Note: Descriptions are shown in the official language in which they were submitted.
~3~
Back~round Art
Solar cells have been de~e.loped ~or g~nera~ing
electrical energy directly from sunllght. Xn
g~neral, these cells can be classified as either
heterojunction devices, which depend upon junctions
such as thos~ ~ormed between two different semi-
conductor materials or between a metal and a semi-
conductor or from a metal~insulator~semiconductor
. sandwich, an~ homojunotion devices which depend
only upon junctions formed between layers of the
same ~emiconductor material doped to different
impurity le~s to provide different electrical
propert.ies.
~37~
-- 2 --
~ 3eretofore, homojunction cells using direct-g~p
semiconductor materials have generally exhibited dis-
appointing efficiencies. One reason for the relatively
low efficiencies in homojunction solar cells is be-
5 lieved to be the high absorption coefficient whichis inherent in direct gap semiconductor materials
such as gallium arsenide. For example, approximately
half of the carriers due to AM 1 radiation are gen-
erated within 0.2 ~m of the surface of gallium ar-
10 senide. Therefore, for materials such as GaAs, whichalso has a high surface recombination velocity, most
of the carriers generated by solar radiation recombine
before they reach the junction causing a significant
decrease in conversion efficiency.
One approach which has been used to overcome
this problem has been the use of a thin window layer
of gallium aluminum arsenide (Gal_xAlxAs) grown
over the GaAs wafer by liquid phase epitaxy.
Such cells may be referred to as heteroface cells.
20 Because the recombination velocity is much less at
a Gal_xAl ~s/GaAs interface than at a GaAs sur-
face, higher conversion efficiencies have been
achieved. Thus, Hovel and Woodall report conversion
efficiencies of up to 22~ for Gal_xAlxAs/GaAs hetero-
25 face solar cells but only up to 14~ for GaAshomojunction solar cells for air mass i (AM 1~
radiation. See Hovel and Woodall, J. M., 12th IEEE
Photovoltaic Specialists Conf., 1976 (Institute of
Electrical and Electronic Engineers, New York, 1976),
30 p. 945.
Nevertheless, aluminum is so reactive in the
vapor phase that it is difficult to prepare hiyh
quality Gal xAlxAs layers by conventional chemi-
cal vapox deposition, which is a highly preferred
~l3~
fabrication method. Because of this, it has been
necessary to grow Gal_xAlxAsx layers by metal-
organic chemical vapor deposition. See Dupuis,
R. D., Dapkus, P. D., Yingling, R. D. and Moody,
5 L. A., Appl. Phys. Lett., 31, 201 (1977). This
method can be both more expensive and more time
consuming than conventional chemical vapor depo-
sition.
Disclosure of the Invention
This invention relates to improved shallow-
homojunction photovoltaic devices and methods for
their fabrication. These shallow-homojunction de-
vices are based upon a plurality of layers of direct
gap semiconductor materials suitably doped to pro-
15 vide an n /p/p structurè. An antireflection coat-
ing is applied over the n top layer and the nt
top semiconductor layer is also limited to a thick-
ness within the range which permits, upon light
irradiation, significant carrier generation to
20 occur in the p layer below the n+ top layer.
Thus, the top junction is referred to as a shallow-
junction. p~ef~ r'ec~
In a particularly ~r~ fabrication method,
the antireflection layer deposited on the n+ top
25 semiconductor layer is formed by anodization. The
anodic layer forms an excellent antireflection
coating and requires no ~acuum processing. Addi-
tionally, the anodization process serves to reduce
the thickness of the top semiconductor layer without
30resulting in significant surface degradation.
Thus, homojunction solar cells can be provided
which have conversion efficiencies approaching
those obtainable with heteroface cells. Fur-
-- 4 --
thermore, the use of a Gal_xAlxAs layer is avoided
which eliminates a relatively complicated and ex-
pensive step in the overall fabrication of a
solar cell.
Still another significant advantage of the
photovoltaic devices described herein is the ease
with which ohmic contacts can be applied to them
since they have high concentrations of dopants in
their outer layers. This is particularly important
10 because shallow homojunctions can have their junction
qualities destroyed by attempts to form ohmic con-
tacts at elevated temperatures. With this device,
ohmic contacts can be applied by directly plating
a metal layer on the semiconductor surfaces without
15 elevated temperatures so that the junction quality
is preserved.
These shallow homojunction solar cells addi-
tionally possess vastly superior resistancé to
degradation by electron bombardment than either
20 Gal xAlxAs heterojunction cells or shallow homojunc-
tion GaAs cells having a p top layer. Because of
this, the shallow homojunction cells described herein
have great potential for use in space applications.
The heavily doped p+ layer additionally enables
25 excellent solar cells to be grown on substrate mat-
erials other than the host semiconductor materials,
i.e., the semiconductor material which absorbs sun-
light and generates electrical current.
Gallium arsenide solar cells, according to this
30 invention, have been grown on gallium arsenide and
germanium substrates with equally outstanding effi~
ciencies o~ over 20% at AM 1.
