Bioadaptive Ni single atoms unlock high rate microbial electrosynthesis of isopropanol from CO₂

Abstract

Hybrid systems that integrate electrochemical CO₂ reduction with microbial upgrading offer a viable route to high value organic compounds from CO₂ at ambient conditions. However, electrocatalyst deactivation in microbial growth media remains a key barrier, limiting efficiency and increasing cost. Here we show that a bioadaptive single-atom nickel catalyst (Ni SAC), coupled with genetically engineered Clostridium ljungdahlii, enables robust electrosynthesis of isopropanol (IPA) from CO₂ via a CO-mediated pathway. Instead of relying on H₂ as an electron carrier, the system applies high-rate CO formation in complex growth media, maintaining a tunable CO Faradaic efficiency up to 92%, which is 9.4 to 52.7 times greater than conventional Ag catalysts. This performance supports stable IPA production at current density of 10.8 A/m² and production rate of 161.3 mg/L/day. In situ Raman and X-ray absorption spectroscopy, together with theoretical calculations, indicate that the Ni SAC can resist competing organic adsorption and retain its coordination structure during CO₂ reduction in bioelectrolytes, providing a mechanistic basis for the catalyst stability and integrated process performance.

The deactivation of CO₂ reduction electrocatalyst in microbial media remains a key barrier for hybrid bio-electrochemical systems. Here, the authors present a bioadaptive nickel single atom catalyst that resists organic poisoning to enable high-rate CO-mediated isopropanol production from CO₂.

Introduction

Isopropanol (IPA) is a critical platform chemical that is widely used in personal healthcare products and in industrial production of fuels and chemicals. Because IPA has a higher octane rating than conventional fuels, it is regarded as a promising fuel additive for enhancing combustion efficiency¹. Its disinfectant properties also prompted a dramatic increase in global demand and prices during the COVID-19 pandemic. Moreover, IPA’s critical role in semiconductor cleaning has fueled overall IPA market expansion, as rising chip demand from AI technologies driven increasing market share within this segment^2,3. The global IPA market was valued at $4.8 billion in 2024 and is projected to grow significantly over the next decade to more than $7 billion by 2033⁴. Current industrial IPA production predominantly relies on energy-intensive hydration of propylene (two-thirds of supply) or hydrogenation of acetone. Both petro-based routes depend on fossil propylene or H₂, as well as face challenging separations from IPA/H₂O/diisopropylether azeotropes⁵. Renewable energy-driven electrochemical CO₂ conversion offers a sustainable approach to both reduce carbon emissions and produce high-value chemicals^6–9. However, prior research on C₃ alcohol electrosynthesis has only achieved a maximum Faradic efficiency (FE) of 30-50% for n-propanol with pure CO supply^10–13 and was only able to realize IPA generation under high-pressure CO₂ conditions¹⁴. The synthesis of IPA via CO₂ electrolysis at ambient pressure remains a major barrier.

Anaerobic microbial fermentation of syngas recently emerged as a powerful platform to convert C₁ substrates into multi-carbon alcohols^15,16. In particular, genetic engineering of Clostridium ljunghdalii has enabled the production of IPA from syngas, demonstrating the feasibility of tailoring microbial pathways toward target alcohols¹⁷. However, commercial syngas production still depends on coal gasification and natural gas refining, both of which are highly carbon-intensive¹⁸. To address this, a variety of high-selectivity, tunable CO₂ electrolysis methods have been reported for on-demand syngas generation, offering a cleaner syngas source for bioconversion¹⁹. Integrated microbial electrochemical systems (iMES) based on gas diffusion electrodes have been shown to enhance kinetic mass transfer during microbial fermentation and potentially overcome limitations associated with separated abiotic electrolysis and fermenentation^20,21. Moreover, iMES systems enable microbials to convert simple electrode-derived intermediates (e.g., H₂, CO, and formate) into a wide range of high-value products such as alcohols, carboxylic acids and esters through their native or engineered metabolic pathways²², which are nearly impossible on traditional inorganic catalyst surfaces. This material-microbe synergy can thus expand the accessible electrochemical product space beyond the limits of conventional electrocatalysis.

Considering the energy demand and carbon footprint, the iMES process may decouple IPA synthesis from fossil feedstocks by using renewable electricity to generate electron donors. This shifts energy and carbon inputs from fossil feedstocks to electrons and waste CO₂, avoids steam reforming emissions, and is operationally attractive for distributed, point-source CO₂ valorization. Notably, fermentation of syngas has already achieved carbon-negative IPA at pilot scale (−1.17 kg CO₂ per kg IPA) when powered by low-carbon energy and waste-gas feeds²³, which further highlights the decarbonization potential by interlinking the microbial conversion with renewable syngas supply. Despite such promise, electrochemical syngas-mediated IPA synthesis via iMES has not yet been reported, primarily because traditional abiotic catalysts that were optimized for abiotic electrolyte lose significant activities when exposed to complex, nutrient-rich biological growth media^24,25. The fundamental causes of catalyst deactivation in such mixed electrochemical-microbial environments remain poorly understood, impeding the development of an integrated, syngas-fed IPA production platform.

In this study, we designed a single-site Ni catalyst that can exhibit high tolerance to bioelectrolyte components (i.e., YT medium for microbial growth) and enable selective CO production, thereby significantly enhancing the CO supply rate in biological electrolytes. The synthesized Ni single-atom catalyst (Ni SAC) demonstrated increased resistance to complex organic components in the electrolyte than conventional Ag catalyst, maintaining CO FE nearly identical to that in ideal electrolytes while allowing tunable syngas ratios. When we combined this catalyst with a genetically engineered IPA-producing C. Ljungdahlii biosynthetic platform, we achieved the CO-mediated CO₂-to-IPA microbial electrochemical synthesis with an average rate of 161.3 mg/L/day. Additionally, we employed multiple in-situ characterization techniques to deeply probe the electrode-electrolyte interfacial properties in the bioelectrolyte and elucidated the deactivation mechanism of conventional electrocatalysts. We discovered that on continuous polycrystalline metal surfaces (e.g., Ag), extensive coverage of organics from the microbial growth medium disturbed the electrode-electrolyte interface and competitively occupied reactive sites, deactivating CO₂ adsorption and conversion. We hope this study will provide different insights for future optimization of interfacial properties and the expansion of material–microbial hybrid systems.

