Heterostructured perovskite nanocrystals for water stable plasmon-enhanced photoelectrocatalysis

As a result of globalisation and industrialisation, the concentration of carbon dioxide (CO₂) in the atmosphere has increased significantly in recent years¹, drawing attention to the urgent need for effective approaches to reduce CO₂ emissions. When produced through sustainable processes, hydrocarbons derived from CO₂ can serve as carbon-neutral fuels, enabling their seamless integration into the existing energy infrastructure, thereby reducing the reliance on non-renewable fossil fuels².

One important technique that has a lot of potential for solving this problem is catalysis. Metal oxides (TiO₂, Cu₂O, Co₃O₄, etc.)^3–5, 2D materials (MXene, graphene oxide, MoS₂, etc.)^6–8, and metal-organic frameworks (MOFs)^9–11 are among the many catalysts that have been studied for CO₂ conversion. Despite their effectiveness, traditional catalysts have significant disadvantages such as poor visible light-harvesting capability, insufficient charge potential for reactions, and low selectivity^10,12. These challenges highlight the need for innovative catalytic strategies to improve the effectiveness and scalability of the CO₂ reduction reactions.

In recent years, halide perovskite-based catalysts have attracted considerable interest because of their unique properties, including tunable band gap, high charge carrier mobility for efficient charge separation, strong light absorption, and high surface area for perovskites in the form of nanocrystals (NCs)^13–17. Owing to these properties, perovskites have demonstrated considerable potential in photocatalysis (PC), in which visible light is used to drive selective chemical reactions, particularly CO₂ reduction, for efficient conversion into valuable products such as methane and formic acid¹⁸. Moreover, they are particularly suitable for solar-driven catalysis because sunlight is a readily available renewable energy source.

For example, the incorporation of halide perovskites significantly enhances the catalytic properties of TiO₂. Specifically, hybrid TiO₂/CsPbBr₃ nanofibres have demonstrated superior performance in photocatalytic CO₂ reduction, achieving a higher production rate of 9.02 μmol·g⁻¹·h⁻¹ compared to 4.68 μmol·g⁻¹·h⁻¹ for pristine TiO₂ nanofibres, thereby confirming the enhanced catalytic activity¹⁸. In another study, a more complex heterostructured CsPbBr₃/Au/TiO₂ catalyst showed a 5.4-fold increase in performance in comparison to the CsPbBr₃/TiO₂ structure due to the formation of an ohmic contact, enabling fast charge transfer across the CsPbBr₃ and TiO₂ interface¹⁹. Au@CdS/CsPbBr₃ catalyst was also employed for the CO₂ photoreduction reaction, achieving a rate of approximately 3 μmol·g⁻¹·h⁻¹ with an apparent quantum yield of about 0.27%, where CdS served as a mediator layer between the CsPbBr₃ and gold (Au) plasmonic antennas, which generate hot electrons²⁰. Thus, the usage of lead-halide perovskites provides an improvement over traditional catalysts. However, these studies have some limitations, such as selectivity because the main products of PC are hydrogen (H₂), carbon monoxide (CO), and methane (CH₄). In this regard, several studies have demonstrated that the application of protective shell coatings is an effective strategy for enhancing the stability of halide perovskites in various applications^21–23. Nevertheless, achieving a long-term stable photocatalytic performance using perovskite-based nanomaterials remains a challenge.

In this study, we develop water-stable and highly selective photoelectrocatalytic materials for efficient CO₂ reduction reactions. For this purpose, we modify the hot-injection protocol for the synthesis of core-shell CsPbBr₃@TiO₂ heterostructures, which are stable not only in polar liquids but also in water. The CsPbBr₃@TiO₂ heterostructures allow us to achieve high photocatalytic activity for the conversion of CO₂ to CH₄. Subsequently, we show that the catalytic performance of CsPbBr₃@TiO₂ perovskite heterostructures can be enhanced by designing CsPbBr₃@TiO₂ core-shell nanoparticles (NPs) deposited on a gold nanoparticle (NP) layer, leveraging the phenomena of localised surface plasmon resonance (LSPR) and the generation of hot carriers. Thus, we achieve the conversion of CO₂ into valuable products, such as ethylene (C₂H₄) and propylene (C₃H₆), with high selectivity. Water is used as the reaction medium to facilitate the catalytic process, which significantly reduce the overall cost by eliminating the need for expensive supporting electrolytes and solvents. These findings indicate that perovskite-based nanomaterials are suitable for large-scale and sustainable CO₂ utilisation and conversion into fuel.

In this study, we present a novel approach to CO₂ conversion by combining the advantages of perovskite-based PEC catalysis in water media with LSPR enhancement. By addressing the problem of lead halide perovskite stability in water, we establish a foundation for the development of a robust and scalable photoelectrocatalytic system for efficient CO₂ reduction. In this study, we aimed to control the selectivity of the CO₂ reduction reaction using various processes (electrocatalytic, photocatalytic, and photoelectrocatalytic) in perovskite-based heterostructures. To achieve this, we synthesised CsPbBr₃@TiO₂ core NPs and investigated their structures and optical properties. We then prepared the catalysts by stepwise impregnation of copper foam with colloidal gold and CsPbBr₃@TiO₂ core-shell NPs.

