Corrosion Behavior and Microstructure Influence of Glass-Ceramic Nuclear Waste Forms

In many countries that utilize atomic energy, such as the United States,¹  France,²  United Kingdom,³  India,⁴  Japan,⁵  South Korea,⁶  Belgium,⁷  and Germany,⁸  borosilicate glass is either being utilized or under study as a waste form for the immobilization of high-level waste (HLW) generated from aqueous reprocessing of used nuclear fuel. The borosilicate glass is a desirable waste form matrix as the glass structure can easily incorporate a wide range of fission products with high durability in expected disposal conditions.^4,9-15  Increasing the waste loading of borosilicate glass can reduce costs, time, and volume associated with waste cleanup. However, the limited solubility of HLW components such as Mo (as seen in testing of borosilicate glasses with high Mo¹⁶ ), the lanthanides (Ln), and noble metals in the glass phase can place limits on waste loading of ∼18 wt%.¹⁷  Above these solubility limits, crystalline phases can form upon cooling from the melt as nondurable weak points of the glass waste form. As such, the formation of such phases is commonly avoided in borosilicate glass development for HLW.

However, by controlling the chemistry and growth of crystalline phases in borosilicate glasses, insoluble species, and radioactive fission products can be incorporated into durable crystalline phases in the waste form. This process creates glass-ceramic waste forms that can achieve higher waste loadings (∼45 wt%), higher heat tolerances, and ease of production using baseline melter technology.¹⁷  In this paper, data related to a glass-ceramic system with targeted phases including powellite ([XMoO₄], where X = alkali, alkaline earth, and/or Ln, and CaMoO₄ are observed in previous high Mo additions to borosilicate glasses¹⁸ ) and oxyapatite ([Y₂Ln₈Si₆O₂₆], where Y = alkali and/or alkaline earth) are presented.¹⁹  Despite the challenges in attaining consistent control over chemistry and microstructure in glass-ceramic fabrication, targeted compositions have been attained over a range of experimental conditions and processed on an engineering scale cold crucible induction melter.¹⁹ 

Generation of phases that have low-chemical durability, or that lead to mechanical failure and separation from the glass phase, would risk the integrity of the waste form. The presence of multiple phases is also a problem for modeling because the behavior of each phase should be known and accelerated corrosion can occur along grain boundaries.²⁰  Understanding the corrosion behavior of the individual phases in the glass-ceramic and the influence of changing microstructure on their dissolution is imperative for further development of this technology. In static product consistency testing (PCT) of a matrix of glass-ceramic compositions (including the 1× sample in this report), low normalized releases of B, Na, Li, Mo, and Si at 90°C were measured with static conditions at time frames between 3 d and 28 d²¹  and the values were close to the trends observed on SON68, a simulated high-level waste glass. Previous work has used a modified single-pass flow-through (MSPFT) test in which a compositional range of glass-ceramic samples (as coupons) were corroded together in a single vessel. Post-mortem surface analyses of these coupon samples showed enhanced corrosion attack in some cases at the crystalline/glass interface, possibly resulting from mechanical stresses induced by differences in the coefficient of thermal expansion (CTE) between the phases. These effects may be enhanced with varying microstructure size of the crystalline phases within the glass-ceramic. No significant chemical differences were identified at the interface of glass and crystalline phases compared with the bulk phases, although preferential corrosion of one phase over another can also result in enhanced corrosion and cannot be ruled out yet with currently available data. It was also suggested from the MSPFT data that the continuous glass phase experienced the highest corrosion rate compared with the crystalline phases. However, because of the multi-sample nature of the MSPFT, no solutions analyses were performed and thus individual corrosion rates from the phases could not be determined.²² 

