Cristobalite-Related Phases in the KAlO2-KAlSiO4 System

Article abstract of DOI:10.1006/jssc.1999.8426

A study of the K-rich end of the potassium aluminate-silica system (K_1-xAl_1-xSi_xO₂; 0≤x≤0.25) at temperatures up to 1500°C has confirmed the existence of a cubic cristobalite-type solid solution extending to x≈0.235 at 1500°C. The end member, KAlO₂, transforms on cooling to an orthorhombic KGaO₂-type structure at 531°C (orthorhombic Pbca: a=5.4327(8), b=10.924(2), c=15.469(2) A, Z=16). The room temperature structure has been modeled using chemically restrained refinement of Guinier-Haegg X-ray powder diffraction (XRD) data by the Rietveld method (reduced χ²=0.074, wRp=0.065, Rp=0.049). With progressive substitution of SiO₂ into KAlO₂ the cubic-to-orthorhombic phase transition temperature is lowered. A new cristobalite-related phase field has been identified as existing close to room temperature between x≈0.10 and 0.125. The XRD data for this phase can be indexed to a tetragonal superstructure with a_r=2√2a_p and c_r=2c_p (r=resultant, p=parent). Analysis of K-rich specimens by XRD and differential scanning calorimetry required extreme care to avoid rapid degradation by atmospheric moisture.

Full text of DOI:10.1006/jssc.1999.8426

Journal of Solid State Chemistry 147, 624}630 (1999)
Article ID jssc.1999.8426, available online at http://www.idealibrary.com on
Cristobalite-Related Phases in the KAlO2+KAlSiO4 System
Shamus L. G. Husheer, John G. Thompson, and Alexandra Melnitchenko
Research School of Chemistry, Australian National University, Canberra, ACT 0200, Australia
Received March 29, 1999; in revised form June 4, 1999; accepted June 8, 1999
More detail was subsequently added by these authors and
A study of the K-rich end of the potassium aluminate+silica
system (K1؊xAl1؊xSixO2; 04x40.25) at temperatures up to
15003C has con5rmed the existence of a cubic cristobalite-type
solid solution extending to x+0.235 at 15003C. The end mem-
ber, KAlO2, transforms on cooling to an orthorhombic KGaO2-
type structure at 5313C (orthorhombic Pbca: a ؍ 5.4327(8),
b ؍ 10.924(2), c ؍ 15.469(2) A, Z ؍ 16). The room temperature
structure has been modeled using chemically restrained re5ne-
ment of Guinier-Hagg X-ray powder di4raction (XRD) data by
the Rietveld method (reduced v2 ؍ 0.074, wRp ؍ 0.065, Rp ؍
0.049). With progressive substitution of SiO2 into KAlO2 the
reported in a later review paper (6).
In the present study our focus is on the cristobalite-
related solid solution K Al Si O (04x40.22) at
ꢄ\V ꢄ\V V ꢀ
15003C (6) at the K-rich end of the KAlO }KAlSiO
ꢀ
ꢂ
pseudobinary. In a recent transmission electron microscopy
(TEM) and X-ray powder di!raction (XRD) study of the
analogous NaAlO }NaAlSiO pseudobinary at about
ꢀ
ꢂ
12003C (7) we identi"ed "ve new cristobalite-related
phases at intermediate compositions in what was pre-
viously thought to be a continuous solid solution. On this
cubic-to-orthorhombic phase transition temperature is lowered. basis we considered that closer investigation of the cris-
A new cristobalite-related phase 5eld has been identi5ed as
existing close to room temperature between x+0.10 and 0.125.
The XRD data for this phase can be indexed to a tetragonal
superstructure with ar ؍ 2(2ap and cr ؍ 2cp (r ؍ resultant,
p ؍ parent). Analysis of K-rich specimens by XRD and di4eren-
tial scanning calorimetry required extreme care to avoid rapid
degradation by atmospheric moisture. ( 1999 Academic Press
Key Words: potassium aluminate; potassium aluminosilicate;
cristobalite-type; KGaO2-type; synthesis; Rietveld re5nement;
cristobalite-related superstructure,; solid solution, XRD.
tobalite-related solid solution in the K-containing system
was justi"ed.
KAlO , the end member of the subject &&cubic'' solid
ꢀ
solution between 04x40.22 in the system K Al Si
ꢄ\V ꢄ\V V
O , has been studied in its own right. XRD data were "rst
ꢀ
reported for material synthesized at 16503C by Brownmiller
(8), and Barth (9) identi"ed cubic KAlO as being cris-
ꢀ
tobalite-related. Two more recent thermodynamic studies
(10, 11) identi"ed a reversible phase transition for KAlO at
ꢀ
522}5353C with the higher temperature form having cubic
symmetry and being unquenchable and the lower temper-
ature form having lowered symmetry. Otsubo et al. (10) also
reported the d-spacing for the 220 re#ection as a function of
INTRODUCTION
The ternary system K O}Al O }SiO has been carefully temperature up to 10503C.
