Photocatalytic Reduction of CO 2 over Iron-Modiﬁed g-C 3 N 4 Photocatalysts

: Pure g-C 3 N 4 sample was prepared by thermal treatment of melamine at 520 ◦ C, and iron-modiﬁed samples (0.1, 0.3 and 1.1 wt.%) were prepared by mixing g-C 3 N 4 with iron nitrate and calcination at 520 ◦ C. The photocatalytic activity of the prepared materials was investigated based on the photocatalytic reduction of CO 2 , which was conducted in a homemade batch reactor that had been irradiated from the top using a 365 nm Hg lamp. The photocatalyst with the lowest amount of iron ions exhibited an extraordinary methane and hydrogen evolution in comparison with the pure g-C 3 N 4 and g-C 3 N 4 with higher iron amounts. A higher amount of iron ions was not a beneﬁcial for CO 2 photoreduction because the iron ions consumed too many photogenerated electrons and generated hydroxyl radicals, which oxidized organic products from the CO 2 reduction. It is clear that there are numerous reactions that occur simultaneously during the photocatalytic process, with several of them competing with CO 2 reduction. in-plane structural packing motif of the tri-s-triazine units with an interplanar distance of d = 0.675 nm [34,35]. The strong peak at 27.5 ◦ corresponds to the (002) plane, which is attributed to the interlayer stacking of aromatic rings with a distance of d = 0.324 nm [36,37]. After the modiﬁcation, no additional peaks corresponding to iron are observed, which was probably due to the low initial amount of iron nitrate. Furthermore, there are no signiﬁcant changes to the position or the intensity of the peaks, indicating that the crystalline structure of the materials is preserved [38]. region corresponds to the stretching mode of all of the OH-containing species and N-H stretching from the residual amino groups [40,41]. The shoulder located at 1630 cm − 1 represents the H-O-H bending mode of all of the water molecules present at the surface of the material. The peaks in the 1628–1232 cm − 1 region correspond to the characteristic stretching of the C-N heterocycles, including the trigonal N-(C) 3 and bridging H-N-(C) 2 units [42]. Finally, the sharp absorption peak at 805 cm − 1 can be attributed to the breathing mode of the triazine units [43]. The FT-IR spectra of the modiﬁed materials demonstrate no signiﬁcant changes. More speciﬁcally, there is no observable shift in the peaks, indicating that the chemical bonding of the main g-C 3 N 4 network is unaffected [38].


Introduction
The most pressing problems of the present time are undoubtedly the increasing energy requirements that are closely connected to fossil fuel combustion and the increasing concentration of greenhouse gases in the atmosphere. These problems are related, and finding a new clean source of energy is becoming increasingly important. Developed countries are trying to regulate greenhouse gas emissions; for example, the European Union has a long-term goal to reduce its carbon footprint. The first step towards this goal was to reduce greenhouse gas emissions by 20% (from 1990 levels) by 2020. This goal was successfully achieved, and emissions were reduced by 23% in 2018. However, a new legislation was put in place and added to the Paris Agreement. This legislation aims to reduce greenhouse gas emissions by 40% (from 1990 levels) by 2030. According to Energy in figures-Statistical Pocket Book 2019 issued by European Union, world energy production is increasing every year, and it is predicted that it will continue to rise in the years to come [1]. The highest annual increase of world energy production still comes from solid fossil fuels and natural gas compared to renewable or nuclear sources. Fossil fuel reserves are not infinite, and when considering the rate at which the world's energy consumption is increasing, it is clear that more focus has to be placed on finding new sources of energy. The photocatalytic reduction of carbon dioxide (CO 2 ) is currently one of the hottest candidates

