Skip to main content
NCBI home page
JACS Au logoLink to JACS Au
. 2021 Apr 26;1(5):550–559. doi: 10.1021/jacsau.1c00004

Water Splitting with a Single-Atom Cu/TiO2 Photocatalyst: Atomistic Origin of High Efficiency and Proposed Enhancement by Spin Selection

Cheng Cheng †, Wei-Hai Fang †, Run Long †,*, Oleg V Prezhdo ‡
PMCID: PMC8395698  PMID: 34467318

Abstract

graphic file with name au1c00004_0008.jpg

Anatase TiO2 is an intensely investigated photocatalytic material due to its abundance and chemical stability. However, it suffers from weak light harvesting and low photocatalytic efficiency. Experiments show that light absorption and photocatalytic properties can be enhanced simultaneously by TiO2 doping with well-dispersed Cu atoms, forming a single-atom catalyst (Cu/TiO2) that can be used for solar water splitting and other applications. By performing ab initio nonadiabatic molecular dynamics simulations, we demonstrate that Cu/TiO2 is inactive before light irradiation due to rapid electron–hole recombination via both shallow and deep traps. Surprisingly, the shallow trap is more detrimental to the Cu/TiO2 performance than the deep trap because it couples better to free carriers. After light irradiation, leading to electron transfer and Cu/TiO2 protonation, the shallow trap is eliminated, and a local distortion around the Cu atom stabilizes the deep trap state on the Cu d-orbital, decoupling it from free charges and giving rise to high photocatalytic hydrogen generation activity. We further demonstrate that the photocatalytic performance of Cu/TiO2 can be enhanced by spin selection, achievable experimentally via optical intersite spin transfer or chiral semiconductor coating. Both H adsorption and spin selection enhance charge carrier lifetimes by an order of magnitude. The spin selection mechanism does not require formation of the H species, which necessitates concurrent sources of electrons and protons and which is intrinsically unstable because water splitting involves frequent proton shuffling. Our results rationalize the experimental observations at the atomistic level, provide mechanistic insights into operation of single atom photocatalysis, and demonstrate that spin selection can be used to develop advanced and efficient systems for solar energy conversion.

Keywords: TiO2 photocatalysis, photoelectrochemical water splitting, single-atom catalysts, spin-selective dynamics, dopant and passivation, nonadiabatic molecular dynamics, time-dependent density functional theory

1. Introduction

Photoelectrochemical (PEC) water splitting has attracted significant attention because of its promise in hydrogen production as an alternative and environmentally friendly fuel.1−7 Among various photoelectrodes, heterogeneous photocatalysts have the potential to achieve high efficiency and selectivity.8,9 However, it is still difficult to control the energies of the frontier orbitals of catalysts and cocatalysts to avoid unnecessary energy losses. In the past few years, single-atom catalysts (SACs) started to draw increasing attention. Representing the ultimate miniaturization of metal particles, SACs present well-defined reaction centers and exhibit high chemical activity and selectivity.10,11 The local atomic configuration of supported single atoms can be finely tuned, and their energy levels can be precisely aligned to achieve high performance. The rapid progress of modern science and technology, including accurate characterization and deep theoretical understanding of catalytic mechanisms, has led to development of increasing numbers of efficient SACs.11,12 The experimental and theoretical advances allow one to design rationally and at the atomic level substrate-supported SACs to enhance the PEC reactivity.

TiO2 is one of the most popular oxide semiconductors used in photocatalytic and photovoltaic applications due its chemical and biological stability and low cost.13 However, TiO2 has a wide band gap (∼3.2 eV for anatase TiO2), which limits light absorption to only a small portion (∼5%) of the solar spectrum.14 Numerous efforts have been made to improve the photocatalytic performance of anatase TiO2, including chemical doping,15 cocatalyst development,16 heterostructure design,14 etc.17 Doping of a substrate constitutes an effective strategy to extend the spectral response into the visible region to enhance light harvesting, as confirmed by both experiments and theory.15,18 Photoactivation achieved by introduction of Cu dopants into the anatase TiO2 lattice has been reported in recent years.15,18,19 For example, Colón et al. found that photocatalytic oxidation of phenol over copper doped TiO2 was enhanced compared with the undoped system.20 Navas et al. synthesized Cu-doped anatase TiO2, confirmed with X-ray diffraction (XRD), X-ray photoelectron spectroscopy (XPS), and Raman spectroscopies that Cu atoms were incorporated into the lattice by substitution of Ti atoms, and demonstrated that a new Cu 3d–Ti 3d optical transition enhanced visible absorption in the 400–800 nm wavelength range.18 Alotaibi et al. demonstrated that the much enhanced antibacterial photocatalytic activity of Cu-doped anatase TiO2 could be attributed to the extended lifetime of separated charge carriers.19 Lee et al. synthesized site specific anatase TiO2-based SACs by incorporating Co, Fe, Ni, Cu, and Rh atoms. The photocatalytic H2 production rate was enhanced compared to pure TiO2 with the highest activity (34 times higher) observed with Cu/TiO2.15 Cu/TiO2 remained inactive prior to light irradiation and showed a sharp absorption onset at ∼390 nm and a broad band at 700 nm, corresponding to the band-to-band transition of anatase TiO2 and d-d transition of Cu, respectively. The low-temperature electron paramagnetic resonance spectra of the initial state sample exhibited a characteristic Cu2+ signal. After exposure to light and surface protonation, photogenerated electrons transferred from the conduction band (CB) of TiO2 to the d-orbitals of the copper atoms, forming an active state that exhibited strong absorption in the whole range of measured wavelengths (300–800 nm). The electrons trapped in the Cu d-orbitals induced a polarization field, resulting in a local TiO2 lattice distortion around the copper atoms. The system changed color from white to black and exhibited strong light absorption in the whole range of measured wavelengths, making it comparable to the intensely studied black TiO2 synthesized under harsh conditions to induce irreversible structural distortion.17 The static density functional theory (DFT) calculations that accompanied the experimental results supported the experimental findings.15 Photoluminescence spectroscopy was employed to characterize the rate of recombination of charge carriers. The behavior of the doped Cu/TiO2 was almost identical initially to that of bare TiO2. Following light absorption and surface protonation, the luminescence yield decreased, which was attributed to enhanced photocatalytic activity.15 Lee et al. rationalized the experimental data theoretically. They found a local structural distortion around the Cu atom and disappearance of the spin-down Cu dz2 antibonding state within the gap upon light absorption and surface protonation.15

Because doped systems such as Cu/TiO2 can contain an odd number of electrons, spin selection provides an additional handle on the photocatalytic activity. Noguchi et al. found that Fe2+ and Fe3+ coexisted in Fe-doped BaTiO3, introducing both electron donor and acceptor levels, and demonstrated that spin-selective charge carrier transfer can be controlled by the pump energy. Under visible light at energies from 1.5 to 1.9 eV, the spin-down electron in the Fe2+-3dz2 state is pumped into the CB; meanwhile, the spin-down hole in the empty Fe3+ state is injected into the valence band (VB). When the pumped energy is above 2.3 eV, the spin-up electron of the Fe2+-3dx2–j2 state is additionally injected into the CB.21 Mtangi et al. enhanced water splitting and hydrogen production by eliminating formation of hydrogen peroxide. The goal was achieved by imposing electron spin selectivity by coating the anode with chiral organic semiconductors from helically aggregated dyes as sensitizers.22 Recently, Willems et al. demonstrated that the spin-down electrons can be selectively transferred from Pt to Co in a CoPt alloy by laser pulses through controlling spin selectivity by optical intersite spin transfer.23 These examples demonstrate that spin selectivity can be achieved using various experimental techniques.

