Phase Transformations and Phase Segregation during Potassiation of Sn x P y Anodes.

K-ion batteries (KIBs) have the potential to offer a cheaper alternative to Li-ion batteries (LIBs) using widely abundant materials. Conversion/alloying anodes have high theoretical capacities in KIBs, but it is believed that electrode damage from volume expansion and phase segregation by the accommodation of large K-ions leads to capacity loss during electrochemical cycling. To date, the exact phase transformations that occur during potassiation and depotassiation of conversion/alloying anodes are relatively unexplored. In this work, we synthesize two distinct compositions of tin phosphides, Sn4P3 and SnP3, and compare their conversion/alloying mechanisms with solid-state nuclear magnetic resonance (SSNMR) spectroscopy, powder X-ray diffraction (XRD), and density functional theory (DFT) calculations. Ex situ 31P and 119Sn SSNMR analyses reveal that while both Sn4P3 and SnP3 exhibit phase separation of elemental P and the formation of KSnP-type environments (which are predicted to be stable based on DFT calculations) during potassiation, only Sn4P3 produces metallic Sn as a byproduct. In both anode materials, K reacts with elemental P to form K-rich compounds containing isolated P sites that resemble K3P but K does not alloy with Sn during potassiation of Sn4P3. During charge, K is only fully removed from the K3P-type structures, suggesting that the formation of ternary regions in the anode and phase separation contribute to capacity loss upon reaction of K with tin phosphides.

Introduction

Conversion- and alloying-type anodes offer high capacities in alkali metal batteries, such as commercially successful Si anodes in Li-ion batteries (LIBs). Due to the greater abundance of K compared to Li (and the ability to use aluminum current collectors), K-ion batteries (KIBs) have the potential to lower the cost of grid-scale electrochemical storage[1−3] and reduce demand pressure on Li. However, KIBs are still a nascent technology compared to LIBs, requiring new avenues of electrode and electrolyte materials design.[4−7] K-ion conversion electrodes that can accommodate the large volume changes associated with K insertion/removal while simultaneously delivering practical energy densities are being actively developed. Phosphorus- and phosphide-based compounds display high theoretical specific capacity for KIBs,[8−12] making these materials attractive to study as potential anodes. In particular, binary metal phosphides provide a route to mitigate the large volume expansion observed upon conversion reactions with phosphorus (up to 600% for the formation of K3P) by producing ternary intermediates and/or displacement and subsequent conversion reactions during potassiation of the parent phase.[13−18]

Of the binary compounds, tin phosphides are readily synthesized in a variety of different stoichiometries (e.g., Sn4P3, SnP, Sn3P4, and SnP3[19,200]) offering a rich phase space to study the mechanisms underpinning potassiation/depotassiation in high-capacity (>600 mAh g–1) anodes. Yet, many of the intermediates generated upon K reaction with tin phosphides are highly disordered and separate into multiple phases, making structural assignment challenging. For example, Guo and co-workers used operando X-ray diffraction (XRD) to examine the potassiation behavior in Sn4P3, proposing a mechanism by which the Sn4P3 phase-separates into metallic Sn and amorphous P upon discharge to ∼0.21 V vs K+/K, allowing the formation of K3P11.[20] After further discharge to potentials below 0.17 V vs K+/K, K3P11 reacts with additional K to form K3P and metallic Sn alloys with K to form KSn. However, each of these assignments relied on a singular, low-intensity reflection in the XRD pattern (12.71° (for KSn), 12.14° (for K3P11), and 14.27° (for K3P)). Li and co-workers also report segregation of Sn4P3 and Sn, followed by conversion and alloying reactions that form K3P, KSn, and K4Sn23, although these assignments rely on minor, low-intensity XRD reflections or lattice spacing measurements from high-resolution transmission electron microscopy (HRTEM) of the crystalline components.[21] In another study, Park and co-workers reported on the final discharge products formed at the end of potassiation of SnP3 using ex situ XRD, HRTEM, and X-ray photoelectron spectroscopy (XPS).[22] The authors proposed a reaction pathway that included phase separation of Sn particles from bulk SnP3 and a conversion reaction between the remaining P and K to form K3P. The Sn followed a two-step alloying reaction to first form K4Sn23 and then KSn. Yet only one reflection could be observed for each intermediate in the diffraction data, indicating that heterogeneity and disorder in the anode complicate structural analyses.

