Polymorphism in micro-, Submicro-, and nanocrystalline NaNbO3

Article abstract of DOI:10.1021/jp052974p

NaNbO₃ powders with various particle sizes (ranging from 30 nm to several microns) and well-controlled stoichiometry were obtained through microemulsion-mediated synthesis. The effect of particle size on the phase transformation of the prepared NaNbO₃ powders was studied using X-ray powder diffraction, Raman spectroscopy, and nuclear site group analysis based on these spectroscopic data. Coarsened particles exhibit an orthorhombic Pbcm (D_2h¹¹, no. 57) structure corresponding to the bulk structure, as observed for single crystals or powders prepared by conventional solid-state reaction. The crystal symmetry of submicron powders was refined with the space group Pmc2₁ (C_2v², no. 26). The reduced perovskite cell volumes of these submicron powders were most expanded compared to all the other structures. Fine particles with a diameter of less than 70 nm as measured from SEM observations showed an orthorhombic Pmma (D _2h⁵, no. 51) crystal symmetry. The perovskite formula cell of this structure was pseudocubic and was the most compact one. A possible mechanism of the phase transformation is suggested. ? 2005 American Chemical Society.

Full text of DOI:10.1021/jp052974p

20122
J. Phys. Chem. B 2005, 109, 20122-20130
Polymorphism in Micro-, Submicro-, and Nanocrystalline NaNbO3
,
†
‡
§
|
Yosuke Shiratori,* Arnaud Magrez, J u1 rgen Dornseiffer, Franz-Hubert Haegel,
†
†
Christian Pithan, and Rainer Waser
Institut f u¨ r Elektronische Materialien, Institut f u¨ r Festk o¨ rperforschung (IFF), Forschungszentrum J u¨ lich GmbH,
D-52425 J u¨ lich, Germany, Laboratoire des Nanostructures et des NouVeaux Mat e´ riaux Electroniques
(LNNME), Ecole Polytechnique F e´ d e´ rale de Lausanne (EPFL), CH-1015 Lausanne-EPFL, Switzerland, Institut
f u¨ r Chemie und Dynamik der Geosph a¨ re II (ICG II), Forschungszentrum J u¨ lich GmbH, D-52425 J u¨ lich,
Germany, and NanDOx, Goldenbergstrasse 2, D-50354 H u¨ rth, Germany
ReceiVed: June 3, 2005; In Final Form: August 23, 2005
NaNbO
stoichiometry were obtained through microemulsion-mediated synthesis. The effect of particle size on the
phase transformation of the prepared NaNbO powders was studied using X-ray powder diffraction, Raman
spectroscopy, and nuclear site group analysis based on these spectroscopic data. Coarsened particles exhibit
3
powders with various particle sizes (ranging from 30 nm to several microns) and well-controlled
3
11
an orthorhombic Pbcm (D2h , no. 57) structure corresponding to the bulk structure, as observed for single
crystals or powders prepared by conventional solid-state reaction. The crystal symmetry of submicron powders
2
was refined with the space group Pmc2
1
(C2V , no. 26). The reduced perovskite cell volumes of these submicron
powders were most expanded compared to all the other structures. Fine particles with a diameter of less than
5
7
0 nm as measured from SEM observations showed an orthorhombic Pmma (D2h , no. 51) crystal symmetry.
The perovskite formula cell of this structure was pseudocubic and was the most compact one. A possible
mechanism of the phase transformation is suggested.
+
+
1
. Introduction
is believed to be related to the cationic ordering of K and Na
1
8
in the lattice. In this study, we investigated one of the end
members of such a solid solution, NaNbO3, containing only one
of the alkaline cations, in order to describe the influence of the
particle size by excluding that of atomic distribution.
Lead-free and environmentally friendly NaNbO3-based solid
solutions such as (Kx, Na1-x)NbO3 or (Lix, Na1-x)NbO3 have
presently been drawing interest as piezoelectric materials.
1
-4
Sodium niobate (NaNbO3), being one of the end members of
such solid solutions, is known as an antiferroelectric perovskite-
type compound, which may transform into a ferroelectric one
Bulk NaNbO3 materials (single crystals and ceramics) reveal
5
,6,15,17
a much more complex phase-transition behavior
than the
by chemical doping5 or by an imposed bias electrical dc field.
,6
6
macroscopic KNbO3 system. KNbO3 undergoes several struc-
which are typical for many perovs-
kite-type phases such as BaTiO3, from cubic to tetragonal at
435 °C, and then to orthorhombic and rhombohedral at 225 and
-10 °C, respectively. At room temperature, NaNbO3 is
antiferroelectric. Darlington et al. described the crystal-
lographic structure as a monoclinic one. Later, Xu et al.21
reported the stable NaNbO3 modification at room temperature
as an orthorhombic one with the space group Pbcm (D2h , no.
7). The structure of NaNbO3 has 8 perovskite formulas per
unit cell at room temperature, and consequently, 117 optical
phonons exist. Raman scattering studies successfully reflected
the phase-transition behavior of NaNbO3 associated with the
tilts of NbO6 octahedra, dependent on the thermodynamical
and pressure. Softening of the
librational modes of NbO6 octahedra on cooling from room
temperature was found to be a result of octahedra reorientation.
Pressure-induced spectral changes at 7 GPa have been discussed
as a result of a first-order phase transition and a major structural
change accompanying a transformation of the crystallographic
symmetry. Dielectric properties and Raman scatterings of
NaNbO3 ceramics, which were prepared from powders obtained
through the thermal decomposition of a precursor derived from
a complex niobium salt and sodium nitrate, were also reported,
and it has been concluded that the orientation of NbO6 groups
is a dominant phase-transformation mechanism at temperatures
5
,6,12
Some of the NaNbO3-based solid solutions serve as relaxor-
type dielectric materials, and also, NaNbO3 itself indicates a
relaxor-like dielectric behavior. Recently, nanosized powders
of NaNbO3-based lead-free oxides are believed to potentially
tural phase transitions,
7
19
8
5
9-11
20
gain importance to meet the demand of improved sinterability.
