While commonly used piezoelectric materials contain lead, non-hazardous, high-performance piezoelectrics are yet to be discovered. Charged domain walls in ferroelectrics are considered inactive with regards to the piezoelectric response and, therefore, are largely ignored in this search. Here we demonstrate a mechanism that leads to a strong enhancement of the dielectric and piezoelectric properties in ferroelectrics with increasing density of charged domain walls. We show that an incomplete compensation of bound polarization charge at these walls creates a stable built-in depolarizing field across each domain leading to increased electromechanical response. Our model clarifies a long-standing unexplained effect of domain wall density on macroscopic properties of domain-engineered ferroelectrics. We show that non-toxic ferroelectrics like BaTiO(3) with dense patterns of charged domain walls are expected to have strongly enhanced piezoelectric properties, thus suggesting a new route to high-performance, lead-free ferroelectrics.
While commonly used piezoelectric materials contain lead, non-hazardous, high-performance piezoelectrics are yet to be discovered. Charged domain walls in ferroelectrics are considered inactive with regards to the piezoelectric response and, therefore, are largely ignored in this search. Here we demonstrate a mechanism that leads to a strong enhancement of the dielectric and piezoelectric properties in ferroelectrics with increasing density of charged domain walls. We show that an incomplete compensation of bound polarization charge at these walls creates a stable built-in depolarizing field across each domain leading to increased electromechanical response. Our model clarifies a long-standing unexplained effect of domain wall density on macroscopic properties of domain-engineered ferroelectrics. We show that non-toxic ferroelectrics like BaTiO(3) with dense patterns of charged domain walls are expected to have strongly enhanced piezoelectric properties, thus suggesting a new route to high-performance, lead-free ferroelectrics.
In the last two decades, sensors and actuators based on lead-containing ferroelectrics
gained a dominant position in applications ranging from delicate positioning systems,
ultrasonic medical diagnostic tools and microsurgery devices to robust fuel injectors.
These lead-containing materials have to be replaced by environmentally benign
alternatives12. So far, the search for lead-free ferroelectrics was
mainly focused on the synthesis of new compositions exhibiting high-intrinsic piezoelectric
properties345. The possibility to enhance piezoelectric properties of
known lead-free ferroelectric crystals by so-called frustrated poling and domain
engineering has also been explored678910. In this approach, the
poling electric field is applied along a nonpolar direction, which forms equal angles with
at least two possible orientations of spontaneous polarization. During this poling, two or
more ferroelectric states are equally preferred, creating an artificial domain structure
with several coexisting domain states. Such a structure is macroscopically polar (has
non-zero net polarization) and includes a certain volume fraction of domain walls.It has been found that, in the classical lead-free perovskite ferroelectric BaTiO3, a high density of domain walls
with spacing in the micrometre range can result in up to a fourfold enhancement of
functional properties9, while it was established that the enhancement is not
associated with the domain wall motion. However, the body of available experimental data is
insufficient to clearly relate the observed effects to features of the domain patterns in
the samples. This experimental activity has drawn much attention by theorists1112131415 who were unable to explain the experimental observations
unless unrealistic assumptions were taken. Thus, despite intriguing experimental findings,
the future of fine-scale domain patterns as systems with enhanced electromechanical
properties remains obscure.Here we demonstrate the existence of an enhancement mechanism of the dielectric and
piezoelectric response in ferroelectrics with fine-scale charged domain-wall patterns. This
mechanism provides a realistic interpretation of the experimental observations and outlines
a new direction in the search for high electromechanical response in lead-free
ferroelectrics.In the experimental reports on enhanced properties due to engineered domain structures, we
recognized that head-to-head and tail-to-tail domain configurations16171819, such as those shown in Fig. 1, must have
been formed during the frustrated poling process of BaTiO3 (refs 8,10). The reports810 describe poling techniques
leading to the controlled formation of domain structures with various domain-wall
densities. Analysis of the domain walls and crystal orientations leads to indirect
identification of 90°-charged domain walls10 which occupy an
unspecified fraction of crystal volume. However, the significance of this formation of
charged domain walls passed entirely un-noticed.
Figure 1
Free carrier concentration and band bending at charged domain walls.
