# Fileset

[s41467-022-32808-w.pdf](https://mdr.nims.go.jp/filesets/e01658ab-c63a-455b-8df8-374f63a5d5bc/download)

## Creator

Junxue Li, Mina Rashetnia, Mark Lohmann, Jahyun Koo, Youming Xu, Xiao Zhang, [Kenji Watanabe](https://orcid.org/0000-0003-3701-8119), [Takashi Taniguchi](https://orcid.org/0000-0002-1467-3105), Shuang Jia, Xi Chen, Binghai Yan, Yong-Tao Cui, Jing Shi

## Rights

[Creative Commons BY Attribution 4.0 International](https://creativecommons.org/licenses/by/4.0/)

## Other metadata

[Proximity-magnetized quantum spin Hall insulator: monolayer 1 T’ WTe2/Cr2Ge2Te6](https://mdr.nims.go.jp/datasets/434a4820-0bb7-488c-8943-4c6758c72c23)

## Fulltext

Proximity-magnetized quantum spin Hall insulator: monolayer 1 T’ WTe2/Cr2Ge2Te6nature communicationsArticle https://doi.org/10.1038/s41467-022-32808-wProximity-magnetized quantum spin Hallinsulator: monolayer 1 T’ WTe2/Cr2Ge2Te6Junxue Li1,2,8, Mina Rashetnia1,8, Mark Lohmann 1, Jahyun Koo3, Youming Xu4,Xiao Zhang5, Kenji Watanabe 6, Takashi Taniguchi 7, Shuang Jia 5, Xi Chen4,Binghai Yan 3, Yong-Tao Cui 1 & Jing Shi 1Van der Waals heterostructures offer great versatility to tailor unique inter-actions at the atomically flat interfaces between dissimilar layered materialsand induce novel physical phenomena. By bringing monolayer 1 T’ WTe2, atwo-dimensional quantum spin Hall insulator, and few-layer Cr2Ge2Te6, aninsulating ferromagnet, into close proximity in an heterostructure, we intro-duce a ferromagnetic order in the former via the interfacial exchange inter-action. The ferromagnetism inWTe2manifests in the anomalous Nernst effect,anomalous Hall effect as well as anisotropic magnetoresistance effect. Usinglocal electrodes, we identify separate transport contributions from themetallic edge and insulating bulk. When driven by an AC current, the secondharmonic voltage responses closely resemble the anomalousNernst responsesto AC temperature gradient generated by nonlocal heater, which appear asnonreciprocal signals with respect to the induced magnetization orientation.Our results from different electrodes reveal spin-polarized edge states in themagnetized quantum spin Hall insulator.Conventional ferromagnets are known to exhibit a series ofmagnetization-dependent transverse transport phenomena as linearresponses to electric field or temperature gradient, many of which areintimately interrelated. Inmetals and semiconductors for example, theanomalous Hall effect (AHE) and anomalous Nernst effect (ANE) areconnected by the Mott relation1–4 through the anomalous Hall con-ductivity’s energy derivative, and their respective unquantized coeffi-cients are quantitatively determinedbyeither intrinsic and/or extrinsicmechanisms. In contrast towell-studied ferromagnetic systems, little isknown about these properties in two-dimensional topological systemswith edge states. In quantum anomalous Hall insulators (QAHI), forexample, the anomalous Hall conductance is quantized to e2h due to theone-dimensional (1D) ballistic chiral edge transport, the hallmark ofthe quantumanomalousHall effect5,6. According to theMott relation, a1D ballistic chiral edge is not expected to generate any ANE. Experi-mentally it is challenging to investigate the ANE in QAHIs well below1 K. In ideal quantum spin Hall insulators (QSHI)7–14, on the other hand,the two counter-propagating helical edge currents produce neitherHall nor Nernst signal due to time reversal symmetry (TRS), althougheach edge channel has the same but opposite quantized Hall con-ductance. To understand the edge current transport between the fullyspin-polarized in QAHIs and spin unpolarized in QSHIs, here weintroduce a ferromagnetic order in a QSHI to study the behaviors ofpartially spin-polarized edge current transport.Among existing 2D QSHIs including HgTe/CdTe10,11 and InAs/GaSb14 quantum wells, 1 T’ monolayer (ML)-WTe2 has recently attrac-ted much attention for its well-defined bulk and gapless edge bandstructure with a large QSHI gap15–17, spatially resolved metallic edgeReceived: 11 March 2022Accepted: 18 August 2022Check for updates1Department of Physics and Astronomy, University of California, Riverside, CA 92521, USA. 2Department of Physics, Southern University of Science andTechnology, Shenzhen 518055, China. 3Department of Condensed Matter Physics, Weizmann Institute of Science, Rehovot, Israel. 4Department of Electricaland Computer Engineering, University of California, Riverside, CA 92521, USA. 5International Center for Quantum Materials, School of Physics, PekingUniversity, Beijing 100871, China. 6Research Center for Functional Materials, National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan.7International Center for Materials Nanoarchitectonics, National Institute for Materials Science, 1-1 Namiki, Tsukuba 305-0044, Japan. 