Note: Descriptions are shown in the official language in which they were submitted.
1alti7610
The present invention relates to a method of
producing p-n junction semi-conductor solar cells. More
particularly, the invention relates to a method of
producing solar cells with the principal objective of
sharply reducing production costs by depositing poly-
crystalline silicon on a relatively cheap substrate such
as metallurgical-grade silicon, graphite or steel.
The problem of uncovering new, abundant, cheap
and non-polluting sources of energy is a problem of vital
national importance. Of all energy sources, solar energy
is one of the more attractive sources because of its
abundant supply and because it is completely non-polluting.
An indication of the abundance of solar energy is evident
by the fact that the solar power on the surface of the
earth is approximately one hundred thousand times greater
than the current power consumption from all energy sources.
Presently, solar energy is utilized by converting
` solar energy to thermal energy and by converting solar energy
to electricity which is known as the photovoltaic system.
Both methods of utilizing solar energy are expected to aid
in meeting the ever-increasing demand for clean solar
energy. Currently, the silicon solar cell is the most well-
;, known device in the photovoltaic system. Further, technology
has advanced to the point where silicon solar cell panels
which are capable of producing several kilowatts of power
have been ~ed reliably in all types of space craft for manv years.
-- 2 --
. ' ~
10~7610
Currently, silicon solar cells are manufactured
by preparing polycrystalline silicon by reducing trichloro-
silane with hydrogen, growing single crystals of silicon of
controlled purity from the polycrystalline material,
preparing silicon wafers by cutting the single crystal
ingot to a thickness of at least 0.25 mm followed by polishing
and etching, diffusing a dopant into the silicon wafers to
form a shallow p-n junction, applying ohmic contacts to the
rear surface and grid contact to the diffused surface,
applying antireflecting and protective coatings to the
diffused surface and finally mounting the cell into position.
This rather intricate procedure results in the current
high costs of manufacturing silicon solar cells. Although
the costs of production for single crystalline solar cells
has recently been reduced from about $100/peak watt to about
$20/peak watt, further reduction in coct of about one order
of magnitude is necessary if widespread utility of solar
cells is to be realized in large-scale terrestrial applicationsO
One prior art process of manufacturing semi-
conductor solar cells as shown by Tarneja, et al, U.S. Patent
3,460,240 issuedAugust, 1969involves epitaxially depositingsili-
con ona quaxtzsubstrate toform an N-type layer over which
is epitaxially deposited two P-type silicon layers. However,
this process has the disadvantage that the overall process
requires the rather detailed and expensive sequence of
'
iO~7~1V
steps necessary to deposit epitaxial silicon so that no
significant decrease in cost of manufacture is observed.
The Jones reference, U.S. Patent 3,078,328,issued
February,1963shows a methodof manufacturingsolar cellsin which
a layerof siliconis grownonto a graphitesurface froma silicon
melt and doped to form an N~type layer. In this growth
step, silicon and carbon at the interface of the silicon and
graphite layers mix to form an intermediata layer of
silicon carbide. The device is completed by formation of a
top p-type layer of silicon by diffusion. The reference
again is disadvantaged by the complicated fabrication
procedure. Thus, the cost of manufacture is unattractive
. ~.,
from a commercial viewpoint.
Small-area polycrystalline silicon solar cells
have also been fabricated by the deposition of silicon from
a vapor state reactant. A polycrystalline silicon layer of ;~
a thickness of 25 - 50 ~m was deposited on silicon substrates
at 900C by the reduction of trichlorosilane with hydrogen.
In this method, silicon substrates were used for convenience
,:
~:~ in order to eliminate the cracking of silicon which has been
deposited on other substrates. By this procedure, 1 cm
solar cells were fabricated by the successive diffusion of
gallium and phosphorous to form a p-n junction about 2.5 ,~m
below the surface of the device. The device had a maximum
" ,, :
~ open-circuit voltage of about 0.3 V, and the greatest
:. ,,
~ ~ efficiency was about O.9o/O.
.
; - 4 -
, . .
"~
,' .
1()~i7610
A need continues to exist, therefore, for
a method by which silicon semiconductor solar cells
can be easily and c~eaply produced.
Accordingly, one object of the present
invention is to provide a method of manufacturing
polycrystalline silicon semiconductor solar cells
readily and cheaply.
Another object of the invention is to
provide a method of bonding polycrystalline sili-
con to common substrates of substantially different
chemical composition so as to form silicon semi-
, ~
~ conductor solar cells.
. .:
;~ Briefly, these objects and other objects
~ of the invention, as hereinafter will become more
.
readily apparent, may be obtained by a low-cost
polycrystalline silicon solar cell, which compris-
.
es: a steel base; at least one overlying layer of
polycrystalline silicon of p-type conductivity and
at least one overlying layer of polycrystalline
silicon of n-type conductivity so as to define a
, .............................. .
