Synthesis of CD and CD-AQ. CDs were initially prepared via a hydrothermal reaction using citric acid (CA) as a carbon source in the presence of ethylene diamine (EDA). EDA acts as both a nitrogen source and a surface-passivating agent in this reaction (Fig. 1a). The as-synthesized CDs showed high aqueous stability with a ζ-potential of +5.2 mV at pH 7 owing to the presence of functional amine groups on the CD surface derived from EDA. In order to introduce AQ molecules as a co-catalyst, we employed N-ethyl-N’-(3-dimethylaminopropyl)carbodiimide methiodide (EDC) to mediate the surface functionalization of CDs through the reaction between amines and carboxylic acid-containing AQ molecules. Considering the solubility of all reactants, ethanol was used as a suitable solvent for the functionalization.
High-resolution transmission electron microscopy (HR-TEM) analysis revealed the nearly spherical morphology of CD and CD-AQ with an average diameter of 4.5 ± 0.8 nm and 9.9 ± 2.3 nm (Fig. 1b and Fig. S1 in the Supplementary Information). The bandgap energy (Eg) and energy levels of the conduction band (CB) and valence band (VB) of CD and CD-AQ were respectively determined by using Tauc’s plot from UV/vis spectra and ultraviolet photoelectron spectroscopy (UPS) (Fig. S2 and S3). Interestingly, there was no significant difference in the bandgap energies (3.2–3.3 eV) or in the energy level of CB and VB regardless of the types of CDs used.
Successful functionalization by AQ was also monitored by UV/vis and photoluminescence (PL) spectroscopy (Fig. 1c and 1d). The CDs exhibited characteristic absorption peaks at 237 nm (π–π* transition of sp2-carbon network) and 341 nm (n−π* transition of carbonyl groups), confirming the presence of various surface functional groups. The AQ molecules used for the functionalization displayed distinct absorbance at 258 nm and 335 nm owing to the π–π* transition of benzoid and quinoid.21 The spectra of the CD-AQ displayed the presence of absorption peaks of both CD and AQ molecules, indicating successful surface functionalization. The characteristic PL spectra of both CD and CD-AQ were found to depend on the excitation wavelength (Fig. S4). The maximum emission of CD was observed at 429 nm upon excitation at 335 nm, while that of CD-AQ was located at 434 nm upon excitation at 345 nm. In addition to shorter PL lifetime of CD-AQ than CD obtained by time-correlated single photon counting (TCSPC) measurements (Fig. S5), the PL intensity of CD-AQ was decreased by 87% upon conjugation with AQ on the surface of CD (Fig. 1d), indicating the efficient transfer of the photoexcited charge carriers from CD to AQ.
The successful functionalization of CD-AQ was further investigated by Fourier transform infrared (FT-IR) spectroscopy and X-ray photoelectron spectroscopy (XPS) (Fig. 2). In the FT-IR spectra of both CD and AQ molecules, a peak at 1700 cm-1 was observed corresponding to the stretching vibration of C=O (Fig. 2a). However, the C=O stretching vibration peak of CD-AQ shifted slightly to 1656 cm-1, indicating the successful amide coupling between CD and AQ.22 In addition, a new peak attributed to the amide bond was also observed in 3300 to 3480 cm-1 in the XPS spectrum of the CD-AQ.23 In the C 1s XPS spectra, the peak intensity of the oxygen-containing groups at a higher binding energy such as C-OH, C=O, and COOH decreased, while the peaks at 284.2 and 285.6 eV assigned to the C=C and C-N, respectively, increased after functionalization of AQ on CD owing to the formation of an amide bond. However, a simple mixture of CD and AQ (in the absence of the EDC agent) did not indicate the formation of an amide group, suggesting a critical role of EDC in the coupling reaction of CD-AQ (Fig. 2b). In addition, a new characteristic N 1s peak at 402 eV was observed for the CD-AQ, which was assigned to the amide bond without noticeable changes in the other N-configurations of graphitic-, pyridinic-, and pyrrolic-N (Fig. 2c). The peak shift of N1s of CD-AQ to a lower energy compared to pristine CD and CD+AQ (mix) is also attributed to the electron delocalization by amide coupling with AQ molecules, which can also offer a strong evidence for successful functionalization of AQ on CDs. Taken together, the successful chemical functionalization of AQ on CD to afford a CD-AQ was confirmed. Meanwhile, the relative mass ratio of AQ to CD was determined to be 1.7 in the CD-AQ nanocomposite based on UV/vis spectroscopy (Fig. S6).