Experimental and computational tools that can identify and parse amorphous, multicomponent species with high chemical resolution are required to understand and control complex potassiation behavior in binary anodes. In this study, we use a combination of 31P and 119Sn solid-state nuclear magnetic resonance (SSNMR) spectroscopy and density functional theory (DFT) calculations of stability and simulated NMR spectra for a large set of hypothetical structures to elucidate the amorphous and short-range crystalline structure in products that form during potassiation/depotassiation of SnP3 and Sn4P3. Assignments of crystalline structures are supported via ex situ XRD at different states of discharge/charge of KIBs assembled with both anode materials. Both experiment and theory suggest that SnP3 and Sn4P3 exhibit distinct potassiation mechanisms. K insertion into Sn4P3 and SnP3 generates local environments resembling K-Sn-P ternary phases that are predicted to be stable based on DFT calculations. Displacement of metallic Sn is observed in the case of Sn4P3, but no Sn phase separates upon potassiation of SnP3. Unambiguous assignment of metallic Sn during K insertion is achieved by monitoring the metallic Knight shift in 119Sn NMR (metals give rise to unique NMR shifts that are well resolved from diamagnetic compounds due to coupling between nuclei and conduction electrons[23]). When metallic Sn is present, little to no K–Sn alloying is observed.

Experimental Section

Materials and Methods

Potassium metal (chunks in mineral oil, 98% trace metals basis), potassium hexafluorophosphate (KPF6, > 99.5%), propylene carbonate (PC, anhydrous, > 99%), dimethyl carbonate (DMC, anhydrous, > 99%), hexanes (anhydrous, > 99%), Sn (> 99%), and sodium carboxymethyl cellulose (CMC) were purchased from Sigma-Aldrich. Red phosphorous powder (98.9%) was purchased from Alfa Aesar. Carbon Super P was purchased from MTI Corporation. Prior to use, KPF6 salt was dried in vacuo overnight at 100 °C before bringing into an Ar-filled glovebox (O2 < 0.1 ppm, H2O < 0.5 ppm). All other materials were used as received.

Synthesis of SnP3 and Sn4P3

SnP3 and Sn4P3 were prepared by first mixing Sn and red P in 1:3 and 4:3 molar ratios, respectively, in a stainless-steel ball mill in an Ar-filled glovebox. Powder mixtures were ball-milled in a SPEX 8000M Mixer/Mill for 8 h (SnP3) and 1 h (Sn4P3). Product purity was confirmed with XRD and SSNMR.

Electrode Fabrication

Electrode films were created by mixing an 8:1:1 mass ratio of SnP:carbon Super P:CMC binder. First, the SnP and carbon Super P were mixed in a stainless-steel ball mill under Ar for 30 min. In a mortar and pestle, water was added dropwise (∼10 drops per 100 mg of dry mixture) to CMC binder until a slurry was formed. The SnP and carbon mixture was then added and mixed to make a uniform slurry. The slurry was cast onto a Cu current collector (6 μm thick, MTI) using a 150 μm doctor blade and dried at 100 °C under vacuum overnight. The dried film was punched into 12.7 mm diameter disks to use in cell assembly. Typical mass loadings of active material (SnP) per anode were 5–15 mg cm–2. To assemble the opposing K electrodes, K metal was first treated by rinsing thoroughly in hexanes, then removing the external oxide layer with a razor blade. Small pieces of potassium were then placed in a bag coated with hexanes and rolled into thin sheets (∼0.25 mm thick) using a cylindrical weight. The K sheet was then removed from the bag and, after waiting for the hexanes to evaporate, stamped into 12.7 mm diameter disks. These electrodes were used for all electrochemical testing and NMR characterization.

Electrochemical Cycling

2032-type coin cell casings were used to assemble K/SnP3 and K/Sn4P3 half-cells with 15 mm diameter glass microfiber separators (purchased from GE Life Sciences) soaked with 0.8 M KPF6 in PC. Galvanostatic cycling experiments were performed at C/200 and C/100 for SnP3 and Sn4P3, respectively (where nC refers to full theoretical discharge in 1/n h). C-rates were calculated from the theoretical capacities of SnP3 and Sn4P3 for the formation of K3P and KSn (1266 and 613 mAh g–1, respectively). Cells were discharged to 0.05 V and charged to 2.00 V vs K+/K.

