One step electrodeposition of CuInSe2: Improved structural, electronic, and photovoltaic properties by annealing under high selenium pressure

One step electrodeposition of CuInSe 2: Improved structural, electronic,and photovoltaic properties by annealing under high selenium pressure

Jean-Francois Guillemoles,a) Pierre Cowache, Alain Lusson,b) Kamel Fezzaa,Frederic Boisivon, Jacques Vedel, and Daniel LincotLaboratoire d’Electrochimie et de Chimie Analytique, CNRS, URA 216, ENSCP,11 rue Pierre et Marie Curie, 75231 Paris Cedex 05, France

~Received 2 August 1995; accepted for publication 24 January 1996!

Films of Cu–In–Se alloys can be electrodeposited in a wide range of controlled composition.Annealing treatments under Se pressure transform these precursor films in large grain CuInSe2 filmswith improved electronic properties. These modifications are shown to depend on the Se pressureimposed during the treatment allowing a certain tailoring of the electronic properties of the films.The properties of electrodeposited/selenized films are presented as obtained from luminescencemeasurements, Hall effect, and photoelectrochemical characterization. An efficiency of 6.5%~totalarea, without antireflecting coating! is reported for the best CuInSe2/CdS/ZnO solar cell. An analysisof the device is also presented where limitations by interface recombination are shown to be thedominant loss mechanism. ©1996 American Institute of Physics.@S0021-8979~96!07009-9#

t

l

r

nn

e

e

w

e

na

epo-tinge

en-f aa

sby

be-in

anro-ell-

m-era-on-the.hisenti-se-

ntal

in.

rgy.on-

I. INTRODUCTION

CuInSe2/ZnO cells with efficiencies close to 17% werobtained with CuInSe2 ~CIS! films prepared by coevaporation of the elements.1 The development of terrestrial applications is dependent not only on efficiency, but also ondevelopment of suitable elaboration techniques. Electrodesition is an excellent technique for large area processing,cost, and high material yield. Electrodeposition is alreaused on a development stage in the preparation of polyctalline CdTe layer for efficient CdS/CdTe cells.2,3 With con-cern to CIS, there is a wide gap with the efficiencies obtainwith solar cells based on electrodeposited films in spitenumerous attempts.4–7 Despite a value of 5.2% reported i19898 for a device based on electrodeposited CIS, it is orecently that improved values~around 7%! have been re-ported by our group and the former [email protected]%,9 6.5% ~thiswork!, and by the former group10#. These recent progresseopen new perspectives.

To our opinion the postannealing treatment neededimprove the properties of the as grown films is a key stnot yet fully controlled. From our experience as well as froa literature survey, annealing treatments performedvacuum or in inert gas atmospheres lead generally, in spitthe improvement of crystalline and optical properties,11,12 toelectronic properties unsatisfactory for devices. We hashown that a dramatic improvement of these propertiespossible by combining electrodeposition and annealing trements under a sufficient elemental selenium pressure, ling to an efficiency of 5.6%.9,13,14In this article we detail theproperties of electrodeposited/selenized films, as obtaifrom photoluminescence measurements, Hall effect,photoelectrochemical characterization. An efficiency of 6.5~total area, without antireflection coating! is reported for thebest device.

a!For correspondence, author can also be reached via Electronic [email protected]

b!Also with Laboratoire de Physique des Solides, CNRS, 1 place A. Bria92195 Meudon, France.

J. Appl. Phys. 79 (9), 1 May 1996 0021-8979/96/79(9)/7293

Downloaded¬25¬Sep¬2002¬to¬134.157.181.198.¬Redistribution¬subj

e--hepo-owdyys-

edof

ly

s

top,minof

veasat-ad-

ednd%

II. EXPERIMENT

A. Sample preparation

The CIS samples were prepared by one step electrodsition on soda lime glass substrates covered by a conducMo layer ~'0.5 mm, deposited by gun evaporation at thInstitut fur Physikalische Elektronik, Stuttgart University!.

The electrodeposition of the CIS films was made pottiostatically on the molybdenum substrates by means otrielectrode setup~PAR 273!. The reference electrode wasmercury sulphate electrode~MSE,E050.65 V/NHE, the nor-mal hydrogen electrode!. Before deposition the solution wadeaerated by argon bubbling. The films were preparedreducing copper~II !, selenium~IV !, and indium~III ! speciesin an acidic sulphate solution at a fixed potential value,tween20.7 V and21.1 V/MSE, as reported elsewheredetail.15

After deposition the films were thermally annealed inelemental selenium atmosphere. The originality of our pcess is to use a closed reactor allowing us to impose wcontrolled Se pressures in a wide range from 1025 to 1021

atm by means of a fixed temperature in the selenium copartment. The substrate and the selenium source temptures are fixed by using a two-zone furnace. This applicatioriented approach has the great advantage to processfilms in a uniform way, near thermodynamic equilibriumThere are also no losses of selenium from the reactor. Tprocedure is also suitable for a multisubstrate treatm~batch! since it is a nondirectional process. Typical condtions were a substrate temperature of 400–450 °C and alenium source temperature of 300–370 °C~lower than thesubstrate temperature to avoid the condensation of elemeselenium on the films!. The background vacuum was 1023

Torr. The duration of the treatment was typically 20–30 m

B. Thin-film characterization

The composition of the films was measured by enedispersive spectroscopy~EDS! and by atomic absorptionThe composition profiles were obtained by secondary-i

ail:

nd,

7293/10/$10.00 © 1996 American Institute of Physics

ect¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp

e

b

n

-p

u

r

a

m

t

be

o

-y

i

u

cein

ent

on-there

eat

ur-

g-inlly

-f

-n

hehe

or

mass spectroscopy~SIMS! ~with Cs atoms! at the Labora-toire de Physique des Solides of Meudon–Bellevue. Sevmass channels were simultaneously recorded for eachment~Cu, In, Se, Mo, O, Cu1Cs, In1Cs,...!. For clarity andto avoid interferences, only the channels with the single ements were presented. The crystalline structure and orietion were determined by x-ray diffraction on au22u setupwith a Co anticathode~l51.789 Å!.

