Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Journal of Non-Crystalline Solids 357 (2011) 1647–1656

Contents lists available at ScienceDirect

Journal of Non-Crystalline Solids


j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j n o n c r y s o l

The structural role of Zr within alkali borosilicate glasses for nuclear


waste immobilisation
A.J. Connelly a, N.C. Hyatt a,⁎, K.P. Travis a, R.J. Hand a, E.R. Maddrell b, R.J. Short b
a
Immobilisation Science Laboratory, Department of Materials Science and Engineering, Sir Robert Hadfield Building, Mappin Street, University of Sheffield, S1 3JD, UK
b
National Nuclear Laboratory Ltd., Sellafield, Seascale, Cumbria, CA20 1PG, UK

a r t i c l e i n f o a b s t r a c t

Article history: Zirconium is a key constituent element of High Level nuclear Waste (HLW) glasses, occurring both as a fission
Received 7 July 2010 product and a fuel cladding component. As part of a wider research program aimed at optimising the solubility
Received in revised form 24 December 2010 of zirconium in HLW glasses, we have investigated the structural chemistry of zirconium in such materials
using X-ray Absorption Spectroscopy (XAS). Zirconium K-edge XAS data were acquired from several inactive
Keywords:
simulant and simplified waste glass compositions, including a specimen of blended Magnox/UO2 fuel waste
XAS;
Glass;
glass. These data demonstrate that zirconium is immobilised as (octahedral) six-fold coordinate ZrO6 species
Zirconia; in these glasses, with a Zr–O contact distance of 2.09 Å. The next nearest neighbours of the Zr species are Si at
Nuclear; 3.42 Å and possibly Na at 3.44 Å, no next nearest neighbour Zr could be resolved.
Borosiliate © 2011 Elsevier B.V. All rights reserved.

1. Introduction ZrSiO4 (zircon), which may cause cracking of waste glasses due to
thermal expansion mismatch. This work uses results from X-ray
In the UK and France, spent UO2 fuel rods are reprocessed using the absorption spectroscopy and bond valence modelling to propose a
Purex process [1] the (Zircaloy) fuel cladding is mechanically structural model for Zr within alkali borosilicate glasses.
removed and then the spent nuclear fuel dissolved in concentrated XAS investigation by Farges et al. [3] showed dominant ZrO6
nitric acid. The uranium and plutonium are separated from the fission species (i.e. VIZr) within silicate glass (similar to the environment in
products by solvent extraction. High Level Waste (HLW) arising from the mineral zektzerite). Farges et al. also suggested the presence of
this process, comprising the fission products and minor actinides, is ZrO8 (VIIIZr—similar to the environment in zircon) within highly
vitrified in an alkali borosilicate glass matrix. polymerised glasses (i.e. few Non Bridging Oxygens (NBO)).
An alternative to the mechanical removal of fuel cladding material McKeown et al. [4] published similar results with alkali borosilicate
is full chemical dissolution of the entire fuel rod. This leads to glasses showing VIZr and suggested that there was a higher
significant amounts of fuel assembly components being taken into coordination VIIZr species present (similar to the environment in
solution; giving a waste stream high in zirconium, iron, chromium and baddeleylite (ZrO2)), which increased with increasing Zr concentra-
nickel (accounting for 80–90 wt.% on an oxide basis) as well as fission tions. In a study of alkali borosilicate glasses Ferlat et al. [5] suggest
products and non-recycled actinides [2]. Vitrification of such a waste that the VIZr species is highly symmetrical.
stream, with a reasonable waste loading in excess of 10 wt.%, is not The next nearest neighbour (NNN) of Zr has not been clearly
possible within the constraints of the current UK process. This is due defined from experiment. Based on the study of a limited number of
to constraints of processing temperature (requiring temperature aluminosilicate and borosilicate compositions, Farges et al. [3]
b1060 °C), the refractory nature of zirconium oxide (Tm = 2710 ± suggested Si and Al (or Na) as NNNs, and Galoisy et al. [6] proposed
25 °C), and the low solubility of ZrO2 in borosilicate glass. To improve Si next nearest neighbours. Quantitative determination of these NNN
the zirconium solubility, without compromising wasteform durabil- from EXAFS data is a difficult and nontrivial problem: since there are
ity, reformulation of the current alkali borosilicate glass composition several potential backscattering species in a disordered system, the
may be desirable. To achieve this efficiently an understanding of the number of desired structural parameters exceeds the number of data
structural role of Zr within such glasses is required. This knowledge and the problem is therefore mathematically ill posed and the
may also lead to an improved understanding of the role of Zr in the solution non-unique. This is of particular relevance to the case here
formation of crystalline phases such as NaLiZrSi6O15 (zektzerite) and due to the large number of possible solutions for the NNN in glasses
Na, Si, B, etc., and the fact that these shells are expected to overlap in
the same region (3.4 and 4.0 Å). This issue may be mitigated, to some
⁎ Corresponding author. extent, by appropriate parameterisation of the local cluster involved
E-mail address: n.c.hyatt@sheffield.ac.uk (N.C. Hyatt). in the scattering process using selected scattering paths.

0022-3093/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnoncrysol.2011.01.005
1648 A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656

