Hot-stage transmission electron microscopy study of phase
transformations in hexacelsian (BaAl2Si2O8)
Zhengkui Xu,a) James L. Shull, Jr.,b) and Waltraud M. Kriven
Department of Materials Science and Engineering, University of Illinois at Urbana-Champaign,
Urbana, Illinois 61801
(Received 19 July 2001; accepted 7 March 2002)
Phase transformations in synthetic hexacelsian were investigated by hot-stage
transmission electron microscopy. A second phase transformation from an
orthorhombic to hexagonal structure was identified in the synthetic hexacelsian at
approximately 700 °C. The hexacelsian was found to exhibit a sequence of phase
transformations on heating of hexagonal (P63/mcm)–orthorhombic (Immm)–hexagonal
(P6/mmm). Antiphase domain boundaries, which were observed in P63/mcm and Immm
phases, were absent in the P6/mmm phase. Crystal symmetries of the three phases were
determined by convergent beam electron diffraction, and space group symmetries
were derived by comparison of experimental selected area electron diffraction patterns
with computer-simulated patterns.
I. INTRODUCTION
Recently, BaAl2Si2O8 was identified as one of the materials with displacive phase transformation, which show
potential for use as transformation weakeners.1,2 Experiments with composites of BaAl2Si2O8 and alumina revealed a new mechanism for achieving debonding in
composites. This mechanism, called “reconstructive
transformation toughening,” uses a volume reducing,
reconstructive phase transformation to generate tensile stresses at the interfaces in composites. Excellent debonding behavior was observed in layered
BaAl2Si2O8/Al2O3 composites tested at room temperature and at 850 °C. The most significant advantage of this
new mechanism is that it is relatively insensitive to
changes in temperature. The phase transformation in
hexacelsian BaAl2Si2O8 at approximately 300 °C has an
associated volume change of 0.43% with a decrease in
volume on cooling.1,2 It is this transformation which is of
interest for shear-induced, transformation weakening.1
It has been known for many years that there are four
polymorphs of BaAl2Si2O8: celsian; paracelsian; ␣- and
␤-hexacelsian. Celsian and paracelsian occur naturally as
minerals. Celsian is the thermodynamically stable phase
under ambient conditions. It is a framework aluminosilicate and a member of the feldspar family.3 Paracelsian is
a less common mineral and is believed to be metastable
at all temperatures under ambient pressure. The structure
consists of chains of tetrahedra similar to those found in
a)
Now at Department of Physics and Materials Science, City University of Hong Kong, 83 Tat Chee Ave., Kowloon, Hong Kong.
b)
Now at Siemens Westinghouse, Pittsburgh, PA 15235-5098.
J. Mater. Res., Vol. 17, No. 6, Jun 2002
the feldspar structures, but in paracelsian they are linked
together in a different manner.4 On the other hand,
hexacelsian is the stable high-temperature phase of
BaAl2Si2O8 existing under equilibrium conditions between 1590 °C and the melting point, 1760 °C. However,
hexacelsian persists metastably when cooled below
1590 °C.5–7 Hexacelsian is structurally quite different
from celsian and paracelsian. It is composed of double
sheets of Al2Si2O8 with the barium cations located between the sheets.8 The presence of two polymorphs of
hexacelsian, ␣- and ␤-hexacelsian, was first observed by
Takéuchi using x-ray diffraction (XRD).9 ␣-Hexacelsian
consists of pseudohexagonal Al2Si2O8 sheets, which are
trigonally distorted, with Ba2+ ions between the layers.
