Introduction

Carbon materials have played a pivotal role in human innovation, from ancient civilizations to modern nanotechnology1,2. Recent advances in energy conversion and storage technologies (e.g., fuel cells, metal-air batteries, water or CO2 electrolysis, biomass valorization) have renewed interest in carbons due to their enabling roles as electrocatalytic materials3. Conventional synthesis of these carbon-based electrocatalysts, however, typically involves a multi-step process. This often includes energy-intensive, high-temperature pyrolysis (800-1200 °C) in a tube furnace under controlled atmospheres for several hours, followed by acid leaching and sometimes a second pyrolysis step to enhance active site density4,5. Alternative methods, such as chemical vapor deposition, flash Joule heating, plasma carbonization, and ultrasonic spray pyrolysis, often have to be coupled with additional thermal treatment or require complex and costly equipment6,7,8,9. Recent approaches have sought to circumvent these limitations by harnessing chemical energy for transformations. Among these, hypergolic reactions, where separate fuel and oxidizer components spontaneously ignite upon contact, represent an effective strategy for the rapid, ambient-condition synthesis of carbon nanostructures10,11,12. In parallel, explosion- or shock-wave-assisted methods have also been developed, though they typically employ a top-down approach starting from bulk graphite, which can limit chemical tunability9,13,14,15.

Building on the concept of chemical energy-driven synthesis, we introduce a rapid and energy-autonomous strategy using a polyaniline-HClO4 composite as a single, self-contained precursor. The stored chemical energy is liberated through brief ignition triggered by mild heating (≈120 °C), microwave irradiation, or mechanical stimulus at ambient temperature. This ignition drives the simultaneous exfoliation and carbonization of the polyaniline matrix, directly yielding two-dimensional (2D) amorphous carbon nanosheets with a high surface area and an interconnected porous network. Unlike typical bipropellant hypergolic systems that require separate fuel and oxidizer, our monopropellant-inspired design integrates both roles into one stable solid. Upon triggering, the decomposition of HClO4 generates intense local heat for instantaneous carbonization while simultaneously producing a violent gas release that acts as a dynamic exfoliation force, forcibly separating the polymer into large-area nanosheets. The entire process is simple, conducted in a round-bottom flask without complex instrumentation or post-treatment, and completes within 2-7 min under ambient conditions. Moreover, in contrast to top-down approaches that limit chemical tunability, our bottom-up precursor synthesis enables facile molecular-level doping. We demonstrate this versatility by flexibly incorporating transition metal dopants (e.g., Fe, Co, Ni, Cu) during precursor preparation, creating single-atom-site tailored for high-performance electrocatalysis in the oxygen reduction, H2O2 production, and CO2 reduction reactions.

Results and discussion

Energy-autonomous carbonization of PANI into POP-C

In a typical synthesis, polyaniline in its emeraldine salt form (PANI) is first prepared using an organic/aqueous interfacial polymerization method with HClO4 as the acid dopant (Supplementary Fig. 1)16,17. The obtained PANI is then loaded into a round bottom flask fitted with a thermometer, covered with quartz wool to permit outgassing (Supplementary Fig. 2). Subsequent mild heating of the sample for 2-7 min initiates the energy-autonomous carbonization process, which culminates in a brief ignition (lasting ≈0.4 s) and a violent release of gas (Fig. 1b). The process is like making popcorn, as shown in Fig. 1a–c and Supplementary Movies 1 and 2. The minimum temperature required to trigger this popping reaction depends on the heating rate, being around 86 and 121 °C at heating rates of 30 and 15 °C min−1, respectively (Supplementary Fig. 2). The instantaneous local reaction temperature exceeds 1400 °C, as confirmed by the melting of a K-type thermocouple (Supplementary Fig. 3). The mass of the sample decreases by ≈90% after popping, while its volume expands significantly (Fig. 1a, c). This popping reaction can also be induced by microwave irradiation (Fig. 1d–f, Supplementary Movie 3 and Supplementary Note 1) or even grinding at room temperature (Supplementary Fig. 4 and Supplementary Note 1). Scanning electron microscopy (SEM) analysis indicates that the dense aggregate of PANI precursor are exfoliated into interconnected 2D nanosheets after the rapid popping reaction (Fig. 1g, h and Supplementary Fig. 5). Raman spectra confirm that the PANI precursor is completely converted into carbon (denoted as POP-C), as evidenced by the clear D and G bands characteristic of carbon, without any trace signal corresponding to PANI (Fig. 1i)18,19. The broaden and overlapping D and G bands suggest that POP-C possesses abundant defects and a highly disordered structure, likely due to the incorporation of N in the carbon sp2 networks20.

Fig. 1: Synthetic process of POP-C from PANI.
Fig. 1: Synthetic process of POP-C from PANI.The alternative text for this image may have been generated using AI.
Full size image

a–c Optical images of the PANI precursor (a) before, (b) during and (c) after the popping reaction triggered by thermal heating. d–f Optical images of PANI precursor (d) before, (e) during and (f) after the popping reaction induced by microwave. g–h, SEM images of (g) pristine PANI and the resulting (h) POP-C. Images in (g) and (h) are representative of 3 independent measurements with similar results. i Raman spectra of PANI (green line) and POP-C (black line). Source data are provided as a Source Data file.

