CdS thin film post-annealing and Te–S interdiffusion in a CdTe/CdS solar cell

Available online at www.sciencedirect.com

www.elsevier.com/locate/solener

Solar Energy 86 (2012) 1023–1028

CdS thin film post-annealing and Te–S interdiffusionin a CdTe/CdS solar cell

Xavier Mathew a,⇑, Jose S. Cruz b, David R. Coronado a, Aduljay R. Millan a,Gildardo C. Segura a, Erik R. Morales a,c, Omar S. Martınez a,c, Christian C. Garcia a,

Eduardo P. Landa a

a Centro de Investigacion en Energıa, Universidad Nacional Autonoma de Mexico, 62580 Temixco, Morelos, Mexicob Universidad Autonoma de Queretaro, Facultad de Quımica-Materiales, Queretaro 76010, Mexico

c Centro de Investigacion en Materiales Avanzados S.C. Miguel de Cervantes 120, Chihuahua, Chih, CP 31109, Mexico

Available online 29 July 2011

Abstract

Small area CdTe/CdS solar cells were fabricated using chemical bath deposited CdS and CSS deposited CdTe thin films to investigatethe interface properties related to the CdS processing. The effect of post deposition annealing of CdS on the junction properties and thepossible interdiffusion at the interface is discussed. The hypothesis of hardening of CdS due to annealing against the diffusion of “S” and“Te” is discussed using the quantum efficiency data in the blue and red regions. Devices prepared using as-deposited CdS films exhibitedevidences of higher “S” and “Te” diffusion compared to devices made using CdS films annealed in oxygen. The maximum efficiency of thedevices used in this study was 9.8%.� 2011 Elsevier Ltd. All rights reserved.

Keywords: CdTe/CdS; CSS; CdS annealing; Vapor chloride treatments; Quantum efficiency

1. Introduction

Increased awareness of the limitations of exploitable oilreserves, and the search for green energy solutions promptedan intense search for alternate energy pathways whichresulted in great interest for photovoltaic technology inrecent years. Today, in thin film PV market the leadingand proven technologies are CdTe and a-Si, with CIGSand dye-sensitized (DSSC) solar cells are entering into themarket. Though the laboratory level efficiency of CdTe solarcells is lower than that of the CIGS champion cells, CdTeposses several advantages which made it the frontrunner inthin-film solar cells. CdTe is a robust compound material,which can be prepared by several techniques in a highlyreproducible manner, and is tolerable to high processingtemperatures, making it ideal for large scale industrial

0038-092X/$ - see front matter � 2011 Elsevier Ltd. All rights reserved.

doi:10.1016/j.solener.2011.06.024

⇑ Corresponding author.E-mail address: [email protected] (X. Mathew).

production. Among the various CdTe thin-film depositiontechniques close spaced sublimation (CSS) or its modifiedversion known as close spaced vapor transport (CSVT) isthe most convenient method for growing uniform films onlarge area substrates for superstrate devices. Typicalsubstrate temperature in CSS process is in the range450–600 �C, and large grain films can be deposited at a rateof 1–5 lm/min depending on the temperature gradientbetween source and substrate. The CSS process appliedto the fabrication of CdTe thin films and devices arereviewed in the literature (Xuanzhi, 2004; Britt andFerekides, 1993; Moutinho et al., 2000; Omura and Muroz-ono, 1997; Rose et al., 1999).

To date the well-known and most successful hetero-junction partner to CdTe is CdS, and the highest efficiencyfor a CdTe/CdS device was obtained with chemical bathdeposited (CBD) CdS thin films (Wu et al., 2001). Theunique advantage of using CBD technique for depositingCdS is that very thin films with thickness in the range

Table 1Chemical composition of CBD bath used to deposit CdS thin films.

