Articles

https://doi.org/10.1038/s41563-018-0034-4

Ultrahigh piezoelectricity in ferroelectric

ceramics by design
Fei Li   1,2,5*, Dabin Lin1,5, Zibin Chen3,5, Zhenxiang Cheng4,5, Jianli Wang   4, ChunChun Li1, Zhuo Xu2,
Qianwei Huang3, Xiaozhou Liao   3, Long-Qing Chen1*, Thomas R. Shrout1 and Shujun Zhang   1,4*

Piezoelectric materials, which respond mechanically to applied electric field and vice versa, are essential for electromechanical
transducers. Previous theoretical analyses have shown that high piezoelectricity in perovskite oxides is associated with a flat
thermodynamic energy landscape connecting two or more ferroelectric phases. Here, guided by phenomenological theories and
phase-field simulations, we propose an alternative design strategy to commonly used morphotropic phase boundaries to fur-
ther flatten the energy landscape, by judiciously introducing local structural heterogeneity to manipulate interfacial energies
(that is, extra interaction energies, such as electrostatic and elastic energies associated with the interfaces). To validate this,
we synthesize rare-earth-doped Pb(Mg1/3Nb2/3)O3–PbTiO3 (PMN–PT), as rare-earth dopants tend to change the local structure
of Pb-based perovskite ferroelectrics. We achieve ultrahigh piezoelectric coefficients d33 of up to 1,500 pC N−1 and dielectric
permittivity ε33/ε0 above 13,000 in a Sm-doped PMN–PT ceramic with a Curie temperature of 89 °C. Our research provides a
new paradigm for designing material properties through engineering local structural heterogeneity, expected to benefit a wide
range of functional materials.

F
erroelectric crystals possess spontaneous polarization that can plateau over the years. Alternative ways to further flatten the ther-
be switched by an electric field from one direction to another. modynamic energy profile have been proposed. For example, on
Among all known piezoelectrics, perovskite oxide ferroelec- the basis of phenomenological and first-principles calculations, the
trics exhibit the highest piezoelectric charge coefficient d (which authors of ref. 11 considered utilizing depolarization electric fields
quantifies the ability to transform force into electrical charge) generated by charged domain walls, and the authors of refs 8,12,13
and dielectric permittivity and thus are the materials of choice for suggested using external stresses. However, there have been no
many piezoelectric devices, including medical ultrasonics for imag- breakthrough advances to enhance the piezoelectricity of practical
ing, diagnostics and therapy, underwater acoustics and ultrasonic materials beyond the MPB approach.
motors, to name a few1–3. In perovskite ferroelectrics, the piezoelec-
tric coefficient d can be expressed as d =​  2PSQε (ref. 4), where PS is High piezoelectricity from local structural heterogeneity
the spontaneous polarization, Q is the electrostrictive coefficient Here, we propose to flatten the thermodynamic energy profile to
and ε is the dielectric permittivity. In contrast to Q and PS, ε can be enhance the piezoelectricity of perovskite ferroelectrics by utilizing
readily tuned for perovskite ferroelectrics; thus, an increase in d is interfacial energies through judiciously introducing local structural
generally associated with an increase in permittivity5,6. From ther- heterogeneity. This work is motivated by recent investigations by us
modynamic theory, the dielectric permittivity is associated with the and others on the classical relaxor ferroelectrics14,15 PMN–PT and
curvature of the free-energy density profile of a system with respect Pb(Zn1/3Nb2/3)O3–PbTiO3 (PZN–PT) crystals developed about 30
to the polarization (∂ 2G ∕∂P 2), where a small curvature (a so-called years ago, which suggested that local structural heterogeneity may
flat energy profile) corresponds to a high ε, consequently leading play an important role in the piezoelectricity of ferroelectrics16–20. To
to a high piezoelectric coefficient d7–10. The thermodynamic energy illustrate this concept, we consider a tetragonal ferroelectric domain
profile of a ferroelectric solid solution is expected to be flattened with locally introduced heterogeneous polar regions, as shown
when the composition approaches a morphotropic phase boundary in Fig. 1b. In such a system, the bulk energy of those local polar
(MPB), yielding both drastically enhanced ε and d, as depicted in regions, which is related to the local chemical composition, favours
Fig. 1a. Therefore, the long-standing approach to achieving a flat
energy profile has been to focus on the design of an MPB. However,
the orthorhombic phase (polar vectors along the diagonal P X Y or
PXY direction). Owing to the polarization discontinuity, interfacial
̄
there is a limit on the piezoelectricity that can be achieved at the energies, including electrostatic, elastic and gradient energies, are
MPB compositions. As shown in Fig. 1a, it is not possible to achieve present. The interfacial energies favour the transformation of those
an ideally flat energy profile at the MPB composition in which local regions to the tetragonal state (polar vector along PY), to mini-
∂ 2G ∕∂P 2 approaches zero at the polarization of the energy minima, mize the polarization discontinuity. Competition between the bulk
since the compositionally induced ferroelectric phase transition is and interfacial energy may lead to a highly flattened free-energy
first order in nature for most perovskite ferroelectrics1,4. This is the profile, as shown in cases II and III in Fig. 1b. If the impact of these
main reason why the piezoelectricity in ferroelectrics has reached a interfacial energies is sufficient, significantly higher piezoelectricity

1
Materials Research Institute, Pennsylvania State University, University Park, PA, USA. 2Electronic Materials Research Laboratory, Key Laboratory of the
Ministry of Education, Xi’an Jiaotong University, Xi’an, China. 3School of Aerospace, Mechanical and Mechatronic Engineering, University of Sydney,
Sydney, New South Wales, Australia. 4Institute for Superconducting and Electronic Materials, Australian Institute of Innovative Materials, University of
Wollongong, Wollongong, New South Wales, Australia. 5These authors contributed equally: Fei Li, Dabin Lin, Zibin Chen and Zhenxiang Cheng.
*e-mail: [email protected]; [email protected]; [email protected]

