Solid-State

Electronics

Pergamon Press 1962. Vol. 5, pp. 341-351.

EXPERIMENTS

Printed in Great Britain

ON Ge-GaAs HETEROJUNCTIONS
R. L. ANDERSON*

International

Business Machines Corporation,

(Received

29 January;

Yorktown Heights, New York

in revised form 23 April

1962)

Abstract-The
electrical characteristics
of Ge-GaAs heterojunctions, made by depositing Ge
epitaxially on GaAs substrates, are described. 1-v and electro-optical characteristics are consistent
with a model in which the conduction- and valence-band edges at the interface are discontinuous.
The forbidden band in heavily doped (n-type) germanium appears to shift to lower energy values.
R&sum&-Les
CaractCristiques Clectriques des h&rojunctions
de Ge-GaAs produites en dbposant
le Ge Cpitaxiallement sur les couches inferieures de 1AsGa sont d&rites. Les CaractCristiques 1-F
et electrooptiques sont consistantes avec un modhle dans lequel les bords des bandes de conduction
et de valence g linterface sont discontinus. La bande defendue dans le germanium (type-n) fortement dope semble se dCplacer vers des valeurs &energie plus basse.
Zusammenfassung-Die
elektrischen Kenngrsssen von Ge-GaAs Hetero-Uberg%ngen, die man
durch epitaxiale Ablagerung von Ge auf GaAs-Substraten herstellt, werden beschrieben. 1-F und
elektro-optische Kenndaten entsprechen einem Modell, in dem die Blinder des Leitungs- und
Valenzbandes an der Grenzfllche diskontinuierlich sind. Das verbotene Band in stark dotiertem
(n-Typ) Germanium scheint sich nach niedrigeren Energiewerten zu verlagem.

1. INTRODUCTION

JUNCTIONS between
two semiconductors
of the
same element but with different impurities
present
have been studied
extensively.
These junctions
are reasonably
well understood.
The periodicity
of the lattice is not disturbed
at the junction
and
so the properties
of the semiconductors
at the
junction can be expected to be the bulk properties.
Metal-semiconductor
contacts,
on the other
hand, are not well understood.
The chief difficulty
is usually
attributed
to interface
effects. Even
though
the semiconductor
and the metal may
each be monocrystalline,
the crystal structures
and lattice constants
in general are different and
so an expitaxial contact is not formed. Because of
the abrupt change in the structure
and periodicity
of the lattice and the resultant
disorder
in the
region near the interface,
material properties
are
not the same here as they are in the bulk.
The theoretical
voltage-current
characteristic
of a p-n junction or a metal semiconductor
contact
_
* Now at the Department of Electrical Engineering,
Syracuse University, Syracuse, New York.

is derived

for most models

to be of the form(l)

I = Io[exp(qV/lkT)-

11

(1)

where I is the current due to an applied voltage V,

10 is the saturation
current or the current for large
negative voltage, q is the electronic
charge, K is
Boltzmanns
constant
and T is the absolute
temperature.
The
value
of 10 is reasonably
independent
of voltage
in most
derivations.
The diode formula is often written in the form
I = Io[exp(qV/+T)-

11

(2)

where 7 is an empirical factor which describes the

disagreement
between
simple theory and experiment for forward bias (V > 0). The value of +j is
commonly
about 2.3 for gallium
arsenide
p-n
junctions,
is between 2 and 4 in Ge point-contact
diodes and approaches
the theoretical
value of
unity only in Ge p-n junctions
(and in silicon
p-n junctions
at elevated
temperatures).
The
variation of current with reverse voltage is usually
accounted
for by permitting
the term 10 to vary
slowly with voltage. These deviations
from the
theory have not been adequately
explained.
341

342

R.

L.