In this work, corrosion studies of a glass-ceramic, identified as C1 (which is the centroid composition of a statically-designed test matrix for a variety of radioactive waste streams from reprocessing, further described in a previous report²² ) fabricated using three different cooling rates, using single-pass flow-through (SPFT) testing at 90°C and room temperature pH (pH(RT)) 7 and 9 are presented. The standard SPFT method allows for determination of bulk glass-ceramic corrosion rates, as well as qualitative corrosion rates of individual phases, using solution analysis from a single glass-ceramic sample (as both powders and coupons) corroded in an individual reactor. Glass-ceramic monoliths subjected to corrosion by the SPFT system were also characterized with scanning electron microscopy (SEM) and electron dispersive spectroscopy (EDS), x-ray diffraction (XRD), and atomic force microscopy (AFM). This combination of both solutions and solids analysis allows for a more thorough understanding of the processes responsible for the degradation of glass-ceramics, including the pH-dependent durability of the individual phases in the glass-ceramic that are formed to encapsulate various families of radionuclides. Additionally, investigation of the samples formed at varying cooling rates also allows the testing of a range of microstructures and determination of the effect they have on corrosion behavior. Increasing cooling rate (by a 5× factor) has previously shown to impact the release rate in PCT testing of glass-ceramics materials developed for immobilization of reprocessed UMo fuel waste in France.²³  The 5× change in cooling rate led to a 40% decrease in B release (from the glass) and a 58% increase in Mo release (from a powellite phase).²³  The measured corrosion rates will ultimately be used to tailor the waste form to reduce overall corrosion rates. This is important because the canister will see a range of cooling profiles depending on the distance from the outer canister walls. Thus the microstructure will vary radially in a canister. To capture the expected variability of microstructure caused by changes in cooling rate, the expected canister centerline cooling (CCC) profile (1×) and two extremes (fastest [4× the CCC profile] and slow [¼× the CCC profile]) were examined. Overall, the results highlight the ability of glass-ceramics to provide a durable matrix for radioactive waste streams generated from used nuclear fuel reprocessing.

The C1 glass-ceramic has a target composition given in Table 1. The “Others” component listed in Table 1 was composed of (mol fraction) reagent grade 0.017 PdO, 0.065 RhO₂, 0.167 RuO₂, 0.054 Ag₂O, 0.098 CdO, 0.084 SeO₂, 0.052 SnO₂, and 0.463 TeO₂. This elemental set is a subset of waste components.²²  The glass-ceramics were made from a 500-g batch of reagent grade oxides, carbonates, and boric acid, which were homogenized for 4 min in an agate milling chamber. The batch was melted twice in a lidded Pt/10% Rh crucible between 1,250°C and 1,450°C for 1 h and quenched on an Inconel^®† plate in air; between the first and second melts, the glass was ground in a tungsten carbide mill for 4 min to ensure homogenization of all components. The quenched glass was then ground to a powder, loaded into a Pt boat with a lid, and reheated to the melting temperature, 1,300°C, for 30 min, then slowly cooled down to 400°C, based on the CCC temperature profile of a 0.61-m (2-ft) diameter canister.¹⁶  This sample was denoted “1×” slow cooling rate. Further information on the heat treatment can be found elsewhere.¹⁷  Two additional samples were fabricated at cooling rates 4× CCC rate (fast) and ¼× the CCC rate (slow). The phase assemblages of the glass-ceramic samples fabricated at the various cooling rates were characterized by XRD and SEM-EDS.

XRD was performed to determine the crystalline phases and their concentrations in the glass-ceramic specimens cooled at various rates. Specimens were prepared for XRD by grinding them to a very fine powder and doping the powder with a known concentration (mass%) of Standard Reference Material #674b (TiO₂, rutile) to facilitate phase quantification. The XRD patterns were collected with a Bruker D8 Advance^† XRD system (Bruker AXS) equipped with a Cu target (K_α1 = 0.15406 nm) over a scan range of 5° 2θ to 75° 2θ using a step size of 0.015° 2θ and a hold time of 4 s per step. The scans were analyzed with TOPAS^† (v4.2) whole-pattern fitting software according to the fundamental parameters approach.²⁴  Structure patterns were selected from the Inorganic Crystal Structure Database (release 2013) with unit cell dimensions refined in the fitting process of each pattern. The amorphous content of each specimen was determined by difference after crystalline phases were quantified and renormalized based on the concentration of the known internal standard (rutile) in the specimens.