For comprehensive structural analysis, we employed techniques such as transmission electron microscopy (TEM), scanning electron microscopy (SEM), X-ray photoelectron spectroscopy (XPS), and X-ray diffraction (XRD). Their optical properties were studied using absorption and photoluminescence (PL) spectroscopy. Time-resolved photoluminescence (TRPL), electron paramagnetic resonance (EPR), electrochemical impedance spectroscopy (EIS), and transient photocurrent (TPC) measurements were performed to analyse the charge separation in the CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au heterostructures. Gas chromatography was used to analyse the products of the CO₂ reduction reaction and the stability of the catalysts was studied using chronoamperometry. We present experimental data showing how we can control the selectivity of the CO₂ reduction reaction to form H₂, CO, CH₄, C₂H₄, or C₃H₆. To provide a deeper understanding of these processes, we discuss possible reaction pathways for CO₂ reduction based on the experimental results.

For the synthesis (Fig. 1) of the CsPbBr₃@TiO₂ core-shell NPs, we modified the standard hot-injection protocol²⁴. First, PbBr₂ was mixed with the 1-octadecene (ODE) and dried under vacuum. Subsequently, oleic acid (OA) and oleylamine (OAm) were added to the PbBr₂ suspension in the ODE until the solution became clear. Then, Cs-oleate was injected into the flask, and TBT and water, pre-mixed in ODE, were added stepwise. After 10 min, the reaction mixture was cooled, and the CsPbBr₃@TiO₂ core-shell NPs were purified.

Fig. 1  Scheme of modified hot-injection synthesis of CsPbBr₃@TiO₂ NPs.

The resulting CsPbBr₃@TiO₂ core-shell NPs are shown in the HRTEM images (Fig. 2a, c, d). The elemental composition, including Cs, Pb, Br, and Ti (Fig. 2b), was determined using EDX mapping. Interplanar distances corresponding to 4.2 Å (100) were defined using fast Fourier transform (FFT) analysis of the HRTEM images (Fig. 2c, d). Perovskite NCs are known to form after the injection of Cs-oleate during synthesis. In our modified protocol (Fig. 2), TBT was first added after Cs-oleate injection, followed by the addition of water to initiate the hydrolysis reaction, resulting in the coating of the particles. Based on this approach, we conclude that the CsPbBr₃ NCs were successfully covered with TiO₂ shells. The presence of the shell is supported by the contrast difference and amorphous halo observed in the HRTEM images (Fig. 2a, c, d) as well as by EDX mapping showing the presence of Ti (Fig. 2b). The XRD diffractograms of the CsPbBr₃ and CsPbBr₃@TiO₂ core-shell NPs exhibited strong diffraction peaks corresponding to the (100), (110), (111), (200), (210), (211), (220), and (300) planes of the cubic perovskite phase. The thin TiO₂ shell complicates definitive structural determination using HRTEM alone. In the XRD pattern obtained for the CsPbBr₃@TiO₂ system (Fig. 2e), additional weak reflections are marked with stars. While the peaks observed at ~25° and ~27° may superficially resemble those of the anatase and rutile phases, their assignment to crystalline TiO₂ is unlikely, considering the low-temperature, ambient conditions used during shell formation. Instead, these peaks precisely match the reference pattern of the Cs₄PbBr₆ phase (PDF-2 ICDD card No. 01-073-2478), which is a known secondary phase that may form through the partial degradation of CsPbBr₃ under the processing conditions.

Fig. 2  Structural characterization of CsPbBr₃@TiO₂ NPs. a TEM image of CsPbBr₃@TiO₂; b HAADF-STEM and STEM-EDX mapping of Br, Cs, Pb, Ti; c-d HRTEM images of individual CsPbBr₃@TiO₂ NPs with the corresponding fast Fourier transformation images; e XRD patterns of pristine CsPbBr₃ NCs and CsPbBr₃@TiO₂ NPs. Scale bars correspond to 50 nm (a,b); 10 nm (c,d); f XPS Depth profiling of Ti 2p region; g XPS depth profiling of Pb 4f region.

To further evaluate the presence and integrity of the TiO₂ shell, we conducted controlled Ar⁺-ion sputtering combined with high-resolution XPS measurements of the Ti 2p and Pb 4f core levels (Fig. 2f, g). The sputtering rate was adjusted to approximately 0.5 nm/min to enable gradual removal of surface layers without immediate disruption of the core. During the initial stages of sputtering, the Ti 2p spectra (Fig. 2f) exhibited progressive changes in the peak shape and the appearance of lower binding energy shoulders corresponding to Ti³⁺ and Ti²⁺ species, indicating partial reduction of the shell under ion bombardment. In contrast, the Pb 4f spectra (Fig. 2g) remained unchanged in shape and position, whereas their intensity increased with etching time, suggesting that the perovskite core was still chemically intact and protected by the TiO₂ layer. Subtle asymmetry in the Pb 4f line appeared only after an estimated etching depth of ~4 nm, and a distinct shoulder at ~137 eV, indicative of Pb⁰ formation, emerged beyond ~7 nm of sputtering. These results are in excellent agreement with the HRTEM data showing a TiO₂ shell thickness of 4–5 nm, and they confirm that the outer shell effectively prevents the immediate degradation of the perovskite during ion exposure. This experiment provides complementary evidence for the core–shell architecture and demonstrates the utility of differential XPS profiling as a non-trivial but informative diagnostic tool for heterostructured nanocomposites.

