Visible-light photocatalytic CO₂ hydrogenation using surface-alloyed plasmonic AgPt nanoprisms

Abstract

Development of suitable catalysts for light-driven CO₂ hydrogenation is an alluring goal in catalysis. In this study, plasmonic Ag nanoprisms were combined with Pt to make surface-alloyed nanoparticles for aqueous-phase CO₂ hydrogenation. The Pt loading favoured the product selectivity towards multi-electron C₁ products and promoted acetic acid production via C–C coupling. Increasing the reaction pressure further improved acetic acid production where the highest yield of 0.491 mmol g_cat⁻¹ was achieved at 20 bar. Within the visible-light region, the in-plane dipole resonance peak of Ag₉₁Pt₉ at 600 nm contributed the highest apparent quantum yield of 26.7%. These investigations demonstrated the significance of designer plasmonic catalysts and highlighted their photocatalytic enhancement towards CO₂ conversion.

---

Broader context

Photocatalytic CO₂ conversion to valuable C₁–C₂ chemicals offers a low carbon intensity pathway to combat climate change and replace fossil fuels as feedstock. This study focuses on enhancing photocatalytic activity by developing plasmonic surface-alloyed AgPt nanoprisms, to convert CO₂ in an aqueous environment at room temperature. The addition of Pt enhances productivity towards a variety of chemicals, particularly acetic acid. The investigations on using visible-light illumination and achieving high apparent quantum yield underscore the ability of plasmon-mediated photocatalysis to enhance CO₂ conversion. These findings highlight the potential of using plasmonic materials as photocatalysts.

Introduction

Carbon capture, utilisation and storage (CCUS) technologies have expanded over the last decade, supporting the emissions reduction target towards creating a circular economy.¹ Bulk chemical production from CO₂ offers an opportunity to displace conventionally used fossil-derived chemicals. Carbon-neutral chemicals can be produced by hydrogenating the captured CO₂ using green hydrogen.² Photocatalytic CO₂ hydrogenation is an emerging area of interest due to its potential to utilise sunlight.³

Semiconductor photocatalysts have been studied extensively for CO₂ hydrogenation to produce C₁ chemicals such as carbon monoxide (CO),^4,5 methane (CH₄),^4–7 formic acid (HCOOH),^8–10 formaldehyde (HCHO)^11,12 and methanol (CH₃OH)^13–15 and C₂ chemicals including acetic acid (CH₃COOH),⁸ acetaldehyde (CH₃CHO),¹⁶ ethane (C₂H₆)^5,12 and ethanol (C₂H₅OH).^16,17 The major challenges of photocatalytic CO₂ reduction and hydrogenation can be attributed to (1) poor multi-component catalyst design hindering light-to-energy transfer, (2) limited light absorption range, and (3) limited control of product selectivity because of a lack of understanding of reaction pathway. Plasmonic nanomaterials have garnered interest for various applications,^18–20 including their potential as photocatalysts.^21–24 Noble metal nanoparticles (NPs) made of silver (Ag), gold (Au) and copper (Cu) can capture the incident light through the collective excitation of conduction electrons. The spatial confinement of this excitation because of a NP's exterior boundary results in a resonant oscillation, known as the localised surface plasmon resonance (LSPR).^25,26 LSPR produces a local electric field amplification and decay by creating hot charge carriers and phonon vibrations.^27,28

Ag provides several advantages over other plasmonic metals (e.g. Au, Cu) including high plasmon quality factor, tunable optical properties and distinguishable interband and intraband transitions.²⁹ Various factors affect LSPR such as particle size, morphology, composition and surrounding environment.³⁰ Among these, silver nanoprism-like morphology has multiple tunable LSPR modes in the visible region and higher photocatalytic performance.^31,32 Ag-based catalysts have shown CO₂ activation. However, co-catalysts are required to enhance the photocatalytic properties. Combining Ag with a catalytically active metal such as platinum (Pt) can overcome the challenges with photocatalysts in hydrogenating CO₂.^33–37 Researchers have also found the plasmon energy transfer within Ag–Pt nanocubes useful for photochemical reactions.^38,39

Here, we show plasmon-mediated photocatalytic capability of surface-alloyed AgPt nanoprisms (NPrs) in and aqueous-phase CO₂ hydrogenation reaction to produce HCOOH, HCHO and CH₃COOH. The AgPt NPrs were synthesised through galvanic replacement reaction (GRR) of Pt with Ag NPrs. The deposition of Pt on the Ag NPrs was surface-limited, which formed Ag–Pt alloys. The catalyst performance improved significantly with increasing Pt loading and reaction pressure. Control experiments confirmed the visible-light-driven photocatalysis consuming CO₂ and H₂. The in-plane dipole resonance peak was majorly responsible for the catalytic performance. We also investigated catalyst deactivation through colloidal instability. The reaction pathway was interpreted based on the catalyst selectivity and hydrogenation of selected intermediates.

Results

Materials characterisation

Ag nanoprisms (Ag NPrs) were synthesised according to the method developed by Aherne and co-authors with modifications to increase sample concentration.⁴⁰ The colloidal synthesis produced a high proportion of platelet nanoprisms (NPr) (77%), which were categorised into triangles (29%) with an edge length of 63 ± 10 nm, and discs (48%) with a diameter of 42 ± 6 nm (Fig. S.1, S.2 and Table S.1, ESI†). Pt was incorporated in the Ag NPr by a galvanic replacement reaction (GRR) using K₂PtCl₄. The Pt loadings were calculated from the total moles of Ag and Pt used in synthesis, and samples were labelled Ag_100−xPt_x, where x corresponded with loadings from 1.5–9 mol% Pt, assuming complete GRR. The UV-vis spectra of Ag NPrs and AgPt NPrs showed both red-shift and dampening of the in-plane dipole resonance peak of Ag NPr at 600 nm, proportional to the Pt loadings (Fig. 1 and Table S.2, ESI†). This was attributed to the galvanic etching of Ag NPr and Ag–Pt alloy formation on the NPr surface.^41–45 In comparison, the Pt NP spectra did not show any appreciable plasmon resonance peaks in the wavelength region (Fig. 1).

Fig. 1 UV-vis spectra of AgPt surface alloyed catalyst with different Pt loading.