Brief Description of the Drawings
FIG. 1 is a cross sectional elevational view of
a typical prior art solar cell employing a thin Ga1_x
AlxAs window;
FIG. 2 is a cross-sectional elevation view il-
lustrating one embodiment of a solar cell according
to this invention;
FIG. 3 i~ a cross-sectional elevation view of
another embodiment of a solar cell according to this
10 invention;
FIGS. 4(a) and 4(b) are schematic illustrations
of the application of gold contacts to a solar cell
of this invention;
FIGS. 5(a) and 5(b~ are schematic illustrations
15 of the application of tin contacts to a solar cell
of this invention;
FIG. 6 is an exploded cross-sectional view
illustrating the area around one contact finger of
a solar cell Labricated according to this invention;
FIG. 7 is a plot of data illustrating the
quantum efficiency at varying wavelengths for a
solar cell fabricated according to this invention;
FIG. 8 is a plot comparing measured reflectivity
to theoretical reflectivity for an anodic anti-
25 reflection la~er applied to a GaAs shallow-homojunc-
tion device according to this invention;
FIG. 9 is a plot of data illustrating the
quantum efficiency at varying wavelengths for a
solar cell of this invention having an anodic anti-
30 reflection coating;
FIG. 10 is a graphical presentation of the
impurity profile for a GaAs shallow-homojunction
photovoltaic device of this invention grown on a
germanium substrate;
FIG. 11 is a plot of data illustrating the
~3~
spectral response of a GaAs solar cell of this
invention having an n+ layer thinned by an anodi-
zation-strip cycle;
FIG. 12 is a plot of data for the spectral
response of a GaAs solar cell of this inventio~, and
comparin~ internal and external quantum effici~ncy;
FIG. 13 is a plot of data illustrating th~
power conversion efficiency as a function of n~
layer thickness for four GaAs solar cells of this
invention;
FIG. 14 is a plot of data illustrating the
decrease in IsC with increasing electron irradi-
ation fluences for a GaAs shallow-homojunction
solar cell of this ir~vention;
FIG. 15 is a plot of data illustrating the
decrease in IsC, VOC, fill factor and power
conversion efficiency with increasing electron
irradiation fluences for another GaAs shallow-
homojunction solar cell of this invention.
20 Best Mode of Carrying Out the Invention
The invention will now be further described
with particular reference to the Figures.
A prior art homojunction photovoltaic device
10 is illustrated in FIG. 1. The substrate 12 i5
25 formed from an n GaAs wafer. Typically, substrate
12 might be doped to a carrier concentration of
1016-1017 carriers/cm3. Layer 14 is formed over
layer 12 and comprises p GaAs. 1ayer 16 is formed
from p+ Ga1_xAlxAs and might have a carrier con-
30 centration of 1018 carriers/cm3. A typical thicknessfor layer 16 is less than one micrometer. This
Gal_xAlxAs window has been used because the re-
combination velocity fox carriers generated upon
solar illuminhtion i5 much less at the Gal_xAlxAs/
~l~3'76~3~
GaAs interface than it is within GaAs itself. In
practice, a device of FIG. 1 can be formed by
depositing a thin layer 16 of p+ Gal_xAlxAs over
an n GaAs substrate and subsequently diffusing some
5 of the p-dopants from the Gal_xAlxAs layer into
the GaAs substrate to form layer 14.
Unfortunately, Gal_xAlxAs coatings are
difficult to apply using chemical vapor deposi-
tion and are relatively difficult to use in the
10 formation of an ohmic contact. Because of this,
ohmic contacts are typically applied to devices
such as that shown in FIG. 1 by employing vacuum
coating techniques followed by alloying. These
techni~ues are relatively expensive and detrimental
15 to shallow-homojunctions.
FIG. 2 illustrates one embodiment of an im-
proved photovoltaic device of this invention.
Therein, substrate 22 is formed from a wafer of
p+ GaAs. Substrate 22 might be formed, for
20 example, from GaAs suitably doped with p-dopants
such as zinc, cadmium, berylium or magnesium to a
carrier concentration of at least about 1018 carriers/
cm3. The thickness is not critical for substrate 22,
but might be between about 1 and about 500 um. Layer
25 24 is formed on the upper surface of layer 22 and
is a layer of GaAs suitably doped with p-dopants
to a caxrier concentration of about 1014-101~ carriers/
cm3. The thickness of layer 24 depends upon the
minority carrier diffusion length and absorption
30 coefficient, with a typical range for GaAs of from
about 1 to about 5 ,um.
Layer 26, formed from n+ GaAs, is epitaxially
deposited upon layer 24. Layer 26 may be formed from
GaAs suitably doped with n-dopants, such as sulphur,
35 selenium or silicon to a carrier concentration of at
~l~3~
-- 8 --
least about lol7 carr ers/cm3. It ls critical to limit
the thickness Qf lay~r 26 to one which allows sig-
nificant carrier generation within layex ~4. Thus,
layer 26 would typically be limited to a
5 maximum thickness of 1500 A, and preferably less.
Great care is necessary to assure that such
thin layers are uniform, and not all deposition
techniques are suitable. The techniques believed
to be suitable include chemical vapor deposition,
10 molecular beam epitaxy, liquid phase epitaxy and ion
beam implantation~ In addition, if care is taken
to dope the top layer 26 to a sufficiently high
concentration, an ohmic contact can be formed on
the surface without degrading junction character-
15 istics.
