Sliding Ferroelectrics: Interlayer Charge Switching

• Sliding ferroelectrics are layered van der Waals materials that generate spontaneous out-of-plane polarization through stacking-dependent interfacial charge transfer and lateral sliding of layers.
• Their switching mechanism exploits symmetry-controlled energy landscapes and domain wall dynamics, enabling multistate nonvolatile memory and tunable topological, magnetic, and optical properties.
• Experimental and computational studies reveal superlubric, low-barrier switching behavior with promising applications in energy-efficient nanoelectronic and optoelectronic devices.

Sliding ferroelectrics are layered, usually van der Waals, systems in which a spontaneous out-of-plane polarization is generated by stacking-dependent interfacial charge transfer and reversed by in-plane relative sliding of the constituent layers rather than by the conventional intralayer ion-displacement mechanism. The class now spans bilayer and multilayer inorganic materials such as $h$-BN, rhombohedral $MX_2$, MBi$_2$Te$_4$, CuF$_2$, and Fe$_5$GeTe$_2$, as well as an amphidynamic coordination polymer, $(15\text{-Crown-5})$Cd$_3$Cl$_6$, in which direct macroscopic $MX_2$0–$MX_2$1 hysteresis was reported (Wang et al., 14 May 2025, Meng et al., 2022, Dong et al., 11 Jun 2025, Miao et al., 2023).

1. Structural origin and polarization formalism

The defining structural motif is a stack of nonpolar or weakly polar monolayers whose relative registry removes inversion or mirror symmetry and creates a vertical dipole. In bilayer MBi$MX_2$2Te$MX_2$3 $MX_2$4, the centrosymmetric AA′ bilayer is unstable and spontaneously slides into two equivalent symmetry-broken minima, AB′$MX_2$5 and AB′$MX_2$6, carrying $MX_2$7 and $MX_2$8, respectively; the two states are reversibly connected by in-plane sliding along either $MX_2$9 or $_2$0 (Dong et al., 11 Jun 2025). In 3R-bilayer MoS$_2$1, two energy-degenerate stackings, XM and MX, are related by a shear translation $_2$2, and the interface dipole reverses when that lateral displacement is executed (Li et al., 2024). In Janus TMD bilayers $_2$3, the sliding-induced interlayer polarization coexists with an intrinsic monolayer electric field set by the electronegativity difference between the two chalcogens, so the monolayer polarity itself becomes a control parameter for the bilayer ferroelectric response (Mahajan et al., 28 May 2025).

The polarization is typically evaluated with the modern Berry-phase formalism. In MBi$_2$4Te$_2$5, the polarization is written as

$_2$6

and, for the sliding path from AA′ to a polar AB′ state,

$_2$7

The same systems admit an equivalent charge-density picture in which

$_2$8

making explicit that the dipole originates from layer-resolved charge redistribution across the vdW gap (Dong et al., 11 Jun 2025). Closely related Berry-phase and charge-integration forms are used in bilayer MoS$_2$9, Janus TMDs, and $_4$0-ZrI$_4$1 (Gao et al., 2024, Mahajan et al., 28 May 2025, Wan et al., 4 Oct 2025).

Representative computed values show that the scale of polarization is system dependent rather than universal. For MBi$_4$2Te$_4$3, $_4$4, $_4$5, and $_4$6 pC/m for GeBi$_4$7Te$_4$8, SnBi$_4$9Te$_2$0, and PbBi$_2$1Te$_2$2, respectively (Dong et al., 11 Jun 2025). Bilayer CuF$_2$3 reaches $_2$4 pC/m (Peng et al., 11 Mar 2026), whereas $_2$5-ZrI$_2$6 is reported as $_2$7 (Wan et al., 4 Oct 2025). This spread suggests that sliding ferroelectricity is best viewed as a mechanism class rather than a single materials family with one characteristic polarization scale.

2. Symmetry, energy landscapes, and switching coordinates

The symmetry reduction induced by sliding controls which polarization components are allowed and how electric fields couple to the sliding coordinate. In rhombohedral $_2$8 bilayers, the two ferroelectric ground states M and N have point group $_2$9, which allows only $_5$0, while the midpoint A along the sliding path has $_5$1, which forbids $_5$2 but allows an in-plane $_5$3. For a generic intermediate registry of point group $_5$4, both $_5$5 and $_5$6 are permitted, while $_5$7 remains forbidden for sliding along $_5$8. For MoS$_5$9, the calculated evolution is $_2$0 pC/m, $_2$1, $_2$2 pC/m, with $_2$3 pC/m (Jafari et al., 2023). The existence of a large in-plane dipole at the intermediate state is a recurring feature of sliding pathways.