In comparison with as-synthesised Ag NPrs (Fig. 2A), edge deformation was observed in bright field scanning transmission electron microscopy (BF-STEM) micrographs for all AgPt NPrs with varying Pt loadings (Fig. 2C, E, G and I). Between samples, the dimensions of the NPrs (Fig. S.2C, ESI†) and the degree of edge deformation did not vary significantly, but Kirkendall voids were present in at higher Pt loadings (Fig. 2I and J) which may contribute to the non-linear relationship between Pt concentration and in-plane dipole resonance wavelength (Table S.2, ESI†).

Fig. 2 (A)–(L) STEM-EDS micrographs of (A) and (B) Ag NPr (C) and (D) Ag_98.5Pt_1.5 (E) and (F) Ag₉₈Pt₂ (G) and (H) Ag_96.5Pt_3.5 (I) and (J) Ag₉₁Pt₉ (K) and (L) Pt NP (inset of STEM micrographs shows corresponding bulk sample solutions) with corresponding elemental maps of Ag (red) and Pt (green) (M) rotational averages of selected area diffraction (SAED) patterns of corresponding catalysts.

STEM energy dispersive spectroscopy (STEM-EDS) mapping of clusters of NPrs was used to quantify Ag/Pt composition of the samples (Table 1) and confirmed that free Pt NPs were not present. Mapping of individual NPrs at all loadings revealed Pt was distributed across the entire surface but was concentrated at the edges of the triangular face (Fig. 2C–J). As STEM-EDS is a depth-insensitive technique, EDS line scans of Ag₉₁Pt₉ at different orientations were collected (Fig. S.3, ESI†). Pt was most abundant at the edges of the NPrs in both orientations, and a pronounced minimum of Pt occurred at the centre of the NPr in the edge-on case. Given these results and previous literature showing the surface-limited nature of alloy formation through galvanic processes, the Pt deposition was determined to be surface-limited.^42,46

Table 1 Composition of Ag and Pt (mol%) characterised from STEM-EDS, ICP-MS and XPS analysis

Sample	STEM-EDS		ICP-MS		XPS
	Ag (mol%)	Pt (mol%)	Ag (mol%)	Pt (mol%)	Ag^0 (mol%)	Pt^0 (mol%)	Pt^2+ (mol%)
Ag NPr	100	—	100	—	100	—	—
Ag98.5Pt1.5	98.9 ± 5.0	1.1 ± 0.1	98.5 ± 0.7	1.46 ± 0.07	97.6	2.25	0.10
Ag98Pt2	98.8 ± 5.1	1.2 ± 0.2	97.9 ± 0.3	2.07 ± 0.30	95.5	3.85	0.67
Ag96.5Pt3.5	97.4 ± 4.9	2.6 ± 0.2	96.4 ± 1.3	3.59 ± 0.12	94.2	4.89	0.93
Ag91Pt9	92.6 ± 4.2	7.4 ± 0.4	92.2 ± 0.5	7.81 ± 0.46	87.2	7.04	5.73

---

The crystal structure of the NPrs was studied using selected area electron diffraction (SAED) (Fig. 2M). In agreement with Aherne and co-authors, peaks for HCP {1010}, FCC {111} and {220} planes were identified in the Ag NPr sample.⁴⁰ Based on Vegard's Law and the lattice parameters of Ag (4.08 Å) and Pt (3.92 Å), a linear expansion of the real space inter-planar distances with increasing Pt concentration was expected. As Pt concentration increased up to Ag_96.5Pt_3.5, the HCP {1010} and FCC {111} peak positions remained unchanged, while the FCC {220} peak underwent a small reciprocal space expansion consistent with Ag–Pt alloying under Vegard's Law. HCP {1010} and FCC {111} planes occurred on the flat triangular faces and through the interior of the NPr, while FCC {220} planes were predominantly present at the edges and tips. The change to only the {220} peak position suggests that, at concentrations at or below 3.5%, Pt was mainly deposited along the NPr edges.

However, at Ag₉₁Pt₉, the crystallography of the sample changed considerably. The HCP {1010} peak was still present, though shifted from 4.06 nm⁻¹ in Ag to 3.99 nm⁻¹ in Ag₉₁Pt₉. Similarly, the FCC {220} peak shifted from 7.03 nm⁻¹ in Ag to 6.87 nm⁻¹ in Ag₉₁Pt₉ in a reversal of trend observed at lower Pt concentrations (where the peak gradually shifted from 7.03 nm⁻¹ to 7.05 nm⁻¹). The large peak at 4.18 nm⁻¹ did not align with any of the expected reflections. Based on the lattice parameter of Ag, the FCC {111} reflection can be calculated to be 4.23 nm⁻¹, but in the Ag NPr sample, FCC {111} was observed at 4.30 nm⁻¹. It was unlikely that the 4.18 nm⁻¹ peak is FCC {111}, as this would require a significant shift from the corresponding Ag NPr peak in the opposite direction to what was expected from AgPt alloying. However, given similar shifts occurred for the HCP {1010} and FCC {220} peaks in this sample, the FCC {111} assignment cannot be ruled out. The other explanation for the 4.18 nm⁻¹ peak is an Ag–Pt alloy phase of HCP {1010}. Pt has been observed to adopt an HCP lattice structure when deposited on another HCP surface and may be responsible for the anomalous peak.^47,48 High resolution TEM (HR-TEM) for single particle measurement do not give us conclusive indication of the existing changes as the lattice parameters of Ag and Pt are within the range of ∼5% error (Fig. S.4, ESI†). The atom-to-atom distance varied from 0.22 to 0.32 nm. These variations resulted in an average of 0.27 nm which are significantly higher than a 5% error (Fig. S.5, ESI†).

XPS analysis was conducted to investigate the efficiency of Pt deposition by GRR and determine the surface composition of the Ag Pt NPs (Fig. 3). Ag 3d spectra of Ag NPr showed binding energy peaks at 367.7 eV (3d_5/2 spin–orbit component), corresponding to Ag⁰⁴⁹ (Fig. 3A). The Ag 3d_5/2 peaks in AgPt NPrs also showed similar peak profiles at ∼367.7 eV, while no signals were observed for Ag⁺. For AgPt, Pt 4f spectra showed both Pt⁰ and Pt²⁺ signals with binding energies at 70.1 and 72 eV of 4f_7/2 spin orbital^50,51 (Fig. 3B). The intensity of the Pt²⁺ signal was proportional to Pt loading, thus showing the limited extent of GRR. The XPS results were used to quantify the elemental composition of Ag⁰, Pt⁰ and Pt²⁺ in the different samples (Table 1). C 1s spectra for Ag NPr showed peaks at 285 eV (C–H), 286.5 eV (C–O) and 288 eV (C=O), which likely results from the citrate used as capping agent.⁵² Similarly, O 1s spectra for Ag NPr showed peaks at ∼531 eV (C=O) and 533 eV (C–O/C–OH) (Fig. S.6, ESI†).⁵³ The peak signals corresponding to C=O were significantly lower on Ag₉₁Pt₉, resulting from the loss of capping agent on the surface.⁴⁹

Fig. 3 XPS spectra of AgPt NPrs in (A) Ag 3d region and (B) Pt 4f region.

