In this paper
AbstractAcknowledgementsMachine-Learned Interatomic Potential DetailsBibliographyDefinition of Structural Order ParametersComputational MethodsDETERMINATION OF ReferencesWorksPhysics and Astronomy
Engineering ferroelectric nanocapacitors through interfacial phonon bias
Abstract
As ferroelectric devices approach the nanometer scale, their behavior becomes increasingly sensitive to boundary conditions. In perovskite oxide nanocapacitors, epitaxial strain and electrostatic screening are well-established control parameters. Then, while expected to contribute to the epitaxial stabilization of ferroelectric ultrathin films, the lattice-dynamical state of the electrodes has not been isolated as an independent tuning knob. Here we call this poorly explored coupling interfacial phonon bias, whereby the electrodes' phonon energy landscape reshapes that of the adjacent ferroelectric across their shared interface. Using first-principles-accurate machine-learned interatomic potentials and molecular-dynamics simulations, we examine this mechanism in symmetric BiFeO nanocapacitors in contact with a metallic perovskite oxide. By independently controlling the orientatio and amplitude of the electrodes' octahedral-tilt landscape, while keeping conventional elastic and electric boundary conditions fixed, we show that the electrodes' phonon energy landscape can modify the stability of competing BiFeO phases. In particular, this allows us to shift the ferroelectric transition temperature and even induce antiferroelectric-like switching at ambient conditions. We thus showcase interfacial phonon bias as a complementary and powerful strategy to modify and optimize the properties of ultrathin epitaxial films.
Boundary conditions govern the behavior of ultrathin films of functional materials, ferroelectric oxides being a notable example. In thin films and nanocapacitors, the best-established control parameters for epitaxial stabilization are epitaxial strain, which reshapes the structural free-energy landscape through lattice mismatch, and electrostatic boundary conditions, which determine the screening of the spontaneous polarization and the associated depolarizing fields [1, 2, 3, 4, 5]. These two effects have been studied extensively and are now central to the design of ferroelectric heterostructures [2, 3]. However, epitaxial stabilization is not governed by elastic and electric boundary conditions alone: the structural connectivity across a coherent interface can also modify the relative stability of competing phases. Yet, the internal structure and lattice-dynamical features of the surrounding oxide environment have remained largely unexplored as an independent control knob.
The ability to independently control structural boundary conditions is especially valuable in ferroelectrics with competing polymorphs, where relatively small changes in the free-energy landscape can select polar, antipolar, or nonpolar states [6, 7, 8]. Such phase selection can determine, for example, whether a nanocapacitor exhibits conventional ferroelectric switching [9, 10], a field-induced transition from an antipolar to a polar state [11, 12, 13], or a more complex multistep polarization reversal [14, 15]. It can therefore modify key functional properties, including the switching barrier and coercive field [9, 10], dielectric and magnetoelectric responses [16, 17], and energy-storage behavior [18, 19]. A structural control parameter that operates through the electrodes’ internal structure and phonon energy landscape, rather than through elastic mismatch or electrostatic screening, thus provides an additional route to engineer phase stability and switching behavior in nanoscale ferroelectric devices.
This mechanism can be especially important in heterostructures of structurally compatible materials, where structural distortions can couple across the interface [20, 21, 22, 23]. In perovskite oxides, for example, the corner-sharing oxygen-octahedral network provides a direct link between the structures of adjacent layers [24, 25, 26]. The importance of this connectivity is illustrated by the pronounced growth-orientation dependence of the metal-to-insulator transition temperature in films on [27, 28], and by the suppression of oxygen-octahedral breathing distortions reported in -based heterostructures [29, 30]. In these systems, however, interfacial structural coupling is generally intertwined with epitaxial strain, confinement, interface chemistry, and electrostatic effects, making its individual contribution difficult to isolate. Here, we use predictive atomistic simulations to isolate and explore this mechanism, which we call interfacial phonon bias (see Figure 1 for a schematic representation). Our results show it provides us with a powerful knob to engineer the behavior of ferroelectric nanocapacitors.

Figure 1. Concept of interfacial phonon bias and symmetric nanocapacitor architecture. Solid black and white dashed lines represent the intrinsic phonon energy landscape and the biased one, respectively. Electrode: the octahedral-tilt energy landscape is modified to favor a selected structural state. Interface: the shared interface (e.g., oxygen-octahedral network in oxide perovskites) provides the pathway for structural coupling across layers. Ferroelectric: the modified electrodes’ landscape reshapes the free energy and phase competition of the nanocapacitor.
This effect may be particularly relevant for materials like , where polarization is strongly coupled to oxygen-octahedral rotations [8, 15, 31], so that changes in the tilts’ free-energy landscape are expected to substantially affect the behavior of the ferroelectric polarization. Consistently, recent experimental and theoretical work has shown that switching in cannot be understood solely in terms of polarization and strain: the evolution of octahedral tilts during reversal is essential to the switching barrier and kinetics [10, 32, 33]. The competition between the polar and antipolar phases in BiFeO has also attracted considerable interest as a possible route to achieve antiferroelectric-like behavior and energy-storage functionality [8, 12, 34, 35]. Controlling this – phase competition through the interfacial phonon bias would therefore offer an alternative to the strain-, electrostatic- and composition-based approaches commonly used to stabilize competing nonpolar or antipolar phases [36, 37, 38].
