Reactivity of V2O5, MoO3 and WO3 in molten KNO3, studied by mass spectrometry

Article abstract of DOI:10.1016/S0925-8388(01)00955-0

The reactivity of transition metal oxides MoO₃, WO₃ and V₂O₅ in molten KNO₃ was studied by mass spectrometry of the gases evolved during heating, and X-ray diffraction of the solidified melts. The rea

Full text of DOI:10.1016/S0925-8388(01)00955-0

Journal of Alloys and Compounds 322 (2001) 97–102
L
www.elsevier.com/locate/jallcom
Reactivity of V O5, MoO3 and WO3 in molten KNO3, studied by mass
2
spectrometry
P. Afanasieva , , D.H. Kerridge
b
*
a
´
Institut de Recherche sur la Catalyse, 2 Avenue A. Einstein, 69626, Villeurbanne, Cedex, France
bDepartment of Engineering Materials, University of Southampton, Southampton SO17 1BJ, UK
Received 31 August 2000; received in revised form 18 January 2001; accepted 25 January 2001
Abstract
The reactivity of transition metal oxides MoO3, WO3 and V O5 in molten KNO3 was studied by mass spectrometry of the gases
2
evolved during heating, and X-ray diffraction of the solidified melts. The reaction stoichiometry showed that multistep reactions of the
polyoxoanions occur. Intermediate products identified in the reactions of the oxides are hexa- and pentavanadates, tri- and dimolybdates,
and ditungstate. After prolonged heating at 5508C, the reactions proceeded to form the monooxoanions: VO23 , MoO422 and WO422. The
reactivity of the oxides decrease in the sequence: V O5.MoO3.WO3, the inverse of the oxoacidity sequence deduced from the equilibria
2
in ionic melts (WO3.MoO3.V O5). The first reaction peak is determined by kinetics of the solid–liquid reaction, whereas further
2
transformations of polyoxospecies are determined by their oxoacidity.
2001 Elsevier Science B.V. All rights reserved.
Keywords: Molten nitrate; Molybdenum; Tungsten vanadium; Oxoacidity
1. Introduction
Generally, the products were oxospecies such as soluble
oxoanions or insoluble oxides.
Reactions of oxides with molten salts are technically
important since they are involved in the hot corrosion of
alloys [1] and in the syntheses of new materials from ionic
fluxes [2].
Soluble oxoanions can be produced from molybdenum
oxide which undergoes stepwise transformations leading to
alkali metal molybdates A2MoO4 (A5K, Na)
MoO3 1 2NO23 → MoO422 1 2NO2 1 O2
(3)
1
2
]
Molten salts as reaction media provide an alternative to
aqueous chemistry by offering different solubilities and/or
reactivity of species [3–6]. Molten nitrates act as basic
and/or oxidising media towards the majority of chemical
compounds [7]. These properties follow respectively from
the ability of nitrate anion to be a donor of oxygen anion
O22, or a donor of oxygen atom O?. The following
equations can be proposed to describe the Lux–Flood
(L–F) interaction of molten nitrates with ionic or neutral
oxoacids:
Stepwise transformation through intermediate di- and
trimolybdates was suggested [9]. Coumert et al. [10,11]
reported that the dimolybdate was formed rapidly at 3508C
from the trioxide and the monomolybdate more slowly, on
the basis of polarography and electrochemical titration. In
contrast, Kust [12] reported monomolybdate was in
equilibrium with dimolybdate at 282–3448C, while
Schlegel and Bauer [13] found using emf measurements
that the trioxide reacted only slowly from 330 to 6038C to
form dimolybdate.
Tungsten trioxide has been much less studied, though El
Din [14] has reported that it is stable in potassium nitrate
at 3508C. In contrast, Hassanien and Kordes [15] found
that cryoscopy suggested the formation of monotungstate.
However, it is not clear what the maximum temperatures
were to which the solutions had been heated. As with the
molybdates, Kust [12] found the monotungstate to be in
equilibrium with ditungstate and oxide.
NO23 1 Acid NO12 1 Acid.O22
(1)
(2)
NO12 1 NO23 2NO2 1 O2
1
]
2
Reactions of the main classes of inorganic compounds
were extensively studied by Kerridge and coworkers [8,9].
*Corresponding author. Tel.: 133-4-72-44-53-39; fax: 133-4-72-44-
53-99.
E-mail address: afanas@catalyse.univ-lyonl.fr (P. Afanasiev).
