Trilayer WTe₂: Structure, Semimetallicity & More

• Trilayer WTe₂ is a three-monolayer film in the distorted Td phase, defined by a preserved four-mode Raman fingerprint even as it transitions from insulating to semimetallic behavior.
• It exhibits an electronic reconstruction with valence and conduction bands overlapping by approximately 10–20 meV, marking the critical crossover from bilayer gaps to semimetallicity with a Z₂=1 topological signature.
• Its anisotropic crystal structure and self-limiting surface oxidation—confined to a ~0.5 nm layer—provide clear experimental identifiers and practical insights for device fabrication.

Searching arXiv for relevant papers on trilayer WTe₂ to ground the article in the literature. Trilayer WTe$_2$ is a three-monolayer film of tungsten ditelluride in the distorted orthorhombic $T_d$ phase whose properties occupy an intermediate regime between atomically thin insulating layers and bulk semimetallic or Weyl behavior. In this thickness range, the material retains a distinct layer-resolved Raman fingerprint, exhibits ambient degradation that is slower than in monolayer and bilayer flakes because oxidation is confined to a self-limiting surface region, and undergoes an electronic reconstruction in which the indirect gap present in thinner films is suppressed so that valence and conduction states overlap at the Fermi level (Kim et al., 2015, Ye et al., 2016, Sakano et al., 2021, Qiao et al., 18 Apr 2026). Trilayer WTe$_2$ is therefore a focal thickness for studies of dimensional crossover, interlayer-coupling-driven topology, Berry-curvature-related transport, and environmental stability.

1. Crystal structure and layer-specific identity

Trilayer WTe$_2$ consists of three $T_d$-phase WTe$_2$ sheets stacked along the $c$-axis by van-der-Waals forces. Each monolayer is a W plane sandwiched by two Te planes in a distorted octahedral coordination (Ye et al., 2016). In the structural description used for few-layer electronic studies, each WTe$_2$ monolayer in the $T_d$ phase is a distorted $1T'$ unit with local inversion symmetry, and adjacent layers stack with a $T_d$0 ($T_d$1) rotation about the $T_d$2-axis and no net in-plane shift (Sakano et al., 2021). Bulk $T_d$3-WTe$T_d$4 has space group $T_d$5; for $T_d$6, the real structure has point group $T_d$7 with a single $T_d$8 mirror, while the effective in-plane potential gradient alternates in sign layer by layer (Sakano et al., 2021).

The lattice is strongly anisotropic. W atoms displace from ideal octahedral sites to form zigzag chains along the $T_d$9-axis, yielding quasi-1D bonding and weaker coupling along the $_2$0- and $_2$1-axes (Kim et al., 2015). This anisotropy is central to the lattice dynamics of the trilayer because one prominent Raman mode corresponds to vibrations along the tungsten-chain direction and is nearly insensitive to thickness reduction (Kim et al., 2015).

A practical consequence is that trilayer WTe$_2$2 can be distinguished spectroscopically from thinner films. In the Raman study of Kim et al., the four principal first-order Raman peaks observed in bulk WTe$_2$3 are fully conserved down to three layers, whereas bilayer and monolayer progressively lose one or two of these peaks (Kim et al., 2015). This makes trilayer a spectroscopically well-defined thickness even before considering its electronic structure.

2. Raman and lattice-dynamical signatures

Two complementary Raman descriptions appear in the literature. In one, the trilayer vacuum Raman fingerprint at room temperature shows four robust first-order peaks labeled $_2$4–$_2$5: $_2$6, $_2$7, $_2$8, and $_2$9, assigned respectively to an in-plane W–Te vibration, an out-of-plane Te–Te/W–Te mixed mode, an in-plane W–Te bending mode, and an out-of-plane breathing mode (Ye et al., 2016). In the layer-resolved Raman work of Kim et al., the four principal first-order peaks for the trilayer are reported as $_2$0 at $_2$1, $_2$2 at $_2$3, $_2$4 at $_2$5, and $_2$6 at $_2$7, with relative intensities of approximately $_2$8, $_2$9, $T_d$0, and $T_d$1 of the bulk values, respectively (Kim et al., 2015). Both descriptions agree that trilayer WTe$T_d$2 retains four clear first-order Raman modes.

