The atomic structure of niobium and tantalum containing borophosphate glasses
This article has been downloaded from IOPscience. Please scroll down to see the full text article.2009 J. Phys.: Condens. Matter 21 375106(http://iopscience.iop.org/09538984/21/37/375106)Download details:IP Address: 129.12.11.80The article was downloaded on 21/01/2011 at 09:16Please note that terms and conditions apply.View the table of contents for this issue, or go to the journal homepage for more
HomeSearchCollectionsJournalsAboutContact usMy IOPscience
IOP P
UBLISHING
J
OURNAL OF
P
HYSICS:
C
ONDENSED
M
ATTER
J. Phys.: Condens. Matter
21
(2009) 375106 (11pp) doi:10.1088/09538984/21/37/375106
The atomic structure of niobium andtantalum containing borophosphateglasses
K M Wetherall
1
, P Doughty
1
, G Mountjoy
1
, M Bettinelli
2
,A Speghini
3
, M F Casula
4
, F CesareMarincola
4
, E Locci
4
andR J Newport
1
1
School of Physical Sciences, University of Kent, Ingram Building, Canterbury CT2 7NH, UK
2
Laboratorio di Chimica dello Stato Solido, DB, Universit`a di Verona and INSTM,UdR Verona, Strada Le Grazie 15, 37134 Verona, Italy
3
DiSTeMeV, Universit`a di Verona and INSTM, UdR Verona, Via della Pieve 70,37029 Verona, Italy
4
Dipartimento di Scienze Chimiche, Universita di Cagliari, SS 554 Bivio per Sestu,09042 Monserrato (CA), Italy
Received 16 June 2009, in ﬁnal form 24 July 2009Published 17 August 2009Online at stacks.iop.org/JPhysCM/21/375106
Abstract
A complete structural study has been carried out on sodium borophosphate glass containingincreasing amounts of either niobium or tantalum. A combination of high energy xraydiffraction, neutron diffraction, extended xray absorption ﬁne structure, nuclear magneticresonance, and infrared and Raman spectroscopy has been used to discern the local atomicstructure of each component and the changes with M content, where M is either niobium ortantalum. The glasses are found to consist of tetrahedral borate and phosphate with octahedralMO
6
. As expected, B and P play the roles of tetrahedral network formers. At low M contentthere are isolated MO
6
units with P
···
M and B
···
M linkages that contribute to the glassnetwork. As the M content increases, the number of M
···
M links increases, and at the highestM content each MO
6
unit is connected to several others. The octahedra become signiﬁcantlydistorted as the niobium content increases, an effect that is not seen for tantalum.(Some ﬁgures in this article are in colour only in the electronic version)
1. Introduction
The origin of nonlinear optical (NLO) effects in glassesis a ﬁeld of current interest due to their optoelectronicapplication. For example, glasses with elevated NLO effectsare promising candidates for ultrafast optical switches [1].Large values of NLO susceptibility
χ
(
3
)
are observed inglasses containing empty d shell transition metal ions such asTi
4
+
and Nb
5
+
[2]. One such group of glasses are sodiumborophosphates containing niobium or tantalum, with thecomposition
x
M
2
O
5
(
1
−
x
)(
Na
2
B
4
O
7
)
0
.
05
(
1
−
x
)(
NaPO
3
)
0
.
95
,where M
5
+
is either Nb
5
+
or Ta
5
+
ions. NLO effects inthese glasses have recently been investigated, with the aimof explaining the phenomena in terms of the local atomicenvironmentofNb
5
+
andTa
5
+
ions[3–6]. However, there have
been very few studies of the structure, in particular of the Nbor Ta environments.Given the interest in the NLO properties of these glasses,there have been several previous attempts to characterize theatomic structure that mainly utilize vibrational spectroscopy.Previous spectroscopic studies of Nb
2
O
5
–NaPO
3
–Na
2
B
4
O
7
glasses have shown that as the M
5
+
concentration is increased,its role changes [7]. When M
2
O
5
is added in small quantitiesit forms isolated MO
6
octahedra. As the concentration isincreased there begins to be corner sharing of octahedra.Eventually, at the highest content, a threedimensional network of corner sharing MO
6
takes on the role of a network former, strengthening the glass network by eliminating nonbridging oxygen atoms [7]. Previous studies have indicatedthat this clustering of the MO
6
octahedra results in the
09538984/09/375106+11
$
30.00
©
2009 IOP Publishing Ltd Printed in the UK
1
J. Phys.: Condens. Matter
21
(2009) 375106 K M Wetherall
et al
Table 1.
