17O solid-state NMR and first-principles calculations of sodium trimetaphosphate (Na3P3O9), tripolyphosphate (Na5P3O10), and pyrophosphate (Na 4P2O7)

Article abstract of DOI:10.1021/ic800637p

The assignment of high-field (18.8 T) ¹⁷O MAS and 3QMAS spectra has been completed by use of first-principles calculations for three crystalline sodium phosphates, Na₃P₃O₉, Na ₅P₃O₁₀, and Na₄P₂O ₇. In Na₃P₃O₉, the calculated parameters, quadrupolar constant (C_Q), quadrupolar asymmetry (η_Q), and the isotropic chemical shift (δ_cs) correspond to those deduced experimentally, and the calculation is mandatory to achieve a complete assignment. For the sodium tripolyphosphate Na ₅P₃O₁₀, the situation is more complex because of the free rotation of the end-chain phosphate groups. The assignment obtained with ab initio calculations can however be confirmed by the ¹⁷O{ ³¹P} MAS-J-HMQC spectrum. Na₄P₂O₇ ¹⁷O MAS and 3QMAS spectra show a complex pattern in agreement with the computed NMR parameters, which indicate that all of the oxygens exhibit very similar values. These results are related to structural data to better understand the influence of the oxygen environment on the NMR parameters. The findings are used to interpret those results observed on a binary sodium phosphate glass.

Full text of DOI:10.1021/ic800637p

Inorg. Chem. 2008, 47, 7327-7337
17O Solid-State NMR and First-Principles Calculations of Sodium
Trimetaphosphate (Na P O ), Tripolyphosphate (Na P O ), and
3
3
9
5 3 10
Pyrophosphate (Na P O )
4 2 7
†
†
†
†
Filipe Vasconcelos, Sylvain Cristol, Jean-Francois Paul, Gr e´ gory Tricot, Jean-Paul Amoureux,
†
†
Lionel Montagne, Francesco Mauri, and Laurent Delevoye*
,†
´
UCCS-Unit e´ de Catalyse et Chimie du Solide, UMR CNRS, 8181, Ecole Nationale
Sup e´ rieure de Chimie de Lille, UniVersit e´ des Sciences et Technologies de Lille, BP 108,
5
9652 VilleneuVe d’Ascq Cedex, France, and Institut de Min e´ ralogie et Physique des Milieux
Condens e´ s, UniVersit e´ Pierre et Marie Curie, Campus Boucicaut, 140 rue de Lourmel,
7
5015 Paris
Received April 8, 2008
17
The assignment of high-field (18.8 T) O MAS and 3QMAS spectra has been completed by use of first-principles
calculations for three crystalline sodium phosphates, Na , Na 10, and Na . In Na , the calculated
parameters, quadrupolar constant (C ), quadrupolar asymmetry (η ), and the isotropic chemical shift (δcs) correspond
to those deduced experimentally, and the calculation is mandatory to achieve a complete assignment. For the
sodium tripolyphosphate Na 10, the situation is more complex because of the free rotation of the end-chain
phosphate groups. The assignment obtained with ab initio calculations can however be confirmed by the O{ P}
MAS-J-HMQC spectrum. Na
17O MAS and 3QMAS spectra show a complex pattern in agreement with the
P O
3 3 9
5 3
P O
P O
4 2 7
3 3 9
P O
Q
Q
5 3
P O
17
31
4 2 7
P O
computed NMR parameters, which indicate that all of the oxygens exhibit very similar values. These results are
related to structural data to better understand the influence of the oxygen environment on the NMR parameters.
The findings are used to interpret those results observed on a binary sodium phosphate glass.
1
. Introduction
Nuclear magnetic resonance (NMR) together with X-ray
spinning (MAS). All of these elements make the 31P an
attractive nucleus for the structural study of phosphate
compounds with NMR. The usual strategy includes the
preliminary determination of NMR parameters (chemical
shifts) for a series of similar crystalline phosphate compounds
to further interpret the broad resonances generally observed
in amorphous and disordered materials.
1
diffraction (XRD) are among the most powerful techniques
to study the structure of phosphate-based compounds.
Phosphates occupy an important place in oxide material
solid-state chemistry. Indeed, phosphates can be found as
glasses, ceramics, organic/inorganic hybrids, catalysts, gels,
and so forth. In the case of amorphous forms of phosphates,
NMR has the great advantage to probe the local environment
of the nuclei and does not require any long-range order. P
nucleus is commonly used to characterize the coordination
and cationic surroundings as its chemical shift strongly
depends on these parameters. Moreover, P isotope has
1
2
MHz/T), and is a spin / nucleus, so that a complete
averaging of the anisotropic broadenings resulting from
powder samples can be obtained by using magic-angle
Oxygen has a predominant role in the structure of
1
7
inorganic materials. Its NMR-active isotope, O, is a
5
2
quadrupolar nucleus (with a nuclear spin I ) / ). This means
3
1
that the nuclear nonspherical charge distribution interacts
with the electric field gradient (EFG) at the nucleus. The
knowledge of the EFG can help elucidate the structural
characteristics such as coordination numbers and the geom-
etry of the neighboring nuclei. Then, in the case of a
negligeable chemical shift anisotropy with respect to the
3
1
γ
00% natural abundance, is NMR sensitive ( /2π ) 17.235
1
1
7
MAS speed, the O MAS NMR can be characterized by
(
1) Tricot, G.; Montagne, L.; Delevoye, L.; Palavit, G. In Recent AdVances
in Solid-State NMR of Inorganic Polymeric Phosphate Materials; De
Jaeger, R., Gleria, M., Eds.; Nova Science Publishers, Inc.: New York,
USA, 2007; Chapter 13, pp 663-710.
*
To whom correspondence should be addressed. E-mail:
laurent.delevoye@ensc-lille.fr.
†
Universit e´ des Sciences et Technologies de Lille.
1
0.1021/ic800637p CCC: $40.75  2008 American Chemical Society
Inorganic Chemistry, Vol. 47, No. 16, 2008 7327
Published on Web 07/22/2008

Vasconcelos et al.
three NMR parameters: the isotropic chemical shift (δcs), the
quadrupolar constant (C ), and the quadrupolar asymmetry
).
However, 17O NMR is much less popular than 31P NMR
literature must be used with caution when they are considered
out of their initial range of application. In amorphous
systems, the distribution of NMR parameters is often related
Q
(η
Q
2
to a distribution in bond lengths and angles. They are well
1
6
in solid phosphates because of several drawbacks: i) its low
characterized in silicate materials but much less in
phosphates.
γ
natural abundance (0.038%), ii) its low sensitivity ( /2π
)
-
5.7716 MHz/T), and iii) the fact that it shows, even under
First-principles calculations are often carried out as a
complement to the experimental and empirical approaches.
