University of Nebraska - Lincoln

DigitalCommons@University of Nebraska - Lincoln

Alexei Gruverman Publications Research Papers in Physics and Astronomy

4-6-2012

Mechanical Writing of Ferroelectric Polarization

Haidong Lu
University of Nebraska-Lincoln, s-hlu4@unl.edu

C.-W. Bark
University of Wisconsin–Madison

D. Esque de los Ojos

Universitat Politecnica de Catalunya, Barcelona, Spain

J. Alcala
Universitat Politecnica de Catalunya, Barcelona, Spain

Chang-Beom Eom
University of Wisconsin-Madison, eom@engr.wisc.edu

See next page for additional authors

Follow this and additional works at: https://digitalcommons.unl.edu/physicsgruverman

Part of the Physics Commons

Lu, Haidong; Bark, C.-W.; Esque de los Ojos, D.; Alcala, J.; Eom, Chang-Beom; Catalan, G.; and Gruverman,
Alexei, "Mechanical Writing of Ferroelectric Polarization" (2012). Alexei Gruverman Publications. 56.
https://digitalcommons.unl.edu/physicsgruverman/56

This Article is brought to you for free and open access by the Research Papers in Physics and Astronomy at
DigitalCommons@University of Nebraska - Lincoln. It has been accepted for inclusion in Alexei Gruverman
Publications by an authorized administrator of DigitalCommons@University of Nebraska - Lincoln.
Authors
Haidong Lu, C.-W. Bark, D. Esque de los Ojos, J. Alcala, Chang-Beom Eom, G. Catalan, and Alexei
Gruverman

This article is available at DigitalCommons@University of Nebraska - Lincoln: https://digitalcommons.unl.edu/

physicsgruverman/56
Published in Science vol. 336, no. 6077 (April 6, 2012), pp. 59–61; doi: 10.1126/science.1218693 Online at http://www.sciencemag.org/con-
tent/336/6077/59.abstract Copyright © 2012 American Association for the Advancement of Science. Used by permission.
Submitted January 4, 2012; accepted for publication February 23, 2012.

Mechanical Writing of Ferroelectric Polarization

H. Lu,1 C.-W. Bark,2 D. Esque de los Ojos,3 J. Alcala,4 C. B. Eom,2 G. Catalan,5,6 and A. Gruverman 1
1. Department of Physics and Astronomy, University of Nebraska–Lincoln, Lincoln, NE 68588, USA
2. Department of Materials Science and Engineering, University of Wisconsin–Madison, Madison, WI 53706, USA
3. Department of Fluid Mechanics, Grupo Interdepartamental para la Colaboración Científica Aplicada (GRICCA), Universitat Politecnica de Catalunya,
Barcelona, Spain
4. Department of Materials Science and Metallurgical Engineering, GRICCA, Universitat Politecnica de Catalunya, Barcelona, Spain
5. Institut Catala de Recerca i Estudis Avançats, (ICREA) Catalunya, Spain
6. Centre for Investigations in Nanoscience and Nanotechnology (CIN2), Consejo Superior de Investigaciones Cientificas (CSIC) and Institut Catala de
Nanotecnologia (ICN), Campus de Bellaterra, Barcelona, Spain
Corresponding authors — G. Catalan & A. Gruverman; gustau.catalan@cin2.es ; agruverman2@unl.edu

Abstract
Ferroelectric materials are characterized by a permanent electric dipole that can be reversed through the application of an external voltage, but a
strong intrinsic coupling between polarization and deformation also causes all ferroelectrics to be piezoelectric, leading to applications in sensors and
high-displacement actuators. A less explored property is flexoelectricity, the coupling between polarization and a strain gradient. We demonstrate that
the stress gradient generated by the tip of an atomic force microscope can mechanically switch the polarization in the nanoscale volume of a ferroelec-
tric film. Pure mechanical force can therefore be used as a dynamic tool for polarization control and may enable applications in which memory bits are
written mechanically and read electrically.

