# Fileset

[s41467-025-66594-y.pdf](https://mdr.nims.go.jp/filesets/f27ea5a9-c6aa-47b5-b937-0f188f8a72b2/download)

## Creator

Keyu Wei, Yixuan Luo, [Kenji Watanabe](https://orcid.org/0000-0003-3701-8119), [Takashi Taniguchi](https://orcid.org/0000-0002-1467-3105), Yanfeng Guo, Xiaoxiang Xi

## Rights

[Creative Commons BY-NC-ND Attribution-NonCommercial-NoDerivs 4.0 International](https://creativecommons.org/licenses/by-nc-nd/4.0/)

## Other metadata

[Gate tuning of coupled electronic and structural phase transition in atomically thin Ta2NiSe5](https://mdr.nims.go.jp/datasets/00a40184-54d7-4084-9b8c-d61c4c0c5231)

## Fulltext

Gate tuning of coupled electronic and structural phase transition in atomically thin Ta2NiSe5Article https://doi.org/10.1038/s41467-025-66594-yGate tuning of coupled electronic andstructural phase transition in atomically thinTa2NiSe5Keyu Wei 1, Yixuan Luo2, Kenji Watanabe 3, Takashi Taniguchi 4,Yanfeng Guo 2,5 & Xiaoxiang Xi 1,6,7Realizing an excitonic insulator phase from narrow-gap semiconductorsremains challenging, as unambiguous experimental signatures are difficult toestablish. Ta2NiSe5 has been widely regarded as a leading candidate, yet thenature of its phase transition and insulating state remains controversial. Here,we report a systematic Raman spectroscopy study of Ta2NiSe5 as a function ofthickness and field-effect doping, complemented by electrical transportmeasurements. The phase transition persists down to the monolayer limit,with the critical temperature increasing as thickness decreases. In bilayersamples, both electron and hole doping suppress the insulating state, withelectron doping lowering and hole doping raising the transition temperature.Importantly, the quasi-elastic scattering, previously attributed to excitonicfluctuations, evolves monotonically across the entire doping range, incon-sistent with the expected suppression of excitonic correlations by Coulombscreening. These findings rule out a dominant excitonic mechanism andinstead point to a coupled electronic and structural phase transition, whosestability is tunable by carrier doping. Our doping-based approach offers ageneral strategy for evaluating the role of excitonic effects in candidate exci-tonic insulators.In semiconductors or semimetals, unscreened Coulomb interactioncan lead to the formation of bound electron-hole pairs known asexcitons. If the exciton binding energy exceeds the band gap or bandoverlap, these composite bosons may condense at low temperatureinto a macroscopic quantum state—an excitonic insulator (EI)1–5. As adistinct type of correlated insulator, the EI has attracted sustainedinterest both as a platform to study electronically driven phase tran-sitions and as a potential host for macroscopic quantum coherence,including superfluid-like behavior6,7. While significant advances havebeen made in engineered bilayer systems with spatially separatedelectrons and holes8–18, realizing an EI phase in natural bulk crystalsremains experimentally challenging. In addition to the scarcity ofcandidate materials, a major obstacle is the lack of unambiguousexperimental signatures unique to exciton condensation19–27.A key complication in bulk EI candidates is the ubiquitous pre-sence of electron-phonon coupling, which makes it difficult to isolatethe electronic origin of the insulating ground state. For example, in theprototypical EI candidate 1T–TiSe2, exciton condensation has beenproposed to drive the emergence of a charge-density wave (CDW)19,20.However, the CDW ordering wavevector coincides with a latticeReceived: 28 May 2025Accepted: 10 November 2025Check for updates1National Laboratory of Solid State Microstructures and Department of Physics, Nanjing University, Nanjing, China. 2State Key Laboratory of QuantumFunctional Materials, School of Physical Science and Technology, ShanghaiTech University, Shanghai, China. 3Research Center for Electronic and OpticalMaterials, National Institute forMaterials Science, Tsukuba, Japan. 4ResearchCenter forMaterials Nanoarchitectonics, National Institute forMaterials Science,Tsukuba, Japan. 5ShanghaiTech Laboratory for Topological Physics, ShanghaiTechUniversity,Shanghai, China. 6Collaborative InnovationCenter of AdvancedMicrostructures, Nanjing University, Nanjing, China. 7Jiangsu Physical Science Research Center, Nanjing, China. e-mail: guoyf@shanghaitech.edu.cn;xxi@nju.edu.cnNature Communications |        (2025) 16:10999 11234567890():,;1234567890():,;http://orcid.org/0009-0008-4039-4936http://orcid.org/0009-0008-4039-4936http://orcid.org/0009-0008-4039-4936http://orcid.org/0009-0008-4039-4936http://orcid.org/0009-0008-4039-4936http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0003-3701-8119http://orcid.org/0000-0002-1467-3105http://orcid.org/0000-0002-1467-3105http://orcid.org/0000-0002-1467-3105http://orcid.org/0000-0002-1467-3105http://orcid.org/0000-0002-1467-3105http://orcid.org/0000-0002-9386-4857http://orcid.org/0000-0002-9386-4857http://orcid.org/0000-0002-9386-4857http://orcid.org/0000-0002-9386-4857http://orcid.org/0000-0002-9386-4857http://orcid.org/0000-0002-8685-9267http://orcid.org/0000-0002-8685-9267http://orcid.org/0000-0002-8685-9267http://orcid.org/0000-0002-8685-9267http://orcid.org/0000-0002-8685-9267http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-025-66594-y&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-025-66594-y&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-025-66594-y&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-025-66594-y&domain=pdfmailto:guoyf@shanghaitech.edu.cnmailto:xxi@nju.edu.cnwww.nature.com/naturecommunicationsdistortion