In $_2$4-BN bilayers, the coupling between a vertical electric field and lateral motion is governed by the off-diagonal Born effective charge tensor. At AB or BA, $_2$5 by $_2$6 symmetry, but along the AB$_2$7SP$_2$8BA path the off-diagonal components become sharply nonzero and peak near the saddle point, so an out-of-plane field $_2$9 generates in-plane forces $(15\text{-Crown-5})$0 and $(15\text{-Crown-5})$1 that drive the required sliding (Wang et al., 14 May 2025). This is a symmetry-mediated electromechanical coupling rather than the direct longitudinal force familiar from oxide ferroelectrics.

The sliding potential is commonly represented by a double well. For BN, a fitted one-dimensional form is

$(15\text{-Crown-5})$2

with $(15\text{-Crown-5})$3, $(15\text{-Crown-5})$4, and $(15\text{-Crown-5})$5 (Deng et al., 2024). Because the saddle state also carries a finite in-plane dipole, an inclined electric field lowers the switching threshold more efficiently than a purely vertical field. The critical field is

$(15\text{-Crown-5})$6

and, for BN, it is minimized at $(15\text{-Crown-5})$7 below the normal, giving a reduction of over $(15\text{-Crown-5})$8 relative to purely vertical switching (Deng et al., 2024). This result is presented as universal for sliding ferroelectric systems whose monolayers belong to the symmetry group $(15\text{-Crown-5})$9.

3. Domain walls, nonuniform switching, and superlubric motion

A central controversy concerns whether polarization reversal proceeds by homogeneous rigid translation of an entire atomic layer or by domain-wall-mediated collective motion. For Bernal $_3$0-BN bilayers, the synchronized long-distance translation picture has been explicitly challenged: it is described as a “spinodal decomposition-like homogeneous switching process” that violates Neumann’s principle and is unlikely to occur because, in a pure AB or BA domain, $_3$1 symmetry forces the unit-cell-averaged in-plane force under $_3$2 to vanish (Ke et al., 3 Feb 2025). In the same system, any perturbation breaking the in-plane symmetry is reported to trigger avalanche-like switching, so the practical coercive field is governed by fluctuations and symmetry breaking rather than by a conventional intrinsic maximum of the restoring-force curve (Wang et al., 14 May 2025).

The alternative picture centers on broad domain walls. In $_3$3-BN, a “wave-like” wall of width $_3$4 nm is described by an order parameter

$_3$5

with free-energy density

$_3$6

equilibrium width $_3$7, and profile $_3$8 (Ke et al., 3 Feb 2025). The key claim is that coherent propagation of this wide interface has an almost zero activation barrier because the local energy increase at the leading edge is balanced by the local decrease at the trailing edge. In the same material, the off-diagonal Born charges peak near the $_3$9 region inside the wall, producing a nonuniform driving force and a “wriggling” wall motion under $_6$0 (Wang et al., 14 May 2025).

The resulting dynamics are superlubric or soliton-like rather than creep-like. Finite-field MD for $_6$1-BN gives $_6$2 m/s at $_6$3 K and $_6$4 V/nm for the $_6$5 wall, with extrapolated velocities up to $_6$6 m/s and an anomalous cooling-promoted increase of velocity (Ke et al., 3 Feb 2025). In bilayer 3R-MoS$_6$7, the domain wall width is $_6$8, the characteristic speed is $_6$9 m/s, and the translational coordinate obeys $MX_2$00, yielding uniformly accelerated motion under field and constant coasting velocity after field removal (Shi et al., 4 Feb 2025). Machine-learning molecular dynamics on ferroelectric MoS$MX_2$01 moiré superlattices sharpen this distinction further: the rigid-sliding barrier is $MX_2$02 meV/atom, whereas the domain-wall-mediated barrier is $MX_2$03 meV/atom, sufficient to explain thermally driven interlayer drift of order $MX_2$04 m/s at 300 K; sulfur vacancies of about $MX_2$05 already induce a sliding-to-pinning transition (Li et al., 22 Apr 2026).

