Influence of The Ferroelectric Polarization On The Electronic Structure of Batio Thin Films

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Influence of the ferroelectric polarization on the electronic structure of BaTiO3 thin films

N. Barrett1,*, J. Rault1, I. Krug2, B. Vilquin3, G. Niu3, B. Gautier4, D. Albertini4 and O. Renault5


1
CEA, IRAMIS/LENSIS/SPCSI, Bâtiment 462, 91191 Gif sur Yvette cedex, France
2
Institut f. Festkörperforschung IFF-9, Forschungszentrum Jülich, 52425 Jülich, Germany
3
Institut of Nanoscience de Lyon (INL), 36, avenue Guy de Collongue, 69134 Ecully, France
4
Institut National des Sciences Appliquées de Lyon (INSA), 20, avenue Albert Einstein, 69621 Villeurbanne, France
5
CEA LETI/MINATEC, SPCIO, 17 avenue des Martyrs, 38054 Grenoble cedex, France

* corresponding author: [email protected]

Abstract: Micrometric domains of precise ferroelectric polarization have been written into a 20 nm thick epitaxial
thin film of BaTiO3(001) (BTO) on a Nb doped SrTiO3 (STO) substrate using PiezoForce Microscopy (PFM). The
domain dependent electronic structure has been studied using fully energy-filtered PhotoEmission Microscopy
(PEEM) and synchrotron radiation. Shifts, induced by ferroelectric polarization, of up to 300 meV are observed in the
work function of the sample. The surface is Ba-O terminated. Polarization-induced distortion of the electronic
structure is observed in the valence band and on the Ba 3d, Ti 2p and O 1s core levels of BTO. Polarization
dependent surface adsorption is observed. A simple electrostatic model based on net surface charge is not sufficient
to explain the observed modifications in the electronic levels.

Keywords: Ferroelectric, oxide surface, PEEM, photoelectron spectroscopy, synchrotron radiation

Introduction
Thin film ferroelectrics are an active field of
research due to their promising technological
applications in electronic devices such as ferroelectrics
memories. Their surfaces also show fascinating
properties such as catalysis enhancement or possible
magnetoelectric coupling with magnetic overlayers. In a