Glass compositions, the label indicates the mol% M
2
O
5
. A reasonable estimate of error in these values is
±
2%.SampleM
2
O
5
(mol%)Na
2
B
4
O
7
(mol%)NaPO
3
(mol%)Fe
2
O
3
(mol%)Density(g cm
−
3
)Density
(
atoms ˚A
−
3
)
Ta 9.3 9.3 4.4 84.3 2.0Ta19.6 19.6 3.9 74.5 2.0 4.04 0.0799Ta35.0 35.0 3.2 59.8 2.0 4.70 0.0747Nb10.8 10.8 4.6 84.1 0.5Nb21.1 21.1 4.7 73.7 0.5 2.92 0.0720Nb36.5 36.5 3.2 59.8 0.5 3.48 0.0759
increase in the third order nonlinear susceptibility due to thehyperpolarizability of an extended structure of MO
6
units [7].In particular, the short M–O bond is important as it increasestheinﬂuence ofthedorbitalsonthenonlinearopticalresponseof the glass.The conclusions above have been drawn from spectroscopic techniques that observe the movement of the bondsrather than direct observation of the correlations in positionsbetween M
5
+
ions and neighbouring ions. There have alsobeen extended xray absorption ﬁne structure (EXAFS) [7]and xray diffraction (XRD) [8] studies reported on glasseswith the same composition as those studied here, and nuclearmagnetic resonance (NMR) [9] has been performed on relatedbut simpler glass compositions. Presented here if a full studyof the atomicscale structure that uses data taken some timeago. It includes the techniques of high energy xray diffraction(HEXRD), neutron diffraction (ND) and NMR which are newto the literature on these glasses, as well as EXAFS, Ramanand IR spectroscopy data not previously published.
2. Method
2.1. Glass preparation
Glasses were made with batched compositionslisted in table 1.They contain small amounts of iron for the purposes of theMossbauer study presented in [5]. As the Fe content is less
than 1
/
3rd of B content, Fe is neglected in the present study.The glasses were prepared by following the method describedin [5]. Analytical grade Nb
2
O
5
, Ta
2
O
5
, Na
2
B
4
O
7
·
H
2
O,NaH
2
PO
4
·
H
2
O and Fe
(
NO
3
)
3
for the niobium containing glassand Fe
2
O
3
for the tantalum containing glass were used. Theappropriate mixtures were melted in an electric furnace underoxygen in a platinum crucible at 1250
◦
C for 30 min. Themelts were quenched by pouring on to a cold copper plateat room temperature and annealed overnight under an oxygenatmosphere, for further details see [5]. Their densities were
measured using Archimedes’ method and are also listed intable 1.
2.2. Raman and IR spectroscopy
Raman spectra were measured using a Jobin Yvon modelHR640 Raman spectrometer, ﬁtted with a 25 mW, 623.8 nmhelium neon laser and a liquid nitrogen cooled CCDmultichannel detector. Repeated measurements with improvedresolution and less noise were taken for the samples containing9.3%Ta, 35.0%Taand36.5%NbusingaLabRAM300systemconﬁgured with a 50 mW 532 nm laser excitation. Spectrawere recorded by a TE air cooled CCD detector cooled to
−
70
◦
C. (Comparison between the two measurements showedthe peak positions and heights were consistent.)Infrared spectra were measured by a Biorad FTS175Cspectrometer controlled by WinIR software. Samples werediluted in dry KBr and measured for 64 scans in diffusereﬂectance mode over the range 4000–400 cm
−
1
with aresolution of 4 cm
−
1
. The spectrum of blank KBr was alsomeasured to allow background subtraction. As the spectrumfor the glass containing 21.1% Nb
2
O
5
was recorded at a laterdate, the background has a slight inﬂuence from adsorbedwater around 1500 cm
−
1
.