Results

Ni single-atom catalyst enables CO generation in microbial media for IPA synthesis

We employed a genetically tractable acetogenic bacterium, Clostridium Ljungdahlii, in the experiments due to its capacity to convert CO₂/H₂ or syngas into value-added biochemicals²⁶. Through systematic thermodynamic optimization and theoretical modeling, we previously demonstrated that overexpression of acetoacetyl-CoA transferase (AACT) and acetoacetate decarboxylase (AADC) enabled robust conversion of CO₂/CO/H₂ mixtures into IPA (Fig. 1a), while synergistically enhancing ethanol biosynthesis¹⁷. In this study, our comparative fermentation analyses identified CO as a critical determinant for IPA synthesis. Under identical cultivation conditions, fermentation without CO feeding exhibited poor microbial growth (Supplementary Fig. S1) and yielded negligible IPA production (<0.1 mM). The maximum acetate and ethanol titers were constrained to ~42.9 mM (2531.8 mg/L) and ~8.9 mM (409.4 mg/L), respectively (Fig. 1b, c). In contrast, CO-supplemented fermentation produced ~11.4 mM IPA (681.3 mg/L), representing a 142-times enhancement, which was also accompanied by a 1.7 times (73.9 mM, 4359.3 mg/L) increase in acetate and a 3.3 times boost (29.3 mM, 1347.8 mg/L) in ethanol production (Fig. 1b, c). The cell viability results showed that the engineered C. Ljungdahlii strain cannot grow without CO/H₂/CO₂ feeding (Supplementary Fig. S2), but the introduction of CO/H₂/CO₂ apparently promoted viability during the logarithmic phase in the fermentation experiments, indicating that the organic components in YT medium alone should be insufficient to support IPA, acetate, or ethanol production. A drop in cell viability from day 3 to 5 was also observed, signifying that biological growth and metabolism had ceased and cells were entering apoptosis, which corresponds to the plateau observed in both metabolite concentrations and OD₆₀₀. During this period, dead cells remained in suspension and continued to contribute to the optical density. As a result, the OD₆₀₀ was distorted and could not serve as a reliable indicator of actual cell viability.

Fig. 1. Syngas fermentation and electrocatalyst bioadaptivity.

a Scheme of proposed IPA synthesis by bioadaptive Ni SAC and engineered C. ljungdahlii (with detailed metabolism pathway shown). Gas fermentation results for b IPA (red lines) and c Acetate (orange lines), and EtOH (blue lines) production with CO/H₂/CO₂ (solid lines) or H₂/CO₂ (dashed lines) feeding at a temperature of 37 °C. Here, 20%/80% H₂/CO₂ and 50%/20%/30% CO/H₂/CO₂ were used for the gas fermentation experiments. FE (left Y axis, red bars: CO₂RR, gray bars: HER) and current densities (right Y axis, red lines: CO₂RR, gray lines: HER) of d Ag NPs and e Ni SAC in bioelectrolyte at a temperature of 37 °C without iR compensation. The error bars represent the standard deviations from at least three independent tests. Source data are provided as a Source Data file.

This CO-dependent metabolic shift arises from its dual role in precursor and energy metabolism (Fig. 1a). Beyond serving as a carbon source for acetyl-CoA formation, CO oxidation via CO dehydrogenase (CODH) generates reduced ferredoxin (Fd_red). This electron carrier drives the Rnf complex to establish a transmembrane H⁺ gradient, thereby generating ATP to sustain growth and NADH to drive IPA biosynthesis²⁶. H₂ metabolism of C. ljungdahlii is less efficient, although it can also supply reducing equivalents via hydrogenases²⁷. In H₂/CO₂ fermentation, H₂ must simultaneously support both branches of the Wood-Ljungdahl pathway (WLP). In the methyl branch, CO₂ is assimilated via the formate dehydrogenase (Fdh) and H₂ is used by a bifurcating hydrogenase (in a complex with the Fdh) to generate Fd_red for formate formation. While in the carbonyl branch, Fd_red is required again for CO₂-to-CO reduction by CODH. In addition, Fdred is required for ATP generation by the Rnf complex/ATPase, which is essential for autotrophic growth by these organisms. As a result, when C. ljungdahlii grows on H₂/CO₂, the available H₂ pool is divided between carbon assimilation and energy conservation, leaving far fewer reducing equivalents available for product (e.g., IPA) formation. In contrast, CO oxidation directly provides Fdred via CODH, which can be efficiently utilized by the Rnf complex and other enzymes (e.g., Nfn) to generate NAD(P)H. Thus, in CO/CO₂/H₂ fermentation, the pool of Fdred is significantly larger, supporting both robust autotrophic growth and enhanced IPA production²⁸. Consequently, microbial electrochemical systems targeting highly reduced compounds like IPA must prioritize sustained CO generation for C. ljungdahlii, ensuring both carbon flux and energy conservation requirements are met.

However, most microbial electrosynthesis studies supply H₂ as the primary electron donor rather than generating CO from CO₂ reduction. Although the thermodynamic standard potential for H₂ evolution is only 0.1 V more positive than that for CO formation, media complexity (ones containing yeast extract and tryptone)²⁹, low conductivity, and low alkali-cation concentration of bioelectrolyte make hydrogen evolution reaction (HER) kinetically more favorable and technically more convenient. Moreover, previous works have shown that CO₂ reduction performance is severely impaired in microbial growth medium^24,25. To address this challenge, we first evaluated Ag, a benchmark catalyst for CO₂-to-CO electroconversion, in the C. ljungdahlii growth medium. Even at a deep reduction potential of −1.05 V vs. reversible hydrogen electrode (RHE), the FE for CO production remained below 25%, with a partial current density of only 2.7 mA/cm2 (Fig. 1d), representing a 1 ~ 2 order-of-magnitude decrease compared to previously reported Ag performances in abiotic electrolyte^30–32. Resolving the deactivation of CO₂ reduction catalysts in bioelectrolyte is therefore a critical step toward an integrated microbial electrochemical process for IPA production (Fig. 1a).