Until recently, the first-principles calculations of O NMR
MAS conditions, a second-order quadrupolar broadening. An
isotopic O enrichment, combined with the use of a very
high magnetic field, can partly circumvent these drawbacks.
Furthermore, modern 2D experiments, that is, the multiple-
1
7
2,3
17
4
parameters had been applied to cluster systems (i.e., isolated
1
7,18
molecules or solid aggregates) in silicates,
aluminosili-
5
,6
19
20,21
quantum MAS
(MQMAS) and the satellite-transition
cates, and phosphates.
These studies reveal that the
7
17
MAS (STMAS), which are dedicated to quadrupolar nuclei
with half-integer spin quantum number, were developed to
eliminate completely the residual second-order quadrupolar
broadening and to give high-resolution spectra. The MQMAS
O quadrupolar parameters depend mainly on the first
coordination sphere surrounding the oxygen. For the chemi-
cal shift parameters, the first coordination sphere is not
sufficient especially in ionic solids where the long-range
electrostatic potential has a large influence. In such systems,
the correct description of the charge density requires an
increasing number of atoms to be defined in the cluster,
resulting in a large increase of the required computational
resources. Different alternatives can be used to circumvent
8
experiment has the advantage of being easy to implement
even if it suffers from a lack of sensitivity of the triple-
quantum excitation and reconversion transfers. STMAS can
yield substantial signal enhancements compared to MQMAS,
but the presence of the first-order quadrupolar interaction in
the satellite transitions imposes a high degree of accuracy
for the magic-angle setting. In recent years, the efficiency
of these sequences has been greatly improved in many ways
such that it is now possible to explore the structure through
2
2
this problem. Di Fiori et al. have proposed an embedding
scheme to take into account the electrostatic potential in
cluster calculations of NMR shielding tensors. Another
approach is to use the 3D periodicity of the solid with
periodic Density Functional Theory (DFT) calculations, as
95
9
39
25
10
the NMR study of low-γ nuclei like Mo, K, Mg, and
of course O (for a good review of O NMR developments
1
7
17
23
proposed by Mauri et al. to estimate the solid response to
see ref 11).
the presence of a uniform magnetic field. Later, the Gauge
Including Projector Augmented Waves (GIPAW) approach
1
7
The assignment and interpretation of the O MAS spectra,
even with a good signal-to-noise ratio and the use of high-
resolution sequences at a very high magnetic field, may still
be ambiguous due to the lack of an extensive database of
NMR parameters for this nucleus. If such a O database
exists for a series of organic compounds,
2
4
by Pickard and Mauri added, to this initial development,
2
5
the PAW method. This development gave access to the
all-electron magnetic response in a pseudopotential formal-
1
7
17
ism. This method has been confronted to experimental
O
1
2,13
26
it is much more
NMR for different materials including zeolites, sodium
silicates, crystalline aluminosilicates, calcium alumino-
2
7
28
time demanding to develop such a database for inorganic
materials, as it is specific to the nature of the cationic
2
9
30,31
silicate glasses, and minerals.
1
7
environment. Since the pioneering studies on O solid-state
In the present article, we will show how periodical DFT
14,15
NMR,
empirical assignments have been used to interpret
the NMR response in terms of the local environment.
However, these empirical correlations collected in the
(16) Clark, T. M.; Grandinetti, P. J.; Florian, P.; Stebbins, J. F. Phys. ReV.
B 2004, 70, 064202.
(
17) Clark, T. M.; Grandinetti, P. J. Solid State Nucl. Magn. Reson. 2005,
2
7, 233.
(
(
(
2) Zeyer, M.; Montagne, L.; Kostoj, V.; Palavit, G.; Prochnow, D.; Jaeger,
(18) Xue, X.; Kanzaki, M. Phys. Chem. Minerals 1998, 26, 14.
(19) Kubicki, J.; Toplis, M. Am. Mineral. 2002, 87, 668.
(20) Alam, T. M.; Segall, J. M. J. Mol. Struc.: Theochem 2004, 674, 167.
(21) Cherry, B. R.; Alam, T. M.; Click, C.; Brow, R. K.; Gan, Z. J. Phys.
Chem. B 2003, 107, 4894.
C. J. Non-Cryst. Solids 2002, 311, 223.
3) Flambard, A.; Montagne, L.; Delevoye, L. Chem. Commun. 2006,
3
426.
4) Gan, Z.; Gorkov, P.; Cross, T. A.; Samoson, A.; Massiot, D. J. Am.
Chem. Soc. 2002, 124, 5634.
(22) Di Fiori, N.; Orendt, A. M.; Caputo, M. C.; Ferraro, M. B.; Facelli,
J. C. Magn. Reson. Chem. 2004, 42, S41.
(
(
5) Frydman, L.; Hardwood, J. J. Am. Chem. Soc. 1995, 117, 5367.
6) Medek, A.; Harwood, J. S.; Frydman, L. J. Am. Chem. Soc. 1995,
(23) Mauri, F.; Pfrommer, B. G.; Louie, S. G. Phys. ReV. Lett. 1996, 77,
5300.
1
17, 12779.
(
(
7) Gan, Z. J. Am. Chem. Soc. 2000, 122, 3242.
(24) Pickard, C. J.; Mauri, F. Phys. ReV. B 2001, 63, 245101.
(25) Blöchl, P. E. Phys. ReV. B 1994, 50, 17953–17979.
(26) Profeta, M.; Mauri, F.; Pickard, C. J. J. Am. Chem. Soc. 2003, 125,
541–548.
8) Amoureux, J.-P.; Fernandez, C.; Steuernagel, S. J. Magn. Reson. 1996,
A123, 116.
(
9) d’Espinose de Lacaillerie, J.-B.; Barberon, F.; Romanenko, K. V.;
Lapina, O. B.; Le Pollès, L.; Gautier, R.; Gan, Z. J. Phys. Chem. B
(27) Charpentier, T.; Ispas, S.; Profeta, M.; Mauri, F.; Pickard, C. J. J.
Phys. Chem. B 2004, 108, 4147–4161.
2
005, 109, 14033.
(
10) Dowell, N. G.; Ashbrook, S. E.; Wimperis, S. J. Phys. Chem. B 2004,
(28) Gervais, C.; Profeta, M.; Babonneau, F.; Pickard, C. J.; Mauri, F. J.
Phys. Chem. B 2004, 108, 13249–13253.
1
08, 13292.
(
(
(
11) Ashbrook, S.; Smith, M. E. Chem. Soc. ReV. 2006, 35, 718.
12) Wu, G. Prog. Nucl. Magn. Spect. 2008, 52, 118.
(29) Benoit, M.; Profeta, M.; Mauri, F.; Pickard, C. J.; Tuckerman, M. E.
J. Phys. Chem. B 2005, 109, 6052–6060.
13) Lema ˆı tre, V.; Smith, M. E.; Watts, A. Solid State Nucl. Magn. Reson.