Nature Materials | VOL 17 | APRIL 2018 | 349–354 | www.nature.com/naturematerials 349

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles Nature Materials

a c
12
PbZr0.60Ti0.40O3 0% –40 –20 0 20 40

Relative dielectric permittivity ε/ε0 (×103)

Piezoelectricity increases as
PbZr0.55Ti0.45O3 5% Angle (deg.)
composition approaches MPB 10
Landau energy density
PbZr0.52Ti0.48O3 10%
15%
8 300 K
Approaching
y
MPB
6 200 K

x
Become flat 4 50 K
at MPB
2

PR PT PR
Polarization 0
0 50 100 150 200 250 300
Temperature (K)
b d
Case I Case II Case III
0.3
0%
5%
PXY PXY
10%
0.2 15%

Loss factor
Y

X
0.1
Free energy

PXY PY PXY PXY PY PXY PXY PY PXY 0.0

0 50 100 150 200 250 300
Increased impact of interfacial energies Temperature (K)

Fig. 1 | Phenomenological illustration of the enhanced piezoelectricity in a homogeneous ferroelectric system and a ferroelectric system with local
structural heterogeneity. a, Landau energy of a homogeneous system, where rhombohedral Pb(Zr,Ti)O3 is taken as the example (polarization variation is
̄ crystallographic plane). The two PR schematics represent the polarizations along the [111] and [1̄ 11]
plotted on a (110) ̄ directions, respectively, while PT is
along the [001] direction. As the composition approaches MPB, the curvature of the energy profile (at the polarization of energy minima (PR)) decreases,
leading to increased piezoelectricity. b, A heterogeneous ferroelectric system, in which heterogeneous polar regions (cyan) are embedded in a tetragonal
ferroelectric domain (yellow). The impact of interfacial energies increases from case I to III. In case I, the impact of interfacial energies is minimal. Cases
II and III show two possible states for the local regions when the impact of interfacial energies is prominent. c,d, Phase-field simulation of the transverse
dielectric permittivities (c) and loss factors (d) for the heterogeneous system depicted in b. The volume fractions of the heterogeneous regions are set to
be from 0% to 15%. For a specific volume fraction, the error bars represent the standard deviations of the calculated permittivities and losses, which are
obtained from five different random distributions of local polar regions. The detailed calculations are given in the Methods. Taking a 10% volume fraction
for example, the simulated microstructures are given in the inset of c, where the colour denotes the angle (degrees) between the polar vector and the
y direction. The green colour represents the polar vector along the y direction, while the red and blue colours represent the polar vectors along the x ȳ and
xy directions, respectively.

can be realized. For a specific volume fraction of ‘local regions’, a temperatures, a system with more heterogeneous regions exhibits
smaller size corresponds to a higher specific surface area and con- a higher dielectric permittivity. In addition, due to the enhanced
sequently a higher interfacial energy. Therefore, to achieve effective impact of the interfacial energies, a higher volume fraction of the
interfacial energies, nanoscale structural heterogeneity is desired. tetragonal phase is favoured at elevated temperatures (see inset of
Fig. 1c). Figure 1d shows the simulated dielectric loss factors with
Phase-field simulations. To validate this concept, we first performed respect to temperature, and reveals a distinctive signature for con-
phase-field simulations on the dielectric response of heteroge- tributions from the heterogeneous regions to the dielectric response
neous systems depicted in Fig. 1b. The thermodynamic parameters (that is, the enhanced dielectric loss factors at cryogenic tempera-
employed in the simulation are shown in the Methods, where the fer- tures), and a system with a larger volume fraction of heterogeneous
roelectric Landau energy (that is, bulk energy) of the locally embed- regions producing a higher loss maximum. The enhanced loss factor
ded heterogeneous regions favours an orthorhombic phase. As shown is associated with the polarization switching of heterogeneous regions.
in Fig. 1c, when we increase the volume fraction of the local hetero- As depicted in Fig. 1b (case II), local regions with the polarization
geneous regions, the room-temperature dielectric permittivity is
drastically enhanced. At low temperature, the Landau energy well is
̄
of P X Y may switch their polar vectors to the electric-field-favoured
polar direction PXY by applying an electric field along the X direction.
deep, and thus the impact of the interfacial energies associated with During this process, energy dissipation is expected since switching
the structural heterogeneity is minimal. With increasing temperature, between two stable states requires overcoming an energy barrier.
the Landau energy well becomes shallow, and the impact of the inter-
facial energies becomes prominent, giving rise to a flattened average Ferroelectric ceramics design
thermodynamic energy density profile for the whole heterogeneous To experimentally realize our proposed idea, we chose PMN–PT,
system as a function of apparent polarization. Therefore, at elevated which possesses the highest piezoelectric activity among all binary