(a) An electroded (110)c plate of tetragonal BaTiO3. (b) A periodic
structure of charged 90° domain walls where bound polarization charge (+,
−) induced by divergence of polarization, P0, is almost
perfectly compensated by free charge (, ) whereas the imperfection of compensation creates
built-in depolarizing field Edep. (c) Phase-field
simulation-calculated band bending induced by Edep. The bending
causes the edges of the conduction EC or valence
EV bands (solid black lines) to approach the Fermi level
EF (dashed black line) where high concentration of free
electrons (red line) or holes (blue line) are generated and become available for
compensation of the bound charge. The bandgap between the conduction and valence
bands is assumed 3 eV. The competing band bending and charge compensation are in
equilibrium when conduction and valence bands cross the Fermi level about
~0.22 eV at head-to-head and tail-to-tail domain walls, respectively.
Here we report an analysis, based on Landau–Ginzburg–Devonshire theory and
the theory of semiconductors, which establishes that the dielectric and piezoelectric
response of such systems with charged domain walls increase strongly on the reduction of
the domain period. This is expected to happen under two conditions: the ferroelectric phase
transition in the material should be of first order and a strong dielectric anisotropy must
exist, namely the dielectric constant in the direction perpendicular to the spontaneous
polarization, a, should be much larger than that
in the direction parallel to it, c. Both these
conditions are met for BaTiO3
(ref. 20) and KNbO3 (ref. 21).
Results
Charged domain walls
Charged domain walls contain a high density of bound charges, which must be strongly
compensated by free carriers to prevent a huge depolarizing field that would prohibit
formation of dense periodic domain structures. For realistic parameters of
ferroelectric perovskites (unless these are heavily doped), such compensation cannot
be provided with the typical carrier concentration in these materials1622. Most of the screening charge originates from electron transfer
across the forbidden energy gap, Eg. This means that, at the
head-to-head wall (positive bound charge), the bottom of the conduction band of the
material should approach the Fermi level, providing electrons for the screening.
Analogously, at the tail-to-tail wall (negative bound charge), the top of the valence
band approaches the Fermi level, providing screening holes. This implies the
existence of a potential difference of about Eg/q (q
is the elementary charge) between such couple of walls. As a result, the domain
between these walls is exposed to an electric field, Edep, which
depends on the forbidden gap and the domain width, w, as expressed in equation (1),This field is a result of the incomplete compensation of bound polarization charge
with free carriers at the domain walls. The imperfect compensation is a principal
feature of the system because a perfect compensation would result in zero-band
bending, and then the free carriers needed for compensation would not be available at
all. Therefore, the system keeps an optimal trade-off between imperfect screening and
a minimal necessary band bending. In the case of an electroded (110)c (the
subscript c denotes the pseudocubic orientation) plate of tetragonal
BaTiO3 (Fig. 1a) containing a pattern of charged domain walls (Fig.
1b), Eg=3 eV and with domain width w of several
micrometres, the depolarizing field Edep can readily reach values
of 10–20 kV cm−1 (equation (1)). It
is this built-in field (growing with domain-wall density) that induces a polarization
rotation leading to enhanced dielectric and piezoelectric response of each
domain.
Polarization rotation
The mechanism of polarization rotation is qualitatively illustrated in Fig. 2 for the case of tetragonal BaTiO3. The contour plot of bulk free-energy
density, Fig. 2a, shows two global minima corresponding to two
different orientations of polarization in the tetragonal phase and one local minimum
for the orthorhombic phase. The minimal energy path for polarization switching
between the two tetragonal states inevitably follows a rotation-like path across the
orthorhombic minimum. The key factor responsible for the enhanced properties is that
polarization rotated by Edep approaches a point of instability
between tetragonal and orthorhombic phases. At this point, a small excitation by an
external electric field in the [110]c direction is able to induce a large
change of polarization along the dash-dotted line in Fig. 2.
The essential characteristic of a material that allows the enhancement effect is
illustrated in Fig. 2b. It shows a surface that encloses a
region where values of polarization lead to energy lower than −2.18 MJ
m−3. Polarization is stable in the tetragonal minima (white
circles), but follows a minimal energy path during switching between the tetragonal
states. The important feature is the non-zero change of the
P3-component during switching. The energy profile corresponds to
BaTiO3, but a
qualitatively equivalent situation is available, for example, in a (111)c
plate of KNbO3.
Figure 2
Enhancement of the dielectric response by polarization rotation.