Ferroelectrics strongly couple changes in dients, rather than homogeneous strain, can be We have explored whether flexoelectric-
polarization of a material with its deforma- exploited for switching polarization and “writ- ity can actively switch ferroelectric polarization
tion, which gives rise to useful electromechan- ing” the domain bits into ferroelectric memories. by mechanically pushing AFM tips onto the sur-
ical phenomena including piezoelectric, elec- This process is allowed by symmetry because a face of epitaxial single-crystalline BaTiO3 films,
trostrictive, and flexoelectric effects. Coupling strain gradient, unlike a homogeneous strain, is thereby inducing large and localized stresses.
between polarization and homogeneous in- an odd-parity tensor without inversion symme- The films were fabricated by atomic layer con-
plane strain can also be used to tune the sym- try—strain gradients have directionality and po- trolled growth on atomically smooth (001) Sr-
metry and properties of ferroelectric thin films larity (8, 9). Although flexoelectricity is gener- TiO3 substrates with La0.67Sr0.33MnO3 conduc-
via strain engineering using different substrates ally weaker than piezoelectricity, gradients grow tive buffers that served as bottom electrodes
(1, 2). For a ferroelectric, however, the spon- in inverse proportion to the relaxation length, so (13, 14). Compressive stress induced by the sub-
taneous strain (unit-cell deformation) is iden- very large flexoelectric effects can be achieved at strate ensured that polarization was aligned in
tical whether the polarization is pointing up the nanoscale (10–12). the direction perpendicular to the surface, and
or down, so one does not immediately think
of mechanical deformation as a viable mecha-
nism to invert the polarization in a ferroelec-
tric memory.
Stress can nevertheless be used to influence
polarization. A large local deformation can be re-
alized by pressing the sharp tip of an atomic force
microscope (AFM) against the surface of a film,
thereby causing a large stress concentration near
the tip-sample contact. Previous studies showed
suppression of the piezoelectric response under
a sufficiently high loading force (3, 4), with al-
most complete recovery after stress release.
Also, polarization vector rotation, caused by si-
multaneous application of an external electrical
bias and tip-induced stress, has been observed in
polycrystalline thin films (4, 5), although in these
studies the random orientation of the polariza-
tion in the crystals allowed for partial ferroelas-
tic rotation rather than pure 180° inversion of
polarization. The stress from adjacent grains also
meant that the original polar configuration was
recovered upon release of the tip pressure, so no
permanent writing was achieved.
Recent reports by Lee et al. (6) show that the Figure 1. Mechanically induced reversal of ferroelectric polarization. (A and B) PFM phase (A) and
amplitude (B) images of the bidomain pattern electrically written in the BaTiO3 film. (C) Single-
flexoelectric effect caused by strain gradients can
point PFM hysteresis loops of the BaTiO3 film. (D and E) PFM phase (D) and amplitude (E) images
create a strong imprint in uniaxial, perfectly ori- of the same area after the 1-by-1–μm2 area in the center (denoted by a dashed-line frame) has been
ented ferroelectric thin films, and flexoelectric- scanned with the tip under an incrementally increasing loading force. The loading force was in-
ity due to substrate bending was also invoked in creasing in the bottom-up direction [denoted by a black arrow in (E)] from 150 to 1500 nN. (F) PFM
the imprint of polarization in polycrystalline fer- amplitude as a function of the loading force obtained by cross-section analysis along the white ver-
roelectric films (7). This suggests that strain gra- tical line in (E).

59
60 H. Lu et al. in Science 336 (2012)