Transport properties in the films were determinedHall effect and resistivity measurements in the van der Paconfiguration. As the films were necessarily deposited oconducting substrate, the conducting layer had to be remoprior to measurement. Then, the measurement is madelayers grown on the relevant conditions~similar to thoseused for device elaboration!, in opposition to characterization of layers grown on an insulating substrate. The extralation of the results may be problematic because, as is wknown, properties of thin films are dependent on the natof the substrate. We developed two methods,16 one chemicaland the other mechanical, to investigate the transport perties of our films. In both cases the CIS is encapsulatedan epoxy resin~Epofix, Struers!, leaving the glass side bare

In the first method the glass substrate is dissolved usconcentrated hydrofluoric acid. We operated in such contions that the underlying Mo would be passivated, i.e., inslightly acidic solution atpH'3.17 This was necessary inorder to prevent formation of pinholes in the Mo film thwould allow some local CIS etching. The Mo substrate wthen dissolved using a solution of K2HPO4 ~0.5 M! and ofH2O2 ~0.1 M!. In these conditions the solution reacts veslowly with the CIS and the Mo layer can be removed copletely with minimal damage on the CIS layer. In the secomethod, the CIS was stripped off by applying a tangenforce as in Ref. 18. Both methods showed similar results

Photoluminescence spectra were recorded at 1.6 K wan excitation power of 10 mW~unless otherwise specified!focused on a 0.2-mm-diam spot~from an argon laser line 488nm, analyzer: BOMEM DA8 Fourier transform interferometer!. Specification of the power density is essential herecause of the known dependence of the luminescence pposition with this parameter in this material.19

To assess the photovoltaic properties of the films befjunction completion, a photoelectrochemical technique wused. This technique has been previously used successby us to characterize CuInSe2 and CuGaSe2 coevaporatedfilms and electrodeposited films.7,20 The semiconductor electrolyte junction is made by dipping the films in a slightlacidic solution@H2SO4/K2SO4 ~0.1 M! at pH52#. The ex-perimental setup consisted of a potensiostat PAR 273 anconventional electrochemical glass cell. A low-intensmonochromatic illumination was provided by a monochrmator, Jobin & Yvon H20. The flux was measured with cabrated Si and Ge detectors. The photocurrent was detewith a lock-in technique. Spectral responses were measat reverse bias~21 V versus MSE reference electrode!.

C. Devices elaboration and characterization

Devices with the Mo/CIS/CdS/ZnO structure have beelaborated with these films.9,13,26,39 After selenization, the

7294 J. Appl. Phys., Vol. 79, No. 9, 1 May 1996


ralele-

le-nta-

yuwa

vedon

o-ellre

op-in.ingdi-a

tas

ry-

ndial.ith

-e-aks

reasfully

d atyo-li-ctedred

en

CIS films were first covered by a thin CdS film~about 20 nmthick! deposited from solution as described previously.21 AZnO window layer was then deposited on top of the deviby rf sputtering in two steps, according to Ref. 22: first a th~200 nm! undoped layer and second a thicker~1.4–2.5mm!conducting ZnO:Al layer. These ZnO bilayers are transpar~80%–95% of transmittance in the visible range! and con-ducting~r51.431023 V cm!. Hall-effect measurements givea doping level of'531017 cm23 for the intrinsic layer.

The cell area~typically 0.1 cm2! is defined by mechani-cal scribing. For the optoelectrical measurements the ctacts are taken on the molybdenum back contact and onZnO by using In–Ga-coated tungsten wires. The cells wecharacterized by their [email protected] ~Ref. 23!# and darkI –Vcurves and by their spectral responses. Pre- or post-htreatment in air~200 °C, during a few minutes! was used toimprove the characteristics. The dependence of the photocrent with applied bias~such as shown on Fig. 10! was ob-tained with a slightly modified setup, the light source beinmodulated and the photocurrent being detected by a lockanalyzer. The intensity of the light source was eventuavaried with neutral density filters.

III. RESULTS

A. Electrochemically deposited precursor films withcontrolled composition

As demonstrated by electrochemical studies24–26 theoverall composition of the film is controlled by the ratiobetween Cu~II ! and Se~IV ! fluxes at the electrode surface~a5JSe/JCu!. This ratio is proportional to the ratio of concentrations in the solution Se~IV !:Cu~II ! because the deposition othose two elements is limited by diffusion,

a5K@Se~IV !#/@Cu~II !#. ~1!

The proportionality factorK, related to the diffusion co-efficients of Se~IV ! and Cu~II ! and to the hydrodynamic con-ditions in the deposition cell, is in our caseK'1.4.

This is summarized by the reaction

Cu~II !1aSe~IV !1~214a!e2→CuSea . ~2!

At potentials,20.85 V ~versus MSE!, or ,20.6 Vwhen In~III ! is present in solution~the free energy of forma-tion of In—Se bonds lowers the potential of reaction!, theselenium deposited can be further reduced as Se~2II ! whilereacting with In~III !,

3Se~2II !12In~III !→In2Se3 . ~3!

In the presence of excess In~III ! in solution, the In2Se3formation is limited by the kinetics of the reaction, i.e., controlled by the deposition potential. The CIS formation theoccurs according to

In2Se31Cu2Se→CuInSe2 . ~4!

In Fig. 1 is shown the zone diagram representing tcomposition of the electrodeposited films as a function of tapplied potential and of the solution composition. Fora,2the film is copper rich and fora.2 the film becomes indiumrich in a given range of deposition potential. The precurs

Guillemoles et al.


o

-le

m.

.ss

-n

ern

os

films studied in this work were prepared at21 V ~versusMSE!, with a spanning a range from 2.1 for In-rich to 1.9 foCu-rich films.

B. Composition of the films

After the Se treatment, in opposition to what was somtimes observed for annealing in inert atmospheres at tperatures above 400 °C, EDS measurements showed n

FIG. 1. Qualitative phase diagram of electrochemically deposited cpounds in the Cu–In–Se system.a is the ratio of Se over Cu fluxes of ionto the electrode.

