Path Less Traveled: A Contemporary Twist on Synthesis and Traditional Structure Solution of Metastable LiNi12B8.

Achieving kinetic control to synthesize metastable compounds is a challenging task, especially in solid-state reactions where the diffusion is slow. Another challenge is the unambiguous crystal structure determination for metastable compounds when high-quality single crystals suitable for single-crystal X-ray diffraction are inaccessible. In this work, we report an unconventional means of synthesis and an effective strategy to solve the crystal structure of an unprecedented metastable compound LiNi12B8. This compound can only be produced upon heating a metastable layered boride, HT-Li0.4NiB (HT: high temperature), in a sealed niobium container. A conventional heating and annealing of elements do not yield the title compound, which is consistent with the metastable nature of LiNi12B8. The process to crystallize this compound is sensitive to the annealing temperature and dwelling time, a testament to the complex kinetics involved in the formation of the product. The unavailability of crystals suitable for single-crystal X-ray diffraction experiments prompted solving the crystal structure from high-resolution synchrotron powder X-ray diffraction data. This compound crystallizes in a new structure type with space group I4/mmm (a = 10.55673(9) Å, c = 10.00982(8) Å, V = 1115.54(3) Å3, Z = 6). The resulting complex crystal structure of LiNi12B8 is confirmed by scanning transmission electron microscopy and solid-state 11B and 7Li NMR spectroscopy analyses. The extended Ni framework with Li/Ni disorder in its crystal structure resulted in the spin-glass or cluster glass type magnetic ordering below 24 K. This report illustrates a "contemporary twist" to traditional methodologies toward synthesizing a metastable compound and provides a recipe for solving structures by combining the complementary characterization techniques in the cases where the traditionally used single-crystal X-ray diffraction method is nonapplicable.

Introduction

The outcome of a chemical reaction depends on whether the reaction pathway is thermodynamically or kinetically controlled. Thermodynamically controlled pathways often have higher activation energies and yield products with the lowest free energies. On the contrary, the activation energy for a kinetically controlled pathway is typically lower, thus leading to products that rapidly form but do not have the lowest energies. A traditional solid-state synthesis that involves the direct combination of solid elemental precursors has to be performed at elevated temperatures for an extended time to overcome the high diffusion barriers between inorganic solids. Such prolonged heat treatment often directs the reactions along a thermodynamically controlled pathway hindering access to metastable (kinetically stabilized) compounds. Therefore, for the synthesis of metastable compounds, ″turning down the heat″ and employing unconventional synthetic routes are shown to be successful strategies to access metastable compounds with potentially unprecedented topology and properties.[1−12]

Recent advancements toward gaining kinetic control through metathesis,[12−20] low-temperature solution synthesis,[21−28] and chimie douce (soft chemistry)[29−35] have allowed metastable compounds to become readily accessible. When it comes to borides, the challenge of synthesizing metastable borides is even greater due to the high melting point of elemental boron (∼2350 K), thus requiring unconventional synthetic methods. Rare examples of metastable borides include Na2BP2 obtained via an oxidative elimination pathway,[36] β-NdCo2B2 synthesized via rapid cooling of an arc-melted ingot,[37] and Ni7B3 from low-temperature solution synthesis,[22] making metastable borides an underexplored class of materials. Motivated by these studies, we pursued the preparation of metastable borides in the Li–Ni–B system through soft topochemical deintercalation route.

Our recent report on two metastable borides, RT-Li0.6NiB and HT-Li0.4NiB (RT: room temperature; HT: high temperature), synthesized upon partial topochemical deintercalation of Li from the layered RT-LiNiB and HT-LiNiB polymorphs[38] laid a foundation for this study. Crystal structures of both RT- and HT-LiNiB feature alternating, one-atom-thick [NiB] and Li layers with different topologies.[39] Given their layered structure and labile Li cations, LiNiB polymorphs are suitable precursors for preparing two-dimensional (2D) single-layer borides, MBenes (M = early transition metals) via Li deintercalation. MBenes are boron analogues of 2D carbides/nitrides, MXenes. MXenes are obtained by topochemical deintercalation of loosely bonded “A” atoms from the parent MAX (MAX) phases (M = early transition metals, A = elements of groups 13–14 (mainly Al or Si), and X = C or N).[40−43] Unlike graphene and MXenes, bulk preparation of individual MBenes remains in its infancy, although there are a few appropriate deintercalable precursors. Nevertheless, MBenes have been theoretically predicted to have outstanding properties as battery materials, catalysts, and magnetic materials,[44−47] and several attempts have been made to synthesize them.[48−52]