0925-8388/01/$ – see front matter
PII: S0925-8388(01)00955-0
2001 Elsevier Science B.V. All rights reserved.

98
P. Afanasiev, D.H. Kerridge / Journal of Alloys and Compounds 322 (2001) 97–102
The existing literature on reactions of vanadium(V)
were broken and the solidified melts ground in a mortar.
X-Ray diffraction patterns of solidified melts were ob-
tained on a Siemens diffractometer with Cu Ka emission.
The diffractograms were analysed using the standard
JCPDS files.
oxide is somewhat similar. El Din and Hosary [16] found a
slow reaction at 3508C in potassium nitrate to form
metavanadate (VO23 ), which proceeded more quickly
when sodium peroxide was added. However metavanadate
was reported to form from the oxide at 3208C in the
sodium nitrate–potassium nitrate eutectic and then reacted
at higher temperatures to orthovanadate [9], while Hassa-
nien [17] found both the oxide and a condensed vanadate
to give freezing point depressions appropriate to formation
of orthovanadate, but again the solution is likely to have
been heated to a higher temperature than 3208C at some
point. Tridot et al. [18] working with the sodium nitrate–
potassium nitrate eutectic found three vanadium oxides to
be acidic (i.e. NO2 and O2 were evolved), but in contrast
Mokhasaev et al. [19] determining the phase diagram of
the oxide with potassium nitrate found no gas evolution
and a 2:3 compound, which could be a hexavanadate if the
reaction was actually that of an acid–base type.
Many of the differences in these sometimes conflicting
reports arise because the experimental conditions have not
been completely specified and others can be confidently
attributed to the very varied reactant concentrations neces-
sary for the different techniques used, yet others can be
attributed to error. However a comprehensive and sys-
tematic knowledge of these reactions, namely the exact
composition of polyoxospecies formed at the different
stages of reaction is important for prediction of corrosion
products as well as the results of chemical syntheses of
solids in these media, since the species from the melt
solution are usually included in the corresponding precipi-
tates.
The gaseous products evolved upon heating of the
samples were studied using a Gas Trace A mass-spec-
trometer (Fison Instruments) equipped with a quadrupole
analyser (VG analyser) working in Faraday mode. The
ionisation was done by electron impact with an electron
energy of 65 eV. The samples (ca 0.1 g) were heated from
room temperature to 5508C in a glass cell at the heating
rate of 1.58 min21. A silica capillary tube heated at 453 K
continuously bled off a proportion of the gaseous reaction
products. Several signals were registered, those with m/e5
18, 30, 32, 44 and 46 corresponded, respectively, to
ionized species H2O, NO, O2, CO2, and NO2. Calibration
of sensitivity for semi-quantitative estimations of gaseous
products amounts was done using a reaction of known
stoichiometry, the decomposition of Pb(NO3)2. Though
the sensitivity for different compounds is not the same, for
any given signal we achieved a good linearity of the
detector response. For different amounts (20, 50, 100 and
300 mg) of Pb(NO3)2, the signals of NO, O2 and NO2
varied all in step, proportionally to the compound loading.
Since the linearity of detector is satisfactory, we concluded
that the signals can be integrated and the amounts of
ionized species for different domains of temperature can be
compared.
3. Results
The creation of scales of acidic/basic strength in melts,
similar to those for water, still await suitable systematic
studies. For although studies of the oxobasicity of nitrate
melts have been carried out for several decades, it is
evident that there is still much contradiction in this field,
caused by the varying observations and the absence of a
uniform approach. A thorough comparison of the reactivity
of different oxospecies would be useful.
Though the experimental device was simple, several
points of the technique needed to be established to be sure
of correct interpretation of the data obtained. Analysis time
was less than 1 s as verified by opening of a capillary
valve to air. The rate of air pumping by the capillary, if
opened at atmospheric pressure, was 15 cm3 min21, with a
reactor volume of less than 5 cm3. If connected to an
empty closed reactor, the signal level reached a steady
state value in vacuum in ca. 30 s. Thus at a dwell time
between analyses of 30 s and a heating rate 1.58 min21, the
distance between subsequent points was 0.758, seemingly
suitable for the reactions studied, which gave extended
peaks on the temperature scale.