; p-n junction; and a diffusion barrier of borosili-
? cate interposed between the steel base and the first
polycrystalline layer.
;
Preferably, said p-n junction is formed
R by first and second overlying polycrystalline sili-
.
, ,
~ 30
' ,. _ S r~
.:
iO~;7610
con layers both of p-type conductivity and a
third overlying layer of polycrystalline sili-
cone of n-type conductivity.
It is also preferred that the first
overlying layer is lQ - 40~m thick and has a
resistivity of 0.002 - 0.003 ohm-cm, the second
overlying layer is 8 - lO~m thick and has a
resistivity of 0.2 - 2 ohm-cm and the third over-
lylng layer is 0.2 - 0.4~ m thick and has a re-
sistivity of 0.001 - 0.002 ohm-cm.
A more complete appreciation of the
invention and many of the attendant advantages
thereof will be readily obtained as the same
becomes better understood by reference to the
following detailed description when considered
: i
~ in connection with the accompanying drawings,
;~ wherein:-
Figure 1 shows the resistivity pro-
file of the p-region of polycrystalline silicon
:. ,.:.
20 solar cell device supported on a graphite
.i. :.::
substrate;
Figure 2 shows the current-voltage
. . .
~` characteristics of a mesa diode formed from a poly-
crystalline silicon, p-n junction device supported
on a graphite substrate;
Figures 3A to 3F show one embodiment of
~` polycrystalline silicon, p-n junction devices support-
~ ed on a graphite substrate;
.,:
'`''
- - 6 -
.~ .
. .
-- .
: : . ,
. ~: ' , . . .
:,, , ' ~ . . :
llD67610
Figure 4 is a graph of the current-
voltage characteristics of one of the devices
of Figure 3;
Figure 5 is the resistivity profile of
the p-region of a polycrystalline silicon junction
device supported on a borosilicate coated steel
substrate;
- Figures 6A to 6F show one embodiment of
the configuration of polycrystalline silicon, p-n
.~ 10 junction solar cell devices supported on a sub- strate of borosilicate coated steel;
Figure 7 is a schematic diagram of an
apparatus for the zone-melting of silicon on gra-
phite substrates;
; Figure 8 is a schematic diagram of an
apparatus for the unidirectional solidification of
:
silicon on graphite substrates;
: Figure 9 is a graph of the dark current-
: voltage characteristics of a n -silicon/p-silicon/p -
~: 20 silicon (unidiretionally recrystallized)/graphite
~ solar cell;
: Figure 10 is a graph of the current-
~: voltage characteristics of a n -silicon/p-silicon/
.~ p -silicon (unidirectionally recrystallized)/
: graphite solar cell under illumination with a
quartz-halogen lamp equivalent to AMO
.. i
,: ` ,
.,: .
~ ~ - 7 ~
,; ,, .
,
. . , , ' .
10~761V
.
conditions;
Figure 11 is a graph of the current-voltage
characteristics of a n -silicon/p-silicon/p+-silicon
(unidirectionally recrystallized)/graphite solar cell
under illumination with a quartz-halogen lamp equivalent
~- to AMI conditions; and
Figure 12 is a graph showing the relation
between short-circuit current density and open-circuit
.:
; voltage of a solar cell on metallurgical silicon/graphite
~ 10 measured under different illumination levels.
,
An important consideration in the manufacture
of silicon semiconductor solar cells is the type of support
. .,
- structure employed. The ideal support should be one which
has the characteristics of being low in cost, light-weight,
.,
chemically inert to silicon at high temperatures, having
; a high electrical conductivity and a coefficient of
-: .
s expansion similar to that of silicon. Furthermore, the
substrate should be such that silicon strongly adheres to
- the surface of the substrate, and the interface between the
~-~ 20 substrate and the silicon layer should be of low electrical
resistance. In the past, refractory metals such as tantalum
: .
have been used as substrates for the deposition of poly-
~` crystalline silicon of high purity. Such substrates,
.::
,:
however, have not been satisfactory because, from an
economical point in view, they are not competitive. The
`, choice of an appropriate substrate having the above-mentioned
. :.
:'
:- ~
.. :~ - .
:.
. .. . ,. ,, - . -. , . . ~ ~
iO67610
advantages would be a significant factor in lowering the
costs of silicon solar cells, thus rendering their use
more attractive.
The high cost of manufacturing silicon solar cells
from semiconductor grade silicon has already been discussed.
