Facet-Specific Liquid-Phase Exfoliation of an Ionic 1D Crystal, (NbSe₄)₃I, into Ultrathin Nanoribbons

Abstract

Liquid-phase exfoliation (LPE) of emergent materials composed of weakly bound one-dimensional (1D) and quasi-1D (q-1D) building blocks presents a straightforward route not only for the discovery of confined physical states in 1D but also for the realization of scalable functional devices. However, compared to the more established routes in two-dimensional (2D) crystals, the nature of LPE in 1D and q-1D crystals presents a more random process. This distinction arises from the various available interchain directions across several crystallographic facets unique to 1D and q-1D solids, from which the chains can be cleaved apart into a stochastic combination of nanowires, nanoribbons, and nanosheets. Using the 1D ionic phase comprised of ∼4.3 Å thin chains, (NbSe₄)₃I, we demonstrate herein the profound influence of crystal morphology, exposed facets, and their degree of wettability, passivation, and surface roughness in directing the LPE behavior of 1D crystals. Through the growth of bulk crystals as long needles with exposed (hk0) facets or as quasi-2D flakes with exposed (00l) facets susceptible to passivation, we show that these two distinct precursor morphologies display divergent behaviorboth in solvent preference and quality of resulting nanostructures. Under optimal conditions involving bulk needles and tetrahydrofuran as solvent, we show that the LPE of (NbSe₄)₃I results in ultrathin nanoribbons with high aspect ratios bearing lengths >5 μm, thicknesses down to 7.2 ± 2.6 nm, and widths of 26.4 ± 10.9 nm. The nanoribbons, solution processable as thin films, retain their native crystal structure and semiconducting character. Moreover, the nanoribbons also manifest pronounced degrees of bending and substrate-driven twisting at the nanoscale while maintaining long-range order. These results highlight a means to understand the fundamental chemical and physical behavior of noncovalently bound 1D solids through the realization of solution-processable 1D nanoribbons and nanowires that also have the potential as components for next-generation devices that approach the atomic scale.

Introduction

Low-dimensional materials, owing to their physical tunability, chemical modularity, and ultrathin form factors, are uniquely positioned to become building blocks for next-generation devices that approach the atomic scale. − These materials are intrinsically defined by distinct chemical bonding motifs: materials with strong and covalent bonding in two directions and weak interactions in one direction are referred to as two-dimensional (2D) materials, while materials with strong covalent bonding in only one direction are denoted as one-dimensional (1D). These materials comprise building blocks that are weakly held together by noncovalent interactions such as van der Waals (vdW) or ionic bonding, which readily allow for their exfoliation into few or single building units via top-down approaches. − Unlike the nanomaterials derived from three-dimensional (3D) crystals that currently constitute common functional devices, low-dimensional materials have well-defined surface chemistry that can preserve their native crystalline order even at thickness that approaches the single atom limitwithout the introduction of dangling bonds or major passivation surfaces.

In the search for low-dimensional materials with smaller form factors and dimensionalities, the recent years have seen a resurging interest in crystalline inorganic materials that are comprised of atomically precise and weakly bound 1D and quasi-1D (q-1D) chains. , The continuing discovery of new classes of these materials that have been experimentally demonstrated to reach the single chain regime now cover semiconductors, − unique semimetals, − ordered magnets, − and structures that display helical motifs − and have even been shown to manifest emergent properties that are more pronounced in 1D such as topological states − and robust charge density waves. , Beyond their discovery, advances in high-accuracy and high-resolution metrology have allowed for the comprehensive study and understanding of these ultranarrow structures. From a synthesis, fabrication, and processing standpoint, while bottom-up methods such as chemical vapor deposition, , vapor–solid or vapor–liquid–solid growth, − substrate-directed growth, and growth within nanotubes have been employed to access nanoscale analogues of 1D and q-1D crystals, these approaches have been currently limited to nanostructures that have thicker diameters, sporadic, nonuniform, or complicated by the interfaces from which they are formed, thereby currently limiting their applications in substrate-scale devices. Furthermore, the bottom-up synthesis of such 1D and q-1D nanostructures is further complicated by the complex interplay between their nucleation and growth dynamics, as well as the unique capability of these solids to grow along the intrachain and the various interchain axes. While more stochastic, the top-down exfoliation of bulk material is generally considered to be the most scalable method for processing large batches of nanomaterials. If controlled, this approach has the potential to access a narrow distribution of nanostructures, as the size of exfoliated structures is always capped by the thickness of the building unit.

To date, liquid-phase exfoliation (LPE) has been the most promising method for scalable isolation of nanosheets from two-dimensional (2D) materials, such as in graphene and transition metal dichalcogenides. ,, The mechanism of exfoliation in these materials is well understood and generally arises from the microscale-to-nanoscale fracturing and delamination of the layers caused by sonication-driven microcavitation of a suitable solvent. Existing knowledge of LPE relies on surface tension matching between the material of interest and the solvent, a bulk property that is imposed by atomic-scale intermolecular forces that are intrinsic to the surface character of the material and are not often thought to be influenced by the morphology of the precursors. This understanding applies well to 2D materials, where there is a continuous and uniform surface exposed along the basal plane of the material. However, crystals composed of highly anisotropic and noncovalently bound 1D and q-1D chains present an intrinsic challenge and opportunity in that covalently bound chains that comprise the bulk crystals can be cleaved along various crystallographic axes perpendicular to the chain direction. This is in stark contrast to 2D materials that are generally cleavable only between the layers, and it becomes more important to define what a 1D or q-1D “bulk” material specifically refers to. Such a distinct feature also opens several pathways to which the exfoliation solvent can permeate the crystal, which, while it enables a more efficient interfacing with the crystal, could introduce several exfoliation products if not optimized. To date, only a few select studies have investigated LPE of 1D or q-1D materials ,,, and, as such, the material basis of how the additional structural degree of freedom afforded by these materials may affect their behavior in top-down LPE has not yet been fully understood.