In this work, we isolate the interfacial-phonon effect by design. We choose LaNiO perovskite electrodes whose interfaces with BiFeO do not involve any nominal-charge discontinuity that could obscure the structural physics of interest. We work with symmetric BiFeO nanocapacitors, considering the NiO/BiO composition for both interfaces and in all calculations, so that the electric boundary conditions are kept constant. The in-plane lattice parameter is fixed to that of a canonical substrate, i.e. SrTiO oriented along the [1] pseudocubic direction, to impose the epitaxial constraint. (See Figure 1 and Supplemental for full structural details.) We then tune the electrodes’ structural state by controlling the orientation, amplitude, and stiffness of its octahedral tilts.
We track the antiphase and in-phase octahedral-tilt amplitudes about the pseudocubic axes by and , respectively (). We quantify these and relevant antipolar distortions [39] ( being the direction of the anti-phase modulation) by projecting atomic displacements onto symmetry-adapted modes of the cubic perovskite reference structure. We also track an approximate polarization computed using nominal charges. Full definitions and normalization conventions are provided in the Supplemental.
To isolate the role of the electrodes’ tilt energetics, we require a way to vary its tilt pattern and stiffness without changing its chemistry or the epitaxial and electrostatic boundary conditions. A so-called harmonic bias provides this control while allowing the distortions of the electrodes and the ferroelectric to relax together. Following similar approaches found in the literature [40, 41], we add a harmonic bias term to the electrodes’ tilt degrees of freedom of the form
where , are the target amplitudes, and is the stiffness (see Supplemental for full details on our implementation of the harmonic bias).
To simulate these nanocapacitors at realistic sizes (thousands of atoms) and over timescales beyond the reach of direct first-principles methods (few to tens of nanoseconds), we employ a machine-learned interatomic potential based on the neuroevolution potential formalism [42], trained on first-principles data for BiFeO/LaNiO heterostructures and individual bulks (see Supplemental for validation of the model).
We engineer our model electrodes in the following way. First, we flatten the native LaNiO tilt energy landscape: we add a harmonic compensating term that counteracts the intrinsic instability of the LaNiO’s antiphase tilts, bringing the electrodes close to the threshold between untilted and tilted configurations. We estimate a critical compensating stiffness of eV/(Årad), consistent in magnitude with the quadratic coefficient obtained from Density Functional Theory (DFT) for the Landau model description of bulk LaNiO eV/(Årad) [39].
A second harmonic term then defines a preferred finite-amplitude tilt pattern in our model electrodes. We select the tilt configuration in Glazer’s notation [43]. Specifically, we use the structure of bulk SrRuO (SRO) as a relevant reference, preserving the relative amplitudes of the antiphase and in-phase rotations while varying their overall magnitude. The full bias term thus reads
where , are a scaling factor controlling the electrodes’ target tilt amplitudes and target stiffness, while are the equilibrium tilt amplitudes of bulk SRO (i.e., and , all other tilt components being zero). This is a particularly relevant octahedral tilt pattern because it corresponds to the most common one found in perovskites [39, 44], and in particular to the antipolar phase of . It is therefore very common among potential perovskite electrodes and well suited to bias the competition between the ferroelectric and the antipolar phases.
We start our analysis by computing the energy difference between the and polymorphs for the nanocapacitor, , which is a proxy for the transition temperature [34, 45], as a function of the scaling factor and stiffness . Figure 2 shows the resulting landscape, where we mark points at which the electrodes exhibit the same in-phase tilt amplitude after structural relaxation of the entire nanocapacitor (the antiphase tilts follow the same trend). The target amplitudes define the preferred electrode tilt pattern, while the stiffness sets the energetic penalty for deviations from it, as given in Eq. (2). The relaxed electrode amplitudes are therefore not imposed directly, but emerge from the balance between this bias, the intrinsic electrode energy landscape, and the interfacial coupling to the layer. The lines of constant are approximately parallel to those of constant , indicating that the realized average electrode tilt amplitude is the key structural variable controlling the relative stability of ’s polymorphs.
Figure 2. Energy difference between the and polymorphs of a 3.2 nm nanocapacitor as a function of the stiffness () and target amplitude () defining the harmonic bias applied to the electrodes. The scattered points mark regions where the resulting electrode tilt amplitude () is constant. The green dashed line marks the cases explored in Figure 4.
We then use molecular-dynamics (MD) simulations with a linear heating protocol to locate the phase transition temperature. We fix eV/(Å rad) for simplicity [46] and vary the target tilt amplitude.
Starting from the phase of in the orientation such that the polarization points in the direction and the antiphase tilts are parallel to it, we first thermalize the system for 20 ps at 1 K and, subsequently, heat it at a rate of K/fs until the transition to occurs. Comparison with equilibrium simulations (see Supplementary) shows that this protocol provides sufficiently converged transition temperatures.
As a reference calculation, we first simulate bulk under the epitaxial constraint imposed by a (001)-oriented substrate (1% compressive epitaxial strain). We use an 8 × 8 × 8 supercell, where the 1 × 1 × 1 cell corresponds to one perovskite formula unit. The -to- transition is identified at approximately 850 K. The in-phase tilt develops in-plane, which reflects the symmetry breaking due to the epitaxial strain.
We next examine how the biased electrode modifies this transition in symmetric nanocapacitors with ferroelectric thicknesses of 1.6 nm, 3.2 nm, and 6.4 nm. The simulated electrode is always 3.2 nm thick, and the considered supercell comprises perovskite units in plane. We vary , producing electrode in-phase tilts spanning –.