The crystals occur in a trilling form made up of three
orthorhombic individuals simulating a hexagonal symmetry. Due to the distortion, the true structure contains
two such layers in a body-centered orthorhombic cell
with a ⳱ 5.3 Å, b ⳱ 9.2 Å, and c ⳱ 15.6 Å. The space
group of ␣-hexacelsian was not specified. At temperatures above 300 °C, it changed into a hexagonal
structure, ␤-hexacelsian, due to a restoration of the distorted structure of the Al2Si2O8 sheets to the ideal hexagonal structure. ␤-Hexacelsian has a space group of
P6/mmm with a ⳱ 5.3 Å and c ⳱ 7.8 Å. This ␣ ↔ ␤
hexacelsian transformation was also indicated by a discontinuous change in the thermal expansion and specific
heat at approximately 300 °C.10
A more detailed investigation of phase transformation
in BaAl2Si2O8 was carried out by Müller11,12 using hotstage transmission electron microscopy (up to approximately 600 °C), combined with energy dispersive x-ray
microanalysis. The phases were identified by electron
© 2002 Materials Research Society
1287
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
diffraction, using bright- and dark-field imaging. Space
groups of the phases were derived from systematic absences observed in the corresponding electron diffraction patterns. Two different polymorphs were found at
room temperature: a hexagonal polymorph with a space
group of P63/mcm (a approximately 5.3 Å and c approximately 15.6 Å); a pseudohexagonal, orthorhombic
polymorph with a space group of Immm (a approximately
5.3 Å, b approximately 9.2 Å, and c approximately
15.6 Å). This is different from what was reported by
Takéuchi earlier.9 However, in Takéuchi’s XRD work,
there were some weak extra reflections remaining unexplained. It should be mentioned that hexacelsian analogs
in other compounds with feldspar compositions such as
CaAl2Si2O8 and SrAl2Si2O8 also have the same space
group of P63/mcm.13,14
Antiphase domains with the displacement vector 㛳 c
were observed in both P63/mcm and Immm hexacelsians,
and twins were only observed in Immm-hexacelsian. These observed, distinct, microstructural features of
P63/mcm- and Immm-hexacelsians are expected in terms
of group-theoretical considerations. No twins can occur
in P63/mcm hexacelsian but antiphase domains are to be
expected, and both twins and antiphase domains are possible in Immm hexacelsian. 15 In addition, energydispersive x-ray microanalysis indicated that the Si/Al
ratio of Immm-hexacelsian is a few percent higher than
that of P63/mcm-hexacelsian which is stoichiometric
about 1:1. The coexistence of the two polymorphs at
room temperature was attributed to observed difference
in chemical composition between P63/mcm- and Immmhexacelsians. When P63/mcm-hexacelsian is heated it
transforms rapidly and reversibly to Immm-hexacelsian.
The temperature of this transformation was found to depend on the chemical composition and cooling history.
Similar composition dependence of the phase transformation behavior in hexacelsian was also reported by Kremenovic et al.,16,17 who also found that the phase
transformation is dependent of ordering of Si and Al
atoms on equivalent crystallographic sites in (Al,Si)O4
tetrahedra. Ordered and disordered hexacelsians were
prepared by different synthesis routes. The phase transformation at approximately 307 °C was only observed in
ordered stoichiometric hexacelsian but not in disordered
nonstoichiometric hexacelsian. However, the crystal
structures of the ␣ (at 25 °C) and ␤ forms (at 362 °C) of
hexacelsians were refined in the trigonal space group P3
(space group No. 147), from synchrotron XRD data. No
space group change was detected during the ␣ ↔ ␤ phase
transformation. The effect of ordering on the phase transformation was also reported recently by Colomban
et al.18 in sol-gel-prepared, pure and lithium-doped hexacelsians by using infrared, Raman, and thermal expansion studies. It should be mentioned that Tabira et al.19
recently observed an extremely strong and characteristic
1288
diffuse intensity distribution perpendicular to the 〈110〉
directions in the high-temperature polymorph (Immm) of
stoichiometric hexacelsian and Cs- and Rb-doped hexacelsian. The diffuse intensity distribution was attributed
to coupled tetrahedral rotation of 〈110〉 columns of
corner-connected (Al,Si)O4 tetrahedra about the [001]
axis. However, they did not find any evidence for the
existence of either the P3̄ structure reported by Kremenovic or the P63/mcm polymorph reported by Müller in
their specimen. In addition, they did not detect the P6/
mmm since their heating experiment stopped at approximately 600 °C.
No further transformation was observed by Müller
during heating experiments up to approximately 600 °C.
P6/mmm hexacelsian with c ⳱ 7.8 Å reported by Takéuchi was not detected in any transmission electron microscopy (TEM) sample at room temperature or on heating.
Since both Immm and P63/mcm are subgroups of P6/mmm,15
Müller proposed that the stable stoichiometric hexacelsian at high temperatures probably has the space group
of P6/mmm with c ⳱ 7.8 Å. This follows from the
occurrence of twins and antiphase domains in Immmhexacelsian on cooling, originated from a phase transformation P6/mmm → Immm. However, the temperature of
this transformation is not known. With further cooling,
twins disappeared and another phase transformation
Immm → P63/mcm occurred. Therefore, the following
phase transformation sequence in hexacelsian was
proposed:
melt → P6/mmm (assumed) → Immm (␣-hexacelsian)
→ P63/mcm (␤-hexacelsian) .