Motivated by the intriguing properties of MBenes and the successful partial Li-deintercalation from LiNiB, we attempted the complete Li deintercalation at elevated temperatures to obtain individual MBene (NiBene) layers. Unexpectedly, heating HT-Li0.4NiB leads to a novel metastable boride LiNi12B8 with a new structure type. A thorough analysis of in situ high-temperature synchrotron and ex situ laboratory powder X-ray diffraction (PXRD) data reveals LiNi12B8 to be metastable. Solving the crystal structure of this new compound via the commonly used single-crystal X-ray diffraction method presented a challenge due to the unavailability of large enough and good quality single crystals and the presence of light elements (Li and B) that are weak X-ray scatterers. Therefore, to elucidate the structure of this compound, we used complementary characterization techniques—high-resolution synchrotron PXRD, 7Li and 11B solid-state NMR spectroscopy, scanning transmission electron microscopy (STEM), and density functional theory (DFT). The approach presented here that includes solving structure from PXRD and backing it up by an array of other characterization methods illustrates a “contemporary twist” to traditionally used single-crystal XRD as the only way to determine the structure of a new compound. Finally, magnetic property studies of LiNi12B8 reveal spin glass or cluster glass type magnetic ordering at low temperatures. Overall, this report provides a successful example of a synergistic approach used for the discovery and characterization of a metastable compound and signals about the ample phase space of the metastable compounds that are inaccessible via traditional synthetic routes.

Experimental Section

Synthesis

Details about the synthesis of RT-LiNiB, HT-LiNiB,[39] and their partially Li-deintercalated counterparts RT-Li0.6NiB and HT-Li0.4NiB[38] can be found in our recent work. To synthesize nearly phase-pure LiNi12B8, HT-LiNiB was first exposed to air for 24 h, washed with deionized (DI) water to remove residual LiOH·H2O, and dried under vacuum, followed by repeating this treatment but after 48 h air exposure. Then, 30–60 mg of the dried sample was sealed in an Nb tube, heated to 973 K at a heating rate of 1.4 K/min, dwelled for 12 h, and quenched into cold water.

Characterization

All polycrystalline samples were characterized by PXRD using a Rigaku MiniFlex600 powder diffractometer with Cu Kα radiation (λ = 1.540593 Å). The thermal stability of RT-Li0.6NiB, HT-Li0.4NiB, and LiNi12B8 was investigated by in situ high-temperature synchrotron PXRD collected at the 17-BM beamline at the Advanced Photon Source Argonne National Laboratory (APS ANL) with an average wavelength λ = 0.24141 Å. High-resolution synchrotron PXRD data collected at 11-BM APS ANL at 298 K (λ = 0.45784 Å) were utilized to solve the crystal structure of LiNi12B8. Solid-state 7Li and 11B NMR spectra were acquired using a Bruker Avance III HD spectrometer with a widebore 9.4 T (ν1H = 400 MHz) NMR magnet. Quantitative elemental analysis was carried out via inductively coupled plasma mass spectrometry (ICP-MS) using a Thermo Fisher Scientific X Series 2 ICP-MS. An FEI Quanta-250 field emission scanning electron microscope (SEM) equipped with an Oxford X-Max 80 detector and an Oxford Aztec energy-dispersive X-ray spectroscopy analysis system was utilized for the elemental analysis of the samples. High-angle annular dark field and bright field scanning transmission electron microscopy (HAADF & ABF-STEM) images and electron diffraction (ED) data were acquired using a JEM ARM200F cold FEG double aberration-corrected microscope operated at 200 kV. Magnetic property measurements were performed on the polycrystalline samples using Quantum Design MPMS XL and MPMS3 SQUID magnetometers. First-principles calculations were carried out using DFT within the spin-polarized generalized gradient approximation by the VASP code. Detailed information about the synthesis and characterization methods utilized in this work can be found in the Supporting Information.

Results and Discussion

NiBene Layers in the Structures of Lithium Nickel Borides

The layered lithium nickel borides RT-LiNiB, HT-LiNiB, RT-Li1+NiB, and HT-Li1+NiB[39,53] represent a family of layered compounds whose crystal structures are built up of alternating [NiB] and Li layers (Figure ). These compounds feature different topologies of individual MBene (NiB) and Li layers depending on the temperature and Li “chemical” pressure. We have recently reported[38] on the topochemical deintercalation of Li from both RT- and HT-LiNiB polymorphs. Images obtained by HAADF-STEM and SEM revealed that deintercalation progresses layer by layer following a “ziplock” mechanism: upon deintercalation of a single Li layer, the “ziplock” is closed, and the adjacent [NiB] layers are forced to condense. However, complete deintercalation of Li could not be achieved by exposure to air, water, ethanol, and dilute HCl at room temperature. Instead, novel metastable borides RT-Li0.6NiB and HT-Li0.4NiB form, with complex crystal structures best described as a random intergrowth of ordered Li[NiB] (n = 1, 2, 3) substructures with the substructure Li[NiB]2 being the most abundant, accounting for up to 90 mol % in the case of HT-Li0.4NiB. Considering the metastable nature of the Li-deintercalated products, we hypothesized[54,55] that additional thermal energy might lead to further Li deintercalation, accompanied by a possible restructuring or transformation of NiBene layers.

Figure 1

LiNiB phases as flexible platforms to access a plethora of unique layered and nonlayered materials.