Angeli et al. [7] used MAS-NMR to confirm the presence of Zr–O–Si Table 1
bonds and suggested that Zr–O–B bonds may also be present. There Showing batch compositions of simplified glasses (mol%), analysed compositions do
not deviate significantly from these values. High Zr glass composition is a 15 wt.% waste
has been little evidence of Zr–O–Zr bonds within glasses, only Farges loaded full chemical dissolution composition. The Blend and Magnox waste glasses are
[8] suggested their presence from EXAFS spectra with a separation of 25 wt.% loaded with ISL standard UK waste composition (See Table 2). FS blend is a full
ca. 3.67 Å which is characteristic of edge shared ZrO8 polyhedra as in scale blend stimulant glass provided by Nexia solutions. *not including ZrO2.
zircon.
SiO2 B2O3 Na2O Li2O ZrO2 Waste*
Due to the low backscattering amplitude of B-atoms it is difficult
Hli-BG 63.9% 19.6% 11.0% 5.5% 0.0% 0.0%
for XAS to give detailed information as to the possibility of Zr–O–B
Fli-BG 60.6% 18.5% 10.5% 10.5% 0.0% 0.0%
bonds. In fact, there has been little investigation on the question of Hli-1Zr 63.5% 19.5% 11.0% 5.5% 0.5% 0.0%
such linkages, although there is some evidence for existence in Fli-1Zr 60.3% 18.4% 10.4% 10.4% 0.5% 0.0%
leached phase-separated glasses (e.g. Vycor [9]). Chemical analysis of Hli-5Zr 62.2% 19.1% 10.8% 5.4% 2.5% 0.0%
the leached high borate region in such glasses, and the remaining high Fli-5Zr 59.1% 18.1% 10.2% 10.2% 2.5% 0.0%
Hli-8Zr 61.3% 18.8% 10.6% 5.3% 4.1% 0.0%
silica regions, showed that ZrO2 is found in both the high borate
Fli-8Zr 58.2% 17.8% 10.0% 10.0% 4.0% 0.0%
region and the high silica region, implying formation of Zr–O–B bonds. Simplified R7T7 67.3% 17.4% 13.6% 0.0% 1.8% 0.0%
Molecular Dynamics (MD) calculations of ZrO2 bearing alkali High Zr 56.5% 17.3% 11.0% 7.5% 5.7% 2.0%
borosilicate glasses performed by the authors (Connelly et al. 2010 Blend 53.5% 16.4% 9.2% 9.2% 1.4% 10.3%
Magnox 51.1% 15.6% 8.8% 8.8% 0.8% 14.9%
in press) showed good agreement with the XAS results discussed
FS Blend Composition similar to Blend sample
above [4–6] and additionally showed the presence of B within the
next nearest neighbour coordination shell of these glasses. It also
showed that IIIB was more likely to bond to VIZr than IVB.
The importance of the coordination environment in nuclear waste 2. Experimental
glasses is primarily related to the need to increase solubility and so
reduce the crystallisation of undesirable phases. The compositional Glass melts were undertaken in Pt crucibles with a 5 h melt time,
dependence on ZrO2 solubility has been reported by various authors including 1 h initial melting time and 4 h stirring (with a Pt stirrer)
[10–15] and there appears to be an additive effect. The information rotating at 60 rpm; glasses were then poured into preheated steel
can be presented as 4 principles: block moulds. Glass blocks were annealed for 1 h at 500 °C and cooled
to room temperature at 1 °C/min. Batches were well mixed before
1. The amount of ZrO2 dissolved is approximately proportional to the melting and raw materials from the same source were used for all
alkali content with the effect increasing in the order Li b Na b K; glass melts. All raw materials used were high purity, with the
2. CaO, SrO, and BaO increase the solubility about 1 part for every 5 following major constituents: SiO2 from SiO2 (Sigma, ≈99%, purity,
parts added; particle size 0.5–10 μm); B2O3 from Na2B4O7.10H2O (Fisons, 99.5%
3. Ratio of B2O3 to SiO2 shows relatively small effect on the solubility purity); Na2O from Na2CO3 (Sigma, N99% purity); Li2O from Li2CO3
of ZrO2; (Sigma, N99% purity); and ZrO2 from ZrO2 (Aldrich, 99% purity, 5 μm
4. Al2O3, ZnO and MgO decrease the amount of ZrO2 that may be particle size). Table 1 gives the simplified glass compositions prepared
dissolved (see below). in this study with Table 2 showing the composition of the glasses
containing full stimulant waste streams. All glasses were examined
It has been shown that change in solubility of ZrO2 varies in a with scanning electron microscopy and X-ray powder diffraction to
fixed stoichiometry with the number of alkali ions in excess of that confirm the absence of crystalline inclusions and residual batch
needed to charge compensate aluminium1 [15,16]. This indicates contents.
the need for alkali ions to be available for Zr to be incorporated in X-ray Absorption Spectroscopy (XAS) is sensitive to the short-
the glass. In a study of zircon solubility in felsic melts (water range atomic structure and oxidation state of an absorber atom
saturated analogues of granitic liquids) Watson [17] concluded within a material. In this work zirconium K-edge (17,997.6 eV) XAS
that: “Zircon solubility in felsic melts appears to be controlled by data was collected at room temperature, and 80 K, in transmission
the formation of alkali zirconosilicate complexes of simple (2:1) mode on Station 16.5 of the CCLRC Synchrotron Radiation Source
M2O:ZrO2 stoichiometry.” However, these experiments were car- (Daresbury, UK), using a Si(220) monochromator with horizontal
ried out under 2 kbar pressure at 700–800 °C. This result was focusing [19]. Data were also acquired in transmission mode at room
supported in a Raman study of potassium silicate glasses by Ellison
and Hess [12]. However, they suggested a M2O:ZrO2 ratio of 1:1.
Batch pre-sintering (preventing settling) and increased tempera-
ture also increase the solubility of ZrO2 [15]. Using temperatures up to Table 2
ISL standard waste compositions in mol% [31,32]. Pd and Tc replaced with Ag and Re
1450 °C Wang et al. [18] found that 23.5 wt.% (12.2 mol%) ZrO2 was
respectively.
soluble in a glass of composition 56.9SiO2-30.9Na2O-12.2ZrO2. Similar
solubilities were found in soda-lithia-silicate and soda-lithia-calcia- Oxide Blend Magnox High Zr Oxide Blend Magnox High Zr
silicate glasses. This shows that high levels of alkali are required to Ag2O 0.84% 1.21% NiO 0.94% 0.96% 13.5%
dissolve high ZrO2 concentrations. Work by the present authors using Al2O3 6.86% 16.62% P2O5 0.45% 0.36%
MD showed that each mole of ZrO2 added to a glass required 1 mole of BaO 3.08% 4.21% 0.4% Pr6O11 0.43% 0.17% 0.1%
CdO 0.24% Rb2O 0.78% 0.31%
alkali oxide (Connelly et al. 2010 in press). CeO2 5.25% 2.14% 0.6% Re2O7 0.19% 0.32%
This contribution aims to make direct observations of the structure Cr2O3 1.53% 1.44% 3.5% RuO2 6.38% 2.65% 0.7%
around Zr within alkali borosilicate glasses leading to an improved Cs2O 3.04% 1.44% 0.3% SeO2 0.20% 0.08%
understanding of the structural role of Zr, factors that determine the Eu2O3 0.14% 0.04% SiO2 0.35% 0.58%
Fe2O3 3.44% 4.31% 5.0% Sm2O3 0.83% 0.35% 0.1%
solubility of ZrO2, and the effect of ZrO2 on the physical properties
Gd2O3 4.33% 0.49% 0.6% SnO2 0.13%
(e.g. viscosity) of glasses. La2O3 9.55% 3.89% 0.2% SrO 2.96% 1.18% 0.3%
MgO 18.56% 45.43% TeO2 1.03% 0.43% 0.1%
MnO2 0.10% Y2O3 0.79% 0.31% 0.1%
1 MoO3 10.81% 4.38% 1.2% ZnO 0.16%
This is because Al is preferentially charge compensated. Interestingly the addition
Nd2O3 4.31% 1.73% 0.5% ZrO2 12.3% 4.95% 72.8%
of B2O3 appears to have little effect on this relationship.
A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656 1649

