PolarMem: Engineering Polar Textures in Metals

• PolarMem is a concept that exploits domain topology and multicomponent tilt symmetry to stabilize emergent nanoscale polar textures in conductive ferromagnets.
• The approach uses freestanding SrRuO₃ membranes to concentrate strain gradients at specific defects, enabling polar antiphase boundaries and nanoclusters with measurable polar displacements.
• Advanced microscopy and DFT+U calculations validate that Néel-like interpolation at hard APBs generates a strong polar response, distinctly differing from nonpolar easy APBs.

PolarMem is a membrane-based polar metal concept in which domain topology and multicomponent tilt symmetry are leveraged to stabilize, reconfigure, and utilize nanoscale polar textures within a conducting ferromagnet. In the reported realization, freestanding $34\ \mathrm{nm}$ SrRuO$_3$ membranes with orthorhombic $Pbnm$ symmetry and Glazer tilt pattern $a^{-}a^{-}c^{+}$ develop a hierarchy of ferroelastic domain refinement after release from the substrate, and this structural reorganization spontaneously generates two distinct classes of emergent polar texture: polar antiphase boundaries at translation-inequivalent “hard” APBs and polar nanoclusters at embedded $90^\circ$ ferroelastic walls. Translation-equivalent “easy” APBs remain nonpolar. The concept addresses a central obstacle in polar metals—screening of long-range dipoles by itinerant electrons—by localizing symmetry breaking and strain gradients at specific walls and defects rather than relying on a homogeneous polar phase (Haria et al., 30 Apr 2026).

1. Material system and membrane transformation

The material platform is epitaxial SrRuO$_3$ (SRO), a conductive, ferromagnetic perovskite that in bulk form is centrosymmetric yet already hosts the multicomponent antiferrodistortive tilt field characteristic of $a^{-}a^{-}c^{+}$ octahedral order. Releasing epitaxial SRO films from their substrates removes long-range clamping, drives ferroelastic domain refinement from tens of micrometres to tens of nanometres, and produces membrane wrinkling due to mechanical relaxation. In the clamped state, ferroelastic $X/X^\*$ and $Y/Y^\*$ domains exhibit lateral dimensions of tens of micrometres with uniform interiors; in the freestanding state, dense nanoscale contrast variations indicate nanodomain refinement, and atomic-resolution mapping additionally reveals $Z/Z^\*$ regions that are not accessible from mesoscale contrast alone (Haria et al., 30 Apr 2026).

This membrane transformation is the enabling step in PolarMem. The freestanding geometry concentrates strain gradients at topological boundaries and embedded walls, creating local conditions under which electric polarization can emerge even though the bulk metallic state disfavors long-range dipoles. The resulting polar textures are not interfacial artifacts of heterostructure design and do not require chemical asymmetry. Instead, they are generated internally by symmetry-selective boundary interpolation and elastic accommodation within a single metallic oxide membrane.

2. Symmetry, order parameters, and roto-flexoelectric coupling

The primary structural order parameter is the multicomponent tilt field $_3$0, where $_3$1 are the out-of-phase $_3$2 tilts and $_3$3 is the in-phase $_3$4 tilt, all referred to pseudocubic axes. Within this framework, APBs in $_3$5 SRO split into two symmetry classes. Easy APBs are translation-equivalent and related by the parent-cubic half-translation $_3$6, which flips all three tilt components. Hard APBs are translation-inequivalent and related by a “c-flip” mirror that reverses $_3$7 but preserves $_3$8, giving $_3$9. Because this second relation cannot be realized by a primitive translation, the wall must interpolate continuously and inversion symmetry is locally broken (Haria et al., 30 Apr 2026).

The coupling framework is expressed in Landau–Ginzburg–Devonshire form. The relevant constitutive relations are $Pbnm$0 for flexoelectricity and $Pbnm$1 for rotostriction, while minimizing the free energy gives induced polarization of the form $Pbnm$2. A commonly used compact form is $Pbnm$3. The key consequence is that large tilt gradients alone are insufficient: polarization is amplified only if a finite tilt component survives where the gradients are maximal. This is exactly what distinguishes hard from easy APBs.