In Situ and Ex Situ Heating of RT-Li0.6NiB and HT-Li0.4NiB

In order to study thermal effects and possible complete deintercalation of Li atoms from RT-Li0.6NiB and HT-Li0.4NiB, we acquired in situ and ex situ PXRD data. Both compounds undergo similar transformations under in situ heating (Figures , S1, and S2). Upon initial heating, RT-Li0.6NiB and HT-Li0.4NiB disproportionate into a Li-rich, RT-LiNiB, or HT-LiNiB and the Li-depleted compound LiNi3B1.8. Further heating increases the relative peak intensities for LiNi3B1.8 and eventually produces the stable binary borides of nickel (Ni3B and Ni4B3). Interestingly, a few unknown peaks are detected upon heating RT-Li0.6NiB above 649 K and below 715 K. Rietveld refinement of the PXRD pattern using the RT-LiNiB structure as the initial model leads to a new layered RT*-LiNiB structure (a = 4.552(19) Å, b = 4.844(8) Å, c = 6.38(3) Å, β = 105.80(7)°, and P2), which differs from RT-LiNiB (a = 4.61170 Å, b = 4.83330 Å, c = 6.15700 Å, β = 109.61200°, and P21/c) by the way [NiB] layers are stacked (Figure S3). However, this structure neither could be observed in the ex situ heating of RT-Li0.6NiB nor quenched to room temperature by rapid cooling. We presume that a side reaction with the silica capillary drives the gradual ″leaching″ of Li from RT-Li0.6NiB and HT-Li0.4NiB phases, leading to the reorganization of the [NiB] layers (Figures , S1, and S2):

Figure 2

In situ high-temperature synchrotron powder X-ray diffraction data (λ = 0.24141 Å) of HT-Li0.4NiB show gradual leaching of Li due to reaction of Li with the silica capillary. Increasing the temperature causes the transformation sequence HT-Li0.4NiB → HT-LiNiB → LiNi3B1.8 → Ni4B3.

To study only the effect of temperature without a change in composition due to the side reaction, we performed ex situ heating of RT-Li0.6NiB and HT-Li0.4NiB in sealed Nb tubes at various temperatures and collected PXRD data after cooling to room temperature. Similar to the in situ heat treatment, both compounds undergo alike transformation upon ex situ heating except for a major difference for HT-Li0.4NiB between 673 and 1073 K. When the HT-Li0.4NiB phase is heated at 673 K for 12 h, peaks corresponding to the known ternary compound LiNi3B1.8 along with a few unknown peaks appear in the PXRD pattern (Figure ). However, when it is heated to 773 K for 12 h, a plethora of unknown peaks are detected, implying a significant transformation of the [NiB] layers within the narrow temperature range of 673–773 K. Because these are broad peaks, indexing of the PXRD pattern is challenging and will be the subject of a future study. Upon heating HT-Li0.4NiB to 973 K, a new set of unknown peaks appear, different from the peaks observed at 773 K (highlighted in yellow in Figure ), in addition to minor amounts of LiNi3B1.8 and Ni2B. Upon solving the crystal structure, these unknown peaks are attributed to the novel compound LiNi12B8. Although ex situ heating of HT-Li0.4NiB leads to crystalline LiNi12B8 with a high yield, only traces of it form upon heating RT-Li0.6NiB at 973 K (Figure S4). Such yield differences most likely arise from the different topologies of the [NiB] layers in the partially Li-deintercalated precursors.[38,39]

Figure 3

PXRD patterns (λ = 1.540593 Å) of the ex situ heat treatment of HT-Li0.4NiB at different temperatures with a 12 h dwelling time. Dotted boxes are placed around the most intense peaks of phases present at specific temperatures.

Interestingly, the formation of LiNi12B8 is observed only during the ex situ experiment and not upon in situ heating, suggesting that the dwelling time plays a crucial role in stabilizing the target compound. The importance of time–temperature regimes for crystallization processes was previously established.[54,55] Therefore, we have investigated the impact of dwelling time on the preparation of LiNi12B8 to shed light on the reaction kinetics and optimize the yield. For this, HT-Li0.4NiB was sealed in Nb tubes and heated to 973 K (heating rate of 1.4 K/min) with variable dwelling times (Figure S5). Ex situ dwelling at 973 K for 8 h results in the target LiNi12B8 phase with LiNi3B1.8 as a side product (∼10 wt % from the Rietveld refinement of PXRD data). The significant amount of LiNi3B1.8 indicates that a longer dwelling time might be required. When the dwelling time is increased to 12 h, LiNi12B8 becomes the primary phase with LiNi3B1.8 (∼3 wt %) and Ni2B (∼0.5 wt %) as minor impurities. When the dwelling time is set to 16 h, a significant amount of Ni2B is detected (∼22 wt %) along with the target phase and traces of LiNi3B1.8. The concurrent appearance of LiNi12B8 and Ni2B from LiNi3B1.8 implies that LiNi3B1.8 (Ni/B = 1.67:1) disproportionates into Ni-rich Ni2B (Ni/B = 2:1) and Ni-deficient LiNi12B8 (Ni/B = 1.5:1). Our results suggest that a 12 h dwelling time is an adequate compromise to obtain the highest yield of the target LiNi12B8 compound (Figures and S5). Hence, the crystallization of LiNi12B8 from the semicrystalline HT-Li0.4NiB phase depends not only on the dwelling temperature but also on the dwelling time. Even though the formation of the LiNi12B8 compound is witnessed upon ex situ heating of HT-Li0.4NiB, this compound cannot be synthesized via direct synthesis from the elements, further emphasizing the metastable nature of LiNi12B8 (Figure S6).