It has been found that the n /p/p structurehas significant advantages over previously employed
structures. The p+ layer, for example, forms a
back surface ield junction with the p layer to
20 provide high efficiency current collection. The
high doping level in the p~ layer also simplifies
the application of an ohmic contact thereto and
this high doping level also allows the shallow-
homojunction cell to be formed on a different sub-
25 strate material which reduces the cost or provides
other advantages. In ~his regardl the heavily doped p
layer allows tunneling between any heterojunction
formed between dissimilar materials thereby making
ohmic contact feasible.
The n+ top layer reduces the series resistance
of solar cells having this structure. It also sim-
plifies the formation of an ohmic contact to the
cell because of its high doping level.
The p layer also provides a major advantage in
~3~
the n /p/p structure. Greater cell efficiencies
are possible with this structure compared to other
shallow homo~unction structures, such as p /n,
because of the greatly increased diffusion length
of minority carriers, i.e., electxons, in a p layer
compared to the diffusion length of minority carriers
in an n layer, i.e., holes,
The overall cell structure of n /p/p thus
provides a shallow-homojunction device which can
10 be manufactureed inexpensively, is capable of pro-
viding high efficiencies, can be deposited on sub-
strates formed from different materials, and has
outstanding resistance to degradation by electron
bombardment which is a severe problem encountered
15 in space applications.
Top layer 28 is an antireflection coating which
reduces the reflection of GaAs and thus increases
absorption of solar energy. The antireflection
coating might be, for example, successive layers
20 o~ transparent materials having relativ~ly high and
relatively low indices of refract:ion, respectively.
For example, an antireflection coating might be
prepared by electron-beam evaporation of titanium
dioxide and magnesium fluoride.
It is particularly preferred to apply the anti-
reflection coating by anodization. Application of
anodic coatings can be done without any vacuum
processing and is an inexpensive way of producing
excellent antireflection coatings. In addition,
the application of an anodic layer necessarily
reduces the thickness of the top GaAs layer. For
example~ it has been found that application of an
anodic layer typically reduces the thickness of the
GaAs layer at the rate of about 2/3 part of the
35 volume of GaAs to one part by volume of anodic layer.
~l37~
-- 10 --
Typically, the thickness of the antireflection layer
would be based upon quarterwave theory.
Anodic coatings can be formed by employing
the device as the anode in an electrolytic cell. By
proper selection and control over the cell parameters,
including the electrolyte and voltage applied, thin
uniform anodic coatings can be formed Suitable
electrolytes are well known and a specific one found
to be suitable is a solution formed by mixing 3 grams
of tartaric acid into 100 ml of water, adding suffi-
cient N~40~ to adjust the pH to about 6 2, and then
adding 250 ml of propylene glycol. The voltage
applied should be sufficient to produce the anodic
coating in the thickness desired.
Although the description above has been
limited to GaAs cells, other direct bandgap semi-
conductors such as InP and CdTe are also suitable.
Additionally, the direct bandgap semiconductor layers
can be deposited on substrates formed from different
materials.
FIG. 3 illustrates a shallow-homojunction
device 30 ~o~med on p+ germanium substrate 32. The
thickness of substrate 32 might be from 0.1 ~m to
500 ~m and it might be ~ormed from single crystal
Ge doped with p-dopants to a carrier level of 1013
carriers~cm or greater. GaAs layers are then applied
to Ge substrate 32, by chemical vapor deposition or
other techniques, to form the desired shallow-homo-
junction device from p GaAs layer 22, p GaAs layer
24 and thin n~ GaAs layer 26, which are similar to
layers 22, 24 and 26, respectively, in FIG. 2,
Antireflection coating 28 is subsequently applied
over thin n+ GaAs layer 26 to complete this embodiment.
There is a major cost advantage possible in
the manufacture of GaAs shallow homojunction solar
cells when the actual gallium arsenide employed can
be minimi2ed by depositing the cell on a substrate
S of less costly material, such as germanium or silicon.
Gallium arsenide solar cells theoretically have
higher conversion eficiencies than cells formed
from indirect bandgap materials, such as silicon and
germanium. In addition, gallium arsenide cells poten-
tially should bP more radiation resistant in space en~vironments. By growing gallium arsenide cells on
materials such as silicon or germanium, which have
lower costs, the advantagesof both types of
materials can be achieved.
Substrates formed from materiàls differen~ from
the host semiconductor can offer other advantages in
addition to cost advantages. Germanium~ for ex-
ample, has higher thermal conductivity than GaAs
which is an advantage in heat dissipation. Germanium
20 also has a lower melting point and lower vapor
pressure than Ga~s, which might allow easier laser
crystalliæation on a substrate such as graphite.
Laser crystallized germanium having large grains
would pro~ide a good substrate for chemical ~apor
25 deposition of GaAs.
The use of substrates which are diffèrent lrom
the host semiconductor is possible because of th~
heavy doping of the substrate and the p layer o
the device. This permits tunneling to occur around
30 the heterojunction between the substrate and p
layer so that it does not act as a barrier. Thus,
a good ohmic contact oan be formed.
~37~
- 12 -
It should be understood that the embodimenk
shown in FIG. 3 is still considered to be a homo-
jun~tion cell, even though it technically contains
a boundary betwPen dissimilar materials, namely the
boundary between p 5e substrate 32 and p GaAs
layer 22. Although this boundary might techni-
cally be referred to as a heterojunction, the heavy
doping allows ohmic contact. This should be con-
trasted with the heteroface between p GaAs layer 14
and thin p Gal xAlxAs layer 16 in FI~. 1, which
serves the function of reducing surface recombina-
tion velocity of carriers generated in layer 14
upon solar irradiation. It should also be con-
trasted with a typical heterojunction between dis-
similar materials which is used to create a barrierto current flow in heterojunction devices. Because
of these differences in purpose, solar cell 30 il-
lustrated in FXG. 3, and other similar cells, will
be referred to herein as shallow-homojunction solar
cells formed from direct gap semiconductors deposited
on different substrate materials.
