Ana
Garcia-Mulero
a,
Abdullah M.
Asiri
b,
Josep
Albero
a,
Ana
Primo
*a and
Hermenegildo
Garcia
*ab
aInstituto Universitario de Tecnología Química, Consejo Superior de Investigaciones Científicas-Universitat Politecnica de Valencia, Universitat Politecnica de Valencia, 46022 Valencia, Spain. E-mail: aprimoar@itq.upv.es; hgarcia@qim.upv.es
bCenter of Excellence in Advanced Materials Research, King Abdullah University, Jeddah, Saudi Arabia
First published on 15th July 2022
Microporous graphitic carbon (mp-C) derived from the pyrolysis of α-, β-, and γ-cyclodextrins exhibited photocatalytic activity in CO2-saturated acetonitrile–water upon irradiation with UV-Vis light and in the presence of triethanolamine, forming H2 (19 μmol h−1) and CO (23 μmol h−1) accompanied by a lesser proportion of CH4 (4 μmol h−1). The most efficient was the mp-C material derived from α-cyclodextrin (mp-Cα) and having a pore dimension of 0.68 nm. The process also occured, although to a much lesser extent, under simulated sunlight or with UV-Vis irradiation in the absence of a sacrificial agent, with H2O being the electron donor. The origin of the CO was proved by isotopic 13C labelling experiments. Photocurrent measurements proved the occurrence of charge separation and the increase in photocurrent intensity in the presence of CO2. Transient absorption spectroscopy was used to detect the charge separate state decay in the microsecond time scale and proved that a fraction of the photogenerated electrons were able to react with CO2.
In this context, defective doped graphenes have shown activity for H2 generation.11 More recently, microporous graphitic carbon (mp-C) derived from cyclodextrins has also shown photocatalytic activity for hydrogen generation and oxygen evolution.12 Theoretical calculations indicate that oxidation sites can be associated with the residual population of oxygen atoms in microporous graphitic materials.13 In addition, these calculations also indicate that confinement inside the micropores can facilitate bond cleavage in a kind of preactivation when the size of the molecules fit tight within the graphenic wall of the microporous carbon.13 Photocatalytic hydrogen evolution activity was also found to be increased upon N- or P-doping as a consequence of band alignment, thus the resulting mp-(N)C and mp-(P)G powders were able to even promote overall H2O splitting.
Going further in the study of the photocatalytic activity of this type of metal-free microporous graphitic carbon, it would be of interest to determine if these novel materials can also promote the photocatalytic CO2 reduction, a reaction that is more challenging than hydrogen generation from water due to the need for larger reduction potentials and the large range of products that can be formed. Herein, it is reported that, in the absence of any metal, microporous graphitic carbon materials exhibited photocatalytic activity for selective CO2 reduction to CO in the presence of sacrificial electron donors and that this photocatalytic activity followed the trend observed for overall water splitting, with the most efficient material being the one derived from the smallest α-cyclodextrin.
The specific surface area of microporous carbons was measured by isothermal CO2 adsorption at 273 K. At this point, it should be noted that in contrast to CO2 mp-Cα did not adsorb N2, a fact that was attributed to its pore size (0.64 nm) being close to the N2 kinetic diameter in comparison with the pore sizes of mp-Cβ and mp-Cγ, which were 0.78 and 0.99 nm, respectively. In the cases of mp-Cβ and mp-Cγ, isothermal N2 adsorption was used to estimate the pore-size distribution (see Fig. S1†), which corresponded in both cases predominantly to dimensions below 20 nm. A remarkable revelation was the significantly much larger area of mp-(P)Cα. It is proposed that the acidity of H3PO4 used as the P precursor was responsible for some additional corrosion of the graphenic walls during the pyrolysis process, leading to a three-fold specific surface area increase.
