Mixing, Domains, and Fast Li-ion Dynamics in Ternary Li-Sb-Bi Battery Anode Alloys

Antimony and bismuth can both alloy with up to three molar equivalents of lithium and are therefore attractive candidates for replacing graphite in Li-ion battery anodes. Li3Sb and Li3Bi have the same cubic structure (Fm3 ̅m), but the ternary Li-Sb-Bi system has not been studied. We synthesized Li3(SbxBi1-x) with different Sb mole fractions at room temperature by ball-milling. These ternary alloys all have cubic crystal structures, as determined by XRD, but show a tendency towards phase segregation for x = 0.25 and 0.50. For x = 0.25, the lattice parameter presents a clear positive deviation from Vegard’s law in XRD, while for x = 0.50, XRD reveals two phases after milling, with the Bi-rich minority phase diminishing after thermal annealing. Solid-state nuclear magnetic resonance spectroscopy provides evidence for a Sb-enriched environment around the Li atoms for Li3Sb0.25Bi0.75, and nuclear spin-lattice relaxation measurements of the binary and ternary alloy phases point to low activation energies and rapid Li ion diffusion in Li3Bi.


INTRODUCTION
Antimony and its alloys have been intensively studied in recent years as high-capacity anode materials in alkali metal-ion based rechargeable batteries. 1 10 It has a theoretical storage capacity of 660 mAh/g, which is almost twice as high as graphite (372 mAh/g) that is currently used in commercial lithium-ion batteries (LIB), and double that of hard carbon in emerging sodium-ion batteries. 11 13 Recently, antimony-based anodes have been extended to include potassium-ion batteries. 14,15 The range of antimony-alkali compounds that has been characterized is, however, considerably broader than just these two elements, and can include up to three alkali metal ions from group I, from lithium to cesium, with the number of intermediate compounds increasing as one moves down the group of the periodic table. 16,17 In contrast to the more widely studied tin anodes, antimony shows good performance in conventional commercial carbonate-based electrolytes, making it a more suitable candidate to replace graphite in LIBs. 18 20 Bismuth is isostructural with antimony and the elements have full mutual solid solubility above 200 °C. 17 However, thermodynamic calculations suggest there is a miscibility gap below 140 °C. 21 Like antimony, bismuth also alloys with up to three molar equivalents of alkali metals from lithium to cesium. 16,17 Since bismuth is much more ductile than antimony, it makes the material less susceptible to fracture during electrochemical cycling with lithium, and there are several examples of bismuth addition to antimony improving the cycling stability relative to either element by itself. 2,8 Although in-situ diffraction data of the individual elements during lithiation are available and show that A3Bi and A3Sb (A = Li, Na, K) are isostructural, 15,22 25 no such data exist on binary Sb-Bi solid solutions. In addition, no ternary phase diagrams obtained via metallurgical routes have been reported. 17 Solid-state nuclear magnetic resonance (NMR) spectroscopy is a highly sensitive method to interrogate local atomic chemical environments and is therefore suitable to study subtle compositional variations on short length scales (< 10 Å) in a multicomponent alloy. Such socalled incipien phase segrega ion can be de ec ed b NMR more readil han b o her techniques that represent an average long-range crystalline structure, as is the case for X-ray diffraction (XRD). 26 31 The lithiation potential of antimony is approximately 200 mV more positive than that of bismuth, 8 indicating higher thermodynamic stability, which could result in the formation of Sb-rich clusters. Therefore, diffraction and NMR methods are complementary echniq es ha , oge her, pro ide a de ailed and n anced ie of a ma erial s s r c re. NMR also provides insightful information about dynamics, and the onset of ion dynamics within conductive materials, which is important information with regards to battery applications, and in particular rate-capability. 32 41 In the present work, we study a series of ternary Li3SbxBi1-x compounds using XRD and NMR spectroscopy. The Li-rich alloys were synthesized using thermal annealing, ball milling, or a combination thereof, rather than electrochemical lithiation in order to more accurately control the Li-to-metal stoichiometry and eliminate any quantitative errors or additional NMR signals from Li salts formed by electrolyte decomposition. 39,42 Antimony and bismuth show close-toideal solid solution behavior in binary alloys. Ternary Li3SbxBi1-x alloys show evidence of phase segregation for x = 0.25 and 0.50 (prior to annealing). A minority phase that is enriched in Sb is detected by NMR for x = 0.25. Our findings provide future directions for in-situ and operando diffraction experiments tracking phase transformations during lithiation of these binary metal Sb-Bi alloys.

