QUANTUM WELL DEVICES FOR MICROWAVE AND

MILLIMETER WAVE OSCILLATOR

APPLICATIONS

 

Vijay P. Kesan, Ph.D.

The University of Texas at Austin, 1989

Supervising Professor: Dean P. Neikirk

 

Copyright

by

Vijay P. Kesan

1989

 

DISSERTATION

Presented to the Faculty of the Graduate School of

The University of Texas at Austin

in Partial Fulfillment

of the Requirements

for the Degree of

DOCTOR OF PHILOSOPHY

THE UNIVERSITY OF TEXAS AT AUSTIN

May, 1989

 

ABSTRACT

The analysis, design, fabrication, and testing of a QWITT (quantum well injection transit time) diode oscillator and self-oscillating mixer are presented.

Small signal and large signal models for the QWITT device that relate physical device parameters to the dc terminal characteristics and predicted rf performance were developed. These models were then used to provide optimum device dimensions to maximize rf performance at a desired frequency of operation. The large signal device model was also used to design a planar QWITT diode oscillator.

Initially, GaAs/AlAs resonant tunneling diode structures were grown by MBE and devices with good room temperature negative differential resistance (NDR) characteristics were fabricated. A number of QWITT devices were then fabricated and planar and waveguide oscillator circuits were designed to test these devices at frequencies between 1-35 GHz. The rf performance of QWITT devices was shown to be significantly better than resonant tunneling diodes. An output power of 1 mW, corresponding to an output power density of 3.5-5 kW/cm2 in the frequency range of 5-8 GHz has been obtained from a planar QWITT oscillator. This is the highest output power obtained from any quantum well oscillator at any frequency and is approximately 5 times higher power and 2-3 higher output power density than reported in the literature for a comparable frequency. This result also represents the first planar circuit implementation of a quantum well oscillator. By comparison, the cw output power density obtained from an IMPATT diode at these frequencies is 10-30 kW/cm2. Good qualitative agreement between the dc and rf characteristics of QWITT devices and theoretical predictions based on small signal and large signal analyses has been achieved. We also present results on improving dc-to-rf conversion efficiency by optimizing the design of the drift region in the device through the use of a doping spike. By optimizing the doping concentration of the spike, an increase in efficiency from 3% to 5% has been obtained, without compromising the output power at X-band.

We have demonstrated that self-oscillating QWITT diode mixers have the ability to produce conversion gain in both waveguide and planar circuits. A maximum conversion gain of 10 dB was obtained over a narrow band width of 10-20 MHz at X-band frequencies. If broad band operation (around 100-200 MHz) is desired, then this may be achieved with an average conversion loss of about 3-5 dB. To the best of our knowledge, this is the first report of conversion gain obtained from a self-oscillating mixer using any quantum well device and is also the highest conversion gain reported for a self-oscillating mixer circuit using any semiconductor device.

 

ACKNOWLEDGEMENTS

In an experimental project such as this, one becomes quickly indebted to a large number of people. I would like to express my deep gratitude and appreciation to Dr. Dean Neikirk for providing guidance and enthusiastically supporting my work. I am indebted to Dr. Ben Streetman for sharing his thoughts freely, and for his guidance, support, and friendship. My heartfelt thanks to Dr. Tatsuo Itoh for providing many discussions on microwave circuit issues and for contributing to my non-technical life by pulling me away to the tennis courts. I am grateful to Dr. Peter Blakey for lending his insight on transit time devices. To Dr. Chris Maziar, I am thankful for discussions on transport theory and for her friendship. I would like to thank my good friend Terry Mattord for sharing his experience with MBE and UHV equipment. I would also like to thank Drs. Al Tasch and Michael Downer for taking time off from their busy schedules to serve on my supervisory committee.

To my collaborators, Andy Campbell, Gentry Crook, Ananth Dodabalapur, Tom Linton, Doug Miller, Amir Mortazawi, and Vijay Reddy, I would like to express my sincere thanks for their assistance and discussions. To my fourth floor mates and fellow members of Club MED (Micro-Electromagnetic Devices) who have never been dull, a special word of thanks. I would like to thank Terrace Demerjian and Bernice Wootton for really keeping things moving and for helping me conquer the bureaucracy. I would also like to acknowledge the financial assistance of the Texas Advanced Technology Program and the Joint Services Electronics Program without whose support this work would not have been possible. Finally, I would like to thank my parents for their patience and support over the years and for not asking when I was going to graduate.