Electrode Extraction

Table summarizes the states of charge at which galvanostatic cycling was stopped for electrode extraction and characterization. After cycling, the SnP electrode was removed from the coin cell and submerged in DMC for 30 s before drying at room temperature under vacuum for 30 min. Extraction was performed in an Ar-filled glovebox within, at most, 4 h after cells completed cycling. To collect data for the pristine electrodes, SnP electrodes were assembled into coin cells, then immediately removed, washed, and dried as described.

Table 1

States of Charge (V vs K+/K) Where SnP Electrodes Were Extracted for Structural Analysis with XRD and NMR

		XRD	NMR
SnP3	discharge	1.50, 0.67, 0.53, 0.47, 0.35, 0.06	0.53, 0.06
	charge	1.77, 2.00	1.77, 2.00
Sn4P3	discharge	0.90, 0.26, 0.21, 0.18, 0.11, 0.06	0.18, 0.06
	charge	1.50, 2.00	1.50, 2.00

X-ray Diffraction

XRD patterns were collected on a PANalytical XPERT3 powder diffractometer with Cu Kα radiation. To confirm synthesis, pristine SnP3 and Sn4P3 powders were removed from the Cu current collector and placed directly on a zero-background Si plate in the glovebox and sealed with Kapton polyimide film (Chemplex) in an air-free sample holder for data collection. To conduct ex situ measurements, individual SnP electrodes (including the Cu current collector) were placed on a zero-background Si plate and sealed in the same manner. Reference powder diffraction patterns for SnP3 and Sn4P3 are from the Inorganic Crystal Structure Database (Collection Codes 16293 and 15014, respectively). Rietveld refinement was performed using the TOPAS Academic software package (version 7.12). The crystallite sizes were estimated by a Voigt-convolution approach according to Balzar et al., assuming a lognormal size distribution.[24,25]

One-Dimensional (1D) NMR Measurements

SSNMR experiments were performed at room temperature on a Bruker Avance NEO 600 MHz spectrometer equipped with a 1.6 mm HFXY magic-angle spinning (MAS) Phoenix NMR probehead. SnP electrodes at various states of charge were washed and dried as described above. The SnP active material was then scraped from the Cu current collector, ground in a mortar and pestle, and packed in a 1.6 mm o.d. ZrO2 rotor in an Ar-filled glovebox. 1D 31P NMR (90° single pulse, 10 s recycle delay, 2048 scans) experiments were referenced to the 31P shift of solid ammonium dihydrogen phosphate at 0.8 ppm. The recycle delay for 31P NMR experiments was optimized on the pure phase binary compounds. 119Sn NMR was collected using a rotor-synchronized Hahn echo (90°−τ–180°−τ–acquire, where τ = 55.6 and 35.7 μs for 18 and 28 kHz MAS frequency, respectively) pulse sequence. 119Sn NMR was collected at two separate offset frequencies: one corresponding to metallic Sn at 6894 ppm with 46,080 scans and 0.5 s recycle delay, and a second corresponding to diamagnetic Sn compounds and alloys at −153 ppm with 512 scans and 60 s recycle delay, unless otherwise noted. 119Sn NMR shifts were externally referenced to SnO2 at −604.3 ppm.

Density Functional Theory Calculations

The ternary K–Sn–P composition space has been recently explored using various crystal structure prediction techniques, namely, ab initio random structure searching (AIRSS), evolutionary algorithms, and structural prototyping,[26,27] yielding a wealth of predicted stable and metastable phases that could form during electrochemical cycling. Gauge Including Projector Augmented Waves (GIPAW) calculations were performed with CASTEP (v20)[28,29] to predict the 31P chemical shifts for these stable and low-lying metastable phases of K–Sn–P reported in previous work.

