SON68 glass alteration under Si-rich solutions at low temperature (35–90 °C): kinetics, secondary phases and isotopic exchange studies

Abstract

Pristine and ²⁹Si-doped SON68 glass was leached in dynamic mode under silica-rich synthetic Callovo-Oxfordian (COx) groundwater at pH 8, a glass surface-to-liquid volume (S/V) ratio (900–1800 m⁻¹) and at 35, 50 and 90 °C. The solutions were analysed by ICP-MS and ion chromatography, the alteration products were studied by electron microscopy and Raman spectroscopy, and the Si-isotopes profiles were obtained by time-of-flight secondary ion mass spectrometry (TOF-SIMS). The glass alteration seems to be governed by both diffusion and surface reaction processes. After 653 days of alteration the normalized leaching rates were 2.6 (±0.2) × 10⁻⁵, 9.9 (±0.8) × 10⁻⁵ and 2.4 (±0.2) × 10⁻³ g m⁻² d⁻¹ at 35, 50 and 90 °C, respectively. The major alteration secondary phase is clay-type Mg-silicates at all temperatures in addition to powellite and apatite at 90 °C. SIMS studies clearly showed uptake of ²⁹Si by the surface gel via condensation and in the Mg-rich phyllosilicates via precipitation from solution. The precipitation of phyllosilicates at all temperatures constitutes the main process which destabilises the gel layer, thus maintaining a long-term glass dissolution rate in COx groundwater, higher than in pure water.

---

1. Introduction

In France, fission products and minor actinides arising from spent fuel reprocessing are vitrified in a borosilicate glass matrix. The resulting glass composed of about 30 different oxides, named R7T7, conditioned in stainless steel container is expected to be stored in a deep geological disposal located between 420 and 550 m depth in an argillaceous Callovo-Oxfordian (COx) formation. This multi-barrier containment concept is designed to minimize as far as possible the dispersion of radionuclides (RN) into the biosphere. Nevertheless, the alteration of the glass by groundwater cannot be excluded for long-term. Alteration may initially (some hundreds to thousands years) occur by vapor as long as void spaces are not saturated by water followed by alteration in groundwater. In order to prove the safety of disposal, the long-term behaviour of nuclear glass in contact with aqueous solutions have been extensively published.^1–5 The first approach was to conduct in-lab leaching experiments on simplified systems. Experiments were used to determine the glass alteration mechanisms in the short-term and to evaluate the effects of various parameters such as pH, temperature and glass composition on the glass alteration. On the basis of this approach, scientists established models^2,6–8 to predict the long-term glass alteration, based on thermodynamic and kinetic considerations. In parallel, in order to validate these models, the analogy of nuclear glasses with basaltic glasses,^9–14 and archaeological glasses,^15–20 has been studied, bridging the gap between short term data obtained in well constrained conditions and long term data in natural environments.

Many different studies under water saturated conditions have demonstrated that glass corrosion is a result of various processes. The first one is the hydration of the glass and the diffusion of water into the glass network followed by ion exchange between protons in solution and alkali metals located at network terminal sites. This process leads to the formation of a hydrated glass surface region.²¹ The release of glass elements follows a slope of −1/2 in double logarithmic diagrams, indicating a diffusion process.¹ A recent study by Neeway et al.²² using Atom Probe Tomography (APT) clearly developed an interdiffusion model based on Fick’s second law to derive diffusion coefficients of alkalis. In silica saturated conditions, the mass loss of glass tracers follows also an inverse square root of time dependence.²³ Many studies showed that the hydration and interdiffusion processes also involves the hydrolysis of the bonds of boron in the hydrated glass, its dissolution and release of B species although it is a glass former.^24,25 As a consequence of these reactions, the remaining silicate network undergoes deep reorganizations to balance the charges and minimize the free energy of the system.²⁵ The second key process is the hydrolysis, which implies the attack of the Si–O–M bonds (M = Si, Al, Zr,…) by OH⁻, H₂O or H₃O⁺, leading to a further disturbance of the silica network connectivity.²⁶ This process leads to dissolution of silica. A recent study by Hellmann et al.²⁷ proposed another mechanism for glass corrosion using a variety of characterization tools including APT. The mechanism, denoted as ‘interfacial dissolution–precipitation’, describes the glass dissolution as an interaction between water molecules and nanometre-scale level glass surface leaving the soluble species in solution (Na, Li, B,…) while the less soluble species (Si, Fe, Al,…) precipitate.

In systems with limited water volume, accumulation of dissolved Si will occur until a pH dependent certain concentration limit, which is often termed “saturation concentration”. For Si concentrations far below saturation, the glass dissolution rate is the fastest and depends essentially on the temperature, pH and glass composition^28–31 and to a lesser extent to the presence of some inorganic³² or organic species.³³ Saturation occurs due to the existence of a backward reaction of hydrolysis, i.e. the condensation reaction of silica. The backward reaction occurs in a wide range of pH and involves partly or totally detached species from the glass surface.³⁴ This process leads to the formation of a gel at the glass surface. This step is tied to a slowing of the dissolution rate. The drop of the rate is due to (1) an affinity process (i.e. saturation of the solution with dissolved species;⁶), (2) the growth of the gel layer that forms a diffusion barrier between the pristine glass and the leachate.^7,35 After this step, the dissolution rate of the glass continues to drop to reach the residual rate. For the SON68 glass a residual rate of 10⁻⁴ g m⁻² d⁻¹ is assumed in deionized water at 90 °C, a value about four orders of magnitude smaller than the initial rate.^36,37 This regime corresponds to the so called residual rate. It has been observed under specific conditions (pH > 10 and or T > 100 °C) that a resumption of alteration may occur after some time of glass corrosion with the residual rate. This effect was attributed to the precipitation of zeolite minerals such as analcime or philipsite at the gel surface^2,38,39 changing the solution chemistry and the transport properties of the gel. Fig. 1 reminds the relationships between the basic mechanisms and kinetic regimes of glass dissolution. It appears that the study of the formation of the gel layer is crucial to better understand and model the mechanisms of glass alteration. The gel is amorphous, porous and rich in silica, in very sparingly soluble elements (Al, Fe, Zr,…), in alkaline-earth elements and in actinides.⁴⁰ However, it is depleted in soluble glass constituents (Na, Li, B, Mo,…) and alkali elements.⁴¹ The gel structure (e.g. mechanical cohesion, porosity,…) are greatly influenced by several parameters such as pH, temperature, length of experience, solution composition and the glass composition.⁴² At present, the mechanism of gel formation is controversial. Two major hypotheses have been proposed. For basaltic glasses, many authors consider a mechanism of dissolution/precipitation.^12,43,44 This process is often thought of requiring a thermodynamic equilibrium with the entire solution (case of static test conditions). The layer formed on the basalt glass surface, denoted palagonite, has a different nature from that of the pristine glass. This phenomenon has been observed in the alteration of basaltic glass in seawater.⁴⁵ Concerning nuclear waste glasses, some authors^46,47 studied the structure of the gels by Raman spectroscopy and nuclear magnetic resonance (NMR) for ²⁹Si and ¹⁷O isotopes, and highlighted the hypothesis of in situ silica condensation. Unlike the mechanism of dissolution/precipitation, this process involves a thermodynamic equilibrium with local chemical environments, which can be different from that of the entire solution. Both proposed mechanisms are not incompatible and could coexist. In diluted solutions they may be linked by the rate of transport of dissolved silica in the aqueous solution, while in saturated solutions both processes appear to be identical. To study the mechanism of dissolution/precipitation versus Si condensation, previous experiments have been done using ²⁹Si-rich tracing solutions: (1) in static mode at a high S/V ratio,⁴⁸ and (2) in dynamic mode at low S/V ratio.^49,50 The isotopic analysis of altered glass layers by ionic microprobe (SIMS) revealed that the gel was formed by hydrolysis/condensation mechanism whereas phyllosilicates were formed by dissolution/precipitation mechanism.⁴⁹ A recent study performed with the international simple glass (ISG), a 6 oxide glass with the same molar ratio as the R7T7 reference glass,⁵¹ showed that under silica saturation conditions, the gel is only formed by in situ condensation of Si–O–M bonds (M = Si, Al, Zr) following the release of Na, Ca and B atoms.²⁵

Fig. 1 A schematic representation of the main kinetic regimes of the glass corrosion process (adapted from Poinssot and Gin⁵³).

The present study was carried out with two main goals: (a) to investigate the dissolution rate of SON68 glass in flowing solutions through experiments at S/V ratio between 900 and 1800 m⁻¹ and at 35, 50 and 90 °C. The choice of these temperature values was motivated by the wish to cover several situations: in the French design 90 °C is the expected temperature shortly after the closing of the disposal galleries; 50 °C is the expected temperature after the breach of the overpack containers and 35 °C is close to the temperature of Callovo-Oxfordian formation at 500 m depth.⁵² Many experiments have been conducted at 90 °C but only few data are available at lower temperature. (b) To better understand the mechanisms responsible for the formation of the alteration layer (gel + secondary crystalline phases) by tracing isotopic exchanges between the solution and the glass. To do so, we performed two additional experiments using ²⁹Si doped SON68 and ²⁹Si enriched COx water. Despite the advantages of using ²⁹Si as a tracer, we are not aware of literature work on ²⁹Si doped SON68 glass. Here, insight into glass corrosion mechanisms relies on the use of isotope sensitive analytical techniques in the solid and liquid phase.

