WorksPhysics and Astronomy
Polar meron-antimeron networks in strained and twisted bilayers
Abstract
Out-of-plane polar domain structures have recently been discovered in strained and twisted bilayers of inversion symmetry broken systems such as hexagonal boron nitride. Here we show that this symmetry breaking also gives rise to an in-plane component of polarization, and the form of the total polarization is determined purely from symmetry considerations. The in-plane component of the polarization makes the polar domains in strained and twisted bilayers topologically non-trivial, forming a network of merons and antimerons (half-skyrmions and half-antiskyrmions). For twisted systems, the merons are of Bloch type whereas for strained systems they are of Néel type. We propose that the polar domains in strained or twisted bilayers may serve as a platform for exploring topological physics in layered materials and discuss how control over topological phases and phase transitions may be achieved in such systems.
Introduction
Recently it has been realized that ferroelectricity can occur in layered systems comprised of stacks of two-dimensional (2D) materials such as hexagonal boron nitride (hBN), see Figure 1 (a), provided the stack of layers does not have inversion symmetry [1]. In an aligned stack of hBN (3R stacking), which has four non-orthogonal mirror planes and is therefore non-centrosymmetric but still non-polar, sliding one layer over the other breaks the mirror symmetry about the plane which is parallel to and half-way between the layers, resulting in an interlayer transfer of electronic charge and an out-of-plane polarization [1, 2, 3], see Figure 1 (b). For anti-aligned hBN (2H stacking), there is an inversion center for every stacking, and the system is nonpolar. Applying an electric field to aligned hBN, the polarization can be inverted via a relative sliding between the layers (van der Waals sliding) [4, 5] in order to align the polarization with the field, see Figure 1 (c). This mechanism is highly unconventional when compared to the ferroelectricity observed in ABO oxide perovskites, in particular because the polarization generated is perpendicular to the atomic motion.
Figure 1. Ferroelectricity in twisted bilayers. (a): Sketch of a monolayer of hBN. (b): Illustration of the interlayer charge transfer and resulting polarization arising from relative sliding in bilayer hBN, where and are the charges of the B and N atoms, respectively, and is the equilibrium layer separation. (c): Out-of-plane polarization and (d) stacking energy in bilayer hBN as a function of relative stacking along the unit cell diagonal, from first-principles calculations in Ref. [3]. The stacking configurations at the extrema, AA, AB, SP and BA are labeled and sketched above the plots, and their stacking energies , and are marked on the vertical axis of (d). Changing between the energetically stable AB and BA stacking configurations via vdW sliding inverts the polarization between the minimum and maximum values and . (e)–(g): Sketch of red and blue hexagonal bilayers with relative twist angles of, , respectively.
In a twisted bilayer, two layers are twisted with respect to one another, forming a supercell known as a moiré superlattice, see Figure 1 (d)–(Figure 1f). A moiré superlattice can also be generated by introducing a small relative strain or lattice mismatch between the layers. Twisting has been shown to result in novel phenomena such as superconducting [6] and insulating [7] behavior in bilayer ‘magic angle graphene’, and recently ferroelectricity in hBN [5, 8, 9]. A small lattice mismatch moiré superlattice formed between non-Bravais lattice monolayers has local regions with different stacking configurations, which may locally break mirror symmetry. This symmetry breaking in conjugation with absence of inversion center in the monolayer leads to local out-of-plane polarization with stacking dependent direction [2, 3]. Thus, the stacking domains in strained and twisted bilayers can be identified as out-of-plane ‘moiré polar domains’ (MPDs). The experimentally observed ferroelectricity has been attributed to the motion of the domain walls separating the MPDs in response to an applied out-of-plane electric field. As a result, the MPDs with polarization (anti-)aligned to the field (shrink) grow in size.
Something which to our knowledge has not been considered is the possibility of an in-plane polarization, both in moiré superlattices and commensurate layered systems. Because layered systems are periodic in the in-plane directions, the in-plane polarization is a lattice valued quantity, and only changes in the in-plane polarization are well defined, modulo a quantum of polarization [10]. 2D honeycomb compounds with an AB sublattice structure have a triangular in-plane polarization lattice [11], and it is natural to expect that changing the stacking configuration in a bilayer may result in a continuous change in the in-plane polarization.
We show with first-principles calculations of bilayer hBN that an in-plane polarization is indeed generated in layered system when one layer slides over the other. As a consequence, the MPDs do not just point in the out-of-plane direction, but also have an intricate in-plane component, such that the polarization vector has topologically non-trivial winding and the MPDs form a network of merons and antimerons (winding numbers ). This indicates that the polar properties of layered systems can have rich topological structure. Topological polar structures such as skyrmions [12, 13] and merons [14] have been observed in ferroelectric materials such as oxide perovskites, and have been shown to result in novel physics of interest for future applications in nanotechnology, such as negative capacitance [15] and high density information processing [13]. So far, band topology in the moiré systems has appeared in the electronic structure of magic angle graphene [16, 17], Chern bands in twisted topological insulators [18], and topological superconductivity in twisted cuprates [19]. The polar meron-antimeron network suggests that moiré materials also exhibit real space topology, echoing recent similar discoveries of topologically nontrivial strain fields in twisted bilayers [20] and magnetic textures in moiré patterned topological insulators [21].