Figure 3 shows the temperature evolution of the structural order parameters for two contrasting cases: a cubic electrode and an electrode whose realized tilt amplitudes are close to those of bulk SrRuO. In both cases the transition to is marked by the emergence of in-phase tilts and antipolar distortions, together with the loss of polarization. For the cubic electrode, the two in-plane orientations for in-phase tilts are symmetry equivalent; the tilt develops along in the simulation shown. By contrast, the tilted electrode selects the orientation of the in-phase tilt in the BiFeO layer, which develops parallel to that in the electrode. Thus, the electrode provides an orientational preference in addition to modifying the relative stability of the competing phases.

Figure 3. Temperature-driven phase transition of 3.2 nm nanocapacitors, with in-plane strain fixed to that of a (001)-oriented substrate, for eV/(Årad). a) (final , cubic electrode) and b) (final , close to bulk ’s tilts). The abrupt change from the polymorph (antiphase tilts and polarization fully developed) to the polymorph (one antiphase tilt becomes in-phase, polarization is lost and antipolar mode develops) marks the transition temperature.
Figure 4 shows the computed -to- transition temperature as a function of the realized electrode tilt amplitude for the three ferroelectric thicknesses here considered. In the limit of vanishing electrode tilt amplitude, the transition temperature is close to the value obtained for strained bulk BiFeO. (The difference likely reflects the realistic treatment of screening in the nanocapacitor simulations, which penalizes the phase compared to the ideal short-circuit conditions assumed in the calculations of pure BiFeO.) Increasing the electrode tilt amplitude systematically lowers the transition temperature, consistent with an enhanced preference for the electrode-favored tilt pattern. This reduction is more pronounced for thinner ferroelectric layers, which have a larger interface-to-volume ratio, supporting the interfacial origin of the effect. At sufficiently large tilt amplitudes, the transition occurs below room temperature. In particular, for amplitudes comparable to those
Figure 4. Computed to transition temperatures for BiFeO nanocapacitors of different thicknesses, as a function of electrode tilt amplitude ( orientation).
of bulk SrRuO (–), all nanocapacitors considered here stabilize in the phase at room temperature.
The same interfacial bias that stabilizes the phase at room temperature may also reshape functional properties such as electric-field-driven transitions. We thus investigate how the electrodes’ phonon energy landscape affects field-driven switching in the 3.2 nm BiFeO nanocapacitor. This thickness is selected as a representative intermediate case, for which the interfacial phonon bias remains substantial while the ferroelectric layer is sufficiently thick to be experimentally convenient. We focus on three final electrode tilt amplitudes, , , and .
To probe the resulting switching loops, we apply a sinusoidal electric field along the vertical stacking direction, with a maximum amplitude of 10 MV/cm and a frequency of 1 GHz, over two full cycles. The field is incorporated as effective forces exerted on nominally charged ions in the ferroelectric layer ( for Bi and Fe, for O). The short time scales accessible to MD motivate the high driving frequency and the use of large field amplitudes: field-driven switching can involve thermally activated nucleation, and stronger fields lower the relevant free-energy barriers, making these events accessible within the simulated cycles. These calculations therefore aim to understand qualitatively how the structure changes under an applied electric field and how these changes depend on interfacial phonon bias, rather than to predict experimental coercive fields quantitatively. The resulting Polarization-Electric field (–) loops are shown in Figure 5.
Figure 5. Calculated out-of-plane polarization–field loops for the 3.2 nm BiFeO nanocapacitor with realized electrode tilt amplitudes , , and at 300 K.
In the limit of zero electrode tilts (), the nanocapacitor displays the usual ferroelectric loop. For the largest , the calculated – curve displays the characteristic double-loop shape of antiferroelectric switching. At zero field, the nanocapacitor stabilizes in the antipolar phase, whereas a sufficiently large applied field induces the polar phase. The reproducibility of this response over the simulated cycles indicates a reversible field-driven transition between these structural phases. For the intermediate electrode tilt amplitude, the two lobes of the double loop become connected. In this case, the field-driven switching pathway allows the nanocapacitor to transform from the polar phase to the antipolar phase at comparatively small fields. Once the nanocapacitor switches to , reducing the field to zero, rather than continuing the cycle toward the next transition, allows the antipolar phase to be retained at zero field. This behavior therefore suggests that, for electrodes with intermediate tilt amplitudes, interfacial phonon bias can provide access to a metastable antipolar zero-field state, potentially enabling multistate functionality.
Interfacial phonon bias thus endows a structural dimension to epitaxial stabilization beyond elastic and electric boundary conditions. While epitaxial strain modifies phase stability by constraining the average in-plane lattice parameters, coherent interfaces impose further structural boundary conditions through the connectivity of the atomic network. We expect this mechanism to be most effective in heterostructures whose structural networks remain connected across the interface, allowing distortions in one material to influence those in the other. Perovskite oxides provide a natural platform because their corner-sharing oxygen octahedra couple tilt patterns across interfaces. Moreover, the strong (competitive) coupling between polarization and octahedral tilts in most perovskites [47, 48, 49] means that this interfacial influence can extend directly to polar stability and switching, making the mechanism relevant well beyond BiFeO. For controlling ferroelectric phase stability, particularly promising systems are those in which polar and antipolar or nonpolar phases compete and respond differently to the transmitted distortion. Beyond perovskites, we expect the mechanism to apply whenever interfacial bonding couples compatible distortions in adjacent materials.