The Raman spectra of the mp-C samples were almost coincident. As an example, Fig. 1 shows a representative example, while Fig. S2 in the ESI† gathers the Raman spectra of all the samples under study. Three signals corresponding to 2D (broad), G, and D bands appeared at about 2700, 1590, and 1350 cm−1, respectively. These three signatures were characteristic of defective graphenes derived from oligo or polysaccharides. Doping was not reflected in any detectable change in the Raman spectra.
The high crystallinity of mp-C was reflected in the XRD patterns of the powders. Fig. 1 also presents a selected example, while the collection of XRD patterns are gathered in Fig. S3 in the ESI.† Broad peaks corresponding to the loose stacking of graphene sheets at about 26°, 45°, and 61°, corresponding to the 002, 004, and 110 face diffractions were recorded. In addition, a new diffraction peak at short angles that was attributed to the presence of the micropores was also recorded for mp-C.
The morphology of the particles was determined by field-emission scanning electron microscopy (Fig. 2 and Fig. S4 in the ESI†). It was observed that upon pyrolysis, the cyclodextrin powders formed a thick crust carbon residue with a smooth surface due to the melting of the cyclodextrin powder and transformation into graphitic carbon. Inspection at a higher magnification of the crust revealed the presence of granules arranged into the form of tubular agglomerates, with most of them longer than one micron length and about 80 nm wide, aligned perpendicular to the surface. Transmission electron microscopy revealed the presence of small particles of about 10–20 nm and microporous channels. Fig. 2 shows a selection of these TEM images. From the contrast between the walls and the voids, the dimensions of the cyclodextrin precursor were used to determine the dimensions of the pores in the resulting graphitic carbon.
The pore sizes were estimated as 0.68, 0.86, and 0.97 nm for mp-Cα, mp-Cβ and mp-Cγ, respectively. These values were in good accordance with the data previously reported, showing that the diffuse-reflectance-UV-Vis absorption spectra of the samples under study were also very similar, with a continuous absorption over the whole range of UV and visible wavelengths and then decreasing gradually in absorptivity towards longer wavelengths. The conjugated condensed polycyclic aromatic domains present in mp-C should be the chromophores responsible for this absorption in the visible region. Fig. 1 also includes a representative DR-UV-Vis spectrum, while the collection for all the samples is presented in Fig. S5 in the ESI.† Some relative absorption maxima were measured at 280, 400, 550, and 620 nm, coincident for all the mp-C samples. N- and P-doping modified the DR-UV-Vis spectra mostly in the UV region by introducing a more intense absorption band with a tail extending to 350 and 300 nm for mp-(N)Cα and mp-(P)Cα, respectively, while no influence was observed at longer wavelengths. From the optical absorption spectroscopy analysis, the bandgap of the samples could be estimated using Tauc plots. The optical bandgap for each mp-C sample is provided also in Table 1.
The photocatalytic activity for H2 or CO evolution followed the order mp-Cα > mp-Cβ ≈ mp-Cγ. This order is similar to what has been observed for H2 generation, and it has been interpreted as reflecting the beneficial influence of the spatial confinement and small pore size commensurate with the substrate dimensions to enhance the photocatalytic activity.12 Thus, the sample with the smallest pore size was the most active of the series since the tight fitting of CO2 inside the micropores of mp-Cα should favour electron transfer from the mp-Cα photocatalyst to the substrate. The control experiments, as shown in Fig. S7,† indicated there was negligible gas evolution in the absence of photocatalyst.
No beneficial influence of doping on the photocatalytic CO2 reduction to CO was observed, and the performance of mp-Cα was also higher than those of mp-(N)Cα and mp-(P)Cα (see Fig. 4). This was probably due to the band energy of the various samples, which were in all cases above the thermodynamic potential required for CO2 reduction to CO (see Fig. S8†).
