Carbonyl-staggered covalent organic frameworks for high-rate and stable rechargeable calcium-organic batteries

Abstract

Rechargeable calcium batteries are promising for next-generation energy storage devices due to the high crustal abundance of calcium (Ca) and the high-capacity and dendrite-free nature of Ca metal. However, their implementation is hindered by the lack of high-power and high-capacity positive electrodes, largely due to sluggish Ca²⁺ diffusion in inorganic hosts and limited active-site utilization in organic electrodes. Here we design a series of covalent organic frameworks incorporating different quinone units to construct carbonyl-staggered channels that accelerate Ca²⁺ hopping, enhance accessibility of deep-layer active sites even at high rates, and facilitate electronic charge transport through in situ conjugation reconstruction at ortho-carbonyl sites during repeated Ca²⁺ insertion. These combined effects contribute to fast charge-transfer kinetics, endowing the optimized heterogeneous framework with high-rate capability (56.4% capacity retention from 0.05 A g⁻¹ to 2 A g⁻¹) and long cyclability (71.7% capacity retention after 5,000 cycles at 2 A g⁻¹), with further validation in Ca-metal pouch cells. This work highlights carbonyl staggering as a molecular design strategy for regulating Ca-ion transport and improving active-site utilization in calcium-organic batteries.

Calcium batteries require high capacity cathodes with fast ion diffusion kinetics. Here, authors reported a carbonyl staggered covalent organic framework that yields alternating configuration of active sites in flexible skeleton for rapid Ca²⁺ hopping and high capacity utilization.

Introduction

While lithium-ion batteries (LIBs) have dominated the electrochemical energy storage market for decades, the intrinsic lithium scarcity (~20 ppm crustal abundance) and thermal runaway risks arising from lithium dendrite formation in high-energy lithium metal batteries have driven the exploration of alternatives with better affordability and safety^1–3. Calcium metal batteries (CMBs) have emerged as a promising candidate because of Ca’s natural abundance in the Earth’s crust (41,500 ppm, over 2000 times more plentiful than Li), high theoretical capacities (2,073 mAh cm⁻³), low redox potential (−2.87 V vs. standard hydrogen electrode), and lower tendency for dendritic metal deposition in organic electrolytes^4–6. Nevertheless, the strong Coulombic interactions between inorganic host lattices and divalent Ca ions pose substantial challenges to developing suitable positive electrodes with fast ion diffusion kinetics, high capacities, and long cycle life^7–9.

Unlike the smooth diffusion of monovalent Li⁺ ions in intercalation-type positive electrodes, such as Li_xCoO₂, Li_xFePO₄ and Li_xNi_yCo_zMn_wO₂ (0 <x < 1; y + z + w = 1), Ca²⁺ with a larger ionic radius (1 Å for Ca²⁺ vs. 0.76 Å for Li⁺) and stronger electronic polarization encounter high diffusion energy barriers in inorganic lattices (i.e., >1 eV for MnO₂¹⁰, V₂O₅¹¹). Although several polyanionic compounds with three-dimensional frameworks (i.e., Na_0.5VPO_4.8F_0.7¹², Ca_xNaV_1.5Cr_0.5(PO₄)₃¹³) can support reversible Ca²⁺ intercalation with appealing cyclability, their achievable capacities and rate performance remain limited (<100 mAh g⁻¹)¹². Organic redox-active compounds^14,15, composed of lightweight, abundant elements and flexible skeletons, can in principle offer higher capacities and better rate capability than inorganic hosts, as evidenced by the high theoretical capacities (>300 mAh g⁻¹) for small-molecule quinone crystals. However, their implementation has been plagued by low cycling capacities, poor rate capability, and unsatisfactory cycling stability arising from the inevitable dissolution of small organic molecules in nonaqueous polar aprotic electrolyte solutions compatible with metal negative electrodes. The intrinsically poor electronic conductivity and tortuous ion-transport pathways further lead to low redox-site utilization and limited rate capabilities¹⁶. The use of polymeric frameworks has been widely adopted to suppress the dissolution of organic active materials by tethering redox-active motifs along polymer chains^17,18. Yet the presence of abundant inactive skeletons and uncontrollable long-chain configurations compromises the specific capacities and obscures mechanistic understanding.

Redox-active covalent organic frameworks (COFs) have recently emerged as a versatile platform to overcome these limitations. Constructed by linking functional motifs through strong covalent bonds into crystalline porous architectures, COFs feature intrinsic insolubility and structural order suitable for mechanistic investigations^19–21. The well-defined two-dimensional (2D) COFs make them promising candidates for accommodating large cations, such as Ca²⁺. Recent studies have focused on incorporating more functional groups^22,23 or extending the π-conjugation^24,25 to improve the theoretical capacities and working voltages of positive electrode materials based on these moieties. Nevertheless, most reported COFs suffer from low redox utilization, with practical capacities substantially below the theoretical values especially those intended to be used in multivalent metal-ion battery systems, for example, 253 mAh g⁻¹_practical vs. 773 mAh g⁻¹_theoretical for TB-COF in aqueous Ca-ion battery²⁶, 161 mAh g⁻¹_practical vs. 271 mAh g⁻¹_theoretical for PTO-COF in Zn-ion battery²⁷, 180 mAh g⁻¹_practical vs. 296 mAh g⁻¹_theoretical for DATP-CNT in Al-ion battery²⁵. Such discrepancies arise from two main factors: the high resistance to multivalent metal-ion diffusion through the one-dimensional channels of these organic frameworks^28,29 and the dense π–π stacking between 2D polygonal COF layers, which bury active sites deep within the framework, resulting in extended ion-transport paths and poor active-site accessibility. Exfoliated COF nanosheets^30,31 can partially alleviate these limitations by exposing active sites on the 2D layers. Yet such a “top-down” approach induces severe restacking issues and structural instability during cycling. Therefore, developing a molecular-level design strategy to enhance ion accessibility while preserving framework integrity upon repeated Ca²⁺ insertion/extraction remains a critical challenge. Recent computational studies have recommended the importance of theoretical approaches for elucidating ion-storage mechanisms in COF electrodes^32–35, but a fundamental understanding of how compositional and motif arrangement regulate Ca-ion transport and storage in multicomponent COFs has not been systematically investigated.

In this work, we introduce a carbonyl-staggered COF structure incorporating dual ortho-quinone moieties to enhance the accessibility of redox-active sites for high-capacity and long-life Ca-ion batteries. Unlike conventional mono-quinone COFs with perfectly eclipsed carbonyl groups along the b-axis, our carbonyl-staggered COF adopts a bilayer-periodicity AA-stacked configuration. This specific alternating configuration naturally yields a carbonyl-staggered environment that mitigates adjacent Coulombic interactions and enables a ‘monkey-bar’-like Ca²⁺ hopping mechanism, thereby facilitating rapid ion diffusion to active sites deep within the COFs (Fig. 1i). During Ca²⁺ insertion, the ortho-carbonyl groups undergo conjugation-assisted reconfiguration to form C = C bonds (Fig. 1ii), thereby reinforcing the electronic conduction along the COF backbones. Combined experimental and theoretical investigations elucidated a ‘two-step insertion and one-step extraction’ pathway and a radical-assisted multielectron redox mechanism for Ca²⁺ storage in the carbonyl-staggered COFs. In comparison to the small molecule crystals (i.e., 9,10-phenanthrenequinone PQ, 4,5,9,10-pyrenetetraone PT), the homogeneous COFs (PTCOF, PQCOF), and the conventional eclipsed AA-stacked COFs (PQ_0.33PT_0.67COF, PQ_0.67PT_0.33COF), the optimized carbonyl-staggered PQ_0.5PT_0.5COF delivers a high specific capacity of 225 mAh g⁻¹ (204 mAh g⁻¹ after practical capacity correction, corresponding to 94.9% theoretical utilization), long-term cycling stability (over 150 mAh g⁻¹ after 800 cycles at 0.2 A g⁻¹, 71.7% capacity retention at 2 A g⁻¹ after 5000 cycles), and a high specific energy of 499 Wh kg⁻¹_COF in Ca metal pouch cells. Our findings highlight the role of stacking-mode engineering in regulating ion transport and redox-site accessibility in COFs, providing insights into the design of high-performance organic electrodes for Ca and potentially other multivalent metal batteries.

