Stearic acid gel derived MgTiO3 nanoparticles: A low temperature intermediate phase of Mg2TiO4

Article abstract of DOI:10.1016/j.jallcom.2009.11.181

Stearic acid gel is employed to study the phase evolution of MgTiO₃ nanoparticles by thermal gravimetric analysis, X-ray diffraction, and Fourier transform infrared. During the preparation of stearic acid gel, tetrabutyl titanate easily absorbed moisture to hydrolyze into Ti(OH)₄ firstly, and then reacts with stearic acid and magnesium stearate to form magnesium-titanium oxide network polymer gel, meanwhile n-butanol is generated. When stearic acid gel is calcined in air, a series of oxidation and combustion reactions occur, meanwhile apparent heat is given off. The results show that a metastable intermediate phase Mg₂TiO₄ is generated at 450 °C and nearly disappeared at 550 °C. Simultaneously, a new solid phase of MgTiO₃ appears. The metastable intermediate phase Mg₂TiO₄ is successfully identified in the current work.

Full text of DOI:10.1016/j.jallcom.2009.11.181

Journal of Alloys and Compounds 492 (2010) 564–569
Contents lists available at ScienceDirect
Journal of Alloys and Compounds
journal homepage: www.elsevier.com/locate/jallcom
Stearic acid gel derived MgTiO3 nanoparticles: A low temperature intermediate
phase of Mg2TiO4
Dandan Lia,b, Liqiu Wanga,b,∗, Dongfeng Xuea,b
a State Key Laboratory of Fine Chemicals, Department of Materials Science and Chemical Engineering, Dalian University of Technology, Dalian, 116012, PR China
b Key Laboratory for Micro/Nano Technology and System of Liaoning Province, Dalian University of Technology, Dalian, 116023, PR China
a r t i c l e i n f o
a b s t r a c t
Article history:
Stearic acid gel is employed to study the phase evolution of MgTiO3 nanoparticles by thermal gravimetric
analysis, X-ray diffraction, and Fourier transform infrared. During the preparation of stearic acid gel,
tetrabutyl titanate easily absorbed moisture to hydrolyze into Ti(OH)4 firstly, and then reacts with stearic
acid and magnesium stearate to form magnesium–titanium oxide network polymer gel, meanwhile n-
butanol is generated. When stearic acid gel is calcined in air, a series of oxidation and combustion reactions
occur, meanwhile apparent heat is given off. The results show that a metastable intermediate phase
Mg2TiO4 is generated at 450 ◦C and nearly disappeared at 550 ◦C. Simultaneously, a new solid phase of
MgTiO3 appears. The metastable intermediate phase Mg2TiO4 is successfully identified in the current
work.
Received 25 July 2009
Received in revised form
25 November 2009
Accepted 26 November 2009
Available online 2 December 2009
Keywords:
Nanostructured materials
Sol–gel processes
© 2009 Elsevier B.V. All rights reserved.
Crystal structure
X-ray diffraction
Transmission electron microscopy
1. Introduction
Recently, most studies mainly focus on improving the purity
and reducing sintering temperature of MgTiO3 ceramic powder.
Magnesium titanate (MgTiO3), with ilmenite structure, has the
dielectric properties of low dielectric loss (high quality factor), high
dielectric constant and near zero temperature coefficient of res-
onant frequency [1]. Therefore, MgTiO3 has wide applications in
capacitors, resonators, filters, antennas for communication, radar,
direct broadcasting satellite and global positioning system operat-
ing at microwave frequencies [2,3]. In the past thirty years, much
attention has been focused on the synthesis of MgTiO3. Various
synthesis methods are reported, including physical and chemical
methods [4–9]. In particular, stearic acid gel (SAG) method is a use-
ful and attractive method for synthesis of MgTiO3 nanoparticles. In
a recent report on SAG synthesis of MgTiO3 powders by Kang et
al. [10], the tetragonal flake-like nanocrystallites of MgTiO3 is syn-
thesized by SAG method. SAG method has also many distinctive
advantages to prepare nanoparticles of BaTiO3 [11,12] and ZnTiO3
[13], such as decreasing significantly the crystallization tempera-
ture, producing high purity nanopowders, easy control of the M
(M = Ba, Zn and Mg) and Ti stoichiometric ratio and low cost in
comparison with other sol–gel methods [11,12].
