Three step h2searh2s reaction of cu in ga precursors for controlled composition and adhesion of cu(in,ga)(se,s)2 thin films

In this work, CIGSS films with homogenous Ga distribution and good adhesion were formed using a three-step reaction involving: 1 selenization in H2Se at 400 C for 60 min, 2 temperature r

Three-step H2Se/Ar/H2S reaction of Cu-In-Ga precursors for controlled composition

and adhesion of Cu(In,Ga)(Se,S)2 thin films

Kihwan Kim, Gregory M Hanket, Trang Huynh, and William N Shafarman

Citation: Journal of Applied Physics 111, 083710 (2012); doi: 10.1063/1.4704390

View online: http://dx.doi.org/10.1063/1.4704390

View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/111/8?ver=pdfcov

Published by the AIP Publishing

Advertisement:

Three-step H2Se/Ar/H2S reaction of Cu-In-Ga precursors for controlled

Kihwan Kim, Gregory M Hanket, Trang Huynh, and William N Shafarmana) Institute of Energy Conversion, University of Delaware, Newark, Delaware 19716, USA (Received 6 March 2012; accepted 14 March 2012; published online 19 April 2012) Control of the through-film composition and adhesion are critical issues for Cu(In,Ga)Se2(CIGS) and/or Cu(In,Ga)(Se,S)2(CIGSS) films formed by the reaction of Cu–In–Ga metal precursor films

in H2Se or H2S In this work, CIGSS films with homogenous Ga distribution and good adhesion were formed using a three-step reaction involving: (1) selenization in H2Se at 400

C for 60 min, (2) temperature ramp-up to 550

C and annealing in Ar for 20 min, and (3) sulfization in H2S at 550

C for 10 min The 1st selenization step led to fine grain microstructure with Ga accumulation near the

Mo back contact, primarily in a Cu9(In1xGax)4phase The 2nd Ar anneal step produces significant grain growth with homogenous through-film Ga distribution and the formation of an InSe binary compound near the Mo back contact The 3rd sulfization step did not result in any additional change in Ga distribution or film microstructure but a small S incorporation near the CIGSS film surface and complete reaction of InSe to form CIGSS were observed The three-step process facilitates good control of the film properties by separating different effects of the reaction process and a film growth model is proposed Finally, CIGSS solar cells with the three-step reaction were fabricated and devices with efficiency ¼ 14.2% and VOC¼599 mV were obtained V C 2012 American Institute of Physics [http://dx.doi.org/10.1063/1.4704390]

I INTRODUCTION

The growing demand for low cost Cu(In,Ga)Se2(CIGS)

or Cu(In,Ga)(Se,S)2(CIGSS) solar cells has led to increased

interest in the reaction of metal precursors in Se- and/or

S-containing atmospheres which may have advantages for

large-scale manufacturing.1With precursor reaction in H2Se

and H2S, 17.2% efficiency has been demonstrated on

30  30 cm2 submodules and large scale manufacturing is

underway.2,3While a number of options have been

demon-strated for deposition of the precursor films, and for the

reac-tion, a common problem is that the reacted CIGS films have

Ga accumulation near back contact which results in lower

bandgap near the front of the film and, consequently, lower

than expected VOC.46Another issue with precursor reaction

is the poor quality of the Mo/CIGSS interface which may

yield poor adhesion This narrows the process window and

causes yield issues with large area processing The adhesion

issues have been ascribed to stress built-up in the CIGSS

film due to the volume expansion from the metal precursor

to CIGSS (Ref.7) or to the formation of a MoSe2layer on

the Mo surface It has been suggested that the MoSe2

orienta-tion may determine adhesion, and the c-axis should be

ori-ented perpendicular to the Mo surface.8,9

Hanket et al.10 characterized the reaction pathways to

form Cu(InGa)Se2 or Cu(InGa)S2films from metal

precur-sors using H2Se and H2S, respectively According to their

study, there are preferential and/or fast reactions of In with

Se, and Ga with S, that yield an unfavorable Ga grading

through the CIGSS films formed by the sequential process of

selenization/sulfization The Ga and In could be

homoge-nized but slow interdiffusion required annealing at 600

C for 1 h.5Other studies have successfully demonstrated a Ga homogenization by adjusting the controlling reaction,11–14 but often with poor adhesion Another approach by Kim

et al.15

introduced a three-step reaction process which included selenization at 400

C, annealing at 550

C and, finally selenization at 500

C They realized partial Ga ho-mogenization with improved adhesion but the obtained