In this article, using the 1D ionic crystal (NbSe₄)₃I as a model phase, we explore how different forms of 1D “bulk” crystalline precursors of the same composition exfoliate in drastically different ways. We present results that illustrate how the LPE behavior of the model phase drastically varies depending on several factors such as the overall bulk crystal morphology (e.g., needles, flakes, or polycrystalline powders), exposed crystallographic facets, and the surface roughness. Beyond these key material insights, upon ideal and optimized LPE conditions, we demonstrate that (NbSe₄)₃I is exfoliable into solution-processable, uniform, and ultrathin nanoribbons that show pronounced nanoscale bending and twisting and preserve its native atomic structure and semiconducting character. These results demonstrate a novel tunable parameter in the form of bulk crystal morphology to achieve effective LPE that may be applicable to other 1D solids. In particular, these results fundamentally highlight the distinct exfoliation behavior of noncovalently bound 1D and quasi-1D chains whose top-down exfoliation behavior into ultrathin nanowires, nanoribbons, or even nanosheets is deeply tied to the degree and, sometimes, anisotropic strength of the axis-dependent interchain interactions in these phases. The approach that we present herein may be especially effective in closely related phases within the (MSe₄)_nI (M = Nb, Ta; n = 2, 3, 3.33) class of solids, as synthetically accessing the needle- or flake-like crystals from these phases is highly likely owing to their well-defined and axis-specific coupling across the cationic (MSe₄) _n units. We expect that this facet-specific exfoliation, which may ultimately alter the LPE dynamics in 1D and quasi-1D materials, can be extended to other noncovalently bound solids such as the quasi-1D Chevrel-type crystals and related phases, − given that the strength of the interchain interactions is precisely tuned to be weaker via specific compositional substitution.

Results and Discussion

Our desire to investigate the LPE behavior of (NbSe₄)₃I was motivated by its distinct crystal structure that is characterized by unique crystalline cavities that we speculated could allow for efficient solvent permeation. The crystal structure of (NbSe₄)₃I is comprised of ∼4.3 Å-wide NbSe₄ chains with cationic character, which are bound together by iodide anions (Figure A). The unique structure of this 1D ionic phase arises from the distinctive electronic nature of the Nb atoms in the lattice wherein 1/3 of the Nb atoms are postulated to bear a “Nb⁵⁺” oxidation state, while 2/3 of the atoms are expected to be in the “Nb⁴⁺” state, although the direct assignment of these oxidation states from prior reports is complicated by the expected degree of delocalization and the presumed insensitivity of the probed core electron state to changes in oxidation state. Regardless, this hypothesized mixed valency of Nb is consistently manifested in its crystal structure, wherein two different spacings corresponding to the “Nb⁴⁺Nb⁴⁺” spacing of 3.252 Å and an “Nb⁴⁺Nb⁵⁺” spacing of 3.061 Å are apparent and are stacked in an “Nb⁴⁺Nb⁴⁺Nb⁵⁺” repeating pattern. These “Nb⁵⁺” atoms have been demonstrated to host the structural antiferroelectric transition, as they were found to glide antiparallel below the antiferroelectric Néel transition temperature at around 0 °C (274 K). Based on charge balance, 2/3 of the Nb sites along the NbSe₄ chains can be thought of as being neutral and rationalizes the formation of site cavities along iodide channels as visible in the [100] projection in the structure (Figure B). We posited that the manifestation of these cavities, not often observed in 1D and quasi-1D ionic lattices, could be leveraged for efficient LPE.

1.

1D ionic structure of (NbSe₄)₃I and its bulk crystal growth behavior. (A) Perspective crystal structure model of (NbSe₄)₃I down the chain direction, depicting its 1D ionic nature. (B) Crystal structure of (NbSe₄)₃I viewed along the [100] (left) and [001] (right) directions. (C) Typical CVT reaction derived from elemental precursors and resulting in pure phase (NbSe₄)₃I along a temperature gradient. (D & E) The resulting temperature-regio-dependent needle- or flake-like morphology of (NbSe₄)₃I single crystals.

In our efforts to grow the bulk crystal precursors of (NbSe₄)₃I, we found that the resulting bulk single crystals grown via chemical vapor transport (CVT) intriguingly varied in morphology along the temperature gradient imposed along the reaction tube (Figure C). From the reaction tubes, we found that needle-like crystals resulting from primarily covalent growth along the chain direction were predominantly grown in the center of the tube between the hot (T _h = 730 °C) and cold (T _c = 650 °C) zone regions (Figure C,D) while flake-like or prismatic quasi-2D crystals with faceting distinct from the needles formed in the cold zone (Figure C,E). Between these distinct growth zones, both small needle-like crystals and other nonuniform small crystals form and are indicative of the prominent influence of growth temperatures in driving the morphologies of bulk (NbSe₄)₃I single crystals (Figure S3). Qualitatively, as expected from the strong covalency along the chain direction, we observe the dominance of the needle-like (both short and long) over the flake-like bulk crystals. Both crystal morphologies are composed of exclusively pure phase (NbSe₄)₃I: when combined and finely ground, powder X-ray diffraction (PXRD) patterns showed no competing impurity phases upon Le Bail fitting (Figure S1). Le Bail fitting was employed to mainly confirm the phase identity and purity of the ensemble, as any further refinement of the structure based on PXRD pattern fitting is complicated by the propensity of the anisotropic 1D crystals to preferentially align in the measurement configuration. Similarly, energy-dispersive X-ray spectroscopy (EDS) of single isolated crystals also corroborated the elemental ratio of 3:12:1 Nb/Se/I in both flakes and needles (Figure S4). This significant difference observed in the shape and morphology of bulk (NbSe₄)₃I crystals has prompted us to first investigate the origins and nature of the divergent crystal faceting before attempting the LPE of these crystals.

