Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Fabrication and modeling of composite-collector heterojunction bipolar transistors 1993

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata

Download

Media
ubc_1994-0152.pdf [ 2MB ]
ubc_1994-0152.pdf
Metadata
JSON: 1.0065280.json
JSON-LD: 1.0065280+ld.json
RDF/XML (Pretty): 1.0065280.xml
RDF/JSON: 1.0065280+rdf.json
Turtle: 1.0065280+rdf-turtle.txt
N-Triples: 1.0065280+rdf-ntriples.txt
Citation
1.0065280.ris

Full Text

Fabrication and Modeling of Composite- Collector Heterojunction Bipolar Transistors By Bahram Ghodsian B.ENG. in Electronic & Electrical Engineering (Hons.) University of London (King’s College London), U.K. (1991) A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF ELECTRICAL ENGINEERING We accept this thesis as conforming to the re uired standard The University of British Columbia November 1993 © Bahram Ghodsian 1993 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives, It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. (Signature) _______ ELECTI’)CAL £W. Department of ____________________ The University of British Columbia Vancouver, Canada Date ____ DE-6 (2/88) Abstract This thesis is concerned with processing and modeling aspects of heterojunction bipolar transistors (HBTs) with composite-collectors. HBTs with InGaAs/InP composite collectors were designed, fabricated and measured. Their dc characteristics are compared with those of a device which was structurally similar but with a conventional n-InGaAs collector. The measured data for both devices can be well-described by an analytical model. The model indicates the need to ensure that the InGaAs layer in the composite collector exceed a critical thickness if the current gain is to be preserved. It is also shown that the spacer layer between the emitter and base, which was used to prevent zinc diffusion into the emitter, must be considered in the model if the collector current is to be correctly predicted. The comparison of the experimental data also suggests that surface recombination is not a dominant base recombination current in the devices studied. 11 Contents Abstract ii List of Figures vi List of Tables xiii Acknowledgment xv 1 Introduction 1 1.1 Background 1 1.2 Importance of the Composite-Collector HBT 4 1.3 Overview 5 2 Wafer Design Considerations for the Composite-Collector HBT(CCHBT) 6 2.1 High Breakdown Voltage and High Speed in an HBT 7 2.2 HBT Structure 9 2.2.1 Emitter Configuration 9 2.2.2 Base Layer 11 2.2.3 Spacer (Base Setback layer) 12 2.2.4 Collector Configuration 15 3 Composite-Collector Heterojunction Bipolar Transistor (CCHBT) Analytical Model Development 25 3.1 Boundary Conditions for the E-B junction 25 3.2 Composite-Collector Heterojunction Bipolar Transistor Model 29 3.2.1 Base-Collector Junction 32 111 3.3 3.4 3.4.1 3.4.2 3.4.3 3.4.4 3.4.5 3.4.6 3.4.7 3.4.8 3.4.9 38 41 42 43 43 44 45 46 47 48 4.1 4.2 4.2.1 4.2.2 4.2.2.1 4.2.2.2 4.2.2.3 4.2.2.4 4.2.2.5 4.2.2.6 5.1 5.2 5.2.1 5.2.2 5.2.3 5.2.4 5.2.5 for lnGaAs#2 and lnP#2 Process for the Fabrication of Devices Scribing and Cleaning the Wafers Fabrication steps Spinning the Photoresist Wet Chemical Etch for the Emitter Mesa Photoresist Stripping and Cleaning Wet Chemical Etch for the Base Mesa Wet Chemical Etch for the Collector Mesa Pattern for Metal Contact 65 65 67 67 72 80 86 90 Recombination Current in HBTs Material Parameters of Ini_GaAsPi_. Ratio of x to y for lattice-matching to lnP . Bandgaps and Electron Affinity Effective Masses Low-Field Low-Doping Majority Carrier Mobility Doping Concentration Dependency of Mobility Auger Coefficients of Ini_o.47yGaAsyPi_y Radiative Coefficient Dielectric Constants Shockley-Read-Hall Lifetime 47 4 5 Non-Self-Aligned Method for Device Fabrication 50 SIMS Plots of wafer lnGaAs#1 and lnP#1 and Modifications 50 55 55 56 56 57 57 57 58 58 Results and Discussion Experimental Procedure DC Characteristics Common-Emitter and Base Characteristics Gummel Plots Parasitic Resistance Measurement Emitter Contact Geometries Model and Experimental Comparison iv 6 Conclusion and Recommendation for Future Research . 95 6.1 Conclusion 95 6.2 Recommendation for Future Research 96 Bibliography 97 V List of Figures Figure 2.1 The basic structure of a composite-collector HBT 6 Figure 2.2 Conduction band profile for a two-layer emitter at equilibrium 10 Figure 2.3 Conduction band profile for a three-layer emitter at equilibrium 11 Figure 2.4 Conduction band profile simulated by LUMIN at equilibrium. The setback layer lies between 0.3 and 0.3O8tm, the emitter is to the left and the base to the right 14 Figure 2.5 Valence band profile simulated by LUMIN at equilibrium. The setback layer lies between 0.3 and 0.308m, the emitter is to the left and the base to the right 14 Figure 2.6 Calculated energy band profile for conventional collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 19 Figure 2.7 Calculated electric field profile for conventional collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 19 Figure 2.8 Calculated energy band profile for inverted field collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 20 Figure 2.9 Calculated electric field profile for inverted field collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 20 vi Figure 2.10 Calculated energy band profile for launcher collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 21 Figure 2.11 Calculated electric field profile for launcher collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 21 Figure 2.12 Calculated energy band profile for undoped collector performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 22 Figure 2.13 Calculated electric field profile for undoped collector performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 22 Figure 2.14 Calculated energy band profile for composite collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 23 Figure 2.15 Calculated electric field profile for composite collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 23 Figure 3.16 Energy-band diagram of an ideal abrupt N-p heterojunction at thermal equilibrium 26 Figure 3.17 Illustrating the conduction band profile and flow of electrons from the base to the collector under three conditions: (a) regular DHBT with no undoped lnGaAs layer, (b) CCHBT with a narrow undoped lnGaAs layer, and (c) CCHBT with a wide undoped lnGaAs layer 30 Figure 3.18 Schematic illustration of HBT structure and the conduction band profile after including a lightly-doped layer in the Collector 32 vii Figure 3.19 The plot of z.Ec as function of A for the CCHBT device for VCB=O 36 Figure 3.20 The plot of I as function of A for the CCHBT device for VBE=O.4V and VCB=O 37 Figure 3.21 The plot of ‘B as function of A for the CCHBT device for VBE=O.4V and VCB=O 37 Figure 3.22 The various components of base current for an emitter area of 40 x 40tm2 39 Figure 3.23 The experimental data for majority carrier mobility in n and p-type lattice-matched Ini_o.47GaAsPi_, together with the best fit curves as a function of arsenic composition. . . . 45 Figure 3.24 The experimental data for hole and electron mobility in In053Ga47As and InP as a function of doping concentration, together with the best fit curves 46 Figure 4.25 The ln(113), As and Zn secondary ion count profiles for lnGaAs#1. The vertical dashed lines define the metallurgical base boundaries 51 Figure 4.26 The ln(1 13), As and Zn secondary ion count profiles for lnP#1. The vertical dashed lines define the metallurgical base boundaries 51 Figure 4.27 Calculated energy band profile, before and after outdiffusion of Zn as performed by LUMIN for lnP#1 based on results of SIMS plot in Figure 4.26, and the original specification, respectively. Refer to “Cumulative Depth” column in Table 2.1 for layer locations 52 viii Figure 4.28 The ln(113), As and Zn secondary ion count profiles for lnGaAs#2. The vertical dashed lines define the metallurgical base boundaries 54 Figure 4.29 The ln(113), As and Zn secondary ion count profiles for lnP#2. The vertical dashed lines define the metallurgical base boundaries 54 Figure 4.30 The HBT fabrication sequence: (1) Spin the PR on wafer; (2) Pattern the PR and then develop it; (3) etch the emitter mesa; (4) Spin, pattern, develop the PR and then etch the base mesa; (5) Spin, pattern, develop the PR and then etch the collector mesa; (6) Spin, pattern, develop the PR and then evaporate Ti/Pt/Au 61 Figure 4.31 Schematic diagram of the mask for the non-self-aligned HBT 62 Figure 4.32 Optical micrograph of a non-self aligned HBT (emitter area of 60 x 60 tm2). The mangification factor is 1044 62 Figure 4.33 The SEM micrograph, showing the emitter mesa etch profile. . 63 Figure 4.34 SEM micrograph before lift-off, showing the metal lip resulting from the chlorobenzene process 63 Figure 4.35 SEM micrograph before lift-off, showing another view of the same effect as shown in Figure 4.34 64 Figure 5.36 Measured output characteristics of lnGaAs#2 and lnP#2 HBT for an emitter area of 60 x 80im2. The base current starts at 5pA and increases in steps of 20A 69 Figure 5.37 Measured ouput characteristics of lnGaAs#1 HBT for an emitter area of 60 x 80irn2. The base current starts at 5pA and increases in steps of 20tA 69 ix Figure 5.38 Magnified offset region of the measured output characteristics shown in Figure 5.36. The base current starts at 51zA and increases in steps of 20tA 70 Figure 5.39 Measured reverse characteristics of lnP#2 HBT for an emitter area of 60 x 80im2. The base current starts at 5tA and increases in steps of 20tA 70 Figure 5.40 Measured reverse mode Gummel plot of lnP#2 HBT for an emitter area of 60 x 80m2 71 Figure 5.41 The measured common-base characteristics of InGaAs#2 and lnP#2 HBTs for an emitter area of 40 x 40m2. The emitter current starts at OA and increases in steps of 2mA. . 71 Figure 5.42 The set up for devices, measured by HP4145B; (a) Gummel plots, (b) Common emitter I-V characteristics, (c) Common base I-V characteristics 72 Figure 5.43 Non-alloyed Gummel plot (magnitude of current) of lnGaAs#2 at Vcb=OV for emitter area 40 x 40m2 73 Figure 5.44 Non-alloyed Gummel plots (magnitude of current) of InP#2 at Vcb=3V for emitter area 40 x 40m2 73 Figure 5.45 Non-alloyed Gummel plot (magnitude of current) of lnGaAs#1 at Vcb=OV for emitter area 40 x 40m2 74 Figure 5.46 The measured dc gain 3 versus the collector current in lnGaAs#1, lnGaAs#2 and InP#2 for an emitter area of 40 x 40m2 79 Figure 5.47 The measured small-signal gain hfe versus the collector current in lnGaAs#1, InGaAs#2 and lnP#2 for an emitter area of 40 x 40im2 79 x Figure 5.48 Transmission line pattern used to experimentally determine the emitter, base and collector contact resistances. The area of each pad is 60 x 80um2 and the separation between them starts at 21um and increase in steps of 1im 80 Figure 5.49 Equivalent resistor network representing the end effect and the contact resistance 81 Figure 5.50 Plot of total contact to contact resistance as a function of L to obtain transfer length and contact resistance values 82 Figure 5.51 Resistance measurements data for the emitter contact (area = 60 x 80im2) 84 Figure 5.52 Resistance measurements data for the base contact (area 60 x 80im2) 84 Figure 5.53 Resistance measurements data for the collector contact (area = 60 x 80tm2) 85 Figure 5.54 The measured common-base Tc vs VCB characteristics of lnGaAs#2 HBT for the cases of alloyed and non-alloyed contacts, for emitter area 40 x 40pm2. The emitter current starts at OA and increases in steps of 2mA 85 Figure 5.55 Dependence of collector and base current density on the emitter length in lnGaAs#2 87 Figure 5.56 Dependence of collector and base current density on the base emitter separation for devices with emitter area =60 x 80rni2.. 88 Figure 5.57 Dependence of collector and base current density on the emitter width 88 Figure 5.58 Dependence of collector and base current density on the emitter area 89 xi Figure 5.59 The dc current gain at Vbe = 0.4V vs emitter area to periphery ratio (In case of lnGaAs#2) 89 Figure 5.60 Schematic illustration of the emitter-base junction energy band profile after including a lightly-doped layer between the emitter and base layers 91 Figure 5.61 Comparison of experimental and analytical model data for the Gummel Plot of lnGaAs#2 HBT (emitter area 40 x 40tm2)at Vcb=OV. The effect of spacer and parasitic resistances are not modeled 93 Figure 5.62 Comparison of experimental and analytical model data for the Gummel Plot of lnGaAs#2 HBT (emitter area 40 x 40,um2) at Vcb=OV, taking into account the effect of the emitter-base spacer layer and parasitic resistances 93 Figure 5.63 Comparison of experimental and analytical model data for the Gummel Plot of lnP#2 HBT (emitter area 40 x 40tm2)at Vcb=3.OV, taking into account the effect of the emitter-base spacer layer and parasitic resistances 94 xii List of Tables Table 2.1 Layer specification for the HBTs used in this work. The two collector layers are further specified in Tables 2.2 and 2.3. . 16 Table 2.2 Collector configuration of SHBT 24 Table 2.3 Collector configuration of CCHBT 24 Table 3.4 Lattice constants of the four binary compounds 42 Table 3.5 Hole effective masses of the four binary compound semiconductors 44 Table 3.6 Radiative coefficients of four binary compound semiconductors 47 Table 3.7 Dielectric constant of four binary compound semiconductors 48 Table 3.8 Summary of material parameters used in the CCHBT model. . 49 Table 5.9 The specification given by EPI for the lnGaAs#2 HBT 66 Table 5.10 The specification given by EPI for the lnP#2 HBT 66 Table 5.11 The collector and the base current ideality factors for the three structures. The ideality factors were measured by taking the average value at five different points on the linear region of the Gummel plot 74 Table 5.12 Summary of the resistance measurements of a non-alloyed contact of lnGaAs#2 82 xlii Table 5.13 Summary of the resistance measurements of an alloyed (at temperature 300°C) contact of InGaAs#2 83 Table 5.14 Summary of the resistance measurements of an alloyed (at temperature 350°C) contact of InGaAs#2 83 xiv Acknowledgment Firstly, I would like to thank my parents for their ultimate patience, support, under standing and encouragement throughout the years, specially during the course of this work. Secondly, I would like to express my sincere gratitude to my thesis supervisor Professor Dave Puifrey, for his generous financial support and guidance during the course of this work. His enthusiasm and fresh insights have provided constant encouragement in my work. Very special thanks go to my co-supervisor, Dr. Sean McAlister, who is the leader of the Device Physics group at the National Research Council, for invaluable assistance and helpful discussions on HBT design and simulations, and for his efforts in purchasing wafers from EPI1 in time for this work to be carried out. It has been a valuable experience to work with him. A special note of thanks to other members of the Device Physics group, Dr. Zine-Eddine Abid for his continuous assistance in the cleanroom, Dr. Ross McKinnon for use of his photolithography mask plate and Dr. Zhan-Ming Li for permitting me to use his program LUMIN. There are a number of other people in the Microfabrication group at NRC to whom I am indepted for their technical assistance, in particular, Dr. Mike Davies for his wet chemical etch recipe and the ohmic contact structure, Dr. Mahmoud Fallahi and Dr. Margaret Buchanan for their photolithography process, Mr. Richard Barber for RIE, Mr. Philip Chow-Chong for wafer cleaning procedures, Mr Jeff Fraser for taking the SEM micrographs, Mr. Stephen Rolfe for performing the SIMS analysis and Mr. Paul Marshall for performing the metal deposition. Finally, I would like to thank my colleagues Mr. Jun Xiong Feng and Mr. Shawn Searles and all my friends in the Solid-State Electronics group within the Department of Electrical Engineering at UBC for the discussions we have had on HBTs and other matters. Epitaxial Product International Ltd., Cypress Drive, St. Mellons, Cardiff, U.K., CF3 OEG. xv To my parents and my sisters. xvi Chapter 1 Introduction 1.1 Background The idea of having a wide bandgap material for the emitter goes back to the early years of the transistor. The announcement of the first transistor was made in 1948 [1], and the first-ever heterojunction device was proposed by Shockley in 1951 [2], and was later developed by Kroemer in 1957 [3], 1982 [4], 1983 [5]. Since then there have been a number of attempts to make heterojunction devices. The first attempts were made by growing Ge on GaAs by a vapor deposition method [6] or Ge on Si by an alloying process [7]. However, the material that was made by these methods, using the primitive fabrication equipment of the day, had high dislocation densities, particulary at the interfaces, which ultimately restricted the usefulness of the device. There was not much development until the mid-70’s and the emergence of two new crystal growth technologies, namely: MOCVD (Metal Organic Chemical Vapor Deposition)[8] and MBE (Molecular Beam Epitaxy)[9]. The emergence of these new technologies not only improved the interface problem to some extent , but also it marked the birth of a new branch of engineering which is now known as “Bandgap Engineering”. It is the ability to manipulate the properties of individual layers during crystal growth which gives engineers the power to make new devices which would be unobtainable otherwise. Many new heterojunction electrical devices have emerged, among them are the High Electron Mobility Transistor (I-IEMT) [10] and the Heterojunction Bipolar Transistor (HBT). In recent years the HBT has attracted much attention. One of the reasons for this sudden interest is the fact that the HBT is much easier to fabricate than the FIEMT. In order to fabricate a high quality HEMT one needs to have access to electron beam lithography to pattern the gate finger. Whereas for an HBT, although to have access to such equipment would be an advantage, it is not necessary. With optical lithography it is possible to make very competitive devices. Another reason for the current popularity of HBTs is that they possess a number of advantages over conventional silicon Bipolar Junction Transistors (BJT), some of which are listed below: 1. One of the main advantages of a heterojunction is that the difference between the energy bandgaps of the emitter and base layer can produce a different energy barrier for the transport of electrons and holes across the emitter-base junction. This means that by having a higher hole barrier in the valence band, one can confine the holes to the base. This leads to a suppression of the reverse injection current from the base to the emitter. This greatly enhances the emitter injection efficiency and ultimately the gain. 2. Due to the fact stated above, a high base doping concentration can be tolerated, which is essential in order to reduce the resistance of the thin base layer (as required for high fmaz). The emitter doping can also be reduced, which consequently reduces the depletion-region capacitance of the emitter-base junction. These measures will improve the emitter charging time (as required for high ft). 