only 180° inversion of polarization would be induced by the AFM tip when pressed against the The flexoelectric field was also incorpo-
allowed. BaTiO3 films with a thickness of 12 BaTiO3 film surface (the structure is sketched in rated into free-energy calculations of the Ba-
unit cells, or ~4.8 nm, have been chosen so as Figure 2A). The calculations are for a tip force of TiO3 thin film epitaxially clamped on the
to ensure epitaxial clamping and prevent mis- 1000 nN over a tip-sample contact area of 10-nm SrTiO3 substrate (Figure 2C). Under a homo-
match strain relaxation (15). radius, and the calculated strain distributions for geneous compressive uniaxial stress (23), the
Initial testing of the films by means of piezo- all the strain components are mapped (Figure S2) height of the barrier separating the two en-
response force microscopy (PFM) show that as- (18). Using the calculated strain gradients, the ergy minima is decreased, but the double well
grown BaTiO3 films are in a single-domain state flexoelectric field is obtained by multiplying the remains symmetric, and thus no specific po-
with out-of-plane polarization, indicating effec- strain gradient times the flexoelectric tensor and larity is favored, as expected from symmetry.
tive screening of the depolarizing field by sur- dividing by the dielectric constant (9, 19). To be Flexoelectricity, on the other hand, generates
face adsorbates (16). Bipolar domain patterns conservative in our calculations, we use the the- a polar bias consistent with the experimen-
can be generated conventionally with an electri- oretical values for the flexoelectric tensor coeffi- tal observation of mechanical switching. When
cally biased PFM tip: the film surface is scanned cients (20), which are a thousand times smaller incorporated to the free energy, the flexo-
with a tip under ±4-V bias exceeding the co- than the experimental ones (21, 22). The integral electric bias destabilizes the positive side (up-
ercive voltage. The 2-by-2–μm2 PFM images of of the flexoelectric field is the flexoelectric po- ward-pointing polarization) of the double well
this electrically written domain structure are tential, which we have calculated and is mapped and forces the switch to the downward-point-
shown in Figure 1, A and B. A typical value of in Figure 2B [details of the calculations are pro- ing polarization state, as observed. This is sup-
the contact force during conventional PFM im- vided in (18)]. ported by PFM hysteresis loops measured
aging is ~30 nN. A stable and uniform PFM am-
plitude signal across the domain boundary illus-
trates effective electric switchability of the film
and strong polarization retention. Local PFM
spectroscopic measurements (Figure 1C) ex-
hibit bipolar piezoelectric hysteresis loops.
The mechanical switching has been inves-
tigated by scanning a 1-by-1–μm2 area of the
bipolar domain pattern with the electrically
grounded tip under an incrementally increas-
ing loading force from 150 to 1500 nN, with a
corresponding change in the applied stress from
0.5 to 5 GPa (approximating the tip-surface
contact area as a disk of 10 nm in radius). Note Figure 2. (A) Sketch of the strain gradient
and associated flexoelectric field (arrows) in-
that, although the maximum local stress is very duced by the AFM tip pushing on the surface
large, it is still well below the threshold (~20 of the BaTiO3/La0.67Sr0.33MnO3 heterostruc-
GPa) for irreversible plastic damage of the Ba- ture. (B) Variation of the flexoelectric voltage
TiO3 surface (17). After that, a larger area of 2 in the film region under the tip, assuming a
by 2 μm2 is imaged by conventional PFM with a compressive uniaxial stress of 3.2 GPa, equiv-
low load of 30 nN (Figure 1, D and E). alent to 1 μN of force over a circular area of 10
The tip-induced stress reverses the PFM nm in radius. The field is perpendicular to the
phase contrast in the left half of the image in voltage gradient and reaches a strength of ~2
MV/cm near the surface. (C) Free-energy cal-
Figure 1D, from dark to bright, indicating in- culations for the epitaxially clamped BaTiO3
version of the polarization from up to down. film without any tip pressure (blue curve),
Figure 1E shows a nonmonotonous change in with homogeneous compressive stress
the corresponding PFM amplitude image of the of 3.2 GPa (red curve), and with the calculated flexoelectricity from the tip-induced strain gradient
flexoelectrically switched domain: Initially, the (green curve). Flexoelectricity skews the double well, forcing polarization switching toward the sta-
amplitude decreases as load increases and then, ble downward state.
at an applied force of ~750 nN, it increases Figure 3. Fabrication of na-
again (Figure 1F). This type of behavior is analo- noscale domain patterns by
gous to the polarization-reversal process in con- mechanical means. (A) Do-
ventional (voltage-induced) PFM (Figure S1) main lines mechanically
(18) in which the electromechanical amplitude written in the BaTiO3 film
signal passes through a minimum during switch- by scanning the film with a
ing, as in Figure 1C. This is caused by the for- tip under a loading force of
mation of 180° domains (antiparallel polariza- 1500 nN. (B) The same do-
main structure modified by
tion) so that the net polarization and associated electrical erasure of the me-
piezoelectric signal go through zero when the chanically written domains.
volume fractions of domains with opposite po- Erasure has been performed
larization become equal. Beyond that, the PFM by scanning the central seg-
amplitude grows again while the PFM phase is ment with the tip under a dc
changed by 180°, indicating that the polariza- –3-V bias. (C) Topographic
tion has been inverted. For reference, the PFM image of the same area ac-
images of conventionally (that is, electrically) quired after mechanical writ-
ing showing that the film sur-
switched domains in the same BaTiO3 film are face was not affected by the
shown in the supplementary materials (Figure writing process. (D) An array
S1) (18). The polarization patterns generated by of flexoelectrically written
an electrical bias (Figure S1) (18) and by me- dot domains illustrating the
chanical load (Figure 1) are identical. possibility of using mechan-
To rationalize the obtained results, we per- ical writing for high-density
formed finite-element calculations of the strain data-storage application.
Mechanical Writing of Ferroelectric Polarization 61

