Relaxor behavior in rocksalt cation-ordered material induced by (anti)ferroelectric phase competition Authors: Yu Yun, Liyan Wu, Drew Behrendt, Hao Pan, Menglin Zhu, Michael Xu, Anthony J. Ruffino, John Carroll, Irina Baraban, Zishen Tian, Xianfei Xu, Dongfang Chen, Brendan M. Hanrahan, Lane W. Martin, James M. LeBeau, Andrew M. Rappe, Ilya Grinberg, Jonathan E. Spanier Published: 2026 ABSTRACT Relaxor ferroelectrics are characterized by dispersion of the temperature-dependent dielectric constant with frequency and enhanced electromechanical coupling. These properties arise from the dynamic polar response of correlated nanodomains that are strongly associated with atomic-scale compositional disorder, which disrupts long-range ferroelectric ordering and enables nanodomain formation. Here, we report relaxor properties originating from spontaneous low temperature phase competition in fully cation-ordered antiferroelectric PbMg0.5W0.5O3 epitaxial films, including the identification of a new low-energy polar phase. Unlike prototypical relaxors, the B-site cations in coherently strained PbMg0.5W0.5O3 films exhibit long-range rocksalt chemical ordering. Temperature-dependent polarization studies reveal the switching behaviors associated with the phase transitions from paraelectric to antiferroelectric to ferroelectric, and the characteristic dielectric relaxation is ascribed instead to phase competition between the polar and antipolar phases mediated by temperature and substrate clamping. This phase competition breaks long-range dipole correlation and leads to dielectric dispersion and relaxor behavior. These findings demonstrate a new paradigm for designing relaxor material properties through engineered phase competition. Introduction Relaxor ferroelectrics (FE) are prized for their exceptional electromechanical and dielectric responses, making them vital for sensors, transducers, and energy storage1–3. Their unique properties are related to fascinating, complex polar dynamics and a flat Landau energy profile, which arise from competition of multiple polar phases4,5 induced by local chemical and structural inhomogeneity6–9. This nanoscale disorder, often attributed to cation disorder, gives rise to the hallmark of relaxor FE behavior: a strong dependence of the dielectric constant on frequency (dielectric dispersion) tied to evolution of polar states with temperature10–13, in contrast to weak frequency dispersion reported in standard FE solid solutions that is attributed to domain wall and phase boundary effects14. The degree of compositional ordering and the size of the resulting polar domains further modulates relaxor behavior15. For instance, the 1:1 type FE relaxors Pb(Sc1/2Nb1/2)O3 and Pb(Sc1/2Ta1/2)O3 can be either a classic or relaxor FE solely depending on the extent of B-site cation ordering, which can be kinetically controlled by annealing15–18. Numerous models of relaxor behavior have been proposed8,19–21 which are all based on the concept that the long-range ordering of polar domains must be limited by nanoscale disorder, while no ordered systems exhibiting relaxor behavior have been reported to date. Here, we demonstrate that relaxor behavior can be realized without compositional randomness, owing to phase competition between the polar and antipolar states in a long-range cation-ordered perovskite. We stabilize highly cation-ordered films and demonstrate the coexistence of antiferroelectric (AFE) and FE phases in perovskite PbMg0.5W0.5O3. Using density functional theory (DFT), we identify a novel FE phase that is less than 1 meV/atom above the ground state AFE phase and is stabilized through compressive strain. The observed frequency-dependent dielectric relaxation correlates with the formation of mixed FE and AFE phases arising from compressive strain caused by the cooling substrate and the kinetic trapping of these phases at low temperatures. Our results show that relaxor behavior can be obtained in an elegantly simple material system distinguished by chemical composition homogeneity—paving the way for completely new relaxor FE materials that are not constrained by the limitations of the atomic disorder associated with conventional relaxor FEs. Results Structural characterization and B-site cation ordering in PbMg0.5W0.5O3 Bulk PbMg0.5W0.5O3 has a double-perovskite (rocksalt) cubic structure and undergoes a first-order phase transition from cubic structure with the Fm3m space group (paraelectric phase) to orthorhombic structure with the Pmcn (Pnma) space group (AFE phase) at T≈≈311 K, in close proximity to room temperature22–28 (Fig. 1a). Due to the large charge difference between Mg2+ and W6+, PbMg0.5W0.5O3 is well-known to display long-range rocksalt B-site cation ordering27–30 regardless of the synthesis protocol, in contrast with the Pb(Sc1/2Nb1/2)O3 and Pb(Sc1/2Ta1/2)O3 systems for which the smaller charge difference of 2 between Sc3+ and Nb/Ta5+ leads to the absence (presence) of long-range B-site cation ordering for samples obtained by quenching (annealing).Fig. 1Thin film growth and X-ray diffraction (XRD) characterization of B-site ordered PbMg0.5W0.5O3 on NdScO3 substrate.a Atomic structure models of PbMg0.5W0.5O3 with PE phase (cubic) and AFE phase (orthorhombic). b Representative XRD θ−2θ scan of 100 nm-thick PbMg0.5W0.5O3/NdScO3 heterostructure. Inset: RHEED patterns of PbMg0.5W0.5O3 and NdScO3. c A higher resolution plot of the PbMg0.5W0.5O3 024-diffraction condition with clear Laue oscillations. d The rocking curve of the PbMg0.5W0.5O3 (024) plane with FWHM of 0.01°. e RSM of NdScO3 240-diffraction condition, exhibiting coherent epitaxial growth without strain relaxation. f RSM of PbMg0.5W0.5O3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{1}\,\bar{3}$$\end{document}1¯3¯ 5-diffraction, indicating B-site cation ordering. Thin film growth and X-ray diffraction (XRD) characterization of B-site ordered PbMg0.5W0.5O3 on NdScO3 substrate. a Atomic structure models of PbMg0.5W0.5O3 with PE phase (cubic) and AFE phase (orthorhombic). b Representative XRD θ−2θ scan of 100 nm-thick PbMg0.5W0.5O3/NdScO3 heterostructure. Inset: RHEED patterns of PbMg0.5W0.5O3 and NdScO3. c A higher resolution plot of the PbMg0.5W0.5O3 024-diffraction condition with clear Laue oscillations. d The rocking curve of the PbMg0.5W0.5O3 (024) plane with FWHM of 0.01°. e RSM of NdScO3 240-diffraction condition, exhibiting coherent epitaxial growth without strain relaxation. f RSM of PbMg0.5W0.5O3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{1}\,\bar{3}$$\end{document}1¯3¯ 5-diffraction, indicating B-site cation ordering. The high-quality epitaxial PbMg0.5W0.5O3 thin films were deposited by pulsed laser deposition (“Methods”). The 100-nm-thick PbMg0.5W0.5O3 film was grown on an NdScO3 (110) substrate with an excellent lattice matching and a representative x-ray diffraction θ−2θ scan (Fig. 1b) indicates the films are single-phase and epitaxial. The reflection high-energy electron diffraction (RHEED) patterns of the NdScO3 and PbMg0.5W0.5O3 film (inset of Fig. 1b), the surface morphology with very low roughness measured by atomic force microscopy (Fig. S1), and clear Laue oscillations (used to determine film thickness) of PbMg0.5W0.5O3 024-diffraction conditions (Fig. 1c), indicate the two-dimensional (2D) growth mode, a smooth film surface and film-substrate interface. The full width at half maximum (FWHM) of rocking curves of the PbMg0.5W0.5O3 (024) plane is about 0.01° (Fig. 1d), indicating the high crystallinity and lateral coherence of PbMg0.5W0.5O3 films. The in-plane strain state of PbMg0.5W0.5O3/NdScO3 heterostructure, analyzed with reciprocal space mapping (RSM) of NdScO3 (240) (Fig. 1e), shows that the PbMg0.5W0.5O3 film is coherently strained to the substrates without strain relaxation and exhibits clear Laue oscillations. The appearance of PbMg0.5W0.5O3 \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\bar{1}\bar{3}5$$\end{document}1¯3¯5-reflection with d-spacing of 1.353 Å indicates the presence of the expected long-range B-site cation ordering (Fig. 1f). To further reveal the atomic-level cation-ordered structure and antipolar Pb displacements, we performed scanning transmission electron microscopy (STEM) characterization at both ambient (298 K) and liquid nitrogen (approximately 103 K) temperatures. Viewed along the pseudocubic NdScO3 [1–10] (Fig. 2a) and [100] (Fig. S2) zone axis at room temperature, the cross-sectional high-angle annular dark-field (HAADF) STEM image of a 200-nm-thick PbMg0.5W0.5O3/NdScO3 film reveals alternating atom column intensity on the B-site sub-lattice. This atomic-number sensitive HAADF contrast is consistent with the expected ordering of Mg and W, as indicated by the overlaid atomic model (Fig. 2b). In addition, the nano-beam electron diffraction (NBED) pattern collected across a larger region of the film shows defined ½ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{111\right\}$$\end{document}111-type superlattice reflections (marked by white circle in Fig. 2c), consistent with a doubling of the repeat unit associated with rock-salt ordering of Mg and W. Subsequently, a cross-sectional sample viewed along