To understand the nature and faceting of the surfaces of the (NbSe₄)₃I needles and flakes in more detail, we performed scanning electron microscopy (SEM) and atomic force microscopy (AFM) on single crystals specific to each morphology (Figure ). The needles, while predominantly flat macroscopically, presented distinct and well-defined striations and channels that are indicative of long-range fractures between the covalent chains of the crystal. This observation gave us a strong indication that the covalent chain axis ran parallel to the long axis of the crystal, as expected from its 1D structural nature. On the other hand, the surface of the flakes did not show any striations or fracturing and was relatively smoother compared to the needles, albeit presenting many rounded bumps scattered throughout the surface (Figure B). This observed topographical nonuniformity and bump-like features of the flakes were surprising and prompted us to investigate these further by intentionally fracturing the crystals mechanically. From these qualitative fracturing experiments, we found that the needles fractured along the length of the crystal while the flakes, instead, cleaved more easily orthogonal to the surface (Figure C,D). To rationalize this morphology, we posit that the weak ionic interactions between chains may be overcome more easily than the covalent interaction along the chains, giving rise to the facile cleavage of (NbSe₄)₃I crystal between chains. This presented a strong indication that the covalent axis was normal to the surface of the flake, as if these were short chains that bundled side-by-side to form the quasi-2D flake surface. The realization of this orientation-driven morphology also explains the nonuniformity of the flake surface, as it was apparent from SEM imaging that it is comprised of a passivation layer necessary to terminate the exposed dangling bonds at the ends from which the covalent chains terminate.

2.

Surfaces and facets of (NbSe₄)₃I single crystals from two distinct morphologies. (A) Low-magnification SEM images of the pristine crystal surface of the needles (left) and flakes (right). (B) AFM surface topography of the needles (left) and flakes (right). Needles reveal 1D-striations along the surface but are otherwise flat, while flakes have a nonuniform surface with bump-like nanoscale features. (C) SEM of a mechanically fractured (NbSe₄)₃I needle, revealing the facile cleavage of fiber-like structures parallel to the length of the needle. (D) SEM of a mechanically fractured (NbSe₄)₃I flake with facile cleavage of significantly shorter fiber-like structures orthogonal to the surface of the flake. (E) Orientation-dependent Raman spectroscopy of (NbSe₄)₃I crystals. “Perpendicular” denotes that the polarization of the Raman laser is oriented perpendicular to the long axis of the needle or flake, while parallel indicates that the polarization of the Raman laser is oriented in line with the long axis needle or flake.

We also performed orientation-dependent Raman spectroscopy to further establish the orientation and structural anisotropy of the resulting (NbSe₄)₃I crystals, allowing us to directly correlate the symmetry in the crystals to their exposed surface facets. (Figures E and S5 and Table S1). Looking into the (NbSe₄)₃I crystals with needle morphology, we found that when the long axis of the needles was oriented either perpendicular or parallel to the polarization of the incident laser beam, the resulting Raman spectra revealed different intensity ratios of the apparent Raman-active modes. In the perpendicular orientation of the needles, characteristic peaks appeared at 141 (A_1g), 179 (A_1g), and 231 (B_1g) cm^–1. While these same modes are not as apparent in the parallel orientation, a weak but observable mode at 297 cm^–1 (E_g/B_2g) appeared instead. Both spectra showed the most intense major peak at 278 cm^–1 (A_1g). These results are consistent with previous polarization studies on needle-like (NbSe₄)₃I crystals and further validate that the long axis of the needle corresponds to the covalent [001] direction of the constituent chains. On the other hand, looking into the (NbSe₄)₃I crystals with flake morphology, we found that all five modes observed in the crystals with needle morphology (one major peak, one minor peak from parallel orientation, and three minor peaks from perpendicular orientation) appeared in the spectra. These peaks, regardless of the crystal orientation with respect to the incident laser, appeared at qualitatively equivalent relative intensities from both orientations and validated that each of these crystal orientations (and their corresponding crystallographic directions) are symmetrically equivalent. This observation is expected based on the tetragonal unit cell of (NbSe₄)₃I, in which the [100] and [010] directions are symmetrically identical. Beyond our electron microscopy imaging, these results gave further validation that the covalent [001] axis was indeed perpendicular to the flake surface.

While surprising, the origins of the differing anisotropic growth within the (NbSe₄)₃I crystals can be reasoned in terms of the two distinct growth regions: (1) as higher temperatures induce the more thermodynamically favored growth conditions, similar to other 1D and quasi-1D crystals, the growth of (NbSe₄)₃I crystals along the strongly bonded covalent chain direction is promoted and results in the formation of needle-like crystals; and (2) as slightly lower temperatures are more likely to promote kinetic-like conditions, the reaction proceeds to grow flake-like crystals from the lateral crystallization of the chains through the weaker noncovalent bonding along the basal plane direction. The ability to synthetically direct the formation of (NbSe₄)₃I crystals to form from either the covalent or noncovalent bonding direction is common at the nanoscale under highly tunable parameters and conditions like chemical vapor deposition (CVD), , but uncommon and less trivial in bulk-scale crystal growth as we observed here. Given the synthetic realization of highly distinct surfaces and crystallographic faceting of the flakes and needles of (NbSe₄)₃I, this created a unique opportunity to compare the LPE dynamics between the exposed surfaces and observe how solvent-surface interfaces may differ.

