Available online at www.sciencedirect.

com

ScienceDirect
Journal of the European Ceramic Society 35 (2015) 2479–2488

Structure–property relationships in (1 − x)BaTiO3–xBiGdO3 ceramics

Giorgio Schileo a,∗ , Antonio Feteira a , Klaus Reichmann b , Ming Li c , Derek C. Sinclair c
a Christian Doppler Laboratory for Advanced Ferroic Oxides, Sheffield Hallam University, Howard Street, S1 1WB Sheffield, UK
b Christian Doppler Laboratory for Advanced Ferroic Oxides, Institute for Chemistry and Technology of Materials, Graz University of Technology,
Stremayrgasse 9, Graz A-8010, Austria
c Department of Materials Science and Engineering, The University of Sheffield, Mappin Street, S1 3JD Sheffield, UK

Received 16 December 2014; received in revised form 1 March 2015; accepted 2 March 2015
Available online 1 April 2015

Abstract
(1 − x)BaTiO3 –xBiGdO3 ceramics were prepared by the solid state reaction method. X-ray diffraction and Raman spectroscopy indicate a maximum
co-solubility of Bi/Gd in BaTiO3 at x = 0.10 with a change of symmetry from tetragonal to pseudo-cubic at x = 0.08. Backscattered electron images,
however, reveal the presence of a secondary phase in x ≥ 0.06. The dielectric behaviour evolves continuously with x from a classical ferroelectric to
a typical relaxor and this transition is accompanied by a shift in the permittivity maxima towards lower temperatures. The presence of two dielectric
anomalies for x ≥ 0.06 is associated with residual core–shell structures, as revealed by transmission electron microscopy. The dielectric anomaly
associated with the core regions remains at ∼120 ◦ C, whereas the other anomaly decreases continuously towards lower temperature with x. This
study shows that chemical equilibrium is much more difficult to achieve than in other (1 − x)BaTiO3 –xBi[Me]O3 systems, where Me is Yb or Sc.
© 2015 Elsevier Ltd. All rights reserved.

Keywords: BaTiO3 ; Bi; Gd; Relaxor; Core–shell

1. Introduction themselves as dielectric anomalies (maxima) on the temperature

dependence of the relative permittivity, εr . The ferroelectric-to-
BaTiO3 (BT) is a well-known ferroelectric perovskite [1] that paraelectric transition at 130 ◦ C is accompanied by a sharp peak
is widely used in the manufacture of multilayer ceramic capaci- in εr , which reach values as high as ∼10,000–12,000, depend-
tors mainly due to its inherent large relative permittivity at room ing on microstructure [5,6]. This large temperature dependence
temperature (εr ∼1600) [2]. Above its Curie temperature, TC of εr is undesirable for most of its technological applications.
(∼130 ◦ C) the crystal structure symmetry is pseudo-cubic (space Chemical doping can be used to shift TC and to broaden the tem-
group Pm3̄m). On cooling, it undergoes three consecutive struc- perature dependence of εr . Indeed, the vast majority of dopants
tural phase transitions, at 130 ◦ C from pseudo-cubic to tetragonal lower and broaden the TC maximum (with a few exceptions
(P4mm), at 0 ◦ C from tetragonal to orthorhombic (Amm2) and at like Pb2+ and Ca2+ ) because they interfere with the long-range
−90 ◦ C from orthorhombic to rhombohedral (R3c). All transi- ordering of dipoles [7].
tions encompass a second-order Jahn–Teller distortion involving Simultaneous replacement of Ba2+ by Bi3+ and Ti4+ by a
hybridisation between the Ti 3d and O 2p orbitals, causing a dis- trivalent metal for charge compensation has been studied in
placement of Ti4+ with respect to the anions, thereby leading recent years. Bi3+ , like Pb2+ , possesses a 6s2 lone pair of elec-
to the appearance of spontaneous polarisation along differ- trons which is not involved in any chemical bonds. This lone
ent crystallographic directions [3,4]. These transitions manifest pair is highly polarisable and it has been demonstrated, e.g. in
compounds such as BiFeO3 and BiMnO3 , a long-range ordering
of lone pairs can provide an alternative source for the occurrence
∗ Corresponding author. Tel.: +44 7468425319. of ferroelectricity [4]. Another reason why Bi doping of BT is
E-mail address: giorgio.schileo@gmail.com (G. Schileo). currently under investigation concerns the search for lead-free

http://dx.doi.org/10.1016/j.jeurceramsoc.2015.03.004
0955-2219/© 2015 Elsevier Ltd. All rights reserved.
2480 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488

