UNIVERSITY OF CINCINNATI September 23 03 _____________ , 20 _____ Pradeep A. Balaraman I,______________________________________________, hereby submit this as part of the requirements for the degree of: Master of Science ________________________________________________ in: Electrical Engineering ________________________________________________ It is entitled: Design, Simulation and Modeling of InP/GaAsSb/InP ________________________________________________ Double Heterojunction Bipolar Transistors ________________________________________________ ________________________________________________ ________________________________________________ Approved by: Dr. Kenneth Roenker ________________________ Dr. Marc Cahay ________________________ Dr. Punit Boolchand ________________________ ________________________ ________________________ DESIGN, SIMULATION AND MODELING OF InP/GaAsSb/InP DOUBLE HETEROJUNCTION BIPOLAR TRANSISTORS A Thesis Submitted to the Division of Research and Advanced Studies of the University of Cincinnati in partial fulfillment of the requirement for the degree of MASTER OF SCIENCE in the Department of Electrical and Computer Engineering and Computer Science of the College of Engineering September 2003 by Pradeep A. Balaraman B.E. in Electronics and Communications Engineering, Tamilnadu College of Engineering, Bharathiyar University, Coimbatore, India, April 2000. Thesis Advisor and Committee Chair : Dr. Kenneth P. Roenker Committee Members: Dr. Marc Cahay, Dr. Punit Boolchand Abstract Device modeling using a two dimensional, drift-diffusion approach utilizing a commercial numerical device simulator has been used to investigate the operation and performance of InP/GaAsSb heterojunction bipolar transistors (HBTs). GaAsSb lattice matched to InP has an energy bandgap (0.72 eV) that is similar to that of InGaAs (0.75 eV) so that Sb-based HBTs have been proposed as a replacement for InGaAs-based HBTs. In particular, the conduction band lineup is more favorable at the base-collector, which makes the GaAsSb-based HBTs especially attractive for double heterojunction bipolar transistors (DHBTs) where higher breakdown voltages are desired. In this work, the results of device modeling will be compared initially with recent experimental reports to validate the modeling approach. Then the design and operation of the devices will be examined to investigate the factors controlling device performance in order to facilitate improvements in device design. The degradation of device performance at high currents due to the formation of a parasitic barrier in the collector region and the base push out effects is examined. Finally, a device structure with improved high frequency performance is presented. Acknowledgements I would like to sincerely thank my thesis advisor Dr. Roenker for his constant support and guidance. I thank him in particular for his patience and brilliant ideas to help move the research along when it got stuck many times. I thank Dr. Cahay and Dr. Boolchand for their participation in my thesis defense. I would like to thank my lab mates, Nitish and Donka, for helping me learn the simulation software and Mohan, Aniket, Shivani, Dave, and Rama, for the their help and support. Many thanks to my friends, Sathya, Prashanth, Mukesh, Priya, Chandy, Sheeda, Vignesh, and the “Gang” at Tech office for making my stay at UC a pleasant and memorable one. Special thanks to friends back home, Bhama, Gopal, Saravana, Senthil, Subha, and Vinesh for their awesome and endearing company. Finally, I would like to thank my family for their boundless love and support. ii Contents Title Page Abstract................................................................................................................................i Acknowledgements.............................................................................................................ii Contents..............................................................................................................................iii List of Figures....................................................................................................................vii List of Tables.....................................................................................................................xv Chapter I Introduction 1.1 Introduction..................................................................................................1 1.2 Bipolar Junction Transistor..........................................................................2 1.3 Heterojunction Bipolar Transistors..............................................................6 . 1.4 Double Heterojunction Bipolar Transistors.................................................9 1.5 InP-Based Heterojunction Devices..........................................................10 1.6 InP/GaAsSb/InP Double Heterojunction Bipolar Transistors...................15 1.7 Advantages of InP/GaAsSb/InP DHBTs...................................................18 1.8 Motivation and Organization of the Thesis...............................................22 Reference:..........................................................................................................................23 Chapter II Simulation Software, Techniques and Material Properties 2.1 Introduction………………………………………………………….…28 2.2 Simulation Technique...…………………...……………….…………. 29 2.3 Device Physics and Models………………..…………….…………… 30 iii 2.3.1 Poisson’s Equation……..………………………………………... 30 2.3.2 Carrier Continuity Equations……………………………………..31 2.3.3 Transport Equations………..…………………………………... 32 2.4 Material Parameter Models………………………….………………... .33 2.4.1 Mobility Modeling……………………………………………...33 2.4.1.1 The Caughey-Thomas Model……….………………………... 34 2.4.1.2 Parallel Electric Field Dependent Mobility…….…………….35 2.4.2 Carrier Generation-Recombination Models…………………....35 2.5 Material Properties of GaAsSb………………………..………………..37 2.5.1 Crystal Structure and Lattice Matching…….………………….... 38 2.5.2 GaAsSb Energy Band Gap……………...…………..………….....40 2.5.3 Bandgap Offsets for InP/GaAsSb Heterojunctions……………....45 2.5.4 Effective Masses and Density of States………………….…….....49 2.5.5 Mobility………………………………………………….…….....53 2.5.5.1 Hole Mobility………………………………………………...... 53 2.5.6 Electron Mobility……………………………………………....... 57 2.5.7 Carrier Lifetimes……………………………………………........ 58 2.6 Material Properties of InP and InGaAs…………………...…………....60 2.6.1 Indium Phosphide…………………………………………….......60 2.6.1.1 Heavy Doping Effects in InP…………………….…………......61 2.6.2 InGaAs…………………………………………………………....63 2.7 Conclusion……………………………………………………………...66 Reference.....................................................................................................……..…....…67 iv Chapter III Simulation of InP/GaAsSb/InP HBTs and basic results 3.1 Introduction………………………………………..……………….......70 3.2 Software Description…………………………....…..……………........70 3.2.1 Definition of HBT structure and Mesh Specification.……….......72 3.2.2 Physics-Based Models…….…………………………………......73 3.2.2.1 Hot-Electron Model………………………………………........74 3.2.3 Input Bias Setup and Output Result Extraction…..………….......76 3.3 Structure and Modeling of InP/GaAsSb/InP HBTs…………….….......77 3.4 High Current Effects………………………………………….…..........95 3.5 Current Gain Collapse at High Frequency………………………..........98 3.6 Conclusion………………………………………………………........105 Reference:………………………………………………………………………………106 Chapter IV Optimization of InP/GaAsSb/InP HBTs 4.1 HBT Performance Parameters..............................................................110 4.1.1 AC Current Gain..........................................................................110 4.1.2 Current Gain Cutoff Frequency (fT ).............................................111 4.1.3 Unilateral Power Gain and Maximum Frequency of Operation..114 4.2 Optimization of InP/GaAsSb/InP HBTs...............................................115 4.2.1 Optimization of the Base Layer...................................................116 4.2.2 Optimization of the Emitter Layer...............................................123 4.2.3 Collector Layer Optimization......................................................128 4.3 Optimized Structure..............................................................................134 4.4 Conclusions...........................................................................................143 v Reference…………….....................................................................................................143 Chapter V Conclusions and Future Work 5.1 Conclusion............................................................................................144 5.2 Future Work..........................................................................................146 Reference:………………………………………………………………………………147 vi List of Figures Figure 1.1 Band diagrams of an npn homojunction bipolar transistor under (a) thermal equilibrium (b) and in the normal, forward-active mode .................................4 Figure 1.2 Band diagrams of an Npn HBT under (a) thermal equilibrium and (b) in the normal forward-active mode .............................................................................7 Figure 1.3 Comparison of material properties of InP, GaAs and Silicon .........................12 Figure 1.4 Conduction band profiles for a few design options for InP-based HBTs. [6]..14 Figure 1.5 Band lineup of InP/GaAsSb heterojunction ....................................................16 Figure 1.6 Comparison of reported fT and fmax values for InP HEMTs, GaInAs DHBTs and InP DHBTs................................................................................................17 Figure 1.7 fT values vs. BVceo for InP collector DHBTs from literature (€).....................20 Figure 1.8 Normalized cut-off frequency dependency on collector current density for various DHBT and SHBT technologies ..........................................................21 Figure 2.1 GaAsSb crystal structure where Sb is incorporated in some of the As sites....39 Figure 2.2 Lattice parameter versus compositio n plot for In1-x Gax Asy Sb1-y and compositions (dotted line) lattice matched to InP...........................................40 Figure 2.3 Energy Band structure of the binaries GaAs (a) and GaSb (b)........................41 Figure 2.4 Plot of energy band gap versus composition for the quaternary In1-x Gax Asy Sb1-y for the material lattice matched to InP (dotted line).......................................42 Figure 2.5 Variation of GaAs1-x Sbx bandgap with Sb content • Klem et al et al. [6] and ♦Merkel et al...................................................................................................43 Figure 2.6 Energy separation between L- valley of conduction band and top of the valence band....................................................................................................44 vii Figure 2.7 PL spectra for GaAsSb for various Sb concentrations...................…..........................45 Figure 2.8 Staggered Band line-up at InP/GaAsSb Heterojunctions.................................46 Figure 2.9 Band offsets of GaAsSb on InP........................................................................48 Figure 2.10 Calculated Band offsets in GaAsSb on InP based on results of Merkel et al [6] (•) Ec b and (€) E vb.....................................................................................49 Figure 2.11 Empirical variation of hole effective mass with Sb content [9] (♦) Heavy hole mass (€) Light hole mass.........................................................................50 Figure 2.12 Variation of the electron effective mass with Sb Content .............................51 Figure 2.13 Temperature dependence of intrinsic carrier concentration: 1) x = 0(GaAs), 2) x = 0.3, 3) x = 0.8, 4) x = 1 (GaSb)...........................................................53 Figure 2.14 Hole Hall Mobility as a function of Sb in GaAsSb........................................54 Figure 2.15 Hall Mobility for holes as a function of carrier concentration for various Sb compositions. Where different symbols correspond to the indicated Sb composition ranges..........................................................................................55 Figure 2.16 Hole mobility variation with doping: experimental values (+) [14] and calculated Caughe y-Thomson values (•)......................................................56 Figure 2.17 Electron mobility variation with doping: experimental values (+) [13, 16] and calculated Caughey-Thomson values (•)......................................................58 Figure 2.18 Calculated carrier lifetime variation with doping (€) Hole lifetime, (•) Electron life time..........................................................................................60 Figure 2.19 Photoluminescence spectra of InP..................................................................61 Figure 2.20 Bandgap narrowing in InP due to heavy doping............................................63 viii Figure 2.21 Electron ( ) and hole (O) minority carrier lifetime models as a function of doping level......................................................................................................64 Figure 3.1 An InP/GaAsSb/InP DHBT structure shown with mesh grid..........................73 Figure 3.2 The Staggered Bandgap line-up of InP/GaAsSb/InP HBTs.............................75 Figure 3.3 Schematic of the simulated InP/GaAsSb/InP DHBT structure........................79 Figure 3.4 Simulated structure of InP/GaAsSb/InP HBT..................................................80 Figure 3.5 Simulated conduction and valence band profiles for InP/GaAsSb/InP DHBT with 200A0 base width and no applied biases..................................................83 Figure 3.6(a) Gummel-Poon plot of the InP/GaAsSb DHBT structure from Dvorak et al for 400 A0 base width (C: 4 x 10-19 cm-3 ) and 1.5x24 µm2 ..............................85 Figure 3.6(b) Simulated Gummel-Poon characteristics of InP/GaAsSb/InP structure for the device structure described in Figures 3.4 and 3.3 for Vcb=2.0 V.........86 Figure 3.7(a) Typical common-emitter characteristics for a small area 0.4x11µm2 emitter device with a 250 Å base and a 2000 Å InP collector.....................................87 Figure 3.7(b) Simulated common-emitter characteristics for a 0.5x1 µm emitter device (Fig 3.3) with a 200 A0 base and 2000 A0 InP collector...........................87 Figure 3.8(a) Frequency response of fabricated structure reported by Bolognesi et al. with a 200 A0 thick base with a doping of 8x1019 /cm3 ..........................................89 Figure 3.8(b) Frequency response of the simulated InP/GaAsSb/InP HBTs (•) current gain, (X) unilateral power gain.......................................................................89 Figure 3.9(a) Evolution of fT with JC for the InP/GaAsSb DHBT with a base width of 200A0 and collector width of 2000 A0 for a VCE=0.4V to 1.8V as reported by Bolognesi et al.................................................................................................91 ix Figure 3.9(b) Evolution of fT with J C for collector-emitter bias VCE = 0.4 to 1.8V for simulated structure...........................................................................................91 Figure 3.9(c) Peak fT variation with VCE for simulated (•) and fabricated (s) InP/GaAsSb/InP DHBTs.................................................................................92 Figure 3.10(a) fmax and fT dependence on JC for a 250 A0 base and 2000 A0 InP/GaAsSb DHBT for various VCE biases..........................................................................94 Figure 3.10(b) Evolution of fmax vs collector current density JC for 200A0 bases and 2000A0 collector for the simulated InP/GaAsSb DHBT for various VCE........94 Figure 3.10(c) Reported fmax variation with collector-emitter bias VCE (•) for a 4x11 µm2 emitter, 250 A0 base and 2000 A0 Collector [9] and simulated peak fmax values (? ) ....................................................................................................95 Figure 3.11 High frequency and Cbe behavior of a 200 A0 base and 2000 A0 InP/GaAsSb/InP device...................................................................................98 Figure 3.12 Variation of current gain with collector current density for VCE=1.8 V at 2 GHz...............................................................................................................99 Figure 3.13 Variation of maximum unilateral power gain with collector current density for VCE=1.8 V at 2 GHz................................................................................99 Figure 3.14 Collapse of electric field and subsequent parasitic barrier formation in the collector of InP/GaAsSb/InP DHBTs with Vbe varied from 0.6 V to 1.0 V and VCE = 1.8V.....................................................................................................101 Figure 3.15 Critical current density for the onset of high current induced device degradation versus based-collector voltage for the simulated and calculated device behavior..............................................................................................102 x Figure 3.16(a) Hole concentration for low and high current operation of a 200 A0 base and 2000 A0 collector simulated InP/GaAsSb/InP DHBT, ‘B’ and ‘C’ indicate the base-collector regions for VCE = 1.8V and for VBE from 0.6V to 1.1V...104 Figure 3.17 Conduction band barrier height versus collector current density for simulated InP/GaAsSb structure given in Figure 3.3 for VCE =1.8 V............................104 Figure 4.1 Parameters affecting the high frequency performance of the HBT................112 Figure 4.2(a) fT Vs base-emitter voltage as a functio n of the base width for VCE=1.8V for a fixed base doping of 4x1019 cm-3 ........................................................114 Figure 4.2(b) ft vs Collector Current density as a function of base width for VCE=1.8V and VBE from 0.6V to 1.07V for a fixed base doping of 4x1019 cm-3 ......114 Figure 4.2(c) fmax vs base-emitter voltage for various base widths for VCE=1.8V and VBE from 0.6V to 1.07 Volts for a fixed base doping of 8x1019 cm-3 ...............119 Figure 4.2(d) fmax vs collector current density for various base widths for VCE=1.8V and VBE from 0.6V to 1.07 Volts for a fixed base doping of 8x1019 cm-3 .........119 Figure 4.2(e) Variation of peak fT and fmax with base width for a base doping of 8x1019 cm-3 , (+) Peak fT fmax from results by Bolognesi et al. [2] for similar device structure.......................................................................................................120 Figure 4.3(a) fT versus emitter-base voltage for various doping levels for VCE=1.8V for a fixed base width of 20nm............................................................................121 Figure 4.3(b) fmax versus emitter-base voltage for various doping levels for VCE=1.8V for a fixed base width of 20nm............................................................................122 xi Figure 4.3(c) Variation of peak fT and fmax with base doping for a fixed base width of 20nm, (+) indicates peak fT and fmax results from Bolognesi et al. [2] for similar device structure..................................................................................122 Figure 4.4(a) Effect of emitter width variation on fT for VCE=1.8V for a fixed emitter doping of 3x1017 cm-3 ....................................................................................124 Figure 4.4(b) Effect of emitter width variation on fmax for VCE=1.8V for a fixed emitter doping of 3x1017 cm-3 ....................................................................................124 Figure 4.4(c) Variation of peak fT and fmax with emitter width for a fixed emitter doping of 3x1017 /cm3 , (+) indicates peak fT and fmax results by Bolognesi et al. [2] for similar device structure..................................................................................125 Figure 4.5(a) fT versus base-emitter voltage for various doping levels for VCE=1.8V for a fixed emitter thickness of 700 A0 ...................................................................126 Figure 4.5(b) fmax versus base-emitter voltage for various doping levels for VCE=1.8V for a fixed emitter length of 700 A0 ..................................................................127 Figure 4.5(c) Variation of peak fT and fmax with emitter doping for a fixed emitter width of 700 A0 ........................................................................................................127 Figure 4.6 Calculated collector depletion thickness, fT and fmax, as a function of VCE...129 Figure 4.7(a) fT vs base-emitter voltage for several collector widths for VCE=1.8V for a fixed collector doping of 3x1016 /cm3 .........................................................130 Figure 4.7(b) fmax vs base-emitter voltage for several collector widths for VCE=1.8V for a fixed collector doping of 3x1016 /cm3 .........................................................130 Figure 4.7(c) Variation of peak fT and fmax with collector width for a fixed collector doping of 3x1016 /cm3 .................................................................................131 xii Figure 4.8(a) fT Vs Base emitter Voltage for various collector doping for a fixed collector width of 2000A0 ..........................................................................................133 Figure 4.8(b) fmax Vs Base emitter voltage for various collector doping for a fixed collector width of 2000A0 ...........................................................................133 Figure 4.8(c) Variation of peak fT and fmax with collector doping for a fixed collector width of 2000A0 ............................................................................................134 Figure 4.9(a) Schematic of the Optimized InP/GaAsSb/InP HBT..................................135 Figure 4.9(b) Schematic of the InP/GaAsSb/InP HBT structure by Bolognesi et al.......135 Figure 4.10 Simulated Structure of Optimized InP/GaAsSb/InP HBT...........................136 Figure 4.11 Gummel-Poon simulations of optimized InP/GaAsSb/InP Structure..........136 Figure 4.12 Frequency response of the optimized InP/GaAsSb/InP HBT for VCE = 1.8V.............................................................................................................137 Figure 4.13 Frequency response of the optimized InP/GaAsSb/InP HBT for VCE = 1.8V.............................................................................................................138 Figure 4.14(a) Evolution of fT with JC for various VCE from 0.8 to 1.8 Volts.................140 Figure 4.14(b) Evolution of fmax with JC for various VCE from 0.8 to 1.8 Volts..............140 Figure 4.15 Maximum power and AC current Gain versus collector current density (l) Current gain (n) Power gain for VCE = 1.8V...........................……............141 Figure 4.16(a) Peak fT variation with VCE for InP/GaAsSb/InP DHBT structure of Bolognesi et al. [2] (g ), simulated fT for the same structure ( n) and peak fT for the optimized device structure (o) given in Figure 4.9.........................142 xiii Figure 4.16(b) Peak fmax variation with VCE for InP/GaAsSb/InP DHBT structure of Bolognesi et al. [2] (g ), simulated fT for the same structure ( n) and peak fT for the optimized device structure (o) given in Figure 4.9.........................142 xiv List of Tables Table 1.1 Evolution of InP/GaAsSb/InP DHBT device development...............................17 Table 2.1 Bandgap of GaAs0.51Sb0.49 44 Table 2.2 Effective Masses of Electrons and Holes for GaAs, GaSb and GaAsSb...........51 Table 2.3 Caughey-Thomas Parameters for Holes and Electrons.....................................58 Table 2.4 Electron and Hole Minority Carrier Lifetimes for GaAs, GaSb and Latticematched GaAsSb.................................................................................................59 Table 2.5 Minority Carrier Lifetime Model Parameters for InGaAs.................................64 Table 2.6 Caughey-Thomas Parameters for InP, InGaAs, GaAsSb..................................65 Table 2.7 Summary of Important Material Parameters for GaAs0.51 Sb0.49 InP In0.53Ga0.47 As.....................................................................................................66 Table 3.1 Evolution of InP/GaAsSb/InP DHBT Device Development.............................78 Table 3.2 Summary of Important Material Parameters for GaAs0.51 Sb0.49 InP In0.53Ga0.47 As..................................................................................................