during application of mechanical stress: As the 10-nm in radius, so switching can be highly lo- 17. Y. Gaillard, A. Hurtado Macías, J. Muñoz-
tip pressure is increased, the positive coercive calized, allowing fabrication of high-density do- Saldaña, M. Anglada, G. Trápaga, J. Phys.
bias decreases, whereas the negative coercive main patterns. Because no voltage is applied D Appl. Phys. 42, 085502 (2009).
bias remains intact (Figure S3) (18). during mechanical switching, leakage and/or 18. Supplementary materials follow the Notes
section.
For the loading force of 1000 nN, the calcu- dielectric breakdown problems are minimized
19. W. Ma, Phys. Status Solidi B 245, 761 (2008).
lated flexoelectric field reaches a maximum of [in fact, even insulating tips can be used to me- 20. R. Maranganti, P. Sharma, Phys. Rev. B 80,
2 MV/cm, comparable with the intrinsic coer- chanically write the domains, as shown in Fig- 054109 (2009).
cive field (theoretical) and that extracted from ure S4 (18)]. Because electrodes are not re- 21. W. Ma, L. E. Cross, Appl. Phys. Lett. 88,
the piezoelectric hysteresis loops (experimen- quired, the problems caused by their finite 232902 (2006).
tal). Note that such a large flexoelectric field screening length (25) are also removed. 22. J. Hong, G. Catalan, J. F. Scott, E. Artacho,
has been obtained in spite of our conservative Conversely, if top electrodes were used, J. Phys. Condens. Matter 22, 112201 (2010).
choice of flexoelectric coefficients, support- mechanical writing would enable the targeted 23. A. Yu. Emelyanov, N. A. Pertsev, A. L.
Kholkin, Phys. Rev. B 66, 214108 (2002).
ing the feasibility of the flexoelectric switching poling of localized areas under the electrodes—
24. P. Vettiger et al., IEEE Trans. NanoTechnol.
mechanism. We stress that the Landau formal- which is impossible using voltage, as the electric 1, 39 (2002).
ism provides an upper limit for the ideal (in- field is homogeneous in a parallel-plate capac- 25. V. Nagarajan et al., J. Appl. Phys. 100,
trinsic) switching barrier; in practice, there will itor. This suggests the possibility of controlled 051609 (2006).
be defects that act as nucleation sites facilitating fabrication of domain walls underneath top 26. G. Catalan, J. Seidel, R. Ramesh, J. F. Scott,
the switching at lower coercive fields. In real electrodes, useful for electronic device applica- Rev. Mod. Phys. 84, 119 (2012).
devices, therefore, the effective coercive field tions employing physical properties of domain 27. D. A. Bonnell, S. V. Kalinin, A. Kholkin, A.
may be considerably lower than calculated here, walls (26) that could be read in a nondestruc- Gruverman, MRS Bull. 34, 648 (2009).
28. V. Garcia et al., Nature 460, 81 (2009).
meaning that the flexoelectric field will be ca- tive manner by PFM imaging (27) or by mea-
pable of inducing switching at lower loads or suring the electroresistive effect (28). Acknowledgments — A.G and G.C. conceived
larger thicknesses than assumed here. the idea, designed the experiment, and wrote
There are several useful features of mechan- References and Notes the paper; H.L. implemented experimental
ical switching: (i) It generates stable domain measurements; C.-W.B. fabricated the sam-