Above all, it is difficult for me to imagine an environment that would have enabled me to work better, and to all those who contributed to it, my heartfelt thanks.

TABLE OF CONTENTS

Acknowledgements v

Abstract vi

Table of Contents viii

List of Tables x

List of Figures xi

Chapter 1. INTRODUCTION 1

Chapter 2. MODELING OF THE QWITT DIODE 20

Chapter 3. MOLECULAR BEAM EPITAXY AND DEVICE FABRICATION 56

Chapter 4. DC CHARACTERISTICS OF QWITT DEVICES 64

Chapter 5. RF CHARACTERISTICS OF QWITT DEVICES 77

Chapter 6. RECOMMENDATIONS FOR FUTURE WORK 97

Appendices 107

Bibliography 113

Vita

LIST OF TABLES

 

Table 4.1: DC characteristics for QWITT devices with uniformly doped depletion regions, A through D. 70

Table 4.2: DC characteristics for QWITT devices with a doping spike, E through H. 74

Table 5.1: Microwave and millimeter-wave performance of QWITT diode oscillators, A through D, in both waveguide and planar circuits. 8

Table 5.2: Microwave and millimeter-wave performance of QWITT diode oscillators, E through H, in both planar microstrip and waveguide circuits. 84

LIST OF FIGURES

 

Fig. 1.1: Pattern of current research in semiconductor heterostructures. Adapted from Ref. 45. 3

Fig. 1.2 (a): Electron energy as a function of position in the double barrier resonant tunneling structure. The energy level E1 occurs above the bottom of the bulk conduction band because of confinement in the x direction. (b) Diode current as a function of incident electron energy. From Ref. 78. 7

Fig. 1.3: Physical structure of the QWITT diode and energy band diagram of the device when no bias is applied. The length of the transit time region, W, would be much greater than the quantum well thickness for typical millimeter wave frequencies. 10

Fig. 1.4: a) The bias point for the QWITT diode is adjusted so the bias voltage across the quantum well injection region, Vi, is less than the resonant bias by an amount equal to the rf voltage across the well, VRF; b) current injection through the well will then peak at wt = /2 when the total instantaneous voltage across the well is equal to 2E1/2. Note that the I-V curve shown is for the quantum well injection region only, and does not represent the I-V curve for the complete QWITT diode. 12

Fig. 1.5: (a) Schematic energy band diagram at wt=0 when the dc voltage drop across the quantum well Vi is below resonance by an amount equal to the amplitude of the rf voltage VRF; (b) At /2 in the rf cycle the well will now be at resonance, causing current to be injected into the transit time region of the device. In actual operation the device would not be at equilibrium, and the electric field in the drift region would be changed by the presence of the drifting current pulse. 13

Fig. 2.1: (a) Quantum well diode structure which exhibits significant transit-time effects. The GaAs spacer layer on the cathode side of the quantum well region is made thin to reduce parasitic series resistance. A thick, undoped GaAs spacer layer is used on the anode side to produce a depletion region of length W, much longer than the thickness of the quantum well region l. The transit-time through this layer is much larger than that through the quantum well region; this device forms a quantum well injection transit-time (QWITT) diode. (b) Small-signal equivalent circuit for the structure shown in (a). ZQW is the specific impedance, e' is the effective dielectric constant, and s is the injection conductance, of the quantum well; Ztt is the specific impedance, and e is the dielectric constant of the depletion region. Rprstc represents the parasitic series resistance due to any undepleted spacer regions, highly doped contact regions, and ohmic contacts. 23

Fig. 2.2: Transit angle optimized small-signal specific negative resistance curves for fixed values of quantum well injection conductance (dashed curves, given by (2.5)), and the performance envelope for both injection conductance and transit angle optimization (solid curve, given by (2.9)). The depletion region is assumed to be GaAs, with a saturation velocity vs of 6 x 106 cm/sec. The optimized depletion layer thickness corresponding to each point on the curves is found from (2.7); for the solid curve the injection conductance s at each frequency is found from (2.8), and the transit angle is /3. 28

Fig. 2.3: (a) Injection conductance and transit angle optimized envelope of specific negative resistance, given by (2.9); (b) transit angle optimized specific negative resistance for devices with a positive quantum well injection conductance s of 0.5 -1cm-1, given by (2.5) and (2.7); and band-limited characteristics for devices with fixed depletion layer thicknesses W of (c) 6.0 µm, (d) 0.4 µm, and (e) 0.09 µm, given by (2.5). 31

Fig. 2.4: Specific negative resistance for devices with a negative injection conductance s of -0.3 -1cm-1 and depletion layer thicknesses of 0.23 µm, 0.14 µm, and 0.07 µm. For such a device |s|/e = 41 GHz.