2.3. High energy xray diffraction measurements
The xray diffraction data was collected on Station 9.1 of the synchrotron radiation source (SRS), Daresbury Laboratory.The ﬁnely powdered samples were enclosed inside a 0.5 mmthick circular aluminium annulus with kapton windows. Theywere mounted in
θ/
2
θ
ﬂat plate transmission geometry withscanning of 2
θ
from 2
◦
to 126
◦
. The incident beam size was1 mm
×
10 mm. For the samples containing tantalum, thexray wavelength was set at
λ
=
0
.
4858 ˚A (calibrated usingthe Kedge of a Ag foil); this wavelength provides data to ahigh value of wavevector transfer (
Q
max
∼
23 ˚A
−
1
where
Q
=
4
π
sin
θ/λ
). For the samples containing niobium, thewavelength had to be changed to 0.6569 ˚A (again calibratedusing the Kedge of a Ag foil) to avoid ﬂuorescence. An ionchamber detector was used with a 1 mm slit at 30 cm to retainacceptable resolution.The ﬁrst stage of the data analysis is the application of corrections for the polarization of the xray beam, the variationin sample thickness with incident angle, and backgroundscattering. Corrections for the absorption, Compton scattering,the selfscattering and the sharpening function (where thesignal is divided by the average form factor per electronsquared)are thenmade. Theresultantscatteringintensity,
i
(
Q
)
reveals structural information
via
a Fourier transform to obtainthe total pair correlation function, as in (1).
T
X
(
r
)
=
2
π
2
ρ
N
r
+
Q
max
Q
min
M
(
Q
)
Qi
X
(
Q
)
sin
(
Qr
)
d
Q
(1)where
r
is the atomic separation between atom pairs,
ρ
N
is thebulk number density and
M
(
Q
)
is a window function appliedto reduce the Fourier transform termination artefacts that arise2
J. Phys.: Condens. Matter
21
(2009) 375106 K M Wetherall
et al
from the ﬁnite range of
Q
. A Hanning window function hasbeen used here.In order to obtain the structural information directly fromthe experimental data, each possible pairwise combination of elements
i
and
j
is represented by a pair correlation function
p
ij
(
r
)
that is then ﬁtted to the data. The
Q
space simulation of each pair correlation function
p
ij
(
Q
)
is generated using (2).
p
ij
(
Q
)
=
N
ij
ω
ij
sin
Qr
ij
exp
[−
0
.
5
Q
2
σ
2
ij
]
(
c
j
Qr
ij
)
−
1
.
(2)The sum of these
p
ij
(
Q
)
is equivalent to the
i
X
(
Q
)
in (1)and is Fourier transformed using (1) for comparison to the
experimental real space data
T
X
(
r
)
, having been subjected toexactly the same Fourier transform effects.The parameters in (2) are the averaged coordination
number of atom type
j
around an atom of type
i
,
N
ij
, thepairwise atomic separation,
r
ij
, and the disorder factor (ameasure of thermal and static disorder)
σ
ij
. These are variedin the ﬁtting procedure. The parameter
c
j
is the concentrationof atom type
j
and
ω
ij
is the weighting function that accountsfor the variation in scattering strength
f
i
(
Q
)
of different atomtypes, as deﬁned in (3).
ω
ij
=
(
2
−
δ
ij
)
c
i
c
j
f
i
(
Q
)
f
j
(
Q
)
[
f
(
Q
)
]
2
.
(3)
2.4. Neutron diffraction measurements
The neutron diffraction data was collected on the GEMdiffractometer on the ISIS spallation neutron source at theRutherford Appleton Laboratory, UK. Timeofﬂight data wascollectedovera widerange of
Q
(upto40 ˚A
−
1
). The powderedsamples were placed in an 8.3 mm vanadium can. The datawas analysed using the ATLAS suite of programs [10]. An
additional correction was made to the data from the samplecontaining 19.6% Ta
2
O
5
as an insufﬁcient amount of samplein the vanadium can meant the sample did not ﬁll the verticalheight of the beam (a polynomial was subtracted from the low
Q
part of the
Q
space data).The principles of neutron diffraction have much incommon with those for xrays, and equations of the same formare used in the analysis. A key difference srcinates from theneutron’s interacting with nuclei as scattering centres ratherthan electron density as for xrays. In neutron diffractionthe real space total correlation function
T
(
r
)
is given by (4),
where
b
i
isthescatteringlengththat describeshow stronglythetarget nuclei scatter neutrons. The value of
b
i
used for a givenelement is an average over the natural abundance of isotopesfor that element. Here a Lorch window function is used.