To improve the purity of MgTiO3, investigating the phase evo-
lution of MgTiO3 nanoparticles is very meaningful. In the past
decades, there were some reports related to intermediate phases
to produce MgTiO3 by various methods. For the first kind of the
solid-state reaction method, an intermediate phase of MgTi2O5
was observed during the reaction process [14,15]. For the sec-
ond kind of the acetic acid sol–gel method, the intermediate
phases like MgTi2O5, Mg2TiO4 and TiO2 were appeared dur-
ing the thermal decomposition of the gel [2,16]. For the third
kind of the coprecipitation method, the product was never com-
pletely free from the phase of MgTi2O5 [17]. For the last kind
of the mechano-chemical complexation method, the phase of
MgTi2O5 was observed in the final product of MgTiO3 [18]. Up
to now, no work was on the phase evolution of SAG to produce
MgTiO3.
The aim of this work is to determine the mechanism of SAG
thermal decomposition and to obtain the optimum conditions for
producing high purity MgTiO3 nanoparticles. Different measures
including calcination of SAG, blank experiment and quenching
method are employed to uncover the mechanism of MgTiO3
nanoparticles derived from SAG. Especially, MgTiO3 nanoparticles
are evolved through a metastable intermediate phase Mg2TiO4,
which appears at 450 ◦C and disappears at 550 ◦C. Furthermore,
600 ◦C is the optimal temperature to mildly synthesize MgTiO3
nanoparticles.
∗
Corresponding author at: State Key Laboratory of Fine Chemicals, Department of
Materials Science and Chemical Engineering, Dalian University of Technology, 158
Zhongshan Road 43# Postbox, Dalian, Liaoning 116012, PR China.
E-mail address: Lqwang218cn@163.com (L. Wang).
0925-8388/$ – see front matter © 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jallcom.2009.11.181

D. Li et al. / Journal of Alloys and Compounds 492 (2010) 564–569
565
Scheme 1. All steps of investigating the phase evolution of MgTiO3 nanoparticles.
2. Experimental
using KBr as a binding material in the range of 400–4000 cm−1. The general mor-
phology of MgTiO3 nanoparticles was detected by transmission electron microscopy
(TEM, Tecnai G220 S-Twin, America). The average and distribution sizes of MgTiO3
nanoparticles were calculated from MgTiO3 transmission electron images by the
software of Gatan Digital Micrograph.
2.1. Gel preparation procedure
The gel was prepared by using available manufacture of analytical grade mag-
nesium stearate, tetrabutyl titanate, and stearic acid. An appropriate amount of
stearic acid was first melted in a beaker at about 100 ◦C, and then a fixed amount
of magnesium stearate was added to the melted stearic acid and dissolved to form
a yellow transparent solution. Subsequently, the stoichiometric tetrabutyl titanate
was added to this solution under stirring to form a homogeneous brown sol. At the
same time some little bubbles were generated in the brown sol. Treated the brown
sol at high temperature for short time to remove the little bubbles into the air. And
then naturally cooled down to room temperature, finally, dried for 12 h to obtain the
gel denoted as “Mg–Ti gel”. The Mg–Ti gel was calcined at different temperatures
in air to obtain samples. In a typical synthesis procedure, stearic acid, magnesium
stearate and tetrabutyl titanate were mixed according to the molar ratio [stearic
acid]:[magnesium stearate]:[tetrabutyl titanate] = 2.1:1.0:1.0.
2.3. Scheme on investigating the phase evolution of MgTiO3 nanoparticles
In the current work, many measures were taken to investigate the phase evolu-
tion of MgTiO3 nanoparticles, to clearly show all steps here, they were summarized
in Scheme 1. Scheme 1 clearly shows that three kinds of gel were prepared, and a
series of characterizations were carefully carried out.
3. Results and discussion
3.1. Thermal decomposition process
To investigate the possible reactions occurred in the preparation of Mg–Ti gel,
the generated little bubbles caused by adding tetrabutyl titanate were cooled into
distillate by a condenser in the other similar preparation of Mg–Ti gel experiment,
and analyzed the transparent distillate by FTIR.
Keeping the same conditions as the above “Mg–Ti gel” preparation, blank exper-
iments were carried out for only the Mg or Ti component dissolved in stearic acid
to form the corresponding gels, denoted as “Mg gel” or “Ti gel”.
