Insights into Charge Transfer at an Atomically Precise Nanocluster/Semiconductor Interface.

The deposition of an atomically precise nanocluster, for example, Ag44 (SR)30 , onto a large-band-gap semiconductor such as TiO2 allows a clear interface to be obtained to study charge transfer at the interface. Changing the light source from visible light to simulated sunlight led to a three orders of magnitude enhancement in the photocatalytic H2 generation, with the H2 production rate reaching 7.4 mmol h-1  gcatalyst -1 . This is five times higher than that of TiO2 modified with Ag nanoparticles and even comparable to that of TiO2 modified with Pt nanoparticles under similar conditions. Energy band alignment and transient absorption spectroscopy reveal that the role of the metal clusters is different from that of both organometallic complexes and plasmonic nanoparticles: A type II heterojunction charge-transfer route is achieved under UV/Vis irradiation, with the cluster serving as a small-band-gap semiconductor. This results in the clusters acting as co-catalysts rather than merely photosensitizers.

Ligand‐protected metal nanoclusters (NCs) have shown great potentials in the field of energy conversion, as their electronic structures can be precisely tuned by varying their size (i.e. number of metal atoms), composition, and nanostructure.1, 2, 3, 4, 5 Current studies on the utilization and conversion of light energy focus on the sensitization ability of metal NCs and aim to exploit the part of the solar spectrum that large band‐gap semiconductors (i.e. TiO2, ZnO) are not capable of.6, 7, 8, 9, 10 The energy absorbed from visible photons is temporarily stored within the NCs before being released or transferred into the semiconductor.11, 12, 13, 14 By varying the atomic packing of gold NCs, the lifetime of the excited state can be as long as 5 μs, which is comparable to bulk silicon.15 Extending the photoresponse of semiconductors from the UV to the Vis/IR region enables visible‐light‐driven photocatalytic and photoelectrochemical reactions, such as hydrogen generation and CO2 reduction.16, 17, 18, 19, 20, 21, 22 Despite advances in research, the role of NCs in hybrid materials is still in the early stages compared to organometallic dyes.23 Considering that metal NCs with distinct gaps between the highest occupied molecular orbital (HOMO) and the lowest unoccupied molecular orbital (LUMO) can produce excitons (electron–hole pairs) after absorbing light, it is reasonable to treat them as small‐band‐gap semiconductors.24, 25, 26 Under simulated sunlight, both the NCs and the semiconductor will be excited. Then, comes an essential question: Does the combination of metal NCs with a large‐band‐gap semiconductor show a synergistic effect on light harvesting and charge separation at the binary semiconductor heterojunction, as is evident with the natural photosynthesis system? Addressing this question could also lead to practical applications, as sunlight is the most abundant and also free renewable energy resource.

Plasmonic metal nanoparticles (NPs) represent another kind of sensitization materials which could extend the photoresponse to longer wavelengths.27, 28, 29 The position of the surface plasmonic resonance (SPR) peak can be tuned by controlling the shape and size of the NPs.29 Noble metals (e.g. Pt) with a high work function can readily extract the excited electrons from the semiconductor through the Schottky junction and use these electrons to reduce the protons adsorbed on their surface, thereby resulting in an enhanced photocatalytic performance.30, 31 However, NPs with a certain shape and size only harvest the photon energy in a narrow range within the visible and infrared region.29 To improve the utilization of sunlight, NPs with different sizes need to be integrated, but this results in complications in the system. Moreover, the size of the NPs affects their ability to extract electrons from the semiconductor.32, 33, 34 The smaller the size of the NPs, the higher the energy level of the NPs.32 Once the energy levels of the NPs become higher than the conduction band (CB) of the semiconductor, the NPs will lose their ability to extract electrons.33 Once the NPs reach a critical size, they will capture both electrons and holes, thereby acting as a recombination center.34 Furthermore, the larger the particle size, the smaller the specific area, which leads to a lower exposure to the active center.2 The discrete energy levels of metal NCs endow them with multiple absorptions within the UV/Vis/IR region.14 Moreover, they have already proven to be catalytically active.35, 36, 37, 38, 39 This provides an opportunity in photocatalytic systems for the metal NCs to serve as both the light absorber and co‐catalyst.