2. Experimental procedure

2.1. Glass specimen preparation

In this work experiments were performed with the simulated French inactive reference nuclear glass denoted as SON68. Additional tests were conducted with ²⁹Si-doped SON68 glass ([²⁹Si] = 14%) which corresponds to a ratio of ²⁹Si/²⁸Si = 0.16. The glass composition is indicated in Table 1. The glass provided by the French Atomic Commission (CEA) was ground with a mortar and pestle to minimize the production of fine particles. The glass powder was fractionated by sieving. The 32–100 μm fraction was chosen for all experiments to avoid its complete dissolution for the long-term experiments. Thin glass monoliths (1 × 1 × 0.1 cm) were cut from a glass block and polished to 1 μm. Glass powder and monoliths were then cleaned in ethanol during 1 h using the ultrasonic cleaner in order to remove fine particles. Specific surface area measured by the BET method gives values of in the range of 0.37 ± 0.1 to 0.73 ± 0.07 m² g⁻¹. These values seem to be very high compared to the calculated geometric surface area of 0.047 m² g⁻¹ using McGrail et al.⁵⁴ relationship:

S_geo = (3/ρ × r)

where S_geo is the surface area (m² g⁻¹), ρ is the glass density (2.63 g cm⁻³) and r is average radius (m).

Table 1 Chemical composition in weight of the SON68 nuclear glass

Oxide	wt%	Oxide	wt%	Oxide	wt%
SiO2	45.85	ZnO	2.53	Nd2O3	2.04
B2O3	14.14	P2O5	0.29	Pr2O3	0.46
Na2O	10.22	SrO	0.35	Ag2O	0.03
Al2O3	5.00	ZrO2	2.75	CdO	0.03
CaO	4.07	MoO3	1.78	SnO2	0.02
Li2O	1.99	Cs2O	1.12	TeO2	0.23
Fe2O3	3.03	BaO	0.62	Ce2O3	0.97
NiO	0.43	Y2O3	0.20	Others	0.39
Cr2O3	0.53	La2O3	0.93

---

Recent publications suggested the use of geometric surface area rather than BET data because of high uncertainties for glass powder.^55,56 Fournier et al.⁵⁵ conducted dissolution experiments with a soda lime glass in the form of monoliths with a known geometric surface area and glass powder. They found that the rates measured on powders using geometric surface area (S_geo) instead of BET surface area (S_BET) are closest to those found for monoliths. They suggested that features (e.g. defects) at the atomic scale could contribute to overestimation of the S_BET. Thereby, we chose to use the geometric surface area for glass dissolution kinetics calculations.

2.2. Solution preparations

Initially Euriso-Top amorphous silica ²⁹SiO₂ or Aldrich SiO₂ powder were dissolved in 0.2 M NaOH solution to obtain a 1500 ppm ²⁹Si or ²⁸Si stock solutions, respectively. ICP-MS and HR-ICP-MS analyses of stock solutions gave values of 1500 ± 60 ppm of ²⁸Si and 1500 ± 15 ppm of ²⁹Si. Tests were conducted with synthetic Callovo-Oxfordian (COx) water. The chemical composition of pore waters equilibrated with the clay stone at different temperatures is shown in Table 2. The synthetic water in equilibrium with the claystone was prepared according to the procedure described by Gaucher et al.⁵⁷ In order to work under saturation conditions, COx water initially containing ²⁸Si was adjusted to 42 ppm by addition of the ²⁸Si stock solution, giving pH ranging from 6.6 to 7.5 and which was adjusted to 8 using 2 M NaOH solution. Leaching solutions doped in ²⁹Si were prepared by diluting a ²⁹Si stock solution in COx water free of silicon. Chemical analyses of inlet solutions gave values ranging from 38 to 40 ppm of silicon, a small part of this latter concentration precipitated at pH 8. Leaching solutions were changed every month to avoid possible degradation. All solutions were prepared with ultrapure water.

Table 2 Nominal composition of water in equilibrium with the clay stone at varying temperatures⁵⁷

Component	Concentration (mmol L^−1)
	35 °C	50 °C	90 °C
Cl	41	41	41
S(VI)	15.6	14	10
Na	45.6	42	39
K	1	1	0.96
Ca	7.36	9.9	10
Mg	6.67	4.1	2.5
Sr	0.17	0.2	0.17
Si	0.18	0.35	0.84
TIC	2.62	3.05	3.3

---

2.3. Glass leaching procedure

The experimental setup is shown schematically in Fig. 2. All batch experiments were carried out in dynamic mode and in poly(tetrafluoroethylene) (PTFE) reactors (V = 30 ± 0.5 mL) at temperatures of 35, 50 and 90 °C. To assess repeatability each manipulation was repeated three times. A Watson Marlow 205 CA peristaltic pump was used to ensure a continuous flow of leaching solution in equilibrium with the air from inlet reservoirs to the reactors via Teflon tubing. The flow rate was 4 ± 1 mL per day (replacement of ∼10–16% reactor volume per day). The amount of glass powder introduced into the reactor was determined from the target S/V of the experiment and the specific surface area of the powder. For the experiments using ²⁹Si as tracer, a polished glass monolith was placed on a holder in the vials in order to allow an isotopic analysis of the solid by TOF-SIMS at the end of the experiment. At the beginning of the test the glass surface area/solution volume ratio (S/V) was 900 m⁻¹ for non-doped SON68 glass and 1800 m⁻¹ for ²⁹Si-doped SON68 glass, respectively. Outlet solutions were sampled after 1, 3, 7, 14, 28, 56, 100, 163, 202, 257, 362, 415, 482 and 653 days at room temperature, and collected in vials filled with nitrogen gas. The pH was measured after removing of the vial for analyses. Samples were passed through a Whatman inorganic membrane filter with pore size about 0.2 μm, and acidified with ultra-high-purity concentrated HNO₃ and then stored in a cool chamber before analyses. Table 3 summarizes all the experiments performed in this work with the parameters used. At the end of experiment, glass samples were taken from the reactor, carefully rinsed with ultrapure water, and then dried in laboratory atmosphere before characterization.

Fig. 2 Schematic diagram of the experimental setup used in this work.

Table 3 Parameters used for the different leaching experiments performed in this work

Parameter	Leaching experiments
	Non-doped SON68 + non doped COx water	^29Si-doped SON68 + non-doped COx water	Non-doped SON68 + ^29Si-doped COx water
Temperature	35, 50 and 90 °C	90 °C	90 °C
Flow rate	4 ± 1 mL per day	4 ± 1 mL per day	4 ± 1 mL per day
Mass of the glass	0.5773 g	1.0598 g	0.5575 g
Surface	0.027134 m^2	(0.0515 + 0.00022) m^2	(0.0262 + 0.00022) m^2
Volume	30 mL	29 mL	29 mL
S/V	∼900 m^−1	∼1800 m^−1	∼900 m^−1
Experiment period	653 days	485 days	456 days

---

2.4. Analytical methods

Leaching solution analyses were performed at SUBATECH laboratory using a PQ-Excel VG-Elemental ICP-MS for Si, B, Li, Cs and Mo on acidified samples with analytical uncertainty between 4 and 9% in most cases. ²⁹Si/²⁸Si ratios were measured using HR-ICP-MS with a relative uncertainty less than 1.5%.

The morphological and chemical analyses of corroded monoliths and powders were achieved using a JSM 5800 LV scanning electron microscope (SEM) coupled with energy-dispersive X-ray spectroscopy (EDX). The samples were coated with a thin carbon layer in order to get a better resolution of the K-lines for lighter glass components (Si, Al and Ca). Analyses were done under an accelerating voltage of 15 kV. The final compositions of analysed areas were calculated assuming oxide stoichiometry and normalisation to 100%. Altered samples were also analysed by X-ray diffractometry (XRD). For all samples, XRD patterns were characteristic of amorphous materials with the absence of diffraction peaks.

Corrosion products were identified by micro-Raman spectroscopy. Measurements were performed at room temperature using a T64000 Jobin-Yvon/LABRAM spectrometer equipped with a 600 lines mm⁻¹ diffraction grating. The instrument is equipped with an Olympus microscope (×100 objectives) and Ar⁺–Kr⁺ laser (514 nm exciting line). The power of illumination was about 5 mW at the sample in order to prevent any deterioration of the material. Single spectra were obtained in the 100–2200 cm⁻¹ range with an integration time of 600 s.