The GIPAW-NMR calculations employed the PBE functional,[30] Vanderbilt ultrasoft pseudopotentials,[31] and a plane-wave cutoff energy of 650 eV. The Brillouin zone was consistently sampled with Monkhorst–Pack grids of maximum spacing of 2π × 0.02 Å–1.[32] All structures were relaxed until the forces on each ion fell below 0.01 eV/Å–1 prior to performing the chemical shielding calculations; any structures exhibiting metallicity at the PBE level of theory were excluded from the subsequent GIPAW calculations due to the difficulty of reliably simulating chemical shielding of paramagnetic systems.[33] The computed chemical shielding tensors were referenced against reported experimental 31P shifts for black P, LiP5, Li3P, NaP, and Na3P to provide chemical shift tensors (Figure S9), showing good agreement (fit gradient of −0.997, R2, of 0.996).[34,35] All of the described calculations and analyses were performed at high throughput with the open-source Python library matador.[26]

Results

Synthesis and Characterization of Tin Phosphide Anodes

SnP3 and Sn4P3 were synthesized by ball-milling stoichiometric amounts of Sn and P. The structural integrity of SnP3 and Sn4P3 after ball-milling with conductive carbon for electrode fabrication was confirmed by XRD (Figure ). Rietveld refinement of XRD patterns from the SnP3 and Sn4P3 films indicates that the average grain sizes of the active materials are 9 and 16 nm, respectively (assuming a lognormal distribution). Nano-sized active materials are consistent with previously reported syntheses for use in KIBs.[22,36] Particles in this size range are expected to improve transport through the active material and may be less prone to fracture and pulverization during electrochemical cycling compared to micron-sized particles.[37]

Figure 1

Rietveld refinement of the XRD pattern corresponding to (a) SnP3 and (b) Sn4P3 films after ball-milling with conductive carbon, mixing with CMC, and coating on Cu foil. The crystallite sizes were estimated using a Voigt-convolution approach according to ref (24), assuming a lognormal size distribution.

Electrochemical Characterization of Tin Phosphides and Possible Potassiation Pathways

The experimental galvanostatic discharge/charge plot (Figure ) shows that tin phosphides exhibit first-cycle discharge capacities of 176 (for SnP3) and 235 mAh g–1 (for Sn4P3). The lower discharge capacity compared to the theoretical capacity is consistent with either deviation from the predicted potassiation pathway and/or sluggish kinetics that prevent K insertion in the active material. The shape of the voltage profile observed upon the charge of SnP3 is similar to the discharge curve, albeit at lower capacity, suggesting that potassiation may be only partially reversible and/or that side reactions occur on the first discharge. In contrast, the voltage profile for K removal from Sn4P3 is substantially more sloped, indicating that different phase transformations may occur during depotassiation. Voltage profiles for later cycles retain these features (Figure S1).

Figure 2

Voltage profiles of first-cycle discharge/charge for SnP3 (black line) and Sn4P3 (blue line) during galvanostatic cycling at C/200 and C/100, respectively.

To understand the possible phase transformations taking place during potassiation/depotassiation of the tin phosphide anodes, we first turn to DFT. DFT calculations describe the thermodynamic stability of several phases in the relevant composition space, many of which are not currently reported in crystal structure databases, computational or otherwise (indicated with an asterisk (*) in the following lists). The full predicted phase space is elaborated elsewhere;[38] in this work, the local environments from low-lying trial structures are used to provide insight into the short-range structural motifs formed during cycling and probed with SSNMR and XRD (vide infra).

No crystal structures for ternary K–Sn–P compounds exist in current databases. Our DFT calculations predict that KSn2P2, KSn3P3, KSnP, K5SnP3, and K8SnP4 are all stable phases that may be formed during potassiation or depotassiation of either Sn4P3 or SnP3 anodes (shown in Figure S2). For the K–P edge, each of KP15, KP7*, K3P11, K3P7*, K2P3, KP, and K5P4* are predicted to be stable. Most notably, calculations performed with the PBE functional also indicate that two possible charging endpoints, K4P3 and K3P, are both only metastable at finite temperatures; however, they both lie very close to the convex hull (+2 and +5 meV/atom, respectively). Applying instead the rSCAN functional,[39,40] K3P is found to be thermodynamically stable, yet lies almost precisely on the existing tie-line between K5P4 and K, indicating that this phase may not form unless cycled to very low voltages (0.01 V vs K+/K). Similar calculations for the K–Sn phase space indicate that there are three stable K–Sn phases, namely, K3Sn17 (which destabilizes the known K4Sn23 phase), K4Sn9, and KSn.

