Introduction

The amount of solar energy incident on the Earth in a single hour exceeds the total annual energy consumption of the global population. Consequently, solar energy represents a viable alternative to fossil fuels1,2. In order to facilitate the wider implementation of solar energy, it is essential to develop a framework for the sustainable and environmentally friendly storage of chemical fuels. Hydrogen is a candidate renewable fuel; it is an energy-dense carrier that produces only water upon combustion with oxygen3,4,5,6.

The electrocatalytic hydrogen evolution reaction (HER) requires an external electrical input, the sustainability of which depends on its primary generation source. On the other hand, dye-sensitized photocatalysis (DSP) has been proven to be a promising method for directly converting solar energy into H2 without using electric power. In our specific system for photocatalytic HER, tetrabromofluorescein (EY) is employed as the photosensitizer to absorb visible light7,8. For a standard DSP process, photoexcitation of the sensitizer is followed by electron–hole separation and transfer to the catalysts and sacrificial agents to initiate reductive and oxidative reactions, respectively9,10. The charge separation and transfer, which are crucial for the performance of DSP systems, require suitable catalysts with fast charge transfer kinetics, both internally and interfacially11,12,13,14,15.

Until now, noble metals (and noble-metal-containing materials), such as Pt, Ir, and Pd, as well as the state-of-the-art HER catalysts showing high electrocatalytic HER performance16,17,18,19,20, have been well studied, and some well-known concepts such as d-band theory have been proposed21,22,23,24,25. While noble metals exhibit favorable electron conductivity, their scarcity and cost limit large-scale application26,27,28. Consequently, a central challenge is the development of non-noble metal catalysts that exhibit comparable activity. Various low-cost catalysts based on earth-abundant materials have been explored for HER, such as transition metal chalcogenides29,30, transition metal phosphides31,32, transition metal carbides33,34, transition metal nitrides35,36, and so on. Non-noble metal photocatalysts with high carrier mobility and density are worth investigating for photocatalytic hydrogen production under visible light.

The possibilities for developing high-performance catalysts have been expanded after the discovery of topological materials37,38. Topological materials, which include topological insulators (TIs), topological semimetals (e.g., Dirac, Weyl, and nodal-line semimetals), and topological superconductors, are characterized by electronic boundary states protected by their band topology39. The protected surface states in topological materials host mobile charge carriers, a property expected to promote efficient charge transport and long-term stability, ultimately enhancing photocatalytic activity37,38,39,40. Several topological semimetals like (Nb,Ta)/(As,P) and MIrTe4 (M = Nb,Ta) have been investigated for dye-sensitized HER, and the hydrogen production rate of NbIrTe4 was reported to be 18,000 μmol g−1 (10 h)−1 41. However, the range of topological materials applied to photocatalytic systems is still limited. Moreover, the relationship between topological properties and the photocatalytic HER reaction process requires further investigation41,42. Further progress in photocatalytic HER is limited by two key challenges: inefficient utilization of solar light and the recombination of charge carriers. These challenges often cause low solar-to-hydrogen conversion efficiency and necessitate the exploration of photocatalysis with earth-abundant elements.

Our work aims to design an effective photocatalyst by exploring the properties of topological materials. The transition metal pnictide CoAs3 is one of the promising candidates, as its predicted topological electronic structure is expected to yield favorable charge transport properties. In this work, we prepared the topological semimetal CoAs3 for dye-sensitized HER, as well as other topological materials, including the semimetals WTe2, HfP2, and the insulators Mo2C, Bi2Te3, and Bi2Se3. We evaluated the photocatalytic activity of these materials, found that the noble-metal-free photocatalyst CoAs3 achieves a hydrogen production rate of 2688 μmol h−1 g−1 under visible light irradiation (λ ≥ 420 nm). Furthermore, transient absorption (TA) spectroscopy and density functional theory (DFT) calculations indicate that this high catalytic activity can be attributed to effective electron transfer at the dye/CoAs3 interface and the material’s topological semimetal nature. Our results suggest that the carrier dynamics in such materials can be utilized to address the existing challenges, providing a foundation for designing improved photocatalysts.