All SEM analyses were performed with a JSM-7001F^† microscope (JEOL USA, Inc.) with a backscattered electron detector. The SEM analyses were performed on the glass-ceramics before testing, and on cross-sectioned post-test coupons in corroded (surface) and uncorroded (bulk) regions to determine corrosion behavior and how it relates to the glass-ceramic microstructure. This was done in an attempt to understand the relationship between the corrosion behaviors of the individual phases in the glass-ceramic. Cross sections were made through samples where a masked spot was also intersected to provide a pristine coupon surface to reference all corrosion measurements. The SEM was also coupled to an XFlash 6|60^† EDS Si-drift detector (SDD; Bruker) that was used to perform elemental mapping and spot analysis.

The SPFT test method allows for measurement of dissolution rates of minerals and glasses under controlled conditions, and described fully in ASTM C1622-10.²⁵  The design of the system, shown in Figure 1, enables a continuous flow of fresh solution into a reaction vessel containing the glass-ceramic. A syringe pump (Norgren Kloehn) is used to pump the buffered solution from the input reservoir to the reactor vessel inside the oven through polytetrafluoroethylene (PTFE) tubing. The buffered solution was 0.05 M tris(hydroxymethyl)aminomethane (Tris; Fisher) adjusted to pH(RT) 7 and 9 through the addition of concentrated HNO₃. The calculated pH of the solutions at the experimental temperature of 90°C are shown in Table 2.²⁶  The PTFE tubing connects the syringe pump to a perfluoroalkoxy (PFA) Teflon^† reactor containing the glass-ceramic powder or coupons. The reactor is a two-piece system consisting of the reactor base threaded together with the lid containing the inlet and outlet ports. The SPFT reactors have a diameter of 47.5 mm and height of 63.6 mm with a total inner volume of ∼60 mL. The glass-ceramic powder is placed in the reactor and forms a thin layer at the bottom of the reactor to interact with the contact solution. The effluent solution then flows out of the reactor to a collection bottle placed outside of the oven. Flow rates are determined gravimetrically at defined sampling intervals. Three blank effluent samples were collected from each reactor prior to the addition of the glass to determine background concentrations of the analytes.

The collected effluent was analyzed with inductively coupled plasma optical emission spectroscopy (ICP-OES) for Al (detection limit [DL] 49.4 μg/L), Ba (DL 9.85 μg/L), B (DL 75.6 μg/L), Ca (DL 101 μg/L), Li (DL 71.9 μg/L), Mo (DL 29.7 μg/L), Si (DL 164 μg/L), Na (DL 134 μg/L), and Sr (DL 18.8 μg/L) and ICP-mass spectrometry (ICP-MS) for Cs (DL 2.55 μg/L), Nd (DL 4.45 μg/L), and La (DL 5.63 μg/L).

Tests were performed on glass-ceramic samples of fabricated at three cooling rates (1/4×, 1×, and 4×). Powdered samples were prepared by crushing the glass-ceramic in a tungsten carbide mill and then sieving to a −100 + 200 (149 μm to 75 μm diameter) size fraction, rinsing with deionized (DI) water, sonicating in DI water, rinsing in ethanol, sonicating in ethanol, and drying in a 90°C oven according to ASTM C1285-14.²⁷  The prepared glass-ceramic powder was placed into the 60 mL reactor after collection of blank samples to determine background concentrations of various analytes in the influent solution.

In the pH(RT) 9 SPFT tests, coupons (∼1 cm × 1 cm × 1 mm) of the C1 glass-ceramic were placed in the reactor on a Teflon^† basket. The coupons were polished on both sides to a 0.25-μm finish. The polished surface was masked with ∼2-mm diameter spots of a room-temperature vulcanization (RTV) silicone (Loctite 595 Silicone Sealant^†). This masking creates a reference height for the uncorroded glass-ceramic surface. This enables the determination of corrosion rates using solution analyses and height changes in the coupons.²⁸  Because of limited material fabricated, only a single reactor could be performed at each q/S value. However, the use of a single reactor does not impede quantification of error associated with the measurement. Three steps are used to ensure validity of the data: (1) testing different q/S ratios at each condition to determine conditions at which forward dissolution rate is present, (2) running tests for sufficient durations to ensure steady-state dissolution is achieved, and (3) implementation of analytical QA procedures to quantify error. The quantification of error is discussed in the Single-Pass Flow-Through Results section. The test specifics for each reactor, including glass-ceramic masses, surface areas, and flow rates, are provided in Table 3.

