Charge density wave (CDW) materials host correlated electronic states typified by periodic lattice distortions and static modulations of conduction electrons1. Non-volatile memory and computing devices based on the principle of phase change memory (PCM)2,3 may leverage the intrinsic resistivity changes associated with CDW phase transitions4,5,6,7,8,9. 1T-TaS2, a van der Waals (vdW) layered solid, is a prototypical CDW material in which the atomic lattice distorts in-plane to form 13-atom star-shaped clusters10,11. The tiling of these clusters and the extent of commensuration with the underlying atomic lattice define the CDW phases and the electronic properties of 1T-TaS210,11,12,13,14,15. Notwithstanding the in-plane nature of this CDW lattice distortion, interlayer coupling plays a key role in stabilizing intralayer charge order in 1T-TaS2. Accordingly, together with flake thickness16,17,18 and doping levels16,19, vertical heterostructuring is a powerful route for engineering CDW transitions20,21,22,23,24,25. For example, whereas in pristine, bulk 1T-TaS2 the commensurate (C) CDW phase only forms below about 180 K11,12 (and is only observed at much lower temperatures in exfoliated thin flakes15), electronically isolating monolayer 1T-TaS2 with thicker metallic slabs of H-TaS2 has been shown to stabilize the C-CDW state in monolayer 1T-TaS2 at room temperature23,24. To this end, the recent synthesis of endotaxial TaS2 offers new mechanisms for accessing modular CDW systems23.

In this work, we demonstrate an approach converse to preceding literature—employing moderate thermal annealing to interdisperse monolayer H-TaS2 between few-layer 1T-TaS2 lamellae. In the resulting verti-lateral 1T-TaS2/H-TaS2 heterostructures, decoupled 1T-TaS2 fragments undergo independent transitions from the disordered incommensurate (IC) to ordered commensurate CDW phase above room temperature. These transitions are hallmarked by synchronous stepwise switching of chirality and resistance with high predictability; the number of steps is encoded by the quantity and arrangement of H-TaS2 layers. Thus, the developed materials represent a distinctive framework for deterministic engineering of multistate resistance and chirality changes in 1T-TaS2. Additionally, we find that the nucleation of H-TaS2 polytype initiates at macrosocopic flake defects, a mechanistic insight we harness to showcase the potential of strain engineering for the rational design of verti-lateral TaS2 heterostructures. Moreover, modulating the H-TaS2 content tunes the proportion of heterochiral CDW superlattices, resulting in a range of optically detectable net chiralities. Therefore, our work provides an adaptable roadmap for design of versatile optoelectronic phase change materials with reliable, multilevel and multifunctional switching.


Commensuration in verti-lateral TaS2 heterostructures

Exfoliated 1T-TaS2 crystals were annealed in high vacuum at 350 °C for 30 mins and then rapidly cooled to room temperature (see “Methods" section), engendering changes in optical contrast, indicative of structural transformations (Fig. 1a). This thermally induced phase evolution was characterized by selected area electron diffraction (SAED) of TaS2 flakes before and after annealing. Before annealing, SAED patterns of 1T-TaS2 flakes are consistent with the expected nearly commensurate (NC)-CDW structure (Fig. 1b)11. In contrast, SAED data from annealed samples (Fig. 1c) reveal the presence of two \(\sqrt{13}\times \sqrt{13}\) heterochiral superlattices. The CDW enantiomorphs, α (L) and β (R), possess supercells that are rotated ± 13.9 degrees relative to the unit cell of 1T-TaS2 (Fig. 1d)11,26,27. Additionally, the SAED data of annealed 1T-TaS2 flakes are consistent with a C-CDW phase at room temperature23,24. These ensemble changes in extent of CDW commensuration in heat-treated flakes were also evinced by linearly polarized Raman spectroscopy (Fig. 1e and Supplementary Figs. 15 and 16). Raman spectra obtained in visually distinct regions of annealed flake S1 (Fig. 1e) show a strong correlation between the red-channel optical contrast change upon annealing (ΔOCR) and the sharpness of low-frequency (40–100 cm−1) Raman modes related to CDWs28,29,30. The CDW spectral peaks become increasingly well-defined from regions 3 to 1 (R3–R1), a behavior that is tightly correlated with progressive ΔOCR. Furthermore, as ΔOCR becomes more negative, new Raman spectral features emerge at 64 cm−1, 89 cm−1, 123 cm−1, and 230 cm−1. The manifestation of new peaks and general sharpening of Raman features is consistent with Brillouin zone folding of 1T-TaS2 as it undergoes the transition from the nearly commensurate CDW (NC-CDW) to C-CDW28,29,30.

