Thermal evolution of metakaolin geopolymers:

Part 1 Physical evolution

Peter Duxson, Grant C. Lukey, Jannie S.J. van Deventer
*
Department of Chemical and Biomolecular Engineering, The University of Melbourne, Vic. 3010, Australia
Received 16 February 2006; received in revised form 4 September 2006
Abstract
The physical evolution of materials during heating is a critical factor in determining their suitability and performance for applications
ranging from construction to refractories and adhesives. The eect of dierent cations (sodium and potassium) on the physical evolution
of geopolymeric materials derived from metakaolin is investigated for a range of specimens with Si/Al ratios between 1.15 and 2.15. It is
observed that the eect of potassium is to reduce the thermal shrinkage, while thermal shrinkage increases with increasing Si/Al ratio in
the presence of each alkali type. The thermal shrinkage behavior of mixed-alkali specimens is observed to change from a mean of the
sodium and potassium specimens at low Si/Al ratio to behave similarly to sodium specimens at high Si/Al ratios. It is clear from this
investigation that alkali cations only have a signicant eect on thermal shrinkage of geopolymer at low Si/Al ratios (61.65), while both
Si/Al ratio and alkali cation have little eect on the extent of thermal shrinkage at Si/Al P1.65.
2006 Elsevier B.V. All rights reserved.
PACS: 65.60.+a
Keywords: Microstructure; Alkali silicates; Aluminosilicates; Thermal properties
1. Introduction
The alkali aluminosilicate structure of geopolymeric
gels renders them intrinsically re resistant. The potential
of using geopolymers as a re resistant material with
mechanical properties superior to traditional cements has
been known for over two decades [1]. More recently,
geopolymers have been suggested as a low cost castable
ceramic binder [2], with application in some ceramic and
high-technology applications. The thermal properties of
geopolymeric gels have not been fundamentally investi-
gated; therefore the eect of gel composition on thermal
properties is unknown yet critical to the tailor-design of
materials to suit applications. The ultimate objective of
this research resides in predictive response modelling of
geopolymeric structural members under load subjected to
re. Such models exist for cements and other binders,
and require detailed thermophysical and thermomechani-
cal data for their accuracy [3]. Therefore, it is important
to understand rstly the evolution of geopolymer during
thermal exposure, including thermal shrinkage, crystalliza-
tion, thermal conductivity and mechanical strength at ele-
vated temperature.
Initial investigations of the thermal properties of geo-
polymers have explored some of the basic responses of
the material to exposure to elevated temperatures [47].
The literature centers around simple measures of the abil-
ity of geopolymeric materials to resist structural degrada-
tion, as measured by crystallization observed in XRD
diractograms [4,5]. The changes in structure induced by
elevated temperature have also been observed to be minor
0022-3093/$ - see front matter 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2006.09.019
*
Corresponding author. Tel.: +61 3 83446619; fax: +61 3 83444153.
E-mail address: jannie@unimelb.edu.au (J.S.J. van Deventer).
www.elsevier.com/locate/jnoncrysol
Journal of Non-Crystalline Solids 352 (2006) 55415555
by NMR [4,5]. Barbosa and MacKenzie [4] suggest that
the structure of geopolymeric materials is incredibly resil-
ient to exposure to high temperatures. Little to no change
in the XRD and NMR structure was observed up to
1400 C [4]. In contrast, investigation of Na-geopolymers
observed a large extent of linear shrinkage when exposed
to elevated temperature [57]. Two distinct regions of
shrinkage have been observed, beginning at approximately
100 C and 600 C. These regions of physical change are
indicative of structural changes occurring within the
material.
The physical evolution of geopolymers has been charac-
terized recently by a combination of methods including
TMA, DTA, TGA and nitrogen porosimetry [8]. The ther-
mal shrinkage and weight loss of Na-geopolymer was
shown to vary signicantly with Si/Al, porosity and heat-
ing rate. The behavior of Na-geopolymer was characterized
by four regions of thermal shrinkage and weight loss
observed in all specimens. The initial shrinkage observed
by Rahier et al. [7] was linked with a region of capillary
strain resulting from dehydration. The second large charac-
teristic region of shrinkage was linked to structural densi-
cation by viscous sintering [8]. The region of slow
shrinkage and weight loss between the two densication
regions was identied as dehydroxylation and condensa-
tion of silanol and aluminol groups on the surface of the
gel combined with structural relaxation. The systematic
study of specimens with 1.15 6 Si/Al 6 2.15 showed that
increasing the Si/Al ratio decreased the onset temperature
of densication and substantially increased the extent of
densication [8]. Investigation of physical evolution at
heating rates between 1 C min
1
and 20 C min
1
deter-
mined that densication was increased as the heating rate
increased, likely due to entrapment of water in the gel,
which reduces the energy barrier to viscous sintering.
Although alkali is known to aect the structure and
mechanical properties of geopolymers [911], the eect of
alkali has not yet been investigated on any aspect of the
physical evolution of geopolymers synthesized from
metakaolin. As alkali type is one of the most easily varied
compositional constituents of geopolymer, investigation of
the eect of the most common alkali on thermal properties
is critical.
This investigation elucidates the physical evolution of
K-geopolymers with chemical composition KAlO(SiO)
z