the pseudocubic [100] zone axis was cooled to liquid nitrogen temperature and imaged in situ to fully recover the AFE state31. From the HAADF STEM image, the relative displacements between Pb and Mg/W atomic columns were mapped, revealing the characteristic antiparallel cation displacements associated with the AFE structure (Fig. 2d–f). The displacement pattern in experiments is consistent with that of the atomic model for AFE PbMg0.5W0.5O3 (Fig. 2g), in which four unique Pb displacement directions comprise the antipolar [110]-aligned dipole planes, as shown in Fig. 2f.Fig. 2STEM characterization of PbMg0.5W0.5O3 thin films.a STEM-HAADF image of the PbMg0.5W0.5O3 film interface along the pseudocubic <110> projection taken at 298 K. b A magnified subregion is overlaid with the projected atomic model with Pb, Mg, and W cations. Mg (Z = 12) and O (Z = 8) are not visible, while Pb (Z = 82) and W (Z = 74) are clearly defined. c STEM-NBED pattern averaged from the film region along the pseudocubic <110> projection. The ½ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{111\right\}$$\end{document}111 superlattice reflection is circled and consistent with the rock-salt ordered structure. d Cryogenic STEM-HAADF image along the pseudocubic <100> projection taken at 103 K. e The subregion and its corresponding (f) ordered Pb vs. Mg/W cation displacements are consistent with AFE structure. The color wheel saturates at a displacement magnitude of 25 pm. g Atomic model of PMW, with AFE displacement pattern indicated and matching experiment. STEM characterization of PbMg0.5W0.5O3 thin films. a STEM-HAADF image of the PbMg0.5W0.5O3 film interface along the pseudocubic <110> projection taken at 298 K. b A magnified subregion is overlaid with the projected atomic model with Pb, Mg, and W cations. Mg (Z = 12) and O (Z = 8) are not visible, while Pb (Z = 82) and W (Z = 74) are clearly defined. c STEM-NBED pattern averaged from the film region along the pseudocubic <110> projection. The ½ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left\{111\right\}$$\end{document}111 superlattice reflection is circled and consistent with the rock-salt ordered structure. d Cryogenic STEM-HAADF image along the pseudocubic <100> projection taken at 103 K. e The subregion and its corresponding (f) ordered Pb vs. Mg/W cation displacements are consistent with AFE structure. The color wheel saturates at a displacement magnitude of 25 pm. g Atomic model of PMW, with AFE displacement pattern indicated and matching experiment. Temperature evolution of switching behavior To assess the electrical properties, Ba0.6Sr0.4RuO3 layers were deposited on NdScO3 as bottom electrodes (“Methods”) with the results of the corresponding structural characterizations of the PbMg0.5W0.5O3/Ba0.6Sr0.4RuO3/NdScO3 heterostructure displayed in Fig. S3. The out-of-plane polarization-electric-field (P-E) hysteresis loops and their derivative dP/dE plotted vs electric field of a 200-nm-thick PbMg0.5W0.5O3 film are presented in Fig. 3a, b for temperatures ranging from 293 to 100 K.Fig. 3Temperature dependent characterization of polarization switching for PbMg0.5W0.5O3 thin films.a, b Polarization-electric field (P-E) loops and their derivative dP/dE vs electric field at the temperature of 293, 250, 200, 150, and 100 K, respectively. Extrapolation of the FE phase polarization to zero field is shown by dotted lines. c P-E loops and d dP/dE via electric field at the temperature from 10 to 80 K. e Temperature evolution of critical fields for the phase transition from AFE to FE and from FE to AFE extracted from temperature-dependent P-E loops. f Temperature evolution is shown for Pr, PFE,0, and Pmixed. g Temperature evolution of fFE,mixed, fFE,switch, and their product fFE,mixed* fFE,switch representing the fraction of FE phase responsive to the applied E field. Temperature dependent characterization of polarization switching for PbMg0.5W0.5O3 thin films. a, b Polarization-electric field (P-E) loops and their derivative dP/dE vs electric field at the temperature of 293, 250, 200, 150, and 100 K, respectively. Extrapolation of the FE phase polarization to zero field is shown by dotted lines. c P-E loops and d dP/dE via electric field at the temperature from 10 to 80 K. e Temperature evolution of critical fields for the phase transition from AFE to FE and from FE to AFE extracted from temperature-dependent P-E loops. f Temperature evolution is shown for Pr, PFE,0, and Pmixed. g Temperature evolution of fFE,mixed, fFE,switch, and their product fFE,mixed* fFE,switch representing the fraction of FE phase