Fig. 1. Dynamic π-conjugated covalent organic frameworks (COFs) design with staggered carbonyl channels for fast Ca²⁺ diffusion.

i Stacking-dependent Ca²⁺ diffusion: eclipsed carbonyl alignment restricts Ca²⁺ transport (green spheres), whereas staggered carbonyls facilitate Ca²⁺ hopping (yellow spheres). ii Dynamic π-conjugated design: Ca²⁺ insertion induces enolization and C = C bond formation in 9,10-phenanthrenequinone (PQ), and 4,5,9,10-pyrenetetraone (PT), the extended π-conjugation in the COFs increases electronic conductivity. iii The optimized carbonyl-staggered PQ_0.5PT_0.5COF model with alternating PQ/PT layers and staggered carbonyl channels at every edge for efficient Ca²⁺ hopping and deeper Ca²⁺ diffusion.

Results and Discussion

Rational design and synthesis of carbonyl-staggered COFs

To simultaneously address two critical challenges in COF-based organic positive electrodes, namely, the sluggish Ca²⁺ diffusion kinetics and the limited accessibility of redox-active sites throughout the active mass, we selected the high-capacity ortho-quinone monomers PQ and PT to construct carbonyl-staggered COFs (Fig. 1iii). The PQ and PT precursors were first functionalized with 2,7-diamino groups to yield diaminoquinone derivatives (DAPQ and DAPT), which were subsequently linked with 1,3,5-triformylphloroglucinol (TFG) via a Schiff-base condensation reaction to form crystalline COFs with periodically arranged quinone ligands (Supplementary Fig. 1). The successful synthesis of phase-pure DAPQ and DAPT monomers was confirmed by ¹H nuclear magnetic resonance (NMR) spectra (Supplementary Figs. 2, 3). A series of carbonyl-based COFs, including PQCOF, PTCOF, and PQPTCOF with different PQ:PT stoichiometric ratios, were synthesized to systematically investigate how carbonyl stacking configurations influence the Ca-ion storage behavior. Notably, the selection of PQ and PT from a range of quinone candidates was guided by their relatively small highest occupied molecular orbital (HOMO)-lowest unoccupied molecular orbital (LUMO) energy gaps (3.56 eV for PQ and 3.50 eV for PT), which are favorable for electron transfer and redox activity. Electrostatic potential (ESP) mapping revealed pronounced negative potential at the carbonyl oxygen atoms of PQ and PT (ESP_min: PQ = −49.31 kcal mol⁻¹, PT = −40.27 kcal mol⁻¹), indicating an enhanced tendency for Ca²⁺ coordination. Additionally, the adjacent carbonyl groups in the ortho-quinone configuration enable bidentate coordination with Ca²⁺ (Supplementary Fig. 4, Supplementary Data 1). The formation of PQPTCOF was confirmed by Fourier-transform infrared (FTIR) spectroscopy (Supplementary Fig. 5). Distinct vibrational bands corresponding to aromatic C = C, carbonyl (C = O), C–N, and N–H groups are clearly identified in the precursors. The disappearance of N–H stretching bands from primary amine groups in the DAPQ and DAPT monomers (3200–3500 cm⁻¹) and the O–H stretching in TFG (3200–3350 cm⁻¹) confirmed the consumption of reacting groups during polycondensation, thereby demonstrating the successful formation of PQPTCOF.

The synthesized PQ_xPT_1−xCOFs were denoted as PQ_0.33PT_0.67COF, PQ_0.5PT_0.5COF, and PQ_0.67PT_0.33COF according to the monomer ratios. Comparative analysis of the Rietveld-refined powder X-ray diffraction (PXRD) patterns (Fig. 2a–c and Supplementary Data 2) elucidates the correlation among PQ/PT ratio, crystal structure, and stacking configuration. Both PQ_0.33PT_0.67COF and PQ_0.67PT_0.33COF adopt triclinic symmetry with a conventional AA stacking mode with an interlayer spacing of 3.53 Å. In contrast, PQ_0.5PT_0.5COF adopts a bilayer-periodicity AA stacking mode with an interlayer distance of 3.43 Å. All three PQ_xPT_1−xCOFs exhibit a discernible stacking-related (001) reflection, indicating interlayer ordering. The position of this characteristic reflection is directly correlated with the interlayer spacing of the layered COF framework, which can be quantitatively calculated via Bragg’s law. Among them, PQ_0.5PT_0.5COF shows the highest diffraction intensity at low-angle (100) reflection, suggesting a higher degree of in-plane structural ordering than PQ_0.33PT_0.67COF and PQ_0.67PT_0.33COF. The ordered stacking structure is further corroborated by high-resolution transmission electron microscopy (HRTEM) observations with distinct lattice fringes (Supplementary Fig. 6).

Fig. 2. Chemical structure and electrochemical performance of PQ_xPT_1−xCOF positive electrodes.

a-c Schematic structures and Powder X-ray diffraction (PXRD) analyses of PQ_xPT_1−xCOF with varying PQ/PT ratios, with PQ and PT shown in pink and blue, respectively. Left: simulated structural models, including top-views and side-view insets (for the orange block part) that illustrate the staggered and eclipsed oxygen (red atoms) stacking and interlayer distances; Atomic coordinates of the optimized structural models are provided in Supplementary Data 2. Right: the corresponding PXRD patterns. d Pore size distribution of PQ_xPT_1−xCOFs. e Discharge/charge potential profiles of carbonyl-eclipsed COFs (homogeneous PTCOF and heterogeneous PQ_0.67PT_0.33COF) and carbonyl-staggered PQ_0.5PT_0.5COF positive electrodes at 0.05 and 2 A g⁻¹. f Rate performance of PTCOF, PQCOF, PQCOF/PTCOF, PQ_xPT_1−xCOF positive electrodes at specific currents from 0.2 to 2 A g⁻¹. All electrochemical measurements in this figure are performed using activated carbon (AC) counter electrodes and 1 M Ca(TFSI)₂ in TEGDME/DOL electrolyte. All electrochemical and characterization tests were conducted at 25 ± 1 °C.

The homogeneous ligand COFs (PQCOF and PTCOF) exhibit well-defined PXRD patterns and type-II N₂ adsorption isotherms with uniform pore size distributions (Supplementary Fig. 7), collectively supporting their ordered porous structures. These distinct stacking modes give rise to different carbonyl group alignments: the monolayer-periodicity AA stacking in PQ_0.33PT_0.67COF and PQ_0.67PT_0.33COF leads to vertical superimposition of carbonyl groups from eclipsed layers, whereas the bilayer-periodicity AA stacking in PQ_0.5PT_0.5COF indicates spatially staggered carbonyl arrays across interlayers. Detailed lattice parameters are listed in Supplementary Table 1. Pore size distributions further support the bilayer-periodicity AA stacking of PQ_0.5PT_0.5COF with a smaller dominant pore aperture of 1.04 nm, in contrast to the larger pores of 2.0 and 2.25 nm observed for PQ_0.33PT_0.67COF and PQ_0.67PT_0.33COF, which can be attributed to the interlayer staggering that creates pore constrictions (Fig. 2d, Supplementary Fig. 8). Density functional theory (DFT) calculations further support this interpretation, indicating that when PQ and PT motifs are integrated within the same framework, the bilayer-periodicity carbonyl-staggered stacking model of PQ_0.5PT_0.5COF is energetically more favorable than the corresponding carbonyl-eclipsed PQCOF and PTCOF reference models (Supplementary Fig. 9 and Supplementary Data 3). FTIR, Raman and scanning electron microscopy (SEM) characterizations (Supplementary Figs. 10, 11) further corroborate the structural integrity and microparticle morphologies of the PQ_xPT_1−xCOF series, indicating that despite variations in ligand compositions and stacking modes, the COFs retain similar chemical environments and framework structures. In addition, a 1:1 mechanical mixture of PQCOF and PTCOF (denoted as PQCOF/PTCOF) was prepared as a control. Its PXRD patterns and Brunauer–Emmett–Teller (BET) characteristics are different from those of PQ_0.5PT_0.5COF, supporting its distinction from a simple physical mixing of PQCOF and PTCOF (Supplementary Fig. 12).

