Reduced graphene oxide supported C3N4 nanoflakes and quantum dots as metal-free catalysts for visible light assisted CO2 reduction

The visible light photocatalytic reduction of CO2 to fuel is crucial for the sustainable development of energy resources. In our present work, we report the synthesis of novel reduced graphene oxide (rGO)-supported C3N4 nanoflake (NF) and quantum dot (QD) hybrid materials (GCN) for visible light induced reduction of CO2. The C3N4 NFs and QDs are prepared by acid treatment of C3N4 nanosheets followed by ultrasonication and hydrothermal heating at 130–190 °C for 5−20 h. It is observed that hydrothermal exposure of acid-treated graphitic carbon nitride (g-C3N4) nanosheets at low temperature generated larger NFs, whereas QDs are formed at higher temperatures. The formation of GCN hybrid materials was confirmed by powder X-ray diffraction, X-ray photoelectron spectroscopy, Fourier transform infrared spectroscopy, field emission scanning electron microscopy, transmission electron microscopy (TEM), and UV–vis spectroscopy. High-resolution TEM images clearly show that C3N4 QDs (average diameter of 2–3 nm) and NFs (≈20–45 nm) are distributed on the rGO surface within the GCN hybrid material. Among the as-prepared GCN hybrid materials, GCN-5 QDs exhibit excellent CO2 reductive activity for the generation of formaldehyde, HCHO (10.3 mmol h−1 g−1). Therefore, utilization of metal-free carbon-based GCN hybrid materials could be very promising for CO2 photoreduction because of their excellent activity and environmental sustainability.


Introduction
The solar-light-assisted photocatalytic reduction of CO 2 into useful chemicals, such as HCOOH, HCHO, CH 4 , and CH 3 OH is one of the sustainable ways to address the issues of both global warming and the energy crisis [1][2][3][4][5][6]. So far, a variety of semiconductor photocatalysts, such as ZnO, TiO 2 , WO 3 , and CdS have been developed for the photoreduction of CO 2 [7][8][9][10]. However, poor separation of photo-induced electron-hole pairs and insufficient adsorption of CO 2 at the catalyst surface are crucial problems preventing effective catalyst performance and CO 2 reduction [11]. An ideal photocatalyst for CO 2 conversion should possess a narrow bandgap and good light-harvesting properties, proper conduction band (CB) and valence band (VB) edge positions, exhibit efficient charge separation, have a large surface area, and it must be cost effective. Considering the above factors, nontoxic metal-free catalysts, such as graphitic carbon nitride (g-C 3 N 4 ) and reduced graphene oxide (rGO) have received wide attention in recent years for CO 2 reduction and water splitting applications [12][13][14][15][16][17]. Both g-C 3 N 4 and rGO have a two-dimensional sheet structure with high surface area and possess appropriate band edges for CO 2 reduction. Also, g-C 3 N 4 and rGO are inexpensive and easy to synthesize.
Despite all these interesting properties, pure g-C 3 N 4 only weakly absorbs visible light due to its wide band gap and also has poor electrical conductivity [18]. An efficient way to increase the charge separation and electrical conductivity of g-C 3 N 4 is to modify it with rGO. Besides the structural and electronic modification of the g-C 3 N 4 material with rGO, another interesting strategy is to increase the number of catalytic active sites (pyridinic N, graphitic N, and edge amine groups) in g-C 3 N 4 [19]. This can be achieved by generating zero-dimensional (0D) quantum dots (QDs) and nanoflakes (NFs) of g-C 3 N 4 from 2D sheets by facile hydrothermal reactions. Recently, solar energy conversion using g-C 3 N 4 QDs has attracted significant attention [20][21][22]. Besides, g-C 3 N 4 has also been coupled with noble-metal-free compounds for higher catalytic activity [23][24][25][26][27]. Despite all these significant findings, there have been few studies focused on the improvement of the visible light absorption of g-C 3 N 4 /rGO hybrid materials [28][29][30][31] and g-C 3 N 4 QDs, and their CO 2 photoreduction ability has not yet been reported.
Hence, in our present study, metal-free hybrid catalysts consisting of rGO-supported C 3 N 4 (GCN) NFs and QDs are prepared by a hydrothermal method. The formation of the GCN hybrid is controlled over temperature during hydrothermal heating. The as-synthesized GCN hybrids are then characterized and applied to the photoreduction of CO 2 under visible light. The concentration of the photocatalytic product, formaldehyde (HCHO), is monitored spectrophotometrically using the Nash reagent, and also confirmed by gas chromatography.