responsive to the applied E field. At T > 150 K, the hysteresis loops show classic AFE behavior where large polarization is observed at large positive E field and then shows five stages of decrease: slow, rapid, slow, rapid, and slow (Fig. 3a). The slow changes correspond to the changes in the intrinsic P of the two FE phases (Stages I and V) and the AFE phase (Stage III), while the rapid changes correspond to the switching of the material from FE to AFE and from AFE to FE (Stages II and IV). As the temperature decreases, the hysteresis loops become wider with greater difference in the E-fields of the AFE→FE (EAFE→FE) and FE→AFE (EFE→AFE) transitions (Supplementary Materials). For T = 100 and 80 K, similar behavior is observed with five stages of decrease in P (Fig. 3c, d); however, a full switching from the FE phase to the AFE phase is not achieved and some amount of the FE phase remains even at zero field as indicated by non-zero remanent P (Pr). Since at EAFE→FE a kinetic barrier no longer prevents the transformation of the FE phase to the AFE phase, the retention of P at low E fields, including at zero field means that a mixed state comprised by both AFE and FE phases is thermodynamically preferred at these conditions with Pr equal to the polarization of the thermodynamically preferred mixed state at zero field (Pmixed). The fraction of the FE phase in the mixed state at these temperatures can be estimated using the ratio between the zero-field P of the FE phase (PFE,0) and Pr, where PFE,0 can be obtained by extrapolating the E-field dependence of P at high fields when the system is fully in the FE state (the dotted line in Fig. 3c). Due to the coexistence of the FE and AFE phases, the slow change of polarization in Stage III involves both the change in the intrinsic P of the AFE phase and the switching of the remaining FE phase from Stage II to the AFE phase. The temperature evolution of FE and AFE phases, and their apparent coexistence, is also revealed in the inelastic light scattering (Supplementary Materials), including the emergence of the cubic T2g(2) mode ≈≈ 65 cm−1 (Fig. S4) corresponding to antiphase vibration between A-site (Pb) and O octahedra, which splits into multiple lines at temperatures <100 K, similar to bulk28. Comparison of the loops for 60 ≤ T ≤ 100 K (Fig. 3c) shows that as the temperature decreases, Pmixed is equal to Pr and increases strongly (Fig. 3f), indicating a rise in the fraction of the FE phase in the thermodynamically favored mixed phase. For T < 60 K, the transition to the thermodynamically preferred mixed state only occurs at negative E fields, indicating that the FE phase is kinetically trapped at E = 0. For these temperatures, Pmixed can be estimated by the extrapolation of the P dependence in Stage III to E = 0 (the dashed line in Fig. 3c). Thus, at these temperatures, the kinetic trapping of the FE phase becomes important, leading to wider FE loops and inability of the FE phase to switch to the AFE phase at zero field. To quantitatively describe the changes in the switching characteristics of the PbMg0.5W0.5O3 films, we plot the characteristic parameters of the loops in Fig. 3e–g. A plot of EAFE→FE and EFE→AFE versus temperature (Fig. 3e) shows that for T > 250 K, EAFE→FE = EFE→AFE, whereas for T < 250 K, EAFE→FE exhibits the expected behavior of rising with decreasing temperature while EFE→AFE first increases and then starts to decrease at ≈150 K, due to the increasingly strong stabilization of the FE phase. The plot of Pr, Pmixed and PFE,0 (Fig. 3f) shows that PFE,0 exhibits a gradual decline from 0.255 C m−2 at 10 K to ~0 at 300 K, while Pr decreases rapidly from 0.233 C m−2 at 10 K to 0.04 C m−2 at 60 K and then decreases slowly to 0 at 150 K. Based on the values of PFE,0, Pr, and Pmixed we can estimate the fraction of PbMg0.5W0.5O3 films that is thermodynamically preferred to be in the FE phase at E = 0 (fFE,mixed = Pmixed/PFE,0), the fraction of the FE phase that is switchable to the AFE phase at E = 0 (fFE,switch = (PFE,0–Pr)/PFE,0). The total amount of the FE phase that can be switchable in a film at thermodynamic equilibrium (non-poled) at zero-field can be estimated as the product of fFE,mixed and fFE,switch (Fig. 3g). It can be seen fFE,mixed*fFE,switch starts rising at 125 K, peaks at 60 K and then decreases to zero at 10 K. This suggests that the response to the applied E field will depend most strongly on the E field frequency in the region at ~60 K. At higher temperatures, the switching is very fast due to low kinetic barriers at E = 0, whereas at lower temperature the