ꢀ ꢁ
investigated over many decades because of its fundamental
ꢀ
ꢀ
In a study of KFeO , Pistorius and de Vries (12) stated
ꢀ
importance to mineralogy and silicate ceramics. The equi- that &&KAlO probably has the same structure'' as KGaO ,
librium phase "elds in the pseudobinary KAlSiO }SiO as but this comment does not appear to be based on any
a function of temperature and pressure were established in experimental evidence. While other authors have accepted
ꢀ
ꢀ
ꢂ
ꢀ
the period from the 1920s to the late 1950s and are sum- that the room-temperature form of KAlO has a cris-
ꢀ
marized in Fig. 407 of Ref. (1). While there were several tobalite-related structure with symmetry lower than cubic,
attempts to synthesize a compound of stoichiometry, XRD data for the lower symmetry phase, other than those
K Al SiO (2, 3), it was only more recently that phase rela- of Brownmiller (8) from 1935, and determination of the unit
ꢀ
ꢀ
ꢃ
tionships in the KAlO }KAlSiO pseudobinary were sys- cell or space group symmetry appear not to have been
ꢀ
ꢂ
tematically worked out.
reported.
Consequently the aims of this study are
(i) to reinvestigate the cristobalite-related solid solution
Li et al. (4) concentrated on the K-rich end of the
KAlO }KAlSiO pseudobinary and reported a maximum
solubility of SiO in KAlO of 20 mol% at 9003C. Cook et at the K-rich end of the KAlO }KAlSiO pseudobinary to
al. (5) proposed a tentative equilibrium phase diagram for look for new cristobalite-related phases as observed in the
the KAlO }KAlSiO join as a function of temperature. NaAlO }NaAlSiO pseudobinary,
ꢀ
ꢂ
ꢀ
ꢀ
ꢀ
ꢂ
ꢀ
ꢂ
ꢀ
ꢂ
624
0022-4596/99 $30.00
Copyright ( 1999 by Academic Press
All rights of reproduction in any form reserved.

CRISTOBALITE-RELATED POTASSIUM ALUMINOSILICATES
625
(ii) to determine the unit cell, space group symmetry, and and decomposed in a furnace with temperature ramped to
crystal structure of the end member KAlO at room temper- around 3503C over a 2-h period. The sample, now a low-
ꢀ
ature, and
density crystalline solid, was ground to a "ne powder and
(iii) to de"ne the phase "eld for the low temperature then ramped to 8503C over 3 h. The sample was then
structure as a function of temperature and composition.
reground and annealed at 9003C overnight.
In the second method reactive alumina, produced by
decomposing Al(NO ) )9H O at 3503C overnight, was
SYNTHESIS
ꢁ ꢁ
ꢀ
mixed with a stoichiometric amount of K CO ) 1.5H O,
ꢀ
ꢁ
ꢀ
K
Al Si O Specimens with x'0
ꢄ\V ꢄ\V V ꢀ
"red at 3503C, and then ramped to 8503C over 2 h. The
Potassium aluminosilicate specimens were synthesized sample was reground and annealed at 9003C overnight.
from colloidal silica containing 32 wt% SiO (Ludox AM,
du Pont), potassium nitrate (AnalaR, BDH), and aluminium lized potassium aluminate hydrate (K Al O )3H O), which
The third method required the synthesis of well-crystal-
ꢀ
ꢀ
ꢀ ꢂ
ꢀ
nitrate nonahydrate (AnalaR, BDH) for compositions was prepared according to the method of Allen and Rogers
x"0.025, 0.05, 0.075, 0.1, 0.125, 0.15, 0.175, 0.2, and 0.25. In (13), under an argon atmosphere and using dry ethanol and
each case specimens were prepared by mixing KNO and then diethyl ether to wash the product. The K Al O )
ꢁ
ꢀ ꢀ ꢂ
Al(NO ) ) 9H O in a 1:1 mole ratio, dissolving the mixture 3H O reagent was stored under argon to prevent reaction
ꢁ ꢁ
in a minimum quantity of warm (30}353C) distilled water, with atmospheric CO and H O. Con"rmation of pure
ꢀ
ꢀ
ꢀ
ꢀ
and then adding the stoichiometrically required quantity of K Al O )3H O was obtained by XRD and chemical analy-
colloidal silica with stirring while maintaining the temper- sis, both of which were consistent with the results of Johan-
ꢀ
ꢀ ꢂ
ꢀ
ature between 30 and 353C. The solutions were dehydrated sson (14). Dehydration of K Al O )3H O to generate
ꢀ
ꢀ ꢂ
ꢀ
in an oven at 1103C overnight. The resultant solids were KAlO was performed by ramping to 9003C over a period
ꢀ
transferred to platinum crucibles and then progressively of 6 h with a "nal annealing at 9003C overnight.