Material characterization
The preparation of g-C 3 N 4 NFs and QDs from nanosheets and the formation of the GCN hybrid material are shown schematically in Figure 1. When GO is subjected to hydrothermal treatment with CN NFs and QDs, GO converts to rGO (Supporting Information File 1, Figure S1) and GCN hybrid materials are formed. The phase purity of the synthesized g-C 3 N 4 nanosheet, GCN-5, GCN-10, and GCN-20 hybrid materials are first examined by powder X-ray diffraction (PXRD). Figure 2 shows the PXRD patterns of pure g-C 3 N 4 nanosheets and GCN hybrids. A broad peak at 2θ = 27.5° is observed for the g-C 3 N 4 nanosheets, which is due to the interplanar stacking (002 plane) of aromatic C-N heterocycles present in g-C 3 N 4 [32]. However, the intensity of both (100) and (002) peaks is tremendously decreased for the nanosheets compared to the bulk C 3 N 4 and is also found to be shifted to a lower angle (Supporting Information File 1, Figure S2). For the GCN hybrid materials, the intensity of the (002) peak of g-C 3 N 4 nanosheets is significantly decreased. The shifting and decrease in intensity of the (002) peak indicates structural changes of the 2D g-C 3 N 4 nanosheets to 1D NFs and 0D QDs. The XRD peak intensity of a plane is largely dependent on its internal structure. When the nanosheet transforms into QDs/NFs by acid cutting and hydrothermal heating, it undergoes many structural changes (breaking of aromatic planes along (002) directions) internally. Thus, the size of the nanosheet is drastically decreased when it transforms into quantum dots. Consequently, the intensity of the planes is expected to largely decrease. In the GCN hybrid material, the peaks related to rGO are not observed in the hybrid materials, which may be due to the low amount of rGO incorporation.
Moreover, the peaks related to other impurities are not found in the pattern, indicating the formation of pure hybrid composite materials. The presence of rGO is confirmed by X-ray photoelectron spectroscopy (XPS) analysis, which is carried out to study the surface composition and the interactions and valence states of the elements present in the hybrid materials. The full XPS survey spectrum of the GCN-5 hybrid material reveals the presence of C, N and O, while no impurity elements are found ( Figure 3a). Figure 3b shows the fit to the C 1s peak of the  GCN-5 hybrid. The fitting of the C 1s spectrum shows four major deconvoluted peaks related to the carbon states of rGO and g-C 3 N 4 . The sharp peaks at binding energies of 285.04, 287.03, 288.52, and 289.6 eV observed in the C 1s spectrum correspond to C-C bonds in rGO, N-C=N/C-O bonds in g-C 3 N 4 and rGO, and C=O and O−C=O bonds in rGO, respectively [33,34]. The core-level N 1s profile shows (Figure 3c) three deconvoluted peaks at binding energies of 398.8, 400.6, and 401.8 eV, which are attributed to the sp 2 -hybridized N (C-N=C), the tertiary N, and the N-H group of g-C 3 N 4 , respectively, which are present in the GCN-5 hybrid material [35].
Interestingly, the N peak intensities are significantly decreased compared to those of the pure g-C 3 N 4 nanosheet (Supporting Information File 1, Figure S3), which is due to the breaking of most C-N bonds due to formation of NFs and QDs. Figure 3d shows a high-resolution XPS spectrum of O 1s present in the GCN-5 hybrid. The peaks at binding energies of 531.7 and 532.8 eV are attributed to carbonyl and epoxy C-O groups of rGO, respectively, which are still present in the rGO [34]. The XPS peaks of g-C 3 N 4 and rGO are shifted slightly to higher and lower binding energies in the GCN-5 hybrid, respectively, indicating possible charge transfer between g-C 3 N 4 and rGO in the heterostructure. Hence, the XPS results confirm the successful preparation of the composite and the existence of an interaction between rGO and g-C 3 N 4 inside the composite.