…....81 Table 4.1 Simulation results for Optimized InP/GaAsSb/InP structure..........................138 xv Chapter I 1.1 Introduction Silicon has been the dominant material of the electronics industry for the past five decades and will continue to remain so in the foreseeable future. However, silicon-based devices come under severe limitations like current gain degradation and lack of linearity when they are extended to microwave frequencies. The silicon-based Bipolar Junction Transistor commonly referred to as the BJT, can only be used in applications in the frequency range up to a few GHz. At these freque ncies and above is where the excellent high frequency performance characteristics of III-V based devices can be harnessed, particularly in satellite, mobile and radar communications. With more of these applications being adopted for consumer uses, the market for III-V based devices is expected to continue to grow. III-V-based npn bipolar transistors have achieved tremendous performance in terms of the microwave frequencies of operation and higher current gains because of their superior material properties, such as higher electron mobility, and unique bandgap alignment when two different materials are used to form a heterojunction. Npn Heterojunction Bipolar Transistors (HBTs) differ from the BJTs in the fact that the emitter is replaced by a material with a larger bandgap than the base. This leads to a lowering of the barrier for electron injection into the base, but an increased barrier for the holes which prevents their back injection into the emitter, resulting in higher current gains. As a result, the base can be made thinner and at the same doped higher than the emitter to reduce the base resistance and further increase its high frequency performance. The idea of the HBT is not new, but was actually conceived a long time ago by Shockley 1 and co-workers in 1948 [1, 2]. Historically, difficulties in growing high quality, relatively defect free III-V materials have slowed the exploitation of this concept. However, with recent advances in epitaxial growth technologies, such as Molecular Beam Epitaxy (MBE) and Metal Organic Chemical Vapor Deposition (MOCVD), the full potential of HBTs is now being realized. While III-V based devices cannot hope to replace silicon in the sub-GHz range or in terms of cost effectiveness, they have carved out niche markets above 10 GHz, where higher speed and performance factors outweigh cost. Not only are III-V devices a competition for silicon in these areas, but they are also useful in applications where it would be impossible for silicon-based devices to compete, such as for very high frequency (>100 GHz) applications. Recently, InP/GaAsSb/InP HBTs have demonstrated state of the art performances, in fact one of the highest for any solid-state device with cutoff and maximum oscillation frequencies of 300 GHz [3]. Numerical Computer Assisted Design (CAD) simulation and modeling of such InP/GaAsSb/InP HBTs is the focus of this thesis. In this chapter, we provide a brief introduction to HBTs and their advantages over BJTs. We introduce the InP/GaAsSb material system, its advantages and their applications in HBTs. 1.2 Bipolar Junction Transistors It is important that we review the basic operation of the homojunction BJT before we study the InP/GaAsSb/InP HBTs since understanding the operation of the BJT is essential for discussing the performance of HBTs. The BJT consists of three main layers, namely the emitter, the base and the collector all made of the same silicon material. The emitter 2 and the collector are doped of the same type, whereas the base layer sandwiched in between is of the opposite type. The basic principle behind a BJT is that the current flowing from the emitter to the collector can be controlled by small changes in the base current [4]. For our current discussion here, we will assume an npn transistor, with the emitter and the collector being of n-type and the base being p-type. One of the dominant features of p-n junctions, i.e., the injection of minority carriers across the junction under forward biasing, forms the basis for operation of the BJTs. In the normal operating condition for npn transistors, the base-emitter junction is forward biased at VBE (VBE>0) and the base-collector junction is reverse biased at VBC (VBC<0). Under a forward bias, the minority carrier concentration in the base at the base-emitter junction is amplified by the law of junction by a factor given by exp (qVBE/kT), where VBE is the applied voltage to the base-emitter junction. For a transistor to operate properly, the width of the base region WB should be should less than the diffusion length Ln of the electrons [4, 5]. This is because the electrons injected into the base from the emitter must diffuse across base to the collector without excessive recombination with the majority carrier holes in the base. These minority carrier electrons when they reach the base-collector junction are quickly swept across the collector by the high electric field in the reverse-biased collector-base junction and constitute the collector current (see Figure 1.1). 3 Figure 1.1 Band dia grams of an npn homojunction bipolar transistor under (a) thermal equilibrium (b) and in the normal, forward-active mode [4]. The band diagram of a BJT in the rmal equilibrium and normal operation are shown in Figure 1.1. The dotted line in the figure shows the Fermi level in each region for the transistor. We can see that in a normal BJT the emitter region is doped heavily and the collector region is lightly doped. Under normal operation, two currents dominate the transistor operation, the collector current that flows from the emitter to the collector and the base current. This base current is normally due to the back injection of holes into the emitter from the base. Solving the carrier continuity equations, the base current and the collector current are given as follows [4, 5] qAE DpE niE2 qV exp( BE ) X E NE kT (1.1) qAE DnB niB2 qV IC = exp( BE ) X B NB kT (1.2) IB = 4 where AE is the emitter junction area, NE is the emitter doping level, XE is the emitter thickness, DpE is the minority hole diffusion coefficient in the emitter, niE is the intrinsic carrier concentration in the emitter, NB is the base doping level, XB is the base thickness, DnB is the minority electron diffusion coefficient in the base and niB is the intrinsic carrier concentration in the base. A simple analysis of the above two expressions will reveal that in order for the collector current to be greater than the base current, it is necessary that the emitter doping level exceed the base doping level. In doing so, we have traded off an important transistor design parameter. For keeping the base doping at a lower value means that the base resistance will be high. This is can be compensated by widening the base, however, this will in turn contribute significantly to base transit time of the electrons, thus lowering the device’s maximum frequency of operation (fmax). Therefore, for a fixed base doping level, the current gain of the BJT can be increased, initially, by increasing the emitter doping. But, significant emitter bandgap narrowing occurs when emitter doping is around NE~1018 /cm3 , causing the current gain to degrade. The very nature of the homojunction of the BJT therefore gives us a mutually conflicting set of requirements, where the base doping cannot be increased without sacrificing current gain. This limits the operation of the BJTs to low frequency amp lifiers, filters and oscillators. The performance of the HBT can be significantly better than that of the BJT, if we dope the base high and at the same time comparatively reduce the emitter doping, which is more feasible because of the wide band gap emitter in the HBT. This is where HBTs have made tremendous progress in terms of frequency of operation in the range of 100’s of Giga Hertz. 5 1.3 Heterojunction Bipolar Transistors The idea for HBTs is not new, but actually as old as the bipolar transistor itself, originally put forward by Shockley and co-workers along with his claim for the transistor in 1948 [1, 2]. However, as Kroemer suggested in his now famous paper [5], the idea of HBTs was ahead of its time waiting for the development of modern epitaxial growth techniques for defect free growth of III-V compound semiconductor heterostructures. Heterojunction Bipolar Transistors (HBTs), when appropriately biased, provide good current amplification and are highly desirable for high-speed operation. In comparison with conventional silicon Bipolar Junction Transistors (BJTs), HBTs exhibit higher unity gain cutoff frequencies fT , lower base resistance, lower base-emitter capacitance and higher Early Voltages [4]. One of the reasons HBTs ha ve these advantages over a simple BJT is that the emitter injection efficiency no longer depends only on the relative doping levels of the emitter and the base region. The presence of a base/emitter bandgap difference adds a term that depends exponentially on the energy gap difference, which dominates the emitter injection efficiency. For abrupt heterostructures, the current gain has an exponential dependence on the valence band ∆Ev discontinuity for n-p-n devices [4]. As a result, the doping of the base region can actually be increased above the emitter layer doping and the base thickness reduced which results in the dramatic improvement in the microwave performance (fT , fmax) of an HBT compared to a BJT. Spectacular advances in material growth technologies have enabled precise control of layer thickness, doping and the ability to change composition from one III-V semiconductor to a different III-V semiconductor material system. Shown in Figure1.2 is an abrupt emitter-base heterojunction in an Npn HBT [4]. Heterostructures utilize the 6 energy gap variations between two different semiconductors, such as the emitter and the base in the HBT, in addition to the electric fields and forces acting on the electrons and holes to control and distribute the flow of carriers. For the Npn HBT, electrons are injected thermally into the base from the emitter, diffuse across the base and electric field in the base-collector depletion regions sweep them into the collector at high speed. The bandgap lineup of heterojunctions is such that the valence band discontinuity prevents back injection of holes from the base into the emitter. This results in significant increase in current gain without sacrificing the high frequency performance. Figure 1.2 Band diagrams of an Npn HBT under (a) thermal equilibrium and (b) in the normal forward-active mode [4]. Heterojunctions can be of two types, graded and abrupt. For abrupt heterojunctions, two different semiconductors with different bandgaps are brought together and the 7 interface is abrupt, i.e. the composition changes abruptly. In graded heterojunctions, the composition is gradually changed across the heterojunction and there is a suppression of energy band spikes due the energy band edge discontinuity seen at the abrupt heterojunction interfaces. However, we will discuss only abrupt junctions, as it directly relates to this work. The alignment of the conduction and valence bands in Npn HBTs is such that by limiting hole back injection into the emitter, the emitter injection efficiency is close to the ideal value of unity, regardless of the doping levels of the emitter and the base [5]. In most cases, the conduction band of the emitter lies above that of the base and the current gain of the device is given by equation 1.3. [4] The valence band discontinuity at the heterojunction is also important, as it controls the amount of back injection into the emitter. The ratio of the electron to hole currents across the emitter-base heterojunction is given approximately by [4] In N v = βmax = E nb exp( ∆Eg / kT ) Ip NB v pe (1.3) where ∆Eg is the difference in bandgap between the emitter and base, where NE is the emitter doping, NB is the base doping and vnb and vpe are the thermal velocities of electrons and holes in the base and emitter, respectively. This result assumes negligible recombination in the base and ∆Ev = ∆Eg, so it gives the maximum possible current gain. When ∆Ev ≠ ∆Eg then the emitter base energy bandgap difference increases the In /Ip ratio by a factor of exp( ∆Ev /kT) relative to that for BJTs and the valence band ∆Ev discontinuity at the heterojunction replaces ∆Eg in (1.3) and is directly related to the current gain as follows [4] 8 In : exp( ∆Ev / kT ) Ip (1.4) So if the valence band discontinuity is sufficiently large, as it is for the InP/GaAsSb HBT, a significant current gain can be expected even if NE=NB. For an abrupt heterojunction bipolar transistor, an advantage of the potential barrier (spike) in the conduction band at the emitter-base interface is that the barrier facilitates the injection of the electrons into the base region with a high velocity (~108 cm/s) that exceeds the saturation velocity. The result is a highly efficient and very fast, near-ballistic electron transport through the base. Thus, an HBT can be designed for high frequency operation using a highly doped base without compromising injection efficiency or current gain. 1.4 Double Heterojunction Transistors While Shockley’s initial description of the HBTs stopped with having a wide bandgap emitter, now we know that having a wide bandgap material for the collector can have significant advantages as well, a few of which are listed below [5]: 1) The hole injection from the base to the collector is suppressed because of the wide bandgap collector. This is particularly important if the transistors are being used for digital switching operations. Use of a wide gap collector facilitates the use of higher base doping and at the same time keeping the collector doping low. In addition, there is a complete suppression of parasitic charge storage, combined with greatly reduced RCtime constant due to lower base resistance as a result of higher base doping. 2) Using the same material for the emitter and the collector produces a double heterojunction transistor (DHBT), one at the emitter and the other at the collector junction. This translates into a huge advantage when HBTs are incorporated into ICs 9 since the collector and emitter can be interchanged. Implementation of Emitter Coupled Logic (ECL) becomes easie r by the use of DHBTs. 3) Specific material systems like the InP-based materials offer superior transport and thermal properties than GaAs based materials. In fact, InP-based HBTs have demonstrated some of the highest microwave performances and InP-based DHBTs have additional benefits like enhanced breakdown voltages. 4) One of the most important advantages of the DHBTs, including InP/GaAsSb/InP DHBT that we are going to study, is that they have a low turn-on voltage of ~0.4V translating into higher power gain and power added efficiency, which is extremely important for portable devices like mobile phones running on limited power sources like batteries. 5) However, a significant drawback for npn DHBTs, is the fact that the ∆Ec at the basecollector junction can form a barrier to electron injection from the base into the collector for material systems such as InP/InGaAs. This degrades the current gain and device speed. One of the advantages of the InP/GaAsSb/InP system is that this barrier is nonexistent [6]. 1.5 InP-Based Heterojunction Devices The emergence of the 10 Gb/s optical communication systems and the drive for hardware operating at 40 Gb/s has pushed the existing III-V technologies towards their limits [6]. However, continuing growth in the field of broadband communications, fibre optic transmissions and microwave direct digital frequency synthesis has demanded continuing advances in compact, highly efficient and cost effective solid-state solutions. 10 To date, InP-based HBTs have demonstrated some of the highest fT and fmax frequencies for any solid-state devices [7]. Even though the initial thrust in heterojunction technology was based on GaAs devices, recent advances in MBE and MOCVD have enabled InP devices to dominate at millimeter wave frequencies because of the following reasons: 1) Substrate thermal conductivity: InP has a higher substrate thermal conductivity than GaAs (0.68 versus 0.46 W-cm/K) resulting in lower device heating [8]. 2) Optical Integration: InP devices are compatible with long wavelength (1.3 and 1.55 µm) laser and LED sources. This allows for direct integration and tight coupling of light sources and transmitter/receiver circuitry [8]. 3) The lower surface recombination velocity of the InP-based HBTs (three orders of magnitud e lower than that of GaAs) results in a reduction of surface leakage currents and improved gain at low current densities, enhancing the ability to scale down devices to smaller dimensions for LSI implementation [7] 4) InP based devices have lower turn-on voltages, significantly lower 1/f noise and higher power added efficiency (PAE) than GaAs devices. [7] 5) At a clock rate of 50 GHz, for reasonable levels of power dissipation on-chip, InP HBTs provide significantly more integration capability because of their ol wer power dissipation per gate than GaAs and SiGe HBT technology [9] . The following Figure 1.3 gives us a quick overview of the advantages of InP over conventional semiconductor materials like GaAs and Si [9]. Figure 1.3(a) shows InPbased HBT, have lower turn-on voltages than Si BJTs and GaAs HBTs due to the smaller 11 bandgap of InGaAs. Figure 1.3(b) shows that InP HBTs have a higher breakdown voltage at a given collector doping, while Figure 1.3(c) shows that InGaAs and InP have higher electron mobilities than GaAs. Finally, Figure 1.3(d) shows that InP has a larger thermal conductivity than GaAs, though less than that of Silicon. This entails huge advantages for InP-based devices over GaAs-based ones in high frequency (>10 GHz) RF power devices where most of the III-V devices find application, in terms of better thermal management and long term device reliability. Figure 1.3 Comparison of material properties of InP, GaAs and Silicon [9] 12 Almost all of the InP-based heterojunction devices are implemented using InGaAs as the narrow gap material for the base semiconductor. InP/InGaAs-based single HBTs have demonstrated outstanding properties with cut-off frequencies more than 200 GHz and with extrapolated cutoff frequencies exceeding 1 THz [10]. This impressive performance of the InP-based HBTs is mainly because of aggressive scaling and reduction in the collector junction area by backside processing [10] and because of excellent material properties of InGaAs in terms of mobility and electron saturation velocity. This poses a very good question as to why switch to GaAsSb, when InGaAs is doing the job admirably. However, we will shortly see in the following discussion where InGaAs fails and GaAsSb steps in to save the day. The main drawback with the InP/InGaAs is that it suffers from a low breakdown voltage because of the narrow InGaAs collector energy gap of 0.75 eV [11]. Kurishima et al and Chor et al, [12, 13] proposed enhancing the breakdown voltages by using an wide bandgap AlInAs or InP collector layers in the double heterojunction transistor. But this advantage comes at the cost of a collector current blocking effect caused by the positive conduction band discontinuity between the InGaAs base and the wider band gap InP collector. For an InP collector, a blocking barrier of 0.25 eV must be overcome between the base and the collector as shown by curve (a) in Figure 1.4 where the base-collector discontinuity ∆Ec impedes the electron minority carriers, dramatically reducing the device’s fT and the current gain. Various doping and /or composition grading schemes have been developed to alleviate this blocking effect [6]. In practice, sometimes the blocking effect is reduced and the current gain is improved by using an undoped InGaAs spacer layer between the base and the collector as in curve 13 (b) in Figure 1.4 and curve (c) where compositional grading is employed. Curve (d) shows the favorable band lineup for the InP/GaAsSb DHBT of interest here. Figure 1.4 Conduction band profiles for a few design options for InP-based HBTs. [6] The grading schemes set forth in [12,13,14] are difficult in the sense that they introduce significant epitaxial growth complexity and impose stringent uniformity and repeatability requirements on the epitaxial growth. Even these grading schemes do not lend an easy or foolproof method for eliminating the collector-blocking effect at the basecollector interface, because high collector current densities cause a retarding potential, eventually degrading the transistor performance [15]. The specifics of any grading and/or doping schemes at the base/collector junction need to be considered carefully because they can have a dominant impact on DHBT performance at high current densities [16]. In 14 contrast, the use of a GaAsSb base layer provides an elegant solution to the collector blocking effect plaguing InGaAs-based DHBT designs. The staggered type II heterojunction band lineup at GaAsSb/InP heterojunctions allows unhindered injection of electrons from the GaAsSb base into the InP collector since the GaAsSb conduction band lines up 0.10–0.15 eV above the InP conduction band edge near the lattice- matched composition of GaAsSb [17], [18]. As a result, InP/GaAsSb DHBTs benefit from two fundamental advantages over InP/InGaAs DHBTs : 1) The structure requires no graded transition layers at the collector- or emitter-base junctions, thus simplifying the growth of the epitaxial layers and 2) C-doped GaAsSb features high doping efficiencies [19] and displays little or no H-passivation effects [15]. Also the low diffusivity and the good donor properties of Carbon doping ensures that the base-collector junction is precisely and permanently self-aligned to GaAsSb/InP interface leading to stability and repeatability of device characteristics. 1.6 InP/GaAsSb/InP Double Heterojunction Bipolar Transistors As discussed in the previous section, an excellent solution for the collector blocking effect seen in InP/InGaAs DHBTs has been found by utilizing GaAsSb in place of InGaAs for the base layer. The staggered (type-II) bandgap lineup at the base-collector junction for GaAsSb/InP heterojunction shown in Fig.1.4 (curve d) is free from the difficulties discussed above [20, 21]. The energy band lineup of the GaAsSb with InP provides an elegant solution to the collector-blocking problem. At 300 K, the GaAs0.51Sb0.49 conduction band edge is above that of InP by 0.15 eV and the valence band 15 discontinuity is 0.78 eV as shown in Figure 1.5 [6]. This results in a non-blocking conduction band profile at the base-collector heterojunction, eliminating the current blocking effect faced in conventional InGaAs DHBTs. There is no need for compositional grading at all! Figure 1.5 Band lineup of InP/GaAsSb heterojunction [3] The initial thrust in exploiting this system was delayed due to concerns with low mobility and high resistivity in GaAsSb. The earliest reported device using this material system was by Bhat et al. [22] and McDermott et al. [15] in 1996. The ir devices had fT and fmax performance of around 30GHz and 45GHz, respectively. These results discouraged further development using this material system for a while, considering the difficulty in growing GaAsSb layers with low resistivity. Also the mobility of GaAsSb is roughly 50-60% of that in InGaAs for a given concentration. While there is a conduction band barrier at the emitter-base junction impeding electron injection into the base from the emitter, the effect is not fatal and high performance InP/GaAsSb DHBTs have been reported [3]. Recent developments have shown impressive performance gains in terms of 16 cut-off and maximum frequency of operation. The following table gives a series of performance benchmarks in the development of the device. Table 1.1 Evo lution of InP/GaAsSb/InP DHBT device development AC Gain dB fT (GHz) fmax (GHz) Base/Collector Thickness A0 Emitter Size µm2 45 N/A N/A 350-1500/3000 70X70 27 30 45 500/3000 2X20 50* 75 23 400/3000 5X12 >100 78 N/A 400/3000 5X12 39 305 230 200/2000 5X0.5 39 270 >300 250/2000 5X0.5 42 300 300 200/2000 0.4X11 >80 N/A N/A 500/2000 20X20 Group McDermott et al 1996 [15] Bhatt et al 1996 [22] Bolognesi et al 1998 [23] Hu et al 1999 [24] Dvorak and Bolgonesi et al 2000 [6] Dvorak and Bolgonesi et al 2000 [6] Dvorak and Bolognesi et al 2001 [3] Rajavel et al 2003 [25] * DC Gain The MOCVD grown, Carbon doped, abrupt heterojunction, InP/GaAsSb/InP DHBTs feature a very small VCE offset voltage < 0.1V and a low turn-on voltage of VBE of 0.4 V at a current density of JC =1 A/cm2 , by taking advantage of the staggered band lineup at the InP/GaAsSb interfaces [11]. Electrons are injected ballistically from the GaAsSb base into the wideband gap InP collector where they travel with a saturated high drift velocity and low rates of impact ionization. This is especially important as the minority carrier 17 electron mobility in the p-doped base is the order of 600-800 cm2 /Vs. This is low when compared to similarly doped InGaAs layers with electron mobilities of 2000-3000 cm2 / Vs in InP/InGaAs HBTs. This lower mobility necessitates the use of thinner bases for GaAsSb than InGaAs to achieve comparable base transit times. However, to maintain an acceptable base sheet resistance and high fmax a higher base doping is necessary. This need for higher doping is solved by GaAsSb’s affinity for Carbon doping and low Hpassivation effects [20]. 