Reaction Mechanism of the Sn2Fe Anode in Lithium-Ion Batteries.

Sn2Fe anode materials were synthesized by a solvothermal route, and their electrochemical performance and reaction mechanism were evaluated. The structural evolution in the first two lithium cycles was investigated by X-ray absorption spectroscopy (XAS), synchrotron X-ray diffraction (XRD), and magnetic studies. In the first cycle, progressive alloying of Sn with Li accompanied by metallic iron displacement occurs upon lithiation, and the delithiation proceeds by Li x Sn dealloying and recovery of the Sn2Fe phase. In the second cycle, both XRD and XAS identify Li-Sn alloying at earlier lithiation stages than in the first cycle, with low-Li-content alloys evident in the beginning of the lithiation process. In the fully lithiated state, XAS analysis reveals higher coordination numbers in both the Li x Sn and Fe phases, which points toward more complete reaction and higher crystallinity of the products. Upon second delithiation, the Sn2Fe phase is generally reformed as evidenced by XRD. However, XAS indicates somewhat reduced Sn-Fe coordination and shorter Fe-Fe distance, which indicates incomplete reconversion and metallic Fe retention, which is also evident in the magnetic studies. Thus, a combination of long-range (XRD, magnetic) and local (XAS) techniques has revealed differences between the first and the second Li cycles relevant to the understanding of the capacity fading mechanisms.

Introduction

Lithium-ion batteries have been successfully applied in various fields, such as portable electronic devices, medical devices, electric and hybrid electric vehicles (EV and HEV), and others.[1,2] Currently graphite is widely used as an anode for lithium-ion batteries due to its low cost, stability, and decent cyclability. However, the low operating voltage of graphite, which is close to the lithium-plating voltage, potentially causing safety issues,[3] and its limited theoretical capacity of 372 mAh/g makes researchers look for safer and higher capacity anode candidates. Among such candidates are Si (4200 mAh/g),[4,5] Sn (993 mAh/g),[6] and Ge (1620 mAh/g).[7] A Sn-based anode is one of the most promising candidates because of its high capacity, high packing density, and safe working voltage.[8] Despite a higher theoretical capacity of the Sn-based anode as compared with a graphite anode, its practical use is hindered by the huge volume change (about 260%)[9] of Sn during lithiation and delithiation. This volume change results in the building and destruction of the solid electrolyte interphase (SEI) layer on each cycle, which, in turn, causes a continuous increase in the cell impedance and subsequent fast capacity failure.[10] Generally, there are two strategies to mitigate this volume change. One is to downsize the particles to nanoscale, which helps to release the stress on the particles when the volume change occurs and, thus, improve the cycling performance.[11,12] Downsizing to nanoscale also shortens the lithium diffusion path.[13] However, associated issues such as low tap density, high surface reactivity, as well as the flammable or explosive tendency should not be neglected.[14] The other strategy is to make Sn–M alloys or Sn–M–C composite materials, in which M can be either electrochemically active or inactive, such as Fe,[15−19] Co,[20,21] Mn,[22] Ni,[23,24] Cu,[25−27] Ti,[28] and so on.[29,30] The M component can act as a matrix to accommodate the volume change during cycling and hold the integrity of the active material, so that the capacity retention can be improved. Enhanced electronic conductivity and prevention of Sn particles’ aggregation are other benefits of introducing M to form Sn–M or Sn–M–C composites.[26,29] The first successful commercialization of a Sn-based anode was in the SONY Nexelion battery first released in 2004.[31] Analysis of the Nexelion Sn anode showed that it contained an essentially nanosize amorphous CoSn alloy embedded in carbon[32] and could deliver a reversible capacity of more than 500 mAh/g at 1 mA/cm2 for over 30 cycles,[33] significantly higher than the 350 mAh/g of graphite-based anodes. This success reignited interest in related materials as it is necessary to replace the expensive and toxic Co by other elements. Therefore, we targeted Sn–Fe-based alloys because Fe is a cost-effective, earth-abundant, and environmentally benign material, and it does not react with lithium, hence it can provide an inactive cushion effect.[18,19]

Among all of the reported Sn–Fe alloys, Sn2Fe is the most stable Sn-rich phase at room temperature according to equilibrium binary phase diagrams,[34] and it also exhibits the highest reversible capacity.[16] With respect to the reaction mechanism of the Sn2Fe anode, it has been reported that Li–Sn alloys are formed during lithiation while Fe is being extruded, even though Fe is hard to be detected.[16,35] However, there is a debate for the delithiation process, for which some researchers claim that Sn2Fe is reformed upon delithiation, while others argue contrarily that the “liberated” Fe particles remain inert so that Sn2Fe could not be recreated. For example, Chamas et al.[36] claim that the first discharge could be considered as an irreversible transformation from Sn2Fe into a α-Fe/Li7Sn2 nanocomposite by combining operando Sn Mössbauer spectroscopy and ex situ magnetic measurements, while the first charge process is a progressive delithation of Li7Sn2 and a back reaction of poorly lithiated Li–Sn phases with the iron nanoparticle generated at the first discharge. Yoon and co-workers propose that Sn2Fe is decomposed and Li4.4Sn is formed on reaction with lithium; the reaction is reversed during lithium removal.[35] They also point out that the second cycle is similar to the first cycle by X-ray diffraction data.[35] However, no iron phase was clearly identified in their results. Mao et al.[37,38] claim that the reaction 8.8Li + Sn2Fe → 2Li4.4Sn + Fe occurs during the first discharge, and alloying/dealloying of Li with Sn was the primary reaction in the subsequent charge–discharge cycles. They also claim that the “rejected” Fe is inert so that Sn2Fe could not be reformed during charge.[37,38] In another paper, they point out that crystalline Sn or Sn2Fe could not be detected by X-ray diffraction, but a singlet similar to that of ultrafine Sn2Fe was found in Mössbauer spectra, which probably lead to the conclusion that Sn2Fe could be reformed during the charging process.[15] Moreover, they observe peak shifts of Li4.4Sn but no iron showing in the diffraction pattern, which is explained by the small solubility of Fe in the Li–Sn alloy.[15] They have also studied the reaction mechanism of other Sn–Fe alloys such as SnFe and attributed the reason that liberated Fe was not found to small grain effect or overlapping with SnFe peaks.[16] As for the formation of Fe during this electrode reaction, Nwokeke and co-workers have detected superparamagnetic iron (and/or tin-doped iron) nanoparticles during discharge by both electron paramagnetic resonance (EPR) and Mössbauer spectroscopy, which they thought would be preserved even after the reverse charge process.[39]