Results
We first calculate the out-of-plane and in-plane polarization of 3R-stacked bilayer hBN, using the first-principles SIESTA [22] and ABINIT [23] codes (see Methods). The polarization is calculated in commensurate 3R-stacked bilayer hBN as a function of relative displacement between the layers, i.e. in configuration space [24], which can be used to estimate the polarization in real space for arbitrary strains and twist angles, provided the mismatch is small enough that the local stackings are well approximated by a commensurate bilayer plus a relative translation (see Supplementary Material, Section I). A changing in-plane polarization was found, of the same order of magnitude as the out-of-plane polarization, see Figure 2 (a) and (Figure 2b). The shape of the polarization field as a function of relative stacking is determined purely from symmetry considerations, although the magnitude is material specific. The aligned AA stacking with space group (#187) has three out-of-plane mirror planes running through rotations of the unit cell diagonal, where and , and an in-plane mirror plane half way through the layers. The bilayer is therefore non-polar for this stacking, but because the mirror planes are not orthogonal, it is not centrosymmetric. Sliding one layer over the other by or along the unit cell diagonal (or one of its rotations), the energetically favourable AB and BA stackings are realized, both with the polar space group (#156). The three mirror symmetries through rotations of the unit cell diagonal are preserved, but the in-plane mirror symmetry is broken, allowing for a polarization only in the out-of-plane direction. Half way between the AB and BA stacking configurations, at the saddle point (SP, ), the (#39) space group is realized, which only has an out-of-plane mirror symmetry through the unit cell diagonal, assuming the relative translation of the layers is along this diagonal. Additionally, the system is left invariant after mirroring about the plane half-way between the layers plus a non-symorphic translation of , preventing any out-of-plane polarization. Thus, only an in-plane polarization along the unit cell diagonal is allowed. For any other translation along the unit cell diagonal or one of its rotations, the bilayer has the Cm (#8) space group, with only the mirror plane running through that diagonal. The polarization is then confined to that mirror plane, but can have both in-plane and out-of-plane components. Finally, for a translation not along the unit cell diagonal, the P1 (#1) space group is realized, and the polarization can point in any direction.
Figure 2. Total polarization in bilayer hBN. (a) Out-of-plane polarization and (b) change in in-plane polarization in fractional coordinates along the configuration space diagonal, calculated with SIESTA (black) and ABINIT (red), where and are the normalized lattice vectors of the bilayer, and is the unit vector normal to the bilayer. The hollow points are first-principles measurements and the solid lines are fitting to symmetric basis functions. The winding of the total polarization along the configuration space diagonal is sketched above. 2D plot of (c) out-of-plane and (d) in-plane polarization in configuration space, in Cartesian coordinates and is the bilayer lattice constant. The black arrows represent the commensurate bilayer lattice vectors.
Using the in-plane mirror symmetry of the AA stacking configuration, we can further deduce that the out-of-plane polarization must be an odd function of in-plane translations (see Supplementary Material, Section I). Requiring also that transforms as a scalar field with respect to rotations about the out-of-plane axes through AA, AB and BA, it can be shown to be of the form
in fractional coordinates in configuration space, where are fractions of the lattice vectors and ; the fractional and Cartesian coordinates in configuration space are related via , where
We can also show that the in-plane polarization must be even with respect to in-plane translations. Additionally, it should transform like a vector with with respect to rotations about the out-of-plane axes, and therefore must be of the form
Eqs. (1) and (2) were fit to the 1D data along the unit cell diagonal in Figure 2 (a) and (Figure 2b). The 1D data were sufficient to obtain a good fit to the 2D functions, which was verified by fitting to a larger set of displacements parametrizing the entire unit cell in configuration space. The out-of-plane and in-plane polarization in Cartesian coordinates, are obtained by the transformations and and are shown in Figure 2 (c) and (Figure 2d), respectively. The shape of the out-of-plane polarization in bilayer hBN as a function of relative stacking is well known [2, 3]: forms a triangular domain structure, with each domain having three neighboring domains of opposite polarization. The in-plane polarization has a remarkable structure: flows into and out of the centers of the AB and BA domains, at which it is zero. The magnitude of is maximal along the lines joining the centers of the AB and BA domains. A six-pointed star forms around the AA stacking configuration, at which both in-plane and out-of-plane components of the polarization are zero, modulo a quantum of polarization. Also, we note from symmetry that .
Mapping from configuration space to real space, we see that the total polarization has a different form, depending on whether the moiré superlattice is induced via relative straining or twisting. For a relative strain between the layers, the mapping between configuration space and real space is . In this case, configuration space and real space are related via a simple scaling by . In Figure 3 (a) and (Figure 3b) we show the out-of-plane and in-plane polarization in a strained bilayer, which are identical to the polarization in configuration space. For a relative twist , the mapping between configuration space and real space is , for , (see Supplementary Material, Section I). For scalar fields, the quantities are related via a scaling by and a reorientation of the cell vectors. For vector fields, the general form is different in both spaces. In Figure 3 (d) and (Figure 3e) we show the out-of-plane and in-plane polarization in a twisted bilayer. We can see that the out-of-plane polarization has the same form, but the in-plane polarization curls in a clockwise manner around the centers of the AB/BA domains.
