Experimental

To ball mill samples in N₂(g), a high-energy ball mill (SPEX Model 8000-D, Spex CertiPrep, Metuchen, NJ) was placed in a N₂-filled glove box. 65 mL hardened steel vials (also from Spex CertiPrep) were used in milling containers. All powders were loaded and all ball milling was conducted in the N₂-filled glove box which contained the ball mill. Free access of ambient N₂ gas into and out of the milling vial during the milling process was accomplished by removing the o-ring seal of the vial and drilling a 0.125 inch hole into the vial clamping ring, as shown in Fig. S2. Using this set-up, Si-Ti-N alloys were synthesized by ball milling under constant 1 bar pressure N₂(g), 0.5 mL total (according to bulk densities) of various ratios of Si powder (Sigma Aldrich, 325 mesh, 99%) and Ti powder (−325 mesh, 99% metals basis from Alfa Aesar) with 180 g of 0.125″ stainless-steel balls. This powder and ball loading is optimized for rapid milling with little iron contamination.^[Reference Hatchard, Genkin and Obrovac^16] A sample was also made by ball milling 0.5 mL of TiSi₂ powder (Alfa Aesar, 99%) in N₂ gas under the same conditions. Unless otherwise stated, all alloys were ball milled for 16 h.

X-ray diffraction (XRD) patterns were measured using a Rigaku Ultima IV diffractometer equipped with a Cu Kα x-ray source and a graphite diffracted beam monochromator. Each XRD scan was collected from 10° to 60° 2-theta in 0.05° increments for 3 seconds per step. Single point BET measurements were conducted using a Micromeritics FlowSorb ΙΙ 2300 using a nitrogen/helium gas mixture (28.6% N₂). Samples were degassed at 150°C for 30 min to remove moisture before the measurement. True sample densities were measured with He gas using a Micromeritics AccuPyc II 1340 gas pycnometer. Nitrogen and oxygen contents were determined by the LECO method (NSL Analytical Services, Inc, Cleveland, OH). Scanning electron microscope (SEM) images were measured using a TESCAN MIRA 3 LMU Variable Pressure Schottky field emission scanning electron microscope. Transmission electron microscopy (TEM) images were taken using a Philips CM30 TEM. TEM specimens were made by suspending sample powders in methanol, sonicating 10 min, and placing a drop onto a lacey carbon-coated TEM grid.

Electrode slurries were prepared by mixing Si-Ti-N alloys, carbon black (Imerys Graphite and Carbon, Super C65) and a 10 wt.% aqueous solution of lithium polyacrylate (LiPAA) with a volumetric ratio of 70/5/25 in distilled water. Slurries were mixed for 1 h using a Retsch PM200 planetary mill at 100 rpm with three 13 mm tungsten carbide balls, spread onto copper foil (Furukawa Electric, Japan) with a 0.004 inch gap coating bar, and dried in air for 1 h at 120°C. This resulted in a typical electrode loading of 1 mg of active alloy per square centimeter and an average coating thickness of 15 µm. Circular working electrodes, 1.35 cm² in area were cut from the coated foils using a circular punch.

Electrodes were assembled in 2325-type coin cells with a lithium foil counter/reference electrode. Two layers of Celgard 2300 separator were used in each coin cell. 1 M LiPF₆ (BASF) in a solution of ethylene carbonate, diethyl carbonate and monofluoroethylene carbonate (volume ratio 3:6:1, all from BASF) was used as electrolyte. Cell assembly was carried out in an Ar-filled glove box. Cells were cycled galvanostatically at 30.0 ± 0.1°C between 5 mV and 0.9 V using a Maccor Series 4000 Automated Test System at a C/20 rate for the 1st cycle and a C/5 rate for the following cycles. To simulate the “constant current constant voltage” (CCCV) cycling protocol used in commercial cells, after each discharge (lithiation) cells were held at 5 mV until the current dropped to a C/40 rate during the first cycle and to a C/20 rate during subsequent cycles prior to commencing the next charge step. Here, C-rate is the current required to fully charge or discharge cells in 1 h and was calculated based on the amount of Si in the alloy and its theoretical capacity of 3578 mAh/g.^[Reference Obrovac and Christensen^17] All specific capacities were calculated with respect to the weight of the Si-alloy in the electrode.