8These authors con-tributed equally: Junxue Li, Mina Rashetnia. e-mail: jing.shi@ucr.eduNature Communications |         (2022) 13:5134 11234567890():,;1234567890():,;http://orcid.org/0000-0001-7186-2044http://orcid.org/0000-0001-7186-2044http://orcid.org/0000-0001-7186-2044http://orcid.org/0000-0001-7186-2044http://orcid.org/0000-0001-7186-2044http://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/0000-0001-8705-6055http://orcid.org/0000-0001-8705-6055http://orcid.org/0000-0001-8705-6055http://orcid.org/0000-0001-8705-6055http://orcid.org/0000-0001-8705-6055http://orcid.org/0000-0003-2164-5839http://orcid.org/0000-0003-2164-5839http://orcid.org/0000-0003-2164-5839http://orcid.org/0000-0003-2164-5839http://orcid.org/0000-0003-2164-5839http://orcid.org/0000-0002-8015-1049http://orcid.org/0000-0002-8015-1049http://orcid.org/0000-0002-8015-1049http://orcid.org/0000-0002-8015-1049http://orcid.org/0000-0002-8015-1049http://orcid.org/0000-0002-9395-8482http://orcid.org/0000-0002-9395-8482http://orcid.org/0000-0002-9395-8482http://orcid.org/0000-0002-9395-8482http://orcid.org/0000-0002-9395-8482http://crossmark.crossref.org/dialog/?doi=10.1038/s41467-022-32808-w&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-022-32808-w&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-022-32808-w&domain=pdfhttp://crossmark.crossref.org/dialog/?doi=10.1038/s41467-022-32808-w&domain=pdfmailto:jing.shi@ucr.edustates18, possible exciton insulating state19,20, and close-to-quantizedconductance21,22. In general, ferromagnetic order can be introduced viadoping or interfacial proximity coupling in heterostructures contain-ing a ferromagnet. The latter approach has been demonstrated inseveral material systems including graphene23 and topologicalinsulators24,25. For ML-WTe2, the heterostructure approach has a par-ticular appeal because van der Waals (vdW) heterostructures such asCrI3/WSe226 and CrI3/WTe227 have proven very effective in creatingproximity coupling due to the atomically flat interfaces. In CrI3/WTe227,edge current transport revealed fascinating nonlinear and non-reciprocal characteristics which were attributed to electron-magnoninteraction between themetallic edge inWTe2 and themagnetic CrI3. Itgave rise to interestingquestions suchas themagnetic state of thebulkand the nature of the nonreciprocity.In this work, we fabricate high-quality vdW heterostructurescomprised of ML-WTe2 and few-layer vdW ferromagnet Cr2Ge2Te6(CGT) and probe edge and bulk transport responses to both AC tem-perature gradient and electric field. ANE and AHE as well as the ani-sotropic magnetoresistance (AMR) unequivocally affirm theproximity-induced ferromagnetism in the entire atomic layer of ML-WTe2. At low temperatures, the two-component transport, i.e., theedge and bulk, can be cleanly disentangled. The unquantized AHE anddefinitive edge ANE responses are indicative of partially spin-polarizededge channel, distinguishing itself from the ideal 1D chiral or 1D helicaledge channels.Results and discussionWe use a glove-box transfer/pickup technique to fabricate our het-erostructure devices which is described in the Methods section andSupplementary Section 1. Figure 1a shows a schematic illustration ofthe ML-WTe2/CGT vdW heterostructure. CGT is an insulatingferromagnet (resistance well above ∼10GΩ in thin flakes) below itsCurie temperature Tc of 61 K with the magnetic anisotropy perpendi-cular to its atomic layers28–31. We expect a strong exchange interactionbetween ML-WTe2 and CGT at the atomically flat interface, and con-sequently, for the former to acquire ferromagnetism via proximitycoupling. We electrically probe the induced ferromagnetism in ML-WTe2 by measuring the ANE, AMR and AHE responses. Since thetransfermethod is a low temperature process, it should not cause CGTto become conductive31. We also exclude the formation of a con-ductive surface layer of CGTbypossible charge transfer fromML-WTe2because its resistance is found to increase after it is put in contact withCGT as will be discussed later. Due to the larger than four orders ofmagnitude difference in resistance between ML-WTe2 and CGT, thetransport responses should be solely from ML-WTe2. The ANE devicestructure is shown in Fig. 1b. Figure 1c is the optical image of the device(D1) prior to transfer of ML-WTe2/CGT composite layer by the pickup/transfer technique.To measure ANE responses, we fabricate a heater for generatingmainly a lateral temperature gradient ∇T perpendicular to both theheater and the voltage channel. While passing an AC current throughthe heater (via 1–2 electrode pair in Fig. 1c), we record the secondharmonic voltage responses from electrode pairs 3–8, 4–7 and 4–5 (asshown in Fig. 1c) as anout-of-planemagneticfieldHz is swept. Figure 1dsummarizes the Hz-dependence of the voltage signals from the threeelectrode pairs at the nominal system temperature of 4.0 K. Hysteresisbehaviors are observed in all three channels which resemble theanomalousHall loop in CGT/Pt31, except that here the slanted loops arenearly closed. We note that the shape of the hysteresis loop dependson CGT flake thickness and it can be completely collapsed above acertain thickness due to stronger dipolar interaction (see Fig. S-2 ofref. 31). The linear Hz-dependent background is often caused by theFig. 