Our confirmation of the origin of the two (NbSe₄)₃I single-crystal morphologies presented us an opportunity to investigate how these bulk precursors that bear distinct surface characters and faceting would diverge in their behavior toward LPE (Figures and S6). Based on the ionic nature of these crystals, we expected that polar solvents, such as water (H₂O), would be the most efficacious for the LPE and unbundling of the crystals into thinner 1D structures. However, we found a range of exfoliation results that strongly depended on the crystalline morphology and exposed surface of the bulk crystals used as precursors instead (Figure A). Our LPE experiments involved a range of solvents with various values of surface tension, emphasizing more on the range of its polar component than the dispersive component, as follows (arranged in decreasing polar component): water (H₂O), acetone ((CH₃)₂CO; propan-2-one), tetrahydrofuran ((CH₂)₄O; oxolane; THF), isopropyl alcohol ((CH₃)₂CHOH; propan-2-ol), chloroform (CHCl₃; trichloromethane), diethyl ether ((CH₃CH₂)₂O; ethoxyethane), and hexane (C₆H₁₄). We highlight the specific detail of our exfoliation procedure wherein we employed one single crystal of a needle or a flake of (NbSe₄)₃I to avoid any unexpected intercrystal interference or nonuniformity during the exfoliation process. Thus, the resulting suspensions appear faint not because of the lack of material suspended but because it was intentionally made dilute for the specificity sought after in our study. We also note that these suspensions, at this point, were not centrifuged to preserve the nature and distribution of the exfoliated nanostructures.

3.

Liquid-phase exfoliation of (NbSe₄)₃I. (A) Surface-dependent LPE of (NbSe₄)₃I using various forms as bulk precursors: needles (top), flakes (middle), and powder (bottom). (B) Dark-field optical micrographs of the resulting (NbSe₄)₃I suspensions from LPE, dropcasted onto SiO₂/Si substrates. The solvent labels correspond, with their preferred IUPAC names, to water (H₂O), acetone ((CH₃)₂CO; propan-2-one), isopropanol ((CH₃)₂CHOH; propan-2-ol), chloroform (CHCl₃; trichloromethane), hexane (C₆H₁₄), THF ((CH₂)₄O; tetrahydrofuran; oxolane), and diethyl ether ((CH₃CH₂)₂O; ethoxyethane). Note that the suspensions are faint as these were intentionally loaded at low concentrations, with one single crystal for the case of the bulk needle and flake.

We found that for thin, needle-like crystals using tetrahydrofuran (THF) as a solvent resulted in the most effective LPE (Figure A, top). This combination resulted in a stable suspension that is visually more concentrated compared to the other solvents and yielded many thin, well-defined, and high aspect ratio wire-like structures that did not form aggregates upon dropcasting as observed from dark-field optical microscopy (Figure B). Upon exfoliation, we observed complete exfoliation of the precursors and found that no polycrystalline powder remained unsuspended. The (NbSe₄)₃I bulk needles, owing to the pronounced intrachain covalency, were fully exfoliated into thin nanoribbons. We note that the slight red-orange tint found in the suspension is due to the trace I₂ that is unavoidable even with careful synthesis due to its use as one of the precursors and its solubility in THF (Figure S7). We further elected not to rinse the trace I₂ off the crystals in any solvent such as acetone to avoid any confounding variables that could be introduced to the surface by a non-native washing solvent. Unexpectedly, using water as a solvent for the LPE had no effect on these bulk (NbSe₄)₃I needles. Keeping the conditions the same as with the LPE using other solvents, we found that crystals would float on the surface of water after sonication, indicating the surface tension of the solvent was too high to exfoliate the crystal into nanostructures effectively.

Using the same set of solvents, the exfoliation behavior of the (NbSe₄)₃I flake was markedly distinct from its needle-like counterparts (Figure A, middle). We found that using chloroform as a solvent for LPE was the most effective, yielding exfoliated wire-like structures that suspended well but aggregated together upon dropcasting (Figure B). In stark contrast to the needles, water appeared to induce some degree of exfoliation, while THF did not show any signs of exfoliation for these large flakes. We performed many trials using either flakes or needles of (NbSe₄)₃I that were sonicated in THF, which consistently resulted in the highly divergent exfoliation behavior. Finally, we performed the same LPE process using the same set of solvents on the powder formed by indiscriminately grinding flakes and needles together to further understand the impact of the crystal morphology and faceting on the LPE process as well as to disambiguate any morphology or crystallinity effects on the exfoliation of the crystals (Figure A, bottom; Figure S8). We found that powdered (NbSe₄)₃I crystals appeared to give the best result upon exfoliation in water, consistent with our initial hypothesis based on an exposed ionic surface that would favor the use of a more polar LPE solvent. Still, while significant amounts of exfoliated nanocrystals were suspended in water, dark-field optical microscopy imaging of drop-cast samples revealed little morphological control, forming a rough, aggregated film instead of pristine, isolated 1D nanostructures expected from the exfoliation of (NbSe₄)₃I (Figure B). Many other solvents, including acetone, isopropanol, and THF, exhibited some degree of exfoliation, albeit significantly less pronounced compared to water. These results are unusual, especially when considering the existing literature on top-down LPE, which typically do not differentiate between the form of the starting materials and simply refer to these as “bulk” or call for the process of grinding into powder prior to exfoliation. As a result, we sought to more comprehensively confirm that there was no structural difference between the flakes and needles other than the nature of the available exposed surface.

To evaluate the success of (NbSe₄)₃I LPE at the nanoscale and to simultaneously gather structural information on the resulting nanocrystals, we performed high-resolution electron microscopy on each of the three successful LPE conditions previously discussed (Figure ). We first performed aberration-corrected high-angle annular dark-field (HAADF) scanning transmission electron microscopy (STEM) imaging on the THF-exfoliated needlesthe conditions that gave the most uniform and well-defined samples, to obtain atomic resolution images and elemental mapping (Figures A and S9). Complementarily, we also performed selected-area electron diffraction (SAED) on these exfoliated structures as a baseline for comparison between samples (Figure B). From these experiments, both the atomically resolved HAADF-STEM images and collected SAED patterns were in close agreement with the bulk crystal structure of (NbSe₄)₃I and preserved the elemental composition post-exfoliation (Figure S9). We then performed transmission electron microscopy (TEM) on all three LPE samples (needle, flake, and powder precursors) exfoliated using the best solvents (THF, chloroform, and water, respectively) to investigate the distinct features of the exfoliated nanostructures resulting from these conditions. The results from these experiments were consistent with the observations from optical microscopy but yielded additional morphological, structural, and crystallographic insights. From TEM, we found that the bulk needle-like crystals exfoliated using THF produced long, thin nanoribbons with an anisotropic rectangular cross section, that did not appear to aggregate on the grid (Figure C). The bulk flake-like crystals exfoliated in CHCl₃ produced long nanoribbons but also produced shorter wires alongside and appeared to result in more significant aggregation (Figure F). The precursors in ground powder from exfoliation in water showed exfoliated structures that have some degree of 1D anisotropic character at the nanoscale, which could not be observed with optical microscopy, although it remained a poor method of accessing long, thin, and uniform nanowires or nanoribbons (Figure I).