(30) Ashbrook, S. E.; Le Poll e` s, L.; Pickard, C. J.; Berry, A. J.; Wimperis,
S.; Farnan, I. Phys. Chem. Chem. Phys. 2007, 9, 1587.
(31) Ashbrook, S. E.; Berry, A. J.; Frost, D. J.; Gregorovic, A.; Pickard,
C. J.; Readman, J. E.; Wimperis, S. J. Am. Chem. Soc. 2007, 129,
13213.
2
004, 26–215.
(
(
14) Schramm, S.; Oldfield, E. J. Am. Chem. Soc. 1984, 106, 2502.
15) Timkem, H. C.; Turner, G. L.; Gilson, J. P.; Welsh, L. B.; Oldfield,
E. J. Am. Chem. Soc. 1986, 108, 7231.
7328 Inorganic Chemistry, Vol. 47, No. 16, 2008

17O Solid-State NMR and First-Principles Calculations
calculations can help in the assignment and interpretation
3. Calculation Methods
1
7
of O solid-state NMR spectra of crystalline sodium
phosphate compounds. We will then discuss the influence
of short- and long-distance ordering in phosphates on the
Calculations were performed using the PARATEC code
3
4,35
36
at DFT level of theory.
We used the PBE functional
for the generalized gradient approximation (GGA) of the
exchange-correlation functional. The potentials due to the
ions are represented by norm-conserving Troullier-Martins
quadrupolar parameters, C
Q
Q
and η , and the isotropic
chemical shift δcs. The results will finally be used for a better
understanding of the long-range disorder in a sodium
phosphate glass system.
3
7
pseudopotentials. The electronical configuration involved
in the construction of the pseudopotentials for the different
2
3
31
17
2
6
0
nuclei Na, P, and O are respectively { 2p 2p 3d }, {
3
2
. Experimental Section
s 2p 3d0.2 }, and { 2s 2p } with respective core radii
2
1.8
2
3
2
.1. Samples preparation. In a first step, two sodium phosphate
glasses (50Na O-50P and 62.5Na O-37.5P ) were prepared
by quenching a mixture of Graham’s salt ((NaPO , Prolabo) and
sodium carbonate (Na CO , Prolabo), melted at 900 °C for 1 h.
Then, the O enrichment process was performed by heating the
(
in atomic unit) {1.8 1.49 1.8}, {2.0 2.0 2.0}, and {1.45
2
2
O
5
2
2 5
O
1
.45}.
3 n
)
The electronic structure gives access to the electric field
gradient (EFG) tensor through the reconstruction of the all-
electron wave function that is obtained with the projector
2
3
1
7
1
7
glasses at 700 °C for 8 h in the presence of O-enriched water
2
5,26
3
augmented wave (PAW) approach.
is then diagonalized to obtain the principal components Vxx
yy, and Vzz where by convention |Vxx| < |Vyy| < |Vzz|. These
components are related to the quadrupolar coupling constant
and the asymmetry parameter η by C ) eQ|Vzz|/h and
) (Vxx - Vyy)/Vzz, respectively, where Q is the nuclear
quadrupole moment and h is the Planck constant. In the
The EFG tensor V
vapor in a closed tubular furnace. A statistical exchange between
17
the oxygen atoms of the material and the O atoms of the enriched
vapor is achieved through hydrolysis and condensation reactions,
summarized by the following equations:
,
V
C
Q
Q
Q
1
7
17
P - O - P + H O T P- O - H + P - O - H (1)
2
η
Q
17
17
-
31
2
P- O - H + P - O - H T P- O - P + H O (2)
present work, we used Q ) -25.58 × 10
m , which is
2
3
8
the value proposed by Pyykk o¨ .
Finally, pure crystalline sodium trimeta- and tripolyphosphates,
The calculation of the absolute chemical shielding was
Na
of the O enriched glasses at 550 °C for 8 h. Sodium pyrophos-
phate, Na , was prepared through a solid-state reaction between
diammonium hydrogen phosphate and sodium carbonate at 800 °C.
3 3 9 5 3
P O and Na P O10, were prepared through the devitrification
1
7
performed using GIPAW algorithms developed by Pickard
2
4
P O
4 2 7
and Mauri. Experimental isotropic chemical shifts δcs and
absolute isotropic chemical shielding σ are related through
the definition of an isotropic reference shielding σref defined
by δcs ) σref - σ. In the NMR community, it is commonly
accepted to use liquid reference compounds to compare
different experiments. In a computing approach, it would
not be conceivable to achieve accurate results using a liquid
reference compound. In most cases, an arbitrary choice of a
solid reference compound or an isolated resonance from the
unassigned spectrum is preferred. In the present work, we
set the absolute chemical shift value to an unambiguous
isolated resonance, as presented subsequently. The conver-
gence test in absolute chemical shielding on simple model
1
7
17
The sample was O-enriched under O water vapor at 700 °C for
5 h.
The crystallinity and purity of the three samples were checked
1
3
1
through X-ray powder diffraction and P MAS NMR, in which
the absence of broad signals indicates a complete crystallization.
1
7
2
.2. NMR details. All O MAS NMR experiments were
performed at 18.8 T on a Bruker Avance II spectrometer with a
Bruker 3.2 mm (rotor diameter) tri-γ probehead operating at 108.47
17
31
17
and 323.90 MHz for O and P, respectively. The O MAS spectra
were obtained using a selective pulse of 11 µs at a radio frequency
(
RF) field strength of 7.5 kHz. The excitation pulse was preceded
32
by a Double Frequency Sweep (DFS) pulse to enhance the central
transition polarization for half-integer quadrupolar nuclei. The signal
enhancement was achieved using a 3 ms DFS pulse sweeping
between 200 and 800 kHz at an RF field amplitude of 50 kHz. The
1
7
(sodium orthophosphate) shows that the convergence in O
NMR parameters is achieved for an energy cutoff of 100
Ry. For the integration on reciprocal space, we used a dense
3
QMAS spectra were collected at a MAS speed of 23 kHz, using
k-point grid (3×2×3 for Na
×4×2 for Na ).
The calculation of NMR parameters requires the accurate
knowledge of the structure. However, the published crystal-
lographic data for Na P O and Na P O were obtained with
3
P O
3 9
5 3
, 3×4×1 for Na P O10, and
8
the Z-filter sequence, which consists of two hard pulses of 3.5
and 1.25 µs at an RF field of 100 kHz, for triple-quantum excitation
and reconversion, respectively, followed by a soft pulse of 11 µs
2
4 2 7
P O
1
at an RF field of 7.5 kHz. The t step was set to the MAS period.
1
7
31
The O{ P} heteronuclear multiple quantum coherence (HMQC)
spectrum was acquired using the pulse sequence introduced by
3
3
9
5
3
10
3
9,40
a poor precision using the Weissenberg film technique.