J. Appl. Phys., Vol. 79, No. 9, 1 May 1996

Downloaded¬25¬Sep¬2002¬to¬134.157.181.198.¬Redistribution¬sub

r

e-em-ap-

preciable change in the Cu/In ratio in the film. This isimportant since this means that this ratio is already fixedfrom the electrochemical step.

SIMS experiments have been performed on the films before and after the annealing under Se pressure. An exampis shown in Fig. 2. Before treatment@Fig. 2~a!# it immedi-ately appears that the signals are not constant across the filThis indicates that there is a gradient along the directionperpendicular to the substrate. It is difficult to define pre-cisely its origin, but it may be due to a change either in thecomposition of the film, or to a change in its structure, orboth. A remarkable change takes place after [email protected]~b! and 2~c!#: All the signals become flat within the film.This indicates a complete homogenization of the film, incomposition and structure, perpendicularly to the substrateOne other important feature is the increase of the sharpneof the transitions between the different layers. This alsoshows the homogenization of the film since nonabrupt transitions are generally associated with local inhomogeneities icomposition or structure.

The global Se content has a tendency to increase afttreatment. This has been attributed to the formation of ainterfacial layer of MoSe2 between the Mo substrate and theCIS film. The presence of this interfacial film has been evi-denced by x-ray measurements27 showing characteristicpeaks of MoSe2 on selenized samples only@see also Fig.4~c!#. This MoSe2 interfacial layer is composed of smallgrains ~'50 nm! preferentially oriented with the metallicplanes perpendicular to the substrate.27

m-

FIG. 2. SIMS profiles vs the sputter etching time of a nearly stoichiometricfilm ~a! before and~b!, ~c! after selenization at 450 °C,PSe'1022 atm.

7295Guillemoles et al.

ject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp

ne

ar-

r-l-s

rt

io

rtaeicllic-utof

fi-ithv-

r-ta-

ne

-willir-i-ngralhate-

The presence of the interfacial MoSe2 layer is also vis-ible on the SIMS profiles, with an increase of the Se signnear the Mo interface together with a decrease of Cu andsignals. Little SIMS data on CIS is available now, but aincrease of Se near the Mo interface was also foundothers.1~b! The thickness of this layer~up to 400 nm! could beobtained from the SIMS profiles and was found to dependthe treatment and the thickness of the CIS film.

Some samples@Fig. 2~b!# gave a significant Cu contentinside the MoSe2 region. While no Cu–Mo–Se ternary couldbe identified by x-ray diffraction~XRD!, we think this isprobably due to some migration of Cu into the lamellaMoSe2. This feature is not systematic, as seen in Fig. 2~c!.We also present in this figure the profile of oxygen concetration. It is seen that O is present on the surface, probablyoxides,53 and in the Mo film.

C. Structural studies and sintering mechanism

Before annealing, the films are composed of very smagrains~'50 nm!. They are polyphasic with the presence oCuInSe2 and excess In2Se3 or Cu2Se. Guillen and Herrerohave shown that the presence of In2Se3 and Cu2Se simulta-neously was also possible in as-grown films.28 Annealingtreatments in inert atmosphere or vacuum improve markedthe crystallinity of the films as shown by various groups;4–6

however, as shown previously the presence of selenium fther improves dramatically the recrystallization.14 Largegrains with size up to 10mm, and extending over the wholefilm thickness, were formed under high selenium pressurAll the films were very adherent.

The composition of the films, regarded as the Cu/In rtio, has also a dramatic influence on the recrystallization prcess and the final morphology of the film. Figures 3~a!, 3~b!,and 4~a! and 4~b! show the scanning electron microscop~SEM! and XRD results obtained with two films treated together under moderate selenium pressure conditions~1022

atm! with one slightly indium rich~a! and the other slightlycopper rich~b!. The indium rich film@Fig. 3~a!# remains very

FIG. 3. SEM micrographs of CuInSe2 films after the selenium treatment~430 °C,PSe'1022 atm! for ~a! a slightly indium-rich film and~b! a slightlycopper-rich film.


Downloaded¬25¬Sep¬2002¬to¬134.157.181.198.¬Redistribution¬subje

alInnby

on

r

n-as

llf

ly

ur-

es.

a-o-

y-

compact, the effect of recrystallization appearing clearly othe superficial texture. The film is strongly oriented with th^112& axis perpendicular to the substrate@Fig. 4~a!#. The ex-tent of preferential orientation can be measured by comping the ratio between the intensities of the^112& and the^204& peaks, a factor of 10 is obtained in this case. An inteesting point is that the characteristic reflections of the Chacopyrite structure are now obtained for indium-rich film@Fig. 4~c!#, and not only for copper-rich films as reportedpreviously for annealing treatments carried out in ineatmospheres.11,35 For the copper-rich films, large well-shaped grains are formed, with a more open [email protected]~b!#. Interestingly, contrary to indium-rich films these filmsgive a much less marked preferential orientation with rataround 2 between the intensities of the^112& and the^204&peaks@Fig. 4~b!#.

D. Transport properties

Resistivity measurements of films annealed in an ineatmosphere for different Cu/In ratios are shown in Fig. 5 asfunction of temperature. For Cu-rich films a decrease of thconductivity with the temperature is observed. This metallbehavior is probably due to the presence of a quasimetaCuxSe phase. For indium-rich or stoichiometric films, an increase of the conductivity with temperature is observed, bthe slope is very low. This behavior corresponds to thatdegenerate semiconductors (NA.1019 cm23! explaining thepoor photovoltaic properties of these films.

