Atomically dispersed Ni in cadmium-zinc sulfide quantum dots for high-performance visible-light photocatalytic hydrogen production

See allHide authors and affiliations

Science Advances  14 Aug 2020:
Vol. 6, no. 33, eaaz8447
DOI: 10.1126/sciadv.aaz8447


Catalysts with a single atom site allow highly tuning of the activity, stability, and reactivity of heterogeneous catalysts. Therefore, atomistic understanding of the pertinent mechanism is essential to simultaneously boost the intrinsic activity, site density, electron transport, and stability. Here, we report that atomically dispersed nickel (Ni) in zincblende cadmium–zinc sulfide quantum dots (ZCS QDs) delivers an efficient and durable photocatalytic performance for water splitting under sunlight. The finely tuned Ni atoms dispersed in ZCS QDs exhibit an ultrahigh photocatalytic H2 production activity of 18.87 mmol hour−1 g−1. It could be ascribed to the favorable surface engineering to achieve highly active sites of monovalent Ni(I) and the surface heterojunctions to reinforce the carrier separation owing to the suitable energy band structures, built-in electric field, and optimized surface H2 adsorption thermodynamics. This work demonstrates a synergistic regulation of the physicochemical properties of QDs for high-efficiency photocatalytic H2 production.


The design of catalysts with suitable chemical and physical properties is essential for many energy conversion and storage technologies, such as water electrolyzers, photocatalytic water splitting, metal-air batteries, and fuel cells. Introducing active single atoms into the catalysts has been the most widely used strategy to tune the electronic structure of the catalysts to optimize their intrinsic activities, electronic transport, active site densities, and stability (13). Furthermore, downsizing catalyst particles or clusters to single atoms is highly desirable for maximizing catalytic efficiency. It has been well recognized that the isolated active atoms singly dispersed in matrices can maximize the efficiency of atom utilization (hence, single-atom catalysts) in acidic (4) and alkaline (5) hydrogen evolution reaction (HER) (6), electrochemical reduction of CO2 (7), CO oxidation (2), etc.

The structurally uniform and well-defined single atomic sites in a simple structure provide atomic-level insight and the corresponding catalytic reaction mechanism (2, 8, 9). However, it is still a big challenge to understand the catalytically active center in “atomic interfaces” (typically with different charges or even different chemical identities) (7) within higher complexities. Obviously, the identification of an active site structure with more sophisticated functionalities is crucial for catalytic reactions that simultaneously require different kinds of functional active sites. Grimaud  and Xu et al. (10) found that substituting Al of inactive but low-cost CoAl2O4 with a small amount of Fe can activate the preoxidation of Co and optimize the O 2p level of oxide for greater structural flexibility to facilitating the surface reconstruction of CoAl2O4. They revealed that on the reconstructed surface, the Fe substitution facilitates a two-step deprotonation process, which leads to the formation of active oxygen sites at a low overpotential and thus greatly promotes the oxygen evolution reaction (OER) (10). The Cd1−xZnxS solid solution rather than bare CdS and ZnS demonstrated efficient photocatalytic properties for hydrogen (H2) evolution in the presence of sacrificial electron donors. Its photocatalytic activities can be optimized by adjusting the x value for achieving appropriate band structure for the photocatalytic H2 production (1113). Xu and Wang et al. (14) also unveiled the accurate amount of catalytically inactive Zn2+ incorporating into CoOOH that can give rise to oxygen nonbonding states with different local configurations, which is critical to regulating the OER mechanism. They found that if two neighboring nonbonding oxygens with the partially filled O(2p) band can hybridize their oxygen holes without sacrificing metal-oxygen hybridization substantially, the OER proceeds via the lattice oxygen oxidation mechanism pathway on the metal oxyhydroxides, discovering that Zn0.2Co0.8OOH has optimum activity (14). Identification of active site structures is even more vital for the photocatalytic H2 production through water splitting driven by visible light because of their complex multiple reaction processes: from light harvesting to charge (electron and hole pairs) excitation, further to photocatalyst surface engineering for both HER and OER (1517). Therefore, it is critical to investigate the efficiency of the introduced single atoms on a per-atom basis and further systematically evaluate the electronic and structural properties of the photocatalysts to realize synergetic integration of photocatalytic constituents into a functional heterostructure. One of the challenges to unveil the synergistic effect of the single atoms with their coordinating elements is the irregular structure of the bulk catalysts, resulting in the different exposed facets with anisotropic physical and chemical properties.

In contrast to the elusive irregular bulk materials, here we use a crystallographic and structural chemistry design approach to manipulate the atomically dispersed nickel (Ni) in zincblende cadmium–zinc sulfide quantum dots (ZCS QDs), maximizing the photocatalytic properties for water splitting driven by visible light. The colloidal QDs, such as ZnS-coated InP QDs (15) with tunable bandgap and versatile surface properties, remain among the most promising photocatalysis for the HER. Here, experimental investigation and theoretical density function theory (DFT) calculations revealed that with continuous adjustment on the concentration of atomically dispersed Ni, the anisotropy of different crystalline facets [(111) and (110)] of the ZCS QDs can be fine-tuned. In particular, the (111) facets consisting of isolated, high-density, low-valence Ni(I) as verified by the x-ray absorption fine structure analysis (XAFS), electron energy-loss spectroscopy (EELS) spectra, etc., show a highly catalytic activity for HER. Furthermore, surface heterojunctions due to the different exposed crystal planes [(111) and (110)] in the same phase maximize the charge carrier separation. Meanwhile, the built-in electric field (BIEF) can further facilitate the charge carrier migration to the surface, leading to enhanced electronic conductivity. Together with proper surface H2 adsorption thermodynamics on the atomically dispersed Ni(I) active sites, the optimized Ni0.03125Zn0.25Cd0.75S QDs achieved a remarkable photocatalytic H2 production rate of 18.87 mmol hour−1 g−1. This work could pave a new avenue for the synergistic modulation of the physicochemical properties of nanocrystals to substantially improve their performances for various applications, e.g., catalysis, electronics, and optoelectronics.


Phase structures and morphology

Stoichiometric ZCS QDs were synthesized through the hot injection approach, using metal chloride and thioacetamide (TAA) as precursors and oleylamine (OLA) as the reducing and stabilizing agent. The field emission scanning electron microscopy (FESEM), the high-resolution transmission electron microscopy (HRTEM), and the high-angle annular dark field scanning TEM (HAADF-STEM) images were used to identify the morphology of the as-prepared QDs. It shows that the monodisperse ZCS QDs with the uniform tetrahedral shape and a narrow size distribution (~8.5 nm) were obtained (fig. S1, A to D). They exhibit an equilateral triangle shape with a 60° angulus parietalis, and the fast Fourier transform (FFT) plots can be indexed as (111) crystal planes of the zincblende structure (fig. S1, E to G). The ZCS QDs are enclosed by (111) facets, as identified by the angles of the simulated outlines of the triangular pyramids sitting on the tetrahedral facets together with the HAADF-STEM (fig. S1, H to L). After dispersing Ni into ZCS QDs, we obtained four NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs. All these QDs are monodisperse (with an average size of ~8.0 nm). They self-assemble to form a superlattice structure (Fig. 1, A1 to D1 and A2 to D2, and fig. S2, A to H), as confirmed by the regular hexagonal arrangement FFT patterns (insets of Fig. 1, A1 to D1). The low polydispersity in size can be ascribed to the temperature-dependent release of reactive sulfur species from TAA and the optimized reactivity of metal-OLA complexes, which facilitates the better separation of nucleation and growth stages (fig. S2, I to R) (18). With increasing Ni content, the color of the as-prepared QDs gradually changes from yellow to dark green (fig. S2, S to W), indicating the composition variation over the entire series. The HAADF-STEM images (Fig. 1, A3 to D3) show that the vertices of the tetrahedra are truncated with ~120° facets, and the vertices become more truncated with increased Ni in NixZCS QDs, resulting in the shape evolution from tetrahedra to truncated tetrahedra as demonstrated by the geometrical models (Fig. 1, A6 to D6). The FFT patterns (Fig. 1, A5 to D5) of the atomically resolved HAADF-STEM images (Fig. 1, A4 to D4) verify their dominant (111) facets. Moreover, the Miller indices of the angulus parietalis on facets of the truncated tetrahedra are determined through analyzing the projection angles (19). It proves that only the (110) cobundled facets can contribute to 120° angulus parietalis for the truncated tetrahedral architecture (fig. S2X). Therefore, the as-prepared NixZn0.25Cd0.75S1+x QDs are enclosed by the (111) and (110) facets only. Figure S3 shows more detailed shape evolution images of the NixZCS QDs along [111], [100], [110], and [112] orientations, confirming the role of Ni in the evolution of the enclosed facets. For comparison, the bare CdS and ZnS QDs were also synthesized through the same process. They also present the uniform size distribution and the tetrahedral morphology (fig. S2, Y and Z). As characterized by FESEM, HRTEM, and HAADF-STEM, it confirms the uniform size, surface morphology, and crystallography of the as-prepared NixZCS QDs. Furthermore, the clearly identified anisotropy crystal facets of the NixZCS QDs could present the tunable electron structure to tune the photocatalytic activity.