Table 1 summarises the metal composition measured through various techniques. The measurements from STEM-EDS of particle maps and ICP-MS of bulk solution were in good agreement and were close to theoretical estimates. Results from XPS quantification corresponded to the surface composition, which showed higher Pt loading. This was expected as XPS is a surface-sensitive technique. This is in agreement with the surface-limited bimetallic structure seen in STEM line scans (Fig. S.3, ESI†). Of the total Pt signals, Pt²⁺ increased from ∼4.3% to ∼45% with increasing Pt loading. This indicated that the degree of galvanic reduction was inversely correlated with the Pt loading.

Photocatalytic CO₂ hydrogenation

Effect of Pt loading. We investigated the addition of Pt towards the product selectivity in aqueous-phase CO₂ hydrogenation operated in a batch photo reactor (Fig. S.7A and B, ESI†) at 10 bar. The catalyst loading was selected based on wavelength absorbance to reduce optical saturation (Fig S.7D and Table S.3, ESI†). HCOOH, HCHO and CH₃COOH were detected as liquid products, whereas no gaseous products were obtained. As seen in Fig. 4A, adding 1.5 mol% Pt significantly improved the HCOOH yield by approximately 10-fold, to 0.42 ± 0.12 mmol g_cat⁻¹. A maximum HCOOH yield of 0.52 ± 0.08 mmol g_cat⁻¹ was obtained at 2 mol% Pt. Similarly, the highest HCHO yield of 0.16 ± 0.01 mmol g_cat⁻¹ was obtained at 3.5 mol% Pt. With 9 mol% Pt loading, we observed the highest CH₃COOH yield of 0.43 ± 0.03 mmol g_cat⁻¹. This trend revealed that the yield of products requiring higher electron consumption increased with the Pt loading. To validate that this reaction was not driven by Pt solely, we performed the reaction with Pt NPs, which yielded comparably low amounts of HCOOH and CH₃COOH.

Fig. 4 (A) Product yield of AgPt surface alloyed catalysts with different Pt loadings (B) in situ temperature profile of the reactor solution. Reaction conditions – pressure = 10 bar (H₂/CO₂ = 1), filtered Xe lamp illumination, t = 2 h.

In situ temperature profile was measured during the reaction to investigate the effect of Pt loading on the plasmon decay pathway (Fig. 4B). The trend of photothermal response indicated that low Pt loading did not significantly affect the photothermal response, suggesting that electron relaxation to thermalisation was not affected. The highest in situ temperature of ~0.35 °C/mg_cat was observed in Ag₉₈Pt₂ whereas Ag₉₁Pt₉ showed a relatively lower rate and steady-state temperature. The correlation between dampened extinction intensity of the plasmon resonance peaks and temperature increase was similarly reported for Au/Pd system.⁵⁴ The plasmon decay for Ag–Pt system has been reported to under electron and energy transfer, therefore reducing heat generation.^41,55

A series of control reactions verified the visible-light photocatalytic activity of AgPt NPrs. Catalysts were irradiated with a 300 W Xe Lamp in unfiltered and filtered modes corresponding to different spectral intensity profiles (Fig. S.7C, ESI†). As seen in Fig. 5, the product selectivity between the illumination modes varied at lower Pt loadings. This difference in the selectivity at lower Pt loading was due to electrons generated by the interband transitions in Ag that lied in the UV region (Fig. 1 and 5A, B). As the Pt loading increased, the product selectivity was relatively similar as the plasmon decay pathway dissipated towards surface AgPt sites, thus governing the catalytic pathway (Fig. 5C and D).³⁸ A control reaction with no illumination (reaction in dark) was also conducted for Ag₉₁Pt₉, which produced 0.11 mmol g_cat⁻¹ HCOOH and <0.01 mmol g_cat⁻¹ HCHO. The aqueous production of HCOOH has a negative Gibbs free energy and is favourable at these conditions.⁵⁶ The apparent quantum yield (AQY) was calculated for the cumulative production of HCOOH, HCHO and CH₃COOH against the photons emitted in the two illumination modes. The increase in AQY in the filtered mode increased proportional to the Pt loading, thus confirming the visible-light driven nature of the Ag_100−xPt_x nanoprisms.

Fig. 5 Product yields and AQY under unfiltered Xe lamp illumination (320–2500 nm) and Filtered Xe lamp illumination (380–780 nm) of (A) Ag_98.5Pt_1.5 (B) Ag₉₈Pt₂ (C) Ag_95.5Pt_3.5 (D) Ag₉₁Pt₉ reaction conditions – ∼10 mg catalyst, P = 10 bar, H₂/CO₂ (1 : 1 v/v), t = 2 h.

Further control reactions were conducted using Ag₉₁Pt₉ to validate the consumption of CO₂ and H₂ (Fig. S.13, ESI†). The reaction did not proceed without any catalyst. A 100% CO₂ environment yielded 0.012 mmol g_cat⁻¹ HCOOH and 0.053 mmol g_cat⁻¹ HCHO. While Ag-based catalysts have shown HCOOH and HCHO production in similar applications using H₂O as the electron-donating species, this reaction relied on the presence of H₂ as the electron donor.^57–59 To confirm the absence of any other carbon sources in the reaction, a 100% H₂ environment was used. We observed the formation of HCOOH at 0.11 mmol g_cat⁻¹, HCHO at 0.026 mmol g_cat⁻¹ and CH₃COOH at 0.01 mmol g_cat⁻¹. This contradicted our initial assumption. However, we detected ~280 μM bicarbonate (HCO₃⁻) from HPLC analysis in the solution before the reaction commenced which decreased to 190 μM, indicating that CO₂ from the air was absorbed and converted into aqueous species, such as HCO₃⁻ and carbonate (CO₃²⁻), in the feed solution.

