Photoelectrochemical CO2 Reduction at a Direct CuInGaS2/Electrolyte Junction

Photoelectrochemical (PEC) CO2 reduction has received considerable attention given the inherent sustainability and simplicity of directly converting solar energy into carbon-based chemical fuels. However, complex photocathode architectures with protecting layers and cocatalysts are typically needed for selective and stable operation. We report herein that bare CuIn0.3Ga0.7S2 photocathodes can drive the PEC CO2 reduction with a benchmarking 1 Sun photocurrent density of over 2 mA/cm2 (at −2 V vs Fc+/Fc) and a product selectivity of up to 87% for CO (CO/all products) production while also displaying long-term stability for syngas production (over 44 h). Importantly, spectroelectrochemical analysis using PEC impedance spectroscopy (PEIS) and intensity-modulated photocurrent spectroscopy (IMPS) complements PEC data to reveal that tailoring the proton donor ability of the electrolyte is crucial for enhancing the performance, selectivity, and durability of the photocathode. When a moderate amount of protons is present, the density of photogenerated charges accumulated at the interface drops significantly, suggesting a faster charge transfer process. However, with a high concentration of proton donors, the H2 evolution reaction is preferred.

redispersing them in toluene, adding ethanol again to precipitate them and centrifuging the mixture to separate them. This procedure was repeated 3 times. Finally, the nanocrystal pellet was redispersed in 1-hexanethiol (CH 3 (CH 2 ) 5 SH, 95%, Sigma-Aldrich) to prepare the nanocrystal ink (200 mg mL −1 ). The nanocrystal ink was tape-casted onto flexible molybdenum foils (0.1 mm, 99.95%, chemPUR) and annealed at 250°C for 4 min in air in a hot plate (VWR 10027-246). This procedure was repeated twice to reach a film thickness of around 800 nm. The films were spray-coated with a 6.5 M SbCl 3 (99.9%, Aldrich) methanol solution (0.35 mmol antimony cm −2 , geometric area of electrode), flame-sealed in an ampoule containing 6 mg of sulfur powder and annealed in an oven pre-heated at 550°C for 30 min.
Finally, the ampoules were removed from the oven and let cool down naturally.

S1.3 Surface Roughness Factor Determination
The roughness factor of CIGS film was determined by the atomic force microscopy (AFM).
An average roughness factor of 1.22±0.02 was obtained by measuring 5 different samples.

S1.4 Photoelectrochemical Characterizations
Photoelectrochemical measurements were carried out in a cappuccino-type electrochemical cell (0.238 cm 2 active geometric area, Figure S1, full description can be found in S2 ) with a 3-electrode configuration: a CIGS working electrode, a Pt mesh counter electrode and a Ag/Ag + (acetonitrile/0.1 M tetrabutylammonium perchlorate/0.01 M AgNO 3 ) non-aqueous reference electrode (RE-7, ALS Co., Ltd). The 3-electrode configuration was controlled by a Bio-Logic SP-300 potentiostat. The applied potential was calibrated to ferrocene redox couple (Fc + /Fc, 98%, Acros). A 450 W xenon arc lamp (Newport 66921), that was calibrated to AM 1.5G, was used as light source. Linear sweep voltammetry (LSV) was recorded at a scan rate of 20 mV/s. The incident photon-to-current efficiency (IPCE) was measured on a tunable light source platform (TLS-300XU, Newport) including a 300 W xenon arc lamp S5 (Newport 6258), a DC arc lamp power supply (OPS-A500 DC), a cornerstone 130 monochromator (CS130-USB-3-FH) and a research arc lamp housing (Newport 66902). The photon flux was calibrated by a calibrated Si photodiode (FDS100-CAL, Thorlabs). Impedance measurements were carried out on the Bio-Logic SP-300 potentiostat with frequency ranges from 5 MHz to 0.2 Hz and a 25 mV sinusoidal amplitude. Impedance data were fitted with equivalent circuits using modeling software ZView (Scribner Associates). Figure S1: Schematic of a cappuccino-type electrochemical cell.

