The violet cell: An improved silicon solar cell

Solar Cells, 29 (1990) 151 - 166

151

THE VIOLET CELL: AN IMPROVED SILICON SOLAR CELL* J. LINDMAYER and J. F. ALLISON

Communications Satellite Corporation, 950 L 'Enfant Plaza, S. W., Washington, DC 20024 (U.S.A.)

Summary State-of-the-art Si solar cells exhibit a poor q u a n t u m yield at short wavelengths, below 0.5 /~m the typical response drops sharply. Extensive work has resulted in an extension of the response to wavelengths as short as 0.3 ~um, significantly improving the solar cell current. The conversion efficiency has been further improved by an increased fill factor. The combination of a short-wavelength response and a sharper I - V curve has produced a conversion efficiency which is about 30% higher than that of state-of-the-art cells for space applications. The improved solar cell is called "violet cell."

1. Introduction

State-of-the-art Si solar cells perform considerably below predicted conversion efficiency limits. The projected efficiency of converting solar energy to electrical energy varies widely, depending on the parameter values assumed. For relatively long-lived minority carriers and thick layers, an upper bound of about 20% has been projected [1 - 3]. By contrast, we find that real space-quality n÷-p junction solar cells 300 /~m thick with a base resistivity of 10 ~2 cm exhibit a typical efficiency somewhat above 10% outside of the atmosphere. After these cells are irradiated by 1-MeV electrons to a level of 3 × 1014 electrons cm -2, the conversion efficiency falls to the neighborhood of 8.5%. Conventional cells are very limited in the short-wavelength region [4] and their diode characteristics are far from ideal. In a recent review paper, Wolf [5] calculated that, with a junction depth of 2000 A and a front surface recombination velocity reduced to the order of 102 cm s-1, the q u a n t u m yield could be significantly improved in the 0.45 - 0.6 pm range, resulting in a 17% improvement in the photocurrent. *This material w a s presented in part at the 9th IEEE Photovoltaic Specialists' Conference, Silver Spring, MD, May 2 - 4 , 1972. Reprinted with permission from Comsat Technical Review, 3 (1) 1973.

152 In fact, it can be shown that recombination in the Si crystal assumes controlling importance. There are four basic regions in which recombination mechanisms can be identified: (1) in the diffusion layer, recombination of photocarriers generated by short-wavelength light limits the blue-violet response of the cell; (2) in the space charge layer of the junction, recombination primarily affects the sharpness of the junction (fill factor); (3) in the bulk region, recombination of photocarriers generated by penetrating light affects the red response and limits photovoltage; (4) at the rear contact, interface recombination limits the IR response; the significance of this effect depends on cell thickness. This paper reports that, as a result of an independent study concerning the first two recombination mechanisms, the short-wavelength response can be extended and the diode characteristics improved to near-theoretical values. During this study it was found that recombination of photocarriers generated by blue-violet light is not controlled by front surface recombination. Instead, the front regime should be broken up into three regions: a shallow region with an extremely short lifetime, called the "dead layer"; a high-field region maintained by the impurity profile; and the actual space charge region. This model indicates the importance of minimizing the dead layer thickness and the recombination states appearing in the space charge layer so that the short-wavelength response can be extended over the entire solar spectral range and the diode characteristics can be made nearly ideal. The associated gain in current and fill factor are not subject to degradation from high-energy electron irradiation. The average conversion efficiency, specified with respect to total area, has been raised above 13% {With respect to active area, the corresponding efficiency is a b o u t 14%). This increased efficiency has not been obtained by increasing the minority carrier lifetime with associated radiation-sensitive red response (effective lifetime less than 10/~s). The operating efficiency after irradiation by 1-MeV electron to 3 × 1014 electrons cm -2 is 11.5%. This latter figure is quite independent of Si thickness, at least down to thicknesses of 150 pm. In view of the increased blue-violet response, proper simulation of the solar spectrum required special attention. In addition, as conventional antireflective coatings of SiOx and TiOx show absorption at short wavelengths, it was necessary to develop fully transparent coatings with appropriate refractive indices. In this respect, the dead layer model shows that the problem of varying surface recombination velocities associated with various coatings can be ignored. Instead, stress and optical requirements are the main considerations. The cells described here use tantalum oxide antireflective coatings, which have the required transparency and refractive index. The reduced depth of impurity diffusion used here to minimize the dead layer thickness significantly increases the lateral resistance of the cell's n-diffused layer. In addition, the better basic diode requires a lower than usual series resistance. To respond to these demands, the collecting metal