High costs are involved by the series of process steps
involved and the high cost of some of the manufacturing
steps in particular. In view of these problems, the use of
, , ,
~ a thin layer of polycrystalline silicon containing a p-n
~, ~
junction deposited on a suitable substrate should
substantially reduce material and process costs. Although
the efficiency of polycrystalline silicon solar cells cannot
compete with the efficiency of single crystalline cells, the
. .
....
unit power costs would be many times less than that of
:`
present single cyrstal cells.
It has now been discovered that polycrystalline
,
silicon can be satisfactorily deposited on a steel, graphite
`~ or metallurgical polycrystalline silicon substrate to form
a suitable silicon solar cell. The p-n junction must be
~,. .
very shallow; the n-layer (or the upper layer) is 0.2 - 0,4~m
in thickness, and the p-layer (or the inner layer) is
10 - lOO~m in thickness. The order of preference of
substrates is: metallurgical-grade silicon, graphite, and
steel~ Although most steels will serve the purpose,
metallurgical grade silicon and graphite are preferred.
x In the preparation of solar cells using either a graphite
:.''
g _
' :',
: :;
~ ~ .
.. . .
~C~t;7610
or polycrystalline silicon substrate, polycrystalline
silicon is deposited on an appropriate substrate by any
acceptable procedure normally used for the deposition
of polycrystalline silicon, such as by the thermal
decomposition of silane, by the reduction of trichloro-
silane or silicon tetrachloride with hydrogen at
temperatures ranging from 900 C - 1200C or by the thermal
decomposition of dichlorosilane. Several such layers are
deposited and appropriately doped so as to form a p-n
junction device by any conventional procedure known to
one skilled in the art. In one embodiment, the first
deposited silicon layer is doped with a p+- type dopant
to achieve a p+-type layer 10 - 30~m thick and having
a resistance of 0.007-0.003 ohm-cm. Doping occurs
; simultaneously with the deposition of silicon by introducing
the dopant gas, e.g., diborane for p-type conductive layers
and phosphine for n-type conductive layers, in proper amount
into the reactant mixture. Thereafter, two successive
silicon lay~rs are deposited which are p-type and then ;
n-type successively. The p-type layer is 8 - 10,4m thick
:
~,~ and has a resistance of 0.2 - 2 ohm-cm, while the n-type
~i layer is 0.2 - 0.4~m thick and has a resistance of 0.001 -
. ~:
~ 0.002 ohm-cm. FIGURE 1 shows the resistivity profile of
i~ the structured device, except for the n-layer, which is
` obtained by the spreading resistance technique. The plot
:
shows a high resistivity area in the silicon adjacent the
,.
- lQ _
,~: ..
: ........ ' ' . :
. . . . .
10~;7fà~(1
graphite-silicon interface and i~ apparently caused by
the formation of silicon carbide.
In one embodiment of the invention, the silicon
devices above can be converted into mesa diodes by
.
depositing titanium-silver or gold dots on the silicon
~:.
surface followed by masking and etching of the surface.
~ .
- Titanium-silver is the preferred contact material.
FIGURE 2 shows the current-voltage characteristics of
one of these diodes which is exemplary of the values
obtained for these diodes, wherein the "n" value calculated
from the forward characteri9tics is 1.9, which is very
similar to that for single crystal silicon p-n junctions.
This value indicates that the grain boundaries in
polycrystalline silicon deposited on graphite supports
are not significant in current conduction.
.:,
By the procedure shown above, so7ar cells can be
fabricated wherein the thickness of the polycrystalline
silicon layer of the cell ranges from 0.001 to 0O005 cm
which is at least 100 times less than the amount of silicon
~.:
employed in single crystal solar cells. The most significant
advantagesare that the energy-, labor-, and material-
.:. ~ .
~ consuming steps of the process of producing single crystal
. .~
cells, i.e., the growth of single crystals,the
preparation of the silicon wafers and the diffusion process,
: ~ .
are eliminated. Although the efficiency of the poly-
crystalline solar cells is less than that of single crystal
-- 1 1 --
iOti7610
cells, the disadvantage i9 more than countered by the
many times lower unit power costs of the polycrystalline
cells.
A similar procedure to that shown above can be
,
used to prepare polycrystalline silicon solar cells having
a support structure of metallurgical polycrystalline silicon
by any method known to those skilled in the art of
establishing p-n junctions in deposited polycrystalline
silicon. In a representative procedure, a p-type poly-
crystalline layer is deposited on a wafer of silicon,
which has been pretreated with hydrogen or with hydrogen
. . ~ ,
chloride at an elevated temperature ranging from 1000C
to 1100C by the decomposition of silane with diborane as
~: ........................................................................ .
~ the dopant source. Thereafter, an upper n-type silicon layer
r '. :~ ~
~ is deposited using phosphine or arsine as a dopant to yield
,
a structure having a shallow p-n junction. The device is
completed by attaching leads to the n- and p-regions.
