Introduction

Urea, a critical nitrogen fertilizer that supports ~ 40% of global agricultural productivity, is recognized as one of the most essential chemicals1,2. In industry, urea production relies heavily on upstream ammonia synthesis via the Haber-Bosch process and hydrogen generation through steam methane reforming—both of which are associated with high energy consumption and considerable CO2 emissions3,4. Electrocatalytic urea synthesis utilizing nitrate (NO3−) pollutants—or NO3− obtained via sustainable plasma activation of N2—along with CO2 from anthropogenic activities, simultaneously mitigates environmental pollution and enables sustainable chemical production, offering a promising alternative strategy5,6,7,8,9,10,11. However, the intricate multistep reaction mechanism of urea electrosynthesis that involves a 16-electron transfer presents significant limitations in urea production and induces severe side reactions12,13,14. Consequently, the rational design and development of electrocatalysts with high performance is urgently required.

The C-N intermediate formation (e.g., OCNO*) and protonation are the key elementary steps in urea electrosynthesis7,15,16,17,18,19. Despite extensive efforts, simultaneously optimizing these substeps on a single catalyst remains highly challenging due to scaling relations that correlate the adsorption free energies of relevant C-N intermediates and their protonated species7,15,16,17,18,19. This correlation often leads to suboptimal urea production efficiency. Tandem catalysts, which spatially separate the C-N coupling and protonation steps onto distinct components and facilitate the transport of intermediates between them, offer a promising strategy to independently tailor the adsorption behavior of key species12,18. However, the remote distance between C-N coupling sites and protonation sites, combined with the significant steric hindrance of C-N intermediates, hinders efficient C-N intermediate transfer and subsequent protonation. As a result, the overall urea production efficiency often falls short of expectations. Therefore, the design of tandem urea electrosynthesis catalysts with a better reaction path to overcome the above challenges is highly desirable.

In this work, we propose an optimized tandem urea electrosynthesis pathway, leveraging dual spillover of the sterically unhindered CO* and H* intermediates. Palladium hydrides (PdHx) are paired with copper (Cu) to separate efficient CO2-to-CO* and NO3−-to-NO* conversions on PdHx and Cu, respectively. CO* spillover from PdHx surface to Cu facilitates the first C-N coupling (intermediate OCNO*), while H* spillover from PdHx lattice to Cu promotes the protonation of the resulting C-N intermediates. Our catalyst system achieved a high urea production rate (rurea) of 236.5 ± 8.9 mmol gcat−1 h−1 with a Faradaic efficiency (FEurea) of 62.6 ± 1.8% at – 0.7 V vs. RHE. To demonstrate practical applicability, we designed an enlarged two-electrode flow cell using PdHx/Cu catalysts as both the cathode and anode. This system enabled continuous urea electrosynthesis at the cathode and methanol-to-formate conversion at the anode, operating stably over 200 h with high efficiency (470.5 ± 30.1 mmol gcat−1 h−1 for urea and 2850.1 ± 51.1 mmol gcat−1 h−1 for formic acid). A techno-economic analysis and life cycle assessment for this route indicated consistent profitability and significantly reduced CO2 emissions, compared with these for the conventional route.

Results

Catalyst design

Current tandem catalysts for urea electrosynthesis typically consist of component (i) for efficient C-N coupling and component (ii) for facile C-N intermediate protonation (Fig. 1a). However, due to the restricted diffusion of these sterically hindered C-N intermediates (e.g., barrier: ~ 0.83 − 3.61 eV)20,21, the successful transfer of C-N intermediates from component (i) to component (ii) is infrequent, severely limiting subsequent C-N protonation and thus urea electrosynthesis12.

Fig. 1: Design conception.
Fig. 1: Design conception.
Full size image

a Schematic diagram of the conventional design conception for tandem urea electrosynthesis12,18. b Schematic diagram of our design conception for tandem urea electrosynthesis through dual CO* and H* spillover over the PdHx/Cu catalysts. The energy barrier values are based on the previous results20,21,23,24.

To address this limitation, we propose a tandem electrosynthesis pathway that leverages the efficient dual spillover of CO* and H* intermediates, which exhibit significantly reduced steric hindrance (Fig. 1b). In our design, metallic Cu with exposed (111) crystal planes is selected as component (ii) owing to its well-established selectivity in converting NO3− to NO* intermediates, along with its ability to accumulate NO* due to the high energy barrier associated with the NO* → N* step22. Subsequently, PdHx is identified as the optimal choice for component (i) based on its dual functionality in enabling two efficient and non-interfering spillover processes: (i) surface-mediated CO* migration (e.g., barrier: ~ 0.16 eV)23 and (ii) lattice-incorporated H* (de)intercalation (e.g., barrier: ~ 0.06 eV)24. In the resulting PdHx/Cu hybrid system, CO* species generated on Pd surfaces undergo rapid migration to adjacent Cu surfaces through a well-documented spillover mechanism25,26. Concurrently, PdHx phases serve as dynamic H* reservoirs, facilitating reversible (de)intercalation and subsequent H* spillover to neighboring Cu surfaces27. This design creates spatial proximity between spillover-derived CO* (COSp*) and NO* species, thereby significantly enhancing their collision probability and accelerating C-N coupling. Furthermore, the close proximity of spillover-derived H* (HSp*) to the resulting C-N intermediates (e.g., OCNO*) ensures high collision probability, enabling efficient protonation of C-N intermediates. By simultaneously optimizing both C-N coupling and protonation processes through this dual spillover-driven tandem pathway, we anticipate that PdHx/Cu hybrid catalysts exhibit high capability for significantly improving urea electrosynthesis performance.

Catalyst synthesis and characterization

Considering that Cu nanosheets offer a large surface area with exposed (111) crystal planes and that decorating them with PdHx nanoparticles can generate abundant PdHx–Cu heterogeneous interfaces28,29, we synthesized PdHx/Cu catalysts via a two-step wet-chemical method: (i) preparation of PdHx nanoparticles, and (ii) synthesis of Cu nanosheets with simultaneous in situ loading of PdHx. Through the inductively coupled plasma-mass spectrometry (ICP-MS), the weight loading of PdHx in PdHx/Cu catalysts was determined to be ~ 7.0 wt.% (Supplementary Table 1). The transmission electron microscopy (TEM) image revealed the uniform distribution of small nanoparticles (~ 3.81 nm) on the nanosheets (Fig. 2a). High-resolution TEM analysis revealed lattice spacings of 2.09 Å and 2.31 Å for the nanosheets and nanoparticles, respectively. The former corresponds to the (111) plane of bulk Cu28, while the latter was consistent with the (111) spacing of the reported PdHx (Fig. 2b)30,31. TEM elemental X-ray mapping revealed a concentrated distribution of Pd in the nanoparticles and a uniform distribution of Cu in the nanosheets (Fig. 2c). These results indicate the formation of PdHx/Cu hybrids.

Fig. 2: Characterizations of model catalysts.
Fig. 2: Characterizations of model catalysts.
Full size image

a TEM image of the PdHx/Cu catalysts. b High-resolution TEM image of the PdHx/Cu catalysts. c Elemental X-ray mapping of the PdHx/Cu catalysts. d XRD pattern of the PdHx/Cu catalysts. e XPS valence band of PdHx/Cu and Pd/Cu. f FT-IR spectra of PdHx/Cu and Pd/Cu. g 1H ssNMR spectra of PdHx/Cu and Pd/Cu.

