ETH Library

Kinetics of ferroelectric switching

in poled barium titanate ceramics:
Effects of electrical cycling rate

Journal Article

Author(s):
Kannan, Vignesh ; Trassin, Morgan ; Kochmann, Dennis M.

Publication date:
2022-09

Permanent link:
https://doi.org/10.3929/ethz-b-000569294

Rights / license:
Creative Commons Attribution 4.0 International

Originally published in:

Materialia 25, https://doi.org/10.1016/j.mtla.2022.101553

Funding acknowledgement:
178747 - Understanding and improving the long-term performance of ferroelectric ceramics (SNF)
188414 - Multifunctional oxide electronics using natural ferroelectric superlattices (SNF)

This page was generated automatically upon download from the ETH Zurich Research Collection.
For more information, please consult the Terms of use.
 Materialia 25 (2022) 101553

Contents lists available at ScienceDirect

Materialia
journal homepage: www.elsevier.com/locate/mtla

Kinetics of ferroelectric switching in poled barium titanate ceramics:

Effects of electrical cycling rate
Vignesh Kannan a,∗, Morgan Trassin b, Dennis M. Kochmann a
a
Mechanics & Materials Laboratory, Department of Mechanical and Process Engineering, ETH Zürich, Switzerland
b
Laboratory for Multifunctional Ferroic Materials, Department of Materials, ETH Zürich, Switzerland

a r t i c l e i n f o a b s t r a c t

Keywords: Our understanding of the rate-dependent electro-mechanical response of ferroelectric ceramics is incomplete –
Ferroelectrics primarily due to limited experimental data characterizing the material behavior at high rates (or short time
Hysteresis scales). Therefore, we experimentally study the effect of cycling rate on polarization switching in poled barium
Rate dependence
titanate (BaTiO3 ) ceramics across five orders of magnitude in cycling rate. Mechanical strain data as a function of
Domain wall
the applied electric field were collected across three orders of magnitude. We quantify the rate-dependent coercive
Electro-mechanical coupling
field, remanent and peak polarization, apparent permittivity and apparent piezoelectric coefficient, and actuation
strain. Results reveal a reduction in polarizability of the material with increasing rate and a strong asymmetry in
the electromechanical hystereses, which comes with differences in the rate dependence when loading parallel vs.
anti-parallel to the direction of poling. Supported by a simple model and ex-situ piezoresponse force microscopy,
we conclude that rate effects arise from the mobility of 90◦ domain walls and the competition between nucleation
and growth of domains. The asymmetric hysteresis highlights the importance of point charge and dipole defects,
which affect the domain wall kinetics and hence the rate effects of polarization switching through the competing
time scales of space charge migration, dipole reorientation, and domain wall activity.

1. Introduction occurs through the heterogeneous nucleation and growth of nano- to

micron-scale domains of constant dipole moment, separated by domain
Ferroelectric materials are among the most studied multi-functional walls. Understanding the dissipative processes driving domain evolution
active materials, which exhibit a strong coupling between electrical, me- is central to our ability to design and control the electro-mechanical re-
chanical, and thermal fields. This multi-physical coupling has made this sponse, which has remained an active subject of research since the 1950s
class of materials relevant for a wide range of applications including [20,21]. The coupling between mechanical and electrical fields in fer-
infrared detection [1,2], energy harvesting [3–5], acoustics, sensor and roelectric crystals is linear under small applied fields (the piezoelectric
transducer technologies [6], active structures [7–11], electro-optic tech- effect) but involves nonlinear, hysteretic processes when applying large
nology [12], and more recently in nano-scale devices such as memory electric fields, by which the permanent electrical polarization can be ir-
storage [13–15]. reversibly altered while inducing mechanical strain [22,23]. Conversely,
With historical origins in the classical phenomenon of pyroelectricity the application of large mechanical stresses can induce changes in elec-
[16], followed by the discovery of piezoelectricity [17], ferroelectric- trical polarization [24].
ity was discovered in the 1920s [18]. However, widespread industrial In-situ studies of polarization switching in bulk single-crystal fer-
and academic interest in ferroelectric materials did not begin until the roelectrics have offered useful but limited insight into the kinetics of
1940s with the identification of ferroelectric perovskite ceramics [16]. domain wall evolution and its effects on the electro-mechanical perfor-
While a broad range of ferroelectric materials exists with different mech- mance [25–29]. As may be expected, this problem proves to be consid-
anisms contributing to their spontaneous polarization, our focus is on erably more challenging when dealing with bulk polycrystalline ceram-
crystalline ferroelectric materials, specifically perovskites. At the atomic ics, which have significant technological relevance – despite the recent
scale, ferroelectricity in these crystals is characterized by a permanent rise in interest in thin-film technology. In materials with pure tetrag-
electrical dipole moment due to a non-centrosymmetric crystal structure onal phases such as barium titanate (BaTiO3 referred to as BTO) at
that remains stable below the Curie temperature [19]. Upon application room temperature, observed microstructures typically involve an intri-
of large macroscopic electric fields, switching of electrical polarization cate network of 90◦ and 180◦ domain walls. Depending on the applied

∗
Corresponding author.
E-mail addresses: [email protected] (V. Kannan), [email protected] (M. Trassin), [email protected] (D.M. Kochmann).

https://doi.org/10.1016/j.mtla.2022.101553
Received 16 August 2022; Accepted 23 August 2022
2589-1529/© 2022 The Author(s). Published by Elsevier B.V. on behalf of Acta Materialia Inc. This is an open access article under the CC BY license
(http://creativecommons.org/licenses/by/4.0/)
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

electrical/mechanical loads and their rates, polarization switching oc- coelastic response of ferroelectric ceramics [11] by characterizing both
curs through a competition between domain nucleation and the motion the ferroelectric behavior and the coupled mechanical deformation of
of 90◦ and 180◦ domain walls [30,31]. Recent experiments [31] hinted cantilever beam samples. In this study, we focus on polarization switch-
at a complex sequence of active mechanisms during polarization switch- ing and the resulting mechanical straining of the sample, and we do not
ing of bulk lead zirconate titanate (PZT) ceramics under different am- exploit the viscoelastic measurement capabilities of the setup. The indi-
plitudes of electric field. vidual components of the BES setup are illustrated in Fig. 1. Cantilever
As limited as present understanding of the rate-dependent switching beam samples of dimensions 40 mm × 3 mm × 1 mm are mounted be-
mechanisms across scales is the current ability to accurately model and tween custom-designed Macor clamps. The sample geometry was cho-
predict the rate-dependent switching response of ferroelectric ceram- sen due to its versatility for structural characterization of viscoelastic
ics. While equilibrium properties are well studied, robust models that properties. The sample surfaces (40 mm × 3 mm) are coated with a thin
capture the complex time-dependent phenomena across scales in bulk film electrode (tens to hundred nanometer thickness), and copper tape
ferroelectric ceramics are a rare find [32]. Existing models range from is glued to the Macor clamps at the contact area for the application of
phenomenological ones at continuum length scales [33–38] to micro- an electric field (see Section 2.2) through multi-strand copper wires sol-
mechanics-based phase field and lamination models [39–42] to molec- dered to each of these clamps. Surface mechanical strains are measured
ular dynamics calculations that probe domain kinetics at atomic scales using a stereo-camera system, and the images are processed using stereo-
[43,44] to electronic structure calculations of domain walls [45]. Most scopic DIC (see Section 2.3). Instrumentation setup, loading conditions,
models fail to accurately predict the effects of loading rate under large and data acquisition were controlled remotely using a Python script.
amplitude electro-mechanical loading. This is in part due to the current
lack of macroscopic hysteresis data across a wide range of time scales,
2.2. Electrical loading and electrical displacement measurements
and the present lack of knowledge of the microscopic kinetics of domain
walls. This study seeks to address the former problem and offers insight
Electrical loads were applied using a high-voltage amplifier (Trek
into the latter. Finite time scales associated with domain wall nucleation
10/10B-HS) capable of maximum output voltages of ±10 kV with a slew
and growth [25,26,28,31] imply that, at limiting rates of electric-field
rate greater than 700 V/μs and bandwidth of ∼ 20 kHz. The input to the
loading, the underlying governing mechanisms may change, resulting
amplifier was provided by a remotely-controlled arbitrary function gen-
in a change in electro-mechanical performance.
erator (Tektronix AFG31022). The waveform input to the high-voltage
The frequency-dependent response of bulk ferroelectrics ceramics
amplifier was amplified by a constant gain of 1000 V/V and applied to
has received some attention in experimental studies that aimed at
the sample through high-voltage cables soldered to electrodes on the
understanding macroscopic hysteresis effects [46–50] and frequency-
Macor clamps. The average electric displacement across the sample was
dependent coercive fields at different length scales [51,52]. Note that
measured using an in-house charge amplifier circuit with a gain of 105 .
all of those studies focused on lead-based ferroelectrics (most impor-
Details of the measurement circuit are described in Appendix B.
tantly PZT). Therefore, the available data on purely tetragonal ferro-
electrics (such as BTO) have remained largely unexplored, with some
exceptions [53–55]. More importantly, the kinetics of domain switch- 2.3. Actuation strain measurements: Digital Image Correlation
ing mechanisms that contribute to the macroscopic rate effects have
remained widely unknown despite significant efforts to improve the res- Conventional experimental methods have used electro-mechanical
olution in length and time scales for in-situ measurements [31,53–58]. sensors (e.g., Linear Variable Differential Transformers) and other opti-
We further note that a large volume of research has focused on thin-film cal interferometry-based techniques [67] to measure electro-mechanical
ferroelectrics, especially in recent times [59]. Advancements in manip- actuation strains in ferroelectrics. Since the unique sample geometry in
ulation and characterization techniques at small length and time scales our setup (Fig. 1) renders those measurement techniques challenging,
have opened up the range of functionalities and control for their de- we use stereoscopic DIC to extract spatially-resolved surface strains on
sign [60,61]. Research ranges from in-situ control of thin-film proper- the sample during electrical switching. Two cameras (Basler acA 4112)
ties for ferroelectric performance [62,63] to utilizing domain walls as with resolutions of 4096 × 3000 pixels and pixel sizes of 3.5 μm were
active nano-electronic components [64,65]. Nevertheless, the intricate mounted in a stereo-configuration at an angle of ∼ 15◦ with respect
domain kinetics in thin films, like in bulk ferroelectrics, have remained to each other. Camera optics were chosen to achieve lateral resolu-
widely unexplored. tions of ∼ 4 μm per pixel. Correlation-based displacement calculations
In this study, we experimentally investigate the effect of electric were performed using the commercial code Vic-3D (Correlated Solu-
loading rate on the nonlinear electro-mechanical response of poled BTO, tions). The fields of view achieved using the optics were approximately
using cyclic experiments under large-amplitude electric fields across five 20 mm × 3 mm.
orders of magnitude in cycling frequency. In addition, stereoscopic Dig- A suitable region of interest (ROI) within this field of view was cho-
ital Image Correlation (DIC) data offer insight into the corresponding sen such that the signal-to-noise ratio of captured images was well-suited
rate dependence of mechanical actuation strains (across three orders of for correlation. The ROI was chosen to be at least 5 mm × 2 mm to ob-
magnitude). The combination of these different measurements presents tain statistically significant strain measures for each experiment, while
a unique perspective on the rate-dependent mechanisms of ferroelec- decreasing the noise due to numerical differentiation of displacement
tric switching and their kinetics. Using ex-situ Piezoresponse Force Mi- data. For the displacement correlation, a subset size of ∼ 100–150 px
croscopy (PFM) measurements and limit cases from a simple empirical was chosen with a step size of 21 px, thus ensuring on the order of 103
model with linear switching kinetics, we present the micro-mechanical data points for each experiment. Strains were calculated from displace-
switching mechanisms that support our macroscopic measurements of ment data using a filter size of ∼ 300–400 px. To plot the mean and
the rate-dependent electro-mechanical behavior. standard deviations of time-resolved strain fields, the edges of the ROI
were ignored due to potential errors during numerical differentiation.
2. Experimental methods and materials The main objective of the DIC measurements was to extract the me-
chanical strains during polarization switching. Note that the 3D DIC data
2.1. Experimental setup: Broadband Electromechanical Spectroscopy only provides in-plane strains on the surface of the sample, 𝜀11 , 𝜀22 , and
𝜀12 (see Fig. 1 for coordinate axes), where the 𝑒1 − 𝑒2 -plane lies on the
We use a custom-built experimental setup named Broadband Elec- specimen surface and its axes are aligned with the specimen edges. To
tromechanical Spectroscopy (BES) [66], which was originally developed calculate out-of-plane strain component 𝜀33 along the direction of the
to investigate the effect of electrical polarization on the broadband vis- applied electric field, we use zero stress boundary conditions on the

2
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 1. Schematic of the Broadband Electromechanical Spectroscopy (BES) experiment.

