Aging and Charge Compensation Effects of the Rechargeable Aqueous Zinc/Copper Hexacyanoferrate Battery Elucidated Using In Situ X-ray Techniques

The zinc/copper hexacyanoferrate (Zn/CuHCF) cell has gained attention as an aqueous rechargeable zinc-ion battery (ZIB) owing to its open framework, excellent rate capability, and high safety. However, both the Zn anode and the CuHCF cathode show unavoidable signs of aging during cycling, though the underlying mechanisms have remained somewhat ambiguous. Here, we present an in-depth study of the CuHCF cathode by employing various X-ray spectroscopic techniques. This allows us to distinguish between structure-related aging effects and charge compensation processes associated with electroactive metal centers upon Zn2+ ion insertion/deinsertion. By combining high-angle annular dark-field-scanning electron transmission microscopy, X-ray absorption spectroscopy (XAS), X-ray photoelectron spectroscopy, and elemental analysis, we reconstruct the picture of both the bulk and the surface. First, we identify a set of previously debated X-ray diffraction peaks appearing at early stages of cycling (below 200 cycles) in CuHCF. Our data suggest that these peaks are unrelated to hypothetical ZnxCu1–xHCF phases or to oxidic phases, but are caused by partial intercalation of ZnSO4 into graphitic carbon. We further conclude that Cu is the unstable species during aging, whose dissolution is significant at the surface of the CuHCF particles. This triggers Zn2+ ions to enter newly formed Cu vacancies, in addition to native Fe vacancies already present in the bulk, which causes a reduction of nearby metal sites. This is distinct from the charge compensation process where both the Cu2+/Cu+ and Fe3+/Fe2+ redox couples participate throughout the bulk. By tracking the K-edge fluorescence using operando XAS coupled with cyclic voltammetry, we successfully link the aging effect to the activation of the Fe3+/Fe2+ redox couple as a consequence of Cu dissolution. This explains the progressive increase in the voltage of the charge/discharge plateaus upon repeated cycling. We also find that SO42– anions reversibly insert into CuHCF during charge. Our work clarifies several intriguing structural and redox-mediated aging mechanisms in the CuHCF cathode and pinpoints parameters that correlate with the performance, which will hold importance for the development of future Prussian blue analogue-type cathodes for aqueous rechargeable ZIBs.

the cell is 10 Ω. If not stated otherwise, the cells were opened under ambient conditions before analyses, and the anode and cathode were washed thoroughly by dipping them into fresh ultrapure water ca. 5 times for a few seconds, and then immediately dried with Ar gas. A few cells were also analyzed without this washing step, which is then explicitly stated in the text.

Physical characterization
The water content and the molecular weight of the CuHCF material were determined using a TGA-Q500 thermogravimetric analyzer (TA Instruments). An amount of ca. 5-6 mg of the pristine CuHCF powder (without binder or carbon) was placed in aluminum pans. The samples were equilibrated at 30 °C prior to ramping the temperature to 500 °C at a heating rate of 10 °C min -1 under N2 flow.
Grazing incidence X-ray diffraction (XRD) of the CuHCF cathodes was performed using a Siemens D5000 diffractometer in parallel beam mode with Soller slits (0.40°), and 1 mm in and outgoing slit. The diffractometer was operated at 45 kV and 40 mA. The diffractograms were recorded in an angular range between 5-70 ° (2) with a step size of 0.035°, and 28 s per step. The samples were placed onto dedicated monocrystalline Si plates with zero-background.
Scanning electron microscopy (SEM) was carried out using a LEO 1550 microscope (ZEISS), equipped with an energy dispersive X-ray spectrometer (EDS) from Oxford Instruments. Images were recorded with a typical acceleration voltage of 3 -5 kV, while the EDS spectra were conducted with a voltage of 15 kV. Post processing of SEM-EDS elemental maps was performed using the AZtec software (Oxford Instruments).