Figure 6 shows the resistivity curve as a function otemperature for an indium-rich film before and after selenzation. In that case the larger dependence of resistivity wtemperature indicates a well defined semiconducting behaior. From Hall-effect measurements these films arep typewith acceptor densities in the 1016–1017 cm23 range and mo-bilities comprised between 10 and 50 cm2/V/s. The curvatureof the resistivity versus temperature plot is usually intepreted for polycrystalline materials by assuming differenconduction mechanisms in different ranges of temperture.29,30 An activation energy of'85 meV in the high-temperature range and of'40 meV in the low-temperaturerange can be derived. However, in the case of polycrystallimaterials, with grain sizes in themm range and active impu-rity concentrations in the 1016–1017 cm23 range, it isexpected31,32 that the transport will be controlled by the barriers at grain boundaries. Because these grain boundarieshave different barrier height due to the local variation of thecrystalline orientation or the impurity concentration fluctuations, following Ref. 33, we will assume a Gaussian distrbution of barrier heights. This assumption, besides beinatural, relieves us from the necessity of assuming sevetransport mechanisms at temperatures above 77 K. In tcase, the expression of the thermally activated resistivity bcomes

r5r0 expFEb

kT21

2 S d

kTD2G , ~5!

whereEb is the average barrier height andd is its standarddeviation.

Guillemoles et al.

ct¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp

l

J. App

Down

FIG. 4. XRD spectra corresponding to the samples of Fig. 3:~a! In-richfilm; ~b! Cu-rich film, showing the strong dependence of the preferentiaorientation on the Cu/In ratio; and of Figs. 2~a! and 2~b!: ~c! In-rich film,with an expansion~top! of the spectra showing the characteristic chalcopy-rite reflections~CIS-c! and the MoSe2 peaks.

t

s

it

c-

i-e-

The best fit of the experimental data~Fig. 6! gives adistribution centered atE0'150 meV, with a standard deviation of 40 meV. This is in good agreement with the modelSeto34 which relates grain size, surface states density, doplevel, and barriers energy. The corresponding surface sdensity is 1012 cm22.

For comparison the resistivity/temperature behaviorthe as-grown film is shown in Fig. 6. It appears also a

FIG. 5. Resistivity vs temperature plots for~a! stoichiometric and In-richfilms, squares and triangles, respectively, and~b! Cu-rich film, annealed at400 °C in a nitrogen atmosphere.

l. Phys., Vol. 79, No. 9, 1 May 1996

loaded¬25¬Sep¬2002¬to¬134.157.181.198.¬Redistribution¬sub

-ofingates

ofa

degenerate semiconductor. This explains by the way whycan be deposited cathodically.

E. Photoluminescence studies

Photoluminescence is a useful tool to assess the eletronic quality of CIS or CIGS~Cu~In,Ga!Se2! films.9,19,34

Films annealed in neutral atmospheres have shown no lumnescence signal but once annealing is carried out in a sel

FIG. 6. Resistivity vs temperature plots for a stoichiometric film before andafter selenization~PSe'1022 atm!.


ject¬to¬AIP¬license¬or¬copyright,¬see¬http://ojps.aip.org/japo/japcr.jsp

ln

l

rtrs

l

s

.t-s

te

nium atmosphere, a photoluminescence signal appears, inpendently of the initial composition of the film. It increasewith the selenium pressure, and the position of the dominapeak is displaced toward higher energy.14 In that case theshape of the luminescence signal is similar to that of a cevaporated film.

We have studied in detail the influence of temperatuand light intensity on the luminescence spectra of stoichimetric or indium rich films. In Fig. 7~a! we show for instancethe influence of the laser intensity. It appears that the pemaximum is moving to higher energies with increasing lasintensity whereas the slope in the low energies remainsmost intensity independent. This behavior gives insight inthe mechanism of emission in these films. We interpret it arelated to donor–acceptor pair recombination as in Refs.and 34. From the temperature dependence of the intensitythe main band@Fig. 7~b!#, we infer that the donor states arelocated at about 5–10 meV from the mobility edge, if weassume the donor states to be shallower, in accordance wthe smaller effective masses of electrons.

F. Photoelectrochemical characterization

To assess the photovoltaic quality of the films, measurments have been carried out with an electrolyte junctiousing an acidic electrolyte. A negative photocurrent was oserved for applied potentials more cathodic than20.8

FIG. 7. Photoluminescence of a stoichiometric film after selenizatio~PSe'1022 atm!, ~a! under different power excitation intensities and~b! atdifferent temperatures. The measurement conditions are reported abovecurves.



de-snt

o-

reo-

akeral-tos19of

ith

e-n,b-

V/MSE. This indicates ap-type conductivity. For potentialsmore cathodic than21 V/MSE, an important increase of thedark current is observed. Hence, the choice of the potentiaof 21 V to carry out the spectral response measurements. IFig. 8 it shown a typical spectral response obtained withthese films. It is plotted in relative units to correct for thesurface recombination losses still present in the potentiarange where the measurements are done~21 V/MSE!. Thespectral response is characterized by a decay toward showavelengths which is not expected by the classical model fophotoresponse in a Schottky diode. This behavior which waalso observed for solid-state devices9 can be attributed to thepresence of a dead layer at the surface. A simple modeassuming a semi-infinite semiconductor accounts for thisphenomenon,

QE~l!5e2adS 12e2aW

11aLnD , ~6!

whered is the thickness of the dead layer,a the absorptioncoefficient, andLn the diffusion length of minority carriers~electrons here!.

Fitting procedures with the above relation, on the rela-tive shape of the curves and using absorption coefficientpreviously determined9 have been carried out. Assuming avalue of the space charge width of 0.3mm ~obtained fromC–V measurements!, effective values of the dead layerthickness and the diffusion length can be determined. In Fig8 an experimental and a calculated spectral response are ploted together for comparison. A set of representative results ishown in Table I. It appears that effective diffusion length

TABLE I. Determination of diffusion lengthLn and dead layer thicknessd,with a semi-infinite approximation and a more complete thin-film model, onelectrodeposited/selenized CuInSe2 layers.

Sample

Semi-infinite model Thin-film model

Ln d ~mm! Ln d ~mm!

A 0.5 0.06 0.4 0.06B 0.7 0.07 0.6 0.07C 2.1 0.17 1.8 0.17D 2.3 0.14 2 0.14

n

the

FIG. 8. Spectral response of a Se-annealed film measured in the electrolyat 20.95 V/MSE ~solid line!. Calculated curve~dotted line! with Ln anddead layer of 2mm and 150 nm, respectively.