ANDERSON

Little work has been done on junctions between

two semiconductors. GUBANOVhas suggested that
the I-V characteristics of copper oxide rectifiers
might
be indicative
of semiconductor-semiconductor contacts.* SHOCKLEY(3) and KR&UER@)
suggested using a semiconductor
with a wide
forbidden region as an emitter for a transistor
which has base and collector of a narrower-gap
semiconductor. The purpose of this is to obtain a
high injection efficiency. JENNY(~) has described
attempts to fabricate
a Gap-GaAs
wide-gap
emitter by diffusing phosphorus into gallium
arsenide. Little success has been reported.
This paper discusses the electrical characteristics
of junctions formed between Ge and GaAs. These
junctions
are contained within a monocrystal.
Ge was deposited epitiaxially onto GaAs seeds by
the Iodide Process.(s-s) These two materials have
similar crystal structure,
and virtually equal
lattice constants (5.62 A). As a result, it is expected
that strain at the interface is negligible.
Junctions
between two dissimilar materials
will be referred to as heterojunctions
in contrast
to homojunctions where only one semiconductor
is involved.
ENERGY-BAND
PROFILE
OF HETEROJUNCTIONS
Consider the energy-band profile of two isolated
pieces of semiconductor shown in Fig. 1. The two
semiconductors
are assumed to have different
1 erent dielectric constants (E),
band gaps (EB), dff
different work functions (4) and different electron
affinities (0). Work function and electron affinity
are defined, respectively, as that energy required
to remove an electron from the Fermi level (Ef)
and from the bottom of the conduction band (EC)
to a position just outside of the material (vacuum
level). The top of the valence band is represented
by Ev. The subscripts 1 and 2 refer to the narrowgap and wide-gap semiconductors, respectively.
In Fig. 1, the band-edge profiles (&I, Ee2, Evl,
E,z) are shown to be horizontal.
This
is
equivalent to assuming that space-charge neutrality exists in every region. The difference in
energy of the conduction-band
edges in the two
materials is represented by AE, and that in the
valence-band edges by AE,.

A junction formed between an n-type narrowgap semiconductor and a P-type wide-gap semiconductor is considered first. This is referred to
_

VACUUM

-------a--

LEVEL

92

0,
ELECTRON
ENERGY

92

4
I

Ec?