1 | Device structure and anomalous Nernst signals in monolayer 1 T’WTe2/Cr2Ge2Te6 heterostructure. a Schematic of monolayer (ML) 1 T’ WTe2/Cr2Ge2-Te6(CGT) vdW heterostructure. b Schematic of the ANE device structure. BNstands for hexagonal boronnitride. Thenonlocal heater for generating in-planetemperature gradient ∇Tip is underneath the bottom BN. Open circuit voltageVANE due to ANE is measured in sweeping out-of-plane magnetic field Hz.c Device optical image before transfer of ML-WTe2/CGT composite layer onpre-patterned Pt electrodes. The scale bar is 20 μm. The polygons with yellow,green and white dashed boundaries indicate the profiles of ML-WTe2, CGT andtop-BN, respectively. d Magnetic field dependence of 20-loop averaged ANEsignal fromelectrodepairs 3–8, 4–7 and4–5. The heating current in channel 1–2is 12 mA and the system temperature is set to 4.0 K. e 20-loop averaged ANEsignal from channel 3–8 at different sample temperatures ranging from 14.9 Kto 67.4 K. The vertical axis is the ANE voltage normalized by the heating powerP. The actual sample temperature Ts is indicated in each panel which is cali-brated with the temperature dependence of ML-WTe2’s four-terminal resis-tance measured with a small current. f Heating-power-normalized ANE signalsfrom channels 3–8, 4–7 and 4–5 vs. Ts. Inset shows the data between 40 K and70 K, where the ANE signals in all three channels vanish around the Curietemperature Tc = 61 K of CGT.Article https://doi.org/10.1038/s41467-022-32808-wNature Communications |         (2022) 13:5134 2ordinary Nernst effect frommagnetic field but here it is much smallerthan the magnitude of the total signal at saturation; therefore, thehysteresis loops cannot be produced by the stray field from CGT withsaturationmagnetizationMs (<4πMs ∼2 kOe), which strongly indicatestheir acquired origin fromCGT, i.e., the anomalous Nernst effect (ANE)arising from the spin-orbit coupling (SOC). The same argument holdsfor the SOC origin of the AHE in ferromagnets3 simply because theordinary Hall effect cannot account for the largemagnitude of the Hallhysteresis loops if it was from a stray field-generated ordinary Hallsignal. Additionally, in Supplementary Section 2, we present a discus-sion to exclude the spinSeebeckeffect (SSE), the competing effect thatcan arise from unmagnetized WTe2. The ANE voltage, i.e., VANE ∼L ∇T ×Mz ẑ� �x , here L being the channel length for each pair of elec-trodes pair that are separated along the x-direction, andMz the out-of-plane component of the induced magnetization in ML-WTe2. Theinduced Mz should follow that of the CGT surface layer, the source ofthe induced ferromagnetism. The slanted VANE loops with nearlyvanishing remanence at Hz = 0 are results of multidomain formationwhich produces stochastic responses in each single field sweep. Thedomain formation was previously studied in the anomalous Hall andmagnetic force microscopy work of CGT thin flakes31. Linear heatingpower dependence of VANE magnitude is consistent with the expectedlinear ∇T -dependence (see Supplementary Sections 3 and 4).To further investigate the correlation between themagnetic orderin CGT and the induced ferromagnetism in ML-WTe2, we carry outtemperature dependence measurements of VANE . Figure 1e (and moredata in other channels in Supplementary Section 5) shows the loops ofnormalized ANE voltage by power P, VANE/P, at selected temperatures.Because of the slanted loop shape, we take the saturation values ofVANE/P on both positive and negative field and plot the half differencebetween them in Fig. 1f for the three pairs. They all decrease as thedevice warms up before disappearing as the sample temperature Tsapproaches 61 K, the Tc of CGT. Ts is obtained by using the calibratedWTe2 sample resistance as a sensitive thermometer. It is clear that theANE signal is closely related to the ferromagnetic order para-meter of CGT.After experimentally establishing the induced ferromagnetism inML-WTe2, we turn to magneto-transport properties of the magnetizedML-WTe2. By passing an AC current inML-WTe2 with the frequency of f(13 Hz) and the root-mean-square (rms) amplitude of Irms (3 μA), wesimultaneously measure the first- and second-harmonic longitudinalvoltage responses, i.e., 1f and 2f voltages (measurement geometrysketched in Fig. 2a). Figure 2b plots theHz-dependence of the 1f and 2 fvoltages from the 4–7 channel measured at 4 K. The raw 1f signal (toppanel in Fig. 2b) clearly contains both Hz-symmetric (V 1f�S4�7 ) and Hz-antisymmetric (V 1f�AS4�7 ) components with comparable magnitude. Thelarge Hz-antisymmetric signal mixed in the longitudinal channel iscaused by the irregular WTe2 shape, which is not etched into a regularHall bar to avoid possible damages. After symmetrization and anti-symmetrization of the longitudinal voltage, we obtainV 1f�S4�7 and V 1f�AS4�7hysteresis loops that are characteristic of the AMR and AHE responsesof ferromagnetic conductors, respectively (as shown in middle panelof Fig. 2b). As discussed in Supplementary Section 6, we exclude thespin Hall magnetoresistance mechanism that could arise from strongSOC