Fig. 1 Morphologies of the metal sulfide QDs.

TEM (A1, B1, C1, and D1), HRTEM (A2, B2, C2, and D2), and HAADF-STEM (A3, B3, C3, and D3) images of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs. (A4), (B4), (C4), and (D4) are atomically resolved HAADF-STEM images of the typical freestanding NixZCS QDs viewed along the [111] orientation, and (A5), (B5), (C5), and (D5) are their corresponding FFT patterns. (A6), (B6), (C6), and (D6) are the simulated atomic models of the as-prepared NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs. Insets in (A1) to (D1) are the FFT patterns of the TEM images, showing their regular hexagonal arrangement. Insets in (A4) to (D4) are the simulated atom arrangement of the (111) crystal planes of the zincblende structure. Scale bars, 50 nm (A1, B1, C1, and D1), 10 nm (A2, B2, C2, and D2). 2 nm (A3, B3, C3, and D3), 5 Å (A4, B4, C4, and D4), and 5 1/nm (A5, B5, C5, and D5).

To gain deeper insight into the phase structures of the bulk and surface compositions of the NixZCS QDs, inductively coupled plasma mass spectrometry (ICP-MS), energy-dispersive x-ray spectroscopy (EDS), the Rietveld refinement x-ray diffraction (XRD) patterns, and the selected area electron diffraction (SAED) analyses were explored. ICP-MS and EDS measurements confirm that the chemical compositions of NixZCS QDs are close to the molar ratio of Ni, Zn, and Cd in the initial solution (table S1). Owing to the similar covalent radii of Ni, Zn, and Cd (121, 131, and 148 pm, respectively), Zn and Ni can readily substitute Cd within the NixZCS QDs, which was confirmed by the calculated lowest free energy of the substitute dopants (fig. S4, A to G, and table S2). Moreover, the formation energy of an impurity Ni is 0.45 eV in the Ni0.125ZCS when nickel chloride is used as Ni source, implying the high solubility of Ni in the ZCS matrix. This is also identified by the Rietveld refinement XRD patterns (fig. S4, H to L). The lattice parameters decreased from 5.712 Å for bare ZCS QDs to 5.706, 5.694, 5.665, and 5.631 Å for NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, respectively, which is well consistent with the DFT calculation results and Vegard’s law plot (fig. S4, M and N) (20). The main XRD peaks [(111), (220), and (311)] continuously shift toward higher angles without phase separation along with increasing Ni content, indicating the homogeneous structure of the NixZCS solid solutions (fig. S4l). As calculated (fig. S4, O to S), with Zn and Ni substitute, the Td symmetry of the [CdS4] tetrahedron is preserved, but the atomic relaxation results in an isotropic elongation of the four Cd─S bonds, indicating that the [Ni/ZnS4] tetrahedron experiences a local inward relaxation with shorter Ni─S and Zn─S bond length (~2.33 Å) than the original Cd─S bond (~2.53 Å). It implies the effect of the atomically dispersed Ni on the volume shrinkage of the supercell along with increasing Ni content. The SAED ring patterns are consistent with XRD results (fig. S4, T to W). Together with the ICP-MS and EDS measurements, it elucidates the solid solution feature of the as-prepared NixZCS QDs, which could benefit the electron conductivity for their catalytic applications.

Depth profiling of the crystal structure and electronic structure was carried out using x-ray photoelectron spectroscopy (XPS). The atomic ratios of Ni/Zn/Cd on the surface are 3.13%:28.18%:68.69%, 9.50%:26.71%:63.79%, 11.75%:23.23%:65.02%, and 12.65%:21.53%:65.82% for NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, respectively, as measured by XPS. The metal versus S ratio was confirmed to be around 62%:38%, indicating the loss of the S on the surface. The dangling S on the (111) crystal plane of zincblende presents a much higher activity, implying their instability [~2 eV higher surface energy of (111) facets with dangling S than the one without dangling S]. The high-resolution XPS results of Zn 2p and Cd 3d agree well with the values reported for the divalent Zn and Cd in pure metal sulfides (fig. S5, A and B) (21, 22). Two subpeaks in S 2p 3/2 XPS spectra (Fig. 2A) can be ascribed to the lattice S2− (at ~161.1 eV) and the oxidized states of S (at ~163.2 eV), respectively (23). The gradually decreased subpeak at ~162.9 eV indicates the lower oxidization of surface S along with increased concentration of Ni in NixZCS QDs. This is consistent with the geometry evolution of the NixZCS, because more (110) facets increase the Bader charge of surface S [−0.92 to −0.96 V for (110) facets and −0.85 to −0.91 V for (111) facets, table S3]. The Ni 2p XPS profile shows the main peaks of Ni 2p3/2 and Ni 2p1/2 at ~855.3 and ~872.6 eV, respectively (Fig. 2B) (24). The former main peak can be ascribed to the direct charge transfer from the nearest-neighbor S 2p to Ni core hole (25, 26). It can be fitted into a doublet (~855.6 and ~856.9 eV, peaks 1 and 2) with the increased Ni concentration, which is related to the two different photoemission channels: e.g., a single Ni site screened by a shift of electron density from the surrounding nearest-neighbor S 2p (peak 1) and a nonlocal effect from an adjacent cation occupancy (peak 2) (27). Here, peak 2 can only be ascribed to the adjacent Ni cation occupancy, which is evidenced by the local density of state (LDOS) of the Ni, Zn, Cd, and S in the NixZCS (fig. S5, C to F). The d10 metals Cd and Zn show a tight and localized DOS at the bottom edge of the valence band (VB) in NixZCS crystals and do not participate in bonding. Only Ni 3d levels occupied the top of the VB and bottom of the conduction band (CB), together with substantial S 2p charge-transfer character to form the bandgap. It indicates that Ni 2p3/2 peak 2 can only be derived from the adjacent Ni nonlocal effect. Thus, owing to the low Ni concentration within Ni0.015625ZCS and Ni0.03125ZCS QDs, the adjacent Ni nonlocal screening is eliminated, resulting in the broader peak 2 within Ni 2p3/2 peaks, which is hard to distinguish, compared with the ones in the higher Ni doping levels samples. Furthermore, the 2p53d8 state associates with the satellite feature, leading to the broad peak at ~863 eV, which agrees with the single metal site character even for very dilute Ni (28). Such broad satellite peak also decreased in NixZCS QDs within less Ni because of the less Ni d-d nonlocal screening transition contributions (25, 26), confirming the low-level Ni distribution within the Ni0.015625ZCS and Ni0.03125ZCS QDs. As revealed by the XPS, the Ni atoms are highly uniformly distributed within the NixZCS QDs host crystals, which can benefit the usage of Ni active sites.

Fig. 2 Physicochemical characterization of the NixZCS QDs.

High-resolution S 2p spectra (A) and Ni 2p spectra (B) of NixZCS QDs (x = 0.015625, 0.03125, 0.0625, and 0.125). The XPS spectra are normalized to the Cd intensity. a.u., arbitrary units. (C) The EELS spectra of Ni in the NixZCS (x = 0.125 0.03125, 0.0625, and 0.125) QDs. (D) Schematic diagraph of (111) and (110) facet crystals. (E) The Ni K-edge XANES spectra of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, where peak A represents 1s → 3d transition and peak B represents 1s → 4p transition (the inset shows the expanded pre-edge region). (F) Fourier transform [k3 χ(k)] of the phase-uncorrected extended x-ray absorption fine structure spectra (EXAFS) of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs. (G) Four-shell fitting of Fourier transformations of EXAFS spectra for Ni0.03125ZCS. EXAFS spectra were fitted using the FEFF 8.0 code [the inset shows the structure of the Ni site in DFT optimized (111) and (110) facet crystals; the balls in green, gray, purple, and yellow represent Ni, Zn, Cd, and S atoms, respectively]. (H to M) Wavelet transform plots of Ni foil, NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, and NiS. (N) Schematic diagraph of the crystal structures of the as-prepared NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs.

