LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) CHARACTERIZATION
OF Si/GexSilx/Si HETEROJUNCTION BIPOLAR TRANSISTORS
By
GUOXIN LI
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 1998
ACKNOWLEDGEMENTS
I would like to thank my supervising professors Dr. Arnost Neugroschel and Dr. ChihTang Sah for their guidance during my doctoral research. I would also like to thank Dr. Peter Zory, Dr. Vladimir Rakov and Dr. Mark Orazem for serving on my supervisory committee. I also thank Jin Cai, WanPeng Cao, Derek Martin, and Yi Wang for helpful discussions.
ii
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ............................................................................................. ii
A B S T R A C T ..................................................................................................................... v
CHAPTERS
1. INTRODUCTION TO GeSi HETEROJUNCTION BIPOLAR TRANSISTOR....... 1
1.1 Introduction.................................................................................................... 1
1.2 GeSi HBT Fabrication...................................... ............................ 3
1.3 Pseudomorphic Growth, Critical Layer and Geinduced Effects..................... 7
1.4 GeSi HBT Characteristics............................................................................ 8
1.5 Monte Carlo Simulation of Mobilities................................ .............. 16
1.6 Energy Band Diagram..................................................... 19
1.7 NcNv and ni Model for Linear Ge Distribution........................... ........ 24
1.8 S u m m ary ............................................................................................................. 2 6
2. LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) METHODOLOGY.. 28
2 .1 Introduction ......................................................................................................... 28
2.2 Base Thickness Modulation (BTM)..................................... ............. 30
2.3 LFGV M ethodology................................................................................... 33
2.4 Derivation of LFGV Methodology and Parameter Extraction......................... 41
2.5 Experiments and Results.......................................................................... 45
2.6 Sensitivity of tB and rl to T12 and XR/XB......................................................... 55
2.7 GeSi HBT Performance Enhancement....................................................... 61
2.8 Sum m ary............................................................................................................. 62
3. LFGV ANALYSIS OF GeSi HBT WITH NO REFERENCE Si BJT.................... 70
3.1 Introduction......................................................................................................... 70
3 .2 T h eory ................................................................................................................. 7 1
3.3 Experiments and Results.................................................... 74
3.4 LFGV Analysis with no Si Reference BJT.................................... ....... 84
3.5 Sum m ary ............................................................................................................. 86
111
4. SUMMARY AND CONCLUSIONS................................................................. 87
4 .1 S um m ary ............................................................................................................. 87
4.2 GeSi HBT Modeling and LFGV Methodology........................... ....... 88
APPENDIX A GENERAL EQUATIONS FOR BJT................................... ...... 89
APPENDIX B TRANSISTOR RELATION FOR A GAUSSIAN DOPANT IMPURITY
CONCENTRATION PROFILE IN THE BASE OF BJT.................... 92
APPENDIX C LIST OF SYMBOLS............................................... 98
R EFER E N C E S................................................................................................................. 100
BIOGRAPHICAL SKETCH.........................................................107
iv
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
LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) CHARACTERISTICS
OF Si/GexSil_x/Si HETEROJUNCTION BIPOLAR TRANSISTORS By
Guoxin Li
December 1998
Chairman: Dr. Arnost Neugroschel Cochairman: Dr. ChihTang Sah
Major Department: Electrical and Computer Engineering
Low Frequency Conductance Voltage (LFGV) methodology for analysis of heterojunction bipolar transistors (HBTs) is presented. The LFGV methodology is based on a combined measurement of the dc collector and base currents and low frequency (lkHz) smallsignal conductance resulting from the base thickness modulation. It gives accurate and independent determination of the builtin electric field resulting from the spatially varying energy gap and dopant impurity concentration profile in the base layer. The LFGV methodology also gives accurate quantitative values for the minoritycarrier transport parameters, including the minority carrier diffusion length and recombination lifetime, and a complete charge control analysis of the recombination losses in the emitter and base region of the HBT.
v
The LFGV methodology is used here to study the performance improvements of Si/GexSil_/Si HBT over that of the Si BJT, including current gain, switch speed, and Early voltage which are primarily due to the aiding electric field.
Analytical models for the effective densityofstates NcNv and intrinsic carrier concentration ni as a function of Ge mole concentration for a linear Ge distribution in the base region are developed. Analytical solutions for the LFGV methodology are then derived for an arbitrary Ge and dopant impurity concentration profile. The analysis showed that the main effect of Ge in the base region with a linearly retrograded Ge concentration profile comes from the aiding electric field which reduces the base transit time and increases the collector current and the cutoff frequency fT. Accurate measurements of the base minority carrier diffusion length and recombination lifetime indicate large performance improvement of GeSi HBT over Si BJT. Design considerations for optimal performance of GeSi HBT are also investigated, such as the optimal Ge concentraction profile in the base region. The LFGV methodology is applicable to other bipolar junction transistors and HBTs, including those fabricated in EIIV materials.
vi
CHAPTER 1
INTRODUCTION TO GeSi HETEROJUNCTION BIPOLAR TRANSISTOR
1.1 Introduction
The concept of heterojunction bipolar junction transistor (HBT) was first proposed by Shockley in 1948 [12]. The purpose was to build bipolar junction transistors with larger energy gap in the emitter than in the base, which improves the emitter injection efficiency over that of the homogeneous bipolar junction transistors. The first detailed analysis of the HBT characteristics and performance was given by Kroemer [3].
Although the advantages of the heterostructures were known, practical applications were not realized until the 1970s when the epitaxial technology was developed. The first successful HBT was fabricated in 1972 by IBM [4] using AlGaAs/GaAs heterostructure to give the larger emitter energy gap. The anticipated superior performance of the HBT was demonstrated by these and other IIIV compound HBTs. Until recently, the AlGaAs/GaAs HBTs have been the most successful. This is due to the excellent lattice matching between the two material which gives a nearly defectfree heterointerface. But GaAs or other HBTs based on IIIV semiconductor materials are difficult to integrate with the existing Si integrated circuits due to the large lattice mismatch. Also IIIV HBTs are difficult to fabricate and thus not cost efficient. GaAs HBT has been more successful in discrete devices, such as high frequency oscillators and amplifiers.
1
2
The breakthrough for an Sicompatible HBT was accomplished again by IBM in 1988 [5]. They used Germanium Silicon (GeSi) as the base material which has a smaller energy gap than that of the Si emitter and collector. A thin strained single crystal GexSilx film on Si substrate was grown pseudomorphically using ultrahigh vacuum/chemical vapor deposition/low temperature epitaxy (UHV/CVD/LTE) process [68]. The energy gap of the base can be tailored by varying the Ge mole content in the base. Furthermore, the Ge distribution in the base region can be graded which will produce an aiding electric field for the minority carrier transport through the base. Ge incorporation in the base will also modify the effective mass and mobility of both majority and minority carriers. This results in a number of advantages of Si/GexSilx/Si HBT over Si BJT, such as higher collector current, higher current gain 3, superior high frequency performance (higher unit gain frequency fT and maximum oscillation frequency fmax), and excellent low noise behavior. The best high frequency performance was obtained at low voltages (VcE=13V), which is very useful for portable communication applications.
Since the introduction of the pseudomorphic growth mechanism in 1972 [1], the techniques of growing a thin base GeSi layer have been highly developed to accommodate the lattice mismatch between the GeSi layer and Si substrate. Technology has been developed for GeSi layer deposition in commercial epitaxial reactors. During the pseudomorphic growth, the lattice of the GeSi layer has to conform with that of the Si substrate; thus, the lattice of the GeSi layer is strained and distorted compared with that of the unstrained GeSi layer. This lattice strain is responsible for
3
the changes in the physical parameters that give the performance improvements of GeSi HBT over Si BJT. The performance of the GeSi HBT demonstrated so far is superior to, or comparable with, that of the GaAs HBTs [67]. The most important advantage of using thin GeSi layer is that its epitaxial deposition is fully compatible with Si integratedcircuit manufacturing technology; thus, monolithic integration of GeSi HBT with Si integrated circuit can be very easily accomplished. Also, the thermal conductivity of Si is three times higher than that of GaAs which is very important in power amplifiers and very large scale integrated circuit (VLSI). The GeSi HBT has very important applications in portable communication systems, such as mobile phones, pagers, DACs and others.
1.2 GeSi HBT Fabrication
Silicon and germanium are completely miscible in any mole fraction ratio. The GeSi material has a diamond lattice structure. The lattice constant can be calculated from the Vegard rule [5]:
a(GexSilx) = a(Si) + x(a(Ge) a(Si)) (1.1) where x is the Ge mole fraction and a is the lattice constant. The lattice constant of Ge is 4.17 percent higher than that of Si. When Si and Ge atoms are mixed during the growth of the GeSi layer, the Si lattice is tetragonally distorted and strained. The lattice mismatch between the GeSi layer and the Si substrate can be accommodated by two ways. One is a generation of misfit dislocations at the GeSi/Si interface which may also propagate through the entire layer. This occurs when the GeSi layer is very thick and the strain is released by creation of misfit dislocations. The second possibility is
4
pseudomorphic growth. In pseudomorphic growth, if the GeSi layer thickness is below a critical thickness, the lattice mismatch is accommodated elastically with no dislocation created. The GeSi layer is under biaxial strain both in the xy plane and in the z or growth direction.
Several epitaxial techniques for GeSi layer deposition were developed, such as the molecular beam epitaxy (MBE) [910] and limited reaction process (LRPCVD) [11]. Figure 1.1 shows a HBT transistor structure used in this thesis which was fabricated by chemical vapor deposition (CVD) [11]. The singlecrystal strained GeSi layer was grown over the nSi collector to give the intrinsic base, and also on the LOCOS SiO2 layer which becomes polycrystalline and is the extrinsic base regions. This results in an interface between the intrinsic and extrinsic base regions. Germane is the source of Ge during the growth and boron impurity doping was obtained with diborane. Boron was then implanted to give the extrinsic base. The oxide/nitride layer were deposited and etched to open the emitter region. The polysilicon layer was then deposited and implanted by a high concentration of arsenic and patterned to form the emitter. The impurity and GeSi concentrations were measured by SIMS (secondary ion mass spectroscopy) as shown in Fig. 1.2. The impurity and Ge concentration profiles can be precisely controlled during film growth to give specified Ge distribution in the base, such as a linearly retrograded Ge gradient. This Ge gradient produces an aiding electric field which accelerates the minority electrons in the base layer.
5
Emitter Boron n+poly ptype polySi base
/GeSi epi
nSi Collecto n+ emitter
ptype intrinsic base
Fig. 1.1 Cross section view of GeSi HBT. Adapted from Neugroschel et al. [1112].
6
10 22
E As 12 Ge:
B8 O__c 8 E 1018
a
0
o 1016i I a
Emitter Base: Collector 0
1015 09
0 XR XB
Depth
Fig. 1.2 Secondary ion mass spectroscopy (SIMS) profile of GeSi HBT. The base doping level is about 3x1018 cm3 and the Ge mole fraction in the base is about 10%. Adapted from Ron Tang et al. [13].
7
1.3 Pseudomorphic Growth, Critical Layer and Geinduced Effects
During the pseudomorphic epitaxial growth of GeSi layer, the lattice mismatch between GeSi layer and Si substrate was accommodated by biaxial strain in the GeSi layer. A twodimensional illustration of the pseudomorphic growth of GeSi layer is shown in Fig. 1.3. As is shown in Fig. 1.3b, under conditions that favor pseudomorphic growth, the GeSi layer is biaxially strained and its lattice assumes that of the substrate. On the other hand, lattice defects such as misfit dislocations occur at the GeSi/Si interface when the GeSi layer becomes relaxed (no strain) as shown in Fig. 1.3c. There is a critical layer thickness beyond which the layer will relax and create the misfit dislocations at the GeSi/Si interface. The dislocations may also penetrate the entire layer. This critical layer thickness can be calculated based on the strain of the GeSi layer and the thermal dynamics theory [57]. Figure 1.4 shows the theoretical and experimental critical layer thickness versus the Ge mole fraction [1]. As is shown in the figure, based on the model of People and Bean, for XGe=10% in the GexSilx layer, the critical layer thickness is almost 1 gm which is enough for high performance GeSi HBT whose base thickness is usually in the range of <1000A. The Geinduced biaxial strain in the GeSi layer changes the physical and chemical properties of the layer, such as the energy band structure [1416], electron and hole effective masses, and electron and hole mobility gn and g [1718]. Figure 1.5 shows the constant energy surfaces in the GeSi layer under biaxial strain [17, 19]. As can be seen from the figure, the two surfaces in the zdirection (growth direction) are deformed which gives a smaller electron effective mass mn and also a higher electron concentration in the z direction, while the energy of
8
the four surfaces on the kx and ky axes is lowered. The resulting reduction in the energy gap and lowering of the effective mass are very beneficial for improved performance of GeSi HBT. The smaller energy gap increases the collector current and gain, while the smaller mn in the vertical minoritycarrier electron flow direction (in npn HBT) increases the minoritycarrier mobility which reduces the base transit time. Figure 1.6 illustrates the inplane (x) of growth direction and normal to the growth plane or the growth (z) direction. Figure 1.7 shows the calculated values of energy gap reduction of strained GeSi layer and unstrained (bulk) material versus the Ge mole fraction. It shows that, for the strained GeSi layer, the energy gap reduction AEGGe is almost a linear function of the increasing Ge mole fraction XGe for XGe<40%.
1.4 GeSi HBT Characteristics
Typical collector current and base current dependencies on VBE of GeSi HBT and Si BJT are shown in Fig. 1.8. The collector current of GeSi HBT, Ic(GeSi), shows large improvement over that in Si BJTs. The base current of GeSi HBT, IB(GeSi), is also slightly smaller than that in Si BJTs. This gives a much larger common emitter forward current gain 3F of GeSi HBT compared to that in Si BJT. Figure 1.9 shows a typical high frequency response versus collector current. Improvement by almost a factor of 2 in fT for GeSi HBT is shown. Recently, fT values as large as 116GHz or higher were reported [20]. This improvement of fT is very important for RF wireless communication applications. The improvement of fT is mainly due to the Geinduced aiding electric field for the base minority carriers which reduces the minority carrier transit time through the base region which still is a limiting factor of fT. Other performance
9
Unstrained GexSi1x Strained GexSilx
 
Si Si
(a) Relaxed GexSi1lx (b)
Si
(c)
Fig. 1.3 Two dimensional illustration of GeSi pseudomorphic growth. (a). Unstrained GeSi layer and Si substrate. (b). Strained GeSi layer. The lattice mismatch is accommodated by the strain in the GeSi layer. (c). Relaxed GeSi layer. There are misfit dislocations at the GeSi/Si interface. Adapted from Sah [1, Fig. 774.1, p. 947].
10
Wm Relaxed c 1 00nm
People and Bean
lOnm
O nm Pseudomorphic
0 20 40 60 80 100 Ge fraction /(1%)
Fig. 1.4 Critical layer thickness of GeSi layer versus Ge fraction. The dashed line shows the theoretical results calculated from the thermal dynamics. The solid line gives the experimental results. Adapted from Sah [1, p. 949].
kZ
k
Fig. 1.5 Constant energy surfaces in the conduction band in GexSil_x under biaxial strain [17, 19]. The z direction corresponds to the growth direction. The two surfaces in the zdirection (growth direction) are raised in energy and deformed such that they give a smaller electron effective mass mn. The four x and y surfaces are lowered in energy. The overall energy gap is smaller and the electrons preferentially populate the two valleys in the zdirection in which mn is low. This is also the direction of the minoritycarrier flow in HBT.
12
Emitter PGeSi base Collector
Fig. 1.6 Direction illustration in vertical GexSil_x HBT. The z direction corresponds to the growth direction and is also the direction of the minority electron flow in a vertical BJT. The inplane direction is important and applicable for lateral MOS transistors.
13
1.2
Unstrained
1.0
0.8
Strained
LIU
0.6
0 0.2 0.4 0.6 0.8 1.0 Si Ge Atomic Fraction, XGe Ge
Fig. 1.7 Energy gap reduction for strained and unstrained GeSi layer versus Ge mole fraction in the layer. Adapted from Sah [1, p. 951].
14
101
GeSi HBT 103  Si BJT VCB=1V, T=296K IC,,,
107
109
1011
0 0.2 0.4 0.6 0.8 1.0 VBE /(1 V)
Fig. 1.8 Typical IV characteristics of GeSi HBT and Si BJT at VCB=1V. The GeSi HBT and Si BJT have the same dopant impurity concentration profile and geometry except the Ge in the base for GeSi HBT. The solid line is GeSi HBT and dashed line is Si BJT.
15
60
I
50
>' 40 GeSiHBT
30
u 20
o 10 SiBJT
0 I I I I I I I II I
102 101 100 Collector Current Ic /(1 mA)
Fig. 1.9 Typical unit gain frequency fT versus Ic characteristics of GeSi HBT (full circles) and Si BJT (open circles). The GeSi HBT and Si BJT have the same dopant impurity concentration profile and geometry except the Ge in the base for GeSi HBT. Adapted from Merit Hong et al. [11].
16
improvements of GeSi HBT, such as higher Early voltage VA will be discussed in detail in later chapters.
1.5 Monte Carlo Simulation of Mobilities
The Geinduced strain in the GeSi layer will change the Ek relationship in both the conduction and valence band, and thus change the effective masses of electrons and holes, mn and mp. The electron and hole mobilities, pn and gp, will be also affected by the change in the effective masses and become anisotropic. Monte Carlo simulations were used to calculate the electron and hole mobilities in strained and relaxed GeSi layers for different Ge mole fraction and dopant impurity concentrations [1718]. Figure 1.10 and Fig. 1.11 show the calculated results for electron mobilities versus Ge mole fraction in ptype and ntype GeSi layer, respectively, for doping level of 1018 cm3 close to that in a typical GeSi HBT [17]. The longitudinal mobility (upward triangles) is in the growth or z direction, the transverse mobility (downward triangles) is in the xy plane. Figure 1.10 and Fig. 1.11 shows only a small (2%) increase in the longitudinal mobility with Ge mole fraction at 300K. However, a much larger improvement (50%) was calculated at 77K for xGe~510%. The transverse mobility shows a large reduction with increasing XGe which is always smaller than that in Si (XGe=O) at any Ge concentration level. The results for the majority and minority hole mobilities are similar to those for the electron mobilities in Fig. 1.10 and Fig. 1.11. In summary, the expected minority electron mobility improvement for the doping level of about 1018 cm3 is very small, gn(GeSi)/4tn(Si)=l.0. This agrees very well with our experimental results as will be shown in chapter 2.
17
co) 600
NAA=10 cm3, T=300K
E 500
longitudinal
400
E 300
C 0
200
., transverse
100
0
0.
0 0.05 0.10 0.15 0.20
Ge mole fraction /(1%)
Fig. 1.10 Minority electron mobility gn in ptype GexSil_ layer. The upward triangles are in the perpendicular z direction which is the growth direction and also the electron movement direction in vertical BJT. The downward triangles are in the xy plane. Adapted from L.E. Kay and T.W. Tang [17].
18
co 1.5
= 1 NDD=1018 cm3, T=300K
E
 longitudinal
0
E
0
a 0.5 transverse
a
0
0 0.05 0.10 0.15 0.20 Ge mole fraction 1/(1%)
Fig. 1.11 Majority electron mobility En in ntype GexSil_ layer. The upward triangles are in the perpendicular z direction which is also the growth direction, and the downward triangles are in the xy plane. Adapted from L.E. Kay and T.W. Tang [17].
19
1.6 Energy Band Diagram
In Si/GexSil_x/Si HBT, a GexSil_x base layer with a smaller energy gap than that of pure Si is sandwiched between the Si emitter and collector. The energy band diagram is very important for analysis and design optimization of the GeSi HBT. Figure 1.12, Fig. 1.13, Fig. 1.14, and Fig. 1.15 show the energy band diagrams for triangular Ge distribution and trapezoidal Ge distribution in the base region at equilibrium (zero bias) and forward active mode (VBE>OV, VCB
XGe(X) = XGe(O) + [XGe(XR) XGe(O)].X/XR (O
Ebi = (1/q)(dEc/dx) = (1/q)(AEGGe/XR) = independent of x
= (kT/q)(r/xR) O
20
AEGGe(O) AEGGe(XB)
Ec Base E
n+Si /nSi
Emitter Collector
Ev
Ge,
NAA .
0 XB
Fig. 1.12 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with a triangular Ge distribution in the base region at equilibrium (zero bias). EF is the Fermi level at equilibrium. AEGGe(O) is the Geinduced energy gap reduction at x=O in the quasineutral base region. AEGGe(XB) is the Geinduced energy gap reduction at x=XB in the quasineutral base region. AEGGe = AEGGe(XB) AEGGe(O) is the total Geinduced energy gap reduction through the base region.
21
AEGGe(O) AEGGe(XB)
VbB  F
Ec n "PGeSi Base FP
FN    Vbi+VCB
n+Si qVBE
Ev Emitter
Ge nSi F NAA Collector
0 XB
Fig. 1.13 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with a triangular Ge distribution in the base region under forward active operation, VBE>OV, VCB>OV. FN is the quasiFermi level for electrons and Fp is the quasiFermi level for holes. Vbi is the builtin potential of the p/n junction. AEGGe(O) is the Geinduced energy gap reduction at x=O in the quasineutral base region. AEGGe(XB) is the Geinduced energy gap reduction at X=XB in the quasineutral base region. AEGGe = AEGGe(XB) AEG.Ge(O) is the total Geinduced energy gap reduction through the base region.
22
AEGGe(O) AEGGe(XB)
Ec Base
EF
Ev Emitter Collector
v, Ge
 I
0 XR XB
Fig. 1.14 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with a trapezoidal Ge distribution in the base region at equilibrium (zero bias). EF is the Fermi level at equilibrium. AEGGe(O) is the Geinduced energy gap reduction at x=O in the quasineutral base region. AEGGe(XB) is the Geinduced energy gap reduction at X=XB in the quasineutral base region. AEGGe = AGGe(XB) AEGGe(o) is the total Geinduced energy gap reduction through the base region.
23
AEGGe(O) AEGGe(XB)
VbiVBEL ae
F _   FN Vbi+VCB
n+Si qVBE/\
E, Emitter //
Ge Ev ",  n~iFN
nSi FN N ,,"" ",Collector
0 XR XB
Fig. 1.15 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with a trapezoidal Ge distribution in the base region under forward active operation, VBE>0V, VcB>0V. FN is the quasiFermi level for electrons and Fp is the quasiFermi level for holes. Vbi is the builtin potential of the p/n junction. AEGGe(O) is the Geinduced energy gap reduction at x=O in the quasineutral base region. AEGGe(XB) is the Geinduced energy gap reduction at x=XB in the quasineutral base region. AEGGe = AEGGe(XB)AEGGe(O) is the total Geinduced energy gap reduction through the base region.
24
1.7 NcNv and n. Model for Linear Ge Distribution
The Ge in the base region changes not only the energy gap, but also other electronic material parameters. The effective density of states in the conduction band and valence band, Nc and Nv, will change because they are a function of the electron and hole effective masses me and mh, respectively which change with the Geinduced strain and change of the Ek relationship in the epitaxial GeSi layer [1718]. Nc and Nv are given by [1]
Nc = 2(27rmekT/h2)3/2 (1.6a) N, = 2(2rmhkT/h2)3/2. (1.6b) Figure 1.16 shows the theoretical calculations of the normalized density of states product NcNv versus Ge mole fraction [1722]. In order to obtain analytical models for the GeSi HBT, we approximate the NcNv product by an exponential function leastsquares fitted to the theoretical data,
Nc(x)Nv(x) = Nc(O)Nv(O).exp(C.x/XR) (1.7)
where the constant C is a nonlinear least squares fit (NLSF) parameter. The model fits the theory well for XGe < 10%, with C = 0.770.1, in particular to the theoretical models of Prinz et al. [2021] and Jain et al. [22] which are also supported by the photoluminescence data and electrical measurements in Si/GexSil_/Si HBTs. The Nc(x)Nv(x) model given by (1.7) enables the decoupling of the diffusivity DB from the NcNv product, as will be shown later.
The Geinduced energy gap reduction AEGGe in the base region will change the intrinsic carrier concentration ni which can be expressed as [1]
25
1.0
1 .0 1 I I I I I 1 1 k. Jain et al. [22] c 0.8 Prinz et al. [21]
Z C=0.66
o 0.6
Z .. C=0.76
S 0.4
> Theory
Z 0.2 NLSF
0 T=300K
0.0
0 2 4 6 8 10 12 14
Ge fraction XGe /(1%)
Fig. 1.16 The theoretical calculations of the normalized density of states product NcNv versus Ge mole fraction [1722]. The solid lines are from calculations in [2122]. The dashed lines are the NLSFs to (1.7) for C=0.66 and C=0.76.
26
nf(x) = Nc(x)Nv(x).exp(EG(x)/kT). (1.8) For a linear Ge distribution using results from section (1.6), we have
EG(X) = EGO AEGHD AEGGe(O) AEGGe'X/XR (1.3) where AEGGe = AEGGe(XR) AEGGe() is the total energy gap reduction across the quasineutral base region. AEGHD is the energy gap reduction due to heavy doping effect, and AEGGe(O) is the Geinduced energy gap reduction at x=O in the quasineutral base region. Inserting equations (1.3) and (1.7) into equation (1.8), gives an analytical expression for the intrinsic carrier concentration ni for a linear Ge distribution,
n2(x) = ni0 2.exp(AEGGe(O)/kT). exp(AEGHD/kT).exp(x/xl), (1.9) where nio is the intrinsic carrier concentration of Si
ni2 = Nc(O)Nv(O).exp(EGo/kT) (1.10) The constant x1 is defined as 1/Xl=(1/XB)[(AEGGe/kT)C]. For a simplicity, we define =AEGGe(O)/kT and y=AEGHD/kT, then equation (1.9) becomes
n?(x) = nio 2*expa.exp y.exp(x/xl). (1.11)
1.8 Summary
In conclusion, the basic parameters governing the GeSi HBT performance were discussed in this chapter, including fabrication by pseudomorphic epitaxial growth, critical layer thickness, and electronic properties of the GeSi layer. The lattice mismatch between the GeSi layer and the Si substrate is accommodated by the biaxial strain in the GeSi layer. This Geinduced strain in the GeSi layer is responsible for changes in electronic material parameters important for HBT performance, in particular, the energy gap, density of states, and mobility. For the Ge mole fraction XGe <~40%,
27
the energy gap in strained GeSi layer decreases linearly with XGe. Monte Carlo simulation results for minority and majority electron and hole mobilities were also discussed. Analytical approximations for the intrinsic carrier concentration n; and density of states dependence on the Ge mole fraction were developed here in this thesis for an idealized linearlygraded Ge concentration profile in the GeSi base layer.
Measured IBVBE and IcVBE curves and high frequency responses of GeSi HBT and Si reference BJT have demonstrated the performance improvements in GeSi HBT compared to Si BJT.
CHAPTER 2
LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) METHODOLOGY
2.1 Introduction
In the last chapter, we presented a preliminary qualitative analysis of the GeSi HBT and its comparison with the Si BJT. In this chapter, we describe a lowfrequencyconductancevoltage (LFGV) methodology developed in order to obtain accurate quantitative values for the dominant minoritycarrier transport parameters that underlie the GeSi HBTs performance enhancement, such as the diffusion length, recombination lifetime, and diffusion coefficient in the base layer. The LFGV method also gives the analysis methodology of current gain and emitter injection efficiency based on a separation of the total base current into the recombination losses in the emitter and base. Thus, the LFGV method gives a complete chargecontrol analysis of the bipolar transistor performance. Analytical models are developed for the LFGV method for a trapezoidal and linearly graded Ge concentration profile in the base. However, the LFGV method is general and can be applied to any bipolar transistor with an arbitrary dopant impurity and materialcomposition profile, including those using the IIIV compound semiconductors.
Comprehensive reviews of the Si/GexSilx/Si HBT technology, including fabrication, Geinduced effects in the base layer, currentvoltage characteristics, modeling, and performance comparison with Si BJT are discussed in chapter 1. Performance
28
29
improvements in collector current, transit time and output conductance in GeSi transistors over those in similar Si transistors were demonstrated qualitatively and graphically. In this chapter, we present detailed analysis based on an analytical model, to extract reliable and accurate values of the minoritycarrier parameters in the GeSi material. These parameters are essential to assess the Geinduced effects on transistor performance, design optimization and to verify theoretical predictions about the expected performance benefits due to Ge.
The key point in obtaining these quantitative values is a separation of the various Geinduced effects from the measured data. A lack of an experimental methodology for this analysis is the main reason why these values are still not available in the literature. For example, in GeSi transistors, the increase in the collector current Ic can arise from the aiding buildin electric field in the base due to the Ge retrograded profile, the base minoritycarrier diffusivity enhancement, or both. An accurate separation of the performance improvement between the two parameters requires independent measurements of the builtin field and mobility. Yet another factor that affects the collector current is the reduction of the effective density of states, Nc and Nv, which are functions of the effective masses mn and mp, respectively. Theoretical calculations show that the density of states product, NcNv, in thin pseudomorphic GexSil_x layers grown on (100) Si substrate is significantly decreased compared to that in Si due to Geinduced strain [1718]. This gives (NcNv)GeSi/(NcNv)si 0.6 for GeSi layer with xGe = 5% [1718]. The minoritycarrier mobility was predicted to increase with Ge concentration for xGe,<10% especially at higher base impurity concentration (1019
30
cm3) and at low temperatures (77K) [17]. But, the mobility enhancement has not yet been experimentally confirmed in fabricated GeSi HBTs. The builtin electric field and higher minority carrier mobility increases the collector current IC and current gain 1F, but this is compensated by the decrease in the density of states. Thus, a methodology to separate these Geinduced effects is the key to a comprehensive quantitative analysis of the Si/GexSilx/Si HBT.
In this work, we present the LFGV methodology for a complete chargecontrol analysis of HBTs, including a quantitative comparison with Si reference transistors, and an experimental demonstration giving the first quantitative measurements of the builtin electric field and minoritycarrier recombination lifetime in the GeSi base.
2.2 Base Thickness Modulation (BTM)
Lowfrequencyconductancevoltage (LFGV) methodology is based on the combination of the dc IcVBE and IBVBE measurements with the lowfrequency (lkHz) smallsignal conductance measurements resulting from the base thickness modulation [2324]. The base thickness modulation (BTM) effect is illustrated in Fig. 2.1 for a Si BJT. The collector current IC is proportional to the gradient of the base minority carrier distribution in the base layer at a certain forward bias emitterbase voltage VBE and reverse bias collectorbase voltage VCB. When a smallsignal voltage Vcb is superimposed upon the dc basecollector junction voltage VCB, the quasineutral base thickness XB will be modulated by AXB, as shown in Fig. 2.1. The shadowed area in Fig. 2.1(b) shows the change in the minority carrier distribution in the quasineutral base due to the base thickness modulation, and the base minority carrier charge QB
31
AXB 4H4
n+ p n
0 jib Xs
SII II II II
VBE vVc
N(0) AQB
B::
01 Xa
(b)
superimposed on dc value VCB.
(b) noritycarrier distribution in the base.
II II II II II
(b)
32
change by AQB with the smallsignal Vcb. The minority carrier concentration at the emitter boundary of the quasineutral base at x=O, N(O), remains constant for a fixed VBE and is given by [n2/P(0)].exp(qVBE/kT), where P(O)=NAA(0) is the holeemitter boundary of the quasineutral base at x=O, N(O) remains constant for a fixed VBE and is given by [n2/P(0)].exp(qVBE/kT), where P(O)=NAA(O) is the hole concentration at x=O, k is the Boltzmann constant, and T is the temperature. The change in the slope of N(x) gives the smallsignal collector current ic. The output conductance go is defined by
c10
go (2.1)
Vcb VBE
The base current IB also changes due to a change in the base recombination current, QB/TB, with the base thickness modulation by the smallsignal voltage Vcb. This produces the reverse transconductance gr which is defined by
Ib
gr (2.2)
Vcb VBE
where ib is the smallsignal base current due to Vcb. Analytical expressions for go and gr can be derived (see appendix A),
ic Ic dXB
go90 (2.3)
Vcb VBE XB dVCB VBE
ib QB dXB
gr = 0 (2.4)
Vcb VBE TB XB dVcB VBE
where XB is the quasineutral base thickness, QB is the base minority carrier charge corresponding to VBE and VCB, and B is the base minority carrier recombination lifetime. Since Ic and QB are exponential functions of the forward baseemitter voltage
33
VBE, go and gr are also exponential functions of VBE. Taking the ratio of equations (2.3) and (2.4), we obtain a very important relation for the analysis of the bipolar junction transistor [1, p. 822, 2425]:
go Ic QB/tB TB
(2.5)
gr QB/TB QB/TB tB
where tB is the base minority carrier transit time.
In order to accurately measure the output conductance go and especially the reverse transconductance gr (gr<
2.3 LFGV Methodology
Two LFGV methodologies were developed. The first LFGV methodology described in this section is based on a comparison of the transistor characteristics measured in Si/GexSil_/Si HBT and reference Si BJT where the difference between the measured characteristics is due to Ge only [2627]. By analyzing this difference, we can then calculate the various Geinduced material parameter changes. This approach has been commonly used by other investigators [2834]. The basic assumption in this comparative approach is that the two transistors have nearly identical geometry and dopant impurity concentration profiles except the Ge incorporation in the GeSi HBT
34
base layer. The second method using only the GeSi HBT without the reference Si BJT will be discussed in chapter 3.
There are several performance enhancement gauges of GeSi HBT over Si BJT. The most important and also most commonly used by investigators is the dc collector current Ic enhancement [3536]. For example, for a uniform base doping and uniform Ge concentration in the ptype base of npn transistor, the ratio of the collector currents can be written as
Ic(GeSi) (NcNV)GeSi Dn(GeSi)
exp(Ay).exp(AEGGe/kT) (2.6) Ic(Si) (NcNv)si Dn(Si )
where NC, Nv are the density of states of the conduction band and valence band, respectively, Dn is the minoritycarrier electron diffusivity, AEGGe is the Geinduced energygap reduction in the base, and Ay = AEGHD/kT=[AEGHD(GeSi) AEGHD(Si)]/kT is the normalized difference between the heavy dopinginduced energy gap reduction in GeSi and Si base layers.
There are four unknowns in (2.6) all related to the Ge in the GeSi base layer. The key issue is how to separate these Geinduced effects. Two approaches have been presented in the literature using the Ic ratio in (2.6) [2932]. The first approach uses the measurement of the temperature dependence of Ic(GeSi)/Ic(Si), assuming that the density of states and mobility ratios are temperature independent. Then by plotting the Ic ratio versus 1/T, equation (2.6) yields an activation energy EA = AEGGe + AEGHD = AEGGe. This AEGGe can then be used to calculate the improvements in the current gain, transit time, and other characteristics as a function of the Ge mole fraction and
35
temperature. But theoretical calculations show that the density of states and mobility have very complicated temperature dependence, so the assumption of the temperature independence of the two preexponential factors in (2.6) is unreliable and can lead to large errors in AEGGe, as also noted previously [32]. The second approach is to assume theoretically predicted values of (NcNv)GeSi and Dn(GeSi) and then calculate EA from (2.6). Apparently, there is another unknown, Ay, and also the theoretical results are strongly dependent on models and input parameters. Furthermore, both approaches give an estimate for the energy gap reduction term only and do not provide values or estimates for the mobility or recombination lifetime in the GeSi base.
The above discussion shows that an analysis based on the measurement of Ic(GeSi) only is not sufficient. The LFGV methodology, however, will use a combination of the dc IBVBE and IcVBE characteristics with the smallsignal output conductance go and reverse transconductance gr measurements. This novel methodology can overcome the limitations of the above and other approaches.
The LFGV methodology will be presented here for the idealized trapezoidal Ge concentration profile in the base which is shown in Fig. 2.2. In this model, the base dopant impurity concentration profile and the Ge distribution are divided into two regions. In the first region, O
36
16
1022
E NDD 12
C)
Ge
c 1020 N, (
,AA i E
o E
IZ 8
0 1018
4 4
Emitter Base: Collector 1016 0
0 XR XB
Depth
Fig. 2.2 Idealized model for the Ge and dopant impurity concentration profiles in GeSi HBT. The dopant impurity concentration profiles for Si BJT (no Ge in the base) are the same as those in GeSi HBT.
37
 AEG(O) is the total Geinduced energy gap variation through O
Ebi = (1/q)(dEc/dx) = (1/q)(AEGGe)/XR = independent of x. (2.7a)
= (kT/q)(R/X,), (2.7b) where ir = AEGGe/kT is the normalized Geinduced energy gap reduction through the base.
For XR
Ebi2 =(kT/q)[n2/(X,XR)]. (2.8)
The LFGV methodology also uses the collector current ratio Ic(GeSi)/Ic(Si), but this ratio is modified. The Geinduced NcNv reduction term NcNv(GeSi)/NcNv(Si) is lumped together with the Geinduced energy gap reduction term AEGGe=(kT/q)r where the lumped parameter is described by a factor, rll. This modification gives several benefits. First, rlI can be measured directly and accurately, while the original two factors are not measurable. Secondly, all other GeSi HBTs parameters can be calculated in terms of rl. This modification also enables the decoupling of the effects of NcNv from that of the mobility and heavy doping on Ic(GeSi).
38
The collector current density for nonuniform base can be expressed as [38]
Jc(GeSi)=[exp(qVBE/kT)1]/fB [PB/qDn(x)ni(x)]dx (2.9) Using the analytical model for a trapezoidal Ge distribution in the base region as discussed in section 1.7, the intrinsic carrier density in the GeSi base is given by nm(x) = NcNv(x)exp(EGo/kT).expa.expy.exp(r x/XR), (O
NBO (O
P(x) = (2.11)
NBO.exp[(xxR)/x2] (XR
Ic(GeSi) NcNV(GeSi) DB(GeSi) a+Ay+q A + (lA)(1exp(q2))/12
e (2.12) Ic(Si) NcN,(Si) D,(Si) expr1 1exp(R2) A +(1A)
where Ay = [AEGHD(GeSi)AEGHD(Si)]/kT and A=XR/XB. (2.12) can then be used to evaluate the effect of AEGGe from the temperature dependence measurement or calculation as discussed in section 2.1. But in LFGV method, we choose to use a simple model for the NcNv(x), as is discussed in section 1.7, and then lump this NcNv variation with the AEGGe.
39
Nc(x)Nv(x) = Nc(O)Nv(O)exp(Cx/XR) (2.13)
where C is a LSF parameter. Then the intrinsic carrier concentration is given by
ni 2. expa. exp y. exp (x/x ) (O
n =x (2.14)
n(x) ni expa.expy exp(xR/X1), (XR
where ni is the intrinsic carrier concentration in low doped Si given by
ni2 = Nc(O)Nv(O).exp(EGO/kT), (2.14a) and XI=XB/[(AEGGe/kT)C]=XB/ 1. Combining (2.9), (2.11) and (2.14) yields the ratio of the collector current in GeSi and Si Ic(GeSi) DB(GeSi) A+(1A)[1exp(q2) /2]
exp(R1+a+Ay). (2.15) Ic(Si) D,(Si) A(expy11)/nI+(1A)[1exp(q2) ]/2 where l = rlC is the normalized effective energy gap variation through the base layer that includes the effects of variable Ge concentration and variation of density of states with XGe. In (2.15), DB(GeSi) = IB [P/n2(x)]dx / fB Pdx/Dn(x)n2(x) is the spatiallyaveraged electron diffusivity in the base. The weighting factor of DB(GeSi) is ni2(x) which increases with x or XGe. Thus, the value of DB(GeSi) is weighted toward the region close to X=XB. Since Dn(GeSi) is expected to be a very weak function of XGe for the average XGE<5% [28], defining DB(GeSi) in (2.7) causes only a negligible error. For a purely linearlygraded Ge concentration profile, XR = XB and A=1, which simplifies (2.15) to [27]
Ic(GeSi) Ri DB(GeSi)
.expa expAy (2.16) Ic(Si) 1exp(nj) D,(Si)
One reason for lumping the spatial variation of NcNv with the spatial variation of the energy gap AEGGe is that these two parameters always appear together as a product in
40
the calculation of other GeSi parameters, such as the transit time. Thus, the analysis of the GeSi HBT with a trapezoidal or linearlygraded Ge concentration profiles can be conveniently made in terms of a single parameter ll. It is important to note here that, as will be shown later, the value of the fitting parameter C = (r rl), which is model dependent, does not affect the determination of the base diffusion length, recombination lifetime, transit time, or diffusivity. It only affects the calculation of the builtin electric field Ebi and AEGGe, which may give a small error due to the uncertainty of the fitting parameter C. However, this analytical approach is much more accurate than using the spatially averaged value of NcNv in the integral equation (2.9), which allows taking NcNv outside the integral. Another very important feature of the new expression (2.15) is that it decouples the (NcNv) product from mobility in (2.12) and allows determination of the mobility ratio.
In order to analyze the Geinduced Ic enhancement, we have to separate the measured ratio in (2.15) or (2.16) into five or four terms. A rough approximation for practical profiles with xGe(O)=0 is to assume that the enhancement in Ic is dominated by the builtin field which leaves only the first term in (2.16) and leads directly to the estimation of r 1. However, for an accurate analysis of the transistor and for purpose of acquiring a material database, more accurate approach is necessary. For example, an uncertainty in AEGGe(0) or AEGHD of 10meV gives an error of exp(10/25)= 1.5. Thus, an analysis of the performance of Si/GexSilx/Si HBT based on measurement of Ic only is clearly not sufficient.
41
The LFGV methodology described in the next section will give an accurate extraction of rl.
2.4 Derivation of LFGV Methodology and Parameter Extraction
Below we present detailed derivation and description of the LFGV methodology. As is discussed in section 2.2, the base thickness modulation gives rise to two smallsignal conductances, an output conductance, go, and a reverse transconductance, g,. In the first step of the LFGV methodology, we need the go only. It can be derived from the Ic equations (2.9) and (2.15) and is given by (see appendix A)
Ic 1 aXB
go ... (2.17)
XB A(expr11)expl2/01 + (lA)(expR2l)/q2 aVcB VBE
From (2.17), we can see immediately that the expression (2.17) for go is simpler than (2.15) for Ic. The reason for this is that the unknown parameters, DB(GeSi)/DB(Si), a, and y are lumped into Ic which is a measurable quantity. This is one of the advantages of go compared with Ic and is one important aspect of the LFGV methodology. Useful relations (2.18a) and (2.18b) are derived by taking the ratios of the collector currents and the output conductances of the GeSi HBT and Si reference BJT, Ic(GeSi)/Ic(Si) A[(expn11)expR2]/R1 + (1A)(expn21)/n2 (2.18a)
go(GeSi)/go(Si) Aexp2 + (1A)(expR21)/ 2 Ic(GeSi)/Ic(Si) expRm1
XR = XB (2.18b) go(GeSi )/go(Si) R
The only assumption in (2.18a) and (2.18b) is that aXB/aVCB is the same for GeSi HBT and Si reference BJT. This assumption is proven to be a very good approximation from measuring the junction capacitance CBC. These relations are used to calculate rl
42
uniquely and independently of diffusivity, density of states and heavydoping effects which cancel out in the ratios of Ic and go.