The path for the rotation of the polarization vector P0, which
is induced by electric field Edep, is given by the profile of
the bulk free energy (a). The dash-dotted line illustrates a path with
minimal energy barrier for switching between domain states of the tetragonal phase
(end points of the dash-dotted line). The key feature of the illustrated mechanism
is the Edep-induced polarization rotation by an angle
θ that is accompanied by a change of
P3-component of the polarization, approaching the point of
thermodynamic instability. The three-dimensional contour plot (b) shows the
surface of a region where the values of spontaneous polarization give energy lower
than −2.18 MJ m−3. The enhancement effect is
available when the minimal energy paths between the tetragonal minima (white
circles) are curved into the P3 direction.
In view of the strong dielectric anisotropy of BaTiO3 (experimentally at room temperature
a/c30) (ref. 20), it is the value of a that controls the total dielectric response of
the system. For this reason, we discuss the impact of the built-in field
Edep on this quantity alone. Let us consider the case where the
application of the field leads to small changes of the ferroelectric polarization. In
this approximation, the relative reduction of the absolute value of the polarization
can be evaluated as is (P0 is the
spontaneous polarization) with the transverse component being
γδP0 where γ=
a/c is the dielectric anisotropy
factor. Using the polarization equation of state in the X–Z plane, one can
readily describe an impact of the built-in field on a in terms of the following relation between the modified value
of the transverse component of the dielectric constant a and δ:Here factors A and B can be readily expressed in terms of the
coefficient of the Landau expansion23: for BaTiO3 at room temperature
A≈0.3 and B≈0.6. The sign of the coefficient B has
an essential role in the enhancement mechanism and it can only be positive in
materials that exhibit a first-order phase transition. The A-containing term
in equation (2) describes the elongation and contraction of
polarization, whereas the B-containing term controls the effect of
polarization rotation. In the linear approximation, the rotation angle θ
can be expressed in terms of relative change of the polarization δ :
θ=γδ. Keeping this in mind, one sees that the
positive sign of B implies that the polarization rotation leads to a
permittivity enhancement. Conversely, if the ferroelectric transition in the material
were of the second order, B would be negative and the polarization rotation
would not lead to enhancement of permittivity. With a large value of γ,
one concludes from the equation (2) that polarization rotation
is the leading effect. The angle corresponding to substantial enhancement of
a can be evaluated from the condition
Bγθ21, yielding
|θ|13° for BaTiO3. Such a level of
polarization rotation can be reached under realistic values of the built-in field.
This is clear from Fig. 3a, which shows the calculated
dependence of θ on Edep. Thus for the domain pattern
addressed, the electric field arising from incomplete compensation of the bound
charge on head-to-head and tail-to-tail walls is strong enough to dramatically
enhance the dielectric response of the system. Similar effects of enhancement of the
dielectric and piezoelectric properties can be expected also in KNbO3 and in ferroelectrics at
their morphotropic boundary.
Figure 3
Depolarizing field-induced polarization rotation and impact on
permittivity.
The angle θ growing with Edep (a) which
results in an increase of the transverse dielectric response a (b). Displayed curves correspond to
temperatures 278 K (blue), 288 K (green), 298 K (red), and 308 K (orange).
The semi-quantitative arguments following from equation (2) are
supported by more detailed Landau-theory calculations (see Methods) that show the unlimited growth of the transverse component of
the dielectric constant when approaching a critical angle as shown in Fig. 3b.
Properties of ferroelectrics with charged domain walls
We verified the enhancement mechanism described above by means of phase-field
simulations of a poly-domain system, where the ferroelectric was treated as a
wide-bandgap semiconductor in the effective mass approximation (see Methods). First, stability of 90°-charged domain walls was
confirmed. Then, we have calculated piezoelectric, d33 and
d31, and dielectric 33
coefficients as functions of domain-wall spacing w, which are experimentally
measurable parameters of the system. The phase field results (circles and squares in
Fig. 4a–c) include contributions of domain response
and intrinsic response of domain walls. To identify the magnitudes of these
contributions, we performed the Landau-theory calculations for a single domain that
is artificially exposed to depolarizing electric field applied along x axis
(Fig. 1b). The magnitude of this electric field was obtained
from phase-field simulation as approximately
Edep=ΔVeff/w where
ΔVeff=3.5 V is an 'effective' potential difference between
domain walls. This potential difference is larger than Eg /q = 3 V
due to the shape of the band bending (Fig. 1c). However, as
seen in Fig. 4a–c, the intrinsic domain wall
contribution, which is the difference between single-domain and poly-domain response,
is negligible. This conclusion is surprising with respect to the huge domain-wall
thickness (Fig. 4d), but it agrees with studies done in the
past111213, where intrinsic domain-wall response failed to
explain the experimental observations78910. The large thickness
of charged domain walls, DW>100 nm, compared with neutral ones,
DW~1 nm, is qualitatively consistent with a recently published
theory16 that explains the widening mechanism of charged domain
walls in detail. Our calculations also show temperature T and
Eg dependence of the smallest theoretically possible domain-size
wlim (Fig. 4e).