Fig. 1: Polytype and charge density wave (CDW) transformations in TaS2 flakes.
figure 1

a Optical micrographs of a 1T-TaS2 flake, labeled S1, before and after thermal annealing on 90 nm SiO2/Si substrate. Regions of interest are marked. Scale bars: 15 μm. b, c Room temperature selected area electron diffraction (SAED) patterns of a representative 1T-TaS2 flake before (b) and after (c) thermal annealing. A representative set of first-order superlattice peaks is marked for each sample. Scale bars: 2 nm−1. d Schematic of real space α and β CDW domains. Unit cells of α, β, and 1T-TaS2 are shaded in violet, pink and blue, respectively. Angle ϕ represents the rotational misalignment between the unit cell of 1T-TaS2 and the unit cells of the CDW superlattices. Schematic diffraction patterns for α, β, and α + β patterns are shown at the bottom. e Linearly polarized Raman spectra in selected regions of S1 after thermal annealing. Vertical dashed line represents the energy cutoff after which the Raman intensity was scaled by a factor of 2. f Relationship between the red-channel optical contrast change upon annealing (ΔOCR) and the number of H-TaS2 layers in distinct regions of S1 and S2 (a 1T-TaS2 sample) after annealing. This data is overlaid with the thickness dependence of OCR for 2H-TaS2 on 90 nm SiO2/Si substrate (ref. 31). gj Atomic-resolution differential-phase-contrast scanning transmission electron microscopy (DPC-STEM) images along the \([10\overline{1}0]\) zone axis of: region R1 of S1 (g), area R2 of S1 (h), regions R3--R4 of S2 (i), and region R3 of S1 (j) overlaid with structures of 1T-TaS2 (ref. 63) and H-TaS2 (ref. 64). H-TaS2 layers are highlighted and Ta ions aligned along the c-axis are connected with pink lines. Scale bars in gj are 1 nm.

To unveil the relationship between ΔOCR, CDW commensuration, and atomic lattice structure of these annealed 1T-TaS2 crystals, atomic resolution differential-phase-contrast scanning transmission electron microscopy (DPC-STEM) imaging and analysis (Fig. 1f–j) was performed on cross-sectional samples made from optically distinct regions of flakes S1 (Fig. 1a) and S2 (Supplementary Fig. 15b). Representative DPC-STEM micrographs, depicted in Fig. 1g–j, reveal that thermal annealing induces a partial transition from the 1T (octahedrally coordinated Ta) to the H (trigonal prismatic Ta) structure. The H polytype forms within the 1T-TaS2 matrix (i.e., endotaxially), overwhelmingly as monolayers, separating generally thicker fragments of 1T-TaS2. At boundaries between regions with dissimilar ΔOCR, the number of H-TaS2 layers varies across a single 1T-TaS2 flake, constructing lateral heterostructures with atomically sharp interfaces (Fig. 1i). Furthermore, these DPC-STEM data show how 1T-TaS2 slabs separated by a H-TaS2 layer slip relative to each other, resulting in the misalignment of the Ta centers (vertical lines in Fig. 1g–j) in mixed polymorph heterostructures.

We find that optical contrast measurements are a powerful and convenient tool for identifying polymorph composition, owing to the linear relationship between ΔOCR upon annealing and the number of H-TaS2 layers, determined from DPC-STEM (Fig. 1f). This relationship mirrors, and stems from, the linear scaling between layer count and OCR for freestanding few-layer H-TaS231 (see Supplementary Note 2 for details). Therefore, crystal sections with a more negative ΔOCR (i.e., appearing increasingly blue) contain a greater number of layers of the H polymorph. Taken together with the increased sharpness and number of CDW Raman modes (Fig. 1e), these data establish that the formation of monolayer H structures leads to increased order of 1T-TaS2 C-CDW domains, consistent with prior work with thicker H-TaS2 slabs23,24 and bulk mixed polytype TaS2 phases12,26,32,33.

The ensemble room-temperature ordering of CDW domains in TaS2 heterostructures with different polytype compositions was further probed using SAED. Our findings support that samples comprising ≥ 20% of H-TaS2 manifest an ordered C-CDW phase, characterized by sharp CDW reflections (Supplementary Fig. 4b–d). In contrast, the CDW reflections observed in samples with < 20% H-TaS2 appear less well-defined and exhibit peak splitting and angular blurring (Supplementary Fig. 4a). Accordingly, we infer that crystals containing less than 20% H-TaS2 content host a disordered C-CDW phase with likely coexistence of some NC-CDW domains. Notably, the chirality of CDW domains (Fig. 1d) could be exploited for next-generation switching devices, making the nano-scale structural understanding of C-CDWs in TaS2 heterostructures integral for their potential applications.