system is kinetically trapped by a very high barrier so that switching does not occur on the timescale of the hysteresis loop measurements. Relaxor behavior in PbMg0.5W0.5O3 To show the relaxor characteristics, we examine the temperature-dependent dielectric permittivity (ɛ´) and loss (tanδ) at various frequencies (Fig. 4a). As the temperature decreases, a dielectric enhancement without frequency dispersion is observed near room temperature, corresponding to a sharp first-order phase transition between PE and AFE phases (Fig. 1a) at the Curie temperature (Tc). As the temperature is reduced to around 130 K, a distinct dielectric relaxation behavior is identified, where the drop in permittivity and a concomitant peak in the dielectric loss demonstrate a large variation as a function of driving frequency, similar to that observed in relaxor FEs11,12. The relationship between the temperature of maximum dielectric loss (Tm) of the low-temperature relaxor phase and the frequency (f) can be described by Arrhenius law11:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f={f}_{0}\exp \left[-{E}_{{{\rm{a}}}}/{k}_{{{\rm{B}}}}{T}_{{{\rm{m}}}}\right]$$\end{document}f=f0exp−Ea/kBTmwhere f is the frequency of measurement, f0 is the attempt jump frequency for dipoles, Ea is the activation energy, and kB is the Boltzmann constant. The linear relationship between ln (f) and 1/Tm and the fitted parameters using Eq. (1) are shown in Fig. 4b. Ea ≈ 0.095 eV is similar to the activation energy in Pb-relaxors11,32,33. However, in contrast to Pb(Mg1/3Nb2/3)O3-PbTiO3, the relaxation cannot be attributed to the weak coupling at Mg-rich local environments8 due to the full B-cation order in the PbMg0.5W0.5O3 films. The similarity between the temperature of the onset of the relaxor behavior and the temperature at which EAFE starts to deviate from the expected monotonic temperature evolution and Pr begins to increase suggest that both of these effects are caused by the same structural phenomenon that occurs at T ≈ 100–130 K. Furthermore, as observed from Fig. 4a, after the onset at 130 K, the difference between ε(T) at different frequencies first increases with lower T until T ≈ 60 K and then decreases. The temperature range between 60 and 130 K corresponds to the temperature range of rapid rise in Pmixed (FE phase stabilization) and weak kinetic trapping of the FE phase observed in the hysteresis loop, whereas the strong kinetic trapping of the FE phase corresponds to weaker dispersion observed for T < 60 K. This suggests that the relaxor behavior is due to the partial stabilization of the FE phase and coexistence of the FE and AFE phases; the coexistence of the phases breaks the long-range polar ordering and leads to the formation of nanopolar domains that respond differently to different frequencies of applied field. The maximum dispersion also corresponds to the maximum of fFE,mixed*fFE,switch (purple shadow in Fig. 3g), suggesting that at these temperatures the FE and AFE phases achieve both relatively high mixing due to a high fFE,mixed which creates a high density of FE/AFE interfaces/domain walls, similar to the high density of domain walls found in relaxor FEs while retaining the ability to respond to the applied E-field (high fFE,switch).Fig. 4Dielectric relaxation of PbMg0.5W0.5O3 thin films.a Dielectric permittivity and loss factor as a function of temperature and frequency (2–200 kHz). b The frequency dependence of Tm for PbMg0.5W0.5O3 films. Tm is the temperature for the maximum loss factor for a specific frequency. Dielectric relaxation of PbMg0.5W0.5O3 thin films. a Dielectric permittivity and loss factor as a function of temperature and frequency (2–200 kHz). b The frequency dependence of Tm for PbMg0.5W0.5O3 films. Tm is the temperature for the maximum loss factor for a specific frequency. Defects such as oxygen vacancies often have complex effects on conventional FEs and in some cases their presence at high concentration gives rise to relaxor behavior as well as high leakage and asymmetric hysteresis loops34,35. In this work, to reduce the effects of defects, high-quality epitaxial films were obtained by optimizing growth conditions followed by post-annealing in a high-oxygen atmosphere. These films exhibit low leakage and symmetric hysteresis loops, indicating a low concentration of defects such as oxygen vacancies in the films. Such low concentrations of defects do not give rise to relaxor behavior in films of AFE materials such as PbZrO3. Therefore, it is unlikely that the observed relaxor behavior is due to the presence of oxygen vacancies and other defects, and rather arises from the competition between the AFE and FE phase as discussed above. Relaxor dispersion in the vicinity of AFE-FE phase transition has been previously observed in Na₀.