T
N
(
r
)
=
4
πρ
N
r
i
c
i
b
i
2
+
2
π
Q
max
Q
min
M
(
Q
)
Qi
N
(
Q
)
sin
(
Qr
)
d
Q
.
(4)Structural information is then obtained from the experimentaldata using the same method as for the xray diffraction data.Each type of atom pair is modelled by a pair correlationfunctionasdeﬁnedin(2). Forneutrondiffractiontheweighting
factors are deﬁned in (5).
ω
ij
=
(
2
−
δ
ij
)
c
i
c
j
b
i
b
j
.
(5)The accuracy of information obtained from ﬁtting paircorrelation functions depends on correctly assigning the atomtypes to
i
and
j
for all the peaks
p
ij
(
r
)
, including those thatoverlap. Accurate results can be obtained once all peaks areidentiﬁed and accounted for. The ﬁtting was then optimizedby using a nonlinear least squares ﬁtting algorithm to ﬁnd theparameters
R
ij
,
N
ij
and
σ
ij
which give best agreement withexperiment [11].
2.5. Xray absorption measurements
The EXAFS data was collected on Station 9.2 of theSynchrotron Radiation Source (SRS), Daresbury Laboratory.The spectra were recorded in transmission mode usinga doublecrystal Si(220) monochromator, and ionizationchambers to detect the incident and transmitted beamintensities,
I
0
and
I
t
, respectively. Data were recorded aroundthe Ta L
III
edge at 9881 eV before changing to the Nb Kedge at 18986 eV. Finely ground samples were diluted withpolyethylene and pressed into pellets to give a satisfactoryabsorption and edge absorption step. The monochromatorwas adjusted to give 50% harmonic rejection. The EXAFSspectra were collected over the range
k
=
3–18 ˚A
−
1
with astep of 0
.
045 ˚A
−
1
, where
k
is the photoelectron wavevector.Multiple spectra were recorded and summed for each sample.The programs EXCALIB, VIPER and EXCURV98 wereused to analyse the data [12]. The
k
3
weighted EXAFSsignal is ﬁtted by structural parameters to obtain values forcoordination number,
N
, interatomic distance,
R
, and theDebye–Waller factor,
A
=
2
σ
2
. The magnitudeof the EXAFS,
χ(
k
)
, is:
χ(
k
)
=
AFAC
j
N
j
kR
2
j

f
(π,
k
,
R
j
)

e
−
2
R
j
/λ(
k
)
e
−
2
σ
2
j
k
2
×
sin
(
2
kR
j
+
2
δ(
k
)
+
ψ(
k
,
R
j
))
(6)where
f
(π,
k
,
R
j
)
describes the photoelectron backscattering,
λ(
k
)
is the electron mean free path, and
δ(
k
)
and
ψ(
k
,
R
j
)
are the phaseshifts experienced by the photoelectron dueto the potentials of the emitting and backscattering atomsrespectively.
AFAC
is the proportion of electrons that arescattered elastically and is reﬁned by analysis of referencematerials with known structure, i.e. M
2
O
5
. Least squaresreﬁnements of the structural parameters were carried outagainst the
k
3
weighted EXAFS signal to minimize the ﬁtindex, FI, as in (7).
FI
=
i
(
k
3
(χ
T
i
−
χ
E
i
))
2
(7)where
χ
T
i
and
χ
E
i
are the modelled theoretical andexperimental EXAFS, respectively. The results of thereﬁnements are reported in terms of the discrepancy index,
R
di
.
R
di
=

(χ
T
(
k
)
−
χ
E
(
k
))

k
3
d
k

χ
E
(
k
)

k
3
d
k
×
100%
.
(8)3
J. Phys.: Condens. Matter
21
(2009) 375106 K M Wetherall
et al
Figure 1.
Raman spectra showing evolution of the peaks as afunction of Ta and Nb content.