3. It is also possible to use a rn-v semiconductor compound with high electron mobility in the base. This means faster transit across the base, which will improve the frequency 2 response of the device. These advantages can only be achieved by use of an appropriate material system for the HBT. There are a few of them to choose from. The one which is used in our model and in fabricated devices is the lattice-matched combination of JnP/Ino53Gao47As. Unless otherwise stated, from now on in this thesis In0•53Ga47As will be referred to simply as IriGaAs. Although the bulk of HBT research for the past ten years has been focused on the A1GaAs/GaAs material system, there has been growing interest in structures composed of indium materials. These structures are strong challengers for high speed and optoelec tronic applications [11,12]. Material systems such as A1o.481n0.52s/Ino.3Ga7sand InP/InGaAs are very suited to HBT technology, because they take advantage of the very high electron mobility of InGaAs (11000 cm2/V-s) as compared to GaAs (8500 cm2IV-s). The use of an InGaAs base is additionally attractive for applications in long wavelength fiber optic communication systems since the energy bandgap is well matched to the 1.3—1.6tm spectral range of contemporary low-loss and low-dispersion silicon fibers [13]. Recent de scriptions of InP/InGaAs ITBTs reported cut-off frequencies, ft, as high as 110GHz [14] and small-signal current gain, life, as large as 24,000 [15]. AlInAs/InGaAs is equally impres sive, with ft of 78GHz [16] and hfe of 1500 [17]. These performance figures are especially impressive in light of the relative immaturity of both the process technology and device models for these systems. To summarize some of the important advantages of the InP/InGaAs system (from a device point of view) which have become very apparent, we note: 3 1. The InP/InGaAs material system has a larger valence band discontinuity (LE 0.34eV) [18] at the emitter-base junction, as opposed to the AIGaAs/GaAs (0.l5eV)[18] or AlTnAs/InGaAs (0.24eV) [18] systems. 2. There is a larger energy separation between the F-L valleys for both InP and InGaAs (0.60eV, 0.56eV respectively) [19] as compared to A1GaAs and GaAs (0.33eV, 0.30eV respectively) [191. This, of course, means that carriers can be kept in the high-mobility F-valley for larger distances, so improving the device’s speed. 3. GaAs suffers from a high surface recombination velocity (106 cms’ vs i03 cms’ in InGaAs) as evidenced by the dependence of 3 on the emitter perimeter/area (P/A) ratio. This becomes an important factor when trying to maintain a high current gain on scaling down devices below the submicron level [201. 4. Extremely low contact resistance, which is required for high-speed devices, can be achieved on InGaAs. This is often difficult to do on GaAs and AIGaAs[21]. 1.2 Importance of the Composite-Collector HBT The most important disadvantage of InP-based single heteroj unction bipolar transistors (SHBT) is the narrow bandgap material (InGaAs) in the collector. Because of the narrow bandgap of InGaAs (0.75eV), there will be a high impact ionization rate, which results in a low breakdown voltage. The low breakdown voltage can be improved by replacing the InGaAs with wider bandgap material such as InP (1.35eV), as in the case of double heterojunction bipolar transistors (DHBT). But the disadvantage of this arrangement is the fact that there is a conduction band off-set at the base-collector junction which can cause electrons to be reflected back into the base and contribute to the base recombination. This 4 ultimately lowers the device gain. By arranging the collector to be composed of two layers (InGaAs and InP), it is possible to make sure the impact ionization takes place in the wider bandgap material (InP) and at the same time the conduction band off-set does not limit the flow of electrons from the base to the collector. This is the essence of the composite- collector HBT (CCHBT). 1.3 Overview In Chapter 1, we have briefly stated the general advantages of the Hi3T over the BJT. The advantages of the CCHBT, especially for power applications, and the preferred material system have been identified. In Chapter 2, the design considerations for the CCHBT are discussed and the wafer specifications for the first fabrication attempt are presented. In Chapter 3, the model for analysis of the CCHBT is presented in detail. In Chapter 4, the growth problem encountered in the first run and the modification for the second fabrication run is discussed. The fabrication procedure of the HBTs is also described in this chapter. In Chapter 5, the experimental measurements made on the devices are presented and the experimental data are compared with theoretical values calculated from our model. Finally, conclusions and recommendations are presented in Chapter 6. 5 Chapter 2 Wafer Design Considerations for the Composite-Collector HBT (CCHBT) The prime step toward the realization of any ITT-V compound semiconductor device is the design of the wafer. The selection of the HBT layers imposes an intrinsic limit on the way the device is going to perform in a practical situation. In this crucial step the objective is to select the appropriate values for the thicknesses and doping concentrations of the layers and, in addition, the materials for the layers. Furthermore, the way the actual device characteristics agree with the intrinsic behavior depends on both the HBT’s layers and the technology used in their production. In this chapter we address some of the important issues that have to be considered very carefully for optimizing the HBT structure for a particular application. The basic layer structure for the HBT is shown in Figure 2.1. L(((((((((((cci.(((((((((((C(cX(((cxccccc Emitter CapI Emitter Capil Emitter Spacer Layer Base Collector I __________________________________ Etch S Layer Collector II Suboollector Substrate Figure 2.1 The basic structure of a composite-collector HBT. 6 The ultimate aim of UBC’s InP/InGaAs project is to integrate HBTs with a semiconductor laser on a semi-insulating InP (S.I. InP) substrate. Although we are not concerned with the actual integration of devices in this thesis, it has to be kept in mind when optimizing the HBT structure. This leads to a number of contradictory requirements that have to be met by the HBT, e.g.: • The HBT has to have relatively high speed to be able to modulate the laser. • It has to have sufficient gain to drive the laser. • It has to have a high breakdown voltage to allow a wide operating range. 2.1 High Breakdown Voltage and High Speed in an HBT The basic requirement for a high-speed n-p-n HBT is to reduce the overall transit time taken for electrons moving from the emitter to the collector. Two major figures of merit which are used widely to measure high-speed response of an HBT are the current gain cut off frequency (fe) and the maximum frequency of ocscillation (fmaz), which are expressed as follows [22]: 1 (1) 27r(TE + TB + TSCR + Tc) frnax = (2) where -r is the emitter-base charging time, TB is the base transit time, TscR is the transit time across the collector space charge region (SCR), ‘i-c is the collector charging time, RB is the base resistance and Cc is the base-collector junction capacitance. As we can see from (1) the total time taken to travel from emitter to collector consists of four major individual components. For instance, the base transit time can be reduced by 7 just reducing the base layer thickness. But the reduction in the base thickness is going to increase the base resistance which degrades fmax. This is because the time delay associated with charging the base-collector capacitance depends on the base resistance. But no matter how small the base transit time is, there is not going to be a major reduction in the total transit time, unless there is an equal attempt to reduce all other components which contribute to the total time delay. One of the main limitations for HBTs is the collector transport. There has been a great deal of interest in recent years to modify the traditional n — collector in order to achieve the optimum design for the collector configuration [23,241. One can, of course, simply reduce the thickness of the collector-base space charge re gion (SCR) to improve the carrier transit time across the collector. In practice, however, the collector-base SCR thickness is limited by the base-collector breakdown voltage. The break down voltage is inversely proportional to the collector doping concentration and proportional to the square of the collector-base SCR thickness. Therefore it is relatively common to use a low-doped collector layer. However low collector doping concentration is not very desirable, because it results in base-push-out effects at large current levels, and the resulting wide SCR increases the distance that electrons have to travel. Therefore a low doping concentration will decrease the device speed. The actual collector thickness, as opposed to the SCR thickness, must also be large so that the resistance (Re) of the collector is kept low. This resistance appears in the collector capacitance charging time constant (Tc = RC). 8 2.2 HBT Structure 2.2.1 Emitter Configuration As we have mentioned earlier, one of the components of the cut-off frequency (ft) is the emitter charging time ‘re. The idea behind designing the emitter is to keep this value as low as possible and at the same time maintain a sufficiently high gain. This charging time basically depends on the time delay associated with the emitter and collector capacitances which are being charged through the emitter dynamic resistance, i.e. TE rC (3) where C = CEJ + CCJ i.e., the sum of the two junction capacitances. To be accurate the emitter series resistance should also be included, thus the emitter charging time becomes [25] r(Cj + Cj) + RECCJ (4) where RE Emitter resistance. CEJ Emitter junction capacitance. Cj Collector junction capacitance. re Emitter-base junction dynamic resistance. In our case RE is the resistance of the undepleted portion of the InP emitter layer, plus the resistances of the emitter caps and the metal contact. The large bandgap of InP ( 1.35eV) means that contacts will have a high ohmic resistance, unless the emitter layer is doped to such an extent that the electrons can easily tunnel through the barrier between 9 the metal and the InP layer. Because of the large bandgap, the doping concentration of the emitter has to be very high (of the order of 1 x 109cm3 [26, on p. 220]) before there is a significant tunneling effect. In practice, a high doping concentration is not desirable since it creates a large capacitance at the E-B junction, which slows down the device. Hence to over come this obstacle, we have to design a multi-layer emitter with two or three layers. The layer under the metal contact has to be InGaAs because it has a small band gap ( 0.75eV), and, by making the doping concentration high enough, the metal placed in contact with the surface will result in an ohmic contact. Once we try to grow this layer Ef Figure 2.2 Conduction band profile for a two-layer emitter at equilibrium. on top of a moderately-doped layer of InP, there will be a barrier to the flow of electrons, as shown in Figure 2.2. Again this barrier creates an unwanted resistance. The solution to this problem, which is often implemented in practice, is to insert between these two layers a third layer which has the same high doping concentration as the InGaAs layer, with the result as shown in Figure 2.3. The values of doping concentration and thickness for each - - - flt InGaAs N-InP 10 layer used in this work were chosen by combining features from earlier devices [27,281. They appear in Table 2.1. 2.2.2 Base Layer Figure 2.3 Conduction band profile for a three-layer emitter at equilibrium. w2 TB = 2DB - - + n - InGaAs N-InP N-InP As far as the base layer is concerned there are two things that have to be compromised, i.e., the base transit time (TB) and the base resistance (RB). The time taken for electrons to travel the quasi-neutral base region can be calculated from the linear distribution of the excess electron density tapering to zero at the base-collector depletion region boundary. TB is given by [25] (5) where W Width of the quasi-neutral base layer. DB Diffusion coefficient of minority-carrier electrons in the base. 11 This implies that the shorter the base width, the faster the carriers traverse the base. But the thinner the base width the higher the resistance of the base layer. The base resistance RB in an n-p-n bipolar transistor may be reduced by increasing the p-type impurity level, because the decrease in majority carrier mobility, which opposes the direct effect of the increased doping level, is sublinear at high doping levels. The resultant reduction in charging time constant RBC is important for applications in high-speed electronics. In fact, base resistance is a critical issue in microwave bipolar transistors and was the motivation for the original heterojunction bipolar transistor (HBT) concept, in which the base doping could be increased somewhat without decreasing current gain /3. Recent work has shown that growth of high quality layers of InGaAs lattice-matched to InP with p-type doping 1020cm3 is possible [29,30]. The quality of these layers depends on the technology used to grow the layers and on the choice of p-type impurity atoms. In practice there are only three types of p-type dopant available; they are beryllium (Be), carbon (C) and zinc (Zn). 