Reversible metal cluster formation on Nitrogen-doped carbon controlling electrocatalyst particle size with subnanometer accuracy

The study addresses how to achieve on-demand control of product selectivity in electrocatalytic CO2 reduction on copper-based catalysts by dynamically steering the active Cu species between single atoms, subnanometer clusters, and larger nanoparticles. Cu-N-C systems feature Cu single sites that reversibly aggregate into metallic particles under CO2RR and redisperse under anodic conditions. The authors hypothesize that by applying pulsed potentials and tracking structural evolution operando, one can finely tune the particle size distribution with subnanometer accuracy and thereby rationally direct CO2RR selectivity among H2, CH4, CO, and C2+ products.

Prior work established that Cu-N-C catalysts can undergo reversible transformation between cationic single sites and metallic particles during operation, including systems derived from Cu phthalocyanine and MOF/COF precursors. Particle size effects in CO2RR are well documented for nanoparticles (2–15 nm), with smaller sizes favoring H2 and CO formation due to more low-coordinated sites; for ultrasmall clusters and single atoms, reports showed high selectivity toward methane or ethanol, but conclusions have varied. Pulsed electrolysis has been suggested as a tool to influence adsorbates, local environment, and catalyst structure. The present study integrates these strands by dynamically creating distinct Cu species in situ and correlating them to selectivity, reconciling previous conflicting size–selectivity reports.

Catalyst synthesis: A ZIF-8 precursor was synthesized in methanol from 2-methylimidazole and Zn(NO3)2·6H2O, refluxed 24 h, washed, dried, and pyrolyzed in Ar at 1000 °C for 1 h to form N-doped carbon (N-C). The N-C was acid-washed in 20 wt% HNO3, then impregnated with Cu2+ by dispersing in 6 mM Cu(NO3)2·3H2O in isopropanol with sonication (2 h at −40 °C) and stirring (2 h, RT), followed by drying and annealing in Ar at 700 °C for 1 h, acid-wash, and drying, yielding Cu-N-C with singly dispersed Cu. Ex-situ XPS, ICP-MS, XRD, and HAADF-STEM confirmed Cu incorporation and absence of clusters in as-prepared samples. Electrochemical conditions: Static CO2RR in CO2-saturated 0.1 M KHCO3 (pH 6.8) at −1.35 V vs RHE. Pulsed CO2RR alternating Ec = −1.35 V and Ea = 0.44 V with varied pulse durations Δtc and Δta (typically 2–30 s). Potentials referenced to RHE (ERHE = EAg/AgCl + 0.242 V + 0.059·pH). Potential pulses were chosen to ensure cluster formation under Ec and redispersion under Ea. Operando spectroscopy: Quick XAFS at Cu K-edge (8979 eV) with time resolution up to 2 s per spectrum at SuperXAS (SLS), P64 (PETRA III), and CryoEXAFS/KMC-3 (BESSY II) beamlines. Fluorescence mode detection (PIPS or 13-element SDD). XANES/EXAFS processed via beamline software, in-house Mathematica scripts, LARCH and FEFFIT. Linear combination analysis (LCA) of XANES using Cu foil and as-prepared Cu-N-C as references; PCA to verify no copper oxide formation during anodic pulses. EXAFS fitted in R-space (R = 1.0–2.8 Å, k = 2–8 Å−1) including Cu–O and Cu–Cu paths; refined R, N, σ2 and ΔE0; amplitude reduction factors S0^2 of 0.85 (Cu–Cu) and 0.68 (Cu–O). Apparent coordination numbers corrected for the fraction of singly dispersed Cu (χSAC) to obtain true Cu–Cu CN for metallic species. Kinetic/size modeling: Minimal two-population model (ultrasmall clusters and larger particles) incorporating four processes: (i) reduction of single sites during Ec; (ii) growth of metallic particles during Ec; (iii) fragmentation of particles during Ea; (iv) conversion of small clusters to single sites during Ea. Transition rates and sizes estimated from early-cycle and steady-state χSAC and Cu–Cu CN at Δtc = Δta = 30 s. Particle sizes inferred from Cu–Cu CN and literature CN–size relations. Electrocatalytic testing: H-type two-compartment cell with anion-exchange membrane, CO2-saturated 0.1 M KHCO3, Toray carbon paper (0.5 cm2) spray-coated on one side with catalyst; Ag/AgCl reference (checked vs RHE) and Pt gauze counter. CO2 flow 20 mL min−1 to both compartments. Potentiostat: Autolab PGSTAT 302N. Online GC (Agilent 7890B; TCD for H2, FID for C-products) every 15 min up to 4000 s per condition; liquid products by L-GC and HPLC. For some pulse series, the same sample was reused sequentially with 30 min OCP rests; only gaseous products monitored and selectivities/partial currents reported. For very short Δtc (<10 s), capacitive currents dominate; a procedure assuming total FE = 100% for detected products was used to deconvolute partial currents and compute FEs.