Transmission electron microscopy (TEM) was carried out using a Thermo Fisher Scientific Themis Z microscope at the Electron Microscopy Centre (EMC) at the department of Materials and Environmental Chemistry (MMK), Arrhenius Laboratory, Stockholm University. The samples were scratched off from the substrate with a diamond scriber onto a TEM grid (SPI Supplies 200 mesh Cu grid with holey carbon supporting films). The microscope has both probe and image spherical aberration (Cs) correctors and Super-X Si drift detectors for EDS analysis that cover all angles of the sample. High-angle annular dark-field (HAADF) and bright-field (BF) scanning TEM (STEM) images were acquired simultaneously. The STEM and EDS analyses were performed at 60 kV. The probe current was 150 pA. The convergent angle of electron probe was 18.3 mrad. The collection angle of the BF and HAADF detector was up to 22 mrad and from 50 to 200 mrad, respectively. Post processing of the images and EDS elemental maps was performed in Velox (Thermo Fisher Scientific B. V.).

X-ray photoelectron spectroscopy
X-ray photoelectron spectroscopy (XPS) spectra were acquired with a PHI 5500 spectrometer (Physical Electronics) using monochromatic Al Kα radiation (1486.7 eV) and an electron emission angle of 45 °. XPS spectra were collected post mortem at the Cu 2p, Fe 2p, Zn 2p, C 1s, O 1s, N 1s, and S 1s edges, for pristine electrodes (not exposed to electrolyte), for electrodes cycled for a different number of cycles and stopped at OCP (for "aging" studies), and for electrodes cycled for 200 cycles and then stopped in either the charged (CHA) or the discharged (DCH) states (for studies of the charge compensation process). XPS spectra were also recorded for carbon "blank" cells, in which the active material (CuHCF) was left out. The spectra containing graphitic carbon were energy calibrated to the sp 2 -C peak of the carbon at 284.3 eV, while other spectra were calibrated to the adventitious carbon peak at 284.8 eV. 2,3 The XPS chamber pressure was kept at 1 x 10 -7 Torr or lower pressures during the sample analysis. The acquired spectra were analyzed using the CasaXPS software and considering typical a Gaussian/Lorentzian lineshape (30 % Lorentzian contribution) and employing a (Shirley-type) background. The sp 2 component in graphite was fitted using an asymmetric finite Lorentzian (LF) lineshape numerically convoluted with a Gaussian, LF(α, β, w, m), where α and β are parameters that modify the curvature of the Lorentzian, w is the width parameter of the smooth function, and m is the width parameter of the Gaussian. The parameters were minimized in order to achieve the best fit with the data, and were following; α =0.95, β = 1.6, w = 100, m = 80. The LF lineshape is equivalent to the Doniach-Sunjic lineshape, except that it has a finite background which is practical when using standard backgrounds such as Shirley type.

Elemental analysis
The elemental compositions and metal loadings were determined using inductively coupled plasmaoptical emission spectroscopy (ICP-OES), using an Avio 200 spectrometer (Perkin-Elmer). The sample preparation was carried out as follows; the CuHCF samples (ca. 10 mg powder or the circular disc electrode containing ca. 1.5 mg of active material) was first heated to 500 °C in a furnace (the temperature was ramped at 5 °C min -1 ) for 6 h to convert the copper hexacyanoferrate phase into an oxide phase to facilitate digestion with acid. Afterwards, the samples were digested in a mixture of 0.5 mL HNO3 (65%, EMSURE, Merck), 1 mL H2SO4 (ISO, 95-97%, Merck) and 1.5 mL HCl (37 %, ACS reagent Merck) into their metallic components. The samples were then diluted with ultrapure water up to a final volume of 200 mL. The emission wavelengths were selected as follows; Cu 327.393 nm, Fe 238.204 nm, Zn 206.200 nm, and K 766.490 nm. A multi-element standard solution containing 10 mg L -1 of the elements dissolved in 5% HNO3 (Perkin-Elmer) was used for calibration curve.