Motivated by the experimental and theoretical works showing the enhanced photocatalytic activity, extended visible absorption, increased charge carrier separation efficiency, and controlled spin selectivity,15,18,19,21−24 we carried out time-dependent DFT (TD-DFT)25−27 simulations combined with nonadiabatic (NA) molecular dynamics (MD)28 to study nonradiative charge trapping and recombination in the Cu/TiO2 SAC. The simulations provide a detailed atomistic description of the mechanisms underlying the PEC water splitting process, establish that formation of a stable Hads_Cu/TiO2 species is needed to extend charge carrier lifetime for efficient water splitting, and demonstrate that, alternatively, high PEC performance of the Cu/TiO2 SAC can be achieved by spin selectivity that is independent of the Hads_Cu/TiO2 structure that is intrinsically unstable.

2. Computational Methods

We employed real-time TDDFT25−27 combined with NA-MD28 to simulate the nonradiative charge trapping and recombination dynamics in Cu/TiO2 and Hads_Cu/TiO2 using the decoherence-induced surface hopping (DISH) algorithm.29 In the approach, the electronic system is treated quantum mechanically, while lattice vibrations are described semiclassically. The methodology has been applied successfully to study photoexcitation dynamics in a broad range of systems, including TiO2,30,31 and it interfaces with electron donors,32−35 metal halide perovskites,36−38 and so on.39−42

Spin-unrestricted DFT calculations were employed for geometry optimization, electronic structure calculations, adiabatic MD, and NA coupling (NAC) calculations, performed with the Vienna Ab Initio Simulation Package.43 The Perdew–Burke–Ernzerhof (PBE)44 functional was adopted to describe electronic exchange-correction interactions, and the projector-augmented-wave method45 was used to treat electron–ion core interactions. The kinetic energy cutoff for the plane wave basis was set to 500 eV. A Γ-centered 3 × 3 × 1 Monkhorst–Pack k-point mesh46 was used to optimize geometry, and a much denser Γ-centered 9 × 9 × 1 k-point mesh was used for electronic structure calculations. The energy convergence criterion of the electronic self-consistent field was 10–5 eV, and the structures were relaxed until the ionic forces were less than 10–3 eV·Å–1. To correct the self-interaction error of the DFT functional, on-site Coulomb corrections47Ueff = 4.2 and 5.2 eV were applied to the Ti 3d and Cu 3d electrons according to the previous reports.48,49 The van der Waals interactions were included using the Grimme DFT-D3 method.50,51 Only the Γ-point is used to sample the Brillouin zone in the NAMD simulations because anatase TiO2 has a direct band gap at the Γ-point, in agreement with the previous PBE calculations.52 After geometry optimization, the systems were heated to 300 K with repeated velocity rescaling for 2 picoseconds (ps) to equilibrate the geometries in the canonical ensemble. Then, 4 ps adiabatic MD trajectories were generated in a microcanonical ensemble with a 1 fs time step. These trajectories were used for NAMD simulations of the charge trapping and recombination dynamics. To obtain long time quantum dynamics data, the 4 ps long NA Hamiltonians were iterated multiple times under the classical path approximation. Additional details about the NAMD methodology can be found in the literature.28,53

The calculated anatase TiO2 bulk lattice parameters of a = b = 3.89 Å and c = 9.65 Å agree well with the experiment values of a = b = 3.79 Å and c = 9.52 Å.54 The anatase TiO2 (101) surface containing 72 atoms was constructed as a slab model with the addition of a 15 Å vacuum layer. The atoms of the bottom three layers were fixed to maintain the bulk properties. Replacing the highlighted five-coordinated surface Ti atom with a Cu atom gave the Cu-doped TiO2 system, marked as Cu_TiO2 (Figure 1a). This configuration has the smallest formation energy compared to substitution of a six-coordinated Ti surface atom or a surface O atom, or adsorption on the surface at both hollow and bridge sites.15,55 Besides, Lee et al. demonstrated that metal dopant atoms incorporate exclusively into the most stable Ti vacancies on TiO2.15 Hereafter, the properties of the Cu atom doped TiO2 SAC were investigated. H adsorption (protonation in the presence of an excited electron) has a strong influence on H+ reduction in an acidic solution with a hole scavenger.56 The most stable structure of H adsorption on Cu/TiO2 is shown in Figure 1b. The structure was selected from a series of structures considering various choices of the H adsorption site in the Cu/TiO2 system, summarized in Figure S1. The configuration used for the NAMD calculations has the lowest total energy, in agreement with the previous work.15

Figure 1.

Figure 1

Open in a new tab

Top (top row) and side (bottom row) views of the optimized geometries of (a) Cu/TiO2 and (b) Hads_Cu/TiO2. The double-sided arrows in the bottom panel of (b) show the elongated Ti–O (green) and Cu–O (blue) bonds in the AC plane.

3. Results and Discussion

Control over charge carriers in the Cu/TiO2 photocatalytic system can be achieved either by spin selection or H adsorption leading to Hads_Cu/TiO2. The two spin channels are identical in Hads_Cu/TiO2 because it contains an even number of electrons. Section 3.1 focuses on the spin-down channel which exhibits the faster dynamics in Cu/TiO2 and therefore determines the overall charge carrier lifetimes in the absence of spin selection. We consider the geometric and electronic structures of Cu/TiO2 and Hads_Cu/TiO2, electron-vibrational interactions and charge trapping and recombination processes. Section 3.2 considers the corresponding properties for the spin-up channel of Cu/TiO2.

3.1. Geometric and Electronic Structure

We first consider the geometric and electronic structure of the Cu/TiO2 and Hads_Cu/TiO2 systems at 0 K. The top and side views of the optimized geometries are shown in Figure 1. The substitution of the five-coordinated Ti atom of the topmost surface of the pristine anatase TiO2 (101) surface with a Cu atom, Figure 1a, leads to a local distortion. Compared to the Ti–O bond length 1.87 Å in the ac plane in the pristine TiO2 system, the Cu–O bond length (blue arrow in Figure 1b) increases to 2.16 Å, and the four in-plane Cu–O bond lengths decrease due to ionic radii mismatch between the Ti and Cu atoms. Adsorption of a H atop O next to the Cu dopant increases the Ti–O bond length, marked by the green arrow in Figure 1b, to 2.12 Å from 1.92 Å in the pristine TiO2 and 2.07 Å in the Cu/TiO2 systems. Meanwhile, the length of the Cu–O bond in the ac plane increases to 2.20 Å. The changes in the bond length are relatively small. Even though the system undergoes a noticeable distortion upon doping, the Cu doped TiO2 photocatalyst remains stable during the water splitting process.