153 grid was changed profoundly to result in a new pattern containing about 60 fine lines over a 2-cm length. The new grid configuration is called fine geometry. As the short-wavelength response of the present cell is much better than that of conventional cells, it is called the "violet cell."

2. Fine geometry and diffusion The benefits offered by the violet cell cannot be realized w i t h o u t a major change in the grid collection pattern because of the high lateral resistance of the thin diffusion layer. The degradation of efficiency with series resistance can be estimated readily from the I - V characteristic: I = lo[exp(V/Vo)

--

i ] --Isc

(I)

where I 0 is the theoretical reverse current (a constant truly applicable in the forward direction only), V is the photovoltage, Isc is the short-circuit photocurrent, and V0 is the thermal voltage, which is equal to k T / q for the ideal diode. We note that, in the power-producing quadrant, I is negative, whereas Isc is positive by definition. From eqn. (1), the internal conductance of the cell is d/

Gi -

dY

-

Isc + I + I

V0

0 ~

-

I.¢--]I]

(2)

V0

We note also that the highest conductance occurs near the open-circuit voltage (I = 0). For high fill factors the current at the m a x i m u m power point is about Im ~ 0.95 Isc and V0 = 26 mV at room temperature. Hence, the conductance at the m a x i m u m power point is ... Isc --I/m[ --, 0.05 Isc _ Isc (mA) Gm V0 V0 520

(3)

The relative power loss, as opposed to the useful power in the load, is (for small losses). By restricting the loss to 1%, we find an approximate series resistance m a x i m u m of 0.032 ~ . This figure indicates a degradation rate of 12.5%1/0.1 ~2 in the fill factor. Hence, it can be seen t h a t a very low series resistance is needed if the ideal diode curve is to be approached. To optimize the fine geometry pattern, a simple parallel bar pickup pattern, shown in Fig. 1, will be analyzed. The bars are spaced at a distance d on a cell which has linear dimensions of do. It is assumed t h a t the width of a line is much smaller than the spacing between lines. Figure 1 shows two neighboring lines and the equivalent current source lines. If the resistance of the surface layer is denoted as R [] (~2 D-I), the series resistance becomes RsG m

R, =

d 1 1 R ° 4 d o 2m

where m is the number of lines (having two edges). As spaced lines,

(4) dold = m

for equally

154

Zero current line

I

Equivalent current source

do

u L [do ~1 Fig. 1. Simple illustrative aid for calculation of series resistance as a function of grid spacing. RD 1 R s - ~m2

(5)

The i m p o r t a n t conclusion is that the series resistance decreases with the square o f the n u m b e r o f lines used. In the conventional technology, six pick-up bars axe used (on the 2-cm × 2-cm cell) and the usual diffusion creates a resistance of a b o u t 50 ~2/ [3 -1 in the n+-p cell. F r o m eqn. (5), the series resistance would be about 0.17 ~2. In reality, however, experimental measurements indicate a resistance of about 0.25 ~2, suggesting additional c o n t a c t resistance. Clearly, the conventional g e o m e t r y c a n n o t be used when, for example, the sheet resistance is increased by an order of magnitude; in such cases, the series resistance would be prohibitively high. This work used a geom e t r y consisting of a b o u t 60 lines. An orderof-magnitude increase in the n u m b e r o f lines would provide a new degree of freedom. If, for example, the lateral resistance were an order o f magnitude higher ( ab o u t 500 ~2/D-1), this structure would allow for a series resistance in the h u n d r ed t hs of an ohm. With the pattern shown in Fig. 2, it was possible to hold the grid area obstruction loss to only 5 -7%. In addition, it was possible to maintain a negligible series resistance for the metal pattern itself.