With these motivations, we deposited atomically precise silver NCs Ag44(SR)30 (denoted as 1, SR=thiolate) on the large‐band‐gap metal oxide semiconductor TiO2 and studied the photocatalytic H2 production performance. By changing the excitation energy from only the visible to the full solar spectrum, the charge‐transfer pathway of the system can be readily tuned to a type II photocatalysis mechanism. In addition to the extension of the photoresponse, the synergistic effect between the NCs and TiO2 was found to contribute significantly to the enhanced catalytic performance. Thus, the bifunctional role of NCs as the light absorber and co‐catalyst for photocatalytic H2 production was revealed.

The synthesis of 1 is described in the Experimental Section. Five apparent absorption bands at λ=412, 483, 535, 641, and 833 nm are observed (Figure 1 a).40 The high molar extinction coefficient (ϵ) of 1 indicates its good light‐harvesting ability (Figure S1).14 For example, the ϵ value for the absorption at λ=483 nm is 1.1×105 cm−1 M−1, four times higher than that of commonly employed gold NCs.41 It should be noted that previous studies show that the HOMO–LUMO gap of 1 (0.77 eV) is smaller than the optical band gap, because the electron transition from the HOMO to the LUMO (superatom 1D to 2S) is optically forbidden.40, 42, 43 The electrospray ionization (ESI) mass spectrum (see Figure S3) of 1 showed a peak at m/z 2274.2, and its isotopic patterns were in perfect agreement with the simulated pattern, thus indicating 1 carries four negative charges (Figure S4).40 Figure 1 b shows the UV/Vis spectra of the TiO2 film before and after the deposition of 1. The enhanced absorption of 1‐TiO2 in the visible region is attributed to 1. Multiple bands at λ=380, 500, and 600 nm, rather than an SPR absorption band of Ag NPs, indicate that the NCs maintain their size after deposition onto a TiO2 film.40

Figure 1

a) UV/Vis absorption spectrum of Ag44 NCs (1) in solution. b) UV/Vis absorption spectrum of the TiO2 film before and after deposition of 1. c) Secondary electron cutoff for the TiO2 film before (black) and after (red) deposition of 1. d) The shift of Ti 2p3/2 after the deposition of 1. e) Valence band XPS spectrum of pristine TiO2 film. f) Fluorescence spectra of TiO2 film (black), 1‐TiO2 (red), and 1 (blue) in terephthalate solution irradiated by UV/Vis light.

To assess the solid‐state properties of the cluster/TiO2 composite, the energy band alignment of 1‐TiO2 was measured with ultraviolet photoelectron spectroscopy (UPS), which is a helpful way to study whether metal NCs with distinct energy gaps can be treated as small‐band‐gap semiconductors. The secondary electron cutoffs (SECOs) were measured to be 3.9 eV and 4.7 eV before and after deposition of the clusters, respectively. This indicates an increase in the work function (Δϕ) of 0.8 eV (Figure 1 c). The Ti 2p3/2 core level band shifts to a lower binding energy upon deposition, thus pointing to an upward bending of the band (ΔϕBB of 0.7 eV (Figure 1 d). Thus, the contribution to the change in the work function from the formation of an interface dipole (ΔϕID) is small (0.1 eV).44 From the UPS spectra (Figure 1 e), the valence band maximum (E VBM) of TiO2 was found to be located at 3.2 eV (below the Fermi level, E F). The HOMO of 1 was measured to be at 0.7 eV (Figure S5). Application of the equation E LUMO=E HOMO–LUMO gap−E HOMO therefore computes the LUMO of 1 to be 0.07 eV above E F.40

Figure 2 a shows the energy level diagram of the cluster/TiO2 system deduced from the above results. The contact of 1 to TiO2 leads to bending of the TiO2 surface band and thus a change in the near‐surface space‐charge region. Electrons in TiO2 will transfer to 1 until their Fermi levels are electrically aligned. Consequently, the formation of an internal built‐in electric field is beneficial for the separation and transport of carriers. The interaction between 1 and TiO2 alters the energy levels of 1, which accounts for the broadening of the bands of 1‐TiO2 in the visible region, compared with pure 1 (Figure S6).45 When both TiO2 and 1 are excited, the photoinduced holes in the valence band of TiO2 will be readily transferred to the HOMO of 1, thereby separating them from the photoinduced electrons. This is similar to the scenario in a straddling semiconductor heterojunction.46, 47 Note that the observed absorption band of 1 at λ=833 nm is attributed to electron promotion from the HOMO to LUMO+1.40 Thus, the position of LUMO+1 is calculated to be 0.8 eV above the E F, which is higher than the conduction band minimum (CBM) of TiO2.40 Therefore, it is expected that when electrons in 1 are photoexcited to LUMO+m (m≥1) levels, electron–hole separation is processed through a staggered type II pathway. In this case, electrons in 1 will transfer to the CB of TiO2 and, simultaneously, holes at TiO2 will inject into the HOMO of 1. This process can reduce the resistance to charge transfer and improve the efficiency of the charge separation, which will further result in an improved photocatalytic performance.47

