Report

Emergence of room-temperature ferroelectricity at reduced dimensions

See allHide authors and affiliations

Science  18 Sep 2015:
Vol. 349, Issue 6254, pp. 1314-1317
DOI: 10.1126/science.aaa6442

Thinning films induces ferroelectricity

Thin ferroelectric films are needed in computers and medical devices. However, traditional ferroelectric films typically become less and less polarized the thinner the films become. Instead of using a good ferroelectric and making it thinner, Lee et al. started with SrTiO3, which in its bulk form is not ferroelectric. This material does have naturally occurring nanosized polarized regions. and when the thickness of the SrTiO3 films reaches the typical size of these regions, the whole film aligns and becomes ferroelectric.

Science, this issue p. 1314

Abstract

The enhancement of the functional properties of materials at reduced dimensions is crucial for continuous advancements in nanoelectronic applications. Here, we report that the scale reduction leads to the emergence of an important functional property, ferroelectricity, challenging the long-standing notion that ferroelectricity is inevitably suppressed at the scale of a few nanometers. A combination of theoretical calculations, electrical measurements, and structural analyses provides evidence of room-temperature ferroelectricity in strain-free epitaxial nanometer-thick films of otherwise nonferroelectric strontium titanate (SrTiO3). We show that electrically induced alignment of naturally existing polar nanoregions is responsible for the appearance of a stable net ferroelectric polarization in these films. This finding can be useful for the development of low-dimensional material systems with enhanced functional properties relevant to emerging nanoelectronic devices.

Low-dimensional ferroelectric structures hold a great potential for scientific and technological endeavors (1). Reducing size while retaining ferroelectric properties enables an increase in the storage capacity of nonvolatile ferroelectric memories (2), exploration of diverse nanoelectronic functions (37), and discovery of exotic physical phenomena (8, 9). However, maintaining the ferroelectricity in low-dimensional structures, such as ultrathin films, has been hampered by depolarization effects (1012), which arise from the uncompensated charges at the interface. The strong scaling effect seems to inevitably suppress ferroelectricity and its functions below a critical dimension (1013). A recent theoretical work suggested an intriguing concept for reversibly enhancing ferroelectricity in ultrathin ferroelectric capacitors via the tailoring of chemical bonds at the metal/oxide interface (14), but this mechanism has not yet been experimentally confirmed.

Here, we describe a different mechanism, which enables enhancement of ferroelectricity as the thickness of the system is decreased. In our approach, we use naturally existing polar nanoregions (PNRs)—local nanometer-sized polar clusters—in an archetype dielectric material with perovskite structure: strontium titanate (SrTiO3). PNRs are generally believed to arise from local nanoscale inhomogeneities (such as chemical or structural disorder) (15, 16), which exist in every material (17). For example, Sr vacancies are intrinsic point defects in SrTiO3 because of their small formation energy (18, 19), comparable with that of oxygen vacancies, which are likely to act as a natural source of PNRs (Fig. 1A) (20, 21). It has been previously shown (20) that relatively thick (tens of nanometers) films of SrTiO3 exhibit relaxor behavior at low temperatures because of the presence of the PNRs. We demonstrate that electrically induced alignment and stabilization of PNRs in nanometer-thick SrTiO3 films results in the emergence of net ferroelectric polarization at room temperature.

Fig. 1 Theoretical calculations showing the emergence of ferroelectricity in ultrathin films of otherwise nonferroelectric SrTiO3.

(A and B) Dimensional engineering of the polarization (P) stability in PNRs. Shown are schematics of PNRs and their P in (A) thick SrTiO3 and (B) ultrathin SrTiO3. Blue and white arrows indicate P and Ed, respectively. (C) Calculated atomic structure near the off-center antisite Ti atom, which induces P along the [001] (or its equivalent) direction. (D) Profile of local P around the antisite Ti atom in a 3 by 3 by 3 supercell obtained from DFT calculations. (E) Calculated energy barrier between the polarization states. Polarization switching from [001] to Embedded Image direction can be achieved via the metastable polarization states with [101] and Embedded Image directions. (F) Remnant polarization of the model system, in which a single spherical PNR is embedded in SrTiO3 with a thickness t, obtained with phase-field simulations. The lateral dimension of SrTiO3 was fixed as 64 by 64 unit cells (23).