Photocatalytic production of H2O2 with CDs. The molar ratio of the CD precursors CA to EDA, for instance, significantly affects the light absorption and photocatalytic properties of CDs.24 Thus, before investigating the photocatalytic performance of as-synthesized CD-AQ, we first evaluated the performance of CDs while varying the molar ratio of their precursors. All CDs were successfully synthesized, showing a characteristic n−π* transition peak of carbonyl groups at 341 nm (Fig. S7). Under solar-simulated light illumination, H2O2 was produced with an initial production rate of 148 (CA:EDA ratio of 5:2), 423 (5:5), and 89 μmol g-1 h-1 (5:7) for CDs prepared with a different molar ratio of CA to EDA, respectively (Fig. 3a). Interestingly, the highest photocatalytic performance was observed for CD with an equimolar ratio of CA to EDA (Fig. S8). This can be accounted for by considering the high degree of carbonization with a higher fraction of C=C bond in reference to C-N in C 1s XPS spectra than those of CDs prepared from other precursor ratios (Fig. S9), resulting in enhanced light absorption under a wide range of wavelengths. Hereafter, CD prepared at a ratio of 5:5 was selected for further experiments owing to its high photocatalytic H2O2 production.
In the absence of light irradiation and O2 source, CD showed a negligible photocatalytic performance for H2O2 generation, indicating that the reaction required to produce H2O2 is a photoresponsive catalytic reaction through the ORR and consumes O2 (Fig. 3b and Fig. S10). We also conducted the photocatalytic H2O2 production on CD in the presence of 1 mM of CA and EDA to evaluate the effect of the residual precursors in as-prepared CDs because these precursors are also known to serve as electron donors in conventional H2O2 production processes.25 However, the amount of H2O2 generated over 2 h decreased notably in the presence of each and both precursors (i.e., CA and EDA), with corresponding performances of 699.8 μmol g-1 (CA only), 711.2 μmol g-1 (EDA only), and 177.8 μmol g-1 (both CA and EDA), respectively, compared to 845 μmol g-1 for the bare CDs. The results indicate that the presence of residual CA and EDA interferes with the photocatalytic performance of CD instead of accelerating H2O2 production by acting as electron donors. It is inferred that CA and EDA are adsorbed on CDs by interacting with functional groups may interrupt surface reactions (e.g., substrate adsorption, electron transfer from donors and to acceptors) by covering the active reaction sites on the surfaces of the CDs.
Most semiconductor photocatalysts cannot achieve the H2O2 production and water oxidation simultaneously in the absence of sacrificial electron donors due to insufficient oxidation power and inefficient charge carrier separation.26 Thus, much effort has been made towards developing photocatalysts that enable the efficient H2O2 production only from water and molecular oxygen by tuning their structural, chemical, and electrical properties.27-30 As one of the most efficient photocatalysts for H2O2 production owing to its unique chemical process through the formation of a superoxo radical and 1,4-endoperoxide species,9, 10 C3N4 generated H2O2 at a rate of 606.7 μmol g-1 h-1. However, the CD prepared in this study demonstrated an even higher H2O2 product rate of 631.8 μmol g-1 h-1 in the presence of ethanol as a sacrificial electron donor. Most interestingly, there was negligible production of H2O2 in C3N4 (69.6 μmol g-1 h-1), in the absence of ethanol electron donor, while CD could efficiently produce H2O2 (609.4 μmol g-1 h-1) in the same conditions (Fig. 3c). The significant decrease in H2O2 production on CD in the presence of both CA and EDA supports the finding that these potential impurities merely act as retardants for H2O2 production, rather than as electron donors (Fig. 3d).
Photocatalytic production of H2O2 with CD-AQ. We further functionalized CDs with AQ molecules, which improved the photocatalytic effect of hybrid CD-AQ. In addition, we optimized the photocatalytic performance of CD-AQ via varying degrees of surface functionalization of AQ molecules on the surfaces of the former (Fig. S11). Surprisingly, the kinetics of H2O2 production improved significantly for highly functionalized CD-AQ (1187.8 μmol g-1) compared to bare CD (845.4 μmol g-1) under solar-simulated light illumination for 2 h (Fig. 4a). Although AQ itself can produce H2O2 under UV illumination via an irreversible photochemical reaction,31 the photochemical production of H2O2 using bare AQ was merely 237.2 μmol g-1 even at a much higher concentration of 0.5 g L-1. This enhanced performance by functionalization with AQ agrees with our previous finding that the AQ accelerated the photocatalytic H2O2 production of C3N4 by acting as a co-catalyst.16 The combination of CD and AQ in the CD-AQ hybrid is unique and beneficial for efficient dual-photocatalytic process for H2O2 production, not only because it introduces other catalytic active sites, but also because it promotes efficient charge carrier separation, as demonstrated by the decreased PL intensity and lifetime of CD-AQ (Fig. 1d and Fig. S5) as well as the enhanced apparent quantum yield (AQY) of CDs upon functionalization with AQ; the AQY profile of CD-AQ analogously resembles its absorption spectrum and is more pronounced than that of CD regardless of the excitation wavelengths (Fig. S12).