patterns exhibiting no relaxation for days after 1. K. J. Choi et al., Science 306, 1005 (2004). ples; C.-B.E. supervised sample preparation
2. G. Catalan et al., Phys. Rev. Lett. 96, 127602 and reviewed the paper; and D.E.O., J.A., and
switching, (ii) mechanically written domain pat- (2006).
terns are electrically erasable, (iii) no damage to G.C. performed finite-element and free-energy
3. G. Zavala, J. H. Fendler, S. Trolier-McKinstry, calculations. G.C. and A.G. thank the Lever-
the sample surface caused by a high loading force J. Appl. Phys. 81, 7480 (1997). hulme Trust for international network funding
was observed, and (iv) the mechanically written 4. A. Kholkin et al., Appl. Phys. Lett. 82, 2127 (F/00 203/V) for the funds that have enabled
domains are nanoscopic. These features are illus- (2003). this collaboration. G.C. acknowledges finan-
trated in Figure 3. Mechanically written parallel 5. A. Gruverman, A. Kholkin, A. Kingon, H. To- cial support from grants MAT2010-10067-E and
linear domains, shown in Figure 3A, have been kumoto, Appl. Phys. Lett. 78, 2751 (2001). MAT2010-17771, and J.A. acknowledges sup-
subsequently transformed into the pattern in Fig- 6. D. Lee et al., Phys. Rev. Lett. 107, 057602 port from grant MAT2011-23375 (Ministerio de
(2011). Educación, Ciencia e Innovación). The work at
ure 3B by electrically erasing central domain seg- 7. A. Gruverman et al., Appl. Phys. Lett. 83, 728
ments with a tip under a dc –3-V bias. The top- University of Nebraska-Lincoln was supported
(2003). by the Materials Research Science and Engi-
ographic image of the same area of the BaTiO3 8. S. M. Kogan, Sov. Phys. Solid State 5, 2069 neering Center (NSF grant DMR-0820521) and
film (Figure 3C), acquired after this procedure, (1964). by the U.S. Department of Energy (DOE), Of-
does not exhibit any traces of surface deforma- 9. J. F. Scott, J. Chem. Phys. 48, 874 (1968). fice of Basic Energy Sciences, Division of Ma-
tion. Finally, Figure 3D shows an array of dot do- 10. G. Catalan, L. J. Sinnamon, J. M. Gregg, J. terials Sciences and Engineering (DOE grant
mains only 30 nm in size, written by abruptly al- Phys. Condens. Matter 16, 2253 (2004). DE-SC0004876). The work at Univ. of Wiscon-
ternating the tip load between 30 and 1500 nN 11. G. Catalan et al., Nat. Mater. 10, 963 (2011). sin–Madison was supported by the NSF under
12. M. S. Majdoub, P. Sharma, T. Çağin, Phys. grant ECCS-0708759.
during scanning. Rev. B 78, 121407 (2008).
These results open up a way to write ferro- 13. C. B. Eom et al., Science 258, 1766 (1992).
electric memory bits using mechanical force in- 14. L. J. Belenky, X. Ke, M. Rzchowski, C. B.
Supplementary Materials, including Ma-
stead of electrical bias in data-storage devices. Eom, J. Appl. Phys. 97, 10J107 (2005). terials and Methods, Supplementary Text,
By converting mechanical stress into readable 15. In our studies, the flexoelectric switching Figures S1 to S4, and References, follow
information, such devices would operate as a has been demonstrated in films with thick-
nanoscopic analog of typewriters that could be ness of up to 48 unit cells.
scaled up using a millipede-like scheme (24). 16. D. D. Fong et al., Phys. Rev. Lett. 96, 127601
(2006).
The tip-sample contact area is typically less than
 www.sciencemag.org/cgi/content/full/336/6077/59/DC1

