Band structure engineering of carbon nitride hybrid photocatalysts for CO 2 reduction in aqueous solutions †

Through the co-polymerisation of dicyandiamide and barbituric acid precursors, a series of visible light active carbon nitride photocatalysts has been prepared and characterized, and their photocatalytic activity has been evaluated. Structural and electronic characterisation has enabled variations in observed activity towards water splitting and CO 2 reduction to be understood, both in the presence and absence of the iron porphyrin co-catalyst Fe III tetra(4-carboxylphenyl)porphyrin (FeTCPP). A combination of the most active carbon nitride catalyst using 5 wt% barbituric acid and FeTCPP provides a hybrid system where the alignment of band structure with appropriate reduction potentials and enhanced carrier lifetimes is capable of CO 2 reduction in an aqueous solution with >60% selectivity for CO production. This study is one of only a few that achieves selective CO 2 reduction using a hybrid molecular catalyst-carbon nitride photocatalyst in aqueous solution.


Introduction
Carbon nitride (CN) is a promising material in the eld of photocatalysis and a wide range of synthetic routes exist but developing an understanding of how these different procedures modify electronic structure and in turn the effect on photocatalytic activity is limited. 1,2Of particular interest is the use of molecular doping, or copolymerisation with additional cyclic monomers, in an attempt to improve photocatalytic activity.The rst compound used for copolymerisation with any carbon nitride precursor was barbituric acid (BA) in 2010. 3Wang and co-workers synthesised a series of carbon nitrides by mixing dicyandiamide (DCDA) with different amounts of BA.The resultant materials showed trends with the quantity of BA used; with increasing amounts of barbituric acid used, the materials carbon content increased which in turn lead to a decrease in the materials band gap and shis in the band structure.The materials were tested for photocatalytic H 2 evolution under UV and visible illumination, in the presence of triethanolamine (TEOA) hole scavenger and chloroplatinic acid (H 2 PtCl 6 ) for in situ Pt deposition.It was found that copolymerisation with BA resulted in enhanced photoactivity for hydrogen evolution, but only up to 5 wt% doping; past this point, activity decreased signicantly which was taken to be due to excessive doping likely introducing defects (Fig. 1) acting as recombination centres and loss of conjugation along the carbon nitride backbone hindering photoactivity.It has also previously been claimed that through the incorporation of BA in carbon nitride that heterojunction formation can be achieved which leads to improved photocatalytic activity. 4ere we examine in detail the electronic structure and photophysics of carbon nitrides with different C : N ratios prepared through the co-polymerisation of DCDA and BA in order to rationalise the behaviour of this widely studied class of materials.Due to the understanding gained on this class of BA doped carbon nitrides, we are able to rationalise the behaviour of an effective co-catalyst/carbon nitride combination for CO 2 reduction in water and infer how the control of the band structure by BA doping can control the yield of charge transfer to the co-catalyst.Oen photocatalytic materials can suffer from low activity due to high charge recombination rates.Addition of a co-catalyst to the surface can aid in charge separation improving photocatalytic activity and a wide range of ways to improve charge transfer kinetics have been studied, including by addition of dopants such as S, P, 5 carbon domains 6 and transition metal phosphides 7 amongst others.3][14][15][16][17][18][19] Of particular relevance is the work of He and co-workers who have published on the use of Fe III tetra(4-carboxylphenyl)porphyrin (FeTCPP) as a stable cocatalyst on an undoped carbon nitride for CO 2 reduction. 20,21n 2018, they found that the carbon nitride/FeTCPP hybrid was capable of reducing CO 2 to CO with 98% selectivity when the photocatalytic experiment was performed in acetonitrile/water/ triethanolamine (3 : 1 : 1, v : v : v).A subsequent study presents a proposed carbon dot modied carbon nitride further enhanced CO evolution rates using the same co-catalyst.These are important studies demonstrating the promise of this cocatalyst, along with the mechanism and intermediates for Fe porphyrin reduction of CO 2 discussed by Bonetto et al., 22 but to date it has not been used for CO 2 reduction to CO in water with a carbon nitride absorber with recent reports focussing on photodegradation of p-nitrophenol. 23The data presented here show these hybrid systems are capable of performing CO 2 reduction in an aqueous solution with >60% selectivity for CO production over H 2 .