To validate the source of carbon species responsible for forming the C₁ and C₂ products in this study, the feed solution was dosed with 0.5 mM HCO₃⁻ and H₂. The results showed increased product yields, with the maximum yields of HCOOH, HCHO and CH₃COOH at 0.19 mmol g_cat⁻¹, 0.072 mmol g_cat⁻¹, and 0.125 mmol g_cat⁻¹, respectively. This confirmed that HCO₃⁻ was a precursor to the hydrogenation reaction in a similar yield trend to that of H₂ + CO₂ environment. Photo-oxidation of adsorbed citrate capping on Ag NPr to CO₂ has been observed.⁶⁰ Hence, we dosed a reaction containing 0.5 mM sodium tricitrate and H₂. No products were detected until 1 h, similar to the 100% H₂ environment case, which showed negligible contribution from citrate.

The optical properties of the NPrs were analysed ex situ at different reaction times (Fig. S.8, ESI†). The plasmon resonance peaks observed in Ag NPr did not change. However, a blue shift was observed in the AgPt NPrs. This suggested changes in the adsorbed surface species on the NPrs, although further investigations were conducted to determine possible causes.^58,61 XPS analysis of the NPrs post-reaction showed some BE shifts for Ag 3d and Pt 4f regions compared to as-synthesised NPrs (Fig. S.9, ESI†). The Ag 3d_5/2 peak for Ag NPr shifted the most from 367.7 eV to 368 eV. The BE shift was reduced with increasing Pt loading. Similarly, the highest BE shift in Pt 4f_7/2 was observed for Ag_98.5Pt_1.5 from 70.1 eV to 70.5 eV. We deduce these shifts in binding energy indicate diffusion of Ag to form an Ag–Pt surface alloy adlayer.⁶²

Aggregation of NPrs was observed among all samples under reaction conditions (Fig. S.10A, ESI†), which was hypothesised to be because of either the loss of capping agent or structural changes. Upon the introduction of reaction gases, a decrease in pH occurred due to. the solvation of CO₂, which created an acidic medium. The decrease in pH increases the protonation of trisodium citrate, which could cause NPrs aggregation.⁶³ Analysis of the solutions pre- and post-reaction revealed an average pH reduction from 6.9 to 4.9 (Fig. S.10B, ESI†). The shift in pH would cause protonation of [Citrate]³⁻ to H₂[Citrate]⁻ and affect the surface charge of the NPr. Therefore, we investigated the change in surface charge by measuring the zeta potential and hydrodynamic radius of the NPrs post-reaction (Fig. S.10C, ESI†). The zeta potential of as-synthesised AgPt NPrs averaged at −28 mV with a mean hydrodynamic radius of 48 nm. The zeta potential of the NPrs post-reaction increased compared to the initial measurement, where this increase was proportional to the Pt loading. Similarly, the hydrodynamic radius increased proportionally to the Pt loading with a mean hydrodynamic radius of ∼80 nm for Ag₉₁Pt₉. XPS analysis of post-reaction AgPt NPrs in C 1s region (Fig. S.9C, ESI†) also showed a significant decrease in the signal area for C=O (288 eV), confirming the cause of aggregation from loss of citrate from the particle surface. This implied that the acidic environment formation of acidic products further caused citrate protonation and particle aggregation.⁶⁴

On the nanoscale, examples of NPrs with structural deformations, such as holes on particle faces, were seen in pre- and post-reaction samples. The frequency of hole formation was higher in the post-reaction samples (Table S.4 and Fig. S.11, ESI†). Catalyst composition was also quantified post-reaction (Table S.5, ESI†). The composition changed compared to as-synthesised samples. However, there were discrepancies among the measurements. We considered these discrepancies as part of sample preparation techniques and analysed the data separately. In comparison, a relative increase in Pt loading of samples post-reaction was observed from ICP-MS and XPS. The increase in Pt loading implied a loss of Ag, which was not detected as Ag⁺ in XPS measurements. The total catalyst recovered was also calculated using ICP-MS which showed significant Ag loss of up to 81.7% while only 10.8% Pt loss occurred in the same sample (Table S.6, ESI†). Hence, we proposed this as an unrecoverable catalyst lost during the reaction by aggregation and colloidal instability. Measurements from STEM-EDS indicated no significant compositional change occurred to the imaged NPrs (Table S.5, ESI†).

Impact of reaction pressure. The effect of pressure on the product yields was investigated because of the low solubility of H₂ at ambient pressures (Fig. 6A). With increasing pressure, HCOOH production was favoured with the maximum yield of 0.46 mmol g_cat⁻¹ at 20 bar. This confirmed the pressure dependence of the photocatalytic reactions and was likely produced by the hydrogenation of dissolved CO₂.⁶⁵ The HCHO yield was not affected significantly by increasing pressure. CH₃COOH was not observed at 1 bar, followed by a significant increase to 0.36 mmol g_cat⁻¹ at 5 bar and reached a maximum yield of 0.49 mmol g_cat⁻¹ at 20 bar. The yield profile of CH₃COOH over the reaction time showed a trend where the production of CH₃COOH production was only observed after 1 h at 5 bar, but reached a near-linear productivity of ~0.196 mmol g_cat⁻¹ h⁻¹ at 20 bar (Fig. 6B). We deduced that the hydrogenation of CO₂ was limited by mass transfer below 5 bar. This was overcome at higher pressures but reached a kinetic limitation of the rate-determining step, as observed by the incremental change in CH₃COOH yield between 5–20 bar (Fig. S.12A, ESI†). We also investigated the colloidal stability as a function of pressure. The pH of the solution decreased at higher reaction pressures owing to the increased solubility of CO₂. As a result, the zeta potential and hydrodynamic radius increased to −0.25 mV and 144 nm respectively, implying increased extent of aggregation (Fig. S.12B and C, ESI†).

Fig. 6 (A) Product yield of Ag₉₁Pt₉ as a function of pressure (B) CH₃COOH yield profile at different pressures. Reaction conditions – ∼10 mg catalyst, H₂/CO₂ (1 : 1), room temperature, filtered illumination mode.