Guillemoles et al.


ee

.

-

elh

r

e

0

i

g

o-

esc-as

hatatichentto

e-f

b-of

t

he
values are in the 0.5–2.3mm range, a confirmation of thegood electronic behavior of the films.
G. Device analysis

Devices made on the films give reproducible efficienciin the range of 3%–4%, with a value of 6.5% for the best cto date, associated with an open-circuit voltage of 350 mVshort-circuit currentISC'28 mA cm22, and a fill factor FF'59%. TheI –V characteristic of this cell is shown in Fig. 9Short-circuit photocurrent values as high as 35 mA/cm2 areobtained in some cases.35 As the efficiencies are low as compared to those expected from the film properties~transport,diffusion length!, a precise analysis has been carried outdetermine the origin of losses in these devices.

Spectral responses, recorded at different biases, are slar. The factor by which they decrease under forward biasindependent on the wavelength:13 Unlike space-charge re-combination, the recombination affects the carriers indepdently of the position of their generation. This is generalindicative of interface recombination. To check this hypotesis further, the dependence of the normalized photocurras a function of potential has been studied for various illmination levels~the photocurrent is normalized to its largevalue, i.e., at reverse bias!. The results are shown in Fig. 10for three different illumination levels. From reverse to foward bias the photocurrent is first almost independent of ptential and then decreases abruptly. The transition potenbecomes more and more positive with increasing illumintion. These behaviors are characteristic of potential depdent photocurrent collection due to interface recombintion.13,36

Direct evidence of the presence of interface states canobtained by using photocapacitance measurements. Fig11 presents~capacitance!22–voltage curves measured at 5kHz, in darkness and under weak illumination~3 mA/cm2!.From negative to positive potentials, the apparent doplevel decreases from'231016 to '231015 cm23. The firstvalue is indicative of the bulk doping level in the CIS filmthe second one is generally attributed to an interfacial copensated layer containing deep traps.36 Illumination inducesa negative shift of the curve, indicating also the presenceinterface states. The corresponding photocapacitance si

FIG. 9. I –V plot of the best cell elaborated under light~87.5 mW/cm2! anddark conditions. Total area: 0.1 cm2. Efficiency is 6.5% under AM1.5 with-out antireflective coating.



sll, a

to

imi-is

n-y-entu-r

-o-tiala-n-a-

beure

ng

,m-

ofnal

presents a maximum at about 0.2 V~inset!. The shape is alsosimilar to that obtained with the first generations of evaprated cells36 which was also limited by interfacial recombi-nation.

We observed an important amelioration of the devicwhen the CIS was annealed in air at 200 °C before the juntion realization. A second heat treatment in air at 200 °C walso found to improve the devices.

IV. DISCUSSION

From the results presented above, it appears clearly tannealing under high selenium pressure leads to dramchanges in the structural and electronic properties of tfilms. Those modifications are exemplified in the importaimprovement of the performances of the solar cells duethis treatment.

The recrystallization of the films is spectacular, espcially if we keep in mind the relative low temperature oannealing~400–450 °C! much below the melting point ofCIS ~1000 °C!. The sintering process also takes place at sustrate temperatures much lower than the melting point

FIG. 10. Normalized photocurrent as a function of voltage for differenillumination levels: ~1! no light bias; ~2! light bias leading toISC53mA/cm2: ~3! light bias leading toISC525 mA/cm2. The photocurrent wasobtained with a white source and detected with a lock-in analyzer. Tcurrent is normalized to its saturation value, in reverse bias.

FIG. 11. Capacitance22 vs voltage of a 5.5% effective device~a! in darknessand~b! under illumination withISC53 mA/cm2. The frequency was 50 kHz.Curve ~c! corresponds to the photocapacitance signal~Cill2Cdark!/Cdark,from curves~a! and ~b!.



o

a

ah

l

fa

u

o

s

o

b

tsh

r

e

e

y-w

he

ts

s-the-s.s.o-icr-

psn

eis

ofthutitye

ee-asb-te

Cu2Se~T5525 °C!. As a consequence the recrystallizationnot assisted by molten Cu2Se as assumed in the caseevaporated films,37 but more likely to a sintering effect in-volving elemental selenium, as for instance a liquid phasurrounding the CIS grains and promoting the coalescebetween the grains by liquid phase sintering. This is asupported by SEM observations showing rounded grainsdeep grooves between the grains. Moreover, the grain sare larger than those reported on Cu-rich evaporated CNevertheless the differences observed between copper-indium-rich films still indicate a specific influence of copperich films on the formation of the liquid phase.

In any case, the recrystallization process is concomitwith a higher mobility of the ions on the cation sublatticeshown, for instance, by the appearance of Chalcopyrite cacteristic peaks on XRD spectra. This can be expectedcause the principal effect of an annealing under high chagen pressure of a Chalcopyrite-type compound is to crecation vacancies.38 These explain the improved mobility ocations in the CIS for annealing under Se pressure compto annealing in vacuum or neutral gas atmospheres.

Photoluminescence characterizations of our films gaanother occasion to illustrate this feature. As shown in F7~a!, there is an exponential tail in the PL spectra that shoreflect the density of states in the compound. A disordparameter can be extracted from the slope of the low-enebranch of the luminescence peak, corresponding to thebach slope in absorption studies,39,40since emission, like ab-sorption, is related to the density of states of the semicductor. We define it by

I5I 0 expS EEuD . ~7!

