You are on page 1of 4

J. Am. Ceram. Soc.

, 85 [6] 1581– 84 (2002)


journal
Grain Growth and Twin Formation in 0.74PMN䡠0.26PT

Jay S. Wallace,* Jong-Moo Huh,* John E. Blendell,* and Carol A. Handwerker*


National Institute of Standards and Technology, Gaithersburg, Maryland 20899

The mechanisms controlling normal and exaggerated grain effect of grain-to-grain impingement. Improved understanding of
growth in lead magnesium niobate–lead titanate (PMN–PT) the grain growth mechanisms operating in this system can be used
ceramics have been investigated by varying the PbO-based to define the microstructural evolution paths and processing
liquid-phase volume fraction from 0.03 to 0.6 and sintering parameters for increasing the size of PMN–PT single crystals.
temperature from 900° to 1100°C. There is a transition in
matrix grain growth rate and matrix grain shape with liquid
fraction; samples with liquid volume fractions less than ⬃0.15 II. Experimental Procedure
show relatively small equiaxed grains resulting from grain-to-
grain impingement. Samples with higher liquid fractions show PMN–PT powders were formed from PbO, Nb2O5, MgCO3,
significantly larger, facetted, cube-shaped grains, whose size is and PbTiO3 starting powders using a one-step process.6 Thermo-
independent of liquid fraction, indicating that a surface nucle- gravimetric analysis was performed on each of the constituents to
ation rate mechanism controls growth in this regime. Exagger- assure a precise stoichiometry of 0.74PMN䡠0.26PT (mole fraction)
ated grains were found in the high liquid fraction samples. of the final body. The three oxides were attrition milled for 1 h
Electron backscatter diffraction showed that all of the exag- using 2 mm diameter Al2O3 media and isopropyl alcohol. MgCO3
gerated grains contained 60° 具111典 twins but none of the was dissolved in 3M HNO3 and mixed with the isopropyl alcohol
normal matrix grains contained twins. The reentrant angles in slurry. The slurry was rotary evaporated to dryness and the
the twinned grains give them a growth advantage over un- resulting powder was calcined in air for 0.5 h at 1050°C in a MgO
twinned grains, resulting in a population of exaggerated crucible to decompose the nitrate and form PMN–PT solid
grains. solution. The calcined powder was attrition milled for 1 h to reduce
the PMN–PT particle size and dried. Finally, the powder was dry
mixed with PbO powder to obtain the desired quantity of liquid
I. Introduction phase. The liquid-phase volume fraction, Vᐉ, between 0.03 and 0.6,
was calculated using data for the equilibrium PMN–PbO liquid
composition,7 assuming that PbTiO3 partitions equally in the solid
L EAD MAGNESIUM NIOBATE, Pb(Mg1/3Nb2/3)O3 (PMN), is a rel-
axor ferroelectric that shows excellent piezoelectric response
and is a good candidate for transducers.1 These properties can be
and liquid and that the density of the liquid can be calculated from
the rule of mixtures. All samples were isostatically pressed at 350
enhanced when lead titanate, PbTiO3 (PT), is present in solid MPa and sintered for 2 h in an oxygen atmosphere on MgO setter
solution in amounts below that of the morphotropic phase bound- plates using a double MgO crucible technique to minimize PbO
ary, ⬇0.33 mol fraction.2 However, conventionally sintered poly- loss. Sample cross sections were polished with 0.05 ␮m silica
crystalline PMN–PT ceramic materials exhibit inferior piezoelec- slurry for scanning electron microscopy (SEM) and electron
tric properties compared with single crystals. This is due in part to backscatter diffraction (EBSD). Etching, when needed, was per-
the anisotropy of the piezoelectric properties,3 requiring a high formed with a 5% HNO3 ⫹ 1% HF solution for 2 to 10 s. Grain
degree of alignment to approach single-crystal properties. In mean intercept length (L៮ ) was measured using line intercepts of
addition, polycrystalline materials have pores, cracks, and remnant 100 –300 grains on polished cross sections.8 The relative standard
second phase inclusions, all of which degrade the properties. deviation of L៮ between images of the same specimen was typically
To bridge the performance gap between single-crystal and between 5% and 10%. Individual grains were removed from the
polycrystalline PMN, “in situ ” crystal growth of a single large sintered body by wicking the PbO liquid from the sintered sample
seed4 or from multiple, aligned “template” seeds5 that are larger with a ZrO2 powder bed for 1 h at 1000°C followed by etching for
than the matrix grains has been applied. Initial growth of the 3 h with 3M HNO3, then sonication and filtration.
single-crystal seeds is dominated by growth in the 具111典 direction
until either impingement on large matrix grains occurs or the
growth surfaces become bounded by slow-growing (100) facets.4 III. Results and Discussion
In the case of templated growth,5 normal and exaggerated grain
growth in the matrix lowers the driving force of the template The matrix grains showed a shape transition with liquid frac-
grains, and impingement of growing grains occurs, resulting in a tion. Samples sintered with low liquid fractions, Vᐉ ⬍ 0.15,
further decrease in driving force. These effects create a practical showed fine, equiaxed, space-filling grains (Fig. 1(A)) which
limit for the maximum crystal size achievable. generally did not appear facetted. For these low liquid fractions,
The present work focuses on the investigation of grain growth shape accommodation occurred under the influence of the capil-
mechanisms of PMN–PT grains in a PbO-rich liquid. Measure- lary forces of the wetting liquid and the equilibrium shape of an
ments of growth rate with varying amounts of liquid phase were isolated grain surrounded by liquid was not observed. Samples
used to determine the rate-limiting growth mechanism and the with high liquid fractions, Vᐉ ⬎ 0.15, had facetted, cube-shaped
matrix grains (Fig. 1(B)) which resulted from relieving the
constraint of shape accommodation with increasing liquid fraction.
This cube shape persisted to higher liquid fractions and long
sintering times, indicating that it is the equilibrium grain shape.
M. P. Harmer—contributing editor
Accompanying the change in grain shape with increasing liquid
fraction was a transition to larger grain sizes, by a factor of
approximately 4 (Fig. 2) similar to those found in simulations.9 In
Manuscript No. 187543. Received August 3, 2001; approved March 8, 2002. this regime the grain size for a given set of annealing conditions
*Member, American Ceramic Society. was insensitive to liquid fraction. Since the average separation

