1246

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 38. NO. 6, J U N E 1991

Simulation of Quantum Transport in Quantum Devices with Spatially Varying Effective Mass
Hideaki Tsuchiya, Matsuto Ogawa, Member, IEEE, and Tanroku Miyoshi, Member, IEEE
Abstract-The precise simulation of quantum transport in quantum size structures has become more important with the recent progress of crystal growth technology. In this paper, we present Some novel progress in quantum-mechanical simulation based on the wigner function model. A,, exact nonlocal formulation in the Wigner representation due to a spatially varying effective mass and its discretization for numerical calculation are discussed for the first time. To verify the validity of such a formulation, the current-voltage characteristics of resonant tunneling diodes are simulated to compare with the conventional Wigner function model. We also point out the importance of the self-consistent calculation in the electrostatic potential for precise device simulation. Finally, we emphasize that the Wigner function model is superior to the alternative method based on the transmission probability method even for the static simulation of quantum transport.

I. INTRODUCTION HE precise simulation of quantum transport in quantum-size structures has become more important with recent progress in crystal growth technology. The Wigner function model has been a powerful method to analyze the dynamic behavior of quantum size optical and electronic devices [ 11-[9]. However, there still remain some essential problems to be solved in the Wigner function model. In this paper, we will present some successful progress in the model taking the quantum-well resonant tunneling diode (RTD) as the prototype quantum-size device. First, we will discuss an exact nonlocal formulation in the Wigner representation due to a spatially varying effective mass. When we permit the change of energy-band structure with position to be simulated by a spatially varying effective mass, the quantum Liouville equation governing the density matrix is readily derived from the effective mass Hamiltonian [ 101. Carefully performing a Fourier transform of the Liouville equation, we will mathematically obtain the exact nonlocal Wigner function formulation for spatially varying effective mass. In the finite difference discretization for the Wigner representation, the current density is successfully defined to satisfy the discretized continuity equation. This is the crucial point in the numerical calculation. Further, it should be emphasized that the novel Liouville equation including the spatial variation in effective mass can be discretized into a

block-tridiagonal matrix equation as well as a conventional one. so far, a nonlocal formulation of the Wigner representation was discussed for the abrupt change Of effective mass at a heterojunction based on classical considerations [ 1 11, where a step-like change of effective mass is considered to lead to additional impulsive force at the heterointerface that contributes to a classical drivin the Liouville equation' Unfortunately~ the paing per was primarily concerned with the formulation of the problem and its solution was not discussed. TO verify the validity of the novel formulation derived in this paper, the cument-voltage characteristics of the RTD are simulated and compared with the conventional Wigner function model where the spatial variation of effective mass is neglected. It is found from the simulation that the positional variation of effective mass plays an important role in the transport characteristics of quantum devices. In addition, we also point out that the self-consistent calculation of the electrostatic potential should be included to model the space charge in a quantum well. To this end, Poisson's equation is solved simultaneously. In the last part of this paper, we emphasize that even for the static simulation of quantum transport, the Wigner function model is superior to the alternative simulation of Tsu and Esaki [ 121 based on the transmission probability method. In that method, the transport characteristics are determined by the transmission probability of the electron plane wave assuming an equilibrium distribution of carriers over the electrodes. Consequently, the calculated characteristics never depend upon the dimension of the electrodes. On the other hand, in the Wigner function model, the distribution function excited at one reservoir attached to the electrode is simulated spread in the k (wave-number) space as it approaches a barrier through the electrode due to the barrier potential repulsion. Thus the Wigner function method is more likely to be a powerful tool in the static and dynamic simulation of quantum transport in quantum-size devices if the positional variation of effective mass is considered as discussed in this paper.
FOR 11. QUANTUM-MECHANICAL LIOUVILLE EQUATION SPATIALLY VARYING EFFECTIVE MASS TO simulate the quantum transport in quantum devices with a spatially varying effective mass, an exact nonlocal formulation in the Wigner representation is derived mathematically starting from the effective mass Hamiltonian.

Manuscript received July 28, 1990; revised January 15, 1991. The review of this paper was arranged by Associate Editor M . Shur. The authors are with the Department of Electronic Engineering, Kobe University. Rokko-dai, Nada-ku. Kobe 657. Japan. IEEE Log Number 9144383.