An in-depth understanding of the crystallographic structure
of nanosized crystallites is fundamental in order to maintain or
even improve material functions such as piezoelectric and
dielectric response. The study of size-induced phase transforma-
tions is an essential topic within the establishment of the next
generation of nanotechnology. Recently, (K0.5, Na0.5)NbO3 was
found to exhibit a phase transformation driven by particle size
reduction.10 The transformation was successfully monitored by
X-ray powder diffraction (XRPD) combined with Raman
scattering spectroscopy. The latter is at present the most
powerful method to detect local distortions in a crystal lattice
11
5
15,17
16
variables temperature
15
12-17
and eventually phase transitions.
Actually, (K0.5, Na0.5)-
NbO3 transforms from a monoclinic structure with a space group
of either Pm or P2 to a triclinic one with P1 or P-1 symmetry
at a critical Brunauer-Emmett-Teller (BET) equivalent particle
size of about 200 nm with decreasing particle size.10 This feature
1
6
*
Corresponding author. E-mail: yo.shiratori@fz-juelich.de. Tel: +49-
2
461-61-4389. Fax: +49-2461-61-2550.
†
IFF-Forschungszentrum J u¨ lich.
LNNME-EPFL.
ICG II-Forschungszentrum J u¨ lich.
NanDOx.
‡
§
|
22
below room temperature.
1
0.1021/jp052974p CCC: $30.25 © 2005 American Chemical Society
Published on Web 10/04/2005

Polymorphism in Crystalline NaNbO3
J. Phys. Chem. B, Vol. 109, No. 43, 2005 20123
Particle size, which influences the surface energy and the
internal pressure of a particle,13,23,24 is another important
thermodynamical parameter. Chemical aspects such as homo-
geneity and lattice defects, depending on the synthesis route,
may also play a role. In the present study, we focused on
NaNbO3 powders produced by a microemulsion-mediated
approach allowing control of nucleation and growth during the
formation of ultrafine nanosized particles.11,25 Particle size
induced polymorphism of the prepared NaNbO3 powders has
been investigated by XRPD and Raman scattering spectroscopy.
2
. Experimental Section
.1. Powder Preparation. Stoichiometric NaNbO3 powders
2
were prepared through the hydrolysis of mixed ethanolic
solutions of sodium and niobium ethoxides using a water in oil
microemulsion as the reaction medium. The synthesis process
has already been described in detail elsewhere.1 As the
precursor, an ethanolic solution of sodium-niobium ethoxide
was prepared from pure niobium ethoxide mixed with pure
sodium ethoxide (ABCR). Niobium ethoxide was obtained from
NbCl5 (Alfa Aesar) suspended in toluene (Merck, analytical
grade) by the addition of ethanol and ammonia (Linde), filtering
off ammonium chloride, and subsequent vacuum distillation (158
1
°
C, 0.1 mbar). After the preparation of the precursor solutions,
a stoichiometric amount of the microemulsion was added in
order to form the nanosized powders. Raw powders were
annealed in air for 1 h at various temperatures between 200
and 1000 °C. After cooling to room temperature, a part of all
these powders was heat-treated for 12 h at the same temperatures
in air again.
2.2. Characterization. The stoichiometry of all powders was
evaluated from the results of inductively coupled plasma with
optical emission spectroscopy (ICP-OES) for the metallic ions
and of infrared spectroscopy for oxygen extracted by heating
under helium gas flow. Samples for ICP-OES were prepared
as solutions in H2SO4, which were diluted to a total volume of
5
0 mL by dissolving 10 mg of the respective powders in 3 mL
Figure 1. SEM images of (a) the raw NN powder and the powders
annealed for 12 h at (b) 400, (c) 500, (d) 600, (e) 700, (f) 800, (g) 900,
and (h) 1000 °C.
of heated H2SO4. The carbon content was measured by detecting
the CO2 amount evolving from the samples burnt in flowing
oxygen using infrared spectroscopy. The morphologies and
particle sizes of all powders were studied by scanning electron
microscopy (SEM, Hitachi S-4100). The average particle size
of the powders was also evaluated from their specific surface
areas measured by means of N2 adsorption BET method using
a surface area analyzer (Micromeritics, Gemini 2360). XRPD
patterns and Raman spectra were collected at room temperature
using a Philips X’Pert diffractometer (Cu KR radiation) and a
Jobin Yvon T64000 spectrometer (Ar laser excitation with
5
respectively. Refinement of the XRPD patterns was performed
according to the Rietveld method by using the Fullprof software
package.26 A program in the Grams/AI software package
individual particles is around 30 nm, and the particle shape is
spherical. The powders were coarsened up to the micron order
(see Figure 1h). Figure 2 shows some selected XRPD patterns
measured for the raw and annealed powders. Figure 2a reveals
that the as-synthesized particles are amorphous. The XRPD
patterns for NN @ 500-1 and NN @ 400-12 showed sharp
and intense diffraction patterns (see Figure 2b and c) reflecting
the crystallization of single-phase NaNbO3. This was also
monitored by Raman spectroscopy (see Figure 3). The Raman
spectrum of the raw powder (see Figure 3a) and those for the
samples annealed for 1 h at temperatures up to 400 °C are
mainly composed of carbon chain deformation modes (200-
+
14.5 nm wavelength and <50 mW power at the sample),
-
1
-1
3
00 cm ), C-O and C-C stretching modes (around 900 cm ),
(Thermo Galactic) was used to analyze the Raman scattering
which are ascribed to the surfactant molecules, and their
profiles.