Table 3
Bond lengths (Å) and bond valences for a selection of cations. Bond length (diO)
calculated as sum of ionic radii. RiO (used to calculate viO) values taken from Ref [26].
The syntax used for the cations is such that the superscript Roman numeral before the
element symbol refers to the coordination and the superscript number after the
valence.

Cation diO viO Cation diO viO


IV
Si4+ 1.62 1.01 VI
Na+ 2.38 0.21
III 3+ IV +
B 1.37 1.00 Li 1.95 0.27
IV 3+
B 1.47 0.77

temperature using beamline X23A2 of the National Synchrotron


Light Source (NSLS), Brookhaven National Laboratory (BNL), USA.
This beamline uses an upwards reflecting Si (311) monochromator
of a fixed exit Golovchenko-Cowan design. Glass samples and
crystalline standards (see Tables 1 and 2) were powdered and
sieved to b52 μm and attached to layers of adhesive tape (for highly
absorbing standards boron nitride was intimately mixed to reduce
absorption). Samples were visually checked for inhomogeneities
prior to data acquisition. The thickness of each sample was tuned to Fig. 1. Proposed structure of atoms around the Zr atom. The alkali ions are included in a
charge balancing role. For clarity not all Zr coordinated oxygens are shown as BO, in a
give an optimum absorbance and edge step. Selected samples were
glass it is thought that all Zr coordinated oxygens would be BO.
run both at room temperature and at 80 K using a liquid nitrogen
cold finger.
XAS data were processed and normalised using Athena [20].
Extended X-ray Absorption Fine Structure (EXAFS) data were R, and the χ2v value. The R-factor represents the mean square misfit
analysed used Artemis [20]. The analysis of EXAFS data was carried between the data and the fit for both the real and imaginary parts of
out by creating a structural model [21] (FEFF7 [22]) composed of X- the Fourier transform, a good fit is normally considered to be a few
ray scattering paths from crystalline reference structures (taken from percent. χ2v takes account of the degree of over determinacy in the
the ICSD database) [23]. The structural EXAFS parameters that are system, however, the absolute value of χ2v is not meaningful only the
often determined by a fit to data are Ndegen (degeneracy of the path), relative value on addition of new fitting parameters, compared to the
δri (change in the half-path length), σ 2 (relative mean square previous fit value (significant reductions are of the order of factors
displacement about the equilibrium path length), S20 (passive electron of 2). Values of χ2v are reported as ×103. The value for the passive
reduction factor), and ΔE0i (energy shift of the photoelectron). Fits to electron reduction parameter S20 was found to converge to 1.05 ± 0.1
the EXAFS data were made in R-space, the data sets were aligned and for all Zr K-edge EXAFS data acquired on crystalline standard
the backgrounds were removed by the AUTOBK program (Rbkg, set to materials, at both beamlines. Therefore, S20 = 1.05 was applied to all
0.8 Å) [22]. The goodness-of-fit values are the EXAFS reliability factor, models fitted to data acquired from glass samples. A Hanning window
was used in all fits with a value of dk = 1. The boundaries of the
windows are indicated in figures and in the text where appropriate.
Table 4 FEFFIT can also adjust the background in the fit and report the
Bond valence values for species of relevance to this work. correlation between the background and the structural-fit para-
Species BV (vu) Charge comp. BV (vu) meters. We found that this correlation was always less than 60%,
VI
indicating that the background did not significantly affect the
Zr–O–IVSi 1.66 VI
Na 1.87
IV
Li 1.93
determined structural parameters.
VI
Zr–O–IIIB 1.65 VI
Na 1.86
IV
Li 1.92
VI IV
Zr–O– B 1.42 2(VINa) 1.84
2(IVLi) 1.96
VII
Zr–O–IVSi 1.57 2(VINa) 1.99
2(IVLi) 2.11
VII
Zr–O–IIIB 1.56 2(VINa) 1.98
2(IVLi) 2.10
VII IV
Zr–O– B 1.33 3(VINa) 1.96
2(IVLi) 1.87
VIII IV
Zr–O– Si 1.49 2(VINa) 1.91
2(IVLi) 2.03
VIII
Zr–O–IIIB 1.48 2(VINa) 1.90
2(IVLi) 2.02
VIII
Zr–O–IVB 1.25 4(VINa) 2.09
3(IVLi) 2.06
VI IV
(2 Zr)–O– Si 2.31
(2VIZr)–O–IIIB 2.3
(2VIZr)–O–IVB 2.07
(2VIIZr)–O–IVSi 2.13
(2VIIZr)–O–IIIB 2.12
(2VIIZr)–O–IVB 1.89
(2VIIIZr)–O–IVSi 1.97
(2VIIIZr)–O–IIIB 1.96
(2VIIIZr)–O–IVB 1.73 VI
Na 1.94
IV Fig. 2. XANES data for the standards and 2 representative glass samples. Lines are as a
Li 2.00
guide to the eye only.
1650 A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656

Table 5
Description of the paths included in the models and final fit results for zektzerite and catapleiite. Atomic structures from References [33] and [34] respectively.