For easy APBs, the interpolation is Ising-like: all tilt components collapse at the wall center, so $Pbnm$4 precisely where $Pbnm$5 is largest, suppressing $Pbnm$6. For hard APBs, the interpolation is Néel-like: $Pbnm$7 pass through zero with continuous rotation while $Pbnm$8 remains finite. The preserved $Pbnm$9 component then provides a nonzero prefactor where gradients of $a^{-}a^{-}c^{+}$0 are steepest, amplifying the roto-flexoelectric response.

3. Boundary classes and emergent polar textures

The PolarMem mechanism is organized by boundary type rather than by bulk phase transition. Two classes of wall-localized polar texture and one nonpolar reference class are identified.

Boundary type	Interpolation or strain mechanism	Polar response
Easy APB	Translation-equivalent; Ising-like collapse of all tilt components	Nonpolar; no response above the $a^{-}a^{-}c^{+}$1 STEM sensitivity
Hard APB	Translation-inequivalent; Néel-like interpolation with preserved $a^{-}a^{-}c^{+}$2	Polar displacements exceeding $a^{-}a^{-}c^{+}$3 and extending over $a^{-}a^{-}c^{+}$4
Embedded $a^{-}a^{-}c^{+}$5 ferroelastic wall	Elastic accommodation of strain mismatch plus rotostriction	Polar nanoclusters of characteristic $a^{-}a^{-}c^{+}$6

At a hard APB spanning roughly four atomic planes, or about $a^{-}a^{-}c^{+}$7–$a^{-}a^{-}c^{+}$8 in width, the hydrostatic strain changes sharply from $a^{-}a^{-}c^{+}$9 to $90^\circ$0 over about three pseudocubic unit cells, corresponding to $90^\circ$1. Polar displacements exceed $90^\circ$2, are oriented away from the wall on both faces, and extend over roughly $90^\circ$3. By contrast, easy APBs show only mild, non-local tensile strain of about $90^\circ$4 unrelated to the APB itself and no detectable polar response. Embedded $90^\circ$5 ferroelastic walls constitute a distinct source of polarity: elastic accommodation of deviatoric eigenstrain mismatch, together with rotostriction as the tilt field interpolates across the wall, produces localized hydrostatic strain gradients of order $90^\circ$6 and yields clustered polar regions of characteristic size about $90^\circ$7 (Haria et al., 30 Apr 2026).

This boundary selectivity is the defining feature of PolarMem. Polarization is not ubiquitous at all topological defects; it appears only where symmetry and interpolation character preserve a finite roto-flexoelectric prefactor or where elastic accommodation generates strongly localized strain gradients.

4. Correlative experimental evidence across length scales

The experimental case rests on correlative microscopy extending from mesoscale electron channelling contrast imaging to atomic-resolution scanning transmission electron microscopy. ECCI establishes the macroscopic change in domain architecture upon membrane release, while annular dark-field STEM combined with Fourier filtering resolves $90^\circ$8, $90^\circ$9, and $_3$0 variants at nanometre scales. Because ECCI alone cannot assign refined membrane domains without an epitaxial reference, the STEM analysis is essential for identifying the actual variant hierarchy (Haria et al., 30 Apr 2026).

Phase lock-in strain mapping quantifies local strain tensor components and the hydrostatic strain $_3$1. Polarization is inferred from in-plane polar vectors constructed from Ru off-centering relative to the surrounding Sr cage. The decisive observation is the co-localization of polar displacements with strain gradients and suppressed variant amplitude in Fourier maps. At hard APBs, this co-localization confirms a boundary-local origin and a flexoelectric drive. At embedded $_3$2 walls, polar vectors align along $_3$3, flowing from tensile to compressive zones. The polarization is spatially heterogeneous yet coherent over several nanometres, exceeding the sub-unit-cell screening length of SRO and matching the spatial extent of the strain-gradient field.

The screening context is central. SRO has $_3$4, $_3$5, and $_3$6, yet the local polar textures persist over approximately $_3$7–$_3$8. The reported interpretation is that these textures are mechanically driven and flexoelectric rather than stabilized by long-range electrostatics.