155

Fig. 2. Fine geometry contact pattern.

The n ÷ layer was formed by diffusion of P into the Si. Diffusion time and temperature were mapped and correlated with the resulting junction depth. First, the usual junction depth of 4000 A was reproduced, and then cells were made with progressively decreasing junction depths of 2900, 1500, and approximately 1000 A. The majority of the cells finally used this last thickness, and it was found that with proper precautions a 1000-A junction depth is practicable and results in stable cells. The diffusion studies showed that the lateral conductivity of the diffused layer drops much faster with junction depth than linearly. At a depth of 1500 A, the sheet resistance is a b o u t 500 ~ / D -I. Another parameter monitored carefully during these studies was the sharpness of the I - V curve in the forward direction. Stresses and defects originating at the Si surface propagate into the crystal and create recombination sites in the space charge layer, causing deviation from the ideal diode characteristics with an attendant reduction in the fill factor [6]. It is well k n o w n [7, 8] that, in the usual diffusion process, the distribution of P does n o t follow a complementary error function distribution characteristic of simple diffusion processes. Instead, a nearly constant impurity concentration regime arises near the surface at the solid solubility level. Actually, the constant-concentration regime may be characterized b y a mixture of substitutional and interstitial P, with an accompanying very short minority carrier lifetime. Figure 3 shows typical P distributions for junction depths of 4000, 2900, and 1200 A. It can be seen that the 4000 A diffusion has an attendant heavily damaged layer of a b o u t 1500 A, b u t this "dead layer" diminishes in thickness as the diffusion depth becomes shallower. The critical concentration at which the dead layer begins to develop is the same order of

156 iO zl 4.2 x 10 20l 2,5 x I O Z O ~ - - - . 1.7 x I0 zO' iOgo N*= 6x 1015/cm2

1019

N*: 2 x 1015/cm2

N*: 4 1 1014/¢mz ]NJ IO is

iO 17

1016

zO~s O

I000

2000

3000 Depth (X)

4000

5000

Fig. 3. Diffusion profiles for P in Si for three junction depths (N* denotes the integrated impurity concentration).

magnitude as the surface density of silicon atoms, i.e. of the order of 10 is atoms am -2. Shockley has shown that dislocation generation is n o t only dependent on surface concentration, but is also affected by the total number of impurities (N*) found in a unit surface area [9]. A model advanced by Van Der Merwe demonstrates that above a critical misfit dislocations are created; the critical value for P in Si appears to be N* = 1.2 X 10 Is atoms cm -2 [10]. As indicated in Fig. 3, a great decrease in dislocation density will occur for the shallower junctions. Although it is generally true that the dislocation density decreases sharply toward the diffusion front, the total density of such dislocations is minimized by shallow junctions. Fine geometry allows for shallow junctions, thus improving the blueviolet response and also increasing the perfection of the diode by reducing the dislocations.

157

3. Model for short-wavelength response To compute the short-wavelength response, it is usually assumed t h a t a lifetime can be assigned to minority carriers in the diffused layer and that the presence of the surface can be taken into account by a surface recombination velocity. For the defects introduced by diffusion, the situation is somewhat different. Figure 4 is an energy diagram for an n+-p junction with a distance-dependent recombination state density. (The density of recombination states decreases with increasing depth into the bulk beyond the junction.) There are three easily distinguishable areas of importance: (1) Dead layer. In a conventional solar cell with a junction depth of 3500 - 4000 A, the width of the interstitial P layer is at least 1000 A. Photocarriers cannot be collected from this region where the diffusion transit time is longer than the lifetime

Dp

> Tp(n+-p cell)

