Introduction

Topological polar textures are topological defects and domain patterns intrinsically embedded in ferroelectric materials. Their recent advances have demonstrated flux-closure domains1,2,3,4, vortices/antivortices5,6,7,8,9,10,11,12,13,14, electrical bubbles15,16,17,18,19, merons20,21, polar skyrmions22,23,24,25,26,27,28,29,30, hopfions31, and others32,33,34,35,36, which provide an excellent opportunity for studying nanoscale topology and emergent polar properties. These complex polar textures are condensed by the delicate balances between elastic, electrostatic, and gradient energies37,38. Notably, via the elastic energy penalty of substrate epitaxial strains, particle-like polar skyrmions with continuous polarization rotation across the cores and peripheries can be stabilized in oxide ferroelectric heterostructures23,24,25,26,27,28,29. Such skyrmionic textures present rich physics (e.g., negative permittivity24 and chirality29), and prospective device applications, including ultrahigh-density memristors above a gigabit per square inch27, both of which so far have garnered considerable interest.

Although van der Waals ferroelectric crystals provide a new ingredient beyond oxide ferroelectric heterostructures for fundamental and technological studies, their progress in polar skyrmions remains elusive due to the challenges of manipulating intricate boundary conditions. Here, we observe switchable skyrmions in layered ferroelectric CuInP2S6 crystal with thicknesses ranging from the bulk (i.e., 180 nm) to the atomic layers of about 8 nm. By reducing the crystal thickness, a topological transition occurs and the skyrmions incrementally evolve from elliptical to circular textures. Unlike previously reported polar skyrmions24,25,26,27,28,29, we do not intentionally introduce additional elastic energy from the substrates to support the presence of CuInP2S6 skyrmions. Instead, material intrinsic properties from CuInP2S6 can be leveraged to modulate the boundary conditions and stabilize polar textures. As a typical ferroionic material, CuInP2S6 is usually coupled with the migration of copper ions39,40, leading to the charge accumulation. Phase-field simulations and experimental evidence show that the charge-related energy penalty (e.g., copper ion-induced electrostatic energy) contributes to the stabilization of polar skyrmions in CuInP2S6 crystals.

Results

In this work, we engineered the composition of van der Waals CuInP2S6 material to increase the amount of copper ions, and attempted to approach the stoichiometric composition. This design can suppress phase separation from CuInP2S6 and In4/3P2S641,42 and thus induce the evolution of ferroelectric domain configurations (Supplementary Fig. 1 and Note 1). Bulk CuInP2S6 crystals were synthesized using the chemical vapor transport (CVT) method, and thin flakes were exfoliated onto gold-layer-covered silicon substrates (see Methods). Basic characterization by second harmonic generation (SHG), Raman spectra, single-crystal X-ray diffraction, and piezoelectric force microscopy (PFM) switching loops were systematically presented in Supplementary Fig. 1, 2 and Tables 1, 2, confirming the stoichiometry, structural asymmetry, and ferroelectricity. Remarkably, the SHG results for the CuInP2S6 crystal used in this work (i.e., S3: Cu0.90In0.99P2S5.90) show highly polar anisotropy, reflecting the distinct noncentrosymmetry that supports our following observations43. Figure 1a shows typically monoclinic CuInP2S6 crystal structures with highlighting two polarization orientations (POOP and PIP). Unlike common ferroelectrics44,45, CuInP2S6 crystals combine both properties of ferroelectric polarization and ionic conductivity (i.e., ferroionic), in which ions serve as a new knob to interact with polarization switching. Under external stimuli, Cu ions can accumulate over the crystal surface and increase the interfacial charge density39,40. Despite many reports on the out-of-plane polarization of CuInP2S6 crystal42,43,46, small in-plane polarization does exist, which has been revealed by PFM measurement and theoretical calculations47. Following previous works47,48,49, in-plane and out-of-plane polarization arises from the opposite motions between the In and Cu atoms. We conjecture that the orthogonal polarization directions (i.e., in-plane and out-of-plane dipoles) and the energy from abundant charges, including copper ions, likely benefit the polarization curling of polar textures.