Results and discussion

Characterizations of CoAs3 compounds

The cubic crystalline structure of the skutterudite-type compounds CoAs3 in the Im\(\bar{3}\) symmetry is shown in the insert image in Fig. 1a, where the Co (deep blue) and As (green) atoms occupy the 8c and 24g Wyckoff positions, respectively. The six-coordinated octahedron is formed by one cobalt atom and six arsenic atoms, with a Co-As bond length of 2.345 Å. The powder X-ray diffraction (XRD) pattern (Fig. 1a) shows sharp diffraction peaks, all of which can be indexed to the cubic skutterudite phase, indicating a high crystalline quality and phase-pure sample. The diffraction patterns were indexed to the cubic space group Im\(\bar{3}\) (Fig. 1a and Table s2). Rietveld refinement of the XRD data for CoAs3 yielded lattice parameters consistent with previously reported values43. Scanning electron microscopy images (Fig. S13) show that the CoAs3 crystals have an average size of 1.2 μm, with individual crystals ranging from 0.5 to 3 μm. Energy dispersive X-ray (EDX) spectroscopy mapping of a CoAs3 microparticle (Fig. 1c–e) reveals a spatially uniform distribution of Co and As, with no other elements detected. The atomic structure of the CoAs3 microcrystals was further investigated using high-resolution scanning transmission electron microscopy (HR-STEM) and selected-area electron diffraction (SAED). The HR-STEM images (Fig. 1h, i) and SAED pattern (Fig. 1g) confirm the crystal structure, and the measured lattice constants are consistent with reported values44. X-ray photoelectron spectroscopy (XPS) was used to analyze the chemical states of the CoAs3 polycrystals (Fig. 1j, k). The Co 2p spectrum shows two main peaks at 779.3 eV (2p3/2) and 794.3 eV (2p1/2), which correspond to a mixed valence state of Co2+ and Co3+. The As 3d spectrum exhibits peaks at 42.2 and 41.4 eV, corresponding to Co-As bonds. These binding energies are all consistent with previously reported values45,46.

Fig. 1: Crystalline structure, X-ray diffraction, SEM images, element mappings, and XPS characterization of CoAs3 microcrystal.
figure 1

a The final Rietveld refinement of XRD spectrum of CoAs3 polycrystals. (left insert) Structural model of CoAs3 (space group Im\(\bar{3}\)). Blue and green balls represent Co and As atoms, respectively. (right insert) SEM image of CoAs3 microcrystal. b–e SEM images (b) and element mappings of CoAs3 flake, showing a uniform distribution of Co + As (c), Co (d), and As (e) elements. The average Co:As atomic ratio determined by the energy-dispersive X-ray spectroscopy (EDX) spectra is close to 1:3, and no foreign elements were detected within the detection limit of the instrument; f–i Atomic-level TEM image and low-magnification scanning transmission electron microscopy (STEM) image. (insert) of CoAs3 microcrystals. The yellow rectangle at the top right of the insert graph indicates the area where the atomic-level TEM (f) and the selected-area electron diffraction (SAED) pattern (g) were measured. Red circles in (g) mark the primary diffraction spots. h, i Atomic-level TEM images of a CoAs3 crystal taken at different locations, where 4.2 Å corresponds to the distance to the nearest neighboring Co. j, k X-ray photoemission spectroscopy (XPS) spectra of j Co 2p and k As 3d for freshly prepared CoAs3 polycrystals. Source data for Fig. 1 are provided as a Source data file.

Full size image

Photocatalytic activity of CoAs3

To identify the topological properties beneficial for photocatalytic HER, we compared the dye-sensitized HER performance of CoAs3 with that of other topological materials, including WTe2, HfP2, Mo2C, Bi2Te3, and Bi2Se3. These materials were selected to represent a range of topological properties and band structures near the Fermi level (EF) and were tested under identical experimental conditions. EY and triethanolamine (TEOA) were used as the photosensitizer and the sacrificial electron donor, respectively. The photocatalytic activity of these materials remains stable over 5 h (Fig. 2a); such stability may be attributed to their protected topological surface states47,48. Among these materials, CoAs3 achieves a photocatalytic HER rate of 684 μmol h−1 g−1, 14.3 times that of Bi2Se3. The electrochemically active surface area (ECSA) of each photocatalyst was determined from cyclic voltammetry measurements (Supplementary Figs. S1–S3). Although the ECSA of CoAs3 (1.61 × 10−4 mF cm−2) is approximately half that of Mo2C, its photocatalytic activity exceeds that of other materials tested. This discrepancy points to the contribution of the electronic property of CoAs3 to its catalytic activity, which we investigate in the following sections.