heated to higher temperatures to remove water and decom-
Samples prepared by each of these methods were sealed in
pose the nitrate. Samples were heated sequentially at 100, Pt tubes, "red at 15003C overnight, and then quenched in
150, 200, 250, 350, 500, and 7003C, typically for several water. Again, failure of a tube was indicated by absorption
hours or overnight. After each heating the samples were of water upon quenching.
rapidly cooled to room temperature and then remixed in
To test whether cristobalite-related potassium aluminate
a mortar and pestle. One gram of each sample was pressed was truly a stoichiometric single phase, specimens were
into a pellet for "nal overnight annealing of the samples at prepared according to the nitrate synthesis described above
9003C. Cook et al. (5) noted that above about 10003C it was with $10% potassium relative to the formula KAlO .
ꢀ
necessary to perform reactions in sealed vessels to prevent
volatilization of K O.
For each of x"0.1, 0.2, and 0.25 three portions of each
(&50 mg) were sealed in separate 4 mm in o.d. Pt tubes and
CHARACTERIZATION AND DATA COLLECTION
ꢀ
Specimens were examined after each "ring by X-ray pow-
"red at 15003C: One was annealed overnight at this temper- der di!raction (XRD) using a Siemens D5000 di!ractometer
ature and then quenched in water; a second was cooled to with CuKa radiation (j"1.5418 A) or using a Guinier-
12003C, held at this temperature for 16 h, and then quen- Hagg camera with monochromated CuKa radiation
ꢄ
ched in water; and a third was slowly cooled over 4 days (j"1.5406 A) with Si (NBS No. 640) as an internal stan-
from 15003C to room temperature. As for Cook et al. (5) dard. Unit cell dimensions of equilibrium specimens were
failure of the Pt tube during "ring for quenched specimens obtained from Guinier-Hagg data. As the KAlO and
ꢀ
was indicated by absorption of water.
KAlO -rich specimens were extremely hygroscopic it was
ꢀ
For each of x"0.025, 0.05, 0.075, 0.125, 0.15, and 0.175 necessary to open the sealed Pt tubes under argon or dry
small portions of each were sealed in Pt tubes as above, nitrogen in a glove box, seal the Guinier-Hagg specimen
annealed at 15003C for 16 h, and then quenched in water. between two pieces of support tape, and record the "lm with
the camera purged by dry nitrogen. In the air KAlO and
ꢀ
KAlO -rich specimens decomposed within tens of seconds
of exposure to atmospheric moisture.
ꢀ
KAlO End Member
ꢀ
KAlO was prepared by three methods:
(1) decomposition of nitrates, as described above,
The XRD pro"le of room-temperature KAlO used for
Rietveld re"nement was obtained by scanning a Guinier-
ꢀ
ꢀ
(2) solid state reaction of K CO ) 1.5H O and reactive Hagg "lm of the material without Si standard. The resultant
ꢀ
ꢁ
ꢀ
Al O , the method used by Cook et al. (5), and
step size was &0.0183 2h and the extracted pro"le between
193 and 723 2h, spanning 309 re#ections, was used.
Di!erential scanning calorimetry (DSC) data were re-
ꢀ ꢁ
(3) dehydration of K Al O ) 3H O.
In the "rst method a stoichiometric mixture of KNO
and Al(NO ) ) 9H O was ground in a mortar and pestle corded for specimens from x"0.0 to 0.2 using a TA
ꢀ
ꢀ ꢂ
ꢀ
ꢁ
ꢁ ꢁ
ꢀ

626
HUSHEER, THOMPSON, AND MELNITCHENKO
Instruments 2100 di!erential scanning calorimeter. Speci- representative XRD pro"les for the three di!erent phases
mens were transferred from freshly opened Pt tubes to Al between 253 and 46.53 2h juxtaposed.
DSC pans and sealed under dry nitrogen in a glove box.
The specimens quenched from 12003C and slowly cooled
Heating and cooling rates of 203C per minute were used and to room temperature were essentially the same as those
the furnace chamber was purged with dry nitrogen.
quenched from 15003C except that the end member com-
position of the cubic phase, as indicated by the unit cell
ꢁꢁ
RESULTS AND DISCUSSION
dimension, seemed to vary.
Hereafter we refer to the low-temperature form of KAlO
ꢀ
X-Ray Powder Diwraction
as a-KAlO , the high-temperature form as b-KAlO , and
ꢀ
ꢀ
All specimens quenched from 15003C to room temper- their respective solid solutions as a and b . The new phase
ꢁꢁ ꢁꢁ
ature produced sharp Guinier-Hagg "lms. The specimen at that occurs at room temperature between about x"0.10
ꢀ
x"0.25 contained a two-phase mixture of the cubic end and 0.125 is labeled a .
