- r- r- m r r -- .


r- I C




t ...


LO -zj
--l co


00 L)
o o

! I





co r-

o o


-- Ln 0o

It C0
.l .

r~- co

in Co

I C- I I

I ',D D

0) 00


I *r 0

i CM a 114

C- o O

IN 01
I -- NC
I N r-


*- On

:r:I- -- 0- 1- r- r- rN-
0cococo OCo co Nj C\iJ Nj


O %.0 In

CO r- 1-

I CO C- O -N

S .0

+ i+
0 In



C0 0-

usk r

4Cm N

Ar- ,r-- -
cO C)
O r-~
CO 0

0 o
mC 1 +1
+1 L +
+1 :r

C 01 ')
In m 1n

00 00

00 0 00
+1+1 +1+1
+1+1 +1
co In + r-
Ln N- C0 N 0
N-1'Co 0) 01) C;

oIor-c'.i + 1+ I -+-
-z3- in r~- r-~ CMi CMj n

rN r-- r- CM iaj .-- 0 CM C
CM N C O r- I I CO C- m I I m
r- l- -ic r--- 1 I r- r- 4 r- I l-


+C l

,- Ci


n L-






+o+i+ i
+ 1+ 1+I





m o


-..---j -f-- e
-0C-) L. L. ) Mi -
^- C-- *- n c* -- -'
^-'.-- 0) 0) Nj C') 0 cm .-a- --- -- C) -to 0
C--- en i-0 ) m m C) 0-)-' U Cm-- -N mm -- ^ m- 0
U an _- v "- -- c-j m cnr o en.-- o < ^- --j cm ----
t N) Jn n n C n ln) l In V) :c


0 E


co -
O 0



Y a





f- -a- :Zr CIM L mn co r-, r- i' ~r- r r" .z' r r r a rN- r-N co

r- c 0 i- O M M r- r i r- r- i- C N r- r- i-- r- --

r --r.- -- -C CJ -i -

CM )
0 I~ I I I

I- II C) i I
- I I CO
C0 0


Lz i ID I-

C ID I -

,CM o I Lm
O '- n o 00 C0

N 0o c Dm n
0 aC ac
C+ + -I- -- M r-
N- CO CO in In o

.D r- + +
n- -
C) cM CD

0-- Ln q CO0

CD h


CO 1^-
tD 0

- 0C-

m o ,--



01 CMO
+ rs r-

+ I -- In

CO 0

CM0 0C

r- -z-C

I I I I CM r-O
I I I I I 0
C) C

I I 0

D tl CO
N-o a o

o r~


ID .0
0 >






1 1 X
I i C

II ro
I *- C

O M m



I 0 I I I
I 1- I I I
CM 0


m O CO C
0 I I

r- LD C 0
on o n On i
I -- c c--


N- CM N- co an CM

I ID an i-

r-r : r-- r- CO r- r- r- co CM ~N- r- CO
r- co r N- r N C- rC r- rN r- r- -co r-

Ln m

+1 +0 0
+N OcaOn
a 000- a- 0 a 0 00n
c*4** *0 0 00 C 0 C O C .
000 0 L0 a -- 0 --Lni

+1+1+1 +1 +1 +| +i +|+|
+1 +1 +|+[ oo m
cM co- 01 r-- <3 0- cCMCM
CO 03-- CO L03 ID 01CO CM CMCOl- CO 0
O1COM Uo D-y 00 o Do 0 0 *0 ors
SOLD LD LO [r. O L3D LD D^b3LD 13.0D

CM m rN CM r-- rN -~ 0a n r' 00 CO
- I *KL- O I I CO 0 CO ( LO I -L-3

CM 01 '*
0 000 0 C iD n
0 0 c 0 .- D
+|1 --- 0 +i CM co C 0 D00
r + 1+ O + + D -
Ln 0 + + 1|+ + C)O +1+0 + + + +1+ o- +i+
01 +i cM co r- + +i
Nr ma -O o)m %Din O t cj o + 1-t -
ID in
- rNDo CO O i Dz t- Cn L- U L D CO rh O
IO ID-- an o o r-0m O Co InCM t m- LD 0) -z
I I I I I I -


-M 01 c 0- 3 --- --3 3N
-- Co-- -- in -0-, cj 0 ~ c0--
01 -- U------ CM 0 i- -- -- --- 0o --- 0- --- 0- c --- CM
-5U C"- -- i- --- ----- -- i -- C U CD '-- --
C '----- L) L) CM 0c-- U 0-- c- CM 0) -C-)--)---) CJ =
=v -C c CM ------ CM < (_ Q a C a 0 *i- *C- *- -V
I l- l -- cC) l-0c 00 X .C. O. .0 Q. 0. it v.

