n-CuInSe2 photoelectrochemical cells

Solar Cel~, 16 (1986) 529 -548

529

n-CuInSe2 PHOTOELECTROCHEMICAL CELLS

DAVID CAHEN The Weizmann Institute o f Science, Rehovot 76100 (Israel)

Y. W. CHEN*, R. NOUFI, R. AHRENKIEL and R. MATSON Solar Energy Research Institute, Golden, CO (U.S.A.)

MICHA TOMKIEWICZand WU-MIAN SHEN Brooklyn College, City University o f New York, Brooklyn, N Y 'U.S.A.)

(Received July 15, 1985)

Summary The results of a wide research program, aimed at improving the conversion efficiency of n-CuInSez/liquid electrolyte solar cells, while maintaining the stability exhibited by this material in aqueous polysulfide, are described. Aqueous polyiodide is chosen as the electrolyte, on the basis of comparative studies with a number of aqueous and non-aqueous electrolytes. By making the polyiodide slightly acidic, both basic and acid electrolyte photo-oxidation can be avoided. Brz/MeOH is shown to be a good etchant for n-CuInSe2 and o p t i m u m conditions for its use are determined. Its chemical and physical effects axe studied. Chemically, a near-stoichiometric top layer is removed, leaving an O-rich surface. Physically, this treatment removes Fermi level pinning, possibly due to Se-related defect states. Subsequent air oxidation further improves cell performance and conditions for it are optimized. Chemically this treatment leads to formation of In-O bonds. Evidence is found for a low conductivity top layer, and for a ten-fold decrease in doping density, as compared to the etched sample, which is sufficient to explain the improved performance. Only the chemical etch has a significant effect on the charge collection efficiency, as measured by EBIC. From electrochemical decomposition, solid state chemical and surface composition studies we formulate stabilization strategies: adding Cu ÷ and In 3÷ to the solution and forming an additional indium oxide film on the electrode surface. In this way, stable, and 12% efficient cells are formed.

1. Introduction During the last few years the conversion efficiencies of photoelectrochemical cells (PEC) (also called electrochemical photovoltaic cells), have *Present address: InterNorth Research Center, Omaha, Nebraska, U.S.A. 0379-6787/86/$3.50

© Elsevier Sequoia]Printed in The Netherlands

530 risen steadily and a num be r of systems with > 1 0 ~ efficiency and reasonable to excellent stability have been reported [1]. Among PECs those based on n-CuInSe 2 (and n-CuInS2) o c c u p y a special position for a num ber of reasons. Their most outstanding feature is u n d o u b t e d l y their o u t p u t stability [ 2 - 5 ] , which, with them, is m uch less of an issue than with most ot her systems. But it was n o t until the ubiquitous aqueous polysulfide system, modified so successfully for high efficiency Cd(Se,Te) cells [6], was replaced by aqueous polyiodide (a redox couple that has been used also for high efficiency layered dichalcogenide PECs [1]) that > 1 0 ~ CuInSe 2 PECs could be constructed [7-12]. A n o t h e r remarkable feature of the n-CuInSe2 cells is that when singlecrystal devices are compared, the performance of the PECs is, at the time of writing, even be t t e r than that of p-CuInSe2-based solid state cells. While this is n o t the only case where a PEC out-performs a solid state device, it is special because considerable efforts are spent on the solid state cells, something that ca nnot be said for Cd(Se,Te) or WSea-based solid state devices. It would be simplistic to state that just changing the redox electrolyte led to the near-doubling of n-CuInSe: PEC performance. In addition, surface optimization was a prerequisite. Here we will present a summary of our work th at led to the n-CuInSez/aqueous polyiodide PEC (up to 12. 5~ efficient) and describe the present status of our understanding of this system. We believe that, notwithstanding the opposite semiconducting behavior of CuInSe 2 in PEC and solid state devices, such a summary will also be of use in attempts to improve our understanding of the solid state device. We will limit the scope of this report to our own work and refer the reader to the original articles by Menezes and coworkers [10 - 12], who have carried o u t in parallel to ours, some elegant work using a c i d i c iodide solution. The differences that result from the slight (as little as 2 - 3 pH units seems sufficient) changes in redox electrolyte show on the one hand the versatility of PECs and on the ot her hand their problems. Suffice it to mention here th at the electrochemical kinetics at the n-CuInSe2 surface layer are better at lower pH values, all ot her factors being equal, while the solution chemical stability is decreased.

2. Experimental details n-CuInSe 2 crystals, with 1016- 1017 cm -3 nominal doping levels, 6 0 0 850 cm z V -1 s-1 Hall mobilities and 1 - 10 ~2 cm resistivities were used. They were obtained from Rincon e t al. [13] and Bachmann e t al. [14]. Most of the crystals were cut along their (112) planes. T hey were polished with A120 3 powder, down to 0.05 pm. Ohmic contacts were made to one side by rubbing I n - G a alloy o n t o the surface. A copper wire was attached to this surface using Ag epoxy. E xcept for the surface to be exposed to the electrolyte, the crystals were covered by Torr-seal epoxy.