The survey spectra (Fig. 3g) identified the composition of freshly synthesised CsPbBr₃@TiO₂ core-shell NPs and CsPbBr₃@TiO₂/Au heterostructures, as well as control samples of CsPbBr₃ NСs and TiO₂ synthesised under identical conditions. High-resolution spectra of O1s, Ti2p, Cs3d, Pb4f, Br3d, and Au4f were acquired (Figure 3a-f) to refine the chemical state analysis. The high-resolution spectra of Cs3d5 (724 eV), Pb4f7 (138.1 eV), and Br3d5 (68 eV) for the control CsPbBr₃ NСs (Fig. 3a-c) revealed chemical states of Cs⁺, Pb²⁺, and Br⁻, consistent with reported data for bromide perovskites²⁵. The binding energies of O1s (530 eV) and Ti2p3 (458.53 eV) for the TiO₂ control sample (Fig. 4d, e) confirmed the oxidation states of O²⁻ and Ti⁴⁺, supporting the formation of titanium dioxide²⁶. Quantitative XPS analysis, presented in Table S1 indicates that the stoichiometric ratio for perovskite is Cs:Pb:Br at approximately 1:1:3, and that for TiO₂, Ti:O is 1:2. Comparisons of the high-resolution spectra (Fig. 4a-c) for the CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au structures with control samples revealed shifts and changes in the emission line shapes. The peak positions and FWHM values are detailed in Table S2.

Fig. 3  XPS spectra of TiO₂, CsPbBr₃ NCs, CsPbBr₃@TiO₂ core-shell NPs and CsPbBr₃@TiO₂ core-shell NPs deposited on gold substrate: a Cs 3d; b Pb 4f; c Br 3d; d O 1s; e Ti 2p; f Au 4f; g XPS survey spectrum of CsPbBr₃ NCs, TiO₂ control, CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au; h High-resolution valence band XPS spectra of TiO₂, CsPbBr₃ NCs as control samples, and CsPbBr₃@TiO₂; i The secondary electron cut-off in XPS spectra (applied -10 V bias) of TiO₂ control, Au control, CsPbBr₃ NCs.

Fig. 4  Band diagrams of a CsPbBr₃ and TiO₂ before contact, b CsPbBr₃@TiO₂ core-shell NPs and c CsPbBr₃@TiO₂ core-shell NPs deposited on Au NPs layer.

The reduced photoelectron signal from the perovskite in CsPbBr₃@TiO₂ was attributed to the formation of a TiO₂ shell, whereas for CsPbBr₃@TiO₂/Au, it was ascribed to the partial coverage of CsPbBr₃@TiO₂ with Au NPs (Fig. S1) shows the SEM image, size distribution, and UV-Vis absorption spectra. This finding convincingly demonstrates the successful formation of a shell structure encapsulating the perovskite NCs, as corroborated by HRTEM analysis.

Surface charging effects were ruled out by analyzing the C1s spectra (Fig. S2), which exhibited identical binding energies and FWHM values for all the synthesised samples. Despite the changes in the binding energies of the O1s, Ti2p3, Cs3d5, Pb4f, and Br3d5 photoelectron lines, the stoichiometric ratios for Cs:Pb:Br and Ti:O remained consistent at approximately 1:1:3 and 1:2 (Table S1) for both the CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au structures, suggesting no significant stoichiometric deviations in the perovskite core or the TiO₂ shell. The observed peak shifts are likely due to interface formation between materials with different band structures^26,27.

As shown in Table S2, the photoelectron lines Cs 3d5, Pb 4f7, and Br 3d5 for the CsPbBr₃@TiO₂ sample shifted towards higher binding energies, whereas the Ti 2p3 and O 1s lines shift to lower values (~0.2 eV). This indicates significant electron charge transfer from the CsPbBr₃ NCs to the TiO₂ shell. For the CsPbBr₃@TiO₂/Au sample, the Ti 2p3 and O 1s lines exhibit opposite trends, shifting towards higher binding energies. This suggests the formation of an interface between the TiO₂ and Au. The Au 4f7 photoelectron line at 83.8 eV is shifted 0.2 eV below the standard reference, indicating electron transfer from the TiO₂ shell to the Au NPs. The Cs 3d5, Pb 4f7, and Br 3d5 lines continue to shift towards higher binding energies, which is consistent with the enhanced charge transfer from the CsPbBr₃ core to the TiO₂ shell and the formation of a double interface that alters the band alignment.

Notably, the Pb 4f7 peak retains its symmetric shape and exhibits only minor and uniform broadening (FWHM ~1.15–1.18 eV) with no observable shoulders or additional components. Furthermore, no signals corresponding to Pb–O, Pb–OH, or carbonate species are detected in the Pb 4f, O 1s, and C 1s spectra (Fig. 3, Fig. S2). This indicates that the observed shifts are not due to the degradation of the perovskite lattice, rather they arise from interfacial effects and dipole formation at the CsPbBr₃/TiO₂ and TiO₂/Au boundaries. These features underscore the role of Au in tuning the heterojunction structure and facilitating interfacial charge separation, which contributes to enhanced photocatalytic activity.