Eu is the pendant of Urbach slope in emission procesand can be related to the local fluctuations with respect toperfect crystalline lattice.41 Temperature variation ofEu

shows that the thermal disorder~phonons! is not responsiblefor the tail in the emission, but that it is rather causedalloy disorder.42

Values ofEu'25–35 meV are obtained for our films, tbe compared with'20 meV for coevaporated films. Thesvalues, larger than for classical semiconductors~a few meV!,indicate some alloy disorder effects in the films. It is tostressed thatEu will depend on the Se pressure for a givesample~Fig. 12!. This, together with the appearance of Chacopyrite peaks on XRD spectra shows that the annealingder Se pressure reduces the disorder on the cation subla

From this and from the dependence of emission intenwith Se pressure, we can infer that this treatment diminisdrastically the number of defects present in the material; bthere is more than this in it: Not only is there a healing effeon the defects by the Se treatment~as indicated by the in-crease in intensity and shown above by other methods!, butthe nature of dominant defect type is also changed. Fortermediate Se pressures the dominant transition is diffefrom the one obtained at strong Se pressures, i.e., theyvolve different defect levels.14

It is now well documented that the electronic propertiof CIS are dominated by its intrinsic points defects.43–45 It



isf

sencelsondizesIS.andr-

antsar-be-co-ate

red

veig.lderrgyUr-

n-

esthe

by

e

enl-un-tice.ityesut,ct

in-entin-

s

has been widely acknowledged38,46–48that the point defectsmost likely to occur during an annealing under strong Sactivity areVCu andVIn . The former should dominate in theindium-rich samples~that are also the layers important forelectronic applications!. A cationic disorder, not precludedby the existence of the XRD peaks of the ordered Chalcoprite phase, is to be expected in this material due to the loformation energy of the related point defect InCu and CuIn .For obvious reasons, the former should predominate in tindium-rich samples. Moreover, it is the only donor of thefour quoted point defects. Thus, from likeliness argumenwe will attribute the high-energy transition emerging withstrong Se pressure in our material to an InCu–VCu pair.

These results demonstrate that by varying the Se presure during the annealing it becomes possible to changedominant defect levels in this material, and to tailor its electronic properties via the concentration of its intrinsic defectThis is also shown by the analysis of the transport propertieAs a consequence, after heat treatment in selenium atmsphere the film becomes apparently suitable for photovoltaapplications. This we have checked by looking at the chaacteristics of liquid and solid junctions.

Photoelectrochemical characterization gave two grouof samples. In the first group represented by A and B iTable I, the diffusion length is around 0.5mm, but the deadlayer thickness is small, of the order of tens of nm. In thsecond group represented by B and C, the diffusion lengthmuch higher, around 2mm, but the drawback is that the deadlayer thickness is also much more important, in the range0.15 mm. It seems that a correlation exists between bovalues. In the second group as the diffusion length is abothe thickness of the film, questions appear about the validof the application of the semi-infinite hypothesis under thesconditions. An extended model was thus used49 to considerthe finite thickness of the layer, the possibility of reflectionon the back surface, and the recombination velocity on thback contact. The results are shown in Table I. The effectivdiffusion length is only slightly reduced by using the extended model. These values, which have to be consideredindicative since uncertainties exist on the values of the asorption coefficients, indicate that the films really presengood quality. It can be noted that the best films are thos

FIG. 12. Variation of the disorder parameterEu as a function of Se sourcetemperature during the annealing treatment~temperature of substrate:450 °C! for identical samples.

Guillemoles et al.


u

e

e

r

.6o-

ene-

-

h.a

ead

lefur-ug-s-er.c-tralthethofdyur-

.ersi-

ri-tchSeofas-reas

’’e,er,ghk--m-

ors torlorl-

which present a thicker dead layer~above 100 nm!. The ex-istence of the dead layer and the consistency of the reshave been checked by measuring its real thickness by meof etching experiments. The removal of the dead layer, asciated with the appearance of an increase of the quantefficiency at short wavelengths, takes place after a 25 s ein a 0.5% bromine in methanol solution@Fig. 13~a!#. Fromthe calibration with a profilometer, this corresponds to thremoval of about 0.15mm @Fig. 13~b!#, which is in goodagreement with the values obtained for the dead layer frothe model.

Concerning the values obtained by the extended modit is also interesting to note that the back contact recombintion velocity is best fitted by zero, contrary to the infinitvalue generally assumed for an ohmic contact. This suggethe possibility of a back surface field effect in this structurIt could possibly be associated with the presence of tMoSe2 layer at the interface

50 since this compound isp typewhen produced under excess Se pressure and has a widethan CuInSe2 @Eg ind.'1.1 eV, dir.51.4 eV ~Ref. 51!#.Moreover, because of its preferential orientation, this layshould induce no additional series resistance in the dev~the transverse conductivity is large!. Thus, in oppositionwith the conclusion of other authors,52 we think that thisinterfacial layer might improve the quality of the devices.

It can be noted that the analysis made previously on5.6% efficient solid-state device leads to a value of 0.035mm

FIG. 13. ~a! Evolution of the spectral response of a Se-annealed film mesured in the electrolyte at20.95 V/MSE after different etching times in a0.5% Br–methanol solution. Before etching the decrease at short walength could be fitted to a dead layer of 0.14mm. After 25 s, there is nolonger a decrease at short wavelength.~b! The amount of compound re-moved by the etch was calibrated with a profilometer and gives 0.16mm, inclose agreement with the model.



ltsansso-umtch

e

m

el,a-

sts.he

gap

erice

a

for the dead layer thickness and to a diffusion length of 0mm.9 This is consistent with the present results from photelectrochemistry and indicates that the CuInSe2 layer in thatcase belongs more likely to the first group. We have sethat our cells are prone to interface recombination. This rcombination is very likely to occur in the very defective‘‘dead layer,’’ which explains why samples with smaller diffusion length~and thinner dead layer! gave better solid-statejunctions than the samples with the highest diffusion lengtThe dead layer, present on the surface of the grains, ishighly defective layer and thus very conductive~this is anecessary condition for the carriers generated inside the dlayer not to be collected!. A thick dead layer would then giveshunted devices, as is indeed observed.