To get more evidences, multiple spectroscopic analyses were carried out and similar catalysts with Pd nanoparticles loaded on Cu nanosheets (Pd/Cu, Supplementary Fig. 1) were also prepared for comparison. The powder X-ray diffraction (XRD) pattern of PdHx/Cu was indexed to standard PdHx (JCPDS #87-0641) and Cu (JCPDS #85-1326) phases (Fig. 2d). The observed diffraction peak downshift of PdHx/Cu relative to that of Pd/Cu is due to lattice expansion resulting from H intercalation in PdHx31. The X-ray photoelectron spectroscopy (XPS) results indicated that Pd and Cu in both PdHx/Cu and Pd/Cu are mainly in the metallic state (Supplementary Fig. 2). The Pd 3 d peak of the PdHx/Cu shift to higher binding energies by 0.3 eV compared with that of the Pd/Cu, demonstrating the electronic Pd−H interaction in PdHx/Cu32. XPS valence-band spectra (Fig. 2e) revealed a narrower bandwidth for PdHx/Cu compared to Pd/Cu, consistent with characteristic features of Pd hydride systems31. The appearance of ~ 6 and ~ 9 eV shake-up satellites in PdHx/Cu indicates strong hybridization between Pd 4 d and H 1 s orbitals33. These results support the formation of PdHx/Cu hybrids. The Fourier transform-infrared (FT-IR) transmission spectra were performed to examine the vibration of Pd−H in PdHx lattice (Fig. 2f). Four new characteristic peaks located at 893, 1025, 1079, and 1130 cm−1 were observed for PdHx/Cu compared with that for Pd/Cu. These peaks match well with those of the previously reported ones for Pd hydrides32, confirming the successful formation of PdHx in PdHx/Cu. The 1H solid-state nuclear magnetic resonance (ssNMR) analysis revealed the appearance of a new peak at ~ 26 ppm for PdHx/Cu (Fig. 2g), which was absent for Pd/Cu. This again confirms the existence of a Pd−H bond in PdHx34. CO stripping voltammetry further excluded the formation of Pd-Cu alloys on PdHx/Cu surface, as reflected by the presence of two typical CO stripping peaks at ~ 0.52 and ~ 0.76 V vs. RHE, consistent with those of PdHx and Cu benchmarks (Supplementary Fig. 3)28,35. All the above results confirm the successful synthesis of the PdHx/Cu catalysts.

Catalytic evaluation

The catalytic performance for urea electrosynthesis from CO2 and NO3− was investigated using a three-electrode flow cell (Supplementary Fig. 4). To better evaluate the performance of PdHx/Cu, we then synthesized the Cu catalysts through similar synthetic strategy without addition of PdHx (Supplementary Fig. 5). Linear sweep voltammetry (LSV) revealed that the PdHx/Cu catalysts exhibited high urea electrosynthesis activity, as evidenced by the significantly higher current densities in the presence of both CO2 and NO3− compared to environments missing either or both of these reactants (Supplementary Fig. 6). The urea production yield was driven by chronoamperometric conditions (applied potential range: − 0.4 to − 0.8 V vs. RHE) and the catalytic current density was extracted from the chronoamperometry profiles (Supplementary Fig. 7). A 95% on-the-fly iR compensation method was applied during these measurements36, and neither the uncompensated nor the compensated chronoamperometry profiles exhibited potentiostat oscillations (Supplementary Fig. 8), confirming the appropriateness of the compensation protocol. The liquid and gas products were quantified using 1H NMR spectroscopy and gas chromatography (GC). The PdHx/Cu catalyst exhibited optimal urea electrosynthesis performance at − 0.7 V vs. RHE, achieving a Faradaic efficiency (FEurea) of 62.6 ± 1.8% and a mass-normalized urea production rate (rurea) of 236.5 ± 8.9 mmol gcat−1 h−1 (Fig. 3a and see 1H-NMR quantification details in Supplementary Fig. 9). These urea product quantification was further examined via spectrophotometric methods (see details in Methods section)4,12, which was consistent with the results obtained from 1H-NMR (Supplementary Fig. 10). Under identical conditions, the individual Cu showed significantly lower performance, with an rurea of 19.8 ± 1.5 mmol gcat−1 h−1 and FEurea of 17.3 ± 0.8% (Fig. 3b and Supplementary Fig. 11). For both catalysts, no NO2−, formamide (HCONH2), N2, formate (HCOO−) and C2+ products were detected. These comparative results unequivocally demonstrate that tandem PdHx/Cu catalysts outperform Cu in catalytic activity. After supplementation with 15NO3−, the products exhibited 1H NMR signals nearly identical to those of the 15N-urea standard and these signals disappeared when operated at open-circuit potential (OCP) (Supplementary Fig. 12). Similarly, when 13CO2 was supplied, the 13C NMR spectrum of the products matched that of 13C-urea and vanished at OCP (Supplementary Fig. 12). These results provide strong evidence that the N in urea is derived from NO3− in the electrolyte, while the C originates from the fed CO2. The PdHx/Cu catalysts also demonstrated long-term stability in urea electrosynthesis (Fig. 3c), maintaining consistent catalytic activity over an extended period of 200 h.

Fig. 3: Catalytic evaluation.
Fig. 3: Catalytic evaluation.
Full size image

a FE distributions of the products and corresponding rurea for PdHx/Cu at different potentials. b Comparisons of urea electrosynthesis performance of PdHx/Cu and Cu. c Durability and FEurea of PdHx/Cu through chronopotentiometry at − 0.7 V vs. RHE. d FE distributions of the products and corresponding j for PdHx/Cu(+)||PdHx/Cu(−) configuration at different Ecell values. e Continuous urea production at the cathode and formic acid production at the anode for 200 h. Every day, an appropriate amount of NO3− was added to the electrolyte. f Calculated cost for urea electrosynthesis in cthe athode over the flow cell with PdHx/Cu(+)||PdHx/Cu(−) electrode configuration at different Ecell. g Calculated revenue and cost for formate generation and methanol consumption in anode over the flow cell with PdHx/Cu(+)||PdHx/Cu(−) electrode configuration at different Ecell. Error bars stand for the standard deviation of three independent measurements.

To systematically assess catalytic performance, we conducted a comparative analysis of key performance parameters (rurea, FEurea, and catalytic durability) of the PdHx/Cu catalysts, benchmarking against state-of-the-art urea electrosynthesis catalysts reported in the literature. As illustrated in Supplementary Table 2, these catalysts exhibit high rurea, FEurea, and long-term stability, outperforming all recently reported representative electrocatalysts12,37,38,39,40,41. These results highlight the significant advancement achieved by the tandem PdHx/Cu system for urea electrosynthesis.

To evaluate the practical potential of the PdHx/Cu catalysts, we scaled up a two-electrode PdHx/Cu(+)||PdHx/Cu(−) flow cell system. The methanol oxidation reaction (MOR) was employed as the anodic reaction instead of the oxygen evolution reaction (OER), owing to the high MOR catalytic activity of PdHx/Cu (Supplementary Fig. 13), the significantly lower thermodynamic potential of MOR compared to OER, and the possibility for simultaneous production of valuable chemicals such as formate42,43. To ensure optimal conditions for MOR, the catholyte was composed of 1.0 M KOH and 0.1 M KNO3 and continuously purged with CO2 at a flow rate of 20 sccm. The anolyte consisted of 1.0 M KOH and 2.0 M methanol. It is noted that the presence of methanol had a negligible effect on the reported electrode potentials throughout the electrocatalytic measurements (Supplementary Fig. 14). Using this optimized system (Supplementary Fig. 15), we achieved the highest FEurea of 66.6 ± 1.3% and FEformate of 88.6 ± 1.4% for formate under cell voltages (Ecell) ranging from 2.2 to 2.8 V (Fig. 3d). Remarkably, the electrolyzer demonstrated long-term stability over 200 hours of continuous operation (Fig. 3e), yielding a urea production rate of 470.5 ± 30.1 mmol gcat−1 h−1 (see 1H-NMR quantification details in Supplementary Fig. 16) and a formate production rate of 2850.1 ± 51.1 mmol gcat−1 h−1. The urea quantification through spectrophotometric methods further confirmed the obtained urea production performance (Supplementary Fig. 17). To the best of our knowledge, this urea production efficiency is comparable to the best results reported to date (Supplementary Table 2).