1.7 Advantages of InP/GaAsSb/InP DHBTs The InP/GaAsSb staggered bandgap line-up has the added advantage, besides that of having no collector blocking, the electrons from the base are launched into the basecollector depletion region with a high velocity exceeding their saturation velocity, a phenomenon known as the velocity overshoot [4]. The massive valence band discontinuity (~0.8 eV) at the emitter-base junction increases the emitter injection efficiency by reducing the hole back injection from the base into the emitter. In addition, the small energy gap of GaAsSb results in low emitter-base turn-on voltages, which make the devices attractive for low-power, long-talk-time portable communication systems. Analyzed from multiple points of view, the InP/GaAsSb/InP DHBT offers excellent performance capabilities, by taking advantage of several bandgap engineering techniques. The interesting part is that all these advantages come without resorting to any dopant or compositional grading. The challenge is that the GaAsSb material system has not been extensively studied and only a few have reported success in growing it [3,11,15,17,18]. Recently, Bolognesi et al. [3] have demonstrated InP/GaAsSb/InP HBTs with excellent 18 high frequency performance. These HBTs feature the best performance ever reported for a DHBT in any material system with fT and fmax over 300 GHz with a AC current gain of 42 dB. They have overcome the speed advantage traditionally held by InP HEMTs and InP/GaInAs HBTs, while maintaining good breakdown voltages and superior current drive capability. A competitive analysis of the performance of the devices is given Figure 1.6 [26]. The figure illustrates that HEMTs generally lose their fmax advantage over HBTs around 250 GHz, while InP/GaAsSb DHBTs close the significant cutoff frequency gap between HEMTs and HBTs. The above consideration leads to the conclusion that HEMTs may no longer offer significant intrinsic performance advantages over HBTs for ultrahigh speed circuits at 80 Gb/s and beyond [26]. Figure 1.6 Comparison of reported fT and fmax values for InP HEMTs, GaInAs DHBTs and InP DHBTs [26] 19 The collector breakdown voltage BVCEO is a major concern when evaluating the performance of the HBTs. The use of a wide bandgap collector in InP/GaAsSb/InP HBTs results in good breakdown voltages. Figure 1.7 shows that InP/GaAsSb/InP HBTs feature some of the best fT vs breakdown voltage product for any of the InP based device [6]. Figure 1.7 fT values vs. BVceo for InP collector DHBTs from literature (n) M.W. Dvorak et al (6) One other big advantage of the InP/GaAsSb/InP HBTs is that it does not suffer from the conventional Kirk effect and alloy potential blocking effect that leads to the current degradation in other InP based HBTs. The use of a C-doped GaAsSb base layer enables the fabrication of ultrahigh-speed InP collector DHBTs without the additional complication of graded layers at the emitter-base and base-collector heterojunctions. The favorable band alignment in InP/GaAsSb/InP DHBTs allows high collector current densities required for high-speed logic applications, and eliminates the alloy potential 20 problems associated with graded interfaces under high current operation [27]. Figure 1.8 shows a comparison of the fT fall-off rates normalized to the peak fT of 300 GHz of InP/GaAsSb/InP DHBTs [26]. In our case, the high-current degradation follows a different mechanism from that of Kirk effect normally seen in AlInAs/GaInAs DHBTs and GaInP/GaAs SHBTs. A high current induced thermionic barrier at the base-collector junction of the InP/GaAsSb HBT reduces the electron exit velocity in the collector leading to a fall in the fT and fmax performance [28]. However, this roll-off is pushed to higher current densities in InP/GaAsSb DHBTs and the resulting DHBTs can be operated at current densities approaching those found in high-performance SiGe/Si bipolar transistors with the attendant reduction in dynamic emitter resistances and associated charging delays [26]. Figure 1.8 Normalized cut-off frequency dependency on collector current density for various DHBT and SHBT technologies [26] More recently, Bove et al. [29] have reported a proof-of-concept demonstration DHBTs using GaAlInAs as the emitter layer. This eliminates the small conduction band 21 spike seen in InP/GaAsSb/InP HBTs at the emitter/base interface. No microwave performance evaluations have been reported. Even though using a GaAlInAs might have superior transport properties in terms of its bandgap line- up with GaAsSb, the maturity of InP growth techniques and its material advantages will likely allow it to outperform any other device configurations for some time to come. 1.8 Motivation and Organization of the Thesis With the advantages for GaAsSb-based HBTs as enumerated in the preceding paragraphs, it is desirable to study and quantify the performance capabilities of the InP/GaAsSb/InP DHBT and to optimize its design. Computer based simulation tools have become an integral part of the modern design process, helping speed up the time to market of integrated circuit products. CAD tools are useful for resolving design issues like doping and layer thickness. Moreover, they can help us understand transistor performance and offer deep insight into the physics behind their operation. By far, the most important advantage is that simulation tools can drastically reduce fabrication costs by providing the manufacturer with the optimum structured device. Ultimately, device fabrication and characterization are required to finalize the device design. This thesis is focused on the design and simulation of InP/GaAsSb/InP Do uble Heterojunction Bipolar Transistors (DHBTs). The first chapter, as described above, has given a brief overview of the operation of BJTs and HBTs, their differences in operation, the advantages of a double heterojunction bipolar transistor as applied to InP/GaAsSb material system and its advantages over InP/InGaAs DHBTs. Chapter 2 gives a brief overview of the software use for simulation and summarizes the properties of InP, 22 GaAsSb and InGaAs used in the device modeling. In Chapter 3 an InP/GaAsSb/InP HBT structure is defined, simulated and its performance compared with the published results. The effects of various changes, like doping, base width and structure on the device performance is analyzed in Chapter 4, and the optimum structure for best performance is determined. Performance degradation due to high current effects is also examined. Chapter 5 summarizes the results of the thesis and suggests future improvements and progress that can be made. References: [1] J. Bardeen and W.H. Brattain, “ The Transistor, a Semi-conductor Triode”, Phys. Rev., vol. 74, pp.230-231, Jul. 1948 [2] Shockley, W., US patent No. 2.569,347, 1951 [3] M.W. Dvorak, C.R. Bolognesi, O. J. Pitts and S.P. Watkins, “300 GHz InP/GaAsSb/InP Double HBTs with High Current Capability”, IEEE Electron Device Letters, vol. 22, pp 361-363, 2001 [4] Liu, W., “Handbook of III-V Heterojunction Bipolar Transistors”, John Wiley and Sons Inc., New York, pp 26-27, 1998 [5] H. Kroemer, “Heterostructure Bipolar Transistors and Integrated Circuits”, Proc. Of the IEEE, vol. 70, No 1. pp. 13-25, 1982 [6] M. W. Dvorak, O.J. Pitts, S.P. Watkins and C.R. Bolognesi, “Abrupt Junction InP/GaAsSb/InP Double Heterojunction Transistors with FT as High as 250 GHz and BVCEO >6 V, IEDM, Tech. Digest, pp. 178-182, 2000 23 [7] O. Osamu, H.Hasegawa, “ InP-Based Materials and Devices-Physics and Technology” John Wiley & Sons, NewYork pp.53-55, 1999 [8] P.M. Asbeck, “Bipolar Transistors” High Speed Semiconductor Devices, S.M. Sze editor, New York, Wiley, 1990, pp.335-397 [9] G. Raghavan, M. Sokolich and W.E. Stanchina, “Indium Phosphide ICs Unleash the High Frequency Spectrum” IEEE Spectrum, vol. 37 No 10, pp 45-49 2000 [10] J. W. Rodwell, M. Urteaga, T.Mathew, D.Scott, D.Mensa, Q. Lee, J.Guthrie, Y. Betser, Suzanne C. Martin, R. P. Smith, S. Jaganathan, S. Krishnan, S.I. Long, R. Pullela, B.Agarwal, U.Bhattacharya, L. Samoska, and M. Dahlstrom, “Submicron Scaling of HBTs”, IEEE Trans. On Electron Devices, vol.48 pp.2606-2624, 2001. [11] C.R. Bolognesi, M.W. Dvorak, N. Matine, P. Yeo, X.G. Xu and S.P Watkins, “InP/GaAsSb/InP Double HBTs: A New Alternative for InP-Based DHBTs”, IEEE. Trans on Electron Devices, vol 48, No 11, pp. 2631-2639, 2001 [12] K. Kurishima, H. Nakajima, T. Kobayashi, Y. Matsuoka, and T. Ishibashi, “ HighSpeed InP/InGaAs Double-Heterostructure Bipolar Transistors with Suppressed Collector Current Blocking”, Appl. Phys. Lett., vol 62 pp. 77-79, 1998 [13] E.F. Chor and C. J Peng, “Composite Step-Graded Collector of InP/InGaAs/InP DHBT for Minimized Carrier Blocking”, Electron. Lett., vol 32, pp. 1409-1411, 1996 [14] C .Nguyen, T. Liu, M. Chen, H.C Sun and D Rensch, “AlInAs/GaInAs/InP Double Heterojunction Bipolar Transistor with a Novel Base-Collector Design for Power Applications” , IEEE Electron Devices Lett., vol 17, pp.133-135, 1996 24 [15] B. T. McDermott, E. R. Gertner, S. Pittman, C. W. Seabury, and M. F. Chang, “Growth and doping of GaAsSb via metalorganic chemical vapor deposition for InP heterojunction bipolar transistors,” Appl. Phys. Lett., vol. 68, pp. 1386–1388, 1996. [16] C.R Bolognesi, M.W Dvorak, O.J Pitts, S.J Watkins, T.W Macelwee, “Investigation of High Current Effects in Staggered Line- up InP/GaAsSb/InP Heterostructure Bipolar Transistors”, Presented IEEE Electron Devices Meeting, 2000. [17] M. Peter, N. Herres, F. Fuchc, K. Winkler, K.H. Bachemand J. Wagner, “Band Gaps and Band Offsets in Strained GaAs1-y Sby on InP Grown by Metalorganic Chemical Vapor Deposition.” Appl. Phys. Lett. vol 74. pp. 410-412, 1999. [18] J. Hu, X. G. Xu, J. A. H. Stotz, S. P. Watkins, A. E. Curzon, M. L. W. Thewalt, N. Matine, and C. R. Bolognesi, “Type II photoluminescence and conduction band offsets of GaAsSb/InGaAs and GaAsSb/InP heterostructures grown by metalorganic vapor phase epitaxy,” Appl. Phys. Lett., vol. 73, pp. 2799–2801, 1998. [19] S. P. Watkins, O. J. Pitts, C. Dale, X. G. Xu, M. W. Dvorak, N. Matine, and C. R. Bolognesi, “Heavily carbon-doped GaAsSb grown for HBT applications,” J. Cryst. Growth, vol. 221, pp. 59–65, 2000 [20] K.I Anastasiou, “ GaAsSb for Heterojunction Bipolar Transistors”, IEEE Trans Elect. Devices, vol. 40, pp. 878-884, 1993 [21] C. R. Bolognesi and S.P. Watkins, “ InP-based Double Heterojunction Bipolar Transistors: It May Not Have to be GaInAs”, Compound Semiconductors, pp. 94-99, 2000. [22] R. Bhat, W-P. Hong, C. Caneau, M. A. Koza, C-K. Nguyen, and S. Goswami, “InP/GaAsSb/InP and InP/GaAsSb/InGaAsP Double Heterojunction Bipolar Transistors 25 with a Carbon-doped Base Grown by Organo- metallic Chemical Vapor Deposition”, Appl. Phys. Lett. ,vol 68, No (7), pp 985-987,1996. [23] C.R. Bolognesi, N. Matine, M.W. Dvorak, X.G. Xu, J.Hu, S.P. Watkins, and M.L. W Thewalt, “Nearly Ideal InP/GaAsSb/InP Double Heterojunction Transistors with Ballistically Launched Collector Electrons”, Electronics Letters, vol. 34, no. 17, pp. 1700-1702, 1998. [24] X. G. Xu, J. Hu, and S. P. Watkins N. Matine, M. W. Dvorak, and C. R. Bolognesi “Metalorganic Vapor Phase Epitaxy of High-quality GaAs0.5Sb0.5 and its Application to Heterostructure Bipolar Transistors”, Appl.Phys.Lett., vol. 74, pp. 776-778, 1999. [25] R.D. Rajavel, T. Hussain, M.C. Montes, M.W. Sawins, S. Thomas III, and D.H. Chow, “Molecular Beam Epitaxial Growth and Characterization of InP/GaAsSb/InP Double Heterojunction Bipolar Transistors”, J. of Crystal Growth, vol. 251, pp. 848–851, 2003. [26] C.R. Bolognesi, N. Matine, M. W. Dvorak, O.J. Pitts, and S.P. Watkins, “Ultrahigh Performance Staggered Line InP/GaAsSb/InP NPN Double Heterojunction Transistors”, Jpn. J. Appl. Phys., vol. 41, pp1131-1135, prt.1, No.2B, 2002 [27] C.R. Bolognesi, N. Matine, M.W. Dvorak, X.G. Xu, J.Hu and S.P. Watkins, “ NonBlocking Collector InP/GaAsSb/InP Double Heterojunction Bipolar Transistors with a Staggered Lineup Base-Collector Junction”, IEEE Electron Device Lett., vol 20, pp 155157 , 1999. [28] C.R. Bolognesi, M. W. Dvorak, O.J. Pitts, S.P. Watkins, and T.W MacElwee, “Investigation of High Current Effects in Staggered Lineup InP/GaAsSb/InP 26 Heterostructure Bipolar Transistor: Temperature Characterization and Comparison to conventional Type-I HBTs and DHBTs” IEDM 2001, pp. 768-771, 2001 [29] P. Bove, H. Lareche, and R. Langer, “ InP/GaAsSb and (Al, Ga) InAs/GaAsSb DHBT Material Grown in a Four Inches Multi-wafer MBE Machine” Indium Phosphide and Related Materials Conference, 2002, IPRM 14th , pp. 607 -610, 2002. 27 Chapter II Simulation Software, Techniques and Material Properties 2.1 Introduction The core theme for this thesis is the use of Computer Aided Design (CAD) tools for the simulation and modeling of HBTs. Simulation allows us to design, test, and evaluate quickly semiconductor devices without physically fabricating them. This translates into huge advantages in terms of cost reduction and vastly reduced time to market, as different devices structures can be evaluated in a very short time. In addition, simulation packages are powerful tools that can help us to obtain device performance parameters such as potential, carrier and current profiles within the device during operation, which cannot be measured experimentally. The simulation procedure followed for this work is physics-based, that is based on the processes occurring within the device and their description mathematically. Semiconductor devices can be modeled in two ways; one is to determine the terminal electrical properties of a device based on empirical results and model the device using techniques like spline fitting. The other method is to study the carrier transport processes taking place within the device. The later, physics-based simulation provides three major advantages: 1) it is predictive, 2) it provides insight, and 3) it captures knowledge in a way that makes it easier to understand the various phenomena acting within the device. [1]. We use a physics-based approach to modeling of InP/GaAsSb DHBTs. In this chapter we give a description of the software used for the simulation, the physics behind it and 28 the actual device physics models used. The material parameters used for the different layers of the device are also summarized. In particular, the material properties of GaAsSb needed for the simulation were derived from the published literature and are reported here. 2.2 Simulation Technique The simulation software package used for this work is a commercial device simulator ATLAS, from Silvaco International [1]. ATLAS is a versatile, modular and extendable software product for two and three dimens ional device simulation. Simulation is done by discretizing the equations that describe the operation of a device onto to a two or three dimensional grid, consisting of a large number grid points called nodes. The ATLAS simulation package is built around physics-based models. These physics-based models are basically a mathematical description of the various phenomena, such as carrier generation, recombination and transport that are responsible for the operation of the device. The models describe the dynamics of carrier generation, recombination and injection as well as transport within different material regions of the device and coupling between the regions. These models are quantified in differential equations and transposed onto the grid to simulate the transport of carriers through a device structure. Electron and hole concentrations are calculated, along with their spatial variation and dependence on material properties as well as biasing. The simulator also predicts the electrical characteristics of the device, for example the terminal currents that are associated with the specified physical structures and different bias conditions. These results can be used to characterize the DC and AC operation of the device and to provide insights into the physical aspects of device operation, provided they include all the important device 29 physics. These models consist of a set of fundamental equations, which link together the electrostatic potential, charge density and the carrier densities within a simulation domain. An important advantage of this type of model is that it can not only be used to predict the terminal current-voltage characteristics of the devices for comparison with device measurements, but also to investigate the origins of the observed behavior and the factors limiting the device performance. 2.3 Device Physics and Models The simulation technique is a numerical one, which arrives at the terminal characteristics of the device by solving the partial differential equations describing the physics of the materials and the effects of potentials and heterojunctions on carrier transport. These equations include Poisson’s equation, the carrier continuity equations and the transport equations for each carrier (electrons and holes). These equations must be solved simultaneously and self consistently in each region of the device. Here, a brief description of the basic semiconductor equations involved is provided. 2.3.1 Poisson’s Equation Poisson’s equation relates the electric field distributio n with the distribution in space of the charge and is given in one dimension by [2] dE( x) q = ( p − n + Nd+ − N a− ) dx εs (2.1) where E(x) is the electric field, ε s the dielectric constant of the semiconductor, p and n are the carrier concentrations and N d+ , N a− are the concentrations due to ionized donor and acceptor impurities, respectively all of which can be a function of position. Clearly, 30 the doping in each region of the device must be specified at the outset in order to describe the electric field everywhere within the device. An alternate form of the Poisson equation is obtained when the electric field E(x) in (2.1) is replaced by –dV/dx as given in (2.2). Generally both the equations are referred to as the Poisson’s equation. E (x ) = − dV dx (2.2) An important point that should be noted when working with HBTs is that the device is composed of many layers of different semiconductor materials. So we have to take into account the “quasi-electric” field contributed by the gradient of the electron affinity in the case of compositional grading. Then the total electric field inside the space charge region is the sum of the electric field contribution due to Poisson’s field and the quasi-electric field. E ( x ) total = E ( x ) poisson + E ( x ) q u a s i (2.3) 2.3.2 Carrier Continuity Equations The continuity equations for the electrons and holes are given by [1, 2] ∂n 1 ∂ = • J n + Gn − R n ∂t q ∂x (2.4) ∂p 1 ∂ =− • J p +Gp − R p ∂t q ∂x (2.5) where n and p are the electron and hole concentrations, J p and J n are the total electron and hole current densities, G p and Gn are the generation rates for electrons and holes due to optical absorption, Rp and Rn are the net recombination rates for electrons and holes, respectively, and q is the magnitude of the charge on an electron. The rates of 31 recombination, Rn and Rp are proportional to the excess carrier concentration and are given by Rn = (n-n0 /τn ) and Rp = (p-p0 /τp ) where τn and τp are lifetimes of the electrons and holes, respectively. 2.3.3 Transport Equations The hole and electron current densities J p and J n are important terms in (2.4) and (2.5) and therefore they justify further elaboration. These two terms are derived from the drift-diffusion model relating the carrier concentrations to the current density and are given as follows [1, 2] J p = q µ p pE ( x) − qDp J n = q µn nE ( x ) + qDn dp( x) dx dn ( x ) dx (2.6) (2.7) where µn and µ p are electron and hole mobilities, and D p and Dn are hole and electron diffusion constants. The first term involving the E(x) is called the drift current while the second term refers to the diffusion current. By default, ATLAS solves both equations. However, for some specific circumstances, it is sufficient to solve for only one carrier. Device modeling consists of the solution to the above defined five equations (Poisson, continuity and current density equations) subject to boundary conditions imposed by the device geometry and biasing. The previously discussed set of five equations provides a basic framework for device simulation once a device structure and its biasing have been specified. These current density equations are based on transport models that are obtained by applying approximations and simplifications to the Boltzman transport equation. However, to completely and accurately define the device model, it is essential that each of the parameters in (2.1) to (2.7) be further defined by secondary equations. For 32 example, the electron mobility maybe a function of position x if the doping is not uniform, since the mobility varies with the doping level. In the simplest cases the material parameters are just constants, but they must be specified for each material incorporated in the device. Other models incorporated in the simulation include ones for the carrier generationrecombination mechanism, band gap narrowing model at high doping density, density of states for different materials and velocity saturation effects at high electric fields [1]. 2.4 Material Parameter Models Many properties of the materials are important for device modeling such as energy bandgap, the effective mass, mobility, saturation velocity, and recombination lifetimes. These parameters are typically dependent on the local concentration of the dopants and the temperature as well as the composition. Material models are needed for device modeling that incorporate these effects and are briefly discussed here. 2.4.1 Mobility Modeling The carrier mobility is a parameter that characterizes the drift motion of the carriers in the presence of an electric field. The electrons and holes are accelerated by electric fields, but lose momentum because of various scattering process involving lattice vibrations, impurity ions, material interfaces and other imperfections [3, 4]. The effects of these microscopic phenomena are averaged and lumped together into the macroscopic parameter called the mobility, which is a function of the local electric field, doping concentration, and temperature. Though a variety of models are available for 33 characterizing the mobility’s behavior in low and high electric fields, we use two empirical models that are widely used. These are 1) Caughey-Thomas doping dependent mobility model and 2) Parallel electric field dependent mobility model Each of these models is described in more detail below. 2.4.1.1 The Caughey-Thomas Model Also known as the analytic low field mobility model, this model specifies the mobility dependence on doping and its variation with temperature. However, our simulations are done at 300 K, so the temperature dependence of the mobility is not employed. In its simplest form, the model is given by [3] µn ( N ) = µmin + µ max − µ min N a 1+ ( ) Ncr (2.8) where N is the total dopant density, µ max , µ min are maximum and minimum possible mobility, at low and high doping, respectively, NCr is the critical doping above which the mobility degrades, and α is a fitting parameter. However, the Caughey-Thomas model [3] does not take into account the mobility difference between majority and minority carriers, as these undergo somewhat different scattering mechanisms. However, there is little data to distinguish these, so a single mobility will be used. For each of the materials in the InP/GaAsSb/InP HBT these parameters, for the mobility model for electrons and holes for (2.8) must be specified. These are described for InP and GaAsSb later in this chapter. 34 2.4.1.2 Parallel Electric Field Dependent Mobility As the electrons and holes are accelerated by the electric field, their drift velocities initially increase linearly. However, these velocities begin to saturate at higher electric fields ~105 V/cm. So the mobility becomes a function of the electric field strength. The following expression is an empirical one widely used to describe the mobility’s variation between low and high electric fields [3] 1 1 µn ( E ) = µn 0[ ] BETAN µ E 1 + ( n0 ) BETAN VSATN (2.9) 1 1 µ p ( E ) = µ p 0[ ] BETAP µ p 0E BETAP 1+ ( ) VSATP (2.10) where E is the parallel electric field, µn0 , µp0 are the low field electron and hole mobilities, respectively, derived from the Caughey-Thomas model, VSATN and VSATP are the saturation velocities, and BETAN and BETAP are fitting parameters. The values for these for each of the semiconductor materials of interest are given in the following sections. 