0018-9383/91/0600-1246$01 .OO 0 1991 IEEE

. .

TSUCHIYA

er a l . :

QUANTUM TRANSPORT IN QUANTUM DEVICES

1247

Now, we consider the following effective mass Hamiltonian which guarantees probability current conservation at any position with variable effective mass [ 101:
(1)

where

- sin [(k - k ' ) 4 ]

. COS [(k - k ' ) t ]

(7)

where A , m ( x ) , and v(x) are Planck's constant divided by 2 a , the effective mass, and the potential, respectively. For the Hamiltonian, the quantum-mechanical Liouville equation for the density matrix is written as

(8)

where p is the density matrix represented in the coordinate space and defined as [ 2 ]
P@,

(x
1
1

x')

c P,Cp,(x) Cpfw

:)I

sin [(k - k ' ) 4 ]

(11)

(3)
1
-

where cpl (x)represents the wave function in the i state and P, is the probability of occupying the i state. When the independent variables are changed to x = (x x') and 4 = x - x ' , the Wigner function [ 1 3 ] is defined as a Fourier transform of the density matrix with respect to the relative coordinates 4

m-(x, 4 )
--

m(x +
-

it)
it)

1 m(x -

44)
(12)

"(x, 4)

m(x +

+ m(x 1 . - ;E)

f ( x , k) =

In the derivation of the Liouville equation (6), the following relations are used to simplify the representation:'

-W

dEejkEP(x+

it, x

it).

(4)

On the other hand, the density matrix is represented by the inverse Fourier transform of the Wigner function with respect to the momentum variable k dke-Jkv(X,k).
(5)

Performing the Fourier transformation of the Liouville equation (2), the following Liouville equation for the Wigner function is obtained.
-=

af at

OD

-a

dk' -MMI(x, k - k') 8a

af ( x , k ' )
~

It is readily found that (6) includes the conventional Liouville equation for constant effective mass. If the effective mass is forced to the constant value of m , because M , ( x , k - k ' ) , M 2 ( x , k - k ' ) , and M 3 ( x , k - k ' ) are zero and M 4 ( x , k - k ' ) is simplified as ( 4 a / m ) 6 ( k k ' ) , where 6 ( k - k ' ) is the Dirac delta function, (6) leads to

ax

-h

+A

im iW
-a -a

4 a k'M2(x, k - k ' ) f ( x , k ' ) Equation (14) is precisely identical with the conventional Liouville equation [ 2 ] ,and it is well known that it reduces to the Boltzmann transport equation in the classical limit. As for the novel Liouville equation ( 6 ) , it is found to correspond to the following transport equation in the classical limit:

c 16a M 3 ( x ,k - k ' ) [ a i f E 2 k ' )

I248

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 38, NO. 6, JUNE 1991

It is worth noting that a new force term is added in (15) due to the variable effective mass as is expected through a p / a t = - a H / a x and the effective mass Hamiltonian (1). The Wigner function has the following relations with the carrier density n ( x ) and the current density j ( x ):

On the other hand, the boundary conditions for potential are given for the bias condition V, as [ 141

where w ( x ) is the band grading function introduced to represent the bandgap variation with distance and given by where e is the electron charge. Here, the carrier density n ( x ) is found from ( 5 ) and the current density j (x) is defined so as to satisfy the continuity equation derived from the Liouville equation (6) by integrating with respect to k . To include the self-consistency in potential, the following Poissons equation must be solved simultaneously:

where N , and N,, denote effective densities of states in the conduction band and the valence band, respectively, and EG is the energy gap.

where E denotes the spatially varying dielectric constant and J? is the doping density. The Liouville equation (6) and Poissons equation (18) are mutually related through

u ( x > = - [ x , ( x > - x m 1 - e [ r l . ( x ) - rl.(O>l (19) where xe is the electron affinity of semiconductor material. According to Frensley [2], the boundary conditions for the Wigner function are given at the reservoirs attached to the quantum device as

f(0, k) = f , ( k ) ,

k >0 k <0

111. DISCRETIZATION OF THE LIOUVILLE EQUATION Equations (6) and (1 8) will be solved numerically using the finite-difference method. When the position coordinates x ( 0 Ix IL ) are represented by N , mesh points with a uniform mesh spacing A , ( = L / ( N , - l)), the relative coordinates E are represented by Nt mesh points with a uniform mesh spacing 2Ax, and the k dimension (-n/2AX Ik 5 n / 2 6 , ) is discretized into Nk mesh points with a uniform mesh spacing A k = a/NkA,, (6)-(11) are transformed into the following discretized equations:
M I ( x ,k - k ) at
k

f(L k)

