Water driven phase transitions in Prussian white cathode materials

Synthesis of each hydrated Prussian white (H-PW) batch proceeded by preparing a solution of 0.44 M Na₄[Fe(CN)₆]·10H₂O (Sisco Research Laboratories Pvt. Ltd; extra pure AR, 99%) in deionised and deoxygenated water and heating to 80 °C. An approximately two-fold molar excess of HCl (37%) was added over 12 h using a syringe pump to produce the acid-facilitated self-decomposition of Na₄[Fe(CN)₆]·10H₂O. N₂ was flowing during the reaction process that lasted 24 h. The obtained precipitate was filtered under inert conditions, washed four times with deionized and deoxygenated water, and finally dried under flowing N₂ within the filtration setup for 12–24 h. Dehydrated Prussian White (D-PW) was prepared by drying a batch of H-PW at 160 °C for 48 h under vacuum. The resulting powders were stored in an Ar glovebox for further studies.

Synthesis of PW via co-precipitation (CP-PW) proceeded by preparing two solutions in deoxygenated water stirring for 1 h. Solution A contained 0.1 M FeSO₄, 0.51 M sodium citrate, and 0.225 M ascorbic acid. Solution B contained 0.1 M Na₄Fe(CN)₆. The two solutions were added to a reaction bottle containing deionized and deoxygenated water heated to 80 °C at a constant rate of 4 ml h⁻¹ over 13.5 h with stirring under flowing N₂. The reaction solution was further aged for 8 h before filtration, washing with deionized and deoxygenated water, and finally drying at ambient temperature for 24 h under inert conditions. The resulting powder was stored in an Ar glovebox for further studies.

Sample composition was determined using inductively coupled plasma-optical emission spectroscopy (ICP-OES) to determine Na and Fe contents, while elemental analysis was used to determine C, N, H content. Measurements were performed by Medac Ltd, United Kingdom. Water content and dehydration temperature of the samples were studied with thermogravimetric analysis (TGA) using TA Instruments TGA Q500 by loading ∼5 mg of powder in an Al pan in air before placing the sample into the TGA furnace under flowing N₂ (60 ml min⁻¹). The sample was heated to 500 °C with a heating rate of 5 °C min⁻¹. Particle size and morphology were investigated using scanning electron microscopy (SEM) on a SEM/EDS-Zeiss 1550 instrument working with an acceleration voltage of 3 kV using an in-lens detector. The sample was loaded onto carbon tape in a glovebox and quickly transferred into the vacuum chamber of the microscope.

Mössbauer measurements were carried out at RT using constant acceleration type of vibrator and a ⁵⁷CoRh source. The samples were enclosed in a sealed Al cover. The so-formed absorber had a sample concentration of ∼10 mg cm⁻². Calibration spectra were recorded at 295 K using natural Fe metal foil as a reference absorber. The spectra were folded and fitted using the least square Mössbauer fitting program Recoil to obtain the values of the center shift CS, the magnitude of the electric quadrupole splitting IQSI, the full-width at half maxima W of the Lorentzian absorption lines, and the spectral areas A.

Samples were prepared for laboratory XRD experiments by dispersing the powder on a single crystal Si disc. Powder XRD measurements were performed on a Bruker D8 TwinTwin diffractometer (double Cu Kα radiation λ₁ = 1.540600 Å, λ₂ = 1.544390 Å) with a Lynx-eye XE-T position sensitive detector. The experiments were conducted between 10 and 80° in steps of 0.02°. Synchrotron powder XRD data were collected on the high-resolution powder diffraction beamline P02.1 at PETRA III, Hamburg, Germany [24]. The wavelength was determined to be 0.20734 Å from measurements of a LaB₆ standard (NIST660c). The samples were filled into 0.3 mm glass capillaries in an argon-filled glovebox for high temperature measurements using a capillary spinner and Oxford hot–air blower with a variable heating rate. The diffraction patterns were collected using a Varex XRD 4343CT detector (150 × 150 µm² pixel size, 2880 × 2880 pixel area). The obtained 2D patterns were integrated into 1D diffraction patterns using Dawn [25]. Neutron powder diffraction (ND) data were collected on the time-of-flight instrument POLARIS [26] at the ISIS Pulsed Neutron and Muon Source [27]. 1.6960 g of H-PW was loaded into an open cylindrical vanadium can and measured continously from RT to 250 °C with a heating rate of 0.4 °C min⁻¹ with a 1 h collection time at RT, 1 h and 50 min collection time at 250 °C and a collection time of 10 min for each of the patterns during heating. Upon cooling, diffraction data was measured with a collection time of 4 min/pattern with a cooling rate of ∼1.25 °C min⁻¹. 1.6095 g of D-PW was loaded into a open cylindrial vanadium can, cooled to 4 K, and measured for 12 h at 4 K. All refinements were performed using the Rietveld method [28] with symmetry-mode refinements using ISODISTORT [29] in the TOPAS software [30] where scale factors, cell parameters, peak shape, and atomic displacement factors (ADPs) were refined.

