IEEE Trans UFFC 589 1867 2011
IEEE Trans UFFC 589 1867 2011
IEEE Trans UFFC 589 1867 2011
net/publication/51661064
CITATIONS READS
20 535
5 authors, including:
Kok-Boon Chong
30 PUBLICATIONS 524 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Development of epitaxial growth and fabrication technologies for AlGaN-based UV-LEDs View project
All content following this page was uploaded by Kok-Boon Chong on 06 March 2015.
Abstract—Time- and temperature-dependent effects are models is the failure to predict a relationship between the
critical for the operation of non-volatile memories based on fer- switching time to the applied electric field and tempera-
roelectrics. In this paper, we assume a domain nucleation pro-
cess of the polarization reversal and we discuss the polariza-
ture. Most of the published relations are based on empiri-
tion dynamics in the framework of a non-equilibrium statistical cal expressions derived from the study of domain wall mo-
model. This approach yields analytical expressions which can tion. Moreover, (1) contains no explicit dependence on the
be used to explain a wide range of time- and temperature-de- E field or temperature. Nucleation processes, as described
pendent effects in ferroelectrics. Domain wall velocity derived by KAI models, assume, in general, that they are gener-
in this work is consistent with a domain wall creep behavior in
ferroelectrics. In the limiting case of para-electric equilibrium,
ated with a certain probability followed by the fast rever-
the model yields the well-known Curie law. We also present sal of polarization, without giving a relationship between
experimental P-E loops data obtained for soft ferroelectrics the probability of nuclei formation and the driving forces
at various temperatures. The experimental coercive fields at (electric field, temperature, etc.). We represent the ferro-
various temperatures are well predicted by the coercive field electric sample as an assembly of N regions, which we call
formula derived in our theory.
elementary polar regions. Applying a statistical treatment
to the microscopic mechanisms responsible for nucleation
I. Introduction inside an elementary polar region, is our approach to de-
W hen an electric field is applied to a ferroelectric ma- rive the kinetic equations of the polarization switching in
terial, the reversal process occurs via a nucleation this paper. This theoretical approach was first proposed in
of domains, movement of domain walls, and subsequently [10] and applied to ferroelectric systems that display depo-
expansion and growth at the expense of the existing do- larization field effects. In this paper, we used a simplified
mains [1]. One of the fundamental models of the polariza- version of the model to derive the kinetic equations of
tion switching is the Kalmogorov-Avrami-Ishibashi (KAI) polarization reversal and to explain the experimental P-E
domain nucleation-switching model, which treats polar- loop data acquired at various temperatures. The model
ization reversal as a nucleation process [2] and it predicts makes the following assumptions: 1) the polarization
the following temporal dependence of polarization/electric reversal occurs via a domain nucleation process; 2) the
displacement during the switching process: nucleation process occurs independently inside regions of
material called elementary polar regions; 3) the nucleation
n
∆D(t) = 2p s ⋅ (1 − e −(t/t sw) ), (1) process is considered to be activated when the region of
reversed polarization reaches a critical volume V*; and 4)
where t is the experimental time; tsw is the switching time; the switching rate is controlled by the switching rate of
2ps is the amount of reversed polarization, where ps is the critical volume V*, rather than the rate of expansion
the spontaneous polarization at zero applied field E; and of the domain. If the volume of this region of reversed
n is the Avrami exponent. The KAI model of polariza- polarization is below a critical level, V*, the nucleus is
tion switching has successfully given a good description unstable and decays. Above the critical volume, it is ener-
of polarization kinetics of ferroelectric single crystals [3] getically favorable for the nucleus to grow. It is important
and some epitaxial thin films [4], but it is not fully ap- to mention that these assumptions are fully valid for re-
plicable to the reversal over larger time periods or the versal produced by applied electric fields much larger than
reversal of polycrystalline thin films [5]. Several studies the coercive field. When applied E fields are comparable
have attempted to improve the KAI model to increase to or smaller than the coercive field, reversal via domain
its applicability. The introduction of a distribution of re- wall movement becomes dominant and the total switching
laxation times [6], a nucleation-limited switching model time includes also the time taken for the domain wall to
[5], a statistical time-dependent depolarization field [7], expand to full reversal state. However, the proposed nucle-
size effects [8], and thermal effects [9] are a few of the ation model remains valid, because considerations of the
most notable published innovations. Despite all of these domain wall dynamics are complementary to it. Although
advances, one of the main limitations of the nucleation outside the scope of this model, this important issue is
also briefly discussed in Section IV.
Manuscript received October 14, 2010; accepted April 12, 2011.