amplitude of the this poling pulse was set to 3.5 MV/m for a duration of
103 s with rise and fall times of 25 s each. As the coercive field of barium
titanate is typically < 1 kV/mm, we assume that polarization saturation
was achieved during this step, which is also shown by the excellent re-
peatability of data. The profiles for the cyclic loading (referred to as the
switching step) were of constant slope to investigate and interpret the
data and multi-scale polarization switching kinetics as a function of a
constant loading rate.
The synchronization of data acquisition across all instrumentation
Fig. 2. Schematic of electric loading history used in experiments: an initial pol- is essential, as two independent measurements are used simultane-
ing step precedes each cycling experiment to ensure comparable initial condi- ously to characterize polarization switching and mechanical strains (see
tions of the poled sample. Every cycling step begins with electric field applica-
Section 2.1). Calibration of the stereoscopic DIC system was performed
tion in the direction opposite to poling (negative switching).
a priori to recording a dataset. Calibration was repeated multiple times
during an experimental cycle, as experiments utilizing a single setup and
surface of our sample, so that linear elasticity yields the same specimen were performed over a time span of multiple days.
1 − 2𝜈 Instrumentation setup and data acquisition were controlled remotely
𝜀33 = − (𝜀 + 𝜀22 ), (1) using a Python script. Before each experiment, parameters for loading
1 − 𝜈 11
and data acquisition were set and transferred to the arbitrary function
where 𝜈 is Poisson’s ratio. A value of 𝜈 = 0.23 was chosen for our mate-
generator, oscilloscope and cameras. All measurement systems were si-
rial (poled polycrystalline BTO) [68]. Strains inside the ceramic samples
multaneously triggered using a 5 V TTL pulse from an Arduino board
are sufficiently small for the assumption of linear elasticity to hold. We
(Fig. 1).
further neglect changes in Poisson’s ratio due to polarization switching
(any such changes would be significant only during switching, whereas
differences in Poisson’s ratio before and after switching are negligible 2.5. Data postprocessing
due to crystallography).
For various data, including the ferroelectric hysteresis and eletro-
2.4. Load histories and experiment control mechanical butterfly curves, we compute the slopes of measured hys-
teresis loops, which requires a series of filtering steps. First, a higher-
Fig. 2 shows schematically the electric load history for the experi- order low pass Butterworth filter was applied to all time-resolved sig-
ments used in this study. Each experiment was precluded by a pre-poling nals recorded from the oscilloscope. As the sampling frequencies were
pulse to ensure consistent initial conditions before each experiment. The at least six orders of magnitude larger than the electric field input fre-

3
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

quency, cut-off frequencies were set to low values (two orders of magni- In ferroelectric ceramics, the statistical variability of polarization evo-
tude smaller than the sampling frequency). The data was then fitted us- lution across samples relates primarily to variability in crystallographic
ing B-spline interpolation to obtain smooth curves for numerical differ- orientation distribution and defect distribution, influenced by the man-
entiation, using a central-difference scheme (with a window size equiv- ufacturing process. In this study, we investigate the effects of loading
alent to an electric field of 200 V/mm and step size of one data point). rate on the kinetics of polarization switching without considering vari-
Choosing a window size with respect to the electric field (and not in sam- ability in the intrinsic material microstructure. Hence, data presented
pling points) ensured that the slopes were comparable across different was collected from the same sample. The number of electrical load-
electrical cycling frequencies, irrespective of changes in numerical res- ing cycles was sufficiently low to avoid ferroelectric fatigue – observed
olution. This numerical differentiation scheme was applied for all data to occur beyond 103 cycles [71]. For reference, a subset of measure-
of this study. ments and loading conditions were performed on two alternative sam-
ples. While some quantitative differences were observed, the trends in
2.6. Piezoresponse Force Microscopy rate-dependent switching were consistent across all samples.

Piezoresponse Force Microscopy (PFM) is a modification of contact- 3.1. Electro-mechanical hysteresis: cyclic loading
based Atomic Force Microscopy, which applies a bias field to a conduc-
tive probe tip, so the local polarization is probed using the converse Fig. 3 presents the time histories of simultaneous in-situ measure-
piezoelectric effect [69,70]. We use PFM to spatially resolve the ferro- ments during a representative experiment at a cycling frequency of
electric domain configuration and obtain insight into the nature of mi- 2 × 10−3 Hz. The first row (Fig. 3(a)) shows the applied electric field his-
croscopic domain walls in the pre-poled samples for a mechanism-based tory, the second (Fig. 3(b)) shows the measured change in the volume-
interpretation of our macroscopic rate-dependent measurements. averaged electric displacement with respect to the initial state, and the
Samples were prepared for microscopy using mechanical polishing third (Fig. 3(c)) shows the time history of each in-plane strain compo-
down to 50 nm (colloidal silica suspension) surface finish. Care was nent on the sample surface, calculated from stereoscopic DIC data (mean
taken to minimize load on the sample during polishing. The samples values are marked by circles, while shaded bands present the standard
were mounted on a conductive chuck using silver paste for PFM imaging. deviation of the corresponding strain components calculated over the
Measurements were performed using a Bruker MultiMode atomic force region of interest on the sample). The electric field is applied along the
microscope at room temperature with an AC bias of 6 V. 𝑒3 -axis, while strains are measured in the 𝑒1 − 𝑒2 -plane. The out-of-plane
PFM is a surface-sensitive measurement, which is not necessarily rep- strain is calculated using Eq. (1). For all experiments presented in this
resentative of the domain structure throughout the volume. Therefore, study, a positive electric field refers to a field applied along the direc-
prior to PFM, samples were polished such that measurements were col- tion of pre-poling, and consequently a negative field is applied opposite
lected at least a few hundred micrometers below the surface of the sam- to the direction of pre-poling. In Fig. 3, 𝑡0 marks the time at which all
ple. The thus-obtained data allows us to acquire a qualitative picture of instrumentation was triggered. The first cycle (𝑡0 , 𝑡2 ) is ignored when
the domain configuration prior to the application of electric field (not presenting hysteresis data, which is a standard process that accounts for
to extract the quantitative statistics of domain wall distribution nor of the lack of knowledge of the initial polarization state of the sample. We
defects). note that the phase offset between the first cycle and the second cycle
is less that 1%, indicating the effectiveness of the poling step in ensur-
ing consistent initial configurations (close to a poled state) before each
2.7. Material
experiment.
We used commercially manufactured 3%-doped barium titanate
(BTO) ceramics (DelPiezo). The choice of composition was beyond the 3.1.1. Phenomenology of the electro-mechanical hysteresis
scope and control of the current study due to limited availability of high- Fig. 4 presents the electric displacement hysteresis (𝑑 − 𝑒-loop) and
quality BTO ceramic samples (and data regarding their composition). the electro-mechanical coupling hysteresis (𝜀 − 𝑒-loop) obtained from
SEM-EDS (Electron Dispersive Spectroscopy) analysis however showed the time-resolved data shown in Fig. 3. The first loop is also included
that the primary dopant was Ca2+ , commonly used for manufacturing in this figure. The green circle in the 𝑑 − 𝑒-loop marks the trigger time
BTO ceramics. 𝑡0 shown in Fig. 3. Notice that, while the repeatability of the 𝑑 − 𝑒-loop
The samples were pre-poled, and silver electrodes were fired on the across both cycles is excellent, the 𝜀 − 𝑒-loops indicate drift from the
front and back 40 mm × 3 mm surfaces of samples. Scanning Electron first to the second cycle. Correcting for this initial strain offset at each
Microscopy (SEM) measurements revealed grain sizes on the order of cycle significantly improves the repeatability of 𝜀 − 𝑒-loops. Hence, the
3 μm. The initial polarization state of a sample before cycling is gener- strains will be denoted by Δ𝜀33 to indicate that the strain loops are offset
ally unknown. In-situ investigations of the absolute polarization state of by the initial strain at the start of the cycle.
the sample are complex and may require specialized facilities. Instead, Fig. 4 also indicates the time sequence of the polarization and strain
we applied the pre-poling step, for which we have achieved peak elec- evolution by the arrows labeled sequentially from ○ 1 to ○.
8 The total time

tric fields of 3.5 MV/m (beyond which electric arcing occurs around the for one cycle in this experiment was 500 s (frequency 𝜈 = 2 × 10−3 Hz).
sample, as the experimental geometry does not allow for immersing the Starting from ○, 1 the 𝑑 − 𝑒-loop’s increased slope between ○ 1 and ○ 3

entire sample in a medium of high dielectric strength like silicone oil). indicates the maximum rate of polarization switching, which comes
By performing the pre-poling step at peak electric field over long times, with an initial contractile strain followed by strain reversal. At ○, 3 the

we ensure consistent initial configurations for every measurement. As polarization has approximately saturated with the peak polarization
the coercive fields reported for barium titanate [22] are significantly 𝑝(𝑝) measured at the maximum applied electric field, while the 𝜀 − 𝑒-
lower than the peak applied electric field, we assume that samples are loop becomes close to linear. Upon reversal of electric field, the rema-
close to being fully poled at the start of a measurement. nent polarization 𝑝𝑟 is reached at ○ 5 (at zero electric field). The analo-
gous electro-mechanical behavior is observed in the positive direction
(○5 to ○).
8
3. Results and discussion
While the calculation of polarization from our data is nontrivial with-
out knowledge of the in-situ domain structure, remanent and peak po-
Cyclic loads with constant electric field rates (cf. Fig. 2) were ap-
larization are computed from the constitutive relation
plied across five orders of magnitude of cycling frequencies (2 × 10−3 to
20 Hz). For each experiment, at least four complete cycles were applied. 𝒅 = 𝝐𝒆 + 𝒑, (2)

4
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 3. Time-resolved data for cyclic loading applied along the 𝒆3 -direction at a constant rate (frequency = 2 × 10−3 Hz). (a) Applied electric field, (b) change in
electric displacement Δ𝑑, (c) in-plane strains. Strain components are shown with their mean (solid circles) and standard deviations (shaded bands). The inset shows
a sample schematic with the coordinate axes.

where 𝒅 , 𝒆 and 𝒑 are, respectively, the (measured) electric displace- electric effect due to migration of charges within the material, and the
ment vector, the applied electric field vector, and the sought polariza- polarization evolution due to switching activity (i.e., nucleation and
tion vector. 𝝐 is the dielectric permittivity tensor. Using this relation, growth of ferroelectric domains). Leakage currents [72] are assumed
the remanent polarization is calculated as the electric displacement at small in comparison to electric displacement measurements in the non-
zero electric field, and the peak polarization is calculated by extrapo- linear polarization switching regime; yet, they can have an effect on the
lating a straight line to the electric displacement axis (shown in Fig. 4). extracted apparent permittivity (while ideal ferroelectrics are assumed
In the following, we present the difference in polarization between the to be perfectly insulating, most technologically relevant ferroelectrics
positive and negative directions as Δ𝑝𝑟 and Δ𝑝(𝑝) (also shown in Fig. 4). are semiconductors [73]).
The fact that Δ𝑝𝑟 ≠ Δ𝑝(𝑝) indicates that a small fraction of the switched Starting from 𝜖𝑟∗ ≃ 6500 at the onset of loading (state ○),
1 the appar-
polarization at saturation is reversible (discussed in Section 3). ent permittivity increases with increasing applied electric field by an
For a quantitative analysis, we extract from Fig. 4 the apparent rela- order of magnitude (state ○),2 where peak polarization switching oc-
tive permittivity (written in 1D for simplicity) curs. At maximum applied electric field (between ○ 3 and ○),
4 polariza-
( ) tion switching has slowed down and 𝜖𝑟∗ varies between ∼ 1000 and 2500
1 𝜕𝑑 1 𝜕𝑝
𝜖𝑟 =
∗
= 𝜖+ , (3) (during electrical loading in both directions). Overall, these values are
𝜖0 𝜕𝑒 𝜖0 𝜕𝑒 in the range of those typically reported for BTO [22] and the value of
where 𝑒, 𝑑, and 𝑝 are, respectively the applied electric field, electric dis- 1150 indicated by the manufacturer. The large variability is attributed to
placement, and polarization components in the direction of the applied the nature of the two measurements (dielectric constant measurements
electric field, and 𝜖0 = 8.854 × 10−12 F/m, the electrical permittivity of are performed within the linear piezoelectric regime at high frequen-
vacuum. Plotted in Fig. 5 along with the 𝑑 − 𝑒-loop, 𝜖𝑟∗ is a measure of cies) and numerical differentiation errors. Given these sources of error,
the material’s ability to polarize in response to an external electric field we can safely assume that the samples were driven to near saturation
(relative to vacuum). during switching with little domain activity at large electric fields (for
We use the term apparent, since multiple mechanisms contributing the experiments presented in Fig. 5). The behavior upon loading in the
to the polarization evolution, as evident from Eq. (3): the classical di- reverse direction (from ○ 5 and ○)
8 is qualitatively analogous. During re-

5
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 4. (a) Electric displacement hysteresis (𝑑 − 𝑒-loop), and (b) actuation strain hysteresis (𝜀 − 𝑒-loop) at 2 × 10−3 Hz electric cycling frequency. Labels 1 through 8
indicate the time sequence of measurements.