2.4.2 Carrier Generation-Recombination Models When we consider a semiconductor material whether doped or undoped, there always exists a thermal equilibrium concentration of holes and electrons due to a balance between thermal generation and electron-hole pair recombination due to continual excitation due to processes like phonon and photon emissions. These processes are responsible for the generation-recombination of electron hole pairs, and are the mechanisms by which the carrier concentrations return to equilibrium. For undoped material, the electron and hole concentrations are equal and given by 35 n0 = p0 = nie (2.11) where nie is the intrinsic carrier concentration given by − Eg nie = N c N v e 2KT (2.12) where Eg is the semicond uctor energy bandgap. Nc and Nv are the effective density of states in the conduction and valence bands, respectively, given by N c = 2( N v = 2( 2π kT 3/2 *3/2 ) mde 2 h 2π kT 32 *3/2 ) ( mlh + m*3/2 hh ) 2 h (2.13) (2.14) where h is the Planck constant, mde is the density of states effective mass of the electrons, and mlh and mhh are the light hole and heavy hole effective masses, respectively. A detailed description of the effective mass values for the GaAsSb material system is given section 2.5.4. For an n-type semiconductor, the thermal equilibrium concentrations for electron and holes, respectively are given by n0 ; N D+ ; N D (2.15) p0 = nie2 n0 For heavy doping, the energy bandgap is reduced so nie is enlarged and the minority carrier concentration p0 is increased [1]. The duration the electron hole pairs remain without recombining determines their lifetime in the material. It is an important parameter because it can influence the current gain in an HBT. ATLAS implements carrier lifetime modeling by using the ShockleyRead-Hall (SRH) recombination model where the lifetime is dependent on the impurity concentration [1,4], 36 RSRH = pn − n 2ie ETRAP − ETRAP τ p [n + nie exp( )] + τ n[ p + nie exp( )] kTL kTL TAUN 0 1 + N /( NSRHN ) TAUP 0 τp = 1 + N /( NSRHP ) τn = (2.16) where ETRAP is the difference between the trap energy level and the intrinsic Fermi level, TL is the lattice temperature in degree Kelvin, TAUN0 and TAUP0 are the electron and hole lifetimes at very low doping, N is the local doping, and NSRHN and NSRHP are the doping levels beyond which the carrier lifetimes decrease. The parameters τn and τp correspond to the carrier lifetime in the material for a doping level N. 2.5 Material Properties of GaAsSb In order to precisely simulate and predict device behavior, it is necessary that we work with accurate values of the material parameters. For the HBT under consideration, three different materials are involved: InP, GaAs0.51 Sb0.49 and In0.53Ga0.47 As, the last two being lattice matched to InP. The material parameters of InP and InGaAs have been well documented, and included in the ma terials library of Silvaco, whereas it is not so in the case of new materials like GaAsSb. As in the case of InP and InGaAs, GaAsSb is an IIIV compound semiconductor material. The material properties of Gallium Arsenide Antimonide (GaAsSb) depend upon its composition, i.e. Antimonide content. The most commonly used composition for HBTs is GaAs1-x Sbx where X =0.49, which gives the composition that is lattice matched to InP. The advantage of this system over other ternary material systems is due to its large valence and small conduction band discontinuity when it forms a heterojunction with InP. When used as a base in n-p-n 37 InP/GaAsSb/InP HBTs, the large valence band offset of 0.81 eV and the favorable conduction band offset of 0.15 eV eliminates the collector current blocking effect at the collector base junction that degrades the performance in the InP/GaAsSb/InP that degrades the performance in InP/InGaAs/InP double heterojunction HBTs [5]. The interest in GaAsSb is relatively new and it has only recently been seriously exploited as a material for HBTs. As a result, its material parameters are not well known or widely reported in the literature. When studying the material properties of GaAsSb, it is frequently necessary that we start with the properties of the better known binaries of GaAs and GaSb and infer the properties of GaAsSb by interpolation. The material properties summarized here are not comprehensive and only those properties needed for device simulation are included. For example, the temperature dependence of the properties is not described as all simulations are done at 300K with no device self- heating due to current flow assumed. However, the temperature dependent properties of the material have been reported elsewhere [6,7]. 2.5.1 Crystal Structure and Lattice Matching GaAsSb has a Zinc Blende crystal structure as shown in Figure 2.1. This follows from that of the binaries that constitute GaAsSb, since GaAs and GaSb have the same Zinc Blende structure. In order to grow defect free layers of GaAsSb, it is a necessary and important condition that GaAsSb be latticed matched to InP. The lattice constant of InP is 5.8687A. The lattice constant of GaAs1-x Sbx varies with the composition x, where GaAs1x Sbx can be considered as a ternary as part of the InGaAsSb material system. 38 Figure 2.1 GaAsSb crystal structure where Sb is incorporated in some of the As sites [2] Figure 2.2 shows the plot of the lattice parameter versus composition for the quaternary In1-x Gax Asy Sb1-y . For deducing the lattice constant of GaAsSb, we can start by neglecting the composition of Indium in the quaternary to simplify to the ternary and use Vegard’s law to arrive at the composition of Sb, i.e. x in GaAs1-x Sbx, that is lattice matched to InP. Then we have, a(x) = 5.6532(1-x) + 6.09593x, [7] where 5.6532A and 6.09593A are the lattice constants of GaAs and GaSb, respectively. For the lattice matched condition where a(x) =5.8687A, matched to the lattice constant of InP, the value of x would be = 0.49, i.e. the composition of Sb in GaAsSb. Similarly, In0.53Ga0.47 As is the composition of the InGaAs ternary which is lattice matched to InP. 39 Figure 2.2 Lattice parameter versus composition plot for In1-x Gax Asy Sb1-y and compositions (dotted line) lattice matched to InP [8] 2.5.2 GaAsSb Energy Band Gap The energy band diagram of the binaries GaAs and GaSb are shown in Figure 2.3 [7]. From the diagram we can see that both of them have a direct energy band gap. The energy band profile for GaAsSb will be similar but with a different bandgap. Due to the relatively new interest in the material, few references agree on the precise value of the energy bandgap. Theoretically, the bandgap of GaAsSb can be deduced empirically using Vegard’s law from that of the quaternary In1-x Gax Asy Sb1-y as shown in Figure 2.4 based on the composition determined from the lattice parameter as described above. The energy gaps of GaAs and GaSb are 1.42eV and 0.70eV, respectively. For our case neglecting the Indium composition, we arrive at the value of 1.09eV for GaAs0.51 Sb0.49. 40 Figure 2.3 Energy Band structure of the binaries GaAs (a) and GaSb (b) [8] Figure 2.4 Plot of energy bandgap versus composition for the quaternary In1-x Gax Asy Sb1-y for the material lattice matched to InP (dotted line) [8] 41 The empirical reported variation of the bandgap for GaAs1-x Sbx with the composition of Sb is shown in Figure 2.5 [6, 9]. The empirical relations reported for the bandgap variations are shown in equations (2.17) (Klem et al [6]) and (2.18) Merkel et al [9]). Eg (x) =1.35x2 -2.05x+1.42 (2.17) Eg (x) =1.2x2 -1.9x+1.52 (2.18) Considering the above two equations, the one given by equation 2.17 is more acceptable when determining the bandgap variation with composition, because it’s closer to experimental values reported by Bo lognesi et al [5]. The experimentally measured bandgap of MOCVD grown, undoped GaAs51 Sb49 has been reported as 0.72 ±0.01eV at 300K [5] where the energy band gap was measured using optical absorption. This measurement agrees with the calculated value of Eg=0.720eV obtained from Klem et al’s [6] empirical expression for Eg (x) at T=300K. Figure 2.6 shows the experimental confirmation for the variation of bandgap with composition at different temperatures. Curve 1 shows the energy separation between the L-valley and top of the valence band, while curves 2, 3, 4 shows the variation in bandgap with composition at 100K, 210K and 300K respectively. The bandgap value reported in [5] is given preference because the group has built actual working devices using the above composition. In the following Table 2.1 the different bandgap values reported by various sources are summarized. 42 Figure 2.5 Variation of GaAs1-x Sbx bandgap with Sb content • Klem et al et al. [6] and ♦Merkel et al [9] Figure 2.6 Energy separation between L- valley of conduction band and top of the valence band Curve 1 [Rosenbaum and Woolley ], Curves 2-4; [Taylor and Fortin], versus composition x. 2 - T=100 K, 3 - T=210 K, 1, 4 -T=300 K. [7] 43 Table 2.1 Bandgap of GaAs0.51Sb0.49 Energy gap eV Source 1.09 Vegard’s Law 0.72 Bolognesi et al (5) 0.72 Klem et al (6) 0.8772 Merkel et al(9) 0.7396 Matine et al(10) Figure 2.7 shows the 10 K PL peak positions of all investigated samples plotted versus the Sb concentration [11]. It provides direct experimental evidence for the bandgap of GaAsSb. Except for the lattice matched composition when Sb content = 0.49, the material is either under tensile or compressive stress. The band energy diagram is also a good way of predicting the lattice matched condition of GaAsSb grown on InP. For example, GaAs1-x Sbx under tensile strain (x<0.49), the light hole band (lh) lies above the heavy hole band (hh) at the center of the Brillouin zone, and therefore the transition between the conduction band (cb) and the lh band forms the lowest band gap. For compressively strained GaAs1-x Sbx, for values of x>0.49 the hh band is located above the lh band which makes the cb-hh transition the lowest Bandgap. In the lattice matched case the hh band and lh band are degenerated and band gap is identical to that of unstrained GaAsSb, thereby establishing the lattice matched condition of GaAs0.51Sb0.49 . 44 Figure 2.7 PL spectra for GaAsSb for various Sb concentrations [11] 2.5.3 Bandgap Offsets for InP/GaAsSb Heterojunctions We are very much interested in the bandgap offsets at the InP-GaAsSb heterojunctions because it plays a major part in determining the performance parameters of the npn HBTs, such as the cut-off frequency, the turn-on voltage, and the current gain. Figure 2.8 shows that the InP/GaAsSb heterojunction is a type II heterojunction with a conduction band offset ∆EC as 0.15 eV and the valence band offset ∆EV as 0.78eV as reported by Bolognesi et al [5]. However, we also have slightly different values reported by the same group [6], for ∆EC and ∆EV as 0.18eV and 0.81eV, respectively [5]. We follow the later values for simulation, because they have been the most recently reported 45 values for the offsets reported by this group. Type II band alignment in GaAsSb/InP heterojunctions has been proved experimentally by Hu et al. [12]. With a band gap alignment as shown in the Figure 2.8 the electrons are injected into the GaAsSb base thermally, where as they are launched into the InP collector layer with high initial velocity corresponding to the conduction band discontinuity ∆EC. Such a mechanism is a beneficial factor affecting the electron transport in the base-collector depletion layer and ultimately the high frequency performance of these devices. Knowledge of the bandgap offsets is important because it helps us to model the device accurately in numerical simulations. However, none of the recent reports for this material system considers bandgap narrowing due to heavy doping effects. Because of bandgap narrowing, the bandgap offsets might change resulting in somewhat different performance of these devices. These effects become important considering the higher operating current of these devices particularly at high frequencies. Figure 2.8 Staggered Band line-up at InP/GaAsSb Heterojunctions [10] 46 Figure 2.9 shows how the offsets differ when GaAsSb is under tensile or compressive stress due to non- latticed matched composition and conditions with InP [11]. This is interesting to note because by varying the band offsets we can control the turn-on and the collector offset voltage for the HBT to some extent. The offsets are nothing but energy differences between the corresponding conduction and valence band lineups at the heterojunction interface. For Sb composition x>0.27, the conduction band offset becomes negative i.e., Ec (InP) < Ec (GaAsSb) and thus GaAs1-x Sbx on InP shows a staggered typeII band alignment, as shown in Figure 2.9. The conduction band (cb) offset and valence band (vb) offset offsets of strained GaAs1-x Sbx relative to InP can be curve fitted to ∆Ecb(x) = 0.13-0.49x (eV) and ∆Evb(x) = 0.43+0.64x (eV) for 0.2<x<0.7 [6], where ∆Ecb = Ec (InP)-Ec (GaAsSb) and ∆Evb = Ev (GaAsSb)-Ec (InP). The band offsets are defined as the energy difference between the respective conduction and valence band of the semiconductor materials, when the heterojunctions are lined-up. plotted in Figure 2.10. 47 The equations are Figure 2.9 Band offsets of GaAsSb on InP [11] We can see in Figure 2.10 that the ∆Evb varies more rapidly than ∆Ecb for variations in Sb composition. The highest valence band discontinuity that can be reached is 0.878eV, whereas for conduction band discontinuity it is 0.213eV. While base push out is certainly eliminated due to the large ∆Evb, its effect on high current fT roll-off is to be studied for the specific case of InP/GaAsSb HBTs. If the large ∆Evb helps in achieving higher current densities and so higher current gain and fT , then lattice mismatched GaAsSb can be grown for the base region, keeping in mind that for achieving a minimum base transit time, the thickness of the base must be small, i.e., 200-400A0 . Better device performance certainly can be achieved by exploiting the unique bandgap lineup. 48 Figure 2.10 Calculated Band offsets in GaAsSb on InP based on results of Merkel et al [6] (•) Ecb and (€) E vb 2.5.4 Effective Masses and Density of States Very few results have been published reporting the hole and electron effective masses for GaAsSb. The values presented here have been interpolated in part from that of the binaries GaAs and GaSb. From these hole and electron masses, the density of states can be deduced as a function of the Sb content. The empirical relationship for the electron and the hole effective masses for GaAsSb reported by Ya Vul et al [7] are given as follows mΓ = (0.063-0.0495x+0.0258x2 ).m0 (2.19) mh = (0.51-0.11x)m0 (2.20) mlp = (0.082-0.32x) m0 (2.21) where mΓ, mh, and mlp are the electron, heavy hole and light hole masses, respectively and x denotes the Sb alloy content. Figure 2.11 shows the variation of the hole effective 49 masses with the Sb content. The curves were plotted us ing the empirical relation shown in the above equations (2.20) and (2.21) [7]. Figure 2.12 shows the empirical change in the electron effective mass with the Sb content as a function of the relation in equation (2.19). The curve corresponds to the values traced using equation (2.19) and the dots represent the experimental values. We can see that the theoretical and experimental values for the electron effective mass are close together. However, there have been no reported experimental values for the hole effective mass in the literature. Figure 2.11 Empirical variation of hole effective mass with Sb content [9] (♦) Heavy hole mass (€) Light hole mass. 50 Figure 2.12 Variation of the electron effective mass with Sb Content [7] Table 2.2 Effective Masses of Electrons and Holes for GaAs, GaSb and GaAsSb Parameter GaAs [7] GaSb [7] GaAs 0.51 Sb0.49 [7] Electron effective Mass m0 0.063 0.041 0.0449 Heavy Hole Mass m0 0.51 0.4 0.4561 Light Hole Mass m0 0.082 0.05 0.0663 Table 2.2 summarizes the effective masses of the binaries GaAs and GaSb and the lattice matched composition of GaAsSb. It should be noted the values in Table 2.2 are all empirical ones calculated using equations (2.19), (2.20), and (2.21). We can see from the table that GaAs has higher electron effective mass than other two materials listed above. However, this does not result in higher mobility for GaAsSb, as we will see because it is 51 reduced due to alloy scattering effects in the material. The hole effective masses fall approximately midway between the binaries. The results for the lattice matched GaAsSb shown in Table 2.2 were used in our device simulation. Effective Density of States and Intrinsic Carrier Concentration The effective density of states of the carriers in the conduction and valence bands are essential parameters for device modeling since they determine along with the Fermi energy the equilibrium in density of electrons and holes. For the lattice matched case of x=0.49 using the electron and the hole effective masses given in Table 2.2, the effective density of states can be determined as Nc=2.39X1017 /cm3 and Nv =4.65X1018 /cm3 for the conduction and valence bands, respectively [7]. We can infer from the equations (2.13) and (2.14) that these parameters vary with temperature. The intrinsic carrier density is given by equation (2.12) at 300 K is given as 9.61X1011 /cm3 . This compares very well with experimental values of around 1012 /cm3 [7]. From the equation (2.12) we can predict an increase of intrinsic concentration ni with increasing temperature, as the effective densities increase with temperature as seen in equations (2.13) and (2.14). In addition, the energy bandgap also typically decreases as the temperature rises. This rise in ni has been experimentally confirmed by results from Ya Vul et al. [7] as shown in Figure 2.13. 52 Figure 2.13 Temperature dependence of intrinsic carrier concentration: 1) x = 0(GaAs), 2) x = 0.3, 3) x = 0.8, 4) x = 1 (GaSb) [7] 2.5.5 Mobility 2.5.5.1 Hole Mobility Carrier mobilities are important parameters for device modeling as they ultimately decide how “fast” the device operates. For our case where the base of the InP/GaAsSb HBT is P type, both the electron and hole mobilities are important. Measurements of hole mobility in lattice matched GaAsSb by Watkins et al. [13] indicate a very low value (~50 cm2 /v.sec). This has been attributed to strong alloy scattering mechanisms which dominates the hole mobility at doping levels up to the 1019 range. This importance of alloy scattering is supported by the experimental Hall data for a series of layers grown in the entire composition range from x=0 to x=1 which are summarized in Fig 2.14 [13]. All the samples had measured Hall concentrations in the range from 1x1019 to 1x1020 cm-3 . We can infer that the mobility is much lower for intermediate ranges of Sb content than the binary endpoints shown in Figure 2.14. The hole Hall mobility i.e. the majority carrier 53 mobility, for the p-type base in an InP/GaAsSb/InP HBT is 40cm2 /Vs at 1x1019 /cm3 for the lattice matched composition. For intermediate compositions, there is a slight tendency towards somewhat higher mobilities on the Sb rich side. [13], [14]. Figure 2.14 Hole Hall Mobility as a function of Sb in GaAsSb [13] Figure 2.15 shows the dependence of Hall mobility for holes on the carrier concentration for a number of samples grown at various temperatures and Sb mole fractions. The dependence of the Hall mobility on doping le vel is weak, because of the strong alloy scattering effect. This is interesting because the mobility does not change significantly even at high doping levels, though these results are for a limited range of doping. As a result, the base can be doped highly to reduce the base sheet resistance, without any significant reduction in the base transit time. 54 Figure 2.15 Hall Mobility for holes as a function of carrier concentration for various Sb compositions. Where different symbols correspond to the indicated Sb composition ranges [13]. For a given doping level and Sb composition, there are significant variations in the values of hole mobility, reported in the literature. However, for a given Sb composition, there are very few experimental results and over a limited range of doping, so a rigorous fit could not be attempted to the Caughey-Thomas mobility model for the doping dependence given by equation (2.8). As a result, the fitting parameters were chosen so that the derived Caughey-Thomas model fits reasonably well with the experimental values at the high doping end where the material will be used. The maximum and minimum reported values for hole mobilities are 60cm2 /v-s at a doping of 5X1016 /cm3 [14] and 20cm2 /v-s at 1X1020 /cm2 [15], respectively. Using this and other intermediate doping values from experimental results as 55 a starting point, the Caughey-Thomas equation can be solved for critical doping Nc and the constant α in the equation (2.8), repeated here for convenience, µ p ( N ) = µ min + µ max − µ min N 1 + ( )a Nc (2.8) The curve fitted Caughey- Thomas model for mobility variation with doping is shown in Figure 2.16. The data points with the (+) symbol represent a few of experimental values chosen from the data in Figure 2.15 for doping greater than 1019 /cm3 . There is good agreement with the experimental values towards higher doping levels, where the doping is comparable to the doping values usually used for our base in InP/GaAsSb HBTs. Figure 2.16 Hole mobility variation with doping: experimental values (+) [14] and calculated Caughey-Thomson values (•) 56 2.5.6 Electron Mobility The electron mobility in GaAsSb differs depending upon the type of dopant used. When N-type doping is attempted using diethyl telluride, the electron mobility varied from 2246 to 750 cm2 / V s for a doping concentration from 2X1017 to 1.5X1019 /cm3 [15]. GaAsSb is usually doped n-type using Silicon with majority carrier mobilities around 500-700cm2 /V s, significantly lower than for a material doped to a comparable level using diethyl telluride (DETe) [15]. However, it should be noted that, going by published results, DETe doping was not done for any fabricated InP/GaAsSb HBT. This might be due to concerns of dopant out diffusion from the thin base (200-400 Ao thick). The variation of mobility with doping for n-type dopants using the Caughey-Thomas model is shown in Fig 2.17. The model was fitted using the same principles outlined for determining hole mobility. The data points with the (+) symbol indicate the experimental values from the literature. No data was found for the minority carrier mobilities for electrons in p-type material, so the majority carrier mobility was used. Table 2.3 summarizes the Caughey-Thomas mobility parameters used in the device simulation. 57 Figure 2.17 Electron mobility variation with doping: experimental values (+) [13, 16] and calculated Caughey-Thomson values (•) Table 2.3 Caughey-Thomas Parameters for Holes and Electrons Hole µmax cm2 /V s 60 µmin cm2 /V s 20 7.935x1018 Fitting parameter ∝ 0.45244 Electron 2246 750 1.8199x1017 1.0577 Mobility Nc cm-3 2.5.7 Carrier Lifetimes The minority carrier lifetime in the base region is also critically important to the HBT device’s performance, influencing both current gain and cut-off frequency. The minority carrier lifetimes for GaAsSb were calculated from that of the binaries GaAs and GaSb [16, 17], since no experimental values have been reported for lattice matched GaAsSb The values adopted for the binaries are the longest reported lifetimes in the literature for 58 these binary materials. For our composition of the alloy lattice matched to InP, the values were calculated using Vegard’s law for n-type and p-type material, respectively. The values used are incorporated in the Shockley-Read-Hall recombination model (RSH Model) are listed in Table 2.4. Table 2.4 Electron and Hole Minority Carrier Lifetimes for GaAs, GaSb and Latticematched GaAsSb Material Electron Lifetime TAUN0(s) Hole Lifetime TAUNP0 (s) GaAs [16] 5x10-9 3x10-6 GaSb [17] 2x10-7 1x10-8 1.005x10-7 1.53x10-6 GaAs0.51Sb0.49 (Vegard’s Law) The critical doping NSHRN and NSHRP for which the lifetime falls-off in n and p type GaAsSb is 2e17 and 5e17 cm-3 , respectively. Then the variation of the lifetimes τn and τp with doping for latticed matched GaAsSb was calculated using equation (2.14) (RSH Model) and are plotted in Fig. 2.18. RSRH = pn − n 2ie ETRAP − ETRAP τ p [n + nie exp( )] + τ n[ p + nie exp( )] kTL kTL TAUN 0 1 + N /( NSRHN ) TAUP 0 τp = 1 + N /( NSRHP ) τn = 59 (2.14) Figure 2.18 Calculated carrier lifetime variation with doping (€) Hole lifetime, (•) Electron life time 2.6 Material Properties of InP and InGaAs 2.6.1 Indium Phosphide Indium Phosphide (InP) forms the collector and emitter material of the InP/GaAsSb/InP HBT studied here. InP is a well known direct bandgap material with a bandgap energy of 1.344eV [18]. Because of its application in almost all the III-V based devices and lightwave circuits, its properties have been extensively documented. We will not go into the details, but we will briefly summarize the mobility parameters relevant to our simulation along with that for InGaAs. Deducing the minority carrier lifetime for InP poses more of a challenge because it is a strong function of material quality and processing history. Hence, the default Silvaco material library values have been used for simulation. These values are summarized in the table to follow. 