t
AE,
Ed
ELI ----

--E,l

Eg2

EI
AE,

______L__.

__

~~~
E 2

FIG. 1. Energy-band
diagram for two isolated semiconductors in which space-charge neutrality is assumed
to exist in every region.

as an n-p heterojunction.
The
energy-band
profile of such a junction at equilibrium is shown
in Fig. 2.
Wiihin any single semiconductor the electrostatic potential difference between any two points

2.

* For a review of Gubanovs work see Ref. 2.

I,

x.3

11

FIG. 2. Energy-band diagram of n--p heterojunction at

equilibrium.
can be represented by the vertical displacement
of the band edges between these two points, and
the electrostatic field can be represented by the
slope of the band edges on a diagram such as
Fig. 2. Then the difference in the work functions
of the two materials is the total built-in voltage
( VD). Vn is equal to the sum of the partial built-in

EXPERIMENTS

ON

Ge-GaAs

voltages (Vol+ Vns) where Vnl and Vns are the

electrostatic
potentials
supported
at equilibrium
by semiconductors
1 and 2, respectively.
Since
voltage is continuous
in the absence
of dipole
layers, and since the vacuum level is parallel to
the band edges, the electrostatic
potential difference
(I/J) between any two points is represented
by the
vertical displacement
of the vacuum level between
these two points.
Because of the difference
in
dielectric
constants
in the two materials,
the
electrostatic
field is discontinuous
at the interface.
Since the vacuum level is everywhere
parallel
to the band edges and is continuous,
the discontinuity
in conduction-band
edges (AEC) and
valence-band
edges (AEV) is invariant with doping
in those cases where the electron affinity and band
gap (E,) are not functions
of doping (i.e. nondegenerate
material).
Solutions
to Poissons equation,
with the usual
assumptions
of a Schottky
barrier,*
give, for the
transition widths on either side of the interface for
a step junction,

(X2-X0)

2
-

NDlW(

[q

~A2(d!Dl+

and the total width

W =
=

occurs in the most lightly doped region for nearly

equal dielectric
constants.
The transition
capacitance
is given by a generalization of the result for homojunctions
:

C=

~(&DI

+ ~2Nz42)

l/2

qNDlNAw2
(VD-

v)

(6)

It can be seen from Fig. 2 that the barrier to

electrons
is considerably
greater
than that to
holes, and so hole current will predominate.
The case of an n-n junction
of the above two
materials
is somewhat
different.
Since the work
function
of the wide-gap
semiconductor
is the
smaller, the energy bands will be bent oppositely
to the n-p case (See Fig. 3). However,
there are

l/2

V)

(3b)

l2NA2) I

W of the transition

region

is

(x2-x0)+(x0-x1)
h2(

VP-

[ (qom

The relative
semiconductors

VD -

343

HETEROJUNCTIONS

voltages
are

Q(NA2

+ &A2)NDl~A2

supported

VDl--1

G2 -

+ &)2

=v2

12

in each

1 A2<2
NDI~I

(4)
FIG. 3. Energy-band

at

of the

(5)

where VI and Vs are the portions of the applied

voltage I/ supported
by materials 1 and 2 respectively. Of course VI+ Vz = V. Then (VnlVI)
and (Fns - Vs) are the total voltages (built in plus
applied) for material 1 and material 2, respectively.
We can see that most of the potential
difference
* See Ref. 9 for details for calculations
junctions.

diagram of n-n heterojunction

equilibrium.

for homo-

a negligible
number
of states available
in the
valence band and so the excess electrons
in the
material
of greater
work function
will occupy
states in the conduction
band. Since there are a
large number of states available in the conduction
band, the transition
region extends only a small
distance
into the narrow-band
material and the
voltage is supported
mainly by the material with
the smaller work function.
The voltage profile in the interface
region can
be determined
by solving for the electric
field
strength
(F) on either side of the interface
and
using the condition
that the electric displacement
(D = &) is continuous
at the interface. Assuming

R.

344

Boltzman

L.

ANDERSON

statistics in region 1,

-(Vo1-

3. PREDICTED

V)]Jlr2

In region 2 the electric displacement

face is
E2FZ(XO)

[2EZpYD2(

VD2

(7)

at the inter-

v2p2

(8)

Equating equations (7) and (8) gives a relation

between (Vnr- VI) and (Vna- Va) which is quite
complicated.
However, it is reasonably easy to
get an upper limit of (Vor- VI). If the exponential
in equation (1) is expanded in a Taylor series,
the following inequality is obtained :
1)

2kT E&&
(VDl_

Vl)

<

__

~$7;

vD2

v-2)

(9)

From equation (9) we can see that the electrostatic potential will be supported mainly by semiconductor
2 unless Nns $ Nur, or for high
forward bias.
n-n
heterojunctions
the
transition
For
capacitance
is difficult to calculate. However,
above, the
except for the cases mentioned
capacitance