An equilibrium is established at the interface between semiconductors having different work functions (WF) and Fermi levels through charge transfer. This gives rise to band bending and the formation of depletion regions. These phenomena occur as the system adjusts to align the Fermi levels of the two materials, thereby minimising the energy difference at the interface. The redistribution of charge modifies the local electrostatic potential, resulting in shifting of the positions of the energy bands relative to the vacuum level, which, in turn, alters the binding energies of the emitted photoelectrons.

Although ultraviolet photoelectron spectroscopy (UPS) is widely used to probe the valence band edges and work functions, its application to nanocomposite systems such as CsPbBr₃@TiO₂/Au is limited because of several critical factors. These materials exhibit strong adsorption of adventitious carbon (Fig. S2) from ambient exposure and possess a rough and developed heterogeneous morphology. Consequently, UPS measurements are highly susceptible to surface contamination, which distorts the secondary electron cutoff and valence band edge, making accurate data acquisition unreliable. Although ion sputtering is often employed to clean surfaces prior to UPS measurements, this step is not suitable in our case because it irreversibly damages the composite structure and alters the interfacial electronic properties. In contrast, XPS with Al Kα 1486.6 eV excitation enables the analysis of subsurface layers up to ~5 nm in depth, directly corresponding to the shell thickness and interfacial region of our heterostructures. Combined with its sufficient energy resolution (~0.01 eV), XPS is a more appropriate technique for assessing interfacial charge redistribution in CsPbBr₃@TiO₂/Au systems.

Thus, to confirm the presence of interfaces in the CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au structures, the valence band maximum (VBM), WF, and bandgap (E_g) were determined using XPS and UV-Vis spectroscopy.

The high-resolution valence band (VB) spectra of the CsPbBr₃ NCs, TiO₂ control samples, and CsPbBr₃@TiO₂ structures are shown in (Fig. 3h). Because of the significant overlap between the VB of the CsPbBr₃ NCs and TiO₂, determining the level shifts in the CsPbBr₃@TiO₂ structure was challenging. Nevertheless, VBM values of 1.45 eV for NPs and 2.51 eV for TiO₂ suggest that the primary contribution to VBM shifts in the CsPbBr₃@TiO₂ structure comes from the CsPbBr₃ NCs. This red shift (~0.12 eV) indicates interface formation between the CsPbBr₃ NCs and TiO₂, leading to charge redistribution, Fermi level alignment, and band bending. Given the dominant contribution of CsPbBr₃ to the VBM of the CsPbBr₃@TiO₂ system, the perovskite likely acts as an electron donor, with its bands bending downward as follows:

Using VBM and band gap values, assuming E_f = 0, the Fermi level relative to the conduction band minimum (CBM) is 0.91 eV for CsPbBr₃ NPs and 0.46 eV for TiO₂, consistent with n-type conductivity in both materials. The work functions of $ {WF}_{CsPbBr3}=4.1\;{\rm{eV}} $ and $ {WF}_{TiO2}= 4.5\;{\rm{eV}} $ (Fig. 3i). These results align with the hypothesis of downward band bending in CsPbBr₃ relative to the Fermi level, whereas the TiO₂ bands bend upward for Fermi level alignment. The VBM shift of 0.12 eV and a work function difference of 0.4 eV yield a TiO₂ band bending estimate, relative to the Fermi level, of

indicating asymmetric Fermi level shifts within the bandgaps of CsPbBr₃ and TiO₂ relative to the vacuum level. The Fermi level shift upon forming semiconductor contacts depends on the carrier concentration, as described by

where $ {N}_{CsPbBr3} $ and $ {N}_{TiO2} $ are the carrier concentrations in CsPbBr₃ and TiO₂, respectively.

The band diagrams were constructed based on the obtained VBM, E_f, and WF values (Fig. 4). In the initial stage (before contact), the energy levels of CsPbBr₃ and TiO₂ are independent, with their conduction bands (CB) and VB separated and their Fermi levels located at different energy positions. After the formation of the CsPbBr₃@TiO₂ structure, Fermi-level alignment occurs, causing band bending. The CB and VB of CsPbBr₃ shift downward by approximately 0.12 eV, while the bands of TiO₂ shift upward by about 0.28 eV, creating a barrier height of 0.4 eV. The overall Fermi level of the system is established at approximately −4.22 eV. This band bending leads to the realisation of an “S-scheme” heterojunction, where photoexcited electrons from TiO₂ recombine with holes from CsPbBr₃ at the interface. Simultaneously, charge separation is effectively maintained, with electrons remaining in CsPbBr₃ and holes in TiO₂, facilitating the photocatalytic processes.