inML-WTe2. Therefore, themagnetoresistance loops are possibleonly when the ML-WTe2 layer is proximity-magnetized. Additionally,the 1f Hz-antisymmetric or the AHE signal in the longitudinal resistancechannel provides another proof of the induced ferromagnetism inML-WTe2. In the Hall channel (between electrodes 7 and 9), we observesimilar AHE signal mixed with a Hz-symmetric AMR signal but noobservable quantized Hall signal is present. Similar proximity-inducedAHE has also been observed in other quantum materials includinggraphene23 and topological insulators25,32.Besides these 1 f hysteresis loops, there is also a large Hz-anti-symmetric hysteresis loop in the 2 f response (Fig. 2b-bottom), whichhas the same but inverted shape as the 1 f AHE loop. In YIG/Pt het-erostructures, passing a large AC current in Pt can indeed produceboth 1f and 2f responses, but normally the former dominates;33 andthe 2 f signal is attributed to SSE due to the Joule heating generated bythe AC current in Pt. Similar to the nonlocal heating case, here the SSEmechanism by the sample self-heating induced an 2 f out-of-plane ∇Tcomponent can also be excluded with the same reasons (see discus-sion in Supplementary Section 2). However, the sample self-heatinggenerates an AC temperature rise in the ML-WTe2 flake with the fre-quency of 2f . The net outward heat dissipation into the surroundingmedium results in a 2 f in-plane ∇T component around the edge of theflake (see ourCOMSOL simulation results in Supplementary Section 4).This lateral ∇T component at the lower boundary is orthogonal to thelongitudinal voltage channel as shown in Fig. 2a, and thus produces anANE-like 2 f response that is sensitive toMz. In fact, the 2f signal has thesame polarity as that of the ANE signal produced by the nonlocalheater in Fig. 1d with the same in-plane ∇T direction, which is con-sistent with the sample self-heating scenario. Hence, we attribute the2f hysteresis to the ANE response to the sample Joule heating. Becauseof the hysteretic characteristic of the AHE responses, the 2f voltage isclearly unidirectional or nonreciprocal, depending on the direction ofthe magnetization. The nonreciprocity in 2 f voltage may share thesame origin with the nonreciprocal transport phenomena reported inother systems 27,34–36.As both 1f and 2f signals are all consistent with the induced fer-romagnetism in ML-WTe2, they should vanish when CGT turns para-magnetic. Figure 2d–f show the respective Hz-dependence of AMR,AHE and ANE signals at selected temperatures. Indeed, they all dis-appear at and above the Tc of CGT (as summarized in Fig. 2c).Even though the lower electrodes in device D1 are placed close tothe sample edge (as shown in Figs. 1c and 2a), they also contact thesample interior due to the transfer resolution limit; therefore, themeasured signals contain contributions from both edge and bulkcomponents. The pristineML-WTe2 is known to be aQSHIwith gaplessedge states; we also expect the magnetized ML-WTe2 to contain bothbulk and edge states. To understand their respective contributions, wefabricate another type of devices represented by D7 as illustrated inFig. 3a, in which one set of multiple electrodes (from #3 to #6 inFig. 3b)make electrical contact only with the bulk, called the Bulk-onlyelectrodes, but the other set (from #11 to #14 in Fig. 3b) with both thebulk and edge, called the Bulk + Edge electrodes. The Bulk-only elec-trodes are electrically insulated from the sample edge using an addi-tional thin layer of BN (∼35 nm thick) to cover the pre-patterned Ptelectrodes except for their far ends before transfer ofML-WTe2 so thatthe uncovered ends of the Pt electrodes can only probe the bulkchannel of ML-WTe2. With these two sets of electrodes, we first mea-sure the two-terminal resistance as a function of temperature for theBulk-only and the Bulk + Edge electrodes, i.e., R5-6 and R13-14, respec-tively. Figure 3c reveals a striking contrast between these two sets.First, R5–6 (Bulk-only) is always greater than R13–14 (Bulk + Edge) overthe entire temperature range, indicating that the edge is more con-ductive than bulk. Second, while the difference between them remainsrelatively small at high temperatures, the Bulk-only resistanceR5-6 risessteeply below 10K while the Bulk + Edge resistance R13-14 only shows amoderate increase. This is a sign of bulk carrier freezing indicative of abulk gap. From the temperature dependence, we determine activationenergy of 3.16meV (Supplementary Section 7). This value is compar-able to that found in CrI3/WTe2 (2.5meV)27 but much smaller than theQSHI bulk gap in ML-WTe2 grown on a bilayer graphene substrate(∼45meV)17. Compared to pristine QSHI, magnetizedQSHI is expectedto exhibit a smaller gap due to spin splitting. For example, the spinsplitting in conduction bands is about 30meV for CGT/WTe2 accord-ing to our density functional theory calculations (see SupplementarySection 8).We also note that a wide discrepancy in the band gap existsin the literature among various theoretical and experimentalArticle https://doi.org/10.1038/s41467-022-32808-wNature