Distinguishing plasmon-mediated catalysis. Because of the anisotropic shape of Ag NPr, multiple LSPR peaks were observed corresponding to in-plane and out-of-plane dipole and quadrupole plasmon resonances (Fig. 1). To distinguish the effect of LSPR peak on catalyst performance, the impact of illumination wavelength was investigated on the product yield of CO₂ hydrogenation using Ag₉₁Pt₉ (Fig. 7). Using a supercontinuum laser with a contrast filter, the excitation wavelengths between 400–900 nm could be controlled to a bandwidth of less than 2.5 nm. Product yields were normalised using the irradiation power (mW). The highest yield contribution of 0.57 mmol g_cat⁻¹ mW⁻¹, was revealed by trends in HCOOH yield at the in-plane dipole resonance peak (600 nm) of Ag₉₁Pt₉. This also coincided with the highest AQY of 26.7%. CH₃COOH was not detected at wavelengths higher than 700 nm. At 600 nm, the highest in situ temperature of 1.7 °C was observed (Fig. S.14A, ESI†). The peak temperature profile correlated well with the yield profile, confirming that plasmon generation was the highest at the in-plane dipole and significantly increased the catalytic performance. HCHO and CH₃COOH exhibited higher selectivity at lower wavelengths (400–500 nm), possibly because of the energy of hot carriers, which is proportional to the energy of the illuminating photon (Fig. S.14B, ESI†). Interestingly, we observed that the yield of HCHO and CH₃COOH were in the same magnitudes across all wavelengths except >700 nm where no CH₃COOH was produced. This implied a rate-limiting step that affected the C–C coupling to CH₃COOH.

Fig. 7 Wavelength-dependent product yields normalised by laser power. Black line shows extinction spectra for Ag₉₁Pt₉ post-reaction.

Reaction pathway investigation. Based on a study conducted by Genovese and co-authors,⁶⁶ the reaction pathway to CH₃COOH was investigated by using several intermediates (NaHCO₃, HCOOH, HCHO, CH₃OH) observed during CO₂ hydrogenation (Table S.7, ESI†). Using NaHCO₃ produces HCO₃⁻ as a reactant that produced all liquid products. Based on previous literature findings, HCO₃⁻ approaches equilibrium with aqueous CO₂ and create COO⁻ on the catalyst surface that was a key reaction intermediate. Comparatively, using HCOOH and HCHO did not yield CH₃COOH significantly higher than control reaction 2 (Fig. S.13, ESI†). This implied that *HCOO and *CHO were not the intermediates involved in C–C coupling towards CH₃COOH. This also suggested that adsorption of these radicals inhibited CH₃COOH production, as the oxidation of these intermediates was not favourable. Using CH₃OH, however, produced a relatively higher CH₃COOH yield, implying *CH₃ as a key intermediate for the reaction. Contrarily, the oxidation of CH₃OH as a hole scavenger could also produce CO₂, contributing to the reaction pathway.⁶⁷ Therefore, we conducted a control reaction involving NaHCO₃ and CH₃OH that produced a higher yield of CH₃COOH. From these reactions, we deduce that CO₂/COO⁻ and *CH₃ were key reaction intermediates in the C–C coupling to CH₃COOH and *HCOO/*CHO resulted from CO₂ hydrogenation but did not participate directly in C–C coupling.

Based on the liquid products observed, the reaction pathway was elucidated based on similar findings from previous studies.^66,68 As shown in Scheme 1, the reaction initiates with the adsorption of CO₂ resulting from dissolved CO₂ and HCO₃⁻. The formation of HCOOH was followed by a two-step electron reduction involving COO⁻ and *HCOO⁻. HCHO was produced by further hydrogenating *HCOO⁻ to *CHO and *CH₂O. *CH₃ was produced from *CH₂O, undergoing a nucleophilic reaction from COO⁻ to produce CH₃COO⁻/CH₃COOH.

Scheme 1 Proposed reaction pathway of CO₂ to HCOOH, HCHO and CH₃COOH on AgPt NPr.

Discussion

The CO₂ hydrogenation performance of Ag₉₁Pt₉ NPrs was comparable with other metal-based photocatalysts reported in the literature (Table S.8, ESI†). While HCOOH and HCHO were produced in similar yields, CH₃COOH was produced at a much higher yield. This could be attributed to the favourability of C–C coupling on AgPt NPrs and optimal reaction conditions. Although most reaction conditions involved H₂O as the electron donor to reduce CO₂ to CO, using H₂ in our experiments significantly improved the product yields. H₂ is used in CO₂-to-CH₃OH, which favours *CH₃ formation. Using a liquid-phase reaction also favoured certain products, and the reaction pressure improved the mass transfer of reactants in the solution. This study conducted the reactions at room temperature and confirmed the catalyst was primarily active in the visible light region. Therefore, the reaction was considered to be plasmon-mediated and driven by visible-light illumination. We observed correlations between the LSPR peaks and in situ temperature; however, this did not confirm the exact mechanism of plasmon decay. With the addition of Pt, the plasmon decay mechanism shifted towards energy transfer and hot-electron generation. Further investigation would provide insight into this mechanism. Increasing Pt loading and reaction pressure affected the colloidal stability. The aggregation presented a limitation on the parameters that could be investigated. An alternative capping agent such as polyvinylpyrrolidone (PVP) and cetyltrimethylammonium bromide (CTAB) could improve the colloidal stability. In order to create a scalable synthesis, catalyst immobilisation on a support or encapsulation in microparticle frameworks could help overcome this limitation. Further improvement in achieving high AQY could be approached by designing a modular reactor system which ideal wavelength illumination in a continuous mode.

Conclusions

We have developed a bimetallic plasmon-active catalyst consisting of a combination of the Ag–Pt system for photocatalysis. The incorporation of Pt into Ag NPrs as surface-limited alloys significantly improved the product yield and selectivity by altering the plasmon decay pathway towards energy transfer from Ag to Pt. The maximum yields of HCOOH (0.52 mmol g_cat⁻¹ with Ag₉₈Pt₂), HCHO (0.16 mmol g_cat⁻¹ with Ag_96.5Pt_3.5 and CH₃COOH (0.43 mmol g_cat⁻¹ with Ag₉₁Pt₉) suggested multi-electron hydrogenation and C–C coupling of *CH₃ and COO⁻ species. Notably, a peak AQY of 26.7% was observed at the in-dipole resonance (600 nm) for Ag₉₁Pt₉, highlighting the wavelength dependence in driving the reaction. Catalyst deactivation due to colloidal instability and long-term performance validation remains a challenge. Therefore, future work on improving catalyst stability and investigating surface species will provide deeper insight on the viability for scale-up. These findings highlight the role of designer plasmonic catalysts in enhancing photocatalytic CO₂ conversion. The results demonstrate that careful tuning of catalyst composition and morphology significantly improve the light-to-energy transfer and product selectivity.