Equivalent relations to those in (2.17), (2.18) can be derived for the reverseactive operation mode with the BC junction forwardbiased and the EB junction reversebiased. For this case exp(rl) and exp(r2) are replaced by exp(ll) and exp(r2) and rl and rl2 are replaced by T11 and r12*
From (2.18b), we can see that for a triangular Ge distribution, we can calculate the rl directly from the measured collector current and output conductance. But for a trapezoidal Ge distribution in (2.18a), we need to know A=XR/XB and r2. In the later part of this section, we will show that r12 can be calculated from a combination of forward active measurement and reverse active measurement of Ic(Si), go(Si) and gr(Si) of the Si reference BJT. Also we will show that the Ic ratio is not sensitive to XR/XB and t12, since they are the same for GeSi HBT and Si BJT and nearly canceled out in the Ic ratio. After an independent determination of rll, the determination of other parameters is straightforward and will be discussed together with the experimental results in the next section.
The total base recombination current IB is composed of the recombination losses in the emitter and base regions,
QB QE
IB = + (2.19)
TB TE
where QE and QB are the stored charges, and TE and TB are the chargecontrol minoritycarrier recombination lifetime or time constant in the emitter and base, respectively. In
43
(2.19) we neglected the spacecharge region recombination components that can be easily separated out from the measured IgB In order to separate the recombination losses in (2.19), we need the gr measurement which results from the modulation of the base recombination loss. For the GeSi HBT, gr is derived as
QB X 1 aXB
gr (2.20)
TBXB Y A[(expql1)/q1)]expq2 + (1A)[(expq21)/q2] aVB "VBE In (2.20), the terms X and Y are given below
X = A2(exp1+i11)/012 +
A(1A)[(expq11)/ 1].[(expn21)/02] +
(1 A)2(exPq 21)02 / 022 (2.21a)
Y = A2[exp(1)+11)]/q12 +
A(1A)[1exp(q1)].[1exp(q2)]/R102 +
(1A)2[exp(q2)+02]/022. (2.21b) When we take the ratio of gr and go, a useful expression is obtained:
gr tB
B (2.22)
90 TB
where tB is the base minority carrier transit time, TB is the chargecontrol minority carrier base recombination lifetime, and B is a factor related to the builtin electric field in the base due to both Ge and doping nonuniformity (B=1.0 in fieldfree base). B is a function of rl, r12, ni(x), and hole concentration in the base P(x) and is given by (2.23) below [39]
1 XB ni 2(x) XB dx
B dx 1 (2.23)
tBDB 0O PB(x) 0 ni 2(x)/P(x)
44
= X/Y (2.24a)
expn111
X= = XB. (2.24b)
exp( q)+Ri1
A combination of the dc IBVBE and smallsignal measurements gives the expressions for the base diffusion length and ratio of the recombination losses in the emitter and base region:
QE/TE B(go/9r)
1 (2.25) QB/TB PF
L2 = X2(g0/gr){A2(expr1)/R12 +
A(1A)[(exprl1)/om].[(exp21)/2] +
(1A)2(expnr2 21)/ 22} (2.26a)
0 +exp(qi)1
L2 = XB2( o/g ). XR = XB. (2.26b) r12
A combination of (2.19) and (2.25) gives (QB/TB)GeSi and (QB/TB)Si from the measured data. Using (2.26a) or (2.26b) we then calculate directly the diffusion length where the only unknown parameter is XB which can be measured by SIMS.
Expression (2.22) is also used to calculate the parameter r12 which determines the builtin field for XR
goF/goR BF tBF
(2.27)
9rF/9rR BR tBR
45
The ratio in (2.27) is a function of A=XR/XB and 12 only. If XR/XB is known from SIMS or fabrication conditions, r12 can be calculated from (2.27).
2.5 Experiments and Results
High performance GeSi HBT and reference Si BJT are used to demonstrate the LFGV methodology. SIMS profile of the completed GeSi HBT is shown in Fig. 1.2. The reference Si BJT is fabricated with the identical fabrication steps except it has no Ge in the base layer. Measurements of the EB junction capacitance (4.44fF/Lm2 for GeSi and 4.37fF/gm2 for Si), breakdown voltage (5.2V for GeSi and 5.1 for Si), and base sheet resistance under the emitter (6.2kG/square for GeSi and 6kKG/square for Si) confirm the similarity of the Si and GeSi base concentration and thickness, which is important for a quantitative performance analysis to delineate the effects of Ge. All the measurements are taken on large transistors with the EB junction area of 100.4x100.8gpm2 in order to eliminate the edge and sidewall effects.
Figure 2.3 shows typical IBVBE and IcVBE characteristics of GeSi and Si transistors. Note the significant increase in Ic and slight decrease in IB for the GeSi transistor. The measured enhancement in the collector current Ic(GeSi)/Ic(Si) = 3.86. The insensitivity of Ic(GeSi) to the reverse basecollector voltage up to VBc=5V shows the absence of parasitic energy barrier at the BC junction that may reduce Ic [4041]. Figure 2.4 gives the measured smallsignal output conductance, go, reverse transconductance, gr, for both the forward active mode and reverse active mode, and the collector current Ic of GeSi HBT, as a function of VBE, showing the expected exp(qVBE/kT) dependence [42]. Figure 2.5 gives the measured smallsignal output
46
conductance, go, reverse transconductance, gr, for both the forward active mode and reverse active mode and the collector current Ic of Si BJT. The measured dependencies in Figs. 3.43.6 are used for the LFGV analysis. The results for large square device (100.4x100.8lm2) are summarized in Table 2.1. A detailed discussion of the analysis and its implications follows.
The first step in the analysis is to find the parameter r12 using (2.27) which determines the builtin field for XR
With the known r12, we can then calculate the key parameter rl from the measured Ic and go for the GeSi and Si transistors using (2.18a). This gives rl1 = 1.98, and r = (r11 + C) = 2.7+0.1 using C = 0.70.1. The Geinduced energy gap reduction at XGe=9.8% is AEGGe = (kT)r = 705meV. This corresponds to AEGGE = 7XGE meV which is in very good agreement with the theoretical calculations and other experiments in GeSi as a function of XGe. This good agreement demonstrates the accuracy of our model that accounts for the dependence of the NcNv on Ge concentration. The value of the fitting parameter, C, causes only small uncertainty in the results. The Geinduced aiding electric field within the linear Ge range, O
The mobility ratio DB(GeSi)/DB(Si) can be evaluated from (2.15). The remaining unknowns in (2.15) are a = AEGGe(O)/kT and Ay = AEEHD/kT. SIMS profiles indicate a = 0 at VBE=0.52V. This gives [DB(GeSi)/DB(Si)].exp(Ay) = 1.6. This product cannot be separated without additional independent measurement of either the mobility ratio or
47
10'1
GeSi HBT 103  Si BJT ,, VCB=1 V, T=296K ,
S107/
109 / /
10 /11
0 0.2 0.4 0.6 0.8 1.0 VBE /(1 V)
Fig. 2.3 ICVBE and IBVBE characteristics of GeSi HBT and Si BJT. The measurements were taken on large transistors with the EB junction area of 100.4x100.8tm2 in order to eliminate the edge and sidewall effects. Note the significant increase in IC and slight decrease in IB for the GeSi transistor. The measured enhancement in the collector current Ic(GeSi)/Ic(Si) = 3.86.
48
104 Si/GexSil.x/Si HBT .oI 10.
T=296K ICF
10. go 90F 10grR <
S1010 10~9
1012 I I I I 1011
0.3 0.4 0.5 0.6 0.7 VBE /(1 V)
Fig. 2.4 The measured forward active mode and reverse active mode smallsignal output conductances, go, reverse transconductance, gr, and the collector current, Ic, of GeSi HBT, as a function of VBE. All measured curves show the expected exp(qVBE/kT) dependence.
49
I I I
04 Si BJT 10T=296K goR
106 _g 10.5
0rF 10 2
0 10 10
1012 I I I I 1011
0.3 0.4 0.5 0.6 0.7 VBE /(1lV)
Fig. 2.5 The measured forward active mode and reverse active mode smallsignal output conductances, go, reverse transconductance, gr, and the collector current, Ic, of Si BJT, as a function of VBE. All measured curves show the expected exp(qVBE/kT) dependence.
50
Table 2.1 Summary of experimental data and calculated parameters for Si/GexSilx/Si HBT and reference Si BJT. The top six lines give the measured dc and BTM ac data obtained at 23"C, VBE = 0.52V and VCB = 1V. The computed parameters using LFGV method are shown in the next eighteen lines. The first column is for Si/GexSilx/Si HBT, the second for Si baseline BJT, and the third column gives the comparison ratio of the two previous columns. ABE=100.4x100.8pm2, XR=680A, X,=800A.
Parameter GeSi HBT Si BJT GeSi/Si
Ic (gA) 7.74 200.5 3.86 IB (nA) 59.4 69.9 0.85 OF 130.3 28.6 4.54 go (1/2) 2.125x108 1.67x108 1.27 gr (1/Q) 3.0x10' 1.8x1010 0.17 go/gr 708.3 92.78 7.63
102 2.7 2.7 I'l 1.98
AEGGe (meV) 68.6
Ebi (kV/cm) 10
B 12.6 2.6 4.8 LB (9m) 3.6 0.8 4.5 QB/TB (nA) 0.87 8.4 0.1 QE/'E (nA) 58.5 50.6 1.16 (QE/'E)/(QB/'B) 67 6 11.2 (QE/TE)/IB (%) 98.5 88 1.12 (QB/'B)/IB (%)) 1.5 12 0.13 QB (C) 21x108 10x1018 2.1 DB (cm2/V.s) 5.5 5.5 1.0 tB (ps) 2.7 4.8 0.56 TB (ns) 24 1.1 22 Ay 0.47 AEGHD (meV) 12 VA (V) 364 120 3.0
51
Ay. Figure 2.6 shows the measured temperature dependence of ln[exp(Ay).D B(GeSi)/DB(Si)] in the range of 200K300K. The temperature dependence shows a small activation energy of EA12meV. This is consistent with Ay = AEGHD/kT = 0.47 or AEGHD = 12meV which is calculated from the measured room temperature Ic ratio assuming a unity mobility ratio which is independent of temperature. A negligible improvement in the minoritycarrier mobility for the average 5% Ge concentration and
3x1018cm3 base doping at room temperature was also predicted theoretically as is discussed in section 1.5. The conclusion that DB(GeSi) = DB(Si) is also supported by the sparameter measurements of the base transit time for HBTs. Using the measured data in Fig. 2.6, we can calculate the temperature dependence of the minority carrier diffusivity ratio assuming that AEGHD= 12meV is constant with the temperature range of 200K296K. The results are shown in Fig. 2.7 and are in good agreement with theoretical predictions [17].
The recombination components of IB are calculated from (2.19) and (2.25). The calculation shows that the recombination losses in the emitter and base constitute 98.5 and 1.5% of Ig, respectively in GeSi HBT and 88% and 12% in Si BJT. An important selfconsistency result is that (QE/'E)GeSi = (QE/CE)Si, as anticipated since xGe=0 at the EB junction and, thus, the hole injection from the base into the emitter and the emitter recombination are expected to be the same in both transistors. The small differences can be explained by differences in the emitter profile and recombination properties at the n+polysilicon/n+Si emitter interface. The very significant reduction in the recombination losses in the base, (QB/tB)GeSi/(QB/TB)Si = 0.1 is due to a significantly
52
2 .0 1 1 1 1 1 1 1 1
2.0
m
1.5
a,
1.0
0.5 Experiment
AEGHD=12meV C0 I I I
3 4 5 6 1000 /T /(1/1 K)
Fig. 2.6 The measured temperature dependence of ln[exp(Ay).D B(GeSi)/DB(Si)] in the
range of 200K300K normalized to 296K in large 100.4x100.8gm2 transistor.
53
2.0
Co
1.5
0 1.5
0.5 I I
200 220 240 260 280 300 Temperature /(1 K)
Fig. 2.7 The measured temperature dependence of DB(GeSi)/DB(Si) in the range of
200K300K normalized to 296K in large 100.4x100.8um2 transistor.
54
larger base minority recombination lifetime zB ratio 'B(GeSi)/TB(Si) = 24ns/1.lns = 21. The reasons for the significant improvement of the minoritycarrier recombination lifetime in thin strained GeSi layers is not known and requires further study. The diffusion length ratio can be calculated from (2.26) and gives LB(GeSi)/LB(Si) = 3.6ptm/0.8 = 4.5 in the GeSi layer. This ratio is calculated directly from the smallsignal BTM measurements without using the dc IBVBE curve. This is an advantage for HBTs that have a large BE junction leakage or large spacecharge layer recombination component. TB was calculated taking DB=5.5cm2/s [27] for both transistors. Although ZB is relatively short (=20ns), it has only a small effect on the gain since IB in the large 100.4x100.80m2 transistors is dominated by the emitter. Rough estimates of TB in GeSi HBTs were published previously [4244], but our present measurements in large transistors are the first rigorous accurate data.
The minority carrier base charge QB is larger in GeSi transistor, QB(GeSi)/QB(Si) = 2.1 because of smaller energy gap in the base which increases ni2. This increase in the stored charge should be accounted for in speed calculations. The electron transit time across the quasineutral base layer is then calculated and gives 2.7ps/4.8ps=0.56, a factor of about two improvement for the GeSi HBT, as expected. This is due almost entirely to the builtin field from the graded Ge concentration profile rather than the mobility enhancement which is negligible. The Early voltage for GeSi HBT is significantly higher which is important for analog circuit application.
The results for a stripe transistor with narrow emitter fingers (0.8x26gm2 with 88 fingers) are summarized in Table 2.2. The major difference between large square
55
transistor and stripe transistor is due to the perimeter recombination at the edges of the stripe transistor. But, the narrowemitter transistor is used in the circuit applications because the narrowemitter geometry minimizes the base series resistance. So the measurements for stripe transistor are also important. Due to the edge effect, some measured parameters, especially those associated with the base current, such as IB, 3F, gr and go/gr, are different from those of the large square device. Two key parameters, rlI and tB, are however very similar for both transistors geometries. This is because rll and tB are calculated from Ic and go only which are essentially onedimensional even for the narrow emitter transistor with a very thin base. Figure 2.8 shows the measured temperature dependence of ln[exp(Ay)DB(GeSi)/DB(Si)] in the range of 150K300K. The temperature dependence shows a small activation energy of EA8meV. For this transistor pair, the measured [DB(GeSi)/DB(Si)]exp(Ay) = 1.0 compared to 1.6 for the large transistor. This gives DB(GeSi)/DB(Si)=0.7. The disagreement with the DB(GeSi)/DB(Si)= 1.0 result obtained from the large transistor is due to the narrowfinger geometry.
2.6 Sensitivity of tB and il to n, and XR/XB
An idealized base dopant impurity concentration profile P(x) and Ge distribution in the base used for the analysis are shown in Fig. 2.2. The base region is divided into two regions, derived from the actual SIMS profile shown in Fig. 1.2. For O
56
Table 2.2 Summary of experimental data and calculated parameters for Si/GexSilx/Si HBT and reference Si BJT. The top six lines give the measured dc and BTM ac data obtained at 23"C, VBE = 0.58V and VCB = 1V. The computed parameters using LFGV method are shown in the next eighteen lines. The first column is for Si/GexSilx/Si HBT, the second for Si baseline BJT, and the third column gives the comparison ratio of the two previous columns. ABE=(0.8x26)x88pm2, XR=680A, XB=800A.
Parameter GeSi HBT Si BJT GeSi/Si
Ic (ptA) 14.42 5.16 2.79 IB (nA) 161.6 173.2 0.93 OF 89.2 29.8 3.0 go (1/92) 2.687x108 4.08x108 0.66 gr (1/Q2) 3.5x1010 5.2x1010 0.67 go/gr 76.77 78.46 0.98
T12 2.2 2.2 11 2.5
AEGGe (meV) 79
Ebi (kV/cm) 12
B 15 2 7.5 LB (pm) 1.2 0.7 1.7 QB/ZB (nA) 12 32 0.38 QE/'E (nA) 149 141 1.06 (QE/'E)/(QB/'B) 12 4.4 2.7 (QE/'E)/IB (%) 92 82 1.1 (QB/tB)/IB (%)) 8 18 0.4 QB (C) 35x108 25x1018 1.4 DB (cm2/Vs) 4 5.5 0.7 tB (ps) 2.4 4.9 0.5 TB (ns) 2.8 0.8 3.5 Ay 0.3 AEGHD (meV) 8 VA (V) 536 127 4.2
57
0.5 lI" "
C) Ex eriment
S0.4 LSF
co 0.3o 0.2O. 0.1
2 3 4 5 6 7 8
1 000/T /(1/1 K)
Fig. 2.8 The measured temperature dependence of ln[exp(Ay).D B(GeSi)/DB(Si)] in the range of 150K300K normalized to 296K for stripe transistor. Least squares fit to the data is shown by the solid line and gives an activation energy of EA = AEGHD = 8+0.2meV.
58
will certainly cause some errors since the SIMS profiles may not be very accurate for very thin base layers, especially near the collector junction. Note that the base dopant impurity concentration profile NAA(X) can be better approximated by a Gaussian distribution (see Appendix B for Gaussian distribution model). This results from the boron outdiffusion from the GeSi layer during subsequent fabrication. The increasing doping NAA(x) at x=0 shown in Fig. 1.2 will produce a retarding electric field which will decelerate the base minority carrier movement across this region. The decreasing doping density will produce an aiding electric field. Thus, an uncertainty in a precise location of XR can cause errors in calculation of the transit time and other parameters. In order to estimate the influence of this uncertainty on the GeSi HBT analysis, we investigate the dependence of r 1, tB on XR/XB and r12. Figure 2.9 shows the normalized Geinduced electric field rl versus XR/XB with r12 as a parameter. r~l is the key parameter underlying the operation of GeSi HBT. Figure 2.10 shows the transit time ratio tB(GeSi)/tB(Si) versus XR/XB with '12 as a parameter for r11=2.0. As can be seen from Fig. 2.9 and Fig. 2.10, rl and tB(GeSi)/tB(Si) are rather insensitive to the location of the transition point XR. This is expected, since rl and tB(GeSi)/tB(Si) are calculated from the collector current ratio of the two transistors with nearly identical geometry and dopant impurity concentration profiles. The above sensitivity analysis shows that taking x=XR as the transition point where the Ge linear profile intersects the flat Ge concentration profile in the measured SIMS dependencies, Fig. 1.2, will result in a negligible error.
59
3.0 1 1 1
2.5
:. 2=1'6 1.8
2.03 2.7
1.5
1 .0 a .I I I I I I I I 1
0.6 0.7 0.8 0.9 1.0 XR/XB
Fig. 2.9 Sensitivity of TI, to XR/XB calculated from (2.18a) with r12 as a variable parameter.
60
1 .0 I I I
1=2.0
0.8
0.6
9 2=1.4 1.6 1.8
( 0.4
(.
m 0.2
0.5 0.6 0.7 0.8 0.9
XR/XB
Fig. 2.10 Sensitivity of tB(GeSi)/tB(Si) to XR/XB calculated from (A.12) with r12 as a variable parameter for 1=2.0.
61
2.7 GeSi HBT Performance Enhancement
In this section, we will plot the important GeSi HBT parameters as a function of rl normalized to that of Si BJT in order to show the performance enhancement of GeSi HBT over Si BJT. Since the triangular Ge distribution in the base region is a simple and also a very good approximation for the trapezoidal Ge distribution, we will use the triangular Ge distribution as a demonstration. The extension to the trapezoidal Ge concentration profile is straightforward. In fact, most of the important performance enhancements can be accomplished by the triangular Ge distribution in the base region.
The most important performance enhancement of GeSi HBT over Si BJT is the increase in the collector current Ic(GeSi), as is shown in Fig. 2.11. The collector current ratio Ic(GeSi)/Ic(Si) varies almost linearly with rl for r1>3.0. For 1l>3.0, the collector current ratio Ic(GeSi)/Ic(Si) can be modeled by a linear function, Ic(GeSi)/Ic(Si) o r oc (AEGGekT)C.
The output conductance ratio go(GeSi)/go(Si) vs rl is shown in Fig. 2.12. As can be seen from the figure, the go ratio shows a rapid decrease for 1.0
62
The base minority electron distribution for 1q1=2.0 is shown in Fig. 2.13. As can be seen from the figure, n(x) throughout the base is larger for GeSi HBT. This is because the intrinsic carrier concentration n2 of GeSi HBT is larger than that of Si BJT.
The base minority carrier charge ratio QB(GeSi)/QB(Si)>1.0 vs rl is shown in Fig. 2.14. The increase of base minority carrier charge will not affect the performance of GeSi HBT significantly because the aiding electric field in the base region increases Ic. Thus, the minoritycarrier transit time through the base, tB=QB/Ic is still smaller for the GeSi HBT. This is shown in Fig. 2.15 for the forward active mode.
The transit time improvement is very important for high frequency performance. As is shown in the figure, the base transit time can give a factor of two improvement for r11=2.0, which corresponds to about 10% Ge in the base. The improvement of tB(GeSi) is due to the Geinduced aiding electric field Ebi, rather than the enhancement of diffusivity Dn, which is negligible according to our measurements and theoretical calculations.
The base transit time ratio tB(GeSi)/tB(Si) for the reverse active mode is shown in Fig. 2.16. As expected for the Ge concentration profile in Fig. 1.2, tBR(GeSi)/tBR(Si) >>
1.0.
2.8 Summary
In conclusion, a new LFGV methodology was developed and a complete analysis of the effects of Ge in the base of Si/GexSil_/Si HBT was presented. The LFGV methodology is based on a onedimensional BJT theory and accurate dc and lowfrequency measurements of similar GeSi and Si transistors. It is accurate, expedient and
63
10
T=300K
O
0
C)
0 . i I, 1, I, I I, I, I, 1 ,
0 2 4 6 8 10
Fig. 2.11 Collector current ratio Ic(GeSi)/Ic(Si) vs r11 (normalized to [DB(GeSi)/DB(Si)].exp(a+Ay)). For r1>3.0, the collector current ratio Ic(GeSi)/Ic(Si) can be modeled by a linear function, Ic(GeSi)/Ic(Si) o i11 o (AEGGe/kT)C.
64
1.0
T=300K
0.8
0
0 0.6
a 0.4
o 0.2
0
0 2 4 6 8 10
Fig. 2.12 Output Conductance ratio go(GeSi)/go(Si) vs rll for the forwardactive mode (normalized to [DB(GeSi)/DB(Si)].exp(a+Ay)). The go ratio shows a rapid decrease for 1.0
65
1.0
T=300K
0.8 ""
0.6 "0.4
 n(x) GeSi HBT
0.2  n(x) Si BJT
0 0.2 0.4 0.6 0.8 1.0
x/XB
Fig. 2.13 Base minority carrier distribution of GeSi HBT and Si BJT vs x/XB for r1=2.0 in forward active mode (normalized to ni.[exp(qVBE) 1]/NB).
66
2.0
0 1.8% T=300K
C 1.8
O
1.6
C)
a) 1.4
m 1.2
1.0
0 2 4 6 8 10 r1i
Fig. 2.14 Base minority carrier charge ratio QB(GeSi)/QB(Si) vs rl for the forward active mode (normalized to exp(a+Ay)).
67
1.0
T=300K
0.8
U)
(3
0.6
0.4
m 0.20 2 4 6 8 10
Fig. 2.15 Base minority carrier transit time ratio tB(GeSi)/tB(Si) in forward active mode vs rl (normalized to DB(GeSi)/DB(Si)). The base transit time can give a factor of two improvement for 111=2.0 which corresponds to about 10% Ge in the base.
68
100
 T=300K
C/) 80
C 60 a 40
a20
0 2 4 6 8 10 rT
Fig. 2.16 Base minority carrier transit time ratio tB(GeSi)/tB(Si) in reverse active mode vs 1I (normalized to DB(GeSi)/DB(Si)). The reverse base transit time is much larger than in the forward active mode.
69
general. The analytical model for the Geconcentrationdependence of the density of states product decouples it from the minoritycarrier mobility and heavydoping effects. A combination of Ic and go measurements gives a unique determination of the critical parameter rll. All the important minoritycarrier transport parameters underlying the improved performance of GeSi HBT compared to similar Si BJT were determined and quantified.
The results show that the reduction in the base transit time is due to the Geinduced aiding electric field in the region with linearlyvariable Ge concentration. The minority carrier mobility or diffusivity enhancement was found to be negligible compared to Si BJT. The energy gap reduction due to the heavy doping (NBase=3x1018cm3) is about 10meV larger in GeSi HBT than in Si BJT at room temperature. First accurate measurements of the minority carrier diffusion length and recombination lifetime in the base show greatly improved values in the GeSi base compared to Si. The base current was found to be dominated almost completely by the recombination within the emitter in the large 100.4x100.8gm2 transistors, while in circuit transistors with a narrow emitter stripe (<1m), the base recombination is more important due to increased importance of the sidewall and surface effects. Since the dc and lowfrequency smallsignal measurements are independent of parasitics, accurate highfrequency and highspeed parameters can be readily obtained which are difficult to measure directly such as the lps base transit time. The performance improvements of GeSi HBT over Si BJT were shown graphically as a function of rl1.
CHAPTER 3
LFGV ANALYSIS OF GeSi HBT WITH NO REFERENCE Si BJT
3.1 Introduction
In the previous chapters, the performance improvements of GeSi HBT over Si BJT are discussed and demonstrated. The LFGV methodology for accurate determination of parameters of GeSi HBT was also developed. This methodology is based on simultaneous measurement of GeSi HBT and Si BJT with nearly identical dopant impurity concentration profiles such that the difference in the performance is due to Ge only. The disadvantage of this approach is that a reference Si BJT is not always available. Furthermore, it is very difficult to fabricate GeSi and Si transistors with nearly identical dopant impurity concentration profiles. In this chapter, we describe another LFGV method for GeSi HBT analysis that does not use a reference Si BJT.
One of the key parameters in the LFGV methodology is the normalized Geinduced energy gap reduction, i=AEGGe/kT, which is due to the compositional grading of Ge in the base. This energy gap variation produces an aiding electric field, Ebi which significantly decreases the electron transit time through the base layer that yields higher speed and fT [4546]. Thus, accurate and independent measurement of the rl and the resulting Ebi is essential for the design, optimization and analysis of GeSi HBT, and for the verification of theoretical predictions about the expected performance benefits due to the compositional grading. The determination of rl is also a basis for a complete
70
71
chargecontrol analysis of the recombination components in the base current [47]. The new method is demonstrated for Si/GexSil_/Si HBT with a linear or trapezoidal Ge concentration profile in the base. The Geinduced energygap reduction parameter r is determined solely from the terminal measurement of GeSi HBT without any assumptions. The builtin electric field Ebi can be then calculated, independent of the minority carrier mobility and heavydoping effects. The method also does not require a knowledge of the dopant impurity concentration profile in the base.
3.2 Theory
The SIMS profile for the analyzed GeSi HBT is shown in Fig. 1.2 [48]. The Ge is linearly graded within O
The collector current can be expressed for any base dopant impurity concentration profile and compositional variation in the GeSi base as [1, p. 760, 38]
q exp(qVBE/kT)
JC = (3.1)
XB P(x)
0 Dn(x) ni2( dx
0 D0(x).n12(x)
72
qDB exp(qVBE/kT)
(3.2)
X0 nP(x) dx
where DB = IP(x)/n2(x)/IP(x)/Dn(x)ni(x)dx is the average minority electron diffusivity in the base. If we superimpose a small sinusoidal signal on the dc reverse BC junction bias, with the BE junction forwardbiased, the smallsignal output conductance, goF, due to the basethickness modulation can be obtained by taking the derivative of JC or IC with respect to VCB,
alC aXB Ic 1 P(XB) .aXB
g9oF = = (3.3)
aXB aVB IXB P(x) ni2(XB) aVcB dx
S ni2(x)
In (3.3), P(XB).a XB)/aVCB = (1/q)(aQB/aVcB) is the smallsignal basecollector junction capacitance per unit area, CBC. Thus, (3.3) can be written as
ICF CBC
goF = (3.4)
0F XB P(x) n2(XB) dx
JO ni2(x)
Similarly, we can obtain goR for the reverseactive operation with the EB junction reversebiased and the BC junction forwardbiased,
ICR CBE
goR = (3.5)
90R XB P(x) ni2(0) (3.5)
0 ni2(x) dx
The integral in the denominator is the same both for the forwardactive and reverseactive operation modes. Thus, the ratio of equations (3.4) and (3.5) gives
73
goR ICR n12(XB) CBE
= (3.6)
90F ICF ni2(0) CBC
From (3.6) we obtain the ratio of the intrinsic carrier concentrations at the two edges of the quasineutral base layer which is given below in (3.7)
n2(XB) ICF g0R CBC
= (3.7a)
n (O0) ICR goF CBE
NC(XB)NV(XB)
= .exp(AEGGe/kT). (3.7b)
Nc(O)Nv(O)
The result in equation (3.7) is completely general and applies to any HBT or BJT with any compositional grading and dopant impurity concentration profile, including IIIV HBTs.
Equation (3.7) uses no assumptions about the material composition or impurity doping in the base and gives the ratio of the intrinsic carrier concentrations at the two edges of the quasineutral base layer directly from the measured collector current, output conductance, and junction capacitances for any HBT or BJT. The approximation using the average base diffusivity in (3.2) does not affect the ratio in (3.7), since the integral relation for DB is the same for the forwardactive and reverseactive operation modes.
One possible application of the result in (3.7) is measurement of the energy gap in GeSi as a function of Ge concentration using test structures with different Ge linearlygraded profiles as shown in Fig. 3.1. Equation (3.7) can be also used to measure the intrinsic carrier concentration ratio due to heavydopinginduced energy gap reduction using structures with a highlow dopant impurity concentration profiles as shown in Fig.
3.2.
74
3.3 Experiments and Results
A high performance Si/GexSilx/Si HBT is used to demonstrate the application of the result in equation (3.7). The Ge concentration profile is shown schematically in Fig. 1.2. The Ge gradient from SIMS showed a linearlygraded Ge concentration from about XGe=0% at the EB junction to a peak of about 10% at the BC junction. The base dopant density was nearly constant at =3x1018cm3, and the quasineutral base thickness was about 800A. The experimental data is tabulated in Table 3.1. All the measurements were performed on large area 100.4x100.8 pm2 transistors to eliminate the perimeter edge and twodimensional effects. The difference between ICF and ICR is due to slightly larger base thickness for the reverse mode, which is expected for very thin base. The output conductances goF and goR, and the junction capacitances CBC and CBE were measured at 1kHz using precise bridge WK B214.
With reference to the profile in Fig. 1.2 and the intrinsic carrier concentration model for linear Ge concentration profile in equations (2.10), and assuming that the difference in the intrinsic carrier density at x=0 and x=XB is due to Ge grading, we obtain from (3.7),
n? (XB)
n?() expq1 = exp(qr C) = exp [(AEGGe/kT) C], (3.8a)
n2(O)
ICF goR CBC
(3.8b)
ICR goF CBE
where C is the least squares fitting parameter for the density of states product NCNv for linear Ge distribution discussed in detail in section 1.7. Using the data in Table 3.1 we obtain, n (XB)/n?(0) = 7.68 from (3.7). This gives T11=2.04 and AEGGe = 70meV where
75
G14
S 2Ge
E 1022
o 12
S102 E
83
CN C c 1018 NAA ( o0 4 E
O 1016 Emitter Base Collector 0
0 XR XB
Fig. 3.1 GeSi HBT with different Ge concentration profiles in the base region.
76
CO
1022
E ND
I
10, NAA(X)
C
10
S10 Emitter Base Collector
0 XR XB Depth
Fig. 3.2 Bipolar transistor with different highlow dopant impurity concentration profiles in the base region.
77
C=0.7 is used. The aiding electric field Ebi = (kT/q)(rj/XR) = (0.0255)(2.74/680x108) = 10.3kV/cm. This value compares very well to Ebi= 10kV/cm which is obtained by the comparative method discussed in chapter 2 (section 2.5) using the comparison LFGV methodology.
In conclusion, the above example demonstrates one application of the new measurement method, namely the measurement of the energygap variation due to the compositional grading across the GeSi base. Similar results can be obtained for IIIV HBTs with graded base. The new nonreference methodology can be also used to measure the temperature dependence of the parameter ll. GeSi HBT shows a superior low temperature Ic enhancement compared to Si BJT. This is because the builtin electric field Ebi is inversely proportional to the temperature, Ebi" GGe/kT, where k is the Boltzmann constant and AEGGe is the Geinduced energy gap reduction across the quasineutral base region. Thus, at lower temperature, the builtin electric field is greatly enhanced. All other parameters related to Ebi will be also enhanced. For example, the collector current Ic(GeSi) will increase, the output conductance go(GeSi) will decrease, and the base minority carrier transit time tB will decrease significantly which gives higher speed and fT. Theoretical calculations of the minority mobility as a function of temperature and doping were presented in the literature [17, 5253]. In order to verify these theoretical predictions and analyze the GeSi HBT performance at lower temperature, the determination of the temperature dependence of 1 is very important. This was investigated by measuring the GeSi HBT characteristics at several different low temperatures. Using the models and equations given in the section 3.2, we can
78
calculate the important GeSi HBT parameters as a function of temperature. The results are shown in Table 3.2 and Table 3.3. From Table 3.2, we can see that the basecollector junction capacitance CBC remains constant as temperature decreases while the baseemitter junction capacitance CBE shows a slight decrease with decreasing temperature. This is because the thickness of the basecollector junction SCR, XBC, is much larger than XBE due to lower doping at the collector junction. This gives CBc<
From Fig. 3.3, we can see that rl obtained by a combination of (3.7) and (3.8), and thus the builtin electric field Ebi, increase almost linearly with 1/T. The dashed line is calculated using II=AEGGe/kT C, where AEGGe=70meV as measured at 297.15K, and C=0.7. The deviation of the theoretical curve from the experimental dependence may come from two sources: (1) decrease of the AEGGe with decreasing temperature, and (2) the increase of the fitting parameter C with decreasing T. The latter one is more reasonable since the NcNv product is a strong function of temperature as is shown in equations (1.6a) and (1.6b). The increase of C at low temperature means that the NcNv(GeSi)/NcNv(Si) is a decreasing function of temperature.
Figure 3.4 shows the measured diffusion length LB(GeSi) vs T. The diffusion length can be expressed as
LB(GeSi)= Dr = (kT/q)Pr (3.10)
79
Since the gt' product is not strongly temperature dependent [5455], LB(GeSi)o4T, or increasing with increasing T, as shown in Fig. 3.4.
From Fig. 3.5, we can see that the base recombination component (QB/TB)GeSi/IB increases with decreasing temperature. Thus, the base recombination becomes more important as temperature decreases.
Table 3.1 Summary of measured parameters of GeSi HBT.
goF 2.125x108 (S) goR 6.549x106 (S) ICF 7.739 (gA) ICR 12.45 (A) CBC 1.446 (pF) CBE 39.676 (pF)
80
Table 3.2 Temperature dependence of measured parameters of GeSi HBT
T(K) goR/goF ICR/ICF CBC(pF) CBE(pF)
297.2 285.2 1.63 1.99 39.4 250 495.9 1.81 1.99 39.0 225 630.7 1.85 1.99 38.6 200 967.3 2.29 1.99 38.4 175 2792 3.94 1.99 38.2 150 4000 2.88 1.99 38.0 125 13,000 4.81 1.99 37.7
Note: AE=100.4x100.8jtm2, T=297.15K, VBE=0.52V, VBc=.0OV.
The junction capacitances CBC and CBE were measured at f=lkHz.
Table 3.3 Temperature dependence of GeSi HBT parameters
calculated from the measured data in Table 3.2
T(K) l1 LB [QB/TB] /B
(gm)
297.2 2.04 3.9 1.2% 250 2.60 4.3 0.9% 225 2.83 4.0 1.0% 200 3.05 3.2 1.5% 175 3.58 2.6 2.3% 150 4.25 1.8 4.4% 125 4.93 1.5 4.2%
81
8
Experiment
6 Theory
6
2
0 I i I I I I
2 4 6 8 10
1000/ /(1/1 K)
Fig. 3.3 Normalized energy gap reduction parameter rll vs 1000/T.
82
6
E
A *
() 2
0
J
0 I I
100 150 200 250 300
Temperature /(1 K)
Fig. 3.4 Diffusion length LB(GeSi) T.
83
6
O
CO
0
0 I I
2 4 6 8 10 100ooo/T /(1/K)
Fig. 3.5 Base recombination component QB(GeSi)/IB(GeSi) vs 1000/T.
84
3.4 LFGV Analysis of GeSi HBT with no Si Reference
After the determination of rl described in the previous section, a complete chargecontrol analysis of the GeSi HBT can be made. The only other parameters needed for this analysis are 112 which determines the builtin electric field in the region of XR
goF/goR BF tBF
S (3.10)
grF/grR BR tBR
The ratio in (3.10) is a function of rll, A=XR/XB and 112 only. Since 111 can be determined independently as described in section 3.2 and XR/XB is known from SIMS or estimated from fabrication conditions, r12 can be then calculated from (3.10). Using the data in Table 2.1 and r1=2.04, we obtained 112=2.6 which agrees very well with the comparative method in section 2.5 which gives 12=2.7. We assume that DB(GeSi)=DB(Si) as determined in chapter 2. The results of a complete charge control analysis for the GeSi HBT following the procedure described in chapter 2 are summarized in table 3.4. Since 11=2.04 and r12=2.6 obtained using the nonreference analysis are almost the same as those obtained using the comparative LFGV methodology in section 2.5 (r11=1.98 and r12=2.7), the results in table 3.4 are very similar to those in table 2.1.
85
Table 3.4 Summary of experimental data and calculated parameters for Si/GexSilx/Si HBT by LFGV methodology with no Si reference BJT. The top six lines give the measured dc and BTM ac data obtained at 23C, VBE = 0.52V and VCB = 1V. The computed parameters are shown in the next sixteen lines. ABE=100.4x100.8pm2, XR=680A, X,=800A.
Parameter GeSi HBT
Ic (gA) 7.74 Ig (nA) 59.4 F 130.3 go (1/ 2) 2.125x108 gr (1/Q) 3.0x10" go/gr 708.3
12 2.6 rll 2.0 AEGGe (meV) 69 Ebi (kV/cm) 10 B 12
LB (gm) 3.5 QB/'B (nA) 0.91 QE/TE (nA) 58.5 (QE/TE)/(QB/TB) 65 (QE/TE)/IB (%) 98.5 (QB/ZB)/IB (%)) 1.5 QB (C) 21x108 Dg (cm2/Vs) 5.5 tB (ps) 2.8 rB (ns) 23 VA (V) 364
86
3.5 Summary
In conclusion, a simple and unique methodology is presented for the direct measurement of the ratio of the intrinsic carrier densities at the two edges of the quasineutral base layer in HBT or BJT. The method is based on a combination of dc and smallsignal lowfrequency (1kHz) measurement of the transistor characteristics. The normalized energy gap parameter il, the energy gap reduction AEGGe, and the builtin electric field Ebi are determined independently of the base dopant impurity concentration profile, minority carrier mobility and heavydoping effects in GeSi HBT with a linear or trapezoidal Ge concentration profile. Then a complete charge control analysis of GeSi HBT is made without using a Si reference BJT. The temperature dependence of rl, diffusion length LB(GeSi), and base recombination component (QB/TB)GeSi were measured to investigate the performance of GeSi HBT as a function of temperature.
CHAPTER 4
SUMMARY AND CONCLUSIONS
4.1 Summary
In chapter one, we gave a general review of the GeSi HBT fabrication, including the discussion of pseudomorphic growth of GeSi layer and its critical layer thickness. A comparison of the Si/GexSilx/Si HBT characteristics with those for a Si reference BJT, Monte Carlo simulations of the mobility improvement and other Geinduced changes in the material properties were also discussed. An exponential approximation model for the density of states NcNv dependence on the Ge concentration was given for a linear Ge distribution in the base. In chapter two, a lowfrequencyconductancevoltage (LFGV) methodology was developed for detailed analysis of the Si/GexSilx/Si HBT. This methodology is based on a comparison of the GeSi HBT performance with that of a reference Si BJT. A brief review of the base thickness modulation method which underlies the LFGV method is also given. Experimental data were presented to demonstrate the LFGV methodology in highperformance Si/GelSil_x/Si HBTs. Important performance enhancements of GeSi HBT over Si BJT were measured and analyzed. In chapter three, we introduced a methodology for GeSi HBT analysis without using a reference Si BJT. In this nonreference analysis, the key parameter rl of GeSi HBT can be determined directly and a complete charge control analysis can be made. Experimental demonstration of the nonreference method was given.
87
88
4.2 GeSi HBT Modeling and LFGV Methodology
The LFGV methodology is applicable to an analysis of any heterojunction bipolar transistor (HBT), including those based on IIIV materials. The LFGV methodology uses a combination of the dc and smallsignal measurements of the transistor characteristics. The dc characteristics are the collector and base current as a function of the emitterbase junction bias. The smallsignal characteristics are the output conductance and reverse transconductance resulting from the base thickness modulation. They are measured at low frequency (lkHz) to avoid parasitic capacitance and inductance effects.
The LFGV methodology was applied for the analysis of Si/GexSil_/Si HBT with a linearlygraded or trapezoidal Ge concentration profiles in the base. Analytical expressions for the dc and smallsignal parameters were developed. Important minority carrier transport parameters, such as the base diffusion length, recombination lifetime, diffusivity, builtin electric field, and base transit time were determined. The LFGV methodology also allows a separation of the base current into the components recombining in the base and emitter region, respectively.
APPENDIX A
GENERAL EQUATIONS FOR BJT A. The general integral relations for calculation of important parameters in bipolar junction transistor are given below. Equations for a trapezoidal Ge distribution in the base and constantexponential dopant impurity concentration profile in the base are given in section B.
1. Collector current Ic(GeSi) [1, p. 760, 38]
q exp(qVBE/kT)
IC = (A.1)
XB P ( x)
0 Dn(x).ni2(x)
2. Minority carrier distribution in the base [38]
Jc n (x) XB (Y)
n(x) dy (A.2)
q P(x) x Dn.ni2(y)
3. Minority carrier charge in the base
X,
QB= n(x)dx (A.3)
4. Minority carrier base transit time [38]
1 X, n?(x) XB P(y)
tB = . dx (A.4)
Dn 0 P(x) x n2(y)
5. Correction factor B [27]
1 XB P(x) XB n (x)
B dx dx 1 (A.5)
DntB lo n2(x) P(x) 89
90
6. Output conductance go
alc aXB IC 1 P(XB) .aXB
go = = (A.6)
aXB aVcB [XB P(x) ni2(XB) aVCB
dx
0 ni2(x)
7. Reverse transconductance gr
a(QB/TB) aXB
gr = (A.7)
aXB aVcB
8. Current gain
IC q 1
p = (A.8)
IB IBO XB P(x)
dx
0 Dn(x) ni 2(x)
where IBO is the base saturation current.
9. Early voltage VA and OVA product
Ic qni2( ()Dn(X) XB P(x)
VA = dx (A.9)
go CBC 0 Dn(x). ni2(x)
q2
pVA n= ( (XB) .Dn(XB) (A.10)
IBO CBC
B. Equations for a trapezoidal Ge distribution and constantexponential dopant
impurity concentration profile in the base of GeSi HBT (Fig. 2.2) obtained from the
general integral relation (A. 1)(A. 10) are given below.