M. Vopsaroiu, P. M. Weaver, and M. G. Cain are with the National
Physical Laboratory Teddington, UK (e-mail: mv1@npl.co.uk). II. Model Description
M. Vopsaroiu is currently known as Melvin M. Vopson.
M. J. Reece and K. B. Chong are with the Centre for Materials Re-
search, Queen Mary University of London, London, E1 4NS, UK. We assume that the switching of elementary polar re-
Digital Object Identifier 10.1109/TUFFC.2011.2025 gions is provided by the nucleation of a domain per el-
t sw = υ 0 −1 ⋅ e (W B −p sE )/k BT . (7)
III. Switching Kinetics and Coercive Field barrier of 0.9 eV (1.44 × 10−19 J) for zero applied field
at room temperature and a spontaneous polarization of
For N elementary polar regions, we define N1(t) = 0.74 C/m2. For all calculations, the energy barrier has
P1(t) ⋅ N as the number of sites in state 1 at time t and been kept constant with the shifts in the energy resulting
N2(t) = P2(t) ⋅ N as the number of sites in state 2 at time only from the applied electric field. This is a rather rough
t, with N = N1 + N2. We now assume that the nucleation approximation because the energy barrier has, in fact, a
process occurs over a very short period of time and simul- strong dependence on the temperature and other intrinsic
taneously in all the elementary polar regions and that (4) or extrinsic factors.
and (5) describe the state of each polar nucleation region Modeling data obtained using (9) is presented in Fig.
at time t. We also assume that the sample has been poled 2 and shows that a faster polarization reversal occurs at
so that all polar sites are reversed into state 1 at the initial higher temperatures. The results of this study are very
state, P1(0) = 1, which results in N = N1(0) and N2(0) = important because they suggest a substantial thermal de-
0. We now apply instantaneously a constant electric field pendence of the switching time, which should be care-
E to reverse the polarization into state 2. Over a period of fully considered in the design, operation and stability
time t, the total amount of polarization (electric displace- considerations of memory devices based on ferroelectrics.
ment) switched from state 1 into state 2 is equal to: The coercive field derived from the standard Landau-
Ginzburg-Devonshire (LGD) model overestimates most of
∆D(t) = 2p s ⋅ N 2(t)/N = 2p s ⋅ (1 − P1(t)), (8) the experimental data. Many experimental studies showed
that the coercive field is very susceptible to the measure-
where ps is the spontaneous polarization. Using (4), the ment frequency [15] or the measurement temperature [1],
time-dependent expression for the switched polarization/ [16]. Our approach to the switching can explain such ef-
electric displacement is obtained: fects. Let’s assume again that the sample has been poled
at time t = 0 into state 1 [P1(0) = 1, positive polariza-
∆D(t) = 2p s ⋅ (1 − e −(t /t sw)) ⋅ (1 − P1(∞)). (9) tion]. We now apply an arbitrary constant negative E field
to the sample, opposite to the polarization direction. The
Eq. (9) is in good agreement to the KAI model when field will induce the nucleation and switching of elementa-
the Avrami exponent is n = 1. However, the expressions ry sites into state 2 (negative polarization). Under the ac-
derived in our model contain explicit dependence on the tion of the negative field, in time, enough sites will switch
E field and temperature via P1(∞) and tsw. Moreover, our into state 2 so that the total net polarization of the sample
relations, in comparison to those derived in the KAI mod- is zero. In terms of the hysteresis loop, the applied field at
el, contain the factor (1 − P1(∞)), which can take posi- which this condition is met is called the coercive field (Ec).
tive values from 0 to 1. This accounts for the amount of However, in terms of our switching model, the occupation
reversed polarization, which can be 0 if P1(∞) = 1 and 2ps probabilities of the two possible states are equal at the
if P1(∞) = 0 (i.e., full reversal). Fractional values 2ps(1 − coercive field: P1(t) = P2(t) at Ec. Because P1(t) + P2(t)
P1(∞)), are obtained for P1(∞) ≠ 0 or 1, which accounts = 1, P1(t) = 0.5. Using (4), the condition P1(t) = 0.5 and
for the fact that at equilibrium not all the polar sites are the fact that P1(∞) ≪ 1 for a large negative E field, we
necessarily reversed. For numerical simulations, we have determined that the coercive field can be written as [10]
used the following parameters: α = −11.57 × 107 V·m/C;
β = 2.1 × 108 V·m5/C3; kB = 8.61 × 10−5 eV/K (1.38
× 10−23 J/K); V* = 10−26 m3. This results in an energy E c(t,T ) ≅ W B/p s − k BT /V *p s ⋅ ln(υot ln(2)−1). (10)
1870 IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control , vol. 58, no. 9, September 2011
t = t sw + t d, (11)
Fig. 3. Time-dependent coercive field for different temperatures. The
where tsw is the switching nucleation time given in (7) and
increase in both the measurement time and the temperature has the ef-
fect of lowering the Ec. td is the expansion time of the reversed domain. If vd is
the velocity of the domain wall, then in the time interval
td = t − tsw the domain will expand by the distance vd(t
This analytical expression allows studying the time and − tsw). The area of the expanded domain can be written
