Observation of the ammonium salt of 12-molybdophosphoric acid by in situ raman spectroscopy during solid-state synthesis: Spectral analysis and reconstruction using the band-target entropy minimization (BTEM) algorithm

Article abstract of DOI:10.1021/ic801609d

The solid-state reaction between ammonium heptamolybdate (AHM) and zirconium phosphate (ZrP) to give the ammonium salt of 12-molybdophosphoric acid (AMPA) was performed at 25-400 °C and monitored using in situ Raman spectroscopy. Spectral analysis of the

Full text of DOI:10.1021/ic801609d

Inorg. Chem. 2009, 48, 1967-1973
Observation of the Ammonium Salt of 12-Molybdophosphoric Acid by in
Situ Raman Spectroscopy during Solid-State Synthesis: Spectral
Analysis and Reconstruction Using the Band-Target Entropy
Minimization (BTEM) Algorithm
Chilukoti Srilakshmi,* Wee Chew, Kanaparthi Ramesh, and Marc Garland*
Institute of Chemical and Engineering Sciences (ICES), 1, Pesek Road, Jurong Island, Singapore, 627833
Received August 24, 2008
The solid-state reaction between ammonium heptamolybdate (AHM) and zirconium phosphate (ZrP) to give the
ammonium salt of 12-molybdophosphoric acid (AMPA) was performed at 25-400 °C and monitored using in situ
Raman spectroscopy. Spectral analysis of the Raman data using the band-target entropy minimization (BTEM)
algorithm resulted in spectral estimates for the starting materials and product, AHM, ZrP, and AMPA, as well as
the byproduct MoO3 and an intermediate 11(NH4)2O · 4(MoO3)7. The time-dependent relative concentration profiles
were obtained, and the contributions of the individual signal intensities of each constituent to the total measured
signal intensity were determined (range: 8.4-27.2%). The present results are important since the synthesis of
AMPA is normally performed in buffered aqueous solution and not in the solid state. The present study also
indicates that a maximum yield of the desired ammonium salt of 12-molybdophosphoric acid is achieved by stopping
the solid-state reaction at ca. 350 °C. The combined spectroscopic and chemometric approach used in this contribution
appears applicable to other solid-state synthetic studies in order to reveal more detailed time-dependent information
on the species present.
1. Introduction
dissolution of the appropriate amounts of the precursors, that
is, HPA and ammonium carbonate/nitrate in water, followed
by precipitation at controlled pH, and the extraction of the
solvent,4 and (2) direct precipitation by dissolution of the
appropriate amounts of the cations in the solution containing
the dissolved acids.5
We previously reported an alternate synthesis, by interact-
ing the phosphate ion of the metal phosphate with molyb-
denum species of the AHM, thus avoiding corrosive reagents
such as H3PO4.6-9 A few other studies have also reported
the formation of heteropoly compounds by reacting the fluid
12-Molybdophosphoric acid (MPA), an important het-
eropolycompound, has been extensively used as a catalyst
for a variety of chemical reactions.1 Bulk MPA exhibits the
well-known Keggin structure2 with 12 edge-sharing moly-
bdenum(VI) MoO6 octahedra surrounding the central PO4
tetrahedron. In spite of the importance of MPA as a catalyst,
its wider application is restricted because of its complexity,
poor reproducibility of the methods of synthesis, and low
thermal stability. These drawbacks limit the practical ap-
plication of MPA. For example, apart from the calcination
temperature, the rate at which this temperature is reached
also affects the final structure.3 In order to overcome these
difficulties, several methods have been reported for the
preparation of the salts of heteropolyacid exhibiting the
Keggin structure. Two of the important routes are (1)
(4) McGarvey, G. B.; Moffat, J. B. J. Catal. 1991, 130, 483.
(5) Van Veen, J. A. R.; Hendriks, P. A. J. M.; Andrea, R. R.; Romers,
E. J. G. M.; Wilson, A. E. J. Phys. Chem. 1990, 94, 5282.
(6) Sai Prasad, P. S.; Raghavan, K. V.; Suryanarayana, I.; Rao, K. N.;
Srilakshmi, C.; Santosh kumar, M.; Reddy, K. B. U.S. Patent
6,534,435, 2003.
* To whom correspondence should be addressed. E-mail: c_srilakshmi@
ices.a-star.edu.sg (C.S.), marc_garland@ices.a-star.edu.sg (M.G.).