Communications |         (2022) 13:5134 3groups15–17,27. On the other hand, R13-14 is essentially dominated by edgeconduction below 10K where the bulk carriers freeze out. In themeantime, R13-14 approaches 303.7 kΩ at 2 K, which is over two ordersmaller than R5-6 but at least a factor of 10 larger than h/e2 (=25.8 kΩ),the quantized resistance for a single 1D ballistic edge channel. Thislarge edge resistance, which corresponds to an order of magnitudesmaller than the quantized edge conductance, was also previouslyreported in similar heterostructures27 or pristine ML-WTe221 andexplained by an edge state gap or backscattering in the edge channels.The large resistance also excludes the possibility of any chargetransfer-induced conduction in CGT, which would decrease the resis-tance instead. We also perform gate voltage Vg dependence mea-surements of the Bulk + Edge resistance. It generally shows a broad andshallow maximum centered around Vg = 0 at 4 K (SupplementarySection 9), which is likely due to the overlap between the small edgegap and bulk valence band.Because of the small bulk gap, we need to stay below 10K intemperature to access the QSHI edge transport by the Bulk+ Edgeelectrodes. To avoid sample heating and thus achieve lower sampletemperature Ts than in Figs. 1 and 2, we use a much smaller AC current(Irms = 0.5mA) in the nonlocal heater and simultaneously record ANEvoltages from the Bulk +Edge13,14 and Bulk-only5,6 electrodes. Fig-ures 3d, e show the Ts-dependent ANE voltages from the Bulk + Edgeand the Bulk-only electrodes scaled by the heating power P of thenonlocal heater. The following surprising contrast stands out from theside-by-side comparison. First, the edge ANE signal does not vanish atlow temperatures as one would expect for ideal 1D edge channels, butrather increases in its magnitude. Second, the edge ANE changes thesign as Ts decreases, whereas the bulk ANE remains the same sign,highlighting distinctly different characteristics between the edge andbulk channels. To compare with the nonlocal heating case, we alsodeliberately inject smaller AC currents in ML-WTe2 to reduce the self-heating power in order to achieve lower sample temperatures. Inter-estingly, the opposite signs between 4K (Fig. 3f, left) and 14 K (Fig. 3g,left) of the 2 f signals in the Bulk+ Edge electrodes also indicate a signchange between these two temperatures. In contrast, the 2 f signals inthe Bulk-only electrodes remain the same sign (Fig. 3f, g, right). Thecorrespondence between the nonlocal- and self-heating cases furthersupports that the nonreciprocal 2 f responses originate from ANE.With the resistance and ANE data from the two sets of electrodes,we can readily separate the bulk and edge contributions using a simpletwo-component circuit model (as sketched in Fig. 4a, see details inFig. 2 | Temperature dependence of linear and non-linear responses in ML-WTe2/Cr2Ge2Te6 heterostructure. a Schematic illustration of measurement geo-metry. An AC current of 3 μA flows in the ML-WTe2 flake through channel 3–8 (x-direction), and the 1f and 2f voltage responses are measured from channel 4-7 insweeping Hz with f = 13Hz here. The thick arrows with color gradient across thedevice boundary indicate thedirectionof heat dissipation, i.e.,�∇Tip, where∇Tip isthe in-plane component of the temperature gradient.bMagnetic field dependenceof the raw 1f , V 1f4�7 (top), 1f Hz-symmetric, V 1f�S4�7 and Hz-antisymmetric, V 1f�AS4�7(middle), and 2 f V2f4�7voltage (bottom) at 4 K. c Temperature dependence of themagnetoresistance (MR) ratio (top), AHE voltage (middle) and 2f voltage (bottom)taken from the data in d–f below. MR ratio magnitude is defined as [V 1f4�7(5 kOe)-V 1f4�7(0Oe)]/V 1f4�7(0Oe), and 4V 2f4�7 =V2f4�7 Hs� �� V 2f4�7 0ð Þ. Hs is the saturationmagnetic field of V 2f4�7. d–f Magnetic field dependence of MR ratio (d), V 1f�AS4�7 (e)and V2f4�7 (f) voltage at various temperatures. The temperatures denoted here areread from the thermometer of the measurement system and the actual sampletemperature should be higher due to local heating.Article https://doi.org/10.1038/s41467-022-32808-wNature Communications |         (2022) 13:5134 4Fig. 4 | Two-component transport from edge and bulk channels of ML-WTe2.a Illustration of the “parallel battery-resistor” model with ANE voltages from Edge(VANEE ) and Bulk (VANEB ) channels. GE(GB) is the conductance of the Edge (Bulk)channel. VANEB+ E is the total ANE signal from both edge and bulk. b Temperaturedependence of the ANE signal fromBulk, Bulk + Edge, and Edge channels. The EdgeANE is calculated using the “parallel battery-resistor”model. Inset shows a zoom-inplot of the high-temperature data. c Ratio of ANE from Edge (Bulk + Edge) channelto that from Bulk-only channel as a function of Ts.Fig. 3 | Anomalous Nernst effect and anomalous Hall effect from edge and bulkchannels of ML-WTe2/Cr2Ge2Te6. a Schematic diagram of device structure.Electrodes on the left side probe combined edge and bulk signal of ML-WTe2,and those on the right side are partly covered with BN to prevent edge contactwith