1581
1582 Journal of the American Ceramic Society—Wallace et al. Vol. 85, No. 6

Fig. 1. Scanning electron microscope images of sintered PMN–PT samples: (A) rounded matrix grains result from shape accommodation in a sample with
0.1 liquid fraction; (B) facetted, cube-shaped matrix grains in a sample with 0.4 liquid fraction.

between grains, thus the length of the diffusion path, increased as regime, growth is limited by the nucleation of new ledges on
the liquid fraction increased, a smaller average grain size would singular grain facets; once nucleated, the growth ledge moves
have resulted if diffusion were the growth-rate-limiting step.10,11 rapidly to complete the facet.
Therefore, the independence of L៮ with liquid fraction indicates that The apparent temperature dependence of the transition region
diffusion is not the growth-rate-limiting step. Such behavior has between low and high liquid fractions is an artifact of the
been observed in Si3N412,13 and other facetted ceramic systems.11 experimental conditions in which high and low liquid fraction
A further indication that diffusion was not rate limiting was samples were sintered together in a single crucible. Samples with
found in the matrix grain shape. As opposed to a rounded grain high liquid fractions had sufficient liquid to fill all the pore
morphology for rough surfaces and diffusion-controlled growth,14 volume. Thus all liquid–vapor interfaces were flat. For lower
facetted grain shapes at high liquid fractions (Fig. 1(B)) indicate liquid fractions capillary pores with negative radii of curvature
that interface attachment is the growth-limiting step.15 In this existed, lowering the PbO vapor pressure above the liquid. This
activity difference resulted in transport of PbO from the high liquid
fraction to the low liquid fraction samples, with mass gain of up to
30% for low Vᐉ samples and mass loss of 40% for high Vᐉ samples
sintered at 1100°C. With further time at elevated temperature all of
the samples lost PbO by vaporization. Since vapor transport to and
from samples was transient and not readily quantifiable, no attempt
was made to correct for the changing liquid fractions; the value of
volume fraction for matrix grain size transition must be viewed as
approximate.
Accompanying the matrix grain size and shape transitions at
Vᐉ ⬇ 0.15 was the development of facetted exaggerated grains,
i.e., those larger than 2.25 times the average grain size16 (Figs. 3
and 4). The exaggerated grains did not have the cube shape of the
matrix grains and always contained reentrant edges in cross
section. EBSD determined that each exaggerated grain contained
one or more 60° 具111典 twins. The twin boundaries were not straight
and were not necessarily aligned with either of the adjacent crystal
orientations (Fig. 3(B)), indicative of low twin boundary energy.
Twinned matrix grains were not found by EBSD. The existence of
twins in the exaggerated grains, but not the matrix grains, indicates
that the exaggerated grains form a separate grain size population as
a result of a different grain growth mechanism, twin plane
reentrant edge (TPRE) growth.17 When a twin is present, there is
a significantly reduced nucleation barrier for the formation of a
Fig. 2. Mean intercept length of matrix grains was independent of liquid stable new ledge at the twin site, compared with nucleation of a
fraction for Vᐉ ⬎ 0.2, indicating that growth is not diffusion controlled. For stable ledge on a singular interface, resulting in a growth advan-
Vᐉ ⬍ 0.15 impingement limits grain size. Note: the curves are drawn to aid tage for the twinned grain. This grain growth behavior is well
in visualizing the data groups. known for BaTiO318,19 where it has been attributed to the
June 2002 Grain Growth and Twin Formation in 0.74PMN䡠0.26PT 1583