ICP-OES analysis of H-PW revealed the composition Na_2.14(7)Fe[Fe(CN)₆]·zH₂O indicating a successful synthesis of Na-rich, low vacancy PW (table S1). TGA measurement of H-PW (figure S1) shows a weight loss of ∼10.54(3) wt% when T < 225 °C, which corresponds to the release of surface absorbed water and interstitial water (=2.07(2) H₂O per formula unit (f.u.)). Similarly, the composition of the PW sample synthesized by co-precipitation (CP-PW) was determined to be Na_2.17(5)Fe[Fe(CN)₆]·1.96(2)H₂O from ICP-OES and TGA analysis (figure S1, table S1) and the composition of D-PW was determined to be Na_2.08(7)Fe[Fe(CN)₆]_0.98(2) (table S1).

All samples have similar cubic morphology with H-PW and D-PW being micron-sized particles (as previous reported [6]) and CP-PW being multifaceted intergrown micron-sized particles (figure S2).

The Mössbauer spectrum at 295 K of H-PW (figure S3) shows a three-line structure, which was fitted using two quadrupole split doublets. The central unresolved line originates from low spin Fe²⁺ _C and the well-resolved doublet originates from high spin Fe²⁺ _N. The fitting parameters are presented in table S2. The hyperfine parameters are in good agreement with the results from earlier studies of PW [5, 6, 31]. The nominal value of Fe vacancies at the low spin site was verified from the Mössbauer spectral areas corresponding to low spin Fe_C and the high spin Fe_N. The areas were 47(3):53(3) indicating a vacancy-free structure.

Laboratory XRD was performed immediately following the synthesis of H-PW and revealed that the material adopts a monoclinic structure with space group $P{2_1}/n$ (figure S4).

Neutron powder diffraction data were collected from RT up to 250 °C from H-PW. In the RT pattern, two phases were observed due to an incomplete phase transition which occurs between 30 to 40 °C from $P{2_1}/n$ to $R\bar 3$. The presence of the $R\bar 3$ phase can be seen in figure 2 by the appearance of the $20\bar 4$ reflection between the $022$ and $400$ reflections of the $P{2_1}/n$ phase in the peak at 2.6 Å. A small amount of intensity of the $P{2_1}/n$ reflections was observed at 33 °C with all the intensity gone by 40 °C. Thus, the transition is likely complete at ∼35 °C. This phase transition was also observed for a sample with similar composition Na_2.17(5)Fe[Fe(CN)₆]·1.96(2)H₂O, CP-PW, prepared by co-precipitation. From synchrotron XRD (figures S5 and S6), CP-PW undergoes the same phase transition from $P{2_1}/n$ to $R\bar 3$ between 30 to 40 °C. Hence, the phase transition near RT for Na-rich PW is independent of the synthesis method. The existence of both structures for the same compound near to RT has not been reported previously, although the two structures have been reported independently at RT [7, 13, 15].