Figure 4

Contents of phases LiNi3B1.8, Ni2B, and LiNi12B8 upon ex situ heating of HT-Li0.4NiB at 973 K for variable dwelling times; 12 h dwelling time is optimal with respect to the content of side phases.

From the analysis of both in situ and ex situ experiments, we can conclude that thermal energy increases the mobility of Li ions in both RT-Li0.6NiB and HT-Li0.4NiB and stabilizes the Li depleted compounds, LiNi3B1.8, LiNi12B8, and finally stable binary borides of nickel (Ni3B and Ni4B3). The basic pH of the solution obtained upon washing the samples with DI water supports the release of elemental Li with the subsequent formation of LiOH. However, complete deintercalation of Li to form NiBene layers cannot be achieved under the thermal methods employed.

In situ high-temperature PXRD data are utilized to reveal the metastable nature of LiNi12B8. Upon heating LiNi12B8, no visible changes in the PXRD patterns are detected (Figure S7) until 815 K, when diffraction peaks for Ni4B3 appear. At 983 K, LiNi12B8 irreversibly decomposes into Ni4B3 and does not reform on cooling to room temperature. Similarly, the formation of Ni4B3 is observed when the DI water-washed LiNi12B8 phase is heated ex situ at 973 K for 12 h in a sealed niobium container. The formation of only Ni4B3 (Ni/B = 1.33:1) upon ex situ heating is in line with the Ni/B ratio in the LiNi12B8 compound (Ni/B = 1.5:1), further supporting the proposed (vide infra) structural model. Therefore, upon exposing LiNiB (RT- and HT-) compounds to the air, deintercalation of Li atoms takes place, stabilizing metastable Li∼0.5NiB compounds, which then form LiNi12B8 upon heating and finally decomposing completely to Ni4B3 (Scheme ).

Scheme 1

Schematic Showing That LiNi12B8 Can Be Only Obtained via Gradual Leaching of Li upon Heating of the Layered Li∼0.5NiB Precursor

The latter is a result of partial Li deintercalation from LiNiB.

Crystal Structure

The formation of LiNi12B8 is sensitive to annealing temperature and dwelling time. LiNi12B8 compounds can only be accessed by ex situ heating of a microcrystalline HT-Li0.4NiB precursor, yielding microcrystalline powder of the target compound. Several attempts to grow crystals large enough for single-crystal XRD were unsuccessful. As a result, the crystal structure of LiNi12B8 was solved from the high-resolution synchrotron PXRD data, using the FOX software[56] and further refined with GSAS II software.[57] LiNi12B8 crystallizes in a tetragonal unit cell, space group I4/mmm (a = 10.55673(9) Å, c = 10.00982(8) Å, V = 1115.54(3) Å3, Z = 6, Pearson symbol tI126) (Table ). Searches for the Pearson symbol, unit cell metric, and Wyckoff sequence (omn) in the Inorganic Crystal Structure Database (ICSD)[58] indicate that LiNi12B8 forms a new structure type. Upon least-square refinements, the occupancy factors of selected Ni sites deviate from unity, suggesting mixed Ni/B or Ni/Li site occupancies. From the crystal structure analysis, the refined interatomic distances between Ni sites (∼2.5–2.6 Å) are too long for the small B atoms; thus, mixed occupied Ni/Li sites are considered. The least-square refinements suggested five Ni/Li mixed occupied sites, 32o, 16m, 8h, 4d, and 4e, with Li occupancies ranging 2–8% (Table S1). Refinement of this model from the high-resolution synchrotron PXRD data resulted in RB = 6.6% and G.O.F = 1.196 (Figure S8, Table ). With the consideration of five Ni/Li mixed occupied sites, the formula refines to Li1.317(2)Ni11.683(2)B8 or Li(Ni12–Li)B8 (x = 0.317(2)), which is in excellent agreement with the Li/Ni ratio of Li1.4(3)Ni11.6(2) determined from the ICP-MS data (see the Experimental Section in Supporting information). For the sake of clarity, we will refer to the title compound as LiNi12B8. Several reactions targeting the refined composition of LiNi12B8 were attempted by utilizing elemental precursors and LiH or Li as the lithium sources (Figure S6), but the LiNi12B8 phase could not be obtained, thereby emphasizing its metastability.

Table 1

Experimental Details and Crystallographic Data for the Li(Ni12–Li)B8 (x = 0.317(2)) Compounda

empirical formula	Li(Ni12–xLix)B8 (x = 0.317(2))
formula weight, g/mol	4688.42
space group, Z	I4/mmm, 6
cell parameters
a, Å	10.55673(9)
c, Å	10.00982(8)
V, Å3	1115.54(3)
temperature, K	295
wavelength, Å	0.45784
step scan, °	0.001
2θ° range	2.5–50
program	GSAS II
RB	0.066
RP	0.063
G.O.F.	1.196

Rietveld refinement was done using high-resolution synchrotron powder X-ray diffraction data (11-BM APS ANL). Further details of the crystal structure refinement can be obtained from the Inorganic Crystal Structure Database (ICSD) produced by FIZ Karlsruhe – Leibniz Institute for Information Infrastructure (www.fiz-karlsruhe.de) by quoting the CSD-2158021 deposition number at www.ccdc.cam.ac.uk/structures.