TOF-SIMS technique was used to obtain isotopic Si ratios and elemental depth profiles in the corrosion layers. Measurements were performed at the University of Lyon using an IONTOF GmbH TOF SIMS instrument. Profiles were achieved by alternating abrasion and analyses cycles. Abrasions were run with a 2 keV primary O₂⁺ ion beam on an area measuring 300 × 300 μm². Analyses were carried out using a focused Bi⁺ primary beam at 25 keV on an area measuring 100 × 100 μm².

2.5. Dissolution rate determination

Based on the concentration of mobile elements in the solution, the normalized concentration NC_i in g m⁻³ was determined using the formula:where C_i is the concentration of the component i in the effluent in g m⁻³ and X_i is the mass fraction of component i in the glass.

Normalized dissolution rate NLR_i+1 was calculated in units of grams of glass dissolved per square metre per day (g m⁻² d⁻¹) from the following relation:

where t is the alteration time in day, i and i + 1 are two consecutive sampling intervals, F is the alteration solution flow in m³ d⁻¹, V is the solution volume in m³ and (S/V) is the glass surface area to solution volume ratio in m⁻¹. Decreasing the surface area of the glass during the experiment was calculated according to the method described by Neeway et al.³ In all cases the variation was negligible.

Experimental uncertainty of the glass dissolution rate was determined by calculating the standard deviation of the function f(x1, x2,⋯, xn) from the following equation:

where δ_f is the standard deviation of the function f, x_i is the parameter i and δ_i is the standard deviation of the parameter i. The substitution of eqn (2) into (3) gives:

Resulting errors of dissolution rates were near 17% and were dependent essentially on the errors associated with the tracer concentrations in the solution.

3. Results

3.1. Corrosion kinetics

3.1.1. Evolution of pH and element concentrations. For all experiments the pH of the input solution was 8 ± 0.05 at room temperature. It has been shown that the pH of the leachates was slightly changed over the alteration time at 35 and 50 °C, with values of 8.11 and 8.01, respectively (Fig. 3(a)). The pH of the leachates at 90 °C remains high during the first two months and then slightly evolved towards that of the inlet solution. Then its value decreases significantly and stabilizes near 7.7 until the end of the experiment.

Fig. 3 (a) Evolution of pH in the leachates with time at 35, 50 and 90 °C. The pH measurements were performed at room temperature. (b) Evolution of element concentrations as a function of alteration time for the leaching experiment performed 90 °C. Errors in concentration values are between 8 and 11%. Si concentrations represent the fraction of Si released from the glass (measured total concentration minus input concentration).

The concentrations of the leached glass elements in the outlet solution at 90 °C are plotted versus the alteration time in Fig. 3(b), and the same trends are observed at 35 and 50 °C. During the first 14 days, concentrations increase rapidly reaching maximum values, followed by a decrease before reaching a steady state. This phenomenon may be due to the transient reactor operation. The behaviours of B, Li, Cs and Mo are similar. These elements are released from the glass until 653 days. The increase in the concentration of B towards 260 days is an experimental artefact likely due to solution filtration (filter failure) or apparatus malfunctioning. The concentration of Si represents only a calculated concentration value corresponding to the release by the glass, calculated by subtracting Si in the input solution from Si in the leachate. It must be kept in mind that: (1) a fraction of the leached Si is retained in both alteration gel and secondary precipitates, and (2) a part of Si from the input solution can also form secondary phases on the glass surface. This calculation leads to negative concentrations after 163 days, suggesting that some Si is consumed during the reaction. The ²⁹Si isotopic tracing allows us to accurately perform mass balance calculation and specifically determine the fraction of Si released by the glass (see below).

3.1.2. Evolution of normalized leaching rates. The measured normalized dissolution rates of the glass elements at different temperatures as a function of corrosion time are presented in double logarithmic graphs in Fig. 4(a)–(d). The measured values obtained at 653 days can be seen in Table 4. The release rates of B, Li and Cs are similar indicating that they are not incorporated in secondary phases and are released nearly congruently. However, Mo seems to have the same behaviour as the other tracers until 60, 200 and 260 days at 90, 50 and 35 °C, respectively, and then the rate of Mo drops significantly compared to other elements. As observed in the case of Si this change is affected by the temperature. This could be explained by the formation of secondary phases containing Mo such as calcium molybdate (CaMoO₄) as previously reported by Neeway et al.³ who studied the SON68 glass in super-saturation conditions (Si = 125 ppm) at 90 °C and pH 9.5.

Fig. 4 Evolution of elements dissolution rates (NLR, g m⁻² d⁻¹) as a function of corrosion time at 35 °C (a), 50 °C (b) and 90 °C (c and d). Errors in NLR values are near 17%.

Table 4 Normalized release rates of glass elements in g m⁻² d⁻¹ measured at 653 days

Element	NLR (g m^−2 d^−1)
	35 °C	50 °C	90 °C
B	2.6 (±0.2) × 10^−5	9.9 (±0.8) × 10^−5	2.4 (±0.2) × 10^−3
Li	2.3 (±0.3) × 10^−5	6.6 (±0.9) × 10^−5	1.7 (±0.3) × 10^−3
Cs	1.9 (±0.3) × 10^−5	4.6 (±0.6) × 10^−5	1.3 (±0.2) × 10^−3
Mo	8.7 (±1.4) × 10^−6	1.9 (±0.2) × 10^−5	4.7 (±0.7) × 10^−4

---

The normalized leaching rate of Si is lower than that of B. This is much more remarkable at 90 °C where NLR B is about 1.5 orders of magnitude higher than NLR Si. This may be explained by the incorporation of silica in secondary phases at this temperature. Other sparingly soluble glass constituents are probably as well incorporated in this phase.

As shown in Fig. 4(a) and (b), the release of tracers at 35 and 50 °C seems to follow a −0.8 gradient. This value represents neither a diffusion-controlled process (−0.5 gradient) or stop of reaction and dilution (−1 gradient),¹ and a detailed interpretation is not available. For the experiment at 90 °C, the first 30 days of glass alteration show rather a −0.5 gradient (Fig. 4(c)), which indicates a diffusion process. Whereas for longer durations, a gradient of −1 suggests strong slow down of reaction rates or even a stop of reaction (the gradient of −1 is an indicator of a simple dilution process of initially leached glass constituents in the continuously flowing water, Fig. 4(d)). The results in Table 4 show that after 653 days of alteration, the residual rate of SON68 glass at 90 °C is reached (2.4 × 10⁻³ g m⁻² d⁻¹). This value is about 10 times higher than that obtained for glass dissolution in pure water in static mode.^36,37 It should be noted that the rate of the glass altered at 35 °C is about two orders of magnitude smaller than the glass altered at 90 °C.

Fig. 5 shows the normalized dissolution rate for boron as a function of time at 90 °C obtained in this work and compared to previous results realised by Neeway.⁵⁸ The experiments by Neeway⁵⁸ were run under the same conditions as this work (pH, temperature and flow rate) by varying the solution composition (pure water), the silica concentration in inlet solutions (25 and 75 ppm) and the glass powder sizes (Ø = l μm). These experiments were stopped after 170 and 300 days for solutions with initial silica concentration of 75 and 25 ppm, respectively. This is due to the complete dissolution of the glass because the use of very fine powder. In our work the experiments were conducted up to 650 days. Also the initial glass dissolution rate obtained in this work is higher than that obtained by Neeway⁵⁸ even for experiments using lower initial silica concentration (25 ppm). This is due to the presence of Mg in COx water used in this work, which leads to silica consumption via Mg-silicates formation and thus increases glass dissolution rate. Also, the high S/V used in Neeway⁵⁸ work is expected to lead to a rapid saturation with regard to silica for instance and also a lower dissolution rate.

Fig. 5 Normalized leaching rates of boron (NLR, g m⁻² d⁻¹) versus time at 90 °C, pH 8, and by using silica saturated COx water, results obtained in this work are compared to those obtained by Neeway⁵⁸ under the same conditions by using silica saturated pure water.

3.1.3. Temperature dependence of glass alteration rate. The normalized leaching rates calculated from the release of glass constituents at different temperatures are given in Fig. 6(a)–(d). NLR value at 90 °C is around 1.5 and two orders of magnitude higher than the NLR values obtained at 50 and 35 °C, respectively. Pierce et al.⁵⁹ have also demonstrated that the dissolution rate based on boron release increases by two order of magnitude with increasing the temperature from 23 to 90 °C. By comparing the two experiments at 35 and 50 °C, we can see that there is generally no significant difference in the leaching rates calculated from B, Li and Cs release until nearly 60 days. The two curves can be distinguished starting from 100 days. This can be explained by the low diffusion of water through the glass matrix at low temperatures. However, in the case of Mo, the two curves at 35 and 50 °C seem to separate from 480 days (Fig. 6(d)). This behaviour can be related to the precipitation of this element in secondary phases.

Fig. 6 Normalized dissolution rates of B (a), Li (b), Cs (c) and Mo (d) (NLR, g m⁻² d⁻¹) as a function of time at different temperatures.