5.5H
2
O, where 1.15 6 z 6 2.15, to determine whether the
characteristic behavior of Na-geopolymers investigated pre-
viously applies to geopolymers synthesized with dierent
cations. Dilatometry, DTA and microscopy are used to
relate the dierent extent and regions of thermal shrinkage
exhibited by K-geopolymer to changes in microstructure
during thermal exposure. The thermal shrinkage and
weight loss of geopolymers with chemical composition
Na
y
K
1y
AlO(SiO)
z
5.5H
2
O, where 0 6 y 6 1, are then
compared to determine the eect of dierent alkali cations
and heating rates on the physical evolution and the extent
of thermal shrinkage of geopolymers.
2. Experimental procedure
2.1. Materials
Metakaolin was purchased from Imerys (UK) under the
brand name of Metastar 402. The molar composition of
metakaolin determined by X-ray uorescence (XRF) was
(2.3:1)SiO
2
Al
2
O
3
with small amounts of a high tempera-
ture form of muscovite as an inert impurity. The Brunauer
EmmettTeller (BET) surface area [12] of the metakaolin,
as determined by nitrogen adsorption on a Micromeritics
ASAP2000 instrument, was 12.7 m
2
/g, and the mean parti-
cle size (d
50
) was 1.58 lm.
Alkaline silicate solutions based on three diering ratios
of alkali metal Na/(Na + K) = M (0.0, 0.5 and 1.0) with
composition SiO
2
/M
2
O = R (0.0, 0.5, 1.0, 1.5 and 2.0)
and H
2
O/M
2
O = 11 were prepared by dissolving amor-
phous silica in appropriate alkaline solutions until clear.
Solutions were stored for a minimum of 24 h prior to use
to allow equilibration.
2.2. Geopolymer synthesis
Geopolymer samples were prepared by mechanically
mixing stoichiometric amounts of metakaolin and alkaline
silicate solution to allow Al
2
O
3
/M
2
O = 1 to form a homog-
enous slurry. After 15 min of mechanical mixing the slurry
was vibrated for a further 15 min to remove entrained air
before being transferred to cylindrical polyethylene moulds
and sealed from the atmosphere. Samples were cured in a
laboratory oven at 40 C and ambient pressure for 20 h
before storage at ambient temperatures in sealed vessels
before use in DTA/TGA, dilatometric and Nitrogen adsorp-
tion experiments. Geopolymers synthesized with sodium,
mixed-alkali and potassiumactivating solutions are referred
to as NaNaK- and K-geopolymers, respectively.
2.3. Analytical techniques
Simultaneous DTA and TGA measurements were per-
formed on a PerkinElmer Diamond DTA/TGA with plat-
inum sample crucibles. Experiments were performed
between 25 C and 1050 C at a heating and cooling scan
rate of 10 C/min with a Nitrogen purge rate of 200 mL/
min. TMA measurements were performed on a Perkin
Elmer Diamond TMA using samples with 5 mm diameter
at a constant heating rate of 10 C/min with a nitrogen
purge rate of 200 mL/min. All specimens were run in ran-
dom order, with a control specimen run multiple times
throughout the duration to test reproducibility, which
was found to be approximately 0.2%.
Microstructural analysis was performed using an FEI
XL-30 FEG-SEM. Samples were polished using consecu-
tively ner media, prior to nal preparation using 1 lm
diamond paste on cloth. As geopolymers are intrinsically
non-conductive, samples were coated using a gold/palla-
dium sputter coater to ensure that there was no arching
5542 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
or image instability during micrograph collection. Micro-
graphs of annealed specimens were taken of specimens that
were polished prior to annealing, so that any topographical
changes resulting from thermal exposure could be related
to structural evolution.
3. Results
Fig. 1 shows the thermal shrinkage of K-geopolymers
with 1.15 6 Si/Al 6 2.15. The shrinkage of K-geopolymers
exhibits the same characteristic trends as those of Na-geo-
polymers reported previously [5,7,8]. In particular, the
K1.15 specimen exhibits a remarkably small extent of
shrinkage beyond 300 C, as observed for Na-geopolymer
of the same Si/Al ratio [8]. Therefore, the shrinkage behav-
ior of K-geopolymer specimens can be separated into the
same four characteristic regions outlined in the observa-
tions of Na-geopolymer [8]. Nominal shrinkage and large
fractional weight loss are observed in Region I; Region II
begins with the onset of initial shrinkage at approximately
100 C, lasting until the rate of shrinkage decreases and
evaporation of free water is complete at about 300 C;
Region III is demarcated by gradual weight loss and
shrinkage from dehydroxylation; and Region IV begins
with the onset of densication by viscous sintering. The
onset temperature of Region II for K-geopolymer occurs
at approximately 70 C, 90 C and 115 C for specimens
with Si/Al ratios of 1.15, 1.40 and P1.65 respectively.
Region II shrinkage can be observed from the onset of
shrinkage until about 250300 C. The thermal shrinkage
of the K1.15 specimen in Region III is nominal. A slight
region of expansion observed at approximately 700 C,
which is discussed later in this article, may be related to
expansion as a result of crystallization. The specimen is
observed to undergo only a very small extent of densica-
tion in Region IV (from 700 C to 1000 C). The specimens
with Si/Al P1.40 exhibit almost identical thermal shrink-
age in Region III, before the onset of a signicant extent
of shrinkage during densication in Region IV. The onset
temperature of densication can be clearly observed to
decrease with increasing Si/Al ratio. The overall shrinkage
of the specimens also increases with Si/Al ratio, although
specimens with Si/Al 6 1.65 exhibit a similar extent of den-
sication after heating to 1000 C.
DTA thermograms of K-geopolymer are presented in
Fig. 2. A large endotherm appears from ambient tempera-
ture until approximately 300 C in all specimens. The endo-
therm can be attributed to evaporation of free pore water
and is also observed for Na-geopolymer [8]. The tempera-
ture span of the endotherm increases with decreasing Si/
Al ratio. In addition, distinct minima appear in the endo-
therms at approximately 70 C with increasing Si/Al ratio.
The thermogravimetric data of K-geopolymers are pre-
sented in Fig. 3. The weight loss data indicate that the rate
of water loss increases with Si/Al ratio, which is observed
across the entire temperature range to 1000 C. The upper
bound temperature of weight loss decreases as the Si/Al
ratio of the specimens increases for 1.15 6 Si/Al 6 1.65,
from approximately 750 C to 700 C and 600 C, respec-
tively. However, the temperature of nal weight loss of
specimens with Si/Al P1.65 is similar. The loss of weight
from geopolymer can be assumed to be entirely from water
loss, by either evaporation of free water or condensation of
hydroxyl groups. The thermogravimetric data in Fig. 3
contain two predominant events of evaporation of uncon-
strained pore water (<300 C) and liberation of water by
-20
-16
-12
-8
-4
0
0 200 400 600 800 1000
Temperature C Temperature C
(a)
(b)
(c)
(d)
(e)
(i)
-5
-4
-3
-2
-1
0
0 100 200 300 400
(a)
(b)
(c)
(d)
(e)
(ii)
Fig. 1. Thermal shrinkage of K-geopolymer from (i) ambient to 1000 C and (ii) Regions I and II, with Si/Al ratios of (a) 1.15, (b) 1.40, (c) 1.65, (d) 1.90
and (e) 2.15.
0 200 400 600 800 1000
Temperature (C)

C
m
g
-
1
(e)
(d)
(c)
(b)
(a)
Fig. 2. DTA thermograms of K-geopolymer with Si/Al ratios of (a) 1.15,
(b) 1.40, (c) 1.65, (d) 1.90 and (e) 2.15.
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5543
condensation/polymerization (>300 C). The temperature
range of dehydration can be correlated with the endotherm
in the DTA thermograms in Fig. 2, with the minima in the
endotherms of high Si/Al ratio specimens relating to the
increased rate of dehydration in these specimens (Fig. 3).
Fig. 4 shows the microstructure of K-geopolymer speci-
mens with 1.15 6 Si/Al 6 2.15. The microstructure of K-
geopolymer contains large pores in specimens with Si/
Al 6 1.40. The large pores in the microstructure of the
low Si/Al ratio K-geopolymer specimens may be attributed
to a high degree of gel reorganization during formation,
similar to Na-geopolymer [13]. Specimens with Si/Al P
1.65 exhibit a more homogeneous microstructure, despite
having similar nominal porosity, taken up by water. The
change in microstructure is due to large reductions in the
pore size of these specimens as the Si/Al ratio is increased,
as a result of the dierences in solution chemistry during
0
20
40
60
80
100
0 200 400 600 800 1000
Temperature