0 CL E





n --

C1 E


0 E




S CO r r- h rh r-- I3LO C ( -

CM \ -

e- 1r~
r- r-


r- C-M


i [ )n co -a- ^o cc 0
i(n c0 CD i Z Co c0 1
I O0 C" LO M "O :- I

1 O 3 CO i- CM CO I
I 1.00OCLO

cocn i-n zr 0 in) Ln
OalCj m ;- rcu :T '
o-- o c

CNjr- Cj LC)n -d- C O C
-zl- '- c0 C) M r- CD
)- 0CM CO U L -- CM i-

mcu C c M CO I D ,n
* c e 1O 00 CML
U3 rI IO I -- cI

Sr- i i I M

C0 0C

L. I V -
0C I1- O
- cci-



I I 0



CM 0 CM 1.

r'- CO. rO- r. 0 CM C
r- --- 1 r

uin O O o in
Ln o o o o- -
O CM CM **C0 *
,- 0O -- 0 00
+1 +1 +C1 CM m 1.

10 + 1- + 1 +1 +1 +
r- r -C N m -" O-- 0-. 0
CD- OY C)r O CM n o 0 LOM C O O MN

Lo oi r 4 CO O Co 01 O z lC 1n Iz;
r- r- 3- <3 r- t.0 L 0 LO C C-

Cc CO (,. C,-. d CM C ,. O CC
co c- r- r-- -) 3 LO O r r--cM r- .-0 C- o --
u- r r c r r- -- -

.0 Cm oCoo) Co Ln CM *O

C C CMC ,- CM CM CO CM O C0 c m r

+1 +1 +1+1 +i +1 + +1+1 +1 +1 + +1
m <3 ON Ln U-) C. r m U3 3- 3 0O
. CM C *
o0 cO Oco oi in CMN I 0 I MN r-. o
0' 1.0 01 ,- ,-. M -- I r- r- r-

0n C- M- U N0 u
-- -- en o ,- i- -- u n to u

i- r- CM c -- CM CM L
C 1 IIOO CD N < r- *- *r- *r- *-- *- *r- *-- r- *- *r- *- *o C
LV) Ln V) V1 ) V) V) VU) 1 ) (V) CDu



+ +

T- (-

+ +

o 010


"--. --




0 E







co E
CM r-
04- U




20 1

0 -Ga(c)
In c


-20 -
2 GaAs(c)


- -60 -





873 973 1073 1173 1273

Temperature (K)

Figure 4-1
Gibbs Energies of Formation for Gallium and Indium Species
(at 100kPa Pressure)



-60 AsC

Figure 4-2



-60 AsCT3

873 973 1073 1173 1273

Temperature (K)

Figure 4-2
Gibbs Energies of Formation for Arsenic Species
(at 100kPa Pressure)








-60 PC13

873 973 1073 1173 1273

Temperature (K)

Figure 4-3
Gibbs Energies of Formation for Phosphorus Species
(at 100 kPa Pressure)






0 Si(c)

-40 SiO

0 Si0
e SiC13



-200 I I
873 973 1073 1173 1273

Temperature (K)

Figure 4-4
Gibbs Energies of Formation for Silicon Species
(at 100 kPa Pressure)

40 H




-20 I I-


873 973 1073 1173 1273

Temperature (K)

Figure 4-5
Gibbs Energies of Formation in the H-C1-O System
(at 100 kPa Pressure)



The decomposition of the trihydride of the group V elements N, P

and As has, in the past, been studied using manometric methods in

closed systems [50-55, 57-59] and also by infrared spectrometry in

open systems [56, 60, 61, 64-66]. A major disadvantage which is en-

countered with measurements based on manometric techniques is that the

system total pressure depends on all of the species present. It is

therefore difficult to remove the influence of other reactions in the

system from the observed data. Usually this difficulty is addressed

by assuming that a single reaction step is rate limiting and that the

remaining reaction products are at equilibrium. This technique has

been employed for the decomposition of PH3 and AsH3 by employing the

overall reaction [57]

4VH3 ---> V4 + 6H2 5-1

This reaction is not applicable to the NH3 system, however, since no

known tetramers of N exist. For NH3 decomposition, the following over-

all reaction has been applied [51]

2VH3 ---> V2 + 3H2 5-2

The existence of one or more slow reactions in the sequence between the

disappearance of VH3 and the formation of V2 or V4 can cause the ini-

tiating step of the reaction to appear to be slower than it actually is.