temperature dependence of the coercive field of ferroelec- in general as [25]
tric materials. It is important to mention that this relation
can be applied to any type of energy barrier WB, including Ad = C d[v d(t − t sw)]n, (12)
complex energy barrier distributions. V* is the volume of
the elementary nucleation site, which is taken as 10−26 m3 where n is related to the dimensionality of the shape of the
in this work (although it doesn’t appear in previous equa- reversed domain. For a 1-D domain, the reversed area is
tions, any relation containing an energy term is, in fact, Ad = 2vd(t − tsw) and for a 2-D domain, it is Ad = πvd2(t
the product of energy ⋅ V*, because the energies are ex- − tsw)2. It has been shown that in the case of inhomoge-
pressed in energy per unit volume). neous nucleation, the domain wall velocity is proportional
Fig. 3 shows the calculated time dependence of the co- to tsw (see [20], [25]):
ercive field at different temperatures, when the time var-
ied from 1 ms to 5 s (frequency of 1 kHz to 0.2 Hz). For v d ∝ 1/t sw. (13)
the calculations, we have used the same numerical param-
eters defined previously in this section. Using relations (7) and (13), we obtain the wall veloc-
The data shows a consistent decrease of the Ec when ity as
the measurement time increases (i.e., faster measurement
v d ∝ e −W B/k BT × e γ ⋅E , (14)
results in larger Ec) and a reduction of the Ec at higher
temperatures. For example, we calculated a reduction of
where γ = ps/kBT. Eq. (14) is consistent with a creep be-
almost 50% of the coercive field when the measurement
havior of the domain wall movement [21]–[24]. The wall
time increased from 1 ms to 5 s at room temperature.
velocity depends strongly on the activation energy (i.e.,
More substantial variations are also observed when the
energy barrier), temperature, and increases rapidly with E
measurement temperature increases, with reductions of up
field. For very small E fields, the velocity reduces dra-
to 500% when the temperature was increased from 200K
matically, reaching the lower limit of vd ∝ e −W B/k BT for E
to 350K at constant measurement frequency.
= 0. This shows that in the low-field regime, thermal ac-
tivation plays an important role in the domain wall dy-
namics and, consequently, in the polarization dynamics.
IV. Domain Wall Velocity
Energy terms related to the domain wall energy and elas-
tic interactions at pinning sites can be incorporated into
The existing model makes the assumption that the po-
the energy barrier expression as a further development of
larization switching time is dominated by the nucleation
the model by including such energy terms in the initial
time (i.e., the time taken by the critical volume V* to
energy functional equation (2).
reverse its polarization in our model) and that the time
taken by the domain wall expansion to reach full polariza-
tion reversal is negligible. This assumption is only correct
for applied E fields much larger than the coercive field and V. The Equilibrium Susceptibility
for cases of ideal materials with perfect crystal structures
and no impurities. In reality, this is not the case at all, We now impose the equilibrium para-electric conditions
and the movement of the domain wall is an important pro- to equations (3): dP1/dt = dP2/dt = 0, resulting in
vopsaroiu et al.: polarization dynamics and non-equilibrium switching processes in ferroelectrics 1871
Melvin Vopson (formerly known as Marian the development of a prototype SEM-based instrumented indentation
Vopsaroiu) received his B.Sc. and M.Sc. degrees in system, and he joined NPL in 1997 to lead the Functional Materials
physics from the University of Bucharest. Melvin Research group. Research activity includes the development of measure-
received his Ph.D. degree in physics from the Uni- ment methods to elucidate materials behavior in ferroelectric and piezo-
versity of Central Lancashire in 2002. He worked electric ceramics and thin film materials, and more recently in multifer-
as postdoctoral research fellow of the University of roic materials and materials metrology for Spintronics and Energy
York for almost 4 years, followed by an appoint- Harvesting. The focus of his research is materials metrology, and he has
ment as senior R&D engineer at Seagate Technol- published more than 80 peer-reviewed scientific papers in the field. He
ogy. Melvin joined NPL in 2006, and his research chairs the IOM3 Smart Materials and Systems Committee. He is Knowl-
interest currently revolves around experimental edge Leader for the Materials Division at NPL and Principal Research
and theoretical studies of solid-state physics with Scientist for the Multifunctional Materials technical area and also a
emphasis on ordered and multifunctional systems. Melvin has authored member of the Institute of Physics. In 2009, he was awarded the Institute
or co-authored more than 35 scientific articles with major contributions of Materials’ Verulam medal for outstanding contributions to ceramic
in the fields of thin film growth technologies, optical techniques of char- science, and is Visiting Professor at Queen Mary, University of London.
acterization of solids, and development of novel metrologies and innova-
tions involving ferroic materials.