S-F. Wang et al.: Liquid-phase sintering and chemical inhomogeneity in the BaTiO3–BaCO3–LiF system
was found later that a very long soak time (≈265 h) was
was then dry-milled and sieved through a 325 mesh
screen. X-ray powder diffraction (XRD) analysis con-
firmed BaLiF3 formation.
necessary to achieve a K ≈ 4000 and an acceptably
low dissipation factor for sintering temperatures
<900 °C.24,25 In a more comprehensive study, Haussonne
et al.26 reported the effect of 1–2 wt% LiF additions on
the densification and dielectric properties of nonstoichio-
metric BaTiO3. They pointed out that with the addition of
excess BaCO3, a pseudocubic perovskite phase formed
during liquid-phase sintering, which was responsible for
the densification behavior and the resulting dielectric
properties. The effect of Ba/Ti ratio on the shrinkage and
densification behavior of LiF-fluxed BaTiO3 was also
studied by Tolino et al.27 They confirmed that the excess
Ba reduced the sintering temperature and increased the
fired bulk density. However, the different shrinkage rates
exhibited by the dilatometric curves were left in question.
Desgardin et al.28 extended the study of the BaTiO3–LiF
system to the BaTiO3–BaLiF3 system. A mechanism in-
volving the formation of a Ba(Ti1−xLix)O3−3xF3x solid
solution was used to explain low-temperature sintering.
Later, Guha et al.29 and Potten et al.30 reported that the
cubic solid solution coexists with Li2TiO3 during liquid-
phase sintering. In contrast, Lin et al.31 reported that
LiTiO2 is present owing to the lower solubility of Ti in
the liquid.
From the aforementioned literature, it is apparent that
BaTiO3 with the addition of LiF can densify at very low
temperatures and lead to dielectrics with acceptable di-
electric properties. However, many observations remain
unexplained. The objective of this investigation was to
classify the following: (i) the role of Li and F on the
densification during liquid phase sintering of BaTiO3; (ii)
the nature of the liquid phase which forms during liquid
phase sintering; (iii) the chemical inhomogeneity within
the microstructure; (iv) the microstructural–dielectric
property relationship.
Various amounts of LiF (0.1–0.2 wt%) and BaLiF3
(3–7 wt%) were mixed/milled with the BaTiO3 powders
for 24 h using a ball mill. Light scattering (Horiba In-
struments, Irvine, CA) measurements indicated that mill-
ing altered only the particle size of the HPB-A powder,
reducing it to an average of 0.7 m. To prepare the
powders for pressing, they were mixed with DuPont
5200 binder using acetone as a solvent, dried, pulverized
using a mortar and pestle, and then sieved through a 120
mesh screen. Disks were prepared by uniaxial pressing at
15000 psi. Green densities differed slightly for the three
powder types, decreasing with increasing surface area:
HPB-A, 3.61 g/cm3 (60% TD); HPB-B, 3.35 g/cm3 (56%
TD); Cabot, 3.17 g/cm3 (53% TD). These disks were
calcined for 2 h at 500 °C, followed by sintering in a
closed crucible at temperatures ranging from 800 to
1150 °C for 2 h. A 6 °C/min heating rate and 3 °C/min
cooling rate were used throughout.
Bulk density measurements were made using Archi-
medes technique. Xylene was used as the liquid medium
for specimens of 90% TD or higher, and distilled water
was used for more porous specimens. The main source of
the experimental error for measuring the high-porosity
sample was originated from the evaporation loss of the
immersion liquid. Water which has higher surface ten-
sion but lower volatility, therefore, was used for speci-
mens with density less than 90% TD For each type of
sample, density results were obtained by averaging the
data of at least three specimens. It was found that the
density measurement was very reproducible and the de-
viation of the results was less than 1%.
Analyses of Li and F content were carried out on
samples before and after sintering. Sintered pellets were
ground with a mortar and pestle into very fine particles.
For Li analysis, ≈0.1 g of each sample was dissolved in
concentrated H2SO4 and heated at 80–90 °C for 48 h.
These solutions were then diluted to 50 ml with H2O and
HCl. Analyses were performed using a direct-current (dc)-
plasma emission spectrometer (Spectrametrics Inc.). As a
check, microwave digestion was also used to dissolve
BaTiO3 in the solution. The results from these two methods
were consistent. Fluorine analysis was performed by boil-
ing 0.1 g of each sample in de-ionized water for 18 h. The
solution was carefully handled to prevent any acid contami-
nation. Any HF formation, which will evaporate imme-
diately, will result in F losses. The solutions were then
diluted into 50 ml with H2O. Quantitative analysis was
performed using a Dionex ion chromatography unit.
X-ray diffraction patterns using Cu K␣ radiation and a
Ni filter (Scintag DMC 105 x-ray diffractometer) were
obtained from the as-fired and polished surfaces of the
sintered samples to determine the effects of LiF additions
II. EXPERIMENTAL PROCEDURE
Three types of high-purity BaTiO3 (HPB) with varying
particle sizes were used in this study, including two
TAM-Ticon (TAM Ceramics, Niagara Falls, NY) HPB
powders (denoted HPB-A, Ba/Ti ס
0.999, 0.8 m,
2.96 m2/g; HPB-B, and Ba/Ti ס
1.000, 0.44 m,
5.96 m2/g) and a Cabot (Cabot Corp., Boyertown, PA)
hydrothermal powder (Ba/Ti ס
1.000, 0.13 m,
9.40 m2/g). Two flux systems were used in this study:
LiF–BaCO3 and BaLiF3. Reagent grade LiF, BaCO3, and
BaF2 powder were obtained from Aldrich (Aldrich
Chemical Co., Milwaukee, WI). The BaLiF3 was pre-
pared by mixing stoichiometric amounts of BaF2 and LiF
for 24 h with a vibratory mill. The powders were milled
in methyl alcohol using polyethylene jars and ZrO2 me-
dia. After drying, these powders were transferred to a Pt
crucible, reacted at 850 °C, and quenched. The powder
408
J. Mater. Res., Vol. 15, No. 2, Feb 2000