Supplementary Materials for

Mechanical Writing of Ferroelectric Polarization
H. Lu, C.-W. Bark, D. Esque de los Ojos, J. Alcala, C. B. Eom, G. Catalan,* A.
Gruverman1*

*To whom correspondence should be addressed. E-mail: gustau.catalan@cin2.es (G.C.);

agruverman2@unl.edu (A.G.)

Published 6 April 2012, Science 336, 59 (2012)

DOI: 10.1126/science.1218693

This PDF file includes:

Materials and Methods

Supplementary Text
Figs. S1 to S4
References
Materials and Methods

Single-crystalline epitaxial BaTiO3 heterostructures have been grown by pulsed

laser deposition (PLD) on atomically smooth (001) SrTiO3 substrates with 30-nm-thick

La0.67Sr0.33MnO3 electrode. Reflection high-energy electron diffraction (RHEED) has been

used for in-situ monitoring of the layer-by-layer growth process. Before deposition, low

angle miscut (<0.1°) SrTiO3 substrates were etched using buffered HF acid for 90 seconds

to maintain Ti-termination and then were annealed in oxygen at 1000°C for 12 hours to

create atomically smooth surfaces with single-unit-cell-height steps. During deposition of

all the layers substrate temperature was maintained at 680 °C with chamber oxygen

pressure kept at 150 mTorr. The samples were annealed at growth temperature and 1 atm

oxygen pressure for 30 minutes and then cooled down to room temperature.

Polarization imaging and local switching spectroscopy has been performed using a

resonant-enhanced Piezoresponse Force Microscopy (MFP-3D, Asylum Research).

Conductive silicon cantilevers (PPP-EFM, Nanosensors) have been used in this study. PFM

hysteresis loops were obtained at fixed locations on the BaTiO3 electrodes as a function of

switching pulse amplitude (pulse duration was 25 ms) superimposed on ac modulation bias

with amplitude of 0.8Vp-p at 320 kHz. Tip contact forces have been calibrated by measuring

force-distance curves.

The finite element simulations were performed for an AFM tip in contact with a

BaTiO3 film of 5 nm thickness attached to a semi-infinite SrTiO3 substrate. The full

anisotropy of the contact behavior was taken into account through the following

coefficients for the stiffness matrices: (1) for BaTiO3, C11 = 358.1 GPa, C12 = 115.2 GPa

and C14 = 149.8 GPa; (2) for the SrTiO3 substrate, C11= 421 GPa, C12 = 122.1 GPa, C14 =

2
133.2 GPa [i]. As compared to an ideal spherical shape, a blunt punch configuration better

reproduces the nanoscale surface irregularities present in a real AFM tip.

Supplementary Text

I. Piezoresponse signal variation due to electrically-induced polarization reversal

Figure S1 shows conventional PFM images of the BaTiO3 film where an area in the

center has been scanned with a tip under an incrementally increasing electrical dc bias

before PFM imaging. The PFM amplitude contrast in Fig S1(A) gradually changes with the

applied electric dc bias. Specifically, initially the PFM amplitude signal decreases and then,

after reaching a minimum value at the applied bias of 1.25 V, it starts to increase (Fig

S1(C)). This change in PFM amplitude is analogous to what was observed in the

measurements as a function of mechanical stress (Figs 1D and 1F of the main text). A

corresponding change in the PFM phase image (Fig S1(B)) shows contrast inversion at the

same dc bias at which the PFM amplitude reaches its minimum, also in complete analogy

with the mechanically induced phase change in Fig 1E. In both cases, the PFM response

can be interpreted in terms of nucleation and growth of 180o domains with polarization

antiparallel to the initial state, whose volume fraction grows under increasing dc bias. A

minimum in the PFM amplitude corresponds to the situation at which there are equal

fractions of upward and downward polarization, so that the piezoelectric responses cancel

each-other.

The results show that there is no fundamental difference in the switching response

induced by conventional electric biasing (Fig S1) and that induced by mechanical loading

(Figs 1D and 1E of the main text), indicating that flexoelectricity is an effective substitute

for voltage in ferroelectric switching.