This results in an over estimation of the activation energy associated


with the reaction. A major advantage inherent in manometric methods is

that pressure can be measured to a very high degree of accuracy.

Spectrometric investigations provide a means for measuring, direct-

ly, the rate of appearance or disappearance of individual chemical

species in the system. Frequently, more than one signal can be moni-

tored during the course of the experiment and, therefore, the opportunity

for determining the entire kinetic sequence is greatly enhanced. The

sensitivity of spectrometer instruments varies depending on the type of

spectrometer and supporting equipment employed, but is is not unusual to

find mass spectrometers which have detection limits below 1 ppm.

A major disadvantage connected with quantitative composition meas-

urements based on spectrometric instruments is that of calibration. Most

spectrometers provide an output signal which is proportional to the

amount of the species which is present at the detector. The value of the

proportionality constant is rarely known and generally depends on the

specific chemical species and the energy (i.e. temperature) of that

species. Moreover, if a sample must be removed from the system for

analysis, a method of sampling must be chosen, such that the sample

composition accurately represents the system from which it was removed.

Also, the sampling technique must either not perturb the system signi-

ficantly, or it must perturb the system in a way which is known and

can be corrected for during the data reduction.

Even though there are many variables regarding the application of

spectrometric techniques for the measurement of composition, these

techniques are highly desirable because of the ability to follow sig-

nals representative of individual chemical species. Thus, the system

employed here for the investigation of NH3, PH3 and AsH3 decomposition

is based on a quadrupole mass spectrometer coupled to a constant

volume reaction tube through a sampling orifice. This technique

provided real time monitoring of the reaction gas phase composition

while perturbing the reacting system in a known manner which was

easily corrected for during data analysis.

Experimental Apparatus and Method

A schematic representation of the equipment used during the in-

vestigation of NH3, PH3 and AsH3 thermal decomposition is shown in

Figure 5-1. Due to the extremely toxic and flammable nature of the

gases involved, hooded enclosures were constructed around the storage

area for the gas cylinders and the reaction chamber. These enclosures,

along with the vacuum pump exhaust from the gas sampling system, were

vented through the laboratory exhaust hood. The exhaust gas from the

reaction rube was first passed through a Draeger class B3-P filter

before being vented through the laboratory exhaust hood. This was done

to remove any residual VH3 and its toxic reaction products from the

vented gases.

High pressure gas cylinders containing 4.3% NH3, 10.07% PH3 and

10.03% AsH3 in H2 were connected to a common stainless steel gas line

and solenoid operated valve for inlet to the reaction tube. Hydrogen

was provided as an additional inlet to the reaction tube through a

separate gas line and valve. This arrangement allowed the reactor to

be charged with any of the available gases and allowed the gas in the

reaction chamber to be diluted with H2 if desired. Purge gas consisting

of N2 or He was available through the AsH3, PH3 or H2 purge assemblies

and gas lines. A capacitance manometer was used to monitor the pressure

in the reaction tube over a range of 100 Pa to 105 Pa with a precision

of 0.1%.




S.- c

LL. a



The reaction tube consisted of a 54 cm long by 6 cm O.D. quartz

tube placed inside of a three zone Marshall furnace (model 1169-5) and

is shown in greater detail in Figure 5-2. A temperature profile

which was constant to within 2 K across the length of the reaction

tube was obtained by placing the tube inside of two Dynatherm liquid

sodium furnace liners and by controlling each of the electrically

heated furnace zones with individual Linberg Model 59344 heater con-

trollers. The latest temperature profiles were obtained when furnace

zones 2 and 3 were operated with identical setpoints and zone 1 was

operated at a setpoint 20 K below that of the other two zones.