2.2.3 Spacer (Base Setback layer) For high-quality crystal growth, and for a well-defined device structure, a precise control over the diffusion of dopants during growth is needed. Usually, Si and Zn are used as n- and p-type dopants, respectively in the MOCVD method [31], which is the growth technology used in this work. It is known that the diffusion coefficients of Zn are quite high in both InP and InGaAs, especially at high doping concentrations and temperature [32]. Thus zinc tends to diffuse really deeply into the adjacent layers [33]. In an HBT structure, the shift of the pn junction position into the InP layer due to significant Zn diffusion leads to reduction in the emitter injection efficiency, and ultimately to a reduction in the current gain of the 12 device. The thickness of the spacer layer which absorbs this Zn diffusion is one of the most important design features in the structure of the HBT and it can mean the difference between a good and bad device, as we will see in the next chapter. Figures 2.4 and 2.5 illustrate the state of the conduction and the valence band respectively, in three typical situations, as simulated using LUMIN2. The dashed line illustrates the preferred situation, where some diffusion has taken place in the spacer layer, but the diffusion has stopped at the interface between the wide and narrow bandgap material (E-B junction). This results in an abrupt junction. If little or no out-diffusion occurs (solid line), the undoped spacer increases the device base resistance and also probably allows more recombination to occur via electrons trapped in the notch or well in the conduction band. On the other hand if there is diffusion through the spacer and into the emitter (dotted line in Figures 2.4 and 2.5), the movement of the junction into the emitter would drastically reduce the emitter injection efficiency, since the emitter junction would be formed in the wider-gap emitter layer and the base would contain a P-pt heteroj unction. The undesirable effect is that the asymmetry in the barrier height is removed, i.e., both electrons and holes see the same barrier. This is something which has to be avoided by all means, because it prevents the base layer from being highly doped. Therefore the thickness of the spacer layer has to be right. If it is too thick there will be some undoped layer after out-diffusion from the base and if it is too thin the valence band offset will disappear. Since the wafers were purchased from EPI, we followed their suggestion of a thickness of O.O8m for a base doping of 1.5 x 109cm3. 2 LUMIN is a 2—D drift-diffusion model, developed at NRC, that solves Possion’ s equation self-consistently with the continuity equation. 13 ci) > c,) ci) w 0 C-) ci) w ci) > c,) ci) w C . 0 ci) w Figure 2.5 Valence band profile simulated by LUMIN at equilibrium. The setback layer lies between 0.3 and 0.308#m, the emitter is to the left and the base to the right. 1.2- 1.0- 0.8 0.6 0.4- 0.2- 0.0- — -0.2- 0.20 ,/ /_ With No Diffusion in Spacer — - — With Some Diffusion in Spacer/ With Diffusion Beyond the Spacer / 0.22 0.24 0.26 0.28 Depth (jim) Ô.3d Figure 2.4 Conduction band proffle simulated by LUMIN at equilibrium. The setback layer lies between 0.3 and 0.308gm, the emitter is to the left and the base to the right. -0.1 -0.3 -0.5- -0.7- -0.9- —1.1 - -1.3- -1.5- — 0.20 With No Diffusion in Spacer — — — With Some Diffusion in Spacer ..‘ ‘.. ‘ / With Diffusion Beyond Spacer..” / / / 0.22 0.24 0.26 0.28 0.30 0.32 Depth [jim] 14 2.2.4 Collector Configuration Here we discuss the design of a collector configuration that will give a high breakdown voltage and at the same time keep the SCR transit time and collector charging time to a minimum. In order to compare a few known configurations, the thickness, doping and configuration of emitter and base were kept the same for all designs. All the collector configurations used for simulation had a thickness of 3450A and were either doped at 1 x 106cm3or undoped. Because the collector layer is usually undoped or lightly doped, it is not possible to make a good ohmic contact directly to it. The usual method is to use a subcollector layer which is thick and highly doped. This method allows good ohmic contact to be made to the collector. The thickness and doping of the layers are shown in Table 2.1. There are two important mechanisms of breakdown in silicon bipolar junction transistors, namely: punchthrough breakdown and avalanche breakdown. Punchthrough breakdown occurs when the reverse-bias collector-base voltage becomes so large that the collector-base depletion region merges with the emitter-base depletion region. This particular mechanism is very unlikely to occur in HBTs simply because the base layer is so highly doped that almost the entire depletion-region width exists on the collector side of the junction. Therefore the dominant breakdown mechanism is avalanche breakdown. When the reverse bias voltage applied to the collector-base junction exceeds some critical value, the reverse current rises rapidly with further increase in the applied reverse bias voltage. This rapid current growth or avalanche is caused by an impact ionization process. During this process an electron (or hole) gains so much energy from the electric field in the collector-base depletion region that, on colliding with a lattice atom, it can excite another electron from 15 the valence band into the conduction band. Newly-created carriers are in turn accelerated by the electric field and create new electron-hole pairs via impact ionization. The actual breakdown voltage of the junction is defined as the voltage at which the total number of carriers generated as result of this process leads to a large specified current. Layer Material Thickness(A) Cumu. Depth (gm) Doping(cm1) Emitter CapI n+ InGaAs 1500 0.150 1.0 x 109cm3 Emitter Capil N+ IflP 500 0.200 1.0 x 109cm3 Emitter N InP 1000 0.300 1.0 x 10’8cm3 Spacer InGaAs 80 0.308 UID Base p+ InGaAs 500 0.358 1.5 x 109cm3 Collector I x 450 0.403 x Collector II x 3000 0.703 x Etch Stop InP 100 0.713 1.0 x 109cm3 Subcollector InGa.As 4000 1.113 2.0 x 109cm3 Substrate S.I. InP 35Oitm — — Table 2.1 Layer specification for the HBTs used in this work. The two collector layers are further specified in Tables 2.2 and 2.3. The problem with the conventional collector design is that because there is a large electric field at the interface between the base region and the collector’s SCR (see Figures 2.6 and 2.7), electrons enter this region with an excess energy. The excess energy of the incoming electrons together with the rate at which their energy increases are two of the most important reasons that lead to a poor breakdown voltage in an HBT. For the purpose of this section, we have compared the electric field profile of other known collector designs to see which has the potential for high breakdown voltage. For simplicity, the simulation was performed 16 with a one-dimensional drift-diffusion model, LUMIN, which gave the band and electric field profiles across the entire structure. Although LUM1N is not capable of predicting the breakdown voltage caused by impact ionization, by examining the electric field profile in the SCR region of each design, it is possible to conclude which design will have the highest breakdown voltage. The collector structures that were considered are as follows: A. Conventional Collector (nInGaAs). B. Inverted Field Collector (pInGaAs). C. Launcher (pn InGaAs). D. Undoped Collector (U/D InGaAs). In the case of the inverted field collector [34,35] (Figures 2.8 and 2.9), a small electric field is created at the base-collector junction, so that the carriers enter the SCR with a lower energy than in the conventional design and travel deep into the region before they actually experience a gradual increase in the electric field. This design will improve the breakdown voltage. In the case of the launcher design [34,35] shown in Figures 2.10 and 2.11, there is a sudden increase in the carrier’s energy as they enter the SCR from the base region which is not any better than in the conventional design. The undoped collector (Figures 2.12 and 2.13) creates a low electric field at the base-collector junction and the magnitude of the electric field stays constant all the way across the SCR. Although the undoped collector has the potential for high breakdown voltage, it does not necessarily lead to the highest breakdown voltage that is possible for the InP/InGaAs material system. A low electric field in the SCR is one factor which can improve the breakdown voltage, but it is also possible to improve it by replacing the collector with a large bandgap 17 material like InP for which the impact ionization rate is much smaller than InGaAs. The disadvantage of having the conduction band offset at the base-collector is that it limits the flow of carriers across the junction. The carriers that do not have enough energy to overcome the barrier are going to recombine and result in an increase in the base current and a reduction of the DC gain as compared to a single heterojunction bipolar transistor (SHBT). Although double heterojunction bipolar transistors (DHBT) with high gain and large current-handling capabilities have been demonstrated [36,37], it is a common belief that such structures are inherently slower than SHBTs [36]. It has been shown that by inserting a delta-doped (thickness of 30A) region at the base-collector region of InPIInGaAs DHBTs, the effective barrier height at the base-collector interface can be reduced [38]. Again this can improve the low gain of DHBTs, but at the expense of lowering the breakdown voltage. There has been a lot of interest in composite-collector HBTs (CCHBTs) lately [39—43]. This sudden interest comes from the fact that by having a composite-collector one can retain the advantages of both the SHBT (high gain) and the DHBT (high breakdown voltage). The energy band profile and the electric field of such devices have been calculated by LUMIN and are shown in Figures 2.14 and 2.15. The collector is composed of two layers: an undoped layer of InGaAs and a relatively thick layer of undoped InP. In such a configuration it is rather important to get the thickness of undoped InGaAs layer correct. If this undoped layer is too thick the carriers that are traveling through it will gain enough energy to initiate impact ionization in the InGaAs layer, whereas the aim of the composite collector is to have impact ionization take place in the InP layer. If the thickness is too small the collection efficiency is degraded. The electrons are reflected back into the base to increase the bulk recombination 18 At Equlibrium Low-Level Injection Electron Quasi-Fermi Level Hole Quasi-Fermi Level 0.2 0.4 Depth [rim] Figure 2.6 Calculated energy band profile for conventional collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. E C) > - 0 0 a) LI C) •10 a) w Figure 2.7 Calculated electric field profile for conventional collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. > a) > 9 ci) C w C 0 .1 0 ci) w 1.0- 0.0 -1.0 -2.0- -3.0-— 0.0 0.6 0.8 1.0 Depth [jim] 19 > C) > ci) w .C) C) w 1.0 0.0 -1.0 -2.0 -3.0 — 0.0 At Equilibrium Low-Level Injection Electron Quasi-Fermi Level Hole Quasi-Fermi Level 0.60.2 0.4 0.8 1.0 Depth [tim] Figure 2.8 Calculated energy band profile for inverted field collector design performed by LUMEN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. 2 C) 0 0 1 ci) U- C) .1C) ci) w Figure 2.9 Calculated electric field profile for inverted field collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. Depth [urn] 20 Depth [j.tm] Figure 2.10 Calculated energy band profile for launcher collector design performed by LUIvIIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. E C) 0 0 ci) LL 0 4-0 1) w 1.0• 0.0 -1.0 -2.0 > ci) > ci) w 2 U) w At Equilibrium Low-Level Injection Electron Quasi-Fermi Level — — Hole Quasi-Fermi Level -3.0 I 0.0 0.2 0.4 0.6 0.8 1.0 Depth [pPm] Figure 2.11 Calculated electric field profile for launcher collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. 21 > ci) > ci) w 0 .10 ci) w Depth [jim] Figure 2.12 Calculated energy band profile for undoped collector performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. 1.0- 0.0 -1.0 -3.0-— 0.0 At Equilibrium Low-Level Injection — — — Electron Quasi-Fermi Level — — Hole Quasi-Fermi Level 0.2 0.4 0.6 0.8 1.0 2 C) > C 0 ci) U- C.) .10 C) w Figure 2.13 Calculated electric field profile for undoped collector performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. Depth [jim] 22 1.0 > > U) C w C C) ci) w Depth [jim] Figure 2.14 Calculated energy band profile for composite collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. 0.4 0.6 0.8 1.0 Depth [jim] Figure 2.15 Calculated electric field profile for composite collector design performed by LUMIN. Refer to “Cumulative Depth” column in Table 2.1 for layer locations. 0.0 -1.0 -2.0 -3.0 0.0 0.2 0.4 0.6 0.8 1.0 E 0 0 0 7- - ci LL 0 ci) w 6.0 4.0- 2.0- 0.0 -2.0 -4.0-— 0.0 At Equilibrium Low-Level Injection V 0.2 23 which ultimately means a poor gain for the device. The first devices of this kind were fabricated and tested by Feygenson et al. [39]. They studied the importance of the thickness of the undoped InGaAs layer, by varying it from lOnm to l6Onm. For layer thicknesses less than 4Onm poor collector efficiencies were observed, whereas for thicker layers the collector breakdown voltage was reduced. For 4Onm thick InGaAs and a 500nm InP layer, they obtained a breakdown voltage of more than 1OV with ft and frnax of 64 and 34 GHz, respectively [391. Taking these results into consideration, the chosen thicknesses of these layers are presented in Tables 2.2 and 2.3. The reason that 45rim was chosen is to allow for any possible out-diffusion from the base and we reduced the 500nm to 300nm to improve the speed. For the purpose of this thesis we fabricated a SHBT and CCHBT to compare their characteristics, as described in the following chapters. Material Layer Thickness (A) Doping (cm3) Collector I InGaAs 450 UID Collector II InGaAs 3000 UID Table 2.2 Collector configuration of SHBT. Material Layer Thickness (A) Doping (cm3) Collector I InGaAs 450 U/D Collector II InP 3000 U/D Table 2.3 Collector configuration of CCHBT. 24 Chapter 3 Composite-Collector Heterojunction Bipolar Transistor (CCHBT) Analytical Model Development In this chapter we develop an analytical model for the composite-collector HBT (CCHBT). In recent years there has been a number of analytical models which predict the electrical characteristics of single [44,45] and double [46] heterojunction bipolar transistors. But, as has already been discussed in previous chapters, the advantages of the two structures can be maintained by using a composite-collector structure. For this reason there is a need for an analytical model that can describe the behavior of the CCHBT. 3.1 Boundary Conditions for the E-B junction To evaluate the currents in an HBT, the excess carrier densities in the base must be known. The usual boundary condition at the base edge of the emitter-base SCR which is applicable to a homojunction cannot be applied to a heterojunction. This is because the large band spike leads to a violation of quasi-neutrality and also of the assumption of a constant electron quasi-Fermi level across the SCR in an npn device [47]. Instead there is a split in the quasi-Fermi level [47]. Under these circumstances, the excess carrier concentration at the edge of the base-emitter SCR can be computed by a current balancing technique in which the thermionic-tunneling current across the junction is equated to the drift-diffusion current in the base [45]. The electrostatics and the current transport at the base-emitter junction are unaffected by the nature of the collector. The equations for current transport across a heterojunction 25 (6) (7) have been derived elsewhere [25,48]. Figure 3.16 shows the schematic energy band diagram of a CCHBT in a non-equilibrium situation, where there is a wider bandgap material in the emitter and collector. By solving Poisson’s equation on either side of the E-B interface, an expression for the depletion region width on either side of the junction as a function of the the total potential can be obtained. The assumption is that there are no mobile carriers in the depletion regions. The potential across the depletion regions is then given by Ji J2 XflE XpB Xp X Figure 3.16 Energy-band diagram of an ideal abrupt N-p heterojunction at thermal equilibrium. V(x)=_-Y(’_+xflEx) for XnE<X<O2 qN (x2 V(x)=___--_xBx) for OXXpB 26 Defining VT as the total potential difference across the N-p heterojunction at the E-B in the non-equilibrium condition, VTVbiVaVbE+VbB = V(xnE) — V(xB) = q (NexE + (8) 2\ EE B J where V is the built-in potential across the emitter-base junction, Va is the applied voltage across the E-B junction and VbE and VbB are the portions of VT supported by the depletion regions in the emitter and the base respectively. Having a continuous displacement at the interface, EEE(O) = EBEB(0), gives NDXE=NAXPB (9) Substituting (9) into (8) and solving for XnE and XpTh we get 1/2 2VTE EB Nb XnE = qN(ENe + €BNb) 1/2 (10) 2VT€E€BNe XpB = qNb(fN + fBNb) (11) The total depletion width is WT = XnE + XpB 1/2 — 2VTEB(Nb+Ne) (12) — qNbNe(ENe + BNb) 27 From Ref. [25], the built-in potential is given by qV = EgB + zE — q(V + Vp) (13) where EgB is the energy bandgap of the semiconductor in the base, LECE = XB — XE is the conduction band offset, and V and V are the separations of the electron and hole quasi Fermi levels from, respectively, the conduction band in the emitter and the valence band in the base. V and V are also given by [44] “NCEqV = kTln \ nO /NVBNqV=kT1n1—j (14) \ PpO I where no and PpO are the equilibrium majority carrier concentrations in the emitter and base respectively. Combining (13), and (14) qVfB = kTln (NCBB) + (XB — XE) + kTln ( OPpO )NCBNVB = (XB — XE) + kTln () + kTln () (15)NE The model for current transport across the B-B heterojunction is based on the thermionic diffusion model of Grinberg et at. [45). The net electron current density injected from the emitter into the base under low-level injection (the difference between the J1 and J2 in Figure 3.16) is given by [25] JNE —qVIZESNE [floB (exp (VBE) — i) — (xPB)] I’ qJXEE)NE=7exPI\\— kT 28 where /EE is the electron potential energy barrier seen by electrons trying to “back-inject” into the emitter (see Figure 3.16), SNE is effective the junction velocity of electrons, Bo is the equilibrium electron concentration in the p-type semiconductor, VBE is the applied forward bias, iI (XpB) is the excess electron concentration at the depletion edge of the p-type semiconductor, -y is the tunneling factor to account for the transport of electrons through the narrow spike [25] and VE, the electron thermal velocity, is given by VnE/ kT (17) V 2rm where mE is the effective mass of electrons in the emitter. 