(𝑎)
Fig. 5. (a) Apparent relative permittivity 𝜖𝑟∗ (proportional to the slope 𝜕 𝑑∕𝜕 𝑒) and (b) apparent piezoelectric coefficient 𝑑33 (proportional to the slope 𝜕 𝜀33 ∕𝜕 𝑒) vs.
applied electric field 𝑒 at 2 × 10−3 Hz cycling frequency (positive electric fields corresponding to the pre-poled direction). Data is presented for one cycle. Number
markers correspond to those in Fig. 4.

versal of electric field, 𝜖𝑟∗ ≃ 4800 at zero field followed by a peak during metric hysteresis has been observed in ferroelectric ceramics [49,75–78]
switching. Note that the peak apparent permittivity in the positive and and is attributed to the kinetics of internal defects [23], which may arise
negative directions (○6 and ○) 2 are different, reflecting the asymmetry in from the manufacturing process [78] and is typically linked to doping el-
the 𝑑 − 𝑒-hystereis (and also the 𝜀 − 𝑒-loop). ements [49,79] and composition. (This will be discussed in more detail
Analogously, we extract from the 𝜀 − 𝑒-loop and illustrate in in Section 3.2.) Fig. 5 indicates that switching and mechanical actua-
Fig. 5 the apparent piezoelectric coefficient. tion is slightly preferred along the negative direction, i.e., opposite to
the direction of pre-poling in this case.
(𝑎) 𝜕𝜀33 𝜕𝜀
𝑑33 = = 𝑑33 + 𝑡 , (4) Fig. 5 also lets us introduce the coercive field, which is classically
𝜕𝑒 𝜕𝑒
defined as the electric field at which the polarization vanishes during
This measure entails the combined effect of the large-signal linear piezo- switching. However, since measurements of the nonlinear electric dis-
electric coefficient 𝑑33
∗ and the sensitivity of the switching-induced trans-
placements are relative, zero polarization is calculated by creating an
formation strain 𝜀𝑡 to the applied electric field. Consequently, when no offset of the 𝑑 − 𝑒-loop along the electric displacement axis by the mean
polarization switching occurs (at saturation), this is the same as the of the remanent polarization in the positive and negative directions.
large-signal piezoelectric coefficient. This is valid only under the assumption that the maximum electric field
Importantly, in tetragonal systems like BTO, spontaneous transfor- amplitudes are sufficiently high to cause polarization saturation during
mation strains are non-zero only for a 90◦ switching process, whereas the positive and negative switching steps. While this may be true at low
180◦ switching does not significantly alter the transformation strain. At cycling frequencies, intrinsic material and rate effects prevent the sam-
zero applied field (state ○),
1 the calculated 𝑑33 is ∼ 175 pC/N – which fits ple from achieving a completely poled state at moderate to high cycling
data reported by the manufacturer (140 pC/N) and in literature [74]. It frequencies. The result would distort calculations of coercive fields from
decreases rapidly during switching and reaches zero at the reversal of data and subsequent interpretations of rate-dependent switching. Using
strain. Notice a small increase in the apparent piezoelectric coefficient an alternative kinetics perspective, we define coercive field 𝑒𝑐 as the elec-
(𝑎)
between regions ○ 2 and ○ 3 in the negative direction (𝑑
33
≃ −280 pC/N). tric field at which the driving force on domain switching – and hence
At large electric fields, i.e., upon saturation, the apparent piezoelectric the apparent electrical permittivity – is maximal within a load cycle:
coefficient reduces to ∼ 150 pC/N. When loading in the opposite direc-
𝑒𝑐 = arg max 𝜖𝑟∗ (𝑒). (5)
tion, the apparent piezoelectric coefficient increases to ∼ 225 pC/N af-
ter peak switching at ○ 6 (slightly less than the observed ∼ 280 pC/N in This ensures a unique and consistent interpretation across loading rates
the negative direction, again reflecting the asymmetric behavior). Asym- in both the positive and negative loading directions. In the positive load-

6
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

ing direction, Fig. 5 reveals 𝑒𝑐 ≃ 0.4 MV/m. Assuming that the loading change in the slope of the back-switched polarization (followed by satu-
rates in Fig. 5 are sufficiently slow to achieve saturation, we can verify ration, Fig. 7) indicates that the former mechanism is dominant at lower
the definition (Eq. (5)) by comparison to the classical definition (cal- rates, while the latter may be dominant beyond frequencies of 0.1 Hz.
culating the coercive field as the electric field at which the change in This also points to a critical time scale for charged defect migration.
electric displacement Δ𝑑 is half the net difference in remanent polar- Fig. 8 presents the coercive field amplitudes along the positive and
ization Δ𝑝𝑟 ∕2). The comparison between the two definitions matches to negative directions (denoted by 𝑒+ 𝑐 and 𝑒𝑐 , respectively), and the mean
−

within 2%. coercive field normalized by that at the lowest rate. The coercive field
during negative switching increases by nearly 100% with increasing cy-
3.1.2. Rate effects in the electro-mechanical hysteresis cling rate. The coercive field in the positive direction, however, shows
The experimental setup and postprocessing steps described in the opposite trend, although the reduction in coercive fields is less pro-
Section 3.1.1 have been applied across the entire frequency range nounced. The mean coercive field shows an increase greater than 20%.
probed in this study. Fig. 6 shows 𝑑 − 𝑒-hystereses with the extracted (The asymmetry in the coercive field will be addressed in Section 3.2.)
apparent permittivity (top row), as well as the corresponding 𝜀 − 𝑒- Based on the interpretation that the coercive field is linked to the driv-
hystereses and the extracted apparent piezoelectric coefficient (bottom ing force for macroscopic polarization switching at peak switching rates,
row). At least three 𝑑 − 𝑒- and 𝜀 − 𝑒-hystereses are shown for each cy- we conclude that the (collective) mobility1 of macroscopic polarization
cling frequency (with the exception of the 𝜀 − 𝑒-loop at the lowest fre- switching decreases with an increase in loading rate.
quency of 𝜈 = 2 × 10−3 Hz). (Strain data were not collected beyond fre- To support this observation, we plot in Fig. 9 the maximum apparent
quencies of 1 Hz due limitations in time resolution of our cameras). permittivity normalized by the values at the slowest rate as a function
Since the cyclic electric fields were of constant slope (Fig. 3), the corre- of cycling frequency. The trends are qualitatively similar along both di-
sponding electric field loading rates are also indicated. Error bars in the rections albeit much more pronounced in the negative direction. While
𝜀 − 𝑒-data represent the standard deviation of strains measured within no significant change in the maximum permittivity is evident for fre-
the region of interest (ROI) on the sample. quencies below 10−2 Hz, a significant decrease of up to ∼ 40% in the
The peak apparent permittivity in the negative direction of the ap- negative direction and ∼ 20% in the positive direction is observed at
plied electric field axis gradually decreases with increasing cycling fre- higher frequencies. This confirms a decreasing polarizability of the ma-
quency. Consequently, the peak strains in this direction decrease with terial with increasing loading rate. The differences in the positive and
increasing cycling frequency. In addition, while the peak apparent per- negative directions can be explained as follows. The initial configura-
mittivity is relatively rate insensitive along the pre-poling direction (in tion for all frequencies was the same in the negative directions (viz.
comparison to the negative switching direction), the reversal of polariza- the pre-poled microstructure after the quasistatic pre-poling step). By
tion to the pre-poled state is preferred as the loading rate is increased (at contrast, loading in the positive direction succeeded the loading step in
least up to 0.1 Hz, beyond which the apparent permittivity decreases). It the negative direction, so that the initial configuration for the positive
is important to note that the positive switching step follows the negative direction varies based on the frequency used during loading in the neg-
switching step (not the pre-poling step), and hence the rate-effects dis- ative direction. Fig. 7 indicates that, while this effect is negligible at low
cussed here may be history-dependent. In other words, the microstruc- frequencies, it becomes appreciable at higher loading rates.
ture of the material prior to reversal of applied electric field (or the ref- Fig. 10 presents the peak apparent piezoelectric coefficient as a func-
erence configuration) is an important consideration while understand- tion of rate, and the data indicates a steady decrease of the peak ap-
ing rate effects along the positive switching direction. We also conclude parent piezoelectric coefficient for electrical loading along the negative
that the asymmetry of the hysteresis loops is a function of loading rate. direction (labelled 𝑑33∗− ), while the opposite (i.e., a steady increase) is ob-

Going beyond previous observations of asymmetric hysteresis [75–78], served for loading in the positive direction (labelled 𝑑33 ∗+
). Note that data
our results indicate that rate effects in polarization switching may be at higher frequencies are missing due to limitations in the captured DIC
enhanced or inhibited by the kinetics of extrinsic defects like vacancies data at high rates (which does not yield reliable data for taking deriva-
and defect dipoles. tives of measured curves).
For a more quantitative analysis of the cyclic data in Figs. 6 and 7 Fig. 11 shows the effect of loading rate on the net change in strain
presents the change in remanent Δpr and peak polarization Δp(p) across (𝛿𝜀33 = 𝜀max
33
− 𝜀min
33
) along each direction of applied electric field, as ex-
the full range of cycling frequencies. It is evident that increasing the tracted from experimental data. A steady decrease in maximum strains
loading rate makes the material less susceptible to polarization switch- is evident along the negative switching direction. (As we will argue be-
ing. A significant decrease with increasing frequency is apparent in both low, this decrease in strain is likely associated with greater residual 90◦
the remanent polarization (by ∼ 30%) and the peak polarization (by domain volume fractions at the end of the negative switching stage. On
∼ 17%). We conclude that intrinsic rate effects prevent the sample from the contrary, the maximum change in strains along the positive direc-
achieving complete switching at shorter time scales. The difference be- tions are relatively rate-independent, implying that complete switching
tween remanent and peak polarization also shows that the polarization occurred during the subsequent positive electric field step.)
state of the material changes during reduction from peak electric fields Fig. 12 presents the asymmetry of the 𝑑 − 𝑒-hysteresis as a function
to zero electric fields; i.e., some fraction of the switched polarization is of cycling frequency, showing both the internal bias field
reversible—we refer to this as ‘back-switching’ (see right-hand side axis |𝑒+
𝑐 | − |𝑒𝑐 |
−
in Fig. 7). We introduce the back-switched fraction as [49] 𝑒𝑏𝑖𝑎𝑠 = − (7)
2
( Δ𝑝r )
𝛿𝑝(b) = 1 − × 100 (6) where 𝑒+𝑐 and 𝑒𝑐 are the measured coercive fields along the positive and
−
Δ𝑝(p) negative directions, respectively, and the ratio of maximum apparent
Physically, back-switching is a measure of the stability of the sponta- permittivity values between the negative and positive directions. The
neous domain configuration towards the end of the hysteresis loop. Our bias fields indicate that the asymmetry of the hysteresis loops increases
data shows increased back-switching at higher rates, which is attributed gradually towards the negative direction at low cycling rates and rapidly
to two mechanisms: (i) reduction of domain wall mobility and a con- saturates with an increase in loading rate.
sequent increase in domain nucleation at higher rates, resulting in a
higher volume fraction of unstable domains; (ii) a lower fraction of de- 1
We use the term mobility here and in the following as a macroscopic analog
fect dipole reorientation at higher rates results in unstable domain con- of the ratio of velocity to driving force of domain walls on the microscale (see
figurations, effectively driving polarization switching towards the pre- Appendix A.3). It is a measure of the sensitivity of ferroelectric domain switching
poled direction as observed in the hysteresis loops (see Fig. 6). The sharp with respect to an applied electric field.

7
 V. Kannan, M. Trassin and D.M. Kochmann
8

Materialia 25 (2022) 101553

Fig. 6. Effect of the electric loading rate on the 𝑑 − 𝑒 and 𝜀 − 𝑒 hysteresis loops (positive electric fields corresponding to the pre-poled direction).
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 7. Effect of cycling rate on remanent, peak, and back-switched polarization.

Fig. 10. Effect of cycling rate on the apparent piezoelectric coefficient.

Fig. 8. Effect of cycling rate on the coercive field.

Fig. 11. Effect of cycling rate on the net change in mechanical strain.

Fig. 9. Effect of cycling rate on the maximum apparent permittivity. Fig. 12. Effect of cycling rate on asymmetry of 𝑑 − 𝑒 hysteresis.

9
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 13. Numerically predicted electromechanical response based on different assumptions for 180◦ vs. 90◦ domain wall activity in the model of Appendix A. The
increase in volume fraction of 90◦ domains results in greater contractile strains, while 180◦ switching causes very little strain. The apparent permittivity decreases
with decreasing 90◦ domain wall activity. The arrow on the electric field axis indicates direction of the applied electric field.