₅Bi₀.₅TiO₃ (NBT) and NaNbO₃; however, both of these systems are distinct from the PMW films examined in this study. NBT films show strong cation disorder on the A-site, so that relaxor behavior observed in this system is considered to arise from cation disorder, similar to the classic PMN material. While NaNbO₃ is nominally ion-disorder-free, it is a highly complex material that is known to be extremely sensitive to a wide variety of factors during synthesis, including small deviations from stoichiometry and homogeneity; thus, due to its extreme sensitivity, even low concentrations of defects may affect NaNbO₃ strongly enough to give rise to relaxor behavior. Furthermore, the FE and AFE phase in NaNbO₃ are due to the off-center displacements of O anions rather than the cation as is the case in PMW and other standard FEs such as PMN and PZT. Therefore, the relaxor behavior in NaNbO₃ can be considered to be a result of the unusual structural properties of NaNbO₃. Thus, due to these distinctions between our systems and the previously reported NBT and NaNbO₃, our work on high-quality, single-crystal PMW films demonstrates a new paradigm for designing relaxor behavior in cation-driven perovskite FEs through engineered phase competition. To characterize the atomic configuration of this newly observed FE state, we used DFT calculations to find the most stable FE phase for bulk PbMg0.5W0.5O3. We found that instead of homogeneous displacement of Pb atoms along a single axis, this phase has a polarization along (110) generated by alternating Pb displacements along (100) and (010), similar to the AFE ground state. These results are quite different from the FE phase of the prototypical AFE PbZrO3, which shows a standard uniform displacement for all Pb and Zr ions (Fig. 5). Interestingly, the FE phase was found to be only 0.5 meV per atom higher in energy than the AFE phase for bulk PbMg0.5W0.5O3, while for PbZrO3 the energy difference is over 2 meV per atom. The small energy difference between the AFE and FE phases is consistent with the observation of the stabilization of the FE phase at low temperatures revealed by the hysteresis loops.Fig. 5Comparison of antiferroelectric and ferroelectric atomic configurations in PbMg0.5W0.5O3 and PbZrO3.Most stable AFE (a, b) and FE (c, d) atomic configurations of bulk PbMg0.5W0.5O3, showing the atomic structures (a, c) and corresponding Pb displacement patterns (b, d). Most stable AFE (e, f) and FE (g, h) configurations of bulk PbZrO3, showing the atomic structures (e, g) and corresponding Pb displacement patterns (f, h). Arrows indicate the directions of Pb displacements. Comparison of antiferroelectric and ferroelectric atomic configurations in PbMg0.5W0.5O3 and PbZrO3. Most stable AFE (a, b) and FE (c, d) atomic configurations of bulk PbMg0.5W0.5O3, showing the atomic structures (a, c) and corresponding Pb displacement patterns (b, d). Most stable AFE (e, f) and FE (g, h) configurations of bulk PbZrO3, showing the atomic structures (e, g) and corresponding Pb displacement patterns (f, h). Arrows indicate the directions of Pb displacements. To shed light on the appearance of the FE phase at low temperatures (<150 K), we examine the influence of strain changes on the relative stabilities of the FE and AFE phases (Fig. 6). Strain changes will appear due to the contraction of the lattice for both substrate and PbMg0.5W0.5O3 upon cooling. Our calculations show that the two phases change thermodynamic ordering with strain, with the equilibrium volume of the FE phase being 1% smaller than that of the AFE phase for relaxations at 0 K (Fig. 6a). Calculations of both the phase energies and the homogeneous switching barriers (triple wells) at different strain states show that compressive strain promotes the formation of the FE phase, with the FE phase becoming the ground state with a moderate (≈ −0.3%) change in the strain (Fig. 6b) similar to the strain imposed by the substrate on PbMg0.5W0.5O3 at low temperatures36,37 without strain relaxations (Fig. S5 and Supplementary Materials). This suggests that the FE phase will be stabilized by the shrinkage of the lattice parameters of the substrates during cooling, consistent with the appearance of the FE phase at T < 150 K observed from the measured hysteresis loops.Fig. 6The influence of strain changes on the relative stabilities of the FE and AFE phases of