Calculations predict that when SrTiO3 is deficient in Sr, antisite Ti defects could instantaneously form and generate local dipole moments by an off-centering displacement (Fig. 1C) (22). Our first-principles density-functional theory (DFT) modeling (23) shows that the energy gain from this Ti off-centering is as large as ~0.5 eV, originating from structural distortion driven by an ionic radii difference between Ti2+ (0.86 Å) and Sr2+ (1.44 Å). A local polarization profile around the antisite Ti atom (Fig. 1D) indicates that the off-centered antisite Ti atom induces a large local polarization in its residing unit cell and coherently polarizes the surrounding region. The polarization switchability follows from a calculated energy barrier of ~0.1 eV between the polarization states (Fig. 1E), which is comparable with a double-well potential barrier in conventional ferroelectric materials (24). Thus, although pure bulk SrTiO3 is centrosymmetric and nonpolar (25), the PNRs of nanometer-characteristic size can naturally form because of the intrinsic Sr deficiency in SrTiO3.

We have previously observed Sr deficiency and associated PNRs even in nominally stoichiometric SrTiO3 bulk single crystals and films (20, 21). These small-sized PNRs, however, do not necessarily generate ferroelectricity. When the film thickness t is much larger than the average PNR size ξ (Fig. 1A), PNRs are isolated in an insulating matrix. The depolarization field Ed in PNRs cannot be effectively screened and thus destabilizes the polarization of PNRs. On the other hand, as t is decreased the PNRs can play a vital role in the emergence of ferroelectricity (Fig. 1B). When t becomes comparable with or smaller than ξ, the electrical boundary conditions for PNRs drastically change as their interfaces come in contact with metallic electrodes and/or become exposed to surface adsorbates. The external charges screen the Ed by compensating for the polarization charge and thus can allow a switchable and stable polarization in PNRs. Such dimensional engineering of polarization stability in PNRs would provide an unconventional way to create and enhance ferroelectricity at reduced dimensions, distinct from methods such as strain (2628) and interface (13, 14) engineering.

Phase-field simulations were used to model polarization in a representative PNR region with a dimension of 7.5 nm embedded in the SrTiO3 film (23). The thickness of the SrTiO3 film in the simulations was varied within the range of 64 unit cells (or 25 nm) to 8 unit cells (3.2 nm) (23). We introduced external charges on the top and bottom SrTiO3 interfaces and investigated the stability of the PNR polarization after poling by an external electric field. After the electric field was removed and the system was relaxed to equilibrium, we evaluated the remnant polarization over the whole SrTiO3 film (by the average over the volume). We found that a decrease in the SrTiO3 thickness not only results in a more stable polar state but also greatly enhances the remnant polarization of the system (Fig. 1F). This implies a possible creation and enhancement of ferroelectricity at the reduced scale via the PNR-related mechanism.

We experimentally tested these theoretical predictions in strain-free (001) single-crystal SrTiO3 films (23), grown on a (001) SrTiO3 substrate with a single-crystal conductive oxide SrRuO3 bottom electrode (29). Film thickness was atomically controlled in the range of 120 to 6 unit cells (fig. S1). The nominally stoichiometric films had a normal unit-cell volume, almost the same as that of the substrate, indicating that the films were nearly free of excessive point defects (fig. S2). Only minute amounts of Sr deficiency (~1 atomic % at most) were present, possibly because of small formation energy (figs. S3 and S4) (18, 19). As mentioned above, Sr deficiency can naturally generate PNRs (Fig. 1, C to E) without compromising crystalline quality of the film, according to the theoretical calculations. Second harmonic generation and Raman spectroscopy measurements provide further evidence of the PNR formation in our SrTiO3 films (20, 21), with a nonpolar-to-polar transition at ~400 K (fig. S3).

