<%BANNER%>

Boron Interaction with Germanium and Self-Instertitials in Silicon


PAGE 1

BORON INTERACTION WITH GERMANI UM AND SELF-INTERSTITIALS IN SILICON By LJUBO RADIC A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Ljubo Radic

PAGE 3

This document is dedicated to the professo rs that perpetually ask the question: And why is that?

PAGE 4

iv ACKNOWLEDGMENTS I would like to thank my advisor, Dr. Ma rk E. Law, for the support and guidance throughout this work. This section would be incomplete without a mention of my colleagues from the SWAMP group, whom I ow e many an interesting discussion. This includes Ibrahim Avci, Tony Saavedra, Renata Camillo-Castillo, Robert Crosby, Robert Robison, Chad Lindfors, Michelle Phen and others.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES............................................................................................................vii LIST OF FIGURES.........................................................................................................viii ABSTRACT.....................................................................................................................xi ii CHAPTER 1 INTRODUCTION........................................................................................................1 1.1 Brief Historical Overview.......................................................................................1 1.2 MOS Transistor Structure.......................................................................................3 1.3 MOS Transistor Operation.....................................................................................3 1.3.1 Limiting Factors in MOS Performance........................................................4 1.3.2 MOS Transistor Scaling...............................................................................4 1.4 Semiconductor Processing Technologies...............................................................5 1.4.1 Ion Implantation...........................................................................................6 1.4.2 Diffusion.......................................................................................................9 2 LITERATURE REVIEW...........................................................................................22 2.1 Dopant Diffusion in Silicon..................................................................................22 2.2 Self-diffusion in Silicon........................................................................................24 2.3 Germanium Diffusion in Si1-xGex.........................................................................25 2.4 Boron Diffusion in Silicon....................................................................................27 2.4.1 Boron as Interstitial Diffuser in Silicon......................................................27 2.4.2 Pre-amorphization and B............................................................................28 2.5 Boron-interstitial Clusters.....................................................................................29 2.6 Boron Diffusion in Si1-xGex..................................................................................31 2.7 Modeling Boron Diffusion in Si1-xGex.................................................................35 3 BORON CLUSTERING IN SILICON......................................................................62 3.1 Boron-interstitial Cluster Dissolution...................................................................62 3.1.1 Introduction................................................................................................62 3.1.2 Experimental...............................................................................................64

PAGE 6

vi 3.1.3 Results and Discussion...............................................................................65 3.1.4 Modeling.....................................................................................................67 3.2 Boron Cluster Formation and Preamorphization..................................................68 3.2.1 Experiment and Findings............................................................................68 3.2.2 Modeling.....................................................................................................70 3.3 Conclusion............................................................................................................75 4 BORON CAPTURE AT THE OXIDE/ SILICON INTERFACE AND END-OFRANGE DISLOCATION LOOPS.............................................................................98 4.1 B Interaction with Oxide/ silicon Interface Under TED........................................98 4.1.1 Experimental Conditions............................................................................98 4.1.2 Simulations...............................................................................................102 4.2 Boron Segregation to the EOR Loops................................................................104 4.2.1 Experimental.............................................................................................105 4.2.2 Simulations...............................................................................................105 4.3 Conclusions.........................................................................................................106 5 BORON INTERACTION WITH GERMANIUM UNDER INTERSTITIAL SUPERSATURATION............................................................................................127 5.1 Prior Work Mode l Implementation..................................................................127 5.1.1 Bandgap Narrowing..................................................................................127 5.1.2 Boron Diffusion in Si and Interaction with Ge........................................128 5.2 Experiment Design and Considerations..............................................................129 5.2.1 Experimental Conditions..........................................................................131 5.2.2 Experimental Results................................................................................132 5.2.3 Simulations...............................................................................................133 5.3 Constant B Concentration Experiment...............................................................134 5.3.1 Experimental conditions and results.........................................................134 5.3.2 Simulations...............................................................................................136 5.4 Conclusions.........................................................................................................136 6 SUMMARY AND FUTURE WORK......................................................................151 6.1 Summary.............................................................................................................151 6.2 Future Work........................................................................................................153 6.2.1 Material Specification B Ma rker Layers in Relaxed Si1-xGex...............153 6.2.2 Information Expected...............................................................................154 LIST OF REFERENCES.................................................................................................157 BIOGRAPHICAL SKETCH...........................................................................................162

PAGE 7

vii LIST OF TABLES Table page 2-1 Extracted B diffusivity prefactor ( D0) and activation energy ( Ea) from experiment of Zangenberg et al................................................................................38 3-1 Retained boron dose [cm-2] in different processing steps........................................77 4-1 Sample labels and implant sequence, showing Ge implant dose as the only variable in the sequence.........................................................................................108 4-2 Active dose and carrier type meas ured by Hall-van der Pauw system..................109 4-3 Integrated dose from the measured B pr ofiles in the control sample annealed at 825oC, taking full profile (Q1) and ignoring the surface spike (Q2).......................110 4-4 Integrated dose from the measured B profiles in the 4x1015 cm-2 Ge implanted sample annealed at 825oC, taking full profile (Q1) and ignoring the surface spike (Q2).........................................................................................................................111

PAGE 8

viii LIST OF FIGURES Figure page 1-1 Triode electron tube schematic ...............................................................................14 1-2 Number of components per integrated circuit vs. time, later known as Moores law .........................................................................................................................1 5 1-3 Number of transistors vs. time, shown to realize the original prediction of G. E. Moore, as well as updated trends ............................................................................16 1-4 Schematic cross-section of a nMOSFET transistor..................................................17 1-5 Schematic of an ion implanter .................................................................................18 1-6 Illustration of an ion implantation proce ss, in particular (a) a single ion path and (b) the resulting damage cascade ............................................................................19 1.7 Silicon crystal viewed from (a) < 110> direction and (b) tilted ~10o off the <110> direction .......................................................................................................20 1.8 Illustration of fluxes enteri ng and exiting a given volume.......................................21 2-1 Diffusion mechanisms in silicon: a) interstitialcy diffusion, b) kick-out, c) vacancy diffusion, and d) concerted exchange ........................................................39 2-2 Self-diffusion in Si as measured by Bracht with symbols denoting different samples.....................................................................................................................40 2-3 Self-diffusion in Si as measured by Ural et al. Dots represent measured diffusivities, solid line is the best fit expression for self-diffusion..........................41 2-4 Interstital fraction, fSiI of the self-diffusion from Ural et al. The lines represent predictions from several metal diffusion experiments.............................................42 2-5 Germanium diffusivities for various temperatures and Ge contents, after Zangenberg et al ......................................................................................................43 2-6 Activation energies and prefactors for various Ge contents, after Zangenberg.......44 2-7 Activation energies and prefactors fo r various Ge contents, after Strohm..............45

PAGE 9

ix 2-8 Comparing diffusivities of Ge in Si1-xGex as measured by Zangenberg and Strohm......................................................................................................................46 2-9 Comparison of Ge diffusion with respect to ambient and consequential interstitial or vacancy supersaturation .....................................................................47 2-10 B profiles after an implant of 2x1014 at 60keV, annealed at a) 800 oC and b) 950oC .......................................................................................................................48 2-11 B profiles, as-grown and after 1h at 670oC .............................................................49 2-12 Fractional activation and sheet re sistance of samples annealed at 750oC for: a) varying B dose at 80keV, and b) 4x1014 cm-2 at varying energy ............................50 2-13 B profiles used in experiment by Mi rabella et al. showing the as-grown, annealed for 2 min at 815oC, as well as illustrating the methodology for determining the active fraction ................................................................................51 2-14 Time evolution of the clustered B dose during annealing at 815 to 950 oC, with extracted time constants. .........................................................................................52 2-15 Structure used by Kuo et al. with a B marker layer in the Si0.83Ge0.17 strained layer.......................................................................................................................... 53 2-16 Diffused B profiles after 30 minute anneal at 860oC in a) Si and b) Si0.83Ge0.17 ....54 2-17 Effective B diffusivity as a function of B concentration .........................................55 2-18 Diffusivity of B in Si0.7Ge0.3 layer ...........................................................................56 2-19 B profiles in multiplayer structure before (dashed line) and after (solid line) annealing at 975oC ..................................................................................................57 2-20 Illustration of a structure used by Kuo et al. showing relaxed Si1-yGey and pseudomorphic Si1-xGex...........................................................................................58 2-21 B diffusivity at 800oC as a function of strain, with the numbers in the parenthesis are ( x,y ) Ge content. Positive strain represents biaxial tension and negative strain bi axial compression. .......................................................................59 2-22 Schematic description of the structur e used by Zangenberg et al. to measure B diffusion in Si1-xGex.................................................................................................60 2-23 B profiles before (thin line) and after (thick) 850oC annealing of a given B well concentration............................................................................................................61 3-1 Secondary ion mass spectrometry (SIMS) measured boron profiles of investigated implant conditions after first anneal step, 750oC for 30 minutes in inert ambient.............................................................................................................78

PAGE 10

x 3-2 Secondary ion mass spectrometry (SIM S) measured boron profiles. Initial condition is after 750oC, 30 minutes inert anneal. Subsequent anneal is 850oC, 60 minutes in respective ambient.............................................................................79 3-3 Active dose measured by Hall-van der Pauw method, during the annealing at 850oC. The time zero measurement corres ponds to the condition after a 750oC, 30 minutes inert anneal............................................................................................80 3-4 Active fraction, the ratio of active to retained dose (integra ted SIMS profile), compared for the different annealing ambient at the end of the 850C, 60 minutes anneal. The initial condition is after a 750oC, 30 minutes inert anneal.....81 3-5 Major cluster formation and disso lution paths of Liu et al. with B3I and B2I3 containing most of the B clustered dos e in figure a) and simulation of B clustering and dissolution dur ing thermal processing used in experiment in figure b)....................................................................................................................82 3-6 Cluster formation and dissolution paths of modified Liu et al. energetics, with B4I4 and B2I3 containing most of B clustered do se, are shown in figure a), with simulation of B clustering and dissoluti on (modified energetics) during thermal processing used in experiment is shown in figure b)...............................................83 3-7 Boron profiles of material used in the study of Jones et al., unimplanted and annealed in inert ambient at 800oC..........................................................................84 3-8 Boron profiles after a 5x1015 Si implant at 146keV, annealed in inert ambient at 800oC .......................................................................................................................85 3-9 Boron marker layers and excess in terstitial damage following a 146keV Si implant shows proximity of deepest marker layer to the damage............................86 3-10 Doses of important cluste rs in a simulation of a B 4x1014 20keV implant into crystalline Si, during a 30 minutes 750oC anneal, resulting in activation ~18%.....87 3-11 Doses of important cluste rs in a simulation of a B 4x1014 20keV implant into preamorphized Si (with excess interstiti al dose increased by a factor of 6), during a 30 minutes 750oC anneal, resulting in activation >85%............................88 3-12 Doses of important cluste rs in a simulation of a B 4x1014 20keV implant into preamorphized Si (at liquid nitrogen te mperatures), during a 30 minutes 750oC anneal, resulting in activation >90%........................................................................89 3-13 Simulated boron profiles after 5 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in the amorphous material............................................................................................90

PAGE 11

xi 3-14 Simulated boron profiles after 30 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in the amorphous material............................................................................................91 3-15 Simulated boron profiles after 3 minutes at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in the amorphous material............................................................................................92 3-16 Simulated boron profiles after 5 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in the amorphous material, and the binding en ergies of small BICs are reduced........93 3-17 Simulated boron profiles after 30 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in the amorphous material, and the binding en ergies of small BICs are reduced........94 3-18 Simulated boron profiles after 3 minutes at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in the amorphous material, and the binding en ergies of small BICs are reduced........95 3-19 Simulated clustered doses in BIC experiment for B implant 4x1014@ 20keV, model presented earlier in the chapter......................................................................96 3-20 Simulated clustered doses in BIC experiment for B implant 4x1014@ 20keV, model parameters adjusted to cover regrowth behavior ..........................................97 4-1 Profiles of B and Ge in as-implanted sample A.....................................................112 4-2 Cross-section TEM (XTEM) images of samples (a) A and (b) D under 100000 magnification..........................................................................................................113 4-3 B profiles from Ge implanted samples during annealing at 700oC........................114 4-4 B profiles from Ge implanted samples during annealing at 825oC........................115 4-5 SIMS profiles of Ge, As and B in samples: A, with 4x1015 cm-2 Ge (1min@825oC), B with 1.2x1015 cm-2 Ge (as-implanted) and C with 4x1014 cm-2 Ge (as-implanted) per figure a, b, c, respectively..................................................116 4-6 B profiles from control sample during annealing at 700oC....................................117 4-7 B profiles from control sample during the annealing at 825oC..............................118 4-8 Simulated B profiles of control sample during anneals at 825oC...........................119 4-9 Simulated B profiles of control sample during anneals at 825oC, diffusion enhancement factor reduced five times..................................................................120

PAGE 12

xii 4-10 Simulated B profiles of control sample during anneals at 825oC, assuming surface trapping of BI.............................................................................................121 4-11 Simulated B profiles of 4x1015 cm-2 Ge implanted sample (As contaminated) during anneals at 825oC, assuming electric fi eld effect dominant.........................122 4-12 B profiles from sample E1 during annealing at 750 oC..........................................123 4-13 B profiles from sample E1 during annealing at 825oC...........................................124 4-14 Simulated B profiles of B segregatio n to the loops in sample E1 during annealing at 750oC.................................................................................................125 4-15 Simulated B profiles of B segregatio n to the loops in sample E1 during annealing at 825oC.................................................................................................126 5-1 Profiles of B and Ge in as-implanted sample 2A...................................................138 5-2 Cross-section TEM (XTEM) image of sample 2A under 100000 magnification..139 5-3 Boron profiles duri ng the annealing at 780oC for: a) Ge implanted sample, b) control sample........................................................................................................140 5-4 Simulated B profiles of Ge impl anted sample during anneals at 780oC................141 5-5 Simulated B profiles of control sample during anneals at 780oC...........................142 5-6 B profiles from control sample during annealing at 750 oC...................................143 5-7 B profiles from control sample during annealing at 825oC....................................144 5-8 B profiles from Ge implanted sample during annealing at 750 oC........................145 5-9 B profiles from Ge implanted sample during annealing at 825 oC........................146 5-10 B profiles from Ge implanted sample during annealing at 560 oC........................147 5-11. B profiles from Ge implante d sample during annealing at 600 oC.........................148 5-12 Simulated B profiles from Ge impl anted samples during annealing at 750oC.......149 5-13 Simulated B profiles from Ge impl anted samples during annealing at 825oC.......150 6-1 Dopant profiles in a Si1-xGex material. Figure a) cont ains Ge profile for Si0.8Ge0.2 material, while figure b) has B profile...................................................................155 6-2 Boron profile in Si0.8Ge0.2 material........................................................................156

PAGE 13

xiii Abstract of Dissertation Pres ented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy BORON INTERACTION WITH GERMANI UM AND SELF-INTERSTITIALS IN SILICON By Ljubo Radic August 2006 Chair: Mark E. Law Major Department: Electrical and Computer Engineering Understanding the diffusion phenomena in sili con is imperative for the rapid and efficient development of the future semiconduc tor processes. Boron is pertinent as the dopant of choice in formation of a p-type semiconductor, yielding higher activation and carrier mobilities than the alternatives. In order to overcome certain transient behavior, such as transient diffusion enhancement a nd clustering, these phenomena have to be characterized and modeled. One part of this dissertation describes th e investigation of th e properties of the significant boron-interstitial clusters (BICs). Previous models descri bed boron clustering through the single dominant cluste r, assumed to be interstiti ally poor. An experimental investigation of the BIC dissolution with resp ect to annealing ambient revealed a reaction contrary to the expectations from that model, suggesting that one of the dominant clusters may be interstitially rich. Additionally, the ti me dependence of the reactivation indicated the reactivation process was coming from two sources. Using a set of ab-initio

PAGE 14

xiv calculations as a starting point adjusting to dependence on the annealing ambient, the model showed appropriate react ion. An earlier experiment probed the formation of the BICs with a separation by amorphous/crystallin e interface. Assuming th e inability of the self-interstitial or boron-inter stitial pair to ente r the amorphous layer si gnificantly reduces the cluster formation. However, further adju stments were necessary to properly capture the difference between the enhanced di ffusion phenomena in the regrown amorphous layer, and a distinct clustering pe ak in the crystalline material. Another part of the dissertation examines the interaction of Ge and B in silicon under interstitial s upersaturation. Experimentally observed segregation of B onto oxide/silicon interface in samples with no Ge im plant did not occur in the presence of Ge. In conjunction with increase in peak concentration of B in the presence of Ge, this experimental evidence supports an earlier hypoth esis of the Ge-B pair ing as the cause of the diffusivity reduction of B in Si1-xGex. Modifying the model to account for the interstitial supersaturation and extracting the pair formation energy from a relative increase in the diffusion activation energy in higher Ge concentration, the model qualitatively captures the phenomena.

PAGE 15

1 CHAPTER 1 INTRODUCTION 1.1 Brief Historical Overview The twentieth century saw several significan t changes in the fields of engineering. The advent of the vacuum tube gave rise to the field of electronics It consisted of the filament as the emissive element, cathode, a node and the control grid (Fig. 1-1). With negative voltage applied to the grid the cathode current could be cut off, while increasing the control voltage amplified the current. Amp lification of the signal made long distance phone and radio communication practical. One of the significant benefits of th e vacuum tubes is the simplicity of construction. In the 1930s, Julius Edgar Lili enfeld filed two patents for controlling electric current [Lil30] which would today be described as MESFET (metalsemiconductor field effect transistor) a nd depletion-mode MOSFET (metal-oxidesemiconductor field effect transistor). Sin ce the clean environment and surface control required to produce viable semiconductor materi al were not available before the middle of the twentieth century, his inventions we re forgotten until twenty years later. Another application of cathode tubes was in early computers. Electronic Numerical Integrator And Computer (ENIAC) was the fi rst electronic computer, built from 1943 to 1946, and it contained 17,468 vacuum tubes, 7,200 crystal diodes, 1,500 relays, 70,000 resistors, 10,000 capacitors and around 5 m illion hand-soldered joints. It required rewiring for execution of a new program, and could perform 5000 additions per second while consuming 160kW of power.

PAGE 16

2 The first fully functional transistor was developed at Bell Laboratories, by John Bardeen, Walter Houser Brattain, and William Bradford Shockley, who were awarded the Nobel Prize in physics in 1956. Though their goal was to use the high quality germanium crystals to form a field-effect transistor (FET) predicted by Lilienfeld, but they actually found a bipolar junction transi stor (BJT). After some time, and a number of solved problems, the percentage of functional transistors (yield) was still not very high and production of high purity germanium stil l posed a problem. In the 1960s, Texas Instruments tried using silicon instead of germanium, which was easier to work with. Several factors led to the ascent of the si licon based MOSFETs: wa ter insoluble native oxide (SiO2), physical and electrical properties of the oxide, no interface states between the silicon and the oxide that co uld serve as electron traps or recombination sites, as well as improvements in production facilities. By that time, transistors replaced majority of the vacuum tubes. The trend of miniaturization continued as Jack Kilby [Kil59] at Texas Instruments patented a “solid circuit” in germanium. Soon afterwards, Robert Noyce of Fairchild Semiconductor was awarded a patent for a "uni tary circuit" made of silicon. Planar MOSFET technology scaling has been drivi ng the semiconductor technologies since then. As expressed in an observation by Gordon E. Moore [Moo65], the number of transistors per square inch on integrated circuits had doubled every year since the integrated circuit was invented (Fig. 1-2). Th is growth continues, as shown in Fig. 1-3 [Moo03], to this day. The rest of the introdu ction will concentrate on MOS technology as the dominant technology today.

PAGE 17

3 1.2 MOS Transistor Structure MOSFET transistors fall into the category of unipolar transistors, meaning that only one carrier type is respons ible for conduction. While diffe rent types of FETs control the conduction between the source and the drain (S /D) electrodes in different ways, all of them operate with either electrons or holes. This is not the case for the BJT, where both carrier types are involved. MOS transistors can be divided into several types. First, the tr ansistor can have a n-type or a p-type channel. S econd, the transistor can have a channel existing or not, with no voltage applied to the gate electrode. In case there is an electrical connection existing between the source and the drain region, the devi ce is said to be a depletion mode device. Otherwise, there is no connection unless the prop er voltage is applied to the gate, such as in enhancement mode device. Most transistors in digital circuits are used as switches, for which the enhancement mode device is more suitable. 1.3 MOS Transistor Operation The schematic cross-section of an enha ncement mode nMOSFET is shown in Fig. 1-4. The external contacts ar e source, drain, and gate. Body contact is usually connected to the source, thus physically differentiating between the source and the drain. This also forms a reverse polarized pn junction at the drain-body boundary, as VDS>0 is necessary for nMOS operation. For a VGS lower than the threshold voltage VT, the p-type region separates the source and the drai n allowing no current flow, and the device is said to be turned off. As the VGS increases to values over the VT, the electron rich region forms a channel near the surface between the source and the drain. Th is facilitates the current flow from the drain to the s ource, and the device is said to be turned on. Therefore, the

PAGE 18

4 gate voltage VGS controls the current flow between the drain and the source with the transistor acting as a switch. 1.3.1 Limiting Factors in MOS Performance The MOS device performance is limited by the design and th e parasitic effects. The design parameters, such as gate length, widt h, and oxide thickness, ga te material, source and drain region thickness, factor in to de termine the transistor’s performance. The parasitic effects are leakage currents, pa rasitic resistances and capacitances. The technologically necessary overl ap of the gate over the sour ce and the drain region forms two capacitors: CGS and CGD. Even though the capacitance amounts are usually similar, the latter is electrically more significant. During the switching from an off state into an on state, the transistor operates as a small signa l amplifier. The increase in the small signal difference between the gate and the drain effectively increases CGD by the voltage amplification factor. The parasitic resistances include the resistances in the S/D regions, their extensions, transition resistances towards metal contact, and inte rnal resistance of the poly-silicon gate electrode. 1.3.2 MOS Transistor Scaling The continued advances of semiconducto r technology, expressed in transistor density and speed, are due to the scaling of the planar technology on silicon and the properties of silicon dioxide. Reduction in gate length had increased the current and reduced the input capacitance, both of which reduce switc hing time. Decreasing the dimensions of the transistor also meant that more transistors were av ailable per chip, thus increasing the computing power. The decrease of the dimensions of th e transistor does pose some technological difficulties. The breakdown electric field of the silicon dioxide is a material property one

PAGE 19

5 can not circumvent. Thus the scaling limite d by the maximum electric field is labeled constant field scaling, requiring the reduction of the gate voltage in proportion to the reduction of the oxide thickness. As the s ource and the drain regions come closer together, the depletion regions could touch or merge, givi ng rise to a leakage current irrespective of the gate vo ltage. This is known as subthreshold leakage, causing problems with the power consumption and the detection whether the transistor is on or off. Similarly, the reduction of the lateral dimensions increases the capacitances between metal lines. The reduction of physical dimens ions also requires reduction in operating voltages in order to keep below the brea kdown electric field of the oxide. Though the operating voltage is reduced, the extremely small oxide th ickness poses a problem as tunneling becomes significant with several m onolayer thick oxides. New materials with higher dielectric constants, allowing for thicke r layer, have also been introduced. In order to keep the conduction of the transistor under the control of the ga te, the conduction must take place on or near the surface. With reduction of operating vo ltages that area is reduced as well. This means that source and dr ain region depths have to be scaled with the lateral dimensions. The parasitic resistan ces of S/D are invers e proportional to the depth, thus requiring increased doping to k eep them fixed. The contact material and resistance also play a role in the parasitics, somewhat mitigated by the recent switch from Al to Cu for metal interconnects. 1.4 Semiconductor Processing Technologies Semiconductor integrated ci rcuits are formed by an application of following processes: (1) growth or a deposition of a la yer, (2) exposure to light, (3) patterning or etching, (4) dopant implantation, and (5) th ermal annealing. The latter two are the

PAGE 20

6 frequently used to form the n-type or ptype semiconductor regions, later to become a part of the device, and will be discussed in further detail. 1.4.1 Ion Implantation Ion implantation is the dominant met hod of introducing dopant into silicon material. The accuracy of the implanted dose and the dopant distribution in depth are the reason for its widespread use. The process of ion implantation begins with a gas or a solid source providing an ionized beam (Fig. 1-5). The ion beam then goes through a magnet and an aperture, providing the mass separation of elements and isotopes extracted from the source. The voltage used to accelerate the ions from the source is called the extraction voltage. The extraction voltage and the accelera tion voltage both contribute to the final ion’s implant energy. The ion beam pa sses through the scanning plates in x and y directions, and the deflection be fore reaching the target wafer. The wafer sits in a Faraday cage that repels the incidental electrons and integrates the incoming charge from the ion beam. The direct and accurate measurement of the incoming ion beam current, also the implanting dose rate, provides the accur acy in the total implanted dose. The aforementioned deflection serves to remove ne utrals, ions that lost the charge between the aperture and the scan plates, from the beam reaching the target wafer and introducing an error in the dose measurement. The two major mechanisms of the energy lo ss of an ion entering the crystal lattice are electron ( Se) and nuclear ( Sn) stopping (Eq. 1-1), with N as the lattice density. ) ( ) ( E S E S N ds dEn e (1-1)

PAGE 21

7 As the incoming ions enter the crystal lattic e, they are stripped of the outer electron shell and slowed by the interaction with (“drag” of) the bound and free electrons. The stopping energy can be approximated by a velo city proportional, and expressed as E k E Se ) ( (1-2) Nuclear stopping involves Coulombic inte raction between the ion and the host lattice atom. As the ion approaches the target atom, the electrostatic repulsion deflects the ions path, and/or removes the host atom from its lattice position. In the case neither the ion nor the host atom occupy the lattice pos ition, the collision has produced a vacancy and an interstitial atom (the dislodged host atom). Altern atively, the ion performs a replacement collision, by removing the host atom from a lattice position and occupying it, thus becoming substitutional a nd creating a self-interstitial. For practical calculations, nuclear st opping can be approximated [Zie00a] by ) / /( ) ( 10 462 8 ) (2 23 0 2 23 0 1 2 1 1 2 1 15cm atom eV S Z Z M M M Z Z E Sn n (1-3) where the Z and M are atomic number and mass, subscripts 1 and 2 denote ion and target atoms, and the reduced energy is 23 0 23 0 1 2 1 2 1 0 253 32zZ Z M M Z Z E M (1-4) and the reduced nuclear stopping Sn() can be calculated as for e<30: 5 0 21226 019593 0 01321 0 2 1383 1 1 ln ) ( nS (1-5) for e>30: 2 ln ) ( nS (1-6)

PAGE 22

8 The occurrence of collision events is ra ndom and makes the individual ion path rather difficult to predict (Fig. 1-6a). Though the individual ion travels a distance R the peak of ion distribution is the mean projected range Rp, a far more interesting value. The projected range is defined as i i px N R 1 (1-7) where xi is the depth of a given ion i It can also be a pproximated [Lin63] by 1 23 1 M M R Rp (1-8) In ideal case of uniform and amorphous ta rget, with collision frequency and energy transfer per collision being random, the implan ted ion distribution can be described by a Gaussian function N(x) Implant dose is Q and Rp is the vertical distribution spread. 2 22 ) ( exp 2 ) (p p pR R x R Q x N (1-9) For a crystalline target, the distribution is somewhat different. Certain directions in the crystal structure allow for less collisions and deeper penetration, commonly referred to as channeling. The nuclear stopping is minimal in the case of channeling, as the angle of incidence is rather small and the Coulom bic repulsion keeps the ion in the channel. The electron stopping is smalle r, thus allowing the ion to travel signifi cantly longer distances. Fig. 1.7 illustrates the difference between looking into a channel and when the tilt is applied. Wafer tilt and twist are used to prevent channeling by providing seemingly random distribution of silicon atoms on the surf ace plane. Also, a sacrificial oxide layer

PAGE 23

9 called screening oxide is sometimes used to randomize the incoming ion’s trajectories before entering the silicon lattice. During a single ion’s path, it may displace as much as 104 silicon atoms by nuclear collisions before it comes to rest (Fig. 1-6b). If the primary ion collides with a host lattice atom with more than th e energy of displacement Ed ( Ed is 14 eV for silicon [Bau69]), the now interstitial atom continues to travel th rough the lattice until it collides with another lattice atom, thus creating additional damage cascades. The damage depends on the atomic species, as heavier atoms lose most of their energy in collisions, while lighter atoms initially lose a significant portion of their energy due to electronic stopping. In case of high implant doses, the damage can displa ce a percentage of the lattice atoms (over 10% [Chr81]), creating amorphous zones. The amorphization is sometimes used as an introductory step in low energy implants to prevent channeling, called pre-amorphization implant (PAI). After the implantation step is complete, th ere can be a significant interstitial and vacancy concentrations in the implanted re gion. Only a small portion of the implanted dopant atoms has occupied the lattice positio ns rendering them electrically inactive. These damaged regions need to be restor ed to a monocrytallin e lattice and dopants activated, which is accomplished by a high temperature thermal treatment. 1.4.2 Diffusion Fick’s laws describe diffusion of given speci e, as the change of concentration with respect to position and time. Diffusion is a thermally activated process, so one can write the probability p of an atom species X hop between two lattice positions as kT G pm Xexp (1-10)

PAGE 24

10 where m XG is migration energy of a species X and k is the Boltzmann constant. The frequency of hop attempts could be approximated by the Debye frequency which makes the frequency of hops species X makes to a neighboring site p (1-11) Now, assuming two lattice plan es separated by a distance a, with concentrations n1 and n2, jump frequency of 2 1 in each direction, during a small interval of time t the atom X hops can be written as plane 1 -> plane 2 t n 12 1 (1-12) plane 2 -> plane 1 t n 22 1 (1-13) thus resulting in the flux J over an area A 2 12 1n n A J (1-14) If one expressed the above equa tion in terms of concentration c a A n c (1-15) 2 12 1c c a J (1-16) and applied it over the incremental distance x it gives the formula of the Fick’s first law. x c D x c a J 22 1 (1-17) Given the above equation, one can express the diffusivity as

