Metal-center electron affinity modulates multicolor electrochromism in 2D conjugated metal-organic frameworks Authors: Qi Zhao, Jing Yang, Zhendong Liu, Xiaohan Chen, Jinkai Wang, Yulin Gao, Xue Chen, Xingyang Wang, Wanwan Wang, Fei Wang, Chengzhi Zhang, Yanfeng Gao, Yong-Wei Zhang, Dan Zhao, John Wang Published: 2026 ABSTRACT Two-dimensional conjugated metal-organic frameworks are emerging platforms for electrochromic devices, yet the current stage is limited by monotonous color and an unclear coloration mechanism. While prior studies emphasize the ligand-centered redox process, the role of metal centers is often overlooked. Herein, we report a metal-center electron-affinity mechanism that is critical to electrochromic performance. We first employ an unsupervised machine learning method to screen divalent metal-catecholate (CAT) systems, identifying Zn, Ni, and Cu as optimal candidates. All three systems exhibit multicolor transitions from green to blue and eventually to purple; however, Zn- and Ni-CAT-1 significantly outperform Cu-CAT-1, with nearly threefold faster switching speeds (4.8/5.4 vs. 14.4 s) and cycling stability enhanced by two orders of magnitude ( > 20000 vs. < 300 s). Cu2+ (3d9) possesses a redox-active configuration that favors electron affinity, thereby requiring additional electron participation and a higher energy demand than the relatively inert Ni2+ (3d8) and Zn2+ (3d10). This valence variation also induces stronger metal-oxygen coupling, ultimately leading to slower redox kinetics and structural degradation. As proof of concept, we demonstrate the applicability of these findings to camouflage, displays, and wearables for designing multicolor electrochromic devices. Introduction Electrochromic (EC) materials reversibly change their optical properties under an applied electric field1,2. This unique property has enabled the commercialization of energy-efficient devices, including smart windows and light-modulating systems like anti-glare mirrors, which operate via blue-to-transparent switching3,4. Over the recent years, considerable research and industrial efforts have shifted toward realizing multicolor functionality to support emerging applications5,6. Tech companies such as Apple, Google, Huawei, and Xiaomi have integrated color-tunable or black-state features into augmented reality glasses and automotive skylights. However, specific applications, including adaptive camouflage7, pattern displays8, wearable electronics9,10, and sensors11, require multicolor capability to enhance functionality and improve visual experience. Thus far, efforts to realize multicolor electrochromism have largely focused on two main classes of materials: transition metal oxides (TMOs) and polymers8,12. On the one hand, due to their intrinsic structural rigidity and narrow valence-state ranges, TMOs are fundamentally constrained in achieving multicolor tunability13. To compensate, optical engineering approaches, including resonant cavities and plasmonic architectures, are commonly employed to broaden their spectral response14,15. Resonant cavities provide strong optical confinement and tunable light-matter interactions but demand rather strict control over film thickness and refractive index16. Plasmonic resonances enhance local electromagnetic intensities, yet their modulation is generally limited to a single wavelength17. On the other hand, polymers with multiple redox states, such as viologens, polyaniline, and polypyrrole, can undergo reversible structural and energy-gap transitions through electrochemical (de)doping, enabling multicolor switching18,19. Nevertheless, these systems frequently exhibit charge localization, backbone degradation, and loss of conductivity under cycling, limiting the long-term performance and environmental endurance20. Such challenges necessitate alternative material platforms that simultaneously ensure chromatic versatility and electrochemical stability. Two-dimensional conjugated metal-organic frameworks (2D c-MOFs) exhibit relatively high electronic conductivity, tunable pore size, large surface area, and multitype redox states, exceeding the kinetic requirements for multicolor switching21,22. Early studies on MOF-74, NBU-3, and HKUST-1 demonstrated their feasibility, paving the way for the development of 2D c-MOFs23,24. Despite the growing interest, multicolor MOFs remain uncommon, and their reported performance varies widely: some systems exhibit sub-second switching with long-term cycling stability (thousand cycles), while others degrade within a few cycles and require over 30 s for coloration (Table S1). This inconsistency indicates the limited mechanistic understanding of electrochromic behavior in MOFs. Prevailing interpretations attribute color modulation primarily to ligand-centered redox transitions, such as the [NDI]−/[NDI]2− or [PDI]− and [PDI]2− redox couples interconversion, with metal nodes commonly regarded as coordination support25,26. Several fundamental issues concerning the role of metal centers remain insufficiently explored (Fig. 1a): (i) How metal-valence-state transitions affect charge transfer pathways and charge kinetics. Charge transfer typically occurs through the π-conjugated ligand network. However, when the metal center undergoes redox transitions, electrons are partially transferred via the metal site. It remains to be seen whether charge pathways can influence charge mobility and modulate the energy barriers of redox processes. (ii) How local coordination and electronic distribution impact