of a metal-semiconductor
contact
is a good approximation.
In the heterojunctions discussed here, the energy
gap of the wide-gap material (Ga-As) overlaps
that of the narrow-gap material, and the polarity
of the built-in field (and of rectification)
is
dependent on the conductivity type of the widegap semiconductor. Fig. 4 shows the equilibrium
energy-band
diagrams for p-n and p-p heterojunctions.

I-V

CHARACTERISTICS

Because of the discontinuities

in the band
edges at the interface, the barriers to the two types
of carriers have different magnitudes,
and so
current in a heterojunction
will in most cases
consist almost entirely of electrons or of holes.
The variation of current with applied voltage
for these heterojunctions
(neglecting generationrecombination current) is
I = A exp(-qVBZ,kT)--B

exp(-qVgl/kT)
(10)

where VB~ is the barrier that carriers in semiconductor 1 must overcome to reach semiconductor
2, and Vga is the barrier to the carriers moving
the opposite direction. The coefficients A and I?
depend on doping levels, on carrier effective mass
and on the mechanism of current flow.
In the junctions
depicted in Fig. 24,
Vjjr
exists for the predominant current carrier and so
I = A exp[ -q( VD$kT]
[exp(qVdkT)-

exp(-qVl/kT)J
(11)

where Va and li are those portions of applied

voltage appearing in materials 2 and 1, respectively.
The first term in the brackets is important for
forward bias and the second for reverse bias. If
12 = V/y then 1-1 = (1 - 1/7)V and the current
varies approximately exponentially with voltage
in both forward and reverse directions. It should
be noticed how-ever that at increased reverse
voltage
V& disappears, i.e. (Vu1 - by) > Al?,
(for the case of the p-n junction), and the current is
expected to saturate. If Vor > .Wc (again for a
p-n heterojunction---see
Fig. S), 1Tnr = 0 and the

.Ef

a
FIG. 4. Energy-band
diagrams in the interface region for p-n and p-p
heterojunctions. Electron energy is plotted vertically.

FIG. 6. Cross-sectional
view of a wafer of GaAs on which Ge has been deposited.
The thickness of the deposit is about 0.03 cm.

[facing

p. 344

I:Ic. i. Plot of photovoltage and thermoelectric

voltage against distance normal to the surface
for an no p Ge-CaAs hcterojunction.
The surface is indicated by the extreme left of each trace.
The junction position is indicated hy a zero thermoelectric
voltap_c.

TlP#
DEPOSlTlON

2
3

VOLTAOE

THERMQ-ELECTRIC

PHOTO-VOLTAGE
7.

FIG. 8. Region near interface of Fig. 4 on expanded scale. The shape of the photovoltage
indicates the transition region to be predominantly in the gallium arsenide.

plot

FIG.

9. V-l
characteristics
of a Ge-GaAs
n-n heterojunction.
The
ordinate scale is 0.1 mA/div while the abscissa scale is 1-O V/div.

EXPERIMENTS

entire applied
barrier height:

voltage

is effective

ON

in

Ge-GaAs

varying

I = A exp[-q(Vn-AE,)/UJ[exp(qV//lKT)-l]
(12)
Above a critical forward voltage in such a diode,
Vsr will become finite [(vol- VI) < AE,] and
the current will vary exponentially with I2 = I/7
(see Section 4.2).
Since in the n-p heterojunction,
the current is
limited by the rate at which holes can diffuse in
the narrow-gap material,(rsJ
A = XaqN~z(DP/rP)l~2

(134

where the transmission

coefficient X represents
the fraction of those carriers having sufficient
energy to cross the barrier which actually do so.

k___
t

FIG. 5. Band diagram of p-n heterojunction in which no

barrier exists for electrons going from Ge to GaAs (solid
line) and for applied forward bias where now the barrier
does exist (dashed line). The expected I-V characteristics
are considerably different in the two regions.

D, and 7P are diffusion constant and lifetime,

respectively, for holes in the narrow-gap material,
and a represents junction area.
The case of the p-n heterojunction is analogous.
In the case of n-n and p-p heterojunctions,
SinCe
VDl
and vi are small with respect to voz
and Va, respectively,
and because the current
is carried by majority carriers, we have, in analogy
with the emission theory for metal-semiconductor
diodes,(ri)
W)
where Ns and m* are, respectively, net impurity
density and carrier effective mass in semiconductor
2.

HETEROJUNCTIONS

The above formulae would be modified someand

what
by
generation-recombination(12)
leakage currents, by image and tunnel effects,
and by interface states.
4. EXPERIMENTAL

RESULTS

In this section the electrical characteristics of

n-p, n-n, p-n and p-p heterojunctions
are reported
and interpreted
with respect to the theory of
Sections 2 and 3. It must be emphasized that the
junctions
reported here were made in two depositions. The n-type germanium in the n-p and
n-n junctions is expected to be similar since this