3.2 Composite-Collector Heterojunction Bipolar Transistor Model The CCHBT is basically a double heterojunction device (DHBT) with an extra layer in the collector (see Figure 3.17). We propose to model the current transport in the CCHBT by using the regular expressions for the DHBT [46] and [48], but with modifications to account for the fact that the effective height of the “spike” at the base-collector interface depends on the width X of the additional layer in the collector. With reference to Figure 3.1 7a, in the case of a regular DHBT, the barrier LEc for the electrons passing into the collector is given by the usual expression [49]. With a new additional layer, the barrier LEc is reduced (see Figure 3. 17b). As the thickness of the layer increases, LEc diminishes and eventually disappears (see Figure 3. 17c). Thus, over some range of thickness )., zE110 will vary from a maximum value to a minimum value of zero. The objective of the next subsections is to find the A dependence of 29 The resulting expression for LEc will then be used in the equation for the collector current in the Ebers-Moll relations as previously derived for a regular DFIBT [48], i.e. Undoped InGaAs Figure 3.17 Illustrating the conduction band profile and flow of electrons from the base to the collector under three conditions: (a) regular DHBT with no undoped InGaAs layer, (b) CCHBT with a narrow undoped InGaAs layer, and (c) CCHBT with a wide undoped InGaAs layer. = A11 (exp (VBE) — i) + A12 (exp (VBc) — i) + (18) = A21(exp (VBE) — i) +A22(ex (VBc) — i) +J (19) In order to derive the above equations, it is assumed that there is a uniform base doping, no compositional grading in the base, abrupt junctions between the narrow and wide bandgap (c) OA 30 qDE75E - A11 = — qySNEnBo LE tanh () A21 = —A12 = [ 2 /WB a = — cosh — LB \LB PE Pc are the equilibrium hole concentrations in the emitter and collector region, respectively; LE, L, LB are the minority carrier diffusion lengths in the emitter, collector and base, respectively; DE, Dc, DB are the minority carrier diffusion coefficients in the emitter, collector and base, respectively; WE, W, WB are the lengths of the quasi-neutral emitter, collector and base layers, respectively; R and JQ are the recombination current at the emitter-base junction and the recombination/generation current in the SCR of the base-collector junction, respectively; materials and infinite surface recombination velocities at the emitter and collector contacts. Then the A coefficients are given by 4Yn (i_ i)— a qSNEB 2qy SNEñBO LB [(qSNE + anyn) (i + — 4yqSNC) qLSNC — an(ynan + qSNE)qDcp ___________________________ C BA22 = ]Lctanh () — YnBO [(qSNE + anyn) (1 + 4yqSNc) — qLSNC qDUB yn= 2cosh (!:-) (20) (21) (22) (23) (24) where 31 3.2.1 Base-Collector Junction Consider an N-p-n InP/lnGaAs HBT in which a lightly-doped InGaAs layer is inserted between the base and collector layers, i.e., between x = 0 and x = \ in Figure 3.18. The spacer layer used in InP/InGaAs HBT fabrication are nominally undoped but in fact, have a residual n-type doping of about 104cm3.In deriving the model for transport in this region of the device, Boltzmann statistics are used throughout and the free carriers are assumed to reside and be transported primarily in the F valley. Emitter Base Collector Figure 3.18 Schematic illustration of HBT structure and the conduction band profile after including a lightly-doped layer in the Collector. Solving Poisson’s equation, assuming the lightly-doped InGaAs layer is fully depleted and its doping concentration is smaller than the collector doping concentration, the electric fields E(x) in the space-charge region are computed with the boundary condition that the -Xp 0 Lightly-Doped lnGaAs V2 V3 Xn Ef 32 electric field is zero at x and xi,, i.e., qN E(x)=—-——(x+x) —xx<O (25) E(x) = ___ — x) A <x <x (26) To find the electric field in the undoped InGaAs layer, we use the boundary condition E(O) = — so we can write the expression for the electric field as E(x) = — 0 x A (27) where Nb, N8, are the doping concentrations in the base, spacer and collector layers, respectively; e, Es, e are the permittivities for base, spacer and collector, respectively; Xp, Xn are the depletion-region widths in the base and collector, respectively; A is the spacer thickness; Integrating the electric field in each region to find the potential across each region, the total potential drop across the wide bandgap collector material is given by V3= x_A)2 Ax<x (28) The potential difference across the narrow bandgap material comprises a portion across the space charge region in the base and another across the lightly-undopecl InGaAs collector layer, i.e., v1=Yxi —xp<x<O (29) 33 v2 = _2 + (30) Therefore the total potential difference (VT) is VT—Vl+V2+3 = ,\)2qNb(x)) _2 (31) 2-j B j 2es and, in terms of the built-in potential, is VT Vb — VBG (32) where TVbBC is the built-in potential of the base-collector junction; VBC is the applied voltage across the junction; The built-in potential of the base-collector junction is the same as the built-in potential across an abrupt heterojunction. Therefore it is given by [25] qBC = (xB — xc) + in (NCCNb) + — in () (33) where XB, xc are the electron affinities of the base and collector, respectively; N, N are the effective densities of states in the conduction bands of the base and collector, respectively; iB is the intrisic carrier concentration in the base. 34 In order to have charge neutrality, and again assuming that there are no mobile carriers in the space charge regions, we have xNb = AN3 + (x — A)N (34) Now solving the (31) and (34) to find expressions for x and x, we find that -B+/B24AC = 2A (35) where A = qEs(NbNCCfc+NB) B = 2qN€sA(N€c—N3EB) C = (—qNSNCCEBECA + qNEBSA2 — 2VTNCCBSG) and x can be written, from (34), as = x1,N — A(N8 — N) (36) The expression for the electron current at the base-collector junction of a DHBT structure is similar to that at the emitter-base junction and is given by - - ( (qVBC’\ ‘\ NC = qVCSNC Bo exp kT ) — 1) — n(x) SNC=7flcexP(_ kT) As we can see from Figure 3.17, in a composite collector, /Ec is a function of the lightly-doped InGaAs layer thickness (A). As this thickness increases from zero, the conduction band offset is moved away from the base-collector interface and therefore the 35 0.20 > ci) C) C w 0.10 0.00 0.0 Figure 3.19 The plot of Ec as function of A for the CCHBT device for VCB=O. barrier seen by the electrons (/Ec) is decreased. The expression for /Ec as a function of A follows from Figure 3.16 and equations (29, 30) ZE - q(VT - V3) Ecc_q((+xpA’ _iYA2’ (38) j 2 j where /Ec is the conduction band offset at the base-collector junction. Plots of /Ec,Ic, and ‘B vs A are shown in Figures 3.19, 3.20 and 3.21. The emitter area, used to convert J and JB to current was 40 x 40tm2. To be consistent with the physics of the situation and the implementation of this relationship, the minimum value of ZEc is set to zero. It can be seen that a spacer thickness of at least 700A is needed if the effect of the conduction band spike on the current characteristics is to be eliminated. [A} 36 0.030 SHBT DHBT 0.000 I I I I 0.0 200.0 400.0 600.0 800.0 Figure 3.20 The plot of 1c as function of A for the CCHBT device for VBE=O.4V and VCB=O. 1.8- DHBT 0.0 200.0 400.0 600.0 800.0 2 [A] Figure 3.21 The plot of ‘B as function of A for the CCHBT device for VBE=O.4 and VCB=O. 37 3.3 Recombination Current in HBTs In this section we compute all components of the base current density JB in an InP/InGaAs ITBT for a wide range of bias conditions and various device parameters. The resulting information should be useful in determining the relative importance of the JB components and in accurately estimating the base current for predicting /3 in the fabricated HBTs. Recombination current can be divided into two main components: recombination current in the quasi-neutral base (JQNB), recombination current in the emitter-base SCR (JSCR). It has been assumed that the surface recombination current, the recombination current due to interface states at the E-B and the reverse injected current into the emitter are small and can be ignored. There are three type of recombination processes which are important in the device: Shocldey-Read-Hall, Auger and radiative. It has recently been confirmed [50] that the quasi-Fermi level splitting (/.Ef) at the emitter-base junction has to be calculated and be included into the diode expressions of the base-side recombination currents to avoid overestimating the gain of the device. As shown by Searles and Pulfrey [511, these currents can be written as Neni, LE — NratVbi NratVBE — JSRH,B Cs exp q exp qkT kT _______ BE JSRH,E Cs exp q— 2.Irorpo 2kT 2 VBE—/Ef JAtig,B Csn:&CpONb exp q kT (40) 2 VBEJAug,E CsnjCnoNeexp q- 38 o1 0 Collector Current - - - Quasi-Neutral Base — — SRH Current in the E-B junction -210 [E—E]Auger Current in the E-B Junction — )I( )< Radiative Current in the E-B Junction Z -‘ TotalCurrent s,-’ 1010 Base-Emitter Voltage [V] Figure 3.22 The various components of base current for an emitter area of 40 x 40pm2. 2 qV2 VBE —JRad,B CsnB(l — Nrat)1exp q kT (41) 2 qV2 VBERad,E CsnieBNratexp q- where TnO and ‘rpo are the Shockley-Read-Hall carrier lifetimes and Cs = kTq(1 Nrat)Vbj N — Nb (42)ra — Nb + Ne zE = + EE where LSE is the discontinuity in the intrinsic Fermi level at the interface, /.1Eg is the difference between the bandgaps in the emitter and the base, and is the permittivity. The Auger 39 and radiative coefficients, together with the minority carrier lifetimes used in this work are discussed in Section 3.4. The various recombination current components for our devices have been plotted in Figure 3.22. Searles et al. [51] have shown that in typical AJGaAs/GaAs devices SRH recombination in the emitter side of the SCR is the dominant component of the base current at low values of VBE, and at high VBE the dominant component becomes the quasi-neutral base recombination. But as we can see from Figure 3.22, the dominant component of base current over almost the full range of VBE in our devices is the quasi- neutral base recombination. This is because of the very short lifetime (5Ops) that we have used for the minority carriers in the base. Such a value is appropriate because it gives an excellent fit to the gain in InPfInGaAs HBTs which have been used in front-end receivers [52], and which have been successfully modeled by UBC’s device analysis program [53]. 40 3.4 Material Parameters of Ini_GaAsPi_ Although this thesis is concerned mainly with InGaAs in the form ofIn053Ga47As, other alloy compositions are of interest in electronic and optoelectronic applications. In fact, the ternary compound InGaAs can be recognized as a special case of the quaternary compound IniGaAsPi_, in which the bandgap can change all the way from 0.75eV to 1.35eV. This energy range corresponds to a wavelength range of 0.91—1.65gm, which covers the 1.3—1.6gm band which is suited to fiber-optic communication systems on account of the low dispersion and attenuation in typical fiber-optic cables in this wavelenght range. To model devices, such as HBTs and lasers, which would be used in transmitters for such systems, it is necessary to know how various electronic and optical properties vary with the mole fractions x and y. As such information is not generally available, we present here a mixture of data from the literature and computed values which should serve as a basis for model calculations for devices based on InGaAsP. Because the ratio of the total number of group Ill atoms to the number of group V atoms is unity, then the composition of any Ini_GaAsPi_ compound is uniquely represented by two parameters. One variable (x) gives the function of group III sites occupied by Ga and the other (y) gives the fraction of group V sites filled by As. In deriving many physical parameters 0 of the alloys when specific experimental data are unavailable, a linear interpolation scheme is generally adopted using the values of the related binary compounds. The parameters of a quaternary compound such as Ini_GaAsPi_y can be obtained from the respective values of the four binaries, InP, GaP, GaAs and InAs 41 in accordance with [54] O(Ini_xGaxAsyPi_y) = XYOGaAs + x(1 — Y)OGaP + (1 — X)yOInAs + (1 — x)(1 — y)Oip (43) If the parameters for the binary compound are available, they can be used in the above equation to estimate the parameter of the quaternary for a particular x and y. 3.4.1 Ratio of x to y for lattice-matching to InP Equation (43) can be used to work out the ratio of x to y at which the quaternary is lattice-matched to InP. Using the lattice constants for the binary compounds as given in Table 3.4, the relation between x and y for lattice-matching to InP can be summarized as Compound Lattice Constant(A) GaP 5.4512 GaAs 5.6532 InAs 6.0583 InP 5.8687 Table 3.4 Lattice constants of the four binary compounds [54]. — O.l896y — 0.4175 — O.0l24y (44) o.47y As we are interested only in lattice-matched systems in this thesis, we can express the parameters of interest solely in terms of “y”, and then use (44) to find the corresponding value of “x”. 42 3.4.2 Bandgaps and Electron Affinity The minimum bandgaps (eV) for the F, L and X valleys are given by Egr = 1.35 — O.’720y + 0.120y2 Egx = 5.04 + O.39y+ 0.149y2 (45) EgL = 3.14 — O.739y+ 0.149y2 The expression for the bandgap of the F valley is based on experimental data [55,p. 295], whereas the expressions for the X and L valleys are based on theoretical considerations [55,p. 295]. The conduction band discontnuity is assumed to be about 42% of the total energy bandgap difference between the emitter and the base [53]. Knowing that x(Ini.-o.47yGaoAsyPi_y) increases as y increases, we can obtain an expression for the electron affinity of Ini_o.47GaAsPi_ by taking a value of 4.50 eV for InP [53]. (Ini_o.47yGaAsyPi_y) = (InP) + 0.417[Egr(InP) — Egr(y)] (46) = 4.50 + 0.42(0.720y — 0.120y2) Bandgap narrowing due to heavy doping is also included in our model. But because of the lack of data in the literature, it assumed that the effective bandgap narrowing in the p-type InGaAs base is similar to that in p-type GaAs [56]. For the base doping density of about 109cm3used in this work, the bandgap narrowing is about 70meV. 3.4.3 Effective Masses The electron effective mass in lattice-matched Ini_o.47GaAsPi_ is given by Tflneff = (0.08 — 0.039y)mo (47) This is based on experimental data [54]. 43 The expression for the effective mass of holes in lattice-matched Ini_0.47yGao.AsyPi_y is obtained by interpolating using (43) and the data for the four binary compound semiconductors shown in Table 3.5. Compound mpeff (/mo) GaP 0.79 GaAs 0.62 InAs 0.60 InP 0.85 Table 3.5 Hole effective masses of the four binary compound semiconductors [54]. mpeff = (0.85 + O.28y — 0.04y2)mo (48) 3.4.4 Low-Field Low-Doping Majority Carrier Mobility The low-field, low-doping mobility of electrons in lattice-matched Ini_o.4’yGa7AsPi_ (cm2V1s)is obtained from a best fit to experimental data [SS,p. 202] (see Figure 3.23) and is given by = 1000(1.2085y3 — 0.6735y2+ O.15’78y + 0.4086) (49) The low-field, low-doping mobility of holes in lattice-matched ‘fll—O.47yGao.47AsP1 —y (cm2V’s) is obtained from a best fit to experimental data [S5,p. 2021 (see Figure 3.23) and is given by ILpO = 1000(0.8182y4— 1.4655y3 + 1.2127y — O.4’758y + 0.1412) (50) 44 250.0 0.0 I I 0.0 0.4 0.6 0.8 1.0 Arsenic Composition (y) Figure 3.23 The experimental data for majority carrier mobility inn and p-type lattice-matchedIni_o.47GaAsPi_, together with the best fit curves as a function of arsenic composition. 