In view of these results, one aspect that is remarkable is the higher efficiency of the materials, such as mp-Cα, constituting greater than 98% C element in the absence of any metal or even dopant element. In the state of the art, it is frequently assumed that doping in the optimal proportion should increase the photocatalytic activity of graphitic carbons, but this assumption did not seem to apply in the present set of samples. It could be that the mp-Cα contained an excessive dopant amount. Also the simultaneous co-doping with two elements could result in an enhancement of the photocatalytic activity. It appears that spatial confinement of the photocatalytic reaction inside small pores was the key factor leading to the higher photocatalytic activity of mp-Cα. The apparent quantum yield of CO formation measured for mp-Cα at 380 nm was determined as 0.02%.
The photocatalyst stability was checked by performing a set of consecutive uses with the same mp-Cα sample, whereby similar temporal evolution profiles were observed for H2, CO, and CH4 in four consecutive uses (Fig. S9†). Furthermore, TEM characterization of the mp-Cα sample after exhaustive use showed that the porosity and morphology of the sample was maintained in the process. These photocatalytic data and characterization support the photocatalyst stability under the irradiation conditions.
Under the same solvent mixture and conditions, but using solar simulated sunlight, mp-Cα was also revealed to be active, although with lesser gas evolution. The results are presented in Fig. S6 in the ESI.† Notably, CH4 (8 μmol h−1) was the major product formed under these conditions, with a significantly lesser evolution of H2 (1.3 μmol h−1) and CO (2 μmol h−1). This product dependence with the irradiation wavelength would indicate the existence of different sites in mp-C. It is proposed that CH4 was the product formed when there was strong CO adsorption as a consequence of the deeper photocatalytic reduction. If this were the case, then CO would be formed in those sites with less CO affinity, while CH4 would be on those other with stronger CO adsorption. It seems that those sites with stronger CO affinity were responsive to longer wavelengths.
To determine the origin of CO and CH4, blank controls experiments under the same conditions by replacing CO2 by Ar were performed, whereby the formation of CO or CH4 was not observed, except in the case of mp-(N)Cα and mp-(P)Cα, for which the presence of CO in minute amounts was observed. Further confirmation of CO2 as the source of CO and CH4 in the photocatalytic reactions was obtained by using 13C-labelled CO2 in the reaction with the mp-Cα photocatalyst and by analyzing the products by mass spectrometry. This technique revealed the presence of a peak at m/z 29 amu and the complete absence of a peak at m/z 28 amu (Fig. 3), providing firm evidence that all the CO was derived from 13CO2. The photocatalytic activity of mp-C to promote artificial photosynthesis, promoting CO2 reduction by H2O was finally also explored. In these measurements, the experimental conditions were similar except that the triethanolamine volume was replaced by additional H2O. In these reactions, CH3CN (12 mL) and H2O (8 mL) were used as the solvent mixture using the same mp-Cα weight (15 mg). In these reactions, the products observed were H2 (30 μmol h−1), CO (2 μmol h−1), and O2. The formation of CH4 could not be detected under these conditions. The temporal product evolution is presented in Fig. S9.† These results were in line with our previous reports showing that mp-Cα was able to photocatalytically generate H2 and O2 from H2O.12 A control experiment, in the absence of CO2 did not show CO evolution. Now if CO2 was also present, then, photocatalytic CO2 reduction also occurred concomitantly with overall water splitting. Again, these results are remarkable and in our knowledge unprecedented, since mp-Cα is composed almost purely of C and no metals or doping elements are present.
Prior characterization of the polarization curves for the mp-Cα/FTO photoelectrodes indicated that they could be used in the range of potential from 1 to −0.5 V vs. Ag/AgCl, since at −1 V vs. Ag/AgCl there was a strong discharge due to the hydrogen evolution reaction (HER). Measurements were carried out under N2- and CO2-saturated conditions (Fig. 5). Large currents were observed at 1 V vs. Ag/AgCl bias potential with no difference in the gas present in the measurements. This indicates that all the charge carriers were mobilized under these extreme conditions. In contrast, at more moderate bias potentials of 0.5 and 0.25 V vs. Ag/AgCl, although the absolute value of the photocurrents were obviously lower depending on the extraction voltage, much large enhancements by light were observed upon CO2 saturation of the electrolyte in comparison to N2 (Fig. 5). At 0 V bias potential vs. Ag/AgCl bias, no charge extraction was clearly observed and a change from positive to negative photocurrent beyond −0.25 V vs. Ag/AgCl that should be close to the conduction band potential of mp-C was measured.