Patha R (Å)b Ndegenc Ri σ2 Parameters

Catapleiite
C1 Zr → O1 2.14 6 2.09 (0.01) 0.003 R 0.66%
C2 Zr → Si 3.59 6 3.44 (0.10) 0.008 Nipd 22.5
C3 Zr → Si → O1 3.62 12 3.57 (0.06) 0.005 Nvar 9
C4 Zr → O2 3.70 6 3.59 (0.06) 0.0004 E0 3.58 (± 1.1)
C5 Zr → O1 → Zr → O1 4.28 6 2 × RC1 2 × σ2C1

Zektzerite
Z1 Zr → O1 2.06 6 2.08 (0.01) 0.003 R 1.4%
Z2 Zr → Si 3.44 6 3.46 (0.02) 0.004 Nipd 22.7
Z3 Zr → Si → O1 3.55 12 3.57 (0.03) 0.003 Nvar 9
Z4 Zr → Li 2.97 2 2.92 (0.09) 0.002 E0 1.95 (± 1.11)
a
Atom types in the scattering path.
b
Initial path length from the theoretical models.
c
Degeneracy from the theoretical models.

3. Results and discussion relating to oxygen must be close to the theoretical value of 2.0 valence
units (vu) [25], with an acceptable tolerance of ±0.1 vu [27]. The
3.1. Bond valence analysis oxygen bond valence is given:

Although originally devised for crystalline materials, bond valence vi0 = exp½ Ri0  di0 Þ = b  ð1Þ
(BV) calculations can be used to provide an insight to the validity of
certain structural arrangements within glasses. This analysis is based where viO is the valence of the bond between cation i and oxygen,
on Pauling's second rule [24], that the sum of the bond valences RiO is the bond valence parameter for oxides of cation i, diO is the

Fig. 3. Data (line) and best-fit model (circles) for zektzerite. (a) Magnitude of Fourier Fig. 4. Data (line) and best-fit model (circles) for catapleiite. (a) Magnitude of Fourier
transform, with inset showing real part of Fourier transform (b) χ(k) k3. transform, with inset showing real part of Fourier transform (b) χ(k) k3.
A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656 1651

cation i—oxygen bond length (Å), and b is a constant generally taken provided by Si and IIIB. Thus, ca. 2.1 vu are required to completely
as 0.37 Å [26]. charge balance the 6 bridging oxygen. These 2.1 vu are provided by all
It is thought that VIZr is the primary species present in silicate glass the contributions from ca. 2.1 alkali ions (see Fig. 1).
with Na and Si cations as next nearest neighbours, evidence has also The coordination of VIIZr requires a larger contribution from alkali
been shown for the presence of higher coordinated VIIZr and/or VIIIZr. ions for charge balancing than VIZr due to the low BV of VIIZr (see Tables 3
Previous work by the authors showed the presence of B and Li also as and 4). For silicate bonding, more than one alkali ion contribution is
next nearest neighbours (Connelly et al. 2010 in press). The values for required: VIIZr–O–IVSi+ 2(IVLi) = 2.11 vu and VIIZr–O–IVSi + 2(VINa)=
bond valence sums of relevance to the discussion below can be found 1.99 vu. This is also the case for the borate bonding: VIIZr–O–IIIB =
in Tables 3 and 4. 1.56 vu and VIIZr–O–IVB = 1.33 vu; in this case it is even more difficult to
The presence of Si as a next nearest neighbour to Zr would require achieve charge compensation (see Table 4). In a similar manner to
VI
Zr–O–IVSi bonds giving a BV, for this arrangement, of 1.66 vu. This above, this indicates that a VIIZr species requires ca. 3.1 alkali ions to
value is much less than 2.00 vu and so this bond requires a charge balance 7 bridging oxygen ions to either Si or IIIB.
contribution from an alkali cation for charge compensation. Consid- For VIIIZr the low BV may be too low to consider alkali charge
ering VIZr–O–IVSi + IVLi = 1.93 vu and VIZr–O–IVSi + VINa = 1.87 vu, it compensation. From Table 4 it can be shown that VIIIZr requires the
is apparent that IVLi is marginally the preferred charge balancing contribution of ca. 4.1 alkali ions to charge balance 8 bridging
cation as the resulting BV is closer to 2.00. The presence of VIZr–O–IIIB oxygen ions to either Si or IIIB. For this reason, Zr–O–Zr bonding may
bonds gives a BV for the oxygen of 1.65 vu and for VIZr–O–IVB a BV of occur. For example 2(VIIIZr)–O–IVSi gives a bond valence of 1.97 vu
1.42 vu. This indicates, again, that a contribution from alkali ions is (as is seen in zircon). Consistent with this hypothesis, Watson [17]
required for charge compensation. In fact, it indicates that IIIB is more showed that for highly polymerised glasses there was a propensity
likely to bond to VIZr than IVB, as a lower contribution from the alkali for zircon crystallisation. Farges et al. [3] suggest the presence of
ions is required to increase the value of the oxygen BV to ca. 2. This zircon like units in highly polymerised aluminosilicate glasses with
result is supported by recent MD calculations (Connelly et al. 2010 in the presence of Zr–O–Zr bonds. The bond valence model presented
press). Based on the BV model presented above, assuming that for gives a good model for the propensity of zircon crystallisation from
each Zr–O–X BV must equal 2 vu, then 12 vu are required per VIZr. highly polymerised glasses and also the low solubility of ZrO2 in
Assuming three Zr–O–Si and three Zr–O–IIIB per VIZr then 9.9 vu are such glasses.