Interface orientation provides an additional handle on this interfacial coupling. For example, NdNiO films grown on NdGaO exhibit a metal-to-insulator transition near 200 K for the (001) orientation, compared with 335 K for the (111) orientation [27, 28]. This orientation dependence has been attributed to substrate-imposed octahedral tilts, illustrating how interfacial connectivity can substantially modify a phase-transition temperature. (Each interfacial octahedron shares one oxygen across a (001) interface but three across a (111) interface, imposing stronger constraints on the tilt pattern in the latter.) Further control could be achieved by tuning the phonon energy landscape of the substrate.
In practice, electrode composition, interface orientation, epitaxial strain, layer thickness, and temperature offer distinct routes to tune interfacial phonon bias. These parameters can modify the preferred electrode distortion, its stiffness, and the extent of its influence on the ferroelectric layer. In real heterostructures, such changes will generally interplay with other boundary conditions. The bias-potential approach used here provides a controlled model to isolate the impact of the electrodes’ internal energy landscape while elastic, electric, and chemical constraints remain the same. This separation can guide the choice of electrode materials and interface geometries that favor a desired ferroelectric phase or switching response. We hope that the striking behaviors revealed and explained here (including the ability to induce distinct electrical responses in BiFeO nanocapacitors at ambient conditions) will stimulate further theoretical and experimental exploration of this general design strategy.
Acknowledgements
Work supported by the Luxembourg National Research Fund (FNR) through Grant No. BRIDGES/18421428/SWITCHON. U.D. also acknowledges FNR’s support through Grant No. C23/MS/17909853/BUBBLACED. Work supported in part by Intel.
Machine-Learned Interatomic Potential Details
We construct the nanocapacitor machine-learned interatomic potential (MLIP) using the neuroevolution potential architecture as implemented in GPUMD [51]. We gather the DFT data required for training (crystal structures with energies, forces and stresses) using VASP’s on-the-fly algorithm. The usual approach consists on running ab initio molecular dynamics (AIMD) of the structures to be trained on. In our case, the required size of the nanocapacitors makes it impossible to run AIMD of the full nanocapacitor. We go around this issue by “breaking up” the nanocapacitor in parts. As the NEP model captures the energetics based on the environment of each atom, we can decompose these environments into three: bulk , bulk and interface (see Figure 6a). These cells are small enough that we can run AIMD on them and span the local environments of atoms in any nanocapacitor size. The bulk environment training was done using a perovskite supercell with AIMD at 300 K starting from the relaxed and polymorphs of and polymorph of , and it was extended with a high temperature training for bulk using a temperature ramp from 300 K to 1000 K. The training on the interface environment was done at 300 K, starting from the relaxed structure of two different orientations of with the polarization along the (refer to as the configuration) and along , with the anti-phase tilts () parallel to the polarization. We extended the training of the nanocapacitor by training in the presence of an effective constant negative electric field, starting from the configuration, with amplitudes 2, 4 and 6 MV/cm. In this way, we can explore the potential well away from the minimum towards switching. We gather a total of 2426 structures which we use to train the neuroevolution potential (NEP) MLIP (training and validation data are available in Zenodo [52]), split into 1436 of bulk 75 of bulk and 915 of the nanocapacitor.
Figure 6. a) Illustration of nanocapacitor architecture and training structures emulating the different environments (crystal structures figures obtained using VESTA software [50]). b) Histogram of errors in energies, forces and stresses on the validation set.
To validate the model, we extract 100 equispaced snapshots from all the different AIMD simulations and run single-point DFT calculations. These calculations are not used to train the model, thus they serve as a measure of the goodness of the model to interpolate. We show the histogram for the errors in the prediction of energies, forces and stresses in Figure 6b. The root mean squared errors (RMSE) for the validation set are 5.4 meV/atom for the energies, 64 meV/Å for the forces and 4.2 kbar for the stresses, well within the usual errors reported in the literature.
We are mainly interested in the structural properties as a function of temperature. We thus further test the model using bulk for benchmarking purposes as described in the main text. Notice that the transition to the -like polymorph happens with the in-phase tilt axis in-plane. This implies that this configuration has lower energy than the one with in-phase tilt out-of-plane. We confirm that this is the case by comparing the relative energy of both configurations after a full relaxation. The MLIP predicts this energy difference to be 16 meV/fu, while DFT predicts 18 meV/fu. Not having trained on any constrained configuration, this further validates that our model performs well at extrapolating. From these tests we conclude that the MLIP for the nanocapacitors is accurate both for interpolation and extrapolation.
In this work we opted for symmetric nanocapacitors. First, by symmetric we refer to nanocapacitors where both interfaces are chemically equivalent. A usual way of understanding the perovskite structure with chemical formula is as a stacking of alternating AO and layers. Constructing a superstructure mixing two different perovskites, and , results in two interfaces, and , which are usually different. This is the source of interesting phenomena like that of 2 dimensional electron gases (2DEGs) at the interface of differently nominally charged layers as in, e.g., superstructures. Even in systems where there is no charge discontinuity in neighboring layers, this asymmetry always introduces a built-in field which makes nanocapacitor states with polarization up and down (in the direction orthogonal to the interface) different, making it harder to extract what may be subtle energy differences due to other effects, such as the ones we are focusing on in this work. A symmetric interface would be one where we add one more layer of one of the perovskites, breaking stoichiometry, but removing the built-in fields from interface asymmetry. In what follows, we symmetrize the interface by adding one extra BiO layer, so that our symmetric interfaces are .