Fig. 3. Scanning electron microscope images of sintered PMN–PT samples: (A) exaggerated grain with reentrant angles in a sample with 0.4 liquid fraction;
(B) EBSD image of an exaggerated grain and matrix grains. The dark lines within the exaggerated grain are 60° 具111典 twin boundaries.

formation of a double twin. The 60° 具111典 twins in PMN–PT create Several possible mechanisms for twin formation have been
a stable geometric configuration during growth, providing a considered: twinning resulting from mechanical strains, grain
continuing source of ledges as the grains grow and a nearly linear coalescence, and chemical effects. First, since the processing route
growth rate of the exaggerated grains until impingement on other for the powders included mechanical comminution of the prere-
grains of similar or greater size occurs.11,13,20 acted PMN–PT powders, mechanical twinning of some fraction of
the precursor grains could result in twinned “nuclei” for the
exaggerated grains. However, when PMN–PT was made from
starting oxides, PbO, MgCO3, Nb2O5, and TiO2 without a milling
step after the constituents were reacted and PMN formed, twinned
exaggerated grains were also found. Therefore, the twins do not
appear to form as a result of milling.
Coalescence of facetted grains in systems with an incompletely
wetting liquid to form twinned grains or grains with low-angle
boundaries has also been proposed,21,22 particularly for BaTiO3.
Calculations of the likelihood of forming single and double twins
based on geometric probabilities of twin formation seem plausible,
depending on the angular acceptance criteria chosen. Such a model
would predict a continuously increasing number of twinned grains
with time with the rate of coalescence highest for the smallest
grain sizes. However, the estimated number of exaggerated grains
per unit volume was nearly constant for samples annealed between
0.5 and 8 h at 1050°C, indicating that exaggerated grains did not
continuously nucleate during annealing. In addition, there is a
nucleation incubation time for formation of exaggerated grains
which depends on the PbTiO3 and MgO concentrations.23 During
this incubation period there can be significant grain growth
without exaggerated grain formation. These observations lead to
the tentative conclusion that nucleation of twins does not occur as
the result of grain coalescence.
Twin formation has also been shown to result from composi-
tional changes.24,25 In MgO, for example, 具111典 twins are stabi-
lized by a thin platelet of Mg(OH)2, formed in the presence of
water vapor, which serves as the template for the twin.26 Twinning
has also been found in crystals that were formed by solidification
from a binary liquid where compositional changes occurred during
cooling, but coalescence did not occur.27 The hypothesis that
compositional changes could affect twinning in this system is
Fig. 4. SEM image of a single exaggerated grain that was extracted from supported by three observations. First, there are compositional
a sintered body. The (100) faces, 60° 具111典 twin and reentrant angle of the variations within the PMN–PT grains as a result of growth
exaggerated grain are evident. conditions. The initially homogeneous PMN–PT solid solution
1584 Journal of the American Ceramic Society—Wallace et al. Vol. 85, No. 6
3
particles dissolve in the PbO liquid phase and precipitate on the S.-E. Park and T. R. Schrout, “Characteristics of Relaxor-Based Piezoelectric
