Charge storage mechanisms of a π–d conjugated polymer for advanced alkali-ion battery anodes

The demand for fast-charging metal-ion batteries underlines the importance of anodes that work at high currents with no risk of dendrite formation. NiBTA, a one-dimensional Ni-based polymer derived from benzenetetramine (BTA), is a recently proposed promising material for safe fast-charging batteries. However, its charge–discharge mechanisms remained unclear and controversial. Here we solve the controversies by providing the first rigorous study using a combination of advanced theoretical and experimental techniques, including operando and ex situ X-ray diffraction, operando Raman spectroscopy and ex situ X-ray absorption near-edge spectroscopy (XANES). In safe potential ranges (0.5–2.0 V vs. M+/M, M = Li, Na or K), NiBTA offers high capacities, fast charge–discharge kinetics, high cycling stability and compatibility with various cations (Li+, Na+, K+). In the Na- and K-based cells, fast bulk faradaic processes are manifested for partially reduced states. Atomistic simulations explain the fast kinetics by facile rotations and displacements of the macromolecules in the crystal, opening channels for fast ion insertion. The material undergoes distinct crystal structure rearrangements in the Li-, Na- and K-based systems, which explains different electrochemical features. At the molecular level, the charge storage mechanism involves reversible two-electron reduction of the repeating units accompanied by a change of the absorption bandgap. The reversible reduction involves filling of the orbitals localized at the ligand moieties. No reduction of NiBTA beyond two electrons per repeating unit is observed at potentials down to 0 V vs. M+/M.


List of
. Tight-binding Hamiltonian (in eV) and overlap matrices for the nearest and next nearest neighbor LMOs shown in Figure S2 Table S4. Natural atomic orbital (NAO) populations for pristine and doubly charged oligomer shown in Figure S12, compared also with the BTA molecule shown in Figure S6 Figure S31 Table S13. Bader atomic charges calculated for crystalline NiBTA (cryst-hc, Figure S23) and Li2NiBTA (crystLiLi-h, Figure S25) Cell assembling. All procedures were performed in an argon-filled glovebox with O2 and H2O levels below 1 and 0.1 ppm, respectively. For studying the electrochemical properties, coin cells (CR2032) were assembled. Alkali metals (Li, Na or K) were used as counter-electrodes, glass fiber was used for the separators. For the Li-based cells, the electrolytes were 1M LiPF6 in EC:DMC (1:1 v/v) or 1M LiTFSI in DME:DOL (1:1 v/v). For the Na-and K-based cells, the electrolytes were 1.5M NaPF6 in DME and 1.5M KPF6 in DME, respectively. Each cell contained 70-80 µL of the electrolyte. For the ex situ XRD and XANES measurements, disassemblable cells (ECC-Ref from EL-CELL) were assembled in a two-electrode configuration. Li, Na or K metals were used as the counter electrodes. Glass fiber separators were used, and additional layers of polyolefinbased separators were placed on top of the working electrodes to avoid sticking of the glass fiber to the electrodes. The electrolytes for the Li-based, Na-based and K-based cells were 1M LiPF6 in EC:DMC (1:1 v/v), 1.5M NaPF6 in DME and 1.5M KPF6 in DME, respectively. For the operando XRD measurements, a Swagelok-type cell with a Be window [Leriche2010] was assembled. Li, Na or K metals were used as the counter electrodes. Three layers of the glass fiber separators were placed between the electrodes to avoid shortcircuiting of the cell because of the alkali metal dendrite growth. The electrolytes for the Li-based, Na-based and K-based cells were 1M LiPF6 in EC:DMC (1:1 v/v), 1.5M NaPF6 in diglyme and 1.5M KPF6 in diglyme, respectively. The cell contained 240 µL of the electrolyte. For the operando Raman spectroscopy measurements, a cell with an optically transparent window (ECC-Opto-Std from EL-CELL) was assembled in a two-electrode configuration. The electrodes were placed in direct contact with the optically transparent window (sapphire of mineral glass), with the active material facing the window. Li, Na or K metals were used as the counter electrodes. Three layers of glass fiber separators were used. The electrolytes for the Li-based, Na-based and K-based cells were 1M LiPF6 in EC:DMC (1:1 v/v), 1.5M NaPF6 in diglyme and 1.5M KPF6 in diglyme, respectively. Galvanostatic cycling and capacity calculations. The experiments were performed with Neware BTS-4000 stations at room temperature. The current densities were set to 0.1, 0.2, 0.5, 1, 2 or 5 A g −1 . The potential ranges were 0.5-2.0 V or 0.01-2.0 V vs. M + /M (M = Li, Na or K). Prior to the prolonged cycling at 2 A g −1 , the cells were subjected to five cycles at 0.1 A g −1 to avoid low initial capacities associated with the activation effects [Kapaev2019]. The current densities and capacities were calculated basing on the mass of NiBTA unless stated otherwise.