For all model systems, we here compared in a systematic
3
3
17
π
Massiot et al. The O selective /
2
and π pulses were set to 11
3
1
and 22 µs, respectively and the P π pulse to 10.5 µs. The total
evolution time under O- P J coupling was set to 6 ms. All other
17
31
(
34) Pfrommer, B.; Raczkowski, D.; Canning, A.; Louie, S. G.; Mauri, F.;
Cote, M.; Yoon, Y.; Pickard, C. J.; Heynes, P. PARATEC (PARAllel
Total Energy Code).
relevant experimental parameters are given in the figure captions.
1
7
31
The O and P chemical shifts are given in ppm with respect to
tap water and 85% of H PO in solution, respectively.
(
(
35) Kohn, W.; Sham, L. Phys. ReV. 1965, 140, 1133.
36) Perdew, J. P.; Burke, K.; Ernzerhof, M. Phys. ReV. Lett. 1996, 77,
3
4
3
865.
(
(
32) Iuga, D.; Schafer, H.; Verhagen, R.; Kentgens, A. J. Magn. Reson.
(37) Trouillier, N.; Martins, J. L. Phys. ReV. B 1991, 43, 1993.
(38) Pyykk o¨ , P. Mol. Phys. 2001, 99, 1617.
2
000, 147, 192.
33) Massiot, D.; Fayon, F.; Alonso, B.; Trebosc, J.; Amoureux, J.-P. J.
Magn. Reson. 2003, 164, 160.
(39) Ondik, H. Acta Crystallogr. 1967, 18, 226.
(40) Corbridge, D. Acta Crystallogr. 1967, 13, 263.
Inorganic Chemistry, Vol. 47, No. 16, 2008 7329

Vasconcelos et al.
line in part d of Figure 1. These sites have a multiplicity of
1
. The other sites, O4, O5, O6, which are located on either
side of the symmetry plane, show a multiplicity of 2. Two
types of oxygen sites can be distinguished: bridging oxygens
(O3 and O4) that denote oxygen atoms connected to two
phosphorus, and nonbridging oxygens (O1, O2, O5 and O6)
that are connected to one phosphorus, only. Following the
3 3 9
conventional notation, all phosphate entities in Na P O are
2
noted Q because they are connected to two other phosphate
groups.
The structure of the high-temperature form of sodium
5 3
tripolyphosphate Na P O10 (phase I), was determined by
4
0
Corbridge in 1960 using single-crystal XRD. The com-
pound crystallizes in a monoclinic C2/c space group with
3
1
four formula units per unit cell (part b of Figure 1). The
P
4
7
31
MAS NMR reveals the presence of two P sites at 1.3
Figure 1. Unit cells of (a) Na3P3O9, (b) Na5P3O10, (c) Na4P2O7. The basic
anionic unit is shown in (d), (e), and (f) for Na3P3O9, Na5P3O10, and
Na4P2O7, respectively. Cell parameters and the notation employed for
nonequivalent oxygen sites are those given in the original papers.
For Na3P3O9, the mirror plane is indicated by a dashed line.
and -7.5 ppm with relative intensities 2:1 in agreement with
1
2
23
the presence of Q and Q phosphate groups. The Na
3
9,40,48
47
23
MQMAS NMR reveals the presence of three Na sites,
which is in agreement with the crystal structure. The anionic
unit is formed of a chain of three phosphate groups (part e
way the NMR parameters calculated directly from the
original XRD structure with the ones obtained after DFT
1
of Figure 1). Two of these phosphate groups are noted Q
4
1–43
as they are end-chain phosphates, and the third phosphate
structure optimization. We used the VASP code
for the
2
44
group is a central phosphate, Q , connected to two other
structure optimization, with GGA-PW91 functional and the
PAW scheme to describe the electrons-ions interaction
with Ecut ) 600 eV and same k-points grid as for the NMR
4
5
phosphate units. Five crystallographically distinct oxygen
sites can be distinguished. One site, denoted O2, is a bridging
oxygen and the others are nonbridging oxygen sites, one of
(
calculations). We took into account all the degrees of
freedom (ionic internal coordinates, cell shape, and cell
volume) for structure optimization, and we rescaled cell
parameters to the experimental volume to avoid the DFT
tendency to overestimate both the distances and the cell
volume. This overestimation only affects the quadrupolar
2
them being bonded to the Q phosphorus (O1), the three
1
others bonded to a Q phosphorus (O3, O4, O5).
The structure of the sodium pyrophosphate (or tetrasodium
4
8
diphosphate) Na
972 using single-crystal XRD. The structure is described
in an orthorhombic P2 space group with four formula
units per unit cell (part c of Figure 1). The P MAS NMR
4 2 7
P O was determined by Leung et al. in
1
1 1 1
2 2
constant C
parameter upon the absolute Vzz value. On the other hand,
the chemical shift and the asymmetry parameter η are only
weakly affected because they are relative values by definition.
Q
because of the direct dependence of this
3
1
2
3
spectrum and Na MQMAS NMR spectra respectively
reveals the presence of two and four crystallographically
nonequivalent sites as expected from the structure. The two
Q
3
1
P sites are separated by 0.5 ppm. The anionic unit is formed
4
. Structural Data
1
of two Q phosphate groups (P1 and P2), connected through
a bridging oxygen denoted O1. The other six oxygen sites
are nonbridging oxygens (O2 to O7).
3 3 9
The structure of sodium trimetaphosphate Na P O , was
3
9
solved by Ondik in 1965 using single-crystal XRD (part a
of Figure 1). It crystallizes in the orthorhombic Pmcn space
group with four formula units per unit cell. The basic unit,
5. Results
Na
3 3 9
P O , is composed of two crystallographically distinct
5
.1. Trimetaphosphate Na
3
P
3
O
9
. Figure 2 brings together
3 3 9
sodium sites, two phosphorus sites, and six oxygen sites. It
1
7
all O NMR measurements of Na P O . The top spectrum
2
3
has been studied previously through Na MAS NMR by
1
7
4
6
31
in part a of Figure 2 shows the 1D O MAS spectrum. The
line shape is characteristic of half-integer quadrupolar nuclei
MAS spectra broadened by second-order quadrupolar inter-
action. The sharp discontinuities reveal the high degree of
crystallinity of the sample. To increase the resolution, a 2D
Koller et al. The P MAS spectrum exhibits two reso-
nances at -16.2 and -19.2 ppm, in a 1:2 ratio, in good
agreement with crystallographic data. The basic anionic unit,
as depicted in part d of Figure 1, is composed of a ring of
three phosphate groups. Three oxygen sites, O1, O2, O3,
are located on a symmetry plane represented by a dashed
3
QMAS spectrum (part b of Figure 2) was acquired. In the
MAS dimension of part b of Figure 2, nonbridging oxygens
are in the 50 to 85 ppm chemical shift range, whereas the
large resonance on the left, which spans from 60 to 120 ppm,
(
(
(
(
(
(
41) Kresse, G.; Hafner, J. Phys. ReV. B 1993, 48, 13115.