It turns out that three questions still remain. First,
whether there is a second phase transformation in hexacelsian at T > 600 °C? Second, if there is a second transformation, what is the temperature of the second
transformation? Third, what is the stable high temperature phase?
Recently, DSC and dilatometry analysis of hexacelsian
below 1500 °C by Shull1 found not one, but two firstorder transformations: the previously reported ␣- to
␤-hexacelsian transformation at approximately 300 °C
and a second transformation at approximately 690 °C, as
illustrated in Figs. 1 and 2.1 The transformation at approximately 320 °C has an associated volume change of
0.43% with a decrease in volume on cooling. The second
transformation has a negligible volume change and a
small enthalpy of transformation. The very small enthalpy of the transformation is likely the reason that the
transformation at approximately 690 °C has not been previously reported. In keeping with previous conventions,
the phases are identified as ␣- and ␤-, and ␥-hexacelsian.
High-temperature indentation studies revealed that there
was a significant decrease in hardness associated with the
J. Mater. Res., Vol. 17, No. 6, Jun 2002
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
␣- to ␤-hexacelsian transformation on heating and then
there was little change in the hardness of ␤-hexacelsian
as the temperature was increased further (see Fig. 3).
Once transformed to ␥-hexacelsian, however, the hardness began to decrease again with increasing temperature. All of these results support the phase transformation
sequence proposed by Müller.12 However no significant
differences in crystal structure or symmetry could be
readily detected among the three phases by a conventional high-temperature x-ray powder diffraction study.
Only a noticeable discontinuity in the c parameter in
conjunction with ␣ ↔ ␤ transformation was observed.
Shull’s thesis work partly answered the aforementioned
three questions.1 There is a second phase transformation
in hexacelsian at approximately 690 °C detected by DSC,
dilatometry, and indentation analysis. However, whether
there is a real crystal structure or symmetry change at
approximately 690 °C and what is the crystal structure of
high-temperature ␥-hexacelsian still remain unanswered.
FIG. 1. DSC trace of BaAl2Si2O8 hexacelsian taken in the range from
200 up to 1400 °C. It clearly shows the existence of two phase transformations: a sharp ␣- ↔ ␤-hexacelsian transformation at approximately 310 °C and a second transformation with a small enthalpy of
transformation at approximately 700 °C.
FIG 2. Thermal expansion coefficient as a function of temperature for
BaAl2Si2O8 hexacelsian showing two phase transformations: 1st at
approximately 310 °C and 2nd at approximately 700 °C.
In this work, therefore, a systematic investigation of
phase transformations in hexacelsian were carried out
by using hot-stage TEM up to approximately 800 °C.
Crystal symmetries of three phases were to be determined by convergent beam electron diffraction (CBED),
and space group symmetries would be derived by comparison of experimental selected area electron diffraction
(SAED) patterns with corresponding computer-simulated
patterns.
II. EXPERIMENTAL PROCEDURE
Powders of the materials used in this study were prepared using a polymerized complex technique similar to
the Pechini process.20 The precursors were calcined at
various times and temperatures to remove the organics
and to crystallize the desired phases. The powders were
formed into solid bodies by dry pressing. Sintering experiments were conducted in electric furnaces equipped
with MoSi2 heating elements. Samples were typically
heated and cooled at a rate of 5 °C/min. The sintering
temperatures employed were generally between 1400
and 1600 °C. The samples heated at 1550 °C for 24 h
were used for TEM studies of phase transformations.
TEM specimens were obtained by ultrasonically drilling 3-mm disks and polishing to approximately 100-␮m
thick. The central portion of the disks was further reduced to approximately 10 ␮m by dimpling, with final
perforation by an argon-ion beam. TEM studies were
carried out in a Phillips CM-12 electron microscope
(120 kV) (Philips Instruments, Inc., Mahwah, NJ) using a
Gatan double-tilt hot stage. Electron diffraction patterns
(and all other images) were recorded using a Gatan slow
scan CCD camera along a number of zone axes as a
function of temperature. Accurate values of temperature
were difficult to obtain, since the specimens tended to
FIG. 3. High-temperature indentation data of BaAl2Si2O8 hexacelsian
obtained in the range from 20 up to 850 °C.