In this work, we discuss the variation of reactivity of
three oxides, i.e. MoO3, WO3 and V O5 with the same
2
oxide donor (KNO3), with the technique of evolved gas
mass spectrometry which has already shed fresh light on
reactions of Zr(IV) chloride [20,21] together with XRD on
the solid products.
Another important parameter was measurement of back-
ground gases from the potassium nitrate used as the
solvent/reactant. The temperature at which KNO3 com-
mences to decompose is reported to be 4008C [22]. Thus
blank measurements were made with pure KNO3, to
ascertain whether the rate of its decomposition was
measurable under the conditions used here and also to
discover if features arose from impurities in the commer-
cial nitrate and/or any residual gases from the experimen-
tal device. The corresponding mass spectra (Fig. 1) show
2. Experimental
Samples of solidified melts for the XRD study have
been prepared in Pyrex test tubes. Mixtures of transition
metal oxides of approximately molal concentration in
KNO3 were heated at a rate 1.58 min21. When the desired
temperature was achieved the test tubes were removed
from the furnace and cooled in air. Then the test tubes

P. Afanasiev, D.H. Kerridge / Journal of Alloys and Compounds 322 (2001) 97–102
99
react slowly with residual water. Carbon dioxide may also
be due to oxidation of any impurity organic species, but
the latter might have been expected to begin reaction at
lower temperatures, whereas there was a progressive
increase in CO2 signal from 300 to 5008C.
The nitrogen dioxide signal was quite negligible (even
on the 10211 scale), being at the limit of the device
sensitivity.
It must be emphasized that pure NO2 in the mass
spectrometer is always registered as two signals: that of
NO2 itself (m/e546) and a five times larger NO signal
(m/e)530. The observation of NO arose probably from
decomposition of NO2 in the mass spectrometer. We could
not see any considerable variations of NO/NO2 ratio as a
function of the transfer line temperature. The same results
were obtained at capillary temperatures of 453 and 298 K.
We suggest therefore that decomposition occurs in the ion
source itself. Separate experiments using gaseous NO2
showed that when this gas was injected the response at
m/e530 was five times that at m/e546 (NO2), which
confirms the ratio found here. It is important to note that
the ratio of signals observed from pure KNO3 corres-
ponded to formation of NO only.
Since the mass spectrometer worked in the Faraday
mode, and linearity was verified, semi-quantitative analysis
of the intensity of peaks was possible, though secondary
processes in the mass spectrometer occurred, as illustrated
above for nitrogen dioxide. The analysis is quantitative
provided the final stoichiometry of the reaction is known
from another technique, i.e. here X-ray diffraction. The
reaction peaks can then be integrated and their proportions
give the values of degree of reaction advancement, as a
part from the final stoichiometry.
Fig. 1. Mass-spectroscopic signals of gases evolved during heating of
pure KNO3.
that below 5508C, the decomposition rate was very slow.
Only water gave a considerable intensity, at a 1028 order
of scale (note for comparison, that if the measuring
capillary was opened to air, the signals of oxygen and
nitrogen were about 1026). We observed slow dehydration
of salt (background water) which continued up to tempera-
tures as high as 5008C. The nitric oxide signal at mass 30
was very low but could be clearly seen at the 1029 scale,
and probably arose because of the presence of some nitrite
impurity in the nitrate which decomposed above 3008C:
The results for the oxides studied were as follows:
3.1. V O5
2
V O5 reaction began slowly at 2008C and accelerated
2
markedly at around the melting point of KNO3 (3368C).
Mass spectroscopy showed NO, NO2 and O2 to be more
freely evolved at that temperature (Fig. 2). In contrast,
smaller quantities of hydrated water were evolved at a
fairly steady rate on heating above room temperature and
its evolution was almost complete on melting of the
solvent salt.
1
]
2KNO2 K2O 1 2NO 1 2 O2
(4)
Pure NO2 was formed in all the stages of reaction,
which was of acid–base type (i.e. Eqs. 1 and 2) since the
oxygen mass peak (32) moved very much in step with
those of NO and NO2, all three gases having maximum
release rates at 365, 424 and 4658C. Small amounts of
carbon dioxide were evolved above 2508C and were
thought to be formed from the small impurity of carbonate.
Comparison of the CO2 signal from this reaction and the
This nitrite could well arise from thermal decomposi-
tion. Because the nitrite impurity concentration was so
low, the amount of oxygen which could be evolved by
reaction (4) was below the limit of definite detection, since
while the NO signal could readily be followed at the 1029
scale, that of oxygen would be similar in size to its
background signal from residual air.
