Rational Design of Carbon Nitride Photoelectrodes with High Activity Toward Organic Oxidations

Abstract Carbon nitride (CN x ) is a light‐absorber with excellent performance in photocatalytic suspension systems, but the activity of CN x photoelectrodes has remained low. Here, cyanamide‐functionalized CN x (NCNCN x ) was co‐deposited with ITO nanoparticles on a 1.8 Å thick alumina‐coated FTO electrode. Transient absorption spectroscopy and impedance measurements support that ITO acts as a conductive binder and improves electron extraction from the NCNCN x , whilst the alumina underlayer reduces recombination losses between the ITO and the FTO glass. The Al2O3|ITO : NCNCN x film displays a benchmark performance for CN x ‐based photoanodes with an onset of −0.4 V vs a reversible hydrogen electrode (RHE), and 1.4±0.2 mA cm−2 at 1.23 V vs RHE during AM1.5G irradiation for the selective oxidation of 4‐methylbenzyl alcohol. This assembly strategy will improve the exploration of CN x in fundamental and applied photoelectrochemical (PEC) studies.


Carbon nitride (CNx) synthesis
Unfunctionalized CNx was synthesized following a previously reported procedure. [1] Melamine (5 g) was heated at 550 °C for three hours (ramp rate 1 °C min −1 ) under air. The obtained yellow powder (50% yield) was ground using a pestle and mortar.

NCN CNx synthesis
Cyanamide-functionalized CNx was synthesized by mixing the CNx powder with potassium thiocyanate (weight ratio 1:2) and by heating the mixture firstly at 400 °C for an hour and then at 500 °C for 30 min (ramp rate 30°C min −1 ) under Ar following a published procedure. [2] The powder was allowed to cool to room temperature, washed twice with H2O and dried overnight at room temperature. NCN CNx was co-deposited with commercial ITO nanoparticles (~50 nm) on FTO-coated glass, adapting a literature procedure. [3] Different masks were prepared and used as a template for the deposition of the photo(electro)catalyst mix, either by cutting Parafilm with a drilling bill to obtain a 0.25 cm 2 , 0.33 cm 2 , and 0.5 cm 2 circle. The Parafilm template was pressed onto the FTO side (1×2 cm) and slightly heated (45 s in a 120°C drying oven) to ensure uniform electrode size. NCN CNx and ITO nanoparticles were combined at different weight ratios (50 mg total), dispersed in ethanol (1.1 mL) and ultrasonicated for 30 min. 5 μL of the mixture were drop-casted onto the templated FTO glass and allowed to dry in air producing 1 layer of catalyst. In this work, up to 2 layers of catalyst were studied. Once the last layer had dried, the mask was removed, and the electrodes were annealed at 250°C for 1 h under Ar (ramp rate 10°C min −1 ).

ALD deposition of Al2O3 on FTO glass
A thin Al2O3 layer was grown by a commercial (AT-410, Anric Technologies) atomic layer deposition (ALD) on the FTO coated glass as a hole blocking layer. The precursors used were trimethylaluminum (Al(CH3)3) and water (H2O) that were maintained at room temperature. A constant flow of 29 sccm of N2 was used to carry the precursors to the reaction chamber preheated at 150 ºC. Each ALD cycle consisted of 3 pulses of Al(CH3)3, 11 s of N2 purge, 2 pulses of H2O and 13 s of N2 purge. Each cycle was calibrated to produce an Al2O3 layer of 0.91 Å thick. Three different thickness were tested by depositing 2 layers (1.8 Å), 3 layers (2.7 Å), and 4 layers (3.6 Å) of Al2O3. The highest performance was achieved with the thinner layer (see Figure S3b), which was employed for all sample in this work, unless stated otherwise.