Fig. 1: Observation of elongated polar skyrmions in CuInP2S6 crystals.
figure 1

a Schematics of the CuInP2S6 crystal structure. The antiparallel motions of Cu and In atoms result in out-of-plane and in-plane dipoles. b Large-area lateral piezoelectric force microscopy (L-PFM) amplitude and topographic mapping of coffee-bean-like polar textures on the surface of a 160-nm-thick CuInP2S6 crystal. The sample was scanned along the horizontal direction, and the detected polarization by PFM probe is along the arrow (bottom-left). c L-PFM and vertical piezoelectric force microscopy (V-PFM) mapping for the top-left region in b. d Enlarged dipolar configurations marked in red in (c). Schematic of the dipolar configurations is depicted at the bottom of (d). e Annular dark-field STEM image for identifying skyrmionic features. f Phase-field simulation of in-plane and out-of-plane polarization configurations for elongated polar skyrmions. The results resemble L-PFM and V-PFM responses. P represents normalized polarization.

Full size image

For exploring topological polar textures, we used the high-resolution vector PFM technique, simultaneously detecting vertical and lateral electromechanical responses, to scan CuInP2S6 flakes. As shown in Fig. 1b and Supplementary Fig. 3, we observe spontaneous particle-like nanodomains which are orderly and self-assembled over a large area. Such an ordered arrangement may stem from material anisotropy, as revealed by the phase-field simulations (Supplementary Note 2). To capture the detailed domain texture, we selected a small region and the acquired vector PFM responses are presented in Fig. 1c. From the lateral PFM (L-PFM) amplitude images, it is seen that the nanodomain structures present a regular coffee-bean-like shape, whose sizes range from ~250 nm in length and ~30 nm in width to a smaller length of less than 50 nm. These different sizes are similar to those of magnetic skyrmions50,51,52, and suggest the possibility of polar texture propagation. Remarkably, L-PFM phase mapping shows a strong contrast of nanodomain responses against the background, in which all nanodomains exhibit distinguishable upper and lower portions with antiparallel in-plane polarizations. However, the nanodomain structures in vertical PFM (V-PFM) phase mapping demonstrate a slightly smaller size than that of L-PFM responses, manifesting polarization curling at the core area of the particle domains. The L-PFM and V-PFM observations are crucial to envisioning and constructing the rotated polarization vector of these nanodomains. We note that these PFM results are not entirely associated with the topographic variation, ruling out the influence of the sample thickness in PFM phase mapping. In addition, we estimated the effective piezocoefficients from PFM amplitude mapping as shown in Supplementary Fig. 3b, in which in-plane and d33 piezocoefficients are around 1–3 and 4–6 pm/V, respectively. This d33 magnitude roughly coincides with previous works, e.g., 5–12 pm/V53, and 2.5 pm/V49; the commensurate in-plane piezocoefficient is likely caused by our composition engineering of CuInP2S6 crystal.

When looking into a single nanodomain (Fig. 1d), the in-plane polarization can be considered a head-to-head type (i.e., an upward-to-downward configuration). Their contour size in the L-PFM phase closely resembles the coffee-bean-like amplitude. From V-PFM phase and amplitude, we deduce that the detected polarization should orient inward due to the specific piezoelectric effect in CuInP2S652,54. These observations indicate that the ordered nanodomains are probably in a hedgehog-like state with Néel-type polarization rotating gradually from the core to the periphery (see the following images for more evidence). To intuitively visualize the texture, we show the schematic of an elliptical Néel-type skyrmion in the bottom panel of Fig. 1d, which is analogous to those observed in ferromagnetic systems52,55. This emergent topological domain can match with the PFM results well, and has never been observed in conventional ferroelectric thin films. The elliptical shape arises from copper-ion induced anisotropy noncentrosymmetry (See Supplementary Fig. 1 and the following theoretical explanation). Strikingly, over a large area of 35 × 40 μm2, such polar skyrmionic solitons can uniformly exist (Supplementary Fig. 3c), which is promising for creating ultrahigh-density memory devices for neuromorphic devices with thousands of resistance states. Moreover, we probed the presence of topological textures by scanning transmission electron microscopy (STEM). From the dark-field STEM image in Fig. 1e, orderly elongated shapes are observed with a length of up to three hundred nanometers, which are analogous to our PFM finding. Selected area electron diffraction (Supplementary Fig. 3d) shows a single-crystal pattern and completely rules out the generation of a new phase at the elliptical area.