ceramics to comply with new European regulations aiming at 800, 900 and 1000 ◦ C for 8 h, with intermediate milling cycles
the gradual replacement of lead in electronics, due to its toxic- between each calcination. Calcination at 1000 ◦ C was repeated
ity. Bi has been used as dopant in BT in combination with Al, until no changes were observed in X-ray diffraction (XRD) data.
Sc, In and Yb [8–11]. The calcined powders were mixed with 5 wt% polyethylene
BiAlO3 is thermally unstable and decomposes at 550 ◦ C, just glycol to improve the mechanical strength of the green bod-
30 C above its TC of 520 ◦ C. BiAlO3 solid solution limit with
◦ ies. Pellets of 13 mm in diameter were uniaxially pressed in
BaTiO3 (rAl = 0.535 Å, 11% smaller than Ti [12]) was reported a stainless steel die (Specac, Kent, UK) under an applied pres-
to be x ∼ 0.12, with a phase transition from tetragonal to rhom- sure of 50–100 MPa. The final sintering was carried out between
bohedral at x ∼ 0.10. Nevertheless, the composition x = 0.15 was 1350 ◦ C (for 0.00 ≤ x ≤ 0.06) and 1400 ◦ C (for 0.08 ≤ x ≤ 0.15)
also investigated and described as a typical relaxor, even if a more for 2 h, in a closed crucible and covering the pellets with
diffuse character of the dielectric peak starts to appear at x = 0.10, source powder of the same composition. The firing profile
together with the so-called “pinching” of phase transitions (the included a de-binding step using a heating rate of 1 ◦ C/min
gradual merging of all three phase transitions in pure BT in one up to 350 ◦ C followed by heating at 5 ◦ C/min to the sintering
single broad peak in doped BT). (1 − x)BaTiO3 –xBiScO3 solid temperature.
solution (rSc = 0.745 Å [12]) presents a phase transition from Room-temperature XRD patterns were recorded with an
tetragonal to pseudocubic at x ∼ 0.05–0.075. The lattice param- X-ray diffractometer in transmission geometry (model: AXS
eters follow the empirical Vegard’s law and no signs of secondary D8 Advance, BrukerTM , Coventry, UK) using K␣-Cu line at
phases up to x = 0.40; TC decreases slowly with composition λ = 1.54059 Å and a step size of 0.02◦ . Rietveld refinements
until x = 0.06, then more steeply where clear relaxor behaviour is were performed with the General Structure Analysis System
established. With increasing x, the permittivity maximum moves (GSAS) suite of programs [14]; a polynomial function of up to
towards higher temperatures. seven terms was used to fit the background, whereas a pseudo-
BiInO3 cannot be prepared at ambient pressure, however Bi Voigt function was chosen to fit the peaks. Thermal displacement
(rBi = 1.45 Å, extrapolated [12]), and In (rIn = 0.80 Å [12]) can parameters were set to isotropic. Silicon powder (99.999%,
substitute for Ba and Ti in BaTiO3 , respectively, up to x = 0.25. In Alpha Aesar, 325 mesh, lot no. 5001L26T) was mixed together
this case, the tetragonal to cubic transition lies between x = 0.10 with the samples and used as reference material. The Si cell
and x = 0.12, with the two phases clearly coexisting at x = 0.10. parameters were fixed while the zero correction in GSAS was
Finally, the (1 − x)BaTiO3 –xBiYbO3 system adopts a varied to best fit them [15]. Raman spectra were obtained with
tetragonal structure which continuously decreases in tetrago- a Raman Microscope (model: inVia, RenishawTM , New Mills,
nality until x = 0.06, above which it can be fully indexed as UK) in backscattering geometry and a 532 nm non-polarised
pseudocubic. TC and the degree of tetragonality decrease accord- Argon laser light using an objective lens of 20× and 50× mag-
ingly, but above x = 0.06 the permittivity maximum shifts again nification.
towards higher temperatures. Relative permittivity measurements above room tempera-
To the best of our knowledge, the impact of Bi and Gd ture were carried out with an Impedance/Gain Phase Analyser
co-doping on the structure and properties of BaTiO3 has not (model: 1260, Solartron Instruments, Farnborough, UK) cou-
been reported. Here we use a combination of X-ray diffraction pled to a tube furnace (model: MTF, Carbolite, Hope Valley,
(XRD), Raman spectroscopy, scanning and transmission elec- UK). Sub-ambient measurements were performed using an
tron microscopy, and dielectric measurements to establish the LCR meter (Agilent E4980A, Agilent, USA) in a closed-cycle
structure–property relationships in ceramics prepared according He refrigerator (Oxford Instruments Ltd., Oxfordshire, UK).
to the formula Ba1−x Bix Ti1−x Gdx O3 . BiGdO3 is a putative end- The sintered pellets were coated with Pt or Ag paste. Capac-
member, as it has never been synthesised under normal ambient itance was measured versus temperature (from 300 to 520 K
pressure. It has been proposed that a hypothetical compound and from 10 to 300 K at 2 K intervals) at four different fre-
may be stable with the perovskite structure for a tolerance fac- quencies (1, 10, 100 and 1000 kHz). Sintered pellets were
tor, t with 0.88 < t < 1.09 [13]. The value of t for the end-member polished using SiC sandpaper and 6 and 1 ␮m diamond paste
BiGdO3 is below the lower limit (0.862). In fact, it is much polishing pads, and subsequently thermally etched at 1200 ◦ C
smaller than the t factor of prototypical perovskites such as for 30 min.
BaTiO3 and SrTiO3 (1.062 and 1.002, respectively). Ceramic microstructures were investigated using scanning
electron microscopes (SEM) (Model: XL30 with LaB6 crystal,
2. Experimental Philips, The Netherlands and model: Nova Nano 200 with a
Field Emission Gun, FEI, Czech Republic) operated at 20 kV
BaCO3 (Sigma–Aldrich, UK, ACS reagent, >99%), and using a 5–10 mm working distance. SEM micrographs were
TiO2 (Sigma–Aldrich, UK, ACS reagent, >99%), Gd2 O3 taken from both polished and fractured surfaces. Samples were
(Sigma–Aldrich, UK, ACS reagent, 99.9%), Bi2 O3 (Sigma– examined in both secondary electron (SE) and backscattered
Aldrich, UK, ACS reagent, 99.9%) were weighed according electron (BSE) imaging modes. Chemical microanalysis were
to the (1 − x)BaTiO3 –xBiGdO3 stoichiometry to obtain solid carried out by energy dispersive X-ray spectroscopy (EDS).
solutions with 0.00 ≤ x ≤ 0.15, and ball milled overnight with EDS signals were optimised for the best signal-to-noise ratio
propan-2-ol using yttria-stabilised zirconia milling media. The and collected for 60 s (at each point). For TEM analysis, the
obtained slurries were dried at 80 ◦ C and fired consecutively at ceramics were ground manually with SiC sandpaper until about
 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488 2481