Calculating capacity contributions from carbon black. Specific capacities of Super P carbon black are shown in Figure S14, Figure S15, Figure S16 and Figure S17. Contributions from carbon to the capacities of the NiBTA-based electrodes were calculated as follows: where Qcorr is the capacity after subtracting the contribution from Super P, QNiBTA is the capacity per NiBTA mass, QSP is the capacity of Super P in the control experiments measured at the same current density in the same potential range, ωSP and ωNiBTA are mass fractions of Super P and NiBTA in the electrode, respectively. Cyclic voltammetry. The experiments were performed with BioLogic VMP3 at room temperature. The potential ranges were 0.5-2.0 V vs. M + /M (M = Li, Na or K). Prior to the main experiments, ten scans at 1 mV s −1 were carried out to eliminate irreversible processes. The current densities were calculated basing on the mass of NiBTA. Sample preparation for the ex situ XRD measurements. The electrodes were discharged at room temperature to 0.5 V or 0.01 V vs. M + /M (M = Li, Na or K) at a constant current density of 30 mA g −1 . The cells were disassembled in an argon-filled glovebox, the electrodes were then washed with ~4 mL of DME and dried in the glovebox environment at room temperature. The electrodes were placed onto a Mylar film in a sample holder for the Huber Guinier Camera 670, and the holder was sealed with a Kapton tape to prevent decomposition in air during the measurements. The sample was then taken out from the glovebox. Ex situ XRD measurements. The XRD patterns were measured with a Huber Guinier Camera 670 that operates with CoKα1 radiation (λ = 1.78892 Å). The measurements were carried out directly after the Kapton-sealed electrodes were removed from the glovebox. Single scans were measured in the 2θ range of 4-100° each ten minutes to monitor possible changes caused by oxidation. The resolution was 0.005°. The measurements continued overnight. The scans that were measured before changes in the XRD patterns started to occur were then averaged.
Operando XRD measurements. The cell for the XRD measurements was placed inside a Bruker D8 ADVANCE powder X-ray diffractometer that operates with CuKα radiation. The cell was connected to BioLogic SP-150. The XRD patterns were collected at room temperature during galvanostatic discharging to 0.5 V vs. M + /M (M = Li, Na or K) and subsequent charging to 2.0 V M + /M. The current density was 30 mA g −1 basing on the mass of NiBTA. Single scans were measured each 10 minutes, the resolution was 0.04°, the 2θ range was 12-32° for the Li-based cell and 14-38° for the Na-and K-based cells. The XRD patterns were baseline-corrected with Bruker DIFFRAC.EVA software. Sample preparation for the ex situ XANES measurements. The electrodes were discharged at room temperature to 0.5 V or 0.01 V vs. M + /M (M = Li, Na or K) at a constant current density of 30 mA g −1 . The cells were disassembled in an argon-filled glovebox and sealed between two layers of Kapton tape. Pristine electrode was sealed between Kapton layers as well. Ex situ XANES measurements. The XANES spectra were obtained with a laboratory spectrometer of the Department of Radiochemistry, Moscow State University. An X-ray tube with a silver anode with a power of 1.5 kW was used as an X-ray source. The X-ray tube, monochromator crystal, and silicon drift detector (Amptek) were placed in a Rowland circle, 0.5 m in diameter. The radiation energy was chosen using the reflection [444] of a spherically bent silicon crystal (with a bend radius of 0.5 m) oriented at a Bragg angle of 71.6°. To avoid damage to the samples by the white beam, the samples were placed in front of the detector where they were exposed only to the monochromatic beam. The beam size was 5 mm by 5 mm. The data were collected in the transmission mode, at each energy point of the XANES region the signal was accumulated for 5 s. For all samples, 5 spectra were collected and combined using the IFEFFIT software. Operando Raman spectroscopy measurements. The cell for Raman spectroscopy measurements was placed in a Thermo Scientific DXRxi Raman Imaging microscope and connected to BioLogic SP-150. The spectra were collected at room temperature during CV measurements in the potential ranges of 0.5-2.0 V or 0.01-2.0 V vs. M + /M (M = Li, Na or K). The potential scan rate was 0.047 mV s −1 . Prior to the CV measurements, linear sweep voltammetry from open-circuit voltage to 2.0 V vs. M + /M was performed with the scan rate of 1 mV s −1 . The laser excitation wavelength λ was 532 nm or 780 nm, the laser power was 1 mW. Single scans were measured every 7 minutes (~20 mV). The range of Raman shifts was 50-2,200 cm −1 (for λ = 780 nm) or 