Fig. 2: Photocatalytic activity of various topological materials.
figure 2

a Hydrogen production rates of CoAs3, WTe2, HfP2, Mo2C, Bi2Te3 and Bi2Se3. b Eosin Y dye-sensitized photocatalytic HER activity of CoAs3 after 5 h of irradiation, normalized with respect to the catalyst amount in grams, with different amounts of catalyst. c Hydrogen production rate of CoAs3 at different amounts of dye. d Control experiments of hydrogen production in the dye sensitization system. P25 represents for TiO2 nanoparticles photocatalyst of ca. 25 nm. CoAs3-Vis represents photocatalysis under visible light, while CoAs3-FS refers to photocatalysis under a full-spectrum light. e The variation of the quantum apparent efficiency (AQE) of photocatalyst CoAs3 vs. radiation wavelength (blue dots) and the UV-vis absorption spectrum of EY (red curve). f The reproducibility of the EY dye-sensitized photocatalytic hydrogen evolution reaction (HER) activity of CoAs3 for the optimum dye/catalyst ratio. In (b, c, f), data are presented as mean ± s.d. from three independent experiments (n = 3). Source data for Fig. 2 are provided as a Source data file.

Full size image

The photocatalytic activity of CoAs3 was investigated by varying the catalyst mass and photosensitizer concentration. To determine the effect of catalyst mass, the concentrations of the EY and the TEOA were kept constant at 10 mg and 15 mL, respectively. Figure S4 shows that H2 can be produced in the absence of CoAs3 catalysts, and the activity increased with the addition of CoAs3. Therefore, the hydrogen evolution rate attributed to CoAs3 can be calculated by subtracting the rate of the control experiment from the total measured rate. As shown in Fig. 2b, the normalized activity of CoAs3 remains stable between 1 and 25 mg, which well reflects the intrinsic activity of CoAs3. When the catalyst mass was increased beyond 30 mg, the hydrogen evolution rate decreased. This rate decrease is attributed to the large amount of catalyst masking the active sites and blocking the light absorption49. Therefore, in subsequent experiments, the amount of CoAs3 was fixed at 25 mg.

The dosage of photosensitizer was then adjusted from 10 to 80 mg, while keeping the catalyst and sacrificial agent constant (Fig. 2c). The hydrogen evolution rate initially increased as the amount of photosensitizer was raised. The rate reached a maximum value of 2697 μmol h−1 g−1 at a photosensitizer amount of 70 mg, and subsequently decreased when the amount was increased to 80 mg (Fig. 2c). This decrease at higher concentrations is attributed to two factors: (1) a shading effect caused by excess photosensitizer, and (2) collisional quenching of the excited photosensitizer (*EY) by ground-state EY molecules, which can deactivate *EY before interfacial electron transfer occurs11,50,51.

Control experiments show that hydrogen evolution is negligible in the absence of any of the three components: the CoAs3 catalyst, the EY photosensitizer, or the TEOA sacrificial agent (Fig. 2d). Under full-spectrum irradiation, the hydrogen evolution rate increased to 3232 μmol h−1 g−1. In contrast, graphene with super electron transfer efficiency was also used for comparison, which exhibits a lower activity compared with CoAs3. These results suggest that CoAs3 not only serves as an effective charge transfer material but also provides abundant HER active sites in this system. It is important to note that the rate of photocatalysis is likely to vary substantially depending on the effects of a given temperature, pressure, light intensity, and other factors. The ultraviolet–visible absorption spectra of CoAs3 and Bi2Se3 exhibit minimal absorption in the visible range (Fig. S12). This property suggests they do not function as the primary light absorbers in this photosensitized system. The absorption peaks of pure EY solution are at 480 and 518 nm due to π-π* excitation41, which is consistent with previous results52. When EY powder is mixed with CoAs3 and Bi2Se3 powders, respectively, absorption peaks of 487 and 526 nm appear near the characteristic peaks of the dye, indicating a redshift. This redshift suggests an electronic interaction between EY and the CoAs3 or Bi2Se3 materials. Such an interaction may facilitate the transfer of photogenerated electrons from the excited photosensitizer to the catalyst53. Additionally, a color change in the solution from red–orange to saffron was observed during the photocatalytic process, suggesting degradation of the EY photosensitizer over the reaction time.