Although substrate 32 has been illustrated
to be single crystal Ge, other substrates could be
employed. Single crystal silicon, for example,
could also be employ~d. In fact, substrate 32 might
also be formed from polycrystalline or amorphous
materials, including silicon and germanium.
Electrical contact to the thin n+ layer can be
easily made because of the high doping level therein.
30 Contacts can be formed by electroplating metals
such as gold, tin, etc. Specific procedures for the
electroplating of these metals differ, however, and
are respectively illustrated in FIGS. 4 and 5. In
~376~
- 13 -
both of these figures, the shallow-homojunction solar
cell has the structure illustrated in FIG. 3.
In FIGS. 4(a) and 4(b), a typical application
of gold contacts is illustrated. In this technique,
5 the thin n+ layer 26 is first anodized to form
anodic coating 28 while simultaneously thinning n+
layer 26.
The anodization potential can be set to achieve
the appropriate thickness for the antireflection co~ting -
and n layer. A photoresist mask 30 is then placedover anodic coating 28 and finger openings 32 (FIG.
4[a]) are etched through anodic coating 28 employ-
ing an etch such as dilute hydrochloric acid. Gold
contacts 34 are then electroplated onto n layer
26 through photoresist mask 30. Photoresist mask
30 is then removed by dissolving it in acetone
to produce the device of FIG. 4(b~.
The application of tin contacts is illustrated
in FIGS. 5(a) and 5(b). Photoresist mask 30 is
directly applied to thin n~ layer 25 and tin con-
tacts 36 are electroplated onto layer 26 through
mask 30 (FIG. 5[a]). Photoresist mask 30 is then
removed and the thin n layer 26 is anodized (FIG. 5
~b]). When this procedure is employed, the thin n
layer 26 remains thicker under tin contacts 36 than
under the remainder of anodic coating 28. Thus, layer
2~ has raised shoulders 38 directly beneath tin
contacts 36 as can be seen in FIG~ 5tb). Thus,
there is a larger separation between the metal
contacts and the p-n junction for tin contacts
than for gold contacts. Because the n+ layer
is extremely thin after anodization, this in-
creased separation should imp~ove device
yield and reliability.
,,
.. .
':
~l37'6~L
- 14 -
The use of tin contacts is also advantageous
for optimizing the n~~ layer thickness since the
anodic oxide formed on gallium arsenide can be
stripped with dilue HCl and the cell reanodized
5 without removing the contacts. Because the
thickness of the oxide layer is very uniform
and easily controlled by adjusting the anodizing
voltage, a series of alternating anodization and
stripping steps can therefore be used for con-
10 trolled reduction of n+ layer thickness. Thethickness of the gallium arsenide removed during
each anodization can be accurately determined by
using elliposometry to measure the anodic oxide
thickness and multiplying this value by an appro-
15 priate factor
Some anodic antireflection coatings may besomewhat unstable in harsh environments. If this
is a problem, it can be overcome by application
of a thin (e.g., 100 A), trans~arent, protective
20 coating of a material such as SiO2 or phosphosilicate
glass. Such protective coati~c~s can be applied by
pyrolytic deposition techniques.
FIG. 6 is a cross-sectional view illustrating
one finger of a solar cell having such a protective
~5 SiO2 coating 40. ~evice fabrication is similar to
that described above for FIGS. 4(a) and 4(b), except
that a hydrofluoric acid etch is employed prior to
the hydrochloric acid etch. Contact finger 34 can
be formed from gold plated to a thic~ness of about
30 4 ~m. The back contact 42 can also be formed from
plated gold~ Although the SiO~ protective layer
was described as being applied prior to contact
formation, it could also be applied after the con-
tacts have been formed.
~3~
- 15 -
As those skilled in the art will recognize,
other metals could be used in place of gold and tin
for purposes of establishing electrical contact with
the photovoltaic device which is described herein,
In devices fabricated as illustrated in FIG. 4, any
metal could be employed including gold, silver, plati-
num, tin, aluminum, copper, etc. In devices fabricated
as illustrated in FIG. 5, those metals can be used
which form a sufficiently thick oxide layer during
anodization such that current leakage through metal
contacts is reduced enough to allow the semiconductor
surface to be anodized. Tin, aluminum and copper are
examples.
Devices of this invention have at least one
n-p homojunction and at least one other junction suffi-
cient to increase current collection. This other
junction requires an impurity profile wherein the
majority carriers all have the same charge and wherein
an electrical field is created by the impurity profile
which aids in collecting minority carriers. The p
layer also allows the use of substrate materials other
than the host semiconductor material. Specific
examples of suitable junctions include high/low
homojuncti~ns and graded pro~ile junctions where the
impurity doping level increases with distance from
the n-p junction.
This invention canbe further specifically
illustrated by the following Examp~es.
EXAMPLE 1
GALLIUM ARSENIDE SHALLOW-HOMOJUNCTION PHOTOVOLTAIC DEVICE
GaAs layers were grown in an AsC13-Ga-H2 system.