To directly visualize the PNRs in our films, we used aberration-corrected scanning transmission electron microscopy (STEM) with high-angle annular-dark-field (HAADF) and annular-bright-field (ABF) techniques (30, 31). We detected the atomic positions of Sr, Ti, and O in the 12-unit-cell-thick film from the ABF-STEM image (Fig. 2A) and the simultaneously obtained HAADF-STEM image (23), which allowed us to determine the polar atomic displacements. The magnitude of atomic displacements in polar regions [Fig. 2A, (i)] was as large as ~0.1 Å [comparable with that of conventional ferroelectric BaTiO3 (24)] but was negligible in other regions [Fig. 2A, (ii)]. The measured magnitude of atomic displacements is comparable with the calculated average magnitude (~0.06 Å) of atomic displacements in our theoretical model (Fig. 1C). Polarization maps obtained from the Ti and O displacements (23) indicate the presence of PNRs with a size of a few nanometers, whereas no clear polarization pattern is observed in the SrRuO3 region (Fig. 2B). PNRs have a downward polarization state in the as-grown film, which is consistent with ferroelectric domain measurements made with piezoresponse force microscopy (PFM) (fig. S6). Following the theoretical modeling and calculations (Fig. 1, B and F), these PNRs may enable ferroelectricity in ultrathin SrTiO3 films, although they cannot generate a macroscopic net polarization in bulk or thick films.

Fig. 2 Atomic-scale imaging of polar nanoregions.

(A) Filtered ABF-STEM image, including all atomic positions. The (i) and (ii) regions are examples of polar and nonpolar regions, respectively, and are enlarged for clear view (insets). They are schematically drawn on the right; δTi and δO denote the atomic displacement of Ti and O, respectively. (B) Polarization vectors for each unit cell, estimated from atomic displacements in (A). Arrows denote the polarization direction; the stronger the polarization, the darker the arrow color. Strength of polarization is also expressed as a color map, ranging from white (weak) to red (strong).

Using a PFM approach, we found that the stable and switchable polarization could indeed be realized at room temperature in ultrathin SrTiO3 films (Fig. 3) (23). Bipolar domain patterns, similar to those generated in conventional ferroelectric BaTiO3 films (5, 32), have been created in the SrTiO3 films with a thickness of less than 60 unit cells by scanning the film surface with an electrically biased PFM tip. Bulk-like 120-unit-cell-thick SrTiO3 film did not show any PFM contrast after poling, which is consistent with the notion that PNRs are effective only in the thinnest films. The stability of the written domain patterns was thickness-dependent: Whereas the PFM contrast disappeared within a few minutes in the 60-unit-cell-thick film, the bipolar domain patterns were distinct and stable for several hours or more in thinner films (such as 12-unit-cell-thick film) that exhibited a polarization stability as good as that of ferroelectric 12-unit-cell-thick BaTiO3 films (fig. S8). The PFM result demonstrates that the switchable and stable polarization, which is the signature of ferroelectricity, emerges at room temperature in ultrathin films of otherwise nonferroelectric SrTiO3.

Fig. 3 Visualization of electrically written polarization states in strain-free SrTiO3 films of different thicknesses by means of PFM.

(A to D) Bipolar domain patterns of [(A) and (B)] 12-unit-cell-thick and [(C) and (D)] 60-unit-cell-thick SrTiO3 films. The upward polarization in (C) and (D) spontaneously relaxes within several minutes. The downward polarization in (C) and (D) is characterized by a weaker amplitude PFM signal. The 24- and 6-unit-cell-thick SrTiO3 films also show a bipolar domain pattern with clear contrast (23). (E and F) The thickest SrTiO3 film of 120 unit cells does not exhibit any contrast after domain writing. The PFM amplitude signal is weak, which is consistent with the nonpolar nature of the bulk-like SrTiO3 films.