After 200 hours of continuous operation, comprehensive post-mortem characterizations of the cathodic PdHx/Cu catalysts were performed using XRD, TEM, XPS, FT-IR, and 1H ssNMR (Supplementary Fig. 18). The analyses revealed no detectable structural or compositional changes, demonstrating long-term catalyst durability during cathodic urea electrosynthesis. In contrast, characterizations of the anodic PdHx/Cu catalysts through XRD, FT-IR, 1H ssNMR, XPS, and Raman spectroscopy showed complete lattice H* de-intercalation from PdHx and concomitant surface oxidation of Cu during methanol electrooxidation (Supplementary Fig. 19). This transformation led to the in situ formation of catalytically active and stable Pd/CuOx species, which maintained long-term activity for the anodic reaction−a phenomenon consistent with previous reports44,45,46. Importantly, ICP-MS analysis of the post-reaction electrolyte revealed minimal metal leaching, with total Pd and Cu concentrations measured at 2.8 ppm (Supplementary Fig. 20). This corresponds to a catalyst weight loss of only 0.42 wt.%, effectively ruling out catalyst dissolution as a significant degradation pathway. Throughout the 200-hour electrosynthesis, no detectable crossover of NO3− or HCOO− from the catholyte to the anolyte through the anion-exchange membrane (AEM) was observed (Supplementary Figs. 21, 22). SEM and atomic force microscopy images of the AEM before and after 200 h of continuous electrolysis through our two-electrode flow cell system revealed no significant changes in the morphology of AEM (Supplementary Fig. 23), confirming its structural stability throughout the electrosynthesis. Owing to the hydrophobic treatment of the carbon cloth gas diffusion electrode (GDE), no cathode flooding occurred throughout the 200-hour electrosynthesis period (Supplementary Fig. 24). Correspondingly, the dissolution of CO2 into the electrolyte was significantly suppressed, which maintained a nearly constant electrolyte pH and effectively prevented salt precipitation throughout the 200-hour electrosynthesis process (Supplementary Figs. 25, 26). These findings underscore the potential robustness of the flow cell reactor based on the PdHx/Cu(+)||PdHx/Cu(−) configuration for practical applications. While the reported results demonstrate promising potential, both the efficiency and operational stability of the current system remain below the threshold required for practical application. As a result, challenges such as ion crossover through the AEM, membrane stability, electrode flooding, pH variation, and salt precipitation did not pose immediate concerns under the present experimental conditions. Nevertheless, these issues are expected to become critically important as the technology progresses toward real-world implementation and therefore warrant further investigation.

We then performed a preliminary economic assessment of the system’s potential profitability using established models from prior researches25,47. The analysis incorporated both the total cost of urea electrosynthesis and the net profit from formate co-production (see Supplementary Note 1 for full methodology). The cost estimation for urea synthesis included capital investment, balance of plant, installation, maintenance, raw materials, electricity consumption, and product purification, evaluated over cell voltages (Ecell) ranging from 2.2 to 2.8 V. At an electricity price of US$0.01 kWh−1 (reflecting current rates in the Middle East), the estimated production costs slightly exceeded the current market price of urea (~US$650 per ton)48, with a minimum achievable cost of US$935.7 per ton (Fig. 3f and Supplementary Table 3). Notably, the co-production of formate yielded net profits between US$3029.4 and US$3561.5 per ton of urea produced—calculated as formate revenue minus methanol consumption costs (Fig. 3g and Supplementary Table 4)—thereby rendering the overall process economically viable.

We then performed the preliminary evaluation of the carbon footprint for our co-production system, accounting for the fate of biogenic carbon (see details in Supplementary Note 2). When generated urea is applied as a fertilizer in agriculture—where CO2 is ultimately released during end use—the electrochemical pathway yields life-cycle emissions of 0.923 kg CO2 per kg of urea, significantly lower than the 1.8 kg CO2 per kg of urea from the conventional production route12. These results highlight the strong potential of the PdHₓ/Cu-based flow electrolyzer system for industrial-scale implementation, offering a sustainable and profitable approach for urea synthesis.

Catalytic contribution identification

The enhanced performance of PdHx/Cu in comparison to that of Cu can be attributed to the following factors: (i) PdHx/Cu exhibits increased active sites owing to morphological change; (ii) the PdHx itself demonstrates high catalytic activity for urea electrosynthesis; (iii) the Cu itself serves as the primary contributor due to the electronic metal-support interaction; (iv) the interfacial PdHx−Cu sites show pronounced catalytic synergy for urea generation; and (v) effective CO* and H* spillover promotes C-N intermediate formation and protonation for tandem urea electrosynthesis.

To determine whether the improved urea electrosynthesis activity of PdHx/Cu originates from an increase in catalytic sites, we compared the physical surface areas of PdHx/Cu and pure Cu using double-layer capacitance (Cdl) measurements derived from CV. As shown in Supplementary Fig. 27, the nearly identical Cdl values (82.6 mF vs. 84.4 mF) indicate comparable physical surface areas between the two catalysts. This similarity can be attributed to the low weight loading and small size (~ 3.81 nm) of the PdHx nanoparticles, which contribute little additional surface area, combined with the partial blockage of Cu sites during deposition that may even result in a net reduction. These findings rule out the possibility that increased physical surface area is responsible for the enhanced activity. As shown in Supplementary Fig. 28, PdHx/Cu (53.0 m2 g−1) and Cu (51.7 m2 g−1) showed the close Brunauer–Emmett–Teller (BET) surface areas, again confirming the above fact49. Control experiments using PdHx nanoparticles (~ 3.70 nm) on reduced graphene oxide (PdHx/rGO, Supplementary Fig. 29) showed significantly lower urea electrosynthesis activity (Supplementary Fig. 30), indicating the PdHx itself is not the primary active component. We then probed whether Cu itself dominates the catalysis by selectively poisoning PdHx with triphenylphosphine (PPh3)50,51. If Cu itself were the main active phase, the activity should remain largely unaffected. Instead, a sharp decline in urea production was observed (Supplementary Fig. 31), demonstrating that the enhanced performance does not result from independent contributions from either component.

The catalytic role of interfacial sites might also dominate the urea electrosynthesis performance of PdHx/Cu. If this were the case, a positive linear correlation should be observed among PdHx loading, the density of PdHx–Cu interfacial sites, and urea formation activity (Supplementary Fig. 32). To test this hypothesis, we investigated the relationship between PdHx loading and rurea (Supplementary Figs. 33, 34), with particular emphasis on low loading levels to minimize interference from particle aggregation. Interestingly, rurea reached a saturation point at a PdHx loading of 7.0 wt.%, indicating that the capacity of the Cu domains to accommodate spillovered H* (HSp*) and CO* (COSp*) species was likely maximized at this level. These results clearly demonstrate that the enhanced urea electrosynthesis performance of PdHx/Cu is not primarily attributable to the PdHx–Cu interfacial sites. After ruling out the above factors (i–iv), the improvement is most plausibly attributed to a tandem mechanism involving CO* and H* spillover.