3.4.5 Doping Concentration Dependency of Mobility gao (51) = 1 + \ ref) 2 C) > .1 -Q 0 Best Fit Curve for Holes 200.0 A Experimental Data for Holes / pi ‘.7 J’ A 150.0 100.Oi i__•&._ ____ AA A A 50.0 - - - - Best Fit curve for Electrons fl Experimental Data for Electrons 0.2 10000.0 8000.0 6000.0 4000.0 2000.0 0.0 The dependence of the electron and hole mobilities ofIn0•53Ga47Asand InP on doping concentration is obtained from a best fit to experimental data (see Figure 3.24) [55,p.2O2], and is given by where NT is the doping concentration. Table 3.8 gives the corresponding values of c and Nref. ga is the low-filed low-doping mobility of the materials given in Section 3.4.4. 45 102 C) > .4- — — - lnP Fitted Data for Electron - .4 •lnP Exp. Data for Electron I V lnP & lnGaAs Fitted Data for Hole •InP & InGaAs Exp. Data for Hole — — InGaAs Fitted Data for Electron o •InGaAs Exp. Data for Electron10 14 ‘‘‘‘‘15 ‘‘‘‘‘‘16 ‘‘‘‘‘‘17 ‘‘‘‘‘‘18 ‘‘‘‘‘‘‘‘19’’’’ ‘010 10 10 10 10 10 10 Doping Concentration [cmj Figure 3.24 The experimental data for hole and electron mobility in ‘no.53 Gao.47As and InP as a function of doping concentration, together with the best fit curves. 3.4.6 Auger Coefficients of Ini_o.47GaAsyPi_ The Auger coefficient (cm6s1) for the process where a conduction band electron recombines with a heavy-hole, transferring it to the light-hole band can be obtained from a best fit to experimental data [57], and is given by log Tap = 3.5854y4 — 8.0619y3+ 7.9209y2 — ‘7.4853y — 4.8851 10—36 (52) GPO = Tap where Tap is the lifetime and G,o the Auger coefficient. The Auger coefficient (cm6s1)for the process where a hole recombines with a conduc tion band electron, and the energy is transferred to another conduction band electron can be obtained from a best fit to experimental data [57], and is given by log Tan 4.2933y — 9.9298y3 + 1O.514y2 — 9.3l2’Ty — 4.6021 1036 (53) Cf0 = Tan 46 where Tam is the lifetime and Co the Auger coefficient. Sometime after the above equations were developed, new data on Auger coefficients for lattice-matched InGaAs appeared in the literature [58]. The new values (C0 = Co = 5 x 10—30 cm6s) were used in this work. 3.4.7 Radiative Coefficient The radiative coefficient for lattice-matched Ini_o.47Ga ,AsPi_ (cm3s’) is ob tained by interpolating using (43 and 44) with the four binary compound values shown in Table 3.6. The expression is given by Compound B(cm3s) GaP 5.370 x 10—14 GaAs 7.210 x 10—10 InAs 8.500 x lO InP 1.260 x Table 3.6 Radiative coefficients of four binary compound semiconductors [59]. B = 1.26 x i0 — l.742y x 10 + 8.659y2 x 10_b (54) Although this equation might be useful in describing the general dependence of B on y, the experimental data on which it is based is old, and does not agree particularly well with more recent data for the specific compound 1n0•53Ga047As [53]. We use the more recent data (B = 4.0 x 10_li cm3s1)in this work. 3.4.8 Dielectric Constants The dielectric constant for lattice-matched In1 o.4iGao.47AsPi _ is obtained by interpolating using (43 and 44) and the four binary compound values shown in Table 3.7. 47 The expression is given by Compound GaP 11.1 GaAs 13.1 InAs 14.6 InP 12.4 Table 3.7 Dielectric constant of four binary compound semiconductors [54]. (12.4 + l.5y) 3.4.9 Shockley-Read-Hall Lifetime (55) The Shocldey-Read-Hall lifetime of electrons in lattice-matched Ini_o.47GaAsPi_ is assumed to be the same as has been measured for highly-doped GaAs [60], i.e.. TO = 5Ops (56) For holes there is not much data available so we take the value often quoted for moderately doped GaAs [61], i.e. Tpo = 2Ons (57) 48 A summary of material parameters is presented in Table 3.8. Parameters 1n053 Ga0 47As InP Lattice Constant (A) 5.87 5.87 Bandgap F [eVI 0.75 1.35 Bandgap X [eV] 5.50 5.04 Bandgap L [eV] 2.55 3.14 mneff 0.04 1 0.08 mpeff 0.61 0.85 /-n0 [cm2Vs] 11000 4086 ILpO {cm2Vs] 231 141 a(n — type) 0.56 0.68 (p — type) 0.71 0.71 Nref(fl — type) 1.1 x 1018 1.7 x 1018 Nref(p — type) 1.8 x i0’ 1.6 x 1017 C11 {cm6s] 5.0 x i0° 5.0 x i0° Cp [cm6s] 5.0 x i0° 5.0 x i0° B [cm3s] 4.0 x 10 4.0 x 10_li 13.9 12.4 r11j [s] 50 x 10_12 50 x 10—12 rj [sI 20 x i0 20 x Table 3.8 Summary of material parameters used in the CCHBT model. 49 Chapter 4 Non-Self-Aligned Method for Device Fabrication In this chapter the non-self-aligned technology used to fabricate HBTs is described. There were two batches of wafers, the first batch contained the two wafer structures with the specifications presented in Tables 2.1, 2.2 and 2.3 of Chapter 2. Hereafter the device structure with the InGaAs/InGaAs collector is called “InGaAs” and the one with the InGaAsIInP collector is called “InP”. The number after this designation indicates the wafer number. The second batch contained the modified version of the first batch’s wafers. Some of the reasons for the poor performance of the devices in the first batch, together with reasons for the modifications employed in the second batch, are discussed in this chapter. The wafers were fabricated in the clean room facility of the Microfabrication group in the Institute for Microstructural Sciences (IMS) at the National Research Council, Ottawa. 4.1 SIMS Plots of wafer InGaAs#1 and InP#1 and Modifications for InGaAs#2 and InP#2 In order to analyze the doping concentration profiles of the p-type dopant atoms in our wafers, it was necessary to look at the secondary ion mass spectroscopy (SIMS) plot of each wafer. This is a useful analysis tool to measure not only the p-type dopant concentration but also other specified atomic species as a function of depth in the wafer. Because it is a destructive technique, a small piece (0.5 x 0.5cm2) of the wafer was used for such measurements. The process consists of milling the surface with an ion beam and analyzing by mass spectrometry the released secondary ions. A cesium (Cs) ion beam was used for 50 C I 0 U Z 0 H 5 0 Z 0 U w C PROCESSED DATA 25 Jan 93 02 CANMET Metals Technology Laboratories FILE: 1030—2 Figure 4.25 The ln(l 13), As and Zn secondary ion count profiles for lnGaAs#l. The vertical dashed lines define the metallurgical base boundaries. C I Z 3 0 U Z 0 H C 4 0 U C C 102 PROCESSED DATA 25 Jan 93 02 CANMET Metals Tachnology Laboratontes FILE: 2530—I 2.0 iiSZn+0 As Zn Figure 4.26 The In(l 13), As and Zn secondary ion count profiles for metallurgical base boundaries. InP#l. The vertical dashed lines define the to4 50 101 DCPTh (microns) i0 101 1.0 OEPTH (microns) 51 1.5 0.5 ci) >% ci C -0.5 -1.5 0.0 1.0 Figure 4.27 Calculated energy band profile, before and after outdiffusion of Zn as performed by LUMIN for InP#1 based on results of SIMS plot in Figure 4.26, and the original specification, respectively. Refer to “Cumulative Depth” column in Table 2.1 for layer locations, sulfur (S) and gallium (Ga), and an oxygen (Or) ion beam for indium (In), arsenic (As) and zinc (Zn). The SIMS plots for the first batch are shown in Figures 4.25 and 4.26, for counts of three atomic species, namely: indium, arsenic and zinc. The base metallurgical boundaries have been marked on the graphs. As we can see, in the case of InGaAs#1 the Zn atoms have diffused deep into the emitter and collector layers, which suggests the emitter-base junction is now formed in the wide bandgap material (InP) and the undoped collector has shrunk. As discussed in Chapter 2, these effects are very undesirable as far as the performance of the devices is concerned. The diffusion of Zn into the collector has a more severe impact on InP#l’s performance than InGaAs#l’s. This is because InP#1 is essentially a DHBT, so any out-diffusion of Zn into the collector, would create a barrier to the flow of electrons from the base to the collector. Figure 4.26 suggests that the entire undoped InGaAs layer and part of the InP layer in the collector of InP#l has been converted into p-type material. This certainly 0.2 0.4 0.6 0.8 Depth [urn] 52 creates a huge barrier at the base-collector junction which limits the flow of electrons. The InP#1 device has been simulated by LUMIN and the result is shown in Figure 4.27. The actual reason for the out-diffusion of zinc atoms into the emitter and collector of our devices is difficult to explain, but it has recently been shown by Kurishima et al. [33], that high doping (> 109cm3of Si) in the subcollector of InP/InGaAs heterojunction bipolar transistor structures is responsible for high Zn diffusivity and associated broadening of the base layers. This abnormal redistribution of Zn has been explained by Deppe [62]. He has described that at high subcollector concentration there will be an increase in the number of column ifi interstitial sites. So when a p+.doped layer is grown after it, the Zn atoms tend to diffuse into the subcollector to occupy the vacant interstitial sites. He has also shown that this out-diffusion can be reduced by decreasing the n-type doping concentration. In order to made sure the next batch of wafers resulted in operational devices, three modifications were made to the original specification, 1. The base doping concentration level was reduced from 1.5 x 109cm3 to 1.0 x 109cm3. 2. The spacer layer thickness at the emitter-base junction was increased from 80A to bOA. 3. The n-type doping concentration level in the subcollector was reduced from 2.0 x 109cm3 to 5.0 x 108cm3. These measures were taken to limit the out-diffusion of Zn into adjacent layers as much as possible. Figures 4.28 and 4.29 indicate that the Zn was confined into the base layer as hoped-for. The base metallurgical boundaries have been marked on the graph to indicate this. 53 0.5 2.0 2.5 DEPTH (micron.) Figure 4.28 The In(l 13), As and Zn secondary ion count profiles for lnGaAs#2. The vertical dashed lines define the metallurgical base boundaries. PROCESSED DATA CANHET M.t.1. TechfloloQy Lfloratoriea 19 May 93 02 FILE: 403—OA PROCESSSO DATA 19 May 93 02 CANHET Matala Tacflnolaoy L.boratortaa FILE: 7S3—OA P S 1 z 0 LI Z 0 I. C 4 0 1 0 U S S Aa Zn ‘OS 10’ 102 101 I, I Z 0 U Z 0 C 4 0 1 0 U S S 2.0 DEPTH (micron.) Figure 4.29 The In(1 13), As and Zn secondary ion count profiles for InP#2. The vertical dashed lines define the metallurgical base boundaries. 54 4.2 Process for the Fabrication of Devices A simple fabrication process technique was used to make the devices in order to maximize yield and minimize the time taken. The process sequence is outlined in the following sub sections. The process was developed by the Microfabrication group at NRC. 4.2.1 Scribing and Cleaning the Wafers The diamond scribe technique was used to cut the wafers into 1.2 x 1.2 cm2 pieces or tiles. The original wafers were 2 in diameter with primary and secondary flats. After cutting, the tiles were cleaned using the following procedure to remove dirt and grease from the surfaces. Two beakers of lOOml of trichloroethylene were put in an ultra sonic bath at 80°C for a few minutes. Tiles were first immersed in the first beaker for 2.5 minutes, and then transferred to the second beaker quickly (to avoid drying of the tiles) for another 2.5 minutes. During this time, two other beakers were prepared, one with lOOml of acetone and the other with lOOml of methanol, and kept outside of the hot-tub. After this the tiles were quickly transferred into the acetone beaker for 2.5 minutes, and then to the methanol beaker for another 2.5 minutes. Meanwhile, two beakers with lOOmi of isopropylalcohol were put in the hot-tub, first the tiles were transferred into the first beaker for 2.5 minutes, then into the second beaker for another 2.5 minutes. After that, the tiles were rinsed with deionized (DI) water for 5 minutes in a water cascade. In this cleaning procedure, trichloroethylene was used to degrease the surface while acetone was used to rinse off the trichioroethylene. Methanol was used to rinse off the acetone, isopropylalcohol to rinse off the methanol and, finally, DI water to rinse off the 55 isopropylalcohol. Now to remove the native surface oxide, firstly the tiles were dipped into a solution of HC1 : 1120 (1: 1) for 30 seconds and then rinsed with DI water for 1 minute. Second, the tiles were dipped into a solution of NH4O : 1120 (1: 10) for 30 seconds and again rinsed with DI water for another 1 minute. Each time the tiles were taken out of the DI water, they were blown dry quickly by nitrogen gas so as not to leave any stain on the surface. Finally, to minimize re-oxidation of the surface, the tiles were dehydrated by baking them on a hot-plate for 2 minutes at 90° C. 4.2.2 Fabrication steps In the following paragraphs the steps that were taken to fabricate the non-self-aligned devices are described in detail. 4.2.2.1 Spinning the Photoresist The first step after the initial cleaning of the tiles was to apply positive photoresist (PR) using a spin-on technique. In order to get an approximately 1tm-thick PR film, a speed of 5000 rpm was used for 30 seconds. The tiles were then prepared for exposure by soft baking at 90°C for 1 minute. The state of the wafer after this step is illustrated in Figure 4.30. Next a mask aligner was used to transfer the mask pattern from a photolithographic mask to the photoresist. A schematic diagram of the masks is shown in Figure 4.31 and the optical micrograph is shown in Figure 4.32. Ultraviolet light (wavelength 365nm) was used to expose the PR for 2.7 seconds. The developing solution was diluted with DI water in the ratio of 5:1 DI:developer and the patterned tiles were immersed in it for 45 seconds at room temperature and then rinsed off in flowing DI water for several minutes. Step 2 in Figure 4.30 shows the results. After that the tiles were 56 given a hard bake at 120° C for 2 minutes to make the remaining PR layers more resilient to subsequent chemical etch solutions. 4.2.2.2 Wet Chemical Etch for the Emitter Mesa Having patterned the PR, the chemical solution for etching the emitter mesa was prepared. The emitter mesa was the same for both HBT structures. The wet chemical etch used for the InGaAs emitter cap layer was citric acid (50% solution by weight) and hydrogen peroxide with ratio of 3:1 (Citric Acid : 11202) for 3 minutes. This is a reaction-limited etch. The etch rate of this solution is about 1200A per minute and it stops at the InP layer of the emitter. After that the tiles were rinsed with DI water and blown dry with nitrogen gas. The mesa height, as measured using a DEKTAK surface profilometer, was found to be between 1.14 and 1.16tm. The wet chemical etch used for the InP emitter layers was a solution of phosphoric acid and hydrochloric acid (H3P04 : HEll) in the ratio of 3:1. This solution is a selective etch and does not react with the InGaAs layer of the base layer. The time of the etch was 2 minutes and violent agitation was employed to dissipate the hydrogen evolved in the reaction. Step 3 in Figure 4.30 shows the state of the wafer. 4.2.2.3 Photoresist Stripping and Cleaning After the height of the mesa was measured with the DEKTAK, the tiles were put into a PR-stripping solution for 10 minutes. Then the tiles were cleaned using the sequence of solvents listed in subsection 4.2.1. 4.2.2.4 Wet Chemical Etch for the Base Mesa PR was applied and developed as de scribed in subsection 4.2.2.1. First the base mesa was etched with the InGaAs wet chemical etch then followed by the InP wet chemical etch. The InGaAs part of the mesa etch thick 57 ness, in the case of the InGaAs#2 wafer was about 4500A, whereas in the case of the InP#2 wafer, the InGaAs thickness was about 1500A and InP 3000A. The corresponding etching times for InGaAs#2 and InP#2 were 5 minutes (for InP#2, 3 minutes to etch the InGaAs and 2 minutes the InP). Then the tiles were put into the InP chemical etch for 10 second to etch the InP etch-stop layer. Then the height of the mesa was inspected by the DEKTAK. Once the height was satisfactory, the PR was stripped and the tiles solvent-cleaned in preparation for the next step. 4.2.2.5 Wet Chemical Etch for the Collector Mesa Following a repeat of the previously- described procedure for PR application, exposure and development, and after using the sub- collector mask, the tiles were put into the InGaAs etching solution for 5 minutes. The sub-collector thickness is the same for both structures so the same etching time was used for both of them. The tiles were then stripped of PR and solvent-cleaned. 