The use of spherical colloidal gold NPs as plasmonic enhancers has been widely reported because of their well-defined optical properties and ability to promote hot-electron generation^28–32. The addition of Au NPs to the CsPbBr₃@TiO₂ structure creates a second interface at the TiO₂/Au boundary, resulting in the formation of a Schottky barrier. Under photoexcitation, hot electrons generated in the Au NPs can overcome this barrier and enter the conduction band of TiO₂, where they participate in interfacial redox reactions. Simultaneously, charge separation is further improved owing to the presence of this energy barrier, which suppresses electron backflow and promotes directional carrier flow. Band diagrams (Fig. 4), including the observed energy shifts (0.12 eV for CsPbBr₃ and 0.28 eV for TiO₂), are consistent with the XPS analysis and confirm the formation of both S-scheme and Schottky interfaces, highlighting the cooperative role of each component in optimizing the charge transfer and in enhancing the photocatalytic performance.

We performed UV-Vis absorption, PL, and TRPL measurements to gain a deeper understanding of the optical properties and recombination processes (Fig. 5). The absorption properties (Fig. 5a, c) of CsPbBr₃@TiO₂ were higher than those of the pure CsPbBr₃ NCs, particularly in the UV region. According to the PL measurements (Fig. 5b), the peaks of the CsPbBr₃ NCs of the CsPbBr₃@TiO₂ NPs correspond to 511 and 514 nm. This red shift can occur for TiO₂ coated NPs because of the removal of the lowest fraction of NPs during the purification process. At the same time, a drastic decrease in the PL intensity is observed, suggesting that the catalytic efficiency increased. As the radiation recombination rate decreases, charge separation improves, potentially increasing the amount of charge directed towards the reaction.¹⁸ As a result of the excitation mapping of the CsPbBr₃ NCs and CsPbBr₃@TiO₂ NPs, strong quenching of the PL Intensity was observed in the range 265–480 nm. This range covers the excitation range of the UV-Vis lamp used in the catalytic experiments. Fig. 5d, e present the Tauc plots for the CsPbBr₃ NCs and CsPbBr₃@TiO₂ structures (d) and the TiO₂ control sample (e). The CsPbBr₃@TiO₂ structure’s band gap value of 2.36 eV matches that of the CsPbBr₃ NCs, indicating that the perovskite structure remains intact after TiO₂ shell formation, as confirmed by the HRTEM, XRD, and XPS data.

Fig. 5  Optical characterization of CsPbBr₃ NCs and CsPbBr₃@TiO₂ core-shell NPs. a Absorption spectra of CsPbBr₃ NCs and CsPbBr₃@TiO₂ core-shell NPs. b PL spectra of CsPbBr₃ NCs and CsPbBr₃@TiO₂ core-shell NPs. Excitation wavelength corresponds to 375 nm c Excitation mapping of CsPbBr₃ NCs and CsPbBr₃@TiO₂ core-shell NPs. d-e Tauc plots of CsPbBr₃ NCs and CsPbBr₃@TiO₂ NPs. f TRPL measurements of CsPbBr₃ NCs and CsPbBr₃@TiO₂ core-shell NPs. Excitation wavelength corresponds to 375 nm, frequency corresponds to 100 kHz, power corresponds to 6.3 µW.

TRPL spectrophotometry (Fig. 5f) was used to examine the recombination process. During this analysis, we estimated the initial charge carrier density for TiO₂ (1.2 × 10¹⁷ cm⁻³) and CsPbBr₃ (2.77 × 10¹⁷ cm⁻³) (Table S3), which are necessary for adjusting the energy levels and their bending in the band structures of the diagram (Fig. 5). By substituting the calculated carrier concentration values into (Eq. 2), the Fermi level shifts for each material were determined as approximately 0.11 eV downward for CsPbBr₃ and 0.29 eV upward for TiO₂. These shifts reflect the redistribution of charge carriers at the interface and are consistent with the VBM positions obtained from the XPS analysis, confirming the alignment of the Fermi levels in the system.

The ABC model (Eq. 4) was used for this calculation because it is suitable for perovskites with low charge carrier concentrations³³

Term A is responsible for trap-assisted recombination, B for bimolecular recombination, and C for Auger recombination. At low fluence, Auger recombination does not contribute to charge carrier recombination;³⁴ thus, the normalised PL (Eq. 5) decay curve exhibits bimolecular behaviour and can be approximated using the following equation:

where n₀ is the maximum charge carrier density.

As mentioned previously, the PL intensity of CsPbBr₃@TiO₂ is much weaker than that of pristine CsPbBr₃ NCs (Fig. 5b), confirming that radiative recombination is suppressed because of the feasibility of charge transfer across the interface of the CsPbBr₃@TiO₂ heterostructure. This point of view is also confirmed by the results of TRPL (Fig. 5f and Fig. S4) spectrophotometry: The PL lifetime of CsPbBr₃ NCs (31.5 ns) is twice that of CsPbBr₃@TiO₂ (17.4 ns), which once again indicates that part of the radiative recombination is reduced by the addition of TiO₂ (1.7 ns) (Table S3-S4). The CsPbBr₃@TiO₂ NPs exhibited a faster PL decay than the CsPbBr₃ NCs, which can be attributed to the increase in carrier migration in this S-scheme heterostructure¹⁷. This suggests an improvement in the catalytic activity. Moreover, we observed that the PL lifetime of CsPbBr₃@TiO₂/Au (12.1 ns) (Table S4) exhibited an even faster decay time than the CsPbBr₃@TiO₂ NPs, corresponding to the idea that reducing the radiation recombination rate enhances charge separation, thereby improving the catalytic performance of the system. Using the TRPL spectrophotometry results, we calculated the electron transfer rate (K_et = 0.25 × 10⁸ ns⁻¹) using the following equation³⁵:

The value obtained by He et al. corresponds to 5.57 × 10⁸ ns⁻¹. In that study, the major component of the particles was attributed to TiO₂ demonstrating a low difference between the well-known TiO₂ catalysts³⁵. In contrast, in our study, a lower TiO₂ content relative to that of CsPbBr₃ resulted in a lower electron transfer rate. However, this difference does not fully reflect the charge-separation ability of the catalyst. In this study, we used a smaller amount of TiO₂ to achieve greater stability while still demonstrating the full range of advantages of perovskite-based catalysis, such as tunable reaction energies for various reactions.

To further investigate the role of the heterojunction in the interfacial charge separation and transport, TPC measurements were performed under periodic illumination (Fig. S6). The CsPbBr₃@TiO₂/Au heterostructure exhibited a markedly enhanced photocurrent response compared to the CsPbBr₃@TiO₂ and pristine CsPbBr₃ NCs, both in amplitude and stability. This enhancement is indicative of more efficient photoinduced charge separation and improved carrier extraction across the dual-interface architecture comprising the CsPbBr₃/TiO₂ and TiO₂/Au junctions.

Complementary EIS measurements (Fig. S7) performed under both dark and illuminated conditions reveal a significant reduction in charge transfer resistance (Rct) upon illumination. In particular, the CsPbBr₃@TiO₂/Au sample demonstrates the most pronounced decrease in Nyquist semicircle diameter under light exposure, confirming the facilitation of interfacial electron transfer processes in the presence of the plasmonic Au component.

These photoelectrochemical results, together with spectroscopic and structural analyses, provide compelling evidence that the formation of dual heterojunctions and the integration of Au NPs synergistically promote directional charge transport and suppress interfacial recombination, thereby enhancing the overall photoelectrocatalytic performance.

For the CO₂ reduction reaction, copper (Cu) foam was used as a reference, along with copper foam impregnated with pristine CsPbBr₃ NCs, CsPbBr₃@TiO₂ core-shell NPs (CT), and CsPbBr₃@TiO₂ core-shell NPs deposited on Au NPs (CsPbBr₃@TiO₂/Au, CTA) (Fig. 6a, b). We suggest the use of copper foam because it provides a large surface area for the developed heterostructure designs. Three distinct electrochemical processes were investigated in the electrochemical cell, as detailed in Fig. S8. The results for the copper foam impregnated with CsPbBr₃ NCs are not presented, as they are nearly identical to those obtained from catalytic experiments performed on clean copper foam. This similarity is attributed to the low stability of the pristine CsPbBr₃ NCs and their fast degradation in KHCO₃-water solution. Subsequent experiments demonstrated the stability of the water-resistant CsPbBr₃@TiO₂ core-shell NPs.

Fig. 6  Catalytic performance of copper foam (Cu) as a reference, copper foam Impregnated with pristine CsPbBr₃@TiO₂ (CT), and CsPbBr₃@TiO₂ core-shell NPs deposited on Au NPs (CTA) utilized for a electrocatalysis (EC); b PC; c PEC. d Tafel plot of Cu, CT and CTA for 60 min. e Chronoamperometric measurements of Cu, CT and CTA.

Detailed analysis of the reaction products (over 60 min) using gas chromatography revealed the formation of H₂, CO, CH₄, C₂H₄, C₃H₆ (Fig. 6a-c). O₂ was detected only during EC operation and is not included in the PC/PEC product analysis. In this study, we demonstrate the high efficacy of different processes and catalysts in the formation of various products. In electrochemical catalysis, CO was predominantly formed on Cu, CT, and CTA (Fig. 6a), with molar fractions ranging from 70.6% to 77%. Notably, when comparing CT with Cu, there was a 1.5-fold increase in the CH₄ production (from 19.6% to 30.4%). Similarly, on comparing CTA with Cu, we observed nearly the same CH₄ yield (19.2%), along with the formation of a small amount of C₂H₄ (3.8%). In contrast, Cu exhibited limited reduction ability, with H₂ formation observed at 9.8%. Therefore, in the electrochemical process, we observed a shift in product formation from H₂ (on Cu) to CH₄ (on CT) and C₂H₄ (on CTA). In PC (Fig. 6b), 42.4% CO, 42.4% CH₄, and 15.2% C₃H₆ were produced on Cu. For CT, the product distribution shifted significantly with a decrease in CO formation (17%) and an increase in CH₄ production (70.6%), accompanied by a small amount of C₃H₆ (12.4%). For CTA, there was a marked shift towards C₂H₄ formation (22%) and dominant production of C₃H₆ (70.7%), with a smaller yield of CH4 (7.3%). Finally, during photoelectrocatalysis (PEC) (Fig. 6c), Cu produced H₂ (11.1%), CO (47.2%), and CH₄ (41.7%). For CT, we observed a high selectivity for CH₄ (94.1%) and a low yield of C₂H₄ (5.9%). For the CTA, we achieved high yields of CH₄ (41.2%) and C₂H₄ (58.8%). Thus, the most remarkable processes involve the formation of C₃H₆ on CTA in PC and C₂H₄ on CTA in PEC. These products are of significant value because of their potential applications as energy sources for chemical syntheses³⁶. Our results indicate that the number of electrons can be influenced, enabling different activation energies involved in the chemical transformations using two different heterostructured designs (CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au) and three processes (EC, PC, and PEC).