PAGE 25

11 kT G a g a g Dm Xexp2 2 (1-18) with g as the geometric factor, 2 1 g for one dimensional diffusion, and 6 1 g for a three dimensional problem. The migration energy consists of enthalpy H and entropy S as in the following equation m X m X m XS T H G (1-19) Incorporating that into the diffusivity equation yields th e final expression, kT Q D kT H k S a g DX X m X m Xexp exp exp0 2 (1-20) with D0X as the prefactor, and the QX as the activation energy. The first Fick’s law (derived above) describes the relation of flux with concentration, though it has no dependence on time. The second Fick’s law defines the relationship of the concentra tion with time and position. If a flux J1 entering the box is greater than the flux J2 exiting the box, the concentration in the box must increase in accord with the conservation of matter principle. The principle of Fick’s sec ond law is described in equations below x J x J J x A J A J A t c 2 1 2 1 (1-21) The number of dopants entering the box per unit time from the left is AJ1, dopants leaving the box on the right is AJ2, and their difference divided by the volume is equal to concentration per unit time. Using the equati on from the first Fick’s law, the equation governing the diffusion of dopant look like this in one dimension. 2 2x c D x c D x x J t c (1-22)

PAGE 26

12 In three dimensions it yiel ds a bit longer expression. c D c D c D J z J y J x J t c2) ( ) ( (1-23) 2 2 2 2 2 2 2z y x (1-24) There are several conditions where the solution of Fick’s laws provides general and frequently used solutions. In a limited s ource diffusion, assuming an instantaneous source depositing a fixed amount Q of impurity atoms on the su rface of otherwise pristine silicon, the diffused profile would be in the form of Dt x Dt Q t x c4 exp ) (2 (1-25) Another interesting case is diffusion from a constant source, similar to predeposition step, when impurities are introduced into silicon while maintaining constant surface concentration cS. The constant surface concentration is generally maintained through abundant source, such as polysilicon or oxide doped above solid solubility at the predeposition temperature. The diffusing part of the dopant concentration would then be at the solid so lubility. The solution of Fi ck’s laws with these boundary conditions results in comp lementary error function, Dt x erfc c t x cS2 (1-26) A more common case in semiconductor proces sing is the combination of implant and anneal. Presuming a dose Q of a dopant was implanted with an energy E with a mean projected depth Rp with vertical straggle Rp, the distribution after an implant would be

PAGE 27

13 2 22 ) ( exp 2 ) (p p pR R x R Q x c (1-27) Following anneal at temperature T with diffusivity D for a time t the distribution would be Dt R R x Dt R Q x cp p p4 2 ) ( exp 2 2 ) (2 2 2 (1-28) provided the profile was suffici ently far away from the surf ace, satisfying the following: Dt R Rp p4 22 2 (1-29)

PAGE 28

14 Figure 1-1. Triode electr on tube schematic [Psi06]

PAGE 29

15 Figure 1-2. Number of components per inte grated circuit vs. tim e, later known as “Moore’s law” [Moo65]

PAGE 30

16 Figure 1-3. Number of transist ors vs. time, shown to realize the original prediction of G. E. Moore, as well as updated trends [Moo03]

PAGE 31

17 Figure 1-4 Schematic cross-sec tion of a nMOSFET transistor

PAGE 32

18 Figure 1-5. Schematic of an ion implanter [Imp06]

PAGE 33

19 Figure 1-6. Illustration of an ion implantation process, in pa rticular (a) a single ion path and (b) the resulting damage cascade [Wil84]

PAGE 34

20 Figure 1.7 Silicon crystal viewed from (a) <110> direction and (b) tilted ~10o off the <110> direction [May70]

PAGE 35

21 Figure 1.8. Illustration of fluxes entering and exiting a given volume

PAGE 36

22 CHAPTER 2 LITERATURE REVIEW 2.1 Dopant Diffusion in Silicon This section outlines some significant quantities influencing the diffusion of a given dopant species in silicon. Diffusion in the silicon crystal, according to the standing theory [Fah89a], occurs through four mechanisms. Namely: interstitia lcy and kick-out mechanism, which are interstitially mediated, vacan cy mediated and concerted exchange. The former three require a point defect, an interstitial or a v acancy, respectively. Point defects interact with the dopant atom to form a dopant-defect pair These pairs then diffuse by a random walk mechanism to generate net dopant diffusion. The kick-out mechanism involves an interstitial that does not form a pair with the dopant atom. Rather, the dopant atom is kicked out into the void, becomes an interst itial and diffuses through the channels in the crystal lattice. Eventually the dopant atom kicks out a silicon atom forming a silicon interstitial and occupies its lattice position. The fourth mechanism, concerted exchange, does not require point defect presence. Rather it consists of two atoms, a dopant atom and a silicon atom, simultaneously exchanging po sitions. It is frequently neglected, as theoretical studies suggest the activation energy for concerted exchange would be prohibitively high [Nic89], compared to poi nt defect mediated mechanisms. Recent experiments attempted to estimate the con certed exchange fractional contribution in diffusion [Ura99a]. The results of thos e experiments were somewhat ambiguous,

PAGE 37

23 allowing for the existence of concerted exch ange mechanism, but also allowing for its fraction contribution of zero. These four diffu sion mechanisms are illustrated in Fig. 2-1. AX X A (2-1) ,b AX f X f AXE E E (2-2) m AX b AX f X m AX f AX AE E E E E E (2-3) Equation 2-1 describes the formation of a dopant-defect pair. The dopant is denoted as A and the defect as X which can be a silicon interstitial, I or a vacancy, V Formation of a dopant-defect pair, AX leads to a lower energy state, as given in the Eq. 2-2. Breaking of the pair requires a binding energy, b AXE which is provided by the lattice. Motion of the pair occurs when a migration energy barrier, m AXE is overcome. Therefore, the dopant diffusion process via a point defect requires the formation of the defect, its reaction with a dopant atom in a substitutional site, to form a dopant-defect pair, which then migrates contributing to macroscopical ly observable diffusion behaviour. Activation energy for diffusion, in that case, can be writ ten as a function of the formation energy of the defect X f XE the binding energy of the dopant-defect pait AX b AXE and the migration energy of a dopant defect pair AX m AXE as shown in the Eq. 2-3. In single crystal silicon, those energi es have fixed values. In Si1-xGex alloys, those values can vary with Ge content. For a complete picture, one should have information on both defect and dopant pair diffusion, with resp ect to Ge content. Also, Ge diffusion could be a function of Ge content, and should be investigated to allow simulation of spatially varying Ge profiles. The experimental data available in the literature will be discussed later in this chapter.

PAGE 38

24 2.2 Self-diffusion in Silicon Despite numerous theoretical and experiment al studies on self-diffusion in silicon, contributions from interstitial and vacancy components are uncertain. Radioactive tracer technique results are limite d by a 2.6 hour lifetime of 31Si, so monitoring heavy metal diffusion is frequently used in attempting to extract self-diffusion parameters. With the availability of the isotopically pure silic on sources, structures could be grown (by molecular beam epitaxy (MBE) or chemical vapor deposition (CVD)) with isotopically enriched or depleted layers. This appr oach requires no assumptions in measuring prefactors and activation energies for interstitial and vacancy diffusion. The study of Bracht et al. [Bra98] measured se lf-diffusion of Si in an inert ambient. The samples were sealed in Ar filled capsule s and annealed at temperatures ranging from 855oC to 1388oC. The extracted diffusivity DSi (Eq. 2-4) contains both interstitial and vacancy diffusion components. Taking interstiti al diffusivity measured in a Zn diffusion experiment [Bra95] as CI *DI (Eq. 2-5), they extracted CV *DV (Eq. 2-6). s cm T k eV DB Si 2 250 17004 0 75 4 exp 530 (2-4) s cm T k eV D CB I I 2 *95 4 exp 2980 (2-5) s cm T k eV D CB V V 2 *) 14 4 exp( 92 0 (2-6) Ural et al. [Ura98] performed a sim ilar study measuring self-diffusion through diffusion of 30Si from naturally abunda nt, into isotopically depleted layers. The natural abundance of 3.1% was reduced to 0.002% in isotopically depleted layers. During the annealing, samples were subjected to inert, oxidizing and nitridizing ambient. Oxidizing ambient injects silicon inters titials [Hu74], while nitridizing ambient injects vacancies

PAGE 39

25 [Fah85] from the surface. This perturbation of the point defe ct concentrations allows the extraction of the fractional interstitial, fI and vacancy, fV contributions to self-diffusion, without assumptions. Self-diffusivity measured in inert ambient was DSi (Eq. 2-7), matching the values of Brach t et al. [Bra98]. Extracted fI was in a range of 0.550.6(0.1), while interstitial a nd vacancy diffusivities were DSiI (Eq. 2-8) and DSiV (Eq. 2-9), respectively. s cm T k eV DB Si 276 4 exp 560 (2-7) s cm T k eV DB SiI 268 4 exp 149 (2-8) s cm T k eV DB SiV 286 4 exp 636 (2-9) 2.3 Germanium Diffusion in Si1-xGex Determining Ge diffusivity with respect to Ge content is interesting, particularly for a spatially varying Ge profile, such as base of a BJT or 2D hole gas (2DHG) transistors. Zangenberg et al. [Zan01] m easured Ge diffusion in Si1-xGex (x<0.5), MBE grown layers. Diffusion was measured using 70Ge as a tracer into 72Ge rich layers, during anneals at temperatures from 850 to 1050 oC. Transmission electron microscopy was used to confirm that the dislocation density was low enough not to influence diffusivity measurement. The diffusivities measured are shown in Fig. 2-5, with the extraction of prefactor and activation energies in Fig 2-6. A more comprehensive study was performe d by Strohm et al. [Str02]. Various material types used include MBE, CVD, Cz and FZ grown materi al, collected from a number of academic and industrial sources. 31Si and 71Ge isotopes were used as tracers, introduced into the material by ion implantation. Tracer dose below the threshold for formation of extended defects ensured neglig ible influence of implantation process.

PAGE 40

26 Transient electron microscopy (TEM) and etch pit measurements verified that the dislocation density was under 107 cm-2. The samples were ann ealed at temperatures of 650 to 1250 oC in inert ambient. Their extracted pr efactors and activation energies for Ge diffusion are shown in Fig. 2-7. Despite the difference in prefactors and activation energies, the diffusivities agree quite well to the values provided by Zangenberg, as shown in Fig. 2-8. Large number of samples having different Ge concentrations provided a rather small stepping in Ge content. It is interesti ng that such variation of sources produced such a nicely behaving plot. A distin ct change in the prefactor dependence with increasing Ge content ocurrs around 35% Ge, not as visibl e in the activation en ergy dependence. The authors attribute this to the change in the diffusion mechanism, from interstitial to vacancy dominated, as one increases Ge content. The situation is not as clear when it co mes to fractional interstitial and vacancy contributions to Ge diffusion. Initial st udies of Fahey et al. [Fah89b] estimated fI and fV using ammonia ambient anneals. Ammonia (NH3) annealing ambient, reacting with a bare silicon surface, injects vacancies into the bulk [Fah85][Kri96][Kri97][Mog96]. The same ambient, reacting with oxide (SiO2) injects interstitials. Reference was provided by a sample coated in oxide and nitride, ensu ring no surface reaction, and no point defect injection. Samples contained a Ge marker layer grown by MBE. Peak Ge concentration was ~5%. Anneals conducted under vacancy an d interstitial inject ion both diffused the Ge profile more than the referen ce sample. At a temperature of 1050oC, they estimated fI = 0.3-0.4. Unfortunately, data for ot her temperatures was not provided.

PAGE 41

27 Studies of Griglione et al. [Gri00, Gri01] concentrated on extraction of fI and fV by annealing Si0.85Ge0.15 layers in inert and oxidizing am bient at temperatures ranging from 900 to 1200 oC. Diffusion of buried B marker layer was monitored to verify the oxidizing reaction on the surface injected interstitia ls and estimate interstitial diffusivity enhancement. The enhanced diffusion of the bur ied B marker layer also showed that the Si0.85Ge0.15 layer does not significantly perturb inte rstitial population, meaning there was no pile-up of silicon interstitials in GeI pa ir. Since the oxidizing ambient resulted in reduced diffusion (Fig 2-9.) of Ge from the Si0.85Ge0.15 layer, it was inferred that the diffusion was vacancy mediated. To verify th is, anneals in nitridizing ambient were performed at temperatures of 1100 and 1200 oC. Instead of the expected increase in diffusivity, a considerable reduction in diffus ivity was observed, thus making it difficult to draw a firm conclusion on the available data. 2.4 Boron Diffusion in Silicon 2.4.1 Boron as Interstitial Diffuser in Silicon Boron is generally accepted to be an interstitially diffusing species [See68][Fah85][Fan96][Gos97]. More recent measurements [U ra99b] of B diffusion in interstitial and vacancy supersaturation, gave fI values higher than 0.84, assuming the vacancy concentration was unperturbed. A mo re strict assumption on the recombination of equilibrium point defects 1 1 eq I nit I nit V eq V eq V ox V ox I eq IC C C C C C C C yields fI values above 0.94. The exact mechanism of B microscopic diffusi on is not certain, between interstitial pair diffusion and kick-out mechanism. Recent ab-initio calculations [Win99] suggest B diffusion is mediated by a BI pair. The pair binding energy was calculated to be 0.8eV, while the migration energy was 0.2 eV. Accord ing to the Eq. 2.5, the activation energy

PAGE 42

28 for B diffusion would also depend on the forma tion energy of an interstitial. Since the values of point defect formation and migra tion energies used in FLOOPS are somewhat different than the ones used in this study, a free parameter is necessary to match the macroscopically observed activation energy. In this case, the binding energy of a BI pair is used as a free parameter. The value of 0.5 eV was used to match the B diffusivity values of FLOOPS defaults. 2.4.2 Pre-amorphization and B B implants into crystalline silicon materi al seldom provide the optimal results, due to channeling, solid solubility limit and the d eactivation to form boroninterstitial clusters (BICs). Low energy implants require implantation in deceleration mode in order to provide viable implant currents. Even then, th ere is a possibility of energy contamination due to the neutralization of the charged speci es before reaching the decelerating stage. Once the B ion enters the silicon lattice, a certain portion of the im planted dose channels, increasing the pn junction depth. Increasing the dose to reduce the parasitic resistances can result in BIC formation, thus decreasing active B concentration below solid solubility limit. Some of these detrimental effects can be mitigated by pre-amorphizing the silicon crystal. The disordered structure of the am orphous silicon prevents channeling [Zie00b], thus reducing the pn junction depth. Activation can also benefit from the preamorphization, as the amorphous layers ca n be regrown at low temperatures (450600oC) incorporating dopant concen trations higher than solid solubility at the regrowth temperature. The spatial separation of the end of range (EOR) damage and the dopant suppresses the formation of BICs [Jon96a ]. Germanium is the popular choice of

PAGE 43

29 amorphizing implant species due to producing le ss damage than the silicon implant for the same implant dose [Cla02]. 2.5 Boron-interstitial Clusters Dopants are mostly introduced into a Si wafer via implantation process, which is followed by an annealing process. During the thermal process, the damaged crystalline lattice is repaired and the dopants fall in to the substitutional positions, thus becoming electrically active. However, the annealing process follow ing a B implant under certain conditions, such as yielding a peak concentration above 1x1019 cm-3, does not activate all the dopants. The implanted profile splits into two regions, the i mmobile peak and the rapidly diffusing tail, as first observed by Michel et al. [Mic87]. The immobile portion of the profile is particularly interesting as it is at a concentration much lower than the solid solubility of B in Si, above 4x1019 cm-3 for temperatures over 800oC [Vic69]. In the comparison between the anneal ed profiles at 800 and 950 oC, immobile peak is far more pronounced at the lower temperature (Fig. 2-10) The same is the case with the diffusion in the tail regions, as the transient enhanced diffusion (TED) is reverse activated, having higher diffusion enhancement at lower temperatures. Stolk et al. [Sto95] investigated the fo rmation of the immobile peak portion, by implanting Si into a wafer containing B -doped layers. The Si 5x1013 cm-2 implant at 40 keV contained the damage into the top 0.1 m layer, directly aff ecting only the B layer marked as 1 (Fig. 2-11). The annealed B profile s show several interes ting characteristics. First and second -doped layers have signifi cantly diffused tails, indi cating the interstitial supersaturation as expected after a Si implant and consequential TED. But, these -doped

PAGE 44

30 layers also have immobile peaks, indicating th at the majority of B in them is inactive. The shouldering concentration (the intersection of the immobile peak and he diffused tail) of the 1st and 2nd -doped layer is significantly lower than B solid solubility at the anneal temperature, indicating that some mechanism other than the solubility limit is preventing B diffusion. The 3rd -doped layer exhibits lower diffu sion enhancement than the first two, but it does not have a visible immobile portion. All the other -doped layers diffuse closer to the equilibrium diffusivities. This experiment shows that the interaction between self-interstitials and substitutional B can form immobile peaks. Furthermore, it shows the excess interstitials from the implant damage can diffuse and interact with substitiutional B forming metastable clusters visible as immobile B peak s after an anneal. As these clusters are the product of the interaction of self-interstitials and boron, thus called boron-interstitial clusters. To probe the dissolution kine tics of BICs, Lilak et al [Lil02] used thirteen B implant conditions. The samples were annealed at 750oC for 30 minutes in order to deactivate the majority of the B dose. Then the samples were followed through a number of times during anneals at 750 or 850 oC, using the Hall effect measurement and spreading resistance profiling (SRP) to meas ure activation. Some measurements are shown in the figure 2-12. These measurements indicate that the increase in dose with constant implant energy, as well as the decreasing implant energy with constant dose, decrease the activation upon annealing at temperatures ~750oC. Another important parameter, the energy required for the reac tivation of B was extracted from these measurements to a value around 3eV.

PAGE 45

31 A similar experiment was done by Mirabella et al. [Mir03], using the MBE grown B marker layers. The shallowest marker layer was used to monitor the clustering process, while the deeper marker layers were used to monitor the interstitial supersaturation. These samples were implanted with 5x1013 cm-2 Si at 20keV, providing the excess interstitials for the clustering process. The active dose was measured after several anneals with temperatures in the range from 815 to 950 oC, and compared to the as-grown dose of 3x1013 cm-2. The methodology was different than th e one used in Lilak et al., as the active dose was extrapolated from SIMS meas urement. It was assumed the diffusing tails form a Gaussian distribution within the entire profile, and representative of the active fraction of the B profile (Fig. 2-13). Such an indirect approach elim inates the possibility of measuring electrically partially active cl usters, rather measuring the substitutional B concentration through its availability to partake in a diffusion process. Having determined the active fraction in a number of anneals, they extracted the activation energy for reactivation as 3.2 0.4eV, in agreement with the previous measurement of Lilak et al. The B cluste red doses and time constants used in the extraction are shown in figure 2-14. 2.6 Boron Diffusion in Si1-xGex There are several reports [Kuo93][Kuo95a][Zan01] on th e reduction of B diffusion in Si1-xGex compared to Si. The first report of reduced B diffusivity in Si1-xGex was by Kuo et al. [Kuo93]. The structure used in th e measurement is shown the Fig. 2-15. The layers were grown on <100> n-type Cz ochralski Si wafers in a CVD process. Initially, a layer of undoped Si was depos ited. On top of that a 60 nm thick Si0.83Ge0.17 layer, in-situ doped with B concentrations from 1018 to 3x1019 cm-3 in the center 20 nm.

PAGE 46

32 A silicon capping layer of 60 nm was deposited to enhance the stability of the Si0.83Ge0.17 layer. Control samples contained only th e B layer. Secondary ion mass spectrometry (SIMS) was used to extract dopant profiles. Initial profiles, as well as profiles after a 30 minute anneal at 860oC in nitrogen ambient, are shown in Fig. 2-16. At the top of Fig. 2-16 profiles for boron in the sample without the Ge layer can be seen, while the plot at the bottom indicates a B layer contained in 17% Ge layer. Square symbols are used to plot the as-grown profile s, while the plus (+) symbol demonstrates the annealed profiles. The extr acted diffusivity was found to be approximately an order of magnitude lower in Si1-xGex sample, as shown in the Fig. 2-17. Moriya et al. [Mor93] also report on the reduction of B diffusivity in Si1-xGex layers. In one experiment, 15 nm Si0.7Ge0.3 layers were grown by rapid thermal chemical vapor deposition (RTCVD), within wh ich a B peak concentration of 8x1019 cm-3 was introduced in the middle 5 nm of the Si0.7Ge0.3 layer. Ge rich layer was grown on a Si buffer layer and capped with Si layer, a ll containing a background B concentration of 1018 cm-3 to avoid electric field effect influen ce on diffusion. Si control samples only differed in the fact they contained no Ge. Th e samples were annealed in a rapid thermal annealing system (RTA) at temperatures from 850oC to 1000oC, in an inert ambient. The Si control showed no significant deviation from the accepted values for diffusivity of B, thus verifying no anomalous material effects were present. This was not the case for B diffusion in Si0.7Ge0.3 layer. The diffusivity of B in Si0.7Ge0.3 layer was significantly lower than the control, as shown in Fig. 218. Activation energy extracted was about 1 eV higher than the accepted values for B diffusion in Si.

PAGE 47

33 A separate experiment in the same publication [Mor93] tried to measure B diffusion as a function of Ge content (10%-50%). The multilayer structure was grown MBE and annealed at 975oC for 30 seconds. As Fig. 2-19. shows, increase in the Ge content reduces B diffusivity. Another experiment by Kuo et al. [Kuo95a] a ttempts to isolate the strain influence on diffusion. As Ge has 4.2% la rger lattice constant than Si, it introduces a strain into the lattice. Separating the influence of strain from the Ge content was achieved using the structure shown in Fig. 2-20. First, a grad ed buffer layer was grown on <100> p-type Si wafer was grown. The buffer layer contained linearly increasing Ge content, serving to relax the strain with a minimum of thread ing dislocations. Second, a 0.25um relaxed Si1-yGey layer was grown to serve as a substrat e for a pseudomorphic diffusion structures. Third, pseudomorphic 60 nm thick Si1-xGex layer was grown, containing a 20 nm B layer. Fourth and last, the capping 40 nm Si1-yGey layer was grown to ensure thermal stability. Having an independent control over x y Ge contents, one can engineer the strain for a given x percent Ge. In order to investigate the dependence of di ffusion irrespective of strain, x has to be equal to y Another application could be to investigate the diffusion as a function of Ge content, with a constant strain. That can be accomplished for x=y+z where z is any Ge content delivering the desire d strain. Some of these combinations are shown in the Fig. 2-21. There are several things to note. First, di ffusivity around 0% strain is a function of Ge content. Second, strain dependence for 10% and 20% Ge is rather weak, indicating that the strain is not the do minant component in reduction of B diffusivity with increasing Ge content. Rather, it would seem the chemical effect of Ge is more significant. It is

PAGE 48

34 interesting that the Ge presence reduces the B diffusivitiy, particularly since the self-diffusion of Ge and Si increase with Ge content in Si1-xGex alloys [Str02]. Presuming that B diffusion remains interstitially mediated in Si1-xGex (confirmed for 10% [Wil01] and 18% Ge [Kuo95b]), and that the properties of the BI pair remain unchanged with Ge content, the increase in self-diffusion should increase the effective B diffusivity. Similar structure (Fig. 2-22) to that of Kuo et al. [Kuo95a] was used in an experiment of Zangenberg et al. [Zan03] to investigate the diffusi on of B and P in MBE grown Si1-xGex. The Si1-xGex layers were grown on the (100) Czochralski Si wafers, using the graded buffer layer to prevent the di slocation growth towards the surface. TEM verified that samples before and after annealing had dislocat ion density under 106 cm-2, and that no dislocations formed at the ps eudomorphic layer interfaces, confirming that dislocations had minimal influence on dopant di ffusion. The samples used in the study of B diffusion were annealed at temperatures between 800 and 925 oC, with Ge contents of 0%, 1%, 12%, and 24%. The experimental results we re quite unexpected. Desp ite the previous body of literature on reduction of B diffusion in Si1-xGex alloys, the study found a small difference between B diffusion in relaxed Si1-xGex and the control samples. The extracted parameters for B diffusion in relaxed Si1-xGex are shown in Table 2-1. Even though the measured control diffusivities fit the certain m easured values for B diff usion in Si in that temperature range, the extracted activation ener gy of 2.68 eV is quite different from the values of Fair (3.46 eV) [Fai81] or Haddara et al. (3.75 eV) [Had00]. The other interesting point in this measurement is the change of the B activation energy with increasing Ge content. The addition of 1% Ge in the relaxed Si1-xGex alloy

PAGE 49

35 increases the activation energy of diffu sion by ~0.5eV, which does not change significantly in higher Ge cont ents, as demonstrated in Si0.88Ge0.12 and Si0.76Ge0.24. These values would suggest, provided the properties of the BI pair and self-i nterstitial diffusion were unchanged by Ge content, that the addi tion of Ge adds anot her potential well of about 0.5eV to the B diffusion mechanism. Having reviewed all of these experi mental data of B diffusion in Si1-xGex, one can notice there is no clean study of B diffusion with a numbe r of samples over various temperatures, like Strohm et al. [Str02] for Ge self-diffusion. The closest to the detailed measurement of B diffusion in relaxed Si1-xGex is the experiment of Zangenberg et al. [Zan03] with a temp erature range of 125oC and a problematic c ontrol. Most studies concentrated on one temperature, or one time, or one Ge content. 2.7 Modeling Boron Diffusion in Si1-xGex The first B diffusion model to be discusse d was published by Lever et al [Lev98]. In their experiment B layer was sandwiched between two Ge layers, both of 10% or 3% Ge content. One set of Lever’s samples wa s grown via low pressure CVD (LPCVD) on <100> FZ silicon wafers. They used B wells with concentrations of 3x1018, 6x1018, and 1.5x1019 cm-3. Initially, a 100 nm thick silicon buff er layer was grown on a wafer. It was followed by growth of undoped 40 nm thick Si0.9Ge0.1 layer, a 250nm thick B well, another 40 nm, and finally capped with 100 nm undoped Si. In another set of Lever’s samples, a silicon buffer layer was grown on a Czochralski silicon wafer. It was followed by a 80 nm thick undoped Si0.97Ge0.03 layer, 200 nm thick B well with concentration of 1019 or 4x1019 cm-3, another 80 nm Si0.97Ge0.03 layer, and capped with a silicon layer. All samples were capped with 40 nm SiO2 and 160 nm Si3N4, to ensure the inert surface

PAGE 50

36 condition. Anneals were performed at 850oC for times of 4, 24 or 96 hours. Fig. 2-23 shows some of the resulting profiles. In their simulations, Lever et al. considered B diffusion as perturbed by Ge content. This simplification disregards possibility that B activation depends on Ge content. Influence of strain was also neglected, as study by Kuo et al [Kuo95a] showed it to be small. Change in bandgap was considered to be linear with Ge content. Vacancy mediated diffusion of B was al so neglected, which also limits this model to lower Ge content. Higher Ge content might have si gnificant vacancy self-diffusion component, which could influence B diffusion, as well as fI. The B flux equation (Eq. 2-10) consists of th ree terms. The first term is Fick’s law term for diffusion of boron, denoted as CB. Second term is impurity related electric field diffusivity enhancement. Third term is resu lt of energy bandgap variation due to Ge content. Diffusivity is descri bed as sum diffusivity coming fr om interaction with neutral, D1 and ionized, D2 point defects (Eq 2-11), w ith factor beta as ratio of those two. The last equation (Eq 2-12) is the Arrhenius expr ession for diffusivity as a function of temperature. x n p n n c x N N p n c x c D Ji B D A B B B Bln 20 (2-10) 1 2 2 1 2 1 0, 1 1 D D n p D D n p D D Di i B (2-11) kT k DB5 3 exp0 (2-12)

PAGE 51

37 Ge B act B Ge act B Bc s s c c s c c c _1 (2-13) The pairing between Ge and B is assumed to form a GeB cluster, immobile, but electrically active. The rest of B is available for diffusion ( CB_act), as described by Eq. 2-13. The fitting parameters of the model are b k and s Although the parameters have been individually fitted for each annealing condition, they do not vary significantly. The good features of this model include: simplicit y, few parameters, and good fit to several Ge contents with the same value of s This model has been implemented in FLOOPS, and used as an initial guess in experiment design.

PAGE 52

38 Table 2-1. Extracted B diffusivity prefactor ( D0) and activation energy ( Ea) from experiment of Zangenberg et al. [Zan03] Ge [%] Temp range [oC] D0 [cm2/s] Ea [eV] 0 800-900 3.4(2.3)x10-4 2.68(7) 1 800-925 3.4(2.0)x10-2 3.13(6) 12 800-925 2.4(2.4)x10-1 3.30(10) 24 800-925 5.7(7.0)x10-2 3.18(13)

PAGE 53

39 a) b) c) d) Figure 2-1. Diffusion mechanisms in silicon: a) interstitialcy diffusion, b) kick-out, c) vacancy diffusion, and d) concerted exchange [Fah89a]

PAGE 54

40 Figure 2-2. Self-diffusion in Si as measur ed by Bracht [Bra98], w ith symbols denoting different samples

PAGE 55

41 Figure 2-3. Self-diffusion in Si as measured by Ural et al. [Ura 99b]. Dots represent measured diffusivities, solid line is th e best fit expression for self-diffusion. Dashed lines represent bounds of self-d iffusion as reported by Bracht [Bra98], showing good agreement between measurements.

PAGE 56

42 Figure 2-4. Interstital fraction, fSiI of the self-diffusion from Ural et al. [Ura99b]. The lines represent predictions from se veral metal diffusion experiments.