4.

Nanoscale morphology and structure of exfoliated (NbSe₄)₃I nanocrystals. (A) Atomic resolution C_S-corrected HAADF-STEM image of (NbSe₄)₃I acquired from nanoribbons exfoliated from bulk needles in THF (left) and the corresponding simulated image (right). (B) SAED patterns of (NbSe₄)₃I acquired from nanoribbons exfoliated from bulk needles in THF along the [110] (left) and [100] (right) zone axes. (C) Wide-field TEM image of (NbSe₄)₃I nanoribbons from the exfoliation of bulk needles in THF. (D) HRTEM image of the resulting nanoribbons from the exfoliation of bulk needles in THF. (E) FFT of the micrograph in part D. (F) Wide-field TEM image of (NbSe₄)₃I nanoribbons from the exfoliation of bulk flakes in chloroform. (G) HRTEM image of the resulting nanoribbons from the exfoliation of bulk flakes in chloroform. (H) FFT of the micrograph in G. (I) Wide-field TEM image of nonuniform (NbSe₄)₃I nanocrystals from the exfoliation of powdered samples in water. (J) HRTEM image of the resulting nanocrystals from the exfoliation of powdered samples in water. (K) FFT of the micrograph in J.

In all three cases involving the needle, flake, and powder precursors, the nanoscale structure of the resulting exfoliated samples was found to be unperturbed from the native structure of (NbSe₄)₃I (Figure D,G,J) and the ensuing fast Fourier transformations (FFTs) of the images (Figure E,H,K). Each FFT is indexable to the native (NbSe₄)₃I crystal structure along the directions perpendicular to the long axis of the crystal either in the [100] or predominantly in the [110] zone axis. The covalent axis of the crystal, apparent as the [002] direction (corresponding d ₀₀₁ = 9.6 Å measured, 9.565 Å actual), ran parallel along the long axis of the nanoribbons in all cases. We note that the spots corresponding to the odd-numbered (00l) series (e.g., (001), (003), (005), etc.) were not present in any FFT or SAED pattern, as these are the expected systematic absences. Furthermore, the long, thin wires exfoliated from needles in THF maintain the native structure even after 6 months of exposure to ambient air conditions, as determined by high-resolution TEM (HRTEM) (Figure S10). Additionally, we complemented this evidence of impressive material stability with complementary exposure-dependent Raman spectroscopy studies (Figure S11). Based on these experiments, we found no noticeable change in the peak positions and full width at half-maximum (7.7 cm^–1 and 8.0 cm^–1 for the pristine and 2-month-old films, respectively) and no signs of Raman modes indicative of bulk oxide formation (e.g., NbO or Se–O) in the spectra even after 2 months of air exposure. These results confirmed that there was no atomic structural difference between the flake-like crystals, needle-like crystals, or ground powder that could have driven the observed differences in the LPE behavior of (NbSe₄)₃I crystals in various forms.

Given the identical crystal structure and chemical composition of the flakes, needles, and ground powder samples, we infer that the observed differences in their LPE dynamics are, primarily, a product of their bulk crystal attributes such as wettability and apparent surface tension. We note that the key difference between the bulk needles and flakes is suggestive of the facet- and surface-sensitive LPE behavior of the (NbSe₄)₃I crystals. In the case of the bulk needles, the crystal surfaces are predominantly defined by (hk0) facets (including the ⟨100⟩ and ⟨110⟩ families) that correspond to the noncovalent interchain direction in the crystal, suggesting that the vast portion of the exposed surface is expected to be ionic in nature. We believe that this distinct feature allows for efficient permeation of the solvent between the constituent chains during the LPE process. Interestingly, the structural cavities along the iodine columns in the crystal lattice are large enough to host a full THF molecule, which may have increased its likelihood of permeation between the chains and the efficiency of the subsequent delamination step. In the case of the bulk flakes, we found that the majority of the exposed surface is characterized by (00l) facets belonging to the ⟨001⟩ family. This intuitively means that the surface is passivated, as the covalent chains are predisposed to terminate in dangling bonds at the exposed ends of the crystal. We posit that this intrinsic and amorphous passivation layer presents a different interaction with the LPE solvents compared to the native ionic surface along the length of the needle-like crystals, which, as we have established herein, is only expressed as a minor facet component of the flakes. Finally, the observed differences in the LPE behavior between the bulk crystals (needles and flakes) and the ground powder emerge due to the roughness of the surface. Because the ground powder has a significantly rougher surface, its propensity for wettability is expected to be far greater than for the bulk crystals with relatively smoother surfaces. , Thus, the combination of the relatively higher apparent surface area and more exposed ionic facets of the powders, together with the high surface tension and polar character of water, allows for the more efficient but also more random, exfoliation of the powdered samples. Taken altogether, these results indicate that the differences in the observed LPE behavior across the three precursor forms arose from the difference in the morphology of the starting crystals (i.e., faceting, exposed surface area, and surface roughness) and how the structure chemically and physically interacts with solvent (i.e., surface tension and its polar component) at the bulk crystal length scalea feature distinct to exfoliable 1D and quasi-1D crystals that have large, exposed surfaces or edges and multiple crystallographic axes available for cleavage or exfoliation.

