Water‐Soluble Polymeric Carbon Nitride Colloidal Nanoparticles for Highly Selective Quasi‐Homogeneous Photocatalysis

Abstract Heptazine‐based polymeric carbon nitrides (PCN) are promising photocatalysts for light‐driven redox transformations. However, their activity is hampered by low surface area resulting in low concentration of accessible active sites. Herein, we report a bottom‐up preparation of PCN nanoparticles with a narrow size distribution (ca. 10±3 nm), which are fully soluble in water showing no gelation or precipitation over several months. They allow photocatalysis to be carried out under quasi‐homogeneous conditions. The superior performance of water‐soluble PCN, compared to conventional solid PCN, is shown in photocatalytic H2O2 production via reduction of oxygen accompanied by highly selective photooxidation of 4‐methoxybenzyl alcohol and benzyl alcohol or lignocellulose‐derived feedstock (ethanol, glycerol, glucose). The dissolved photocatalyst can be easily recovered and re‐dissolved by simple modulation of the ionic strength of the medium, without any loss of activity and selectivity.

μs (45° flip angle) corresponding to 49 kHz, with 1 H SPINAL-64 decoupling at 80 kHz, recycle delay (RD) was 60 s. The 1 H-13 C CPMAS spectra were acquired at a MAS rate of 12 kHz using a 1 H 90° pulse length set to 3.2 μs (78 kHz); the CP step was performed with a contact time (CT) of 3500 μs using a 50−100% RAMP shape pulse on the 1 H (66 kHz) channel and using a 45 kHz square pulse on the 13 C channel; RD was 5 s. During the acquisition, a SPINAL-64 decoupling scheme was employed using a pulse length for the basic decoupling units of 6 μs at a rf field strength of 80 kHz. 1 H-15 N CPMAS spectra were typically acquired with the following parameters: MAS rate of 5 kHz, 1 H 90° pulse length of 4 μs employing a rf field strength of 63 kHz, the CP step was performed with a CT of 2000 μs using a 50−100% RAMP shape on the 1 H (57 kHz) channel and using a 34 kHz square shape pulse on the 15 N channel; decoupling with SPINAL-64 using a pulse length of 6.8 μs using a rf field strength of 63 kHz was applied, RD was 5 s. All spectra were acquired at ambient probe temperature. MagicPlot Pro software was used for the deconvolution of the chosen spectra. The R 2 values for fitting curves were always higher than 0.97.
In order to perform X-ray photoelectron spectroscopy (XPS), the powder samples K-PHI-S and K,Na-PHI were placed on top of a conductive double-sided carbon tape (Ted Pella Inc., USA) and fixed to the XPS sample holder.
The measurements were performed using a UHV Multiprobe system (Scienta Omicron, Germany) with a monochromatic X-ray source (Al Kα) and an electron analyzer (Argus CU) with 0.6 eV energy resolution. Charge compensation during data acquisition was realized by an electron flood gun (NEK 150, Staib, Germany) at 6 eV and 50 μA. The background was subtracted and spectra were calibrated using the C1s peak (284.6 eV) before undergoing fitting using Voigt functions (30:70). A scanning electron microscope JEOL-6610LV equipped with an Oxford Instruments EDX detector was used to collect SEM images of the dry poly(heptazine imide) samples and determine their elemental composition. Particle size distribution of the water-soluble poly(heptazine imide) samples and their zeta-potentials were determined by means of dynamic light scattering (DSL) and electrophoretic mobility methods using a Zetasizer Nano ZS (Malvern Instruments Ltd., UK) operating at a detection angle of 173.5°.
Diffusive-reflectance UV-vis spectra of solids were taken by a Shimadzu UV2600 UV-vis spectrophotometer. UVvis spectra of the poly(heptazine imide) nanoparticles solutions were recorded using a Cary 60 (Agilent Technologies) spectrophotometer. Cyclic voltammogram of the K,Na-PHI material was recorded in the dark in acetonitrile solution of tetrabutylammonium hexafluorophosphate (TBAFP 6 , 0.1 M) using a SP-300 BioLogic potentiostat (Caix, France) and a three-electrode cell composed of a platinum counter electrode and an Ag/AgCl (sat. KCl) reference electrode. The working electrode was prepared by drop casting of the K,Na-PHI nanoparticles solution on a glassy carbon substrate followed by gelation by adding conc. HCl. The potential values were recalculated and reported with respect to the normal hydrogen electrode (NHE). The images were recorded using the SALVE at 80kV acceleration voltage (where indicated) and the Philips CM20 at 200kV, both in HRTEM mode.
The liquid sample was diluted (where indicated) in purified water (10 µl of sample in 1 mL of water), then drop cast onto copper TEM support grids with holey carbon film. The undiluted liquid samples were prepared by dipping of the TEM grid into the solution and drying. The solid samples were sonicated in ethanol and then drop cast onto TEM grids.