The crystal structure of LiNi12B8 can be described by either a framework of Li- or B-centered Ni polyhedra. In addition to five Ni/Li mixed occupied sites, there are two fully occupied Li sites (Figure ), Li1 at the 4c position and Li2 at the 2b position, each inside cuboctahedra of 12 Ni atoms. The Li–Ni bond distances are in the range of 2.50–2.66 Å, comparable to Li–Ni distances in the known Li–Ni–B compounds.[39,53,59] The Li-centered nickel polyhedra are connected by corner sharing; however, Li2-centered polyhedra are separated from each other by two face-shared Ni-centered rectangular prisms (Figure d).

Figure 5

(a) Polyhedral representation of the LiNi12B8 crystal structure in terms of Li-centered Ni polyhedra (deep blue and magenta) and Ni-centered rectangular prisms (orange); (b) Li1-centered nickel polyhedron; (c) Li2-centered nickel polyhedron; (d) two Ni-centered rectangular prisms, connected by face sharing, separate the Li2-centered polyhedra; and (e–g) different building blocks are highlighted within the structure. B atoms are omitted for clarity.

There are four different types of B atoms present in the crystal structure: B1 atoms in the 16n position are coordinated by eight Ni atoms in a square antiprismatic environment; B2 atoms occupy the 16n site that has seven nearest Ni atoms forming a capped trigonal prism; B3 atoms occupy 8h sites with six nearest Ni atoms in a trigonal prism, and B4 atoms at 8i sites are coordinated by eight Ni atoms forming a square antiprism (Figure a–e). The Ni–B bond distances are in the range of 2.04–2.26 Å and are comparable to the Ni–B distances found in binary nickel borides[60−62] and ternary Li–Ni–B borides.[39,53,59] The crystal structure of the LiNi12B8 compound can be represented as a framework of B-centered 6-, 7-, and 8-vertex polyhedra of Ni atoms, as shown in Figure a–e. Here, the B4-centered polyhedra are linked together by two Ni–Ni bridges (Figure g). B1-, B2-, and B3-centered polyhedra are connected by face sharing, resulting in B5 zig-zag chains (B1–B2–B3–B2–B1) (Figures f and S9) with B–B distances of 1.80–1.82 Å. These B–B distances are comparable to that found in elemental boron and other binary and ternary borides.[39,53,59,63−66]

Figure 6

(a) Crystal structure of LiNi12B8 represented as boron-centered Ni polyhedra. Coordination environment of (b) B1; (c) B2; (d) B3; and (e) B4. (f) B1-, B2-, B3-centered polyhedra are connected by face sharing, resulting in B5 chains (one such chain is drawn in red on top of the polyhedra, right); (g) B4-centered polyhedra are linked by corner sharing.

Interestingly, a similar kind of zig-zag [B5] chain is observed in the crystal structure of Ni12AlB8,[67] which is orthorhombic (space group Cmce, a = 10.5265 Å, b = 14.527 Å, c = 14.5540 Å, Z = 12) with nine distinct B sites as opposed to tetragonal LiNi12B8 with four distinct B sites (Figure S10). Like LiNi12B8, the different B sites in the structure of Ni12AlB8 are surrounded by six, seven, and eight Ni atoms, with similar coordination environments. Comparison of the Li and Al-centered Ni polyhedra in LiNi12B8 and Ni12AlB8 reveals similar coordination environments (Figures and S10); however, the building blocks are oriented differently, differentiating one structure from the other (Figures and S11). Besides their structural dissimilarities, the synthetic procedures to obtain LiNi12B8 and Ni12AlB8 also differ. Ni12AlB8 can be obtained by a direct reaction from the element, but lower temperature annealing (1173–1223 K) is needed to stabilize Ni12AlB8, while higher temperatures (>1223 K) favor the formation of the τ-boride Ni20Al3–B6+4, x = 1.5–2.[67]

Electron Microscopy

The complex structure of LiNi12B8 was solved from the synchrotron PXRD data. An excellent fit of PXRD data to the structural model (Figure S8), sensible interatomic distances and typical coordination environments in the obtained structural model upholds the structural solution. To further confirm the structural model of LiNi12B8, we used ED and STEM. ED patterns collected from the main crystallographic zones [001], [100], [110], and [111] (Figure a) confirm the tetragonal symmetry with a ∼ 10.5 Å and c ∼ 10.0 Å. Although the values of a and c are close, the high-angle annular dark field (HAADF) STEM images of [001] and [100] zones and their corresponding simulated images differ significantly (Figure ), in agreement with tetragonal symmetry. Similar values of a and c parameters in a tetragonal structure can render this structure as “pseudo-cubic”, therefore causing twinning. Such twinning and corresponding twin boundaries are observed in ED patterns and HAADF-STEM images (Figure b) for the [100] zone, where two grains are flipped along the <011> plane resulting in the splitting of diffraction spots in the [100] ED pattern (Figure b, inset).

Figure 7

(a) Electron diffraction patterns of the four main zones of tetragonal LiNi12B8 structure. (b) Antiphase boundary defect along the <011> plane visible from the ED pattern of the [100] zone and HAADF image. Ni and Ni/Li sites are shown in blue, Li sites in yellow, and B sites in red.