After experimentally observing CuInP2S6 polar skyrmions, we next rationalize their existence using a phase-field model based on the time-dependent Ginzburg-Landau equation. In order to model surface charges of CuInP2S6 flakes induced by the migration of Cu ions39,40, we employed a charge density of −0.0255 C/m2 (Supplementary Note 3), below which the skyrmionic patterns cannot be intrinsically condensed but above which topological patterns undergo a transition (see theoretical simulations in Supplementary Fig. 7). Figure 1f and Supplementary Fig. 3e show the successful acquisition of simulated elliptical polar skyrmions that are orderly aligned in one direction. In skyrmion area, the elongated configurations of in-plane and out-of-plane polarization highly coincide with the experimental L-PFM and V-PFM findings in Fig. 1c. However, away from skyrmionic area, zero in-plane polarization was setted for the simplicity of the simulation. As expected, these results reveal that surface charges likely induced by Cu ions play a crucial role in the formation of polar skrymions, which is experimentally substantiated by Raman spectra (Supplementary Fig. 1). Thus, the canted polarization over the skyrmionic area can be mainly ascribed to the electrostatic energy penalty. In addition to the influence of surface charges, the strong anisotropy in CuInP2S6 flakes is responsible for the ordered arrangement of elliptical polar skyrmions, which is induced due to the anisotropic material parameters in phase-field simulations (see Supplementary Note 2 and Supplementary Table 3). Moreover, the topological nature of skyrmionic polar textures is commonly described by an invariant called the skyrmion number (N)37:

$$N=\frac{1}{4{{\rm{\pi }}}}\iint {{\bf{n}}} \cdot \left(\frac{\partial {{\bf{n}}}}{\partial x} \, \times \, \frac{\partial {{\bf{n}}}}{\partial y}\right){dxdy}$$
(1)

where n denotes the unit polarization vector, and the integrand is the Pontryagin charge density in the xy plane. The Pontryagin charge density mapping of the elliptical polar texture is displayed in Supplementary Fig. 3e, which results in an integer skyrmion number of −1. From this perspective, the topological classification of the observed elliptical polar skyrmions is equivalent to that of magnetic56 and other polar skyrmions22,23,24,25,26,27,28,29 in previous studies.

To further confirm the elliptical texture of polar solitons, we performed angle-dependent PFM mappings by rotating the CuInP2S6 crystal with a clockwise direction. Figure 2 exhibits the V-PFM response at 0° and the L-PFM responses at 0°, 45°, and 90°, respectively, in which the images were adjusted to accommodate the 0°-scanned results. Related raw PFM images under different rotation angles are shown in Supplementary Fig. 4, and the rotation can be clearly identified by the two ends of coffee-bean-like shape. High-frequency torsion of the PFM probes allows the precise detection of in-plane electromechanical responses, indicative of in-plane polarization. In particular, for 45° and 90° scanning, the detected in-plane amplitude responses deviate slightly from the 0° coffee-bean-like pattern. For example, the 45° L-PFM mapping shows rectangle-like contours in which the bright pattern is divided by a dark diagonal. The 90° L-PFM mapping displays two contrasted ends of a rectangle-like shape while most areas exhibit dark responses. As illustrated in Fig. 1d, such divergences in shapes can be well interpreted by the polarization curling for elongated polar skyrmions. Following previous studies27,57,58, we plotted the polarization vector based on PFM mapping with respect to the scanning angles (Fig. 2b and Supplementary Note 4). Five center-convergent configurations can be clearly identified, demonstrating the polarization rotation and vertices over the surfaces. This also signifies the formation of polar skyrmions in ferroelectric CuInP2S6 crystals. Apart from angle-dependent measurement, we also scanned topological solitons with different spring-constant probes and different laser spots to exclude the impact of scanning artifacts. As shown in Supplementary Fig. 5, all scenarios can reproduce the elliptical skyrmions.

Fig. 2: Confirmation of the 3D structure of CuInP2S6 polar skyrmions.
figure 2

a. Cantilever-angle dependent PFM mappings with five center-convergent polar skyrmions. The V-PFM responses at 0° and L-PFM responses at 0°, 45°, and 90° are depicted, and α indicates the azimuth angle between cantilever and sample orientations. b Polarization vector mapping constructed by the L-PFM responses in (a). See Supplementary Information for construction details. c. MAADF-STEM image showing similar skyrmion shapes with PFM results. d. Atomic resolved HAADF-STEM image showing the boundary between skyrmion and non-skyrmion area. The dashed line indicates the transitional area.