Transmission electron spectroscopy (TEM) analysis (Fig. 2a) reveals a wrinkled, interconnected network structure of POP-C formed by thin 2D nanosheets. Selected area electron diffraction (SAED) analysis of POP-C shows diffuse ring patterns typical of long-range disordered amorphous materials (inset in Fig. 2a). High resolution TEM (HRTEM) imaging (Fig. 2b) further confirm the ultrathin monolayer structure of POP-C, while a fast Fourier-transformation (FFT) analysis of the HRTEM image (Fig. 2c) identifies distinct of 5-, 6-, and 7-membered carbon rings within the nanosheets. X-ray diffraction (XRD) pattern (Supplementary Fig. 6) shows broadening of the (002) diffraction peaks at a diffraction angle of 2θ = 21.5°, corresponding to an interlayer spacing of 0.41 nm, confirming the short-range ordered and expanded structure of POP-C21. These results suggest that the as-synthesized POP-C would feature a short-range ordered and long-range disordered structure, consisting of wavy 2D amorphous carbon nanosheets with some stacking7,22,23. This loose and open structure endows POP-C with a large surface area, as confirmed by both Brunauer-Emmett-Teller (BET) surface area analysis (> 900 m2 g−1) (Fig. 2d) and electrochemically active surface area analysis (400 m2 g−1) (Supplementary Fig. 7). Argon sorption isotherm and pore size distribution analyses indicate the presence of micro- and mesopores (Fig. 2d).

Fig. 2: Morphological and structural characterizations of POP-C.
Fig. 2: Morphological and structural characterizations of POP-C.The alternative text for this image may have been generated using AI.
Full size image

a TEM image with a SAED pattern (inset). b HRTEM image of the 2D carbon nanosheet. Images in (a) and (b) are representative of 3 independent measurements with similar results. c FFT filtered HRTEM image of the region outlined in (b). d Argon sorption isotherm and corresponding pore size distribution (inset). e Neutron pair distribution function G(r). f XPS spectrum of the C 1 s core level. g XPS spectrum of N 1 s core level. h Carbon K-edge EELS. i Carbon K-edge XANES spectrum. Source data are provided as a Source Data file.

The atomic structure of POP-C was further studied using neutron pair distribution function (PDF) analysis24,25,26, which resolves the average local environment of carbon atoms. As shown in Fig. 2e, the first peak in the PDF occurs at 1.41 Å, close to the value found for sp2 carbon. There is no evidence of a peak around 1.54 Å corresponding to four-fold coordinated carbon atoms24, implying that three-fold coordinated carbon bonding is dominant in the as-synthesized POP-C. The C − C − C bond angle obtained from the ratio of the first and second neighbor distances is determined to be 120.6o, again confirming the dominant sp2 carbon nature of POC-C27. Nevertheless, the vanishing peak in the long-distance range suggests the long-range disordered structure of POP-C. These results imply that the POP-C would likely possess a honeycomb lattice structure comprising three-fold coordinated sp2 carbons, while also featuring significant structural disorder due to defects or vacancies. The surface composition of POP-C was investigated by X-ray photoelectron spectroscopy (XPS). The XPS spectra (Supplementary Fig. 8) reveal a high surface content of nitrogen (6.8 at.%) in POP-C. The deconvoluted high-resolution XPS spectra of C 1 s suggests a dominant sp2 carbon with C − C bonds (Fig. 2f), while the N 1 s spectra can be deconvoluted into three major peaks corresponding to pyridinic, pyrrolic and graphitic nitrogen (Fig. 2g)28,29. The electronic structure of the POP-C was also investigated by electron energy loss spectroscopy (EELS) and X-ray adsorption near edge spectroscopy (XANES). The carbon K-edge EELS spectra shows a clear peak at 285 eV corresponding to the electron transition from the 1 s to π* orbital (Fig. 2h), characteristic of sp2 carbon30. A band starting at 292.5 eV is assigned to transitions to the p-orbitals merging with part of the broad σ* band in a graphitic structure. Similar result can also be obtained with XANES. As shown in the carbon K-edge XANES spectra of POP-C (Fig. 2i), the peak of 285.5 related to the C − C π* transition from sp2 hybrids and a peak around 292.4 eV related to the C − C σ* transition from sp2 carbon28,31,32. Different from the EELS spectra, an additional peak at around 288.3 eV was observed, which can be attributed to π* C − O/C − N, π* N − C = N resonances33,34,35. A distinction between the C K-edge EELS (Fig. 2h) and XANES (Fig. 2i) spectra is evident in the sharpness of the sp2 C = C π* feature near 285 eV. EELS reveals a well-defined π* peak, whereas XANES shows a broadened signal. This difference arises from the distinct probing scales and averaging effects of these two techniques. TEM-EELS probes local, well-ordered sp2 domains within individual nanosheets. In contrast, XANES provides a macroscopic average, where the sp2 C = C π* transition (≈285.5 eV) overlaps with contributions from abundant C − N and C − O groups (≈288.3 eV) in disordered regions. Therefore, the sharp EELS signal confirms the existence of well-developed sp2 networks at the nanoscale, while the broader XANES signal reflects the chemically and structurally diverse, functionalized nature of the material on a macroscopic scale. Nitrogen K-edge spectra shows three pronounced π* resonances at 398.4, 399.8 and 401.2 eV (Supplementary Fig. 9), which are generally attributed to pyridinic, pyrrolic and graphitic forms of nitrogen, respectively28,29, consistent with the XPS measurements. The oxygen K-edge spectra exhibit two main features, resonance at 531.9 eV and 536.7 eV corresponding to π* and σ* renounces of C = O and C − O groups, respectively36. These results confirm that the POP-C would feature a predominant sp2 carbon structure, with nitrogen dopants and oxygen functional groups in diverse configurations.