VOC500 mV was low relative to that expected for the bandgap estimated by the Ga alloying This was attributed to

Ga depletion near the surface which apparently occurred dur-ing the 3rd step selenization at 500

C

In this work, we report a three-step process for reaction

of Cu-Ga-In precursor films that gives good control of adhe-sion and through-film composition with enhanced grain size and produces high VOCin solar cells made from these films The process includes an initial 400

C reaction in H2Se fol-lowed by an anneal step in Ar at 550

C and finally a reaction

in H2S at 550

C Material characterizations for each step have been carried out, and we describe the film formation pro-cess in terms of grain growth and Ga distribution Thin film solar cells were fabricated with the CIGSS absorber layer and the devices exhibited efficiency >14% with VOC600 mV

II EXPERIMENTAL

Mo (700 nm) and metal precursors were sputter depos-ited from In and Cu0.8Ga0.2 targets onto 100

100

soda lime glass substrates The sputtering system was configured with

a rotating (5 rpm) platen so that 350 alternating layers of In and Cu0.8Ga0.2 were deposited corresponding to 700 nm which is sufficient to form a fully reacted CIGSS film with thickness 1.6  1.8 lm

a) Electronic mail: wns@udel.edu.

The metal precursors were reacted and annealed in a

200

diameter quartz tube with heating jacket which enables

tem-perature up to 600

C and a push-pull rod that enables samples

to be inserted or removed from the hot reaction zone.10Before

reaction, the reactor was evacuated to 5  106Torr to remove

moisture and impurities Then, Ar gas was introduced in the

reactor and the pressure was allowed to reach atmosphere Ar

flow was maintained through the entire reaction with a

nomi-nal turnover time of about 1 min Prior to heating, samples

were held in a room temperature end of tube

For the reaction process, the reactor hot zone was first

heated to 400

C in flowing H2Se(0.35%)/O2(0.0035%)/

Ar(balance) Then, the samples were pushed in the hot zone

and the 1st step selenization was carried out for 60 min After

the 1st step, the samples were pulled out of the hot zone and

the temperature was increased to 550

C in flowing Ar

When the temperature stabilized, the sample was moved into

the hot zone again and the 2nd step Ar annealing was

per-formed for 20 min Finally, with the samples remaining in

the hot zone, the 3rd step sulfization was carried out in

flow-ing H2S(0.35%)/O2(0.0035%)/Ar(balance) at 550

C for

10 min After the sulfization, the samples were pulled back

to the end of the tube and cooled down to T < 100

C in

60 minutes The role of O2in both selenization and

sulfiza-tion is believed to form a thin oxide layer on the top of the

precursor during the initial reaction to prevent excessive

agglomeration10and to react in the gas phase with H2Se and

H2S to form elemental Se and S.16

Solar cells were fabricated with a conventional Mo/

CIGSS/CdS/i-ZnO/ITO/Ni-Al structure A 50 nm thick CdS

layer was deposited on the CIGSS absorber layer by chemical

bath deposition (CBD) and i-ZnO (50 nm)/ITO (150 nm) layers

with sheet resistance 30X/sq were deposited by RF magnetron

sputtering Finally, a grid of Ni (50 nm)/Al (3000 nm) was

deposited by e-beam evaporation Cells were delineated by

mechanical scribing with area 0.47 cm2

Film characterization included scanning electron

mi-croscopy (SEM) images with EDS analysis using an Amray

model 1810 microscope with an Oxford Instruments

PentaFETVR

6900 EDS detector All EDS measurements were taken with a 20 kV accelerating voltage, so the

mea-surement probes approximately 1 lm or half of a 2 lm

thick CIGSS film For improved image resolution, some

SEM images were acquired using a JEOL JSM-7400 F

Asymmetric (2h scan) X-ray diffraction (XRD) analysis

was performed to characterize phases in the metal

precur-sors and reacted films using a Rigaku D/Max 2500 With

incident angles from 0.5

to 8