(6)

where x0 is the width of the dead layer, Dp is the diffusion constant of holes, and rp is the lifetime of holes. If it is assumed t h a t the diffusion constant is about equal to one, the transit time is of the order of 10 -1° s. This layer is dead for lifetimes less than 100 ps. Such a short lifetime is possible if we remember that this is a degenerate region with a recombination state density of the order of the atomic density. (2) Field layer. In the region where the impurity concentration decreases with distance, the electric field is well above 10 kV cm -1. At such fields, the carriers approach saturation drift velocities [11] (107 cm s-1 for

dead layer

field layer

depletion layer

t

,

I

I

Bulk

---

--

-

-

--

~ :----_-:

-_:_.-_

-_-_._-__-.

1~

-_-_

-

-

)-_-_z : --=-_:

-.__ - - -

-_ _

~,,:

-

-

-

-

Intrinsic point

F i g . 4. E n e r g y b a n d d i a g r a m f o r n + - p j u n c t i o n .

-_-_

-

--

-

--

Ferm~level -

-

158 electrons and somewhat less for holes). Now the drift transit time may be compared with the lifetime. For the collection of carriers one needs XF - - < Tp(n+--p cell) p

(7)

where xp is the width of the field region and p is the drift velocity. The width of this region is about 2000 A in the conventional cell so that the drift transit time is of the order of 10 -12 s. In addition, the lifetime in this region is rapidly increasing as the associated state density drops sharply with decreasing impurity concentration. The conditions for collection are easily met, particularly if we realize that at the lower impurity concentration the lifetime may be of the order of 10 -6 s. (3) Depletion layer. In this region the field is very large, the drift transit is again of the order of 10 -12 s, and the lifetime may be as long as 10 -6 s. Although the conditions for photocarrier collection are clearly met, it must be recognized that the dislocations always advance ahead of the diffusion profile. This means that there are more recombination states in the depletion layer than in the bulk. The effect of such states is to increase the space charge recombination current of the diode; although their effect on the short-circuit current is insignificant, the fill factor is reduced [6]. The model was tested experimentally by matching the predictable dead layer quantum yield behavior with actual measurements and by enforcing different surface recombination velocities externally. Let us first describe some of the studies related to the surface recombination velocity. In many of the initial experiments an antireflective coating was grown by thermal oxidation of Si. Thermally grown SiO2 has an extremely good transparency (that of fused quartz); however, its index of refraction is low (n = 1.46) and therefore a 9% - 10% reflection remains at the quarter-wave minimum point. It is known from MOS field effect transistor studies that the surface state density changes with crystal orientation; it is highest on the (111) plane and lowest on the (100) plane. Surface recombination velocities as low as 100 cm s-1 have been reported for oxidized Si surfaces. If surface recombination is an important minority carrier loss mechanism in the Si cell, thermal oxidation and different surface treatments should affect the blueviolet response. Figure 5 shows some results obtained on the (100) plane. This figure indicates that the thinned oxide results in about 10% reflection at the matching wavelength and an improved short-wavelength response. After the SiO2 was completely removed, the q u a n t u m yield dropped to a level controlled by the reflection coefficient of Si. Exposure of the bare surface to moisture temporarily raised the q u a n t u m yield for all wavelengths with no change in the short-wavelength characteristics. Apparently moisture changes only the optical properties. Exposing the bare Si surface to a variety of coatings caused no significant change in the short-wavelength response. Similar results were obtained on the (111) plane. Because these results were obtained under very different surface conditions,

159 1.0

.

.