C
Si/Al Ratio
Fig. 3. Weight loss of K-geopolymers measured by TGA from ambient to
1000 C. The arrow indicates the increased rate of weight loss with
increasing Si/Al ratios of specimens from 1.15 to 2.15.
Fig. 4. SEM micrographs of K-geopolymer with Si/Al ratios of (a) 1.15, (b) 1.40, (c) 1.65 (d) 1.90 and (e) 2.15.
5544 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
reaction [13]. The change in the pore size distribution
observed in Fig. 4 can be correlated with the large increase
of Youngs modulus of K-geopolymer for Si/Al 6 1.65
observed in mechanical data presented elsewhere [9].
Fig. 5 shows the microstructures of K1.15 geopolymer
after being exposed to temperatures of 300 C, 600 C
and 1000 C. The microstructure of the K1.15 specimen
does not appear to undergo signicant microstructural evo-
lution as a result of exposure to temperatures up to
1000 C, which may be expected from the small extent of
shrinkage observed in Fig. 1. The microstructures of
K1.65 geopolymer specimens after being exposed to the
same temperatures are presented in Fig. 6. The K1.65 spec-
imen undergoes signicant extent of thermal shrinkage
after exposure to 300 C and 1000 C (Fig. 1), and can be
observed to undergo some extent of microstructural evolu-
tion during heating, especially at 1000 C (Fig. 6c). The
K1.65 specimen without heat treatment (Fig. 4c) can be
observed to exhibit a largely homogeneous microstructure.
After heating to 300 C (Fig. 6a), the polished surface of
the specimen can be observed to exhibit a less smooth sur-
face, with numerous small pores and small cracks observed
across the cross-section of the microstructure. The pores
observed after heating to 300 C are in the order of
100 nm in size, while the cracks are similarly wide, with
lengths ranging from several hundred nanometers to sev-
eral microns (Fig. 6a).
After heating to 600 C the K1.65 specimen can be
observed to exhibit a reduced number of cracks in the
microstructure, though the remaining cracks are larger
(Fig. 6b). The small pores readily seen in the microstruc-
ture at 300 C are no longer observed, and the topography
of the gel phase appears smoother. The reduction in the
number of cracks, pores and smooth topography of the
microstructure observed in Fig. 6b implies that the gel
undergoes a signicant level of thermal relaxation and
healing of small cracks, which is consistent with reduction
in surface area and joining of surfaces that occur as a result
of condensation during dehydroxylation [14]. After heating
to 1000 C (Fig. 6c), the microstructure of the K1.65 spec-
imen can be observed to exhibit a more textured surface,
with dark regions several hundred nanometers in size seen
in the gel phase. The exposed edges of large pores appear
smoothed in comparison to the pores in the microstruc-
tures of specimens exposed to lower temperatures, implying
the material has softened and the surface roughness has
been reduced, driven by surface tension. The dark regions
in the gel appear evenly dispersed in the gel phase, but
do not appear within unreacted material. Therefore, the
dark regions suggest a phase separation occurs within the
gel, which may be indicative of nucleated crystallization.
Fig. 7 shows XRD diractograms of the K1.15 and
K1.65 specimens subjected to the same conditions as those
in the micrographs in Figs. 5 and 6. It can be observed
clearly in Fig. 6b that the K1.65 specimen appears amor-
phous at 600 C, but peaks correlating to leucite (KAlSiO
6
)
and kaliophilite (KAlSiO
4
) appear in the specimen exposed
to 1000 C. The appearance of leucite and kaliophilite in
the K1.65 specimen only after exposure to 1000 C
supports the interpretation of the dark regions in the
Fig. 5. SEM micrographs of K1.15 geopolymer after heating to (a) 300 C, (b) 600 C and (c) 1000 C at a constant heating rate of 10 C min
1
before
quenching.
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5545
micrographs in Fig. 7b. Although it can not be observed in
the micrograph presented in Fig. 5c, from Fig. 7a it can be
observed that the K1.15 specimen also develops crystalline
content correlating to kaliophilite after exposure to
1000 C. Crystallization in geopolymeric gel and the identi-
cation and quantity of phases formed at high temperature
are analysed in greater detail in the second part of this arti-
cle by quantitative XRD [15].
Fig. 8 presents comparisons of the thermal shrinkage of
Na-, NaK- and K-geopolymer with 1.15 6 Si/Al 6 2.15.
The specimens generally exhibit thermal shrinkage in the
order Na > NaK > K in Region II, with similar rates of
shrinkage in Region III, while the onset temperature of
Region IV appears to follow K > NaK Na. The
NaK1.15 specimen appears to exhibit thermal shrinkage
that is the mean of the Na1.15 and K1.15 specimens. In
general, the behavior of NaK-specimens tends more to
Na-geopolymer than K-geopolymer as the Si/Al is
increased.
The derivative of the dilatometric data is shown in
Fig. 9, which allows for clearer observation of changes in
the rate of axial shrinkage and the onset temperatures of
each region of thermal shrinkage. The onset temperature
of Region II is known to increase with Si/Al ratio for K-
geopolymer (Fig. 1) and Na-geopolymer [8], but the eect
of alkali on the onset temperature of Region II is yet to
be identied. The onset temperature of Region II is higher
for K-geopolymer than Na-geopolymer at all Si/Al ratios.
Fig. 6. SEM micrographs of K1.65 geopolymer after heating to (a) 300 C, (b) 600 C and (c) 1000 C at a constant heating rate of 10 C min
1
before
quenching.
Fig. 7. XRD diractograms of (a) K1.15 and (b) K1.65 geopolymer at (i) ambient temperature and after annealing at (ii) 300 C, (iii) 600 C and (iv)
1000 C.
5546 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
Also, the rate of shrinkage in Region II is lower for K-geo-
polymer than for Na-specimens (Fig. 9). The NaK-speci-
mens exhibit onset temperatures to Region II that are
similar to K-geopolymer for Si/Al 6 1.40 (Fig. 9a and b),
while specimens with Si/Al P1.65 exhibit onset tempera-
tures more like Na-geopolymer (Fig. 9ce). The rate of
shrinkage observed for the NaK-1.15 specimen in Region
II is the mean of K1.15 and Na1.15 specimens (Fig. 9a).
The rate of shrinkage observed of the NaK-specimens pro-
gressively increases to be similar to the rate of the Na-spec-
imens with increasing Si/Al ratio, for Si/Al 6 1.90
(Fig. 9bd). The NaK2.15 specimen exhibits the greatest
rate of shrinkage compared to the Na- and K-specimens
(Fig. 9e), the peak of which also occurs at a temperature
below that of the Na2.15 specimen. At lower Si/Al ratios
the temperature of peak shrinkage in Region II can be
observed to follow the order K > NaK > Na.
The extent of thermal shrinkage of specimens may be
quantied by separating the dilatometric data into the four
characteristic regions described earlier, and is presented in
Fig. 10. Fig. 10a shows that the thermal shrinkage in
Region II resulting from capillary strain increases in the
order Na > NaK > K. The extent of shrinkage in Region
II also increases slightly with Si/Al ratio. However, the
increase in shrinkage for Na-, NaK-, and K-specimens with
increasing Si/Al ratio is nominal for Si/Al P1.40. A com-
parison of the weight loss of Na-, NaK-, and K-specimens
with Si/Al ratio of 1.65 is presented in Fig. 11, which
-8
-6
-4
-2
0
0 200 400 600 800 1000
Temperature (C)
0 200 400 600 800 1000
Temperature (C)
0 200 400 600 800 1000
Temperature (C)
0 200 400 600 800 1000
Temperature (C)
0 200 400 600 800 1000
Temperature (C)