4.2.2.6 Pattern for Metal Contact With the three mesas established, it remained to apply contact metal to the emitter, the base and the collector. To pattern the three metal contacts on the surface of the tiles, a chlorobenzene lift-off process was used. In lift-off, the PR pattern is first established and the metal to be lifted-off is evaporated on top of the PR. In areas where there is no PR, the metal is also deposited. The tile is later immersed in a chemical which dissolves the PR (typically acetone). The metal deposited on clear areas remains. This procedure is facilitated by the fact that the metal forms a discontinuous film on the tile and no tearing of the metal is required. The metal is deposited in a highly directional manner and the edge of the PR pattern is shaped into a form which 58 incorporates an overhanging lip. The lip and the vertical nature of the pattern edge creates a space where no metal can be deposited. The recipe for the chlorobenzene liftoff process is as follows. The photolithographic process relies on contact image transfer. Good transfer requires that the mask be in intimate contact with the PR on the tile. Spinning PR creates an edge bead on the tile which inhibits good contact. The edge bead is a natural consequence of surface tension as PR spins off the edge. The bead can rise up very high above the surface of the tile which is a problem when working with smaller pieces. It was found necessary that the edge bead be removed using the following procedure. After applying the PR in the standard manner, the tiles were first exposed through a mask that was transparent only at the edges. The exposure was extended to 2 minutes after a soft bake of 70°C for 1 minute. This long exposure allowed a very short developing time of 10 seconds in the standard developing solution. The tiles were then rinsed in DI water and blown dry. After removing the bead, the tiles were patterned for metal contact in the same manner. Prior to developing the exposed pattern, the surface of the PR was made less sensitive to developer by immersing the tiles in the chlorobenzene for 6 minutes. The chlorobenzene was rinsed off with tricloro-trifloroethane (freon) for 15 seconds and the tiles were blown dry with nitrogen gas. The freon was further driven off the surface of the PR by baking the tiles on a hot plate for 45 seconds at 70° C. Development took place in the standard development solution but required 2—3 minutes of immersion and gentle agitation. Visual inspection of the pattern was made and, if required, more development was allowed to take place. Following the development of the PR pattern, the tiles were subjected to a 59 descumming procedure. The objective of this procedure is to remove any organic residue from the surface in areas where the PR has been removed. The descum procedure uses a microwave source to excite an oxygen plasma at low pressure ( 3 Torr). The excited oxygen species is effective in removing carbon-based molecules from the surface. For example, PR is removed at the rate of 50A/min in the system which was employed. After descumming the tiles, they were dipped into a solution of HC1: 1120 in the ratio of 1:1 for 30 seconds to remove any native oxide from the surface. Then the metal was evaporated on the tiles. The metal deposition system utilizes an electron beam to heat and evaporate the metals. A three-layer metal system was deposited, 500A of titanium (Ti), 750A of platinum (Pt) and 2500A of gold (Au). Figures 4.34 and 4.35 show the SEM micrographs of part of the device prior to lift-off in acetone. 60 Figure 4.30 The HBT fabrication sequence: (1) Spin the PR on wafer; (2) Pattern the PR and then develop it; (3) etch the emitter mesa; (4) Spin, pattern, develop the PR and then etch the base mesa; (5) Spin, pattern, develop the PR and then etch the collector mesa; (6) Spin, pattern, develop the PR and then evaporate Ti/Pt/Au. PR1 2 3[ 4 Emitter Base + Collector Sub-Collector Ti/Pt/Au 5r Y 6 61 Collector Mesa (Mask 3) Emitter Mesa (Mask 1) Base Mesa (Mask 2) _ Metal L Figure 4.31 Schematic diagram of the mask for the non-self-aligned HBT. Figure 4.32 Optical micrograph of a non-self aligned HBT (emitter area of 60 x 60 pm2). The mangification factor is 1044. 62 Figure 4.33 The SEM micrograph, showing the emitter mesa etch profile. Figure 4.34 SEM micrograph before lift-off, showing the metal lip resulting from the chlorobenzene process. 63 Figure 4.35 SEM micrograph before lift-off, showing another view of the same effect as shown in Figure 4.34. 64 Chapter 5 Results and Discussion In this chapter, the experimental dc characteristics of a single heterojunction bipolar transistor (SHBT) InGaAs#2, and a composite collector heterojunction transistor (CCHBT) InP#2 are presented. Tables 5.9 and 5.10 show the specifications for the wafers as given by the manufacture (EPI). Preceding the presentation of these characteristics is a brief description of the experimental setup for the dc measurements. The dc measurements include the current- voltage characteristics for both the common-base and common-emitter connections, Gummel plots for forward and reverse modes of operation, dc and differential gain. An investigation of the alloying temperature for the metallic contacts on the device performance is also presented. On each piece of wafer, a variety of transistors with emitter areas ranging from 1Otm x 1Om to 8Opm x 8Oim was fabricated. Some analysis and comparison of the devices based on the emitter dimensions is also presented in this chapter. Also we present results from the analytical model which was discussed in Chapter 3, and make, where pos sible, comparisons between the simulated and measured results. The device were characterized at the Device Physics group in the Institute for Microstruc tural Sciences (IMS) at the National Research Council, Ottawa. 5.1 Experimental Procedure To avoid wire-bonding, electrical contacts to the transistor were made by probing the device with tungsten needle probes mounted on fine xyz positioners. All of the electrical measurements were performed using an HP4145B Semiconductor Parameter Analyzer, with 65 Layer Material Thickness (A) Doping (cm3) Emitter CapT InGaAs (n+) 1500 1.1 x 1019 Emitter Capli InP (N+) 500 9.0 x 1018 Emitter InP (N) 1000 1.0 x 1018 Spacer InGaAs 100 UID Base InGaAs (p+) 500 1.0 x i019 Collectorl InGaAs (n-) 900 1.0 x 1016 Collectoril InGaAs (n-) 3000 1.0 x 1016 Etch Stop InP (N+) 100 5.0 x 1018 Subcollector InGaAs (N+) 4000 5.7 x 10 Substrate S.I. InP Table 5.9 The specification given by EPI for the InGaAs#2 HBT. Layer Material Thickness (A) Doping (cm3) Emitter CapI InGaAs (n+) 1500 9.7 x 10 Emitter Capli InP (N+) 500 7.8 x 1018 Emitter InP (N) 1000 1.3 x 1018 Spacer InGaAs 100 U/D Base InGaAs (p+) 500 9.8 x 1018 Collectorl InGaAs (n-) 900 1.0 x 1016 Collectorli InP (n-) 3000 1.0 x 1016 Etch Stop InP (N+) 100 5.0 x 1018 Subcollector InGaAs (N+) 4000 4.6 x 1018 Substrate S.I. InP Table 5.10 The specification given by EPI for the InP#2 HBT. 66 source measuring unit (SMU) triaxial cables, SMU1, SMU2 and SMU3, connected to the emitter, base and collector, respectively. The way the variables (VAR1, VAR2 and COM) were assigned to each metal contact depended on the intended electrical measurements. 5.2 DC Characteristics In this section the dc characteristics of the InGaAs#l, InGaAs#2 and InP#2 HBTs are discussed in detail. 5.2.1 Common-Emitter and Base Characteristics Figures 5.36 and 5.37 illustrate the common-emitter output characteristics of InGaAs#1, InGaAs#2 and InP#2. One of the most remarkable features that can be observed from Figure 5.36 is the fact that the common-emitter breakdown voltage (BVCEO) of InP#2 is approximately three times greater than that of InGaAs#2. This confirms the role played by the composite-collector structure in suppressing impact ionization breakdown. Figures 5.38, illustrates the offset region of the InGaAs#2 and InP#2 structures. As it can be seen, the collector current is negative and is equal (in the case of InGaAs#l) to or less (in the case of InGaAs#2 and InP#2) to the base current. Obviously in this region the collector injects more heavily than the emitter and leads to a negative collector current. It is also observable, from Figure 5.37 and 5.38, that the offset-voltage of InGaAs#1 is much larger than in the other two structures (InGaAs#2 and InP#2). The VCE offset voltage is approximately equal to the difference between the effective turn-on voltages of the two junctions. It basically means that at a certain value of VCE the two junctions are forward biased such that the injection of electrons from each junction into 67 the base is the same. This results in zero effective collector current. As we can see from Figure 5.38 the offset voltage is independent of base current. When InP#2 was operated in the inverted mode, a negligible offset voltage was observed, as shown in Figure 5.39, and the emitter current increased more gradually than in the forward mode. Notice also that the breakdown voltage of the device has deteriorated. Because of the high-doping concentration in the emitter, there will be a higher electric field which lowers the breakdown voltage. The devices in the inverted mode do not have any useful gain, as can be seen from Figure 5.40. This is because for both devices the base-collector junction is effectively a homojunction, for which the injection effeciency is very low due to the low doping density of the collector. A low output conductance and good device linearity are important in minimizing harmonic distortion. The differential output conductance of an HBT is defined as go = dI/dVcE IC/(VA + VCE), where the Early voltage VA is the 1c — VCE slope extrapolated to the VCE axis. The fabricated HBTs present excellent output characteristics compared to previously reported results from similar wafer structures [43]. The negative output conductance observed in AlGaAs/GaAs HBTs results from device heating and decreasing current gain with temperature. This was not observed in our InP/InGaAs devices presumably because of their better thermal dissipation, due to the good thermal conductivity of the InP substrate (0.44—0.455 W/cm K for GaAs [63, on p.24.7] as opposed to 0.66—0.7 W/cm K for InP [63, on p.S’7l]). The common-base characteristics of InP#2 show also an improved breakdown voltage over the InGaAs#2 devices (see Figure 5.41). Both devices show high output conductance in the common-base mode. 68 E •1 I Figure 5.37 Measured ouput characteristics of InGaAs#1 HBT for an emitter area of 60 x 80m2. The base current starts at 51zA and increases in steps of 20A. 6.0 4.0 2.0 0.0 0.0 4.0 8.0 12.0 Collector-Emitter Bias Voltage [V] Figure 5.36 Measured output characteristics of InGaAs#2 and InP#2 HBT for an emitter area of 60 x 80m2. The base current starts at 5A and increases in steps of 20A. — 1.60 E .1-’ 0.40 Collector-Emitter Bias Voltage [V] 69 Figure 5.38 Magnified offset region of the measured output characteristics shown in Figure 5.36. The base current starts at 5A and increases in steps of 20A. E •1- Figure 5.39 Measured reverse characteristics of InP#2 HBT for an emitter area of 60 x 80m2. The base current starts at 5iA and increases in steps of 2OzA. E .1 0 0 4- 0 ci) 0 0 Collector-Emitter Bias Voltage [V] 0.05 0.04 0.03 0.02 0.01 0.00 Emitter-Collector Voltage [V] 70 101 102 11004 Emitter Current 10 - - - - Base Current 10b0 - I I I 0.0 0.2 0.4 0.6 0.8 1.0 Collector-Base Voltage [V] Figure 5.40 Measured reverse mode Gummel plot of InP#2 HBT for an emitter area of 60 x 80jm2. 11.0- InGaAs/lnGaAs Collector _ s.c’ - - - - - lnGaAs/lnP Collector —;_=__________ --I . 7.0- / rrQ.V I/ 3.0- ,// 1.0- —1.0-- -1.0 1.0 3.0 5.0 7.0 9.0 11.0 13.0 15.0 Collector-Base Bias Voltage [V] Figure 5.41 The measured common-base characteristics of InGaAs#2 and InP#2 HBTs for an emitter area of 40 x 40m2. The emitter current starts at OA and increases in steps of 2mA. 71 5.2.2 Gummel Plots As already discussed in the previous chapters, the quality of the junctions in an HBT play a vital role in the performance of the device. This quality depends to a large extent on how abrupt the junctions can be made. This abruptness in turn depends on how well the p-type dopant can be confined in the base layer. One can extract a lot of information about the junction quality from measurements of the dependence of the logarithm of the collector (Is) and the base current (Ib) on the emitter-base bias voltage, i.e. from Gummel plots. Such plots were examined for the three fabricated devices. Figure 5.42(a) illustrates how the device was configured for a Gummel plot measurement by the HP4145B. This is usually done with the base and collector short-circuited or VCB = OV. The measured Gummel plots for the three structures are shown in Figures 5.43 - 5.45. ‘c (a) (b) (c) Figure 5.42 The set up for devices, measured by HP4145B; (a) Gummel plots, (b) Common emitter I-V characteristics, (c) Common base I-V characteristics. All three Gummel plots exhibit a low current non-ideal region. For the present devices, the current in this region is actually photo-excited current. This was established by workers at the Device Physics group at NRC, who showed that this current could be eliminated by 72 I___ 10 Collector Current - - - - - Base Current -1210 - I I I I I 0.0 0.2 0.4 0.6 0.8 1.0 Emitter-Base Bias Voltage [V] Figure 5.43 Non-alloyed Gummel plot (magnitude of current) of InGaAs#2 at Vcb=OV for emitter area 40 x 40jzm2. I ___ 10b0 - Collector Current 12 — — - Base Current 1 0 - I I I I 0.0 0.2 0.4 0.6 0.8 1.0 Emitter-Base Bias Voltage [V] Figure 5.44 Non-alloyed Gummel plots (magnitude of current) of InP#2 at Vcb=3V for emitter area 40 x 40m2. 73 1 02 1 4 — -6 C -8o 10 10b0 1012 Figure 5.45 Non-alloyed Gummel plot (magnitude of current) of InGaAs#l at Vcb=OV for emitter area 40 x 40m2. performing the measurements in the dark. The optically generated electrons are swept across the base-collector junction and out of the collector terminal by the electric field. At the same time, photogenerated holes are swept into the base, and are apparently of such magnitude that they cause a reversal of the base current from the direction it has at higher biases where recombination in the base dominates and is supplied by holes flowing into the base from the base contact. Structure flc InGaAs#1 1.82 2.73 InGaAs#2 1.02 1,13 InP#2 at Vcb=3 1.02 1.16 Table 5.11 The collector and the base current ideality factors for the three structures. The ideality factors were measured by taking the average value at five different points on the linear region of the Gummel plot. Emitter-Base Bias Voltage [V] 74 The ideal linear region of the Gummel plot is the region immediately after the base current reversal. In this region both the collector and the base currents increase exponentially with emitter-base bias voltage. The slopes of the curves yield the ideality factor of the base current (which gives information on the type of recombination which is dominant) and the collector current. These have been worked out and presented in Table 5.11. In order to explain the source of the ideality factor of the collector current in our HBT devices, it is informative to examine the general expression (58) for the collector current under low-level injection conditions [49]. A comprehensive analysis of the approach used to derive the expression (58) has been presented in Ref. [64]. The formulation neglects any hot-electron effects as it assumes that all electrons injected into the base are immediately thermalized, and that transport proceeds by diffusion. It is also assumed that the diffusion coefficient is constant across the base layer and that there is no base recombination. JNG = —qn2 exp (4y) — exp (4) (58) Nb TB+TEB+TBC where 1 TB=— w 1 TEB=Y VnE)NE TBC= - 1 59 VnCSNC (-ZEESNE = 7nE exp kT /-LEc SNC = 7nC exp kT The normalized junction velocity SNE can be represented by a simple exponential expression, 75 i.e. [65], * ( qVSNE = A exp — (60) SIC’ J where A* and s are strong functions of the doping concentration in the emitter and base and also a function of the conduction band offset. The values of A* and s for the devices considered here are 0.375 and 6.55 respectively. The three terms in the denominator of (58) can be viewed as being related to the transport of electrons across the three important regions of the device, namely: the E-B junction (TEE), the quasi-neutral base region (TB) and the B-C junction (TBC). This interpretation has been suggested before [49], and is convenient for assessing which part of the device is the bottleneck for the carrier transport. For example in a situation where we have a homojunction at the E-B, the LEE term in (59) disappears. This implies that the electrons are injected across the E-B junction with an infinte velocity, such that the limiting factor in (59) is the diffusion transport across the base (TB). This results in a collector current with an ideality factor of unity. Now in a situation where there is an abrupt heterojunction at the E-B interface, the electrons are injected across the junction with a finite velocity. It is possible that TEB becomes the dominant term in the denominator of (58) and the collector current then becomes thermionic/tunnel limited. From (58)-(60) we find that the collector current can be written as ib * ( VBE(1—)”\ JNC = —qjç--A exp q kT (61) i.e. the collector ideality factor is n = (1 — For our value of s 6.55, this gives 1.18. This is a little higher than observed experimentally. One explanation of the 76 difference could be that the spacer layer at the emitter-base junction is playing an important role in the current transport process. This possiblity is examined in Section 5.2.5. On the other hand the base ideality factor depends on the different base recombination mechanisms. The base current in HBT’s generally consists of four major current compo nents: (a) bulk recombination current in the quasi-neutral base region (b) Space Charge Region (SCR) recombination current in the emitter-base depletion region, (c) reverse injec tion current of holes into the quasi-neutral emitter (d) surface recombination current at the exposed extrinsic base surface. It is well known that the reverse injection current increases exponentially with the emitter-base bias voltage with an ideality factor of “-i 1, and both the space charge region recombination current and the surface recombination current often increase with an ideality factor of “.