The CO dependency of engineered C. ljungdahlii necessitates a bioelectrolyte-compatible CO₂ reduction catalyst capable of circumventing organic poisoning. Notably, recent advances in electrocatalysis revealed that atomically dispersed Ni-N-C configurations surpass conventional Ag catalysts in the abiotic medium³³. In addition, the elevated potential of zero charge (PZC) of Ni single sites sustained increased surface cation charge density than silver, which stabilized the *CO₂^- intermediate and provided high cation insensitivity, enabling sustained CO FE even in cation-insufficient organic amine electrolytes³⁴. While this frontier remained underexplored, we hypothesized that a structurally minimalist Ni single-atom structure could potentially exhibit high biocompatibility and overcome the limitations of conventional catalysts in biological medium. Hence, we engineered a N-coordinated Ni single-atom catalyst (Ni SAC) on N-doped carbon using a modified pyrolysis approach (synthesis details and structure analysis in Methods and Materials). For conventional abiotic electrolysis, Ni SAC achieved >95% of CO FE at industrial current densities (50–500 mA/cm2, Supplementary Fig. S3). In comparison to polycrystalline Ag NPs, Ni SAC exhibited a notable enhancement in CO selectivity and activity in the bioelectrolyte (Fig. 1e, Supplementary Fig. S4). Its CO FE was ~25.3 times that of Ag NPs at −0.85 V vs. RHE and can exceed 92% at higher overpotentials. Meanwhile, the maximum CO partial current density of Ni SAC in the concentrated YT medium surpassed 25.3 mA/cm² at −1.05 V vs. RHE. In addition, the performance and controllable CO₂/H₂ ratio of Ni SAC can also be extended to other commonly used microbial growth media (e.g. PETC medium, Supplementary Fig. S5), establishing it as a competitive candidate for integrated microbial electrochemical IPA production.

Adsorption-induced deactivation in bioelectrolyte

Unlike conventional abiotic electrolytes that facilitate CO₂ reduction through optimized alkali cation effects^35,36, the alkali ion-depleted biological medium (pH ≈ 6) presents a fundamentally distinct interfacial environment. The insufficient ionic strength may change the interfacial properties and render both CO₂ mass transport, adsorption, and subsequent intermediate stabilization thermodynamically unfavorable on the Ag surface³⁷. Ionic chromatography (IC) analysis revealed that the bioelectrolyte contains approximately 93.3 mM Na⁺ and 13.3 mM K⁺ (Supplementary Table S1), and thus we used a 100 mM Na⁺ solution as an ideal electrolyte for comparative studies. Particularly, Ag NPs exhibited severe deactivation in the bioelectrolyte, with CO FE and partial current density reduced by >75% at identical potentials compared to the ideal electrolyte (Supplementary Fig. S6). In contrast, Ni SAC maintained nearly identical CO generation activity, FE, and performance trends in both electrolytes (Supplementary Fig. S7). This phenomenon indicated that the complex medium may induce catalytic deactivation on Ag NPs via non-canonical mechanisms surpassing simple alkali cation depletion effects, while Ni SAC can inherently resist this. The bioelectrolyte contains dozens of organic compounds, including amino acids, vitamins, purines/pyrimidines and fragments of nucleotides/polypeptides, all of which may influence the catalyst interface properties and potentially interact with each other³⁸. The complex composition of bioelectrolyte precludes deconvolution of individual organic species contributions.

To establish a mechanistic understanding of interfacial properties, we systematically characterized and compared the electrochemical behaviors of Ag NPs and Ni SAC. Electrochemical kinetics indicated that the rate-determining step for CO₂RR on both Ag NPs and Ni SACs was the first electron transfer during CO₂ adsorption³⁹, and Ni SAC always exhibited CO₂RR kinetics better than Ag NPs (Supplementary Fig. S8). Cyclic voltammetry (CV) tests showed the double-layer capacitance (Cdl) of both catalysts decreased by ~30% in bioelectrolyte versus ideal conditions (Supplementary Fig. S9, 10), demonstrating the accumulation of hydrophilic organic macromolecules within the electrochemical double layer (EDL), inducing Helmholtz layer thickening, and thereby impeding mass transport of dissolving CO₂^40,41. Conventionally, Cdl variations among electrodes in identical electrolytes were attributed to differences in electrochemically active surface area^10,42,43. However, due to the significant changes in the electrolyte environment in this study, we directly normalized the partial current densities by Cdl. As shown in Fig. 2a, b, the normalized CO production rates of Ni SAC in either ideal or bioelectrolyte were significantly higher than Ag NPs, indicating the greatly promoted CO₂RR kinetics on each reactive site. More importantly, even excluding the effect of Cdl, Ag NPs exhibited a substantial intrinsic activity loss for CO production in the bioelectrolyte, whereas the performance of Ni SACs remained nearly unchanged compared to that in the ideal electrolyte. This suggested that the presence or absence of adsorptive poisoning by organic species on the reactive sites, thereby affecting CO₂ adsorption, may play a dominant role in altering CO₂RR activity on Ag NPs and Ni SACs.

Fig. 2. Electrochemical performance and bioadaptive mechanisms.

Normalized CO (red lines) / H₂ (blue lines) partial current densities of a Ag NPs and b Ni SAC in ideal (dash lines) and bioelectrolyte (solid lines) at a temperature of 37 °C without iR compensation. In-situ Raman spectrum in bioelectrolyte on the surface of c Ag NPs and d Ni SAC. e DFT calculations of adsorption free energies of representative organics. f Scheme of different bioadaptive behaviors of Ag and Ni SAC electrodes. The error bars represent the standard deviations from at least three independent tests. Source data are provided as a Source Data file.