involving a soft phonon mode, making it difficult to disen-tangle electronic and lattice contributions28,29. In contrast, Ta2NiSe5 is adirect-gap semiconductor, avoiding complications from finite-wavevector CDWs. Several experimental works support an EI state inthis material, including the flattening of the valence band top revealedby angle-resolved photoemission spectroscopy (ARPES)30,31, a dome-shaped electronic phase diagram centered at zero band gap32, opticalsignature of exciton-phonon coupling33, and ultrafast optical mod-ulation of the insulating state34–37. Nevertheless, a structural phasetransition also occurs in this material38–40, raising the possibility thatthe insulating state is of structural rather than electronic origin41–44.Some theoretical works suggested that exciton condensation45 or evenan excitonic instability46 can drive this structural transition, whereasothers showed that the structural transition, rather than the excitoniceffect, is required to open the experimentally observed bandgap42,43,47,48. This entanglement of electronic and structural effects isnot unique to Ta2NiSe5 and continues to pose a challenge for the studyof EIs.In this work, we propose an experimental approach to assess theexcitonic nature of an insulating state by tracking its evolution undercarrier doping. Since excitons are central to the EI state, and sincecarrier doping enhances Coulomb screening, possible hallmarks of anEI should be suppressed by both electron and hole doping. Thisanticipated doping dependence serves as a useful criterion for evalu-ating the role of excitonic effects. We implement this approach inatomically thin Ta2NiSe5, employing field-effect gating to achievecontinuous and reversible doping. Bymonitoring the Raman response,we find that the quasi-elastic scattering (QES), previously attributed toexcitonic fluctuations, evolves monotonically with doping across awide range on both the electron and hole sides. This observation isinconsistent with expectations from an excitonic mechanism butsupports a scenario in which the insulating state originates from astructural phase transition driven by electron-phonon coupling47 andtunable via carrier density.ResultsCoupled electronic and phononic excitationsBulk Ta2NiSe5 consists of monolayers stacked along the b-axis via vander Waals interactions (Fig. 1a), crystallizing in an orthorhombicstructure (space group Cmcm) at high temperature49,50. Within eachmonolayer, Ta and Ni atoms are coordinated by Se atoms in approxi-mately octahedral and tetrahedral geometries, respectively, formingchain-like motifs along the a-axis. Below TC = 328K, the systemundergoes a second-order phase transition into a monoclinic phase(space group C2/c), associated with a putative EI state. The two Tachains flanking each Ni chain exhibit shear-like displacements inopposite directions (see arrows in Fig. 1a), breaking the in-planemirrorsymmetries and giving rise to an antiferroelectric or ferro-rotationalorder38–40.This phase transition is clearly manifested in the temperature-dependent Raman spectra (Fig. 1b, c), which show drastic changesboth in the low-energy QES and in selected phonon modes. Near TC,the QES is strongly enhanced and hybridizes with the phononmode atapproximately 70 cm−1, and the coupling of these two features isenabled by their compatible symmetry (B2g in the orthorhombic phaseand Ag in the monoclinic phase)51–53, as further confirmed bypolarization-angle dependent Raman measurements (Fig. 1d). Follow-ing ref. 51, we label the phonon modes numerically in order ofincreasing frequency. The crossed polarization (ac) configuration isadoptedwith the polarization angle thatmaximizes the intensity of theQES and mode 2, as they are particularly sensitive to the phase tran-sition. Indeed, both mode 2 and mode 5 have been shown to exhibitTaNiSebcca1 2 3 4 5546632sssFig. 1 | Crystal structure and Raman characteristics of Ta2NiSe5. a Crystalstructure of Ta2NiSe5 in the orthorhombicphase. The arrows in the lower part indicateatomic displacement in the monoclinic phase. b, c Temperature-dependent Ramanspectra (b) and the corresponding intensity color plot (c) of bulk Ta2NiSe5.d Polarization-angle-dependent Raman intensity color plot of bulk Ta2NiSe5 at 300K.e Fitting analysis of the low-energy Raman conductivity of bulk Ta2NiSe5 at 300K.f Thickness-dependent Raman spectra of Ta2NiSe5 at 150K. All data except those ind were collected in the crossed polarization (ac) configuration.Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 2www.nature.com/naturecommunicationsshearing vibration of the Ta atoms along the chain direction, whichmatches the form of the monoclinic lattice distortion47.QES was previously observed in iron pnictide superconductors54and magnetic materials55–57, where it was attributed to charge nematicfluctuations and magnetic fluctuations, respectively. In Ta2NiSe5,however, since magnetic order is absent50, no evidence of electronicnematicity was reported, and all expected Raman-active phononmodes can be accounted for51, the QES was attributed to an electronicorigin51,52. The QES exhibits a Lorentzian lineshape centered at zerofrequency, resembling that observed in nematic and magnetic sys-tems, whichmotivates a similar analysis framework54–57. The measuredRaman intensity I(ω) is converted to the Raman susceptibility,χ″(ω)∝ I(ω)/[n(ω, T) + 1], where n(ω, T) is the Bose-Einstein distributionat temperature T. The Raman conductivity, defined as χ″(ω)/ω, ismodeled as a Lorentzian function