350 Nature Materials | VOL 17 | APRIL 2018 | 349–354 | www.nature.com/naturematerials

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Nature Materials Articles
MPB ferroelectric systems21,22, and further enhanced its piezoelectric tetragonal side of the MPB region. An extraordinarily large d33 is
responses using the nanoscale structural heterogeneity introduced observed for 2.5Sm-PMN–29PT ceramics, reaching the order of
through a doping strategy. According to previous experiments on 1,500 pC N−1—the highest value ever reported in polycrystalline
Pb-based perovskite ferroelectrics, rare-earth dopants can introduce ceramics (to the best of our knowledge, the previously reported
effective random fields/bonds and/or change the ordering degree of highest d33 was in a Pb(Hf,Ti)O3–Pb(Ni,Nb)O3 (PHT–PNN) sys-
B-site cations23–25. Therefore, the addition of rare-earth dopants into tem26, being on the order of 970 pC N−1), which is significantly
PMN–PT is considered to be an effective approach to enhance its higher than that of commercial ‘soft’ PZT ceramics (for example,
local structural heterogeneity. We employed low-temperature dielec- PZT5H, PZT5), as shown in Table 1. To verify the large piezoelec-
tric experiments by comparing the measured losses with the signa- tric coefficient of 2.5Sm-PMN–29PT ceramics, we employed three
tures obtained from phase-field simulations to evaluate the impact different piezoelectric measurements: the Berlincourt method (d33-
of various rare-earth dopants on the local structural heterogeneity meter, which is a quasi-static measurement based on the direct
in 0.72PMN–0.28PT (PMN–28PT), and the results are shown in piezoelectric effect, with values given in Table 1); electric-field-
Supplementary Fig. 1. We down-selected the Sm-PMN–PT system induced strain based on converse piezoelectric effect (Fig. 2b);
to provide a detailed discussion, since this system exhibits the high- and the electrical impedance method following the IEEE Standard
est loss factor at cryogenic temperatures compared to other doped on Piezoelectricity (Fig. 2c). The minimal difference in d33 values
PMN–PT systems. We synthesized Sm-doped PMN–PT ceram- measured by the three methods is thought to be associated with the
ics through high-temperature solid-state processing (see details in contribution of domain-wall motion, which increases with increas-
the Methods). The Sm-doping concentrations are in the range of ing measurement field amplitude and/or decreasing measurement
1–3 mol%. The detailed dielectric and piezoelectric properties of the frequency27. Notably, the combination of large piezoelectric coef-
Sm-doped PMN–PT ceramics are listed in Supplementary Table 1. ficient (~1,250 pC N−1) and ultrahigh dielectric permittivity (free
In the following discussion, we focus on 2.5 mol% Sm-doped dielectric permittivity ~10,000, clamped dielectric permittivity
0.71PMN–0.29PT (2.5Sm-PMN–29PT) and 2.5 mol% Sm-doped ~3,600), together with a good temperature stability, is achieved in
0.69PMN–0.31PT (2.5Sm-PMN–31PT) ceramics, owing to their 2.5Sm-PMN–31PT ceramics. As shown in Fig. 2d, only a minimal
representative properties. piezoelectric variation of 11% over the temperature range of 0–90 °C
is observed. Of particular significance is that the 2.5Sm-PMN–31PT
Piezoelectricity of Sm-doped PMN–PT ceramics ceramics exhibit improved thermal stability of the dielectric permit-
Figure 2a shows that the piezoelectric coefficients d33 for the tivity compared to commercial PZT5H ceramics and PMN–30PT
PMN–PT system are significantly enhanced on Sm doping. By crystals over the temperature range of 0–60 °C (Fig. 2e). All of these
adding 2.5 mol% Sm, the pseudocubic-tetragonal phase boundary merits make the 2.5Sm-PMN–31PT ceramic a promising candidate
of PMN–xPT is shifted from the PT content of 0.35 to 0.28 (see for numerous electromechanical applications at room temperature,
Supplementary Fig. 2). Thus, the maximum piezoelectric response especially for high-frequency transducer arrays, where the ultra-
of 2.5 mol% Sm-doped PMN–xPT is observed at x =​ 0.29 on the high clamped dielectric permittivity in 2.5Sm-PMN–31PT ceramic

a b c
106
2.5Sm-PMN–29PT ceramic sE33 ~28.8 pm2N–1 306 kHz 100
Undoped 2.5% Sm doped ε33/ε0 ~13,000
0.03 PZT5H ceramic
d33 ~1,420 pC N–1
PZT8 ceramic
1,500 k33 ~78% 50
105
Impedance (Ω)
d33 (pC N–1)

Phase (deg.)
Strain (%)

–1
1,000 0.02 d33 = 1,530 pm V
0
500 4
10
0.01
0 –1 –50
d33 = 660 pm V
T

PM 4PT

T
8P

9P

0P

1P

PM 2P

6P

d33 = 210 pm V
–1
202 kHz
–2

–2

–3

–3

–3

–3

–3

0.00 103 –100

N

0.0 0.5 1.0 1.5 2.0 150 200 250 300 350
PM

PM

PM

PM

PM

Electric field (kV cm–1) Frequency (kHz)

d 0.04 e 0.8
0 °C 70 °C 2.5Sm-PMN–31PT ∆ > 100%
Variation of dielectric permittivity

2.5Sm-PMN–31PT
20 °C 80 °C PHT–PNN
40 °C 90 °C 0.6
0.03 PZT5H ∆ = 80%
60 °C Variation = 11%
PMN–30PT crystal
Strain (%)

0.4
0.02
∆ = 37%
0.2

0.01 ∆ = 27%
0.0
–1
1 kV cm
0.00 –0.2
–1 0 10 20 30 40 50 60
Electric field (kV cm /scale)
Temperature (°C)

Fig. 2 | Piezoelectric and dielectric properties of Sm-PMN–PT ceramics. a, The piezoelectric coefficient d33 of 2.5 mol% Sm-modified PMN–xPT and
PMN–xPT ceramics, measured using a Berlincourt d33-meter. b, Electric-field-induced strains of the 2.5Sm-PMN–29PT, PZT5H and PZT8 ceramics,
measured at 1 Hz. c, IEEE standard d33 measurements for a longitudinal bar (1.5 ×​ 1.5 ×​ 5.5 mm3) of the 2.5Sm-PMN–29PT ceramic. d, Electric-field-induced
strains of 2.5Sm-PMN–31PT at different temperatures. The frequency and amplitude of the electric-field are 1 Hz and 2 kV cm−1, respectively. e, Relative
variations of the dielectric permittivity with respect to temperature for 2.5Sm-PMN–31PT ceramics and three state-of-the-art piezoelectric materials (that
is, the commercial PZT5H ceramic, the [001]-oriented PMN–30PT crystal and the literature-reported PHT–PNN ceramic26 with d33 of 970 pC N−1).