, XRD patterns sensitive to sampling depth were acquired Broad spectrum XRD

pat-terns were obtained with a step size of 0.05

and scan speed

of 0.5

/min, and fine XRD line profiles were taken with a

step size of 0.02

and a scan speed of 0.05

/min Auger electron spectroscopy (AES) measurements for

composi-tion depth profiling were conducted using Physical

Elec-tronics 660 Scanning Auger Microprobe Current-voltage

(J-V) characteristics including total area conversion

effi-ciency were measured under AM 1.5 illumination at 25

C and quantum efficiency (QE) was performed after solar cell

fabrication

III RESULTS

A Cu-In-Ga metal precursor SEM images of as-sputtered precursor in Fig 1 show nodules distributed on a smooth background Although the metal precursor was prepared by repetition of In and Cu-Ga layers, it seems significant agglomeration occurs and a smooth layered structure is not obtained

Table I shows wide-area and spot-EDS measurements taken from Fig.1 The wide area average EDS measurements give composition ratios Cu/(In þ Ga) ¼ 0.82 and Ga/(In þ Ga)

¼0.19 Spot-EDS measurements of the “nodules” and the smooth “background” show significant differences in compo-sition The nodules are In-rich, while the smooth background

is Cu-rich and Ga-rich compared to the average composition

XRD patterns taken with incident beam angles of 0.5

and 4

are shown in Fig.2 Since both Cu and Ga solubilities

in In are low,17 almost pure In and intermetallics such as

Cu3Ga, Cu4In, and Cu9(In1  xGax)4are observed.10,18–22The value of x in the Cu9(In1xGax)4phase is determined to be

0.4 by applying Vegard’s law.23Comparison of the different incident angles shows that the surface is relatively In-rich and the bulk film contains the Cu-rich intermetallic phases There-fore, wide-area EDS measurements lead to overestimation of the relative In composition Even though the measured aver-age Cu/(In þ Ga) ratio of the metal precursor is 0.82, the reacted film is expected to yield higher Cu/(In þ Ga)

B Film formation by three-step reaction Fig 3 shows cross sectional SEM images of reacted films after each step The 1st step (H2Se at 400

C for

60 min) leads to fine microstructure and flat voids between

FIG 1 Plan-view SEM images of a Cu-In-Ga metal precursor.

TABLE I Compositional values of wide-area and spot-EDS measurements taken from Fig 1

EDS measurement Location

Cu (at %)

In (at %)

Ga (at %)

Mo (at %)

Cu/

(In þ Ga)

Ga/

(In þ Ga)

the CIGS thin film and Mo back contact as shown in Fig.

3(a) The bottom 500 nm of the film shows finer structure

However, after the 2nd step (Ar anneal at 550

C for 20 min), the film exhibits a significant grain growth with an

agglomer-ation of voids as shown in Fig.3(b) Even though there are

still the voids in the CIGS film, the contact area between the

Mo and CIGS is effectively increased and thus might lead to

improvement of film adhesion Finally, after the 3rd step

(H2S at 550

C for 10 min), it appears that the films do not undergo any significant change as shown in Fig.3(c)

Plan-view SEM images of the reacted films from both

front and back sides along with associated EDS data are

shown in Fig.4 The images from the films back sides were

taken after delamination from the Mo/SLG using the

tech-nique described elsewhere.10 The front surfaces of the

reacted films show a similar trend to the cross sectional

observations in Fig.3 The transition between the 1st step

and the 2nd step causes a significant change in the

morphol-ogy, while the transition between the 2nd step and the 3rd step does not The EDS analysis reveals that the ratio of Ga/ (In þ Ga) also changes considerably from the 1st step to the 2nd step After H2Se reaction, the Ga content is much higher

at the back half of the film consistent with the commonly observed accumulation of Ga near the Mo contact.10,12,15 However, front and back EDS measurements gave compara-ble Ga/(In þ Ga) ratios as shown in Figs.4(c)and4(d)so the