‘ 2. In homojunctions, the quasi-neutral base recombi nation current has an ideality factor of unity. However, in heterojunctions, the splitting of the electron quasi-Fermi level means that the ideality factor for this recombination current will be the same as that of the collector current. As we can see from Table 5.11 the base ideality factors and the magnitude of the base current (from Figures 5.43 and 5.44) of the two structures, InGaAs#2 (Vb = OV) and InP#2 (Vb = 3.OV) are very close to each other. Firstly this suggests that the dominant source of recombination is the same for both structures. Secondly, it appears that the conduction band offset at the B-C junction does not block the transport of carriers; if it did, the magnitude of the base current in InP#2 would have been larger than in InGaAs#2. Following the argument previously presented for the collector current ideality factor, in the case of thermionic/tunnel limited injection, we would expect that the base current would have the same ideality factor 77 as the collector current if the base current were dominated by neutral-base recombination. Figure 3.22 suggests that this is the major source of the base current in our devices, so a value of 1.18 for b is to be expected. This is very close to the value measured for InGaAs#2 and InP#2. The Gummel plot of InGaAs#1 shows a large shift along the positive direction on the VBE axis. This can be explained by looking at the analytical expression of the built- in potential of an abrupt heterojunction (15) in Chapter 3. This expression suggests that V(NP)62 > V(Np)1, > V(np), where the capital letters represent the wide bandgap and the lower case the narrow bandgap semiconductor, respectively. So as the junction is moved into the wide bandgap semiconductor, the built-in voltage becomes large too, and as a result of this there will be a positive shift on VBE. The measured dc (/3) and small-signal (hfe) current gains versus the collector current are plotted in Figures 5.46-5.47 for the InGaA#1, InGaAs#2 and InP#2 HBT structures. Notice that the InGaAs#2 device exhibits a gain of more than unity for the collector-current range of 10—10 A to 10—2 A, which could be very desirable from the circuit designer’s point of view. 78 102 \ \ — — —— / / \ ——— / 101 100 — — InP#2 at VCb=OV InGaAs#1 at VCb=OV - - - InGaAs#2 at VCb=OV 1 O’ - ‘-6 ‘“‘““‘-4‘‘“‘‘-2’10 10 10 10 1 [A] Figure 5.46 The measured dc gain / versus the collector cuffent in InGaAs#1, InGaAs#2 and InP#2 for an emitter area of 40 x 40m2. 210 --—- 7— 1 /10 / z -C r__J_f 100 I V — - - InP#2 at VCb=3VInGaAs#1 at VCb=OV- - - InGaAs#2 at VCb=OV10 -10’’’’’’ ““-8‘‘‘‘‘‘-6‘‘“‘-4‘‘‘““‘‘‘““‘-2’’’”10 10 10 10 10 1 [A] Figure 5.47 The measured small-signal gain life versus the collector current in InGaAs#l, InGaAs#2 and InP#2 for an emitter area of 40 x 40m2. 79 5.2.3 Parasitic Resistance Measurement In order to obtain good high-frequency performance from an HBT, it is critical to make reliable, low resistance ohmic contacts to the active layers. The penetration depth of the conventional ohmic system Au/Ge/Ni after annealing is on the order of l000—4000A in GaAs [66,67], which is not suitable for HBTs with thin base layers. HBTs also need a base metal system that can survive annealing and dielectric deposition/passivation without degradation. The alternative system for contacting a thin layer is the TiJPtJAu system which wrr•••• _ - Epitaxial Layer -- Mesa Edge %Z55i Metal Figure 5.48 Transmission line pattern used to experimentally determine the emitter, base and collector contact resistances. The area of each pad is 60 x 80cm2 and the separation between them starts at 2pm and increase in steps of 1pm. has a shallow penetration depth in InGaAs and low resistance ohmic contact, even without heat treatment [68]. We have found that alloying can lower the resistance of the contact, but it also degradates the performance of HBT. The basic technique used to measure the resistance of ohmic contacts employs a test pattern composed of differently spaced ohmic contacts [26,p. 234], as illustrated in Figure 5.48. Ohmic contacts were formed on the InGaAs cap of the emitter, base and the subcollector collector layers and separated by a distance L. The contacts have a width W and the pattern is isolated to restrict the current to flow only across the distance L. The resistance between two 80 such contacts consists of the two contact resistances plus the resistance of the semiconductor layer between the two contacts (see Figure 5.49). Hence, the total resistance is [68] RT = 2R + R2 L (62) where R is the resistance between the metal and an imaginary plane at the edge perpendicular to the metallization; R8h2 is the sheet resistance of the semiconductor layer between the two metal contacts; Figure 5.49 Equivalent resistor network representing the end effect and the contact resistance. The total resistance (RT) is measured experimentally for each value of spacing (L) and then plotted on a linear scale graph. Figures 5.51-5.53 illustrate such graphs for the emitter, base and collector mesas. The slope of the line gives the value 4* and the intercept with the R axis gives the value 2R. The results for non-alloyed contacts are presented in Table 5.12, using Figures 5.51 - 5.53. As we can see, the contact resistances of the emitter and the collector metal contacts are significantly smaller than that of the base. This is mainly due to the fact that electrons have Rsh2 81 higher mobility than the holes. As we can also see by comparing the results of non-alloyed Figure 5.50 Plot of total contact to contact resistance as a function of L to obtain transfer length and contact resistance values. and alloyed samples (see Tables 5.12 - 5.13), 30s alloying improves the results considerably. Although more data is necessary to reach a conclusive outcome, it appears that the optimum temperature to alloy Ti/Pt/Au on InGaAs layers, is about 300° C. It was found that at an alloying temperature of 350° C, the output characteristic of devices was affected severely (the output plot for 300°C is not shown because no significant change was observed). This can be seen from Figure 5.54. The breakdown voltage of InGaAs#2 has deteriorated, such that BVCBO is only about 3.5V. This also means that these devices cannot operate normally at temperatures as high as 350°C. The actual reason for this severe effect on the output characteristic, as the temperature is increased, is not obvious yet. More experimental data is needed to understand what is actually happening. Emitter Base Collector Rc [0] 0.43 35 0.32 Rs2 [O/D] 15.2 1200 7.2 Table 5.12 Summary of the resistance measurements of a non-alloyed contact of InGaAs#2. RT RSH2L + w L 82 Emitter Base Collector Rc [l] 0.21 9.88 0.25 R5h2[cl/D] 16.8 1343 9.1 Table 5.13 Summary of the resistance measurements of an alloyed (at temperature 300°C) contact of InGaAs#2. Emitter Base Collector Rc [el] 0.28 33 0.30 Rsh2 [11/0] 16.0 1530 8.64 Table 5.14 Summary of the resistance measurements of an alloyed (at temperature 350°C) contact of InGaAs#2. 83 8.0 — Non-alloyed 7.0 — — — Alloyed at 300°C TE 0.0 5.0 10.0 15.0 20.0 25.0 Pad Separation [j.tm] Figure 5.51 Resistance measurements data for the emitter contact (area = 60 x 80jim2). 500 - — Non-alloyed 400 - — — — Alloyed at 300°C 300 - — Alloyed at 350°C 200- 100 0 I I I 0.0 5.0 10.0 15.0 20.0 25.0 Pad Separation [jim] Figure 5.52 Resistance measurements data for the base contact (area = 60 x 80m2). 84 5.0 Non-alloyed i° :zo%z 0.0 5.0 10.0 15.0 20.0 25.0 Pad Separation [jim] Figure 5.53 Resistance measurements data for the collector contact (area = 60 x 80m2). 9.0 7.0- 5.0- 3.0- ‘ o 1.0 - -1.0- — -1.0 1.0 3.0 5.0 7.0 Collector-Base Voltage [V] Figure 5.54 The measured common-base Ic vs VCB characteristics of InGaAs#2 HBT for the cases of alloyed and non-alloyed contacts, for emitter area 40 x 40m2. The emitter current starts at OA and increases in steps of 2mA. 85 5.2.4 Emitter Contact Geometries Returning now to the unalloyed InGaAs#2 devices, it was mentioned in subsection 5.2.2 that the experimental results show evidence of the base ideality factor sightly more than unity. As it was explained, this could be due to recombination in the emitter-base depletion region, at the emitter-base interface, or it could equally be due to surface recombination at the surface of the base between the emitter stack and the base contact. As the latter mechanism is often important in A1GaAs/GaAs devices, some information on its relevance to InPITnGaAs devices was sought here by performing measurements on devices with different emitter contact geometries and different base-emitter spacing. Results are shown in Figures 5.55 - 5.59. The emitter length/width dimensions in microns are given beside each data point. Except where noted, the emitter-base separation was 5tm. In all cases, currents were recorded at Vbe = O.4V and Vb = OV, i.e., well-away from the region where high-level injection or series resistance effects may be important. To analyze the results, note that the total base current can be conveniently represented as the sum of the current in the intrinsic base-emitter junction ‘BE, and the current at the extrinsic surface ‘Bs. Dividing each of these by the emitter area gives a base current density of = JBE + JBSLBE (63) where LBE is the emitter-base separation and WE is the emitter finger width. Note that the emitter finger length affects ‘BE and ‘Bs equally, and so does not appear in (63). In light of (63), changing the emitter area should only have an effect through the associated change in emitter width. Thus, the decreasing JB shown in Figure 5.58 does indicate the presence of 86 101 20/80 10° p pj 10 10 Emitter length [tim] Figure 5.55 Dependence of collector and base current density on the emitter length in InGaAs#2. significant surface recombination current. Contrarily, changing the emitter width only, with the emitter length held constant at 2Oim, appears to indicate otherwise, as the base current is practically constant (see Figure 5.57). Further evidence that surface recombination is not significant comes from Figure 5.56, which shows a near-constant JB as LBE is increased. The interpretation of the results is further confused by the effect of changing the emitter length LE. This should have no effect, but Figure 5.55 indicates otherwise. Clearly more data is needed before the question of the importance of the surface recombination can be resolved. However, even though the current gain does increase slightly as the area/perimeter ratio is increased (see Figure 5.59),it does appear that the surface recombination current is not as prominent as in A1GaAs/GaAs HBTs, presumably because the surface recombination velocity of InGaAs is much lower than that of GaAs or A1GaAs [201. 87 101 c.J E 0 () >% 100. I p pj 101. I I I 3.0 3.5 4.0 4.5 5.0 Emitter-Base Metal Separation [jim] Figure 5.56 Dependence of collector and base cuffent density on the base-emitter separation for devices with emitter area = 60 x 80m2. 101 20/60 0I40 20/80 > C “-‘ U .1 C p pj D’ 101 I I 40.0 50.0 60.0 70.0 80.0 Emitter width [jim] Figure 5.57 Dependence of collector and base current density on the emitter width. 88 101 10/10 E C) > 80/80 °) 10°-C ci) U .1 C c_) 101 102 ió io Emitter Area [jim2] Figure 5.58 Dependence of collector and base current density on the emitter area. 102 101 0.0 5.0 10.0 15.0 20.0 Emitter Area to Periphery ratio [jim] Figure 5.59 The dc current gain at Vbe = O.4V vs emitter area to periphery ratio (In case of InGaAs#2). 89 5.2.5 Model and Experimental Comparison In this section, the analytical model derived in Chapter 3 is used to simulate the Gummel plots of the InGaAs#2 and InP#2 HBTs with the layer specifications given in Tables 5.9 and 5.10. Consider the Ebers-Moll equations, as derived for the emitter and the collector current densities in Chapters 3. JE A11 (exp () — i) + A12 (exp () — i) + (64) = A21 (exp (VBE) — i) + A22 (exp (VBc) — i) + JQ (65) JREB is the sum of the recombination current densities in the emitter-base SCR. So the base current calculated from JB JE — J includes both the recombination in the quasi- neutral region of the base and ,JREB The recombination/generation (J) current in the base-collector SCR is usually small enough to be ignored. For the purpose of our simulation this current was extracted from the measured Gummel plot and used in our simulation. As it can be seen from Figure 5.61, the Gummel plot of the InGaAs#2 device, there is a large discrepancy between the measured and simulated data. The low collector- current ideality factor from the measured data suggests that the collector current is limited by diffusion in the base layer. Whereas the model data suggests that it is limited by thermionic/tunneling across the E-B junction. As discussed in Section 5.2.2, the discrepancy is probably due to the fact that conduction across the emitter-base junction is influenced by the presence of the spacer layer. The effect of the spacer layer can be taken into account by assuming that its main influence on carrier transport is via a modification of the barrier 90 height /EUE (see Figure 5.60). The effect is exactly analogous to that developed in Section 3.2.1 to model the reduction in the collector barrier height /Ec due to the presence of the n InGaAs layer in the composite collector structure. Thus, with reference to Figure 5.60 and from (38), the barrier height LEE is related to the spacer thickness 6 by = ZECE - q(VT - V3) = ECE — q((k + XB6 — Y62” (66) 2 J 2sE ) Ec Ev Figure 5.60 Schematic illustration of the emitter-base junction energy band profile after including a lightly-doped layer between the emitter and base layers. It can be seen that the spacer layer leads to a reduction in LEUE, and this lessens the importance of the emitter-base barrier in determining the collector current. I I I I I I I I I I nE Emitter Base I I I I 91 A further change made to the program was to allow for the voltage drops across the base and emitter resistances (measured in Section 5.2.3) so that the computed plot could be based on the terminal base-emitter voltage, and not just the intrinsic junction voltage. The values of RE and RB were 0.43 and 185 ohm (for an emitter area of 40 x 40jm2), respectively. It was assumed the values are true for both structures. The new simulated Gummel plots are shown in Figures 5.62 and 5.63 and show a much improved agreement between the measured and the computed values of the collector and base current. Note that the presence of the spacer layer has lowered the barrier at the emitter-base junction to such a degree that the n-factor for the collector current is now 1.02. This is the same as for the measured current, and indicates that conduction in the device is now being conducted by transport through the base, rather than by transport across the junction. The small deviation between the base current values at low bias suggests the presence of an extra recombination current which has not been accounted for in our model. Recombination in the actual spacer layer is one possibility [691. Finally, we note that the good fit to the InP#2 device has been achieved without having to invoke any modification to the transport properties at the collector junction of the composite-collector structure. Recall that our analysis of Section 3.2.1 indicated that the composite-collector layer should not not affect the dc characteristics below breakdown so long as the lightly-doped InGaAs layer was thicker than 700A. The actual thickness in InP#2 was 900A so the result gives some confidence in the analysis presented in Section 3.2.1. 