In another aspect of the invention, the fabrication
of the cheap silicon solar cells by the decomposition of
polycrystalline silicon on a steel substrate was attempted.
The problem of using steel as a substrate for polycrystalline
silicon cells is complicated by several problems. The first
,
, ; is that a large and therefore significant difference exists
. . .:;
in the thermal expansion coefficients of the materials.
; Silicon has an expansion coefficient of 4 x 10 C-l, while
::
most steels have coefficients of 14 - 16 x 10-6OC
1 ,
- 12 -
.:.
,' ~
:. - :
10167610
Another factor is that high rates of diffusion of silicon
and iron atoms occur into each other between adjacent
layers of steel and polycrystalline silicon at the
temperatures employed in the deposition process. It is
known from the phase diagrams of mixtures of iron and
silicon that the following iron silicides form: Fe3Si,
Fe5Si3, FeSi and FeSi2. Still another factor is that
at temperatures above 800C, a solid solution of iron
silicides forms when silicon is deposited on steel
substrates by any of the conventional silicon deposition
procedures. This latter fact is substantiated by the
. .
high electrical conductivity of the region.
In view of the above problems with steel as
a substrate for silicon solar cells, it is apparent that
another material must be applied between the deposited
silicon layer and the steel substrate. Any such material
(hereinafter referred to as the "diffusion barrier"~ must
.,
have the ability to minimize the difference in thermal
expansion characteristics of the two layers and should be
~;20 chemically inert to steel and silicon at high temperatures.
' Further, the intervening layer should strongly adhere to
,;, .
both the steel and polycrystalline layers and be compatible
with the manufacturing technology of the polycrystalline
silicon. Thus, for example, the diffusion barrier may be
`~ deposited by in-situ chemical vapor deposition prior to
. .~
'
- 13 -
, ~ .
'' '"~;
.":,- .
. . ., ~ , .
. :. , ^ .. ..
~0~;7~i10
application of the silicon layer.
It has now been found that layers of films of
silica, borosilicate, phosphosilicate, aluminosilicate
and combinations thereof form suitable diffusion barriers
between steel and silicon. If the diffusion of boron is
objectionable from a borosilicate layer to the silicon
layer, a borosilicate-silica double layer is necessary
between the steel and si1icon layers.
Silica, borosilicate and phosphosilicate are
.. . .
all inert toward silicon and steel at high temperatures,
and have the added advantage of being relatively soft.
~ For instance, silica and silicon have hardness values on
-~ the Moh's scale of 5 and 7, respectively. ~he thermal
expansion coefficient of borosilicate can be changed
as a function of its composition which makes its use
:
particularly attractive. For example, a borosilicate
, .
- composition containing 18 mole % boron oxide has a
thermal expansion coefficient similar to that of silicon.
In polycrystalline silicon devices which have a
borosilicate/steel base, borosilicate is amorphous and
is not expected to significantly influence the structure
, j ,
,~. of the silicon deposit. Silicon which is deposited over
.
borosilicate has properties similar to silicon which is
deposited over silicon nitride or silicon dioxide.
Various factors such as substrate temperature, deposition
:: .
~ rate and the amount of boron incorporated substantially
.:
:
. ,~.:
~ - 14 -
::
;
10f~76~V
influence the micro-structure of the overlying silicon
layer. Generally, silicon which is deposited at low
temperatures and high rates without intentional doping
consists of small crystallites, the size of which
increase with increasing deposition temperature,
decreasing decomposition rate and the incorporation of
high concentration of boron, for example, a concentration
of about 10 boron atoms/cm3. Undoped silicon film
deposited at 900C at a rate of 2~m/min reveals a
fiber-like structure. However, as the deposition rate is
decreased to 0.2~m/min, small crystallites of silicon
less than l,~m in size become dominant. Further, a
notable increase in the size of the crystallites occurs by
further increasing the deposition temperature to 1000 C.
Both silica and borosilicate can be readily
.
deposited on steel substrates by a variety of known
chemical deposition techniques of which the following
are exemplary: Silica can be deposited by (1) the oxidation
of silane, or (2) by the pyrolysis of a tetraalkoxysilane,
. " ~
;- 20 such as tetraethoxysilane. Borosilicate can be deposited
(1) by the oxidation of a mixture of silane and diborane,
or (2) by the pyrolysis of a tetraalkoxysilane-trialkoxy-
;
borane mixture such as tetraethoxysilane-trimethoxyborane.
Naturally, the amounts of boron and silicon in the
.. .
borosilicate are determined by the mole ratios of the
silicon reactant and the boron reactant. Other suitable
- 15 -
: .~
.~ . I
. .