Experimental evidence for CO* spillover

To establish conclusive evidence for CO* spillover and identify its role, we systematically designed and carried control experiments. We first performed CO temperature-programmed desorption (CO-TPD) measurements for PdHx/rGO, Cu and PdHx/Cu (Supplementary Fig. 35). PdHx displayed a TPD peak at ~ 509 °C, corresponding to CO* desorption from Pd25, whereas Cu delivered a TPD peak at ~ 160 °C, corresponding to CO* desorption from Cu52. Comparatively, PdHx/Cu presented a similar CO* desorption peak to that of PdHx (513 °C) and a sharply increased CO* desorption peak compared to that of bare Cu (165 °C), revealing the strong probability for the occurrence of CO* spillover from PdHx to Cu in PdHx/Cu during urea electrosynthesis. In situ differential electrochemical mass spectrometry (DEMS) was employed to monitor the m/z = 28 signal to investigate CO* generation behavior on PdHx/rGO, Cu, and PdHx/Cu catalysts during reaction in 0.2 M KHCO3 electrolyte fed with CO2 (Fig. 4a). The weak CO* signal on Cu indicates its low CO* coverage, consistent with previous reports53. In contrast, the exponentially enhanced CO* signal of PdHx/Cu − PdHx/rGO demonstrated a substantial increase in CO* coverage on the Cu component, providing direct evidence for CO* spillover from PdHx to Cu as an additional CO* supply pathway on Cu. Further validation was obtained by PPh3 treatment, which selectively blocked PdHx-to-Cu CO* transfer50,51, resulting in near-complete suppression of the CO* signal. CO-stripping cyclic voltammetry (CV) was performed on PdHx/rGO, PdHx/Cu, and Cu catalysts in 0.2 M KHCO3 electrolyte to investigate CO* oxidation behavior (Fig. 4b). PdHx/rGO exhibited a CO* oxidation peak at ~ 0.89 V vs. RHE, consistent with reported Pd-based catalysts35, while Cu showed a peak at ~ 0.52 V vs. RHE, in agreement with literature28. Notably, the CO* oxidation peak intensity at ~ 0.52 V vs. RHE was significantly enhanced for PdHx/Cu compared to Cu, demonstrating efficient CO* spillover from PdHx to Cu. This conclusion was further supported by the sharp decrease in these CO* oxidation peaks at ~ 0.89 and ~ 0.52 V vs. RHE after PPh3 treatment, which blocked the PdHx and inhibited PdHx-to-Cu CO* spillover50,51. These results provide clear clues for active CO* spillover from PdHx to Cu in PdHx/Cu.

Fig. 4: Experimental evidence for CO* spillover.
Fig. 4: Experimental evidence for CO* spillover.
Full size image

a In situ DEMS investigations of various model catalysts during urea electrosynthesis without NO3− feeding at − 0.7 V vs. RHE. The cyan curve represents the DEMS signal of Cu. The blue curve represents the difference in the DEMS signals of PdHx/Cu and PdHx/rGO. The gray curve represents the DEMS signal of PPh3-treated PdHx/Cu. The presented ion current data represent raw signals without any post-treatment. b CO-stripping CVs of Cu, PdHx/Cu and PPh3-treated PdHx/Cu in 0.2 M KHCO3 electrolyte at a scan rate of 20 mV s−1. CO(g) was pre-adsorbed from the CO-saturated electrolyte at − 0.1 V vs. RHE for 2 min. In situ ATR-SEIRAS spectra for the (c) Cu and (d) PdHx/Cu catalysts during reaction at the potential range of − 0.4 to − 0.7 V vs. RHE in KHCO3 electrolyte only fed with CO2. e In situ ATR-SEIRAS spectra for the PdHx/Cu catalysts at − 0.7 V vs. RHE in KHCO3 electrolyte with different feedings.

To obtain direct evidence, in situ attenuated total reflectance-surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS) was employed to monitor the reaction intermediates on PdHx/rGO, PdHx/Cu, and Cu catalysts during the reaction in 0.2 M KHCO3 electrolyte (Fig. 4c–e). When only CO2 was introduced, PdHx/rGO exhibited a typical peak corresponding to CO* vertically adsorbed on the Pd-Pd bridge sites (PdHx−CObridge, 1958 cm−1)54, which emerged at − 0.4 V vs. RHE and increased rapidly with negative potential shift (Supplementary Fig. 36). This observation confirms the facile potential-dependent electrochemical reduction of CO2 to CO* on PdHx. In contrast, Cu catalysts displayed only a weak peak for CO* adsorbed on atop Cu sites (Cu−COatop, 2076 cm−1)54 within the same window (Fig. 4c), indicating sluggish CO2-to-CO* conversion. Notably, in the PdHx/Cu system, the PdHx−CObridge peak also showed significant potential-dependent enhancement, reaffirming efficient CO* generation on PdHx (Fig. 4d). Strikingly, a much more pronounced Cu−COatop signal was observed even at minimal overpotentials, providing direct spectroscopic evidence for rapid CO* spillover from PdHx to Cu.

When CO2 and NO3− were co-fed, the prominent Cu−COatop peak disappeared, while a new characteristic peak corresponding to OCNO* adsorbed on Cu (Cu−OCNO) emerged at 2151 cm−1 (Fig. 4e)17. Subsequent isotopic switching from 12CO to 13CO in the feed resulted in the disappearance of both PdHx−12CObridge and Cu−O12CNO peaks, with concomitant appearance of redshifted peaks at 1905 cm−1 and 2110 cm−1, respectively. The measured frequency shifts between the original and newly formed peaks showed excellent agreement with the expected 12C/13C isotopic shift values55,56,57, confirming that the new peaks should be assigned to PdHx−13CObridge and Cu−O13CNO. These findings strongly confirm that CO* adsorbed on Cu originates exclusively from PdHx via spillover, and the spillovered COSp* actively participates in the first C-N coupling to form OCNO*. Identical spectral results were obtained from the cathodic PdHx/Cu catalysts in the PdHx/Cu(+)||PdHx/Cu(−) flow cell system with 1.0 M KOH electrolyte (Supplementary Fig. 37), thereby confirming the effective operation of the CO* spillover mechanism for urea electrosynthesis in this two-electrode system.

Experimentalevidence for H* spillover

To investigate the occurrence and efficiency of H* spillover, in situ characterizations were employed to monitor H* species on the catalyst surface during the reaction. Electrochemical impedance spectroscopy (EIS) was employed to quantify the amount of H* on Cu in the PdHx/Cu system (Supplementary Fig. 38). The EIS spectra, exhibiting two semicircles, were fitted using a dual-parallel equivalent circuit model (inset, Supplementary Fig. 38)58; the corresponding fitted parameters are summarized in Supplementary Table 5. The solution resistance (Rs) showed minor variations due to molecular fluctuations in the electrolyte. The first parallel element (T and R1) corresponds to the charge-transfer kinetics of urea electrosynthesis59. The consistently low R1 values across all catalysts indicate high conductivity and rapid charge-transfer capability. The second parallel element (C and R2) characterizes H* adsorption on the catalyst surface, with R2 and C representing the adsorption resistance and pseudocapacitance, respectively58. The H* adsorption charge (Q) was determined by integrating the H* adsorption capacitance (C) as a function of potential (η) derived from Nyquist plots (Fig. 5a)60. Control experiments were performed using bare Cu and PdHx/rGO. Comparative analysis revealed a 6.2-fold increase in the differential H* adsorption charge (Q[PdHx/Cu] − Q[PdHx/rGO] = 3131 μC) relative to that of bare Cu (Q[Cu] = 502 μC), demonstrating a significant enhancement in H* generation on the Cu surface of the PdHx/Cu system. These results provide direct evidence for efficient PdHx-to-Cu H* spillover, which ensures a substantial H* supply on Cu.

Fig. 5: Experimental evidence for H* spillover.
Fig. 5: Experimental evidence for H* spillover.
Full size image

a Plots of C vs. η of the Cu, PdHx/rGO and PdHx/Cu catalysts. b CVs of the Cu, PdHx/rGO and PdHx/Cu catalysts in 0.2 M KHCO3 electrolyte at a scan rate of 20 mV s−1. In situ ATR-SEIRAS spectra for the (c) Cu and (d) PdHx/Cu catalysts during reaction at the potential range of − 0.4 to − 0.7 V vs. RHE in KHCO3 electrolyte without feeding. e In situ ATR-SEIRAS and FT-IR spectra for the PdHx/Cu catalysts at − 0.7 V vs. RHE in KHCO3/H2O or KDCO3/D2O solution. f In situ ATR-SEIRAS for the PdHx/Cu catalysts at − 0.7 V vs. RHE in KDCO3/D2O solution with different feedings.