4. Thickness dependence and multistate ferroelectricity

Layer number introduces an additional discrete degree of freedom absent in ordinary bilayer descriptions. In multilayer 3R-MoS$MX_2$06, each MoS$MX_2$07/MoS$MX_2$08 interface is modeled by an interfacial dipole $MX_2$09, with free energy

$MX_2$10

This construction yields $MX_2$11 distinct net-polarization values in an $MX_2$12-layer stack and explains the anomalous intermediate polarization states observed in dual-gate 3R-MoS$MX_2$13 devices (Meng et al., 2022). In trilayers, the reported switching path consists of two sequential interfacial flips, each with a barrier of $MX_2$14 meV per unit cell, separated by an intermediate valley of about $MX_2$15 meV per cell. Bilayer, tetralayer, and pentalayer barriers remain in the narrow range $MX_2$16–$MX_2$17 meV per unit cell, consistent with predominantly local interfacial energetics (Meng et al., 2022).

A distinct multilayer realization appears in altermagnetic CuF$MX_2$18. Bilayer CuF$MX_2$19 has two ferroelectric minima, FE-I and FE-II, at sliding displacements $MX_2$20, separated by a nonpolar intermediate state at $MX_2$21, with $MX_2$22 meV/atom and $MX_2$23 pC/m (Peng et al., 11 Mar 2026). In quadrilayer CuF$MX_2$24, four inequivalent ferroelectric states are identified: $MX_2$25

$MX_2$26

Neighboring-state switching requires sliding one bilayer by $MX_2$27 with a barrier of about $MX_2$28 meV/atom (Peng et al., 11 Mar 2026). This suggests a general multilayer design rule: once interfacial dipoles become individually addressable, sliding ferroelectrics can support multilevel nonvolatile states without changing composition.

5. Coupling to topology, magnetism, and spin/layer transport

Sliding ferroelectricity has become a practical route to electrically tunable topological phases. In bilayer MBi$MX_2$29Te$MX_2$30, sliding between oppositely polarized states induces reversible band inversion accompanied by topological phase transitions. Without SOC, the AB′$MX_2$31 bilayers are ordinary insulators; with SOC, a Bi-$MX_2$32/Te-$MX_2$33 inversion closes and reopens the gap, producing indirect gaps of 31 meV in GeBi$MX_2$34Te$MX_2$35, 36 meV in SnBi$MX_2$36Te$MX_2$37, and 35 meV in PbBi$MX_2$38Te$MX_2$39. The nontrivial phase is confirmed both by the WCC-derived $MX_2$40 invariant and by gapless helical edge states spanning the bulk gap (Dong et al., 11 Jun 2025). The corresponding low-energy description,

$MX_2$41

encodes the central switching effect: sliding changes the interlayer potential and hybridization such that $MX_2$42, toggling band inversion (Dong et al., 11 Jun 2025). A related lattice-model and first-principles program on bilayer ScI$MX_2$43 predicts a sliding-induced sequence of second-order TI, spin-hybrid-order TI, quantum spin Hall, quantum anomalous Hall, and trivial phases, with distinct anomalous Nernst signatures (Yang et al., 2 Jun 2025).

Magnetic and metallic extensions show that polarization reversal can be coupled directly to spin splitting and transport. In bilayer Fe$MX_2$44GeTe$MX_2$45, the nonpolar AFM AA′ phase converts, by interlayer sliding, into FE AB or BA ferrimagnetic metallic states with $MX_2$46 pC/m, sliding barrier $MX_2$47 meV per unit cell, and net magnetization $MX_2$48 per cell (Guo et al., 11 Aug 2025). Broken in-plane mirror symmetry lifts the Kramers degeneracy, generating $MX_2$49-dependent spin splitting of $MX_2$50–$MX_2$51 meV near the Fermi crossings, while FE switching reverses polarization, spin splitting, and magnetization simultaneously. The same system exhibits a linear magnetoelectric coefficient $MX_2$52, anomalous Hall conductivity $MX_2$53, and anomalous Nernst conductivity $MX_2$54 (Guo et al., 11 Aug 2025).

In altermagnetic CuF$MX_2$55, the ferroelectric polarization is directly coupled to layer-locked $MX_2$56-wave spin splitting. The data are summarized by

$MX_2$57

so reversing $MX_2$58 reverses the sign of the spin splitting and the layer localization of the spin-polarized carriers (Peng et al., 11 Mar 2026). More generally, rhombohedral $MX_2$59 sliding ferroelectrics preserve a large Zeeman-type splitting throughout the entire sliding path; reported $MX_2$60-valley values include 148 meV in MoS$MX_2$61, 184 meV in MoSe$MX_2$62, 417 meV in WS$MX_2$63, and 453 meV in WSe$MX_2$64 (Jafari et al., 2023). This indicates that the sliding coordinate can reconfigure polarization while leaving strong SOC-driven spin polarization intact.