Fig. 1 shows the TGA-SDTA analysis for the dried Mg–Ti gel. For
the SDTA curve in Fig. 1, the endothermic peak at 110 ◦C is assigned
to the volatilization of generated n-butanol. All other peaks are
exothermal over 170 ◦C. This means that the decomposition process
of the Mg–Ti gel consists of a series of oxidation and combustion
reactions. The TGA curve in Fig. 1 indicates that the weight loss
occurs in three steps corresponding to their exothermal peaks (at
322, 418 and 465 ◦C in the SDTA curve): (i) the initial weight loss
of about 46% occurs from 170 to 330 ◦C, which is assigned to com-
bustion of organic substances in the gel; (ii) the second weight loss
of about 24% occurs in the range of 330–420 ◦C due to the combus-
tion of organic substances further in the gel; (iii) the final weight
2.2. Characterization
The thermal decomposition of the dried Mg–Ti gel was analyzed by thermogravi-
metric analyzer (TGA, TGA/SDTA 851e, Switzerland). TGA/SDTA measurements were
done from room temperature to 600 ◦C with a heating rate of 10 ◦C/min in air. The
calcined powders were characterized by X-ray diffraction (XRD, D/Max-3C, Japan),
using Cu K␣ radiation. FTIR spectra of the obtained samples were characterized
by Fourier transform infrared spectrometry (FTIR, Bruker Equinox 55, Germany),

566
D. Li et al. / Journal of Alloys and Compounds 492 (2010) 564–569
Fig. 1. TGA and SDTA curves of Mg–Ti gel.
loss of about 13% occurs between 420 and 500 ◦C, corresponding
to the SDTA peak at 465 ◦C, is assigned to generation of Mg2TiO4
solid phase, simultaneously, accompanied by some organic residue
oxidation and combustion reactions which release CO2 into the air.
No apparent peak and weight loss can be observed over 500 ◦C. All
possible crystallite species generated in each step are identified by
XRD measurements in Fig. 2.
3.2. Crystal phases of calcined Mg–Ti gel at different temperatures
Fig. 2(a) and (b) shows the XRD patterns of all samples calcined
inairat differenttemperatures for 2 h. As shown in Fig. 2(b), the XRD
patterns of 250 and 300 ◦C samples are similar to that of gel sample,
which indicate that the network structure of magnesium–titanium
oxide polymer is destroyed gradually from room temperature to
300 ◦C, and it is almost destroyed completely over 350 ◦C. Mean-
while, there is no significant crystal phase observed between 350
and 400 ◦C, which indicates that some amorphous phases are ini-
tially generated. In Fig. 2(a), the trace of metastable phase Mg2TiO4
(JCPDS, 25-1157) peaks can be found from XRD pattern of 450 ◦C.
When the temperature is up to 500 ◦C, the apparent peaks of
Mg2TiO4 are found. With increasing the calcination temperature,
the peaks of Mg2TiO4 nearly disappears at 550 ◦C. Simultaneously,
new solid phase of MgTiO3 (JCPDS, 06-0494) is generated. The
peaks of MgTiO3 get stronger and stronger with increasing the cal-
cined temperature. All measured XRD patterns strongly confirm
our TG/SDTA analysis for the dried gel.
Fig. 2. XRD patterns of Mg–Ti gel calcined at different temperatures, (a) 450–650 ◦C;
(b) gel-400 ◦C.
RCOOH + C4H9OH → RCOOC4H9 + H2O
(3)
Because the whole preparation process is in the open air, tetra-
butyl titanate easily absorbs moisture from air, and then hydrolyzes
to generate n-butanol. The distillate FTIR in Fig. 3 shows that reac-
tion (1) is actually occurred. Subsequently, the melt stearic acid can
3.3. FTIR spectrometry
3.3.1. FTIR of distillate and Mg–Ti gel
The collected distillate was analyzed by FTIR shown in Fig. 3.
All absorption bands in the FTIR spectrum are in good agreement
with the standard spectrum of n-butanol [19]. Hence, the above
FTIR spectrum proves that n-butanol is generated during the prepa-
ration of Mg–Ti gel. Meanwhile, according to the experimental
phenomena of the little bubbles generated by adding tetrabutyl
titanate in the preparation of Mg–Ti gel, the possible reactions in
the preparation of Mg–Ti gel can be written as follows: R = C17H35
Ti(OC4H9)4 + 4H2O → Ti(OH)4+4C4H9OH
(1)
Fig. 3. FTIR spectrum of distillate collected from the Mg–Ti gel.