The apparent quantum efficiency (AQE) was measured at several incident light wavelengths (Fig. 2e) (See the “Methods” section for detailed experimental and calculation procedures.). The CoAs3-EY system exhibited a maximum AQE of 15.2% at an incident wavelength of 500 nm. Relatively high AQE values can also be obtained at 450 and 480 nm with lower EY absorption intensity. The system remains active at shorter wavelengths despite the reduced light absorption by the photosensitizer.

To investigate the stability of the CoAs3 photocatalyst at the optimal ratio of 25 mg CoAs3 and 70 mg photosensitizer, a set of cycling experiments with a unit cycle of 5 h was conducted. Over eight consecutive 5-h cycles, the hydrogen evolution rate showed a minor decrease in each cycle (Fig. 2f). In the recovery experiment (ninth cycle), the catalysts were cleaned and centrifuged before being reused in a fresh experimental solution. Upon reuse, the catalyst’s activity was restored to its initial level, yielding 13,692 μmol g−1 of H2 in 5 h. Due to the inferior stability of EY54, the decreased activity of the DSP system may be ascribed to the reduced concentration of EY. Therefore, after the recovery of EY concentration, the activity can be well-restored, indicating a stable performance of CoAs3 catalyst. The EDX and XPS analyses (Supplementary Fig. S5) of the catalysts performed after the cycling experiments showed no observable changes in their components or morphology. Collectively, these findings indicate that CoAs3 is a structurally and functionally stable catalyst for photocatalytic hydrogen evolution.

Electron transfer efficiency at the EY/CoAs3 interface

To investigate the electron-transfer process during DSP photocatalytic HER, amperometric current density I–t curves were measured (Supplementary Fig. S6) in a 0.1 M Na2SO4 and TEOA aqueous solution. In the absence of a catalyst, the EY-sensitized electrode exhibited a transient anodic photocurrent spike upon illumination. This feature is characteristic of charge accumulation at the electrode-sensitizer interface followed by rapid recombination, as the photogenerated electrons are not transferred to an acceptor55,56,57. This behavior indicates fast charge recombination when no effective electron acceptor presents58,59.

In contrast, the introduction of CoAs3 catalyst changes the carrier dynamics. For the EY/CoAs3 system, the photocurrent increases throughout the illumination period. This photoactivation is attributed to light-induced effects that occurred at the dye-catalyst interface, such as a stable dye adsorption and a dynamic adsorption–desorption equilibrium, which enhance photosensitization efficiency60. Furthermore, the slow current decay after illumination ceases indicates that CoAs3 possesses charge-storage capacity, allowing trapped electrons to subsequently participate in the HER58. Notably, while the EY/Bi2Se3 system shows a similar photoactivation trend, its total photocurrent density is much lower. This lower photocurrent is observed despite Bi2Se3 being a TI with surface states reported to support efficient charge transport. Therefore, this result strongly suggests that the rate-determining step is not charge transport but the slow kinetics of the surface HER. These comparative results indicate that the overall photoelectrochemical efficiency depends on a combination of interfacial charge separation, bulk transport, and surface reaction kinetics. The high performance of the EY/ CoAs3 system should be due to a better balance of many factors, particularly the higher intrinsic HER activity and favorable photoactivation properties of the CoAs3 surface.

The kinetics of interfacial charge transfer were investigated using electrochemical impedance spectroscopy. The charge transfer resistance (Rct) of the EY-sensitized CoAs3 and Bi2Se3 systems was compared. Both the EY/CoAs3 and EY/Bi2Se3 systems exhibited their lowest Rct at 520 nm, as shown in Fig. S11. This wavelength corresponds to the primary absorption peak of the EY photosensitizer, suggesting that the rate of electron injection from the excited dye to the catalyst is dependent on the amount of light absorbed. Moreover, under all comparable illumination wavelengths, the EY/CoAs3 system consistently displayed a smaller Rct than the EY/Bi2Se3 system (specific values summarized in Table S5). This lower Rct value for the CoAs3 system indicates more rapid interfacial electron transfer kinetics compared to the Bi2Se3 system. This process includes both electron injection from the dye to the catalyst and the subsequent utilization of charge in the HER.

