Reuse of AIP Publishing content is subject to the terms at: <a href="https://publishing.aip.org/authors/rights-and-permissions">https://publishing.aip.org/authors/rights-and-permissions</a>. Downloaded to: 192.249.3.188 on 30 November 2018, At: 12:30

Local probing of the interaction between intrinsic defects and ferroelectric domainwalls in lithium niobateGreg Stone, Donghwa Lee, Haixuan Xu, Simon R. Phillpot, and Volkmar Dierolf

Citation: Appl. Phys. Lett. 102, 042905 (2013); doi: 10.1063/1.4789779View online: https://doi.org/10.1063/1.4789779View Table of Contents: http://aip.scitation.org/toc/apl/102/4Published by the American Institute of Physics

Articles you may be interested inMicrostructure and defects probed by Raman spectroscopy in lithium niobate crystals and devicesApplied Physics Reviews 2, 040602 (2015); 10.1063/1.4934203

Growth, defect structure, and THz application of stoichiometric lithium niobateApplied Physics Reviews 2, 040601 (2015); 10.1063/1.4929917

Periodic domain inversion in x-cut single-crystal lithium niobate thin filmApplied Physics Letters 108, 152902 (2016); 10.1063/1.4946010

Influence of the artificial surface dielectric layer on domain patterning by ion beam in MgO-doped lithium niobatesingle crystalsApplied Physics Letters 110, 082903 (2017); 10.1063/1.4977043

Interactions of defect complexes and domain walls in CuO-doped ferroelectric (K,Na)NbO3Applied Physics Letters 102, 242908 (2013); 10.1063/1.4811268

Formation of dendrite domain structures in stoichiometric lithium niobate at elevated temperaturesJournal of Applied Physics 112, 104113 (2012); 10.1063/1.4766308

Local probing of the interaction between intrinsic defects and ferroelectricdomain walls in lithium niobate

Greg Stone,1,a) Donghwa Lee,2 Haixuan Xu,3 Simon R. Phillpot,4 and Volkmar Dierolf1,b)

1Department of Physics, Lehigh University, Bethlehem, Pennsylvania 18015, USA2Condensed Matter and Materials Division, Lawrence Livermore National Laboratory, Livermore,California 94550, USA3Center for Defect Physics, Oak Ridge National Laboratory, Oak Ridge, Tennessee 37831, USA4Department of Materials science and Engineering, University of Florida, Gainesville, Florida 32611, USA

(Received 18 December 2012; accepted 15 January 2013; published online 29 January 2013)

We demonstrate the capability of confocal Raman spectroscopy to characterize nanoscale

interactions of defects with ferroelectric domain walls by identifying defect-related frequency shifts

in congruent lithium niobate. These shifts resemble those observed for an external field applied

anti-parallel to the ferroelectric axis, suggesting a small reduction of the electric polarization.

Density functional theory calculations suggest that this reduction results from a change in the

intrinsic defect cluster structure and polarization at the domain wall. VC 2013 American Institute ofPhysics. [http://dx.doi.org/10.1063/1.4789779]

The ferroelectric properties of lithium niobate are

widely exploited in nonlinear frequency conversion devices

and a wide variety of other technologies. In such devices, the

orientation of the ferroelectric polarization is manipulated,

which produces domain walls that delineate domains with

opposite polarization. These devices commonly use the lith-

ium deficient congruent composition, which displays signifi-

cant differences in nucleation, growth, shape, and smallest

stable size of ferroelectric domains compared to defect free

stoichiometric lithium niobate. These observations point to a

strong interaction between the intrinsic defects associated

with this off-stoichiometry and the ferroelectric domain

walls; the nature of this interaction is not well understood.

Moreover, the characterization of such interactions experi-

mentally poses a significant challenge as they occur on a

nanometer length scale, magnifying the difficulties that al-

ready exist in probing the detailed structure of domain walls

and of defects individually. In this letter, we demonstrate the

successful use of confocal Raman spectroscopy as a sensitive

high-resolution probe of electric fields in ferroelectric mate-

rials and show that the presence of defects at the domain

wall reduces the electric polarization. Correlating this finding

with results from density functional theory (DFT), we char-

acterize a distinct, mutual accommodation effect between

domain walls and defects.