The spin-polarized projected density of states (PDOS) is shown in Figure 2 for the Cu/TiO2 and Hads_Cu/TiO2 systems. The PDOS is split into contributions from Ti, O, Cu, and H. The CB minimum (CBM) and VB maximum (VBM) are primarily composed of the Ti and O atomic orbitals, respectively. The CBM and VBM are separated by 2.37 and 2.35 eV in both spin channels for the Cu/TiO2 and Hads_Cu/TiO2 systems. This value is close to 2.33 eV for the calculated bandgap of the pristine TiO2 surface and is notably lower than the experimental bandgap (3.2 eV) of pure anatase TiO2 due to the self-interaction error of the PBE functional. To account for the systematic error, we scaled uniformly the calculated energy gaps between pairs of electronic states by the ratio of the experimental and calculated bandgaps, as indicated in Table 1. TiO2 doping with a low concentration of Cu does not change the bandgap but creates one electron trap state in the spin-up channel, and two electron trap states in the spin-down channel, Figure 2a. The deep midgap states in both spin-up and spin-down channels are marked as trap, while the shallow electron trap state located just above the Fermi energy is denoted as trap′. Addition of the H atom to Cu/TiO2 eliminates the trap′ state, such that the spin-up and spin-down channels of Hads_Cu/TiO2 exhibit identical PDOS with one trap state within the bandgap, Figure 2b. The midgap trap state arise from 3d orbitals of Cu and 2p orbitals of O in the ab plane near the Cu dopant in both Cu/TiO2 and Hads_Cu/TiO2. The shallow trap′ state arises predominantly from 2p orbitals of O with a minor contribution from 3d orbitals of Cu. The orbital resolved PDOS on the Cu atom of Cu/TiO2 and Hads_Cu/TiO2 systems (Figure S2) further shows that the two spin–orbitals forming the trap state inside the TiO2 bandgap originate primarily from the Cu dx2–y2 orbital with a minor contribution from the dxz orbital. The trap′ state has a minor contribution of the dxy orbital. The rest of the Cu d orbitals are inside the TiO2 VB and are spread in energy due to mixing with the VB states. Note that five ligands (O atoms) are bound to the Cu atom, and the Cu complex is not octahedral. Therefore, the octahedral crystal field theory cannot fully describe the Cu d orbital splitting. The trap states strongly influence nonradiative charge dynamics. In particular, the involvement of the trap and trap′ states in the spin-down channel of the Cu/TiO2 system result in much more complicated charge dynamics compared to that in the spin-up channel. Therefore, the charge carrier dynamics in the Cu/TiO2 photocatalysts can be controlled by spin selection, which can be realized experimentally.21−23

Figure 2.

Figure 2

Open in a new tab

Spin-up and spin-down PDOS of (a) Cu/TiO2 and (b) Hads_Cu/TiO2. The dashed line represents the Fermi energy. The Cu/TiO2 system has one midgap trap state in the spin-up channel and two trap states in the spin-down channel. Addition of H passivates and eliminates the trap state near the Fermi energy. The Hads_Cu/TiO2 system has an even number of electrons, and the spin-up and spin-down channels have identical PDOS with one midgap trap state.

Table 1. Canonically Averaged Energy Gaps, Pure-Dephasing Time, Absolute NAC and State-to-State Transition Rates for the Pairs of States in the Spin-Down Channel in Cu/TiO2 and Hads_Cu/TiO2.

    energy gap (eV) NAC (meV) dephasing (fs) rate (ps–1)
Cu/TiO2 CBM → trap 2.07 2.71 2.68 0.0045
CBM → trap′ 3.06 5.95 4.03 0.182
CBM → VBM 3.20 3.36 9.57 0.0595
trap → trap′ 1.00 8.30 2.75 0.417
trap → VBM 1.13 3.91 2.80 0.0575
trap′ → VBM 0.14 17.4 5.49 0.561
Hads_Cu/TiO2 CBM → trap 2.23 1.66 1.13 0.0032
CBM → VBM 3.20 1.99 8.53 0.0211
trap → VBM 0.97 6.40 1.06 0.147
Open in a new tab

Cu K-edge extended X-ray absorption fine structure analysis indicated that the Cu atoms were located in the Ti sites,15 and no characteristic peak corresponding to metallic bonding was observed, as corroborated by the XRD data. Low-temperature electron paramagnetic resonance spectra and near-edge X-ray absorption spectroscopy demonstrated a +2 oxidation state of the Cu atom in Cu-doped TiO2.15 Furthermore, according to the ionic/atomic radii of Ti(IV) (0.60 Å), Cu(0) (1.45 Å), Cu(I) (0.77 Å), and Cu(II) (0.73 Å), it is expected that the Cu atom in the +2 oxidation state will form the most stable configuration when doped into in TiO2 in place of Ti because of the smallest mismatch between the ionic radii of Ti(IV) and Cu(II). We calculated the Bader atomic charge,57 which indicated that the Cu atom lost 1.22 e in the neutral Cu/TiO2 system. Previous calculations demonstrated that the replacement of Ti(IV) with Cu(II) in anatase TiO2 had the lowest formation energy among various Cu oxidation states.19 Thus, both the experimental and theoretical results have demonstrated that substitution of Ti with Cu in anatase TiO2 produces Cu in the +2 oxidation state.

The charge trapping and recombination rates are determined by the NAC between the corresponding pairs of states, and the NAC depends on overlap between the wave functions, Inline graphic. Figure 3 shows charge densities of the key orbitals involved in charge trapping and recombination dynamics in the Cu/TiO2 and Hads_Cu/TiO2 systems. The VBM and CBM are supported by on the O and Ti atoms. The trap′ state is localized around the Cu dopant, but it is supported by 2p orbitals of the O atoms, Figure 2a. In comparison, the midgap trap states are localized almost equally on the Cu dopant and the surrounding O atoms in the ab plane. The results are consistent with the PDOSs analysis for the two systems, Figure 2. Generally, localized states couple to high-frequency phonon modes, generate strong NAC, and accelerate charge dynamics. The presence of two trap states in the spin-down channel of the Cu/TiO2 systems provide more nonradiative recombination channels than the singe trap state in the Hads_Cu/TiO2 system. Therefore, H adsorption by Cu/TiO2 can enhance carrier lifetimes and photocatalytic performance.

Figure 3.

Figure 3

Open in a new tab

Spin-down charge densities for VBM, trap′, trap, and CBM of the Cu/TiO2 system, and VBM, trap, and CBM of the Hads_Cu/TiO2 systems. The trap′ state arises primarily from 2p orbitals of O atoms next to the Cu dopant. The midgap trap states in both Cu/TiO2 and Hads_Cu/TiO2 are localized on the Cu dopant and neighboring O atoms in the ab plane.

3.2. Electron-Vibrational Interactions

Figure 4 shows evolution of energies of the key states in the Cu/TiO2 and Hads_Cu/TiO2 systems at 300 K. The trap state fluctuates more than the trap′ state in the Cu/TiO2 system, Figure 4a, because trap is well-separated energetically and decoupled from the band edges. In contrast, trap′ couples strongly to the VBM due to the very small energy gap, and such coupling dampens the trap′ state energy oscillation. The NAC between trap′ and the VBM is large because of the small energy gap, Table 1. The NAC between trap′ and trap is also large because they are relatively close to each other and often oscillate in phase. The presence of the H atom in the Hads_Cu/TiO2 system enhances fluctuations in the energies of the band edge and the trap states. H is a light atom and moves significantly. As a result, both amplitude and frequency of the energy fluctuations increase, Figure 4b. Strong thermal fluctuations lead to a dynamic balance of adsorption and desorption of protons around the active site on the Cu/TiO2 surface and drive water splitting to proceed persistently. The trap state can approach closely the VBM of the Hads_Cu/TiO2 system, and as a result, the NAC is significant, Table 1. On the other hand, trap is far in energy from the CBM, and the corresponding NAC is small. Large fluctuations of the trap state energy in the Hads_Cu/TiO2 system leads to rapid coherence loss (pure-dephasing) for transitions involving the trap state, Table 1. The interplay between energy gaps, NAC, and decoherence governs the charge trapping and recombination dynamics.