Fig. 1. (a) Room temperature XRD patterns for (1-x)BaTiO3 -xBiGdO3 solid Fig. 2. Compositional dependence of lattice parameters and unit cell volume.
solutions (logarithmic scale): (1) x = 0.00, (2) x = 0.02, (3) x = 0.04, (4) x = 0.06,
(5) x = 0.08, (6) x = 0.10, (7) x = 0.15 (* = Bi1.55 Gd0.45 O3 – PDF no. 00-048-
0351). (b) Evolution of (2 0 0)/(0 0 2) peak splitting with composition.

80 ␮m, mounted on copper rings, fixed with epoxy resin and

bombarded with Ar ions for 3 h using a Precision Ion Polishing
System (PIPSTM ) (model 691, Gatan, Abingdon, UK) operating
at a milling angle of ±8◦ and beam energy of 4 keV. A TEM
(model CM20, Philips, The Netherlands) equipped with a tung-
sten filament and operated at an accelerating voltage of 200 kV
was used for examining the grain substructure.

3. Results

3.1. Structure and purity

Fig. 3. Comparison of unit cell volume between (1 − x)BaTiO3 –xBiYbO3 [13]
and (1 − x)BaTiO3 –xBiGdO3 for 0 ≤ x ≤0.08.
Room-temperature XRD patterns for (1 − x)BaTiO3 –
xBiGdO3 (0 ≤ x ≤ 0.15) powders are shown in Fig. 1a. For
x ≤ 0.10, all reflections shift towards lower 2θ with increas- Lattice parameters calculated from the Rietveld refinements
ing x, indicating an increase in unit cell volume. In addition, are listed in Table 1. The unit cell volume increases almost lin-
splitting of doublet reflections such as (2 0 0)/(0 0 2) gradually early up to x = 0.10 and then reaches a plateau, suggesting the
decreases with increasing x, indicating a reduction in the degree solid solution limit has been achieved, Fig. 2. The composi-
of the tetragonal distortion, Fig. 1b. For x ≤ 0.06, diffraction data tional variation of the unit cell volume is compared with results
can be indexed on a tetragonal structure (space group P4mm), for (1 − x)BaTiO3 –xBiYbO3 [11] in Fig. 3. The volumes start to
whereas for x ≥ 0.10 the average crystal symmetry appears to be diverge at x ≥ 0.02. Moreover and rather surprising, the unit cell
consistent with a cubic perovskite (space group Pm3̄m). Tetrago- volume for the Gd-series appears smaller than that for the Yb-
nal and cubic symmetries coexist for x = 0.08. No super-lattice series counterparts. This result is contrary to expectation because
reflections that may indicate the presence of octahedral tilting Gd has a larger ionic radius than Yb, and its incorporation on the
or cation ordering were detected, but a minor secondary phase B-site should produce a larger unit cell. Nevertheless, Rietveld
was detected, as indicated by the asterisk in Fig. 1. The extra refinement shows evidence for Bi (A-site) vacancies, especially
peaks match well the XRD pattern of Bi1.55 Gd0.45 O3 (PDF no. for higher dopants concentrations. χ2 (goodness of fit) slightly
00-048-0351). improves if the occupancy of A-sites is lower than its theoretical

Table 1
Lattice parameters, unit cell volume and c/a ratio for (1-x)BaTiO3 -xBiGdO3 , obtained from Rietveld refinements of room temperature XRD patterns (sintered pellets).

x a (Å) c (Å) V (Å3 ) c/a ratio

0.00 3.9910(2) 4.0319(2) 64.22(1) 1.01025

0.02 4.0032(1) 4.0323(2) 64.62(1) 1.00727
0.04 4.0076(1) 4.0354(1) 64.81(1) 1.00694
0.06 4.0104(1) 4.0372(2) 64.93(1) 1.00668
0.08 4.0198(2) 4.0291(1) 4.0375(2) 65.24(1) 65.41(1) 1.00440
tetragonal cubic tetragonal cubic
0.10 4.0344(9) 65.66(4) 1
0.15 4.0339(4) 65.64(2) 1
2482 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488

Fig. 5. Compositional dependence (x) of selected Raman modes.