The interfacial charge transfer mechanism between photoexcited EY and the catalysts was investigated using in situ electron paramagnetic resonance (EPR) spectroscopy measured under illumination. In contrast to the weak intrinsic signal observed for pure EY under illumination, attributed to the formation of the excited state, the EY/catalyst mixtures exhibited a dramatic increase in signal intensity at g = 2.001 upon illumination (Fig. S15). This prominent g-value is characteristic of an organic radical, consistent with the formation of the EY radical cation (EY+). Quantitatively, the EPR signal intensity for the EY/Bi2Se3 and EY/CoAs3 mixtures was enhanced by factors of 6.1 and 26, respectively, relative to that of pure EY under the same illumination. This signal enhancement provides the spectroscopic evidence for efficient electron transfer from photoexcited EY to the catalyst, leading to the generation of the EY+ radical cation. The larger signal enhancement factor for CoAs3 (26) compared to that for Bi2Se3 (6.1) indicates that CoAs3 more readily accepts an electron from photoexcited EY, leading to more effective charge separation.

Femtosecond TA spectroscopy was used to investigate the ultrafast charge transfer dynamics between the EY photosensitizer and the prepared catalysts with an excitation wavelength of 570 nm. By systematically investigating the ultrafast carrier dynamic processes of pure EY solution and the EY/Cat composites, we have obtained direct spectroscopic evidence of photoinduced electron transfer (PET) and resolved the subsequent multi-step kinetic pathways. The two-dimensional pseudo-color maps for pristine EY/CoAs3 and EY composite are presented in Fig. 3a, d, respectively, with the data for EY/Bi2Se3 provided in the Supporting Information (Fig. S7a).

Fig. 3: Ultrafast carrier dynamics at the EY/catalyst interface.
figure 3

Time slices of the TA spectra of a, b EY/CoAs3 composite and d, e pure EY solution under excitation at 570 nm. Comparison of the TA kinetic probed at different wavelengths in EY/CoAs3 (c) and EY (f). g Comparison of the TA kinetic probed at 420 nm between in EY solution (green), EY/CoAs3 (red), and EY/Bi2Se3 (blue). Solid lines are their tri-exponential decay fitting. Source data for Fig. 3 are provided as a Source data file.

Full size image

To establish a benchmark, the transient behavior of pristine EY was characterized to understand its intrinsic photophysical processes. The TA spectrum of pristine EY (Fig. 3e) exhibits a positive photoinduced absorption (PIA) band centered at 420 nm and a negative bleach signal centered at 510 nm. According to the previous report, the PIA signal is assigned to the absorption of singlet *EY, while the bleach signal comprises the broad ground-state bleach and stimulated emission of EY61,62. In stark contrast, the introduction of the catalysts in EY/Cat composites leads to an obvious change in their TA spectra. Besides the PIA band centered at 420 nm (for *EY) and the negative bleach signal at 510 nm present in pure EY, both the TA spectra of EY/CoAs3 and EY/Bi2Se3 (Figs. 3b and S7a) still show an additional PIA signal with its peak at 470 nm, which is the definitive signature of the EY radical cation (EY+). This emergence of EY+ provides direct spectroscopic evidence for electron transfer from the excited EY to the catalyst, which can also be clearly reflected in their respective TA kinetics.