Ge (phosphorus doped) was deposited simultaneously on n- and p-type GaAs. Likewise the
p-type Ge (gallium doped) in the p-n and p-p
junction is expected to be similar. The p-type
GaAs seeds in the n-p and p-p heterojunctions
were cut from adjacent slices of a monocrystal.
The same is true for the n-type GaAs seeds in the
n-n and p-n junctions.
Fig. 6 shows a crosssectional view of a GaAs substrate surrounded by
deposited Ge.
To fabricate a diode from such a wafer, the
deposited Ge was removed from one side, and the
wafer was then broken into chips. Ohmic contacts
were made to both sides of the chip, and the
chip was then mounted in a transistor header and
etched to remove surface damage.
All heterojunctions
tested showed rectification.
For forward bias, the GaAs was biased negative
(with respect to the Ge) for the n-n and p-n
junctions and positive for n-p and p-p junctions.
This is in agreement with the proposed model.
The junctions studied can be classified as being
good diodes or bad diodes. The good units
of each junction
type all behave very nearly
identically. The built-in voltages are equal and the
electrical characteristics vary only slightly among
units. The bad units, however, all appear to have
somewhat lower built-in
voltages which vary
from unit to unit. Although the bad units have not
been studied as intensively as the good units, it
appears that if the reduced built-in voltage is
taken into account, the electrical characteristics
are similar to those of the good units. It is thought
that the bad units contain defects at the interface
which lower the barrier height. Many such bad
units were transformed into good units by reducing
the junction
area and presumably
eliminating

346

R. L.

ANDERSON

defects.
Only the good units will be discussed
further.
That rectification
actually occurs at the interface
was determined
by probing the material.(ls)
Fig. 7
shows a plot of thermoelectric
and of photovoltaic
potential versus distance from germanium
surface
of an n-p heterojunction.
The germanium
surface
position
is represented
by the extreme
left of
either trace. The thermoelectric
voltage null is
indicated to be about 46.8 p below the germanium
surface.
In the photovoltage
plot, the position
46.8 p from the Ge surface is as indicated. That this
position indeed corresponds
to the interface
can
be seen from Fig. 8 where the transition
region is
expanded
to show that the junction
(position
of
maximum
slope in the photovoltage
plot) is as
indicated in Fig. 7. For the case of the n-n or p-p
junctions,
a similar method was used. Instead of
the change
in polarity
for the thermoelectric
voltage,
an abrupt
change
in magnitude
was
observed at the junction.
The deposited
n-type Ge (in the n-p and n-n
junctions)
was much more heavily doped than was
the GaAs. The net donor concentration
in the Ge
was determined
by resistivity
measurements
and
was found
to be about
lOls/cma.
Capacitance
measurements
on n-p and p-p heterojunctions
indicate that the net acceptor concentration
in the
GaAs is constant
for distance from the junction
greater
than 0.25 p and is equal to 1.5 x 101s
atoms/cma.
This is in agreement
with Hall-effect
data. In the n-n heterojunctions,
capacity measurements indicate a net donor concentration
in the
GaAs varying as ~(417) where x is the distance
from the interface.
At the edge of the transition
region at equilibrium,
the net donor concentration
is about 4 x 101s atoms/cma.
The resistivity of the
p-type
Ge (in p-n and p-p heterojunctions)
was
not measured.
However,
electrical characteristics
indicate
a net acceptor
concentration
in the
neighborhood
of 101s atoms/ems.
In the n-n andp-p junctions,
the space charge in
the Ge is composed
of mobile carriers and so the
voltage supported
at the junction is expected to be
almost
entirely
in the GaAs
in these
cases.
However,
since the n-type
Ge is more heavily
doped than the GaAs, and the p-type Ge is more
lightly doped, the built-in voltage and transition
region occur predominantly
in the GaAs for n-p
junctions
and in the Ge for p-n junctions.
This

can be seen for an n-p junction

in a plot of photo
voltage ZX. position (see Fig. 8) where the position
of maximum
slope indicates
an undetectable
voltage is supported
by the Ge.
4.1 Alignment of bands at interface
The built-in
voltages at room temperature
as
determined
from 1-I and from C-V characteristics are presented
(Table 1) for representative
n-n, n-p, p-p and p-n heterojunctions.
The

Table 1
~____

I_-.

VD

Heterojunction

I-V

n-?z

0.47* 0.02
0.62f 0.02

0.48* 0.05
0+35+ 0.05

0.56
0.53

0.70 _t 0.05
0.55 rt: 0.05

Z-P
P-P
P*

c-v
--__

rt 0.03
+ 0.03
_

agreement
between methods
is good for n-n and
p-n junctions
but not for n-p or p-p heterojunctions.
Since similar germanium