Bibliography
Definition of Structural Order Parameters
The structural order parameters are extracted from each MD snapshot by projecting atomic displacements onto a set of symmetry-adapted distortion modes defined for the perovskite reference unit cell. For each snapshot, the fractional displacement of every atom from its reference position is computed as . We use the usual cubic perovskite positions as reference. The six octahedral tilt amplitudes — three R-point (, , ) and three M-point (, , ) —, and the 6 antipolar modes are obtained as inner products of the displacement field with the corresponding symmetry-mode vectors. Since antipolar modes modulated along X, Y and Z-points are degenerate, we plot the amplitude of each pair of modes. This amplitude is then transformed to physical units (C/m) using the average displacement per Bi atom and nominal charges. The polarization “proxy” is computed as
where are isotropic nominal charges ( for Bi, Fe, for O), is the number of formula units, and is a conversion factor to physical units (C/m). Even though this is a rough approximation, it is known to work well for BiFeO.
Computational Methods
All first-principles calculations were performed using the projector augmented-wave (PAW) method [53] as implemented in VASP [54, 55], with the PBEsol exchange-correlation functional [56]. We solve explicitly for 6 electrons of O (), 15 of Bi (), 14 of Fe (), 16 of Ni () and 11 of La (). The plane-wave energy cutoff was set to 500 eV in all cases. We use a -point grid of for bulk BiFeO, for bulk LaNiO, and for the nanocapacitors. We add a Hubbard correction of to the Fe atoms as usual for BiFeO magnetic calculations, and the magnetic state of BiFeO is always converged to the G-type antiferromagnetic state.
The molecular dynamics simulations were performed using GPUMD software. We use the Martyna-Tuckerman-Tobias-Klein ensemble [57], with a Nosé-Hoover thermostat [58] and a Parrinello-Rahman barostat [59]. We modify the GPUMD code to fix the in-plane strain and allow for the third lattice vector to move according to the barostat. We use a timestep of 1 fs and a constant temperature ramp of 0-1000 K during 10 ns, which represents a heating of K/fs. We set the thermostat coupling to default and the barostat period to 10000 fs.
The phase transition we are targeting is between BiFeO and polymorphs. The polymorph has a polarization in the direction, with anti-phase tilts parallel. The phase has anti-phase tilts in the direction, in-phase in the and an antipolar distortion modulated parallel to the in-phase tilts, with no net polarization. Since it is a first order phase transition, we can easily capture the transition temperature by computing a running average of the polarization and finding the abrupt change from finite average polarization to 0. We checked several values for the running average window and threshold for the transition temperature determination, with minimal impact on the numerical value of the transition temperature itself.
One may argue that our MD simulations are never at equilibrium, due to the slow heating up protocol we follow to capture the transition temperature. On the one hand, this method should only overestimate the transition temperature due to not spending enough time at a particular temperature. In that sense, our results serve as a higher bound for the transition temperature. We run a further test comparing two MD simulations, a usual fixed-target-temperature at 300 K run for 10 ns (see Figure 7a and Figure 7b) and our temperature ramp protocol (see Figure 7c and Figure 7d). From the constant temperature run, we extract the average () and standard deviation () of temperatures. We use them to extract a window of data from the ramp protocol from to . We compare the resulting histograms in Figure 7. We can see that the temperature histogram for the temperature ramp protocol is basically flat due to the constant heating and the polarization histograms are very similar, with an error of less than 0.03% in the average value.
Figure 7. Comparison between usual equilibrium MD methods and the temperature ramp approach. a) Polarization histograms for the constant temperature run. b) Temperature histogram for the constant temperature run. c) Polarization histogram for a similar temperature window in the temperature ramp protocol. Temperature histogram of the same temperature window as in c). d) Temperature histogram for the temperature ramp approach.
DETERMINATION OF
We determine the amplitude of harmonic bias required to “flatten” the energy landscape of by running MD simulations at 10 K on a supercell with a time step of 1 fs, and we increase the tilt bias by 0.5 every 10 ps (see Figure 8). We can see clearly the transition between the native tetragonal structure () and the target cubic structure (). We thus choose .
Figure 8. R tilt average amplitude and standard deviation for different tilt bias amplitudes. We can see how the transition between tetragonal and cubic lies around .
References
- [1]N. A. Pertsev, A. G. Zembilgotov, and A. K. Tagantsev, “Effect of mechanical boundary conditions on phase diagrams of epitaxial ferroelectric thin films,” Physical Review Letters 80, 1988–1991 (1998).DOI
- [2]Darrell G. Schlom, Long-Qing Chen, Chang-Beom Eom, Karin M. Rabe, Stephen K Streiffer, and Jean-Marc Triscone, “Strain tuning of ferroelectric thin films,” Annual Review of Materials Research 37, 589 (2007).DOI
- [3]M. Dawber, K. M. Rabe, and J. F. Scott, “Physics of thin-film ferroelectric oxides,” Rev. Mod. Phys. 77, 1083–1130 (2005).DOI
- [4]Massimiliano Stengel and Nicola A. Spaldin, “Origin of the dielectric dead layer in nanoscale capacitors,” Nature 443, 679–682 (2006).pubmed.ncbi.nlm.nih.gov/17036000
- [5]Céline Lichtensteiger, Stéphanie Fernandez-Pena, Christian Weymann, Pavlo Zubko, and Jean-Marc Triscone, “Tuning of the depolarization field and nanodomain structure in ferroelectric thin films,” Nano Letters 14, 4205–4211 (2014).DOI
- [6]Karin M. Rabe, “Antiferroelectricity in oxides: A reexamination,” in Functional Metal Oxides (John Wiley & Sons, Ltd, 2013) Chap. 7, pp. 221–244.DOI
- [7]Akansha Singh, Vivek N. Singh, Enric Canadell, Jorge Íñiguez, and Oswaldo Diéguez, “Polymorphism in bi-based perovskite oxides: A first-principles study,” Phys. Rev. Materials 2, 104417 (2018).DOI
- [8]Oswaldo Diéguez, O. E. González-Vázquez, Jacek C. Wojdeł, and Jorge Íñiguez, “First-principles predictions of low-energy phases of multiferroic BiFeO₃,” Physical Review B 83, 094105 (2011).