Ar anneal caused a significant Ga diffusion through the film The H2S reaction step did not cause a significant change

in composition except S incorporation to the film surface (Figs.4(e)and4(f))

Fig.5shows the (112) peaks of XRD spectra taken from the film surface with incident angles of (a) 0.5

and (b) 8

cor-responding to sampling depths of 110 nm and 1.8 lm, respec-tively.24The (112) peak positions are summarized in TableII The results from the XRD analyses are in good agreement with the data obtained by the SEM/EDS analyses The (112) peak position after the 1st step is comparable to that of pure CuInSe2consistent with a lack of Ga incorporation into the chalcopyrite phase This shifts due to Ga homogenization after the 2nd step and the peak position 2h ¼ 26.82

(8

incident angle) corresponds to Ga/(In þ Ga) ¼ 0.19 when considering Vegard’s law and the JCPDS database.23,25,26 After the 3rd step, a broad second peak at 2h ¼ 27.70

is observed only for the scan with 0.5

incident angle and this is attributed to S dif-fusion at the front of the film

The (112) peaks taken from the film back side are shown

in Fig.6and their positions are listed in TableII The surface

of the film back side after the 1st step [Fig.6(a)] shows the

FIG 2 Asymmetric XRD patterns of a Cu-In-Ga metal precursor XRD

pat-terns were taken from incident angles of 0.5 

(black) and 4 

(red) Symbols of

n,, ~, and ! indicate In, Cu 3 Ga, Cu 4 In, Cu 9 (In 1x ,Ga x ) 4 , respectively.

FIG 3 Cross-sectional SEM images of reacted CIGS or CIGSS films after

(a) 1st step, (b) 2nd step, and (c) 3rd step.

FIG 4 Plan-view images of the front and backside of reacted CIGS or CIGSS films after each step with associated EDS data: (a) front side and (b) back side of film after 1st step, (c) front side and (d) back side of film after 2nd step, and (e) front side and (f) back side of film after 3rd step.

(112) peak at 26.86

with shoulder peaks at 27.68

and 28.04

which correspond to CuGaSe2(CGS) and CuGa3Se5,

respectively.27,28 After the 2nd step, the shoulder peaks are

absent and the difference in the (112) peak positions between

the back and front surface measurements is significantly

nar-rowed It seems that at 550

C, there is sufficient energy for the selenized film to be recrystallized with the Ga

homogeni-zation No noticeable change was observed from the film

back side after the 3rd step

C Residual intermetallics and binary compounds

SEM images of the Mo surface after delamination are

shown in Fig.7with the associated EDS data in Table III

Spot EDS on a nodule in Fig.7(a)reveals Cu and Ga while

only Mo was found from the background Most of these nod-ules (except residual CIGS particles) were not present on the

Mo surface after the 2nd step as shown in Fig 7(b) EDS data show only Mo in this case After the 3rd step, some Se and a roughened Mo surface were observed as shown in Fig

7(c)and the EDS data

Fig.8shows XRD patterns of the Mo surface as observed

in Fig 7taken with the incident beam angle ¼ 0.5

for great-est surface sensitivity The XRD pattern after the 1st step shows a prominent peak at 44.05

Considering the EDS data from the “nodule,” the peak was identified as due to the

Cu9(In1xGax)4intermetallic compound with x  0.95.10,21–23 However, no peak from an intermetallic compound is detected after the 2nd step, in accord with the SEM observation It is noteworthy that a MoSe2phase was found on the Mo surface only after the 3rd step29,30showing that Se from the film bulk diffused to the Mo to form the MoSe2 during the 3rd step sulfization

Finally, an InSe binary compound was also observed in the CIGS film after the 2nd step as shown in Fig.9.31Peaks from the InSe phase were more intense from the back side of the film than from the front side It seems that the Cu9(In

1x-Gax)4intermetallic compound reacted with the CIGS under Se- or S- starved condition, and as a result, some CIGS decomposed to form the InSe binary However, it was fully eliminated during the 3rd step

FIG 5 Asymmetric XRD patterns of the front side of reacted CIGS or

CIGSS films after each step with the incident beam angles of (a) 0.5  and (b)

8  Sampling depths with 0.5  and 8  are 110 nm and 1.8 lm, respectively.

TABLE II (112) peak position with each step and different incident beam

angles.