(2)

D. Li et al. / Journal of Alloys and Compounds 492 (2010) 564–569
567
Fig. 4. FTIR spectrum of the Mg–Ti gel.
Fig. 5. FTIR spectra of Mg–Ti gel calcined at different temperatures.
naturally react with the products generated by reaction (1). So the
reactions (2) and (3) should be existed in the SAG system. Vice versa,
the generated water in reactions (2) and (3) can promote reaction
(1). That is to say, the reactions of (1) and (2) or (1) and (3) promote
each other. As a result, all reactions are carried out completely in
the preparation of Mg–Ti gel.
3.4. Intermediate phase detected from Mg–Ti gel to form MgTiO3
To understand well the process of MgTiO3 formation in Mg–Ti
gel, some blank experiments were carried out for Mg gel and Ti
gel. The phase development in Mg gel and Ti gel was character-
ized by XRD under the same conditions as the previous Mg–Ti gel
(Figs. 6 and 7).
In order to confirm that the possible reactions (2) and (3)
occurred in the preparation process of Mg–Ti gel, FTIR is performed
to trace the organic species of Ti(IV)–stearic complexes and butyl
stearate in the Mg–Ti gel. Fig. 4 shows FTIR spectrum of Mg–Ti
gel. In Fig. 4, the strong peaks at 2920 and 2852 cm−1 can be
assigned to the symmetric and asymmetric vibrations of –CH2– and
–CH3 groups, respectively. The CH2 scissoring and in-phase rock-
ing bands appear at around 1462 and 719 cm−1, respectively. All
these observations confirm the existence of long organic chains
in the Mg–Ti gel [20]. Peaks at 1745 and 1178 cm−1 represent
the stretching vibration band of C O and C–O–C groups belong-
ing to ester, which indicate that the reaction (3) may occur to
form butyl stearate in the Mg–Ti gel. According to some docu-
ments [21], the absorbance bands are attributable to the bidentate
of Ti(IV)–carboxylic complexes are observed at 1550–1590 cm−1
(antisymmetric) and 1430–1470 cm−1 (symmetric). The latter is
overlapped by the scissoring frequency of –CH2– at 1464 cm−1
and, consequently, forms a broaden band. The corresponding large
absorbance bands at around 1462 and 1570 cm−1 in Fig. 4 indicate
that the Ti(IV)–stearic complexes are formed in the Mg–Ti gel. The
proposed reaction (2) occurs indeed. Therefore, the above assump-
tion of reactions (1)–(3) occurred in the preparation of Mg–Ti gel
are credible.
In Fig. 6, no MgO phase can be observed for 250 and 300 ◦C XRD
patterns. Referring to the gel XRD pattern in Fig. 2(b), the peaks at
around 2ꢀ = 10◦ and 21◦ can be assigned as that the Mg gel network
structure is not destroyed completely. When the temperature is
over 350 ◦C, the gel network structure is destroyed completely and
MgO (JCPDS, 45-0946) phase is formed. With increasing the cal-
cined temperature, the intensities of MgO peaks become stronger
and stronger.
In Fig. 7, no apparent peaks can be observed for 300 ◦C XRD pat-
tern, which indicates that the Ti gel network structure is almost
destroyed completely, and all possible generated phases are amor-
phous. When the temperature is over 350 ◦C, the mixture of anatase
and rutile phases is formed. Anatase is the dominant phase at low
temperatures (e.g., <500 ◦C), while by further increasing the tem-
perature, rutile becomes the dominant crystal phase in the calcined
mixture.
3.3.2. FTIR of calcined samples
To trace some possibly organic species existed in the calcined
gels, the FTIR spectra for the as-prepared samples at different cal-
cination temperatures are shown in Fig. 5. The small absorption
bands around 2800–3000 cm−1 and 1400–1600 cm−1 in the 350 ◦C
FTIR spectrum indicate that little amount of organic substance still
exists. With increasing the temperature, i.e., over 400 ◦C, no appar-
ent absorption bands for organic substance can be observed, while
the absorption band between 500 and 700 cm−1 gets stronger and
stronger, which is assigned to the formation of TiO6 octahedra. In
addition, a broad band centered at 3460 cm−1 is corresponded to
the water adsorbed on the surface of nanoparticles.