91
1. Collector current Ic
qDnnio2 .exp(a)exp(y)exp(01) .exp(qVBE/kT)
JC= (A.11)
NBoXB[(XR/XB)(expqll)/l+(lXR/XB)(1exp(2) )/R2]
2. Minority carrier base transit time tB
X 2
tB = Y (A.12)
Dn
3. Output conductance go
Ic 1 aXB
go (A.13)
XB A(expq11)/n)expn2 + (1A)(expp21)/q2 aVCB VBE
4. Reverse transconductance gr
QB X 1 aX,
gr= *. (A.14)
TBXB Y A[(expnl1)/q1)]expq2 + (1A)[(expn21)/02] aVcB VBE In (A. 1.9) and (A.1.11), the terms X and Y are given below X = A2(expn1+n11)/012 + A(1A)[(expno1)/ll].[(expp21)/2I +
+ (1A)2(expn2n021)/q22 (A.15) Y = A2[exp(n1)+n11)1]/12 + A(1A)[1exp(0 )].[1(expq2)]/ 192+
+ (1A)2[exp(02)+n211/n22. (A.16)
5. Correction factor B
B = X/Y (A.17)
6. PVA product
[ expl11 expn21
VA = XB A expn2 + (1A) (A.18)
01 r2 aXB/aVcB
q2
PVA = ni02.Dn(XB) .expnl (A.19)
IBO CBC
APPENDIX B
TRANSISTOR RELATION FOR A GAUSSIAN
DOPANT IMPURITY CONCENTRATION PROFILE IN THE BASE OF BJT
I. Most bipolar junction transistors are fabricated using ion implantation to dope the base which gives a Gaussian dopant impurity concentration profile. For GeSi HBT, a Gaussian profile may result from the impurity doping outdiffusion from the epitaxially deposited base layer during subsequent fabrication. Thus, the relevant LFGV equations for the Gaussian distribution are important. Two different Gaussian distribution are considered and shown in Fig. B.1 and Fig. B.2. The Gaussian concentration profile P(x) shown in Fig. B.1 has a peak value at x=O. The profile shown in Fig. B.2 has a peak value inside the base layer. The dashed line shows the exponential approximation for P(x).
Following are the analytical derivations of the equations for the two Gaussian distributions P(x).
1. Base dopant impurity concentration profile P(x) shown in Fig. B. 1 P(x) = NAA.exp(x2/r2) (B.1) where a is
U = XB/[In[NAA/P(XB)]'/2 (B.2) XB
f P(x)dx = 7/2 N.oAAN erf(XB/U) (B.3)
92
93
10'
Gaussian Distribution
 Exponential Distribution c)
1016
0 400 800 1200 1600 2000
x/(1A)
Fig. B.1 Gaussian distribution and exponential approximation.
94
2. Collector current
qDnn2 [exp(qVBE/kT)1]
I, = (B.4)
n/2 NAA erf(XB/o)
3. Minority carrier base transit time tB= [erf(XB/) fBexp(x2/U2)dx fBexp(x2/a2)erf(x/a)1dx (B.5)
2 Dn
4. Output conductance
21c exp(XB2/g2) aXB
go = (B.6)
~ir. erf(XB/a) aVCB
5. Reverse transconductance
Ic exp(XB2/a2) aXB
gr = [1Bexp(x2/U2)erf(x/a)dx]. (B.7)
DnTB erf(XB/) aVB
6. Correction factor
fBexp(x2/2) erf(x/)dx
B = (B.8)
erf(XB/o) fexp(x2/U2)dx fBexp(x2/g2)erf(x/u)dx
For the profile in Fig. B.1, the value for the correction factor is obtained numerically using (B.8) which gives B=7.0. The exponential approximation gives B=6.0 using much simpler analytical expression (A. 17). Thus, the exponential approximation of the Gaussian profile in the base of a bipolar junction transistor gives only a small error for the value of B. This allows a simple charge control analysis of bipolar junction transistor using the exponential profile approximation shown in sections 2.3 for

86
3.5 Summary
In conclusion, a simple and unique methodology is presented for the direct
measurement of the ratio of the intrinsic carrier densities at the two edges of the
quasineutral base layer in HBT or BJT. The method is based on a combination of dc
and smallsignal lowfrequency (~lkHz) measurement of the transistor characteristics.
The normalized energy gap parameter r^, the energy gap reduction AEG_Ge, and the
builtin electric field Ebi are determined independently of the base dopant impurity
concentration profile, minority carrier mobility and heavydoping effects in GeSi HBT
with a linear or trapezoidal Ge concentration profile. Then a complete charge control
analysis of GeSi HBT is made without using a Si reference BJT. The temperature
dependence of rj Â¡, diffusion length LB(GeSi), and base recombination component
(QB^B^GeSi were measured to investigate the performance of GeSi HBT as a function of
temperature.
The LFGV methodology is used here to study the performance improvements of
Si/GexSij_x/Si HBT over that of the Si BJT, including current gain, switch speed, and
Early voltage which are primarily due to the aiding electric field.
Analytical models for the effective densityofstates NCNV and intrinsic carrier
concentration n as a function of Ge mole concentration for a linear Ge distribution in
the base region are developed. Analytical solutions for the LFGV methodology are then
derived for an arbitrary Ge and dopant impurity concentration profile. The analysis
showed that the main effect of Ge in the base region with a linearly retrograded Ge
concentration profile comes from the aiding electric field which reduces the base transit
time and increases the collector current and the cutoff frequency fT. Accurate
measurements of the base minority carrier diffusion length and recombination lifetime
indicate large performance improvement of GeSi HBT over Si BJT. Design
considerations for optimal performance of GeSi HBT are also investigated, such as the
optimal Ge concentraction profile in the base region. The LFGV methodology is
applicable to other bipolar junction transistors and HBTs, including those fabricated in
IIIV materials.
vi
27
the energy gap in strained GeSi layer decreases linearly with xGe. Monte Carlo
simulation results for minority and majority electron and hole mobilities were also
discussed. Analytical approximations for the intrinsic carrier concentration nÂ¡ and
density of states dependence on the Ge mole fraction were developed here in this thesis
for an idealized linearlygraded Ge concentration profile in the GeSi base layer.
Measured IBVBE and ICVBE curves and high frequency responses of GeSi HBT and
Si reference BJT have demonstrated the performance improvements in GeSi HBT
compared to Si BJT.
NcNv(GeSÂ¡)/NcNv(Si)
25
Fig. 1.16 The theoretical calculations of the normalized density of states product NCNV
versus Ge mole fraction [1722], The solid lines are from calculations in [2122], The
dashed lines are the NLSFs to (1.7) for C=0.66 and C=0.76.
38
The collector current density for nonuniform base can be expressed as [38]
Jc (GeSi) = [exp (qVBE/kT)1] / JB [PB/qDn (x) n? (x) ] dx (2.9)
Using the analytical model for a trapezoidal Ge distribution in the base region as
discussed in section 1.7, the intrinsic carrier density in the GeSi base is given by
nf(x) = NcNv(x)exp(EGo/kT) .expa.expy.exp(rjx/XR), (0
nf(x) = NcNv(x)exp(EGo/kT) expa.expy .exp(r?), (XR
where a=AEG_Ge(0)/kT and y=AEG_HD(GeSi)/kT and have been defined in section 1.7.
The base doping in the GeSi base layer is given by
nbo
(0
P(x) =
(2.11)
lNB0.exp[(xxR)/x2] (xR
where x2 is defined as x2=(XBXR)/ln[NAA0/P(XB)].
If we use average values for the NCNV and the diffusivity Dn, then we can take these
two quantities out of the integral in (2.9) and obtain the following expression for the Ic
ratio
Ic(GeSi) NcNv(GeSi) DB(GeSi) a+Ay+q A+ (1A)(lexp(i72))//?2
= e (2.12)
Ic(Si) NCNV(Si) Db(Si) exprjl lexp(^2)
A +(1A)
1 12
where Ay = [AEG_HD(GeSi)AEG_HD(Si)]/kT and A=XR/XB. (2.12) can then be used to
evaluate the effect of AEG_Ge from the temperature dependence measurement or
calculation as discussed in section 2.1. But in LFGV method, we choose to use a simple
model for the NcNv(x), as is discussed in section 1.7, and then lump this NCNV
variation with the AEG.Ge.
42
uniquely and independently of diffusivity, density of states and heavydoping effects
which cancel out in the ratios of Ic and g0.
Equivalent relations to those in (2.17), (2.18) can be derived for the reverseactive
operation mode with the BC junction forwardbiased and the EB junction reverse
biased. For this case exp(rj) and exp(r2) are replaced by explr^) and exp(T]2) and rij
and r2 are replaced by iq x and r2.
From (2.18b), we can see that for a triangular Ge distribution, we can calculate the rj
directly from the measured collector current and output conductance. But for a
trapezoidal Ge distribution in (2.18a), we need to know A=XR/XB and r2. In the later
part of this section, we will show that r\2 can be calculated from a combination of
forward active measurement and reverse active measurement of Ic(Si), g0(Si) and gr(Si)
of the Si reference BJT. Also we will show that the Ic ratio is not sensitive to XR/XB
and r2, since they are the same for GeSi HBT and Si BJT and nearly canceled out in the
Ic ratio. After an independent determination of iq x, the determination of other
parameters is straightforward and will be discussed together with the experimental
results in the next section.
The total base recombination current IB is composed of the recombination losses in
the emitter and base regions,
Qb Qe
IB = + (2.19)
tb te
where QE and QB are the stored charges, and xE and xB are the chargecontrol minority
carrier recombination lifetime or time constant in the emitter and base, respectively. In
PAGE 1
/2: )5(48(1&< &21'8&7$1&( 92/7$*( /)*9f &+$5$&7(5,=$7,21 2) 6L*HALA6L +(7(5281&7,21 %,32/$5 75$16,67256 %\ *82;,1 /, $ ',66(57$7,21 35(6(17(' 72 7+( *5$'8$7( 6&+22/ 2) 7+( 81,9(56,7< 2) )/25,'$ ,1 3$57,$/ )8/),//0(17 2) 7+( 5(48,5(0(176 )25 7+( '(*5(( 2) '2&725 2) 3+,/2623+< 81,9(56,7< 2) )/25,'$
PAGE 2
$&.12:/('*(0(176 ZRXOG OLNH WR WKDQN P\ VXSHUYLVLQJ SURIHVVRUV 'U $UQRVW 1HXJURVFKHO DQG 'U &KLK7DQJ 6DK IRU WKHLU JXLGDQFH GXULQJ P\ GRFWRUDO UHVHDUFK ZRXOG DOVR OLNH WR WKDQN 'U 3HWHU =RU\ 'U 9ODGLPLU 5DNRY DQG 'U 0DUN 2UD]HP IRU VHUYLQJ RQ P\ VXSHUYLVRU\ FRPPLWWHH DOVR WKDQN LQ &DL :DQ3HQJ &DR 'HUHN 0DUWLQ DQG
PAGE 3
7$%/( 2) &217(176 $&.12:/('*(0(176 LL $%675$&7 Y &+$37(56 ,1752'8&7,21 72 *H6L +(7(5281&7,21 %,32/$5 75$16,6725 ,QWURGXFWLRQ *H6L +%7 )DEULFDWLRQ 3VHXGRPRUSKLF *URZWK &ULWLFDO /D\HU DQG *HLQGXFHG (IIHFWV *H6L +%7 &KDUDFWHULVWLFV 0RQWH &DUOR 6LPXODWLRQ RI 0RELOLWLHV (QHUJ\ %DQG 'LDJUDP 1&19 DQG QI 0RGHO IRU /LQHDU *H 'LVWULEXWLRQ 6XPPDU\ /2: )5(48(1&< &21'8&7$1&( 92/7$*( /)*9f 0(7+2'2/2*< ,QWURGXFWLRQ %DVH 7KLFNQHVV 0RGXODWLRQ %70f /)*9 0HWKRGRORJ\ 'HULYDWLRQ RI /)*9 0HWKRGRORJ\ DQG 3DUDPHWHU ([WUDFWLRQ ([SHULPHQWV DQG 5HVXOWV 6HQVLWLYLW\ RI W% DQG UM [ WR U? DQG ;5;% *H6L +%7 3HUIRUPDQFH (QKDQFHPHQW 6XPPDU\ /)*9 $1$/<6,6 2) *H6L +%7 :,7+ 12 5()(5(1&( 6L %7 ,QWURGXFWLRQ 7KHRU\ ([SHULPHQWV DQG 5HVXOWV /)*9 $QDO\VLV ZLWK QR 6L 5HIHUHQFH %7 6XPPDU\ LLL
PAGE 4
6800$5< $1' &21&/86,216 6XPPDU\ *H6L +%7 0RGHOLQJ DQG /)*9 0HWKRGRORJ\ $33(1',; $ *(1(5$/ (48$7,216 )25 % 7 $33(1',; % 75$16,6725 5(/$7,21 )25 $ *$866,$1 '23$17 ,0385,7< &21&(175$7,21 352),/( ,1 7+( %$6( 2) %7 $33(1',; & /,67 2) 6<0%2/6 5()(5(1&(6 %,2*5$3+,&$/ 6.(7&+ ,9
PAGE 5
$EVWUDFW RI 'LVVHUWDWLRQ 3UHVHQWHG WR WKH *UDGXDWH 6FKRRO RI WKH 8QLYHUVLW\ RI )ORULGD LQ 3DUWLDO )XOILOOPHQW RI WKH 5HTXLUHPHQWV IRU WKH GHJUHH RI 'RFWRU RI 3KLORVRSK\ /2: )5(48(1&< &21'8&7$1&( 92/7$*( /)*9f &+$5$&7(5,67,&6 2) 6L*H[6L[6L +(7(5281&7,21 %,32/$5 75$16,67256 %\ *XR[LQ /L 'HFHPEHU &KDLUPDQ 'U $PRVW 1HXJURVFKHO &RFKDLUPDQ 'U &KLK7DQJ 6DK 0DMRU 'HSDUWPHQW (OHFWULFDO DQG &RPSXWHU (QJLQHHULQJ /RZ )UHTXHQF\ &RQGXFWDQFH 9ROWDJH /)*9f PHWKRGRORJ\ IRU DQDO\VLV RI KHWHURMXQFWLRQ ELSRODU WUDQVLVWRUV +%7Vf LV SUHVHQWHG 7KH /)*9 PHWKRGRORJ\ LV EDVHG RQ D FRPELQHG PHDVXUHPHQW RI WKH GF FROOHFWRU DQG EDVH FXUUHQWV DQG ORZ IUHTXHQF\ aON+]f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
PAGE 6
7KH /)*9 PHWKRGRORJ\ LV XVHG KHUH WR VWXG\ WKH SHUIRUPDQFH LPSURYHPHQWV RI 6L*H[6LMB[6L +%7 RYHU WKDW RI WKH 6L %7 LQFOXGLQJ FXUUHQW JDLQ VZLWFK VSHHG DQG (DUO\ YROWDJH ZKLFK DUH SULPDULO\ GXH WR WKH DLGLQJ HOHFWULF ILHOG $QDO\WLFDO PRGHOV IRU WKH HIIHFWLYH GHQVLW\RIVWDWHV 1&19 DQG LQWULQVLF FDUULHU FRQFHQWUDWLRQ QÂ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
PAGE 7
&+$37(5 ,1752'8&7,21 72 *H6L +(7(5281&7,21 %,32/$5 75$16,6725 ,QWURGXFWLRQ 7KH FRQFHSW RI KHWHURMXQFWLRQ ELSRODU MXQFWLRQ WUDQVLVWRU +%7f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
PAGE 8
7KH EUHDNWKURXJK IRU DQ 6LFRPSDWLEOH +%7 ZDV DFFRPSOLVKHG DJDLQ E\ ,%0 LQ >@ 7KH\ XVHG *HUPDQLXP 6LOLFRQ *H6Lf DV WKH EDVH PDWHULDO ZKLFK KDV D VPDOOHU HQHUJ\ JDS WKDQ WKDW RI WKH 6L HPLWWHU DQG FROOHFWRU $ WKLQ VWUDLQHG VLQJOH FU\VWDO *H[6LM[ ILOP RQ 6L VXEVWUDWH ZDV JURZQ SVHXGRPRUSKLFDOO\ XVLQJ XOWUDKLJK YDFXXPFKHPLFDO YDSRU GHSRVLWLRQORZ WHPSHUDWXUH HSLWD[\ 8+9&9'/7(f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f DQG H[FHOOHQW ORZ QRLVH EHKDYLRU 7KH EHVW KLJK IUHTXHQF\ SHUIRUPDQFH ZDV REWDLQHG DW ORZ YROWDJHV 9&( 9f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
PAGE 9
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f 7KH *H6L +%7 KDV YHU\ LPSRUWDQW DSSOLFDWLRQV LQ SRUWDEOH FRPPXQLFDWLRQ V\VWHPV VXFK DV PRELOH SKRQHV SDJHUV '$&V DQG RWKHUV *H6L +%7 )DEULFDWLRQ 6LOLFRQ DQG JHUPDQLXP DUH FRPSOHWHO\ PLVFLEOH LQ DQ\ PROH IUDFWLRQ UDWLR 7KH *H6L PDWHULDO KDV D GLDPRQG ODWWLFH VWUXFWXUH 7KH ODWWLFH FRQVWDQW FDQ EH FDOFXODWHG IURP WKH 9HJDUG UXOH >@ D*H[6Â[f D6Lf [D*Hf D6Lff f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
PAGE 10
SVHXGRPRUSKLF JURZWK ,Q SVHXGRPRUSKLF JURZWK LI WKH *H6L OD\HU WKLFNQHVV LV EHORZ D FULWLFDO WKLFNQHVV WKH ODWWLFH PLVPDWFK LV DFFRPPRGDWHG HODVWLFDOO\ ZLWK QR GLVORFDWLRQ FUHDWHG 7KH *H6L OD\HU LV XQGHU ELD[LDO VWUDLQ ERWK LQ WKH [\ SODQH DQG LQ WKH ] RU JURZWK GLUHFWLRQ 6HYHUDO HSLWD[LDO WHFKQLTXHV IRU *H6L OD\HU GHSRVLWLRQ ZHUH GHYHORSHG VXFK DV WKH PROHFXODU EHDP HSLWD[\ 0%(f >@ DQG OLPLWHG UHDFWLRQ SURFHVV /53&9'f >@ )LJXUH VKRZV D +%7 WUDQVLVWRU VWUXFWXUH XVHG LQ WKLV WKHVLV ZKLFK ZDV IDEULFDWHG E\ FKHPLFDO YDSRU GHSRVLWLRQ &9'f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f DV VKRZQ LQ )LJ 7KH LPSXULW\ DQG *H FRQFHQWUDWLRQ SURILOHV FDQ EH SUHFLVHO\ FRQWUROOHG GXULQJ ILOP JURZWK WR JLYH VSHFLILHG *H GLVWULEXWLRQ LQ WKH EDVH VXFK DV D OLQHDUO\ UHWURJUDGHG *H JUDGLHQW 7KLV *H JUDGLHQW SURGXFHV DQ DLGLQJ HOHFWULF ILHOG ZKLFK DFFHOHUDWHV WKH PLQRULW\ HOHFWURQV LQ WKH EDVH OD\HU
PAGE 11
(PLWWHU )LJ &URVV VHFWLRQ YLHZ RI *H6L +%7 $GDSWHG IURP 1HXJURVFKHO HW DO >@
PAGE 12
&RQFHQWUDWLRQ FPnf 'HSWK )LJ 6HFRQGDU\ LRQ PDVV VSHFWURVFRS\ 6,06f SURILOH RI *H6L +%7 7KH EDVH GRSLQJ OHYHO LV DERXW [O FPn DQG WKH *H PROH IUDFWLRQ LQ WKH EDVH LV DERXW b $GDSWHG IURP 5RQ 7DQJ HW DO >@ *HUPDQLXP bf
PAGE 13
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f DV VKRZQ LQ )LJ F 7KHUH LV D FULWLFDO OD\HU WKLFNQHVV EH\RQG ZKLFK WKH OD\HU ZLOO UHOD[ DQG FUHDWH WKH PLVILW GLVORFDWLRQV DW WKH *H6L6L LQWHUIDFH 7KH GLVORFDWLRQV PD\ DOVR SHQHWUDWH WKH HQWLUH OD\HU 7KLV FULWLFDO OD\HU WKLFNQHVV FDQ EH FDOFXODWHG EDVHG RQ WKH VWUDLQ RI WKH *H6L OD\HU DQG WKH WKHUPDO G\QDPLFV WKHRU\ >@ )LJXUH VKRZV WKH WKHRUHWLFDO DQG H[SHULPHQWDO FULWLFDO OD\HU WKLFNQHVV YHUVXV WKH *H PROH IUDFWLRQ >@ $V LV VKRZQ LQ WKH ILJXUH EDVHG RQ WKH PRGHO RI 3HRSOH DQG %HDQ IRU [*H b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f DUH GHIRUPHG ZKLFK JLYHV D VPDOOHU HOHFWURQ HIIHFWLYH PDVV PQ DQG DOVR D KLJKHU HOHFWURQ FRQFHQWUDWLRQ LQ WKH ] GLUHFWLRQ ZKLOH WKH HQHUJ\ RI
PAGE 14
WKH IRXU VXUIDFHV RQ WKH N[ DQG N\ D[HV LV ORZHUHG 7KH UHVXOWLQJ UHGXFWLRQ LQ WKH HQHUJ\ JDS DQG ORZHULQJ RI WKH HIIHFWLYH PDVV DUH YHU\ EHQHILFLDO IRU LPSURYHG SHUIRUPDQFH RI *H6L +%7 7KH VPDOOHU HQHUJ\ JDS LQFUHDVHV WKH FROOHFWRU FXUUHQW DQG JDLQ ZKLOH WKH VPDOOHU PQ LQ WKH YHUWLFDO PLQRULW\FDUULHU HOHFWURQ IORZ GLUHFWLRQ LQ QSQ +%7f LQFUHDVHV WKH PLQRULW\FDUULHU PRELOLW\ ZKLFK UHGXFHV WKH EDVH WUDQVLW WLPH )LJXUH LOOXVWUDWHV WKH LQSODQH [f RI JURZWK GLUHFWLRQ DQG QRUPDO WR WKH JURZWK SODQH RU WKH JURZWK ]f GLUHFWLRQ )LJXUH VKRZV WKH FDOFXODWHG YDOXHV RI HQHUJ\ JDS UHGXFWLRQ RI VWUDLQHG *H6L OD\HU DQG XQVWUDLQHG EXONf PDWHULDO YHUVXV WKH *H PROH IUDFWLRQ ,W VKRZV WKDW IRU WKH VWUDLQHG *H6L OD\HU WKH HQHUJ\ JDS UHGXFWLRQ $(*B*H LV DOPRVW D OLQHDU IXQFWLRQ RI WKH LQFUHDVLQJ *H PROH IUDFWLRQ [*H IRU [*Hab *H6L +%7 &KDUDFWHULVWLFV 7\SLFDO FROOHFWRU FXUUHQW DQG EDVH FXUUHQW GHSHQGHQFLHV RQ 9%( RI *H6L +%7 DQG 6L %7 DUH VKRZQ LQ )LJ 7KH FROOHFWRU FXUUHQW RI *H6L +%7 ,F*H6Lf VKRZV ODUJH LPSURYHPHQW RYHU WKDW LQ 6L %7V 7KH EDVH FXUUHQW RI *H6L +%7 ,%*H6Lf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
PAGE 15
8QVWUDLQHG *H[68[ 6WUDLQHG *H[6LA 6L 6L Df 5HOD[HG *H[68[ W Q ? 6L )LJ 7ZR GLPHQVLRQDO LOOXVWUDWLRQ RI *H6L SVHXGRPRUSKLF JURZWK Df 8QVWUDLQHG *H6L OD\HU DQG 6L VXEVWUDWH Ef 6WUDLQHG *H6L OD\HU 7KH ODWWLFH PLVPDWFK LV DFFRPPRGDWHG E\ WKH VWUDLQ LQ WKH *H6L OD\HU Ff 5HOD[HG *H6L OD\HU 7KHUH DUH PLVILW GLVORFDWLRQV DW WKH *H6L6L LQWHUIDFH $GDSWHG IURP 6DK > )LJ S @
PAGE 16
&ULWLFDO /D\HU 7KLFNQHVV )LJ &ULWLFDO OD\HU WKLFNQHVV RI *H6L OD\HU YHUVXV *H IUDFWLRQ 7KH GDVKHG OLQH VKRZV WKH WKHRUHWLFDO UHVXOWV FDOFXODWHG IURP WKH WKHUPDO G\QDPLFV 7KH VROLG OLQH JLYHV WKH H[SHULPHQWDO UHVXOWV $GDSWHG IURP 6DK > S @
PAGE 17
)LJ &RQVWDQW HQHUJ\ VXUIDFHV LQ WKH FRQGXFWLRQ EDQG LQ *H[6ÂB[ XQGHU ELD[LDO VWUDLQ > @ 7KH ] GLUHFWLRQ FRUUHVSRQGV WR WKH JURZWK GLUHFWLRQ 7KH WZR VXUIDFHV LQ WKH ]GLUHFWLRQ JURZWK GLUHFWLRQf DUH UDLVHG LQ HQHUJ\ DQG GHIRUPHG VXFK WKDW WKH\ JLYH D VPDOOHU HOHFWURQ HIIHFWLYH PDVV PQ 7KH IRXU [ DQG \ VXUIDFHV DUH ORZHUHG LQ HQHUJ\ 7KH RYHUDOO HQHUJ\ JDS LV VPDOOHU DQG WKH HOHFWURQV SUHIHUHQWLDOO\ SRSXODWH WKH WZR YDOOH\V LQ WKH ]GLUHFWLRQ LQ ZKLFK PQ LV ORZ 7KLV LV DOVR WKH GLUHFWLRQ RI WKH PLQRULW\FDUULHU IORZ LQ )(%7
PAGE 18
]r ; (PLWWHU I IA]] 3*H6L EDVH A0[[ &ROOHFWRU )LJ 'LUHFWLRQ LOOXVWUDWLRQ LQ YHUWLFDO *H[6LMB[ +%7 7KH ] GLUHFWLRQ FRUUHVSRQGV WR WKH JURZWK GLUHFWLRQ DQG LV DOVR WKH GLUHFWLRQ RI WKH PLQRULW\ HOHFWURQ IORZ LQ D YHUWLFDO %7 7KH LQSODQH GLUHFWLRQ LV LPSRUWDQW DQG DSSOLFDEOH IRU ODWHUDO 026 WUDQVLVWRUV
PAGE 19
(QHUJ\ *DS H9f )LJ (QHUJ\ JDS UHGXFWLRQ IRU VWUDLQHG DQG XQVWUDLQHG *H6L OD\HU YHUVXV *H PROH IUDFWLRQ LQ WKH OD\HU $GDSWHG IURP 6DK > S @
PAGE 20
9%(9f )LJ 7\SLFDO ,9 FKDUDFWHULVWLFV RI *H6L +%7 DQG 6L %7 DW 9&% 9 7KH *H6L +%7 DQG 6L %7 KDYH WKH VDPH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH DQG JHRPHWU\ H[FHSW WKH *H LQ WKH EDVH IRU *H6L +%7 7KH VROLG OLQH LV *H6L +%7 DQG GDVKHG OLQH LV 6L %7
PAGE 21
&XWRII )UHTXHQF\ I7 *+]f &ROOHFWRU &XUUHQW OF P$f )LJ 7\SLFDO XQLW JDLQ IUHTXHQF\ I7 YHUVXV ,F FKDUDFWHULVWLFV RI *H6L +%7 IXOO FLUFOHVf DQG 6L %7 RSHQ FLUFOHVf 7KH *H6L +%7 DQG 6L %7 KDYH WKH VDPH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH DQG JHRPHWU\ H[FHSW WKH *H LQ WKH EDVH IRU *H6L +%7 $GDSWHG IURP 0HULW +RQJ HW DO >@
PAGE 22
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f LV LQ WKH JURZWK RU ] GLUHFWLRQ WKH WUDQVYHUVH PRELOLW\ GRZQZDUG WULDQJOHVf LV LQ WKH [\ SODQH )LJXUH DQG )LJ VKRZV RQO\ D VPDOO abf LQFUHDVH LQ WKH ORQJLWXGLQDO PRELOLW\ ZLWK *H PROH IUDFWLRQ DW +RZHYHU D PXFK ODUJHU LPSURYHPHQW bf ZDV FDOFXODWHG DW IRU [*Hab 7KH WUDQVYHUVH PRELOLW\ VKRZV D ODUJH UHGXFWLRQ ZLWK LQFUHDVLQJ [*H ZKLFK LV DOZD\V VPDOOHU WKDQ WKDW LQ 6L [*H f DW DQ\ *H FRQFHQWUDWLRQ OHYHO 7KH UHVXOWV IRU WKH PDMRULW\ DQG PLQRULW\ KROH PRELOLWLHV DUH VLPLODU WR WKRVH IRU WKH HOHFWURQ PRELOLWLHV LQ )LJ DQG )LJ ,Q VXPPDU\ WKH H[SHFWHG PLQRULW\ HOHFWURQ PRELOLW\ LPSURYHPHQW IRU WKH GRSLQJ OHYHO RI DERXW FPn LV YHU\ VPDOO _[Q*H6LfSQ6Lf 7KLV DJUHHV YHU\ ZHOO ZLWK RXU H[SHULPHQWDO UHVXOWV DV ZLOO EH VKRZQ LQ FKDSWHU
PAGE 23
0LQRULW\ HOHFWURQ PRELOLW\ FP9r6f *H PROH IUDFWLRQ bf )LJ 0LQRULW\ HOHFWURQ PRELOLW\ SQ LQ SW\SH *H[6LLB[ OD\HU 7KH XSZDUG WULDQJOHV DUH LQ WKH SHUSHQGLFXODU ] GLUHFWLRQ ZKLFK LV WKH JURZWK GLUHFWLRQ DQG DOVR WKH HOHFWURQ PRYHPHQW GLUHFWLRQ LQ YHUWLFDO %7 7KH GRZQZDUG WULDQJOHV DUH LQ WKH [\ SODQH $GDSWHG IURP /( .D\ DQG 7: 7DQJ >@
PAGE 24
0DMRULW\ HOHFWURQ PRELOLW\ FP9r6f )LJ 0DMRULW\ HOHFWURQ PRELOLW\ SQ LQ QW\SH *HALA OD\HU 7KH XSZDUG WULDQJOHV DUH LQ WKH SHUSHQGLFXODU ] GLUHFWLRQ ZKLFK LV DOVR WKH JURZWK GLUHFWLRQ DQG WKH GRZQZDUG WULDQJOHV DUH LQ WKH [\ SODQH $GDSWHG IURP /( .D\ DQG 7: 7DQJ >@
PAGE 25
(QHUJ\ %DQG 'LDJUDP ,Q 6L*HALA6L +%7 D *HALA EDVH OD\HU ZLWK D VPDOOHU HQHUJ\ JDS WKDQ WKDW RI SXUH 6L LV VDQGZLFKHG EHWZHHQ WKH 6L HPLWWHU DQG FROOHFWRU 7KH HQHUJ\ EDQG GLDJUDP LV YHU\ LPSRUWDQW IRU DQDO\VLV DQG GHVLJQ RSWLPL]DWLRQ RI WKH *H6L +%7 )LJXUH )LJ )LJ DQG )LJ VKRZ WKH HQHUJ\ EDQG GLDJUDPV IRU WULDQJXODU *H GLVWULEXWLRQ DQG WUDSH]RLGDO *H GLVWULEXWLRQ LQ WKH EDVH UHJLRQ DW HTXLOLEULXP ]HUR ELDVf DQG IRUZDUG DFWLYH PRGH 9%(!9 9&%9f $OVR IURP )LJ ZH FDQ VHH WKDW IRU *H PROH IUDFWLRQ [*Hab WKH HQHUJ\ JDS UHGXFWLRQ $(*B*H LV DOPRVW D OLQHDU IXQFWLRQ RI [*H )RU WUDSH]RLGDO *H GLVWULEXWLRQ ZH KDYH [*H[f [*Hf >[*H;5f [*Hf@}[;5 [;5f f (J[f (JR $(JBKG $(*B*Hf $(*B*Hm[;5 [;5f f ZKHUH $(*B*H $(*B*H[5f $(*B*Hf f LV WKH WRWDO *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ WKURXJK WKH EDVH UHJLRQ $(*B+' LV WKH HQHUJ\ JDS UHGXFWLRQ GXH WR KHDY\ GRSLQJ DQG $(**Hf LV WKH HQHUJ\ JDS UHGXFWLRQ GXH WR *H DW [ LQ WKH TXDVLQHXWUDO EDVH UHJLRQ 7KH *H JUDGLHQW ZLOO SURGXFH DQ DLGLQJ HOHFWULF ILHOG (EL (EL OTfG(FG[f OTf$(*B*H;5f LQGHSHQGHQW RI [ N7TfWM[5f [;5 f ZKHUH Uf $(* *HAN7n
PAGE 26
$(4*H2f $(*B*H;%f 3*H6L %DVH Q6L (PLWWHU Y 1 *Hf $$ ; % Q6L &ROOHFWRU )LJ (QHUJ\ EDQG GLDJUDP RI *H6L +%7 GDVKHG OLQHf DQG 6L %7 VROLG OLQHf ZLWK D WULDQJXODU *H GLVWULEXWLRQ LQ WKH EDVH UHJLRQ DW HTXLOLEULXP ]HUR ELDVf () LV WKH )HUPL OHYHO DW HTXLOLEULXP $(*B*Hf LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(**H;%f LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ ;% LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(**H $(*B*H;%f $(**Hf LV WKH WRWDO *H LQGXFHG HQHUJ\ JDS UHGXFWLRQ WKURXJK WKH EDVH UHJLRQ
PAGE 27
$(R*H2f $(*B*H;%f )LJ (QHUJ\ EDQG GLDJUDP RI *H6L +%7 GDVKHG OLQHf DQG 6L %7 VROLG OLQHf ZLWK D WULDQJXODU *H GLVWULEXWLRQ LQ WKH EDVH UHJLRQ XQGHU IRUZDUG DFWLYH RSHUDWLRQ 9%(!9 9&E!9 )Q LV WKH TXDVL)HUPL OHYHO IRU HOHFWURQV DQG )S LV WKH TXDVL)HUPL OHYHO IRU KROHV 9EL LV WKH EXLOWLQ SRWHQWLDO RI WKH SQ MXQFWLRQ $(*B*Hf LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(*B*H;%f LV WKH *H LQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ ;% LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(*B*H $(**H;%f $(**H2f LV WRWDO *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ WKURXJK WKH EDVH UHJLRQ
PAGE 28
(F (Y $(*B*Hf $(*B*H;%f 3*H6L %DVH ;5 ;% ( ) )LJ (QHUJ\ EDQG GLDJUDP RI *H6L +%7 GDVKHG OLQHf DQG 6L %7 VROLG OLQHf ZLWK D WUDSH]RLGDO *H GLVWULEXWLRQ LQ WKH EDVH UHJLRQ DW HTXLOLEULXP ]HUR ELDVf () LV WKH )HUPL OHYHO DW HTXLOLEULXP $(**Hf LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(*B*H;%f LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ ;% LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(*B*H $(*B*H;%f $(*B*Hf LV WKH WRWDO *H LQGXFHG HQHUJ\ JDS UHGXFWLRQ WKURXJK WKH EDVH UHJLRQ
PAGE 29
$(*B4Hf $(*B*H;%f )LJ (QHUJ\ EDQG GLDJUDP RI *H6L +%7 GDVKHG OLQHf DQG 6L %7 VROLG OLQHf ZLWK D WUDSH]RLGDO *H GLVWULEXWLRQ LQ WKH EDVH UHJLRQ XQGHU IRUZDUG DFWLYH RSHUDWLRQ Y%(!RY 9&%!9 )Q LV WKH TXDVL)HUPL OHYHO IRU HOHFWURQV DQG )3 LV WKH TXDVL)HUPL OHYHO IRU KROHV 9EL LV WKH EXLOWLQ SRWHQWLDO RI WKH SQ MXQFWLRQ $(*B*Hf LV WKH *H LQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(**H;%f LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ ;% LQ WKH TXDVLQHXWUDO EDVH UHJLRQ $(*B*H $(*B*H;%f$(*B*Hf LV WKH WRWDO *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ WKURXJK WKH EDVH UHJLRQ
PAGE 30
1&19 DQG Q" 0RGHO IRU /LQHDU *H 'LVWULEXWLRQ 7KH *H LQ WKH EDVH UHJLRQ FKDQJHV QRW RQO\ WKH HQHUJ\ JDS EXW DOVR RWKHU HOHFWURQLF PDWHULDO SDUDPHWHUV 7KH HIIHFWLYH GHQVLW\ RI VWDWHV LQ WKH FRQGXFWLRQ EDQG DQG YDOHQFH EDQG 1F DQG 1Y ZLOO FKDQJH EHFDXVH WKH\ DUH D IXQFWLRQ RI WKH HOHFWURQ DQG KROH HIIHFWLYH PDVVHV PH DQG PK UHVSHFWLYHO\ ZKLFK FKDQJH ZLWK WKH *HLQGXFHG VWUDLQ DQG FKDQJH RI WKH (N UHODWLRQVKLS LQ WKH HSLWD[LDO *H6L OD\HU >@ 1F DQG 1Y DUH JLYHQ E\ >@ 1F MUPHN7Kf Df 1Y UPKN7Kf Ef )LJXUH VKRZV WKH WKHRUHWLFDO FDOFXODWLRQV RI WKH QRUPDOL]HG GHQVLW\ RI VWDWHV SURGXFW 1&19 YHUVXV *H PROH IUDFWLRQ >@ ,Q RUGHU WR REWDLQ DQDO\WLFDO PRGHOV IRU WKH *H6L +%7 ZH DSSUR[LPDWH WKH 1&19 SURGXFW E\ DQ H[SRQHQWLDO IXQFWLRQ OHDVW VTXDUHV ILWWHG WR WKH WKHRUHWLFDO GDWD 1F[f1Y[f 1Ff1YfH[S&[;5f f ZKHUH WKH FRQVWDQW & LV D QRQOLQHDU OHDVW VTXDUHV ILW 1/6)f SDUDPHWHU 7KH PRGHO ILWV WKH WKHRU\ ZHOO IRU [*H b ZLWK & s LQ SDUWLFXODU WR WKH WKHRUHWLFDO PRGHOV RI 3ULQ] HW DO >@ DQG DLQ HW DO >@ ZKLFK DUH DOVR VXSSRUWHG E\ WKH SKRWROXPLQHVFHQFH GDWD DQG HOHFWULFDO PHDVXUHPHQWV LQ 6L*HALA6L +%7V 7KH 1F[f1Y[f PRGHO JLYHQ E\ f HQDEOHV WKH GHFRXSOLQJ RI WKH GLIIXVLYLW\ '% IURP WKH 1&19 SURGXFW DV ZLOO EH VKRZQ ODWHU 7KH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ $(*B*H LQ WKH EDVH UHJLRQ ZLOO FKDQJH WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ Qc ZKLFK FDQ EH H[SUHVVHG DV >@
PAGE 31
1F1Y*H6cf1F1Y6Lf )LJ 7KH WKHRUHWLFDO FDOFXODWLRQV RI WKH QRUPDOL]HG GHQVLW\ RI VWDWHV SURGXFW 1&19 YHUVXV *H PROH IUDFWLRQ >@ 7KH VROLG OLQHV DUH IURP FDOFXODWLRQV LQ >@ 7KH GDVKHG OLQHV DUH WKH 1/6)V WR f IRU & DQG &
PAGE 32