Fig. 7(a) shows a plot of the logarithm of release rates for B, Li, Cs and Mo versus the inverse of the absolute temperature. The rates data are the normalized release rates calculated at 482 days leaching. The apparent activation energy (E_a, J mol⁻¹) may be obtained from these plots using the Arrhenius equation:

where k is the rate constant (in this case the rate of glass dissolution NLR, g m⁻² d⁻¹), A is the Arrhenius parameter (g m⁻² d⁻¹), R is the gas constant (8.314 J K⁻¹ mol⁻¹) and T is the temperature in Kelvin (K). Applying the integrated form of eqn (5) the gradient of these plots has the value (–E_a/R). Activation energy was calculated for 3, 257 and 482 days leaching. An example of B release may be seen in Fig. 7(b). The same is observed for all tracers (not shown here). Numerical results are shown in Table 5.

Fig. 7 (a) The normalized release rates for the glass tracers calculated on 482 days leaching versus the inverse temperature in K. (b) The normalized leach rates calculated from boron release at 3, 257 and 482 days versus the inverse temperature in K.

Table 5 Apparent activation energies (kJ mol⁻¹) calculated after 3 and 482 days of alteration based on glass tracers (B, Li, Cs and Mo) release at temperatures ranging from 35 to 90 °C

Elements	Activation energy (kJ mol^−1)
	Time (days)
	3	482
B	75 (±10)	70 (±11)
Li	73 (±11)	70 (±12)
Cs	71 (±10)	69 (±12)
Mo	71 (±12)	69 (±13)

---

The E_a values calculated in the current work are in good agreement with results published in the literature. Ferrand et al.¹ calculated E_a from the water diffusion coefficients in the SON68 glass powder (Ø = 20 μm). Experiments were conducted in dynamic systems at 50 and 90 °C with silica rich solution at pH 4.8, 7.2 and 9.8, giving E_a values of 49, 52 and 85 kJ mol⁻¹, respectively. Many other works^11,12,54 conducted with different glass compositions, including basaltic glasses, give similar activation energy values (72–77 kJ mol⁻¹).

3.1.4. Surface analyses by SEM/EDX and micro-Raman. SEM images of the glass grains leached at 35, 50 and 90 °C after 653 days of alteration are shown in Fig. 8(a)–(c). As the temperature increases, the size and quantity of the precipitates on the glass surface increases. In all cases, in areas not covered by surface precipitates, a clay-like morphology can be observed. X-Ray diffraction performed on glass samples altered at varying temperatures reveals an important amount of amorphous materials with the absence of diffraction peaks, suggesting that there are not enough crystalline products to detect. EDX spectra of phyllosilicates formed at 90 °C reveal the presence of elements originating from the glass (Al, Mo) from the solution (Mg) and/or from both of them (Si, Ca, Na). Several types of precipitates have been formed. A large amount of magnesium silicates with chemical composition similar to Mg₃Si₄O₁₀(OH)₂ has been identified at 90 °C (Fig. 8(d)), and to a lesser extent at 50 and 35 °C. The Raman vibration modes of the surface layer covering the glass (clay-like) can be attributed to phyllosilicate type magnesium silicate (talc)⁶⁰ (Fig. 8(e)). The Raman analysis is in good agreement with the phase determined from EDX analysis.

Fig. 8 SEM micrographs of glass powders altered at 35 (a), 50 (b) and 90 °C (c) during 653 days. SEM micrograph of the SON68 glass altered for 653 days at 90 °C with the corresponding EDX (d) and Raman spectra compatible with hydrated magnesium silicate (talc)⁶⁰ (e).

At 90 °C, the precipitation of Mg-silicate was important enough to drive the pH to a lower value than at lower temperatures. According to Debure et al.⁶¹ the reactions of Mg-silicate formation can be written as:

(≡Si–O–Si(OH)₃) + OH⁻ ↔ (≡Si–O⁻) + H₄SiO_4(aq) (a)

3Mg²⁺ + 4H₄SiO_4(aq) + 2H₂O ↔ Mg₃Si₄O₁₀(OH)₂ + 6H₃O⁺ (b)

According to this reaction, the release of magnesium into the pore water by Mg bearing minerals of the clay stone could have a negative effect on the long term behaviour of the glass through consumption of silica in solution, thus sustaining glass dissolution. Conversely, the long-term behaviour may be somewhat stabilized by the release of protons in this reaction, lowering the pH. Our results suggest that the corresponding pH is around 7.7 at 90 °C (Fig. 3(a)).

Jollivet et al.⁶² have also demonstrated, by modelling of the SON68 glass alteration in COx ground water with GRAAL model,² that the precipitation of secondary Mg phyllosilicates lowers the pH of the leachate and enhances the dissolution of the passivating layer, increasing consequently the diffusivity of this layer. In static conditions, when all the magnesium in solution has precipitated, the simulations showed that the pH returns slowly to the values usually measured for the alteration of glass in pure water.

Fig. 9(a)–(c) shows the evolution of Si and Mg concentration as a function of time at different temperatures. We can notice that the initial measured concentration of Mg, and to a lesser extent of Si, are lower than the nominal concentrations indicated in Table 2. This can be explained by the partial precipitation of these elements during water storage. The first instants of glass dissolution are characterized by a significant increase in the Si concentration combined with a decrease in Mg concentration. Thereby, the concentration of Mg reaches a minimum value for a maximum Si concentration after 14 days of alteration. Gradually, the glass releases less Si and Mg concentration increases to reach a steady state. In the first step of alteration where the glass releases much Si, the pH increases and Mg reacts to form magnesium silicates. For the experiment carried out at 90 °C, the concentrations of Si and Mg in the input solution are about 38.5 and 55.9 ppm, respectively. At steady state, their concentrations are about 30.1 and 50.3 ppm, respectively (Fig. 9(c)). The difference in concentration of each element between the input solution and the leachate represents the precipitated concentration of element in the reactor. Therefore, the Si/Mg mass ratio of precipitated elements is equal to 1.37, which is not so far from that of talc stoichiometry (1.53). The difference could be attributed to Si consumption by the gel. For experiments carried out at 35 and 50 °C the Si/Mg mass ratios were 1.44 and 1.22, respectively.

Fig. 9 Evolution with time of Si and Mg concentrations for leaching experiments performed at 35 °C (a), 50 °C (b) and 90 °C (c). Errors are near 10% for both elements. The dashed lines correspond to the measured concentration of elements in inlet solutions (green Mg, red Si). Evolution of Si concentration in the leachates as a function of time at 35, 50 and 90 °C (d).

The evolution of Si concentration as a function of time at different temperatures is presented in Fig. 9(d). The decrease of Si concentration is high at 90 °C, this is probably due to the fact that Mg silicate precipitation is favoured at high temperature. Thus the pH is controlled by both the formation this phase and the release of the glass alkali elements in solution. Indeed, during the first two months of alteration at 90 °C the pH remains relatively high (Fig. 3(a)) despite a precipitation of a large amount of Mg silicates which coincides with a sharp decrease in Mg concentration (Fig. 9(c)). During this period the glass releases the maximum of its components including the Si ([Si] > 42 ppm). A competition seems to occur between the two processes. The precipitation of Mg silicates becomes dominant and controls the pH of the solution when the release of the glass alkali elements drops. It should nevertheless be emphasized that the pH values in the leachate do not reflect the local pH values of water in the altered layer pores that can be higher.⁶³

SEM/EDX analyses revealed the formation of a rich Mo and Ba phase on the glass sample altered at 90 °C, suggesting the precipitation of powellite. The presence of this phase is confirmed with micro-Raman analyses (Fig. 10). The retention of Mo in alteration products was also observed by Neeway et al.³ It is known that (Ba,Ca)MoO₄ powellite phase is the primary Mo containing crystalline phase that forms in the alteration of nuclear waste glass.⁶⁴ The powellite mineral is capable of incorporating a wide range of trivalent actinides (Am, Cm and Pu) and lanthanides (La, Nd).^10,65,66 Powellite phase was observed only for experiments conducted at 90 °C. The drop in Mo release rate after 200 and 260 days at 50 and 35 °C, respectively, indicates that a significant fraction of Mo is retained in the alteration products, although the Mo bearing phase has not been identified. Two other Ca-containing phases were observed at 90 °C by SEM/EDX analyses. Micro-Raman analyses confirm the precipitation of apatite and calcite. These minerals can also incorporate trivalent actinides and lanthanides.^10,65 Only calcite was observed at 35 and 50 °C.