10~7~i10
alkoxyboranes include triethoxyborane and tripropoxyborane,
and other suitable tetraalkoxysilanes include tetramethoxy-
silane and tetrapropoxysilane. The methods by which the
silica and borosilicate diffusion barriers are deposited
on the steel substrates are not critical, and any deposition
method well known to those skilled in the art is applicable.
Phosphosilicate layers can be deposited by any suitable
technique such as by the oxidation of silane and phosphine
mixtures or by the hydrolysis of a silicon tetrachloride-
phosphorous trichloride mixture. All of the above methods
~, . .
for depositing the above diffusion layers are compatible
with current methods of depositing polycrystalline layers.
Normally, the thickness of the deposited diffusion barrier
is 1 - 5,um. Normally, the borosilicate diffusion layers
contain from 10 - 18% boron while phosphosilicate layers
contain from 10 - 2~/o phosphorous.
By using the above diffusion barrier materials
it is possible to prepare integral and tightly formed
structures such as silicon/silica/steel, silicon/borosilicate/
: .;
steel and silicon/silica-borosilicate/steel which completely
avoid contamination by highly conductive iron silicide.
. ~, .
Suitable steels used for the substrates of the present
solar cells include Armco silicon steel and USS Vitrenamel 1,
which has a carbon content less than 0.008%. Prior to
. ,.:
deposition of the first polycrystalline silicon layer to
eventually form the necessary p-n junction, the surface of
the steel can be scavenged of oxygen by treatment with a
. .~
- 16 -
~ . ~ . ., - :
10~7ti10
reducing agent such as hydrogen at a temperature from
600 to 1000C, preferably 900 to 1000C. Thereafter,
the appropriate silica, borosilicate or phosphosilicate
layer can be deposited on the steel surface.
Having generally described one aspect of the
invention, a further understanding can be obtained by
reference to certain specific examples which are provided
herein for purposes of illustration only and are not
intended to be limiting unless otherwise specified.
- 10 EX~MPLE 1
' Figure 3A shows a graphite base and Figure 3B
shows the structural configuration of a solar cell having
;~ a graphite base 1, over which was deposited a p+- silicon
layer 2 by the thermal decomposition of silane and diborane.
Thereafter, p-silicon layer 3 and n+-silicon layer 4 were
deposited by the thermal decomposition of silane using
diborane and silane as dopants, respectively. Grid contacts
of Ti-Ag-Al 5 were evaporated onto the n+-siiicon surface
as shown in Figures 3C and 3D. Figures 3E and 3F show
another view of the device in which excess silicon deposits
: ,:
were removed from the periphery of the device. The current-
; voltage characteristics of the device were measured with an
Aerospace Control Corp. Model 302+AMO solar simulator at
room temperature, and the efficiencies achieved for a
series of cells were in the range of 1.0 - 1.6%.
'
- 17 -
: ' -
',
. .
10~7~10
Figure 4 is a plot of current density versus
voltage for one of the above polycrystalline silicon on
graphite solar cells. The open-circuit voltage, short-
circuit current density, and fill factor obtained are
0.3~ V, 13 m~/Cm , and 0.44, respectively. A value of
0.85 ohms was obtained for its series resistance from the
current-voltage relations under several different levels
of illumination.
EX~MPLE 2
A 6 cm2 disc of p-type polycrystalline silicon
having a resistivity of 0.001 ohm-cm was obtained from
metallurgical grade silicon. The disc was heated under
a hydrogen atmosphere at 1000C. A p-type, 10,~ m thick
:;
layer of polycrystalline silicon was deposited by thermally
decomposing silane with diborane as the dopant to produce
a resistivity of 1 ohm-cm. Thereafter, an n-type layer
:;
.....
0.2 - 0.4,~m thick and having a resistivity of 0.002 ohm-cm
was deposited by thermally decomposing silane and phosphine
to yield a shallow p-n junction device. Aluminum contacts
.:~
were attached to the p- and n-regions. The device was
subjected to light from an AM0 solar simulator and an open
:. .~,,
`~ circuit voltage and short-circuit current density of 0.52 V
and 15 mA/cm2, respectively, were obtained, corresponding to
a conversion efficiency of 3%.
; Deposition of Borosilicate of Steel Substrates:
.,
~ - 18 -
..
.
",.. - ~ ~ ,,
.~.. . .. .