ML-WTe2, thus only detect the bulk ANE. b Optical image of device D7prior to transfer of ML-WTe2/Cr2Ge2Te6 composite layer. The ML-WTe2 flake isindicated by the purple dashed polygon. Electrodes from 11 to14 probe theANE signal from both edge and bulk, and electrodes from 3 to 6 probe thebulk-only ANE signal. The scale bar is 10 μm. c Temperature dependence ofresistance from 5–6 (Bulk) and 13-14 (Bulk + Edge). d, e ANE signals from 13–14(d) and 5–6 (e) at selected temperatures. For Ts < 20 K, 40 loops are used foraveraging; for Ts > 20 K, 20 loops are used for averaging. f, g 2 f currentresponse from Bulk + Edge (11–14 and 11–12) and Bulk-only (5–6 and 5-4)channels to AC voltage applied between the two outermost horizontal elec-trodes on ML-WTe2 vs. Hz at Ts = 4 K (f) and 14 K (g). 60 loops are used foraveraging in f, and 10 loops for other curves than the 11–14 electrodes (3loops) in g, respectively. The rms magnitudes of the AC voltage for 4 K and14 K measurements are 20 and 200mV, respectively.Article https://doi.org/10.1038/s41467-022-32808-wNature Communications |         (2022) 13:5134 5Supplementary Section 10). Figure 4b plots the temperature depen-dence of ANE signals from bulk and edge channels, the latter of whichis calculated using the circuit model. Clearly, both edge and bulk ANEsignals increase in magnitude as Ts approaches zero. This apparentlow-temperature magnitude increase in both channels is possiblycaused by the rapidly decreasing thermal conductivity due to phononfreezing, which greatly enlarges the actual ∇T (Supplementary Sec-tion 11). Since both channels are equally affected by the same thermalconductivity, we plot the ratio of the edge to bulk ANE voltages inFig. 4c. Interestingly, aside from the sign change, the ratio shows aquickdive at low-Ts, whichmaybe causedby the faster decreasingbulkANE due to carrier freezing.After successfully disentangling the two-component transportbehavior, now we discuss implications of these observations. It waspreviously known that in channels longer than 100nm21,22, the smallerthan quantized 1D conductance indicates that the QSHI edge statessuffer from backscattering possibly due to inhomogeneous bulkstates37,38 and thus the 1D Dirac edge transport is diffusive. Here wepropose an edge transport picture (see Fig. S9 in SupplementaryNote 8) to explain the observed edge AHE signal (this mechanism canbe modified to explain the edge ANE signal driven by a temperaturegradient). Under broken time-reversal symmetry, the two counter-propagating helical Dirac edge states in magnetized QSHI do not havethe same conductivity. In fact, the unquantized AHE must arise from anet charge current composed of two unequal counter-propagatingflows of charge carriers, or a spin-polarized current. Under a voltagebias, the two longitudinal edges do not conduct the same amount ofcurrent, which leads to a net transverse or Hall voltage. The sameinequivalent longitudinal edges, if they acquire energy-dependence inthe conductivity due to the hybridization of the Dirac edge states withthe bulk and/or ordinary edge states, also produce a net transversevoltage under a temperature gradient, which is the Nernst voltage.Such hybridization leads to quasi-1D edge states. The polarity of thevoltages depends on the direction of themagnetization, which explainsthe observed 2 f hysteresis. If the temperature varies, the chemicalpotential can intercept both the Dirac and ordinary edge states.Because the asymmetry between the both edge types can be different,it can result in different voltage responses. Although the actual out-come depends on the band structure and scattering details, the com-petition between the two types of edge states can in principle lead to asign change in the transverse voltages. The low-temperature 2 f signchange observed in our experimentmay be a result of this competition.In summary, we have demonstrated proximity-induced ferro-magnetism in ML-WTe2 using the vdW heterostructure approach. Inthis magnetized QSHI, we have unequivocally disentangled the edgeandbulk transport from local transport probes. The nonzero edgeAHEand ANE responses indicate that the edge states of the magnetizedQSHI are partially spin-polarized, qualitatively different from the 1Dballistic chiral edges in QAHIs or helical edges in QSHIs.MethodsDevice fabricationWe fabricate our devices using the following processes (from bottomto top)1. The few-layer graphene (FLG) gate is transferred (with the BNusing Polycarbonates pickup procedure) on a SiO2/Si substrate withprepatterned Cr(5 nm)/Au(30 nm) heater using electron beam litho-graphy (EBL)2. A thin hexagonal boron nitride (BN) layer (∼35 nm) istransferred to electrically isolate the FLG gate and the pre-patternedheater from themonolayer (ML) 1 T’WTe2 to be transferred on top3. Pt-electrodes are patterned on the top of the bottom BN using EBL andlift-off4. For devices likeD1, a suitableML-WTe2 flake is identifiedunderopticalmicroscope (as shown in Supplementary Fig. S1a, b) and pickedupusing a thinflake of CGT, then theML-WTe2/CGTheterostructure