Small amounts of CO2 (m/e544) were detected above
the melting temperature suggesting the possible presence
of some alkali metal carbonates which would probably
blank experiment showed that this Lux–Flood acid (V O5)
considerably lowered the temperature of carbon dioxide
evolution due to the reaction
2

100
P. Afanasiev, D.H. Kerridge / Journal of Alloys and Compounds 322 (2001) 97–102
363, 404 and 424 having the areas of 12.6, 16.2 and 22%
of total curve area, before the more clearly separated peak
of 4638C which provided 41% of the total area. The first
three peaks were rather poorly separated, the peak at 4048C
not being visible on the experimental curve as a visible
maximum. By contrast, the area of the sharp peak at 4638C
could be reliably estimated, and corresponded well with
the value of 40% of total gas production in reactions
(6)–(9) inclusive. Thus the last peak at 4638C is that of
reaction (9), whereas reactions (6)–(8) cannot be attribu-
ted to individual reaction maxima.
3.2. MoO3
MoO3 showed similar effects. NO, NO2 and O2 evolved
above the melting point with very similar gaseous ratios
and with two maxima at 384 and 4558C (Fig. 3) while
again water was mostly evolved at temperatures below the
melting point of the solvent/reactant. XRD indicated the
initial products were K2Mo3O10 at 4008C, K2Mo2O7 at
5008C but K2MoO4 at 5508C (Table 1), corresponding to
the sequence of equations:
Fig. 2. Mass-spectroscopic signals of gases evolved during heating of a
mixture of KNO3 and V O5 in molar ratio 10:1.
2
2K2CO3 1 V O5 2KVO3 1 2CO2
(5)
2
However, only an insignificant proportion of V O5
2
reacted this way.
X-Ray diffraction indicated that there was little reaction
of V O5 until 3368C but at 4008C after 2 h, a complex
2
vanadate had formed (K2V O16 with probably some
6
K2V O21), that this took up more oxide to form K3V O14
8
5
and then KVO3 was formed at 5508C (after 8 h), corre-
sponding to the reactions:
1
2
]
4V O5 1 2KNO3 → K2V O21 1 2NO2 1 O2
(6)
(7)
(8)
(9)
2
8
1
2
]
3K2V O21 1 2KNO3 → 4K2V O16 1 2NO2 1 O2
8
6
5K2V O16 1 8KNO3 → 6K3V O14 1 8NO2 1 2O2
6
5
1
2
]
K3V O14 1 2KNO3 → 5KVO3 1 2NO2 1 O2
5
Separation of the NO pattern using a PEAKFIT utility
with Gaussian–Lorentzian area profiles as profile fitting
functions, showed that the curve could be described with
r250.99 as a superposition of peaks with the maxima at
Fig. 3. Mass-spectroscopic signals of gases evolved during heating of a
mixture of KNO3 and MoO3 in molar ratio 10:1.

P. Afanasiev, D.H. Kerridge / Journal of Alloys and Compounds 322 (2001) 97–102
101
Table 1
Phase composition of solidified melts as determineda
Reaction T (8C)
Reacted oxide
V
Mo
W
400
500
550b
K2V O16 (37–175)
K2Mo3O10 (37–1467)1
eK2Mo2O7 (36–34)
K2Mo2O7 (36–34)
eK2MoO4 (29–1021)
K2MoO4 (29–1021)
No reaction
6
eK3V O14 (27–453)
5
eK3V O14 (27–453)
K2W2O7
(19–1007)
K2WO4 (24–905)
5
1KVO3 (33–1052)
KVO3 (33–1052)
a e means small relative intensity of peaks of the corresponding phase.
b After 8 h at this temperature.
1
]
3MoO3 1 2KNO3 → K2Mo3O10 1 2NO2 1 2 O2
(10)
values 66 and 33% for reactions (10) and (11) provided
reaction (12) was negligible at 5508C. Thus the maxima at
384 and 4558C corresponded to the formation of tri-
molybdate and dimolybdate, respectively. Dimolybdate
was persistent, and disappeared only after prolonged
heating (8 h) at 5508C when it formed monomolybdate by
reaction (12).