was used for jz-n and
n-p junctions,
the model proposed
predicts
that
the sum of the built-in voltages (vn) for the two
types of junctions
plus the energy between
the
appropriate
band
edge and the
Fermi
level
(A,, AD) adds up to the band gap of the GaAs.
The same is true for p-n and p-p junctions.
The
values of A, and A, are calculated to be 0.19 and
0.07 eV assuming
the magnitude
of the hole
effective mass is equal to that of a free electron
(ma) and using the published
value of 0.07&a
for the electron effective mass.(r)
Then, with VJJ obtained from I-I- data,
0~62+0~47-t0~19+0~07
for n-p and n-n junctions

= 1.35 eV

and

0~53+0*56-t0~19+0*07

= 1.35 eV

for p-p and p-n junctions,

which
is in good
agreement
with the published
value of 1.36 eV for
the band gap of GaAs.
The magnitude
of AEe and AE, can be obtained
only approximately
from the data, because the
position
of the Fermi level with respect
to the

EXPERIMENTS

ON Ge-GaAs

HETEROJUNCTIONS

347

conduction band edge and the band gap of this

degenerate germanium
can only be estimated.
If the density of states in the conduction band for
degenerate Ge is that for non-degenerate Ge, and if
the band gap of this Ge is assumed to be 0.48 eV
as suggested by PANKOVE,(~~)values of 0.56 and
0.32 eV are obtained for AE, and AE,, respectively,
for degenerate n-type Ge and non-degenerate
GaAs.
A calculation of the band edge discontinuities
for non-degenerate p-type Ge and non-degenerate
GaAs gives values of 0.15 andO.
eV, respectively,
for AE, and AE,. These measurements
indicate
that with increased doping of germanium with
phosphorus, the entire forbidden band is depressed
to lower energies.
4.2 I-V characteristics
The heterojunctions
studied have static I-V
characteristics reasonably typical of those reported
for homojunctions.
The forward current varies
approximately exponentially with applied voltage,
and the reverse characteristics show a soft breakdown. The n-p junctions show an additional abrupt
breakdown which is believed to be due to the
avalanche effect. Fig. 9 shows the I-V characteristics of an n-n heterojunction at room temperature.
The I-V
characteristics
can generally
be
written as in equation (2) where the value of 7
indicates the deviation from ideal forward rectifier
characteristics.
For n-n and p-p junctions,
the
applied voltage is supported almost entirely by
the GaAs, and as a result the factor q is expected to
approach unity. The value of 7 is also expected to
approach unity for the n-p junctions, because the
Ge is so much more heavily doped than is the
GaAs so that again the applied voltage is almost
entirely supported by the GaAs.
However, in the p-n heterojunctions
studied,
the relative dopings indicate that 77should equal
unity for VD~- VI > AE, and approach V/V2
for Vol- Vl < AE,. The plot of In I vs. V is
shown for respective junctions of these four classes
in Fig. 10. The data was taken at elevated temperatures to reduce the influence of surface leakage
and generation-recombination
currents.
The plots of the n-n, n-p and p-p junctions can
all be expressed by equation (2) where r] is just
slightly greater than unity as is expected. The
I-V characteristic
for the p-n heterojunction,

0.08

0.16
Forward

0.24
Volts

0.32

0.40

FIG. 10. Forward I-V characteristics for p-n, n-n, n-p

and p-p heterojunctions. The indicated value of 7 is
fitting
the
found
by empirically
expression
I = lo expCqV/$sT).

however, has a sloppy characteristic

and a
value of 7 of approximately
3.5, although it is
not a constant. At 78K the I-V characteristics of
this diode are as shown in Fig. 11. There are three
straight-line
regions of this plot corresponding
to three distinct values of r] (equation 2). In region
a, for applied voltage V < 0.16 V, 7 = 2.1. In
the range 0.16 < V < O-7 V (region b), equation
(2) is satisfied with 77 = 16.7. For V > 0.7 V
(region c), the value of 7 doubles and becomes
7 = 8.3. These characteristics are interpreted as
follows :
The barrier is decreased by the amount of the
applied voltage in region a. A value of 77 = 2.1
results from recombination
of carriers in the
transition region. In regions b and c, the conduction-band
edge in the Ge is lower than its

348

R.

L.

ANDERSON

peak in the GaAs (see Fig. 5). As a result, only that

portion of applied voltage (Ps) appearing
in the
GaAs lowers the barrier.
Then effectively
11 is
increased.
The halving of 7 at about V = 0.7 V
results from the predominance
of injected current
above this voltage.

1 !

Id7

for the n-p and p-p homojunctions,

and about
10-s for the n-n junction.
The value of X for the
p-n junction
was more difficult
to determine.
However,
a value less than IO-3 with a value of
AE, in the neighborhood
of 0.5 eV seems necessary