centered at zero frequency, σ0Γ/(ω2 + Γ2), where σ0 represents the spectralweight and Γ is the half-widthof the QES peak. The coupling between the QES and phonon mode 2leads to a Fano lineshape for the latter, modeled by the expressionA1 +q2ðϵ +qÞ2ϵ2 + 1 , where ϵ = (ω −ω2)/γ, and A, ω2, and γ denote the amplitude,frequency, and half-width of the phonon mode, respectively. Theparameter 1/∣q∣ quantifies the coupling strength between the phononand the QES. A model incorporating both the Lorentzian and the Fanoterms provides an excellent fit to the spectra at all temperatures whereQES is present. A representative fit at 300K is shown in Fig. 1e, withresults at other temperatures provided in Supplementary Note 1. Thisanalysis offers a robust method for quantitatively tracking the evolu-tion of both the QES and mode 2 under varying experimentalconditions.Thickness dependenceAtomically thin Ta2NiSe5 samples were obtained by mechanical exfo-liation. Thenumber of layerswasdeterminedbasedonoptical contrastand verified using atomic force microscopy (Supplementary Note 2).The frequency of mode 5 varies systematically with thickness (Fig. 1f),assisting the identification of the layer number. The persistence ofsharp phonon peaks in samples down to the monolayer limit indicatesgood sample quality, in contrast to an earlier study where Ramansignals were undetectable for samples thinner than four layers, likelydue to degradation58.We performed systematic temperature-dependent Raman mea-surements on samples with varying thickness. Figure 2a–d reveals thatthe phase transition persists down to the monolayer limit. Notably, TCincreaseswithdecreasing thickness, as indicatedby the temperature atwhich the QES reaches its maximum intensity and by the coexistenceof modes 4 and 5 persisting to higher temperatures (see Supplemen-tary Note 3 for mode analysis). Analysis of the QES and mode 2 usingthe method described above yields the parameters shown in Fig. 2e–i.For the bulk sample, the QES spectral weight σ0 is maximized at the TCdetermined from resistancemeasurements32,50, strongly supporting itselectronic origin. This maximum is retained in atomically thin samplesand is used to extract a thickness-dependent TC by fitting the QEStemperature dependence above 250K to a Gaussian function, whichshows a systematic enhancement as the thickness decreases (bluecircles in Fig. 2j). The monolayer sample exhibits enhanced back-ground scattering when heated above 400K, likely due to laser-induced degradation at elevated temperatures. Its spectral weight fortheQES at high temperature and TC cannot be reliably analyzed for thisreason. The data shown in Fig. 2d have been corrected for this back-ground contribution, as detailed in Supplementary Note 4. This issuewas not observed in bilayer or thicker samples.Fitting parameters associated with mode 2, including its ampli-tudeA, frequencyω2, linewidth2γ, andFanoasymmetry 1/∣q∣ are shownin Fig. 2f–i. The amplitude shows a minimum, whereas the other threeparameters showamaximumnearTC. These trends are consistentwithprevious bulk measurements51. The characteristic temperatures cor-responding to the extrema in theparameters systematically increase asthickness is reduced, asmarked by arrows in Fig. 2f–i and summarizedin Fig. 2j, corroborating the enhancement of TC in atomically thinsamples. This finding contrasts with an earlier report of TC reductionupon decreasing thickness58. Quantifying TC frommode 2 parameters,however, is less reliable, as they show different characteristic tem-peratures even for the same sample. This could be due to theFig. 2 | Thickness and temperature dependent Raman conductivity.a–d Temperature-dependent Raman conductivity color plot for bulk, trilayer,bilayer, and monolayer samples. e Temperature dependence of the QES spectralweight. High-temperature data points aremissing because the QES signal becomestoo weak for reliable fitting. The solid lines are Gaussian fits. f–i Temperaturedependenceof the amplitude, frequency, full width, and 1/∣q∣ofmode2. jThicknessdependence of the temperature corresponding to the maximum or minimumvalues marked by arrows in (e–i) (filled symbols). The open squares are TC valuesestimated from electrical transport measurements. Error bars in (e–i) are standarddeviations obtained from fitting analysis, and those in (j) represent uncertainties indetermining the characteristic temperatures. All data were collected in the crossedpolarization (ac) configuration.Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 3www.nature.com/naturecommunicationsadditional effect of lattice anharmonicity47, which becomes enhancedat high temperature and near structural phase transitions. Anothersource of discrepancy between the different criteria for determiningTC using mode 2 parameters arises from the data analysis itself. Inprinciple, this could be remedied by employing a fully self-consistentanalysis that couples the QES and the bare phonon52. However, forTa2NiSe5, such an approach has been shown to have only a minimalinfluence on the analyzed results51. We further performed resistancemeasurements on samples with varying thickness (SupplementaryNote 5). The kink in the temperature derivative of the resistance as anindication of the phase transition58 was observed in all measuredsamples, although it becomes weaker in atomically thin samples. Thecorresponding TC values are plotted in Fig. 2j as open symbols, whichconfirm the increased TC as the sample thickness decreases.Effect of doping on electrical transportHaving established the phase transition in