PAGE 57

43 Figure 2-5. Germanium diffusivities for various temperatures and Ge contents, after Zangenberg et al [Zan01]

PAGE 58

44 Figure 2-6. Activation energies and prefactors for various Ge contents, after Zangenberg

PAGE 59

45 Figure 2-7. Activation energies and prefactors for various Ge contents, after Strohm

PAGE 60

46 Figure 2-8. Comparing diffu sivities of Ge in Si1-xGex as measured by Zangenberg and Strohm

PAGE 61

47 Figure 2-9. Comparison of Ge diffusion with respect to ambient and consequential interstitial or vacancy supersaturation [Gri01]

PAGE 62

48 a) b) Figure 2-10. B profiles after an implant of 2x1014 at 60keV, annealed at a) 800 oC and b) 950oC [Mic87]

PAGE 63

49 Figure 2-11. B profiles, as -grown and after 1h at 670oC [Sto95]

PAGE 64

50 a) b) Figure 2-12. Fractional activation and sheet resistance of samples annealed at 750oC for: a) varying B dose at 80keV, and b) 4x1014 cm-2 at varying energy [Lil02]

PAGE 65

51 Figure 2-13. B profiles used in experiment by Mirabella et al. showing the as-grown, annealed for 2 min at 815oC, as well as illustrating the methodology for determining the active fraction [Mir03]

PAGE 66

52 Figure 2-14. Time evolution of the cluste red B dose during annealing at 815 to 950 oC, with extracted time constants. [Mir03]

PAGE 67

53 Figure 2-15. Structure used by Kuo et al [Kuo93] with a B marker layer in the Si0.83Ge0.17 strained layer

PAGE 68

54 Figure 2-16. Diffused B profiles after 30 minute anneal at 860oC in a) Si and b) Si0.83Ge0.17 [Kuo93]

PAGE 69

55 Figure 2-17. Effective B diffusivity as a function of B concentration [Kuo93]

PAGE 70

56 Figure 2-18. Diffusivity of B in Si0.7Ge0.3 layer [Mor93]

PAGE 71

57 Figure 2-19. B profiles in multiplayer structure before (dashed line) and after (solid line) annealing at 975oC [Mor93]

PAGE 72

58 Figure 2-20. Illustration of a structure us ed by Kuo et al. [Kuo95a] showing relaxed Si1-yGey and pseudomorphic Si1-xGex

PAGE 73

59 Figure 2-21. B diffusivity at 800oC as a function of strain, with the numbers in the parenthesis are ( x,y ) Ge content. Positive strain represents biaxial tension and negative strain biax ial compression. [Kuo95a]

PAGE 74

60 Figure 2-22. Schematic description of the stru cture used by Zangenberg et al. [Zan03] to measure B diffusion in Si1-xGex

PAGE 75

61 Figure 2-23. B profiles before (t hin line) and after (thick) 850oC annealing of a given B well concentration. Figure a) B we ll with a concentration of 3x1018 cm-3, sandwiched in Si0.9Ge0.1 layers, annealed for 96 hour s. Figure b) B well with a concentration of 1.5x1019 cm-3, sandwiched in Si0.9Ge0.1 layers, annealed for 96 hours. Figure c) B well wi th a concentration of 4x1019 cm-3, sandwiched in Si0.97Ge0.03 layers, annealed for 24 hours. [Lev98]

PAGE 76

62 CHAPTER 3 BORON CLUSTERING IN SILICON Overview This chapter discusses the behavior of boron-interstitial clusters (BIC) and its influence on activation of B in silicon. Experimental evidence shows that the previous model and its behavior with respect to a nnealing ambient do not capture the observed cluster dissolution phenomena. The most r ecent set of ab-initio calculated formation energies is utilized in attempt to capture this phenomenon, which is achieved with certain changes within the ab-initio energetics. The cluster formation is examined in the experiment of Jones et al. [Jon96a], which indicates that the formation is a rapid process and is completely prevented by preamorphization. Assuming the inability of si licon interstitials entering the amorphous layer significantly reduces the percentage of boron tied in the clusters. Additional, small changes of the cluster properties further redu ce the formation probability of the clusters, thus matching the experimentally observed behavior. 3.1 Boron-interstitial Cluster Dissolution 3.1.1 Introduction Ion implantation is the preferred met hod of introducing dopants due to the precision of implanted dose, control and repeatability. Future technology nodes will require highly doped shallow junctions pr ocessed with a low thermal budget. Such conditions emphasize the role of point def ects generated by the implantation process. Recombination of point defects and dopant activation take place during annealing post

PAGE 77

63 implantation. For nonamorphizing implants, the annealing step can fo rm extended defects and dislocation loops, which are responsib le for enhanced diffusion [Eag94][Jon96b]. Besides effects on dopant diffusion, high inters titial concentrations are also known to deactivate boron, through the formation of boron-interstitial clus ters (BIC)[Sto95]. Ab initio studies provided information on the stability of some clusters, used in several implementations of BIC models in Monte Carlo [Cat98] a nd continuum process simulators [Pel97][Lil97]. It is unclear if boron will cluster in an interstitial supersaturation lower than that found in i on implantation, and if measurement methods would be able to distinguish such a presum ably low concentration of clustered boron from the noise of the active boron profile. Sp ecifically, characterization of low cluster densities is difficult, as low cluster densities are not detectable as spikes in the profile acquired by secondary ion mass spectrometr y (SIMS) profiles and could be below detection limit of spreading resistance or Hall effect measurements. The behavior of BICs during dissolution, with respec t to interstitial supersatura tion, has not been verified experimentally. The model proposed by Pel az [Pel99] suggests that break up of the clusters requires an interstit ial to drive the pro cess along an in terstitial lean path. This theory says that the boron is driven to clus ter in a large interstitia l supersaturation, but a lower supersaturation might provide excess interstitials that will speed cluster dissolution and boron reactivation. Notwithst anding, experimental evidence suggests that BICs can serve as a source of interstitials for transi ent enhanced diffusion ( TED), by dissolution of boron clusters in later stag es of annealing [Sol00]. To test the theory of BIC cluster dissolu tion through interstitial lean path, implants of boron were performed to achieve peak boron concentration on the order of 1019 cm-3.

PAGE 78

64 Annealing is performed in two steps. Boron in terstitial clusters form during first anneal step and deactivate a significant part of the dose. The second anneal is used to determine the influence of annealing ambient on cluste r formation or dissolution. Clustering effects are investigated via SIMS and Hall effect measurements. 3.1.2 Experimental Czochralski <100> n-type silicon wafers were implanted with boron. Three implant conditions with nominal boron peak concentrations of 1019 cm-3 were investigated. Implant doses and energies were 1x1014 cm-2 at 5 keV, 2x1014 cm-2 at 10 keV and 4x1014 cm-2 at 20keV, respectively. The first anneal step was performed at 750oC for 30 minutes in inert N2 ambient. To investigat e the influence of ambien t on B cluster dissolution, subsequent anneals were done in inert (N2 flow) or oxidizing (dry O2) ambient at 850oC for 10, 20, 30 and 60 minutes. Oxidation of silico n is known to inject interstitials [Hu74], allowing investigation of BICs in a low supersaturated interstitial environment. Samples were analyzed by SIMS, Hall e ffect and ellipsometry measurements. Boron depth profiles were measured by CAMECA IMS-3f instrument using O2+ primary ion beam and magnetic sector analyzer. Raster size and analyzed area were 200 m and 60 m in diameter, respectively. Samples were biased at 4.5 kV with an effective impact energy of 5.5 keV. The Woollam EC110 ellipsometer was used for oxide thickness measurements, showing no difference between initial condition and inert annealed samples. Oxide thickness measurement of samples annealed in oxidizing ambient was used in aligning SIMS profiles, due to th e consumption of silicon. Measured oxide thickness agreed well with standard predicti ons at these temperatures. The active dose was measured using a MMR Technologies Inc. Hall-van der Pauw system. Measurements

PAGE 79

65 were performed using a magnetic field of 300 mT with the current ranging from 1 A to 1 mA at room temperature. Square shaped samples with 14 mm sides had 1 mm diameter e-beam deposited Al contacts placed symmetrica lly near the corners. Factors contributing to the inaccuracy of the measurement include nonuniform dopant distribution and heavy to light hole ratio. Assuming measured pr ofiles were steep enough to neglect the influence of deeper layers with lower dopant concentration and highe r mobility, we chose the value of 0.7 for the Hall mobility factor [Sas88]. 3.1.3 Results and Discussion First annealing step provides the initia l condition with BICs of similar peak concentrations for studied implant conditi ons (Fig. 3-1). During this anneal excess interstitials generated by the implantation process interact with boron forming BICs. These clustered boron samples ar e referred to as the initial condition from this point on. Fig. 3-2 shows boron profiles at critical stages of thermal processing for each of the studied implant conditions. Comparison of oxi dized and inert annealed profiles show oxidation enhanced diffusion for the boron as exp ected for an interstitial diffusing species [Pac90]. Direct comparison of the clustered fr action is difficult from the SIMS results because of the extra diffusion. Although the shallowest implant exhibits the highest diffusion enhancement, we do not believe the distance to the surface to be significant. Different diffusion enhancements may be a c onsequence of a difference in the initial clustered dose, as oxidation is known to create enhancements uniformly deep in the bulk [Gri85]. The surface influence in these sa mples is mostly through segregation and outdiffusion of boron, resulting in dose loss (Tab le I). Dose loss is most apparent in

PAGE 80

66 lowest implant dose/energy sample, since it ha s the shallowest boron profile in presented sample set (Fig. 3-2c). Hall effect measurements dur ing various stages of ann ealing are shown in Fig. 3-3. The initial anneal deactivates a large fracti on of the dose, creating the initial clustered concentration that establishes our initial c ondition. Subsequent annealing activates the boron by dissolving the clusters, with inert anneals consistently having a higher active dose than oxidizing anneals. Considering dose loss and reactivation effects are mixed in the Hall effect measurement, defining activ ation as ratio of act ive to retained dose (integrated SIMS profile) is used to distinguish them. Activation fractions for known boron profiles are presented in Fig. 3-4. Increase of activation by 15-40% for inert annealed over oxidized annealed samples c onfirms the trend demonstrated by the Hall effect measurement. That leads to the conc lusion that boron clusters dissolve slower in the oxidizing ambient than in the inert ambient. Provided that one of the do minant BICs has a self-int erstitial release reaction during dissolution, its dissoluti on rate would be diminished or annulled with the increase in the interstitial concentrat ion. Since the oxidation induced interstitia l injection reduces reactivation (Fig. 3-3) when compared to th e inert ambient anneal, which leads to the conclusion that one of the significant clusters has an interstitial release reaction during dissolution. Examining the time dependence of the r eactivation shows disagreement with the idea of one dominant BIC. In the course of the second stage anneal, one can observe the initial increase in activation in almost all samples irrespective of the annealing ambient. After the initial increase, active dose in oxi dizing ambient stays constant, while inert

PAGE 81

67 ambient continues to reactivate B at longe r anneal times. Supposing the reactivation processes occurring under oxidati on are also occurring in iner t ambient as well, one could consider inert reactivation as a superposition of oxidizing and an additional long-term reactivation. This would indi cate there are at least two clusters holding a significant B dose. The one responsible for the short time reactivation would be ra ther insensitive to the ambient. On the other hand, the long-te rm reactivation would be coming from a cluster whose dissolution is aff ected by interstitial injection. 3.1.4 Modeling In attempt to model the oxidation behavi or we utilized the energetics published by Liu et al. [Liu00] The model is impl emented in model description language ALAGATOR, assuming diffusion limited reacti on rates, and simulated using the process simulator FLOOPS [FLO02]. Implanted ion and damage profiles are generated by UT-Marlowe [UTM00]. Boron clustering and cl uster dissolution paths are shown in Fig. 3-5a. The clustering process is different from previously described model [Pel99]. During high interstitial supersaturation, BICs grow towards B and SiInt rich, which later release interstitials to form B rich, SiInt poor clusters. However, there are two clusters, B3I and B2I3, containing the majority of the B dose, as opposed to exclusively B4I [Pel99]. Regarding the dissolution path s, they would indicate oppos ite reaction to oxidizing ambient than seen in experiment. Oxidation injected interstitials would be captured by B3I enhancing the transition towards B3I2. This is the rate limiting step, since the migration energy of a BI is lower than that of SiInt, and B3I2 can subsequently dissolve with the release of BI solely. Simulation of the two-step anneal used in the experiment resulted in Fig. 3-5b and shows the simulated activation during second anneal step. One can observe a good fit to

PAGE 82

68 clustered condition after the first anneal ste p. The reactivation during second anneal step is rather insensitive to oxidizing ambient. This is the consequence of similar B2I3 and B3I doses after the first anneal step, as B2I3 can provide more than sufficient number of interstitials available for B3I reactivation. In order for a given BIC to dissolve slower under oxidation, it would have to release an interstitial along its path to B s ubstitutional. Then the interstitial injection would be able to slow down that reaction, as observed in experiment. Therefore, we propose that the ambient sensitive BIC, BmIn, would have at least as many Si interstitials as B atoms (n m). We modified the energetics in an attempt to model the oxidation behavior using B4I4 and B2I3 as the two dominant clusters. The resulting cluster formation and dissolution paths using the modified energetics are shown in Fig. 3-6a. In this proposed model, B2I3 provides the reactivati on at short times (less than 10 minutes) and is insensitive to oxida tion. On the other hand, B4I4 dissolution is aff ected by interstitial injection and provides reactivat ion at longer times. Simulation of the two-step anneal resulted in active dose plot shown in Fig. 3-6b. One can observe a good match to the initial clustered condi tion. The model also shows appropr iate behavior under oxidation, which is the first such model according to th e literature. Two-stage reactivation is also visible, as described previously. 3.2 Boron Cluster Formation and Preamorphization 3.2.1 Experiment and Findings One of the frequently applied methods for increasing activation of B implants is preamorphization. There are multiple benefits and some disadvantages to this method. Important benefits include: reduction in boron clustering, hi gh activation at very low temperatures during the solid phase regrow th, and reduction of the junction depth by

PAGE 83

69 reduction of the channeling tail. The main disadvantage is the formation of extended defects in the EOR region, which is detrimen tal to the electrical characteristics of the pn junction. In an investigation of B clustering Jone s et al. [Jon96a] implanted a Si dose of 5x1015 cm-2 at energy of 146 keV, at liquid ni trogen temperatures. The implanted material contained three B marker layers, grown via MBE, as seen in Fig. 3-7. The Si implant at 146 keV produced an amorphous layer 0.324 m thick, having the damage peak concentration coincide with the middle marker laye r. Upon annealing at 800oC, the excess interstitials in the damaged re gion interact with B in the marker layers beneath the amorphous/crystalline ( a/c ) interface. This can be observed at longer annealing times as the immobile B spikes, at the position of the B peak, as seen in Fig. 3-8. The center B marker layer is almost co mpletely immobile dur ing the 3 minutes of annealing. At that time the shallowest peak is diffused out showing no signs of clustering, while the deepest marker layer exhibits an apparent clustering peak. The difference between the outer two peaks is interesting si nce they are almost the same distance from the a/c interface. Assuming the damage is concen trated in the region just beneath the a/c interface, the shallowest and the deepest marker layers should observe the same interstitial supersaturations. Th is assumption is partially validated from the similarity in the observed profile broadening shown in Fig. 3-8. Although the deepest marker layer has higher peak concentrations dur ing the anneal when compared to the shallowest marker layer, it still experiences similar diffusion e nhancement, i.e. interstitial supersaturation. The apparent clustering peak revealed at l onger annealing times i ndicates the difference in the material structure, between deepest ma rker layer in the crystalline silicon and the

PAGE 84

70 shallowest marker layers in amorphous silic on, might be responsible for the existence of the BICs. The question then becomes if it is a mate rial structure, or just an observable consequence of the different interstitial s upersaturations each of the marker layers experiences at short times. Even though the regrowth of the 0.15 m between the shallowest and middle marker layer takes le ss than 0.25 seconds (according to [Ols88]), the supersaturation could drop by a factor of 2 or more during th at time [Cow99]. The delay of the a/c interface reaching the sh allowest marker layer facilitates avoiding the initial high interstitial supersaturations fo rming BICs in the middle marker layer, allowing only the lower interstitial concentra tion reaching the shallowest marker layer, and could also be contributing to the reduction in the BIC formation. 3.2.2 Modeling A UT-Marlowe simulation used to verify that the assumed distribution of excess interstitials following an amorphizing implant is shown in Fig. 3-9. Th e excess interstitial profiles obtained from the UT-Marlowe implan t simulation show that the damage is not concentrated solely around the middle marker layer and limited to the region just beneath the a/c interface. Rather, there are excess intersti tials that could provide the interstitials and boron in a sufficient proximity for cluster formation in the deepest marker layer. Assuming that the silicon and boron intersti tials are not able to enter the amorphous layers inserts a delay in the cluster form ation of the shallowest B marker layer. Presuming that the boron cluster formation is fast, the delay could prevent the BIC formation since most of the available Si inte rstitials would already be tied in clusters, either BICs or self-interstitial clusters. Th e delay would also mean that SMICs could be

PAGE 85

71 dissolved by the time the a/c interface reaches the shallowe st marker layer, thus the interstitial supersaturation would be signifi cantly reduced. The reduction of the interstitial supersaturation affects the BIC formati on in two ways, through reduction of self-interstitials and boron interst itials, again limiting the BIC formation. To test this hypothesis, an additional equation was used to calculate the amorphous layer depth in the simulation. The self-interst itial and boron in terstitial diffusivities were annulled in the region above the a/c interface, thus disabling their diffusion and their reactions to any other species. This forms a reflecting interface boundary at the position of the a/c interface, as it moves towards the surface. The a/c interface depth equation (Eq. 3-1) uses the regrowth velocity measur ed by Olson & Roth [Ols88] in units of m/s. The diffusivity of self-interstitial and boron in terstitials were limited to the crystalline material, as shown in the Eq. 3-2 and 3-3, with the reaction term omitted for brevity. kT eV t depth aSi 7 2 exp 10 1 3 _12 (3-1) terms reaction x Int depth aSi x D t IntInt_ _2 2 (3-2) terms reaction x BI depth aSi x D t BIBI_ _2 2 (3-3) The following simulation results illustrate the influence of spatial separation between the boron profile and the excess interstitials. The case chosen for this demonstration is the an nealing a boron dose of 4x1014 cm-2 implanted at 20keV. Figure 3-10 shows the simulated doses of pertinen t boron and interstitial clusters during a 30 minute anneal at 750oC, with boron implant performed without a preamorphization. Figure 3-11 shows the same boron implant, with a silicon preamorphization of 1x1015

PAGE 86

72 cm-2 implanted at 100keV, annealed under the same conditions. One of the benefits of the preamorphization is the smaller amount of excess interstitials remaining in the EOR region, as is apparent in this case, with excess intersti tial dose of ~8x1013 cm-2. In order to be able to compare the results, the ex cess interstitial dose in the preamorphized case was ad hoc increased by a factor of six, matching the excess interstitials available in the case without preamorphization. There are several important values indicating the domi nant processes in these simulations, namely Smic, C311 and Bsub. Initia l Smic dose is the in itial dose of trapped interstitials that evolve into {311}’s duri ng the course of the anneal. Without boron present, most of the excess interstiti als in Smic’s are captured by {311}’s. In the case of boron implant into crysta lline silicon (Fig. 3-10), only ~15% of excess interstitials from Smic’s are captured by {311}’s. The rest is trapped in various BIC configurations, reducing the substitutio nal boron dose to ~18%. Displacing the excess interstitials by a preamorphization Si im plant changes all of these values (Fig. 3-11). As the boron is relatively far away from the Smic profile, most of the excess interstitial dose in the EO R region is captured in a form of a {311} defect. Interstitial supersaturation supported by the {311} defects ev entually forms some BICs, still leaving >85% boron in the substitutiona l position after the 30 minutes 750oC anneal. This demonstrates the increase in the substitutional boron dose by a factor of four using preamorphization with the equivalent excess interstitial dose, solely by spatial separation of the boron and excess interstitial regions. Considering the actual preamorphization damage of a Si 1x1015cm-2 100keV implant performed at liquid nitrogen temperat ures, simulated excess interstitial dose is

PAGE 87

73 quite lower than that of the B implant. Simulated doses (Fig. 3-12) show a certain insensitivity of the clustering process to the EOR trapped interstitial dose, as the boron substitutional dose drops to ~90%. This may not be unreasonable when compared to the previous case (Fig. 3-11) having artificially increased excess inters titial dose, as the {311} supported interstitial supers aturation does not depend on th e defect density, size or trapped interstitial dose. Ther efore, the boron profile shoul d experience roughly the same interstitial supersaturati on irrespective of the {311} dos e, prior to its complete dissolution. This confirms the displacement of the excess interstitial makes a difference between 18% activation (Fig. 3-10) in non-preamorphized case versus over 85% activation (Fig. 3-11, 3-12) in the preamorphi zed case. Coupled with the lower amount of excess interstitials in th e realistic preamorphized case, such as a Si 1x1015 cm-2 implant at 100keV with liquid nitrogen cooling, the activati on can rise over 90% as seen in the Fig. 3-12. Even though the previously described bl ocking of the interstitial and boron interstitial pair diffusion into the amorphous layer significantly reduces the clustered portion of the profile, it does not yield good fits to the profiles in Fig. 3-8. The simulated profiles for the annealing of 5 seconds, 30 seconds and 3 minutes at 800oC are shown in Fig. 3-13, 3-14, and 3-15, respectively. The SIMS profiles in these fi gures are the SIMS profiles from Fig. 3-8. In these figures, the Bsub represents the substitutional portion of boron profile, the only species assume d to be electrically active. The Total_B profile represents all boron, irre spective of its configuration or pos ition in the crystal lattice. The clustered boron concentrati on is not plotted for clarit y, since it is inferred as Total_B-Bsub.

PAGE 88

74 One can notice that the shallowest and the deepest B marker peak magnitude are very similar throughout the simulation, in c ontrast to the SIMS profiles. This is interesting, particularly considering the diffe rent portions of the pr ofile in a clustered state, visible as the different Bsub peak concentrations in Fig. 3-15. The time dependence is not captured well, as the shallowest and the deepest marker layers do not diffuse as seen in the SIMS. Consideri ng the activation of boron in th e shallowest peak drops only 14% from the full activation, this make s the significant reduction in diffusion enhancement interesting. Presuming the formation of BICs is still too strong in the regrown amorphous layer, the model parameters were adjusted. The binding energy of the small clusters was reduced without affecting equilibrium di ffusion, in an attempt to reduce their concentrations. This effectively prevents small BICs from serving as trapping and nucleation sites for larger BICs. The results from the simulations using the modified BIC model parameters are shown in Fig. 3-16, 3-17, and 3-18. In these figures, one can observe the fairly good match for the middle B marker layer. At longer annealing times, it would s eem that the clustered peak does not release enough mobile boron to retain the diffu sed tails around the clustered peak. The shallowest and the deepest marker layers ha ve a similar peak concentration at shorter times despite the different clustered portions. Peak concentrations differ at longer times, revealing a clustered peak in the deepest marker layer that is not present in the shallowest marker layer. The shallowest marker layer does form clusters, but since the clustered dose dropped to around 1.7%, they ar e not visible in the profile.

PAGE 89

75 The binding energies of the small BICs that were reduced to f acilitate the reduction of clustering in regrown materi al did not significantly affect the other parameters of the model. That is the due to the fact that th e reactivation properties are the consequence of the dissolving cluster, in this case B4I4. The changes did seem to accelerate the B2I3 dissolution, which was compensated by increa sing its binding energies without the adverse effects on either clus tering and reactivation simulatio ns, or the simulations of BIC clustering in amorphous material. The overa ll increase in activa tion is partially due to the change in the formulation of in terstitial supersatur ation supported by {311}’s (C311 BindI) to match enhanced diffusion with the interstitial equilibrium concentration (Int Cstar) values used in the Li u et al. [Liu00][Win99], and pa rtially due to the change in the model parameters. Two figures (Fig. 3-19, 3-20) compare the reactivation during second annealing stage at 850oC, in inert ambient, between the model presented earlier in the chapter and the changes introduced to capture the behavior in regrown silicon. 3.3 Conclusion Boron interstitial clusters formed by a 750C anneal afte r implantation deactivated the majority of implanted boron. These sample s allow a study of the reactivation rate in two different ambient conditions. The oxidi zing ambient shows consistently lower activation at a slower rate when compar ed to the inert ambient, which strongly demonstrates that the presence of excess in terstitials slows cluster dissolution. This conclusion is contrary to the accepted BIC model, though it is supported by experimental evidence [Sol00] suggesting that BIC disso lution involves the release of silicon interstitials. To model the oxidation behavior, we utili zed the energetics publis hed by Liu et al. [Liu00], with some modifications. The mode l implemented in the description language

PAGE 90

76 ALAGATOR, and simulated using the proce ss simulator FLOOPS [FLO02] captured the reactivation behavior, correctly distinguishing the reaction to the ambient atmosphere. An implementation of the re flective boundary at the a/c interface reduces the simulated clustered portion of the B implanted into amor phous silicon, as seen in the work of Jones et al. [Jon96a]. However, the reductions in binding energies of small BICs were necessary to model the clusteri ng behavior in the crystallin e, versus amorphous silicon.

PAGE 91

77 Table 3-1. Retained boron dose [cm-2] in different processing step s. First anneal step is 750oC, 30 min in inert ambient. Second anneal step is 850oC, 60 min in respective ambient. Dose loss in lowest implanted dose sample is comparable to active dose measured by Hall effect measurement. Second anneal step Implanted dose [cm-2] First anneal step [cm-2] Inert [cm-2] Oxidizing [cm-2] 4x1014 3.69x1014 3.58x1014 3.33x1014 2x1014 1.75x1014 1.58x1014 1.25x1014 1014 7.48x1013 7.01x1013 4.48x1013

PAGE 92

78 Figure 3-1. Secondary ion mass spectrome try (SIMS) measured boron profiles of investigated implant conditions (B implant 1x1014 cm-2 at 5 keV, 2x1014 cm-2 at 10 keV and 4x1014 cm-2 at 20keV) after firs t anneal step, 750oC for 30 minutes in inert ambient.

PAGE 93

79 Figure 3-2. Secondary ion mass spectrometry (SIMS) measured boron profiles. Initial condition is after 750oC, 30 minutes inert anneal. Subsequent anneal is 850oC, 60 minutes in respective ambient. Figure parts a), b) and c) shows the result for a 4x1014 cm-2/ 20keV, a 2x1014 cm-2/ 10keV, and a 1x1014 cm-2/ 5keV B implant, respectively. These conditi ons are chosen because the peak concentrations and initial clustered concentrations are similar.

PAGE 94

80 Figure 3-3. Active dose measured by Hall-van der Pauw method, during the annealing at 850oC. The time zero measurement corresponds to the condition after a 750oC, 30 minutes inert anneal.

PAGE 95

81 Figure 3-4. Active fraction, the ra tio of active to retained dos e (integrated SIMS profile), compared for the different annealing ambient at the end of the 850C, 60 minutes anneal. The initial condition is after a 750oC, 30 minutes inert anneal.

PAGE 96

82 a) b) Figure 3-5. Major cluster formation and dissolu tion paths of Liu et al. [Liu00] with B3I and B2I3 containing most of the B clustered dose in figure a) and simulation of B clustering and dissolution during thermal processing used in experiment in figure b)

PAGE 97

83 a) b) Figure 3-6. Cluster formation and dissoluti on paths of modified Liu et al. [Liu00] energetics, with B4I4 and B2I3 containing most of B cl ustered dose, are shown in figure a), with simulation of B clustering and dissolution (modified energetics) during thermal processing used in experiment is shown in figure b)

PAGE 98

84 Figure 3-7. Boron profiles of material us ed in the study of Jones et al. [Jon96a], unimplanted and annealed in inert ambient at 800oC

PAGE 99

85 Figure 3-8. Boron profiles after a 5x1015 Si implant at 146keV, annealed in inert ambient at 800oC [Jon96a]

PAGE 100

86 Figure 3-9. Boron marker layers and excess interstitial damage following a 146keV Si implant shows proximity of deepest marker layer to the damage

PAGE 101

87 Figure 3-10. Doses of important cl usters in a simulation of a B 4x1014 20keV implant into crystalline Si, during a 30 minutes 750oC anneal, resulting in activation ~18%

PAGE 102

88 Figure 3-11. Doses of important cl usters in a simulation of a B 4x1014 20keV implant into preamorphized Si (with excess interstiti al dose increased by a factor of 6), during a 30 minutes 750oC anneal, resulting in activation >85%

PAGE 103

89 Figure 3-12. Doses of important cl usters in a simulation of a B 4x1014 20keV implant into preamorphized Si (at liquid nitrogen temperatures), during a 30 minutes 750oC anneal, resulting in activation >90%

PAGE 104

90 Figure 3-13. Simulated boron pr ofiles after 5 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-inter stitials and BI pairs are annulled in the amorphous material.

PAGE 105

91 Figure 3-14. Simulated boron pr ofiles after 30 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-i nterstitials and BI pairs are annulled in the amorphous material.

PAGE 106

92 Figure 3-15. Simulated boron profiles after 3 minutes at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-inter stitials and BI pairs are annulled in the amorphous material.

PAGE 107

93 Figure 3-16. Simulated boron pr ofiles after 5 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-inter stitials and BI pairs are annulled in the amorphous material, and the binding energies of small BICs are reduced.

PAGE 108

94 Figure 3-17. Simulated boron pr ofiles after 30 seconds at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-i nterstitials and BI pairs are annulled in the amorphous material, and the binding energies of small BICs are reduced.

PAGE 109

95 Figure 3-18. Simulated boron profiles after 3 minutes at 800oC, following a Si implant of 5x1015 cm-2 at 146keV. Diffusion of self-inter stitials and BI pairs are annulled in the amorphous material, and the binding energies of small BICs are reduced.