Fig. 5. Data (line) and best-fit Model 1 (circles) for glass HLi-8Zr. (a) Magnitude of Fig. 6. Data (line) and best-fit Model 2 (circles) for glass HLi-8Zr. (a) Magnitude of
Fourier transform, with inset showing real part of Fourier transform (b) χ(k) k3. Fourier transform, with inset showing real part of Fourier transform (b) χ(k) k3.
1652 A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656

4. XANES more closely reproduced by zektzerite. However, feature D is at


higher energy for zektzerite than the glasses or catapleiite. This
The complex nature of the XANES process makes deconvolution of indicates shorter Zr–O bond in the zektzerite than in the glasses and
the experimental data, in a similar manner to EXAFS, very difficult for catapleiite.
chemically complex (disordered) systems. However, comparison
between XANES data may show differences in local coordination very 5. EXAFS
effectively. Differences in Zr speciation between the standards and
glasses give rise to the variation in the XANES spectra shown in Fig. 2. XANES spectra presented above show that the structural environ-
The spectra for the two representative glass samples presented in ment of Zr in both zektzerite and catapleiite is very similar to that seen
Fig. 2 closely resemble each other and the spectra for zektzerite and in the glasses. Thus, as a starting point, the structural environment of
catapleiite, indicating a similar Zr coordination environment in these Zr in these minerals was analysed. The best models fitted to the
materials. The features identified in Fig. 2 can be assigned to various spectra, determined by R-factor and chemical plausibility, are
electronic transitions and scattering effects [3,8,27,28]: presented in Table 5, Fig. 3, and Fig. 4. These models are the result
of adding individual paths to the model with each path only being
A. Attributed to 1s–4d transition involving Zr4d/O2p molecular allowed to remain in the model if it sufficiently improved goodness-
orbitals, indicating a VIZr site without a centre of symmetry. The of-fit parameters. These models do not include all of the paths
height and position of pre-edge feature is a function of the degree available from the FEFF calculation of the structures of zektzerite and
of p–d mixing and oxidation state (constant here at + 4). catapleiite because of the highly complex nature of these structures
B. Due primarily to single scattering from second neighbour element and the inherent limited data range of EXAFS. The final models give
(e.g. Si); very good fits with low R values and small errors on determined path
C. Due primarily to multiple scattering involving first neighbour lengths and E0 (Table 5). Quantitative analysis of EXAFS data is in
O around Zr; accordance with qualitative analysis of XANES data, showing that the
VI
D. First EXAFS oscillation—position dependent on Zr–O bond length. Zr–O bond in catapleiite Zr–O bond length is longer than that in
zektzerite. Table 5 shows that the average Zr–O bond length for
If these features are compared for the two glasses, zektzerite and zektzerite from crystallography is ca. 2.08 Å which is in good
catapleiite it can be seen that features A, B and C for the glasses are agreement with the value from EXAFS (2.06 Å).

Fig. 7. Data (line) and best-fit Model 3 (circles) for glass HLi-8Zr. (a) Magnitude of Fig. 8. Data (line) and best-fit Model 4 (circles) for glass HLi-8Zr. (a) Magnitude of
Fourier transform, with inset showing real part of Fourier transform (b) χ(k) k3. Fourier transform, with inset showing real part of Fourier transform (b) χ(k) k3.
A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656 1653

For catapleiite there is a large difference in Zr-O distance, with number being ±20%. Relative errors on bond lengths measured by
crystallographic data a distance of 2.14 Å and EXAFS giving 2.09 Å. EXAFS are much less (generally ±0.01 Å) and the determined Zr–O
Analysis of calcium catapleiite by Merlino et al [29] gave a Zr–O distance bond lengths are consistent with a majority of 6 coordinated Zr
of 2.08 Å which is a an excellent agreement with that determined in this species. First shell fits of ZrO2 and zircon (not shown) gave Zr–O bond
paper. A Zr–O distance of 2.14 Å is significantly greater than the typical lengths of ca. 2.14 Å for 7 coordinate ZrO2 and 2.24 Å for 8 coordinate
Zr–O distance for VIZr–O and suggests that the published crystallo- zircon, in good agreement with published crystallographic data. If
graphic parameters for catapleiite may warrant further investigation. these values are compared to those of zektzerite it can be seen that the
The fitting of the glasses is more complex as there is no relevant Zr–O bond length in the glasses is consistent with VIZr. NZ1 values for
analogue crystal structure available from which to build a model, other glasses measured in this work are all close to 6 (see Table 8), and
thus, the model was built using the structure of zektzerite as a basis, hence the restraint NZ1 = 6 was imposed for all subsequent fitting
given the similarity of XANES spectra. Due to the large number of models.
glass compositions investigated, we discuss in detail the analysis of In the presence of a symmetrical octahedral ZrO6 unit, multiple
fitting to the EXAFS data for a single representative model glass scattering (ms) paths would be expected to be detectable as collinear
composition (HLi-8Zr), for sake of brevity. The fits of these models to or nearly collinear multiple scattering paths generally give have high
the data are presented in Figs. 5–10, and the fitted structural backscattering amplitude. The effect of ms paths was investigated in
parameters in Tables 6–7. Results for the other glasses are presented Model 3 (Fig. 7 and Table 6); the addition of this parameter to the
in Tables 8 and 9 and in Figs. 11 and 12. model reduced both R and χ2v significantly and so this additional
In Model 1, the first shell included 6 oxygen atoms with the parameter was retained. The presence of a symmetrical ZrO6 unit was
average Zr–O distance present in the crystal structure of zektzerite also suggested by Ferlat et al. [5].
(see Figs. 5 and 6). This initial model afforded a good fit to the data As discussed above, BV theory can be used to identify plausible
with Zr–O bond lengths of 2.09 Å (±0.01), which agrees well with intermediate range structural order around Zr within alkali borosil-
literature data for VIZr in silicate glasses [3]. In Model 2, the number of icate glasses In this work models containing paths for various different
oxygen atoms in the first shell (NZ1) was allowed to refine (with fixed species at the intermediate range were compared. Attempts to fit Zr,
S20 = 1.05, see Fig. 6 and Table 6). This gave a NZ1 of 6.6(±1.0). This Li, and B next nearest neighbours led to either only marginal
represents an error of ca. ±15% which is reasonable for coordination improvement or significantly inferior fits, as determined from the
number measured with EXAFS, a standard error for coordination goodness-of-fit parameters. Attempts to fit Si alone also afforded a

Fig. 9. Data (line) and best-fit Model 5 (circles) for glass HLi-8Zr. (a) Magnitude of Fig. 10. Data (line) and best-fit Model 6 (circles) for glass HLi-8Zr. (a) Magnitude of
Fourier transform, with inset showing real part of Fourier transform (b) χ(k) k3. Fourier transform, with inset showing real part of Fourier transform (b) χ(k) k3.
1654 A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656

Table 6
Description of the paths included in the glass model and final fit results. K-range [2.8–11.5], R-range [1.2–4.5].