(1) Mizuno, N.; Misono, M. Chem. ReV. 1998, 98, 199.
(7) Srilakshmi, C.; Narasimha Rao, K.; Lingaiah, N.; Suryanarayana, I.;
Sai Prasad, P. S. Catal. Lett. 2002, 83, 127.
(8) Srilakshmi, C.; Ramesh, K.; Anderson, B. G.; Niemantsverdriet, J. W.;
Sai Prasad, P. S. Appl. Catal. A 2005, 296, 54.
(9) Srilakshmi, C.; Lingaiah, N.; Sai prasad, P. S.; Narayana, K. V.; Martin,
A.; Lucke, B. Catal. Commun. 2004, 5, 199.
(2) Kozhevnikov, I. V. Chem. ReV. 1998, 98, 171.
(3) Albonetti, S.; Cavani, F.; Trifiro, F.; Gazzano, M.; Koutyrev, M.; Aissi,
F. C.; Aboukais, A.; Guelton, M. J. Catal. 1994, 146, 491.
10.1021/ic801609d CCC: $40.75  2009 American Chemical Society
Inorganic Chemistry, Vol. 48, No. 5, 2009 1967
Published on Web 01/12/2009

Srilakshmi et al.
phase with one or more of the species in the solid phase.10,11
Hence, it is interesting to examine the possibility of generat-
ing the salts of MPA, in their Keggin structure, by solid-solid
interaction between a phosphate support and a molybdenum
salt.
In addition, in situ Raman spectroscopy of this solid-state
reaction can provide interesting information concerning the
reaction chemistry. Indeed, Raman spectroscopy has, over
the past few years, contributed to considerable progress in
the area of heterogeneous catalysis because of its in situ
capabilities and its ability to identify different metal oxide
structures. The novelty of the present investigation is the
formation of the ammonium salt of MPA (AMPA), by
solid-solid interaction between (NH4)6Mo7O24 · 4H2O and
zirconium phosphate, and the use of advanced chemometric
techniques in order to better understand the solid-state
syntheses. In this manner, issues pertaining to pH control
are avoided. To the best of our knowledge, this approach
had not been previously reported.
With regard to the solid-state reaction chemistry, we have
successfully applied band-target entropy minimization (BTEM)
spectral reconstruction methods to the in situ Raman data in
order to reconstruct the pure component spectra of the
components observed during the AMPA formation. BTEM
is a self-modeling curve resolution computational method
that has recently been developed. It has been shown to be
capable of separating the spectra of “pure components” from
matrices of mixture spectra12-18 without any recourse to a
priori information such as spectral libraries. The BTEM
algorithm has been successfully applied to the analysis of
data matrices generated by a number of different spectro-
scopic techniques including IR, infrared reflection absorption
spectroscopy, X-ray diffraction (XRD), NMR, and FT-
Raman.19-23 Although originally developed for vibrational
spectroscopic studies of liquid-phase reactions, BTEM has
recently been applied to heterogeneously catalytic systems.
Thus, application of BTEM to DRIFTS data from CO and
NO adsorption studies on Pt and Pd has resulted in the
reconstruction of pure component spectra of minor species,
Figure 1. Reference Raman spectra of the starting complexes and product
ammonium heptamolybdate, zirconium phosphate, and AMPA used in this
study.
that is, those having less than 0.05% of the total signal
intensity such as isocyanate species.24-26 In addition, some
previously unreported species were identified.
The present study represents the first application of in situ
Raman spectroscopic data and BTEM analysis to study a
solid-state synthesis. These experiments cover a variety of
temperatures and resulted in the reconstruction of the pure
component spectra obtained during solid-state reaction
between AHM and ZrP. It appears that the present combined
approach should be applicable to the study of other solid-
state syntheses.
2. Experimental Methods
Zirconium phosphate (P/Zr ratio of 2.0) was prepared by adding
zirconium nitrate to a vigorously stirred aqueous solution of HCl
(2 mol) and H3PO4 (4 mol) to yield a gelatinous precipitate which
was washed with 2% H3PO4 (aq). Upon drying the sample at 120
°C overnight, amorphous ZrP was obtained. The obtained amor-
phous ZrP was calcined at 400 °C for 4 h. Ammonium heptamo-
lybdate (Sigma-Aldrich) with a purity of 99.98% was used without
any further purification. Purified air was used during the solid-state
reactions (ca. 99.98% Soxal, Singapore). The measured reference
Raman spectra of the starting complexes are shown in Figure 1.