Figure 8

High-resolution HAADF-STEM images of the five main zones with magnified images, simulated patterns of the LiNi12B8 structure, and the structure overlays. Ni and Ni/Li sites are shown in blue, and Li sites, in yellow.

The observed intensity for a HAADF-STEM image is directly proportional to the atomic number squared so that heavier atoms (Ni, in this case) appear brighter, while light atoms (Li and B) are essentially invisible. On the contrary, light atoms (Li and B) can be visualized in ABF-STEM images and appear as dark gray dots together with darker Ni atoms. HAADF-STEM images for the five main crystallographic zones ([001], [100], [101], [110], and [111]) are shown in Figure along with the simulated images and corresponding structure overlays, where Ni or Ni/Li sites are depicted as dark-blue spheres and Li as yellow spheres. The simulated and experimental HAADF-STEM images match extremely well, confirming the positions of Ni atoms in the structural model obtained from synchrotron PXRD data. In addition, the columns built from Ni atoms are emphasized in Figure . The characteristic distances between these columns along the different directions determined from the HAADF-STEM images are 0.73 nm for [001], 0.75 nm for [100], 0.5 nm for [101], 0.5 nm for [110], and 0.26 nm for [111] zones, which align excellently with the expected values from the structural model. Since HAADF-STEM images are dominated by heavier atoms (here Ni), structural models with a similar Ni environment cannot be distinguished. Owing to this, HAADF-STEM images were not useful in differentiating ideal (LiNi12B8, without Li/Ni mixed occupied sites) and disordered (Li(Ni12–Li)B8 with Li/Ni mixed occupied sites) structural models.

On the other hand, the ABF STEM analysis is suitable for imaging light elements. Thus, ABF-STEM images were acquired to better visualize positions of lighter atoms (Li and B) in the structure. The ABF-STEM image in the [110] crystallographic zone is consistent with the simulated image (Figure S12), confirming the proposed structural model. However, given the complexity of the structure and that the STEM images are 2D projections of the structure, we cannot resolve the intricate Ni/Li disorder using STEM since for all of the crystallographic viewing zones, the Ni and Li sites are located within the same column when viewed in the 2D projections.

A plate-like crystallite morphology of LiNi12B8 is evident from SEM images (Figure S13). Energy-dispersive X-ray analysis (EDX) verifies the presence of only Ni and B, whereas detection of Li and accurate quantification of B cannot be performed due to limitations of the EDX technique. Signals for Nb (container used for synthesis) or any other heavy elements besides Ni are not detected in the sample, proving that LiNi12B8 is not a Nb-stabilized phase. The homogeneous distribution of Ni and B atoms can be clearly seen from the false-color elemental maps collected from different sample areas (Figure S13).

Solid-State NMR Spectroscopy

HAADF-STEM was extremely useful in confirming the proposed structural model but only with respect to the atomic positions of heavy Ni atoms. We further utilized 7Li and 11B solid-state magic angle spinning (MAS) NMR spectroscopy to confirm the positions of light atoms in the structure. There are seven crystallographically independent Li sites, with five of those sites being mixed occupied by Li/Ni. There are four crystallographically independent B sites in the structure. Importantly, both MAS 7Li and 11B NMR spectra of LiNi12B8 are distinct from the NMR spectra of other known ternary borides in the Li–Ni–B system (Figure S14).[39,53,59] In the MAS 7Li solid-state spectra (Figure ), a total of eight peaks are observed, two of them are attributed to the two fully occupied Li sites in the LiNi12B8 structure (marked in red). The peak around 22 ppm (marked with *) corresponds to the small amount of the impurity phase LiNi3B1.8 (Figures and S14).[67] The remaining five peaks (marked in blue) in the 7Li NMR spectra are assigned to the five Li/Ni mixed occupied sites within the LiNi12B8 phase. Considering the multiplicity of each Li/Ni (M) occupied site and the two fully occupied Li sites (Table S1), the expected integrated intensity values for LiNi12B8 are M1/M2/M3/M4/M5/Li1/Li2 = 0.896(3):0.352(4):0.152(3):0.34(5):0.152(5):4: 2 or 0.224:0.088:0.038:0.085:0.038:1:0.5, when scaled to unity. From the MAS NMR spectrum, the peak intensities’ qualitative ratio is 0.41:0.24:0.08:0.18:0.06:1:0.59 (Figure S15), which is in somewhat reasonable agreement with the proposed model obtained by XRD refinements. However, given the significant overlap of 7Li peaks in the spectrum, it is challenging to perform quantitative integration/deconvolution. Yet, the presence of more than two MAS NMR 7Li peaks is in line with the disorder associated with Li/Ni mixed occupancies, rather than the fully ordered LiNi12B8 structure with only two Li sites. The MAS 7Li NMR chemical shifts for all eight peaks are in the range of −20 to +70 ppm, indicating a Knight shift from the conduction electrons and suggesting the metallic conductivity of the LiNi12B8 phase.[59] This behavior is in good agreement with the predicted metallic behavior from the band structure and electronic density of states (vide infra) (Figure ). Therefore, the solid-state 7Li NMR spectrum supports the proposed structure and metallic behavior of LiNi12B8.