Path R (Å) Ndegen Ri σ2 Parameters

Model 1
Z1 Zr → O1 2.06 6 2.09 (± 0.01) 0.005 R 1.9%
χ2v 15.1
Nipd 17.9
Nvar 3
E0 1.23 (± 2.13)

Model 2
Z1 Zr → O1 2.06 NZ1 (6.6(± 1.0)) 2.08 (± 0.02) 0.006 R 1.9%
χ2v 15.6
Nipd 17.9
Nvar 4
E0 0.86 (± 2.40)

Model 3
Z1 Zr → O1 2.06 6 2.09 (± 0.01) 0.004 R 0.95%
Z5 Zr → O1 → Zr → O1 4.15 6 2xRZ1 2x σ2Z1 χ2v 10.0
Nipd 3
Nvar 17.9
E0 (2.16 ± 1.68)

small improvement in goodness-of-fit parameters; however, addition neighbour species to the Zr K-edge EXAFS should be from coordinated
of Si with the relevant ms path gives a significant improvement in fit Si atoms. The presence of B as a NNN to Zr is also limited due, in part,
(Model 4 Fig. 8 and Table 7). Model 4 shows that the presence of Si to the low B2O3 content of the glasses (see Table 1) and due to the
atoms greatly improves the fit quality, although the precision on coordination of the B that is present. As discussed above, Bond Valence
refined coordination number is poor. Theoretical bond valence model models show that 4-coordinated B (IVB) is unlikely to form Zr–O–B
also suggests the presences of Si as a next nearest neighbour (NNN). and that IIIB bond are more likely. The high alkali oxide to B2O3 ratio in
The other glasses measured in this work all show a similar the glasses under consideration means that the fraction of IIIB
contribution of NNN Si (see Table 8). The large error in the value of available for bonding is significantly smaller than the overall content
NSi is due to the strong correlation between NZ2 and σ2Z2 which gives a of B2O3. In fact, approximately 60% of the B-atoms are IVB [30] leaving
high error on this value. To investigate the potential for Si NNN only 40% IIIB available to form Zr–O–IIIB. Thus, assuming that Si and IIIB
further, the model was refined with the number of Si NNN (with ms are equally likely to form Zr–O–X bonds then the 6 oxygens of a VIZr
paths) fixed at integral values between 1 and 6: the best fit was species would be bonded to ca. 5 Si and only ca. 1 IIIB. This would make
generally achieved with 4 Si atoms (see Model 5). The presence of the detection of B as a NNN to VIZr very difficult.
ms paths along with the Si as NNN suggests a highly regular Model 6 shows that trying to fit Na within the local structure
environment for the Zr which increases the probability of crystal- around Zr gives a significant improvement in fit quality, this is not as
lisation from the melt. significant as for Si. However, if Na is fitted without holding the
Ferlat et al. [5] showed that B and Na have only a small impact on number of Na atoms being fixed the fit does not converge on a
Zr EXAFS signal, Li would also be expected to have only a small effect physically realistic model. Thus, Model 6 uses a fixed number of Na
of the EXAFS spectra. This is due to the low back scattering amplitudes atoms. The number of Na atoms in the model was increased as
of these elements. In the case of Na and Li there is also an inherent integers and NZ2 = 3 chosen giving the best fit for a physically realistic
disorder in the position of these network modifier elements. It is model. From Bond Valence analysis there must be the equivalent
therefore expected that the dominant contribution of next nearest charge compensation from 2 alkali ions in the coordination shell of Zr.

Table 7
Description of the paths included in the models and final fit results. aThis path was not used in zektzerite fit (Table 5) as it did not lead to significant improvement.

Path R (Å) Ndegen Ri σ2 Parameters

Model 4
Z1 Zr → O1 2.06 6 2.09 (± 0.01) 0.005 R 0.71%
Z5 Zr → O1 → Zr → O1 4.15 6 2xRZ1 2x σ2Z1 χ2v 5.3
Z2 Zr → Si 3.44 NZ2 (3.7 ± 6.0) 3.42(± 0.07) 0.006 Nipd 17.9
Z3 Zr → Si → O1 3.55 2*NZ2 3.58 (± 0.06) 0.003 Nvar 8
E0 1.52(± 1.44)

Model 5
Z1 Zr → O1 2.06 6 2.09 (± 0.01) 0.005 R 0.71%
Z5 Zr → O1 → Zr → O1 4.15 6 2xRZ1 2x σ2Z1 χ2v 4.9
Z2 Zr → Si 3.44 4 3.42(± 0.06) 0.007 Nipd 17.9
Z3 Zr → Si → O1 3.55 2*NZ2 3.58 (± 0.04) 0.004 Nvar 8
E0 1.51(± 1.36)

Model 6
Z1 Zr → O1 2.06 6 2.09 (± 0.01) 0.005 R 1.3%
Z5 Zr → O1 → Zr → O1 4.15 6 2xRZ1 2x σ2Z1 χ2v 6.7
Z6a Zr → Na 3.39 3 3.44(± 0.03) 0.006 Nipd 17.9
Nvar 5
E0 (1.88 ± 1.33)
A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656 1655

Table 8
Goodness of fit values of Models 1, 2, and 3 described above to other glasses measured
in this project, where R indicates R-factor (%). K-range [2.8–11], R-range [1.2–4.5].
‘–’ indicates an unphysical result. FS indicates full scale simulated HLW glass.