1
1
decomposition products. However, upon annealing for 1 h
above 500 °C or for 12 h above 400 °C, these specific bands
disappeared (see Figure 3b and c), and typical perovskite-type
spectra with intense peaks in the 50-300 cm and 500-700
cm regions
3
. Results and Discussion
.1. Chemical Analysis of the Prepared Powders. Figure
represents the SEM images of the as-synthesized powder used
-
1
3
-
1
10-17
1
have been obtained for these powders. The
as raw material for further annealing and of the powders
contents of organic and inorganic carbon, such as carbonates,
in these powders were less than 0.1 wt %, respectively. Figure
4 shows the molar percentages of the elements dependent on
annealing temperature for the prepared powders as a result of
ICP-OES and infrared spectroscopy. As shown, the stoichio-
metry of Na/Nb/O ) 1:1:3 was achieved for the powders
annealed for 12 h. We will use the terms NN @ T-1 and NN
@
T-12 when the raw powder is annealed for 1 h or for 12 h,
respectively, at the temperature T in Celsius. Particles of the
raw powder (see Figure 1a) are connected by surfactant residues
and form loose agglomerates. The average diameter of the

20124 J. Phys. Chem. B, Vol. 109, No. 43, 2005
Shiratori et al.
Figure 4. Annealing temperature dependence of the molar percentages
Figure 2. XRPD patterns recorded for (a) the raw NN powder, (b)
NN @ 500-1, (c) NN @ 400-12, (d) NN @ 500-12, (e) NN @
of Na (O), Nb (4) and O (0) in the NaNbO
h and (b) 12 h.
3
powders annealed for (a)
1
8
00-12, and (f) NN @ 1000-12.
Figure 5. BET plots obtained for (a) the raw NN powder and the
powders annealed for 12 h at (b) 400, (c) 500, (d) 600, (e) 700, (f)
Figure 3. Raman spectra collected for (a) the raw NN powder, (b)
NN @ 500-1, and (c) NN @ 400-12.
8
00, (g) 900, and (h) 1000 °C.
annealed at temperatures above 400 °C. In summary, powders,
annealed at temperatures above (i) 400 °C for 12 h or (ii) 500
adsorbed in a unimolecular layer, and the BET constant.
Therefore, in the case of unimolecular adsorption, the plots of
p/p0 and p/V(p0 - p) should give a linear correlation (in the
range of p/p0 from 0.05 to 0.35 ). BET plots for the powders
annealed for 12 h are shown in Figure 5. All powders except
for NN @ 1000-12 (see Figure 5h) showed excellent linear
correlations of p/V(p0 - p) as a function of p/p0 from which Vm
values were obtained. The poor linear correlation for NN @
°
C for 1 h, contain particles without residual surfactant. XRPD
combined with Raman scattering confirm that the samples are
single-phase NaNbO3.
2
7
3
.2. Particle Size Determination. To evaluate the average
27
particle size, the BET method was applied in combination with
direct observation of the morphology and the particle dimensions
by SEM.
1
000-12 is probably due to a very small surface area of the
In the case of unimolecuar adsorption, the relationship
between the adsorption equilibrium pressure p and the volume
sample. The specific surface areas (SSA) were calculated, their
errors being less than 5% except for NN @ 1000-12.
Subsequently, BET equivalent particle diameters (dBET) were
calculated by the relationship
27
of gas adsorbed at p (BET equation ) is
p
c - 1
p
1
)
‚
+
V(p - p)
V c p0 V c
0
m m
6
dBET
)
SSA × F
where F is the theoretical density of NaNbO3, for which the
under an adsorption equilibrium state at a constant temperature,
where p0, Vm, and c are the saturation pressure, the volume

Polymorphism in Crystalline NaNbO3
J. Phys. Chem. B, Vol. 109, No. 43, 2005 20125
Figure 7. Annealing temperature dependence of (a) the reduced
perovskite cell volume and (b) the lattice parameters obtained for the
powders annealed for 12 h. The average dSEM values are indicated in
parentheses in graph (a).
Figure 6. Annealing temperature dependence of (a) specific surface
area (SSA: 4, 1 h treatment; 2, 12 h treatment) and BET equivalent
particle diameters (dBET: O, 1 h treatment; b, 12 h treatment) and (b)
particle sizes determined by the line intercept method averaged for about
1
1
as Pbcm (D2h , no. 57) space group with a ) 5.5071(1) Å, b
1
00 particles, which were randomly selected (dSEM: O, 1 h treatment;
)
5.5698(1) Å, and c ) 15.5245(4) Å (hereafter referred to as
b, 12 h treatment). The vertical bars in the graphs (a) and (b) indicate
the errors evaluated from linear regressions for the BET results and
the standard deviations obtained from the quantitative evaluation of
SEM images, respectively.
O1 structure).
For intermediate annealing temperatures 700 °C < T < 900
C, the hkl reflections with odd l values, indexed in the O1
°
symmetry, vanish. This indicates that the unit cell now only
contains two instead of four octahedral units along the c axis.