The layer dependence is anomalous relative to hexagonal dichalcogenides. As the layer number decreases from bulk to trilayer to monolayer, the $T_d$3, $T_d$4, and $T_d$5 modes blueshift, while the $T_d$6 mode remains almost fixed in frequency (Kim et al., 2015). Specifically, the measured bulk-to-3L-to-2L-to-1L evolution is $T_d$7 for $T_d$8, $T_d$9 for $_2$0, $_2$1 for $_2$2, and $_2$3 for $_2$4 (Kim et al., 2015). The negligible shift of $_2$5 is attributed to its atomic-displacement pattern along the one-dimensional tungsten chains, so it is almost unaffected by interlayer dielectric screening (Kim et al., 2015).

First-principles phonon calculations support this interpretation. Bulk, bilayer, and trilayer phonons were computed using density functional theory as implemented in VASP + PHONON, with LDA for exchange-correlation, PAW pseudopotentials, cutoff energy $_2$6, and a Monkhorst–Pack $_2$7 mesh for dielectric and Raman calculations (Kim et al., 2015). The phonon eigenvalue problem at $_2$8 is written as

$_2$9

and Raman intensities are extracted from derivatives of the dielectric tensor:

$c$0

The calculations reproduce the observed trend that all modes blueshift except $c$1 (Kim et al., 2015).

For identification purposes, trilayer WTe$c$2 is characterized by the simultaneous presence of four peaks near $c$3, $c$4, $c$5, and $c$6, with $c$7 nearly unshifted from bulk and $c$8 and $c$9 reaching maximum integrated intensities at $_2$0 (Kim et al., 2015).

3. Electronic structure and the insulator–semimetal transition

Trilayer WTe$_2$1 is the critical thickness at which few-layer WTe$_2$2 becomes semimetallic. Laser-$_2$3-ARPES on exfoliated, graphene-encapsulated flakes strictly sorted by layer number shows that, along $_2$4–$_2$5, the highest valence band and lowest conduction band in 3-layer WTe$_2$6 approach and overlap at $_2$7 (Sakano et al., 2021). The momentum-dependent overlap or gap is defined as

$_2$8

Experimentally, $_2$9 changes sign as a function of $T_d$0: near $T_d$1 ($T_d$2), the valence-band maximum lies just below $T_d$3 and $T_d$4, whereas at $T_d$5 to $T_d$6, weak ARPES intensity from the conduction-band bottom crosses $T_d$7 and $T_d$8 (Sakano et al., 2021).

The numerical scale of the overlap is small. At $T_d$9, $1T'$0 is $1T'$1 below $1T'$2, with no conduction band observed up to $1T'$3, so $1T'$4 to $1T'$5 (Sakano et al., 2021). At $1T'$6,

$1T'$7

that is, effectively zero or slightly negative overlap of order $1T'$8–$1T'$9, with a maximum semimetallic overlap of $T_d$00–$T_d$01 where both bands are best resolved (Sakano et al., 2021). This establishes trilayer WTe$T_d$02 as a compensated semimetal at the onset of the $T_d$03 layer transition.

ARPES with $T_d$04 and $T_d$05 reports the same qualitative electronic reconstruction. Along the high-symmetry $T_d$06–$T_d$07–$T_d$08 and $T_d$09–$T_d$10–$T_d$11 directions, the top of the valence band crosses $T_d$12 in the trilayer, with no obvious gap, and second-derivative spectra confirm that the indirect gap closes between valence-band maximum and conduction-band minimum (Qiao et al., 18 Apr 2026). Energy-distribution-curve fitting yields no resolvable gap at $T_d$13 or along $T_d$14, in contrast to the $T_d$15 indirect gap in the monolayer and $T_d$16 gap in the bilayer (Qiao et al., 18 Apr 2026).