Generic synthesis of POP-C with single-atom-site

To demonstrate the generic applicability of the proposed approach, POP-C with single-atom-site (M-POP-C) was synthesized through the popping reaction. Four typical transition metals (M = Fe, Co, Ni, Cu) were selected to incorporate into the PANI to form M-PANI precursors. The detailed preparation procedure is illustrated in Supplementary Fig. 10. Infrared spectroscopy (Supplementary Fig. 11) indicates that metal ions coordinate with nitrogen atoms of PANI to form M-PANI complexes37, rather than simply undergoing physical mixing, as elaborated in Supplementary Note 2. Analogous to the preparation of POP-C, heating (15 °C min−1) triggers the popping of M-PANI complexes at ≈120 °C, resulting in the formation of M-POP-C (M = Fe, Co, Ni, Cu). The metal contents in M-POP-C are less than 1 wt.% according to inductively coupled plasma mass spectrometry (ICP-MS) analysis (Supplementary Table 1). Raman spectra (Supplementary Fig. 12) show that M-POP-C samples exhibit comparable characteristics to POP-C, confirming their defect-rich carbonaceous nature. SEM and TEM images (Fig. 3a–c and Supplementary Figs. 13 and 14) reveal that M-POP-C retains the 2D nanosheet morphology of POP-C. XRD patterns (Supplementary Fig. 15) show no diffraction peaks of either metal or metal oxide, indicating the absence of large metal or metal oxide crystals in M-POP-C. The double layer capacitance of four M-POP-C is comparable to POP-C, indicating that incorporating metal has little impact on the exposed surface area (Supplementary Fig. 16). XPS analysis (Supplementary Figs. 17-20) indicates that the major content of carbon is in the sp2 state, and nitrogen has been successfully doped into the carbon. The XPS N 1 s spectra (Supplementary Figs. 17-20) shows a higher peak density at 398.8 eV compared to POP-C (Supplementary Fig. 8), suggesting the presence of metal-nitrogen bonds29. The chemical and electronic structure of the metals in M-POP-C were probed by synchrotron-based X-ray adsorption (XAS) analysis. The Co K-edge XANES spectra of Co-POP-C and reference samples (Fig. 3d), and the first derivative XANES for Co-POP-C (Supplementary Fig. 21) clearly indicate that the valence state of Co in Co-POP-C is ≈+238. The Fourier transform (FT) k2-weighted extended XAFS (EXAFS) spectrum of Co-POP-C (Fig. 3e) features one primary peak at 1.72 Å, attributed to the backscattering between Co and light atoms. No obvious Co−Co signal (2.19 Å) or other high-shell peaks can be found, verifying that Co atoms exist exclusively in an atomically dispersed state38,39. Wavelet transforms (WT) analysis of the Co K-edge EXAFS oscillations of Co-POP-C and reference samples (Fig. 3f) reveals a single intensity maximum at ≈5.8 Å−1, corresponding to Co−O or Co−N contribution in the first shell. No intensity maximum corresponding to Co−Co was detected, further confirming the atomic dispersion of cobalt atoms38,40. Similar results were obtained for POP-C doped with other metals, as detailed in Supplementary Note 3. These results establish the general applicability of the energy-autonomous carbonization for synthesizing carbon materials incorporating diverse single-atom-site.

Fig. 3: Morphological and structural characterizations of Co-POP-C.
Fig. 3: Morphological and structural characterizations of Co-POP-C.The alternative text for this image may have been generated using AI.
Full size image

a SEM image of Co-POP-C. b TEM image of Co-POP-C. c HRTEM image of Co-POP-C. Images in (a)–(c) are representative of 3 independent measurements with similar results. d Normalized Co K-edge XANES spectra of Co-POP-C and reference samples. e k2-weighted Fourier-transform (FT) Co K-edge EXAFS spectra of Co-POP-C and reference samples. In (d) and (e), colors denote: Co foil (yellow), CoO (pink), CoPc (green) and Co-POP-C (blue). f Wavelet transforms for the k3-weighted Co K-edge EXAFS signals of Co-POP-C and reference samples. Source data are provided as a Source Data file.