3
II. Flexoelectric calculations

The flexoelectric field is given by [ii, iii]:

f ijkl ∂ekl
Eh = (1)
εih ∂x j

where fijkl is the flexoelectric tensor, εih the dielectric constant, ejk the strain and xl are the

spatial coordinates of the film. The vertical component of the tip-induced flexoelectric field

is therefore:

f 3311 ∂e11 f 3322 ∂e22 f1313 ∂e13 f 2323 ∂e23 f 3333 ∂e33
E3 = + + + + (2)
ε 33 ∂x3 ε 33 ∂x3 ε 33 ∂x1 ε 33 ∂x2 ε 33 ∂x3

The dielectric constant is ε 3 = ε 0ε r , where ε 0 is the permittivity of vacuum and

εr=30 for compressively strained BaTiO3 on SrTiO3. Though the shear flexoelectric

coefficients (f1313 and f2323) are comparable to the transverse ones (f3311 and f3322) [iv], the

finite element calculations (Fig S2) show that the shear strain gradients are an order of

magnitude smaller than the transverse ones, so we can neglect the shear terms. Conversely,

though the longitudinal flexoelectric coefficient (f3333) is an order of magnitude smaller than

the others [iv], the longitudinal deformation is an order of magnitude larger than the

transverse one, so it cannot be neglected. Thus, the flexoelectric field under the tip is:

f 3311 ∂e11 f 3322 ∂e22 f 3333 ∂e33

E3 = + + (3)
ε 33 ∂x3 ε 33 ∂x3 ε 33 ∂x3

The strain fields e11, e22 and e33 have been calculated by finite element methods

using parameters listed below in the Materials and Methods section. Once the flexoelectric

field E3 is known, its effect is calculated by adding the term − E3 P to the standard Landau-

Devonshire free energy [v] of an epitaxial film under uniaxial compression.

4
 To visualize the results and facilitate comparison with experiment, it is useful to

introduce a concept of a “flexoelectric potential”, V, whose gradient is the flexoelectric

∂V
field: E3 ≡ . Comparing this with Eq. 3, we obtain the flexoelectric potential as:
∂x3

f 3311 f 3322 f 3333

V = e11 + e22 + e33 (4)
ε 33 ε 33 ε 33

The flexoelectric potential under the tip, measured in Volts, is shown in Figure 2b of the

article. Note that though Eq. 4 would imply that a homogeneous deformation also causes a

non-zero voltage, this would however be constant across the film and would therefore have

no physical consequence. Only potential differences matter, i.e. there is only a field if there

is a potential gradient, which is the case when the induced deformation ininhomogeneous.

III. Stress and piezoelectricity

The mechanical boundary conditions are asymmetric in our film, as there is a large

stress concentration at the top surface and virtually no stress at the bottom surface.

Standard piezoelectric coupling will therefore lead to a polarization gradient directly

proportional to the strain gradient, and therefore to a depolarization field that, if

sufficiently large and unscreened may conceivably be able to cause switching. However,

this mechanism cannot be invoked when the local depolarizing field is compensated, as is

the case in our experiments, where we have used grounded metallic tips and/or metallic

electrodes. More generally, piezoelectric switching can also be discarded because the

piezoelectric effect should lead to the back-switching of domains of both polarization

directions, up and down, whereas our experimental data show the switching only of the

5
domains oriented up but not the opposite. This is incompatible with the piezoelectric

mechanism and fully consistent with the flexoelectric mechanism.

Another, more relevant effect of compressive vertical stress is related to decrease of

tetragonality of the film and thus its coercivity. This can be seen in Fig 2C of the main

article, showing that the energy barrier between the two potential wells is lower when the

film is under uniaxial compression. This lower barrier means that smaller biases are

sufficient to switch it. Figure S3 shows several PFM hysteresis loops acquired in the

BaTiO3 film at different tip-applied loading forces. It can be seen that, under increasing

load, the hysteresis loops become both narrower, signaling reduced coercivity, and

increasingly asymmetric, due to the flexoelectric bias. We can quantify this:

(1) We define the coercivity as the half-width of the hysteresis loop, measured as

the difference between positive and negative coercive voltages divided by two: Vc=(Vc+-

Vc-)/2. The coercivities for the films under increasing vertical force of 50nN, 250nN,

500nN and 1000nN are, respectively, 1.6V, 1.3V, 1.0V, 0.6V. This monotonical decrease

in coercivity reflects the decrease in the average tetragonality of the film due to the vertical

compression.