The in situ Raman study involved two different initial ratios of
reagents, namely, (1) AHM/ZrP ) 0.7 and (2) AHM/ZrP ) 1.0.
The solids were mixed thoroughly using a mortar and pestle to a
fine powder and loaded into the in situ Raman cell (Linkam Model
CCR 1000, U.K.) fitted with ZnSe windows for measurements.
2.1. Raman Microscopy. Raman spectra were measured using
a Raman microscope (In Via Reflex, Reni-Shaw) equipped with a
Peltier cooled deep-depleted CCD array detector (576 × 384 pixels)
and a high grade Leica microscope. The Raman scattering was
(10) Kasztelan, S.; Payen, E.; Moffat, J. B. J. Catal. 1988, 112, 320.
(11) Banares, M. A.; Fierro, J. L. G.; Moffat, J. B. J. Catal. 1993, 142,
406.
(12) Chew, W.; Widjaja, E.; Garland, M. Organometallics. 2002, 21, 1982.
(13) Li, C.; Widjaja, E.; Chew, W.; Garland, M. Angew. Chem., Int. Ed.
2002, 41, 3785.
(14) Li, C.; Widjaja, E.; Garland, M. J. Am. Chem. Soc. 2003, 125, 5540.
(15) Widjaja, E.; Li, C.; Chew, W.; Garland, M. Anal. Chem., 2003, 75,
4499.
(16) Widjaja, E.; Crane, N.; Chen, T.-S.; Morris, M. D.; Ignelzi, M. A.,
Jr.; McCreadie, B. R. Appl. Spectrosc. 2003, 57, 1353.
(17) Srilakshmi, Ch.; Widjaja, E.; Anderson, B. G.; Garland, M. Talanta
2007, 72, 847.
(18) Srilakshmi, Ch.; Widjaja, E.; Garland, M.; Anderson, B. G. J. Raman.
Spectrosc. 2007, 38, 349.
(19) Liangfeng, G.; Kooli, F.; Garland, M. Anal. Chim. Acta 2004, 517,
229.
(20) Liangfeng, G.; Wiesmath, A.; Sprenger, P.; Garland, M. Anal. Chem.
2005, 77, 655.
(21) Ong, L. R.; Widjaja, E.; Stanforth, R.; Garland, M. J. Raman Spectrosc.
2003, 34, 282.
(22) Cheng, S.; Rajarathnam, D.; Meiling, T.; Garland, M. Appl. Spectrosc.
2006, 60 (5), 521.
(24) Li, C.; Widjaja, E.; Garland, M. J. Catal. 2003, 213, 126.
(25) Srilakshmi, C.; Widjaja, E.; Feng, G.; Huajun, Z.; Anderson, B. G.;
Hans Niemantsverdriet, J. W. Phys. Chem. Chem. Phys., 2008, 10,
3535.
(23) Boon Hong, K.; Wee-Sun, S.; Chew, W. Anal. Chim. Acta 2006, 571
(1), 113.
(26) Srilakshmi, C.; Feng, G.; Anderson, B. G.; Hans Niemantsverdriet,
J. W. Phys. Chem. Chem. Phys. 2008, 10, 5510.
1968 Inorganic Chemistry, Vol. 48, No. 5, 2009

Ammonium Salt of 12-Molybdophosphoric Acid
excited with a red laser (785 nm), and a 20× objective lens was
used to collect the backscattered light. A spot size of ca. 2 µm
was used. All of the spectra were acquired with a laser power of
ca. 111-230 µW with an exposure time of 10 s over the range of
100-1500 cm-1
.
2.2. In Situ Raman Spectroscopic Studies. A 30 mg sample
of powder (AHM and ZrP) mixed thoroughly in an agate mortar
was loaded into the in situ Raman cell, and the temperature was
raised from room temperature (Rt) to 400 °C in a purified air flow
of 20 mL/min. The temperature was increased in 50 °C intervals
at a rate of 10 °C/min and then held constant for ca. 30 min. This
was repeated until the final temperature of 400 °C was reached.
The sample was kept isothermal for another 120 min at 400 °C.
In experiment 1 with AHM/ZrP ) 0.7, 33 spectra were acquired,
and in experiment 2 with AHM/ZrP ) 1.0, 62 spectra were
acquired.