Experimental
CN samples were synthesised by stirring dicyandiamide (3 g) in water with different amounts of barbituric acid; 0, 0.15, 0.3 and 0.6 g, for CN-DCDA, CN-BA(5), CN-BA (10) and CN-BA (20)  respectively, where the number in parenthesis refers to the BA% in the precursor mixture.Mixtures were then dried at 60 °C and then heated to 550 °C in a muffle furnace in covered crucibles for 4 hours in air with a heating and cooling rate of 5 °C min −1 .Higher surface area carbon nitrides were prepared via thermal oxidation etching of the bulk materials by heating in open crucibles to 500 °C for 2 hours with a ramp rate of 5 °C min −1 in a muffle furnace, in air.FeTCPP modied carbon nitrides were prepared by mechanical mixing of 50 mg carbon nitride in 50 mL 90 mM FeTCPP (4 mg) ethanolic solution, le stirring for 24 hours in the dark.
Elemental Analysis (CHN) and Inductively Coupled Plasma Optical Emission Spectroscopy (ICP OES) were performed by the University of Liverpool analytical services using a ThermoFisher FlashEA 1112 CHN analyser and an Agilent 5110 ICP-OES spectrometer, respectively.N 2 adsorption-desorption isotherms at 77 K were collected on a Micromeritics Tristar instrument.Samples were degassed under vacuum at 130 °C for 20 h before measurement.Powder X-ray diffraction (PXRD) was carried out using a Bruker D8-Advance X-ray diffractometer operating with Cu Ka 1 of l = 0.15418 nm for carbon nitride samples.Carbon nitride samples were studied by Diffuse Reectance Infrared Fourier Transform Spectroscopy (DRIFTS) on a Bruker Vertex 70 spectrometer, tted with a detachable DRIFTS attachment.All Nuclear Magnetic Resonance (NMR) spectra were recorded under Cross Polarisation (CP) Magic Angle Spinning (MAS) conditions on a Bruker DSX 400 MHz NMR spectrometer equipped with a 4 mm HXY probe tuned to X = 13 C at 101 MHz and Y = 15 N at 40 MHz.UV-vis-NIR diffuse reectance spectroscopy (DRS) data was obtained with a Shimadzu UV-2550 UV/ vis spectrometer, equipped with an integrating sphere, over the spectral range of 200-1400 nm and BaSO 4 was used as a reectance standard.X-ray photoelectron spectroscopy (XPS) valence band spectra were measured using a Mg Ka (1253.6 eV) X-ray source operating at 144 W and a hemispherical PSP Vacuum Technology electron energy analyser.XPS core level spectra was carried out with a two chamber Thermo Theta Probe spectrometer using a monochromated Al Ka X-ray source (1486.6 eV) in constant analyser energy mode.X-rays were focused to a 400 micron spot at the sample surface, which dened the analysis area.Sample charging was prevented by use of a dual beam ood gun.High resolution core line spectra, were recorded at 50 eV pass energy, and survey spectra were recorded at 150 eV pass energy.Steady-State Photoluminescence (PL) and Time-Resolved Emission Lifetime Spectroscopy measurements were recorded on an Edinburgh Instruments FLS980-D2S2-STM spectrophotometer, equipped with a 450 W Ozone free Xe arc lamp, excitation and emission monochromators and a photomultiplier tube detector.FT-Raman was performed on a Bruker MultiRAM, IRFS27.Full details of the characterisation of materials can be found in the ESI.† Photocatalytic experiments were carried out using 2 mg of photocatalyst dispersed in 2 mL solution in the presence of a range of hole scavengers (EDTA, TEOA) in a 4 mL vial.Vials were purged with N 2 or CO 2 for 30 minutes prior to being placed under illumination.The tests were carried out using a 300 W Xe lamp (Newport) with a 375 nm long pass lter or a Schott glass KG1 lter (>50% transmission between 375 and 690 nm) at an intensity of ∼100 mW cm −2 , unless stated otherwise.For the photoresponse experiments a 75 W Xe lamp (OBB Corp.) equipped with a monochromator and was focussed onto the cell at an intensity of ∼0.16 mW cm −2 .Gases evolved were detected by an Agilent technologies 6890N instrument equipped with a pulsed discharge detector (D-3-I-HP, Valco Vici) and a 5 Å molecular sieve column (ValcoPLOT, 30 m length, 0.53 mm ID) with N6 helium as the carrier gas (5 mL min −1 ).For isotopic labelling experiments, systems were rst purged with N 2 for 30 minutes and then purged with 13 CO 2 for several minutes before being placed under illumination.Gases evolved were collected and injected into a special made pre-purged cell which was then placed in the FTIR equipment.