Accordingly, we employed in-situ Raman to systematically investigate adsorption behavior and local environment on Ag NPs and Ni SAC reactive sites (Fig. 2c, d). During potential sweeps from −0.5 V to −0.9 V vs. RHE, the Ag electrode surface exhibited numerous complex organic adsorption signatures. Compared to previous studies using Ag colloids for yeast extract detection, the bioelectrolyte adsorbed on Ag nanoparticles exhibited changed and more complex Raman spectra⁴⁴. For example, the adenine ring-breathing peak (~730 cm⁻¹) was replaced by an in-phase ring stretching peak of guanine at 680 cm⁻¹, revealing a shift in adsorption preference under negative potentials. At higher shifts (e.g., 1000–2000 cm⁻¹), overlapping peaks from purines (e.g., various C–N vibrations of adenine or guanine from 1200–1500 cm⁻¹)⁴⁵ and amino acids (e.g., −COO⁻ of alanine, ~1398 cm⁻¹)⁴⁶ complicated precise assignments. Despite this complexity, the spectra confirmed significant adsorption of organic compounds from bioelectrolyte on Ag NPs within the CO₂RR potential window. Notably, all organic adsorption signals decreased with increasing overpotential, culminating in significant desorption beyond −0.8 V vs. RHE where Raman intensities diminish markedly. This potential-dependent desorption correlates well with the observed recovery of CO₂RR activity at −0.85 V vs. RHE in our electrochemical measurements, suggesting that the competitive adsorption of organics from the bioelectrolyte and the resulting changes in interfacial properties would be the main reasons for the deactivation of the CO₂RR on Ag NPs. Nevertheless, due to the rapid development of competing H₂ evolution at higher overpotentials^47,48, the higher energy and catalyst costs, as well as the increased risk of Ag⁺ leaching during long-time operation³⁰, Ag is still a less suitable candidate than Ni SAC for integration into iMES systems targeting IPA production.

In contrast, the Ni SAC surface did not exhibit any complex organic adsorption peaks like those on Ag, and only the D/G bands fingerprints of carbon were present⁴⁹. It is noteworthy that although we enhanced the Raman response of Ni SAC by using Au/SiO₂ core-shell nanoparticles⁵⁰, the signals from organic substances in the bioelectrolyte on Ni SAC should be much weaker than those on Ag particles with an inherent surface-enhancement effect⁵¹. Therefore, we cannot rule out the possibility of a small amount of organic adsorption on the Ni SAC sites only based on in-situ Raman results. Nevertheless, compared with Ag NPs, the extent of organic poisoning on the Ni SAC surface was significantly reduced or even absent, thus preserving a sufficient number of sites for CO₂ adsorption and activation.

While distinguishing the individual roles of all organic species remains challenging, we employed Ag(111) and Ni–N₄ as catalyst models and conducted DFT calculations to investigate differences in thermodynamic adsorption and interfacial properties (Supplementary Data 1). The enrichment of hydrophilic organic molecules within the double layer modified interfacial characteristics including ion concentration and electric field strength. CO₂RR intermediates (*CO₂, *COOH, *CO) responded differently to variations in electric field strength. The adsorption of CO₂ with large dipole moment exhibited the strongest field dependence (Supplementary Fig. S11), and CO₂ adsorption at Ni–N₄ sites demonstrated greater electric field sensitivity compared to that on Ag(111), enabling effective CO₂ activation under weaker fields. This suggested that Ni-N₄ single-atom sites, which has discrete and narrow d state³⁹, can maintain effective CO₂ activation capability under demanding interfacial conditions. Guided by in-situ Raman results and the composition profile of microbial growth medium⁵², we selected adenine, guanine, alanine, and aspartic acid as representative organics to assess their adsorption energies differences. As shown in Fig. 2e, Supplementary Fig. S12, the adsorption of all the tested organics was feasible on Ag(111) but thermodynamically unfavorable on Ni–N₄, and the adsorption free energy difference between these two sites all exceeded 0.3 eV, indicating that Ni–N₄ sites are relatively repulsive toward these organic than Ag (Fig. 2f), aligning well with the proposed mechanism.

Atomic structure dynamics of Ni SAC

Compared to Ag NPs, the promoted adaptability and activity of Ni SAC in the bioelectrolyte were expected to be correlated with its distinctive atomic architecture. To elucidate the structure-performance correlation and monitor the dynamic structure evolution of Ni SAC during electrolysis, comprehensive in-situ/ex-situ characterizations were conducted. Scanning transmission electron microscopy (STEM) images revealed that the carbon substrate in Ni SAC exhibits an uneven, wrinkled morphology (Supplementary Fig. S13). Elemental mapping (Supplementary Fig. S14) and energy dispersive X-ray spectroscopy (EDS, Supplementary Fig. S15) confirmed the uniform distribution of Ni, N, C, and O elements with Ni loading amount of 0.33 at%, while X-ray diffraction (XRD, Supplementary Fig. S16) showed no characteristic peaks of Ni NPs. High-resolution high-angle annular dark-field STEM (HAADF-STEM, Fig. 3a) quantitatively verified the single atom dispersion with interatomic distances measuring 5.5 Å, which exceeds the theoretical Ni–Ni bond length (2.5 Å) and excludes cluster formation. The absence of a continuous metallic Ni surface may obstruct large organic molecules from achieving multidentate adsorption, thereby making the interaction between atomically dispersed Ni sites and organics more difficult^53,54. Variations in the vertical height of these Ni atoms can be attributed to the unevenness of the carbon substrate. The locally uneven carbon framework surrounding the single-atom Ni sites can impose considerable steric hindrance, preventing large organic molecules, especially those with rigid and planar purine rings, from approaching the Ni center and forming strong bonding interactions, while these molecules can adjust their orientation to seek the most stable adsorption configuration on the broad Ag surface (Supplementary Fig. S12)⁵⁴. High-resolution X-ray photoelectron spectroscopy (XPS) revealed a characteristic Ni 2p_3/2 peak centered at 855.2 eV with a Ni atomic ratio of 0.34 at% (Fig. 3b, Supplementary Fig. S17), consistent with EDS results. Deconvolution of N 1 s spectra identified catalytically active pyridinic (398.2 eV) and pyrrolic (400.8 eV) nitrogen as the dominant species, with a distinct Ni-N contribution (~399.5 eV) of ~1.93 at% (Supplementary Fig. S18). Comparative analysis with NiPc demonstrated a 0.7 eV negative shift in Ni binding energy (Fig. 3b), indicating the formation of an electron-rich Ni center with valence state lower than +2 ⁵⁵.

Fig. 3. Structural and electronic characterizations of Ni SAC.

a HAADF-STEM image of Ni SAC. The illustration shows a 3D data map of two randomly selected individual atomic sites under dark-field mode, where the vertical intensity represents differences in contrast. Scale bar: 5 nm. b High-resolution XPS of Ni SAC (red line), Ni NPs (gray line) and NiPc (blue line). Ex-situ c XANES and d EXAFS comparisons between Ni SAC (red lines) and reference materials (gray lines: Ni foil, blue lines: NiPc). In-situ e XANES and f EXAFS comparisons of Ni SAC in CO₂ saturated ideal (blue lines) and bioelectrolyte (red lines). Source data are provided as a Source Data file.