NaNbO3 exhibits a new structure (hereafter referred to as O2
3
9
theoretical value for single crystals (4.575 g/cm ) was used,
since the changes of density evaluated from the representative
XRPD refinements and the corresponding deviations of dBET
were less than 0.3%. Figure 6 shows the resulting dependence
of SSA, dBET, and dSEM on the annealing temperatures, where
dSEM is the average particle size determined through SEM
analysis. It could be concluded from the BET results and direct
SEM observations that particles in the powders annealed at
temperatures below 500 °C are smaller than 200 nm in dBET
and 100 nm in dSEM, respectively, and that particles coarsen to
a size of >1.2 µm in dBET (dSEM ≈ 1 µm) by an annealing
treatment at 1000 °C. The values of dBET include the effect of
loose agglomeration of particles, and the dSEM values represent
the intrinsic particle sizes. Figure 6a indicates that agglomeration
of particles starts during removal of surfactant at 400 °C and
agglomeration is promoted at 600 °C. In the following part of
this report, we basically use dSEM values to indicate particle
sizes.
2
structure) with Pmc21 space group (C2V , no. 26) and lattice
constants corresponding to a ) 7.7673(3) Å, b ) 5.5150(2) Å,
and c ) 5.5651(2) Å (determined for NN @ 800-12). In
comparison with the O1 structure, the primitive volume is
expanded by up to 0.2%, while particle size is reduced from
1.1 µm to 580 nm through the O1 to O2 transition (NN @ 1000-
12 f NN @ 900-12). Such a behavior is not expected, since
a reduction of the particle size should induce an increase of the
internal pressure resulting from an increase of surface curvature,
which should lead to a reduction of the unit cell volume or a
1
3
transformation to a more symmetrical structure. On the other
hand, the effect of reduced cell volume with decreasing particle
size is clearly observed for the O2 structures: NN @ 900-12
(580 nm) f NN @ 700-12 (180 nm).
For lower temperatures 400 °C < T < 500 °C, a third
polymorph, the O3 structure, was described in the Pmma space
5
3
.3. Size-Induced Phase Transformation. 3.3.1. X-ray
group (D2h , no. 51) with unit cell constants differing from the
Powder Diffraction. On annealing raw powders above 400 °C
for 12 h, several structural changes have been observed (see
Figure 2). The lattice constants could be refined by applying
an orthorhombic symmetry for all samples. The evolution of
the primitive volume V/Z (V and Z are the volume of the lattice
and the number of formula per unit cell, respectively) and the
dimension of a perovskite formula are plotted as a function of
the annealing temperature, i.e., particle size (see Figure 7). Two
discontinuities separating three different temperature ranges, and
thus particle size regions, can be distinguished.
ones of the O2 structure (a ) 5.5211(7) Å, b ) 7.7951(6) Å,
and c ) 5.5235(5) Å). The symmetry increased and the lattice
volume shrunk compared to O2, which could be attributed to
the increase of the internal pressure in the particles. The three
perovskite formula parameters become closer for the O3
structure, which can be considered as a pseudocubic structure.
Nevertheless, the primitive volume of the O3 structure, again,
interestingly expands in the case of smaller particles annealed
at temperatures below 600 °C down to 400 °C.
Moreover, one should note that, for the NN @ 600-12
sample, the refinement of the XRPD pattern was performed by
considering a mixture of two phases. Both O2 and O3 structures
The lattice parameters refined for NN @ 1000-12 were close
to those refined for bulk material.21 The structure is described

20126 J. Phys. Chem. B, Vol. 109, No. 43, 2005
Shiratori et al.
TABLE 1: Results of the Rietveld Refinements and the Nuclear Site Group Analysis Performed for the NaNbO
3
Coarse
Powder (NN @ 1000-12), Intermediate Powder (NN @ 800-12), and Fine Powder (NN @ 500-12)
atom
x
y
z
biso (Å2) Wyck site sym.
irreducible representations
NN @ 1000-12, O
1
, Pbcm (D2h11, no. 57)
Nb
0.2429(4)
0.262(2)
0.258(3)
0.301(3)
0.174(4)
0.516(2)
0.967(2)
0.2627(4)
0.75
0.777(2)
0.25
0.242(4)
0.019(4)
0.466(3)
0.1252(5)
0
0.25
0
0.25
0.140(1)
0.113(1)
0.3(1)
0.7(1)
0.7(1)
0.3(1)
0.3(1)
0.3(1)
0.3(1)
8e
4c
4d
4c
4d
8e
8e
C
1
3A
+ A
2A + A
+ A
g
+ 3A
u
u
+ 3B1g + 3B1u + 3B2g + 3B2u + 3B3g + 3B3u
+ 2B1g + 2B1u + 2B2g + 2B2u + B3g + B3u
+ 2B1g + B1u + B2g + 2B2u + B3g + 2B3u
+ 2B1g + 2B1u + 2B2g + 2B2u + B3g + B3u
x
Na1
Na2
O1
O2
O3
C
C
C
C
C
C
2
A
g
xy
x
s
g
u
2
A
g
u
xy
s
2A
3A
3A
g
g
g
+ A
u
+ 2B1g + B1u + B2g + 2B2u + B3g + 2B3u
1
1
+ 3A
+ 3A
u
+ 3B1g + 3B1u + 3B2g + 3B2u + 3B3g + 3B3u
+ 3B1g + 3B1u + 3B2g + 3B2u + 3B3g + 3B3u
O4
u
presence of an inversion center
Γ
Γ
Γ
Γ
Γ
Γ
Total
15A
B
15A
g