60 2.6.1.1 Heavy Doping Effects in InP Here we will briefly take a look at the effects of heavy doping in InP. This is important because of its effect on gain and reliability of the device. Heavily doped n-InP has been extensively studied by optical techniques, particularly the method of photoluminescence spectroscopy. The dominant feature in these spectra is the Burstein-Moss shift. Band filing manifests itself in the photoluminescence measurements as a broadening of bandto-band transition peak and an accompanying shift of that peak to higher energies as the (~1016 /cm3 ) doping level increases [18]. This effect can become visible even at low doping levels because of the low electron effective mass. Qualitatively, the cause of the shifting and widening of the band-to-band peak is due to the electronic transitions from states deep within the conduction band to states near the top of the valence band. Figure 2.19 PL spectra of InP [18]. 61 Here we characterize a series of five samples n-InP doped in the range of 2x1016 cm-3 to 4X1018 cm-3 to study the effect of doping such as bandgap narrowing. Photoluminescence measurements were made at room temperature (RT=294 K). The photoluminescence spectra are shown in Fig. 2.19. The RT data shows a single broad peak located slightly above the intrinsic bandgap. As the doping increases, the band filing effects manifests itself as described earlier as a broadening of the B-B peak coupled with a shift to higher energy. Detailed knowledge of the bandgap narrowing is important because it affects many critical parameters such as intrinsic carrier concentration, absorption co-efficient, emission spectrum and heterojunction band discontinuity, all factors used in device design. The band gap reduction often used in electrical device design is a non-physical quantity, which accounts for the degeneracy effects as well as physical band structure changes. In view of the use of hetero- interfacial band offsets for barrier formation, an understanding of the effects of doping upon band structure, and especially upon the fundamental bandgap, is essential for optimal device design. A significant modification of the InP band structure occurs due to carrier-carrier, carrier- impurity, and impurityimpurity interactions. The net result is to lower the energies of the conduction band states and raise the energies of the valence band states: the effect is therefore referred to as bandgap narrowing. However, practically no work has been done on the heavy doping effects in GaAsSb because of the relatively new interest in this material system. There are seve ral methods by which the extent of bandgap narrowing can be obtained from experimental photoluminescence data. There are several method by which this can be done namely by studying the luminescence peak energy, spectral line-shape fitting, 62 extrapolation of the low energy side of the spectrum, and direct observation of the extent of the band tail [18]. We will now characterize the bandgap narrowing only by using the last method. Fig. 2.20 shows the values obtained. Figure 2.20 Bandgap narrowing in InP due to heavy doping [18]. The bandgap narrowing is in the range of 0.08eV at ~5x1018 /cm3 for heavily doped samples when calculated by band tailing method. However, the exact allocations of the narrowing in the valence and conduction bands are not known. We can make an educated guess as to where this occurs. As a result of increasing doping, the valence band moves up and the conduction band comes down, we can expect a decrease in valence band offset ∆Evb and an increase in the conduction band offset ∆Ecb. 2.6.2 InGaAs InGaAs serves as the material for the emitter cap and sub-collector region. InGaAs has some of the highest reported mobilities among III-V materials. Detailed description of 63 these parameters is elaborated in the reference by Levinshtein et al [18], and Datta et al [19]. As discussed before, the minority carrier lifetimes are a strong function of doping and is modeled using the empirical expression [19], τ ( N ) = 10 β −γ log N (2.22) where ß and γ are fitting parameters summarized in Table 2.5 for InGaAs along with maximum lifetime values for electron and holes. Figure 2.21 shows the plot of electron and hole lifetimes as a .function of the doping level [19]. Table 2.5 Minority Carrier Lifetime Model Parameters for InGaAs Parameters Holes Electrons τmax (ns) 10 0.3 ß 22.4 12.6 γ 1.2 0.73 NC (cm-3 ) 8X1017 8X1017 Figure 2.21 Electron ( ) and hole (O) minority carrier lifetime models as a function of doping level [19]. 64 The mobility parameters of InP [17, 19] for the Caughey-Thomas model described in section 2.4.1.1 are given in the following Table 2.6 along with that for InGaAs [18, 19] and GaAsSb. The mobility values for GaAsSb follow from the discussions earlier in the chapter. Table 2.6 Caughey-Thomas Parameters for InP, InGaAs, GaAsSb. Mobility Parameters MU1N.CAUGH (cm2 /v-s) InP [18,19] 300 InGaAs [18,19] 3372 GaAsSb MU2N.CAUGH (cm2 /v-s) 4917 11599 750 MU1P.CAUGH (cm2 /v-s) 20 75 20 MU2P.CAUGH (cm2 /v-s) 151 331 60 ALPHAN.CAUGH 0.0 0.0 0.0 ALPHAP.CAUGH 0.0 0.0 0.0 BETAN.CAUGH -2.3 -2.3 -2.3 BETAP.CAUGH -2.2 -2.2 -2.2 GAMMAN.CAUGH -3.8 -3.8 -3.8 GAMMAP.CAUGH -3.7 -3.7 -3.7 DELTAN.CAUGH 0.46 0.76 1.057 DELTAP.CAUGH 0.96 1.37 0.45244 574 NCRITN.CAUGH(cm-3 ) 6.4e17 8.9e16 1.8190e17 NCRITP.CAUGH(cm-3 ) 7.4e17 1.0e18 7.935e18 65 Table 2.7 Summary of Important Material Parameters for GaAs0.51Sb0.49 InP In0.53Ga0.47 As Parameter Symbol GaAs 0.51 Sb0.49 InP In0.53Ga0.47 As Permittivity ε r/ε0 14.4 [7] 12.35 13.88 Energy Gap Eg(eV) 0.72 [5] 1.344 0.74 Electron affinity χ(eV) 4.09 [7] 4.38 4.58 Electron effective mass Light hole m */m n o 0.04494[7] 0.08 0.041 m */m lh o 0.06632[7] 0.089 0.45 Heavy hole m */m hh o 0.4561 [7] 0.6 0.052 Electron µ n 5400cm2 /Vs 12000cm2 /Vs 200cm2 /Vs 300cm2 /Vs 2.3875x1017 cm-3 [7] 1.1x1019 cm-3 7.7x1018 cm-3 4.6456x1018 cm-3 [7] 5.7x1017 cm-3 2.1x1017 cm-3 Mobility Hole Mobility µp 700cm2 /Vs @3x1017 cm3 [7] 40cm2 /Vs@5x1017 cm-3 [14] Valence band density of states Conduction band density of States Nv Nc 2.7 Conclusion In this chapter, a brief overview of the simulation software was given along with the physics-based models used in the simulation and analysis of the performance of the device. The material properties of GaAsSb were investigated in detail, such as the band energy gap, mobility, effective masses, lifetime and density of states. The bandgap offsets 66 of GaAsSb with InP was discussed with emphasis of its effects on device performance. Finally important properties of InP and InGaAs are summarized. References [1] ATLAS User Manual, Vol. 1, Silvaco International, Santa Clara, CA, Feb’ 2000 [2] W. Liu, Handbook of III-V Heterojunction Bipolar Transistors, John Wiley and Sons Inc., New York, pp 26-27 [3] D.M. Caughey and R.E. Thomas, “Carrier Mobilities in Silicon Empirically Related to Doping and Fie ld”, Proc. IEEE 55, pp.2192-2193 1967 [4] D.J. Roulston, N.D.Arora, and S.G.Chamberlain, “Modeling and Measurement of Minority Carrier Lifetimes Versus Doping in Diffused Layers of n +/- p Silicon Diodes”, IEEE Trans. of Electron Devices, pp 284-291, Feb 1982. [5] C.R. Bolognesi, N. Matine, M.W. Dvorak, X.G. Xu, J.Hu and S.P. Watkins, “ NonBlocking Collector InP/GaAsSb/InP Double Heterojunction Bipolar Transistors with a Staggered Lineup Base-Collector Junction”, IEEE Electron Device Lett., vol 20 pp 155157 , 1999. [6] J. F. Klem, D. Huang, H. Morkoc, Y. E. Ihm, and N. Otsuka, Proc. SPIE 877, 28 1988 [7] A. Ya Vul,”Handbook Series on Semiconductor Parameters” Vol 2, New Jersey, World Scientific, 1999, 111-131. [8] K.P. Roenker, “ECES 750 Semiconductors and Heterojunctions”, University of Cincinnati, Class Notes, Fall 2000. [9] K.G. Merkel, V.M Bright, M.A. Marciniak, C.L.A Cerny, M.O. Manasreh, “Temperature Dependence of the Direct Band Energy Gap and Donor-Acceptor 67 Transition Energies in Be-doped GaAsSb Lattice Matched to InP”, Appl. Phys. Lett., Vol 65, pp 2442-2443, 1994. [10] N. Matine, M.W. Dvorak, S. Lam, C.R. Bolognesi, “Demo nstration of GSMBE Grown InP/GaAsSb/InP DHBTs”, IEEE Indium Phosphide and Related Materials Conference Proceedings, pp 239-242, 2000 [11] M. Peter, N. Herres, F. Fuchc, K. Winkler, K.H. Bachemand J. Wagner, “Band Gaps and Band Offsets in Strained GaAs1-y Sby on InP Grown by Metalorganic Chemical Vapor deposition.” Appl. Phys. Lett., Vol 74. pp. 410-412, 1999 [12] J. Hu, X. G. Xu, J. A. H. Stotz, S. P. Watkins, A. E. Curzon, M. L. W. Thewalt, N. Matine, and C. R. Bolognesi, “Type II Photoluminescence and Conduction Band Offsets of GaAsSb/InGaAs and GaAsSb/InP Heterostructures Grown by Metalorganic Vapor Phase Epitaxy,” Appl. Phys. Lett., vol. 73, pp. 2799–2801, 1998 [13] S.P. Watkins, O.J. Pitts, C. Dale, X.G. Xu, M.W. Dvorak, N.Matine, and C.R. Bolognesi, “ Heavily Carbon-Doped GaAsSb Grown in InP for HBT Applications”, J. Crys. Growth, Vol 221. pp. 59-65, 2000 [14] V. Fink, E. Chevalier, O.J Pitts, M.W. Dvorak, K.L. Kavanagh, C.R. Bolognesi and S.P. Watkins, “Anisotropic Resistivity Correlated with Atomic Ordering in P-type GaAsSb”, J. Crys. Growth, vol 79, pp 2384-2386, 2001 [15] B.T. Mc Dermott, E.R. Gertner, S. Pittman, C.W. Seabury and M.F. Chang, “ Growth and Doping of GaAsSb via Metalorganic Chemical Vapor Deposition for InP Heterojunction Bipolar Transistors”, Applied Physics Letters, vol 68, pp1389-1388, 1996. [16] Goldberg Yu. A. and N.M. Schmidt, “Handbook Series on Semiconductor Parameters”, Vol. 1 World Scientific, London, 1999, pp. 40-97. 68 [17] M.Levinshtein, S. Rumyantsev and M. Shur, Handbook Series on Semiconductor Parameters, Vol.2, World Scientific, London, 1996, pp. 169-190. [18] R.M. Sieg, “A Photoluminescence Study of Heavily Doped Indium Phosphide”, M.S thesis, Ohio State University, 1994. [19] S. Datta, S. Shi, K. P. Roenker, M. M. Cahay, and W. E. Stanchina, "Simulation and Design of InAlAs/InGaAs Pnp Heterojunction Bipolar Transistors," IEEE Trans. on Electron Devices: Vol. 45, No. 8, Aug. 98, pp. 1634 - 1641. 69 Chapter III Simulation of InP/GaAsSb/InP HBTs and basic results 3.1 Introduction In this chapter, we model and analyze the structure and performance of InP/GaAsSb/InP double heterojunction bipolar transistors (DHBTs) using a commercial numerical device simulator ATLAS from Silvaco Inc [1]. In the previous chapter, we described the material parameters and models. Here we briefly explore how these models are incorporated in the simulation software, which is followed by a brief discussion on how the actual simulation is done. Next, we present some basic simulation results for the InP/GaAsSb DHBT. In this chapter, the D.C, small signal AC and microwave performance of a specific transistor structure is evaluated and the results are compared with the published results by Bolognesi et al. [2], thereby establishing the reliability of the simulations. In the following chapter, we report simulation results for the modification and performance optimization of the device structure. 3.2 Software Description The ATLAS software package is a comprehensive and flexible tool for simulating the electronic performance of arbitrary two dimensional device structures, which can be composed of semiconductors, insulators, conductors and terminals. The simulation can be performed to obtain not only DC, small signal AC and microwave analysis of the device terminal behavior, but also analysis of the internal device parameter distributions, such as the conduction and valence band energies, electrostatic potential, electric fields, electron 70 and hole concentrations and the current components. It also includes the interface physics models, which are provided for all topology combinations, such as heterojunctions, metal-semiconductor contacts and surfaces. These capabilities enable accurate simulation of device parasitics, leakage currents and trapped charges. The simulator consists of a device simulation sub framework ATLAS and a modular set of application oriented tools like BLAZE, which is a 2-D device simulator for III-V material devices and devices with position dependent band structure (heterojunctions). The software also includes a Graphic User Interface (GUI) called TONYPLOT for plotting the results. In addition, it has a two-dimensional fabrication process simulator called ATHENA. Simulation of a device for its characteristics involves the following methodology. Initially the device’s structure in two dimensions is described including all the layer compositions, dopings and thickness as well as the location of the device electrical contacts. Physical models are incorporated that specify the dependence of one parameter on the other, such as carrier mobility on doping levels. Appropriate physicsbased models are then selected by including recombination models (Auger, ConSRH), mobility models (Klaassen, Fldmob, Conmob), and bandgap narrowing models (BGN). Then, the bias conditions are specified for performing the AC, DC and microwave simulation. The two dimensional grid of mesh points for the calculations and the appropriate numerical techniques for solving the differential equations can also be specified. The device simulation process can be summarized as follows: 1. The physical HBT structure to be simulated is described, 2. The physical models to be used are specified, and 3. Input bias specification and the desired output device characterization extraction values are set. 71 Each of these are discussed in more detail in the following sections and illustrated for our InP/GaAsSb DHBT. 3.2.1 Definition of HBT structure and Mesh Specification The physical structure of the device to be simulated is specified initially along with the material regions and their properties. This includes specifying the composition, doping and thickness of each region and the size and location of each of the device contacts. We then proceed by specifying a two dimensional mesh of grid points over the device structure. The meshes are defined in the ATLAS run file by using the statement MESH for specifying the grid lines in the x and y direction. Defining of the mesh is a very important part in the simulation procedure of the device, because the mesh grid points are the places where the simulator solves the device equations and calculates the parameter such as the electric potential and carrier concentration. Most of the characteristics of the device are determined by what happens around the junctions and the interfaces in the structure of the device. So it is important that our simulation be accurate in these areas. For our simulation, this is achieved by making the mesh denser around the emitter-base and base-collector junction areas where there is a rapid change in carrier concentration and potentials as shown in Figure 3.1 for our InP/GaAsSb DHBT. However for an excessive number of grid points, the simulator takes a much longer time to arrive at a solution. So a balance must be found between excessive simulation time and too coarse a mesh. Care should also be taken that the mesh is not too dense, because it can overwhelm the simulations, resulting in convergence problems for the equations. The optimum way for defining the mesh is that, it should provide 72 reasonably accurate results, whereas at the same time the simulator should be able to solve the equations within a reasonable time and without any convergence issues. Figure 3.1 An InP/GaAsSb/InP DHBT structure shown with mesh grid 3.2.2 Physics-Based Models As described earlier in Chapter 2, the device characteristics are calculated using physics-based models. The type of models used depends on the nature of the device being simulated. In our case, for the HBTs, the most important models are the field and doping dependent models for the mobility. The models used in the simulation are specified in the MODELS statement of the ATLAS run file. The Caughey-Thomas model for the doping dependency of the mobility is specified by the parameters “analytic” on the Model statement. Similarly, the field dependent mobility model is activated by using the 73 parameter “fldmob”. The carrier recombination models (RSH Model) and the doping dependent carrier lifetime models are activated by specifying “rsh” and “conrsh” respectively on the models statement. All the previously mentioned models except for the concentration dependent carrier recombination model “consrh” have been used in the simulations. These models along with the parameters have been described in Chapter 2; for a more detailed view of the syntax and the structure of the run file please refer to Appendix A. 3.2.2.1 Hot-Electron Model Some of the initial difficulties faced during the simulation runs were that of the low values of fT and fmax for the simulated device when compared with the published experimental results [2]. This we attributed to the low values of carrier mobility in the GaAsSb base. However, the impressive high frequency performance of the device reported in the literature has been attributed to the high velocity with which the electrons are ballistically injected into the collector from the base [2]. This ballistic effect is due to the small device dimensions (base width), high internal fields, and the conduction band alignment at the base-collector heterojunction as shown in Figure 3.2. Because of these properties the electrons injected into the collector have kinetic energies substantially above kT, with a effective electron temperature Te larger than the lattice temperature T. Though the Hot-Electron model incorporated into the simulation software was primarily meant for modeling short channel effects in MOSFETs, we have included hot-electron effects in our modeling of the InP/GaAsSb HBTs, since the basic underlying device physics is the same. 74 Figure 3.2 The Staggered Bandgap line-up of InP/GaAsSb/InP HBTs [3] Typically, hot-electron effects are included in HBTs by designing structures with a wider band gap emitter than the base, as it is in our case. For very narrow base devices, the transport process may not be diffusive since the carriers undergo relatively few collisions in crossing the base. When the electrons are injected into the base over the conduction band spike, they enter with excessive energy and velocity. Hence, the term hot-electron effect is used. Normally incorporating the ballistic injection effects through the hot-electron model reduces the base transit time by replacing the slow diffusive motion by the faster ballistic propagation [3]. Hot-electron effect in InP/InGaAs HBTs has been experimentally verified by Teissier et al. [4]. Electroluminescence spectroscopy was used to directly investigate the transport of electrons through a relatively thicker base (~150nm). Ballistic as well as quasi-ballistic transport was found to dominate the 75 transport process through the base even at room temperature. The ballistic effect was found to become more pronounced as the base thickness was reduced. This observation conforms to the reasoning that in a thinner base, the electrons undergo less collisions, thus retaining their higher transport velocity. In our case, the base thickness used in all the devices reported, is in the range of 200-500 A0 , much smaller than the 1500 A0 base thickness used in the experimental device by Teissier et al. [4] to demonstrate the hotelectron effect. However, for our InP/GaAsSb DHBT there is no conduction band spike, since the band alignment has the cond uction band edge for the GaAsSb base at a higher energy than that for the implementation as shown in Figure 3.2. So there will be no hotelectron injection into the base. For our InP/GaAsSb DHBT, however, the hot-electron effect will modify the transit time across the base-collector space charge region, since the conduction band edge for the base is above that of the collector layer as seen in Figure 3.2. We can conclude that the dominant transport mechanism through the base of the InP/GaAsSb DHBT and the reason for impressive fT and fmax performance in spite of the low carrier mobility in the GaAsSb base is due the hot-electron effect in the basecollector space charge region. The hot-electron model is activated in the simulations by using the parameter “hcte.el” in the models statement. 3.2.3 Input Bias Setup and Output Result Extraction The bias conditions on the terminals of the device must be specified for each of the electrodes. The simulation is initially done with no applied bias, so as to obtain an initial, thermal equilibrium solution. The voltages applied on the terminals are then increased in small steps so that the simulation proceeds without any convergence problems. The DC biases are swept for the collector and the base terminal to obtain the Gummel plot of IB, 76 IC versus Vbe. For AC analysis, after obtaining the initial solution using the DC biases, the AC bias is applied with a specific frequency. The frequency is then increased and the AC analysis repeated. Subsequently, the microwave S-parameters can be calculated and the performance results are extracted. Finally, the frequency of the analysis can be swept and the device performance as a function of the frequency can be examined and the cutoff frequency fT and the maximum frequency of oscillation fmax determined. For detailed instructions on applying bias and parameter extraction the reader is referred to the ATLAS user manual [1]. 3.3 Structure and Modeling of InP/GaAsSb/InP HBTs The structure of the transistor examined in this simulatio n and studied is the one reported by Bolognesi et al. [2]. This group has demonstrated devices with one of the highest reported fT , fmax and breakdown voltages for the InP/GaAsSb/InP heterostructure. The main reason for choosing their work as a bench mark is that they have reported a series of fabricated devices with increasing performance measures. Table 3.1 shows a series of reported results in the development of InP/GaAsSb DHBTs from several groups. Bolognesi et al. [2] have reported partial results for different devices with various base widths, doping, growth methods, and different emitter finger widths from 0.4µm to several µm and different base junction areas. So an exact comparison of our simulations results with their reported results is not possible. However, we will draw upon several experimental results for comparison and validation of our simulation model. We examine here in particular, a recent structure for which they have reported a more comprehensive set of results [2]. 77 Table 3.1 Evolution of InP/GaAsSb/InP DHBT Device Development. AC Gain Base/Collector fT GHz fmax GHz Emitter Size µm2 dB Thickness A0 45 N/A N/A 350-1500/3000 70X70 27 30 45 500/3000 2X20 50* 75 23 400/3000 5X12 >100 78 N/A 400/3000 5X12 39 305 230 200/2000 5X0.5 39 270 >300 250/2000 5X0.5 42 300 300 200/2000 0.4X11 >80 N/A N/A 500/2000 20X20 Group McDermott et al. 1996 [5] Bhatt et al. 1996 [6] Bolognesi et al. 1998 [7] Hu et al. 1999 [8] Dvorak and Bolgonesi et al. 2000 [9] Dvorak and Bolgonesi et al. 2000 [9] Dvorak and Bolognesi et al. 2001 [2] Rajavel et al. 2003 [10] * DC Gain The layer structure for the MOCVD-grown NpN InP/GaAsSb/InP abrupt double heterojunction bipolar transistors (DHBTs) is shown in Figure 3.3. Bolognesi et al. [2] have reported current gains of 40 dB, an fT of 300 GHz and an fmax of 300 GHz. The devices have demonstrated outstanding dynamic performances over a wide range of biases including the saturation mode [2]. The 2000 Å InP collector provides good breakdown voltages of BVCEO = 6V; thinner (≈1500 Å) collectors allow operation at still higher currents with fT = 200 GHz at a collector current density JC of 650 kA/cm2 . 