Experimental

Chemicals

Silver nitrate (AgNO₃), trisodium citrate dihydrate, sodium borohydride (NaBH₄), poly(4-styrenesulfonic acid)sodium salt (PSSS, 1000 kDa), L-ascorbic acid, potassium tetrachloroplatinate(II) (K₂PtCl₄), formic acid (reagent grade), acetic acid (reagent grade) and methanol (ACS reagent) were purchased from Sigma Aldrich and used as received. CO₂ (ultra-high purity) and H₂ (ultra-high purity) was purchased from BOC. MilliQ water (18.2 mΩ) was used for all the experiments.

Synthesis of Ag NPr and Ag_100−xPt_x nanoparticles

Ag seed synthesis. 5 mL of 2.5 mM trisodium citrate, 0.3 mL of 10 mM NaBH₄ and 0.25 mL of 500 mg L⁻¹ PSSS were stirred until mixed under ambient conditions. 5 mL AgNO₃ was added at a rate of 2 mL minute⁻¹ under stirring. The Ag seed solution was aged at room temperature without protection from light for 4–6 hours.

Ag NPr synthesis. 75 μL of 10 mM ascorbic acid, 5 mL Milli-Q H₂O and 47 μL Ag seed solution were mixed under ambient conditions. Under stirring, 3 mL AgNO₃ was added at a rate of 1 mL minute⁻¹. After 2 additional minutes of stirring, 5 mL of 25 mM trisodium citrate was added. The sample was mixed for another 2 minutes before the synthesis was complete.

Ag_100−xPt_x synthesis. A chosen volume of Ag NPr solution was stirred in dark. The molar mass of Ag present in this solution was calculated based upon the synthesis process. The appropriate mass of Pt precursor for the desired Pt loading was then calculated (e.g. for 500 mL of Ag₉₁Pt₉, 3.64 mg K₂PtCl₄ was needed). The Pt precursor was dissolved in a small volume of Milli-Q H₂O before being added to the stirring Ag NPr solution. The volume of H₂O used was arbitrarily set to 5 mL but did not affect the synthesis.

Characterisations

UV-vis extinction spectra of the samples were obtained using a PerkinElmer Lambda 365 UV-vis spectrophotomer. Samples to be imaged in STEM were centrifuged three times at 10k RPM for 30 minutes. The supernatant was removed and replaced with Milli-Q H₂O. 1.4 μL of sample solution was drop cast on plasma-cleaned carbon-film-coated Cu grids. STEM imaging was captured on both the FEI Tecnai G2 T-20 TWIN TEM and F20 S-TWIN TEM at Monash Centre for Electron Microscopy (MCEM). STEM-EDS was conducted on the F20 using a Bruker 30 mm² 123 eV windowless SDD and Quantax analysis system. ¹⁰⁷Ag and ¹⁹⁵Pt isotope concentrations were analysed in the bulk catalyst samples using Inductively Coupled Plasma – Mass Spectroscopy (Thermo iCAP TQe Triple Quadrupole). 2 mL of catalyst samples were diluted in 70% v/v HNO₃ followed by serial dilutions in 2% v/v HNO₃. Sample zeta potentials and hydrodynamic radius were measured and recorded by Malvern Nano Zetasizer using disposable folded capillary cells (DTS1070). XPS analysis of as-synthesised nanoparticles was conducted on Thermo Scientific K-alpha XPS (monochromated Al Kα, E_photon = 1487.6 eV, 1 × 10⁻⁹ torr). Scans were obtained using a pass energy of 50 eV with 0.1 eV resolution. Post-reaction samples were analysed on the Thermo Nesa Surface Analysis System (monochromated Al Kα, E_photon = 1487.6 eV). The spectra were charge corrected to adventitious carbon at 285 eV in the C 1s region using Avantage software. Ag 3d and P 4f regions were background corrected with Shirley algorithm. Ag⁰ and Pt⁰ peaks were fitted with a Doniac-Sunjic asymmetry model. The oxidation states (Pt²⁺), C 1s and O 1s regions were equipped with Gaussian–Lorentzian function in Origin software (2022b).

Photocatalytic CO₂ hydrogenation

The reaction was carried out in batch mode in a PARR (A239M) photoreactor with a K-type thermocouple inserted at the base. 5–10 mg of as-synthesised catalysts were concentrated, re-dispersed in 50 mL Milli-Q H₂O and added to the reaction vessel. The reactor was sealed, and H₂ was introduced for 15 min to flush the reactor. This was followed by venting the reactor and introducing CO₂ and H₂ (1 : 1) to the specified pressure. A 300 W Xe lamp (Zolix 300P) irradiated the sample through the photoreactor window. For visible light (380–780 nm) irradiation, a UV + IR cutoff filter was used. Liquid products were analysed using high-performance liquid chromatography (Agilent 1220 Infinity). An ion exclusion column (Rezex RHM-Monosaccharide H+ 8%) with refractive index detector was used to detect HCOOH and CH₃COOH). A C18 column (Xterra MS C18) with diode array detector was used to detect HCHO by derivatisation with Nash reagent. Gaseous products (CO, CO₂, CH₄ and H₂) were analysed using gas chromatography (Agilent MicroGC 990) equipped with a TCD detector. Wavelength-dependent studies were performed using the same photoreactor setup as NKT Supercontinuum source (SuperK Fianium FIU-15) with a LTTF Contrast VIS giving a bandwidth <2.5 nm.