Photoelectrochemical Set-Up
A single compartment cell in a three-electrode configuration was filled with Na2SO4 electrolyte (0.1 M, 9 mL, pH 7) and the substrate of interest (i.e., 4-methylbenzyl alcohol (4-MBA), ethylene glycol (EG)) in a 50 mM concentration. For methanol, ethanol, and glycerol a 10% v/v solution was tested. Ag/AgCl in saturated KCl was used as reference electrode and Pt (Pt mesh supported on a Pt wire) as the counter electrode. After purging the cell with N2 (~15 min), linear sweep voltammograms were performed (LSVs), applying a potential from -0.2 to +1.6 V vs. RHE at a rate of 10 mV s −1 under chopped simulated solar light (AM 1.5G, 100 mW cm −2 , 10 s on/off intervals). All measurements were conducted on three separate batches of samples, to verify reproducibility and to calculate standard deviation. Photoelectrochemical impedance spectroscopy (PEIS) measurements were performed on a PGSTAT302N potentiostat (Metrohm-Autolab, The Netherlands) using the same conditions as the photoelectrochemical tests, but under a constant flow of Ar. PEIS measurements were carried out at selected applied potentials with a sinusoidal perturbation of 10 mV and a frequency range from 100 kHz to 0.1 Hz.

Characterization
UV-Vis spectroscopy. UV-vis spectra were recorded on a Varian Cary 60 UV-Vis spectrophotometer using a diffuse reflectance accessory (for powder and panel samples). The measured diffuse reflectance of the mixture is then inverted directly by the program using the Kubelka-Munk theory.
FT-IR spectroscopy. FT-IR spectra were recorded on a ThermoScientific Nicolet iS50 spectrometer. The Omnic software was used for analysis.
Electron microscopy. Scanning electron microscopy (SEM) was conducted on a TESCAN MIRA3 FEG-SEM. Samples were sputter-coated with a 10 nm layer of Cr (for SEM coupled with energy-dispersive X-ray spectroscopy, or EDX) prior to microscopy. Transmission electron microscopy (TEM) was conducted on a Thermo Schientific (FEI) Talos F200X G2 TEM. All samples were prepared by scratching the drop casted catalyst mixture from the FTO glass post-annealing. The mixture was then dispersed in ethanol (low concentration) and drop-cast onto carbon-coated Cu TEM grids and allowed to dry before use.
X-ray photoelectron spectroscopy (XPS). XPS analysis was carried out using a Escalab 250XI spectrometer from Thermo Fisher Scientific (West Sussex, UK). The instrument was operating in constant analyzer energy mode. A monochromatic Al-Kα source (1486.74 eV), a flood gun for charge neutralization and 300um of spot size were used. Survey scans were acquired using pass energy of 100 eV, 3 scans were recorded using 0.5 eV step size and dwell of 50ms. Two spectra at room temperature from different areas of the sample were acquired and then results were averaged.
Incident photon-to-current efficiency (IPCE) measurements. IPCE of the selected electrodes was determined using a New-port Oriel 66881 setup, on Oriel 74000 Cornerstone monochromator to selector the wavelength, with a full width at half maximum (FWHM) of 15 nm. The wavelength was typically varied between 300-500nm in 25 nm steps. An Air Mass 1.5 Global (AM 1.5G) solar filter was employed. The light intensity at different wavelengths was determined using a Thorlabs PM100D thermal power meter with a Thorlabs S302C thermal power sensor. The resulting photocurrent was measured on an IviumStat potentiate, with the sample as the working electrode, an Ag/AgCl/KCl (sat) reference electrode and platinum as counter electrode. All the measurements reported herein were performed in a 50 mM 4-MBA 0.1 M Na2SO4, with an external applied bias of 0.65 V vs Ag/AgCl (1.25 V vs RHE, following Equation

Equation 1
The IPCE values were determined using Equation 2, where h is the Planck constant, c is the speed of light, J is the photocurrent density, e is the electron charge and P is the wavelength-dependent light intensity flux.

Equation 2
Faradaic efficiency (FE%). The FE of the photoelectrodes was determined by comparing the moles of products (moli) to the total charge passed (Q), as shown in Equation 3. The equivalent charge used per molecule converted is noted z=2, while F is the Faraday constant.

Equation3
The total charge was obtained by integration of the recorded current (I) over the duration (t) of the chronoamperometry, illustrated in Eq. 4, where t0 and te a represent the beginning and end-points of the measurement.