Fig. 4. Room temperature Raman spectra for (1 − x)BaTiO3 –xBiGdO3 ,

x = 0.00–0.15 (from bottom to top). chemical composition(s), as appropriate. In undoped BaTiO3
the grain size reaches 10–20 ␮m [18], Fig. 6a; grain growth
value. However, χ2 for some compositions could not be reduced is inhibited upon incorporation of small amount of dopants:
below 3, therefore any conclusion from the fitting of the data for 0.02 ≤ x ≤ 0.06, the grain size distribution appears quite
must be drawn with caution. narrow with most grains ranging from 1 ␮m (or even less) to
about 3 ␮m, Fig. 6b–d. Although x = 0.02 appears dense, the
3.2. Raman spectroscopy number and the size of pores in x ≥ 0.04 ceramics appears to
increase. BSE images revealed a very small amount of parasitic
Raman spectroscopy has a shorter length of scale than XRD, phase in x = 0.04 (Fig. 6c) that could not be detected by XRD.
therefore it allows the identification of local deformations aris- EDS analysis performed on 10 points starting from the central
ing from the difference between the ionic radii of the dopants plane of the cross-section towards the surface for x = 0.02 and
and of the ions that they replace [16]. In pure, polycrys- 0.08 did not show any evidence of compositional gradients. In
talline BaTiO3 there are three broad modes at about 260 cm−1 both samples, the composition in the centre matched the com-
[A1 (TO)], 520 cm−1 [A1 (TO)] and 720 cm−1 [A1 (LO)], a dip position near the surface, especially for the value of Bi. Pores
at 180 cm−1 [A1 + E] and a sharp peak at 307 cm−1 [E(TO)] >5 ␮m were observed on the polished cross-section of x = 0.08
[11]. The sharp peak at 307 cm−1 is commonly regarded as a ceramics, whereas the grain size distribution remained between
“signature” of ferroelectric behaviour (long-range ordering of 1 and 5 ␮m. For 0.10 ≤ x ≤ 0.15, large pores (5–10 ␮m) were
dipoles) in BaTiO3 [17]. The room temperature Raman spectra visible across the surface of as-fired ceramics; after grinding
for Ba1−x Bix Ti1−x Gdx O3 , Fig. 4, show that small concentra- those surfaces samples were re-examined and the high degree of
tions of Bi and Gd have a dramatic impact on the observable porosity remained, indicating that bulk porosity is an inherent
modes. In particular, the sharp peak at ∼307 cm−1 starts to feature in those ceramics. This observation corroborates the low
broaden for x = 0.02 and eventually disappears at x = 0.08. The values of relative density measured for these ceramics (∼90%).
dip around 182 cm−1 in pure BaTiO3 is also affected by small Also, the presence of the parasitic phase (bright areas between
amounts of Bi/Gd and disappears at x = 0.04. Probably the most grains) was more evident in 0.08 ≤ x ≤ 0.15, as shown by the
important feature is the extra band at about 830 cm−1 , whose BSE images. For 0.10 ≤ x ≤ 0.15, the grain size distribution
intensity is proportional to the dopant concentration, and a new narrowed slightly, as smaller grains coalesced into ∼5 ␮m
mode above 180 cm−1 ; the latter was previously observed in the grains, and the intergranular space between the grains became
BaTiO3 –BiYbO3 when the average crystal symmetry changed almost completely filled by the Bi and Gd enriched secondary
from tetragonal to pseudocubic [11]. The compositional depend- phase, Fig. 6e–g. EDS analysis on the secondary phase is listed
ence of the phonon modes is illustrated in Fig. 5. The A1 (TO) in Table 2.
mode at 260 cm−1 continuously hardens up to x = 0.08, whereas Bright-field TEM of x = 0.10 revealed the presence of a
the A1 (LO) mode at 720 cm−1 shows a frequency inflection at core/shell structure, as illustrated in Fig. 7. The outer region of
this concentration. This variation reflects the change of the aver- the grains showed the typical mottled contrast commonly exhib-
age crystal symmetry observed from the XRD data. The new ited by relaxors. A similar observation was reported by Ogihara
modes, 1 and 2, soften continuously with increasing x, which et al. [9] for (1 − x)BaTiO3 –xBiScO3 ceramics.
may be associated with the increase in the unit cell volume.
3.4. Dielectric properties
3.3. Ceramic microstructure and phase assemblage
The temperature dependence of εr for (1 − x)
Microstructure and phase assemblage were investi- BaTiO3 –xBiGdO3 ceramics is shown in Fig. 8. The per-
gated by SEM with BSE images used to identify any mittivity maximum, εmax , broadens and shifts towards lower
compositional inhomogeneity and EDS employed to estimate temperature with increasing x; up to x = 0.06 (Fig. 8d) typical
 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488 2483

Fig. 6. SEM micrographs of ceramics sintered at (a–d) 1350◦ C and (e and f) 1400◦ C for 2 h, polished and thermally etched at 1200◦ C for 30 minutes: (a) x = 0.00,
(b) x = 0.02, (c) x = 0.04, (d) x = 0.06, (e) x = 0.08, (f) x = 0.10 and (g) x = 0.15.
2484 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488