QI [f 1F[f1Y[fH[S(*[fN7f f )RU D OLQHDU *H GLVWULEXWLRQ XVLQJ UHVXOWV IURP VHFWLRQ f ZH KDYH (J[f (* $(*B+' $(*B*Hf $(*B*H[;5 f ZKHUH $(**H $(*B*H;5f $(**Hf LV WKH WRWDO HQHUJ\ JDS UHGXFWLRQ DFURVV WKH TXDVLQHXWUDO EDVH UHJLRQ $(*+' LV WKH HQHUJ\ JDS UHGXFWLRQ GXH WR KHDY\ GRSLQJ HIIHFW DQG $(*B*Hf LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [ LQ WKH TXDVLQHXWUDO EDVH UHJLRQ ,QVHUWLQJ HTXDWLRQV f DQG f LQWR HTXDWLRQ f JLYHV DQ DQDO\WLFDO H[SUHVVLRQ IRU WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ Qc IRU D OLQHDU *H GLVWULEXWLRQ QI[f QL4 f H[S$(*B*HfN7fH[S$(*B+'N7f f H[S[[A f ZKHUH QL LV WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ RI 6L QLR 1Ff1YfH[S(*2N7f f 7KH FRQVWDQW ;M LV GHILQHG DV O[ O;%f>$(**HN7f&@ )RU D VLPSOLFLW\ ZH GHILQH D $(*B*HfN7 DQG \ $(*+'N7 WKHQ HTXDWLRQ f EHFRPHV Q[f QL H[SDH[S\H[SO[[A f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ab
PAGE 33
WKH HQHUJ\ JDS LQ VWUDLQHG *H6L OD\HU GHFUHDVHV OLQHDUO\ ZLWK [*H 0RQWH &DUOR VLPXODWLRQ UHVXOWV IRU PLQRULW\ DQG PDMRULW\ HOHFWURQ DQG KROH PRELOLWLHV ZHUH DOVR GLVFXVVHG $QDO\WLFDO DSSUR[LPDWLRQV IRU WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ Qc DQG GHQVLW\ RI VWDWHV GHSHQGHQFH RQ WKH *H PROH IUDFWLRQ ZHUH GHYHORSHG KHUH LQ WKLV WKHVLV IRU DQ LGHDOL]HG OLQHDUO\JUDGHG *H FRQFHQWUDWLRQ SURILOH LQ WKH *H6L EDVH OD\HU 0HDVXUHG ,%9%( DQG ,&9%( FXUYHV DQG KLJK IUHTXHQF\ UHVSRQVHV RI *H6L +%7 DQG 6L UHIHUHQFH %7 KDYH GHPRQVWUDWHG WKH SHUIRUPDQFH LPSURYHPHQWV LQ *H6L +%7 FRPSDUHG WR 6L %7
PAGE 34
&+$37(5 /2: )5(48(1&< &21'8&7$1&( 92/7$*( /)*9f 0(7+2'2/2*< ,QWURGXFWLRQ ,Q WKH ODVW FKDSWHU ZH SUHVHQWHG D SUHOLPLQDU\ TXDOLWDWLYH DQDO\VLV RI WKH *H6L +%7 DQG LWV FRPSDULVRQ ZLWK WKH 6L %7 ,Q WKLV FKDSWHU ZH GHVFULEH D ORZIUHTXHQF\ FRQGXFWDQFHYROWDJH /)*9f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
PAGE 35
LPSURYHPHQWV LQ FROOHFWRU FXUUHQW WUDQVLW WLPH DQG RXWSXW FRQGXFWDQFH LQ *H6L WUDQVLVWRUV RYHU WKRVH LQ VLPLODU 6L WUDQVLVWRUV ZHUH GHPRQVWUDWHG TXDOLWDWLYHO\ DQG JUDSKLFDOO\ ,Q WKLV FKDSWHU ZH SUHVHQW GHWDLOHG DQDO\VLV EDVHG RQ DQ DQDO\WLFDO PRGHO WR H[WUDFW UHOLDEOH DQG DFFXUDWH YDOXHV RI WKH PLQRULW\FDUULHU SDUDPHWHUV LQ WKH *H6L PDWHULDO 7KHVH SDUDPHWHUV DUH HVVHQWLDO WR DVVHVV WKH *HLQGXFHG HIIHFWV RQ WUDQVLVWRU SHUIRUPDQFH GHVLJQ RSWLPL]DWLRQ DQG WR YHULI\ WKHRUHWLFDO SUHGLFWLRQV DERXW WKH H[SHFWHG SHUIRUPDQFH EHQHILWV GXH WR *H 7KH NH\ SRLQW LQ REWDLQLQJ WKHVH TXDQWLWDWLYH YDOXHV LV D VHSDUDWLRQ RI WKH YDULRXV *H LQGXFHG HIIHFWV IURP WKH PHDVXUHG GDWD $ ODFN RI DQ H[SHULPHQWDO PHWKRGRORJ\ IRU WKLV DQDO\VLV LV WKH PDLQ UHDVRQ ZK\ WKHVH YDOXHV DUH VWLOO QRW DYDLODEOH LQ WKH OLWHUDWXUH )RU H[DPSOH LQ *H6L WUDQVLVWRUV WKH LQFUHDVH LQ WKH FROOHFWRU FXUUHQW ,F FDQ DULVH IURP WKH DLGLQJ EXLOGLQ HOHFWULF ILHOG LQ WKH EDVH GXH WR WKH *H UHWURJUDGHG SURILOH WKH EDVH PLQRULW\FDUULHU GLIIXVLYLW\ HQKDQFHPHQW RU ERWK $Q DFFXUDWH VHSDUDWLRQ RI WKH SHUIRUPDQFH LPSURYHPHQW EHWZHHQ WKH WZR SDUDPHWHUV UHTXLUHV LQGHSHQGHQW PHDVXUHPHQWV RI WKH EXLOWLQ ILHOG DQG PRELOLW\ @ 7KLV JLYHV 1F1Yf*H6L1F1Yf6L a IRU *H6L OD\HU ZLWK [*H b >@ 7KH PLQRULW\FDUULHU PRELOLW\ ZDV SUHGLFWHG WR LQFUHDVH ZLWK *H FRQFHQWUDWLRQ IRU [*Hab HVSHFLDOO\ DW KLJKHU EDVH LPSXULW\ FRQFHQWUDWLRQ
PAGE 36
FPf DQG DW ORZ WHPSHUDWXUHV a.f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f /RZIUHTXHQF\FRQGXFWDQFHYROWDJH /)*9f PHWKRGRORJ\ LV EDVHG RQ WKH FRPELQDWLRQ RI WKH GF ,FB9%( DQG AEnAEH PHDVXUHPHQWV ZLWK WKH ORZIUHTXHQF\ aON+]f VPDOOVLJQDO FRQGXFWDQFH PHDVXUHPHQWV UHVXOWLQJ IURP WKH EDVH WKLFNQHVV PRGXODWLRQ >@ 7KH EDVH WKLFNQHVV PRGXODWLRQ %70f HIIHFW LV LOOXVWUDWHG LQ )LJ IRU D 6L %7 7KH FROOHFWRU FXUUHQW ,F LV SURSRUWLRQDO WR WKH JUDGLHQW RI WKH EDVH PLQRULW\ FDUULHU GLVWULEXWLRQ LQ WKH EDVH OD\HU DW D FHUWDLQ IRUZDUG ELDV HPLWWHUEDVH YROWDJH 9%( DQG UHYHUVH ELDV FROOHFWRUEDVH YROWDJH 9&% :KHQ D VPDOOVLJQDO YROWDJH YFE LV VXSHULPSRVHG XSRQ WKH GF EDVHFROOHFWRU MXQFWLRQ YROWDJH 9&% WKH TXDVLQHXWUDO EDVH WKLFNQHVV ;% ZLOO EH PRGXODWHG E\ $;% DV VKRZQ LQ )LJ 7KH VKDGRZHG DUHD LQ )LJ Ef VKRZV WKH FKDQJH LQ WKH PLQRULW\ FDUULHU GLVWULEXWLRQ LQ WKH TXDVLQHXWUDO EDVH GXH WR WKH EDVH WKLFNQHVV PRGXODWLRQ DQG WKH EDVH PLQRULW\ FDUULHU FKDUJH 4%
PAGE 37
$;% Ef )LJ Df &URVVVHFWLRQDO YLHZ RI DQ QSQ %7 6PDOOVLJQDO YROWDJH YFE LV VXSHULPSRVHG RQ GF YDOXH 9&% Ef 0LQRULW\FDUULHU GLVWULEXWLRQ LQ WKH EDVH
PAGE 38
FKDQJH E\ $4% ZLWK WKH VPDOOVLJQDO YFE 7KH PLQRULW\ FDUULHU FRQFHQWUDWLRQ DW WKH HPLWWHU ERXQGDU\ RI WKH TXDVLQHXWUDO EDVH DW [ 1f UHPDLQV FRQVWDQW IRU D IL[HG 9%( DQG LV JLYHQ E\ >QA3f@}H[ST9%(N7f ZKHUH 3f 1$$f LV WKH KROHHPLWWHU ERXQGDU\ RI WKH TXDVLQHXWUDO EDVH DW [ 1f UHPDLQV FRQVWDQW IRU D IL[HG 9%( DQG LV JLYHQ E\ >Q"3f@mH[ST9%(N7f ZKHUH 3f 1$$f LV WKH KROH FRQFHQWUDWLRQ DW [ N LV WKH %ROW]PDQQ FRQVWDQW DQG 7 LV WKH WHPSHUDWXUH 7KH FKDQJH LQ WKH VORSH RI 1[f JLYHV WKH VPDOOVLJQDO FROOHFWRU FXUUHQW LF 7KH RXWSXW FRQGXFWDQFH J LV GHILQHG E\ R nFE f %( 7KH EDVH FXUUHQW ,% DOVR FKDQJHV GXH WR D FKDQJH LQ WKH EDVH UHFRPELQDWLRQ FXUUHQW 4EWE WKH EDVH WKLFNQHVV PRGXODWLRQ E\ WKH VPDOOVLJQDO YROWDJH YFE 7KLV SURGXFHV WKH UHYHUVH WUDQVFRQGXFWDQFH JU ZKLFK LV GHILQHG E\ U nFE f %( ZKHUH LE LV WKH VPDOOVLJQDO EDVH FXUUHQW GXH WR YFE $QDO\WLFDO H[SUHVVLRQV IRU J DQG JU FDQ EH GHULYHG VHH DSSHQGL[ $f ,& G;% AEH R nFE A%( ;% G9 U nFE &% G;5 f A%( W%r;% GA&% f n%( ZKHUH ;% LV WKH TXDVLQHXWUDO EDVH WKLFNQHVV 4% LV WKH EDVH PLQRULW\ FDUULHU FKDUJH FRUUHVSRQGLQJ WR 9%( DQG 9&% DQG [% LV WKH EDVH PLQRULW\ FDUULHU UHFRPELQDWLRQ OLIHWLPH 6LQFH ,F DQG 4% DUH H[SRQHQWLDO IXQFWLRQV RI WKH IRUZDUG EDVHHPLWWHU YROWDJH
PAGE 39
9%( J DQG JU DUH DOVR H[SRQHQWLDO IXQFWLRQV RI 9%( 7DNLQJ WKH UDWLR RI HTXDWLRQV f DQG f ZH REWDLQ D YHU\ LPSRUWDQW UHODWLRQ IRU WKH DQDO\VLV RI WKH ELSRODU MXQFWLRQ WUDQVLVWRU > S @ ,F 4EAE WE fÂ§ f U 4EA W E 4EA WE W% ZKHUH W% LV WKH EDVH PLQRULW\ FDUULHU WUDQVLW WLPH ,Q RUGHU WR DFFXUDWHO\ PHDVXUH WKH RXWSXW FRQGXFWDQFH J DQG HVSHFLDOO\ WKH UHYHUVH WUDQVFRQGXFWDQFH JU JUmJf D SUHFLVH HTXLSPHQW LV QHHGHG 7ZR LQVWUXPHQWV ZHUH XVHG WR PHDVXUH J DQG JU D PDQXDO :D\QH .HUU EULGJH % DQG DQ DXWRPDWHG ,((( EXV FRQWUROOHG +3$ /&5 PHWHU 0RVW RI WKH GDWD ZHUH REWDLQHG XVLQJ :. % EHFDXVH RI LWV KLJK FRQGXFWDQFH VHQVLWLYLW\ a Oef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
PAGE 40
EDVH OD\HU 7KH VHFRQG PHWKRG XVLQJ RQO\ WKH *H6L +%7 ZLWKRXW WKH UHIHUHQFH 6L %7 ZLOO EH GLVFXVVHG LQ FKDSWHU 7KHUH DUH VHYHUDO SHUIRUPDQFH HQKDQFHPHQW JDXJHV RI *H6L +%7 RYHU 6L %7 7KH PRVW LPSRUWDQW DQG DOVR PRVW FRPPRQO\ XVHG E\ LQYHVWLJDWRUV LV WKH GF FROOHFWRU FXUUHQW ,F HQKDQFHPHQW >@ )RU H[DPSOH IRU D XQLIRUP EDVH GRSLQJ DQG XQLIRUP *H FRQFHQWUDWLRQ LQ WKH SW\SH EDVH RI QSQ WUDQVLVWRU WKH UDWLR RI WKH FROOHFWRU FXUUHQWV FDQ EH ZULWWHQ DV ,F*H6Lf 1F1Yf*H6L 'Q*H6Lf f f H[S$\f f H[S$(*B*HN7f f ,&6Lf 1&19f6 'Q6Lf ZKHUH 1F 1Y DUH WKH GHQVLW\ RI VWDWHV RI WKH FRQGXFWLRQ EDQG DQG YDOHQFH EDQG UHVSHFWLYHO\ 'Q LV WKH PLQRULW\FDUULHU HOHFWURQ GLIIXVLYLW\ $(*B*H LV WKH *HLQGXFHG HQHUJ\JDS UHGXFWLRQ LQ WKH EDVH DQG $\ $(*B+'N7 >$(*B+'*H6Lf $(*+R6Lf@N7 LV WKH QRUPDOL]HG GLIIHUHQFH EHWZHHQ WKH KHDY\ GRSLQJLQGXFHG HQHUJ\ JDS UHGXFWLRQ LQ *H6L DQG 6L EDVH OD\HUV 7KHUH DUH IRXU XQNQRZQV LQ f DOO UHODWHG WR WKH *H LQ WKH *H6L EDVH OD\HU 7KH NH\ LVVXH LV KRZ WR VHSDUDWH WKHVH *HLQGXFHG HIIHFWV 7ZR DSSURDFKHV KDYH EHHQ SUHVHQWHG LQ WKH OLWHUDWXUH XVLQJ WKH ,F UDWLR LQ f >@ 7KH ILUVW DSSURDFK XVHV WKH PHDVXUHPHQW RI WKH WHPSHUDWXUH GHSHQGHQFH RI ,F*H6Lf,F6Lf DVVXPLQJ WKDW WKH GHQVLW\ RI VWDWHV DQG PRELOLW\ UDWLRV DUH WHPSHUDWXUH LQGHSHQGHQW 7KHQ E\ SORWWLQJ WKH ,F UDWLR YHUVXV 7 HTXDWLRQ f \LHOGV DQ DFWLYDWLRQ HQHUJ\ ($ $(*B*H $(*B+' $(**H 7KLV $(&&Â FDQ WKHQ EH XVHG WR FDOFXODWH WKH LPSURYHPHQWV LQ WKH FXUUHQW JDLQ WUDQVLW WLPH DQG RWKHU FKDUDFWHULVWLFV DV D IXQFWLRQ RI WKH *H PROH IUDFWLRQ DQG
PAGE 41
WHPSHUDWXUH %XW WKHRUHWLFDO FDOFXODWLRQV VKRZ WKDW WKH GHQVLW\ RI VWDWHV DQG PRELOLW\ KDYH YHU\ FRPSOLFDWHG WHPSHUDWXUH GHSHQGHQFH VR WKH DVVXPSWLRQ RI WKH WHPSHUDWXUH LQGHSHQGHQFH RI WKH WZR SUHH[SRQHQWLDO IDFWRUV LQ f LV XQUHOLDEOH DQG FDQ OHDG WR ODUJH HUURUV LQ $(**H DV DOVR QRWHG SUHYLRXVO\ >@ 7KH VHFRQG DSSURDFK LV WR DVVXPH WKHRUHWLFDOO\ SUHGLFWHG YDOXHV RI 1F1Yf*H6L DQG 'Q*H6Lf DQG WKHQ FDOFXODWH ($ IURP f $SSDUHQWO\ WKHUH LV DQRWKHU XQNQRZQ $\ DQG DOVR WKH WKHRUHWLFDO UHVXOWV DUH VWURQJO\ GHSHQGHQW RQ PRGHOV DQG LQSXW SDUDPHWHUV )XUWKHUPRUH ERWK DSSURDFKHV JLYH DQ HVWLPDWH IRU WKH HQHUJ\ JDS UHGXFWLRQ WHUP RQO\ DQG GR QRW SURYLGH YDOXHV RU HVWLPDWHV IRU WKH PRELOLW\ RU UHFRPELQDWLRQ OLIHWLPH LQ WKH *H6L EDVH 7KH DERYH GLVFXVVLRQ VKRZV WKDW DQ DQDO\VLV EDVHG RQ WKH PHDVXUHPHQW RI ,F*H6Lf RQO\ LV QRW VXIILFLHQW 7KH /)*9 PHWKRGRORJ\ KRZHYHU ZLOO XVH D FRPELQDWLRQ RI WKH GF ,%9%( DQG ,*9%( FKDUDFWHULVWLFV ZLWK WKH VPDOOVLJQDO RXWSXW FRQGXFWDQFH J DQG UHYHUVH WUDQVFRQGXFWDQFH JU PHDVXUHPHQWV 7KLV QRYHO PHWKRGRORJ\ FDQ RYHUFRPH WKH OLPLWDWLRQV RI WKH DERYH DQG RWKHU DSSURDFKHV 7KH /)*9 PHWKRGRORJ\ ZLOO EH SUHVHQWHG KHUH IRU WKH LGHDOL]HG WUDSH]RLGDO *H FRQFHQWUDWLRQ SURILOH LQ WKH EDVH ZKLFK LV VKRZQ LQ )LJ ,Q WKLV PRGHO WKH EDVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH DQG WKH *H GLVWULEXWLRQ DUH GLYLGHG LQWR WZR UHJLRQV ,Q WKH ILUVW UHJLRQ [;5 WKH EDVH GRSLQJ LV DVVXPHG WR EH FRQVWDQW DQG WKH *H GLVWULEXWLRQ LV DVVXPHG WR EH UHWURJUDGHG OLQHDU ,Q WKH VHFRQG UHJLRQ ;5[;% WKH EDVH GRSLQJ LV DSSUR[LPDWHG E\ DQ H[SRQHQWLDO IXQFWLRQ DQG WKH *H GLVWULEXWLRQ LV DVVXPHG WR EH FRQVWDQW ,Q WKH UHJLRQ ZLWK WKH OLQHDU *H GLVWULEXWLRQ [;5f (*[f (Jf $(*B*H[;5f >@ ZKHUH (*f LV WKH HQHUJ\ JDS DW [ DQG $(**H $(*;5f
PAGE 42
&RQFHQWUDWLRQ FP f 'HSWK )LJ ,GHDOL]HG PRGHO IRU WKH *H DQG GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOHV LQ *H6L +%7 7KH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOHV IRU 6L %7 QR *H LQ WKH EDVHf DUH WKH VDPH DV WKRVH LQ *H6L +%7 *HUPDQLXP bf
PAGE 43
$(Jf LV WKH WRWDO *HLQGXFHG HQHUJ\ JDS YDULDWLRQ WKURXJK [;5 7KH EXLOWLQ DLGLQJ HOHFWULF ILHOG (EL IRU WKH HOHFWURQV LQ WKH SW\SH *H6L EDVH ZLWK D XQLIRUP ERURQ GRSLQJ DQG SRVLWLRQ GHSHQGHQW HQHUJ\ JDS (*[f IURP WKH JUDGHG *H FRQFHQWUDWLRQ SURILOH LV JLYHQ E\ >@ (EL Tf&G(FG[f Tf$(*B*Hf;5 LQGHSHQGHQW RI [ Df N7TfS;5f Ef ZKHUH U_ $(**HN7 LV WKH QRUPDOL]HG *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ WKURXJK WKH EDVH )RU ;5[;% WKH EXLOWLQ DLGLQJ ILHOG LV GXH WR WKH QRQXQLIRUP EDVH GRSLQJ 7KLV FDQ EH DSSUR[LPDWHG DVVXPLQJ DQ H[SRQHQWLDO SURILOH 1$$[f 1$$fH[S>U_[ ;Uf;E;Uf@ ZKHUH U_ ,Q >1$$;5f1$$;%f@ 7KH FRQVWDQW EXLOWLQ DLGLQJ ILHOG IRU ;5[;5 LV JLYHQ E\ (EL N7Tf>W";%;5f@ f 7KH /)*9 PHWKRGRORJ\ DOVR XVHV WKH FROOHFWRU FXUUHQW UDWLR ,F*H6Lf,F6Lf EXW WKLV UDWLR LV PRGLILHG 7KH *HLQGXFHG 1&19 UHGXFWLRQ WHUP 1F1Y*H6Lf1F1Y6Lf LV OXPSHG WRJHWKHU ZLWK WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ WHUP $(*B*H N7TfU_ ZKHUH WKH OXPSHG SDUDPHWHU LV GHVFULEHG E\ D IDFWRU ULS 7KLV PRGLILFDWLRQ JLYHV VHYHUDO EHQHILWV )LUVW ULM FDQ EH PHDVXUHG GLUHFWO\ DQG DFFXUDWHO\ ZKLOH WKH RULJLQDO WZR IDFWRUV DUH QRW PHDVXUDEOH 6HFRQGO\ DOO RWKHU *H6L +%7V SDUDPHWHUV FDQ EH FDOFXODWHG LQ WHUPV RI ULM 7KLV PRGLILFDWLRQ DOVR HQDEOHV WKH GHFRXSOLQJ RI WKH HIIHFWV RI 1&19 IURP WKDW RI WKH PRELOLW\ DQG KHDY\ GRSLQJ RQ ,F*H6Lf
PAGE 44
7KH FROOHFWRU FXUUHQW GHQVLW\ IRU QRQXQLIRUP EDVH FDQ EH H[SUHVVHG DV >@ F *H6Lf >H[S T9%(N7f@ % >3%T'Q [f Q" [f @ G[ f 8VLQJ WKH DQDO\WLFDO PRGHO IRU D WUDSH]RLGDO *H GLVWULEXWLRQ LQ WKH EDVH UHJLRQ DV GLVFXVVHG LQ VHFWLRQ WKH LQWULQVLF FDUULHU GHQVLW\ LQ WKH *H6L EDVH LV JLYHQ E\ QI[f 1F1Y[fH[S(*RN7f H[SDH[S\H[SUM[;5f [;5f Df QI[f 1F1Y[fH[S(*RN7f f H[SDH[S\ H[SU"f ;5[;%f Ef ZKHUH D $(*B*HfN7 DQG \ $(*B+'*H6LfN7 DQG KDYH EHHQ GHILQHG LQ VHFWLRQ 7KH EDVH GRSLQJ LQ WKH *H6L EDVH OD\HU LV JLYHQ E\ QER [[Uf 3[f f O1%H[S>[[5f[@ [5[[%f ZKHUH [ LV GHILQHG DV [ ;%;5fOQ>1$$3;%f@ ,I ZH XVH DYHUDJH YDOXHV IRU WKH 1&19 DQG WKH GLIIXVLYLW\ 'Q WKHQ ZH FDQ WDNH WKHVH WZR TXDQWLWLHV RXW RI WKH LQWHJUDO LQ f DQG REWDLQ WKH IROORZLQJ H[SUHVVLRQ IRU WKH ,F UDWLR ,F*H6Lf 1F1Y*H6Lf '%*H6Lf D$\T $ $fOH[SLff" H f ,F6Lf 1&196Lf 'E6Lf H[SUMO OH[SAf $ $f ZKHUH $\ >$(*B+'*H6Lf$(*B+'6Lf@N7 DQG $ ;5;% f FDQ WKHQ EH XVHG WR HYDOXDWH WKH HIIHFW RI $(*B*H IURP WKH WHPSHUDWXUH GHSHQGHQFH PHDVXUHPHQW RU FDOFXODWLRQ DV GLVFXVVHG LQ VHFWLRQ %XW LQ /)*9 PHWKRG ZH FKRRVH WR XVH D VLPSOH PRGHO IRU WKH 1F1Y[f DV LV GLVFXVVHG LQ VHFWLRQ DQG WKHQ OXPS WKLV 1&19 YDULDWLRQ ZLWK WKH $(**H
PAGE 45
QI [f 1F[f1Y[f 1Ff1YfH[S&[;5f f ZKHUH & LV D /6) SDUDPHWHU 7KHQ WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ LV JLYHQ E\ QL R H[SD H[S \ H[S [[[f [[5f f QL H[SDr H[S \ H[SA[A [5[[%f ZKHUH Qc4 LV WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ LQ ORZ GRSHG 6L JLYHQ E\ QL 1Ff1YfH[S(*2N7f Df DQG [ ;%>$(*B*HN7f&@ ;%UL &RPELQLQJ f f DQG f \LHOGV WKH UDWLR RI WKH FROOHFWRU FXUUHQW LQ *H6L DQG 6L ,F*H6Lf '%*H6Lf $O$f>OH[SU"f@@ f H[SWAD$\f f ,F6Lf '%6Lf $H[SAOfAO$f>OH[SAf@UM ZKHUH Uf [ U_& LV WKH QRUPDOL]HG HIIHFWLYH HQHUJ\ JDS YDULDWLRQ WKURXJK WKH EDVH OD\HU WKDW LQFOXGHV WKH HIIHFWV RI YDULDEOH *H FRQFHQWUDWLRQ DQG YDULDWLRQ RI GHQVLW\ RI VWDWHV ZLWK [*H ,Q f '%*H6Lf % >3Q[f@G[ % 3G['Q[fQ[f LV WKH VSDWLDOO\ DYHUDJHG HOHFWURQ GLIIXVLYLW\ LQ WKH EDVH 7KH ZHLJKWLQJ IDFWRU RI '%*H6Lf LV Qc[f ZKLFK LQFUHDVHV ZLWK [ RU [*H 7KXV WKH YDOXH RI '%*H6Lf LV ZHLJKWHG WRZDUG WKH UHJLRQ FORVH WR [ ;% 6LQFH 'Q*H6Lf LV H[SHFWHG WR EH D YHU\ ZHDN IXQFWLRQ RI [*H IRU WKH DYHUDJH [*(b >@ GHILQLQJ '%*H6Lf LQ f FDXVHV RQO\ D QHJOLJLEOH HUURU )RU D SXUHO\ OLQHDUO\JUDGHG *H FRQFHQWUDWLRQ SURILOH ;5 ;% DQG $ O ZKLFK VLPSOLILHV fWR >@ ,F*H6Lf @L '%*H6Lf H[SD f H[S$\ f ,F6Lf OH[SIQf 'J6Lf 2QH UHDVRQ IRU OXPSLQJ WKH VSDWLDO YDULDWLRQ RI 1&19 ZLWK WKH VSDWLDO YDULDWLRQ RI WKH HQHUJ\ JDS $(*B*H LV WKDW WKHVH WZR SDUDPHWHUV DOZD\V DSSHDU WRJHWKHU DV D SURGXFW LQ
PAGE 46
WKH FDOFXODWLRQ RI RWKHU *H6L SDUDPHWHUV VXFK DV WKH WUDQVLW WLPH 7KXV WKH DQDO\VLV RI WKH *H6L +%7 ZLWK D WUDSH]RLGDO RU OLQHDUO\JUDGHG *H FRQFHQWUDWLRQ SURILOHV FDQ EH FRQYHQLHQWO\ PDGH LQ WHUPV RI D VLQJOH SDUDPHWHU UA ,W LV LPSRUWDQW WR QRWH KHUH WKDW DV ZLOO EH VKRZQ ODWHU WKH YDOXH RI WKH ILWWLQJ SDUDPHWHU & UM UM[f ZKLFK LV PRGHO GHSHQGHQW GRHV QRW DIIHFW WKH GHWHUPLQDWLRQ RI WKH EDVH GLIIXVLRQ OHQJWK UHFRPELQDWLRQ OLIHWLPH WUDQVLW WLPH RU GLIIXVLYLW\ ,W RQO\ DIIHFWV WKH FDOFXODWLRQ RI WKH EXLOWLQ HOHFWULF ILHOG (EL DQG $(*B*H ZKLFK PD\ JLYH D VPDOO HUURU GXH WR WKH XQFHUWDLQW\ RI WKH ILWWLQJ SDUDPHWHU & +RZHYHU WKLV DQDO\WLFDO DSSURDFK LV PXFK PRUH DFFXUDWH WKDQ XVLQJ WKH VSDWLDOO\ DYHUDJHG YDOXH RI 1&19 LQ WKH LQWHJUDO HTXDWLRQ f ZKLFK DOORZV WDNLQJ 1&19 RXWVLGH WKH LQWHJUDO $QRWKHU YHU\ LPSRUWDQW IHDWXUH RI WKH QHZ H[SUHVVLRQ f LV WKDW LW GHFRXSOHV WKH 1&19f SURGXFW IURP PRELOLW\ LQ f DQG DOORZV GHWHUPLQDWLRQ RI WKH PRELOLW\ UDWLR ,Q RUGHU WR DQDO\]H WKH *HLQGXFHG ,F HQKDQFHPHQW ZH KDYH WR VHSDUDWH WKH PHDVXUHG UDWLR LQ f RU f LQWR ILYH RU IRXU WHUPV $ URXJK DSSUR[LPDWLRQ IRU SUDFWLFDO SURILOHV ZLWK [*Hf LV WR DVVXPH WKDW WKH HQKDQFHPHQW LQ ,F LV GRPLQDWHG E\ WKH EXLOWLQ ILHOG ZKLFK OHDYHV RQO\ WKH ILUVW WHUP LQ f DQG OHDGV GLUHFWO\ WR WKH HVWLPDWLRQ RI UA +RZHYHU IRU DQ DFFXUDWH DQDO\VLV RI WKH WUDQVLVWRU DQG IRU SXUSRVH RI DFTXLULQJ D PDWHULDO GDWDEDVH PRUH DFFXUDWH DSSURDFK LV QHFHVVDU\ )RU H[DPSOH DQ XQFHUWDLQW\ LQ $(*B*Hf RU $(*+' RI O2PH9 JLYHV DQ HUURU RI aH[Sf 7KXV DQ DQDO\VLV RI WKH SHUIRUPDQFH RI 6L*H[6L[6L +%7 EDVHG RQ PHDVXUHPHQW RI ,F RQO\ LV FOHDUO\ QRW VXIILFLHQW
PAGE 47
7KH /)*9 PHWKRGRORJ\ GHVFULEHG LQ WKH QH[W VHFWLRQ ZLOO JLYH DQ DFFXUDWH H[WUDFWLRQ RI ULM 'HULYDWLRQ RI /)*9 0HWKRGRORJ\ DQG 3DUDPHWHU ([WUDFWLRQ %HORZ ZH SUHVHQW GHWDLOHG GHULYDWLRQ DQG GHVFULSWLRQ RI WKH /)*9 PHWKRGRORJ\ $V LV GLVFXVVHG LQ VHFWLRQ WKH EDVH WKLFNQHVV PRGXODWLRQ JLYHV ULVH WR WZR VPDOOVLJQDO FRQGXFWDQFHV DQ RXWSXW FRQGXFWDQFH J DQG D UHYHUVH WUDQVFRQGXFWDQFH JU ,Q WKH ILUVW VWHS RI WKH /)*9 PHWKRGRORJ\ ZH QHHG WKH J RQO\ ,W FDQ EH GHULYHG IURP WKH ,F HTXDWLRQV f DQG f DQG LV JLYHQ E\ VHH DSSHQGL[ $f G ;5 R ;% $H[SSOfH[SSKL O$fH[SSOfQ 9 &% f n%( )URP f ZH FDQ VHH LPPHGLDWHO\ WKDW WKH H[SUHVVLRQ f IRU J LV VLPSOHU WKDQ f IRU ,F 7KH UHDVRQ IRU WKLV LV WKDW WKH XQNQRZQ SDUDPHWHUV '%*H6Lf'%6Lf D DQG \ DUH OXPSHG LQWR ,F ZKLFK LV D PHDVXUDEOH TXDQWLW\ 7KLV LV RQH RI WKH DGYDQWDJHV RI J FRPSDUHG ZLWK ,F DQG LV RQH LPSRUWDQW DVSHFW RI WKH /)*9 PHWKRGRORJ\ 8VHIXO UHODWLRQV Df DQG Ef DUH GHULYHG E\ WDNLQJ WKH UDWLRV RI WKH FROOHFWRU FXUUHQWV DQG WKH RXWSXW FRQGXFWDQFHV RI WKH *H6L +%7 DQG 6L UHIHUHQFH %7 ,F*H6Lf,F6Lf $>H[SSOfH[SS@S O$fH[SSOfS Df J*H6LfJ6Lf ,F*H6Lf,F6Lf J*H6LfJ6Lf $H[SS O$fH[SSOfS H[SUAO Ef KL 7KH RQO\ DVVXPSWLRQ LQ Df DQG Ef LV WKDW ;%9&% LV WKH VDPH IRU *H6L +%7 DQG 6L UHIHUHQFH %7 7KLV DVVXPSWLRQ LV SURYHQ WR EH D YHU\ JRRG DSSUR[LPDWLRQ IURP PHDVXULQJ WKH MXQFWLRQ FDSDFLWDQFH &%& 7KHVH UHODWLRQV DUH XVHG WR FDOFXODWH Uf W
PAGE 48
XQLTXHO\ DQG LQGHSHQGHQWO\ RI GLIIXVLYLW\ GHQVLW\ RI VWDWHV DQG KHDY\GRSLQJ HIIHFWV ZKLFK FDQFHO RXW LQ WKH UDWLRV RI ,F DQG J (TXLYDOHQW UHODWLRQV WR WKRVH LQ f f FDQ EH GHULYHG IRU WKH UHYHUVHDFWLYH RSHUDWLRQ PRGH ZLWK WKH %& MXQFWLRQ IRUZDUGELDVHG DQG WKH (% MXQFWLRQ UHYHUVH ELDVHG )RU WKLV FDVH H[SU_Mf DQG H[SU_f DUH UHSODFHG E\ H[SOUAf DQG H[S7@f DQG ULM DQG U_ DUH UHSODFHG E\ LT [ DQG U_ )URP Ef ZH FDQ VHH WKDW IRU D WULDQJXODU *H GLVWULEXWLRQ ZH FDQ FDOFXODWH WKH U_M GLUHFWO\ IURP WKH PHDVXUHG FROOHFWRU FXUUHQW DQG RXWSXW FRQGXFWDQFH %XW IRU D WUDSH]RLGDO *H GLVWULEXWLRQ LQ Df ZH QHHG WR NQRZ $ ;5;% DQG U_ ,Q WKH ODWHU SDUW RI WKLV VHFWLRQ ZH ZLOO VKRZ WKDW U? FDQ EH FDOFXODWHG IURP D FRPELQDWLRQ RI IRUZDUG DFWLYH PHDVXUHPHQW DQG UHYHUVH DFWLYH PHDVXUHPHQW RI ,F6Lf J6Lf DQG JU6Lf RI WKH 6L UHIHUHQFH %7 $OVR ZH ZLOO VKRZ WKDW WKH ,F UDWLR LV QRW VHQVLWLYH WR ;5;% DQG U_ VLQFH WKH\ DUH WKH VDPH IRU *H6L +%7 DQG 6L %7 DQG QHDUO\ FDQFHOHG RXW LQ WKH ,F UDWLR $IWHU DQ LQGHSHQGHQW GHWHUPLQDWLRQ RI LT [ WKH GHWHUPLQDWLRQ RI RWKHU SDUDPHWHUV LV VWUDLJKWIRUZDUG DQG ZLOO EH GLVFXVVHG WRJHWKHU ZLWK WKH H[SHULPHQWDO UHVXOWV LQ WKH QH[W VHFWLRQ 7KH WRWDO EDVH UHFRPELQDWLRQ FXUUHQW ,% LV FRPSRVHG RI WKH UHFRPELQDWLRQ ORVVHV LQ WKH HPLWWHU DQG EDVH UHJLRQV 4E 4H ,% f WE WH ZKHUH 4( DQG 4% DUH WKH VWRUHG FKDUJHV DQG [( DQG [% DUH WKH FKDUJHFRQWURO PLQRULW\ FDUULHU UHFRPELQDWLRQ OLIHWLPH RU WLPH FRQVWDQW LQ WKH HPLWWHU DQG EDVH UHVSHFWLYHO\ ,Q
PAGE 49
f ZH QHJOHFWHG WKH VSDFHFKDUJH UHJLRQ UHFRPELQDWLRQ FRPSRQHQWV WKDW FDQ EH HDVLO\ VHSDUDWHG RXW IURP WKH PHDVXUHG ,% ,Q RUGHU WR VHSDUDWH WKH UHFRPELQDWLRQ ORVVHV LQ f ZH QHHG WKH JU PHDVXUHPHQW ZKLFK UHVXOWV IURP WKH PRGXODWLRQ RI WKH EDVH UHFRPELQDWLRQ ORVV )RU WKH *H6L +%7 JU LV GHULYHG DV 4J ; G;J JU f LJ;J < $>H[SIMAOfAf@V[SUM O$f>V[SIOfIM@ AFE AEH ,Q f WKH WHUPV ; DQG < DUH JLYHQ EHORZ ; $H[SAAOfM $G$O&LH[SA'UML@ >H[SAOfS@ O$fH[SSSOfQ Df < $>H[SU"fSOf@S $ $f >OH[S ;f >OH[SUMf@SS O$f>H[SSfQO@S Ef :KHQ ZH WDNH WKH UDWLR RI JU DQG J D XVHIXO H[SUHVVLRQ LV REWDLQHG U W% % f R WE ZKHUH W% LV WKH EDVH PLQRULW\ FDUULHU WUDQVLW WLPH [% LV WKH FKDUJHFRQWURO PLQRULW\ FDUULHU EDVH UHFRPELQDWLRQ OLIHWLPH DQG %LVD IDFWRU UHODWHG WR WKH EXLOWLQ HOHFWULF ILHOG LQ WKH EDVH GXH WR ERWK *H DQG GRSLQJ QRQXQLIRUPLW\ % LQ ILHOGIUHH EDVHf % LV D IXQFWLRQ RI ULM U_ QI[f DQG KROH FRQFHQWUDWLRQ LQ WKH EDVH 3[f DQG LV JLYHQ E\ f EHORZ >@ % L9[f G[ G[ 3%[f QL[f3[f f
PAGE 50
;< Df H[SUMLLQ ;5 ;% Ef H[SU"LfMLO $ FRPELQDWLRQ RI WKH GF ,%9%( DQG VPDOOVLJQDO PHDVXUHPHQWV JLYHV WKH H[SUHVVLRQV IRU WKH EDVH GLIIXVLRQ OHQJWK DQG UDWLR RI WKH UHFRPELQDWLRQ ORVVHV LQ WKH HPLWWHU DQG EDVH UHJLRQ 4HWH %JJUf f 4EA W E eI /_ ;%JJUf^$H[SWMWMOfI@ $O$f>H[SAOfQL@ f >H[SUMOfQ@ O$fH[SUQOfK` Df UH[SU[f /E ;%JJUf ;5 ;% Ef QL $ FRPELQDWLRQ RI f DQG f JLYHV 4VARH6L DUrG 4EWEfV IURP PHDVXUHG GDWD 8VLQJ Df RU Ef ZH WKHQ FDOFXODWH GLUHFWO\ WKH GLIIXVLRQ OHQJWK ZKHUH WKH RQO\ XQNQRZQ SDUDPHWHU LV ;% ZKLFK FDQ EH PHDVXUHG E\ 6,06 ([SUHVVLRQ f LV DOVR XVHG WR FDOFXODWH WKH SDUDPHWHU U_ ZKLFK GHWHUPLQHV WKH EXLOWLQ ILHOG IRU ;5[;% 7KLV FDQ EH IRXQG IURP WKH 6L %7 IRU ZKLFK UMA2 0HDVXULQJ WKH RXWSXW DQG UHYHUVH FRQGXFWDQFHV IRU WKH IRUZDUG VXEVFULSW )f DQG UHYHUVH VXEVFULSW 5f PRGH RI RSHUDWLRQV ZH REWDLQ WKH IROORZLQJ UDWLR 6RI4R5 %S WJS f %U W%5 6W)A2S5 f
PAGE 51
7KH UDWLR LQ f LV D IXQFWLRQ RI $ ;5;% DQG Uf RQO\ ,I ;5;% LV NQRZQ IURP 6,06 RU IDEULFDWLRQ FRQGLWLRQV U_ FDQ EH FDOFXODWHG IURP f ([SHULPHQWV DQG 5HVXOWV +LJK SHUIRUPDQFH *H6L +%7 DQG UHIHUHQFH 6L %7 DUH XVHG WR GHPRQVWUDWH WKH /)*9 PHWKRGRORJ\ 6,06 SURILOH RI WKH FRPSOHWHG *H6L +%7 LV VKRZQ LQ )LJ 7KH UHIHUHQFH 6L %7 LV IDEULFDWHG ZLWK WKH LGHQWLFDO IDEULFDWLRQ VWHSV H[FHSW LW KDV QR *H LQ WKH EDVH OD\HU 0HDVXUHPHQWV RI WKH (% MXQFWLRQ FDSDFLWDQFH I)SP IRU *H6L DQG I)SP IRU 6Lf EUHDNGRZQ YROWDJH 9 IRU *H6L DQG IRU 6Lf DQG EDVH VKHHW UHVLVWDQFH XQGHU WKH HPLWWHU NeVTXDUH IRU *H6L DQG NeVTXDUH IRU 6Lf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f,F6Lf 7KH LQVHQVLWLYLW\ RI ,F*H6Lf WR WKH UHYHUVH EDVHFROOHFWRU YROWDJH XS WR 9%& 9 VKRZV WKH DEVHQFH RI SDUDVLWLF HQHUJ\ EDUULHU DW WKH %& MXQFWLRQ WKDW PD\ UHGXFH ,F >@ )LJXUH JLYHV WKH PHDVXUHG VPDOOVLJQDO RXWSXW FRQGXFWDQFH J UHYHUVH WUDQVFRQGXFWDQFH JU IRU ERWK WKH IRUZDUG DFWLYH PRGH DQG UHYHUVH DFWLYH PRGH DQG WKH FROOHFWRU FXUUHQW ,F RI *H6L +%7 DV D IXQFWLRQ RI 9%( VKRZLQJ WKH H[SHFWHG H[ST9%(N7f GHSHQGHQFH >@ )LJXUH JLYHV WKH PHDVXUHG VPDOOVLJQDO RXWSXW
PAGE 52
FRQGXFWDQFH J UHYHUVH WUDQVFRQGXFWDQFH JU IRU ERWK WKH IRUZDUG DFWLYH PRGH DQG UHYHUVH DFWLYH PRGH DQG WKH FROOHFWRU FXUUHQW ,F RI 6L %7 7KH PHDVXUHG GHSHQGHQFLHV LQ )LJV DUH XVHG IRU WKH /)*9 DQDO\VLV 7KH UHVXOWV IRU ODUJH VTXDUH GHYLFH [SPf DUH VXPPDUL]HG LQ 7DEOH $ GHWDLOHG GLVFXVVLRQ RI WKH DQDO\VLV DQG LWV LPSOLFDWLRQV IROORZV 7KH ILUVW VWHS LQ WKH DQDO\VLV LV WR ILQG WKH SDUDPHWHU U_ XVLQJ f ZKLFK GHWHUPLQHV WKH EXLOWLQ ILHOG IRU ;5[;% )RU NQRZQ ;5;% f \LHOGV U? OQ>1$$;5f1$$;%f@ IRU ERWK 6L %7 DQG *H6L +%7 7KLV FRUUHVSRQGV WR 1$D;Uf1DD;Ef ZKLFK DJUHHV ZHOO ZLWK WKH 6,06 SURILOHV LQ )LJ :LWK WKH NQRZQ U_ ZH FDQ WKHQ FDOFXODWH WKH NH\ SDUDPHWHU U_ M IURP WKH PHDVXUHG ,F DQG J IRU WKH *H6L DQG 6L WUDQVLVWRUV XVLQJ Df 7KLV JLYHV DQG U? UA &f s XVLQJ & s 7KH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DW [*H b LV A**H N7fUf sPH9 7KLV FRUUHVSRQGV WR $(JBJH [*( PH9 ZKLFK LV LQ YHU\ JRRG DJUHHPHQW ZLWK WKH WKHRUHWLFDO FDOFXODWLRQV DQG RWKHU H[SHULPHQWV LQ *H6L DV D IXQFWLRQ RI [*H 7KLV JRRG DJUHHPHQW GHPRQVWUDWHV WKH DFFXUDF\ RI RXU PRGHO WKDW DFFRXQWV IRU WKH GHSHQGHQFH RI WKH 1&19 RQ *H FRQFHQWUDWLRQ 7KH YDOXH RI WKH ILWWLQJ SDUDPHWHU & FDXVHV RQO\ VPDOO XQFHUWDLQW\ LQ WKH UHVXOWV 7KH *HLQGXFHG DLGLQJ HOHFWULF ILHOG ZLWKLQ WKH OLQHDU *H UDQJH [;5 (EL N7TfU@;5f sN9FP 7KH PRELOLW\ UDWLR '%*H6Lf'%6Lf FDQ EH HYDOXDWHG IURP f 7KH UHPDLQLQJ XQNQRZQV LQ f DUH D $(*B*HfN7 DQG $\ $((B+'N7 6,06 SURILOHV LQGLFDWH D DW 9EH 9 7KLV JLYHV >'%*H6Lf'%6Lf@}H[S$\f 7KLV SURGXFW FDQQRW EH VHSDUDWHG ZLWKRXW DGGLWLRQDO LQGHSHQGHQW PHDVXUHPHQW RI HLWKHU WKH PRELOLW\ UDWLR RU
PAGE 53
9EH9f )LJ ,F9%( DQG ,%9%( FKDUDFWHULVWLFV RI *H6L +%7 DQG 6L %7 7KH PHDVXUHPHQWV ZHUH WDNHQ RQ ODUJH WUDQVLVWRUV ZLWK WKH (% MXQFWLRQ DUHD RI [_XP LQ RUGHU WR HOLPLQDWH WKH HGJH DQG VLGHZDOO HIIHFWV 1RWH WKH VLJQLILFDQW LQFUHDVH LQ ,F DQG VOLJKW GHFUHDVH LQ ,% IRU WKH *H6L WUDQVLVWRU 7KH PHDVXUHG HQKDQFHPHQW LQ WKH FROOHFWRU FXUUHQW ,F*H6Lf,F6Lf
PAGE 54
9EH 9f )LJ 7KH PHDVXUHG IRUZDUG DFWLYH PRGH DQG UHYHUVH DFWLYH PRGH VPDOOVLJQDO RXWSXW FRQGXFWDQFHV J UHYHUVH WUDQVFRQGXFWDQFH JU DQG WKH FROOHFWRU FXUUHQW ,F RI *H6L +%7 DV D IXQFWLRQ RI 9%( $OO PHDVXUHG FXUYHV VKRZ WKH H[SHFWHG H[ST9%(N7f GHSHQGHQFH 98f rO
PAGE 55
)LJ 7KH PHDVXUHG IRUZDUG DFWLYH PRGH DQG UHYHUVH DFWLYH PRGH VPDOOVLJQDO RXWSXW FRQGXFWDQFHV J UHYHUVH WUDQVFRQGXFWDQFH JU DQG WKH FROOHFWRU FXUUHQW ,F RI 6L %7 DV D IXQFWLRQ RI 9%( $OO PHDVXUHG FXUYHV VKRZ WKH H[SHFWHG H[ST9%(N7f GHSHQGHQFH ,&$f
PAGE 56
7DEOH 6XPPDU\ RI H[SHULPHQWDO GDWD DQG FDOFXODWHG SDUDPHWHUV IRU 6L*H[6LB[6L +%7 DQG UHIHUHQFH 6L %7 7KH WRS VL[ OLQHV JLYH WKH PHDVXUHG GF DQG %70 DF GDWD REWDLQHG DW r& 9%( 9 DQG 9&% ,9 7KH FRPSXWHG SDUDPHWHUV XVLQJ /)*9 PHWKRG DUH VKRZQ LQ WKH QH[W HLJKWHHQ OLQHV 7KH ILUVW FROXPQ LV IRU 6L*H[6LM[6L +%7 WKH VHFRQG IRU 6L EDVHOLQH %7 DQG WKH WKLUG FROXPQ JLYHV WKH FRPSDULVRQ UDWLR RI WKH WZR SUHYLRXV FROXPQV $%( [OSP ;5 Âƒ ;% Âƒ 3DUDPHWHU *H6L +%7 6L %7 *H6L6L ,F 3$f ,% Q$f 3I JR ,2f [R [O JU *f [O [O JRAJU A %L $(*B*H PH9f (EL N9FPf % /E LPf 4EAE r$f 4HWH Q$f 4HA:Ef 4HWHf,E bf 4EWEf,E bff 4E &f [O [O 'E FP9rVf W% 3Vf WE QVf $\ $(JKG PH9f 9$ 9f
PAGE 57
$\ )LJXUH VKRZV WKH PHDVXUHG WHPSHUDWXUH GHSHQGHQFH RI OQ>H[S$\fm' %*H6Lf'%6Lf@ LQ WKH UDQJH RI .. 7KH WHPSHUDWXUH GHSHQGHQFH VKRZV D VPDOO DFWLYDWLRQ HQHUJ\ RI ($aPH9 7KLV LV FRQVLVWHQW ZLWK $\ $(*+'N7 RU $(JKG PH9 ZKLFK LV FDOFXODWHG IURP WKH PHDVXUHG URRP WHPSHUDWXUH ,F UDWLR DVVXPLQJ D XQLW\ PRELOLW\ UDWLR ZKLFK LV LQGHSHQGHQW RI WHPSHUDWXUH $ QHJOLJLEOH LPSURYHPHQW LQ WKH PLQRULW\FDUULHU PRELOLW\ IRU WKH DYHUDJH ab *H FRQFHQWUDWLRQ DQG a[OFP EDVH GRSLQJ DW URRP WHPSHUDWXUH ZDV DOVR SUHGLFWHG WKHRUHWLFDOO\ DV LV GLVFXVVHG LQ VHFWLRQ 7KH FRQFOXVLRQ WKDW '%*H6Lf '%6Lf LV DOVR VXSSRUWHG E\ WKH VSDUDPHWHU PHDVXUHPHQWV RI WKH EDVH WUDQVLW WLPH IRU +%7V 8VLQJ WKH PHDVXUHG GDWD LQ )LJ ZH FDQ FDOFXODWH WKH WHPSHUDWXUH GHSHQGHQFH RI WKH PLQRULW\ FDUULHU GLIIXVLYLW\ UDWLR DVVXPLQJ WKDW $(*B+' PH9 LV FRQVWDQW ZLWK WKH WHPSHUDWXUH UDQJH RI .. 7KH UHVXOWV DUH VKRZQ LQ )LJ DQG DUH LQ JRRG DJUHHPHQW ZLWK WKHRUHWLFDO SUHGLFWLRQV >@ 7KH UHFRPELQDWLRQ FRPSRQHQWV RI ,% DUH FDOFXODWHG IURP f DQG f 7KH FDOFXODWLRQ VKRZV WKDW WKH UHFRPELQDWLRQ ORVVHV LQ WKH HPLWWHU DQG EDVH FRQVWLWXWH DQG b RI ,% UHVSHFWLYHO\ LQ *H6L +%7 DQG b DQG b LQ 6L %7 $Q LPSRUWDQW VHOIFRQVLVWHQF\ UHVXOW LV WKDW 4(7(f*HVL 4HWHfVS DV DQWLFLSDWHG VLQFH [*H DW WKH (% MXQFWLRQ DQG WKXV WKH KROH LQMHFWLRQ IURP WKH EDVH LQWR WKH HPLWWHU DQG WKH HPLWWHU UHFRPELQDWLRQ DUH H[SHFWHG WR EH WKH VDPH LQ ERWK WUDQVLVWRUV 7KH VPDOO GLIIHUHQFHV FDQ EH H[SODLQHG E\ GLIIHUHQFHV LQ WKH HPLWWHU SURILOH DQG UHFRPELQDWLRQ SURSHUWLHV DW WKH QSRO\VLOLFRQQ6L HPLWWHU LQWHUIDFH 7KH YHU\ VLJQLILFDQW UHGXFWLRQ LQ WKH UHFRPELQDWLRQ ORVVHV LQ WKH EDVH 4EA*F6A4EAEA fÂ§ LV GXH WR D VLJQLILFDQWO\