Figure 4.

Figure 4

Open in a new tab

Evolution of energies of the band edge and trap states for the spin-down channel in (a) Cu/TiO2 and (b) Hads_Cu/TiO2 at 300 K. The data are taken from the last 2 ps of the 4 ps microcanonical MD trajectories.

Electron-vibrational interactions induce both inelastic and elastic electron–phonon scattering, and both types of scattering influence nonradiative charge trapping and recombination dynamics. Inelastic scattering leads to energy transfer from electrons to phonons, while elastic scattering determines lifetimes of coherent quantum-mechanical superpositions between initial and final states. Inelastic scattering is quantified by the magnitude of the NAC, Table 1, which depends on both the phonon modes that couple to the electronic subsystem, and overlap of initial and final wave functions, Figure 3. Elastic electron–phonon scattering is quantified by the pure-dephasing time, Table 1, which is computed using the optical response theory.58 The pure-dephasing time depends on the amplitude and memory of the phonon-induced fluctuations of corresponding energy gap.59

To identify the phonon modes that couple to the electronic subsystem, we computed the influence spectra by performing Fourier transforms (FTs) of the autocorrelation functions (ACF) for the phonon-induced energy gap fluctuations of the pairs of states, as shown in Figure S3. Both low-frequency acoustic modes and high-frequency optical modes contribute to the spectral densities of the two systems. Anatase TiO2 exhibits six active Raman modes: A1g (513 cm–1), B1g (399 and 519 cm–1), and Eg (144, 197, and 639 cm–1).18,60,61 The peaks at 399, 197, and 144 cm–1 can be attributed to the O–Ti–O bending motions.60 The peaks around 639, 513, and 519 cm–1 are associated with the Ti–O bond stretching modes.60 TiO2 doping with Cu and subsequent protonation break anatase TiO2 symmetry and relax electron–phonon coupling selection rules. The doping and protonation also give rise to new phonon modes. As a result, many modes contribute to the electron–phonon coupling, and the key peaks are shifted and broadened, in agreement with the previous work.18 The O–Ti–O bending Eg phonon at 144 cm–1 and the Ti–O stretching A1g phonon at 513 cm–1 are particularly prominent in Figure S3. The modes below 50 cm–1 and around 300 and 600 cm–1 in the Cu/TiO2 system and the modes around 80, 270, and 730 cm–1 in the Hads_Cu/TiO2 system can be regarded as the new modes. Generally, the spectral density of the Cu/TiO2 system exhibits more and higher peaks, indicating stronger electron–phonon coupling and faster charge trapping and recombination dynamics.

Figure S4 shows the pure-dephasing functions. Fitting each of the curve to a Gaussian, exp(−0.5(t/τ)2), gives the pure-dephasing times reported in Table 1. The pure-dephasing times are short, sub-10 fs, due to participation of multiple phonon modes, Figure S3. The coherence times are shorter in Hads_Cu/TiO2 than Cu/TiO2. Generally, short-lived coherence favors long-lived excited states, while the long-lived coherence favors fast charge recombination. The relative values of the coherence times can be rationalized by the ACFs shown in Figure S4. The initial values of the unnormalized ACF represent squares of the energy gap fluctuations, corresponding to the data shown in Figure 4. Larger initial values and faster decay of the ACFs result in shorter pure-dephasing times.59 The CBM-VBM transitions are characterized by the smallest initial values and slowest decay of the ACFs, Figure S4, resulting in longest coherence times, Table 1. The initial values of the unnormalized ACFs for transitions involving the trap states are much larger in Hads_Cu/TiO2 than Cu/TiO2, Figure S4, because the trap energy level oscillates much more due to presence of the H atom. As a result, the corresponding coherence times are very short, on the order of 1 fs, Table 1.

3.3. Charge Dynamics in the Spin-Down Channel

Figure 5 shows the charge trapping and recombination dynamics. Fitting the data using an exponential function P(t) = Aexp(−t/τ) gives the time scales, τ, reported in Figure 5. The rise and decay components are fitted separately. To facilitate interpretation of the full quantum dynamics simulations involving multiple states, Figure 5, we also report state-to-state transition rate constants obtained by NAMD calculations with only two states at a time, Table 1. The NAMD data are shown in Figures S6 and S7. The rate constants given in Table 1 are obtained by exponential fitting. Such state-to-state analysis provides valuable insights, because, for instance, the small magnitude of the trap state population maximum observed in both Cu/TiO2 and Hads_Cu/TiO2 can be due to either its slow population or fast depopulation, and the depopulation can occur into either trap′ or VBM. The rate of transition of the photogenerated electron in CBM into trap is slow in both systems, Table 1, i.e., the electron bypasses the trap state and goes directly into VBM or trap′ if available.

Figure 5.

Figure 5

Open in a new tab

Charge carrier trapping and recombination dynamics of (a) Cu/TiO2 and (b) Hads_Cu/TiO2 in the spin-down channel. The insets zoom onto the population of the trap state.

The fact that the deep midgap trap state has little influence on the charge recombination dynamics in both systems is quite surprising. The result can be understood by considering the NAC and pure-dephasing times, Table 1. The NAC for the CBM-trap transition is smaller than the NAC for the CBM-VBM and CBM-trap′ transitions. The strong localization on the Cu dopant and adjacent O1–O5 of the trap state, Figure S9, decouples it from the free electron states. In comparison, trap′ has little contribution from Cu. It is energetically close to and mixes with VB states. As result, trap′ couples more strongly the VBM than trap does, Table 1. The extent of nuclear motion influences the NAC magnitude. To this end, we computed the standard deviations of the positions of the Cu dopant and the O1–O5 atoms contributing to the trap state (0.227 Å) and of the O1 atoms contributing to the trap′ state (0.303 Å), Figure S9. The larger amplitude of atomic motions also rationalizes the larger NAC between CBM and trap′. Further, the pure-dephasing times for the CBM-trap transition are small, Table 1. This is because the trap state energy oscillates significantly, Figure 4, and large energy oscillation results in rapid coherence loss.59 Thus, localization of the trap state on the Cu dopant reduces its correlation with free electron states and decreases the corresponding wave function overlap, resulting in small NAC and fast coherence loss. Both factors lead to the small CBM-trap transition rates. In contrast, the transition rates between the trap′ state in the Cu/TiO2 system and the free electron states are large, and this shallow trap has a strong influence on the charge carrier lifetimes.

The state-to-state analysis is supported by the comprehensive quantum dynamics simulations, Figure 5. The trap state is never populated in either system, while the trap′ state is populated by near 20%. Note that fitting the rise of the trap′ state population gives a 1 ps time constant, Figure 5a, which is 5 times shorter than the inverse of the CBM-trap′ rate constant, Table 1. This apparent contradiction arises from the fact that the rate of the trap′-VBM transition is large. Counterintuitively, the smaller CBM-trap′ and larger trap′-VBM rates produce fast rise and slow decay of the trap′ population with a small maximum amplitude. The electron–hole recombination is faster nearly by a factor of 10 in Cu/TiO2 compared to Hads_Cu/TiO2, 5.59 ps vs 43.4 ps, Figure 5. Elimination of the trap′ state either by spin selection or protonation will have a strong influence on performance of Cu doped TiO2. The presence of the Hads_Cu/TiO2 structure rationalizes the high PEC water splitting activity observed in the experiment.15

3.4. Charge Dynamics in the Spin-Up Channel

The Cu/TiO2 system has an odd number of electrons, and the electronic properties and charge dynamics in the spin-up channel differ from those in the spin-down channel. The Hads_Cu/TiO2 system has an even number of electrons, and the two spin channels exhibit similar properties. Therefore, we consider explicitly the spin-up channel for the Cu/TiO2 system only.