Table 2
EDS analysis on (a) matrix grains and on (b) the secondary phase (white areas in BSE images); composition is given in at.% excluding oxygen.
Bulk (a) 0.02 0.04 0.06 0.08 0.10Sint. t = 2 h 0.10Sint. t = 10 h 0.15

Ti 47.9 ± 0.5 48.6 ± 1.1 46.2 ± 0.9 45 ± 0.3 44.0 ± 0.8 45.4 ± 0.3 43.9 ± 0.7
Ba 49.8 ± 0.5 47.9 ± 0.5 48.2 ± 0.6 49 ± 0.6 48.8 ± 0.5 47.7 ± 0.5 49.3 ± 0.5
Gd 1.43 ± 0.21 2.1 ± 0.2 3.6 ± 0.6 4.3 ± 0.7 5.3 ± 0.7 5.3 ± 0.3 5.2 ± 0.4
Bi 0.85 ± 0.12 1.4 ± 0.2 2.1 ± 0.9 1.3 ± 0.4 1.8 ± 1.0 1.7 ± 0.3 1.6 ± 0.2

Secondary phase (b) 0.08 0.10Sint. t = 10 h 0.15

Ti 23.8 ± 1.6 4.7 ± 1.8 3.2 ± 0.1

Ba 24.9 ± 2.0 4.5 ± 2.2 2.5 ± 0.2
Gd 16.2 ± 1.1 33.7 ± 1.4 23.9 ± 0.1
Bi 35.1 ± 2.5 57.1 ± 2.7 70.5 ± 0.1

ferroelectric behaviour is observed, whereas the system evolves 4. Discussion

into a relaxor for x ≥ 0.10; for x = 0.08 there are two dielectric
anomalies, ascribable to the ferroelectric and the relaxor 4.1. Structure and purity
phase (see Fig. 8e). This is consistent with the XRD patterns,
where the system transforms from tetragonal to pseudocubic XRD data indicate x = 0.10 as the solid solution limit for this
with the coexistence of the two phases, Fig. 1b. The value system. Based on reported data for Ba1−x Bix Ti1−x [MeIII ]x O3
of the dielectric permittivity maximum, εmax , decreases from solid solutions (where MeIII is a trivalent metal) the solid solu-
∼4000 for 0.02 ≤ x ≤ 0.06 to ∼2000 for x ≥ 0.10, whereas tion limit may depend (among other factors) on the ionic radius
the dielectric loss is 10% or lower for most frequencies of the B site dopant: for Al, due to its small size, it is only 12%, it
and compositions. Thus, according to the overall dielectric then reaches a maximum at 40% for Sc before decreasing again
behaviour, (1 − x)BaTiO3 –xBiGdO3 ceramics can be divided for the larger Yb (30%), Table 3.
into two groups: for 0.02 ≤ x ≤ 0.06 only one dielectric From additional XRD patterns taken at intermediate stages
anomaly is detectable, whereas for x ≥ 0.08 a second maximum during the fabrication process it was possible to follow the evo-
is visible. The first anomaly is composition-dependent and lution of parasitic phases with processing temperature, Fig. 9.
therefore shifts to lower temperature with increasing x at At 800 ◦ C unreacted Bi2 O3 (PDF no. 01-076-2478) was still
∼10 ◦ C/mol, the other anomaly remains constant at T ∼ 130 ◦ C. present, alongside with Gd2 TiO5 (PDF no. 00-034-1307). After
The latter is smaller in magnitude due to the small amount milling and re-firing up to 1000 ◦ C, a stable phase forms:
of ferroelectric phase still present in x = 0.10 and x = 0.15, Bi1.208 Gd0.792 O3 (PDF no. 00-040-0318). To promote com-
Fig. 8g. plete reaction into a single phase, the sintering temperature was
increased to 1400 ◦ C. At this stage however, a small amount
of parasitic phase remained observable on a logarithmic scale
(Figs. 1 and 9). This phase was identified as Bi1.55 Gd0.45 O3
(PDF no. 00-048-0351), which – compared to Bi1.208 Gd0.792 O3
– presents a higher Bi-to-Gd ratio; the peak positions of the
BaTiO3 doped lattice shift towards lower angles after sinter-
ing at 1400 ◦ C: the (0 0 1)/(1 0 0) reflection shifts from 22.101◦
to 21.980◦ , (1 1 0) from 31.349◦ to 31.295◦ , and (1 1 1) from
38.712◦ to 38.586◦ . This indicates that at least some Gd from
the parasitic phase has become incorporated into the BaTiO3
lattice.

Table 3
Comparison of ionic radii and solid solution limits for (1 − x)BaTiO3 –
xBi[MeIII ]O3 .
Radius (nm) Solid solution limit Ref.