While investigating the morphologies of the exfoliated nanostructures under TEM, we also observed that both the (NbSe₄)₃I needles (exfoliated in THF) and flakes (exfoliated in chloroform) were presented as ultrathin nanoribbons with sub-100 nm widths that showed pronounced bending and twisting (Figures and S12). The observed twisting and bending were more apparent in the THF-exfoliated needles, as these resulted in a more consistent and uniform exfoliation into nanoribbons. These nanoribbons were found to adhere to the edges of the carbon support, bending and twisting to follow the path of the lacey carbon surface (Figure A). This behavior indicates that the flexion observed is likely a result of the interfacial interaction rather than internal strain within the nanoribbons. These twisted nanoribbons typically lie flat along or close to the [110] zone axis (Figure B,C). Given the dominance of the highly ionic [110] facet in the nanoribbons, we suspect that the twisting is induced by the interfacial strain that arises when the highly charged surfaces of the nanoribbons that are stabilized in the suspension are liberated from the solvent as these are cast and dried onto the lacey carbon grid.

5.

Pronounced twisting and bending of long, ultrathin (NbSe₄)₃I nanoribbons. (A) Low-magnification TEM images of ultrathin (NbSe₄)₃I nanoribbons that show significant degrees of bending and nanoscale twisting across the lacey carbon support. (B) HRTEM image depicting a well-resolved full 180° twist in an (NbSe₄)₃I nanoribbon. (C) FFTs of the specific regions (indicated as I, II, III in the micrograph) of the nanoribbon shown in B.

Often, thin (NbSe₄)₃I nanoribbons are observed to twist to a full 180° rotation (Figure B,C). Presented with the high-resolution micrographs of the twisted samples, we looked into the apparent long-range structure in more detail, starting from the indexable [110] zone axis on the left side of the nanoribbon. The representative nanoribbon can be observed twisting inward to a region of higher electron density, indicating that the nanoribbon is twisting and is not simply thinning. Furthermore, the corresponding FFT shows significant rotational disorder at the twisted region, with spots smearing into rings. Following the twist, the nanoribbon turns to the opposite [−1–10] zone axis after a full 180° twist. As a result of the twisting strain, (001) and (003) spots are visible in the FFT pattern, indicating a symmetry reduction due to the twisting. This is distinct from the ferroelectric transition’s symmetry reduction, which preserves the (001) and (003) systematic absences. While stochastic, the twisting of the exfoliated (NbSe₄)₃I nanoribbons is a highly reproducible and very common result of our LPE process and serendipitously presents a unique method to access surface-driven chiral nanoribbons based on an achiral parent phase.

After observing the distinct bending and twisting behavior from the long and ultrathin (NbSe₄)₃I nanoribbons, we next sought to sort and isolate these high aspect ratio nanoribbons from the larger, more rigid nanoribbons via centrifugation (Figure ). Note that the presented structures and results prior to this section represent the structure and distribution of the resulting nanocrystals postexfoliation, without any centrifugation step. Given that the combination of the bulk needles of (NbSe₄)₃I sonicated in THF resulted in the highest concentration of exfoliated material, high apparent aspect ratios, most pristine wires, and minimal aggregation, we employed these specific conditions to optimize the sorting of the exfoliated nanoribbons via high-speed centrifugation. Based on our results, we found that centrifuging the resulting suspension from exfoliation at 7500 rpm enables access to long, ultrathin nanoribbons with high aspect ratios, lengths reaching the micrometer scale, and widths that are well under 50 nm. To more accurately describe the morphology of the exfoliated nanoribbons postcentrifugation, we employed a combination of atomic force microscopy (AFM) and scanning electron microscopy (SEM) (Figures B,C and S13) techniques, which allowed for a statistical distribution of nanoribbon heights and widths (Figure D,E). The nanoribbon heights were found to have an average size of 7.6 ± 2.6 nm (n = 125) and displayed a Gaussian distribution, which is usually atypical for a nanomaterial ensemble resulting from top-down techniques such as LPE. We found that the widths of the nanoribbons presented more variance in their distribution but yielded values that remained under ∼50 nm, resulting in an average value of 26.4 ± 10.9 nm (n = 112). Moreover, these nanoribbons were surprisingly uniform, generally presenting with unusually high aspect ratios on the order of 100 (dimensionless length/width). The highest aspect ratio nanoribbon from our samples was found to be 337, which is based on the 6.75 μm-long and 20 nm-wide nanoribbon shown in Figure .

6.

Efforts toward achieving high aspect ratio nanoribbons from exfoliated (NbSe₄)₃I via centrifugation. (A) Low-magnification TEM images of high aspect ratio nanoribbons of (NbSe₄)₃I. (B) Representative AFM micrograph of a thin film comprised of exfoliated (NbSe₄)₃I nanoribbons dropcasted onto a substrate. (C) Representative SEM micrograph of a thin film comprised of exfoliated (NbSe₄)₃I nanoribbons dropcasted onto a substrate. (D) Height frequency distribution of exfoliated (NbSe₄)₃I nanoribbons obtained from AFM, n = 125. (E) Width frequency distribution of exfoliated (NbSe₄)₃I nanoribbons obtained from SEM, n = 112.