Nature Materials | VOL 17 | APRIL 2018 | 349–354 | www.nature.com/naturematerials 351

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles Nature Materials

Understanding of high piezoelectricity in Sm-PMN–PT

Table 1 | Comparison of the piezoelectric coefficients and
To better understand the underlying mechanism(s) contributing to
relative dielectric permittivities of the 2.5Sm-PMN–29PT and
the high dielectric and piezoelectric properties, we investigated the
2.5Sm-PMN–31PT ceramics with PMN–36PT and state-of-the-
microstructures and temperature-dependent dielectric properties for
art ‘soft’ PZT ceramics
2.5Sm-PMN–31PT ceramics. It is found that the impacts of Sm on
Ceramics ε33/ε0 d33 (pC N−1) the density (Supplementary Table 2) and grain size (Supplementary
2.5Sm-PMN–29PT 13,000 1,510
Fig. 3) of the ceramics are minimal; thus, the high dielectric and
piezoelectric properties of Sm-doped PMN–PT are not due to the
2.5Sm-PMN–31PT 10,000 1,250
changes in the density and grain size associated with Sm-doping.
PMN–36PT 5,100 620 Figure 3a,b shows the synchrotron X-ray diffraction (XRD) results
PHT–PNN 6,000 970 for 2.5Sm-PMN–31PT. Below the Curie temperature (~377  K),
Commercial PZT5H 3,400 650 the patterns of the three main perovskite diffraction peaks (that is, 
<​100>​,  <​110>​ and <​111>​
) exhibit minimal variations, indicat-
Commercial PZT5 1,700 500
ing that no long-range ferroelectric phase transition occurs over
The piezoelectric coefficients d33 are measured by a Berlincourt d33-meter. The errors in the data this temperature range. On the basis of Rietveld-refinement analy-
are within ±​5%.
sis, 2.5Sm-PMN–31PT can be approximately treated as a tetragonal
phase with a minor orthorhombic component (detailed refinement
results and structural parameters are given in Supplementary Fig.
(that is, ~3,600, versus PZT5H: 1,400–1,700 and PMN–PT crystals: 4 and Supplementary Table 3). The volume fraction of the minor-
800–1,000) guarantees the electrical impedance matching, leading ity orthorhombic phase is found to increase from 8% to 24% as the
to enhanced sensitivity and broadened bandwidth, thus allowing temperature decreases from 310 K to 200 K, while maintaining almost
the miniaturization of transducers28. the same value over the temperature range of 20–200 K. The ferro-

a b
Measured data 100
2.5Sm-PMN–31PT 2.5Sm-PMN–31PT

Phase volume fraction (%)

Refinement results
80
<100> <110> <111>
Intensity (a.u.)

450 K 60 P4mm
Amm2
310 K
40 Pm-3m
200 K
20
20 K
TC
0
9.7 9.8 9.9 10.0 13.8 13.9 14.0 14.1 17.0 17.1 17.2 0 100 200 300 400
2θ (deg.) T (K)
c

1 μm

[100]
A-site cations
2 nm
[010] B-site cations

Fig. 3 | Microstructure of 2.5Sm-PMN–31PT. a, Synchrotron XRD and refinement results of the <​100>​, <​110>​ and <​111>​peaks for 2.5Sm-PMN–31PT
powders. The Rietveld-refinement parameters and full set of XRD data are given in Supplementary Fig. 4 and Supplementary Table 3. Note that 2.5Sm-
PMN–31PT is a heterogeneous ferroelectric system, in which considerable lower-symmetry (for example, monoclinic or triclinic) phases may exist due
to the presence of random fields, cation order–disorder and interfacial energies. For the Rietveld refinement, after various possibilities had been tried,
acceptable fitting results are obtained by approximating 2.5Sm-PMN–31PT as a mixture of tetragonal and orthorhombic phases. b, The volume fraction of
different phases as a function of temperature, obtained by Rietveld refinement. c, Dark-field TEM (upper left corner) and atomic-resolution TEM images
at room temperature, recorded along the crystallographic [001] direction. The polar vectors (arrows) are given for each unit-cell column in the atomic-
resolution TEM. The positions of the A-site and B-site atomic columns are indicated in the enlarged images on the right.

