Highly active oxide photocathode for. photoelectrochemical water reduction

SUPPLEMENTARY INFORMATION Highly active oxide photocathode for photoelectrochemical water reduction Adriana Paracchino 1, Vincent Laporte 2, Kevin Sivula 1, Michael Grätzel 1 and Elijah Thimsen 1 1 Institute of Chemical Sciences and Engineering, Ecole Polytechnique Fédérale de Lausanne, Laboratory of Photonics and Interfaces, Station 6, CH-1015 Lausanne, Switzerland. 2 Interdisciplinary Centre for Electron Microscopy, Ecole Polytechnique Fédérale de Lausanne, Station 12, CH-1015 Lausanne, Switzerland. Cu(0) formation on bare Cu 2 O Electrodes Evidence of Cu formation in the illuminated area of the measured bare Cu 2 O electrodes is provided by SEM (Figure S1), X-ray photelectron spectroscopy (XPS) spectroscopy (Figure S2), and X- ray diffraction (Figure S4). First, bright nanoparticles were observed on the surface of the Cu 2 O grains by SEM after PEC measurement, which appeared by visual inspection as a black circle in the illuminated area (Figure S1). Depth profile XPS analysis was performed on bare Cu 2 O samples before and after PEC measurement. Before PEC measurement, a native CuO surface oxide was indicated in the Cu 2p region of the XPS spectrum by a shoulder at 933.8 ev and a satellite structure at the higher binding energy side (not shown). The CuO signal disappeared after 30 seconds of sputtering, indicating that the native oxide was less than 2 nm in thickness. The quantification analysis after the first sputtering cycle for the as-deposited sample from the Cu 2p and the O 2p peaks showed a Cu/O ratio that was constant with sputtering time and close to the expected value of 2.77 due to a preferential nature materials www.nature.com/naturematerials 1

supplementary information oxygen sputtering 1. After PEC stability measurement, the quantification analysis showed Cu/O values between 3.5 and 7.5, according to the duration of sputtering, indicating enrichment of copper relative to the as-deposited Cu 2 O. We therefore conclude that the sample measured after PEC is a mixture of Cu (formed via reaction (1) of the main text) and Cu 2 O. An XRD peak for Cu metal was also observed after PEC measurement (Figure S4), which was absent in XRD spectra of the as-deposited sample. a b Figure S1. SEM images of a bare Cu 2 O electrode before (a) and after (b) PEC stability measurement. Cu nanoparticles are visible on the surface of the Cu 2 O grains after PEC measurement. The insets are 2 nature MATERIALS www.nature.com/naturematerials

supplementary information digital images of the electrodes: the black circle visible on the measured electrode is the illuminated area, where photocathodic decomposition occurred to form Cu(0). Figure S2. Atomic concentrations from XPS analysis for Cu and O in a Cu 2 O electrode as-deposited and after PEC measurement. The horizontal lines indicate the expected Cu (dark green) and O (light green) concentrations according to the preferential oxygen sputtering in Cu 2 O under Ar + bombardment 1. Role of the ZnO buffer layer and Al 2 O 3 doping (this discussion refers to Table 1 of the main text) The ZnO buffer layer between the Cu 2 O and TiO 2 was critical to improve the stability of the photocurrent. This can be seen by comparing the bare electrode (sample 1), Cu 2 O/11 nm TiO 2 electrode (sample 2) and the Cu 2 O/20 nm ZnO/11 nm TiO 2 electrode (sample 3). For the bare electrode, the photocurrent was very unstable and decayed to 0% of its initial value after 5 minutes (Figure 1 a). Depositing 11 nm of TiO 2 on the surface did little to enhance the stability, as the photocurrent again decayed to 0% of its initial value after 20 minutes. It should be noted that increasing the TiO 2 thickness did not increase the stability, and electrodes with 30 nm TiO 2 protective layers still decayed to 0% of nature materials www.nature.com/naturematerials 3