Fig. 10 Micro-Raman spectra obtained on the surface of the SON68 glass altered during 653 days at 90 °C, showing the various Raman bonds of calcite, apatite and powellite.^67–70

3.2. Isotopic studies

For the two additional experiments performed at 90 °C and pH 8 in different combinations: (²⁹Si doped SON68 + non-doped COx water) and (non-doped SON68 + ²⁹Si enriched COx water) the release of glass tracers (B, Na,…) is similar as in the experiments without isotopic tracing. Fig. 11(a) shows the evolution with time of the ²⁹Si/²⁸Si ratio in outlet solutions for the (²⁹Si doped SON68 + non-doped COx water) experiment. The red horizontal line corresponds to the initial ²⁹Si/²⁸Si isotopic ratio of the alteration solution. The experimental value measured by HR-ICP-MS confirms the natural isotopic ratio (0.0508). The curve is similar in shape to those seen in the release of the glass tracer elements. After 458 days of corrosion the ratio becomes equal to that in inlet solution indicating that at this time, no significant amount of Si is released from the glass to influence the Si isotopic ratio in solution.

Fig. 11 (a) Evolution with time of the ²⁹Si/²⁸Si ratio in outlet solutions for the (²⁹Si doped SON68 + non-doped COx water) experiment. The red horizontal line corresponds to the initial ²⁹Si/²⁸Si isotopic ratio of the alteration solution. HR-ICP-MS measurements give a value of 0.0508 which represents the natural isotopic ratio. Errors stay under 2%. (b) Evolution with time of the total Si released for the same experiment compared to the similar experiment with non-doped glass at 90 °C. Errors are about 10%.

The total concentrations of silicon released from the glass (Si_released) were calculated from the elemental analyses of Si and isotopic analyses of ²⁹Si/²⁸Si ratios. The ³⁰Si/²⁸Si ratios were not measured and assumed to be constant 0.0335 (natural isotopic ratio), because the low enrichment of ²⁹Si-doped SON68 glass in ³⁰Si compared to ²⁹Si.

The total concentration of silicon (Si_total) = (²⁸Si + ²⁹Si + ³⁰Si).

²⁸Si = Si_total/[1 + (³⁰Si/²⁸Si) + (²⁹Si/²⁸Si)]

²⁸Si_inlet = Si_total,inlet/[1 + 0.0335 + (²⁹Si/²⁸Si)_inlet]

where ²⁸Si_inlet is the concentration of ²⁸Si in the inlet solutions; Si_total,inlet is the total concentration of silicon in the inlet solutions and (²⁹Si/²⁸Si)_inlet the isotopic ratio in the inlet solutions.

²⁹Si_inlet = [(²⁸Si_inlet) × (²⁹Si/²⁸Si)_inlet].

where ²⁹Si_inlet is the concentration of ²⁹Si in the inlet solutions.

The concentration of ²⁹Si in the outlet solutions (²⁹Si_outlet) is obtained using the same method.

²⁹Si_outlet = [(²⁸Si_outlet) × (²⁹Si/²⁸Si)_outlet], the (²⁹Si/²⁸Si)_outlet ratio is presented by the points on the curve in Fig. 11(a).

Therefore, the concentration of ²⁹Si_released (released from the glass) is equal to (²⁹Si_outlet − ²⁹S_inlet) by assuming that ²⁹Si and ²⁸Si are released together with the initial isotopic ratio in the glass (²⁹Si/²⁸Si = 0.16).

The total silicon released from the glass Si_{total,released} = [(²⁹Si_released × 100)/14]. The weight percentage of ²⁹Si in total silicon of the glass is 14.

The evolution with time of the total Si released (Si_{total,released}) for the (²⁹Si doped SON68 + non-doped COx water) experiment is compared to that from the experiment with non-doped glass (SON68 + COx, 90 °C) (Fig. 11(b)). Both curves follow the same shape. The isotopic tracing of the glass allows us to monitor the evolution of Si released from the glass until the end of the experiment. For the experiment (non-doped SON68 + ²⁹Si enriched COx water) the evolution of the Si concentration with time could not be monitored because the use of a very rich ²⁹Si solution (99.9%). In fact, the results of measurement of the isotopic ²⁹Si/²⁸Si ratios give values equal to the background.

Fig. 12 shows the normalized leaching rates of glass elements for: (a) ²⁹Si doped SON68 + non-doped COx water experiment, and (b) non-doped SON68 + ²⁹Si enriched COx water experiments. The curves follow the same trend as that in Fig. 4(d). The dissolution rates tend towards a value of about 10⁻³ g m⁻² d⁻¹. Fig. 12(a) also shows the evolution of Si dissolution rate based on isotopic ²⁹Si/²⁸Si ratio calculations described above. Compared to the other glass tracers, the Si is strongly retained in the alteration layer. Fig. 12(c) shows the evolution with time of silicon retention factor f_Si (expressed in %) in the gel. The value of f_Si is calculated from the normalized mass loss of silicon NLSi (g m⁻²) and boron NLB (g m⁻²): f_Si = (1 − (NLSi/NLB)) × 100. The silicon retention factor remains high during the experiment, it is around 96% at the beginning of the experiment and then decreases to 94% after 14 days (transient reactor operation) and after it increases to 100% after 458 days. The very high f_Si value is due to the use of silica saturated solutions and particularly higher S/V ratio (1800 m⁻¹). Indeed, S/V ratio is the determining parameter in the increase of silica retention factor.⁷¹ Even though low silicon diffusion coefficients are associated with high silicon retention factors, the authors showed that a variation in the S/V ratio between 1 and 2000 cm⁻¹ results in a decrease of the apparent silicon diffusion coefficient by eight orders of magnitude. The very high f_Si value confirms the very low silicon diffusion coefficient (see Section 3.2.2.) and indicates either that the gel formed is dense and protective or that the dissolution affinity, due to high silica concentrations is very low. The normalized dissolution rates of boron for different leaching experiments carried out at 90 °C, pH 8 and [Si]_init = 42 ppm can be seen in Fig. 12(d). Whatever the combination, a decrease in the alteration rate is observed to achieve a value of about 10⁻³ g m⁻² d⁻¹ after 653 days of alteration. The slight difference in the dissolution rate between the non-doped and ²⁹Si-doped SON68 glass is likely due to the difference in S/V ratio. The silicon retention in the gel is known to play a major role in the porosity decrease.^72,73

Fig. 12 Evolution of element dissolution rates (NLR, g m⁻² d⁻¹) as a function of corrosion time at 90 °C for the (²⁹Si doped SON68 + non-doped COx water) experiment (a) and the (non-doped SON68 + ²⁹Si enriched COx water) experiment (b). Evolution of Si retention factor f_Si (%) as a function of corrosion time at 90 °C for the (²⁹Si doped SON68 + non-doped COx water) experiment (c). Evolution of the boron dissolution rates (NLR, g m⁻² d⁻¹) versus time at 90 °C in various combinations (d). pH 8 and [Si]_init = 42 ppm. Errors in NLR values stay under 17%.

3.2.1. Surface analyses by SEM/EDX and micro-Raman. Analyses of glass powders in the two additional experiments using SEM/EDX and micro-Raman spectroscopy reveal the presence of the same phases observed in the leaching experiment of the non-doped SON68 glass in non-doped COx water at 90 °C. A profile of altered glass monolith and the distribution of selected elements by EDX topography for the (²⁹Si doped SON68 + non-doped COx water) experiment after 485 days of alteration can be seen in Fig. 13. The alteration layer thickness is about 1 μm, this value is close to the depletion depth calculated from the mass loss of boron for the same period (1.22 μm).

Fig. 13 Profile of the layer formed after the alteration of the ²⁹Si doped SON68 glass in non-doped COx water for 485 days at 90 °C, pH 8 and [Si]_init = 42 ppm. The EDS topography shows the distribution of selected elements.

In agreement with what is described above, the EDX mapping shows that the Mg (initially absent in the glass) is almost completely absent in the gel, though it is highly concentrated in the phyllosilicates. This suggests a super-saturation of solution towards the magnesium silicates which are massively precipitated on the glass surface, impeding the migration of Si through the gel. This phenomenon suggests that the precipitation of phyllosilicates is not kinetically limited but it is rather limited by Si-availability as shown by Jollivet et al.⁶² and Fleury et al.⁷⁴

The Ca is highly enriched in the gel and in spherical structures in the glass/phyllosilicates interface. This enrichment occurs simultaneously with enrichment in phosphate, which suggests the formation of a calcium phosphate phase (apatite). The apatite mineral may be beneficial to storage because of its capacity to retain the actinides present in the glass. Valle et al.⁴⁹ have also observed by TOF-SIMS analysis a high concentration of Ca at the glass/phyllosilicates interface after 6 months of SON68 glass alteration, as previously observed by Gin et al.⁷² who indicated the presence of a thin alteration layer rich in calcium and phosphates.

Phosphates have the tendency to incorporate the rare earth elements REE,⁷⁵ this may explain the presence of La in the spherical structures. It has been shown that the protective properties of the amorphous gel are directly linked to the distribution of REE and Ca within the alteration layer,⁷² the authors have shown that the precipitation of rare-earth phosphates and Ca tend to maintain alteration rates substantially higher than when the same elements are uniformly distributed in the gel.