X-ray absorption spectroscopy (XAS)
X-ray absorption spectroscopy (XAS) was measured at the Cu, Fe, and Zn K-edges of the CuHCF cathode at the KMC-3 beamline at the BESSY II synchrotron facility, Helmholtz Zentrum Berlin (HZB), Germany. A scintillation detector coupled to a photomultiplier was employed to collect the fluorescence. The detector was covered with a foil one element below the investigated K-edge (Z-1) to reduce scattered light and to improve the signal. The Zn/CuHCF cell was investigated "ex situ" and "operando". The cathodes investigated ex situ comprised the electrode in its pristine state, and electrodes cycled for 0, 100, and 200 cycles in 1 M ZnSO4 in pouch cells, and then stopped at OCP, which corresponds to ca. ~1.7 V vs. Zn 2+ /Zn. The electrode parts were taken out immediately after stopping the cells, rinsed with ultrapure water, and immediately dried with Ar-gas. The operando measurements were performed in real-time conditions while cycling the Zn/CuHCF in a pouch cell. The pouch cell had been modified with an X-ray transparent Kapton window, and the fluorescence was monitored from the backside of the CuHCF cathode through the graphite current collector. Spectra were collected while holding the potential in either the discharged state (DCH, 1.00 V) or in the charged state (CHA, 2.15 V). Potentiodynamic measurements were also performed, where we followed the fluorescence during cyclic voltammetry. For these dynamic measurements, the fluorescence was recorded by connecting the output of the Keithley instrument to the analog input of the Bio-Logic SP-200 potentiostat, to ensure accurate translation of the fluorescence signal and the voltage in time. The fluorescence was monitored around the K-edge position, where largest changes were observed; 8989 eV for the Cu Kedge, 7126 eV for the Fe K-edge, and 9660 eV for the Zn K-edge. The local atomic structure was determined by simulations of the extended X-ray absorption fine structure (EXAFS) using scattering function generated in FEFF 9.1. 4 The simulation approach was adopted from earlier studies, and will be described briefly hetein. [5][6][7] Given that the Cu/Fe ratio is ~1.5 in the pristine CuHCF powder samples, there should be 1/3 rd of Fe(CN)6 vacancies in the native CuHCF structure. These vacancies were S5 confirmed to be occupied by zeolitic and/or coordinating water´s in neutron diffraction experiments by Wardecki et al. 5 In our EXAFS simulations on the Cu K-edge, we therefore represented these Fe(CN)6 vacancies with a Cu-O shell, where the O represent the water molecules. Multiple scattering was considered up to 4-leg path due to the strong focusing effect from the linear Fe-CN-Cu chains.

S6
Calculations of gravimetric capacity, Coulombic efficiency, round-trip efficiency, and electrons transferred in the electrochemical process The practical charge/discharge capacity (Qcha/dch) of the active material was determined from the galvanostatic charge/discharge voltage profiles according to Equation (S1): where Qcha/dch is the capacity during charge or discharge expressed in mAh g -1 , i is the applied current in mA, t is the time in hours, and m is the mass of the active material in grams determined by ICP-OES.
The theoretical capacity (Qtheoretical) is calculated according to Equation (S2): where n is the "electron transfer number" (i.e., moles of etransferred in the redox-process per moles of redox-active metal sites (in this case (Cu+Fe)), 8 F is the Faraday constant (96485.33 C mol -1 ), and Mw is the molecular weight of CuHCF determined from TGA analysis (267.07 g mol -1 ) assuming a molar formula of K:Cu:Fe:C:N of 0.05:1.00:0.66:4:4 and 3.3 water molecules per unit formula. 9 Since the capacity strongly depends on the value of n (i.e., the electrons transfer number), which varies for different cycling rates, we instead use the term "expected" capacity at different cycling rates, which is calculated according to Equations (S3)-(S4): where n is the electron transfer number (see Equation S5 below), Mw is the molecular weight of CuHCF determined by TGA analysis (267.07 g mol -1 ), and F is the Faraday constant. The difference in expected capacities between 8C and 1C rates is thus ~25 mAh g -1 , which is in good agreement with our measured practical charge/discharge capacities at these rates. The C-rate (i.e., "N·C-rate") refers to the current (i) applied to withdraw the entire electrode capacity in (1/N) hours, and is evaluated based on the initial reported theoretical capacity of CuHCF (i.e., 60 mAh g -1 ) and its active mass found in the coated electrodes. 10 The electron transfer number (n) used in Equation (S3-(S4), i.e., the number of electrons transferred per of (Cu+Fe) sites, was determined from the cyclic voltammograms according to Equation (S5): where i · dt is the integrated area under the redox peaks (i.e., the charge in units C), i is the current in A, F is the Faraday constant, and nCu+Fe is the moles of total metal (i.e., moles of Cu and Fe) on the S7 electrode determined from ICP-OES. At 1C and 8C, we estimate that 0.78 ± 0.09 and 0.54 ± 0.10 eare transferred across the interface (see Figure S2c-d).