supplementary information their initial value after 20 minutes. However, if 20 nm of ZnO was used as a buffer layer between the Cu 2 O and TiO 2, the initial photocurrent was greatly improved to 7.8 ma cm -2 and retained 14% of its initial value after 20 minutes (sample 3), a vast improvement over both the bare and TiO 2 -only samples. The hypothesis is that the ZnO buffer is acting as a nucleation layer to control the mechanism of TiO 2 growth. The TiO 2 growth is presumably layer-by-layer on the fresh ZnO and island growth on the electrodeposited Cu 2 O. The stability of the ZnO layer was further improved by inserting periodic subnanometer layers of Al 2 O 3 into the ZnO buffer layer. Monolayers of Al 2 O 3 were inserted into the layered structure with different configurations by alternating TMA/H 2 O cycles with the DEZ/H 2 O cycles during ZnO deposition, which is a well known route to synthesizing Al-doped ZnO by ALD 2. First, one Al 2 O 3 layer approximately 0.17 nm in thickness was placed periodically every 4 nm in the ZnO, with the final Al 2 O 3 layer at the ZnO/TiO 2 interface (sample 4). Second, one Al 2 O 3 layer approximately 0.17 nm in thickness was placed every 2 nm in the ZnO with the final layer at the ZnO/TiO 2 interface (sample 5). The overall thickness of the ZnO:Al layer was constant at approximately 21 nm. For the sample with Al 2 O 3 layers spaced every 4 nm (sample 4), the photocurrent remained the same as the ZnO-only sample, but the stability improved to 33% of the initial photocurrent after 20 minutes. When the Al 2 O 3 layers were placed every 2 nm, the stability further improved to retain 53% of its initial value after 20 minutes of illumination at 0 V vs. RHE, but was accompanied by a drop in the initial photocurrent to 4.7 ma cm 2, presumably due to the high aluminum content that increased the resistance of the protective layer either through tunnel barriers or through reduced electron mobility. To further elucidate the effect of Al, Al 2 O 3 layers 0.9 nm in thickness were placed at the ZnO/TiO 2 interface (sample 6) and Cu 2 O/ZnO interface (sample 7). In both samples, the initial photocurrent at 0 V vs. RHE dropped significantly, to 2.7 ma cm 2 for sample 6 and 2.4 ma cm 2 for sample 7, presumably due to the rather thick Al 2 O 3 layer that presented a large barrier to electron tunneling. The interesting trend is in the stability. The sample with 0.9 nm of Al 2 O 3 at the ZnO/TiO 2 interface (sample 6) had approximately the same photocurrent stability after 20 minutes as the sample 4 nature MATERIALS www.nature.com/naturematerials

supplementary information with Al 2 O 3 layers distributed periodically every 2 nm (sample 5). However, the electrode with the 0.9 nm Al 2 O 3 at the Cu 2 O/ZnO interface had approximately the same stability as the electrode with no Al 2 O 3 (sample 3). We therefore conclude that the Al 2 O 3 is not stabilizing the Cu 2 O, but instead stabilizing the ZnO. Hall effect measurements were carried out at room temperature using an Ecopia HMS-3000 to determine any electronic role Al 3+ might be playing in the Al-doped ZnO layers through its effect on the majority carrier density. Films were deposited on square quartz substrates at identical conditions to the protective layers, and the Al content was controlled by varying the TMA/H 2 O to DEZ/H 2 O cycle ratios. The thickness of the Al-doped ZnO layers was 21 nm in every case, as measured by elipsometry on "witness" samples deposited on optically polished silicon. The Al atom fraction was varied from 0.0 to 4.3 %, as calculated by the rule of mixtures 2. The electron mobility, resistivity and electron density are plotted in Figure S3. For the as-deposited films, the carrier densities for the Al 3+ atom fractions of 1.4%, 2.2% and 4.3% were 5.0 x 10 19 cm 3, 7.9x10 19 cm 3 and 1.9x10 20 cm 3 respectively; which was much lower than the expected Al 3+ atom concentration of 5.7x10 20 cm cm 3, 9.3x10 20 cm 3 and 1.8x10 21 cm 3 respectively. After heat treating the films for 60 min at 200 o C in air, the carrier concentration in the ZnO dropped dramatically to 3.6x10 16 cm 3, 3.8x10 17 cm 3 and 1.1x10 18 cm 3 for respective Al 3+ atom fractions of 1.4%, 2.2% and 4.3%. In the as-deposited case, the carrier concentration was approximately 10% of the expected Al 3+ atom concentration, indicating that 90% of the aluminum was not serving as an electron donor, but was instead serving a different role. The effect was even more dramatic after heat treatment, where less than 0.1% of the Al 3+ was serving as an electron donor. We believe this indicates that the Al 3+ was serving a structural role, perhaps as an isolated Al 2 O 3 phase in the periodic layers where it was deposited. This conclusion is supported by the observation that for Al 3+ -containing samples, the electron mobility decreases monotonically with increasing Al 3+ atom concentration (Figure S3 a). nature materials www.nature.com/naturematerials 5