Sample Model 1 Model 2 Model 3

R χ 2v NZ1 R χ 2
v R χ 2v

FLi-1Zr 1.5 10.9 6.3 (± 0.9) 1.5 11.7 0.58 5.8


FLi-5Zr 1.6 67.8 6.5 (± 1.0) 1.6 71.0 0.67 41.8
HLi-1Zr 1.6 2.9 6.3 (± 0.9) 1.6 3.1 0.64 1.5
HLi-5Zr 1.6 18.7 6.4 (± 0.9) 1.6 19.8 0.67 10.2
HLi-5Zr (80 K) 1.6 45.5 6.4 (± 1.0) 1.6 48.1 0.66 26.7
Simplified R7T7 3.2 38.0 6.1 (± 1.1) 3.3 40.9 1.2 19.8
High Zr 1.6 67.6 6.4 (± 1.0) 1.6 71.4 0.66 39.6
Magnox 1.4 2.0 6.2 (± 0.8) 1.4 2.1 0.55 0.95
FS Thorp 1.4 0.61 6.2 (± 0.8) 1.4 0.65 0.50 0.27
FS Blend 1.4 3.9 6.2 (± 0.8) 1.5 4.1 0.61 2.0

However, this may come as contributions from many alkali ions. It is Fig. 11. Data (line) and best-fit Model 5 (circles) for other glasses: magnitude of Fourier
difficult to predict how many alkali ions would be associated with a Zr transform.
oxycation. Thus, the presence of 3 Na atoms is not unrealistic.
The results of applying the Models 1–6 to all of the data considered
in this work are shown in Tables 8 and 9 and Figs. 11 and 12. These (except Mg), the smaller effect being due to steric hindrance in the
data show that the spectra for all of the samples are very similar and NNN of the Zr oxyanion. The structure present with Si (and B) as NNN
the fits are very good for all of the models. The change in χ2v between would reduce the space available for large alkaline earth ions.
the different models is similar for all of the samples and agrees well The third principle suggests that little change is seen in solubility
with HLi-8Zr discussed in detail above. This shows that the Zr with changing B2O3:SiO2 ratio within the glass. This can be
environment of fully loaded and simplified glasses is very similar up to understood as bond valence analysis shows that there is little
a Zr loading of ca. 4 mol% (ca. 8 wt.%). difference in stability with the presence of 3 coordinated B or Si as
From the data presented in this paper it can be said that the NNN. Some variation might be expected where an increase in 3
majority of Zr atoms within the alkali borosilicate glasses have 6 coordinated B (over 4 coordinated) is found. The final principle where
oxygen ions in the first coordination shell. With a Zr–O distance of ca. Al2O3, ZnO and MgO are suggested the decrease ZrO2 solubility is
2.09 Å. This agrees well with literature data [3–5]. The next nearest more complex. It would suggest that Al2O3, ZnO and MgO are all
neighbour species are more difficult to ascertain from EXAFS. Fig. 1 preferentially charge compensated over ZrO2 so removing alkali ions
showed the environment of Zr developed from Bond Valence analysis which could potentially have charge compensated ZrO2. This has been
where there is thought to be Si and IIIB (less likely IVB) as NNN with at observed with Al2O3 [15,16] but is not expected for ZnO or MgO,
least 2 alkali ions to provide charge compensation. Evidence from however, the effect of these elements of bulk glass structure is not
EXAFS analysis shown in this work suggests that the presence of Si in well understood at present.
very regular arrangements (with ms paths) and Na in a similar
manner to the model. However, EXAFS analysis alone cannot confirm 6. Conclusions
or refute the model further for reasons described in the text.
Taking a general overview of the literature and the work presented Zr K-edge XAS data indicate that Zr is primarily present in all of the
here, Fig. 1 represents the state of knowledge of the environment of alkali borosilicate glasses studies here as ZrO6 species with a Zr–O
the majority of Zr within alkali borosilicate and silicate glasses [3–6]. contact distance of 2.09 Å, although a small fraction of ZrO7 and/or
This being the case the general principles for the compositional ZrO8 species may also be present.
dependence of the solubility of ZrO2 set out in the introduction can be The fits to glass structure presented in this work suggest that the
understood from a structural standpoint. The first principle was that local environment of Zr within both model and full scale (simulant)
increasing alkali increased ZrO2 solubility. This can be understood in
terms of increased alkali availability for charge compensation. The
same can be said of the second principle with alkaline earth elements

Table 9
Goodness of fit values of Models 4, 5, and 6 described above to other glasses measured
in this project, where R indicates R-factor (%). K-range [2.8–11], R-range [1.2–4.5].
‘–’ indicates an unphysical result. FS indicates full scale simulated HLW glass.

Sample Model 4 Model 5 Model 6


2 2
NZ2 R χ v R χ v R χ2v

FLi-1Zr 3.7 (± 4.4) 0.27 2.7 0.27 2.4 0.72 4.9


FLi-5Zr 3.6 (± 4.7) 0.33 17.8 0.33 16.0 0.85 28.8
HLi-1Zr – – – 0.27 0.61 0.67 1.1
HLi-5Zr 3.1 (± 4.4) 0.31 4.9 0.31 4.4 0.74 8.0
HLi-5Zr (80 K) 3.2 (± 4.3) 0.29 9.9 0.29 8.9 0.75 16.9
Simplified R7T7 – – – 0.54 11.1 1.1 15.8
High Zr 3.9 (± 5.1) 0.32 18.4 0.32 16.5 0.78 28.9
Magnox 3.6 (± 5.1) 0.29 0.64 0.29 0.58 0.66 1.0
FS Thorp 3.9 (± 4.8) 0.25 0.17 0.25 0.15 0.61 0.30
FS Blend 4.1 (± 4.6) 0.32 1.2 0.32 1.1 0.74 2.1
Fig. 12. Data (line) and best-fit Model 5 (circles) for other glasses: χ(k) k3.
1656 A.J. Connelly et al. / Journal of Non-Crystalline Solids 357 (2011) 1647–1656