thin film the polarization state gives rise to a fixed

surface charge,   P.n , where P is the polarization
 Figure 1. Distortion for a Ba-O terminated unit cell
vector and n is the surface normal. [SNGR09] This
gives rise to a depolarizing field which reduces the FE X-ray photoemission microscopy (XPEEM)
polarization in the material. The surface state of thin conserves the spatial origin of the photo-emitted
films is therefore crucial in determining the electrical electrons [EWRB09]. Domains with different
properties of real thin film based devices. The surface ferroelectric polarization should show shifts in the
charge may be screened either by internal charge carriers positions of the electronic levels. Local photoelectron
and defects in the ferroelectric or externally by spectra from the different FE domains will contain this
adsorbates. Thus, surveying the polarization dependence information. XPEEM collects electrons with a small
of electronic structure is a fundamental step towards the inelastic mean free path, it is thus surface sensitive. It is
understanding of some very promising phenomena. non destructive, crucial if one wishes to carry out other
Epitaxially strained BaTiO3 (BTO) films show imaging, structural or electrical analyses of the same
interesting ferroelectric properties compared to those of sample. In a simple electrostatic model, the net surface
bulk BTO. The maximum in-plane strain can attain 1-2%. charge should modify the kinetic energy of the
Theory has predicted that the Curie temperature can photoelectrons.
attain 550°C. Ferroelectric (FE) BTO(001) is an ABO3
Experimental
perovskite oxide with a tetragonal structure. The
ferroelectric distortion lies along the c axis. For example, A 20 nm thin film of BTO has been grown
Fig 1 shows the expected distortion for a Ba-O epitaxially on an Nb-doped SrTiO3 (001) substrate by
terminated surface unit cell, with P- polarisation giving a Pulsed Laser Deposition (PLD). Thanks to niobium
net negative surface fixed charge, as indicated. doping, the substrate is conducting and works as an
electrode. It also prevents significant charging effects
during photoemission. Micron scale polarized
ferroelectric domains have been written by Piezo Force
Microscopy (PFM, an Atomic Force Microscope with a Results
conductive tip) in d.c. (writing) mode as shown on the Work function
Fig. 2 together with a typical phase signal as observed by
With photon energy of 575 eV, the threshold
PFM in the a.c. (reading) mode. The topography image
represents a true secondary electron (SE) peak and direct
is uniform. Thus the written region is flat but the phase
transitions play no part in the intensity position. Fig. 3
signal shows FE domains with a clear electrical contrast.
shows a selection of images from the total image series.
Contrast inversions between differently FE domains as a
function of E  E F are immediately visible. The
position of the threshold in the local spectra extracted
from the image series is equivalent to the work function.
After correction for the Schottky effect due to the high
extractor field, E = 98 meV for 12 kV, [RBRH06] the
position of the photoemission threshold corresponds
exactly to the work function. The threshold spectra were
fitted with a complementary error function
  E
I E   A  erfc 0   I min
 2 
Figure 2. (a) schematic of a PFM. (b) PFM phase image of the
BTO thin film
where Φ0 is the work function and σ the half-width of
The thin film deposition process includes an the rising side of the secondary electron peak. The work
oxygen plasma treatment at 620°C and a partial pressure function shifts up to 300 meV depending on the domain
of 3 Pa. After plasma treatment, the sample was cooled polarization strength.
in 4×104 Pa oxygen. It is thought that the film has
undergone a phase transition to a single ferroelectric
domain with a P- imprint at RT. This is confirmed by the
PFM writing which was easier for P+ followed by P-
poling, showing that the film has actually a P- imprint.
Two domains with strong negative polarization, called
P-- and P---, were written by PFM in the film.
The energy filtered XPEEM experiments have
been conducted on TEMPO beam line of the SOLEIL
synchrotron (Saint Aubin, France). The temporary end
station is a NanoESCA XPEEM (Omicron
Nanotechnology GmbH) [EWNZ05, RBBZ07]. This
aberration-compensated energy-filtered PEEM offers a
lateral resolution of ~ 100 nm for core level emission
with, simultaneously, an energy resolution of 0.15-0.2
eV [EWRB09]. Experiments were conducted in Ultra
High Vacuum (UHV) at a pressure of 2.10 -8 Pa. To avoid Figure 3. (upper) images from the threshold image series as a
a transition to the paraelectric phase, the sample was not function of (E – EF); (below) threshold spectra extracted from
regions (a), (b) and (c) and the complementary error function fits
heated after introduction into the XPEEM chamber.
to the rising side of the SE peak.
Unfortunately, some surface contamination is therefore
inevitable, and as we will see, may reduce the observed Valence band
FE contrast.
The local valence band spectra extracted from the
Image series were acquired over the photoemission three differently polarized regions are shown in Fig. 4.
threshold region (photon energy h = 575 eV), the Ba 3d, The Ba 5p emission is always centred at ~ 15 eV and
Ti 2p and O 1s core levels (h = 898, 575 and 647 eV does not change in intensity as a function of FE
respectively) and the valence band (h = 80 eV). The polarization. The weak component centred between 9.0
photon energy was tuned with respect to each core level and 9.5 eV increases steadily in intensity from zone (a)
so as to maintain the same surface sensitivity and to to (c). The energy position of the intensity maximum of
avoid Auger lines interfering with the core level spectra the valence band shifts slightly to higher binding
(photoelectron kinetic energy between 100-120 eV). A energies as the strength of the polarization increases,
64.5 m Field of View (FoV) was used. 12 kV however, the shift is well below the experimental
extraction voltage. Dark images were also acquired with resolution. A zoom on the leading edge of the valence
the beam shutter closed in order to remove camera noise. band is shown in the inset. A 200 meV shift between the
The non-isochromaticity of the XPEEM is smaller than position in zone (a) and that for (b) and (c) is observed.
60 meV. This translates into a shift in the valence band maximum
(VBM), as obtained from a linear extrapolation. A
non-zero density of states is always observed in the gap
region above the valence band maximum up to the Fermi
level (EF = 0 eV in all figures). Figure 5. (upper left) SE image of the FE domain structure;
(upper right) Ba 3d5/2 core level shifts; (lower) best fits to local
spectra extracted from zones a, b and c. Surface components in
dark grey, bulk in light grey.

There are three components in the O 1s spectrum,


two of which, at higher binding energies (HBE), are
again surface related. The first surface component
reflects surface, or near-surface oxygen in the BTO,
while the second is linked to adsorbed contaminants
(most probably, CO2). Fig. 6 shows the O 1s peak (h =
647 eV) for three different polarizations. The observed
shifts are similar to those of the Ba 3d5/2 core level and
the work function.

Figure 4. Valence band spectra for the three polarized regions


((a) blue; (b) red; (c) green). (Inset) enlargement of the valence
band leading edge. The thin black and grey lines are two
theoretical DOS calculations, see text.