atomically thin samples, wenow investigate how it evolves under carrier doping. To this end, weintegrated bilayer Ta2NiSe5 into a dual-gate field-effect transistor con-figuration, as illustrated schematically in Fig. 3a. A typical device image isshown in Supplementary Note 5. Two-probe resistance measurementswere performed at 150K while sweeping both the top-gate voltage (Vt)and bottom-gate voltage (Vb). As shown in Fig. 3b, the resistance peaksalong a line corresponding to charge neutrality. This charge-neutralityline is slightly offset from theorigin, possibly due to extrinsic doping. Tofocus on the effect of pure carrier doping, we follow the Vt-Vb relationthat ensures cancellation of the gate-induced electric field (see “Meth-ods”), as marked by the arrowed line in Fig. 3b. Both electron and holedoping progressively suppress the insulating state, as seen in the resis-tance versus doping curve in Fig. 3c. Here, n denotes the sheet carrierdensity, with positive (negative) values representing electron (hole)doping. The resistance exhibits a maximum at charge neutrality, whichpersists to high temperature but becomes thermally broadened. Themaximum doping achieved amounts to approximately 0.06e/unit cell,estimated using the applied gate voltages and the thickness of thehexagonal boron nitride (hBN) dielectric layers.The temperature dependence of resistance at selected hole dop-ing densities is shown in Fig. 3d, revealing insulating behavior acrossthe accessible doping range. The electron-doped side exhibits quali-tatively similar behavior. An Arrhenius plot of the resistance (Fig. 3e)reveals two distinct thermal activation regimes. By fitting thetemperature-dependent resistance to R / expðΔ=2kBTÞ, where kB isthe Boltzmann constant, we extract two energy gaps: a larger gap, Δ1,which dominates above 250 K, and a smaller gap, Δ2, which governstransport at lower temperatures. Their doping dependence is sum-marized in Fig. 3f. While Δ1 is strongly doping-dependent and decrea-ses with increasing carrier density, Δ2 remains largely unchangedabove 1013 cm−2 for both electron and hole doping. The smaller gaplikely arises from defect-related in-gap states, consistent with priorscanning tunneling spectroscopy (STM) studies of exfoliatedTa2NiSe558. The values of Δ1 = 0.3 eV and Δ2 = 0.1 eV at charge neutralityare comparable to those reported in bulk samples32,58,59.At first glance, the suppression of insulating behavior by bothelectron and hole doping is consistent with the screening of an exci-tonic gap by doped carriers21. However, such ambipolar dopingdependence is not unique to EIs. For instance, in Bernal-stackedbilayergraphene featuring a non-excitonic band gap opened by an electricfield, the insulating state is similarly suppressed by both electron andhole doping60. In that case, the evolution of the activation gap iseffectively associated with a shift in the chemical potential. In bilayerTa2NiSe5, the persistence of Δ1 up to doping levels of 1013 cm−2, alongwith the gradual slowing of its suppression at higher doping, suggeststhat the gap formation is unlikely to be dominated by excitoniceffects47.Effect of doping on Raman scatteringTo further elucidate the effect of carrier doping, we examine the cor-responding evolution in the Raman spectra. Figure 4a–d showsdoping-dependent Raman conductivity maps at selected tempera-tures, with representative spectra displayed in Fig. 4e–h. At 150K, theQES is nearly absent at zero doping but is induced by electron doping.At 300K, QES is already present at charge neutrality and becomesGraphitehBNhBNGraphiteGrTa2NiSe5GrVtVbVsFig. 3 | Doping dependent electrical transport properties of bilayer Ta2NiSe5.a Schematic structure of the dual-gate bilayer Ta2NiSe5 device D17. Gr: graphiteelectrodes. b Dual-gate mapping of the resistance measured at 150 K. The dashedline marks the charge neutrality and the arrows denote pure doping. c Doping-dependent resistance at selected temperatures. d Temperature-dependentresistance at charge neutrality and selected hole dopings. e Arrhenius plot of theresistance at typical doping levels. The solid lines are fits to the thermally activatedtemperature dependence over two temperature ranges. f Doping dependence ofthe activation gaps. The shaded areas represent error bars obtained from thefitting analysis.Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 4www.nature.com/naturecommunicationsclearly enhanced (suppressed) by electron (hole) doping. At 350K,close to the TC estimated earlier, the doping dependence of the QESweakens. At 400K, the QES exhibits a doping dependence opposite tothat observed at lower temperatures. The doping tunability of the QESaffirms its electronic origin. Previous Raman studies attributed theQESat high temperatures to excitonic fluctuations51,52, an interpretationsupported by theoretical predictions61 and ARPES evidence for pre-formed excitons above TC62. However, our results challenge this view.If theQES indeed originated from excitonic fluctuations, both electronand hole doping would be expected to suppress it, as enhancedCoulomb screening weakens the excitonic bound states. Contrary tothis expectation, we observe a monotonic doping dependence,inconsistent with the excitonic fluctuation scenario.Recent x-ray scattering experiments revealed pronounced diffusescattering along the inter-chain direction in bulk Ta2NiSe5, peaking atTC and persisting over a broad temperature range, which suggestsdynamic lattice fluctuations associated with an inter-chain shearmodethat freeze into a static distortion in the monoclinic phase43. Thesefluctuations enable hybridization between Ta-derived conduction andNi-derived valence bands, otherwise symmetry-forbidden in theFig. 