Equations

where,

N_{formic acid} = 2

N_formaldehyde = 4

N_{acetic acid} = 8

Data availability

All data has been included in the manuscript and ESI.† Raw data may be provided upon request.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This research is supported by the Australian Research Council (ARC) Research Hub for Carbon Utilisation & Recycling (RECARB Hub) (IH220100012), Monash University, BASF Australia Ltd and Woodside Energy Technologies Pty Ltd. The authors would like to thank Woodside Monash Energy Partnership for financial support. The authors acknowledge the use of the instruments and scientific and technical assistance at the Monash Centre for Electron Microscopy, a node of Microscopy Australia. This research used equipment funded by the Australian Research Council grant LE110100223. The authors acknowledge the use of facilities within the Monash X-ray Platform. The authors acknowledge the facilities, and the scientific and technical assistance of the RMIT University's Microscopy & Microanalysis Facility, a linked laboratory of Microscopy Australia, enabled by NCRIS The authors acknowledge the assistance of Prof. Terence Turney and thank him for his advice and feedback. Prof. Stefan A. Maier acknowledges the Lee-Lucas Chair in Physics.

Notes and references

1. G. Garcia-Garcia, M. C. Fernandez, K. Armstrong, S. Woolass and P. Styring, ChemSusChem, 2021, 14, 995–1015.
2. W. Ahmad, G. Bhardwaj, R. Lakshman, P. Koley and A. Tanksale, Energy Fuels, 2023, 37, 19377–19399.
3. Y. Wang, E. Chen and J. Tang, ACS Catal., 2022, 12, 7300–7316.
4. M. Tahir and N. A. S. Amin, Int. J. Hydrogen Energy, 2017, 42, 15507–15522.
5. J. Zhao, B. Liu, L. Meng, S. He, R. Yuan, Y. Hou, Z. Ding, H. Lin, Z. Zhang, X. Wang and J. Long, Appl. Catal., B, 2019, 256, 117823.
6. G. Chen, R. Gao, Y. Zhao, Z. Li, G. I. N. Waterhouse, R. Shi, J. Zhao, M. Zhang, L. Shang, G. Sheng, X. Zhang, X. Wen, L.-Z. Wu, C.-H. Tung and T. Zhang, Adv. Mater., 2018, 30, 1704663.
7. S. Xue, C. Wei, M. Shen, X. Liang, J. Wang, C. Yang, W. Xing, S. Wang, W. Lin, Z. Yu, Y. Hou, J. C. Yu and X. Wang, Proc. Natl. Acad. Sci. U. S. A., 2024, 121, e2319751121.
8. D. Zeng, H. Wang, X. Zhu, H. Cao, W. Wang, Y. Zhang, J. Wang, L. Zhang and W. Wang, Appl. Catal., B, 2023, 323, 122177.
9. K. Tamaki, P. Verma, T. Yoshii, T. Shimojitosho, Y. Kuwahara, K. Mori and H. Yamashita, Catal. Today, 2023, 411–412, 113795.
10. R. Kuriki, K. Sekizawa, O. Ishitani and K. Maeda, Angew. Chem., Int. Ed., 2015, 54, 2406–2409.
11. L. F. Garay-Rodríguez and L. M. Torres-Martínez, J. Mater. Sci.: Mater. Electron., 2020, 31, 19248–19265.
12. W. Hou, W. H. Hung, P. Pavaskar, A. Goeppert, M. Aykol and S. B. Cronin, ACS Catal., 2011, 1, 929–936.
13. Z.-j Wang, H. Song, H. Pang, Y. Ning, T. D. Dao, Z. Wang, H. Chen, Y. Weng, Q. Fu, T. Nagao, Y. Fang and J. Ye, Appl. Catal., B, 2019, 250, 10–16.
14. D. Wu, K. Deng, B. Hu, Q. Lu, G. Liu and X. Hong, ChemCatChem, 2019, 11, 1598–1601.
15. M. Izadpanah Ostad, M. Niknam Shahrak and F. Galli, J. CO2 Util., 2021, 43, 101373.
16. N. Li, X. Liu, J. Zhou, W. Chen and M. Liu, Chem. Eng. J., 2020, 399, 125782.
17. Y. Meng, L. Zhang, H. Jiu, Q. Zhang, H. Zhang, W. Ren, Y. Sun and D. Li, Mater. Sci. Semicond. Process., 2019, 95, 35–41.
18. S. Unser, I. Bruzas, J. He and L. Sagle, Sensors, 2015, 15, 15684–15716.
19. Z. Zhang, W. Shen, J. Xue, Y. Liu, Y. Liu, P. Yan, J. Liu and J. Tang, Nanoscale Res. Lett., 2018, 13, 54.
20. C. Clavero, Nat. Photonics, 2014, 8, 95–103.
21. C. Hu, X. Chen, J. Jin, Y. Han, S. Chen, H. Ju, J. Cai, Y. Qiu, C. Gao, C. Wang, Z. Qi, R. Long, L. Song, Z. Liu and Y. Xiong, J. Am. Chem. Soc., 2019, 141, 7807–7814.
22. H. Yin, Z. Sun, K. Liu, A. A. Wibowo, J. Langley, C. Zhang, S. E. Saji, F. Kremer, D. Golberg, H. T. Nguyen, N. Cox and Z. Yin, Nanoscale Horiz., 2023, 8, 1695–1699.
23. P. Christopher, H. Xin and S. Linic, Nat. Chem., 2011, 3, 467–472.
24. I. García-García, E. C. Lovell, R. J. Wong, V. L. Barrio, J. Scott, J. F. Cambra and R. Amal, ACS Sustainable Chem. Eng., 2020, 8, 1879–1887.
25. U. Kreibig and M. Vollmer, Optical Properties of Metal Clusters, Springer Berlin/Heidelberg, Berlin, Heidelberg, Berlin, Heidelberg, 1995.
26. S. A. Maier, Plasmonics: Fundamentals and Applications, Imprint, Springer, New York, NY: Springer US, 1st edn, 2007.
27. A. Manjavacas, J. G. Liu, V. Kulkarni and P. Nordlander, ACS Nano, 2014, 8, 7630–7638.