Chemicals Concentration Exact amount for450 ml bath

Cadmium acetate 0.12 M 57.5 mgAmmonium acetate 0.74 M 334.2 mgThiourea 0.074 M 33.4 mgAmmonium hydroxide 28–30% 12.3 ml

1024 X. Mathew et al. / Solar Energy 86 (2012) 1023–1028

60–80 nm can be deposited with uniform surface coverageand minimum pin-holes. In addition, the absorptioncoefficient of CBD CdS is lower than that of the CdS filmsprepared by other techniques having the same thickness. Itis believed that the unintentional incorporation of impuri-ties from the bath contributes to the thickness of the filmsand the actual amount of absorbing CdS is less (Albinet al., 2002), which is an advantage of CBD over otherphysical methods, however, the inherent slow growth rateof CBD and a solution based process may not be as attrac-tive as physical methods such as CSS or sputtering in a pro-duction line. In CdTe/CdS solar cells, the most crucial stepin the fabrication process is the post-deposition cadmiumchloride annealing known as the junction activation pro-cess. The major effects of this annealing process in presenceof Cl species are to promote recrystallization leading tograin growth, grain boundary passivation, doping of theCdTe, decrease in resistivity of CdTe film, and alloying atthe metallurgical interface forming CdTe1�xSx andCdTeyS1�y. The principal effect of this alloying is to lowerthe lattice mis-match between CdTe and CdS and reducethe recombination centers (McCandless et al., 1999;Xuanzhi, 2004).

In the case of both CdTe and CdS oxygen is known toplay a role in its morphology (Flores Mendoza et al.,2011), electrical properties (Rios-Flores et al., 2010) andhardening against diffusion. In CdTe the effects of oxygenis manifested in grain size, enhanced p-doping, and passiv-ation of the grain boundaries. In CdS, the prominent effectis to make it resistant to in-diffusion of Te and out-diffu-sion of S (Albin et al., 2002). In this paper we are discussingsome of the experimental results obtained on the postdeposition annealing of CdS and its effects on the CdTe/CdS device performance; the quantum efficiency (QE) dataof the devices were analyzed, and the features in the shortand long wavelength regions are discussed in the light ofconcepts in the literature regarding the interdiffusion atthe CdTe–CdS interface. Even though CdTe and CdS arestudied for decades, and many properties of CdTe/CdSinterface are reported in the literature, a specific discussionfocused on the effect of post-deposition annealing of CdSand its effect on CdTe/CdS interface is not well-docu-mented in literature. This paper will update the knowledgeabout the role of post-deposition thermal treatment of CdSin controlling the properties of CdTe/CdS interface.

2. Experimental

2.1. Thin films and devices

The CdS films were deposited using CBD technique at90 �C from a bath containing cadmium acetate (Cd(OOCCH3)2.2H2O), ammonium acetate (CH3CO2NH4), thio-urea (H2NCSNH2), and 28–30% ammonium hydroxide.The transparent conducting oxide (TCO) glasses used assubstrates in this study were Tec 7 glasses obtained fromPilkington. TCO substrates were cleaned following our

standard procedure and inserted in a vertical position intothe CBD container filled with appropriate amount ofde-ionized water and allowed to stabilize the temperaturefor 30 min. After stabilizing the temperature cadmium ace-tate, ammonium acetate, and NH4OH were added to thede-ionized water. In order to minimize the homogeneousprecipitation, the thiourea was added in small quantitiesat regular intervals. The exact amounts of the chemicalsand the molarities are given in Table 1. Approximate thick-nesses of the as-deposited films were in the range of 100–110 nm. The films were annealed in oxygen environmentat 450 �C for 45 min or at 550 �C for 5 min, in some casesthe films were etched using very dilute HCl. The CdTe filmswere deposited on top of the CdS by CSS technique in anatmosphere of oxygen and argon, and the total pressure inthe CSS chamber was 15 mbar. The source and substratetemperatures were 600 and 550 �C respectively, and thedeposition rate was higher than 1 lm/min. Thickness ofthe CdTe films used in this study was 4.5–5 lm. The hetero-structures were vapor chloride treated in a separate CSSchamber using pure CdCl2 as the chloride source. Thevapor chloride treatment parameters were: substrate tem-perature = 390 �C, source temperature = 395 �C, and timeduration = 20 min. The total pressure in the treatmentchamber was 75 mbar, which contains 50% oxygen and50% argon. The devices were completed by sequentiallydepositing 3 nm of Cu and 30 nm of Au in a high vacuumchamber, and the devices were Cu diffused at 150 �C for10 min.