EXPERIMENTAL SECTION
Antimony and bismuth metal powders were purchased from SigmaAldrich (St. Louis, MO).
Lithium foil was obtained from MTI Corporation (Richmond, CA).
The Sb-Bi solid solutions with compositions Sb0.25Bi0.75, Sb0.50Bi0.50 and Sb0.75Bi0.25 were prepared as starting materials by melt spinning followed by annealing. Melt spinning was performed on an Edmund Buehler GMBH SC10 single-roller melt spinner. Compacted powder pellets of the desired overall composition were loaded into a BN crucible and induction melted using a 5 kW HU5 RF generator. The melting process was monitored using a pyrometer. The melt was ejected at a temperature of 700 750 °C onto a 30 cm diameter copper wheel spinning at 40 Hz, corresponding to a surface velocity of 38 m/s. The resulting materials were not yet singlephase and were crushed in an agate mortar, milled for 30 minutes using 304 stainless steel milling tools at 500 rpm and annealed on a hotplate. The annealing temperature was set to the solidus temperature for that particular composition, ranging from 300 °C for Sb0.25Bi0.75 to 380 °C for Sb0.75Bi0.25, for 16 hours, under argon.
Lithiation was performed at room temperature, as it would be in a battery, by ball milling the powders with a stoichiometric amount of Li foil for 5 hours at 500 rpm on a TenCan XQM-2 planetary mill using 50 ml 304 stainless steel jars and balls. The Sb-Bi starting alloys were processed in 2 g batches using 20 g of balls. Reactants were weighed inside the glovebox on a Mettler Toledo balance with 1 mg readability. Elemental Bi turned out to be too ductile to be milled effectively with Li. To synthesize Li3Bi, a pressed pellet of Bi powder was wrapped in a stoichiometric amount of Li foil and fired on a hotplate set at 300 °C for 16 h in an alumina crucible covered with a quartz glass plate inside an Ar-filled glovebox.
Powder XRD was performed on an Ultima IV multipurpose diffractometer (Rigaku) with a 285 mm goniometer radius or on an AXS diffractometer (Discover 8, Bruker, Madison, WI) i h C K radia ion ( = 1.5406 ). The AXS diffrac ome er as eq ipped i h a odimensional general-area diffraction detection system (GADDs) using a Vantec-500 detector with a sample detector distance of 22 cm. To protect the lithiated materials from air and moisture, double-sided adhesive Kapton tape with a round cutout was laminated onto a glass slide. The powder was added into the compartment left by the cutout and then covered with a polypropylene foil, all while handling the powder under an inert Ar atmosphere. A photo of the resulting window is included as Figure S1.
Solid-state NMR experiments were performed on a Bruker 500 MHz (B0= 11.75 T) Avance NEO NMR spectrometer, equipped with a 4 mm double resonance HX magic-angle spinning (MAS) Bruker probe. Powdered samples were diluted (50% by mass with dry silica) and packed into 4 mm o.d. zirconia ceramic rotors and sealed with a Vespel ® cap inside an Ar glove box. 7 Li MAS NMR data were acquired using a Bloch decay pulse sequence with optimized pulse widths between 3.3 and 4.0 µs (calibrated for each sample, B1/2 = 71 kH (sol ion)), rec cle delays of 5 20 s, 16 co-added transients and a spinning frequency of 10 kHz. Variable temperature non-spinning 7 Li NMR using a Bloch decay and inversion recovery pulse sequences were used to measure the linewidth (fwhm) and nuclear spin-lattice relaxation (T1), respectively, between 203 and 333 K. Acquisition parameters were identical to those reported above. The 7 Li NMR spectra were referenced to 0 ppm using 1 M LiCl solution. Non-spinning 121 Sb NMR meas remen s ere acq ired sing a Hahn echo ( /2 acq.) pulse sequence, with 2 and 4 µs pulses ( B1/2 = 42 kH (sol ion)), be een 2k and 100k co-added transients and recycle delays of 0.25 1 s. For 121 Sb NMR spectra of Li2Sb and Li3SbxBi1-x compounds the variable offset cumulative spectrum (VOCS) 43 approach was applied using 50-100 kHz offsets and between 3 and 10 steps. The 121 Sb NMR spectra were referenced to 0 ppm using 0.5 M KSbF6 in acetonitrile. The non-spinning 209 Bi NMR spectrum of Li3Bi was collected using a Bloch decay p lse seq ence i h a /2 = 3.2 s, 128 co-added transients and a 1 s recycle delay. The 209 Bi NMR data was referenced to 0 ppm using a saturated solution of Bi(NO3)3 in HNO3 and D2O.