Thermocouples were located at positions 12 cm, 25 cm, 50 cm and 62

cm into the furnace. These positions were chosen based on previous

measurements which demonstrated that the temperature profile from 16 cm

through 54 cm was flat to within 0.5 K. At positions less than 16 cm

into the furnace, the temperature drops off due to heat losses near

this end of the furnace. Positions from 54 cm to the end of the re-

action tube typically showed the highest temperatures in the furnace,

but this was compensated for by lowering the setpoint in zone 1. Fire-

brick insulation was placed at each end of the furnace in order to

minimize the heat losses. The system was limited to an operating range

of 673 K to 1273 K due to functional and safety constraints imposed by

the furnace liners.

The inlet tube to the reactor consisted of a 40 cm long, 6 mm o.d.

quartz tube which extended outside of the furnace and was mated to a

stainless steel tube through a stainless steel fitting using viton o-

rings as seals. Sampling of the gas in the reaction tube was accom-

plished by continuously drawing a sample through a small orifice

(n U

0 C 0



- *'r-


(C\J LL-





(nominally 0.1 mm in diameter) at the end of the reaction tube.

Further discussion of the gas sampling system and mass spectrometer

is provided in Appendix B.

The use of an orifice for obtaining continuous gas samples from

the reaction tube represents a significant perturbation on the re-

acting system and, therefore, must be considered in the data analysis.

As is described in Appendix B, the mass spectrometer provides an output

signal which is proportional to the partial pressure of the chemical

species present in the reaction chamber. The partial pressure of each

species changes due to participation in chemical reactions and the

continual bleed on the reactor caused by the gas sampling system. A

species balance on the reaction volume yields the following equation

S= ri V-


where: P. = pressure of species i
r. = chemical reaction rate for species i

n = rate of molar loss of species i through the sampling


V = system volume

Coulson et al. [174] have analyzed the flow of a compressible fluid

through an orifice and have shown that for an isentropic process, crit-

ical flow is obtained when

Pd 2 k/(k-1)
P k+1

where Pd is the downstream pressure, Pu is the upstream pressure and k

is the ratio of the heat capacity at constant pressure to the heat

capacity at constant volume. The heat capacity ratio ranges from 1.13

(PH3) to 1.66 (He) for the gases used in these experiments. Therefore,

keeping the ratio of the reaction tube pressure to the first vacuum

stage pressure below 0.49 will assure that all of the gases flowing

through the sample orifice are at critical flow. During the experi-

ments, the partial pressure of any measurable gas in the reaction

chamber was greater than 500 Pa. As is described in Appendix B, the

operating pressure of the first vacuum stage was between 10 and 50 Pa.

Thus, the requirement for critical flow was always met.

Applying the results of Coulson et al. [174] to equation 5-3

yields the following relationship

dPi A Ke RT 1/2 55
dt i V M. 1

where Ae = CDA0

CD = orifice discharge coefficient

A0 = orifice diameter

M. = molecular weight of species i


Ke = [k(2 (k+l)/(k-1)]1/2 5-6
e k+l

The variable Ae which represents the product of the orifice dia-

meter and discharge coefficient, is unknown since a sufficiently accu-

rate value for the orifice diameter is not known and the discharge co-

efficient is a function of the orifice Reynolds number. A relationship

between Ae, gas molecular weight and system temperature was therefore

determined experimentally using H2, He, N2 and CO2.

It has long been recognized that catalytic surfaces become less

active the longer they are in use, but usually exhibit a relatively

constant period of activity following an initial period of deactivation.

The quartz reactor surface was pretreated prior to the experiments by

pressurizing the tube to 130000 Pa with hydrogen at a temperature of

1073 K for 48 hours. The NH3 and AsH3 decomposition experiments were

repeated several times (days apart) in order to look for changes in the

catalytic activity. No changes in activity were observed. Between ex-

periments, the reaction tube was either maintained under vacuum or a

helium pressure of approximately 105 Pa.

The procedure employed for the collection of rate data was the same

for each group V trihydride. Each trihydride was investigated over a

full range of temperatures before the next was admitted to the system.

Ammonia was studied first followed by phosphine and finally arsine.

