Thin Solid Films, Volume 535, 15 May 2013, Pages 348–352

Electrical characterization of Cu2ZnSnSe4 solar cells from selenization of sputtered metal layers

Guy Brammertza,*, Yi Rena,b, Marie Buffièrea,b, Sofie Mertensa, Jurgen Hendrickxa, Hakim Markoa, Armin E. Zaghia,b, Nick Lenaersa,b, Christine Köblec, Jef Vleugelsb, Marc Meurisa, Jef Poortmansa

a

imec - partner of Solliance, Kapeldreef 75, 3001 Heverlee, Belgium.

b

Department of Metallurgy and Materials Engineering, K.U. Leuven, Kasteelpark Arenberg 44, 3001 Heverlee, Belgium. c

Helmholtz-Zentrum Berlin für Materialien und Energie GmbH, Hahn-Meitner-Platz 1, 14109 Berlin, Germany.

* Corresponding author: E-mail address: [email protected] Phone: +32 16 28 8120

Abstract

We report on the electrical and physical properties of Cu2ZnSnSe4 (CZTSe) solar cells consisting of an absorber layer fabricated by selenization of sputtered Cu, Zn, Sn multilayers. Cross-section scanning electron microscopy images show that the polycrystalline absorber layers are approximately 1 µm thick and that the typical grain size is of the order of one µm. Energy-dispersive X-ray spectroscopy measurements show Cu-poor and Zn-rich composition with Cu/(Zn+Sn) ~ 0.8 and Zn/Sn ~ 1.2. Solar cells are fabricated out of this absorber material using a standard process flow for chalcogenide solar cells. Under AM1.5G illumination, the best 1x1 cm2 CZTSe solar cell shows an efficiency of 6.3 % with a maximum short circuit current of 31.3 mA/cm2, an open circuit voltage of 0.39 V and a fill factor of 52%. Doping density of the absorber layers is derived using the Drivel Level Capacitance Profiling (DLCP) technique, showing low p-type doping density which seems to increase exponentially with the Zn/Sn ratio. Comparing the values obtained from DLCP to the ones derived from MottSchottky plots of the same devices, it is shown that for CZTSe care has to be taken when deriving the doping density. Similarly to copper indium gallium selenide junctions, Mott-Schottky plots overestimate the amount of free carriers in the buffer due to the presence of fast defect states inside the bandgap.

1. Introduction Cu2ZnSn(S,Se)4 (CZTSSe) materials are being developed for application as thin film absorber layers in solar cells [1-4]. 10.1% efficient solar cells have already been realized on the basis of this type of absorber layer [5], which consists essentially of relatively abundant elements. Further improvements in the layer and cell structure are necessary nevertheless, in order to achieve efficiencies comparable to Cu(In,Ga)(S,Se)2 (CIGSSe) based solar cells, which are nowadays already in production and have shown on research cell level efficiencies of the order of 20% [6]. CZTSSe layer fabrication can be performed using different production processes ranging from wet deposition using solutions and/or particle suspension methods [7-8] to co-evaporation of the constituent elements [9] and sulfurization/selenization of different types of precursor layers [10-11]. In the present study, we have fabricated CZTSe layers by sputtering of Cu, Zn and Sn metal layers, followed by annealing in an H2Se containing atmosphere. The fabricated layers were then processed into solar cells. We first present the physical properties of the fabricated cell structures, including compositional and structural analysis by scanning electron microscopy (SEM) energy dispersive X-ray spectroscopy (EDX), followed by electrical and optical characterization with illuminated and dark currentvoltage measurements and external quantum efficiency measurements. The doping density of the absorber layers is then derived from Mott-Schottky plots and from Drive Level Capacitance Profiling (DLCP). 2. Experimental details First, Cu, Zn and Sn are DC sputtered at room temperature onto 5x5 cm2 Mo on soda lime glass substrates in an Ar atmosphere with a sputtering rate of 1 nm per second. In the present study we show the properties of four different layer structures: Our basic reference layer consists of a 110 nm Cu / 150 nm Zn / 170 nm Sn multilayer deposited onto Mo/glass substrates, with Sn being the bottom layer of the multi-stack. This layer structure will be referenced as Sample A. A second layer consists of a similar layer structure, but this time with only 110 nm of Zn as the middle layer. This layer will be referenced as sample B. A third layer, sample C, consists of a similar layer structure as sample A, but with only 125 nm of Sn as the bottom layer. And finally, sample D, which consists of a more complex layer structure: 55 nm Cu / 115 nm Zn / 255 nm Sn / 110 nm Cu / 110 nm Zn. A summary of the layer structures and compositions before anneal can be found in table 1. The four different layers are then annealed in vacuum for 15 minutes at 450°C in a continuous flow of 20 sccm of pure H2Se directed onto the sample surface. At the end of the run, the total pressure in the 1liter volume isolated chamber is 30000 Pa, consisting mainly of H2 released by the decomposition of H2Se at high temperature, the Se being quickly absorbed by the sample and by the cold sidewalls of the reactor. The selenized samples are then sent to Helmholtz Zentrum Berlin for processing into solar cells. A standard procedure for CIGS based solar cells was used, consisting of