Al 0.535 12% [8]

In 0.68 25% [10]
Sc 0.745 40% [9]
Yb 0.868 30% [11]
Fig. 7. TEM micrograph of composition x = 0.10 sintered at 1400 ◦ C for 2 h, Gd 0.938 10% [This study]
showing a core/shell structure.
 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488 2485

Fig. 8. Temperature dependence of the relative permittivity in the range 10–523 K for ceramics: x = 0.00 (a), x = 0.02 (b), x = 0.04 (c), x = 0.06 (d), x = 0.08 (e), x = 0.10
(f) and x = 0.15 (g).

Although the unit cell volume increases with x, as illustrated In the present case, Gd3+ may enter the A site, due to a relatively
in Fig. 2, based on ion size arguments this is not as high as slow incorporation of Bi and/or its volatilisation at high temper-
expected when compared to the compositional dependence of atures, however Rietveld refinement also shows evidence for Bi
Bi/Yb-doped BaTiO3 , given that Gd3+ in 6-fold coordination is (A site) vacancies, especially for higher dopants concentrations.
∼8% larger than Yb3+ [12], Fig. 3. This can be interpreted by χ2 (goodness of fit) for some compositions could not be reduced
incomplete incorporation of Bi and/or Gd. Alternatively, Gd may below 3, maybe due to unaccounted residual lattice stress, the
occupy simultaneously both the A and B sites of the BaTiO3 lat- small size of the nanodomains in the relaxor compositions or the
tice. In fact, ab initio calculations indicate that Gd3+ can occupy presence of the secondary parasitic phase. All these factors can
both A and B sites, leading to a self-compensation mechanism. affect peak shape and therefore the quality of the fit.
2486 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488

[19] and (1 − x)BaTiO3 –xLaYbO3 solid solutions [5], since its

intensity increases with the dopant level.

4.3. Phase assemblage

EDS results are presented in Table 2. The values of Bi con-

centration in the matrix are systematically below the theoretical
value; a second phase appears between the grains with a high
concentration of Bi and Gd; the results of EDS analysis on these
areas in different samples (Table 2b) do not match with each
other, possibly because of the small size of the impurity phase
grains and/or local inhomogeneity within the impurity itself,
nonetheless they indicate that Bi is far more abundant than Gd,
Fig. 9. Evolution of parasitic phases with sintering temperature (x = 0.10); which supports the identification of this phase as Bi1.55 Gd0.45 O3
j = Bi2 O3 (PDF no. 01-076-2478), k = Gd2 TiO5 (PDF no. 00-034-1307), by XRD. It can be concluded that a certain amount of Bi and
l = Bi1.208 Gd0.792 O3 (PDF no. 00-040-0318), m = Bi1.55 Gd0.45 O3 (PDF no. 00- Gd, rather than diffusing in the BaTiO3 matrix, remains in inter-
048-0351).
granular spaces, even with multiple calcination and intermediate
milling steps; the incorporation of Bi is slowed by the presence
4.2. Dielectric properties of a rather stable secondary phase, a fact which is corroborated
by the EDS results on a x = 0.10 sample sintered at 1200 ◦ C for
The ferroelectric-to-relaxor crossover due to the increasing 2 and 10 h (Table 2a). Rietveld refinements indicate only a loss
amount of dopants in the BaTiO3 matrix can be followed by the of Bi, since the X-ray diffraction intensity from these impuri-
evolution in the dielectric behaviour, Raman spectra and XRD ties is too weak to be detected. The presence of impurity peaks
patterns. All these techniques allowed two regimes to be identi- for 0.04 ≤ x ≤ 0.06 is only possible to be observed in logarith-
fied: in the first one, up to x = 0.06, the global crystal structure mic scale. The presence of this secondary phase is therefore
remains tetragonal at room temperature, as confirmed by the undetected, or underestimated, by XRD. Increasing the time
presence of a sharp peak at 307 cm−1 in the Raman spectra, of sintering at 1200 ◦ C from 2 to 10 h increased the density
the splitting of the (2 0 0)/(0 0 2) doublet in XRD data and from by reducing the porosity and by increasing the grain size but
the temperature dependence of the permittivity. Higher levels nonetheless a complete homogenisation of Bi and Gd into the
of dopants however cause a series of effects on both the overall BT lattice was not possible. This might be explained with the
and local crystal structure: typical relaxor dielectric behaviour fact that, unlike other perovskite systems used in solid solutions
is observed, and both Raman spectra and XRD data indicate a with BaTiO3 , BiGdO3 does not exist as a pure end member with
change from tetragonal to cubic symmetry starting at x = 0.08, a perovskite structure. However, Su and Virkar [20] in their study
where these two phases coexist (see Fig. 1b); the structure is of the Bi2 O3 –Gd2 O3 phase diagram reported the existence of a
referred to as pseudo-cubic, since it has been demonstrated that rhombohedral single phase of Bi2 O3 –Gd2 O3 when the Gd2 O3
(even if on a time-averaged scale) the position of the Ti4+ cation concentration in Bi2 O3 is between 18 and 30% at room temper-
within the oxygen octahedra is in the centre, this situation is ature. The XRD pattern reported for 15%-Gd2 O3 –75% Bi2 O3
better described as a double-well potential, i.e., at T > TC the ther- present a strong peak at about 28◦ (2θ), and the EDS results indi-
mal energy is sufficient to allow the Ti atom to switch between cate a relative amount of 23.9 ± 0.1% of Gd and 70.5 ± 0.1%
eight degenerate positions, off-centre with respect to the oxy- of Bi, which supports the identification of the parasitic phase
gen framework [18]. This is the reason why the two A1 bands as Bi1.55 Gd0.45 O3 (Fig. 1). Observing EDS results (Table 2a) a
at 260 and 520 cm−1 are still present above TC even though clear trend can be observed, which is consistent with the lattice
they should be prohibited in a true cubic phase by Raman selec- parameters trend: for all compositions up to x = 0.10, the mea-
tion rules based only on symmetry. It is worth noticing that the sured content of Gd always matches the theoretical value within
A1 (TO) mode at 250–270 cm−1 , like – to a less extent – the experimental error (standard deviation), even if it is not possible
A1 , E(TO) mode at 510–520 cm−1 in undoped BaTiO3 shifts to state from EDS if Gd occupy B sites only or also the A site. The
towards higher wavenumbers with Bi3+ addition, as reported by amount of Bi is increasing up to x = 0.06, to remain stable even
Strathdee et al. [11], whereas it shifts towards lower wavenum- for higher concentrations, and it is below the expected values,
bers upon the addition of La3+ as pointed out by Feteira and which corroborates the hypothesis that part of it remains outside
Sinclair [5]. This effect may be due to the lone pair of elec- the grains in the secondary phase. At x = 0.15, the unit cell vol-
trons on the Bi3+ ions. The two modes at 513 and 718 cm−1 ume is not increasing and the values of Gd and Bi in the matrix
shift towards higher wavenumbers as already reported for similar are the same as those for x = 0.10, whereas the secondary phase
perovskite solid solutions [5,11], and therefore may be depend- in the x = 0.15 sample is enriched in Bi. This explains why the
ent on the simple dimensions of the unit cell, regardless of the increase in unit cell volume is smaller than the increase in vol-
nature of the dopants. The appearance of the new mode (no. 3, ume for the Bi–Yb solid solutions with BaTiO3 (Fig. 3) starting
see Fig. 4) at about 830 cm−1 is clearly related to doping on the exactly from x = 0.04 when the first evidence of the secondary
B site, as previously reported for Zr-doped BT (BaZrx Ti1−x O3 ) phase appears, even though Yb is ∼8% smaller than Gd.
 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488 2487