Further investigation into the hydrogen adsorption kinetics on Cu and PdHx/Cu during the reaction was conducted via Tafel analysis of log R2 versus applied potential (Supplementary Fig. 39)59, derived from EIS data (Supplementary Fig. 38 and Table 5). The Cu exhibited showed a large EIS-derived Tafel slope (136.5 mV dec−1), indicative of its limited H* adsorption kinetics. In contrast, PdHx/Cu showcased a much lower EIS-derived Tafel slope (90.1 mV dec−1). This finding revealed that the intrinsic kinetic bottleneck of H* adsorption on the Cu side was overcome with sufficient PdHx-to-Cu hydrogen spillover functioning as a much faster pathway for H* supply. CV investigations were then performed on the PdHx/rGO, PdHx/Cu and Cu catalysts at a potential range from 0 to 0.3 V vs. RHE without the feeding of CO2 and NO3− (Fig. 5b), enabling the in situ monitoring of H* during the reaction. PdHx/rGO exhibited an oxidation CV peak at ~ 0.05 V vs. RHE, indicative of the typical desorption behavior of lattice H* in PdHx27. Conversely, no CV peak was observed for Cu within this potential range, consistent with the intrinsically limited availability of adsorbed H* on the Cu surface. In comparison, PdHx/Cu showed a similar lattice H* desorption signal at ~ 0.05 V vs. RHE, along with a sharply raised oxidation CV peak at 0.26 V vs. RHE. Previous reports and various control experiments (Supplementary Figs. 40–42) confirmed that this oxidation CV peak corresponded to the H* desorption on Cu rather than OH* or metal/hydroxide redox behavior28,61,62. These observations provide evidence for efficient H* spillover from PdHx lattice to Cu surface and thus sufficient spillovered HSp* for desorption.

To demonstrate H* spillover mechanism, we conducted in situ ATR-SEIRAS measurements on PdHx/Cu and Cu catalysts in 0.2 M KHCO3/H2O electrolyte (Fig. 5c, d). Upon sweeping the potential from − 0.4 to − 0.7 V vs. RHE, the Cu catalyst exhibited weak vibrational signature of adsorbed H* (Cu–H, 2102 cm−1)63, consistent with its inherently low hydrogen adsorption capability. In stark contrast, PdHx/Cu displayed an intense Cu–H signal even at minimal potential, indicating a H* spillover process that facilitates rapid H* transfer from PdHx to Cu28. Further mechanistic validation was achieved through isotopic exchange experiments (Fig. 5e), where replacing KHCO3/H2O with KDCO3/D2O resulted in the complete disappearance of the Cu–H signal and the emergence of a redshifted Cu–D signal at 1500 cm−1, matching the theoretical H/D isotopic shift ratio (1/√2)64,65. Complementary in situ FT-IR transmission spectra confirmed the concurrent depletion of the lattice Pd–H signals (893, 1025, 1079 and 1130 cm−1) in PdHx and the formation of lattice Pd–D signals (638, 732, 771 and 807 cm−1), directly linking the HSp* to the dynamic (de)intercalation of lattice H* in PdHx24. These spectroscopic observations provide unequivocal evidence for a lattice H* spillover mechanism, wherein PdHx acts as an H* reservoir, supplying lattice H* to Cu via a (de)intercalation cycle and subsequent spillover.

Following the introduction of CO2 or NO3−, Cu–D vibrational signal remained largely unaltered (Fig. 5f), indicating that the spillovered D* (or H* under non-isotopic conditions) on Cu less participates in the electroreduction of CO2 or NO3− during urea electrosynthesis. Notably, when both CO2 and NO3− were introduced simultaneously, Cu–D signal rapidly diminished and Cu–OCNO signal came into being. Further mechanistic insight was gained by adding the D* trap tertiary butanol (t-BuOH) to the electrolyte66, which resulted in a marked increase in the Cu–OCNO signal. These observations provide compelling evidence that the HSp* actively facilitates the protonation of key C-N coupling intermediates (e.g., OCNO*). The cathodic PdHx/Cu catalysts in the PdHx/Cu(+)||PdHx/Cu(−) flow cell system exhibited identical spectral results in 1.0 M KOH/H2O electrolyte (Supplementary Fig. 43). This consistency provides direct evidence for the efficacy of the H* spillover mechanism on these catalysts during urea electrosynthesis within this two-electrode configuration. Collectively, these findings confirm the occurrence of a tandem urea electrosynthesis mechanism over PdHx/Cu as initially hypothesized, in which surface CO* and lattice H* spillover from PdHx-to-Cu drive first C-N coupling and protonation of the resulting C-N intermediates during urea electrosynthesis.

Mechanistic insights

To elucidate the role of the tandem mechanism in urea electrosynthesis, we first experimentally identified the reaction pathway for urea electrosynthesis over the Cu and PdHx/Cu catalysts through online DEMS. When co-fed with CO2 and NO3−, signals at m/z = 58 and 44 were detected (Supplementary Fig. 44). Control experiments with only CO2 or only NO3− showed no significant signals at these m/z values (Supplementary Fig. 45), confirming that these fragments do not originate from the electroreduction of CO2 or NO3− alone. Therefore, the signals at m/z = 58 and 44 can be attributed to the key OCNO* and OCNH2* intermediates, respectively, during urea electrosynthesis. Taken together, the following reaction pathways were identified: (i) first C-N coupling (CO* + NO* → OCNO*) and protonation to OCNH2*; (ii) second C-N coupling (OCNH2* + NO* → OCNO*) and protonation to urea16. Accordingly, we examined the free energy profiles of key reaction steps on the Cu model. The CO2 reduction pathway (CO2(g)-to-CO*) exhibits a high barrier of 0.46 eV in the step of CO2(g) → COOH* (Fig. 6a), making CO* formation energetically unfavorable. In contrast, NO3− reduction (NO3−(g)-to-NO*) is energetically favorable (Fig. 6b). However, the first C-N coupling (CO* + NO* → OCNO*) is severely limited by both the scarcity of CO* intermediates and a formidable coupling barrier of 0.70 eV (Fig. 6c). In addition, Cu model demonstrates stagnant water dissociation process (Supplementary Fig. 46), resulting in an insufficient supply of H*. Consequently, the protonation of the resulting OCNO* through the proton-coupled electron transfer (PCET) pathway (OCNO* → OCNOH* → OCN* → OCNH2*) face a notably high barrier of 0.70 eV at the OCN* → OCNH* step (Fig. 6c). These combined limitations in first C-N intermediate formation and protonation explain the experimentally observed negligible urea production on the Cu catalysts.

Fig. 6: Mechanistic insights.
Fig. 6: Mechanistic insights.
Full size image

a Calculated free energy diagrams of the CO2 reduction on Cu model. b Calculated free energy diagrams of the NO3− reduction on Cu model. c Calculated free energy diagrams of C-N coupling and protonation on Cu model. d) Calculated free energy diagrams of the CO2 reduction on the PdHx side of the PdHx/Cu model. e Calculated free energy diagrams of CO* spillover from PdHx surface to Cu surface on PdHx/Cu model. f Calculated free energy diagrams of the NO3− reduction on the Cu side of PdHx/Cu model. g Calculated free energy diagrams of C-N coupling and protonation on the Cu side of PdHx/Cu model. h Calculated free energy diagrams of H* spillover from PdHx lattice to Cu surface on PdHx/Cu model.