J. Mater. Res., Vol. 17, No. 6, Jun 2002
1289
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
heat in the electron beam. Reported values of temperature are used as relative indications, since the experiments were carried out under the same conditions of
illumination and similar specimen thickness. Computer
simulations of SAED patterns and real crystal structure
were carried out using “Desktop Microscopist 2.0” (dynamic simulation) and “Crystal Design” programs, respectively. Thickness of samples used for computer
simulations is 100 nm.
III. RESULTS AND DISCUSSION
Two polymorphs of hexacelsian, hexagonal (or ␣) and
orthorhombic (or ␤) hexacelsian, were found in synthetic
polycrystalline samples at room temperature. The coexistence of two polymorphs was previously reported by
Müller also.11,12 Hexagonal and orthorhombic hexacelsians can be easily distinguished by CBED and SAED
patterns when grains were tilted to certain zone axes. One
example is illustrated in Fig. 4, which shows CBED (a–c)
and SAED (d–f) patterns as a function of temperature obtained from a [0001]-oriented hexagonal grain. Figure 4(a)
shows a room-temperature CBED [0001] zone axis pattern in which a six-fold rotation axis perpendicular to the
pattern (or parallel to [0001]) and six mirror planes can
be easily appreciated. This 6 mm diffraction symmetry
indicates that the examined grain has a hexagonal structure.21 The 6 mm diffraction symmetry remained unchanged until when the sample was heated to T > 300 °C,
as shown in Fig. 4(b), which was obtained from the same
region in the same grain but at 350 °C. In this CBED
pattern only a two-fold rotation axis and two mirror
planes, which are perpendicular to each other (as
marked), can be appreciated. This 2 mm diffraction symmetry indicates an orthorhombic structure of the examined grain. The two-fold rotation axis is parallel to the
[001] axis and two mirror planes are parallel to [100] and
[010] axes of the orthorhombic unit cell, respectively, as
expected. The CBED pattern in Fig. 4(b) was also observed in some hexacelsian grains at room temperature,
indicating the coexistence of ␣- and ␤-hexacelsians.
It turns out that the first phase transformation from a
hexagonal to orthorhombic structure in hexacelsian occurred at approximately 300 °C and that the orthorhombic symmetry can be easily distinguished from hexagonal
symmetry by checking the symmetry of CBED [0001]
(or [001]) zone axis pattern. A second phase transformation was detected with further increasing temperature
FIG. 4. [0001] (a–c) CBED and (d–f ) SAED patterns of BaAl2Si2O8 hexacelsian as a function of temperature; (a,d) 20 °C; (b,e) 350 °C; (c,f )
800 °C.
1290
J. Mater. Res., Vol. 17, No. 6, Jun 2002
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
above 700 °C, as shown in Fig. 4(c), which was obtained
from the same oriented grain but at 800 °C. This CBED
pattern exhibits a six-fold rotation axis and six mirror
planes again, similar to Fig. 4(a), indicating that the hightemperature phase also has a hexagonal symmetry.
Therefore, a sequence of phase transformations of hexagonal (␣) → orthorhombic (␤) → hexagonal (␥) hexacelsian on heating was identified by a set of CBED
patterns shown in Figs. 4(a)–4(c), respectively. For the
first time, a second structural phase transformation at
approximately 700 °C in BaAl2Si2O8 hexacelsian was
unambiguously identified by CBED, which confirms
Shull’s DSC, thermal expansion, and indentation results
mentioned earlier.1 Figures 4(d)–4(f) are corresponding
SAED patterns obtained at 20, 350, and 800 °C, respectively. Unlike CBED patterns, these three SAED patterns
look similar to each other except for the intensity distribution of diffraction spots. For a hexagonal structure, the
intensity of the first ring of spots is weaker than that of
the second ring of spots [Figs. 4(d) and 4(f)]. On the
other hand, for an orthorhombic structure, no distinct
difference in the intensity between the two rings of spots
was observed [Fig. 4(e)]. The observed difference in intensity distribution is attributed to the difference in crystal structure. This is confirmed later by computer
simulation of SAED patterns of the three crystal structures. No significant change in lattice parameters among
three phases was detected by SAED patterns, due to the
low accuracy of SAED technique (approximately 1%).