Although the energies of the trap state in Cu/TiO2 are not degenerate in the spin-up and spin-down components, the energy difference is small, Figure 2a, and other properties are also similar. The charge densities of CBM, trap, and VBM are almost identical for the two spin channels, and the phonon-induced fluctuations are the state energies are similar, compare Figure 3 and Figure 6. Due to the similarity in the energy fluctuations, the pure-dephasing times for transitions involving CBM, trap, and VBM are comparable between the spin-up and spin-down channels, compare Table 1 and Table 2. In particular, the coherence is maintained longer for the CBM-VBM transition than for CBM-trap and trap-VBM transitions, because the trap state energy fluctuates much more than the energies of the band edges. The NAC matrix elements involving the trap state are slightly smaller in the spin-up channel than the spin-down channel, while they follow the same trend. The NAC for the trap-VBM transition is larger, because the trap state is closer in energy to the VBM and has a large contribution from O 2p orbitals, similarly to the VBM.

Figure 6.

Figure 6

Open in a new tab

(a) Spin-up charge densities for VBM, trap, and CBM of the Cu/TiO2 system. The trap state originates from the Cu dopant and neighboring oxygen atoms in the ab plane. (b) Evolution of the state energies in the last 2 ps of the 4 ps microcanonical MD trajectories at 300 K.

Table 2. Canonically Averaged Energy Gap, Pure-Dephasing Time, Absolute NAC, and State-to-State Transition Rates for the Pairs of States in the Spin-Up Channel for Cu/TiO2.

  energy gap (eV) NAC (meV) dephasing (fs) rate (ps–1)
CBM → trap 1.96 1.58 3.30 0.0015
CBM → VBM 3.20 1.69 7.23 0.0108
trap → VBM 1.24 3.66 3.47 0.0582
Open in a new tab

Comparing the spin-up component of Cu/TiO2 with the properties of the Hads_Cu/TiO2 system, we observe similar trends, i.e., spin selection results in similar effects as protonation. Both spin selection and protonation eliminate the trap′ state, reducing the number of nonradiative decay channels and extending carrier lifetimes. The main difference arises from the larger fluctuation of the trap state energy in the protonated system due to presence of the light H atom. As a result, the pure-dephasing times involving the trap state are shorter in the Hads_Cu/TiO2 system than in the spin-up channel of Cu/TiO2, and the NAC for the trap-VBM transition is larger, Table 1 and Table 2.

The electron–hole recombination dynamics in the spin-up channel of Cu/TiO2 proceeds directly between the CBM and VBM, bypassing the trap state, Figure 7. The trap state is effectively decoupled from the CBM, because it is highly localized around Cu and is formed by Cu and O orbitals, while the CBM arises from Ti orbitals, Figure 2. The conclusion is supported by the state-to-state transition rates, shown in Table 2 and Figure S10. The CBM-trap rate is an order of magnitude smaller than the CBM-VBM rate. The population of the trap state never reaches more than a few percent. The charge carrier dynamics in the spin-up channel of Cu/TiO2 is similar to the dynamics in the Hads_Cu/TiO2 system, Figure 5b. The electron–hole recombination time is even longer in the spin-up channel of Cu/TiO2 because of the smaller NAC in the absence of the H atom. To test convergence of the NAMD results obtained using the 4 ps NA Hamiltonians, we repeated the calculation for both spin down and spin up channels using the last 3 ps of the data. The test results, Figure S11, agree with the original data, Figures 5 and 7, within a few percent.

Figure 7.

Figure 7

Open in a new tab

Charge trapping and recombination dynamics in the spin-up channel of Cu/TiO2. The inset zooms onto the population of the trap state.

The above results indicate that Cu/TiO2 is inactive due to the rapid charge recombination in the spin-down channel. The recombination is facilitated by the trap′ state that is present only in one spin channel and is located close to the VBM. Surprisingly, the deep midgap trap state has little influence on the charge carrier dynamics. To make the Cu/TiO2 system catalytically active, the decay channel created by the trap′ state should be eliminated. This can be done either via spin selection or H adsorption. Interestingly, spin selection extends carrier lifetimes more than H adsorption. Spin selection blocks the trap′ decay channel, while H adsorption eliminated the trap′ state. After photoexcitation and protonation, the trap′ state in the spin-down channel is pushed into the VB and leaves only the Cu dx2–y2 state in the band gap. The fact that the photocatalytic H2 production is attributed to Hads_Cu/TiO2 rather than Cu/TiO2 agrees well with the experimental report.15 The additional means of controlling the Cu/TiO2 catalytic performance by spin selectivity can be used in SAC reactions that do not involve formation of the stable Hads_Cu/TiO2 species, which requires concurrent sources of electrons and protons.

4. Conclusions

Using a combination of ab initio real-time TD-DFT and NAMD, we investigated nonradiative charge trapping and recombination in Cu-doped anatase TiO2 that acts as a single atom photocatalyst and shows superior activity in PEC water splitting experiments. The atomistic analysis shows that the activity arises from the Hads_Cu/TiO2 species rather than Cu/TiO2. Without the coadsorbed H, the Cu dopant creates two trap states inside the TiO2 bandgap, facilitating rapid nonradiative losses. A shallow trap near the VBM arises from 2p orbitals of O atoms next to the Cu dopant but has no contributions from the dopant itself. A deep midgap trap is supported by the dx2–y2 orbital of the Cu dopant and neighboring O atoms in the ab plane. Surprisingly, it is the shallow trap that is responsible for the fast charge losses, while the midgap state is largely decoupled from free charge carriers and contributes little to the nonradiative relaxation dynamics. Localization of the midgap trap on the Cu dopant reduces its correlation with free electron states, decreases the corresponding wave function overlap, and results in small NAC and fast coherence loss. The NAC between the shallow trap state and the VBM is large because of the small energy gap, and the NAC between the shallow and deep traps is also large because their wave functions overlap well. Thus, the presence of the shallow trap activates the deep trap as well.

Photoinduced electron transfer and protonation of the Cu dopant site on the TiO2 surface pushes the shallow trap state into the VB and extends the charge carrier lifetime by nearly an order of magnitude. The simulations show that the local structure undergoes a noticeable distortion upon doping. Nevertheless, the Cu doped TiO2 photocatalyst remains stable during the water splitting process. Thermal fluctuations lead to a dynamic balance of adsorption and desorption of proton around the active site on the Cu/TiO2 surface and drive water splitting to proceed persistently. Although H adsorption eliminates the shallow trap, it introduces faster motions and enhances fluctuations of the midgap state energy, making the midgap trap a stronger nonradiative relaxation center. The fact that the photocatalytic H2 production is attributed to Hads_Cu/TiO2 rather than Cu/TiO2 agrees well with the experimental results.