3. Computational Aspects
3.1. Band-Target Entropy Minimization (BTEM) Algorithm.
As previously mentioned, the primary use of BTEM is to extract
the pure component spectra from a set of multicomponent mixture
spectra, without recourse to a priori information of any kind,
including spectral libraries. Afterward, the contributions of each
individual pure component spectrum to the measured experimental
spectra can be determined using multilinear regression in order to
estimate the signal contribution of each species as well as the
relative concentrations.
Let Ik×ν represent the Raman intensity in the consolidated
spectroscopic data matrix, where k denotes the number of spectra
taken and ν denotes the number of data channels associated with
the spectroscopic range. The experimental intensities Ik×ν have a
bilinear data structure and can be described as a product of two
matrices, namely, the concentration matrix Ck×s and the Raman
pure component scattering coefficient matrix Js×ν (where s denotes
the number of observable species in the chemical mixture).
Accordingly, the associated error matrix εk×ν contains both experi-
mental error and model error (nonlinearities).27
Figure 2. Selected in situ Raman spectra of the solid-state reaction of
ammonium heptamolybdate and zirconium phosphate at various tempera-
tures.
equation representing the transformation of the right singular vectors
ˆ
into each spectral estimate is given by eq 3, where J1×ν is the BTEM
estimated pure component Raman spectrum, T1×z is the transforma-
tion vector, and VT is the set of basis vectors used as the basis
z×ν
set.
T
ˆ
J
1×ν ) T1×zVz×ν
(3)
ˆ
Each pure Raman spectral estimate J1×ν is obtained by targeting
bands in the first few VT vectors. The BTEM algorithm then retains
this selected feature and, at the same time, returns an entire spectrum
which has minimum entropy. Accordingly, the resulting spectral
estimate is the simplest pattern present. This procedure is repeated
for all important observable physical features in the selected VT
vectors. A superset of reconstructed pure component spectra is
obtained. This superset is then reduced to a smaller set of
representative spectral patterns to eliminate spectral redundancies.
This results in an enumeration of all observable pure component
spectra.
I
k×ν ) Ck×s
J
s×ν + εk×ν
(1)
The BTEM algorithm proceeds by first decomposing Ik×ν into its
singular vectors using singular value decomposition (SVD), as
shown in eq 2.28 This is then followed by the appropriate
transformation of the right singular vectors, VT, into physically
meaningful pure component spectra. See section 4.2.1 for more
details.
In the present study, a total of 95 spectra were measured, 33 in
the first experimental run and then 62 in the second experiment.
I
k×ν ) Uk×k
Σ
k×νVTν×ν
(2)
The combined intensity matrix had dimensions I95×1414
.
In contrast to other self-modeling curve resolution methods, BTEM
was uniquely developed to resolve one pure component spectrum
at a time. The number of right singular vectors, z, taken for inclusion
in the transformation, is usually much larger than s, to account for
any nonlinearities present (nonstationary spectral characteristics).
The number z is usually determined by identifying the vectors which
possess localized spectral bands or features of clear physical
significance and retaining these, while discarding the vectors that
are associated with randomly distributed noise. The principal
4. Results and Discussion
4.1. Raman Spectroscopy. Some selected and representa-
tive in situ Raman spectra from the solid-state reaction
between AHM and ZrP, measured at various temperatures,
are presented in Figure 2. These spectra show that some
considerable baseline shifts occur. The low-temperature
spectra show some relatively significant background or
fluorescence. Above ca. 250 °C, this background more-or-
less disappears. In addition, spikes (due to cosmic rays) can
be seen in a few of the spectra. Clearly, there is considerable
change in the prominent Raman bands as the solid-state
reaction proceeds.
(27) Garland, M.; Visser, E.; Terwiesch, P.; Rippin, D. W. T. Anal. Chim.
Acta 1997, 35, 1337.
(28) Golub G. H.; Van Loan, C. F. Matrix Computations; Johns Hopkins
University Press: Baltimore, 1996.
Inorganic Chemistry, Vol. 48, No. 5, 2009 1969

Srilakshmi et al.
Table 1. Characteristic Raman Peak Assignments of the Species
Observed in the Present Study8,29-38
Spectra measured of the initial powder mixture at Rt
showed bands at 1091 and 943 cm-1. The peak at 1091
corresponds to ZrP, and the peak at 943 cm-1 corresponds
to 11(NH4)2O·4(MoO3)7. The latter was apparently generated
during the solid-phase grinding and sample preparation.