10~i7~;10
0~6 - 1.5 mm thick low carbon U.S. Steel Vitrenamel steel
plates having a linear thermal expansion coefficient of
1.4 x 10 5OC in the temperature range of 25 - 700C were
used as su~strates (Note that silicon has an expansion
coefficient of 3.6 x 10-6C 1,), The steel substrates
were placed in and supported on a silicon car~ide coated
graphite susceptor and the susceptor was heated exothermally
by an rf generator. Before silicon was deposited on
the wafers, the steel substrates were heated under a
- 10 hydrogen atmosphere at 1000C to remove oxygen from the
,
steel surface. A film of borosilicate of a thickness
ranging from 3 - 5~4m was then deposited at 900C from a
reactant mixture of 20 l/min hydrogen, 25 ml/min silane,
6 ml/min diborane and 250 ml/min oxygen.
Several polycrystalline solar cells were then
prepared by depositing silicon on steel substrates coated
with 3 - 5,~m of borosilicate. The first silicon layer was
;,' ~
~ p-type of 10 - 40~m thickness and of a resistivity of
; 0.002 - 0.003 ohm-cm, while the second layer was of p-type
--, 20 silicon of 8 - 10,4m thickness having a resistivity of
:
0.2 - 2 ohm-cm. The final layer was a layer of n-type
silicon of 0.2 - 0.4,~m thickness and 0.001 - 0.002 ohm-cm
resistivity. The resistivity profile of a device having
a first p-silicon layer of 35f~m thickness of 0.002 ohm-cm
: :.
;` resistivity, a second p-silicon layer of lO,~m thickness of
0.2 ohm/cm resistivity and a final n-silicon layer of
', .
.
.~ - - 19 -
;......... . .. .
, ' ' ~ ' '' - '
,
~0~;7~1V
0~2f~m thickness of 0.001 ohm/cm resistivity on a
borosilicate/Vitrenamel I ~ub~trate i9 shown in Figure 5.
EX~MPLE _3
Several solar cells were fabricated from the
steel/borosilicate structures de~cribed above a~ shown
in Figure 6A. Because borosilicate was used as the
diffusion barrier, it was necessary to place the ohmic
. .
contacts in the n- and p-regions of the device on the front
surface of the silicon. A 2 mm wide strip 15 of the upper
L0 n - and p-silicon layers was removed from the periphery
of the device as shown to expose the p+- layer 12, which is
.
above the borosilicate layer 11 and steel substrate 10, but
below p-type layer 13 and n-type layer 14 (Figure 6B). A
:. ~.. . . .
~ thick aluminum film 16, 1 mm wide was then deposited onto
.
~ the exposed p+-layer, Figures 6C and 6D. The device was
:-~
:~ then annealed at 500C for 20 min. to form a low-resistance
ohmic contact. Thereafter, an aluminum grid contact 17 was
... . .
~ evaporated through a metal mask onto the n+-layer, Figures
:
6E and 6F.
O The electrical properties of the cells are
:
exemplified by a particular cell having the following
. .
configuration: 0.2~hm thick - 0.001 ohm-cm n-silicon/5~m
thick - 3 ohm-cm p-silicon/15,4m thick - O.003 ohm-cm p-
silicon on borosilicate/Vitrenamel I. The p~-layer was
::. i,'
~' deposited at 1000C, and the other layers were deposited
at 900C. The average grain size was 2,5~m. The p-n+junction
,:
- 20 -
.,-
' ~
..
iOtj7610
was 4.4 cm in area The current-voltage characteristics
were determined under illumination by a tungsten lamp.
The power density of the surface of the device was estimated
to be 80 mV/cm2. The open-circuit voltage was about
0.1 V, and the conversion efficiency was about 0.05%.
EX~MPLE 4
A USS Vitrenamel I steel plate 1 mm thick was
supported on a silicon carbide coated graphite susceptor in a
55 mm ID fused silicon reaction tube. An rf generator was
used to heat the susceptor externally. A 3 ~m thick
; borosilicate layer was deposited on the substrate by
- decomposing a mixture of hydrogen, silane, diborane and
oxygen at flow rates of 20 l/min, 25 ml/min, 6 ml/min
and 250 ml/min, respectively, at 900C. Thereafter, a
layer of silicon 55~m thick was deposited over the
borosilicate by decomposing a mixture of hydrogen and silane
at flow rates of 20 ml/min and 50 ml/min, respectively,
at 900C. Both the silicon and borosilicate layers tightly
adhered to the underlying substrates. Further, metallurgical
~20 examination of the cross-section of the deposited layers
:~:
~; showed that no diffusion of iron occured into the silicon,
~; and similarly, no diffusion of silicon occurred into the
~ ':
steel substrate. The silicon layer overlying the boro-
silicate layer was of the n-type having an electrical
?
resistivity of about 450 ohm-cm. The resistivity data
. .
indicated that borosilicate effectively functions as a
, .
''
_ 21 -
. '
,'`.'~
,: .
: 10~7~0
diffusion barrier against the diffusion of iron from the
steel substrate into silicon.