Synchrotron X-ray absorption spectroscopy (XAS) enables highly precise characterization of the valence state and coordination environment of single atoms, leading to a deeper understanding of the local environment of Ni single sites. Ex-situ X-ray absorption near edge structure (XANES) analysis revealed that the Ni K-edge absorption energy of the Ni SAC exhibited a shift to higher energy positions compared to Ni foil, resembling the XANES spectrum of NiPc (Fig. 3c). This suggested that the Ni SAC likely adopted a square planar-like structure with a valence state just below +2⁵⁵, supported by the first derivative analysis of the XANES spectrum (Supplementary Fig. S19) as well as XPS results. Fourier-transformed extended X-ray absorption fine structure (EXAFS, Fig. 3d) indicated exclusive Ni–N coordination at 1.87 Å (Supplementary Table S3), with no detectable Ni–Ni metallic bonds (2.49 Å) observed. Quantitative fitting yields an average coordination number of 3.5, implying a predominant planar Ni–N₄ configuration with partial nitrogen loss during calcination. Previous studies on Ni–N₃ structures have demonstrated that, compared to Ni–N₄ structures, Ni–N₃ configuration primarily reduced the energy barrier in the hydrogenation step (*CO₂ → *COOH) of CO₂RR, with minor impact on the CO₂ adsorption energy^56,57. Our DFT simulations that employed the Ni-N₄ model as a representative configuration can therefore balance computational accuracy with experimental relevance.

The Ni K-edge XANES spectra reveal three primary molecular orbital transitions: 1 s → 3d (3d_z2 back-bonding/3d_x2_-y2), 1 s→4p_z, and 1 s→4p_xy/continuum (Fig. 3c)⁵⁸. While NiPc retained high D₄h symmetry, the synthesized Ni SAC sample exhibited structural distortion, facilitating hybridization between Ni 3d and 4p orbitals⁵⁵. This manifests as intensified 1 s → 3 d transition features and attenuated 1 s→4p_z transitions. Prior investigations of SCN⁻/NO₃⁻ adsorbate poisoning mechanisms suggested that, originating from the electron donation of the Ni 3d_z2 orbital to the unoccupied orbitals of reactants, any stronger impurity adsorption than CO₂ occupying Ni active sites could induce more pronounced splitting between 4p_z and 4p_xy/continuum orbitals⁵⁸. Such interactions produce a redshift in the 1 s→4p_z transition and a concurrent blueshift in the 1 s→4p_xy/continuum transition within XANES profiles. To probe these effects, we performed in-situ XAS measurements on Ni SACs under CO₂RR conditions in both ideal and bio-electrolytes (Supplementary Fig. S20). Notably, the XANES spectra remained invariant under all tested conditions, irrespective of electrolyte composition (Fig. 3e). This indicates that the various complex organic molecules from the bioelectrolyte, unlike other reducible (i.e., NO₃⁻/CO₂) or poisoning (i.e., SCN^-) species, cannot achieve stabilization on the Ni single sites via strong interactions with the Ni 3d_z2 orbital⁵⁸. In other words, their competitive adsorption with CO₂ is likely negligible or, at least, cannot disrupt the orbital hybridization between CO₂ and the Ni single sites under reducing potentials. Given the dynamic nature of electrocatalytic processes and controlled experimental parameters, enhanced EXAFS signal intensity would signify competitive adsorption of any transient poison at Ni sites, while no signal amplification of Ni SAC was detected in the bioelectrolyte system (Fig. 3f). These observations further illustrate that organic constituents from the bioelectrolyte, including those not modeled in our DFT simulations, may accumulate in the electrochemical double layer without forming strong bonds to Ni active sites. Consequently, the adsorption competition between CO₂ and organics from bioelectrolyte on Ni single sites is not significant or even absent, preserving the local coordination geometry and bond lengths of Ni SAC, which aligns with in-situ Raman results and electrochemical performance.

Proof of an integrated electrobiosystem for IPA synthesis

The high activity and CO selectivity of Ni SAC in the bioelectrolyte were validated through both experimental results and mechanistic insights. Building on this, we integrated Ni SAC with the engineered C. ljungdahlii strain within a hybrid iMES system. As illustrated in Fig. 4a, Supplementary Fig. S21, to mitigate potential evaporation losses of EtOH/IPA during extended operation, the bioelectrolyte reservoir was connected to a cold trap to capture any volatilized alcohol products. To counteract pH fluctuations that may occur during long-term iMES operation, 100 mM phosphate-buffered saline (PBS) buffer was added to the bioelectrolyte. The introduction of PBS also enhanced the accessibility of alkali metal cations, thereby reducing the overpotential required to achieve similar levels of CO₂RR activity and selectivity (Supplementary Fig. S22). The introduction of Ni SAC electrode into the hybrid system will not have a significant effect on the cell viability of C. ljungdahlii strain (Supplementary Fig. S2). In addition, the Ni SAC electrode exhibited stable current density and a consistent CO/H₂ ratio for over 100 h in the bioelectrolyte containing 100 mM PBS (Supplementary Fig. S23). After 100 h of operation, post-characterizations (STEM, XPS and ICP, Supplementary Fig. S24–26) demonstrated that the overall structure of the Ni SAC remained stable, and no apparent Ni²⁺ ion leaching was detected in the electrolyte, indicating the high biocompatibility and stability of Ni SAC. Applying lower overpotentials may improve the FE of C₂₊ oxygenates but can also decrease the reducing equivalents supply rate and consequently diminish the overall production rate of C₂₊ oxygenates⁵⁹. To ensure sufficient CO supply and a high CO/H₂ ratio, the iMES system was operated continuously at −0.75 V vs. RHE (Supplementary Note S1).

Fig. 4. Proof concept of iMES for IPA production from CO₂.

a Scheme of a designed iMES system for IPA production. Created in BioRender. Zhou, G. (2025) https://BioRender.com/uaa1ssg. b Current density and OD₆₀₀ during four-day operation. c molar concentrations and d transferred charge and corresponding ERE (orange line) of IPA (red bars), EtOH (blue bars) and Acetate (gray bars). e Performance comparison of IPA production between this work and previous reports. The iMES tests were carried out at a temperature of 37 °C without iR compensation. The error bars represent the standard deviations from at least three independent tests. Source data are provided as a Source Data file.

