SUPPLEMENTARY INFORMATION Three dimensional nanopillar array photovoltaics on low cost and flexible substrates Zhiyong Fan 1,2,4, Haleh Razavi 1,2,4, Jae-won Do 1,2,4, Aimee Moriwaki 1,2,4, Onur Ergen 1,2,4, Yu-Lun Chueh 1,2,4, Paul W. Leu 1,2,4, Johnny C. Ho 1,2,4, Toshitake Takahashi 1,2,4, Lothar A. Reichertz 2, Gregory F. Brown 2,3, Steven Neale 1,4, Kyoungsik Yu 1,4, Ming Wu 1,4, Junqiao Wu 2,3, Joel W. Ager 2, Ali Javey 1,2,4* 1 Department of Electrical Engineering and Computer Sciences, University of California at Berkeley, Berkeley, CA, 94720, USA. 2 Materials Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA. 3 Department of Materials Science and Engineering, University of California at Berkeley, Berkeley, CA, 94720, USA. 4 Berkeley Sensor and Actuator Center, University of California at Berkeley, Berkeley, CA, 94720, USA. * Corresponding author: ajavey@eecs.berkeley.edu nature materials www.nature.com/naturematerials 1
Nanoimprint for highly regular AAM fabrication a b d c Figure S1: The first (a) and second (b) imprint on an Al substrate with a straight line optical diffraction grating. (c) and (d) are SEM images of the AAM after first and second anodization steps, respectively. 2 nature MATERIALS www.nature.com/naturematerials
Nanopillar exposure by wet etching of AAM Figure S2: Nanopillar exposure length as a function of the AAM etching time in 1 N NaOH solution at room temperature. nature materials www.nature.com/naturematerials 3
Characterization of CdTe thin film a b Figure S3: (a) SEM image of CdTe film after ion milling. (b) X-ray diffraction patterns confirm that as-grown CdTe films are polycrystalline with mixed phase of hexagonal close packed and cubic structures. 4 nature MATERIALS www.nature.com/naturematerials
Transmission spectrum of the Cu/Au top contact The transmission spectrum of Cu/Au (1/13 nm) thin films on a glass substrate was measured by a grating-based spectrometer (SP2360, Princeton Instruments). The sample was illuminated by a high-power halogen lamp using a condenser lens, and the transmitted light through the electrode was collected by a 10x objective. The transmission spectrum of a bare glass substrate was also collected and used for the background correction. The measurement result shown below suggests an average T ~ 50% transmissions for the give wavelength range, which is significantly lower than that of high quality indium-tin-oxide transparent conductive oxide (T ~ 85%). Figure S4 Optical transmission spectrum of the Cu/Au top contacts. nature materials www.nature.com/naturematerials 5
Crystal structure and optical property of CdS nanopillars To characterize the crystal structure of the grown CdS NPLs, an AAM substrate with CdS NPL grown inside was etched fully for ~1.5 hours in 1 N NaOH. Then the substrate was ultra-sonicated in ethanol to disperse the NPLs into the solution. The NPL suspension was dropped onto a copper grid. Then the crystal structure of the NPLs was characterized with a transmission electron microscope (TEM) (JEM-2100F), which was operated at 200 kv, with a point-to-point resolution of 0.17 nm. As shown in Figure S4a and b, CdS NPLs are single crystalline with preferred [110] growth direction. Figure S4c shows the energy dispersive x-ray spectroscopy (EDS) taken from the center part of NPL, revealing that the atomic compositions of Cd and S are 51 % and 49 %, respectively. Figure S4d shows the room temperature photoluminescence (PL) of a single CdS NPL measured by exciting the NPL with a He-Cd laser (8mW of power at 325nm wavelength, IK series from Kimmon. The measured spectrum demonstrates a peak intensity at ~500nm, corresponding to ~2.4 ev band-to-band emission from CdS. 6 nature MATERIALS www.nature.com/naturematerials
a b c d Figure S5: (a) Low magnification TEM image of a CdS NPL and the selected area electron diffraction pattern (inset), showing its single crystalline nature. (b) High resolution TEM image resolves lattice fringes, indicating [110] growth direction. (c) Energy dispersive x-ray spectroscopy (EDS) taken from center part of a NPL. (d) Roomtemperature photoluminescence from a single CdS NPL. nature materials www.nature.com/naturematerials 7
Light trapping with 3D CdS NPL array a b Figure S6: (a) SEM image of ordered CdS NPL array after partial etching of AAM. The inset is a photograph of four substrates with exposed CdS NPL heights of 0, 60, 163 and 231 nm (left to right, respectively). (b) Reflectance spectra of the four substrates shown in the inset of (a). Compared with blank AAM on Al substrate, the reflectance is greatly reduced, with 231 nm exposed NPL height, resulting in a reflectance minima of ~1.6%. 8 nature MATERIALS www.nature.com/naturematerials