Figure 2

a) Schematic energy level diagram of TiO2 before (black) and after (orange) deposition of the NCs. All numbers are given in eV. b) Catalytic hydrogen evolution by photocatalysts under UV/Vis light (100 mW cm−2). TiO2 NPs: black, TiO2 NPs modified with Ag NPs: orange, and 1‐TiO2(p) NPs: red. c) Cycling test over 1‐TiO2(p) NPs under UV/Vis light (100 mW cm−2).

The distribution of holes under UV/Vis irradiation was studied by probe experiments. Nonfluorescent terephthalate (TPA) was used as a probe for detecting the formation of hydroxyl radicals, which are one of the products of water oxidation (.OH/H2O, 2.3 V versus NHE).48 For a solution of TPA with TiO2, a photoluminescence (PL) band at λ=435 nm appears and increases gradually as the illumination continues (Figure S7). As the edge of the TiO2 valence band (VB) is more positive than the potential of .OH/H2O, it can be deduced that, under irradiation, the holes in the VB of TiO2 oxidize H2O to .OH, which further reacts with TPA to produce highly fluorescent hydroxyterephthalate (HTPA).48 The solution of TPA with clusters shows negligible PL after irradiation, in good agreement with the observations that the HOMO of 1 is less positive than the VB of TiO2 (Figure 1 f). The solution of TPA with 1‐TiO2 shows a much weaker PL band at λ=435 nm compared to TiO2. This supports the hypothesis that photoinduced holes in TiO2 transfer to the HOMO of 1.

To verify the superiority of the composite in photocatalytic reactions, the NCs were loaded on commercial TiO2 NPs at a weight percentage of 2 %. For distinction, the particulate 1‐TiO2 photocatalyst is defined as 1‐TiO2(p). The matrix‐assisted laser desorption ionization (MALDI) mass spectrum of 1‐TiO2(p) shows a broad signal around m/z 8900 (assigned to fragments of 1), thus indicating the successful deposition of 1 on the TiO2 NPs (Figure S8). The photocatalytic hydrogen evolution performance of the 1‐TiO2(p) NPs was first examined under irradiation with visible light, with reference to the pure TiO2 NPs (Figure S9). Whereas the latter shows no hydrogen production, the activity increases when 1 is coupled to the TiO2 NPs, with 0.07 μmol H2 detected after 1 h (see Figure S9). The H2 production rate increases over the next three hours at an average rate of 8.4 μmol h−1 gcatalyst −1. Considering the fact that no Pt NP co‐catalysts are employed in this case, 1 seems to be a promising photosensitizer compared to the previously reported gold NCs.18 Surprisingly, the 1‐TiO2(p) NPs exhibit a three orders of magnitude enhancement of H2 production when the light is switched from visible light to simulated sunlight with equivalent intensity (Figure 2 b). The rate of hydrogen production under simulated sunlight reaches an average of 5.5 mmol h−1 gcatalyst −1 in 6 h, with a maximum of 7.4 mmol h−1 gcatalyst −1. This performance is ten times higher than that of the pure TiO2 NPs and five times higher than that of TiO2 NPs modified with Ag NPs. The rate of hydrogen production by the 1‐TiO2(p) photocatalytic system is, in fact, comparable to that of the TiO2 NPs modified with Pt NPs under the same conditions (Figure S10). Such a large improvement in the photocatalytic performance can not be attributed solely to the photoresponse extension of the TiO2 NPs alone. In addition, 1‐TiO2(p) NPs exhibit good stability, with 83 % of the photocatalytic activity remaining after five cycles (Figure 2 c). The Ag 3d XPS spectra of 1‐TiO2(p) before and after the reaction show no apparent change (Figure S11).49, 50 Furthermore, no larger nanoclusters were observed in a wider range mass spectrometric analysis (e.g. from m/z 10 000 up to 35 000) after the reaction (Figure S8).