L
/
L

(
%
)

L
/
L

(
%
)

L
/
L

(
%
)

L
/
L

(
%
)

L
/
L

(
%
)
K
NaK
Na
-12
-10
-8
-6
-4
-2
0
K
NaK
Na
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
K
NaKNa
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
K
NaK
Na
-20
-18
-16
-14
-12
-10
-8
-6
-4
-2
0
K
Na
NaK
Fig. 8. Thermal shrinkage of Na-, NaK- and K-geopolymers with Si/Al ratios of (a) 1.15, (b) 1.40, (c) 1.65, (d) 1.90 and (e) 2.15. Data for Na-geopolymers
taken from [8].
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5547
exhibits characteristics typical of comparisons of specimens
at all Si/Al ratios. The rate of dehydration decreases mar-
ginally in the order of K > NaK > Na.
Fig. 10b indicates that the shrinkage of K-geopolymer
during dehydroxylation in Region III is marginally greater
than that of the Na- and NaK-specimens, which exhibit
similar values across the Si/Al ratios investigated in the
current work, especially at Si/Al ratio of 1.65. Fig. 10c
shows the extent of thermal shrinkage occurring during
densication and viscous sintering in Region IV. The
amount of shrinkage during densication can be observed
to generally increase with Si/Al ratio in all alkali series,
with the exception of the Na1.90 specimen that exhibits a
small degree of thermal expansion prior to 1000 C. At
high Si/Al ratio the extent of densication in NaK- and
K-specimens appears to reach a maximum (Fig. 10c),
whereas the Na2.15 specimen exhibits a large degree of
shrinkage, characterized by a constant rate of shrinkage
at high temperature (>900 C), which is typical of viscous
ow rather than densication. The NaK-specimens gener-
ally exhibit the highest extents of densication for Si/
Al 6 1.90, which correlates with the observations in
Fig. 8. K-geopolymers exhibit a reduced extent of densi-
cation compared to the Na- and NaK-specimens
(Fig. 10), despite the rate of densication in these speci-
mens being greater than for the Na- and NaK-specimens
(Fig. 9).
Fig. 12 shows the linear shrinkage of K1.65 and
NaK1.65 for heating rates of 1, 2, 5, 10 and 20 C min
1
to allow comparison of previous results for the Na1.65
Fig. 9. Derivative thermal shrinkage of ( )Na-, ( )NaK- and ( )K-geopolymer with Si/Al ratios of (a) 1.15, (b) 1.40, (c) 1.65, (d) 1.90 and (e) 2.15.
Data for Na-geopolymers taken from Duxson et al. [8].
5548 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
specimen from the previous study [8]. The extent of thermal
shrinkage reduces with increasing constant heating rate in
Region I and II for both K- and NaK-geopolymer. How-
ever, there is no readily observable correlation between
the nal extent of thermal shrinkage after heating to
1000 C and heating rate. The thermal shrinkage of
NaK-geopolymer is observed to be greater than for K-
geopolymer.
Fig. 13 shows the degree of linear shrinkage observed in
the four regions of thermal shrinkage for Na-, NaK- and
K-geopolymer with Si/Al ratio of 1.65 subjected to dier-
ent constant heating rates. The amount of shrinkage in
Regions I and II decreases slightly with increased heating
rate for each alkali composition (ie. Na-, NaK- and K-geo-
polymer) (Fig. 13a). However, the extent of shrinkage
observed in Region II can be observed to follow
Na > NaK > K, as also observed readily in Fig. 10. The
extent of thermal shrinkage in Region III is similar for
Na- and NaK-specimens and appears to be independent
of heating rate (Fig. 13b). The extent of thermal shrinkage
for the K1.65 specimen is comparatively higher in Region
III, and displays some level of dependence on heating rate,
though there is no clear trend.
Thermal shrinkage observed in Region IV during densi-
cation and sintering increases considerably with increased
heating rate in all specimens (Fig. 13c), though the trend
and absolute extent of densication are dierent in each
specimen. In general, the extent of shrinkage follows the
order NaK > Na > K. However, in Fig. 13c the extent of
densication observed in Region IV for geopolymer speci-
mens heated at 5 C min
1
is similar, but large dierences
in the extent of shrinkage are observed between these spec-
imens at both low and high heating rates. Therefore,
despite the similar appearance of the extent of densication
-14
-12
-10
-8
-6
-4
-2
0
1.15 1.40 1.65 1.90 2.15
Si/Al ratio

-14
-12
-10
-8
-6
-4
-2
0
2
1.15 1.40 1.65 1.90 2.15
Si/Al ratio
-14
-12
-10
-8
-6
-4
-2
0
1.15 1.40 1.65 1.90 2.15
Si/Al ratio
a b
c
Fig. 10. Thermal shrinkage of (j) Na-, (m) NaK-, and () K-geopolymers in (a) Region II, (b) Region III, and (c) Region IV. Data for Na-geopolymers
taken from Duxson et al. [8].
0
20
40
60
80
100
0 200 400 600 800 1000
Temperature C
Fig. 11. Comparison of the weight loss of Na- (thin line), NaK- (dotted
line), and K- (bold line). geopolymers with Si/Al ratios of 1.65.
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5549
for Na1.65 and NaK1.65 specimens at a heating rate of
5 C min
1
, the use of dierent heating rates is able to iden-
tify signicantly dierent densication characteristics of
these specimens, which would not be observed by analysis
at a single heating rate of 5 C min
1
.
4. Discussion
4.1. Physical evolution of K-geopolymer
The increase in the onset temperature of Region II
(Fig. 1) with increasing Si/Al ratio has been linked with
the increase of the Youngs modulus of Na-geopolymers
[8]. The Youngs modulus of K-geopolymers is similar to
Na-geopolymers [9], suggesting the trend of increased onset
temperature of Region II with increasing Si/Al ratio in
Fig. 1 may also be related to structural rigidity. Despite
large compositional dierences, specimens with Si/Al P
1.65 exhibit almost identical thermal shrinkage with respect
to temperature in Region II (up to approximately 300 C).
Furthermore, at the upper temperature of Region II the
K1.40 specimen displays a similar extent of thermal shrink-
age as the specimens with Si/Al P1.65. In comparison to
all other K-geopolymers, the K1.15 specimen clearly exhib-
its a reduced extent of shrinkage in Region II despite hav-
ing the lowest onset temperature (Fig. 1ii).
The trends observed in the DTA thermograms of Fig. 2
have also been observed during the dehydration of Na-geo-
polymer [8], suggesting the characteristics of geopolymer
dehydration are largely independent of alkali. However,
the upper bound temperature of dehydration in K-geopoly-
mers with low Si/Al ratio is lower than for Na-geopolymer.
This is likely to be a reection of the comparatively
decreased ordering of K-geopolymer with low Si/Al ratio
[11]. For instance, the K1.15 specimen exhibits an amor-
phous XRD diractogram after 7-days ageing, while the
Na1.15 specimens are partially crystalline after the same
period of ageing [9]. Therefore, the K1.15 specimen is unli-
kely to be able to retard dehydration compared to Na1.15
specimens, which contain intercrystalline water. Further-
more, the temperature of the minima in the endotherm of
Na-geopolymer appears at approximately 100 C, which
is signicantly higher than that observed in Fig. 2 for K-
geopolymer (i.e. 70 C). The decrease in the temperature
of the minima is most likely a result of the decreased energy
of hydration of K