integrity and charge transport. Different metal ions give rise to distinct coordination environments and electron distributions, which in turn can influence structural robustness and charge-transfer efficiency. (iii) How metal-ligand orbital coupling governs redox behavior. The strength of π-d interactions, especially between metal centers and oxygen atoms, determines redox-state transitions, with stronger interactions potentially hindering reversibility and increasing the energy barrier for switching. Critically, no established principles currently exist to guide the metal selection or design for optimal performance.Fig. 1Concept and AI-assisted method.a Schematic comparison between the conventional understanding (left) and the proposed hypothesis (right) regarding the role of metal centers in electrochromism. b Workflow of the unsupervised machine learning approach based on correlation and clustering analysis. c Standardized heatmap visualization comparing divalent metals based on cost, toxicity, catecholate binding energy, Lewis acidity, optical contrast, and cycling life. Concept and AI-assisted method. a Schematic comparison between the conventional understanding (left) and the proposed hypothesis (right) regarding the role of metal centers in electrochromism. b Workflow of the unsupervised machine learning approach based on correlation and clustering analysis. c Standardized heatmap visualization comparing divalent metals based on cost, toxicity, catecholate binding energy, Lewis acidity, optical contrast, and cycling life. Herein, using an unsupervised machine learning approach, a standardized heatmap visualization identifies Zn, Ni, and Cu as the most representative metal candidates across a series of 2D c-MOFs. The different-sized flexible multicolor electrochromic devices (FMEDs) are developed for adaptive camouflage and display applications. Contrary to conventional empiricism favoring redox-active metals, we prioritize the low electron-affinity selection principle as critical to performance. Both experimental and simulation results reveal a Cu+/Cu2+ transition and associated metal-to-ligand charge transfer, whereas Zn- and Ni-based analogs maintain their charge states. Consequently, Zn- and Ni-CAT-1 exhibit faster switching (4.8/5.4 vs. 14.4 s) and superior cycling stability (>20,000 vs. <300 s). These variations in the local metal coordination environment increase energy demand, modify metal-oxygen interaction, hinder charge transport, and compromise the structural stability, collectively contributing to performance differences. Results Unsupervised machine learning-guided screening of divalent metals via standardized heatmap We first confirmed the ligand, 2,3,6,7,10,11-hexahydroxytriphenylene (H12C18O6, HHTP), as a fixed organic linker prior to metal ion selection. HHTP is a typical redox-active linker widely used in redox-driven energy systems27. As a well-established catecholate unit, it supports multiple redox states, providing a robust foundation for multicolor electrochromic switching. In addition, HHTP readily coordinates with metal centers via a simple hydrothermal process28. Its hexa-anionic character makes HHTP well-suited for coordination with divalent metal ions (M2+), facilitating the charge-balanced formation of the prototypical M3(HHTP)2 architecture (Supplementary Note 1). Despite being divalent, ions, such as Be2+, Sr2+, Ba2+, Pb2+, Cd2+, Sn2+, Cr2+, and Ra²⁺ are rarely used due to their unfavorable chemical properties, high toxicity, and high cost (Supplementary Note 2). The remaining divalent metal ions exhibit no apparent drawbacks and show broadly comparable characteristics. Manual ranking, however, is limited in multi-dimensional analysis, as reliance on a single indicator or a narrow subset of metrics, thereby hindering the identification of materials with comparable performance across the full indicator set. The assignment of metric importance is inherently subjective, leading to variability in rankings. Manual sorting also struggles to capture complex patterns. For example, two materials may differ markedly in cost yet exhibit comparable overall performance when stability, optical contrast, and toxicity are considered jointly. Such relationships are often obscured in single-metric rankings. In such a case, we employed an unsupervised machine learning approach based on correlation and clustering analysis to rapidly and accurately screen optimal metal candidates (Fig. 1b and detailed procedures are presented in Supplementary Note 3). The key indicators (Supplementary Note 4 and Table S1), including cost, toxicity, binding energy with catecholate, Lewis acid suitability, optical contrast, and cycling life, are standardized on a comparable scale using the Z-score normalization (implemented via StandardScaler)29. Hierarchical clustering analysis was applied to all standardized metrics in parallel to compute pairwise similarities in a multi-dimensional space, and objectively cluster both metals and metrics without prior assumptions. The resulting dendrogram determines the optimal ordering of rows (metals) and columns (metrics) in the heatmap, yielding a structured visualization. As indicated by strong red signals in the heatmap (Fig. 1c), Ni, Cu, and Zn demonstrate optimal performance, especially in optical contrast, cycle life, and catecholate binding energy, and rank near the top across most metrics. Structural and morphological characterization of selected metal-cat frameworks Zn/Cu/Ni-HHTP samples were prepared