(a) Semi-logarithmic scale emphasizing the low frequency (f << 41 GHz) behavior of these devices. For low frequencies the specific negative resistance is given by (2.10), and is frequency independent.

(b) Log-log scale emphasizing the high frequency (f >> 41 GHz) behavior of these devices. For high frequencies the device resistance changes from negative to positive at the intrinsic transit-time cut-off frequency, given by (2.14). A more realistic estimate of cut-off frequency is obtained by finding the intersection of the device specific negative resistance curve and a curve of constant specific contact resistance, shown at 10-6 -cm2. 34

Fig. 2.5: Output power at two representative frequencies as a function of the rf voltage modulation factor for a QWITT diode. 49

Fig. 2.6: Output power density as a function of frequency for the two bias conditions in a QWITT diode: (a) below the current peak, i.e, below resonance (/2 mode); (b) above the current peak, i.e., above resonance (3/2 mode). 50

Fig. 2.7: Efficiency as a function of the rf voltage modulation factor at 125 GHz for the two bias conditions for a QWITT diode. 51

Fig. 2.8: Output power and optimum device area as a function of frequency for a QWITT diode. For the quantum well dc I-V characteristic considered in this analysis, the peak current density is 40 kA/cm2 at 0.8 V and the current density at the valley is 12 kA/cm2 at 1.2 V. 53

Fig. 2.9: Output power as a function of frequency for two QWITT diodes with different quantum well characteristics (i.e., different values of s). The dc negative differential resistance for one quantum well was chosen to be much higher than the other. 54

Fig. 3.1: RHEED measurements of the growth rate of GaAs and AlAs as a function of gallium and aluminum effusion cell temperatures. 57

Fig. 3.2: Carrier concentration at 300K, as determined by Hall measurements, for MBE grown silicon doped GaAs layers versus silicon effusion cell temperature. 58

Fig. 3.3: A cross sectional SEM photograph of a quantum well mesa device typically used in this study. 60

Fig. 4.1: A schematic cross section of the QWITT diode structures, A through D, examined in this study. 65

Fig. 4.2: Room temperature dc I-V curves in both bias directions for a resonant tunneling diode, device D. 67

Fig. 4.3: Room temperature dc I-V curves for QWITT devices A and C with 500 Å and 2000 Å depletion region length respectively. 68

Fig. 4.4: A schematic cross section of the QWITT diode structures with a doping spike, E through H, examined in this study. 73

Fig. 4.5: Room temperature dc I-V curves for QWITT devices E and F with a doping spike of 5x1016 cm-3 and 1x1017 cm-3 respectively. 75

Fig. 5.1: Block diagram of the waveguide circuit used at microwave and millimeter wave frequencies. 78

Fig. 5.2: Photograph of a WR-10 waveguide, a GaAs chip containing QWITT diodes of varying diameters mounted on a micrometer controlled post, and a diode whisker contact. 79

Fig. 5.3: A planar microstrip QWITT diode oscillator at X-band. 81

Fig. 5.4: Close up view of the diode chip showing devices of various diameters contacted by a whisker in the planar microstrip circuit. 82

Fig. 5.5: Spectrum of a QWITT diode (device F) oscillating at 31 GHz in a waveguide circuit. 86

Fig. 5.6: Block diagram of both waveguide and planar self oscillating mixer circuits. 90

Fig. 5.7: Conversion gain as a function of IF in the X-band waveguide self oscillating mixer using the QWITT diode. 91

Fig. 5.8: Photograph of broadband IF signal from 2 to 100 MHz with an average power of -52 dBm. The rf signal power is -49 dBm centered at 9.3 GHz. 92

Fig. 6.1: Schematic representation showing top and side views of the parallel plate waveguide circuit periodically loaded by QWITT diodes or resonant tunneling diodes. 102

Fig. 6.2: Imaginary part of the input admittance as a function of frequency for a periodic oscillator circuit designed to operate at 94 GHz. 104

Fig. 6.3: Oscillation frequency as a function of the distance between adjacent diodes. At any desired oscillation frequency the distance between the diodes must be half the guide wavelength. 105