78 Figure 3.3 Schematic of the simulated InP/GaAsSb/InP DHBT structure [2] The device consists of a heavily doped, 3000 Å thick, InP substrate, an n+ (1 x 1019 /cm-3 ) 500 Å In0.53 Ga0.47As sub-collector, a 2000 Å n- InP collector (S: 3 x 1016 cm3 ), a 200 (or 250) Å thick GaAs0.51Sb0.49 p+ base (C: 8 x 1019 cm-3 ), a 700 Å n InP emitter (S: 3 x 1017 cm-3 ) and a heavily doped 500 Å InP/1000 Å InGaAs emitter cap as shown in Figure 3.3 [2]. An important note here is that Bolognesi et al. [2] reported results for two different structures, one with InGaAs layers for the sub-collector and the emitter cap and the other structure with InP layers used throughout except for the base. It is not clear what specific materials they used for the published results. For the purpose of simulation we used InGaAs layers as shown in Figure 3.3. The fabricated device was reported to have a double mesa structure with self-aligned base contacts and an emitter area of 0.4 x 11µm2 as shown in Figure 3.4. For the purpose of simulation, Silvaco assumes a default stripe length of 1µm. During extraction of parameters from the simulations, the current values are properly scaled for comparison with published results. The heavily doped InGaAs layers have been used for the SubCollector and the emitter contact layers to reduce contact resistance. 79 Figure 3.4 Simulated structure of InP/GaAsSb/InP HBT [after Bolognesi et al [2]]. The device contacts have been modeled as ohmic and with a base contact resistivity of 2x10-7 Ωcm2 . This resistivity has been reported by Bolognesi et al [2] and has been used in the simulations. The emitter and collector contact resitivities have not been reported by the group and hence default resistivity values of 5x10-8 Ωcm2 have been used in the simulations. The metal contacts have been placed 100 nm from the edges of the mesas. The InGaAs Sub-collector region actually sits on a InP substrate (not shown here), which was neglected for the purpose of simulation. Only half of the structure is taken into account for simulation, because of the symmetry of the whole structure and the fact that simulating the structure would take twice as much processing time. Due to the lack of 80 data on lateral contact dimensions, they have been chosen to conform to state of the art values with 0.5µm emitter and base widths and 2µm wide collector contacts as shown in Figure 3.4 [11]. It should be noted that the emitter width is slightly bigger than 0.4 µm length used by Bolognesi et al. [2] in their device structure. However, as mentioned earlier, current values are scaled to account for the difference in emitter widths. The material parameters used in the simulation of the device are repeated from Chapter 2 for convenience in Table 3.2. Table 3.2 Summary of Important Material Parameters for GaAs0.51Sb0.49 InP In0.53Ga0.47 As Parameter Symbol GaAs 0.51 Sb0.49 InP In0.53Ga0.47 As Permittivity ε r/ε0 14.4 12.35 13.88 Energy Gap Eg(eV) 0.72 1.344 0.74 Electron affinity χ(eV) 4.19 4.38 4.58 Electron effective mass Light hole m */m n o 0.04494 0.08 0.041 m */m lh o 0.06632 0.089 0.45 Heavy hole m */m hh o 0.4561 0.6 0.052 Electron µ n 700cm2 /Vs @3x1017 cm-3 5400cm2 /Vs 12000cm2 /Vs 40cm2 /Vs@5x1017 cm-3 200cm2 /Vs 300cm2 /Vs 2.3875x1017 cm-3 1.1x1019 cm-3 7.7x1018 cm-3 4.6456x1018 cm-3 5.7x1017 cm-3 2.1x1017 cm-3 Mobility Hole Mobility Valence band density of states Conduction band density of States µp Nv Nc 81 The unique bandgap line- up in the InP/GaAsSb DHBTs is exploited to achieve high performance without grading or using different doping profiles to remove the conduction band barrier at the collector junction seen in InP/InGaAs DHBTs. The bandgap profile reported by Bolognesi et al. [3] is shown in Fig 3.2 for the InP/GaAsSb DHBT with 400 A0 base and a 3000 A0 collector. Since the conduction band edge for the GaAsSb is above that for InP at the base-collector junction, electrons are injected into the collector with an excess energy. These electrons are called hot electrons, which contribute to faster than normal transport across the base-collector space charge region resulting in ballistic transport. The conduction and the valence band discontinuity are ∆EC = 0.18 eV and ∆EV = 0.78eV, respectively, at the base-collector heterojunction. Using the layer structure given in Figure 3.3 following Bolognesi et al. [2], the InP/GaAsSb/InP DHBT structure with self- aligned base and collector contacts was simulated. The simulated energy band structure for a device with 200 A0 base width and 2000 A0 is shown in Figure 3.5. As can be seen, by comparison with that reported by Bolognesi et al. [3] (see Fig 3.2), there is no conduction band spike at the base-collector interface and there is a large valence band discontinuity at the emitter-base heterojunction which effectively blocks hole back injection into the emitter. It is also apparent that the lightly doped 2000A0 n-InP collector is completely depleted for zero VCE. 82 Figure 3.5 Simulated conduction and valence band profiles for InP/GaAsSb/InP DHBT with 200A0 base width and no applied biases. The Gummel plot for the InP/GaAsSb DHBT as reported by Dvorak et al. [12] is shown in Fig 3.6(a). The Gummel plot is a plot of the base and collector currents as a function of Vbe when Vcb=0 V. We consider a slightly different structure by Dvorak et al. [12] for comparison because of the non-availability of Gummel plots for the original structure by Bolognesi et al. [2]. The primary difference in the structure is the larger emitter area of 1.5X24 µm2 and the 400A0 GaAsSb base (C: 4x 1019 cm-3 ) of Dvorak et al. [12]. For simulation, Silvaco assumes a emitter strip length of 1µm and hence the simulated emitter area for our device is 0.5X1 µm2 . The simulated Gummel plot is 83 shown in Fig 3.6(b). The maximum collector current for the simulated structure was of the order of a few milli-amperes, whereas for the fabricated structure it was of the order of few tens of milli-amperes, which is expected due to the differences in emitter areas. The trends of the curves are similar in these two plots. The maximum collector current is reached for a base voltage of 0.8 Volts in both the cases. For our simulation results, the base ideality factor ηb is 1.01 and the collector ideality factor ηc is 1.115 for a forward voltage of Vbe=0.5V. This compares favorably with the published results for the fabricated structure [12], which had ideality factors of ηb=1.09 and ηc=1.01. The ideality factor is a measure of which current component dominates the total forward current. When the ideality factor is closer to 1, it means that the diffusion current dominates, whereas when the recombination current dominates, η=2. Usually the forward current consists of both the currents. For InP/GaAsSb DHBTs we see that the ideality is closer to 1, meaning that the diffusion current dominates in the base and reflecting the absence of a conduction band discontinuity at the base-emitter and basecollector heterojunctions. Recombination current usually dominates at smaller forward biases (and current levels). At large biases (as 0.8 V), parasitic resistance and high current effects causes the currents to tend to saturate. 84 Figure 3.6(a) Gummel-Poon plot of the InP/GaAsSb DHBT structure from Dvorak et al for 400 A0 base width (C: 4 x 10-19 cm-3 ) and 1.5x24 µm2 [12] Figure 3.6(b) Simulated Gummel-Poon characteristics of InP/GaAsSb/InP structure for the device structure described in Figures 3.4 and 3.3 for Vcb=2.0 V 85 The common emitter characteristics reported for a 250 A0 base, 2000 A0 collector InP/GaAsSb DHBT [2] are shown Figure 3.7(a). The transistor has common emitter current gain of β=40-50dB, with breakdown voltages BV CEO > 6V, which can attributed to the low electron multiplication co-efficient in the InP collector. Transistors fabricated with 200 A0 base have similar characteristics, but with a higher current gain of around 80 dB. The simulated common emitter characteristics for the structure in Figure 3.3 are shown in Figure 3.7(b). The simulations have been done in steps of 5µA base current, considering the fact that the simulated device has a emitter stripe length of 1µm compared to the 11µm used in the fabricated device [2]. The non- linearity of base current in the saturation region of the Gummel plot, makes its difficult to simulated with 5µA steps, hence at higher Ib bias we resort to a stepping of 10µA. The reported current of 490 kA/cm2 [2] could be reached by operating the device at higher base current bias, however that would mean the onset of high current effects in the simulated device, considering its smaller emitter stripe and the thinner 200A0 base. The calculated current gain is around 34 dB for the simulated device. This compares reasonably well with the 40-50 dB Bolognesi et al. [2] reported for the 250A0 device, but not so for the device with a 200A0 thick base which has a 80 dB gain. The characteristics display a low collector offset voltage which indicates that no blocking can take place at the base-collector junction, as also evident from the device band diagram in Figure 3.5. Since, self- heating effects have not been used in the simulation model, the transistor breakdown cannot be observed in the characteristics. 86 Figure 3.7 (a) Typical common-emitter characteristics for a small area 0.4x11µm2 emitter device with a 250 Å base and a 2000 Å InP collector [2]. Figure 3.7(b) Simulated common-emitter characteristics for a 0.5x1 µm emitter device (Fig 3.3) with a 200 A0 base and 2000 A0 InP collector 87 The frequency response of the current and power gain of the fabricated structure reported by Bolognesi et al. [2] are shown in Figure 3.8(a) for a device with a 200 Å thick base and 0.4x11µm2 emitter. The transistor shows an fT =300GHz and fmax=300 GHz obtained by extrapolating the current gain and Mason’s unilateral power gain U assuming a -20dB/oct roll-off, with the peak AC current gain being 42 dB. The simulated performance of the structure is shown in Figure 3.8(b). For comparison, the simulated structure had a maximum fT of 344 GHz and the peak current gain was 34 dB, similar to Bolognesi’s results. The discrepancy in gain might be due to the fact that, the DC bias point for the measurement was not specified in [2] and hence the characterization could have been done at different bias points. The simulated structure was biased at VBE=0.8V and VCE=1.8V. Also, it should be noted that fT and fmax for the fabricated devices were extrapolated and frequencies from beyond 30GHz were not actually measured by Bolognesi et al [2]. Also shown in Figure 3.8(b) is the simulated power gain versus frequency from which we obtain an fmax of 295 GHz. This compares with the fmax of 300 GHz from the results by Bolognesi et al [2] for the same device, so there is good agreement. 88 Figure 3.8(a) Frequency response of fabricated structure reported by Bolognesi et al. with a 200 A0 thick base with a doping of 8x1019 /cm3 [2] Figure 3.8(b) Frequency response of the simulated InP/GaAsSb/InP HBTs (•) current gain, (X) unilateral power gain. 89 Figure 3.9(a) shows the variation of fT with DC biases (JC ) for various collectoremitter voltages VCE, as reported by Bolognesi et al. [2] for the fabricated device. A peak fT of 300 GHz is obtained at JC=410 KA/cm2 for VCE = 1.8 V [2]. The cut-off frequency goes through a maximum as function of JC because the base collector charging time τc has a inverse dependence on the collector current. As a result, a transistor designed for high frequency performance has to operate at high, but not necessarily excessive collector current levels. At excessively large JC, the device performance degrades due to high current effects (discussed below). Note that the peak fT increases with increasing VCE since a larger VCE delays the fall-off at high current densities. Figure 3.9(b) shows the evolution of fT with JC obtained for the similar, simulated device as described in Figure 3.3, the primary difference being the use of a 0.5µm emitter contact, while the fabricated structure had a 0.4µm emitter contact. The current density was scaled using the emitter area of the simulated device to enable comparison with the published results. While the simulated results overestimated the cut-off frequency at lower current density, they realistically described the peak fT and its roll-off at high JC. Similar to Figure 3.9(a), the peak fT increases with VCE and moves to a higher JC. A peak fT of 344 GHz was achieved at a collector current density of 432 KA/cm2 for VCE = 1.8 V. It can be seen that even at low collector-emitter bias of 0.6 Volts, we are able to achieve impressive high frequency performance of fT greater than 100 GHz. This low VCE case corresponds to operating the transistor in saturation mode. 90 Figure 3.9(a) Evolution of fT with JC for the InP/GaAsSb DHBT with a base width of 200A0 and collector width of 2000 A0 for a VCE=0.4V to 1.8V as reported by Bolognesi et al. [2]. Figure 3.9(b) Evolution of fT with JC for collector-emitter bias VCE = 0.4 to 1.8V for simulated structure. 91 Figure 3.9 (c) Peak fT variation with VCE for simulated (•) and fabricated (s) InP/GaAsSb/InP DHBTs. Figure 3.9 (c) shows the plot of the peak fT as a function of JC versus VCE for both the simulated and the fabricated structures. We can see that for both sets of results the fT increases for increasing values of VCE and there is reasonable agreement between the two sets of results. The excellent dynamic performance in saturation can be attributed to the confinement of holes in the base because of the large valence band discontinuity ∆EV = 0.7–0.8 eV at the InP/GaAsSb interface at the base-collector junction, which prevents hole injection from the P-type GaAsSb base into the collector when the base/collector junction is forward biased [2]. 92 The fmax characterization for the fabricated structure by Dvorak and Bolognesi et al. [9] is shown in Figure 3.10(a) as a function of JC and VCE. This device has the same structure for which the fT characterization was done earlier except for the 250 A0 base was used [9]. The reported fmax reaches a peak of 300 GHz for a collector current density of 250 KA/cm2 and VCE = 1.8 V. Figure 3.10(b) shows the variation of fmax with JC for different base emitter voltages for the simulated structure with 200A0 base width. The simulated fmax reaches peak value of 295 GHz at JC = 313 KA/cm2 for a collector bias of 1.8 Volts. The fall-off in fmax is related to that seen for fT and is predictable due to the high current effects. The responses of peak fmax as a function of JC to different collector bias VCE for the fabricated and the simulated structure are illustrated in Figure 3.10 (c). The simulated structure reaches a peak of 295 GHz for VCE = 1.8V, whereas, the fabricated structure has a peak performance of 300 GHz at the same VCE. These simulation results are in excellent agreement with the published results [2]. The increase in fmax at higher VCE biases is normal, keeping with the fact that the collector-base capacitance decreases with increasing VCE and fmax’s inverse dependence on Cjc. 93 Figure 3.10(a) fmax and fT dependence on JC for a 250 A0 base and 2000 A0 InP/GaAsSb DHBT for various VCE biases [9] Figure 3.10(b) Evolution of fmax vs collector current density JC for 200A0 bases and 2000A0 collector for the simulated InP/GaAsSb DHBT for various VCE. 94 Figure 3.10 (c) Reported fmax variation with collector-emitter bias VCE (•) and simulated peak fmax values (? ) for a 4x11 µm2 emitter, 250 A0 base and 2000 A0 Collector [9] 3.3 High Current Effects Although no classical base push out can occur in the InP/GaAsSb system due to the hole confinement in the base region by the large valence band discontinuity ∆EV at the InP/GaAsSb heterojunction, the onset of fT and fmax roll-offs seen in Fig 3.9 and Fig. 3.10 at very high JC are due to the band flattening in the collector [2]. However, this can be offset to certain extent by operating the device at a highe r collector-base VCB and VCE voltages. We can understand this trend by briefly analyzing the various transit time components that contribute to the cut-off frequency fT . The cutoff frequency of the transistor is related to the collector-emitter transit time τec by the equation [13] 95 fT = 1 ,τ ec = τ e + τb + τsc + τc 2πτ ec (3.1) where, τec is the total HBT input response delay time, τe is the emitter charging time, τb is the base transit time, τsc is the base-collector space charge transit time, and τc is the collector charging time. The space-charge transit time, τsc, is the transit time for the carriers to drift through the depletion region of the base-collector junction, quantitatively given as in τ sc = X dep (3.2) 2vsat where Xdep is the base-collector depletion width where this assumes electron velocity saturation. The emitter charging time τe is the time required to change the base potential by charging up the capacitances through the differential base-emitter junction resistance, τe = C je (3.3) IC where Cje is the emitter junction capacitance. The collector charging time τc is the time required to charge up the base-collector junction capacitance, is given by, τc = C jc (3.4) IC where Cjc is the collector junction capacitance. For JC corresponding to peak fT , the (τe + τc) terms are small, so the τec term can be simplified as τec ≈ τb + τ sc. Now the dependence of peak fT on VCE varies by the factor that dominates the remaining terms in the emittercollector transit time. The improvement in fT could be due to several things 1) τb may go down somewhat since the larger Xdep will make the neutral base region slightly smaller. 96 2) If the space charge component τsc dominates the collector current charging τc, then we would see a fall-off in fT with increasing collector-emitter voltage. Since this is not the case, we could assume that the decrease in collector capacitance reduces collector current charging time significantly to offset any increase in τsc. This assumes that the collector region is fully depleted. The conventional understanding is that peak fT is directly proportional to VCE. An increase in VCE results in an increase in the basecollector depth Xdep, which reduces the collector junction capacitance, leading to an increase in fT . For our device, the collector is fully depleted even at zero bias because of the collector region is narrow (2000A0 ), so this is not the explanation. The origin of fT increase with VCE is likely due to an increase in the average electron velocity in the B-C space charge region. This may also be related to the hot-electron effect described above. 3) Figure 3.11 from Bolognesi et al. [14] shows that CBE is nearly constant until the onset of high current effects at JC ≈ 3µA/(µm)2 corresponding to the occurrence of peak fT . 97 Figure 3.11 High frequency and Cbe behavior of a 200 A0 base and 2000 A0 InP/GaAsSb/InP device [16]. 3.4 Current Gain Collapse at High Frequency Figures 3.12 and 3.13 show the variation of the AC current gain and unilateral power gain, respectively, with collector current density. The gain values measured were the maximum gains measured for VCE=1.8V at 2 GHz. The characterization was done at 2 GHz because this frequency is the upper limit of the band used in consumer applications like mobile phones and automotive radars. The power gain reaches a peak of 34 dB at a current density of 0.3 KA/cm2 before sharply degrading above 100 kA/cm2 . Similarly, the AC current gain has a value of 34 dB that remains constant over several orders of magnitude before falling off above 100 KA/cm2 . 98 Figure 3.12 Variation of current gain with collector current density for VCE=1.8 V at 2 GHz Figure 3.13 Variation of maximum unilateral power gain with collector current density for VCE=1.8 V at 2 GHz 99 The high current degradation in the type-II InP/GaAsSb DHBTs at high current densities is not due to the conventional base push-out effect [15]. Base push-out does not occur at the high current density because hole injection from the base into the collector is blocked by the high valence band discontinuity (~0.78eV). As a consequence, a parasitic barrier forms in the conduction band at the base-collector heterojunction (16, 17) that impedes electron flow causing the collector current to saturate and the device gain and speed to degrade. In brief, the origin of fall in current gain and the fT roll-off can be described as follows: 1) The onset of fT roll-off at high collector current density is attributed to the collapse of the electric field at the base-collector heterojunction as shown in Figure 3.14. The ‘critical current density’ at which the electric field collapses (and also at which fT reaches its peak value) is given by [15] 2ε s (Vbc +Vbi ) J Crit = vs qNC + WC2 (3.5) where NC is the collector doping level, WC is the collector thickness, vs is the average electron velocity through the collector, Vbc is the base collector reverse bias, Vbi is the collector junction built- in potential, q the electron charge and ε s is the permittivity of the material. This equation shows that the current density which collapses the electric field at the base-collector junction is linearly dependent on the base collector voltage Vbc. Figure 3.15 shows the dependence of the critical current density JC on Vbc for the simulated device. The plot is reasonably described by the linear-like relation of 3.5, at least for lower values of Vbe, suggesting that the onset of fT roll-off is associated with the collapse of electric field. This has been experimentally confirmed by Bolognesi et al. [14, 15]. 100 Complicating this phenomenon is the fact that because of the hole confinement in the base due to the larger ∆EV at the base-collector junction and the excess electron build up in the collector, at high current densities a charge dipole is induced which reverses the electric field at the base-collector heterojunction which is seen in Figure 3.14. Figure 3.14 Collapse of electric field and subsequent parasitic barrier formation in the collector of InP/GaAsSb/InP DHBTs with Vbe varied from 0.6 V to 1.0 V and VCE = 1.8V. 101 Figure 3.15 Critical current density for the onset of high current induced device degradation versus based-collector voltage for the simulated and calculated device behavior [15]. 2) The hole and electron build up in the base-collector heterojunction is shown on Figure 3.16 (a) and (b). This hole build up in the base causes the formation of a parasitic barrier in the collector region of the InP/GaAsSb DHBT as shown in Figure 3.14 at higher current densities after the electric field collapses. This barrier greatly reduces the exit velocity of electrons leaving the base and leads to a drop in fT (15, 16, 17). The parasitic barrier height rises with the current density as 133mV at JC = 550 kA/cm2 (see Fig. 3.17). This barrier leads to electron charge storage in the base (see Fig 3.16(b)), resulting in the reduction of the base exit velocity by a factor of ~ exp(-EB /K BT). These in turn leads to a sharp increase in CBE as shown in Figure 3.13 and a drop in fT . The measure of the parasitic barrier height in the conduction band for various values of JC is plotted in Figure 3.17. The barrier does not occur at emitter-base biases Vbe up to 0.9V. 102 The effect can be partially offset by using higher doping levels in the collector [15]. However, the gains achieved by doping the collector higher are offset by considerations of lower breakdown voltage at the junction and speed-related effects of the collector-base capacitance. Although Bolognesi et al [15] explain that the high current degradation is partly due to “Kirk- like effects”, in reality there is hole accumulation in the collector as can be observed in Figure 3.16 (a), which is characteristic of the classical base-pushout or Kirk effect. Therefore, the phenomenon here causing fall-off in the device performance is a combination of parasitic barrier formation and the Kirk effects. Figure 3.16(a) Hole concentration for low and high current operation of a 200 A0 base and 2000 A0 collector simulated InP/GaAsSb/InP DHBT, ‘B’ and ‘C’ indicate the base-collector regions for VCE = 1.8V and for VBE from 0.6V to 1.1V. 103 Figure 3.16(b) Electron (b) concentration for low and high current operation of a 200 A0 base and 2000 A0 collector simulated InP/GaAsSb/InP DHBT, ‘B’ and ‘C’ indicate the base-collector regions, for VCE = 1.8V and for VBE from 0.6V to 1.1V. Figure 3.17 Conduction band barrier height versus collector current density for simulated InP/GaAsSb structure given in Figure 3.3 for VCE =1.8 V. 104 3.5 Conclusion In this chapter we have demonstrated reasonable agreement in performance of the simulation results with the reported experimental results on the fabricated devices from Bolognesi et al. [2, 3] for InP/GaAsSb DHBTs. The accuracy of the simulations depends in part on the precision of the material properties adopted during the simulation. The high frequency performance of the device matched closely the published results with the simulation results being fT =344 GHz and fMAX=295GHz, with the published experimental values of 300 GHz for fT and fMAX, respectively [2]. The variation of the simulated fT with the collector current density agrees very well with the published values showing a fall-off above 100 kA/cm2 . The current density for the peak fT was found to be 410 KA/cm2 for the device reported by Bolognesi et al [2,3], whereas in our case it was 432 KA/cm2 . The variation of the current gain and unilateral power gain with collector current density were also along expected lines. The peak fT and fmax as a function of JC were examined versus VCE and found to be in good agreement with the results reported for the experimental devices. The reason for the fall off in device performance at high currents was shown to be related to the formation of the parasitic barrier in the conduction band. Therefore, the high current degradation is not due solely to the conventional Kirk (base pushout) effect, but also due to the formation of a parasitic barrier in the conduction band of the collector. The parasitic barrier height is a function of the collector current density, collector doping and VCE. Having verified the validity of the model of the InP/GaAsSb DHBT structure being simulated, in the next chapter we investigate the optimization of the performance of the device for high frequency operations. Also, we will briefly examine the physics behind the factors affecting fT and fMAX. In summary, initial comparison of the model’s results 105 with the limited available experimental results shows good agreement so that use of the model is appropriate for investigating epitaxial design for the device. References: [1] ATLAS User Manual, Vol. 1, Silvaco International, Santa Clara, CA, Feb’ 2000. [2] C.R Bolognesi, M.W Dvorak, O.J Pitts & S.J Watkins, “300 GHz InP/GaAsSb/InP Double HBTs with High Current Capability and BVCEO 6 V”, IEEE Electron Device Letters, vol. 22, pp 361-364, 2001. [3] C.R. Bolognesi, N. Matine, M.W. Dvorak, X.G. Xu, J.Hu and S.P. Watkins, “ NonBlocking Collector InP/GaAsSb/InP Double Heterojunction Bipolar Transistors with a Staggered Lineup Base-Collector Junction”, IEEE Electron Device Lett., vol 20 pp 155157 , 1999. [4] R. Teissier, J.L. Pelouard, F. Mollot, “Direct Measurement of Ballistic Electron Distribution and Relaxation Length in InP-based Heterojunction Bipolar Transistors using Ele ctroluminescence Spectroscopy”, Appl. Phys. Lett., Vol. 72, pp. 2730-2732, 1998. [5] B. T. McDermott, E. R. Gertner, S. Pittman, C. W. Seabury, and M. F. Chang, “Growth and Doping of GaAsSb via Metalorganic Chemical Vapor Deposition for InP Heterojunction Bipolar Transistors,” Appl. Phys. Lett., vol. 68, pp. 1386–1388, 1996. [6] R. Bhat, W-P. Hong, C. Caneau, M. A. Koza, C-K. Nguyen, and S. Goswami, “InP/GaAsSb/InP and InP/GaAsSb/InGaAsP Double Heterojunction Bipolar Transistors with a Carbon-doped Base Grown by Organo- metallic Chemical Vapor Deposition”, Appl. Phys. Lett., vol 68, No (7), pp 985-987,1996. 