11(NH4)2O · 4(MoO3)7 is one of the intermediates found
during the thermal decomposition of ammonium heptamo-
lybdate to MoO3 using Raman spectroscopy by Murugan and
Chang.29 A similar type of intermediate has been observed
by others using IR, XRD, and thermogravimetric-differential
thermal analysis.30-33 There are no significant changes
observed in the spectra measured from 100 to 200 °C.
The spectra observed from 250-400 °C indicated major
changes. At 250 °C, the spectra show a mixture of phases
containing both AMPA at 1089, 975, 965, 920, 911, 887,
632, and 247 cm-1 and less intense bands of 11(NH4)2O ·
4(MoO3)7 at 945, 932, 844, and 220 cm-1. It can be seen
that the band at 942 cm-1 for 11(NH4)2O · 4(MoO3)7 shifts
to 945 cm-1 as the temperature is increased from Rt to 250
°C. Salts of heteropolyacids can be present as aggregates of
mono-oxo species. The main spectroscopic features of
AMPA having a Keggin structure were observed at 1089
cm-1 (νas P-O, weak), 975 cm-1 (νs Mo-Od), 965 cm-1
(νas Mo-Od), 920-877 cm-1 (νas Mo-Ob-Mo), 632
cm-1(νas Mo-Oc-Mo), and 247 cm-1 (νs Mo-Oa). The
spectral features observed for ZrP, 11(NH4)2O · 4(MoO3)7,
and AMPA agree with those reported in the literature.29,34-36
A summary of the observed bands and their assignments are
given in Table 1. Above 300 °C, bands corresponding to
11(NH4)2O · 4(MoO3)7 disappear, and bands due to AMPA
continue to increase up to 400 °C.
Species
Raman shift
assignment
AHM
935
902
892
885
859
840
1091
ν(Mo-O)
ν(Mo-O-Mo)
ν(Mo-O-Mo)
ν(P-O)
ZrP
11(NH4)2O·4(MoO3)7 945
ν(Mo-O)
932
844
975, 965
920-877 νas(Mo-Od)
632
247
ν(Mo-O-Mo)
νs(Mo-Od)
AMPA
νas(Mo-Ob-Mo)
νs(Mo-Oa) with bridging character
νas(OdModO) twist
νs(OdModO) twist
MoO3 (orthorhombic) 994
820
663
471
378
338
285
243
977
899
882
695
283
248
220
ν(O-Mo-O)
ν Mo-O-Mo stretch and bend
δ(OdModO) scissors
δ(Mo-O-Mo) bend
δ(OdModO) wagging
τs(OdModO) twist
MoO3 (hexagonal)
ν(OMo)
corner modes of ν (Mo-O-Mo)
edge mode of ν (OMo3)
δ(OMo3)
δ(OMo2)
δ(OMo2)
(normally called the right singular vectors) represent a set
of basis vectors which contain abstract information on the
pure component spectra of the observable components
present in the system. These basis vectors are ordered
according to their contribution to the total variance in the
observations. Hence, the initial few vectors are associated
with the chemically significant information in the system,
and the remaining vectors are primarily associated with the
random instrumental error. In addition, it should be noted
that, after decomposition, there are a total of 1414 basis
vectors in VT, but only 95 vectors are physically meaningful
(possessing either chemical or random error signals). This
is due to the fact that there are only 95 spectra in the original
data set.
Since not all 95 right singular vectors are needed in further
analysis (many of the last vectors are primarily noise), the
set of vectors can be reduced, and only a subset will be used
in the BTEM spectral reconstructions. Figure 3 shows some
of the first 20 right singular vectors. The first two vectors
show many localized signals superimposed on large back-
ground signals. The next eight vectors have clear localized
signals, but this is often accompanied by a wavy baseline
(see vector 8). Vectors 11-20 are primarily noise, but they
are heteroscedasticly distributed. The remaining vectors were
essentially white noise. Accordingly, 20 VT right singular
vectors were used in the subsequent BTEM spectral recon-
structions.