Electrocatalytic performance of POP-C and M-POC-C

The structure of POP-C and the single-atom-site characteristic of M-POP-C, which are potential active sites for a range of catalytic processes, prompted us to assess their electrocatalytic capabilities. We conducted an evaluation of the synthesized POP-C and M-POP-C in relation to their performance in two prominent electrocatalytic reactions: the oxygen reduction reaction (ORR) and the carbon dioxide reduction reaction (CO2RR). The electrocatalytic performance of POP-C and M-POP-C for ORR was evaluated in both acidic and alkaline electrolyte (Fig. 4a–d). It is well documented that ORR can proceed either through a 4-electron pathway to produce water, or a 2-electron pathway to produce H2O2. We first measured the ORR performance in 0.1 M KOH electrolyte (Fig. 4a, b), revealing that ORR proceeds predominantly through the 2-electron pathway over metal-free POP-C, as reflected by its high H2O2 selectivity (70%-90%), which is significantly higher than M-POP-C. Long-time bulk electrolysis over POP-C at 0.67 V for 15 h (Supplementary Fig. 25a) shows an average Faraday efficiency (FE) of 87% and a high H2O2 production rate of 175 mmol g−1 h−1. Incorporating transition metal dopants significantly impacts the ORR pathway. Specifically, both Co-POP-C and Fe-POP-C exhibit low selectivity toward H2O2 production (<25%), indicating that ORR would proceed mainly through the 4-electron transfer pathway. The polarization curves (Fig. 4a) disclose that the Co-POP-C and Fe-POP-C display a half-wave potential of 0.80 and 0.78 V, respectively. The ORR activity of M-POP-C can be further improved by increasing the metal content. As shown in Supplementary Fig. 26, the half-wave potential of Co-POP-C with a higher Co content (Co-POP-C(H)) reaches 0.82 V. To evaluate the 4-electron ORR activity in a practical device, Co-POP-C was assembled into an Al-air battery as cathode catalyst. As shown in Supplementary Fig. 27, this battery demonstrates an open circuit voltage of 1.9 V and a peak power density of 160 mW cm−2. During discharge at 20 mA cm−2, the battery maintains a stable discharge voltage of 1.5 V and delivers a specific capacity of 950 m Ah g−1. This performance surpasses commercial Pt/C and is comparable to recently reported non-precious metal catalysts. In 0.1 M HClO4 electrolyte (Fig. 4c, d), only Fe-POP-C can catalyze the 4-electron ORR, while Co-POP-C produces H2O2 over a wide potential range with maximum H2O2 selectivity close to 100%. Long-time bulk electrolysis at 0.51 V over Co-POP-C (Supplementary Fig. 25b) shows an average FE of 91% and a H2O2 production rate of 128 mmol g−1 h−1.

Fig. 4: Electrocatalytic properties of POC and M-POP-C.
Fig. 4: Electrocatalytic properties of POC and M-POP-C.The alternative text for this image may have been generated using AI.
Full size image

a–d ORR performance evaluated via rotating ring disk electrode (RRDE) measurements. Polarization curves for the ORR and H2O2 production in O2 saturated (a) 0.1 M KOH and (c) 0.1 M HClO4 solutions. Corresponding H2O2 selectivity derived from the ring and disk current densities in (b) 0.1 M KOH and (d) 0.1 M HClO4 solutions. In (a)–(d), colors denote: POP-C (orange), Fe-POP-C (yellow), Co-POP-C (blue), Ni-POP-C (green), and Cu-POP-C (purple). e, f CO2RR performance evaluated by 1 h electrolysis in CO2 saturated 0.5 M KHCO3 solution. e FE of the primary products for POP-C and M-POP-C at optimized potentials. f Faraday efficiency (FE) of CO and H2 for Ni-POP-C across varying potentials. In (e) and (f), colors denote: H2 (blue), CO (orange), HCOOH (green) and C2H5OH (pink). In (e) and (f), data are presented as mean ± standard deviation. The center of the error bars represents the mean of 3 independent samples. Source data are provided as a Source Data file.

We further investigated the electrocatalytic performance of POP-C and M-POP-C for CO2RR. As shown by the linear scan voltammograms (Supplementary Fig. 28), POP-C presents relatively low reduction current density, with hydrogen as the main product over a wide potential range (Supplementary Fig. 29). In comparison, M-POP-C shows higher reduction currents, accompanied by pronounced selectivity toward various C1 and C2+ products. As shown in Fig. 4e and Supplementary Fig. 30, Cu-POP-C exhibits higher selectivity toward formate formation. CO is a major gas product over both Fe-POP-C and Co-POP-C. For liquid products, Fe-POP-C tends to promote ethanol formation, while Co-POP-C exhibits activity toward ethanol and formate formation. Among various M-POP-C catalysts, Ni-POP-C presents high CO selectivity over a wide potential range (over 90% from −0.7 to −0.85 V), with a maximum CO FE of 98.1% at −0.8 V (Fig. 4f). These findings clearly demonstrate the electrocatalytic performance of POP-C materials, and more importantly their tunability by varying the metal dopants’ content or identity in a task-specific manner.