Results and discussion
A series of carbon nitrides were prepared using up to 20% (wt.)BA content based on an existing synthetic method 3 as shown in Fig. 1, with the resultant materials being labelled CN-DCDA, CN-BA(5), CN-BA (10) and CN-BA (20) where the number in parenthesis gives the BA wt%.All materials had relatively low surface area ranging from 12.81 m 2 g −1 for CN-DCDA to 4.41 m 2 g −1 for CN-BA(20) (Fig. S1, Table S2 †).PXRD shows the characteristic peaks of carbon nitride as discussed in the ESI (Fig. S2 †).CHN conrmed that by increasing the amount of BA in the precursor mixture from CN-DCDA to CN-BA (20) led to increased carbon content demonstrated by an increase in the C : N from 0.663 to 0.736 (Table S1 †).][40][41][42][43][44] Past studies on co-polymerisation of BA with DCDA suggested that BA would be directly incorporated into the carbon nitride structure and likely lead to the replacement of one of the N ring atoms with C, leading to C]C-H functionality. 3 While the 13 C CP MAS NMR spectra (Fig. 2, Table S6 and S7  ), 163 (C a ) and 157 ppm (C e ) as expected, 45 here we also nd that the 13 C spectrum of CN-BA(20) reveals a new peak at 94 ppm signalling a new carbon C d environment which is not present in the CN-DCDA sample.7][48][49] The new peak at 94 ppm is rather broad indicating the presence of a range of chemical shis and local environments for the C]C-H functionality. 50The 13 C and 15 N MAS NMR, DRIFTS, Raman and XPS spectra (Fig. 2, S4-S7, Table S4-S9 †) also show the presence of -C^N, -NH 2 and -N(H)-groups in samples prepared in both the presence and absence of copolymerised BA, in line with there being a range of defect sites present.A further defect arising as a result of BA Fig. 2 (Left) 13 C and (right) 15  doping is identied through deconvolution of the high resolution C 1s XPS (Fig. S7 †).As the BA concentration increases, a peak around 290 eV increases in intensity.This has been tentatively assigned to O-C]O (C g ), likely arising from the C]O groups present in the barbituric acid or bound CO 2 .This could not be conrmed by 13 C MAS NMR as it lies at a similar shi to the C-N heterocycle peaks and, if present, are likely below the detection limit of 13 C NMR or too mobile for CP detection.
The addition of BA is therefore shown to lead to the generation of a number of defects, including, C^N, C]C-H, -NH x , (C e , N f , C f , N d and N e ) in-line with past studies on BA doped carbon nitrides and these have a great effect on the band gap of the carbon nitride materials.For CN-DCDA, CN-BA(5), CN-BA (10) and CN-BA (20), the indirect band gaps for these materials are 2.69 eV, 2.40 eV, 2.21 eV and 1.92 eV, respectively (Fig. 3 and Table S10 †).Previously, the composition of the bands has been determined computationally, revealing that the CB is made up of C p z orbitals, whilst the VB is made up of N p z orbitals. 51Others have stated that observed shis in band positions were due to the inclusion of NH x . 50Within this work it is likely that the addition of a variety of different carbon states alters the make-up of the VB and CB causing them to shi in different directions, leading to a decrease in the band gap.