asa composite layer is transferred onto the pre-patterned Pt-electrodes(as shown in Supplementary Fig. S1d). The WTe2 bulk crystals in ourexperiments are purchased from 2D Semiconductor and the CGTcrystals are provided by Prof. S. Jia’s group at Peking University5. Fordevice D7 with the edge insulated from the ML-WTe2 interior (bulk)channel, a thin BN layer is transferred to cover the Pt electrodes exceptthe far ends (Supplementary Fig. S1e) prior to the transfer ofML-WTe2/CGT composite layer on to the pre-patterned Pt electrodes6. The ML-WTe2/CGT heterostructure device including Pt electrodes is encapsu-lated by a top BN layer to prevent degradation and maintain devicestability. The above exfoliation and transfer processes are carried outin a glove box in the atmosphere of argon to avoid device degradation.We have fabricated and studied seven devices in this work and they allshow the same qualitative transport behaviors.Anomalous Nernst effect (ANE) measurements using nonlocalheaterAn AC current with the frequency of 13Hz is fed to the nonlocal heaterusing Keithley 6221 current source, and the second harmonic (2 f) ANEvoltage from different channels of ML-WTe2 is detected using thestandard lock-in technique. A magnetic field is swept perpendicular tothe layers of the devices. All measurements for devices are performedusing either a homemade closed-cycle system or Quantum Design’sDynaCool Physical Property Measurement System (PPMS).Measurements of 1 f and 2 f responses to AC current in ML-WTe2We perform the linear (1 f) and non-linear (2 f) voltage measurementsby feeding an AC current (voltage) into ML-WTe2 of device D1 (D7)using Keithley 6221 current source (lock-in amplifier SR830), and 1 fand 2 f voltage (current) signals are probed using the standard lock-inmethod with the frequency of 13Hz or 11 Hz. All measurements fordevices are performed using either a homemade closed-cycle systemor Quantum Design’s DynaCool PPMS.Reporting summaryFurther information on research design is available in the NatureResearch Reporting Summary linked to this article.Data availabilityThe datasets generated during and/or analysed during the currentstudy are available from the corresponding author on reasonablerequest.References1. Xiao, D., Yao, Y., Fang, Z. & Niu, Q. Berry-phase effect in anomalousthermoelectric transport. Phys. Rev. Lett. 97, 026603 (2006).2. Pu, Y., Chiba, D., Matsukura, F., Ohno, H. & Shi, J. Mott relation foranomalous Hall and Nernst effects in Ga1−xMnxAs ferromagneticsemiconductors. Phys. Rev. Lett. 101, 117208 (2008).3. Nagaosa, N., Sinova, J., Onoda, S., MacDonald, A. H. & Ong, N. P.Anomalous Hall effect. Rev. Mod. Phys. 82, 1539–1592 (2010).4. Mott, N. F. & Jones, H. The Theory of the properties of metals andalloys (Dover, New York, 1958).5. Yu, R. et al. Quantized anomalous Hall effect in magnetic topolo-gical insulators. Science 329, 61–64 (2010).6. Chang, C.-Z. et al. Experimental observation of the quantumanomalous Hall effect in a magnetic topological insulator. Science340, 167–170 (2013).7. Kane, C. L. &Mele, E. J. Quantum spin Hall effect in graphene. Phys.Rev. Lett. 95, 226801 (2005).8. Bernevig, B. A. & Zhang, S. C. Quantum spin Hall effect. Phys. Rev.Lett. 96, 106802 (2006).9. Bernevig, B. A., Hughes, T. L. & Zhang, S. C. Quantum spin halleffect and topological phase transition in HgTe quantum wells.Science 314, 1757–1761 (2006).10. König, M. et al. Quantum spin Hall insulator state in HgTe quantumwells. Science 318, 766–770 (2007).Article https://doi.org/10.1038/s41467-022-32808-wNature Communications |         (2022) 13:5134 611. Roth, A. et al. Nonlocal transport in the quantum spin Hall state.Science 325, 294–297 (2009).12. Hasan, M. & Kane, C. Colloquium: topological insulators. Rev. Mod.Phys. 82, 3045–3067 (2010).13. Qi, X.-L. & Zhang, S.-C. Topological insulators and super-conductors. Rev. Mod. Phys. 83, 1057–1110 (2011).14. Knez, I., Du, R.-R. & Sullivan, G. Evidence for helical edge modes ininverted InAs/GaSb quantum wells. Phys. Rev. Lett. 107,136603 (2011).15. Qian, X., Liu, J., Fu, L. & Li, J. Quantum spin Hall effect in two-dimensional transition metal dichalcogenides. Science 346,1344–1347 (2014).16. Zheng, F. et al. On the quantum spin Hall gap of monolayer 1T’-WTe2. Adv. Mater. 28, 4845–4851 (2016).17. Tang, S. et al. Quantum spin Hall state in monolayer 1T′-WTe2. Nat.Phys. 13, 683 (2017).18. Shi, Y. et al. Imaging quantum spin Hall edges in monolayer WTe2.Sci. Adv. 5, eaat8799 (2019).19. Jia, Y. et al. Evidence for a monolayer excitonic insulator. Nat. Phys.18, 87–93 (2022).20. Sun, B. et al. Evidence for equilibrium exciton condensation inmonolayer WTe2. Nat. Phys. 18, 94–99 (2022).21. Fei, Z. et al. Edge conduction in monolayer WTe2. Nat. Phys. 13,677–682 (2017).22. Wu, S. et al. Observation of the quantum spin Hall effect up to 100kelvin in a monolayer crystal. Science 359, 76–79 (2018).23. Wang, Z., Tang, C., Sachs, R., Barlas, Y. & Shi, J. Proximity-inducedferromagnetism ingraphene revealedby the anomalousHall effect.Phys. Rev. Lett. 114, 016603 (2015).24. Wei, P. et al. Strong interfacial exchange field in the graphene/EuSheterostructure. Nat. Mater. 