Preparation of SbxBi1-x solid solutions
Powder XRD patterns of the SbxBi1-x alloys in the as-prepared and annealed states are shown in Figure 1. The unlithiated starting alloys were prepared via melt spinning followed by thermal annealing. The cooling rate during melt spinning is insufficient to cross the liquidus-solidus gap vertically and produce a single-phase material, which is why a subsequent annealing step is required to form a single-phase compound. As illustrated on the right-hand side in Figure 1, all XRD patterns are shifted towards lower angles, i.e. larger d-spacing, as the Bi content is increased, which is expected based on the larger atomic radius (148 pm (Bi) vs 139 pm (Sb)) of this element. Although thermodynamic calculations point to the possible existence of a miscibility gap below 140 °C, 21 experimental evidence for this gap is difficult to interrogate experimentally as solid-state diffusion is very slow at these low temperatures. Although the samples described here ere no q enched in an a af er he annealing s ep, here is no evidence for phase separation and complete solid solubility appears to be achieved. As an additional illustration, the XRD patterns in Figure S2 show a narrower angular range where the peak positions for all three compositions fall between the Sb and Bi reference patterns. The unit cell volumes obtained from refinement of the diffraction patterns support a linear relationship between the unit cell volume and xSb (R 2 =0.99), as shown in Figure S3, indicating ideal solidsolution behavior in the binary Sb-Bi host materials.

Li3(SbxBi1-x) alloy preparation
The lithiated alloys were synthesized by taking the SbxBi1-x alloys, prepared via melt-spinning, ball milling and annealing, and then ball-milling them with Li metal. To synthesize Li3Bi, a Bi pellet was wrapped in Li foil and fired at 300 °C, because all the material was observed to stick to the milling balls and vial when ball milling was attempted. The powder XRD patterns for the Li3(SbxBi1-x) are shown in Figure 2. The lattice parameters, grain size and microstrain are summarized in Table 1. As is the case for the metals, Li3Bi has a larger lattice parameter and unit cell volume than Li3Sb and the peak positions of the intermediate compositions are clearly shifted from Li3Sb towards Li3Bi as the Bi content increases. For all compositions, it appears that a single-phase, face-centered-cubic (fcc) compound can be formed. Li3Sb is formed during the milling step between Sb powder and Li metal. The sample contains traces of Sb metal and Li 2 Sb, as confirmed by NMR (vide infra), indicating a slight under-stoichiometry due to weighing errors or selective cold welding of Li to the milling tools due to its ductility. The lattice parameter of Li3Sb appears to be slightly larger than that found in the literature, 6.58 vs. 6.56 Å, even though Li3Sb is considered to be a line compound. 17 Li3Bi has the largest grain size and lowest microstrain, which is not surprising since it is the only one in the series to have been synthesized by thermal annealing rather than ball milling. Although most Li- is shifted much closer to that of Li3Bi than expected based on the nominal composition, which is already somewhat obvious from the position of the (111) peak in Figure 2. The lattice parameters, grain size, and microstrain are summarized in Table 1 Figure 2 is that of the material after annealing at 350 °C for 16 hours. As shown in Figure S4, the as-milled material clearly consists of multiple fcc phases. At higher angles in particular, the peaks are very wide and appear to be composed of at least two separate peaks. Annealing narrows all the peaks considerably, indicative of grain growth, and the peaks of any secondary fcc phases clearly diminish in intensity. As highlighted in the inset of Figure 2 showing the (422) peaks, a small shoulder is still visible on the low-angle side for Li3(Sb0.50Bi0.50), but not for any of the other compositions. The lattice parameter of the secondary phase is close to that of Li3Bi. When attempting to synthesize ternary alloys at a higher temperature (400 °C), disproportionation of the material and segregation of high-temperature (hexagonal) Li3Sb were observed as shown in Figure S5 for Li3(Sb0.25Bi0.75). Therefore, 350 °C is probably close to the limit that these materials can withstand before decomposing into the respective binary compounds.
From the diffraction data, it appears that at 3 Li/(Sb+Bi), a single phase fcc material is thermodynamically favored for the full compositional range, although for x = 0.50, an annealing step was necessary. It is important to consider whether this behavior would be representative of Operando NMR measurements have also confirmed the formation of Li2Sb during lithiation. 39,42 When we attempted to synthesize pure Li2Sb by ball milling, the as-milled powder consisted of Sb metal, cubic Li3Sb, and only a trace amount of Li2Sb (see Figure S6). Incidentally, we observed that this phenomenon is not confined to the Li-Sb system Li15Si4 is formed when a Li-Si mixture is milled as well, as per Figure S7. Directly contacting Li metal and Sb is equivalent to short-circuiting a half-cell battery and setting the working electrode potential to 0 V vs. Li. As a result, the most concentrated lithiated phase is formed first and Li2Sb is obtained only after annealing, as shown in Figure S6 and Figure