(2) We define the offset bias as the shift of the centre of the hysteresis loop, which

is half the sum of positive and negative coercive voltages: Voff=(Vc++Vc-)/2. The offset for

the films under increasing force is 0V, -0.1V, -0.4V, -0.7V. There is a monotonical shift of

the hysteresis loops towards negative values as the pressure on the tip is increased,

consistent with an increasing flexoelectric bias.

Note also that the positive coercive field is reduced while the negative coercive

field seemingly does not. This is a direct consequence of the simultaneous narrowing and

biasing of the loops. As the hysteresis loops get narrower, both the negative and positive

6
coercive fields move towards zero. For the negative coercive field, the negative shift

caused by the bias is compensated by the positive shift caused by the reduction in bias,

with the net effect being that the negative coercive field barely moves. Meanwhile, for the

positive coercive field, the negative bias and reduction in coercivity both act in the same

direction and thus the positive side of the loop rapidly shifts left. At a sufficiently high load

(>1000 nN), the positive coercive field is shifted to the negative voltage range, indicating

that without external voltage only the downward polarization state is stable.

IV. Mechanical switching with an insulating tip

Figure S4 illustrates that mechanical switching is possible even with an insulating

tip (PFM readout of course still requires a conducting tip). Mechanical switching was

performed with a non-conductive silicon tip by applying a force of 1400 nN during

scanning. The resulting polarization pattern was then visualized in a conventional PFM

mode using a metal-coated tip. These results demonstrate that the mechanical switching is

impervious to electrical boundary conditions (we have now done it both with grounded

conducting tips and with insulating tips). The demonstration of switching with

mechanically tougher non-metallic tips (Si or diamond) that are not prone to mechanical

wear and accompanying electrical degradation as metal-coated tips is helpful in the

implementation in more robust devices.

7
Figure S1. Electrically-induced reversal of ferroelectric polarization in the BaTiO3
film. (A) PFM amplitude and (B) phase images of the film acquired after the central
area has been scanned with the tip under an incrementally increasing dc bias (image
size is 3x3 µm2). The dc bias was increasing in the bottom-up direction (marked by a
black arrow in (A)) from 0 V to 2.3 V. (C) PFM amplitude as a function of a dc bias
obtained by cross-section analysis of the poled area in (A).

8
Figure S2. Finite element calculations of the deformations fields along the (A) x

(e11), (B) y (e22), and (C) z (e33) directions under pressure from the AFM tip. (D)

Finite element calculation map of the shear strain (e13=e23). All the strain

components show gradients, with the shear strain gradient being the smallest one.

9
Figure S3. PFM hysteresis loops acquired at different loading forces. Note a shift

of the loops towards the negative bias upon the load increase.

10
 Figure S4. PFM amplitude (A) and phase (B) images of the domain pattern

generated by mechanical switching using a non-conductive Si tip under a 1400 nN

load and visualized in a conventional PFM mode using a PtIr-coated Si tip.

i
Piskunov, S., Heifets, E., Eglitis, R.I., Borstel, Bulk properties and electronic structure of

SrTiO3, BaTiO3, PbTiO3 perovskites: an ab inition HF/DFT study. Comp. Mater. Sci. 29,

165 (2004).
ii
Ma, W. A study of flexoelectric coupling associated internal electric field and stress in

thin film ferroelectrics. Phys. Stat. Sol. (b) 245, 761-768 (2008).
iii
Lee, D. et al. Giant Flexoelectricity Effect in Ferroelectric Epitaxial Thin Films. Phys.

Rev. Lett. 107, 057602 (2011).

11
iv
Maranganti, R. and Sharma, P. Atomistic determination of flexoelectric properties of

crystalline dielectrics. Phys. Rev. B 80, 054109 (2009).

v
Yu. Emelyanov, A., Pertsev, N. A., and Kholkin, A. L., Effect of external stress on

ferroelectricity in epitaxial thin films, Phys. Rev. B 66, 214108 (2002).

12