To further probe the local hybridization states of Ni in the NixZCS QDs on high spatial resolution down to the nanometer level, the L-edge EELS was collected to measure the Ni dipole transitions from 2p orbitals to unoccupied 3d orbitals. The L-edge EELS was recorded to evaluate the local electron distributions and magnetic moments of Ni (Fig. 2C) (29, 30). The L3/L2 white-line ratio of Ni is increased from 1.124 for Ni0.015625ZCS to 1.21 for Ni0.03125ZCS, 1.420 for Ni0.0625ZCS, and 1.573 for Ni0.125ZCS QDs, indicating the gradually lowered occupied Ni d orbit and more unpaired electrons (enhanced magnetic moments) within Ni (29, 30). We studied the anisotropy crystal facets of the orthorhombic nonpolar (110) surface and hexagonal polar (111) surfaces of the NixZCS to reveal the local electronic structure of Ni in NixCZS QDs. As confirmed by both surface energy and experimental M:S molar ratio results (XPS and EDS, table S1), the metal cations on the polar (111) surface coordinate with three S rather than four. The relaxed top MS (M = Ni, Zn, and Cd) layers of both (111) and (110) crystal planes appear corrugated with respect to the bulk termination, and all cations in the outermost layer tend to downshift toward the inner layer (S tends to upshift) (Fig. 2D and fig. S5, G and H). The Ni and Zn ions become almost coplanar with their three coordinated S atoms (Cd even further downshifts to form the invagination pyramid), and shorter M─S bond lengths (~2.25 Å for both Zn─S and Ni─S and ~2.48 for Cd─S) were achieved. The (111) surfaces demonstrated a flatter feature (almost parallel to the ab plane) than the (110) ones (~137° to the ab plane). Being consistent with the Bader charge [0.95 to 0.98 for (110) and 0.54 to 0.58 for (111) facets] and the unpaired electron number of Ni [1.61 to 1.69 for (110) and 0.95 to 0.99 for (111) facets], the magnetic moments of Ni atoms were calculated to be 1.61 to 1.69 μB for (110) facets and 0.95 to 0.99 μB for (111) facets (table S4). These results unveil that the Ni in the (111) facets shows one unpaired electron with a 3d9, S = 1/2 electronic configuration and monovalent oxidation state, whereas the Ni in (110) facets features two unpaired electrons and a 3d8, S = 2/2 electronic configuration. Therefore, as characterized by the Ni EELS spectra, the higher L3/L2 white-line ratio derived from more Ni dipoles transitioning from 2p to 3d orbits can be ascribed to the more unoccupied Ni 3d orbits on the (110) facets. Meanwhile, as unveiled by the EELS, the unpaired electrons on the (110) and (111) facets can contribute to more interaction with the H.

The localized coordination environments of induced Ni within NixZCS QDs crystals were confirmed by the XAFS (Fig. 2, E to M). All four K-edge Ni x-ray absorption near-edge structure (XANES) spectra present the prepeak at ~8348 eV (peak A in Fig. 2E), suggesting the dipole-forbidden but quadrupole-allowed transition and the 3d empty metal states and 4p orbit hybridization of the Ni central atoms (31). Furthermore, along with the increase of Ni in NixZCS QDs, the pre-edge peak intensity increased (inset of Fig. 2E), indicating the more empty p component in p-d hybridized orbits. These are consistent with the EELS spectra that presented more (110) facets, resulting in the less occupied p component. Figure S6 simulated the XANES of NiSx (x = 6, 4, and 3) complexes to study their symmetrical dependency. The shoulder near 8348 eV appears in the sixfold coordinated NiS6 octahedron (Oh), fourfold coordinated NiS4 tetrahedra (Td), threefold coordinated NiS3 triangular pyramid, and the trigonal planar complex (C3v), but not in the square planar complexes (C4v) (32). For NiS6 Oh symmetry, the eg (dz2 and dx2−y2) mainly contributes to the electric dipole and quadrupole transitions, whereas for the NiS4 Td, it mainly comes from the t2g (dxy, dyz, and dxz) (fig. S6, A to C). Losing one more axial to form the threefold coordinated NiS3 triangular pyramid, it presented a more intense pre-edge peak, mainly contributed by eg, and the intensity increased along with the Ni moving toward the planar center, reaching the maximum with the trigonal planar complex (C3v). This is consistent with the pre-edge peak intensity changing from Ni0.125ZCS to Ni0.015625ZCS QDs. These transitions can serve as a fingerprint for identifying the trigonal planar Ni-S3 (25, 29). After calculation of the two shells [NiS4M12 and NiS3M9 for (111) facets and NiS3M7 for (110) facets, fig. S6D], the calculated XANES of NiS4M12 presented the highest white line (1s → 4p dipole transfer) and NiS3M9 mainly contributed to the prepeak (1s → 3d transfer). The experimental XANES of NixZCS QDs shows higher white line intensity than Ni foil and NiS. The intensities were enhanced with increased Ni in NixZCS QDs, confirming the NixZCS QDs mixed with the threefold coordinated NiS3 and fourfold coordinated NiS4 tetrahedra (Td). Furthermore, the XANES edge positions of NixZCS QDs lie between those of Ni foil and NiS, indicating that their oxidization state is between 0 and +2 (inset in Fig. 2E). This also confirms the Ni mixed with the threefold coordinated NiS3 and Td symmetry as demonstrated in Fig. 2D. Figure 2F displays the Fourier transform (FT) k3 χ(k) of the phase-uncorrected extended x-ray absorption fine structure spectra (EXAFS). The FT k3 χ(k) of Ni foil displays peak at 2.15 Å, corresponding to Ni-Ni interaction and does not appear in any of the NixZCS QDs, confirming isolation of Ni in NixZCS QDs. The main peak at ~1.58 Å of NixZCS QDs is attributed to the scattering interaction between Ni and the first shell S (Ni-S). Figure 2G and fig. S6 (E to G) show the four-shell fitted FT k3 χ(k) EXAFS spectra of Ni0.03125ZCS QDs and NixZCS (x = 0.015625, 0.0625, and 0.125) QDs, respectively. It is well matched with DFT-optimized (111) and (110) facet crystals (inset of Fig. 2G and table S5 summarize the fitting results). The wavelet transform plot of Ni foil shows the maximum at around 7 Å−1 (Fig. 2H), corresponding to Ni─Ni bonding, while for NiS, there are two wavelet transform plot maximal centers, related to Ni─S (5 Å−1) and Ni─Ni (7 Å−1) bonding, respectively (Fig. 2M). The maximum wavelet transforms of NixZCS QDs are all around 5 Å−1 (Fig. 2, I to L), aligning with Ni─S bonding only. All these XAFS results elucidate that the Ni coordinates with the threefold S on (110) and (111) surfaces and fourfold S within the Td symmetry, suggesting the different localized electron structures on the surface, profiting the tunable catalytic property, as illustrated in Fig. 2N. Furthermore, DFT calculations on surface energy and the adsorption energies of OLA on the crystal planes (fig. S3, I and J) were applied for disclosing the morphology evolution on NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, as shown in fig. S3 (E to L).

The sub-angström-resolution aberration-corrected HAADF-STEM and corresponding STEM-EDS spectrum were conducted to observe the highly uniformly distributed Ni atom within the NixZCS host crystals. The clean surfaces and edges of the NixZCS QDs can be observed by the HAADF-STEM (Fig. 3, A to D). The STEM-EDS elemental mapping of the Ni0.03125ZCS QDs readily reveals the homogeneous distribution of Ni, Zn, Cd, and S in the integrated QDs (Fig. 3, E to H). HAADF images represent the projection of atoms along the incident beam direction. Therefore, the sub-angström-resolution aberration-corrected HAADF-STEM images of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs along the [111] zone axis with different intensities for atomic columns (Fig. 3, I to L) were used to characterize the dispersion and configuration of the Ni/Zn and Cd atoms on the basis of Z-difference of the individual heavy atoms (33). The individual Ni atoms occupy exactly the positions of the Cd and/or Zn atoms within the surface of NixZCS QDs without the interstitial sites, which is comparable with the simulated Ni-substituted scanning tunneling microscope (STM) image of the (111) surface of Ni0.03125ZCS (Fig. 3M). The line profiles (Fig. 3N) of the (111) surface of Ni0.03125ZCS marked in Fig. 3J, showing different intensities, could identify the different elements (higher intensity represents Cd-rich atomic columns and lower intensity denotes Zn/Ni-rich atomic columns). The measured distances between two adjacent atoms (Fig. 3N) can be fitted with the relaxed NixZCS crystals (fig. S6, H to J). The atomically resolved HAADF-STEM confirms the uniformly distributed Ni, Zn, and Cd in the as-prepared NixZCS QDs, which can profit the catalytic properties.

Fig. 3 HAADF-STEM and EDS spectrum.

(A to D) Low-magnification HAADF-STEM images of Ni0.015625ZCS QDs (A), Ni0.03125ZCS QDs (B), Ni0.0625ZCS QDs (C), and Ni0.125ZCS QDs (D). (E to H) The EELS mapping of Ni0.03125ZCS QDs. (E) S, (F) Ni, (G) Zn, and (H) Cd. (I to L) Sub-angström-resolution aberration-corrected HAADF-STEM images of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs along the [111] zone axis with different intensities for atomic columns: higher intensity for Cd-rich atomic columns and lower intensity for Zn/Ni-rich atomic columns; (I) Ni0.015625ZCS QDs, (J) Ni0.03125ZCS QDs, (K) Ni0.0625ZCS QDs, and (L) Ni0.125ZCS QDs. (M) Simulated scanning tunneling microscope (STM) image of the (111) surface of Ni0.03125ZCS. The inset shows the corresponding atom arrangement along the [111] orientation. The yellow, dark pink, gray, and light gray colors represent the S, Cd, Zn, and Ni atoms, respectively. (N) Line profiles show the variation in intensities along the directions marked in (J).