106 [7] C.R. Bolognesi, N. Matine, M.W. Dvorak, X.G. Xu, J.Hu, S.P. Watkins, and M.L. W Thewalt, “Nearly Ideal InP/GaAsSb/InP Double Heterojunction Transistors with Ballistically Launched Collector Electrons”, Electronics Letters, vol. 34, no. 17, pp. 1700-1702, 1998. [8] X. G. Xu, J. Hu, and S. P. Watkins N. Matine, M. W. Dvorak, and C. R. Bolognesi “Metalorganic Vapor Phase Epitaxy of High-quality GaAs0.5Sb0.5 and its Application to Heterostructure Bipolar Transistors”, Appl.Phys.Lett., vol. 74, pp. 776-778, 1999. [9] M.W Dvorak, O.J. Pitts, S.P Watkins, C.R Bolognesi, “Abrupt Junction InP/GaAsSb/InP HBTs with FT as high as 250 GHz and BVceo >6 V”, IEDM Tech Digest, pp 178-181, 2000. [10] R.D. Rajavel, T. Hussain, M.C. Montes, M.W. Sawins, S. Thomas III, and D.H. Chow, “Molecular Beam Epitaxial Growth and Characterization of InP/GaAsSb/InP Double Heterojunction Bipolar Transistors”, J. of Crystal Growth, vol. 251, pp. 848–851, 2003. [11] J. W. Rodwell, M. Urteaga, T. Mathew, D. Scott, D. Mensa, Q. Lee, J. Guthrie, Y. Etser, Suzanne C. Martin, R. P. Smith, S. Jaganathan, S. Krishnan, S.I. Long, R. Pullela, B.Agarwal, U.Bhattacharya, L. Samoska, and M. Dahlstrom, “Submicron Scaling of HBTs”, IEEE Trans. On Electron Devices, vol.48, pp 2606-2624, 2001. [12] M .W Dvorak, T. Matine, C.R Bolognesi, “Design and Performance of InP/GaAsSb/InP Double Heterojunction Bipolar Transistors” Journal of Vacuum Science Technology A, vol. 18, pp 761-764, 2000. [13] Liu, W.,“Handbook of III-V Heterojunction Bipolar Transistors”, John Wiley and Sons Inc., New York, pp 26-27, 1998. 107 [14] C.R Bolognesi, M.W Dvorak, S.J Watkins, “InP/GaAsSb/InP Heterojunction Bipolar Transistors”, IEDM Tech. Digest, pp. 343-351, 2002. [15] C.R Bolognesi, M.W Dvorak, O.J Pitts, S.J Watkins, T.W Macelwee, “Investigation of High Current Effects in Staggered Line- up InP/GaAsSb/InP Heterostructure Bipolar Transistors”, IEDM Tech. Digest, pp 341-350, 2000. [16] S. Tiwari and D. J. Frank, “Analysis of the Operation of GaAlAs/GaAs HBT’s,” IEEE Trans. Electron Devices, vol. 36, pp. 2105–2121, 1989. [17] S. Tiwari, “A New Effect at High Currents in Heterostructure Bipolar Transistor”, IEEE Electron Device Letters, vol. 9, pp. 142-144, 1988. 108 Chapter IV Optimization of InP/GaAsSb/InP HBTs The basic modeling and simulation of the InP/GaAsSb/InP DHBTs was discussed in the previous chapter. In this chapter, we build on the work done in earlier chapters and optimize the InP/GaAsSb DHBT structure in terms of doping and epitaxial layer thickness for better high frequency performance and current gain. The optimization of the two parameters i.e., doping and layer thickness, is a challenging one, since obtaining a better performance on one set of parameters, i.e. fT or fmax, doesn’t necessarily mean optimum performance in terms of the other. We arrive at the best epitaxial design for maximizing fT , fmax, power gain, and current gain by optimizing the conflicting effects of varying the doping concentration and epitaxial layer widths. In this chapter, we investigate the high frequency performance of the InP/GaAsSb/InP DHBTs and briefly explain the various factors that affect fT and fmax. Device structures with different epitaxial thickness and doping are simulated to arrive at the optimum device for high frequency performance. The low and high current performances of the devices are examined as well as the effects of the operating frequency. In addition, the effects of the velocity overshoot on high frequency performance and the incorporation of it into the device model are elucidated. Finally, we simulate the optimized device structure and present performance measures with improved performance over the latest published results. 109 4.1 HBT Performance Parameters In this section, we briefly summarize the parameters that characterize the high frequency performance of the HBTs such as the cutoff frequency fT and maximum oscillation frequency fmax and describe their relationship to the device’s structure. The various factors that affect these parameters and the underlying physics are discussed. We are interested in the above parameters because fT determines the switching speed of digital circuits, whereas fmax is an important figure of merit in transistors designed for power applications. Peak cut-off frequency is an important benchmark when HBTs are used to design high-speed digital circuits. HBTs operating at high frequencies usually operate at a high current density to obtain sufficient current gain and fT . However, for low noise applications, such as LNAs in a wireless transceiver, lower current performance of HBTs without excessively low gain and fT is critical. This is because the noise minima in a HBT occur at low collector current densities. So it is essential that the HBTs have sufficient gain and fT at low as well as high current densities. 4.1.1 AC Current Gain The small signal current β ac is defined as the small signal collector current divided by the small-signal base current, as given by (4.1). The small signals are nothing but AC signals applied as input and measured at the output whose magnitude is very small when compared to the DC operation point. The β ac is given as follows [1], βac = ic βdc = ib 1 + jωβ dc (Cπ + C jc ) / gm 110 (4.1) where ic is the small-signal collector current, ib the corresponding base current, β dc is the DC gain of the transistor when no small signals are applied, Cp is the total input capacitance, Cjc is the base-collector junction capacitance, gm is the transconductance of the transistor and ω is the angular frequency of device operation. Clearly at high enough frequencies, the AC current gain degrades due to the device’s internal capacitances. 4.1.2 Current Gain Cutoff Frequency (f T) The cutoff frequency fT is defined as the frequency at which the AC current gain ßac becomes unity. Also, the fT can be defined as the magnitude at which the forward current gain h-parameter h21 reaches unity. However, it should be noted that the former definition is tied to the applied collector-base voltage Vcb and more realistically represents the performance of the transistor under different bias conditions. This definition is used in the present work to extract the value of fT from the simulation runs. The cutoff frequency of the transistor is related to the collector-emitter transit time by the equation [1] fT = 1 ,τ ec = τ e + τb + τsc + τc 2πτ ec (4.2) Where, τec is the total HBT input response delay time, τe is the emitter charging time, τb is the base transit time, τsc is the space charge transit time, and τc is the collector charging time. We will briefly discuss each of these transit times, and the weight of their influence on the overall cutoff frequency. These transit times are affected by the various capacitances and resistances of the transistor as shown in the equivalent small signal circuit in Fig 4.1. [1] 111 Figure 4.1 Parameters affecting the high frequency performance of the HBT [1] The emitter charging time τe is the time required to change the base potential by charging up the capacitances through the differential base-emitter junction resistance, τe = ηkT .(C je + C jc ) qI C (4.3) where, Cje and Cjc denote the junction capacitances for the base-emitter and basecollector junctions, respectively, η is the ideality factor for the collector current, IC is the D.C collector current, and kT/q is the voltage equivalent of the temperature. We can see from the above equation that the emitter charging time has an inverse dependence on the collector current, as result of which the HBTs designed for high frequency operate at high current levels. Also, the emitter charging time scales directly with capacitances Cje and Cjc, meaning smaller device structures result in higher fT . These capacitances Cje and Cjc also are related to the emitter and collector doping respectively, varying as the square root of doping. In 4.1 Cje is contained in the total input capacitance Cπ and is usually lower than Cjc because the collector junction area is usually larger than the emitter junction area. 112 The base transit time τb is defined as the time required to discharge the excess minority carriers in the base in the form of the collector current and is given as [1], τb = WB2 ηb .DnB (4.4) where WB is the width of the base region, DnB is the electron diffusion constant in the base, ηb is a dimensionless parameter whose value is approximately 2, which depends on the magnitude of the base quasi-electric field for compositional or dopant grading. This transit time is important for PNP HBTs, since the low minority diffusion coefficient of holes results in a significant base transit time. For the NPN HBTs, the base transit time may be negligible when compared to the overall transit time. However, it becomes significant when optimizing fT for high frequency operation since the peak fT occurs when τec≈ τb+ τbc. The space-charge transit time, τsc, is the transit time for the carriers to drift through the depletion region of the base-collector junction, and is quantitatively given as [1] τ sc = X dep (4.5) 2Vsat Where Xdep is the thickness of the base-collector depletion region and vsat is the saturation velocity of the electrons. When analyzing the transit times in a transistor, τsc is of particular importance because it is frequently the largest contributor to the overall transit time at peak fT . One of the reasons is that the collector is often the thickest part of the transistor in order to maintain a reasonable breakdown voltage. Various schemes exist to reduce τsc such as base-collector grading and band gap engineering. All these schemes 113 serve to keep the electric field below the critical level, so the electrons traveling through the space-charge region exceed the saturation velocity vsat through a phenomenon known as the velocity overshoot. In our case the electron velocity is established by the unique band gap line up of the GaAsSb base and the InP collector, whereby the electron is ballistically launched with a high energy into the collector. Therefore τsc can be significantly reduced. Finally, the collector charging time τc is the time required to charge up the basecollector junction capacitance, and is given by [1], τ c = ( RE + RC )C jc (4.6) where RE and RC are the emitter and collector series resistances, respectively. If the HBT is not designed properly, these resistances can significantly increase τc causing degradation in device performance. The overall transit time can be summed up as follows τ ec = X ηkT WB2 .(C je + C jc ) + + dep + ( RE + RC )C jc qI C ηb .DnB 2vsat (4.7) We will examine in detail the effect of the various parameters in (4.7) on fT and fmax while optimizing the different regions of the device structure. 4.1.3 Unilateral Power Gain and Maximum Frequency of Operation (fmax ) A transistor whose output is completely isolated from its input is said to be unilateral and the process by which a lossless feedback network is added to the transistor such that the overall two-port is unilateral is called unilateralization [1]. The unilateral 114 power gain of the transistor is the power gain after the transistor plus the lossless network is made unilateral [1]. The maximum oscillation frequency is the frequency at which the unilateral power gain of the transistor rolls-off to unity and it depends on the value of fT as [1] fmax = fT 8π rbC jc (4.8) where rb is the base series resistance, Cjc is the collector junction capacitance and fT is the cut-off frequency. So fmax rises with fT , but may have a different point of optimization for a devices structure, for example base doping, than fmax. 4.2 Optimization of InP/GaAsSb/InP HBTs The semiconductor industry has been able to make impressive advances in technology, because it has been able to continuously improve the performance of its devices. Faster and more efficient devices owe their existence to improvement in process technology that has facilitated the fabrication of devices with ever-smaller dimensions. For an III-V device to be more useful, maximizing its high frequency performance is a key factor. With this perspective in mind, we proceed to optimize the device structure, scaling the device’s dimension in the vertical and horizontal directions to maximize both fT and fmax. With these primary objectives in sight, the device’s performance in terms of the current and power gain as a function of the current density are also important. This is particularly true for the kind of applications for which InP DHBTs will be used, i.e. in LNAs and power amplifiers. 115 4.2.1 Optimization of the Base Layer The base region of the transistor is the most important layer for the DHBT, in the context that significant improvements in high frequency performance can be made by efficient design of the base region. The fT of a transistor is affected by the transistor base layer design mainly through the base transit time as given in (4.4). Considering this quantitatively, τb is proportional to the square of WB. Therefore, maximization of fT calls for reducing the base thickness towards zero. For zero base thickness, there is no base region and contact, and so no bipolar transistor can be formed. As a result, from the fT consideration alone, the goal is to make the base as thin as possible, keeping within the epitaxial processing capabilities. However, decreasing the base thickness increases the base resistance rb. This reduces fmax because of its direct dependence on the reciprocal of the base resistance as given in (4.8). Therefore, optimization of the base layer involves a compromise between maximizing fmax and fT . A balanced trade-off needs to be made between the two parameters when deciding on the value of the base width. Keeping all other design parameters constant as given in Table 3.3 in chapter 3, the base width of the InP/GaAsSb/InP transistor was varied to study the effect on fT and fmax. Figure 4.2(a) and 4.2(b) show the performance of the transistor in terms of the cut-off frequency as a function of Vbe and JC, as the base width is varied from 10nm to 60nm for a fixed doping of 4x1019 /cm3 . As we can see, the variation in fT is along expected lines, going through a maximum as Vbe and JC increase. The peak fT achieved was 388 GHz at a collector current density of 262 KA/cm2 and Vbe=0.92V for a base width of 30nm. As a point of reference, Bolognesi et al [2] have reported a InP/GaAsSb DHBT with base width as small as 20 nm with a fT of 300 GHz 116 Figure 4.2(a) fT Vs base-emitter voltage as a function of the base width for VCE=1.8V for a fixed base doping of 4x1019 cm-3 Figure 4.2(b) ft vs Collector Current density as a function of base width for VCE=1.8V and VBE from 0.6V to 1.07V for a fixed base doping of 4x1019 cm-3 117 Figure 4.2(c) and 4.2(d) shows the behavior of the fmax versus the base-emitter voltage and collector current density, as the base width is varied. The peak value of fmax rises as the base width is varied from 10nm to 60nm as shown in Figure 4.2(e). The initial steep rise in fmax at smaller base widths (< 30 nm) is due to the fact that it is directly related to fT as in (4.8). Hence, an increase in fT leads to a corresponding increase in fmax. The peak fmax of 325 GHz observed here was achieved for a base width of 60 nm at 194.73 kA/cm2 . Keeping in mind, the objective is the maximization of both fT and fmax, the base width of 30nm serves as an excellent solution. The peak fT that can be achieved at this base width is 388 GHz, whereas the fmax reaches a maximum of 320 GHz. Figure 4.2(e) shows the variation of peak fT and fmax with base width. The fT actually dips for a base width of 20nm. At lower base widths, the base resistance dominates, leading to a decrease in fmax. It should be noted that other active layers have not yet been optimized. 118 Figure 4.2(c) fmax vs base-emitter voltage for various base widths for VCE=1.8V and VBE from 0.6V to 1.07 Vo lts for a fixed base doping of 8x1019 cm-3 Figure 4.2(d) fmax vs collector current density for various base widths for VCE=1.8V and VBE from 0.6V to 1.07 Volts for a fixed base doping of 8x1019 cm-3 119 Figure 4.2(e) Variation of peak fT and fmax with base width for a base doping of 8x1019 cm-3 , (+) Peak fT fmax from results by Bolognesi et al. [2] for similar device structure. The same epi-structure as before given in Table 3.3 is now investigated for the effects of base doping on high frequency parameters. The base transit time is not a very strong function of the base doping NB as given by the diffusion coefficients appearing in τb [1], DnB = where µ kT µ q (4.9) = µ(N) as given chapter 2. The diffusion coefficient in the base layer of the HBTs saturates to a relatively low constant values at high doping levels. As a result, increasing the doping has the advantage of decreasing the sheet resistance and at the same time without subsequent increase in the base transit time. However, fT does increase somewhat 120 as the doping level gets higher as shown in Figure 4.3(c). This could be due to impurity scattering in the base when it is doped to the material solubility levels. This is confirmed by the simulations results for fT versus base doping shown in Figure 4.3(a). The peak fT increases with doping and shifts to higher collector current density. This effect is pronounced for doping of 3x1019 cm-3 , 7x1019 cm-3 and 1x1020 cm-3 . Similarly, fmax increases with doping as predicted by (4.9) and is verified by the simulations as shown in Figure 4.3(b). The peak fmax levels shift to lower current density as the doping is increased. A doping of 7x1019 cm-3 is seen as optimum value for maximizing both fT and fmax. Figure 4.3(a) fT versus emitter-base voltage for various doping levels for VCE=1.8V for a fixed base width of 20nm 121 Figure 4.3(b) fmax versus emitter-base voltage for various doping levels for VCE=1.8V for a fixed base width of 20nm Figure 4.3(c) Variation of peak fT and fmax with base doping for a fixed base width of 20nm, (+) indicates peak fT and fmax results from Bolognesi et al. [2] for similar device structure. 122 4.2.2 Optimization of the Emitter Layer Design of the emitter layer can be improved in a number of ways, such as by using a graded or abrupt base-emitter heterojunction, by varying the layer thickness (XE) and finally, by changing the doping level (N E). For the HBT under consideration, InPGaAsSb forms an abrupt heterojunction at the emitter-base region with a staggered band gap line up. This bestows an advantage as previously discussed in chapter 1, by blocking the hole back injection into the emitter from the base which improves the current gain. Design of the emitter layer thickness is rather straightforward in the sense that it should be thick enough to limit hole back injection and at the same time be able to keep the emitter resistance RE within reasonable limits [1]. The impact of the emitter resistance on τe is to increase τe and reduce both fT and fmax [1]. We can expect that the present transistor would follow the same general trends, since the basic underlying physics remains unchanged. Figure 4.4(a) shows the simulation results for the effect of change in emitter width on fT . The cut-off frequency shows some modest increase as the emitter width is increased. Figure 4.4(b) shows the variation of fmax with emitter width. The increase in emitter width increases the emitter transit time τe. Since τe constitutes a small proportion of the overall transit time, the increase has minimal effects on fT and fmax. Figure 4.4(c) shows the variation of peak fT and fmax with emitter width. The optimum emitter width based on the simulation and experimental [2] results is chosen as 90nm. 123 Figure 4.4(a) Effect of emitter width variation on fT for VCE=1.8V for a fixed emitter doping of 3x1017 cm-3 . Figure 4.4(b) Effect of emitter width variation on fmax for VCE=1.8V for a fixed emitter doping of 3x1017 cm-3 . 124 Figure 4.4(c) Variation of peak fT and fmax with emitter width for a fixed emitter doping of 3x1017 /cm3 , (+) indicates peak fT and fmax results by Bolognesi et al. [2] for similar device structure Emitter doping plays a more significant role in the high frequency performance of the device than emitter width. Emitter doping directly influences the base-emitter junction capacitance Cje which appears in the fT equation in (4.3). Typically for a HBT Cje >> Cjc, so at lower doping levels, Cje is lower and hence charges faster. Figure 4.5(a) shows the effect of emitter doping on cut-off frequency as a function of collector current JC. From a simple analysis of the graph we might infer that the lowest possible emitter doping is desirable. However, at low emitter doping the transistor is susceptible to high current effects. Hence, a higher doping is chosen so that the current density is reasonably high and at the same time keeping τe low. From (4.8) we find that the fmax dependence of on the emitter design is only through fT . Neither the base resistance nor the base-collector 125 capacitance is affected by variations in the emitter doping. Therefore, once the fT is optimized for an emitter layer design, fmax is automatically optimized. The behavior of the InP/GaAsSb/InP transistor was investigated for high frequency performance for various values of emitter doping. Figure 4.5(b) shows the variation of fmax with emitter doping. An emitter doping of 3x1017 /cm3 seems like a good value from the plot of peak fT and fmax in Figure 4.5(c). However, using this value of doping in the final structure resulted in high current degradation occurring at lower current densities than normal. The simulations show that both fT and fmax are optimized for a emitter doping of 7x1017 cm-3 . As discussed before, though the high frequency performance is better at the lower doping levels, we settle upon a higher doping level to delay high current effects. Figure 4.5(a) fT versus base-emitter voltage for various doping levels for VCE=1.8V for a fixed emitter thickness of 700 A0 126 Figure 4.5(b) fmax versus base-emitter voltage for various doping levels for VCE=1.8V for a fixed emitter length of 700 A0 Figure 4.5(c) Variation of peak fT and fmax with emitter doping for a fixed emitter width of 700 A0 . 