Over four days of uninterrupted operation, the average current density remained stable at ~11.4 mA/cm², and the optical density (OD₆₀₀) of C. ljungdahlii steadily increased over the first three days, peaking at ~1.36 (Fig. 4b). The slightly decreased current density could be attributed to the loss of gas-diffusion layer (GDL) hydrophobicity over long-term operation⁶⁰. Similar electrochemical performance observed before and after inoculating C. ljungdahlii further illustrated the stable operation and high bioadaptivity of the Ni SAC in the flow cell (Fig. 4b, Supplementary Fig. S23). After accounting the evaporated amount, the average IPA titer reached ~10.8 mM over four days, with a total C₂₊ oxygenate (EtOH + IPA + Acetate) concentration of ~91.1 mM (Fig. 4c). The total electron recovery efficiency (ERE) for C₂₊ oxygenates reached 47.7 ± 6.9% (Fig. 4d). Specifically, the combined ERE for IPA and EtOH exceeded 26.4% (10.5% for IPA and 15.9% for EtOH).

Reports on iMES for IPA production remain scarce, positioning this field as a largely unexplored area. Here, we present the demonstration of CO-mediated IPA iMES. Due to the sufficient CO supply and strong reducing power relative to H₂, our system achieved areal production rate (3.23 mg/cm²/day) and volumetric production rates (161.3 mg/L/day) for IPA, which is competitive to prior studies (Fig. 5e, Supplementary Table S2)^61–63. It is worth noting that a recent work by Liu et al. reported H₂-mediated CO₂ reduction using a different microbial strain and metabolic pathway⁶⁴. The system achieved high rates of PHB generation ( ~ 116.7 mg/L/day) along with C_4-5 alcohols coproduction, and a byproduct of IPA (~100 mg/L/day), highlighting a promising direction that combining alternative strains and metabolic pathways with bioadaptive catalysts to access higher-value products.

Discussion

In this study, we unlocked previously unattainable syngas-mediated CO₂-to-IPA conversion by marrying poison-resistant Ni SAC with engineered microbial fermentation. This study provided proof that Ni SAC can maintain the high CO₂RR selectivity and activity in the complex bioelectrolyte, continuously generating CO to drive downstream IPA biosynthesis. Both In-situ Raman and XAS results illustrated that, compared with conventional Ag NPs, Ni SAC can withstand the competitive adsorption of organics, maintain the atomic coordination structure, as well as preserve active sites for CO₂ adsorption and activation in bioelectrolyte. DFT adsorption simulations also revealed that Ni SAC exhibited relative repulsion toward representative organics than Ag NPs in the microbial growth medium. All the in-situ spectrums and simulations provided a mechanistic insight that can expand fundamental understanding of microbial electrochemical interfaces and future applications of CO₂RR in different electrolytes. The hybrid iMES system achieved an IPA production rate of 161.3 mg/L/day (10.8 A/m²) under ambient conditions, delivering performance that is competitive with previously reported H₂-mediated bioelectrolysis processes, and demonstrating a viable pathway to sustainable chemical manufacturing. Our findings pointed out the priority of “bioadaptive” catalysts for bridging the gap between abiotic electrochemistry and microbial conversion. Looking forward, this strategy can be extended to other valuable products and integrated systems, marking a significant step toward sustainable industrial bioprocessing powered by renewable electricity.

Despite the stable current density observed over a four-day operation, the efficiency and electrolyzer design of iMES in this work need further improvement for large-scale deployment⁶⁵. Notably, at −0.75 V vs. RHE, the total FE of liquid products did not exceed 50%, indicating that nearly half of the electrons remain in CO/H₂. This is primarily because of a mismatch between the generation rate of CO/H₂ and the slower microbial consumption rate^59,66. Looking ahead, future iMES designs employing Ni SAC or other bioadaptive electrodes should prioritize optimizing electron donor mass transfer to microbes, thereby bridging the gap between CO/H₂ production and microbial uptake as well as improving electron utilization efficiency and overall FE toward target products⁶⁶. Besides, another challenge is the unavoidable flooding of GDL, which can disrupt the gas-liquid-solid triple-phase interface, limit CO/H₂ delivery, and reduce long-term electrochemical efficiency/stability⁶⁷. Simultaneously, the co-production of IPA, EtOH, and acetate can lead to product mixing in the bioelectrolyte, which will be continuously flushed by flowing media. This not only dilutes the target products but also imposes significant energy cost and financial pressure on downstream separation^10,68. Therefore, future directions should also aim to enhance electrode hydrophobicity to mitigate flooding and explore integrated one-pot separation or microbial recycling strategies to improve product recovery and enable continuous operation, while above mentioned optimizations are not the focus of this current work.

Methods

Catalyst synthesis

Ni SAC was synthesized by a modified two-step pyrolysis method^55,57. All the chemicals for Ni SAC synthesis, including nickel (II) nitrate hexahydrate (Ni(NO₃)₂ ⋅ 6H₂O, 99.999%), polyvinylpyrrolidone (PVP, (C₆H₉NO)_n, average M_w ~ 1300000), melamine (C₃H₆N₆, >99%), cyanuric acid (CA, C₃H₃N₃O₃, >98%), and isopropanol (IPA, 99.9%), were purchased from Sigma-Aldrich and used without further purification. To synthesize Ni SAC, 6 g melamine, 3 g CA, 10 mg Ni(NO₃)₂ ⋅ 6H₂O, and 1 g PVP were dispersed into 25 mL IPA and ultrasonicated for 1 h. The white dispersion was then dried by a freeze-dryer, and the solid mixture was further ground for 1 h. Thereafter, the precursor powder was carbonized in a tube furnace under Ar atmosphere by a two-stage calcination process: the calcination temperature was first raised from 25 °C to 600 °C with an increasing rate of 3 °C/min and hold for 2 h; Then, the temperature was increased to 900 °C with a rate of 2 °C/min and hold for 1 h.