2.2. Characterization techniques

The transmittance measurements were performed on aShimadzu UV-3101PC UV–Vis spectrophotometer usinga substrate glass in the reference beam. The XRD datawas collected at 0.5� angle of incidence on a RigakuDMAX 2200 X-ray diffractometer with Cu Ka radiationand the data analysis was performed using JADE� V6.5.The surface and cross-sectional SEM images were takenusing a Hitachi Field Emission SEM S5500. The I–V

measurements were performed in a home assembled I–V

system including Oriel solar simulator, and a Keithley236 SMU; the current and voltage data were collectedusing an I–V test program which can give all the deviceparameters once the measurement run is completed. Thesolar simulator was calibrated with a Dexter thermopileusing the calibration data provided by the manufacturer(Dexter Research Center. Inc., Michigan). The QE system

X. Mathew et al. / Solar Energy 86 (2012) 1023–1028 1025

consisted of a tungsten halogen lamp, oriel monochroma-tor, optical chopper, and a lock-in amplifier (modelSRS). The data was collected at 5 nm interval in the wave-length region 350–1000 nm. The QE system was calibratedusing a Si photodiode standard calibrated at NREL.

3. Results and discussion

3.1. CdTe and CdS thin films

Thin films of CdS and CdTe were developed asexplained in Section 2. The thickness of the as-depositedCdS films used in this study was in the range of 100 nm,and that of the CdTe film was about 4.5 lm. Fig. 1 showsthe transmittance spectra of the as-deposited and annealedCdS films in the wavelength range 350–1000 nm. A bandgap value of 2.39 eV was calculated for the as-depositedfilm from the transmittance data, the position of the bandgap in the wavelength scale is shown with dotted lines. Inthis case the film was annealed at 550 �C for 5 min in pres-ence of pure oxygen in a furnace. The purpose of selectingthis annealing temperature was to investigate the transmit-tance of the CdS films which were exposed to the samedeposition conditions exist in the CSS chamber of CdTe.In our CSS process for CdTe, the deposition temperaturewas 550 �C, and we used a mixture of pure argon and

400 600 800 10000

30

60

90

annealed

T (%

)

λ (nm)

as-deposited

Eg = 2.39 eV

Fig. 1. Line of sight transmittance spectra of the as-deposited andannealed CdS films. The annealing was at 550 �C in presence of oxygen for5 min. The dotted line corresponds to the band gap position in eV.

Fig. 2. SEM images of the CdS films presented in Fig. 1; (

oxygen. Generally the ramp up period of the CSS furnacewas 5–7 min to reach 550 �C for the substrate and 600 �Cfor the CdTe source. During this period, the CdS filmundergoes an un-intentional annealing in oxygen at ele-vated temperatures inside the CSS chamber.

From Fig. 1 it can be observed that the transmittance ofthe as-deposited and annealed films maintains more or lessthe same characteristics except that in the case of annealedfilm the absorption edge corresponding to the band gap ofthe film is sharper indicating significant recrystallization.Also the transmittance is slightly higher compared to theas-deposited film which may be due to the thinning of thefilm. The SEM images of the above films are presented inFig. 2, signs of recrystallization such as coalescence andcluster formation are evident, and however, no significantgrain growth is observed. Apart from the physical proper-ties discussed here, the oxygen in the annealing environ-ment plays an important role in controlling theinterdiffusion of S and Te, the effect of this interdiffusioncan be observed in the QE data discussed in Section 3.2.

The high resolution SEM images showing the surfacemorphologies of the CSS deposited CdTe films and cross-sections of the Tec 7/CdS/CdTe heterostructures are pre-sented in Fig. 3. The as-deposited film (Fig. 3a) shows avery compact and rough surface without voids. Few pin-holes at the grain boundaries are visible but it cannot causeshunting between front and back electrodes since the filmthickness is much larger than the average grain size. Theaverage grain size is in the order of 0.9 lm. We haveobtained similar values of grain size for our films depositedon metallic substrates, but with different morphologicalfeatures (Hernandez et al., 2004). Fig. 3b is the SEM imageof the same film vapor chloride treated at 390 �C for20 min. The chloride treated film surface shows morphol-ogy distinct from that of as-deposited film; small graincoalescence and effective sealing of the grain boundarypin-holes can be observed. The average grain size is largerthan 1.5 lm. The cross-sections of the above films areshown in Fig. 3c and d, the SEM images (c and d) corre-sponds to the as-deposited and vapor chloride treated het-erostructure. The regions correspond to the TCO, CdTe/CdS/TCO interface, and CdTe are clearly marked in theimage. The XRD data of the films show only negligibleeffects for the annealing on the structural parameters

a) as-deposited, and (b) annealed at 550 �C in oxygen.