The simulations show that control over charge carrier dynamics in the Cu/TiO2 photocatalytic system can be achieved by spin selection because the second trap state responsible for the majority of carrier losses appears only in one spin channel. Spin selection can be implemented experimentally by several techniques, such as coating with chiral semiconductors and optical intersite spin transfer. The charge carrier dynamics in the spin-selected channel of Cu/TiO2 is similar to the dynamics in the Hads_Cu/TiO2 system. The charge recombination time is even longer in the spin-selected Cu/TiO2 because the NAC is smaller in the absence of the H atom. Further, the spin selection mechanism does not require H to be present next to Cu. H adsorption and desorption occur dynamically during the PEC water splitting, and when the H atom is not adsorbed at the catalytic site, carrier losses are fast. The additional means of controlling the Cu/TiO2 catalytic performance by spin selectivity can be used in SAC reactions that do not involve formation of a stable protonated species.

The simulations establish the mechanisms responsible for the high H2 production activity of Cu doped TiO2, advance our understanding of excited-state dynamics in photocatalytic and photovoltaic systems, and suggest that spin selectivity can be used to design better single-atom photocatalysts.

Acknowledgments

This work was supported by the National Science Foundation of China, Grants 51861135101, 21973006, 21688102, and 21590801. R.L. acknowledges financial support from the Recruitment Program of Global Youth Experts of China and the Beijing Normal University Startup. O.V.P. acknowledges support of the U.S. Department of Energy, Grant DE-SC0014429.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.1c00004.

  • Total energy of various configurations of H adsorbed on the Cu/TiO2 system, orbital resolved PDOS, spectral densities, pure-dephasing functions, state-to-state transition dynamics, PDOSs and local geometry that contribute to the trap′ and trap states in the Cu/TiO2 system, charge trapping and recombination dynamics using shorter NA coupling sampling, and the optimized coordinates of the Cu/TiO2 and Hads_Cu/TiO2 systems (PDF)

The authors declare no competing financial interest.

Supplementary Material

au1c00004_si_001.pdf (1.5MB, pdf)