=f,(k),

(20)

AX

f ( x + Ax, k ) - f ( x , k ) , f ( x , k ) -f(x - A,, k ) ,

-

>0

k M 2 ( ~k, - k ) f ( x , k )

where the subscripts 1 and r express the left-hand side and the right-hand side boundary, respectively, L is the device length, and kB is the Boltzmann constant. fi,, . ( k ) is derived by integrating the Fermi-Dirac distribution function represented by the carriers quasi-Fermi level Efi,rand temperature T over the transverse momenta. The above boundary conditions mean that the distribution of carriers injected into the device from a reservoir is characterized by the thermal equilibrium distribution of carriers at the reservoir, and all carriers impinging upon a reservoir from the device are absorbed by the reservoir. The carriers quasi-Fermi level E f r will be determined by the carrier densities n (0) and n ( L ) at the boundaries so as to satisfy the charge neutrality condition, and p (0) n (0) = n?(O)and p ( L ) n ( L ) = n(L), where ni denotes the intrinsic carrier density.

- 2(1

+ 2A;k*)f(x,k) + f ( x
kM4(x, k - k )
AX

Ax, k ) ]

-AC
k

-h k V ( X ,k - k ) f ( x , k )

TSUCHIYA et al.: QUANTUM TRANSPORT IN QUANTUM DEVICES

1249

from the discretized Liouville equation ( 2 4 ) by summing with respect to k

--

2NkA,

c sin [ ( k - k ) E ]

--

C COS [(k - k ) E ]
1

NkAx

It is easily found from ( 3 0 ) that the current density is independent of x for a steady-state solution using this definition o f j . Here, it is worth noting that the number of Nk is exactly limited to 2 N E from the requirement that the discretized equations (24)-(29) should be identical with the discretized equations of the conventional Liouville equation ( 1 4 ) when the effective mass is forced constant. The discretization of Poissons equation (1 8) is straightforward. However, for the iterative calculation, the following transformed matrix equation describing the potential change SJ/is required at each iteration:

+ 173(~) W ( X + A,)
-

- V I ( X ) J/(x

A,)

r2(x) $ ( x ) - 173(x) $ ( x + A x )
-

- e[F(x)
where

n(x)l

(33)

V I ( X )= ( X - ~ A X ) / ( A , > ~

2 V(x, k - k ) = Nk

sin [(k - k ) ( ]
5

173(~)= E ( X + ; A x ) / ( A , ) ~ 172(x)
= -[E(X

[u(x +

4)

- Y(X

- ; A x ) + E(X + ; A x ) l / ( A x ) 2 . ( 3 4 )

4).

(29) In the discretized Liouville equation (24), the forward difference and the backward difference are used for k < 0 and k > 0, respectively, to satisfy the boundary conditions given by ( 2 0 ) and ( 2 1 ) . When the current densities are defined at intervals between mesh points, we expect the discretized continuity equation to have the form

In the numerical calculation, the Wigner functionf( x, k ) is computed iteratively for each bias condition until the self-consistent solution is obtained. In this paper, the iteration is continued until the change of the potential energy Y ( x ) becomes less than 10 meV at any position. To obtain the transient response of a quantum size structure, the further discretization with respect to time is performed based upon the following approximation method with mesh spacing A , :
- -f(x,

af
at

k, t

+ A,)
At

-f(x,

k , t)

The above form of approximation for the current density is commonly used in the classical device simulation to improve the computational accuracy. If the carrier density is defined as

then the current density is successfully defined to satisfy the discretized continuity equation ( 3 0 ) that is derived

where d : is the operator defined in ( 2 4 ) . Although the matrix equation derived from the discretization of the novel Liouville equation is much less sparse than that of the conventional one, it can be transformed into the block-tridiagonal matrix with (Nk X Nk) nondiagonal matrix blocks. Thus the matrix equation will be solved by calculating only the (Nk X Nk) inverse matrices.