2.6. NMR measurements
Solidstate NMR experiments were performed at 9.39 T usinga Varian 400 UNITY INOVA spectrometer operating at theLarmor frequency of 161.90 and 128.32 Hz for
31
P and
11
B,respectively, and equipped with a 4 mm probe head. Thesamples were packed into cylindrical Si
3
N
4
rotors. All spectrawere recorded under magic angle spinning (MAS) conditionsat a spinning speed of 12 kHz.The
31
P MAS experiments were run with a pulse length of 5
µ
s (90
◦
tip angle), a recycle time of 1 s, 120 kHz bandwidth,and 512 scans.
31
P chemical shifts were referenced to 85%H
3
PO
4
at 0 ppm. The
11
B MAS spectra were obtained using ashortpulse of 1.2
µ
s, a recycle time of 1 s, 100 kHz bandwidth,and 2048 scans. The pulse length was chosen to be 1
/
6 of the
π/
2 pulse length, calibrated using a 1 M H
3
BO
3
aqueoussolution.
11
B chemical shifts were referenced to aqueous boricacid which resonates at 19.6 ppm relative to boron triﬂuorideetherate.
3. Results
3.1. Raman and IR spectroscopy
The spectra recorded from all six samples are shown inﬁgures 1 and 2, the peak positions are listed in table 2. The
spectra exhibit the anticipated features as are highlighted inthe literature [2, 5, 7, 8]. At the lowest M
2
O
5
content, afeature at 900 cm
−
1
is dominant that is assigned as isolatedMO
6
octahedra. As the concentration increases this banddecreases signiﬁcantly and the dominant feature becomes aband at 810 cm
−
1
that relates to corner sharing octahedralchains. There is a strong secondary feature around 650 cm
−
1
that implies a threedimensional network of MO
6
octahedra isbeginning to form. This interpretation is consistent with theidentiﬁcation of phosphate bands. A band at 1270 cm
−
1
and anintense but broad band between 1050 and 1120 cm
−
1
, whichboth decrease, indicate a shortening of the metaphosphatechains, and a band at
∼
980 cm
−
1
, which increases, indicatesan increase in isolated PO
4
units.
Figure 2.
IR spectra showing evolution of the peaks as a function of Ta and Nb content.
These assignments are further reinforced by the comparison of the Nb and Ta samples. Bands that experience a shiftin frequency and a difference in intensity between the Nb andTa samples almost certainly involve an M–O bond vibration,whereas bandswhich donotchange withMtypeprobably havea different srcin. This is particularly evident with the band at910 cm
−
1
as it is much stronger in the niobium samples thanin the tantalum samples (where it appears at 890 cm
−
1
). Thisis because, within the octahedra of NbO
6
, the niobium is ableto move offcentre, giving rise to a large variation in Nb–Obond lengths. There is therefore a much larger resonance withthe stretching vibration occurring at 910 cm
−
1
. This shows thegreater degree of distortion that occurs within the niobium octahedra compared to that found within the tantalum octahedra.Bands have been assigned to
Q
0
,
Q
1
and
Q
2
, where
Q
n
deﬁnes the connectivity of the phosphate units by
n
bridgingoxygen atoms.
Q
0
is therefore isolated PO
4
tetrahedra,
Q
1
aredimers and
Q
2
are chains. The identiﬁcation of all three typesimplies that there is complexity in the P connectivity: thereare vibrations occurring in the sample that are present in thephosphate units connected to two, one or no other phosphateunits. The presence of all three types of band suggests that P isconnected to networkforming units other than phosphate.
3.2. High energy xray and neutron diffraction measurements
The
i
(
Q
)
data for the four samples Ta19.6, Ta35.0, Nb21.1and Nb36.5 from neutron and xray diffraction are shown inﬁgure 3. The corresponding
T
(
r
)
is shown in ﬁgure 4 forneutron and xray diffraction. Table 3 shows the sequence of interatomic correlations (between atom types
i
and
j
) that wasusedtoﬁtthexray andneutrondiffractiondata simultaneouslyto obtain a single set of structural parameters for each samplethat agree with both techniques. The values obtained by theﬁtting are listed in the table as
R
, the pairwise interatomicdistance,
N
, the number of neighbouring atoms, and
σ
, thestandard deviation in distance.Multicomponent systems such as these are complex tomodel with individual pair correlations due to the presenceof overlapping correlations at
r
values greater than
∼
3 ˚A. It4