aq
compared to Na

aq
[16]. Therefore,
water is more easily liberated from the hydration shell of
alkali cations associated with aluminum in K-geopolymer
than Na-geopolymer.
Similar magnitudes of shrinkage and weight loss
observed in the TGA thermograms in Fig. 3 for specimens
with Si/Al P1.65 up to approximately 700 C imply that
the physical and chemical distribution of hydroxyl sites in
these specimens is similar. The specimens with Si/Al 6
! 1.40 lose water over a greater temperature region, indicat-
ing that some of the hydroxyl groups on the surface of
these specimens are more tightly bound than those on cor-
responding specimens at higher Si/Al ratio. The nominal
Si/Al ratio of geopolymer specimens determine the propor-
tion of silanol and aluminol groups on the surface of the
gel, with higher Si/Al ratio specimens containing a greater
proportion of silanol groups. Condensation of silanol or
aluminol groups on the surface of the geopolymeric gel
proceeds according to the following generalized exothermic
reaction:
TOHHOT ! TOT H
2
O 1
where: T is Al or Si. The energy of condensation for tetra-
hedral linkages follows the order SiOAl > SiOSi > Al
OAl [17]. Therefore, it may be expected that compara-
tively higher temperatures will be required to fully dehydr-
oxylate geopolymeric gel containing a greater proportion
of aluminol groups, than those with larger amounts of sil-
anol groups (i.e. lower Si/Al ratio). A greater fraction of
weight loss can be observed to occur during dehydroxyla-
tion (i.e. >250 C) in the specimens with Si/Al 6 1.40, com-
pared to higher Si/Al ratio specimens. The dierences in
the proportion of water liberated during dehydroxylation
-20
-16
-12
-8
-4
0
0 200 400 600 800 1000
Temperature (C)
Temperature (C)
-20
-16
-12
-8
-4
0
0 200 400 600 800 1000
a
b
Fig. 12. Linear shrinkage of (a) K- and (b) NaK-geopolymer with Si/Al
ratios of 1.65 measured at heating rates of 1 C min
1
, 2 C min
1
,
5 C min
1
, 10 C min
1
, and 20 C min
1
. The arrow indicates increasing
constant heating rate.
5550 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
may be indicative of dierences in the porosity, microstruc-
ture and Al/Si ordering of the specimens with varying Si/Al
ratios [13]. Larger proportions of the water in low Si/Al ra-
tio specimens are likely to be a result of more water being
incorporated into the material as hydroxyl groups or inter-
crystalline water in zeolite phases.
The onset of Region IV in Fig. 1 is observed to decrease
with increasing Si/Al ratio, similar to Na-geopolymer [8].
The onset temperature of densication occurs approxi-
mately at 730 C, 780 C, 880 C and 930 C for specimens
with Si/Al of 2.15, 1.90, 1.65 and 1.40, respectively. The
K1.15 specimen is essentially dimensionally stable up to
750 C, beyond which only a small amount of shrinkage
is observed. The minor amount of linear shrinkage
observed in Region IV (ie. viscous sintering) of the K1.15
specimen is an order of magnitude less than that of the
higher Si/Al ratio K-geopolymers, indicating that the
mechanisms of thermal shrinkage responsible for densica-
tion in the higher Si/Al ratio specimens is not signicant in
this specimen. In contrast, the specimens with Si/Al 6 1.65
exhibit remarkably similar extents of densication, despite
signicant dierences in chemical composition and onset
temperature. The magnitude of thermal shrinkage of the
K1.40 specimen is close to the average of the K1.15 and
K1.65 specimens. The large increase in thermal shrinkage
observed in the specimens with 1.40 6 Si/Al 6 1.65 occurs
in Region IV, and indicates that there are large dierences
in the structure of these specimens that only become signif-
icant during viscous sintering, despite the small composi-
tional dierence. Despite this, almost identical thermal
shrinkage behavior is exhibited by K-geopolymers over a
wider compositional range (i.e. 1.65 6 Si/Al 6 2.15),
though there is approximately 150 C dierence in the
onset temperatures observed for these specimens. There-
fore, it appears that composition correlates strongly to
the onset temperature for sintering, while the extent of
shrinkage may be more closely related to microstructure.
Large changes in the microstructure and pore distribution
of Na-geopolymer have been observed in the composition
region 1.65 6 Si/Al 6 2.15 [13], and may provide some idea
to the dierence in structure of the low Si/Al ratio speci-
mens (i.e. K1.15 and K1.40) and specimens with Si/
Al P1.65.
The progressive change in microstructure for Si/
Al 6 1.65 (Fig. 4), correlates with the increase of the initial
temperature of shrinkage at the beginning of Region II
observed in Fig. 1. The similar appearance of the micro-
structures for specimens with Si/Al P1.65 also correlates
with the similar onset temperature to Region II. However,
at high temperatures the K-geopolymers with microstruc-
tures comprised of large pores (K1.15 and K1.40) are sig-
nicantly more resilient to structural densication than
those with small pores (Si/Al P1.65), which is the opposite
trend than observed for the onset temperature of Region II.
0
2
4
6
8
10
12
14
Na NaK K
Na NaK K
Na NaK K
1C min
-1
2C min
-1
5C min
-1
10C min
-1
20C min
-1
1C min
-1
2C min
-1
5C min
-1
10C min
-1
20C min
-1
1C min
-1
2C min
-1
5C min
-1
10C min
-1
20C min
-1
0
2
4
6
8
10
12
14
0
2
4
6
8
10
12
14
Fig. 13. Linear shrinkage of Na-, NaK- and K-geopolymer with Si/Al ratios of 1.65 subjected to dierent heating rates in (a) Regions I and II, (b) Region
III and (c) Region IV. Data for Na1.65 taken from Duxson et al. [8].
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5551
The specimens with homogeneous microstructures exhibit
large extents of shrinkage, implying that the distribution
of porosity in these specimens and microstructural evolu-
tion with temperature may be related to the extent of den-
sication observed.
The pore volume of Na-geopolymer is known to
increase upon heating to 300 C [8], and correlates with
the more open and porous appearance of the microstruc-
ture in Fig. 6a. This suggests that dehydration causes con-
traction of the gel, resulting in aggregation of pores, which
appears as pores that can be observed by SEM, and the
development of cracks where capillary strain forces exceed
the tensile strength of the gel. These observations are con-
sistent with capillary strain causing the shrinkage observed
in Region II and the localized microcracking of the gel to
release stress. The evolution of K-geopolymer microstruc-
ture in Fig. 6 is consistent with the theoretical model for
thermal shrinkage of geopolymers proposed previously
based on nitrogen porosimetry and dilatometric results
[8]. The observation of gel softening in the K1.65 specimen
between 600 C and 1000 C provides evidence to explain
the large extent of shrinkage observed only where the
homogeneous gel microstructure is observed (Fig. 4ce).