First-principles calculations reproduce the measured semimetallicity. HSE06 in the $T_d$17-ARPES study confirms the small overlap in $T_d$18 (Sakano et al., 2021), while GGA+$T_d$19 with $T_d$20 and HSE06 in the later thickness-evolution study find that the valence and conduction bands overlap by $T_d$21 near $T_d$22, with a finite density of states $T_d$23, characteristic of a semimetal (Qiao et al., 18 Apr 2026). The valence band is composed primarily of W $T_d$24 and Te $T_d$25 orbitals, and the conduction-band bottom derives from W $T_d$26 states (Qiao et al., 18 Apr 2026).

4. Interlayer coupling, symmetry, and topological characterization

The thickness evolution of WTe$T_d$27 is non-monotonic. Monolayer WTe$T_d$28 is reported as insulating with indirect QSH gap $T_d$29 and $T_d$30, bilayer WTe$T_d$31 as a reduced-gap trivial insulator with $T_d$32, and trilayer WTe$T_d$33 as semimetallic with $T_d$34 in the sense that valence and conduction bands remain separated in $T_d$35-space near the time-reversal invariant momenta even though there is no global full gap (Qiao et al., 18 Apr 2026). In the bulk limit, the system becomes a 3D type-II Weyl semimetal with chiral Weyl nodes of $T_d$36 and Fermi-arc surface states (Qiao et al., 18 Apr 2026).

The mechanism is interlayer-coupling-driven band reconfiguration. The van der Waals interlayer hopping $T_d$37 between W $T_d$38-orbitals in adjacent layers is reported as $T_d$39 (Qiao et al., 18 Apr 2026). As the layer number $T_d$40 increases, quantum-well subbands form, driving the valence band up and the conduction band down; a critical coupling occurs when the $T_d$41-dependent mass term

$T_d$42

changes sign, closing the gap at $T_d$43 (Qiao et al., 18 Apr 2026). In the effective Hamiltonian near $T_d$44,

$T_d$45

with $T_d$46 acting on orbital parity and $T_d$47 on layer pseudospin, the mass term $T_d$48 changes sign when $T_d$49, leading to gap closing at finite $T_d$50 (Qiao et al., 18 Apr 2026).

For the trilayer, the $T_d$51 invariant is defined through parity products at the four time-reversal invariant momenta $T_d$52:

$T_d$53

with

$T_d$54

where $T_d$55 is the parity eigenvalue of the $T_d$56-th occupied band (Qiao et al., 18 Apr 2026). The reported DFT result is $T_d$57, $T_d$58, $T_d$59, and $T_d$60, implying $T_d$61, and hybrid Wannier-center evolution along $T_d$62–$T_d$63 crosses the reference line an odd number of times, confirming the nontrivial invariant (Qiao et al., 18 Apr 2026). Because the trilayer is globally gapless, global Chern numbers are ill-defined, but local Berry curvature can still be computed around avoided crossings to track the approach to a Weyl transition (Qiao et al., 18 Apr 2026).

The even–odd effect in layer number is a recurrent theme. The reported interpretation is that odd $T_d$64 yields net inverted band ordering and $T_d$65, while even $T_d$66 yields $T_d$67, before semimetallicity emerges when $T_d$68 and eventually connects to bulk Weyl physics (Qiao et al., 18 Apr 2026). This places trilayer WTe$T_d$69 at the first odd-layer semimetallic member of the sequence.