42) Kresse, G.; Furthmuller, J. Phys. ReV. B 1996, 54, 11169.
43) See http://cms.mpi.univie.ac.at/vasp/.
44) Perdew, J. P.; Burke, K.; Wang, Y. Phys. ReV. B 1996, 54, 16533.
45) Kresse, G.; Joubert, D. Phys. ReV. B 1999, 59, 1758.
46) Koller, H.; Engelhardt, G.; Kentgens, A. P. M.; Sauer, J. J. Phys. Chem.
(47) Fyfe, C.; zu Altenschildesche, H. M.; Skibsted, J. Inorg. Chem. 1999,
38, 84.
1
994, 98, 1544.
(48) Leung, K. Y.; Calvo, C. Can. J. Chem. 1972, 50, 2519–2526.
7330 Inorganic Chemistry, Vol. 47, No. 16, 2008

17O Solid-State NMR and First-Principles Calculations
Figure 2. 17O NMR of Na3P3O9 (18.8 T). (a) The 17O MAS spectrum was obtained after the accumulation of 832 scans with a recycling delay of 60 s. (b)
1
7
A total of 96 scans was accumulated with a recycling delay of 20 s and 125 transients in the t1 domain for an experimental time of ∼66 h. (c) O isotropic
spectrum is shown together with the extracted individual slices and corresponding fit simulations.
Table 1. NMR Parameters Extracted by Fitting Each Individual
Table 2. Relative Intensities Extracted from the Fit of the 1D Spectra
a
Resonance of the MQMAS Spectrum of Na3P3O9 and Na5P3O10
compound
site
relative intensity (%)
compound
Na3P3O9
CQ (MHz)
ηQ
δcs (ppm)
Na3P3O9
O6
O1
O2, O5
O3
O4
O1
22.5
11.6
33.5
11.3
21.1
20.9
59.5
19.6
4.5
0.3
0.4
0.1
0.6
0.2
0.2
∼0.6
75.5
77.5
84.0
120.2
79.5
85.3
∼126
4
4
.3
.3
∼
7.0
4.5
.5
6.5
Na5P3O10
Na P O
10
5
3
4
O3, O4, O5
O2
∼
a
The sites are organized from low to high chemical shifts. The accuracy
Table 3. DFT-GIPAW Calculated 17O Quadrupolar Coupling Constant
CQ), Quadrupolar Asymmetry (ηQ) and Isotropic Chemical Shift (δcs)
Parameters Using DFT-Optimized Structure of Na3P3O9 and Na5P3O10
of the measurement varies significantly depending on the intensity of the
individual resonance. Nevertheless, the uncertainty does not exceed the
following values: CQ ( 50 kHz; δcs ( 0.5 ppm; ηQ ( 0.05).
(
a
compound
Na3P3O9
site
CQ (MHz)
ηQ
δcs (ppm)
typically corresponds to bridging oxygens. The isotropic
projection (δiso scale) of the 3QMAS spectrum reveals the
presence of two groups of nonbridging oxygens, around δiso
O6
O1
O5
O2
O3
O4
O1
O3
O4
O5
O2
O1
O2
O3
O4
O5
O6
O7
4.78
4.65
4.62
4.74
7.72
7.51
4.87
4.91
4.89
4.85
7.36
7.31
4.84
4.80
4.85
4.93
5.03
5.03
0.26
0.31
0.09
0.07
0.61
0.66
0.15
0.12
0.19
0.25
0.66
0.62
0.10
0.17
0.12
0.16
0.21
0.18
75.5
78.2
84.3
84.4
124.8
126.9
79.9
89.3
86.0
)
82 and 90 ppm, and a unique resonance at δiso ) 138
ppm corresponding to bridging oxygens. In this first ap-
proach, the empirical attribution was made based on the work
by Zeyer et al. who have shown that, in phosphate glass
Na5P3O10
Na4P2O7
2
systems, bridging oxygens can be separated from nonbridging
oxygens through their quadrupolar constant C and their
Q
chemical shift δcs. The individual anisotropic cross section,
together with their best-fit simulations, are shown on part c
of Figure 2. Table 1 summarizes the NMR parameters
resulting from the deconvolution of each resonance of the
88.8
136.7
132.67
81.24
80.94
75.83
78.38
74.37
75.25
3
QMAS spectrum. One can notice that the shoulder on the
isotropic resonance at δiso ) 85 ppm has been considered as
an additional site, even though the low signal intensity
renders difficult the determination of the NMR parameters
a
The sites are organized from low to high chemical shifts.
NMR (4) and the ones expected from the XRD structure
(6).
17O NMR parameters resulting from our DFT-GIPAW
calculation are given in Table 3. All calculated chemical
shifts are shifted by a constant value to match the calculated
value for O6 with the lowest experimental resonance
(
Q
especially η , which is directly related to the shape of the
resonance). The present results finally reveal the presence
of a total of four oxygen sites that are nonequivalent from a
pure NMR point of view. There is a clear difference between
1
7
the number of sites experimentally observed through
O
Inorganic Chemistry, Vol. 47, No. 16, 2008 7331

Vasconcelos et al.
observed on the 3QMAS spectrum (δcs ) 75.5 ppm). All
values are consistent with those extracted from the 3QMAS
spectrum (Table 1). The calculated and experimental qua-
drupolar asymmetry parameters η are close, or at least
Q
within an acceptable accuracy if we consider the low signal
intensity of sites O1 and O3/O4. The calculated quadrupolar
coupling constants C
assignment of the bridging (∼7 MHz) and nonbridging sites
∼4.5 MHz), even if they are found slightly larger by
.3-0.5 MHz than expected for such site environments. It
must be noticed that such deviations (<0.2 MHz) were
Q
are also consistent with the empirical
(
0
2
6,28,31
previously observed for silicates.
This result can be
partly due to the fact that the DFT calculated quadrupolar
constants, which are calculated at 0 K, are obtained without
taking into account the thermal motions that occur during
4
9,50
the NMR experiment.
The calculated chemical shifts must be considered care-
fully. From the 3QMAS spectrum, the presence of an
additional intermediate nonbridging site with δcs ) 77.5 ppm
is suggested. It is clearly confirmed by first-principles
calculations as the site denoted O1. The DFT calculation
further brings to light the similarity of sites O2 and O5 in
terms of the NMR parameters. These two sites correspond
to the single resonance observed at δcs ) 84.0 ppm on the
3
QMAS spectrum. The two bridging oxygen sites, O3 and
O4, are found to be separated by ∼2 ppm even though a
single oxygen site is detected at first sight on the MQMAS
spectrum. Furthermore, it appears that calculated bridging
oxygen chemical shifts are overestimated by ∼5-7 ppm with
respect to the experimental values. This deviation for
bridging oxygen is only due to the fact that here, a
nonbridging oxygen site, O6, is chosen as chemical shift
reference.