3.2. Microstructural mechanisms of domain switching and rate effects three markers represent 90◦ switching phenomena with varying criti-
cal volume fractions (𝜗𝑐𝑟𝑖𝑡 ) for sequential switching (see the legend in
Multiscale measurements to probe domain wall kinetics under static Fig. 13(a)). The choice of 𝜗𝑐𝑟𝑖𝑡 allow us to simulate three cases, in which
[25,27,80] and quasistatic conditions [81,82] have revealed polariza- the nucleation of 180◦ domains within pre-existing 90◦ domains (for
tion reversal by sequential 90◦ or 180◦ domain activity – with greater reversal of polarization) is progressively delayed with increasing 𝜗𝑐𝑟𝑖𝑡 .
evidence for the former, especially in polycrystals. At short time scales, Fig. 13(a-c) show the normalized 𝑝 − 𝑒-, 𝑑 − 𝑒-, and 𝜀 − 𝑒-data, while (d-
the grain orientation distribution and extrinsic defects favor sequential f) show the volume fraction evolution for the three types of domains
90◦ over 180◦ switching – as confirmed by recent experiments [31]. vs. the applied electric field. Based on the rationale that sequential 90◦
The growth of domain walls at short time scales have also revealed switching has a lower energy barrier than single-step 180◦ switching
evidence for stick-slip-type kinetics in single-crystal BTO (observed as [26,31,80], the electric field for nucleation of initial 90◦ domains is
‘Barkhausen pulses’) [26,83–85], which indicates rate effects in macro- set to a lower value than for 180◦ domains, and the mobility of do-
scopic polarization switching. Unfortunately, a full micromechanical un- main walls during 90◦ switching is higher. As a consequence, the 180◦
derstanding is amiss [20] due to few in-situ measurements to probe the switching case shows the lowest apparent permittivity and insignificant
rate-dependent kinetic evolution of domain walls [86], especially so in contractile strains (relative to the other three cases), as no transforma-
polycrystalline ferroelectrics – with the exception of in-situ synchrotron tion strain is associated with this mechanism. For 90◦ switching, the
X-ray diffraction studies [53–55]. The latter studies have offered un- apparent permittivity (i.e., the slope of the 𝑑 − 𝑒-curve in Fig. 13(b))
precedented insight into domain volume fraction evolution in polycrys- decreases, and contractile strains (Fig. 13(c)) increase with decreasing
talline tetragonal ferroelectrics, and offer great opportunity to extract 90◦ domain wall activity (and higher 90◦ domain wall fractions during
detailed kinetics data. However, modifications of the technique [58], switching).
and combination with other spectroscopy-based methods [66,87,88] are While Fig. 13 presents the effect of varying domain structures on
critical to understand the effects of 180◦ domains and domain wall nu- macroscopic response, these calculations assume that the initial config-
cleation rates. uration is a single-crystal oriented along the direction of applied electric
To shed light onto the microstructural origin of the rate dependence field. This is different from the ceramic samples in experiments, whose
in our experimental data, we use a simple, qualitative model for ferro- initial poled state contains complex domain patterns due to polycrys-
electric switching along with ex-situ microscopy, which allows us to un- talline texture and defects. Fig. 14 presents two Piezoresponse Force
derstand the microstructural evolution of 180◦ vs. 90◦ domain patterns Microscopy (PFM) datasets from pre-poled samples, which allow us to
across the investigated range of frequencies. The model is presented in identify the domain wall distribution in the sample [70]. Images reveal
detail in Appendix A. In a nutshell, we consider a simplified 2D scenario, an abundance of 90◦ domain walls (laminar microstructure indicated
in which 180◦ polarization reversal under an applied electric field can by red arrows) in the sample at zero applied electric field, which can
occur in a number of ways, via nucleation of 180◦ domains and sequen- be expected, as the formation of domain patterns (specifically 90◦ do-
tial 90◦ domains, and variants thereof. Of critical importance is the evo- main walls) is a dominant mechanism for poling a grain whose sponta-
lution of the mechanical strain during switching, which – better than neous polarization vector is not aligned with the poling direction. While
the average polarization – can help distinguish between 180◦ and 90◦ care was taken to minimize mechanical load during polishing, some fer-
switching. roelastic domain evolution is inevitable. Hence, future studies require
To highlight the distinct signature of microstructural mechanisms in elimination of mechanical pre-processing steps for reliable quantitative
the average macroscopic measurements, we compare in Fig. 13 model measures of microscopic domain switching.
predictions for different domain switching mechanisms. Open circles The pre-poled microstructure implies that polarization reversal likely
represent switching purely by 180◦ domain wall activity, while the other occurs by the growth and shrinkage of these pre-existing 90◦ domains.

10
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 14. Piezoresponse Force Microscopy (PFM) data from pre-poled samples. (a,b) Two different tip-sample configurations are shown (inset). Red arrows indicate
characteristic stripe-like 90◦ domain walls. As expected, data shows a larger fraction of 90◦ domain walls in the pre-poled state. Left images show the in-plane signal,
which characterizes one in-plane component of the polarization (torsion of the cantilever is maximum when the polarization points along the 𝑦(𝑡) direction and 𝑥(𝑡)
direction in a, b, respectively), while the right ones present the out-of-plane component of the polarization (note that a small component may arise from buckling
from in-plane polarized domains with a polarization pointing along 𝑥(𝑡) and 𝑦(𝑡) , respectively along with deflection of the cantilever from domains with polarization
pointing along the 𝑧(𝑡) direction). Scale bars are in units characterizing the photodiode output used for data collection. Note that a single scan does not fully resolve
the 3D polarization vectors [89].

Fig. 15. Numerical results describing the effects of the kinetic mobility coefficient 𝜇 on the electro-mechanical response (for switching by 90◦ domain wall activity).
The arrow on the electric field axis indicates the direction of the applied electric field.

This has also been observed by in-situ X-ray diffraction [31]. With in- presented in Figs. 9 and 10. The latter shows a steady decrease of the
creasing loading rate, the polarizability of the sample decreases (as seen peak apparent piezoelectric coefficient with increasing rate for electrical
in Figs. 8 and 9). Fig. 15 presents results of the model for 90◦ switching loading in the negative direction. (Note that the numerical model can-
by varying the (collective) mobility 𝜇 of domains (see Appendix A.3). A not predict the limiting polarization due to intrinsic material or kinetic
decrease in apparent permittivity is observed with decreasing mobility. effects. The limiting polarization at the least mobility in Fig. 15 is due
While the overall slope of the 𝜀 − 𝑒-curve decreases, maximum strains to the limited calculation time; for longer times the sample would be al-
remain the same, as the final polarization state is constant across all lowed to reach saturation.) We hence conclude that the collective mobility
cases. Both these observations are consistent with our experimental data of 90◦ domains decreases with increasing loading rate.

11
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. 16. Numerical results describing the effects of initial polarization state on electro-mechanical behavior (for switching by 90◦ domain wall activity). The arrow
on the electric field axis indicates direction of applied electric field.

In summary, polarization switching (along the negative direction) volume fractions for domains were calculated from these initial polar-
begins by the motion of pre-existing 90◦ domain walls. As the electric ization values and driven to saturation by applying an electric field at
field is increased, 90◦ domain wall activity is likely to increase along constant rate in the positive direction. Results show a significant in-
with some 180◦ domain wall activity. At higher loading rates, the mo- crease in the net change in strain and a slight decrease in coercive fields
bility of 90◦ domain walls limits domain wall activity associated with with decreasing initial polarization (corresponding to greater initial 90◦
an increase in coercive fields (Fig. 8). Reduced peak strains at higher domain volume fraction). Both these trends are reflected in the exper-
loading rates indicate enhanced 90◦ domain activity towards the end of imental data. An increasing loading rate corresponds to a decrease in
loading, which is supported by the reduced remanent and peak polar- residual polarization (Fig. 7) after negative switching, and a subsequent
ization measurements at higher rates. increase in the net change in strain (𝛿𝜀+ 33
in Fig. 11) as well as a slight
The limiting mobility of 90◦ domain walls at higher loading rates reduction in coercive field along the positive direction (Fig. 8). From
could enhance the nucleation rate of additional 90◦ and 180◦ domain a microscopic viewpoint, this reduction may occur by the growth of a
walls. Using the Rayleigh model for domain switching, it was shown larger fraction of domains or by the nucleation of additional domains
[90] that, while applying sub-coercive electric fields, the presence of to reorient the polarization to the pre-poled state. Our rate-dependent
180◦ domain walls introduces piezoelectric nonlinearity in the second experimental data hence suggests that, as the loading rate is increased, a
harmonic of the strain. An interesting implication of that study was larger density of domain walls becomes mobile, which may also explain the
the correlation between the stability of the polarization state to do- reduction in coercive field. Notice, however, that the net change in strain
main wall content; specifically, an unstable polarization state was at- (Fig. 11) decreases slightly at 10−1 Hz, whose origin is unclear but may
tributed to higher 180◦ domain wall density. Our data (Figs. 6 and 7) be related to 180◦ domain wall activity (which has little effect on the
indicates that the domain configuration becomes less stable at higher macroscopic strain; see Fig. 13). The strain data at 1 Hz is limited due
rates. Combining this with our conclusions of reduced 90◦ domain wall to the number of sampling points (limited temporal resolution of the
motion, this implies more significant 180◦ domain wall activity at higher imaging system).
rates. Our measurements also show strong evidence for an evolving asym-
Further, the aforementioned study [90] found that a greater fraction metry in the 𝑑 − 𝑒- and 𝜀 − 𝑒-hysteresis loops with loading rate. Such
of unstable domains could imply a higher density of 180◦ domain walls. asymmetry of the hysteresis loop has been reported experimentally be-
This may be used in the interpretation of our data during switching along fore [75,91–94]. The nature of asymmetry has broadly been described as
the positive direction. As compared to the negative poling direction, the a deformation [95] or an asymmetric displacement [93] of the hystere-
coercive field decreases significantly less with increasing loading rate sis loop. In contrast to the classical phenomenon of aging which occurs
(Fig. 8), which also holds for the apparent permittivity (Fig. 9). The re- under no applied electric field, a compelling argument for this asymme-
duction in coercive fields indicates that, while domain evolution is less try is the presence of defect dipoles and space charges due to doping of
mobile, lesser driving force is required for switching along the positive ferroelectrics.
direction. The relative insensitivity of strain to cycling frequency along Defect dipoles are caused by acceptor dopants at the B-site of the per-
the positive direction (relative to the negative switching direction) con- ovskite crystal (Ti4+ site) [79]. These defects tend to align with the po-
firms our hypothesis that complete switching to saturation occurs pri- larization direction and stabilize the domain configuration in the poled
marily by 180◦ domain switching (Fig. 11). In addition, this implies that state. The reorientation of defect dipoles under applied electric fields di-
domain nucleation is preferred over growth at shorter time scales. rectly relates to the kinetics of pinning and depinning of domain walls.
To test this hypothesis, Fig. 16 presents numerical calculations with Some reports indicate that the time scales required for dipole reorien-
four different initial polarization values 𝑝𝑖𝑛𝑖𝑡 (all other parameters kept tation are on the order of seconds at room temperature [96]—although
constant) to mimic the different residual polarization states at the end of this relies on the specific composition and density of defects in the sys-
the negative switching step, as observed in experiments (Fig. 7). Initial tem. Point space charges are commonly found in the form of oxygen

12
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

vacancies in perovskites [97] and have been associated with the sta- 4. Conclusion
bilization of domains [94,98]. The rationale for this mechanism lies
in charged domain walls (CDWs), which allow for relaxation of elec- We have presented an in-situ experimental study of the effects of
trostatic charge compatibility constrains across the interface. The for- electric cycling rate on the macroscopic switching response of poled
mation of CDWs has been reported to be energetically favorable [99], BTO ceramics. Our data present evidence for a strong rate dependence
especially in bulk ferroelectrics with electrodes. A crucial criterion for in the polarizability, mechanical actuation, and symmetry of the electro-
the stability of CDWs is the presence of compensating charges around mechanical hysteresis. An in-depth understanding of the interaction
the bound charges on the CDW, also known as screening charges. A between domain walls, charged defects, and their respective kinetics
higher concentration of vacancies induced by doping concentrations through hybrid in-situ experiments and models is crucial to capture
[91] improves the potential stabilization of CDWs [98,99] and has been the effect of electric field loading rate on the macroscopic electro-
reported to suppress switching [100,101] due to pinning of domain mechanical response of ferroelectric ceramics and to inform meso- to
walls. macroscale models for ferroelectrics.
SEM-EDS analysis showed that the primary dopant in our samples is The following are key conclusions obtained from this study:
Ca2+ . The substitution of these ions at the Ba2+ A-site and Ti4+ B-site
are known to have different effects on defect generation. The former is 1. Polarization and mechanical hysteresis are strongly asymmetric
known to have little effect on the generation of charged defects (due to in poled, Ca-doped BaTiO3 ceramics. This asymmetry—classically
an isovalent substitution), while the latter results in the formation of de- linked to the existence of dipole defects in doped ferro-
fect dipoles [102]. It is safe to assume that two types of defects—elastic electrics—evolves with electric field loading rate potentially due to
dipole defects and point defects in the form of oxygen vacancies—are domain wall stabilization and pinning by point defects at long time
present in our samples. scales and dipole defects at shorter time scales.
Defect dipoles are known as one major cause for asymmetric hystere- 2. Remanent and peak polarization are inversely related to loading
sis [79,103], which is also observed in our experiments (Figs. 6 and 7). rate—decreasing by, respectively, ∼ 30% and ∼ 17% from 10−3 to
However, prior studies have linked asymmetric hysteresis to an asym- 10 Hz. This confirms incomplete switching at shorter time scales.
metric distribution of space charges [101,104], which, in turn, was at- The back-switched polarization increases as a function of loading
tributed to internal bias fields, while a symmetric (or pseudo-random) rate (from 18% to over 30%), and indicates less stable spontaneous
distribution of charges was linked to constriction of hysteresis. Given domain configurations at shorter time scales. This reduced stability
that our samples were pre-poled during manufacturing, as well as poled is attributed to a decreasing 90◦ domain wall activity, followed by
prior to every experiment, we assume the existence of internal bias fields enhanced nucleation of 180◦ domains.
[78] and hence an asymmetric charge distribution. The relative domi- 3. The mean coercive field increases by ∼ 20% within the probed fre-
nance of these defects in our samples may be resolved by changing the quency range. This implies an increase in driving forces required for
direction of poling in our cyclic experiments. Direction-specific poling polarization switching. While space charges are known to increase
and switching experiments (not part of this study) showed that, while the coercive fields, the time scales of applied electric fields were too
the degree of asymmetry was reduced, some asymmetry still persisted. short for long-range vacancy migration – particularly at high fre-
Hence, the presence of dipole defects may not be eliminated. quencies. We hence present two hypotheses for this increase in coer-
Note that all experiments presented in this study were performed on cive field: (i) intrinsic rate-dependent mobility of 90◦ domain walls,
the same sample to avoid sample-to-sample variability in vacancy con- and (ii) clamping of domain walls due to local dipole defects.
centration. Moreover, the total number of cycles over the entire exper- 4. Asymmetry in the coercive field—referred to as internal bias
imental run was not sufficiently high to induce fatigue [71]. We hence field—increases with increasing loading rate (by ∼ 0.2 MV/m across
assume similar concentration and spatial distribution of space charges the probed frequencies). Three mechanisms for domain switching
at the start of each experiment. mechanisms at shorter time scales are proposed: (i) reduced mobil-
Our data offers insight into the kinetics of charged defects and their ity of 90◦ domain walls due to intrinsic domain wall kinetics and
interaction with domain walls. The increasing asymmetry with loading defect pinning, (ii) subsequent enhancement of domain nucleation
rate (Fig. 12) directly refers to critical time scales for the migration of due to higher bias fields and, (iii) nucleation of 180◦ over 90◦ do-
space charges and the reorientation of dipole defects. At low loading mains, potentially due to the reduction in mobile space charges for
rates (each cycle around 500 s), diffusive migration of charges to do- domain wall stability.
main walls allow for the stabilization of the evolving microstructure
quasistatically. Additionally, the applied fields are sufficiently strong to Declaration of Competing Interest
reorient defect dipoles. The apparent permittivity is therefore higher as
the polarizability increases. At higher loading rates (𝑒̇ > 0.1 MV/m s), The authors declare that they have no known competing financial
time scales for domain evolution are likely shorter than those for space interests or personal relationships that could have appeared to influence
charge migration (see Fig. 7), resulting in domain wall pinning due to the work reported in this paper.
dipole defects. Furthermore, the presence of 90◦ CDWs is less probable
due to the lack of point screening charges in the bulk (note that defect
Acknowledgments
dipoles are neutral on average). Hence, an additional nucleation of 180◦
domains may be preferred.
The authors acknowledge funding from the Swiss National Sci-
We conclude this discussion by summarizing important mechanisms
ence Foundation. V.K. and D.M.K. acknowledge funding through grant
influencing the domain wall kinetics, which require further investiga-
200021-178747, and M.T. through grant 200021-188414. Regular dis-
tion: (1) the time scales for space charge migration (for both point
cussions through the course of this study with L. Guin and R. Indergand
charges and defect dipoles) and diffusivity at domain walls; (2) the rel-
are also gratefully acknowledged.
ative activity of 90◦ vs. 180◦ domain walls as a function of internal bias
fields and loading history; (3) the effect of space charges on domain wall
kinetics (especially the kinetics of charged domain walls); (4) the kinet- Statement of data availability
ics of domain nucleation. Studies of these mechanisms promise refined
kinetic models for domain wall kinetics and consequently robust predic- The data that support the findings of this study are freely available in
tions for rate effects in the electromechanical response of ferroelectric the ETH Zürich Research Collection under the https://doi.org/10.3929/
ceramics. ethz-b-000536681.