FTIR spectra further confirm ( Figure 4) the formation of C 3 N 4 NFs and QDs, as well as the structural changes of C 3 N 4 nanosheets. The peaks at around 3000-3110 cm −1 , 1200-1650 cm −1 and 810 cm −1 are due to the N-H stretching vibration, aromatic CN heterocycles, and the s-triazine ring of the g-C 3 N 4 nanosheet, respectively [36,37]. However, after acid treatment and hydrothermal heating, most of the peaks attributed to CN heterocycles have vanished, indicating the structural transformation of g-C 3 N 4 nanosheets to C 3 N 4 NFs and QDs. Besides, a sharp peak at 1380 cm −1 is observed, which can be ascribed to -C-O stretching vibrations of carboxylate groups, which are formed due to the breaking of some C-N bonds of triazine rings and their oxidation to carboxylic groups in the C 3 N 4 NFs and QDs [19]. After the introduction of rGO to form GCN-5, the characteristic peaks of C 3 N 4 NFs and QDs (CN-5, CN-10, and CN-20) could still be observed along with some smaller intense peaks (epoxy group, 1150-1250 cm −1 ) attributed to rGO [38]. The peaks related to O-H groups in the GCN-5 sample are also noticed at around 3200 cm −1 due to the presence of a small percentage of rGO, which is shifted to lower wavenumber, indicating H-bonding interactions between CN-5 and rGO. The structural transformation of C 3 N 4 nanosheets to QDs is also verified by UV-visible spectroscopy (Supporting Information File 1, Figure S4). The C 3 N 4 nanosheet exhibits a sharp peak at ≈316 nm, which is shifted to ≈285 nm (higher energy) after the formation of QDs (CN-5 QD) due to the breaking of C=N links [39]. Hence, XRD, XPS, IR and UV studies show the successful structural transformation of g-C 3 N 4 nanosheets to QDs and the presence of possible organic functionalities in the GCN-5.
The size, morphology, and distribution of the synthesized NFs and QDs were investigated by TEM and FESEM studies. Figure 5a,b shows TEM images of CN-5 and GCN-5 samples. It is observed that after the acid treatment and hydrothermal heating of g-C 3 N 4 nanosheets at 190 °C for about 5 h, the g-C 3 N 4 nanosheets are transformed into small nanoparticles along with QDs about 2-3 nm in diameter (Figure 5a). However, with a decrease in hydrothermal heating temperature (130 °C) and an increase in heating time (10 h and 20 h), the g-C 3 N 4 nanosheets then produced a flake-like substance with an average size of 20-45 nm (Figure 5c and 5e). However, all particles are not completely transformed to QDs; there are also some larger particles along the QDs, as indicated by TEM. The TEM images clearly show (Figure 5b,d,f) that these NFs and QDs are well distributed and decorated with a thin rGO layer after the introduction of GO into the hybrid. Particle size distributions of the CN samples are also given in Figure 6. It is noticed that the average particle diameters of 45, 20, and 2 nm are obtained for CN-10, CN-20, and CN-5 samples, respectively. It is also observed that the size of the particles for 20 h heating (at 130 °C) is smaller compared to 10 h heating. However, for 20 h of heating, some nanoflakes are found to be agglomerated, which is probably due to the extended heating time. However, for 10 h of heating, more uniform particles/ flakes are observed as indicated by the TEM results. The TEM results suggest that the formation of QDs from g-C 3 N 4 is temperature dependent. To further verify this, TEM images of the samples were taken at two different heating temperatures (150 and 170 °C) within the range 130 to 190 °C.
The results confirm that the limiting temperature at which the nanoflakes are converted to quantum dots is around 170 °C   Figure 7). With an increase of temperature from 130 to 170 °C, the particle size starts to become smaller, and at above 170 °C, QDs are predominantly formed.
When a g-C 3 N 4 nanosheet is subjected to acid etching followed by long term (10-20 h) hydrothermal heating at low temperature (130 °C), the g-C 3 N 4 sheet breaks into pieces in different orientations and generates flake-like shapes of the material. Upon further increase of the heating temperature, the process proceeds very fast and cuts the flakes into dot-like structures (i.e., QDs) even within a small span of time (5 h). During the acid treatment process, some C-N bonds of the s-triazine units of the g-C 3 N 4 sheet are oxidized and oxygen-containing carboxylate functional groups are generated at the edge and on the basal plane [19] as indicated by FTIR spectra (Figure 4). This results in orientational cleavage of g-C 3 N 4 nanosheets and generation of some spherical particles of 300-500 nm as indicated by FESEM. Finally, CN QDs and NFs are formed after the hydrothermal treatment of these spherical particles at 130 to 190 °C. The surface of these CN QDs finally contains amino and carbonyl-functional groups, as indicated by XPS also.