Uptake of Ca²⁺ in COFs with staggered versus eclipsed carbonyls

To evaluate the Ca²⁺ storage behavior in COFs with distinct ligand ratios and stacking modes, we assembled homogeneous PQCOF and PTCOF, together with heterogeneous PQPTCOF electrodes in Ca-ion cells using 1 M Ca(TFSI)₂ tetraethylene glycol dimethyl ether (TEGDME)/1,3-dioxolane (DOL) (1:1, v/v) as the electrolyte and activated carbon (AC) as the counter electrodes. The AC electrodes were also employed as the reference electrodes to avoid complications arising from the irreversibility of Ca metal negative electrodes in most organic electrolytes. AC electrodes have been shown to be electrochemically active in nonaqueous electrolytes via ion adsorption/desorption interactions³⁶, enabling their use as both counter and reference electrodes. The potential of the AC electrodes was calibrated to be around 3.22 V vs. Ca²⁺/Ca in our Ca-ion electrochemical system (Supplementary Fig. 13). Cyclic voltammetry (CV) scanning at 0.5 mV s⁻¹ (Supplementary Fig. 14) revealed two distinct reduction peaks (approximately −0.35 V and −1.05 V) and a single oxidation peak (−0.2 V) for all these COF electrodes, consistent with Ca²⁺ insertion/extraction via reversible enolization of quinone carbonyls. Galvanostatic discharge/charge potential profiles (Fig. 2e and Supplementary Fig. 15) obtained at 0.05 and 2 A g⁻¹ between −1.5 V and 0.5 V vs. AC delineate their Ca ion storage capacities. At a low rate of 0.05 A g⁻¹, PTCOF with the highest density of carbonyl groups indicated the highest specific capacity of 234 mAh g⁻¹ followed by the 225 mAh g⁻¹ for PQ_0.5PT_0.5COF, 220 mAh g⁻¹ for PQ_0.33PT_0.67COF, 216 mAh g⁻¹ for PQ_0.67PT_0.33COF, and 215 mAh g⁻¹ for PQCOF, all approaching their theoretical capacities. However, PQ_0.5PT_0.5COF maintained the highest capacity of 127 mAh g⁻¹ at 2 A g⁻¹, in contrast to the less than 80 mAh g⁻¹ capacity retention in other COFs, thus indicating improved Ca²⁺ transport kinetics and redox reversibility for PQ_0.5PT_0.5COF. The narrowest discharging/charging potential gap in the dQ/dV vs. AC curves (Supplementary Fig. 16) further suggests the lowest polarization and improved Ca-ion uptake of PQ_0.5PT_0.5COF positive electrodes.

When cycling from 0.2 A g⁻¹ to 2 A g⁻¹, all COF positive electrodes demonstrated better cyclability than those comprising PQ and PT small-molecule crystals (Fig. 2f, Supplementary Fig. 17), which is attributed to the structural robustness and insolubility of COFs in the electrolyte solutions during cycling (Supplementary Fig. 18). PQ_0.5PT_0.5COF delivered a specific capacity of 175 mAh g⁻¹ at 0.2 A g⁻¹ and retained 127 mAh g⁻¹ at 2 A g⁻¹. When the current was returned to 0.2 A g⁻¹, the capacity was recovered to 175 mAh g⁻¹, indicating good reversibility. In contrast, homogeneous COFs exhibited inferior rate performance; for instance, PQCOF retained only 75 mAh g⁻¹ at 2 A g⁻¹ despite an initial capacity of 148 mAh g⁻¹ at 0.2 A g⁻¹. The carbonyl-eclipsed PQ_0.33PT_0.67COF and PQ_0.67PT_0.33COF exhibited moderate capacities of 101 and 115 mA h g⁻¹ at 0.2 A g⁻¹, far below their achievable capacities of ~220 mAh g⁻¹ at 0.05 A g⁻¹. Notably, although the high carbonyl-group density of PTCOF provides a higher theoretical capacity, PT-rich COFs did not show a proportional improvement in capacity at high rates. This discrepancy highlights that carbonyl density alone cannot fully determine the electrochemical activity. It is hypothesized that excessive PT incorporation in aligned AA-stacked COFs may intensify interlayer electrostatic interactions, thereby hindering Ca²⁺ diffusion. Although the higher crystallinity of PQ_0.5PT_0.5COF supports the presence of ordered ion-transport channels, its high-rate capability reflects the combined effects of structural regularity, carbonyl arrangement, and interlayer transport accessibility. Conversely, the moderate carbonyl density and carbonyl-staggered stacking configuration of PQ_0.5PT_0.5COF alleviate electrostatic hindrance and enable faster Ca²⁺ transport, contributing to its high-rate capability compared with previously reported COF positive electrodes in various rechargeable battery systems (Supplementary Table 2). The inferior kinetics of the PQCOF/PTCOF mixture support the view that the enhanced Ca²⁺ transport in PQ_0.5PT_0.5COF primarily arises from the specific carbonyl-staggered stacking configuration rather than chemical composition alone. Therefore, the stacking configuration and local coordination environment play important roles in regulating Ca ion-transport kinetics and active-site accessibility.

The accessibility of active sites in COFs is closely related to the diffusion kinetics of Ca²⁺. To elucidate the Ca²⁺ transport behavior, electrochemical measurements and DFT calculations were conducted. Diffusion coefficients (D) derived from the CV curves (Fig. 3a and Supplementary Fig. 19) revealed faster ion transport during oxidation than upon reduction across all COF samples, consistent with the faster decalciation kinetics. PQ_0.5PT_0.5COF positive electrodes showed the most favorable ion transport kinetics with a diffusion coefficient D_Red ≈ 1.2 × 10⁻⁶ cm² s⁻¹, which substantially exceeded those of carbonyl-eclipsed heterogeneous PQ_0.33PT_0.67COF and the homogeneous PTCOF electrodes (D_Red = 8.1 × 10⁻⁷ cm² s⁻¹ and D_Red1 = 4.14 × 10⁻⁷ cm² s⁻¹, respectively). The enhanced diffusion behavior of PQ_0.5PT_0.5COF can be attributed to its staggered carbonyl framework, which effectively reduces dense carbonyl stacking and facilitates rapid Ca²⁺ hopping pathways. Consequently, PQ_0.5PT_0.5COF achieves an oxidation diffusion coefficient of 1.64 × 10⁻⁶ cm² s⁻¹, comparable to that of PQCOF but exceeding those of other PQPTCOFs. Crucially, the diffusion coefficient determined under demanding high-scan-rate conditions (1–32 mV s⁻¹) further supports the favorable Ca-ion transport kinetics. This enhanced diffusion capability facilitates access to redox sites in deeper layers and helps maintain high capacity at high specific currents. The eclipsed stacking of densely packed active sites in PTCOF restricts ion migration despite the abundance of carbonyl groups, while PQCOF enables rapid diffusion but lacks sufficient redox-active sites. The staggered architecture in PQ_0.5PT_0.5COF thus reconciles these trade-offs by improving ion transport pathways and the spatial accessibility of carbonyl groups, yielding a favorable balance between ion transport and redox density. Galvanostatic intermittent titration technique (GITT) analysis further supports the favorable ion-transport kinetics of the carbonyl-staggered PQ_0.5PT_0.5COF positive electrode, with Ca²⁺ diffusion coefficients of 10⁻⁸–10⁻⁵ cm² s⁻¹ during calciation/decalciation (Supplementary Fig. 20). The faster decalciation kinetics may be associated with the rational ortho-carbonyl design, in which enolization-driven conjugation reconstruction during calcination facilitates charge transport.

Fig. 3. Ca²⁺ ions diffusion kinetics in PQ_xPT_1−xCOF structures.

a Quantitative comparison of the apparent diffusion coefficients at four characteristic redox peaks (Red₁, Red₂, Red₃, and O_x) for PQCOF, PTCOF, and PQ_xPT_1−xCOF positive electrodes at 25 ± 1 °C. b Schematic illustration of four types of Ca²⁺ diffusion pathways in PQ_xPT_1−xCOF: (i) intra-layer path 1, (ii) intra-layer path 2, (iii) eclipsed inter-layer path, and (iv) staggered inter-layer path. (PQ and PT units are colored pink and blue, respectively.) Atomic coordinates are provided in Supplementary Data 4. c Density functional theory (DFT)-calculated energy barriers for Ca²⁺ diffusion along the routes in (b).