Mechanism of the energy-autonomous carbonization

Our systematic investigation (Supplementary Note 4) identified the HClO4 content within the PANI precursor as the key factor governing the energy-autonomous popping reaction. HClO4 exists in two distinct states within PANI (Fig. 5a), i.e., doped HClO4, that is ionically bound to the PANI backbone, and free HClO4 that is physically trapped within the polymer matrix. The total HClO4 content can be tuned by adjusting the acid concentration during PANI synthesis (Fig. 5b) and precisely quantified using a titration method (Supplementary Note 4). Crucially, PANI containing only doped HClO4 or a very low free HClO4 content (<4%) fails to undergo the popping reaction (Supplementary Figs. 44 and 48). Quantitative analysis confirms the role of free HClO4 as the primary driver, as the reaction initiation temperature decreases inversely with increasing free HClO4 content (Fig. 5c and Supplementary Fig. 45). The collected product yield also demonstrates a trade-off, decreasing at excessively high free HClO4 contents (Fig. 5c) due to the increased reaction violence (Supplementary Fig. 45), which physically expels material from the reaction vessel. Raman spectroscopy confirms the formation of carbonaceous materials across the investigated range of free HClO4 contents (Supplementary Fig. 46). However, the resulting carbon morphology is critically dependent on the concentration of free HClO4. Precursors with low free HClO4 content (12%) only yield agglomerated carbon nanoflakes (Fig. 5d). As the content increases to 17%, the generation of internal gas pressure is sufficient to initiate partial exfoliation, resulting in the appearance of exfoliated 2D nanosheets (Fig. 5e). When the free HClO4 content exceeds 21%, the products adopt a fully developed, interconnected 2D nanosheet morphology (Fig. 5f-h). This morphological progression provides direct visual evidence that free HClO4 is the critical component that both drives the explosive reaction and generates the critical temperature/pressure required for forming the 2D carbon nanosheets.

Fig. 5: Key factor governing the energy-autonomous popping reaction.
Fig. 5: Key factor governing the energy-autonomous popping reaction.The alternative text for this image may have been generated using AI.
Full size image

a Schematic of the reactive PANI precursor, illustrating the chemical distinction between doped and free HClO4. b Quantification of total HClO4 (brown), free HClO4 (orange) and doped HClO4 (blue) content within the PANI precursor synthesized using HClO4 solutions with different concentrations. c Reaction triggering temperature (yellow) and POP-C yield (green) as a function of free HClO4 content. In (b) and (c), data are presented as mean ± standard deviation. The center of the error bars represents the mean of 3 independent samples. d–h TEM images of POP-C products synthesized from PANI with free HClO4 contents of (d) 12%, (e) 17%, (f) 21%, (g) 33%, and (h) 47%. Images in (d)–(h) are representative of 3 independent measurements with similar results. Source data are provided as a Source Data file.

Based on these findings, we propose a synergistic mechanism for the energy-autonomous carbonization and exfoliation of PANI. As illustrated in Supplementary Fig. 53, the transformation requires a PANI emeraldine salt containing a critical amount of free HClO4. Upon heating, HClO4 decomposes, initiating a sequence of events. First, the decomposition generates potent radical species and oxidants, such as ClO2, ClO, and O241, triggering a violent, highly exothermic oxidation of PANI, which has long been recognized as an effective radical scavenger/heat insulator, and can be used to desensitize nanothermites42,43,44. The decomposition of perchlorate species also concomitantly releases a large volume of gases (e.g., Cl2, O2, CO2, H2O). The instantaneous release of energy drives rapid dehydrogenation and carbonization, with local temperature confirmed to exceed 1400 °C (Supplementary Note 4). Simultaneously, the violent gas evolution creates a critical internal pressure that physically exfoliates the carbonizing polymer matrix from the inside out, while the in-situ gas-bubble templating directly yields the characteristic 2D porous nanosheet morphology with an open, interconnected pore network (Supplementary Fig. 47). This synergy between intense localized heating and explosive gas expansion culminates in a thermal runaway event, observed as a rapid “pop”. While the reaction initiation temperature and overall POP-C yield are governed by external factors (e.g., PANI packing density and amount, heating rate, water content) as detailed in Supplementary Note 4, the formation of the desired 2D nanosheet morphology critically requires a high free HClO4 content (mass content >30%), which ensures sufficient gas volume and exothermicity for complete carbonization and exfoliation.