6. Experimental manifestations, optical responses, and device implementations

Direct experimental access to sliding ferroelectricity has been hindered by weak polarization and leakage in many 2D systems, but an important exception is the amphidynamic crystal $MX_2$65Cd$MX_2$66Cl$MX_2$67. Its large band gap $MX_2$68 eV permits direct macroscopic ferroelectric measurements, with $MX_2$69–$MX_2$70 loops showing $MX_2$71–$MX_2$72 and $MX_2$73 rising from $MX_2$74 kV/cm at 293 K to $MX_2$75 kV/cm at 273 K (Miao et al., 2023). The same material exhibits SHG intensity reaching $MX_2$76 quartz at 300 K, pyroelectric polarization $MX_2$77, and switchable $MX_2$78 PFM domains under $MX_2$79 V tip bias (Miao et al., 2023). Because the net polarization decomposes into geometric and sliding contributions,

$MX_2$80

it also demonstrates that sliding ferroelectricity can coexist with other polar mechanisms in a single lattice (Miao et al., 2023).

Optical and electronic control channels are now diverse. In 3R-bilayer MoS$MX_2$81, photoexcitation tunes the sliding polarization over a large range at fixed sliding coordinate, with

$MX_2$82

$MX_2$83 pC/m, quenching to $MX_2$84 pC/m at $MX_2$85 e/u.c. and recovering to $MX_2$86 pC/m at $MX_2$87 e/u.c.; the switching barrier changes only modestly, from $MX_2$88 meV/u.c. in the dark to $MX_2$89 meV/u.c. at $MX_2$90 and $MX_2$91 meV/u.c. at $MX_2$92 (Gao et al., 2024). A separate nonlinear-optical line predicts photo-injected spin-plus-orbital magnetization of order $MX_2$93–$MX_2$94 in ZrI$MX_2$95, WTe$MX_2$96, and MoS$MX_2$97, with the sign reversing when the sliding dipole reverses (Zhou, 2022). In $MX_2$98-ZrI$MX_2$99, the linear electro-optic response is unusually electronic rather than ionic, with $_2$00 pm/V and $_2$01 pm/V; under biaxial strain, $_2$02 varies nearly linearly from 16.1 to 43.2 pm/V and follows $_2$03 (Wan et al., 4 Oct 2025). In WTe$_2$04, exciton condensation has been proposed to stabilize the sliding-ferroelectric state beyond single-particle DFT, increasing the barrier from $_2$05 meV to $_2$06–$_2$07 meV and yielding excitonic gaps of $_2$08–$_2$09 meV (D'Alessio et al., 1 Oct 2025).

Device demonstrations establish that the phenomenon is technologically operative rather than merely model-level. Bilayer-MoS$_2$10 sliding ferroelectric memories exhibit a memory window $_2$11 V, conductance ratio above $_2$12, retention time $_2$13 years, and programming endurance $_2$14 cycles at room temperature; flexible devices retain their performance after $_2$15 bending cycles and support 32 conductance levels, Hebbian plasticity, and 97.81% MNIST accuracy when mapped into a five-layer CNN (Li et al., 2024). The sliding ferroelectric resonant tunnel junction extends the concept to two-terminal transport, combining atomically thin polar barriers, superlubric shear-soliton switching, and momentum-conserving resonant tunneling to realize writing voltages below $_2$16 V, reading biases under $_2$17 V, current densities exceeding $_2$18 nA $_2$19m$_2$20, a room-temperature ON/OFF ratio $_2$21, and switching energies below $_2$22 fJ within the SLAP architecture (Raab et al., 28 Feb 2026).

Sliding ferroelectrics therefore occupy a distinct place within ferroic condensed matter. Their order parameter is set by interlayer registry, their switching can be domain-wall-mediated and superlubric rather than nucleation-limited, and their polarization readily hybridizes with topological, magnetic, optical, and resonant-transport phenomena. A plausible implication is that the most consequential future developments will come from co-designing these couplings—rather than optimizing polarization alone—because the literature already shows nonvolatile electrical control of $_2$23 topology, anomalous Hall transport, altermagnetic spin splitting, electro-optic response, and multistate memory in the same sliding-based framework (Dong et al., 11 Jun 2025, Guo et al., 11 Aug 2025, Peng et al., 11 Mar 2026, Wan et al., 4 Oct 2025).