PAGE 58
OQ>H[S$\f'%*H6Lf'%6Lf@ 7 .f )LJ 7KH PHDVXUHG WHPSHUDWXUH GHSHQGHQFH RI OQ>H[S$\fm' %*H6Lf'%6Lf@ LQ WKH UDQJH RI .. QRUPDOL]HG WR LQ ODUJH [SP WUDQVLVWRU
PAGE 59
'%*H6Lf'%6Lf 7HPSHUDWXUH .f )LJ 7KH PHDVXUHG WHPSHUDWXUH GHSHQGHQFH RI '%*H6Lf'%6Lf LQ WKH UDQJH RI .. QRUPDOL]HG WR LQ ODUJH [SP WUDQVLVWRU
PAGE 60
ODUJHU EDVH PLQRULW\ UHFRPELQDWLRQ OLIHWLPH WE UDWLR 7%*H6Lf7%6Lf QVOOQV 7KH UHDVRQV IRU WKH VLJQLILFDQW LPSURYHPHQW RI WKH PLQRULW\FDUULHU UHFRPELQDWLRQ OLIHWLPH LQ WKLQ VWUDLQHG *H6L OD\HUV LV QRW NQRZQ DQG UHTXLUHV IXUWKHU VWXG\ 7KH GLIIXVLRQ OHQJWK UDWLR FDQ EH FDOFXODWHG IURP f DQG JLYHV /%*H6Lf/%6Lf SP LQ WKH *H6L OD\HU 7KLV UDWLR LV FDOFXODWHG GLUHFWO\ IURP WKH VPDOOVLJQDO %70 PHDVXUHPHQWV ZLWKRXW XVLQJ WKH GF ,E9EH FXUYH 7KLV LV DQ DGYDQWDJH IRU +%7V WKDW KDYH D ODUJH %( MXQFWLRQ OHDNDJH RU ODUJH VSDFHFKDUJH OD\HU UHFRPELQDWLRQ FRPSRQHQW [% ZDV FDOFXODWHG WDNLQJ '% FPV >@ IRU ERWK WUDQVLVWRUV $OWKRXJK WE LV UHODWLYHO\ VKRUW QVf LW KDV RQO\ D VPDOO HIIHFW RQ WKH JDLQ VLQFH ,% LQ WKH ODUJH [SP WUDQVLVWRUV LV GRPLQDWHG E\ WKH HPLWWHU 5RXJK HVWLPDWHV RI WE LQ *H6L +%7V ZHUH SXEOLVKHG SUHYLRXVO\ >@ EXW RXU SUHVHQW PHDVXUHPHQWV LQ ODUJH WUDQVLVWRUV DUH WKH ILUVW ULJRURXV DFFXUDWH GDWD 7KH PLQRULW\ FDUULHU EDVH FKDUJH 4% LV ODUJHU LQ *H6L WUDQVLVWRU 4%*H6Lf4%6Lf EHFDXVH RI VPDOOHU HQHUJ\ JDS LQ WKH EDVH ZKLFK LQFUHDVHV Qc 7KLV LQFUHDVH LQ WKH VWRUHG FKDUJH VKRXOG EH DFFRXQWHG IRU LQ VSHHG FDOFXODWLRQV 7KH HOHFWURQ WUDQVLW WLPH DFURVV WKH TXDVLQHXWUDO EDVH OD\HU LV WKHQ FDOFXODWHG DQG JLYHV SVSV D IDFWRU RI DERXW WZR LPSURYHPHQW IRU WKH *H6L +%7 DV H[SHFWHG 7KLV LV GXH DOPRVW HQWLUHO\ WR WKH EXLOWLQ ILHOG IURP WKH JUDGHG *H FRQFHQWUDWLRQ SURILOH UDWKHU WKDQ WKH PRELOLW\ HQKDQFHPHQW ZKLFK LV QHJOLJLEOH 7KH (DUO\ YROWDJH IRU *H6L +%7 LV VLJQLILFDQWO\ KLJKHU ZKLFK LV LPSRUWDQW IRU DQDORJ FLUFXLW DSSOLFDWLRQ 7KH UHVXOWV IRU D VWULSH WUDQVLVWRU ZLWK QDUURZ HPLWWHU ILQJHUV [_LP ZLWK ILQJHUVf DUH VXPPDUL]HG LQ 7DEOH 7KH PDMRU GLIIHUHQFH EHWZHHQ ODUJH VTXDUH
PAGE 61
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f'%*H6Lf'%6Lf@ LQ WKH UDQJH RI .. 7KH WHPSHUDWXUH GHSHQGHQFH VKRZV D VPDOO DFWLYDWLRQ HQHUJ\ RI ($aPH9 )RU WKLV WUDQVLVWRU SDLU WKH PHDVXUHG >'%*H6Lf'%6Lf@H[S$\f FRPSDUHG WR IRU WKH ODUJH WUDQVLVWRU 7KLV JLYHV '%*H6Lf'%6Lf 7KH GLVDJUHHPHQW ZLWK WKH '%*H6Lf'%6Lf UHVXOW REWDLQHG IURP WKH ODUJH WUDQVLVWRU LV GXH WR WKH QDUURZ ILQJHU JHRPHWU\ 6HQVLWLYLW\ RI W% DQG U_ WR U_ DQG ;5;% $Q LGHDOL]HG EDVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH 3[f DQG *H GLVWULEXWLRQ LQ WKH EDVH XVHG IRU WKH DQDO\VLV DUH VKRZQ LQ )LJ 7KH EDVH UHJLRQ LV GLYLGHG LQWR WZR UHJLRQV GHULYHG IURP WKH DFWXDO 6,06 SURILOH VKRZQ LQ )LJ )RU [;5 WKH EDVH GRSLQJ 3[f LV IODW DQG WKH *H GLVWULEXWLRQ LV UHWURJUDGHG OLQHDU DQG IRU ;5[;% WKH EDVH GRSLQJ GHFUHDVHV H[SRQHQWLDOO\ DQG WKH *H GLVWULEXWLRQ LV IODW 7KH WUDQVLWLRQ SRLQW LV GHILQHG DV ;5 $OWKRXJK ;5 FDQ EH REWDLQHG IURP WKH PHDVXUHG 6,06 SURILOHV WKLV
PAGE 62
7DEOH 6XPPDU\ RI H[SHULPHQWDO GDWD DQG FDOFXODWHG SDUDPHWHUV IRU 6L*H[6LM[6L +%7 DQG UHIHUHQFH 6L %7 7KH WRS VL[ OLQHV JLYH WKH PHDVXUHG GF DQG %70 DF GDWD REWDLQHG DW r& 9%( 9 DQG 9&% ,9 7KH FRPSXWHG SDUDPHWHUV XVLQJ /)*9 PHWKRG DUH VKRZQ LQ WKH QH[W HLJKWHHQ OLQHV 7KH ILUVW FROXPQ LV IRU 6L*H[6LB[6L +%7 WKH VHFRQG IRU 6L EDVHOLQH %7 DQG WKH WKLUG FROXPQ JLYHV WKH FRPSDULVRQ UDWLR RI WKH WZR SUHYLRXV FROXPQV $%( [f[SP ;5 Âƒ ;% Âƒ 3DUDPHWHU *H6L +%7 6L %7 *H6L6L ,F 3$f ,% Q$f 3I J LDf [Of [Of JU O4f [O2a [O2 JRAJU r $(*B*H PH9f (EL N9FPf % /E SPf 4EWE Q$f 4HWH Q$f 4HAH9&4L$Ef 4HWHf,E bf 4EWEf,E bff 4E &f [Oa [Of 'E FP9Vf W% 3Vf WE QVf $\ $(JKG PH9f 9$ 9f
PAGE 63
7 .f )LJ 7KH PHDVXUHG WHPSHUDWXUH GHSHQGHQFH RI OQ>H[S$\f'%*H6Lf'%6Lf@ LQ WKH UDQJH RI .. QRUPDOL]HG WR IRU VWULSH WUDQVLVWRU /HDVW VTXDUHV ILW WR WKH GDWD LV VKRZQ E\ WKH VROLG OLQH DQG JLYHV DQ DFWLYDWLRQ HQHUJ\ RI (D $(JKG sPH9
PAGE 64
ZLOO FHUWDLQO\ FDXVH VRPH HUURUV VLQFH WKH 6,06 SURILOHV PD\ QRW EH YHU\ DFFXUDWH IRU YHU\ WKLQ EDVH OD\HUV HVSHFLDOO\ QHDU WKH FROOHFWRU MXQFWLRQ 1RWH WKDW WKH EDVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH 1$$;f FDQ EH EHWWHU DSSUR[LPDWHG E\ D *DXVVLDQ GLVWULEXWLRQ VHH $SSHQGL[ % IRU *DXVVLDQ GLVWULEXWLRQ PRGHOf 7KLV UHVXOWV IURP WKH ERURQ RXWGLIIXVLRQ IURP WKH *H6L OD\HU GXULQJ VXEVHTXHQW IDEULFDWLRQ 7KH LQFUHDVLQJ GRSLQJ 1DD[f DW [ VKRZQ LQ )LJ ZLOO SURGXFH D UHWDUGLQJ HOHFWULF ILHOG ZKLFK ZLOO GHFHOHUDWH WKH EDVH PLQRULW\ FDUULHU PRYHPHQW DFURVV WKLV UHJLRQ 7KH GHFUHDVLQJ GRSLQJ GHQVLW\ ZLOO SURGXFH DQ DLGLQJ HOHFWULF ILHOG 7KXV DQ XQFHUWDLQW\ LQ D SUHFLVH ORFDWLRQ RI ;5 FDQ FDXVH HUURUV LQ FDOFXODWLRQ RI WKH WUDQVLW WLPH DQG RWKHU SDUDPHWHUV ,Q RUGHU WR HVWLPDWH WKH LQIOXHQFH RI WKLV XQFHUWDLQW\ RQ WKH *H6L +%7 DQDO\VLV ZH LQYHVWLJDWH WKH GHSHQGHQFH RI 7f M W% RQ ;5;% DQG U_ )LJXUH VKRZV WKH QRUPDOL]HG *HLQGXFHG HOHFWULF ILHOG U_ M YHUVXV ;5;% ZLWK U_ DV D SDUDPHWHU U_ M LV WKH NH\ SDUDPHWHU XQGHUO\LQJ WKH RSHUDWLRQ RI *H6L +%7 )LJXUH VKRZV WKH WUDQVLW WLPH UDWLR W%*H6LfW%6Lf YHUVXV ;5;% ZLWK U_ DV D SDUDPHWHU IRU A $V FDQ EH VHHQ IURP )LJ DQG )LJ DQG W%*H6LfW%6Lf DUH UDWKHU LQVHQVLWLYH WR WKH ORFDWLRQ RI WKH WUDQVLWLRQ SRLQW ;5 7KLV LV H[SHFWHG VLQFH WL DQG W%*H6LfW%6Lf DUH FDOFXODWHG IURP WKH FROOHFWRU FXUUHQW UDWLR RI WKH WZR WUDQVLVWRUV ZLWK QHDUO\ LGHQWLFDO JHRPHWU\ DQG GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOHV 7KH DERYH VHQVLWLYLW\ DQDO\VLV VKRZV WKDW WDNLQJ [ ;5 DV WKH WUDQVLWLRQ SRLQW ZKHUH WKH *H OLQHDU SURILOH LQWHUVHFWV WKH IODW *H FRQFHQWUDWLRQ SURILOH LQ WKH PHDVXUHG 6,06 GHSHQGHQFLHV )LJ ZLOO UHVXOW LQ D QHJOLJLEOH HUURU
PAGE 65
)LJ 6HQVLWLYLW\ RI Uf WR ;5;% FDOFXODWHG IURP Df ZLWK U? DV D YDULDEOH SDUDPHWHU
PAGE 66
W%*H6LfW%6Lf )LJ 6HQVLWLYLW\ RI W%*H6LfW%6Lf WR ;5;% FDOFXODWHG IURP $ f ZLWK U? DV D YDULDEOH SDUDPHWHU IRU AA
PAGE 67
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f DV LV VKRZQ LQ )LJ 7KH FROOHFWRU FXUUHQW UDWLR ,F*H6Lf,F6Lf YDULHV DOPRVW OLQHDUO\ ZLWK U_M IRU U_M!a )RU AA WKH FROOHFWRU FXUUHQW UDWLR ,F*H6Lf,F6Lf FDQ EH PRGHOHG E\ D OLQHDU IXQFWLRQ ,F*H6Lf,F6Lf rF UfM rF $(*B*HN7f& 7KH RXWSXW FRQGXFWDQFH UDWLR J*H6LfJ6Lf YV LV VKRZQ LQ )LJ $V FDQ EH VHHQ IURP WKH ILJXUH WKH J UDWLR VKRZV D UDSLG GHFUHDVH IRU O2GMA2 DQG WHQGV WR ]HUR IRU YHU\ ODUJH UM M 7KLV PHDQV WKDW WKH RXWSXW UHVLVWDQFH RI *H6L +%7 LV PXFK ODUJHU WKDQ WKDW RI 6L %7 IRU ODUJH SM DQG WKH (DUO\ YROWDJH 9$ DOVR LQFUHDVHV UDSLGO\ ZLWK U_ 6R WKH ILJXUHRIPHULW RI 9$ LPSURYHV JUHDWO\ IRU ODUJH Uf M )RU H[DPSOH IRU Uf J*H6LfJ6Lf ,F*H6Lf,F6Lf DQG 9$*H6Lf9$6Lf 7KH FXUUHQW JDLQ ZLOO DOVR LQFUHDVH URXJKO\ WLPHV IRU WKH VDPH EDVH FXUUHQW 6R WKH WRWDO 9$ LPSURYHPHQW LV a ff
PAGE 68
7KH EDVH PLQRULW\ HOHFWURQ GLVWULEXWLRQ IRU U_ LV VKRZQ LQ )LJ $V FDQ EH VHHQ IURP WKH ILJXUH Q[f WKURXJKRXW WKH EDVH LV ODUJHU IRU *H6L +%7 7KLV LV EHFDXVH WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ UA RI *H6L +%7 LV ODUJHU WKDQ WKDW RI 6L % 7 7KH EDVH PLQRULW\ FDUULHU FKDUJH UDWLR 4%*H6Lf4%6Lf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b *H LQ WKH EDVH 7KH LPSURYHPHQW RI W%*H6Lf LV GXH WR WKH *HLQGXFHG DLGLQJ HOHFWULF ILHOG (EL UDWKHU WKDQ WKH HQKDQFHPHQW RI GLIIXVLYLW\ 'Q ZKLFK LV QHJOLJLEOH DFFRUGLQJ WR RXU PHDVXUHPHQWV DQG WKHRUHWLFDO FDOFXODWLRQV 7KH EDVH WUDQVLW WLPH UDWLR W%*H6LfW%6Lf IRU WKH UHYHUVH DFWLYH PRGH LV VKRZQ LQ )LJ $V H[SHFWHG IRU WKH *H FRQFHQWUDWLRQ SURILOH LQ )LJ W%5*H6LfW%56Lf } 6XPPDU\ ,Q FRQFOXVLRQ D QHZ /)*9 PHWKRGRORJ\ ZDV GHYHORSHG DQG D FRPSOHWH DQDO\VLV RI WKH HIIHFWV RI *H LQ WKH EDVH RI 6L*HALA6L +%7 ZDV SUHVHQWHG 7KH /)*9 PHWKRGRORJ\ LV EDVHG RQ D RQHGLPHQVLRQDO %7 WKHRU\ DQG DFFXUDWH GF DQG ORZ IUHTXHQF\ PHDVXUHPHQWV RI VLPLODU *H6L DQG 6L WUDQVLVWRUV ,W LV DFFXUDWH H[SHGLHQW DQG
PAGE 69
2 7LO )LJ &ROOHFWRU FXUUHQW UDWLR ,F*H6Lf,F6Lf YV U_ M QRUPDOL]HG WR 3%*H6Lf'%6Lf@}H[SD$\ff )RU UM !a WKH FROOHFWRU FXUUHQW UDWLR ,F*H6Lf,F6Lf FDQ EH PRGHOHG E\ D OLQHDU IXQFWLRQ ,F*H6Lf,F6Lf RF $(**HN7f&
PAGE 70
7LO )LJ 2XWSXW &RQGXFWDQFH UDWLR J*H6LfJ6Lf YV U M IRU WKH IRUZDUGDFWLYH PRGH QRUPDOL]HG WR >'%*H6Lf'%6Lf@}H[SD$\ff 7KH J UDWLR VKRZV D UDSLG GHFUHDVH IRU Uf DQG WHQGV WR ]HUR IRU YHU\ ODUJH UM 7KLV PHDQV WKDW WKH RXWSXW UHVLVWDQFH RI *H6L +%7 LV PXFK KLJKHU WKDQ WKDW RI 6L % 7 IRU ODUJH U? W DQG WKH (DUO\ YROWDJH 9$ DOVR LQFUHDVHV UDSLGO\ ZLWK Uf M 7KXV WKH ILJXUHRIPHULW 9$ LPSURYHV VLJQLILFDQWO\ IRU ODUJH U_ M IRU WKH *H6L +%7
PAGE 71
)LJ %DVH PLQRULW\ FDUULHU GLVWULEXWLRQ RI *H6L )%7 DQG 6L %7 YV [;% IRU U_ LQ IRUZDUG DFWLYH PRGH QRUPDOL]HG WR Q}>H[ST9%(fO@1%f
PAGE 72
7LO )LJ %DVH PLQRULW\ FDUULHU FKDUJH UDWLR 4%*H6Lf4%6Lf YV Uf M IRU WKH IRUZDUG DFWLYH PRGH QRUPDOL]HG WR H[SD$\ff
PAGE 73
f+L )LJ %DVH PLQRULW\ FDUULHU WUDQVLW WLPH UDWLR W%*H6LfW%6Lf LQ IRUZDUG DFWLYH PRGH YV UM M QRUPDOL]HG WR '%*H6Lf'%6Lff 7KH EDVH WUDQVLW WLPH FDQ JLYH D IDFWRU RI WZR LPSURYHPHQW IRU A ZKLFK FRUUHVSRQGV WR DERXW b *H LQ WKH EDVH
PAGE 74
L L ,fÂ§WfÂ§U77O L L L L L L L 7LO )LJ %DVH PLQRULW\ FDUULHU WUDQVLW WLPH UDWLR W%*H6LfW%6Lf LQ UHYHUVH DFWLYH PRGH YV U_ M QRUPDOL]HG WR '%*H6Lf'%6Lff 7KH UHYHUVH EDVH WUDQVLW WLPH LV PXFK ODUJHU WKDQ LQ WKH IRUZDUG DFWLYH PRGH
PAGE 75
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f LV DERXW O2PH9 ODUJHU LQ *H6L +%7 WKDQ LQ 6L %7 DW URRP WHPSHUDWXUH )LUVW DFFXUDWH PHDVXUHPHQWV RI WKH PLQRULW\ FDUULHU GLIIXVLRQ OHQJWK DQG UHFRPELQDWLRQ OLIHWLPH LQ WKH EDVH VKRZ JUHDWO\ LPSURYHG YDOXHV LQ WKH *H6L EDVH FRPSDUHG WR 6L 7KH EDVH FXUUHQW ZDV IRXQG WR EH GRPLQDWHG DOPRVW FRPSOHWHO\ E\ WKH UHFRPELQDWLRQ ZLWKLQ WKH HPLWWHU LQ WKH ODUJH [SP WUDQVLVWRUV ZKLOH LQ FLUFXLW WUDQVLVWRUV ZLWK D QDUURZ HPLWWHU VWULSH OSPf WKH EDVH UHFRPELQDWLRQ LV PRUH LPSRUWDQW GXH WR LQFUHDVHG LPSRUWDQFH RI WKH VLGHZDOO DQG VXUIDFH HIIHFWV 6LQFH WKH GF DQG ORZIUHTXHQF\ VPDOO VLJQDO PHDVXUHPHQWV DUH LQGHSHQGHQW RI SDUDVLWLFV DFFXUDWH KLJKIUHTXHQF\ DQG KLJKn VSHHG SDUDPHWHUV FDQ EH UHDGLO\ REWDLQHG ZKLFK DUH GLIILFXOW WR PHDVXUH GLUHFWO\ VXFK DV WKH aOSV EDVH WUDQVLW WLPH 7KH SHUIRUPDQFH LPSURYHPHQWV RI *H6L +%7 RYHU 6L %7 ZHUH VKRZQ JUDSKLFDOO\ DV D IXQFWLRQ RI U_M
PAGE 76
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
PAGE 77
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f (* *Hf$(**H[;5f >@ ZKHUH (*f LV WKH HQHUJ\ JDS PDJQLWXGH DW WKH (% MXQFWLRQ HGJH RI WKH TXDVLQHXWUDO EDVH [ ;% LV WKH EDVH WKLFNQHVV DQG $(*B*H $(**H;%f$(*B*Hf LV WKH WRWDO YDULDWLRQ RI WKH HQHUJ\ JDS GXH WR WKH *H JUDGLHQW 7KLV JLYHV D VSDWLDOO\FRQVWDQW DLGLQJ HOHFWULF ILHOG IRU WKH HOHFWURQV LQ WKH SW\SH EDVH LQ WKH OLQHDUO\JUDGHG UHJLRQ (EL[f OTf$(*;5f N7TfUL;5f LQGHSHQGHQW RI [ ZKHUH U? LV WKH QRUPDOL]HG HQHUJ\ JDS GURS WKURXJK WKH EDVH U_ $(*B*HN7 7KH FROOHFWRU FXUUHQW FDQ EH H[SUHVVHG IRU DQ\ EDVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH DQG FRPSRVLWLRQDO YDULDWLRQ LQ WKH *H6L EDVH DV > S @ TrH[ST9%(N7f F f ;% 3[f G[ 'Q[f Q[f
PAGE 78
T'% H[ST9%(Nf f 3[f G[ QL[f ZKHUH '% 3[fQ[f3[f'Q[fQ"[fG[ LV WKH DYHUDJH PLQRULW\ HOHFWURQ GLIIXVLYLW\ LQ WKH EDVH ,I ZH VXSHULPSRVH D VPDOO VLQXVRLGDO VLJQDO RQ WKH GF UHYHUVH %& MXQFWLRQ ELDV ZLWK WKH %( MXQFWLRQ IRUZDUGELDVHG WKH VPDOOVLJQDO RXWSXW FRQGXFWDQFH JR) GXH WR WKH EDVHWKLFNQHVV PRGXODWLRQ FDQ EH REWDLQHG E\ WDNLQJ WKH GHULYDWLYH RI F RU ,F ZLWK UHVSHFW WR 9&% DLU D;5 ,U 3;%f f D[% R) f D[% DY&% % 3;f G[ QL;% DY &% QA&[f ,Q f 3;%fG;%f9&% OTf4%9&%f LV WKH VPDOOVLJQDO EDVHFROOHFWRU MXQFWLRQ FDSDFLWDQFH SHU XQLW DUHD &%& 7KXV f FDQ EH ZULWWHQ DV b &%& R) 3;f G[ QL ;%f f QL[f 6LPLODUO\ ZH FDQ REWDLQ JR5 IRU WKH UHYHUVHDFWLYH RSHUDWLRQ ZLWK WKH (% MXQFWLRQ UHYHUVHELDVHG DQG WKH %& MXQFWLRQ IRUZDUGELDVHG &5 %( R5 3[f QLf G[ f QL[f 7KH LQWHJUDO LQ WKH GHQRPLQDWRU LV WKH VDPH ERWK IRU WKH IRUZDUGDFWLYH DQG UHYHUVH DFWLYH RSHUDWLRQ PRGHV 7KXV WKH UDWLR RI HTXDWLRQV f DQG f JLYHV
PAGE 79
R5 F5 QLA;%f &%( f ) &) QLf &%& )URP f ZH REWDLQ WKH UDWLR RI WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQV DW WKH WZR HGJHV RI WKH TXDVLQHXWUDO EDVH OD\HU ZKLFK LV JLYHQ EHORZ LQ f QL;%A b 5 &%& f Df Q"f &5 I FEH 1F;Ef1Y;Ef H[S$(*BJHN7f Ef 1Ff1Yf 7KH UHVXOW LQ HTXDWLRQ f LV FRPSOHWHO\ JHQHUDO DQG DSSOLHV WR DQ\ +%7 RU %7 ZLWK DQ\ FRPSRVLWLRQDO JUDGLQJ DQG GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH LQFOXGLQJ P9 +%7V (TXDWLRQ f XVHV QR DVVXPSWLRQV DERXW WKH PDWHULDO FRPSRVLWLRQ RU LPSXULW\ GRSLQJ LQ WKH EDVH DQG JLYHV WKH UDWLR RI WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQV DW WKH WZR HGJHV RI WKH TXDVLQHXWUDO EDVH OD\HU GLUHFWO\ IURP WKH PHDVXUHG FROOHFWRU FXUUHQW RXWSXW FRQGXFWDQFH DQG MXQFWLRQ FDSDFLWDQFHV IRU DQ\ +%7 RU %7 7KH DSSUR[LPDWLRQ XVLQJ WKH DYHUDJH EDVH GLIIXVLYLW\ LQ f GRHV QRW DIIHFW WKH UDWLR LQ f VLQFH WKH LQWHJUDO UHODWLRQ IRU '% LV WKH VDPH IRU WKH IRUZDUGDFWLYH DQG UHYHUVHDFWLYH RSHUDWLRQ PRGHV 2QH SRVVLEOH DSSOLFDWLRQ RI WKH UHVXOW LQ f LV PHDVXUHPHQW RI WKH HQHUJ\ JDS LQ *H6L DV D IXQFWLRQ RI *H FRQFHQWUDWLRQ XVLQJ WHVW VWUXFWXUHV ZLWK GLIIHUHQW *H OLQHDUO\ JUDGHG SURILOHV DV VKRZQ LQ )LJ (TXDWLRQ f FDQ EH DOVR XVHG WR PHDVXUH WKH LQWULQVLF FDUULHU FRQFHQWUDWLRQ UDWLR GXH WR KHDY\GRSLQJLQGXFHG HQHUJ\ JDS UHGXFWLRQ XVLQJ VWUXFWXUHV ZLWK D KLJKORZ GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOHV DV VKRZQ LQ )LJ
PAGE 80
([SHULPHQWV DQG 5HVXOWV $ KLJK SHUIRUPDQFH 6L*HALA6L +%7 LV XVHG WR GHPRQVWUDWH WKH DSSOLFDWLRQ RI WKH UHVXOW LQ HTXDWLRQ f 7KH *H FRQFHQWUDWLRQ SURILOH LV VKRZQ VFKHPDWLFDOO\ LQ )LJ 7KH *H JUDGLHQW IURP 6,06 VKRZHG D OLQHDUO\JUDGHG *H FRQFHQWUDWLRQ IURP DERXW [*H b DW WKH (% MXQFWLRQ WR D SHDN RI DERXW b DW WKH %& MXQFWLRQ 7KH EDVH GRSDQW GHQVLW\ ZDV QHDUO\ FRQVWDQW DW [OLVFPf DQG WKH TXDVLQHXWUDO EDVH WKLFNQHVV ZDV DERXW Âƒ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f DQG DVVXPLQJ WKDW WKH GLIIHUHQFH LQ WKH LQWULQVLF FDUULHU GHQVLW\ DW [ DQG [ ;% LV GXH WR *H JUDGLQJ ZH REWDLQ IURP f Q";%f Q"f H[S"A H[SU" &f H[S >$(*B*HN7f &@ b U FEF f f r&5 R) &%( Df Ef ZKHUH & LV WKH OHDVW VTXDUHV ILWWLQJ SDUDPHWHU IRU WKH GHQVLW\ RI VWDWHV SURGXFW 1&19 IRU OLQHDU *H GLVWULEXWLRQ GLVFXVVHG LQ GHWDLO LQ VHFWLRQ 8VLQJ WKH GDWD LQ 7DEOH ZH REWDLQ QI;%fQIf IURP f 7KLV JLYHV A DQG $(**H PH9 ZKHUH
PAGE 81
&RQFHQWUDWLRQ FP f )LJ *H6L +%7 ZLWK GLIIHUHQW *H FRQFHQWUDWLRQ SURILOHV LQ WKH EDVH UHJLRQ *HUPDQLXP bf
PAGE 82
&RQFHQWUDWLRQ FP f 'HSWK )LJ %LSRODU WUDQVLVWRU ZLWK GLIIHUHQW KLJKORZ GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOHV LQ WKH EDVH UHJLRQ
PAGE 83
& LV XVHG 7KH DLGLQJ HOHFWULF ILHOG (EL N7TfU_;5f f[O7f N9FP 7KLV YDOXH FRPSDUHV YHU\ ZHOO WR (EL N9FP ZKLFK LV REWDLQHG E\ WKH FRPSDUDWLYH PHWKRG GLVFXVVHG LQ FKDSWHU VHFWLRQ f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r $(*B*HN7 ZKHUH N LV WKH %ROW]PDQQ FRQVWDQW DQG $(*B*H LV WKH *HLQGXFHG HQHUJ\ JDS UHGXFWLRQ DFURVV WKH TXDVLQHXWUDO EDVH UHJLRQ 7KXV DW ORZHU WHPSHUDWXUH WKH EXLOWLQ HOHFWULF ILHOG LV JUHDWO\ HQKDQFHG $OO RWKHU SDUDPHWHUV UHODWHG WR (EL ZLOO EH DOVR HQKDQFHG )RU H[DPSOH WKH FROOHFWRU FXUUHQW ,F*H6Lf ZLOO LQFUHDVH WKH RXWSXW FRQGXFWDQFH J*H6Lf ZLOO GHFUHDVH DQG WKH EDVH PLQRULW\ FDUULHU WUDQVLW WLPH W% ZLOO GHFUHDVH VLJQLILFDQWO\ ZKLFK JLYHV KLJKHU VSHHG DQG I7 7KHRUHWLFDO FDOFXODWLRQV RI WKH PLQRULW\ PRELOLW\ DV D IXQFWLRQ RI WHPSHUDWXUH DQG GRSLQJ ZHUH SUHVHQWHG LQ WKH OLWHUDWXUH > @ ,Q RUGHU WR YHULI\ WKHVH WKHRUHWLFDO SUHGLFWLRQV DQG DQDO\]H WKH *H6L +%7 SHUIRUPDQFH DW ORZHU WHPSHUDWXUH WKH GHWHUPLQDWLRQ RI WKH WHPSHUDWXUH GHSHQGHQFH RI Uf M LV YHU\ LPSRUWDQW 7KLV ZDV LQYHVWLJDWHG E\ PHDVXULQJ WKH *H6L +%7 FKDUDFWHULVWLFV DW VHYHUDO GLIIHUHQW ORZ WHPSHUDWXUHV 8VLQJ WKH PRGHOV DQG HTXDWLRQV JLYHQ LQ WKH VHFWLRQ ZH FDQ
PAGE 84
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f DQG WKH EDVH UHFRPELQDWLRQ FRPSRQHQW 4%[%*H6Lf,% YV WHPSHUDWXUH UHVSHFWLYHO\ )URP )LJ ZH FDQ VHH WKDW U_ M REWDLQHG E\ D FRPELQDWLRQ RI f DQG f DQG WKXV WKH EXLOWLQ HOHFWULF ILHOG (EL LQFUHDVH DOPRVW OLQHDUO\ ZLWK 7 7KH GDVKHG OLQH LV FDOFXODWHG XVLQJ TS$(A4JN7 & ZKHUH $(*B*H PH9 DV PHDVXUHG DW DQG & 7KH GHYLDWLRQ RI WKH WKHRUHWLFDO FXUYH IURP WKH H[SHULPHQWDO GHSHQGHQFH PD\ FRPH IURP WZR VRXUFHV f GHFUHDVH RI WKH $(*B*H ZLWK GHFUHDVLQJ WHPSHUDWXUH DQG f WKH LQFUHDVH RI WKH ILWWLQJ SDUDPHWHU & ZLWK GHFUHDVLQJ 7 7KH ODWWHU RQH LV PRUH UHDVRQDEOH VLQFH WKH 1&19 SURGXFW LV D VWURQJ IXQFWLRQ RI WHPSHUDWXUH DV LV VKRZQ LQ HTXDWLRQV Df DQG Ef 7KH LQFUHDVH RI & DW ORZ WHPSHUDWXUH PHDQV WKDW WKH 1F1Y*H6Lf1F1Y6Lf LV D GHFUHDVLQJ IXQFWLRQ RI WHPSHUDWXUH )LJXUH VKRZV WKH PHDVXUHG GLIIXVLRQ OHQJWK /%*H6Lf YV 7 7KH GLIIXVLRQ OHQJWK FDQ EH H[SUHVVHG DV /%*H6L f O 'U ?N7TfAU f
PAGE 85
6LQFH WKH S[ SURGXFW LV QRW VWURQJO\ WHPSHUDWXUH GHSHQGHQW >@ /%*H6Lfr:7 RU LQFUHDVLQJ ZLWK LQFUHDVLQJ 7 DV VKRZQ LQ )LJ )URP )LJ ZH FDQ VHH WKDW WKH EDVH UHFRPELQDWLRQ FRPSRQHQW 4%[%f*H6L,% LQFUHDVHV ZLWK GHFUHDVLQJ WHPSHUDWXUH 7KXV WKH EDVH UHFRPELQDWLRQ EHFRPHV PRUH LPSRUWDQW DV WHPSHUDWXUH GHFUHDVHV 7DEOH 6XPPDU\ RI PHDVXUHG SDUDPHWHUV RI *H6L +%7 iR) [O6f iR5 [O6f ,F) S$f FU S$f &EF S)f &EH S)f
PAGE 86
7DEOH 7HPSHUDWXUH GHSHQGHQFH RI PHDVXUHG SDUDPHWHUV RI *H6L +%7 7.f 6R5A6R) A&5A&) &EFS)f &%(3)f 1RWH $( [cDP 7 9%( 9 9%& 9 7KH MXQFWLRQ FDSDFLWDQFHV &%& DQG &%( ZHUH PHDVXUHG DW I ON+] 7DEOH 7HPSHUDWXUH GHSHQGHQFH RI *H6L +%7 SDUDPHWHUV FDOFXODWHG IURP WKH PHDVXUHG GDWD LQ 7DEOH 7.f KL SPf >4EAE,AE b b b b b b b
PAGE 87
2 7 .f )LJ 1RUPDOL]HG HQHUJ\ JDS UHGXFWLRQ SDUDPHWHU UM YV 7
PAGE 88
7HPSHUDWXUH .f )LJ 'LIIXVLRQ OHQJWK /%*H6Lf 7
PAGE 89
)LJ %DVH UHFRPELQDWLRQ FRPSRQHQW 4%*H6LfW%*H6Lf YV 7
PAGE 90
/)*9 $QDO\VLV RI *H6L +%7 ZLWK QR 6L 5HIHUHQFH $IWHU WKH GHWHUPLQDWLRQ RI Uf GHVFULEHG LQ WKH SUHYLRXV VHFWLRQ D FRPSOHWH FKDUJH FRQWURO DQDO\VLV RI WKH *H6L +%7 FDQ EH PDGH 7KH RQO\ RWKHU SDUDPHWHUV QHHGHG IRU WKLV DQDO\VLV DUH U? ZKLFK GHWHUPLQHV WKH EXLOWLQ HOHFWULF ILHOG LQ WKH UHJLRQ RI ;5[;% DQG WKH GLIIXVLYLW\ '%*H6Lf LQ WKH EDVH UHJLRQ $V GHULYHG LQ VHFWLRQ HTXDWLRQ f LV XVHG WR FDOFXODWH WKH SDUDPHWHU U? )URP WKH PHDVXUHPHQW RI WKH RXWSXW FRQGXFWDQFH J DQG UHYHUVH WUDQVFRQGXFWDQFHV JU IRU WKH IRUZDUG VXEVFULSW )f DQG UHYHUVH VXEVFULSW 5f PRGHV RI RSHUDWLRQV DQG WDNLQJ WKH UDWLR RI HTXDWLRQ f IRU WKHVH WZR PRGHV IU W%) f U)U5 %U W%5 7KH UDWLR LQ f LV D IXQFWLRQ RI UM M $ ;5;% DQG U? RQO\ 6LQFH UMM FDQ EH GHWHUPLQHG LQGHSHQGHQWO\ DV GHVFULEHG LQ VHFWLRQ DQG ;5;% LV NQRZQ IURP 6,06 RU HVWLPDWHG IURP IDEULFDWLRQ FRQGLWLRQV U? FDQ EH WKHQ FDOFXODWHG IURP f 8VLQJ WKH GDWD LQ 7DEOH DQG Ac ZH REWDLQHG Uf ZKLFK DJUHHV YHU\ ZHOO ZLWK WKH FRPSDUDWLYH PHWKRG LQ VHFWLRQ ZKLFK JLYHV U? :H DVVXPH WKDW '%*H6Lf '%6Lf DV GHWHUPLQHG LQ FKDSWHU 7KH UHVXOWV RI D FRPSOHWH FKDUJH FRQWURO DQDO\VLV IRU WKH *H6L +%7 IROORZLQJ WKH SURFHGXUH GHVFULEHG LQ FKDSWHU DUH VXPPDUL]HG LQ WDEOH 6LQFH UfM DQG UL REWDLQHG XVLQJ WKH QRQUHIHUHQFH DQDO\VLV DUH DOPRVW WKH VDPH DV WKRVH REWDLQHG XVLQJ WKH FRPSDUDWLYH /)*9 PHWKRGRORJ\ LQ VHFWLRQ UfM DQG Uf f WKH UHVXOWV LQ WDEOH DUH YHU\ VLPLODU WR WKRVH LQ WDEOH
PAGE 91
7DEOH 6XPPDU\ RI H[SHULPHQWDO GDWD DQG FDOFXODWHG SDUDPHWHUV IRU 6L*H[6LMB[6L +%7 E\ /)*9 PHWKRGRORJ\ ZLWK QR 6L UHIHUHQFH %7 7KH WRS VL[ OLQHV JLYH WKH PHDVXUHG GF DQG %70 DF GDWD REWDLQHG DW r& 9%( 9 DQG 9&% ,9 7KH FRPSXWHG SDUDPHWHUV DUH VKRZQ LQ WKH QH[W VL[WHHQ OLQHV $%( [OSP ;5 Âƒ ;% Âƒ 3DUDPHWHU *H6L +%7 ,F 3$f ,% Q$f I J P [O JU O4f [OaQ JRAJU 7LO $(4B*H PH9f (EL N9FPf % /E SPf 4EAE WW$f 4HWH Q$f 4H[Hf4EWEf 4HWHf,E bf 4EAEf$E bff 4E &f [Of 'E FP9Vf W% 36f WE QVf 9$ 9f
PAGE 92
6XPPDU\ ,Q FRQFOXVLRQ D VLPSOH DQG XQLTXH PHWKRGRORJ\ LV SUHVHQWHG IRU WKH GLUHFW PHDVXUHPHQW RI WKH UDWLR RI WKH LQWULQVLF FDUULHU GHQVLWLHV DW WKH WZR HGJHV RI WKH TXDVLQHXWUDO EDVH OD\HU LQ +%7 RU %7 7KH PHWKRG LV EDVHG RQ D FRPELQDWLRQ RI GF DQG VPDOOVLJQDO ORZIUHTXHQF\ aON+]f PHDVXUHPHQW RI WKH WUDQVLVWRU FKDUDFWHULVWLFV 7KH QRUPDOL]HG HQHUJ\ JDS SDUDPHWHU UA WKH HQHUJ\ JDS UHGXFWLRQ $(*B*H DQG WKH EXLOWLQ HOHFWULF ILHOG (EL DUH GHWHUPLQHG LQGHSHQGHQWO\ RI WKH EDVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH PLQRULW\ FDUULHU PRELOLW\ DQG KHDY\GRSLQJ HIIHFWV LQ *H6L +%7 ZLWK D OLQHDU RU WUDSH]RLGDO *H FRQFHQWUDWLRQ SURILOH 7KHQ D FRPSOHWH FKDUJH FRQWURO DQDO\VLV RI *H6L +%7 LV PDGH ZLWKRXW XVLQJ D 6L UHIHUHQFH %7 7KH WHPSHUDWXUH GHSHQGHQFH RI UM c GLIIXVLRQ OHQJWK /%*H6Lf DQG EDVH UHFRPELQDWLRQ FRPSRQHQW 4%A%A*H6L ZHUH PHDVXUHG WR LQYHVWLJDWH WKH SHUIRUPDQFH RI *H6L +%7 DV D IXQFWLRQ RI WHPSHUDWXUH
PAGE 93
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f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
PAGE 94
*H6L +%7 0RGHOLQJ DQG /)*9 0HWKRGRORJ\ 7KH /)*9 PHWKRGRORJ\ LV DSSOLFDEOH WR DQ DQDO\VLV RI DQ\ KHWHURMXQFWLRQ ELSRODU WUDQVLVWRU +%7f LQFOXGLQJ WKRVH EDVHG RQ ,,,9 PDWHULDOV 7KH /)*9 PHWKRGRORJ\ XVHV D FRPELQDWLRQ RI WKH GF DQG VPDOOVLJQDO PHDVXUHPHQWV RI WKH WUDQVLVWRU FKDUDFWHULVWLFV 7KH GF FKDUDFWHULVWLFV DUH WKH FROOHFWRU DQG EDVH FXUUHQW DV D IXQFWLRQ RI WKH HPLWWHUEDVH MXQFWLRQ ELDV 7KH VPDOOVLJQDO FKDUDFWHULVWLFV DUH WKH RXWSXW FRQGXFWDQFH DQG UHYHUVH WUDQVFRQGXFWDQFH UHVXOWLQJ IURP WKH EDVH WKLFNQHVV PRGXODWLRQ 7KH\ DUH PHDVXUHG DW ORZ IUHTXHQF\ aON+]f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
PAGE 95
$33(1',; $ *(1(5$/ (48$7,216 )25 %7 $ 7KH JHQHUDO LQWHJUDO UHODWLRQV IRU FDOFXODWLRQ RI LPSRUWDQW SDUDPHWHUV LQ ELSRODU MXQFWLRQ WUDQVLVWRU DUH JLYHQ EHORZ (TXDWLRQV IRU D WUDSH]RLGDO *H GLVWULEXWLRQ LQ WKH EDVH DQG FRQVWDQWH[SRQHQWLDO GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH LQ WKH EDVH DUH JLYHQ LQ VHFWLRQ % &ROOHFWRU FXUUHQW ,F*H6Lf > S @ TH[ST9%(N7f ,U ;U 3[f G[ 'Q[f QL[f 0LQRULW\ FDUULHU GLVWULEXWLRQ LQ WKH EDVH >@ F QI[f Q[f T 3[f ;% 3\f [ 'QQL\f G\ 0LQRULW\ FDUULHU FKDUJH LQ WKH EDVH $ f $f n% Q[fG[ 0LQRULW\ FDUULHU EDVH WUDQVLW WLPH >@ W% a ;% QI[f 3[f 3\f &RUUHFWLRQ IDFWRU % >@ G[ % 'QW% ;% 3;f Q[f G[ [ Q"\f ;% Q"[f G[ 3[f $f $f $f
PAGE 96
2XWSXW FRQGXFWDQFH J ,& G;J R D[% DY &% ;% 3[f G[ QL[f 5HYHUVH WUDQVFRQGXFWDQFH JU D 4E WJf ;% ;U DY &% O 3;%f D[% f QL;%f DY&% $f $f &XUUHQW JDLQ T %2 3;f 'Q[fQL[f G[ ZKHUH ,% LV WKH EDVH VDWXUDWLRQ FXUUHQW (DUO\ YROWDJH 9$ DQG 9$ SURGXFW 9D TQL;%f'Q;%f R %& 3[f 'Q[f f QL[f G[ ERrFEF Q";%f'Q;Jf $f $f $f % (TXDWLRQV IRU D WUDSH]RLGDO *H GLVWULEXWLRQ DQG FRQVWDQWH[SRQHQWLDO GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH LQ WKH EDVH RI *H6L +%7 )LJ f REWDLQHG IURP WKH JHQHUDO LQWHJUDO UHODWLRQ $Of$,2f DUH JLYHQ EHORZ
PAGE 97
&ROOHFWRU FXUUHQW ,F T'QQO f H[SDfH[S YfH[Sf f H[ST9%(N7f MF 1%R;%>;5;%f H[SUMLf ;U;Ef OH[SAf fM@ 0LQRULW\ FDUULHU EDVH WUDQVLW WLPH W% ;E W% < 'Q 2XWSXW FRQGXFWDQFH J ;% $H[SU"OfL"fH[SA O$.H[SAOfA 5HYHUVH WUDQVFRQGXFWDQFH JU 4E ; ;U DY &% D[% U r fÂ§ r r WE;E < $>H[SAOfAf@H[SQ O$f>H[SMOfII@ D9&% ,Q $ f DQG $ f WKH WHUPV ; DQG < DUH JLYHQ EHORZ ; $H[SAOfQ $O$f>H[S+LOf"@ f >H[S"OfU"@ O$fH[SULULOfUL < $>H[SfOf@U $ $f>H[S&AMBf@ f >OH[SU`fOUcLQ] O$f>H[SULfULO@LL &RUUHFWLRQ IDFWRU % b % ;< 39$ SURGXFW 9D ;% f H[SAO H[SQO $ H[SUMJ $f nOO Â%2 r &%& f QLA 'Q;%f f H[SU"L D[%DY&% $f $f $f $f $f $f $f $f $f
PAGE 98
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f VKRZQ LQ )LJ %O KDV D SHDN YDOXH DW [ 7KH SURILOH VKRZQ LQ )LJ % KDV D SHDN YDOXH LQVLGH WKH EDVH OD\HU 7KH GDVKHG OLQH VKRZV WKH H[SRQHQWLDO DSSUR[LPDWLRQ IRU 3[f )ROORZLQJ DUH WKH DQDO\WLFDO GHULYDWLRQV RI WKH HTXDWLRQV IRU WKH WZR *DXVVLDQ GLVWULEXWLRQV 3[f %DVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH 3[f VKRZQ LQ )LJ %O 3[f 10H[S[Df ZKHUH D LV %Of D ;%>OQ>103;%f@ ; %f %f
PAGE 99
;Âƒf )LJ %O *DXVVLDQ GLVWULEXWLRQ DQG H[SRQHQWLDO DSSUR[LPDWLRQ
PAGE 100
&ROOHFWRU FXUUHQW T'QQc >H[ST9%(N7fO@ ,F %f Â AÂ‘FU1$$HUI;%FUf 0LQRULW\ FDUULHU EDVH WUDQVLW WLPH ? Q}D W% >HUI;%FUf %H[S[UfG[ %H[S[FUfHUI[FUf@G[ %f 'Q 2XWSXW FRQGXFWDQFH ,& H[S;EFWf D;% S %f L HUI;%FUf D9&% 5HYHUVH WUDQVFRQGXFWDQFH ,F H[S;JFUf D;% JU f >%H[S[FUfHUI[FUfG[@ %f 'QW% HUI;%FUf D9&% &RUUHFWLRQ IDFWRU %H[S[DfHUI[FUfG[ % %f HUI;%FUf M%H[S[fG[ M%H[S[fHUI[RfG[ )RU WKH SURILOH LQ )LJ %O WKH YDOXH IRU WKH FRUUHFWLRQ IDFWRU LV REWDLQHG QXPHULFDOO\ XVLQJ %f ZKLFK JLYHV % 7KH H[SRQHQWLDO DSSUR[LPDWLRQ JLYHV % XVLQJ PXFK VLPSOHU DQDO\WLFDO H[SUHVVLRQ $ f 7KXV WKH H[SRQHQWLDO DSSUR[LPDWLRQ RI WKH *DXVVLDQ SURILOH LQ WKH EDVH RI D ELSRODU MXQFWLRQ WUDQVLVWRU JLYHV RQO\ D VPDOO HUURU IRU WKH YDOXH RI % 7KLV DOORZV D VLPSOH FKDUJH FRQWURO DQDO\VLV RI ELSRODU MXQFWLRQ WUDQVLVWRU XVLQJ WKH H[SRQHQWLDO SURILOH DSSUR[LPDWLRQ VKRZQ LQ VHFWLRQV IRU
PAGE 101