Electrode fabrication

Considering the long-term operation time of microbial electrosynthesis, a commercial gas diffusion carbon layer (Sigracet 39BB, FuelCell Store) was first modified by the previously reported method⁶⁸ to improve hydrophobicity. 100 mg carbon black and 20 mg Teflon dispersion were dispersed into a 1:1 mixture of IPA/DI water. The mixture was sprayed onto the Sigracet 39BB electrode until a 10 wt% loading was achieved. Then, the modified electrode was placed in a 350 °C furnace for 15 min and cooling to room temperature (20–25 °C). Commercial Ag NPs (US Research Nanomaterials Inc) or synthesized Ni SAC, and 20% Nafion ionomer were first dispersed into a 1:1 mixture of IPA/DI water to make a 12.5 mg/mL catalyst ink. Then, the catalyst ink was sprayed onto the modified gas-diffusion carbon layer to fabricate the working electrode with a catalyst loading of 0.5 mg/cm².

Electrochemical tests

All electrochemical measurements were performed by an electrochemical potentiostat (BioLogic, VMP3). All the electrochemical results were collected by a chronoamperometry mode at a temperature of 37 °C. For the integrated iMES or abiotic CO₂RR test, a purchased mixed metal oxide mesh (MMO, 2 cm * 2 cm, Standard Anode) was used as counter electrode and an Ag/AgCl electrode was used as reference electrode. All the potentials used in this work were converted to the RHE scale by the equation:

$$E\left(vs.RHE\right)=E\left(Ag/AgCl\right)+0.21V+0.0591V*pH$$ 1

Whenever calibration of the reference electrode was required, its potential was calibrated against RHE by measuring the open-circuit potential of a Pt electrode in a H₂-saturated electrolyte. The stabilized open-circuit potential was taken as 0 V vs. RHE, and the resulting offset was used to convert all measured potentials to the RHE scale.

The active area of the working electrode was sealed to 1 cm² by Kapton tape. Nafion 212 membrane (Fuel Cell Earth, 2.54 * 2.54 cm²) was pre-activated and used as CEM to separate the cathodic and anodic chambers. To activate the Nation membrane, the membrane was rinsed with DI water and then was immersed in 3 wt% H₂O₂ solution at 80 °C for 1 h. Thereafter, the membrane was sequentially soaked in DI water and 0.5 M H₂SO₄ at 80 °C for 1 h. Finally, the activated Nation membrane was repeatedly cleaned and stored in DI water. CO₂ (99.999%, Airgas) was introduced to the electrolyzer at 5 sccm controlled by a mass flow controller. Acidic Na₂SO₄ solution (pH ≈ 2) was used as an anolyte. Both catholyte and anolyte were flowed at 1 mL/min controlled by a peristaltic pump. Unless otherwise specified, bioelectrolyte refers to the YT medium, and the ideal electrolyte only contains 100 mM NaCl without any organics. All bioelectrolytes/ideal electrolytes (details as claimed below) were freshly prepared before testing. pH and conductivity of different electrolytes were tested by a pH meter (Fisher Scientific) and a conductivity meter (Fisher Scientific). All the electrochemical and microbial electrochemical tests were independently performed at least three times.

Products quantification

All gas products were tested by a gas chromatograph (GC, Agilent 8890). The liquid products were collected from the cathodic chamber or cold trap and detected by a high-performance liquid chromatograph (HPLC, Agilent 1260). To trap the evaporated liquid products as much as possible, a 15 mL glass tube containing 4 mL DI water at 0 °C was connected to the bioelectrolyte reservoir outlet.

The FE or ERE of a specific product was calculated by the equation:

$${FE}_{x}or{ERE}_x=\frac{n_x*electrontransfernumber*F}{Electrodearea*I*t}*100\%$$ 2

Where n_x is the production amount (mol) of product X, determined by GC or HPLC; F is the Faradaic constant, 96485.33 C/mol; I is the total current density (mA/cm²). Although the equation is the same, we use FE and ERE to indicate direct electrochemical products (CO/H₂) and indirect microbial products (IPA/EtOH/Acetate).

Bacterial strains and growth media

The IPA-producing C. ljungdahlii strain used in this study was created by Lo et al.¹⁷. Growth of C. ljungdahlii was in YTF-rich medium consisting of 10 g/L Bacto yeast extract, 16 g/L Bacto tryptone, 4 g/L NaCl, 0.5 g/L cysteine-HCl, with the addition of 5 g/L fructose when strains were grown from frozen stocks. Cell growth was monitored at 600 nm using a NanoDrop spectrophotometer (ThermoFisher Scientific, Waltham, USA).

Autotrophic growth of C. ljungdahlii for syngas vs CO₂/H₂ comparisons

Frozen stocks of the IPA-producing strain of C. ljungdahlii were inoculated into 10 mL of YTF containing fructose in a 15 mL Falcon tube. Growth overnight was at 37 °C inside a COY (Grass Lake, MI) anaerobic growth chamber, flushed with 95% N₂ and 5% H2 and maintained anaerobic by palladium catalyst. Overnight cultures were added to 50 mL of YT (no fructose) in a 250-mL Duran pressure plus bottle (DWK Life Sciences, USA), pressurized to an additional 10 PSI above ambient with a syngas mix of 50%/20%/30% CO/H₂/CO₂ and left to grow shaking at 200 RPM at 37 °C. Bottles were regularly pressurized back until the OD₆₀₀ reached 1. Afterwards, C. ljungdahlii was transferred to fresh 50 mL YT medium in triplicate and grown either in a syngas mix of 50%/20%/30% CO/H₂/CO₂ or 20%/80% CO₂/H₂ pressurized to an additional 10 PSI above ambient again in 250-mL Duran pressure plus bottles. Bottles were routinely repressurized until growth ceased. For the cell viability assays, C. ljungdahlii was grown in 250-mL Duran pressure plus bottles in a syngas mix of 50%/20%/30% CO/H₂/CO₂ as previously described. Colony-forming units (CFU) per mL were assessed by taking 1 mL of culture from the bottles on selected days, serially diluting the culture to provide sufficient resolution to determine colonies and plating on YT + 5 g/L fructose medium. Colonies were counted after 5 days of growth on plates.