FIG 6 Asymmetric XRD patterns of the back side of reacted CIGS or CIGSS films after each step with incident beam angles of (a) 0.5  and (b) 8 

D Device results

Solar cells were fabricated from the reacted films and

their J-V curves with device parameters are shown in Fig.10

and Table IV With the absorber after the 1st step, the J-V

behavior is very leaky and we believe that poor crystallinity

and an incomplete reaction degrade the device performance The cell with CIGS film after the 2nd step is fully shunted Presumably, the shunting could be caused by defects formed

at the same time as the InSe such as Se-vacancies associated with Se liberation6,32 and/or Cu2Se, even though it was not identified by XRD The solar cell after the 3rd step has an efficiency of 14.2% (without AR coating) Since the Ga ho-mogenization through-film is realized and there is no interme-diate product such as InSe, Voc¼599 mV was obtained with high fill factor of 73.5%

QE measurements are shown in Fig 11 An estimated bandgap of the CIGSS determined by the inflection point at long wavelength33gives 1.09 eV CIGSS films finished with sulfization can have a bias dependent collection due to for-mation of a conduction band barrier with excess S at the sur-face.12,34,35 This can be assessed by bias dependent QE.36 The ratio QE(0 V)/QE(1V) shown in Fig.11is flat indicat-ing negligible bias dependence and, with the high FF, shows that there is no loss due to such a collection barrier

FIG 7 Plan-view images of the surface of Mo back contact after

delamina-tion of the CIGS film after (a) 1st step, (b) 2nd step, and (c) 3rd step.

TABLE III Compositional values of wide-area and spot-EDS

measure-ments taken from Fig 7

EDS measurement location

Cu (at %)

In (at %)

Ga (at %)

Se (at %)

Mo (at %)

After 1st step: Background < 0.5 — < 0.5 — 99.4

FIG 8 Asymmetric XRD patterns of the surface of Mo back contact after each step: (a) 1st step, (b) 2nd step, and (c) 3rd step All the patterns were taken with 0.5  incident angle corresponding to a sampling depth of 110 nm Symbols of n, , and ~ indicate CIGS, Cu 9 (In 1x Ga x ) 4 , and MoSe 2 , respectively.

FIG 9 Asymmetric XRD patterns of the back side of delaminated CIGS films after (a) 2nd step and (b) 3rd step The InSe phase observed after the 2nd step was completely removed by the 3rd step.