WUDQVLVWRUV ZLWK D *DXVVLDQ RU *DXVVLDQOLNH SURILOH LQ WKH EDVH ZLWK D VPDOO HUURU FRPSDUHG ZLWK DFFXUDWH QXPHULFDO VROXWLRQ ,, )ROORZLQJ DUH WKH HTXDWLRQV IRU WKH *DXVVLDQ GLVWULEXWLRQ 3[f VKRZQ LQ )LJ % %DVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH 3[f 3[f 10H[S>[;fFU@ ZKHUH [ LV WKH SHDN YDOXH SRVLWLRQ RI WKH 3[f DQG FU LV HTXDO WR ;;Qf &7 OQ>103;%f@ ;% U 3[fG[ R A WWFW1DD^HUI>;%;fR@ HUI;FUf` %f %f %f &ROOHFWRU FXUUHQW T'A >H[ST9%(N7f @ ,F WW] % f OAU1$$^HUI>;%;fF@ HUI;FUf` 0LQRULW\ EDVH WUDQVLW WLPH L QFU W% ^HUI>;%;fM@% H[S>[;fFM@G[ 'E % H[S>[;f&7@HUI>[;fFU@G[` %f &RUUHFWLRQ IDFWRU HUI;FMf M%H[S>[;fFU@G[ %H[S>[;fFU@HUI>[;fR@G[ % % f HUI>;%;fFU@M%H[S>[;f7@G[ M%H[S>[;fR@HUI>[;fFU@G[ 2XWSXW FRQGXFWDQFH ,& H[S>;%;fFU@ ;% J fÂ§ f f 7FU HUI>;%;fFU@ HUI;Uf D9&% %f
PAGE 102
5HYHUVH WUDQVFRQGXFWDQFH ,F H[S;%;fUf U 'Q7% HUI ;%fÂ§;fHUf HUI;FUf ^HUI ;JFUf M%H[S>[;fFU@G[ M%H[S>[;fRf@HUI>[;FMf@G[` %f
PAGE 103
)LJ % *DXVVLDQ GLVWULEXWLRQ DQG LWV H[SRQHQWLDO DSSUR[LPDWLRQ
PAGE 104
$ % & %7 &3 %& %( 'L 'f (EL (EL $( $( **H *+' Â‘ 1 L7 i 6U *H K +%7 LE E ,F N OE $33(1',; & /,67 2) 6<0%2/6 ;5;% &RUUHFWLRQ IDFWRU /HDVW VTXDUHV ILWWLQJ SDUDPHWHU IRU 1&19 SURGXFW %LSRODU MXQFWLRQ WUDQVLVWRU %DVH FROOHFWRU MXQFWLRQ FDSDFLWDQFH %DVH HPLWWHU MXQFWLRQ FDSDFLWDQFH $YHUDJHG EDVH PLQRULW\ FDUULHU GLIIXVLRQ FRHIILFLHQW %DVH PLQRULW\ HDUQHU GLIIXVLRQ FRHIILFLHQW %XLOWLQ HOHFWULF ILHOG LQ WKH EDVH UHJLRQ GXH WR *H %XLOWLQ HOHFWULF ILHOG LQ WKH EDVH UHJLRQ GXH WR 3[f &RQGXFWLRQ EDQG 9DOHQFH EDQG (QHUJ\ JDS (QHUJ\ JDS UHGXFWLRQ LQ WKH EDVH UHJLRQ GXH WR *H (QHUJ\ JDS UHGXFWLRQ LQ WKH EDVH UHJLRQ GXH WR KHDY\ GRSLQJ (OHFWURQ TXDVL)HUPL OHYHO +ROH TXDVL)HUPL OHYHO &XWRII IUHTXHQF\ 2XWSXW FRQGXFWDQFH 5HYHUVH WUDQVFRQGXFWDQFH *HUPDQLXP 3ODQFN FRQVWDQW +HWHURMXQFWLRQ ELSRODU WUDQVLVWRU 6PDOOVLJQDO EDVH FXUUHQW GXH WR YFE 6PDOOVLJQDO FROOHFWRU FXUUHQW GXH WR YFE %DVH FXUUHQW &ROOHFWRU FXUUHQW %ROW]PDQQ FRQVWDQW 'LIIXVLRQ OHQJWK
PAGE 105
PH (OHFWURQ HIIHFWLYH PDVV PK +ROH HIIHFWLYH PDVV Q[f %DVH PLQRULW\ FDUULHU GLVWULEXWLRQ QÂ ,QWULQVLF FDUULHU FRQFHQWUDWLRQ 1F (IIHFWLYH GHQVLW\ RI VWDWH LQ FRQGXFWLRQ EDQG 1Y (IIHFWLYH GHQVLW\ RI VWDWH LQ YDOHQFH EDQG 3[f %DVH GRSDQW LPSXULW\ FRQFHQWUDWLRQ SURILOH 4% %DVH PLQRULW\ FDUULHU HOHFWULF FKDUJH 4H (PLWWHU PLQRULW\ FDUULHU HOHFWULF FKDUJH 6L 6LOLFRQ W% %DVH PLQRULW\ FDUULHU WUDQVLW WLPH YFE 6PDOOVLJQDO YROWDJH VXSHULPSRVHG XSRQ 9&% aON+]f 9$ (DUO\ YROWDJH [*H *HUPDQLXP FRQFHQWUDWLRQ LQ WKH EDVH UHJLRQ ;5 7UDQVLWLRQ SRVLWLRQ RI WKH *H GLVWULEXWLRQ LQ WKH EDVH ;% %DVH WKLFNQHVV 9EL %XLOWLQ SRWHQWLDO RI WKH SQ MXQFWLRQ 9%( %DVH HPLWWHU MXQFWLRQ ELDV 9&% %DVH FROOHFWRU MXQFWLRQ ELDV D $(*B*HfN7 QRUPDOL]HG HQHUJ\ JDS UHGXFWLRQ DW [ LQ WKH EDVH UHJLRQf 3) &RPPRQ HPLWWHU FXUUHQW JDLQ < $(*B+'N7 QRUPDOL]HG HQHUJ\ JDS UHGXFWLRQ LQ WKH EDVH UHJLRQ GXH WR KHDY\ GRSLQJf $\ >$(*B+'*H6Lf$(*B+'*H6Lf@N7 Uf 1RUPDOL]HG WR WKHUPDO YROWDJH N7Tf HQHUJ\ JDS UHGXFWLRQ GXH WR *H LQ WKH EDVH UHJLRQ U? )LHOG IDFWRU GXH WR QRQXQLIRUP EDVH GRSLQJ U@M 1RUPDOL]HG HQHUJ\ JDS UHGXFWLRQ LQ WKH EDVH UHJLRQ 7J %DVH PLQRULW\ UHFRPELQDWLRQ OLIHWLPH [( 5HFRPELQDWLRQ WLPH FRQVWDQW LQ WKH HPLWWHU SQ %DVH PLQRULW\ FDUULHU PRELOLW\
PAGE 106
5()(5(1&(6 >@ &7 6DK )XQGDPHQWDOV RI 6ROLG6WDWH (OHFWURQLFV 5LYHU (GJH 1DQG 6LQJDSRUH :RUOG 6FLHQWLILF 1RY 6HH VHFWLRQV RQ SS RQ KLVWRU\ DQG RSHUDWLRQ SULQFLSOH RI *H6L ELSRODU MXQFWLRQ WUDQVLVWRUV >@ :%6KRFNOH\ &LUFXLW HOHPHQW XWLOL]LQJ VHPLFRQGXFWLYH PDWHULDOV 86 3DWHQW )LOHG XQ JUDQWHG 6HS H[SLUHG 6HS \HDU VWDWXH RI OLPLWDWLRQf >@ +HUEHUW .URHPHU +HWHURVWUXFWXUH ELSRODU WUDQVLVWRUV DQG LQWHJUDWHG FLUFXLWV 3URF ,((( YRO 1R SS DQ >@ :3 'XPNH 0 :RRGDO DQG 9/ 5LGHRXW ,%0f *D$V*D$V$O KHWHURMXQFWLRQ WUDQVLVWRU IRU KLJK IUHTXHQF\ RSHUDWLRQ 6ROLG6WDWH (OHFWURQLFV f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
PAGE 107
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
PAGE 108
>@ 5 3HRSOH ,QGLUHFW EDQG JDS RI FRKHUHQWO\ VWUDLQHG *HALA EXON DOOR\V RQ VLOLFRQ VXEVWUDWHV 7KH $PHULFDQ 3K\VLFDO 6RFLHW\ YRO 1R SS XO\ >@ /( .D\ DQG 7: 7DQJ 0RQWH &DUOR FDOFXODWLRQ RI VWUDLQHG DQG XQVWUDLQHG HOHFWURQ PRELOLWLHV LQ *H[6LB[f XVLQJ DQ LPSURYHG LRQL]HGLPSXULW\ PRGHO $SSO 3K\V YRO 1R SS $XJ >@ 6. &KXQ DQG ./ :DQJ (IIHFWLYH PDVV DQG PRELOLW\ RI KROHV LQ VWUDLQHG 6LMB[*H[ OD\HUV RQ f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
PAGE 109
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
PAGE 110
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
PAGE 111
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f KHWHURVWUXFWXUH ,((( 7UDQV (OHFWURQ 'HYLFHV YRO SS 'HF >@ 6 7DNDJL / +R\W 5LP :HOVHU DQG ) *LEERQV (YDOXDWLRQ RI WKH YDOHQFH EDQG GLVFRQWLQXLW\ RI 6L6L MB[*H[6L KHWHURVWUXFWXUHV E\ DSSOLFDWLRQ RI DGPLWWDQFH VSHFWURVFRS\ WR 026 FDSDFLWRUV ,((( 7UDQV (OHFWURQ 'HYLFHV YRO SS )HE >@ = 0DWXWLQRYLF.UVWHOM 9 9HQNDWDUDPDQ ( 3ULQ] & 6WXUP DQG &+ 0DJHH %DVH UHVLVWDQFH DQG HIIHFWLYH EDQGJDS UHGXFWLRQ LQ QSQ 6L6LM[*H[6L +%7V ZLWK KHDY\ EDVH GRSLQJ ,((( 7UDQV (OHFWURQ 'HYLFHV YRO SS 0DU
PAGE 112
>@ 3( &RWWUHOO DQG =3 @ =3 @ &7 .LUN U $ WKHRU\ RI WUDQVLVWRU FXWRII IUHTXHQF\ I7f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
PAGE 113
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
PAGE 114
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
PAGE 115
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
45
The ratio in (2.27) is a function of A=XR/XB and r)2 only. If XR/XB is known from
SIMS or fabrication conditions, r2 can be calculated from (2.27).
2,5 Experiments and Results
High performance GeSi HBT and reference Si BJT are used to demonstrate the
LFGV methodology. SIMS profile of the completed GeSi HBT is shown in Fig. 1.2.
The reference Si BJT is fabricated with the identical fabrication steps except it has no
Ge in the base layer. Measurements of the EB junction capacitance (4.44fF/pm2 for
GeSi and 4.37fF/pm2 for Si), breakdown voltage (5.2V for GeSi and 5.1 for Si), and
base sheet resistance under the emitter (6.2kÂ£2/square for GeSi and 6kÂ£2/square for Si)
confirm the similarity of the Si and GeSi base concentration and thickness, which is
important for a quantitative performance analysis to delineate the effects of Ge. All the
measurements are taken on large transistors with the EB junction area of
100.4x100.8pm2 in order to eliminate the edge and sidewall effects.
Figure 2.3 shows typical Ib_Vbe and IcVBE characteristics of GeSi and Si
transistors. Note the significant increase in Ic and slight decrease in IB for the GeSi
transistor. The measured enhancement in the collector current Ic(GeSi)/Ic(Si) = 3.86.
The insensitivity of Ic(GeSi) to the reverse basecollector voltage up to VBC=5V shows
the absence of parasitic energy barrier at the BC junction that may reduce Ic [4041].
Figure 2.4 gives the measured smallsignal output conductance, g0, reverse
transconductance, gr, for both the forward active mode and reverse active mode, and the
collector current Ic of GeSi HBT, as a function of VBE, showing the expected
exp(qVBE/kT) dependence [42]. Figure 2.5 gives the measured smallsignal output
63
O 2 4 6 8 10
Til
Fig. 2.11 Collector current ratio Ic(GeSi)/Ic(Si) vs r j (normalized to
PB(GeSi)/DB(Si)]exp(a+Ay)). For rj 1>~3.0, the collector current ratio Ic(GeSi)/Ic(Si)
can be modeled by a linear function, Ic(GeSi)/Ic(Si) oc (AEG.Ge/kT)C.
14
VBE/(1V)
Fig. 1.8 Typical IV characteristics of GeSi HBT and Si BJT at VCB=1V. The GeSi
HBT and Si BJT have the same dopant impurity concentration profile and geometry
except the Ge in the base for GeSi HBT. The solid line is GeSi HBT and dashed line is
Si BJT.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS ii
ABSTRACT v
CHAPTERS
1. INTRODUCTION TO GeSi HETEROJUNCTION BIPOLAR TRANSISTOR 1
1.1 Introduction 1
1.2 GeSi HBT Fabrication 3
1.3 Pseudomorphic Growth, Critical Layer and Geinduced Effects 7
1.4 GeSi HBT Characteristics 8
1.5 Monte Carlo Simulation of Mobilities 16
1.6 Energy Band Diagram 19
1.7 NCNV and nf Model for Linear Ge Distribution 24
1.8 Summary 26
2. LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) METHODOLOGY.. 28
2.1 Introduction 28
2.2 Base Thickness Modulation (BTM) 30
2.3 LFGV Methodology 33
2.4 Derivation of LFGV Methodology and Parameter Extraction 41
2.5 Experiments and Results 45
2.6 Sensitivity of tB and rj x to r\2 and XR/XB 55
2.7 GeSi HBT Performance Enhancement 61
2.8 Summary 62
3. LFGV ANALYSIS OF GeSi HBT WITH NO REFERENCE Si BJT 70
3.1 Introduction 70
3.2 Theory 71
3.3 Experiments and Results 74
3.4 LFGV Analysis with no Si Reference BJT 84
3.5 Summary 86
iii
7
1.3 Pseudomorphic Growth, Critical Layer and Geinduced Effects
During the pseudomorphic epitaxial growth of GeSi layer, the lattice mismatch
between GeSi layer and Si substrate was accommodated by biaxial strain in the GeSi
layer. A twodimensional illustration of the pseudomorphic growth of GeSi layer is
shown in Fig. 1.3. As is shown in Fig. 1.3b, under conditions that favor pseudomorphic
growth, the GeSi layer is biaxially strained and its lattice assumes that of the substrate.
On the other hand, lattice defects such as misfit dislocations occur at the GeSi/Si
interface when the GeSi layer becomes relaxed (no strain) as shown in Fig. 1.3c. There
is a critical layer thickness beyond which the layer will relax and create the misfit
dislocations at the GeSi/Si interface. The dislocations may also penetrate the entire
layer. This critical layer thickness can be calculated based on the strain of the GeSi
layer and the thermal dynamics theory [57]. Figure 1.4 shows the theoretical and
experimental critical layer thickness versus the Ge mole fraction [1], As is shown in the
figure, based on the model of People and Bean, for xGe=10% in the Ge^i^ layer, the
critical layer thickness is almost 1 p.m which is enough for high performance GeSi HBT
whose base thickness is usually in the range of <1000A. The Geinduced biaxial strain
in the GeSi layer changes the physical and chemical properties of the layer, such as the
energy band structure [1416], electron and hole effective masses, and electron and hole
mobility pn and pp [1718]. Figure 1.5 shows the constant energy surfaces in the GeSi
layer under biaxial strain [17, 19]. As can be seen from the figure, the two surfaces in
the zdirection (growth direction) are deformed which gives a smaller electron effective
mass mn and also a higher electron concentration in the z direction, while the energy of
Energy Gap /(1eV)
13
Fig. 1.7 Energy gap reduction for strained and unstrained GeSi layer versus Ge mole
fraction in the layer. Adapted from Sah [1, p. 951].
72
qDB exp(qVBE/k)
(3.2)
P(x)
dx
0 ni2(x)
where DB = JP(x)/n2(x)/JP(x)/Dn(x)n?(x)dx is the average minority electron diffusivity
in the base. If we superimpose a small sinusoidal signal on the dc reverse BC junction
bias, with the BE junction forwardbiased, the smallsignal output conductance, goF, due
to the basethickness modulation can be obtained by taking the derivative of Jc or Ic
with respect to VCB,
air
aXR
Ir
1
P(XB) axB
9oF
(3.3)
axB avCB
dx
ni2(XB
av,
CB
0 n^Cx)
In (3.3), P(XB).dXB)/3VCB = (l/q)(0QB/3VCB) is the smallsignal basecollector
junction capacitance per unit area, CBC. Thus, (3.3) can be written as
% CBC
9oF
P(X)
dx
ni (XB)
(3.4)
0 ni2(x)
Similarly, we can obtain goR for the reverseactive operation with the EB junction
reversebiased and the BC junction forwardbiased,
CR
BE
9oR
P(x) ni2(0)
 dx
(3.5)
0 ni2(x)
The integral in the denominator is the same both for the forwardactive and reverse
active operation modes. Thus, the ratio of equations (3.4) and (3.5) gives
tB(GeSi)/tB(Si)
60
Fig. 2.10 Sensitivity of tB(GeSi)/tB(Si) to XR/XB calculated from (A. 12) with r\2 as a
variable parameter for ^^2.0.
APPENDIX B
TRANSISTOR RELATION FOR A GAUSSIAN
DOPANT IMPURITY CONCENTRATION PROFILE IN THE BASE OF BJT
I. Most bipolar junction transistors are fabricated using ion implantation to dope the
base which gives a Gaussian dopant impurity concentration profile. For GeSi HBT, a
Gaussian profile may result from the impurity doping outdiffusion from the epitaxially
deposited base layer during subsequent fabrication. Thus, the relevant LFGV equations
for the Gaussian distribution are important. Two different Gaussian distribution are
considered and shown in Fig. B.l and Fig. B.2. The Gaussian concentration profile P(x)
shown in Fig. B.l has a peak value at x=0. The profile shown in Fig. B.2 has a peak
value inside the base layer. The dashed line shows the exponential approximation for
P(x).
Following are the analytical derivations of the equations for the two Gaussian
distributions P(x).
1. Base dopant impurity concentration profile P(x) shown in Fig. B.l
P(x) = NM.exp(x2/a2)
where a is
(B.l)
a = XB/[ln[NM/P(XB)]1/2
X.
(B.2)
(B.3)
92
82
100 150 200 250
Temperature /(1K)
Fig. 3.4 Diffusion length LB(GeSi) T.
300
106
[48] P.E. Cottrell, and Z.P. Yu, "Velocity saturation in the collector of Si/Ge^i^/Si
HBTs," EEEE Electron Device Lett., vol. 11, No. 10, pp. 431433, Oct. 1990.
[49] Z.P. Yu, P.E. Cottrell, and R.W. Dutton, "Modeling and simulation of highlevel
injection behavior in double heterojunction bipolar transistors," IEEE 1990
Bipolar Circuits and Technology Meetings, No. 8.5, pp. 192194.
[50] C.T. Kirk, Jr., "A theory of transistor cutoff frequency (fT) falloff at high current
densities," IRE Trans. Electron Devices, pp. 164174, Mar. 1962.
[51] E.J. Prinz, and J. C. Sturm, "Current gainEarly voltage products in heterojunction
bipolar transistors with nonuniform base bandgaps," IEEE Electron Device Lett.,
vol. 12, No. 12, pp. 661663, 1991.
[52] E.J. Prinz, and J.C. Sturm, "Analytical modeling of current gainEarly voltage
products in Si/Si^Ge^Si Heterojunction Bipolar Transistors," IEDM, vol. 33,
No. 2, pp. 853856, 1991.
[53]T. Manku and A. Nathan, "Effective mass for strained ptype Si^Ge^.," J. Appl.
Phys., vol. 69, pp. 84148416, June 1991.
[54]I.Yun Leu and A. Neugroschel, "Minoritycarrier transport parameters in heavily
doped ptype silicon at 296 and 77K," IEEE Trans. Electron Devices, vol. 40, pp.
18721875, Oct. 1993.
[55]Chih Hsin Wang, Konstantinos Misiakos, and A. Neugroschel, "Temperature
dependence of minority hole mobility in heavily doped silicon," Appl. Phys. Lett.,
vol. 57, No. 2, pp. 159161, July. 1990.
Concentration /(cm 3)
36
Depth
Fig. 2.2 Idealized model for the Ge and dopant impurity concentration profiles in GeSi
HBT. The dopant impurity concentration profiles for Si BJT (no Ge in the base) are the
same as those in GeSi HBT.
Germanium /(1%)
xml version 1.0 encoding UTF8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchemainstance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E5RFSDSDJ_2E4RW8 INGEST_TIME 20141208T22:07:54Z PACKAGE AA00024487_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
103
[25] A. Neugroschel, "Determination of recombination lifetimes and recombination
currents in pn junction solar cells, diodes, and transistors," IEEE Trans. Electron
Devices, vol. 28, pp. 108115, Jan. 1981.
[26] A. Neugroschel and ChihTang Sah,"Measurement of builtin electric field in
base of Si/Ge^i^/Si HBT with linearlygraded Ge concentration profile,"
Electronics Letters, vol. 32, No. 24, pp. 22802282, Nov. 1996.
[27]A. Neugroschel, C.T. Sah, J.M. Ford, J. Steele, R. Tang, C. Stein, P. Welch, and
J. Watanabe, "Performance comparison analysis of GeSi and Si bipolar
transistors," Electronics Lett., pp. 12391241, June 20, 1996.
[28] D.M. Richey, J.D. Cressler, A.J. Joseph, "Scaling Issues and Ge concentration
profile Optimization in Advanced UHV/CVD SiGe HBTs," IEEE Electron
Device Letters, vol. 44, No. 3, pp. 431440, Mar. 1997.
[29] J.D. Cressler, J.H. Comfort, E.F. Crabb, G.L. Patton, J.M.C. Stork, and J.Y.C.
Sun, "On the profile design and optimization of epitaxial Si and SiGebase
bipolar technology for 77K applicationsPart I: transistor dc design
considerations," IEEE Trans. Electron Devices, vol. 40, pp. 525541.
[30] J.D. Cressler, E.F. Crabb, J.H. Comfort, J.M.C. Stork, and J.Y.C. Sun, "On the
profile design and optimization of epitaxial Si and SiGebase bipolar technology
for 77K applicationsPart II: circuit performance issues, IEEE Trans. Electron
Devices, vol. 40, pp. 542555.
[31] C.A. King, J.L. Hoyt, and J.F. Gibbons, "Bandgap and transport properties of
Sij^Gex by analysis of nearly ideal Si/Si^Gex/Si heterojunction bipolar
transistor, IEEE Trans. Electron Devices, vol. 36, pp. 20932104, Oct. 1989.
[32] M. Schroter, M. Friedrich and HM. Rein, "A Generalized Integral Charge
Control Relation and its Application to Compact Models for SiliconBased
HBTs," IEEE Trans. Electron Devices, vol. 40, No. 11, pp. 20362046, 1993.
39
nf (x)
Nc(x)Nv(x) Nc(0)Nv(0)exp(Cx/XR) (2.13)
where C is a LSF parameter. Then the intrinsic carrier concentration is given by
ni o2. expa. exp y. exp (x/xx) (0
(2.14)
ni02. expa* exp y. exp^/x^. (xR
where nÂ¡Q is the intrinsic carrier concentration in low doped Si given by
ni02 = Nc(0)Nv(0).exp(EGO/kT), (2.14a)
and x1=XB/[(AEG_Ge/kT)C]=XB/ri1. Combining (2.9), (2.11) and (2.14) yields the
ratio of the collector current in GeSi and Si
Ic(GeSi) DB(GeSi) A+(lA)[lexp(r?2)]//72]
= expt^+a+Ay).
(2.15)
Ic(Si) DB(Si) A(exp^1l)/^1+(lA)[lexp(^2)]/rj2
where r) x = rC is the normalized effective energy gap variation through the base layer
that includes the effects of variable Ge concentration and variation of density of states
with xGe. In (2.15), DB(GeSi) = /B [P/n2(x)]dx / JB Pdx/Dn(x)n2(x) is the spatially
averaged electron diffusivity in the base. The weighting factor of DB(GeSi) is nÂ¡2(x)
which increases with x or xGe. Thus, the value of DB(GeSi) is weighted toward the
region close to x=XB. Since Dn(GeSi) is expected to be a very weak function of xGe for
the average xGE<5% [28], defining DB(GeSi) in (2.7) causes only a negligible error. For
a purely linearlygraded Ge concentration profile, XR = XB and A=l, which simplifies
(2.15)to [27]
Ic(GeSi) 7]i DB(GeSi)
= .expa expAy (2.16)
Ic(Si) lexp(fn) Dg(Si)
One reason for lumping the spatial variation of NCNV with the spatial variation of the
energy gap AEG_Ge is that these two parameters always appear together as a product in
11
Fig. 1.5 Constant energy surfaces in the conduction band in GexS!_x under biaxial
strain [17, 19]. The z direction corresponds to the growth direction. The two surfaces
in the zdirection (growth direction) are raised in energy and deformed such that they
give a smaller electron effective mass mn. The four x and y surfaces are lowered in
energy. The overall energy gap is smaller and the electrons preferentially populate the
two valleys in the zdirection in which mn is low. This is also the direction of the
minoritycarrier flow in FEBT.
24
1.7 NCNV and n? Model for Linear Ge Distribution
The Ge in the base region changes not only the energy gap, but also other electronic
material parameters. The effective density of states in the conduction band and valence
band, Nc and Nv, will change because they are a function of the electron and hole
effective masses me and mh, respectively which change with the Geinduced strain and
change of the Ek relationship in the epitaxial GeSi layer [1718]. Nc and Nv are given
by [1]
Nc = 2(2jrmekT/h2)3/2 (1.6a)
Nv = 2(27rmhkT/h2)3/2. (1.6b)
Figure 1.16 shows the theoretical calculations of the normalized density of states
product NCNV versus Ge mole fraction [1722], In order to obtain analytical models for
the GeSi HBT, we approximate the NCNV product by an exponential function least
squares fitted to the theoretical data,
Nc(x)Nv(x) = Nc(0)Nv(0).exp(C.x/XR) (1.7)
where the constant C is a nonlinear least squares fit (NLSF) parameter. The model fits
the theory well for xGe < 10%, with C = 0.70.1, in particular to the theoretical models
of Prinz et al. [2021] and Jain et al. [22] which are also supported by the
photoluminescence data and electrical measurements in Si/Ge^i^/Si HBTs. The
Nc(x)Nv(x) model given by (1.7) enables the decoupling of the diffusivity DB from the
NCNV product, as will be shown later.
The Geinduced energy gap reduction AEG_Ge in the base region will change the
intrinsic carrier concentration nÂ¡ which can be expressed as [1]
CHAPTER 1
INTRODUCTION TO GeSi HETEROJUNCTION BIPOLAR TRANSISTOR
1.1 Introduction
The concept of heterojunction bipolar junction transistor (HBT) was first proposed
by Shockley in 1948 [12]. The purpose was to build bipolar junction transistors with
larger energy gap in the emitter than in the base, which improves the emitter injection
efficiency over that of the homogeneous bipolar junction transistors. The first detailed
analysis of the HBT characteristics and performance was given by Kroemer [3].
Although the advantages of the heterostructures were known, practical applications
were not realized until the 1970s when the epitaxial technology was developed. The
first successful HBT was fabricated in 1972 by IBM [4] using AlGaAs/GaAs
heterostructure to give the larger emitter energy gap. The anticipated superior
performance of the HBT was demonstrated by these and other IIIV compound HBTs.
Until recently, the AlGaAs/GaAs HBTs have been the most successful. This is due to
the excellent lattice matching between the two material which gives a nearly defectfree
heterointerface. But GaAs or other HBTs based on IIIV semiconductor materials are
difficult to integrate with the existing Si integrated circuits due to the large lattice
mismatch. Also IIIV HBTs are difficult to fabricate and thus not cost efficient. GaAs
HBT has been more successful in discrete devices, such as high frequency oscillators
and amplifiers.
1
A
B
C
BJT
CP
BC
"BE
Di
D
Ebi
Ebi2
AE(
AE,
GGe
GHD
N
iT
Â§0
Sr
Ge
h
HBT
ib
!b
Ic
k
lb
APPENDIX C
LIST OF SYMBOLS
=XR/XB
Correction factor
Least squares fitting parameter for NCNV product
Bipolar junction transistor
Base collector junction capacitance
Base emitter junction capacitance
Averaged base minority carrier diffusion coefficient
Base minority earner diffusion coefficient
Builtin electric field in the base region due to Ge
Builtin electric field in the base region due to P(x)
Conduction band
Valence band
Energy gap
Energy gap reduction in the base region due to Ge
Energy gap reduction in the base region due to heavy doping
Electron quasiFermi level
Hole quasiFermi level
Cutoff frequency
Output conductance
Reverse transconductance
Germanium
Planck constant
Heterojunction bipolar transistor
Smallsignal base current due to vcb
Smallsignal collector current due to vcb
Base current
Collector current
Boltzmann constant
Diffusion length
98
105
[41]E.J. Prinz, P.M. Garone, P.V. Schwartz, X. Xiao, and J. C. Sturm, "The effect of
base dopant outdiffusion and undoped Si[.xGex junction spacer layers in
Si/Sij.xGex/Si heterojunction bipolar transistors," IEEE Electron Device Lett.,
vol. 12, No. 2, pp. 4244, Feb, 1991.
[42] E.L. Heasell, "The derivation of a relationship for the chargecontrol base
recombination lifetime in Terms of the base and collector smallsignal
conductances," IEEE Trans. Electron Devices, vol. Ed28,No. 5, pp. 595596,
May. 1981.
[43] A.J. Joseph, J.D. Cressler, R.C. Jaeger, D.M. Richey, and D.L. Harame, "Neutral
base recombination in advanced SiGe HBTs and its impact on the temperature
characteristics of precision analog circuits," IEDM Tech. Digest, pp. 755758,
1995.
[44]Z.A. Shafi, C.J. Gibbings, P. Ashburn, I.R.C. Post, C.G. Tuppen, and D.J.
Godfrey, "The importance of neutral base recombination in compromising the
gain of Si/SiGe heterojunction bipolar transistors," IEEE Trans. Electron Devices,
vol. 38, pp. 19731976, Aug. 1991.
[45] C.H. Gan, J.A. del Alamo, B.R. Bennett, B.S. Meyerson, E.F. Crabb, C.G.
Sodini, and R. Reif, "Si/Sij.xGex valence band discontinuity measurements using
a semiconductorinsulatorsemiconductor (SIS) heterostructure," IEEE Trans.
Electron Devices, vol. 41, pp. 24302439, Dec. 1994.
[46] S. Takagi, J.L. Hoyt, K. Rim, J.J. Welser, and J.F. Gibbons, "Evaluation of the
valence band discontinuity of Si/Si j_xGex/Si heterostructures by application of
admittance spectroscopy to MOS capacitors," IEEE Trans. Electron Devices, vol.
45, pp. 494501, Feb. 1998.
[47] Z. MatutinovicKrstelj, V. Venkataraman, E. J. Prinz, J.C. Sturm, and C.H.
Magee, "Base resistance and effective bandgap reduction in npn Si/Sij.xGex/Si
HBTs with heavy base doping," IEEE Trans. Electron Devices, vol. 43, pp. 457
466, Mar. 1966.
58
will certainly cause some errors since the SIMS profiles may not be very accurate for
very thin base layers, especially near the collector junction. Note that the base dopant
impurity concentration profile NAA(X) can be better approximated by a Gaussian
distribution (see Appendix B for Gaussian distribution model). This results from the
boron outdiffusion from the GeSi layer during subsequent fabrication. The increasing
doping Naa(x) at x=0 shown in Fig. 1.2 will produce a retarding electric field which
will decelerate the base minority carrier movement across this region. The decreasing
doping density will produce an aiding electric field. Thus, an uncertainty in a precise
location of XR can cause errors in calculation of the transit time and other parameters.