Isolated domain walls and their impact on the surround-

ing regions have been studied by a wide range of methods

both experimentally1–3 and theoretically.4–6 While it is

widely accepted that the polarization reversal at the domain

wall between two anti-parallel domain states can take place

over only a few lattice sites, defects and strain have been

shown to extend the perturbation over several microns,

which is most notable in samples that have not been

annealed after domain inversion by electric field poling.7

Several models have been proposed regarding the

composition and structure of defect clusters that accommodate

lithium deficiency in congruent lithium niobate. DFT calcula-

tions,8 x-ray diffraction, and neutron scattering9,10 favor defect

complexes that consist of a niobium antisite that is charge

compensated by four lithium vacancies. Nuclear magnetic res-

onance studies have shown three of these lithium vacancies

are in the nearest neighborhood position next to the niobium

antisite while the fourth lithium vacancy lies on one of the cat-

ion sites along the ferroelectric axis.11 Configurations involv-

ing different locations of this fourth vacancy have similar

formation energies but have different impacts on the local ion

displacements and different contribution to the net polariza-

tion of the crystal.12 Therefore, it is expected that a thermally

equilibrated system will have an ensemble of such defect

complexes, with concentrations according to their thermody-

namic weights.

To investigate the mutual interaction effect of defects

with a domain wall experimentally, we utilized Raman spec-

troscopy, which is sensitive to both changes in the crystal

structure and internal fields in ferroelectric materials, such as

lithium niobate, through the converse piezoelectric effect.

The Raman measurements were performed using a custom

built confocal Raman microscope incorporating a 1.32 nu-

merical aperture oil immersion objective providing a lateral

spatial resolution as low as 250 nm. Raman spectra were col-

lected by a spectrometer equipped with a liquid nitrogen

cooled CCD array, while the domain wall region within the

sample was scanned through the laser focus using a XYZ

nanopositioning stage. The setup incorporated two single

mode fibers to act as the pinholes of a confocal microscope.

The same objective was used to focus a 488 nm laser beam

onto the sample and to collect the Raman scattering from the

sample in the backscattering collection geometry. For the

results reported here, the polarization of the laser light and

the collected Raman scattering were chosen to be parallel to

each other. A dichroic mirror and Raman long pass filter

removed any back scattered laser light, while a linear polar-

izer selected the orientation of the Raman scattering imaged

a)Current address: Department of Materials Science and Engineering, Penn-

sylvania State University, University Park, Pennsylvania 16802, USA.b)Author to whom correspondence should be addressed. Electronic mail:

vod2@lehigh.edu.

0003-6951/2013/102(4)/042905/4/$30.00 VC 2013 American Institute of Physics102, 042905-1

APPLIED PHYSICS LETTERS 102, 042905 (2013)

by the detector. The samples were oriented in the microscope

such that the laser propagated along the ferroelectric axis.

To identify and isolate the processes caused by the intrin-

sic defects, we compared the results obtained for congruent

lithium niobate with those obtained for near-stoichiometric

lithium niobate and stoichiometric lithium tantalate. The later

sample was included because it is possible to produce virtu-

ally defect free lithium tantalate by the vapor transport equili-

brated (VTE) process, while some defects remain in

nominally stoichiometric lithium niobate. After the domain

inversion process, all samples were annealed at 200 �C for

several hours to alleviate any strain and redistribute any poling

induced charges, which influence the crystal properties near

the domain wall.13–15 Additionally, annealing the sample

results in a reconfiguration of the intrinsic defects in the do-

main inverted region back to their thermal equilibrium distri-

bution. Consistent with the recovery of the thermodynamic

equilibrium of the defects after annealing, we find that the vir-

gin and domain inverted states are indistinguishable both in an

optical microscope and from their Raman spectrum.

A comparison of the Raman spectra shows several key

differences between the bulk domain and domain wall region,

as seen in Figure 1. Some of these differences include inten-

sity changes in spectra regions near several Raman modes,

which are highlighted by the solid arrows in Figure 1(a) and

are consistent for all of the samples with different defect con-

centrations. They also spectrally overlap with the intensity

changes reported due a modification of the Raman scattering

selection rules by the presence of a domain wall.16 Therefore,

we do not attribute the intensity changes present at the domain

wall to the presence of intrinsic defects. More importantly, we

find additional changes in the Raman spectrum that corre-

spond to frequency shifts of several modes highlighted in Fig-

ure 1 by the dotted arrows. We attribute these shifts to the

presence of intrinsic defects since they are most pronounced

in the congruent lithium niobate material, very small for the

near-stoichiometric lithium niobate, and completely absent for

intrinsic defect free stoichiometric lithium tantalate, which is

shown in Figures 1(b)–1(e) for two selected modes.