50-3,400 cm −1 (for λ = 532 nm). The spectra were baseline-corrected with Bruker OPUS software. Sample preparation for UV-Vis-NIR spectroscopy measurements. To prepare the NiBTA-based films, 36 mg of NiBTA and 4 mg of sodium carboxymethylcellulose were dispersed in 4 mL of deionized H2O via ultrasonication. The suspension was deposited with a spin-coater onto glass substrates (~2x2 cm pieces, 600 µL per piece). The rotation rate of the spin-coater was 2,000 rpm. The alkaliation solution was prepared in an argon-filled glovebox. Naphthalene (0.8 mmol, 103 mg) was dissolved in diglyme (8 mL). Metallic potassium (~50-100 mg) was introduced to the solution. The mixture was stirred at room temperature for 2 h, resulting in a dark-green solution of potassium naphthalenide. Excess of potassium was removed afterwards. The NiBTA-based film was introduced to the prealkaliation solution and kept at room temperature for 2 days in an argon-filled glovebox. The film was removed from the solution, washed with ~4 mL of DME and dried in an argonfilled glovebox at room temperature.
UV-Vis-NIR spectroscopy measurements. The spectra were collected in transmittance mode using an AvaSpec-2048-2 fiber-optic spectrometer that was placed in a N2-filled glovebox. The samples were transferred directly from an Ar-filled glovebox in a sealed vial. Pellet preparation and conductivity measurements. Cylindrical pellets of NiBTA (d = 10 mm) were prepared by cold pressing of the powder (146 mg) with commercial pressing equipment (Carver). Applied load was 5 metric tons. Prior to the conductivity measurements, top and bottom of the pellet were coated with 30-50 nm of gold using a Quorum Q150T ES magnetron. Sides of the pellet were covered with an adhesive tape before the sputtering. The coated pellet was placed in a symmetrical cell between two copper disks. Direct-current polarization was applied using BioLogic VMP3. The voltage was changed in 50 mV steps between −1 and 1 V, the current was measured over 5 s at each step. Microscopy studies and elemental analysis. The microscopy and energy-dispersive X-ray spectroscopy measurements were performed with a Titan Themis Z transmission electron microscope at 200 kV. Samples were ground in a mortar in an organic solvent (ethanol for the pristine electrode, anhydrous dimethyl carbonate for the cycled electrodes) and dispersed onto Cu square mesh grid covered with holey carbon. The electrodes after cycling were prepared in an Ar-filled glovebox and transferred to the TEM chamber with a short air exposure of 1-2 seconds. Standard double tilt TEM holder was used. Imaging mode was high-angle annular dark field (HAADF) STEM, using HAADF detector at an effective camera length of 115 mm. Super-X EDX detector was used for the EDX mapping. Thermo Fisher Scientific Velox software was used both for visualization and analysis of EDX data. Computations. For calculations of oligomers and standalone polymer, we used CAM-B3LYPp2p/a3p method as implemented in Gaussian 16 package. The CAM-B3LYP functional [Yanai2004] was chosen because it has the lowest mismatch between orbital and total energies, see Table S14, and proven reliability for π-conjugated molecules [Tukachev2019, Zhugayevych2018]. Comparison with other functionals was performed for critical calculations, at least with PBE0 which is more reliable beyond the class of πconjugated semiconductors [Vasilchenko2021]. The basis 6-31G*, referred here as 'p2p', commonly used for light elements is unreliable for heavier ones. For them we used a more robust Def2-TZVP basis abbreviated as 'a3p'. Because scalability is critical, we used 'p2p' basis for light elements up to Ar and 'a3p' for heavier ones, denoting this combination as 'p2p/a3p'. It should be noted that the standard 'p2p' basis uses 6 d-orbitals, whereas in the 'p2p/a3p' combination every l-shell contains 2l + 1 orbitals. For calculations of crystals, we used PBE-D3/PAW600 method as implemented in VASP 5.4 package, here 600 means 600 eV plane wave energy cutoff. Despite PBE functional gives typically inaccurate electronic structure of π-conjugated systems, the combination PBE-D3 is robust for prediction of intermolecular packing [Halaby2021, Zhugayevych2018]. Use of hybrid functionals is limited due to poor convergence of wavefunction for all considered basis sets and programs. PBE/PAW600 was also used for calculation of differential charge density, Bader charges [Tang2009] and band structure [Ong2013] of crystal structures. Localization of molecular orbitals is performed by projection onto basis of selected block of atoms using "MolMod:-LocalizeMO" program which code is available at http://zhugayevych.me/maple/MolMod/MolMod.ini, see Figure S11 for details.