92 12 0 Measured data Model Data 1o0,0I’,”, Vbe [V] Figure 5.61 Comparison of experimental and analytical model data for the Gummel Plot of InGaAs#2 HBT (emitter area 40 x 4Om2)at Vcb=OV. The effect of spacer and parasitic resistances are not modeled. 1 O2 0 Measured Data Model Data Vbe [V] Figure 5.62 Comparison of experimental and analytical model data for the Gummel Plot of InGaAs#2 HBT (emitter area 40 x 40m2)at Vcb=OV, taking into account the effect of the emitter-base spacer layer and parasitic resistances. 93 1 02 0 ooc:oo:yD Vbe [‘] Figure 5.63 Comparison of experimental and analytical model data for the Gummel Plot of InP#2 HBT (emitter area 40 x 40zm2)at Vcb=3.OV, taking into account the effect of the emitter-base spacer layer and parasitic resistances. 94 Chapter 6 Conclusion and Recommendation for Future Research 6.1 Conclusion Large-area InP/InGaAs/InGaAs HBTs using MOCVD-grown layers on InP substrates have been fabricated by a non-self-aligned technology. The effect of including a wide bandgap InP layer in the collector has been investigated both experimentally and theoretically. The conclusions that can be drawn from this work are as follows. 1. The presence of a wide-bandgap InP layer in the composite-collector devices gives a sig nificant improvement (2—3 times) in the common-emitter and common-base breakdown voltages. This may prove useful for employment of HBTs in power applications. 2. The dc Gummel plots and the SIMS profiles of the SHBTs and CCHBTs indicate that the 100A-thick spacer layer at the E-B junction is sufficient to prevent any out-diffusion of Zn atoms (for a base doping of 1O’9cm3)into the wide-bandgap emitter. The presence of this spacer affects the junction height and, therefore, the collector current. 3. Both structures demonstrate an adequate dc current gain even at a collector current as low as 1tA (15 and 20 for the SHBT and CCITBT, respectively). The slight difference in gain can be attributed to the fact that the ratio of the emitter to base doping concentration for the CCHBT is slightly larger than for the SHBT. 4. The use of Ti/Pt/Au as a metal contact appears to be effective in improving the ohmic contact to both n- and p-type In-based material, even when it is not alloyed at high 95 temperature. The optimum alloying temperature is suggested to be 300°C, as opposed to 350°C as used commonly for GaAs devices. 5. A model has been developed for the CCHBT which appears to satisfactorily predict the below-beakdown behavior of both the collector and the base current of both fabricated structures. The model suggests that, for the CCHBT devices studied, the thickness of the narrow bandgap layer in the composite-collector structure should be greater than 700A. In addition, from various comparisons of the model results and the experimental data, the major effect of the spacer layer on transport in HBTs has been demonstrated. 6.2 Recommendation for Future Research Now that large-area HBT fabrication and modeling capabilities have been established. it is proposed that the next phase of the joint UBC/NRC project be the development of a fabrication procedure to make small-area devices for high frequency applications. This would be particularly interesting to see whether any compromise needs to be reached in meeting the design requirements for high gain, high breakdown voltage and high speed in CCHBTs, such as would be necessary, for example, in the integration of HBTs and a laser on the same substrate to form a monolithically-integrated laser transmitter circuit. The inclusion of impact ionization breakdown in the model is a step worth pursuing. To do this, it will be necessary to make some improvement to the classical depletion approximation. Neglect of mobile charge carriers in the space charge regions at high current levels in the SCR of the collector may overestimate the electric field and so lead to large errors in the estimation of the breakdown voltage. 96 Bibliography [1] J.Bardeen and W.H.Brattain, “The Transistor, a Semiconductor Triode,” Phys. Rev., vol. 74, p. 435, 1948. [2] W.Shockley, “U.s. patent, 2569347,” 1951. [3] H.Kroemer, “Theory of Wide-Gap Emitter for Transistors,” Proc. IRE., vol. 45, pp. 1535—1537, 1957. [4] H.Kroemer, “Heterojunction Bipolar Transistor and Integrated Circuits,” Proc. IEEE, vol. 70, pp. 13—25, 1982. [5] H.Kroemer, “Heterostructure Bipolar Transistors: What Should We Build?,” Vac. Sci. Tech., vol. Bi, no. 2, pp. 126—130, 1980. [6] R.A.Anderson, “Experiments on Ge-GaAs Heterojunctions,” Solid State Electron., vol. 5, pp. 341—35 1, 1962. [7] J.Shewchun and L.Y.Wei, “Germanium-Silicon Alloy Heterojunction,” J. Electrochem. Soc., vol. 111, pp. 1145—1149, 1964. [8] R.D.Dupuis, L.A.Moudy, and P.D.Dapkus, “Preparation and Properties of GaA1As/GaAs Heterojunctions Grown by Metalorganic Chemical Vapour Deposition,” mt. Sym. on GaAs and Related Compounds, pp. 1—9, 1978. [9] A.Y.Cho and J.R.Artur, “Molecular Beam Epitaxy,” Frog. Solid State Chem., vol. 10, pp. 157—191, 1975. [10] T.Mimura, S.Higmizu, T.Fuji, and K.Nanbu, “A New Field Effect Transistor with Selective Doped GaAs/n-A1GaAs Heterojunctions,” Japanese Journal of Applied Physics, vol. 19, pp. L225—L227, 1980. [11] R.N.Nottenburg, J.C.Bischoff, and J.H.Abeles, “Base Doping Effects in InGaAs/InP Double Heterojunction Bipolar Transistors,” Tech. Dig. IEEE IEDM, p. 278, 1989. [12] W.Lee and C.Fonstad,“1n052Al48s/1n3Ga7IsAbrupt Double Heterojunction Bipolar Transistor,” IEEE Elect. Dev. Lett., vol. 12, no. 7, pp. 683— 97 685, 1986. [13] T.Miya, Y.Terunama, T.Hosaka, and T.Miyashitta, “Ultimate Low-Loss Single-Mode Fibre at 1.55tm,” Electron. Lett., vol. 15, p. 106, 1976. [14] R.N.Nottenburg, Y.K.Chen, M.B.Panish, D.A.Humphrey, and R.Hamm, “Hot-electron InGaAs/InP Heterojunction Bipolar Transistors with of 110GHz,” IEEE EDL, vol. 10, no. 1, pp. 30—32, 1989. [15] C.S.Kyono, N.D.Gerrard, C.J.Pinzone, C.M.Maziar, and R.D.Dupuis, “Very High Current Gain InGaAs/InP Heterojunction Bipolar Transistors Grown by Metaloganic Chemical Vapor Deposition,” Electron. Lett., vol. 27, no. 1, pp. 40—41, 1991. [16] C.F.Farley, M.F.Chang, P.M.Asbeck, N.H.Sheng, R.Pierson, G.J.Sullivan, K.C.Wang, and R.B.Nubling, “High Speed (ft=78GHz) AlInAs/GaInAs Single Heterojunction HBT,” Electron. Lett., vol. 25, no. 13, pp. 845—847, 1989. [17] C.K.Peng, T.Won, C.W.Litton, and H.Morkoc, “A High Performance InGaAs/InAlAs Double Heterojunction Bipolar Transistor with Non-alloyed n+ InAs layers on InP(n) grown by MBE,” IEEE EDL, vol. 9, no. 7, pp. 33 1—333, 1988. [18] B.Jalali, R.N.Nottenburg, M.Banu, and R.K.Montgomery, “High Speed InGaAs HBT Devices and Circuits,” in Proc. mt. Conf InP and Related Mterials, pp. 228—233, 1992. [19] C.M.Maziar, M.H.Sommerville, and C.S.Kyono, “Extended Velocity in InGaAs Collec tor for High Speed Heterojunction Bipolar Transistors,” SPIE, High-Speed Electronics and Devices Scaling, vol. 1288, pp. 69—77, 1990. [20] R.N.Nottenburg, Y.K.Chen, M.B .Panish, R.Hamm, and D.A.Humphrey, “High-Current Gain Submicrometer InGaAs/InP Heterojunction Bipolar Transistors,” IEEE EDL, vol. 9, no. 10, pp. 524—526, 1988. [21] Y.Ota, T.Hirose, M.Yanagihara, A.Ryoji, T.Kato, and M.Inada, “A1GaAs/GaAs HBT with GaInAs Cap Layer Fabricated By Multiple-Self-Alignment Process Using One Mask,” Electron. Lett., vol. 25, no. 9, pp. 610—612, 1989. [22] H.F.Cook, “Microwave Transistors: Theory and Design,” Proc. IEEE, vol. 59, pp. 1163— 1181, Aug. 1971. 98 [23] C.M.Maziar, M.E.Klausmeier-Brown, and M.S.Lundstorm, “A Proposed Structure for Collector Transit-Time Reduction in A1GaAs/GaAs Bipolar Transistors,” IEEE EDL., vol. 7, no. 8, pp. 483—485, 1986. [24] T.Ishibashi and Y.Yamauchi, “A Possible Near-Ballistic Collector in An A1GaAs/GaAs HBT with a Modified Collector Structure,” IEEE ED., vol. 35, no.4, pp. 401—404, 1988. [25] S.C.M.Ho, “The Effect of Base Grading on the Gain and High Frequency Performance of A1GaAs/GaAs Heterojunction Bipolar Transistors.” M.A.Sc. thesis, The University of British Columbia, Vancouver, Canada, 1989. [26] R.Williams, Modern GaAs Processing Methods. Artech House Inc. Norewood, MA, 1990. [27] M.Parrilla, D.Newson, D.Skellern, and M.MaCbean, “Modeling, Design and Perfor mance of InP/InGaAs Double-Heterojunction Bipolar Transistors,” in Proc. mt. Conf InP and Related Materials, p. 414, 1992. [28] Z.E.Abid, W.R.McKinnon, S.P.McAlister, and M.Davies, “InP Double Heterstructure Bipolar Transistors with A Quaternary Collector for Improved Breakdown Behavior,” in Proc. mt. Conf InP and Related Mterials, pp. 432—434, 1993. [29] B.Jalali, Y.K.Chen, R.N.Nottenburg, D.Sivco, D.A.Humphrey, and A.Y.Cho, “Influence of Base Thickness on Collector Breakdown in Abrupt AlInAs/InGaAs Heterojunction Bipolar Transistors,” IEEE EDL., vol. 11, no. 9, pp. 400—402, 1990. [30] B.Jalali, R.N.Nottenburg, A.F.J.Levi, R.A.Hamm, M.B.Panish, D.Sivco, and A.Y.Cho, “Base Doping Limits in Heterostructure Bipolar Transistors,” Appi. Phys. Lett., vol. 56, pp. 1460—1462, 1990. [31] T.Kobayashi and K.Kurishima, “Suppression of Abnormal Zn Diffusion in InPIInGaAs Heterojunction Bipolar Transistor Structures,” Appi. Phys. Lett., vol. 62, no. 3, pp. 284— 285, 1993. [32] K. Kurishima, T.Makimoto, T. Kobayashi, and T.Ishibashi, “InPfInGaAs Heterostructure Bipolar Transistors Grown at Low Temperature by Metalorganic Chemical Vapor Deposition,” Jap. Jou. of Appl. Phys., vol. 30, no. 2B, pp. L258—L261, 1991. 99 [33] K. Kurishima, T. Kobayashi, and U. Gosele, “Abnormal Redistribution of Zn in InPfJnGaAs Heterojunction Bipolar Transistor Structures,” Appi. Phys. Lett., vol. 60, no. 20, pp. 2496—2498, 1992. [34] H.-F. Chau, J.Hu, D.Pavlidis, and K.Tomizawa, “Breakdown-Speed Considerations in A1GaAs/GaAs Heterojunction Bipolar Transistor with Special Collector Designs,” IEEE Trans. ED, vol. 39, no. 12, 1992. [35] H.F.Chau and E.A.Beam, “High-Speed InPIInGaAs Heterojunction Bipolar Transistors,” IEEE EDL., vol. 14, no. 8, pp. 388—390, 1993. [36] A.W.Hanson, S.A.Stockman, and G.E.Stillman, “Comparison of InGaP/GaAs Single- and Double Heterojunction Bipolar Transistors with a Carbon-Doped Base,” IEEE EDL., vol. 14, no. 1, pp. 25—28, 1993. [37] J.-L. Pelouard, P.Hesto, J-P.Praseuth, and L.Goldstein, “Double-Heterojunction GaAlI nAs/GaInAs Bipolar Transistor Grown by Molecular Beam Epitaxy,” IEEE EDL., vol. 7, no. 9, pp. 516—518, 1986. [38] E.Tokumitsu, A.G.Dentai, C.H.Joyner, and S.Chandrasekhar, “InP/InGaAs Double Heterojunction Bipolar Transistors grown by Metalorganic Vapor Phase Epitaxy with Sulfur Delta Doping in the Collector Region,” Appl. Phys. Lett., vol. 57, no. 26, pp. 2841—2843, 1990. [39] A.Feygenson, D.Ritter, R.A.Hamm, P.R.Smith, R.K.Montgomery, R.D.Yadvish, H.Temkin, and M.B.Panish, “InGaAs/InP Composite-Collector Heterojunction Bipo lar Transistors,” Electron. Lett.,, vol. 28, no. 7, pp. 607—609, 1992. [40] A.Feygenson, R.A.Hamm, D.Ritter, P.R.Smith, R.K.Montgomery, R.D.Yadvish, and H.Temkin, “High Speed InGaAs/InP Composite Collector Bipolar Transistors,” in Proc. mt. Conf DRC, 1992. [41] A.Feygenson, R.A.Hamm, P.R.Smith, M.R.Pinto, R.K.Montgomery, R.D.Yadvish, and H.Temkin, “A 144GHz InP/InGaAs Composite Collector Heterojunction Bipolar Tran sistor,” in IEDM, 1992. [42] D.Ritter, R.A.Hamm, A.Feygenson, H.Temkin, and M.B.Panish, “Bistable Hot Electron Transport in InP/GaInAs Composite Collector Heterojunction Bipolar Transistors,” Appl. 100 Phys. Lett., vol. 61, no. 1, pp. 70—72, 1992. [43] A.Feygenson, R.K.Montgomery, P.R.Smith, R.A.Hamm, M.Haner, R.D.Yadvish, and M.B.Panish, “InP/GaInAs Composite Collector Heterojunction Bipolar Transistors and Circuits,” in Proc. mt. Conf InP and Related Materials, pp. 572—575, 1993. [44] S.C.M.Ho and D.L.Pulfrey, “The Effect of Base Grading on the Gain and High Frequency Performance of A1GaAs/GaAs Heterojunction Bipolar Transistors,” IEEE Trans. ED, vol. 36, p. 2173, 1989. [45] A.A.Grinberg, M.S.Shur, R.J.Fischer, and H.Morkoc, “An Investigation of the Effect of Graded-Layers and Tunneling on the Performance of A1GaAs/GaAs Heterojunction Bipolar Transistors,” IEEE ED., vol. 31, pp. 1758—1765, 1984. [46] O.S.Ang and D.L.Pulfrey, “The cut-off frequency of base-graded and junction-graded AlGaAs DHBTs,” Solid-State Electronic, vol. 34, pp. 1325—1328, 1992. [47] M.S.Lundstorm, “An Ebers-Moll Model for the Structure Bipolar Transistor,” Solid- State Electronics, vol. 26, p. 1173, 1986. [48] S.O.Ang, “Double Heterojunction Bipolar Transistor Model.” M.A.Sc. thesis, The University of British Columbia, Vancouver, Canada, 1991. [49] A.St.Denis, D.L.Pulfrey, and A.Marty, “Reciprocity in Heterojunction Bipolar Transis tors,” Solid-State Electronics, vol. 35, no. 11, pp. 1633—1637, 1993. [50] D.L.Pulfrey and S.Searles, “Electron Quasi-Fermi Level Splitting at the Base-Emitter Junction of AlGaAs/GaAs HBT’s,” IEEE ED., vol. 40, no. 6, pp. 1183—1185, 1993. [51] S.Searles and D.L.Pulfrey, “An Analysis of Space-Charge-Region Recombination in HBTs,” IEEE ED., submitted for publication. [52] S.Chandrasekhar, B.C.Johnson, E.Tokumitsu, A.G.Dentai, C.H.Joyner, A.H.Gnauck, J.S.Perino, and G.J.Qua, “A Monolithic Long Wavelength Photoreceiver Using Het erojunction Bipolar Transistors,” IEEE QE., vol. 27, no. 3, p. 773, 1991. [53] Q.Z.Liu, D.L.Pulfrey, and M.K.Jackson, “Analysis of the Transistor-Related Noise in Integrated pin-I{BT Optical Receiver Front Ends,” IEEE ED., to appear December 1993. 101 [54] S.Adachi, “Material Parameters of InGaAsP and Related Binaries,” J. Appl. Phys., vol. 53, pp. 8775—8792, Dec. 1982. [55] T.P.Pearsall, GaInAsP Alloy Semiconductors. J.Wiley, 1982. [56] M.S.Lundstrom, M.E.Klausmeier-Brown and M.R.Melloch and R.K.Ahrenkiel and B.M.Keyes, “Device-Related Material Properties of Heavily Doped Gallium Arsenide,” Solid-State Electronics, vol. 33, no. 6, pp. 693—704, 1990. [57] N.K.Dutta and R.J.Nelson, “The Case for Auger Recombination in InGaAsP,” J. Appl. Phys., vol. 53, no. 1, pp. 74—92, 1982. [58] L.Davis and Y.Lam and D.Nichols amd J.Singh and P.K.Bhattacharya, “Auger Re- combination Rates in Compressively Strained InGai...As/InGaAspIInP (0.53x0.73) Multiquantum Well Lasers,” IEEE PTL,, vol. 5, no. 2, pp. 1163—1181, 1993. [59] Y.P.Varshni, “Band-to-Band Radiative Recombination in Groups IV, VI, and ffl-V Semiconductors (I),” Phys. Stat. Sol., vol. 19, pp. 459—514, 1967. [60] A.P.Heberle, U.Strauss, W.W.Ruhle, K.H.Bachem, T.Lauterbach, and N.Haegle, “Minority-carrier lifetime in heavily doped GaAs:C,” Jap. J. Appl. Phys., Part], no. 1B, pp. 495—497, 1993. [61] D.R.Wright, P.E.Oliver, T.Prentice, and V.W.Steward, “Diffusion Length in p-Type MOCVD GaAs,” J. Ciystal Growth, vol. 55, pp. 183—191, Oct. 1981. [62] D.G.Deppe, “Thermodynamic Explanation to the Enhanced Diffusion of Base Dopant in A1GaAs-GaAs npn Bipolar Transistors,” Appi. Phys. Lett., vol. 56, no. 4, pp. 370—372, 1990. [63] K.-H.Hellwege, ed., Landolt-Bornstein “Numerical Data and Functional Relationships in Science and Technology”, vol. 17A. Springer-Verlag, 1982. [64] M.S.Lundstrom, “An Ebers-Moll Model For Heterojunction Semiconductor Devices,” Solid-State Electronics, vol. 29, no. 11, pp. 1173—1179, 1986. [65] J.Feng, M.A.Sc thesis, The University of British Columbia, Vancouver, Canada,. to appear in 1993. 102 [66] J.R.Shappirio, R.T.Lareau, R.A.Lux, J.J.Finnegan, D.D.Smith, L.S.Heath, and M.Taysing-Lara, “Metal Penetration and Dopant Redistribution Beneath Alloyed Ohmic Contacts to n-GaAs,” J. Vac. Sci. Tech., vol. A 5, no. 4, pp. 1503—1507, 1987. [67] M. Ogawa, “Alloying Behavior of Ni/Au-Ge Films on GaAs,” J. Appi. Phys., vol. 51, no. 1, pp. 406—412, 1980. [68] “Private communication with Dr. Z.Abid at NRC.” [69] H.Ito, “Generation-Recombination Current in the Emitter-Base Junction of Al GaAs/GaAs HBTs,” Jap. J. ofAppi. Phys., vol. 25, no. 9, pp. 1400—1404, 1986. 103

Cite

Citation Scheme:

    

Usage Statistics

Country Views Downloads
United States 12 2
India 9 9
China 2 0
Taiwan 2 0
Japan 2 0
City Views Downloads
Kolkata 9 9
Ashburn 8 0
Unknown 3 5
Redmond 2 0
Beijing 2 0
Tokyo 2 0
Sunnyvale 1 0

{[{ mDataHeader[type] }]} {[{ month[type] }]} {[{ tData[type] }]}

Share

Share to:

Comment

Related Items