4 | Doping dependence of the Raman conductivity in bilayer Ta2NiSe5. a–d Doping-dependent Raman conductivity color plots for the bilayer sample D17 atselected temperatures. e–hRepresentative spectra corresponding to (a–d) at selected doping levels. All data were collected in the crossed polarization (ac) configuration.Fig. 5 | Doping and temperature dependent QES and mode 2 parameters forbilayer Ta2NiSe5 device D50. a Color map showing the doping and temperaturedependence of the QES spectral weight. b Line cuts along the temperature axis atzero doping (undoped), the maximum electron doping (n-doped), and the max-imum hole-doping (p-doped) in (a). The solid lines are Gaussian fits to the dataabove 250K, with the fitted peak center corresponding to the TC and the full widthat half maximum characterizing the temperature spread (ΔT) of the QES. c Dopingdependence of TC (left axis) and ΔT (right axis). d–f, The upper panels show thedoping and temperature dependent colormaps of the frequency, amplitude, and 1/∣q∣ formode 2. The lower panels are the corresponding line cuts at the samedopinglevels as those in (b). Error bars in (b–f) are standard deviations obtained fromfitting analysis.Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 5www.nature.com/naturecommunicationsorthorhombic phase63. Resonant inelastic x-ray scattering furthersupports this picture by revealing substantial Ta-Ni orbital hybridiza-tion above TC41, which can manifest as a pseudogap in the orthor-hombic phase, as indicated by ARPES43 and infrared spectroscopy33measurements. The QES intensity in our Raman data also peaks at TC,indicating a direct link to suchfluctuations and justifying using theQESto quantify them, analogous to its application for quantifying chargenematic fluctuations in iron-pnictide superconductors54.More systematic doping-dependent Raman measurements wereperformed at a finer temperature step in another bilayer device, whichallows us to extract the temperature dependence of the QES spectralweight at different doping levels. The results, shown in Fig. 5a, areconsistent with the doping dependence at selected temperaturesdescribed above. By fitting the temperature dependence of the QESabove 250K to a Gaussian function (Fig. 5b), TC is quantified as thecenter of the Gaussian peak, and its doping dependence is shown inFig. 5c. We estimate an enhancement of 5 K and a reduction of 34 Kunder the maximum hole and electron doping of 2.6 × 1013 cm−2,respectively. The reduction of TC under electron doping is consistentwith previous potassium dosing studies and DFT calculations, whichshow that electrondoping tends to stabilize the orthorhombic phase43.Thedopingdependenceof theQES at selected temperatures, as shownin Fig. 4, can therefore be attributed to the doping-induced modifica-tion of TC, which shifts the peak in the temperature dependence of theQES spectral weight (Fig. 5a, b). The spread of the QES temperaturedependence, quantified as the full width at half maximum of theGaussian peak, grows as TC decreases (Fig. 5c), indicating a moreextended fluctuation regime under electron doping.Phononmode 2 also exhibits clear doping dependence, especiallybelow TC (Fig. 4). At 150 K, where the QES is weak, mode 2 softens andbroadens with doping (Supplementary Note 6), which could be due todoping-induced enhancement of either the electron-phonon couplingor the electronic density of states at the chemical potential. Above~250K and below TC, when the QES becomes more prominent, thedoping dependence of mode 2 parameters becomes monotonic(Fig. 5d–f). Electron doping reduces the mode amplitude (Fig. 5e) butincreases its frequency (Fig. 5d), 1/∣q∣ (Fig. 5f) and linewidth (Supple-mentary Note 6), resembling the effect of increasing temperature justbelow TC (Fig. 2). A similar trend is observed for mode 5 (Supple-mentary Note 6), further corroborating electron-doping-enhancedlattice fluctuations when approaching TC from below. Both modesare sensitive to the structural phase transition and dominated by Taatomic vibration47. They couple efficiently with electronic states at thebottom of the conduction band formed by Ta-5d orbitals45, explainingthe more significant doping effect on the electron-doped side (seemore discussions in Supplementary Note 6).DiscussionThe phase transition in bulk Ta2NiSe5 has been attributed to a softzone-center B2g phonon, as predicted by DFT39. Raman spectroscopycan, in principle, detect such a mode. While K. Kim et al. observed nosoft-mode behavior51, M.-J. Kim et al. identified mode 2 as a soft modebased on measurements up to 800K64. In our data, mode 2 showsslight softening and broadening between 550K and 450K (Fig. 2a, g,h), but this trend is interrupted by the enhancement of the QES.Moreover, mode 2 retains a Fano lineshape across the orthorhombicphase, even when the QES is not well resolved (Fig. 1, SupplementaryNote 1), indicating persistent electronic fluctuations beyond the scopeof DFT. These observations support a phase transition driven byentangled electronic and structural instabilities47, which persist tohigher temperature at reduced dimensionality. This enhanced TC uponthickness reduction cannot be attributed to enhancedexcitonic effectsin reduced dimensions, because otherwise TC should have been sup-pressed by both electron and hole doping, contrary to the monotonicdoping dependence observed here. Instead, the enhanced TC could bedue to increased