To achieve a comprehensive understanding of the reaction mechanism of this material, Sn2Fe was prepared solvothermally. The solvothermal method was chosen due to its low energy consumption and scalability and to avoid the high-energy ball-milling step necessary to achieve small particle size in the high-temperature synthesized alloy, thus eliminating the impurities introduced by the ball-milling medium. The reaction mechanism of this solvothermally synthesized Sn2Fe during the first two cycles has been thoroughly investigated through local and long-range characterization techniques such as X-ray absorption spectroscopy (XAS), powder X-ray diffraction (XRD), and magnetic studies.

Experimental Section

Sn2Fe was synthesized via the solvothermal method modified from a previous report.[40] SnCl2 (99%, Sigma-Aldrich) and FeCl3 (anhydrous, Sigma-Aldrich) with a molar ratio of 2:1 were put into a 125 mL Teflon-lined autoclave with 80 mL of ethanol. After stirring the as-prepared suspension for 1 h, a sufficient amount of NaBH4 (99%, Fisher Scientific) was added. The autoclave was then sealed and heated to 200 °C with a heating rate of 5 °C/min. After keeping at 200 °C for 20 h, the autoclave was naturally cooled down to room temperature. The obtained precipitate was washed by deionized water and ethanol for five times and then dried in the vacuum oven at 80 °C overnight. To investigate the influence of the precursor’s ratio on the resulting product as well as the cycling performances, different ratios (5:1 and 10:1) between SnCl2 and FeCl3 were used.

The material’s morphology was studied by a Zeiss Supra 55 VP field emission scanning electron microscope (SEM) operating at 5 kV. The phase composition was initially determined by powder X-ray diffraction (XRD) using a Scintag XDS2000 θ-θ diffractometer equipped with a Ge(Li) solid-state detector and Cu Kα sealed tube (λ = 1.54178 Å). The data were collected in the range of 2θ = 10–80° with a step size of 0.02° while spinning the sample to minimize preferred orientation.

For the electrode preparation, 80 wt % active material, 10 wt % carbon black, and 10 wt % poly(vinylidene fluoride) (PVDF) binder were mixed with an appropriate amount of N-methyl-2-pyrrolidone (NMP) solvent to form a slurry. The obtained slurry was spread onto the copper foil by a doctor-blade and then dried in the vacuum oven at 80 °C overnight. The electrodes (each with ∼5 mg of the active material) were assembled into 2325-type coin cells in a He-filled glovebox with a lithium foil (Aldrich, thickness 0.38 mm) as the counter electrodes and Celgard 2400 separator (Hoechst Celanese). The electrolyte was 1 M lithium hexafluorophosphate (LiPF6) dissolved in ethylene carbonate (EC) and dimethyl carbonate (DMC) with a volume ratio of 1:1 and 6% fluoroethylene carbonate (FEC) as the additive. The electrochemical performance was tested on VMP multichannel potentiostat (Biologic). The galvanostatic cycling test was performed at various current density within a voltage range of 0.01–1.5 V. The first cycle was performed at 0.12 mA/cm2, while the following cycles were at 0.20 mA/cm2. For ex situ electrode preparation, the cells cycled to different lithiated/delithiated stages were stopped, and the electrodes loaded with ∼5 mg of actives materials were taken out. Further sample preparation was all done in an Ar-filled glovebox. All of the electrochemistry for the ex situ studies were done using 0.12 mA/cm2 current density.

For synchrotron XRD, the powder samples from as-prepared electrodes were scraped off, filled in separate capillaries, and characterized at the beamline 17BM (wavelength 0.728 Å) at Advanced Photon Source (APS), Argonne National Laboratory (ANL). Data refinement and analysis was done with General Structure Analysis System (GSAS).[41,42]