5. Spin splitting, Berry curvature, and nonlinear transport context

Trilayer WTe$T_d$70 also occupies a distinct symmetry regime in spin-resolved and Berry-curvature-related phenomenology. In the $T_d$71-ARPES measurements, the spin-resolved splitting at momentum $T_d$72 is defined as

$T_d$73

For $T_d$74, the valence bands show only very small or unresolvable splittings; fits of energy-distribution curves at $T_d$75 give

$T_d$76

typically in the $T_d$77–$T_d$78 range, with most bands $T_d$79 (Sakano et al., 2021). This contrasts with even-$T_d$80 films, where splittings up to $T_d$81 are observed (Sakano et al., 2021).

The structural interpretation is that incomplete cancellation of local dipoles in even-$T_d$82 layers gives large spin–orbit splittings, whereas in odd-$T_d$83 trilayers much of the triangular-Te dipole arrangement partially cancels, leaving only small residual $T_d$84 (Sakano et al., 2021). This is consistent with the recovered partial symmetry of the middle layer in trilayer stacking and with the weaker effective structural asymmetry relative to even-$T_d$85 films (Sakano et al., 2021).

Few-layer WTe$T_d$86 permits nonzero Berry curvature because inversion symmetry is broken. The band Berry curvature is written as

$T_d$87

and the Berry-curvature dipole as

$T_d$88

Although $T_d$89 and $T_d$90 are not directly reported for trilayer in the ARPES work, previous transport measurements on 3-layer WTe$T_d$91 are stated to have revealed nonlinear anomalous Hall currents and ferroelectric switching, both interpreted there as fingerprints of a nonzero $T_d$92 (Sakano et al., 2021). This suggests that trilayer WTe$T_d$93 couples its near-compensated semimetallicity to inversion-breaking Berry-curvature physics, albeit with weaker spin splitting than in even-layer films.

A plausible implication is that trilayer WTe$T_d$94 is experimentally important not because it maximizes a single order parameter, but because it combines semimetallic overlap, odd-layer topological character, and finite inversion-breaking responses within the same thickness window (Sakano et al., 2021, Qiao et al., 18 Apr 2026).

6. Environmental instability, oxidation chemistry, and passivation

Trilayer WTe$T_d$95 is substantially more stable in air than monolayer or bilayer WTe$T_d$96, but it still degrades by oxidation. Raman monitoring after exfoliation shows that 3L WTe$T_d$97 is stable in vacuum, whereas in ambient air each Raman mode decays approximately exponentially toward a nonzero plateau (Ye et al., 2016):

$T_d$98

where $T_d$99 is elapsed time since exfoliation, $_2$00–$_2$01 accounts for measurement delay, $_2$02 is the Raman intensity that decays, $_2$03 is the residual saturation intensity, and $_2$04 is the characteristic decay time (Ye et al., 2016). For trilayer WTe$_2$05, $_2$06–$_2$07 ($_2$08–$_2$09 days), and residual intensities $_2$10–$_2$11 remain after two weeks (Ye et al., 2016).

This decay is much slower than in thinner flakes. Monolayer WTe$_2$12 shows complete disappearance of Raman modes within $_2$13 in air, bilayer WTe$_2$14 has $_2$15–$_2$16 with saturation after $_2$17, and trilayer requires $_2$18 weeks for full saturation (Ye et al., 2016). The thickness dependence is explained by the fact that only the topmost layer oxidizes and the oxidized surface constitutes a smaller fraction of the total volume as the layer number increases (Ye et al., 2016).

Surface analysis identifies the chemistry. In 3L WTe$_2$19 aged for $_2$20 days, XPS before etching shows new Te $_2$21 oxide components at $_2$22 and $_2$23, assigned to TeO$_2$24, and W $_2$25 oxide components at $_2$26 and $_2$27, assigned to WO$_2$28 with $_2$29 (Ye et al., 2016). After $_2$30 of Ar$_2$31 etching, all Te–O and W–O peaks vanish and only Te–W and W–Te signatures remain; a second $_2$32 etch confirms that the oxide layer is confined to the top $_2$33 (Ye et al., 2016). AES mapping corroborates that O, W, and Te signals associated with oxides are confined to a nonuniform $_2$34 surface skin, with hot spots at defect sites (Ye et al., 2016).