352 Nature Materials | VOL 17 | APRIL 2018 | 349–354 | www.nature.com/naturematerials

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Nature Materials Articles
electric domain structure of the 2.5Sm-PMN–31PT ceramic is shown the Sm modification. The phase-transition-unrelated dielectric loss
in the upper left corner of Fig. 3c, indicating a lamellar-type domain anomaly, which does not exist in classical ferroelectrics such as
configuration with a domain size of 80–300 nm. Further transmis- Pb(Zr,Ti)O3 (Supplementary Fig. 8), is believed to be associated with
sion electron microscopy (TEM) experiments show that the impacts the energy dissipation induced by the switching of nanoscale polar
of Sm on the domain structure and domain size of PMN–PT ceramics regions17,31,32. On the basis of the simulated results in Fig. 1d, a higher
are minimal if the doping does not change the macroscopic crystal- loss maximum in 2.5Sm-PMN–31PT indicates a higher volume frac-
line symmetry of the ferroelectric phase (see Supplementary Fig. 5), tion of heterogeneous polar regions (that is, a higher level of struc-
indicating that the ultrahigh piezoelectric properties in Sm-doped tural heterogeneity). Therefore, we believe that the addition of Sm in
PMN–PT ceramics are not attributed to the changes in domain struc- PMN–PT creates perturbations in the local order parameters such
ture induced by Sm-doping. The atomic structure within a tetrago- as polarization and strain, leading to heterogeneous polar regions
nal domain is characterized by high-resolution TEM and given in and additional interfacial energies, including polarization gradient
Fig. 3c. The positions of the A-site (Pb/Sm) and B-site (Mg/Nb/Ti) energy, elastic energy and electrostatic energy, when compared to the
atomic columns are indicated in the enlarged images on the right. undoped PMN–PT system. As a result, according to the phenom-
According to the positions, the atomic displacements are presented enological analysis and phase-field simulations presented in Fig. 1b,c,
as vectors pointing from the centre of a B-site cation to the centre of these additional energies flatten the averaged free-energy profile of
its four nearest neighbouring A-site cations. These atomic displace- the whole heterogeneous Sm-doped PMN–PT system, thus enhanc-
ments represent the magnitudes and directions of the polar vectors ing the piezoelectric and dielectric properties.
for each unit-cell column29. As shown in Fig. 3c, although most polar Note that the Curie temperature TC of 2.5Sm-PMN–31PT is 377 K
vectors are along the tetragonal [010] direction, orthorhombic-like (see inset of Fig. 4a), which is 70 K lower than that of PMN–36PT
nano-regions (polar vector along the [110] direction) are observed in (TC =​ 445 K). The lower TC of Sm-PMN–31PT may also contribute to
the tetragonal matrix (also see Supplementary Fig. 6). Here, the TEM the high room-temperature dielectric permittivity; however, it is not
and XRD results clearly reveal that 2.5Sm-PMN–31PT is a ferroelec- the dominant factor. The dielectric permittivities of the 2.5Sm-PMN–
tric system with local structural heterogeneity at the nanoscale and is 31PT and PMN–36PT ceramics with respect to the temperature ∆​T
analogous to the schematic picture hypothesized in Fig. 1b. (that is, T−​TC) are compared and given in Supplementary Fig. 9. At
Figure 4a,b shows the temperature dependence of the dielectric a specific ∆​T (in the ferroelectric phase, that is, ∆​T < 0), the dielec-
permittivity and loss factor for the 2.5Sm-PMN–31PT ceramic. The tric permittivity of 2.5Sm-PMN–31PT is >​ 80% higher than that of
PMN–36PT (0.64PMN–0.36PT) ceramic, which exhibits the high- PMN–36PT, indicating that the TC effect cannot explain the high
est room-temperature dielectric permittivity and piezoelectric coef- room-temperature dielectric permittivity of 2.5Sm-PMN–31PT. To
ficient in the PMN–xPT ceramic system, is selected for comparison. experimentally confirm this, we introduce isovalent A-site dopants
The 2.5Sm-PMN–31PT and PMN–36PT ceramics exhibit similar Ca2+ and Sr2+ in PMN–31PT, which only decrease the Curie tempera-
dielectric permittivities at 20 K, while 2.5Sm-PMN–31PT shows a ture TC but do not introduce additional random fields. As expected,
much larger increase in the dielectric permittivity with increasing no enhancements in the room-temperature dielectric permittivity
temperature. The relative dielectric permittivity ε/ε0 of 2.5Sm-PMN– and low-temperature dielectric loss are observed in these ceramics,
31PT is found to be ~10,000 at room temperature, which is much although the TC is decreased to 390 K (Supplementary Fig. 10).
higher than that of PMN–36PT (~5,100). In PMN–36PT, a long- From a materials design point of view, we suggest that a higher
range ferroelectric (tetragonal-to-monoclinic) phase transition degree of local structural heterogeneity may result in a higher
occurs at about 260 K (ref. 30), which effectively enhances its room- room-temperature dielectric permittivity but at the cost of tem-
temperature permittivity. Notably, no phase transition occurs in perature stability, as simulated in Fig. 1c. In 2.5Sm-PMN–xPT
2.5Sm-PMN–31PT below the Curie temperature, and hence, the (x >​ 28), a higher PT content will stabilize the majority tetragonal
high room-temperature dielectric permittivity of 2.5Sm-PMN–31PT phase, consequently lowering the degree of structural heterogeneity.
cannot be attributed to a ferroelectric phase transition. In addition, as Therefore, it is easy to understand that 2.5Sm-PMN–29PT shows a
shown in Fig. 4b, the maximum dielectric loss factor of 2.5Sm-PMN– higher room-temperature dielectric permittivity than 2.5Sm-PMN–
31PT (~6.5%) is greater than that of PMN–36PT (~4.3%). By study- 31PT, while 2.5Sm-PMN–34PT exhibits better temperature stabil-
ing the loss factor with respect to the dopant concentration for ity, as shown in Supplementary Fig. 11. This may help guide future
Sm-PMN–PT ceramics (Supplementary Fig. 7), the enhanced cryo- material design to achieve specific properties to meet the require-
genic loss factors in 2.5Sm-PMN–31PT are clearly associated with ments of different piezoelectric applications.

a 12.0 b 0.08
50
2.5Sm-PMN–31PT
Relative dielectric permittivity ε/ε0 (×103)

40 PMN–36PT
10.0
ε/ε0 (×103)

30
20 0.06
8.0
10
Loss factor

6.0 0
300 350 400 450 500
Temperature (K) 0.04
4.0
2.5Sm-PMN–31PT
PMN–36PT
2.0 0.02

0.0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Temperature (K) Temperature (K)

Fig. 4 | Temperature-dependent dielectric properties of the 2.5Sm-PMN–31PT and PMN–36PT ceramics. a, Relative dielectric permittivity. b, Loss factor.
The data were measured at 1 kHz.