Figure 3. 17O MAS NMR 1D spectrum of Na P O is shown together
3
3
9
with the individual lineshapes simulated using, as a starting point, NMR
parameters from the DFT-GIPAW calculations for the optimized structure.
1
7
(
Top left inset) Expansion of the O spectrum in the chemical shift region
of bridging oxygens, with the calculated lineshapes for O3 and O4 sites
plain thin lines) and the resulting sum of the two contributions (dashed
(
line).
Finally, Figure 3 shows the fit of the experimental
spectrum using the calculated values as a starting point. We
observed that a scaling factor needs to be applied to the
computed chemical shift to match the experimental ones. The
correlation presented in the discussion section (Figure 9)
nonbridging oxygens can be attributed to the direct observa-
tion of a J coupling between O and P nuclei. This
attribution can be confirmed experimentally by identifying
17
31
1
2 31
the small resonances on either sides of the Q and Q
P
17
31
resonances as a contribution from the O- P J coupling
cf. 1D MAS spectrum on the right of Figure 6). The
simulation of the O MAS spectrum (Figure 4) leads to a
measured splitting between 120 and 130 Hz that is in the
range of published values of nonbridging oxygen to phos-
Q
leads to the value of 0.88. All calculated C were reduced
(
by 0.3 MHz for the reasons given above. The relative ratio
of the different oxygen sites are those expected from the
structure (Table 2). The resulting calculated and experimental
MAS spectra exhibit identical most-intense discontinuities.
The accuracy of the DFT calculation for site O1 also leads
to a good overall picture of the small discontinuities.
Furthermore, it was possible to separate O3 and O4 sites in
the experimental spectrum at the expense of a long experi-
mental acquisition of 12 h. This is exhibited in the inset of
Figure 3, which shows a shoulder around 118 ppm with the
correct relative intensity for O3 and O4 sites.
17
51
52
phorus J coupling in solution or in organic solids.
The 2D 3QMAS spectrum (part b of Figure 5) clearly
shows three resonances: one in the chemical shift range of
bridging oxygens and two in the chemical shift range of
nonbridging oxygens. Once again, we observe a discrepancy
between the NMR results and the structural data as there
are five crystallographically distinct oxygens. The assignment
of the unique bridging oxygen is obvious, taking into account
empirical considerations that have been mentioned above.
5
.2. Tripolyphosphate Na
5
P
3
O
10. The 17O 1D MAS
10 reveals the
spectrum shown in Figure 4 for Na
5
P
3
O
17
The assignment of nonbridging O sites can be achieved
presence of a splitting of less than 1 ppm in the nonbridging
oxygen region, proving the excellent crystallinity of our
sample. This splitting, which leads to an observed spreading
of the 2D 3QMAS lineshapes (Figure 5) in the region of
using the information provided by the correlation spectrum
17
31
of Figure 6. It shows the O{ P}MAS-J-HMQC spectrum
(51) Sammons, R. D.; Frey, P. A.; Bruzik, K.; Tsai, M.-D. J. Am. Chem.
Soc. 1983, 105, 5455.
(
(
49) Kristensen, J. H.; Farnan, I. J. Chem. Phys. 2001, 114, 9608.
(52) Bryce, D. L.; Eichele, K.; Wasylishen, R. E. Inorg. Chem. 2003, 42,
50) Schurko, R. W.; Wi, S.; Frydman, L. J. Phys. Chem. 2002, A106, 51.
5085.
7332 Inorganic Chemistry, Vol. 47, No. 16, 2008

17O Solid-State NMR and First-Principles Calculations
throughout the entire chemical-shift range of both bridging
and nonbridging oxygen sites. Except for tripolyphosphate
sites O3, O4, and O5, the correlation is excellent with a
regression coefficient close to 1. We propose that the
apparent discrepancy between theoretical and experimental
parameters is the result of a dynamical effect through the
free rotation of the end-chain phosphate group around the
P-O2 axis. An experimental confirmation of our interpreta-
tion would be obtained by recording high-resolved 1D
spectrum at low temperature even if that would bring
additional experimental difficulties to reach the same level
of resolution. Such an experiment is beyond the scope of
the present article.
5
.3. Pyrophosphate Na
4
P
2
O
7
. Figure 8 shows the 17O
. The MAS spectrum
a) reveals the presence of bridging oxygen O1 in the
MAS and 3QMAS spectra of Na
4 2 7
P O
(
8
0-120 ppm range. The most intense signal on the right
corresponds to the crystallographically distinct nonbridging
oxygen sites. There is no evidence of second-order quadru-
polar lineshapes, which could help distinguish the six
nonbridging oxygen sites. The region of nonbridging oxygen
sites is exhibited on the 3QMAS spectrum (part b of Figure
8
) (bridging oxygens are barely detected due to unfavorable
O enrichment). Unfortunately, the six nonbridging oxygen
1
7
Figure 4. 17O MAS NMR 1D spectrum of Na5P3O10 (18.8 T) together
with the individual lineshapes. The inset shows the measurement of O- P
J coupling constant on the O3, O4, and O5 sites simulation.
sites are not clearly separated on the 2D spectrum. Neverthe-
less, the apparent broad line shape cannot be attributed to a
possible distribution of chemical shift because the crystal-
linity of the sample was checked through XRD and con-
firmed by the quality of the P and Na MAS spectra (not
shown). In fact, the signal of each component adds up to
form a large line shape that is spread over a chemical shift
range of 10 ppm in the isotropic dimension, reinforcing the
idea that the end-chain oxygens are crystallographically
distinct but very close in terms of the NMR parameters.
17
31
17
31
31
that correlates O sites with chemically bonded P. The P
spectrum of Na 10 (right side of the 2D spectrum) is
composed mainly of two resonances at 1.3 and -7.5 ppm,
3
1
23
5 3
P O
1
2
which correspond to Q and Q phosphate groups, respec-
tively. In Figure 6, only the chemical shift range of
nonbridging oxygen is presented. Indeed, it was not possible
to observe a correlation signal between bridging oxygens
and phosphorus sites because of (i) the low intensity of this
oxygen resonance and (ii) the possibly smaller J-coupling
value for bridging oxygens compared to nonbridging oxy-
gens. We observe two correlation signals for nonbridging
oxygen sites. The top signal shows a correlation between
4 2 7
The calculated NMR parameters for Na P O are given
in Table 3. Once again, bridging and nonbridging oxygen
sites are clearly distinguished by their chemical shift and
quadrupolar coupling constant. Our observation for the
Q
asymmetry parameter η is somewhat different from the
5
3
1
7
2
one nonbridging O site and the Q phosphate group only.