PbMg0.5W0.5O3.a Energy for the AFE and FE phase as a function of strain from equilibrium PbMg0.5W0.5O3 cell. For any compressive modification, the FE phase is preferred (blue) while the opposite is true for AFE (red). There also exists a regime with a small compressive strain where the phase energy is almost indistinguishable (purple) and the phase energies remain close throughout. b Corresponding triple well potential for each of the stability regimes. When either phase is preferred, the other remains close in energy and the barrier for transitioning between them remains similar. c A schematic for the phase evolution and competition accounting for the relaxor behavior at the cryogenic temperature. The influence of strain changes on the relative stabilities of the FE and AFE phases of PbMg0.5W0.5O3. a Energy for the AFE and FE phase as a function of strain from equilibrium PbMg0.5W0.5O3 cell. For any compressive modification, the FE phase is preferred (blue) while the opposite is true for AFE (red). There also exists a regime with a small compressive strain where the phase energy is almost indistinguishable (purple) and the phase energies remain close throughout. b Corresponding triple well potential for each of the stability regimes. When either phase is preferred, the other remains close in energy and the barrier for transitioning between them remains similar. c A schematic for the phase evolution and competition accounting for the relaxor behavior at the cryogenic temperature. We then performed calculations for the intrinsic FE→AFE switching pathway; here as well, we found that PbMg0.5W0.5O3 exhibits unusual behavior where dipoles must only rotate 90° (Fig. S6) to change the polarization by 180°, leading to much lower switching energy, which accounts for the lower EAFE and EFE of PbMg0.5W0.5O3 compared to that of the typical AFE PbZrO3. Figure 6c depicts the schematic of structural evolution associated with relaxor behavior as a function of temperature. At higher temperature, the AFE phase is energetically favorable. With decreasing temperature, the energy difference between the FE and AFE phases decreases and the long-range antipolar ordering is disrupted by the formation of small regions of FE phase (likely at the AFE domain walls) intermixed with the regions of the AFE phase, resulting in a diffuse phase transition behavior at the intermediate temperature. As temperature is further reduced, the energy difference between FE and AFE phases decreases further, and the fraction of the FE polar phase with non-zero macroscopic polarization grows further, but the FE phase is kinetically trapped and cannot switch in response to the applied E-field. Further insights into the connection between the AFE-FE phase competition and observed relaxor behavior would benefit from theoretical modeling of relaxor dynamics in PMW. However, simple models for understanding relaxor dynamics are currently lacking, while MD simulations of coexisting AFE and FE phases in PMW would be highly non-trivial and require parameterization and validation of atomistic potentials as well as simulations at large length- and timescales. Such simulations are outside the scope of this work and will be conducted in future research. While compositional inhomogeneity and the presence of nanoscale regions with short-range structural correlations have long been believed to be two primary prerequisites for a relaxor, our work shows that the competition between the FE and AFE phases can lead to relaxor behavior even in a fully ordered oxide. The thermodynamically determined local polar development associated with (anti)ferroelectric phase competition is promising for realizing and harnessing relaxor behavior without compositional inhomogeneity, akin to the intrinsic phase competition that enables unexpected responses in lattice and charge–orbital ordered antiferromagnetics38 and its temperature evolution of magnetic hysteresis across magnetic phases39. Methods Sample preparation Heterostructures of 100 nm-PbMg0.5W0.5O3/NdScO3 and 30 nm-SrRuO3/200 nm- PbMg0.5W0.5O3/20 nm-Ba0.6Sr0.4RuO3/NdScO3 were synthesized via pulsed-laser deposition, with a target-to-substrate distance of 5.5 cm, using a KrF excimer laser (248 nm, Coherent). The Ba0.6Sr0.4RuO3 bottom electrodes were grown at a heater temperature of 780 °C in a dynamic oxygen pressure of 20 mTorr with a laser fluence of 2 J cm−2 and a laser repetition rate of 3 Hz from a ceramic target of the same composition. The PbMg0.5W0.5O3 layers were grown at a heater temperature of 580 °C in a dynamic oxygen pressure of 200 mTorr with a laser