by a hydrothermal process (Fig. 2a and experimental details are supplemented in Supporting information). These crystalline materials, known as metal-catecholates (M-CATs), demonstrate a layered structure with an extended honeycomb lattice. Previous studies have shown that M-CAT frameworks can exhibit structural variations, including different stacking modes, such as AA-stacked (eclipsed), slipped stacking, alternative stacking of 2D layers, or even 3D frameworks30–33. In such a case, we first performed micro electron diffraction (MicroED) measurements to probe the crystallographic features of the synthesized MOFs. Although the diffraction patterns exhibit well-defined reciprocal lattices, the limited intensity of high-angle reflections prevents full structural solution and precise determination of the stacking configuration (Figs. S1–S3 and Supplementary Note 5). Nevertheless, reliable unit cell parameters could be extracted, and notably, the Zn-, Ni-, and Cu-based samples show highly consistent lattice parameters, indicating similar underlying framework structures (Table S2). To further resolve the stacking mode, we carried out Rietveld refinement of the PXRD patterns (Fig. S4). The refined lattice parameters are in good agreement with the MicroED results within experimental error (Fig. S5 and Table S3). Based on these combined analyses, all three samples can be assigned to an AA-stacked two-dimensional M-CAT framework. They also exhibit nearly identical peaks, corresponding to the (100), (200), (210), and (220) planes, being consistent with the previous reports34. Additional phase characterizations were conducted using X-ray photoelectron spectroscopy (XPS), Fourier transform infrared spectroscopy (FTIR), and Raman spectroscopy, as detailed in Supplementary Note 6 and Figs. S6–S8. Further parameters, such as specific surface area, electronic conductivity, and optical bands are presented in Supplementary Note 7 and Fig. S9, confirming the successful synthesis of metal-CATs and their physicochemical advantages. Scanning electron microscopy (SEM) studies show that metal-CATs form an interlaced nanorod-like network with average lengths of 200–600 nm and diameters of 10–50 nm (Fig. S10). This structure enhances electrolyte penetration and promotes efficient redox processes35. To further determine the atomic arrangement, cryo-transmission electron microscopy was performed at −165 °C to preserve the fragile MOF structure under a high-energy electron beam. Cu-CAT-1 shows clear lattice patterns with a lattice spacing of 1.9 nm, corresponding to the (100) crystal plane (2θ = 4.6°) (Figs. 2b, c, and S11). Similar lattice fringes of 2.0 and 1.8 nm were observed in Zn-CAT-1 (Figs. 2f and S12, S13) and Ni-CAT-1 (Fig. 2j), respectively. Notably, all samples with well-defined honeycomb frameworks feature ordered, spacious channels of approximately 2.0 nm along the [001] direction (Fig. 2d, g, h, k, l), which promote efficient ion diffusion. Such structural advantages surpass those of conventional electrochromic materials, including TMOs and conducting polymers21,36. The corresponding elemental mapping images confirm the uniform distribution of C, O, and Zn/Cu/Ni (Fig. 2e1-4, i1-4, m1-4). Variations in the central metal node have minimal impact on the phase, structure, and morphology of metal-CATs.Fig. 2Phase characterization and morphology of metal-CATs.a Synthetic diagram and molecular packing of metal-CATs. b Morphology of Cu-CAT-1. c Lattice details of Cu-CAT-1. d Honeycomb structure of Cu-CAT-1. The insert shows the inverse FFT images. e 1-4 EDS elemental mapping results of Cu-CAT-1. f Lattice details of Zn-CAT-1. g Honeycomb structure of Zn-CAT-1. h Inverse FFT images of (g). i 1-4 EDS elemental mapping results of Zn-CAT-1. j Lattice details of Ni-CAT-1. k Honeycomb structure of Ni-CAT-1. l Inverse FFT images of (k). m 1-4 EDS elemental mapping results of Ni-CAT-1. Phase characterization and morphology of metal-CATs. a Synthetic diagram and molecular packing of metal-CATs. b Morphology of Cu-CAT-1. c Lattice details of Cu-CAT-1. d Honeycomb structure of Cu-CAT-1. The insert shows the inverse FFT images. e 1-4 EDS elemental mapping results of Cu-CAT-1. f Lattice details of Zn-CAT-1. g Honeycomb structure of Zn-CAT-1. h Inverse FFT images of (g). i 1-4 EDS elemental mapping results of Zn-CAT-1. j Lattice details of Ni-CAT-1. k Honeycomb structure of Ni-CAT-1. l Inverse FFT images of (k). m 1-4 EDS elemental mapping results of Ni-CAT-1. Metal-dependent electrochromic performance and multicolor demonstration in metal-CATs A series of optical and electrochemical measurements was conducted to evaluate the influence of central metal ions on multicolor electrochromic behavior. Metal-CAT electrodes were fabricated by a facile spray-coating method (details are provided in the Supporting Information) and evaluated in a three-electrode system (Ag/AgCl as the reference electrode and Pt as the counter electrode). All samples were prepared under identical film deposition conditions, including the spray concentration, film-heating temperature, and annealing time, to ensure a fair comparison. Cross-sectional SEM analysis confirms that the three samples exhibit comparable film thicknesses of 518.4, 532.7, and 528.0 nm (Fig. S14). All metal-CATs display multicolor behaviors. In situ optical transmittance spectra of Zn/Ni/Cu CAT-1 demonstrate a blue shift from 580 nm to 480 nm and further to 460 nm under applied voltages of −1, +0.3V, and +1 V, respectively, accompanied by color