In order to test if the improved light harvesting of the CN-BA series translated into improved photocatalytic activity the materials were tested for H 2 evolution.We have studied the asprepared (labelled here bulk) samples in the presence of ∼1 wt% Pt co-catalyst (deposited in situ) and a 10 mM EDTA hole scavenger.As the synthesis method gives relatively low surface area materials (see below), we have also prepared thermally exfoliated high surface area (HSA) materials from the bulk samples for photocatalytic testing.Full details of the exfoliation process can be found in the ESI † and these show the same structural properties as the parent bulk materials.From Fig. 4 it can be seen that the highest performing material for photocatalytic hydrogen evolution was CN-BA(5) for both the bulk and HSA materials, with a large increase in activity post-thermal exfoliation.In line with past reports, 3,4,52-60 the addition of BA into the carbon nitride synthesis leads to an improvement in photocatalytic hydrogen evolution rate, despite a large decrease in BET surface area.The surface area decreases with BA addition (12.8 (CN-DCDA) to 4.4 m 2 g −1 (CN-BA( 20)), Fig. S1, Table S2 † and 192.5 m 2 g −1 (HSA-CN-DCDA) to 63.8 m 2 g −1 (HSA-CN-BA( 20)), Fig. S10, Table S13 †).However, there is not a simple correlation between the light harvesting properties of the material or surface area with photocatalytic activity.
We have also evaluated the photocatalytic CO 2 reduction activity of the carbon nitride materials when modied with a FeTCPP co-catalyst in aqueous solutions.High surface area (HSA-CN) materials were soaked in solutions of the FeTCPP catalyst for 24 hours prior to washing to remove unbound catalysts, details can be found in the ESI.† Loading levels of the FeTCPP on each carbon nitride were determined via ICP and by comparison of the soaking solutions pre-and post-soaking via UV-vis (Fig. S11 and Table S15 †).With the exception of FeTCPP/ HSA-CN-BA (20), all of the FeTCPP/HSA-CN samples were found to reduce CO 2 to CO under UV-vis illumination (>375 nm) in the presence of TEOA as a hole scavenger over 4 hours of illumination at signicant levels, Fig. 5 and Table 1.Control experiments under N 2 showed no signicant levels of CO evolution, instead H 2 becomes the dominant product (Table S17 and S18, Fig. S12 †) and experiments using isotopically labelled 13 CO 2 show that 13 CO is formed (Fig. S13 †) conrming that CO is produced from the purged CO 2 .The CO generated while purging with N 2 is due to decomposition of the TEOA holescavenger as discussed in the ESI page S22.† The highest rate of CO evolution was obtained by HSA-CN-BA(5)/FeTCPP, with a rate of 5.7 nmol h −1 (75 mmol g −1 (FeTCPP) h −1 ), which although lower than results reported by others in non-aqueous solvents and higher incident light intensity (6.52 mmol g −1 with 98% CO selectivity), 20,21 represents a relatively rare example of  selective CO 2 reduction using a hybrid molecular catalystcarbon nitride photocatalyst in water, with the molecular catalyst enabling selectivities of greater than 2 : 1 (CO : H 2 ).The rate of CO production followed a similar trend to that observed for H 2 evolution using a Pt co-catalyst, with HSA-CN-BA(5)/FeTCPP out performing both HSA-CN-BA(10)/FeTCPP and HSA-CN-BA(20)/FeTCPP, however it is notable that despite having signicantly worse light harvesting properties HSA-CN-DCDA achieves a higher CO evolution rate than both HSA-CN-BA(10)/ FeTCPP and HSA-CN-BA(20)/FeTCPP.
To assess the role of the molecular catalyst, a number of control experiments were performed.In the absence of the FeTCPP catalyst, negligible CO is evolved (Fig. 5) and H 2 is the sole product (Fig. S12 †).FeTCPP is able to absorb visible light, but we found that in the absence of the carbon nitride catalyst, the CO yield decreases by a factor of ca.67 compared to the HSA-CN-BA(5)-FeTCPP sample (Table S17 †) indicating that the primary photocatalytic pathway for CO formation is photon absorption by the carbon nitride with electron transfer to the iron porphyrin catalyst.Supporting this conclusion is the wavelength dependence of CO evolution for HSA-CN-DCDA-FeTCPP which matches well with the recorded UV-vis spectra of the carbon nitride, conrming that CO reduction was initiated by excitation of the carbon nitride, Fig. 6.][63] To understand the trends in photocatalytic activity with BA doping, we have carried out a detailed study of the electronic structure and the kinetics of the photogenerated charges.Past studies have proposed that excessive BA doping can cause the formation of high concentrations of defect sites which act as recombination centers. 