Figure 4
Enhancement of macroscopic properties with decreasing domain size.
Relative permittivity 33 (a), and
piezoelectric coefficients d33 (b), and
d31 (c), are plotted as a function of domain width
w. The solid lines represent the inner response of a domain that is
exposed to built-in depolarizing field owing to imperfectly compensated charged
domain walls. Displayed curves correspond to temperatures 278 K (blue lines), 288
K (green lines), 298 K (red lines), and 308 K (orange lines). Red circles and
green squares represent phase-field simulation results that include intrinsic
contributions of charged domain walls at temperatures 288 K (green squares) and
298 K (red circles). Domain-wall width DW against domain width w
(d), shows small domain-wall broadening when critical domain width is
approached. (e) The critical (minimal) domain width wlim
against temperature T is shown for bandgaps 2 eV (dash-dotted), 3 eV
(dashed), and 4 eV (solid).
The phase-field calculated band-bending profile and free-charge concentrations are
shown in Fig. 1c, where the difference between the edge of the
conduction (or valence) band and the Fermi level in the wall centre, ΔE,
can be evaluated for the condition that the charge density of the free carriers at
this point, (meff and ħ are
the effective mass of electrones/holes, and the reduced Planck constant,
respectively) is about the mean bound charge density in the walls,
√2P0/w. Such estimates, with
meff=0.12me (me is the free
electron mass), w=100 nm and P0=0.26 C
m−2 yield ΔE=0.24 eV in reasonable agreement
with the results of phase-field simulation where ΔE≈0.22 eV.The large concentration of free carriers at charged domain walls (Fig.
1c) may raise the question as to whether the expected high domain-wall
conductivity cannot short-circuit the electrodes and hence make the enhancement
phenomena practically unavailable. To answer this question, we simulated a
finite–6 μm thick–electroded sample. One can see, in a
colour-scale map in Fig. 5, that the free-carrier concentration
dramatically drops in the vicinity of the electrodes. This results in the formation
of ~300 nm or ~1 μm gaps between electrodes and regions with
higher free-charge concentration. These gaps are formed because of the flat profile
of the electric potential (constant potential) at the electrodes that prohibits
band-bending and, hence, a high concentration of free-carriers in the vicinity of the
electrodes. The results in Fig. 5 correspond to the situation
where the work functions of the electrode and ferroelectric are identical, but, in
general, the size of the gaps is dependent on the difference between the work
functions.
Figure 5
Charged domain walls in the vicinity of electrodes.
Colour-scale graphs show the distribution of free-carrier concentration in a
ferroelectric with charged domain walls. The concentration is maximal in the
centres of charged domain walls, but it dramatically drops in the vicinity of the
electrodes. This results in a formation of ~300 nm (a,d), and
~1 μm (b,c), gaps between the electrodes and regions
with higher free-charge concentration. These gaps are formed due to the flat
profile of the electric potential (constant potential) at the electrodes that
prohibits band-bending and, hence, a high concentration of free-carriers in the
vicinity of the electrodes. The results correspond to the situation where the work
functions of the electrode and ferroelectric are identical, but, in general, the
size of the gaps is dependent on the difference between the work functions.
Discussion
Our analysis shows that the built-in electric field, created in periodic domain patterns
with charged domain walls, grows significantly with increasing domain-wall density. Once
the domain period is in the range of several micrometres, the built-in field can lead to
an appreciable enhancement of the dielectric and piezoelectric response of the
polydomain system. We identify the features that control this phenomenon, namely a very
high dielectric anisotropy factor γ=
a/c and a first-order ferroelectric
phase transition.The presented results are qualitatively in good agreement with experimental observations
of domain-wall density effects on macroscopic properties of [111]c-poled
BaTiO3 crystals78910 (Fig. 6). In these experiments, the
piezoelectric and dielectric constants were subject to a fourfold enhancement when
domain size reached three-to-five microns. We suggest the presented mechanism as a new
degree of freedom in the search for ferroelectrics with high piezoelectric response that
is particularly attractive in light of the need for lead-free piezoelectrics.