1250

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 38, NO. 6, JUNE 1991

IV. COMPARISON BETWEEN THE NOVELMETHOD AND THE CONVENTIONAL MODEL To verify the validity of the novel formulation derived in this paper, the RTD is simulated as the prototype quantum-size device. The RTD model used in the simulation is shown in Fig. 1 , which consists of a 4.5-nm-wide GaAs quantum well and 2.8-nm-wide Alo 3Gao ,AS barrier layers. The GaAs electrode layers of 17.5 nm are included on each side of the device. Outside the electrodes, reservoirs characterized by the thermal equilibrium distributions of carriers are assumed. The conduction band discontinuity is taken to be 0.27 eV. In the calculation, N , = 8!, Nk = 60, N E = 30, and the mesh spacing A x = 5.65 A , that is comparable to the atomic monolayer spacing, are used. The effective mass of electron are 0.067mo in GaAs and 0.092~ in~Alo ~ 3Gao,AS, respectively, where mo is the free electron mass. The doping density in the GaAs electrodes is given as 2 X 10l8cm-3 and the barrier and well layers are assumed to be undoped. The relative dielectric constants are 12.9 in GaAs and 12.03 in Alo3Gao7As, respectively. Scattering processes are neglected and all calculations are performed at a room temperature of 300 K. First, to study the effect of a spatially varying effective mass in quantum transport, the current-voltage characteristics of the RTD are simulated by solving the discretized Liouville equation (24) under the condition df/& = 0. The applied bias voltage is assumed to be dropped uniformly across the well and the barriers (the flat-band model) as shown in Fig. 1. The simulated results are shown in Fig. 2, along with the curve simulated assuming the constant effective mass of 0.067mo throughout the device. The solid and the broken lines correspond to the novel and conventional simulations, respectively. It is found from the figure that for the novel simulation method the resonance takes place at a lower bias voltage and the peak-to-valley ratio in current increases about 1.8 times larger. The above results mean that the positional variation of the effective mass plays an important role in the transport characteristics of quantum devices. Next, to study the space-charge effects, the discretized Liouville equation (24) and the discretized Poisson's equation (33) are solved self-consistently as shown in Fig. 3. The solid line is the self-consistent solution and the dot and dash line is the solution under the flat-band model. The calculated carrier distributions and the conductionband-energy variations are shown in Fig. 4 for the bias conditions (a) 0 V, (b) 0.18 V (peak current), and (c) 0.24 V (valley current). In the figure, the conduction-band-energy variations under the flat-band model are also shown in dotted lines. It is found from Fig. 4(b) that at resonance the electric field strength in the quantum-well layer is reduced compared with that of the flat-band model due to the carrier accumulation in the well. This is the reason why the resonance takes place at a higher bias voltage for the self-consistent solution in Fig. 3. Further, when the bias voltage is applied, the carriers are found accumulated

0.4

r-----l

10

20

30

40

Distance (nm)

Fig. I . Resonant tunneling diode model used in the simulation.

Variable effective mass Constant effective mass

voltage ( v )

Fig. 2. Comparison between the novel and conventional simulation

Voltage(V)

Fig. 3 . Self-consistent simulation of current-voltage characteristics.

in the left electrode and depleted in the right electrode as shown in Fig. 4(b) and (c). Due to such a space-charge distribution, the electric fields are induced in the originally neutral electrodes to accelerate the electrons. Accordingly, the relatively higher current flows when the space-charge effects are considered as shown in Fig. 3. The above results mean that the space charge has important influence on the characteristics of quantum devices, and the self-consistent calculation in potential is required for precise device simulation. It is worth noting that the accumulation and depletion layers are formed although inelastic processes are not included in these calculations [ 151. The velocity of electron is inversely proportional to its effective mass. When an electron enters the barrier region, it speeds down having larger effective mass. On the contrary, when it exits the barrier region, it speeds up with light effective mass. Thus the accumulation layer is formed in the left electrode, and the depletion layer appears in the right electrode.

TSUCHIYA et al.: QUANTUM TRANSPORT IN QUANTUM DEVICES

0.4

0.2

3
v

0.0

;
U C

- 0.2
Distance ( n m )

a 'b
"t A
4 2
I

' : T
Wgner function
.._-Transmission probability
\.-------~'

1251

. '

0.1

0.2
Voltage ( V )

0 . 3

0.4

(a)

Fig. 5 . Comparison between the Wigner function model and the transmission probability method. Self-consistency in potential is not implemented, but variable effective mass is assumed in both models.