PAGE 110

96 Figure 3-19. Simulated clustered doses in BIC experiment for B implant 4x1014@ 20keV, model presented earlier in the chapter

PAGE 111

97 Figure 3-20. Simulated clustered doses in BIC experiment for B implant 4x1014@ 20keV, model parameters adjusted to co ver regrowth behavior [Jon96a]

PAGE 112

98 CHAPTER 4 BORON CAPTURE AT THE OXIDE/SILI CON INTERFACE AND END-OF-RANGE DISLOCATION LOOPS Overview This chapter discusses the phenomena of B diffusion near the oxide/silicon interface and end-of-range (EOR) dislocation loops under the in terstitial supersaturation. The oxide/silicon interface trap s the mobile BI pairs, thus providing an additional mechanism for removal of excess interstitial s from the bulk and effectively reducing the diffusion enhancement. This chapter also discusses the experimentally observed segregation of B in the vicinity of EO R loops that is modeled through a trapping mechanism along the circumference of the loop. 4.1 B Interaction with Oxide/ silicon Interface Under TED This section discusses an experiment de signed to investigate Ge influence on B diffusion under transient enhanced diffusion (TED) conditions, with emphasis on the control samples. The following section focu ses on the experimental phenomena observed in samples receiving no Ge implant, pert inent simulations and conclusions about interaction of B and oxide/silicon interface. 4.1.1 Experimental Conditions Four float-zone (FZ) n-type Si <100> wafers, resistivity 70-130 cm, were preamorphized by silicon implant with a dose of 1x1015 cm-2 at an energy of 100keV. The Si preamorphization implant (PA I) provides an amorphous layer 0.23 m thick, as measured by cross-section transmission el ectron microscopy (XTEM). The implant was

PAGE 113

99 performed at liquid nitrogen temperatures to ensure the surface laye rs would experience the TED. During the annealing following an implant under room te mperature conditions, a loop layer could form at the amorphous/crystalline ( a/c ) interface, capturing the interstitials released from EO R damage and preventing the back flow of interstitials to the surface [Cha96]. Following the Si PAI, one wafer was implanted with Ge dose of 4x1015 cm-2 at 30keV, hereon out referred to as sa mple A. The control wafer (sample D) received an additional Si implant with 1x1015 cm-2 at 30 keV, ensuring amorphization in the near surface region. Finally, all the wafers received the same B implant, with dose of 2x1014 cm-2 at 10 keV. Since the Si PAI and the final B implant are identical for all wafers, the Ge implant dose is the only variab le in the implant sequence. The full implant sequence is listed in table 4-1. The dopant profiles, their relative positions as well as the initial a/c interface position are illustrated in Fig. 4-1. The followi ng figure (Fig. 4-2) shows amorphous layer depth as determined by XTEM, courtesy of An tonio Saavedra. These images confirm that the two-step amorphization formed a con tinuous amorphous layer extending from the surface, as opposed to a buried amorphous layer. Anneals were performed at temperatures of 700 and 825oC in inert N2 ambient. An AG Associates Heat Pulse 210T Rapid Thermal Annealer (RTA) was u tilized for anneals shorter than 15 minutes, with a 125oC/s ramp up rate. Anneals exceeding 15 minutes were conducted in a Thermolyne quartz tube furnace. Sample A, which received the 4x1015 cm-2 Ge implant, was subjected to the annealing conditions at the same time as the control samples. Fig. 4-3 and 4-4 show the B profiles during the 700oC and 825oC anneals, respectively. The interesting feature of

PAGE 114

100 these figures is the increase in the profile peak. The increase and th e shift of the peak seem to occur closer to surface and also closer to the Ge implanted region. The time dependence suggests the phenomena is associ ated with TED, as the increase stops approximately at the same time as the cont rol sample profiles cease motion. One could, therefore, presume a GexByIz cluster might be responsible for such a behavior. Before delving into that hypothesis too much, there are additional measurements one must take into account. At a later stage of the experiment, the active dose measurements via Hall effect consistently failed to get the expected carri er type on Ge implanted samples. Several Hall-van der Pauw measurements are listed in table 4-2, to illust rate the situation. Having the n-type silicon fo rmed by Si PAI, Ge and B implant required an explanation. One possibility was that the EO R damage in the depletion region, known to serve as recombination centers, could be incr easing the leakage current into the bulk. The electrons in the bulk can offset the measurem ent, even yielding incorrect carrier type. Hence the measurement is conducted at ~ 100K to prevent carriers from populating mid-bandgap traps [Jai03] and subsequently r ecombining. In this pa rticular case, the carrier type detected remained unchanged. Th e possibility of surface contamination was discounted based on the fact that the carrier type in the control samples was consistently detected as holes. Since samples A and D were handled in the same manner, annealed in the same furnace/RTA, mostly at the same tim e, it would be difficult to expect only one set of samples to get contaminated in the process. The last two rows in the table show the ne t electrons present even after a very large thermal budget, unexpected considering the SIMS profiles of smaller thermal. At this

PAGE 115

101 point, the implant provider hinted that the implant was not specifically a 74Ge isotope. Furthermore, that the mass spectrometry apertu re was open wide to le t three Ge isotopes (70Ge, 72Ge, and 74Ge) from the source reach the target Providing the next element in atomic weight is 75As, a donor in silicon, this suggeste d there might be a slight problem with the experiment. Figure 4-5 shows the As presence in a ll the samples implanted with Ge. The samples designated to receive 4x1015, 1.2x1015, and 4x1014 cm-2 doses of Ge, also received the 3.3x1014, 5.75x1013 and 2.15x1013 cm-2 doses of As. Having the As projected range and distribution similar to that of Ge would suggest the contamination from the source, or non-discriminant mass spectrometry. The evolution of the B profiles for the experimental control (no Ge implant), sample D, during the 700oC and 825oC anneals are shown in Fig. 4-6 and 4-7. The as-implanted condition was added to the figur es for comparison. It is evident that the profiles exhibit profile broadening consistent with TED. Also, there appears to be no immobile portion of the B profile, generall y associated with the presence of boroninterstitial clusters (BICs). This further asse rts that the experiment al conditions allow for the investigation of Ge and B interaction, w ithout an eventual in terference from other clustering processes. Other inte resting features of these prof iles include the slight motion of the profile peak towards the surface, and the B surface spike. While the assumed reflective boundary at the surface might account for some of the prof ile peak shift, one would not expect a significant dip to still exist between th e surface and the profile peak by the time the depth of the peak shift is observed. As for the B surface spike in the secondary ion mass spectrometry (SIMS) measur ement, it is presumed to be due to the

PAGE 116

102 transition between the oxide and the silicon matri x. It is considered to be an artifact and not related to the possibility of B presence in the thin native oxide or at the interface. However, looking at the figures 4-6 and 4-7, one can observe the steady increase of the surface spike width with annealing time. A ll the SIMS measurements were performed under the same conditions, so the artifact should not change on that account. Interestingly, this does not occur with the Ge implanted samples (Fig. 4-3 and 4-4). In sample A, the B surface spike width doe s not increase with annealing time. The surface spike raises th e question of the dose conservation, and the possible need to normalize the measured profiles, bot h for comparison between the measured and simulated profiles. Table 4-3 compares the total dose measured by SIMS to the dose when the surface spike (first 7nm) is ignor ed, and provides the ratio between the two. Taking the “artifact only” presumption would imply the SIMS dose error would result in a standard deviation of 11.1%, as opposed to the 2.95% when assuming the surface spike is actually B. The trend of dose decreasi ng with annealing time when ignoring surface spike is another argument agai nst the “artifact only” assump tion, as no such trend exists when entire profile is integrated. Again, thes e differences do not occur in sample A to that extent, as shown in table 4-4. 4.1.2 Simulations Initial simulations of control sample a nneals (Fig 4-8) indi cated a problematic mismatch in the profile shape, peak magnit ude, junction depth and surface spike. This would suggest the bulk enhancement factor and surface phenomena might not be properly modeled. The profile broadening appears excessi ve, reducing the peak magnitude and spreading the profile more into the bulk. The convoluted profile is the simulated B profile

PAGE 117

103 convoluted with an exponential function with a decay length of 3nm. The decay length was extracted from the slope of the surf ace peak, presuming it due to the almost a -function of B at the interface or in the native oxide layer. It is interesting to note the simulation fails to match the surface spike for the longer anneal times, although the surface concentration of B in si licon is not very far off the measured values. Also, if the B surface spike were entirely due to the segr egation of B into the oxide, it would not be increasing at longer times. After 2 minutes, th e concentration of B in the near surface region, presumably at the surf ace as well, is declining due to diffusion into the bulk. Therefore, one would expect the surface spike to decrease in concentration, or stay the same for longer anneal times. Leaving aside the surface phenomena for a moment, one might attempt to match the enhanced diffusion by reducing the enha ncement factor. That would reduce the effective diffusion of B, possibly providing a better match to the measured profiles. Having reduced the enhancement factor (C 311 BindI), previously calibrated on the {311}’s defect dissolution and enhanced diffusion of B marker layers by a factor of 5, the resulting simulated profiles are shown in Fig. 4-9. The reduction in the enhancement factor doe s facilitate a better agreement between the simulated and the measured profiles, na mely being closer on the diffusion into the bulk. However, it does not match the profile shape or the surface peak much better than in Fig. 4-8. The slight shift of the profile peak towards the surface does not occur in the simulation until the profile between the peak and the surface becomes completely flat. As for the surface peak, these simulations do not match the surface spike at any point in

PAGE 118

104 time. Rather, the surface spike is shallo wer than measured value for a given concentration. One might consider the possibility of trapping of B on the surface, or the oxide/silicon interface. The interface trap density of 2x1014 cm-2 was taken from Voung et al. [Vou96], previously used to explain threshold voltage shifts in pMOS devices. The simulations seem to match the measur ed profile reasonably well, once the BI pair trapping on the surface is included. This trapping mechanism facilitates the continuous widening of the surface spike, but affects the overall profile shape. The trapping of the BI pair at the interface para llels the same process for self-interstitials, increasing the removal rate of excess inte rstitials. The additional recombination mechanism for interstitials at the oxide/ silicon interface increases the gradient of interstitial towards th e surface, assisting the BI pair tr aps at the interface in shifting the profile peak towards the surface. This is vi sible when comparing the Fig. 4-9 and 4-10, for diffusion times from 1 to 4 minutes. Simulating the annealing of sample A and assuming the interaction of Ge and B negligible compared to the electric field e ffects, results in prof iles seen in Fig. 4-11. 4.2 Boron Segregation to the EOR Loops An additional experiment was performed to further examine the interaction of B with the oxide/silicon interface and EOR loops The use of highly doped p-type wafers simplifies the experiment, as there is no need for B implant. The presence of B in the EOR loop layer shows an increase of B con centration in the loop layer, modeled by the capture of BI pair on the EOR loops.

PAGE 119

105 4.2.1 Experimental The wafers used were float-zone (FZ) p-type Si <100> wafers, resistivity ~0.02 m cm, with the concentration of B ~4x1018 cm-3. Both of these values were confirmed via four-point probe measurement and SIMS, respectively. Both wafers received a Si 1x1015 cm-2 implant at 100keV, with liquid nitrogen cooling. The control wafer (samples E1) received an additional Si 1x1015 cm-2 implant at 30keV to ensure the continuous amorphous layer, while the Ge implanted wafer (samples E2) received a Ge 4x1015 cm-2 at 30keV. The samples were annealed at 750oC for duration of 1 and 4 hours, and at 825oC for 1 and 4 minutes. Figures 4-12 and 4-13 show the boron prof iles of sample E1 annealed at 750 and 825oC, respectively. The profiles exhibit an increase below 0.2 m, indicative of the B trapping on the EOR loop layer. Also, there is a broadening and an increase in magnitude of the B surface spike, coming from the B tr apping at the oxide/silicon interface. The dip in the otherwise flat B profile near the surf ace is the consequence of the B segregation to the oxide and the oxide/silicon interface. 4.2.2 Simulations The phenomenon observed in the sample E1 is associated with the capture of B at the oxide/silicon interface and the EOR l oops. Assuming the characterization of the former from the earlier parts of this ch apter is adequate, a capture reaction on the boundaries of the EOR loops is added (after Xia et al. [Xia99]). The reaction rate is assumed to be diffusion limited along the circ umference of the loop, as shown in Eq. 4-1 and 4-2. The BI represents the mobile boroninterstitial pair concentration, with DBI as its

PAGE 120

106 diffusivity. The RLoop and DLoop are average radius and the loop density, and the BLoop is the concentration of B trapped in the loop. BI Si fD a k 4 (4-1) Loop Loop f LoopD R BI k t B 2 (4-2) The simulated and SIMS measured profiles of the sample E1 are compared in the Fig. 4-14 and 4-15. In the Fig. 4-14 and 4-15, one can observe a relatively good match for the surface spike broadening and the B profile dip near the surface. The capture on the EOR loops has certain discrepancies from the experiment al data. From the experimental data, it would seem the capture is symmetric around th e loop layer and shallo wer than predicted by the model. The modeled capture does not have the same symmetry, since the simulated loop density decays more gradually towards the bulk. Therefore, even though the captured B provides a dip and a sharp increas e on the shallower side, it is not possible to get the same (Fig. 4-12) or even deeper dip (Fig. 4-13) on the deeper side of the segregated peak. The lower de nsities of EOR loops capture B preventing the formation of the observable dip in the regi on deeper than the former a/c interface. Despite these differences, the capture model doe s provide a reasonable match. 4.3 Conclusions An experiment is designed to inve stigate the Ge and B interaction in preamorphized silicon. The Si preamorphization provides the interstiti al supersaturation, as well as significantly reduces the formation of BICs, thus simplifying the diffusion of B in the control samples.

PAGE 121

107 Nonetheless, B profiles in the control sa mples diffused slower than the available models predicted. The simultaneous B dose lo ss in the bulk and the broadening of the surface spike, usually considered a SIMS meas urement artifact, indicated a possibility of B trapping at the oxide/silicon interface. Incl uding an equation to describe the trapping of a BI pair at the interface yielded improved f its, both in the near surface region and the bulk diffusion. These improved fits were achie ved without changes to the other models. Also, the phenomenon of B segregation to EOR loop layer is observed in the Si preamorphized, highly doped p-t ype wafer. Modeling this phenomenon through a BI pair capture at the circumference of the loop provides a qualitati ve fit to the experimental data.

PAGE 122

108 Table 4-1. Sample labels and implant sequen ce, showing Ge implan t dose as the only variable in the sequence Sample Si implant Ge implant B implant A 1x1015 cm-2, 100 keV, LN2 4x1015 cm-2, 30 keV 2x1014 cm-2, 10 keV B 1x1015 cm-2, 100 keV, LN2 1.2x1015 cm-2, 30 keV 2x1014 cm-2, 10 keV C 1x1015 cm-2, 100 keV, LN2 4x1014 cm-2, 30 keV 2x1014 cm-2, 10 keV D 1x1015 cm-2, 100 keV, LN2 1x1015 cm-2, 30 keV none 2x1014 cm-2, 10 keV

PAGE 123

109 Table 4-2. Active dose and carrier type measured by Hall-van der Pauw system Sample Annealing conditions Hall temperature [K] Active dose [cm-2] Carriers Ge implanted 15m@700C 300 1.67x1014 eGe implanted 15m@700C 100 3.95x1014 eGe implanted 4h@700C 300 8.48x1013 eGe implanted 4h@700C 100 1.73x1014 eControl 15m@700C 300 2.78x1014 h+ Control 4h@700C 300 2.23x1014 h+ Ge implanted 15s@825C 300 7.13x1013 eControl 15s@825C 300 1.39x1014 h+ Ge implanted 8m@825C 300 3.89x1013 eControl 8m@825C 300 2.02x1014 h+ Ge implanted 8m@825C, 20s@1100C 300 4.56x1013 eGe implanted 8m@825C, 2h@1100C 300 3.26x1014 e

PAGE 124

110 Table 4-3. Integrated dose from the measured B profiles in the contro l sample annealed at 825oC, taking full profile (Q1) and ignoring the surface spike (Q2) Annealing conditions SIMS integrated dose Q1 [cm-2] SIMS integrated dose Q2 [cm-2], x>7nm Ratio (Q2 reference/Q2) as-implanted 1.67x1014 1.58x1014 1.03 15s@825C 1.78x1014 1.63x1014 1 (reference) 30s@825C 1.79x1014 1.60x1014 1.02 1m@825C 1.82x1014 1.60x1014 1.02 2m@825C 1.75x1014 1.44x1014 1.13 4m@825C 1.80x1014 1.30x1014 1.25 8m@825C 1.73x1014 1.21x1014 1.34 Q [cm-2] 1.77x1014 1.49x1014 [cm-2] 5.21x1012 1.65x1013 Q [%] 2.95 11.1

PAGE 125

111 Table 4-4. Integrated dose from the measured B profiles in the 4x1015 cm-2 Ge implanted sample annealed at 825oC, taking full profile (Q1) and ignoring the surface spike (Q2) Annealing conditions SIMS integrated dose Q1 [cm-2] SIMS integrated dose Q2 [cm-2], x>7nm Ratio (Q2 reference/Q2) as-implanted 1.88x1014 1.61x1014 1.1 15s@825C 2.03x1014 1.77x1014 1 (reference) 30s@825C 2.01x1014 1.76x1014 1 1m@825C 1.87x1014 1.65x1014 1.08 2m@825C 1.88x1014 1.65x1014 1.07 4m@825C 1.82x1014 1.58x1014 1.11 8m@825C 1.88x1014 1.66x1014 1.07 Q [cm-2] 1.91x1014 1.67x1014 [cm-2] 7.81x1012 7.23x1012 Q [%] 4.09 4.32

PAGE 126

112 Figure 4-1. Profiles of B and Ge in as-implanted sample A

PAGE 127

113 a) b) Figure 4-2. Cross-section TEM (XTEM) imag es of samples (a) A and (b) D under 100000 magnification

PAGE 128

114 a) b) Figure 4-3. B profiles from Ge implan ted samples during annealing at 700oC

PAGE 129

115 a) b) Figure 4-4. B profiles from Ge implan ted samples during annealing at 825oC

PAGE 130

116 a) b) c) Figure 4-5. SIMS profiles of Ge, As and B in samples: A, with 4x1015 cm-2 Ge (1min@825oC), B with 1.2x1015 cm-2 Ge (as-implanted) and C with 4x1014 cm-2 Ge (as-implanted) per figure a, b, c, respectively

PAGE 131

117 a) b) Figure 4-6. B profiles from contro l sample during annealing at 700oC

PAGE 132

118 a) b) Figure 4-7. B profiles from control sample during the annealing at 825oC

PAGE 133

119 a) b) c) d) e) f) Figure 4-8. Simulated B profiles of c ontrol sample during anneals at 825oC

PAGE 134

120 a) b) c) d) e) f) Figure 4-9. Simulated B profiles of c ontrol sample during anneals at 825oC, diffusion enhancement factor reduced five times

PAGE 135

121 a) b) c) d) e) f) Figure 4-10. Simulated B profiles of c ontrol sample during anneals at 825oC, assuming surface trapping of BI

PAGE 136

122 a) b) c) d) e) f) Figure 4-11. Simulated B profiles of 4x1015 cm-2 Ge implanted sample (As contaminated) during anneals at 825oC, assuming electric fi eld effect dominant

PAGE 137

123 Figure 4-12. B profiles from samp le E1 during annealing at 750 oC

PAGE 138

124 Figure 4-13. B profiles from samp le E1 during annealing at 825oC

PAGE 139

125 a) b) Figure 4-14. Simulated B profiles of B segreg ation to the loops in sample E1 during annealing at 750oC

PAGE 140

126 a) b) Figure 4-15. Simulated B profiles of B segreg ation to the loops in sample E1 during annealing at 825oC

PAGE 141

127 CHAPTER 5 BORON INTERACTION WITH GERMANIUM UNDER INTERSTITIAL SUPERSATURATION Overview This chapter provides an overview of the earl ier model of Lever et al., as well as an experiment used in attempt to extract parameters for the model. The experiment is designed to observe the B and Ge interacti on in the high interstit ial supersaturation environment to maximize an eventual reaction product. Upon implanting Ge and B, during a subsequent anneal Ge seems to be slowing down B diffusion and retaining higher B concentrations when compared to control samples. Modeling this behavior through Ge-B pairing mechanism provides a re asonable fit to the experimental data. 5.1 Prior Work Model Implementation In an effort to model B diffusion in lowe r Ge contents, the model of Lever et al. [Lev98] is utilized in the experimental de sign. However, some changes are required, in order to account for phenomena such as bandgap narrowing and transient enhanced diffusion (TED). 5.1.1 Bandgap Narrowing Germanium content can vary as a func tion of depth, which means the bandgap narrowing is a function of depth as well. This requires a slight change in the potential equation, which treats bandgap, EG and intrinsic carri er concentration, ni as a function of temperature, but not as a spatial variant. Adding the EG and ni terms allows for a change with respect to depth, or any other axis. So lving the Poisson equation with a spatially

PAGE 142

128 varying bandgap and considering the Fermi level influence on diffusion accounts for the quasi-electric field effects. Although the bandgap narrowing due to the Ge presence provides some segregation of B into Ge ri ch layers, it is not enough to account for the experimentally seen increas es in B concentration. Lever et al. followed Kuo [Kuo93], in re ferencing People’s [Peo85] calculation in determining the bandgap change. Their result s are in agreement with data of Lang [Lan85], Dutartre [Dut 91] and Robbins [Rob92]. In partic ular, the expressions of Weber et al. [Web89] and Dutartre [Dut91] are employed for the bandgap of relaxed and strained layers, resp ectively. Conduction and valence band densities of state are assumed not to change significantly. 5.1.2 Boron Diffusion in Si and Interaction with Ge It is well established that B diffuses almost exclusively through interaction with silicon interstitials [See68][Fah85][Fan96] [Gos97][Ura99b]. Furthermore, there is a concentration dependence of diffusion, wh ich is modeled as diffusion through the interaction with neutral and ch arged interstitials. Since the Poisson equation has taken into account the spatial vari ance of the Ge con centration, its effect s on the diffusion of neutral and negatively charged BI pairs ar e already considered in the Fermi level dependent diffusivity. x n p BI n p n p D D Ji i i BIp BI BI 10 (5-1) x n p BI x BI n p n p D D Ji i i BIp BI BIln 10 (5-2)

PAGE 143

129 x n x p x n pi i ln ln ln (5-3) Further divergence from the model of Lever et al. includes change s to the clustering reaction. The original model considers B as a diffusing species, which is replaced by a BI pair diffusion. The modified model assumes a reaction is occurring be tween a BI pair and a Ge atom, such that a GeBI complex is fo rmed. The GeBI complex can then release an interstitial forming a GeB pair. Alternatively, it can release a BI pair returning to the initial Ge configuration. The original work of Lever et al. suggests that GeB pairing occurs as a consequence of mi croscopic strain relaxation. Du e to lack of an alternative explanation of the GeB pairi ng phenomena that Lever’s mode l describes, this work is continued under the same presumption. BI Int Bsub (5-4) GeBI Ge BI (5-5) GeBI Int GeB (5-6) 5.2 Experiment Design and Considerations This section discusses the experiment de signed to investigat e Ge influence on B diffusion under transient enhanced diffusion (T ED) conditions. TED conditions are used to accentuate the eventual difference in B di ffusion with and without Ge present in the material. In case a difference in the profile ev olution is observed, th at would discern any doubt of the existence of GeB pair, and might offer some insight into its properties. In order to investigate th e influence of Ge on B diffu sion, all other parameters involved should be invariant of experiment al conditions. One of the problems in the experiment design is that B is known to form boron-interstitial clusters (BICs) upon

PAGE 144

130 implant into a crystalline silicon and subs equent annealing [Mic87][Sto95]. This should be avoided in order to incr ease the probability of a B atom reacting with a Ge atom, versus another B atom or silicon self-inter stitials. One way to reduce the formation of BICs is to amorphize the surface silicon layer pr ior to a B implant. The work of Jones et al. [Jon96a] showed BICs do not form in thick amorphous layers, yet B experiences enhanced diffusion. Germanium is a popular choice as an amor phization species. It is an isovalent chemical element in the silicon lattice; its crys tal lattice is 4.2% larg er crystal lattice than that of silicon, and it amorphizes silicon at lower implant doses due to its larger mass. Although there are a number of studies on th e effect of Ge pre-amorphizing implant (PAI) on B diffusion, these investigations are generally directed towards junction formation and have varying doses of Ge PAI. This allows for investigations of different initial amorphous layer depths and end of range (EOR) damage, which may be influencing the B diffusion beha vior and its reaction with Ge However, this precludes the extraction of parameters that determine the Ge-B interaction. Considering the goal of this study is to i nvestigate the interaction of Ge and B, initial conditions should be invariant of di fferent Ge and B implant conditions. The amorphous/crystalline ( a/c ) interface should be defined by an implant common for all the samples in the experiment. Using silicon PAI for the formation of the amorphous layers has several benefits. First, the Si PAI crea tes an amorphous layer without the introduction of dopants. Secondly, the Si PAI allows the a/c interface to be positioned independently of the other implants. Thirdly, if the damage profiles of the other implants are contained within the amorphous layer created by the Si PAI step, those im plants will not contribute

PAGE 145

131 to the EOR damage. The evolution of th e EOR damage and TED during subsequent anneal steps would then solely depend on the Si PAI. Therefore, the Si PAI ensures the independence and invariance of a/c interface position, EOR damage and TED, with respect to Ge and B implant conditions. The diffusion enhancement, due to the intersti tial supersaturation, is also beneficial for the detection of the product in the reac tion described by the Eq. 5-5. For a given interstitial supersaturation, defined as si licon interstitial concentration over the equilibrium interstitial concentration, BI c oncentration is increased by the same factor (Eq. 5-4). This means more B will be in its mobile form, and available for reaction with Ge, resulting in the increase of the GeBI. Du e to the interstitial release in the reaction forming a GeB pair (Eq. 5-6), the increase in the GeBI would not necessarily lead to the increase in a GeB pair concentration. Initial simulations using the model adopted from Lever et al. [L ev98] indicated the possibility of cluster formation under interstit ial supersaturation during TED, without the need for a high Ge content. Hence, an invest igation of lower Ge concentrations could be used to verify the existence of the GeB pair. Should the B diffusion depend on Ge implant conditions, it may be possible to probe the binding energy of the cluster by additional anneal steps. Alternatively, it w ould disprove the existence of a clustering reaction between Ge and B. This would indi cate that the diffusion reduction might be caused by other effects, such as strain or interstitial formation energy variation. 5.2.1 Experimental Conditions Having realized the contamination occurred during the implant (details in Chapter 4.1), the experiment was repeated using new material. The wafers used were float-zone (FZ) p-type Si <100> wafers with resistivity of 140-180 cm. The implant conditions

PAGE 146

132 were identical to the initial experiment from the previous chapter. Sp ecifically, the Si PAI with an implant dose of 1x1015 cm-2 at the energy of 100 keV produced a 0.22 m thick amorphous layer. Subsequently, one wafe r was implanted with Ge dose of 4x1015 cm-2 at 30 keV, labeled sample 2A. The control wafe r (sample 2D) received another Si implant with 1x1015 cm-2 at 30 keV, ensuring the continu ous amorphous layer. Finally, both wafers received the same B implant, with a dose of 2x1014 cm-2 at the energy of 10 keV. The anneals were performed at the temperature of 780oC in a RTA furnace under nitrogen flow. 5.2.2 Experimental Results Figure 5-1 shows the as-implanted profiles of the sample 2A (Ge implanted). The amount of As found in the Ge implanted samples were on the order of 0.5%, the same percentage as detected in the Ge standard used for calibration of the SIMS counts. Presuming the Ge standard is not contaminated would lead one to consider the 75As signal as interference of 74Ge+H. The time dependence of B profile during the 780oC anneals is shown in Fig. 5-3. Figure 5-3 a) shows B profiles for Ge implante d samples. The profile tail diffuses into the bulk, while the profile peak shifts towards the surface. It is also possible that the profile peak is shifting towards the Ge doped region, to form GeB pair. The profiles in the control samples (Fig 5-3.b) show enhanced diffusion, as a consequence of the {311} defect dissolution in the EOR region. In bot h these cases, the diffusion enhancement is dropping as one nears the surface, due to the interstitial concentra tion gradient towards the surface. Figures a) and b) differ mostly in the profile peak and the surface spike. Ge implanted samples seem to shift the pr ofile peak without a significant drop in

PAGE 147

133 concentration, and do not have such a clear in crease in the surface spike. Both of these observations could be due to Ge trapping B, retaining a higher B peak concentration, and preventing B from reaching the surface to contribute to the surface spike. As for the similarities, the junction depth motion in both samples seems parallel. This would indicate a negligible difference in self-intersti tial concentration between the Ge implanted and control samples. Therefor e, neither Ge diffusing through interaction with interstitials, nor any sort of clustering presents an interstitial sink. 5.2.3 Simulations In attempt to simulate these anneals, one can consider the results of Zangenberg et al. [Zan01]. They reported the increase in activation energy for B diffusion in relaxed Si1-xGex layers of ~0.5eV, as a weak function of the Ge content. Relating that to the GeB pair, one could infer that the formation ener gy of GeB pair is similar to the activation energy increase. Based on the similarity of the junction depth during the anneal, the GeBI complex is presumed not to be as signifi cant. Assuming the B diffusion mostly occurs through an interaction with interstitials, fo rming a BI pair; tying the BI in a GeBI complex would reduce the concentration of the mobile B specie. Macroscopically, it would be observed as the reducti on of the effective diffusivity, most notably in the tail region. Since no significant difference in tail diffusion is observed between Ge implanted samples and the control, the concentration of GeBI is assumed to be much lower than that of GeB. This is valid when the binding energy of a BI pair to GeBI is lower than 2.4eV, and for Ef GeB of 0.5eV, which are the values used in the simulations. The binding energy of the GeBI complex would also be rather di fficult to determine, as it is effectively a transient species.

PAGE 148

134 Figures 5-4 and 5-5 show the simulation results for Ge implanted and control samples, respectively. Figure parts a, b, a nd c correspond to annealing times of 1, 4 and 16 minutes at 780oC. 5.3 Constant B Concentration Experiment 5.3.1 Experimental conditions and results To further investigate the properties of th e GeB pair, an additional experiment was performed. The wafers used were float-zone (FZ) p-type Si <100> wafers, resistivity ~0.02 m cm, with the concentration of B ~4x1018 cm-3. Both of these values were confirmed via four-point probe measurement and SIMS, respectively. The use of highly doped wafers simplifies the experiment, as there is no need for B implant. The implant conditions used were similar to the previous experiments. Both wafers received a Si 1x1015 cm-2 implant at 100keV, with liquid nitrogen cooling. The control wafer (samples E1) received an additional Si implant to ensure the continuous amorphous layer, while the Ge implanted wafer (samples E2) received a 4x1015 cm-2 at 30keV. The samples were annealed under nitrogen flow in Thermolyne quartz tube furnace and AG Associates Heat Pulse 210T Rapid Thermal Annealer (RTA). The anneal times and temperatures were: 750oC for a duration of 1 and 4 hours, and at 825oC for 1 and 4 minutes. Figures 5-6 and 5-7 show the boron profiles of control samples annealed at 750 and 825oC, respectively. The profiles exhibit an increase below 0.2 m, indicative of the B trapping on the EOR loop layer. Also, there is a broadening and an increase in magnitude of the B surface spike, coming from the B tr apping at the oxide/silicon interface. The dip in the otherwise flat B profile near the surf ace is the consequence of the B segregation to the oxide and the oxide/silicon interface.