Nature Materials | VOL 17 | APRIL 2018 | 349–354 | www.nature.com/naturematerials 353

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Articles Nature Materials

Perspective 16. Takenaka, H., Grinberg, I., Liu, S. & Rappe, A. M. Slush-like polar structures
This work successfully addresses the long-standing challenge in in single-crystal relaxors. Nature 546, 391–395 (2017).
17. Li, F. et al. The origin of ultrahigh piezoelectricity in relaxor-ferroelectric
synthesizing new piezoelectric ceramics. The record-high dielectric solid solution crystals. Nat. Commun. 7, 13807 (2016).
and piezoelectric properties of the newly designed ceramics are of 18. Manley, M. E. et al. Giant electromechanical coupling of relaxor ferroelectrics
practical importance for high-performance electromechanical appli- controlled by polar nanoregion vibrations. Sci. Adv. 2, e1501814 (2016).
cations. A new methodology for the design of doped piezoelectric 19. Phelan, D. et al. Role of random electric fields in relaxors. Proc. Natl Acad.
ceramics is provided in this work. We propose to first analyse the Sci. USA 111, 1754–1759 (2014).
20. Xu, G. Y., Wen, J. S., Stock, C. & Gehring, P. M. Phase instability induced
impact of dopants on the local structural heterogeneity for a specific by polar nanoregions in a relaxor ferroelectric system. Nat. Mater. 7,
ferroelectric system, to down-select effective dopants by measur- 562–566 (2008).
ing the dielectric loss at cryogenic temperatures, and then in-depth 21. Park, S. E. & Shrout, T. R. Ultrahigh strain and piezoelectric behavior in
research is focused on the processing and composition optimizations. relaxor based ferroelectric single crystals. J. Appl. Phys. 82, 1804–1811 (1997).
To this end, we would like to emphasize that although a relaxor-ferro- 22. Service, R. F. Materials science - shape-changing crystals get shiftier.
Science 275, 1878 (1997).
electric solid solution (PMN–PT) is used in this study, our proposed 23. Kleemann, W. Relaxor ferroelectrics: Cluster glass ground state via random
strategy to enhance the piezoelectricity by deliberately introducing fields and random bonds. Phys. Status Solidi B 251, 1993–2002 (2014).
‘local structural heterogeneity’ is not limited to relaxor ferroelectrics. 24. Samara, G. A. The relaxational properties of compositionally disordered
For instance, based on phase-field simulations, an enhanced piezo- ABO3 perovskites. J. Phys. Condens. Matter 15, R367–R411 (2003).
electricity can be achieved in PZT superlattices by making use of 25. Chen, J., Chan, H. M. & Harmer, M. P. Ordering structure and dielectric
properties of undoped and La/Na‐doped Pb (Mg1/3Nb2/3)O3. J. Am. Ceram.
the interfacial energies between rhombohedral and tetragonal PZT Soc. 72, 593–598 (1989).
layers, as given in Supplementary Fig. 12. It is expected that tailoring 26. Tang, H. et al. Investigation of dielectric and piezoelectric properties
of local structural heterogeneity can be used for optimizing a wide in Pb (Ni1/3Nb2/3)O3–PbHfO3–PbTiO3 ternary system. J. Am. Eur. Soc. 33,
range of functionalities in ferroic materials such as ferromagnetics, 2491–2497 (2013).
ferroelastics and ferroic composites, since many physical properties 27. Damjanovic, D. Stress and frequency dependence of the direct piezoelectric
effect in ferroelectric ceramics. J. Appl. Phys. 82, 1788–1797 (1997).
of a ferroic material are defined by second derivatives of a thermody-
28. Cannata, J. M., Williams, J. A., Zhou, Q., Ritter, T. A. & Shung, K. K.
namic energy with respect to a certain thermodynamic variable (that Development of a 35-MHz piezo-composite ultrasound array for medical
is, curvature of free-energy landscape)33–35. imaging. IEEE Trans. Ultrason. Ferroelectr. Freq. Contr. 53, 224–236 (2006).
29. Jia, C. L. et al. Unit-cell scale mapping of ferroelectricity and tetragonality in
Methods epitaxial ultrathin ferroelectric films. Nat. Mater. 6, 64–69 (2007).
Methods, including statements of data availability and any asso- 30. Noheda, B., Cox, D. E., Shirane, G., Gao, J. & Ye, Z. G. Phase diagram
of the ferroelectric relaxor (1-x)PbMg1/3Nb2/3O3-xPbTiO3. Phys. Rev. B 66,
ciated accession codes and references, are available at https://doi. 054104 (2002).
org/10.1038/s41563-018-0034-4. 31. Bokov, A. A. & Ye, Z.-G. Recent progress in relaxor ferroelectrics with
perovskite structure. J. Mater. Sci. 41, 31–52 (2006).
Received: 16 September 2017; Accepted: 5 February 2018; 32. Viehland, D., Jang, S. J., Cross, L. E. & Wuttig, M. Freezing of the polarization
Published online: 19 March 2018 fluctuations in lead magnesium niobate relaxors. J. Appl. Phys. 68,
2916–2921 (1990).
33. Nan, C. W., Bichurin, M. I., Dong, S., Viehland, D. & Srinivasan, G.
References Multiferroic magnetoelectric composites: Historical perspective, status,
1. Jaffe, B., Cook, W. R. & Jaffe, H. Piezoelectric Ceramics (Academic, and future directions. J. Appl. Phys. 103, 031101 (2008).
London, 1971). 34. Ji, Y. C. et al. Ferroic glasses. npj Comput. Mater. 3, 43 (2017).
2. Zhang, S. J. et al. Advantages and challenges of relaxor-PbTiO3 ferroelectric 35. Eerenstein, W., Mathur, N. D. & Scott, J. F. Multiferroic and magnetoelectric
crystals for electroacoustic transducers–a review. Prog. Mater. Sci. 68, materials. Nature 442, 759–765 (2006).
1–66 (2015).
3. Rödel, J. et al. Perspective on the development of lead‐free piezoceramics.
J. Am. Ceram. Soc. 92, 1153–1177 (2009).
Acknowledgements
F.L. and T.R.S. acknowledge the ONR support. F.L. also acknowledges the support
4. Lines, M. E., & Glass, A. M. Principles and Applications of Ferroelectrics and
by the National Natural Science Foundation of China (grant numbers 51572214 and
Related Materials (Oxford Univ. Press, Oxford, 1977).
51761145024) and the 111 Project (B14040). S.Z. acknowledges the support from
5. Damjanovic, D. Contributions to the piezoelectric effect in ferroelectric single
ONRG (N62909-16-1-2126) and ARC (FT140100698). L.-Q.C. is supported by the US
crystals and ceramics. J. Am. Ceram. Soc. 88, 2663–2676 (2005).
Department of Energy, Office of Basic Energy Sciences, Division of Materials Sciences
6. Li, F., Jin, L., Xu, Z. & Zhang, S. J. Electrostrictive effect in ferroelectrics: An
and Engineering under Award DE-FG02-07ER46417. Z.C. thanks M. Cabral from North
alternative approach to improve piezoelectricity. Appl. Phys. Rev. 1,
Carolina State University for the sample preparation guidance and discussions.
011103 (2014).
7. Fu, H. & Cohen, R. E. Polarization rotation mechanism for ultrahigh
electromechanical response in single-crystal piezoelectrics. Nature 403, Author contributions
281–283 (2000). The project was conceived and designed by F.L., S.Z., L.-Q.C. and T.R.S.; F.L. and
8. Wu, Z. & Cohen, R. E. Pressure-induced anomalous phase transitions and L.-Q.C. performed the phase-field simulations; F.L. and D.L. prepared ceramic samples,
colossal enhancement of piezoelectricity in PbTiO3. Phys. Rev. Lett. 95, and performed the dielectric and piezoelectric measurements; Z. Chen, Q.H. and X.L.
037601 (2005). performed TEM experiments; Z. Cheng., J.W. and C.C.L. performed XRD measurements;
9. Nahas, Y. et al. Microscopic origins of the large piezoelectricity of lead free F.L., S.Z., L.-Q.C. and T.R.S. wrote the manuscript, and all authors discussed the results.
(Ba,Ca)(Zr,Ti) O3. Nat. Commun. 8, 15944 (2017).
10. Liu, W. & Ren, X. Large piezoelectric effect in Pb-free ceramics. Phys. Rev. Competing interests
Lett. 103, 257602 (2009). The authors declare no competing interests.
11. Sluka, T., Tagantsev, A. K., Damjanovic, D., Gureev, M. & Setter, N. Enhanced
electromechanical response of ferroelectrics due to charged domain walls.
Nat. Commun. 3, 748 (2012). Additional information
12. Budimir, M., Damjanovic, D. & Setter, N. Piezoelectric response and Supplementary information is available for this paper at https://doi.org/10.1038/
free-energy instability in the perovskite crystals BaTiO3, PbTiO3, and s41563-018-0034-4.
Pb(Zr,Ti)O3. Phys. Rev. B 73, 174106 (2006).
Reprints and permissions information is available at www.nature.com/reprints.
13. Ahart, M. et al. Origin of morphotropic phase boundaries in ferroelectrics.
Nature 451, 545–548 (2008). Correspondence and requests for materials should be addressed to F.L. or L.-Q.C. or
14. Cross, L. E. Relaxor ferroelectrics. Ferroelectrics 76, 241–276 (1987). S.Z.
15. Kutnjak, Z., Petzelt, J. & Blinc, R. The giant electromechanical response in Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in
ferroelectric relaxors as a critical phenomenon. Nature 441, 956–959 (2006). published maps and institutional affiliations.