For synchrotron XAS, the as-prepared electrodes were press-sealed between thin layers of a Kapton film and stored in the glovebox prior to measurement. The experiments were performed at beamlines X18A, X18B, X19A at the National Synchrotron Light Source (NSLS), Brookhaven National Laboratory and 5BM, 20BM at the Advanced Photon Sources of Argonne National Laboratory. XAS data were collected at both Fe K-edge (7712 eV) and Sn K-edge (29 200 eV), with the respective metal foils (Fe and Sn) measured in the reference mode simultaneously for the X-ray energy calibration and data alignment at each absorption edge. Data processing and analysis were conducted by using the IFEFFIT package.[43] By using the Athena program,[44] all raw spectra were aligned and averaged, which was followed by normalization and background-removal. For the theoretical analysis of the extended region of absorption data (EXAFS), the passive electron reduction factors, So2, were obtained by fits to the reference foils as 0.77 for Fe and 0.87 for Sn and fixed in the analysis of the sample. The simultaneous analysis of Fe and Sn K-edges EXAFS was employed by fitting theoretical FEFF6 signals to the experimental EXAFS in the r-space. Several parameters describing the electronic properties (e.g., correction to the photoelectron energy origin, ΔE) and local structural environment (coordination numbers (N), bond lengths (R), and mean squared disorder parameter σ2) around absorbing atoms were varied in the fits. Physical reasonable constraints (RSn–Fe = RFe–Sn and σSn–Fe2 = σFe–Sn2) were applied to accurately associate the structure information around Fe and Sn.

Magnetic measurements were performed on electrodes cycled to different stages. For this experiment, the active materials from the electrodes at different states of charge were scraped into plastic capsules inside the glovebox and sealed with vacuum grease to prevent air exposure. A SQUID magnetometer (Quantum Design MPMS XL-5) was employed to investigate the magnetic properties using the following protocol. First, the remnant magnetic field was quenched to less than 3 mOe using the ultralow field option, the sample was cooled to 2 K, and at that temperature, the magnetic field of 10 Oe was applied. Zero-field-cooled (ZFC) magnetization was measured while heating the sample from 2 to 400 K, followed by field-cooled (FC) magnetization measurements in the same field cooling of the sample from 400 to 2 K. Magnetization curves were measured at 2 and 298 K in magnetic fields up to 5 T. The sample was zero-field-cooled before the magnetization data at 2 K was taken.

Results and Discussion

We have first investigated the synthesis products obtained with different precursor ratios as these could lead to phases or composites with advantageous electrochemical performances. It can be noticed from the X-ray diffraction patterns (Figure a) that when the initial molar ratio between Sn and Fe is 2:1, mostly Sn2Fe diffraction peaks appear with a small amount of crystalline impurities. Once the initial molar ratio of Sn/Fe increases (e.g., from 2:1 to 5:1), the obtained final product becomes a mixture of Sn and Sn2Fe. The more Sn in the precursor, the higher the percentage of Sn is found in the final product (Figure a). The precursor ratio also affects the product morphology (Figure ). Although, all of the products are composed of primary spherical particles of ∼100 nm in diameter, as the initial Sn/Fe ratio increases from 2:1 to 10:1, more and more agglomeration occurs through smearing the particles’ boundaries. As for the electrochemical performance, the capacity retention is found to drop dramatically upon cycling when the excessive Sn is present (Figure b), which can be associated with the particle agglomeration and with well-known capacity fading of the pure Sn metal.[10,45] Thus, we have chosen a 2:1 product for further structural investigation.

Figure 1

(a) XRD patterns and (b) electrochemical performance (current density of 0.12 mA/cm2 for the first cycle and 0.20 mA/cm2 for the following cycles) of solvothermal Sn2Fe anode materials synthesized with different initial molar ratios between Sn and Fe.

Figure 2

SEM images of solvothermal Sn2Fe anode materials synthesized with different initial Sn/Fe molar ratios: (a) 2:1, (b) 5:1, and (c) 10:1.

First, we have taken a high-resolution X-ray diffraction pattern at APS beamline 17BM (wavelength = 0.728 Å) and performed Rietveld refinement (Figure ). It revealed the expected Sn2Fe phase, space group I4/mcm, lattice parameters a = 6.532(1) Å, c = 5.321(1) Å, V = 281 Å3, consistent with previous reports,[22] along with the small amount of the SnFe phase. Also, a small shoulder is observed at the (130) peak of Sn2Fe (around 20.6° in Figure ). SnFe peaks are expected in this area but do not match exactly with the shoulder position. The more likely candidate is Fe, the most intense (011) diffraction peak of which is close to the (130) peak of Sn2Fe.

Figure 3

(a) High-resolution XRD patterns (wavelength = 0.728 Å) of solvothermal Sn2Fe and (b) the Sn2Fe structure with selected interatomic distances indicated in ångström. Thin gray lines represent a group of eight (4 + 4) Sn–Sn distances of 3.392 and 3.467 Å.

We have performed magnetic studies of the sample and indeed found a behavior atypical of Sn2Fe, which is a collinear antiferromagnet with the Neel temperature TN = 384 K.[46,47] In the ordered state, the magnetic moments of the Fe atoms in Sn2Fe are aligned ferromagnetically in the chains running in z-directions (vertical in Figure b), but the neighboring chains are aligned antiferromagnetically, so that no net magnetic moment is expected. Instead, we found a hysteresis loop typical of ferromagnets, FC and ZFC curves departed already at 400 K, the highest temperature available in our system, no signs of antiferromagnetic ordering at 384 K and susceptibility values significantly exceeding those reported for Sn2Fe (Figure ). This clearly indicates the presence of a ferromagnetic Fe, since the other possible phase, SnFe, is antiferromagnetic.[48] The amount of Fe estimated from the saturation magnetization is about 2 wt %, consistent with the size of the shoulder observed in the high-resolution XRD pattern. Another interesting feature is the magnetization drop below 4 K observed in FC and ZFC curves for some samples, which is attributed to the presence of a small amount of Sn metal undergoing a superconducting transition.[49] The presence of both Fe and Sn impurities in the final product indicates that the formation of Sn2Fe was incomplete, and small quantities of metals formed by the reduction did not form the alloy.