transitions from green to blue then to purple (Fig. 3a), unlike most MOF systems that only show two colors (Table S4). More specifically, optical modulation of Zn-CAT-1 at 580, 480, and 460 nm are 40.7%, 29.0%, and 23.9%, respectively, comparable to those of Ni-CAT-1 (39.0% at 580 nm, 24.2% at 480 nm, and 20.5% at 460 nm) but significantly higher than those of Cu-CAT-1 (17.1% at 580 nm, 8.4% at 480 nm, and 5.7% at 460 nm). The optical energy band gap (Eg) is responsible for these color changes, with the Eg of Zn-CAT-1 shifting from 2.50 to 2.58 eV and further to 2.62 eV under applied voltages of −1, +0.3, and +1 V, respectively (Fig. S15a). The same trends are demonstrated in Ni-CAT-1 and Cu-CAT-1 (Fig. S15b, c). CIE color coordinates provide quantitative color distribution (Fig. S16). Digital photographs further illustrate the color performance of metal-CATs (Fig. 3b1-3–d1-3). Most MOF systems show only two-color transitions. Although organic linkers may theoretically possess multiple oxidation states, only a single ligand redox couple (Ln/Ln-1) is practically reversible and spectroscopically distinguishable within the accessible potential window. Higher-order ligand oxidation states often require excessively high or low potentials, become electrochemically irreversible, or induce structural degradation. Furthermore, intermediate redox states are often unstable due to radical coupling, ligand cleavage, or local coordination rearrangements, causing the system to collapse to a two-state electrochromic response. These limitations are further exacerbated by the intrinsically low electrical conductivity of most conventional MOFs (typically 10−12–10−8 S cm−1), which kinetically suppresses higher-order redox processes37. Moreover, in many cases, different oxidation states mainly modulate absorption intensity rather than peak position, resulting in changes in color depth instead of color types. In contrast, the present HHTP-based framework overcomes these limitations by stabilizing multiple sequential ligand redox couples, enabling a three-color electrochromic response. Cyclic voltammetry (CV) curves also indicate that all three samples exhibit three pairs of redox states. For Zn-CAT-1, the redox pairs at −0.58 V/−0.07 V, −0.24 V/0.2 V, and 0.28 V/0.8 V correspond to green, blue, and purple colors, respectively (Fig. S17). Similar results are observed for Cu-CAT-1 and Ni-CAT-1 (Fig. S17). The redox potentials of these three samples shift slightly, suggesting that the electronic structure of the materials changes during the redox reaction, which leads to variations in electrochemical kinetics38,39. Following, as shown in Figs. 3e and S18, the coloring time/bleaching time of Zn-CAT-1 at 580, 480, and 460 nm are 4.8/5.0, 7.5/4.5, and 9.6/5.8 s, respectively, while for Ni-CAT-1, they are 5.4/6.0, 9.8/3.0, and 11.5/6.0 s. They are significantly faster than Cu-CAT-1 (14.4/22.0, 22.0 /17.0, and 28/18.5 s). More importantly, Zn-CAT-1 maintains 93% of its original optical performance after 20,000 s, while Ni-CAT-1 maintains nearly 90% (Fig. 3f). In sharp contrast, Cu exhibits rapid degradation, falling to a retention level similar to that of Zn- and Ni-CAT-1 after only 300 s. Given the identical testing conditions and comparable crystal structures, the intrinsic metal identity and its coupling with organic ligands are inferred to be the primary contributors to the performance differences. To further validate the unsupervised screening results, non-selected metal centers, namely Co, Mn, and Fe, were examined experimentally. The limited framework formability and poor color-transition behavior observed for the Co-, Mn-, and Fe-based systems are consistent with their positions in the unsupervised screening map, thereby further supporting the selection of Zn, Ni, and Cu as the most promising metal centers (Supplementary Note 8 and Figs. S19–S21).Fig. 3Electrochromic performances and multicolor demonstrations.a In situ transmittance spectra (800–380 nm) of Zn-CAT-1. Digital color-transition photos for b1-3 Zn-CAT-1, c 1-3 Ni-CAT-1, and d 1-3 Cu-CAT-1. e Switching speed at 580 nm. f Cycling life at 580 nm. Biomimetic leaf-like FMEDs for visual and thermal camouflage. g Artificial leaf-shaped FMED. The insert is a real leaf. h Visual camouflage. h1–h3 Infrared thermal images showing FMED embedded among real leaves. i Multicolor displays composed of 1 cm × 1 cm pixels, enabling independent color modulation for numerical patterns. j 1-2 Flexible displays presenting “bird” and “butterfly” patterns with tunable colors. k 1-3 Wrist-worn bracelets displaying “NUS” in different colors. Electrochromic performances and multicolor demonstrations. a In situ transmittance spectra (800–380 nm) of Zn-CAT-1. Digital color-transition photos for b1-3 Zn-CAT-1, c 1-3 Ni-CAT-1, and d 1-3 Cu-CAT-1. e Switching speed at 580 nm. f Cycling life at 580 nm. Biomimetic leaf-like FMEDs for visual and thermal camouflage. g Artificial leaf-shaped FMED. The insert is a real leaf. h Visual camouflage. h1–h3 Infrared thermal images showing FMED embedded among real leaves. i Multicolor displays composed of 1 cm × 1 cm pixels, enabling independent color modulation for numerical patterns. j 1-2 Flexible displays presenting “bird” and “butterfly” patterns with tunable colors. k 1-3 Wrist-worn bracelets displaying “NUS” in different colors. We employed the optimal sample, Zn-CAT-1, as the working electrode and integrated it with lithium-polyacrylamide hydrogel electrolyte and PET/ITO flexible substrates to fabricate