Figure 6
Experimental data for the effect of domain density on macroscopic piezoelectric
response.
Experimental dependence of piezoelectric coefficient d31 on
domain size w in poled (111)c plate of single crystal
BaTiO3 . Data are
combined from refs 9, 10.
Methods
The phase-field model
The presented results were obtained primarily by means of phase-field simulation that
incorporates coupling between ferroelectric and wide-bandgap semiconductor
properties. Our simulation includes elastic interactions that results in stable
90°-charged domain walls.Model equations are obtained by Lagrange principle from Helmholtz free-energy
density24:where Pi is the polarization, Pi,j its
derivatives (the subscript ',j' represents the operator of spatial derivatives
∂/∂xj), Di the electric
displacement and
eij=1/2(ui,j+uj,i) is the
elastic strain where ui is a displacement vector.The bulk free-energy densityis expressed for a zero strain as a six-order polynomial expansion13,
where αi, αij(e),
αijk are parameters fitted to the single-crystal
properties (Table 1). The remaining contributions represent
bilinear forms of densities of elastic energy
fela[{eij}]=1/2cijkleijekl,
where cijkl is the elastic stiffness, electrostriction energy
fes[{Pi,
eij}]=−qijkleijPkPl,
where qijkl are the electrostriction coefficients, gradient energy
fwall[{Pi,j}]=1/2GijklPi,jPk,l,
where Gijkl are the gradient energy coefficients, and electrostatic
energy fele[{Pi,
Di}]=1/(2
0
B)(Di−Pi)2,
where 0 and B are permittivity of vacuum and relative background
permittivity, respectively. The zero-strain coefficients
αij(e) can be expressed in terms of usually
introduced stress-free coefficients αij as follows:
Table 1
Values of material coefficients for BaTiO3 used in the simulations.
Parameter
Value
Unit
α1
(T−381)3.34×105
J m C−2
α11
(T−393)4.69×106−2.02×108
J m5 C−4
α12
3.23×108
J m5 C−4
α111
(393−T)5.52×107+2.76×109
J m9 C−6
α112
4.47×109
J m9 C−6
α123
4.91×109
J m9 C−6
c11
27.5×1010
J m−3
c12
17.9×1010
J m−3
c44
5.43×1010
J m−3
q11
14.2×109
J m C−2
q12
−0.74×109
J m C−2
q44
6.28×109
J m C−2
Q11
0.1104
m4 C−2
Q12
−0.0452
m4 C−2
Q44
0.0578
m4 C−2
G11
51×10−11
J m3 C−2
G12
−2×10−11
J m3 C−2
G44
2×10−11
J m3 C−2
Γ
4×104
C2 J−1 m−1
s−1
B
7.35
1
μn,μp
0.01
cm2 V−1
s−1
τ
100
ps
EC
−3.6
eV
EF
−5.1
eV
Ev
−6.6
eV
N
1×1024
m−3
By using the Legendre transformation to electric enthalpywhere E = -, is the electric field and the electric potential, and using Lagrange principle, we can uniformly
express the set of field equations that govern the kinetics of ferroelectrics:Equation (5) defines the mechanical equilibrium while inertia is
neglected. Equation (6) represents Gauss's law of a dielectric
including a non-zero concentration of free electrons n and holes p.