The above FTIR spectra indicate that MgTiO3 phase is generated
at 550 ◦C, and with increasing the calcination temperature, MgTiO3
crystallites grow better and better.
Fig. 6. XRD patterns of Mg gel calcined at different temperatures.

568
D. Li et al. / Journal of Alloys and Compounds 492 (2010) 564–569
Hence, new solid phase of MgTiO3 is generated at 550 ◦C, and
with increasing the temperature, the MgTiO3 crystal phase grows
better and better (Fig. 2).
Itis wellknown thatthe finalsolid phase reaction product isseri-
ously affected by calcination time at a fixed temperature point. In
order to clearly show the influence of calcination time on the phase
evolution of the final product MgTiO3, the typical calcination tem-
perature of 550 ◦C was chosen for our experiments. We calcined the
Mg–Ti gel for different calcination times. In order to maintain the
existed phase structure in samples as much as possible, all calcined
samples were immediately quenched in atmosphere.
The trace of the phase evolution of the final product MgTiO3 was
finally detected by XRD (Fig. 8). To easily observe the correspond-
ing crystal phases, the original XRD patterns (e.g., 5, 10, 15 and
30 min) in Fig. 8 are smoothed by the software of Origin, using the 5
Fig. 7. XRD patterns of Ti gel calcined at different temperatures.
From Figs. 6 and 7, it is confirmed that TiO2 and MgO crystal
phases can only be formed directly in their corresponding calcined
stearic acid gel. In contrast with the above Ti gel and Mg gel, the
only metastable phase Mg2TiO4 instead of TiO2 and MgO crystal
phases can be detected in the calcined Mg–Ti gel by XRD, which
indicates that the metastable phase Mg2TiO4 is formed by excessive
MgO interacting with TiO2 in the calcined Mg–Ti gel. Otherwise,
TiO2 and MgO crystal phases should be detected by XRD in the cal-
cined Mg–Ti gel. The possible reason is that MgO can be more easily
evolved from the calcined Mg–Ti gel than that of TiO2, so excessive
MgO reacts with TiO2 to form the metastable phase Mg2TiO4. This
evidence can be found in Figs. 6 and 7. The well-grown crystal MgO
can be obtained even at 350 ◦C in Fig. 6, but for TiO2, the corre-
sponding temperature to obtain well-grown crystal TiO2 must be
over 400 ◦C. The reaction can be described as follow:
2MgO + TiO2 → Mg2TiO4
(4)
With increasing temperature, the rest of TiO2 is evolved from the
calcined Mg–Ti gel, and then when the temperature is high enough
(e.g., 550 ◦C), the following solid reaction may happen:
Mg2TiO4 + TiO2 → 2MgTiO3
(5)
Fig. 8. XRD patterns of Mg–Ti gel calcined at 550 ◦C from 5 to 120 min and then
immediately quenched in the atmosphere.
Fig. 9. TEM image of MgTiO3 nanoparticles derived from Mg–Ti gel calcined at
600 ◦C.

D. Li et al. / Journal of Alloys and Compounds 492 (2010) 564–569
569
gels, such as Mg–Ti, Mg and Ti gel. In the preparation of stearic
acid gel, several reactions occur to form magnesium–titanium
oxide network polymer gel, meanwhile n-butanol is generated.
The decomposition process of the dried Mg–Ti gel consists of a
series of oxidation and combustion reactions in air, accompanied
by exothermal phenomena. The metastable intermediate phase of
Mg2TiO4 is identified successfully for the first time. XRD measure-
ment strongly confirms that the metastable phase Mg2TiO4 can be
initially formed at about 450 ◦C, and then react with amorphous
TiO2 to form MgTiO3 phase. With increasing the calcination tem-
perature, the crystallinity of the as-prepared MgTiO3 nanoparticles
can be improved significantly. The mechanism of the phase evo-
lution of MgTiO3 is in favor of forming nanoscale crystallites of
MgTiO3.
Acknowledgments
The authors thank the financial support from the National
Natural Science Foundation of China (Grant Nos. 50872016,
20973033).