According to Shull’s high-temperature XRD data from
20 to 800 °C, the changes of lattice parameters in terms
of a hexagonal structure with a ⳱ 5.295 Å and c ⳱
7.788 Å, ⌬a and ⌬c are approximately 0.8% and approximately 0.6%, respectively,1 which is beyond the
accuracy of the SAED technique. Since it is hard to obtain high-quality CBED patterns at high temperatures
(>700 °C), the space groups of ␣-, ␤-, and ␥-hexacelsian
were determined by matching experimental SAED patterns with corresponding computer simulated patterns.
Space groups P6 3 /mcm, Immm, and P6/mmm,
proposed by Müller,12 were assigned to ␣-, ␤-, and
␥-hexacelsian for computer simulation of SAED
patterns, respectively. Lattice parameters of three
hexacelsians used in the simulation were obtained
from corresponding SAED patterns, and they are the
following:
␣-hexacelsian: a ⳱ 5.3 Å and c ⳱ 15.6 Å ,
␤-hexacelsian: a ⳱ 5.3Å, b ⳱ 9.2 Å, and c ⳱ 15.6 Å ,
␥-hexacelsian: a ⳱ 5.3 Å and c ⳱ 7.8 Å .
Figures 5–7 are (a) computer-simulated and (b) experimental SAED patterns of four distinct zone axes for P63/
mcm-, Immm-, and P6/mmm-hexacelsians, respectively.
It should be mentioned that the four zone axes of hexagonal and orthorhombic structures have the following
relationships: [0001]h 㛳 [001]o, [0 1 1̄ 0]h 㛳 [010]o,
[2 1̄ 1̄ 0]h 㛳 [100]o, and [1 2̄ 1̄ 3̄]h 㛳 [1 1̄ 1̄]o. As mentioned
earlier, the observed difference in intensity distribution
of diffraction spots in the experimental [0001]h or [001]o
patterns [Figs. 4(d)–4(f)] is attributed to the difference
in crystal structure. This can be clearly seen in the
corresponding simulated patterns in Figs. 5(a), 6(a), and
7(a), which show that the intensity of the first ring of
spots is weaker than that of the second ring for hexagonal
hexacelsian [Figs. 5(a) and 7(a)]. On the other hand, no
distinct difference in the intensity between the two
rings of spots is present in orthorhombic hexacelsian
[Fig. 6(a)]. Besides the intensity distribution difference, extra diffraction spots, 1/2{h0l} and 1/2{0kl},
where h ⳱ l and k ⳱ l, were only present in [010], [100],
and [11̄1̄] SAED patterns of Immm-hexacelsian (Fig. 6) and
not in corresponding ones of P63/mcm- and P6/mmmhexacelsians (Figs. 5 and 7). These extra spots are the
result of a body-centered orthorhombic structure. In addition, from simulated and experimental patterns, the
low-temperature hexagonal can be readily distinguished
from the high-temperature one by checking the [011̄0]
SAED patterns in Figs. 5 and 7. The low-temperature
P6 3 /mcm pattern does not have (0001) spots with
l ⳱ odd number, and the distance between two spots in
(2̄111) and other rows of diffraction spots in P63/mcm is
half of that in P6/mmm. This is due to the doubling of the
c axis length of P63/mcm-hexacelsian over P6/mmmhexacelsian.
It turns out that all experimental patterns match well
with corresponding simulated patterns, indicating that
there is a second phase transformation in hexacelsian and
the high-temperature phase has a P6/mmm structure. This
result confirms the sequence of phase transformations in
hexacelsian proposed by Müller.12 In addition, all microstructure features, such as antiphase domains and twins
previously reported by Müller12 in P63/mcm- and Immmhexacelsians, were also observed in this work. Furthermore, it was found that the antiphase domains changed
their shape and position with increasing temperature and
disappeared when Immm-hexacelsian was transformed to
P6/mmm-hexacelsian at T > 700 °C, as shown in Fig. 8.
It is noted that not only antiphase domains disappeared
when Immm-hexacelsian was transformed to P6/mmm
but twins disappeared also, as illustrated in Fig. 9.
[11̄1̄] SAED patterns in Figs. 9(a) and 9(b) were taken
from single twin I and II, in Immm-hexacelsian, respectively, and that in Fig. 9(c) was taken from the region
covered both twins I and II, thus showing 1/2{0kl} twin
spots along two directions. These twin spots were found
J. Mater. Res., Vol. 17, No. 6, Jun 2002
1291
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
FIG. 5. (a) Computer-simulated and (b) experimental SAED patterns of four distinct zone axes for P63/mcm-hexacelsian.