FMEDs in various sizes, catering to diverse application scenarios40. A biomimetic FMED featuring a leaf-like structure (Fig. 3g) was developed to mimic the appearance of natural leaves and to seamlessly integrate into the environment for visual camouflage (Fig. 3h). Typical flexible substrates, such as ITO/PET, show a relatively low emissivity (ɛ) of 0.37, while natural leaves exhibit a much higher value of 0.93, ranging from 2.5 to 20 μm. This mismatch significantly increases the risk of device exposure under thermal imaging. In contrast, our leaf-like device achieves a comparable ɛ of 0.89, closely matching the thermal radiation performance of real foliage (Fig. S22). As a result, the FMED effectively avoids infrared detection, enabling near-indistinguishable concealment within leaf-covered surroundings (Fig. 3h1–h3). We have also fabricated an FMEDs panel containing 36 individual pixels (size: 6 cm × 6 cm), each measuring 1 cm × 1 cm (Fig. 3i). This structure allows for the independent control of each pixel. By applying 1 V to specific pixels, for example, a purple digit “5” can be displayed. Notably, even without an external voltage, a single charge can sustain a stable pattern for nearly 90 min, making it well-suited for applications that do not require frequent refreshes (Figs. S23 and S24). Compared to a commercial white OLED display (172.2 Wh m−2), the energy consumption is significantly reduced to just 0.005%, suggesting ultra-low energy consumption and application potential (Figs. S25 and S26, Supplementary Note 9). Furthermore, under the same voltage conditions, dynamic color patterns of “butterfly” and “bird” can also be achieved (size: 7 cm × 8 cm, Fig. 3j1,2). Wrist-worn bracelets (size: 4 cm × 20 cm, Fig. 3k1-3) demonstrate strong wearable potential, exhibiting different “NUS” patterns. Bending tests were further performed on flexible displays and wrist-worn bracelets (Fig. S27). Flexible displays and wrist-worn devices can retain 76.1% and 97.1% of their original optical contrast after 500 bending cycles, respectively, demonstrating their mechanical robustness. To broaden their practical conditions, FMEDs maintain reliable operation even at temperatures as low as −20 °C, making them suitable for harsh-environment applications (Fig. S28). Mechanistic insight: metal-regulated coloration behavior Before examining the substantial performance differences driven by central metal ions, we first clarify the prerequisites for the overall coloration mechanism. Most prior studies have attributed color-switching behavior to the stepwise oxidation of organic ligands34,35,41–43, wherein each electron loss generates a radical species under an increasing positive voltage (Fig. S29). The formation of C=O bonds has been identified as a critical factor in the oxidized coloration process34,41,43,44. Notably, this C=O formation occurs spontaneously due to the intermediate radical delocalization and π-conjugated nature of the ligand (Supplementary Note 10)45,46. Attributing to coordination saturation at the oxygen sites, distortions or rearrangements of metal-oxygen bonds can facilitate this C=O bond formation. In other words, the acceleration of the metal-oxygen bond transformation from predominantly ionic to coordinative character promotes C=O formation, thereby expediting the oxidized coloration process. Our ex situ FTIR results suggest that C=O bonds are consistently observed regardless of the type of central metal ion (Fig. S30). In situ electron paramagnetic resonance (EPR) results provide direct evidence for the chemical identity of the intermediate redox state (Fig. S31). A strong EPR signal centered at g ≈ 2.00 appears exclusively at the intermediate potential (+0.3 V, blue state), while the reduced (−1 V) and fully oxidized (+1 V) states are largely EPR silent. This behavior is characteristic of a ligand-centered oxygen radical (semiquinone-type) generated during stepwise oxidation of the HHTP linker. The absence of well-defined hyperfine splitting and the lack of a fixed intensity ratio between spectral features are consistent with a framework-bound, π-delocalized oxygen radical rather than a free superoxide species. In addition, Fig. S32 shows an increased concentration of free electrons around the metal ions in the oxidized state, which is consistent with distortions in the metal-oxygen bonds. Therefore, our observation aligns with the prevailing view that ligands are essential in the coloration mechanism. However, the role of metal centers cannot be overlooked. Subsequent analysis elucidates their involvement and reveals how metal-ligand interactions modulate the coloration process. We first examine the valence states of different metal centers during the redox process. Ex situ XPS results demonstrate notable changes in the Cu 2p orbital splitting under an applied voltage ranging from −1 to +1 V (Fig. 4a). The peaks at 952.0 and 932.9 eV gradually diminish and eventually disappear, while those at 954.1 and 934.3 eV become more pronounced, indicating a transition of Cu from +1 to +2. However, the peak profiles of Zn and Ni exhibit negligible changes during the redox process, with only slight shifts and variations in intensity (Figs. S33 and S34). Furthermore, X-ray absorption fine structure (XAFS) spectroscopy offers additional insights into valence states, metal-ligand interactions, and local coordination environments of the metal centers. As shown in Fig. 4b, the Cu K-edge X-ray absorption near-edge structure (XANES) spectra of samples at