Core-levels
A Shirley background was used. Each core level
component was fitted with a Voigt function, which in
fact turned out almost Gaussian. The full width half
maxima (FWHM) were 1.5, 1.2 and 1.4 eV for Ba, O
and Ti core levels, respectively.
Fig. 5 shows the Ba 3d5/2 peak for the three distinct
FE domains. There are two components, one from the
bulk-coordinated barium (light grey), and the other from
the surface barium (dark grey). The surface component Figure 6. (upper left) SE image of the FE domain structure for
is the strongest, confirming that the photoelectron kinetic (E – EF) = 4.10 eV; (upper right) O 1s core level shifts; (lower
energy has been set to minimize the inelastic mean free panels) best fits to local spectrum extracted from zones a, b and
path (~ 0.5 nm). There is some debate as to the exact c. Surface component are in dark grey, bulk in light grey and
origin of the surface peak. It may be due to Ba emission contamination in black.
from the Ba-O termination layer, but its strength
suggests that it could also be due to a space charge Fig. 7 shows the Ti 2p3/2 peak for the three
region affecting the measured Ba 3d5/2 binding energy domains. The core level shift as a function of
over several unit cells. [LLLM08] This will be the polarization is significantly smaller than for the Ba 3d5/2.
subject of a more in-depth study. Both surface and bulk This is also an indication that changes in the electronic
like components are shifted towards the Fermi level by levels will require an explanation which goes beyond the
150 meV (b) and 300 meV (a) with respect to the simple model of a uniform electric field induced by the
unwritten (c). polarization. The shifts of the Ti 2p3/2 with respect to the
position in zone c are 50 meV (b) and 150 meV (a).
Figure 7. Ti 2p spectrum. The central inset shows the threshold observe; the most positive FE domain gives the highest
image. On the left, the local Ti 2p3/2 spectra extracted from the adsorbate signal. A more complete description of
three zones a, b and c. adsorption dynamics should include the chamber
pressure and also the surface desorption frequently
Discussion observed when using high brilliance synchrotron
Work function radiation. The resultant effective electric field would be
A simple explanation for the work function shifts a balance between polarization-enhanced adsorption and
could be the following. For a P- polarization state, the radiation stimulated desorption.
fixed (negative) surface charge will act to reduce the
work function, defined as the vacuum level with respect Conclusion
to the Fermi level. Thus, the more negative the We have used energy filtered X-ray photoemission
polarization, the smaller the effective work function. electron microscopy to explore ferroelectric polarization
Positive polarization, with positive fixed surface charge dependent changes in the electronic structure of
would have the opposite effect. This electrostatic BTO(001) thin film. P- domains were poled using Piezo
interpretation is valid for the surface potential and Force Microscopy. Contrast is observed in the work
therefore the work function. Since there should be no function, core level binding energies and valence band
chemical change in the ferroelectric state, we could maxima positions. A simple electrostatic model
assume that the electron affinity remains the same, thus a assuming a uniform net electric field, predicts the
shift in the work function should be followed by all of direction of the observed energy shifts. However, the
the electronic levels. However, as we will see, this is not model of rigid displacement of all the electron bands is
the case. not correct. The work function shifts by 300 meV for the
Valence band most strongly poled domain, followed by the Ba 3d 5/2
core levels. Differences in the Ba and Ti core level
The intensity centred at 9 eV in the valence band shifts suggest that an interpretation including the level of
can be attributed to either H2O or CO2 adsorption ionicity/covalency is important. The valence band
[MSRM88], but not to the formation of a surface maximum of the most strongly poled domain shifts by
carbonate species which would, for example, shift the Ba 200 meV away from the Fermi level. Polarization