In order to estimate the influence of this uncertainty on the GeSi HBT analysis, we
investigate the dependence of T) j, tB on XR/XB and r2. Figure 2.9 shows the normalized
Geinduced electric field r j versus XR/XB with r2 as a parameter. r j is the key
parameter underlying the operation of GeSi HBT. Figure 2.10 shows the transit time
ratio tB(GeSi)/tB(Si) versus XR/XB with r2 as a parameter for ^=2.0. As can be seen
from Fig. 2.9 and Fig. 2.10, and tB(GeSi)/tB(Si) are rather insensitive to the location
of the transition point XR. This is expected, since ti, and tB(GeSi)/tB(Si) are calculated
from the collector current ratio of the two transistors with nearly identical geometry and
dopant impurity concentration profiles. The above sensitivity analysis shows that taking
x=XR as the transition point where the Ge linear profile intersects the flat Ge
concentration profile in the measured SIMS dependencies, Fig. 1.2, will result in a
negligible error.
16
improvements of GeSi HBT, such as higher Early voltage VA will be discussed in detail
in later chapters.
1.5 Monte Carlo Simulation of Mobilities
The Geinduced strain in the GeSi layer will change the Ek relationship in both the
conduction and valence band, and thus change the effective masses of electrons and
holes, mn and mp. The electron and hole mobilities, p,n and ju.p, will be also affected by
the change in the effective masses and become anisotropic. Monte Carlo simulations
were used to calculate the electron and hole mobilities in strained and relaxed GeSi
layers for different Ge mole fraction and dopant impurity concentrations [1718].
Figure 1.10 and Fig. 1.11 show the calculated results for electron mobilities versus Ge
mole fraction in ptype and ntype GeSi layer, respectively, for doping level of 1018
cm"3 close to that in a typical GeSi HBT [17]. The longitudinal mobility (upward
triangles) is in the growth or z direction, the transverse mobility (downward triangles) is
in the xy plane. Figure 1.10 and Fig. 1.11 shows only a small (~2%) increase in the
longitudinal mobility with Ge mole fraction at 300K. However, a much larger
improvement (50%) was calculated at 77K for xGe~510%. The transverse mobility
shows a large reduction with increasing xGe which is always smaller than that in Si
(xGe=0) at any Ge concentration level. The results for the majority and minority hole
mobilities are similar to those for the electron mobilities in Fig. 1.10 and Fig. 1.11. In
summary, the expected minority electron mobility improvement for the doping level of
about 1018 cm'3 is very small, xn(GeSi)/p,n(Si)=1.0. This agrees very well with our
experimental results as will be shown in chapter 2.
62
The base minority electron distribution for r,=2.0 is shown in Fig. 2.13. As can be
seen from the figure, n(x) throughout the base is larger for GeSi HBT. This is because
the intrinsic carrier concentration r^ of GeSi HBT is larger than that of Si B JT.
The base minority carrier charge ratio QB(GeSi)/QB(Si)>1.0 vs rj is shown in Fig.
2.14. The increase of base minority carrier charge will not affect the performance of
GeSi HBT significantly because the aiding electric field in the base region increases Ic.
Thus, the minoritycarrier transit time through the base, tB=QB/Ic is still smaller for the
GeSi HBT. This is shown in Fig. 2.15 for the forward active mode.
The transit time improvement is very important for high frequency performance. As
is shown in the figure, the base transit time can give a factor of two improvement for
r j=2.0, which corresponds to about 10% Ge in the base. The improvement of tB(GeSi)
is due to the Geinduced aiding electric field Ebi, rather than the enhancement of
diffusivity Dn, which is negligible according to our measurements and theoretical
calculations.
The base transit time ratio tB(GeSi)/tB(Si) for the reverse active mode is shown in
Fig. 2.16. As expected for the Ge concentration profile in Fig. 1.2, tBR(GeSi)/tBR(Si)
1.0.
2.8 Summary
In conclusion, a new LFGV methodology was developed and a complete analysis of
the effects of Ge in the base of Si/Ge^i^/Si HBT was presented. The LFGV
methodology is based on a onedimensional BJT theory and accurate dc and low
frequency measurements of similar GeSi and Si transistors. It is accurate, expedient and
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Amost Neugroschel, Chairman
Professor of Electrical and Computer Engineering
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
ChihTang Sah, CocMirman
Eminent Scholar and Graduate Research
Professor of Electrical and Computer Engineering
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Peter S. Zory
Professor of Electrical
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
Professor of Electrical and Computer Engineering
78
calculate the important GeSi HBT parameters as a function of temperature. The results
are shown in Table 3.2 and Table 3.3. From Table 3.2, we can see that the base
collector junction capacitance CBC remains constant as temperature decreases while the
baseemitter junction capacitance CBE shows a slight decrease with decreasing
temperature. This is because the thickness of the basecollector junction SCR, xBC, is
much larger than xBE due to lower doping at the collector junction. This gives
Cbc<
fraction of xBC. Figure 3.3Fig. 3.5 show the measured temperature dependence of r\{,
the diffusion length LB(GeSi), and the base recombination component QB/xB(GeSi)/IB
vs temperature, respectively.
From Fig. 3.3, we can see that r j obtained by a combination of (3.7) and (3.8), and
thus the builtin electric field Ebi, increase almost linearly with 1/T. The dashed line is
calculated using qpAE^Qg/kT C, where AEG_Ge=70meV as measured at 297.15K,
and C=0.7. The deviation of the theoretical curve from the experimental dependence
may come from two sources: (1) decrease of the AEG_Ge with decreasing temperature,
and (2) the increase of the fitting parameter C with decreasing T. The latter one is more
reasonable since the NCNV product is a strong function of temperature as is shown in
equations (1.6a) and (1.6b). The increase of C at low temperature means that the
NcNv(GeSi)/NcNv(Si) is a decreasing function of temperature.
Figure 3.4 shows the measured diffusion length LB(GeSi) vs T. The diffusion length
can be expressed as
LB(GeSi )=1l Dr
\(kT/q)^r
(3.10)
Concentration /(cm 3)
75
Fig. 3.1 GeSi HBT with different Ge concentration profiles in the base region.
Germanium /(1%)
41
The LFGV methodology described in the next section will give an accurate
extraction of rij.
2.4 Derivation of LFGV Methodology and Parameter Extraction
Below we present detailed derivation and description of the LFGV methodology. As
is discussed in section 2.2, the base thickness modulation gives rise to two smallsignal
conductances, an output conductance, g0, and a reverse transconductance, gr. In the first
step of the LFGV methodology, we need the g0 only. It can be derived from the Ic
equations (2.9) and (2.15) and is given by (see appendix A)
d XR
9o =
XB A(expp1l)expp2/hi + (lA)(expp2l)/n2
3V
CB
(2.17)
'BE
From (2.17), we can see immediately that the expression (2.17) for g0 is simpler than
(2.15) for Ic. The reason for this is that the unknown parameters, DB(GeSi)/DB(Si), a,
and y are lumped into Ic which is a measurable quantity. This is one of the advantages
of g0 compared with Ic and is one important aspect of the LFGV methodology. Useful
relations (2.18a) and (2.18b) are derived by taking the ratios of the collector currents
and the output conductances of the GeSi HBT and Si reference BJT,
Ic(GeSi)/Ic(Si) A[(expp1l)expp2]/p1 + (lA)(expp2l)/p2
= (2.18a)
g0(GeSi)/g0(Si)
Ic(GeSi)/Ic(Si)
g0(GeSi)/g0(Si)
Aexpp2 + (lA)(expp2l)/p2
expr^l
(2.18b)
hi
The only assumption in (2.18a) and (2.18b) is that 3XB/3VCB is the same for GeSi HBT
and Si reference BJT. This assumption is proven to be a very good approximation from
measuring the junction capacitance CBC. These relations are used to calculate r) t
97
Fig. B.2 Gaussian distribution and its exponential approximation.
69
general. The analytical model for the Geconcentrationdependence of the density of
states product decouples it from the minoritycarrier mobility and heavydoping effects.
A combination of Ic and g0 measurements gives a unique determination of the critical
parameter rj. All the important minoritycarrier transport parameters underlying the
improved performance of GeSi HBT compared to similar Si BJT were determined and
quantified.
The results show that the reduction in the base transit time is due to the Geinduced
aiding electric field in the region with linearlyvariable Ge concentration. The minority
carrier mobility or diffusivity enhancement was found to be negligible compared to Si
BJT. The energy gap reduction due to the heavy doping (NBase=3xl018cm"3) is about
lOmeV larger in GeSi HBT than in Si BJT at room temperature. First accurate
measurements of the minority carrier diffusion length and recombination lifetime in the
base show greatly improved values in the GeSi base compared to Si. The base current
was found to be dominated almost completely by the recombination within the emitter
in the large 100.4x100.8pm2 transistors, while in circuit transistors with a narrow
emitter stripe (
importance of the sidewall and surface effects. Since the dc and lowfrequency small
signal measurements are independent of parasitics, accurate highfrequency and high
speed parameters can be readily obtained which are difficult to measure directly such as
the ~lps base transit time. The performance improvements of GeSi HBT over Si BJT
were shown graphically as a function of rj.
Critical Layer Thickness
10
Fig. 1.4 Critical layer thickness of GeSi layer versus Ge fraction. The dashed line
shows the theoretical results calculated from the thermal dynamics. The solid line gives
the experimental results. Adapted from Sah [1, p. 949].
20
AEQGe(O) AEG_Ge(XB)
PGeSi
Base
n+Si
Emitter
v
N
Ge
AA
0
X
B
nSi
Collector
Fig. 1.12 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with
a triangular Ge distribution in the base region at equilibrium (zero bias). EF is the Fermi
level at equilibrium. AEG_Ge(0) is the Geinduced energy gap reduction at x=0 in the
quasineutral base region. AEG.Ge(XB) is the Geinduced energy gap reduction at x=XB
in the quasineutral base region. AEG.Ge = AEG_Ge(XB) AEG.Ge(0) is the total Ge
induced energy gap reduction through the base region.
68
i i ItrTTl i I i I i I i I i I i I i
0 2 4 6 8 10
Til
Fig. 2.16 Base minority carrier transit time ratio tB(GeSi)/tB(Si) in reverse active mode
vs r j (normalized to DB(GeSi)/DB(Si)). The reverse base transit time is much larger
than in the forward active mode.
61
2.1 GeSi HBT Performance Enhancement
In this section, we will plot the important GeSi HBT parameters as a function of r j
normalized to that of Si BJT in order to show the performance enhancement of GeSi
HBT over Si BJT. Since the triangular Ge distribution in the base region is a simple and
also a very good approximation for the trapezoidal Ge distribution, we will use the
triangular Ge distribution as a demonstration. The extension to the trapezoidal Ge
concentration profile is straightforward. In fact, most of the important performance
enhancements can be accomplished by the triangular Ge distribution in the base region.
The most important performance enhancement of GeSi HBT over Si BJT is the
increase in the collector current Ic(GeSi), as is shown in Fig. 2.11. The collector current
ratio Ic(GeSi)/Ic(Si) varies almost linearly with rj for rj>~3.0. For ^^3.0, the
collector current ratio Ic(GeSi)/Ic(Si) can be modeled by a linear function,
Ic(GeSi)/Ic(Si) c r)j c (AEG_Ge/kT)C.
The output conductance ratio g0(GeSi)/g0(Si) vs is shown in Fig. 2.12. As can be
seen from the figure, the g0 ratio shows a rapid decrease for l.Odj^.O and tends to
zero for very large rj j. This means that the output resistance of GeSi HBT is much
larger than that of Si BJT for large pj and the Early voltage VA also increases rapidly
with r!. So the figureofmerit of (3VA improves greatly for large r) j. For example, for
r)1=2.0, g0(GeSi)/g0(Si) = 0.73, Ic(GeSi)/Ic(Si) = 2.3, and VA(GeSi)/VA(Si) = 2.3/0.72
= 3.0. The current gain (3 will also increase roughly 2.3 times for the same base current.
So the total (3VA improvement is ~ (3.0)(2.3)=7.0.
99
me Electron effective mass
mh Hole effective mass
n(x) Base minority carrier distribution
n Intrinsic carrier concentration
Nc Effective density of state in conduction band
Nv Effective density of state in valence band
P(x) Base dopant impurity concentration profile
QB Base minority carrier electric charge
Qe Emitter minority carrier electric charge
Si Silicon
tB Base minority carrier transit time
vcb Smallsignal voltage superimposed upon VCB (~lkHz)
VA Early voltage
xGe Germanium concentration in the base region
XR Transition position of the Ge distribution in the base
XB Base thickness
Vbi Builtin potential of the pn junction
VBE Base emitter junction bias
VCB Base collector junction bias
a AEG_Ge(0)/kT (normalized energy gap reduction at x=0 in the base region)
(3, PF Common emitter current gain
Y AEG_HD/kT (normalized energy gap reduction in the base region
due to heavy doping)
Ay [AEG_HD(GeSi)AEG_HD(GeSi)]/kT
r) Normalized (to thermal voltage kT/q) energy gap reduction due
to Ge in the base region
r\2 Field factor due to nonuniform base doping
r]j Normalized energy gap reduction in the base region
Tg Base minority recombination lifetime
xE Recombination time constant in the emitter
pn Base minority carrier mobility
APPENDIX A
GENERAL EQUATIONS FOR BJT
A. The general integral relations for calculation of important parameters in bipolar
junction transistor are given below. Equations for a trapezoidal Ge distribution in the
base and constantexponential dopant impurity concentration profile in the base are
given in section B.
1. Collector current Ic(GeSi) [1, p. 760, 38]
q.exp(qVBE/kT)
Ir
Xr
P(x)
dx
0 Dn(x). ni2(x)
2. Minority carrier distribution in the base [38]
Jc nf(x)
n(x) =
q P(x)
XB P(y)
x Dn.ni2(y)
dy
3. Minority carrier charge in the base
(A. 1)
(A.2)
'B
n(x)dx
4. Minority carrier base transit time [38]
tB ~
XB nf(x)
P(x)
P(y)
5. Correction factor B [27]
dx
1
B =
Dn,t;B
XB P(X)
n2(x)
dx
x n?(y)
XB n?(x)
dx 1
P(x)
(A.3)
(A.4)
(A.5)
89
35
temperature. But theoretical calculations show that the density of states and mobility
have very complicated temperature dependence, so the assumption of the temperature
independence of the two preexponential factors in (2.6) is unreliable and can lead to
large errors in AEG.Ge, as also noted previously [32], The second approach is to assume
theoretically predicted values of (NcNv)GeSi and Dn(GeSi) and then calculate EA from
(2.6). Apparently, there is another unknown, Ay, and also the theoretical results are
strongly dependent on models and input parameters. Furthermore, both approaches give
an estimate for the energy gap reduction term only and do not provide values or
estimates for the mobility or recombination lifetime in the GeSi base.
The above discussion shows that an analysis based on the measurement of Ic(GeSi)
only is not sufficient. The LFGV methodology, however, will use a combination of the
dc IBVBE and IGVBE characteristics with the smallsignal output conductance g0 and
reverse transconductance gr measurements. This novel methodology can overcome the
limitations of the above and other approaches.
The LFGV methodology will be presented here for the idealized trapezoidal Ge
concentration profile in the base which is shown in Fig. 2.2. In this model, the base
dopant impurity concentration profile and the Ge distribution are divided into two
regions. In the first region, 0
Ge distribution is assumed to be retrograded linear. In the second region, XR
base doping is approximated by an exponential function and the Ge distribution is
assumed to be constant. In the region with the linear Ge distribution (0
Eg(0) AEG_Ge(x/XR) [1], where EG(0) is the energy gap at x=0, and AEG.Ge = AEG(XR)
49
Fig. 2.5 The measured forward active mode and reverse active mode smallsignal
output conductances, g0, reverse transconductance, gr, and the collector current, Ic, of Si
BJT, as a function of VBE. All measured curves show the expected exp(qVBE/kT)
dependence.
IC/(1A)
Concentration /(cm 3)
76
Depth
Fig. 3.2 Bipolar transistor with different highlow dopant impurity concentration
profiles in the base region.
CHAPTER 4
SUMMARY AND CONCLUSIONS
4.1 Summary
In chapter one, we gave a general review of the GeSi HBT fabrication, including the
discussion of pseudomorphic growth of GeSi layer and its critical layer thickness. A
comparison of the Si/GexSi,_x/Si HBT characteristics with those for a Si reference BJT,
Monte Carlo simulations of the mobility improvement and other Geinduced changes in
the material properties were also discussed. An exponential approximation model for
the density of states NCNV dependence on the Ge concentration was given for a linear
Ge distribution in the base. In chapter two, a lowfrequencyconductancevoltage
(LFGV) methodology was developed for detailed analysis of the Si/GexSij.x/Si HBT.
This methodology is based on a comparison of the GeSi HBT performance with that of
a reference Si BJT. A brief review of the base thickness modulation method which
underlies the LFGV method is also given. Experimental data were presented to
demonstrate the LFGV methodology in highperformance Si/GejSi^/Si HBTs.
Important performance enhancements of GeSi HBT over Si BJT were measured and
analyzed. In chapter three, we introduced a methodology for GeSi HBT analysis
without using a reference Si BJT. In this nonreference analysis, the key parameter r j
of GeSi HBT can be determined directly and a complete charge control analysis can be
made. Experimental demonstration of the nonreference method was given.
87
101
[8]B.S. Meyerson, "UHV/CVD growth of Si and Si:Ge alloys: Chemistry, physics,
and device applications," Proceedings of the IEEE, vol. 80, No. 10, pp. 1592
1608, Oct. 1992.
[9]A. Gruhle, H. Kibbel, U. Erben, and E. Kasper, "MBEgrown Si/SiGe HBTs with
high P, fT, and fmax," IEEE Electron Device Letters, vol. 13, No. 4, pp. 206208,
Apr. 1992.
[10]A. Pruijmboom, J.W. Slotboom, D.L. Gravesteijn, C.W. Fredriksz, A.A. van
Gorkum, R.A. van de Heuvel, J.M.L.van RooijMulder, G. Streutker, and G.F.A.
van de Walle, "Heterojunction bipolar transistors with SiGe base grown by
molecular beam epitaxy," IEEE electron Device Lett., vol. 12, pp. 357359, July
1991.
[11]M. Hong, E. de Frsart, J. Steele, A. Zlotnicka, C. Stein, G. Tam, M. Racanelli, L.
Knoch, Y.C. See, and K. Evans, "Highperformance SiGe epitaxial base bipolar
transistors produced by a reducedpressure CVD reactor," IEEE Electron Device
Letters, vol. 14, pp. 450452, Sep. 1993.
[12] A. Neugroschel, C.T. Sah, J.M. Ford, J. Steele, R. Tang, and C. Stein,
"Comparison of timetofailure of GeSi and Si bipolar transistors," IEEE Electron
Device Lett., vol. 17, pp. 211213, May 1996.
[13] R. Tang, J. Ford, B. Pryor, S. Ansndakugan, P. Welch and C. Burt, "Extrinsic
base optimization for highperformance RF SiGe heterojunction bipolar
transistors," IEEE Electron Device Lett., vol. 18, pp. 426428, Sep. 1997.
[14] R. People and J.C. Bean, "Band alignments of coherently stained Ge^i^/Si
heterostructures on <001> GeySij_y substrates," Appl. Phys. Lett., vol. 48, pp.
538540, Feb. 1986.
[15]J.C. Bean, "Siliconbased semiconductor heterostructures: Column IV bandgap
engineering," Proceedings of the IEEE, vol. 80, No. 4, pp. 571587, Apr. 1992.
21
AEoGe(O) AEG_Ge(XB)
Fig. 1.13 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with
a triangular Ge distribution in the base region under forward active operation, VBE>0V,
VCb>0V. Fn is the quasiFermi level for electrons and Fp is the quasiFermi level for
holes. Vbi is the builtin potential of the p/n junction. AEG_Ge(0) is the Geinduced
energy gap reduction at x=0 in the quasineutral base region. AEG_Ge(XB) is the Ge
induced energy gap reduction at x=XB in the quasineutral base region. AEG_Ge =
AEGGe(XB) AEGGe(O) is total Geinduced energy gap reduction through the base
region.
DB(GeSi)/DB(Si)
53
Temperature /(1K)
Fig. 2.7 The measured temperature dependence of DB(GeSi)/DB(Si) in the range of
200K300K normalized to 296K in large 100.4x100.8pm2 transistor.
3
the changes in the physical parameters that give the performance improvements of GeSi
HBT over Si BJT. The performance of the GeSi HBT demonstrated so far is superior
to, or comparable with, that of the GaAs HBTs [67]. The most important advantage of
using thin GeSi layer is that its epitaxial deposition is fully compatible with Si
integratedcircuit manufacturing technology; thus, monolithic integration of GeSi HBT
with Si integrated circuit can be very easily accomplished. Also, the thermal
conductivity of Si is three times higher than that of GaAs which is very important in
power amplifiers and very large scale integrated circuit (VLSI). The GeSi HBT has
very important applications in portable communication systems, such as mobile phones,
pagers, DACs and others.
1.2 GeSi HBT Fabrication
Silicon and germanium are completely miscible in any mole fraction ratio. The GeSi
material has a diamond lattice structure. The lattice constant can be calculated from the
Vegard rule [5]:
a(GexS!.x) = a(Si) + x(a(Ge) a(Si)) (1.1)
where x is the Ge mole fraction and a is the lattice constant. The lattice constant of Ge
is 4.17 percent higher than that of Si. When Si and Ge atoms are mixed during the
growth of the GeSi layer, the Si lattice is tetragonally distorted and strained. The lattice
mismatch between the GeSi layer and the Si substrate can be accommodated by two
ways. One is a generation of misfit dislocations at the GeSi/Si interface which may also
propagate through the entire layer. This occurs when the GeSi layer is very thick and
the strain is released by creation of misfit dislocations. The second possibility is
I certify that I have read this study and that in my opinion it conforms to acceptable
standards of scholarly presentation and is fully adequate, in scope and quality, as a
dissertation for the degree of Doctor of Philosophy.
ical Engineering
1/
This dissertation was submitted to the Graduate Faculty of the College of
Engineering and to the Graduate School and was accepted as partial fulfillment of the
requirements for the degree of Doctor of Philosophy.
December 1998
Winfred M. Phillips
Dean, College of Engineering
M. J. Ohanian
Dean, Graduate School
32
change by AQB with the smallsignal vcb. The minority carrier concentration at the
emitter boundary of the quasineutral base at x=0, N(0), remains constant for a fixed VBE
and is given by [n^/P(0)]exp(qVBE/kT), where P(0)=NAA(0) is the holeemitter
boundary of the quasineutral base at x=0, N(0) remains constant for a fixed VBE and is
given by [n?/P(0)]exp(qVBE/kT), where P(0)=NAA(0) is the hole concentration at x=0,
k is the Boltzmann constant, and T is the temperature. The change in the slope of N(x)
gives the smallsignal collector current ic. The output conductance g0 is defined by
9o=
'cb
(2.1)
BE
The base current IB also changes due to a change in the base recombination current,
Qb/tb, the base thickness modulation by the smallsignal voltage vcb. This
produces the reverse transconductance gr which is defined by
9r=
'cb
(2.2)
BE
where ib is the smallsignal base current due to vcb. Analytical expressions for g0 and gr
can be derived (see appendix A),
IC dXB
^be
9o=
'cb
^BE XB
dV
9r=
'cb
CB
dXR
(2.3)
^BE tB*XB d^CB
(2.4)
'BE
where XB is the quasineutral base thickness, QB is the base minority carrier charge
corresponding to VBE and VCB, and xB is the base minority carrier recombination
lifetime. Since Ic and QB are exponential functions of the forward baseemitter voltage
67
0 2 4 6 8 10
Hi
Fig. 2.15 Base minority carrier transit time ratio tB(GeSi)/tB(Si) in forward active mode
vs rj j (normalized to DB(GeSi)/DB(Si)). The base transit time can give a factor of two
improvement for 1^=2.0 which corresponds to about 10% Ge in the base.
55
transistor and stripe transistor is due to the perimeter recombination at the edges of the
stripe transistor. But, the narrowemitter transistor is used in the circuit applications
because the narrowemitter geometry minimizes the base series resistance. So the
measurements for stripe transistor are also important. Due to the edge effect, some
measured parameters, especially those associated with the base current, such as IB, PF,
gr and g0/gr, are different from those of the large square device. Two key parameters, rj
and tB, are however very similar for both transistors geometries. This is because rj and
tB are calculated from Ic and g0 only which are essentially onedimensional even for the
narrow emitter transistor with a very thin base. Figure 2.8 shows the measured
temperature dependence of ln[exp(Ay)DB(GeSi)/DB(Si)] in the range of 150K300K.
The temperature dependence shows a small activation energy of EA~8meV. For this
transistor pair, the measured [DB(GeSi)/DB(Si)]exp(Ay) = 1.0 compared to 1.6 for the
large transistor. This gives DB(GeSi)/DB(Si)=0.7. The disagreement with the
DB(GeSi)/DB(Si) = 1.0 result obtained from the large transistor is due to the narrow
finger geometry.
2.6 Sensitivity of tB and r1 to r2 and XR/XB
An idealized base dopant impurity concentration profile P(x) and Ge distribution in
the base used for the analysis are shown in Fig. 2.2. The base region is divided into two
regions, derived from the actual SIMS profile shown in Fig. 1.2. For 0
doping P(x) is flat and the Ge distribution is retrograded linear, and for XR
base doping decreases exponentially and the Ge distribution is flat. The transition point
is defined as XR. Although XR can be obtained from the measured SIMS profiles, this
2
The breakthrough for an Sicompatible HBT was accomplished again by IBM in
1988 [5], They used Germanium Silicon (GeSi) as the base material which has a
smaller energy gap than that of the Si emitter and collector. A thin strained single
crystal GexSij.x film on Si substrate was grown pseudomorphically using ultrahigh
vacuum/chemical vapor deposition/low temperature epitaxy (UHV/CVD/LTE) process
[68], The energy gap of the base can be tailored by varying the Ge mole content in the
base. Furthermore, the Ge distribution in the base region can be graded which will
produce an aiding electric field for the minority carrier transport through the base. Ge
incorporation in the base will also modify the effective mass and mobility of both
majority and minority carriers. This results in a number of advantages of Si/GexSij.x/Si
HBT over Si B JT, such as higher collector current, higher current gain (3, superior high
frequency performance (higher unit gain frequency fT and maximum oscillation
frequency fmax), and excellent low noise behavior. The best high frequency
performance was obtained at low voltages (VCE=13V), which is very useful for
portable communication applications.
Since the introduction of the pseudomorphic growth mechanism in 1972 [1], the
techniques of growing a thin base GeSi layer have been highly developed to
accommodate the lattice mismatch between the GeSi layer and Si substrate.
Technology has been developed for GeSi layer deposition in commercial epitaxial
reactors. During the pseudomorphic growth, the lattice of the GeSi layer has to conform
with that of the Si substrate; thus, the lattice of the GeSi layer is strained and distorted
compared with that of the unstrained GeSi layer. This lattice strain is responsible for
8
the four surfaces on the kx and ky axes is lowered. The resulting reduction in the energy
gap and lowering of the effective mass are very beneficial for improved performance of
GeSi HBT. The smaller energy gap increases the collector current and gain, while the
smaller mn in the vertical minoritycarrier electron flow direction (in npn HBT)
increases the minoritycarrier mobility which reduces the base transit time. Figure 1.6
illustrates the inplane (x) of growth direction and normal to the growth plane or the
growth (z) direction. Figure 1.7 shows the calculated values of energy gap reduction of
strained GeSi layer and unstrained (bulk) material versus the Ge mole fraction. It shows
that, for the strained GeSi layer, the energy gap reduction AEG_Ge is almost a linear
function of the increasing Ge mole fraction xGe for xGe<~40%.
1.4 GeSi HBT Characteristics
Typical collector current and base current dependencies on VBE of GeSi HBT and Si
BJT are shown in Fig. 1.8. The collector current of GeSi HBT, Ic(GeSi), shows large
improvement over that in Si BJTs. The base current of GeSi HBT, IB(GeSi), is also
slightly smaller than that in Si BJTs. This gives a much larger common emitter forward
current gain $F of GeSi HBT compared to that in Si BJT. Figure 1.9 shows a typical
high frequency response versus collector current. Improvement by almost a factor of 2
in fT for GeSi HBT is shown. Recently, fT values as large as 116GHz or higher were
reported [20]. This improvement of fT is very important for RF wireless communication
applications. The improvement of fT is mainly due to the Geinduced aiding electric
field for the base minority carriers which reduces the minority carrier transit time
through the base region which still is a limiting factor of fT. Other performance
37
 AEg(0) is the total Geinduced energy gap variation through 0
aiding electric field, Ebi, for the electrons in the ptype GeSi base with a uniform boron
doping and position dependent energy gap EG(x) from the graded Ge concentration
profile is given by [37]
Ebi = (1/q)CdEc/dx) = (1/q)(AEG_Ge)/XR = independent of x. (2.7a)
= (kT/q)(p/XR), (2.7b)
where r = AEG.Ge/kT is the normalized Geinduced energy gap reduction through the
base.
For XR
can be approximated assuming an exponential profile, NAA(x) = NAA(0)exp[r2(x
Xr)/(XbXr)], where r2 = In [NAA(XR)/NAA(XB)]. The constant builtin aiding field
for XR
Ebi2 =(kT/q)[t?2/(XBXR)]. (2.8)
The LFGV methodology also uses the collector current ratio Ic(GeSi)/Ic(Si), but this
ratio is modified. The Geinduced NCNV reduction term NcNv(GeSi)/NcNv(Si) is
lumped together with the Geinduced energy gap reduction term AEG_Ge=(kT/q)r where
the lumped parameter is described by a factor, rip This modification gives several
benefits. First, rij can be measured directly and accurately, while the original two
factors are not measurable. Secondly, all other GeSi HBTs parameters can be calculated
in terms of rij. This modification also enables the decoupling of the effects of NCNV
from that of the mobility and heavy doping on Ic(GeSi).
CHAPTER 2
LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) METHODOLOGY
2.1 Introduction
In the last chapter, we presented a preliminary qualitative analysis of the GeSi HBT
and its comparison with the Si BJT. In this chapter, we describe a lowfrequency
conductancevoltage (LFGV) methodology developed in order to obtain accurate
quantitative values for the dominant minoritycarrier transport parameters that underlie
the GeSi HBTs performance enhancement, such as the diffusion length, recombination
lifetime, and diffusion coefficient in the base layer. The LFGV method also gives the
analysis methodology of current gain and emitter injection efficiency based on a
separation of the total base current into the recombination losses in the emitter and base.
Thus, the LFGV method gives a complete chargecontrol analysis of the bipolar
transistor performance. Analytical models are developed for the LFGV method for a
trapezoidal and linearly graded Ge concentration profile in the base. However, the
LFGV method is general and can be applied to any bipolar transistor with an arbitrary
dopant impurity and materialcomposition profile, including those using the IIIV
compound semiconductors.
Comprehensive reviews of the Si/GexSij_x/Si HBT technology, including fabrication,
Geinduced effects in the base layer, currentvoltage characteristics, modeling, and
performance comparison with Si BJT are discussed in chapter 1. Performance
28
56
Table 2.2 Summary of experimental data and calculated parameters for Si/GexSij.x/Si HBT
and reference Si BJT. The top six lines give the measured dc and BTM ac data obtained at
23C, VBE = 0.58V and VCB = IV. The computed parameters using LFGV method are
shown in the next eighteen lines. The first column is for Si/GexSi1_x/Si HBT, the second
for Si baseline BJT, and the third column gives the comparison ratio of the two previous
columns. ABE=(0.8x26)x88pm2, XR=680, XB=800.
Parameter
GeSi HBT
Si BJT
GeSi/Si
Ic (PA)
14.42
5.16
2.79
IB (nA)
161.6
173.2
0.93
Pf
89.2
29.8
3.0
g0 (i/a)
2.687xl08
4.08xl08
0.66
gr (l/Q)
3.5xlO~10
5.2xlO"10
0.67
go^gr
76.77
78.46
0.98
*12
2.2
2.2
2.5
AEG_Ge (meV)
79
Ebi (kV/cm)
12
B
15
2
7.5
Lb (pm)
1.2
0.7
1.7
Qb/tb (nA)
12
32
0.38
Qe/te (nA)
149
141
1.06
(Qe^eVCQiAb)
12
4.4
2.7
(Qe/te)/Ib (%)
92
82
1.1
(Qb/tb)/Ib (%))
8
18
0.4
Qb (C)
35xl0~18
25xl018
1.4
Db (cm2/Vs)
4
5.5
0.7
tB (Ps)
2.4
4.9
0.5
tb (ns)
2.8
0.8
3.5
Ay
0.3
AEg.hd (meV)
8
VA (V)
536
127
4.2
96
6. Reverse transconductance
Ic exp((XBX0)2/
9r =
DnTB erf ((XBX0)/er) + erf(X0/cr)
{erf (Xg/cr) jBexp[(xX0)2/cr2]dx + jBexp[(xX0)2/o2)]erf[(xX0/cj)]dx}
(B.16)
80
Table 3.2 Temperature dependence of measured parameters of GeSi HBT
T(K)
SoR^SoF
^CR^CF
Cbc(pF)
CBE(PF)
297.2
285.2
1.63
1.99
39.4
250
495.9
1.81
1.99
39.0
225
630.7
1.85
1.99
38.6
200
967.3
2.29
1.99
38.4
175
2792
3.94
1.99
38.2
150
4000
2.88
1.99
38.0
125
13,000
4.81
1.99
37.7
Note: AE= 100.4x100.8Â¡am2, T=297.15K, VBE=0.52V, VBC=1.0V.
The junction capacitances CBC and CBE were measured at f=lkHz.
Table 3.3 Temperature dependence of GeSi HBT parameters
calculated from the measured data in Table 3.2
T(K)
hi
(pm)
[Qb^bI^b
297.2
2.04
3.9
1.2%
250
2.60
4.3
0.9%
225
2.83
4.0
1.0%
200
3.05
3.2
1.5%
175
3.58
2.6
2.3%
150
4.25
1.8
4.4%
125
4.93
1.5
4.2%
26
nf (x) = Nc(x)Nv(x).exp(EG(x)/kT). (1.8)
For a linear Ge distribution using results from section (1.6), we have
Eg(x) = EG0 AEG_HD AEG_Ge(0) AEG_Ge.x/XR (1.3)
where AEG.Ge = AEG_Ge(XR) AEG.Ge(0) is the total energy gap reduction across the
quasineutral base region. AEG.HD is the energy gap reduction due to heavy doping
effect, and AEG_Ge(0) is the Geinduced energy gap reduction at x=0 in the quasineutral
base region. Inserting equations (1.3) and (1.7) into equation (1.8), gives an analytical
expression for the intrinsic carrier concentration nÂ¡ for a linear Ge distribution,
nf(x) = niQ 2 exp(AEG_Ge(0)/kT).exp(AEG_HD/kT) exp(x/x^, (1.9)
where ni0 is the intrinsic carrier concentration of Si
nio2 = Nc(0)Nv(0).exp(EGO/kT) (1.10)
The constant Xj is defined as l/x1=(l/XB)[(AEG.Ge/kT)C]. For a simplicity, we define
a=AEG_Ge(0)/kT and y=AEG.HD/kT, then equation (1.9) becomes
n2(x) = ni0 2 .expa.expy.explx/x^. (1.11)
1.8 Summary
In conclusion, the basic parameters governing the GeSi HBT performance were
discussed in this chapter, including fabrication by pseudomorphic epitaxial growth,
critical layer thickness, and electronic properties of the GeSi layer. The lattice
mismatch between the GeSi layer and the Si substrate is accommodated by the biaxial
strain in the GeSi layer. This Geinduced strain in the GeSi layer is responsible for
changes in electronic material parameters important for HBT performance, in particular,
the energy gap, density of states, and mobility. For the Ge mole fraction xGe <~40%,
88
4.2 GeSi HBT Modeling and LFGV Methodology
The LFGV methodology is applicable to an analysis of any heterojunction bipolar
transistor (HBT), including those based on IIIV materials. The LFGV methodology
uses a combination of the dc and smallsignal measurements of the transistor
characteristics. The dc characteristics are the collector and base current as a function of
the emitterbase junction bias. The smallsignal characteristics are the output
conductance and reverse transconductance resulting from the base thickness modulation.
They are measured at low frequency (~lkHz) to avoid parasitic capacitance and
inductance effects.
The LFGV methodology was applied for the analysis of Si/GexSi1_x/Si HBT with a
linearlygraded or trapezoidal Ge concentration profiles in the base. Analytical
expressions for the dc and smallsignal parameters were developed. Important minority
carrier transport parameters, such as the base diffusion length, recombination lifetime,
diffusivity, builtin electric field, and base transit time were determined. The LFGV
methodology also allows a separation of the base current into the components
recombining in the base and emitter region, respectively.
Concentration /(cm'3)
6
Depth
Fig. 1.2 Secondary ion mass spectroscopy (SIMS) profile of GeSi HBT. The base
doping level is about 3xl018 cm'3 and the Ge mole fraction in the base is about 10%.
Adapted from Ron Tang et al. [13].
Germanium /(1%)
9
Unstrained GexSUx Strained GexSi^
Si Si
(a) Relaxed GexSUx
1
t
J
n
/
\
(
Si
Fig. 1.3 Two dimensional illustration of GeSi pseudomorphic growth, (a). Unstrained
GeSi layer and Si substrate, (b). Strained GeSi layer. The lattice mismatch is
accommodated by the strain in the GeSi layer, (c). Relaxed GeSi layer. There are
misfit dislocations at the GeSi/Si interface. Adapted from Sah [1, Fig. 774.1, p. 947],
104
[33] B. Le Tron, M.D.R. Hashim, P. Ashbum, M. Mouis, A. Chantre, and G. Vincent,
"Determination of bandgap narrowing and parasitic energy barriers in SiGe HBTs
integrated in a bipolar technology," IEEE Trans. Electron Devices, vol. 44, pp.
715722, May 1997.
[34] P. Ashburn, H. Boussetta, M.D.R. Hashim, A. Chantre, M. Mouis, G.J. Parker,
and G. Vincent, "Electrical determination of bandgap narrowing in bipolar
transistors with epitaxial Si, Sij.xGex, and ion implanted bases," IEEE Trans.
Electron Devices, vol. 43, pp. 774783, May 1996.
[35] J.D. Cressler, E.F. Crabbe, J.H. Comfort, J. Wamock, K.A. Jenkins, J.M.C. Stork,
and J.Y.C. Sun, "Profile Scaling constraints for ionimplanted and epitaxial
bipolar technology designed for 77k operation," IEDM, vol. 33, No.4, pp. 861
864, 1991.
[36]D. Vook, T.I. Kamins, G. Burton. P.J. Vande Voorde, H.H. Wang, R. Coen, J.
Lin, D.F. Pettengill, P.K. Yu, S.J. Rosner, J.E. Turner, S.S. Laderman, H.Su Fu,
and A.S. Wang, "Doublediffused graded SiGebase bipolar transistor," IEEE
Trans. Electron Devices, vol. 41, pp. 10131018, June 1994.
[37]D.J. Roulston, and J.M. McGregor, "Effect of bandgap gradient in the base region
of SiGe heterojunction bipolar transistors," SolidState Electronics, vol. 35, No. 7,
pp. 10191020, 1992.
[38]H.Kroemer, "Two integral relations pertaining to the electron transport through a
bipolar transistor with a nonuniform energy gap in the base region," SolidState
Electron., vol. 28, No. 11, pp. 11011103, Nov. 1985.
[39]A. Neugroschel, "Emitter injection efficiency in heterojunction bipolar transistor,"
SolidState Electron., vol. 30, No. 11, pp. 11711173, 1987.
[40]J.W. Slotboom, G. Streutker, A. Pruijmboom, D.J. Gravesteijn, "Parasitic energy
barriers in SiGe HBTs," IEEE Electron Device Lett., vol. 12, No. 9, pp. 486488,
Sep. 1991.
81
8
6
= 4
2
O
2 4 6 8 10
1000/T /(1/1K)
Fig. 3.3 Normalized energy gap reduction parameter rj2 vs 1000/T.
43
(2.19) we neglected the spacecharge region recombination components that can be
easily separated out from the measured IB. In order to separate the recombination losses
in (2.19), we need the gr measurement which results from the modulation of the base
recombination loss. For the GeSi HBT, gr is derived as
Qg X 1 dXg
gr = (2.20)
igXg Y A[(expfj^l)//7^)]sxprj2 + (lA)[(sxpf72l)/fj2] ^cb ^be
In (2.20), the terms X and Y are given below
X = A2(exp^1+^1l)//j12 +
AdAlCiexp^D/rji]. [(exp^2l)/p2] +
(lA)2(expp2p2l)/n22 (2.21a)
Y = A2[exp(r?1)+p1l)]/p12 +
A( 1A) [lexp( /7X)2. [lexp(rj2)]/p1p2 +
(lA)2[exp(p2)+n2l]/p22. (2.21b)
When we take the ratio of gr and g0, a useful expression is obtained:
9r tB
= B (2.22)
9o tb
where tB is the base minority carrier transit time, xB is the chargecontrol minority
carrier base recombination lifetime, and Bisa factor related to the builtin electric field
in the base due to both Ge and doping nonuniformity (B=1.0 in fieldfree base). B is a
function of rij, r2, nf(x), and hole concentration in the base P(x) and is given by (2.23)
below [39]
1
B =
iV(x)
dx
dx
0 PB(x)
0 ni2(x)/P(x)
1
(2.23)
93
X/(1)
Fig. B.l Gaussian distribution and exponential approximation.
LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) CHARACTERIZATION
OF Si/Ge^i^/Si HETEROJUNCTION BIPOLAR TRANSISTORS
By
GUOXIN LI
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
1998
50
Table 2.1 Summary of experimental data and calculated parameters for Si/GexSi1_x/Si HBT
and reference Si BJT. The top six lines give the measured dc and BTM ac data obtained at
23C, VBE = 0.52V and VCB = IV. The computed parameters using LFGV method are
shown in the next eighteen lines. The first column is for Si/GexSij.x/Si HBT, the second
for Si baseline BJT, and the third column gives the comparison ratio of the two previous
columns. ABE=100.4xl00.8pm2, XR=680, XB=800.
Parameter
GeSi HBT
Si BJT
GeSi/Si
Ic (PA)
7.74
200.5
3.86
IB (nA)
59.4
69.9
0.85
Pf
130.3
28.6
4.54
go (I/O)
2.125x1o8
1.67xl08
1.27
gr (1/G)
3.0xl011
1.8xl010
0.17
go^gr
708.3
92.78
7.63
^2
2.7
2.7
Bi
1.98
AEG_Ge (meV)
68.6
Ebi (kV/cm)
10
B
12.6
2.6
4.8
Lb (!im)
3.6
0.8
4.5
Qb^b (A)
0.87
8.4
0.1
Qe/te (nA)
58.5
50.6
1.16
(Qe^Wb)
67
6
11.2
(Qe/te)/Ib (%)
98.5
88
1.12
(Qb/tb)/Ib (%))
1.5
12
0.13
Qb (C)
21xl018
10xl018
2.1
Db (cm2/V*s)
5.5
5.5
1.0
tB (Ps)
2.7
4.8
0.56
tb (ns)
24
1.1
22
Ay
0.47
AEg.hd (meV)
12
VA (V)
364
120
3.0
102
[16] R. People, "Indirect band gap of coherently strained Ge^i^ bulk alloys on
<001> silicon substrates," The American Physical Society, vol. 32, No. 2, pp.
14051408,July. 1985.
[17] L.E. Kay and T.W. Tang, "Monte Carlo calculation of strained and unstrained
electron mobilities in GexSi(1_x) using an improved ionizedimpurity model," J.