To evaluate the reaction rate and catalytic stability, we carried out electrochemical measurements such as Tafel plot analysis (Fig. 6d) and chronoamperometry (Fig. 6e). In the Tafel plot (Fig. 6d) for the EC process, we observe Tafel slopes, reflecting the efficiency of charge transfer, of 50.2 mV/dec (Cu), 64.1 mV/dec (CT), and 44.2 mV/dec (CTA). This indicates that the reaction rate decreased in the order CTA > Cu > CT. The lower charge transfer efficiency of CT can be attributed to the presence of TiO₂ on the surface of CsPbBr₃, because its bandgap is approximately 3.0 eV (Fig. 5). This complicates the charge transfer between the electrode and the material. In contrast, gold-containing CTA showed better charge transfer efficiency. At the same time, the overpotential difference between Cu and CT is approximately 0.15 V, demonstrating the higher catalytic performance of CT³⁷. This difference is much larger for CTA, approximately 0.45 V, further highlighting its enhanced catalytic activity³⁷. Thus, the catalytic efficiency decreased in the following order: CTA > CT > Cu. Moreover, during the EC process, we assessed the stability and efficiency of the materials using chronoamperometry (Fig. 6e). For Cu and CT, we observed an almost constant current density (j) at -50 mA/cm² for 40 min and 60 min, respectively, indicating that the materials are stable. In contrast, for CTA, we observed an increase in current density from −130 mA/cm² to −170 mA/cm² over 60 min, suggesting that the electrochemical reaction becomes more efficient over time³⁸. This improvement can be attributed to catalyst activation, surface modification, or enhanced charge transfer.

Based on the results obtained (Fig. 6a-c), we discuss possible pathways for the chemical transformations. The products of the electro- or photocatalytic conversion of carbon dioxide are usually dominated by CO³⁹, CH₄⁴⁰, and CH₃OH (methanol),⁴¹ with one carbon atom in the structure, called C₁ products. Hydrocarbon compounds with two (C₂) or more carbon atoms can also be formed, but in most cases in much smaller quantities; nevertheless, they are of considerable interest because they are a crucial chemical feedstock^42,43. From this perspective, our results indicate that the conversion of CO₂ on the surface of CsPbBr₃@TiO₂/Au to C₃H₆ (in PC) and C₂H₄ (in PEC) exceeded 70% and 58%, respectively, which is of great importance and requires a description of the corresponding mechanisms.

In the literature, two key mechanisms of CO₂ conversion with the participation of water in the C₁ products are mainly considered: (1) the fast hydrogenation (or formaldehyde) pathway and (2) the fast deoxygenation (or carbene) pathway,^44–48 with *COOH and *CHOO as the most likely first intermediates, respectively (* means surface-bound). Most likely, both pathways are realised in practice; however, the predominant role of one is determined by the formation of products and intermediates, which are observed experimentally. The TiO₂ surface stabilises the reaction intermediate by forming a chemical bond between CO₂ (and HCO₃⁻) and the catalyst, leading to a less negative redox potential⁴⁷. Thus, CH₃OH, along with CH₄, is a reaction product of the carbene pathway; therefore, it should be part of the gas mixture obtained as a result of conversion. In the formaldehyde pathway, methanol is an intermediate surface-bound compound formed before the formation of methane; therefore, it was not detected in the reaction products. Because we did not observe the formation of CH₃OH, the fast hydrogenation mechanism was chosen as the main mechanism to describe CO₂ conversion in our experiments to further propose, on this basis, an idea of how the formation of C₂ products occurs. Notably, the formaldehyde pathway can be accomplished by forming CO or formic acid (HCOOH). HCOOH formation is thermodynamically more favourable, while the carbon monoxide formation is kinetically more favourable⁴⁹. Because we observed the formation of CO in our experiments, the pathway involved was chosen for further discussion (Fig. 7).

Fig. 7  The proposed scheme of chemical transformations on the surface of CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au during CO₂ conversion in the three considered catalytic processes with the predominant formation of the corresponding products. The schematic inset illustrates the underlying reason for the formation of different products when using CsPbBr₃@TiO₂ versus CsPbBr₃@TiO₂/Au. This distinction arises from the difference in the number of electrons involved in the reaction, ultimately determining the product distribution.

The next step was to move from the mechanism of C₁ product formation to the C₂ product formation. It has been previously shown that CO₂ conversion can yield ethylene⁵⁰, ethane⁵¹, and acetic acid⁵² with a relatively low selectivity of less than 50%. The mechanism of C₂ product formation is assumed to be based on a slow multi-electron reduction and a carbon-carbon (C-C) coupling driven by the high energy generated by illumination⁵³. It has been shown that the binding of two CO molecules occurs through the interaction of carbon atoms with metal atoms with increased electron density^13,14. Accordingly, the reaction pathway proposed by Ji et al.⁵⁴ was used as a reference to explain the formation of C₂H₄ from CO₂ on the CsPbBr₃@TiO₂/Au surface. In our study, this pathway was further refined by clearly distinguishing between the photocatalytic, electrocatalytic, and photoelectrocatalytic processes (Fig. 7). We did not include all intermediate steps in the reaction scheme, indicating only those steps demonstrating fundamental importance for understanding the reasons for obtaining the observed products.