In all figures of the current work, the wave-function isovalue is typically 0.05 for LMO and 0.02 for MO. In case of comparison, all figures are plotted with the same isovalue. Powder diffraction patterns for predicted structures were simulated in VESTA program. Computational spectroscopy was performed for oligomers. While intramolecular vibrational frequencies depend on the local atomic environment, allowing reliable modeling with oligomers, absorption spectra and Raman intensities of "soft" materials require statistical sampling over large ensembles. Therefore, considered oligomer models provide only interpretation of the experimental spectra. For excited states, we used TDDFT. Benchmarking of the computations was performed using Ni-OPD2 as a model compound. The benchmarking results are available in Section S20, p. 62. Geometric definitions used for the calculations are provided in Section S21, p. 67. Figure S1. Geometry of isolated NiBTA macromolecules in pristine and lithiated/sodiated/potassiated states from top and side views. Crystal structures are shown in Figure  S23 and Figure S25. S12 S3 Electronic structure of standalone NiBTA macromolecules Figure S2. Schematic electronic structure of the pristine NiBTA macromolecule. See also Figure 1 in the main text. The energy is in eV. The π-conjugated system is represented by DOS and by monomer LMOs: Ni (left) and BTA (right). The symmetric LMOs (black colored) belong to B1u (z) or B2g (xz) representations of mmm group, whereas antisymmetric LMOs (blue colored) belong to B3g (yz) or Au (xyz) representations. The two Ni LPs are shown in green. Their density is excluded from DOS(σ). The "working" LMO is shown in red: it is empty for the pristine polymer, but occupied upon 2e-reduction, and also it is occupied in the BTA molecule. Table S1. Tight-binding Hamiltonian (in eV) and overlap matrices for the nearest and next nearest neighbor LMOs shown in Figure S2. The matrix elements are indexed by the first row and column containing energies of LMOs. For example, the "working" LMOs (0.112 eV) are coupled "through space" by transfer integral -0.167 eV and through Ni Z orbital (6.483 eV) by -1.663 eV. Figure S3. Electronic bands of pristine NiBTA macromolecule. Band structure for all electrons is shown on the left, the bands only for valence π-orbitals and Ni LPs are shown on the right. Color codes correspond to Figure S2. Band structures of crystalline polymers ( Figure S33) reveal that pronounced dispersion of energy bands occurs only along the polymer chains, so considering isolated macromolecules is reliable for assessing the electronic structure of NiBTA. It is seen that width of frontier bands is smaller than ~1 eV, many of them are almost flat, and π-bands are not entangled in the sense that the wave-function pattern weakly depends on the k-vector. This allows to clearly see the constituting LMOs even visually (see Figure S4). Figure S4. Wave-function of the lowest unoccupied band of pristine NiBTA ("working" band) at Γand X-points. Table S2. List of frontier electronic bands of the standalone pristine macromolecule including all πconjugated LMOs and LPs. The double horizontal line separates occupied and unoccupied bands. Bands are numbered from the bottom of the valence band according to their energy at Γ-point. The next three columns include energies in eV: the band energy at Γ-and X-points, EΓ and EX (X means end of the first Brillouin zone), and the LMO energy, ELMO. In 'type' column the antibonding asterisk is omitted for clarity, and prime labels blue-colored orbitals in Figure S2 and Figure S3. LMOs are numbered starting from 1 within their groups: π-BTA, π-Ni, LP-Ni. The π-system includes LMOs of 4 symmetries: Au (xyz), B1u (z), B2g (xz), B3g (yz); the other 4 symmetries describe σ-system: Ag (1), B1g (xy), B2u (y), B3u (x). Symmetry notations for σ-bands correspond to the wave-function symmetry at Γ-point around Ni atom (the same symmetry is at X-point).