different oxidation states, together with reference standards (CuO, Cu2O, and Cu foil), clearly demonstrate that the absorption edge shifts to higher energy upon oxidation, corresponding to an increase in Cu valence state from ~+1.1 to ~+1.7 (Fig. S35). The extended X-ray absorption fine structure (EXAFS) analysis reveals a pronounced Cu-O bond contraction from 1.94 to 1.86 Å upon oxidation, indicating strengthened Cu-O interactions (Fig. 4c). The wavelet transform (WT) EXAFS maps provide complementary information in both k and R domains (Fig. 4d). In the oxidized state (+1 V), the WT intensity extends to higher k values (~15 Å−1), indicating stronger backscattering and a more ordered Cu-O coordination. In the reduced state (−1 V), attenuation occurs earlier (~13–14 Å−1), consistent with increased structure disorder and decreased coordination numbers. Together with the EXAFS fitting results indicating Cu-O bond elongation, these WT features confirm that oxidation strengthens Cu-O interactions, whereas reduction weakens bonds and induces local disorder. However, the Zn and Ni K-edge XANES spectra exhibit nearly identical features across different redox states, indicating stable valence states (Figs. 4e and S36). The Zn-O and Ni-O bond lengths exhibit minimal changes during redox transitions, decreasing slightly from 2.05 to 2.04 Å (Fig. 4f) and from 2.08 to 2.07 Å (Fig. S37), respectively. The valence states remain stable, with Zn at ~+1.6–1.65 and Ni at ~+1.6–1.7 (Fig. S38). The corresponding WT EXAFS maps of Zn-CAT-1 (Fig. 4g) and Ni-CAT-1 (Fig. S39) also confirm this stability, in sharp contrast to the significant structural rearrangements observed in Cu-based systems. All k-space spectra and fitting results are presented in Figs. S40–S42 and Table S5, confirming the reliability of the analysis.Fig. 4Coloring mechanism.a Ex situ XPS results for Cu 2p conducted at +1, +0.3, and −1 V. b Cu K-edge XANES spectra in the reduced and oxidized state, with respect to Cu foil, Cu2O, and CuO standard samples. c EXAFS spectra of Cu-CAT-1 in the reduced and oxidized state. d WT-EXAFS maps of Cu-CAT-1 in the reduced state at -1 V (left) and oxidized state at +1 V (right), with the color gradient indicating WT intensity from low (green) to high (brown). e Zn K-edge XANES spectra in the reduced and oxidized state, with respect to Zn foil and ZnO standard samples. f EXAFS spectra of Zn-CAT-1 in the reduced and oxidized state. g WT EXAFS maps of Zn-CAT-1 in the reduced (−1 V) and oxidized state (+1 V), with the color gradient indicating WT intensity from low (green) to high (brown). h AFM images of Zn-CAT-1 after 20,000 s of cycling and i the corresponding 3D profiling. j AFM images of Cu-CAT-1 after 2500 s of cycling and k the corresponding 3D profiling. TOF results for l Zn-CAT-1 and m Cu-CAT-1 in reduced state and oxidized state. Coloring mechanism. a Ex situ XPS results for Cu 2p conducted at +1, +0.3, and −1 V. b Cu K-edge XANES spectra in the reduced and oxidized state, with respect to Cu foil, Cu2O, and CuO standard samples. c EXAFS spectra of Cu-CAT-1 in the reduced and oxidized state. d WT-EXAFS maps of Cu-CAT-1 in the reduced state at -1 V (left) and oxidized state at +1 V (right), with the color gradient indicating WT intensity from low (green) to high (brown). e Zn K-edge XANES spectra in the reduced and oxidized state, with respect to Zn foil and ZnO standard samples. f EXAFS spectra of Zn-CAT-1 in the reduced and oxidized state. g WT EXAFS maps of Zn-CAT-1 in the reduced (−1 V) and oxidized state (+1 V), with the color gradient indicating WT intensity from low (green) to high (brown). h AFM images of Zn-CAT-1 after 20,000 s of cycling and i the corresponding 3D profiling. j AFM images of Cu-CAT-1 after 2500 s of cycling and k the corresponding 3D profiling. TOF results for l Zn-CAT-1 and m Cu-CAT-1 in reduced state and oxidized state. Electronically, Cu2+ (3d9) has a redox-active electronic configuration with high electron affinity, making it prone to Cu2+/Cu1+ transitions. Compared to the relatively inert Ni2+ (3d8) and Zn2+ (3d10), this behavior requires additional electron participation and higher energy demand. Furthermore, the higher charge density of Cu2+ results in stronger and more stable Cu2+-O bonds than Cu1+-O, increasing electrostatic attraction and hindering metal-oxygen bond transformation. In contrast, Zn-O and Ni-O bond energies remain nearly unchanged during redox processes. This metal-valence-state transition impedes the oxidation-induced coloration and induces structural changes, including potential degradation. In situ XRD analysis shows that Zn/Ni-CAT samples maintain their diffraction peaks during the redox process (Fig. S43), while the (100) plane of Cu-CAT-1, associated with Cu-O bonds, shifts to higher Bragg angles upon oxidation (Fig. S44), suggesting structural changes. The newly observed peaks, particularly those at 2θ values above 20°, are attributed to diffraction from the ITO substrate. XPS and FTIR analyses conducted before and after cycling further support this conclusion (Supplementary Note 11 and Figs. S45–S47). The corresponding digital photographs demonstrate the color reproducibility of the CAT- based samples. Zn-CAT-1 and Ni-CAT-1 films exhibit minimal color shift after prolonged cycling and retain clear, reversible color changes among different states (Figs. S48–S50). By contrast, Cu-CAT-1 is unable to fully recover its original colored state after prolonged operation. The atomic force microscope provides more morphological evidence. After 20,000 s of cycling, the Zn-CAT-1 and Ni-CAT-1 electrodes exhibit a smoother surface, while severe surface cracking and morphological degradation are clearly observed in Cu-CAT-1 after only 2500 s (Figs. 4h, j, and S51a). 