Preparation
Melamine powder 99% was purchased from Alfa Aesar, and iron nitrate 98% was purchased from Chem-Lab.
Bulk g-C 3 N 4 was synthesized by thermal polycondensation of melamine. The conditions were chosen by taking the relevant literature into account [28][29][30][31], and the results of our previous work were also considered [32,33]. Typically, 5 g of melamine was put into a covered alumina crucible and was heated in a muffle furnace in air to 520 • C with a rate of 5 • C/min, where it remained for 2 h. The yellow product (yield~2.4 g) was collected and was ground into fine powder. The sample was called CN.
A similar procedure was used for the preparation of the iron-modified g-C 3 N 4 photocatalysts. In detail, 5 g of melamine was mixed with specific amounts of iron nitrate resulting in iron concentrations of 0.1, 0.3, and 1.1 wt.%. The mixture was stirred in 100 mL H 2 O for 1 h to achieve a homogeneous suspension. After drying at 80 • C in an oven, the powder was put into a covered alumina crucible and was heated in a muffle furnace in air to 520 • C at a rate of 5 • C/min, where it remained for 2 h. The product (yield~2.5 g) was collected and ground into fine powder. The samples were called CN-Fex, where x is the amount (wt.%) of iron.

Characterization
The crystalline structure, surface area, porosity, chemical composition, and optical characteristics of the samples were thoroughly examined, and the nature of the iron ions was also investigated. Details about the characterization techniques and the equipment that was used are presented in the Supplementary Materials.

Photocatalytic Test
The photocatalytic reduction of CO 2 was performed in a homemade batch-stirred stainless steel reactor (volume 348 mL) with a quartz glass window on the top ( Figure S1). A UV 8 W Hg lamp (peak intensity at 365 nm; Ultra-Violet Products Inc., Cambridge, UK) was used as a light source, and it was located over the quartz glass window of the reactor at a specific height so that the intensity on the level of the suspension was 0.833 mW/cm 2 .
The reactor was filled with 0.09 g of each photocatalyst and 100 mL of 0.2 M NaOH. The NaOH was added in order to increase the solubility of CO 2 in water, thus facilitating the photocatalytic reaction. The suspension was stirred with a magnetic stirrer the entire time to prevent the sedimentation of the photocatalyst. Before the photocatalytic reaction began, the reactor was tightly closed and purged with (He or) CO 2 . The pH of the suspension decreased during the purging from~12.5 to~6.9. A detailed description of the experimental procedure is available in the Supplementary Materials.
The gaseous samples were analyzed in 2 h intervals between the duration 0-8 h, where 0 corresponded to the moment prior to UV irradiation. Samples were collected using a gastight syringe (Hamilton Co., Reno, NV, USA) and were analyzed on a gas chromatograph (Shimadzu Tracera GC-2010 Plus) equipped with a BID detector. Each sample was measured at least three times using the same batch (same photocatalyst, same NaOH) in order to confirm the durability of the photocatalyst and the reproducibility of the experiments (within 5% error). The photon flux and apparent quantum yields (AQY) for individual products in the presence of each photocatalyst were calculated. The photocatalysts were first examined in an inert atmosphere to confirm that no products were formed, and after that the photocatalytic reduction of CO 2 was conducted.

Crystalline Structure
The XRD patterns of the g-C 3 N 4 photocatalysts are presented in Figure 1. For all of the samples, the two characteristic diffraction peaks of g-C 3 N 4 are observed (JCPDS, PDF #87-1526). The weak peak at 13.1 • corresponds to the (100) plane, which is related to the in-plane structural packing motif of the tri-s-triazine units with an interplanar distance of d = 0.675 nm [34,35]. The strong peak at 27.5 • corresponds to the (002) plane, which is attributed to the interlayer stacking of aromatic rings with a distance of d = 0.324 nm [36,37]. After the modification, no additional peaks corresponding to iron are observed, which was probably due to the low initial amount of iron nitrate. Furthermore, there are no significant changes to the position or the intensity of the peaks, indicating that the crystalline structure of the materials is preserved [38].
#87-1526). The weak peak at 13.1° corresponds to the (1 in-plane structural packing motif of the tri-s-triazine uni d = 0.675 nm [34,35]. The strong peak at 27.5° correspon tributed to the interlayer stacking of aromatic rings w [36,37]. After the modification, no additional peaks corr which was probably due to the low initial amount of iro no significant changes to the position or the intensity crystalline structure of the materials is preserved [38].