S2 Geometry of isolated macromolecules
# Ni Z 42 6.52 7.00 6.77 π-BTA' 10 B3g Figure S5. Frontier localized molecular orbitals of the BTA monomer of the pristine polymer. The LMO7 is empty for pristine NiBTA and occupied for charged NiBTA as well as for BTA molecule (each nitrogen has two hydrogens).    S19 Figure S10. Electronic band structure of the standalone pristine macromolecule calculated using various methods.  Figure S11. Comparison of LMOs obtained by variation of localization procedure. We can either localize occupied and unoccupied MOs separately or localize all MOs at once; also, we can either maximize projection onto the selected subspace ("project-in") or minimize the projection out of this subspace. The variant (d) produces the most localized LMOs, whereas the variant (a) results in the most compact tight-binding Hamiltonian. In the latter case, the LMO energies are close to energies of those MOs to which they contribute mostly. For this reason, we use the variant (a) as the default in the present work. Figure S12. Orientation and atom indexing of the NiBTA low-molecular model used for population analysis. The analysis for a longer oligomer (shown in Figure S13, data shown in Table S7 and Table S8) reveals the same features as for the short structure (Table S4 and Table S5), indicating that the latter is a suitable model for analyzing qualitative trends. Table S4. Natural atomic orbital (NAO) populations for pristine and doubly charged oligomer shown in Figure S12, compared also with the BTA molecule shown in Figure S6. Here q is natural atomic charge and other columns show population in each electronic subsystem. If we assume that covalently bound C atom is in electronic configuration C2s2p 3 , triply coordinated N atom in N2s2p 4 (a strong σ-acceptor but also a strong π-donor), O atom in O2s 2 p 4 , and Ni in its present coordination is in Ni3d 10 configuration (a weak donor in both channels), then the normal population of the πsystem is 36. The missing 4 electrons are withdrawn from N and O atoms at about 0.5 electron per atom. At the same time, these atoms have excessive population in σ-system resulting in their negative charge. This π-to-σ transfer does not occur in the BTA molecule, therefore the role of transition metal is critical for the process. Indeed, the Ni-N bonds are formed from Ni sp 2 d hybrids, but the population of Ni 4sp orbitals is very low implying extremely polar bonds. However, the total charge on Ni is not very high (+0.5 to +0.7) meaning a strong back-donation via 3dxz and 3dyz orbitals (π-back-bonding), whose population is close to maximum. Finally, the e2-orbitals (Z2 and X2-Y2) remain intact (negligible depletion) as lone pairs deep in the valence band and should be electronically inert.   Table S6. Natural atomic charges for oligomers of various length, termination, and oxidation state.

S4 Population analysis
Here the terminations are according to Figure S52 with the number at the end of the oligomer name indicating the number of Ni atoms (molecOBe is the same as molecOLi but with Be instead of Li  Figure S13. Orientation and atom indexing of the molecO5 used for population analysis. As seen from Table S6, charge distributions at central Ni and BTA blocks are the same for longer oligomers and other chain terminations indicating that length of the selected oligomer is sufficient for adequate modeling of NiBTA. Table S7. Natural atomic charges and orbital (NAO) populations for pristine and doubly fully reduced NiBTA oligomers shown in Figure S13.  Table S8. Natural atomic orbitals for pristine and reduced NiBTA models shown in Figure S13.  Figure S14. Electrochemistry of Super P in lithium-based cells in the 0.5-2.0 V vs. Li + /Li potential range.Charge-discharge curves with the ether-based (a) and carbonate-based (b) electrolytes at 0.1, 0.2, 0.5, 1, 2 and 5 A g −1 ; dependencies of charge and discharge capacities on the cycle number at varying current densities for the cells with ether-based (c) and carbonate-based (d) electrolytes.     