The Coulombic efficiency (Qeff.) was calculated for each cycle according to the expression in Equation (S6): Where Qdch is the discharge capacity and Qcha the charge capacity. The round-trip efficiency (Ueff.) was also calculated for each cycle via an expression analogous to Equation (S6), however, the ratio is instead Udch/Ucha, which represents the respective discharge and charge energies at the end of each half-cycle expressed in units of mWh. Figure S1. Thermogravimetric analysis (TGA) of the pristine CuHCF powder sample without additions of PVA binder or carbon black (CB). Note that the initial mass loss upon heating from room temperature to about 120 °C is due to release of water from the CuHCF structure. The water loss from TGA analysis was estimated to 22.57 wt. %, which corresponds to 3.3 water molecules per unit formula and, a molecular weight of 267.07 g mol -1 , assuming a molar formula of K: Electrochemical impedance spectroscopy (EIS) before and after 200 cycles. Note that at the upper/lower voltages of the charge/discharge curves in (a), the curves start sloping. We regard this to the oxygen evolution reaction (OER) and the oxygen reduction reaction (ORR), respectively, based on the fact that the thermodynamic feasible onset potentials for these reactions lie within the potential window (1.00 V -2.15 V vs Zn 2+ /Zn = 0.52 V -1.67 V vs. RHE in 1M ZnSO4 pH 4.8). 11,12 It is therefore desirable to cycle this system with a slightly narrower potential window, especially at slow cycling rates. . We did not observe any recovery after the pause, whereby we conclude that the capacity loss is permanent. S11 Figure S4. SEM image of the pristine CuHCF particles before electrode coating (i.e., without any addition of CB or PVA binder).   Table S2. Metal loadings of the components of the Zn/CuHCF cell obtained from ICP-OES (cathode, separator, anode). The investigated states are following: pristine (not exposed to electrolyte, not cycled), and after 200 CV cycles between 1.00 -2.15 V at 2.5 mV s -1 in 1M ZnSO4, and stopped at OCP. Note that the cell parts were analyzed without rinsing with water (i.e., "as is"), to facilitate identifying the location of species, also the ionic ones. Thus, the Zn-content here is very high since it also originates from dried ZnSO4 electrolyte. The standard errors given in brackets were obtained from the overall number of measurements. a) The cell parts were analyzed "as is", i.e., they were not rinsed with water prior to analysis, and therefore the high Zn content originates from the dried ZnSO4 electrolyte.        The graphite cathode. All electrodes were left unwashed since the diffraction peaks were much weaker than in their washed counterparts. The pristine electrodes were never exposed to electrolyte.  13 This circumstance is confirmed by the appearance/broadening of the D' feature next to the main G band. The D' feature is characteristic of defected graphite, and the slight broadening of the 2D band at higher wavenumbers shown in (a) points to higher disorder and possible local distortions of the graphitic structure. In all the cycled graphite electrodes, the spectral features associated with the disorder disappear almost completely after rinsing the electrodes with water. This suggests that the contact with 1M ZnSO4 electrolyte and cycling induces a mild intercalation of Zn 2+ and/or SO4 2into the graphite structure. As a result, part of the electrolyte enters the surface structure of the graphite, nevertheless, after washing with ultrapure water the salt is removed from the surface whereby the original spectral features of pristine graphite are resumed. The washed graphite cathode. All EDS spectra were normalized to the total unit area, and have the same y-axis scale to allow for direct comparison between the samples. It is visible that the Zn, S and O signals decrease after washing the cycled electrodes with water, although, a large extent remains in graphite.  