The TEM results are also in good agreement with the FESEM results. The FESEM image (Figure 8) clearly shows the structural transformation of g-C 3 N 4 nanosheets into spherical parti-cles with an average diameter of 300-500 nm after the acid treatment (Figure 8a and 8b).  Figure 8f); however, the exact size of the CN-5 QDs has been confirmed from the HRTEM results. Figure 9 exhibits HRTEM images of the GCN-5 QDs. It can clearly be seen that CN-5 QDs are decorated (marked by circle and arrows) onto the GO surface with an average diameter of 2-3 nm. A clear lattice spacing of 0.336 nm is also observed for the CN-5 QDs, which corresponds to the (002) plane of hexagonal g-C 3 N 4 , indicating crystalline nature of the QDs [40]. Hence, TEM, HRTEM, and FESEM studies confirm the morphology and size of the NFs and QDs and also confirm the presence of rGO in the hybrid material.
The band gaps of the prepared NFs and QDs are also calculated from UV-visible diffuse reflectance spectra (Supporting Information File 1, Figure S5 where E CB is the CB edge potential; X is the geometric mean of the absolute electronegativity of the constituent atoms in the semiconductors, which is defined as arithmetic mean of the atomic electron affinity (E EA ) and first ionization (E ion ) energy; E e is the energy of free electrons on the hydrogen scale (≈4.5 eV vs NHE); E g is the band gap of semiconductors. The conduction and valence band potential value for GCN-5 are −1.01 and 1.05 eV, respectively, and is found to be lower than any CN NFs. For all other QDs/NFs and hybrid materials the CB values are also found to be more positive (lower) than g-C 3 N 4 sheet [32].

Photoreduction of CO 2
For the photocatalytic reduction of CO 2 , catalytic amounts of GCN hybrid materials (0.5 g/L) are added to an aqueous Na 2 CO 3 solution (0.01 M) in a round bottom flask capped with a rubber septum. The solution is then illuminated by 100 W halogen lamps. A sample (5 mL) of the photo-reacted solution is withdrawn after a measured interval and mixed with Nash's reagent. The solution developed a bright yellow color due to the formation of a 3,5-diacetyl-1,4-dihydro-2,6-lutidine (DDL) complex between HCHO and the Nash reagent. The optical intensity of this complex is then measured at λ max = 412 nm using a UV-vis spectrophotometer, and the final concentration of the produced HCHO was measured using a standard calibration curve (Supporting Information File 1, Figure S6). The UV-visible absorption spectra of the product used for the determination of HCHO in the presence of different GCN hybrids are shown in Figure 10.
It is observed from the UV spectra that with increasing irradiation time, the peak of the DDL complex also increased due to effective production of HCHO. The production rates of HCHO for GCN-5, GCN-10, and GCN-20 are found to be 10.1, 7.1, and 6.5 mmol g −1 h −1 , respectively ( Figure 11). Compared to GCN-10 and GCN-20, GCN-5 exhibited the highest production of HCHO (680 μmol in 4 h) under visible light, and the photocatalytic efficiency of GCN-5 towards the reduction of CO 2 is found to be better than many other catalysts reported previously in the literature [41][42][43]. The CO 2 photoreduction efficiency of the GCN-5 sample was also measured by apparent quantum yield (AQY) measurements, and was found to be 22.3%. The AQY value of GCN-5 is higher than many earlier reports [44][45][46]. It is also clearly observed that in the presence of rGO, the photoactivity of CN-5, CN-10, and CN-20 catalysts is significantly enhanced (5 to 6 fold) for the production of HCHO, indicating that rGO provides a very active catalytic surface for the QDs and NFs, which increases the effective charge separation within the hybrids. Interestingly, the results also show that CN-5 exhibited almost two times higher reduction capacity of CO 2 compared to CN-10 and CN-20. This is because of the QD size of CN-5, which increases its surface area (231.3 m 2 /g) compared to CN-10 (207.8 m 2 /g) and CN-20 (195.4 m 2 /g), thus increasing the number of catalytic active sites. The lower surface area of CN-20 was responsible for its poorer activity compared to CN-10, which is probably because of the agglomeration of CN-20, as suggested by TEM. The photoreduction ability of CN-5 is found to be higher than that previously reported for traditional C 3 N 4 sheets [47,48].
We also carried out the reduction test with samples of different amounts of rGO from 2.5 to 7.0 mg with CN QDs (data not shown). However, the maximum activity is only observed for 5 mg rGO, and hence, this quantity was used for all cases in the present study. The higher loading of rGO yielded less activity, which is probably due to less light penetration in the sample since a higher concentration of rGO suspension blocks the effective light.