In the low Si/Al ratio specimens the skeletal density of
the gel is higher [13] with larger pores observed (Fig. 4a
and b) [13,18]. The skeletal density of the gel in the high
Si/Al ratio specimens is lower, with porosity dispersed
within the gel itself. Therefore, in the low Si/Al ratio spec-
imens the gel is unable to undergo the same extent of cap-
illary strain, structural relaxation and sintering upon
heating as the high Si/Al ratio specimens, and less shrink-
age is observed (Fig. 1).
4.2. The eect of alkali cation on physical evolution
From analysis of K-geopolymer above, it is clear that
the main characteristics of geopolymer evolution during
heating are present regardless of alkali cation (sodium of
potassium). However, the structure and mechanical prop-
erties of geopolymers are known to be subtly aected by
both the presence of dierent alkali cations, and some
extent of mixed-alkali interactions [9,11]. Furthermore,
the eect of alkali cation is known to change with Si/Al
ratio in geopolymers [9].
It is thought that the increase in the onset temperature
of Region II is linked to the improved mechanical proper-
ties of geopolymer with increasing Si/Al ratio. However,
the similarity of the mechanical properties of Na- and K-
geopolymer [9] does not correlate with the dierence in
the onset temperature of Region II in Figs. 8 and 9. Despite
the fact that mechanical properties are likely to play the
major role in determining the onset temperature of Region
II with respect to Si/Al ratio, it is clear from Fig. 9 that the
nature of the alkali cation also aects the onset tempera-
ture (independent of Si/Al ratio). Therefore, the behavior
of the NaK-geopolymer specimens should provide an
insight into the way that alkali aects the thermal
shrinkage.
The behavior of the NaK-specimens in Region II
(Fig. 9) indicates that the role of the alkali cation in ther-
mal shrinkage changes with Si/Al ratio. At low Si/Al
ratios, NaK-geopolymer appears to behave as the average
of the pure Na- and K-specimens, while at higher Si/Al
ratios the behavior appears more heavily inuenced by
sodium than potassium cation. The rates of thermal shrink-
age during Region III (dehydroxylation) are similar for
Na-, NaK- and K-geopolymer specimens at each Si/Al
ratio. Therefore, it appears that the alkali cation does not
play a signicant role in the extent of shrinkage of the
gel during dehydroxylation. As only a small proportion
of weight is lost as a result of dehydroxylation, it would
not be expected that any small dierences in shrinkage dur-
ing the condensation of TOT linkages (T is Al or Si) with
dierent alkali cations would greatly inuence the overall
extent of thermal shrinkage.
The onset temperature and degree of thermal shrinkage
in Region IV (viscous sintering) in Figs. 8 and 9 exhibit the
greatest variation with alkali type. The specimens with Si/
Al ratio of 1.15 are observed to undergo only nominal den-
sication above 700 C (Fig. 8a), especially when compared
to higher Si/Al ratio specimens (Fig. 8be). Clear onset
temperatures of viscous sintering can be observed in
Fig. 9be, indicated by the rapid decrease in the value of
the derivative beyond a temperature of approximately
600800 C. As observed in the previous section, the onset
temperature of densication for K-geopolymer occurs
approximately at 730 C, 780 C, 880 C and 930 C for
specimens with Si/Al ratios of 2.15, 1.90, 1.65 and 1.40,
respectively. The onset temperature of sintering for K-geo-
polymer is clearly higher than the Na- and NaK-geopoly-
mer specimens at all Si/Al P1.40 (Fig. 9). The onset
temperature of densication of Region IV is observed to
decrease with increasing Si/Al ratio for all specimens.
The reduction in the onset temperature to densication
with increasing Si/Al ratio is thought to relate to reduction
in the softening temperature of the specimens at high tem-
perature with increasing Si/Al ratio [8].
While the strength of AlO bonds is weaker than SiO
bonds in the presence of sodium, they are expected to be
stronger when potassium is present as the charge-balancing
cation [19]. Therefore, the onset temperature of densica-
tion of K-geopolymer should increase with decreasing Si/
Al ratio, which is observed in Fig. 8. The same trend of
reducing densication temperature with increasing Si/Al
ratio is observed in Na-geopolymer [8], which should exhi-
bit an increase in the onset temperature of densication
with increasing Si/Al ratio if AlO bond strength was to
dominate. Furthermore, the eect of unreacted material
in Na-geopolymer is thought to be important, leading to
amounts of sodium not associated with aluminum in the
gel (i.e. Al/Na < 1) [8]. The reduction in Al/M ratio is
widely known to reduce the softening temperature of alkali
aluminosilicates [1921] and would explain a reduction in
5552 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
the densication temperature with increasing Si/Al ratio in
Na-geopolymer. Therefore, it may be implied that the eect
of unreacted material on the gel composition has a larger
eect than the changes in strength of AlO bonds in the
presence of dierent alkali cations.
Taking into account the possible eects of unreacted
material and bond strengths in K-geopolymer, the onset
temperature to densication should reduce with increasing
Si/Al ratio, though it may be expected that the K-geopoly-
mer specimens exhibit additional thermal stability from the
increased strength of AlO bonds compared to Na-geo-
polymer, which is observed in Fig. 8. Furthermore, the
amount of unreacted material in K-geopolymers is less
than Na-geopolymer [10], so the eect of the free potassium
on softening may be reduced, which would further increase
the softening temperature. The analysis of the geopolymers
in the current work is unable to distinguish between the
eects of AlO bond character and unreacted material on
the basis of the data presented in the current work, and
is worthy of further investigation.
The onset temperature of Region IV is similar for all
NaK-geopolymers and Na-geopolymers (Fig. 9). Indeed,
the onset temperature of Region IV in NaK-geopolymers
with Si/Al P1.40 (Fig. 9b) is observed to be lower than
that of the Na-specimens. For example, the NaK1.40 spec-
imen begins to densify at a lower temperature (700 C) than
the Na-specimen (740 C), while the K-specimen exhibits
rapid shrinkage only at 800 C (Fig. 9b). The similar or
reduced onset temperature to densication of NaK-speci-
mens is reasonable, given that the mechanism of densica-
tion thought to be responsible for thermal shrinkage in
Region IV is viscous sintering [8], which is preceded by
softening of the gel. Therefore, the reduction in the onset
temperature of densication for NaK-geopolymer below
that of either Na- or K-geopolymers is likely to be the
result of a reduction in the viscosity of NaK-specimens
near the eutectic in the quaternary Na
2
OK
2
OAl
2
O
3