Equation 4
All samples were measured in a 3-electrode setup with the sample as working electrode, Pt mesh counter electrode and an Ag/AgCl (3M KCl) reference electrode. Samples were immersed into approximately 10 mL of 0.1 M Na2SO4, with or without 50 mM 4-MBA as described in the text / figure captions. The quartz cell was sealed and degassed for 30 minutes with nitrogen gas prior to measurements.
Spectroelectrochemistry: All SEC experiments were carried out in the absence of 4-MBA. JV curves were measured using an Ivium Vertex potentiostat at a sweep speed of 10 mVs -1 . Absorbance changes were measured using an in-house setup. The probe beam was generated using a Tungsten lamp (Thorlabs SLS201L/M with color-temperature balancing filter FGT05165 and collimation package attached), focused onto the sample and then directed into a liquid light guide (Thorlabs LLG5-4Z, 5 mm diameter, 420-2000 nm). The light was collected using a CCD operating at -80 degrees Celsius (Oxford Instruments, iDUS 420A-BEX2-DD). Absorbance changes and JV curves were measured simultaneously, with time synchronization and data collection managed using an in-house LabView program.
PIA, TAS and TPC: Unless stated otherwise, all measurements were carried out in the presence of 50 mM 4-MBA. Unless stated otherwise, a potential of 1.2 V vs RHE (0.6 V vs Ag/AgCl) was applied using an Metrohm Autolab PGSTAT 101 operating in chronoamperometry mode. PIA and TPC data were collected simultaneously using an in-house setup controlled using a LabView program. TPC data was collected using a Tektronics DPO-2012B digital phosphor oscilloscope connected directly to the potentiostat. PIA/TAS spectra and kinetics were measured in transmission mode using an in-house setup. The probe beam used for PIA measurements was produced using white light from a 100 W Quartz Tungsten Halogen lamp (Bentham IL1). The white light was passed through an adjustable monochromator, followed by collimating, and focusing lenses and an iris. After transmittance through the sample the beam was passed through further collimating and focusing lenses, a second monochromator and a final focusing lens before incidence on a silicon photodiode (Hamamatsu S3071). The probe beam wavelength was incremented in steps of 50 nm by adjusting both monochromators. We note that the amplitudes of the PIA kinetics are not comparable between Figures 13.a and 13.b as the magnitude of the absorbance change is very sensitive to the sample history.
LED excitation: Samples were excited with a 365 nm LED (using a TTi QL564P DC power supply) which was directed onto the sample using a liquid light guide (3.2 mW cm -2 incident on the sample). In all cases, photoelectrodes were illuminated from the front (i.e. from the solvent side, not through the FTO back contact) for two seconds. A software-controlled MOSFET connected to the power supply in series with the LED was used to switch on the LED in 2 second pulses whilst simultaneously triggering data acquisition. Optical data was collected using an NI-USB-6361 National Instrument DAQ card.
Laser excitation: Samples were excited from the front side with a single 355 nm laser pulse from a Nd:YAG laser (OPOTEK Opolette 355 II, 7 ns pulse width) at an incident intensity of 540 µJ cm -2 . The beam was transmitted through a liquid light guide before incidence on the sample. To get the 100 ns -10 s time range shown in the main text, each sample was excited several times with different timebase settings on the scope. The overlapping TPC traces were then stitched together. The TPC was monitored betwe en pulses to ensure that each sample was given sufficient time to return to its unexcited state between excitation pulses. Optical data was collected on the 5 µs -1 ms timescale using a Tektronics DPO-2012B digital phosphor oscilloscope after being passed through an optical transient amplifier (Costronics 2004). Optical data on the 1 ms -0.1 s timescale was collected using a NI-USB-6361 National Instrument DAQ card.