ꢁꢁ
ꢁꢁ
member and the tetragonal phase at x+0.42 reported by
Cook et al. (5). Specimens between x"0.2 and 0.15 com-
Trends in Unit Cell Dimensions
prised the cubic phase and could be indexed to the cubic
ꢁꢁ
ꢁꢁ
Fd3m (15) unit cell. Specimens between x"0.125 and 0.10
The unit cell dimensions for the 15003C quenched speci-
comprised a single phase similar to the cubic phase but mens as a function of composition are presented in Table 1.
with at least "ve additional lines. Specimens between As the XRD data show that these three phases clearly
ꢁꢁ
x"0.0 and 0.05 comprised a single orthorhombic phase, possess a common parent, namely, cristobalite-type with
i.e., the low-temperature form of KAlO and its solid solu- a +7.7 A, it is possible to observe the trends in the parent
ꢀ
ꢀ
tion. The specimen at x"0.075 appeared to consist of the unit cell dimensions across the subject composition range
two phases on either side in equilibrium. Figure 1 shows by scaling the resultant dimensions appropriately. These
normalized unit cell dimensions are plotted as a function of
composition in Fig. 2.
For a there is an overall increase in the unit cell dimen-
ꢁꢁ
ꢀ
sions with increasing silica content, whereas for a and
ꢁꢁ
there is an approximately linear decrease in dimension
with composition to the end member composition for the
b
ꢁꢁ
b phase in the x"0.25 specimen. This allows an estimate
ꢁꢁ
of the end member composition as x+0.235 for the
b phase quenched from 15003C (a"7.7555(6) A). For the
ꢁꢁ
12003C quench series the unit cell dimension for the end
member was smaller at a"7.747(1) A and smaller again
for the specimen slowly cooled to room temperature at
a"7.7368(7) A, suggesting that either the composition
range for b is increased at lower temperatures or that the
ꢁꢁ
cubic unit cell dimension is somehow dependent on thermal
TABLE 1
Re5ned Unit Cell Dimensions
Dimension (A)
Composition x
Phase
a
b
c
0.0
0.025
0.5
0.075
0.1
0.125
0.15
0.175
0.2
a
ꢁꢁ
a
5.4327(8)
5.433(2)
5.444(1)
10.924(2)
10.931(4)
10.938(3)
10.901(9)
15.469(2)
15.451(3)
15.464(7)
15.432(13)
15.381(4)
15.369(2)
ꢁꢁ
a
ꢁꢁ
a
5.444(4)
ꢁꢁ
aꢀ
ꢁꢁ
aꢀ
ꢁꢁ
b
ꢁꢁ
b
ꢁꢁ
b
ꢁꢁ
b
ꢁꢁ
21.767(2)
21.735(3)
7.6799(7)
7.6750(6)
7.6659(7)
7.6555(6)
FIG. 1. X-ray powder di!raction pro"les between 25 and 46.53 2h for
x"0.0, x"0.10, and x"0.25 in the system K Al Si O with trunc-
ꢄ\V ꢄ\V
V
ꢀ
ation of the +220,* (p"parent) re#ections at &32.83 2h. The satellite
ꢀ
re#ections in the x"0.10 pro"le are indexed according to the resultant
unit cell. The three pro"les x"0.0, x"0.10, and x"0.15 are representa-
tive of the three phase "elds a , aꢀ , and b , respectively.
0.25
ꢁꢁ ꢁꢁ
ꢁꢁ

CRISTOBALITE-RELATED POTASSIUM ALUMINOSILICATES
627
a cubic unit cell with Fd3m symmetry and at least "ve
additional lines. The lack of any splitting of the parent unit
cell lines and the small number of additional lines initially
caused us to be sceptical that we had identi"ed a new phase.
However, careful measurement of the additional lines
showed an exact relationship between the parent and addi-
tional re#ections. Unfortunately electron di!raction analy-
sis of the subject materials using a transmission electron
microscope, which we have used successfully in earlier stud-
ies of sodium aluminosilicates (7, 17) and sodium mag-
nesiosilicates (18) to con"rm unit cell dimensions and space
group symmetry of cristobalite-related oxides, was not pos-
sible due to their reactivity in the atmosphere and suscepti-
bility to electron beam damage.