Results and discussion

Figure 1(a) shows XRD patterns of Si and Ti ball milled in a 85:15 molar ratio under N₂(g) for 4, 8, and 16 h. After milling 4 h, the XRD pattern is composed of amorphous Si and the C49 TiSi₂ phase, which is a metastable phase of TiSi₂ formed by ball milling.^[Reference Wang, Cao, Kalinina, Zheng, Li, Zhu and Obrovac^11] In addition, a small amount of intensity is present at about 37°, which is consistent with the (111) reflection of TiN. After 8 h milling, the TiSi₂ peaks reduce in intensity and peaks from TiN increase in intensity. After 16 h milling only peaks from amorphous Si and nanocrystalline TiN are present. This sample is denoted here as Si₈₅Ti₁₅(N₂). The TiN grain size of Si₈₅Ti₁₅(N₂) is about 5 nm. This sample was found to contain 13 at.% nitrogen and 8 at.% oxygen, according to LECO analysis. This corresponds to a composition of Si₆₇Ti₁₂N₁₃O₈. The oxygen in Si₈₅Ti₁₅(N₂) likely comes from the large oxygen impurity in the Si precursor powder, which was measured to be 8 at.%.^[Reference Wang, Cao, Kalinina, Zheng, Li, Zhu and Obrovac^11] The XRD pattern and the nearly equal amounts of nitrogen and titanium in the alloy suggest full reaction of the Ti with the ambient N₂(g) during ball milling to form TiN. No evidence of Si₃N₄ is present in any of the XRD patterns. These results suggest that reaction (1) is kinetically favored during the milling process, resulting in the initial formation of TiSi₂. With continued milling, the TiSi₂ phase reacts with N₂ accordingly to form TiN. Therefore, we believe the complete reaction of Si and Ti powders ball milled in N₂(g) may be represented by

$${\rm Step}\,\, 1\,\,: \,\,\hbox{85Si(s)} + \hbox{15Ti(s)} \buildrel {{\rm N}_2({\rm g})} \over \longrightarrow \,\,\hbox{55Si(s)} + \hbox{15TiSi}_2({\rm s})$$

(4)$${\rm Step}\,\,2\,\,: \hbox{55Si(s)} + \hbox{15TiSi}_2({\rm s}) \buildrel {{\rm N}_2({\rm g})} \over \longrightarrow \hbox{85Si(s)} +\hbox{15TiN(s)}$$

where again a 85:15 Si:Ti mole ratio is used as a representative example. This reaction is contrary to what is predicted by thermodynamics, as shown in Fig. S1. To confirm this reaction, TiSi₂ powder was ball milled under N₂(g) for 16 h. This also resulted in the formation of a phase-pure Si-TiN alloy with an identical XRD pattern, as shown in Fig. 1(a). These results emphasize the non-equilibrium character of the ball milling process. For comparison, a Si-TiN alloy was made by ball milling Si and TiN powders in a 85:15 molar ratio under Ar gas. This sample is denoted Si₈₅(TiN)₁₅(Ar). The XRD pattern of Si₈₅(TiN)₁₅(Ar) is shown in Fig. S3, and, similar to Si₈₅(TiN)₁₅(N₂), the XRD pattern comprises amorphous Si and TiN, however the grain size of TiN is much larger (11.3 nm). This shows that under the same conditions, TiN may be made with a smaller grain size by forming it in-situ via the reaction of Ti and Si with N₂(g) rather than ball milling TiN directly with Si.

Figure 1.

(a) XRD patterns of Si and Ti powder in an 85:15 atomic ratio ball milled in N₂(g) for 4, 8, and 16 h. (b) XRD patterns of different atomic ratios of Si and Ti powders after ball milling in N₂(g) for 16 h.

Figure 1(b) shows XRD patterns of Si-Ti(N₂) alloys made by ball milling different ratios of Si and Ti powders in N₂(g) for 16 h. When the Si content is high (>60%), only amorphous Si and nanocrystalline TiN are apparent in the XRD pattern. Elemental analysis results in Table I show that these samples have a Ti:N ratio of nearly 1:1, confirming that they consist of Si and TiN phases. The XRD pattern of the Si₆₀Ti₄₀(N₂) sample shown in Fig. 1(b) consists of Si, TiN and TiSi₂ and the Ti:N atomic ratio is about 2:1, according to Table I. This indicates an incomplete reaction according to Eq. (3). We believe that a high content of ductile Ti during the initial stages of milling reduces the ball milling reaction kinetics, requiring long milling times. This will be further described in a subsequent publication. The XRD pattern of the Si₆₀Ti₄₀(N₂) sample shown in Fig. 1(b) consists of TiN and some additional intensity near 42°. This peak appears at lower angle than the main peak of TiSi₂. We speculate that this peak is from the formation of Ti₅Si₃. No XRD peaks are present for Si in this sample.

Table I.