The principal surface reaction in air is reported as

$_2$35

Both Te and W oxidize, unlike WSe$_2$36 or MoTe$_2$37, where only one element oxidizes in the cited comparison (Ye et al., 2016). First-principles estimates give essentially zero activation energy for $_2$38 dissociation on WTe$_2$39, $_2$40, compared with $_2$41 for MoTe$_2$42, $_2$43 for WSe$_2$44, and $_2$45 for black phosphorus (Ye et al., 2016). The absence of a barrier implies immediate monolayer reaction, whereas multilayers react only at their top surface (Ye et al., 2016).

The oxide is self-limiting. The in situ grown $_2$46 TeO$_2$47/WO$_2$48 skin on trilayer WTe$_2$49 provides partial passivation and helps prevent inner layers from further degradation (Ye et al., 2016). This is why a Raman-active core survives for days even in ambient conditions.

7. Experimental methodologies, device implications, and recurrent misconceptions

The main spectroscopic probes used for trilayer WTe$_2$50 are Raman spectroscopy, XPS/AES, and ARPES. In the $_2$51-ARPES study, a photon energy of $_2$52 from the fourth harmonic of a Ti:sapphire laser, a spot size of $_2$53, total energy resolution of $_2$54, momentum resolution of $_2$55, and sample temperature below $_2$56 enabled direct measurement of band overlap in graphene-encapsulated few-layer flakes (Sakano et al., 2021). In the thickness-evolution ARPES study, measurements with $_2$57 at $_2$58 along $_2$59–$_2$60–$_2$61 and $_2$62–$_2$63–$_2$64 resolved the suppression of the monolayer gap by three layers (Qiao et al., 18 Apr 2026). For Raman-based layer assignment, laser power must be kept below $_2$65 to avoid local heating or damage (Kim et al., 2015).

Several points are easily misconstrued. First, trilayer WTe$_2$66 is not simply a more stable version of monolayer WTe$_2$67; it remains chemically unstable in air, but its degradation saturates because oxidation is confined to a thin surface region (Ye et al., 2016). Second, trilayer WTe$_2$68 is not a conventional gapped topological insulator in transport terms: although $_2$69 is reported in the thickness-evolution study, the absence of a full global gap means bulk conduction masks metallic edge states (Qiao et al., 18 Apr 2026). Third, trilayer Raman spectra are not merely attenuated bulk spectra; the peak-preserving but frequency-shifted pattern, especially the near-invariance of the in-plane tungsten-chain mode, reflects the quasi-1D structural anisotropy of the $_2$70 lattice (Kim et al., 2015).

For device fabrication, the reported routes to enhanced stability are procedural rather than intrinsic. Recommended approaches include exfoliation and processing in inert gas gloveboxes or under vacuum, encapsulation with hexagonal-BN layers, blending WTe$_2$71 in protective polymers such as PVA, and refreshing degraded surfaces by gentle calibrated Ar$_2$72 ion etching to remove oxide and expose fresh $_2$73-WTe$_2$74 (Ye et al., 2016). These strategies are motivated by the observation that the decisive oxidation occurs at the initial exposed surface and that once a $_2$75 skin forms, the underlying trilayer remains comparatively intact (Ye et al., 2016).

Taken together, the established picture is that trilayer WTe$_2$76 is the first semimetallic member of the few-layer series, identifiable by a four-mode Raman signature, governed electronically by interlayer-coupling-driven band overlap and odd-layer topological character, and chemically limited by rapid but self-terminating surface oxidation (Kim et al., 2015, Ye et al., 2016, Sakano et al., 2021, Qiao et al., 18 Apr 2026). Its significance lies in this concurrence of lattice anisotropy, semimetallicity, finite inversion-breaking responses, and thickness-dependent topology within a single, experimentally accessible film thickness.