13
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Appendix A. Effect of microstructure and kinetics on macroscopic calculated [106] (see the schematic in Fig. A.18 with the non-polar cu-
electro-mechanical response: a simple numerical model bic phase characterized by 𝒑𝑐 = 𝟎). With respect to the cubic reference,
strains in the tetragonal phases are given by the deformation gradients
A1. Constitutive description [106] (superscripts referring to the three domain types of interest here)

𝑡 𝑡
We present a simple qualitative model to describe the microstruc- ⎛𝜂1 0 0⎞ ⎛𝜂1 0 0⎞
tural mechanisms of ferroelectric switching in comparison with experi- 𝑭 (1) = 𝑭 (3) ⎜
= 0 𝜂1𝑡 0 ⎟, 𝑭 (2) = ⎜0 𝜂2𝑡 0⎟ (A.6)
𝑐 𝑐 ⎜ ⎟ 𝑐 ⎜ ⎟
mental data. Within a single crystal, we consider 𝑁 distinct polarization ⎝0 0 𝜂2𝑡 ⎠ ⎝0 0 𝜂1𝑡 ⎠
states (e.g., 𝑁 = 4 in 2D tetragonal perovskite). Rather than capturing
with 𝜂1𝑡 = 0.9958 and 𝜂2𝑡 = 1.0067 [105,106]. Deformation gradients
the detailed domain microstructure, we take a mesoscale perspective
with the tetragonal 0◦ domain as the reference, 𝑭 (𝑡𝑛) = 𝑭 (1) 𝑭 (𝑐𝑛) (see
−1
and describe microstructural effects only through the evolution of the 𝑐
∑ Fig. A.18), are used
volume fractions, denoted by 𝜗(𝑛) (so 0 ≤ 𝜗(𝑛) ≤ 1 and 𝑛 𝜗(𝑛) = 1) the (𝑛)
( to compute 𝑇
) the infinitesimal transformation strain
total volume fraction occupied by the 𝑛th polarization variant. The 𝑛th tensors  𝑡 = 12 𝑭 (𝑡𝑛) + 𝑭 (𝑡𝑛) − 𝑰 as
variant is characterized by a spontaneous polarization vector 𝒑(𝑛) , per-
mittivity tensor 𝝐 (𝑛) , third-order piezoelectric tensor  (𝑛) , and transfor- ⎧ ◦
(𝑛) ⎪𝟎, (2) if 𝑛 = 1 (0 domains)
mation strain tensor  𝑡 (calculated with respect to the polar tetragonal (𝑛) 𝑡
⎪ , if 𝑛 = 2 (90◦ domains)
phase). Under an applied electric field 𝒆, the constitutive relations 𝑡 =⎨ with
⎪ 𝟎, if 𝑛 = 3 (180◦ domains)
𝒅 = 𝝐𝒆 + 𝒑 ⇔ 𝑑𝑖 = 𝜖𝑖𝑗 𝑒𝑗 + 𝑝𝑖 (A.1) ⎪
⎩

𝜺 =  −1 𝒆 +  𝑡 ⇔ 𝜀𝑖𝑗 = −1 𝑡 ⎛0 0 0 ⎞
𝑘𝑖𝑗 𝑒𝑘 + 𝑖𝑗 (A.2)
𝑡 = ⎜0 0 ⎟
(2)
0.0109 (A.7)
⎜ ⎟
provide the electric displacement vector 𝒅 and the mechanical strain ⎝0 0 −0.0108⎠
tensor 𝜺 (using linearized kinematics). Here and in the following we use
classical index notation with Einstein’s summation convention. These reflect that 180◦ switching results in no straining, while 90◦
For the mixture of 𝑁 variants, we define the effective, homogenized switching cause tensile strains perpendicular to the direction of the ap-
electric displacement field plied field and contractile strains along that direction [19,23] (and our
plane strain assumption in 2D neglects strains in the 𝒆1 -direction).
𝑁
∑ 𝑁
∑ 𝑁
∑ To compare with experimental 𝑑 − 𝑒- and 𝜀 − 𝑒-data, we compute the
⟨𝒅 ⟩ = 𝜗(𝑛) 𝒅 (𝑛) = ⟨𝝐⟩𝒆 + ⟨𝒑⟩ with ⟨𝝐⟩ = 𝜗(𝑛) 𝝐 (𝑛) , ⟨𝒑⟩ = 𝜗 ( 𝑛 ) 𝒑( 𝑛 )
𝑛=1 𝑛=1 𝑛=1 measured electric displacement along the direction of applied electric
(A.3) field 𝒆 = 𝑒(𝑡)𝒆3 as
( 3 ) ( 3 )
as the volume average over all domains, writing ⟨⋅⟩ = ∫ (⋅)d𝑉 for the 1 ∑ (𝑛)
∑
𝑉 ⟨𝑑3 ⟩(𝑡) = ⟨𝜖33 ⟩𝑒(𝑡) + ⟨𝑝3 ⟩ = 𝜗(𝑛) (𝑡) 𝜖33 𝑒3 + 𝜗(𝑛) (𝑡) 𝑝(3𝑛) . (A.8)
volume average. This provides an upper bound on the exact effective 𝑛=1 𝑛=1
quantities and is chosen as a simple approximation.
Similarly, the measured strain components are (with 𝑖, 𝑗 = 1, 2)
Analogously, the homogenized piezoelectric and transformation
( 3 ) ( 3 )
strain tensors follow as, respectively, ∑ ∑
𝑡 (𝑛) −1 (𝑛) (𝑛) 𝑡(𝑛)
∑ ∑ (𝑛)
⟨𝜀𝑖𝑗 ⟩ = ⟨3𝑖𝑗 ⟩𝑒(𝑡) + ⟨𝑖𝑗 ⟩ =
−1
𝜗 (𝑡) (3𝑖𝑗 ) 𝑒3 + 𝜗 (𝑡) 𝑖𝑗 ,
𝜗(𝑛)  (𝑛) , ⟨ 𝑡 ⟩ = 𝜗(𝑛)  𝑡 .
−1
⟨ −1 ⟩ = (A.4) 𝑛=1 𝑛=1
𝑛 𝑛 (A.9)
Notice the usage of the Reuss average (as opposed to the Voigt no-
in which all is known except for the evolving volume fractions 𝜗(𝑛) (𝑡) as
tation) for the piezoelectric tensor. This is based on the constitutive re-
functions of time (for 𝑛 = 1, 2, 3).
lation for mechanical strain (Eq. A.1).
A3. Switching kinetics
A2. Reduction to the experimental setting
Our experiments highlight the rate dependence of ferroelectric
We align the coordinate axes with the sides of the samples used in ex-
switching, which is the macroscale result of intricate kinetic relations
periments, so the applied electric field is 𝒆 = −𝑒(𝑡)𝒆𝟑 , while we measure
between the velocity of a domain wall and the driving force [107,108] as
the electric displacement component 𝑑3 along with the in-plane strain
well as nucleation mechanisms. Based on theoretical studies [108], we
components 𝜀11 , 𝜀12 , 𝜀22 . We assume that the sample is initially poled
assume that the driving force is related to the average electric field, so
opposite to the applied electric field, 𝒑 = 𝑝𝑠 𝒆3 , and after switching at-
the kinetic relation takes the form 𝑣 = 𝜑(𝑒), which translates – in our ho-
tains 𝒑 = −𝑝𝑠 𝒆3 , 𝑝𝑠 being the spontaneous polarization (≃ 0.26 C∕m2 for
mogenized toy model – to 𝜗̇ (𝑛) = 𝜑(𝑒) with some function 𝜑(⋅). For a qual-
tetragonal BTO) [29,105]. To further simplify the scenario, we intro-
itative picture, we assume a simple linear kinetic relation, 𝜗̇ (𝑛) = 𝜇𝑒(𝑡)
duce a 2D setting, in which we consider three types of domains (see
with mobility 𝜇 > 0 [39,40,42]. As our objective here is to highlight
Fig. A.17), characterized by
the effect of switching mechanisms on measurable macroscopic quan-
⎧+𝑝 , if 𝑛 = 1 (referred to as 0◦ domains) tities such as the electric displacements and mechanical strains, we do
⎪ 𝑠
𝑝(3𝑛) = ⎨0, if 𝑛 = 2 (referred to as 90◦ domains) (A.5) not presume a-priori knowledge of the underlying kinetics (precise ki-
⎪−𝑝𝑠 , if 𝑛 = 3 (referred to as 180◦ domains). netic laws for nucleation and growth of domain walls present an open
⎩ challenge).
Of course, a realistic switching process in 3D would involve six variants, In the following, we present solutions for the volume fraction evolu-
yet – assuming isotropic conditions and no bias in directions perpen- tion as a function of time (or applied electric field) for each of the test
dicular to the applied field in a single-crystal – the above assumption cases shown in Fig. A.17, which present possible switching mechanisms
properly reflects the scenario for 𝑝3 (i.e., 𝑝3 = 0 for 𝑛 = 2 is taken as the in a tetragonal single-crystalline ferroelectric. Upon application of an
homogenized polarization resulting from randomly oriented domains electric field in the −𝒆3 -direction, reversal of the polarization occurs. In
perpendicular to the applied field). the two-dimensional description of Fig. A.17, we consider only domains
The associated transformation strains follow from crystallography. of type 0◦ , 90◦ and 180◦ , as introduced above. The four shown switch-
Deformation gradients with the cubic phase as the reference have been ing scenarios describe three mechanisms of reorientation chosen as limit

14
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. A.17. A simplified physical description of the 180◦ and 90◦ domain switching mechanisms.

Fig. A.18. Deformation gradients using different reference configurations for tetragonal domains.