The gradual transformation from prototype ferroelectric to at x = 0.08, consistent with XRD patterns and Raman spectra.
relaxor upon doping is also evident from the temperature Co-doping with Bi and Gd thus first disrupts the long range
dependence of the relative permittivity, Fig. 8: consistent with ordering of ferroelectric BaTiO3 and finally leads to typical
XRD and Raman data, two regimes are identified: the former relaxor behaviour, with a high relative permittivity maximum
up to x = 0.06 where there is no frequency dispersion and TC slightly below room temperature.
decreases at approximately 10 ◦ C/mol, the latter starting with
x = 0.10 showing clear relaxor behaviour. The x = 0.08 com-
Acknowledgments
position presents an intermediate dielectric response, which
reflects the coexistence of tetragonal and pseudocubic phases
The XRD and Raman microscope used for this work were
as previously indicated by XRD analysis. The onset of relaxor
obtained through the Birmingham Science City: Creating and
behaviour is generally ascribed to the disruption of the long
Characterising Next Generation Advanced Materials (West Mid-
range ordering of TiO6 octahedra and the formation of polar
lands Centre for Advanced Materials Project 1) with support
nanoclusters embedded in a paraelectric matrix. In a few cases
from Advantage West Midlands and part funded by the Euro-
this is accompanied by the appearance of core/shell structures,
pean Regional Development Fund. This work was also supported
where an almost undoped BaTiO3 core is surrounded by a
by EPCOS OHG, a group company of the TDK-EPC Corpora-
dopant-enriched shell. This inhomogeneity manifests itself in
tion. Additional funding was provided by the Christian Doppler
the temperature dependence of permittivity as also claimed by
Research Association, Austria, and the Federal Ministry of Sci-
Ogihara and Randall for xBiScO3 –(1 − x)BaTiO3 solid solu-
ence, Research and Economy, Austria.
tions [9]. For 0.08 ≤ x ≤ 0.15, the dielectric response shows
two maxima, one staying at an approximately constant T of
120–130 ◦ C and ascribed to the BaTiO3 core, the other shif- References
ting towards lower T with increasing x, as expected for TC upon
the addition of dopants. The size and relative amounts of these [1] von Hippel A, Breckenridge RG, Chesley FG, Tisza L. High dielectric
constant ceramics. Ind Eng Chem 1946;38(11):1097–109.
polar clusters compared to the matrix may explain why they [2] Powles JG, Jackson W. The measurement of the dielectric properties of
are not observed by X-ray diffraction. TEM micrographs con- high-permittivity materials at centimetre wavelengths. Proc IEE III: Radio
firmed the presence of ferroelectric domains in the inner region Commun Eng 1949;96(43):383–9.
of grains (Fig. 7), whereas no domains are observed towards [3] Von Hippel A. Dielectric materials and applications. Cambridge, US: Tech-
the grain boundaries, which might be explained by the fact nology Press of MIT; 1954.
[4] Cohen R. Origin of ferroelectricity in perovskite oxides. Nature
that dopants diffuse into the outer regions only, forming a shell 1992;358(6382):136–8.
and leaving a ferroelectric core, which is responsible for the [5] Feteira A, Sinclair DC. The influence of nanometric phase separation
dielectric anomaly at ∼130 ◦ C. Raman spectroscopy also cor- on the dielectric and magnetic properties of (1 − x)BaTiO3 –xLaYbO3
roborates these findings: a marked difference between the two (0 ≤ x ≤ 0.60) ceramics. J Mater Chem 2009;19(3):356–9.
groups is shown in Fig. 4. The gradual disappearance of the [6] Ferrarelli MC, Tan CC, Sinclair DC. Ferroelectric, electrical, and