supplementary information a b c Figure S3. Resistivity (a), electron mobility (b) and carrier density (c) from Hall effect measurements on 21 nm Al:ZnO films (constant thickness) with different Al 2 O 3 :ZnO cycle ratios (0, 1:10, 1:20, 1:33) deposited on quartz substrates by ALD. The Al fraction was calculated using the rule of mixtures and assuming the bulk density of ZnO 2. 6 nature MATERIALS www.nature.com/naturematerials

a supplementary information b c Figure S4. XRD before (a) and after (b) PEC characterization for the bare Cu 2 O electrode and after PEC characterization on the surface-protected Cu 2 O (c). nature materials www.nature.com/naturematerials 7

supplementary information Figure S5. H 2 bubbles evolving from the illuminated protected photocathode biased at 0 V vs. RHE. a b Figure S6. Photocurrent spectrum (IPCE) at 0 V vs. RHE (a) and current-potential characteristics (b) for a heat treated Cu 2 O / 5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 11 nm TiO 2 /Pt electrode immersed in 1 M Na 2 SO 4 solution under AM1.5 irradiation and N 2 purging. The photocurrent value calculated by integration of the IPCE curve over the AM1.5 solar spectrum was 25% lower than the measured value from the JV plot, but consistent with the observed photocurrent decay during chronoamperometry at 0 V vs. RHE. 8 nature MATERIALS www.nature.com/naturematerials

supplementary information Figure S7. XPS spectra of Cu 2p as a function of sputtering time in the protective layer of a heat treated Cu 2 O / 5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 11 nm TiO 2 /Pt electrode. No Cu was detected in the protective layer. Each spectra is vertically offset for clarity. Chemical stability of Cu 2 O The chemical stabilization of Cu 2 O was confirmed by the position of the Auger Cu LMM signal in XPS. The Cu LMM peak for an electrode measured for 80 minutes under irradiation at 0 V vs. RHE (Figure S8a) showed the same position as for a bare Cu 2 O that wasn t exposed to a PEC measurement (Figure S8b). The reduction of bare Cu 2 O under photocathodic conditions (Figure S8c) is apparent from a peak shift to a higher kinetic energy as well as from the narrowing of the peak. These features were not observed for a Cu 2 O / 5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 11 nm TiO 2 /Pt electrode after PEC measurement, which is consistent with the chemical stabilization of the photoactive semiconductor. nature materials www.nature.com/naturematerials 9

supplementary information Figure S8. Auger Cu LMM spectra obtained after XPS depth profiling for a protected Cu 2 O electrode measured for stability under light at 0 V vs. RHE for 80 minutes (a), a bare Cu 2 O not measured (b) and a PEC-measured bare Cu 2 O electrode (c). 10 nature MATERIALS www.nature.com/naturematerials