Diverse chirality in polytype heterostructures

The heterochirality of CDW superlattices in H-TaS2/1T-TaS2 at room temperature was mapped with nano-scale resolution using four-dimensional scanning transmission electron microscopy (4D-STEM), establishing that the α and β enantiomorphs are stacked along the c-axis to form vertical CDW superstructures. In 4D-STEM, a converged electron probe is scanned across a sample in a 2D array, while recording 2D diffraction data at each probe position (Fig. 2a)34. We obtained 4D-STEM datasets of a 20-layer flake, S3 (Fig. 2b), parallel to the c-axis with ~ 5.5 nm spatial resolution in regions R2–R5. These regions exhibit progressively sharper Raman spectral features (Supplementary Figure 16) and larger H/T ratios from measurements of ΔOCR. For R2–R5, Bragg reflections associated with both commensurate α and β enantiomorphs are present in nanodiffraction patterns (Fig. 2c, d and Supplementary Figure 7). Thus, the CDW enantiomorphs are coexistent in a ~ 5.5 nm area, revealing their formation in the out-of-plane (c-axis) direction. We note that CDW superstructures can be vertically stacked in two configurations across the H-TaS2 interface: heterochiral (αβ) (Fig. 2e) and homochiral (αα) (Fig. 2f), each hosting inequivalent CDW cluster interlayer stackings and charge distributions (Supplementary Figure 19)27. The interlayer arrangement of CDW clusters in polytype heterostructures exhibits notable distinctions compared to both 1T-TaS2 flakes and homointerfaces. In pristine 1T-TaS2, CDW clusters can be perfectly eclipsed because their building blocks—Ta ions—are directly aligned in the out-of-plane direction35,36. In contrast, in polytype heterostructures, Ta ions in 1T-TaS2 slabs across H-TaS2 interfaces must be laterally offset (Fig. 1g–j). Thus, CDW clusters in neighboring 1T-TaS2 slabs must assume a staggered arrangement, resulting in distinct CDW superlattice patterns (2e–f, Supplementary Figure 19).

Fig. 2: Mapping heterochiral CDW domains with four-dimensional scanning transmission electron microscopy (4D-STEM).
figure 2

a Schematic illustrating 4D-STEM of annealed 1T-TaS2 samples. Three compositionally distinct flake regions are labeled as r1, r2 and r3. The two heterochiral superlattices are marked as α and β. b Optical micrograph of a 20-layer annealed 1T-TaS2 sample S3. Regions of interest and their H-TaS2 proportion (x/n, where x = number of H-TaS2 layers and n = total number of layers), calculated from optical contrast measurements, are labeled. Scale bar: 10 μm. c, d The maximal diffraction patterns, displayed on a logarithmic scale, for S3--R3 (c) and S3--R5 (d). Scale bars: 5 nm−1. e, f Illustration of CDW superstructures for α,β (e) and α,α (f). Unit cells of the CDW superstructures are outlined in black. gj (β − α) / (β + α) virtual dark-field images and their histograms for R2--R5 regions of S3. Insets show the proposed sample composition and plausible chirality stacking based on the H-TaS2 content and 4D-STEM analysis of each region. We note that only the net chirality can be determined, and the exact stacking sequence of the chirality shown here is one of several possibilities as the precise sequence cannot be obtained from plan-view 4D-STEM. All 4D-STEM data was acquired at room temperature. Scale bars in gj: 50 nm.

Next, integrated intensities of α and β diffraction spots were evaluated at each probe position (see “Methods" section for details)37 to reconstruct dark-field images associated with the difference in superlattice ratios, defined as: (β − α)/(β + α) (Fig. 2g–j). The resultant enantiomorphic ratio maps reveal a minor extent of in-plane variation within each heterostructure (see histograms in Fig. 2 g–j), potentially arising from domain pinning defects. Nevertheless, 4D-STEM data and high-resolution TEM analysis (Supplementary Figure 5) consistently point to coexisting, out-of-plane α and β superstructures regardless of the H/T composition. However, the polymorph composition appears to influence the α/β proportion. For example, a less than 45% mean difference in ratio of diffraction pattern intensity from enantiomorphic phases was measured for R4 and R5 at room temperature (Fig. 2i, j). Conversely, a larger mean enantiomorphic disproportion exceeding 65% was obtained for R2 and R3 (Fig. 2g, h). This can be understood by considering that changing the polymorph composition alters the size and number of 1T-TaS2 fragments hosting the two heterochiral CDWs, thereby engendering changes in the overall chirality. A high enantiomorphic disproportion is therefore the most likely for less transformed samples with larger 1T-TaS2 fragments, which would then dominate the overall chirality.