KCN etch, chemical bath deposition of a thin n-type CdS buffer layer and AC-sputtering of 120 nm of intrinsic ZnO followed by 250 nm of highly Al-doped ZnO [12-13]. Finally, a 50 nm Ni - 1 µm Al finger contact pattern is evaporated through a shadow mask for top contact formation. Lateral isolation of the 1x1 cm2 area cells is performed by needle scribing. The processed solar cells were then analyzed with light and dark current-voltage (IV) measurements using a Wacom 2 Lamp (Xenon+Halogen) Solar Simulator System with an AM1.5G spectrum with an illumination density of 1000 W/m2 and a Keithley 2600 Sourcemeter for curve tracing. SEM and EDX analysis of the cell cross sections was made with a Philipps XL-30F tool. The operating voltage of the field emission gun for the electron beam was 15 kV. The EDX measurements were performed at an accelerating voltage of 15 kV as well, using an acquisition area corresponding to a spot size of about 1 micron of diameter, focused onto a CZTSe grain on the same cross section images. The average composition of three different grains was determined for each sample. The external quantum efficiency (EQE) measurements are obtained at room temperature using a grating monochromator-based dual-beam setup (model Bentham 605) under chopped light from halogen and xenon lamps. Finally, admittance and DLCP measurements were performed with an HP 4286A LCR-meter and a Carl Zeiss PA300 probe station at room temperature.

Figure 1: Cross-section SEM images of CZTSe solar cells A (a), B (b), C (c) and D (d).

3. Physical properties Cross-section SEM images of the four samples are shown in figure 1. Samples A, B and D show relatively similar morphology with average grain sizes of the order of 1 µm or slightly less. The CZTSe absorber layer thickness is of the order of 1 µm in the case of samples A and B and of the order of 1.5 µm in the case of sample D, as expected from the 1.5 times thicker sputtered metal precursor layers of sample D as compared to sample A and B. What is relatively striking for samples A, B and D are the holes close to the Mo backside contact. It is not clear whether this is due to crystals that were broken out of the layer during cleaving of the glass substrate, or whether this is a real feature due to poor wetting of CZTSe crystals on Mo. No thick MoSe2 layer could be observed between the Mo and the CZTSe, likely due to the relatively modest annealing temperature of 450°C. The thickness of the MoSe2 layer is estimated to be of the order of 50 nm maximum. Sample C on the other hand shows much smaller average crystal size well below 1 µm. EDX measurements were performed on the absorber layers after selenization in order to derive the shoichiometry of the layers. All layers were found to be Cu-poor and Zn-rich, with Cu/(Zn+Sn) ~ 0.8 and Zn/Sn ranging from 1.14 to 1.46. As expected, the higher vapor pressure species SnSe and Zn evaporate partly during the selenization process [14], such that Cu/(Zn+Sn) increases after anneal and Zn/Sn mainly decreases after anneal. The stoichiometry data of the absorber layers before and after anneal can be found in table 1.