supplementary information Formation of Ti 3+ in the amorphous TiO 2 layer The theory of Ti 3+ formation in the TiO 2 layer was tested by a mild in-situ air oxidation cycle at room temperature in the photoelectrochemical cell by the following procedure. Prior to any PEC measurements, the electrolyte was purged with N 2 for at least 15 minutes. A 20 minute stability measurement was carried out under chopped illumination and N 2 purging. After the stability measurement, the electrolyte was purged in the dark with air for 15 minutes, followed by a 15 minute N 2 purge, to complete an air/n 2 purge cycle. Another 20 minute stability measurement was then recorded, followed by another air/n 2 purge cycle. Thus after each 20 minute stability measurement under N 2 purging, an air/n 2 purge cycle was completed before starting the next 20 minute stability measurement. The results for a heat-treated Cu 2 O/5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 11 nm TiO 2 / Pt electrode and a Cu 2 O/5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 20 nm TiO 2 / Pt electrode are presented in Figure S8. The time axis for Figures S9b and S9c is the time under illumination (i.e. the elapsed time during the air/n 2 cycle is not included). After 40 minutes under chopped illumination, the photocurrent of the electrode with 11 nm of TiO 2 decayed and was not reversible after an air/n 2 purge cycle. The photocurrent of the electrode with 20 nm of TiO 2 decayed after 80 minutes of stability measurement and was not reversible by an air/n 2 purge cycle. We believe this result indicates that a critical concentration of Ti 3+ traps was reached that the air/n 2 purge cycle was insufficient to reverse and thus the initial photocurrent could not be restored. The mechanism proposed for the photocurrent decay is consistent with the XPS Ti 2p signals plotted in Figure S11. The peak at 459 ev is due to Ti 4+ and the shoulder around 457 ev is due to Ti 3+. In an untested sample (Figure S11a), Ti 3+ formation is due to the chemical modification induced by the Ar + sputtering required for the XPS depth profiling. The sample PEC-tested for 20 minutes (Figure S11b) presents a large Ti 3+ shoulder at already 30 s of sputtering, which conclusively indicates that it is enriched in Ti 3+ compared to the not measured sample. Owing to the constant total amount of Ti, the intensity of the Ti 4+ peak is decreasing as the Ti 3+ shoulder becomes larger. The untested sample of nature materials www.nature.com/naturematerials 11

supplementary information Figure S11a was analyzed by XPS depth profiling some months after the sample of Figure S11b, inducing possible slightly different sputtering conditions, but the trend in the Ti 3+ and the Ti 4+ signals with increasing sputtering depth was very reproducible with other similar samples PEC-measured for 20 min or more. a b c 12 12 nature MATERIALS www.nature.com/naturematerials

supplementary information Figure S9. Stability of the photocurrent at 0 V vs. RHE for a Cu 2 O/5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 11 nm TiO 2 /Pt electrode and a Cu 2 O/5 x (4 nm ZnO / 0.17 nm Al 2 O 3 ) / 20 nm TiO 2 /Pt electrode. In chronological order, the JV plot (a) shows the initial linear sweep for the 11 nm TiO 2 sample (LS-1), linear sweep after 20 minutes of stability measurement (LS-after stability), after an air/n 2 purge cycle (LS-2) and after 40 minutes of stability measurement (LS-after stability 2). The stability measurement in panel (b) shows that the photocurrent recovers after the first air/n 2 purge cycle at 20 minutes, but after the second air/n 2 purge cycle at 40 minutes the photocurrent decay is irreversible. Panel (c) shows the same stability experiment for the electrode with 20 nm of TiO 2, which results in the photocurrent being restored after air/n 2 purge cycles at 20, 40 and 60 minutes. However, the photocurrent decay eventually becomes irreversible after 80 minutes of light chopping at 0 V vs. RHE. Figure S10: Pourbaix diagram for TiO 2 in 1M Na 2 SO 4 electrolyte at T=25 C, for [Ti 2+ ]=10 M. Generated using Medusa (http://www.kemi.kth.se/medusa/). nature materials www.nature.com/naturematerials 13

supplementary information Figure S11. Ti2p spectra obtained after XPS depth profiling for an untested electrode (a) and for an electrode tested for stability for 20 minutes. The Ti 2p signals are presented for different Ar + sputtering times. The peak at 459 ev is due to Ti 4+ and the shoulder around 457 ev is due to Ti 3+. The data in b) are the same that have been plotted as contour plot in Figure 4 of the main text. 14 nature MATERIALS www.nature.com/naturematerials

supplementary information References 1 2 Malherbe, J.B., Hofman, S., & Sanz, J.M., Preferential sputtering of oxides: A comparison of model predictions with experimental data. Applied Surface Science 27, 355-365 (1986). Elam, J.W., Routkevitch, D., & George, S.M., Properties of ZnO/Al2O3 alloy films grown using atomic layer deposition techniques. Journal of the Electrochemical Society 150 (6), G339-G347 (2003). nature materials www.nature.com/naturematerials 15