The chirality of TaS2 heterostructures can also be assessed optically, as enantiomorphic disproportion engenders Raman optical activity (ROA): a distinct Raman response to right- and left-circularly polarized light (see Supplementary Note 9 for details)36,38. For 1T-TaS2 displaying ROA, the integrated area ratio of Eg(I) and Eg(II) modes differs in the σ+σ and σσ+ Raman polarization configurations (Fig. 3a), where σiσs (i, s = ±) are phonon helicities of the incident and scattered light38. Note, a stronger ROA signals a higher enantiomorphic disproportion and a larger overall chirality. Representative chirality-dependent optical measurements of sample S4 (Supplementary Figure 18c) are shown in Fig. 3. We find that ROA decreases from region R1 (x/n =3/20) to R3 (x/n =5/20), signaling decreasing overall chirality for an increasing H-TaS2 layer count (Fig. 3b–d and Supplementary Fig. 14). These results are consistent with our 4D-STEM findings; highly transformed regions, on average, exhibit a weaker overall chirality compared to medium transformed regions (Fig. 2g–j). Notably, our verti-lateral heterostructures exhibit a broad spectrum of possible overall chiralities, distinguishing them from 1T-TaS2 flakes/homointerfaces and heavily transformed H-TaS2/1T-TaS2 heterostructures, which can only manifest a single enantiomorphic state23,36. Specifically, the former are homochiral, comprising fully of α or β36, while the latter have been shown to be achiral, hosting an equal proportion of α and β23. Accordingly, our verti-lateral heterostructures may enable chiral opto-electronic memory schemes through their wide array of chiral states that can generate strong optical responses at room temperature.

Fig. 3: Optical detection of chirality in TaS2 heterostructures.
figure 3

a Schematic of a polarization-dependent Raman measurement. b, c Raman spectra in the circular contrarotating polarization configurations (σ+σ and σσ+) obtained for region R1 (b) and region R3 (c) of a 20-layer sample S4. Lorentzian peak fits and the cumulative fits are displayed. d Stacked bar charts displaying the normalized area percentage of Eg(I) and Eg(II) modes measured in the two contrarotating Raman polarization configurations (top: σ+σ, bottom: σσ+) for R1–R3 of S4. For bd, x/n denotes the H-TaS2 proportion. Data was obtained at room temperature.

Multistate resistance and chirality switching

Electronic properties of these heterochiral endotaxial polytype heterostructures were probed using variable-temperature transport measurements. In these studies, we monitored the temperature-dependent longitudinal resistance (Rxx) of mesoscopic devices fabricated from S1 (Fig. 4a) and S4 (Supplementary Fig. 18c). Measurements from compositionally distinct regions (Fig. 4b) were used to establish the relationship between polytype composition and the IC-to-C CDW phase transition behavior. For all measured regions, transport below 300 K is dominated by the metallic H-TaS2 layers (decreasing resistance with decreasing temperature), while transport above room temperature traces CDW transitions of 1T-TaS2 lamellae (Fig. 4c,d). These transport data were complemented with temperature-dependent 4D-STEM of representative heterostructures to provide structural insight (Fig. 4e–g). Above room temperature, 1T-TaS2 transforms from the IC (more conductive) to the C (more insulating) CDW phase with a hysteresis between cooling and warming profiles (Fig. 4c–e)11. As a general observation, increasing H-TaS2 content leads to narrowing of the thermal hysteresis (Fig. 4c), which indicates stabilization of the ordered C-CDW state. These observations lie in agreement with our confocal Raman measurements; consistently sharper CDW Raman features are observed for samples with a higher H-TaS2 content (Fig. 1e and Supplementary Figs. 15 and 16).

Fig. 4: Multistate resistance and chirality switching in TaS2 heterostructures.
figure 4

a Optical micrograph of a mesoscopic device fabricated from annealed flake S1. Scale bar: 10 μm. b Diagrams of sample composition for S1 and S4. For S1, diagrams are derived from atomic resolution DPC-STEM data, while for S4 the structure is proposed based on the ΔOCR-derived H-TaS2 content. Note, for S4, the H-TaS2 layers are randomly placed in the model. c Temperature-dependent resistance of S1 in regions R1–R3 and S4 in region R2. The marked x/n denotes the H-TaS2 proportion. d High-temperature section of c. Curves were vertically shifted for clarity. Inflection points in the cooling curve are marked by a circle. Scaling factors of the resistance curves are indicated on the right. Inset shows the resistance modulation of S4--R2 in five thermal cycles. Temperature ramp rate in c, d was 1 K/min. e Representative temperature-dependent 4D-STEM diffraction patterns, displayed on a logarithmic scale, for a 20-layer heterostructure with five H-TaS2 lamella. Insets display a zoomed-in view of a primary Bragg spot with labeled CDW superstructure peaks. Scale bars: 5 nm−1. f Normalized mean intensity of α, β and incommensurate (IC) diffraction peaks in 4D-STEM datasets obtained at different temperatures for the heterostructure in e. g Temperature-dependent (α − β)/(α + β+IC) calculated from f. For eg, heating was performed in situ and the data were acquired upon cooling from 373 K at 1 K/min.