3D surface profiling shows that Zn-CAT-1 and Ni-CAT-1 possess much lower average roughness (65.5 and 65.8 nm) compared to Cu-CAT-1 (131.2 nm), suggesting the enhanced structural stability (Figs. 4i, k, and S51b). In addition, time-of-flight secondary ion mass spectrometry (TOF-SIMS) three-dimensional ion maps reveal the spatial distribution of inserted ions. The measurement was performed in positive mode, and the detected ion signals were assigned based on standard m/z values, including Li+ (m/z = 7), O+/OH+ (m/z = 16/17), CxHy+ (organic fragments from HHTP ligand), and metal+, such as Cu+ (m/z = 63, 65), Zn+ (m/z = 64, 66), and Ni+ (m/z = 58, 60). Noted that TOF-SIMS data show relative secondary-ion intensities rather than absolute concentrations. Due to element-dependent secondary-ion yields and matrix effects (particularly in positive mode for metal+), signal levels are not directly comparable between Zn+, Ni+, and Cu+. Accordingly, these maps are used to track the spatial distribution and the relative evolution of Li+ ions within the same sample upon redox switching. Taking Zn-CAT-1 as an example (Fig. 4l), in the fully reduced (green) state, Li+ ions are uniformly distributed throughout the MOFs, having penetrated the framework while concurrently receiving electrons to maintain charge balance. During oxidation (from green to purple), Li⁺ ions are gradually extracted from the MOF structures, resulting in a decrease in Li+ concentration. A similar phenomenon is observed in Cu/Ni-CAT-1 samples (Figs. 4m and S52). An intriguing observation is that although Cu-CAT-1 undergoes an overall two-electron redox process, the peak areas associated with the blue and purple states in the CV are not approximately twice that of the green state. While the electrochromic switching among the green, blue, and purple states remains fully reversible, the electrochemical response of Cu-CAT-1 is non-ideal. This behavior involves overall multi-electron participation arising from both coupled Cu2+/Cu+ conversion and ligand-centered redox process. These electron-transfer events occur in a stepwise, partially overlapping manner rather than as a single ideal two-electron wave. Besides, strengthened Cu-O interactions and local coordination rearrangements associated with the Cu2+/Cu+ redox couple lead to partial irreversibility or a reduction in the number of electrochemically active sites. These features distinguish Cu-CAT from Ni-CAT and Zn-CAT, where a metal-centered redox process is absent or less disruptive, enabling better cycling stability. Figure S53 summarizes the schematic illustration of the electrochromic coloring transitions. For all samples, although the organic ligands ultimately achieve a similar structure configuration, the valence states of the metal ions either facilitate or impede their structural transformation. Metal-regulated coloration mechanism by DFT Density functional theory (DFT) calculations have been performed to further support the metal-regulated coloration mechanism. The highest occupied molecular orbital (HOMO) displays metal-dependent electron localization, highlighting the role of the metal center in tuning charge distribution (Fig. 5a). In Zn-CAT-1, the HOMO is localized on the HHTP linkers as conjugated π orbitals, indicating that the two electrons originate from the ligand. In Ni-CAT-1, the HOMO is similarly linker-centered, with minimal Ni contribution, confirming the linker as the primary electron donor during oxidation. In contrast, Cu-CAT-1 shows pronounced HOMO localization on the Cu centers, exhibiting clear Cu-O σ-bond character. Cu-CAT-1 requires electrons for π-π conjugated ligand activation, and an additional electron to drive Cu2+/Cu1+ redox reaction. Charge distribution analysis based on the repeating units of M-MOFs further supports these findings (Fig. 5b). Cu undergoes a Cu2+/Cu+ redox transition, whereas Zn and Ni retain their valence state during the coloration process, which is also consistent with our ex situ XPS and XAFS results. Correspondingly, Cu-CAT-1 exhibits a larger energy gap (1.11 eV) than Zn-CAT-1 (0.65 eV) and Ni-CAT-1 (0.93 eV), indicating a higher energy demand for the electrochromic transition (Fig. 5c).Fig. 5DFT simulation results.a HOMO distributions of Zn-CAT-1, Ni-CAT-1, and Cu-CAT-1; dashed circles indicate metal centers. b Charge density differences for the repeating units of Zn-, Ni-, and Cu-based MOFs; navy, gray, red, and white spheres represent metal, C, O, and H atoms, respectively. c Calculated thermodynamic energy required for the purple-to-green color transition. d Geometric analysis of the metal-O bond length under different redox states. e Lattice parameters a and b in fully reduced states and fully oxidized states. f Schematic illustration of metal effects in the redox process. DFT simulation results. a HOMO distributions of Zn-CAT-1, Ni-CAT-1, and Cu-CAT-1; dashed circles indicate metal centers. b Charge density differences for the repeating units of Zn-, Ni-, and Cu-based MOFs; navy, gray, red, and white spheres represent metal, C, O, and H atoms, respectively. c Calculated thermodynamic energy required for the purple-to-green color transition. d Geometric analysis of the metal-O bond length under different redox states. e Lattice parameters a