PAGE 149

135 Comparatively, the Ge implanted samples exhibit an increase in the B profile magnitude in the region implanted with Ge, as shown in Fig. 5-8 and 5-9. These samples also show the evidence of B trapping on th e EOR loops. Both the increase in the Ge implanted region (upper 50 nm) and the incr ease in the EOR region is more pronounced at the temperature of 750oC. This could be due to the hi gher interstitial supersaturation for both regions, as well as the decreasing st ability of GeB pair in the Ge implanted region, at the lower annealing temperature. The increase of B concentration in the Ge implanted region confirms there is a mechanism for retaining B in the vicinity of Ge, such as the proposed GeB pair. Wider and high er B surface spike in the control samples, compared to the Ge implanted samples, also serves to confirm the existence of a trapping mechanism which prevents B from r eaching the oxide/silicon interface. A much stronger increase in the B profile at the lower annealing temperature poses several questions. First, woul d the B profile increase more during an anneal at a lower temperature, such as 600oC? Second, could the increase be forming prior to the regrowth of the amorphous layer? Third, as the increase in the Ge implanted region leaves no B dip around it, could it be that the increase is just a matter of SIMS ioniza tion efficiency as the silicon changes from amorphous to the crystalline phase? In order to answer these que stions, anneals at 560 and 600oC were performed. The times of the 560oC anneals, 45 and 90 minutes, allo wed for observation of B profile shortly after the regrowth of the amorphous layer. The resulting B profiles show no difference from the as-implanted condition (Fig. 5-10). Therefore, if the increase of the B profile magnitude in the presence of Ge is a consequence of a change in the SIMS ionization efficiency, it re quires a certain time to achie ve the proximity between

PAGE 150

136 individual Ge and B atoms in the crystal lat tice. Similar conclusions can be drawn from the samples annealed at 600oC (Fig. 5-11), as there is no in crease of the B profile in the Ge implanted region up to 90 minutes. 5.3.2 Simulations The phenomena observed in the control sample s are associated with the capture of B at the oxide/silicon interface and the EO R loops, as described in chapter 4. Applying the Ge-B pairing model, described earlier in this chapter, with the same parameter set on the Ge implanted samples provides the simulation results shown in figures 5-12 and 5-13. Simulated profiles at 750oC seem to have the appropriate increase in the B profile, though the increase is closer to the surface in the e xperimental profiles. The small increase over as-implanted profiles during the anneal at 825oC is overstated in the simulated profiles, and similarly displaced as the 750oC. Therefore, one can conclude the Ef GeB ~0.5eV is the upper limit of the GeB pair formation energy. 5.4 Conclusions An experiment is designed to inve stigate the Ge and B interaction in preamorphized silicon. The Si preamorphi zation provides a co mmon amorphous layer depth, EOR damage distribution, and c onsequentially, the same interstitial supersaturation in the control and Ge implante d samples, with respect to depth and anneal time. The preamorphization also significantl y reduces the formation of BICs, thus simplifying the diffusion of B in the control samples. Assuming the GeB pair formation energy of 0.5eV, a model was implemented to simulate the implant and annealing conditions yi elding reasonable fits to the experimental profiles. In order to confirm the existence of GeB pair, the same set of preamorphizing implants were performed on highly doped p-ty pe wafers. Upon annealing, the control

PAGE 151

137 samples exhibited a drop of B concentrati on near the surface, a ppropriately modeled by the BI trapping at the interface. In the same conditions, the Ge implanted samples exhibited an increase of B concentration. As the simulation slightly overstated the increase, one can conclude the upper boundary of the GeB pair formation energy is 0.5eV.

PAGE 152

138 Figure 5-1. Profiles of B and Ge in as-implanted sample 2A

PAGE 153

139 Figure 5-2. Cross-section TEM (X TEM) image of sample 2A under 100000 magnification

PAGE 154

140 a) b) Figure 5-3. Boron profiles during the annealing at 780oC for: a) Ge implanted sample, b) control sample

PAGE 155

141 a) b) c) Figure 5-4. Simulated B profiles of Ge implanted sample during anneals at 780oC

PAGE 156

142 a) b) c) Figure 5-5. Simulated B profiles of control sample during anneals at 780oC

PAGE 157

143 Figure 5-6. B profiles from contro l sample during annealing at 750 oC

PAGE 158

144 Figure 5-7. B profiles from contro l sample during annealing at 825oC

PAGE 159

145 Figure 5-8. B profiles from Ge implan ted sample during annealing at 750 oC

PAGE 160

146 Figure 5-9. B profiles from Ge implan ted sample during annealing at 825 oC

PAGE 161

147 Figure 5-10. B profiles from Ge implan ted sample during annealing at 560 oC

PAGE 162

148 Figure 5-11. B profiles from Ge implan ted sample during annealing at 600 oC

PAGE 163

149 a) b) Figure 5-12. Simulated B profiles from Ge implanted samples during annealing at 750oC

PAGE 164

150 a) b) Figure 5-13. Simulated B profiles from Ge implanted samples during annealing at 825oC

PAGE 165

151 CHAPTER 6 SUMMARY AND FUTURE WORK 6.1 Summary The work presented in this dissertation covers phenomena involving the diffusion of boron in silicon. These phenomena include th e interaction of B with self-interstitials, Ge atoms and oxide/silicon interface. The chapter on boron-interstitial clusters (BIC) discusses the formation of BICs in the crystalline and amorphous material. The negligible formation of BICs in the amorphous, and subsequently regrown layer is observed experimentally. Under the same annealing conditions, the B remaining in th e crystalline material clustered almost completely with minimal observable diffusi on. The model developed to capture the reactivation behavior of B captu res several important characte ristics of the experimental observation. First, the B marker immediatel y under the amorphous/c rystalline interface clusters completely, with mi nimal diffusion. Second, the marker layer in the amorphous material experiences the enhanced diffusion w ithout forming a clustering peak. Third, the marker layer removed from the peak of the excess interstitials experiences the enhanced diffusion and forms a small clustering peak. The chapter also discusses the dissolution of BICs with respect to the annealing ambient. The change from the inert to an oxidizing ambient is demonstrated to reduce dissolution of BICs, thereby sl owing the reactivation of B. Additionally, the reactivation appears to have two distinct regimes. The reactivation during the s hort times appears to be faster than at longer times, and fairly insensitive of ambient. At longer annealing

PAGE 166

152 times, the reactivation depends strongly on ambient. The described features are qualitatively captured in the model describing the reactivation time dependence, its dependence on the annealing ambient, as well as the distinction in the cluster formation between the amorphous and crystalline phase. The interaction between the Ge and B is examined unde r the inte rstitial supersaturation. The silicon preamorphizat ion, the end-of-range damage and its consequential interstitial supersaturation se rve to provide a comm on environment across implant conditions. During the course of this investigation, only the samples having no Ge implant exhibited an interesting dose lo ss and broadening of the surface spike. The surface spike is generally assumed to be an artifact of the transition between silicon dioxide and silicon matrix, not dependent on B concentration in oxide or silicon. Thus, the monotonic increase of B surface spike with control sample annealing time, combined with the dose loss at longer times, suggested a trapping mechanism at the oxide/silicon interface. The inclusion of an equation descri bing a trapping process yielded fits to the tail diffusion and surface spike increase, w ithout altering other models or parameters. The Si preamorphized, Ge implanted samp les exhibited no change in the B surface spike, indicating the Ge interaction with B prevented the B diffusion to the surface. The negligible difference in the tail diffusion l eads to conclusion that the BI releases its interstitial upon reacting w ith Ge, forming a GeB pair. Taking an increase in the activation energy of B diffusion as the forma tion energy of the GeB pair, a model yielded a reasonable fit to the experimental data. An additional experiment showed increase of B concentration above the as-implanted level of the highly doped p-type wafer, thus confirming the existence of the GeB pair.

PAGE 167

153 6.2 Future Work To fully characterize the inte raction of B and Ge in sili con, one should investigate the full spectrum of Si1-xGex materials. As the material was not available during the time of this dissertation, an experimental design is presented in this chapter. The experiment aims to separate the influences of strain and the Ge chemical influence. 6.2.1 Material Specification – B Marker Layers in Relaxed Si1-xGex For this experiment, relaxed Si1-xGex material of 1 m thickness with flat Ge concentration, with <1e6 [cm-2] lattice defects is required. Si capping layer should be around 10 nm. Interface Si cap/SiGe is the origin of the coordinate system. Germanium contents desired are 0, 5, 10, 20, 30, 40 and 50 % Ge in Si, one wafer per Ge content. Relaxed Si1-xGex layer should be grown on top of a graded layer. Grading of 10 % Ge/ m mentioned in the literature, is used in Fig. 6-1a. Ge profile in the 1 m relaxed Si1-xGex material is flat, with four exceptions in 5, 10 and 20 % Ge material. For these Ge contents, four strained layers are desired. Two 10 nm thick compressively strained layers are desired at 430-440 and 560-570 nm depth, with double Ge content of surrounding material. Also, two 10 nm thick tens ile strained layers are desired at 460-470 and 530-540 nm depth, with 0% Ge content. The thickness of 10nm is less than the critical thickness of 20% Ge strained layer, according to the Mathew s-Blakeslee criterion. Boron profiles have the sa me specification for all Si1-xGex wafers. Within the flat part of Ge profiles are three B marker layers (delta doped), positioned at depths of 0.2, 0.5 and 0.8 um, as illustrated in Fig. 6-1b. Peak concentrations of these markers are 2x1019, 1x1018 and 1x1018 [cm-3], respectively.

PAGE 168

154 6.2.2 Information Expected This material specification allows the measurement of B diffusivity in intrinsic (deepest B marker) and extrinsic (shallowest B marker) case with respect to Ge content in relaxed Si1-xGex. The middle marker layer diffuses through tensile and compressively strained Si1-xGex layers, which may elucidate the influe nce of strain, and can be directly compared to the deepest B marker. The strained silicon cap allows for the definitive point defect injection through the oxidi zing and nitridizing ambients useful in determining the fractional interstitial and vacancy diffusiviti es of B with respect to Ge content.

PAGE 169

155 a) b) Figure 6-1. Dopant profiles in a Si1-xGex material. Figure a) c ontains Ge profile for Si0.8Ge0.2 material, while figure b) has B profile

PAGE 170

156 Figure 6-2. Boron profile in Si0.8Ge0.2 material

PAGE 171

157 LIST OF REFERENCES [Azi99] M. J. Aziz, Appl. Phys. Lett. 70, 2810 (1997) [Bau69] R. Bauerlein, Radiation damage in solids, D. S. Bellington, Ed., p. 358, Academic Press, New York (1969) [Bra95] H. Bracht, N. A. Stolwijk, a nd H. Mehrer, Phys. Rev. B 52, 16542 (1995) [Bra98] H. Bracht, E. E. Haller, and R. Clark-Phelps, Phys. Rev. Lett. 81, 393 (1998) [Cat98] M. J. Caturla, M. D. Johnson, and T. Diaz de la Rubia, Appl. Phys. Lett. 72, 2736 (1998) [Cha96] H. S. Chao, P. B. Gr iffin, and J. D. Plummer, Appl. Phys. Lett. 68, 3570 (1996) [Chr81] L. A. Christel, J. F. Gibbons, and T. W. Sigmon, J. Appl. Phys. 52, 7143 (1981) [Cla02] M. H. Clark, K. S. Jones, T. E. Haynes, C. J. Barbour, K. G. Minor, and E. Andideh, Appl. Phys. Lett. 80, 4163 (2002) [Cow99] N. E. B. Cowern, G. Ma nnino, P. A. Stolk, F. Roozeboom, H. G. A. Huizing, J. G. M. van Berkum, F. Cristiano, A. Clav erie, and M. Jaraz, Phys. Rev. Lett. 82, 4460 (1999) [Dut91] D. Dutartre, G. Brmond, A. Soui fi, and T. Benyattou, Phys. Rev. B 44, 11525 (1991) [Eag94] D. J. Eaglesham, P. A. Stolk, H.-J. Gossmann, and J. M. Poate, Appl. Phys. Lett. 65, 2305 (1994) [Fah85] P. Fahey, G. Barbuscia, M. Moslehi, and R. W. Dutton, A ppl. Phys. Lett. 46, 784 (1985) [Fah89a] P. M. Fahey, P. B. Griffin, and J. D. Plummer, Rev. Mod. Phys 61, 289 (1989) [Fah89b] P. Fahey, S. S. Iyer, and G. J. Scilla, Appl. Phys. Lett. 54, 843 (1989) [Fai81] R. B. Fair, Impurity Doping Processes in Silicon, edited by F. F. Y. Wang, p. 315, North Holland, Amsterdam, (1981) [Fan96] T. T. Fang, W. T. C. Fang, P. B. Griffin, and J. D. Plummer, Appl. Phys. Lett. 68, 791 (1996)

PAGE 172

158 [FLO02] FLorida Object Oriented Process Si mulator, University of Florida, Florida 2002. [Gos97] H.-J. Gossmann, T. E. Haynes, P. A. St olk, D. C. Jacobson, G. H. Gilmer, J. M. Poate, H. S. Luftman, T. K. Mogi, and M. O. Thompson, Appl. Phys. Lett. 71, 3862 (1997) [Gri00] M. Griglione, T. J. Anderson, Y. M. Ha ddara, M. E. Law, K. S. Jones, and A. van den Bogaard, J. Appl. Phys. 88, 1366 (2000) [Gri01] M. Griglione, T. J. Anderson, M. E. Law, K. S. Jones, A. van den Bogaard, and M. Puga-Lambers, J. Appl. Phys. 89, 2904 (2001) [Gri85] P. B. Griffin, P. M. Fahey, J. D. Plummer, and R. W. Dutton, Appl. Phys. Lett 47, 319 (1985) [Had00] Y. M. Haddara, B. T. Folmer, M. E. Law, and T. Buyuklimanli, Appl. Phys. Lett. 77, 1976 (2000) [Hu74] S. M. Hu, J. Appl. Phys. 45, 1567 (1974) [Imp06] http://eserver.bell.ac.uk/semicd/semi_p/topics/implant/imp_e2.htm, Aug 2006 [Jai03] S. H. Jain, P. B. Griffin and J. D. Plummer. J. Appl. Phys. 93 (2), 1060 (2003) [Jon96a] K. S. Jones, R. G. Elliman, M. M. Petravic, and P. Kringhj, Appl. Phys. Lett. 68, 3111 (1996) [Jon96b] K. S. Jones, J. Liu, and L. Zhang, Proceedings of Spring Electrochemical Soc. Meeting, 96-4, 116 (1996) [Kil59] Jack Kilby, US3138743, US3138747, US3261081, and US3434015 patents [Kri96] P. Kringhj, Appl. Phys. Lett. 68, 247 (1996) [Kri97] P. Kringhj and A. Nylandste d Larsen, Phys. Rev. B 56, 6396 (1997) [Kuo93] P. Kuo, J. L. Hoyt, J. F. Gibbons, J. E. Turner, R. D. Jacowitz, and T. I. Kamins, Appl. Phys. Lett. 62, 612 (1993) [Kuo95a] P. Kuo, J. L. Hoyt, J. F. Gibbons, J. E. Turner, and D. Lefforge, Appl. Phys. Lett. 66, 580 (1995) [Kuo95b] P. Kuo, J. L. Hoyt, J. F. Gibbons, J. E. Turner, and D. Lefforge, Appl. Phys. Lett. 67, 706 (1995) [Lai03] P. Laitinen, I. Riihimaki, and J. Raisanen, Physical Review B 68, 155209 (2003)

PAGE 173

159 [Lan85] D. V. Lang, R. People, J. C. Bean, and A. M. Sergent, Appl. Phys. Lett. 47, 1333 (1985) [Lev98] R. F. Lever, J. M. Bonar, and A. F. W. Willoughby, J. Appl. Phys. 83, 1988 (1998) [Lil02] A. D. Lilak, M. E. Law, L. Radic, K. S. Jones, and M. Clar k, Appl. Phys. Lett. 81, 2244 (2002) [Lil30] J. E. Lilienfeld, Met hod and apparatus for controlling electric currents, US Patent 1,745,175, application filed October 8, 1926, granted January 18, 1930; J.E. Lilienfeld, Device for controlling electri c current, US Patent 1,900,018 application filed March 28, 1928, patented March 7, 1933 [Lil97] A. D. Lilak, M. E. Law, K. S. Jones, M. D. Giles, and S. K. Earles, IEDM, 493 (1997) [Lin63] J. Lindhart, M. Scharff, and H. Schi ott, Mat.-Fys. Med. Dan. Vid. Selsk 33, 1 (1963) [Liu00] X.-Y. Liu, W. Windl, and M. P. Masquelier, Appl. Phys. Lett. 77, 2018 (2000) [May70] J. W. Mayer, L. Erikson and J. A. Davies, Ion Implantation in Semiconductors, p. 24, Academic Press, New York (1970) [Mic87] A. E. Michel, W. Rausch, P. A. Rons heim, and R. H. Kastl, Appl. Phys. Lett. 50, 416 (1987) [Mir03] S. Mirabella, E. Bruno, F. Priolo, D. De Salvador, E. Napolitani, A. V. Drigo, and A. Carnera, Appl. Phys. Lett. 83, 680 (2003) [Mog96] T. K. Mogi, Michael O. Thompson, H.-J. Gossmann, J. M. Poate, H. S. Luftman, Appl. Phys. Lett. 69, 1273 (1996) [Moo03] G. E. Moore, presentation at Inte rnational Solid State Circuits Conference (ISSCC), February 10, 2003 (http://www.intel.com/tec hnology/silicon/mooreslaw/i ndex.htm, Aug 2006) [Moo65] G. E. Moore, Electronics 38 (1965) [Mor93] N. Moriya, L. C. Feldman, H. S. Lu ftman, C. A. King, J. Bevk, and B. Freer, Phys. Rev. Lett. 71, 883 (1993) [Nic89] C. S. Nichols, C. G. van de Walle, S. T. Pantelides, Phys. Rev. Lett. 62 (1989) 1049; C. S. Nichols, C. G. van de Walle, S. T. Pantelides, Phys. Rev. B 40 (1989) 5484 [Ols88] G. L. Olson, J. A. Roth, Material Science Reports 3, 1 (1988)

PAGE 174

160 [Pac90] P. A. Packan and J. D. Pl ummer, J. Appl. Phys. 68, 4327 (1990) [Pel97] L. Pelaz, M. Jaraiz, G. H. Gilmer, H.-J. Gossman, C. S. Rafferty, D. J. Eaglesham, and J. M. Poate, Appl. Phys. Lett. 70, 2285 (1997) [Pel99] L. Pelaz, V. C. Venezia, H.-J. Gossm ann, G. H. Gilmer, A. T. Fiory, C. S. Rafferty, M. Jaraiz, and J. Barbol la, Appl. Phys. Lett. 75, 662 (1999) [Peo85] R. People, Phys Rev. B 32, 1405 (1985) [Psi06] http://www.psiberaudio. com/tube_history_2.htm, Aug 2006 [Rob92] D. J. Robbins, L. T. Canham, S. J. Barn ett, A. D. Pitt, and P. Calcott, J. Appl. Phys. 71, 1407 (1992) [Sas88] Y. Sasaki, K. Itoh, E. Inoue, S. Kishi, T. Mitsuishi, Solid State Electronics 31, 5 (1988) [See68] A. Seeger, K. P. Chik, Physica Status Solidi 29, 455 (1968) [Sol00] S. Solmi, M. Bersani, M. Sbetti, J. Lundsgaard Hansen, A. Nylandsted Larsen, J. Appl. Phys. 88, 4547 (2000) [Sto95] P. A. Stolk, H.-J. Gossmann, D. J. Ea glesham, D. C. Jacobson, J. M. Poate, and H. S. Luftman, Appl. Phys. Lett. 66, 568 (1995) [Str02] A. Strohm, T. Voss, W. Frank, P. Laitinen, J. Raisanen, Zeitschrift fur Metallkunde 93, 737 (2002) [Ura98] A. Ural, P. B. Gri ffin, and J. D. Plummer, A ppl. Phys. Lett. 73, 1706 (1998) [Ura99a] Ant Ural, Peter B. Griffin, and James D. Plummer, J. Appl. Phys. 85, 6440 (1999) [Ura99b] Ant Ural, Peter B. Griffin, and Ja mes D. Plummer, Phys. Rev. Lett. 83, 3454 (1999) [UTM00] UT-Marlowe version 5.0, Univer sity of Texas, Austin, Texas 2000. [Vic69] G. L. Vick and K. M. Whittle J. Electrochem. Soc. 116, 1142 (1969) [Vuo96] H.-H. Vuong, C. S. Rafferty, S. A. Eshr aghi, J. L. Lentz, P. M. Zeitzoff, M. R. Pinto, and S. J. Hillenius, IEEE Transactions on Electron Devices 43, 1144 (1996) [Web89] J. Weber and M. I. Al onso, Phys. Rev. B 40, 5683 (1989) [Wil01] A. F. W. Willoughby, J. M. Bonar, A. H. Dan, and B. M. McGregor, Diffusions in materials: DIMAT2000, Defect And Diffusion Forum 194-1, 717 (2001)

PAGE 175

161 [Wil84] J. S.Williams and J. M. Poate, I on Implantation and Beam Processing, p. 1, Academic Press, New York, (1984) [Win99] W. Windl, M. M. Bunea, R. Stumpf, S. T. Dunham, and M. P. Masquelier, Phys. Rev. Lett. 83, 4345 (1999) [Xia99] J. Xia, T. Saito, R. Kim, T. Aoki, Y. Kamakura, and K. Taniguchi, J. Appl. Phys. 85, 7597 (1999) [Zan01] N. R. Zangenberg, J. Lundsgaard Hans en, J. Fage-Pedersen, and A. Nylandsted Larsen, Phys. Rev. Lett. 87, 125901 (2001) [Zan03] N. R. Zangenberg, J. Fage-Pedersen, J. Lundsgaard Hansen, and A. Nylandsted Larsen, J. Appl. Phys. 94, 3883 (2003) [Zie00a] J. F. Zeigler. Ion Implanta tion Science and Technology, 214 (2000) [Zie00b] J. F. Zeigler. Ion Implanta tion Science and Technology, 303 (2000)

PAGE 176

162 BIOGRAPHICAL SKETCH Ljubo Radic was born in Zagr eb, Croatia, on August 13th, 1975. He attended the XV Gymnasium from 1989 until 1993 prior to st udies of electrical engineering at the University of Zagreb. During the first year studies he received an award for academic excellence “Josip Loncar,” and grad uated as a valedictorian on the 30th of June 1998. The graduate education commenced in August 2000, as he joined the University of Florida to study electron device processing and simulatio n. The graduate research involved boron diffusion in silicon, the transient diffusion and interaction with germanium, clustering and cluster dissolution phenomena. Upon the completion of the graduate education and the receipt of the Doctor of Philosophy degr ee, he will join the Freescale Semiconductor modeling and simulation group.


Permanent Link: http://ufdc.ufl.edu/UFE0015680/00001

Material Information

Title: Boron Interaction with Germanium and Self-Instertitials in Silicon
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015680:00001

Permanent Link: http://ufdc.ufl.edu/UFE0015680/00001

Material Information

Title: Boron Interaction with Germanium and Self-Instertitials in Silicon
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0015680:00001


This item has the following downloads:


Full Text












BORON INTERACTION WITH GERMANIUM AND SELF-INTERSTITIALS IN
SILICON
















By

LJUBO RADIC


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

UNIVERSITY OF FLORIDA


2006





























Copyright 2006

by

Ljubo Radic

































This document is dedicated to the professors that perpetually ask the question:
And why is that?















ACKNOWLEDGMENTS

I would like to thank my advisor, Dr. Mark E. Law, for the support and guidance

throughout this work. This section would be incomplete without a mention of my

colleagues from the SWAMP group, whom I owe many an interesting discussion. This

includes Ibrahim Avci, Tony Saavedra, Renata Camillo-Castillo, Robert Crosby, Robert

Robison, Chad Lindfors, Michelle Phen and others.
















TABLE OF CONTENTS



A C K N O W L E D G M E N T S ................................................................................................. iv

LIST O F TA B LE S ................. .............................................. ... ............ .. vii

LIST OF FIGURES .............................................. .......... ................. viii

A B S T R A C T .......................................... ..................................................x iii

CHAPTER

1 IN TR OD U CTION ............................................... .. ......................... ..

1.1 B rief H historical O verview .......................................................................... ...... 1
1.2 M O S T ransistor Structure.......................................................................... ......3
1.3 M O S Transistor O operation .................................. ............... ............... 3
1.3.1 Limiting Factors in MOS Performance........................................... ........... 4
1.3.2 M O S Transistor Scaling ........................................ .......... ............... 4
1.4 Sem conductor Processing Technologies ........................................ ....................5
1.4.1 Ion Implantation ................................. .. .......... ................. 6
1 .4 .2 D iffu sio n ................................................ ................ .. 9

2 LITERATURE REVIEW ............................................................ ...............22

2.1 D opant D iffusion in Silicon ...................................................................... .. .... 22
2 .2 Self-diffu sion in Silicon .......................................................................... ... .... 24
2.3 Germanium Diffusion in SilxGe ................................................................. 25
2.4 B oron D iffusion in Silicon............................................ ............................ 27
2.4.1 Boron as Interstitial Diffuser in Silicon..........................................27
2.4.2 Pre-amorphization and B .................................................................. 28
2 .5 B oron-interstitial C lu sters.....................................................................................29
2 .6 B oron D iffu sion in Si G e ........................................................ .....................3 1
2.7 M modeling B oron D iffusion in Si G e ...............................................................35

3 BORON CLUSTERING IN SILICON ........................................... ............... 62

3.1 B oron-interstitial Cluster D issolution ................................................ ............... 62
3 .1.1 Introdu action ................................................................... 62
3.1.2 Experim mental .......... .. ............................. ....... .. ............ 64



v









3.1.3 R results and D discussion ...................................................... .................65
3.1.4 M modeling ................. .... ..................................... .................. ..........67
3.2 Boron Cluster Formation and Preamorphization ...........................................68
3.2.1 Experim ent and Findings................................ ......................... ........ 68
3 .2 .2 M o d e lin g ............................................................................................... 7 0
3.3 Conclusion ................................. ............................... ......... 75

4 BORON CAPTURE AT THE OXIDE/SILICON INTERFACE AND END-OF-
RANGE DISLOCATION LOOPS ........................................ ......................... 98

4.1 B Interaction with Oxide/silicon Interface Under TED ..................................98
4.1.1 Experimental Conditions ......... .... ......... ......... .. ............... 98
4.1.2 Simulations ................................................................... ............. ...............102
4.2 Boron Segregation to the EOR Loops ..................................... ............... 104
4.2.1 Experim ental....... .............................. ............ .. .. ............ 105
4.2.2 Simulations .................. .............................. 105
4 .3 C o n c lu sio n s ................................................................................................... 1 0 6

5 BORON INTERACTION WITH GERMANIUM UNDER INTERSTITIAL
SU PER SA TU R A TION .................................................. .............................. 127

5.1 Prior Work Model Implementation ............ .............................................127
5.1.1 B andgap N arrow ing...................................................... ................... 127
5.1.2 Boron Diffusion in Si and Interaction with Ge ............ ................128
5.2 Experiment Design and Considerations...........................................................129
5.2.1 Experim ental C conditions .................................. ..................................... 131
5.2.2 E xperim ental R esults.................................................................... ... .. 132
5.2.3 Simulations .......................... ................................. ........... 133
5.3 Constant B Concentration Experiment ........................................................134
5.3.1 Experimental conditions and results ............. ...... .................. 134
5.3.2 Sim ulations ........................ ................. .... ............ ...... ...... 136
5 .4 C o n clu sio n s................................................. ................ 13 6

6 SUMMARY AND FUTURE WORK ..................................................... 151

6 .1 S u m m a ry ........................ ................. ................................................. 1 5 1
6.2 Future W ork ............... .... ..... ..... ............. ...... .. .. ..... ...............153
6.2.1 Material Specification B Marker Layers in Relaxed Sil-xGe ...............153
6.2.2 Information Expected .............................. ............... 154

LIST OF REFEREN CE S ......... .................................. ........................ ............... 157

B IO G R A PH ICA L SK ETCH ......... ................. ...................................... .....................162















LIST OF TABLES


Table page

2-1 Extracted B diffusivity prefactor (Do) and activation energy (Ea) from
experim ent of Z angenberg et al ..................................................................... .. .. 38

3-1 Retained boron dose [cm-2] in different processing steps. .....................................77

4-1 Sample labels and implant sequence, showing Ge implant dose as the only
v ariab le in th e sequ en ce ........................................ .......................................... 10 8

4-2 Active dose and carrier type measured by Hall-van der Pauw system ................109

4-3 Integrated dose from the measured B profiles in the control sample annealed at
825C, taking full profile (Qi) and ignoring the surface spike (Q2) .......................110

4-4 Integrated dose from the measured B profiles in the 4x1015 cm-2 Ge implanted
sample annealed at 8250C, taking full profile (Qi) and ignoring the surface spike
( Q 2 ) ................................................................................................... 1 1 1
















LIST OF FIGURES


Figure page

1-1 Triode electron tube schem atic ........................................ ......................... 14

1-2 Number of components per integrated circuit vs. time, later known as "Moore's
law ..... ..................................................................15

1-3 Number of transistors vs. time, shown to realize the original prediction of G. E.
M oore, as w ell as updated trends ........................................ ........................ 16

1-4 Schematic cross-section of a nMOSFET transistor...............................................17

1-5 Schem atic of an ion im planter ........................................ ........................... 18

1-6 Illustration of an ion implantation process, in particular (a) a single ion path and
(b) the resulting dam age cascade ........................................ ........................ 19

1.7 Silicon crystal viewed from (a) <110> direction and (b) tilted -100 off the
< 1 10 > d irectio n .................................................... ................ 2 0

1.8 Illustration of fluxes entering and exiting a given volume ................................21

2-1 Diffusion mechanisms in silicon: a) interstitialcy diffusion, b) kick-out, c)
vacancy diffusion, and d) concerted exchange ................................ ............... 39

2-2 Self-diffusion in Si as measured by Bracht, with symbols denoting different
sam ples ..................................... ................................... ........... 40

2-3 Self-diffusion in Si as measured by Ural et al. Dots represent measured
diffusivities, solid line is the best fit expression for self-diffusion ........................41