ꢀ
The XRD data for the a phase could be indexed to
ꢁꢁ
a tetragonal superstructure of a cristobalite parent with
FIG. 2. Plot of the normalized re"ned unit cell dimensions for the
system (K Al Si O , 04x40.25). The data are normalized to their
a "2(2a and c "2c . The re"ned unit cell dimensions
ꢄ
ꢀ
ꢄ
ꢀ
ꢄ\V ꢄ\V
V
ꢀ
common parent, namely, C9 cristobalite-type with a +7.7 A, to facilitate for x"0.10 were a"21.767(2) and c"15.381(4) A. A com-
ꢀ
comparison. For a , a
(᭿); for aꢀ , a
"(2a (᭹), b
ꢂꢃꢄꢅ
"1/(8 a"ꢄc (᭹); and for b , a
ꢂꢃꢄꢅ ꢁꢁ ꢂꢃꢄꢅ
"1/(2 b (᭺), and c
ꢂꢃꢄꢅ
"a (᭹).
"ꢄc
plete listing of the indexed XRD data in terms of this
tetragonal unit cell is given in Table 3. To allow a full
ꢁꢁ ꢂꢃꢄꢅ
"c
ꢁꢁ ꢂꢃꢄꢅ
ꢀ
ꢀ
ꢀ
comparison between the XRD data for a and b we have
ꢁꢁ ꢁꢁ
history. The unit cell dimensions for the 12003C and slow-
cooled samples at the other compositions (x"0.1 and 0.2)
indicated that there was at least some dependence of the
room-temperature unit cell dimension on thermal history.
TABLE 2
a-KAlO2 X-ray Powder Di4raction Data
h k l
I/I
d
d
ꢃ"ꢁ
ꢅ
#!ꢆ#
0 2
0 2
1 2
1 1
1 2
2 0
1 3
2 1
2 0
2 1
1 2
1 3
1 0
2 2
0 4
2 0
2 3
2 3
0 4
1 5
0 0
0 4
1 6
1 2
2 1
1 6
3 3
2 4
4 0
2
3
3
4
4
0
3
1
2
2
5
4
6
2
4
4
0
2
5
1
8
7
4
9
8
5
4
8
0
5
1
5
4.462
3.749
3.086
3.027
2.729
2.716
2.609
2.599
2.563
2.495
2.412
2.383
2.329
2.320
2.231
2.223
2.177
2.096
2.047
2.010
1.934
1.718
1.576
1.570
1.559
1.507
1.495
1.365
1.358
1.222
1.216
4.458
3.742
3.087
3.029
2.730
2.716
2.610
2.599
2.559
2.497
2.412
2.383
2.330
2.321
2.231
2.223
2.178
2.095
2.047
2.011
1.934
1.717
1.577
1.570
1.559
1.507
1.496
1.365
1.358
1.222
1.217
Determination of Unit Cell and Space Group of a-KAlO
ꢀ
5
Careful measurement of Guinier-Hagg "lms of the equi-
librium specimens allowed us to index the XRD lines to an
orthorhombic unit cell with dimensions a"5.4327(8),
b"10.924(2), and c"15.469(2) A, an&0.7% distortion from
a metrically cubic unit cell. The resolution of the ortho-
rhombic line splitting was not observable if the specimen
was insu$ciently annealed or exposed to the atmosphere
even for a few tens of seconds. These data agreed with the
100
48
12
4
5
1
11
7
27
14
9
1
4
1
2
4
14
2
44
22
2
proposition of Pistorius and de Vries (12) that a-KAlO was
ꢀ
probably KGaO -type in that the observed unit cell dimen-
ꢀ
sions and extinction conditions (15) were consistent with the
KGaO -type space group symmetry Pbca. Further con-
ꢀ
"rmation that a-KAlO was indeed KGaO -type was ob-
ꢀ
ꢀ
tained by calculating the XRD pro"le for this unit cell using
the revised fractional coordinates for KGaO listed in Grey
ꢀ
et al. (16) and substituting Al atoms for Ga atoms in the
calculation. The agreement between calculated and ob-
served intensities was remarkably good. A complete listing
of the indexed XRD data is given in Table 2.
3
1
15
2
12
12
ꢀ
Proposed Superstructure for the a Phase
ꢁꢁ
1 2 12
4 4
The Guinier-Hagg "lms for specimens at x"0.10 and
x"0.125 comprised a set of lines that could be indexed to
0

628
HUSHEER, THOMPSON, AND MELNITCHENKO
TABLE 3
1؊xAl1؊xSixO2 (x ؍ 0.10) X-Ray Powder Di4raction Data
It is interesting to note that the modulation wavevector
K
ꢄ
*
12202 is very common among the cristobalite-type oxide
ꢂ
ꢀ
ꢄ
*
derivative structures (see Refs. 19, 21}23), whereas 14202
ꢂ
ꢀ
h k l?