Figure 3. Polar topology in bilayer hBN. (a): out-of-plane polarization , (b): change in in-plane polarization and (c): local topological charge for biaxially strained hBN. (d): out-of-plane polarization, (e): change in in-plane polarization and (f): local topological charge for twisted hBN. Plots are shown in real space , with with length scales of the bilayer lattice constant divided by strain and twist angle for strained and twisted bilayers, respectively. The black arrows represent the Moiré superlattice vectors in each case. (g)-(i): Proposal of a system in which polar topological phase transitions may be driven by an applied electric field . A single cell of a moiré superlattice is embedded in a dielectric medium. (g): at zero electric field, a meron-antimeron pair forms, i.e. the topological charges in the individual domains are . (h): When a positive electric field is applied, the dielectric medium has normalized polarization , changing the winding around the boundary of the cell. As a result, the antimeron turns into an antiskyrmion (), and the meron vanishes (). (i): When a negative field is applied, the reverse occurs: the meron turns into a skyrmion (), and the antimeron vanishes ().
The components of polarization in Figure 2 (a) and (Figure 2b) were reproduced by integrating the dynamical charges, [25], (see Supplementary Material, Section V). This further allows the decomposition of the polarization into the contributions from the displacements of different atoms and in different directions. The polarization is mostly generated by in-plane sliding, with negligible contributions from the out-of-plane displacements, i.e. the rippling of the interlayer separation as one layer slides over the other. As a result, the antisymmetric part of makes a significant contribution to the polarization, which is highly unusual for a ferroelectric material.
The in-plane component suggests that the polarization field in twisted bilayers does not just point in the out-of-plane direction, but exhibits intricate winding which is topologically nontrivial. Topology has played a manifest role in 2D materials, ranging from band theory to skyrmions in magnetic systems. Such skyrmions arise due to a mapping from a periodic unit cell to a classifying space that is topologically a sphere, as quantified via a homotopy winding.
In Figure 3 (c) and (Figure 3f) we show the local topological charge of strained and twisted bilayer hBN, respectively (see Methods). The AB and BA MPDs have equal and opposite winding. The total winding in each moiré cell is zero, but individually the AB and BA domains have winding numbers of , meaning the MPDs form a triangular network of merons and antimerons. The winding is concentrated at the domain centers and along the lines joining the AB/BA domain centers to the AA stacking configurations, and is zero along the domain walls. The magnitude of the winding is the same for strained and twisted bilayers apart from a reorientation of the axes, but the type of winding is different in each case. For strained bilayers the merons are of Néel type, where the polarization flows into and out of the domain centers. For twisted bilayers the merons are of Bloch type, where the polarization curls around the domain centers.
Discussion
In this work, we illustrate that, in addition to the out-of-plane polarization in layered systems like bilayer hBN, there is also an in-plane polarization as a result of a relative displacement between of the layers. This phenomenon is general to all layered systems, provided that the bilayer lacks inversion symmetry. For the case of 3R-stacked hBN and similar materials (MoS, etc.), there are three out-of-plane mirror planes related by rotations plus an in-plane mirror plane half-way between the layers, various combinations of which are broken as one layer slides over the other.
Our findings indicate that the polar properties of layered systems are much richer than previously thought. For un-twisted bilayer hBN, the energetically favorable AB and BA stackings have zero in-plane polarization. However, knowledge of how the in-plane polarization changes during the process of vdW sliding may prove useful. For example, by measuring the change in out-of-plane polarization, through a change in out-of-plane current, it is possible to determine when vdW sliding occurs between AB and BA domains, but it is not possible to determine in which direction the sliding occurs, i.e. to which of the three neighboring domains. By measuring the change in in-plane polarization, it may be possible to distinguish between these three sliding processes, which may enhance the capacity for information processing in ferroelectric layered systems.
Ferroelectric materials have been fabricated in many different geometries, from 2D thin films and FE/PE superlattices to 1D nanowires [26, 27] and nanotubes [28, 29] (in fact, all nanotubes are inherently polar via flexoelectricity [30, 31, 32]), and 0D quantum dots [33, 34]. Lower-dimensional ferroelectric systems typically exhibit size-dependent transitions in which the local polarization is more complex and can exhibit vortices before the polarization eventually vanishes completely [35, 36, 37, 38]. Soon after, it was realized these polar structures with vortices were topologically nontrivial [39], and skyrmion-like polarization structures were then identified, for example in barium titanate (BaTiO) nanowires embedded in a matrix of strontium titanate (SrTiO) [40]. It has also been proposed that skyrmions may be created by controlling domains and domain walls in ferroelectrics, where at low temperatures the domain walls are of Bloch type and contain an in-plane polarization [41]. Ferroelectric skyrmions have recently been experimentally observed in ferroelectric/paraelectric superlattices [12, 13]. In addition to skyrmions, polar merons have also been considered theoretically and signaled experimentally [14].