+ 13A
u
+ 17B1g + 15B1u + 15B2g + 17B2u + 13B3g + 15B3u
1u + B2u + B3u
+ 17B1g + 14B1u + 15B2g + 16B2u + 13B3g + 14B3u
Acoustic
Optical
IR
lattice constants (Å) reliability factors
g
+ 13A
u
a
b
c
5.5071(1)
5.5698(1)
15.5245(4)
R
R
R
p
) 15.0
R
R
b
f
) 5.7
) 9.5
14B1u + 16B2u + 14B3u
13A
15A
wp ) 18.6
exp ) 14.0
Silent
Raman
u
g
+ 17B1g + 15B2g + 13B3g
2 1
NN @ 800-12, O , Pmc2
(C2V2, no. 26)
Nb
0.751(1)
0
0.5
0
0.5
0.221(4)
0.271(4)
0.7541(9)
0.257(4)
0.263(4)
0.191(7)
0.317(6)
0.508(3)
0.942(3)
0.779(3)
0.75
0.777(4)
0.317(5)
0.315(5)
0.082(4)
0.513(5)
0.5(1)
1.2(1)
1.2(1)
0.8(1)
0.8(1)
0.8(1)
0.8(1)
4c
2a
2b
2a
2b
4c
4c
C
C
C
C
C
C
C
1
3A
2A
2A
2A
2A
3A
3A
1
1
1
1
1
1
1
+ 3A
2
+ 3B
1
1
+ 3B
2
2
yz
Na1
Na2
O1
O2
O3
s
s
s
s
+ A + B
2
+ 2B
yz
yz
yz
+ A
+ A
+ A
2
+ B
1
+ 2B
+ 2B
+ 2B
2
2
+ B
+ B
1
2
2
1
2
1
1
+ 3A
+ 3A
2
+ 3B
+ 3B
1
+ 3B
+ 3B
2
2
O4
2
1
absence of an inversion center
Γ
Γ
Γ
Γ
Γ
Total
17A
1
+ 13A
2
+ 13B
1
+ 17B
+ 16B
+ 16B
2
Acoustic
Optical
IR
A
1
+ B + B
1
2
lattice constants (Å) reliability factors
16A
16A
16A
1
1
1
+ 13A
+ 12B
+ 13A
2
1
2
+ 12B
+ 16B
+ 12B
1
2
1
2
a
b
c
7.7673(3)
5.5150(2)
5.5651(2)
R
R
R
p
) 14.7
R
R
b
f
) 4.7
) 7.4
wp ) 17.8
exp ) 14.4
Raman
2
NN @ 500-12, O
3
, Pmma (D2h5, no. 51)
yz
Nb
0.75
0.75
0.75
0.25
0.25
0
0.750(2)
0
0.5
0
0.5
0.209(2)
0.280(2)
0.7511(2)
0.254(4)
0.261(3)
0.153(6)
0.178(5)
0.5
0.3(1)
1.4(1)
1.4(1)
0.7(1)
0.7(1)
0.7(1)
0.7(1)
4k
2e
2f
2e
2f
4h
4g
C
C
C
C
C
C
C
s
2A
g
+ A + B1g + 2B1u + B2g + 2B2u + 2B3g + B3u
u
z
z
z
z
Na1
Na2
O1
O2
O3
2V
2V
2V
2V
A
A
A
A
A
A
g
g
g
g
g
g
+ B1u + B2g + B2u + B3g + B3u
+ B1u+ B2g + B2u + B3g + B3u
+ B1u + B2g + B2u + B3g + B3u
+ B1u + B2g + B2u + B3g + B3u
+ A
+ A
y
2
u
+ 2B1g + 2B1u + B2g + B2u + 2B3g + 2B3u
+ 2B1g + 2B1u + B2g + B2u + 2B3g + 2B3u
y
O4
0.5
0
2
u
presence of an inversion center
Γ
Γ
Γ
Γ
Γ
Γ
Total
8A
B
8A
g
+ 3A
u
+ 5B1g + 10B1u + 7B2g + 8B2u + 10B3g + 9B3u
1u + B2u + B3u
+ 5B1g + 9B1u + 7B2g + 7B2u + 10B3g + 8B3u
Acoustic
Optical
IR
lattice constants (Å) reliability factors
g
+ 3A
u
a
b
c
5.5211(7)
7.7951(6)
5.5235(5)
R
R
R
p
) 13.4
R
R
b
f
) 4.9
) 5.9
9B1u + 7B2u + 8B3u
3A
8A
wp ) 17.9
exp ) 13.7
Silent
Raman
u
g
+ 5B1g + 7B2g + 10B3g
are present, which indicates that the particle size within this
powder is not homogeneous.
a stacking of two different kinds of crystallographic layers.
Although they are built from identical NbO6 octahedrons, the
difference arises from a different rotation angle between two
following layers. Even if the level of distortion of the equatorial
plane of the octahedron is higher within the O2 structure
compared to O1, one can consider that the O2 structure distortion
is lower. The different orientations between neighboring layers
are no longer present in the O2 structure symmetry. The
symmetry of NaNbO3 at a particle size below 70 nm (O3
structure) is closer to the ideal cubic perovskite structure: the
rotation between octahedrons within one crystallographic layer
vanished, resulting in a more cubic like structure.
Table 1 represents the data obtained from the Rietveld
refinement of NN @ 1000-12, NN @ 800-12, and NN @
5
00-12 structures as typical representatives of the O1, O2, and
O3 structures, respectively. No drastic changes have been
observed in the bond lengths. The average Na-O (1.79(2) nm)
interatomic distance is close to the value in the literature, while
the length of the octahedral Nb-O edges (1.98(2) nm) is
compressed by about 3%.28
Sketches of all these structures O1, O2, and O3 are shown in
Figure 8 with the three particle size ranges identified for the
powders annealed for 12 h from the results of XRPD and Raman
spectroscopy, which will be discussed in section 3.3.2. The phase
transitions, on particle size reduction, occur largely on the
modification of the octahedral framework. Major differences
in these structures can be recognized along the stacking axis
3.3.2. Raman Spectroscopy. Figure 9 shows the Raman
spectra and their fit traces obtained for NN @ 1000-12, NN
@ 800-12, and NN @ 500-12. The Raman scattering pattern
measured for the highly coarsened powder (NN @ 1000-12)
could be modeled by a combination of 36 Lorentzian functions
(see Figure 9a). The spectrum profiles and the positions of the
bands for the main components indicated excellent agreement
(defined as the axis perpendicular to the equatorial plane of the
octahedrons (i.e., c, a, and b axes for the O1, O2, and O3
structures, respectively). The zigzag chains of octahedra, running
along the apical direction, are unchanged through both phase
transitions (O1 f O2 and O2 f O3). According to the space
group and the lattice constants, the O1 structure can be seen as
1
5-17,22
with the assignments reported for pure bulk materials.