15
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

cases – from homogeneous switching (Fig. A.17(a)) to pure 180◦ domain assumed to be a microstructure containing 90◦ domain walls (as
wall activity (Fig. A.17(b)) to pure 90◦ domain wall activity (Fig. A.17(c) is the case in any realistic polycrystalline texture). The domain
and (d)). Let us discuss each case in the following. volume fractions in the final state are characterized by a limiting
(a) Homogeneous switching polarization 𝑝𝑚 ≤ 𝑝𝑠 such that {
{
In an idealized single-crystal, the 0◦ domain switches homoge- 1, if 𝑒 < 𝑒𝑛 , 0, if 𝑒 < 𝑒𝑛 ,
𝜗(1) = , 𝜗(2) = 𝑝 ,
neously and instantaneously to a 180◦ domain at a critical electric 0, if 𝑒 ≥ 𝑒𝑛 1 − 𝑝𝑚 , if 𝑒 ≥ 𝑒𝑛
𝑠
field 𝑒𝑛 (see Fig. A.17(a)). Hence, the volume fraction evolution fol- {
0, if 𝑒 < 𝑒𝑛 ,
lows a step function: 𝜗(3) = 𝑝𝑚 (A.13)
𝑝
, if 𝑒 ≥ 𝑒𝑛 .
{ { { 𝑠
1, if 𝑒 < 𝑒𝑛 , 0, if 𝑒 < 𝑒𝑛 , 0, if 𝑒 < 𝑒𝑛 , 2. Kinetic evolution of 90◦ domain walls
𝜗(1) = 𝜗(2) = 𝜗(3) =
0, if 𝑒 ≥ 𝑒𝑛 , 0, if 𝑒 ≥ 𝑒𝑛 , 1, if 𝑒 ≥ 𝑒𝑛 , Polarization reversal in this case occurs by a sequence of two 90◦
(A.10) switching steps [80], which we simulate by two empirical nucle-
ation criteria. The initial nucleation 90◦ domains occurs at a nu-
where 𝑒𝑛 is (close to) the coercive field.
cleation field 𝑒𝑛 , followed by the nucleation of the 180◦ domains
(b) 180◦ domain walls
within the previously created 90◦ domain at a critical volume
While the above homogeneous switching is observed in a theoret-
fraction 𝜗(2) (𝑡) = 𝜗(2)
𝑐 < 1 (Fig. A.17(c)). Fig. A.17(d) presents the
ical single-crystal whose polarization is parallel to the direction of
limit case, in which nucleation of the 180◦ domains occurs only
the applied electric field, in a real-world scenario with misalignment
after saturation of the 90◦ domains (i.e., 𝜗(2) 𝑐 = 1). The second
between polarization and applied field (and, more generally, a poly-
nucleation criterion is used to mimic effects of delayed sequen-
crystalline texture), the resulting microstructural process is not nec-
tial nucleation on the macroscopic measured response in experi-
essarily homogeneous switching but the formation of domains. We
ments. At microscopic scales, this manifests as varying densities
first consider the case shown in Fig. A.17(b), which shows polariza-
of 90◦ domain walls. Again assuming linear kinetics as in case
tion switching only by 180◦ domain wall nucleation and growth.
b2 above, we obtain,
1. Instantaneous switching
⎧1, ⎧0 ,
To mimic the effect of misalignment (as discussed above), we re- ⎪ ⎪ 𝜇𝜔 2
𝜇𝜔 2
strict the maximum achievable polarization in the direction of ⎪1 − 2 (𝑡 − 𝑡𝑛1 ),
2
⎪ 2 (𝑡 − 𝑡2𝑛1 ),
⎪ 𝜇𝜔 ⎪ 𝜇𝜔
applied electric field to a value 𝑝𝑚 < 𝑝𝑠 below the spontaneous 𝜗(1) (𝑡) = ⎨1 − 2 (𝑡2 − 𝑡2𝑛1 ), 𝜗(2) = ⎨ 2 (𝑡2𝑛2 − 𝑡2𝑛1 ),
polarization 𝑝𝑠 . Still assuming instantaneous switching and ig- ⎪ ⎪ 𝜇𝜔 2
⎪0, ⎪1 − 2 ( 𝑡 − 𝑡 𝑛 2 ) ,
2
noring the role of kinetics, the domain volume fractions hence ⎪0, ⎪1 − 𝜇𝜔 (𝑡2 − 𝑡2 ),
evolve {according to ⎩ ⎩ 2 𝑚 𝑛2
{
1,( ) if 𝑒 < 𝑒𝑛 0, if 𝑒 < 𝑒𝑛 𝑒
𝜗(1) = 𝑝𝑚 𝜗(2) = ⎧0, if 𝑡 < 𝑡𝑛1 = 𝜔𝑛
1
1 − , if 𝑒 ≥ 𝑒 𝑛 , 0, if 𝑒 ≥ 𝑒𝑛 , ⎪
𝑝𝑠
{
2
⎪0, if 𝑡𝑛1 ≤ 𝑡 < 𝑡𝑛2
0,( ⎪ 𝜇𝜔 2
𝜗(3) = ) if 𝑒 < 𝑒𝑛 (A.11)
𝜗 (𝑡) = ⎨ 2 (𝑡 − 𝑡𝑛2 ), if 𝑡𝑛2 ≤ 𝑡 < 𝑡𝑒
(3) 2
(A.14)
𝑝𝑚
1
1 + , if 𝑒 ≥ 𝑒𝑛 . ⎪ 𝜇𝜔 2
2 𝑝𝑠 ⎪ 2 ( 𝑡 − 𝑡 2 ),
𝑛2
if 𝑡 𝑒 ≤ 𝑡 < 𝑡 𝑚
Of course, this reduces to (A.10) for 𝑝𝑚 = 𝑝𝑠 . ⎪ 𝜇𝜔 (𝑡2 − 𝑡2 ), if 𝑡 ≥ 𝑡 .
2. Kinetic evolution of 180◦ domain walls ⎩ 2 𝑚 𝑛2 𝑚
Time 𝑡𝑛1 , analogous to 𝑡𝑛 in (A.12), indicates the nucleation of the
As a more realistic scenario, we solve the linear kinetic equa-
90◦ domains at the nucleation electric field 𝑒𝑛 . Time 𝑡𝑛2 marks the
tion 𝜗̇ (3) (𝑡) = 𝜇𝑒(𝑡) by assuming the nucleation of 180◦ domains
nucleation of 180◦ domains, when 𝜗(2) = 𝜗(2) 𝑐 . Both 𝑒𝑛 and 𝜗𝑐 are
(2)
at an electric field 𝑒 = 𝑒𝑛 . The value of the mobility coefficient is
tuning parameters. The intermediate time 𝑡𝑒 > 𝑡𝑛2 is defined as
arbitrary, as our comparisons of different test cases will be nor-
the time scale (before saturation), at which all 0◦ domains have
malized by the mobility of the 180◦ switching case (Fig. A.17).
been removed (it is introduced to avoid negative volume frac-
The choice of the time-varying electric field 𝑒(𝑡) is given by our
tions during calculations). Note that, for 𝑡 ∈ [𝑡𝑛2 , 𝑡𝑒 ), 𝜗(2) = const.
experiments, which hold the rate of the electric field loading con-
due to the assumption that all 90◦ domain walls have equal mo-
stant (denoted by 𝜔). For simplicity, we let 𝑒(𝑡) monotonically
bility (see Fig. A.17), so the rate of increase of 180◦ domains
increase from zero in the negative 𝒆3 -direction (see Fig. A.17).
matches the rate of decrease of 0◦ domains. For 𝑡 > 𝑡𝑒 , the 180◦
The domain volume fraction evolution in this case follows as
⎧1, ⎧0 , domain volume fraction continues to grow with a correspond-
⎪ 𝜇𝜔 ⎪ ing decrease in the 90◦ domain, until ⟨𝑝3 ⟩ = 𝑝𝑚 . This time, at
𝜗(1) (𝑡) = ⎨1 − 2 (𝑡2 − 𝑡2𝑛 ), , 𝜗(2) = ⎨0, ,
which peak polarization is achieved, is denoted by 𝑡𝑚 , beyond
⎪ 𝜇𝜔 2 ⎪
⎩1 − 2 (𝑡𝑚 − 𝑡𝑛 )
2
⎩0 which only linear piezoelectricity of the final microstructure is
⎧0,
𝑒
if 𝑡 < 𝑡𝑛 = 𝜔𝑛 , observed. The specific choice of 𝑝𝑚 determines whether or not 0◦
⎪ 𝜇𝜔 2 domains are exhausted (𝑡𝑒 > 𝑡𝑚 ), when the maximum reversible
𝜗 (𝑡) = ⎨ 2 (𝑡 − 𝑡𝑛 ), if 𝑡𝑛 ≤ 𝑡 ≤ 𝑡𝑚 ,
(3) 2
(A.12)
polarization is too small.
⎪ 𝜇𝜔 2
⎩ 2 (𝑡𝑚 − 𝑡𝑛 ), if 𝑡 > 𝑡𝑚 ,
2

where 𝑡𝑛 = 𝑒𝑛 ∕𝜔 is the nucleation time at which 𝑒 = 𝑒𝑛 , and 𝑡𝑚

is the time at which polarization switching is complete (i.e., Appendix B. Electrical displacement measurements
⟨𝑝3 ⟩ = 𝑝𝑚 , maximum allowed switching polarization chosen as
an arbitrary parameter). Charges accumulated in ferroelectric samples are traditionally mea-
(c) 90◦ domain walls sured using the Sawyer-Tower circuit [109]. This circuit uses a standard
To differentiate 180◦ from 90◦ switching in the interpretation of our capacitor connected in series with the sample. The voltage output from
experimental data, we now consider the case in which polarization this reference capacitor is directly proportional to the charges accummu-
reversal is restricted to occur by 90◦ domain nucleation and growth lated in the series circuit. The time constant of this circuit is, however,
(see Fig. A.17(c)). directly related to the impedance of the measurement device. For stan-
1. Instantaneous switching dard digital oscilloscopes, the impedance is limited to 1 MΩ, resulting
Instantaneous switching is treated analogous to the above exam- in time constants on the order of 100 s. Long-time measurements are
ple of 180◦ switching, with the difference that the final state is hence non-trivial due to charge decay.

16
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

Fig. B.19. Charge amplifier circuit for electric displacement measurements.

One way of measuring charge in the sample over a wide range of circuit. Due to large voltages applied to the samples (on the order of sev-
time scales is by using a charge amplifier classically used in piezoelec- eral kV), minor imperfections in electrical contact or dielectric break-
tric testing [67]. In our experiment, charges accumulated on the sample down of the surrounding medium or sample failure can result in cur-
are measured using a home-built charge amplifier circuit (Fig. B.19). The rent surges through the circuit. The crossed Zener diodes limit the latter
circuit consists of an operational amplifier (Texas Instruments LF356N) through the operational amplifier. The protection resistor (𝑅𝑃 = 100 Ω)
with a feedback capacitor connected from the output to the inverting in- is installed for the same purpose. A reference capacitor (𝐶𝑅 = 100 μF)
put. The feedback capacitor acts as a current integrator, thus measuring admits the option of converting the circuit to the classical Sawyer-Tower
the charge on the specimen electrodes. The output from the charge am- circuit.
plifier is measured using a digital oscilloscope (Tektronix MDO3014).
References
Due to the operational amplifier, the output impedance is increased by
at least three orders of magnitude, thus improving long-time measure- [1] R.W. Whatmore, P.C. Osbond, N.M. Shorrocks, Ferroelectric materi-
ments. The charge 𝑄 is calculated as als for thermal IR detectors, Ferroelectrics 76 (1) (1987) 351–367,
doi:10.1080/00150198708016956.
𝑄 = 𝐶𝐹 𝑉out (B.1) [2] R. Watton, Ferroelectric materials and devices in infrared detection and imaging,
Ferroelectrics 91 (1) (1989) 87–108, doi:10.1080/00150198908015731.
with output voltage 𝑉out and feedback capacitance 𝐶𝐹 . Three different [3] C.R. Bowen, H.A. Kim, P.M. Weaver, S. Dunn, Piezoelectric and ferroelectric mate-
rials and structures for energy harvesting applications, Energy and Environmental
feedback capacitors (1, 10 and 33 μF) may be chosen based on the re- Science 7 (1) (2014) 25–44.
quired gain. The change in electric displacement [4] K.T. Butler, J.M. Frost, A. Walsh, Ferroelectric materials for solar energy conver-
sion: photoferroics revisited, Energy and Environmental Science 8 (3) (2015) 838–
𝑄 𝐶 𝑉
= 𝐹 out
848, doi:10.1039/C4EE03523B.
𝑑= (B.2)
𝐴 𝐴 [5] T.Y. Kim, S.K. Kim, S.W. Kim, Application of ferroelectric materials for im-
proving output power of energy harvesters, Nano Convergence 5 (30) (2015),
is calculated using the cross-sectional area 𝐴 of the sample surface. In doi:10.1186/s40580-018-0163-0.
practice, charge amplifiers are equipped with a feedback resistor in par- [6] D. Damjanovic, P. Muralt, N. Setter, Ferroelectric sensors, IEEE Sens. J. 1 (3) (2001)
191–206.
allel to the feedback capacitor to avoid saturation of charges during [7] E.F. Crawley, J. De Luis, Use of piezoelectric actuators as elements of in-
long-time experiments. This, however, results in a high-pass filter, so telligent structures, AIAA J. 25 (10) (1986) 1373–1385, doi:10.2514/3.9792.
that at low frequencies (governed by the feedback resistance) the am- http://arc.aiaa.org.
[8] T.A. Asare, B.D. Poquette, J.P. Schultz, S.L. Kampe, Investigating the vibration
plifier has no response [67]. To avoid charge accumulation, a switch is damping behavior of barium titanate (BaTiO3 ) ceramics for use as a high damping
connected in parallel to the feedback capacitor to reset the charge af- reinforcement in metal matrix composites, J. Mater. Sci. 47 (6) (2012) 2573–2582,
ter every experiment. Note that 𝑑 is a relative, not an absolute measure doi:10.1007/s10853-011-6080-9.
[9] K.P. Duffy, B.B. Choi, A.J. Provenza, J.B. Min, N. Kray, Active piezoelectric vibra-
of electric displacement in the sample due to the capacitor reset after
tion control of subscale composite fan blades, J. Eng. Gas Turbines Power 135 (1)
every experiment. This is an important detail for the interpretation of (2013), doi:10.1115/1.4007720.
experimental data. [10] K. Patterson, Lightweight Deformable Mirrors for Future Space Telescopes, Califor-
nia Institute of Technology, 2013 Ph.D. thesis.
The circuit diagram shown in Fig. B.19 has some additional com-
[11] C.S. Wojnar, J.-B. le Graverend, D.M. Kochmann, Broadband control of the vis-
ponents in comparison to conventional charge amplifiers. The buffer coelasticity of ferroelectrics via domain switching, Appl. Phys. Lett. 105 (16) (2014)
capacitor (𝐶𝐵 = 2nF) is typically used for high-speed switching mea- 162912, doi:10.1063/1.4899055.
surements on the order of nanoseconds [110]. This capability has not [12] V.I. Pilipovich, V.I. Polyakov, A.I. Konoiko, Electro-optic light modulators, Instrum.
Exp. Tech. New York 30 (1 pt 2) (1987) 199–201, doi:10.1364/ao.5.001612.
been used in the current study. In addition, the buffer capacitor acts [13] D.A. Buck, Ferroelectrics for Digital Information Storage and Switching, Technical
as a filter for noisy measurements. The crossed Zener diodes protect the Report, 1952. http://hdl.handle.net/1721.3/40244.