There is an apparent enrichment of Al in the gel layer in comparison to the pristine glass and a high concentration in the phyllosilicates. This trivalent element conserves its tetrahedral coordination during the alteration⁷⁶ which requires the presence of charge compensators such us Na, Cs and Ca in order to maintain the charge balance.⁷⁷ In this case, it is essentially the Na which provides this role because it is slightly present in the gel and highly concentrated in the phyllosilicates where Al is also concentrated. Caurel⁷⁸ has shown that under extreme conditions of R7T7 glass alteration (pH > 10.5 and T ≥ 150 °C), Al precipitates as zeolite phases. The precipitation of this phase tends to consume Si present within the gel, the latter is thus poorly connected and loses its protective properties, leading to the resumption of alteration.⁷⁹ This phenomenon has not been observed in laboratory experiments under moderated conditions of glass alteration (pH < 10 and T < 90 °C),³⁸ as is the case in our study. Fe shows a slight enrichment in the gel compared to the glass and a strong presence in the phyllosilicates, which explains its very low concentration (sometimes below the detection limit) in the leachate. Fe may be present in the phyllosilicates as iron hydroxide form.

3.2.2. Analysis of the alteration layer by TOF-SIMS technique and determination of the silicon apparent diffusion coefficient. Fig. 14(a) shows the TOF-SIMS profiles of boron and ²⁸Si/²⁹Si isotopic ratio for ²⁹Si doped SON68 glass monolith altered in non-doped COx water after 485 days, and Fig. 14(b) shows the TOF-SIMS profiles of boron and ²⁹Si/²⁸Si isotopic ratio for non-doped SON68 glass monolith altered in ²⁹Si enriched COx water after 456 days. In the present work, profilometer analyses have not been made, the crater depth caused by sputtering is determined by combination of the profile of boron in the alteration layer and the abrasion rate (1.33 nm s⁻¹). This rate was assumed to be the same as that obtained for the case of the pure silica sputtering (which not necessarily the case). The alteration layer thickness observed by SEM microscopy is in good agreement with this method. Fig. 14(a) and (b) show an alteration depth of 1021 and 999 nm for the two corresponding monoliths. The concentration of B is normalized to its initial concentration in the pristine glass. The depth profile of B shown in Fig. 14(a) allows distinguishing four zones: the first zone corresponds to a pristine glass where the boron concentration is maximum and ²⁸Si/²⁹Si value is minimum (6.13), this value is very close to the theoretical value (6.14). The second zone is referred to as the interfacial gradient and is located between 735 and 1021 nm. The C/C₀(B) value falls about 10 times and the ²⁸Si/²⁹Si ratio increases rapidly to reach a value of 8.3, indicating that the gel is formed by a dissolution/precipitation mechanism incorporating ²⁸Si from the solution. The third zone, between 124 and 735 nm, represents the alteration gel in which C/C₀(B) and ²⁸Si/²⁹Si values remain almost constant with a depletion towards the boron and enrichment in Si. The C/C₀(B) value in the gel is stabilized around 0.09. The fourth zone, located between 0 and 124 nm, corresponds to the precipitation of secondary phases zone, it represents the phyllosilicates. Boron is not taken up by the phyllosilicate while ²⁸Si/²⁹Si ratio increases rapidly to reach a maximum value of 11.06. This value is very close to the maximum value measured in the leachate after 14 days of alteration (²⁸Si/²⁹Si_{max leachate} = 10.96), but it remains lower than that measured in the leachate when the sample is removed from the reactor (after 485 days) which is also equal to that in the leaching solution (²⁸Si/²⁹Si_{leaching solution} = 19.86). Thus, phyllosilicates do not keep the isotopic signature of the alteration solution. Their signature is intermediate between the pristine glass (²⁸Si/²⁹Si = 6.14) and the alteration solution (²⁸Si/²⁹Si = 19.86). This result shows that magnesium phyllosilicates require at least a fraction of silica from the glass to precipitate. This clearly suggests that the precipitation of phyllosilicates sustains glass alteration.

Fig. 14 TOF-SIMS profiles of: boron and ²⁸Si/²⁹Si isotopic ratio for ²⁹Si doped SON68 glass monolith altered in non-doped COx water after 485 days (a); boron and ²⁹Si/²⁸Si isotopic ratio for non-doped SON68 glass monolith altered in ²⁹Si enriched COx water after 456 days (b); remainder of elements for ²⁹Si doped SON68 glass monolith altered in non-doped COx water after 485 days (c–e). Modelling of silica diffusion in the interfacial gradient for the alteration experiment of ²⁹Si doped SON68 glass monolith altered in non-doped COx water after 485 days. (f) Experimental and fitting results of the TOF-SIMS profile of ²⁹Si/²⁸Si isotopic ratio

In the work of Valle et al.⁴⁹ the isotopic signature of phyllosilicates found by SIMS method corresponds to that of the solution when the glass monolith is removed from the reactor. The authors suggest that the phyllosilicates are formed by dissolution/precipitation mechanisms while the gel is formed by hydrolysis/condensation reactions because it keeps an intermediate isotopic signature. However, the precipitation of secondary phases in the gel cannot be excluded (i.e. phosphates, molybdates, precursors of magnesium phyllosilicates,…).

It must also be noted the difference between the two works: (1) in our work we have added the glass powder in addition to the monolith, (2) we did not use the same parameters (flow rate = 12 mL per day and low S/V (34 m⁻¹) as in the work of Valle et al.⁴⁹), and (3) we particularly cleaned with ultrapure water the altered glass monoliths before analyzes. In fact, the cleaning performed on altered glass monoliths was able to eliminate or dissolve a portion of the external layer which was in contact with the solution. Therefore, the phyllosilicates keep a different isotopic signature from that of the alteration solution when the monolith is removed from the reactor. This hypothesis is further supported by hydration experiments of SON68 glass monoliths in the presence of vapor D₂¹⁸O.⁸⁰ Unlike this work, the hydrated glass monoliths have not been cleaned with ultrapure water before analyzes. It is seen that the phyllosilicates keep an isotopic signature ¹⁸O/¹⁶O of the water vapor. In any case, all analysis results highlight a mechanism of dissolution/precipitation responsible for the phyllosilicates formation.

In the second experiment (Fig. 14(b)) four different zones are also noticed with the same behaviour of the elements studied. The depth profile of ²⁹Si/²⁸Si shows the penetration of ²⁹Si from the solution into the non-doped SON68 altered glass. The value of ²⁹Si/²⁸Si ratio in the glass measured by TOF-SIMS technique is equal to 0.0509 and it is very close to the theoretical value (natural abundance = 0.05). The ²⁹Si/²⁸Si ratio reaches a maximum value in the phyllosilicates (²⁹Si/²⁸Si_{max phyllosilicates} = 5.2). As in the first experiment, this value is lower than that measured in the leachate when the sample is removed from the reactor (²⁹Si/²⁸Si_{leaching solution} > 40).

It is therefore clear that the phyllosilicates keep an intermediate isotopic signature between the pristine glass and the alteration solution. The fact that the gel keeps a relatively constant isotopic ratio along the profile would lead to the assumption that the condensation phenomena is almost instantaneous and follows the thermodynamic laws of local dissolution/precipitation.

TOF-SIMS profiles of all glass components for ²⁹Si doped SON68 glass monolith altered in non-doped COx water after 485 days can be seen in Fig. 14(c)–(e). The boron profiles show a significant depletion in the entirety of the alteration layer. Li is also depleted in the alteration layer. Cs behaves in the same manner as B and Li in the gel, however a small amount of this element is retained in the phyllosilicates. The retention of Cs is due to its ability to be incorporated into solid phases, in particular the clays.⁸¹ The concentration of Mo drops sharply in the interfacial gradient and linearly in the gel, this behaviour indicates that an amount of Mo is retained in the gel. In agreement with solution analyses and SEM observations, Mo is also detected in the phyllosilicates. Unlike B and Cs, the gel is not completely depleted in Na. The presence of Al in the gel may cause the retention of Na which assumes the role of a charge compensator. Na is largely concentrated in the external layer of altered glass.

The profiles of Al and Fe are identical, they are especially enriched in the phyllosilicates. Ca behaves in the same manner as Fe in the gel, it is very concentrated at the interface between the gel and the phyllosilicates but its concentration is very low in the external zone. The relative enrichment in Ca compared to the pristine glass indicates its high incorporation into the alteration layer and secondary phases. The incorporation of Ca in the gel has been observed by Chave et al.⁸² during the alteration of the French reference nuclear glass in calcium-rich solution, this phenomenon slows the glass corrosion rate. Mg is very concentrated in the phyllosilicates as magnesium silicates. It should be noted that Mg penetrates into the gel from the solution and thus contributes to the destabilization of the alteration layer via its magnesium silicates transformation.^34,61,83

For the alteration experiment of non-doped SON68 glass in ²⁹Si enriched COx water after 456 days, TOF-SIMS profiles show that all elements behave in the same manner.