Table S9. Types of NiBTA structures considered for simulations. Here Z is number of monomers per unit cell. The "shift" is a mutual shift of two polymers along each other; they are given in fraction of the monomer length for the nearest polymers along z-and y-directions respectively ('1/4' shifts are approximate). Subscript x in symmetry group means (a, b, c) → (c, a, b) transformation from the default International Tables for Crystallography settings. name symmetry Z shifts description Herringbone type crystals, see Figure S22a cryst-h P-1 2 0 any low-symmetry herringbone ('h') cryst-ha Pnnmx 2 0 1/2 cryst-hb Pbamx 2 0 0 cryst-hc P21/n 2 0 1/4 cryst-ho -4 1/2 1/4 'hc' doubled in z and kept orthorhombic Channeled crystals, see Figure S22b cryst-s P1121/n 4 ±1/4 1/2 kept orthorhombic, square pattern ('s') π-Stack type crystals, see Figure S22c, d cryst-wa Immm 2 0 1/2 brickwork pattern ('w') cryst-wb Ammm 2 0 0 cryst-pa Fmmm 4 1/2 1/2 in-plane orderding ('p') cryst-pb Bmmm 2 1/2 0 Isolated π-stacks stack-w cmmm 2 1/2 brickwork pattern ('w') stack-wx c2/m11 2 1/2 tilted polymers stack-p pmmm 1 0 stack-px p-1 1 any Isolated macromolecules polym pmmm 1 polymKK p2/m11x 1 alkali ions distort the planarity

S36
(a) herringbone (cryst-hc) (b) channeled (crystKK-s) (c) brickwork (cryst-wa) (d) in-plane (cryst-pa) Figure S22. Different types of simulated crystal structures. Table S10. Structures predicted by PBE-D3/PAW600 method. Here D is number of translation vectors, Z is number of monomers per unit cell, a is the monomer length and ∆a is its difference with respect to the isolated pristine macromolecule (7.674 Å), b and c are lengths of translation vectors perpendicular to the polymer with c being the π-stacking direction, cx and cz are alongpolymer and perpendicular projections of the vector c, ϕ is the polymer setting angle relative to ab plane, V is the volume per atom not counting alkaline atoms, E is the total energy per monomer relative to the lowest energy polymorph. The stability is tested by 10 ps MD, listed is the number of polymers in the supercell ('x2' means double cell in the polymer direction) followed by the polymorph to which the original structure transforms. The last column shows the consistency of the predicted stable structures with experimental XRD data: '+' means consistent, '-' means not expected to be observed or is not observed, '?' indicates some inconsistencies. Multiple '+' for herringbone structures are due to the sliding of polymers along each other at room temperature.            Figure S31 at positions 1-7. The most favorable configuration is with two K atoms attached at positions 3 and 5 at the opposite faces of the oligomer. This configuration is used to set the chemical potential of K atoms when calculating energies of other configurations. In particular, from energy of configuration '3' we see that a separated pair of K atoms is 2 × 0.14 eV higher in energy then a pair localized on the same monomer.
S48 S10 Electronic structure of simulated crystalline polymers Figure S32. Electronic density of states of simulated structures.   S53 S12 Conductivity measurements Figure S37. Direct-current polarization data for a cylindrical pellet of NiBTA. Height and diameter of the pellet are given.
S54 S13 Raman spectra of NiBTA after chemical reduction Figure S38. Raman spectra of NiBTA-based films before and after chemical reduction measured with the green (left) and near-infrared (right) lasers.
S55 S14 Simulations of UV-Vis-NIR spectra Figure S39. Typical spectra of excitations. The spectra were modeled by oligomer of 4 BTA blocks terminated as in Figure S52i ("reduced" state is modeled by negatively charged molecules). The lowest bright transition has B1u symmetry and corresponds to LMO5→LMO7 excitation. For the reduced oligomers the lowest bright transition is blue-shifted and corresponds to LMO7→LMO8 excitation. Figure S40. Calculated bright excitation energy for oligomers of various length. The oligomers are terminated as in Figure S52i. Experimental data are from [Audi2014]. Calculated energies are shifted by a constant to match the experimental value at a single point (pristine oligomer with 2 BTA blocks). The extrapolated energies match experimental UV-Vis peak positions for both pristine and reduced NiBTA. For the pristine polymer, the left flank of the observed absorption "band" (500-900 nm) can be attributed to oligomers since the size-dependent energy of their bright states (1.3-2.4 eV) perfectly fits this range, whereas the right flank of the absorption (>1000 nm) is formed presumably by a multitude of dipole-forbidden transitions and excitations originating from non-passivated polymer terminations. In the reduced state, the π-conjugated system is half-filled, leading to the bandgap