Summarising the mechanisms of CO₂ conversion briefly described above, the following conclusions can be drawn: In all the three catalytic processes considered, the first step was the deoxygenation of CO₂ with the formation of CO bound to the TiO₂ surface. In the case of EC, a major portion of carbon monoxide molecules is desorbed, and the remaining part participates in the formation of CH₄ and, to a lesser extent, in that of C₂H₄ (Fig. 7). In PEC, most CO molecules are coupled to form COCO compounds and the remaining molecules are converted into methane molecules along the formaldehyde pathway. The formation of COCO dimers was facilitated by excess electrons, which was supported by the proposed heterostructured Schottky/S-scheme of CsPbBr₃@TiO₂/Au (Fig. 4). These dimers were predominantly converted into C₂H₄. This interpretation is further supported by the EPR data (Fig. S3), which reveal predominant •OH radical formation, indicative of a high oxidative potential and efficient charge separation. Notably, the EPR spectra of the Au-containing system show a lower relative contribution from O₂⁻• adducts compared to that of the non-Au sample, suggesting that more photogenerated electrons are actively involved in the reduction reactions⁵⁵. This increased electron availability in the CTA system correlates with the observed shift in the product distribution towards multielectron C₂ products. Based on the band-edge positions¹⁹, the observed redox behaviour is consistent with the S-scheme mechanism.

Based on recent studies, it can be concluded that propylene is obtained from the coupling of the vinyl alcohol intermediates *C₂ and *C₁ (and/or *C₁') sorbed on the surface of TiO₂⁴². These studies have shown that allyl alcohol (the product of the coupling of the aforementioned intermediates, *C₁ and *C₂) is responsible for the formation of propylene.

In this study, we demonstrated a novel strategy for CO₂ conversion by integrating perovskite-based photoelectrocatalysis with localised surface plasmon resonance enhancement. To overcome the problem of lead halide perovskite instability in water, we fabricated stable CsPbBr₃@TiO₂ core-shell NPs and investigated their structural and optical properties. Further, we prepared layered materials based on these NPs and combined them with Au NPs. Furthermore, we explored three distinct catalytic processes — electrocatalytic, photocatalytic, and photoelectrocatalytic — to tune the selectivity of CO₂ reduction to a range of valuable products, including H₂, CO, CH₄, C₂H₄, and C₃H₆.

Comprehensive structural and optical analyses of the CsPbBr₃@TiO₂ core-shell NPs revealed successful synthesis and effective coating of TiO₂ on the CsPbBr₃ NPs. Through the use of TEM, SEM, XRD, and XPS the formation of high-quality NPs was confirmed and provided detailed insights into their structural and elemental compositions. Moreover, based on XPS and optical measurements (UV-Vis absorption, PL spectroscopy, and TRPL spectrophotometry), we described the electronic band structure necessary for understanding the charge dynamics. The analysed charge dynamics demonstrated excellent enhancements in charge separation, indicating a promising path for improving catalytic efficiency.

Using CsPbBr₃@TiO₂ and CsPbBr₃@TiO₂/Au heterostructured designs for electrocatalysis, photocatalysis, and photoelectrocatalysis, we achieved high selectivity in the CO₂ reduction reaction. By controlling the selectivity, we demonstrated that CsPbBr₃@TiO₂ NPs and CsPbBr₃@TiO₂/Au heterostructures can be tuned to produce valuable product varieties. Notably, our findings highlight the potential of perovskite-based catalysts with Cu foam as a reference material in electrocatalytic experiments. The introduction of Au into the CsPbBr₃@TiO₂/Au system further enhanced its catalytic performance, demonstrating the beneficial role of localised surface plasmon resonance enhancement in the photoelectrocatalytic process. Under various catalytic conditions, the reaction products shifted towards higher-order hydrocarbons such as C₂H₄ and C₃H₆, which are of significant interest for chemical synthesis as energy sources.

In terms of catalytic stability, chronoamperometry experiments revealed that both CsPbBr₃@TiO₂ core-shell NPs and CsPbBr₃@TiO₂/Au heterostructures exhibit high stability over extended periods of electrochemical measurements. These findings provide a strong foundation for the future development of stable, efficient, and scalable photoelectrocatalytic systems for sustainable CO₂ conversion.

This study was financially supported by the National Natural Science Foundation of China (Project No.62350610272). S.V.M acknowledges the Department of Science and Technology of Shandong Province (Grant KY0020240040) and the Priority 2030 Federal Academic Leadership Program. P.O.K. acknowledges the support from the Ministry of Science and Higher Education of the Russian Federation (Project No. FSRZ 2023-0006). Fig. 1 was drawn using pictures from Servier Medical Art. Servier Medical Art by Servier is licensed under a Creative Commons Attribution 3.0 Unported License.