To isolate and quantify the defect-related frequency

shifts, we performed a multi-peak analysis of Raman spectra

taken from the domain wall and bulk domain. The multi-

peak analysis shows that at the domain wall, the

EðTOÞ1; EðTOÞ3, and EðTOÞ8 modes shift to lower energy,

the AðLOÞ2 mode shifts to higher energy, and the AðLOÞ4mode experiences no measurable shift as seen in Table I.

While the remaining Raman modes may also experience a

frequency shift at the domain wall, their overlap with the

above-mentioned defect independent effect makes an unam-

biguous determination of their shifts impossible.

In order to elucidate the nature of the defect reconfigura-

tion at the domain wall, we compare the defect-related fre-

quency shifts at the domain wall with those shifts caused by

the presence of frustrated misaligned defects present after

the forward poling process17,18 and those introduced by an

applied external electric field.18 As shown in Table II, the

corresponding directions of the frequency shifts for the

EðTOÞ1; EðTOÞ3; AðLOÞ2, and EðTOÞ8 modes at the domain

wall match the shifts observed in the bulk crystal which has

an electric field applied anti-parallel to the ferroelectric axis.

The direction of the shifts in these modes is also identical to

the shifts due to the introduction of frustrated defects by

FIG. 1. (a) Comparison of the Raman spectrum from the domain wall (red

solid line) and the bulk (blue dotted line) in congruent lithium niobate with

the difference spectra between the two magnified 5 times (black solid line).

The frequency shift of the EðTOÞ3 mode in congruent and near-

stoichiometric lithium niobate are highlighted in (b) and (c) respectively,

while the frequency shift of the EðTOÞ8 mode in congruent and near-

stoichiometric lithium niobate are highlighted in (d) and (e), respectively.

The dotted arrows indicate the changes in the spectrum due to frequency

shifts.

TABLE I. Frequency shift of several Raman modes at the domain wall in

congruent and near-stoichiometric lithium niobate with respect to the bulk

domain determined by a multi-peak curve fitting function in cm�1.

Raman mode Congruent Near-stoichiometric

EðTOÞ1 0.28 6 0.12 0.04 6 0.03

EðTOÞ3 0.29 6 0.21 0.10 6 0.05

AðLOÞ2 �0.30 6 0.19 �0.02 6 0.07

EðTOÞ8 0.44 6 0.10 �0.02 6 0.08

AðLOÞ4 �0.06 6 0.10 �0.02 6 0.08

TABLE II. Frequency shifts of several Raman modes at the domain wall

and after forward poling in cm�1, along with their response to an applied

field parallel to the ferroelectric axis in cm�1=ðkVmm�1Þ.

Domain wall Domain inversion Applied field

Raman mode (Present) (Ref. 17) (Ref. 18) (Ref. 18)

EðTOÞ1 0.28 6 0.12 0.44 0.61 �0.024

EðTOÞ3 0.29 6 0.21 0.26 0.27 �0.011

AðLOÞ2 �0.30 6 0.19 �0.45 �0.33 0.010

EðTOÞ8 0.44 6 0.10 0.37 0.46 �0.027

AðLOÞ4 �0.06 6 0.10 0.67 0.78 �0.003

042905-2 Stone et al. Appl. Phys. Lett. 102, 042905 (2013)

forward poling. Unlike the previously discussed modes, the

AðLOÞ4 mode behaves differently for these two latter effects,

making a distinction possible. For applied external electric

fields along the ferroelectric axis, the AðLOÞ4 mode experi-

ences a very small frequency shift, while after forward

poling, the AðLOÞ4 mode has the largest frequency shift

among the Raman modes. At the domain wall, the AðLOÞ4mode has no noticeable shift, excluding the possibility that

frustrated defects play a role. From the measured response of

the Raman modes in lithium niobate due to an applied elec-

tric field along the ferroelectric axis,18 we find that the

Raman shifts in the EðTOÞ1; EðTOÞ3; AðLOÞ2; EðTOÞ8, and

AðLOÞ4 modes at the domain wall are identical to those

observed for a externally applied electric field along the fer-

roelectric axis of �12 6 5, �26 6 19, �30 6 19, �16 6 4,

and 20 6 33 kV/mm, respectively. Analyzing these values,

we find an effective field of �12 to �13 kV/mm that is con-

sistent within error for all the observed shift of the different

Raman modes. This indicates that the presence of defects at

the domain wall lead to a reduction of the effective internal

electric field along the ferroelectric axis at the domain wall.