Lastly, our demonstration of the exfoliation of (NbSe₄)₃I into ultrathin nanoribbons enables us to develop a means not only to probe the persistence of its semiconducting character but also to show its potential toward solution-processability in thin film devices. While the electronic structure of (NbSe₄)₃I has been established to be semiconducting in nature, as has been demonstrated through optical spectroscopy, angle-resolved photoemission spectroscopy, and electrical transport measurements, the 1D, antiferroelectric, and, consequently, highly stimulus-sensitive nature of the material has complicated the precise measurement of its band gap which has been reported in the broad range of 0.2 to 0.75 eV. Still, leveraging the intrinsic conductivity arising from the low band gap semiconducting character of the bulk (NbSe₄)₃I crystals, coupled with the ultrathin nature of the nanoribbons derived from LPE, we fabricated a proof-of-concept two-probe thin film device using exfoliated (NbSe₄)₃I nanoribbons dropcasted from a THF suspension (Figure ). Once again sonicating needles of (NbSe₄)₃I in THF and centrifuging large wires out, we drop cast a film onto a transparent quartz substrate. We found that the thin film, deposited onto a transparent quartz substrate with prefabricated Au contact pads, shows the significant transmission of ∼40–80% in the visible range (Figures A and S14). Measured between the Au contact pads (Figure B), we found that the average film thickness was approximately 173 nm based on AFM with an RMS roughness value of 73 nm (Figure C). We measured the current between the Au contact pads in the pA range, which, given the average thickness of the film and distance between the contacts, allowed us to estimate a reasonable conductivity value of 0.34 mS/cm. Comparing this to the two-probe device based on a bulk crystal device that we fabricated (Figures S15 and 16), we found that this thin film conductivity value is 5 orders of magnitude lower than the bulk but is not surprising due to its aggregated thin film form factor and nanoribbon morphology where the narrow 1D nanoribbon channels and the likelihood of internanoribbon hopping are expected to lower the apparent measured conductivity. , Nonetheless, the measured conductivity value that we were able to measure based on this thin film is on par with other exfoliated low-dimensional material thin film devices that are nonextrinsically doped semiconductors. Similar to electrical devices based on thin and long nanowires, we expect that the device performance may also be improved by unidirectionally aligning the nanoribbons and improving the inter-ribbon contact, potentially by techniques that are applicable to long nanostructures such as stretching or shearing the ensemble using an elastomer such as poly­(dimethylsiloxane), , sintering the film, and directional drying. Owing to the semiconducting nature of (NbSe₄)₃I, we also anticipate that aliovalent doping, especially at the transition metal site, could further improve the electrical conductivity of the phase, but such studies warrant careful consideration due to the narrow band gap of the phase. Moreover, given the well-documented antiferroelectric transition in this material, the LPE process and the proof-of-concept electrical transport results that we presented herein could be further leveraged for the scalable fabrication of thin film device architectures geared toward high-energy storage capacitors, actuators, microwave dielectrics, and dynamic random-access memory.

7.

Transparent conductive film fabricated from (NbSe₄)₃I. (A) Photograph of (NbSe₄)₃I thin films fabricated by dropcasting the THF-suspended nanoribbons onto a quartz substrate with prefabricated Au contact pads. (B) Optical micrograph of the Au contact pads in the thin film device used to measure the electrical conductivity of (NbSe₄)₃I nanoribbons. (C) AFM of the (NbSe₄)₃I thin film between two Au contact pads, measured with an average thickness of 173 nm and an RMS of 73 nm. (D) Two-probe I–V curve taken between the Au contact pads with a measured conductivity of 0.34 mS/cm based on the height determined by AFM.

Conclusions

In this work, we demonstrate the pronounced impact that the bulk morphology, surface character, and faceting of a 1D ionic crystal, (NbSe₄)₃I, have on its behavior toward liquid-phase exfoliation into ultrathin nanostructures. We established this by carefully showing that the crystals that display needle-like morphologies with (hk0)-faceted surfaces and flake-like morphologies with (00l)-faceted surfaces distinctly display divergent behaviorboth in terms of the solvent preference and the nature (morphology, uniformity, thickness, size distribution, etc.) of the resulting exfoliated materials. Under optimized conditions, exfoliation of needle-like bulk crystals in THF resulted in well-defined and ultrathin nanoribbons that feature heights down to 7.2 ± 2.6 nm and widths as thin as 26.4 ± 10.9 nm. Owing to their ultrathin form factor and high aspect ratios, these bendable nanoribbons uniquely show interface-induced twisting and are easily processable into conducting thin films. Our results experimentally demonstrate a key materials parameter for the top-down LPE of 1D and quasi-1D crystals into uniform and well-defined nanoribbons, emphasizing the crystals' distinct feature of weak and noncovalent interchain interactions that allow for exfoliation across various crystallographic axes. Altogether, our study not only underscores a means to understand the fundamental chemical and physical behavior of 1D solids through the realization of ultrathin 1D nanowires and nanoribbons but also presents a viable pathway toward solution-processable functional devices comprised of building blocks in the sub-10 nm regime.

Methods

General Methods

Structures were generated in SingleCrystal 4 software (CrystalMaker Software Suite). Data were plotted using MagicPlot. Figures were put together and are presented in Lunacy.

Synthesis of Bulk (NbSe₄)₃I

Stoichiometric amounts of elemental precursors of Nb, Se, and I₂ were loaded into a quartz tube (10 mm I.D.; 12 mm O.D.) and evacuated to <20 mTorr. The tube was flame-sealed into an ampule approximately 10 cm in length, which spanned the length of a two-zone furnace. This ampule was then loaded into a two-zone furnace (MTI OTF-1200X). Both sides were heated evenly to 650 °C at 5 °C/min, then the side with elemental reagents in the “hot” zone continued to 730 °C while the “cold” zone remained at 650 °C. This temperature gradient was maintained for 3 weeks before the furnace was turned off and allowed to cool naturally. This resulted in a gradient of needles- and flake-like crystals depicted in Figure B of the main text. The resulting crystals were found to be pure phase (NbSe₄)₃I.

Powder X-ray Diffraction (PXRD)

(NbSe₄)₃I crystals (flakes and needles) were combined and finely ground using a mortar and pestle. PXRD was performed using a Rigaku Miniflex XRD. A Le Bail fit was performed on the diffraction pattern to assess its purity and to calculate its crystallographic parameters using Profex with the literature crystal structure of (NbSe₄)₃I. This diffractogram was converted to the subsequent pair distribution function using xPDFsuite.