AMPA was stable at 400 °C for ca. 30-40 min and then
started to decompose to hexagonal MoO3, which shows bands
at 899 and 695 cm-1. With increasing time, hexagonal MoO3
was completely transformed into orthorhombic MoO3, which
shows peaks at 820, 994, 663, and 285 cm-1. This can be
seen in Figure 2 from the spectrum at 400 °C with a 120
min reaction time. It is important to note that a replicate
experiment gave similar spectra.
4.2. Pure Component Spectral Reconstruction. 4.2.1.
Singular Value Decomposition. The 95 raw Raman mixture
spectra from the combined experiments were subjected to a
singular value decomposition, yielding two orthonormal
matrices U95×95 and VT
and the diagonal singular
1414×1414
value matrix Σ95×1414. The row vectors of the VT matrix
(29) Murugan, R.; Chang, H. J. Chem. Soc., Dalton Trans. 2001, 3125.
(30) Hanafi, Z. M.; Khilla, M. A.; Askar, M. H. Thermochim. Acta 1981,
45, 221.
(31) Zhoulan, Y.; Xinhai, L.; Qiyuam, C. Thermochim. Acta 2000, 352,
107.
(32) Ma, E. Bull. Chem. Soc. Jpn. 1964, 37, 171.
(33) Ma, E. Bull. Chem. Soc. Jpn. 1964, 37, 648.
(34) Rocchiccioli-Deltcheff, C.; Aouissi, A.; Bettahar, M. M.; Launay, S.;
Fournier, M. J. Catal. 1996, 164, 16.
(35) Kasprzak, M. S.; Crouch, S. R.; Leroi, G. E. Appl. Spectrosc. 1978,
32, 537.
(36) Cavani, F.; Etienne, E.; Mezzogori, R.; Pigamo, A.; Trifiro, F. Catal.
Lett. 2001, 75, 99.
(37) Mestl, G.; Ilkenhans, T.; Spielbauer, D.; Dieterle, M.; Timpe, O.;
Kro¨hnert, J.; Jentoft, F.; Kno¨zinger, H.; Schlo¨gl, R. Appl. Catal. 2001,
210, 13.
4.2.3. Spectral Reconstruction. It is known that, if all of
the chemically significant extrema (minima and maxima) in
the primary right singular vectors are targeted, then a superset
of observable pure component spectra are obtained, in which
many replicates appear. This is the approach that has been
(38) Yurchenko, E. N. J. Struct. Chem. 1987, 28, 446.
1970 Inorganic Chemistry, Vol. 48, No. 5, 2009

Ammonium Salt of 12-Molybdophosphoric Acid
the starting materials AHM, with a maximum at 935 cm-1
(Figure 4a), and ZrP, with maximum at 1090 cm-1 (Figure
4b) (please compare spectral estimates and relative intensities
of bands to references in Figure 1).
The remaining four spectral estimates are assigned to
AMPA with a maximum at 965 cm-1 (Figure 4c),8,37
hexagonal MoO3 at 899 cm-1 (Figure 4d),38 orthorhombic
MoO3 (Figure 4e) with maximum at 820 cm-1,38 and
11(NH4)2O·4(MoO3)7 with a maximum at 945 cm-1 (Figure
4f).29 It should be noted that some of the full-range pure
component spectral estimates shown in Figure 4 contain
small signal artifacts, as well as a moderate level of noise.
4.2.3.2. Partial Range Analysis. The full range BTEM
estimates were of sufficiently good quality for identification
purposes. However, for quantitative analysis, it is frequently
advantageous to repeat the BTEM analysis over a restricted
spectral range. This usually results in significantly improved
signal-to-noise ratios as well as the suppression of most
artifacts.
Figure 3. Some of the first 20 VT vectors obtained by SVD.
The region 600-1300 cm-1 was selected for the partial
range analysis since this region contains numerous bands
for all of the species present. The BTEM analysis resulted
in considerably improved signal-to-noise ratios for all of the
constituent spectra. In addition, these pure component
spectral estimates were then mapped back onto a baseline-
corrected data set in order to determine the relative contribu-
tions of each spectral estimate to the total signal. Figure 5
shows both the new spectral estimates as well as their
contributions in each spectrum. Spectra 1-33 correspond
to experiment 1, and those from 34-95 correspond to
experiment 2.