3Steady-state photoluminescence data presented in this work shows that the emission peak maxima is red shied, which is consistent with the decrease in band gap, Fig. S9.5][66][67][68] Time correlated single photon counting (TCSPC) (Fig. 7, Table S11 †) shows that addition of barbituric acid clearly increases the photoluminescence lifetime from ca. 3 ns (DCDA, amplitude average weighted lifetime) to ca. 5 ns (CN-BA( 5)) and for all BA concentrations the lifetime exceeds that of the undoped DCDA (Fig. 7a).A similar trend is also observed for the high surface area sample (Table S12 †).Longer excited state lifetimes may increase the probability of charge transfer to a catalyst, and it is reasonable to conclude that the increase in TCSPC lifetimes of BA doped CN samples Fig. 5 Rate of CO evolution in nmol per hour for 4 mg photocatalyst in 4 mL 10% TEOA (v/v) in water.Samples were placed under UV-vis illumination at ∼100 mW cm −2 in a quartz cuvette, for 4 hours.The Xe lamp was equipped with a water filter and a 375 nm long pass filter.Conditions were varied to confirm CO 2 reduction by purging with different gases, CO 2 (A) and N 2 (B), using CN photocatalyst with FeTCPP in solution with no prior soaking (C), the CN photocatalyst by itself (D), just he FeTCPP in solution with no photocatalyst present (E) and the hole scavenger in aqueous solution with no catalysts present.(F) Full conditions of experiments can be found in Table S16.†  To explore further if the TCSPC can be used to directly monitor the rate of electron transfer to the FeTCPP from the HSA-CN, we have measured the photoluminescence lifetime following catalyst deposition and in all cases the lifetime decreases (Table S12 †).In all samples, except those with the highest BA doping (CN-BA(20)), modication with FeTCPP decreases the HSA-CN lifetime and this is clearly seen in the data for HSA-CN-BA(5), Fig. 7b.The decrease in PL lifetime when FeTCPP is present is in-line with fast electron transfer (<10 ns) to the catalyst.A simple kinetic model can be constructed where there are parallel pathways for relaxation of the emissive state, either through charge transfer to the catalyst or by emission as outlined in the ESI.† Although this is a severe over-simplication, not taking into account the potential role of non-emissive trap states in charge transfer and recombination, it does give rise to similar rate constants for electron transfer to the FeTCPP catalyst for the HSA-CN-DCDA-FeTCPP, HSA-CN-BA(5)-FeTCPP and HSA-CN-BA(10)-FeTCPP samples.Notably the 10% BA doped sample did not show a large change in TCPSC lifetime upon addition of the FeTCPP catalyst, in-line with there being minimal photocatalytic CO production.
To rationalise the photocatalytic behaviour observed and the apparent lack of electron transfer from the HSA-CN-BA(20)-FeTCPP sample it is important not only to study the impact of doping on charge carrier dynamics but also the change in thermodynamic driving force for the electron transfer steps brought about due to the shi in VB and CB positions upon doping.Through the combination of both UV-vis and XPS, the band structure of each of the materials was determined.XPS can be used to measure the valence band minimum (VBM) relative to the material's Fermi level (Fig. S8 †).To be able to compare materials, the Fermi level must be determined by measuring the material's work function (WF), which can be done via XPS by measuring the secondary electron cut-off (SEC).In this work, the VBM and SEC of each material has been measured and used in combination with the optical band gap determined from UV-visible diffuse reection spectroscopy to estimate the conduction and valence band positions.The calculated band positions of each material relative to water splitting and CO 2 reduction redox potentials are depicted in Fig. 8 and are stated in Table S10.† The change in band positions for CN-DCDA and CN-BA(20) agree well with the original study on copolymerisation of DCDA with barbituric acid, 44 however only a single barbituric acid content was reported and the wider trend in band structure as the barbituric acid content was changed was not reported.Here we nd that addition of lower levels of barbituric acid (from CN-DCDA to CN-BA( 5)) causes an upshi in the VB (of +0.41 V) with only a small change in the CB to more negative potentials (of −0.13 V), leading to a decreased band gap.This positive shi in band positions continues for the CN-BA(10) sample, with further decrease in the band gap.For CN-BA(20) both bands are shied to more positive potentials.