Interestingly, the formation of H-TaS2 layers dictates the stepwise evolution of resistance and chirality with temperature in these endotaxial polytype heterostructures. Upon cooling, we observe a series of stepped resistance increases (Fig. 4d), with the step count matching the number of 1T-TaS2 slabs determined from DPC-STEM images of device cross-sections. Thus, the H-TaS2-separated 1T-TaS2 lamellae behave as isolated crystals with distinct IC-C CDW transitions, likely due to the electronic decoupling imposed by the metallic (H-TaS2) spacers23. For this reason, the number of H spacer layers deterministically encodes the number of resistance steps, and the profile (magnitude of resistivity change) of each step is governed by the various thicknesses of the 1T-TaS2 segments; thicker 1T-TaS2 slabs exhibit sharper CDW transitions, as observed in freestanding 1T-TaS2 crystals17,18. We note that resistance traces upon warming are noticeably broadened relative to the respective cooling sweeps. This may be understood by considering that strength of defect pinning is contingent on the CDW phase15,17,28,39. Defects may exert stronger pinning effects on the CDW domains in the localized C-CDW state compared to the “melted" IC-CDW phase, leading to less well-defined transitions upon warming. Nevertheless, the stepwise resistance transitions are highly reproducible in subsequent cooling cycles (Fig. 4d and Supplementary Figs. 17 and 18d). Moreover, in addition to the resistance steps upon cooling, temperature-dependent 4D-STEM reveals stepwise appearance of the α (L) and β (R) C-CDW enantiomorphs in concert with vanishing of the achiral IC phase (Fig. 4e-g). Thus, the IC-C CDW transition in our endotaxial heterostructures is marked both by simultaneous evolution of resistance and overall chirality (Fig. 4g), defined by the proportion of chiral superlattices: (α-β)/(α+β+IC). This synchronous switching sets the stage for optoelectronic devices combining charge and chirality degrees of freedom.

Polytype nucleation and designer heterostructures

Lastly, having established the structure and multistate electronic/chiral switching in TaS2 endotaxial heterostructures, we turn to considering the mechanism of nucleation of these polytype transformations, finding them to be facilitated by wrinkles and folds. We observed that polymorph transitions in 1T-TaS2 flakes, evidenced by changes in ΔOCR and sharpness of Raman spectral features associated with CDW modes, generally emanate from wrinkles, tears, and folds in flakes (Fig. 5a–c and Supplementary Fig. 20). These microscale structural defects inevitably introduce differential stress, which can be accommodated by strain (Fig. 5d)40,41,42, and this strain in turn can alter the energetic barrier between polytypes43,44,45,46. Accordingly, one explanation for the emanation of H-TaS2 at these features, is a strain-induced decreased energetic barrier for the H–T transformation, facilitating in nucleation of polytypic domains near stress points. In addition, differential stress in layered materials can also be accommodated by shear and slip between layers (Fig. 5e)42. This leads to the formation of extended shear dislocations42, which as we observed in Fig. 1g–j, appear to be a prerequisite for the formation of 1T-TaS2/1H-TaS2/1T-TaS2 interfaces from 1T-TaS232. Specifically, in native 1T-TaS2, the Ta ions are directly aligned in the out-of-plane direction (Fig. 5f). However, upon transformation of one layer to H-TaS2, crystallographic stacking with direct S–S overlap would be encountered (Fig. 5g). We find this configuration to be, on average, 8 meV per Å2 higher in energy according to density functional theory calculations (see “Methods" section for calculation details). To eliminate this unfavorable interlayer interaction, one of the 1T-TaS2 layers can slip across the trigonal prismatic interface (Fig. 5h), precisely as observed in Fig. 1g–j. Note that interlayer slips are readily present at flake folds, wrinkles and tears in 2D materials. Thus, polytype domains may form more readily in those defect regions of 1T-TaS2. We surmise that polytype transitions nucleate at macroscopic defect points due to the formation of extended dislocations and/or decreasing of the 1T-H energy barrier due to strain.