531 Aqueous polyiodide solutions were prepared using analytical grade KI and I2. Modified solutions were prepared by adding CuI (or CuI2, if no I2 was added to the original solution; CuI2 reacts to give C u + + ½12) and InI3. The pH of the solution was adjusted by adding aqueous NaOH (pH ~ 6) or by adding HC1 (pH ( 6 ) . Most solutions had a pH between 4 and 6. Air heating was done in conventional laboratory furnaces. Indium oxide films on the electrode surface were formed by plating In from an aqueous solution and passing about 0.2 C cm -2 through the electrode at --0.90 V versus SCE, and air-heating the electrode for 3.5 h at 90 °C. All electrochemical measurements were done under potentiostatic control using large-area carbon or platinum foil counterelectrodes and a saturated calomel electrode as the reference electrode. Tungsten-halogen illumination was used. Crystal stoichiometry was checked by wavelengthdispersive X-ray fluorescence in a CAMECA electron microscope. If we consider CuInSe2 as (Cu2Se)x(In2Se3)l_ x with x = 0.50, the sample compositions were typically x = 0 . 5 0 - 0 . 5 1 with an excess Se of 1- 2 at.%. Most scanning electron microscopy (SEM) and EBIC data were obtained with a JEOL microscope as described elsewhere [15]. Details of surface analyses (X-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES)) can be found in ref. 16. Solution analyses were performed by atomic absorption. Details on capacitance measurements are given in refs. 17 and 18.

3. Results and discussion Figure 1 shows I - V curves for surface-optimized and stabilized nCuInSe2 in modified aqueous polyiodide. A light-to-electricity conversion efficiency of 11.5-12% is obtained at these illumination intensities. We discussed similar results elsewhere in some detail [7, 8] and reported up to 12.5% conversion efficiency in a simple neutral polyiodide solution [9]. We will now present data on the way in which such results were obtained, concentrating first on efficiency optimization by solution modification and surface treatments, and subsequently on cell stabilization. 3.1. Solution optimization

We first discuss the choice of solution and the optimization of the solution of choice. 3.1.1. Selection o f redox solution At the start of this research program only aqueous polysulfide was known to give reasonable, but very stable, conversion efficiencies with nCuInSe 2 [4]. While high short-circuit currents could be obtained, the poor fill factors indicated resistance losses and possible kinetic limitations at forward bias. Indeed, capacitance-voltage measurements showed evidence for band-edge movement, at potentials ~ - 0 . 3 V versus the polysulfide

532 60L--

-- ~.._

I

r

I

50

g

g,=

0

0

I

,

200 Photovoltage

L

I

400

\\

(mV)

Fig. 1. P o t e n t i o s t a t i c p h o t o c u r r e n t - p h o t o v o l t a g e curves for n-CulnSe 2 after 30 s etch in 2% Br2/MeOH and 3 h heating in air at 1 5 0 ° C , o n t o w h i c h In was plated as follows: 0.2 C cm -2 was passed at 0.9 V vs. SCE in an a q u e o u s s o l u t i o n c o n t a i n i n g 6 M KI and 0.01 M In 3+. A f t e r plating, the e l e c t r o d e + In film was t r e a t e d in air at 90 °C for 3.5 h. The s o l u t i o n used for t h e I - V curves c o n t a i n e d 6 M KI, 0.1 M Cu 2+ and 0.1 M In 3+. The e l e c t r o d e area was 5 m m 2. Pt c o u n t e r and r e f e r e n c e e l e c t r o d e s were used. Curves are s h o w n for t w o illumination intensities as n o t e d o n the ordinates. W - h a l o g e n illumination was used. F o r m o r e details see ref. 8.

TABLE 1 Measured r e d o x and o n s e t (for p h o t o c u r r e n t ) p o t e n t i a l s and open-circuit voltages for n-CuInSe2 crystal at ca. 1.5 AM 1 w h i t e light illumination in n o n - a q u e o u s r e d o x e l e c t r o l y t e s Electrolyte

E (redox)

E (onset)

Voc

(v

(V vs. SCE)

(V)

vs.