Figure 9

7Li (top) and 11B (bottom) solid-state NMR spectra of the LiNi12B8 compound. In the 7Li NMR spectra, the peaks marked in red are considered for two fully occupied Li sites, five peaks marked in blue are for the five Li/Ni mixed occupied sites, and the peak in black marked with an asterisk is for the impurity LiNi3B1.8 phase. Experiments were performed at 9.4 T and 25 kHz MAS.

Figure 10

(a) Band structure and electronic density of states (red dashed line indicates the fermi level); (b) phonon density of states.

11B NMR spectra were acquired with two MAS frequencies to confirm the assignment of three distinct isotopic peaks in the LiNi12B8 compound (Figures and S16). The proposed structural model has four independent crystallographic B sites. This discrepancy can be accounted for by considering the coordination environment of boron atoms. In the proposed structural model, B1 and B4 atoms, from 16n and 8h Wyckoff sites, respectively, have the same coordination environment (square antiprismatic polyhedra), consisting of eight Ni atoms. Since atoms in a similar chemical environment likely have similar chemical shifts, these two boron atoms probably generate peaks in the same chemical shift range. Taking into account this assumption and the multiplicity of B1 and B4, the expected ratio of integrated intensities of B peaks in the NMR should be (B1 + B4)/B2/B3 = (16 + 8):16:8 or 1:0.667:0.333 (scaling to unity). These values are in good agreement with the qualitative integrations (1:0.51:0.31) obtained by fitting the isotropic 11B peaks in the solid-state 11B NMR spectrum (Figure S17). Due to the complex interplay between 11B quadrupolar coupling, Knight shift anisotropy, and dipolar coupling, quantitative fitting of the spectrum is challenging. Nevertheless, the range of 11B chemical shift of 190–290 ppm for the LiNi12B8 compound is consistent with its metallic behavior, which in turn is in line with the calculated band structure (vide infra) (Figure ) and 7Li NMR spectrum.

Electronic and Phonon Structure

To assess the thermodynamic stability of the ideal (without Li–Ni mixed occupied sites) LiNi12B8 compound, formation energy is calculated by the DFT calculations. The proposed structural model of the LiNi12B8 compound is stable with a negative formation energy of −308 meV/atom. This formation energy is very close to the formation energy value reported for the RT-LiNiB (−339 meV/atom) and HT-LiNiB (RT-LiNiB (−314 meV/atom)).[39] According to the electronic structure calculations (Figure ), the compound is predicted to be metallic, which agrees with the Knight shifts in 7Li and 11B NMR spectra. The 3d atomic orbitals of Ni significantly contribute to the electronic density of states, followed by a contribution from 2p atomic orbitals for B, while the contribution of 2s Li atomic orbitals is negligible (Figure S18). The proposed structure of the LiNi12B8 compound is dynamically stable since no imaginary phonon frequencies are observed in the calculated phonon spectrum (Figure b).

Magnetic Properties

Studies of the magnetic properties of LiNi12B8 are of interest due to an extended Ni framework in its crystal structure with an overall high fraction of the 3d metal. According to the temperature dependence of the molar magnetic susceptibility plot, the title compound displays paramagnetic behavior in the 30–300 K temperature range (Figure a,c,d). An abrupt increase in magnetic susceptibility below ∼24 K with the decreasing temperature suggests magnetic ordering (Figure a). Additionally, in the low-temperature region (<30 K), the divergence of field-cooled (FC) and zero-field-cooled (ZFC) data indicates either a degree of ferromagnetic ordering or a superparamagnetic transition, or a spin-cluster, or a spin-glass-like behavior. Given the similarity of the M(T) and M(H) data to our earlier work on RT-Li0.6NiB and HT-Li0.4NiB,[38] the cluster- or spin-glass scenario seems more likely in this case (Figure S19). Astonishingly similar magnetic behavior of structurally distinct compounds HT-Li0.4NiB and LiNi12B8 may arise due to the disorder in the Ni environment, where disorder in HT-Li0.4NiB arises due to the random distribution of multiple substructures Li[NiB] (n = 1, 2, 3); LiNi12B8 features mixed occupied Ni/Li sites (Figure S19). With an increase in the applied magnetic field from 50 to 1000 and 2000 Oe, the ZFC/FC bifurcation temperature decreases from ∼18 to ∼12 K and finally to ∼9 K.

Figure 11

(a) Temperature-dependent ZFC/FC molar susceptibility of the polycrystalline LiNi12B8 sample in the 2–300 K temperature range. Inset: temperature-dependent molar susceptibility in the 2–50 K temperature range; (b) field-dependent magnetization at 2, 100, 200, and 300 K; (c) using M/H vs T data at 2 and 4 Tesla, the contribution from ferromagnetic impurity (estimated to be <0.01%) was calculated by the Honda-Owen method to obtain the intrinsic susceptibility of the LiNi12B8 compound. (d) Data corrected with the Honda Owen method were fitted with the modified Curie–Weiss equation. Inset: M vs H plots in the 0–4 Tesla range for 100, 200, and 300 K temperatures.