Fig. 5 Photocurrent measurements, with bubbling in the electrolyte N2 shown in black and in CO2 in red. The right part corresponds to an expansion of the −0.5 V bias potential. |
Beyond this crossover polarization potential, the currents were negative and had a higher intensity in the presence of CO2.
While the photocurrents prove that mp-C undergoes charge separation upon illumination, polarization indicated that it is an n-type semiconductor, whereby the somewhat higher negativity in the presence of CO2 as compared to N2 indicated the occurrence of photocatalytic CO2 reductions, contributing to the process.
To gain information on the nature of the signal, particularly if it corresponds to a charge-separated state, quenching of the signal by O2 was studied. Fig. S10 in the ESI† shows the influence of the quenchers on the transient signal monitored at 400 nm. In the case of mp-Cα, the effect of O2 quenching was marginal with some changes in the intensity and kinetics at 400 and 480 nm, but with no appreciable change in the signal at 640 nm. This trend for O2 quenching was also present, but more remarkable for mp-Cβ. We attributed the increase in the signal intensity in the 400–560 nm region to the prevalence of h+ absorption in this region. In the presence of O2, some photogenerated electrons will be quenched by O2, leading to a lesser prompting of e−/h+ recombination and, therefore, to an excess of h+, which will be reflected in an increase in the intensity of h+ from 400 to 560 nm. The fact that mp-Cβ was more affected than mp-Cα was due to the difference in pore size.
As already commented N2 is not a suitable gas to determine the surface area and N2 and O2 have similar kinetic diameters. To further check this proposal, while providing spectroscopic evidence of the reaction of CO2 with photogenerated e−, TAS measurements were also performed in the presence of CO2. A similar behaviour was observed for O2 and CO2 with a change in the intensity of the temporal profiles and kinetics of the signal in the 480–540 nm region, but no influence at longer wavelengths, which should be mostly due to the unreactive, trapped electrons. In good agreement with our proposal, similar quenching behaviours of O2 and CO2 were observed, as presented in Fig. 7. Notably, the effect of CO2 was even higher than that of O2, in spite of it being a worse electron quencher.
Fig. 7 Deactivation kinetics monitored using N2 (black), O2 (red), and CO2 (green) at (a) 400, (b) 500, and (c) 600 nm. |
We attribute the higher influence of CO2 to its better ability to diffuse inside the pores of mp-C, as previously commented on when describing the specific surface area measurements by isothermal CO2 adsorption in comparison to N2 adsorption. Therefore, the even higher increase in the intensity of the transient signal at 400 nm was due to the better ability of CO2 to trap internal electrons in mp-Cα and mp-Cβ.
Regarding doped mp-(N)Cα, surprisingly its transient signal was very weak, meaning that not many e−/h+ reached the microsecond time scale and that the N promoted charge recombination. It is well-known in semiconductors that an excessive dopant population can promote charge recombination. Apparently, this should be the case of mp-(N)Cα. In contrast, although the signal for mp-(P)Cα was more intense than for mp-Cα and the quenching behaviour for O2 followed the same trend as for mp-Cα, no influence of the presence of CO2 could be monitored. This means that CO2 was not able to quench the photogenerated electrons.
Overall the results of the TAS study explain the photocatalytic activity results, particularly the lack of a positive influence of the dopant elements on the photocatalytic activity of mp-Cα, but for different reasons. In the case of N-doping this was due to the short lifetime of the charge separation and in the case of P-doping, it was due to the failure to transfer electrons to CO2.