Solid-state NMR spectroscopy
The diffraction analysis shows that crystalline, single-phase materials are formed through mechanochemical means across the compositional range, with the exception of Li3Sb0.50Bi0.50, which showed clear signs of phase separation that could not be entirely eliminated by annealing (see Figure S4). Though the diffraction method is an excellent tool to assess the average longrange structure of crystalline solids, it is insensitive to short-range variations in composition and/or lattice spacings, reported for several metal-hydrogen systems. 31 Furthermore, light elements such as lithium are harder to detect accurately using XRD because of their low number of electrons. However, NMR spectroscopy is sensitive to both structure (short and medium range) and dynamics in Li-containing materials; therefore, we used NMR spectroscopy to probe both the local chemical structure of Sb/Bi mixing and Li-ion dynamics in these Li3SbxBi1-x ternary alloys. Lithium-7 (I=3/2) is an attractive nucleus for NMR spectroscopy due to its high receptivity (92.4% N.A. and = 38.9%) and extremely small quadrupole moment (Q = -4.01 fm 2 ). These physical properties for 7 Li result in it behaving as a pseudo-nuclear spin-½ NMR nucleus, contributing insignificantly to second-order quadrupolar broadening of the central transition peak. 36 Figure 3   Thermodynamic or DFT calculations assessing the relative stability of a mixture of the binary compounds versus a ternary phase could shed further light on this, but none have been published so far. Only Li3Sb0.5Bi0.5 has been synthesized before by melting the constituent metals. The lattice parameter was slightly lower than the one we found (6.625 Å vs. 6.653 Å) 45 , and no traces of Li3Bi were reported, indicating that the synthesis route and temperature have a big influence on the outcome.   Table S1.