SCE)

0.2 M C o ( b p y ) (C104) 2 0.1 M T E A P in ACN a

0.,l

0.1 M T E A l , 0.1M 12 in ACN b

0.3

.... 0.1

0.4

0.1 M Fe(CIO4)z 0.1 M T E A P in ACN a

0.4

0

0.4

0.1

0.5

-1.0

0.2

0.1 M TEABr; 0 . 1 M Br 2 in ACN c 0.02 M Na2S; 0.03 M S in M e O H / E t O H ( I : I ) d

-

0.25

0.65

a Cf. refs. 7, 20; TEAP, t e t r a e t h y l a m m o n i u m p e r c h l o r a t e ; ACN, acetonitrile.

b TEAI, t e t r a e t h y l a m m o n i u m iodide. c TEABr, t e t r a e t h y l a m m o n i u m b r o m i d e . d MeOH, m e t h a n o l ; EtOH, ethanol.

redox potential, o f + 0 . 3 5 to 0.4 V o n the electrochemical scale [ 1 9 ] . While n o t investigated further, it is likely that such shifts are due to changes in the nature of adsorbed species on the surface of the n-CuInSe/ electrode. Thus we l o o k e d for other redox electrolytes, b o t h a q u e o u s and n o n - a q u e o u s . A series o f systems was c h o s e n on the basis of their k n o w n redox potentials, the flat-band potential o f n-CuInSe2 in polysulfide and the electronegativity

533 TABLE 2 Measured redox and onset (for photocurrent) potentials and open-circuit voltages for n-CuInSe2 crystal at ~AM 1.5 white light illumination in aqueous redox electrolytes Electrolyte

E

(redox) (V vs. SCE)

E (onset) (V vs. SCE)

Voe (V)

0.5 M I2, 6 M KI, pH 6 1 M S2-, 1M S, 1 M OH0.1 M Fe(CN)6a-; 1 M NaOH 0.1 M Fe2+; 1 N HC104 0.4 M V3+; 4 N HC1 1 M $203-; 1 M SO3-; 1 M OH-

0.3 --0.8 0.1 0.4 0.5 --0.6

--0.15 --1.15 --0.3 0 0 --i.I

0.45 0.35 0.4 0.4 0.5 0.5

of n-CuInSe2 (~4.5 eV). Table 1 summarizes results in non-aqueous systems and Table 2 gives results for the aqueous systems that were checked. The onset potential for photocurrent provides a rough estimate for the fiat-band potential, generally a lower limit (i.e. a minimum value on the solid state scale; a m a x i m u m on the electrochemical scale for n-type semiconductors). If we take the standard calomel electrode (SCE) potential to be about --4.75 eV on the solid state scale, assuming a value of --4.5 eV for the normal hydrogen electrode on that scale, then we see from Tables 1 and 2 that only the sulfide (polysulfide) and the (irreversible) sulfite/thiosulfate systems yield onset potentials that undoubtedly deviate significantly from the values that could be estimated. This may be caused by strong adsorption of solution species. The onset potentials in aqueous polysulfide and polyiodide may be compared with fiat-band potentials obtained from MottSchottky plots and from electrolyte electrorefiectance data [19, 21], v i z . --0.5 V v e r s u s SCE for polyiodide and --1.4 V v e r s u s SCE for polysulfide (decreasing to --1.2 V v e r s u s SCE after prolonged use). Part of these discrepancies can be explained by differences in solution composition, but the main cause is the above-mentioned caveat with respect to using onset potentials to estimate fiat-band potentials. Still the values from the tables would make us choose those solutions with which the largest t h e r m o d y n a m i c driving force for charge transfer can be obtained, with [E(onset) -- E(redox) [ a maximum. However, such arguments are inadequate here, as could be seen from the complete photo I - V curves, rather than from only Vo¢. Those showed in aqueous solutions the following trend (decreasing performance): polyiodide ~ polysulfide ~ Fe(CN6) 3j4- ~ Fe 3/2÷, V 4 / 3 + ) SO32-/$2032-, i.e. not at all consistent with expectations based on [E(onset) -E(redox) I or Voc only. In non-aqueous solutions the trend was more like the one to be expected naively: Co(bpy)3 a/2÷ ~ polyiodide ~ Fe 2j3÷ ~ Br-/Br: polysulfide, but better performance was obtained always in aqueous solutions. It is possible that careful optimization of non-aqueous solutions, along the lines of work done on Si- and GaAs-based systems [22], might change this, but we have not pursued this further.

534

On the basis of these results aqueous polyiodide appears to be the solution best suited to CuInSe2. While there exists a slight thermodynamic advantage with this system over polysulfide [19], kinetic factors appear to be dominant [23]. Addition of Cu + and/or H + (cf. the section on stabilization, below) further improves the kinetics of charge transfer in polyiodide, while use of Cs (rather than Na or K) polysulfide improves the performance in aqueous polysulfide (cf. ref. 6). The pH of the polyiodide solution has a dramatic effect on PEC performance as illustrated in Fig. 2. The deterioration of the I V characteristics at higher pH is probably due not only to changing adsorption characteristics on the CuInSe2 surface, but also to IO 3- formation at high pH, which leads to poorer kinetics of the, generally very fast, polyiodide system at inert electrodes [24]. The dark I Vcurves show that the better kinetics of cathodic charge transfer go hand in hand with better anodic ones. Menezes et al. [10] showed that decreasing the pH to strongly acidic values leads to further improvements in PEC performance. While we have been able to confirm this phenomenon, its use leads to a decrease in solution stability. Neutral poly-