5. First-principles corroboration and atomistic interpolation physics

Spin-polarized DFT+$_3$9 calculations in VASP corroborate the symmetry-selective interpolation mechanisms. The calculations use PBEsol, $a^{-}a^{-}c^{+}$0 on Ru, a plane-wave cutoff of $a^{-}a^{-}c^{+}$1, Gaussian smearing of $a^{-}a^{-}c^{+}$2, and structural relaxations to residual forces below $a^{-}a^{-}c^{+}$3. The computed orthorhombic lattice parameters are $a^{-}a^{-}c^{+}$4, $a^{-}a^{-}c^{+}$5, and $a^{-}a^{-}c^{+}$6, in close agreement with experiment (Haria et al., 30 Apr 2026).

The boundary models reproduce the experimentally inferred polar selectivity. An easy APB modeled in a $a^{-}a^{-}c^{+}$7 supercell shows Ising-like collapse of Ru off-centering and all tilt components at the wall center, with only a weak dipole response. A hard APB modeled in an $a^{-}a^{-}c^{+}$8 pseudocubic supercell exhibits Néel-like interpolation, with reversal of the $a^{-}a^{-}c^{+}$9 component while $X/X^\*$0 is preserved; local Ru off-centering increases, consistent with a polar boundary. For $X/X^\*$1 walls, DFT yields Ru off-centering of $X/X^\*$2–$X/X^\*$3 together with a finite strain gradient and dipole across the wall, supporting the flexoelectric origin of the nanoclusters.

These calculations do not merely reproduce polarity qualitatively. They specifically validate the distinction between Ising-like collapse and Néel-like rotation as the atomistic criterion for whether a multicomponent $X/X^\*$4 boundary can amplify roto-flexoelectric coupling.

6. Functional significance, related directions, and open problems

PolarMem positions a freestanding metallic oxide membrane as a platform for reconfigurable polar–magnetic textures. Because SrRuO$X/X^\*$5 is a conductive, ferromagnetic perovskite, local polarization at selected domain walls offers routes to magnetoelectric coupling based on gradient-driven polar states within a single metallic layer. The proposed functionalities include local modulation of Ru–O–Ru bond angles and inversion symmetry, with possible consequences for spin–orbit coupling, Dzyaloshinskii–Moriya interactions, chiral spin textures, domain-wall electronics, and neuromorphic elements that rely on boundary topology rather than bulk ferroelectric switching. A related strand of work on polar metals showed that symmetric polar-metal electrode/ferroelectric capacitors can suppress the ferroelectric critical-thickness limit through interfacial dipolar coherency when a component of the electrode’s polar axis is perpendicular to the interface (Puggioni et al., 2016). PolarMem instead localizes the operative symmetry breaking within the membrane itself, without heterointerfaces or multilayer twisting (Haria et al., 30 Apr 2026).

Several limitations remain explicit. Polarization is inferred from picometre-scale Ru off-centering rather than converted into absolute $X/X^\*$6 in $X/X^\*$7, and the full tensorial values of $X/X^\*$8 and $X/X^\*$9 have not been extracted. Direct measurements linking polar walls or nanoclusters to magnetism, Hall responses, or anisotropy are not reported. Stability under bias, temperature, and cycling remains to be tested, as does deterministic patterning of hard APBs and $Y/Y^\*$0 walls for device architectures. Whether these polar textures are switchable by electric fields in a metal, likely not, or reconfigurable mechanically at device-relevant speeds is unresolved. Advanced probes such as scanning probe microscopy, second-harmonic generation, and nanobeam diffraction are identified as routes for direct verification of local polarity and dynamic reconfiguration.

A plausible implication is that PolarMem supplies a symmetry-based design rule for other metallic perovskites and mixed-valence oxides: translation-inequivalent boundaries that enforce Néel-like interpolation in multicomponent tilt systems should be the natural sites for robust wall-confined polar textures, whereas boundaries that collapse all order-parameter components should remain nonpolar. In that sense, PolarMem is less a single material result than a topological and symmetry-guided blueprint for engineering local polar functionality inside conducting oxide membranes.