increase. At shorter wavelengths, the intensive absorption should originate from higher excitations and from oligomers of various length. The observed decrease of the longwavelength absorption tail is probably associated with passivation of electronic traps (e.g., radicallike terminations of the polymer) upon NiBTA reduction, which produces intense NIR absorption of the pristine material. Figure S41. Calculated Raman activities for oligomers. The oligomers are terminated as in Figure  S52i. The reduced state is simulated by anion in a highly polarizable dielectric medium, where the charge is two times the number of BTA blocks. For the pristine oligomer, the mode at 660 cm −1 represents Ni-N bond stretching whereas high frequency modes at 1560 and 1660 cm −1 are two different BLA modes. For the reduced polymer, high frequency modes at 1450 and 1660 cm −1 are again BLA modes. However, at 790 cm −1 a new mode appears -BTA-breathing mode, whereas the Ni-N stretching mode at 640 cm −1 is less intense. There is also small peak at 570 cm −1 involving angle bending in BTA and NiN4 fragments. Table S13. Bader atomic charges calculated for crystalline NiBTA (cryst-hc, Figure S23) and Li2NiBTA (crystLiLi-h, Figure S25).The nearest neighbor elements are given in parentheses. For C(-N), two non-equivalent positions are provided. It is seen that Li insertion into NiBTA causes noticeable reduction of ligands, which is further confirmed by the differential charge density revealing accumulation of electron density mainly near carbon and nitrogen atoms ( Figure S42).

S16 Charge distributions in crystal structures
Li2NiBTA +0.0 +0.3 +0.8 +0.5 +0.1 −0.1 −1.1 +0.5 Figure S42. Differential charge density dn = n(Li2NiBTA) -n(NiBTA) drawn for isosurface levels of 0.0085 (a) and 0.006 (b). Calculations were done for crystalline structures (cryst-hc for NiBTA and crystLiLi-h for Li2NiBTA, see Figure S23 and Figure S25). Accumulation and depletion of electron density are shown as yellow and cyan isosurfaces, respectively. During lithiation of NiBTA, electron density to is donated primarily to pz states of nitrogen and carbon. Slight accumulation of electron density at π-hybridised dxz and dyz states of Ni is counteracted by density withdrawn from its dxy and dz2 orbitals. For clarity, only central atoms are shown inside the unit cells.
S59 S17 Electrochemistry in the 0.01-2.0 V vs. M + /M potential ranges Figure S43. Electrochemistry of NiBTA in Na-and K-based cells in the 0.01-2.0 V vs. M + /M potential ranges.Charge-discharge curves of NiBTA in sodium-(a) and potassium-based (b) cells at 0.1 A g −1 for initial cycles. The capacities are given per NiBTA mass. Estimated contributions from Super P are 30 and 45 mA h g −1 for the Na-and K-based cells, respectively (see Figure S17). Figure S44. Electrochemistry of NiBTA in Li-based cells in the 0.01-2.0 V vs. Li + /Li potential range.Charge-discharge curves of NiBTA in lithium-based cells with the ether-based (a) and carbonate-based (b) electrolytes at 0.1 A g −1 for initial five cycles. The capacities are given per NiBTA mass. Estimated contributions from Super P are 50 and 55 mA h g −1 for the carbonate-and ether-based electrolytes, respectively (see Figure S16).
S60 S18 Raman spectroscopy in the 0.01-2.0 V vs. M + /M potential ranges  S61 S19 TEM EDX data before and after deep reduction Figure S47. EDX data of NiBTA before/after deep cycling in Li-based cells.EDX spectra for selected sample areas of pristine electrode (a), electrode after lithiation to 0.01 V vs. Li + /Li (reduced) (a), and electrode after lithiation to 0.01 V and delithiation to 2.0 V vs. Li + /Li (reoxidized) (c); estimated mass fractions of elements (d).
HOMO, B1u Z LUMO, B2g XZ Figure S49. Frontier molecular orbitals of Ni-OPD2.    Figure S51. Orientation and atom indexing of NiBTA used for the calculations. Figure S52. Considered terminations of NiBTA oligomers. The electronic levels are given for LMOs localized on the terminal group (left), Ni (center) and BTA monomer (right). Evidently, all considered terminations do not distort LMOs of the BTA substantially. Natural terminations (b, c) induce the smallest distortion but have low symmetry which is inconvenient for calculations. Replacement of nonsymmetric hydrogen by an alkaline atom (d, e, f) is consistent with the chemistry of the studied system, but the valence p-orbitals of the metal stands next to the LUMO