Table 1: Physical properties of the four different layers used for solar cell fabrication.

Sample A Sample B Sample C Sample D

Layer 1

Layer 2

Layer 3

Sn 170 nm Sn 170 nm Sn 125 nm Zn 110 nm

Zn 150 nm Zn 110 nm Zn 150 nm Cu 110 nm

Cu 110 nm Cu 110 nm Cu 110 nm Sn 255 nm

Layer 4

Layer 5

Cu/(Zn+Sn) after sputtering

Zn/Sn after sputtering

Cu/(Zn+Sn) after selenization

Zn/Sn after selenization

Se/(Cu+Zn+Sn) after selenization

/

/

0.57

1.57

0.77

1.26

1.05

/

/

0.7

1.14

0.85

1.21

1.00

/

/

0.64

2.1

0.79

1.46

1.01

Zn 115 nm

Cu 55 nm

0.57

1.57

0.79

1.14

1.13

4. Electrical and optical properties

Figure 2: I-V characteristics of the four different 1 cm2 CZTSe solar cells under AM1.5G illumination (solid lines) and in the dark (dashed lines).

Figure 2 shows the current-voltage characteristics of the four different fabricated solar cells under AM1.5G illumination (solid lines) and in the dark (dashed lines). All cell parameters are summarized in table 2. Table 2: Electrical properties of the four fabricated cells.

Cell A Cell B Cell C Cell D

Efficiency

VOC

JSC

Fill Factor

Rseries

Rshunt

% 6.3 6.2 0.7 4.4

mV 390 360 233 407

mA/cm2 31.3 36.1 9.1 22.1

% 52 48 34 49

Ω cm2 1.1 3.7 9.4 3.4

Ω cm2 70 400 54 76

Doping in absorber from Mott-Schottky at -2V cm-3 8 1016 5 1016 5 1018 6 1015

Doping in absorber from DLCP at -2V cm-3 1.5 1016 6 1015 5 1018 8 1014

Both cells A and B show efficiencies in excess of 6 %, whereas cell D shows an efficiency of 4.4 %. Sample C shows the lowest efficiency of 0.7 %. Especially the short circuit current densities of cells A and B are particularly large with values of 31.3 and 36.1 mA/cm2 respectively, showing the good carrier collection of the cells, as long as band bending is strong and collection is relatively rapid. The open circuit voltage on the other hand is very low of the order of 0.4 V, which is a very low value for an absorber with bandgap energy of about 1 eV, but is in agreement with findings from other reports [5]. Cell B shows the largest shunt resistance with a value of about 400 Ω cm2, whereas

cell A shows the lowest series resistance with a value of about 1.1 Ω cm2. For all cells the external quantum efficiency was derived as a function of incident light wavelength. The results are shown in figure 3. Cells A, B and D show large quantum efficiency approaching 80 to 90% in a large wavelength region, only cell C shows low quantum efficiency particularly at longer wavelength, for photon absorption far away from the p-n junction. The bandgap derived from the absorption threshold of the EQE plots is of the order of 0.9 eV.

Figure 3: External quantum efficiency of the four different 1 cm2 CZTSe solar cells as a function of illumination wavelength.

Figure 4: Capacitance as a function of bias voltage for 6 different frequencies (10, 15.8, 25.1, 39.8, 63.0 and 100 kHz from top to bottom respectively) varying logarithmically from 10 kHz to 100 kHz for cell A with an AC drive level of 30 mV (a) The caption shows the same data plotted as a Mott-Schottky plot. Capacitance as a function of AC drive level for 9 different bias voltages varying from 0.1 V to -2.3 V with a step of -300 mV for cell A, measured at a frequency of 100 kHz (b). Second order polynomial fits to the data are shown as well (dashed lines).