Zoom In Zoom Out Reset image size

The difference between the $P{2_1}/n$ and $R\bar 3$ structures relates to octahedral tilting [32]. Octahedral tilting involves either in-phase or out-of-phase rotation of the transition metal octahedra in PBAs. This is described by Glazer notation with '+' for in-phase and '−' for out-of-phase tilts with the lowercase letter denoting the relative magnitude of the tilt [32]. For PBAs with a high sodium or potassium content, the most prevalent tilt pattern is ${a^ - }{a^ - }{b^ + }$, which drives the monoclinic $P{2_1}/n$ symmetry. This involves an in-phase tilt of the octahedra around one of the pseudo-cubic unit cell axes with out-of-phase tilts along the other two directions [33]. A tilt pattern of ${a^ - }{a^ - }{a^ - }$ gives rise to rhombohedral $R\bar 3$ symmetry, which is often seen for dehydrated sodium-containing PBAs. In this case, the tilting is out-of-phase for all three pseudo-cubic unit cell axes. It is the single in-phase tilt which gives rise to the split peaks in figure 2, which indicates that one of the crystallographic axes differs from the other two ($P{2_1}/n$).

3.2.1. Refinements of the two near room temperature phases

Since the phase transition occurs close to RT, it is difficult to refine the pure $P{2_1}/n$ phase due to the presence of a second $R\bar 3$ phase, which complicates the modeling of the neutron diffraction data. The $R\bar 3$ phase was refined against data collected at 40 °C since the phase transition is complete at this temperature.

The PBA framework model with sodium included was refined prior to calculating Fourier difference maps to locate the water molecules within the structure. A $R\bar 3$ model containing Fe, C, N, and Na was fitted to the 40 °C data. The refined framework structure is shown in figure 3(a) including the out-of-phase tilt along the a direction of the pseudo-cubic unit cell. Water positions from the literature were tested to see if the scattering difference observed in the Fourier difference maps (figure 3(b)) was accounted for. The $R\bar 3$ model from Wang et al [14] was applied by replacing two of the lower occupancy Na⁺ ions with oxygen atoms (located at −0.167, 0.167, 0.167 and 0.167, 0.333, 0.333 (Wyckoff position 18f)). This yielded a reasonable fit with all positive scattering difference accounted for (figure 3(c)), although with high ADPs for the oxygen atoms. Refinement of the oxygen occupancies resulted in zero occupancy of the oxygen site located at −0.167, 0.167, 0.167 (18f), which was subsequently removed from the model. This gave a stable refinement with reasonable ADPs for the oxygen site located at 0.167, 0.333, 0.333 (18f). Hydrogen atoms were not refined due to the positional and orientation disorder of the water molecules. Inclusion of hydrogen barely changes the simulated neutron diffraction pattern and subsequently does not affect the quality of the fit (figure S7). Similar results are reported in previous studies [20, 22, 23].

Zoom In Zoom Out Reset image size

The established $R\bar 3$ model from the 40 °C data was fixed as the second phase in the RT data. The host framework structure of the primary $P{2_1}/n$ phase containing Fe, C, N, and Na was refined from a starting model reported by Song et al [20] and is presented in figure 3(d) including the in-phase tilt along the a direction of the parent cubic unit cell. Similar to the $R\bar 3$ model, Fourier difference maps were generated to determine the position of oxygen atoms. This revealed two possible sites (figure 3(e)), which were compared to sites reported in a previous $P{2_1}/n$ model [6]. Inclusion of two oxygen sites on the 4e Wyckoff position within the bc-plane removed all the positive scattering differences (figure 3(f)), though the ADPs of the oxygen atoms were high. A closer inspection of the Fourier difference maps revealed an elongation of the observed positive scattering differences. Hence, one of the oxygen atoms was split into two sites, which reduced the ADPs and improved the stability of the refinement. The hydrogen positions were not refined for reasons outlined above.

The final refined models with oxygen included are displayed in figures 4(a), (b) and S8, S9 with refinement details in tables S3 and S4. The model fitted to the RT data consists mainly of the $P{2_1}/n$ structure (82.8(5) %) with a minor hydrated $R\bar 3$ phase (17.2(5) %). The data measured at 40 °C is fitted with a single hydrated $R\bar 3$ structure.