In pristine EY, the similar TA kinetics probed, respectively, at 420 and 445 nm are observed (Fig. 3f), while in EY/Cat composites, their TA kinetics probed at 420 and 470 nm exhibit completely different behaviors. The PIA signal at 420 nm shows a fast decay in the initial 2 ps delay, accompanied by a clear rising process of PIA signal at 470 nm in the same time scale (Figs. 3c and S7c), again verifying the PET from EY to the catalyst. Fig. 3g compares the PIA kinetics probed at 420 nm in pure EY and the EY/Cat composites. Compared with pure EY, both the excited states in the two EY/Cat composites exhibit evidently faster decay kinetics, which can be well fitted by a tri-exponential function (Table 1). Obviously, the fastest process (with τ1 = 2.1 ps and 3.2 ps for EY/CoAs3 and EY/Bi2Se3, respectively) is attributable to the PET process, also consistent with the rise time observed in the PIA kinetics at 470 nm, while the slowest time constant exactly represents the lifetime of the charge-separated state. We also note that the faster process with time constant (τ2) about tens of ps, likely originates from the accelerated decay of EY excited states in EY/Cat composites61. The above kinetics comparison between EY/CoAs3 and EY/Bi2Se3 suggests the faster PET rate and longer lifetime of charge-separated state at the EY/CoAs3 interface over the Bi2Se3 interface, which can account for the higher photocatalytic hydrogen production activity of the CoAs3 catalyst.

Table 1 Summary of kinetic parameters obtained from multi-exponential fitting of the transient absorption decays at 420 nm
Full size table

DFT-calculated electronic structure of CoAs3

Following electron transfer from the excited photosensitizer, the subsequent transport of electrons to surface active sites is governed by the electronic structure of the catalyst material. We therefore investigate the electronic structure of CoAs3 by DFT calculations and compare it with that of other photocatalysts investigated in the current work.

The concept of topological catalysis has recently attracted considerable interest38,63,64,65,66,67. In contrast to TIs, the bulk states of topological semimetals are conductive. This conductivity, along with elevated electronic mobility, is thought to be beneficial for catalytic processes37,38. In addition, topological semimetals are found to carry a large number of topological emergent characters, including different categories of nodal points, nodal lines, and nodal surfaces65,68,69. These topological emergent characters also show diverse surface states. Consequently, both theoretical and experimental studies have identified topological semimetals as promising catalysts. Their proposed catalytic advantages are attributed to a combination of properties, including bulk-state conductivity, topologically protected surface states, and favorable molecular adsorption energies6,41,70,71,72,73,74,75,76,77,78,79. However, the mechanism responsible for the photocatalytic activity of CoAs3 remains unclear, and the specific material properties that determine its performance in the HER have yet to be identified.

DFT calculations were performed to investigate the band structure of CoAs3 (Fig. 4a), which provides insight into its photocatalytic behavior. The relaxed crystalline structure of CoAs3 is provided as Supplementary Dataset 1. Without considering the spin–orbit coupling (SOC) effect, a threefold band crossing node appears at the high-symmetric Γ point, characterized by a topological charge c = 0 (Fig. 4b). Upon incorporating SOC, this threefold degeneracy evolves into a sixfold band crossing node. This node further splits into a twofold band crossing (triple contact-point (TCP)) and a fourfold degenerate quadratic contact-point (QCP), a phenomenon attributed to the coexistence of center-reversal symmetry and time-reversal symmetry. The presence of these degenerate points is indicative of the electronic structure of CoAs3, potentially offering high carrier mobility and efficient electron pathways as well as efficient charge separation and transport in photocatalytic hydrogen evolution. In between the TCP and QCP, there is a TI gap and a Dirac-like topological surface state (Fig. 4c). The QCP along with specific surface states constitute a Luttinger semimetal (LSM)80,81,82, where this degeneracy appears as a quadratic band touching of j = 3/2 fermions. LSMs have been studied due to their proximity to Z2 TIs83, Dirac84, and Weyl semimetals85. This nodal point is formed by the crossing of the valence band maximum and two conduction bands near the conduction band minimum and is protected by time-reversal symmetry and spatial-inversion symmetry. Detailed band structure analysis (Fig. 4b) indicates that the fourfold band degeneracy at the Γ point, near the Fermi level, primarily originates from the Co d-orbitals. Concurrently, with the inclusion of SOC, the sixfold nodal point transforms into a fourfold degenerate quadratic Dirac point (Fig. 4c)84. To further investigate the nontrivial band topology of CoAs3, its surface band structure was calculated (Fig. 4d, e). These calculations show both Luttinger surface states (LSS) crossing the EF and the concomitant appearance of trivial surface states. The presence of these conductive surface states, particularly those crossing EF, can serve as efficient channels for transferring photogenerated electrons to the surface active sites, thereby reducing charge recombination and enhancing hydrogen evolution efficiency. Further calculations reveal the spin texture of the LSS (Fig. 4f), exhibiting a characteristic spin-momentum locked pattern. This spin-momentum locked pattern can suppress backscattering of electrons, thereby reducing recombination, a process beneficial for photocatalysis. A magnified view of the region near the Γ point at the Fermi level (EF) shows that the TI surface states become apparent (Fig. 4g). Since both the LSS and the TI surface states are protected by chiral symmetry and time-reversal symmetry, respectively, they are insensitive to disorder or impurities that do not break these specific symmetries. This property helps to ensure that efficient and stable charge carrier transport channels persist even in imperfect or complex catalytic environments, thereby contributing to sustained activity and high overall efficiency. The surface states emanated from the bulk nodal points of CoP3 have been observed by angle-resolved photoemission spectroscopy measurements86 and are seen to cross through EF. Compared to Bi2Se3, CoAs3 possesses both a fourfold degenerate quadratic Dirac point near EF and the LSS crossing the EF. These features contribute to a higher carrier density and high mobility in CoAs3, whereas Bi2Se3 has only conventional topological surface states crossing EF. The band structure calculations (Figs. 4 and S8) indicate that the d-orbitals of Co are the primary contributors to the electronic states near the EF, leading to the formation of a topological Dirac cone. These topological features could enhance the efficiency of catalytic reactions such as the photocatalytic HER.