Photocatalytic performances of H2 production

The activity in the visible-light–powered photocatalytic H2 evolution of NixZCS QDs was evaluated (Fig. 4A). As comparison samples, the bare CdS and ZnS QDs have negligible visible-light photocatalytic activity with very limited H2 production rates of 242 and 207 μmol hour−1 g−1, respectively. The ZCS QDs increase H2 production rate to 1.50 mmol hour−1 g−1. The NixZCS QDs show a remarkable enhancement of the H2 production rates. In particular, Ni0.015625ZCS QDs achieved a H2 production rate of 13.74 mmol hour−1 g−1. With increasing Ni content, Ni0.03125ZCS QDs present the optimal photocatalytic properties and obtained the highest H2 production rate of 18.87 mmol hour−1 g−1, exceeding that of bare CdS QDs by a factor of 78- and 2.7-fold superior activity over the 1.9 weight % (wt %) Pt-loaded CdS QDs (6.88 mmol hour−1 g−1). The stability of the optimized Ni0.03125ZCS QDs was further tested by cycling photocatalytic experiments (Fig. 4B). No substantial deterioration of photocatalytic activity was observed during cycling. A further increase in the Ni content led to deterioration of photocatalytic performance. Ni0.0625ZCS and Ni0.125ZCS QDs delivered a H2 production rate of 12.11 and 6.97 mmol hour−1 g−1, respectively. The stability of the as-prepared Ni0.03125ZCS QDs was further confirmed by the postcharacterization via the STEM, EXAFS, and XPS after the cycling test (fig. S7). It can be seen that after the cycling test, the Ni0.03125ZCS QDs still preserve the truncated tetrahedra shape with the well-defined crystalline feature (fig. S7, A to C). The QDs also mainly exposed with the (111) facets as verified by their corresponding FFT pattern analyses (fig. S7, D to F, taken from marked areas in fig. S7C). As confirmed by the high-resolution Ni 2p spectra of Ni0.03125ZCS QDs after the cycling test (fig. S7G), the sharp Ni 2p3/2 peak and the broad satellite peak at ~863 eV associated with the 2p53d8 state agree with the single metal site character (28). Furthermore, the Ni K-edge XANES spectra and the corresponding EXAFS FT [k3 χ(k)] spectra of Ni0.03125ZCS QDs (fig. S7, H and I) after the cycling test are comparable with those of the fresh sample, identifying the Ni state as well as the surface composition as stable under photocatalytic conditions.

Fig. 4 Photocatalytic performance and spectroscopy/(photo)electrochemical characterization.

(A) A comparison of the photocatalytic H2 production activities of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, Pt-CdS QDs, CdS QDs, ZCS QDs, and ZnS QDs. The error bars denote SD. (B) Time course of photocatalytic H2 production over Ni0.03125ZCS; every 4 hours, the reaction system was purged with Ar for 30 min to remove H2. (C) Ultraviolet-visible (UV-vis) diffuse reflectance spectra of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, CdS QDs, ZCS QDs, and ZnS QDs. (D) The plots of transformed Kubelka-Munk function of the NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, CdS QDs, ZCS QDs, and ZnS QDs. (E) Electronic band structures of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, Pt-CdS QDs, CdS QDs, ZCS QDs, and ZnS QDs. (F) EIS Nyquist plots of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs, CdS QDs, ZCS QDs, and ZnS QDs electrodes measured under the open-circuit potential of electrodes with visible-light irradiation. (G to J) Unfolded band structure of the supercell for bulk NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) crystals, demonstrating the direct bandgap. The Fermi level moves to 0, as indicated by the dashed lines.

The light-harvesting capability, charge separation and transfer ability, and surface redox reactions were studied to understand the intrinsic photocatalytic property of NixZCS QDs. Ultraviolet-visible (UV-vis) diffuse reflectance spectra show that the ZCS QDs have an absorption edge (~465 nm) between those of bare ZnS and CdS QDs (Fig. 4C). After Ni dispersion, a continuous red shift of the absorption edges can be observed [~469, ~476, ~500, and ~528 nm for NixZCS QDs (x = 0.015625, 0.03125, 0.0625, and 0.125), respectively, Fig. 4C]. The bandgap of ZCS QDs was measured to be 2.75 eV, while NixZCS QDs have gradually decreasing bandgaps of 2.72, 2.67, 2.56, and 2.42 eV, respectively (Fig. 4D). The calculated band structures of bulk CdS, ZnS, and ZCS are shown in fig. S8 (D to F), with bandgaps of 2.41, 3.53, and 2.71 eV, respectively, which are well consistent with the experimental results (Fig. 4D) and the previous reported results (1113). They are larger than the values calculated based on the generalized gradient approximation (GGA) exchange-correlation function, due to the underestimation by the GGA function (34). To gain insights into the charge separation and transfer mechanism induced by the Ni dispersion, the unfolded band structures were calculated (Fig. 4, G to J). For the bulk NixZCS crystals, both the CB minimum (CBM) and VB maximum (VBM) are found to be located at the Г point of the Brillouin zone (BZ; fig. S8, A to C) with the direct bandgap transition. Along with the Ni doping level increase, the CB edge was reduced gradually. This indicates that Ni doping induced considerable energetic elevation of the CBM, which could contribute to narrowing the overall energy gap in the doped materials (2.49, 2.43, 2.38, and 2.31 eV for Ni0.015625ZCS, Ni0.03125ZCS, Ni0.0625ZCS, and Ni0.125ZCS, respectively), which is consistent with the Kubelka-Munk (K-M) results and the UV-vis spectra (NixZCS QDs have gradually decreasing bandgaps of 2.72, 2.67, 2.56, and 2.42 eV, respectively, Fig. 4D). The narrowed bandgap with Ni dopant has been reported by Thambidurai et al. (35), Zhang et al. (12), and Wu et al. (36). The strong Ni spin polarizations in NixZCS become more notable with increased Ni concentration. Moreover, besides the Ni spin-down orbit-introduced energy level between VBM and CBM, several flat bands related to the electron localization around the Ni and Zn region can be found, accompanied by breaking the branch of the bonds (as circled in Fig. 4, G to J). The charge densities of CBM and VBM states (spin up and down) at Г indicate that the dominant orbit components are Ni t2g (3dxy, 3dxz, or 3dyz) and S 2pz for CBM, while S 2px, 2py, or 2pz together with a minority of Ni t2g contributed to the VBM (fig. S8, D to V). For the (110) facets, CBM is also ascribed to the Ni spin-down 3d t2g (dxy, dxz, or dyz) state with slightly decreased bandgap (Eg), whereas the (111) facets demonstrated the increased bandgap along with the increased Ni in NixZCS host crystals (fig. S8, w1 to w4 and x1 to x4). The Fermi level (Ef) and work function (Φ) are Ef = −2.99, −2.94, −2.88, and − 2.83 eV and Φ = 5.96, 5.94, 5. 92, and 5.8 versus vacuum level for the (110) facets of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125), respectively. The values are −3.40, −3.33, −3.37, and − 3.62 eV (Ef) and 4.6, 4.48, 4.56, and 4.68 (Φ) for the (111) facets, respectively. Accordingly, Fig. 5A shows the band edge (VBM and CBM) alignment of the (110) and (111) facets of NixZCS, demonstrating the typical type II (staggered) heterojunctions. Electrons in the (111) crystal planes with the higher Fermi levels tend to flow to the (110) crystal planes to flatten the Fermi levels. Therefore, the (110) facets here had considerable excess negative charges, whereas the (111) facets were electropositive to some extent. The adjustment of Fermi levels brought in the rearrangement of overall band structures (Fig. 5B), in which, to accommodate the flattened Fermi levels, both CBM and VBM of (110) parallelly shifted up, whereas CBM and VBM of (111) synchronously descended. Such rearranged band structures are consistent with the experimental Mott-Schottky plots and VB spectra (Fig. 4E and fig. S9, A to D). The CBMs of NixZCS QDs were measured to be at −0.45, −0.98, −0.51, and −0.31 V versus the standard hydrogen electrode (SHE) for NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125), respectively (Fig. 4E). The flattening of Fermi levels and electron flow could generate the BIEF between the (110) and (111) facets, forming a charge disequilibrium in the depth direction perpendicular to the interface (Fig. 5B). Furthermore, when evaluating the energy levels of the (110) and (111) facets of the same NixZCS (Fig. 5B), the difference in energy of the VB and CB between the (110) and (111) facets can be distinct. It suggests that the electron transfer from the (110) and (111) facets is feasible thermodynamically, leading to the accumulation of electrons on the (111) facets. Such spatial charge separation between the (110) and (111) facets is feasible due to their different energy levels, which is responsible for the formation of reduction and oxidation facets. Meanwhile, the opposite BIEF direction toward the electron flow markedly accelerates the migration of photoexcited electrons from the (110) to the (111) facets, whereas the immobilized positive holes near the (110)/(111) interface remain on the (110) facets because of the high repulsive force from the BIEF. It is worth noting that such synergetic effects of various coexposed facets in one crystal and balanced ratio of different exposed facets is crucial for the photocatalytic efficiency. The band bending of each facet of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) and the BIEF are distinct owing to their different surface electronic structures, which is another factor affecting the spatial charge separation. The downward bent (110) VB and the upward bent (111) VB generate discontinuities at their interface, forming a barrier for holes to transfer from the (110) to the (111) facets, in spite of which the location of the (111) VBM is more favorable for hole migration. Thus, because of such favorable band structure and BIEF, photoexcited electrons could smoothly transfer from (110) to (111), whereas the holes stay within their respective facets. Such separated pathways could drastically increase the photoexcited charge carrier separation efficiency (37) as demonstrated by Fig. 5C. Moreover, owing to the different barrier height, the large offsets of VB and CB between the (110) and the (111) facets could enhance the driving force for the separation of photoexcited electrons and holes (38). Ni0.03125ZCS presented the largest ∆ECBM (~0.54 eV), which is consistent with the most negative CBM as measured (Fig. 4E), indicating its strongest charge separation and hydrogen reduction capability. The ∆ECBM decreased to 0.35 eV for Ni0.0625ZCS. For Ni0.125ZCS, owing to the increased Eg of the (111) facets, both the CB and VB of the (110)/(111) interface generate discontinuities, impeding the transfer of both photoexcited electrons from (111) to (110) facets and holes from (110) to (111) facets. Such limited surface heterojunction cannot drive the charge pair separation. The charge carrier separation efficiency was validated by photoluminescence (PL) spectra (fig. S9E) and transient photocurrent (TPC) response. The strong recombination of electrons at the CBM and the holes at VBM produced a broad peak centered at ~533 nm. Ni0.03125ZCS QDs exhibit the lowest PL peak intensity, implying the most efficient charge carriers separation and transfer. The PL intensity of other QDs follow the sequence Ni0.015625ZCS QDs < Ni0.0625ZCS QDs < Ni0.125ZCS QDs < ZCS QDs < CdS QDs < ZnS QDs, suggesting a suppressed photoexcited charge carrier recombination. The TPC densities show the same trend (fig. S9f). The photocurrent response for Ni0.03125ZCS QDs gives the highest current density under intermittent visible-light illumination (fig. S9g), followed by Ni0.015625ZCS QDs, Ni0.0625ZCS QDs, Ni0.125ZCS QDs, ZCS, CdS QDs, and ZnS QDs. The surface charge-transfer efficiencies were measured by electrochemical impedance spectroscopy (EIS) (Fig. 4F). The semicircle diameters follow the sequence Ni0.03125ZCS QDs > Ni0.015625ZCS QDs > Ni0.0625ZCS QDs > Ni0.125ZCS QDs, also indicating much lower interfacial charge-transfer resistance of Ni0.03125ZCS QDs (table S6). All these results confirmed that doped ZCS QDs, especially the Ni0.03125ZCS QDs, show good interfacial charge carrier transfer and superior charge carrier separation ability.