The desire to obtain a temperature profile which was as flat as

possible along the length of the reaction tube required the suppression

of all of the heat losses in the system. Thus, increasing the temper-

ature of the reactor could be accomplished quickly (-200 K/hr) while

decreasing the reactor temperature was a slow process. Therefore, most

of the data taken for each trihydride was acquired in increasing order

of temperature. This procedure was not strictly adhered to however,

since eventually, the next higher temperature investigated resulted in

reaction rates too fast to be followed with the current mass spectro-

meter configuration. It was therefore not unusual to allow the reactor

to cool so that rate data at intermediate temperatures could be gathered.

This occurred most frequently for AsH3 and occasionally for PH3.

The laboratory did not have the capability for automated control

of the mass spectrometer. Each scan of mass to charge ratio, there-

fore, was initiated manually. The fastest rate at which data could be

reliably scanned and averaged by the mass spectrometer and chart re-

corder was one scan every 10 seconds. This made it difficult to follow

reactions with half lives less than 30 seconds since relatively few

data points could be collected before the species signal was comparable

to the background signal or instrument noise.

The decomposition of NH3 was monitored by following the NH2 peak

at m/e = 16. This peak was followed rather than NH3 since the desorp-

tion of H20 off the walls of the second vacuum stage in the mass spec-

trometer caused a large OH+ background peak to be present at m/e = 17.

The decomposition of PH3 and AsH3 were followed by observing the

entire fragment ion pattern V VH+, VH and VH at m/e = 31, 32, 33

and 34 for PH3 and m/e = 75, 76, 77 and 78 for AsH3. No significant

background was observed at these mass to charge ratios. The levels

of confidence for the PH3 and AsH3 results are, therefore, much higher

than that for NH3.

The procedure employed for these experiments was to first evacuate

the reaction tube and then bring the reactor to the desired temperature.

The reactor was then charged with the desired trihydride by opening the

appropriate solenoid operated valves and monitoring the system pressure

by means of the capacitance manometer. The system was charged to a

pressure of 1.3x105 + 103 Pa for the NH3 experiments and 9.2x104 + 103

Pa for the PH3 and AsH3 experiments. The amount of time required to

charge the reactor was between 3 and 5 seconds. The first mass spec-

trometer scan was begun 15 seconds after the valves were closed. This

delay was primarily due to the restrictions imposed by manual operation

of the system (time required to close valves, start the chart recorder

and initiate the mass spectrometer scan), but also provided sufficient

time for thermal equilibrium to be established.

The amount of time required for the radial temperature profile

in the reaction tube to be flat within 0.1 K may be estimated from

the following analysis. Neglecting natural convection and heat losses

from the ends of the reaction tube, the radial temperature profile

as a function of time is found from the solution of

1 dT d2T 1 dT
a dt dr r dr

where a is the thermal diffusivity of the gas in the reaction tube.

Carslaw and Jaeger [175] have solved equation 5-7 subject to the

boundary condition of zero initial temperature, constant wall temper-

ature and radial symmetry. The result is

2 3 Jo(a r) e- a2t5-8
T = Tw(l R) anJl (aR)
n-1 n n

In this equation, T is the reactor wall temperature, R is the reactor

tube radius and the eigenvalues, an are the roots of the equation

Jo(a R) = 0 5-9

Since the reaction tube contained at least 90% H2 at all times, the

thermal diffusivity of H2 was used to evaluate equation 5-8. A further

assumption inherent in equation 5-8 is thatais invariant with respect

to temperature. This is not true for H2 as a goes as approximately T2

[77]. However, a worse case calculation can be performed by evaluating

a at the initial temperature (300 K) of the gas. The average and cen-

terline temperatures in the reaction tube are shown in Figure 5-3.

Five eigenvalues were used in the evaluation of equation 5-8 to achieve

this result. The reactor centerline temperature was found to be within

0.01% of the wall temperature (e.g. 0.1 K at Tw = 1000 K) for times

greater than 7 seconds. Evaluating a at higher temperatures decreased




0.6 --




0 2 4 6 8

Time (s)

Figure 5-3
Reaction Tube Temperature During Heating

this time sharply (e.g. evaluating a at 1000 K required only 1.9 s to

achieve the same results). The influence exerted on the thermal dif-

fusivity by the presence of VH3 in the system is on the order of 10%

to 15%. This is considered insignificant relative to the choice of

an appropriate temperature for evaluating a. Thus, based on the re-

sults depicted in Figure 5-3, it is concluded that the 15 s delay,

between charging the reactor and the initiation of data collection,

was sufficient to allow the gas in the tube to reach thermal equilibrium.