S1.5 Intensity-Modulated Photocurrent Spectroscopy (IMPS)
An array of white LED (Cree XLamp MC-E Color), that was powered by a background DC current of 600 mA, was used as the light source. An arbitrary function generator (Tektronix AFG3021C) was used to sinusoidally-modulate the light intensity in a frequency range of 50 kHz to 0.5 Hz with a 100 mA modulation amplitude (ca. 16% modulation depth). The applied potential on working electrode was controlled by a Keithley 2450 source measure unit (SMU) and the photocurrent response was monitored by a digital phosphor oscillo-S6 scope (Tektronix DPO7254C) through a differential probe (Tektronix TDP3500) in parallel with a 50 Ohm resistor (Velleman ED/E12) which was in series with the counter electrode.
The entire IMPS setup was enclosed with blackout materials (Thorlabs TB4). IMPS data were fitted using custom Python program with lmfit package to perform curve fitting with nonlinear regression.

S1.6 Gas Chromatography Measurements and Product Quantification
Products quantification of the PEC CO 2 reduction on prepared CIGS photocathodes were performed in a homemade single-chamber PEEK cell using a CIGS photocathode as working electrode and a Pt foil as counter electrode, and a Ag/Ag + (acetonitrile/0.1 M tetrabutylammonium perchlorate/0.01 M AgNO 3 ) non-aqueous reference electrode (RE-7, ALS Co., Ltd). LCS-100 solar simulator (Newport, with air mass 1.5 G filter) was used as the light source. The light intensity was controlled with a calibrated silicon diode with KG 3 filter, by adjusting its distance from the light source. Before each test, CO 2 gas (99.999%, Carbagas) was infused into electrolyte for 10 min to saturate the electrolyte (6 mL) and it was continuously bubbled at a flow rate of 10 cm 3 /min during the tests.
Chronoamperometry CO 2 reduction were performed under simulated one sun illumination at selected potentials by a Gamry potentiostat (Interface 1000). The gas products were periodically injected to and detected by an online gas chromatography (GC, Trace ULRTA, Thermo), where a micropacked shincarbon column (Restek) and pulse discharge detector (PDD, Vici) was used for gas separation and detection, separately. A standard gas purchased from Carbagas was used to calibrate the PDD peak signal, which was a mixture of multiple gaseous products (H 2 , CO, CH 4 , C 2 H 4 , and C 2 H 6 ) of known concentration with CO 2 matrix.
A representative GC trance of H 2 and CO in M-ACN is shown in Figure S2. The Faradaic efficiency (FE) for producing each gas product (H 2 and CO) was calculated by the following equation: where n(H 2 ) and n(CO) is the amount of H 2 and CO detected in one GC injection (mol), N A is the Avogadro constant, j t 0 is the photocurrent during the injection, ∆t is the time required to fill the sample loop (100 µL) of GC, and q is the elementary charge.

S2 Scanning Electron Microscope (SEM) Images
Scanning electron microscope (SEM) configurations (Zeiss Gemini) for these images were 3 kV electron high tension (EHT), 150 pA probe current, 3 mm working distance, 20 µm aperture size and In-Lens annular secondary electrons detector. CIGS thin film photocathodes with thickness around 800 nm and average grain size of ca. 450 nm were fabricated by a solution based method as described in the Section S1.1.

S3 Atomic Force Microscopy (AFM) Image
Topographic image was acquired on an Asylum Research Cypher S AFM.  Figure S5.
Note that the EDX analysis corroborates that the bulk and fed stoichiometry match well.
X-ray diffraction (XRD) pattern was recorded on a Bruker D8 Discover diffractometer with a non-monochromatized Cu-source, a Nickel filter and a LYNXEYE XE energydispersive 1-D detector in Bragg-Brentano geometry.