- [9]Shi Liu, Ilya Grinberg, and Andrew M. Rappe, “Intrinsic ferroelectric switching from first principles,” Nature 534, 360–363 (2016).DOI
- [10]Qiwu Shi, Eric Parsonnet, Xiaoxing Cheng, Natalya Fedorova, Ren-Ci Peng, Abel Fernandez, Alexander Qualls, Xiaoxi Huang, Xue Chang, Hongrui Zhang, David Pesquera, Sujit Das, Dmitri Nikonov, Ian Young, Long-Qing Chen, Lane W. Martin, Yen-Lin Huang, Jorge Íñiguez, and Ramamoorthy Ramesh, “The role of lattice dynamics in ferroelectric switching,” Nature Communications 13, 1110 (2022).DOI
- [11]C. Kittel, “Theory of antiferroelectric crystals,” Phys. Rev. 82, 729–732 (1951).DOI
- [12]M. Graf, M. Sepliarsky, S. Tinte, and M. G. Stachiotti, “Phase transitions and antiferroelectricity in BiFeO₃ from atomic-level simulations,” Phys. Rev. B 90, 184108 (2014).DOI
- [13]Haidong Lu, Sebastian Glinsek, Pratyush Buragohain, Emmanuel Defay, Jorge Íñiguez, and Alexei Gruverman, “Probing antiferroelectric-ferroelectric phase transitions in PbZrO₃ capacitors by piezoresponse force microscopy,” Advanced Functional Materials 30, 2003622 (2020).DOI
- [14]Massimiliano Stengel and Jorge Íñiguez, “Electrical phase diagram of bulk BiFeO₃,” Physical Review B 92, 235148 (2015).DOI
- [15]Natalya S. Fedorova, Dmitri E. Nikonov, Hai Li, Ian A. Young, and Jorge Íñiguez, “First-principles landau-like potential for bifeo₃ and related materials,” Phys. Rev. B 106, 165122 (2022).DOI
- [16]Manfred Fiebig, Thomas Lottermoser, Dennis Meier, and Morgan Trassin, “The evolution of multiferroics,” Nature Reviews Materials 1, 16046 (2016).DOI
- [17]Jacek C. Wojdel and Jorge Íñiguez, “Magnetoelectric response of multiferroic BiFeO₃ and related materials from first-principles calculations,” Physical Review Letters 103, 267205 (2009).pubmed.ncbi.nlm.nih.gov/20366342
- [18]Bin Xu, Jorge Íñiguez, and L. Bellaiche, “Designing lead-free antiferroelectrics for energy storage,” Nature Communications 8, 15682 (2017).
- [19]Hugo Aramberri, Natalya S. Fedorova, and Jorge Íñiguez, “Ferroelectric/paraelelectric superlattices for energy storage,” Science Advances 8, eabn4880 (2022).DOI
- [20]A. Y. Borisevich, H. J. Chang, M. Huijben, M. P. Oxley, S. Okamoto, M. K. Niranjan, J. D. Burton, E. Y. Tsymbal, Y. H. Chu, P. Yu, R. Ramesh, S. V. Kalinin, and S. J. Pennycook, “Suppression of octahedral tilts and associated changes in electronic properties at epitaxial oxide heterostructure interfaces,” Phys. Rev. Lett. 105, 087204 (2010).pubmed.ncbi.nlm.nih.gov/20868130
- [21]Wenlai Lu, Wendong Song, Ping Yang, Jun Ding, Gan Moog Chow, and Jingsheng Chen, “Strain engineering of octahedral rotations and physical properties of srruo3 films,” Scientific Reports 5, 10245 (2015).DOI
- [22]James M. Rondinelli and Nicola A. Spaldin, “Substrate coherency driven octahedral rotations in perovskite oxide films,” Phys. Rev. B 82, 113402 (2010).DOI
- [23]Jun He, Albina Borisevich, Sergei V. Kalinin, Stephen J. Pennycook, and Sokrates T. Pantelides, “Control of octahedral tilts and magnetic properties of perovskite oxide heterostructures by substrate symmetry,” Phys. Rev. Lett. 105, 227203 (2010).pubmed.ncbi.nlm.nih.gov/21231419
- [24]P Zubko, N Jecklin, A Torres-Pardo, P Aguado-Puente, A Gloter, C Lichtensteiger, J Junquera, O Stéphan, and J M Triscone, “Electrostatic Coupling and Local Structural Distortions at Interfaces in Ferroelectric/Paraelectric Superlattices,” Nano Letters 12, 2846–2851 (2012).DOI
- [25]Eric Bousquet, Matthew Dawber, Nicolas Stucki, Celine Lichtensteiger, Patrick Hermet, Stefano Gariglio, Jean-Marc Triscone, and Philippe Ghosez, “Improper ferroelectricity in perovskite oxide artificial superlattices,” Nature 452, 732–736 (2008).