Li ion dynamics
Variable temperature non-spinning 7 Li NMR spectroscopy was undertaken to investigate the Li ion dynamic behavior in these compounds. Figure S9 shows the 7 Li NMR spectra obtained between 203 and 333 K. The changes in linewidth, nuclear spin-lattice relaxation and chemical shift are reported in Table S2. At room temperature, the 7 Li NMR linewidths are 0.7 to 1.5 kHz across the series, with Bi-rich materials being narrower (faster Li ion dynamics) than Sb-rich sister compounds. All compounds exhibit decreasing linewidths with increasing temperature.
Likewise, with increasing temperature (to 333 K) the 7 Li NMR linewidths decrease as the increased Li ion motion reduces the residual homonuclear 7 Li-7 Li dipolar coupling. The spinlattice relaxation for 7 Li is influenced by mobile Li ions, thus measuring the T1 relaxation parameters as a function of temperature allows the associated activation energy for Li dynamics for these materials to be determined using the Arrhenius equation. 49 51 In the temperature range studied here, all Sb-containing materials display a decrease in their spin-lattice relaxation time with increasing temperature, with all T1 values located on the low-temperature side of the T1 curve, as shown in Figure S10. Using these data, the Sb-containing materials were found to range from 32 to 75 meV, as shown in Table 2. In contrast to the Sb-containing phases, the T1 minimum for Li3Bi is found at 233 K, revealing that Li ions are moving rapidly within this phase where the correlation time for Li (T1 minimum) can be obtained and an activation energy of 75 ± 5 meV is determined from the more accurate high-temperature side of the T1 curve. The interplay between the fast and slow-motion limits reveals that at 233 K, Li ion diffusion is occurring at approximately the 7 Li nuclear Larmor frequency (ns timescale, ~0.8 ns), increasing at higher temperatures approaching the value reported for fast solid ionic conductors, consistent with the narrow Lorentzian lineshape observed above. The associated Li3Bi activation energy agrees well with prior conductance measurements (Ea < 100 meV), whereas the NMR results appear to underestimate Li3Sb and associated mixtures which is attributed to the inability to reach high temperature side of the T1 curve (due to hardware limitations) resulting in an underestimation of activation energy. 52 The discrepancy for Li3Bi may be attributed to how the techniques respond to Li ion diffusion: NMR spectroscopy is sensitive to medium-and short-range structure, while conductivity measurement methods respond to bulk solid properties, probing long-range macroscopic dynamics whereby grain boundaries, defects or compositional/structural variations, for example that can influence the measurements. For example, an early NMR study provided evidence of lithium ion dynamics within Li3Sb, 53 while lithium diffusion and conductivities were reported on bulk Li3Sb and Li3Bi materials using the galvanostatic intermittent titration technique (GITT). 52,54 Using GITT, the authors demonstrated that variations in activation energy (200-300 meV for Li3Sb, and 130 meV for Li3Bi) were due to small deviations from the 3:1 stoichiometric ratio for the materials that impacted the measured activation energy. 54 Therefore, the new synthetic approach (mechanochemical synthesis) used to provide single-phase crystalline materials that exhibit varying degrees of local order, may contribute to the variations in activation energies. Recently, this phenomenon has also been observed in other ionic conducting materials. 55 58 We also caution that values determined from the GITT method date back nearly four decades and it has since been suggested that the approach may provide spurious results. 59

CONCLUSIONS
The long, and short range order of the Li3(SbxBi1-x) alloys were studied using X-ray diffraction and NMR spectroscopy, respectively. For x = 0.25, the fcc unit cell volume determined by XRD

ASSOCIATED CONTENT Supporting Information
Photo of air-tight PP-covered Kapton windows used for XRD, additional diffraction patterns related to Sb-Bi and Li-Sb-Bi alloy synthesis. 121 Sb NMR spectrum of Li2Sb compound and static variable temperature 7 Li NMR spectra of all compounds. Tables with chemical shifts and line widths for MAS and static 7 Li, 121 Sb and 208 Bi NMR spectra at room temperature and 7 Li at variable temperature.        Figure S7. XRD pattern of as-milled Li9Si4. Despite the overall composition being close to Li7Si3 stoichiometry, which is a stable intermetallic according to the equilibrium phase diagram, the only crystalline phases present are elemental Si and the metastable Li15Si4 phase, 2 underscoring the tendency to preferably form Li-rich phases during milling. Figure S8. Non-spinning 121 Sb NMR spectrum (black) and simulation (blue) of Li2Sb (B0 = 11.7 T; T = 293 K). * refers to a Li3Sb impurity. Figure S9. Non-spinning 7 Li variable temperature (B0 = 11.75 T) NMR of Li3SbxBi1−x alloys. Figure S10. Temperature dependent 7 Li NMR nuclear spin-lattice relaxation rates (1/T1) of the binary and ternary Li3(SbxBi1-x) alloy series. Table S1. Summary of 7 Li MAS NMR and non-spinning 7 Li, 121 Sb and 209 Bi (non-spinning) NMR results for lithium alloyed compounds (B0 = 11.7 T, T= 293 K).