15, 711–716 (2016).25. Tang, C. et al. Above 400-K robust perpendicular ferromagneticphase in a topological insulator. Sci. Adv. 3, e1700307 (2017).26. Zhong, D. et al. Layer-resolvedmagnetic proximity effect in van derWaals heterostructures. Nat. Nanotechnol. 15, 187–191 (2020).27. Zhao, W. et al. Magnetic proximity and nonreciprocal currentswitching in a monolayer WTe2 helical edge. Nat. Mater. 19,503–507 (2020).28. Zhang, X. et al. Magnetic anisotropy of the single-crystalline ferro-magnetic insulator Cr2Ge2Te6. Jpn J. Appl. Phys. 55, 033001 (2016).29. Gong, C. et al. Discovery of intrinsic ferromagnetism in two-dimensional van der Waals crystals. Nature 546, 265–269 (2017).30. Lin, Z. et al. Pressure-induced spin reorientation transition in layeredferromagnetic insulator Cr2Ge2Te6. Phys. Rev. Mater. 2,051004 (2018).31. Lohmann, M. et al. Probing magnetism in insulating Cr2Ge2Te6 byinduced anomalous Hall effect in Pt. Nano Lett. 19, 2397–2403(2019).32. Yao, X. et al. Record high-proximity-induced anomalous Hall effectin (BixSb1−x)2Te3 thin film Grown on CrGeTe3 substrate. Nano Lett.19, 4567 (2019).33. Schreier, M. et al. Current heating induced spin Seebeck effect.Appl. Phys. Lett. 103, 242404 (2013).34. Avci, C. et al. Unidirectional spin Hall magnetoresistance in ferro-magnet/normal metal bilayers. Nat. Phys. 11, 570–575 (2015).35. Yasuda, K. et al. Large unidirectional magnetoresistance in a mag-netic topological insulator. Phys. Rev. Lett. 117, 127202 (2016).36. Fan, Y. B. et al. Unidirectional magneto-resistance in modulation-doped magnetic topological insulators. Nano Lett. 19,692–698 (2019).37. Nowack, K. C. et al. Imaging currents in HgTe quantumwells in thequantum spin Hall regime. Nat. Mater. 12, 787–791 (2013).38. König, M. et al. Spatially resolved study of backscattering in thequantum spin Hall state. Phys. Rev. X 3, 021003 (2013).AcknowledgementsWe thank Tang Su for his technical assistance at the beginning of theproject, and Ady Stern, Roger Lake and Richard Wilson for helpfuldiscussions. J.X.L. M.L. and J.S. acknowledge the support by DOEaward #DE-FG02-07ER46351 and NSF-ECCS-2051450. M.R. and Y.-T.C.acknowledge the support from the NSF award DMR-2004701. B.Y.Acknowledges the financial support by the European Research Council(ERC Consolidator Grant “NonlinearTopo”, No. 815869). K.W. and T.T.acknowledge support from the Elemental Strategy Initiative conductedby the MEXT, Japan (Grant Number JPMXP0112101001) and JSPSKAKENHI (Grant Numbers 19H05790, 20H00354 and 21H05233).Author contributionsJ.S. conceived the idea, designed the experiments and wrote the paper.J.X.L. andM.R. carriedout themain experimentalwork supervisedby J.S.and Y.T.C., respectively. M.L. participated in the early experimentalwork. J.Y.K. performed band structure calculations supervised by B.H.Y.Y.M.X. performed COMSOL calculations supervised by X.C. X.Z. grewthe CGT crystals supervised by S.J. K.W. and T.T. provided BN crystals.J.X.L., B.H.Y. and Y.T.C. contributed to the data analysis and interpreta-tion and manuscript preparation.Competing interestsThe authors declare no competing interests.Additional informationSupplementary information The online version containssupplementary material available athttps://doi.org/10.1038/s41467-022-32808-w.Correspondence and requests for materials should be addressed toJing Shi.Peer review information Nature Communications thanks the anon-ymous reviewer(s) for their contribution to the peer review of thiswork. Peer reviewer reports are available.Reprints and permission 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 4.0 International License, which permits use, sharing,adaptation, distribution and reproduction in any medium or format, aslong as you give appropriate credit to the original author(s) and thesource, provide a link to the Creative Commons license, and indicate ifchanges were made. The images or other third party material in thisarticle are included in the article’s Creative Commons license, unlessindicated otherwise in a credit line to the material. If material is notincluded in the article’s Creative Commons license and your intendeduse is not permitted by statutory regulation or exceeds the permitteduse, you will need to obtain permission directly from the copyrightholder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.© The Author(s) 2022Article https://doi.org/10.1038/s41467-022-32808-wNature Communications |         (2022) 13:5134 7https://doi.org/10.1038/s41467-022-32808-whttp://www.nature.com/reprintshttp://creativecommons.org/licenses/by/4.0/http://creativecommons.org/licenses/by/4.0/ Proximity-magnetized quantum spin Hall insulator: monolayer 1 T’ WTe2/Cr2Ge2Te6 Results and discussion Methods Device fabrication Anomalous Nernst effect (ANE) measurements using nonlocal heater Measurements of 1 f and 2 f responses to AC current in ML-WTe2 Reporting summary Data availability References Acknowledgements Author contributions Competing interests Additional information