Mass spectrometer scans were made every 10, 30, 60 or 120 seconds,

depending on the reaction rate, and data was taken either for 1200 se-

conds or until the signal was too small to be reliably measured. Back-

ground signal measurements were made before and after the decomposition

data were acquired at each temperature in order to determine whether or

not the VH3 background signals were increasing with system exposure to

the VH3 species. A slight increase in PH3 and AsH3 background was no-

ticed over the course of the experiments, but the change at any one

data gathering session was insignificant. Changes in the NH3 background

could not be observed due to the large H20 background signals present.


Chemical Equilibrium Investigation


The product of the equilibrium calculations was the composition

of the vapor phase in the presence of excess condensed phases. The

composition was investigated as a function of temperature, pressure

and inlet gas composition. The usual procedure was to vary one of

the operating conditions while holding the remaining ones at their

base values. The graphical representation of these results illustrates

the equilibrium vapor mole fraction of each of the species versus the

parameter varied. Since the primary objective of this study was to

examine unintentional Si incorporation levels, mole fractions are

shown typically down to a level of 10-10 (0.1 ppb). The Si species

were always found to be below 10-5 mole fraction. Therefore, only

the lower five orders of magnitude were shown in many cases unless the

upper range was necessary to understand the results. A full parametric

analysis was performed and over 160 plots were generated. In many

cases, the results were similar to analogous system, thus, this chap-

ter includes only those graphs necessary for understanding the prin-

cipal phenomena predicted. In interpreting these plots, it should be

realized that an excess specie serves to hold the activity of that

specie at a constant value. For example, with solid SiO2 present, the

activity of SiO2 is fixed at unity and therefore the product of the Si


and 02 partial pressures is also fixed. Thus, changes in operating

parameters that alter the 02 fugacity will alter the Si activity by

the same degree in an inverse fashion.

The GaAs Chloride System

The effect of temperature on the species present in the GaAs

chloride system source zone (100 kPa pressure, inlet composition:

1% AsC13 in H2) is shown in Figures 6-1 through 6-4. Figures 6-1 and

6-2 apply to the system which used a liquid group III source, GaxAslx

and Figures 6-3 and 6-4 represent the results for the system which em-

ployed GaAs(c) as the group III source material. At low temperatures,

GaC12 and GaC13 became relatively important gallium vapor species along

with GaC1 in the solid source system. In the liquid source system,

GaC1 is the dominant gallium species over the entire temperature range

examined (873
arsenic species at low temperature while As2 became important at high

temperatures. In contrast to previous studies [20, 21], the trimer,

As3, mole fraction was not negligible. In general, comparison of the

silicon activity for the two source zones revealed that the silicon

activity associated with the GaAs(c) source material was much lower

than that which resulted when a liquid source material was employed.

The predominant silicon species in the vapor phase of the system which

used a solid source were the higher silicon chlorides in contrast to

the hydrogen rich silicon species found in the system which used a li-

quid source. An additional interesting feature is that the total mole

fraction of silicon compounds in the vapor for the system which employed

a solid source was greater than that for the system which employed a

liquid source. At first glance, this fact seems contradictory to the

lower observed silicon activity.



0 10-2





873 973 1073 1173

Temperature (K)

Figure 6-1
Effect of Temperature in the GaAs Chloride System Source Zone
(Liquid Source)

973 1073

Temperature (K)

Figure 6-2
Effect of Temperature in the GaAs Chloride
(Liquid Source)

System Source Zone






1 0

-1010 L1
0 873


103 o



10-4 I



r 10-2, As




873 973 1073 1173

Temperature (K)

Figure 6-3
Effect of Temperature in the GaAs Chloride System Source Zone
(Solid Source)




873 973

Temperature (K)

Effect of Temperature

Figure 6-4
in the GaAs Chloride
(Solid Source)

System Source Zone






The following reaction equations may be written to describe the

formation of silicon chlorides, chlorosilanes and silane resulting

from reactions with the quartz reactor wall.