Zoom In Zoom Out Reset image size

In the RT $P{2_1}/n$ structure, Na⁺ is coordinated to a random distribution of O atoms within the bc-plane resulting in only four faces of each subcube occupied by O (figure 4(c)). This results in clear Na⁺ channels along a. Interestingly, the in-phase tilt ${b^ + }$ also lies along the a direction implying that the positioning of water might direct the nature of tilting in PBAs. Layering of Na⁺ and O in the bc-plane was also found for the $P{2_1}/n$ structure of Na_1.89Mn[Fe(CN)₆]_0.97·1.87H₂O [20]. For the $P{2_1}/n$ model in the present study, one of the O atoms was included as a split site. This can be explained by considering the ideal Na–O bond distance for water coordinated to Na⁺ ions, which is 2.3 Å [7, 20]. The distance between the center of mass for the split site and a neighboring Na⁺ is 2.6 Å. Thus, in reality water must lie slightly closer to one of the two Na⁺ ions. These displacements are not long-range ordered leading to the appearance of a split site on average. The obtained bond angles for the $P{2_1}/n$ structure reported here agree with those previous reported [6, 14] with the Fe–N–C bond angle being similar to or less than the Fe–C–N bond angles. This is due to the σ-bonding character of the Fe–N bond providing higher flexibility (figure 4(d)) [18, 20].

Transitioning to the hydrated $R\bar 3$ structure determined at 40 °C (figure S9), O instead appears to be randomly distributed over three faces of each subcube. Here, there are no clear Na⁺ channels in any direction. It should be noted, however, that several distributions of the O atoms were tested which resulted in similar ADPs and quality of fits. This can partly be explained by the short collection time (10 min) of the 40 °C data. All models provided equal quality of fits except for one model containing O atoms on both the −0.167, 0.167, 0.167 and 0.167, 0.333, 0.333 sites, which did not calculate any intensity for the $003$ reflection in the bank 2 data. Thus, it is difficult to draw any compelling conclusions about the arrangement of water in the $R\bar 3$ phase, except for the higher degree of disorder compared to the $P{2_1}/n$ structure. Aside from the change in the tilting system between the $P{2_1}/n$ and $R\bar 3$ structures (figure 3) and very minor change of the offset of Na⁺ within a subcube (figure S10), all bond distances and angles remain similar. Insight into the local structure using total scattering techniques would unravel any local order of water in the hydrated $R\bar 3$ structure. In summary, the $P{2_1}/n$ and $R\bar 3$ structures can both exist for a single compound and the structures were determined with higher degree of confidence due to the use of neutron diffraction.

Prior to use in SIBs, PW is dehydrated which results in structural collapse. To investigate the distortions driving this change in detail, neutron diffraction data were collected on a fully dehydrated sample, D-PW. These data were initially modeled using the obtained hydrated ${{R}}\bar 3$ model without the O atoms included (figures 5(a) and (b), tables S5 and S6). The structure agrees with previously reported structures of dehydrated PW from x-ray data and ab initio calculations [6, 16]. Upon dehydration, the volume reduces by 20% relative to the hydrated material, leading to loss of crystallinity (as indicated by peak broadening) in line with previous studies [6, 20]. Symmetry-mode refinement allows the causes of this phase transition to be understood more clearly. The key distortion modes identified were displacement of Na⁺ along c towards the FeN₆ octahedra and the tilting of rigid FeN₆ and FeC₆ octahedra. As expected, due to the π-bonding character of the Fe–C bond, the Fe–C–N bond angle does not deviate significantly from the hydrated material (figure 5(c)) [18, 20]. Finally, Fe-C/N bond distances remain unchanged relative to the hydrated material.

Zoom In Zoom Out Reset image size

Summarizing the observed changes, the mechanism for the volume collapse can be described. The Na⁺ ions displace along the c direction due to Coulombic attraction to the higher electron density on the N-bound Fe atom. The N atoms opposite are pulled with the Na⁺ ion towards the same direction of displacement due to the same Coulombic attraction resulting in short Na-N distances. At the same time, the rigid octahedra are rotated. The structural collapse is also evident looking at the rotation of the octahedra (table 1). In the hydrated ${{R}}\bar 3$ structure of H-PW, the FeC₆ octahedra are rotated 1.69(1)° while the FeN₆ octahedra are rotated 1.52(7)°. Upon dehydration, the FeC₆ octahedra rotate 17.44(6)° while the FeN₆ octahedra rotate 22.60(4)°, thus, the magnitude of the tilt in the ${{{a}}^ - }{{{a}}^ - }{{{a}}^ - }$ tilt pattern increases. It is reasonable that the FeN₆ octahedra rotate to a larger degree due to the Coulomb attraction between Na⁺ and N⁻ and the stronger Fe–C bond providing less flexibility [ 11, 18, 35 ]. Finally, the removal of water does not change the nature of distortions in PBAs, but rather modifies the magnitude of the pre-existing distortions.