2-4 Interstital fraction, fsj of the self-diffusion from Ural et al. The lines represent
predictions from several metal diffusion experiments. .........................................42

2-5 Germanium diffusivities for various temperatures and Ge contents, after
Z angenberg et al ...................................................................... 43

2-6 Activation energies and prefactors for various Ge contents, after Zangenberg.......44

2-7 Activation energies and prefactors for various Ge contents, after Strohm ..............45









2-8 Comparing diffusivities of Ge in Sil-xGex as measured by Zangenberg and
S tro h m ..............................................................................4 6

2-9 Comparison of Ge diffusion with respect to ambient and consequential
interstitial or vacancy supersaturation ........................................ ............... 47

2-10 B profiles after an implant of 2x1014 at 60keV, annealed at a) 800 C and b)
9 5 0 C ................................................................................4 8

2-11 B profiles, as-grown and after Ih at 6700C .................................. .................49

2-12 Fractional activation and sheet resistance of samples annealed at 7500C for: a)
varying B dose at 80keV, and b) 4x1014 cm-2 at varying energy ............................50

2-13 B profiles used in experiment by Mirabella et al. showing the as-grown,
annealed for 2 min at 8150C, as well as illustrating the methodology for
determ ining the active fraction ........................................ .......................... 51

2-14 Time evolution of the clustered B dose during annealing at 815 to 950 oC, with
extracted tim e constants. ........................................ ...................... .....................52

2-15 Structure used by Kuo et al. with a B marker layer in the Si0.83Geo.17 strained
lay er ................. .................................... ...........................53

2-16 Diffused B profiles after 30 minute anneal at 8600C in a) Si and b) Si0.83Ge0.17 ....54

2-17 Effective B diffusivity as a function of B concentration ......................................55

2-18 D iffusivity of B in Sio.7G e0.3 layer ........................................ ....................... 56

2-19 B profiles in multiplayer structure before (dashed line) and after (solid line)
annealing at 9750C ..................................... ................... 57

2-20 Illustration of a structure used by Kuo et al. showing relaxed Sil-yGey and
pseudom orphic Si G e ........................................ .............................................58

2-21 B diffusivity at 8000C as a function of strain, with the numbers in the
parenthesis are (x,y) Ge content. Positive strain represents biaxial tension and
negative strain biaxial compression. ............................... ... ....................... 59

2-22 Schematic description of the structure used by Zangenberg et al. to measure B
diffusion in Si -xG ex .......................................... ................... .. ...... 60

2-23 B profiles before (thin line) and after (thick) 8500C annealing of a given B well
concentration ...................................................... ................. 6 1

3-1 Secondary ion mass spectrometry (SIMS) measured boron profiles of
investigated implant conditions after first anneal step, 750C for 30 minutes in
inert ambient................................... ................................ 78









3-2 Secondary ion mass spectrometry (SIMS) measured boron profiles. Initial
condition is after 750C, 30 minutes inert anneal. Subsequent anneal is 850C,
60 minutes in respective ambient. ........................................ ......................... 79

3-3 Active dose measured by Hall-van der Pauw method, during the annealing at
850C. The time zero measurement corresponds to the condition after a 750C,
30 m minutes inert anneal. ................................................ ............................... 80

3-4 Active fraction, the ratio of active to retained dose (integrated SIMS profile),
compared for the different annealing ambient at the end of the 8500C, 60
minutes anneal. The initial condition is after a 750C, 30 minutes inert anneal.....81

3-5 Major cluster formation and dissolution paths of Liu et al. with B3I and B213
containing most of the B clustered dose in figure a) and simulation of B
clustering and dissolution during thermal processing used in experiment in
figure b) .................................... ............................ ........ .......... 82

3-6 Cluster formation and dissolution paths of modified Liu et al. energetic, with
B414 and B213 containing most of B clustered dose, are shown in figure a), with
simulation of B clustering and dissolution (modified energetic) during thermal
processing used in experiment is shown in figure b) ............................................83

3-7 Boron profiles of material used in the study of Jones et al., unimplanted and
annealed in inert ambient at 800 C ................................................................ 84

3-8 Boron profiles after a 5x1015 Si implant at 146keV, annealed in inert ambient at
8 0 0 0 C ............................................................................ 8 5

3-9 Boron marker layers and excess interstitial damage following a 146keV Si
implant shows proximity of deepest marker layer to the damage............................86

3-10 Doses of important clusters in a simulation of aB 4x1014 20keV implant into
crystalline Si, during a 30 minutes 750C anneal, resulting in activation -18% .....87

3-11 Doses of important clusters in a simulation ofa B 4x1014 20keV implant into
preamorphized Si (with excess interstitial dose increased by a factor of 6),
during a 30 minutes 750C anneal, resulting in activation >85% ............................88

3-12 Doses of important clusters in a simulation of aB 4x1014 20keV implant into
preamorphized Si (at liquid nitrogen temperatures), during a 30 minutes 750C
anneal, resulting in activation >90% ............................... ........................................ 89

3-13 Simulated boron profiles after 5 seconds at 8000C, following a Si implant of
5x105 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in
the am orphous m material. ..... ........................... .........................................90









3-14 Simulated boron profiles after 30 seconds at 8000C, following a Si implant of
5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in
the am orphous m material. ..... ........................... .........................................91

3-15 Simulated boron profiles after 3 minutes at 8000C, following a Si implant of
5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in
the am orphous m material. ..... ........................... .........................................92

3-16 Simulated boron profiles after 5 seconds at 8000C, following a Si implant of
5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in
the amorphous material, and the binding energies of small BICs are reduced. .......93

3-17 Simulated boron profiles after 30 seconds at 8000C, following a Si implant of
5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in
the amorphous material, and the binding energies of small BICs are reduced. .......94

3-18 Simulated boron profiles after 3 minutes at 8000C, following a Si implant of
5x1015 cm-2 at 146keV. Diffusion of self-interstitials and BI pairs are annulled in
the amorphous material, and the binding energies of small BICs are reduced. .......95

3-19 Simulated clustered doses in BIC experiment for B implant 4x1014@ 20keV,
model presented earlier in the chapter .......... ............................... ............... 96

3-20 Simulated clustered doses in BIC experiment for B implant 4x1014@ 20keV,
model parameters adjusted to cover regrowth behavior .......................................97

4-1 Profiles of B and Ge in as-implanted sample A ............................ .................. 112

4-2 Cross-section TEM (XTEM) images of samples (a) A and (b) D under 100000
m agnification ................................ ........................... .. .... ......... 113

4-3 B profiles from Ge implanted samples during annealing at 700C ...................114

4-4 B profiles from Ge implanted samples during annealing at 825C ...................115

4-5 SIMS profiles of Ge, As and B in samples: A, with 4x105 cm-2 Ge
(lmin@825C), B with 1.2x1015 cm-2 Ge (as-implanted) and C with 4x1014 cm-2
Ge (as-implanted) per figure a, b, c, respectively ..........................116

4-6 B profiles from control sample during annealing at 700C ..................................117

4-7 B profiles from control sample during the annealing at 825C .............................118

4-8 Simulated B profiles of control sample during anneals at 825C ...........................119

4-9 Simulated B profiles of control sample during anneals at 8250C, diffusion
enhancement factor reduced five times...... ......... ................................... 120









4-10 Simulated B profiles of control sample during anneals at 8250C, assuming
surface trapping of BI .................. ........................... .. ...... ... ........ .... 121

4-11 Simulated B profiles of 4x1015 cm-2 Ge implanted sample (As contaminated)
during anneals at 8250C, assuming electric field effect dominant.........................122

4-12 B profiles from sample El during annealing at 750 C.......................................123

4-13 B profiles from sample El during annealing at 825C.......................................124

4-14 Simulated B profiles ofB segregation to the loops in sample El during
annealing at 7500 C ...................... .................... .......................... 125

4-15 Simulated B profiles ofB segregation to the loops in sample El during
annealing at 8250C ...................... .................... .......................... 126

5-1 Profiles of B and Ge in as-implanted sample 2A ................................................ 138

5-2 Cross-section TEM (XTEM) image of sample 2A under 100000 magnification..139

5-3 Boron profiles during the annealing at 7800C for: a) Ge implanted sample, b)
co n tro l sam p le .................................................................... 14 0

5-4 Simulated B profiles of Ge implanted sample during anneals at 780C ................141

5-5 Simulated B profiles of control sample during anneals at 780C..........................142

5-6 B profiles from control sample during annealing at 750 C...............................143

5-7 B profiles from control sample during annealing at 825C...............................144

5-8 B profiles from Ge implanted sample during annealing at 750 C ........................145

5-9 B profiles from Ge implanted sample during annealing at 825 C ........................146

5-10 B profiles from Ge implanted sample during annealing at 560 C ........................147

5-11. B profiles from Ge implanted sample during annealing at 600 C.........................148

5-12 Simulated B profiles from Ge implanted samples during annealing at 750C.......149

5-13 Simulated B profiles from Ge implanted samples during annealing at 825C.......150

6-1 Dopant profiles in a Sil-xGex material. Figure a) contains Ge profile for Sio.8Ge0.2
m material, while figure b) has B profile ....................................... ............... 155

6-2 Boron profile in Sio.8Ge0.2 m material ............................................. ............... 156















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

BORON INTERACTION WITH GERMANIUM AND SELF-INTERSTITIALS IN
SILICON

By

Ljubo Radic

August 2006

Chair: Mark E. Law
Major Department: Electrical and Computer Engineering

Understanding the diffusion phenomena in silicon is imperative for the rapid and

efficient development of the future semiconductor processes. Boron is pertinent as the

dopant of choice in formation of a p-type semiconductor, yielding higher activation and

carrier mobilities than the alternatives. In order to overcome certain transient behavior,

such as transient diffusion enhancement and clustering, these phenomena have to be

characterized and modeled.

One part of this dissertation describes the investigation of the properties of the

significant boron-interstitial clusters (BICs). Previous models described boron clustering

through the single dominant cluster, assumed to be interstitially poor. An experimental

investigation of the BIC dissolution with respect to annealing ambient revealed a reaction

contrary to the expectations from that model, suggesting that one of the dominant clusters

may be interstitially rich. Additionally, the time dependence of the reactivation indicated

the reactivation process was coming from two sources. Using a set of ab-initio









calculations as a starting point, adjusting to dependence on the annealing ambient, the

model showed appropriate reaction. An earlier experiment probed the formation of the

BICs with a separation by amorphous/crystalline interface. Assuming the inability of the

self-interstitial or boron-interstitial pair to enter the amorphous layer significantly reduces

the cluster formation. However, further adjustments were necessary to properly capture

the difference between the enhanced diffusion phenomena in the regrown amorphous

layer, and a distinct clustering peak in the crystalline material.

Another part of the dissertation examines the interaction of Ge and B in silicon

under interstitial supersaturation. Experimentally observed segregation of B onto

oxide/silicon interface in samples with no Ge implant did not occur in the presence of Ge.

In conjunction with increase in peak concentration of B in the presence of Ge, this

experimental evidence supports an earlier hypothesis of the Ge-B pairing as the cause of

the diffusivity reduction of B in Sil-xGex. Modifying the model to account for the

interstitial supersaturation and extracting the pair formation energy from a relative

increase in the diffusion activation energy in higher Ge concentration, the model

qualitatively captures the phenomena.














CHAPTER 1
INTRODUCTION

1.1 Brief Historical Overview

The twentieth century saw several significant changes in the fields of engineering.

The advent of the vacuum tube gave rise to the field of electronics. It consisted of the

filament as the emissive element, cathode, anode and the control grid (Fig. 1-1). With

negative voltage applied to the grid the cathode current could be cut off, while increasing

the control voltage amplified the current. Amplification of the signal made long distance

phone and radio communication practical.

One of the significant benefits of the vacuum tubes is the simplicity of

construction. In the 1930s, Julius Edgar Lilienfeld filed two patents for controlling

electric current [Lil30] which would today be described as IMESFET (metal-

semiconductor field effect transistor) and depletion-mode MOSFET (metal-oxide-

semiconductor field effect transistor). Since the clean environment and surface control

required to produce viable semiconductor material were not available before the middle

of the twentieth century, his inventions were forgotten until twenty years later.

Another application of cathode tubes was in early computers. Electronic Numerical

Integrator And Computer (ENIAC) was the first electronic computer, built from 1943 to

1946, and it contained 17,468 vacuum tubes, 7,200 crystal diodes, 1,500 relays, 70,000

resistors, 10,000 capacitors and around 5 million hand-soldered joints. It required

rewiring for execution of a new program, and could perform 5000 additions per second

while consuming 160kW of power.









The first fully functional transistor was developed at Bell Laboratories, by John

Bardeen, Walter Houser Brattain, and William Bradford Shockley, who were awarded the

Nobel Prize in physics in 1956. Though their goal was to use the high quality germanium

crystals to form a field-effect transistor (FET) predicted by Lilienfeld, but they actually

found a bipolar junction transistor (BJT). After some time, and a number of solved

problems, the percentage of functional transistors (yield) was still not very high and

production of high purity germanium still posed a problem. In the 1960s, Texas

Instruments tried using silicon instead of germanium, which was easier to work with.

Several factors led to the ascent of the silicon based MOSFETs: water insoluble native

oxide (Si02), physical and electrical properties of the oxide, no interface states between

the silicon and the oxide that could serve as electron traps or recombination sites, as well

as improvements in production facilities. By that time, transistors replaced majority of the

vacuum tubes.

The trend of miniaturization continued as Jack Kilby [Kil59] at Texas Instruments

patented a "solid circuit" in germanium. Soon afterwards, Robert Noyce of Fairchild

Semiconductor was awarded a patent for a "unitary circuit" made of silicon. Planar

MOSFET technology scaling has been driving the semiconductor technologies since

then. As expressed in an observation by Gordon E. Moore [Moo65], the number of

transistors per square inch on integrated circuits had doubled every year since the

integrated circuit was invented (Fig. 1-2). This growth continues, as shown in Fig. 1-3

[Moo03], to this day. The rest of the introduction will concentrate on MOS technology as

the dominant technology today.









1.2 MOS Transistor Structure

MOSFET transistors fall into the category of unipolar transistors, meaning that

only one carrier type is responsible for conduction. While different types of FETs control

the conduction between the source and the drain (S/D) electrodes in different ways, all of

them operate with either electrons or holes. This is not the case for the BJT, where both

carrier types are involved.

MOS transistors can be divided into several types. First, the transistor can have a

n-type or a p-type channel. Second, the transistor can have a channel existing or not, with

no voltage applied to the gate electrode. In case there is an electrical connection existing

between the source and the drain region, the device is said to be a depletion mode device.

Otherwise, there is no connection unless the proper voltage is applied to the gate, such as

in enhancement mode device. Most transistors in digital circuits are used as switches, for

which the enhancement mode device is more suitable.

1.3 MOS Transistor Operation

The schematic cross-section of an enhancement mode nMOSFET is shown in Fig.

1-4. The external contacts are source, drain, and gate. Body contact is usually connected

to the source, thus physically differentiating between the source and the drain. This also

forms a reverse polarizedpn junction at the drain-body boundary, as VDs>0 is necessary

for nMOS operation. For a VGs lower than the threshold voltage VT, the p-type region

separates the source and the drain allowing no current flow, and the device is said to be

turned off. As the VGS increases to values over the VT, the electron rich region forms a

channel near the surface between the source and the drain. This facilitates the current

flow from the drain to the source, and the device is said to be turned on. Therefore, the









gate voltage VGs controls the current flow between the drain and the source with the

transistor acting as a switch.

1.3.1 Limiting Factors in MOS Performance

The MOS device performance is limited by the design and the parasitic effects. The

design parameters, such as gate length, width, and oxide thickness, gate material, source

and drain region thickness, factor in to determine the transistor's performance. The

parasitic effects are leakage currents, parasitic resistances and capacitances. The

technologically necessary overlap of the gate over the source and the drain region forms

two capacitors: CGS and CGD. Even though the capacitance amounts are usually similar,

the latter is electrically more significant. During the switching from an off state into an on

state, the transistor operates as a small signal amplifier. The increase in the small signal

difference between the gate and the drain effectively increases CGD by the voltage

amplification factor. The parasitic resistances include the resistances in the S/D regions,

their extensions, transition resistances towards metal contact, and internal resistance of

the poly-silicon gate electrode.

1.3.2 MOS Transistor Scaling

The continued advances of semiconductor technology, expressed in transistor

density and speed, are due to the scaling of the planar technology on silicon and the

properties of silicon dioxide. Reduction in gate length had increased the current and

reduced the input capacitance, both of which reduce switching time. Decreasing the

dimensions of the transistor also meant that more transistors were available per chip, thus

increasing the computing power.

The decrease of the dimensions of the transistor does pose some technological

difficulties. The breakdown electric field of the silicon dioxide is a material property one









can not circumvent. Thus the scaling limited by the maximum electric field is labeled

constant field scaling, requiring the reduction of the gate voltage in proportion to the

reduction of the oxide thickness. As the source and the drain regions come closer

together, the depletion regions could touch or merge, giving rise to a leakage current

irrespective of the gate voltage. This is known as sub-threshold leakage, causing

problems with the power consumption and the detection whether the transistor is on or

off. Similarly, the reduction of the lateral dimensions increases the capacitances between

metal lines. The reduction of physical dimensions also requires reduction in operating

voltages in order to keep below the breakdown electric field of the oxide. Though the

operating voltage is reduced, the extremely small oxide thickness poses a problem as

tunneling becomes significant with several monolayer thick oxides. New materials with

higher dielectric constants, allowing for thicker layer, have also been introduced. In order

to keep the conduction of the transistor under the control of the gate, the conduction must

take place on or near the surface. With reduction of operating voltages that area is

reduced as well. This means that source and drain region depths have to be scaled with

the lateral dimensions. The parasitic resistances of S/D are inverse proportional to the

depth, thus requiring increased doping to keep them fixed. The contact material and

resistance also play a role in the parasitics, somewhat mitigated by the recent switch from

Al to Cu for metal interconnects.

1.4 Semiconductor Processing Technologies

Semiconductor integrated circuits are formed by an application of following

processes: (1) growth or a deposition of a layer, (2) exposure to light, (3) patterning or

etching, (4) dopant implantation, and (5) thermal annealing. The latter two are the









frequently used to form the n-type or p-type semiconductor regions, later to become a

part of the device, and will be discussed in further detail.

1.4.1 Ion Implantation

Ion implantation is the dominant method of introducing dopant into silicon

material. The accuracy of the implanted dose and the dopant distribution in depth are the

reason for its widespread use. The process of ion implantation begins with a gas or a solid

source providing an ionized beam (Fig. 1-5). The ion beam then goes through a magnet

and an aperture, providing the mass separation of elements and isotopes extracted from

the source. The voltage used to accelerate the ions from the source is called the extraction

voltage. The extraction voltage and the acceleration voltage both contribute to the final

ion's implant energy. The ion beam passes through the scanning plates in x and y

directions, and the deflection before reaching the target wafer. The wafer sits in a Faraday

cage that repels the incidental electrons and integrates the incoming charge from the ion

beam. The direct and accurate measurement of the incoming ion beam current, also the

implanting dose rate, provides the accuracy in the total implanted dose. The

aforementioned deflection serves to remove neutrals, ions that lost the charge between

the aperture and the scan plates, from the beam reaching the target wafer and introducing

an error in the dose measurement.

The two major mechanisms of the energy loss of an ion entering the crystal lattice

are electron (Se) and nuclear (S,) stopping (Eq. 1-1), with N as the lattice density.

dE
= N (S, (E)+ S, (E)) (1-1)
ds









As the incoming ions enter the crystal lattice, they are stripped of the outer electron

shell and slowed by the interaction with ("drag" of) the bound and free electrons. The

stopping energy can be approximated by a velocity proportional, and expressed as

S, (E)- k-E (1-2)

Nuclear stopping involves Coulombic interaction between the ion and the host

lattice atom. As the ion approaches the target atom, the electrostatic repulsion deflects the

ions path, and/or removes the host atom from its lattice position. In the case neither the

ion nor the host atom occupy the lattice position, the collision has produced a vacancy

and an interstitial atom (the dislodged host atom). Alternatively, the ion performs a

replacement collision, by removing the host atom from a lattice position and occupying

it, thus becoming substitutional and creating a self-interstitial.

For practical calculations, nuclear stopping can be approximated [ZieOOa] by

8.462.10 15 *Z1Z2M1
S, (E) = 462 0-1 Z1 M3 z S()[ecV/(atoml/cm2) (1-3)
O (M, M)(Z102 +Z023)
(Mi + M)[Z +1 2

where the Z and Mare atomic number and mass, subscripts 1 and 2 denote ion and target

atoms, and the reduced energy s is

32.53M,2E
Z Z1Z2M, +MA )(Z2 +Z023) (1-4)

and the reduced nuclear stopping S,(s) can be calculated as

ln(1+1.1383E) (
fore<30: 2(0.01321216 +0.19593E5)




In(s)
for e>30: Sn(e)=ln (1-6)
2s









The occurrence of collision events is random and makes the individual ion path

rather difficult to predict (Fig. 1-6a). Though the individual ion travels a distance R, the

peak of ion distribution is the mean projected range Rp, a far more interesting value. The

projected range is defined as

RP, = X, (1-7)


where x, is the depth of a given ion i. It can also be approximated [Lin63] by

R
R = (1-8)
1+ M
3M1

In ideal case of uniform and amorphous target, with collision frequency and energy

transfer per collision being random, the implanted ion distribution can be described by a

Gaussian function N(x). Implant dose is Q, and ARp is the vertical distribution spread.

Q K (x -R )2
N(x) = Q exp ( R)2 (1-9)
AR -- 2(AR

For a crystalline target, the distribution is somewhat different. Certain directions in

the crystal structure allow for less collisions and deeper penetration, commonly referred

to as channeling. The nuclear stopping is minimal in the case of channeling, as the angle

of incidence is rather small and the Coulombic repulsion keeps the ion in the channel.

The electron stopping is smaller, thus allowing the ion to travel significantly longer

distances. Fig. 1.7 illustrates the difference between looking into a channel and when the

tilt is applied. Wafer tilt and twist are used to prevent channeling by providing seemingly

random distribution of silicon atoms on the surface plane. Also, a sacrificial oxide layer









called screening oxide is sometimes used to randomize the incoming ion's trajectories

before entering the silicon lattice.

During a single ion's path, it may displace as much as 104 silicon atoms by nuclear

collisions before it comes to rest (Fig. 1-6b). If the primary ion collides with a host lattice

atom with more than the energy of displacement Ed (Ed is 14 eVfor silicon [Bau69]), the

now interstitial atom continues to travel through the lattice until it collides with another

lattice atom, thus creating additional damage cascades. The damage depends on the

atomic species, as heavier atoms lose most of their energy in collisions, while lighter

atoms initially lose a significant portion of their energy due to electronic stopping. In case

of high implant doses, the damage can displace a percentage of the lattice atoms (over

10% [Chr81]), creating amorphous zones. The amorphization is sometimes used as an

introductory step in low energy implants to prevent channeling, called pre-amorphization

implant (PAI).

After the implantation step is complete, there can be a significant interstitial and

vacancy concentrations in the implanted region. Only a small portion of the implanted

dopant atoms has occupied the lattice positions rendering them electrically inactive.

These damaged regions need to be restored to a monocrytalline lattice and dopants

activated, which is accomplished by a high temperature thermal treatment.

1.4.2 Diffusion

Fick's laws describe diffusion of given specie, as the change of concentration with

respect to position and time. Diffusion is a thermally activated process, so one can write

the probability p of an atom species Xhop between two lattice positions as


p =exp _AG" (1-10)
kT







10


where AG' is migration energy of a species X, and k is the Boltzmann constant. The

frequency of hop attempts could be approximated by the Debye frequency v, which

makes the frequency of hops species Xmakes to a neighboring site

F=v.p (1-11)

Now, assuming two lattice planes separated by a distance a, with concentrations nl

1
and n2, jump frequency of -F in each direction, during a small interval of time t, the
2

atom Xhops can be written as


plane 1 -> plane 2



plane 2 -> plane 1


1
FT.n, .t
2

1
1F 7-n2 "t
2


(1-12)



(1-13)


thus resulting in the flux J, over an area A.


J = L(n -n2)
2A

If one expressed the above equation in terms of concentration c,

n
C =
A-a


J = -a(c -c2)
2


(1-14)





(1-15)



(1-16)


and applied it over the incremental distance cx, it gives the formula of the Fick's first

law.


1 2c 9c
J =-F -.a2 -D
2 Ox ax

Given the above equation, one can express the diffusivity as


(1-17)









B ]x (1-18)
D= g.F.a2 =g.v.a2 .exp AkGx (1-18)


1 1
with g as the geometric factor, g = -for one dimensional diffusion, and g = -for a
2 6

three dimensional problem. The migration energy consists of enthalpy H and entropy S,

as in the following equation

AGx = AH\ T -AS' (1-19)

Incorporating that into the diffusivity equation yields the final expression,


D =g. v.a2 exp exp X = DO- exp --X (1-20)
k kT kT

with Dox as the prefactor, and the Qx as the activation energy.

The first Fick's law (derived above) describes the relation of flux with

concentration, though it has no dependence on time. The second Fick's law defines the

relationship of the concentration with time and position.

If a flux J1 entering the box is greater than the flux J2 exiting the box, the

concentration in the box must increase in accord with the conservation of matter

principle. The principle of Fick's second law is described in equations below

8c AJ A J2 J1 J2 J(
.... (1-21)
at A-x x ax

The number of dopants entering the box per unit time from the left is AJI, dopants

leaving the box on the right is AJ2, and their difference divided by the volume is equal to

concentration per unit time. Using the equation from the first Fick's law, the equation

governing the diffusion of dopant look like this in one dimension.

ac a J 2 (1-22)2
D ac)= D (1-22)
at ax Wx x ax









In three dimensions it yields a bit longer expression.

ac OJ OJ OJ
=- +-+- =-V J= V (DVc)= DV (Vc)= DV2c (1-23)
at [8x 8y 8z) = '

2 2 a2 a2
V2 + + (1-24)
x 2 ay2 + z2

There are several conditions where the solution of Fick's laws provides general and

frequently used solutions. In a limited source diffusion, assuming an instantaneous source

depositing a fixed amount Q of impurity atoms on the surface of otherwise pristine

silicon, the diffused profile would be in the form of


c(x,t) = exp (1-25)
r-t ep -4Dt

Another interesting case is diffusion from a constant source, similar to

predeposition step, when impurities are introduced into silicon while maintaining

constant surface concentration cs. The constant surface concentration is generally

maintained through abundant source, such as polysilicon or oxide doped above solid

solubility at the predeposition temperature. The diffusing part of the dopant concentration

would then be at the solid solubility. The solution of Fick's laws with these boundary

conditions results in complementary error function,


c(x,t)= cserfc --j (1-26)


A more common case in semiconductor processing is the combination of implant

and anneal. Presuming a dose Q of a dopant was implanted with an energy E, with a

mean projected depth Rp with vertical straggle ARp, the distribution after an implant

would be










c(x)= exp
AR I2


(x R )2
2AR2
p


(1-27)


Following anneal at temperature T, with diffusivity D for a time t, the distribution

would be


(+ tQ (X_-R,)2
c(x)- AR+2D 2A +4Dt
V;ARp +2~


(1-28)


provided the profile was sufficiently far away from the surface, satisfying the following:


R2 >> 2AR2 + 4Dt
p p


(1-29)






14




"plate"
"grid"

"filament"
control
voltage


Figure 1-1. Triode electron tube schematic [Psi06]













Wb


15S







k-
".''" I0a [ .
o /










0 0 PS 'L .O P Q XJ *




Figure 1-2. Number of components per integrated circuit vs. time, later known as

"Moore's law" [Moo65]
































figure i-J. lMumoer or transistors vs. time, snown to realize me onginal prediction or U.
E. Moore, as well as updated trends [Moo03]







17


Gate
Source Drain







SBody
Figure 1-4 Schematic cross-section of a nMOSFET transistor











Yscan


Figure 1-5. Schematic of an ion implanter [Imp06]







19


(a) IMPLANTATKON ;Sdid R
SINGLE ION
Rp
-----------


(b) DAMAGE Jr1d




Figure 1-6. Illustration of an ion implantation process, in particular (a) a single ion path
and (b) the resulting damage cascade [Wil84]






20












(a) (bj
Figure 1.7 Silicon crystal viewed from (a) <110> direction and (b) tilted -10 off the
<110> direction [May70]













J,


2J


Figure 1.8. Illustration of fluxes entering and exiting a given volume














CHAPTER 2
LITERATURE REVIEW

2.1 Dopant Diffusion in Silicon

This section outlines some significant quantities influencing the diffusion of a

given dopant species in silicon.

Diffusion in the silicon crystal, according to the standing theory [Fah89a], occurs

through four mechanisms. Namely: interstitialcy and kick-out mechanism, which are

interstitially mediated, vacancy mediated and concerted exchange. The former three

require a point defect, an interstitial or a vacancy, respectively. Point defects interact with

the dopant atom to form a dopant-defect pair. These pairs then diffuse by a random walk

mechanism to generate net dopant diffusion. The kick-out mechanism involves an

interstitial that does not form a pair with the dopant atom. Rather, the dopant atom is

kicked out into the void, becomes an interstitial and diffuses through the channels in the

crystal lattice. Eventually the dopant atom kicks out a silicon atom forming a silicon

interstitial and occupies its lattice position. The fourth mechanism, concerted exchange,

does not require point defect presence. Rather, it consists of two atoms, a dopant atom

and a silicon atom, simultaneously exchanging positions. It is frequently neglected, as

theoretical studies suggest the activation energy for concerted exchange would be

prohibitively high [Nic89], compared to point defect mediated mechanisms. Recent

experiments attempted to estimate the concerted exchange fractional contribution in

diffusion [Ura99a]. The results of those experiments were somewhat ambiguous,









allowing for the existence of concerted exchange mechanism, but also allowing for its

fraction contribution of zero. These four diffusion mechanisms are illustrated in Fig. 2-1.

A +X -> AX, (2-1)

Ef = Ef E (2-2)
EA =Ef +E" =E -E +E" (2-3)
AX AX ~ AX AX 1 )

Equation 2-1 describes the formation of a dopant-defect pair. The dopant is denoted

as A, and the defect as X, which can be a silicon interstitial, I or a vacancy, V. Formation

of a dopant-defect pair, AXleads to a lower energy state, as given in the Eq. 2-2.