DFT calculations further clarified the effect of stacking configuration on Ca²⁺ migration (Fig. 3b and Supplementary Data 4). The calculated energy barriers for in-plane migration routes, either between PT carbonyls (Path 1) or between PT and TFG carbonyls (Path 2), are high at ~4 eV (Fig. 3c), indicating in-plane diffusion is kinetically unfavorable. Interlayer migration, in contrast, is highly stacking-dependent. In carbonyl-eclipsed COFs, the direct atop orientation of the carbonyl motifs affords a moderate barrier of 2 eV. The carbonyl-staggered PQ_0.5PT_0.5COF exhibits a significantly lower Ca²⁺ migration barrier of ~0.5 eV by introducing a staggered interlayer path in which Ca²⁺ hops vertically across misaligned carbonyl arrays of alternating PQ and PT layers. This low-energy route can be attributed to both the spatial offset of carbonyl coordination sites and the local electrostatic asymmetry imposed by the PT/PQ heterostructure. Overall, these results suggest that active-site staggering may facilitate fast interlayer Ca²⁺ hopping, promoting active-site utilization throughout the deeper COF frameworks and thereby providing a strategy to balance redox-site density and ion mobility in 2D COF positive electrodes.

Ca²⁺ storage mechanisms

To elucidate the Ca-ion storage mechanisms in the optimal PQ_0.5PT_0.5COF, we combined spectroscopic analyses with DFT calculations. To monitor the structural evolution, ex situ FTIR spectra were collected during the first two cycles at 0.05 A g⁻¹ between −1.5 and 0.5 V vs. AC (Fig. 4a). The pristine PQ_0.5PT_0.5COF displays characteristic C = O (1678 cm⁻¹) and C–N (1614 cm⁻¹) bands. Upon discharge, the C = O peak red-shifts along with the appearance of a new peak at 1475 cm⁻¹, indicating Ca²⁺ accommodation and enolization of carbonyl groups (C = O → C–O⁻). During charging, the C = O peak recovers to 1675 cm⁻¹, signifying reversible Ca²⁺ extraction and restoration of the carbonyl groups. The carbonyl shift stabilizes after the first cycle, while the C–N vibration remains unchanged throughout the cycles, indicating that the framework backbone is largely preserved. High-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) elemental mapping shows a uniform Ca distribution in the calciated PQ_0.5PT_0.5COF (Supplementary Fig. 21), and ex situ X-ray photoelectron spectroscopy (XPS) (Supplementary Fig. 22) further supports the reversible C = O/C–O⁻ conversion, which is consistent with the reversible electrochemical behavior of PQ_0.5PT_0.5COF.

Fig. 4. Ca²⁺ storage mechanisms in the PQ_0.5PT_0.5COF structure.

a Discharge/charge potential profiles of PQ_0.5PT_0.5COF at 0.05 A g⁻¹ (left), and the corresponding ex situ Fourier-transform infrared spectroscopy (FTIR) spectra collected at the marked states during the first and second cycles (right). b Discharge/charge potential profiles for the first cycle at 0.2 A g⁻¹(left) and the corresponding in situ Raman spectra with peak deconvolution at different electrochemical states: (i) pristine, (ii) D_{−0.6 V}, (iii) D_{−0.9 V}, (iv) D_{−1.5 V}, (v) C_{−0.2 V}, and (vi) C_{0.5 V}. c Evolution of R_ct for PQ_0.5PT_0.5COF at pristine, calciated, and decalciated states during the first cycle at 0.05 A g⁻¹. Data are presented as mean ± standard deviation from electrode samples collected from three independently assembled and disassembled cells at each state (n = 3; technical replicates). d Highest occupied molecular orbital (HOMO)-lowest unoccupied molecular orbital (LUMO) energy gaps of PQ/PT motifs before and after Ca²⁺ insertion. Atomic coordinates are provided in Supplementary Data 5. e Electron paramagnetic resonance (EPR) spectra collected at the specific discharge and charge states during the first cycle at 0.05 A g⁻¹. f Simulated Ca²⁺ storage processes in 2PQ_0.5PT_0.5COF. Ca₁ (green) represents Ca²⁺ coordinated with the eclipsed ortho-carbonyls, Ca_2-hopping (orange) denotes Ca²⁺ hopping through the staggered ortho-carbonyls, and Ca₃ (purple) indicates Ca²⁺ coordinated at isolated carbonyl sites on gray 1,3,5-triformylphloroglucinol (TFG) knots shown in gray. Atomic coordinates are provided in Supplementary Data 6. All electrochemical and characterization tests were conducted at 25 ± 1 °C.

In situ Raman spectroscopy provides real-time insights into the redox coordination process (Fig. 4b). Raman spectra collected at 0.1 V intervals under 0.2 A g⁻¹ reveal potential-dependent evolution of C = O and aromatic C = C vibrations (1506 − 1762 cm⁻¹). During Stage I (0 to −0.6 V), the C = O bond (1702 cm⁻¹) gradually red-shifts as Ca²⁺ coordinates with high-energy ortho-carbonyl sites. Stage II ( − 0.6 to −0.9 V) is marked by the disappearance of the C = O bands (1656 − 1728 cm⁻¹) and the emergence of a C = C vibration at 1465 cm⁻¹, consistent with a two-electron reduction of ortho-carbonyls to C–O⁻. This conversion restores aromatic conjugation, as evidenced by the intensified delocalized C = C/C–C modes (1528–1692 cm⁻¹), which may facilitate electronic conductivity. Ex situ electrochemical impedance spectroscopy (EIS) measurements (Fig. 4c, Supplementary Fig. 23, and Supplementary Table 3) reveal a pronounced decrease in the charge-transfer resistance (R_ct) upon calcination, indicating enhanced charge transfer kinetics consistent with the formation of enolate-induced C = C conjugation. Consistently, DFT calculations corroborate these observations by displaying a narrowed bandgap and continuous π-orbital overlap after enolization (Fig. 4d, Supplementary Fig. 24, and Supplementary Data 5), which can facilitate electron delocalization and charge transport. Upon further discharge (Stage III, −0.9 to −1.5 V), Ca²⁺ ions continue to occupy available carbonyl sites with the maximum C–O⁻ signal intensity near 1650 cm⁻¹. During the subsequent charging process, the C–O–Ca bonds dissociate with the restoration of C = O bonds in Stage IV ( − 1.5 to −0.2 V) and the gradual disappearance of C = C bonds in Stage V ( − 0.2 to 0.5 V). The reversible sequence demonstrates that Ca-ion storage occurs via a dual mechanism involving carbonyl coordination and dynamic π-conjugation evolution, while other structural motifs remain electrochemically inert. Although carbonyl-to-enolate conversion is known in quinone-based organic electrodes^16,37, its framework-level conjugation reconstruction for enhancing electronic conductivity in COFs remains less explored.

Electron paramagnetic resonance (EPR) spectra provide evidence for radical-mediated electron transfer during Ca²⁺ storage (Fig. 4e). The pristine electrode exhibits a weak peak (g = 2.0044) attributed to the localized π-electron defects. Upon discharge to −0.6 V, a slight increase in intensity (g = 2.0043) is consistent with the formation of transient radicals (C–O•) via Ca²⁺ coordination at carbonyl sites. Upon further discharge to −0.9 V, the EPR signal (g = 2.0046) reaches its maximum intensity, suggesting the highest concentration of radical intermediates. The relatively weak signal at −0.6 V may be attributed to a rapid two-electron process at ortho-carbonyls, while the strong signal at −0.9 V may reflect stepwise one-electron reduction at isolated carbonyls on the TFG knot. The small g shift (Δg = +0.0003) implies a spin-orbit coupling effect associated with the modulation of the spin density distribution of the radical moieties under Ca²⁺ coordination. Upon deep discharge to −1.5 V, the EPR signal decreases markedly (g = 2.0047), consistent with radical annihilation and conjugation recovery. During the charging process, the EPR signals exhibit a reversible single-peak evolution and a transient radical regeneration. The EPR observations collectively support a ‘two-step insertion and one-step extraction’ mechanism with the radical-assisted multi-electron redox behavior.