1292
J. Mater. Res., Vol. 17, No. 6, Jun 2002
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
FIG. 6. (a) Computer-simulated and (b) experimental SAED patterns of four distinct zone axes for Immm-hexacelsian.
J. Mater. Res., Vol. 17, No. 6, Jun 2002
1293
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
FIG. 7. (a) Computer-simulated and (b) experimental SAED patterns of four distinct zone axes for P6/mmm-hexacelsian.
1294
J. Mater. Res., Vol. 17, No. 6, Jun 2002
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
FIG. 8. Bright-field images of BaAl2Si2O8 hexacelsian recorded at (a) 20 °C and (b) 800 °C. Antiphase domains were not observed in P6/mmmhexacelsian as shown in (b).
FIG. 9. SAED patterns of a BaAl2Si2O8 hexacelsian grain recorded at
(a– c) 350 °C and (d) 800 °C.
to disappear when Immm-hexacelsian was transformed to
P6/mmm-hexacelsian at T > 700 °C, as shown in
Fig. 9(d), which was obtained at 800 °C.
Figure 10 shows a schematic diagram showing the
crystallographic relations for the three BaAl2Si2O8 polymorphs on cooling from high temperature and the change
in symmetry elements involved in the phase transformation as previously described by Bärnighausen.15 At T >
700 °C, hexacelsian has an ideal hexagonal structure, as
illustrated in Figs. 11(a) and 11(b), which correspond to
FIG. 10. Schematic diagram of the crystallographic relations for the
three BaAl2Si2O8 hexacelsian polymorphs.
c and a projections of the high-temperature P6/mmm
structure. Figures 11(a) and 11(b) show the ideal hexagonal Al2Si2O8 layer (or two tetrahedron sheets) along the
[001] direction, with Ba2+ ions between the layer (chh ⳱
7.8 Å). As the temperature is decreased to below 700 °C,
J. Mater. Res., Vol. 17, No. 6, Jun 2002
1295
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
hexacelsian undergoes a P6/mmm to Immm phase transformation. It should be mentioned that this hightemperature phase transformation has a negligible
volume change and a small enthalpy of transformation,
as determined by Shull1 using dilatometry and thermal
DSC analysis. In addition, the ratio of bo/co ⳱ 1.736 is
so close to √3, that the transformed orthorhombic structure
can be considered as a pseudohexagonal. As Takéuchi proposed,9 Immm-hexacelsian contains a pseudohexagonal
FIG. 11. High-temperature P6/mmm-hexacelsian structure shown in
projection down the (a) c and (b) a axes. Small open circles are O, and
small solid circles are Si or Al atoms.
Al2Si2O8 layer, which is trigonally distorted [Fig. 12(a)].
Due to the distortion, the true structure contains two
Al2Si2O8 layers in a body-centered orthorhombic cell,
with a doubling of chh, co ⳱ 2chh ⳱ 15.6 Å, as shown in
Fig. 12(b). This illustrates the doubling of c axis caused
by the distortion of tetrahedron sheets in two different
Al2Si2O8 layers. It is believed that this distortion of tetrahedron sheets resulted in formation of antiphase domains with a 1/2co displacement between the domains as
illustrated in Fig. 12(c).
Another consequence of this hexagonal to orthorhombic phase transformation is the formation of mechanical
twins. Both twins and antiphase domains are expected in
an Immm structure in terms of group-theoretical considerations.15 Figure 13 represents the expected twin structures. There are two types of twin planes: One type is the
{110} twin plane, and the other type is the {130} twin
plane, as shown. The presence of twins in Immm and
absence of twins in P6/mmm hexacelsian were confirmed
by [11̄1̄] SAED patterns in Fig. 9. It should be mentioned
that the similar hexagonal to orthorhombic phase transformation accompanying twin formation was also previously reported in (NH4)2SO4, K2SO4, and K2SeO4
crystals.22,23 As temperature is further decreased to
FIG. 12. (a) [001] and (b) [100] projections of Immm-hexacelsian structure and (c) [100] projection of antiphase domain structure.
1296
J. Mater. Res., Vol. 17, No. 6, Jun 2002
Z. Xu et al.: Hot-stage transmission electron microscopy study of phase transformations in hexacelsian (BaAl2Si2O8)
and antiphase domains disappear. The stable hightemperature phase of hexacelsian has an ideal hexagonal
structure with a space group of P6/mmm. This constitutes
the high-temperature prototype structure in the ferroelastic transformation sequence.