28. K. Kolwas, Plasmonics, 2019, 14, 1629–1637.
29. A. G. M. da Silva, T. S. Rodrigues, J. Wang and P. H. C. Camargo, Chem. Commun., 2022, 58, 2055–2074.
30. W. Xie, K. Zhang, R. Grzeschik and S. Schlücker, Plasmonic Catalysis, 2021, pp. 71–108 DOI:10.1002/9783527826971.ch3.
31. V. V. Pham, T. Q. Nguyen, H. V. Le and T. M. Cao, Nanoscale Adv., 2024, 6, 2380–2389.
32. A. G. M. da Silva, T. S. Rodrigues, J. Wang, L. K. Yamada, T. V. Alves, F. R. Ornellas, R. A. Ando and P. H. C. Camargo, Langmuir, 2015, 31, 10272–10278.
33. S. Kattel, B. Yan, J. G. Chen and P. Liu, J. Catal., 2016, 343, 115–126.
34. E. S. Gutterød, A. Lazzarini, T. Fjermestad, G. Kaur, M. Manzoli, S. Bordiga, S. Svelle, K. P. Lillerud, E. Skúlason, S. Øien-Ødegaard, A. Nova and U. Olsbye, J. Am. Chem. Soc., 2020, 142, 999–1009.
35. F. L. Chan, G. Altinkaya, N. Fung and A. Tanksale, Catal. Today, 2018, 309, 242–247.
36. S. Bai, Q. Shao, Y. Feng, L. Bu and X. Huang, Small, 2017, 13, 1604311.
37. Z. He, Q. Qian, J. Ma, Q. Meng, H. Zhou, J. Song, Z. Liu and B. Han, Angew. Chem., Int. Ed., 2016, 55, 737–741.
38. U. Aslam, S. Chavez and S. Linic, Nat. Nanotechnol., 2017, 12, 1000–1005.
39. V. G. Rao, U. Aslam and S. Linic, J. Am. Chem. Soc., 2019, 141, 643–647.
40. D. Aherne, D. M. Ledwith, M. Gara and J. M. Kelly, Adv. Funct. Mater., 2008, 18, 2005–2016.
41. S.-C. Lin, C.-S. Hsu, S.-Y. Chiu, T.-Y. Liao and H. M. Chen, J. Am. Chem. Soc., 2017, 139, 2224–2233.
42. W. He, X. Wu, J. Liu, K. Zhang, W. Chu, L. Feng, X. Hu, W. Zhou and S. Xie, Langmuir, 2010, 26, 4443–4448.
43. J. Chen, B. Wiley, J. McLellan, Y. Xiong, Z.-Y. Li and Y. Xia, Nano Lett., 2005, 5, 2058–2062.
44. L. Ma, S.-J. Ding and D.-J. Yang, Dalton Trans., 2018, 47, 16969–16976.
45. C. Tiburski and C. Langhammer, ACS Photonics, 2023, 10, 253–264.
46. F. Merkoçi, J. Patarroyo, L. Russo, J. Piella, A. Genç, J. Arbiol, N. G. Bastús and V. Puntes, Mater. Today Adv., 2020, 5, 100037.
47. Y. Chen, Z. Lai, X. Zhang, Z. Fan, Q. He, C. Tan and H. Zhang, Nat. Rev. Chem., 2020, 4, 243–256.
48. Z. Cao, Q. Chen, J. Zhang, H. Li, Y. Jiang, S. Shen, G. Fu, B.-A. Lu, Z. Xie and L. Zheng, Nat. Commun., 2017, 8, 15131.
49. K. Qi, Y. Zhang, J. Li, C. Charmette, M. Ramonda, X. Cui, Y. Wang, Y. Zhang, H. Wu, W. Wang, X. Zhang and D. Voiry, ACS Nano, 2021, 15, 7682–7693.
50. E. Fidiani, G. Thirunavukkarasu, Y. Li, Y.-L. Chiu and S. Du, J. Mater. Chem. A, 2020, 8, 11874–11883.
51. S.-S. Chen, X.-X. Lin, A.-J. Wang, H. Huang and J.-J. Feng, Sens. Actuators, B, 2017, 248, 214–222.
52. T. Shao, D. Bai, M. Qiu, Y. Li, Q. Zhang, Z. Xue, S. He, D. Zhang and X. Zhou, J. Ind. Eng. Chem., 2022, 108, 456–465.
53. M. H. Ullah, K. Il and C.-S. Ha, Mater. Lett., 2006, 60, 1496–1501.
54. J. Gargiulo, M. Herran, I. L. Violi, A. Sousa-Castillo, L. P. Martinez, S. Ezendam, M. Barella, H. Giesler, R. Grzeschik, S. Schlücker, S. A. Maier, F. D. Stefani and E. Cortés, Nat. Commun., 2023, 14, 3813.
55. J. Bi, H. Cai, B. Wang, C. Kong and S. Yang, Chem. Commun., 2019, 55, 3943–3946.
56. S. Moret, P. J. Dyson and G. Laurenczy, Nat. Commun., 2014, 5, 4017.
57. Q. Jianping, T. Juntao, S. Jie, W. Cuiwei, Q. Mengqian, H. Zhiqiao, C. Jianmeng and S. Song, Electrochim. Acta, 2016, 203, 99–108.
58. G. Kumari, R. Kamarudheen, E. Zoethout and A. Baldi, ACS Catal., 2021, 11, 3478–3486.
59. D. Devasia, A. J. Wilson, J. Heo, V. Mohan and P. K. Jain, Nat. Commun., 2021, 12, 2612.
60. P. L. Redmond, X. Wu and L. Brus, J. Phys. Chem. C, 2007, 111, 8942–8947.
61. G. Kumari, X. Zhang, D. Devasia, J. Heo and P. K. Jain, ACS Nano, 2018, 12, 8330–8340.
62. M. T. Schaal, M. P. Hyman, M. Rangan, S. Ma, C. T. Williams, J. R. Monnier and J. W. Medlin, Surf. Sci., 2009, 603, 690–696.
63. M. Barbalinardo, G. Ori, L. Lungaro, G. Caio, A. Migliori and D. Gentili, J. Phys. Chem. C, 2024, 128, 16220–16226.
64. T. C. Prathna, N. Chandrasekaran and A. Mukherjee, Colloids Surf., A, 2011, 390, 216–224.
65. P. G. Jessop, F. Joó and C.-C. Tai, Coord. Chem. Rev., 2004, 248, 2425–2442.
66. C. Genovese, C. Ampelli, S. Perathoner and G. Centi, Green Chem., 2017, 19, 2406–2415.
67. C.-T. Lin, M.-H. Shiao, M.-N. Chang, N. Chu, Y.-W. Chen, Y.-H. Peng, B.-H. Liao, H. J. Huang, C.-N. Hsiao and F.-G. Tseng, Nanoscale Res. Lett., 2015, 10, 74.
68. J. Dankar, V. Rouchon, M. Rivallan, C. Pagis and M. El-Roz, ACS Appl. Mater. Interfaces, 2024, 16, 42210–42220.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ey00046g

---