Figure 4

(a) Magnetization of the Sn2Fe sample at 2 K and (b) field-cooled and zero-field-cooled dependences of magnetization.

The structure of the hydrothermal product was further investigated by the X-ray absorption (XAS) technique, through its two modifications, X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS), as these techniques provide local structural information critical in further reaction mechanism studies. The edge-step normalized and background-subtracted EXAFS data in the r-space for the pristine material measured at Fe and Sn K-edges is presented in Figure along with fitting curves. Sn EXAFS shows prominent scattering signals for up to 4 Å from the Sn scattering center, while Fe EXAFS oscillations diminish past 3 Å. This is consistent with the Sn2Fe structure, where Sn is surrounded by four Fe atoms at 2.789 Å, three (1 + 2) Sn atoms at 2.977 and 3.126 Å, and eight (4 + 4) Sn atoms at 3.392 and 3.467 Å (Figure ). Fe, on the other hand, is surrounded by two Fe (at 2.660 Å) and eight Sn (at 2.789 Å) atoms within 3 Å, and the next coordination shell is at more than 4 Å distance. The best fit performed simultaneously at both edges using the FEFF6 code shows that the most prominent features of the Sn and Fe EXAFS data can be adequately described using the coordination distances mentioned above. For the Sn K-edge EXAFS, an interaction between Sn and Fe contributes to the first nearest coordination shell (peak “α”), and the bond distance is calculated to be at 2.770(7) Å (Table ). The Sn–Sn bonding at 3.11(2) Å was used to account for the “β” peak, and a longer Sn–Sn bond (3.43(1) Å) characterizes the third peak (“γ”) depicted in Sn FT spectra (Figure a). The respective coordination numbers, in the order of bond length, are found to be 4.1(4), 5(3), and 9(1). In this fit, we did not fix the coordination numbers to those of Sn2Fe due to the presence of multiple phases in the sample. Nevertheless, the results agree very well with the coordination numbers of 4, (1 + 2), and (4 + 4) for the nearest three coordination shells in the structure of the Sn2Fe alloy confirming that it is the major phase (Figure ). Iron–tin interaction dominates the nearest coordination around the iron, evidenced by the drastic contrast between calculated coordination numbers: NFe–Sn = 5.5(6) and NFe–Fe = 0.8(6), which agrees well with the 8:2 ratio between Sn and Fe in the first coordination shell of the Sn2Fe structure. The fitting results reveal the contractions of the nearest Fe–Fe (2.62(3) Å) and Fe–Sn (2.770(7) Å) distances relative to the theoretical ones in the crystalline Sn2Fe (2.660 and 2.789 Å, respectively). Two of the shortest Fe–Fe distances correspond to the c-lattice parameter in the Sn2Fe structure, but XRD data does not show contraction with respect to the reported values. Thus, we attribute the shorter Fe–Fe distance from EXAFS to the admixture of signals from Fe and SnFe phases, both with shorter Fe–Fe distances, and evidenced by XRD. In contrast, the EXAFS-derived first Sn–Sn bond (3.11(2) Å) is slightly longer, compared to the theoretical average (∼3.08 Å) in the crystal Sn2Fe, which may be attributed to the contribution of Sn impurity found in the magnetic data. Such a multiphase structure is also supported by the large uncertainty associated with the coordination number, i.e., NSn–Sn = 5(3).

Figure 5

Fourier transform magnitude of EXAFS data (black) and nearest shell fit (red) for pristine Sn2Fe at (a) Sn and (b) Fe edge plotted together with individual coordinations.

Table 1

Structure Parameters Obtained by EXAFS Analysis of the Sn2Fe Anode Material at Various Lithiation/Delithation Stages

sample	pristine (A)	first lithiated to 0.12 V	first fully lithiated to 0.01 V (F)	first fully delithiated to 1.5 V (K)	second lithiated to 0.366 V (L)	second fully lithiated to 0.01 V (P)	second fully delithiated to 1.5 V (U)	Sn2Fe theory	Sn foil	Fe foil
NFe–Fe	0.8(6)	3.0(7)	3.2(3)	0.8(2)	4.0(8)	5.5(1.8)	1.4(5)	2		8
NFe–Sn	5.5(6)	3.6(3)	1.6(2)	4.8(2)	2.1(4)	1.3(7)	4.5(4)	8
NSn–Sn	5.4(2.7)	8.3(5.7)	1.0(5)	6.7(2.1)	1.4(3)		6.5(1.6)	1 + 2	4 + 2
NSn–Fe	4.1(4)	2.9(4)	0.5(4)	3.6(3)	0.6(2)	0.10(9)	3.0(4)
NSn–Li			5.6(1.7)		4.3(4)	6.5(5)
RFe–Fe (Å)	2.62(3)	2.515(6)	2.477(6)	2.537(6)	2.476(8)	2.47(2)	2.49(1)	2.660		2.470(3)
RFe–Sn (Å)	2.770(7)	2.744(5)	2.69(1)	2.747(5)	2.68(1)	2.64(3)	2.752(4)	2.789
RSn–Sn (Å)	3.11(2)	3.07(4)	2.92(3)	3.09(2)	2.98(1)		3.09(2)	2.977	3.011(4)
								3.126	3.017(4)
RSn–Fe (Å)	2.770(7)	2.744(5)	2.69(1)	2.747(5)	2.68(1)	2.64(3)	2.752(4)	2.789
RSn–Li (Å)			2.86(5)		2.92(2)	2.88(1)
σFe–Fe2 (Å2)	0.005(5)	0.014(3)	0.007(1)	0.001(2)	0.009(2)	0.010(4)	0.006(3)			0.0049(4)
σFe–Sn2 (Å2)	0.0081(9)	0.0093(8)	0.007(2)	0.0108(8)	0.007(3)	0.004(5)	0.010(1)
σSn–Sn2 (Å2)	0.019(8)	0.028(15)	0.011(7)	0.021(6)	0.014(3)		0.021(5)		0.0096(7)
									0.010(2)
σSn–Fe2 (Å2)	0.0081(9)	0.0093(8)	0.007(2)	0.0108(8)	0.007(3)	0.004(5)	0.010(1)
σSn–Li2 (Å2)			0.011(14)		0.003(2)	0.007(3)	0.015(3)
R, %	0.82	0.99	1.64	0.70	0.25	0.55	0.92		0.27	0.94