SiO
2
system [2023]. A reduction in the viscosity and soft-
ening temperature of specimens is likely to have a greater
eect on densication temperature than changes in the
bond strength due to the presence of potassium. It is also
apparent from Fig. 8b that the NaK1.40 specimen under-
goes a greater degree of densication than the Na1.40 spec-
imen, indicating that the reduction in the barrier to
densication is reduced greatly in the presence of mixed-
alkali, correlating with a reduction in viscosity.
Sintering in specimens with Si/Al 6 1.65 exhibits a two-
step densication process, as indicated by the two minima
in Fig. 9ce. The maximum rate of densication for these
specimens occurs in the order K > NaK > Na, implying
that the addition of potassium hinders the onset of densi-
cation, but also increases the rate of densication once ini-
tiated. Furthermore, the dierence in the densication
temperature of the Na- and NaK-specimens compared to
the K-geopolymers is observed to decrease with increasing
Si/Al ratio in specimens with Si/Al 6 1.65 from 200 C, to
100 C and <100 C, respectively. The reduction in the dif-
ference of the densication temperatures of Na-, NaK- and
K-geopolymers with increasing Si/Al ratio shows that the
inuence of the alkali cation on the structure and thermal
properties of geopolymer is reduced in comparison to other
factors at Si/Al P1.65. This composition range also
correlates with the dierent microstructures observed for
K-geopolymers in Fig. 4 and has been observed also for
Na-geopolymer [13], implying that the change in micro-
structure and skeletal gel density observed in these speci-
mens with increasing Si/Al ratio is the dominant feature
in relation to determining the thermal shrinkage at high
Si/Al ratio. Therefore, use of dierent alkali cation is of
reduced signicance when formulating geopolymers of high
Si/Al ratio, but signicant improvements in both onset
temperature of densication and overall thermal shrinkage
can be made by use of low Si/Al ratios.
The higher rate of dehydration in K-specimens shown in
Fig. 11 is expected due to the reduced charge density of
potassium cations compared to sodium, which results in a
lower binding energy to water and a reduced temperature
of dehydration [24]. However, given the increased rate of
dehydration (Fig. 11), and the marginally lower Youngs
moduli of potassium specimens [9], it would be expected
that potassium containing specimens shrink more during
dehydration from capillary forces [25]. The pore volume
of geopolymer has been observed to increase with potas-
siumcontent [18]. The increase in pore volume of potassium
containing geopolymer implies that the skeletal density of
the gel increases with potassium content. Therefore, the
order of shrinkage exhibited by geopolymers of dierent
alkali type in the current work may reect larger capillary
forces causing a greater extent of skeletal densication dur-
ing dehydration in specimens with higher sodium content.
The greater extent of shrinkage exhibited by K-geopoly-
mers in Region III (Fig. 10) is a result of the greater tem-
perature range of dehydroxylation of K-geopolymer,
relating to the higher onset temperature of densication
rather than an indication of a greater degree of dehydroxy-
lation (Fig. 8). Nonetheless, Fig. 10b indicates that the
extent of shrinkage observed in Region III is small com-
pared to that observed in Regions II and IV. Furthermore,
the extent of shrinkage is essentially independent of Si/Al
ratio and alkali type. Shrinkage during dehydroxylation
is only likely to be aected by variation in the proportion
of hydroxyl sites in the gel. As such, it is of little surprise
then that the shrinkage of the gel is minimized in the spec-
imens with Si/Al ratio of 1.15, which have the most dense
gel of specimens in the current work and the least number
of hydroxyl groups. Therefore, in order to produce a geo-
polymer with low thermal shrinkage in Region III a dense
skeletal gel appears favorable. Despite some variation in
the extent of shrinkage observed in Region IV with alkali
cation, it is clear that the Si/Al ratio is the dominant factor
in determining the extent of thermal shrinkage of geopoly-
mers. Despite this, all specimens in the current work with
analogous Si/Al ratios have the same nominal porosity
(as water). Therefore, the small but signicant increase in
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5553
the extent of thermal shrinkage observed in Region IV of
geopolymers resulting from dierent alkali cations must
arise from dierences in the response of the gel to thermal
treatment. The eect of alkali cation on the thermal shrink-
age can be more clearly analysed by determining the eect
of dierent heating rates of specimens with dierent alkali
cations.
4.3. Eect of heating rate on physical evolution
It is well known that the gel densication process is
kinetically limited [25], and that the rate of constant heat-
ing in dilatometric experiments has a large eect on the
densication process of Na-geopolymer [8]. The indepen-
dence of thermal shrinkage in Region III from heating rate
in Fig. 12 implies that shrinkage in this region is not kinet-
ically limited and is a function of temperature and cation
alone, and is consistent with previous ndings that dehydr-
oxylation is thermodynamically driven by Si/Al ratio.
The variation in the extent of densication at dierent
heating rates in Fig. 12 may be accounted for by consider-
ation of the underlying processes that are likely to control
the extent of densication. For instance, it has been previ-
ously suggested that water may become entrapped in the
specimen at increased heating rates (either as molecular
water in small pores or as hydroxyl groups), which can
reduce viscosity during sintering and increase densication
[8]. Indeed, the increase in heating rate alone may be su-
cient to reduce the viscosity of the gel and result in
increases in densication with heating rate [25]. The extent
to which densication mechanisms are able to aect densi-
cation will be discussed in greater detail with respect to
experimental evidence of structural evolution in Part 2 of
the work [15]. Nonetheless, at high constant heating rates,
the extent of densication will ultimately be limited by
eradication of porosity, which explains the limit of densi-
cation observed in all specimens in Fig. 13. Therefore, it is
conceivable that changes in material structure resulting
from Si/Al ratio and alkali cation will directly aect the
extent to which and rates where geopolymer densication
in Region IV will vary with constant heating rate.
In Fig. 13 the Na1.65 specimen exhibits a signicant
increase in the extent of shrinkage with increase of the heat-
ing rate from 1 C min
1
to 5 C min
1
, while similar
shrinkage is observed for heating rates greater than
5 C min
1
(Fig. 13c). The densication of the K1.65 spec-
imen increases for heating rates up to 10 C min
1
, before
remaining constant between 10 C min
1
and 20 C min
1
.
A higher value of heating rate above which little or no
change in the amount of densication of the K1.65 speci-
men is observed (compared to Na1.65) may relate to a fas-
ter limiting rate of water diusion from polycondensation