Since methanol serves as the sacrificial electron donor, carbon dioxide instead of molecular oxygen is detected as the oxidation product.18 Besides methanol, other hole scavengers such as ethanol and ethylene glycol were employed, but 1‐TiO2(p) exhibits its best performance when in an aqueous methanol solution (see Figure S12). This result is attributed to the easier oxidation of methanol, which captures the holes timely and thus suppresses the charge‐recombination processes.51

Photoelectrochemical (PEC) measurements were further carried out to study the charge generation and separation processes. Figure 3 a shows the transient photocurrent responses of 1‐TiO2(p) under intermittent irradiation with UV and UV/Vis light. For comparison, the light intensity was fixed at 100 mw cm−2. The higher photocurrent density under irradiation with UV/Vis light suggests a higher charge‐separation efficiency.52, 53 An electrochemical impedance spectroscopy (EIS) study further supported this conclusion. As shown in Figure 3 b, the Nyquist plot of 1‐TiO2(p) under UV/Vis light exhibits a more depressed semicircle in the low‐frequency region than under visible light, thus indicating that 1‐TiO2(p) experiences a more efficient interfacial electron‐transfer process under UV/Vis light.20 This indicates that the charge transfer pathways under irradiation with visible and UV/Vis light are different. The smaller charge‐transfer resistance of 1‐TiO2(p) (Figure 3 c), compared with the TiO2 NPs, is consistent with the transient photocurrent response data (Figure S13) and the band alignment discussion above. A comparison of the Mott–Schottky plots for the pure TiO2 NPs and 1‐TiO2(p) are presented in Figure 3 d. Both curves show a positive slope, thus indicating that the deposition of the NCs on the TiO2 NPs does not change the n‐type behavior of the TiO2.20 The flat band potential of TiO2 NPs and 1‐TiO2(p) are estimated to be −0.33 and −0.68 V versus Ag/AgCl, respectively. The cathodic shift of the flat band potential is consistent with the conclusion reached in the photoelectron spectroscopy study that the deposition of 1 results in an upward bending of the TiO2 band. Figure 3 e shows the PL spectra of the pure TiO2 NPs and 1‐TiO2(p). The band from TiO2 at λ=430 nm is almost quenched upon deposition of 1, which suggests that the recombination of the photogenerated electron–hole pairs over 1‐TiO2(p) is effectively suppressed.54

Figure 3

a) Transient photocurrent responses of 1‐TiO2(p) NPs under UV/Vis and visible light. b) Nyquist phots of 1‐TiO2(p) NPs under different light sources. c) Nyquist plots of TiO2 NPs and 1‐TiO2(p) NPs under UV/Vis light. The UV/Vis light and visible light have the same intensity of 100 mw cm−2. d) Mott–Schottky plots of TiO2 NPs and 1‐TiO2(p) NPs in the dark. e) PL spectra of TiO2 NPs before and after the deposition of 1. f) Absorbed photon to current efficiency of pristine TiO2 NPs and 1‐TiO2(p) NPs in the UV region.

The increased light absorption in the UV region definitely contributes to the enhanced photocatalytic performance. The role of 1 in the charge separation should also be considered. To eliminate the influence of 1 absorbing light in the UV region, the absorbed photon to current efficiency (APCE) of 1‐TiO2(p) and the pure TiO2 NPs were compared (Figure 3 f). The APCE at λ=300, 310, 320, 330, and 340 nm are 10.9 %, 8.9 %, 8.4 %, 8.1 %, and 7.8 %, for 1‐TiO2 and 8.0 %, 5.3 %, 4.7 %, 4.4 %, and 3.9 % for TiO2. The higher APCE of 1‐TiO2(p) indicates that it can convert the absorbed photons into current more efficiently, which agrees with the EIS and band alignment study results.14

To gain further insight into the electron dynamics of 1‐TiO2 in real time, we relied on ultrafast transient absorption (TA) spectroscopy. Excitation at λ=380 nm results in the creation of an electron–hole pair on the TiO2 film within the pump‐probe correlation time (Figure S14). The signal is weak, which is why recombination in the next 100 ps is concluded from the recovery of the bleaching upon excitation at λ=325 nm (Figure S14).54 It is assumed that excitation at λexc=380 nm results in higher excited states Sn of 1‐TiO2 being initially populated, as evident by the difference in the extinction spectra of 1‐TiO2 and TiO2 (Figure S14).

The transient absorption spectra of 1‐TiO2 are shown as 2D plots in Figures 4 a,b and Figure S14. The kinetics can be summarized with the band integrals (BIs) from the data and the global kinetic analysis (Figures 4 d,e). A positive signal is caused by excited‐state absorption (ESA), and negative contributions indicate stimulated emission (SE) and bleaching. SE coincides with the S1←S0 transition and shows up within the pump‐probe correlation time; thus indicating there is an ultrafast transition from Sn towards the S1 state on the same fast time scale.