1

I

I a

b e

E

d

0 0 .c ¢L

0.4

Photovoltage

e'

(V)

b'

-2

a'

-4

I

I

I

I

Fig. 2. P h o t o c u r r e n t - p h o t o v o l t a g e and dark c u r r e n t - v o l t a g e plots for n-CuInSe2 crystal ( 0 . 1 4 4 c m z area), e t c h e d a n d air-heated as in Fig. 1, w i t h o u t e x t r a I n ( o x ) layer, in a q u e o u s 6 M KI, 0 . 5 M I2, as a f u n c t i o n of pH. a, at: p h = 7.1; b, b ' : pH = 9 . 1 ; c , c': p H = 10.3; d, dr: pH = 11:3; e, e': p H = 11.8; f, f': a f t e r e, w i t h s o l u t i o n s t a n d i n g in air for 7 min. L i g h t intensity, ~ 1 . 3 x AM 1.

535 iodide solutions were air-stable for over 10 months, with or w i t h o u t added Cu ÷, if kept in the dark. Neutral, Cu÷-containing solutions became slightly air-sensitive when exposed to light. For example, after 7 days' illumination at about 400 mW cm -2 white light, a solution of 6 M KI contained some 1 mM I2. Cu + appears to catalyze the oxidation of iodide, possibly according to : 2 Cu+ + O2 + 2 H20 2Cu 2 + + 2 1 - -

) 2 Cu2+ + 2 H202 -{- 2e-

)2Cu ++12

At pH 1, fast oxidationof iodide occurs also in the dark in air. Consistent with the electrochemical results of Dagan and Cahen [23], H+ appears to be more effective as a catalyst for iodide oxidation than Cu+, in concentrated acid solution (cf. refs. 25 - 27). Thus, especially acidic polyiodide solutions need to be used after they have been purged of oxygen, if long-term stability is desired.

3.2. Electrode optimization From the earlier, polysulfide PEC work [4] it was clear t h a t surface treatments are of paramount importance for CuInSe2 photoelectrode performance. Obviously the results will be most meaningful if crystals of top bulk optoelectronic quality can be used. Unfortunately, no such samples are yet readily available and all our experiments were carried out on crystals that, while having reasonable mobilities, are probably considerably compensated. 3.2.1. Chemical etchants While strong acid etchants were used previously [3, 4], their use does n o t easily lead to reproducible electrode performances. Therefore other etchants were sought. Figure 3a shows p h o t o / - V curves after a number of etching treatments. The KCN dip is used to dissolve off any free Se that might form on the surface. The Br2/MeOH etch was chosen because it led to stable surfaces, as deduced from C - V measurements. Its use was optimized empirically, with the results shown in Fig. 3b. We used both solution (atomic absorption analyses of the etching solutions) and surface analyses (XPS, AES) to study the chemical effects of the Br2/MeOH etch. Table 3 summarizes some solution analyses. The results confirm that KCN dipping dissolves Se, which probably precipitated on the surface after the acid etch. Both the acid and the Br2/MeOH etch lead to removal of roughly 1 : 1 : 2 stoichiometric ratios of Cu : In : Se. The reproducibility with the latter etch is somewhat better than with the former. Elsewhere we have reported on detailed XPS/AES analyses of the surfaces of Br2/MeOH etched crystals [16]. They show the formation of a clean surface insofar that only single types of Cu and In are seen, together with some Cu-depletion of the surface. Some Se ° appears to be left on the surface and the presence of low- or zero-valent oxygen could be detected.

536

<

sI

E. 1 0 ~ ~

E~o ~105

3 \

4

5

4~

5

\

n

0

0

(a)

0.2 Photovoltage (V)

i

0.4

00

0.2 Photovoltage (V)

(b)

0.4

Fig. 3. Photo I - V curves for n-CuInSe2 crystals in 6M KI, 0.5M I2, pH = 7 as a function of etchants after polishing. Polished electrode area: 0.36 cm2; ~1.5 × AM 1 illumination. I-/I3- potential was c a . + 0.3 V v s . SCE. (a) Etched in 2% Br2/MeOH for 0, 5, 10, 15, 20, 30, 60s (curves 1, 2, 3, 4, 5, 6, 7 respectively). (b) Curve 1: etched in 55% H N O a ( 3 0 s ) and dipped in 10% (w/v) KCN for 60s. Curve 2: only dipped in KCN. Curve 3: as curve 6 in Fig. 3a, followed by dipping in KCN. Curve 4: first dipped in KCN, then as curve 6 in Fig. 3a.