Accordingly, the free energy profiles for urea electrosynthesis were computed on the PdHx/Cu model (see Methods for calculation details and model construction). On PdHx/Cu surface, the PdHx domain enables energetically favorable CO2 reduction, producing vertically oriented CO* at surface site-1 (COSf-1*, Fig. 6d). Notably, a surface CO* spillover channel from PdHx (COSf-1*) → interface (COSf-2* and COSf-3*) → Cu (COSp*) with gradually weakened CO* adsorption exists (Fig. 6e). Through this channel, the PdHx-generated CO* can efficiently migrate to Cu sites with a low barrier of 0.26 eV. Concurrently, the Cu domain efficiently catalyzes NO3− reduction, generating abundant NO* intermediates (Fig. 6f). The spatial proximity of COSp* (from PdHx surface to Cu) and NO* (generated on Cu) enables first C-N coupling (COSp* + NO* → OCNO*) with a significantly reduced barrier of 0.30 eV (Fig. 6g). The PdHx further serves as an efficient H* reservoir: (i) lattice H* (HL1* and HL2*) undergoes facile de-intercalation and spillover to Cu sites, and (ii) protons from the electrolyte spontaneously intercalate into the PdHx lattice, maintaining a sustainable H* supply for Cu domain (Fig. 6h). Given the high energy barriers associated with HSp*-mediated CO2 or NO3− electroreduction (Supplementary Figs. 47, 48), the resulting OCNO* can be readily protonated by HSp* to form OCNH2* via a series of steps (OCNO* + HSp* → OCNOH* → OCNOH* + HSp* → OCN* → OCN* + HSp* → OCNH* → OCNH* + HSp* → OCNH2*). With a continuous supply of HSp* (from the PdHx lattice to Cu) and NO* (generated on Cu), the subsequent almost barrierless second C-N coupling (NO* + OCNH2* → ONOCNH2*) and successive protonation steps leading to urea formation (ONOCNH2* + HSp* → HONOCNH2* → HONOCNH2* + HSp* → NOCNH2* → NOCNH2* + HSp* → HNOCNH2* → HNOCNH2* + HSp* → H2NOCNH2(g)) proceed efficiently, as illustrated in Fig. 6g. Consequently, PdHx/Cu establishes an energetically optimized tandem pathway for urea electrosynthesis, where first coupling between COSp* and NO* onto Cu emerges as the rate-determining step (RDS) with a much lower barrier of 0.30 eV compared to that of the Cu counter (0.70 eV), accounting for the significantly enhanced rurea.

In addition, the competing reaction pathways over both PdHx and Cu domains were investigated. The side reactions of CO2 reduction to CO(g) and NO3− reduction to NH3(g) exhibit high energy barriers (minimum 0.56 eV; see Fig. 6d, f, and Supplementary Figs. 49, 50), which effectively suppress the formation of CO(g) and NH3(g) byproducts. The direct coupling between CO* and NO* intermediates on the PdHx domains of the PdHx/Cu model was also found to face a substantial barrier of 1.42 eV in the CO* + NO* → OCNO* step (Supplementary Fig. 51), ruling out this pathway. Further calculations of water dissociation energies on PdHx, the PdHx–Cu interface, and Cu domains revealed high barriers of 0.54 eV, 0.58 eV, and 0.40 eV, respectively (Supplementary Figs. 52–54), indicating that HER is also strongly inhibited. Similarly, the competing HER pathway via the coupling of HSp* species and protons from the electrolyte on the Cu domains of the PdHx/Cu model showed a high barrier of 0.42 eV in the H* + H⁺ + e⁻ → H2* step (Supplementary Fig. 55), further excluding this possibility. These collective mechanistic insights rationalize the observed high FEurea over the PdHx/Cu catalysts. It is important to note that these DFT calculations were performed on idealized models under standard conditions. Accordingly, key practical aspects such as explicit solvent effects and catalyst dynamics are thus not fully described in our current computational setup.

To experimentally support the theoretical results, a series of controlled experiments were systematically conducted. The kinetic isotope effect (KIE) was investigated by comparing rurea at − 0.7 V vs. RHE using 12CO2 and 13CO2 as reactants (Supplementary Fig. 56). The measured KIE value of 1.61 strongly indicated that surface CO* spillover or OCNO* formation constitutes the RDS in the urea electrosynthesis process. Considering that the in situ Fourier transform infrared spectroscopy results for PdHx/Cu showed more significant characteristic peaks as well as increased coverage of adsorbed NO* species with an increase in the NO3− concentration during urea electrosynthesis (Supplementary Fig. 57), we further explored the urea electrosynthesis performance of PdHx/Cu with different NO3− concentrations. The observed significant enhancement in rurea with increasing NO3− concentrations (Supplementary Fig. 58) provides compelling experimental evidence that the C-N coupling constituted the RDS. This conclusion is supported by the premise that more NO3− (or higher coverage of NO* species) would promote more frequent effective collisions with CO*, thereby accelerating the reaction kinetics if C-N coupling is indeed the RDS67. To gain deeper mechanistic insights, we determined the apparent activation energy (Ea) for urea electrosynthesis by analyzing the temperature dependence of rurea at − 0.7 V vs. RHE (Supplementary Fig. 59). The significantly lower Ea observed for PdHx/Cu compared to bare Cu (25.4 kJ mol−1 vs. 36.1 kJ mol−1) provides direct experimental evidence for the accelerated reaction kinetics enabled by CO* and H* spillover effects. To further verify this hypothesis, we measured the Ea of PPh3-treated PdHx/Cu, which selectively poisons the PdHx domain and suppresses spillover processes. The resulting Ea (43.5 kJ mol−1) was substantially higher than that of the untreated catalyst (Supplementary Fig. 60), reinforcing the critical role of CO* and H* spillover in facilitating urea electrosynthesis. These experimental observations—spanning KIE measurements, concentration-dependent kinetics, and Ea analyses—exhibit excellent agreement with our DFT calculations.

Discussion

In summary, we reported a dual spillover mechanism of CO* and H* to initiate an improved tandem urea electrosynthesis pathway over PdHx/Cu, where PdHx and Cu enabled efficient CO2-to-CO* and NO3−-to-NO* conversions, followed by rapid CO* spillover from PdHx surface to Cu for promoted C-N coupling and simultaneous H* spillover from PdHx lattice to Cu for facilitated C-N intermediate protonation. Consequently, these catalysts showed a high performance for urea electrosynthesis. These unique findings regarding CO* and H* spillover will broaden concepts for tandem catalyst design and provide new considerations for industrial urea electrosynthesis.

Methods

Synthesis of PdHx/Cu

Based on previously established methodologies32, the synthesis was carried out as follows. Initially, 50 mg of palladium acetylacetonate (99.9%, Macklin), 160 mg of polyvinylpyrrolidone (99.9%, Macklin), and 185 mg of cetyltrimethylammonium bromide (99.9%, Macklin) were dissolved in a mixed solvent consisting of 10 mL of N,N-dimethylformamide (99.9%, Macklin) and 2 mL of water under magnetic stirring. The resulting homogeneous yellow solution was transferred to a Schlenk flask, charged with CO gas, and immersed in a water bath at 60 °C for 12 h. After this, 10 mL of formaldehyde (37.0 wt.%, Macklin) solution was added, and the mixture was transferred to a 50 mL Teflon-lined stainless steel autoclave and heated at 160 °C for 4 h. The resulting PdHx was collected by centrifugation at a centrifugation speed of approximately 18,900 × g for 10 min and washed three times with ethanol. For the preparation of PdHx/Cu catalysts, 200 mg of Cu(NO3)2·3H2O (99.9%, Energy Chemical), 200 mg of L-ascorbic acid (99.9%, Macklin), and 4.0 mg of as-synthesized PdHx were dispersed in 30 mL of deionized water. Then, 200 mg of hexadecyltrimethylammonium bromide (99.9%, Macklin) and 200 mg of hexamethylenetetramine (99.9%, Macklin) were added to the mixture. The reaction system was sealed and heated at 80 °C for 3 h in an oil bath. The final products were isolated by centrifugation, rinsed repeatedly with distilled water and ethanol, and dried in a vacuum oven at 60 °C for 12 h to obtain the PdHx/Cu catalysts.