The XRD pattern was recorded using a Cubix-pro PANalytical diffractometer in the range from 5° to 90° at a scan rate of 1° s−1.
Diffuse reflectance UV–vis spectra in the range of 200–800 were recorded on a Cary 5000 spectrophotometer from Varian.
Solid-state 31P NMR spectra were measured at room temperature using a Bruker AV400WB with π/2 pulse sequences of τ = 5 μs and a relaxation time of 5 s. The experiments were carried out with magic-angle spinning at the rate of 10 kHz. For obtaining the spectra, between 100 and 400 scans were accumulated.
The combustion elemental analyses were measured with a Euro EA 3000 analyzer.
TEM images were recorded on a JEOL JEM 2100F with a voltage of 200 kV coupled with an X-Max energy-dispersive X-ray detector (EDS). Also, FESEM images were obtained on a ZEISS ULTRA 55. Samples were prepared by casting one drop of the suspended material in acetonitrile (ACN) onto a carbon-coated copper TEM grid and allowing it to dry at room temperature.
The mp-C was dispersed in 30 mL of ACN (>99.9%, Sigma-Aldrich) using a sonic tip (FisherbrandTM Model 705 at 40% of 700 W for 1 h using a pulsation of 1 s on and 1 s off) in a concentration of 1.5 mg mL−1. Next, 10 mL of this suspension was introduced in to the reactor with 4 mL of triethanolamine (Sigma-Aldrich) as an electro donor and 4 mL of MilliQ water. To reach the 20 mL of mixture for the reaction, 2 mL more of ACN was added. The system was purged using pure CO2 for 10 min and pressurized until 1.4 bar (absolute pressure).
The gas products were analyzed using a gas chromatograph (Agilent 490 MicroGC) equipped with a molecular sieve 5 Å column with a TC detector and Ar as the carrier gas. Also for CO, a 7890A gas chromatograph was used equipped with a column Carboxen®-1010 PLOT L × I.D. 30 m × 0.53 mm, average thickness 30 μm with a TC detector and He as the carrier gas.
For the stability tests, the material was removed from the reactor between the reactions and washed three times by centrifugation (6000 rpm, 15 min) using MilliQ water before starting a new reaction.
The apparent quantum yield (AQY) was determined under irradiation with a 150 W Xe lamp equipped with a Czerny Turner monochromator. The AQY value was calculated by the equation:
During the measurements, N2 or CO2 were bubbled in the solution to ensure the electrolyte was completely saturated. Also, each measurement was done twice in two different electrodes to confirm the material behaviour.
Notice that the photocatalyst was deposited on the FTO using a mixture of 15 mg of ultrasonicated photocatalyst (the ultrasonication was developed in 15 mL of ethanol 40%, 1 h, 1 s on, 1 s off and dried at 60 °C) in 300 μL of ethanol and 75 μL of Nafion 5% (w:w). The electrode was dried at room temperature overnight. The characteristic of each electrode are presented in Table 2.
Electrode | Surface area (cm2) | Active mass (mg) |
---|---|---|
E1 | 0.7 | 3.4 |
E2 | 0.65 | 5.3 |
For the chromoamperometric tests, the electrode was allowed to stabilize for 5 min at the selected potential (1 V, 0.75 V, 0.5 V, 0.25 V, 0 V, −0.25 V, −0.5 V, −0.75 V or −1 V vs. the reference electrode). After that, the electrode was irradiated for 1 min using the same Xe lamp that was used in the photocatalytic tests and it was allowed to relax for 2 min. This was repeated five times for each potential to confirm the reproducibility of the tests.
For the cyclic voltammetry tests, the scan speed was 0.02 V s−1, from 1 V to −1 V vs. the reference electrode, for five cycles.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d2nr02655d |
This journal is © The Royal Society of Chemistry 2022 |