G#q
I/I
d
d
ꢃ"ꢁ
has only been observed once before, in the recently reported
ꢅ
#!ꢆ#
2a cubic superstructure in the system Na Al Si O at
4 0 2
1 1 1
2 2 0#ꢄ (204)
2 2 0
32
1
100
9
4.442
2.858
2.721
2.658
2.529
2.319
2.221
2.094
1.970
1.924
1.765
1.571
1.360
1.301
1.217
4.445
2.858
2.722
2.658
2.530
2.319
2.221
2.093
1.969
1.924
1.765
1.571
1.361
1.301
1.216
ꢀ
ꢀ\V ꢀ\V V ꢂ
ꢄ
*
x+0.55 (7, 21). The modulation wavevector 12242 has no
7 1 2
8 0 0
4 1 5
1 1 6
4 0 6
8 0 4
6 0 6
5 5 6
8 8 0@
12 0 2
12 4 4
16 0 0
16 4 2
16 8 0
ꢆ
ꢀ
ꢂ
precedent. Without the satellite re#ection at 33.653 2h the
1 1 3#ꢄ (224
remaining four re#ections could be indexed to a smaller 2a
cubic superstructure. Nevertheless the re#ection at 33.653 2h
ꢆ
ꢀ
0 0 2#ꢄ (024)
5
ꢂ
3 1 1
33
21
3
2
21
4
83
25
5
is real and the strongest of the satellite re#ections and
2 2 2
requires the adoption of the 2(2a ;2c unit cell.
1 1 3#ꢄ (220)
ꢀ
ꢀ
ꢂ
ꢄ
*
While it is highly speculative, the 12242 modulation
0 2 2#ꢄ (024)
ꢂ
ꢆ
ꢀ
4 0 0
that requires this large supercell may be indirectly asso-
3 3 1
4 2 2
4 4 0
5 3 1
6 2 0
ꢀ
ciated with the composition of the a phase that occurs at or
ꢁꢁ
close to 12.5 mol% SiO ; i.e., one-eighth of the tetrahedral
ꢀ
framework sites are occupied by Si. What is more certain is
that the modulation of the underlying cristobalite-type
structure is displacive in origin rather than compositional
due to the absence of any low-angle ((303 2h) satellite
re#ections (cf. the modulated cristobalite-type sodium
aluminosilicates reported in Ref. 18).
25
? As the resultant tetragonal unit cell is metrically cubic all re#ections can
be indexed in many di!erent ways*only one indexation is given for each.
@ Re#ection position interpolated from "lm without Si internal standard
due to overlap with Si line.
Structure Rexnement of a-KAlO
ꢀ
included the indexed data for x"0.15 from within the
The XRD pro"le of a-KAlO for Rietveld re"nement was
extracted from Guinier-Hagg "lm (j"1.5406 A), as de-
ꢀ
b "eld (Table 4).
ꢁꢁ
Using a modulated structure description (see, for scribed above. The program GSAS (23), which allows soft
example, Refs. 18}22) of this cristobalite-type oxide phase it contraints on interatomic distances, was used for structure
is informative to consider the additional re#ections as satel- re"nement.
lite re#ections of the underlying C9 parent structure. The
The starting model was derived from the revised structure
satellite re#ections can then be reindexed as G $q where of KGaO (16) with the framework cation site occupied by
ꢀ
ꢀ
G is an allowed b-cristobalite-type parent structure re#ec- Al instead of Ga. With the atomic positions "xed at these
ꢀ
tion and q is a modulation wavevector. The XRD data for starting positions the pro"le was re"ned using 21 variables:
x"0.10 are also indexed in this way in Table 3. From this scale, pro"le (pseudo Voigt type 2*GW and LY), back-
we can see that the "ve readily observed satellite re#ections ground (cosine Fourier series No. 2}12 coe$cients), unit cell
(all shown in Fig. 1) can be indexed as "rst-order satellites of dimensions, and isotropic thermal parameters (like atoms
ꢄ
* ꢄ
*
three di!erent modulation wavevectors; 14202 , 12242 , constrained to be equal). The positional parameters were
ꢂ
ꢀ ꢆ
ꢀ
ꢄ
*
and 12202 .
then re"ned with soft constraints on d
1.757$0.01 and 2.869$0.02 A, respectively, to force the
model to be chemically plausible. The "nal re"nement stat-
and d
of
ꢂ
ꢀ
ꢇꢆ}- -}-
TABLE 4
1؊xAl1؊xSixO2 (x ؍ 0.15) X-ray Powder Di4raction Data
ꢀ
istics were Rp"0.049, wRp"0.065, and reduced s "0.07.