For a comparative study, the CN samples are also prepared at 130 °C for 5 h and at 190 °C for 10 h. The TEM study suggests (Supporting Information File 1, Figure S7) that there is no significant change in size of the CN QDs at 190 °C for 5 h and 10 h. The photo-reductive capacity of the CN at 190 °C for 10 h is also found to be almost identical to 190 °C for 5 h. This is possibly because of the similar size of the QDs formed at this temperature. However, for the CN sample at 130 °C of 5 h, it is observed that the size and morphology of the produced NFs are almost similar with CN-10, but a small portion of the CN nanosheet is still present with the NFs, indicating that there might be an incomplete transformation of the nanosheet to NFs. This is probably due to the reduced heating time at this low temperature; however, this did not affect the photoreduction capacity of these CN-NFs (130 °C of 5 h), which was found to be nearly similar to CN-10 (data not shown).
The stability of a catalyst is of paramount importance for practical applications. Hence, the prepared GCN-5 sample was recycled several times to check its reduction performance. It is observed that no significant loss in activity is observed even after the fourth cycle ( Figure 12). Hence, the prepared metal-free GCN-5 is highly stable and can be reused for the photoreduction tests.
The photoreduction of CO 2 to HCHO is also verified by gas chromatography mass spectrometry (GC-MS). A sample is examined for GC analysis after 4 h of photoirradiation with GCN-5 under identical conditions, and the spectrum shows a major peak related to HCHO and a minor peak for trace amounts of methanol, and no other peaks are found (Supporting Information File 1, Figure S8a), confirming the successful reduction of CO 2 to HCHO.
Three sets of control experiments were also conducted, and an isotopic experiment to determine the source of carbon in the product: i) with Na 2 CO 3 and GCN-5 under same condition, but without CO 2 , ii) absence of light under identical condition, iii) without GCN-5, and iv) an isotopic experiment under same conditions using 13 CO 2 (99.0% purity) and the irradiated sample are analyzed by GC-MS. In first three cases, no noticeable change in color is observed with Nash reagents, indicating no formation of HCHO. However, for the isotopic experiment, the GC-MS spectrum shows (Supporting Information File 1, Figure  S8b) the major m/z peaks at 31 and 33, which are observed due to H 13 CHO, and 13 CH 3 OH, respectively.
The photoreduction mechanism of CO 2 to HCHO in the presence of the GCN-5 catalyst under visible light illumination is shown in Figure 13. The CN QDs absorb visible wavelengths of sunlight due to their appropriate band edge potential value, thus exciting the electrons. These excited electrons are then transferred to the rGO surface, because it has a lower conduction band (CB) potential than CN-5, which improves the charge separation process in the GCN hybrid [28]. Finally, these electrons are utilized for the reduction of CO 2 to HCHO in the presence of water reductant. At the valence band (VB), water is oxidized by photo-generated holes (h + ), and generates protons (H + ion), which initiate the CO 2 reduction process. Since the reduction potential of CO 2 to HCHO (E 0 (CO2/HCHO) = −0.48 eV) matches with the CB potential of rGO, the CB electrons in rGO can easily reduce the CO 2 [49] with the help of protons.

Conclusion
Reduced graphene oxide supported C 3 N 4 NF and QD hybrid (GCN) materials have been successfully synthesized via a sol-gel and hydrothermal method and are characterized in this work. The formation of g-C 3 N 4 NFs (20-45 nm) and QDs (2-3 nm) can be controlled by varying the temperature during hydrothermal heating. XRD, TEM, IR, UV and XPS studies confirmed the structural transformation of the nanosheet to QDs, and also the presence of rGO in the GCN hybrid. The as-prepared CN-5 (QD) exhibited better photoreduction of CO 2 and generation of HCHO compared to the CN-10 and CN-20 NFs due to its quantum size. Additionally, the production of HCHO is improved almost five-fold (10.3 mmol h −1 g −1 ) in the presence of rGO with CN-5. It is expected that the QD size of the CN-5 sample promotes a higher number of catalytic active sites and enhanced light absorption, and the rGO provides better charge separation, which further enhances its photo-reductive capacity. Therefore, metal-free GCN hybrid materials could be a potential candidate for CO 2 photoreduction and HCHO (fuel) generation because of their excellent activity, stability and environmental sustainability.