A full characterization of topological polarization also provides for an interesting avenue to explore. One major conceptual problem is that, unlike other topological invariants, the description of polarization in terms of exponentially localized Wannier functions requires topologically trivial electronic bands. It may be that topological polarization is a real space analogue to the topology of electronic bands in momentum space, with inversion symmetry and electric fields playing the role of time-reversal symmetry and magnetic fields. However, there is much work to be done in order to obtain a better understanding of topological polarization, and its relation to topological insulators in the presence of symmetries [42, 43, 44, 45] as well as recently discovered multi-gap topological states due to the natural reality condition [46, 47]. Nonetheless, it is evident that the topological polarization in moiré superlattices may serve as a new platform for topological physics in real materials, with great potential for the observation and control of topological phases in 2D materials.
We propose that it may be possible to drive a topological phase transition using a single moiré supercell embedded in a dielectric medium, from a single meron-antimeron pair into a skyrmion or antiskyrmion, for a fixed applied electric field, see Figure 3 (g)-(i). Such a setup may be possible by embedding a moiré quantum dot in a dielectric material, or in an aligned bilayer in which there is local strain or twisting around a defect, such that the bilayer is strained/twisted inside a domain and unstrained/untwisted outside. At zero electric field, a meron-antimeron pair forms in the supercell, and the polarization in the dielectric medium is zero. When an electric field is applied, the normalized polarization in the dielectric medium is , which changes the winding along the boundary of the cell, adding to the winding in one domain, turning
the meron/antimeron into a skyrmion/antiskyrmion, and canceling the winding in the other domain, making it topologically trivial, reminiscent of a bulk–boundary correspondence. A similar effect was also observed in periodic superlattices when the dielectric response of the system was taken into consideration. As a result, the polarization field has an in-plane component everywhere in the moiré superlattice, including the domain walls. This this leads to an induced winding along the domain walls, suggesting that even in periperiodic moiré systems, the merons/antimerons can be promoted to skyrmions/antiskyrmions, making the others topologically trivial, with an applied electric field.
It may also be possible to manipulate the meron–antimeron domain network via lattice reconstruction at different strains or twist angles. For small strains or twist angles, significant lattice relaxation can occur in order to increase the area of the more energetically stable stacking domains [24], making the polar domains sharper [2, 3]. At zero electric field, the AB and BA domains in 3R-stacked hBN relax evenly, leading to sharp triangular polar domains. However, the positions of the AA, AB, BA and SP stackings are all preserved. As a function of strain or twist angle, this is a continuous deformation which does not break any symmetries, and the meron–antimeron network should therefore be robust against lattice relaxation at zero electric field. When a field is applied, one type of domain will grow and the other will shrink, bending the polar domain walls. The robustness/fragility of the meron–antimeron network in response to the motion of the domain walls is not immediately clear. This goes beyond the scope of this work, however, and we leave it as a direction for future research.
In summary, we have illustrated that electronic out-of-plane and in-plane charge transfer and polarization are fundamental properties of layered systems without inversion symmetry. This will have far-reaching consequences both in terms of fundamental physics in strained/twisted and commensurate layered systems, such as ferroelectricity and topology, as well potential applications for ferroelectric-based nanodevices comprised of layers of 2D materials.
Methods
First-principles calculations
First-principles density functional theory (DFT) calculations were performed using the SIESTA [22] and ABINIT [23] codes, using PSML [48] norm-conserving pseudopotentials [49], obtained from Pseudo-Dojo [50]. SIESTA employs a basis set of numerical atomic orbitals (NAOs) [22], and double- polarized (DZP) orbitals were used for all calculations. The basis sets in SIESTA were optimized by hand, following the methodology in Ref. 51. ABINIT employs a plane wave basis set, which was determined using a kinetic energy cutoff of 1000 eV. A mesh cutoff of 1200 Ry was used for the real space grid in all SIESTA calculations. A Monkhorst-Pack -point grid [52] of was used for the initial geometry relaxations, and a mesh of was used to calculate the polarization. Calculations were converged until the relative changes in the Hamiltonian and density matrix were both less than . In both codes, the revPBE exchange-correlation functional was used [53]. The C09 [54, 55] van der Waals correction was used in the SIESTA calculations and the vdW-DFT-D3(BJ) [56] correction was used in ABINIT. In SIESTA, when an out-of-plane electric field was applied, a dipole correction [57, 58] was used in the vacuum region to prevent dipole-dipole interactions between periodic images. A dipole cutoff in slab-like systems has not been implemented in ABINIT, so although a vacuum space of 50 Å was used to separate the periodic images, the polarization is still slightly enhanced due to dipole-dipole interactions.
The top layer was translated along the unit cell diagonal over the bottom layer, which was held fixed. At each point a geometry relaxation was performed to obtain the equilibrium layer separation, while keeping the in-plane lattice vectors fixed. The out-of-plane and in-plane polarization were then obtained by calculating the Berry phases of the Bloch states. The data were fitted to Fourier expansions which respect the rotation symmetry of bilayer hBN. It was found that both the out-of-plane and in-plane polarization were well described by the first order in the expansions, i.e. Eqs. (1) and (2). At each point along the unit cell diagonal, DFT calculations were performed using ABINIT to calculate the dynamical charges.