Fig. 4: The calculated band structure and surface states of CoAs3, and the proposed mechanism of enhanced photocatalytic HER performance.
figure 4

a Electronic band structure of CoAs3 considering the SOC. Insert: bulk Brillouin zones of CoAs3. b, c The enlarged view of the dashed red area for the enlarged band crossing points with and without SOC along the high symmetry path N–Γ–P in (a). The quadratic contact point at Γ above EF. (TCP triple contact-point, QCP quadratic contact-point, LS Luttinger semimetal). d, e Surface band structures of the top and down surfaces. Note, the trivial surface states are induced by unsaturated dangling bonds on the surface, where the surface states of Luttinger semimetal are clearly shown. f Color mapping of the calculated spin texture of Luttinger semimetal (LS) surface states. g Color plot of the calculated topological insulator (TI) surface states. h The proposed mechanism of enhanced visible light photocatalytic HER performance of CoAs3. Source data for Fig. 4 are provided as a Source data file.

Full size image

The intrinsic nature of high dye-sensitized activity of HER using CoAs3 catalyst can be rationally explained in Fig. 4h. Upon light absorption, the EY photosensitizer generates an electron–hole pair. The electron is injected into the CoAs3 catalyst, leaving a hole on the EY molecule. The conductive topological surface states of CoAs3 facilitate the transport of these injected electrons away from the interface, thus suppressing charge recombination. This process contrasts with that in conventional semiconductors, where charge carriers are often annihilated by defects. In CoAs3, the topological surface states allow electrons to bypass such defects, which helps to preserve the population of photo-generated electrons. Then, the surviving surface electrons are captured by protons adsorbed on the surface and generate H2. Furthermore, the Co d-band crosses the EF (Supplementary Fig. S7). These states are closely related to hydrogen interaction and are proposed to act as the active sites for HER. Meanwhile, the sacrificial electron donor (TEOA) contributes electrons to the photosensitizer and maintains the surface redox reactions.

In summary, this work introduces the topological LSM CoAs3 as a catalyst for dye-sensitized photocatalytic hydrogen evolution. A comparison with other topological materials was performed to evaluate its performance. Under visible light (λ ≥ 420 nm), the CoAs3-based DSP system achieved a hydrogen production rate of 2688 μmol h−1 g−1. Ultrafast spectroscopy confirmed rapid electron transfer at the EY/CoAs3 interface, and theoretical calculations pointed to the role of the material’s topological semimetal nature in its photocatalytic activity. This work identifies a topological semimetal as a stable and active catalyst for photocatalytic HER under visible light. The activity of the system under broad-spectrum illumination suggests a potential for hydrogen production using natural sunlight.