Fig. 5: Nucleation of endotaxial polytype transformation.
figure 5

a, b Optical image of a 1T-TaS2 flake before (a) and after (b) thermal annealing. A flake fold is marked with a magenta arrow. c Atomic Force Microscopy (AFM) map of b. d, e Models of 1T-TaS2 bending at a fold or wrinkle that are accommodated by in-plane strain (d) or interlayer shear and slip e. Models in d and e are adapted from ref. 42. fh Stacking configurations of three TaS2 structures: interface of three 1T-TaS2 layers (f), 1T-TaS2/1H-TaS2/1T-TaS2 interface without layer sliding (g) and 1T-TaS2/1H-TaS2/1T-TaS2 interface with shifting of the topmost 1T-TaS2 layer (h). Violet dashed vertical lines in fh indicate the alignment of Ta ions between the three TaS2 layers. Black dashed lines in (g) represent direct vertical overlap of sulfur ions between the topmost and middle layers. i, j Optical micrograph (i) and illustration (j) of an hBN/1T-TaS2 heterostructure before annealing. k Optical micrograph of flake in i after annealing. Yellow arrows indicate the direction of propagation of the blue H-TaS2 domains. l AFM height profile of flake in i, k in the region marked with a magenta arrow. m Map of ΔOCR after annealing overlaid on the optical micrograph (k). Flake outlines are added in i, k, and m. All scale bars are 5 μm.

We build upon these nanoscale insights to demonstrate that mechanical/strain engineering of 1T-TaS2 may be used for rational design of vertical/lateral TaS2 heterostructures. To this end, we stacked a 14-layer 1T-TaS2 crystal onto a ~ 112 nm-thick hexagonal boron nitride (hBN) flake to deliberately impart local stress onto the region of the 1T-TaS2 flake in the vicinity of hBN (Fig. 5i,j). Indeed, after annealing, the H-TaS2 polytype formation, evidenced by ΔOCR, radiates away from the hBN/TaS2 interface (Fig. 5k–m). It is important to note that coincident vertical/lateral heterostrucures are only formed upon annealing at moderate temperatures for short time periods, as extended or repeated heating leads to formation of predominantly homogeneous structures (Supplementary Fig. 21)23. Accordingly, intricate, multi-component device architectures could be realized by combining substrate patterning with moderate thermal annealing conditions.


In conclusion, we have demonstrated that highly tunable, verti-lateral polytype heterostrucures of 1T-TaS2 and H-TaS2 can be synthesized by moderate thermal annealing of nano-thick 1T-TaS2 flakes. Stress points in 1T-TaS2 are nucleation sites for the H-TaS2 domains, resulting in multi-component flakes with coexisting vertical and lateral heterostructures. The polytype composition of these heterostructures can now be conveniently determined using the optical contrast methodology developed in this work, enhancing the accessibility for characterizing and studying complex TaS2 crystals. Further, we use electron microscopy, Raman spectroscopy, and electronic transport studies to show that altering the H/1T ratio opens manifold possibilities for tailoring structural and electronic behavior of mixed polymorph crystals. Interlayer coupling between 1T-TaS2 fragments is interrupted by H-TaS2 layers, and the increasing content of H-TaS2 correlates with greater CDW commensuration, the appearance of optically detectable heterochiral CDW superlattices, and tunable chirality by modulating the α/β ratios. Furthermore, decoupled 1T-TaS2 fragments transition from the IC-CDW to the C-CDW state independently at high temperatures and the resulting temperature-driven, multistate phase transition is highly predictable: the number and size of resistance and steps is defined by the count and placement of H-TaS2 layers within the polytype heterostructure. Moreover, the changes in resistance are accompanied by corresponding changes in chirality, resulting in simultaneous modifications of both optical and electrical properties. Given the precedent for fast switching in 1T-TaS2 using electrical5,7,15,27,47,48,49,50 and optical fields35,51, endotaxial H-TaS2/1T-TaS2 heterostructures offer a rich framework as designer CDW materials that should exhibit multi-level switching between chiral CDW phases. The ability to predictably fabricate and control such well-defined phase change materials using low-energy external stimuli is a highly promising roadmap toward next-generation computing and data storage.


Mechanical exfoliation of 1T-TaS2

The mechanical exfoliation of 1T-TaS2 (HQ Graphene) is done in an Ar glovebox using an adhesive tape (Magic Scotch). The crystals are exfoliated onto 90 nm SiO2/Si wafers, whose oxide layer is formed by dry chlorination followed by annealing in forming gas (Nova Electronic Materials). First, wafers are cut into ~1 × 1 cm pieces and cleaned for 2 mins in an oxygen plasma cleaner. The chips are then heated to 200 °C on the glovebox hotplate while tessellating a sizeable (~3 × 3 mm) 1T-TaS2 crystal with the adhesive tape. Following this, chips are taken off the hotplate, and placed shiny-side-up onto the tape while still warm. The chips are then pressed for 10 mins with finger pressure. Lastly, the tape is swiftly taken off the chips.