Full size image

High resolution STEM was also employed to verify the topological solitons in CuInP2S6 crystals. Medium-angle annular dark-field (MAADF) STEM image (Fig. 2c) shows elliptical skyrmion shapes with a brighter periphery and darker core, similar to those observed by PFM measurement. For these topological structures, dark regions possessing out-of-plane polarization are surrounded by bright walls with in-plane polarization25,26. By magnifying the skyrmions, high-angle annular dark-field (HAADF) STEM successfully captured atomic-resolved structures (Fig. 2d), in which a white dashed line denotes the boundary between skyrmion and non-skyrmion area. Vector mapping of ion displacement, extracted from atomic-scale STEM images, is typically used to demonstrate polarization rotation, serving as strong evidence for studying skyrmions in oxide ferroelectrics24,25,26,27,28. However, for CuInP2S6 crystal, we argue that collecting ion displacement remains challenging due to the relatively small off-center distance. This can be revealed by the small polarization value of 5~11 μC/cm2,43,59 one-order-magnitude smaller than those in oxide ferroelectrics24,25,26,27,28.

Having demonstrated the presence of polar skyrmions, we now explore thickness-dependent topological textures (Fig. 3 and Supplementary Fig. 6). Thickness has been demonstrated as an effective approach to control polar textures in oxide ferroelectric heterostructures due to the strain modulation3,4,5,16,26,27,28. In line with this strategy, we find that, when reducing the CuInP2S6 thicknesses from ~160 nm to ~40 nm and then to −8 nm, polar skyrmions can gradually evolve from elongated shapes to a mixture of elongated and circular shapes (Fig. 3a and b) and then to circular patterns (Fig. 3c and d). For the circular skyrmions in Fig. 3d, the diameter is about 50 nm; center-divergent and center-convergent configurations as illustrated in Fig. 3e can be identified by blue and red dashed lines, respectively. These circular skyrmions can be condensed over a large area (Supplementary Fig. 7) and exhibit a higher density compared to elliptical polar skyrmions. Such a transition of skyrmionic shapes with respect to thicknesses is also substantiated by using STEM imaging. Besides the elliptical configuration (Fig. 1e), we find ultra-small circular skyrmions with a diameter of around ~5 nm in thin CuInP2S6 flakes (Fig. 3g), which are highlighted by white arrows. The atomic-resolved MAADF-STEM image (Fig. 3h) recorded the circle contrast on single crystal background, implying a non-trivial polarization structure as revealed in PFM measurement. A discrepancy between the sizes observed by STEM and PFM characterizations is mainly because the latter has a measuring resolution (−20 nm) restricted by probe radius27.

Fig. 3: Topology evolution with the reduction of CuInP2S6 thicknesses.
figure 3

Topographic, V-PFM, and L-PFM mappings for a 40-nm (a and b) and 8-nm (c and d) thick CuInP2S6. In (b), both elongated and round polar skyrmions have a center-convergent helicity. In (d), most circular polar skyrmions exhibit center-divergent properties, while only one, as marked in red, is identified as the center-convergent type. e, Schematics of center-convergent and center divergent circular skyrmions. f. Polarization configurations for simulated circular polar skyrmions. g Low-magnification STEM images displaying many circular features, which are marked by white arrows. h Atomic-resolved STEM mapping with skyrmionic lattice. White dashed lines represent typical skyrmions. i Statistic skyrmion sizes as a function of sample thicknesses. The data was extracted from PFM amplitude mappings. j Thickness-dependent Raman spectra. Copper vibrational peaks are labeled by blue dashed lines. k Statistic Raman shifts related to the thickness reduction for copper vibration peaks. The error bars in i and k indicate standard deviation.

Full size image

To confirm the thickness-determined topology, we conducted a statistical analysis on the skyrmion sizes taken from PFM mappings. The data includes but is not limited to these from Fig. 1–3 (see Supplementary Fig. 6 for additional PFM images). As expected, Fig. 3i shows a remarkable transition of topological textures, marked by blue dashed line. The underlying physics is systematically investigated by Raman spectra (Fig. 3j and k). With the thicknesses decreasing from 230 nm to 8 nm, Fig. 3j exhibits a pronounced red shift for copper vibrational peaks, particularly at around 316 /cm. In order to unambiguously identify the change, we plotted typical Raman shifts of copper ion-related peaks60,61 (i.e., 316 and 103 /cm) as a function of thicknesses. The Raman peaks become softened, displaying a dramatic shift by over 12 wavenumbers to 302.5 /cm and by over 1 wavenumber to 102 /cm. These are predominantly caused by the migration behaviors of copper ions, constituting possible evidence for charge induced topology change.