354 Nature Materials | VOL 17 | APRIL 2018 | 349–354 | www.nature.com/naturematerials

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.
Nature Materials Articles
Methods ∂Pi (r, t )
= −L
δF
, (i = 1, 2, 3) (1)
Sample preparation. A series of Pb(1−1.5y)Smy[(Mg1/3Nb2/3)(1−x)Tix]O3, with ∂t δPi (r, t )
x =​  0.28–0.36 and y =​ 0–0.03, was synthesized using the B-site cation precursor
method. The MgNb2O6 powders were prepared at 1,200 °C for 6 h. The Pb3O4, where L is the kinetic coefficient, F is the total free energy of the system, r is the
MgNb2O6, TiO2 and Sm2O3 powders were wet-mixed by ball-milling with space position and Pi (r, t ) is the polarization. δF ∕δPi (r, t ) is the thermodynamic
Y-stabilized zirconia balls for 24 h, and then the mixed powders were calcined driving force for the spatial and temporal evolution of polarization.
at 850 °C for 2 h. On milling with Rhoplex binder for 12 h, the powders were The total free energy of the system includes the bulk free energy, elastic energy,
formed into pellets by uniaxial pressing at 30 MPa, followed by binder burn-out electrostatic energy and the gradient energy:
at 550 °C and cold isostatic pressing at 300 MPa. The samples were sintered at
1,170–1,270 °C for 2 h. The average grain size of the samples was found to be
around 10 μ​m, as shown in Supplementary Fig. 3. The relative density of the F= ∫ [fbulk + felas + felec + fgrad ]dV (2)
ceramics was measured by the Archimedes method. All ceramics possess high V
relative densities ( >​ 95%), as shown in Supplementary Table 2. Silver paste
(DuPont 6160) was applied on both sides of the samples at 850 °C for 20 min to where V is the system volume, fbulk denotes the Landau bulk free-energy density,
form the electrodes. For the dielectric and piezoelectric measurements, felas is the elastic energy density, felec is the electrostatic energy density and fgrad is the
the samples were poled at 10 kV cm−1 and 20–60 °C for 10 min in silicon oil. gradient energy density.
All of the samples were characterized at least 24 h after the poling process. As for The bulk free-energy density is expressed by Landau theory:
most of the ‘soft’ PZT ceramics, ageing is minimal for the Sm-doped PMN–PT
ceramics, as shown in Supplementary Fig. 13. The variation of dielectric and
piezoelectric properties from sample to sample is found to be less than 5%, as fbulk = α1 (P12 + P22 + P32) + α11 (P14 + P24 + P34)
shown in Supplementary Tables 4 and 5. + α12 (P12P22 + P22P32 + P12P32)