127 4.2.3 Collector Layer Optimization The design of the collector layer is more complex when compared with the other two layers, because the factors affecting its design are not only thickness and doping but also the collector-emitter bias VCE. The collector depletion layer Xdep, controlled by VCE also plays a major role in determining the high frequency performance of the HBT. The design is further complicated by the fact that the time constants in the total transit time τec have opposite dependencies on Xdep, and that depletion width is both a function of bias voltage and current. However, the final decision boils down to the fact that whether we need to optimize the fT , fmax performance or the collector breakdown voltage. The choice of VCE directly influences the fT and fmax, in that an increase in VCE increases Xdep, which in turn, diminishes the collector charging time τc and results in a rapid increase in fT . Also, the space charge transit time τsc is directly proportional to Xdep. So there is an optimum value of depletion width beyond which, the increase in τ sc more than compensates for the decrease in τc and the cut-off frequency decreases as in Figure 4.7(c). However, fmax continues to increase because its Xdep dependency cancels out at higher VCE bias. This increase is not indefinite and stops as soon as Xdep reaches the collector width. So an optimum value of VCE would be to optimize both fT and fmax. For InP/GaAsSb/InP HBTs the optimum VCE is 1.8V. These processes are illustrated in Fig 4.16 with reference to AlGaAs/GaAs HBTs [1]. The two time constants relating to the collector design are given as τ sc + τ c = X dep 2vsat + ( RE + RC ). 128 Acε s X dep (4.10) An interesting point to be noted here is that, the collector resistance RC is dependent on the Xdep through the epitaxial resistance in the undepleted collector. So by choosing a collector thickness equal to the depletion depth we can eliminate the portion of collector resistance related to the undepleted collector. In view of the conflicting effects of VCE on each of the transit time components in (4.10), we aim to minimize their sum as a whole. As a rule of the thumb reducing the collector thickness, increases fT . We cannot, however design a collector layer with the thinnest possible width, because for many RF and high power applications, it is essential that the collector be thick enough to ensure a high collector breakdown voltage. The fT and fmax simulation results for various collector widths are shown in Figure 4.7(a) and 4.7(b). Figure 4.7(c) shows the variation of peak fT and fmax with the collector width. Keeping in terms with the requirements discussed above, a collector width of 0.2 micron is chosen as the optimum value. Figure 4.6 Calculated collector depletion thickness, fT and fmax, as a function of VCE 129 Figure 4.7(a) fT vs base-emitter voltage for several collector widths for VCE=1.8V for a fixed collector doping of 3x1016 /cm3 Figure 4.7(b) fmax vs base-emitter voltage for several collector widths for VCE=1.8V for a fixed collector doping of 3x1016 /cm3 130 Figure 4.7(c) Variation of peak fT and fmax with collector width for a fixed collector doping of 3x1016 /cm3 The collector region is usually doped lower than the base so that it can be fully depleted for a given VCB. A good choice of NC can obtained from [1] Nc = 2ε s (φcb + VCB ) 2 qX C (4.11) This works out very well for maximizing fmax, because the collector junction capacitance Cjc is directly proportional to the square root of collector doping. Hence a lower collector doping ensures a higher fmax. However, there are two main disadvantages to doping the collector at a lower value. Kirk effects tend to be pronounced at lower collector doping because the effective base width increases at higher collector current densities due to electric field reversal at the base-collector junction. This field reversal can be delayed by higher doping until the higher collector current densities fully deplete 131 the collector layer. The other disadvantage, as previously discussed would be the lowering of the breakdown voltage, if lower doping is employed. At higher doping, there would be more carriers to balance the field produced by a larger BVCEO. As we did with all the design parameters we have to strike a compromise and pick a doping level that ensure a high enough fmax as well as good breakdown voltages. From Figure 4.8(c) we see that contrary to what we discussed fmax continues to increase with higher doping. However, from (4.8) we see that any increase in Cjc is also cancelled by a corresponding increase in fT . We can see from Figures 4.8(a) and 4.8(b) that for a doping of 2x1017 cm-3 both fmax and fT are jointly optimized. Also at this level of doping the current and power gain is better than what could be achieved at a lower doping. The design of collector layer is interesting in the sense that, once we determine the desired Xdep for best high frequency performance, then XC can be set to the desired Xdep. Also from the value of XC, the desired NC can be obtained from equation or through simulations. 132 Figure 4.8(a) fT Vs Base emitter Voltage for various collector doping for a fixed collector width of 2000A0 Figure 4.8(b) fmax Vs Base emitter voltage for various collector doping for a fixed collector width of 2000A0 133 Figure 4.8(c) Variation of peak fT and fmax with collector doping for a fixed collector width of 2000A0 . 4.3 Optimized Structure The final InP/GaAsSb/InP HBT structure optimized for high frequency performance is shown in Figure 4.9. This structure incorporates the best layer thickness and doping concentration values as determined from the simulations discussed in the previous sections. The changes to the transistor structure have been to the main active layers of the transistor. The design values for the contact layers and the sub-collector are the same as that of the original structure by Bolognesi et al [2]. The emitter and base layer thickness have been inc reased to 900 A0 and 300 A0 , respectively. The base la yer doping remains nearly the same at 7x1019 cm-3 vs 8x1019 cm-3 of Bolognesi et al. [2]. Both the emitter and the collector doping have been increased. The collector doping is considerably more than for the structure by Bolognesi et al. [2] with a doping of 3x1016 cm-3 . In order to enable 134 comparison of performance between the structures by Bolognesi et al. [2] at 2x1017 , the bias values for the transistor has been kept at the same values. The structure of the InP/GaAsSb DHBT is repeated here in Figure 4.9(b) for convenience from chapter 3. The simulated layout of the schematic given below complete with the contacts is shown in Figure 4.10. The Gummel characteristics are shown in Figures 4.11. Figure 4.9 (a) Schematic of the Optimized InP/GaAsSb/InP HBT Figure 4.9 (b) Schematic of the InP/GaAsSb/InP HBT structure by Bolognesi et al. [2] 135 Figure 4.10 Simulated Structure of Optimized InP/GaAsSb/InP HBT Figure 4.11 Gummel-Poon simulations of optimized InP/GaAsSb/InP Structure 136 Figure 4.12 shows the frequency performance of the optimized structure at a Vbe and VCE of 0.8 V and 1.8 V, respectively. The peak fT achieved was 440 GHz at 3.7 mA. The AC current gain of 36 dB is closer to that of the original structure. As can be inferred from the graph, we have achieved our optimization objective of increasing the high frequency performance. Figure 4.13 shows the frequency performance of the optimized structure at a Vbe and VCE of 0.8 V and 1.8 V respectively. The fmax performance followed at 396 GHz at 2.62 mA. The performance parameters are summarized in Table 4.1. The increase in fT is due to the cumulative effect of reduced collector and emitter resistance resulting from increased collector and emitter doping. The fmax performance has been enhanced due to the larger base width resulting in lower base resistance. Figure 4.12 Frequency response of the optimized InP/GaAsSb/InP HBT for VCE = 1.8V 137 Figure 4.13 Frequency response of the optimized InP/GaAsSb/InP HBT for VCE = 1.8V Table 4.1 Simulation results for Optimized InP/GaAsSb/InP structure. Performance Parameter Peak values fT , IC 440 GHz, 3.7 mA Fmax , IC 396 GHz, 2.62 mA AC Gain, IC 34 dB, 0.118 mA Power Gain, IC 32 dB, 0.48 mA The evolution of fT with the collector current density for various collector-emitter biases is shown in Figure 4.14(a) for the optimized structure. The general trend of the 138 curves is similar to the ones for the simulation of the original structure in Chapter 3 and the published results [2]. A peak fT of 440 GHz was achieved at a collector current density of 304.55 kA/cm2 for VCE = 1.8 V. The fT roll-off at higher current densities around 300 kA/cm2 can be explained by considering the capacitance at the base-collector junction. Cjc first dips with increasing current density and reaches a minimum value corresponding to the peak fT bias occurring for the condition of zero electric field at the base-collector junction. Further increases in the current density, reverses the electric field at the base-collector junction, resulting in the formation of a small field induced thermionic emission barrier. Even though the magnitude of this barrier EB is small, it has a huge effect on the base charge storage. This reduces the effective base exit velocity by a factor ~exp (-EB/KT) resulting in a sharp rise in Cjc and subsequently a drop in fT [3]. The fmax performance vs the collector current density for various collector-emitter voltages is shown in Figure 4.14(b). The fmax reaches a peak of 396 GHz at JC = 246 kA/cm2 for VCE = 1.8V. fmax is dependent on VCE thru its effect on the depletion width Xdep. As the collector-emitter bias is increased, the depletion width increases leading a drop in the collector charging time τc. 139 Figure 4.14 (a) Evolution of fT with JC for various VCE from 0.8 to 1.8 Volts Figure 4.14 (b) Evolution of fmax with JC for various VCE from 0.8 to 1.8 Volts 140 Figure 4.15 Maximum power and AC current Gain versus collector current density (l) Current gain (n) Power gain for VCE = 1.8V The above observations have interesting implications. The InP/GaAsSb/InP HBTs reach their peak performance at zero electric fields at the base-collector junctions [3]. This suggests that InP/GaAsSb/InP HBTs would perform well even under low VCB biases. Figure 4.15 shows the variation of the AC current gain and unilateral power gain with collector current density. The gain values measured were the maximum gains measured for VCE=1.8V at 2 GHz. The gains remain comparable to that of the original structure discussed in section 3.3., even as we improved the high frequency performance. Figure 4.16(a) and 4.16(b) compares the extracted peak fT and fmax values for the various structures under consideration. The optimized structure shows substantial improvement over that of the HBT fabricated by Bolognesi et al [2]. 141 Figure 4.16(a) Peak fT variation with VCE for InP/GaAsSb/InP DHBT structure of Bolognesi et al. [2] (g ), simulated fT for the same structure ( n) and peak fT for the optimized device structure (o) given in Figure 4.9 Figure 4.16(b) Peak fmax variation with VCE for InP/GaAsSb/InP DHBT structure of Bolognesi et al. [2] (g ), simulated fT for the same structure ( n) and peak fT for the optimized device structure (o) given in Figure 4.9 142 4.4 Conclusions Starting with the basic structure of the HBT suggested by Bolognesi et al. [2], the HBT structure presented in the previous sections of this chapter has been optimized to demonstrate improved fT and fmax performance. The thickness and doping concentration of each layer of the active region of the transistor were investigated for increasing the fmax and fT values. The optimized structure had a peak fT and fmax of 440 GHz and 396 GHz, respectively. The peak current and power gain was 34 dB and 32 dB, respectively. The performance improvement in the cut-off frequency was over 140 GHz above that for the device structure by Bolognesi et al. [2]. However, we have to consider the fact that our model over estimated the cut-off frequency by 44 GHz. The improvement in fmax performance was 99 GHz over that of Bolognesi et al’s structure [2]. The improved performance is mainly due the reduced resistances in the active layers of the device resulting from higher doping used in the collector and emitter layers. The optimum values were chosen without compromising other performance measures like the breakdown voltage and gain. Reference: [1] Liu, W.,“Handbook of III-V Heterojunction Bipolar Transistors”, John Wiley and Sons Inc., New York, pp 26-27, 1998. [2] C.R Bolognesi, M.W Dvorak, O.J Pitts & S.J Watkins, “300 GHz InP/GaAsSb/InP Double HBTs with High Current Capability and BVCEO 6 V”, IEEE Electron Device Letters, Vol. 22, pp361-364, 2001. [3] C.R Bolognesi, M.W Dvorak & S.J Watkins, “InP/GaAsSb/InP Double Heterojunction Bipolar Transistors”, Gallium Arsenide Integrated Circuit (GaAs IC) Symposium, 2002, 24th Annual Technical Digest, pp265-268, 2002. 143 Chapter V Conclusions and Future Work 5.1 Conclusions In this work, we have investigated and simulated the performance of InP/GaAsSb/InP Double Heterojunction Bipolar Transistors (DHBTs). The staggered bandgap line up resulting from the use of GaAsSb as a base in the NPN InP DHBT bestows a huge advantage in terms of removing the collector blocking effect commonly found in other InP-based DHBTs and also in enhancing the collector breakdown voltage. The simulation model for the transistor was developed, complete with its epitaxial structure, terminal contacts, doping, and device physics models, along with the necessary biasing. The material properties of GaAsSb were researched in detail, such as the band energy gap, mobility, effective masses, carrier lifetime, and density of states and incorporated into the device model used for simulation. In order to validate the transistor simulation model, we initially compared the simulation results with the reported experimental results of the fabricated devices from Bolognesi et al. [1, 2] for InP/GaAsSb DHBTs. The simulated model had the same device structure [1] to ensure accurate comparison of the results. The common-emitter characteristics of the device were investigated. One of the unique features of this transistor is its low turn-on voltage. The turn-on voltage of the simulated device matched the published results with a turn-on voltage of 0.4V. For comparison of the commonemitter characteristics, the current values were properly scaled to reflect the difference in the emitter area of the two devices. The high frequency performance of the device matched closely the published results with the simulation results being fT =344 GHz and 144 fMAX=295GHz, in reasonable agreement with the published experimental values of 300 GHz for fT and fMAX, respectively [1]. The variation of the simulated fT with the collector current density also agrees very well with the published values showing a fall-off above 100 kA/cm2 . The current density for the peak fT was found to be 410 KA/cm2 for the device reported by Bolognesi et al [1,2], whereas in our case it was 432 KA/cm2 . The peak fT and fmax as a function of JC were examined versus VCE and found to be in good agreement with the results reported for the experimental devices. The origins of the falloff in the transistors performance at high currents were also investigated. The reason for the fall off in device performance at high currents is due to the formation of the parasitic barrier in the conduction band. Therefore, the high current degradation is not due solely to the conventional Kirk (base pushout) effect, but also due to the formation of a parasitic barrier in the conduction band of the collector. The parasitic barrier height is a function of the collector current density, collector doping and the collector-emitter bias VCE. The parasitic barrier formed is as large as 140 meV. Simulations also show that there is significant hole accumulation in the collector at the higher base-emitter biases. The parasitic barrier increases with VBE because the depletion electric field is opposite to the barrier electric field. The forward bias decreases the depletion field leading to a increase in the parasitic barrier height. Once we validated the transistor model, the HBT structure was optimized to demonstrate improved fT and fmax performance. The thickness and doping concentration of each layer of the active region of the transistor were investigated. The optimized structure had a thicker base layer at 30 nm and higher emitter doping of 7x1017 /cm3 . The collector was also doped somewhat higher at 2x1017 /cm3 . The optimized structure had a 145 peak fT and fmax of 440 GHz and 396 GHz, respectively. The peak current and power gain was 34 dB and 32 dB, respectively. The performance improvement in the cut-off frequency was over 140 GHz above that for the device structure reported by Bolognesi et al. [1]. However, we have to consider the fact that our model over estimated the cut-off frequency for the original structure by 44 GHz. The improvement in fmax performance for the optimized structure was 99 GHz over that of the original structure by Bolognesi et al. [1]. The improved performance is mainly due the reduced resistances in the active layers of the device resulting from highe r doping used in the collector and emitter layers and a wider base width. The optimum values were chosen without compromising other performance measures like the breakdown voltage and gain. The InP/GaAsSb/InP transistors have a number of advantages as elucidated in earlier chapters, ranging from easy integration with optical transceivers to the simplicity of their fabrication. These advantages make it worthwhile to aggressively pursue the development of these devices in spite of stiff competition from other technologies like SiGe. Also, the low noise performance and the high linearity of the devices over a wide range of frequencies make these devices attractive in high growth markets like cellular base stations and multi-band cellular phones. [3] 5.2 Future Work In this work, we have established the basics for the modeling and simulation of the InP/GaAsSb/InP transistor. Other InP-based transistors [4] have been aggressively scaled and have shown impressive performances in terms of high frequency performance. However, the current device is only moderately scaled and hence there is still scope for 146 further performance enhancement. In addition, the simplicity of the device fabrication with no complicated doping or grading schemes at the collector junction makes it very attractive for use in a production system. The performance of a pnp InP/GaAsSb DHBT could also be investigated for use in a complimentary configuration for uses in digital applications. The high current degradation follows a unique phenomenon, in that it’s not entirely due to base out or the formation of parasitic barrier but a combination of the above to factors. Therefore, the main thrust of future investigations would be to analyze the performance fall off at high currents in terms of doping and the biases. Reference: [1] C.R Bolognesi, M.W Dvorak, O.J Pitts & S.J Watkins, “300 GHz InP/GaAsSb/InP Double HBTs with High Current Capability and BVCEO 6 V”, IEEE Electron Device Letters, vol. 22, pp 361-364, 2001. [2] C.R. Bolognesi, N. Matine, M.W. Dvorak, X.G. Xu, J.Hu and S.P. Watkins, “ NonBlocking Collector InP/GaAsSb/InP Double Heterojunction Bipolar Transistors with a Staggered Lineup Base-Collector Junction”, IEEE Electron Device Lett., vol 20 pp 155157 , 1999. [3] M. Inerfield, W. Skones, S. Nelson, D. Ching, P. Cheng, and C. Wong, “High Dynamic Range InP HBT Delta-Sigma Analog-to-Digital Converters”, IEEE J. of SolidState Circuits, Vol. 38, pp 1524-1532, Sept 2003. [4] J. W. Rodwell, M. Urteaga, T. Mathew, D. Scott, D. Mensa, Q. Lee, J. Guthrie, Y. Etser, Suzanne C. Martin, R. P. Smith, S. Jaganathan, S. Krishnan, S.I. Long, R. Pullela, 147 B.Agarwal, U.Bhattacharya, L. Samoska, and M. Dahlstrom, “Submicron Scaling of HBTs”, IEEE Trans. On Electron Devices, vol.48, pp 2606-2624, 2001. 148 Appendix A Run file for Simulation of InP/GaAsSb/InP DHBT go atlas title Emitter-up InP/GaSbAs/InP HBT device simulation (August 2003) # # SECTION 1: Mesh specification # mesh space.mult=1.0 # x.mesh loc=2.3 spac=0.1 x.mesh loc=3.0 spac=0.1 x.mesh loc=3.5 spac=0.1 x.mesh loc=4.0 spac=0.1 # y.mesh loc=0.00 spac=0.03 y.mesh loc=0.18 spac=0.01 y.mesh loc=0.25 spac=0.02 y.mesh loc=0.30 spac=0.02 y.mesh loc=0.37 spac=0.02 y.mesh loc=0.40 spac=0.005 y.mesh loc=0.45 spac=0.005 y.mesh loc=0.64 spac=0.01 y.mesh loc=0.69 spac=0.01 # SECTION 2: # region num=1 region num=2 region num=3 region num=4 region num=5 x.comp=0.49 region num=6 region num=7 x.comp=0.53 Structure definition material=air material=InGaAs material=InP material=InP material=GaSbAs x.min=2.3 x.min=3.5 x.min=3.5 x.min=3.5 x.min=2.9 x.max=4.0 x.max=4.0 x.max=4.0 x.max=4.0 x.max=4.0 y.min=0.0 y.min=0.18 y.min=0.28 y.min=0.33 y.min=0.42 y.max=0.69 y.max=0.28 y.max=0.33 y.max=0.42 y.max=0.44 material=InP material=InGaAs x.min=2.9 x.max=4.0 y.min=0.44 y.max=0.64 x.min=2.3 x.max=4.0 y.min=0.64 y.max=0.69 # elec num=1 name=emitter x.min=3.5 x.max=4.0 y.min=0.16 y.max=0.18 elec num=2 name=base x.min=2.9 x.max=3.4 y.min=0.40 y.max=0.42 elec num=3 name=collector x.min=2.3 x.max=2.8 y.min=0.62 y.max=0.64 CONTACT NAME=emitter CON.RESIST=5e-8 CONTACT NAME=base CON.RESIST=2e-7 CONTACT NAME=collector CON.RESIST=5e-8 # doping uniform region=2 n.type conc=1.0e19 doping uniform region=3 n.type conc=3.0e19 doping uniform region=4 n.type conc=3.0e17 149 doping uniform region=5 p.type conc=4.0e19 doping uniform region=6 n.type conc=3.0e16 doping uniform region=7 n.type conc=1.0e19 # # SECTION 3: Set models and define material parameters ## material material=GaSbAs eg300=0.72 nc300=2.3875e17 nv300=4.6456e18 affinity=4.19 permittivity=14.4 material material=InP eg300=1.35 nc300=5.7e17 nv300=1.1e19 affinity=4.38 permittivity=12.35 material material=InGaAs eg300=0.75 nv300=7.7e18 nc300=2.1e17 affinity=4.5 permittivity=13.88 # Emitter Contact n+ InGaAs region material region=2 taun0=5.0e-9 taup0=1.0e-9 vsatn=2.5e7 vsatp=7.7e6 #Emitter Cap n+ InP region material region=3 taun0=1e-10 taup0=1e-11 vsatn=2.0e7 vsatp=1e6 # Emitter n- InP region material region=4 taun0=1.0e-7 taup0=1.0e-8 vsatn=2.0e7 vsatp=1e6 # Base P+ GaSbAs region material material=GaSbAs taun0=1.005e-7 taup0=1.533e-6 taurel.el=0.48e11 taumob.el=0.3e-10 vsatn=2.0e8 vsatp=6.8e7 # Subcollector n InP region material region=6 taun0=1.0e-8 taup0=1.0e-9 vsatn=2.0e7 vsatp=1e6 # Collector n+ InGaAs region material region=7 taun0=5.0e-9 taup0=1.0e-9 vsatn=2.5e7 vsatp=7.7e6 #Setting Mobility Values mobility material=InP analytic.n mu1n.caug=300 mu2n.caugh=4917 ncritn.caugh=6.4e17 deltan.caugh=0.46 mobility material=InP analytic.p mu1p.caug=20 mu2p.caugh=151 ncritn.caugh=7.4e17 deltap.caugh=0.96 mobility material=GaSbAs analytic.n mu1n.caug=574 mu2n.caugh=750 ncritn.caugh=1.8190e17 deltan.caugh=1.057 mobility material=GaSbAs analytic.p mu1p.caug=20 mu2p.caugh=60 ncritn.caugh=7.935e18 deltap.caugh=0.45244 mobility material=InGaAs analytic.n mu1n.caug=3372 mu2n.caugh=11599 ncritn.caugh=8.9e16 deltan.caugh=0.76 mobility material=InGaAs analytic.p mu1p.caug=75 mu2p.caugh=331 ncritn.caugh=1.0e18 deltap.caugh=1.37 150 #Setting Models models consrh hcte.el fldmob conmob analytic evsatmod=1 temperature=300 fermidirac print output con.band val.band solve init save outf=newhbtecap.str tonyplot newhbtecap.str -set newhbtecap.set # SECTION 4: Initial solution and collector bias ramp method newton trap autonr output e.velocity solve prev solve v3=0.01 ac freq=1e6 direct solve v3=0.025 vstep=0.025 electr=3 nstep=2 solve v3=0.1 vstep=0.05 electr=3 nstep=10 solve v3=0.65 vstep=0.05 electr=3 nstep=23 ## Calculate Gummel plot # 1 - emitter 2 - base 3 - collector log outf=200basehbt_dcgain_90.log method newton trap solve solve solve solve save v2=0.01 ac freq=1e6 direct v2=0.025 vstep=0.025 electr=2 nstep=2 ac freq=1e6 direct v2=0.1 vstep=0.1 electr=2 nstep=5 ac freq=1e6 direct v2=0.65 vstep=0.05 electr=2 nstep=5 ac freq=1e6 direct outf=originalhbt_90.str tonyplot tonyplot originalhbt_90.str 200basehbt_dcgain_90.log # # Frequency domain AC analysis up to 100 GHz # log outf=200basecorrected_90.log s.param gains inport=base outport=emitter width=50 load inf=originalhbt_90.str master.in 151 # solve previous ac freq=1 direct solve ac freq=10 fstep=10 mult.f nfstep=8 direct solve solve solve solve ac ac ac ac freq=2e9 direct freq=5e9 direct freq=1e10 direct freq=2e10 fstep=2e10 nfstep=23 direct # # AC current gain versus frequency # tonyplot 200basecorrected_90.log -set hbtex06_4_log.set quit 152