10

20 30 Distance ( nm )

, . I

1 0.2
I

I k (nm-' )

40

(b)
I
A

Fig. 6. Distribution of the Wigner function at resonance ( V

0.1 V)

".-

where D(E,, V b ) is the transmission probability in terms of the probability current as a function of the longitudinal energy E, and the bias voltage V,. r Fig. 5 shows the simulated results for (36) and the Wig0.0 ; ner function model, where self-consistency in the electrow C static potential is not implemented, but the variable effective mass is assumed for both models. The solid and the broken lines correspond to the Wigner function model and the transmission probability method, respectively. In the Distance ( n m ) transmission probability method, the assumed Fermi(C) Dirac distribution is kept constant along the electrode. In Fig. 4. Camer distributions and conduction-band-energy variations for the other words, the transport characteristics calculated by this bias condition (a) 0 V , (b) 0.18 V , and. (c) 0 . 2 4 V . 0.18 and 0 . 2 4 V correspond to the peak and valley currents of the current-voltage charactermethod never depend upon the electrode dimension. On istics in Fig. 3, respectively. the other hand, in the Wigner function method the distribution function excited at one reservoir is found spread in the k space as it approaches a barrier due to the barrier V . COMPARISON BETWEEN THE WIGNERFUNCTION potential repulsion even if the potential is uniform in the MODELAND THE TRANSMISSION PROBABILITY METHOD electrode as shown in Fig. 6. Because of this spread of As an alternative model to simulate the static behavior the distribution function, the average energy of electrons of quantum transport, the transmission probability method becomes higher. Consequently, electrons tunnel through [12] is known well. In the method, the transport charac- the barriers for a lower bias voltage and a larger number teristics are determined by the transmission probability of of electrons flow over the barriers to contribute the higher the electron plane wave. When the Fermi-Dirac distri- current density. This is the reason for the discrepancy of bution function is assumed in the electrodes, the current the two curves in Fig. 5 . density is expressed by integrating over the transverse energies as [ 121 VI. CONCLUSION
0.24
0.2
v

i 3

A novel progress in the quantum-mechanical simulation based upon the Wigner function has been presented. An exact nonlocal formulation in the Wigner representation due to a spatially varying effective mass is formulated for the first time. The validity of such a formulation is

1252

IEEE TRANSACTIONS ON ELECTRON DEVICES, VOL. 38, NO. 6, JUNE

1991

verified for the resonant tunneling diode model. We also point out that the self-consistent calculation of the electrostatic potential is important for precise device simulation. As a result, the Wigner function model is more likely to be a powerful tool in the simulation of quantum transport in quantum size devices. In this paper, we discussed only the quantum ballistic transport in quantum-well devices. To make the Wigner function model practically useful, we will have to study the precise model including the scattering processes due to phonons, impurities, and intervalley transitions on the basis of a firmer quantumwell physics.

Hideaki Tsuchiya was born in Ehime, Japan, on August 12, 1964. He received the B.E. and M.E. degrees in electronic engineering from Kobe University, Kobe, Japan, in 1987 and 1989, respectively. Currently, he is working toward the Ph.D. degree at Kobe University. His research interest is centered on the quantum transport characteristics in electronic and optical devices with quantum structures. Mr. Tsuchiya is a member of the Institute of Electronics, Informaticin and Communication Engineers of Japan, and the Japan Society of Applied Physics.

*
Matsuto Ogawa (S80-M82-S82-M85) was born in Morioka, Iwate-Prefecture, Japan, on January 24, 1956. He received the B.E., M.E., and Dr. of Eng. degrees in electronic engineering from the University of Tokyo, Tokyo, Japan, in 1980. 1982, and 1985, respectively. In 1985, he joined the Department of Electronic Engineering of Kobe University, Kobe, Japan, as a Research Associate and is now a Lecturer. He has been engaged in research of semiconductor and lightwave electronics. His current research interest deals with the characterization of heterointerface and quantum structures, high-speed optical devices, optical short-pulse generation by modelocked lasers, and the quantum-mechanical simulation of quantum devices such as RTD (resonant tunneling diode), quantum APD, and HEMT. Dr. Ogawa is a member of Institute of Electronics, Information and Communication Engineers of Japan, and the Japan Society of Applied Physics.