Synthesis of PdHx/rGO controls

50 mg rGO (99.9%, XFNANO) was mixed with 200 μL of the aforementioned CO-injected yellow solution (Pd content = 20 mg mL−1) and 10 mL formaldehyde solution, and transferred into a 15 mL Teflon-lined stainless steel autoclave and maintained at 160 °C for 4 h. The PdHx/rGO products were collected by centrifugation (centrifugation speed of 18,900 × g, 10 min) and washed three times with ethanol.

Synthesis of Cu controls

Cu(NO3)2·3H2O (100 mg) and L-ascorbic acid (200 mg) were mixed with 30 mL of deionized water and stirred to form a homogeneous solution. Then, hexadecyltrimethylammonium bromide (200 mg) and hexamethylenetetramine (200 mg) were added followed by 30 min of stirring. The mixture was sealed and heated to 80 °C for 3 h in an oil bath. The resulting products were centrifuged, rinsed with distilled water and ethanol several times, and then dried in a vacuum oven at 60 °C for 12 h.

Synthesis of Pd/Cu controls

The as-prepared Cu nanosheets (50 mg) were suspended in 47 ml of distilled water, and then 70 μL PdCl3 (99.9%, Macklin) solution (Pd content = 5 mg mL−1) was added. Subsequently, 3 mL of the freshly prepared NaBH4 (99.9%, Macklin) aqueous solution (containing 95 mg NaBH4) was quickly added under vigorous stirring. After 3 h reaction under stirring, the products were collected through centrifuging, washed with water/ethanol thoroughly and then dried at 60 °C for 12 h under vacuum to obtain the Pd/Cu.

Characterization

A Bruker D8 Focus Diffraction System was employed to record the XRD patterns of the products using a Cu Kα source (λ = 0.154 nm). TEM and EDX measurements were performed on a JEOL 2100 F TEM with an accelerating voltage of 200 kV. XPS results were recorded on a photoelectron spectrometer using Al Kα radiation as the excitation source (PHI 5000 VersaProbe). The C 1 s spectrum at a binding energy level of 284.8 eV was used to calibrate all the peaks. The metal content of various electrodes was examined by ICP-MS (Agilent 7500CE). All samples for ICP-MS were pre-treated by aqua regia to obtain clear solutions before measurements. FT-IR spectra were recorded with a NICOLET 6700 IR spectrometer (Thermo Fisher U.S.A.). The 1H ssNMR data were collected Bruker AVANCE III HD 400 MHz NMR spectrometer. The test was performed using an H/X dual-resonance solid probe, 4 mm ZrO2 rotor, spinning speed of 10 kHz, resonance frequency of 400.13 MHz, acquisition time of 2.5 µs, and a recycle delay of 3 s. Raman spectra data collected alpha 300 R. The test was performed 532 nm TEM00 laser, with the laser intensity set at 50% and the exposure time set at 15 s.

CO-TPD measurements

CO-TPD experiments were performed on a Chembet Pulsar instrument (Quantachrome Instruments, USA). A 50 mg sample was pretreated at 100 K for 3 h in a CO atmosphere and cleaned with an Ar gas flow for 30 min to remove the weakly adsorbed CO. The TPD process was performed by heating the sample from 0 to 700 °C at a ramp rate of 2.5 °C min−1 under an Ar atmosphere.

In situ IR

In situ IR experiments were carried out by a NICOLET 6700 IR spectrometer (Thermo Fisher U.S.A.) equipped with a chemically deposited ultra-thin Au film as working electrode for IR-signal enhancement (Supplementary Fig. 61). The obtained various catalysts were directly loaded on the working electrode, while Ag/AgCl and Pt foil served as reference electrode and counter electrode, respectively. It should be noted that the Ar gas was continuously bubbled into the electrolyte before and during the test. In situ IR signals were recorded when the electrodes were under potentiostatic tests.

Online DEMS measurements

Online DEMS measurements were conducted using an electrochemical cell coupled to a mass spectrometer (QAS 100, Linglu Instruments). Product detection was enabled by a porous PTFE membrane (Porosity ≥ 50%, pore size ≤ 20 nm; PF-002HS, Hangzhou Cobetter Filtration Equipment) coated on the back side of the working electrode (Supplementary Fig. 62). The experiments employed a 50 mL Ar-saturated 0.2 M KHCO3 electrolyte with different chemical feedstocks. A constant cathodic potential of − 0.7 V vs. RHE was applied during reaction, and the corresponding mass signals of various intermediates for the catalysts were monitored at different m/z ionic signals.

Electrochemical measurements

The catalyst ink was prepared by dispersing 1 mg of catalyst in a mixture of isopropanol (200 μL) and Nafion (40 μL, 5.0%, Macklin). The inclusion of Nafion serves two key purposes: (i) it acts as a binder, securing catalyst particles to GDE and preventing delamination during prolonged operation68; and (ii) it facilitates efficient proton transport between active sites due to its high proton conductivity69.

A CHI 760E workstation (CH Instruments, Inc., Shanghai) over two self-designed flow cells with electrode window areas of 1.0 cm2 and 9.0 cm2 was used for the measurements. For the three-electrode system, the catalyst ink was uniformly deposited onto a carbon paper gas diffusion electrode (GDE, Suzhou Sinero Technology Co., Ltd) using an ultrasonic spray coater (Sono-Tek ExactaCoat system) to achieve a catalyst loading of 0.4 mg cm−2. The synthesized GDE, Ag/AgCl, and nickel foam served as the working, reference, and counter electrodes, respectively. The catholyte and anolyte were formulated by dissolving KHCO3 in deionized water to form 0.2 M solutions (pH = 8.31 ± 0.02). The catholyte (60 mL) containing 0.1 M KNO3 is in a flow rate of 2.0 sccm and was supplied with 20.0 sccm of CO2, and the anolyte (60 mL) is in a flow rate of 2.0 sccm. An anion exchange membrane (AEM, Fumasep FAB-PK-130) was utilized to separate the anode and cathode into different electrolytes. The gaseous products were collected from the gas bag while the liquid products were collected from the cathode electrolyte bottle. For a typical measurement, urea electrosynthesis was driven by a potentiostatic test at the given potential for 1.0 h. The on-the-fly iR-compensation mode with a compensation proportion of 95% was applied in each test through software tools. Electrochemical impedance spectroscopy (EIS) investigations were conducted on the GAMRY Reference 600 electrochemistry workstation in the frequency range of 100 kHz−0.1 Hz at various overpotentials. The potentials reported are relative to the reversible hydrogen electrode (RHE). To realize the calibration, an actual reversible hydrogen electrode (RHE) was submerged in 0.2 M KHCO3 electrolyte and a reductive current density of approximately − 1.0 mA cm−2 was applied to generate H2 inside the RHE. The potential of the Ag/AgCl reference electrodes was determined to be 0.610 V vs. RHE (Supplementary Fig. 63).

For the two-electrode measurements, the catalyst ink was first prepared by dispersing 4 mg of the PdHx/Cu catalysts into a mixed solvent containing isopropanol (1000 μL) and Nafion (100 μL). For the fabrication of the catalytic electrodes, the catalyst ink was applied to a carbon paper GDE (Suzhou Sinero Technology Co., Ltd) at a loading of 0.44 mg cm−2 to form the cathode, and to a titanium felt GDE (Suzhou Sinero Technology Co., Ltd) at the same loading of 0.44 mg cm−2 to form the anode. The anode and cathode were separated by the anion exchange membrane (Fumasep FAB-PK-130). The catholyte and anolyte were formulated by dissolving KOH in deionized water to form 1.0 M solutions (pH = 13.69 ± 0.03). The catholyte (60 mL) containing 0.1 M KNO3 is in a flow rate of 2.0 sccm and was supplied with 20.0 sccm of CO2, and the anolyte (60 mL) and 2.0 M methanol is in a flow rate of 2.0 sccm. For a typical measurement, urea electrosynthesis was driven by a potentiostatic test at the given cell voltage for 1.0 h.