DFT calculations further provide a possible interpretation of the stacking-dependent coordination chemistry. Molecular electrostatic potential (MESP) maps suggest that the ortho-carbonyl groups in PQ and PT units serve as more favorable Ca²⁺ binding sites than the isolated carbonyls on the TFG linkers (Supplementary Fig. 25 and Supplementary Data 5). Bilayer COF models were then constructed to capture interlayer effects. Carbonyl-staggered PQ_0.5PT_0.5COF provides additional Ca²⁺ accommodation sites arising from staggered carbonyl alignment, forming a stable five-membered chelation motif (Supplementary Fig. 27 and Supplementary Data 6) that is not observed in the carbonyl-eclipsed counterparts. The stepwise Ca²⁺ storage process is proposed as follows (Fig. 4f and Supplementary Data 6): (i) six Ca²⁺ ions are initially chelated by eclipsed ortho-carbonyl groups in a tetradentate configuration (denoted as Ca₁); (ii) three more Ca²⁺ ions occupy the staggered interlayer chelation sites unique to the bilayer-periodicity AA stacking (denoted as Ca_2-hopping); and (iii) twelve Ca²⁺ ions are finally accommodated by the interlayer-bridging TFG sites (denoted as Ca₃), forming a 2PQ_0.5PT_0.5COF-21Ca formula. Although the eclipsed carbonyl-stacked PQ_0.67PT_0.33COF shows comparable theoretical Ca²⁺ storage capacities (2PQ_0.67PT_0.33COF-20Ca, Supplementary Fig. 28 and Supplementary Data 6), the higher Ca²⁺ migration barriers may contribute to the limited active-site utilization at high rates. These calculations are based on idealized structural models and do not fully capture solvation or interfacial effects under operating conditions. Therefore, the proposed coordination configurations and migration pathways should be regarded as possible mechanistic interpretations of the real electrode system. The possible Ca²⁺ storage sites and the theoretical capacities for various COF species are summarized in Supplementary Table 4. Together, our findings demonstrate that carbonyl-staggered PQ_0.5PT_0.5COF can balance ion transport kinetics with high redox accessibility.

Application of COF-based positive electrodes for rechargeable Ca batteries

Building upon the Ca²⁺ storage behaviors in the carbonyl-staggered PQ_0.5PT_0.5COF-based positive electrodes, we systematically evaluated their electrochemical performance to assess practical feasibility. PQ_0.5PT_0.5COF positive electrodes delivered a high initial capacity of 225 mAh g⁻¹ at 0.05 A g⁻¹, higher than those of the monomeric precursors DAPQ (32 mAh g⁻¹) and DAPT (74 mAh g⁻¹) (Fig. 5a). This enhancement can be attributed to the combined effects of ortho-carbonyl coordination and π-conjugation reconstruction in the carbonyl-staggered framework, which facilitates Ca²⁺ uptake and reversible multi-electron redox activity. Consequently, PQ_0.5PT_0.5COF achieves 94.9% active-site utilization after accounting for additive contributions (Supplementary Fig. 29, Methods), higher than that of selected representative COF electrodes in various rechargeable batteries (Supplementary Table 5). The presence of well-defined plateaus in the discharge/charge potential profiles at specific currents from 0.05 to 2 A g⁻¹ (Supplementary Fig. 30) further indicates sustained redox activity at high specific currents. Compared with reported representative Ca ion battery (CIB) positive electrodes (i.e., 162 mAh g⁻¹ for Ca_xNaV_1.5Cr_0.5(PO₄)₃¹³, 192 mAh g⁻¹ for Ca_0.13MoO₃·(H₂O)_0.41³⁸, 159 mAh g⁻¹ for PTCDA³⁹ and 116 mAh g⁻¹ for PAQS@CNTs⁴⁰), PQ_0.5PT_0.5COF positive electrodes exhibit higher specific capacities and better rate capability (Fig. 5c and Supplementary Table 6).

Fig. 5. Electrochemical performance of PQ_0.5PT_0.5COF positive electrodes in half cells (vs. AC counter electrodes, 1 M Ca(TFSI)₂ TEGDME/DOL electrolyte) and full (pouch) cells (vs. Ca metal negative electrodes, 0.4 M Ca[B(hfip)₄]₂/DME electrolyte).

a Discharge/charge potential profiles of PQ_0.5PT_0.5COF, DAPT, and DAPQ electrodes at 0.05 A g⁻¹. b Cycling capacities of PQ_0.5PT_0.5COF positive electrodes at 0.2 A g⁻¹ over 800 cycles in a coin cell. c Comparison of rate capability of PQ_0.5PT_0.5COF positive electrodes with reported positive electrode materials, where each type of positive electrode is represented by a unique symbol as defined in the legend of e (see details in Supplementary Table 6). d Cycling capacities of PQ_0.5PT_0.5COF positive electrodes at 2 A g⁻¹ over 5000 cycles in coin cell. e Comparison of cycling capacity retentions of PQ_0.5PT_0.5COF and peer positive electrodes (see details in Supplementary Table 7). f Discharge/charge voltage profiles of Ca | |PQ_0.5PT_0.5COF pouch cells at 0.05, 0.1, 0.2 and 0.5 A g⁻¹. g Cycling capacities of Ca | |PQ_0.5PT_0.5COF pouch cells over 20 cycles. h Ragone plots for PQ_0.5PT_0.5COF positive electrodes in pouch cells and other Ca-ion positive electrodes measured in full cells, with specific energy and specific power calculated based on the mass of the positive-electrode active material. (See details in Supplementary Table 8). All electrochemical measurements were conducted at 25 °C ± 1 °C.

Long-term cycling performance was evaluated at moderate and high specific currents. At 0.2 A g⁻¹ (Fig. 5b), PQ_0.5PT_0.5COF positive electrodes (cycled in half-cells with AC counter electrodes and 1 M Ca(TFSI)₂ in TEGDME/DOL electrolyte) retained a specific capacity of ~140 mAh g⁻¹ after 800 cycles, corresponding to over 80% retention of their initial capacity. Even at a high rate of 2 A g⁻¹, the intrinsic features of rapid Ca²⁺ hopping and high active-sites accessibility of carbonyl-staggered COF contributed to sustaining~100 mAh g⁻¹ over 5000 cycles (Fig. 5d), showing a competitive long cycling lifespan compared with reported Ca-ion positive electrodes (e.g., 84.2% capacity retention after 1000 cycles for K_xVPO₄F⁴¹, 76% after 200 cycles for PQ¹⁶, 84% after 2500 cycles for Mg_0.25V₂O₅ ∙ H₂O⁴²). Notably, a long lifespan exceeding 10,000 cycles was demonstrated with 82 mAh g⁻¹ remaining, comparing favorably with previously reported Ca battery positive electrodes (Supplementary Fig. 31 and Supplementary Table 7), consistent with the favorable Ca²⁺ storage kinetics, reversibility, and structurally stability of carbonyl-staggered COF architecture (Supplementary Fig. 32), as further supported by the stable in situ EIS evolution during cycling (Supplementary Fig. 33). Durable operation was also observed under high mass-loading conditions (Supplementary Fig. 34). To assess the practical feasibility, Ca | |PQ_0.5PT_0.5COF pouch cells were assembled using 0.4 M Ca[B(hfip)₄]₂/1,2-dimethoxyethane (DME) electrolyte solution. This solution is effective since it enables reversible Ca metal deposition and stripping, and its wide electrochemical window of over 4 V also supports compatibility with positive electrodes. The full cells exhibited electrochemical characteristics similar to those of the half-cells, with a reversible specific capacity of 175 mAh g⁻¹ and a discharge voltage of 2.3 V (Supplementary Fig. 35). Pouch cells (3 cm × 3 cm) delivered an initial discharge capacity of 217 mAh g_COF⁻¹ at 50 mA g⁻¹, with a voltage plateau of 2.3 V and an average discharge voltage of 2.0 V (Fig. 5f), and maintained ~155 mAh g_COF⁻¹ after 20 cycles (Fig. 5g and Supplementary Fig. 36). Ragone analysis exhibited a gravimetric specific energy of 499 Wh kg⁻¹_COF at 115 W kg⁻¹, which is higher than those of previously reported Ca metal full-cell systems evaluated under comparable conditions (Fig. 5h and Supplementary Table 8).