For the reaction mechanism studies, the ex situ samples at different lithiation states were taken out of the coin cells stopped at different voltages as indicated on the cycling curves below in Figure . The associated potentials are listed in Table . The first and the second cycles are noticeably different, as a huge irreversible capacity of about 400 mAh/g is observed in the first cycle, which is attributed to side reactions. Here, we will compare the phase evolution in the first and the second cycles.

Figure 6

Charge states of ex situ solvothermal Sn2Fe samples for (a) the first cycle and (b) the second cycle.

Table 2

Stopping Potentials of ex Situ Samples in the First and Second Cycles

sample first/second cycle	A	B/L	C/M	D/N	E/O	F/P	J/Q	H/R	I/S	J/T	K/U
voltage (V)	3.102	0.366	0.125	0.105	0.063	0.010	0.395	0.518	0.585	0.693	1.500

To reveal the phase changes during Li cycling, high-resolution synchrotron X-ray diffraction measurements have been carried out at APS-17BM (wavelength = 0.728 Å) on the ex situ powder samples. During the first discharge (Li insertion), the peak intensity of Sn2Fe decreases and almost disappears at the full lithiation stage. Meanwhile, the formation of Li–Sn alloys, which contribute to the broad peaks at 2θ = ∼11° and ∼18.2°, is observed (Figure a). The XRD patterns also show that the formation of Li–Sn alloys undergoes a continuous phase-evolution process, progressing from low-lithium-content (such as LiSn) to high-lithium-content Li–Sn phases. As shown in Figure b, the peak shoulder, which appears around 2θ = 10–10.5°, keeps increasing upon lithiation and becomes most pronounced in the fully lithiated state; this shoulder is contributed by the high-lithium-content Li–Sn phases of Li3.5Sn and Li4.4Sn. Formation of the Li4.4Sn phase in the first cycle is, however, questionable, as the irreversible capacity of 400 mAh/g suggests that the full lithiation may not be achieved. The Li3.5Sn phase was observed as the first cycle end lithiation product by Chamas et al. using Mössbauer data.[36]

Figure 7

Synchrotron XRD patterns (wavelength = 0.728 Å) of solvothermal Sn2Fe ex situ samples for (a) the first cycle stopped at different voltages indicated by each curve and (b) it’s expanded view in which Li–Sn alloys can be clearly seen.

As mentioned in the Introduction section, although it is commonly believed that the Fe phase should be extruded from Sn2Fe during the lithiation, a clear XRD evidence for the Fe formation is still missing, probably because of the small particle size of Fe as well as the limited resolution of the lab X-ray diffractometers. Owing to the exceptional high-resolution capability of synchrotron X-ray diffraction, we can attempt here to find such XRD evidence in our data. As shown in Figure a, the XRD peak at 2θ = 20.6° attributed to bcc α-Fe splits more noticeably from the nearby Sn2Fe peak upon lithiation, until it becomes a separate peak at points E and F. The peak area, which is comparable to that of the shoulder in the pristine sample and its sharpness, suggests that it most likely belongs to the inert crystalline iron originally present in the sample. On the other hand, a broad amorphous background develops in the XRD pattern between 20 and 22° toward the end of the lithiation, which might indicate the formation of Fe nanoparticles. Further proofs of the formation of bcc α-Fe from XAS and magnetic analyses will be discussed later.

Figure 8

Expanded views of ex situ synchrotron XRD patterns (wavelength = 0.728 Å) of solvothermal Sn2Fe for (a) the first lithiation process and (b) the first delithiation process stopped at different voltages indicated by each curve in which Fe phase can be clearly seen.