to explain the experimental
results.

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.0 0.9 I.0
Forward

-7

10

Volts

FIG. 11. Forward 1-V characteristics at 78K for p-n

heterojunction having p-Ge less heavily doped than
n-GaAs.

It must be pointed
out that the value of AE,
as determined
earlier would be expected to result
in a change from region a to region b of the 1-L
characteristic
at about 0.66 V, or, conversely,
a
value of AE, of about 0.5 eV would be required to
interpret
the data as we have done. The reverse
electrical
1-Y characteristics
of representative
n-n, n-p, p-p and p-n heterojunctions
are shown
in Fig. 12. Although the data was taken at elevated
temperatures
to minimize
generation-recombination current, the reverse current does not saturate.
The origin of this excess current is not known.
From the magnitude
of the current
at a given
voltage, the values of A in equations
(11) and (12)
can be experimentally
determined.
Comparing
these values with equations
(13a, b) give values
of transmission
coefficients
(X) of about 10-s

Reverse

IO

Volts

FIG. 12. Reverse 1-V characteristic

and

for p-n, n-n, p-p

at elevated temperatures. Soft
breakdown is observed.

n-p heterojunction

If the interpretation
is correct, the small value of
transmission
coefficient
X is a result
of the
radically different
Bloch waves on either side of
the interface.
For the particular
case of the n-n
junctions
discussed
here, if it is assumed that all
electrons
in the Ge are reflected
at the interface
except those centered
around
the K = (0, 0,O)
minimum,
and that all these are transmitted,
a
value of X = 3 x 10-3 results. The actual transmission factor would be expected
to be smaller
than this because of additional
reflection
due to
the discontinuities
in band edges and in the
periodicity
of the potential-energy
function
at
the interface.
4.3 Response to monochromatic radiation
A

p-n

heterojunction,

which

the

electrical

EXPERIMENTS

ON

Ge-GaAs

characteristics
suggest has a band profile as in
Fig. 5, was illuminated
with monochromatic
radiation normally incident to the GaAs surface.
The resultant phbtocurrent
response is shown in
Fig. 13 where the short-circuit
photocurrent
per

HETEROJUNCTIONS

.WY

sufficiently low energy, the photons cannot excite

electrons in the Ge and the current is again zero.
It is noticed that both the direct and the
indirect absorption edges of the Ge are visible,
although the data is not sufficiently accurate to

Photon
FIG.

13.

Short-circuit

Energy,

eV.

photocurrent
of a p-n heterojunction
vs. photon energy (see text).

incident photon is plotted against the photon

energy.
The response shows a broad maximum between
about 0.83 and 1.4 eV. This response may be
explained as follows: the higher-energy
photons
are absorbed near the surface of the GaAs and do
not contribute to the photocurrent.
However, the
GaAs is transparent
to photons having energy
less than that of the forbidden gap and these
photons are transmitted
to the interface. These
photons
having sufficient energy will excite
carriers in the Ge and those which excite carriers
in the transition region or within a carrier diffusion
length of the transition
region will contribute
to the photocurrent.
It is these photons which
produce
the photocurrent
(see Fig. 5). At
6

per incident

photon

The flat-top
of
see much fine structure.
this figure probably indicates that the incident
radiation is entirely utilized in producing photocurrent or else that the absorption coefficient is
reasonably constant in this energy range. The
decrease to zero in photocurrent in the high-energy
region occurs at a value of about 1 as.5 eV instead
of the expected value of the GaAs band gap
(1.36 eV). This result is not understood.
5. HETEROJUNCTIONS

AS DEVICES

The static 1-V characteristics

of the diodes
studied are in general poorer than obtainable in
homojunctions-principally
because of the soft
reverse breakdown.
It is expected that this
characteristic may be improved with more work.

350

R.

L.

ANDERSON

An interesting effect in the pulse response is

expected from certain heterojunctions.
When a
diode is abruptly switched from a state of forward
bias to a state of reverse bias, no effects on the
current due to minority carrier storage are expected.
For the n-n and p-p junctions, this is because
current is by majority carriers. For p-n and n-p
junctions, however, minority-carrier storage exists
as in homojunctions.
Here, however, the discontinuity at the interface prevents the injected
minority carriers from re-entering the GaAs when
the diode is abruptly reverse biased.
Preliminary measurements on heterojunctions
have detected no effects on the pulse response
attributable to storage effects.