Single Crystals for Ultrasonic Transducers,” IEEE Trans. Ultrason., Ferroelectr.
growing grains. Depending on the partitioning coefficients of Freq. Control, 44 [5] 1140 – 47 (1997).
PMN and PT, the PMN:PT composition of the precipitating 4
A. Kahn, F. A. Meschke, T. Li, A. M. Scotch, H. M. Chan, and M. P. Harmer,
material will be initially different from the grain on which it is “Growth of Pb(Mg1/3Nb2/3)O3–35 mol% PbTiO3 Single Crystals from (111) Sub-
being deposited. Qualitative SEM energy dispersive spectroscopy strates by Seeded Polycrystal Conversion,” J. Am. Ceram. Soc., 82 [11] 2958 – 62
(1999).
measurements show that the Ti content of PMN–PT grains in 5
E. M. Sabolsky, A. R. James, S. Kwon, S. Trolier-McKinstry, and G. L. Messing,
contact with 0.5 liquid fraction is significantly lower than that of “Piezoelectric Properties of 具001典 Textured Pb(Mg1/3Nb2/3)O3–PbTiO3 Ceramics,”
grains where no excess PbO is present. Second, when PMN App. Phys. Lett., 78 [17] 2551–53 (2001).
6
without PT is sintered for similar times, temperatures and liquid K. R. Han and S. Kim, “New Preparation Method of Low-Temperature Sinterable
fractions, exaggerated grains are not found. Finally, exaggerated Pb(Mg1/3Nb2/3)O3 Powder and Its Dielectric Properties,” J. Mater. Sci., 35 [8]
2055–2059 (2000).
grains tend to nucleate in the PMN–PT samples where they contact 7
Z.-G. Ye, P. Tissot, and H. Schmid, “Pseudo-Binary Pb(Mg1/3Nb2/3)O3–PbO
the MgO setter. Similar chemical gradients have been found in Phase Diagram and Crystal Growth of Pb(Mg1/3Nb2/3)O3 [PMN],” Mater. Res. Bull.,
other liquid-phase sintered systems in which facetted grains 25 [6] 739 – 48 (1990).
8
exhibit nucleation of special boundaries and exaggerated grain E. E. Underwood, Quantitative Stereology; p 40. Addison-Wesley, Reading, MA,
1970.
growth.22 Such chemical variations could affect the energetics of 9
H. Matsubara and R. J. Brook, “Computer Simulation of Grain Growth During
stacking fault and twin formation, as in the case of MgO. While Liquid Phase Sintering”; pp. 415–22 in Sintering Technology. Edited by R. M.
these observations indicate that twin formation is affected by German, G. L. Messing, and R. G. Cornwall. Marcel Dekker, New York, 1996.
10
composition, further examination of twinning in the PMN–PT A. J. Ardell, “The Effect of Volume Fraction on Particle Coarsening: Theoretical
Considerations,” Acta Metall., 20 [1] 61–71 (1972).
system is necessary to completely understand its origin. 11
R. Warren and M. B. Waldron, “Microstructural Development During Liquid
Phase Sintering of Cemented Carbides,” Powder Metall., 15 [30] 166 –201 (1972).
12
D.-D. Lee, S.-J. L. Kang, and D. N. Yoon, “Mechanism of Grain Growth and
IV. Summary ␣–␤⬘ Transformation During Liquid-Phase Sintering of ␤⬘-Sialon,” J. Am. Ceram.
Soc., 71 [9] 803– 806 (1988).
13
The grain growth of lead magnesium niobate–lead titanate has J. S. Wallace and J. F. Kelly, “Grain Growth in Si3N4”; pp. 501–505 in Silicon
been investigated by varying the amount of the PbO-based liquid Nitride 93. Edited by M. Hoffmann, P. F. Becher, and G. Petzow. Trans Tech
Publications, Uetikon-Zurich, Switzerland, 1994.
phase present during sintering. A change in the growth kinetics 14
S. Sarain and H. W. Weart, “Factors Affecting the Morphology of an Array of
was observed at Vᐉ ⬇ 0.15. For Vᐉ ⬍ 0.15 the matrix grain shapes Solid Particles in a Liquid Matrix,” Trans. AIME, 233 [11] 1990 –94 (1965).