Phase-field simulations in Supplementary Fig. 7 also support our experimental observations on the thickness-dependent topology of polar textures. By increasing the surface charge density in the simulation, elongated skyrmions can be transformed from the labyrinth domains. When the charge density further increases with reduced thicknesses, the shape of the skyrmion changes from ellipse to circle. We note that the thinner flakes should have a higher local strain gradient during the mechanical exfoliation process, resulting in a more remarkable Cu migration because of the stronger flexoelectric field. Thus, the thinner flakes possess larger surface charge densities induced by Cu migration. As a result, tunable polar topologies with respect to the thicknesses can occur. For both elliptical and circular textures (Supplementary Fig. 7), they have a skyrmion number of unity, confirming their nontrivial topology.

The switching ability of polar skyrmions is essential for applications in high-density memory devices. We examined the dynamics of switching properties when applying electrical and mechanical stimuli. Figure 4a and b present the evolved dynamics of polar skyrmions with respect to different DC voltages applied onto the PFM probes. When subjected to negative DC voltages, we simultaneously used a small AC driving voltage to capture skyrmion patterns. V-PFM and L-PFM mappings manifest the mixture of elliptical and circular patterns (dashed red lines). We scrutinize the L-PFM phase under a − 2.5-V DC voltage, finding that the typical colors for upper and lower portions of skyrmions are deterministically switched. This signifies that the skyrmion has been reversed to a new state, specifically transitioning center-convergent skyrmion 1 to center-divergent patterns 1´ (Fig. 4d). When removing DC voltages (i.e., DC = 0 V), skyrmions can reversibly switch back to the initial state, indicating the intermediate, metastable properties of pattern 1´. With the electrical stimulus, the almost unchanged V-PFM responses in Fig. 4b suggest that skyrmion state 1 and 1´ only have the switching of polarization-rotation direction over the surface. Referring to magnetic skyrmions62, we term this variation the switching of helicity number from π to 0, as illustrated in Fig. 3d. However, with the application of positive DC voltages, we do not observe the similar switching behaviors (Supplementary Fig. 8a). To understand these differences, schematics of electric field-driven ion migration are depicted in Supplementary Fig. 8b. Negative DC voltages can attract copper ions to migrate towards the surface while positive DC voltages compel copper ions to move away from the surface63. Therefore, considering the migration nature of copper ions, the reversible switching of polar skyrmions shall arise from charge-related energy penalty. During electrical modulation, under low DC voltages topological protection prevents polar skyrmions from collapsing, while under high DC voltages polar skyrmions can be completely erased or destroyed. We note that this switching behavior holds immense potential for crafting high-density ferroelectric memory devices.

Fig. 4: Switchable CuInP2S6 polar skyrmions via the applications of electrical fields and mechanical forces.
figure 4

a, b, Topographic and PFM mappings demonstrating electrically switchable elongated and circular polar skyrmions under different DC biases. c Evolution of the dipolar configuration realized by continuous PFM scanning with an estimated force of −30 nN over 6 times. d Schematics of the transition between different skyrmionic states. The helicity number is switched between 0 and π.

Full size image

Interestingly, as the polar skyrmions were continuously scanned by PFM probes with an estimated force of ~30 nN for over 6 times (see Methods), the phase patterns dramatically change (Fig. 4c). This finding firmly demonstrates that mechanical stimuli can also be used to manipulate the helicity number of polar skyrmions. The underlying mechanism of the mechanical manipulation of polar skyrmions can be explained by the flexoelectric effect64. Due to the large strain gradient under the PFM tip, a localized flexoelectric field and relevant Cu ion migration exists, leading to the polarization switching over the skyrmion area.

Discussion

In conclusion, by increasing the density of copper ions in van der Waals ferroionic CuInP2S6 crystals, we have shown evidence for large-area polar skyrmions with charge-related energy penalty, and demonstrated their switching properties and topological transitions with respect to thickness variations. Our work highlights that 2D van der Waals ferroelectrics, without epitaxial-strain constraints, could be a new platform for exploring unexplored polar topology and neuromorphic devices with multilevel states and linear synaptic update.