Product analysis

The following previously reported methods were employed for detection:

Urea

The urea products were analyzed with an NMR spectrometer (AVANCE III HD 400). Initially, a urea standard solution (500 μL) at different concentrations was mixed with a DMSO-d6 solution (150 μL) as an internal standard for 1H NMR detection to obtain the relationship between the urea−DMSO-d6 peak area ratio and the urea concentration (Supplementary Fig. 64). 500 μL of the post-reaction electrolyte was neutralized to pH ≈ 7 and mixed with 150 μL of DMSO-d6 for 1H NMR detection. Through the obtained urea−DMSO-d6 peak area ratio, the concentration of generated urea can be calculated. The urea products were also analyzed with a UV-Vis spectrometer (Beijing Purkinje GENERAL Instrument Co., Ltd, TU-1950). Initially, the color agent was prepared by dissolving FeCl3 (0.2 mg), diacetylmonoxime (1.0 mg) and thiosemicarbazide (5 mg) into the mixed solution, which contained H3PO4 solution (15.0 M, 200 μL), H2SO4 (18.4 M, 600 μL) solution and deionized water (1200 μL). A series of standard urea solutions (0.5 mL) with different urea concentrations was mixed with the color agent for 30 min for UV-Vis measurements. The absorption intensity at a wavelength of 525 nm was recorded, providing the corresponding concentration-absorbance curve as shown in Supplementary Fig. 65. Then, the electrolyte was collected, diluted and mixed with the above color agent for UV-Vis detection. Through the obtained absorption intensity at 525 nm, the concentration of generated urea can be calculated.

HCONH2

The HCONH2 products were analyzed with an NMR spectrometer (AVANCE III HD 400). Initially, a HCONH2 standard solution (500 μL) at different concentrations was mixed with a DMSO-d6 solution (150 μL) as an internal standard for 1H NMR detection to obtain the relationship between the HCONH2 − DMSO-d6 peak area ratio and the HCONH2 concentration (Supplementary Fig. 66). 500 μL of the post-reaction electrolyte was neutralized to pH ≈ 7 and mixed with 150 μL of DMSO-d6 for 1H NMR detection. Through the obtained HCONH2 − DMSO-d6 peak area ratio, the concentration of generated HCONH2 can be calculated.

NH3

UV-Vis absorbance spectra were obtained to analyze the concentrations of the reactants and products of NH3. All electrolytes were diluted to appropriate concentrations before spectrophotometric analysis. A color agent was prepared by dissolving salicylic acid (0.1 g) and sodium citrate (0.1 g) in NaOH solution (2 mL, 1.0 M) and then mixing with NaClO solution (1 mL, 0.05 M), sodium nitroferricyanide solution (0.2 mL, 1.0 wt%) and diluted electrolytes (5 mL) evenly. After standing for 30 min, the absorption intensity was recorded at a wavelength of 654 nm, and the concentration of produced NH3 could be estimated through the concentration‒absorbance curve using various standard NH4Cl solutions, as shown in Supplementary Fig. 67.

NO2 −

The ultraviolet-visible (UV-Vis) absorbance spectra were employed to analyze the NO2− products. To realize UV-Vis analysis, the color agent was prepared by dissolving p-aminobenzenesulfonamide (4 g) and N-(1-Naphthyl) ethylenediamine dihydrochloride (0.2 g) into ultrapure water (50 mL). Then, a series of standard NaNO2 solutions (5 mL) with different NO2− concentrations was mixed with the color agent (0.1 mL) for 20 min for UV-Vis measurements. The absorption intensity at a wavelength of 540 nm was recorded, providing the corresponding concentration-absorbance curve as shown in Supplementary Fig. 68. The electrolyte (5 mL) in the cathodic chamber was diluted and mixed with the above color agent for UV-Vis detection, and through the obtained absorption intensity at 540 nm, the concentration of generated NO2− can be calculated.

CH3COOH

The CH3COOH products were analyzed via an NMR spectrometer. Initially, different concentrations of the CH3COOH standard solution (500 μL) were mixed with a DMSO-d6 solution (150 μL) for 1H NMR detection to obtain the relationship between the CH3COOH − DMSO-d6 peak area ratio and the CH3COOH concentration (Supplementary Fig. 69). After neutralizing the pH of post-reaction electrolyte to 7, the generated CH3COOH in electrolyte could be estimated through this concentration‒peak area curve.

HCOO−

The HCOO− products were analyzed via an NMR spectrometer. Initially, different concentrations of the HCOO− standard solution (500 μL) were mixed with a DMSO-d6 solution (150 μL) for 1H NMR detection to obtain the relationship between the HCOO− − DMSO-d6 peak area ratio and the HCOO− concentration (Supplementary Fig. 70). After neutralizing the pH of the post-reaction electrolyte to 7, the generated HCOO− in the electrolyte could be estimated through this concentration‒peak area curve.

CO, H2 and N2

The generated CO, H2 and N2 were collected via a gas bag and analyzed via a gas chromatography instrument (North Point GC 901 A) equipped with a thermal conductivity detector and a flame ionization detector.

FE evaluation

The FE of the relevant reactions was calculated as follows:

$${{{\rm{FE}}}}(\%)={{{\rm{nFm}}}}/{{{\rm{Q}}}}$$

where n is the electron-transfer number of relevant reactions, m is the amount of generated products, F is the Faradaic constant (96485 C mol−1), and Q is the total charge passed through the electrode.

Computational details

All spin-polarized DFT calculations were performed using the Vienna Ab-initio Simulation Package (VASP) software70,71. The Perdew-Burke-Ernzerhof (PBE) under the generalized gradient approximation (GGA) was used to describe exchange-correlation functional72,73. In the process of geometric optimization, the 400 eV is set as the plane wave cutoff energy (Supplementary Fig. 71) meanwhile the 0.03 eV/Å and 1 × 10−5 eV are set as the convergence criterion of force and energy, respectively. The dispersion corrections are applied using the Grimme’s DFT-D3 method74. The optimized Monkhorst-Pack mesh of (1 × 2 × 1) is set for our system due to it achieves computational efficiency without compromising accuracy (Supplementary Table 6)75. The Gibbs free energy is corrected to the experimental temperature (298.15 K) using the VASPKIT software package76.

High resolution TEM image of the as-prepared PdHx/Cu catalysts (Fig. 2b) showed a Pd(111) facet for the PdHx in PdHx/Cu, and a Cu(111) facet for the Cu in PdHx/Cu. To simulate the PdHx component and its lattice H* spillover phenomenon, we firstly built a three-layer Pd slab model derived from Pd(111) with two H atoms inserted into its bulk phase similar to our previous work24. Then, we built a two-layer Cu cluster model originated from Cu(111) and placed the Cu cluster on the PdHx slab surface, finally creating the PdHx/Cu model. For comparison, the Cu cluster model was also employed. The optimized Cu cluster and PdHx/Cu geometries are shown in Supplementary Fig. 72.

In order to determine the change of Gibbs free energy (ΔG) of each elemental step, the following equation was employed: ΔG = ΔE + ΔZPE – TΔS, where ΔE refers to the electronic energy difference directly obtained from DFT calculations, ΔZPE represents the change in zero-point energies, T denotes the temperature (T = 298.15 K), ΔS accounts for the entropy change77.