Compared to the commonly accepted lithiation mechanism, the Sn2Fe’s delithiation process is a subject of debate, in which some researchers claim that the Fe particles formed upon lithiation would remain inert so that Sn2Fe could not be recreated, while others argue that Sn2Fe could be reproduced upon delithiation.[15,35,36,38,39] Our XRD results (Figure a) show that the lithium removal process is a continuous phase-evolution dealloying process: first, high-lithium-content Li–Sn phases of Li3.5Sn and possibly Li4.4Sn are delithiated, associated with the disappearing of the characteristic peak shoulder at 2θ = 10–10.5° (sample F to K); second, most remaining Li–Sn alloys (corresponding to two broad peaks of 2θ = ∼11° and ∼18.2°) are gone when the delithiation voltage is over 0.585 V (samples I and J in Figure a), and meanwhile the Sn2Fe phase starts to reform. Finally, in the charged state, the peaks of Sn2Fe are recovered. Rietveld refinement performed at the beginning and at the end of the cycle, where the amount and crystallinity of the Sn2Fe phase allow for the lattice parameter determination, shows only small variations of lattice parameters (Table ). The peak broadening observed in cycled samples makes it difficult to determine the lattice parameters with high precision, therefore local structural details will be revealed using the XAS technique. Also, as shown in Figure b, the broad XRD feature of formed bcc α-Fe also keeps decreasing upon delithiation, but with a small portion of Fe being inactive, thus remaining in the final product (notice the shoulder at 2θ = 20.6° on the XRD pattern of sample K).

Table 3

Lattice Parameters of Sn2Fe at Various States of Charge in the First Cycle

sample	a (Å)	b (Å)	c (Å)	V (Å3)	Rp	Rwp
A	6.532	6.532	5.321	227.03	0.0554	0.0847
B	6.531	6.531	5.321	226.92	0.0706	0.0913
K	6.519	6.519	5.331	226.56	0.0931	0.1130
U	6.533	6.533	5.345	228.11	0.0638	0.0780

To reveal the local structural details of the first lithium cycle,we further studied the samples at the same states of charge as indicated in Table by the X-ray absorption spectroscopy technique. The selected edge-step normalized and background-subtracted EXAFS data in k- and r-spaces for the samples measured at Fe and Sn K-edges are presented in Figure , where the data for the fully lithiated sample F and fully delithiated sample K obtained from the first cycle are compared with the pristine sample. The delithiated sample resembles the pristine material in structure, whereas the atoms are arranged differently in the lithiated one. This is evidenced by its distinct EXAFS oscillations in the k-space and radial distribution structures at each absorption edge. Such findings confirm that the structural transformation of the pristine material is mostly reversible upon delithiation. The lower amplitude of FT-EXAFS peaks exhibited by the delithiated sample, compared with that in the pristine sample, indicates a reduction of alloy’s particle size after the first cycle.

Figure 9

EXAFS data for the Sn2Fe anode material in its pristine, lithiated, and delithiated forms: (a) Fourier transform (FT) magnitude of EXAFS spectra K2χ(k) at the Fe K-edge, k ranges 2–10.5 Å–1; (b) Fourier transform (FT) magnitude of EXAFS spectra K2χ(k) at the Sn K-edge, k ranges 1.5–12 Å–1. The inserts are their respective k-space EXAFS signal χ(k).

A structure involving Sn–Li interaction was attested the most suitable model to fit the Sn-edge EXAFS data for the lithiated sample (Figure ). It is found that Sn–Fe, Sn–Sn, and Sn–Li, at respective distances of 2.69(1)°, 2.92(3)°, and 2.86(5) Å, contribute to the EXAFS signal at the Sn K-edge. The alloying of Sn in the lithiated sample is well illustrated by the spectral difference from that of the Sn foil (Figure ). The Sn–Li bond, on average slightly shorter than 2.9 Å, points to a Li–Sn alloy structure with Li/Sn > 2.5.[50,51] The derived coordination numbers for Sn–Li and Sn–Sn are 6(2) and 1.0 ± 0.5, respectively, which corresponds well to the LiSn (2.5 < x < 4).[50,51]

Figure 10

FT magnitude of EXAFS data (black) and nearest shell fit (red), plotted together with individual coordinations for the fully lithiated sample F (cell stopped at 0.01 V) in the first cycle: (a) Sn and (b) Fe edges.

Figure 11

EXAFS data for lithiated Sn2Fe of first and second electrochemical cycles: (a) K2-weighted background-subtracted EXAFS signal χ(k) and (b) Fourier transform (FT) magnitude of K2χ(k) at the Fe K-edge, k ranges 2–10.5 Å–1. (c) K2-weighted background-subtracted EXAFS signal χ(k) and (d) Fourier transform (FT) magnitude of K2χ(k) at the Sn K-edge, k ranges 1.5–12 Å–1. Foil data for respective edges are included for comparison.

A combination of Fe–Fe and Fe–Sn bonds was employed to model the Fe EXAFS data at low-R regions (Figure ). The Fe–Fe distance is calculated to be almost equivalent in length to that in the Fe foil, which is considerably shorter (by ∼0.16 Å) than Fe–Fe distance in pristine Sn2Fe. This contraction is also accompanied by a 4-fold increase in the coordination number. These quantitative evidences confirm the initial observations of Fe XAS and are in excellent agreement with the XRD-observed formation of the bcc Fe metal. The resulted Fe–Fe coordination number, 3.9(4), is below the average value (8) for a full first coordination sphere in the Fe crystal structure, which points toward the small particle size of the segregated Fe. The average of the heterogeneous bond between Fe and Sn atoms renders ∼0.08 Å contraction from its original length in the pristine sample, corroborating the structure rearrangement involving the transition from tetragonal Sn2Fe to cubic Fe. The breakdown of the Sn–Fe alloy structure is also demonstrated by the declined coordinations between Fe and Sn: NFe–Sn =1.6(2) and NSn–Fe = 0.5(4). Interestingly, though reduced, the contributions of Fe–Sn interaction to the EXAFS data are essential, hinting the presence of the minor SnFe phase or close proximity of nano Fe- and Sn-based phases.