fluence of 2 J cm−2 and a laser repetition rate of 2 Hz from a ceramic target of the same composition with 10% lead excess fabricated by solid state reaction to compensate for lead loss during growth. The SrRuO3 top electrode layers were grown in situ immediately following PbMg0.5W0.5O3 deposition at a heater temperature of 530 °C in a dynamic oxygen pressure of 100 mTorr with a laser fluence of 2 J cm−2 and a laser repetition rate of 6 Hz. Following deposition, the heterostructures were cooled at a rate of 5 °C min−1 in a static oxygen pressure of 300 Torr. Whole growth process is monitored in situ by reflection high-energy electron diffraction (RHEED). The capacitor structures were fabricated via standard photolithography process and wet etching lithography. The top SrRuO3 layer was selectively wet etched by an NaIO4 aqueous solution (0.1 mol L−1) to realize circular capacitor structures with a diameter of 25 μm. X-ray diffraction (XRD) measurements The room-temperature XRD θ−2θ scans along a normal direction, rocking curves (ω scans) of PbMg0.5W0.5O3 (024), and 2D RSM were studied using Malvern Panalytical Empyrean XRD system (PANalytical), with Cu Kα (wavelength ≈ 1.5406 Å) X-ray. Scanning probe microscopy Surface topography of the PMW films was evaluated by scanning probe microscopy (Asylum Research Cypher, Oxford Instruments) using intermittent contact mode. Electrical measurements The P-E hysteresis loops were measured at a frequency of 10 kHz using a Precision Multiferroic Tester (Radiant Technologies) in a probe station. Dielectric response of the heterostructures was measured as a function of frequency (2–200 kHz) using an impedance analyzer (Keysight E4990A). The temperature of electrical measurement is from 10 to 350 K. Electron microscopy Cross-section samples for STEM characterization were prepared by Ga+ focused ion beam (Thermo Fisher Scientific Helios 600) lift out. Thinning to electron transparency was achieved by milling with decreasing Ga+ ion beam energy from 30 to 5 kV. The final thinning to remove Ga+ damage was performed by broad beam Ar+ ion milling (Fischione 1051 TEM Mill) using beam energies of 0.3 and 0.1 kV. Electron microscopy datasets were acquired using a probe aberration-corrected STEM (Thermo Fisher Scientific Themis-Z S/TEM) at an accelerating voltage of 300 kV with probe semi-convergence angle of 21.8 mrad and beam current of ≈10 pA. For in situ cryo-STEM imaging, cross-sectional lamellae were prepared by focused ion beam lift out and mounted on micro-electromechanical chips (DENSsolutions), after which cooling to liquid nitrogen temperature was performed using a dedicated cryogenic cooling holder (DENSsolutions Arctic Lightning). Raman scattering spectroscopy Raman scattering spectroscopy was collected at selected temperatures between 1.6 and 350 K (OptiCool, Quantum Design) in the backscattering geometries \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z({YY})\bar{Z}$$\end{document}Z(YY)Z¯ (parallel) and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Z({XY})\bar{Z}$$\end{document}Z(XY)Z¯ (crossed) under 325 nm HeCd laser excitation (Kimmon-Koha), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx \,$$\end{document}≈ mW, focused to a ≈≈20 μm spot diameter, using a triple subtractive spectrometer (T64000, Horiba Jobin-Yvon) and an air-cooled electron-multiplying charged-coupled detector camera (Andor Newton). The sample was first cooled to 10 K and then spectra were collected from 10 to 100 K in 10 K steps, and from 100 to 350 K in 25 K steps. Theoretical calculations A full switching from the FE phase to the AFE phase is not achieved and some amount of the FE phase remains even at zero field as indicated by non-zero remanent P (Pr). PFE,0 can be obtained by extrapolating the E-field dependence of P at high fields when the system is fully in the FE state (the dotted line in Fig. 3c). Pmixed can be estimated by the extrapolation of the P dependence in Stage III to E = 0 (the dashed line in Fig. 3c). DFT and NEB calculations were performed using the Quantum Espresso40 code with optimized norm-conserving pseudopotentials generated with the OPIUM41,42 code. All calculations were performed at 0 K. The generalized gradient approximation of Perdew, Burke, and Ernzerhof was used to calculate the exchange correlation energy43. Convergence criteria of 1.4e-5 eV per cell and 2.6e-4 eV Å−1 for energy and forces, respectively, were used for all calculations. Supplementary information Supplementary Information Transparent Peer Review file Supplementary Information Transparent Peer Review file