UK nuclear waste glass are very similar to the mineral zektzerite [4] D.A. McKeown, I.S. Muller, A.C. Buechele, I.L. Pegg, J. Non-Cryst. Solids 258 (1999)
98–109.
with a regular octahedral oxygen environment. In the intermediate [5] G. Ferlat, L. Cormier, M.H. Thibault, L. Galoisy, G. Calas, J.M. Delaye, D. Ghaleb, Phys.
range the presence of Si or Na is likely with a Zr–Si distance of Rev. B 73 (2006) 214207.
3.43 Å (± 0.07) and a Zr–Na distance of 3.44 Å (± 0.03). The [6] L. Galoisy, E. Pelegrin, M.A. Arrio, P. Ildefonse, G. Calas, J. Am. Ceram. Soc. 82
(1999) 2219–2224.
presence of Li and B could be not resolved but both bond valence [7] F. Angeli, T. Charpentier, M. Gailard, P. Jollivet, J. Non-Crys. Solids 354 (2008)
modelling and MD (Connelly et al. 2010 in press) would suggest its 3712–3722.
presence. It is thought that there is no Zr within the intermediate [8] F. Farges, Chem. Geophys 127 (1996) 268.
[9] T. Yazawa, H. Tanaka, K. Eguchi, S. Yokoyama, J. Mater. Sci. 29 (1994) 3433–3440.
range structure of Zr within any of the glasses presented here. [10] A.I. Andrews, R.W. Gates, J. Am. Ceram. Soc. 23 (1940) 288.
Taking a simplified view of glass structure; improvement of Zr [11] V. Dimbleby, S. English, E.M. Firth, F.W. Hodkin, W.E.S. Turner, J. Soc. Glass
solubility can be seen as the need to increase the number of “sites” at Technol. 11 (1927) 52.
[12] A.J.G. Ellison, P.C. Hess, Geochim. Cosmochim. Acta 58 (1994) 1877.
which Zr can fit into the borosilicate glass network. The results
[13] B.W. King, A.I. Andrews, J. Am. Cer. Soc. 24 (1941) 367.
presented suggest that the presence of Na (or likely any alkali (ne [14] C.J. Kinzie, C.H. Commons, J. Am. Cer. Soc. 17 (1934) 283.
earth) ions) may have a strong influence on the stabilisation Zr within [15] M.B. Volf, Chemical Approach to Glass, Elsevier Science, New York, 1984.
these glasses. [16] A.J. Ellison, P.C. Hess, Contrib. Min. Pet. 94 (1986) 343.
[17] E.B. Watson, Cont. Min. Pet. 70 (1979) 407.
[18] T. Wang, R. Hand, P. James, C.R. Scales, Glass Technol. 43C (2002) 162.
Acknowledgments [19] R.L. Bilsborrow, P.A. Atkinson, N. Bliss, A.J. Dent, B.R. Dobson, P.C. Stephenson, J.
Sync. Rad 13 (2006) 54–58.
[20] B. Ravel, M. Newville, J. Syn. Rad. 12 (2005) 537–541.
We are grateful to Bob Bilsborow, Bruce Ravel and Joe Woicik, for [21] S.D. Kelly, D. Hesterberg, B. Ravel, Analysis of Soils and Minerals Using X-ray
assistance in the collection of XAS data. We gratefully acknowledge Absorption Spectroscopy, in: A.L. Ulery, L.R. Drees (Eds.), Methods of Soil Analysis,
EPSRC and Nexia Solutions Ltd. for provision of a CASE award. We also Part 5—Mineralogical Methods, Soil Science Society of America, Madison, WI, USA,
2008.
wish to acknowledge the use of the EPSRCs Chemical Database Service [22] S.I. Zabinsky, J.J. Rehr, A. Ankudinov, R.C. Albers, M.J. Eller, Phys. Rev. B 52 (1995)
at Daresbury. Use of the National Synchrotron Light Source, 2995–3009.
Brookhaven National Laboratory, was supported by the US Depart- [23] D.A. Fletcher, R.F. McMeeking, D. Parkin, J. Chem. Inf. Comput. Sci. 36 (1996)
746–749.
ment of Energy, Office of Science, Office of Basic Energy Sciences, [24] L. Pauling, J. Am. Chem. Soc. 51 (1929) 1010.
under Contract No. DE-AC02-98CH10886. We are grateful to CLRC for [25] I.D. Brown, The Chemical Bond in Inorganic Chemistry: The Bond Valence Model,
the award of beamtime at the Synchrotron Radiation Source, Dares- Oxford university pres, Oxford, 2002.
[26] N.E. Brese, M. O'Keeffe, Acta Crys. B 47 (1991) 192.
bury Laboratory. NCH gratefully acknowledges funding from NDA and
[27] G.E. Brown Jr., F. Farges, G. Calas, Rev. Mineralog. 32 (1995).
the Royal Academy of Engineering. [28] P. Li, I.-W. Chen, J.E. Penner-Hahn, Phys. Rev. B 48 (1993) 63.
[29] S. Merlino, et al., Can. Miner 42 (2004) 1037.
References [30] W.J. Dell, P.J. Bray, S.Z. Xiao, J. NonCryst. Solids 58 (1983) 1.
[31] C. Magrabi, Reference Wastes for the Thorp and Thorp/Magnox Blended
[1] P.D. Wilson (Ed.), The Nuclear Fuel Cycle—From ore to Wastes, Oxford University Vitrification Development, BNFL—Research and Development Dept., Sellafield,
Press, 1996. 1988.
[2] N.C. Hyatt, R.J. Short, R.J. Hand, W.E. Lee, F. Livens, J.M. Charnock, R.L. Bilsborrow, [32] K.S. Matlack, I.S. Muller, H. Hojaji, I.L. Pegg, C. Adearn, C.R. Scales, Mat. Res. Soc.
Ceram. Trans. 168 (2005) 179–187. Sym. Pro. (1999) 247–254.
[3] F. Farges, C.W. Ponader, G.E. Brown, Geochim. Cosmochim. Acta 55 (1991) [33] S. Ghose, C. Wan, Am. Mineral. 63 (1978) 304–310.
1563–1574. [34] B. Brunovskii, Acta Phys. 5 (1936) 863–892.

You might also like