Surface Area and Porosity
The N2 adsorption-desorption isotherms and pore photocatalysts are displayed in Figure S2, with the char All of the samples showed characteristic hysteresis loop typical for mesoporous materials [39]. After the modific the photocatalysts decreased, which can be attributed g-C3N4 pores by the iron. On the other hand, the pore s cally unchanged across all of the samples.

Surface Area and Porosity
The N 2 adsorption-desorption isotherms and pore-size distributions of the g-C 3 N 4 photocatalysts are displayed in Figure S2, with the characteristic values given in Table 1. All of the samples showed characteristic hysteresis loops of type IV isotherms, which is typical for mesoporous materials [39]. After the modification, the specific surface area of the photocatalysts decreased, which can be attributed to the partial blocking of the g-C 3 N 4 pores by the iron. On the other hand, the pore size distribution remained practically unchanged across all of the samples. Table 1. BET surface area, average pore size, and total pore volume of the g-C 3 N 4 photocatalysts.

Sample
BET Surface Area (m 2 /g) Average Pore Size (nm) Total Pore Volume (cm 3 /g)

Chemical Composition
The FT-IR analysis results are shown in Figure 2. For all of the samples, the known characteristic peaks of g-C 3 N 4 can be observed. The broad peak at the 3500-3000 cm −1 region corresponds to the stretching mode of all of the OH-containing species and N-H stretching from the residual amino groups [40,41]. The shoulder located at 1630 cm −1 represents the H-O-H bending mode of all of the water molecules present at the surface of the material. The peaks in the 1628-1232 cm −1 region correspond to the characteristic stretching of the C-N heterocycles, including the trigonal N-(C) 3 and bridging H-N-(C) 2 units [42]. Finally, the sharp absorption peak at 805 cm −1 can be attributed to the breathing mode of the triazine units [43]. The FT-IR spectra of the modified materials demonstrate no significant changes. More specifically, there is no observable shift in the peaks, indicating that the chemical bonding of the main g-C 3 N 4 network is unaffected [38]. gion corresponds to the stretching mode of all of the stretching from the residual amino groups [40,41]. T represents the H-O-H bending mode of all of the wate of the material. The peaks in the 1628-1232 cm −1 regio stretching of the C-N heterocycles, including the trigo units [42]. Finally, the sharp absorption peak at 805 cm ing mode of the triazine units [43]. The FT-IR spectra strate no significant changes. More specifically, there indicating that the chemical bonding of the main g-C3N

Optical Characteristics
The measured diffuse reflectance spectra of the g the inset of Figure 3. They were used for the constructi E) 1/2 = f(E) presented in Figure 4, which allowed the de gap energy (Eg), as described in Ref. [44]. It is eviden absorption of the materials is increased. However, th 2.73 eV for bulk g-C3N4, which is in agreement with t the modified samples.

Optical Characteristics
The measured diffuse reflectance spectra of the g-C 3 N 4 photocatalysts are shown in the inset of Figure 3. They were used for the construction of the absorption functions (F × E) 1/2 = f(E) presented in Figure 4, which allowed the determination of the materials' band gap energy (E g ), as described in Ref. [44]. It is evident that after modification, the light absorption of the materials is increased. However, the E g only slightly decreased from 2.73 eV for bulk g-C 3 N 4 , which is in agreement with the literature [16,17], to 2.71 eV for the modified samples.
The properties of the photogenerated charge carriers were evaluated with fluorescence measurements at λ ex = 365 nm ( Figure 4). All of the samples showed a broad band with λ em at around 455 nm, which is in agreement with the absorption edge wavelength measured by UV-vis spectroscopy. Such a signal can be attributed to the band-band PL phenomenon, which mainly results from the n-π* electronic transitions in g-C 3 N 4 [45]. The strongest emission intensity is displayed by bulk g-C 3 N 4 , which suggests a faster electron-hole radiative recombination rate. On the other hand, the modified samples show significantly weaker emission intensity, which becomes even more prominent as the iron content increases. This indicates a lower recombination rate and higher charge transfer efficiency, suggesting a potential improvement in the photocatalytic activity of the materials. gap energy (Eg), as described in Ref. [44]. It is evident tha absorption of the materials is increased. However, the Eg 2.73 eV for bulk g-C3N4, which is in agreement with the lit the modified samples.