5p core levels. Indeed, Baniecki et al [BKY08] show
dependent screening by adsorbates of the fixed surface
convincing evidence based on DFT calculations for charge certainly plays a role in the amplitude of
molecular CO2 adsorption. There is clearly more depolarizing field.
adsorption on the more positively polarized domains.
The sign of the VBM shift follows the electrostatic Acknowledgements
argument given above. The absence of a Ba 5p core level We acknowledge SOLEIL for provision of
shift is proof that different electronic levels do not react synchrotron radiation facilities and we would like to
in the same way to the FE polarization. thank F. Sirotti, the TEMPO beamline staff and B.
Core levels Delomez for technical assistance.
The core levels shift systematically to higher References
binding energies with increasing polarization P -. The Ba [BMFB96] A.E. Bocquet, T. Mizokawa, A. Fujimori, S.R.
3d shift is twice that of the Ti 2p. The formal ionic Barman, K. Maiti, D.D. Sarma, Y. Tokura, M. Onoda,
charges of Ba and Ti are 2+ and 4+, respectively, but in Phys. Rev. B 1996, 53, 1161.
perovskite oxides there is always some degree of [BKY08] J.D Baniecki, M. Ishii, K. Kurihara, K. Yamanaka, T.
covalency. Several theoretical studies [BMFB96, SI94, Yano, K. Shinozaki, T. Imada, K. Nozaki, N. Kin, Phys.
EV08] have shown that the effective valence charge on Rev. B 2008, 78, 195415.
Ti is more strongly reduced than on Ba. The higher [CBLS04] K.J. Choi, M. Biegalski, Y.L. Li, A. Sharan, J.
ionicity of Ba may explain a bigger FE induced core Schubert, R. Uecker, P. Reiche, Y.B.Chen, X.Q. Pan, V.
level shift. However, the notion of valency must be Gopalan, L.Q. Chen, D.G. Schlom, C.B. Eom, Science
treated with care. Ghosez et al [GMG98] show that 2004, 306, 1005.
dynamical atomic charge provides a better description of [CDKG04] S.A. Chambers, T. Broubay, T.C. Kaspar, M.
variations in bonding hybridization and that charge Gutowski, M. van Schilfgaarde, Surf. Sci., 2004, 554, 81.
transfer cannot be treated as a purely local phenomena. [CKL00] C. Cheng, K. Kunc, and M.H. Lee, Phys. Rev. B,
The adsorbate origin of the HBE O 1s component 2000, 62, 10409.
is well documented on STO [MSRM88] and BaO [EV08] R. I. Eglitis and David Vanderbilt, Phys. Rev. B, 2008,
[BKY08]. The local O 1s spectra from our sample 77, 195408.
indicate that the most polarized zones are the least [EWRB09] M. Escher, K. Winkler, O. Renault, and N. Barrett,
J. Elec. Spectros. Relat. Phenom., 2009,
contaminated. It has been shown that both the adsorption
doi:10.1016/j.elspec.2009.01.001
rate and saturation depend on the strength of the surface
[EWNZ05] M. Escher, N. Weber, M. Merkel, C. Ziethen, P.
FE polarization [LZJK08], a higher P+ should increase
Bernhard, G. Schonhense, S. Schmidt, F. Forster, F.
adsorption, improving surface screening and thus Reinert, B. Kromker, D. Funnemann. J. Phys. Condens.
reducing the depolarization field. This is what we Mat., 2005, 17, S1329.
[GMG98] P. Ghosez, J.-P. Michenaud, X. Gonze Phys. Rev. B,
1998, 58 6224.
[LZJK08] D. Li, M. H. Zhao, J. Garra, A. M. Kolpak, A. M.
Rappe, D. A. Bonnell, and J. M. Vohs, Nat. Mat., 2008
7, 473.
[MSRM88] D. Mueller, A. Shih, E. Roman, T. Madey, R. Jurtz,
R. Stockbauer J. Vac. Sci. Technol. A, 1988, 6 1067.
[PISK03] S. Piskunov, PhD thesis, University of Osnabrück
2003.
[RBRH06] O. Renault, R. Brochier, A. Roule, P.H. Haumesser,
B. Krömker, D. Funnemann, Surf. Inter. Anal., 2006, 38,
375.
[RBBZ07] O. Renault, N. Barrett, A. Bailly, L.F. Zagonel, D.
Mariolle, J.C. Cezar, N.B. Brookes, K. Winkler, B.
Kroemker, D. Funnemann, Surf. Sci., 2007, 601, 4727.
[SI94] C. Sousa and F. Illas, Phys. Rev. B, 1994, 50, 13974.
[SSH03] H. Salehi, N. Shahtahmasebi, S.M. Hosseini, Eur.
Phys. J. B, 2003, 32, 177.
[ZBRB08] L.F. Zagonel, N. Barrett, O. Renault, A. Bailly, M.
Baeurer, M. Hoffmann, S.-J. Shih, D. Cockayne, Surf.
Inter. Anal., 2008, 40, 1709.
[LLLM08] X. L. Li, H. B. Lu, M. Li, Z. Mai, H. Kim, Q.J. Jia,
Appl. Phys. Lett., 2008, 92, 012902.
[SNGR09] J. Shin, V.B. Nascimento, G. Geneste, J. Rundgren,
E.W. Plummer, B. Dkhil, S.V. Kalinin, A.P. Baddorf
Nano Letters, 2009, 9 3720.

You might also like