Equation (7) is the time-dependent
Landau–Ginzburg–Devonshire equation25 which governs the
spatiotemporal evolution of spontaneous polarization with kinetics given by
coefficient Γ.Coupling between the ferroelectric/ferroelastic system with its semiconductor
properties is introduced by considering a non-zero density of free carriers
(electron-hole) in the electrostatic equation (6). The
semiconductor properties were introduced under the assumption of zero concentration
of dopants, which was shown to be an acceptable approximation to a small
concentration of defects and dopants17. The distribution of free
carriers is governed by continuity equations:where electron and hole currents Ji( and
Ji(, respectively, are governed by drift
and diffusion as follows: . Here
μn and μp are electron and hole
mobilities, respectively. Because we analyse only the stationary solution in thermal
equilibrium, we can introduce computationally convenient form of recombination rates
Rn and Rp as follows:
Rn=−(n−n0)/τ
and
Rp=−(p−p0)/τ,
where τ is life-time constant and n0 and
p0 are electron and hole concentrations in thermal
equilibrium:Here F1/2 is the Fermi-Dirac integral. Density of states is given
by effective mass approximation:where effective mass meff=cme is assumed equal
for electrons and holes. As the effective mass of free carriers in BaTiO3 varies in literature, we
tested the model with a wide range of effective mass constants, c0.05, 50, showing an insignificant impact on the enhancement
phenomena. However, the constant c controls the screening regime
(classical/degenerate) and width of the domain walls. Results presented in the graphs
correspond to c=0.117, that is, N=1024
m−3 used by Xiao et al.26The ferroelectric-semiconductor model was designed as a subdomain of size
2w×0.25(μm)2 with applied periodic boundary
conditions joining the solution of all solved variables between opposite parallel
boundaries. The numerical solution of equations (5), (6), (7), (8) and (9) on the defined subdomain was performed by a finite element
method with linear triangular elements of size 4 nm in the vicinity of domain walls
and 40 nm inside domains. The simulation starts from initial conditions that are
defined as zero for all variables except polarization which is , for P0=0.262 C
m−2, in regions separated by 70 nm gap, where P=(0,0).
The simulation reaches thermal equilibrium in <5 ns and gives solutions for the
spatial distribution of polarization Pi, mechanical displacement
ui, electric potential ϕ, and concentrations of
electrons n and holes p. We used these to calculate the band-bending
and charge concentrations in Fig. 1c and domain-wall width and
critical domain size in Fig. 4d,e. Then the domain structure is
exposed to (i) small compressive stress, 100 kPa, and (ii) small electric field, 4 V
mm−1, when the average change of electric displacement is
integrated and (i) the direct piezoelectric and (ii) dielectric coefficients are
calculated (Fig. 4a–c). This calculation also shows
almost zero domain-wall widening when small electric field or stress are applied.
The single-domain Landau-theory approximation
The phase-field results are accompanied by Landau-theory calculations of a single
domain, which is subjected to an artificially introduced depolarizing field obtained
from the phase-field model. This approach calculates macroscopic properties of
domains by minimization of Gibbs free energy. It allows domain wall and domain
contributions to be distinguished.The Gibbs free energy of homogeneous stress-free sample is defined asHere the non-indexed symbols P,E represent vectors. The bulk
free-energy density fbulk is defined as fbulk =
fbulk(e) + fela +
fes for stress-free homogenous sample, that is, at
eij=QijklPkPl.
Here Qijkl are the coefficients of the direct electrostriction
effect.The calculation searches for the minimum of G(P, E) for non-zero
electric field E that gives polarization as follows:Using P(E) and strain induced by electrostriction effect
eij(E)=QijklPk(E)Pl(E),
we calculate effective piezoelectric coefficients:where the starred symbols are expressed in the coordinate system of a
(110)c crystal.The presented results correspond to phenomenological parameters as introduced in
Table 1.
Author contributions
N.S. initiated and supervised the project and contributed to the manuscript. A.K.T.
initiated work on charged domain walls. D.D., A.K.T. and T.S. suggested the principle
idea and wrote the manuscript. T.S. designed, performed and analysed simulations. T.S.,
A.K.T. and M.G. performed analytical calculations.
Additional information
How to cite this article: Sluka, T et al. Enhanced electromechanical
response of ferroelectrics due to charged domain walls. Nat. Commun. 3:748 doi:
10.1038/ncomms1751 (2012).
Authors: Ruijuan Xu; Shi Liu; Ilya Grinberg; J Karthik; Anoop R Damodaran; Andrew M Rappe; Lane W Martin Journal: Nat Mater Date: 2014-10-26 Impact factor: 43.841
Authors: Gabriel Sanchez-Santolino; Javier Tornos; David Hernandez-Martin; Juan I Beltran; Carmen Munuera; Mariona Cabero; Ana Perez-Muñoz; Jesus Ricote; Federico Mompean; Mar Garcia-Hernandez; Zouhair Sefrioui; Carlos Leon; Steve J Pennycook; Maria Carmen Muñoz; Maria Varela; Jacobo Santamaria Journal: Nat Nanotechnol Date: 2017-04-10 Impact factor: 39.213