Topological charge
The winding number, or topological charge, of the polarization field in configuration space is
which can mapped to real space in a strained or twisted bilayers as mentioned in the main text. Calculating the topological charge of a polarization field presents two additional complications when compared with magnetic fields. Firstly, the polarization is not of unit length, and must be normalized. Secondly, there are regions in space with zero polarization (modulo a quantum of polarization), i.e. at the AA stacking configurations, near which Eq. (3) diverges. This can be avoided by calculating the topological charge following the methodology in Ref. 59.
The polarization in the unit cell is discretized on a fine grid with spacing . A plaquette is constructed around each grid point, see Figure 4. The plaquettes form a grid which is offset from the original by half a grid spacing (a similar technique is used in first-principles calculations for more efficient Brillouin zone integrations [52]. The zeros in polarization at the AA stacking configurations are thus not included in the offset grid. The local topological charge can then be defined as
Figure 4. Winding with plaquettes. Sketch of the plaquette defined around each .
where is the signed area spanned by three points on a sphere:
The total charge is then
The total in the configuration space unit cell sums to zero, with precision of around even for relatively coarse grids. The winding numbers of the MPDs converge to for grid spacings below , see Supplementary Figure 5.

Supplementary Figure 5. Convergence of the total topological charge / winding number of the AB domain for a strained (solid line) and twisted (dashed line) bilayer as a function of grid spacing . The second y-axis shows the percentage error.
The meron–antimeron pair (anti)skyrmion transition driven by an applied field was identified by calculating the topological charge using fixed boundary conditions outside the moiré cell rather than periodic boundary conditions. The normalized polarization in the cell is the same, and the normalized polarization outside the cell is taken to be , where is an applied field in the out-of-plane direction. The winding in the interior of the cell is unaffected, but an additional winding is induced along the boundary when a field is applied, see Supplementary Fig. 6. The total winding along the boundary sums to . For , the boundary contributes an additional winding of each to the meron/antimeron, promoting one to a skyrmion/antiskyrmion, and making the other topologically trivial.
Acknowledgements
D. B. acknowledges funding from the University of Liège under special funds for research (IPD-STEMA fellowship programme) and St. John’s College, University of Cambridge. R. J. S and G. C. acknowledge funding from a New Investigator Award, EPSRC grant EP/W00187X/1. R. J. S also acknowledges funding from Trinity College, University of Cambridge. E. B. acknowledges the FNRS and the Excellence of Science program (EOS “ShapeME”, No. 40007525) funded by the FWO and F.R.S.-FNRS. Ph. G. acknowledges financial support from F.R.S.-FNRS Belgium (grant PROMOSPAN) and the European Union’s Horizon 2020 research and innovation program under grant agreement number 964931 (TSAR). The authors acknowledge the CECI supercomputer facilities funded by the F.R.S-FNRS (Grant No. 2.5020.1) and the Tier-1 supercomputer of the Fédération Wallonie-Bruxelles funded by the Walloon Region (Grant No. 1117545).
Author Contributions
D. B. initiated the project, performed the first-principles calculations, calculated the topological charge and wrote the paper. E. B. and P. G. developed the understanding of polarization and ferroelectricity with D. B., and provided guidance for the first-principles calculations. G. C. and R. J. S. developed the understanding of topology with D. B. All authors discussed the results and contributed to the writing of the paper.
Data Availability
The data presented in this study were generated using free and open source first-principles packages as described in the Methods section. The datasets generated during and/or analyzed during this study are available from the corresponding author upon request.
Code Availability
Code used to generate the plotted polarization structures is available from the corresponding author upon request.
Competing Interests
The authors declare no competing interests.
References
Measuring polarization in configuration space

Supplementary Figure 1. Sketch of how the configuration space mapping is used to estimate the local polarization in a twisted bilayer for small twist angles.
In this section we briefly describe how the configuration space mapping is used to estimate local properties in bilayers with small strains or twist angles. For a more detailed description, see Refs. 1 and 2. We illustrate the problem for the case of twist, considering two hexagonal monolayers (red and blue), each with a single atom in the unit cell, twisted at an angle with respect to one another, see Figure 1. Suppose we want to estimate the local properties in the middle unit cell, marked in black, using first-principles calculations. Using the configuration space mapping, the relative position of the red atom in this cell is , modulo any lattice vectors, where is the real space position. For small twist angles, we make the approximation that the local changes in environment around the black unit cell are small, and the local properties can be described by a commensurate bilayer with a relative translation :
see Figure 1. For a small homogeneous strain , the equivalent mapping is
This allows the local properties in strained or twisted bilayers to be parameterized with first-principles calculations using a single commensurate cell of a bilayer, and sliding one layer over the other. The total polarization, both out-of-plane and in-plane, can then be calculated from Berry phases:

Supplementary Figure 2. Sketch illustrating the effect of the mirror operation on (a)-(c): the out-of-plane and (d)-(f): the in-plane components of the polarization.
where is the volume of the bilayer, are the in-plane lattice vectors, and is the Berry phase. The sum is over all occupied bands . The Berry phase is routinely calculated with most widely-available first-principles density functional theory (DFT) codes:
where are the reciprocal lattice vectors and are the cell-periodic Bloch functions. Eq. (4) is essentially the position of the Wannier center, but normalized by the length of the in-plane lattice vectors and .