In order to be able to correlate some of these cell results to the electrical properties of the absorber layers, capacitance-voltage (CV) measurements of the 1 cm2 cells were made as a function of bias voltage (-2.5 V to 0.5 V), frequency (10 kHz to 100 kHz) and AC drive voltage (10 mV to 250 mV). A capacitance-voltage profile with an AC drive level of 30 mV (figure 4a) and a capacitance versus AC drive voltage profile at 100 kHz (figure 4b) is shown in figure 4 for the case of cell A. Well behaved curves were obtained in all cases. The results did not vary with frequency in the range 10 kHz – 100 kHz. At frequencies higher than 100 kHz series resistance and/or fast traps started to influence some of the CVs, whereas at lower frequencies (below 10 kHz) leakage currents were hindering correct measurements in some cases. The doping density was then derived from the Mott-Schottky plot [15] and from DLCP measurements [16]. For

the DLCP measurements, care was taken to correctly shift the bias voltage as the AC drive voltage was increased, in order to obtain constant depletion depth in the absorber layer [16]. A second order fit was made to the Capacitance versus AC drive voltage data (figure 4b), in order to derive the slope and intersection of the curve with zero, from which the doping can be derived using [16]: NDLCP = -C03/(2qεA2C1),

(1)

Where C0 is the intersection of the second order fit with the zero AC drive voltage axis, C1 is the slope of the curve, q is the electron charge, ε the dielectric constant and A the area of the cell. The results of these measurements are shown in figure 5.

Figure 5: Doping density in the buffer layers of the four different cells as a function of applied bias voltage as derived from the Mott-Schottky plot (solid lines) and from DLCP measurements (dashed lines) at 100 kHz.

The doping in the absorber is relatively homogeneous as a function of position (bias voltage) for all samples. For cells A and B, which show the highest efficiencies, the free carrier density derived from the DLCP measurements is of the order of 1016 cm-3. For sample D, which shows a slightly lower efficiency, the doping density is lower, of the order of 1015 cm-3. Sample C on the other hand, which shows efficiency below 1% and has high Zn/Sn ratio, shows a doping density of the order of 5 1018 cm-3, which explains both the low efficiency and the low charge collection efficiency for long wavelengths as derived from the EQE measurements. It is difficult to judge from just 4 samples whether the trend is very relevant, but from figure 6 it becomes apparent that the doping density as measured from DLCP seems to depend exponentially on the Zn/Sn ratio of the absorber as measured with EDX. Therefore, it seems that Zn and/or Sn are either involved directly in the defects responsible for doping in CZTSe or at least they strongly influence the energy level of these defects.

Figure 6: Doping level of the four different layers as derived from DLCP measurements plotted against the Zn/Sn ratio measured on the completed absorber after selenization. The dashed line is an exponential fit to the data.

For samples A, B and C, the doping density derived from the DLCP measurement is a factor 5 lower as the one derived from the Mott-Schottky plot. The Mott-Schottky plot also takes into account charges positioned inside the depletion layer or at the heterojunction and responding to the AC drive level frequency, such as shallow defect levels that are deep enough not to give rise to free carriers, but which are shallow enough to be able to respond to the AC drive frequency of the measurement, when the Fermi level crosses the defect level. DLCP on the other hand derives the variation of charge at the depletion layer edge only [16]. We can therefore conclude that, for our CZTSe layers, the Mott-Schottky plot overestimates the free carrier density inside the bulk of the semiconductor because of the presence of a large amount of defects inside the bandgap that are shallow enough to be able to respond to a 100 kHz signal. These defects are likely similar to the defects found at an energy of 150 meV above the valence band edge energy in similar devices [5]. Further characterization of these defects should be performed with measurements at variable temperature in order to derive the exact activation energy.