The diffusion of Si in the gel was determined according to the method described by Abdelouas et al.⁸⁴ Thus the TOF-SIMS profile of ²⁹Si/²⁸Si isotopic ratio in the interfacial gradient was fitted using Origin 8.6 software. The experimental and fitting results are given in Fig. 14(f). The equation is based on Fick’s second law, which predicts how diffusion causes the concentration to change with time. It takes into account de diffusion of ²⁸Si from the alteration solution toward the gel which is considered a semi-infinite medium.

Fick’s second law is given by:

where C is the concentration of the component in the effluent in g m⁻³, x is the distance in meters, D is the diffusion coefficient in m² s⁻¹ and t is the time in seconds.

The solution of this equation is:

This gives the equation:

A is temperature-independent pre-exponential in m² s⁻¹ (the height of the curve’s peak), d is the position of the center of the peak, D is the diffusion coefficient and t is the time of the experiment.

The obtained value of Si apparent diffusion coefficient is about 1.7 × 10⁻²² m² s⁻¹. This value is consistent with the modelled value obtained by Valle⁸⁵ for alteration experiments of SON68 glass in silica concentration solutions at 90 °C and pH 9.2. The modelled value D_Si, based on the Fick’s second law, was seen to range between 3 × 10⁻²⁰ after 1 h and 7 × 10⁻²³ m² s⁻¹ after 20 days. For the same experiments, but under silica unsaturated conditions, the author obtained a value of D_Si between 5 × 10⁻¹⁴ and 4 × 10⁻¹³ m² s⁻¹. The low diffusion of silica through the gels formed under saturation conditions demonstrates the effect of the affinity term.

It should be noted that this method is a simplified approach from the TOF-SIMS measurements. In fact, silicon can be sorbed and/or precipitated in the gel. Therefore, the process of Si transport in the gel layer may be described in the reactive transport way and not the solute transport in the porous media. Thus, from the TOF-SIMS profiles we can provide apparent diffusion coefficients, the very low value of which indicates an important part of chemical reactivity.

4. Conclusions

Leaching experiments of SON68 glass were performed in dynamic mode under silica rich COx water (42 ppm) at pH 8, S/V ratio (900–1800 m⁻¹) and at 35, 50 and 90 °C. The results showed that the pH decreases at 90 °C because of the large precipitation of magnesium silicates. The glass alteration seems to be controlled by both diffusion and surface reaction process. The residual rate measured at 90 °C after 653 days of alteration is about 10⁻³ g m⁻² d⁻¹; it is about 2 and 1.5 orders of magnitude higher than that obtained at 35 and 50 °C, respectively. At 90 °C the dissolution rate is ten times higher in COx groundwater than in pure water. The activation energy for glass dissolution is about 70 kJ mol⁻¹, in good agreement with the results published in the literature. Surface analysis of altered glass revealed the precipitation of magnesium silicates and calcite at 35 and 50 °C. The same phases in addition to powellite and apatite precipitate at 90 °C. The ²⁹Si isotopic tracing allowed us to follow the distribution of Si in the alteration layer. EDX mapping and TOF-SIMS profiles of elements showed that the phyllosilicates are composed of Mg, Si, Al, Ca, Na and Fe. The gel showed depletion in Mg relative to phyllosilicates and enrichment in the rest of elements. Analysis results highlighted a mechanism of dissolution/precipitation responsible for the phyllosilicates formation. The gel is formed by a very local hydrolysis/condensation with precipitation of hydrolyzed species. The diffusion coefficient of silica through the gels is about 1.7 × 10⁻²² m² s⁻¹, and indicates the protective effect of the gel formed under silica saturated conditions. This work clearly shows that Mg plays an important role in the control of SON68 glass dissolution at low temperatures, expected in the French high-level waste repository. Finally, a thorough characterisation of the Mg-rich phyllosilicates, in particular the solubility, is needed for modelling of the long-term glass dissolution.

Acknowledgements

We would like to thank Andra, the French Agency for Radioactive Waste Management, for partial financial support of the work. We would like to thank Karine David of SUBATECH for HR-ICP-MS analyses. We would also like to thank Nicolas Stephant from the Institut des Matériaux de Nantes Jean Rouxel (Nantes, France) for his help with SEM analyses.