-
1
Two strong bands at 61.0 and 74.5 cm can be assigned to the
Na translational modes against the NbO6 octahedrons. The
bands at 92.2, 120.3, 123.5, 142.8, and 154.0 cm are assigned
+
-1

Polymorphism in Crystalline NaNbO3
J. Phys. Chem. B, Vol. 109, No. 43, 2005 20127
structures.15,16,22 Triply degenerated ν6 (F2u) and ν5 (F2g) modes
showed a splitting of the scattered intensities composed by a
-
1
complicated band profile in the region from 170 to 300 cm .
-
1
Weak ν4 (F1u) bands appeared at 377.3 and 433.1 cm . The ν2
Eg), ν1 (A1g), and ν3 (F1u) modes form the second most intense
band region. The ν2 (Eg) and ν1 (A1g) bands were fitted by two
(
-
1
Lorentzian functions at 556.4, 569.4, and 599.4, 613.3 cm ,
respectively, since their band shapes are asymmetric. The ν3
-
1
(
F1u) mode appeared as a shoulder at 670.3 cm with a
-
1
symmetric peak profile. The final weak band at 870.6 cm is
assigned to the ν1 + ν5 combination mode, which has F2g as an
irreducible representation.
The Raman spectrum for NN @ 800-12 was fitted by 31
functions (see Figure 9b). A remarkable difference with respect
to the spectrum for NN @ 1000-12 was observed in the
-
1
external vibrational modes below 165 cm . The Raman
spectrum registered for NN @ 500-12 (see Figure 9c) could
be fitted by only 22 Lorentzian functions compared to the spectra
for NN @ 1000-12 and NN @ 800-12. This indicates that
the crystal symmetry was drastically modified by the annealing
temperature, i.e., particle size. The evolution of Raman spectra
with the annealing temperature is represented in Figure 10.
Figure 8. Illustrations of the differences between (a) NN @ 1000-
1
2, (b) NN @ 800-12, and (c) NN @ 500-12 structures. Particle
Figure 11 shows that the wavenumbers and the full widths
at half-maximum (fwhm) of specific librational and internal (ν5
and ν3) modes of the NbO6 octahedra change with annealing
temperature, dBET and dSEM, respectively. The wavenumber and
fwhm values are influenced by surrounding lattice defects,
size ranges obtained from XRPD and Raman spectroscopy are also
indicated.
2
2,29
Coulomb interactions, and vibrational relaxation.
The an-
nealing temperature characteristics of the wavenumbers and the
fwhm values were shifted toward lower temperatures by longer
annealing time (see upper graphs in Figure 11). This indicates
that the annealing time promotes the transformations of the
vibrational motions. On the other hand, it seems, according to
the plots of the wavenumber and fwhm versus dBET (see middle
graphs in Figure 11), that a shorter annealing time promotes
the vibrational transformation. In the cases of the dSEM-wave-
number and -fwhm plots, both curves obtained for the shorter
and longer annealing times superimposed each other (see lower
graphs in Figure 11). Therefore, a correlation between particle
size of the powders and lattice dynamics such as NbO6 tilting
and internal vibrations is believed to exist. The corresponding
-
1
bands were located at around 136, 275, and 652 cm up to
00 nm in dSEM (see lower graphs in Figure 11) and abruptly
1
shifted toward higher wavenumber above this particle size.
Figure 12 shows the plots of the wavenumbers versus averaged
dSEM values for all the respective phonons. In the region below
-
1
+
1
60 cm , which includes the Na translational modes and the
NbO6 librational modes, several bands, which have been
observed for the well-coarsened powders, disappeared below
+
dSEM ) 500 nm. The remaining three Na translational modes
finally converged to only one band below dSEM ) 70 nm. Below
-
1
7
0 nm, one of the NbO6 librational modes at 154 cm
disappeared, a final convergence of the ν5 and ν6 bands could
be confirmed, one of the ν4 modes completely disappeared, and
the splitting of the ν1 mode converged. Consequently, it can be
concluded from our Raman spectroscopic results that there are
three different types of structures for coarse (dSEM > 600 nm:
O1), intermediate (200 nm < dSEM < 400 nm: O2), and fine
powders (dSEM < 70 nm: O3). These size ranges obtained from
Raman spectroscopy are indicated in Figure 8.
The results of nuclear site group analysis30 for the O1, O2,
and O3 structures are also shown in Table 1. The analysis for
O1 with a Pbcm structure gave 60 Raman active and 44 IR active
normal modes, which are mutually excluded. This large number
Figure 9. Fit traces of the Raman spectra obtained for (a) NN @
1
d
000-12, (b) NN @ 800-12, and (c) NN @ 500-12. The average
SEM values are indicated in parentheses.
to the librational modes of the NbO6 octahedra. All bands above
-
1
1
60 up to 900 cm are generated by the internal vibrations of
the NbO6 octahedra. Strictly speaking, the ν1 to ν6 modes are
defined for the equilateral shape approximation of the free NbO6
octahedron; however, we apply these descriptions following the
literature for the NbO6 octahedron in the perovskite lattice

20128 J. Phys. Chem. B, Vol. 109, No. 43, 2005
Shiratori et al.