21, 1, FOR PEER REVIEW
The properties of the photogenerated charge carrie cence measurements at λex = 365 nm ( Figure 4). All of the with λem at around 455 nm, which is in agreement with t measured by UV-vis spectroscopy. Such a signal can be phenomenon, which mainly results from the n-π* electr The strongest emission intensity is displayed by bulk g electron-hole radiative recombination rate. On the othe show significantly weaker emission intensity, which bec the iron content increases. This indicates a lower recomb transfer efficiency, suggesting a potential improvement the materials. Photoelectrochemical measurements are one of the c allow the prediction of the photocatalyst behavior. Since Photoelectrochemical measurements are one of the characterization techniques that allow the prediction of the photocatalyst behavior. Since the photocatalyst is irradiated under an external potential that has been applied to the working electrode, the recombination of electrons and holes is strongly suppressed. Based on the photocurrent generation ( Figure 5), it is clear that the highest amount of charge carriers is produced in the case of bulk g-C 3 N 4 . This is connected to a higher specific surface area. However, a higher amount of produced charge carriers does not mean higher photocatalytic activity, especially since bulk g-C 3 N 4 has the highest recombination rate of charge carriers (Figure 4). It is known that the redox potential of Fe 3+ /Fe 2+ is below the CB of g-C 3 N 4 . After iron modification, the photogenerated electrons could be trapped by the Fe 2+ sites, leading to the reduced recombination of the photogenerated electron-hole pairs. With a higher iron concentration, more photogenerated electron trapping sites exist, thus leading to the further decrease of the PL intensity.
under an external potential that has been applied to the bination of electrons and holes is strongly suppressed. Ba ation ( Figure 5), it is clear that the highest amount of cha case of bulk g-C3N4. This is connected to a higher specific s amount of produced charge carriers does not mean high cially since bulk g-C3N4 has the highest recombination rate is known that the redox potential of Fe 3+ /Fe 2+ is below the fication, the photogenerated electrons could be trapped reduced recombination of the photogenerated electronconcentration, more photogenerated electron trapping s further decrease of the PL intensity.

Mössbauer and EPR Spectroscopy
Mössbauer spectroscopy was used to investigate the nature of the iron of the modified g-C 3 N 4 photocatalysts. The 80 K spectra of samples CN-Fe0.3 and CN-Fe1.1 are shown in Figure 6. The most prominent characteristic of both spectra are the two distinct doublets that are characteristic of the Fe 2+ ions in two different chemical environments [46]. These environments may be associated with the iron that is present in the -C 3 N 4 network gaps or on the layer surface. The spectra also show a lower intensity doublet showing the presence of Fe 3+ ions. A subspectrum that can be attributed to magnetic Fe 2 O 3 [47], which is most likely due to the reaction of weakly stabilized Fe 3+/2+ ions with air, is also observed in all of the spectra. At 80 K there is no sign of any subspectral broadening that would suggest the existence of superparamagnetic nanoparticles. Thus, the only crystalline iron-containing phase observed through Mössbauer spectroscopy is that of the Fe 2 O 3 oxide. evident that iron in the g-C3N4 matrix is mainly present in the form of Fe 2+ species. This indicates that during melamine polycondensation, the Fe 3+ ions of the iron nitrate precursor are reduced to Fe 2+ . It should be also noted that the Mössbauer peaks of the sample CN-Fe0.3 are less intense than those of sample CN-Fe1.1, which confirms the lower amount of Fe present in the photocatalyst. The EPR spectra of the modified samples ( Figure S3) display a broad line that is characteristic of iron ion-ion interactions [48]. However, as expected, the Fe 2+ ions were not visible in the EPR measurements. Recent studies have revealed that transition metal ions such as Au 3+ , Ag + , Cu 2+ , and Fe 3+ may be reduced during g-C3N4 synthesis at ~550 °C [49,50]. During melamine polycondensation, apart from NH3, reactive species such as CNH2, H2NCN, and CN2 + are also released. These species can progressively reduce Fe 3+ to Fe 2+ until the formed ions are stabilized in the g-C3N4 network [51]. Figure 7 shows the dependence of product yields on the irradiation time (0-8 h). The main products of the photocatalytic reduction of CO2 are methane (Figure 7a) and carbon monoxide (Figure 7b). However, the hydrogen generated from the photocatalytic water splitting is also present in much higher concentrations (Figure 7c).