K
During preliminary re"nement cycles it became evident
that the parent Fd3m cristobalite-type re#ections were on
a signi"cantly di!erent scale than the satellite re#ections;
i.e., all the parent re#ections were undercalculating and all
the satellite re#ections overcalculating. This was attributed
h k l
I/I
d
d
ꢃ"ꢁ
ꢅ
#!ꢆ#
1 1 1
2 2 0
3 1 1
2 2 2
4 0 0
3 3 1
4 2 2
4 4 0
5 3 1
6 2 0
4 4 4
28
100
18
18
12
3
43
14
3
4.434
2.715
2.316
2.217
1.920
1.762
1.568
1.358
1.298
1.214
1.109
4.432
2.717
2.316
2.216
1.919
1.762
1.568
1.357
1.298
1.214
1.107
to disorder and/or stacking faults in a-KAlO . Discussion of
this phenomen in cristobalite-related oxides can be found in
ꢀ
Section 11 of Ref. (22). For this reason the stronger parent
re#ections were excluded during pro"le re"nement to min-
imize this scaling problem. The above scaling problem can
not be attributed to preferred orientation.
The observed, calculated, and di!erence pro"les for
this model are shown in Fig. 3. The "nal re"ned atomic
14
22

CRISTOBALITE-RELATED POTASSIUM ALUMINOSILICATES
629
FIG. 3. Observed, calculated, and di!erence pro"les from the Rietveld re"nement of the XRD data for a-KAlO between 19 and 723 2h. The observed
ꢀ
data points are shown as # and the calculated pro"le is shown as a continuous line. The four excluded regions correspond to the strong parent re#ections
220 , 400 , 422 , and 440 , respectively.
ꢀ
ꢀ
ꢀ
ꢀ
ꢀ
parameters, which are similar to those reported for KGaO
observation of the b %a phase transition for the x"0.10
ꢁꢁ ꢁꢁ
(16), and bond valence sums (24, 25) for atoms in the re"ned specimen. The latter observation gives further weight to the
ꢀ
model are listed in Table 5.
XRD evidence for the existence of a new cristobalite-related
phase "eld in the subject system with a previously un-
reported structure type.
Combining the DSC and XRD data allows us to propose
a temperature vs composition equilibrium phase diagram
Proposed Temperature-Composition Phase Diagram for
K
Al Si O , 04x40.25
ꢄ\V ꢄ\V V ꢀ
Reproducible thermodynamic events observed in the (Fig. 4) for the system K Al Si O for the K-rich re-
ꢄ\V ꢄ\V V ꢀ
DSC data collected for 15003C quenched specimens from gion 04x40.25. The detail provided in this phase dia-
x"0.0 to 0.20 are summarized in Table 6. The two salient gram is fully consistent with and complementary to the
features are (i) the lowering of the b %a phase transition equilibrium phase diagram for the KAlO }KAlSiO join
ꢁꢁ
ꢁꢁ
ꢀ
ꢂ
with progressive substitution of SiO into KAlO (ii) the
ꢀ
ꢀ
TABLE 6
Di4erential Scanning Calorimetric Data for K1؊xAl1؊xSixO2
TABLE 5
Re5ned Atomic Parameters for a-KAlO2
Heating
Cooling
Atom
x
y
z
100
;
Bond valence sum
*
*ꢁꢃ
Endotherm
(3C)
Energy
(Jg\ꢄ)
Exotherm
(3C)
Energy
(Jg\ꢄ)
Al1
Al2
K1
K2
O1
O2
O3
O4
0.259(1) 0.007(1) 0.188(1)
0.280(1) 0.263(1) 0.063(1)
0.750(1) 0.009(1) 0.067(1)
0.795(1) 0.261(1) 0.190(1)
0.571(2) 0.286(1) 0.017(1)
0.170(1) 0.402(1) 0.107(1)
0.298(1) 0.155(1) 0.146(1)
0.046(1) 0.486(1) 0.278(1)
2.3(4)
2.3(4)
3.8(3)
3.8(3)
2.9(6)
2.9(6)
2.9(6)
2.9(6)
3.03
3.00
1.02
0.87
1.92
2.06
1.93
2.01
Composition x
0.0
541
10.2
5.4
4.0
*
2.5
*
531
10.8
6.9
3.5
*
3.0
*
0.025
0.050
0.075
0.10
440
332, 355
*
430
345, 323
*
224
*
229
*
0.125

630
HUSHEER, THOMPSON, AND MELNITCHENKO
to the rather subtle di!erences between this and b it could
ꢁꢁ
easily have been overlooked in earlier studies.
ACKNOWLEDGMENTS
The authors thank Mr. Marek Schmidt of the Research School of
Physical Sciences and Engineering, ANU, for assistance with preliminary
DTA measurements; Mr. Gordon Lockhart, of the Research School of
Chemistry, ANU, for assistance in collecting the DSC data; Dr. Yasunori
Tabira of the Research School of Chemistry, ANU, for digitizing the
Guinier-Hagg "lms; and Dr. Ray Withers of the Research School of
Chemistry, ANU, for helpful discussions.
REFERENCES
1. E. M. Levin, C. R. Robbins, and H. F. McMurdie, &&Phase Diagrams for
Ceramists,'' Vol. 1. (M. K. Reser, Ed.). The American Ceramic Society,
Columbus, OH, 1964.