References

  1. Fujishima A.; Honda K. Electrochemical Photolysis of Water at a Semiconductor Electrode. Nature 1972, 238 (5358), 37–38. 10.1038/238037a0. [DOI] [PubMed] [Google Scholar]
  2. Tachibana Y.; Vayssieres L.; Durrant J. R. Artificial photosynthesis for solar water-splitting. Nat. Photonics 2012, 6 (8), 511–518. 10.1038/nphoton.2012.175. [DOI] [Google Scholar]
  3. Kim T. W.; Choi K.-S. Nanoporous BiVO4 Photoanodes with Dual-Layer Oxygen Evolution Catalysts for Solar Water Splitting. Science 2014, 343 (6174), 990–994. 10.1126/science.1246913. [DOI] [PubMed] [Google Scholar]
  4. Lumley M. A.; Radmilovic A.; Jang Y. J.; Lindberg A. E.; Choi K. S. Perspectives on the Development of Oxide-Based Photocathodes for Solar Fuel Production. J. Am. Chem. Soc. 2019, 141 (46), 18358–18369. 10.1021/jacs.9b07976. [DOI] [PubMed] [Google Scholar]
  5. He Y.; Hamann T.; Wang D. Thin film photoelectrodes for solar water splitting. Chem. Soc. Rev. 2019, 48 (7), 2182–2215. 10.1039/C8CS00868J. [DOI] [PubMed] [Google Scholar]
  6. Zhang K.; Liu J.; Wang L.; Jin B.; Yang X.; Zhang S.; Park J. H. Near-Complete Suppression of Oxygen Evolution for Photoelectrochemical H2O Oxidative H2O2 Synthesis. J. Am. Chem. Soc. 2020, 142 (19), 8641–8648. 10.1021/jacs.9b13410. [DOI] [PubMed] [Google Scholar]
  7. Ruan Q.; Miao T.; Wang H.; Tang J. Insight on Shallow Trap States-Introduced Photocathodic Performance in n-Type Polymer Photocatalysts. J. Am. Chem. Soc. 2020, 142 (6), 2795–2802. 10.1021/jacs.9b10476. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Wang S.; Guan B. Y.; Wang X.; Lou X. W. D. Formation of Hierarchical Co9S8@ZnIn2S4 Heterostructured Cages as an Efficient Photocatalyst for Hydrogen Evolution. J. Am. Chem. Soc. 2018, 140 (45), 15145–15148. 10.1021/jacs.8b07721. [DOI] [PubMed] [Google Scholar]
  9. Kuriki R.; Ichibha T.; Hongo K.; Lu D.; Maezono R.; Kageyama H.; Ishitani O.; Oka K.; Maeda K. A Stable, Narrow-Gap Oxyfluoride Photocatalyst for Visible-Light Hydrogen Evolution and Carbon Dioxide Reduction. J. Am. Chem. Soc. 2018, 140 (21), 6648–6655. 10.1021/jacs.8b02822. [DOI] [PubMed] [Google Scholar]
  10. Yang X.-F.; Wang A.; Qiao B.; Li J.; Liu J.; Zhang T. Single-Atom Catalysts: A New Frontier in Heterogeneous Catalysis. Acc. Chem. Res. 2013, 46 (8), 1740–1748. 10.1021/ar300361m. [DOI] [PubMed] [Google Scholar]
  11. Wang A.; Li J.; Zhang T. Heterogeneous single-atom catalysis. Nat. Rev. Chem. 2018, 2 (6), 65–81. 10.1038/s41570-018-0010-1. [DOI] [Google Scholar]
  12. Tran D. T.; Nguyen D. C.; Le H. T.; Kshetri T.; Hoa V. H.; Doan T. L. L.; Kim N. H.; Lee J. H. Recent progress on single atom/sub-nano electrocatalysts for energy applications. Prog. Mater. Sci. 2021, 115, 100711. 10.1016/j.pmatsci.2020.100711. [DOI] [Google Scholar]
  13. Fujishima A.; Zhang X.; Tryk D. A. TiO2 photocatalysis and related surface phenomena. Surf. Sci. Rep. 2008, 63 (12), 515–582. 10.1016/j.surfrep.2008.10.001. [DOI] [Google Scholar]
  14. Zheng L.; Teng F.; Ye X.; Zheng H.; Fang X. Photo/Electrochemical Applications of Metal Sulfide/TiO2 Heterostructures. Adv. Energy Mater. 2020, 10 (1), 1902355. 10.1002/aenm.201902355. [DOI] [Google Scholar]
  15. Lee B. H.; Park S.; Kim M.; Sinha A. K.; Lee S. C.; Jung E.; Chang W. J.; Lee K. S.; Kim J. H.; Cho S. P.; Kim H.; Nam K. T.; Hyeon T. Reversible and cooperative photoactivation of single-atom Cu/TiO2 photocatalysts. Nat. Mater. 2019, 18 (6), 620–626. 10.1038/s41563-019-0344-1. [DOI] [PubMed] [Google Scholar]
  16. Xiao M.; Zhang L.; Luo B.; Lyu M.; Wang Z.; Huang H.; Wang S.; Du A.; Wang L. Molten-Salt-Mediated Synthesis of an Atomic Nickel Co-catalyst on TiO2 for Improved Photocatalytic H2 Evolution. Angew. Chem., Int. Ed. 2020, 59 (18), 7230–7234. 10.1002/anie.202001148. [DOI] [PubMed] [Google Scholar]
  17. Chen X.; Liu L.; Yu P. Y.; Mao S. S. Increasing Solar Absorption for Photocatalysis with Black Hydrogenated Titanium Dioxide Nanocrystals. Science 2011, 331 (6018), 746–750. 10.1126/science.1200448. [DOI] [PubMed] [Google Scholar]
  18. Navas J.; Sanchez-Coronilla A.; Aguilar T.; Hernandez N. C.; de los Santos D. M.; Sanchez-Marquez J.; Zorrilla D.; Fernandez-Lorenzo C.; Alcantara R.; Martin-Calleja J. Experimental and theoretical study of the electronic properties of Cu-doped anatase TiO2. Phys. Chem. Chem. Phys. 2014, 16 (8), 3835–45. 10.1039/c3cp54273d. [DOI] [PubMed] [Google Scholar]
  19. Alotaibi A. M.; Williamson B. A. D.; Sathasivam S.; Kafizas A.; Alqahtani M.; Sotelo-Vazquez C.; Buckeridge J.; Wu J.; Nair S. P.; Scanlon D. O.; Parkin I. P. Enhanced Photocatalytic and Antibacterial Ability of Cu-Doped Anatase TiO2 Thin Films: Theory and Experiment. ACS Appl. Mater. Interfaces 2020, 12 (13), 15348–15361. 10.1021/acsami.9b22056. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Colón G.; Maicu M.; Hidalgo M. C.; Navío J. A. Cu-doped TiO2 systems with improved photocatalytic activity. Appl. Catal., B 2006, 67 (1), 41–51. 10.1016/j.apcatb.2006.03.019. [DOI] [Google Scholar]
  21. Noguchi Y.; Taniguchi Y.; Inoue R.; Miyayama M. Successive redox-mediated visible-light ferrophotovoltaics. Nat. Commun. 2020, 11 (1), 966. 10.1038/s41467-020-14763-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Mtangi W.; Tassinari F.; Vankayala K.; Vargas Jentzsch A.; Adelizzi B.; Palmans A. R. A.; Fontanesi C.; Meijer E. W.; Naaman R. Control of Electrons’ Spin Eliminates Hydrogen Peroxide Formation During Water Splitting. J. Am. Chem. Soc. 2017, 139 (7), 2794–2798. 10.1021/jacs.6b12971. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Willems F.; von Korff Schmising C.; Struber C.; Schick D.; Engel D. W.; Dewhurst J. K.; Elliott P.; Sharma S.; Eisebitt S. Optical inter-site spin transfer probed by energy and spin-resolved transient absorption spectroscopy. Nat. Commun. 2020, 11 (1), 871. 10.1038/s41467-020-14691-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Wei D.; Tan Y.; Wang Y.; Kong T.; Shen S.; Mao S. S. Function-switchable metal/semiconductor junction enables efficient photocatalytic overall water splitting with selective water oxidation products. Sci. Bull. 2020, 65 (16), 1389–1395. 10.1016/j.scib.2020.04.042. [DOI] [PubMed] [Google Scholar]
  25. Craig C. F.; Duncan W. R.; Prezhdo O. V. Trajectory Surface Hopping in the Time-Dependent Kohn-Sham Approach for Electron-Nuclear Dynamics. Phys. Rev. Lett. 2005, 95 (16), 163001. 10.1103/PhysRevLett.95.163001. [DOI] [PubMed] [Google Scholar]
  26. Marques M. A. L.; Gross E. K. U. Time-Dependent Density Functional Theory. Annu. Rev. Phys. Chem. 2004, 55 (1), 427–455. 10.1146/annurev.physchem.55.091602.094449. [DOI] [PubMed] [Google Scholar]
  27. Runge E.; Gross E. K. U. Density-Functional Theory for Time-Dependent Systems. Phys. Rev. Lett. 1984, 52 (12), 997–1000. 10.1103/PhysRevLett.52.997. [DOI] [Google Scholar]
  28. Akimov A. V.; Prezhdo O. V. The PYXAID Program for Non-Adiabatic Molecular Dynamics in Condensed Matter Systems. J. Chem. Theory Comput. 2013, 9 (11), 4959–4972. 10.1021/ct400641n. [DOI] [PubMed] [Google Scholar]
  29. Jaeger H. M.; Fischer S.; Prezhdo O. V. Decoherence-induced surface hopping. J. Chem. Phys. 2012, 137 (22), 22A545. 10.1063/1.4757100. [DOI] [PubMed] [Google Scholar]
  30. Stier W.; Duncan W. R.; Prezhdo O. V. Thermally Assisted Sub-10 fs Electron Transfer in Dye-Sensitized Nanocrystalline TiO2 Solar Cells. Adv. Mater. 2004, 16 (3), 240–244. 10.1002/adma.200306027. [DOI] [Google Scholar]