The local polarization in the main text was calculated by sliding one layer of hBN over the other, and at each point calculating the total polarization from Berry phases.
Of course, this method of calculating the local polarization in a twisted bilayer is not valid at all twist angles; at larger twist angles, the local changes in the environment cannot be assumed to be constant. We propose two more well-defined methods of measuring the local polarization:
By calculating the Wannier centers. The local polarization in each unit cell is then obtained by summing over the Wannier centers in that cell.
By integrating the dynamical charges.
In Section IV, we show that method 2 yields the exact same polarization obtained from Berry phases in configuration space. In principle, both of these methods should make it possible to directly calculate the local polarization in a twisted bilayer, although the calculations may be prohibitvely expensive.
Symmetry of the polarization field in bilayer hBN
Using the in-plane mirror symmetry of the AA stacking configuration, we can deduce the forms of the out-of-plane and in-plane polarization in 3R-stacked bilayer hBN.
For bilayer hBN with an AA stacking configuration, there is a mirror symmetry about the plane which is half-way between the two layers, and the out-of-plane polarization is zero. Suppose we translate the top layer by , generating an out-of-plane polarization . This is equivalent to translating the bottom layer by . If we now mirror the system about the plane between the two layers, the out-of-plane polarization is inverted. The same configuration can be achieved by sliding the top layer by . Therefore we can deduce that
i.e. the out-of-plane polarization is odd with respect to in-plane translations. This is illustrated in Figure 2 (a)-(Figure 2c).
Repeating this exercise with the in-plane polarization , we find that the in-plane polarization remains the same after applying the mirror symmetry, since it parallel to the mirror plane. Therefore we must have
i.e. the in-plane polarization is even with respect to in-plane translations. This is illustrated in Figure 2 (d)-(Figure 2f).
Polar Mode Phonon Frequencies in Bilayer hBN

Supplementary Figure 3. Lowest Gamma phonon frequencies in 3R-stacked bilayer hBN as a function of slide along the configuration space diagonal. The bands are shown in black when the phonons are stable, and red when they are unstable.
Dynamical Charges
In order to better understand the physical origin and nature of the in-plane and out-of-plane polarization, we calculated the dynamical charges[3] in bilayer hBN as a function of relative stacking:
i.e. the dipole generated in direction in response to the unitary displacement of atom in direction . Since the bilayer is a semi-periodic system, Eq. (13) is measured at zero electric field for the in-plane responses, and with open circuit boundary conditions for the out-of-plane responses. Typically, Eq. (13) is symmetric or has a negligible anti-symmetric part, and can be diagonalized. The effective charge of each atom can therefore be thought of as an ellipsoid, with lengths and orientations of the axes determined by the eigenvalues and eigenvectors[3]. The change in polarization from a reference configuration at (AA stacking) to a general configuration at can be measured by integrating the dynamical charges:

Figure S4. Total polarization in (a) the out-of-plane direction and (b) the in-plane direction, obtained by integrating the dynamical charges (solid), and from Berry phases (dashed). Individual contributions to the total polarization (black) in (c) the out-of-plane direction and (d) the in-plane direction, from the displacement of the B (blue) and N (red) atoms. Total change in dynamical charges (black) and the decomposition into out-of-plane (blue) and in-plane (red) displacements in (e) the out-of-plane direction and (f) the in-plane direction.
The dynamical charges were calculated for bilayer hBN along the diagonal in configuration space using density functional perturbation theory (DFPT) calculations in ABINIT. Integrating the effective charges results in a polarization almost in exact agreement with the polarization obtained by calculating the Berry phases, see Figure 4 (a) and (Figure 4b).
One advantage to calculating the polarization from the effective charges is that it allows the decomposition of the total polarization into contributions from the displacement of individual atoms in different directions. In Figure 4 (c) and (Figure 4d), the separate contributions to the polarization arising from the displacement of the B and N atoms in the top layer are shown. For the out-of-plane polarization, displacing the two atoms generates large, almost equal but opposite polarizations, which cancel to give the dipole generated by the transferred electronic charge. The individual contributions to the in-plane polarization have the same order of magnitude as the total in-plane polarization. Neither contribution by itself is even, but their sum is even. The total change of the effective charges in the top layer are shown in Figure 4 (e) and (Figure 4f), from which it is clear that the sum of the effective charges is the integrand of the total polarization. Notably, the polarization is almost entirely generated from the relative in-plane displacements. Out-of-plane displacements, i.e. the modulation of the interlayer separation, results in a higher-order contribution which is much smaller in magnitude.
Convergence of the topological charge
Winding along the cell boundaries for an applied electric field

Figure 6. Local winding along in a single moiré cell embedded in a dielectric medium for zero field (black) and for out-of-plane electric fields applied in the positive (red) and negative (blue) directions.