Fig. 5 DFT calculation studies.

(A) The calculated electronic band structures of (110) and (111) facets of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125), CdS, and ZCS. (B) The band structure layouts after accommodation of Fermi levels. (C) Schematic diagram of the BIEF at the interface formed by (110) and (111) facets, contributing a surface heterojunction to reduce the electron-hole recombination. (D and E) Free energy of the hydrogen (H*) adsorption on different Ni, Zn, Cd, and S sites on the surface of (111) (D) and (110) (E) crystal planes of the NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125), CdS, and ZCS.

The LDOS shows that the much tighter and sharper 3d t2g and eg orbits of Ni in (110) facets have greater localization state than the ones in (111) facets (fig. S10, A to H). More Ni 3d electrons in (111) facets, especially the spin-down electrons, are involved in the hybridization with the S s and p orbits owing to their flatter surface feature, leading to the shift of weight in intensity from the CB to the VBM. Thus, it results in higher charge density and partial charge of Ni in (111) than in (110) at the VBM (fig. S10, I to L). The unpaired spin-up electron of monovalent Ni(I) in (111) facets was identified to occupy the high–energy level 3dx2y2 and 3dz2 orbits, which could provide high reactivity toward the catalytic reactions (9), whereas the two unpaired electrons of Ni in (110) facets mainly occupy the 3dxy and 3dxz orbits. Moreover, the Ni d-bond center of the (111) facets is much closer to the Fermi level (fig. S10, M and N), exhibiting a strong hydrogen adsorption ability via the higher antibonding states together with a proton from the electrolyte (39). The H*, HO*, and O* species adsorption on different Ni, Zn, Cd, and S sites of the cleavage (110) and (111) surface of NixZCS was studied to verify their HER activity. The corresponding free energy change ∆GH*, ∆GOH*, and ∆GO* are shown in Fig. 5 (D and E) and fig. S10. For both (111) and (110) surfaces, the anion sites (S), supplying electrons, are favored for the H* adsorption (showing the low |∆GH*|, Fig. 5, D and E). S sites on the (111) surface show the more favorable activities for the H* adsorption than the ones on the (110) surface. As shown by the charge density difference figures (fig. S11, A to H), the S on the (111) surface coplanarly coordinate with Ni, Zn, and Cd ions, making the H* perpendicularly absorbed. This benefits the electron transfer from S to H* with appropriate bonding energy. Meanwhile, M metals from the lower layer contribute electrons to avoid too strong bonding energy between S and H*. As compared, H* forms the slanting angle toward S sites on the (110) surfaces (fig. S11, I to P), showing higher free energy than that on the (111) surfaces (Fig. 5, D and E). It is ascribed to the extra charge contribution from M sites bonding with S, owing to their shorter distance.

These results indicate that the (111) surface is more favorable for the HER than the (110) surface. Moreover, because of the BIEF established by the surface heterojunction and rearranged band levels between the (110)/(111) interface, photoexcited electrons transfer from (110) to (111), further enhancing the catalytic activity of the (111) surface for HER. Thus, with less (111) facets, Ni0.0625ZCS and Ni0.125ZCS have a decreased HER ability. Owing to the occurrence of a single eg lone electron of the monovalent Ni(I) d9 structure on the (111) surfaces, Ni sites on (111) surfaces of Ni0.015625ZCS, Ni0.03125ZCS, and Ni0.0625ZCS also show the low |∆GH*| values (0.30, 0.11, and 0.04 eV, respectively, Fig. 5D). Nevertheless, very strong adsorption energy bonds H* tightly on the Ni site in the Ni0.125ZCS (111) surface, contrary to HER properties (fig. S11, Q, S, U, and W). Because of more unpaired electrons on Ni d8 of the (110) surface, both Ni and S sites on the (110) surface contributed large charge density to H* (fig. S11, R, T, V, and X). As shown in fig. S11 (y1 and y2), HO* prefers to be adsorbed on cation sites for both (110) and (111) surfaces. In particular, the Ni0.03125ZCS demonstrated the low |∆GOH*| on all M sites of (111) and (110) surfaces [~0.36, ~0.42, and ~0.40 eV for Ni, Zn, and Cd on the (111) surface and 0.75, ~0.43, and 0.80 eV for Ni, Zn, and Cd on the (110) surface, respectively]. The (110) surfaces show that the lowest |∆GOH*| on S sites coordinated with Cd, Zn, and Ni, which is beneficial for oxygen evolution. O* also prefers the anion and Ni sites (fig. S11, z1 and z2), which can be ascribed to the low magnetic moment of O* on these sites because of the paired occurrence of lone electrons between O* and S and Ni sites in the geometry. Among NixZCS QDs, Ni0.0325ZCS QDs show the lowest |∆GO*| on the active sites of both (110) facets, indicating its OER activity. Obviously, the BIEF established by surface heterojunction and rearranged band levels in the same phase greatly facilitate the charge carriers’ separation and transfer in the reverse direction, which enhances the (111) and (110) surfaces for HER and OER, respectively.