It has been previously reported that the redox couples for Fe III/II , Fe II/I and Fe I/0 for FeTCPP are 0.031 V, −0.61 V and −1.3 V vs. NHE, measured under Ar in dimethylformamide (DMF) and stated that FeTCPP is capable of accepting electrons from carbon nitride. 20Comparison of the band structures obtained for the bulk materials and the reported redox couples for FeTCPP in DMF can allow us to understand variations in activity (Fig. 8), although it must be taken in to consideration that the carbon nitride band structures are calculated under vacuum and the FeTCPP potentials were determined in an organic solvent, and the actual values would differ in solution and at different pH.CN-DCDA, CN-BA(5) and CN-BA (10) have CB that lie at a more negative potential than the Fe I/0 redox couple, which means that electrons can transfer to the FeTCPP co- catalyst resulting in the formation of Fe 0 , which is required for CO 2 reduction.However, CN-BA(20) has a CB which lies at a much more positive potential than the other materials and appears to lie below that of the Fe I/0 redox couple.This means that the excited electron which lies in the conduction band does not have enough energy to transfer over to the FeTCPP and form the Fe 0 species required, so CO 2 reduction is unlikely to occur over that hybrid material, as conrmed by low CO evolution rates obtained for CN-BA(20), Fig. 5 and Table 1.We anticipate that in a complex material such as these carbon nitrides that are rich in defect states a wide variety of trap states will exist and it is reasonable to conclude that the very small amount of CO produced from CN-BA(20)-FeTCPP samples is via a low population of photoelectrons that may be trapped in higher energy states able to transfer to the co-catalyst.

Conclusions
Here we demonstrate how a carbon nitride structure prepared via co-polymerisation of DCDA and BA can be combined with a molecular catalyst to achieve selective carbon dioxide reduction to CO (75 mmol g −1 (FeTCPP) h −1 , 69% selectivity) in an aqueous solution.By carrying out a detailed electronic structure and photophysical study we are able to identify why a 5% BA content provides optimal activity for carbon dioxide evolution.Despite higher dopant levels giving rise to further band gap narrowing and increased light absorption there is a decrease in photocatalytic activity.Structural characterization shows an increased concentration in defect states and time-resolved photoluminescence shows that although incorporation of a low level of BA is advantageous for increasing lifetime the higher BA contents have no further positive effects.Most importantly at the highest BA concentrations, there is a substantial positive shi in the CB edge which leads to there being insufficient driving force for electron transfer to the FeTCPP co-catalyst.This work therefore connects to previous studies that have shown that low levels of BA doping can also enhance photocatalytic activity for hydrogen production, 3,4,[52][53][54][55][56][57][58][59][60][69][70][71][72][73][74][75] and also provides a rationale for the similar trends in activity seen there, highlighting the importance of consideration of driving force for photoelectron transfer in these systems and as the community moves towards systems without sacricial electron donors, 76 the driving force for hole transfer during water oxidation.

Fig. 1
Fig. 1 Labelled representation of carbon nitride synthesis and structure depicting typical carbon nitride functional groups including aromatic carbon and sp 2 -hybridised N (C-N]C) within the heptazine rings (C a and N a ) and tertiary N (N-(C) 3 ) that bridge between and lie at the centre of the heptazine rings (N b and N c ).The figure also shows potential/proposed defects such as C-NH x in primary and secondary amine groups (C b , C c , N d and N e ), C]C-H functionality (C d ), C^N groups (C e and N f ), N-CH]N groups (C f ) and O-C]O groups (C g ).

Fig. 4
Fig. 4 Rate of H 2 evolved in mmol per gram of photocatalyst per hour for 2 mg of photocatalyst in 2 mL 10 mM EDTA with 1 mL H 2 PtCl 6 (8 wt% solution) over 4 hours visible light irradiation at ∼100 mW cm −2 in a glass vial.

Fig. 6
Fig. 6 UV-vis diffuse reflectance spectra (Kubelka-Munk) of HSA-CN-DCDA pre-and post-soaking in FeTCPP solution and UV-vis spectra of 90 mM FeTCPP ethanolic solution overlaid with rate of CO evolution in mmol per gram of FeTCPP per hour for 2 mg photocatalyst in 2 mL 10% TEOA (v/v) in water.The blue arrows indicate the applicable y-axis.Samples were purged with CO 2 and placed under monochromatic illumination (365, 395 and 420 nm) at 0.16 mW cm −2 in a quartzcuvette, for 24 hours on a 150 W Xe lamp with monochromator.

Table 1
Rate of CO and H 2 evolution for HSA-CN/FeTCPP materials