Our approach of charge-related energy penalty can be generalized to a broad range of ferroionic materials, such as metal seleno- and thio-phosphate materials (i.e., ABP2X6, A = Cu and Ag, B = In, Cr, Bi, and Mn, X = S and Se)65 and traditional oxide ferroelectrics with abundant ion vacancies (i.e., BaTiO3)66. By delicately modulating the ion concentrations for achieving a favorable boundary condition, we anticipate that topological defects including polar vortices, polar merons, polar skyrmions, and so on, can also be stabilized in those materials (i.e., ABP2X6 and BaTiO3 with a deficiency of ion vacancies).

Methods

Sample preparation

High-quality CuInP2S6 single crystals were synthesized using the CVT method. Cu powders (99.5%, Sigma-Aldrich), In pieces (99.99%, Alfa Aesar), P chunks (99.99%, Sigma-Aldrich), and S powders (99.5%, Sigma-Aldrich), with different molar ratios, were loaded into a quartz ampoule. The materials were further mixed with a small amount of iodine granules (99.8%, Sigma-Aldrich), which functioned as the transport agent in the CVT process. Then, the ampoule was vacuumized to below 20 mTorr and flame-sealed. During the growth process in a tube furnace, the source zone and growth zone were kept at 750 °C and 700 °C for 7 days, respectively. The CuInP2S6 thin flakes were mechanically exfoliated onto conductive substrates coated with Au layer (30 nm) using Scotch tape.

PFM measurement

The PFM measurements were performed using the Asylum Research MFP-3D and Cypher. Both the dual AC resonance tracking PFM mode and vector PFM mode were used to acquire useful electromechanical responses. Commercial soft conductive probes (e.g., AC 240 TS with a spring constant of 2 N/m and ASYELEC-01-R2 with a spring constant of 2.8 N/m) were adopted to avoid damaging sample surfaces. AC biases of 0.2 to 0.5 V were applied onto the PFM probes to image polar textures with scanning points of 320 to 512 and a scanning rate of 1.5 Hz. Vector PFM mode can simultaneously record out-of-plane and in-plane electromechanical responses at a free-air resonance and contact resonance frequency, respectively. In particular, for in-plane responses, the dipole perpendicular to the PFM cantilever can be precisely detected by the torsional vibration of PFM probes. When conducting the angle-dependent PFM measurement, the CuInP2S6 sample was manually rotated, and the exact angles were calibrated by comparing the initial and rotated PFM mapping. When carrying out the mechanical switching measurement (Fig. 3a), the AC probe biases were turned off except for 1st and 6th scanning.

TEM measurement

For the TEM analysis, the exfoliated thin CuInP2S6 flakes were directly transferred onto a silicon nitride membrane grid (CleanSiN) on an accurate transfer platform (Metatest, E1-G) using polydimethylsiloxane. The observations were performed by spherical aberration-corrected electron microscopy on a FEI Titan G2 80-200 ChemiSTEM (30 mrad convergence angle, 0.8 Å spatial resolution). Annular dark-field scanning transmission electron microscopy was carried out. The collection semi-angles were 37–200 mrad (for MAADF) and 56-200 mrad (for HAADF).

SHG measurement

The SHG measurements were conducted using a 1030 nm femtosecond-pulsed laser (pulse width of −200 fs and pulse frequency of 1 MHz, YactoFiber-FL-20) by a homemade optical microscope with a 50X near-infrared transmission lens with a spot size of less than 1 μm. The polarization of generated second harmonic 515 nm light was selected in parallel conditions relative to the fundamental beams using individual linear polarizers and analyzed using a compact integrated and aberration-free spectrograph (Princeton Instruments, FERGIE). The polarization direction of the incident light field was adjusted by the rotation of the λ/2 waveplate driven by a rotating motor.

Raman measurement

The Raman spectra were carried out using LabRAM HR Evol (Horiba). The used laser and grating are 532 nm and 1800, respectively.

XRD measurement

The XRD θ-2θ scan was performed at room temperature using a Bruker D2 PHASER diffractometer with Cu Kα radiation (λ = 1.54184 Å). The single-crystal XRD experiment was carried out at 120 K on a Bruker D8 Venture diffractometer with Mo Kα radiation (λ = 0.71073 Å).