References
- [1]Li, L. & Wu, M. Binary compound bilayer and multilayer with vertical polarizations: two-dimensional ferroelectrics, multiferroics, and nanogenerators. ACS Nano 11, 6382–6388 (2017).DOI
- [2]Bennett, D. & Remez, B. On electrically tunable stacking domains and ferroelectricity in moiré superlattices. npj 2D Mater. Appl. 6, 1–6 (2022).DOI
- [3]Bennett, D. Theory of polar domains in moiré heterostructures. Phys. Rev. B 105, 235445 (2022).DOI
- [4]Stern, M. V. et al. Interfacial ferroelectricity by van der waals sliding. Science 372, 1462 (2021).
- [5]Yasuda, K., Wang, X., Watanabe, K., Taniguchi, T. & Jarillo-Herrero, P. Stacking-engineered ferroelectricity in bilayer boron nitride. Science 372, 1458 (2021).
- [6]Cao, Y. et al. Unconventional superconductivity in magic-angle graphene superlattices. Nature 556, 43 (2018).DOI
- [7]Cao, Y. et al. Correlated insulator behaviour at half-filling in magic-angle graphene superlattices. Nature 556, 80 (2018).
- [8]Zheng, Z. et al. Unconventional ferroelectricity in moiré heterostructures. Nature 588, 71 (2020).
- [9]Woods, C. et al. Charge-polarized interfacial superlattices in marginally twisted hexagonal boron nitride. Nat. Commun. 12, 1–7 (2021).DOI
- [10]King-Smith, R. & Vanderbilt, D. Theory of polarization of crystalline solids. Phys. Rev. B 47, 1651 (1993).pubmed.ncbi.nlm.nih.gov/10006189
- [11]Bristowe, N., Stengel, M., Littlewood, P., Artacho, E. & Pruneda, J. One-dimensional half-metallic interfaces of two-dimensional honeycomb insulators. Phys. Rev. B 88, 161411 (2013).
- [12]Das, S. et al. Observation of room-temperature polar skyrmions. Nature 568, 368 (2019).DOI
- [13]Han, L. et al. High-density switchable skyrmion-like polar nanodomains integrated on silicon. Nature 603, 603 (2022).DOI
- [14]Wang, Y. et al. Polar meron lattice in strained oxide ferroelectrics. Nat. Mater. 19, 881 (2020).DOI
- [15]Zubko, P. et al. Negative capacitance in multidomain ferroelectric superlattices. Nature 534, 524–528 (2016).DOI
- [16]Song, Z. et al. All magic angles in twisted bilayer graphene are topological. Phys. Rev. Lett. 123, 036401 (2019).DOI
- [17]Lu, X. et al. Multiple flat bands and topological hofstadter butterfly in twisted bilayer graphene close to the second magic angle. Proceedings of the National Academy of Sciences 118, e2100006118 (2021).DOI
- [18]Lian, B., Liu, Z., Zhang, Y. & Wang, J. Flat chern band from twisted bilayer MnBi₂ Te₄. Phys. Rev. Lett. 124, 126402 (2020).
- [19]Can, O. et al. High-temperature topological superconductivity in twisted double-layer copper oxides. Nat. Phys. 17, 519 (2021).DOI
- [20]Engelke, R. et al. Non-abelian topological defects and strain mapping in 2d moiré materials. arXiv:2207.05276 (2022).arxiv.org/abs/2207.05276
- [21]Guercı, D., Wang, J., Pixley, J. & Cano, J. Designer meron lattice on the surface of a topological insulator. arXiv:2203.04986 (2022).
- [22]Soler, J. M. et al. The SIESTA Method for Ab Initio Order-N Materials Simulation. J. Phys. Condens. Matter 14, 2745 (2002).DOI
- [23]Gonze, X. et al. Abinit: First-principles approach to material and nanosystem properties. Comput. Phys. Commun. 180, 2582 (2009).DOI
- [24]Carr, S. et al. Relaxation and domain formation in incommensurate two-dimensional heterostructures. Phys. Rev. B 98, 224102 (2018).DOI
- [25]Ghosez, P., Michenaud, J.-P. & Gonze, X. Dynamical atomic charges: The case of ABO₃ compounds. Phys. Rev. B 58, 6224 (1998).DOI
- [26]Urban, J. J., Yun, W. S., Gu, Q. & Park, H. Synthesis of single-crystalline perovskite nanorods composed of barium titanate and strontium titanate. J. Am. Chem. Soc. 124, 1186 (2002).pubmed.ncbi.nlm.nih.gov/11841284
- [27]Yun, W. S., Urban, J. J., Gu, Q. & Park, H. Ferroelectric properties of individual barium titanate nanowires investigated by scanned probe microscopy. Nano Lett. 2, 447 (2002).DOI
- [28]Luo, Y. et al. Nanoshell tubes of ferroelectric lead zirconate titanate and barium titanate. Appl. Phys. Lett. 83, 440 (2003).DOI
- [29]Mao, Y., Banerjee, S. & Wong, S. S. Hydrothermal synthesis of perovskite nanotubes. Chem Commun 408 (2003).pubmed.ncbi.nlm.nih.gov/12613636
- [30]Artyukhov, V. I., Gupta, S., Kutana, A. & Yakobson, B. I. Flexoelectricity and charge separation in carbon nanotubes. Nano Lett. 20, 3240 (2020).DOI
- [31]Springolo, M., Royo, M. & Stengel, M. Direct and converse flexoelectricity in two-dimensional materials. Phys. Rev. Lett. 127, 26801 (2021).DOI
- [32]Bennett, D. Flexoelectric-like radial polarization of single-walled nanotubes from first-principles. Electron. Struct. 3, 015001 (2021).DOI
- [33]Chu, M.-W. et al. Impact of misfit dislocations on the polarization instability of epitaxial nanostructured ferroelectric perovskites. Nat. Mater. 3, 87 (2004).pubmed.ncbi.nlm.nih.gov/14730357
- [34]Shin, H.-J. et al. Patterning of ferroelectric nanodot arrays using a silicon nitride shadow mask. Appl. Phys. Lett. 87, 113114 (2005).DOI
- [35]Fu, H. & Bellaiche, L. Ferroelectricity in barium titanate quantum dots and wires. Phys. Rev. Lett. 91, 257601 (2003).pubmed.ncbi.nlm.nih.gov/14754156
- [36]Naumov, I. I., Bellaiche, L. & Fu, H. Unusual phase transitions in ferroelectric nanodisks and nanorods. Nature 432, 737 (2004).