5. Conclusions We have fabricated CZTSe solar cells with absorber layers obtained from H2Se selenization of DC-sputtered Cu, Zn, Sn multilayers. Four different cells with absorbers of different stoichiometry were fabricated. The Cu/(Zn+Sn) ratio is about 0.8 for all four cells, whereas the Zn/Sn ratio varies between 1.14 and 1.46. The ideal Zn/Sn ratio seems to be about 1.2 and the highest efficiency obtained on a 1 cm2 cell is 6.3%, with Jsc of 31.3 mA/cm2, Voc of 390 mV and a fill factor of 52%. The doping density in the absorber layer was measured from DLCP. It seems to depend exponentially on the Zn/Sn ratio. The carrier density in the absorber layers was also measured from MottSchottky plots. The results show that the free carrier density is overestimated by MottSchottky plots, because of the presence of large amounts of fast defect states inside the bandgap of the CZTSe, which respond to frequencies up to 100 kHz or higher. Acknowledgments We would like to acknowledge Tom De Geyter and Guido Huyberechts from Flamac in Gent for sputtering of the metal layers. AGC is acknowledged for providing substrates. The Flemish ‘Strategisch Initiatief Materialen’ (SIM) SoPPoM program is acknowledged for their collaboration. References [1] A. Redinger, D.M. Berg, P.J. Dale, R. Djemour, L. Gutay, T. Eisenbarth, N. Valle and S. Siebentritt, IEEE Journal of Photovoltaics 1 (2012) 200. [2] H. Katagiri, K. Jimbo, W. S. Maw, K. Oishi, M. Yamazaki, H. Araki, and A. Takeuchi, Thin Solid Films 517 (2009) 2455. [3] S. Siebentritt and S. Schorr, Progress in Photovoltaics: Research and Applications (2012) DOI: 10.1002/pip.2156. [4] K. Ito, T. Nakazawa, Japanese Journal of Applied Physics 27 (1988) 2094. [5] D. A. R. Barkhouse, O. Gunawan, T. Gokmen, T. K. Todorov and D. B. Mitzi, Progress in Photovoltaics: Research and Applications 20 (2012) 6. [6] P. Jackson, D. Hariskos, E. Lotter, S. Paetel, R. Wuerz, R. Menner, W. Wischmann, M. Powalla, Prog. Photovolt: Res. Appl. 19 (2011) 894. [7] Q. Guo, G. M. Ford, W.-C. Yang, B. C. Walker, E. A. Stach, H. W. Hillhouse and R. Agrawal, J. Am. Chem. Soc. 132 (2010) 17384. [8] D. B. Mitzi, L. L. Kosbar, C. E. Murray, M. Copel and A. Afzali, Nature 428 (2004) 299. [9] T. Tanaka, T. Sueishi, K. Saito, Q. Guo, M. Nishio, K. M. Yu and W. Walukiewicz, J. Appl. Phys. 111 (2012) 053522. [10] H. Araki, A. Mikaduki, Y. Kubo, T Sato, K. Jimbo, W. S. Maw, H. Katagiri, M. Yamazaki, K. Oishi, A. Takeuchi, Thin Solid Films 517 (2008) 1457. [11] P.A. Fernandes, P.M.P. Salomé, A.F. da Cunha, B.-A. Schubert, Thin Solid Films 519 (2011) 7382. [12] J. Klaer, I. Luck, A. Boden, R. Klenk, I. Gavilanes Perez, R. Scheer, Thin Solid Films 431 (2003) 534. [13] Ch. Köble, D. Greiner, J. Klaer, R. Klenk, A. Meeder and F. Ruske, Thin Solid Films 518 (2009) 1204. [14] A. Redinger, S. Siebentritt, Appl. Phys. Lett. 97 (2010) 092111. [15] S.M. Sze, Physics of semiconductor devices, Wiley, New York (1981). [16] J. T. Heat, J. D. Cohen and W. N. Shafarman, J. Appl. Phys. 95 (2004) 1000.