The reactor tube had an inner diameter of 55 mm,
and the H2 flow through the AsC13 evaporator and
.~
37~iO~L
-16-
over the Ga boat was in the range 300-500 cm3/min.
The p and n dopants were introduced in the vapor
phase by using ~C2~s)2Zn and H~S, respectively. The
reactor tube was vertical, allowing rotation of the
substrate, which resulted in greater doping uniformity
in the layersO Use of high purge flows allowed the
reactor tube to be opened at the bottom to load and
unload substrates without losing the ~2 atmosphere
inside the tube. Thus, the furnace could remain
at growth temperature during the loading procedure,
decreasing the cycle time between runs. Once inside
the reactor tube, the substrate could be preheated
in pure H2 just before being introduced into the
reactant gas flow at the growth posit.on. For a
more detailed description, see Bozler, C. O., Solid
State Research Report, 2, 52, Lincoln Laboratories,
M.I.T. (1975).
A p layer, 1.7 ~m thick, was first grown on â
p , Zn-doped (100)-oriented GaAs substrate with a
carrier concentration of 1018 carriers/cm3~followed
by a thin n layerO The p layer (p~ l~cm 3) and n+
layer (n-~ x1018cm 3) were doped with Zn and S,
respectively, by using (C2Hs)2Zn and ~2S sources.
The sheet resistance of the n+ layer w~s 7Q~/c.
To determine the thickness of this layer, the I-V
characteristic between two ohmic contacts to the
layer was measured while a channel was being
etched between the contacts. When the I-V char-
acteristic for back-to-back diodes was observed,
etching was immediatedly stopped, and the channel
depth was measured with a profilometer. The n~
layer thickness measured by this technique was 1300A.
The initial fabrication step following layer
growth was the pyrolytic deposition of SiO2 glass
~L9.37~
-17-
(1000 ~) on the GaAs wafer at 400C, in order
to protect the n+ layer during the succeeding
steps. Openings for ohmic contact fingers were
etched in the ylass coating using photclitho-
graphic techniques. There were 10 openings, 0.5cm long and 12 ym wide, spaced 1 mm apart~ The
wafer was sputter-etched to remove GaAs to a
depth of 40 ~ in the finger openings, then
sputter-coated with successive layers of Au-
12~ Ge (300 ~) and Au (2000 A). The Au/AuGe filmwas defined photolithographically into 25-lum-wide
fingers, interconnected at one end, that overlaid
the openings in the glass. All of the photolitho-
graphic steps were carried out with standard equip-
ment used for silicon wafer processing~ The waferwas then annealed under flowing N2 for one second
at 300C on a graphite heater strip to establish
ohmic contact between the AuGe fingers and the n+
layer, as verified by measurements of test con-
tacts on the wafer. The conventional techniqueof alloying at ~50C was not used because it was
found to cause penetration of t,he n layer, a~d
su~sequent destruction of the homojunction.
The contact ingers to the n~~ layer were
electroplated with Au to a thickness oS 4 ~n.
The back contact to the p+ substrate was made with
sputtered Au. The active area of the cell was de-
fined by etching a 1 cm x 0.5 cm rectangular mesa
in the Ga~s, and the glass layer was removed with
buffered HF. The fingers of the cell at this stage
had a cross sectional configuration similar to that
illustrated in FIG. 6, except that there was no
anodic layer 28. Finally, the cell was antireflec-
tion-coated with successive layers of Si0 (700 A)
~3~
- 18 -
and MgF2 (1200 A) ormed by electron beam evapora-
tion. For GaAs with this two-layer coating, the
average reflectivity measured over the 0.5-0.9 ~m
wavelength band was less than 5%.
An efficiency measurement of the cell was made
by using a high-pressure Xe lamp with a water filter
as a simulated AM 1 solar source. The incident
intensity was adjusted to 100mW/cm2, using a NASA
standard Si solar cell, c~librated for AM 1, as
a reference. The open-circuit voltage was 0.91 V,
the short-circuit current 10.3 mA, and the fill
factor 0.82, giving a measured conversion effi-
ciency of 15.3%. When the contact area is
subtracted, the corrected efficiency is 17~. The
15 n factor at 100 mA/cm2 is 1.25, as obtained from
the dark I-V characteristic, indicating good material
quality with long carrier diffusion lengths. The
series resistance is 0.5~.
The quantum efficienc~r of this cell as a func-
20 tion of wavelength is shown in FXG. 7, and, as can
be seen, quantum efficiency is hlghest at the
longer wavelengths, with a gradual decrease at
shorter wavelengths.
This cell was fabricated by sputtering and
25 alloying techniques, which although possible because
of the relatively thick n layer, are not preferred.
EXAMPLE 2
GALLIUM ARSENIDE SHALLOW-HOMOJUNCTION PHOTOVOLTAIC
DEVICE HAVING ANODIC ANTIREFLECTION COATING
A photovoltaic device was prepared as in Example
1 except that the antireflection coating was an
anodic coating, and all ohmic contacts were electro-
plated.