The best-fitting results (Table ) for the first cycle delithiated sample K reveals a slight reduction of the Fe–Sn coordination number and the considerable shortening of Fe-involved bonds compared to the pristine material. It is proposed that in the delithiated structure, the core is dominated by Sn–Fe alloying with somewhat strained geometry caused by those unrecovered Fe. This irremediable transformation of the structure is expected as it has been established that the first cycle usually involves irreversible structure rearrangement or activation.[52,53]

Table also shows that at approximately 50% lithiated stage (cell stopped at 0.12 V during first cycle lithiation), the respective value of each Fe-based bond parameter, derived from EXAFS fitting, is in-between of those in pristine and fully lithiated sample, indicating the gradual transformation of Fe from the alloyed to the segregated phase in the course of lithiation intercalation. However, no direct evidence is present to confirm Li–Sn alloying at this stage while large uncertainties and high correlation obtained for the coordination numbers and bond disorders for the two Sn–Sn bonds, averaged at 3.07(4)° and 3.41(3) Å, respectively, implicating the possibility for Sn to be in mixed phases of Fe-alloyed Sn and segregated Sn.

Magnetic properties were studied to further investigate the details of Fe separation and reconversion back to Sn2Fe in the first cycle, based on a distinct difference in their magnetic properties. As was mentioned earlier, Sn2Fe is antiferromagnetic at room temperature, while Fe is ferromagnetic in bulk and superparamagnetic if nanosized.[46,47,49]Figure shows magnetization curves of lithiated (point F) and delithiated (point K) Sn2Fe in comparison with that of the pristine sample (point A). Upon discharge to 0.01 V (point F), magnetization increases significantly and attains about 1.7 μB/mol. This value is less than expected for the bulk iron (2.2 μB), which was observed upon lithiation of Sn2Fe by Chamas et al.[36] However, it is consistent with our EXAFS observation of considerable Sn–Fe bonding in this sample, indicating incomplete conversion. It is interesting to notice that the sharp magnetization increase occurs at the very end of the lithiation process, indicating that Fe displacement from the alloy proceeds gradually, and the distinct Fe particles are formed only at the end of discharge. Charge to 1.5 V results in magnetization decrease, but it still remains a bit higher than that of pristine Sn2Fe indicating some remaining Fe, which is also consistent with EXAFS and XRD observations.

Figure 12

(a) Magnetization curves at 2 K and (b) field-cooled (solid symbols) and zero-field-cooled (open symbols) temperature dependences of magnetization of pristine, lithiated to 0.01 V and delithiated to 1.5 V Sn2Fe.

Field-cooled and zero-field-cooled dependences of magnetization were also studied as they can indicate formation of superparamagnetic iron particles and allow determination of their size. Pristine Sn2Fe shows that FC and ZFC curves depart already at 400 K, the highest temperature available in the experiment. Upon lithiation, the ZFC curve develops a peak at 18 K in a sample lithiated to point K, which is attributed to the blocking temperature Tb of superparamagnetic Fe particles. The volume V of iron particles can be estimated from Tb using equation Tb = KV/25kB (K is the magnetocrystalline energy and kB = 1.38 × 10–16 erg/K is the Boltzmann constant). Assuming the magnetocrystalline constant K = 4.8 × 105 erg/cm3 of metallic Fe0 and spherical particle shape, the Fe particle diameter is about 6 nm at the end of the lithiation process. It should be noted that the magnetocrystalline constants up to an order of magnitude higher were reported for the 2–3 nm Fe particles, which would bring the particle size down to 3 nm. It is consistent with the Fe particle size observed by Chamas et al.[36] from the magnetization data and by Mao et al. from the Mössbauer data.[54]

The FC curve of the delithiated sample closely resembles that of the pristine sample; however, ZFC curve still shows a maximum typical of superparamagnetic Fe particles centered at 100 K, which corresponds to 5 or 10 nm particles using two different magnetocrystalline constants mentioned above. This observation is consistent with slightly higher magnetization found for the delithiated sample in comparison with the pristine one. Larger Fe particle size in the delithiated sample points toward Fe particle coarsening during the delithiation process or indicates that larger Fe particles tend to remain unreacted upon delithiation.

For the anode materials, it is known that the first discharge–charge cycle is often associated with some side reactions such as SEI formation, cracking of crystallites, materials activation, structure rearrangement, and so forth. It is evident in the electrochemical data (Figure a) as a large irreversible capacity, which makes it difficult to delineate the Sn2Fe lithiation/delithiation reaction from the SEI formation. Therefore, to have a comprehensive understanding of the reaction mechanism of Sn2Fe, high-resolution synchrotron X-ray diffraction measurements have also been performed on the ex situ powder samples from the second discharge–charge cycle. As shown in Figure , during the second discharge, the transformation of Sn2Fe to Li–Sn alloys (see the two broad peaks of 2θ = ∼11° and ∼18.2°) and bcc α-Fe is similar to that occurring in the first cycle. However, such a phase transformation mainly occurs as early as at 0.366 V (sample L), which is much faster compared to the first cycle (the main phase transformation occurring at 0.125 V (sample C)). This kind of kinetic difference may be ascribed to the cracking of crystallites or breaking up of agglomerates after the first cycle, which would allow a better electrolyte access to the active electrode material and would trigger the conversion reaction earlier. After the second discharge, a small amount of Sn2Fe still remains in the fully lithiated sample P probably as the discrete or inactive particles (Figures and S2).