Methods

Reagents

Cobalt (Sigma Aldrich, 99.99%), Arsenic (Sigma Aldrich, 99.99%), Tungsten (Alfa Aesar, 99.95%), Tellurium (Alfa Aesar, 99.999%), molybdenum (Alfa Aesar, 99.95%), carbon (Alfa Aesar, 99.9%), Bismuth selenide (Sigma Aldrich, 99.99%) Bismuth(III) telluride (Sigma Aldrich, 99.99%), Eosin Y (Sigma Aldrich, 99%), TEOA (ACROS organics, 99+%), TiO2 (Degussa P25).

Sample preparation

CoAs3: The sample were prepared via solid-phase reaction. Co and As were sealed in quartz tubes in a stoichiometric ratio, heated to 800 °C at a rate of 100 °C per hour, held for 48 h, and cooled down with the furnace.

WTe2: W and Te were ground stoichiometrically in an argon-filled glove box and sealed in an ampoule under vacuum, heated to 700 °C for 7 h and held for 1 day, then heated to 750 °C for 1 h and held for a period of time before being cooled in the furnace.

Mo2C: Carbon and molybdenum were mixed and ground in stoichiometric ratio in an argon-filled glove box and pressed into tablets, flame sealed in ampoules under vacuum and cooled with the furnace after 10 h of heating to 950 °C.

HfP2: The ingredients were combined and ground in stoichiometric proportions in an argon-filled glove box, then compressed into tablets. These were subsequently sealed under vacuum in ampoules and cooled with the furnace after 8 h of heating to 750 °C.

Characterization

The element distribution and the chemical composition of the products were determined through EDS mapping and spot analysis (Hitachi S-4800 scanning electron microscope) and TEM (HITACHI HF 5000), respectively. The chemical composition and crystal lattice structure were determined by XRD (Bruker-D8) with CuKα radiation over an angular range of 10–90° in 1 h at room temperature. The chemical valence states were detected using an XPS (Kratos AXIS ULTRA XPS).

DSP HER

The photocatalytic reactions occurring in hydrogen were carried out in a sealed gas–solid atmospheric pressure cylindrical glass reactor. The light source was a 300 W xenon lamp with a 420 nm cut-off filter. In all, 100 mg of catalyst powder was dispersed in 100 ml of freshly prepared aqueous TEOA (15%v/v), and 0.014 mmol (9.07 mg) of EY was added to the glass reactor, followed by sonication for 30 min to ensure homogeneous particle dispersion (ph = 10.15 ± 0.1). The vessel was positioned upon a magnetic stirrer, and the vacuum pump was activated to commence the photocatalytic process. This was followed by the illumination and agitation of the sample to initiate the reaction.

Apparent quantum efficiency

The AQE can be described based on the following formula:

$${AQE} \, (\%)={AR}{N}_{{{\rm{A}}}}/I\times 100$$

where A, R, NA, and I represent a coefficient (2 for H2 evolution), the evolution rate of H2, Avogadro's constant, and the number of photons absorbed by the reaction solution, respectively. For evaluating the AQE, the reactions were performed under similar reaction conditions except that the monochromatic light was used. As the light source, the monochromatic light was generated by Xe lamp equipped with the band-pass filters centered at 380, 420, 450, 500, 550, and 600 nm, respectively.

TA spectroscopy

The femtosecond pump-probe TA measurements were performed using a regenerative amplified Ti:sapphire laser system (Coherent; 800 nm, 70 fs, 6 mJ pulse−1, and 1 kHz repetition rate) as the laser source and a femto-TA100 spectrometer (Time-Tech Spectra). All experiments were performed under ambient conditions.

First-principles calculations

First-principles calculations were performed based on DFT by utilizing the Vienna ab initio simulation package87,88. The plane-wave basis was adopted with an energy cutoff of 350 eV. The electron–ion interactions were modeled with the projector augmented wave potential89, and the exchange–correlation functional was approximated by the Perdew–Burke–Ernzerhof-type generalized gradient approximation90. For optimized lattice constants and atomic positions, the structural relaxation was performed with an energy (force) criterion of 10−8 eV (0.01 eV Å−1) and by utilizing the DFT-D391 method to include van der Waals corrections. Surface state calculations were performed with the Wannier Tools package92, which is based on the tight-binding Hamiltonians constructed from maximally localized Wannier functions93.