We note that while DFT-predicted thermodynamic stability is useful for screening individual trial structures, the formation of a given phase during electrochemical cycling is highly dependent on kinetics.

Experimental Characterization of Potassiation/Depotassiation Mechanisms of SnP3 and Sn4P3

Ex situ31P and 119Sn SSNMR were collected at multiple states of charge to directly measure the local structural environments present in tin phosphide anodes during potassiation/depotassiation (Figures and 4). Pristine SnP3 (Figure , black) exhibits a broad 31P signal ranging from 100 to −210 ppm, in addition to a sharp (fwhm ∼12 ppm) resonance centered at −20 ppm. Since R3̅m-SnP3 only contains one independent P site, the combination of broad and sharp lines for pristine SnP3 likely arises from small particle sizes (∼9 nm) observed in XRD after prolonged ball milling during synthesis (8 h) and/or amorphous SnP3 compounds. For nano-sized materials, atoms in the particle core often display a sharp resonance consistent with the bulk crystalline material, whereas the surface sites exhibit inhomogeneous broadening.[41−43] In the pristine state, 119Sn SSNMR does not show a resonance for SnP3 (Figure c, black), possibly due to long spin-lattice (T1) relaxation times, in combination with inhomogeneous line broadening, making detection prohibitively long[44] (the maximum recycle delay tested in these experiments was 1000 s, and no signal was observed after a frequency sweep from 6400 to −9300 ppm).

Figure 3

(a) Voltage profile of first discharge/charge of SnP3 under galvanostatic conditions at C/200. Colored points mark the states of charge at which cycling was stopped and SnP3 anodes were extracted for structural analysis. Ex situ (b) 31P and (c) 119Sn SSNMR of SnP3 anodes during the initial discharge (blue-hue spectra) and charge (red-hue spectra). The 31P quintet at −146 ppm (JP–F = 675 Hz) is assigned to residual PF6− from the electrolyte. Asterisks indicate spinning sidebands. Experiments were performed at either 18 or 28 kHz MAS frequency.

Figure 4

(a) Voltage profile of first discharge/charge of Sn4P3 under galvanostatic conditions at C/100. Colored points mark the states of charge at which cycling was stopped and Sn4P3 anodes were extracted for structural analysis. Ex situ (b) 31P and (c) 119Sn SSNMR of Sn4P3 anodes during the initial discharge (blue-hue spectra) and charge (red-hue spectra). The 31P quintet at −146 ppm (JP–F = 675 Hz) is assigned to residual PF6− from the electrolyte. All experiments were performed at 18 kHz MAS frequency.

After discharge of SnP3 to the center of the voltage plateau at 0.53 V vs K+/K (Figure , dark blue), a narrow 31P resonance at 2.6 ppm (fwhm ∼23 ppm) appears with a small shoulder at 11 ppm (these peaks are more clearly observed in Figure S6, which depicts the 31P spectra alone). The resonance at 11 ppm is consistent with elemental black P,[45] whereas the peak at 2.6 ppm agrees with the 31P shift predicted for the on-hull 63-KSnP ternary phase discovered with DFT (δiso = 2.3 ppm, Figure S10). Likewise, a broad 119Sn resonance at −66 ppm (fwhm ∼240 ppm) appears at the same voltage, supporting the formation of new Sn environments that may resemble KSnP-type sites. (N.B. Alkali metal insertion into SnP likely decreases T1 relaxation times,[46−48] allowing us to observe 119Sn resonances during potassiation). The lack of new Bragg reflections in XRD (Figure S3) indicates that crystalline KSnP does not form, but rather that K reaction with SnP3 results in local environments similar to those found in the ternary KSnP phase (where P is coordinated to three Sn and three K atoms and exhibits C3v symmetry).

