Unraveling fundamental active units in carbon nitride for photocatalytic oxidation reactions

Covalently bonded carbon nitride (CN) has stimulated extensive attention as a metal-free semiconductor. However, because of the complexity of polymeric structures, the acquisition of critical roles of each molecular constituent in CN for photocatalysis remains elusive. Herein, we clarify the fundamental active units of CN in photocatalysis by synthesizing CN with more detailed molecular structures. Enabled by microwave synthesis, the as-prepared CN consists of distinguishable melem (M1) and its incomplete condensed form (M2). We disclose rather than the traditional opinion of being involved in the whole photocatalytic processes, M1 and M2 make primary contributions in light absorption and charge separation, respectively. Meanwhile, oxygen molecules are unusually observed to be activated by participating in the photoexcited processes via electronic coupling mainly to M2. As a result, such CN has a higher activity, which was up to 8 times that of traditional bulk CN for photocatalytic oxidation of tetracycline in water.

S-2   Table 1 Combustion elemental analysis   Supplementary Table 2 Centroid distance of electron-hole (D), overlap integral (Sr), index of separation degree (t) and coulomb attraction energy (Ec) of CN and O2@CN molecules.

Photoelectrochemical (PEC) experiments
The photoelectrodes were prepared by spreading aqueous slurries of various CNbased materials over 0.25 cm 2 of ITO glass substrate using adhesive tapes as spaces.
The suspension was preparation by grinding 8 mg of CN-based materials with 60 μL of 0.1wt.% Nafion. Then the photoelectrodes were annealed in air at 120 °C and kept at this temperature for half an hour. The photoelectrochemical measurements were performed in a conventional three-electrode system in a quartz tank with a platinum wire as the counter electrode and Ag/AgCl (saturated KCl) electrode as the reference electrode. Illumination through the ITO back-side illumination was used. The power density of the incident light was calibrated to 1000 W/m 2 at the surface of CN-based photoelectrodes by a 150 W Xe lamp (Beijing NBeT Co., Ltd.) with optical filter (>400 nm). All illuminated areas were 0.25 cm 2 . Photocurrent measurements were performed in a 0.1 M KCl solution at a scan rate of 10 mV/s. The current difference between the light and the dark was defined as the net photocurrent. 1

Photocatalytic oxidation of Azure B
Briefly, 100 mg of photocatalyst was added in a quartz tube (5×5×5cm, 20 mL) of 0.05 mg/mL Azure B aqueous solution (AB). Firstly, the suspension was stirred for 20 min in dark to ensure the establishment of adsorption equilibrium. Afterwards, the quartz tube was top-irradiated under full light by using a 300 W Xe lamp (CEL-HXUV300E, China) with a short-pass filter cutting off lights of wavelength less than 400 nm. 0.4 mL of suspension was extracted and centrifuged at certain time intervals.
The photocatalytic oxidation efficiency of AB at 604 nm was analyzed on the UV-vis spectrophotometer.

Recyclability of CNMW
After each photocatalytic oxidation of TC, the equal amount of TC was added for next recycling test.    As disclosed by the trapping techniques, the photooxidation of TC was predominately driven by the activation of O2 into O2 •and H2O2 via the reduction using the photogenerated electrons. In this sense, the higher CB position would be indicative of stronger oxidation ability. As calculated by using UV-Vis spectra and XPS VB scan spectra, the order of the oxidation ability followed as: CNMW-sol > CNMW-ins > bulk CN.
Nonetheless, it should be aware that except for the oxidation ability, the overall photocatalytic oxidation activity is also influenced by the light absorption, electron donating ability of substrates, dispersibility and other factors as well. From the SEM and TEM images, it can be seen that the stacked structure of bulk CN was retained in the as-prepared CNMW. Nonetheless, many macro-holes were found in the CNMW-ins, which was supposed to be generated by extracting the solvable CNMW-sol using the solvent. From another point of view, it also indicated that CNMW-ins and CNMWsol were seamlessly coupled. The surface area of CNMW (6.3 m 2 /g), and CNMW-ins (37.6 m 2 /g) were also measured by N2 absorption/desorption and compared with that of bulk CN (13.3 m 2 /g). It was found that the surface areas of them were far less proportional to the photocatalytic activity, indicating the surface area did not played a key role in this work. Besides, it should be mentioned that the morphology of the solid CNMW-sol largely depended on the precipitation method from the solution. Some decrease in photocatalytic activity was observed in Fig. S9a, but still fair, as remaining 74% after three cycles. It was presumably ascribed to the partial absorption of TC oxidation product on the surface of CNMW (the color of the catalyst gradually became greyish, see the photo in Figure S9b inset), rather than the photocatalytic corrosion of CNMW itself. To verify this assumption, we examined the XRD of the CNMW sample after the recycle. It was found that except for the disappearance of some minor peaks (*) and slight up-shift of (002) diffraction peak, the most representative The similar activity trends among different carbon nitride samples such as CNMW and bulk CN was also observed for the photocatalytic oxidation of other substrates, such as Azure B. Nonetheless, the photocatalytic oxidation activities also correlated to the electron donating ability of substrates and their surface properties, which appeals for more future investigations. To check whether H2O2 was produced or not, the CNMW dispersion (5 mg/mL) was irradiated for 40 min. After that, it was kept in dark for 30 min to eliminate the interference of reactive oxygen species with short lifetime. Horseradish peroxidase (HRP) and 3,3′,5,5′-Tetramethylbenzidine (TMB) was then added and incubated for 5 min after the removal of the catalyst by centrifugation. It was supposed that if H2O2 was present, colorless TMB would be catalytically oxidized into blue TMBox by HRP. 10 As shown in the UV-Vis spectra below, the typical absorbance of TMBox in case of CNMW was improved with respect to the control pure water, which could also be indicated by a visual color change from colorless to blue of the solution (see the inset photo). Thus, H2O2 was produced during the photocatalytic activation of O2.

S-14
S-15 distribution, indicating that M2 with an asymmetric structure is more likely to be oxidized/reduced than M1, and thus charge transfer occurs easily with O2. Consequently, M2 may be more conducive to the activation of O2, which is consistent with our TDDFT analysis of the electron-hole separation in the first excited state (see Supplementary   Figure 16).