17
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

[14] Y. Tarui, T. Hirai, K. Teramoto, H. Koike, K. Nagashima, Application of the fer- [46] D. Viehland, Y.-H. Chen, Random-field model for ferroelectric domain dynamics
roelectric materials to ULSI memories, Appl. Surf. Sci. 113–114 (1997) 656–663, and polarization reversal, J. Appl. Phys. 88 (2000) 6696, doi:10.1063/1.1325001.
doi:10.1016/S0169-4332(96)00963-4. [47] D. Zhou, M. Kamlah, D. Munz, Rate dependence of soft PZT ceramics under electric
[15] O. Auciello, R. Ramesh, The physics of ferroelectric memories, Phys. Today (1998) field loading, in: C.S. Lynch (Ed.), Smart Struct. Mater. 2001 Act. Mater. Behav.
22–27. https://www.researchgate.net/publication/273629413. Mech., vol. 4333, SPIE, 2001, p. 64, doi:10.1117/12.432740.
[16] M.E. Lines, A.M. Glass, Principles and Applications of Ferro- [48] A. Achuthan, C.T. Sun, A study of mechanisms of domain switching in a ferro-
electrics and Related Materials, Oxford University Press, 2010, electric material via loading rate effect, Acta Mater. 57 (13) (2009) 3868–3875,
doi:10.1093/acprof:oso/9780198507789.001.0001. doi:10.1016/j.actamat.2009.04.043.
[17] J. Curie, P. Curie, Développement par compression de l’électricité polaire dans les [49] M.H. Lente, A. Picinin, J.P. Rino, J.A. Eiras, 90◦ domain wall relaxation and fre-
cristaux hémièdres à faces inclinées, Bull. Soci. Minéral. Fr. 3 (4) (1880) 90–93, quency dependence of the coercive field in the ferroelectric switching process, J.
doi:10.3406/bulmi.1880.1564. Appl. Phys. 95 (2004) 2646–2653, doi:10.1063/1.1645980.
[18] J. Valasek, Piezo-electric and allied phenomena in Rochelle salt, Phys. Rev. 17 (4) [50] R. Yimnirun, Y. Laosiritaworn, S. Wongsaenmai, S. Ananta, Scaling behavior of
(1921) 475–481, doi:10.1103/PhysRev.17.475. dynamic hysteresis in soft lead zirconate titanate bulk ceramics, Appl. Phys. Lett.
[19] B. Jaffe, W.R. Cook Jr., H. Jaffe, Piezoelectric Ceramics, Elsevier, 1971, 89 (16) (2006) 162901, doi:10.1063/1.2363143.
doi:10.1016/B978-0-12-379550-2.X5001-7. [51] J. Yin, W. Cao, Coercive field of 0.955Pb(Zn1∕3 Nb2∕3 )O3 -0.045PbTiO3 single crys-
[20] A.K. Tagantsev, L.E. Cross, J. Fousek, Domains in Ferroic Crystals and Thin Films, tal and its frequency dependence, Appl. Phys. Lett. 80 (6) (2002) 1043–1045,
Springer New York, 2010, doi:10.1007/978-1-4419-1417-0. doi:10.1063/1.1448385.
[21] D. Meier, J. Seidel, M. Gregg, R. Ramesh, Domain Walls, Oxford University Press, [52] Q.M. Zhang, W.Y. Pan, S.J. Jang, L.E. Cross, Domain wall excitations and their
2020, doi:10.1093/oso/9780198862499.001.0001. contributions to the weak-signal response of doped lead zirconate titanate ceramics,
[22] B. Jaffe, W.R. Cook, H. Jaffe, Barium titanate, in: Piezoelectric Ceram., Elsevier, J. Appl. Phys. 64 (11) (1988) 6445–6451, doi:10.1063/1.342059.
1971, pp. 53–114. [53] A. Pramanick, A.D. Prewitt, J.S. Forrester, J.L. Jones, Domains, domain walls and
[23] D. Damjanovic, Hysteresis in Piezoelectric and Ferroelectric Materials, Technical defects in perovskite ferroelectric oxides: a review of present understanding and
Report, 2005. https://infoscience.epfl.ch/record/88325. recent contributions, Crit. Rev. Solid State Mater. Sci. 37 (4) (2012) 243–275,
[24] C.S. Lynch, The effect of uniaxial stress on the electro-mechanical re- doi:10.1080/10408436.2012.686891.
sponse of 8/65/35 PLZT, Acta Mater. 44 (10) (1996) 4137–4148, [54] J.E. Daniels, C. Cozzan, S. Ukritnukun, G. Tutuncu, J. Andrieux, J. Glaum, C. Dosch,
doi:10.1016/S1359-6454(96)00062-6. W. Jo, J.L. Jones, Two-step polarization reversal in biased ferroelectrics, J. Appl.
[25] W.J. Merz, Domain formation and domain wall motions in ferroelectric BaTiO3 Phys. 115 (22) (2014) 224104, doi:10.1063/1.4881835.
single crystals, Phys. Rev. 95 (3) (1954) 690–698, doi:10.1103/PhysRev.95.690. [55] T. Iamsasri, G. Esteves, H. Choe, M. Vogt, S. Prasertpalichat, D.P. Cann, S. Gorf-
[26] E.A. Little, Dynamic behavior of domain walls in barium titanate, Phys. Rev. 98 man, J.L. Jones, Time and frequency-dependence of the electric field-induced
(4) (1955) 978–984, doi:10.1103/PhysRev.98.978. phase transition in BaTiO3 -BiZn1∕2 Ti1∕2 O3 , J. Appl. Phys. 122 (6) (2017) 064104,
[27] R.C. Miller, A. Savage, Velocity of sidewise 180 degree domain-wall motion in in doi:10.1063/1.4998163.
BaTiO3 as a function of the applied electric field, Phys. Rev. 112 (3) (1958) 755– [56] J.E. Daniels, A. Pramanick, J.L. Jones, Time-resolved characterization of ferro-
762, doi:10.1103/PhysRev.112.755. electrics using high-energy X-ray diffraction, IEEE Trans. Ultrason., Ferroelectr.,
[28] V. Gopalan, T.E. Mitchell, In situ video observation of 180 degree domain switching Freq, Control 56 (8) (2009) 1539–1545, doi:10.1109/TUFFC.2009.1218.
in LiTaO3 by electro-optic imaging microscopy, J. Appl. Phys. 85 (4) (1999) 2304– [57] J.L. Jones, J.C. Nino, A. Pramanick, J.E. Daniels, Time-Resolved, Electric-Field-
2311, doi:10.1063/1.369542. -Induced Domain Switching and Strain in Ferroelectric Ceramics and Crystals,
[29] E. Burcsu, Investigation of Large Strain Actuation in Barium Ti- Springer Berlin Heidelberg, Berlin, Heidelberg, 2010, pp. 149–175.
tanate, California Institute of Technology, 2001 Ph.D. thesis. [58] S. Gorfman, H. Simons, T. Iamsasri, S. Prasertpalichat, D.P. Cann, H. Choe,
https://resolver.caltech.edu/CaltechETD:etd-10232001-192042. U. Pietsch, Y. Watier, J.L. Jones, Simultaneous resonant X-ray diffraction measure-
[30] G. Arlt, P. Sasko, Domain configuration and equilibrium size of domains in BaTiO3 ment of polarization inversion and lattice strain in polycrystalline ferroelectrics,
ceramics, J. Appl. Phys. 51 (1980) 4956, doi:10.1063/1.328372. Sci. Rep. 6 (2016) 20829, doi:10.1038/srep20829.
[31] J. Schultheiß, L. Liu, H. Kungl, M. Weber, L. Kodumudi Venkataraman, S. Chec- [59] L.W. Martin, A.M. Rappe, Thin-film ferroelectric materials and their applications,
chia, D. Damjanovic, J.E. Daniels, J. Koruza, Revealing the sequence of switching Nat. Rev. Mater. 2 (2) (2016) 1–14, doi:10.1038/natrevmats.2016.87.
mechanisms in polycrystalline ferroelectric/ferroelastic materials, Acta Mater. 157 [60] N. Strkalj, M. Bernet, M.F. Sarott, J. Schaab, T. Weber, M. Fiebig, M. Trassin, Sta-
(2018) 355–363, doi:10.1016/j.actamat.2018.07.018. bilization and manipulation of in-plane polarization in a ferroelectric|dielectric su-
[32] C. Brennan, Model of ferroelectric fatigue due to defect/domain interactions, Fer- perlattice, J. Appl. Phys. 129 (17) (2021) 174104, doi:10.1063/5.0035867.
roelectrics 150 (1) (1993) 199–208, doi:10.1080/00150199308008705. [61] M.F. Sarott, E. Gradauskaite, J. Nordlander, N. Strkalj, M. Trassin, In situ moni-
[33] A.C.F. Cocks, R.M. Mcmeeking, A.C.F. Cocksa, R.M. Mcmeeking’, A phenomeno- toring of epitaxial ferroelectric thin-film growth, J. Phys. Condens. Matter 33 (29)
logical constitutive law for the behaviour of ferroelectric ceramics, Ferroelectrics (2021) 293001, doi:10.1088/1361-648X/ABF979.
228 (1999) 219–228, doi:10.1080/00150199908226136. [62] M.J. Highland, T.T. Fister, D.D. Fong, P.H. Fuoss, C. Thompson, J.A. East-
[34] M. Kamlah, C. Tsakmakis, Phenomenological modeling of the non-linear electro- man, S.K. Streiffer, G.B. Stephenson, Equilibrium polarization of ultrathin
mechanical coupling in ferroelectrics, Int. J. Solids Struct. 36 (5) (1999) 669–695, PbTiO3 with surface compensation controlled by oxygen partial pressure, Phys.
doi:10.1016/S0020-7683(98)00040-7. Rev. Lett. 107 (18) (2011), doi:10.1103/PHYSREVLETT.107.187602/FIGURES/
[35] C.M. Landis, Fully coupled, multi-axial, symmetric constitutive laws for poly- 1/THUMBNAIL.
crystalline ferroelectric ceramics, J. Mech. Phys. Solids 50 (1) (2002) 127–152, [63] G. De Luca, N. Strkalj, S. Manz, C. Bouillet, M. Fiebig, M. Trassin, Nanoscale de-
doi:10.1016/S0022-5096(01)00021-7. sign of polarization in ultrathin ferroelectric heterostructures, Nat. Commun. 8 (1)
[36] Y. Yu, N. Naganathan, R. Dukkipati, Preisach modeling of hysteresis for (2017) 1–7, doi:10.1038/s41467-017-01620-2.
piezoceramic actuator system, Mech. Mach. Theory 37 (1) (2002) 49–59, [64] J.A. Mundy, J. Schaab, Y. Kumagai, A. Cano, M. Stengel, I.P. Krug, D.M. Gottlob,
doi:10.1016/S0094-114X(01)00065-9. H. Doğanay, M.E. Holtz, R. Held, Z. Yan, E. Bourret, C.M. Schneider, D.G. Schlom,
[37] A.T. Bartic, D.J. Wouters, H.E. Maes, J.T. Rickes, R.M. Waser, Preisach model for D.A. Muller, R. Ramesh, N.A. Spaldin, D. Meier, Functional electronic inver-
the simulation of ferroelectric capacitors, J. Appl. Phys. 89 (6) (2001) 3420–3425, sion layers at ferroelectric domain walls, Nat. Mater. 16 (6) (2017) 622–627,
doi:10.1063/1.1335639. doi:10.1038/nmat4878.
[38] S.C. Hwang, G. Arlt, Switching in ferroelectric polycrystals, J. Appl. Phys. 87 (2) [65] J. Schaab, S.H. Skjærvø, S. Krohns, X. Dai, M.E. Holtz, A. Cano, M. Lilienblum,
(2000) 869–875, doi:10.1063/1.371968. Z. Yan, E. Bourret, D.A. Muller, M. Fiebig, S.M. Selbach, D. Meier, Electrical half-
[39] W. Zhang, K. Bhattacharya, A computational model of ferroelectric domains. Part wave rectification at ferroelectric domain walls, Nat. Nanotechnol. 13 (11) (2018)
I: model formulation and domain switching, Acta Mater. 53 (1) (2005) 185–198, 1028–1034, doi:10.1038/s41565-018-0253-5.
doi:10.1016/j.actamat.2004.09.016. [66] J.B. le Graverend, C.S. Wojnar, D.M. Kochmann, Broadband electromechanical
[40] Y. Su, C.M. Landis, Continuum thermodynamics of ferroelectric domain evo- spectroscopy: characterizing the dynamic mechanical response of viscoelastic ma-
lution: theory, finite element implementation, and application to domain terials under temperature and electric field control in a vacuum environment, J.
wall pinning, J. Mech. Phys. Solids 55 (2) (2007) 280–305, Mater. Sci. 50 (10) (2015) 3656–3685, doi:10.1007/s10853-015-8928-x.
doi:10.1016/j.jmps.2006.07.006. [67] M.G. Cain, Characterisation of Ferroelectric Bulk Materials and Thin Films,
[41] W.L. Tan, D.M. Kochmann, An effective constitutive model for polycrystalline ferro- Springer, 2014, doi:10.1007/978-1-4020-9311-1.
electric ceramics: theoretical framework and numerical examples, Comput. Mater. [68] L. Dong, D.S. Stone, R.S. Lakes, Softening of bulk modulus and negative Poisson
Sci. 136 (2017) 223–237, doi:10.1016/j.commatsci.2017.04.032. ratio in barium titanate ceramic near the Curie point, Philos. Mag. Lett. 90 (1)
[42] R. Indergand, A. Vidyasagar, N. Nadkarni, D.M. Kochmann, A phase-field (2010) 23–33, doi:10.1080/09500830903344907.
approach to studying the temperature-dependent ferroelectric response [69] S.V. Kalinin, A.N. Morozovska, L.Q. Chen, B.J. Rodriguez, Local polarization
of bulk polycrystalline PZT, J. Mech. Phys. Solids 144 (2020) 104098, dynamics in ferroelectric materials, Rep. Prog. Phys. 73 (5) (2010) 056502,
doi:10.1016/j.jmps.2020.104098. doi:10.1088/0034-4885/73/5/056502.
[43] S. Liu, I. Grinberg, A.M. Rappe, Intrinsic ferroelectric switching from first princi- [70] A. Gruverman, M. Alexe, D. Meier, Piezoresponse force microscopy and nanoferroic
ples, Nature 534 (7607) (2016) 360–363, doi:10.1038/nature18286. phenomena, Nat. Commun. 10 (1) (2019) 1–9, doi:10.1038/s41467-019-09650-8.
[44] V. Boddu, F. Endres, P. Steinmann, Molecular dynamics study of ferroelectric do- [71] W.L. Tan, K.T. Faber, D.M. Kochmann, In-situ observation of evolving mi-
main nucleation and domain switching dynamics, Sci. Rep. 7 (1) (2017) 1–10, crostructural damage and associated effective electro-mechanical properties
doi:10.1038/s41598-017-01002-0. of PZT during bipolar electrical fatigue, Acta Mater. 164 (2019) 704–713,
[45] B. Völker, P. Marton, C. Elsässer, M. Kamlah, Multiscale modeling for ferroelectric doi:10.1016/j.actamat.2018.10.065.
materials: a transition from the atomic level to phase-field modeling, Contin. Mech. [72] J.F. Scott, Ferroelectrics go bananas, J. Phys. 20 (2) (2007) 021001,
Thermodyn. 23 (5) (2011) 435–451, doi:10.1007/S00161-011-0188-7. doi:10.1088/0953-8984/20/02/021001.