Electronic structures

The bonding nature of H2O on the Ni, Zn, and Cd species of Ni0.03125ZCS (110) and (111) facets was further investigated by the projected densities of states (PDOS) (Fig. 6 and fig. S12). For both (110) and (111) facets, the energy levels of Ni d orbits and H2O 1b1 orbits are matched, leading to partial occupation of the formed d-1b1 orbits (Fig. 6, A and C). The fragment orbit analysis provides more details of the interaction between Ni and H2O (Fig. 6, B and D). It shows that the spin-up Ni dxz orbit on the (110) facet (Fig. 6B) and the spin-down Ni dz2 orbit on the (111) facet (Fig. 6D) can provide large exchange stabilization energy for the 1b1 orbits of H2O, forcing *H2O to be of radical nature for hydrogenation. The partial 1b1 orbit of H2O was enhanced above the Fermi level, leading to the antibonding orbit and lowering the O─H bonding energy level, which further weakens the H─O bond. Therefore, the more interaction and lower occupation of the 1b1 orbit in H2O show a lower H─O bond order, which is responsible for the lower dissociation barrier than that of *H2O. The crystal orbit overlap population (COOP) analyses (Fig. 6, E and F) also show that under the Fermi level, the Ni dz2 on the (111) facet presents more overlap population with both O and H to form the bonding feature. Meanwhile, H─O forms the antibonding feature, indicating the radical nature for hydrogenation. The dxz orbit of Ni on the (110) facet only generated the bonding feature with O. The H forms the antibonding with Ni dxz and H─O still bonds tightly. Obviously, the flatter crystal plane feature of the (111) facet more easily transfers electrons from the Ni dz2 orbit to the H2O 1b1 orbit, weakening the H─O bond, lowering the bond order in H2O, and facilitating the hydrogenation process. For both Zn and Cd, the d orbits match the 1b2 orbit of H2O with much lower energy level (fig. S12, A to D), which could not contribute a large amount of electron transfer from Zn and Cd to H2O (fig. S12, E, F, H, and I). The corresponding COOP analyses also revealed their limited bonding feature with H and O (fig. S12, G and J). Therefore, the Ni dispersion in NixZCS QDs plays a crucial role in HER.

Fig. 6 Electronic structure analysis.

Projected densities of states (PDOS) and schematic illustrations of 3d orbits of Ni on Ni0.03125ZCS (110) (A) and (111) (C), 1s, 2s, and 2p orbits of the H2O molecule, and their interaction within H2O-Ni configuration, Tot represents the sum of PDOS of each orbits. (B and D) PDOS and schematic illustrations for the H2O-Ni on Ni0.03125ZCS (110) (B) and (111) (D) case and charge density differences of H2O adsorption on Ni0.03125ZCS (110) and (111) (cyan stands for holes and yellow stands for electrons). (B) and (D) also demonstrated the major interactions and energy levels of the molecular orbits of H2O on Ni of (110) and (111) facets with correlation to the orbits from Ni and H2O fragments. (E and F) Crystal orbit overlap population (COOP) for H2O on Cd and Zn of Ni0.03125ZCS (110) (A) and (111) (B) models. The balls in green, gray, purple, yellow, red, and white colors represent the Ni, Zn, Cd, S, O, and H atoms, respectively.

The triethanolamine (TEOA) interaction with the Ni0.03125ZCS (110) and (111) facets was also studied to identify the H2 evolution from TEOA (fig. S13). As shown by the charge density difference plots of TEOA adsorption on Ni0.03125ZCS (110) and (111) (fig. S13, B and C), both the (110) and (111) facets demonstrate strong charge transfer between Ni, Cd, and Zn sites with different H sites of TEOA (HO, Hc1, and Hc2, which coordinated with O, C1, and C2 of TEOA, respectively, as shown in fig. S13A). Besides the Ni, Cd, and Zn sites, their coplanarly coordinated S sites on (111) facets or their adjacent and in the same plane S sites on (110) facets also transfer the charge to the TEOA. Such interactions can be verified by the PDOS of Ni, Cd, Zn, and S on Ni0.03125ZCS (110) and (111) facets together with different H sites of TEOA (HO, Hc1, and Hc2), as shown in fig. S13 (D, E, H, and I). For the bare TEOA, the Hc2 shows the PDOS near the Fermi level, followed by Hc1 and HO, respectively (fig. S13D), indicating their dehydrogenation ability from TEOA. For both (110) and (111) facets, the energy levels of Ni orbits and Hc1 and Hc2 orbits are matched, leading to partial occupation of the formed Ni-Hc2 and Ni-Hc1orbits (fig. S13, D and E). Their corresponding COOP plots (fig. S13, F and G) also show that under the Fermi level, the Ni on both (110) and (111) facets presents overlap population with Hc1 and Hc2 to form the bonding feature. In particular, the Ni on (110) facets shows more overlap population with Hc1 than the one on (111) facets, which could be ascribed to the two unpaired electrons of Ni on (110) facets, contributing to more interaction with the H than that on (111) facets. Relative to Cd, Zn on (110) facets demonstrates the bonding feature with Hc1 (fig. S13F), which could be ascribed to more matched energy levels of Zn orbits and Hc1 on (110) facets (fig. S13D), while both Zn and Cd on (111) facets could bond with Hc1 and Hc2 (fig. S13G) due to their lower PDOS energy levels, which matches the Hc2 and Hc1 energy levels (fig. S13E). Meanwhile, because of the strong charge transfer between TEOA and coplanarly coordinated S sites with metals on (111) facets or in the same plane adjacent S sites to metals on (110) facets (fig. S13, B and C, SNi, SCd, and SZn represent the S coordinated with or adjacent to Ni, Cd, and Zn, respectively), the S PDOS of (110) and (111) facets together with HO, Hc1, and Hc2 were also calculated (fig. S13, H and I). It can be seen that the PDOS of S are more overlapped with H from TEOA. Therefore, they form the bonding feature with H (fig. S13, J and K). In particular, the S adjacent to Ni on the (110) facets generates the strong bonding feature with HO, Hc1, and Hc2, while S adjacent to Cd and Zn bonds well with HO than with Hc1 and Hc2. For the coplanarly coordinated S sites on (111) facets, their PDOS are pushed down, showing more matched energy levels with Hc1. Therefore, it results in more bonding features between S and Hc1 as shown in their corresponding COOP analyses (fig. S13K). Obviously, the flatter crystal plane feature of the (111) facets more easily transfer electrons to the TEOA orbit, weakening the Hc1─C1 bond, lowering the bond order in TEOA, and facilitating the hydrogenation process. Furthermore, we also calculated the adsorption energy of TEOA on Ni0.03125ZCS (110) and (111) facets. The results show that the adsorption of TEOA on (111) facets is much stronger than that on (110) facets (with the adsorption energy of −0.6 eV versus −0.2 eV), and the O─H bonds within TEOA are elongated more evidently on the (111) surface (0.99 Å) compared to that on (110) (0.97 Å). Therefore, the photogenerated holes on (111) facets can be more efficiently consumed by TEOA than on (110). The sequential oxidation of TEOA on (111) facets, induced by photogenerated holes, could produce acetaldehyde and diethanolamine on the basis of the reaction NR3 → CH3CHO + HNR2, (R = C2H4OH, fig. S14A) (4042). The COOP between N with C2 of TEOA absorbed on (111) facets demonstrated a much weaker bonding feature compared with the bare TEOA and TEOA absorbed on (110) facets (fig. S14B). Oxidation of acetaldehyde and diethanolamine could produce hydrogen and contribute to the H2 evolution. Therefore, the COOP between H sites on acetaldehyde and diethanolamine and Ni, Cd, and Zn on (111) facets were further calculated. It can be seen that there is no readily bonding feature between H sites on acetaldehyde with Ni, Cd, and Zn (fig. S14C) and H sites on diethanolamine with Zn (fig. S14D), while HN (H coordinated with N of diethanolamine), Hc2, and Hc1 form the bonding feature with Cd and Ni (fig. S14, E and F). Here, the systematic evaluation of electronic structures of the NixZCS QDs actualizes the synergetic integration of photocatalytic constituents into a functional heterostructure to optimize the photocatalytic activity for HER.


In summary, we demonstrate a strategy to optimize the atomically dispersed Ni within the zincblende NixZCS QDs to maximize their efficient and durable photocatalytic performances for water splitting driven by sunlight. The fine-tuned atomically Ni-dispersed NixZCS QDs achieved an ultrahigh photocatalytic H2 production activity of 18.87 mmol hour−1 g−1. The combination of the experimental investigation and DFT calculations revealed the mechanisms of achieving such high photocatalytic performances. These include (i) the favorable surface engineering of the as-prepared Ni0.03125ZCS QDs with the highly active sites of monovalent Ni(I) on the (111) facets, (ii) the surface heterojunctions between the anisotropic (110)/(111) interface to reinforce the carrier separation owing to the BIEF, and (iii) appropriate surface H2 adsorption thermodynamics. This work demonstrates a synergistic regulation of the physicochemical properties of QDs at the atomic level toward high-efficiency photocatalytic H2 production. Therefore, the reported approach could provide an effective avenue for optimizing the accurate amount of the introduced heterogeneous elements into the catalysts to simultaneously boost the intrinsic activity, site density, electrical transport, and stability.