REFERENCES
W. R. Frensley, Transient response of a tunneling device obtained from the Wigner function, f h y s . Rev. Leu., vol. 57, no. 22, pp. 2853-2856, Dec. 1986. -, Wigner-function model of a resonant-tunneling semiconductor device, f h y s . Rev. B., vol. 36, no. 3, pp. 1570-1580, July 1987. -, Quantum transport calculation of the small-signal response of a resonant tunneling diode, Appl. f h y s . Lerr., vol. 51, no. 6, pp. 448-450, Aug. 1987. -, Quantum transport modeling of resonant-tunneling devices, Solid-state Electron., vol. 31, no. 314, pp. 739-742, 1988. U. Ravaioli, M. A. Osman, W. Potz, N. Kluksdahl, and D. K. Ferry, Investigation of ballistic transport through resonant-tunneling quantum wells using Wigner function approach, fhysica, vol. 1349, pp. 36-40, 1985. A. M. Kriman, N. C. Kluksdahl, and D. K. Ferry, Scattering states and distribution functions for microstructures, f h y s . Rev. B, vol. 36, no. 11, pp. 5953-5959, Oct. 1987. N. C. Kluksdahl, A. M. Kriman, C. Ringhofer, and D. K. Ferry, Quantum tunneling properties from a Wigner function study, SolidStare Electron., vol. 31, no. 314, pp. 743-746, 1988. H. Tsuchiya, S . Shimizu, N. Fujino, M. Ogawa, and T. Miyoshi, Quantum mechanical simulation of hole piling and transient response of long wavelength APDs, in OEC90 Tech. Dig.,1990, pp. 226-227. H. Tsuchiya, M. Ogawa, S. Shimizu, N. Fujino, and T . Miyoshi, Quantum mechanical simulation of long wavelength avalanche photo-diodes using Wigner-function, presented at Electrosoft 90 (Lowell, MA), Aug. 1990. D. J. BenDaniel and C. B. Duke, Space-charge effects on electron tunneling, f h y s . Rev., vol. 152, no. 2, pp. 683-692, Dec. 1966. J. R. Barker, D. W. Lowe, and S. Murray, in The Physics ofSubmicron Structures, H. L. Grubin, K. Hess, G. J. Iafrate, and D. K. Ferry, Eds. New York: Plenum, 1984. pp. 277-286. R. Tsu and L. Esaki, Tunneling in a finite superlattice, Appl. f h y s . Lerr., vol. 22, no. 11, pp. 562-564, June 1973. E. Wigner, On the quantum correction for thermodynamic equilibrium, fhys. Rev., vol. 40, pp. 749-759, June 1932. M. Kurata and J. Yoshida, Modeling and characterization for highspeed GaAIAs-GaAs n-p-n heterojunction bipolar transistors, IEEE Trans. Elecrron Devices, vol. ED-31, no. 4 , pp. 467-473, Apr. 1984. W. R. Frensley, Effect of inelastic processes on the self-consistent potential in the resonant-tunneling diode, Solid-Srare Elecrron., vol. 32, no. 12, pp. 1235-1239, 1989.

. ,

*
Tanroku Miyoshi (S67-M72) was born in Osaka, Japan, on January 6, 1944. He received the B.S., M.S., and Ph.D. degrees all in electronic engineering from the University of Tokyo, Tokyo, Japan, in 1967, 1969, and 1972, respectively. In 1972 he was appointed Lecturer, and from 1974 to 1987, he was an Associate Professor in the Department of Electronic Engineering, Kobe University, Kobe, Japan, where he is presently a Professor. He has been engaged - - in research of electromagnetic-wave theory, microwave integrated circuits, and lightwave electronics. In 1976 he was a Visiting Scholar at McGill University, Montreal, Que., Canada. From 1982 to 1984, he was a Visiting Scientist at Bell Laboratories, Holmdel, NJ. Dr. Miyoshi is a member of the Institute of Electronics, Information and Communication Engineers of Japan and the Japan Society of Applied Physics. He received a Yonezawa Award in 1974 and an Outstanding Book Award in 1977 both from the Institute of Electronics, Information and Communication Engineers of Japan.