18
V. Kannan, M. Trassin and D.M. Kochmann Materialia 25 (2022) 101553

[73] M.I. Morozov, D. Damjanovic, Charge migration in ceramics and its rela- [91] M. Takahashi, Space charge effect in lead zirconate titanate ceramics caused
tion to ageing, hardening, and softening, J. Appl. Phys 107 (2010) 34106, by the addition of impurities, Jpn. J. Appl. Phys. 9 (10) (1970) 1236–1246,
doi:10.1063/1.3284954. doi:10.1143/JJAP.9.1236.
[74] G. Liu, S. Zhang, W. Jiang, W. Cao, Losses in ferroelectric materials, Mater. Sci. [92] H. Thomann, Stabilization effects in piezoelectric lead titanate zirconate ceramics,
Eng. R Rep. 89 (2015) 1–48, doi:10.1016/j.mser.2015.01.002. Ferroelectrics 4 (1) (1972) 141–146, doi:10.1080/00150197208235755.
[75] H.G. Baerwald, D.A. Berlincourt, Electromechanical response and dielectric loss of [93] K. Carl, K.H. Hardtl, Electrical after-effects in Pb(Ti, Zr)O3 ceramics, Ferroelectrics
prepolarized barium titanate under maintained electric bias. Part I, J. Acoust. Soc. 17 (1) (1977) 413–486, doi:10.1080/00150197808236770.
Am. 25 (4) (1953) 703–710, doi:10.1121/1.1907164. [94] P.V. Lambeck, G.H. Jonker, The nature of domain stabilization in fer-
[76] H. Neumann, G. Arlt, Deformation of hysteresis curves by an internal bias roelectric perovskites, J. Phys. Chem. Solids 47 (5) (1986) 453–461,
in ferroelectric ceramics, in: Ultrason. Symp. Proc., IEEE, 1987, pp. 671–674, doi:10.1016/0022-3697(86)90042-9.
doi:10.1109/ultsym.1987.199043. [95] H. Neumann, G. Arlt, The deformation of hysteresis curves by an internal
[77] S. Zhang, J.B. Lim, H.J. Lee, T.R. Shrout, Characterization of hard piezoelectric bias in ferroelectric ceramics, in: Ultrason. Symp., IEEE, 1987, pp. 671–674.
lead-free ceramics, in: IEEE Trans. Ultrason. Ferroelectr. Freq. Control, vol. 56, https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=1535983&tag=1.
NIH Public Access, 2009, pp. 1523–1527, doi:10.1109/TUFFC.2009.1215. [96] P. Erhart, P. Träskelin, K. Albe, Formation and switching of defect dipoles in
[78] E.M. Anton, R.E. García, T.S. Key, J.E. Blendell, K.J. Bowman, Domain switching acceptor-doped lead titanate: a kinetic model based on first-principles calculations,
mechanisms in polycrystalline ferroelectrics with asymmetric hysteretic behavior, Phys. Rev. B 88 (2013) 024107, doi:10.1103/PhysRevB.88.024107.
J. Appl. Phys. 105 (2) (2009) 24107, doi:10.1063/1.3068333. [97] J.F. Scott, C.A. Araujo, B.M. Melnick, L.D. Mcmillan, R. Zuleeg, Quantitative mea-
[79] G. Arlt, H. Neumann, Internal bias in ferroelectric ceramics: origin and time depen- surement of space-charge effects in lead zirconate-titanate memories, J. Appl. Phys.
dence, Ferroelectrics 87 (1) (1988) 109–120, doi:10.1080/00150198808201374. 70 (1991) 382, doi:10.1063/1.350286.
[80] B. Jiang, Y. Bai, W. Chu, Y. Su, L. Qiao, Direct observation of two 90◦ steps of 180◦ [98] Y. Zuo, Y.A. Genenko, A. Klein, P. Stein, B. Xu, Domain wall stability in
domain switching in BaTiO3 single crystal under an antiparallel electric field, Appl. ferroelectrics with space charges, J. Appl. Phys. 115 (8) (2014) 084110,
Phys. Lett. 93 (15) (2008) 152905, doi:10.1063/1.3000634. doi:10.1063/1.4866359.
[81] T. Tsurumu, Y. Kumano, N. Ohashi, T. Takenaka, O. Fukunaga, 90◦ domain [99] M.Y. Gureev, A.K. Tagantsev, N. Setter, Head-to-head and tail-to-tail 180◦ domain
reorientation and electric-field-induced strain of tetragonal lead zirconate ti- walls in an isolated ferroelectric, Phys. Rev. B 83 (18) (2011), doi:10.1103/PHYS-
tanate ceramics, Jpn. J. Appl. Phys. 36 (95) (1997) 5970, doi:10.1143/JJAP.36. REVB.83.184104/FIGURES/1/THUMBNAIL.
5970. [100] W.L. Warren, D. Dimos, B.A. Tuttle, R.D. Nasby, G.E. Pike, Electronic domain pin-
[82] Y.W. Li, F.X. Li, The effect of domain patterns on 180◦ domain switching in BaTiO3 ning in Pb(Zr,Ti)O3 thin films and its role in fatigue, Appl. Phys. Lett. 65 (8) (1998)
crystals during antiparallel electric field loading, Appl. Phys. Lett. 104 (4) (2014) 1018, doi:10.1063/1.112211.
042908, doi:10.1063/1.4863672. [101] I.B. Misirlioglu, M.B. Okatan, S.P. Alpay, Effect of asymmetrical interface charges
[83] B. Brezina, J. Fousek, A. Glanc, Barkhausen pulses in BaTiO3 connected on the hysteresis and domain configurations of ferroelectric thin films, Integrated
with 90◦ switching processes, Cechoslov. Fiz. Z 11 (8) (1961) 595–601, Ferroelectrics 126 (1) (2011) 142–154, doi:10.1080/10584587.2011.575017.
doi:10.1007/BF01689156. [102] Y.H. Han, J.B. Appleby, D.M. Smyth, Calcium as an acceptor impurity in BaTiO3 , J.
[84] V.Y. Shur, E.L. Rumyantsev, D.V. Pelegov, V.L. Kozhevnikov, E.V. Nikolaeva, Am. Ceram. Soc. 70 (2) (1987) 96–100, doi:10.1111/j.1151-2916.1987.tb04936.x.
E.L. Shishkin, A.P. Chernykh, R.K. Ivanov, Barkhausen jumps during do- [103] U. Robels, J.H. Calderwood, G. Arlt, Shift and deformation of the hysteresis curve
main wall motion in ferroelectrics, Ferroelectrics 267 (1) (2010) 347–353, of ferroelectrics by defects: an electrostatic model, J. Appl. Phys. 77 (8) (1995)
doi:10.1080/00150190211031. 4002–4008, doi:10.1063/1.359511.
[85] R.J. Harrison, E.K.H. Salje, The noise of the needle: avalanches of a single pro- [104] X.L. Wang, B. Li, X.L. Zhong, Y. Zhang, J.B. Wang, Y.C. Zhou, Effects of space charge
gressing needle domain in LaAlO3 , Appl. Phys. Lett. 97 (2) (2010) 021907, distribution on ferroelectric hysteresis loops considering the inhomogeneous built-
doi:10.1063/1.3460170. in electric field: a phase field simulation, J. Appl. Phys. 112 (11) (2012) 114103,
[86] E.K.H. Salje, X. Wang, X. Ding, J.F. Scott, Ultrafast switching in avalanche-driven doi:10.1063/1.4767702.
ferroelectrics by supersonic kink movements, Adv. Funct. Mater. 27 (21) (2017) [105] A.M.E. Mitsui, T. Hellwege, K.H. Hellwege, Ferroelectrics and Related Substances,
1700367, doi:10.1002/adfm.201700367. Landolt–Boernstein Numerical Data and Functional Relationships in Science and
[87] C. Borderon, R. Renoud, M. Ragheb, H.W. Gundel, Description of the low field non- Technology, Group III, Springer, New York, 1981.
linear dielectric properties of ferroelectric and multiferroic materials, Appl. Phys. [106] Y.C. Shu, K. Bhattacharya, Domain patterns and macroscopic behaviour
Lett. 98 (11) (2011) 112903, doi:10.1063/1.3567777. of ferroelectric materials, Philos. Mag. B 81 (12) (2001) 2021–2054,
[88] K. Nadaud, C. Borderon, R. Renoud, H.W. Gundel, Effect of manganese doping doi:10.1080/1364281011006957.
of BaSrTiO3 on diffusion and domain wall pinning, J. Appl. Phys. 117 (8) (2015) [107] R. Abeyaratne, J.K. Knowles, On the driving traction acting on a surface of strain
084104, doi:10.1063/1.4913694. discontinuity in a continuum, J. Mech. Phys. Solids 38 (3) (1990) 345–360,
[89] S.V. Kalinin, B.J. Rodriguez, S. Jesse, J. Shin, A.P. Baddorf, P. Gupta, H. Jain, doi:10.1016/0022-5096(90)90003-M.
D.B. Williams, A. Gruverman, Vector piezoresponse force microscopy, Microsc. Mi- [108] Q. Jiang, On the driving traction acting on a surface of discontinuity within a con-
croanal. 12 (3) (2006) 206–220, doi:10.1017/S1431927606060156. tinuum in the presence of electromagnetic fields, J. Elast. 34 (1994) 1–21.
[90] S. Trolier-McKinstry, N. Bassiri Gharb, D. Damjanovic, Piezoelectric nonlinear- [109] C.B. Sawyer, C.H. Tower, Rochelle salt as a dielectric, Phys. Rev. 35 (3) (1930)
ity due to motion of 180◦ domain walls in ferroelectric materials at subcoer- 269–273, doi:10.1103/PhysRev.35.269.
cive fields: a dynamic poling model, Appl. Phys. Lett. 88 (20) (2006) 202901, [110] J.E. Schultheiss, Polarization Reversal Dynamics in Polycrystalline Ferroelectric /
doi:10.1063/1.2203750. Ferroelastic Ceramic Materials, Technical University Darmstadt, 2018 Ph.D. thesis.

19