1
]
2K2Mo3O10 1 2KNO3 → 3K2Mo2O7 1 2NO2 1 2 O2 (11)
1
]
K2Mo2O7 1 2KNO3 → K2MoO4 1 2NO2 1 2 O2
(12)
After reaction at 5008C, XRD of the solidified melt
showed the lines of K2Mo2O7. Separation of the ex-
perimental curves (r250.995) showed that the areas of
peaks at 384 and 4558C were 68 and 32% of the total
profile area, respectively. This is close to the theoretical
3.3. WO3
WO3 was less reactive than MoO3 but again gave
similar ratios of NO, NO2 and O2 but only from 4508C
whereas water evolution continued from 100 to 4508C
(Fig. 4). Since the temperature range of reaction was
narrow, the solidified melt after reaction at 5008C con-
tained a complex mixture of products, where only the
presence of ditungstate could be reliably identified. No
peak separation was possible in this case because there
were no distinct reaction maxima. Monotungstate K2WO4
was identified after 8 h at 5508C.
4. Discussion
The oxoacid corresponding to the Lux–Flood equilib-
rium of nitrate giving oxide ions as base, is the nitronium
cation (NO12 ) which is unstable in the melt (Eq. 2),
producing gaseous NO2. Thus no acids stronger than NO21
can exist in nitrate melts, which means that (quasi)equilib-
rium measurements of electrochemical potentials in such
systems are not possible. Similarly titration of strong acids
with bases is not effective since acid strength levels off at
the hypothetical NO12 species. The domain of thermo-
dynamic stability of molten nitrate is as small as 230 mV at
2308C [23–25] and only 120 mV at 4708C [26]. Thus
potentiometric titration of different oxides in molten
nitrates gave virtually the same curves for CrO3, MoO3,
WO3 or PO23 [27,28], whereas their properties are by no
means similar.
Since the properties of species dissolved in ionic melts
are often similar, being determined mostly by electrostatic
interactions, we can estimate the oxoacidity of the oxides
studied here from some equilibrium measurements made in
other ionic melts. The values of pK measured in KCl–
Fig. 4. Mass-spectroscopic signals of gases evolved during heating of a
mixture of KNO3 and WO3 in molar ratio 10:1.

102
P. Afanasiev, D.H. Kerridge / Journal of Alloys and Compounds 322 (2001) 97–102
NaCl melts at 7008C for the following reactions have been
published by Cherginets [27]:
anionic layers or complex chains of transition metal
polymer oxospecies, separated by K1 cations [32]. In
contrast, WO3 has a three-dimensional interconnected
structure [33] and reacts with much greater difficulty,
giving small oligomeric species. This trend can be con-
tinued since niobium oxide is not reactive at all in molten
KNO3 up to 5508C, i.e. in the range where the reaction can
still be distinguished from nitrate decomposition.
MoO3 1 O22 MoO242 pK 5 8.32
(13)
(14)
(15)
WO3 1 O22 WO242 pK 5 9.31
V O5 1 O22 2VO23 pK 5 6.95
2
The order of acidity WO3 .MoO3 .V O5 follows from
2
these data, which is the inverse of the sequence of
reactivity observed here. However, it seems evident that
the difference in the strength of various acids should in
general correspond to the difference in their reactivity
unless some kinetic hindrance is involved.
References
[1] M. Mobin, U. Malik, J. Alloys Comp. 235 (1996) 97.
[2] P. Afanasiev, C. Geantet, Coord. Chem. Rev. 178/180 (1998) 1725.
[3] G. Charlot, B. Tremillon, in: Les reactions chimiques dans les
´
solvants et les sels fondus, Gauthier-Villars, Paris, 1963.
[4] J. Vallet, J. Chem. Thermodyn. 4 (1972) 105.
[5] M.J. Maeso, J. Largo, Thermochim. Acta 223 (1996) 145.
[6] T. Jriri, M. Gilbert, J. Rogez, J.C. Mathieu, Ann. Chim. Fr. 19
(1994) 121.
[7] D.H. Kerridge, in: J.J. Lagovski (Ed.), Chemistry of Non-aqueous
Solvents, Vol. VB, Academic Press, New York, 1978, p. 270.
[8] H.M.A. Abood, D.H. Kerridge, Thermochim. Acta 215 (1993) 183.
[9] D.A. Habboush, D.H. Kerridge, Thermochim. Acta 101 (1974) 87.
[10] N. Coumert, M. Porthault, J.C. Merlin, Bull. Soc. Chim. Fr. 33
(1965) 910.
[11] N. Coumert, M. Porthault, J.C. Merlin, Bull. Soc. Chim. Fr. 35
(1967) 332.
