Temperaturedependent conductivity of polycrystalline Cu2ZnSnS4 thin films

V. Kosyak ,1 M. A. Karmarkar,1 and M. A. Scarpulla,1,2[a])

1Department ofMaterials Science and Engineering, University of Utah, Salt Lake City, Utah USA, UT 84112-9206, USA

2Department of Electrical and Computer Engineering, University of Utah, Salt Lake City, Utah USA, UT 84112-9206, USA

These figures and discussion are provided in support of the main data presented in the manuscript.

Two types of substrates: one from Saint Gobain consisting of 3 mm soda-lime glass (SLG) and 275 nm of Mo and the 2nd being 1 mm SLG with 750 nm Mo direct current (DC)sputtered in-house were used for the composition series and annealing time series set respectively. Preliminary substrate cleaning was done by rinsing in solvents and ultraviolet (UV)ozone cleaning and finally plasma cleaning in the sputtering chamber for lower back contact resistance.[1] The load-locked sputtering chamber was equipped with liquid nitrogen cooled cold-finger and pumped by a turbo molecular pump backed by a roughing pump and has base pressure 2x10-7 Torr or lower. Deposition pressure of 2.5 mTorrargon was used for all films with radio-frequency (RF) deposition powers close to 110 W for Cu, 145 W for ZnS, and 48 W for Sn targets (Plasmaterials Livermore, CA 99.5% purity). The target–substrate distance was 16.5 cm, and the sample was rotated during deposition with no substrate heating. Annealing of the films was done in a small volume, closed box, staticgas tube furnace as described in the paper text. Sample surface morphology, grain size, andcomposition were analyzed in a FEI Quanta 600FEG scanning electron microscope (SEM) equipped with energy dispersive X-ray analysis (EDX). The grain sizes listed in Table Iwere determined by averaging images taken at multiple locations using digital image analysis software. Cross-sectional images at cleaved edges of the films showed equiaxed grains throughout the cross-section for both sample series. All samples in composition series (Fig. S1)exhibit SnS2 surface precipitates but samples 2C and 4C had higher density to be detected in Raman. In figure labeled2Cand 4C these precipitates are encircled to show their position in center and right edge of the micrograph respectively

Figure S1 - SEM images of the composition series samples all annealed for 30 min (clockwise from upper left): 1C {[Cu]/([Zn]+[Sn])=0.9, [Zn]/[Sn]=1.3}, 2C {[Cu]/([Zn]+[Sn])=0.74, [Zn]/[Sn]=1.4}, 3C {[Cu]/([Zn]+[Sn])= 0.73, [Zn]/[Sn]=1.8}, 4C ([Cu]/([Zn]+[Sn])= 0.61, [Zn]/[Sn]=0.71), and 5C ([Cu]/([Zn]+[Sn])= 0.63, [Zn]/[Sn]=1.9). Note the changes of scale amongst the images. Sn phases are encircled in the 2C and 4C samples.

These SnS2 surface precipitates are highly faceted portions of hexagonal plates which appear to grow outwards from the CZTS film surface. Additionally at cleaved edges they are seen not to penetrate into the films suggesting a vapor-phase growth mechanism. Their absence in similarly large densities on the temperature series samples (especially 30T with identical annealing conditions) discounts hypotheses in which they form by condensation/reaction of excess Sn or SnS vapors during cooling in the graphite box. Thus we propose that they form by vapor phase epitaxy from the reaction of SnS vapors emerging from the CZTS film and S from the annealing atmosphere. The areal density and thickness of individual SnS2 platelets are greatest for the composition series samples because of their rather exaggerated Cu-poorness (=0.61-0.74); we assume that some ZnS is probably also present in the films as required from stoichiometry considerations perhaps near the Mo contact but it was not observed in Raman characterization probably because of the probe depth. The standard 1 min KCN etch thins and removes many of these platelets but in the two noted samples they still remain in sufficient density that areas devoid of them larger than the 2 m Raman probe area could not be located.

For all Cu-poor samples, some mixture of ZnS and SnxSy phases have the potential to form assuming all Cu incorporates into CZTS. While it is possible that only Cu-poor CZTS would exist in full thermal equilibrium, it is possible that these samples are not in full equilibrium and that near-stoichiometric CZTS is formed along with other phases. Although sample 2C is Zn-rich, [Cu] limits the amount of stoichiometric CZTS that can form and thus both SnSx and ZnS may be formed along with CZTS. Although somewhat counterintuitive based on cursory examination of the  and  values, limiting reactant calculations may be critical for understanding the appearance of 2nd phases in/on CZTS thin films. All of the annealing time series samples show higher grain size than the composition series due to larger . The annealing time series samples are close to the compositions optimal for solar cells and in this set (Figure S2) we do not find any Sn phases on the surface. Higher [Cu] which makes Zn the limiting reactant can be the reason why such phases are not seen in the annealing time series.

Figure S2 - SEM images of the annealing time series samples annealed at 10, 30, 75 and 120 min respectively (clockwise from upper left). All samples have [Cu]/([Zn]+[Sn])= 0.9 and [Zn]/[Sn]=1.4 as measured by EDS after annealing.

Current-voltage (I-V) measurements were done on etched samples with sputtered gold contacts. The linearity of I-V curves up to 0.1 V (indicating Ohmic contact) was verified for every contact on each sample at multiple temperatures; example data is shown in Fig. S3.

Figure S3.Example Ohmic I-V behaviour of one of the contacts on sample 4C from data taken at 298 K.

Conductivity value was extracted from these I-V dependencies. A comparison of fitting the measured data with two exponential function representing Mott-variable range hopping (M-VRH) and thermionic emission (TE) versus three exponential functions that also includes nearest neighbor hoping (NNH) is shown in Figure S4.

Figure S4. Example showing that fitting the conductivity data from sample 5C using a model including only Mott variable range hopping (M-VRH) and thermionic emission (TE) is inadequate in the intermediate temperature regime (approximately 50-150 K)

1

[a])Author to whom correspondence should be addressed. Electronic mail:

[1]J. Johnson, H. Nukala, E.A. Lund, W.M. HlaingOo, A. Bhatia, L.W. Rieth and M.A. Scarpulla, in Proceedings of the Material Research Society Symposium, SanFrancisco, USA (MRS, 2010),pp. EE03-03.