~3~
-- 19 --
The GaAs layer used was grown in an AsC13-Ga-H2
CVD system on p+ Zn-doped (100)-oriented substrate
with a carrier concentration of 1018cm 3. A p layer
about 2 ~m thick was first grown on the substrate
5 followed by an n+ layer (n~5 ~ 1018cm~3) were
doped with Zn and S, respectively~ by using
(C2H5)2 Zn and H2S sources. Following GaAs
growth, the n+ layer was anodically oxidized as
follows.
The electrolyte solution used for anodization
was prepared by mixing 3 g of tartaric acid with
100 ml of H20, adding sufficient NH40H to adjust the
pH to about 6.2, and then adding 250 ml of propylene
glycol. The final pH was 4.6-5.8. Anodization of
15 the GaAs was performed at room temperature, using
a platinum wire as cathode. A smooth anodic layer
of uniform thickness was obtained by using a
constant current source with a voltage limiter. The
source was set at a current corresponding to a
20 current density of about 750 ~A/cm2 ~or the GaAs
anode, and the maximum output voltage was set at
about 43 V. The current initially remained con-
stant until the voltage increased to its limiting
value, after which the voltage remained constant
~5 and the current decreased. Anodization was termin-
ated when the current fell to one-tenth of its ini-
tial value. The thickness (measured by ellipsometry
using a He-Ne laser) of the anodic layer was abo~t
20 A/V and did not depend strongly on current density.
30 The layer produced took less than 5 minutes, was about
850 A thick, and consumed about 550 A of the GaAs
layer. The anodic layer was stable up to at
least 250C in air. The optical constants were mea-
~7~
- 20
sured by ellipsomtery at 4358 and 5~61 A using a Hg
lamp and at 6328 A using a He-Ne laser. The values
of the refractive index n at these wavelengths are
1.91, 1.85 and 1.33 respectively, as shown in the inset
of FIG. 8. The values of extinction coefficient k
axe very low, and for the thickness of anodic layers
used, absorption of the optical constants (and the
effectiveness of the antireflection coating) is
illustrated in FIG. 8 by the close agreement between
the measured reflectivity spectrum of an 800 A thick
anodic layer on GaAs and -the value for this struc-
ture calculated using values of n obtained from
the curve shown in the inset of FIG. 8, k = 0
for the anodic layer, and bulk optical constants
15 for GaAs.
A layer of Au about 3 ,um thick was then electro-
plated on the p+ substrate as the back contact.
Photoresist AZ 1350J was spun on the anodic layer,
and photolithographic techniques were used to etch
20 openings for ohmic contact fingers in the anodic
layer. (The anodic layer dissolves readily in AZ
photoresist developer, as well as in HCl.) There
were 10 openings, 0.5 cm long and 12 ~un~ wide,
spaced 1 mm apart and interconnected with a bar at
25 one end. A layer of Au about 3 ~m thick was then
electroplated into the openings, and the wafer was
annealed in N2 for 1 sec at 300C on a graphite
heater strip to produce ohmic contact between the
Au fingers and the n+ layers. Finally, the active
30 area of the cell was defined by etching the GaAs
to form a l-cm x 0.5 cm rectangular mesa.
~3~
Efficiency measurements, using a high-pressure
Xe lamp with a water filter as a simulated AM 1
source, werc made. The incident intensity was ad-
justed to 100 mW/cm2, using a NASA-measured GaAs
solar cell as a reference. The cell was also
measured on the roof of the laboratory at an
ambient temperature of about 20C. The solar
flux density measured with a pyranometer was 98
mW/cm2, close to AM 1 conditions. The open-circuit
voltage was found to be 0.9~ V, the short-circuit
current 25.6 mA/cm2, and the fill factor 0.81,
giving a measured conversion efficiency of 20.~ per
cent, without correcting for the area of the
contact fingers. The quantum efficiency of this
cell as a function of wavelength is shown in FIG.
9. The quantum efficiency exceeds 90 percent at
the maximum, but it decreases quite strongly at
shorter wavelengths.
EXAMPLE 30 GALLIUM ARS~NIDE SHALLOW-HOMOJUNCTION PHOTOVOLTAIC
DEVICES ON GERMANIUM SUBSTRATES
.
The growth procedures and apparatus of
Example 1 were used, except as noted. The Ge
substrates were oriented (100) 2 off toward
(110) and were prepared by coating them with
Si02 on the backside to reduce Ge autodoping
of the GaAs 1ayexs during growth. The electron
concentration in nominally undoped layers deposited
on these coated substrates was 5 x 1015cm 3. The
lattice constants and expansion coefficients of
Ge and GaAs are well matched, a favorable condition
for obtaining good ~uality epitaxial layers~
~l37~
The doping profile used for the solar cells is
shown in FIG. 10. The p+ Ge substrate was highly
doped with Ga (8 x 1018cm 3) in order to overdope
any As that might diffuse into the Ge during the
5 deposition of the GaAs and to assure tunneling
through any thin barriers which could arise at
the heterojunction interface. The p~ GaAs buffer
layer was highly doped with ~n, again to assure
tunneling and also to overdope Ge diffusing into
10 the GaAs during growth. The change in hole con-
centration from 5 x 1018cm~3 in the buffer layer
to 1 x 1017cm 3 in the active layer provided a
backsurface field to increase the collection effi-
ciency. The n+ layers were doped to a carrier con-
15 centration of 5 x 1018cm~3 with sulfur, had asheet resistivity in the range 45-100~, and
their electron mobility was ~1000 cm2/V-sec.