The pH-dependent photocatalytic effect of CD-AQ for H2O2 production was investigated because the photocatalytic ORR is governed by the proton-coupled electron transfer (PCET) reaction as in the following reaction (1):
where the photoexcited electrons from the photocatalyst react with molecular oxygen and protons in the solution. As decreasing pH, therefore, H2O2 production rate of CD-AQ increased considerably, to 2512 μmol g-1 h-1 at pH 1, which was an approximately 4-fold increase compared to pH 6 (Fig. 4b and Fig. S13), outperforming other carbon-based photocatalysts for H2O2 production reported to date (Tables S1). This tendency clearly supports the fact that the formation of H2O2 on CD-AQ is based mainly on the photocatalytic ORR under solar-simulated light illuminations. Similar trends have been observed with other photocatalytic systems used for photocatalytic H2O2 production.32
Long-term stability is another critical requirement in the development of efficient photocatalysts. We therefore performed a long-term performance and stability tests of CD-AQ. There was a continuous increase in H2O2 concentration for a 12 h reaction by CD-AQ up to 4,700 µmol g-1 at pH 6 and 28,600 µmol g-1 at pH 3, respectively (Fig. S14). The slight decrease in the H2O2 production rate was probably due to the increase of pH (e.g., from pH 6 to pH 7.26) due to the protons consumed during the photocatalytic formation of H2O2 through the PCET reaction, and in situ photodecomposition of produced H2O2 by the absorption of UV light.33 The decomposition of H2O2 by CD and CD-AQ was further examined because in situ produced H2O2 can be decomposed by both electrons from CB and holes from VB, which are generated during photocatalytic process. Unlike the gradual decrease of initial H2O2 concentration obtained with C3N4 for 2 h of reaction, a negligible decomposition (less than 1%) of H2O2 was observed with CD and CD-AQ (Fig. S15). Furthermore, there was no change in the specific absorbance of CD after the 2 h photocatalytic process. Collectively, these results indicate that the CD-based photocatalysts are an ideal candidate for the photocatalytic generation of H2O2 owing to their outstanding performance for H2O2 evolution in electron donor-free systems as well as their superior photochemical stability with a limited sluggish decomposition of H2O2.
Electrochemical properties of CD and CD-AQ. Electrochemical analysis was conducted to verify the role of AQ molecules in enhancing the catalytic performance of CD-AQ nanocomposites. AQ exhibits distinct oxidation and reduction peaks and an extremely fast two-electron redox reaction with a redox potential that shifts depending on the functional groups on the AQ.34 Therefore, we investigated the cyclic voltammetry (CV) of CD and CD-AQ in the potential range from -1.2 to 0.0 V vs. Ag/AgCl (Fig. S16). The CV curve of CD-AQ showed an anodic peak at -0.52 V and a cathodic peak at -0.60 V, corresponding to the redox peaks of AQ, whereas that of CD showed a rectangle shape associated with a non-Faradaic current (without distinct peaks). These results imply that AQ was successfully functionalized on the CDs, thus CD-AQ showed the higher current density than that of CD for ORR in O2-saturated condition. As a result, the electrochemical kinetics of electrons transferred within CD-AQ nanocomposites can be affected by the functionalized AQ.
To evaluate the electrochemical kinetics of CD and CD-AQ toward the ORR, we used a rotating ring disk electrode (RRDE) to examine linear sweep voltammetry (LSV) (Fig. 5). The CD-AQ exhibited a ring and disk current density that was more than twice the current density of the CD (Fig. 5a and 5b). The average electron transfer number of CD was calculated as 2.4 and the average efficiency of H2O2 production was 78.4% in the potential ranges from -0.80 to -0.40 V (vs. Ag/AgCl), suggesting that CD is responsible for the two-electron ORR pathway (Fig. 5c). The two reduction plateaus observed in LSV curve can be a key evidence for electrochemical H2O2 production. The electrochemical ORR performance of CD is known to be governed through consecutive two-electron steps upon generation of H2O2 as an intermediate by slow kinetics on carbon surfaces containing chemical functional groups.35 In addition, the imbalance in charge density also increases the oxygen adsorption on the surface of CD, thereby increasing oxygen solubility. It also participates in the release of H2O2 as an intermediate. Interestingly, CD-AQ showed not only highly enhanced electrochemical kinetics with a lower charge transfer resistance (Rct) measured by electrochemical impedance spectroscopy (EIS) (Fig. S17 and Table S2), but also increased selectivity compared to CD alone, with an average electron transfer number of 2.3 and an average H2O2 selectivity of 80.6% as the rotating speed in the same potential (Fig. 5d). These results indicate that the chemical functionalization of AQ onto CD creates a more efficient catalyst for generating H2O2 by enhancing the facile charge transfer and facilitating the selective two-electron pathway of the ORR.