electron-phonon coupling that stabilizes monocliniclattice distortion up to higher temperatures, which remains to besubstantiated by further experimental and theoretical studies.Previous doping studies on bulk Ta2NiSe5 using alkali-metaldeposition were limited to electron doping and inevitably introducedsurface electric fields43,65,66. In contrast, our dual-gate field-effect setupenables independent tuning of carrier density and electric field, pro-viding a cleaner probe of intrinsic doping effects. We observed cleardoping-induced changes in the Raman response that are inconsistentwith excitonic condensation, while electric field effects were minimal(Supplementary Note 7). The field-effect doping approach in thebottom-gate geometry is also compatible with techniques such asARPES and STM, offering a versatile route to probe excitonic correla-tions in other van der Waals EI candidates.MethodsSample and device preparationTa2NiSe5 single crystals were synthesized using the flux method.Freshly cleaved surfaces of the bulk crystals were used for Ramancharacterization. Thin flakes of Ta2NiSe5 were exfoliated from the bulkcrystal onto polydimethylsiloxane (PDMS) substrates and identifiedbased on optical reflection contrast, performed in an argon atmo-sphere within a glovebox. The exact thickness was determinedthrough a combination of optical contrast analysis, atomic forcemicroscopy (AFM), and Raman spectroscopy (see SupplementaryNote 1). For Raman measurements, the flakes were encapsulated byhexagonal boron nitride (hBN) layers with a thickness of 5–15 nm onboth sides. Stacks were assembled by using a polycarbonate (PC) filmon a PDMS stamp to sequentially pick up eachmaterial. Afterward, thestacks were released onto a Si/SiO2 substrate at 200 °C, and immersedin chloroform for 10min and isopropanol for 5min to removethe PC film.Dual-gate field-effect transistors were fabricated for doping-dependent studies using the same exfoliation and transfer techni-ques. First, Ti/Au electrodes were patterned on Si/SiO2 substrates viadirect write laser lithography and thermal evaporation. Then, graphiteelectrodes (serving as contacts for Ta2NiSe5), hBN gate dielectric, andthe bottom-gate graphite electrode were sequentially picked up andreleased onto the Ti/Au electrodes. The stack was cleaned using anAFM in contactmode, with a contact force of 5–20 nN. Finally, the top-gate graphite electrode, hBN gate dielectric, and Ta2NiSe5 flake werepicked up and transferred onto the bottom-gate stack.CharacterizationsRaman spectroscopy was conducted in the back-scattering geometrywith 532 nm laser excitation. An incident power of 0.2mW was usedfor all samples, which was confirmed to cause a negligible heatingeffect. The scattered light passed through Bragg notch filters beforebeing collectedby a grating spectrograph and a liquid-nitrogen-cooledcharge-coupled device. Measurements were performed in a cryostatwithin the 150–350K range and in a heating stage between 300–550K.To ensure consistency between datasets, spectra in the overlapping300–350K range were compared and calibrated (see SupplementaryNote 8). Polarization-angle dependent measurements were carriedout using two polarizers and a half-wave plate. Two-probe resistancewas measured with a lock-in amplifier, applying an excitationcurrent of 10 nA. Gate voltages were applied using source meters.Pure doping is achieved by applying gate voltages along the arrowedline shown in Fig. 3b. The doped sheet carrier density is calculatedas n = ϵrϵ0(Vt/dt + Vb/db)/e, in which the dielectric constant of hBNis ϵr = 3, ϵ0 is the vacuum permittivity, e is the electron charge, Vt andVb are the top and bottom gate voltages, respectively, and dt and dbare the thickness of the corresponding hBN dielectric layers. Theelectric field along the charge neutrality line shown in Fig. 3b is E = (Vt/dt − Vb/db)/2.Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 6www.nature.com/naturecommunicationsData availabilityThe data generated in this study, as shown in themainmanuscript, areprovided in the Source Data file. All other data that support the find-ings of this study are available from the corresponding authors uponrequest. Source data are provided with this paper.References1. Mott, N. F. The transition to the metallic state. Philos. Mag. 6,287–309 (1961).2. Blatt, J. M., Böer, K. W. & Brandt, W. Bose-Einstein condensation ofexcitons. Phys. Rev. 126, 1691–1692 (1962).3. Keldysh, L. & Kopaev, Y. Possible instability of semimetallic statetoward Coulomb interaction. Sov. Phys. Solid State, USSR 6,2219–2224 (1965).4. Cloizeaux, J. D. Exciton instability and crystallographic anomalies insemiconductors. J. Phys. Chem. Solids 26, 259–266 (1965).5. Jérome, D., Rice, T.M. &Kohn,W. Excitonic insulator. Phys. Rev. 158,462–475 (1967).6. Eisenstein, J. P. & MacDonald, A. H. Bose–Einstein condensation ofexcitons in bilayer electron systems. Nature 432, 691 (2004).7. Wu, S. et al. Charge-neutral electronic excitations in quantuminsulators. Nature 635, 301–310 (2024).8. Eisenstein, J. P. Exciton condensation in bilayer quantum Hall sys-tems. Annu. Rev. Condens. Matter Phys. 5, 159–181 (2014).9. Fogler, M. M., Butov, L. V. & Novoselov, K. S. High-temperaturesuperfluidity with indirect excitons in van der Waals hetero-structures. Nat. Commun. 5, 4555 (2014).10. Du, L. et al. Evidence for a topological excitonic insulator in InAs/GaSb bilayers. Nat. Commun. 