G2k9-12, as-depo

CdTe

CdTe/CdS/TCOInterfacial region

TCO

(c) G2k9-12, CL(d)

(b)(a) G2k9-12, as-depo G2k9-12, CL

Fig. 3. FESEM images of the CdTe surface deposited on top of CdS/Tec 7 substrates at 550 �C; (a) surface image of as-deposited film, (b) surface image ofvapor chloride treated film, (c) cross-section of the Tec 7/CdS/CdTe as-deposited, and (d) cross-section of the Tec 7/CdS/CdTe vapor chloride treated.

20 30 40 50 60 700

750

1500

2250 20 30 40 50 60 700

750

1500

2250

23.2 23.6 24.0 24.40

1000

2000

(b)

G2k9-12, CL

Cou

nts

(a.u

)

2 theta ( o )

G2k9-12, as-deposited

(a)annealed

Cou

nts

(a.u

)

2 theta (o)

as-deposited

Fig. 4. XRD patterns of the films presented in Fig. 3 above: (a) as-deposited film, and (b) vapor chloride treated. The inset is the (1 1 1)reflection.

1026 X. Mathew et al. / Solar Energy 86 (2012) 1023–1028

(Fig. 4). This can be due to the fact that the as-depositedfilm was deposited at higher temperatures (550 �C) leadingto stress-free film growth. An enlarged view of the (1 1 1)reflection is shown in the inset of Fig. 4, it can be seen thatthe shift in the peak position is negligible (23.643� and23.65�). The average size of the crystallites was estimatedusing the Scherer relation and the obtained values in bothcases are in the same range, 26.4 and 26.7 nm for as-depos-ited and vapor chloride treated films respectively. It shouldbe noted that the grains observed in the SEM image are notsingle crystals, and it explains the difference between thevalues estimated from SEM and XRD.

3.2. Post-deposition annealing and device properties

The heterojunction CdTe/CdS was annealed in an ambi-ent of CdCl2 vapor and oxygen at a pressure of 75 mbar for

20 min at a temperature of 390 �C. This vapor chlorideannealing is expected to provoke alloying at the interfacedue to the diffusion of S towards the CdTe and Te towardsCdS, and in addition promotes the recrystallization as dis-cussed in Section 3.1. To certain extent this alloying isfavorable for CdTe/CdS solar cells in reducing the latticemis-match and hence the interfacial recombination centers,however, excessive alloying can cause decrease in the bandgap of CdS leading to reduction in current and thinning ofCdS layer which can create regions of CdTe/TCO junction(McCandless et al., 1999; Albin et al., 2002; McCandlessand Sites, 2003). In device fabrication steps, it is commonthat the CdS film is annealed prior to the deposition ofCdTe in selected ambient and one of the purposes of thisannealing is to make the film resistant to excess interdiffu-sion (Albin et al., 2002). In the case of a heterostructurefabricated with as-deposited CBD CdS film, the extent ofinterdiffusion can be large depending on the CdTe deposi-tion temperature. In order to investigate the alloying at theCdTe–CdS interface and discuss the role of oxygen in hard-ening CdS against diffusion, we have developed two sets ofCdTe/CdS devices using two types of CdS plates. In CdSplate I, the CdS film was as-deposited and directly loadedinto the CSS chamber and deposited CdTe, in the case ofCdS plate II, the CdS film was annealed at 550 �C for5 min in oxygen, and afterwards deposited CdTe by CSS.The CSS process was identical in both cases, and both het-erostructures were vapor chloride treated at identical con-ditions. The QE of the devices are presented in Fig. 5. Itcan be observed that in general the QE is high for devicemade with annealed CdS in all wavelengths; this can bedue to the thinning of the CdS film during the high temper-ature annealing. However, the interesting point is thefingerprint regions where the effect of alloying can beobserved; the blue and red response regions.