Synthesis of the NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs

The synthetic procedures for NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs are as follows: the OLA (≥98%, Sigma-Aldrich) was firstly degassed under vacuum at about 100°C for 30 min. After cooling to room temperature, the Ni, Zn, and Cd chlorides (98%, Sigma-Aldrich) and TAA (≥99%, Sigma-Aldrich) were stoichiometrically dispersed in 5 ml of OLA, respectively. Then, Ar was bubbled through the solution for 30 min and followed by heating to certain temperatures (for Ni, Zn, Cd, and S, the temperature are 200°, 230°, 160°, and 90°C, respectively) to form a homogeneous and clear solution. After that, the resulting solutions were injected into a three-necked flask and the mixture was flash-heated to 220°C and aged at that temperature for 30 min, resulting in a colloidal solution. Ethanol (≥99%, Sigma-Aldrich), methanol (≥99.9%, Sigma-Aldrich), and chloroform (≥99.5%, Sigma-Aldrich) (2:1:4 in volume ratio) were added to precipitate the QDs. The precipitate was retrieved by centrifugation and was then redispersed in cyclohexane and subjected to another three rounds of purification. The obtained powders were redispersed easily in nonpolar solvents such as hexane, toluene, and cyclohexane.

Structural and physical characterization

The crystal structures and phases of the NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs were examined by powder XRD (Bruker D8 Discover XRD) using Cu Kα radiation. The morphology was characterized by FESEM (Zeiss Supra 55VP), and the details of the crystal structure of the NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs were identified by TEM and HRTEM (JEM-2011, JEOL). The images were collected by a Gatan charge-coupled device (CCD) camera in a digital format. Atomic resolved HRTEM images and EDS mapping were obtained by HAADF-STEM (JEM-ARM200F, JEOL) at an accelerating voltage of 200 keV and the Oxford EDS system. The energy resolution at the zero-loss peak was 0.3 eV. The spectra were acquired from thin regions of ~20 to 100 nm. Care was taken to ensure that samples were not damaged by the electron beam by applying liquid-nitrogen cooling. All operation and postprocessing of the HAADF-STEM images were conducted by Digital Micrograph software. ICP-MS (Agilent 7500 Series) was used to quantify the elemental composition. EELS spectra were acquired with a Gatan 666 parallel EELS spectrometer attached to the JEM-ARM200F (JEOL) operated at an accelerating voltage of 200 keV. The energy resolution at the zero-loss peak was 0.3 eV. The spectra were acquired from thin regions of ~20 to 100 nm. Care was taken to ensure that samples were not damaged by the electron beam by liquid-nitrogen cooling. All operation and postprocessing of the EELS spectra were conducted by Digital Micrograph software. The white-line ratio on the EELS profile was calculated from the area under Ni L-edge each peak by eliminating the background intensity via a method suggested by Pearson et al. (43). XPS measurements were carried out on a Kratos XSAM-800 spectrometer with an Mg Ka radiation source. UV-vis diffuse reflectance spectra were collected for the dry pressed disk samples with a UV-vis spectrophotometer (LAMBDA 950 UV-vis/near infrared spectrometer, PerkinElmer) using BaSO4 as the reflectance standard at room temperature. Steady-state PL spectra were tested on a Varian Cary Eclipse fluorescence spectrometer at room temperature. The optical images were collected with a Canon EOS 660D camera. Here, the bandgaps of the samples were calculated by the K-M method on the basis of the UV-vis diffuse absorption spectra by the following equation (44)αhν=A(hνEg)1/2where α, hν, and Eg are the absorption coefficient, the photon energy, and the direct bandgap, respectively; A is a constant.

XAFS measurements

The XAFS spectra were collected at the 1W1B station of the Beijing Synchrotron Radiation Facility (operated at 2.5 GeV with a maximum current of 250 mA, Pt L3-edge). XAFS measurements at the Ni K-edge were performed in fluorescence mode using a Lytle detector. All samples were pelletized as disks of 13-mm diameter with 1-mm thickness using graphite powder as a binder. All spectra were collected in ambient conditions.

EXAFS analysis

The acquired EXAFS data were processed by postedge background subtraction from the overall absorption and normalizing with respect to the edge-jump step using the ATHENA module implemented in the IFEFFIT (45). Then, χ(k) data in the k-space ranging from 2.6 to 12.6 Å−1 were Fourier-transformed to real (R) space using Hanning windows (dk = 1.0 Å−1) to separate the EXAFS contributions from different coordination shells. The quantitative information can be obtained by the least-squares curve fitting in the R space with an FT k-space range of 3 to 11.26 Å−1. The backscattering amplitude F(k) and phase shift Φ(k) were calculated using FEFF8.0 code (46). First-principles scattering amplitudes and phase shifts for the photoelectron path of Ni-S and Ni-Cd(Zn) were calculated using Hartree-Fock “muffintin” potential accounting for the screened 1s hole of the absorbing Ni atom (47). To examine the effect of the second and third shells on the determined local structure of the absorbing atom, three series of fits were performed in R space including one, two, and three coordination shells of Ni.

Ni K-edge XANES spectra simulation within DFT framework

The projected p state density of electronic states (pDOS) provided by DFT calculation of NiSx (x = 6, 4, and 3) complexes were compared with KXANES spectra since the K-absorption cross section σ(ω) (ω is the energy of x-ray photons) reflects mainly the energy structure of the empty p states on the absorbing Ni atom according to the dipole selection rule. To validate the application of the DFT method to the structural modeling of NiSx (x = 6, 4, and 3) complexes, their corresponding molecular orbital energy levels were analyzed. The dependence of σ(ω) and pDOS is established according to the Fermi’s Golden rule in the dipole and one electron approximations by the expression: σ(ω) = C |〈Ψf|∇|Ψ1s〉|2ρ(ћω + E1s), where С is the constant coefficient, and Ψ1s and Ψf are the wave functions of the 1s state of Ni atom and the final р symmetrical state with energy Еf = ћω + E1s, respectively. ρ(Еf) is the pDOS at Ni atom, and ∇ is the operator of dipole transition. It can be seen that the fine structure of ρ(Еf) or KXANES is determined mainly by the energy dependence of pDOS on the absorbing atom in assuming of the smooth energy dependence of the matrix element.

Electrochemical measurement

The surface charge-transfer efficiencies were investigated by the EIS measurements, which were carried out on a CHI 660C electrochemistry workstation with a standard three-electrode system using the synthesized NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs coated on F-doped SnO2-coated glass (FTO glass) as the working electrodes, and a Pt wire and Ag/AgCl (saturated KCl) as the counter electrode and reference electrode, respectively. The working electrodes were synthesized as follows: 10 mg of the sample and 1 mg of naphthol (99%, as the binder, Sigma-Aldrich) were dispersed in 10 ml of ethanol. After sonification for several minutes to get the uniformly distributed precipitation solution, 1 ml of solution was dropped onto a 1 cm by 1.5 cm FTO glass electrode via a pipette. The obtained electrode was dried in an oven for 0.5 hours. All working electrodes studied were kept at a similar mass loading of 1 mg. For the TPC density measurement, Na2S and Na2SO3 within the electrolyte worked as hole scavengers rapidly capturing the photoinduced holes on the surface of QDs, which can eliminate the surface charge recombination of QDs. Therefore, the collected TPC density reflects the charge separation efficiency in the bulk of the QD samples. The polarization curves were recorded in the abovementioned three-electrode system and the bias sweep range was from −0.4 to −0.9 V versus SHE with a step size of 0.005 V. The Mott-Schottky plots were also measured by the same three-electrode system using the synthesized NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) QDs coated on FTO glass as the working electrodes without naphthol binder over an AC frequency of 1200 Hz in 0.5 M Na2SO4 aqueous solution. A 300-W xenon light with an UV-cutoff filter (λ ≥ 420 nm) was applied as the light source.

The experimental measurements of photocatalysis

The H2 evolution was characterized in a 50-ml Pyrex flask (sealed with silicone rubber septa) at ambient temperature and atmospheric pressure. Before the experiment, all glassware was rinsed carefully with deionized water. A Xe arc lamp (300 W) with a UV-cutoff filter (λ ≥ 420 nm) was applied as a light source to trigger the photocatalytic reaction. The focused intensity on the flask was ca. 80 mW cm−2. In the typical photocatalytic measurement process, 10 mg of photocatalyst was suspended with constant stirring in a mixed solution of TEOA and water (20 volume %, as a sacrificial reagent). Before irradiation, the suspension was sonicated and bubbled with argon gas for half an hour to remove residual and dissolved air and keep the reaction system under anaerobic conditions. During the irradiation process, 0.2 ml of gas was sampled intermittently through the septum, and the H2 content was analyzed by a gas chromatograph (Clarus 480, PerkinElmer, TCD with Ar as a carrier gas and a 5-Å molecular sieve column). For comparison, the CdS QDs, ZnS QDs, and ZCS QDs and the Pt-CdS QDs were also measured for H2 evolution. For loading the cocatalysts, 1 wt % Pt was deposited onto the surface of photocatalysts by the in situ photodeposition approach using H2PtCl6 (≥37.5%, Sigma-Aldrich) under Xe arc lamp (300 W) irradiation for 40 min.