8, 1971 (2017).11. Liu, X.,Watanabe, K., Taniguchi, T., Halperin, B. I. & Kim, P. QuantumHall drag of exciton condensate in graphene. Nat. Phys. 13,746–750 (2017).12. Li, J. I. A., Taniguchi, T., Watanabe, K., Hone, J. & Dean, C. R. Exci-tonic superfluid phase in double bilayer graphene. Nat. Phys. 13,751–755 (2017).13. Wang, Z. et al. Evidence of high-temperature exciton condensationin two-dimensional atomic double layers.Nature574, 76–80 (2019).14. Ma, L. et al. Strongly correlated excitonic insulator in atomic doublelayers. Nature 598, 585–589 (2021).15. Rickhaus, P. et al. Correlated electron-hole state in twisted double-bilayer graphene. Science 373, 1257–1260 (2021).16. Gu, J. et al. Dipolar excitonic insulator in a moiré lattice. Nat. Phys.18, 395–400 (2022).17. Chen, D. et al. Excitonic insulator in a heterojunction moiré super-lattice. Nat. Phys. 18, 1171–1176 (2022).18. Zhang, Z. et al. Correlated interlayer exciton insulator in hetero-structures of monolayer WSe2 and moiré WS2/WSe2. Nat. Phys. 18,1214–1220 (2022).19. Cercellier, H. et al. Evidence for an excitonic insulator phase in1T–TiSe2. Phys. Rev. Lett. 99, 146403 (2007).20. Kogar, A. et al. Signatures of exciton condensation in a transitionmetal dichalcogenide. Science 358, 1314–1317 (2017).21. Jia, Y. et al. Evidence for a monolayer excitonic insulator. Nat. Phys.18, 87–93 (2022).22. Sun, B. et al. Evidence for equilibrium exciton condensation inmonolayer WTe2. Nat. Phys. 18, 94–99 (2022).23. Gao, Q. et al. Evidence of high-temperature exciton con-densation in a two-dimensional semimetal. Nat. Commun. 14,994 (2023).24. Song, Y. et al. Signatures of the exciton gas phase and itscondensation in monolayer 1T–ZrTe2. Nat. Commun. 14, 1116(2023).25. Gao, Q. et al. Observation of possible excitonic charge densitywaves andmetal-insulator transitions in atomically thin semimetals.Nat. Phys. 20, 597–602 (2024).26. Zhang, P. et al. Spontaneous gap opening and potential excitonicstates in an ideal Dirac semimetal Ta2Pd3Te5. Phys. Rev. X 14,011047 (2024).27. Huang, J. et al. Evidence for an excitonic insulator state inTa2Pd3Te5. Phys. Rev. X 14, 011046 (2024).28. vanWezel, J., Nahai-Williamson, P. & Saxena, S. S. Exciton-phonon-driven charge density wave in TiSe2. Phys. Rev. B 81, 165109 (2010).29. Bianco, R., Calandra, M. & Mauri, F. Electronic and vibrationalproperties of TiSe2 in the charge-density-wave phase from firstprinciples. Phys. Rev. B 92, 094107 (2015).30. Wakisaka, Y. et al. Excitonic insulator state in Ta2NiSe5 probed byphotoemission spectroscopy. Phys. Rev. Lett. 103, 026402 (2009).31. Seki, K. et al. Excitonic Bose-Einstein condensation in Ta2NiSe5above room temperature. Phys. Rev. B 90, 155116 (2014).32. Lu, Y. F. et al. Zero-gap semiconductor to excitonic insulator tran-sition in Ta2NiSe5. Nat. Commun. 8, 14408 (2017).33. Larkin, T. I. et al. Giant exciton Fano resonance in quasi-one-dimensional Ta2NiSe5. Phys. Rev. B 95, 195144 (2017).34. Mor, S. et al. Ultrafast electronic band gap control in an excitonicinsulator. Phys. Rev. Lett. 119, 086401 (2017).35. Okazaki, K. et al. Photo-induced semimetallic states realized inelectron-hole-coupled insulators. Nat. Commun. 9, 4322 (2018).36. Bretscher, H. M. et al. Ultrafast melting and recovery of collectiveorder in the excitonic insulator Ta2NiSe5. Nat. Commun. 12,1699 (2021).37. Katsumi, K. et al. Disentangling lattice and electronic instabilities inthe excitonic insulator candidate Ta2Ni(Se,S)5 by nonequilibriumspectroscopy. Phys. Rev. Lett. 130, 106904 (2023).38. Nakano, A. et al. Antiferroelectric distortion with anomalous pho-non softening in the excitonic insulator Ta2NiSe5. Phys. Rev. B 98,045139 (2018).39. Subedi, A. Orthorhombic-to-monoclinic transition in Ta2NiSe5 dueto a zone-center optical phonon instability. Phys. Rev. Mater. 4,083601 (2020).40. Jog, H., Harnagea, L., Mele, E. J. & Agarwal, R. Exchangecoupling–mediated broken symmetries in Ta2NiSe5 revealed fromquadrupolar circular photogalvanic effect. Sci. Adv. 8, eabl9020(2022).41. Lu, H. et al. Evolution of the electronic structure in Ta2NiSe5 acrossthe structural transition revealed by resonant inelastic x-ray scat-tering. Phys. Rev. B 103, 235159 (2021).42. Baldini, E. et al. The spontaneous symmetry breaking in Ta2NiSe5 isstructural in nature.Proc.Natl. Acad. Sci. USA 120, e2221688120 (2023).43. Chen, C. et al. Role of electron-phonon coupling in excitonic insu-lator candidate Ta2NiSe5. Phys. Rev. Res. 5, 043089 (2023).44. Chen, C. et al. Anomalous excitonic phase diagram in band-gap-tuned Ta2Ni(Se,S)5. Nat. Commun. 14, 7512 (2023).45. Kaneko, T., Toriyama, T., Konishi, T. & Ohta, Y. Orthorhombic-to-monoclinic phase transition of Ta2NiSe5 induced by the Bose-Einstein condensation of excitons. Phys. Rev. B 87, 035121 (2013).46. Mazza, G. et al. Nature of symmetry breaking at the excitonicinsulator transition: Ta2NiSe5. Phys. Rev. Lett. 124, 197601 (2020).47. Windgätter, L. et al. Common microscopic origin of the phasetransitions in Ta2NiS5 and the excitonic insulator candidateTa2NiSe5. npj Comput. Mater. 7, 210 (2021).48. Tang, T. et al. Non-Coulomb strong electron-hole binding inTa2NiSe5 revealed by time- and angle-resolved photoemissionspectroscopy. Phys. Rev. B 101, 235148 (2020).49. Sunshine, S. A. & Ibers, J. A. Structure andphysical properties of thenew layered ternary chalcogenides tantalum nickel sulfide Ta2NiS5and tantalum nickel selenide Ta2NiSe5. Inorg. Chem. 24,3611–3614 (1985).50. Di Salvo, F. J. et al. Physical and structural properties of the newlayered compounds Ta2NiS5 and Ta2NiSe5. J. Less Common Met.116, 51–61 (1986).Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 7www.nature.com/naturecommunications51. Kim, K. et al. Direct observation of excitonic instability in Ta2NiSe5.Nat. Commun. 