Mechanical exfoliation of hexagonal boron nitride

The mechanical exfoliation of hBN (used as received from T. Taniguchi and K. Watanabe) is performed in atomospheric conditions. First, a 90 nm SiO2/Si wafer (Nova Electronic Materials) is cut into ~1 × 1 cm pieces and cleaned for 90 mins in an ozone cleaner. Immediately before the chip cleaning is complete, 3 hBN crystals (~ 1.5 mm × 1.5 mm) are tessellated with an adhesive tape (Magic Scotch). Promptly after cleaning the chips, they are placed shiny-side-up onto the tape and pressed for 10 mins with finger pressure. After this, the tape is swiftly taken off the chips.

Thermal annealing of 1T-TaS2 crystals

The 1T-TaS2 flakes were annealed in high-vacuum (approximately 10−7 Torr) by rapidly warming to 120 °C at 60 °C/min with a 5-minute hold. This is followed by heating at 11.5 °C/min to 350 °C and a 30 min hold at 350 °C, before rapidly cooling to room temperature at 13.5 °C/min.

Determination of layer count for TaS2 flakes

For flakes S1 and S2 that were studied by differential-phase-contrast scanning transmission electron microscopy (DPC-STEM)52, we counted the number of layers in the atomic resolution data. For all other flakes, the number of layers was identified using optical contrast (OC) measurements, following the relationship between OC and thickness established in ref. 53. The OC data was commonly obtained in conjunction with atomic force microscopy (AFM) to corroborate the OC-derived thickness.

Preparation of samples for transmission electron microscopy

For c-axis imaging, we prepared our samples within an Ar glovebox using a custom-built transfer stage. This involved using a polymeric stamp composed of a poly(bisphenol A carbonate) (PC) film covering a polydimethylsiloxane (PDMS) square on a glass slide. The PC/PDMS stamp was created by initially preparing a solution of PC in chloroform with a concentration of 5 % w/w. The solution was then dispensed onto a glass slide using a pipette and evenly distributed by placing it between two glass slides, which were promptly separated. Subsequently, the slides were positioned with the PC side facing up on a hotplate set at 120 °C for 5 mins, resulting in the formation of a relatively uniform PC film. This PC film was then precision-cut into small squares, ~2 mm × 2 mm in size, using a razor blade. Additionally, squares of PDMS, measuring approximately 4 mm × 4 mm, were prepared and affixed to glass slides. To complete the stamp, a PC square was centered within the PDMS square. Finally, the stamp was placed with the polymer side facing up on a hotplate set at 120 °C for 2 mins. Next, the PC/PDMS stamp is used to pick up 1T-TaS2 flakes. To this end, 1T-TaS2 flakes of interest were covered in PC for 3 mins at 160 °C, followed by rapid cooling to room temperature. Then, flakes adhered to the PC/PDMS stamp were placed onto a 200 nm silicon nitride holey TEM grid (Norcada) by melting the PC polymer at 160 °C. The TEM grid underwent a 5-minute cleaning process with O2 plasma immediately before stacking to enhance flake adhesion. After stacking, the PC film was dissolved in chloroform for 20 mins under ambient atmosphere, washed in isopropanol and dried with flowing N2.

For imaging along the crystallographic ab-plane, cross-sectional TEM samples were prepared by standard the standard focused ion beam (FIB) lift-out procedure using Thermo Fisher Scientific Helios G4 and Scios 2 FIB-SEM systems. A 200 nm coating of Pt was deposited over the region of interest using an electron beam at 5 kV and 1.6 nA. This was followed by a deposition of 2.5 μ m Pt using a gallium-ion beam at 8 kV and 0.12 nA. The initial lift-out was performed with a 30 kV Ga beam and 3 nA probe current, followed by a 1 nA current for the lift-out cleaning. Next, the sample was milled with 30 kV and 0.5 nA until reaching ~ 1 μm thickness, followed by 16 kV and 0.23 nA thinning to 0.5 μm. Next, a 5 kV beam, operated between 77-48 pA, was used to thin the sample to electron transparency. Lastly, final polishing was done at 2 kV and 43 pA.

Nanofabrication of mesoscopic devices from TaS2 heterostructures

All nanofabrication steps were performed in the Marvell Nanofabrication Laboratory. In a typical procedure, electron beam lithography (100 kV Crestec CABL-UH Series Electron Beam Lithography System) was used to define electrical contacts. PMMA (polymethyl methacrylate) e-Beam resist was used for this purpose (950 PMMA A6, MicroChem). After lithography, reactive ion etching (RIE) with a mixture of 70 sccm CHF3 and 10 sccm O2 (Semigroup RIE Etcher) was used to remove top TaS2 layers and expose a fresh surface immediately before evaporating Cr/Pt (1 nm/100 nm) (NRC thermal evaporator). After overnight metal lift-off in acetone, electron-beam lithography was used to define a Hall bar-shaped etch mask. Etching of the Hall bar was accomplished by reactive ion etching with 70 sccm SHF6 and 8.75 sccm O2 (Semigroup RIE Etcher). After the SHF6/O2 treatment, samples were cleaned for 20 seconds in a 40 sccm O2 plasma and consecutively the PMMA resist was dissolved in acetone for 20–30 mins.