To study the dehydration mechanism, the structural evolution of H-PW was followed during heating using x-ray and neutron diffraction (figures 6(a) and S12). Initially, PW adopts a ${{P}}{2_1}/{{n}}$ structure which converts to a hydrated ${{R}}\bar 3$ structure close to 35 °C. From the synchrotron XRD data, H-PW transforms from the hydrated ${{R}}\bar 3$ phase to a cubic ${{Fm}}\bar 3{{m}}$ phase at ∼120 °C as evidenced by the appearance of the 440 cubic reflection between the $4\bar 20$ and $208$ reflections of the hydrated ${{R}}\bar 3$. This cubic phase coexists with the dehydrated ${{R}}\bar 3$ phase from 220 °C to 245 °C when heated at a rate of 1 °C min⁻¹ (figure S12). Interestingly, the transition to ${{Fm}}\bar 3{{m}}$ begins at 176 °C in the ND data (figure 6(a)) and the transition to cubic did not go to completion before complete conversion to the dehydrated ${{R}}\bar 3$ phase at 230 °C. Here, a heating rate of 0.4 °C min⁻¹ was applied. In this case, the ND experiment used a larger sample volume compared to the XRD experiment (∼2 g vs ∼2 mg) potentially leading to lower heating efficiency through the whole sample. Previous work on Na_1.76Fe[Fe(CN)₆]·2.6H₂O revealed that the cubic ${{Fm}}\bar 3{{m}}$ phase appears at 150 °C when heated at 3 °C min⁻¹ and continues to coexist with a high-temperature ${{R}}\bar 3$ phase until 300 °C [15].

Zoom In Zoom Out Reset image size

It is clear that the dehydration process is kinetically limited and thus dependent on the heating efficiency and heating rate. The sluggish kinetics of dehydration is likely due to a combination of strong bonding of water to sodium within the structure [31], large diffusion distances through the micron-sized particles, and the energy cost associated with induced strain from the phase transition to the dehydrated ${{R}}\bar 3$ structure. It has been previously shown that Na⁺ ion diffusion through the structure is similarly affected by the presence of water [10]. Further insight on the impact of water and the temperature at which it begins to leave the structure can be obtained from the cell parameter changes during the variable temperature experiment (figure 6(c)).

The thermal expansion of H-PW was extracted from the variable temperature data in the range 40–140 °C. The thermal expansion coefficients were calculated using PASCal [36]. In contrast to most PBAs [ 37, 38 ], H-PW exhibits anisotropic thermal expansion and a small positive volumetric expansion upon heating. The coefficients of thermal expansion (CTE) are ${{{\alpha }}_{{{ab}}}} = - 1.2\left( 9 \right){{\,M}}{{{K}}^{ - 1}}$ for the ab-plane and ${{{\alpha }}_{{c}}} = 34\left( 2 \right){{\,M}}{{{K}}^{ - 1}}$ for the c direction. After dehydration, the sample (now called DH-PW) was cooled from 250 °C down to 35 °C. DH-PW also shows anisotropic thermal expansion with the ab-plane expanding and the c direction contracting with corresponding CTEs of ${{{\alpha }}_{{{ab}}}} = 52.2\left( 3 \right){{\,M}}{{{K}}^{ - 1}}$ and ${{{\alpha }}_{{c}}} = - 31.1{{\,}}\left( 2 \right){{\,M}}{{{K}}^{ - 1}}$ over the temperature range measured. DH-PW also exhibits positive volumetric expansion with a higher rate compared to the hydrated phase. Most PBAs show isotropic negative thermal expansion upon heating (due to their cubic symmetry) with very low magnitudes of CTE [ 37 – 40 ]. The different behavior of PW is a result of the lower symmetry, however, the lack of thermal expansion studies of other distorted PBAs prevents more in-depth comparison. However, the difference between H-PW and DH-PW indicates that the presence of water has a strong influence on the sign and magnitude of the expansion.