Figure 13

Ex situ synchrotron XRD patterns (wavelength = 0.728 Å) of solvothermal Sn2Fe for the second cycle stopped at different voltages indicated by each curve.

Similar to the first charge, the recovery of most Sn2Fe as well as the reversible reaction are also observed during the second charge process. Interestingly, at the lithiation stages S or T (above 0.585 V), LiSn, an intermediate state of the Li–Sn alloys’ evolution, has been clearly captured (Figure ). This low-lithium-content alloy is formed when the dealloying process progresses from high-lithium-content Li–Sn phases (such as Li3.5Sn and Li4.4Sn) to low-lithium-content ones. Different from the first charge, the Sn2Fe recovery during the second charge does not proceed so much until the very end of the delithiation; while a clear Sn2Fe formation can be observed starting from the sample I of the first cycle (Figure a). Although most of Sn2Fe can be recovered from the reversible reaction, a small amount of unreacted Li–Sn alloys as well as bcc α-Fe can be clearly identified in the final product (after two cycles), as shown in Figure . It is noteworthy that the amount of these unreacted species increases from the first to the second cycle, which means that inactive particles of Li–Sn alloys and Fe could be accumulating from cycle to cycle. This might be the reason why the solvothermally synthesized Sn2Fe exhibits a bigger capacity fading than the mechanochemically formed Sn2Fe/Sn/C composite.[18,19]

A comparative XAS evaluation of the second lithiation/delithiation process against the first one was conducted, emphasizing that the structure features at fully lithiated and delithiated stages. Analysis shows that a good fit of EXAFS data for the lithiated sample can be achieved also by involving Sn–Li interaction (2.88 ± 0.01 Å), confirming the reformation of LiSn where x is greater than 2.5.[50,51] The coordination number of Sn–Li is calculated to be 6.5 ± 0.5, comparable to that of the fully lithiated sample in the first cycle.

However, no Sn–Sn bond is detected within ∼3.0 Å distant from the central Sn atom, a distinguished difference from that in the first cycle and also clearly exhibited by FT-EXAFS in Figure d. Its absence and the negligible Sn–Fe bond (0.10 ± 0.09) are taken as the supporting evidences for complete conversion upon lithiation in the second cycle to the high-Li-content alloys, for instance, Li4.25Sn or Li4.4Sn, in which the available Sn–Sn scatterings are expected to appear at R > 4.6 Å.[55−57] In addition, the coordination number of the Fe–Fe bond (5.5 ± 1.8) is larger than its first cycle counterpart, reflecting the growth or aggregation of discrete Fe particles. This more complete lithiation reaction in the second cycle results in a higher charge capacity in the second cycle evidenced in Figure .

EXAFS examinations for the delithiated sample for the second cycle suggest that Sn–Fe alloying is mostly restored as the material is delithiated. The resemblance of Sn spectral properties of the two delithiated samples in the first and second cycles and comparable parameters derived from the fits particularly illustrate the similarity. On the other hand, a noticeable difference of the second cycle delithiated sample from that of the first cycle is observed in the Fe–Fe coordination: a decrease in bond distance and an increase in coordination number. This finding may hint an elevated concentration of the segregated Fe phase in the system.

Additionally, the XAFS investigation of the intermediate phases of the lithiation process during the second cycling demonstrates a progressive transformation of structure from mainly the Sn–Fe alloy to the mixture of Sn–Li alloy and segregated Fe. It suggests that Sn–Li alloying can be identified at a much earlier stage of discharge, compared to the first cycle. The Sn–Li bond averaged at 2.92 ± 0.02 Å is involved in the fitting for Sn EXAFS data for approximately 20% lithiated sample B (cell stopped at 0.366 V during lithiation). Also, the Sn–Sn bond becomes much shorter than that in the Sn2Fe alloy or segregated Sn. The resulted value of 2.98 ± 0.01 Å is consistent with formation of lithiated Sn, the composition of which is in-between of Li7Sn3 and LiSn.[50,51]

Conclusions

Sn–Fe anode materials with various ratios (2:1, 5:1, and 10:1) were synthesized successfully via solvothermal route; among them, 2:1 product of mainly the Sn2Fe phase delivers better electrochemical performance than Sn2Fe/Sn. A combination of XRD, XAS, and magnetic studies has revealed Li–Sn alloying and metallic iron formation during the first lithiation. XAS and magnetic data suggest a small Fe particle size (about 3 nm from the magnetic properties) and incomplete conversion. The first delithiation proceeds by LiSn dealloying and Sn2Fe alloy reformation, indicating that the conversion reaction is generally reversible. Magnetic studies show that some Fe particles of a larger size (5–10 nm) remain after the first charge. In the second lithiation, earlier formation of LiSn alloys and more complete conversion is evidenced; however, upon delithiation, the unreacted Fe accumulation continues, as indicated by shorter Fe–Fe distances, approaching those of metallic Fe. Such coarsening, observed also by Mao et al. using the Mössbauer technique,[54] might be a critical factor contributing to the capacity loss upon cycling. One of the ways to prevent this coarsening could be by creation of composites with even smaller, uniform-sized particles, as we have recently demonstrated.[58] Based on our data and analysis of the reaction mechanisms reported in the literature, we believe that the differences in the reaction mechanism are caused by differences in particle size, morphology, conductive additives, and other details affecting the material’s ability to undergo the reversible conversion.