In this work, we present a stacking-engineered design of carbonyl-staggered COF positive electrodes that facilitates rapid Ca²⁺ transport and enhances the accessibility of active sites in deeper layers at high rates. The optimized PQ_0.5PT_0.5COF positive electrodes deliver a high capacity of 225 mAh g⁻¹, good rate performance (56.4% retention from 0.05 to 2 A g⁻¹), and long-term stability over 5000 cycles (71.7% capacity retention). Comprehensive spectroscopic and theoretical investigations revealed that a bilayer-periodicity AA stacking mode between PQ and PT moieties in the COF (1:1 ratio) induces a staggered interlayer arrangement of carbonyl groups, which enables a low-barrier route for Ca²⁺ hopping compared to conventional AA stacking motifs. Meanwhile, the ortho-carbonyl groups undergo conjugation reconstruction during Ca²⁺ insertion, facilitating electronic charge transport. The findings in this work highlight stacking-mode modulation as a promising strategy for tuning bivalent-ion transport and active-site accessibility in an active mass, supporting the development of high-performance positive electrode materials for rechargeable Ca battery systems. Beyond Ca batteries, this work provides design insights for developing organic electrodes in other multivalent metal-ion battery systems, with potential relevance to sustainable energy storage.

Methods

Materials

All reagents and starting materials were used as received without further purification: 9,10-phenanthrenequinone (PQ, 98%, Energy), pyrene (98%, Energy), fuming nitric acid (HNO₃, 95 wt%, prepared in-house), nitric acid (HNO₃, 69%, VWR), sulfuric acid (H₂SO₄, 97% and 98%, VWR), sodium carbonate (Na₂CO₃, 99.8%, Scharlau), sodium bicarbonate (NaHCO₃, AR, Scharlau), sodium sulfide nonahydrate (Na₂S·9H₂O, 96%, Energy), sodium hydroxide (NaOH, 99.8%, Scharlau), sodium dithionite (Na₂S₂O₄, 88%, 3 A), ammonium chloride (NH₄Cl, AR, Scharlau), sodium periodate (NaIO₄, 99%, Energy), ruthenium(III) chloride hydrate (RuCl₃·xH₂O, 99%, Energy), 2,3-dichloro-5,6-dicyano-1,4-benzoquinone (DDQ, 98%, Energy), 1,3,5-triformylphloroglucinol (TFG, 99%, Energy), carbon nanotubes (CNTs, 98 wt%, 3 A), mesitylene (98%, Energy), 1,4-dioxane (98%, Energy), acetic acid (99.7%, RCl Labscan), methanol (AR, RCl Labscan), dichloromethane (DCM, AR, RCl Labscan), acetonitrile (AR, RCl Labscan), chloroform (AR, RCl Labscan), ethyl acetate (AR, RCl Labscan), N,N-dimethylformamide (DMF,AR, RCl Labscan), acetone (AR, RCl Labscan).

Calcium bis(trifluoromethanesulfonyl)imide (Ca(TFSI)₂, 99.5%, DoDoChem), tetraethylene glycol dimethyl ether (TEGDME, ≥ 99%, Sigma-Aldrich), 1,3-dioxolane (DOL, 99.5%, Macklin), calcium borohydride bis(tetrahydrofuran) complex (Ca(BH₄)₂·2THF, >95%, Sigma-Aldrich), 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP, 99.5%, Macklin), anhydrous 1,2-dimethoxyethane (DME, ≥ 99.5%, Sigma-Aldrich), N-methyl-2-pyrrolidone (NMP, 99.9%, Aladdin), Super P carbon black (carbon content > 99.5%, Canrd), polyvinylidene fluoride (PVDF, ≥ 99.5%, ~600,000 g mol⁻¹, Canrd), polytetrafluoroethylene (PTFE) aqueous dispersion (60 wt.% solids, Canrd), carbon-coated aluminum foil current collectors (Al purity: 99.9%, Al foil thickness: 12 μm, carbon-coating thickness: 1 μm, total thickness: ~13 μm, Canrd), carbon cloth (thickness: ~0.33mm, SCC130, Sinoro), Whatman GF/A glass-microfiber separators (product no. 1820-090, binder-free borosilicate glass, nominal thickness: 260 μm, Cytiva) and activated carbon (AC, YP-50F, Kuraray) were also used as received.

Synthesis of monomers

Synthesis of 1,8-dinitro-9,10-phenanthrenequinone (DNPQ)

DNPQ was synthesized according to the reported procedure⁴³. PQ (1.0 g, 4.8 mmol), fuming HNO₃ (14 mL) and H₂SO₄ (97%, 1.4 mL) were added to a 100 mL flask and kept at 90 °C for 3 h. Then, the reaction solution was diluted with 200 mL of deionized water and neutralized with Na₂CO₃ solution. After filtration, the solid product was recrystallized from acetic acid and dried under vacuum at 25 ± 1 °C overnight to get DNPQ powder (0.6 g) with a yield of 39.6%. ¹H NMR (400 MHz, CDCl₃-d₃, δ [ppm]): 8.36 (d, 2H), 8.66 (dd, 2H), 9.08 (d, 2H).

Synthesis of 2,7-diamino-9,10-phenanthrenequinone (DAPQ)

DNPQ (0.3 g), Na₂S·9H₂O (0.21 g) and Na₂CO₃ (0.22 g) were dissolved in methanol (6 mL) and heated to 90 °C under an N₂ atmosphere overnight. The reaction solution was diluted with 15 mL of deionized water and filtered. The solid filtrate was recrystallized from methanol to get DAPQ powder (0.12 g) with a yield of 40%. ¹H NMR (400 MHz, DMSO-d6, δ [ppm]): 7.64 (d, 2H), 7.07 (d, 2H), 6.84 (dd, 2H), 5.53 (s, 4H).

Synthesis of 4,5,9,10-pyrenetetraone (PT)

PT was synthesized according to literature⁴⁴. Pyrene (2.0 g, 10 mmol), sodium periodate (17.5 g, 82 mmol), and Ruthenium (III) chloride hydrate (0.25 g, 0.2 mmol) were dissolved in a mixed solution including dichloromethane (DCM, 40 mL), acetonitrile (40 mL) and deionized water (50 mL). The solution was kept at 38 ± 1 °C and stirred overnight. After that, the solution was extracted with DCM and the extracted solution was further treated by rotary evaporation to remove the organic solvent. The solid product was recrystallized from chloroform to obtain PT powder (0.5 g) with a yield of 19%. ¹H NMR (400 MHz, DMSO-d6, δ [ppm]), 8.33 (d, 4H), 7.73 (t, 2H).

Synthesis of 2,7-dinitropyrene-4,5,9,10-tetraone (DNPT)

PT (0.5 g) was dissolved in 20 mL of mixed acid solution composed of concentrated HNO₃ (70%, 4 mL) and concentrated H₂SO₄ (98%, 16 mL). The solution was stirred at 90 ± 1 °C for 4 h. After that, the solution was diluted with 150 mL of deionized water and neutralized with NaHCO₃ solution. After filtration, the solid product was dried under vacuum at 65 ± 1 °C for 24 h to get DNPT (0.5 g) with a yield of 75%. ¹H NMR (400 MHz, DMSO-d6, δ [ppm]), 8.89 (s, 4H).

Synthesis of 2,7-diaminopyrene-4,5,9,10-tetraone (DAPT)

DNPT (0.5 g), NaOH (4.5 g), and Na₂S₂O₄ (2.3 g) were dissolved in 100 mL of deionized water and kept at 50 ± 1 °C for 15 min. Then, the solution was mixed with 200 mL of saturated NH₄Cl solution and filtered to get the crude product. After drying in vacuum at 25 ± 1 °C overnight, the crude product (0.29 g) and dichlorodicyanobenzoquinone (0.74 g) were dissolved in methanol (20 mL) and then kept at 35 ± 1 °C overnight. After that, the solution was diluted with ethyl acetate filtered and dried to get DAPT powder (0.26 g) with a yield of 49%. ¹H NMR (400 MHz, DMSO-d6, δ [ppm]): 7.35 (s, 4H), 5.97 (s, 4H).

Synthesis of PQ_xPT_1-xCOF

The PQ_xPT_1−xCOF materials were synthesized by a Schiff-base condensation method. Typically, TFG (21 mg, 0.1 mmol), DAPQ (23.9 mg, 17.9 mg, 11.9 mg, corresponding to x = 0.67, 0.50, 0.33 in PQ_xPT_1−xCOF) and DAPT (14.6 mg, 21.9 mg, 29.3 mg, correspondingly), and mesitylene (1.8 mL), 1,4-dioxane (0.6 mL), and aqueous acetic acid (0.4 mL, 6 M) were added in a 15 mL Pyrex tube. After three freeze-pump-thaws, the tube was kept at 120 °C ± 1 °C for 3 days. The black powder was washed with DMF and acetone to get PQ_xPT_1−xCOF (~56 mg). For the PQ_xPT_1−xCOF positive electrode synthesis process, 50 wt.% CNTs were added to enhance the electronic conductivity.