To be noted, conventional pyrolysis, which relies on external energy sources and involves a gradual temperature increase alongside the sequential decomposition of organic precursors, typically results in carbons that retain the initial morphological structure of their precursors, as detailed in Supplementary Note 5. In stark contrast, the popping reaction described herein is characterized by the decomposition of energetic components in the PANI precursor, triggering rapid carbonization in less than one second. Concurrently, vigorous outgassing promotes the formation of 2D carbon nanosheet. The popping process exhibits strong potential for scale-up due to its speed, low external energy demand, and tunable reaction intensity, which can be moderated via precursor water content, as detailed in Supplementary Notes 4. Moreover, the yield of POP-C from the popping reaction ranges from 17% to 40% depending on reaction conditions, which is comparable to that of conventional pyrolysis (37%), as detailed in Supplementary Notes 4, 5.

To elucidate the dynamic growth mechanism of POP-C from PANI, we conducted large-scale atomistic simulations with the reactive force field (ReaxFF) method45,46. The experimental popping reaction was replicated by rapidly heating PANI molecules to 3500 K at ramping rate of 10 K ps−1, followed by a 2.0 ns hold to accelerate carbonization kinetics and structural evolution. Time-lapse simulation snapshots (Fig. 6a–c; Supplementary Fig. 54 and Supplementary Movie 4) reveal that the benzene rings undergo rapid dehydrogenation and ring opening under the simulated high-temperature conditions, followed by cross-linking into carbonaceous networks. The final structure closely resembles the experimentally synthesized POP-C, characterized by a distinctive 2D nanosheet morphology replete with defects. Figure 6d tracks the evolution of 5-, 6-, and 7-membered carbon ring, identifying two distinct reaction phases. During the initial phase (0-20 ps), over 90% of 6-membered ring disintegrates, coinciding with a marked increase in 5- and 7-membered ring. In the subsequent phase, 6-membered ring formation resumes while 5- and 7-membered ring growth plateaus, ultimately yielding carbon nanosheets with a majority of 6-membered ring, as illustrated in Fig. 6e, f. These simulations underscore the critical importance of rapid high-temperature treatment in the synthesis of 2D amorphous carbon architectures.

Fig. 6: Formation mechanism of POP-C.
Fig. 6: Formation mechanism of POP-C.The alternative text for this image may have been generated using AI.
Full size image

a–c Simulated time-lapse snapshots of PANI carbonization: a initial, b intermediate and c final configurations. Scale bar indicates 0.5 nm and is applicable to (a)–(c). d Temporal evolution of 5-membered ring (green), 6-membered ring (orange) and 7-membered ring (blue) as a function of simulation time. e, f Atomic structure of the simulated amorphous carbon layer, visualized from (e) side- and (f) top-view perspectives. Source data are provided as a Source Data file.

In summary, we present a rapid, energy-autonomous synthesis of 2D amorphous carbon nanosheets with high surface area and atomically dispersed metal sites. This method leverages the intrinsic chemical energy stored in a polyaniline-HClO4 composite. Upon mild heating, the composite undergoes a popping reaction that simultaneously generates a large amount of heat and gas, carbonizing and exfoliating the polymer into porous nanosheets almost instantaneously. Furthermore, the precursor design enables precise incorporation of transition metals (e.g., Fe, Co, Ni, Cu) during synthesis, allowing the creation of tailored active sites that enhance performance in electrocatalytic reactions. We believe this versatile, energy-efficient approach establishes a pathway for the rapid fabrication of functional 2D carbons, with applications extending from electrocatalysis to energy storage and beyond.

Methods

Materials and reagents

Aniline (≥ 99.5%), toluene (≥ 99.5%), perchloric acid (70%, ACS reagent), ammonium persulfate (ACS reagent), hydrochloric acid (37%, ACS reagent), sulfuric acid (95-97%), ferric nitrate nonahydrate (≥ 99.0%), cobalt nitrate hexahydrate (≥ 99.0%), nickel nitrate hexahydrate (ACS reagent), copper nitrate trihydrate (≥ 99.5%), acetone (≥ 99.9%), potassium hydroxide (pellets, 85%, ACS reagent), potassium iodide (≥ 99.5%), ammonium heptamolybdate tetrahydrate (≥ 99.0%), ammonium nitrate (≥ 95.0%), ammonia solution (28%-30%, ACS reagent), sodium thiosulfate (≥ 97.0%), Nafion solution (5 wt.%), deuterium oxide (99.9%), malic acid (≥ 99.0%) were purchased from Sigma-Aldrich. Potassium bicarbonate (99.997%) were purchased from Alfa Aesar. Ltd. Gas diffusion layer (GDL, 28BC) was purchased from Ionpower. Anion exchange membrane were purchased from Selemion. All solutions were prepared with ultra-pure water (18.2 MΩ cm) from a HiTech water purification system.