Preparation of g-C 3 N 4 nanoflakes and quantum dots (CN-5, CN-10, and CN-20) Bulk g-C 3 N 4 was first prepared using urea as a precursor according to our previously described method [32]. In brief, an appropriate amount of urea (6.0 g) was placed in an alumina crucible and was calcined at 550 °C for 2 h at a rate of 5 °C/min under air. The as-prepared bulk g-C 3 N 4 was further exfoliated by thermal oxidation at 500 °C for 2 h at a rate of 5 °C/min to obtain g-C 3 N 4 nanosheets. The obtained 2D g-C 3 N 4 nanosheets were then subjected to acid etching followed by hydrothermal treatment to obtain 0D g-C 3 N 4 QDs [19]. In brief, the g-C 3 N 4 nanosheets (0.5 g) were oxidized with concentrated H 2 SO 4 (50 mL) and HNO 3 (100 mL) for 4 h under ultrasonica-tion. A clear solution was formed, which was diluted with deionized (DI) water (100 mL) to produce a cloud-like colloidal suspension of g-C 3 N 4 . The products were filtered through a 0.4 μm microporous membrane to remove the acid, carefully redispersed in deionized water under ultrasonication, and then transferred to an autoclave (90 mL) and heated at 130-190 °C for different times (5, 10, and 20 h) to yield g-C 3 N 4 NFs and QDs. The product obtained after 5 h of heating at 190 °C is denoted as CN-5, and those obtained after 10 and 20 h heating at 130 °C are denoted as CN-10 and CN-20, respectively.
Preparation of rGO@g-C 3 N 4 nanoflake/quantum dot hybrid materials (GCN) GO was synthesized by the modified Hummers' method [50]. The as-prepared GO (5 mg) and CN NF/QDs (95 mg) were dispersed in 100 mL ethanol, and the mixture was sonicated for 30 min to form a homogeneous suspension. After that, the suspension was transferred into a 90 mL Teflon-lined autoclave and heated at 190 °C for 2 h. The resulting black rGO/g-C 3 N 4 NFs/QDs (GCN) hybrid materials were collected by centrifugation and then washed with distilled water several times, and finally dried under vacuum at 50 °C. The final hybrid products are denoted as GCN-5, GCN-10, and GCN-20.

Photocatalytic reduction of CO 2
Photocatalytic reduction of CO 2 was carried out by a spectrophotometric method [51]. GCN hybrid material (15 mg) was placed in 30 mL of an aqueous Na 2 CO 3 solution (0.01 M) in a 50 mL two-necked round bottom (RB) flask which was sealed by a rubber septum. The solution was bubbled with CO 2 (10 mL min −1 ) for 30 min to remove any dissolved oxygen. The reaction mixture was then illuminated by 100 W visible halogen lamps for 4 h. A sample (5 mL) of the mixture was collected every hour. The reduction rate was determined by measuring the concentration of the reduction product (i.e., formaldehyde), the amount of which was determined by a photospectrometric method using Nash's reagent (prepared by 15 g ammonium acetate (2 M), 0.3 mL acetic acid (0.05 M), and 0.2 mL acetylacetone (0.02 M) in 100 mL water) [52]. The reagent was placed in the dark to avoid any photochemical reaction. The produced HCHO solution (0.5 mL) was then added to 5 mL of Nash's reagent. Then the mixture was heated for 5-10 min in a water bath at 60 °C. The solution developed a bright yellow color, and the optical intensity was then measured at λ max = 412 nm using a spectrophotometer. The concentration of produced HCHO was determined from a standard calibration curve.
The photoactivity usually depends on the quantity of catalyst and light intensity, therefore the CO 2 photoreduction efficiency can also be measured by the apparent quantum yield (AQY), where AQY can be calculated according to Equation 1. The number of electrons required (4 e − ) to obtain each HCHO is also considered, and the intensity of the light is measured by following standard potassium ferrioxalate method [53]. (1)

Characterization techniques
PXRD was carried out using an X'pert Pro PANalytical instrument (Cu Kα radiation, λ = 0.154 nm) within a 2θ range of 10-60°. XPS was conducted using a Thermo Kα XPS (Thermo Fisher Scientific). FESEM was carried out using an S-4800, Hitachi Ltd. instrument (acceleration voltage, 5 kV), and TEM was conducted with a Tecnai G 2 F20 S-TWIN microscope. FTIR spectra were recorded with a Perkin Elmer Spectrum 100 spectrometer, and absorbance spectra were measured using a UV-visible spectrophotometer (Thermo Fisher Scientific). Gas chromatography coupled with mass spectrometry (GC-MS) was carried out on an Agilent 7890A instrument with both a thermal conductivity detector and a flame ionization detector and using helium as a carrier gas.

Supporting Information
Supporting Information File 1 Additional figures.