Figure 10. Evolution of the Raman spectra by annealing temperature for (a) 1 h and (b) 12 h treatments. The average dSEM values are indicated
in parentheses.
Figure 11. Annealing temperature, dBET, and dSEM dependence of the wavenumbers (O, 1 h treatment; b, 12 h treatment) and the full widths at
half-maximum (fwhm: 4, 1 h treatment; 2, 12 h treatment) plotted for (a) the NbO
6
librational and (b) ν
5
3
and (c) ν bands. The horizontal bars
in the graphs (b) and (c) indicate the errors estimated from linear regressions for the BET plots and the standard deviations of the particle sizes,
respectively.
of Raman modes expected from the theoretical point of view is
a cause of the complicated scattering profile (see Figure 9a).
Highly coarsened powders showed sharp splittings (61 and 74
reflects 2 major contributions among the different interactions
+
-
between Na and NbO6 ions in the Pbcm structure. With
decreasing annealing temperature, i.e., particle size, these split
bands shifted to higher wavenumbers, and their separation
-
1
+
cm ) of the Na translational mode (see Figure 10). This

Polymorphism in Crystalline NaNbO3
J. Phys. Chem. B, Vol. 109, No. 43, 2005 20129
3
.3.3. Hypothetical Mechanisms. The enhancement of the
internal pressure within fine particles, known as Gibbs-
Thomson effect, is the most extensively assumed mechanism
for the stabilization of more symmetric structures than the
thermodynamically stable ones at ambient temperature and
pressure (accompanied with shrinkage of the lattice).1
However, (K0.5, Na0.5)NbO3 presents a different behavior. This
compound undergoes a phase transformation from a monoclinic
symmetry (M-type) to a triclinic one (T-type) accompanied by
an expansion of the lattice below a critical dBET value of 200
3,23,24
1
0
nm. According to the atomic positions, the unique Na/K
position of the M-type structure is discriminated into two
different ones in the T-type structure, allowing an ordering of
the A-site cations. Such an ordering within Na/K distribution
could occur resulting from possible chemical gradients arising
during the preparation, which could remain in the particle of
the raw powder. The low diffusion rate of Na or K ions due to
low annealing temperature impedes further homogenization. For
NaNbO3 containing only one alkali metal on the A-site of the
perovskite structure, such a phenomenon is not possible. The
distortion related to the rotation of octahedra on different
crystallographic layers is reduced through both phase transitions
with decreasing particle size, i.e., increasing internal pressure.
The expansion observed for the lattice as the particle size is
reduced is not in agreement with the Gibbs-Thomson effect
and not yet elucidated.
Figure 12.
dSEM dependence of the wavenumbers plotted for all
observed vibrational modes.
4
. Conclusions
became smaller. This behavior is quite similar to that observed
A phase transformation induced by reduced particle size was
for Li-doped compositions, as reported for LixNa1-xNbO3
studied by Raman spectroscopy.31
investigated for NaNbO3 powders prepared through microemul-
sion-mediated synthesis and subsequent annealing treatment.
X-ray diffraction and Raman spectroscopy demonstrated that
the orthorhombic Pbcm structure (O1) of coarsened powders in
micron order transforms via the orthorhombic Pmc21 structure
The assigned symmetry Pmc21 for the O2 structure is not
centrosymmetric, and therefore, the Raman-IR mutual exclu-
sion is absent (see Table 1). For this structure, 57 Raman active
and 44 IR active normal modes can be derived. If molecular
dipole-electrostatic field interactions are taken into consider-
ation for the IR active motions, 101 Raman bands are estimated
to exist for the Pmc21 structure due to the LO-TO splittings.
In the present case, only 31 functions instead of the theoretical
number of 57 or 101 bands were necessary in order to model
the Raman spectra for the intermediate powders due to overlap-
ping of many bands, which could not be resolved in the
(O2) to the Pmma symmetry (O3) with decreasing particle size
in submicron (200-400 nm) and nano (<70 nm) orders,
respectively. The O1 f O2 transition causes the disappearance
of the inversion center of the unit cell. The distortion of the
structure is lowered as the particle size is reduced according to
Gibbs-Thompson effect. Nevertheless, the complex lattice
volume evolution of NaNbO3 has not been elucidated yet.
The comparison between NaNbO3 and (K0.5, Na0.5)NbO3
behavior versus particle size lowering indicates that the transition
from monoclinic to triclinic symmetry for (K0.5, Na0.5)NbO3
could originate from the ordering of K and Na cations in the
structure.
5
complicated scattering profile. In fact, although the Pc21b (C2V ,
no. 29) structure of Li0.02Na0.98NbO3 induces LO-TO splitting,
no remarkable band broadening or splitting have been regis-
tered.31 Therefore, long-range static forces are assumed to be
32
weak in the NaNbO3-based materials. The spectra obtained
for NN @ 950-1 downward to NN @ 750-1 and NN @
Acknowledgment. We gratefully thank Mr. Jochen Friedrich
Institut f u¨ r Festk o¨ rperforschung, Forschungszentrum J u¨ lich
9
00-12 downward to NN @ 700-12, which have submicron
(
particle sizes (approximately 600-200 nm), indicated a similar
evolution of Raman spectra compared to the case of the Li-
content dependency (x e 0.3) of the scattering profiles for
GmbH, J u¨ lich, Germany) for SEM studies. The technical support
of Mr. Manfred Michulitz, Mrs. Nadine Merki, and Mrs.
Hannelore Lippert (Zentralabteilung f u¨ r Chemische Analysen,
Forschungszentrum J u¨ lich GmbH, J u¨ lich, Germany) for chemi-
cal analysis is gratefully acknowledged. One of the authors
(Y.S.) sincerely acknowledges the Alexander von Humboldt
Foundation (Bonn, Germany) for granting financial support
through a Research Fellowship.