Photocatalytic measurements
Selective photocatalytic oxidation reactions of 4MBA to 4MBAL and benzyl alcohol to benzaldehyde were carried out for the K-PHI-S, K-PHI, Na-PHI, and K,Na-PHI samples, CN was used as a reference. For this purpose, the solid samples and the solutions of the poly(heptazine imide) nanoparticles were suspended in 20 mL of water or water-acetonitrile solution of 4MBA (or benzyl alcohol) (0.1 mmol). The pH of water solutions or water-acetonitrile mixture was in the range of 6.5-7.0 and did not undergo noticeable changes after irradiation. The reaction mixture was irradiated for 4 h by a UV LED emitting at 365 nm and producing a photon flux of 40 mW cm -2 . The control experiments were carried out in the absence of any photocatalyst under UV-irradiation and in the presence of the K,Na-PHI and CN samples in dark conditions. However, no noticeable reaction took place in such conditions, as the changes in the 4MBA or benzyl alcohol concentrations were below 1 %, and no H 2 O 2 was detected after 4 h.
The photocatalytic oxidation of 4MBA in the absence of air was performed under the optimal conditions (40 % vol. acetonitrile in the reaction mixture), while purging Ar gas through the system. The samples of 0.1 mL were taken every 30 min and analyzed by an HPLC Shimadzu LC-10ADvp equipped with a UV detector SPD-20A and a reversed phase C18 Nucleosil column. Acetonitrile/water (30/70 vol.) was used as a mobile phase at a flow of 1.2 mL min -1 and a column temperature of 60 °C. 4MBA to 4MBAL were detected and quantified at a wavelength of 240 nm, while benzyl alcohol and benzaldehyde were measured at 220 nm. Considering a quasi-homogeneous nature of the irradiated 4MBA (or benzyl alcohol) solutions with the poly(heptazine imide) photocatalysts and the incompatibility of the used HPLC column with some of the organic solvent immiscible with water, which could possibly be used for the organic compounds extraction, the following probe preparation was implemented. The photocatalyst was separated from the reaction medium by means of a salting-out approach. In details, to 0.1 mL of the sample, 0.2 mL of 1M NaCl solution, and 0.8 mL of acetonitrile were added, the salting-out effect resulted in the two-phase system with the photocatalyst in the aqueous phase and 4MBA, 4MBAL (or benzyl alcohol and benzaldehyde) extracted to the organic phase, which was subsequently analysed. The respective volumes of the additives for the probe-preparation were adjusted for the case of acetonitrile/water solutions, as to maintain the H 2 O to acetonitrile ratio in the probe constant. The calibration was carried out in the same way giving R 2 values higher than 0.999 for the substrate and the product. The recyclability test was made possible by the precipitation of the photocatalyst after each cycle by adding 0.25 g of NaCl to the reaction mixture in water solution and 0.125 to the reaction mixture in water/acetonitrile medium, following by its centrifugation and re-dispersion in a fresh solution of 4MBA. The selectivity (S) of the reaction was determined by the formula: where C 0 (4MBA) is the initial concentration of the substrate, C(4MBA) is the substrate concentration at a given reaction time, and C(4MBAL) is the concentration of the reaction product at a given reaction time. Considering a quasi-homogeneous nature of the carried out photocatalytic reaction, the estimation of the quantum yield of the process is more reliable than for the case of heterogeneous systems. For this, the reaction mixture in water(60%)/acetonitrile(40%) medium containing 4MBA (5mM) and the K,Na-PHI photocatalyst having absorbance of 0.1 at 365 nm were put in a quartz cuvette and irradiated with a 365 nm emitting LED for 2 h. After the reaction the concentration of the produced 4MBAL was analysed using the above described HPLC method.
The quantum yield (Q) of the process was calculated considering that two electrons are required to be withdrawn from the substrate (4MBA) to oxidize it to 4MBAL (1).
The photon flux was determined by the formula: where Δ is the light power absorbed by the system (W m -2 ), q is the charge of electron (C) and E is the energy of irradiation (eV).
The quantum yield was determined as: where n(4MBAL) is the number of moles of 4MBAL produced, N A is the Avogadro number, t is the irradiation time in s, and S is the irradiated area of the reactor (m 2 ) Hydrogen peroxide concentration was estimated photometrically. Sample of 5 mL of the reaction solution was withdrawn after 4 h of irradiation, to that volume 1 mL of TiOSO 4 solution (Ti 1 wt%) in sulfuric acid was added.
The photocatalytic hydrogen peroxide production using model lignocellulose-derived substrates such as ethanol, glycerol or glucose was carried out in the same way as that of benzylic alcohols oxidation, however only H

XRD analysis
The XRD patterns of the K-PHI-S sample series, obtained at different loadings of KOH in the melt, show that a part of melamine present in the precursor does not undergo the condensation to carbon nitride polymers, which is due to the fact that KOH is solid in this temperature range and it does not provide a liquid medium for melamine dissolution and condensation (Fig. S3 Left). Although the material recovered after the synthesis in KOH (K-PHI-S) was insoluble in water, its purification by washing and centrifugation or filtering was complicated owing to its high dispersibility in it. This is why the solid was dialysed, which allowed to achieve its higher purity (Fig. S3 Left).