Thick Si02

._e >,

E

0.5

0

0 0.3

I 0.5

0.4

I 0.6

I 0.7

I 0.8

I 0.9

I 1.0

I.I

Wavelength (pro)

Fig. 5. Quantum yield per incident photon for two thicknesses of thermal SiO2 and bare Si surface in dry and moist ambients.

it appears that the magnitude of the surface recombination velocity must have changed greatly w i t h o u t clearly affecting the violet response. Such behavior is expected from the dead layer model. The quantum yield of a cell can be predicted readily from the dead layer model. We will make the simple assumption that all carriers generated b e y o n d x0 are collected, whereas those generated between 0 and x0 are lost. Then the yield for a particular wavelength is ~s

f e ax dx y

=

x9

=

e-aXo

(8)

f e ax dx 0

where ~ is the absorption coefficient for that wavelength. Figure 6 shows c o m p u t e d quantum yield curves for three different dead layer thicknesses (dashed lines) and the characteristics of a good conventional cell. The IR response was c o m p u t e d b y assuming that no recombination occurs at the back contact of the 200 pm cell. The curve associated with 1500 A of the dead layer was corrected for reflections arising from the antireflective coating. The solid line is the actual measurement of the cell, clearly indicating that the limited blue-violet response is the result of a dead layer thickness of 1500 A. It is interesting to note that fairly long wavelengths also suffer some losses. The SiOx coating matched at 0.6 pm holds the quantum yield at a b o u t 0.9, even for longer wavelengths. Figure 6 also shows that extension of sensitivity into the short wavelengths (e.g. below 0.4 pm) requires a very shallow dead layer and, accordingly, a very shallow junction.

160

50~

1.0

/

/

500 ~

No reflection

1500 ~

/~

\

",,z\.

0.5

'z

\t

J,}: 0,3

0.4

\t

0.5

0,6 0.7 0.8 Wovelengfh (Fro)

0.9

l.O

IJ

Fig. 6. Plot of calculated short-wavelength response using dead layer model.

Figure 7 shows the allowable dead layer thickness for a given cut-off wavelength. The cut-off is defined at 0.71 collection efficiency and the absorption coefficients are those available from the literature [12, 13]. Figure 7 indicates that an extended short-wavelength response requires a rapidly decreasing dead layer thickness or a very shallow junction.

4. Quantum yield of violet cell When it became apparent that the short-wavelength response could be extended significantly, the question of antireflective coating had to be reviewed in terms of blue-violet transparency. The widely used SiOx coating is quite absorbent at short wavelengths. When x ~ 2, it becomes less absorbent, but its index of reflection is low (n -+ 1.46). On the other hand, its refractive index can be increased by decreasing x (as x ~ 1, the index approaches two, which is still a marginal match), but its absorption is significant [14], as shown in Fig. 8. Although TiO: is a far better coating (it has a higher refractive index and less absorption), it also has a band gap of about 3.1 eV so that a sharp absorption sets in at about 0.4 pm, which limits quantum yield measurements below this wavelength. To provide the necessary bandpass, tantalum oxide was used on the violet cells reported in this work. The reflection plus transmission (R + T) of these three oxides is given in Fig. 8, together with the reflection curve for tantalum oxide covered with quartz. The q u a n t u m yield of a typical violet cell is shown in Fig. 9. The antireflective coating is matched around 0.5 pm, and as shown in the figure, the quantum yield is nearly equal to one in this region. A second curve represen-

161 ZOO0

1500

Conventional diffusion

_o

15

I00¢

,c

500

,,~m------- Diffusion region

of present work

o 0.2

0.5

I

I

0.4

0.5

Wavelenoth (~L~m)

Fig. 7. Maximum dead layer thickness as a function of cut-off wavelength for 71% response. ting the quantum yield after irradiation by 1-MeV electrons to 3 X 10 !4 electrons cm -~ indicates that the losses are restricted to the longer wavelengths. The short-circuit current adds up to 160 mA (2-cm X 2-cm cell). The bars with points indicate the current n o t collected for each 0.05-pm segment of the solar spectrum. Most of the loss occurs in the red and IR regions, portions of the spectrum which are most readily damaged by ionizing radiation. 5. Characteristics o f completed cell The cell characteristics reported here relate to 2-cm × 2-cm cells. Figure 10 shows the I - V characteristics of a recent cell. The AM0 radiation

162

I00[

/~..~

'°F/

.o//

o fj;