References

1. K. Ferrand, A. Abdelouas and B. Grambow, J. Nucl. Mater., 2006, 355, 54–67.
2. P. Frugier, S. Gin, Y. Minet, T. Chave, B. Bonin, N. Godon, J. E. Lartigue, P. Jollivet, A. Ayral, L. De Windt and G. Santarini, J. Nucl. Mater., 2008, 380, 8–21.
3. J. J. Neeway, A. Abdelouas, B. Grambow and S. Schumacher, J. Nucl. Mater., 2011, 415, 31–37.
4. S. Gin, P. Frugier, P. Jollivet, F. Bruguier and E. Curti, Int. J. Appl. Glass Sci., 2013, 4, 371–382.
5. J. D. Vienna, J. V. Ryan, S. Gin and Y. Inagaki, Int. J. Appl. Glass Sci., 2013, 4, 283–294.
6. B. Grambow, in Mater. Res. Soc., ed. C. M. Jantzen, J. A. Stone and R. C. Ewing, 1985, vol. 44, pp. 15–27.
7. B. Grambow and R. Müller, J. Nucl. Mater., 2001, 298, 112–124.
8. W. L. Bourcier, D. W. Peiffer, K. G. Knauss, K. D. McKeegan and D. K. Smith, MRS Online Proc. LIbr., 1989, 176, 209–216.
9. A. Abdelouas, J. L. Crovisier, W. Lutze, B. Fritz, A. Mosser and R. Müller, Clays Clay Miner., 1994, 42, 526–533.
10. A. Abdelouas, J. L. Crovisier, W. Lutze, B. Grambow, J. C. Dran and R. Müller, J. Nucl. Mater., 1997, 240, 100–111.
11. I. Techer, T. Advocat, J. Lancelot and J.-M. Liotard, Chem. Geol., 2001, 176, 235–263.
12. J. L. Crovisier, T. Advocat and J. L. Dussossoy, J. Nucl. Mater., 2003, 321, 91–109.
13. D. Wolff-Boenisch, S. R. Gislason, E. H. Oelkers and C. V. Putnis, Geochim. Cosmochim. Acta, 2004, 68, 4843–4858.
14. B. Parruzot, P. Jollivet, D. Rébiscoul and S. Gin, Geochim. Cosmochim. Acta, 2015, 154, 28–48.
15. J. Sterpenich and G. Libourel, J. Non-Cryst. Solids, 2006, 352, 5446–5451.
16. A. Verney-Carron, S. Gin and G. Libourel, Geochim. Cosmochim. Acta, 2008, 72, 5372–5385.
17. A. Verney-Carron, S. Gin, P. Frugier and G. Libourel, Geochim. Cosmochim. Acta, 2010, 74, 2291–2315.
18. G. Libourel, A. Verney-Carron, A. Morlok, S. Gin, J. Sterpenich, A. Michelin, D. Neff and P. Dillmann, C. R. Geosci., 2011, 343, 237–245.
19. A. Michelin, E. Burger, E. Leroy, E. Foy, D. Neff, K. Benzerara, P. Dillmann and S. Gin, Corros. Sci., 2013, 76, 403–414.
20. P. Dillmann, S. Gin, D. Neff, L. Gentaz and D. Rebiscoul, Geochim. Cosmochim. Acta, 2016, 172, 287–305.
21. F. Angeli, M. Gaillard, P. Jollivet and T. Charpentier, Geochim. Cosmochim. Acta, 2006, 70, 2577–2590.
22. J. J. Neeway, S. N. Kerisit, J. Liu, J. Zhang, Z. Zhu, B. J. Riley and J. V. Ryan, J. Phys. Chem. C, 2016, 120, 9374–9384.
23. M. I. Ojovan, R. J. Hand, N. V. Ojovan and W. E. Lee, J. Nucl. Mater., 2005, 340, 12–24.
24. S. Gin, J. V. Ryan, D. K. Schreiber, J. Neeway and M. Cabié, Chem. Geol., 2013, 349–350, 99–109.
25. S. Gin, P. Jollivet, M. Fournier, F. Angeli, P. Frugier and T. Charpentier, Nat. Commun., 2015, 6, 6360.
26. L. L. Hench and D. E. Clark, J. Non-Cryst. Solids, 1978, 28, 83–105.
27. R. Hellmann, S. Cotte, E. Cadel, S. Malladi, L. S. Karlsson, S. Lorenzo-Perez, M. Cabié and A. Seyeux, Nat. Mater., 2015, 14, 307–311.
28. C. Cailleteau, F. Angeli, F. Devreux, S. Gin, J. Jestin, P. Jollivet and O. Spalla, Nat. Mater., 2008, 7, 978–983.
29. P. Frugier, C. Martin, I. Ribet, T. Advocat and S. Gin, J. Nucl. Mater., 2005, 346, 194–207.
30. Y. Inagaki, A. Ogata, H. Furuya, K. Idemistu, T. Banba and T. Maeda, Mater. Res. Soc. Symp. Proc., 1996, 412, 257–264.
31. T. Advocat, J. L. Crovisier, E. Vernaz, G. Ehret and H. Charpentier, MRS Online Proc. LIbr., 1990, 212, 57–64.
32. P. Jollivet, S. Gin and M. Schumacher, Chem. Geol., 2012, 330–331, 207–217.
33. S. Gin, N. Godon, J. P. Mestre and E. Y. Vernaz, Appl. Geochem., 1994, 9, 255–269.
34. C. Cailleteau, F. Devreux, O. Spalla, F. Angeli and S. Gin, J. Phys. Chem. C, 2011, 115, 5846–5855.
35. S. Gin, I. Ribet and M. Couillard, J. Nucl. Mater., 2001, 298, 1–10.
36. S. Ribet and B. Grambow, Modelling R7T7 glass behavior and corrosion product properties, in SUBATECH UMR 6457 Scientific Report, 2008.
37. S. Gin, X. Beaudoux, F. Angéli, C. Jégou and N. Godon, J. Non-Cryst. Solids, 2012, 358, 2559–2570.
38. S. Ribet and S. Gin, J. Nucl. Mater., 2004, 324, 152–164.
39. M. Fournier, S. Gin and P. Fruhier, J. Nucl. Mater., 2014, 448, 348–363.
40. D. Rebiscoul, A. Van der Lee, F. Rieutord, F. Né, O. Spalla, A. El Mansouri, P. Frugier, A. Ayral and S. Gin, J. Nucl. Mater., 2004, 326, 9–18.
41. G. Geneste, F. Bouyer and S. Gin, J. Non-Cryst. Solids, 2006, 352, 3147–3152.
42. O. Deruelle, PhD thesis, University of Pierre et Marie Curie, 1997.
43. W. Zhou, D. R. Peacor, J. C. Alt, R. Van der Voo and L. S. Kao, Chem. Geol., 2001, 174, 365–376.
44. N. A. Stroncik and H. U. Schmincke, Geochem., Geophys., Geosyst., 2001, 2, 53–95.
45. G. D. Garlick and J. R. Dymond, Geol. Soc. Am. Bull., 1970, 81, 2137–2142.
46. F. Angeli, T. Charpentier, S. Gin and J. C. Petit, Chem. Phys. Lett., 2001, 341, 23–28.
47. A. Ledieu, F. Devreux, P. Barboux, L. Sicard and O. Spalla, J. Non-Cryst. Solids, 2004, 343, 3–12.
48. P. Jollivet, F. Angeli, C. Cailleteau, F. Dereux, P. Frugier and S. Gin, J. Non-Cryst. Solids, 2008, 354, 4952–4958.
49. N. Valle, A. Verney-Carron, J. Sterpenich, G. Libourel, E. Deloule and P. Jollivet, Geochim. Cosmochim. Acta, 2010, 74, 3412–3431.
50. S. Gin, P. Jollivet, M. Fournier, C. Berthon, Z. Wang, A. Mitroshkov, Z. Zhu and J. V. Ryan, Geochim. Cosmochim. Acta, 2015, 151, 68–85.
51. S. Gin, A. Abdelouas, L. J. Criscenti, W. L. Ebert, K. Ferrand, T. Geisler, M. T. Harrison, Y. Inagaki, S. Mitsui, K. T. Mueller, J. C. Marra, C. G. Pantano, E. M. Pierce, J. V. Ryan, J. M. Schofield, C. I. Steefel and J. D. Vienna, Mater. Today, 2013, 16, 243–248.
52. A. Dossier, 2005 Argile – Référentiel de comportement des colis de déchets à haute activité et à vie longue, C.RP.ASCM.04.0017, 2005.
53. C. Poinssot and S. Gin, J. Nucl. Mater., 2012, 420, 182–192.
54. B. P. McGrail, W. L. Ebert, A. J. Bakel and D. K. Peeler, J. Nucl. Mater., 1997, 249, 175–189.
55. M. Fournier, A. Ull, E. Nicoleau, Y. Inagaki, M. Odorico, P. Frugier and S. Gin, J. Nucl. Mater., 2016, 476, 140–154.
56. J. P. Icenhower and C. I. Steefel, J. Nucl. Mater., 2013, 439, 137–147.
57. E. C. Gaucher, C. Lerouge, P. Blanc and C. Tournassat, Caractérisation géochimique des forages PAC et nouvelles modélisations, THERMOAR. BRGM/RP–5446–FR, 2007.
58. J. J. Neeway, PhD thesis, University of Nantes, 2011.
59. E. M. Pierce, E. A. Rodriguez, L. J. Calligan, W. J. Shaw and B. P. McGrail, Appl. Geochem., 2008, 23, 2559–2573.
60. G. J. Rosasco and J. J. Blaha, Appl. Spectrosc., 1980, 34, 140–144.
61. M. Debure, P. Frugier, L. De Windt and S. Gin, J. Nucl. Mater., 2012, 420, 347–361.
62. P. Jollivet, P. Frugier, G. Parisot, J. P. Mestre, E. Brackx, S. Gin and S. Schumacher, J. Nucl. Mater., 2012, 420, 508–518.
63. W. H. Casey and B. Bunker, Mineral-water interface geochemistry, in Leaching of mineral and glass surfaces during dissolution, ed. M. F. Hochelia and A. F. White, Mineralogical Society of America, Washington, 1990, vol. 23, pp. 397–426.
64. N. Henry, P. Deniard, S. Jobic, R. Brec, C. Fillet, F. Bart, A. Grandjean and O. Pinet, J. Non-Cryst. Solids, 2004, 333, 199–205.
65. D. Bosbach, T. Rabung, F. Brandt and T. Fanghänel, Radiochim. Acta, 2004, 92, 639–643.
66. A. Sneyers, B. Grambow, P. Hernan, H. J. Alheid, J. F. Aranyossy and L. Johnson, 7th European Commission Conference on the Management and Disposal of Radioactive Waste, Luxembourg, 2008.
67. G. Penel, G. Leroy, C. Rey and E. Bres, Calcif. Tissue Int., 1998, 63, 475–481.
68. A. Antonakos, E. Liarokapis and T. Leventouri, Biomaterials, 2007, 28, 3043–3054.
69. H. G. M. Edwards, S. E. J. Villar, J. Jehlicka and T. Munshi, Spectrochim. Acta, Part A, 2005, 61, 2273–2280.
70. R. Frech, E. C. Wang and J. B. Bates, Spectrochim. Acta, Part A, 1980, 36, 915–919.
71. I. Ribet, S. Gin, E. Vernaz and R. Do Quang, GLOBAL international conference on “Back end of the fuel cycle : From research to solutions”, Paris, 2001.
72. S. Gin, C. Jégou and E. Vernaz, Appl. Geochem., 2000, 15, 1505–1525.
73. P. Van Iseghem, M. Aertsens, S. Gin, D. Deneele, B. Grambow, P. McGrail, D. Strachan and G. Wicks, A Critical Evaluation of the Dissolution Mechanisms of High-level Waste Glasses in Conditions of Relevance for Geological Disposal (GLAMOR), Contract FIKW-CT-2001–20140, European Commission, 2006.
74. B. Fleury, N. Godon, A. Ayral and S. Gin, J. Nucl. Mater., 2013, 442, 17–28.
75. S. Fillet, PhD thesis, University of Montpellier, 1987.
76. P. C. Burns, R. A. Olson, J. R. Finch, J. M. Hanchar and Y. Thibault, J. Nucl. Mater., 2000, 278, 290–300.
77. F. Angeli, J. M. Delaye, T. Charpentier, J. C. Petit, D. Ghaleb and P. Faucon, J. Non-Cryst. Solids, 2000, 276, 132–144.
78. J. Caurel, PhD thesis, University of Poitier, 1990.
79. M. Fournier, P. Frugier and S. Gin, Procedia Earth Planet. Sci., 2013, 7, 264–267.
80. R. Bouakkaz, PhD thesis, University of Nantes, 2014.
81. G. Montavon, E. Alhajji and B. Grambow, Environ. Sci. Technol., 2006, 40, 4672–4679.
82. T. Chave, P. Frugier, S. Gin and A. Ayral, Geochim. Cosmochim. Acta, 2011, 75, 4125–4139.
83. A. Abdelouas, J.-L. Crovisier, W. Lutze, R. Müller and W. Bernotat, Eur. J. Mineral., 1995, 7, 1101–1113.
84. A. Abdelouas, Y. EL Mendili, A. Ait Chaou, G. Karakurt, C. Hartnack, J. F. Bardeau, T. Saito and H. Matsuzaki, Int. J. Appl. Glass Sci., 2013, 4, 307–316.
85. N. Valle, PhD thesis, University of Lorraine, 2000.

---