The field dependence of magnetization at 100, 200, and 300 K is linear above 1 T with no tendency for saturation, consistent with paramagnetic behavior (Figure b). The very slight nonlinearity of magnetization in the low magnetic field of 0–0.5 T, even at 300 K (Figure S20), could be attributed to the presence of a trace amount of ferromagnetic impurities. However, fundamentally, these samples appear to have significantly less ferromagnetic contamination than the HT-Li0.4NiB precursor.[38] On the other hand, the hysteresis loop in the magnetization vs field plot at 2 K suggests magnetic ordering at low temperatures. To determine the intrinsic molar magnetic susceptibility of LiNi12B8, χ at T > 100 K, the contribution of ferromagnetic impurities was subtracted by the Honda-Owen method (Figure c).[68,69] For this, the M/H vs T data were collected at a high enough magnetic field (2 and 4 T) to fully saturate ferromagnetic impurities (Figure d inset) and were linearly extrapolated to an infinite field (1/H → 0). The content of the ferromagnetic impurity is estimated to be 0.002 wt % of Ni metal (under the detection limit of the PXRD experiment), again showing LiNi12B8 to be less contaminated than the starting material HT-Li0.4NiB. The intrinsic magnetic susceptibility for LiNi12B8 at T > 100 K was fitted to the modified Curie–Weiss equation (Figure d): χ = χ0 + C/(T – θ) where χ0 = 6.938(3) × 10–4 emu mol–1, C = 1.004(7) × 10–1 emu K mol–1, and θ = 39.2(4) K. The positive value of the temperature-independent contribution, χ0, is consistent with a Pauli-paramagnetic (metallic) contribution, as seen by the calculated electronic band structure and Knight shifts in 7Li and 11B MAS NMR spectra. The positive value of the Weiss constant, θ, indicates weak ferromagnetic coupling between the adjacent Ni spins. The magnetic moment per Ni atom, calculated from the Curie constant, C, amounts to 0.90(1) μB/per f.u., which becomes 0.075(1) μB/Ni atom. This effective magnetic moment is lower than the effective magnetic moment of Ni metal (0.6 μB/Ni) as well as the spin-only magnetic moment of Ni2+ (2.83 μB) and Ni3+ (1.73 μB) ions. Compared to the other known ternary borides in the Li–Ni–B system, the μB/Ni atom for LiNi12B8 is lower than that of RT-Li0.6NiB, HT-Li0.4NiB, and LiNi3B1.8 but higher than that in the Li2.8Ni16B8 phase (Tables S2 and S3).

Interestingly, magnetic ordering manifesting in ZFC/FC splitting at a low temperature and magnetic field is also seen in other ternary compounds in the Li–Ni–B system: RT-Li0.6NiB, HT-Li0.4NiB, LiNi3B1.8, and Li2.8Ni16B8.[38,59] For the Li–Ni–B compounds obtained by high-temperature annealing (LiNi3B1.8, LiNi12B8, and Li2.8Ni16B8), the temperature of ZFC/FC splitting at 50 Oe linearly increases with Ni at.%, while LiNiB polymorphs (33.33 at.% Ni) lack such magnetic transition (temperature-independent paramagnets), and RT-Li0.6NiB and HT-Li0.4NiB obtained by room-temperature deintercalation exhibit a different trend (Figure S21).

Conclusions

In this study, the metastable layered borides RT-Li0.6NiB and HT-Li0.4NiB obtained by partial Li deintercalation from the LiNiB polymorph at room temperature were further subjected to elevated temperatures to investigate the extent of Li deintercalation and to achieve individual, Li-free 2D NiB layers, NiBenes. We found that heating of HT-Li0.4NiB causes the topological transformation of NiBene layers leading to the formation of the novel metastable compound LiNi12B8 with a new tetragonal structure type. After systematic analysis of in situ high-temperature synchrotron and ex situ laboratory PXRD, we determined that the LiNi12B8 phase is extremely sensitive to annealing temperature and dwelling time. We established that LiNi12B8 cannot be synthesized from the direct reaction of elements and irreversibly decomposes above 973 K, indicating that this compound is indeed metastable. The crystal structure of the compound was solved from high-resolution synchrotron PXRD data and further confirmed by HAADF-STEM and ABF-STEM. A synergistic combination of high-resolution synchrotron PXRD, STEM, ICP-MS, and solid-state 7Li NMR spectroscopy revealed the presence of Ni/Li mixed occupied sites, establishing the formula as Li(Ni12–Li)B8 (x = 0.317(2)). Owing to the presence of an extended Ni framework and a high fraction of 3d-metal, the LiNi12B8 compound exhibits magnetic ordering below 24 K.

This report illustrates unconventional means toward the synthesis of a metastable compound LiNi12B8 and introduces a contemporary twist on traditional structure solution by combining the complementary information provided by an array of characterization techniques. This study illustrates a state-of-the-art recipe to unambiguously solve crystal structures of microcrystalline materials comprising light elements (Li and B) by utilizing divergent techniques (PXRD, STEM, NMR, ICP-MS, DFT) in conjunction. These findings further emphasize the importance of unconventional synthesis methods in accessing novel metastable compounds that are hardly attained by traditional high-temperature solid-state synthesis.