Photocatalytic Activity
Photocatalytic CO2 reduction is a complex process with a number of reactions occurring simultaneously. The g-C3N4 band gap energy around 2.7 eV (Figure 3) corresponds to a wavelength of approximately 460 nm, and since the photocatalytic reduction By comparing the relative spectral areas of the corresponding peaks (Table S2), it is evident that iron in the g-C 3 N 4 matrix is mainly present in the form of Fe 2+ species. This indicates that during melamine polycondensation, the Fe 3+ ions of the iron nitrate precursor are reduced to Fe 2+ . It should be also noted that the Mössbauer peaks of the sample CN-Fe0.3 are less intense than those of sample CN-Fe1.1, which confirms the lower amount of Fe present in the photocatalyst. The EPR spectra of the modified samples ( Figure S3) display a broad line that is characteristic of iron ion-ion interactions [48]. However, as expected, the Fe 2+ ions were not visible in the EPR measurements.
Recent studies have revealed that transition metal ions such as Au 3+ , Ag + , Cu 2+ , and Fe 3+ may be reduced during g-C 3 N 4 synthesis at~550 • C [49,50]. During melamine polycondensation, apart from NH 3 , reactive species such as CNH 2 , H 2 NCN, and CN 2 + are also released. These species can progressively reduce Fe 3+ to Fe 2+ until the formed ions are stabilized in the g-C 3 N 4 network [51]. Figure 7 shows the dependence of product yields on the irradiation time (0-8 h). The main products of the photocatalytic reduction of CO 2 are methane ( Figure 7a) and carbon monoxide (Figure 7b). However, the hydrogen generated from the photocatalytic water splitting is also present in much higher concentrations (Figure 7c).  (Table S1). Among the photocatalysts, the CN-Fe0.1 demonstrated the highest CH4 and H2 AQY at 0.0020% and 0.0733%, respectively. One of the most important properties of a photocatalyst is the CB and VB edge potentials. XPS analysis was used to determine the potential of the valence bands of each photocatalyst ( Figure 8).  Photocatalytic CO 2 reduction is a complex process with a number of reactions occurring simultaneously. The g-C 3 N 4 band gap energy around 2.7 eV (Figure 3) corresponds to a wavelength of approximately 460 nm, and since the photocatalytic reduction was conducted under the irradiation of 365 nm, each incident photon should lead to the generation of an electron and hole. The product yields and the apparent quantum yields (AQY) are shown in the Supplementary Materials (Table S1). Among the photocatalysts, the CN-Fe0.1 demonstrated the highest CH 4 and H 2 AQY at 0.0020% and 0.0733%, respectively. One of the most important properties of a photocatalyst is the CB and VB edge potentials. XPS analysis was used to determine the potential of the valence bands of each photocatalyst (Figure 8).  As can be seen, the VB edge potential shifted to more positive values after the mod ification, which indicates the increased oxidation efficiency of the photogenerated hole [21]. By combining the band gap energy and VB edge potential values, the CB edge po tentials of the materials were also calculated. Thus, the electronic band structures of th g-C3N4 photocatalysts were created ( Figure 9). As the Eg of the photocatalysts remain practically unchanged after the modification, the CB edge potential shifts to less negativ values, indicating a slight decrease in the reduction strength of the photogenerated elec trons. However, the potential of CB is still higher than the redox potential that is require for the CO2 reactions (Equations (1)-(5)). As can be seen, the VB edge potential shifted to more positive values after the modification, which indicates the increased oxidation efficiency of the photogenerated holes [21]. By combining the band gap energy and VB edge potential values, the CB edge potentials of the materials were also calculated. Thus, the electronic band structures of the g-C 3 N 4 photocatalysts were created ( Figure 9). As the E g of the photocatalysts remains practically unchanged after the modification, the CB edge potential shifts to less negative values, indicating a slight decrease in the reduction strength of the photogenerated electrons. However, the potential of CB is still higher than the redox potential that is required for the CO 2 reactions (Equations (1)-(5)).