Synthesis of Ca[B(hfip)₄]₂

For Ca[B(hfip)₄]₂ salt preparation, Ca(BH₄)₂·2THF (1.07 g, 5 mmol) was dissolved in 10 mL anhydrous DME in an argon atmosphere. (CF₃)₂CHOH (7.14 g, 42.5 mmol) in DME was added dropwise in above solution over 30 min under vigorous stirring. The reaction mixture was maintained at 25 °C ± 1 °C for 20 h, after which volatiles were removed by vacuum drying at 40 °C. The resultant white solid was dried at 60 ± 1 °C for 24 h to yield Ca[B(hfip)₄]₂ salt (7.34 g, 88%) for 0.4 M Ca[B(hfip)₄]₂/DME electrolyte preparation.

Electrode preparation

PQ_xPT_1−xCOF positive electrodes were synthesized by blending PQ_xPT_1−xCOF, Super P carbon powder and PVDF binder in a 7:2:1 weight ratio in NMP solvent. The mixture was magnetically stirred for 12 h and manually blade-coated onto carbon-coated aluminum foil current collectors using a four-sided film applicator. The coated electrodes were vacuum-dried at 90 ± 1 °C for 12 h and cut into 12 mm disks using a manual electrode disc punch. The PQ_xPT_1−xCOF active material loading was ~2 mg cm⁻². For pouch cells, single-side-coated PQ_0.5PT_0.5COF positive electrodes were prepared using the same procedure and cut into 3 × 3 cm² squares, with an active-material loading of ~1.7 mg cm⁻² based on pure PQ_0.5PT_0.5COF. PQ_0.5PT_0.5COF positive electrodes for in situ Raman measurements were prepared using PQ_0.5PT_0.5COF, Super P, and PTFE in a 7:2:1 weight ratio. The mixture was ground for 20 min, further ground with ethanol until dry, coated onto 100-mesh titanium mesh (≥ 99.5%, Kangwei), vacuum-dried at 90 °C ± 1 °C for 12 h, and cut into 12 mm disks.

AC counter electrodes were prepared by a similar method with AC and PVDF mixture (9:1, wt.%: wt.%) and coated onto carbon cloth, with an AC loading of ~35 mg cm⁻², and cut into 12 mm-diameter disks. Ca-metal electrodes (purity: 99.95%, thickness: 1 mm, ZhongNuo Advanced Material) were punched into 12 mm-diameter disks for coin cells using a manual hollow punch and cut into 3 × 3 cm² foils for pouch cells using scissors. All Ca electrodes were polished on both surfaces with 120-grit sandpaper to a thickness of ~0.3 mm and used immediately after preparation. No additional mechanical pressure was applied during preparation. All procedures were conducted in an Ar-filled glovebox (H₂O and O₂ < 0.1 ppm, 25 ± 1 °C).

Coin cell assembly

CR2032 coin cells (20 mm in diameter and 3.2 mm in height, 304 stainless, Canrd) were assembled in an argon-filled glovebox using PQ_xPT_1−xCOF positive electrodes as the working electrodes, AC (capacitive interactions) electrodes as the counter electrodes, and Whatman GF/A glass-microfiber separators cut into 16 mm-diameter disks. Each cell contained one single-side-coated PQ_0.5PT_0.5COF positive electrode and one AC counter electrode, with a negative-to-positive electrode capacity ratio (N/P ratio) of ~1.7:1. 130 μL of 1 M Ca(TFSI)₂ in TEGDME/DOL (1:1, v/v) was added to each cell. Electrochemical testing was performed in a temperature-controlled environmental chamber maintained at 25 ± 1 °C.

Pouch cell assembly

Pouch cells were assembled in 4 × 6 cm² aluminum-laminated pouches with titanium tabs. Each pouch cell employed a single-layer configuration consisting of one single-side-coated PQ_0.5PT_0.5COF positive electrode and one Ca metal foil negative electrode, both 3 × 3 cm² in size, separated by a 3.5 × 3.5 cm² Whatman GF/A glass-microfiber separator. The Ca metal foil negative electrode was placed directly on an ultrathin high-purity titanium foil current collector (thickness: 1 μm, purity: 99.999%, Saibo). Titanium tabs were spot-welded directly to the carbon-coated aluminum foil current collector of the positive electrode and the titanium foil current collector of the negative electrode using a handheld spot welder. The N/P ratio was ~142. 500 μL of 0.4 M Ca[B(hfip)₄]₂/DME electrolyte was injected into each pouch cell inside an argon-filled glovebox, followed by direct heat sealing without a separate gas-release step. The sealed cells were rested for 2 h before electrochemical testing to ensure sufficient electrolyte wetting. During cycling, the pouch cells were held between rigid plates using a screw-tightened fixture, and the applied external pressure was not quantitatively measured. Electrochemical testing was performed in a temperature-controlled environmental chamber maintained at 25 °C ± 1 °C.

Materials characterizations

Powder X-ray diffraction (PXRD) patterns were acquired using a Rigaku SmartLab 9 kW diffractometer (Cu Kα, λ = 1.5406 Å) in the 2θ range of 3–50°. Morphological analysis of COFs was performed via field-emission scanning electron microscopy (FE-SEM, TESCAN MIRA3). Chemical bonding configurations were verified by Fourier-transform infrared (FTIR, Bruker Vertex 70) with a ZnSe attenuated total reflectance (ATR) accessory (400–4000 cm⁻¹, 4 cm⁻¹ resolution). Ultraviolet-visible (UV-vis) absorption spectra were recorded on a PerkinElmer UV-vis-NIR using quartz cuvettes (path length: 1 mm).

Ex situ sample preparation and characterization

For ex situ characterization, cells were stopped at the designated electrochemical states and disassembled in an argon-filled glovebox (H₂O and O₂ < 0.1 ppm, 25 ± 1 °C). All ex situ samples were rinsed two times in anhydrous TEGDME solvent to eliminate residual Ca(TFSI)₂ electrolyte, subsequently vacuum-dried for 1 h. The samples were sealed under argon and transferred to the corresponding instruments without air exposure.

Ex situ FTIR spectra of the cycled electrodes were recorded on a Bruker Vertex 70 spectrometer with a ZnSe ATR mode (1200–1800cm⁻¹, resolution:4 cm⁻¹, 64scans). Ex situ X-ray photoelectron spectroscopy (XPS) spectra were collected on a Thermo Scientific Nexsa G2 spectrometer (monochromatic Al Kα source). The samples were etched with Ar⁺ ions before measurement (1 keV, 30 s), and the binding energies were calibrated against the C 1 s peak at 284.8 eV. Ex situ electron paramagnetic resonance (EPR) spectra were collected on a Bruker EMXplus-6/1 spectrometer using thin-film samples (25 °C ± 1 °C, without light irradiation).

In situ Raman measurements

In situ Raman measurements were performed using a CR2032 coin cell with a perforated positive electrode-side case, with all other assembly conditions identical to those of the coin cells for electrochemical tests. The spectra were acquired on a WITec alpha300 R confocal microscope (633 nm laser, potential interval: 0.1 V, specific current: 0.2 A g⁻¹, 25 ± 1 °C).

Electrochemical measurements

The potential of the AC electrodes versus Ca²⁺/Ca was calibrated using an Ag⁺/Ag reference electrode and the Fc⁺/Fc redox couple as an internal reference electrode dissolved in the same electrolyte solution. Electrochemical characterization was carried out using CV with a BioLogic VMP3 workstation between −1.5 and 0.5 V vs. AC at various scan rates. Galvanostatic cycling tests were performed with a Neware system to evaluate cell performance within the same voltage range at different specific currents, and specific capacities were calculated based on the mass of the active materials. CV and galvanostatic cycling tests for Ca metal full cells were conducted in a voltage range of 0.5–3.5 V.

For kinetic analysis, the apparent Ca²⁺ diffusion coefficient (D) for each COF positive electrode was calculated from data collected across a range of 1 to 32 mV s⁻¹. Ex situ EIS measurements were performed in potentiostatic mode using symmetric cells assembled from pristine, calciated, and decalciated electrodes at open-circuit voltage after a stabilization period of 1 h, with a single-sine AC perturbation of 10 mV over the frequency range of 100 kHz to 0.01 Hz and 6 logarithmically spaced data points per frequency decade. The impedance spectra were fitted using ZView software with an equivalent circuit of Rs-(Rct | |CPE). Each measurement was repeated three times to ensure reproducibility. All electrochemical tests were performed at 25 ± 1 °C.