These results also indicate that the defect complexes near the

domain wall can reconfigure such that the annealing process

allows the frustrated defects near the domain wall to reach

thermal equilibrium. This equilibrium may, however, be

different from that achieved in the bulk in terms of defect

configuration and relative concentrations of the different

possible configurations.

In Figure 2, we show the spatial extent of the defect-

related shift (at �565 cm�1), which is identical on either side

of the domain wall. A comparison with the spatial profile of

the effect caused by the change in the selection rules (at

�615 cm�1) reveals that the profile of the defect-related shift

is centered on the domain wall. The widths of the spatial pro-

files reflect the resolution limit of our instrument. These

results further confirm that the crystal is in a thermal equilib-

rium state on either side of the domain wall. In other words,

the observed effect is independent of the domain state, as-

grown or domain inverted, and represents the as-grown state.

In further interpreting these results, one needs to con-

sider the relationship between effective internal electric

fields, local ion displacements, and ferroelectric polarization.

With this in mind, we interpret our results as a reduction of

the ferroelectric polarization at the domain wall compared to

the polarization in the bulk that results from the presence of

intrinsic defects. In order to confirm this interpretation, we

employed DFT calculations19,20 to understand how defect

clusters and a domain wall interact. DFT calculations were

performed using the generalized gradient approximation

(GGA) for exchange and correlation potential.21 A

(4� 2� 2) supercell of the non-primitive hexagonal unit cell

containing 480 atoms and Brillouin Zone integration with a

periodic boundary condition and C point sampling was used

in the calculation. The pseudopotentials and methodologies

used here are the same as those used for a previous study.22

Figures 3(a)–3(d) show four structural arrangements of

intrinsic defect cluster configurations that are energetically

favorable in the bulk and can account for the lithium defi-

ciency in these materials. The first three nearest neighbor

arrangements of the defect clusters are shown and denoted as

1NN, 2NN, and 3NN, see Figure 3(e). Strain caused by

defect clusters leads to complicated displacements of the

neighboring ions that can be best quantified by the resulting

change in the polarization show by the variation in polariza-

tion caused by the different defect configurations in Figure 4.

For each configuration, the variation in the polarization is

very similar to that calculated for the bulk except for the first

nearest neighbor (1NN) configuration, where there is a

FIG. 3. Schematic view of four different defect

clusters which can minimize ionic displacement of

non-uniaxial components ((a)-(d), left) and (e) near-

est neighbor arrangements of defect cluster in the

vicinity of the domain wall. The distance between

defect cluster and domain wall is determined by rel-

ative position of cluster center with respect to the

domain wall: 1NN, 2NN, 3NN.

FIG. 2. Spatial profile across the domain wall for the frequency shift of the

EðTOÞ8 Raman mode taken at �565 cm�1 (red solid line) and the spectral

region with a pronounced directional dispersion region connecting the

EðTOÞ8 and AðLOÞ4 modes taken at �615 cm�1 (blue dotted Line). While

the first effect is due to defects, the second is due to a change in the Raman

selection rules.16

042905-3 Stone et al. Appl. Phys. Lett. 102, 042905 (2013)

decrease in polarization for all four defect configurations.

For 1NN, the defect clusters actively participate in accom-

modating the rapid polarization change resulting in a defect/

domain wall interaction effect.

As seen in Figure 5, the DFT calculations also show that

the presence of the domain wall decreases the defect forma-

tion energy (DFE) of all four different defect cluster configu-

rations. This decrease near the domain wall is more gradual

compared to the polarization change and suggests that the

defect clusters are more stable close to the domain wall. For

the annealing temperatures used, the ionic mobility is limited

and a significant accumulation of clusters at the domain wall

is therefore not expected. However, even small changes in

defect concentrations can contribute to frequency shifts in the

Raman modes. More importantly, the relative formation ener-

gies of the defect clusters are modified. Therefore, we expect

that the annealing process changes the relative concentration

of the different defect configurations at the domain wall in

comparison to the bulk. In the resulting new thermal equilib-

rium, defect configuration (a) is most favorable, in contrast to

the bulk where configuration (b) is most favorable. The

smaller variation of polarization associated with configuration

(a) enhances the described frequency shift.

The DFT calculations demonstrate that the effect of a sin-

gle defect close to the domain wall on the polarization is most

pronounced within a 0.5 nm region from the domain wall.

Hence, by itself this single-defect polarization change would

not be detectable in confocal Raman studies because the do-

main wall occupies only a very small fraction of the imaged

volume. However, congruent lithium niobate contains a large

intrinsic defect concentration. It is likely that the defects near

the domain wall interact with other defects further away from

the domain wall, thereby amplifying the change in polariza-

tion and extending its range beyond a few nanometers result-

ing in the observed frequency shifts in the Raman spectrum.