and b in fully reduced states and fully oxidized states. f Schematic illustration of metal effects in the redox process. In addition, we further performed geometric analysis by tracking M-O bond lengths and interactions during the coloration process. Zn-CAT-1 and Ni-CAT-1 show minimal length changes (<0.02 Å), whereas Cu-CAT-1 exhibits a more pronounced variation (up to 0.05 Å) (Fig. 5d and Table S6), aligning with our EXAFS results. Metal-oxygen interactions were further evaluated using the crystal orbital Hamilton population (COHP) analysis. Pronounced bonding states below the Fermi level stabilize the MOF framework, and integrated COHP (ICOHP) values quantify bond strength (Fig. S54). The ∆ICOHP values of Cu-O, Ni-O, and Zn-O bonds between reduced and oxidized states are 0.16, 0.03, and 0.02, respectively, indicating that Cu-O bonds show the largest change and highest resistance to distortion, thus requiring higher energy input for C=O bond formation. Critically, during the purple-to-green transition, both lattice parameters a and b show elongation. Consequently, Zn- and Ni-CAT-1 expanded by only 1% and 2%, respectively, while Cu-CAT-1 undergoes a size increase up to 4%, indicating greater structural distortion than in its Zn- and Ni-based counterparts (Fig. 5e and Table S7). The Cu2+/Cu+ redox process triggers significant charge redistribution and coordination reorganization, undermining transition stability compared to the redox-inactive Zn2+ and Ni2+ centers. To further investigate the kinetics of M-CAT, the migration pathways and energy barriers were analyzed, considering both surface diffusion and vertical migration. Site A is identified as the most favorable adsorption site, either having the lowest energy in Ni-CAT-1 or an energy comparable to the most stable configuration in Cu-CAT-1 and Zn-CAT-1 (Fig. S55). Therefore, Li⁺ diffusion is expected to initiate at site A and proceed along the surface through the A → B → C pathway. At site C, the ion changes direction and migrates vertically into the 2D framework. Barrier analysis shows that the surface migration step from A → B is facile for all three MOFs, with a very small energy barrier of approximately 0.1 eV. In contrast, the barrier for migration from A → C is calculated to be 0.46, 0.25, and 0.32 eV for Cu-CAT-1, Ni- CAT-1, and Zn- CAT-1, respectively. Once Li+ reaches site C, the directional change toward the interior (C → D) occurs with a negligible barrier (<0.1 eV). The subsequent vertical migration step (D → E) exhibits barriers of 0.28, 0.36, and 0.15 eV for Cu-CAT-1, Ni-CAT-1, and Zn-CAT-1, respectively (Table S8). Overall, the calculated rate-limiting barriers are 0.46, 0.36, and 0.32 eV for Cu-CAT-1, Ni-CAT-1, and Zn-CAT-1, respectively (Fig. S56). These values agree well with experimentally observed switching speeds, with Zn-CAT-1 performing the fastest, followed by Ni- and Cu-CAT-1. Overall, color modulation involves ligand redox transitions and metal-ligand interactions, highlighting the critical roles of the metal center and the ligand (Fig. 5f). Metals with low electron affinity are preferred to ensure stability and efficiency. Metal-valence-state transitions involve additional electron participation and are associated with higher energetic costs. The resulting electron redistribution modulates local geometric configurations, ultimately inducing framework-level structural distortion. Furthermore, the identity of the metal center fundamentally defines the local coordination environment and thereby changes its electron-transfer pathways. Lastly, π-d orbital interactions between the metal and oxygen atoms modulate the strength of the metal-oxygen bond. A significant difference in metal-oxygen bonding during redox increases the difficulty of converting ionic to coordination bonds, thereby hindering C=O bond formation and making redox more kinetically demanding. Discussions The present study emphasizes that the central metal plays a crucial role as the ligand and provides the critical impact of the metal-center electron affinity on the electrochromic behavior of 2D c-MOFs. Our findings differ from conventional empiricism that favors redox-active metal centers. Comparative studies reveal that low-electron-affinity metals, such as Zn2⁺ and Ni2+ facilitate stable coloration, while the Cu2+/Cu+ redox pair incurs additional energetic cost and strengthens metal-oxygen interactions, ultimately leading to structural degradation. DFT calculations further elucidate differences in charge transfer, geometry, and energy-level analyses among the three systems. Although this study focuses on CAT-based 2D MOFs as a representative model system, the mechanistic insights obtained here are not necessarily specific to the HHTP linker. Instead, we suggest a more general physical picture for redox-active 2D MOFs, in which electrochromic performance generally involves coupled redox processes of the organic ligands and metal nodes. Metal-dependent switching kinetics and stability are expected whenever metal substitution occurs within an isostructural 2D MOF framework, regardless of the specific redox-active ligand employed. Based on this understanding, the metal-centered screening strategy described here may be transferable to other 2D MOFs with redox-active ligands and conserved architectures, subject to further experimental validation. Supplementary information Supplementary Information Transparent Peer Review File Supplementary Information Transparent Peer Review File Source data Source Data Source Data