The predominant effects of alloying in the blue and redresponses of a CdTe/CdS device can be explained as

400 500 600 700 800 9000

20

40

60

80

800 850 9000

20

40

60

CdS annealedin oxygen

QE

(%)

λ (nm)

as-depositedannealed

CdS, as-deposited

Fig. 5. QE of CdTe/CdS devices demonstrating the effect of CdSannealing prior to the deposition of CdTe, inset is the long wavelengthresponse of the devices. Solid markers corresponds to the QE data ofdevice with annealed CdS, and the hollow markers represent the data ofdevice with as-deposited CdS film. The two arrows in the QE curvesindicate the wavelength region where the light absorption in the windowlayer begins.

0.0 0.2 0.4 0.6

-20

-10

0

10

20

Eff = 6.6%CdS post-annealed

Eff = 5.4%CdS as-depositedJ

(mA/

cm2 )

v (V)

Fig. 6. J–V characteristics of two typical CdTe/CdS solar cells. One ofdevices was fabricated using as-deposited CdS film, and the other withCdS film post-annealed in oxygen.

-0.4 -0.2 0.0 0.2 0.4 0.6 0.8-30

-20

-10

0

10

20

J (m

A/c

m2 )

v (V)

Jsc = 25.12 mA/cm2

Voc = 694 mVFF= 0.56Eff = 9.8%

Fig. 7. J–V characteristics of a CdTe/CdS device in which the CdS filmwas first post-annealed in oxygen and then etched in dilute acid prior tothe deposition CdTe by CSS.

X. Mathew et al. / Solar Energy 86 (2012) 1023–1028 1027

follows; in case of significant diffusion of Te into the CdSfilm, it can form the alloy CdS1�yTey affecting thetransmission properties of CdS (McCandless and Sites,2003). In this case the blue response of the device will berelatively low. Similarly in the event of significant Sdiffusion into the CdTe, there can be the formation of thealloy CdTe1�xSx (Dhere, 1997; Fischer, 1996; McCandlessand Sites, 2003), the effect is to lower the band gap of CdTewhich can be seen as a shift in the absorption edge in QEtowards higher wavelengths. In the present case both theseeffects are observed in the case of device fabricated with as-deposited CdS film, the shift in the long wavelengthresponse which corresponds to the band gap of CdTe isshown in the inset of Fig. 5 indicating the significant S dif-fusion into the CdTe forming CdTe1�xSx in the case of as-deposited CdS film (hollow markers). The Te diffusion intothe CdS and the formation of the alloy CdS1�yTey isevidenced from the low QE in the blue region for the as-deposited film. Further it can be seen from figure that theas-deposited CdS device has a shallow absorption edge(shown with arrows) indicating significant alloying. TheJ–V curves of two identical devices prepared from the samebatch of films is shown in Fig. 6. It is clear that the shortcircuit current (Jsc) is lower for the device fabricated withthe as-deposited CdS film which is in agreement with theQE data presented in Fig. 5. The post annealing of CdSin oxygen ambient is beneficial for enhancing the deviceefficiency and especially the open circuit voltage (Hanet al., 2011).

We have developed small area CdTe/CdS solar cellsusing a CdS film annealed in oxygen and later etched withvery dilute acid to remove any traces of oxides prior to dedeposition of CdTe. The device was tested under100 mW/cm2 illumination and the measured parametersof our best devices are: Voc = 694 mV, Jsc = 25.12 mA/cm2, FF = 56%, and efficiency = 9.8% (Fig. 7). The lower

Voc can be due to poor interface properties such as localshunts, and recombination at interfacial states. We are inthe process of updating our vapor chloride treatment sys-tem, and further improvement in efficiency is expectedafter optimizing the chloride treatments.