3
1
LixNa1-xNbO3.
Finally, for the O3 structure, the splitting of the Na+
translational band could not be resolved by the curve analysis,
and consequently, only one broad band at 61 cm-1 was
observed. This could be caused by a higher crystallographic
+
-
symmetry and almost only one type of Na -NbO6 ionic
interaction in the fine particles. This consideration is consistent
with the pseudocubic structure for the O3 structure (see Figure
References and Notes
7
b). The O3 structure assigned to the fine powders is centrosym-
metric with 30 Raman active and 24 IR active normal modes
see Table 1). This lower number of normal modes yields more
(1) Cross, E. Nature (London) 2004, 432, 24.
(
2) Saito, Y.; Takao, H.; Tani, T.; Nonoyama, T.; Takatori, K.; Homma,
T.; Nagaya, T.; Nakamura, M. Nature (London) 2004, 432, 84.
3) Matsubara, M.; Yamaguchi, T.; Sakamoto, W.; Kikuta, K.; Yogo,
(
(
simple scattering profiles in comparison with the O2 and O3
T.; Hirano, S. J. Am. Ceram. Soc. 2005, 88, 1190.
structures.
(4) Ringgaard, E.; Wurlitzer, T. J. Eur. Ceram. Soc. 2005, 25, 2701.

20130 J. Phys. Chem. B, Vol. 109, No. 43, 2005
Shiratori et al.
(
(
5) Shirane, G.; Newnham, R.; Pepinsky, R. Phys. ReV. 1954, 96, 581.
6) Jaffe, B.; Cook, W. R.; Jaffe, H. Piezoelectric Ceramics; Academic
(19) Merz, W. J. Phys. ReV. 1949, 76, 1221.
(20) Darlington, C. N. W.; Knight K. S. Physica B 1999, 266, 368.
Press: London, 1971; p 185.
7) Raevski, I. P.; Prosandeev, S. A. J. Phys. Chem. Solids 2002, 63,
939.
(21) Xu, H.; Su, Y.; Balmer, M. L.; Navrotsky, A. Chem. Mater. 2003,
(
1
5, 1872.
1
2
1
2
(22) Lima, R. J. C.; Freire, P. T. C.; Sasaki, J. M.; Ayala, A. P.; Melo,
(8) Lanfredi, S.; Lente, M. H.; Eiras, J. A. Appl. Phys. Lett. 2002, 80,
F. E. A.; Filho, J. M.; Serra, K. C.; Lanfredi, S.; Lente, M. H.; Eiras, J. A.
J. Raman Spectrosc. 2002, 33, 669.
(23) Uchino, K.; Sadanaga, E.; Hirose, T. J. Am. Ceram. Soc. 1989, 72,
1555.
(24) Chraska, T.; King, A. H.; Berndt, C. C. Mater. Sci. Eng. 2000,
A286, 169.
731.
(9) Nobre, M. A. L.; Longo, E.; Leite, E. R.; Varela, J. A. Mater. Lett.
996, 28, 215.
(
10) Shiratori, Y.; Magrez, A.; Pithan, C. Chem. Phys. Lett. 2004, 391,
88.
11) Pithan, C.; Shiratori, Y.; Dornseiffer, J.; Haegel, F.-H.; Magrez,
A.; Waser, R. J. Cryst. Growth 2005, 280, 191.
12) Shen, Z. X.; Hu, Z. P.; Chong, T. C.; Beh, C. Y.; Tang, S. H.;
Kuok, M. H. Phys. ReV. B 1995, 52, 3976.
13) Begg, B. D.; Finnie, K. S.; Vance, E. R. J. Am. Ceram. Soc. 1996,
9, 2666.
14) Wada, S.; Suzuki, T.; Osada, M.; Kakihana, M.; Noma, T. Jpn. J.
Appl. Phys. 1998, 37, 5385.
15) Shen, Z. X.; Wang, X. B.; Kuok, M. H.; Tang, S. H. J. Raman
Spectrosc. 1998, 29, 379.
16) Shen, Z. X.; Wang, X. B.; Tang, S. H.; Kuok, M. H.; Malekfar, R.
J. Raman Spectrosc. 2000, 31, 439.
17) Bouziane, E.; Fontana, M. D.; Ayadi, M. J. Phys.: Condens. Matter
003, 15, 1387.
18) Shiratori, Y.; Magrez, A.; Pithan, C. J. Eur. Ceram. Soc. 2005, 25,
075.
(
(25) Pithan, C.; Hennings, D.; Waser, R. Int. J. Appl. Ceram. Technol.
(
2
004, 2, 1.
(26) Roisnel, T.; Rodriguez-Carjaval, J. Physica B 1993, 192, 55.
(
7
(27) Brunauer, S.; Emmett, P. H.; Teller, E. J. Am. Chem. Soc. 1938,
(
60, 309.
(28) Shannon, R. D. Acta Crystallogr. 1976, A32, 751.
29) Noma, T.; Wada, S.; Yano, M.; Suzuki, T. J. Appl. Phys. 1996,
0, 5223.
(
(
8
(
(30) Rousseau, D. L.; Bauman, R. P.; Porto, S. P. S. J. Raman Spectrosc.
1
981, 10, 253.
(
(
31) Juang, Y. D.; Hu, M. L.; Tse, W. S. J. Appl. Phys. 1994, 76, 3746.
(32) Xia, H.-R.; Li, L.-X.; Yu, H.; Meng, X.-L.; Zhu, L.; Hu, L.-J. Cryst.
Res. Technol. 1999, 34, 901.
2
(
2