Calculation of theoretical and practical capacities

The theoretical capacity of COF (C_COF,theo) was calculated using:

$$C_{\mathrm{COF},\mathrm{theo}}=\frac{nF}{3.6M}=\frac{96485n}{3.6M}$$ 1

where n is the number of electrons transferred, F is the Faraday constant (96485 C mol⁻¹), and M is the molar mass of the repeating unit of COFs. A two-layer structural model was used for calculating the theoretical capacities of all PQ_xPT_1−xCOFs in this work. For PQ_0.5PT_0.5COF, the molar mass of the two-layer repeating unit is 5248.64 g mol⁻¹. Assuming the insertion of 21 Ca²⁺ ions (corresponding to 42 e⁻ transfer) to fully enolize the carbonyl groups in PQ_0.5PT_0.5COF, the theoretical capacity is calculated to be 215 mAh g⁻¹. Using the same method, theoretical capacities of other PQ_xPT_1−xCOFs are calculated and listed in Supplementary Table 4.

The specific capacities are normalized to the mass of active materials. The capacity contributions from conductive additives (CNTs and Super P, Supplementary Fig. 29) are negligible in comparison to COFs and are therefore neglected in general performance discussions. In order to provide an accurate evaluation of the theoretical capacity utilization efficiencies, we further corrected the specific capacities from COFs ($C_{\mathrm{COF},\mathrm{corr}}$) by subtracting the contributions from conductive:

$$C_{\mathrm{COF},\mathrm{corr}}=C_{\mathrm{COF},\mathrm{meas}}-\frac{w_{\mathrm{CNTs}}{C}_{\mathrm{CNTs}}+w_{\mathrm{SP}}{C}_{\mathrm{SP}}}{w_{\mathrm{COF}}}$$ 2

where $C_{COF,meas}$, $C_{CNTs}$, and $C_{SP}$ is the measured specific capacities of COF, CNTs, and super P electrodes in Ca ion cells under identical measurement conditions, $w_{\mathrm{COF}}$, $w_{\mathrm{CNTs}}$ and $w_{\mathrm{SP}}$ are the mass fractions of each component in the COF positive electrodes. Then, the theoretical capacity utilization efficiency can be directly obtained by C_{COF, corr}/C_{COF, theo}.

Calculation of specific energy

The specific energy reported in this work is calculated based on the mass of the positive electrode active material. The gravimetric specific energy (E, Wh kg⁻¹) is determined according to the following equation:

$$E=C\times V_{\mathrm{average}}$$ 3

where C (mAh g⁻¹) is the reversible specific capacity of the positive electrode, normalized to the mass of the active material (PQ_0.5PT_0.5COF) only, excluding the conductive additive (CNTs) and other electrode components.$V_{\mathrm{average}}$ is the discharge voltage of the full cell, determined from the galvanostatic discharge profile.

Theoretical calculations

First-principles calculations were performed using the Vienna Ab initio Simulation Package (VASP) with plane-wave basis sets, periodic boundary conditions, and projector augmented wave (PAW) pseudopotentials^31,45,46. The generalized gradient approximation (GGA) with the Perdew-Burke-Ernzerhof (PBE) functional was adopted to describe the exchange and correlation interactions⁴⁷. To properly describe the van der Waals interactions, DFT-D3 corrections were applied⁴⁸. The maximum plane-wave cutoff energy was set to 500 eV. A 1×4×1 Monkhorst-Pack k-point grid was adopted for structural optimizations, single-point calculations, and binding energy calculations. The results (Supplementary Fig. 26) show that this k-point mesh provides sufficiently converged total energy. ensuring the reliability of the reported energetic trends. All COF structures were fully relaxed until the force on each atom was smaller than 0.02 eV Å⁻¹ before conducting further calculations. Single-point calculations were carried out with a convergence threshold of 10⁻⁵ eV to determine the total energy of the optimized structure, which was subsequently used for binding energy calculations.

To identify the most favorable intercalation site, the binding energy, which represents the strength of interaction between Ca and PQ_xPT_1−xCOF, was calculated using the following equation:

$$E_{\mathrm{binding}}=E_{\mathrm{COF}+\mathrm{Ca}}-E_{\mathrm{COF}}-E_{\mathrm{Ca}}$$ 4

where $E_{\mathrm{COF}+\mathrm{Ca}}$ is the total energy of the optimized Ca-intercalated PQ_xPT_1−xCOF; $E_{COF}$ is the total energy of the pristine PQ_xPT_1−xCOF; $E_{Ca}$ corresponds to the energy per atom of a face-centered-cubic (FCC) Ca crystal. In general, negative $E_{\mathrm{binding}}$ is conducive to realizing the Ca-intercalation reactions. The lower the binding energy, the more stable the Ca-intercalation⁴⁹.

All structural relaxations and binding energy calculations, including the reference-state calculations for bulk FCC Ca, were performed with spin polarization. Zero-point energy (ZPE) corrections were neglected because the COF structures have large system sizes (e.g., approximately 200 atoms per COF unit), which renders full vibrational frequency calculations computationally prohibitive. For comparative purposes, the omission of ZPE is not expected to affect relative energetic trends across similar structures. The energy per atom of bulk FCC Ca was used as the reference to calculate binding energies, consistent with the experimental cell configuration. While the choice of reference, such as isolated Ca atoms or FCC Ca, would affect the energetic baseline and the absolute binding energies, consistent referencing throughout this work ensures reliable comparison of relative trends across different COF structures.

To quantitatively describe the diffusion behavior of Ca ions within the PQ_xPT_1−xCOF positive electrode materials, the Climbing Image Nudged Elastic Band (CI-NEB) method⁵⁰ was applied to search for transition states along proposed Ca ion diffusion pathways. 7 to 10 intermediate images were linearly interpolated between each possible state until the forces were less than 0.1 eV Å⁻¹. For diffusion barrier calculations, a Γ-point-only Monkhorst-Pack k-point (1×1×1) was used because the corresponding models are substantially large and complex⁵¹. Considering the large lattice parameters and porous structure of COFs, this approximation is reasonable for evaluating diffusion barriers along different pathways. A force convergence criterion of 0.1 eV Å⁻¹ was used for CI-NEB calculations, considering the large supercell sizes (up to ~400 atoms), the use of multiple intermediate images, and the structural flexibility of COF structures. This choice provides a reasonable balance between computational efficiency and accuracy, in agreement with the recent large-scale CI-NEB calculations of complex systems^52,53. Although this threshold may introduce minor uncertainty in absolute diffusion barrier values, it is not expected to affect the relative comparison of diffusion barriers and our mechanistic insights in this work.

Author contributions

Z.L.X. supervised the project and revised the manuscript. Y.Y.M., Z.Y.Y., and Y.D.C. contributed equally to this work. Y.Y.M. performed the main experiments, analyzed the data, and wrote the manuscript. Z.Y.Y. synthesized the materials. Y.D.C. performed the DFT calculations. Q.M. and Z.Y.Z. assisted with computational modeling. Q.Q. and Y.K.H. performed the in situ Raman measurements. Z.Z. performed TEM characterization. C.B.D. carried out FTIR analysis. J.Y.Y. and M.C.W. assisted with full-cell assembly. C.F.C. and J.Q.H. contributed to data interpretation and discussion. Y.M.W., Y.K., and D.A. revised the manuscript. All authors reviewed and approved the final manuscript.

Peer review

Peer review information

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

Funding

This work was supported by the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. 15305022 and Collaborative Research Fund No. C6047-23GF), the Research Institute for Advanced Manufacturing (Project No. 1-CDLR), the General Program sponsored by the Basic Research Fund of the Shenzhen Natural Science Foundation (Project No. JCYJ20250604184257074), and the Research Committee of The Hong Kong Polytechnic University (project codes G-UARH, 4-ZZSE and U-CDCL).

Data availability

The data supporting the findings of this study are available within the article and its Supplementary Information. Source Data are provided with this paper. The atomic coordinates of all optimized computational models are provided in Supplementary Data 1-6. 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 https://doi.org/10.1038/s41467-026-77167-y.