Transmission electron microscopy

Selected area electron diffraction (SAED) patterns were obtained with a 40 μm diameter aperture aperture (defining a selected diameter of ~ 720 nm on the sample) using FEI TitanX TEM operated at 60–80 kV.

Four-dimensional scanning transmission electron microscopy (4D-STEM) was performed on FEI TitanX (80 keV, 0.3 mrad indicated convergence semi-angle) with the Gatan 652 Heating holder for in-situ heating experiments. The 4D-STEM data was analyzed with the py4DSTEM Python package37. First, peak detection in py4DSTEM was used to identify the primary Bragg peaks and construct the reciprocal vectors of the host lattice. Next, CDW reciprocal vectors were calculated from symmetry relations to the primary Bragg vectors. Subsequently, we constructed virtual apertures that mask the entirety of the diffraction space except for the CDW (α, β or IC) satellite peak regions (Supplementary Fig. 6a, b). These CDW virtual apertures were then applied to the 4D-STEM diffraction data to integrate the CDW intensities at each probe position. Note, we only integrated over the CDW satellite peaks around the second and third-order primary Bragg peaks to minimize the diffuse scattering contribution. Further, we also defined a background (bg) virtual aperture to measure the diffuse scattering background for each diffraction pattern (Supplementary Fig. 6a, b). This background was subtracted from the integrated CDW intensities at each probe position for a more accurate calculation of enantiomorphic disproportion. Lastly, for the in-situ heating data, we constructed small virtual apertures that exclude overlap regions between the IC and the α/β peaks (Supplementary Fig. 6b).

Differential phase contrast (DPC) STEM images52 were collected on a Thermo Fisher Scientific Spectra 300 X-CFEG operating at 300 kV with a probe convergence angle of 21.4 mrad. The inner and outer collection angles of the quadrant detector were 15 and 54 mrad respectively. DPC-STEM images were reconstructed from the component images output by each quadrant using the py4DSTEM package37.

Raman spectroscopy

Ultra-low frequency (ULF) Raman spectra (Horiba Multiline LabRam Evolution) of TaS2 heterostructures were obtained using a 633 nm laser excitation with the corresponding ULF notch filters at a power of 50–80 μW with 10 s acquisition times and 3 accumulations.

Circularly polarized Raman spectroscopy was performed in backscattering configuration along the ZZ direction. A 100 x objective (N.A. = 0.80) was used for focusing the incident beam onto the sample and for collecting the scattered light. The Raman laser spot size was ~ 1 μm. Circularly polarized light was achieved by using two linear polarizers (LP1 and LP2) and a λ/4 waveplate (Supplementary Fig. 12).

Electron transport measurements

Transport measurements were performed using standard lock-in techniques. Briefly, a 1 μA alternating current (17.777 Hz) was applied between the source and drain contacts while sweeping the temperature in the PPMS DynaCool system. Concurrently, the longitudinal (Vxx) voltage was measured with the SR830 lock-in amplifier. All phases were ≤ 5, and the resistances were determined from Ohm’s law. All data displayed in this work displays the four-probe resistance except for the measurement for S1-R1, which was taken in the 3-probe configuration.

Density functional theory calculations

Density Functional Theory (DFT) calculations were carried out using the Vienna Ab initio Simulation Package (VASP)54,55,56,57 with projector augmented wave (PAW) pseudopotentials58,59 including Ta 5pd, 6s, and S 3sp electrons as valence. The plane-wave energy cutoff was set to 400 eV and the k-point grids were Gamma-centered with a k-point spacing of 0.3 Å−1 for the 1T phase and 0.2 Å−1 for the 1H phase, which gave an energy convergence of 1 meV per atom. The convergence criteria for the electronic self-consistent loop was set to 10−7 eV. Structural optimizations were done using the PBEsol60 exchange-correlation functional until the residual forces on the ions were less than 0.001 eV Å−1.

The energy gain for the interlayer slip observed by DPC-STEM was determined by DFT calculations on mixed phase structures. A shift between 1H and 1T layers of (2a/3, b/3), where a and b are the 1H lattice parameters, was introduced to eliminate direct S-S overlap. This shift is comparable to the offset between layers in the 6R-TaS2 phase. In a 6-layer 1H/1T-α/1T-α stack, the energy gain with interlayer slip was 8 meV Å−2.