TABLE 3 Atomic absorption analyses for Cu, In and Se in various etching solutions Sample

Etchant(s)

Time

1 I b and 2 2 b and 3 3 b and 4 5 6 7c

55% HNO3 5% KCN 55% HNO3 5% KCN 35% HNO3 5% KCN 2% Br2 in MeOH 2% Br2 in MeOH 2% Br2 in MeOH 2% Br2 in MeOH

90 +120 120 +120 60 +120 120 30 30 30

(s)

Cu a

In a

Se a

1 1 1 1 1 1 1 1 1 1

1.2 1.2 1.1 1.1 1.2 1.2 1.4 1.0 1.0 0.8

2.2 2.3 2.0 2.2 2.8 3.4 2.1 2.3 2.4 2.6

a Stoichiometric ratios, normalized to one equivalent Cu. Uncertainty ~ 10%. b After the HNO3 etch the sample was rinsed (rinse solutions were added to the HNO3 etch) and then dipped in KCN solution. The combined results from both treatments are given. c Because of the very small sample, a single crystal was used here rather than the multicrystalline samples used otherwise; the error is at least twice as large, i.e. ~ 20%.

T h e r e s u l t s o f t h e B r 2 / M e O H e t c h i n g t r e a t m e n t s h o w n in Fig. 3 w e r e investigated further by several physical techniques. Electron-beam-induced current experiments on Au/n-CuInSe2 devices using samples, part of whose s u r f a c e h a d b e e n e t c h e d , s h o w e d t h e e x p e c t e d l a r g e i n c r e a s e in c h a r g e c o l l e c t i o n efficiency. M a x i m u m g e n e r a t i o n factors o f 180 c o u l d be o b t a i n e d at 35 kV and 1 n A b e a m current.

537 Electrolyte electroreflectance data, discussed in detail elsewhere [21] showed removal of Fermi level pinning. This effect is less pronounced in the polyiodide solution than in polysulfide, but, in both cases appears to be due to slowing down of surface states located about 0.11 eV above the valence band. The density of these states before etching corresponds to monolayer coverage. It is of interest to note that similar states were detected on CdSe and on CdIn2Se4. On the basis of recent photoluminescence data [28, 29] their energy corresponds to an Se-related defect, Vse if a covalent point defect model is assumed, or Se i if an ionic one is operative. Because the latter possibility appears to contradict the dependence of conductivity type on stoichiometry [30], while the former one involves an ad h o c assumption, a third possibility can be suggested based on recent SIMS results [31], namely the occurrence of the isovalent acceptor Ose. Impedance measurements [32] show the formation of an abrupt junction after etching. Less than 1% of a monolayer of two types of states is left on the etched surface, at energies of about 0.17 eV below the conduction band and at a b o u t 0.45 eV below it, i.e. near mid gap. The remaining surface combination occurs only through the first state. By using a simple G ~ t n e r like [17, 33] model to fit the photo I - V curve, a lower limit of 0.04 pm could be derived for the effective minority carrier diffusion length. Although not further invesigated in this study, we note that controlled photoelectrochemical etching can improve the performance of etched electrodes considerably. However, its efficacy is less pronounced if the chemical etching treatment has been performed optimally [5, 23]. 3.2.2. A i r o x i d a t i o n

Figure 4 shows the effect of controlled air oxidation of a Br2/MeOH etched electrode on its I - V characteristics in polyiodide. From the figure we see that optimal conditions exist for this treatment, which has previously been applied successfully to polysulfide-based cells [4, 34]. The chemical effects of this treatment were investigated by bulk and surface analyses. No significant changes in bulk composition were found, indicating that if oxygen is introduced its concentration is well below the detection limits of the microprobe analyses. Combined XPS/AES analyses described in detail elsewhere [16] show the presence of two types of indium. In view of oxygen XPS and AES data it is likely that In-O bonds are formed, i.e. an indium oxide-covered surface is obtained. This result is consistent with those obtained, under slightly different conditions, on thermally and on anodically oxidized CuInSe 2 samples [35, 36]. No more evidence for zero-valent Se is seen and none for SeO2. The Se that is present occurs in a Cu2Se-like environment. The exact stoichiometry of the oxide formed is not known b u t its binding energies resemble closely those of the native oxide on CuInSe2. Thus it appears that thermal oxidation leads mainly to a thicker and more homogeneous oxide film. Figure 5 shows the effect of air oxidation on the Cu-Se ionicity, by comparing the binding energy differences between the Cu 2p3j2 and Se 3d

538

6

5 2

c

P 10 I== 0 0¢-

o.

5

0

0.2 Photovoltage (V)

0

0.4

Fig. 4. P h o t o I - V curves as a f u n c t i o n of air a n n e a l d u r a t i o n , (1, 0.5, 1, 2, 3, a n d 1 6 h ; curves 1, 2, 3, 4, 5, 6 respectively) a f t e r 30 s Br2/MeOH e t c h . F u r t h e r details are as in Fig. 3.