Breaking of the pair requires a binding energy, Eb which is provided by the lattice.

Motion of the pair occurs when a migration energy barrier, E" is overcome. Therefore,

the dopant diffusion process via a point defect requires the formation of the defect, its

reaction with a dopant atom in a substitutional site, to form a dopant-defect pair, which

then migrates contributing to macroscopically observable diffusion behaviour. Activation

energy for diffusion, in that case, can be written as a function of the formation energy of

the defect X, Ef, the binding energy of the dopant-defect pait AX, Eb and the

migration energy of a dopant defect pair AX, E", as shown in the Eq. 2-3. In single

crystal silicon, those energies have fixed values. In Sil-xGex alloys, those values can vary

with Ge content. For a complete picture, one should have information on both defect and

dopant pair diffusion, with respect to Ge content. Also, Ge diffusion could be a function

of Ge content, and should be investigated to allow simulation of spatially varying Ge

profiles. The experimental data available in the literature will be discussed later in this

chapter.









2.2 Self-diffusion in Silicon

Despite numerous theoretical and experimental studies on self-diffusion in silicon,

contributions from interstitial and vacancy components are uncertain. Radioactive tracer

technique results are limited by a 2.6 hour lifetime of 31Si, so monitoring heavy metal

diffusion is frequently used in attempting to extract self-diffusion parameters. With the

availability of the isotopically pure silicon sources, structures could be grown (by

molecular beam epitaxy (MBE) or chemical vapor deposition (CVD)) with isotopically

enriched or depleted layers. This approach requires no assumptions in measuring

prefactors and activation energies for interstitial and vacancy diffusion.

The study ofBracht et al. [Bra98] measured self-diffusion of Si in an inert ambient.

The samples were sealed in Ar filled capsules and annealed at temperatures ranging from

8550C to 13880C. The extracted diffusivity Ds, (Eq. 2-4) contains both interstitial and

vacancy diffusion components. Taking interstitial diffusivity measured in a Zn diffusion

experiment [Bra95] as CI*DI (Eq. 2-5), they extracted Cv*Dv (Eq. 2-6).

S 250 (4.75 .4)e1 (2-4)
D, = (530 o)exp-(4.7504 M2S] (2-4)
170 p BT )L/J


C D, = 2980exp(- 4.95eV/kT)[cm2s] (2-5)

C*Dv =0.92exp(-4.14eV/kT) [cm2/s] (2-6)

Ural et al. [Ura98] performed a similar study measuring self-diffusion through

diffusion of 30Si from naturally abundant, into isotopically depleted layers. The natural

abundance of 3.1% was reduced to 0.002% in isotopically depleted layers. During the

annealing, samples were subjected to inert, oxidizing and nitridizing ambient. Oxidizing

ambient injects silicon interstitials [Hu74], while nitridizing ambient injects vacancies









[Fah85] from the surface. This perturbation of the point defect concentrations allows the

extraction of the fractional interstitial, f and vacancy,fv contributions to self-diffusion,

without assumptions. Self-diffusivity measured in inert ambient was Ds, (Eq. 2-7),

matching the values of Bracht et al. [Bra98]. Extractedfi was in a range of 0.55-

0.6(0.1), while interstitial and vacancy diffusivities were Dsi (Eq. 2-8) and Ds, (Eq.

2-9), respectively.

Ds, = 560exp(- 4.76eV/kBT)[cm2/s] (2-7)

Ds, =149exp(- 4.68eV/k T)[cm2/s] (2-8)

Dsy = 636exp(- 4.86eV/ksT)[cm2/s] (2-9)

2.3 Germanium Diffusion in Si_-xGex

Determining Ge diffusivity with respect to Ge content is interesting, particularly for

a spatially varying Ge profile, such as base of a BJT or 2D hole gas (2DHG) transistors.

Zangenberg et al. [ZanOl] measured Ge diffusion in Sil-xGex (x<0.5), MBE grown

layers. Diffusion was measured using 0Ge as a tracer into 72Ge rich layers, during

anneals at temperatures from 850 to 1050 C. Transmission electron microscopy was

used to confirm that the dislocation density was low enough not to influence diffusivity

measurement. The diffusivities measured are shown in Fig. 2-5, with the extraction of

prefactor and activation energies in Fig 2-6.

A more comprehensive study was performed by Strohm et al. [Str02]. Various

material types used include MBE, CVD, Cz and FZ grown material, collected from a

number of academic and industrial sources. 31Si and 71Ge isotopes were used as tracers,

introduced into the material by ion implantation. Tracer dose below the threshold for

formation of extended defects ensured negligible influence of implantation process.









Transient electron microscopy (TEM) and etch pit measurements verified that the

dislocation density was under 107 cm-2. The samples were annealed at temperatures of

650 to 1250 C in inert ambient. Their extracted prefactors and activation energies for Ge

diffusion are shown in Fig. 2-7. Despite the difference in prefactors and activation

energies, the diffusivities agree quite well to the values provided by Zangenberg, as

shown in Fig. 2-8.

Large number of samples having different Ge concentrations provided a rather

small stepping in Ge content. It is interesting that such variation of sources produced such

a nicely behaving plot. A distinct change in the prefactor dependence with increasing Ge

content ocurrs around 35% Ge, not as visible in the activation energy dependence. The

authors attribute this to the change in the diffusion mechanism, from interstitial to

vacancy dominated, as one increases Ge content.

The situation is not as clear when it comes to fractional interstitial and vacancy

contributions to Ge diffusion. Initial studies of Fahey et al. [Fah89b] estimated andfv

using ammonia ambient anneals. Ammonia (NH3) annealing ambient, reacting with a

bare silicon surface, injects vacancies into the bulk [Fah85][Kri96][Kri97][Mog96]. The

same ambient, reacting with oxide (SiO2) injects interstitials. Reference was provided by

a sample coated in oxide and nitride, ensuring no surface reaction, and no point defect

injection. Samples contained a Ge marker layer grown by MBE. Peak Ge concentration

was -5%. Anneals conducted under vacancy and interstitial injection both diffused the

Ge profile more than the reference sample. At a temperature of 1050C, they estimated

= 0.3-0.4. Unfortunately, data for other temperatures was not provided.









Studies of Griglione et al. [Gri00, Gri01] concentrated on extraction off, andfv by

annealing Sio.85Geo.15 layers in inert and oxidizing ambient at temperatures ranging from

900 to 1200 C. Diffusion of buried B marker layer was monitored to verify the oxidizing

reaction on the surface injected interstitials and estimate interstitial diffusivity

enhancement. The enhanced diffusion of the buried B marker layer also showed that the

Sio.85Geo.15 layer does not significantly perturb interstitial population, meaning there was

no pile-up of silicon interstitials in Gel pair. Since the oxidizing ambient resulted in

reduced diffusion (Fig 2-9.) of Ge from the Sio.85Geo.15 layer, it was inferred that the

diffusion was vacancy mediated. To verify this, anneals in nitridizing ambient were

performed at temperatures of 1100 and 1200 C. Instead of the expected increase in

diffusivity, a considerable reduction in diffusivity was observed, thus making it difficult

to draw a firm conclusion on the available data.

2.4 Boron Diffusion in Silicon

2.4.1 Boron as Interstitial Diffuser in Silicon

Boron is generally accepted to be an interstitially diffusing species

[See68][Fah85][Fan96][Gos97]. More recent measurements [Ura99b] of B diffusion in

interstitial and vacancy supersaturation, gaveft values higher than 0.84, assuming the

vacancy concentration was unperturbed. A more strict assumption on the recombination

eq ox Ceq nit
of equilibrium point defects < V < 1,- < <1 yields values above 0.94.
C1ox Ceq C nit C~eq
I J V V

The exact mechanism of B microscopic diffusion is not certain, between interstitial pair

diffusion and kick-out mechanism. Recent ab-initio calculations [Win99] suggest B

diffusion is mediated by a BI pair. The pair binding energy was calculated to be 0.8eV,

while the migration energy was 0.2 eV. According to the Eq. 2.5, the activation energy









for B diffusion would also depend on the formation energy of an interstitial. Since the

values of point defect formation and migration energies used in FLOOPS are somewhat

different than the ones used in this study, a free parameter is necessary to match the

macroscopically observed activation energy. In this case, the binding energy of a BI pair

is used as a free parameter. The value of 0.5 eV was used to match the B diffusivity

values of FLOOPS defaults.

2.4.2 Pre-amorphization and B

B implants into crystalline silicon material seldom provide the optimal results, due

to channeling, solid solubility limit and the deactivation to form boron-interstitial clusters

(BICs). Low energy implants require implantation in deceleration mode in order to

provide viable implant currents. Even then, there is a possibility of energy contamination

due to the neutralization of the charged species before reaching the decelerating stage.

Once the B ion enters the silicon lattice, a certain portion of the implanted dose channels,

increasing then junction depth. Increasing the dose to reduce the parasitic resistances

can result in BIC formation, thus decreasing active B concentration below solid solubility

limit.

Some of these detrimental effects can be mitigated by pre-amorphizing the silicon

crystal. The disordered structure of the amorphous silicon prevents channeling [ZieOOb],

thus reducing the pn junction depth. Activation can also benefit from the

preamorphization, as the amorphous layers can be regrown at low temperatures (450-

600C) incorporating dopant concentrations higher than solid solubility at the regrowth

temperature. The spatial separation of the end of range (EOR) damage and the dopant

suppresses the formation of BICs [Jon96a]. Germanium is the popular choice of









amorphizing implant species due to producing less damage than the silicon implant for

the same implant dose [Cla02].

2.5 Boron-interstitial Clusters

Dopants are mostly introduced into a Si wafer via implantation process, which is

followed by an annealing process. During the thermal process, the damaged crystalline

lattice is repaired and the dopants fall into the substitutional positions, thus becoming

electrically active.

However, the annealing process following a B implant under certain conditions,

such as yielding a peak concentration above 1x1019 cm-3, does not activate all the

dopants. The implanted profile splits into two regions, the immobile peak and the rapidly

diffusing tail, as first observed by Michel et al. [Mic87]. The immobile portion of the

profile is particularly interesting as it is at a concentration much lower than the solid

solubility of B in Si, above 4x1019 cm-3 for temperatures over 800C [Vic69]. In the

comparison between the annealed profiles at 800 and 950 C, immobile peak is far more

pronounced at the lower temperature (Fig. 2-10). The same is the case with the diffusion

in the tail regions, as the transient enhanced diffusion (TED) is reverse activated, having

higher diffusion enhancement at lower temperatures.

Stolk et al. [Sto95] investigated the formation of the immobile peak portion, by

implanting Si into a wafer containing B 6-doped layers. The Si 5x1013 cm-2 implant at 40

keV contained the damage into the top 0.1 tm layer, directly affecting only the B layer

marked as 1 (Fig. 2-11). The annealed B profiles show several interesting characteristics.

First and second 6-doped layers have significantly diffused tails, indicating the interstitial

supersaturation as expected after a Si implant and consequential TED. But, these 6-doped









layers also have immobile peaks, indicating that the majority of B in them is inactive.

The shouldering concentration (the intersection of the immobile peak and he diffused tail)

of the 1st and 2nd 6-doped layer is significantly lower than B solid solubility at the anneal

temperature, indicating that some mechanism other than the solubility limit is preventing

B diffusion. The 3rd 6-doped layer exhibits lower diffusion enhancement than the first

two, but it does not have a visible immobile portion. All the other 6-doped layers diffuse

closer to the equilibrium diffusivities.

This experiment shows that the interaction between self-interstitials and

substitutional B can form immobile peaks. Furthermore, it shows the excess interstitials

from the implant damage can diffuse and interact with substitiutional B forming

metastable clusters visible as immobile B peaks after an anneal. As these clusters are the

product of the interaction of self-interstitials and boron, thus called boron-interstitial

clusters.

To probe the dissolution kinetics ofBICs, Lilak et al. [Lil02] used thirteen B

implant conditions. The samples were annealed at 750C for 30 minutes in order to

deactivate the majority of the B dose. Then, the samples were followed through a number

of times during anneals at 750 or 850 C, using the Hall effect measurement and

spreading resistance profiling (SRP) to measure activation. Some measurements are

shown in the figure 2-12. These measurements indicate that the increase in dose with

constant implant energy, as well as the decreasing implant energy with constant dose,

decrease the activation upon annealing at temperatures -750C. Another important

parameter, the energy required for the reactivation of B was extracted from these

measurements to a value around 3eV.









A similar experiment was done by Mirabella et al. [Mir03], using the MBE grown

B marker layers. The shallowest marker layer was used to monitor the clustering process,

while the deeper marker layers were used to monitor the interstitial supersaturation.

These samples were implanted with 5x1013 cm-2 Si at 20keV, providing the excess

interstitials for the clustering process. The active dose was measured after several anneals

with temperatures in the range from 815 to 950 C, and compared to the as-grown dose of

3x1013 cm-2. The methodology was different than the one used in Lilak et al., as the

active dose was extrapolated from SIMS measurement. It was assumed the diffusing tails

form a Gaussian distribution within the entire profile, and representative of the active

fraction of the B profile (Fig. 2-13). Such an indirect approach eliminates the possibility

of measuring electrically partially active clusters, rather measuring the substitutional B

concentration through its availability to partake in a diffusion process.

Having determined the active fraction in a number of anneals, they extracted the

activation energy for reactivation as 3.2+0.4eV, in agreement with the previous

measurement of Lilak et al. The B clustered doses and time constants used in the

extraction are shown in figure 2-14.

2.6 Boron Diffusion in Si_-xGex

There are several reports [Kuo93][Kuo95a][Zan01] on the reduction of B diffusion

in Sil-xGex compared to Si. The first report of reduced B diffusivity in Sil-xGex was by

Kuo et al. [Kuo93]. The structure used in the measurement is shown the Fig. 2-15.

The layers were grown on <100> n-type Czochralski Si wafers in a CVD process.

Initially, a layer of undoped Si was deposited. On top of that a 60 nm thick Si0.83Ge0.17

layer, in-situ doped with B concentrations from 1018 to 3x1019 cm-3 in the center 20 nm.









A silicon capping layer of 60 nm was deposited to enhance the stability of the Si0.83Ge0.17

layer. Control samples contained only the B layer. Secondary ion mass spectrometry

(SIMS) was used to extract dopant profiles. Initial profiles, as well as profiles after a 30

minute anneal at 8600C in nitrogen ambient, are shown in Fig. 2-16.

At the top of Fig. 2-16 profiles for boron in the sample without the Ge layer can be

seen, while the plot at the bottom indicates a B layer contained in 17% Ge layer. Square

symbols are used to plot the as-grown profiles, while the plus (+) symbol demonstrates

the annealed profiles. The extracted diffusivity was found to be approximately an order of

magnitude lower in Sil-xGex sample, as shown in the Fig. 2-17.

Moriya et al. [Mor93] also report on the reduction ofB diffusivity in Sil-xGex

layers. In one experiment, 15 nm Sio.7Ge0.3 layers were grown by rapid thermal chemical

vapor deposition (RTCVD), within which a B peak concentration of 8x1019 cm-3 was

introduced in the middle 5 nm of the Sio.7Ge0.3 layer. Ge rich layer was grown on a Si

buffer layer and capped with Si layer, all containing a background B concentration of

1018 cm-3 to avoid electric field effect influence on diffusion. Si control samples only

differed in the fact they contained no Ge. The samples were annealed in a rapid thermal

annealing system (RTA) at temperatures from 8500C to 10000C, in an inert ambient. The

Si control showed no significant deviation from the accepted values for diffusivity of B,

thus verifying no anomalous material effects were present. This was not the case for B

diffusion in Sio.7Ge0.3 layer. The diffusivity of B in Sio.7Ge0.3 layer was significantly

lower than the control, as shown in Fig. 2-18. Activation energy extracted was about 1 eV

higher than the accepted values for B diffusion in Si.









A separate experiment in the same publication [Mor93] tried to measure B

diffusion as a function of Ge content (10%-50%). The multilayer structure was grown

MBE and annealed at 9750C for 30 seconds. As Fig. 2-19. shows, increase in the Ge

content reduces B diffusivity.

Another experiment by Kuo et al. [Kuo95a] attempts to isolate the strain influence

on diffusion. As Ge has 4.2% larger lattice constant than Si, it introduces a strain into the

lattice. Separating the influence of strain from the Ge content was achieved using the

structure shown in Fig. 2-20. First, a graded buffer layer was grown on <100> p-type Si

wafer was grown. The buffer layer contained linearly increasing Ge content, serving to

relax the strain with a minimum of threading dislocations. Second, a 0.25um relaxed

Sil-yGey layer was grown to serve as a substrate for a pseudomorphic diffusion structures.

Third, pseudomorphic 60 nm thick Sil-xGex layer was grown, containing a 20 nm B layer.

Fourth and last, the capping 40 nm Sil-yGey layer was grown to ensure thermal stability.

Having an independent control over x, y Ge contents, one can engineer the strain

for a given x percent Ge. In order to investigate the dependence of diffusion irrespective

of strain, x has to be equal toy. Another application could be to investigate the diffusion

as a function of Ge content, with a constant strain. That can be accomplished for x=y+z,

where z is any Ge content delivering the desired strain. Some of these combinations are

shown in the Fig. 2-21.

There are several things to note. First, diffusivity around 0% strain is a function of

Ge content. Second, strain dependence for 10% and 20% Ge is rather weak, indicating

that the strain is not the dominant component in reduction ofB diffusivity with increasing

Ge content. Rather, it would seem the chemical effect of Ge is more significant. It is









interesting that the Ge presence reduces the B diffusivitiy, particularly since the

self-diffusion of Ge and Si increase with Ge content in Sil-xGex alloys [Str02]. Presuming

that B diffusion remains interstitially mediated in Sil-xGex (confirmed for 10% [WilOl]

and 18% Ge [Kuo95b]), and that the properties of the BI pair remain unchanged with Ge

content, the increase in self-diffusion should increase the effective B diffusivity.

Similar structure (Fig. 2-22) to that of Kuo et al. [Kuo95a] was used in an

experiment ofZangenberg et al. [Zan03] to investigate the diffusion of B and P in MBE

grown Sil-xGex. The Sil-xGex layers were grown on the (100) Czochralski Si wafers, using

the graded buffer layer to prevent the dislocation growth towards the surface. TEM

verified that samples before and after annealing had dislocation density under 106 cm-2,

and that no dislocations formed at the pseudomorphic layer interfaces, confirming that

dislocations had minimal influence on dopant diffusion. The samples used in the study of

B diffusion were annealed at temperatures between 800 and 925 C, with Ge contents of

0%, 1%, 12%, and 24%.

The experimental results were quite unexpected. Despite the previous body of

literature on reduction of B diffusion in Sil-xGex alloys, the study found a small

difference between B diffusion in relaxed Sil-xGex and the control samples. The extracted

parameters for B diffusion in relaxed Sil-xGex are shown in Table 2-1. Even though the

measured control diffusivities fit the certain measured values for B diffusion in Si in that

temperature range, the extracted activation energy of 2.68 eV is quite different from the

values of Fair (3.46 eV) [Fai81] or Haddara et al. (3.75 eV) [HadOO].

The other interesting point in this measurement is the change of the B activation

energy with increasing Ge content. The addition of 1% Ge in the relaxed Sil-xGex alloy









increases the activation energy of diffusion by -0.5eV, which does not change

significantly in higher Ge contents, as demonstrated in Sio.88Ge0.12 and Si0.76Ge0.24. These

values would suggest, provided the properties of the BI pair and self-interstitial diffusion

were unchanged by Ge content, that the addition of Ge adds another potential well of

about 0.5eV to the B diffusion mechanism.

Having reviewed all of these experimental data of B diffusion in Sil-xGex, one can

notice there is no clean study of B diffusion with a number of samples over various

temperatures, like Strohm et al. [Str02] for Ge self-diffusion. The closest to the detailed

measurement of B diffusion in relaxed Sil-xGex is the experiment of Zangenberg et al.

[Zan03] with a temperature range of 125C and a problematic control. Most studies

concentrated on one temperature, or one time, or one Ge content.

2.7 Modeling Boron Diffusion in Si_-xGex

The first B diffusion model to be discussed was published by Lever et al [Lev98].

In their experiment B layer was sandwiched between two Ge layers, both of 10% or 3%

Ge content. One set of Lever's samples was grown via low pressure CVD (LPCVD) on

<100> FZ silicon wafers. They used B wells with concentrations of 3x101, 6x1018, and

1.5x1019 cm-3. Initially, a 100 nm thick silicon buffer layer was grown on a wafer. It was

followed by growth of undoped 40 nm thick Si0.9Geo.1 layer, a 250nm thick B well,

another 40 nm, and finally capped with 100 nm undoped Si. In another set of Lever's

samples, a silicon buffer layer was grown on a Czochralski silicon wafer. It was followed

by a 80 nm thick undoped Si0.97Ge0.03 layer, 200 nm thick B well with concentration of

1019 or 4x1019 cm-3, another 80 nm Si0.97Ge0.03 layer, and capped with a silicon layer. All

samples were capped with 40 nm SiO2 and 160 nm Si3N4, to ensure the inert surface









condition. Anneals were performed at 8500C for times of 4, 24 or 96 hours. Fig. 2-23

shows some of the resulting profiles.

In their simulations, Lever et al. considered B diffusion as perturbed by Ge content.

This simplification disregards possibility that B activation depends on Ge content.

Influence of strain was also neglected, as study by Kuo et al [Kuo95a] showed it to be

small. Change in bandgap was considered to be linear with Ge content. Vacancy

mediated diffusion of B was also neglected, which also limits this model to lower Ge

content. Higher Ge content might have significant vacancy self-diffusion component,

which could influence B diffusion, as well asfi.

The B flux equation (Eq. 2-10) consists of three terms. The first term is Fick's law

term for diffusion of boron, denoted as Cg. Second term is impurity related electric field

diffusivity enhancement. Third term is result of energy bandgap variation due to Ge

content. Diffusivity is described as sum diffusivity coming from interaction with neutral,

D1 and ionized, D2 point defects (Eq 2-11), with factor beta as ratio of those two. The last

equation (Eq 2-12) is the Arrhenius expression for diffusivity as a function of

temperature.

D cB c+ CR (NA -ND) 2cBn aln(n )(
JB = -D B (N, ND 2 n (2-10)
dx n+p ax n+p x


1+8 P
DO D, +D, =D (DI +D ) n = D2 (2-11)
n 1+/7 D,



Do = kexp 3.5) (2-12)
B kT










B- 1+ CGe B act B (2-13)
CB act S S +CGe

The pairing between Ge and B is assumed to form a GeB cluster, immobile, but

electrically active. The rest of B is available for diffusion (Cg act), as described by Eq.

2-13. The fitting parameters of the model are b, k and s. Although the parameters have

been individually fitted for each annealing condition, they do not vary significantly. The

good features of this model include: simplicity, few parameters, and good fit to several

Ge contents with the same value of s. This model has been implemented in FLOOPS, and

used as an initial guess in experiment design.






38


Table 2-1. Extracted B diffusivity prefactor (Do) and activation energy (Ea) from
experiment of Zangenberg et al. [Zan03]
Ge [%] Temp range [C] Do [cm2/s] Ea [eV]
0 800-900 3.4(2.3)x10-4 2.68(7)
1 800-925 3.4(2.0)x10-2 3.13(6)
12 800-925 2.4(2.4)x10-1 3.30(10)
24 800-925 5.7(7.0)x10-2 3.18(13)







k


Figure 2-1. Diffusion mechanisms in silicon: a) interstitialcy diffusion, b) kick-out, c)
vacancy diffusion, and d) concerted exchange [Fah89a]









T (oC)
1400 1200 1000 800

10" -
#4
v #3
10-3
10, #2
a #1






10-19 I I ----I I I II
0.6 0.7 0.8 0.9
103/T (K-')
Figure 2-2. Self-diffusion in Si as measured by Bracht [Bra98], with symbols denoting
different samples











1200 1100


10-14

10-15
10-16


10-18


T (C)
1000


7.0 8.0 9.0
104/T (K-)
Figure 2-3. Self-diffusion in Si as measured by Ural et al. [Ura99b]. Dots represent
measured diffusivities, solid line is the best fit expression for self-diffusion.
Dashed lines represent bounds of self-diffusion as reported by Bracht [Bra98],
showing good agreement between measurements.









1.00


700 800 900 1000 1100 1200
Temperature (OC)
Figure 2-4. Interstital fraction,fsi of the self-diffusion from Ural et al. [Ura99b]. The
lines represent predictions from several metal diffusion experiments.












Temperature ("C)
950


A Si
SSi0.90GeO.
* Si0osoGe0o20
* Sio70Geo30
A Sio.Ge040
SSio.50Geo.50,
I I


1/kT (eV')

Figure 2-5. Germanium diffusivities for various temperatures and Ge contents, after
Zangenberg et al [ZanOl]


S1050
'.-4


Iu


r 16
E

i 10-"
u


e,L
0
1

0



( 10-16


I I I I I I I I
















U '

x N
A' A N


.' U

1 x
A


U



A


0.2
Ge contents


1000



100


-10&^


1 0


0.1



- 0.01
0.6


Figure 2-6. Activation energies and prefactors for various Ge contents, after Zangenberg











5.0- -10000
4.8- Do
4.6- Do fit1000
4.6- .,, E -1000
4.4- -- Efit
> 4.2- -100
S4.0 E
LU 3.8- 10 0
3.6
3.4- 1
3.2
3.0 .... 0.1
0.0 0.2 0.4 0.6 0.8 1.0
Ge contents


Figure 2-7. Activation energies and prefactors for various Ge contents, after Strohm











110
1E-11
1E-12
1E-13
S1E-14
' 1E-15
E 1E-16
S1E-17
.5 1E-18
S1E-19
1E-20
1E-21
1E-22
7,0x1 04


0 1000 900 ]o00


Zangenberg
Strohm


v A&


.* -
"^ ^ A


}xGe= 0.5

xGe= 0.3

}xGe= 0.1


8,0x10-4 9,0x10-4 1,0x10-3 1,1x103


Figure 2-8. Comparing diffusivities of Ge in Sil-xGex as measured by Zangenberg and
Strohm














E

C1
0


0

C 1
0


1 171 I I I I I I I I I I I I I I I I I I I I I I I I I .
0 0.1 0.2 0.3 0.4 0.5 0.6
Depth (im)
Figure 2-9. Comparison of Ge diffusion with respect to ambient and consequential
interstitial or vacancy supersaturation [GriOl]







48



10 Boron Implant
2xO1 4/cm? at 60 keV




108 *


4 "
l; Furnace anneol ot 800TC o
S- None o 7
z -. 3 rnin 0 -
U0 u J m- 65 mrni
0
f n8 min a 'a
S180 min
,016 ---- r -
0 2000 4000 6000

a) DEPTH (ANOSTROMS)
a)
---- ---- I ---- ---- ---- \- -- -----

0 t Boron Implant
2x 1014/c2 at 60 keV


-4 -

,L
UZ -o I*

0 Od*o,
l101 7 R-apd Annec! at 950C
016 I
5 None -
o 5 sec 40
Su 15 sec o '0
A 30 sec *o
%0
01 6 --- -- I--I n- --&
0 2000 4000 6000

b) DEPTH ANGSTROMSS)
Figure 2-10. B profiles after an implant of 2xl014 at 60keV, annealed at a) 800 C and b)
950C [Mic87]



















'I
E
0

, 1(
4r

C
U
0
0


11


Figure 2-11. B profiles, as-grown and after lh at 6700C [Sto95]












600


550


500


450


4001


350


300


0,3



0.2



0.1


0



0.5



0.4



0.3



0.2





0

0


II 1 1 200
0 50 100 150 200 250 300 350 400
Time (mn)


Figure 2-12. Fractional activation and sheet resistance of samples annealed at 7500C for:
a) varying B dose at 80keV, and b) 4x1014 cm-2 at varying energy [Lil02]


I I 1 I I I j 1250
0 50 100 150 200 250 300 350 400
Time (min)
I 1... i ... 1200
.-i-10 lok FA 1 .iV SOhv t RhW
-8 2IkiV FA --B20oVShmt Rho
--40ktV FA -4DOiV She Rho
*r- O FA --Okp-V Shw8~tFho 1000







~c~ 1I










S: o BIC sample i
S----- clustered B
0 1 2x10d8
10 9 Parabola
200 300
10





200 300 400 500
Depth [nm]
Figure 2-13. B profiles used in experiment by Mirabella et al. showing the as-grown,
annealed for 2 min at 815TC, as well as illustrating the methodology for
determining the active fraction [Mir03]
determining the active fraction [Mir03]






52


best fit


.9 t-32M8 s


b 110i30 s\
815 C \ 29260 s
1-
SO 850 *C 2200t300
900C \
0 950 C

1 10 100 1000 10000
Time [seconds]
Figure 2-14. Time evolution of the clustered B dose during annealing at 815 to 950 oC,
with extracted time constants. [Mir03]









Si control sample


i-Si


p-Si


Sil-xGex sample


-200 A


---


S i-Si -_oo A

i-SiGe

i-SiGe
S i-Si


S Si substrate Si substrate
Figure 2-15. Structure used by Kuo et al. [Kuo93] with a B marker layer in the
Sio.83Geo.17 strained layer


i-Si


!











1 020





1019





10186







1019


I 0


10180


(b) i-Si



SUPREME Rl -

SIMS data
Rs-,rown a
anneaald 4.


-Sj i p-SI i-Sl






+


1''4+
I p
*t r


SIl.3GsO .17
Pi





;i 4'


500 1000
depth (A)


i-Si


1500


Figure 2-16. Diffused B profiles after 30 minute anneal at 8600C in a) Si and b)
Sio.83Geo.17 [Kuo93]


SSUPREM Rt

- SiMSl ale:
asnnrewal
annealed


. . .