This signal can be assigned to oxygen site O1. The other
correlation signal shows nonbridging oxygen sites, which
values published by Prochnow et al. Indeed, they report a
value of 0.6, close to what we measure for bridging oxygens
but not for nonbridging oxygens. The values that we propose
are more in line with those experimentally measured on the
1
are chemically bonded to Q phosphorus sites only. They
are assigned to O3, O4, and O5, respectively. This attribution
is in agreement with the relative intensities obtained with
the deconvolution of the 1D spectrum.
other sodium phosphates of the present work (Na
3 3 9
P O ,
Na 10). A close look at the chemical shift values
5 3
P O
calculated for the six nonbridging oxygens confirms that the
width of the nonbridging oxygen resonance, observed on the
3QMAS spectrum in Figure 8, is not due to distribution but
to a juxtaposition of six resonances, spreading from 74.4 to
81.2 ppm (∼7 ppm). The lack of resolution along the
isotropic dimension of the 3QMAS spectrum renders difficult
the comparison of experimental and calculated data for
Na P O . In particular, the quadrupolar parameters cannot
4 2 7
be extracted from the individual slices. Nevertheless, the
accuracy of the calculated data in the two previous cases,
Table 3 summarizes the NMR parameters calculated with
the GIPAW method for Na
5 3
P O10 using the same chemical
3 3 9
shift reference than that used for Na P O . These results
confirm the assignment deduced from the NMR experiments,
especially for sites O1 and O2. The GIPAW results of end-
chain sites O3, O4, and O5 require special attention. Indeed,
one can clearly see that the computed NMR parameters for
these three sites are slightly different and even in disagree-
ment with previous experimental assignment. In part a of
Figure 7, we plotted the experimental and calculated chemical
3 3 9 5 3
Na P O and Na P O10, increases the confidence in the results
shift values for nonbridging oxygens of Na
3 3 9
P O and
(
53) Prochnow, D.; Grimmer, A.-R.; Freude, D. Solid State Nucl. Magn.
Reson. 2006, 30, 69.
Na 10. Part b of Figure 7 shows the same correlation
5 3
P O
Inorganic Chemistry, Vol. 47, No. 16, 2008 7333

Vasconcelos et al.
Figure 5. 17O NMR of Na5P3O10 (18.8 T). (a) The 17O MAS spectrum required 64 scans with a recycling time of 60 s; (b) 3QMAS spectrum, a total of 12
1
7
scans was accumulated with a recycling delay of 20 s and 140 transients in the t1 domain for an experimental time of ∼10 h; (c) O isotropic spectrum
together with the extracted individual slices and corresponding fit simulations.
Figure 6. 17O{ P} MAS-J-HMQC of Na5P3O10. 17O (top) and 31P (right)
31
1
2
5
D MAS spectra are also presented for comparison with 2D spectrum. The
D spectrum required a total of 16 scans with a recycling delay of 60 s and
0 t1 increments (experimental time of 13 h). Note that the signal truncation
Figure 7. Correlation of experimental and calculated chemical shifts (δiso)
for (a) nonbridging oxygens and (b) all sites. For the left solid triangle, the
tripolyphosphate (Na5P3O10) and (2) the trimetaphosphate (Na3P3O9), using
the DFT optimized structures. The exact correlation between experiments
and calculations is represented with a dashed line. The slope of linear
regression is given with the correspondely regression coefficients (R).
3
1
in the t1 dimension leads to a strong broadening of the P line shape, to be
compared with the high resolution of the 1D MAS spectrum (right). The
3
1
stars in the P MAS spectrum denote the multiplet resulting from the
1
7
31
O{ P} J coupling.
obtained for Na
4 2
P
O
7
. It is observed that C
Q
Q
and η are also
distributed over about 0.2 MHz and 0.1, respectively.
Figure 9 integrates all the possible correlations between
experimental and calculated values for the different NMR
6
. Discussion
We have seen in the previous sections that there is
sometimes a discrepancy between experimental NMR results
and structural data for the different model systems (particu-
parameters (δcs, C
and Na 10 used in this work, that is, published structures
and DFT optimized structures. The data obtained for Na
are excluded as they lack experimental resolution. The end-
chain oxygen sites in Na 10 are omitted due to the
dynamical effect, which seems to affect these sites. For
Q
, η
Q 3 3 9
) and for the two structures of Na P O
5 3
P O
4 2 7
P O
3 3 9 5 3
larly Na P O and Na P O10). Our work demonstrates that
1
7
the DFT results provide the way to link the O local
environment to the NMR response.
5 3
P O
7334 Inorganic Chemistry, Vol. 47, No. 16, 2008

17O Solid-State NMR and First-Principles Calculations
Figure 8. 17O NMR of Na4P2O7 (18.8 T). (a) The 17O MAS spectrum
required 4096 scans with a recycling time of 5 s. (b) 2D 3QMAS spectrum.
A total of 432 scans was accumulated with a recycling delay of 5 s and 64
increments in the t1 domain for a total experimental time of 38 h. (c) The
1
7
isotropic O 3QMAS spectrum.
Q
clarity, we represented the chemical shift and the C values
to the limited range of nonbridging oxygens, even though
the linear regression of the correlation is provided for all
sites.
Figure 9. Comparaison of experimental and calculated NMR parameters
δiso, CQ, ηQ) for (a, b, c) the published structures and (d, e, f) the DFT
(
optimized structure of trimetaphosphate (Na3P3O9) and tripolyphosphate
(for Na5P3O10, sites O1, O2, only). For δiso and CQ parameters, the graphs
are restricted to the range of the nonbridging oxygens. The exact correlation
between experimental and calculated values is represented by a dashed line.
The slope of the linear regression is given with corresponding regression
coefficients (R).
For the isotropic chemical shift, the slope of the correlation
is smaller than one for the optimized structures (right) and
close to one for the published structures (left). However, one
can clearly observe that the optimized structure reveals a
better correlation between experimental and theoretical
results, which is fundamental for the assignment. The
regression coefficient is therefore an important and accurate
parameter that helps to select the best structure for an optimal
fit of the experimental NMR results. The regression coef-
ficients are also given in Figure 9 for the quadrupolar
parameters (b, e) C
Q
Q
and (c, f) η . For both parameters, the
DFT-optimized structures provide systematically a better
correlation than the original published structures.