15
were rounded due to shape accommodation in response to the Y. J. Park, N. M. Huang, and D. Y. Yoon, “Abnormal Growth of Faceted (WC)
capillary pressure. For Vᐉ ⬎ 0.2, the grains were facetted cubes Grains in a (Co) Liquid Matrix,” Metall. Mater. Trans. A, 27A [9] 2809 –19 (1996).
16
C. Wagner, “Theory of Precipitate Change by Redissolution (Ostwald Ripening)”
with growth rates much faster than for Vᐉ ⬍ 0.15 but independent (in German), Z. Elektrochem., 65 [7/8] 581–91 (1961).
of Vᐉ. The matrix growth rate is controlled by a surface nucleation 17
R. Wagner, “On the Growth of Germanium Dendrites,” Acta Metall., 8 [1] 57– 60
mechanism rather than diffusion through the liquid as shown by (1960).
18
the independence of the growth rate on volume fraction liquid and D. F. K. Henning, R. Janssen, and P. J. L. Reynen, “Control of Liquid-Phase-
Enhanced Discontinuous Grain Growth in Barium Titanate,” J. Am. Ceram. Soc., 70
the faceted grain shape. In addition to faceting for Vᐉ ⱖ 0.15, [1] 23–27 (1987).
exaggerated grains (L ⬎ 2.25L៮ ) were observed. All the exagger- 19
M.-K. Kang, Y.-S. Yoo, D.-Y. Kim, and N. M. Hwang, “Growth of BaTiO3 Seed
ated grains contained 60° 具111典 twins and grain surfaces had twin Grains by the Twin-Plane Reentrant Edge Mechanism,” J. Am. Ceram. Soc., 83 [2]
plane reentrant angles. These reentrant angles significantly low- 385–90 (2000).
20
D. S. Buist, B. Jackson, and I. M. Stephenson, W. F. Ford, and J. White, “The
ered the ledge nucleation barrier, giving the twinned grains a Kinetics of Grain Growth in Two-Phase (Solid–Liquid) Systems,” Trans. Br. Ceram.
growth advantage over grains where ledge nucleation can only Soc, 64, 173–209 (1965).
occur on defect-free singular surfaces. The number of exaggerated 21
G. Kästner and R. Wagner, “Nucleation of Twins by Grain Coalescence During
grains was constant with time, indicating that the twin formation the Sintering of BaTiO3 Ceramics,” Philos. Mag. A, 69 [6] 1051–1071 (1994).
22
K. Choi, J.-W. Choi, D.-Y. Kim, and N. M. Hwang, “Effect of Coalescence on
occurs at one time early in the growth process. Exaggerated grains the Grain Coarsening During Liquid-Phase Sintering of TaC–TiC–Ni Cermets,” Acta
were not observed for PMN samples without PbTiO3, suggesting Mater., 48 [12] 3125–29 (2000).
that compositional changes during precipitation from the liquid 23
J.-M. Huh, J. S. Wallace, and J. E., Blendell, unpublished research.
24
may control twin formation. A. Rec̆nik, J. Bruley, W. Mader, D. Kolar, and M. Rühle, “Structural and
Spectroscopic Investigation of (111) Twins in Barium Titanate,” Philos. Mag. B, 70
[5] 1021–34 (1994).
References 25
H. Hattori, “Observation of Twin Boundaries in KCl Crystal,” J. Cryst. Growth,
66 [1] 205–14 (1984).
1 26
S.-E. Park and T. R. Schrout, “Relaxor-Based Ferroelectric Single Crystals for R. R. Cowley, R. L. Segall, R. St. C. Smart, and P. S. Turner, “Growth Twinning
Electromechanical Actuation,” Mater. Res. Innovations, 1, 20–25 (1997). in Magnesium Oxide Smoke Crystals,” Philos. Mag. A, 39 [2] 163–72 (1979).
2 27
S. W. Choi, T. W. Schrout, S. J. Chiang, and A. S. Bhalla, “Dielectric and J. W. Faust and H. F. John, “The Growth of Semiconductor Crystals from
Pyroelectric Properties in the Pb(Mg1/3Nb2/3)O3–PbTiO3 System,” Ferroelectrics, Solution Using the Twin-Plane Reentrant-Edge Mechanism,” J. Phys. Chem. Solids,
100, 29 –38 (1989). 25 [12] 1407–15 (1964). 䡺

You might also like