Theoretical calculations

DFT calculations were performed using the Vienna Ab initio Simulation Package (48) on the basis of the GGA with the Perdew-Burke-Ernzerhof function as the exchange-correlation energy function. For the bulk crystals of NixZCS (x = 0.03125, 0.0625, and 0.125), 2 × 2 × 2 unit cells [relative to the face-centered cubic (fcc) zincblende CdS unit cell, 4 Cd and 4 S atoms, respectively] with 64 atoms (with 1, 2, and 4 Ni atom substitution, respectively) were used for the crystal structure and electronic structure calculations. For the Ni0.03125ZCS, 4 × 2 × 2 unit cells (relative to the fcc zincblende CdS unit cell, 4 Cd and 4 S atoms, respectively) with 128 atoms (with 1 Ni substitution) were used for the crystal structure and electronic structure calculations. We used the projector augmented wave potentials with 4s and 4p valence states for Cd, 3d, and 3s for Zn and Ni, and 3s and 3p for S with the cutoff energy of 500 eV. The conjugate gradient scheme is used to optimize the atom coordinates until the Hellmann-Feyman force is less than 0.01 eV Å−1. The number of k-points was carefully optimized to achieve energy convergence, giving 2 × 2 × 2 and 1 × 2 × 2 Monkhorst-Pack BZ calculations for NixZCS (x = 0.03125, 0.0625, and 0.125) and Ni0.015625ZCS, respectively. To address the on-site Coulomb energy interactions in the localized d electrons of the transition metal ions (Ni, Zn, and Cd), an additional Hubbard-type U term was taken into account with the GGA + U approach (49). The U for Ni was fitted to be 5.7 eV. Free energy of different substitutional, interstitial Ni dispersion species as well as the consideration of the S and Cd vacancies in the presence of Ni impurities were calculated on the basis of the energy differences between the Ni-dispersed crystals, for example, Esubstituted Ni on Cd site of ZCS = ENixZCSEZCSEn·Ni, where ENixZCS, EZCS, and En·Ni are the energies of the substituted Ni on Cd site of ZCS, ZCS, and Ni, respectively. Moreover, to identify the solubility of Ni in the ZCS solid solution, the formation energy of an impurity Ni is calculated on the basis of the following: Ef(NixZCS) = ENixZCSEZCS + μCd − μNi, where ENixZCS is the total energy of the supercell containing x Ni and EZCS is the total energy for the equivalent supercell containing only bulk ZCS. μCd and μNi are the chemical potentials of Cd and Ni, respectively, which satisfy the equilibrium constraint: μCd + μS ≤ μZCSbulk, μS ≤ μSbulk, and μCd ≤ μCdbulk, where μZCSbulk is the energy of bulk ZCS. Under S-rich growth conditions here, the μS is calculated as the energy of one S atom in bulk α-S in the orthorhombic structure, then μCd is obtained by μZCSbulk − μS. μNi is equal to the energy of one Ni atom in NiCl2, as it was used here as the Ni precursor.

Surface calculations were conducted via a slab model with periodic boundary conditions. Both the orthorhombic nonpolar (110) surface and hexagonal polar (111) surfaces of the NixZCS were cleaved from the relaxed NixZCS (x = 0.03125, 0.0625, and 0.125) bulk crystals and built with a vacuum of about 20 Å and a five-layer slab of which the three bottom layers were kept fixed during relaxation. The surface energies and the adsorption energies were performed on the basis of the stoichiometric supercell models with 4 × 4 × 1 and 2 × 4 × 1 k-point meshes for NixZCS (x = 0.03125, 0.0625, and 0.125) and Ni0.015625ZCS slabs, respectively. The surface energies were obtained by the slab methodEsurface=(EslabnEbulk)/2nAwhere A, Ebulk, and Eslab are the area of the supercell, the bulk energy per unit cell, and energy per surface unit cell of the specified slab model, respectively, and n is the number of bulk unit cells.

The OLA adsorption energy on (110) and (111) facets was calculated as followsEadsorption=(Eslab+OLAEslabEOLA)where Eslab+OLA is the energy of the adsorption complex including the relaxed surface and the OLA molecule, and EOLA is the energy of OLA.

Here, the band structures of the bulk and (110) and (111) facets of NixZCS (x = 0.015625, 0.03125, 0.0625, and 0.125) were calculated on the basis of the HSE06 hybrid functional (50) and the doped NixZCS was compared with ZCS via the band unfolding technique. It recovers an effective primary cell picture of ZCS band structure from calculations using different supercells, including both intrinsic and extrinsic perturbations. The band unfolding calculations are carried out using the BandUP code (51).

The ECBM and EVBM versus vacuum edge positions of different facets were calculated on the basis of the calculated band structures and the relation between the bandgap (Eg), work function (Φ), and Fermi level (Ef): Work function (Φ) = E(vacuum)Ef.

Reaction free energy

Three reactions were considered on (110) and (111) facets for evaluating their activities toward the HER and OER:

(i) H* adsorption at the cations (M = Ni, Zn, or Cd) or anion (S) sites (i.e., the Volmer reaction)H++e+*(M or S)H*(M or S)

(ii) H2O dissociation with HO* adsorbed at the cation or anion sitesH2O+*(M or S)HO*(M or S)+H++e

(iii) H2O dissociation with O* adsorbed at the cation or anion sitesH2O+*(M or S)O*(M or S)+2H++2ein which * represents an adsorption site on different facets of NixZCS QDs. Therefore, the final adsorption geometries in the above reactions were denoted as H*(M or S), HO*(M or S), and O*(M or S), respectively. The free energy for H* adsorption on each of the active sites of (110) and (111) facets [∆GH*(M or S)] was calculated according to the following equation with determination of zero-point energy (ZPE) (∆EZPE), entropy (TS), and the pH in solution [∆G(pH)] contributionsΔGH*(M or S)=ΔEH*(M or S)+ΔEZPETΔSΔG(pH)+eUwhere ∆EH*(M or S) is the adsorption energy of an H* at the M or S site, which can be calculated withΔEH*(M or S)=(EnH*Eslabn·EH2/2)/nin which the Eslab, EnH*, and EH2 are the energies of the slab before and after adsorbing with nH* and an H2 molecule, respectively. EH2 is the energy of hydrogen molecules in the gas phase. As the vibrational entropy of H* in the adsorbed state is small, the entropy of adsorption of 1/2H2 is the entropy of H2 in the gas phase under standard conditions (39). ∆G(pH) = −kTln 10 × pH, where T is the temperature and k is the Boltzmann constant. U is the potential equivalent to the ECBM, related to the effect of photoexcitation in photocatalysts. Non-Nernstian dependence of the sulfide band edges with a slope of 33 mV pH−1 has been considered (52). To determine the possible (110) and (111) surface terminations under the photocatalytic conditions, the free energy for reactions (ii) and (iii) was calculated on the basis of the following equationsΔGHO*(M or S)=ΔEHO*(M or S)+ΔEZPETΔS+ΔG(pH)eUΔGO*(M or S)=ΔEO*(M or S)+ΔEZPETΔS+2ΔG(pH)2eUwhere ∆EHO* and ∆EO* are the adsorption energies of HO* and O* on (110) and (111) surfaces, which were calculated as the following equationsΔEHO*(M or S)=(EnHO*+n·EH2/2Eslabn·EH2O)/nΔEO*(M or S)=(EnO*+n·EH2Eslabn·EH2O)/nEnHO* and EnO* are the energies of adsorption geometries HO*(M or S) and O*(M or S), respectively. EH2O is the energy of a H2O molecule.


Supplementary material for this article is available at

This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial license, which permits use, distribution, and reproduction in any medium, so long as the resultant use is not for commercial advantage and provided the original work is properly cited.


Acknowledgments: Funding: This original research was supported by the Australia Research Council, Commonwealth of Australia, the Australian Renewable Energy Agency (ARENA), University of Technology Sydney (UTS), through the Discovery Early Career Researcher Award (DECRA DE170101009), ARC Discovery Project (DP170100436), ARENA 2014/RND106, UTS Chancellor’s Postdoctoral Research Fellowship project (PRO16-1893), and UTS Early Career Researcher grant ECRGS PRO16-1304. We also acknowledge the use of the HAADF-STEM facility in the UOW Electron Microscopy Centre. Author contributions: D.W.S. conceived the research and carried out the synthesis, electrochemical tests, characterization, and the DFT calculations. J.R. performed the experimental measurements of photocatalysis. Z.W.Z. and C.C. conducted the x-ray absorption spectroscopy. Y.D.L., S.Z.Q., and G.X.W. discussed the overall research. D.W.S. conceived the project and drafted the manuscript. All authors discussed the experiments and the final manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data needed to evaluate the conclusions in the paper are present in the paper and/or the Supplementary Materials. Additional data related to this paper may be requested from the authors.

Stay Connected to Science Advances

Navigate This Article