within the gel of the K-specimen compared to the Na-spec-
imen. Therefore, the dierence between the rate of diu-
sion/polycondensation and the rate of constant heating is
reduced, which reduces the increase in driving force for
densication. The NaK1.65 specimen also exhibits an
increase in densication with increasing heating rate,
although the extent of shrinkage is constant for heating
rates 65 C min
1
, and increases for 5 C min
1
6 heating
rate 620 C min
1
. The similarity in the extents of densi-
cation for heating rates 65 C min
1
, implies that the den-
sication processes in the NaK1.65 specimen are faster
than for the K-specimen, though little dierence is
observed at high temperature in Fig. 11. However, higher
rates of densication in mixed-alkali aluminosilicates are
feasible, with transport processes in mixed-alkali specimens
being faster than in pure alkali specimens [26]. Further
studies of the structural evolution of geopolymer as a result
of exposure to elevated temperature, including the eect of
heating rate and annealing, are explored in greater detail in
Part 2 of the work [15].
5. Conclusions
It can be observed in the current work that the thermal
shrinkage of geopolymeric materials derived from metaka-
olin is greatly inuenced by the composition of the alkali
activating solution, both in terms of Si/Al ratio and alkali
content (sodium and potassium). These elements appear to
determine largely the characteristics of thermal shrinkage
of geopolymers. The extent of thermal shrinkage observed
in geopolymers based on dierent alkali compositions
increases rapidly for 1.15 6 Si/Al 6 1.65, and is relatively
constant at higher Si/Al ratios. Thermal shrinkage of all
specimens can be categorized by four characteristic regions
of structural resilience, dehydration, dehydroxylation and
sintering. Microstructural analysis of K-geopolymer shows
that the increase in thermal shrinkage with increasing Si/Al
ratio relates to densication by reduction in porosity dur-
ing dehydroxylation and sintering. Evolution of the micro-
structure is not readily observed for low Si/Al ratio
specimens, which exhibit low thermal shrinkage. However,
the microstructure of a high Si/Al ratio K-geopolymer was
observed to exhibit signicant eects of dehydration, dehy-
droxylation and viscous sintering.
The extent of thermal shrinkage observed in the geopoly-
mers in the current work decreases in the order Na >
NaK > Kfor specimens with Si/Al 6 1.65. In contrast, sim-
ilar extents of thermal shrinkage have been observed for all
alkali series at high Si/Al ratio (ie. Si/Al P1.65). The alkali
cation type has a large eect on the onset temperature of vis-
cous sintering observed in dilatometric data at all Si/Al
ratios, with K > NaK Na. The increase in resilience of
K-geopolymer to densication is thought to be due to
increased strength of AlO bonds in the presence of potas-
sium compared to sodium, which manifests itself as an
increase in the softening temperature of the gel. NaK-speci-
mens are observed to behave more similarly to K-geopoly-
mers at low Si/Al ratio, while they behave similarly to Na-
geopolymer at high Si/Al ratio. The onset temperature of
densication for geopolymers with dierent alkali cation is
observed to reduce with increasing Si/Al ratio. The extent
of densication observed during viscous sintering was
5554 P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555
observed to decrease in the order NaK > Na > K, also
thought to be a result of the increase in the viscosity of these
specimens in the same order. The extent of thermal shrink-
age of geopolymers has been observed to increase with
increasing constant heating rate for Na-, NaK- and K-geo-
polymer specimens with Si/Al ratio of 1.65. The increase in
densication with heating rate was observed to be dependent
on alkali cation, implying that the rates of important pro-
cesses occurring during densication are aected by the pres-
ence of dierent alkali cations.
The use of dierent cations in geopolymers intended for
high temperature applications is only signicant when using
low Si/Al ratio compositions, and demonstrates that the
alkali cation and Si/Al ratio play reduced roles in thermal
shrinkage of geopolymer as the Si/Al ratio increases. The
reduction in the eect of alkali and the dominance of nom-
inal Si/Al ratio and microstructure on the thermal shrink-
age properties of geopolymer, particularly at low Si/Al
ratios, correlates with and validates the structural model
of geopolymers which has been introduced recently [10,11].
Acknowledgements
The authors gratefully acknowledge the nancial sup-
port of the Particulate Fluids Processing Centre (PFPC),
a Special Research Centre of the Australian Research
Council (ARC). Dr Tim Sercombe in the Division of Mate-
rials of the School of Engineering at the University of
Queensland is thanked for help in obtaining some of the
dilatometric data.
References
[1] J. Davidovits, J. Therm. Anal. 37 (8) (1991) 1633.
[2] W.M. Kriven, J.L. Bell, M. Gordon, Ceram. Eng. Sci. Proc. 25 (1)
(2004) 57.
[3] Z.P. Bazant, M.F. Kaplan, Concrete at High Temperatures: Material
Properties and Mathematical Models, Longman House, Essex, 1996.
[4] V.F.F. Barbosa, K.J.D. MacKenzie, Mater. Lett. 57 (910) (2003)
1477.
[5] V.F.F. Barbosa, K.J.D. MacKenzie, Mater. Res. Bull. 38 (2) (2003)
319.
[6] H. Rahier, W. Simons, B. Van Mele, M. Biesemans, J. Mater. Sci. 32
(9) (1997) 2237.
[7] H. Rahier, B. Van Mele, J. Wastiels, J. Mater. Sci. 31 (1) (1996) 80.
[8] P. Duxson, G.C. Lukey, J.S.J. van Deventer, J. Mater. Sci. in press.
[9] P. Duxson, G.C. Lukey, S. Mallicoat, W.M. Kriven, J.S.J. van
Deventer, Coll. Surf. A, in press, doi:10.1016/j.colsurfa.2006.05.044.
[10] P. Duxson, G.C. Lukey, F. Separovic, J.S.J. van Deventer, Ind. Eng.
Chem. Res. 44 (4) (2005) 832.
[11] P. Duxson, J.L. Provis, G.C. Lukey, J.S.J. van Deventer, F.
Separovic, Langmuir 21 (7) (2005) 3028.
[12] S. Brunauer, P.H. Emmett, E. Teller, J. Am. Chem. Soc. 60 (1938)
309.
[13] P. Duxson, J.L. Provis, G.C. Lukey, S.W. Mallicoat, W.M. Kriven,
J.S.J. van Deventer, Coll. Surf. A 269 (13) (2005) 47.
[14] C.J. Brinker, G.W. Scherer, SolGel Science: The Chemistry and
Physics of SolGel Processing, Academic Press Inc., London, 1990.
[15] Part 2 of this paper: P. Duxson, G.C. Lukey, J.S.J. van Deventer, J.
Non-Cryst. Solids, submitted for publication.
[16] P.A. Kollman, I.D. Kuntz, J. Am. Chem. Soc. 94 (26) (1972)
9236.
[17] C.R.A. Catlow, A.R. George, C.M. Freeman, Chem. Commun. (11)
(1996) 1311.
[18] J.L. Bell, W.M. Kriven, 62nd Annual Meeting of the Microscopy
Society of America, vol. 10, Microscopy Society of America,
Savannah, 2004, pp. 590591.
[19] V.K. Leko, O.V. Mazurin, Glass Phys. Chem. 29 (1) (2003) 16.
[20] N.L. Bowen, O.F. Tuttle, J. Geol. 58 (5) (1950) 489.
[21] J.F. Schairer, N.L. Bowen, Am. J. Sci. 245 (4) (1947) 193.
[22] J.F. Schairer, J. Geol. 58 (5) (1950) 512.
[23] J.F. Schairer, J. Am. Ceram. Soc. 40 (7) (1957) 215.
[24] B. Coughlan, W.M. Carroll, Faraday Trans. I 72 (1976) 2016.
[25] C.J. Brinker, G.W. Scherer, E.P. Roth, J. Non-Cryst. Solids 72 (23)
(1985) 345.
[26] D.E. Day, J. Non-Cryst. Solids 21 (3) (1976) 343.
P. Duxson et al. / Journal of Non-Crystalline Solids 352 (2006) 55415555 5555