Compositions, specific surface areas and bulk densities of Si-Ti-N alloys made by ball milling different atomic ratios of Si and Ti powders in N₂ for 16 h. The compositions are based on the initial Si and Ti content in the mill and the N and O contents as determined by LECO analysis.

Figure 2 shows SEM images of the alloys listed in Table I. As the Ti content increases, the average particle size increases from about 0.5 µm for Si₈₅Ti₁₅(N₂) to large flakes up to 15 µm in extent for Si₆₀Ti₄₀(N₂). The large Si₆₀Ti₄₀(N₂) particle size results from the increased content of ductile Ti during the initial stages of milling. Commensurate with the increased particle size, the surface area reduces substantially from 18 m²/g for Si₈₅Ti₁₅(N₂) to 4.43 m²/g for Si₆₀Ti₄₀(N₂). Such low surface area could be advantageous for lowering the amount of surface reactions with electrolyte, when such alloys are used in Li cells.

Figure 2.

SEM images of (a) Si₈₅Ti₁₅(N₂), (b) Si₇₀Ti₃₀(N₂), and (c) Si₆₀Ti₄₀(N₂).

The potential versus capacity and differential capacity curves of Si-Ti(N₂) alloys are shown in Fig. 3. Each is characteristic of Si-inactive alloys where the formation of Li₁₅Si₄ (which would be indicated by an anodic peak in the differential capacity at about 0.45 V) is suppressed by the presence of an inactive phase.^[Reference Du, Hatchard, Dunlap and Obrovac^18] Li₁₅Si₄ formation can only be detected in the differential capacity of the highest Si-containing alloy, Si₈₅Ti₁₅(N₂), as the peaks near 0.45 V in Fig. 3(d). The reversible capacities of the Si₈₅Ti₁₅(N₂) (2411 mAh/g) and Si₇₀Ti₃₀(N₂) (1865 mAh/g) are close to what are predicted theoretically (2184 and 1678 mAh/g, respectively) for the compositions of these samples given in Table I, assuming that they consist of active Si (3578 mAh/g) and inactive SiO₂ and TiN phases. These reversible capacities correspond to 2089 and 1957 Ah/L, respectively. The Si₆₀Ti₄₀(N₂) sample has greatly reduced capacity because of its large inactive TiSi₂ content, which removes active Si from the alloy. Nevertheless, the volumetric capacity of this sample is still high (1371 Ah/L). The Si₅₀Ti₅₀(N₂) sample is not shown in Fig. 3 since it was found to have almost no capacity, consistent with is very low active Si content.

Figure 3.

(a–c) Potential versus capacity curves and (d–f) associated differential capacity curves for Si-Ti(N₂) alloys with different compositions, as indicated.

The cycling performance of these alloys are shown in Fig. 4 in terms of volumetric capacities. Here, the volumetric capacities were calculated at full alloy volume expansion as described in Ref.Reference Obrovac, Christensen, Le and Dahn19 using measured bulk densities, which are listed in Table I. Excellent cycling performance is shown over the 50 cycles for all the materials, although with some fade shown for the Si₇₀Ti₃₀(N₂) sample. The initial coulombic efficiency (ICE) of these alloys is also impressive, ranging from 78% for Si₇₀Ti₃₀(N₂) to 85% for Si₈₅Ti₁₅(N₂). These samples have much higher capacities (more than double) than that demonstrated by Kim et al. in Ref.Reference Kim, Kumta and Blomgren3 (~300 mAh/g).

Figure 4.

Cycling performance of Si-Ti(N₂) alloys with different compositions, as indicated, in terms of volumetric capacity.

Figures S4(a)–S4(c) show the potential versus capacity curve, differential capacity and cycling performance of Si₈₅(TiN)₁₅(Ar), respectively. In contrast to the Si₈₅Ti₁₅(N₂) sample, a large 0.45 V plateau is present in the Si₈₅(TiN)₁₅(Ar) potential versus capacity curve during delithiation, with a large corresponding peak in the differential capacity. This indicates that the inactive TiN phase is not effective in suppressing Li₁₅Si₄ formation in the sample milled in Ar(g). This is likely due to the larger grain size of the TiN phase in the Si₈₅(TiN)₁₅(Ar) sample, which may result in less contact area with active Si. Figure S4(c) shows that this sample undergoes severe fade during cycling, which is often synonymous with Li₁₅Si₄ formation.^[Reference Obrovac and Chevrier^1] This shows that ball milling Si and Ti in N₂ is not only a much more convenient and less expensive way to make Si-TiN alloys, but also the method creates a microstructure that results in superior cycling performance than ball milling Si and TiN directly.