Mechanism of the CD-AQ. The structural advantages of versatile CD-based catalysts (containing a unique sp2/sp3 hybrid carbon structure with various oxygen- and nitrogen-functional groups) facilitate and promote various chemical reactions, which suggests their widespread utility in a number of applications.19, 36-39 In this regard, we monitored the time profile of specific chemical compositions of CD during the photocatalytic process to clarify the underlying mechanisms behind the outstanding performance of CD-AQ in the evolution of H2O2 (Fig. 6b and Fig. S18).
The deconvoluted high-resolution C 1s spectra collected from XPS clearly revealed that the oxidative conversion of surface functional groups of CD took place during photocatalytic H2O2 production in which the C-OH group was diminished considerably whereas the C=O group was increased at the same time. This result agrees with our previous finding that the specific transformation of surface functional groups on CD, from hydroquinone to benzoquinone, was involved in and promoted the formation of metal nanoparticles onto the surface of CD by reducing metal ions under UV illumination.20, 40 Moreover, hydroxyl and hydroquinone groups are known to be very effective for the formation of H2O2 through ORR in the presence of molecular oxygen.16
In order to further elucidate the proposed mechanism and active sites on the surface of CDs, we conducted density-functional theory (DFT) calculation to investigate the adsorption energy and free energy of carbon surface like graphene nanoribbon as a model of CD and AQ-anchored carbon surfaces. In general, the two-electron ORR pathway follows two-step reactions for adsorption and desorption of oxygen with the formation of intermediate *OOH as in step 1 and 2:
O2 + H2O + e- + * à *OOH + OH- (step 1)
*OOH + e- à HO2- + * (step 2)
where * is an active site and *OOH is a key intermediate that plays a pivotal role in overall two-electron ORR performance. As a result, the adsorption energy of *OOH (ΔG*OOH) can be a descriptor with corresponding the thermodynamic limiting potential (UL) for determining the active sites. An ideal catalyst for H2O2 production has a ΔG*OOH of 4.22 eV, which provides the highest activity with the thermodynamic equilibrium potential (U0 = 0.70 V) in the activity volcano plot. Thus, we calculated the energy barrier in terms of free energy (Fig. 6c and S19) and adsorption energy (Fig. 6d) of CD and CD-AQ. The corresponding atomic structures of the examined CD and CD-AQ are shown in Fig. 6e. The ORR on AQ showed a thermodynamically uphill reaction by high free energy of 0.728 eV, resulting in low adsorption of OOH on AQ anchored carbon matrix. Although AQ method is commonly used in industrial H2O2 production, it is because AQ mechanism is not a competitive pathway for carbon-based catalysts and AQ follows different catalytic mechanism reported in other studies using various quinone-enriched molecules with similar DFT results of AQ molecules.41 On the other hand, the oxidized CD surface (herein a form of CD-ketone for DFT calculation) exhibited the highest activity compared to bare CD and other carbon matrix, which suggests a plenty of oxygen functional groups on the surface of CD can serve as thermodynamically active sites in accord with experimental XPS results.
From these results, we propose the possible mechanisms that retain a number of elementary reaction steps occurring on the surface of both CD and CD-AQ systems to form H2O2 (Fig. 6a). Specifically, CD initially generates the photoexcited electrons and holes owing to its appropriate bandgap energy. The photoexcited electrons are then transferred to the active sites (hydroxyl groups) on the surface of CDs, resulting in the reduction of O2 that produces H2O2 with the surrounding protons. This step is followed by the conversion of the hydroxyl groups to carbonyl groups on the surface of CDs. These protons can not only be supplied by the dissociation of nearby carboxylic acid groups; they can also be generated by hydroxyl groups from the CDs themselves in the absence of electron- and proton donors such as ethanol (Fig. S20). This interpretation is also based on the poor performance of C3N4 in the absence of ethanol due to the lack of surface functional groups (Fig. 3c).
Considering the results from the reduced PL and electrochemical analysis, the role of AQ in CD-AQ for H2O2 production was proven to be that of a co-catalyst that promotes charge carrier separation and selectivity for H2O2 formation through the sequential hydrogenation- dehydrogenation of AQ. In CD-AQ, photoexcited electrons are transferred efficiently from CD to AQ molecules. O2 is then reduced by a reversible reaction between AQ and AHQ with electrons and protons donated from the CDs or ethanol.