  31. Wei Y.; Zhou Z.; Fang W. H.; Long R. Grain Boundary Facilitates Photocatalytic Reaction in Rutile TiO2 Despite Fast Charge Recombination: A Time-Domain ab Initio Analysis. J. Phys. Chem. Lett. 2018, 9 (19), 5884–5889. 10.1021/acs.jpclett.8b02761. [DOI] [PubMed] [Google Scholar]
  32. Mehdipour H.; Akimov A. V.; Jankowska J.; Rezakhanai A. T.; Tafreshi S. S.; de Leeuw N. H.; Moshfegh A. Z.; Prezhdo O. V. Persistent Quantum Coherence and Strong Coupling Enable Fast Electron Transfer across the CdS/TiO2 Interface: A Time-Domain ab Initio Simulation. J. Phys. Chem. C 2018, 122 (44), 25606–25616. 10.1021/acs.jpcc.8b06425. [DOI] [Google Scholar]
  33. Long R.; Casanova D.; Fang W.-H.; Prezhdo O. V. Donor–Acceptor Interaction Determines the Mechanism of Photoinduced Electron Injection from Graphene Quantum Dots into TiO2: π-Stacking Supersedes Covalent Bonding. J. Am. Chem. Soc. 2017, 139 (7), 2619–2629. 10.1021/jacs.6b09598. [DOI] [PubMed] [Google Scholar]
  34. Long R.; English N. J.; Prezhdo O. V. Minimizing Electron–Hole Recombination on TiO2 Sensitized with PbSe Quantum Dots: Time-Domain Ab Initio Analysis. J. Phys. Chem. Lett. 2014, 5 (17), 2941–2946. 10.1021/jz5013627. [DOI] [PubMed] [Google Scholar]
  35. Long R.; Fang W.-H.; Prezhdo O. V. Strong Interaction at the Perovskite/TiO2 Interface Facilitates Ultrafast Photoinduced Charge Separation: A Nonadiabatic Molecular Dynamics Study. J. Phys. Chem. C 2017, 121 (7), 3797–3806. 10.1021/acs.jpcc.6b12921. [DOI] [Google Scholar]
  36. He J.; Fang W. H.; Long R.; Prezhdo O. V. Superoxide/Peroxide Chemistry Extends Charge Carriers’ Lifetime but Undermines Chemical Stability of CH3NH3PbI3 Exposed to Oxygen: Time-Domain ab Initio Analysis. J. Am. Chem. Soc. 2019, 141 (14), 5798–5807. 10.1021/jacs.8b13392. [DOI] [PubMed] [Google Scholar]
  37. Li W.; Vasenko A. S.; Tang J.; Prezhdo O. V. Anharmonicity Extends Carrier Lifetimes in Lead Halide Perovskites at Elevated Temperatures. J. Phys. Chem. Lett. 2019, 10 (20), 6219–6226. 10.1021/acs.jpclett.9b02553. [DOI] [PubMed] [Google Scholar]
  38. Qiao L.; Fang W. H.; Long R.; Prezhdo O. V. Extending Carrier Lifetimes in Lead Halide Perovskites with Alkali Metals by Passivating and Eliminating Halide Interstitial Defects. Angew. Chem., Int. Ed. 2020, 59 (12), 4684–4690. 10.1002/anie.201911615. [DOI] [PubMed] [Google Scholar]
  39. Zhang Z.; Liu L.; Fang W.-H.; Long R.; Tokina M. V.; Prezhdo O. V. Plasmon-Mediated Electron Injection from Au Nanorods into MoS2: Traditional versus Photoexcitation Mechanism. Chem. 2018, 4 (5), 1112–1127. 10.1016/j.chempr.2018.02.025. [DOI] [Google Scholar]
  40. Chaban V. V.; Prezhdo V. V.; Prezhdo O. V. Covalent Linking Greatly Enhances Photoinduced Electron Transfer in Fullerene-Quantum Dot Nanocomposites: Time-Domain Ab Initio Study. J. Phys. Chem. Lett. 2013, 4 (1), 1–6. 10.1021/jz301878y. [DOI] [PubMed] [Google Scholar]
  41. Li L.; Long R.; Prezhdo O. V. Charge Separation and Recombination in Two-Dimensional MoS2/WS2: Time-Domain ab Initio Modeling. Chem. Mater. 2017, 29 (6), 2466–2473. 10.1021/acs.chemmater.6b03727. [DOI] [Google Scholar]
  42. Zhou Z.; Liu J.; Long R.; Li L.; Guo L.; Prezhdo O. V. Control of Charge Carriers Trapping and Relaxation in Hematite by Oxygen Vacancy Charge: Ab Initio Non-adiabatic Molecular Dynamics. J. Am. Chem. Soc. 2017, 139 (19), 6707–6717. 10.1021/jacs.7b02121. [DOI] [PubMed] [Google Scholar]
  43. Kresse G.; Furthmüller J. Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set. Phys. Rev. B: Condens. Matter Mater. Phys. 1996, 54 (16), 11169–11186. 10.1103/PhysRevB.54.11169. [DOI] [PubMed] [Google Scholar]
  44. Perdew J. P.; Burke K.; Ernzerhof M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1996, 77 (18), 3865–3868. 10.1103/PhysRevLett.77.3865. [DOI] [PubMed] [Google Scholar]
  45. Blöchl P. E. Projector augmented-wave method. Phys. Rev. B: Condens. Matter Mater. Phys. 1994, 50 (24), 17953–17979. 10.1103/PhysRevB.50.17953. [DOI] [PubMed] [Google Scholar]
  46. Chadi D. J.; Cohen M. L. Special Points in the Brillouin Zone. Phys. Rev. B 1973, 8 (12), 5747–5753. 10.1103/PhysRevB.8.5747. [DOI] [Google Scholar]
  47. Liechtenstein A. I.; Anisimov V. I.; Zaanen J. Density-functional theory and strong interactions: Orbital ordering in Mott-Hubbard insulators. Phys. Rev. B: Condens. Matter Mater. Phys. 1995, 52 (8), R5467–R5470. 10.1103/PhysRevB.52.R5467. [DOI] [PubMed] [Google Scholar]
  48. Morgan B. J.; Watson G. W. A DFT+U description of oxygen vacancies at the TiO2 rutile (110) surface. Surf. Sci. 2007, 601 (21), 5034–5041. 10.1016/j.susc.2007.08.025. [DOI] [Google Scholar]
  49. Scanlon D. O.; Morgan B. J.; Watson G. W. Modeling the polaronic nature of p-type defects in Cu2O: The failure of GGA and GGA+U. J. Chem. Phys. 2009, 131 (12), 124703. 10.1063/1.3231869. [DOI] [PubMed] [Google Scholar]
  50. Grimme S.; Antony J.; Ehrlich S.; Krieg H. A consistent and accurate ab initio parametrization of density functional dispersion correction (DFT-D) for the 94 elements H-Pu. J. Chem. Phys. 2010, 132 (15), 154104. 10.1063/1.3382344. [DOI] [PubMed] [Google Scholar]
  51. Grimme S.; Ehrlich S.; Goerigk L. Effect of the damping function in dispersion corrected density functional theory. J. Comput. Chem. 2011, 32 (7), 1456–1465. 10.1002/jcc.21759. [DOI] [PubMed] [Google Scholar]
  52. Di Valentin C.; Pacchioni G.; Selloni A. Reduced and n-Type Doped TiO2: Nature of Ti3+ Species. J. Phys. Chem. C 2009, 113 (48), 20543–20552. 10.1021/jp9061797. [DOI] [Google Scholar]
  53. Akimov A. V.; Prezhdo O. V. Advanced Capabilities of the PYXAID Program: Integration Schemes, Decoherence Effects, Multiexcitonic States, and Field-Matter Interaction. J. Chem. Theory Comput. 2014, 10 (2), 789–804. 10.1021/ct400934c. [DOI] [PubMed] [Google Scholar]
  54. Howard C. J.; Sabine T. M.; Dickson F. Structural and thermal parameters for rutile and anatase. Acta Crystallogr., Sect. B: Struct. Sci. 1991, 47 (4), 462–468. 10.1107/S010876819100335X. [DOI] [Google Scholar]
  55. Zhang W.; Yin J.-R.; Tang X.-Q.; Zhang P.; Ding Y.-H. Density functional theory studies on the structural and physical properties of Cu-doped anatase TiO2(101) surface. Phys. E 2017, 85, 259–263. 10.1016/j.physe.2016.09.008. [DOI] [Google Scholar]
  56. Nørskov J. K.; Bligaard T.; Logadottir A.; Kitchin J. R.; Chen J. G.; Pandelov S.; Stimming U. Trends in the Exchange Current for Hydrogen Evolution. J. Electrochem. Soc. 2005, 152 (3), J23. 10.1149/1.1856988. [DOI] [Google Scholar]
  57. Henkelman G.; Arnaldsson A.; Jónsson H. A fast and robust algorithm for Bader decomposition of charge density. Comput. Mater. Sci. 2006, 36 (3), 354–360. 10.1016/j.commatsci.2005.04.010. [DOI] [Google Scholar]
  58. Mukamel S.Principles of nonlinear optical spectroscopy. Oxford University Press: 1995. [Google Scholar]
  59. Akimov A. V.; Prezhdo O. V. Persistent Electronic Coherence Despite Rapid Loss of Electron–Nuclear Correlation. J. Phys. Chem. Lett. 2013, 4 (22), 3857–3864. 10.1021/jz402035z. [DOI] [Google Scholar]
  60. Ohsaka T.; Izumi F.; Fujiki Y. Raman spectrum of anatase, TiO2. J. Raman Spectrosc. 1978, 7 (6), 321–324. 10.1002/jrs.1250070606. [DOI] [Google Scholar]
  61. Swamy V.; Kuznetsov A.; Dubrovinsky L. S.; Caruso R. A.; Shchukin D. G.; Muddle B. C. Finite-size and pressure effects on the Raman spectrum of nanocrystalline anatase TiO2. Phys. Rev. B: Condens. Matter Mater. Phys. 2005, 71 (18), 184302. 10.1103/PhysRevB.71.184302. [DOI] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

au1c00004_si_001.pdf (1.5MB, pdf)

Articles from JACS Au are provided here courtesy of American Chemical Society

RESOURCES