EX~MPLE 5
Following the procedure of Example 4, a boro-
silicate layer of 3 - 5~4m thickness was deposited over a
, Vitrenamel substrate. A 30~m layer of silicon was
subsequently deposited over the borosilicate layer by
decomposing silane in hydrogen at 1150C. Subsequent
.~ metallurgical examinations showed that the silicon/boro-
silicate/steel structures effectively prevented the diffusion
. :
' of iron into the silicon layer at 1150C.
.
i EX~MPLE 6
; ,,
A 7,um thick film of silica was deposited on a
steel substrate at 900 C by decomposing a mixture of
"~ hydrogen, silane and oxygen at flow rates of 20 l/min,
25 ml/min and 250 ml/min, respectively. ~hereafter, the
temperature of the substrate was increased to 1150C, and
::
about 30~m of silicon was deposited. In Examples 5 and 6,
i metallurgical examinations showed that no iron diffused from
the substrate into the silicon layer.
According to another aspect of the invention,
the conversion efficiency of polycrystalline silicon solar.
cells on graphite substrates can be substantially increased.
Extensive experimentation has been carried out to produce
~, such cells generally having the configuration 0.2 - O.4~4m
` O.001 - O.002 ohm-cm n-silicon/10-15,4m 0.5-2 ohm-cm
~.
- 22 -
.,
10676~0
p-silicon/10-30~m 0.002-0.005 ohm-cm p-silicon/graphite,
i.e., a low resistivity p-type silicon layer was first
deposited on graphite followed by the depositions of a
medium resistivity p-type silicon layer and a low
resistivity n-type silicon layer. Solar cells with open-
circuit voltages of up to 0.35 V and short-circuit current
densities of up ~ 13 mA/cm2 were obtained under illumination
with an air mass zero solar simulator, corresponding to a
conversion efficiency of about 1.5%. This relatively low
- 10 efficiency is associated with the relatively small size of
crystallites, less than 30 micrometers on the average, in
a polycrystalline silicon layer deposited by conventional
chemical vapor deposition techniques. To improve the con-
version efficiency of polycrystalline silicon solar cells,
the crystallite size in silicon must be substantially
increased, and this invention provides a simple technique
to improve the conversion efficiency o polycrystalline
;,.-., :
' silicon solar cells on graphite substrates.
.~11
` Since the solidification of a silicon melt usually
yields relatively large crystallites, the recrystallization
of the low resistivity p-type layer prior to the deposition
of the medium resistivity p-type and low-resistivity n-type
silicon layers can improve considerably the structural
properties of silicon solar cells. ~owever, the recrystalli-
I
~ zation of a large area silicon layer of a few micrometers
- thickness on graphite substrates is extremely difficult
- because of the large surface tension of molten silicon
''', ,
,.
_ 23 -
,,`' '.
10f~76~0
will cause the thin molten silicon layer to breakdown into
discontinuous filaments. To ~tabilize the molten silicon
layer, the silicon melt-substrate interfacial energy must
be increased to overcome the ambient-silicon melt interfacial
:,
energy associated with the large surface tension of molten
silicon. The silicon-substrate interfacial energy can be
increased by increasing the roughness of the substrate
surface, for example, by sandblasting. Thus, the recrystalli-
zation of the silicon layer may be carried out by first
depositing a low resistivity p-type polycrystalline silicon
layer on a roughened graphite surface, followed by melting
and solidification. Two approaches have been used for the
recrystallization of silicon on graphite: (1) zone-melting,
and (2) melting of the entire area followed by unidirectional
solidification. The zone-melting process was carried out
by using the apparatus shown schematically in Fig. 7.
A p+-silicon/graphite specimen of 15 cm. length and 4 cm.
width was placed in a fused silica tube in a hydrogen flow
and was heated at 1200 - 1300 C by an rf generator. The
.~ .
20 spacings between the turns of the rf coil were adjusted
so that a 2.5 cm wide region of the specimen was above the
melting point of silicon. The specimen was then pulled to
allow the molten zone to traverse through the length of
the specimen. Although zone-melting has produced
polycrystalline silicon layers with relatively large
.
~ - 24 -
10fà7610
crystallites, this process involves the movement of the
specimen and may not be economical for the fabrication of
low cost silicon solar cells. The stationary recrystalli-
zation process is therefore more desirable. This method
is complicated by the fact that the density of the liquid
silicon is about l~/o higher than the density of the solid.
The regions which solidify last will have a protruded
surface due to the expansion of silicon on solidification.
It is therefore necessary that the solidification of molten
silicon be controlled to take place from one end of the
specimen to the other. This was achieved by adjusting
the spacings of the rf coil so that there is a unidirectional
temperature gradient along the length of the specimen, as
shown in Fig. 8. The unidirectional recrystallization of
. ., :
silicon on graphite is simple in operation and has produced
silicon sheets with crystallites as large as two centimeters
.: :,1
in length and several milimeters in width. Most crystallites
are of a [110] orientation as shown by x-ray diffraction.