The analysis of both types of junctions gave reasonabto large diffusion length confirming the intrinsic quality othe material, but the correlation between the Se pressure ding the annealing and the thickness of the dead layer sgests a rationalization of the observations. At high Se presures, the recrystallization is more important, due to a highmobility of the cations and to a thicker liquid surface fluxThis liquid surface flux, upon cooling, leaves a very defetive, Se-rich, surface which is seen as a dead layer in specresponse measurements. The better recrystallization inbulk is made sensible by the increase of the diffusion lengand the thicker dead layer is related to the bigger amountflux. We interpret the improvement of the devices obtaineby annealing in air before the realization of the junction, ban oxidation of the excess elemental Se present on the sface and of the cations~Cu, In! which are then removedduring the subsequent chemical bath deposition of CdS53

The improvement of devices due to annealing in air aftcompletion of the devices was already explained by a pasvation of the interface states.54

We have also observed in photoelectrochemical expements that the dead layer is removed by Br–methanol ebut not by heat treatment in air, so it is not an elementallayer. We can propose an explanation for the formationthis layer if we consider the processes occurring at the geous Sep-type CIS during the selenization. The Se pressuwill cause a creation of cations vacancies on the surface,we have seen, but due to the band bending in thep-type CIS,the negatively charged cation vacancies will be ‘‘trappedby the surface. Thus, the surface will be enriched in Scompared to bulk, with increasing Se pressure. This laywith a very high concentration of vacancies, and so a himobility of the elements, could act as the flux we were looing for. Copper being quite mobile in CIS, it is also in agreement with the observation that the phenomenon is more iportant in Cu-rich films.

V. CONCLUSION

Selenization of electrochemically deposited precursfilms of CIS has been demonstrated as an efficient procesprepare high-quality films. Controlling both the precursocomposition and the selenium pressure allowed us to taithe characteristics of the film in a range suited for photovotaic application: 1016–1017 cm23, p type, large diffusionlength ~around 1mm!.

a-

ve-



em.

ite

c.

. J.

k,

n

em.

S.

E

, P.

m.

a-

.

.nto-

Present limitation in cell efficiency appears to be relateto losses at the interface~recombination, dead layer! and notto bulk properties of the CIS film. This result can be furthediscussed in comparison with the present status of higefficiency cells made by evaporation. For these cells, the Csurface is assumed to be naturally covered by a well-defincompound, corresponding to the ordered vacancy copounds CuIn3Se5,

55 leading to a good interfacial structurewith CIS. Moreover this compound isn type, giving as-grownp/n junction formation, which is highly favorable forthe cell. In our case, these processes do not seem to existour results show that the presence of excess Se probainduces the formation of a dead layer with a high densitysurface states. It is interesting to note that this seems tocommon to selenization under high pressure. Recent resuconcerning the selenization of elemental Cu/In layers undehigh selenium pressure, also indicate strong limitationsinterface recombination.56 As a consequence, a key poinnow to further improve the cells is to develop specific chemcal interface engineering for these films, including probaban etching process.

ACKNOWLEDGMENTS

This work is supported by the Commission of the European Communities under a JOULE contract ‘‘EuroCIS~JOUR-0045-C! and the ECOTECH-ADEME~Contract No.94N80/0020!. We thank Dr. A. Forveille Boutry of the Lab-oratoire de Physique des Solides de Bellevue, CNRS, forSIMS measurements and Dr. L. Robbiola~CNRS/ENSCP!for SEM pictures.

1~a! J. Hedstro¨m, H. Ohlsen, M. Bodegard, A. Kylner, L. Stolt, D. Haris-kos, M. Ruckh, and H. W. Schock, in 23rd IEEE Photovoltaic SpecialisConference, Louisville, 1993, p. 364;~b! M. A. Contreras, J. Tuttle, A.Gabor, A. Tennant, K. Ramanathan, S. Asher, A. Franz, J. Keane,Wang, J. Scofield, and R. Noufi, in 1st WEPEC~24th IEEE PhotovoltaicSpecialists Conference!, Hawaii, 1994, p. 68.

2J. M. Woodcock, M. E. Ozsan, A. K. Turner, D. W. Cunningham, D. RJohnson, R. J. Marshall, N. B. Mason, S. Oktik, M. H. Patterson, S.Ransome, S. Roberts, M. Sadeghi, J. M. Sherborne, D. Sivapatsundaram, and I. A. Walls, in Proceedings of the 12th European Phovoltaic Solar Energy Conference, Amsterdam, April 1994, p. 948.

3G. Fullop, M. Doty, P. Meyers, J. Betz, and C. H. Liu, Appl. Phys. Let40, 327 ~1982!.

4F. J. Pern, J. Goral, R. J. Matson, T. A. Gessert, and R. Noufi, Sol. Ce24, 81 ~1988!.

5Y. Sudo, S. Endo, and T. Irie, Jpn. J. Appl. Phys.32, 1562~1993!.6S. N. Sahu, R. D. L. Kristensen, and D. Haneman, So. Energy Mater.18,385 ~1989!.

7C. Guillen, J. Herrero, and D. Lincot, J. Appl. Phys.76, 359 ~1994!.8C. X. Qiu and I. Shih, Can. J. Phys.67, 44 ~1989!.9J. F. Guillemoles, P. Cowache, S. Massaccesi, L. Thouin, S. SanchezLincot, and J. Vedel, Adv. Mater.6, 379 ~1994!.

10S. N. Qiu, L. Li, C. X. Qiu, and I. Shih, inProceedings of the 12thEuropean Photovoltaic Solar Energy Conference, Amsterdam~Stephens,Bedford, 1994!, p. 1539.

11L. Thouin, S. Rouquette-Sanchez, and J. Vedel, inProceedings of the 10thEuropean Photovoltaic Solar Energy Conference, Lisbonne ~Harwood,Chur, 1991!.

12C. Guillen, E. Galiano, and J. Herrero, Thin Solid Films195, 137 ~1991!.13J. F. Guillemoles, P. Cowache, L. Thouin, S. Massaccesi, S. SancheVedel, and D. Lincot, inProceedings of the 12th European PhotovoltaiSolar Energy Conference, Amsterdam~Stephens, Bedford, 1994!, p. 1550.

14J. F. Guillemoles, A. Lusson, P. Cowache, S. Massaccesi, J. Vedel, anLincot, Adv. Mater.6, 376 ~1994!.



d

rh-ISedm-

andblyofbelts,r abyti-ly

-’’

the

ts

L.