2. J. F. Schiarer and N. L. Bowen, Am. J. Sci. 253, 681}746 (1955).
3. A.Yamaguchi, >ogyo Kyokaishi 78, 74}75 (1970).
4. C. Li, A. F. Reid, and S. Saunders, J. Solid State Chem. 3, 614 (1971).
5. L. P. Cook, R. S. Roth, H. S. Parker, and T. Negas, Am. Mineral. 62,
1180}1190 (1977).
FIG. 4. Proposed temperature vs composition phase diagram from
room temperature to 6003C for the K-rich end of the system
K
Al Si O for 04x40.25. Two phase "elds are unshaded.
ꢄ\V ꢄ\V
V
ꢀ
6. R. S. Roth, Adv. Chem. Ser. 186, 391}408 (1980).
7. J. G. Thompson, A. Melnitchenko, S. R. Palethorpe, and R. L. Withers,
J. Solid State Chem. 131, 24 (1997).
proposed by Roth (6). Con"rmation of the proposed phase
"elds could be obtained by high-temperature XRD studies
but, given the very high reactivity of the K-rich specimens
with the atmosphere, achieving such data would be very
challenging.
8. L. T. Brownmiller, Am. J. Sci. 29, 260 (1935).
9. T. F. W. Barth, J. Chem. Phys. 3, 323}325 (1935).
10. Y. Otsubo, K. Yamaguchi, and Y. Kawamura, Nippon Kagaku Zasshi
83, 352}353 (1962).
11. O. I. Arakelyan, Khim. Prakt. Primen. Silik. 63}71 (1960).
12. C. W. F. T. Pistorius and G. F. de Vries, Z. Anorg. Allg. Chem. 395, 119
(1973).
CONCLUSIONS
13. E. T. Allen and H. F. Rogers, Am. Chem. J. 24, 304}318 (1900).
14. G. Johansson, Acta Chem. Scand. 20, 505}515 (1966).
15. &&International Tables for Crystallography,'' (T. Hahn, Ed.), Vol. A, 4th
ed., Kluwer Academic, Dordrecht, 1995.
The equilibrium phase diagram proposed by Roth (6) for
the KAlO }KAlSi O binary join, which corresponds to
ꢀ
ꢀ ꢃ
04x40.67 in our formula K Al Si O , covers the
ꢄ\V ꢄ\V V ꢀ
temperature range 1300}22003C. Our present results are in 16. I. E. Grey, B. F. Hoskins, and I. C. Masden, J. Solid State Chem. 85, 202
(1990).
good agreement with those of Roth (6) in as much as they
17. J. G. Thompson, R. L. Withers, A. K.Whittaker, R. M. Traill, and J. D.
Fitz Gerald, J. Solid State Chem. 104, 59}73 (1993).
18. R. L. Withers, C. Lobo, J. G. Thompson, S. Schmid, and R. Stranger,
overlap. The new details provided by the present study all
relate to temperatures well below 13003C and are restricted
to the cristobalite-related phases that all fall within the
K-rich end of this binary join at 04x40.25.
Acta Crystallogr. Sect. B 53, 203 (1997).
19. R. L. Withers and J. G. Thompson, Acta Crystallogr. Sect. B 49, 614
(1993).
20. J. G. Thompson, R. L. Withers, A. Melnitchenko, and S. R. Palethorpe,
Acta Crystallogr. Sect. B 54, 531}546 (1998).
21. R. L. Withers, J. G. Thompson, A. Melnitchenko, and S. R. Palethorpe,
Acta Crystallogr. Sect. B 54, 547}557 (1998).
The low-temperature form of KAlO , a-KAlO , has the
ꢀ
ꢀ
KGaO -type structure that is in accord with the suggestion
ꢀ
of Pistorius and de Vries (12) in their study of KFeO . a ,
therefore, also has the room-temperature KGaO -type
structure. The high-temperature form of KAlO , b-KAlO , 22. J. G. Thompson, R. L. Withers, S. R. Palethorpe, and A. Melnitchenko,
presumably has the b-cristobalite-type Fd3m structure
ꢀ
ꢁꢁ
ꢀ
ꢀ
ꢀ
J. Solid State Chem. 141, 29}49 (1998).
23. A. C. Larson and R.B. von Dreele, &&GSAS. The General Structure
Analysis System.'' Los Alamos, NM, 1991.
24. I. D. Brown and D. Altermatt, Acta Crystallogr. Sect. B 41, 244
(1985).
based on the observation at room temperature that the
b
phase has this structure. The evidence that we have
ꢁꢁ
presented in this study does point to the existence of a new
ꢀ
cristobalite-related phase that we have called a phase. Due 25. N. E. Brese and M. O'Kee!e, Acta Crystallogr. Sect. B 47, 192 (1991).
ꢁꢁ