- [26]James M. Rondinelli, Steven J. May, and John W. Freeland, “Control of octahedral connectivity in perovskite oxide heterostructures: An emerging route to multifunctional materials discovery,” MRS Bulletin 37, 261–270 (2012).DOI
- [27]S Catalano, M Gibert, J Fowlie, J Íñiguez, J-M Triscone, and J Kreisel, “Rare-earth nickelates rnio3: thin films and heterostructures,” Reports on Progress in Physics 81, 046501 (2018).DOI
- [28]S. Catalano, M. Gibert, V. Bisogni, F. He, R. Sutarto, M. Viret, P. Zubko, R. Schierholz, G. A. Sawatzky, T. Schmitt, and J.-M. Triscone, “Tailoring the electronic transitions of ndnio3 films through (111)pc oriented interfaces,” APL Materials 3, 062506 (2015).DOI
- [29]Kei Inumaru, Hajime Miyata, and Shoji Yamanaka, “Partial suppression of structural distortion in epitaxially grown babio3 thin films,” Phys. Rev. B 78, 132507 (2008).DOI
- [30]Han Gyeol Lee, Rokyeon Kim, Jinkwon Kim, Minu Kim, Tae Heon Kim, Shinhubm Lee, and Tae Won Noh, “Anisotropic suppression of octahedral breathing distortion with the fully strained odabio3/baceo3 heterointerface,” APL Materials 6, 016107 (2018).DOI
- [31]Teng Gu, Timothy Scarbrough, Yurong Yang, Jorge Íñiguez, L. Bellaiche, and H. J. Xiang, “Cooperative Couplings between Octahedral Rotations and Ferroelectricity in Perovskites and Related Materials,” Physical Review Letters 120, 197602 (2018).DOI
- [32]J.T. Heron, J. L. Bosse, Q. He, Y. Gao, M. Trassin, L. Ye, J. D. Clarkson, C. Wang, Jian Liu, S. Salahuddin, D. C. Ralph, D. G. Schlom, J. Íñiguez, B.D. Huey, and R. Ramesh, “Deterministic switching of ferromagnetism at room temperature using an electric field,” Nature 516, pages 370–373 (2014).DOI
- [33]Natalya S. Fedorova, Dmitri E. Nikonov, John M. Mangeri, Hai Li, Ian A. Young, and Jorge Íñiguez, “Understanding magnetoelectric switching in BiFeO₃ thin films,” Phys. Rev. B 109, 085116 (2024).DOI
- [34]Claudio Cazorla and Jorge Íñiguez, “Insights into the phase diagram of bismuth ferrite from quasiharmonic free-energy calculations,” Physical Review B 88, 214430 (2013).DOI
- [35]Julia A. Mundy, Bastien F. Grosso, Colin A. Heikes, Dan Ferenc Segedin, Zhe Wang, Yu-Tsun Shao, Cheng Dai, Berit H. Goodge, Quintin N. Meier, Christopher T. Nelson, Bhagwati Prasad, Fei Xue, Steffen Ganschow, David A. Muller, Lena F. Kourkoutis, Long-Qing Chen, William D. Ratcliff, Nicola A. Spaldin, Ramamoorthy Ramesh, and Darrell G. Schlom, “Liberating a hidden antiferroelectric phase with interfacial electrostatic engineering,” Science Advances 8, eabg5860 (2022), https://www.science.org/doi/pdf/10.1126/sciadv.abg5860.DOI
- [36]I. C. Infante, S. Lisenkov, B. Dupé, M. Bibes, S. Fusil, E. Jacquet, G. Geneste, S. Petit, A. Courtial, J. Juraszek, L. Bellaiche, A. Barthélémy, and B. Dkhil, “Bridging multiferroic phase transitions by epitaxial strain in BiFeO3,” Physical Review Letters 105, 057601 (2010).
- [37]Daisuke Kan, Lucia Pálová, Varatharajan Anbusathaiah, Ching Jung Cheng, Shigehiro Fujino, Valanoor Nagarajan, Karin M. Rabe, and Ichiro Takeuchi, “Universal behavior and electric-field-induced structural transition in rare-earth-substituted BiFeO3,” Advanced Functional Materials 20, 1108–1115 (2010).DOI
- [38]O. E. González-Vázquez, Jacek C. Wojdeł, Oswaldo Diéguez, and Jorge Íñiguez, “First-principles investigation of the structural phases and enhanced response properties of the BiFeO3-LaFeO3 multiferroic solid solution,” Physical Review B 85, 064119 (2012).