.J.ha-to-

t.

lls

, D.

z, J.c

d D.

15L. Thouin, S. Massaccesi, S. Sanchez, and J. Vedel, J. Electroanal. Ch374, 81 ~1994!.

16Kamel Fezzaa, Memoire de DEA en Sciences des Materiaux, Univers´Paris VI, 1993.

17M. Pourbaix, inAtlas d’equilibres Electrochimiques~Gauthier-Villars,Paris, 1963!, p. 276.

18J. A. von Windheim, H. Wynands, and M. Cocivera, J. Electrochem. So138, 11 ~1991!.

19M. Tanda, S. Manaka, A. Yamada, M. Konagai, and K. Takahashi, JpnAppl. Phys.32, 1913~1993!.

20D. Lincot, H. Gomez, J. Kessler, J. Vedel, B. Dimmler, and H. W. SchocSol. Energy Mater.20, 67 ~1990!.

21D. Lincot and R. Ortega-Borges, J. Electrochem. Soc.139, 1880~1992!.22R. H. Mauch and H. W. Schock, inProceedings of the 10th EuropeanPhotovoltaic Solar Energy Conference, Lisbonne~Harwood, Chur, 1991!,p. 88.

23The illumination was performed with an ELH lamp calibrated to give aincident power of 87 mW/cm2.

24L. Thouin, S. Sanchez, and J. Vedel, Electrochim. Acta38, 2387~1993!.25S. Massaccesi, S. Sanchez, and J. Vedel, J. Electrochem. Soc.140, 2540

~1993!.26L. Thouin, S. Massaccesi, S. Sanchez, and J. Vedel, J. Electroanal. Ch374, 81 ~1994!.

27L. Thouin, J. F. Guillemoles, S. Massaccesi, R. Ortega, P. Cowache,Sanchez, D. Lincot, and J. Vedel, inProceedings of the 11th EuropeanPhotovoltaic Solar Energy Conference, Montreux ~Stephens, Bedford,1994!, p. 866.

28C. Guillen and J. Herrero, J. Electrochem. Soc.141, 225 ~1994!.29C. Guillen and J. Herrero, J. Appl. Phys.71, 5479~1992!.30N. Agli, M. Abassi, A. Maasmi, N. Benslim, L. Mahdjoubi, M. Pas-quinelli, and S. Martinuzzi, Solid State Phenom.37-38, 521 ~1994!.

31R. L. Petriz, Phys. Rev.104, 1508~1956!.32J. Y. W. Seto, J. Appl. Phys.49, 5247~1978!.33J. H. Werner, Solid State Phenom.37-38, 213 ~1994!.34R. E. Hollingsworth and J. R. Sites, in Proceedings of the 18th IEEConference, 1985, p. 1409.

35L. Thouin, J. F. Guillemoles, S. Massaccesi, R. Ortega, S. SanchezCowache, D. Lincot, and J. Vedel, Solid State Phenom.37-38, 527~1994!.

36R. K. Arhenkiel, J. Appl. Phys.59, 181~1986!; Sol. Cells16, 549~1986!.37R. Klenk, T. Walter, H. W. Schock, and D. Cahen, Solid State Pheno37-38, 509 ~1994!.

38C. Schwab, Jpn. J. Appl. Phys.19-3, 30 ~1980!.39M. Neumann-Spallart, E. Galun, G. Hodes, C. Levy-Clement, Y. Marfing, E. Muranevitch, and R. Tenne, J. Appl. Phys.73, 7753~1993!.

40H. Nakanishi, T. Sawaya, S. Endo, and T. Irie, Jpn. J. Appl. Phys.32-2,200 ~1993!.

41J. Jackle, Philos. Mag. B41, 681 ~1980!.42J. F. Guillemoles, A. Lusson, P. Cowache, J. Vedel, and D. Lincot~un-published!.

43C. Rincon, C. Bellabarba, J. Gonzalez, and G. S. Perez, Sol. Cells16, 335~1986!.

44S. Nomura, J. Itoh, and T. Takizawa, Jpn. J. Appl. Phys.32-2, 97 ~1993!.45H. Neumann and R. D. Tomlinson, Sol. Cells28, 301 ~1990!.46B. Tell, J. L. Shay, and H. M. Kasper, J. Appl. Phys.43, 2469~1972!.47G. Masse´, J. Appl. Phys.68, 2206~1990!.48D. Y. Chang and H. Y. Ueng, Jpn. J. Appl. Phys.32-3, 198 ~1993!.49See, for instance, Y. Marfaing, inEnergie Solaire Photovoltaique

~UNESCO & Ellipse, Paris, 1993!, Vol. 1, pp. 104–124.50J. F. Guillemoles, Ph.D. thesis, University of Paris VI, Paris, 1994.51W. Kautek, H. Gerisher, and H. Tributsch, J. Electrochem. Soc.127, 2471

~1980!.52E. Niemi, Jpn. J. Appl. Phys.32-3, 50 ~1993!.53J. Kessler, K. Velthaus, M. Ruckh, R. Laichinger, H. W. Schock, DLincot, R. Ortega-Borges, and J. Vedel, in6th International PhotovoltaicScience and Engineering Conference~Oxford University Press, Delhi,1992!, p. 1005.

54D. Cahen and R. Noufi, Appl. Phys. Lett.54, 558 ~1989!.55D. Schmid, M. Ruckh, F. Grunwald, and H. W. Schock, J. Appl. Phys.73,2902 ~1993!.

56G. Attar, A. Muthaiah, H. Natarajan, D. Nierman, S. Karthikeyan, SZafar, C. S. Ferekides, and D. L. Morel, in 1st World Conference oPhotovoltaic Energy Conversion, in Proceedings of the 24th IEEE Phovoltaic Specialist Conference, Hawaii, 1994, p. 182.

Guillemoles et al.


One step electrodeposition of CuInSe2: Improved structural, electronic, and photovoltaic properties by annealing under high selenium pressure

Documents