In the fully discharged state at 0.06 V vs K+/K (Figure , light blue), we continue to observe the 31P peak corresponding to black P, as well as a new peak at −244.5 ppm; no change is observed in 119Sn NMR. The 31P resonance at −244.5 ppm is similar to that observed for both Na3P (δiso = −207 ppm)[45] and Li3P (δiso = −275 ppm).[49,50] Examination of the chemical shift anisotropy (CSA) associated with the 31P resonance at −244.5 ppm shows that there is very little anisotropy at this site, consistent with the formation of isolated P atoms, likely surrounded by K, similar to those expected for K3P (Figure S7, experimental: Δ = 61 ppm, η = 0.07; calculated: Δ = 56 ppm, η = 0.42). These data indicate that phase separation of elemental P enables a conversion reaction between P and K that produces K3P-like structures. Since there are no additional reflections observed in XRD at the end of discharge (Figure S3), we suspect that these K3P-like environments do not exhibit long-range order. Further, we do not observe metallic Sn at any state of charge (Figure c; Figure S8 shows 119Sn NMR of bulk Sn), indicating that only reactions of K with P and SnP lead to the observed capacity.

Ex situ31P and 119Sn NMR of the intermediates formed during potassiation of Sn4P3 are shown in Figure . The two resonances centered at 1050 and 850 ppm in the 31P SSNMR of the pristine sample (Figure b, black) are assigned to the two independent P sites (at the 6c and 3a positions, respectively) in R3̅m-Sn4P3.[49] Minor resonances in 31P and 119Sn NMR spectra are attributed to residual P (at ∼15 ppm) and Sn (center of mass at ∼6870 ppm) from synthesis. Upon discharge to the middle of the voltage plateau at 0.18 V vs K+/K (Figure , dark blue), 31P resonances at 2.6 and approximately 16 ppm are assigned to KSnP-like sites and phase-separated P, respectively. The presence of KSnP-like sites is again reflected in the 119Sn spectra by the resonance centered at −138 ppm; both 31P and 119Sn NMR spectroscopies are consistent with observations from SnP3 upon initial potassiation. Analysis of 119Sn NMR shows an increase in the NMR signal intensity for metallic Sn, indicating that Sn particles are produced during potassiation. The presence of metallic Sn is also confirmed with XRD, in which multiple small Sn reflections appear at 0.18 V (Figures S4 and S5). At full discharge to 0.06 V vs K+/K (Figure , light blue), K reacts with elemental P to form disordered K3P-type structures. During discharge/charge, the 119Sn NMR resonance for metallic Sn does not change in width, nor the center of mass, which strongly suggests that metallic Sn particles do not participate in further reactions. DFT and ex situ studies of 119Sn NMR during Li–Sn alloying reveal that isotropic 119Sn shifts are very sensitive to alloying (e.g., LiSn has two resonances at 5969 and 5429 ppm compared to a single resonance at 6915 ppm for pure β-Sn).[51,52] We would expect similar 119Sn NMR spectrum changes if K–Sn alloying were taking place.

Depotassiation Mechanisms of SnP3 and Sn4P3

Upon charge of SnP3 to 1.77 V vs K+/K (Figure , light red), the 31P resonance corresponding to the sites resembling K3P decreases considerably, while the relative ratio of elemental phosphorus (at 11 ppm) increases, consistent with K removal. At 2.00 V vs K+/K (Figure , dark red), the 31P NMR spectrum of fully charged SnP3 resembles that of the pristine material, as indicated by the shoulder at 40 ppm and the broad 31P NMR lineshape, but resonances corresponding to KSnP-type sites remain in both 31P and 119Sn NMR. The center of mass of the KSnP-type environment in the diamagnetic region of 119Sn NMR shifts to higher ppm (δ2.00 V – δ1.77 V = 150 ppm), likely due to changes in the stoichiometry of KSnP, some of which may regenerate SnP3-like sites. Partial reformation of SnP3 is also consistent with observations from XRD that show increased intensity for SnP3 reflections upon charging relative to the fully discharged electrode (Figure S3).

Upon depotassiation of Sn4P3 to 1.50 V vs K+/K (Figure , light red), K is readily removed from K3P-type environments, whereas KSnP-type sites persist, similar to SnP3. Again, K removal leads to small changes in the peak position of KSnP sites in 119Sn NMR and the emergence of a broad peak extending from 170 to −310 ppm in 31P NMR (center of mass = −110 ppm). DFT simulations predict several KSnP phases with 31P chemical shifts in this range, indicating that both Sn4P3 and SnP3 anodes generate highly disordered KSnP coordination environments, although these sites may overlap with inhomogeneous broadening observed for SnP3 particles. The chemical shift of metallic Sn in 119Sn NMR remains constant (Figure ) during charge, providing evidence that phase-segregated Sn does not alloy with K.