Se Formal V a l e n c e 2 1 I

I

.. T CulnSe2 ' ,-Af"

I "-....CulnSe2 {ox)

772.0

A

t

~" @"I.

Cu3Se2

0

CuCr2Se,

" - "-...

CuSe2-p

CuSe 771.0 ~-

uSe= m u0 Cu-Se

Ionicity (a.u.)

Fig. 5. Plot o f d i f f e r e n c e in Cu 2p 3 / 2 and Se 3d b i n d i n g energies, as m e a s u r e d b y X-ray p h o t o e l e c t r o n spectroscopy, for cleaved, a n d for s u r f a c e - o p t i m i z e d (3 h curve in Fig. 4) CuInSe2 crystal surfaces, vs. Cu Se ionicity. T h e slope of t h e plot is o b t a i n e d f r o m the n o n - C u I n S e 2 e n t r i e s as justified in ref. 37. Only t h e Cu°/Se ° and CuInSez p o i n t s were m e a s u r e d for t h e first t i m e in this study, and a d d e d to t h e plot. Because a f o r m a l Se valency ~ 2 has n o clear physical m e a n i n g , t h e b o t t o m abscissa was a d d e d to t h e original plot. E x p e r i m e n t a l details are given in ref. 16.

539 peaks for a series of Cu-Se compounds, in a manner described by Folmer [37]. We note the high ionicity in cleaved CuInSe2 and the decreased value after air oxidation. This can be understood by considering the high formal charge on In in CuInSe2, which would tend to polarize the Se. After air oxidation, much of the In near the surface is b o u n d to oxygen and thus exerts less of a polarizing force on the Se. Finally we note that a short sputter etch by Ar ions restores the original (as-cleaved) surface. To understand the strong effect of air oxidation on the I - V characteristics (cf. Fig. 4), we carried o u t some further experiments. Except for the fact that much higher modulation voltages were needed to obtain reflectance signals, no meaningful information could be obtained from electrolyte electroreflectance. It is possible that a high resistivity top layer dictates that higher modulation voltages be used [21]. Rather surprisingly, no great increase in generation factor was seen in the EBIC experiments on Au/n-CuInSe2 (air oxidized). A peak value of 260 was obtained, indicating a modestly improved charge collection efficiency compared to a Br 2/MeOH etched surface. Impedance measurements show that the fiat-band potential is unaffected by air oxidation, b u t the effective doping level in the t o p layer is reduced tenfold, as compared to the etched sample. By fitting the photo I - V curves before and after air oxidation, we find that virtually all of the improvement in I - V characteristics can be explained by this reduction in doping. This can be understood by realizing that the space charge layer width will increase by between threefold and fivefold, from an initial value of circa 500 A. This means that now rather less than one-third of the incident white light will be absorbed in the space charge layer. This significantly reduces the effect of bulk recombination, outside the space charge layer [32]. Further studies using relaxation spectrum analysis [17] and EBIC suggest that the oxidized layer is porous and of low p-type conductivity, when measured in a solid state set-up. Because of its porosity in a liquid electrolyte environment, a situation is obtained that is best described by differential effective medium theory of the insulating oxide and the conductive electrolyte.

3.3. PEC stabilization While the I - V results obtained after solution and surface optimization are impressive they do not persist, i.e. the PEC is unstable. Our strategy to stabilize the systems was based on attempting first of all to understand the decomposition paths that may be operative. Of particular importance here was the observation from surface analyses [16, 38] that the surface of n-CuInSe 2 is readily depleted of Cu. From an extensive study of the solid state chemistry of the (Cu2Se)~(In2Se3)l_ ~ series described in detail elsewhere [39], we conclude that the desired chaicopyrite phase occurs for x-values approximately between 0.55 and 0.44; b e y o n d these limits mixed phases (x ~ 0.55) or cubic phases (0.43 ~ x ~ 0.23) are found.