An AR coating was produced on the n+ layer by
anodic oxidation, which consumed a thickness of
20 GaAs equal to 0.66 times the thickness of oxide
produced. The anodizing solution was prepared
by adding 3 g tartaric acid to 100 ml ~2~ adjust-
ing the pH to 6.2 with NH40H, and adding 250 ml
propylene glycol. The thickness of the oxide
25 layer was proportional to the limiting voltage
used for anodi~ation. The thickness required
for an optimum AR coating was 850 A, which waC
obtained for a limiting voltage of 43V~
Contact to the very thin n+ layer was made
30 easily because of its high doping level. Elec-
troplated Au formed ohmic contacts with a spec-
ific resistance of 8 x 10-5J~-cm2. Electroplated
Sn also formed ohmic contacts, although their
resistance was not measured. Sn had the advan-
~3~
tage t~l~.t in the solution used for GaAs anodi-
zation, Sn was also anodized, forming an oxide
resistive enough to allow the GaAs to be anodized
in the presence of Sn contacts.
Two different fabrication procedures were
used for cells with Au and Sn contact fingers,
as illustrated in FIGS. 4 and 5. For devices
with Au contacts, the n+ layer was anodized first,
finger openings were etched through the oxide
10 using a photoresist mask, and Au was plated
using the same mas~. For cells with Sn contacts
the Sn fingers were plated first, using a photo-
resist mask, the photoresist was then removed,
and the n+ layer was anodized. With this pro-
15 cedure the n+ layer was thicker under the Sncontacts than ~Inder the anodic oxide, so that there
was a larger separation of the metal from the p-n
junction than with Au contacts. Because the n+
layer is so thin, the increased separation was
20 believed to be better for device yield and reli-
ability.
A mesa etch of the GaAs was used to define the
active area of the cells, and the back contact to
the Ge substrate was made by Au plating. No alloy-
25 ing or ~acuum processing was used in cell fahrica-
tion.
Measurements of spectral response as a function
of n+ layer thickness were made on small cells, 0.05
cm2 in area, having two Sn contact fingers 0.5 mm
30 apart connected to a Sn bar at one end. The n~
layer, which was initially 2000A thick with a
sheet resistance of 45 Q~, was thinned by alter-
nate anodization and stripping. The external
quantum efficiency, which is the ratio of the
35number of carriers collected (Isc~q) to the number
l3~6C~
- 24 -
of incident photons, was measured after each of
three anodizations at ~3 ~, so that the cells
were antireflection-coated during each measure-
ment. The values of IsC and incident photon flux
5 were measured as a function of wavelength in a
spectrometer which was arranged so that all th`e
light fell between the two contact fingers. The
results for cell 1 are given in FIG. 11, which
shows that thinning the n+ layer results in a
10 marked improvement in quantum efficiency, espec-
ially at shorter wavelengths This is expected
because of the high absorption coefficients for
GaAs (104-105cm 1), which increase with decreas-
ing wavelength, and the high surface recombina-
15 tion velocity, which is believed to be around107cm/sec. The power conversion efficiency for
each n+ layer thickness is also given in FIG. 11.
These values were measured with the cell fully
illuminated by a simulated AMl source, with no
20 correction made for the finger area.
FIG. 12 shows the final spectral response
of cell 2, which was fabricated next to cell 1.
The n~ layer was sligh-ly thinner than that of
cell 1, and the response was therefore slightly
25 improved at the short wavelengths. The curve ~or
internal quantum efficiency, which is the ratio
ph"t,or,S
~;~ of Isc/q to the rate at which ph~t~ enter the
semiconductor, was o~tained from the measured
external efficiency ~y correcting for the spectral
30 reflectivity of the AR-coated cell. This curve
indicates that the cell design is very near the
optimum.
The AMl power efficiencies of cells 1 and 2
~' .
~,
17~
- 25 ~
and two other small cells fabricated side by side
on the same wafer are plotted in FIG. 13 as a
function of n+ layer thickness. In order to ob-
tain additional thickness values, cells 2, 3 and
4 were first anodized to 10, 20 and 30 V, respect-
ively, after which the oxide was stripp~d. All 4
cells were then anodized together to 43V to pro-
vide an AR coating, stripped, anodized to 43 V,
stripped again and once more anodized to 43 V.
10 Efficiency measurements were made after each 43 V
anodization. After the third such anodization~ the
efficient of cell 3 had dropped to 10% while cell
4 had essentially no sensitivity to light.
~ssuming that the cell output drops to zero
15 when removal of the n+ layer is completed, for
cells 1-4 complete removal would occur at a total
anodi2ation voltage between 149 and 159 V, the
~inal values for cells 3 and ~, respectively. It
was assumed for the purposed of plotting the data
20 that complete removal would occur at 154 V. Using
the removal rate of 13.3 A/V, the initial n+ layer
thickness was found to be 2050 A. To obtain the
thickness of the n~ layer remaining after each
successive anodization the total thickness removed
25 was calculated from the sum of the voltages used to
that point and subtracted from the initial thickness.
The thickness values given in FIGS. 11 and 12, as
well as those of FIG. 13, were obtained in this
manner. For all the points in FIG. 13 the fill
factor was 0.82 and Vn~ was 0.97 V. As the thick-
ness of the n layer is reduced below 200 A thick,
the efficiency drops precipitously from the maximum
value of 21.2%.