Photocatalytic Activity
The potentials of valence and conduction bands slightly differ for each g-C 3 N 4 material; however, it is approximately inside the interval −1.5 V (CB potential) and 1.2 V (VB potential) vs. NHE (pH = 7). It is the relatively high reduction strength (negative potential of CB) that makes g-C 3 N 4 such an attractive material.
The potential of CB is negative enough to allow any of the partial redox reactions to proceed (Equations (1)-(5)) [52,53]. However, the direct reduction of a CO 2 molecule by a single electron is not possible due to a very negative required potential of −1.9 V (Equation (6)) [2,5,54,55].
All of the redox potentials are stated vs. NHE at pH = 7.
The potentials of valence and conduction bands slightly d terial; however, it is approximately inside the interval −1.5 V (CB potential) vs. NHE (pH = 7). It is the relatively high reduction str of CB) that makes g-C3N4 such an attractive material.
The potential of CB is negative enough to allow any of the proceed (Equations (1)-(5)) [52,53]. However, the direct reductio single electron is not possible due to a very negative required p tion (6)) [2,5,54,55]  Based on the equations above, it is clear that the photocatalytic reduction of CO 2 is a multielectron process that also requires the presence of a hydrogen cation. The necessity of a hydrogen cation is the reason why the reaction has to be conducted in aqueous phase or in the presence of water vapor.
There are two possible ways that water can be oxidized by photogenerated holes toward H + (Equations (7) and (8)) [56].
It is clear that Equation (8) would be more probable since it requires just one hole and one water molecule; however, its redox potential is much more positive than the VB potential of g-C 3 N 4 , and therefore, this reaction cannot proceed in the presence of g-C 3 N 4 material, leaving Equation (7) as the main source of the required H + ions. This significantly complicates the situation. First of all, the reaction (Equation (7)) is much less probable since it requires two water molecules o and four holes at the same time, but oxygen is also produced along with H + . The reduction of oxygen molecules to a superoxide radical (Equation (9)) is one of the competitive reactions for the reduction of CO 2 [27,57].
The presence of oxygen not only competes with CO 2 molecules to be adsorbed on the surface of the photocatalyst but also consumes the necessary electrons that are needed for the reduction.
Another competitive reaction is the reduction of the generated H + ions to hydrogen (Equation (10)). 2H It is this competitive reaction that is responsible for the presence of hydrogen among the detected products (Figure 7c). Methane and carbon monoxide were other detected reaction products that can be generated according to Equations (2) and (5). Unfortunately, the possible products in the liquid phase were below the detection limit, and therefore, their presence could not be confirmed.
It is clear that the photocatalytic reduction of CO 2 is a very complex process that follows several multielectron steps. Nevertheless, the presence of the Fe 3+ /Fe 2+ couple ( Figure 6) can complicate the situation even more. Many publications say that the presence of iron ions can be beneficial due to the possibility of photo-Fenton process occurring along with the photocatalysis [58][59][60]. What does this mean in the case of a photocatalytic CO 2 reduction? Part of the photogenerated electrons can directly transfer to Fe 3+ ions and form Fe 2+ (Equation (11)). Fe Holes, however, stay in the valence band and can participate in the oxidation reaction. The presence of Fe 3+ can improve the separation of generated charge carriers; however, the negative side effect is the consumption of electrons needed for the photocatalytic reduction of CO 2 ; therefore, finding an optimum amount of iron ions is necessary [15].