A heterojunction
can be used as a photocell
with a built-in filter, as indicated in Section 4.3.
The cell is sensitive for only a narrow band of
photon wavelengths.
With the GaAs as emitter, and Ge as base and
collector,
a wide-gap-emitter
transistor
seems
possible. Such a transistor would be expected to
have a high injection efficiency independent of
impurity concentration ratio in base and collector.
Attempts to construct such a transistor have not
been successful.
Although the measurements reported above are
on units made from two single depositions of Ge
on GaAs, other depositions have been made.
400,

show the typical

tunnel-diode
characteristic
(see Fig. 14). The value of Vo lies between the
values obtained in Ge and in GaAs tunnel diodes
and is approximately what would be expected
from a consideration
of the proposed band
picture. Fig. 15 depicts the band picture suggested
for a tunnel heterodiode.
Tunneling
takes
place between the Ge conduction band and the
GaAs valence band as in tunnel homodiodes.
The peak-to-valley current ratios for the tunnel
heterodiodes
at room temperature
have been
observed to be in excess of 20. Because of the
magnitude of the built-in voltage ID, the valleys
are wider than the Ge units.

--

GoAs

FIG. 15. Energy-band diagram of an n-p tunnel heterodiode at equilibrium.

SUMMARY

0.5

I.0

VOLTS
FIG. 14. I-V characteristics of a Ge-GaAs tunnel n-p
heterodiode. The ordinate scale is 50 mA/div and the
abscissa scale 0.1 V/div.

Degenerate
n-type Ge was deposited on degenerate p-type GaAs and diodes were fabricated
from this material.* The I--Y plots of these units
* This work was carried out by J. C. MAFUNACEand
F. H. DILL.

Germanium
has been deposited on gallium
arsenide by a process involving germaniumiodine compounds. The resultant structure is a
monocrystal in which the junction between the
Ge and GaAs is abrupt.
These junctions rectify. Probing of the junction
region shows that the rectification occurs at the
interface.
The electrical characteristics of these heterojunctions are roughly what is expected, assuming
the conduction and valence band edges are discontinuous at the interface. For the case of degenerate Ge and non-degenerate
GaAs, these
discontinuities are approximately 0.56 and 0.32 eV
respectively. The forbidden band in Ge appears

EXPERIMENTS

ON

Ge-GaAs

to move to a region of higher energy as the doping

decreases. The discontinuities for non-degenerate
Ge and non-degenerate GaAs appear to be 0.15
and 0.55 eV, respectively. There is some evidence,
however, which suggests that the discontinuity
in conduction band is somewhat larger than this.
All the diodes tested had lower rectification
ratios than have available homodiodes. However,
unlike the case of homodiodes,
no minoritycarrier storage effects were observed for these
heterojunctions
upon switching from a state of
forward to reverse bias.
The short-circuit current as a function of input
photon energy shows the Ge absorption spectrum.
Tunnel heterodiodes have been fabricated which
have 1-v characteristics between those of Ge and
GaAs tunnel homodiodes.

Acknowledgements-The
author wishes to thank Miss
ANNE BENORIC of the IBM Corporation,
Yorktown
Heights, New York, who fabricated most of the heterojunctions for his study. Gratitude is also expressed to
J. C. MARINACE, M. J. OROURKE, J. A. SWANSON,
P. J. PRICE, M. I. NATHAN and W. P. DUMKE for many
helpful discussions.
The author wishes also to thank
J. BALTA-ELfAs and J. F. GARcfA DE LA BANDA who made

HETEROJUNCTIONS

351

their laboratories at the Consejo Superior de Investigaciones Cientificas

(Madrid)
available to him, and
Professor BARCEL~ for aid in the electro-optical measurements.

REFERENCES
1. See, for example, A. VAN DER ZIEL, Solid State
Physical Electronics Chaps. 12 and 13. PrenticeHall, Englewood Cliffs (1957).
2. S. POGANSKI, Halbleiterprobleme
I (Ed. by W.
SCHOTTKY) p. 275. Friedr. Vieweg, Braunschweig
(1954).
3. W. SHOCKLEY, U.S. Pat. 2,569, 347.
4. H. KRAMER, PYOC. I.R.E. 45,1535 (1957).
5. D. A. JENNY, hoc. I.R.E. 45, 959 (1957).
6. R. P. RUTH, J. C. MARINACEand W. C. DUNLAP, JR.,
J. Appl. Phys. 31, 995 (1960).
7. J. C. MARINACE, IBMJ.
Res. Dev. 4, 248 (1960).
8. R. L. ANDERSON,IBM 1. Res. Dev. 4. 283 09601.
9. A. VAN DER Z&L, Sol;d State Physical Eiectr&ics
p. 284. Prentice-Hall,
Englewood Cliffs (1957).
10. W. SHOCIUEY, Bell Syst. Tech. J. 28, 235 (1949).
11. H. C. TORREY and C. A. WHITMER, Crystal RectiJiers.
McGraw-Hill,
New York (1948).
12. C. SAH, R. W. NOYCE and W. SHOCKLEY, Proc. I.R.E.
45, 1228 (1957).
13. R. L. ANDERSONand M. J. OROURKE, An. fis. Quim.
57. 3 (1961).
14. W. G. S&TZE& and J. M. WHELAN, Phys. Rev. 114,
59 (1959).
15. J. I. PANKOVE,Phys. Rev. Letters 4,454 (1960).