Si02(c) + nHCl + (4-n)H2 = 2H20 + SiH4nCln n=0,1,2,3,4 6-1

kSi02(c) + kmHC1 + k(2-m/2)H2 = 2kH20 + SikCkm

k=1,2 m=0,1,2,3 6-2

Reactions 6-1 and 6-2 represent a set of independent formation reactions

which describe the interplay between the dominant vapor phase silicon

species present in the system. Assuming ideal gas behavior, the equili-

brium constants for these reactions are as follows:
2 4-n n 6-3
l,n= H20 YSiH nCln YH. HC1 6-3


K = 2k 1-km2 k(2m/2)
2,k,m H20 YSikClkm P- /H2km 6-4

where y = vapor phase mole fraction of species i

PT= system total pressure ratio (total pressure/reference
state pressure)

The activity of silicon residing in a condensed phase which is at

equilibrium with the vapor phase may be calculated from any reaction

using a vapor Si species reactant and solid Si product. For example,

consider the following reactions and subsequent equilibrium expressions

for the activity of Si(c)
YSi H4
Si(c) + 2H2 = SiH4 aSi= 4 6-5
K5 2 PT
Si(c) + 4HC1 = SiC14 + 2H2 ai= YSiCl4 Y2 6
S4 H6-6

Here K. is the equilibrium constant for reaction (i=5,6). Other equi-

valent relations may be written in order to calculate the condensed

phase silicon activity but the models suggested in equations 6-5 and

6-6 serve as convenient points of focus since either SiH4 or SiC14 is

usually a significant silicon vapor species in the systems studied.

In particular, for those systems using H2 as the carrier gas, the mole

fraction of H2 is nearly constant with a value close to unity. There-

fore, aSi will track the SiH4 mole fraction and is inversely propor-

tional to the system pressure. Both models of course yield identical

values for the silicon activity when applied to the same situation. The

activity of Si presented in these plots can be viewed as the value found

in a solid phase in equilibrium with a vapor having the composition

shown. In order to translate this into a solubility, the nature of the

solid phase must be considered (i.e. the activity coefficient must be

known). Due to the low degree of doping encountered in these systems,

(e.g. typically < 1015 cm-3, which is <100 ppb), the activity coeffi-

cient can be represented by Henry's constant, which is invariant with

respect to composition. Therefore, an increase in Si activity corre-

sponds to an increase in solubility. Thus, the models provided by

equations 6-5 and 6-6 may be used to predict the direction of change

in the silicon concentration based on the knowledge of the vapor phase

equilibrium composition.

The lower silicon activity associated with the solid GaAs source

can therefore be viewed as due to a suppressed SiH4 concentration when

compared to the liquid source (equation 6-5). In the source zone,

which employed solid GaAs, the presence of primarily higher chlorides

and chlorosilanes at the lower source zone temperatures was a result

of less gallium being present in the vapor phase than was present when

a liquid source was employed. Since Ga was in excess in both the li-

quid and solid source systems, the activity of Ga was constrained with

the liquid source exhibiting a higher Ga activity. Thus, sufficient

HC1 was available due to the decomposition of AsC13 and the absence of

enough Ga to consume it to enhance reactions 6-1 and 6-2 for large k

and m values. Figure 6-5, which shows the chloride system pre-source

zone (1% AsC13 in H2, no group III source material present), further

supports this analysis. The absence of group III chlorides caused the

total amount of silicon in the vapor to increase above the level ob-

served in the solid source system while the condensed phase silicon

activity decreased even further.

Table 6-1 lists the enthalpy of formation at 298 K and the Gibbs

energy of reaction at 973 K for some of the vapor species described by

reactions 6-1 and 6-2. The large negative enthalpies of formation are

indicative of strong interatomic bonds and therefore stable species.

Since equilibrium represents the state having the lowest value for the

Gibbs energy of the system, species with a lower Gibbs energy of re-

action are favored. Therefore, providing sufficient chlorine to react

with the silicon species results in a higher total silicon concentration

in the vapor phase but, due to the stability of these species, a lower

activity of solid silicon in the condensed phase. The relative stability

of silicon halides, when compared to silicon hydrides was also recog-

nized by Rai-Choudhury [12].

The outlet equilibrium compositions of the source zones using

solid or liquid source materials at 973 K were used as input to the

mixing zone and the effect of mixing zone temperature was investigated.