S7 Control Experiments for CH 3 OH Oxidation
CH 3 OH has been widely used as hole scavenger for photocatalytic water splitting. Under certain circumstances, H 2 could be one by-product due to CH 3 OH oxidation. S4 Likewise, CH 3 OH oxidation has been extensively studied in direct methanol fuel cells as the anode reaction where the CH 3 OH-to-CO 2 conversion is the dominant oxidation reaction. S5 It is essential to validate if CH 3 OH oxidation has any contribution towards the H 2 we observed during PEC CO 2 reduction. With the aim of resolving the origination of produced H 2 , we performed chronoamperometry on two Pt foils at 4 mA/cm 2 (−2.1 V vs. Fc + /Fc) in pure CH 3 OH with 0.1 M TBAP in the same gas tight cell with He as the carrier gas to simulate PEC CO 2 reduction condition on CIGS photocathodes. Considering H 2 production is dominated by proton reduction on the cathode, there are two potential scenarios in such control experiments. Scenario 1 -CH 3 OH oxidation produces H 2 , the Faradaic efficiency (FE) of H 2 should be greater than 100 % because both cathodic and anodic current are contributing to H 2 production. Scenario 2 -CH 3 OH oxidation produces only CO 2 , the FE of H 2 should be close to 100 % because of proton reduction and we should observe the peak of CO 2 in gas chromatography (GC).
GC trace of CH 3 OH oxidation is displayed in Figure S8 where both H 2 and CO 2 are clearly observed. The control experiments have been performed twice and product quantification results are list in Table S1. As we observed, FE of H 2 is close to 100 % and CO 2 peak signal is significant (quantification of CO 2 is not available due to the absence of calibration gas). This is a sign that our experimental condition is close to scenario 2, similar to the case of direct menthol fuel cells, where the dominate oxidation reaction is CH 3 OH-to-CO 2 conversion. Note that we observe trace of CO which is an intermediate during CH   Mott-Schottky (MS) analysis (Figure 4a) based on the value of C bulk is performed in dark condition in Figure S10a: where q is elementary charge, ε is relative permittivity (taking 10 for CuIn 0.3 Ga 0.7 S 2 ), S7 ε 0 is vacuum permittivity, A is the effective surface area that results from multiplying the geometric area by the surface roughness factor (1.22±0.02), N A is acceptor density, V is applied potential, V f b is flat-band potential, k is Boltzmann constant and T is absolute temperature.
Mott-Schottky analysis reveals a V f b of −0.5 V vs. Fc + /Fc and a N A of 1.68×10 18 cm -3 .

S17
Density of surface states (DOSS) is converted from the value of C SS when EIS is performed under illumination in Figure S10b: Total density of surface states (N SS ) is determined by integrating the DOSS as a function of voltage:

S10 Butler Plot
The flat-band potential (V f b ) has also been estimated by the Butler method, S8 where a V f b of −0.504 V vs. Fc + /Fc has been obtained from Figure S11.  Figure S12: Band diagram of CIGS at flat band condition. E 0 CO 2 /CO represents the redox potential of CO 2 /CO in acetonitrile, which is around −1.28 V vs. Fc + /Fc. S9,S10 S20 S12 Intensity-Modulated Photocurrent Spectroscopy

(IMPS) Analysis
Intensity-modulated photocurrent response j(ω) is fit to the following equation: S11 where j e represents the electron current towards interface and τ d is the mean transit time for photogenerated holes. k tran and k rec are pseudo-first order rate constant for charge transfer and recombination, respectively. C SC and C H represent capacitance of the space charge layer and the Helmholtz layer. α (0< α ≤1) is introduced as a non-ideality factor.

S21
The charge transfer efficiency (T E) is determined by the following equation and plotted as a function of potential in Figure S14: with both co-catalyst/overlayer = 1. Figure S15: Comparison of device simplicity, selectivity, Faradaic efficiency (FE), photocurrent density (j ) and stability against representative photocathodes for PEC CO 2 reduction in organic solvent. Detailed comparison can be found in Table S2.