Fig. 10. Size distribution of MgTiO3 nanoparticles.
point Adjacent Averaging Method. Fig. 8 shows that the metastable
Mg2TiO4 is the dominant phase between 5 and 30 min. When the
calcined time is extended to 2 h, the metastable Mg2TiO4 phase
disappears and the MgTiO3 crystal phase becomes the dominant
phase. Therefore, the metastable phase Mg2TiO4 is also identified
as the intermediate phase of MgTiO3 even at a fixed calcination
temperature point.
References
[1] C.L. Huang, Y.B. Chen, M.L. Lee, J. Alloys Compd. 469 (2009) 357–361.
[2] Y.M. Miao, Q.L. Zhang, H. Yang, H.P. Wang, Mater. Sci. Eng. B 128 (2006)
103–106.
[3] C.L. Huang, C.H. Shen, T.C. Lin, J. Alloys Compd. 468 (2009) 516–521.
[4] K.P. Surendran, A. Wu, P.M. Vilarinho, V.M. Ferreira, Chem. Mater. 20 (2008)
4260–4267.
[5] V. Parvanova, M. Maneva, Thermochim. Acta 279 (1996) 137–141.
[6] J. Zeng, H. Wang, S. Song, J. Cheng, Q. Zhang, J. Cryst. Growth 178 (1997)
355–359.
3.5. Morphology observation
[7] B.D. Lee, H.R. Lee, K.H. Yoon, Y.S. Cho, Ceram. Int. 31 (2005) 143–
146.
Fig. 9(a) and (b) shows the transmission electron images of
MgTiO3 nanoparticles prepared by calcined the Mg–Ti gel at 600 ◦C
for 2 h. Fig. 10 shows the size distribution of the MgTiO3 nanopar-
ticles observed from the transmission electron images. The results
clearly show that MgTiO3 nanoparticles only consist of nanoscale
crystallites with the average size of about 60 nm. Most MgTiO3
crystallites do not aggregate to form large particles. Nearly all crys-
tallites are almost ellipse in shape, and well dispersed with a slight
agglomeration. The mechanism of MgTiO3 formation may elucidate
its small crystallite size, i.e., the phase Mg2TiO4, amorphous phase
and MgTiO3 may affect each other and result in a low growth rate
of MgTiO3 crystallites.
[8] W. Zhu, X. Zhang, L. Xiang, S. Zhu, Nanoscale Res. Lett.
731.
4 (2009) 724–
[9] W. Wu, Q. He, C. Jiang, Nanoscale Res. Lett. 3 (2008) 397–415.
[10] H. Kang, L. Wang, D. Xue, K. Li, C. Liu, J. Alloys Compd. 460 (2008) 160–163.
[11] L. Wang, H. Kang, K. Li, D. Xue, C. Liu, J. Phys. Chem. C 112 (2008) 2382–
2388.
[12] L. Wang, L. Liu, D. Xue, H. Kang, C. Liu, J. Alloys Compd. 440 (2007) 78–83.
[13] L. Wang, H. Kang, D. Xue, C. Liu, J. Cryst. Growth 311 (2009) 611–614.
[14] X. Zhou, Y. Yuan, L. Xiang, Y. Huang, J. Mater. Sci. 42 (2007) 6628–6632.
[15] K. Sreedhar, N.R. Pavaskar, Mater. Lett. 53 (2002) 452–455.
[16] X. Bokhimi, J.L. Boldu, E. Munoz, O. Novaro, Chem. Mater. 11 (1999) 2716–
2721.
[17] M.I. Yanovskaya, N.M. Kotava, N.V. Golubko, J. Sol–Gel Sci. Technol. 11 (1998)
23–29.
[18] A.V. Prasada Rao, S. Komarneni, Mater. Lett. 38 (1999) 186–189.
[19] SDBSWeb: http://riodb01.ibase.aist.go.jp/sdbs/ (National Institute of Advanced
Industrial Science and Technology, November 18, 2009).
[20] X. Wu, L. Zou, S. Yang, D. Wang, J. Colloid Interface Sci. 239 (2001) 369–
373.
4. Conclusions
In the present work, the phase evolution of MgTiO3 from the
calcined stearic acid gel in air is investigated by three kinds of
[21] Y. Murakami, T. Matsumoto, Y. Takasu, J. Phys. Chem. B 103 (1999) 1836–1840.