Scanning Electron Microscopy (SEM) and Quantitative Energy-Dispersive X-ray Spectroscopy (EDS)

SEM was performed on either a FEI Quanta 3D FEG Dual Beam or FEI Magellan 400 XHR, each operated at 10–15 kV. For cleaved images, needle or flake crystals of (NbSe₄)₃I were adhered to carbon tape; then, a spatula was firmly pressed onto the crystal until fracture occurred. For quantitative EDS, an FEI Quanta 3D FEG Dual Beam operated at 20 kV with a 50 mm² silicon drift detector was used.

Atomic Force Microscopy (AFM)

AFM was performed using an Anton Paar Tosca 400 AFM instrument operated in tapping mode. Analysis of the micrographs was performed using Gwyddion.

Raman Spectroscopy

Raman spectra were collected on a micro-Raman system based on a Renishaw Inc. inVia microscope was equipped with a 532 nm laser. Crystals were oriented with the longest axis either parallel or perpendicular to the laser polarization. Raman spectra were fit using MagicPlot Pro with a spline background and asymmetric Lorentzian peaks.

Liquid-Phase Exfoliation of (NbSe₄)₃I

For single-crystal LPE, one single crystal, either a needle or a flake, was placed into 2 mL of solvent. The crystals imaged in Figure were directly used for exfoliation depicted in Figure . For the powder samples, the bulk (NbSe₄)₃I needles and flakes were combined and finely ground together. Afterward, the powder was placed in 2 mL of solvent at a concentration of 3 mg/mL.

Sorting of High Aspect Ratio Nanoribbons via Centrifugation of LPE Suspension

After sonicating (NbSe₄)₃I needles in THF for 10 min, the suspension was transferred into Eppendorf tubes and placed in a VWR Mini Centrifuge to sort the sizes of the exfoliated (NbSe₄)₃I nanostructures. The tubes were centrifuged at 7500 rpm. for 1 min, which removed large bulk particles and nonexfoliated crystals from the resulting suspension. The supernatant liquid was aliquoted out and drop-cast for further characterization.

Optical Microscopy of Exfoliated (NbSe₄)₃I

The suspended (NbSe₄)₃I nanoribbons from LPE were dropcasted onto SiO₂/Si substrates. Dark-field optical images were obtained using a CytoViva hyperspectral system that is based on an Olympus BX-43F microscope.

Scanning Transmission Electron Microscopy (STEM), High-Resolution Transmission Electron Microscopy (HRTEM), Selected-Area Electron Diffraction (SAED), and Elemental Mapping by EDS

Aberration-corrected HAADF-STEM was performed on a JEOL JEM-ARM300F instrument operated at 300 kV. HRTEM, SAED, and STEM-EDS mapping were performed on a JEM-2800 operated at 200 kV. Raw images were processed using a Digital Micrograph (Gatan), while distance measurements and FFTs were performed in ImageJ.

Electrical Conductivity of (NbSe₄)₃I

(NbSe₄)₃I needles were sonicated in THF for 10 min and were centrifuged at 7500 rpm. for 1 min to separate the nonexfoliated bulk crystals. The supernatant liquid was then dropcasted onto quartz substrates with prefabricated Au contact pads. The substrates were then placed on a hot plate heated at 100 °C to aid in the solvent drying process. To ensure even thin film coverage, multiple coatings of the (NbSe₄)₃I supernatant solution were dropcasted onto the substrate until it turned a faintly gray hue, at which point AFM measurements were performed to determine the resulting thin film thickness. Current–voltage (I–V) curves were obtained using a probe station equipped with a Keithley 4200 SCS semiconductor parameter analyzer. Conductivity was calculated using resistance measured from I–V, the estimated thickness from AFM measurements, and the width and height obtained from optical imaging.

Ultraviolet–Visible (UV–vis) and Visible Spectroscopy

UV–vis spectroscopy was performed in the solution phase using an Agilent Cary 100 UV–vis spectrophotometer. A single crystalline needle of (NbSe₄)₃I was sonicated in THF and compared to the spectrum of pure I₂ dissolved in THF. Solid-state UV–vis was also performed on a transparent film of (NbSe₄)₃I using a CytoViva hyperspectral microscope equipped with an Xe halogen light source and a CCD detector. The transmission spectrum was converted to an absorbance spectrum by using ENVI software by accounting for the background spectrum. Transmittance and absorbance spectra of the thin films were also taken in the visible range using the CytoViva hyperspectral system based on an Olympus BX-43F microscope.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsmaterialsau.5c00143.

• PXRD of (NbSe₄)₃I, EDS of bulk (NbSe₄)₃I, additional trials of needle and flake exfoliation in THF, UV–vis spectroscopy of exfoliated (NbSe₄)₃I and pure I₂, Raman spectra peak fitting, STEM-EDS maps of (NbSe₄)₃I nanoribbons, additional TEM images of (NbSe₄)₃I showing twisting and bending nanoribbons, additional TEM images of (NbSe₄)₃I nanoribbons taken after 6 months of air exposure, additional AFM and SEM micrographs following centrifugation, transmittance and absorbance spectroscopy of the transparent (NbSe₄)₃I film, and conductivity of a bulk (NbSe₄)₃I crystal (PDF)

G.M.M.: Conceptualization, methodology, investigation, visualization, and writingoriginal draft preparation. C.J.C.: Investigation. K.G.D.: Investigation. D.L.: Investigation. M.Q.A.: Conceptualization, supervision, and writingreviewing and editing. CRediT: Griffin M. Milligan conceptualization, investigation, methodology, visualization, writing - original draft; Cameron Collins investigation; Kaitlyn G. Dold investigation; Diana Lopez investigation; Maxx Q. Arguilla conceptualization, funding acquisition, supervision, writing - review & editing.

The authors declare no competing financial interest.

Published as part of ACS Materials Au special issue “2025 Rising Stars in Materials Science”.