Appl. Phys., vol. 70, No. 3, pp. 14831488, Aug. 1991.
[18] S.K. Chun and K.L. Wang, "Effective mass and mobility of holes in strained
Sij_xGex layers on (001) Sij.yGey substrate," IEEE Trans. Electron Devices, vol
39, pp. 21532164, Sep. 1992.
[19]B. Pejcinovic, L.E. Kay, T.W. Tang and D.H. Navon, "Numerical Simulation and
Comparison of Si BJTS and Sij.xGex HBTs," IEEE Trans, electron Devices, vol
36, pp. 21292137, Oct. 1989.
[20] A. Schiippen, U. Erben, A. Gruhle, H. Kibbel, H. Schumacher and U. Kdnig,
"Enhanced SiGe heterojunction bipolar transistors with 160GHzfmax," IEDM
1995, p. 743746.
[21] E.J. Prinz, P.M. Garone, P.V. Schwartz, X. Xiao, and J. C. Sturm, "The effect of
baseemitter spacers and straindependent density of states in Si/Sij.xGex/Si
heterojunction bipolar transistors," IEDM Tech. Digest, pp. 639642, 1989.
[22] S.C. Jain, J. Poortmans, S.S. Iyer, J.J. Loferski, J. Nijs, R. Mertens, and R. van
Overstraeten, "Electrical and optical bandgaps of GexSij_x strained layers, IEEE
Trans. Electron Devices, vol. 40, pp. 23382343, Dec. 1993.
[23] J.M. Early, "Effects of spacecharge layer widening in junction transistors,"
Proceedings of the IRE, pp. 14011406, Nov. 1952.
[24] M.S. Birrittella, A. Neugroschel, and F.A. Lindholm, "Determination of the
minoritycarrier base recombination lifetime of junction transistors by
measurements of basewidthmodulation conductances, IEEE Electron Device
Lett., vol. 26, pp. 13611363, Sep. 1979.
46
conductance, g0, reverse transconductance, gr, for both the forward active mode and
reverse active mode and the collector current Ic of Si BJT. The measured dependencies
in Figs. 3.43.6 are used for the LFGV analysis. The results for large square device
(100.4x100.8p,m2) are summarized in Table 2.1. A detailed discussion of the analysis
and its implications follows.
The first step in the analysis is to find the parameter r2 using (2.27) which
determines the builtin field for XR
ln[NAA(XR)/NAA(XB)] = 2.7 for both Si BJT and GeSi HBT. This corresponds to
NAa(Xr)/Naa(Xb) = 15, which agrees well with the SIMS profiles in Fig. 1.2.
With the known r2, we can then calculate the key parameter r j from the measured Ic
and g0 for the GeSi and Si transistors using (2.18a). This gives = 1.98, and r\ = (r^ +
C) = 2.70.1 using C = 0.70.1. The Geinduced energy gap reduction at xGe=9.8% is
^GGe = (kT)r) = 705meV. This corresponds to AEg_ge = 7xGE meV which is in very
good agreement with the theoretical calculations and other experiments in GeSi as a
function of xGe. This good agreement demonstrates the accuracy of our model that
accounts for the dependence of the NCNV on Ge concentration. The value of the fitting
parameter, C, causes only small uncertainty in the results. The Geinduced aiding
electric field within the linear Ge range, 0
The mobility ratio DB(GeSi)/DB(Si) can be evaluated from (2.15). The remaining
unknowns in (2.15) are a = AEG_Ge(0)/kT and Ay = AEE_HD/kT. SIMS profiles indicate
a = 0 at Vbe=0.52V. This gives [DB(GeSi)/DB(Si)]exp(Ay) = 1.6. This product cannot
be separated without additional independent measurement of either the mobility ratio or
85
Table 3.4 Summary of experimental data and calculated parameters for Si/GexSij_x/Si HBT
by LFGV methodology with no Si reference BJT. The top six lines give the measured dc
and BTM ac data obtained at 23C, VBE = 0.52V and VCB = IV. The computed parameters
are shown in the next sixteen lines. ABE=100.4xl00.8pm2, XR=680, XB=800.
Parameter
GeSi HBT
Ic (PA)
7.74
IB (nA)
59.4
3f
130.3
g0 (m
2.125xl08
gr (l/Q)
3.0xl0~n
go^gr
708.3
112
2.6
Til
2.0
AEQ_Ge (meV)
69
Ebi (kV/cm)
10
B
12
Lb (pm)
3.5
Qb^b (ttA)
0.91
Qe/te (nA)
58.5
(Qe/xe)/(Qb/t;b)
65
(Qe/te)/Ib (%)
98.5
(Qb^b)Ab (%))
1.5
Qb (C)
21xl018
Db (cm2/Vs)
5.5
tB (PS)
2.8
tb (ns)
23
VA (V)
364
77
C=0.7 is used. The aiding electric field Ebi = (kT/q)(r/XR) = (0.0255)(2.74/680xl(T8) =
10.3kV/cm. This value compares very well to Ebi=10kV/cm which is obtained by the
comparative method discussed in chapter 2 (section 2.5) using the comparison LFGV
methodology.
In conclusion, the above example demonstrates one application of the new
measurement method, namely the measurement of the energygap variation due to the
compositional grading across the GeSi base. Similar results can be obtained for IIIV
HBTs with graded base. The new nonreference methodology can be also used to
measure the temperature dependence of the parameter r\,. GeSi HBT shows a superior
low temperature Ic enhancement compared to Si BJT. This is because the builtin
electric field Ebi is inversely proportional to the temperature, Ebi=AEG_Ge/kT, where k is
the Boltzmann constant and AEG_Ge is the Geinduced energy gap reduction across the
quasineutral base region. Thus, at lower temperature, the builtin electric field is greatly
enhanced. All other parameters related to Ebi will be also enhanced. For example, the
collector current Ic(GeSi) will increase, the output conductance g0(GeSi) will decrease,
and the base minority carrier transit time tB will decrease significantly which gives
higher speed and fT. Theoretical calculations of the minority mobility as a function of
temperature and doping were presented in the literature [17, 5253]. In order to verify
these theoretical predictions and analyze the GeSi HBT performance at lower
temperature, the determination of the temperature dependence of r) j is very important.
This was investigated by measuring the GeSi HBT characteristics at several different
low temperatures. Using the models and equations given in the section 3.2, we can
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
LOW FREQUENCY CONDUCTANCE VOLTAGE (LFGV) CHARACTERISTICS
OF Si/GexSi1.x/Si HETEROJUNCTION BIPOLAR TRANSISTORS
By
Guoxin Li
December 1998
Chairman: Dr. Amost Neugroschel
Cochairman: Dr. ChihTang Sah
Major Department: Electrical and Computer Engineering
Low Frequency Conductance Voltage (LFGV) methodology for analysis of
heterojunction bipolar transistors (HBTs) is presented. The LFGV methodology is
based on a combined measurement of the dc collector and base currents and low
frequency (~lkHz) smallsignal conductance resulting from the base thickness
modulation. It gives accurate and independent determination of the builtin electric
field resulting from the spatially varying energy gap and dopant impurity concentration
profile in the base layer. The LFGV methodology also gives accurate quantitative
values for the minoritycarrier transport parameters, including the minority carrier
diffusion length and recombination lifetime, and a complete charge control analysis of
the recombination losses in the emitter and base region of the HBT.
v
ln[exp(Ay)DB(GeSi)/DB(Si)]
52
1000/T /(1/1K)
Fig. 2.6 The measured temperature dependence of ln[exp(Ay)D B(GeSi)/DB(Si)] in the
range of 200K300K normalized to 296K in large 100.4x100.8pm2 transistor.
66
Til
Fig. 2.14 Base minority carrier charge ratio QB(GeSi)/QB(Si) vs r) j for the forward
active mode (normalized to exp(a+Ay)).
19
1.6 Energy Band Diagram
In Si/Ge^i^/Si HBT, a Ge^i^ base layer with a smaller energy gap than that of
pure Si is sandwiched between the Si emitter and collector. The energy band diagram is
very important for analysis and design optimization of the GeSi HBT. Figure 1.12, Fig.
1.13, Fig. 1.14, and Fig. 1.15 show the energy band diagrams for triangular Ge
distribution and trapezoidal Ge distribution in the base region at equilibrium (zero bias)
and forward active mode (VBE>0V, VCB<0V). Also from Fig. 1.7, we can see that for
Ge mole fraction xGe<~35%, the energy gap reduction AEG_Ge is almost a linear
function of xGe. For trapezoidal Ge distribution, we have
xGe(x) = xGe(0) + [xGe(XR) xGe(0)]x/XR (0
Eg(x) = Ego AEg_hd AEG_Ge(0) AEG_Gex/XR (0
where AEG_Ge = AEG_Ge(xR) AEG_Ge(0) (1.4)
is the total Geinduced energy gap reduction through the base region. AEG_HD is the
energy gap reduction due to heavy doping and AEG.Ge(0) is the energy gap reduction
due to Ge at x=0 in the quasineutral base region. The Ge gradient will produce an
aiding electric field Ebi:
Ebi = (l/q)(dEc/dx) = (l/q)(AEG_Ge/XR) = independent of x
= (kT/q)(tj/xR) 0
where r) = AEG. Ge^kT'
4. SUMMARY AND CONCLUSIONS
87
4.1 Summary 87
4.2 GeSi HBT Modeling and LFGV Methodology 88
APPENDIX A GENERAL EQUATIONS FOR B JT 89
APPENDIX B TRANSISTOR RELATION FOR A GAUSSIAN DOPANT IMPURITY
CONCENTRATION PROFILE IN THE BASE OF BJT 92
APPENDIX C LIST OF SYMBOLS 98
REFERENCES 100
BIOGRAPHICAL SKETCH 107
IV
64
Til
Fig. 2.12 Output Conductance ratio g0(GeSi)/g0(Si) vs r( j for the forwardactive mode
(normalized to [DB(GeSi)/DB(Si)]exp(a+Ay)). The g0 ratio shows a rapid decrease for
1.0
GeSi HBT is much higher than that of Si B JT for large r\ t and the Early voltage VA also
increases rapidly with r) j. Thus the figureofmerit, (3VA, improves significantly for
large r j for the GeSi HBT.
47
Vbe/(1V)
Fig. 2.3 IcVBE and IBVBE characteristics of GeSi HBT and Si BJT. The
measurements were taken on large transistors with the EB junction area of
100.4x100.8um2 in order to eliminate the edge and sidewall effects. Note the
significant increase in Ic and slight decrease in IB for the GeSi transistor. The measured
enhancement in the collector current Ic(GeSi)/Ic(Si) = 3.86.
31
AXB
(b)
Fig. 2.1 (a) Crosssectional view of an n+/p/n BJT. Smallsignal voltage vcb is
superimposed on dc value VCB.
(b) Minoritycarrier distribution in the base.
34
base layer. The second method using only the GeSi HBT without the reference Si BJT
will be discussed in chapter 3.
There are several performance enhancement gauges of GeSi HBT over Si BJT. The
most important and also most commonly used by investigators is the dc collector
current Ic enhancement [3536], For example, for a uniform base doping and uniform
Ge concentration in the ptype base of npn transistor, the ratio of the collector currents
can be written as
Ic(GeSi) (NcNv)GeSi Dn(GeSi)
= exp(Ay) exp(AEG_Ge/kT) (2.6)
IC(Si) (NCNV)S1 Dn(Si)
where Nc, Nv are the density of states of the conduction band and valence band,
respectively, Dn is the minoritycarrier electron diffusivity, AEG_Ge is the Geinduced
energygap reduction in the base, and Ay = AEG_HD/kT=[AEG_HD(GeSi) 
AEGHo(Si)]/kT is the normalized difference between the heavy dopinginduced energy
gap reduction in GeSi and Si base layers.
There are four unknowns in (2.6) all related to the Ge in the GeSi base layer. The
key issue is how to separate these Geinduced effects. Two approaches have been
presented in the literature using the Ic ratio in (2.6) [2932], The first approach uses the
measurement of the temperature dependence of Ic(GeSi)/Ic(Si), assuming that the
density of states and mobility ratios are temperature independent. Then by plotting the
Ic ratio versus 1/T, equation (2.6) yields an activation energy EA = AEG_Ge + AEG_HD =
AEGGe This AEC,C can then be used to calculate the improvements in the current
gain, transit time, and other characteristics as a function of the Ge mole fraction and
12
z*
>
X
Emitter
f f^zz
PGeSi base
^Mxx
Collector
Fig. 1.6 Direction illustration in vertical GexSij_x HBT. The z direction corresponds to
the growth direction and is also the direction of the minority electron flow in a vertical
BJT. The inplane direction is important and applicable for lateral MOS transistors.
84
3.4 LFGV Analysis of GeSi HBT with no Si Reference
After the determination of r), described in the previous section, a complete charge
control analysis of the GeSi HBT can be made. The only other parameters needed for
this analysis are r\2 which determines the builtin electric field in the region of
XR
equation (2.22) is used to calculate the parameter r\2. From the measurement of the
output conductance, g0, and reverse transconductances, gr, for the forward (subscript F)
and reverse (subscript R) modes of operations and taking the ratio of equation (2.22) for
these two modes:
90f/90r tBF
. (3.10)
9rF/,9rR Br tBR
The ratio in (3.10) is a function of rj j, A=XR/XB and r\2 only. Since rjj can be
determined independently as described in section 3.2 and XR/XB is known from SIMS
or estimated from fabrication conditions, r\2 can be then calculated from (3.10). Using
the data in Table 2.1 and ^Â¡=2.04, we obtained r)2=2.6 which agrees very well with the
comparative method in section 2.5 which gives r\2=2.7. We assume that
DB(GeSi)=DB(Si) as determined in chapter 2. The results of a complete charge control
analysis for the GeSi HBT following the procedure described in chapter 2 are
summarized in table 3.4. Since r)j=2.04 and ri2=2.6 obtained using the nonreference
analysis are almost the same as those obtained using the comparative LFGV
methodology in section 2.5 (r)j=1.98 and r)2=2.7), the results in table 3.4 are very
similar to those in table 2.1.
33
VBE, g0 and gr are also exponential functions of VBE. Taking the ratio of equations (2.3)
and (2.4), we obtain a very important relation for the analysis of the bipolar junction
transistor [1, p. 822, 2425]:
90 Ic Qb^b tb
= = = (2.5)
9r Qb^ t b Qb^ tb tB
where tB is the base minority carrier transit time.
In order to accurately measure the output conductance g0 and especially the reverse
transconductance gr (grg0), a precise equipment is needed. Two instruments were
used to measure g0 and gr, a manual Wayne Kerr bridge B214 and an automated IEEE
bus controlled HP7245A LCR meter. Most of the data were obtained using WK B214
because of its high conductance sensitivity (~10"12 l/Â£2) required for gr measurement in
bipolar transistors with high base doping where the BTM is very small.
2.3 LFGV Methodology
Two LFGV methodologies were developed. The first LFGV methodology described
in this section is based on a comparison of the transistor characteristics measured in
Si/GexSij_x/Si HBT and reference Si BJT where the difference between the measured
characteristics is due to Ge only [2627]. By analyzing this difference, we can then
calculate the various Geinduced material parameter changes. This approach has been
commonly used by other investigators [2834], The basic assumption in this
comparative approach is that the two transistors have nearly identical geometry and
dopant impurity concentration profiles except the Ge incorporation in the GeSi HBT
Majority electron mobility /(cm2/V*S)
18
Fig. 1.11 Majority electron mobility pn in ntype Ge^i^ layer. The upward triangles
are in the perpendicular z direction which is also the growth direction, and the
downward triangles are in the xy plane. Adapted from L.E. Kay and T.W. Tang [17].
Minority electron mobility /(cm2/V*S)
17
0 0.05 0.10 0.15 0.20
Ge mole fraction /(1%)
Fig. 1.10 Minority electron mobility p,n in ptype GexSii_x layer. The upward triangles
are in the perpendicular z direction which is the growth direction and also the electron
movement direction in vertical BJT. The downward triangles are in the xy plane.
Adapted from L.E. Kay and T.W. Tang [17].
44
= X/Y (2.24a)
exprjiin1
= XR = XB. (2.24b)
exp(r?i)+/jil
A combination of the dc IBVBE and smallsignal measurements gives the expressions
for the base diffusion length and ratio of the recombination losses in the emitter and
base region:
Qe/te B(g0/gr)
= 1 (2.25)
Qb^ t b Â£f
L = XB2(g0/gr){A2(exptj1tj1l)/f]12 +
A(lA)[(exp^1l)/ni] [(exprj2l)/n2] +
(lA)2(expr72n2l)/h22} (2.26a)
r/1+exp(r7x)1
Lb2 = XB2(g0/gr). XR = XB. (2.26b)
ni2
A combination of (2.19) and (2.25) gives (Qs/^oeSi ar*d (Qb/tb)s from measured
data. Using (2.26a) or (2.26b) we then calculate directly the diffusion length where the
only unknown parameter is XB which can be measured by SIMS.
Expression (2.22) is also used to calculate the parameter r9 which determines the
builtin field for XR
Measuring the output and reverse conductances for the forward (subscript F) and reverse
(subscript R) mode of operations we obtain the following ratio:
Sof/QoR Bp tgp
Br tBR
StF^OpR
(2.27)
65
Fig. 2.13 Base minority carrier distribution of GeSi F1BT and Si BJT vs x/XB for
r1=2.0 in forward active mode (normalized to n[exp(qVBE)l]/NB0).
Cutoff Frequency fT /(1GHz)
15
Collector Current lc /(1mA)
Fig. 1.9 Typical unit gain frequency fT versus Ic characteristics of GeSi HBT (full
circles) and Si BJT (open circles). The GeSi HBT and Si BJT have the same dopant
impurity concentration profile and geometry except the Ge in the base for GeSi HBT.
Adapted from Merit Hong et al. [11].
4
pseudomorphic growth. In pseudomorphic growth, if the GeSi layer thickness is below
a critical thickness, the lattice mismatch is accommodated elastically with no dislocation
created. The GeSi layer is under biaxial strain both in the xy plane and in the z or
growth direction.
Several epitaxial techniques for GeSi layer deposition were developed, such as the
molecular beam epitaxy (MBE) [910] and limited reaction process (LRPCVD) [11].
Figure 1.1 shows a HBT transistor structure used in this thesis which was fabricated by
chemical vapor deposition (CVD) [11]. The singlecrystal strained GeSi layer was
grown over the nSi collector to give the intrinsic base, and also on the LOCOS Si02
layer which becomes polycrystalline and is the extrinsic base regions. This results in
an interface between the intrinsic and extrinsic base regions. Germane is the source of
Ge during the growth and boron impurity doping was obtained with diborane. Boron
was then implanted to give the extrinsic base. The oxide/nitride layer were deposited
and etched to open the emitter region. The polysilicon layer was then deposited and
implanted by a high concentration of arsenic and patterned to form the emitter. The
impurity and GeSi concentrations were measured by SIMS (secondary ion mass
spectroscopy) as shown in Fig. 1.2. The impurity and Ge concentration profiles can be
precisely controlled during film growth to give specified Ge distribution in the base,
such as a linearly retrograded Ge gradient. This Ge gradient produces an aiding electric
field which accelerates the minority electrons in the base layer.
95
transistors with a Gaussian or Gaussianlike profile in the base with a small error
compared with accurate numerical solution.
II. Following are the equations for the Gaussian distribution P(x) shown in Fig. B.2.
1.Base dopant impurity concentration profile P(x)
P(x) = NM.exp[(xX0)2/cr2]
where x0 is the peak value position of the P(x) and cr is equal to
(XXn)2
CT2 =
ln[NM/P(XB)]
XB
r P(x)dx =
Jo
^ tt/2ctNaa{erf[(XBX0)/o] + erf(X0/cr)}
(B.9)
(B.10)
(B.11)
2. Collector current
qD^2 [exp(qVBE/kT) 1 ]
Ic = ttz (B. 12)
l^/2.(rNAA.{erf[(XBX0)/c7] + erf(X0/cr)}
3. Minority base transit time
i ncr
tB= {erf[(XBX0)/(j]JB exp[(xX0)2/cj2]dx 
2Db
JB exp[(xX0)2/CT2]erf[(xX0)/cr]dx} (B.13)
4. Correction factor
erf(X0/cj) jBexp[(xX0)2/cr2]dx + JBexp[(xX0)2/cr2]erf[(xX0)/o]dx
B = (B. 14)
erf[(XBX0)/cr]jBexp[(xX0)2/(T2]dx jBexp[(xX0)2/o2]erf[(xX0)/cr]dx
5. Output conductance
2IC exp[(XBX0)2/cr2] 0XB
g0 =
7Tcr erf[(XBX0)/cr] + erf(X0/
(B.15)
79
Since the px product is not strongly temperature dependent [5455], LB(GeSi)WT, or
increasing with increasing T, as shown in Fig. 3.4.
From Fig. 3.5, we can see that the base recombination component (QB/xB)GeSi/IB
increases with decreasing temperature. Thus, the base recombination becomes more
important as temperature decreases.
Table 3.1 Summary of measured parameters of GeSi HBT.
Â§oF
2.125xl08(S)
Â§oR
6.549xl06(S)
IcF
7.739 (pA)
!cr
12.45 (pA)
Cbc
1.446 (pF)
Cbe
39.676 (pF)
51
Ay. Figure 2.6 shows the measured temperature dependence of
ln[exp(Ay)D B(GeSi)/DB(Si)] in the range of 200K300K. The temperature dependence
shows a small activation energy of EA~12meV. This is consistent with Ay = AEG.HD/kT
= 0.47 or AEg.hd = 12meV which is calculated from the measured room temperature Ic
ratio assuming a unity mobility ratio which is independent of temperature. A negligible
improvement in the minoritycarrier mobility for the average ~5% Ge concentration and
~3xl018cm"3 base doping at room temperature was also predicted theoretically as is
discussed in section 1.5. The conclusion that DB(GeSi) = DB(Si) is also supported by
the sparameter measurements of the base transit time for HBTs. Using the measured
data in Fig. 2.6, we can calculate the temperature dependence of the minority carrier
diffusivity ratio assuming that AEG_HD=12meV is constant with the temperature range
of 200K296K. The results are shown in Fig. 2.7 and are in good agreement with
theoretical predictions [17].
The recombination components of IB are calculated from (2.19) and (2.25). The
calculation shows that the recombination losses in the emitter and base constitute 98.5
and 1.5% of IB, respectively in GeSi HBT and 88% and 12% in Si BJT. An important
selfconsistency result is that (QE/TE)Gesi = (Qe/te)sp as anticipated since xGe=0 at the
EB junction and, thus, the hole injection from the base into the emitter and the emitter
recombination are expected to be the same in both transistors. The small differences
can be explained by differences in the emitter profile and recombination properties at
the n+polysilicon/n+Si emitter interface. The very significant reduction in the
recombination losses in the base, (Qb/^GcS^Qb^b^ 0.1 is due to a significantly
90
6. Output conductance g0
3IC dXg
9o
axB av,
CB
XB P(x)
dx
0 ni2(x)
7. Reverse transconductance gr
a (Qb/ tg) 8XB
8 Xr
av,
CB
l P(XB). axB
ni2(XB) avCB
(A.6)
(A.7)
8. Current gain
q
I BO
P(X)
0 Dn(x).ni2(x)
dx
where IB0 is the base saturation current.
9. Early voltage VA and (3VA product
Va =
qni2(XB)Dn(XB)
9o
JBC
P(x)
0 Dn(x) ni2(x)
dx
!bo*cbc
.n?(XB).Dn(Xg)
(A.8)
(A.9)
(A.10)
B. Equations for a trapezoidal Ge distribution and constantexponential dopant
impurity concentration profile in the base of GeSi HBT (Fig. 2.2) obtained from the
general integral relation (A.l)(A.IO) are given below.
83
Fig. 3.5 Base recombination component QB(GeSi)/tB(GeSi) vs 1000/T.
71
chargecontrol analysis of the recombination components in the base current [47]. The
new method is demonstrated for Si/Ge^i^/Si HBT with a linear or trapezoidal Ge
concentration profile in the base. The Geinduced energygap reduction parameter r\ is
determined solely from the terminal measurement of GeSi HBT without any
assumptions. The builtin electric field Ebi can be then calculated, independent of the
minority carrier mobility and heavydoping effects. The method also does not require a
knowledge of the dopant impurity concentration profile in the base.
3.2 Theory
The SIMS profile for the analyzed GeSi HBT is shown in Fig. 1.2 [48]. The Ge is
linearly graded within 0
varies as EG_Ge(x)=EG .Ge(0)AEG.Ge(x/XR) [4951], where EG(0) is the energy gap
magnitude at the EB junction edge of the quasineutral base, x=0, XB is the base
thickness, and AEG_Ge=AEG.Ge(XB)AEG_Ge(0) is the total variation of the energy gap
due to the Ge gradient. This gives a spatiallyconstant aiding electric field for the
electrons in the ptype base in the linearlygraded region, Ebi(x) = (l/q)(AEG/XR) =
(kT/q)(ri/XR) = independent of x, where r\ is the normalized energy gap drop through
the base, r=AEG_Ge/kT.
The collector current can be expressed for any base dopant impurity concentration
profile and compositional variation in the GeSi base as [1, p. 760, 38]
q*exp(qVBE/kT)
Jc = (3.1)
XB P(x)
dx
.0 Dn(x) n12(x)
91
1. Collector current Ic
qDnnl02 exp(a)exp( v)exp(/71) exp(qVBE/kT)
jc=
NBoXB[(XR/XB) (exprji1)///!+( 1Xr/Xb) (lexp(^2) )//j2]
2. Minority carrier base transit time tB
Xb2
tB = Y
Dn
3. Output conductance g0
XB A(expr?1l)/i?1)exp^2 + (lAKexp^l)/^
4. Reverse transconductance gr
Qb X 1
0Xr
av,
CB
axB
9r *
tbXb Y A[(exp^1l)/^1)]expn2 + (lA)[(exp/j2l)/ff2] aVCB
In (A. 1.9) and (A. 1.11), the terms X and Y are given below
X = A2(exp^1+/71l)/n12 + A(lA)[(expHil)//?!.] [(exp/?2l)/r?2] +
+ (lA)2(expri2ri2l)/ri22
Y = A2[exp(/71)+/71l)]/r712 + A( 1A)[1expC^j_)] [l(expr}2)l/rÂ¡inz+
+ (lA)2[exp(ri2)+ri2l]/ii22.
5. Correction factor B
%
B = X/Y
6. PVA product
Va = XB
exp^l expn2l
A exprjg + (1A)
'll
12
BO CBC
ni^ Dn(XB) expr?i
1
axB/avCB
(A.11)
(A.12)
(A.13)
(A.14)
(A.15)
(A.16)
(A.17)
(A.18)
(A.19)
54
larger base minority recombination lifetime tb ratio TB(GeSi)/TB(Si) = 24ns/l.lns = 21.
The reasons for the significant improvement of the minoritycarrier recombination
lifetime in thin strained GeSi layers is not known and requires further study. The
diffusion length ratio can be calculated from (2.26) and gives LB(GeSi)/LB(Si) =
3.6pm/0.8 = 4.5 in the GeSi layer. This ratio is calculated directly from the smallsignal
BTM measurements without using the dc Ib"Vbe curve. This is an advantage for HBTs
that have a large BE junction leakage or large spacecharge layer recombination
component. xB was calculated taking DB=5.5cm2/s [27] for both transistors. Although
tb is relatively short (=20ns), it has only a small effect on the gain since IB in the large
100.4x100.8p,m2 transistors is dominated by the emitter. Rough estimates of tb in GeSi
HBTs were published previously [4244], but our present measurements in large
transistors are the first rigorous accurate data.
The minority carrier base charge QB is larger in GeSi transistor, QB(GeSi)/QB(Si) =
2.1 because of smaller energy gap in the base which increases nÂ¡2. This increase in the
stored charge should be accounted for in speed calculations. The electron transit time
across the quasineutral base layer is then calculated and gives 2.7ps/4.8ps=0.56, a factor
of about two improvement for the GeSi HBT, as expected. This is due almost entirely
to the builtin field from the graded Ge concentration profile rather than the mobility
enhancement which is negligible. The Early voltage for GeSi HBT is significantly
higher which is important for analog circuit application.
The results for a stripe transistor with narrow emitter fingers (0.8x26im2 with 88
fingers) are summarized in Table 2.2. The major difference between large square
40
the calculation of other GeSi parameters, such as the transit time. Thus, the analysis of
the GeSi HBT with a trapezoidal or linearlygraded Ge concentration profiles can be
conveniently made in terms of a single parameter r^. It is important to note here that, as
will be shown later, the value of the fitting parameter C = (rj rjx), which is model
dependent, does not affect the determination of the base diffusion length, recombination
lifetime, transit time, or diffusivity. It only affects the calculation of the builtin electric
field Ebi and AEG_Ge, which may give a small error due to the uncertainty of the fitting
parameter C. However, this analytical approach is much more accurate than using the
spatially averaged value of NCNV in the integral equation (2.9), which allows taking
NCNV outside the integral. Another very important feature of the new expression (2.15)
is that it decouples the (NCNV) product from mobility in (2.12) and allows
determination of the mobility ratio.
In order to analyze the Geinduced Ic enhancement, we have to separate the
measured ratio in (2.15) or (2.16) into five or four terms. A rough approximation for
practical profiles with xGe(0)=0 is to assume that the enhancement in Ic is dominated by
the builtin field which leaves only the first term in (2.16) and leads directly to the
estimation of r^. However, for an accurate analysis of the transistor and for purpose of
acquiring a material database, more accurate approach is necessary. For example, an
uncertainty in AEG_Ge(0) or AEG.HD of lOmeV gives an error of ~exp(10/25)= 1.5. Thus,
an analysis of the performance of Si/GexSi1.x/Si HBT based on measurement of Ic only
is clearly not sufficient.
29
improvements in collector current, transit time and output conductance in GeSi
transistors over those in similar Si transistors were demonstrated qualitatively and
graphically. In this chapter, we present detailed analysis based on an analytical model,
to extract reliable and accurate values of the minoritycarrier parameters in the GeSi
material. These parameters are essential to assess the Geinduced effects on transistor
performance, design optimization and to verify theoretical predictions about the
expected performance benefits due to Ge.
The key point in obtaining these quantitative values is a separation of the various Ge
induced effects from the measured data. A lack of an experimental methodology for
this analysis is the main reason why these values are still not available in the literature.
For example, in GeSi transistors, the increase in the collector current Ic can arise from
the aiding buildin electric field in the base due to the Ge retrograded profile, the base
minoritycarrier diffusivity enhancement, or both. An accurate separation of the
performance improvement between the two parameters requires independent
measurements of the builtin field and mobility. Yet another factor that affects the
collector current is the reduction of the effective density of states, Nc and Nv, which are
functions of the effective masses mn and mp, respectively. Theoretical calculations
show that the density of states product, NCNV, in thin pseudomorphic Ge^i^ layers
grown on (100) Si substrate is significantly decreased compared to that in Si due to Ge
induced strain [1718]. This gives (NcNv)GeSi/(NcNv)Si ~ 0.6 for GeSi layer with xGe =
5% [1718]. The minoritycarrier mobility was predicted to increase with Ge
concentration for xGe~<10% especially at higher base impurity concentration (1019
94
2. Collector current
qDnn.Â¡2. [exp(qVBE/kT)l]
Ic = 7= (B.4)
^/2.crNAA.erf(XB/cr)
3. Minority carrier base transit time
\ na
tB= [erf(XB/cr) JBexp(x2/
2.Dn
4. Output conductance
2IC exp(Xb2/ct2) aXB
90 = p (B.6)
i;:.<7 erf(XB/cr) aVCB
5. Reverse transconductance
Ic exp(Xg2/cr2) aXB
gr = [JBexp(x2/cr2)erf(x/cr)dx]. (B.7)
DntB erf(XB/cr) aVCB
6. Correction factor
JBexp(x2/a2)erf(x/cr)dx
B = (B.8)
erf(XB/cr) jBexp(x2/<72)dx jBexp(x2/<72)erf(x/o)dx
For the profile in Fig. B.l, the value for the correction factor is obtained numerically
using (B.8) which gives B=7.0. The exponential approximation gives B=6.0 using
much simpler analytical expression (A. 17). Thus, the exponential approximation of the
Gaussian profile in the base of a bipolar junction transistor gives only a small error for
the value of B. This allows a simple charge control analysis of bipolar junction
transistor using the exponential profile approximation shown in sections 2.3 for
48
Vbe /< 1V)
Fig. 2.4 The measured forward active mode and reverse active mode smallsignal
output conductances, g0, reverse transconductance, gr, and the collector current, Ic, of
GeSi HBT, as a function of VBE. All measured curves show the expected exp(qVBE/kT)
dependence.
(VU)/ l
74
3.3 Experiments and Results
A high performance Si/Ge^i^/Si HBT is used to demonstrate the application of the
result in equation (3.7). The Ge concentration profile is shown schematically in Fig.
1.2. The Ge gradient from SIMS showed a linearlygraded Ge concentration from about
xGe=0% at the EB junction to a peak of about 10% at the BC junction. The base dopant
density was nearly constant at =3xl0iscm3, and the quasineutral base thickness was
about 800. The experimental data is tabulated in Table 3.1. All the measurements
were performed on large area 100.4x100.8ptm2 transistors to eliminate the perimeter
edge and twodimensional effects. The difference between ICF and ICR is due to slightly
larger base thickness for the reverse mode, which is expected for very thin base. The
output conductances goF and goR, and the junction capacitances CBC and CBE were
measured at 1kHz using precise bridge WK B214.
With reference to the profile in Fig. 1.2 and the intrinsic carrier concentration model
for linear Ge concentration profile in equations (2.10), and assuming that the difference
in the intrinsic carrier density at x=0 and x=XB is due to Ge grading, we obtain from
(3.7),
n?(XB)
n?(0)
= exp?^ = exp(r? C) = exp [(AEG_Ge/kT) C],
% 90r cbc
*CR 9oF CBE
(3.8a)
(3.8b)
where C is the least squares fitting parameter for the density of states product NCNV for
linear Ge distribution discussed in detail in section 1.7. Using the data in Table 3.1 we
obtain, nf(XB)/nf(0) = 7.68 from (3.7). This gives ^=2.04 and AEG.Ge = 70meV where
Emitter
Fig. 1.1 Cross section view of GeSi HBT. Adapted from Neugroschel et al. [1112].
REFERENCES
[1] C.T. Sah, Fundamentals of SolidState Electronics, River Edge, NJ and
Singapore: World Scientific, Nov. 1991. See sections 770774 on pp. 931954,
on history and operation principle of GeSi bipolar junction transistors.
[2] W.B.Shockley, "Circuit element utilizing semiconductive materials," U.S. Patent
2569347. Filed Jun. 26, 1948, granted Sep. 25, 1951, expired Sep. 24, 1968. (17
year statue of limitation.)
[3] Herbert Kroemer, "Heterostructure bipolar transistors and integrated circuits, "
Proc. IEEE, vol. 70, No. 1, pp. 1325, Jan. 1982.
[4] W.P. Dumke, J.M. Woodal, and V.L. Rideout (IBM), "GaAsGaAsAl
heterojunction transistor for high frequency operation, SolidState Electronics
15(12), 13391345, Dec. 1972.
[5] S.S. Iyer, G.L. Patton, J.M.C. Stork, B.S. Meyerson, and D.L. Harame,
"Heterojunction bipolar transistors using SiGe alloys, "IEEE Trans. Electron
Devices, vol. 36, pp. 20432064, Oct. 1989.
[6] D.L.Harame, J.H. Comfort, J.D. Cressler, E.F. Crabb, J.Y.C. Sun, B.S.
Meyerson, and T. Tice, "Si/SiGe epitaxialbase transistorsPart I: Materials,
physics, and circuits," IEEE Trans. Electron Devices, vol. 42, pp. 455408, Mar.
1995.
[7]D.L.Harame, J.H. Comfort, J.D. Cressler, E.F. Crabb, J.Y.C. Sun, B.S.
Meyerson, and T. Tice, "Si/SiGe epitaxialbase transistorsPart II: Process
integration and analog applications," IEEE Trans. Electron Devices, vol. 42, pp.
469481, Mar. 1995.
100
BIOGRAPHICAL SKETCH
Guoxin Li was born in Beijing, P.R. China, in 1964. He received the Bachelor of
Science degree in electrical engineering with high honors from the TsingHua
University, Beijing, P.R. China, in 1987 and was recommended to enter the Graduate
School of TsingHua University without any examination. He received the Master of
Science degree in electrical engineering in 1989 with high honors and was also awarded
the XinHua Prize for the outstanding graduate electrical engineering student. He will
receive the Doctor of Philosophy degree in electrical and computer engineering at the
University of Florida, Gainesville, in December 1998 under the supervision of Dr.
Amost Neugroschel and Dr. ChihTang Sah. His doctoral research involved advanced
GeSi HBT modeling, MOS transistor modeling and reliability. His interests are in
submicron transistor technology development and modeling.
107
23
AEG_Qe(0) AEG_Ge(XB)
Fig. 1.15 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with
a trapezoidal Ge distribution in the base region under forward active operation,
vBE>ov, VCB>0V. Fn is the quasiFermi level for electrons and FP is the quasiFermi
level for holes. Vbi is the builtin potential of the p/n junction. AEG_Ge(0) is the Ge
induced energy gap reduction at x=0 in the quasineutral base region. AEG.Ge(XB) is the
Geinduced energy gap reduction at x=XB in the quasineutral base region. AEG_Ge =
AEG_Ge(XB)AEG_Ge(0) is the total Geinduced energy gap reduction through the base
region.
CHAPTER 3
LFGV ANALYSIS OF GeSi HBT WITH NO REFERENCE Si BJT
3.1 Introduction
In the previous chapters, the performance improvements of GeSi HBT over Si BJT
are discussed and demonstrated. The LFGV methodology for accurate determination of
parameters of GeSi HBT was also developed. This methodology is based on
simultaneous measurement of GeSi HBT and Si BJT with nearly identical dopant
impurity concentration profiles such that the difference in the performance is due to Ge
only. The disadvantage of this approach is that a reference Si BJT is not always
available. Furthermore, it is very difficult to fabricate GeSi and Si transistors with
nearly identical dopant impurity concentration profiles. In this chapter, we describe
another LFGV method for GeSi HBT analysis that does not use a reference Si BJT.
One of the key parameters in the LFGV methodology is the normalized Geinduced
energy gap reduction, r=AEG_Ge/kT, which is due to the compositional grading of Ge in
the base. This energy gap variation produces an aiding electric field, Ebi which
significantly decreases the electron transit time through the base layer that yields higher
speed and fT [4546], Thus, accurate and independent measurement of the r\ and the
resulting Ebi is essential for the design, optimization and analysis of GeSi HBT, and for
the verification of theoretical predictions about the expected performance benefits due
to the compositional grading. The determination of r\ is also a basis for a complete
70
57
2
3
4
5
6
7
8
1000/T /(1/1K)
Fig. 2.8 The measured temperature dependence of ln[exp(Ay).DB(GeSi)/DB(Si)] in the
range of 150K300K normalized to 296K for stripe transistor. Least squares fit to the
data is shown by the solid line and gives an activation energy of
Ea = AEg.hd = 80.2meV.
30
cm"3) and at low temperatures (~77K) [17]. But, the mobility enhancement has not yet
been experimentally confirmed in fabricated GeSi HBTs. The builtin electric field and
higher minority carrier mobility increases the collector current Ic and current gain 3F,
but this is compensated by the decrease in the density of states. Thus, a methodology to
separate these Geinduced effects is the key to a comprehensive quantitative analysis of
the Si/GexSij.x/Si HBT.
In this work, we present the LFGV methodology for a complete chargecontrol
analysis of HBTs, including a quantitative comparison with Si reference transistors, and
an experimental demonstration giving the first quantitative measurements of the builtin
electric field and minoritycarrier recombination lifetime in the GeSi base.
2.2 Base Thickness Modulation (BTM)
Lowfrequencyconductancevoltage (LFGV) methodology is based on the
combination of the dc Ic_VBE and ^b'^be measurements with the lowfrequency
(~lkHz) smallsignal conductance measurements resulting from the base thickness
modulation [2324], The base thickness modulation (BTM) effect is illustrated in Fig.
2.1 for a Si BJT. The collector current Ic is proportional to the gradient of the base
minority carrier distribution in the base layer at a certain forward bias emitterbase
voltage VBE and reverse bias collectorbase voltage VCB. When a smallsignal voltage
vcb is superimposed upon the dc basecollector junction voltage VCB, the quasineutral
base thickness XB will be modulated by AXB, as shown in Fig. 2.1. The shadowed area
in Fig. 2.1(b) shows the change in the minority carrier distribution in the quasineutral
base due to the base thickness modulation, and the base minority carrier charge QB
59
Fig. 2.9 Sensitivity of r), to XR/XB calculated from (2.18a) with r\2 as a variable
parameter.
ACKNOWLEDGEMENTS
I would like to thank my supervising professors Dr. Arnost Neugroschel and Dr.
ChihTang Sah for their guidance during my doctoral research. I would also like to
thank Dr. Peter Zory, Dr. Vladimir Rakov and Dr. Mark Orazem for serving on my
supervisory committee. I also thank Jin Cai, WanPeng Cao, Derek Martin, and Yi
Wang for helpful discussions.
11
22
Ec
Ev
AEG_Ge(0) AEG_Ge(XB)
PGeSi
Base
0 XR XB
E
F
Fig. 1.14 Energy band diagram of GeSi HBT (dashed line) and Si BJT (solid line) with
a trapezoidal Ge distribution in the base region at equilibrium (zero bias). EF is the
Fermi level at equilibrium. AEG.Ge(0) is the Geinduced energy gap reduction at x=0 in
the quasineutral base region. AEG_Ge(XB) is the Geinduced energy gap reduction at
x=XB in the quasineutral base region. AEG_Ge = AEG_Ge(XB) AEG_Ge(0) is the total Ge
induced energy gap reduction through the base region.
73
9oR !cR ni2^XB) CBE
= . (3.6)
90F JCF ni2(0) CBC
From (3.6) we obtain the ratio of the intrinsic carrier concentrations at the two edges
of the quasineutral base layer which is given below in (3.7)
ni2(XB^ % 90R CBC
= (3.7a)
n?(0) JCR 90f cbe
Nc(Xb)Nv(Xb)
= exp(AEG_ge/kT). (3.7b)
Nc(0)Nv(0)
The result in equation (3.7) is completely general and applies to any HBT or BJT
with any compositional grading and dopant impurity concentration profile, including
mV HBTs.
Equation (3.7) uses no assumptions about the material composition or impurity
doping in the base and gives the ratio of the intrinsic carrier concentrations at the two
edges of the quasineutral base layer directly from the measured collector current, output
conductance, and junction capacitances for any HBT or BJT. The approximation using
the average base diffusivity in (3.2) does not affect the ratio in (3.7), since the integral
relation for DB is the same for the forwardactive and reverseactive operation modes.
One possible application of the result in (3.7) is measurement of the energy gap in
GeSi as a function of Ge concentration using test structures with different Ge linearly
graded profiles as shown in Fig. 3.1. Equation (3.7) can be also used to measure the
intrinsic carrier concentration ratio due to heavydopinginduced energy gap reduction
using structures with a highlow dopant impurity concentration profiles as shown in Fig.
3.2.