Growth of C. ljungdahlii in the biohybrid system

Before starting the iMES process, all the components (e.g., electrolyzer, bottles, connectors, and tubes) were autoclaved and sterilized. CO₂ and electrolytes were flowed for at least 1 h to keep the anaerobic condition in the electrolyzer. A 0.22 μm filter was used between the gas flow controller and the electrolyzer to remove any possible contamination. Before adding the electrochemical potential (−0.75 V vs. RHE), seeds for the iMES system were prepared as above, and the initial OD₆₀₀ was ~0.5. The operational temperature is 37 °C. Liquid samples were collected from the bioelectrolyte reservoir once per day for each experiment. After finishing the long-term operation (4 days), all the liquids from the cold trap were collected for further detection. The calculations of volumetric production rate were based on the bioelectrolyte volume in the reservoir. All the results were averaged by at least three independent experiments.

Material characterizations

High-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) images, energy-dispersive X-ray spectroscopy, and elemental mapping were obtained from Titan Cubed Themis 300 double Cs-corrected STEM operated at 300 kV and equipped with an extreme field emission gun source and a super-X energy-dispersive spectrometry system. X-ray photoelectron spectroscopy (XPS) data were collected by a Thermo Fisher K-alpha XPS/UPS with a monochromatic Al K-alpha x-ray source at 15 kV. All XPS results were corrected by the carbon 1 s (assigned to 284.8 eV). X-ray diffraction (XRD) data were collected by a Bruker D8 Discover X-Ray Diffractometer. ICP-OES was conducted by an Agilent 5800 ICP-OES system.

In-situ Raman spectroscopy measurement

In-situ Raman spectroscopy measurement was carried out in a customized in-situ PTFE cell. A classy carbon electrode was used as the working electrode. The active area of the working electrode is ~0.2 cm² (5 mm diameter). An Ag/AgCl electrode and a Pt wire were used as the reference electrode and counter electrode, respectively. The FAB-PK-130 membrane was used to separate the cathodic and anodic chambers. Au@SiO₂ was used for enhancing the surface-enhanced Raman scattering of Ni SAC⁵⁰. 0.1 mg/cm² of Au@SiO₂ was first dropped onto the working electrode, and 0.5 mg/cm² of Ni SAC was then loaded. The in-situ Raman results were collected by a Horiba Aramis spectrometer (Imagin Analysis Center, Princeton University) with 785 nm excitation. The exposure time was set as 5 s, and all the Raman results were averaged at least 5 times of collection.

In-situ XAS spectroscopy measurement

Ni K-edge X-ray absorption spectroscopy (XAS) was measured at 8-ID at the National Synchrotron Light Source II (NSLS-II) at Brookhaven National Laboratory (BNL). For the ex-situ experiments, the samples were ground into a fine powder, pressed into a pellet, and sealed between Kapton tape before being scanned. The samples were scanned in their received condition at the Ni K edge. Two ion chambers with a Ni foil in between were placed behind the sample to calibrate the energy. For the in-situ XAS experiments, a 2 cm² carbon paper loaded with 0.5 mg/cm² Ni SAC was used as the working electrode. An Ag/AgCl electrode was used as the reference electrode, and a graphite rod was used as the counter electrode. In-situ XAS was tested in a customized single-chamber XAS electrolyzer (Anhui Absorption Spectrum Equipment Co., LTD). Bioelectrolyte or ideal electrolyte was saturated by CO₂ and flowed into the electrolyzer at a rate of 1 sccm. XAS data were fitted using the winXAS 3.1 software package (Supplementary Table 3). Least-squares fit in R-space of the k²-weighted Fourier transform data from 2.7 to 10.0 Å⁻¹ and was used to obtain the EXAFS coordination parameters. The first shell was used to fit the EXAFS spectra. The amplitude reduction factor (S₀²) was determined by fitting a reference spectrum of the Ni foil (S₀² = 0.8). Optimization of sigma-squared was obtained from fitting the isolated peaks in k-space.

DFT methods

All calculations in this work were performed using the Vienna Ab-Initio Simulation Package (VASP)⁶⁹, where the RPBE functional was used⁷⁰. The D3 correction method⁷¹ was employed to illustrate the long-range dispersion interactions between the adsorbates and surfaces. The core electrons were described with the projector augmented wave (PAW) method⁷². The convergence criteria for electronic and force minimization are set to 10⁻⁶ eV and 0.02 eV/Å for structural optimization. The cutoff energy for the kinetic energy of the plane-waves was 450 eV. The Gibbs free energy for gas-phase molecules and that for adsorbed species on the surface are obtained from the DFT energies with ZPE and entropy corrections determined from frequency calculations using the harmonic oscillator approximation. The Ag (111) surface was modeled by a 3-layer (5 × 4) supercell. The bottom two layers of the slab were constrained as the bulk region, and the top layer and adsorbates were allowed to relax as the interface region. The Brillouin zone was sampled using the 3 × 4 × 1 and 4 × 4 × 1 Gamma-centered k-point grids for the Ag(111) surface and the NiNC surface, respectively. The linearized Poisson-Boltzmann implicit solvation model implemented in VASPsol is used to represent the electrolyte region⁷³. The dielectric constant of water, 78.4, and the Debye screening length corresponding to 1 M concentration of electrolytes, 3.0 Å, were used. As the exact catalytic material under operating conditions is difficult to model, the conclusions drawn from our DFT calculations should be treated with caution and regarded as qualitative. Our simplified, static models do not fully account for surface reconstruction, explicit solvent effects like H-bond interactions, or coverage-dependent phenomena; therefore, the DFT results are intended only as supportive guidance and plausible mechanistic scenarios, rather than definitive evidence.

Author contributions

G. Zhou, J.R. Humphreys, and Z.J. Ren. conceived the idea with input from co-authors. G. Zhou and J.R. Humphreys conducted electrochemical and microbial experiments. D. Cheng and G. Zhou contributed to DFT simulations and analyses. G. Zhou, S. Jiang, W.-L. Huang, and J.T. Miller contributed to ex-situ/in-situ XAS experiments and analyses. G. Cheng, G. Zhou, H. Chen, and N. Yao performed other ex-situ/in-situ characterizations and analyses. G. Zhou, J.R. Humphreys, and Z.J.R. wrote the draft, and other authors edited and commented on the manuscript.

Peer review

Peer review information

Nature Communications thanks Dawei Liang and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

The DFT and supporting data generated in this study are provided in the Supplementary Information/Source Data file. Figures 1a, 2f and 4a were created by Microsoft PowerPoint and BioRender, all confirmed for commercial publication. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-026-68358-8.