TG-DTA analysis
The maximum of thermal decomposition rate of the K-PHI-S and K,Na-PHI samples is observed at about 690 °C (Fig. S6). A clear difference of thermal behaviour of the studied materials could be seen at high-temperature range.
K-PHI-S loses 100 % of its initial mass by 800 °C, which it is not the case of K,Na-PHI retaining about 18 % of its initial mass at the same temperature (Fig. S6). The decomposition of the samples might result in the formation of potassium and sodium carbonates, hydroxides or oxides which are all not volatile compounds, however they melt at temperatures above 800 °C and then can be carried away by the gas flow (Fig. S6). Figure S6. Thermoanalytical curves of the K-PHI-S and K,Na-PHI samples decomposition in O2 (50 mL min -1 ).  The 13 C direct excitation MAS spectra provide a quantitative estimation of the PCN species present in the studied samples (Table S3 and Fig. S7). The peak area of resonance 3 (associated to poly(heptazine) units) is used to estimate the relative amount of nitrile, C≡N (1), secondary amines, N=C-NH (4), primary amines, N=C-NH 2 (5), and oxygenated functional groups, N=C-O -(6), between samples K,Na-PHI and K-PHI-S, by taking the ratios 1:3, 4:3, 5:3 and 6:3, respectively (Table S3). The 6:3 ratio is similar in both materials (cf. 0.20 vs. 0.24) indicating that the relative amount of the C-Ospecies is nearly identical for both. In contrast, the number of -NH 2 groups (5:3 ratio) in the water-soluble K,Na-PHI decreases with respect to K-PHI-S (cf. 0.20 vs. 0.50, respectively), suggesting that the KOH/NaOH medium promotes a higher degree of condensation of the -NH 2 groups. Consequently, the formation of heptazine species (Fig. S8) is favoured, resulting in an increased number of N-C=N (3) carbons in K,Na-PHI as compared to K-PHI-S (Table S3, Fig. S7). The lower concentration of terminal -NH 2 moieties in K,Na-PHI will likely lead to an overall decrease of the number of hydrogen bonds in the polymer network, thus exerting a direct influence on aggregation and solubility.  Figure S9. (a, b, c, d) and (e, f, g, h) XP C1s, N1s, O1s and Na1s spectra obtained for the K-PHI-S and K,Na-PHI samples, respectively. is far superior to the theoretically possible for the fully condensed polyheptazine structure (ca. 0.16:1). We attribute this peak also to the contribution from -NHx bearing triazine fragments such as melamine. 11.12 Theoretical study of K,Na-PHI Figure S10. The salting out effect of NaCl addition on the double layer of the K,Na-PHI nanoparticles and their subsequent precipitation.

Elemental analyses
At low ionic strength, the surface of PHI particles is covered with a hydration layer of water molecules, and the electric double layer extends over a relatively large distance, making the inter-particle van-der-Waals attractive forces weak. With increasing Na + concentration, due to the addition of NaCl, the hydration layer will gradually be stripped, disturbed as the water molecules hydrate preferentially the Na + cations; at the same time, the double layer is compressed due to the more effective screening by Na + ions. These effects result in stronger attractive particle-particle van-der-Waals interaction, leading to the coalescence of K,Na-PHI nanoparticles and their precipitation. The process is reversible: after centrifugation and removal of concentrated NaCl solution, the addition of water yields the PHI colloidal particles fully soluble again. M + stands for Na + or K + .
The model system used for calculations is shown in Fig. S11 and is based on our experimental characterization; it exhibits a higher degree of condensation than typical carbon nitride model systems based on a periodic arrangement of melon strings but leaves enough boundaries to accommodate some functional groups to represent different N or O containing moieties.
Results of the solvation stabilisation study using the GLSSA13 implicit solvation model are summarized in Table   S5. The stability of the model system in water, acetonitrile, and methanol is compared. Two main trends can be extracted from this data: (i) the water environment appears to stabilise the model system more strongly than the other two polar solvents, and (ii) stabilisation improves as a function of the number of oxygen-containing moieties. The four -NH 2 groups are subsequently exchanged for -OH for the solvation study.  Figure S12. DLS particle size distribution of the poly(heptazine imide) samples.    [a] The oxidation rate is quantified considering a pseudo-first order model with R 2 always ≥ 0.98