• Under static CO2RR at −1.35 V, singly dispersed Cu2+ (distorted octahedral Cu–N4 with axial O/OH) rapidly transform (≈100 s) into metallic Cu particles with average effective diameter ≈1.3 ± 0.1 nm, coexisting with residual singly dispersed Cu that stabilizes at ≈12%.
• Pulsed CO2RR (Ec = −1.35 V, Ea = 0.44 V) induces periodic and reversible switching between cationic single sites and metallic Cu, with no copper oxides detected during Ea pulses (LCA/PCA of XANES).
• The system reaches a stationary oscillatory state; χSAC oscillates around higher averages for shorter Δtc (≈40% at Δtc = 2 s; ≈18% at Δtc = 30 s). Cu–Cu CN oscillates (e.g., ≈8.0–9.5 at Δtc = Δta = 30 s).
• Average true Cu–Cu CN increases with Δtc, demonstrating tunable size: ≈1.2 ± 0.1 (Δtc = 2 s) to ≈8.8 ± 0.1 (Δtc = 30 s). Due to coexistence of sizes, CNs reflect distributions rather than single diameters.
• Minimal kinetic model reproduces trends: larger particles formed under pulsed CO2RR have Cu–Cu CN ≈10.1 (≈2.5 nm diameter), nearly twice the static case; ultrasmall clusters’ Cu–Cu CN increases from ≈0.4 (Δtc = 2 s; mainly single atoms/dimers) to ≈4.4 (Δtc = 30 s; ≈0.6 nm), below that of a 13-atom cuboctahedron (CN ≈5.5).
• Three structural regimes identified via pulse parameters: (i) short Δta favor singly dispersed cationic species (Δta < tSAC ≈ 165 s), (ii) intermediate Δta yield strong contribution of ultrasmall clusters, (iii) long Δta favor larger ≈2.5 nm particles (Δta > tSL ≈ 18 s). Increasing Δta mainly shifts regime boundaries toward higher Δtc.
• Selectivity: Static at −1.35 V after 4000 s: H2 FE ≈47%, CO ≈22%, CH4 ≈8%, formate ≈8%.
• Pulsed CO2RR (Ec = −1.35 V, Ea = 0.44 V): for Δtc ≤ 1 s, HER dominates (up to ≈85% FE) with CO and small formate; for Δtc = 10–16 s, HER is suppressed and CH4 FE reaches up to ≈25% with detectable C2H4, ethanol, acetaldehyde; at larger Δtc, CH4 decreases while H2 and CO increase, approaching static-like distributions. Non-monotonic selectivity vs Δtc correlates with the evolving fraction of single sites, ultrasmall clusters, and larger particles.
• Mechanistic insights: Growth under Ec likely dominated by diffusion–coalescence of small metallic species; pulsing regenerates mobile ultradispersed species, enabling larger final sizes (higher CN) than static. Strong particle–N-C support interactions limit maximal size to a few nm and facilitate fragmentation under Ea.
• Dissolution is not the primary cause of particle disappearance: Cu loss <30% over 5 h; no correlation between Cu fluorescence intensity and cluster formation/fragmentation under pulsed conditions.

Dynamic potential pulsing decouples contributions from coexisting Cu species and provides a direct link between structure and selectivity, reconciling prior reports on Cu particle size effects in CO2RR. When larger particles dominate (~2.5 nm), selectivity mirrors monodisperse nanoparticle studies (H2 50–60%, CO ~25%, CH4 15–20%, minor C2H4), attributed to more low-coordinated sites enhancing H and CO2 adsorption. However, selectivity cannot be described by size alone due to differences in facets, defects, and size distributions. Ultrasmall Cu clusters, in contrast, show pronounced CH4 selectivity, aligning with CuPc-derived Cu-N-C studies and DFT predictions that low-coordinated sites lower barriers for CO hydrogenation to methane. The work demonstrates a practical means to switch among regimes dominated by single sites (favoring HER), ultrasmall clusters (favoring CH4), and larger nanoparticles (favoring CO and enabling C2+), by tuning pulse durations. Additional pulsed-condition effects (support changes, adsorbate dynamics, local pH/reactant concentration variations) may also contribute. Strong Cu–support interactions in N-C cap particle growth and promote reversible fragmentation under anodic potentials, explaining oxide absence and size limits.

Pulsed electrolysis enables in situ, reversible control of Cu speciation on Cu-N-C with subnanometer precision, allowing on-demand steering of CO2RR selectivity from predominantly H2 to CH4 or to CO/C2+ products. Time-resolved operando QXAFS disentangles the roles of single atoms, ultrasmall clusters, and nanoparticles, and identifies pulse parameter regions that reproducibly generate each active species. A minimal kinetic model captures key trends and quantifies sizes and fractions of coexisting species. The approach offers a generalizable route to probe and exploit structure–reactivity relationships in dynamic catalytic systems and may extend to other electrocatalytic and thermal catalysis contexts involving reversible metal exsolution/fragmentation. Future work should optimize pulse shapes and parameters via computational design, refine kinetic models, and probe additional structural descriptors (facets, defects) to further enhance selectivity control.

• For very short Δtc (<10 s), capacitive currents dominate, complicating quantitative Faradaic analysis; assumptions (e.g., total FE = 100% for detected products) are used to estimate FEs.
• Some product analyses reused the same electrode without electrolyte refresh between pulse conditions, monitoring only gaseous products; sample history and irreversible support changes can influence selectivity.
• LCA of XANES employs fixed references (Cu foil and as-prepared Cu-N-C), neglecting size-dependent XANES changes and possible evolution of single-site spectra under CO2RR.
• EXAFS-based coordination numbers assume 6-coordination for single sites and require averaging across coexisting species; particle size assignments from CN carry uncertainty due to size distributions.
• The kinetic model uses simplified (exponential) rate dependencies and a two-size population approximation, leading to semi-quantitative agreement but discrepancies in details.
• Maximum particle size remains limited by strong Cu–N-C interactions; conclusions pertain to this support and may not directly generalize to other supports or electrolytes.