540 C o n t r o l l e d e l e c t r o c h e m i c a l d e c o m p o s i t i o n e x p e r i m e n t s w e r e carried o u t o n b o t h p- a n d n - t y p e CuInSe2 in acetonitrile. While n o c a t h o d i c d e c o m p o s i t i o n c o u l d be d e t e c t e d o v e r t h e 2 V w o r k i n g range, several a n o d i c waves w e r e o b s e r v e d and c o u l d be identified. In this w a y t h e f o l l o w i n g d e c o m position reactions were formulated. In the dark: 2 (n or p) CuInSe2 ~ 2 Cu 2+ + In2Se 3 + Se ° + 4 e In the light: 2 n - C u I n S e 2 ~ 2 Cu + + In2Se 3 + Se ° + 2 e Surface analyses s h o w e d strong Cu d e p l e t i o n and s o m e In d e p l e t i o n , as well as t h e p r e s e n c e o f l o w e r - v a l e n t Se, p r o b a b l y Se °. F o r d a r k - d e c o m p o s e d s a m p l e s evidence o f Cu 2÷ was seen. A t o m i c a b s o r p t i o n analyses o f the solutions, a f t e r t h e e l e c t r o c h e m i c a l d e c o m p o s i t i o n , s h o w e d large a m o u n t s o f Cu. Because o f t h e high d e t e c t i o n limit f o r In, no m e a n i n g f u l results f o r its p r e s e n c e c o u l d be o b t a i n e d in this way. A f t e r rinsing the sample w i t h KCN, s o m e Se was f o u n d in the c y a n i d e s o l u t i o n ( a b o u t 25 - 30% o f t h e Cu equivalent f o u n d in t h e a c e t o n i t r i l e solution). S u b s e q u e n t l y we a n a l y z e d e l e c t r o d e p h o t o d e c o m p o s i t i o n in a q u e o u s p o l y i o d i d e . F r o m analyses in the scanning e l e c t r o n m i c r o s c o p e , using s e c o n d a r y e l e c t r o n , b a c k - s c a t t e r e d e l e c t r o n , a b s o r b e d c u r r e n t and WDS scans, the p r e s e n c e o f Se n o d u l e s o n t h e surface was derived. D e c o m p o s i t i o n a p p e a r e d to be m o s t p r o n o u n c e d , visually, along scratches left f r o m t h e polishing. X P S / A E S analyses c o n f i r m e d the p r e s e n c e o f Se, and s h o w e d the surface to have b e c o m e virtually Cu-free. B o t h In and iodide w e r e d e t e c t e d , suggesting t h e o c c u r r e n c e o f InI3, s o m e t h i n g t h a t w o u l d fit t h e r m o d y n a m i c e s t i m a t e s [ 1 9 ] . F r o m these e x p e r i m e n t s it a p p e a r e d t h a t Cu is the m o s t labile c o n s t i t u e n t in CuInSe2, a l t h o u g h cyclic v o l t a m m o g r a m s s h o w e d Cu o x i d a t i o n to o c c u r o n l y w i t h Se o x i d a t i o n . T h e a b o v e f o r m u l a t i o n f o r o x i d a t i v e d e c o m p o s i t i o n agrees w i t h t h e r m o d y n a m i c e s t i m a t e s , e x c e p t t h a t those favor Cu2Se r a t h e r t h a n Cu ÷ f o r m a -

64 wE

_Xx k

\~,.,i-- Optimized Etch E v

\,,.

.~ 3 2 u

I n Only

".... ~ N o n - Optlmal Etch

....

Ol6 0

(3_

c

I IO

I 20 Time (minutes)

Fig. 6. Photocurrent (at short-circuit) us. time plots for n-CuInSe2 in 6 M KI, 0.5 M I2, pH 6. The dotted curve shows results for an acid-etched electrode, the dashed ones for electrodes treated as shown in Fig. 4, curve 5. Note the effect of light intensity (initial current density). Light intensities between 1 and 2 × AM 1 were used.

541

tion upon decomposition in polyiodide [19]. These estimates also confirm the validity of two of the steps we t o o k to stabilize the n-CuInSe2/polyiodide cell, viz. addition of Cu ÷ ions to the iodide solution and further surface coverage with indium oxide. Figure 6 shows the short-term stability of shortcircuit photocurrent o u t p u t for etched and air-annealed CuInSe2 at two initial current densities (broken lines). While at lower current density (about 120 mW cm -2 white light) the electrode appears stable, the high current density data show this not to be the case. On the basis of the results presented above, we modified the polyiodide solution by adding Cu ÷ (as Cu 2÷ to an I2-free solution), by adding In 3+ and b y adding both. Much improved stability was obtained, as can be seen from the results given in Table 4. As shown in Fig. 7, under strongly accelerated conditions, just modifying the solution was not sufficient (curve d). Surface analyses showed the presence of two types of In, of oxygen and of iodide. As we suspected that the indium oxide layer formed by air oxidation might n o t have covered the surface completely, we electrodeposited a thin film of In (0.3 pm by geometric area) on top of the etched and air-oxidized crystal, and air-oxidized this In layer. While SEM pictures showed the coverage to be non-uniform, this treatment proved to be effective for complete stabilization, as shown in curve e of Fig. 7. Microprobe analyses of electrodes treated in this way, before and after their use in polyiodide, showed most of the excess In to have disappeared leaving a normal stoichiometric crystal. It is likely that the excess indium oxide was converted to indium iodide and dissolved in the solution, leaving oxide only on areas not previously covered completely by an oxide film. Surface analyses of electrodes covered with extra indium oxide showed no significant changes after use in polyiodide, either for the In peaks or for the oxygen (whose Auger peak partly overlaps with the iodine peak). TABLE 4 O u t p u t stability o f n - C u I n S e 2 / a q u e o u s p o l y i o d i d e PECs as a f u n c t i o n o f s o l u t i o n and surface m o d i f i c a t i o n s

Charge passed

Isc ( m A c m -2)

(C cm-2) 820 14 25 27 1 080 11 000 15 000 b

Initial

Final

49.0 56.0 68.5 53.0 48.3 55.0 52.0

31.0 52.0 62.5 52.0 48.3 55.0 52.0

a As curve 5 in Fig. 4. b At m a x i m u m power.