//

~,..2,

40

Reflection

0.2

0.5

0.4

0.5

0.6

0.7

0.8

/

0.9

1.0

Wavelength (/.Lm)

Fig. 8. Reflection plus transmission for three oxides.

ompleted cell

_

°I

Z

t'°°

~

d C rren

O 0.3

0.4

O. 5

0.6

i, T

0:8 0.7 Wavelength (~m)

i 0.9

0 1.0

I.I

Fig. 9. Q u a n t u m yield of typical violet cell before and after irradiation b y 3 X 1014 electrons c m -2 and a c c u m u l a t e d cell current and current lost per 0 . 0 5 ~tm as f u n c t i o n s o f wavelength.

e f f i c i e n c y o f this cell is 13.5% (over 14% in t e r m s o f t h e active area). A f t e r i r r a d i a t i o n b y 1-MeV e l e c t r o n s t o a f l u e n c e o f 3 X 10 la e l e c t r o n s c m -2, its actual e f f i c i e n c y is 11.5%. Such r a d i a t i o n a f f e c t s t h e cell p a r a m e t e r s as follows: change in c u r r e n t : --10% c h a n g e in voltage: --5% change in fill f a c t o r : 0% T h e fill f a c t o r f o r such a cell is in t h e n e i g h b o r h o o d o f 80% a n d is s o m e w h a t d e p e n d e n t o n t h e s p e c t r u m . I t can b e seen t h a t d i r e c t illumina-

163

180

I

Efficiency (%) 8 !

9

I0

II

12 13 14

Violet cell

160

After 3x 1014e/cmZ

140 Commercial c e l l ~

120

<

IOC

\

II I0

80 I SUN AMO

8

60

40

20

0 0

I I00

I 200

300 400 Voltage (mV)

500

600

Fig. 10. T y p i c a l I - V curves f o r a violet cell a n d a c o m m e r c i a l cell b e f o r e a n d a f t e r i r r a d i a t i o n b y 3 × 1014 e l e c t r o n s c m -2.

tion of the depletion layer b y energetic photons (at AM0) tends to desensitize the recombination levels, thus reducing the space change recombination current. Figure 10 illustrates the improvement in the overall power o u t p u t of an experimental violet cell (solid curves) when compared with a typical space-qualified commercial solar cell {dashed curves). The I - V curves were taken under illumination simulating AM0 conditions with a balloon flown cell used to standardize the simulator (Spectrolab X-25). The violet cell illuminated under terrestrial conditions exhibits an efficiency of 15.5%16%.

164

Figure 11 shows a series resistance of only 0.05 ~ deduced from I - V measurements at several intensities [15]. Figure 12 shows the efficiency as a function of increasing fluence of 1-MeV electron irradiation. The violet cell is again compared with the cell used widely in space applications, and shows an improvement of over 30% for higher fluences of irradiation. 6. Conclusions

A major change in grid geometry, coupled with very shallow junctions, has produced a major improvement in the short-wavelength response and the fill factor. Neither improvement is susceptible to degradation from ionizing radiation. The new cell is called the violet cell. 180

I

I

I

I

I

I

I

I

I SUN

,eo

m

20 mA 140

--

120 ~Rs= 0 . 0 5 ~ , - - ~ "

,

I00 " E o

015 SUN ~

80"--

,

' 20 mA

J 60--

40--

~0.13 SUN

0

I00

200

300 Vo'lt age (mV)

400

500

Fig. ~ 11/ Measurement of total series resistance under A M O density filters.

600

illumination using neutral

165 I

I

I

I IIIIII

I

I

13 ~

12-

II-

i,°I

= w

9

.-....

7

0

,,,~_ v

I

I

f t IIIIII

1012

I I014

1013 Radiation

I 1015

[0 Is

f l u e n c e (e/cm 2)

Fig. 12. Violet cell compared with commercial cell for radiation damage with 1-MeV electrons.