IV DISCUSSION

The most prominent aspect of the three-step reaction is

the Ar anneal step which creates through-film Ga

homogeni-zation with marked grain growth AES depth profiles after

each step, shown in Fig.12, describe the film evolution The

depth profile taken from the 1st step indicates Ga

accumula-tion and high concentraaccumula-tions of Cu and Ga at the film/Mo

interface consistent with a Cu-Ga intermetallic

accumula-tion A considerable Ga redistribution is realized by the 2nd

step and sulfur incorporation is observed with the 3rd step

only in the surface region A film-growth model based on the

results above is depicted in Fig.13and related explanations

are described below

A 1st step: Selenization

Some previous studies have reported that a reaction

preference between In-Se and Ga-Se is likely to be the main

cause for formation of Ga-poor CIGS with residual Cu-Ga

intermetallics at the Mo interface In the present study, the

same phenomena have been observed This step is believed

to be the most crucial aspect of the three-step reaction as the

extent of the selenization directly affects the Ga

homogeni-zation, recrystallihomogeni-zation, sulfur incorporation, and adhesion

in the following steps The aims of this step are to form a

fine crystalline structure with remaining Cu and Ga not

reacted with Se as a means to increase surface/interface

energy and enable Ga homogenization in the subsequent

pro-cess This can be understood more clearly with respect to the

subsequent steps

B 2nd step: Ar anneal Recrystallization with Ga homogenization is the most prominent effect of the 2nd step as residual Cu-Ga interme-tallics are incorporated into the CIGS bulk While the recrys-tallization and Ga homogenization occur concurrently, it is unclear whether they have a direct connection We previ-ously suggested that the residual Cu9Ga4 intermetallics on the Mo back contact appear to provide a rapid diffusion route

FIG 10 Light J-V curves of devices with the reacted CIGS or CIGSS films

after (a) 1st step, (b) 2nd step, and (c) 3rd step.

TABLE IV Summary of device parameters for CIGSS films with each

reac-tion step.

Film (absorber)

condition V oc (V) J sc (mA/cm2) Fill Factor (%) Efficiency (%)

FIG 11 (a) Quantum efficiency of fully processed CIGSS solar cell The estimated bandgap from the quantum efficiency is 1.09 eV The ratio of QE(1 V)/QE(0 V) (b) is close to 1 indicating little voltage dependent collection.

FIG 12 Compositional depth profiles determined by AES of the reacted CIGS (or CIGSS) films after (a) 1st step, (b) 2nd step, and (c) 3rd step The

Ga profiles (blue) are shown on the magnified scale of the right axis Ga ho-mogenization occurred after the 2nd step and S (violet) incorporation is lim-ited to the CIGSS film surface after the 3rd step.

for Ga ions since they are present as a mixture of liquid and

solid phase above 485

C.14

It is noteworthy that the recrystallization with the Ga

ho-mogenization strongly correlates with the extent of

seleniza-tion For instance, a film selenized at 450

C for 60 min in the 1st step (i.e., a more fully selenized film with a larger

grain size and little residual Cu9Ga4phase remaining10,12–14)

did not yield the recrystallization with Ga homogenization in

the 2nd step

It suggests that the extent of selenization should be

con-trolled to maintain sufficient Cu9Ga4and a fine

microstruc-ture which should result in an enough high interface and/or

surface energy to induce the recrystallization during the 2nd

Ar anneal step The sensitivity of Ga homogenization and

adhesion to the reaction conditions has been discussed in our

previous studies.13,14

Formation of the InSe binary may be attributed to a

reaction of the residual Cu9Ga4intermetallic with the CIGS

film As Cu and Ga diffuse, they react with Se ions from

pre-existed CIGS at the backside of the film InSe is formed as

the CIGS decomposes since the In-Se bond is weaker than

the Ga-Se bond.37The quantity of InSe also depends on the

amount of Cu9Ga4and therefore on the degree of

seleniza-tion, and it was shown previously that more InSe could be

obtained with shorter selenization time at 400

C.15 Mean-while, it is not confirmed whether Cu-Se binary compounds

are formed as might be expected Even if present in the films,

their amount would be small and they were not detected by

XRD or Raman spectroscopy

C 3rd step: Sulfization The main features of this step are sulfur incorporation near the surface and removal of InSe in the film bulk The AES analysis reveals that S is only incorporated near the sur-face Nevertheless, the InSe binary near the Mo contact is completely removed by the sulfization and MoSe2is formed

on the Mo surface Thus, it appears that sulfization enhances

Se indiffusion Accordingly, the degree of S incorporation depends on the amount of InSe in the film and is again con-trolled by the extent of selenization in the 1st step More InSe causes greater Se induffusion and hence more available sites for S incorporation near the surface It has been shown that too much S incorporation causes poor device perform-ance so the S incorporation needs to be controlled.12,35