The relative concentration profiles are, at first glance,
somewhat surprising. Indeed, one would expect a monotonic
decrease in the signals of the reagents as a function of time,
and a monotonic increase in the products as a function of
time. In part, the unexpected concentration profiles are
believed to be the result of the significant density differences
in the compounds present coupled with the agitation/
fluidization of the solid particles that occurs in the Linkam
holder. Indeed, the gas flow is vertical and passes through
the solid particle bed. This probably results in some
stratification of the particles according to their densities over
time. The total signal contribution of each component to the
overall measurements is presented in Table 2.
Figure 4. The results for the BTEM analysis for the full-range spectral
patterns. (a) AHM, (b) ZrP, (c) AMPA [(NH4)3PMo12O40 ·4H2O], (d)
hexagonal MoO3, (e) orthorhombic MoO3, (f) 11(NH4)2O·4(MoO3)7, (g)
visible light.
taken with FTIR for new chemically reactive liquid phase
systems with an unknown number of new and previously
unobserved species. In the present test for a solid-phase
reaction, a similar approach is taken. Accordingly, from the
first 10 VT vectors, significant spectral features were selected
as band targets in the spectral reconstructions, and exhaustive
searches of the data were performed.
4.2.3.3. Further Considerations of Phosphate Vibra-
tion. Although the partial range analyses provided quite
adequate spectral estimates and relative concentration pro-
files, the phosphate vibration in the ZrP spectral estimate is
considerably broader than the other vibrations and therefore
requires further consideration. Additional BTEM analyses
were made on the very restricted interval of 1000-1200 cm-1
(not shown here). Although there is still some noise in the
new spectral reconstructions, it becomes evident that the
broad band seen for ZrP in Figure 5 is in fact a superposition
of two signals, with maxima at 1087 and 1091 cm-1. A pure
reference of ZrP shows a maximum only at 1091 cm-1. The
origin of the superimposed signal with a band maximum at
4.2.3.1. Full-Range Analysis. BTEM was applied to the
full spectral range 100-1500 cm-1 in order to obtain pure
component spectral estimates. A total of seven distinct
patterns were reconstructed, as shown in Figure 4. Six of
these patterns correspond to inorganic species, while that last
pattern corresponds to stray light captured by the measure-
ments. Relatively good spectral estimates are obtained for
Inorganic Chemistry, Vol. 48, No. 5, 2009 1971

Srilakshmi et al.
Figure 5. Partial range BTEM spectral estimates for the inorganic compounds as well as their relative contributions to experimental signal intensity.
Table 2. The Total Signal Contribution of Each Normalized Pure
Component Spectrum to the Overall Measurements from 95 in Situ
Spectra
the observable intermediate 11(NH4)2O · 4(MoO3)7 increase
at a lower temperature, 115 °C, go through a maximum, and
then approach zero at ca. 350 °C. As before, the signals of
both hexagonal and orthorhombic MoO3 increase as the
signals of AMPA and 11(NH4)2O·4(MoO3)7 start to decrease
(see spectral number 70-80 in Figure 5).
Experiments 1 and 2 are also similar because the fate of
zirconium could not be determined from the Raman mea-
surements. Indeed, in both experiments, the signals from ZrP
entirely disappear, but no signals from a new zirconium-
containing phase appear. Since the raw experimental reaction
spectra (Figure 2) show a pronounced increase in the
background signal (average of 2000-5000 counts per
second), it is possible that the zirconium (or a new zirconium
compound) is present in an amorphous state, and no distinct
Raman bands are observed. Therefore, the overall reaction
for the formation of AMPA (up to ca. 300 °C) appears to be
best expressed as
pure component
contribution to total signal intensity (%)
AHM
ZrP
17.9
8.4
orthorhombic MoO3
hexagonal MoO3
11(NH4)2O·4(MoO3)7
AMPA
11.0
9.5
19.5
27.2
1087 cm-1 is not entirely clear, but it apparently arises due
to colinearity with some other signal. In this regard, it can
be noted that AMPA has a phosphate vibration at ca. 1089
cm-1.
4.3. Further Comparison of Experiments 1 and 2.
Experimental runs 1 and 2 differ due to the initial AHM/
ZrP ratios used. In the first experiment, the ratio was 0.7,
and in the second experiment, the ratio was 1. These different
initial compositions appear to have affected the formation
and stability of the desired product, AMPA.