4. Conclusions

We have fabricated two sets of CdTe/CdS devices inwhich one set of devices was prepared by as-depositedCdS film, and the other set using CdS films annealed inoxygen. The CdS was prepared by CBD technique. Theinterfacial diffusion of Te and S was investigated usingthe QE data, and the evidences of the formation ofCdS1�yTey, and CdTe1�xSx alloys were observed. Thedevices fabricated with as-deposited CdS showed higherinter-diffusion at the interface leading to lower blueresponse and a small shift in the CdTe absorption edge.The above mentioned features are explained on the basisof the hypothesis that the CdS becomes resistant to inter-diffusion of Te and S at the interface after annealing in oxy-

1028 X. Mathew et al. / Solar Energy 86 (2012) 1023–1028

gen. The efficiency of the devices used in this study was inthe range of 9.8%.

Acknowledgements

This work at CIE-UNAM was partially supported bythe projects CONACyT-PROINOVA 139562, SENER-CONACyT 117891, CONACyT 60762, ICyTDF, andPAPIIT IN 118409. Authors acknowledge the technicalsupport of Maria Luisa Ramon Garcia in XRD measure-ments and Jose Campos for assistance in electricalcharacterizations.

References

Andreas, Fischer., 1996. Thesis: Photoluminescence and Raman Studies ofCdTe, CdSxTe1�x Thin Film Solar Cells. University of Toledo, Ohio,USA.

Albin, D.S. et al., 2002. Prog. Photovolt: Res. Appl. 10, 309.McCandless, Brian E., Sites, James R., 2003. Cadmium Telluride Solar

Cells. Handbook of Photovoltaic Science and Engineering. John Wiley& Sons, Ltd, ISBN: 0-47149196-9.

Britt, J., Ferekides, C., 1993. Appl. Phys. Lett. 62 (22), 2851–2852.Rose, Doug H., Hasoon, Falah S., Dhere, Ramesh G., Albin,

Dave S., Ribelin, Rosine M., Li, Xiaonan S., Mahathongdy,

Yoxa, Gessert, Tin A., Sheldon, Pete, . Prog. Photovolt.: Res.Appl. 7, 331–340.

Flores Mendoza, M.A., Castanedo Perez, R., Torres Delgado, G.,Marquez Marın, J., Cruz Orea, A., Zelaya Angel, O., 2011. Sol.Energy Mater. Sol. Cells 95, 2023–2027.

Hernandez, G.P., Mathew, X., Eriquez, J.P., Morales, B.E., Lira, M.M.,Toledo, J.A., Juarez, A.S., Campos, J., 2004. J. Mater. Sci. 39, 1515–1518.

Han, Junfeng, Liao, Cheng, Jiang, Tao, Fu, Ganhua, Krishnakumar, V.,Spanheimer, C., Haindl, G., Zhao, Kui, Klein, A., Jaegermann, W.,2011. Mater. Res. Bull. 46, 194–198.

McCandless, B.E., Youm, I., Birkmire, R.W., 1999. Prog. Photovolt.: Res.Appl. 7, 21–30.

Moutinho, H.R., Dhere, R.G., Al-Jassim, M.M., Ballif, C., Levi, D.H.,Swartzlander, A.B., Yong, M.R., Kazmerski, L.L., 2000. In: Proceed-ings of 20th IEEE PVSC, pp. 646–649.

Ramesh G. Dhere., 1997. Thesis: Study of the CdS/CdTe Interface and itsRelevance to Solar Cell Properties, Deparment of Electrical andComputer Engineering, University of Colorado, USA.

Rios-Flores, A., Pena, J.L., Castro-Pena, V., Ares, O., Castro-Rodrıguez,R., Bosio, A., 2010. Sol. Energy 84, 1020–1026.

Omura, K., Murozono, M., 1997. In: Proceedings of 26th IEEE PVSC, pp.343–346.

Wu, X., Dhere, R.G., Albin, D.S., Gessert, T.A., DeHart, C., Keane, J.C.,Duda, A., Coutts, T.J., Asher, S., Levi, D.H., Moutinho, H.R., Yan,Y., Moriarty, T., Johnston, S., Emery, K., Sheldon, P., 2001. NCPVProgram Review Meeting Lakewood, Colorado, pp. 14-17. <http://www.nrel.gov/docs/fy02osti/31025.pdf>.

Xuanzhi, Wu., 2004. Sol. Energy 77, 803–814.