Figure 4

a,b) 2D pseudo‐color plots of the transient absorption (TA) spectra of 1‐TiO2 over different time scales. d,e) Summary of the kinetics of the TA. Early relaxation (τ=0.7 ps) is accompanied by an oscillation (ν=85 cm−1). Decay of SE over 2.4 ps might indicate electron transfer from TiO2 towards interface states. The hole–pair recombination of ca. 100 ps is suppressed in 1‐TiO2 and the lifetime exceeds 2 ns. c,f) Schematic illustration of the charge‐transfer pathways in the 1‐TiO2 photocatalytic system for H2 evolution. c) Under irradiation with visible light, only the clusters are excited. f) Under irradiation with UV/Vis light, both components are excited.

After pumping, early evolution of photoexcited 1‐TiO2 is evident within the first 0.7 ps by partial decay of the ESA (Figure 4 d, blue) accompanied by a decaying oscillation (ν=2.5 ps−1, 85 cm−1) visible between λ=414 and 440 nm (Figure 4 d, red). The oscillation frequency matches the acoustic phonons usually observed in moderately sized clusters.55 Subsequently, with a time constant of 2.4 ps, the SE band vanishes and gives rise to residual or new ESA. Previously, such behavior was interpreted as electron transfer from the conduction band of TiO2 into interface states of the metal‐organic framework/TiO2 system.54 Recovery takes place on a time scale of several nanoseconds (Figure 4 e) as can be seen by the TA spectrum not changing after two nanoseconds. The absence of any evolution related to photoexcited TiO2 shows that 1‐TiO2 can effectively shutter the electron–hole recombination.

Based on the above results, we illustrate the mechanism for photocatalytic H2 production by the cluster/semiconductor composite through a HOMO to LUMO+1 transition as an example. Other main transitions under illumination with light are summarized in Table S1. The vacuum energy in absolute values is converted into redox potential with reference to the normal hydrogen electrode (NHE).56 As illustrated in Figure 4 c, only clusters are excited under visible light. The electrons in the HOMO of the clusters are excited to the LUMO+1. As the energy level of LUMO+1 is higher than that of the CB, the photoinduced electrons are spontaneously injected into the TiO2 for H2 production. The holes stay on the clusters for oxidation reactions. In this case, the clusters only serve as a photosensitizer, similar to a dye.18 The charge‐transfer channel switches when the light source changes. Under irradiation with UV/Vis light, both the cluster and semiconductor are excited, thereby generating electron–hole pairs (Figure 4 f). With the driving force from the internal electric field, the holes in the VB of TiO2 transfer to the HOMO of 1, and electrons in 1 are injected into the CB of TiO2, following a type II charge transfer pathway. The separation of photoinduced electron–hole pairs is more efficient in the type II photosystem. Thus, it enables more electrons to reduce protons adsorbed on the photocatalyst. In addition to the photosensitizer, the cluster also serves as co‐catalyst in this case to facilitate charge separation.

The proposed mechanism for the cluster/semiconductor composite is different from the plasmonic metal nanoparticle/semiconductor system.27 Schematic illustrations of the energy level diagram and the charge‐transfer pathways in a typical TiO2 system modified with Ag NPs are given in Figure S15 for comparison. The better charge separation efficiency of 1‐TiO2(p), together with the good photosensitization ability of 1, account for the better performance of 1‐TiO2(p) than Ag NP‐TiO2 in the photocatalytic H2 generation reaction (Figures 2 b and S16).

In summary, a type II photosystem was built by the combination of TiO2 with atomically precise silver NCs. The photocatalytic rate of H2 generation by the composite is 7.4 mmol h−1 gcatalyst −1, which is ten times higher than that of pristine TiO2 NPs under the same conditions. The significantly enhanced performance is attributed to both the extension of the photoresponse and the efficient separation and transport of charge carriers. The role of the NCs as a small‐band‐gap semiconductor was revealed by the different charge‐transfer routes occurring in the composite under different irradiation conditions. Under irradiation with visible light, the NCs only serve as a light‐absorber. Under irradiation with UV/Vis light, the NCs also act as a small‐band‐gap semiconductor to improve the charge‐separation efficiency. This type II photosystem endows the NCs with the ability to function as co‐catalysts rather than merely photosensitizers. We believe that this study will stimulate more studies on using catalytically active metal NCs for the absorption of light and conversion of energy into relevant chemical fuels.

Conflict of interest

The authors declare no conflict of interest.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supplementary

Click here for additional data file.