In experiment 1, the signals from AMPA start to increase
at ca. 250 °C, go through a maximum, and then approach
zero at ca. 400 °C. During the same period, the signals of
the observable intermediate 11(NH4)2O · 4(MoO3)7 also in-
crease and then decrease. As the signals of AMPA and
11(NH4)2O · 4(MoO3)7 start to decrease, the signals of both
hexagonal and orthorhombic MoO3 increase (see spectral
numbers 20-25 in Figure 5). This implies that the degrada-
tion of AMPA (and perhaps 11(NH4)2O·4(MoO3)7) to MoO3
at prolonged exposure at 400 °C.
Many of the observations from experiment 2 are similar
to those from experiment 1. For example, the signals from
AMPA start to increase at ca. 200 °C, go through a
maximum, and approach zero at ca. 350 °C. The signals of
3(NH4)2O · (MoO3)7 · 4H2O(s) + ZrP2O7(s) f
(NH4)3PMo12O40 · 4H2O(s) + 9MoO3(s) + 3NH3(g) +
8H2O + 1.5H2(g) + ZrPOx (amorphous solid)
Previous work has been reported on the solid-state degrada-
tion of AHM alone. Murugan and Chang29 reported the
thermal degradation of AHM, initially to 11(NH4)2O ·
4(MoO3)7 and 7(NH4)2O·4(MoO3)7 and finally orthorhombic
MoO3, at ca. 385 °C. The present experiments involving
AHM and ZrP as starting materials are consistent with the
aforementioned observation. Orthorhombic molybdenum
oxide MoO3 is the most stable molybdenum-containing
compound at higher temperatures.
1972 Inorganic Chemistry, Vol. 48, No. 5, 2009

Ammonium Salt of 12-Molybdophosphoric Acid
The differences in the initial AHM/ZrP ratios for experi-
ments 1 and 2, as well as the differences in the thermal
behavior of the two systems, require further consideration.
At the lower AHM/ZrP ratio of 0.7, the degradation of
AMPA appears to occur at 400 °C. At the higher AHM/ZrP
ratio of 1, the degradation of AMPA appears to occur starting
at 350 °C. There appears to be an optimal AHM/ZrP ratio
for maximum AMPA formation. Indeed, Bondareva et al.39
also expressed a similar opinion that, for wet impregnation
methods, any excess of Mo or P above the stoichiometric
composition (above ca. 0.7) required by the heteropolyacid
salt could lead to the formation of the corresponding oxides.
For example, degradation of AMPA to MoO3 can be
expressed as below:
spectroscopy for the first time. This solid-state procedure
avoided the use of the more commonly needed pH control
during synthesis, particularly wet impregnation ap-
proaches. Six pure component spectra of inorganic
compounds were reconstructed using the BTEM algorithm,
and the contributions of the individual signal intensities
of each constituent to the total measured signal intensity
were determined (range: 8.4-27.2%). By combining in
situ Raman spectroscopy and BTEM curve resolution, it
was possible to identify
a transient intermediate,
11(NH4)2O · 4(MoO3)7, during the solid-state synthesis.
This latter finding helps to explain some of the selectivity
and side-product observations. By studying the reaction
until 400 °C, it was possible to confirm that good yield
of the desired product, AMPA, is achieved at 350 °C, and
any further temperature increase leads to its degradation.
With regard to the signal processing issues, it was found
that full-range spectral analyses were possible. However,
the quality of these spectral estimates could be consider-
ably improved by analyzing the data over a slightly more
restricted spectral range. The improvement in spectral
estimation is a direct result of minimizing the integration
of noise which occurs when large spectral ranges are used.
The present combined spectroscopic and chemometric
approach appears applicable to other solid-state syntheses
as well.
2[(NH4)3PMo12O40 · 4H2O(s)] f 24MoO3(s) + P2O5(s) +
6NH3(g) + 11H2O
The present temperature-ramped experiments using AHM
and ZrP as precursors show that the solid-state reaction
should be conducted with a ca. 0.7 ratio of AHM and ZrP
and should be terminated at ca. 350 °C in order to maximize
the yield of the desired product (NH4)3PMo12O40 · 4H2O.
Above this temperature, degradation to MoO3 occurs.
5. Conclusions
The formation of AMPA during the solid-state reaction
between AHM and ZrP was observed using in situ Raman
(39) Bondareva, V. M.; Andrushkevich, T. V.; Plyasova, L. M.; Masksi-
movskaya, R. I.; Chumachenko, N. N. React. Kinet. Catal. Lett. 1998,
63, 201.
IC801609D
Inorganic Chemistry, Vol. 48, No. 5, 2009 1973