V CONCLUSIONS

In this work, we have described a three-step H2Se/Ar/

H2S reaction of Cu-Ga-In metal precursors to form CIGSS thin films and characterized material properties of the films and solar cells fabricated from them after each step It was shown that this process enables good control of critical film properties including through-film compositional distribution and formation of large grain structure In addition, greater understanding has been gained of mechanisms that cause void formation and poor adhesion at the Mo/Cu(InGa) (SeS)2interface which commonly occur in films formed by precursor reaction processes

Ga accumulation at the back of the reacted film, near the

Mo back contact, is caused by the relative stability of the

Cu9Ga4intermetallic phase A critical aspect of the process is that the H2Se reaction time and/or temperature is restricted, so that a two phase film with Cu9Ga4and CuInSe2is formed af-ter the first step Then annealing at 550

C in Ar provides suf-ficient energy to drive interdiffusion of Ga and In and recrystallization to form films with grain size >1 lm The interdiffusion does not require reaction with S, as previously proposed Voids at the back of the fully reacted CIGSS film have comparable size and density as Cu9Ga4nodules remain-ing after the first reaction step suggestremain-ing that the voids are caused by the agglomeration of the intermetallic phase Dur-ing the Ar anneal step the film has insufficient Se to fully form Cu(InGa)Se2and consequently an InSe phase is formed The primary role of the third-step H2S reaction is to complete the reaction and eliminated the InSe The formation of MoSe2

which may cause poor adhesion only occurs after the film has sufficient chalcogen, Se þ S, to fully form CIGSS In the three-step process, this occurs only during the H2S reaction Thus controlling the time of this step can be used to minimize the MoSe2 formation Good device performance with high

VOCconsistent with the Ga intermixing are achieved once all secondary phases are eliminated in the final reaction step

ACKNOWLEDGMENTS The authors gratefully acknowledge the technical assis-tance of D Ryan and K Hart The authors are also grateful

to R W Birkmire and B E McCandless for valuable techni-cal discussions

FIG 13 Growth model for the three-step reaction (a) 1st step: due to

pref-erential reaction of In-Se, the surface region contains CIGS with little Ga

and the back side has higher Ga content with the Cu 9 Ga 4 intermetallic and

finer microstructure; (b) 2nd step: recrystallization and Ga homogenization

occur to reduce surface energy as the Cu 9 Ga 4 is incorporated into the CIGS

and InSe is formed; (c) 3rd step: S incorporation occurs near the film surface

and stimulates Se in-diffusion toward the Mo back contact leading to

con-sumption of InSe and formation of MoSe 2

1 W Shafarman, S Siebentritt, and L Stolt, in Handbook of Photovoltaic

Science and Engineering, 2nd ed., edited by A Luque and S Hegedus

(John Wiley & Sons, Ltd., 2011), pp 546–599.

2

K Kushiya, Y Tanaka, H Hakuma, Y Goushi, S Kijima, T Aramoto,

and Y Fujiwara, Thin Solid Films 517, 2108 (2009).

3 J Palm, V Probst, W Stetter, R Toelle, S Visbeck, H Calwer, T Niesen,

H Vogt, O Herna´ndez, M Wendl, and F H Karg, Thin Solid Films

451–452, 544 (2004).

4 T Nakada, R Onishi, and A Kunioka, Sol Energy Mater Sol Cells 35,

209 (1994).

5

M Marudachalam, R W Birkmire, H Hichri, J M Schultz, A

Swartz-lander, and M M Al-Jassim, J Appl Phys 82, 2896 (1997).

6 K Kushiya, Sol Energy Mater Sol Cells 93, 1037 (2009).

7 A Gupta and S Isomura, Sol Energy Mater Sol Cells 53, 385 (1998).

8

T Wada, N Kohara, T Negami, and M Nishitani, Jpn J Appl Phys 35,

L1353 (1996).

9 D Abou-Ras, G Kostorz, D Bremaud, M Ka¨lin, F V Kurdesau, A N.

Tiwari, and M Do¨beli, Thin Solid Films 480–481, 433 (2005).

10

G M Hanket, W N Shafarman, B E McCandless, and R W Birkmire,

J Appl Phys 102, 074922 (2007).