The present results clearly demonstrate that the long-
distance order governs the oxygen chemical shift in phos-
phates. Indeed, only a periodic approach reproduced the
correct spectra. A cluster-based calculation, considering either
the anion or the first coordination sphere within the same
systems (not shown), gives incoherent results compared to
the experimental ones. A close look at the chemical shift
3 3 9
values for nonbridging oxygens in Na P O can further
demonstrate which structural feature influences the chemical
shift. As explained in the results section, the oxygen sites
O2 and O5 are completely equivalent in terms of NMR
parameters, whereas they have different first coordination
spheres (Figure 10). Focusing on the first coordination sphere
only, it even appears that site O2 looks similar to O6, and
O5 looks similar to O1. Our experimental and calculated
results of the chemical shift for nonbridging oxygens
Figure 10. First sphere of coordination of different nonbridging oxygens
in trimetaphosphate (Na3P3O9) with the corresponding distances between
sites obtained after the optimized DFT structure. (a) O1, trigonal site; (b)
O6, tetrahedral site; (c) O2, tetrahedral site; and (d) O5, trigonal-planar
site.
therefore clearly show that the long-range structural orga-
1
7
nization prevails. Following this, we conclude that the
O
Inorganic Chemistry, Vol. 47, No. 16, 2008 7335

Vasconcelos et al.
Figure 11. 17O MQMAS NMR of sodium metaphosphate glass (18.8 T). (a) 17O 2D 3QMAS spectrum, a total of 876 scans was accumulated with a
recycling delay of 0.8 s and 40 t1 increments for an experimental time of about 8 h. (b) Five slices are extracted from the 2D spectrum and shown with their
respective best fit simulation.
chemical shift in phosphates cannot be used to differentiate
local structure and geometry within a type of coordination.
This may be due to a large sensitivity of the local electronic
structure of nonbridging oxygen with respect to the large-
range electrostatic potential.
Table 4. NMR Parameters Extracted by Fitting Five Individual
a
Anisotropic Slices of 3Q Filtered Spectrum for NaPO3 Glass
CQ (MHz)
ηQ
δcs (ppm)
4.5
0.25
0.24
0.24
0.24
0.23
78.27
80.07
82.04
85.37
87.29
4
4
4
.4
.4
.4
We also demonstrate that in contrast with δcs, the C
parameters in sodium phosphates are clearly determined
by the nature of the local environment. Indeed, we verified
that for bridging oxygens in Na , it is possible to match
the C and η values with the empirical angular correlation
that was constructed from a cluster-based calculation per-
Q
and
η
Q
4.3
a
The sites are organized from low to high chemical shifts.
3 3 9
P O
a broad range of about 0.3 (from 0.1 to 0.4). In this case, it
seems possible to further correlate a specific nonbridging
Q
Q
2
0
formed on a simple H
The quadrupolar parameters C
interpreted in terms of local coordination and geometry.
Indeed, within the group of nonbridging oxygens in Na
4 2 7
P O .
oxygen environment with the corresponding η
Nevertheless, such correlation would require (i) a better
experimental determination of the η parameter and (ii) a
Q
parameter.
Q
Q
and η can further be
Q
3 3 9
P O ,
more complete database of sodium phosphate reference
one can observe two domains that separate sites O2 (4.74
MHz) and O6 (4.78 MHz) on the one hand and sites O1
compounds.
These conclusions enable the interpretation of 17O NMR
(
4.65 MHz) and O5 (4.62 MHz) on the other hand. These
groups can be chemically distinguished by their first coor-
dination sphere as shown in Figure 10. The C parameter
can therefore be interpreted as a semilocal probe, sensitive
to the oxygen coordination. In the case of the η parameter,
spectrum of a sodium metaphosphate glass (NaPO
NMR. Part a of Figure 11 represents the 17O 3QMAS
spectrum of NaPO glass restricted to the chemical shift range
of nonbridging oxygens. Part b of Figure 11 represents five
arbitrary anisotropic slices and their respective simulation.
Here, we clearly observe a distribution in chemical shift
without line shape modification. This is confirmed by the
result of the best fit simulations summarized in Table 4. It
can be observed that only the δcs parameter is subjected to
a large variation within this compound and the other
3
) by
Q
3
Q
the nature of the first coordination sphere does not seem to
matter. Indeed, nonbridging oxygen sites are grouped into
two groups {O2,O5} and {O1,O6}, which are related to their
axial or equatorial positions on a cyclohexane-like molecule
rather than the nature and number of sodium sites. The
1
7
conclusion is thus that the O η
Q
parameter in phosphates
parameters (C
Q Q
, η ) vary by only 0.1 and 0.02 MHz,
is a local probe, which is sensitive to the geometry of the
1
7
respectively. Note that the typical second-order quadrupolar
feature is still present with sharp discontinuities, close to
those that are observed on MAS spectra of crystalline
phosphates. Therefore, the glass structure is clearly amor-
phous in a long range, as deduced from the large variation
of δcs, the only parameter sensitive to a long-range disorder.
On the other hand, the local environment in the glass does
not seem to change much, as shown by the quadrupolar
parameters - used as a local probe - that are quasi-constant.
The structure description provided here through the inter-
P- O bond in the phosphate group.
It is important to note that the structure optimization is
much more critical for C
Q
Q Q
than for η because C values
for nonbridging oxygens are within a small range of 0.2
MHz. This result points out the fact that nonbridging oxygen
sites in sodium phosphates can not be characterized by their
experimental C
Q
values solely. On the other hand, the η
Q
values can because they spread over a broad range. The
bridging oxygen values clearly appear around 0.6. For
nonbridging oxygens, the η
Q
values are found to be within
7336 Inorganic Chemistry, Vol. 47, No. 16, 2008

17O Solid-State NMR and First-Principles Calculations
pretation of the NMR parameters is consistent with common
description of amorphous systems.
the absence of variation of C
strates the persistence of an order within the oxygen first-
Q
Q
and η parameters demon-
coordination sphere.
7
. Conclusion
The 17O MAS NMR spectra of a series of three sodium
Acknowledgment. The USTL Centre de Ressources
Informatiques and IDRIS are acknowledged for the allocating
of CPU time. F.V. is supported by the Ministère de
phosphate compounds was interpreted through the combina-
tion of high-resolution NMR experiments and periodical DFT
calculations. The correlation of experimental and theoretical
parameters depends on the quality of the initial structure.
The DFT-optimized structures always leads to a better match.
Two of the sodium phosphate systems presented here were
used provide a set of data that help in the interpretation of
´
l′Education Nationale de l’Enseignement Sup e´ rieure et de
la Recherche. The FEDER, R e´ gion Nord Pas-de-Calais,
´
Ministre de l′Education Nationale de l’Enseignement Su-
p e´ rieure et de la Recherche, CNRS, and USTL are acknowl-
edged for financial support.
Q Q
the NMR parameters in terms of local, η , semilocal, C ,
and long-range, δcs, geometry, and coordination. These
observations were finally applied to the structural description
Supporting Information Available: This material is available
free of charge via the Internet at http://pubs.acs.org.
3
of a binary glass, NaPO . The strong distribution of chemical
shift reveals the long-range disorder in the glass, whereas
IC800637P
Inorganic Chemistry, Vol. 47, No. 16, 2008 7337