structural properties of Dy and Sc co-doped BaTiO3 . J Mater Chem
mode at 307 cm−1 (which is commonly regarded as indicator of 2011;21(17):6292–9.
long-range ferroelectric ordering in pure BaTiO3 ) matches the [7] Shirane G, Suzuki K. On the phase transition in barium–lead titanate (1). J
transformation from tetragonal to pseudocubic as indicated by Phys Soc Jpn 1951;6(4):274–8.
the decrease in the c/a ratio and from permittivity data. However, [8] Yu H, Ye Z-G. A new relaxor ferroelectric system of
the core regions of undoped BT are not detected by Raman spec- (1 − x)BaTiO3 –xBiAlO3 solid solution. In: 17th IEEE International
Symposium on the Applications of Ferroelectrics ISAF 2008. 2008. p.
troscopy because Raman is a local probe method, and the relative 1–2.
amounts of these regions of tetragonal BT is small compared to [9] Ogihara H, Randall CA, Trolier-McKinstry S. Weakly coupled
the surrounding pseudocubic phase. relaxor behavior of BaTiO3 –BiScO3 ceramics. J Am Ceram Soc
2009;92(1):110–8.
5. Conclusions [10] Datta K, Suard E, Thomas PA. Compositionally driven ferroelectric phase
transition in xBiInO3 –(1 − x)BaTiO3 : a lead-free perovskite-based piezo-
electric material. Appl Phys Lett 2010;96(22):221902.
A study of the (1 − x)BaTiO3 –xBiGdO3 system has been con- [11] Strathdee T, Luisman L, Feteira A, Reichmann K. Ferroelectric-to-
ducted for the first time. Gd can be incorporated into the BaTiO3 relaxor crossover in (1 − x)BaTiO3 –xBiYbO3 (0 ≤ x ≤ 0.08) ceramics. J
matrix up to x = 0.10, as demonstrated by the trend in the lattice Am Ceram Soc 2011;94(8):2292–5.
parameters and EDS results. However, as suggested by Rietveld [12] Shannon RD. Revised effective ionic radii and systematic studies of
interatomic distances in halides and chalcogenides. Acta Crystallogr
refinements, BSE images and EDS analysis, samples often con- 1976;A32:751.
tain inhomogeneities. Bi and Gd form a stable secondary phase [13] Goldschmidt V. Shrifter Norske Videnskop-Akad; 1926, November.
that fills the space between grains; diffusion of Bi into the [14] Larson AC, Von Dreele RB. General Structure Analysis System
BaTiO3 matrix can be slower than the other two processes and (GSAS). In: Los Alamos National Laboratory Report LAUR, 86. 1994.
incomplete. This inhomogeneity is reflected in the dielectric p. 748.
[15] McCusker LB, Von Dreele RB, Cox DE, Louër D, Scardi P. Rietveld
response, with two maxima suggesting the presence of a resid- refinement guidelines. J Appl Crystallogr 1999;32(1):36–50.
ual core/shell structure where dopants are confined to the outer [16] Schileo G, Luisman L, Feteira A, Deluca M, Reichmann K.
grain regions, as observed directly in TEM. TC decreases with Structure–property relationships in BaTiO3 –BiFeO3 –BiYbO3 ceramics. J
composition, until a ferroelectric-to-relaxor crossover occurs Eur Ceram Soc 2013;33(8):1457–68.
2488 G. Schileo et al. / Journal of the European Ceramic Society 35 (2015) 2479–2488

[17] Shiratori Y, Pithan C, Dornseiffer J, Waser R. Raman scattering studies on [19] Pokorny J, Pasha U, Ben L, Thakur O, Sinclair DC, Reaney IM. Use of
nanocrystalline BaTiO3 . Part II – consolidated polycrystalline ceramics. J Raman spectroscopy to determine the site occupancy of dopants in BaTiO3 .
Raman Spectrosc 2007;38(10):1300–6. J Appl Phys 2011;109(11):114110.
[18] Frey M, Xu Z, Han P, Payne D. The role of interfaces on an apparent [20] Su P, Virkar AV. Subsolidus phase diagram of the Bi2 O3 –Gd2 O3 system
grain size effect on the dielectric properties for ferroelectric barium titanate and the morphology of phase separation. J Am Ceram Soc 1999;82(8):
ceramics. Ferroelectrics 1998;206(1–4):337–53. 2225–32.