Temperature dependency of the SNOP-cell performance Temperature dependent cell performance measurements were performed under ambient conditions. The device was gradually heated up from 297K to 333K, during which dark and light I-V curves were acquired at various temperatures (Fig. S7). To reduce additional heating caused by illumination from the solar simulator, 0.2 sun (20 mw/cm 2 ) intensity was used for the measurements. a b c Figure S7: (a) Dark and light (b) I-V curves of the solar cell obtained at 5 different temperatures from 297K to 333K. (c) Open circuit voltage (V oc ) decreases with temperature. nature materials www.nature.com/naturematerials 9
EBgB (ev) P Simulation of the performance of SNOP-cells and planar cells The conversion efficiencies of the SNOP-cell with varying NPL embedded length in CdTe, H, was simulated by using Sentaurus. Since the goal of the simulation is to qualitatively verify the trend of experimental data shown in Fig. 4b instead of obtaining precise cell performance characteristics, the electrodes are assumed to be transparent to both photon and charge carriers. The materials properties of this simulation were adopted from references 1-3 as summarized in Table S1. The carrier life times for CdTe were purposely chosen to be smaller than that of CdS since the CdS NPLs are single crystalline while the CdTe films are polycrystalline. The Shockley-Read-Hall (SRH) model was chosen as the primary recombination mechanism. Table S1. Materials parameters used for modeling Property CdTe CdS 1.5 2.4 τbe B(ns) 0.1 2.5 τbh B(ns) 0.1 2.5 2-1 -1 1 B(cmP PVP PsP μbh P) 40 25P P 2-1 -1 1 B(cmP PVP PsP μbe P) 100 100P P -3 16 N (cmp P) 1x10P P 5x10P 16 The simulation results shown in Fig. 4b&d suggest that it is beneficial to have CdS NPLs extend into the CdTe film as much as possible to maximize the carrier collection efficiency. However, due to the processing limitations, a maximum H~640 nm was utilized in the experiments, which corresponds to ~6% efficiency extracted from the simulation result shown in Fig. 4b with a CdTe thickness of 1 µm. Additional simulation 10 nature MATERIALS www.nature.com/naturematerials
results using the same material parameters but with a CdTe thickness of 700 nm show a conversion efficiency of ~12 % (results not shown), which suggests a potential direction to improve cell efficiency with even less CdTe material. To further explore the optimal NPL dimensions, the SNOP-cell performance was simulated as a function of the NPL radius while keeping the NPL pitch constant at 500 nm. As shown in Fig. S8, the maximum efficiency was obtained with ~100 nm NPL diameter, which corresponds to the actual NPL dimension used in our experiments. The smaller NPL radius results in reduced carrier collection region. On the other hand, the NPL radius of >100 nm results in a loss of CdTe filling factor, which effectively lowers the absorption efficiency. Figure S8: Simulation of SNOP-cell efficiency versus the radius of CdS NPL. The material parameters are adopted from Table S1 and the device structure is shown in Figure 4d. In order to compare the performance of the SNOP-cell with conventional planar structured CdS/CdTe cell, further simulations were carried out based on the structures shown in Figure S9a and b in which the material parameters are adopted from Table S1, nature materials www.nature.com/naturematerials 11
except that the minority carrier diffusion length, L n is varied from 50 nm to 50 µm. Figure S9c shows the efficiencies of SNOP and planar cells as a function of L n. It is evident that the SNOP-cell is superior to the planar cell due to the improved carrier collection, especially when L n is smaller than the device thickness (2µm). a b c Figure S9: Structures of (a) SNOP-cell and (b) conventional planar cell used for Sentaurus simulation. (c) Conversion efficiencies of the SNOP and planar cells versus the minority carrier (electron) diffusion length of the CdTe film. The total device thickness is fixed at 1.3 µm including electrodes. The inset shows their ratio, depicting the advantage of SNOP-cell, especially when the minority carrier life times are relatively low. 12 nature MATERIALS www.nature.com/naturematerials
References: 1. Gleockler, M., Fahrenbruch, A. L. & Sites, J. R. Numerical modeling of CIGS and CdTe solar cells: setting the baseline. Proc. World Conf. on Photovoltaic Energy Conversion 3, 491-494 (2004). 2. Kosyachenko, L. A., Savchuk, A. I. & Grushko, E. V. Dependence of efficiency of thin-film CdS/CdTe solar cell on parameters of absorber layer and barrier structure. Thin Solid Films 517, 2386-2391 (2009). 3. Castro-Rodriguez, R., Zapata-Torres, M., Zapata-Navarro, A., Oliva, A. I. & Pena, J. L. Heavily doped CdTe films grown by close-spaced vapor transport technique combined with free evaporation. J. Appl. Phys. 79, 184-187 (1996). nature materials www.nature.com/naturematerials 13