Synthesis of PANI and M-PANI nanofibers

PANI nanofibers were synthesized via interfacial polymerization at room temperature (Supplementary Fig. 1). Typically, 12.8 mmol aniline was dissolved in 10 mL toluene (organic phase), while 3.2 mmol ammonium persulfate (APS) was dissolved in 40 mL of 1 M HClO4 (aqueous phase). The organic solution was layered onto the aqueous phase, after which green PANI fibers formed at the interface and migrated into the aqueous phase over 16 h. The PANI fibers were separated from the aqueous phase through centrifugation (2650× g), followed by washing with water, and air-drying at 60 °C for 22 h. The obtained PANI exhibits a dark green color and a sticky state (Supplementary Fig. 1c), containing both residue HClO4 and water. Totally dried PANI fibers without residue HClO4 and water (denoted as PANI-D) was also prepared. To prepare PANI-D, the PANI obtained by interfacial polymerization was separated by centrifugation (2650× g) and then washed with acetone followed by air-drying at 60 °C for 22 h. PANI-D presented a powder-like state with green color (Supplementary Fig. 1d). In order to study the role HClO4 played during the fast exfoliation and carbonization reaction, 1 M HCl and 0.5 M H2SO4 were used as aqueous phase during the interfacial polymerization process to produce polyaniline (Supplementary Fig. 31). The obtained wet samples were denoted as PANI-HCl and PANI-H2SO4, respectively. M-PANI nanofibers were synthesized similarly to PANI, with metal nitrate dissolved in the aqueous phase, as shown in Supplementary Fig. 10. Typically, for the synthesis of Co-PANI, 12.8 mmol aniline was dissolved in 10 mL toluene (organic phase), and 3.2 mmol APS together with 0.4 mmol Co(NO3)2 were dissolved in 40 mL of 1 M HClO4 (aqueous phase) in a 100 mL beaker. Then the organic solution was layered onto the aqueous phase, followed with 16 h reaction. The separation, washing and drying were conducted in the same way as PANI.

Synthesis of POP-C and M-POP-C

A 100 mL three-neck round-bottom flask (Lenz Laborglasinstrumente, DURAN glass; see Supplementary Fig. 2) was loaded with 200 mg of PANI (or M-PANI). A thermocouple was inserted through one side neck to monitor the temperature at the bottom of the flask. The main neck was loosely sealed with quartz wool to allow pressure release, as the reaction generated a sudden pressure increase. The flask was heated in a heating mantle at a rate of 15 °C min−1 from room temperature. Upon reaching the reaction temperature (110–130 °C), an abrupt popping reaction occurred inside the flask, accompanied by a brief flash (duration ≈0.4 s; see Fig. 1b and Supplementary Movies 1, 2). The flask was then removed from the heating mantle and allowed to cool. The popping reaction can be also initiated by microwave radiation at 1000 W, which is detailed in Supplementary Note 1 and Supplementary Movie 3. PANI precursor with high HClO4 content can also undergo popping reaction through mechanical stimulation, which is detailed in Supplementary Note 1.

Structural analysis

SEM images were captured using Philips XL30 FEG microscope operating at 30 kV. TEM images were taken using JEM-2100F (JEOL) microscope operating at 200 kV. The high resolution EELS and TEM tests were conducted on FEI Titan Themis3 300. 300 (200 kV). Raman spectroscopy measurements were conducted on Horiba XploRA Plus Raman microscope spectrometer with 532 nm excitation (laser power: 0.5 mW). X-ray diffraction (XRD) measurements were carried out on an X-ray powder diffractometer (BRUKER, D8 ADVANCE) using Cu Kα1 radiation and Ge [111] monochromator. X-ray photoelectron spectroscopy (XPS) measurements were carried out on a Nexsa XPS System (Thermo Fisher) using monochromatic Al Kα radiation, and the calibration of binding energy (BE) of the spectra was referenced to the C 1 s electron bonding energy at 284.6 eV. Ar adsorption/desorption isotherms were obtained at 77 K using a Quantachrome NOVA gas sorptionanalyzer. BET analysis was used to determine the surface area. Nonlocal density functional theory (NLDFT) was used to analyze the pore size distribution. C, N and O K-edge XANES spectra were collected at the beamline BL08U1 of Shanghai Synchrotron Radiation Facility (SSRF), and the TEY yield was recorded using a microchannel plate detector. Fe, Co, Ni and Cu K-edge XAFS measurements were performed at BL14W1 and BL17B station in SSRF, using a double-crystal Si (111) monochromator. The storage ring of SSRF was operated at 3.5 GeV with the current of 300 mA. Data were acquired in fluorescence excitation mode using a Lytle detector at room temperature. The XAFS data were background subtracted, normalized, and Fourier transformed by standard procedures within the ATHENA program47. Morlet wavelet transform of EXAFS data was processed by hamaFortran program48. Neutron total scattering measurements were performed on a total scattering neutron time-of-flight diffractometer (Multiple Physics Instrument, MPI) at the China Spallation Neutron Source (CSNS)49. The scattering data were collected for 1 h at ambient conditions. The data were processed using the Mantid software and Fourier transformation to obtain the PDF with a maximum momentum transfer Qmax = 30 Å−1.

ORR and H2O2 production measurement

The electrochemical measurements in this study were acquired using a PMC-1000 multichannel potentiostat (AMETEK) operated via Versastudio software, with all potentials referenced to the reversible hydrogen electrode (RHE).