[12] R.N. Kust, Inorg. Chem. 6 (1997) 2239.
[13] J.M. Schlegel, R. Bauer, Inorg. Chem. 11 (1972) 909.
[14] A.M. Shams El Din, A.A.E. Hosary, J. Electroanal. Chem. 7 (1964)
464.
[15] M. Hassanien, E. Kordes, Zeit. Anorg. Chem 387 (1972) 1.
[16] A.M. Shams El Din, A.A.E. Hosary, J. Electroanal. Chem. 9 (1965)
349.
[17] M. Hassanien, Zeit. Anorg. Chem 399 (1973) 125.
[18] G. Tridot, G. Nowokroki, J. Nicole, M. Wozniak, Comptes Rendus
Acad. Sci. Paris Ser. C 270 (1970) 204.
[19] M.V. Mokhasaev, S.M. Aleikina, F.I. Fedorov, Russ. J. Inorg. Chem.
11 (1966) 644.
[20] P. Afanasiev, C. Geantet, D.H. Kerridge, J. Mater. Chem. 5 (2)
(1995) 347.
[21] P. Afanasiev, C. Geantet, M. Lacroix, M. Breysse, J. Catal. 162
(1996) 143.
The electronegativity of the metals is 1.56 for Mo and V
and 1.59 for W [29]. Therefore, the electronegativity
sequence for oxides is WO3 .MoO3 .V O5. It coincides
2
with the row of oxoacidity cited above and does not
correspond to the sequence of reactivity observed here:
V O5 .MoO3 .WO3. To resolve this apparent contradic-
2
tion, the exact mechanism of the reactions must be
considered, which is seemingly not the same for the three
oxides studied. The initial step of the reactions studied is
formation of hexavanadate, trimolybdate and ditungstate,
respectively. Therefore the nuclearity of the first observed
species changes in the sequence V.Mo.W (which is the
same as the observed reactivity). Thus even if the total
reaction for vanadium has a lower equilibrium constant, its
initial step might be energetically more favorable than for
tungsten oxide. As a result we observe that V O5 reacts at
2
a lower temperature than WO3.
Note that in the sequences of transformations considered
here, every product reacted more slowly than its parent
species (that is why separate reaction maxima peaks were
observed). For example, trimolybdate is less reactive than
tetramolybdate, etc. because the oxoacidity of the sub-
sequent species progressively diminishes upon addition of
new oxygen anions.
The question then arises as to what is the reason for this
difference in reaction mechanisms. It can be supposed that
it stems from the difference in the structure of the
[22] D.R. Lide, H.P.R. Frederikse (Eds.), CRC Handbook of Chemistry
and Physics, CRC Press, Boca Raton, FL, 1996, pp. 4–78.
[23] P.G. Zambonin, J. Jordan, J. Am. Chem. Soc. 91 (1969) 2225.
[24] P.G. Zambonin, J. Electroanal. Chem., 24 (1970) 28, 365.
[25] L.G. Boxall, K.E. Johnson, J. Electrochem. Soc. 118 (1971) 885.
[26] G. Picard, T. Flament, B. Tremillon, J. Electrochem. Soc. 132
(1985) 863.
[27] V.L. Cherginets, Russ. Chem. Rev., 66 (1997).
[28] A.M. Shams El Din, A.A.A. Gerges, Electrochim. Acta 9 (1964)
613.
corresponding oxides. Both V O5 and MoO3 are known to
2
have a lamellar structure [30,31] and can be easily
intercalated with many guest species, e.g. organic mole-
cules or cations such as Li1 condensed Al(III) polycations.
Thus it seems possible that when reaction begins in the
solid state, the alkali metal cations penetrate first between
the layers of transition metal oxide, followed by Lux–
Flood donation of oxygen ion from nitrate to the layer of
transition metal oxide to compensate the additional charge.
Consequently the first intermediate products of reaction are
the lamellar or chain polysalts, consisting of bidimensional
`
[29] J.P. Jolivet, De la solition a l’oxyde, CNRS Editions, Paris, 1994.
[30] J.A.A. Ketelaar, Nature 137 (1936) 316.
[31] L. Kihlborg, Arkiv foer Kemi 21 (1963) 357.
[32] M. Seleborg, Acta Chim. Scand. 20 (1996) 2195.
[33] B.O. Loopstra, Acta Crystallogr. B 24 (1982) 1968.