Product analysis
HPLC analysis. All analyses were conducted on a Waters Breeze system equipped with refractive index (RID-2414) and diode array UV-vis (λ = 210 and 254 nm) detectors. The 4-MBA oxidation product was identified and quantified with a C18 column at 40 °C column temperature. Samples were analyzed in the isocratic flow mode (flow rate 0.5 mL min -1 , H2O:MeCN). Glycerol oxidation products were analyzed with an Ion-Exclusion ROA-Organic Acid H+ (8%) column at 70 °C. Samples were analyzed in the isocratic flow mode (flow rate 0.5 mL min -1 , 0.005 M H2SO4 HPLC water). Calibrations were conducted with external standards for both substrates.
Ionic Chromatography (IC). IC was performed to identify and quantify formate and acetate formation for MeOH, EtOH and glycerol oxidation. The analysis was performed with a 882 Metrohm Compact IC Plus using 3.2 mM Na2CO3 and 1 mM NaHCO3 as an eluent. Calibration was conducted with external standards. NMR spectroscopy. 1 H NMR spectroscopy was used to identify oxidation products of ethylene glycol oxidation. NMR spectra were collected with a Bruker 400 MHz Neo Prodigy spectrometer, with deuterated water as solvent at room temperature.               Extended PIA/TAS/TPC discussion: Spectroelectrochemistry was first employed to identify the absorbance spectra of any charged species that might be present in the electrodes: NCN CNx, ITO: NCN CNx and Al2O3|ITO: NCN CNx. Figure S17.  Figure S17). In all four kinetic traces, a substantial proportion of the electron population decays slowly (halflives > 10 seconds) after the LED is switched off. This is consistent with other reports of long-lived, trapped photoelectrons in NCNterminated carbon nitrides. Whilst these electrons have previously been shown to reside in shallow, catalytically active trap sites. [17,18] Transient photocurrent data measured simultaneously with the optical data ( Figure S17c) shows that the slowly-decaying electron population observed in these photoanodes does not contribute to the photocurrent: whilst these electrons may exist primarily in shallow trap states, they are still too deeply trapped to be efficiently electrically extracted. The bias independence of the PIA electron kinetics in Figure S18.a confirms that electron extraction is inefficient in the bare NCN CNx sample. In contrast, the Al2O3|ITO: NCN CNx PIA signal show a strong dependence on the applied potential, being strongly quenched by an applied bias of 1.2 V, which is indicative of efficient electron extraction mediated by the ITO nanoparticles ( Figure S18b). Extended PEIS discussion: Figure S19a shows the Nyquist plot for the NCN CNX blank film as a function of the applied potential. It is apparent from this figure that there is only one semicircle, associated to one RC process. This RC process can be associated with a electron extraction process from NCN CNX to the FTO back contact, given the PEC behavior shown in Figure 1.a in the manuscript. On the contrary, the ITO ( Figure S19b) exhibits one high frequency semicircle and one low frequency semicircle, where the low-frequency semicircle shows a very high resistance that only starts decreasing at high applied potentials once dark current (charge injection) takes place (see dark onset potential in Figure S9). Interestingly, the high-frequency semicircle in ITO shown Figure S18b is independent of the applied potential suggesting that this process is not a charge transfer to the 4-MBA, but rather from ITO to FTO as observed previously in CoFe-prussian blue electrocatalysts on FTO. [19] Figure S19c shows, from -0.2 to 1.2 V vs RHE, 2 RC processes for the composite ITO: NCN CNx photoelectrode similarly to the ITO alone. It is apparent from this figure, that in the potential range from -0.2 to 1.2 V vs RHE, both high and low-frequency RC processes behave similarly in both ITO and ITO: NCN CNx electrodes the electron transfer from ITO to FTO dominates their JV curves. Thus, it can be concluded that the photoelectrochemical (PEC) behavior of the ITO: NCN CNx photoelectrode is determined by the efficiency of charge injection from the ITO to the FTO. Above 1.2 V vs RHE, coinciding with the photocurrent plateau behavior shown in Figure 2a, orange data, in the manuscript, only one semicircle, associated to charge transfer to the 4-MBA, takes place, suggesting that the PEC behavior is no longer limited by the ITO/FTO interface, but rather by the electron extraction from the CNx to the FTO. A similar behavior is observed for the Al2O3| NCN CNx:ITO photoelectrode, shown in Figure S19d, but with the transition from 2 to 1 RC processes limiting the PEC behavior at ~0.3 V vs RHE, suggesting that the role of the Al2O3 thin layer is to improve the electrical contact at the ITO/FTO interface.