Figure S22. XPS spectra of the CuHCF cathode after different stages of cycling between 1.00 -2.15 V at 2.5 mV s -1 in 1M ZnSO4. (a) Survey spectra. (b) High-resolution spectra (C 1s, K 2p, Cu 2p, Fe 2p, and Zn 2p). The cycled states are displayed from top to bottom: Pristine, 0 cycles (1h at OCP), 100 cycles, and 200 cycles. All spectra were energy calibrated to the graphitic sp 2 -C peak (284.3 eV) from the carbon black. The Cu 2p, Fe 2p, and Zn 2p spectra are also shown in the main manuscript, however, here, the entire spectra are displayed including the 2p1/2 and 2p3/2 regions.  0.18 n/a n/a n/a Figure S24. XPS of reference compounds. All samples were calibrated by setting the sp 3 -C peak (adventitious carbon) to 284.8 eV, except for the two CuHCF cathode samples shown at the bottom, which were calibrated to the sp 2 -C peak at 284.3 eV, since these contain graphitic carbon.       1 The simulations were performed between a krange of ~2-13 Å -1 . For the Cu and Fe K-edges, the coordination numbers were fixed, while they were minimized for the Zn K-edge due to diverse structures. An amplitude reduction factor (S0 2 ) of 0.85 was used for all K-edges. The samples stopped at OCP were analyzed in "ex situ" configuration (i.e., the cells had been disassembled and the samples rinsed with water before analysis), and the samples analyzed in the charged (2.15 V) or discharged (1.00 V) states were analyzed in "in situ" configuration (i.e., in real-time while mounted in the cells and applying a potential). Table S10. Debye-Waller parameters, ionization energy correction, and R-factor filtered for the fit parameters presented in Table S9 above. Fit errors were estimated up to a distance of 5 Å useful Rspace.
(b) The Debye-Waller parameters were minimized in a global fit across the Fe-N and Fe-C shells, however, were fitted individually for each electrode since the coordination numbers were fixed.
(c) The Debye-Waller parameters were minimized in a global fit across all electrodes since the coordination numbers were not fixed for the Zn edge, to avoid overparameterization.   All electrodes had been rinsed with ultrapure water prior to analysis to remove excess of the ZnSO4 electrolyte; without this step, there were no differences between the charged/discharged states in the Zn 2p and S 2p spectra. All spectra were energy calibrated to the graphitic sp 2 -C peak at 284.3 eV. 2,3 The blue component at higher B.E. in the Zn 2p spectrum (ZnSO4 * ) we propose might be related to a surface species of ZnSO4, either crystalline ZnSO4 or a smalller fraction of intercalated ZnSO4 into graphite. The same component in the O 1s spectra at high B.E. may as well be this surface species, or water molecules in a different coordination environment. In the Zn 2p spectrum, it is visible that there is a slightly higher signal from Zn 2+ (in ZnSO4) in the discharged state (1.00 V) in both the CB-PVA and graphite cathodes, which is interesting since there is no difference between 1.00 V (DCH) and 2.15 V (CHA) for the washed CuHCF cathode shown in Figure 7a in the main manuscript. In these carbon-containing cathodes, there is on the other hand no difference in the SO4 2signal in the S 2p spectra between 1.00 V and 2.15 V, which is in contrast to CuHCF where there is a significant increase in the SO4 2signal at 2.15 V. CB-PVA, 2.15 V n/a n/a n/a 77.9 (0.5) 22.0 (0.4) 0.03 (0.02) 0.06 (0.03) CB-PVA, 1.00 V n/a n/a n/a 77.5 (0.4) 22.4 (0.4) 0.04 (0.02) 0.05 (0.03) Graphite, 2.15 V n/a n/a n/a 94.5 (0.5) 5.10 (0.50) 0.13 (0.03) 0.32 (0.05) Graphite, 1.00 V n/a n/a n/a 93. 8  The graphite cathode. The electrodes had been washed with ultrapure water prior to analysis to remove excess of ZnSO4 electrolyte. All EDS spectra were normalized to the unit area, and both (a) and (b) have the same y-axis scale for comparison reasons.