In conclusion, we have identified characteristic shifts of

vibrational frequencies observed in Raman spectroscopy due

to the presence of intrinsic defects at ferroelectric domain

walls. These shifts resemble shifts observed for a negative

electric field along the ferroelectric axis. With the support of

DFT calculations, we attribute this effect to a change in the

defect cluster polarization and concentration ratio of the dif-

ferent intrinsic defect cluster configurations at the domain

wall. Combined these effects lead to a reduction or partial

cancelation of the ferroelectric polarization at the domain

wall compared to the bulk.

The authors gratefully acknowledge finical support by

the National Science Foundation through two Materials

World Network grants (Grant Nos. DMR-0602986 and

DMR-1008075). Computational work was performed under

the auspices of the U.S. Department of Energy at Lawrence

Livermore National Laboratory under Contract No. DE-

AC52-07NA27344.

1M. Foeth, A. Sfera, P. Stadelmann, and P. A. Buffat, J. Electron Microsc.

48, 717 (1999).2S. I. Bozhevolnyi, K. Pedersen, S. Torben, X. Zhang, and M. Belmonte,

Opt. Commun. 152, 221 (1998).3D. A. Scrymgeour and V. Gopalan, Phys. Rev. B 72, 024103 (2005).4J. Padilla, W. Zhong, and D. Vanderbilt, Phys. Rev. B 53, R5969 (1996).5B. Meyer and D. Vanderbilt, Phys. Rev. B 65, 104111 (2002).6H. Chaib, T. Otto, and L. M. Eng, Phys. Status Solidi B 233, 250 (2002).7T. Jach, S. Kim, V. Gopalan, S. Durbin, and D. Bright, Phys. Rev. B 69,

64113 (2004).8H. Xu, D. Lee, J. He, S. B. Sinnott, V. Gopalan, V. Dierolf, and S. R. Phil-

lpot, Phys. Rev. B 78, 174103 (2008).9N. Zotov, H. Boysen, F. Frey, T. Metzger, and E. Born, J. Phys. Chem.

Solids 55, 145 (1994).10N. Iyi, F. Kitamura, J. Izumi, K. Yamamoto, T. Hayashi, H. Asano, and

S. Kimura, J. Solid State Chem. 101, 340 (1992).11A. V. Yatsenko, E. N. Ivanova, and N. A. Sergeev, Physica B 240, 254

(1997).12H. Xu, D. Lee, S. B. Sinnott, V. Dierolf, V. Gopalan, and S. R. Phillpot,

J. Phys.: Condens. Matter 22, 135002 (2010).13Y. Kong, J. Xu, B. Li, S. Chen, Z. Huang, L. Zhang, S. Liu, W. Yan,

H. Liu, X. Xie, L. Shi, X. Li, and G. Zhang, Opt. Mater. 27, 471 (2004).14P. Capek, G. Stone, V. Dierolf, C. Althouse, and V. Gopalan, Phys. Status

Solidi C 4, 830 (2007).15P. S. Zelenovskiy, V. Y. Shur, P. Bourson, M. D. Fontana, D. K. Kuznet-

sov, and E. A. Mingaliev, Ferroelectrics 398, 34 (2010).16G. Stone and V. Dierolf, Opt. Lett. 37, 1032 (2012).17J. G. Scott, S. Mailis, C. L. Sones, and R. W. Eason, Appl. Phys. A 79,

691 (2004).18G. Stone, B. Knorr, V. Gopalan, and V. Dierolf, Phys. Rev. B 84, 134303

(2011).19P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964).20W. Kohn and L. J. Sham, Phys. Rev. 140, A1133 (1965).21J. P. Perdew, in Electronic Structure of Solids 91, edited by P. Ziesche and

H. Eschrig (Akademie Verlag, Berlin, 1991), p. 11.22D. Lee, H. Xu, V. Dierolf, V. Gopalan, and S. R. Phillpot, Phys. Rev. B

82, 014104 (2010).

FIG. 4. Change in polarization near the four defect configurations ((a)-(d)).

For each configuration, three different distances from the domain wall

(1NN, 2NN, and 3NN) and the bulk are considered.

FIG. 5. Defect formation energy for the four defect configurations. For each

configuration, three different distances from the domain wall (1NN, 2NN,

and 3NN) and the bulk are considered.

042905-4 Stone et al. Appl. Phys. Lett. 102, 042905 (2013)