12, 1969 (2021).52. Volkov, P. A. et al. Critical charge fluctuations and emergentcoherence in a strongly correlated excitonic insulator. npj Quant.Mater. 6, 52 (2021).53. Yan, J. et al. Strong electron-phonon coupling in the excitonicinsulator Ta2NiSe5. Inorg. Chem. 58, 9036–9042 (2019).54. Gallais, Y. et al. Observation of incipient charge nematicity in Ba(Fe1-xCox)2As2. Phys. Rev. Lett. 111, 267001 (2013).55. Sandilands, L. J., Tian, Y., Plumb, K. W., Kim, Y.-J. & Burch, K. S.Scattering continuum and possible fractionalized excitations in α-RuCl3. Phys. Rev. Lett. 114, 147201 (2015).56. Glamazda, A., Lemmens, P., Do, S. H., Choi, Y. S. &Choi, K. Y. Ramanspectroscopic signature of fractionalized excitations in theharmonic-honeycomb iridates β- and γ-Li2IrO3. Nat. Commun. 7,12286 (2016).57. Kim, K. et al. Suppression of magnetic ordering in XXZ-type anti-ferromagnetic monolayer NiPS3. Nat. Commun. 10, 345 (2019).58. Kim, S. Y. et al. Layer-confined excitonic insulating phase in ultra-thin Ta2NiSe5 crystals. ACS Nano 10, 8888–8894 (2016).59. Lee, J. et al. Strong interband interaction in the excitonic insulatorphase of Ta2NiSe5. Phys. Rev. B 99, 075408 (2019).60. Liu, X. et al. Tuning electron correlation in magic-angle twistedbilayer graphene using Coulomb screening. Science 371,1261–1265 (2021).61. Sugimoto, K., Nishimoto, S., Kaneko, T. & Ohta, Y. Strong couplingnature of the excitonic insulator state in Ta2NiSe5. Phys. Rev. Lett.120, 247602 (2018).62. Fukutani, K. et al. Detecting photoelectrons from spontaneouslyformed excitons. Nat. Phys. 17, 1024–1030 (2021).63. Watson, M. D. et al. Band hybridization at the semimetal-semiconductor transition of Ta2NiSe5 enabled by mirror-symmetrybreaking. Phys. Rev. Res. 2, 013236 (2020).64. Kim, M.-J. et al. Phononic soft mode behavior and a strong elec-tronic background across the structural phase transition in theexcitonic insulator Ta2NiSe5. Phys. Rev. Res. 2, 042039 (2020).65. Fukutani, K. et al. Electrical tuning of the excitonic insulator groundstate of Ta2NiSe5. Phys. Rev. Lett. 123, 206401 (2019).66. Chen, L. et al. Doping-controlled transition from excitonic insulatorto semimetal in Ta2NiSe5. Phys. Rev. B 102, 161116 (2020).AcknowledgmentsWe thank Cheng Chen for helpful discussions. X.X. acknowledgessupport from the National Key Research and Development Program ofChina (Grant No. 2024YFA1409100), the Natural Science Foundation ofJiangsu Province (Grant Nos. BK20231529 and BK20233001), the Fun-damental Research Funds for the Central Universities (Grant No. 0204-14380233), and theNational Natural Science Foundation of China (GrantNos. 12474170). Y.G. acknowledges support from the National KeyResearch and Development Program of China (Grant No.2024YFA1408400). K.W. and T.T. acknowledge support from the JSPSKAKENHI (Grant Nos. 20H00354 and 23H02052) and World PremierInternational Research Center Initiative (WPI), MEXT, Japan.Author contributionsX.X. conceived the project. K.W. (NJU) performed the experiments. Y.L.and Y.G. grew the Ta2NiSe5 crystals. K.W. (NIMS) and T.T. grew the hBNcrystals. K.W. (NJU) and X.X. analysed the experimental data and inter-preted the results. X.X. wrote the paper, with comments from allauthors.Competing interestsThe authors declare no competing interests.Additional informationSupplementary information The online version containssupplementary material available athttps://doi.org/10.1038/s41467-025-66594-y.Correspondence and requests for materials should be addressed toYanfeng Guo or Xiaoxiang Xi.Peer review information Nature Communications thanks the anon-ymous reviewers for their contribution to the peer review of this work. Apeer review file is available.Reprints and permissions information is available athttp://www.nature.com/reprintsPublisher’s note Springer Nature remains neutral with regard to jur-isdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative CommonsAttribution-NonCommercial-NoDerivatives 4.0 International License,which permits any non-commercial use, sharing, distribution andreproduction in any medium or format, as long as you give appropriatecredit to the original author(s) and the source, provide a link to theCreative Commons licence, and indicate if you modified the licensedmaterial. Youdonot havepermissionunder this licence toshare adaptedmaterial derived from this article or parts of it. The images or other thirdparty material in this article are included in the article’s CreativeCommons licence, unless indicated otherwise in a credit line to thematerial. If material is not included in the article’s Creative Commonslicence and your intended use is not permitted by statutory regulation orexceeds the permitted use, you will need to obtain permission directlyfrom the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.© The Author(s) 2025Article https://doi.org/10.1038/s41467-025-66594-yNature Communications |        (2025) 16:10999 8https://doi.org/10.1038/s41467-025-66594-yhttp://www.nature.com/reprintshttp://creativecommons.org/licenses/by-nc-nd/4.0/http://creativecommons.org/licenses/by-nc-nd/4.0/www.nature.com/naturecommunications Gate tuning of coupled electronic and structural phase transition in atomically thin Ta2NiSe5 Results Coupled electronic and phononic excitations Thickness dependence Effect of doping on electrical transport Effect of doping on Raman scattering Discussion Methods Sample and device preparation Characterizations Data availability References Acknowledgments Author contributions Competing interests Additional information