11 Y Nagoya, K Kushiya, M Tachiyuki, and O Yamase, Sol Energy Mater.

Sol Cells 67, 247 (2001).

12

V Alberts, J Titus, and R W Birkmire, Thin Solid Films 451–452, 207

(2004).

13 G M Hanket, W N Shafarman, and R W Birkmire, in Proceeding of

4th WCPEC (Hawaii, USA, 2006), pp 560–563.

14

G M Hanket, R Kamada, W K Kim, and W N Shafarman, in

Proceed-ings of 33rd IEEE PVSC (San Diego, CA, 2008).

15 W K Kim, G M Hanket, and W N Shafarman, in Proceeding of 34th

IEEE PVSC (Philadelphia, PA, 2009), pp 844–847.

16

S Verma, T W F Russell, and R W Birkmire, in Proceeding of 23rd

IEEE PVSC (Louisville, KY, 1993), pp 431–436.

17 H S Liu, X J Liu, Y Cui, C P Wang, I Ohnuma, R Kainuma, Z P.

Jin, and K Ishida, J Phase Equilibria 23, 409 (2002).

18

In, JCPDS card No 05-0642, JCPDS International Centre for Diffraction

Data, Swarthmore, Pennsylvania, USA.

19 Cu 3 Ga, JCPDS card No 44-1118, JCPDS International Centre for Diffrac-tion Data, Swarthmore, Pennsylvania, USA.

20

Cu 4 In, JCPDS card No 42-1477, JCPDS International Centre for Diffrac-tion Data, Swarthmore, Pennsylvania, USA.

21 Cu 9 In 4 , JCPDS card No 42-1476, JCPDS International Centre for Diffrac-tion Data, Swarthmore, Pennsylvania, USA.

22

Cu 9 Ga 4 , JCPDS card No 02-1253, JCPDS International Centre for Dif-fraction Data, Swarthmore, Pennsylvania, USA.

23 R Kamada, W N Shafarman, and R W Birkmire, Sol Energy Mater Sol Cells 94, 451 (2010).

24

B E McCandless, S S Hegedus, R W Birkmire, and D Cunningham, Thin Solid Films 431–432, 249 (2003).

25 D K Suri, K C Nagpal, and C K Chadha, J Appl Cryst 22, 578 (1989).

26 Cu(In 07 Ga 0.3 )Se 2 , JCPDS card No 35-1102, JCPDS International Centre for Diffraction Data, Swarthmore, Pennsylvania, USA.

27

CuGaSe 2 , JCPDS card No 35-1100, JCPDS International Centre for Dif-fraction Data, Swarthmore, Pennsylvania, USA.

28 CuGa 3 Se 5 , JCPDS card No 51-1223, JCPDS International Centre for Dif-fraction Data, Swarthmore, Pennsylvania, USA.

29

MoSe 2 , JCPDS card No 20-0914, JCPDS International Centre for Diffrac-tion Data, Swarthmore, Pennsylvania, USA.

30 MoSe 2 , JCPDS card No 20-0757, JCPDS International Centre for Diffrac-tion Data, Swarthmore, Pennsylvania, USA.

31

InSe, JCPDS card No 42-0919, JCPDS International Centre for Diffrac-tion Data, Swarthmore, Pennsylvania, USA.

32 S Lany and A Zunger, J Appl Phys 100, 113725 (2006).

33

S Merdes, R Kaigawa, J Klaer, R Klenk, R Mainz, A Meeder, N Papa-thanasiou, D Abou-Ras, S Schmidt, in Proceeding of 23rd EU PVSEC (Valencia, Spain, 2008), pp 2588–2591.

34 Y Nagoya, K Kushiya, M Tachiyuki, and O Yamase, Sol Energy Mater Sol Cells 67, 247 (2001).

35

U P Singh, W N Shafarman, and R W Birkmire, Sol Energy Mater Sol Cells 90, 623 (2006).

36 S S Hegedus and W N Shafarman, Prog Photovolt: Res Appl 12, 155 (2004).

37

Y Seki, H Watanabe, and J Matsui, J Appl Phys 49, 822 (1978).