~3~
-26-
Seven larger cells, lcm x 0~49 cm, were made
from three different wafers. Either Au or Sn was
used for the contact finger pattern, which consisted
of 20 fingers 0.5 cm apart with a connecting bar
at one end. The fingers covered 4% of the total
area. One cell with Au contacts was partially
shorted and one with Sn contacts was thinned too
much. Power efficiency measurements, using a high
pressure Xe lamp with a wafer filter as a simulated
AMl source, were made on the other five cells. The
incident intensity was adjusted to 100mW/cm2 using
a NASA-measured GaAs solar cell as a reference.
Table 1 lists the measured values of VOC, ISc, fill
factor and efficiency, as well as the initial sheet
resistance and the total anodization voltage.
`~ Independent measurements ~ NASA Lewis Research Cen-
ter have confirmed these results. The sheet re-
sistance value gives some indication of the initial
thickness of the n+ layer; a value of 100 ~/O corres-
ponds to approximately 1200 A. The total anodization
voltage is the sum of the limiting voltages used in
a series of anodization~strip steps where the thin-
ing ratio is 20 A/V. For each cell the final ano-
dization was carried out at 43 V to provide the
AR coating. The conve-cion efficiency values, not
corrected for contact areas, are all in the 17 20%
range.
3~
--27--
~ o ~
L~ _
_ _ O o o o o o
E o ~ ~ o~
U V~ o~
~ ~ o o~ o~ L o~ o~
q i Z ~ ~ Ln ~ Ln Ln
j UJ j . ~ O `O ~ `O O
z~
-L~ _
>L~l~ L = ~ _ =
~_
J O
~l37~i~4
-28-
EXAMPLE 4
ELECTRON BOMBARD~ENT OF GaAs
_HALLOW-HOMOJUNCTION SOLAR CELLS
It has been reported that irradiation of Gal xAlxAs/
5 GaAs heteroface solar cells with electrons,causes
the conversion efficiencies of these cells to be
dramatically decreased. This is believed to be
partly because of the ~ ~ n lengths of minority
carriers in p and n layers decrease with increasing
10 electron irradiation, and partly because the sur-
face recombination velocity at the Gal Al As/GaAs
interface increases with increasing electron irrad-
iation. It was hypothesized that the n /p/p
structure of the cells described herein would
15 dramatically increase the resistance to cell
desradation under electron irradiation. If so,
the n /p/p GaAs shallow-homojunction solar cells
would be outstanding candidates for space appli-
cations, involving space vehicles, solar powered
20 satellites, etc.
To test the hypothesis, a series of experi-
ments was run in which shallow-homojunction GaAs
solar cells, as described herein, were irradiated
with 1 MeV electrons from electron fluences rang-
~5 ing from 1014 to 1016 electrons/cm2. The 1016electron/cm2 fluence is equivalent to dosages
which electronic devices would be subjected to
over 50 years in synchronous orbit.
In one experiment, a GaAs shallow-homojunction
30 solar cell prepared according to Example 1 and
having a thin n+ layer of about 1000A without an
antireflection coating thereon was subjected to
electron fluences of 0, 8 x 1014, 3.2 x 1015 and
1.0 x 1016 electron/cm2. The change in the short
35 circuit current, ISc, caused by these electron
6~
-29-
fluences was determined by integrating the quantum
efficiencies with the solar spectrum at air mass 0
(AM O), which is representative of space conditions.
FIG. 14 is a plot of Isc/(Isc)Owhere (ISc)o is the
5 original short circuit current before electron bom-
bardment. As can be seen, at a fluence of 1 x 1016
electron/cm2, the decrease in ISc was only about 7%.
The corresponding Gal xAlxAs/GaAs cell is reported
in the literature to undergo a 99% decrease in ISc
10 under a corresponding electron fluence. See l~alker,
C.E., Byvik, C.E., Conway, E.J., ~-leinbockel, J.I~.,
and Doviak, M. J., "Analytical and Experimental Study
of 1 MeV Electron Irradiated GaAl~s/GaAs Heteroface
Solar Cells," J. ~lectrochem. Soc.: Solid-State
15 Science and Technology, 2034-36 (1978).
FIG. 15 illustrates corresponding data for
a GaAs shallow-homojunction solar cell having a
thin n~ layer of about 1400 A with an anodic
AR coating under electron fluences ranging from
20 5 x 1013 to 7 x 1015 electron/cm~. This cell
was also measured under simulate~ AM O conditions
at the above range of dosages. As can be seen
from FIG. 15 where items having the subscript O in-
dicate values prior to electron bombardment, short
25 circuit current (ISc) decreased by about 20% at
7 x 1015 electron/cm2, the open circuit voltage,
VOC, also decreased gradually. The fill factor,
ff, actually increased slightly before it began to
decrease with increasing fluences. The conversion
30 efficiency~~ of the cell, which was about 14% at AM O
before electron irradiation, still had o~er ~0~
of its original efficiency after being bombarded with
7 x 1015 electrons/cm2.
~L~37~
-30-
Industrial Applicability
This invention has industrial applicability in
the fabrication of solar cells, particularly solar
cells for use in space applications.
Equivalents
Those skilled in the art will recogni~e, or
be able to ascertai.n using no more than routine
experimentation, many equivalents to the specific
embodiments of the invention described herein. Such
equivalents are intended to be encompassed by the
following claims.