Dielectric measurements. The temperature dependence of the dielectric + α111 (P16 + P26 + P36) (3)
behaviour was determined using an LCR meter (HP4980, Hewlett Packard) + α112 [P14 (P22 + P32) + P24 (P12 + P32) + P34 (P12 + P22)]
connected to a computer-controlled temperature chamber (high temperature:
home-made temperature controller; low temperature: Model 325 Cryogenic + α123P12P22P32
Temperature Controller, Lake Shore).
where α1, α11, α12, α111, α112 and α123 are the Landau energy coefficients.
Piezoelectric measurements. The piezoelectric coefficients were determined by In this work, the difference between the tetragonal ferroelectric domain
a combination of the impedance method (the IEEE Standard on Piezoelectricity), and local heterogeneous polar regions is described by spatially dependent
a quasi-static d33-meter (PM300, Piezotest) and electric-field-induced strain. For Landau coefficients. An established set of phenomenological parameters
the resonance method, the impedance spectrum, which was used to determine of a tetragonal relaxor-PT crystal (PZN–0.15PT)17 with adjusted Curie
the resonance and anti-resonance frequencies, was measured using an HP4294 temperatures were used in this study. We changed the volume fraction of
impedance analyser. The electric-field-induced strain was determined using a heterogeneous polar regions to study the corresponding impacts on dielectric
linear variable differential transducer driven by a lock-in amplifier (Stanford behaviours of the heterogeneous system. For the tetragonal domain, the Landau
Research Systems, model SR830). parameters are α1 = 2.38 × 105 (T −410)C−2 m2 N, α11 = 6.16 × 107 C−4 m6 N,
α12 = 3.37 × 108 C−4 m6 N, α111 = 3.42 × 108 C−6 m10 N, α112 = 1.02 × 109 C−6 m10 N
Synchrotron X-ray diffraction (XRD) measurements. In situ synchrotron XRD and α123 = −4.84 × 109 C−6 m10 N. For local polar regions, the Landau
experiments were performed at the Powder Diffraction Beamline located at the energy favours the orthorhombic phase, and the corresponding parameters
Australian Synchrotron (λ =​ 0.6885 Å; T =​ 20–450 K). Two sample set-ups were are α1 = 2.2 × 105 (T −560)C−2 m2 N, α11 = −2.67 × 108 C−4 m6 N,
used for this experiment: the cryostat sample set-up for T =​ 20–300 K, in which α12 = 2.67 × 108 C−4 m6 N, α111 = 3.33 × 109 C−6 m10 N, α112 = 3.33 × 108 C−6 m10 N
the samples were mounted in the cryostat within the sample holder (Al capillary); and α123 = 2 × 1010 C−6 m10 N, where T is the temperature in kelvin. Based on these
and the borosilicate capillary sample set-up for measurements above 300 K, in Landau parameters, the intrinsic dielectric permittivity of the heterogeneous polar
which only a small amount (<​1 mm deep) of material was loaded into the capillary regions and the tetragonal matrix are ~700 and ~2,000, respectively, at 300 K.
funnel. The patterns were analysed using the FULLPROF software. The parameters in the elastic, electrostatic and gradient energy density are assumed
to be uniform throughout the system, as detailed in ref. 17. In the simulations,
Transmission electron microscopy (TEM) experiments. Dark-field TEM we employed two-dimensional 128 ×​ 128 discrete grid points and periodic
observations were carried out using a JEOL 2100 microscope operated at 200 kV. boundary conditions. The grid space in real space was Δx = Δy = 1 nm.
High-resolution scanning transmission electron microscopy–high-angle annular The size of the local polar regions was set to 3–6 nm with random distribution. The
dark-field (STEM–HAADF) images were obtained using a JEOL ARM 200 STEM volume fraction of the local regions was set in the range of 0–15%. To calculate
equipped with a cold field-emission gun and a spherical-aberration corrector the dielectric properties, an a.c. electric field was applied to the system, and the
operated at 200 kV, at the Electron Microscopy Centre of the University of corresponding polarization was recorded at each time step. The amplitude of
Wollongong. All STEM images were Fourier-filtered using a lattice mask to remove the a.c. field was 105 V m−1. For a tetragonal domain, there are two independent
noise. Atomic positions were determined by fitting with two-dimensional Gaussian dielectric permittivities, that is, transverse dielectric permittivity ε11 (measured
peaks using the Ranger 2.4 script in Matlab. TEM specimens were prepared using perpendicular to the polar direction) and longitudinal dielectric permittivity ε33
tripod mechanical gliding until thickness fringes were observed at the sample edge. (measured along the polar direction). Both contribute to the dielectric permittivity
The samples were further milled with 2 keV Ar ions for 20 mins with ±​6° incidence of a polycrystalline ceramic. For a composition in proximity of the MPB, ε11
angles using a precise ion polishing system. To reduce contamination and the is generally much larger than ε33 and is the dominant factor for the apparent
thickness of the damaged layer, the precise ion polishing system was operated at permittivity of ceramics5. Furthermore, in this heterogeneous system, it is found
1 keV for 20 min and 0.5 keV for 30 min with the same incidence angles for a final that ε33 is insensitive to the heterogeneous polar regions. Therefore, the calculated
polishing. The thin areas used for STEM–HAADF imaging were measured to be dielectric permittivity shown in Fig. 1c is ε11. For a specific volume fraction,
approximately 20 nm. standard deviations of the calculated permittivities and loss factors were obtained
from five different random distributions of local polar regions.
Phase-field simulations. In the phase-field simulation of a ferroelectric system
with local phase fluctuation, the time-dependent Ginzburg–Landau equation was Data availability. The data that support the findings of this study are available
employed to describe the evolution of the polarization, upon request from the corresponding authors.

Nature Materials | www.nature.com/naturematerials

© 2018 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.