A new model has been developed for the front of such cells, explaining the short-wavelength cut-off in terms of a dead layer. It has been shown t h a t front surface recombination is n o t an important factor limiting the response. The improved diode characteristics (fill factor) have been explained qualitatively in terms of a critical integrated impurity concentration at the cell surface.

Acknowledgments This achievement could not have been accomplished w i t h o u t the invaluable assistance of many people. The authors are indebted to A. G. Revesz and J. H. Reynolds for m a n y helpful discussions and experiments. Special thanks are due to R. J. Dendall for contributions to the technology. The solar cells were fabricated with the assistance of F. Bland, A. Busch, and D. Martin, and the extensive measurements were carried out by I. Szabo. Appreciable contributions were made by C. Maag in terms of many discussions and measurements of the optical properties of the violet cell. The measurements of the impurity profile were done by M. Croset of Sescosem on the CAMECA ion probe, and the cell irradiations were carried o u t by A. Meulenberg. Special effort was expended by D. Curtin in providing an AM0 solar spectrum. Appreciation is due to E. S. Rittner for continuing discussions and encouragement, as well as to W. L. Pritchard, former Director of COMSAT Laboratories, for support t h r o u g h o u t the research and development program.

166

References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

E. S. Rittner, Phys. Rev., 96 (6) (1954) 1708. J.J. Loferski, J. Appl. Phys., 27 (1959) 777. P. Rappaport, RCA Rev., 20 (3) (1959) 373. F.M. Smits, K. D. Smith and W. L. Brown, J. Br. Inst. Radio Eng., 22 (1961) 161. M. Wolf, Conf. Record 8th IEEE Photovoltaic Specialists' Conf., 1970, p. 360. J. Lindmayer, COMSAT Tech. Rev., 2 (1)(1972) 105. J. C. C. Tsai, Proc. IEEE, 57 (9) (1969) 1499. R. A. McDonald et al., Solid State Electron., 9 (8) (1966) 807. H.J. Queisser, J. Appl. Phys., 32 (9) (1961) 1776. S. Dash and M. L. Joshi, Soc. AIME, Proc. Defects in Electronics Materials for Devices Conf., Boston, MA, 1969, 202. C. B. Norris and J. F. Gibbons, IEEE Trans. Electron Devices, ED-14 (1) (1967) 38. W. C. Dash and R. Newman, Phys. Rev., 99 (1955) 1151. H. R. Philipp and E. A. Taft, Phys. Rev., 120 (1960) 37. A.P. Bradford et al., Appl. Optics, 9 (2) (1970) 339. R . J . Handy, Solid State Electron., 10 (1967) 765.

Appendix Recent results indicate that the efficiency can be raised further. The actual efficiency is now 14% for o u t e r space conditions (AM0), when a solar constant o f 140 mW cm -2 is used and when the actual area of the cell is taken into account. It must be pointed out, however, t h a t at present there is no universally accepted qua nt i t y for the Sun's radiant power or agreement on the cell area t o be used in the c o m p u t a t i o n of efficiency. Early measurements o f the solar constant by J o h n s o n [A1] resulted in a value of 139.5 mW cm -2, whereas more recently Thekaekara [A2] has measured 135.3 mW cm -2. Additional confusion is caused by the f r e q u e n t use of only the exposed solar cell area in the c o m p u t a t i o n of efficiency (active area). In view o f this situation, our r ecent results are summarized as follows: AM0 spectrum, 140 mW cm -2, actual area (4 cm2): 14.0% AM0 spectrum, 140 mW cm -2, active area: 15.0% AM0 spectrum, 135 mW cm -~, active area: 15.5% At sea level the solar spectrum is shifted and on a clear day the solar i n p u t power is a ppr oxi m a t e l y 100 mW cm -2. Use of a p y r h e l i o m e t e r to measure the terrestrial solar input power resulted in a conversion efficiency of nearly 18% based on actual area and 19% based on active area.

References A1 F. S. Johnson, J. Met., 2 (6) (1954) 431. A2 M. P. Thekaekara, NASA-TR-R-351, October 1970.