; An alternate and more economical approach for the
:.:
preparation of silicon sheets on graphite substrates is to
use metallurgical-grade silicon, at a cost of about $1/kg,
: .:
as the starting material. Metallurgical-grade silicon is
about 98% purity, and a major portion of the impurities
in metallurgical silicon can be removed readily by the
chemical treatment of the melt with chlorine or other
reagents. Instead of depositing a low resistivity p-type
. .
. .,:
- 25 -
' ,;
. ~. .. . ......
~0~7610
polycrystalline silicon layer on a graphite surface as
discussed a~ove, the purlfied metallurgical silicon is
melted on a graphite plate and unidirectionally soli-
dified as shown in Fig. 8. This approach is the most
inexpensive one at present for providing silicon sheet
with large crystallites suitable for solar cell purposes.
Solar cells can be prepared by depositing a
; silicon film containing a p-n junction on a p+-silicon~recrystallized)/graphite or a chemically treated
metallurgical silicon (recrystallized)/graphite substrate.
Two examples illustrating the solar cell fabrication
techniques discussed are given below.
EXAMPLE 7
~- A p-type silicon layer of 0.002-0.003 ohm-cm
resistivity was deposited on a roughened graphite substrate
at 1150 C by the thermal reduction of trichlorosilane with
hydrogen. This layer was then recrystallized by uni-
directional solidification, and the recrystallized layer
contained elongated crystallites of up to 2 cm in length
and several millimeters in width. Subsequently, a 10-30 ~m
of 0.2-1 ohm-cm p-type silicon layer and a 0.2-0.4,~m of
0.002-0.004 ohm-cm n-type layer were deposited to complete
~ the solar cell structure. The grid contact consisting
;~ of about 500 R titanium and 5,~m silver was then evaporated
onto the as-deposited surface in the usual manner. The
solar cell produced by this technique showed relatively
'.'~
:
- - 26 -
,
10~7610
good dark current-voltage charac~eristics as shown in
Fig. 9, where the area of the cell was about 20 cm2.
These characteristics are considerably better than those
of solar cells described in the previous disclosure.
Figure 10 shows the current-voltage characteristics of
this cell under illumination with a quartz halogen lamp
calibrated to AM0 conditions. The open-circuit voltage,
; short-circuit current density, and fill factor were found
to be 0.52 V, 17 mA/cm2, and 0.55, respectively,
~10 corresponding to a conversion efficiency of about 3.7%.
~` EX~MPLE 8
Metallurgical-grade silicon was purified by
:,,
treating the melt with chlorine to volatilize about 5% of
silicon in the form of silicon tetrachloride During this
treatment, many impurity elements, such as boron and aluminum,
::; .
react preferentially with chlorine to form volatile chlorides.
,~ .
;~ Chemically treated metallurgical silicon was melted on a
; graphite plate and unidirectionally recrystallized. The
resulting metallurgical silicon/graphite substrates were then
used for the deposition of silicon solar cell structures by
; ~ the thermal reduction of trichlorosilane, as discussed in
. ....................................................................... . .
Example 7. The characteristics of a solar cell of about
32 cm area with the configuration 1000 A SiO2/0.3-0.4 ~m
0.003 ohm-cm n-silicon/30,hm 1 ohm-cm p-silicon/metallurgical
: .:
silicon (unidirectionally recrystallized)/graphite are shown
` ~'
- ::
-- 2 7
' ~
: .
. ~ ... . .. ...
7610
in Figs. 11 and 12. Yigure 11 shows the characteristics
of the cell under illumination equivalent to AMI conditions.
The open-circuit voltage, short circuit current density,
and fill factor were found to be 0.53 V~ 17.5 mA/cm2, and
0.56, respectively, corresponding to an AMI efficiency
of 5%. The relation between the short-circuit current
density and the open circuit voltage measured under
various intensities of illumination is shown in Fig. 12.
The "n" value in the diode equation and the saturation
current density calculated from this plot are approximately
1.76 and 1.2 x 10-7 A/cm2, respectively. In some smaller
- area solar cells (10 cm2 for example), AMI efficiencies
higher than 6% have been observed.
.. ~.~ . .
Having now fully described the invention, it
will be apparent to one of ordinary skill in the art that
- many changes and modifications can be made thereto without
departing from the spirit or scope of the invention as set
forth herein.
.
',: .
:
:
: ,.
, ~ ,
, . ,
:,
~ - 28 _
':-~
:
; :.' ~ : .. -