Since the Fe 2+ ions are unstable, they easily oxidize back into Fe 3+ in the presence of oxygen and form either O − 2 (Equation (12)) or H 2 O 2 (Equation (13)).
The above-mentioned equations clearly suggest the consumption of electrons in both redox Fe 3+ /Fe 2+ reactions. Higher amounts of iron ions would clearly explain the lower yields of the CO 2 reduction products. On the other hand, the oxygen used in Equations (12) and (13) does not compete with CO 2 molecules needed for adsorption on the surface of the photocatalyst.
In addition, the holes in g-C 3 N 4 VB have sufficient potential to oxidize H 2 O to H 2 O 2 (+0.69 V vs. NHE) [27,61] and the potential of H 2 O 2 /·OH (1.07 V vs. NHE) [27] is more positive than the redox potential of Fe 3+ /Fe 2+ (0.77 V vs. NHE) [27,61,62]; the produced H 2 O 2 reacts with Fe 2+ to produce ·OH species. Hydroxyl radical species are very strong oxidizing agents and are very beneficial when the removal of organics in water is the goal. However, the goal is to create organic compounds for the photocatalytic reduction of CO 2 . Therefore, a higher amount of Fe would lead to a higher amount of hydroxyl radicals that would be able to oxidize the already generated organic products (CH 4 ) [27]. This is the reason why the CH 4 yields were lower in the case of photocatalysts containing 0.3 and 1.1 wt.% of Fe (Table S1). Furthermore, when the iron content exceeds an optimal value, the photocatalytic performance decreases, which is possibly due to the competitive capture between the adsorbed CO 2 and Fe 2+ sites. In this case, an excess of Fe 2+ sites could trap more photogenerated electrons, leading to fewer electrons being available to react with the adsorbed CO 2 molecules. On the other hand, the higher Fe content in the photocatalyst clearly leads to a higher selectivity toward CO production (Figure 7b). This suggests higher selectivity for CO 2 reduction to CO. Higher amounts of iron ions are incorporated into the g-C 3 N 4 lattice and affect its electronic properties, which were confirmed by the valence band position shift after the addition of Fe (Figure 9), forming impurities that lead to an enhanced separation of charge carriers. On the other hand, this means a higher amount of hydroxyl radicals and a higher rate of reverse oxidation for the produced hydrocarbons [63].

Conclusions
Iron-modified g-C 3 N 4 photocatalysts were prepared using a simple thermal treatment of melamine mixed with specific amounts of iron nitrate. The XRD and FT-IR measurements did not reveal any iron in the photocatalysts due to its very low concentrations (0.1, 0.3 and 1.1 wt.%). However, its presence in the form of Fe 2+ , Fe 3+ , and Fe 2 O 3 was demonstrated by Mössbauer spectroscopy, with Fe 2+ being the main species.
The addition of iron resulted in a decrease in the photocurrent and specific surface area due to the partial blocking of the g-C 3 N 4 pores by iron. However, this led to the reduced recombination of the photogenerated electron-hole pairs, which is important for high photocatalytic efficiency.
The photocatalytic activity measurements of the prepared materials in the CO 2 reduction showed that the photocatalyst with the lowest amount of iron exhibited the biggest methane and hydrogen evolution in comparison with pure g-C 3 N 4 and g-C 3 N 4 with higher iron amounts. Excess iron consumes electrons that are needed for the reduction reactions and generates hydroxyl radicals that oxidize organic products from the CO 2 reduction. The iron-modified (0.1 wt.%) g-C 3 N 4 proved to be a promising photocatalyst for CO 2 conversion into valuable hydrocarbons.