Further evidence for room-temperature ferroelectricity in ultrathin SrTiO3 films was obtained directly from polarization hysteresis (PE) loops. The 24-unit-cell-thick SrTiO3 films showed a clear polarization hysteresis with nonzero remnant polarization, whereas bulk-like 120-unit-cell-thick SrTiO3 films showed no hysteresis with a paraelectric-like behavior (Fig. 4A). The measured PE curves included a strong nonlinear dielectric contribution. To eliminate this response and to enable a more accurate measurement, we used the double-wave method (33), giving a switched polarization ΔP = P+P of ~1.4 μC/cm2 for the 24-unit-cell-thick SrTiO3 films but ΔP ≈ 0 for the 120-unit-cell-thick films (Fig. 4B). Considering the low density of PNRs in our films, the measured ΔP value seems reasonable, but we believe that it can be increased through interfacial engineering for more efficient charge compensation (13, 14). Our PFM and hysteresis measurements unambiguously confirm room-temperature ferroelectricity in ultrathin SrTiO3 films as well as the enhancement of ferroelectricity at the nanoscale range in agreement with the theoretical predictions (Fig. 1).

Fig. 4 Polarization hysteresis in ultrathin SrTiO3 films.

(A) Polarization versus electric field curves measured at room temperature for 24- and 120-unit-cell-thick SrTiO3 films. We performed hysteresis measurements at the frequency of 20 kHz. (B) The polarization hysteresis of 24- and 120-unit-cell-thick SrTiO3 films at room temperature, measured by using the double-wave method with a triangular ac electric field of 10 kHz. (Inset) Schematic of applied waveform. We obtained the pure hysteresis component by subtracting the nonhysteresis polarization [up (U) and down (D) runs] from the total [positive (P) and negative (N) runs].

This study demonstrates that size reduction does not necessarily lead to the deterioration of ferroelectric properties but in fact could enhance them. Although this mechanism might be limited to the ferroelectrics with a relatively low value of polarization and relaxors, it provides a path toward devices with reduced dimensions in which ferroelectricity is coupled to other functional properties, such as two-dimensional conductivity (34), superconductivity (35), and magnetism (36). In particular, we envision nonvolatile devices with ferroelectric polarization controlling interfacial carrier concentrations. Not limited to SrTiO3, our approach can be applied to other perovskite dielectrics (37), in which PNRs are controlled through defect engineering, as well as artificially layered superlattices (38, 39), in which PNRs can influence multiple interfaces.

Supplementary Materials

www.sciencemag.org/content/349/6254/1314/suppl/DC1

Materials and Methods

Figs. S1 to S9

References (4053)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: This work was supported by the National Science Foundation (NSF) under Designing Materials to Revolutionize and Engineer our Future (DMREF) grant DMR-1234096. The research at University of Nebraska–Lincoln was supported by NSF through the Materials Research Science and Engineering Center (MRSEC) under grant DMR-1420645. S.Y.C. acknowledges the support of the Global Frontier Hybrid Interface Materials of the National Research Foundation of Korea funded by the Korea Government (2013M3A6B1078872). The work at Penn State is partially supported by Penn State NSF-MRSEC Center for Nanoscale Science grant DMR-1420620 and by NSF grant DMR-1410714. The work at Temple University was supported as part of the Center for the Computational Design of Functional Layered Materials, an Energy Frontier Research Center funded by the U.S. Department of Energy (DOE), Office of Science, Basic Energy Sciences under award DE-SC0012575 and used resources of the National Energy Research Scientific Computing Center (NERSC), a User Facility supported by the DOE Office of Science. S.H.O. and K.S. acknowledge the support of the Asian Office of Aerospace Research and Development (AOARD) under grant FA2386-15-1-4046, Brain Korea 21 PLUS project for Center for Creative Industrial Materials (grant F14SN02D1707), and the National Research Foundation (NRF) of Korea funded by the Korean Government (grant 2015R1A2A2A01007904). Raman studies at Boise State University have been supported by NSF under grant DMR-1006136. The work at University of California–Santa Barbara was partially supported by the MRSEC Program of the National Science Foundation under award DMR-1121053.
View Abstract

Navigate This Article