~~-~~~~~ ~ ~ ~ ~~--~~~~~~~~~~~~~~~ --


h


-L l--I..L -k. l












0 Si ..""
10-1
1 0" S

o- /


10-17

1016 1017 10Is 10W9 1020
B concentration (cm3)
Figure 2-17. Effective B diffusivity as a function of B concentration [Kuo93]









T [C]
1000 900 800

1-13 Si0.7 Ge o.3: B
10-



w 10-5 o-5 ++ Boron in Si (Lit.)
EI -E =3.46 eV)
t" ,-7 Ge/Si Intermixing ++
10- | ,


10-19 L 4.42 eV

7 7.5 8 8.5 9 9,5 1
T-I[K-1] x 1C
Figure 2-18. Diffusivity of B in Sio.7Geo.3 layer [Mor93]


0
-4






57


2.5 ,J-._
40% Ge content
6 2 30%
20%
k 1.5
r 50%





0 1

0 1 2 3 4 5
Depth [xlO cm]
Figure 2-19. B profiles in multiplayer structure before (dashed line) and after (solid line)
annealing at 975C [Mor93]










S400 A



S600 A



S0. 5 pm


i


Figure 2-20. Illustration of a structure used by Kuo et al. [Kuo95a] showing relaxed
Sii-yGey and pseudomorphic Sil-xGex


pseudomorphic
Sil_,Gex: x, 0.10, 0.20
or Si: x= 0



y= 0 (no SiGe)
or ys 0.10, 0.15, 0.20






59



T= 800 C (
x x= 0 (Si control)
10-16 -o- x-0.10
x x= 0.20

fully strand



S(0.115,0.195)

5 1 (0.21,0.185)
(0.21.0) (0.225.0.105) (0.21.0. 85
-(022,0.152)

-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4
strain (%)
Figure 2-21. B diffusivity at 800C as a function of strain, with the numbers in the
parenthesis are (x,y) Ge content. Positive strain represents biaxial tension and
negative strain biaxial compression. [Kuo95a]







60

SSi IGe SiGe S Si GOe
1-y y -x Xl-y y
i I Y Ge

S Subfirale
B (enlarged)


1oooA 650A 250A 650A >6000A Depth
Figure 2-22. Schematic description of the structure used by Zangenberg et al. [Zan03] to
measure B diffusion in Sil-xGex









1019 ,0(b

1019



m 10 (





0.0 0.2 0.4 0.6 0.0 0.2 0.4 0.6 0.8
depth (microns) depth (microns)
S100




1018

m 10,8
S1017



0.0 0.2 0.4 0.6 0.8
depth (microns)
Figure 2-23. B profiles before (thin line) and after (thick) 850C annealing of a given B
well concentration. Figure a) B well with a concentration of 3x1018 cm-3,
sandwiched in Sio.9gGeo.i layers, annealed for 96 hours. Figure b) B well with a
concentration of 1.5xl019 cm-3, sandwiched in Sio0.Geo., layers, annealed for
96 hours. Figure c) B well with a concentration of 4x1019 cm-3, sandwiched in
Si0.97Geo.03 layers, annealed for 24 hours. [Lev98]














CHAPTER 3
BORON CLUSTERING IN SILICON

Overview

This chapter discusses the behavior of boron-interstitial clusters (BIC) and its

influence on activation of B in silicon. Experimental evidence shows that the previous

model and its behavior with respect to annealing ambient do not capture the observed

cluster dissolution phenomena. The most recent set of ab-initio calculated formation

energies is utilized in attempt to capture this phenomenon, which is achieved with certain

changes within the ab-initio energetic.

The cluster formation is examined in the experiment of Jones et al. [Jon96a], which

indicates that the formation is a rapid process and is completely prevented by

preamorphization. Assuming the inability of silicon interstitials entering the amorphous

layer significantly reduces the percentage of boron tied in the clusters. Additional, small

changes of the cluster properties further reduce the formation probability of the clusters,

thus matching the experimentally observed behavior.

3.1 Boron-interstitial Cluster Dissolution

3.1.1 Introduction

Ion implantation is the preferred method of introducing dopants due to the

precision of implanted dose, control and repeatability. Future technology nodes will

require highly doped shallow junctions processed with a low thermal budget. Such

conditions emphasize the role of point defects generated by the implantation process.

Recombination of point defects and dopant activation take place during annealing post









implantation. For nonamorphizing implants, the annealing step can form extended defects

and dislocation loops, which are responsible for enhanced diffusion [Eag94][Jon96b].

Besides effects on dopant diffusion, high interstitial concentrations are also known to

deactivate boron, through the formation of boron-interstitial clusters (BIC)[Sto95].

Ab initio studies provided information on the stability of some clusters, used in

several implementations of BIC models in Monte Carlo [Cat98] and continuum process

simulators [Pel97][Lil97]. It is unclear if boron will cluster in an interstitial

supersaturation lower than that found in ion implantation, and if measurement methods

would be able to distinguish such a presumably low concentration of clustered boron

from the noise of the active boron profile. Specifically, characterization of low cluster

densities is difficult, as low cluster densities are not detectable as spikes in the profile

acquired by secondary ion mass spectrometry (SIMS) profiles and could be below

detection limit of spreading resistance or Hall effect measurements. The behavior of BICs

during dissolution, with respect to interstitial supersaturation, has not been verified

experimentally. The model proposed by Pelaz [Pel99] suggests that break up of the

clusters requires an interstitial to drive the process along an interstitial lean path. This

theory says that the boron is driven to cluster in a large interstitial supersaturation, but a

lower supersaturation might provide excess interstitials that will speed cluster dissolution

and boron reactivation. Notwithstanding, experimental evidence suggests that BICs can

serve as a source of interstitials for transient enhanced diffusion (TED), by dissolution of

boron clusters in later stages of annealing [Sol00].

To test the theory of BIC cluster dissolution through interstitial lean path, implants

of boron were performed to achieve peak boron concentration on the order of 1019 cm3.









Annealing is performed in two steps. Boron interstitial clusters form during first anneal

step and deactivate a significant part of the dose. The second anneal is used to determine

the influence of annealing ambient on cluster formation or dissolution. Clustering effects

are investigated via SIMS and Hall effect measurements.

3.1.2 Experimental

Czochralski <100> n-type silicon wafers were implanted with boron. Three implant

conditions with nominal boron peak concentrations of 1019 cm-3 were investigated.

Implant doses and energies were lxl014 cm-2 at 5 keV, 2x1014 cm-2 at 10 keV and 4x1014

cm-2 at 20keV, respectively. The first anneal step was performed at 7500C for 30 minutes

in inert N2 ambient. To investigate the influence of ambient on B cluster dissolution,

subsequent anneals were done in inert (N2 flow) or oxidizing (dry 02) ambient at 850C

for 10, 20, 30 and 60 minutes. Oxidation of silicon is known to inject interstitials [Hu74],

allowing investigation ofBICs in a low supersaturated interstitial environment.

Samples were analyzed by SIMS, Hall effect and ellipsometry measurements.

Boron depth profiles were measured by CAMECA IMS-3f instrument using 02+ primary

ion beam and magnetic sector analyzer. Raster size and analyzed area were 200 |tm and

60 |tm in diameter, respectively. Samples were biased at 4.5 kV with an effective impact

energy of 5.5 keV. The Woollam EC110 ellipsometer was used for oxide thickness

measurements, showing no difference between initial condition and inert annealed

samples. Oxide thickness measurement of samples annealed in oxidizing ambient was

used in aligning SIMS profiles, due to the consumption of silicon. Measured oxide

thickness agreed well with standard predictions at these temperatures. The active dose

was measured using a MMR Technologies Inc. Hall-van der Pauw system. Measurements









were performed using a magnetic field of 300 mT with the current ranging from 1 ptA to

1 mA at room temperature. Square shaped samples with 14 mm sides had 1 mm diameter

e-beam deposited Al contacts placed symmetrically near the corners. Factors contributing

to the inaccuracy of the measurement include nonuniform dopant distribution and heavy

to light hole ratio. Assuming measured profiles were steep enough to neglect the

influence of deeper layers with lower dopant concentration and higher mobility, we chose

the value of 0.7 for the Hall mobility factor [Sas88].

3.1.3 Results and Discussion

First annealing step provides the initial condition with BICs of similar peak

concentrations for studied implant conditions (Fig. 3-1). During this anneal excess

interstitials generated by the implantation process interact with boron forming BICs.

These clustered boron samples are referred to as the initial condition from this point on.

Fig. 3-2 shows boron profiles at critical stages of thermal processing for each of the

studied implant conditions. Comparison of oxidized and inert annealed profiles show

oxidation enhanced diffusion for the boron as expected for an interstitial diffusing species

[Pac90]. Direct comparison of the clustered fraction is difficult from the SIMS results

because of the extra diffusion. Although the shallowest implant exhibits the highest

diffusion enhancement, we do not believe the distance to the surface to be significant.

Different diffusion enhancements may be a consequence of a difference in the initial

clustered dose, as oxidation is known to create enhancements uniformly deep in the bulk

[Gri85]. The surface influence in these samples is mostly through segregation and

outdiffusion of boron, resulting in dose loss (Table I). Dose loss is most apparent in









lowest implant dose/energy sample, since it has the shallowest boron profile in presented

sample set (Fig. 3-2c).

Hall effect measurements during various stages of annealing are shown in Fig. 3-3.

The initial anneal deactivates a large fraction of the dose, creating the initial clustered

concentration that establishes our initial condition. Subsequent annealing activates the

boron by dissolving the clusters, with inert anneals consistently having a higher active

dose than oxidizing anneals. Considering dose loss and reactivation effects are mixed in

the Hall effect measurement, defining activation as ratio of active to retained dose

(integrated SIMS profile) is used to distinguish them. Activation fractions for known

boron profiles are presented in Fig. 3-4. Increase of activation by 15-40% for inert

annealed over oxidized annealed samples confirms the trend demonstrated by the Hall

effect measurement. That leads to the conclusion that boron clusters dissolve slower in

the oxidizing ambient than in the inert ambient.

Provided that one of the dominant BICs has a self-interstitial release reaction

during dissolution, its dissolution rate would be diminished or annulled with the increase

in the interstitial concentration. Since the oxidation induced interstitial injection reduces

reactivation (Fig. 3-3) when compared to the inert ambient anneal, which leads to the

conclusion that one of the significant clusters has an interstitial release reaction during

dissolution.

Examining the time dependence of the reactivation shows disagreement with the

idea of one dominant BIC. In the course of the second stage anneal, one can observe the

initial increase in activation in almost all samples irrespective of the annealing ambient.

After the initial increase, active dose in oxidizing ambient stays constant, while inert









ambient continues to reactivate B at longer anneal times. Supposing the reactivation

processes occurring under oxidation are also occurring in inert ambient as well, one could

consider inert reactivation as a superposition of oxidizing and an additional long-term

reactivation. This would indicate there are at least two clusters holding a significant B

dose. The one responsible for the short time reactivation would be rather insensitive to

the ambient. On the other hand, the long-term reactivation would be coming from a

cluster whose dissolution is affected by interstitial injection.

3.1.4 Modeling

In attempt to model the oxidation behavior we utilized the energetic published by

Liu et al. [LiuOO] The model is implemented in model description language

ALAGATOR, assuming diffusion limited reaction rates, and simulated using the process

simulator FLOOPS [FLO02]. Implanted ion and damage profiles are generated by

UT-Marlowe [UTMOO]. Boron clustering and cluster dissolution paths are shown in Fig.

3-5a. The clustering process is different from previously described model [Pel99]. During

high interstitial supersaturation, BICs grow towards B and Siint rich, which later release

interstitials to form B rich, Siint poor clusters. However, there are two clusters, B3I and

B213, containing the majority of the B dose, as opposed to exclusively B4I [Pel99].

Regarding the dissolution paths, they would indicate opposite reaction to oxidizing

ambient than seen in experiment. Oxidation injected interstitials would be captured by

B3I enhancing the transition towards B312. This is the rate limiting step, since the

migration energy of a BI is lower than that of Siint, and B312 can subsequently dissolve

with the release of BI solely.

Simulation of the two-step anneal used in the experiment resulted in Fig. 3-5b and

shows the simulated activation during second anneal step. One can observe a good fit to









clustered condition after the first anneal step. The reactivation during second anneal step

is rather insensitive to oxidizing ambient. This is the consequence of similar B213 and B3I

doses after the first anneal step, as B213 can provide more than sufficient number of

interstitials available for B3I reactivation.

In order for a given BIC to dissolve slower under oxidation, it would have to

release an interstitial along its path to B substitutional. Then the interstitial injection

would be able to slow down that reaction, as observed in experiment. Therefore, we

propose that the ambient sensitive BIC, BmIn, would have at least as many Si interstitials

as B atoms (n>m). We modified the energetic in an attempt to model the oxidation

behavior using B414 and B213 as the two dominant clusters. The resulting cluster formation

and dissolution paths using the modified energetic are shown in Fig. 3-6a. In this

proposed model, B213 provides the reactivation at short times (less than 10 minutes) and

is insensitive to oxidation. On the other hand, B414 dissolution is affected by interstitial

injection and provides reactivation at longer times. Simulation of the two-step anneal

resulted in active dose plot shown in Fig. 3-6b. One can observe a good match to the

initial clustered condition. The model also shows appropriate behavior under oxidation,

which is the first such model according to the literature. Two-stage reactivation is also

visible, as described previously.

3.2 Boron Cluster Formation and Preamorphization

3.2.1 Experiment and Findings

One of the frequently applied methods for increasing activation of B implants is

preamorphization. There are multiple benefits and some disadvantages to this method.

Important benefits include: reduction in boron clustering, high activation at very low

temperatures during the solid phase regrowth, and reduction of the junction depth by









reduction of the channeling tail. The main disadvantage is the formation of extended

defects in the EOR region, which is detrimental to the electrical characteristics of then

junction.

In an investigation of B clustering Jones et al. [Jon96a] implanted a Si dose of

5x105 cm-2 at energy of 146 keV, at liquid nitrogen temperatures. The implanted

material contained three B marker layers, grown via MBE, as seen in Fig. 3-7.

The Si implant at 146 keV produced an amorphous layer 0.324 |tm thick, having

the damage peak concentration coincide with the middle marker layer. Upon annealing at

800C, the excess interstitials in the damaged region interact with B in the marker layers

beneath the amorphous/crystalline (a c) interface. This can be observed at longer

annealing times as the immobile B spikes, at the position of the B peak, as seen in Fig.

3-8. The center B marker layer is almost completely immobile during the 3 minutes of

annealing. At that time the shallowest peak is diffused out showing no signs of clustering,

while the deepest marker layer exhibits an apparent clustering peak. The difference

between the outer two peaks is interesting since they are almost the same distance from

the a/c interface. Assuming the damage is concentrated in the region just beneath the a/c

interface, the shallowest and the deepest marker layers should observe the same

interstitial supersaturations. This assumption is partially validated from the similarity in

the observed profile broadening shown in Fig. 3-8. Although the deepest marker layer has

higher peak concentrations during the anneal when compared to the shallowest marker

layer, it still experiences similar diffusion enhancement, i.e. interstitial supersaturation.

The apparent clustering peak revealed at longer annealing times indicates the difference

in the material structure, between deepest marker layer in the crystalline silicon and the









shallowest marker layers in amorphous silicon, might be responsible for the existence of

the BICs.

The question then becomes if it is a material structure, or just an observable

consequence of the different interstitial supersaturations each of the marker layers

experiences at short times. Even though the regrowth of the 0.15 |tm between the

shallowest and middle marker layer takes less than 0.25 seconds (according to [Ols88]),

the supersaturation could drop by a factor of 2 or more during that time [Cow99]. The

delay of the a/c interface reaching the shallowest marker layer facilitates avoiding the

initial high interstitial supersaturations forming BICs in the middle marker layer,

allowing only the lower interstitial concentration reaching the shallowest marker layer,

and could also be contributing to the reduction in the BIC formation.

3.2.2 Modeling

A UT-Marlowe simulation used to verify that the assumed distribution of excess

interstitials following an amorphizing implant is shown in Fig. 3-9. The excess interstitial

profiles obtained from the UT-Marlowe implant simulation show that the damage is not

concentrated solely around the middle marker layer and limited to the region just beneath

the a/c interface. Rather, there are excess interstitials that could provide the interstitials

and boron in a sufficient proximity for cluster formation in the deepest marker layer.

Assuming that the silicon and boron interstitials are not able to enter the amorphous

layers inserts a delay in the cluster formation of the shallowest B marker layer.

Presuming that the boron cluster formation is fast, the delay could prevent the BIC

formation since most of the available Si interstitials would already be tied in clusters,

either BICs or self-interstitial clusters. The delay would also mean that SMICs could be









dissolved by the time the a/c interface reaches the shallowest marker layer, thus the

interstitial supersaturation would be significantly reduced. The reduction of the interstitial

supersaturation affects the BIC formation in two ways, through reduction of

self-interstitials and boron interstitials, again limiting the BIC formation.

To test this hypothesis, an additional equation was used to calculate the amorphous

layer depth in the simulation. The self-interstitial and boron interstitial diffusivities were

annulled in the region above the a/c interface, thus disabling their diffusion and their

reactions to any other species. This forms a reflecting interface boundary at the position

of the a/c interface, as it moves towards the surface. The a/c interface depth equation (Eq.

3-1) uses the regrowth velocity measured by Olson & Roth [Ols88] in units of |tm/s. The

diffusivity of self-interstitial and boron interstitials were limited to the crystalline

material, as shown in the Eq. 3-2 and 3-3, with the reaction term omitted for brevity.

(aSi depth) 2.7eV
a(as h) = -3.1.1012 -exp (3-1)
at kT

d0nt 02jnt
S= D, (x > aSi depth) + reaction terms (3-2)
at ax

BI a 2BI
= D,, (x > aSi depth). + reaction terms (3-3)
at ax

The following simulation results illustrate the influence of spatial separation

between the boron profile and the excess interstitials. The case chosen for this

demonstration is the annealing a boron dose of 4x1014 cm-2 implanted at 20keV. Figure

3-10 shows the simulated doses of pertinent boron and interstitial clusters during a 30

minute anneal at 7500C, with boron implant performed without a preamorphization.

Figure 3-11 shows the same boron implant, with a silicon preamorphization of 1x1015









cm-2 implanted at 100keV, annealed under the same conditions. One of the benefits of the

preamorphization is the smaller amount of excess interstitials remaining in the EOR

region, as is apparent in this case, with excess interstitial dose of 8xl013 cm-2. In order

to be able to compare the results, the excess interstitial dose in the preamorphized case

was ad hoc increased by a factor of six, matching the excess interstitials available in the

case without preamorphization.

There are several important values indicating the dominant processes in these

simulations, namely Smic, C311 and Bsub. Initial Smic dose is the initial dose of trapped

interstitials that evolve into {311 }'s during the course of the anneal. Without boron

present, most of the excess interstitials in Smic's are captured by {311 }'s.

In the case of boron implant into crystalline silicon (Fig. 3-10), only -15% of

excess interstitials from Smic's are captured by {311}'s. The rest is trapped in various

BIC configurations, reducing the substitutional boron dose to -18%. Displacing the

excess interstitials by a preamorphization Si implant changes all of these values (Fig.

3-11). As the boron is relatively far away from the Smic profile, most of the excess

interstitial dose in the EOR region is captured in a form of a {311} defect. Interstitial

supersaturation supported by the {311} defects eventually forms some BICs, still leaving

>85% boron in the substitutional position after the 30 minutes 750C anneal. This

demonstrates the increase in the substitutional boron dose by a factor of four using

preamorphization with the equivalent excess interstitial dose, solely by spatial separation

of the boron and excess interstitial regions.

Considering the actual preamorphization damage of a Si 1x1015cm-2 100keV

implant performed at liquid nitrogen temperatures, simulated excess interstitial dose is









quite lower than that of the B implant. Simulated doses (Fig. 3-12) show a certain

insensitivity of the clustering process to the EOR trapped interstitial dose, as the boron

substitutional dose drops to -90%. This may not be unreasonable when compared to the

previous case (Fig. 3-11) having artificially increased excess interstitial dose, as the

{311 supported interstitial supersaturation does not depend on the defect density, size or

trapped interstitial dose. Therefore, the boron profile should experience roughly the same

interstitial supersaturation irrespective of the {311 dose, prior to its complete

dissolution.

This confirms the displacement of the excess interstitial makes a difference

between 18% activation (Fig. 3-10) in non-preamorphized case versus over 85%

activation (Fig. 3-11, 3-12) in the preamorphized case. Coupled with the lower amount of

excess interstitials in the realistic preamorphized case, such as a Si xl1015 cm-2 implant at

100keV with liquid nitrogen cooling, the activation can rise over 90% as seen in the Fig.

3-12.

Even though the previously described blocking of the interstitial and boron

interstitial pair diffusion into the amorphous layer significantly reduces the clustered

portion of the profile, it does not yield good fits to the profiles in Fig. 3-8. The simulated

profiles for the annealing of 5 seconds, 30 seconds and 3 minutes at 800C are shown in

Fig. 3-13, 3-14, and 3-15, respectively. The SIMS profiles in these figures are the SIMS

profiles from Fig. 3-8. In these figures, the Bsub represents the substitutional portion of

boron profile, the only species assumed to be electrically active. The Total B profile

represents all boron, irrespective of its configuration or position in the crystal lattice. The

clustered boron concentration is not plotted for clarity, since it is inferred as Total B-B,,b.









One can notice that the shallowest and the deepest B marker peak magnitude are

very similar throughout the simulation, in contrast to the SIMS profiles. This is

interesting, particularly considering the different portions of the profile in a clustered

state, visible as the different Bub peak concentrations in Fig. 3-15. The time dependence

is not captured well, as the shallowest and the deepest marker layers do not diffuse as

seen in the SIMS. Considering the activation of boron in the shallowest peak drops only

14% from the full activation, this makes the significant reduction in diffusion

enhancement interesting.

Presuming the formation of BICs is still too strong in the regrown amorphous layer,

the model parameters were adjusted. The binding energy of the small clusters was

reduced without affecting equilibrium diffusion, in an attempt to reduce their

concentrations. This effectively prevents small BICs from serving as trapping and

nucleation sites for larger BICs. The results from the simulations using the modified BIC

model parameters are shown in Fig. 3-16, 3-17, and 3-18.

In these figures, one can observe the fairly good match for the middle B marker

layer. At longer annealing times, it would seem that the clustered peak does not release

enough mobile boron to retain the diffused tails around the clustered peak. The

shallowest and the deepest marker layers have a similar peak concentration at shorter

times despite the different clustered portions. Peak concentrations differ at longer times,

revealing a clustered peak in the deepest marker layer that is not present in the shallowest

marker layer. The shallowest marker layer does form clusters, but since the clustered

dose dropped to around 1.7%, they are not visible in the profile.









The binding energies of the small BICs that were reduced to facilitate the reduction

of clustering in regrown material did not significantly affect the other parameters of the

model. That is the due to the fact that the reactivation properties are the consequence of

the dissolving cluster, in this case B414. The changes did seem to accelerate the B213

dissolution, which was compensated by increasing its binding energies without the

adverse effects on either clustering and reactivation simulations, or the simulations of

BIC clustering in amorphous material. The overall increase in activation is partially due

to the change in the formulation of interstitial supersaturation supported by {311}'s

(C311 Bindl) to match enhanced diffusion with the interstitial equilibrium concentration

(Int Cstar) values used in the Liu et al. [LiuOO][Win99], and partially due to the change in

the model parameters. Two figures (Fig. 3-19, 3-20) compare the reactivation during

second annealing stage at 8500C, in inert ambient, between the model presented earlier in

the chapter and the changes introduced to capture the behavior in regrown silicon.

3.3 Conclusion

Boron interstitial clusters formed by a 750C anneal after implantation deactivated

the majority of implanted boron. These samples allow a study of the reactivation rate in

two different ambient conditions. The oxidizing ambient shows consistently lower

activation at a slower rate when compared to the inert ambient, which strongly

demonstrates that the presence of excess interstitials slows cluster dissolution. This

conclusion is contrary to the accepted BIC model, though it is supported by experimental

evidence [Sol00] suggesting that BIC dissolution involves the release of silicon

interstitials.

To model the oxidation behavior, we utilized the energetic published by Liu et al.

[LiuOO], with some modifications. The model implemented in the description language






76


ALAGATOR, and simulated using the process simulator FLOOPS [FLO002] captured the

reactivation behavior, correctly distinguishing the reaction to the ambient atmosphere. An

implementation of the reflective boundary at the a/c interface reduces the simulated

clustered portion of the B implanted into amorphous silicon, as seen in the work of Jones

et al. [Jon96a]. However, the reductions in binding energies of small BICs were

necessary to model the clustering behavior in the crystalline, versus amorphous silicon.






77


Table 3-1. Retained boron dose [cm-2] in different processing steps. First anneal step is
750C, 30 min in inert ambient. Second anneal step is 850C, 60 min in
respective ambient. Dose loss in lowest implanted dose sample is comparable
to active dose measured by Hall effect measurement.
Implanted dose First anneal step Second anneal step
[cm-2] [cm-2] Inert [cm-2] Oxidizing [cm2]

4x1014 3.69x1014 3.58x1014 3.33x1014

2x1014 1.75x1014 1.58x1014 1.25x1014

1014 7.48x1013 7.01x1013 4.48x1013












4x1019



"3x1019-
E


S2x1019
-I-

8
8 1x1019
0
O
cc

n


4x1014 cm2, 20 keV
c-2x1014 cm2, 10ke
1x1014 cm-2, 5 keV


U


A U?
rU
*cc:, )-t- -t
A-&.A Ar-a-


depth [pm]


Figure 3-1. Secondary ion mass spectrometry (SIMS) measured boron profiles of
investigated implant conditions (B implant lxl014 cm-2 at 5 keV, 2x1014 cm-2
at 10 keV and 4x1014 cm-2 at 20keV) after first anneal step, 750C for 30
minutes in inert ambient.


I -- L


I








79




a) 4x1014 cm2, 20 keV
initial cond.
----dry 02
....... inert

E 10



10




1017
0.0 0.1 0.2 03
depth [tim]


b) 2x1014 cm2, 10keV
initial cond.
--dry 0
S ....... inert






.'..,,


10
10 17
0.0 0.1 0.2 0.3
depth ["m]

c) 1x104 cm-2, 5 keV
initial cond.
--- dry 02
... inert

E 1






1 i "



0.0 0.1 0.2 0.3
depth [pm]

Figure 3-2. Secondary ion mass spectrometry (SIMS) measured boron profiles. Initial
condition is after 750C, 30 minutes inert anneal. Subsequent anneal is 850C,
60 minutes in respective ambient. Figure parts a), b) and c) shows the result
for a 4x1014 cm-2/ 20keV, a 2x1014 cm-2/ 10keV, and a lxl014 cm-2/ 5keV B
implant, respectively. These conditions are chosen because the peak
concentrations and initial clustered concentrations are similar.










-0- 4x1014 cm-2 20keV, dry 02
3.0x1014- 4x1014 m-2 20keV, inert
2x1014 cm2, 10keV, dry 02
2.5x1014 2x1014 cm-2, 10keV, inert
-- 1x1014 cm-2, 5keV, dry 02
14 2
0x014- 1x1014 cm, 5keV, inert
2.OxlO -


a 1.5x1014 r A
/[0 -, m



14
1.0x1014 A -


5.0x1013- 0-

0.0-
0 10 20 30 40 50 60
Anneal time [min]

Figure 3-3. Active dose measured by Hall-van der Pauw method, during the annealing at
850C. The time zero measurement corresponds to the condition after a
750C, 30 minutes inert anneal.












0.9-

0.8-

0.7-

0.6-
/-- z


S04-

0.3- Z 4x1014 cm2, 20keV
.2 -A- 2x1014 -2, 10keV
0.2-
x101xO4 c-2 5keV
0.1-
initial dry 02 inert

Annealing condition

Figure 3-4. Active fraction, the ratio of active to retained dose (integrated SIMS profile),
compared for the different annealing ambient at the end of the 8500C, 60
minutes anneal. The initial condition is after a 750C, 30 minutes inert anneal.








-I

-BI + +BI

+I1


B4


B3


B41


B2 B3I B412
+ +
B B2I- B312 B413
+, +, + 4
BI B B2IL--. B3I3 B414
+ +B
BI24- BI13


2.5x1 014

2.x10 14

E 14
0 1.5x104
Ca)
O 14
"O 1.0x101

< 5.0x1013


V V I II I I I I
0 10 20 30 40 50 60
b) Anneal time [min]
Figure 3-5. Major cluster formation and dissolution paths of Liu et al. [LiuOO] with B31
and B213 containing most of the B clustered dose in figure a) and simulation of
B clustering and dissolution during thermal processing used in experiment in
figure b)


* 4x10 'cm, 20keV, dry 0
u 4x10'4 cm2, 20keV, inert ]
A 2x10 cmn2, 10keV, dly 0
A 2x104 cm"2, 10keV, inert
* 1x10 cm2, 5keV, dry 0
0 1x10 cm2, 5keV, inert 0



A O








-I

-BI +BI B4

+I
B3 B41

B2 B3I -- B412

B B214 B312 B413

BI B2I- B313 B -1 B414
+ +
BI 4- BJ,
a) B 2

2.5x1014
4x10"cm2, 20keV, dry O
D 4x10" cm2, 20keV, inert C
2. 14 A 2x1014 cm,10keV, ky 02-
X 0 A 2x10" cm"2, 10keV, inert --
1x10"4 cm2,5keV, dryO -
E0 1x10"4 crr", 5keV, inert ,_
E 1.5x10 -




5.0x1013

0.0
0 10 20 30 40 50 60
b) Anneal time [min]
Figure 3-6. Cluster formation and dissolution paths of modified Liu et al. [LiuOO]
energetic, with B414 and B213 containing most of B clustered dose, are shown
in figure a), with simulation ofB clustering and dissolution (modified
energetic) during thermal processing used in experiment is shown in figure b)












.' .- I -- --0(OC 3 min.


1019-




107 1 .





0 1000 2000 3000 4000 5000 6000 7000 8000
Depth (A)
Figure 3-7. Boron profiles of material used in the study of Jones et al. [Jon96a],
unimplanted and annealed in inert ambient at 800C











1020 unannealed
---- s(00rc 5 sec.
800*C 30 sec.
---c-- 800C 3 min.


I 1019










0 1000 2000 3000 4000 5000 6000 7000 8000
Depth (A)
Figure 3-8. Boron profiles after a 5x1015 Si implant at 146keV, annealed in inert ambient
at 800C [Jon96a]












1E20
--B_ init
Excess Int


S1E19
'OE

.0


G 1E18-
0



1E17
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
depth [jm]

Figure 3-9. Boron marker layers and excess interstitial damage following a 146keV Si
implant shows proximity of deepest marker layer to the damage