- [37]Geneste, G., Bousquet, E., Junquera, J. & Ghosez, P. Finite-size effects in BaTiO₃ nanowires. Appl. Phys. Lett. 88, 112906 (2006).DOI
- [38]Morozovska, A. N., Eliseev, E. A. & Glinchuk, M. D. Ferroelectricity enhancement in confined nanorods: Direct variational method. Phys. Rev. B 73, 214106 (2006).DOI
- [39]Hong, J., Catalan, G., Fang, D., Artacho, E. & Scott, J. Topology of the polarization field in ferroelectric nanowires from first principles. Phys. Rev. B 81, 172101 (2010).DOI
- [40]Nahas, Y. et al. Discovery of stable skyrmionic state in ferroelectric nanocomposites. Nat. Commun. 6, 1 (2015).DOI
- [41]Pereira Gonçalves, M. A., Escrihuela-Sayalero, C., Garca-Fernández, P., Junquera, J. & Íñiguez, J. Theoretical guidelines to create and tune electric skyrmion bubbles. Sci. Adv. 5, eaau7023 (2019).DOI
- [42]Qi, X.-L. & Zhang, S.-C. Topological insulators and superconductors. Rev. Mod. Phys. 83, 1057 (2011).DOI
- [43]Hasan, M. Z. & Kane, C. L. Colloquium: Topological insulators. Rev. Mod. Phys. 82, 3045 (2010).
- [44]Slager, R.-J., Mesaros, A., Juricic, V. & Zaanen, J. The space group classification of topological band-insulators. Nat. Phys. 9, 98 (2013).
- [45]Bernevig, B. A. Topological insulators and topological superconductors (Princeton university press, 2013).
- [46]Bouhon, A., Bzdušek, T. & Slager, R.-J. Geometric approach to fragile topology beyond symmetry indicators. Phys. Rev. B 102, 115135 (2020).DOI
- [47]Ünal, F. N., Bouhon, A. & Slager, R.-J. Topological euler class as a dynamical observable in optical lattices. Phys. Rev. Lett. 125, 053601 (2020).
- [48]García, A., Verstraete, M. J., Pouillon, Y. & Junquera, J. The psml format and library for norm-conserving pseudopotential data curation and interoperability. Comput. Phys. Commun. 227, 51 (2018).DOI
- [49]Hamann, D. Optimized norm-conserving Vanderbilt pseudopotentials. Phys. Rev. B 88, 085117 (2013).DOI
- [50]Van Setten, M. et al. The pseudodojo: Training and grading a 85 element optimized norm-conserving pseudopotential table. Comput. Phys. Commun. 226, 39 (2018).DOI
- [52]Monkhorst, H. J. & Pack, J. D. Special points for brillouin-zone integrations. Phys. Rev. B 13, 5188 (1976).DOI
- [53]Zhang, Y. & Yang, W. Comment on “generalized gradient approximation made simple”. Phys. Rev. Lett. 80, 890 (1998).DOI
- [54]Dion, M., Rydberg, H., Schröder, E., Langreth, D. & Lundqvist, B. Erratum: van der waals density functional for general geometries [physical review letters 92, 246401 (2004)]. Phys. Rev. Lett. 95, 109902 (2005).
- [55]Cooper, V. R. Van der waals density functional: An appropriate exchange functional. Phys. Rev. B 81, 161104 (2010).
- [56]Becke, A. D. & Johnson, E. R. A simple effective potential for exchange. J. Chem. Phys. 124, 221101 (2006).pubmed.ncbi.nlm.nih.gov/16784253
- [57]Neugebauer, J. & Scheffler, M. Adsorbate-substrate and adsorbate-adsorbate interactions of Na and K adlayers on Al (111). Phys. Rev. B 46, 16067 (1992).pubmed.ncbi.nlm.nih.gov/10003746
- [58]Bengtsson, L. Dipole correction for surface supercell calculations. Phys. Rev. B 59, 12301 (1999).DOI