- [39]Peng Chen, Mathieu N. Grisolia, Hong Jian Zhao, Otto E. González-Vázquez, L. Bellaiche, Manuel Bibes, Bang-Gui Liu, and Jorge Íñiguez, “Energetics of oxygen-octahedra rotations in perovskite oxides from first principles,” Physical Review B 97, 024113 (2018).DOI
- [40]Ulf R. Pedersen, Felix Hummel, Georg Kresse, Gerhard Kahl, and Christoph Dellago, “Computing gibbs free energy differences by interface pinning,” Phys. Rev. B 88, 094101 (2013).DOI
- [41]Luca Maragliano and Eric Vanden-Eijnden, “A temperature accelerated method for sampling free energy and determining reaction pathways in rare events simulations,” Chemical Physics Letters 426, 168–175 (2006).DOI
- [42]Zheyong Fan, Zhezhu Zeng, Cunzhi Zhang, Yanzhou Wang, Keke Song, Haikuan Dong, Yue Chen, and Tapio Ala-Nissila, “Neuroevolution machine learning potentials: Combining high accuracy and low cost in atomistic simulations and application to heat transport,” Physical Review B 104, 104309 (2021).DOI
- [43]A. M. Glazer, “Simple ways of determining perovskite structures,” Acta Crystallographica Section A 31, 756–762 (1975).DOI
- [44]Michael W. Lufaso and Patrick M. Woodward, “Prediction of the crystal structures of perovskites using the software program SPuDS,” Acta Crystallographica Section B 57, 725–738 (2001).pubmed.ncbi.nlm.nih.gov/11717471
- [45]Ryosuke Jinnouchi, Ferenc Karsai, and Georg Kresse, “On-the-fly machine learning force field generation: Application to melting points,” Phys. Rev. B 100, 014105 (2019).DOI
- [46]In a simple landau model with a free-energy landscape of the form F(R) = −1/2αR² + 1/4βR⁴, the curvature around the minimum R₀ (that satisfies ∂F/∂R|R=R₀ = 0) is precisely α, which is twice the magnitude of the quadratic curvature of the double well at R = 0.
- [47]W. Zhong and David Vanderbilt, “Competing structural instabilities in cubic perovskites,” Physical Review Letters 74, 2587 (1995).pubmed.ncbi.nlm.nih.gov/10057965
- [48]Jacek C Wojdeł, Patrick Hermet, Mathias P Ljungberg, Philippe Ghosez, and Jorge Íñiguez, “First-principles model potentials for lattice-dynamical studies: general methodology and example of application to ferroic perovskite oxides,” Journal of Physics: Condensed Matter 25, 305401 (2013).DOI
- [49]Nicole A. Benedek and Craig J. Fennie, “Why are there so few perovskite ferroelectrics?” The Journal of Physical Chemistry C 117, 13339–13349 (2013).arxiv.org/abs/1307.8103
- [50]Koichi Momma and Fujio Izumi, “Vesta3 for three-dimensional visualization of crystal, volumetric and morphology data,” Journal of Applied Crystallography 44, 1272–1276 (2011).DOI
- [51]Ke Xu, Hekai Bu, Shuning Pan, Eric Lindgren, Yongchao Wu, Yong Wang, Jiahui Liu, Keke Song, Bin Xu, Yi-fan Li, Tobias Hainer, Lucas Svensson, Julia Wiktor, Rui Zhao, Hongfu Huang, Cheng Qian, Shuo Zhang, Zhezhu Zeng, Bohan Zhang, Benrui Tang, Yang Xiao, Zihan Yan, Jiuyang Shi, Zhixin Liang, Junjie Wang, Ting Liang, Shuo Cao, Yanzhou Wang, Penghua Ying, Nan Xu, Chengbing Chen, Yuwen Zhang, Zherui Chen, Xin Wu, Wenwu Jiang, Esme Berger, Yanlong Li, Shunda Chen, Alexander J. Gabourie, Haikuan Dong, Shiyun Xiong, Ning Wei, Yue Chen, Jianbin Xu, Feng Ding, Zimei Sun, Tapio Ala-Nissila, Ari Harju, Jincheng Zheng, Pengfei Guan, Paul Erhart, Jian Sun, Wengen Ouyang, Yanjing Su, and Zheyong Fan, “Gpumd 4.0: A high-performance molecular dynamics package for versatile materials simulations with machine-learned potentials,” Materials Genome Engineering Advances 3, e70028 (2025).DOI
- [52]All data ara available at 10.5281/zenodo.22769801.
- [53]P. E. Blöchl, “Projector augmented-wave method,” Physical Review B 50, 17953–17979 (1994).pubmed.ncbi.nlm.nih.gov/9976227
- [54]G. Kresse and J. Furthmüller, “Efficient iterative schemes for ab initio total-energy calculations using a plane-wave basis set,” Physical Review B 54, 11169–11186 (1996).pubmed.ncbi.nlm.nih.gov/9984901
- [55]G. Kresse and D. Joubert, “From ultrasoft pseudopotentials to the projector augmented-wave method,” Physical Review B 59, 1758–1775 (1999).DOI
- [56]John P. Perdew, Kieron Burke, and Matthias Ernzerhof, “Generalized gradient approximation made simple,” Physical Review Letters 77, 3865 (1996).pubmed.ncbi.nlm.nih.gov/10062328
- [57]Glenn J. Martyna, Douglas J. Tobias, and Michael L. Klein, “Constant pressure molecular dynamics algorithms,” The Journal of Chemical Physics 101, 4177–4189 (1994).DOI
- [58]William G. Hoover and Brad Lee Holian, “Kinetic moments method for the canonical ensemble distribution,” Physics Letters A 211, 253–257 (1996).DOI
- [59]M. Parrinello and A. Rahman, “Polymorphic transitions in single crystals: A new molecular dynamics method,” Journal of Applied Physics 52, 7182–7190 (1981).DOI