ACKNOWLEDGMENTS
The work of J.L. Shull was supported by Air Force
Office of Scientific Research (AFOSR) AASERT Grant
No. F49620-93-1-0562. The work of Z. Xu was supported by the NSF under Grant No. DMR-9972114. This
work was carried out at the Center for Microanalysis of
Materials at the University of Illinois, which is partially
supported by the United States Department of Energy
under Grant No. DEFG02-96-ER45439.
REFERENCES
FIG. 13. [001] projections of (a) {110} and (b) {130} twin structure
in Immm-hexacelsian. (Al, Si, and O atoms are omitted for clarity).
below 300 °C, hexacelsian undergoes an Immm to P63/
mcm phase transformation accompanying a significant
volume reduction and loss of twins. However, antiphase
domains remain since in P63/mcm-hexacelsian Al2Si2O8
layers or double tetrahedron sheets remain distorted
slightly with respect to the ideal hexagonal structure.
IV. CONCLUSIONS
Hot-stage TEM (up to 800 °C) was used to investigate
phase transformations in synthetic hexacelsian (BaAl2Si2O8).
For the first time the second high-temperature phase
transformation at T > 700 °C was identified by CBED
and SAED techniques in the TEM. Accompanying this
orthorhombic to hexagonal transformation, both twins
1. J.L. Shull, A Investigation of Several “Transformation Weakeners” for Ceramic Composite Interphases, Ph.D. Thesis,
W.M. Kriven, advisor, The University of Illinois at UrbanaChampaign (1997).
2. W.M. Kriven, J. Phys. Colloq. 5(C8), C8-101–110 (1995).
3. R.E. Newnham and H.D. Megaw, Acta Crystallogr. 13, 303
(1960).
4. V.V. Bakakin and N.V. Belov, Soviet. Phys. Crystallogr. 5, 826
(1961).
5. H.C. Lin and W.R. Forster, Am. Mineral. 53, 134 (1968).
6. C.E. Semler, A. Lin, and W.R. Forster, J. Am. Ceram. Soc. 52,
679 (1969).
7. C.E. Semler and W.R. Forster, J. Am. Ceram. Soc. 53, 549 (1970).
8. T. Ito, X-ray Studies on Polymorphism (Maruzen: Tokyo, Japan,
1950).
9. Y. Takéuchi, Mineral. J. (Sapporo) 2, 311 (1958).
10. B. Yoshiki and K. Matsumoto, J. Am. Ceram. Soc. 34, 283
(1951).
11. W.F. Müller, in Electron Microscopy in Mineralogy, edited by
H.R. Wenk (Springer-Verlag, Berlin and Heidelberg, Germany,
New York, 1976), pp. 354–360.
12. W.F. Müller, Phys. Chem. Miner. 1, 71 (1977).
13. Y. Takéuchi and G. Donnay, Acta Crystallogr. 12, 465 (1959).
14. J.A. Gard, in Electron Microscopy in Mineralogy, edited by
H.R. Wenk (Springer-Verlag, Berlin and Heidelberg, Germany,
New York, 1976), p. 52.
15. H. Bärnighausen, Acta Crystallogr. A31, Suppl. S3 (1975).
16. A. Kremenovic, P. Norby, R. Dimitrijevic, and V. Dondur, Solid
State Ionics 101, 611 (1997).
17. A. Kremenovic, P. Norby, R. Dimitrijevic, and V. Dondur, Phase
Transitions 68, 587 (1999).
18. P. Colomban, H. Courret, F. Romain, G. Gouadec, and D. Michel,
J. Am. Ceram. Soc. 83, 2974 (2000).
19. Y. Tabira, R.L. Withers, Y. Takéuchi, and F. Marumo, Phys.
Chem. Miner. 27, 194 (2000).
20. M. Pechini, Method of Preparing Lead and Alkaline-Earth Titanates and Niobates and Coating Method Using the Same to Form
a Capacitor, U.S. Patent No. 3330697 (1967).
21. B.F. Buxton, J.A. Eades, J.W. Steeds, and B.M. Rackham, Philos.
Trans. R. Soc. A 281, 171 (1976).
22. A. Sawada, Y. Makita, and Y. Takagi, J. Phys. Soc. Jpn. 41, 174
(1976).
23. S. Shiozaki and A. Sawada, J. Phys. Soc. Jpn. 43, 1314 (1977).
J. Mater. Res., Vol. 17, No. 6, Jun 2002
1297