Discussion

The combination of SSNMR and DFT provides unique insight into the potassiation/depotassiation behavior of tin phosphide anodes. First, we assign a new ternary KSnP-type environment that forms upon discharge of the battery. The formation of KSnP-like coordination environments in both Sn4P3 and SnP3 is consistent with our predictions from DFT that indicate 63-KSnP is thermodynamically stable upon potassiation of both parent compounds. However, differences in stoichiometry and structure between the two tin phosphides suggest that each has a distinct potassiation mechanism that forms these environments. In Sn4P3, the two-dimensional structure allows for topotactic intercalation and coordination of K with the alternating rows of Sn and P along the c-axis, forming Sn and P coordination environments similar to those found in the KSnP-like sites in NMR. This process leads to displacement of elemental P and concurrent reduction of Sn, but only P undergoes a conversion reaction upon subsequent potassiation. In contrast, the formation of KSnP-like sites in the three-dimensional SnP3 structure requires cleavage of P–P bonds during reduction. P–P bond cleavage during K insertion results in the phase separation of P observed in 31P NMR, ultimately producing local 31P sites that can coordinate to three nearby Sn atoms (bond distances of 2.72, 2.97, and 3.77 Å) and K to produce the C3v symmetry about the P atom observed in KSnP. This mechanism resembles that proposed by Nazar and co-workers for Li insertion in MnP4, where reduction is facilitated by P–P bond breakage rather than redox chemistry associated with the metal center.[16] In both tin phosphides, residual KSnP-like sites are present at full charge. This suggests kinetic or transport limitations on K removal and presents a mechanism of capacity fade by which K is locked in the tin phosphide structure. We note that this degradation is in addition to electrode cracking observed in scanning electron microscopy (SEM) of both tin phosphide anodes (Figure S11).

The second major finding of this work is that metallic Sn is only generated in the case of Sn4P3, and not in SnP3 as previously suggested.[22] While this phase-separated Sn can, in principle, provide capacity in terms of K–Sn alloying (analogous to the behavior observed in tin phosphides for Li- and Na-ion batteries[49,53−57]), no evidence for the electrochemical reaction between K and Sn is seen in 119Sn NMR nor XRD. The lack of capacity from metallic Sn is consistent with a loss of electrical contact upon phase separation.[58] Metallic Sn is never reincorporated in the anode, which is consistent with displacement reactions observed in LIB systems producing electrically isolated metal particles upon conversion and displacement reactions (e.g., InSb,[59] Cu2Sb,[60] CoP3[17]). Thus, most of the observed capacity in both Sn4P3 and SnP3 is due to the formation of disordered sites resembling K3P and likely explains the rapid capacity loss in these systems. Efforts to enable alloying anodes in KIBs must consider this phase separation and find a way to tether the components to the electrode during potassiation/depotassiation.

Conclusions

The combination of SSNMR and DFT enabled the detection and assignment of short-range and amorphous intermediates previously invisible in diffraction analyses of tin phosphide anodes for KIBs. While the intention of binary phosphide anodes, such as tin phosphides, is to enable high capacity and mitigate volume expansion, we find that upon phase separation of elemental P (in both Sn4P3 and SnP3) and metallic Sn (in the case of Sn4P3), only P-containing compounds react reversibly with K. Local structures resembling KSnP persist after charge, suggesting transport and/or kinetic limitations for removing K from within the layered tin phosphides (regardless of starting stoichiometry). In the case of Sn4P3, the formation of electrically isolated Sn prevents total recovery of the pristine material and does not appear to add reversible capacity via alloying. The irreversible phase separation and structural rearrangements inherent to tin phosphides motivate new approaches to enabling P or binary phosphides as a high-capacity KIB anode material. We suspect that tin phosphides in KIBs may benefit from the approaches taken for Si–C anodes in LIBs,[61−63] where nanoparticles of the high-capacity anode material (for KIBs, P) are embedded in a graphitic matrix to accommodate volume expansion.