Stabilization strategy

HNO3 etch ÷ air o x i d a t i o n Br2/MeOH e t c h q- air o x i d a t i o n a A d d 0.1 M In 3+ A d d 0.1 M Cu 2+ A d d 0 . 1 M In~ + + 0 . 1 M Cu 2+ As above, t o I2-free s o l u t i o n As a b o v e + air o x i d i z e d In film, d e p o s i t e d o n CuInSe 2

542 ]

I

i

E 120

E

_

i

e [b+c+ln(ox)] d [b+c] c [+Cull

~

e"

a b [ ~InCI~]

0 0 en

40--

I

I

400

I 1200

Time (sec) Fig. 7. As Fig. 6, at 3- 4 x AM 1 illumination. Curve (a) refers to a surface-optimized electrode (Fig. 4, curve 5). Curves b, c, and d refer to the same surface (after repolishing, re-etching and re-heating) in modified solutions. Curve e refers to an electrode, as in a, onto which In was deposited and subsequently air-oxidized.

T h e p h o t o I - V characteristics o f i n d i u m o x i d e covered e l e c t r o d e s in the m o d i f i e d e l e c t r o l y t e (Fig. 1) were essentially u n c h a n g e d f r o m those o b t a i n e d f o r e l e c t r o d e s w i t h o u t the e x t r a film in simple p o l y i o d i d e solution. A d d i t i o n a l e x p e r i m e n t s were p e r f o r m e d to clarify the effect(s) o f the e x t r a layer o f o x i d i z e d indium. E l e c t r o l y t e e l e c t r o r e f l e c t a n c e results are essentially t h e same as those o b t a i n e d w i t h o u t e x t r a film, o n an e t c h e d and air-oxidized sample [ 2 1 ] , s u p p o r t i n g the c o n c l u s i o n t h a t air o x i d a t i o n o f CuInSez leads to f o r m a t i o n o f indium o x i d e o n the surface. Because o f the relatively high bandgap o f In203 ( ~ 2 eV) the presence o f such a layer is n o t e x p e c t e d to a f f e c t the e l e c t r o d e p e r f o r m a n c e significantly, as it will be t r a n s p a r e n t to m o s t o f the i n c i d e n t radiation. The resistivity o f the indium o x i d e film is p r o b a b l y high, also because some charging o c c u r r e d during SEM experiments. A central q u e s t i o n is the location o f the p h o t o v o l t a l c a l l y active junct i o n in the C u I n S e 2 - I n ( o x ) / p o l y i o d i d e system. Figure 8 shows the j u n c t i o n EBIC response, i.e. scans across a cleaved j u n c t i o n o f a solid state device A u / C u I n S e z / I n ( o x ) / T i . No j u n c t i o n was f o u n d at the large-area Au back c o n t a c t . T h e EBIC response suggests t h a t m o s t o f the barrier occurs at the inner interface, w i t h a smaller o n e at the o u t e r interface. T o verify the c o m p o s i t i o n o f the interface a n u m b e r o f Auger e x p e r i m e n t s were perf o r m e d . Line scans across the cleaved j u n c t i o n s h o w e d considerable Ti in-diffusion, into the I n ( o x ) layer, o x i d a t i o n o f m u c h o f the Ti, but, at a r e s o l u t i o n o f several microns, n o evidence f o r the indium o x i d e film. S p u t t e r profiles o n the c o n t a c t area also s h o w e d some Ti in-diffusion and simult a n e o u s presence o f Ti and o x y g e n . Also evidence for a layer relatively rich in In and o x y g e n was seen, b e t w e e n the CuInSe2 bulk and the T i ( + O) covered t o p layer. It is t h u s t e m p t i n g to ascribe the inner interface in Fig. 8 to the

543

t

r,~ T

(a)

(b)

Fig. 8. Junction (cross-sectional) electron beam-induced current (charge collection) picture (left) and traces (one in the left, Y-modulated in the right picture) for Au/nCuInSe2/In(ox)/Ti-Au device. Note 2.5× higher magnification for left-hand picture. From this type of experiment a ~1.0 pm diffusion length (field-assisted) in the CuInSe2 could be derived.

CuInSe2/In(ox) and the outer interface to that between Ti and In(ox). As seen from Fig. 8b, the EBIC " k n e e " at the outer interface is not always very clear. This figure also shows the response to be quite uniform along the junction at this magnification. However, at lower magnification this uniformity was not preserved and regions w i t h o u t any EBIC response could be seen. This observation agrees with the previously mentioned conclusions from frequency