A standard three-electrode configuration was utilized at ambient temperature, featuring a graphite rod as the counter electrode and a rotating ring-disk electrode (RRDE, Pine Research Instrument, glassy carbon disk area: 0.237 cm2) as the working electrode. Reference electrodes consisted of Ag/AgCl for acidic media (0.1 M HClO4) and Hg/HgO for alkaline media (0.1 M KOH). To prepare the working electrode, a 5 mg mL−1 catalyst ink was formulated by dispersing the solid material in a mixture of 95 vol.% water and 5 vol.% Nafion (5 wt.%). After depositing a calibrated volume of ink onto the RRDE, the electrode was air-dried. Surface stabilization was achieved through cyclic voltammetry (CV) in O2 saturated electrolyte (0.05-1.2 V) at 100 mV s−1 for at least of 20 cycles. To account for ohmic resistance, electrochemical impedance spectroscopy (EIS) was performed at 0.78 V across a frequency range of 105 Hz to 0.02 Hz (10 mV amplitude), with the resulting data used for iR-compensation. Baseline correction was performed by recording linear scan voltammetry (LSV) in N2 saturated electrolyte at 10 mV s−1. The Pt ring was activated via 20 CV cycles (0.05-1.2 V) at 500 mV s−1, following the protocol established by Sa et al.50. ORR activity and H2O2 generation were quantified using LSV from 1.1 V to 0.05 V (10 mV s−1 scan rate) at a rotation speed of 1600 rpm. During these measurements, the Pt ring was maintained at a constant potential of 1.3 V. Following background subtraction, both disk and ring currents were normalized to the geometric surface area. The H2O2 selectivity was determined using the following expression:

$${{{{\rm{H}}}}}_{2}{{{{\rm{O}}}}}_{2}\,{{{\rm{selectivity}}}}=2\frac{{I}_{{{{\rm{R}}}}}/\eta }{({I}_{{{{\rm{D}}}}}+{I}_{{{{\rm{R}}}}}/\eta )}$$
(1)

where \({I}_{{{{\rm{R}}}}}\) and \({I}_{{{{\rm{D}}}}\,}\) is the ring current density (mA cm−2) and disk current density (mA cm−2), respectively. The ring current collect efficiency \(\eta\) is 38.3% according to the specification of the supplier.

Long-term stability and the H2O2 production capability toward real application were carried in an H-cell. Detailed information about the measurement and the quantitation of FE and the H2O2 production rate can be found in the Supplementary Note 6.

CO2RR measurement

The electrolysis cell was made of polyether ether ketone (PEEK) based on the design proposed by Kuhl et al.51 as shown in Supplementary Fig. 58. An anion exchange membrane was used to separate the anode and cathode compartments. RuO2/IrO2 coated titanium plate was employed as counter electrode for oxygen evolution reaction. Ag/AgCl (3 M NaCl) electrode was used as reference electrode. Catalyst ink (5 mg mL−1) was firstly prepared by dispersing the solid catalyst into solution containing 80 vol.% water, 17 vol.% isopropanol and 3 vol.% Nafion solution (5 wt.%). Working electrode was prepared by dropping catalyst ink (80 µL) onto carbon paper (1×1 cm2) and then air-drying at 60 °C for 15 min. After cell assembling, CO2 flowed into both anode and cathode compartment with flow rate of 7 and 30 cm3 min−1, respectively. Then 8 mL and 5 mL 0.25 M K2CO3 solution was injected into anode and cathode compartment, respectively. The electrolyte during electrolysis was CO2 saturated 0.5 M KHCO3 solution.

CV was carried within the potential range from 0.1 to −1.2 V at a scan rate of 200 mV s−1 for 10 cycles until the electrode surface become electrochemically stable. EIS test was carried at a potential of −0.3 V to measure the Ohmic drop, which is used for iR-compensation. The EIS spectra were collected in a frequency range of 0.01 to 100 Hz with an amplitude of 10 mV. Electrolysis was then carried by chronoamperometry for 1 h. The quantitation of the CO2RR products can be found in the Supplementary Note 7.

Atomistic simulation

The C/H/O/N ReaxFF force filed52, developed to simulate the carbon fiber formation from polymers, was used in all MD simulations53,54. The initial amorphous configurations were constructed by placing 80 PANI octamers randomly in a big periodic simulation box at a lower density (i.e., 0.1 g cm−3). Frist, the systems were minimized and compressed to the target densities of 1.5 g cm−3 within 125 ps. The systems were then equilibrated at 300 K for 100 ps. Afterward, they were heated to the target annealing temperatures of 3500 K with a heating rate of 10 K ps−1 and held for a total duration of 2.0 ns. All the MD simulations were conducted using an NVT ensemble with a time step of 0.25 fs, and the temperature was controlled by the Berendsen thermostat with a damping constant of 100 fs. To accelerate the graphitization, the released H2 molecules were removed during a course of simulations.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.