The silicone masks were removed from the reacted glass-ceramic coupon surface mechanically, followed by rubbing the surface with a cotton swab wetted with isopropanol. This cleaning step was found to be necessary to remove residue at the edges of the grease dots which interfered with subsequent AFM measurements of step height. The AFM images were collected in tapping mode on a Concept Scientific Instruments Nano-Observer AFM^† or a Veeco Dimension Icon AFM^†. The height accuracy of these instruments was confirmed by measuring a standard sample during initial instrument installation. Step edges between corroded and protected areas of the glass coupon surface were identified on a digital video display, and the AFM images were collected over a 50 μm × 50 μm or 100 μm × 100 μm area, with the fast scan direction perpendicular to the step edge. Images were analyzed after collection with the scanning probe microscopy data visualization and analysis software Gwyddion^† (version 2.45). Images were leveled by first subtracting a linear background row-by-row, followed by a plane level procedure to level either the protected or corroded region. In some cases, a median or median of differences procedure was necessary to align the height of the rows. Line profiles were then collected across the step edge in the image; each line profile was only one pixel thick and perfectly horizontal, such that it sampled only one scan row. Line profiles were then fitted with least-squares fit lines to both the protected and the corroded areas near the step edge; the fits were considered acceptable if the slopes of these two lines were similar (even if they were not perfectly horizontal). The top portion of the step edge, which was typically the steepest portion, was then fit with a least-squares fit line as well. The step height was taken as the difference in the height values where the fitted step edge line intersected the protected and corroded fit lines. Typically, three line profiles were collected from each AFM image, and the three resultant step heights were averaged. A Dektak 150^† stylus profilometer was used to collect additional line profiles across the corroded and protected regions of the glass coupons. Line profiles at a resolution of 0.167 μm/point and 4 mm in length were collected, which was sufficient to encompass the protected region and corroded regions on either side in a single scan. A stylus tip of 5 μm radius and force of 15.00 mg were used.

The C1 glass-ceramics formed through the three cooling rates produce drastically different microstructures, as expected. SEM micrographs of the polished surface of the glass-ceramic are shown in Figure 2. The fastest cooled 4× samples show the finest crystal size (Figure 2[a]) with increasing crystal size as the cooling rate decreases (Figures 2[b] and [c]). Multiple phases were observed in the SEM images, and the phase assemblages of the samples, as determined by XRD, are listed in Table 4. In the 1× sample oxyapatite (Ca₂Ln₈Si₆O₂₆) was the most abundant crystalline phase at 18.6 wt% and appears as the white regions in the SEM images. With a slowed cooling rate in the ¼× sample, the amount of oxyapatite decreased to 17.2%, while the oxyapatite increased in the 4× sample to 21.2%. The glass remainder phase had a reverse trend comprising 72.1% of the ¼× sample, 71.9% of the 1× sample, and 67.6% of the 4× sample. The powellite structure (light gray in the SEM image) was also present in two different unit cells, primarily Ca-Sr-rich (Ca_xSr_1-xMoO₄) at 9.1 wt% and a minor phase of Ca-Ba-rich (Ca_xBa_1-xMoO₄) at 0.4 wt% in the 1× sample. Spot analyses using EDS were performed to determine the elemental distribution in each crystal phase on the 1× sample, as summarized in Table 5. The powellite contained the Mo, alkaline earths, and lesser amounts of Na and Ce. The oxyapatite phase contained a large amount of the Ln, Ca, and Si. Similar distribution was found in the ¼× and 4× samples. However, these are not pure phases and most elements are expected to be distributed at some levels across all of the phases present in the sample.

The calculated final composition of the glass fraction of the 1× sample is given in Table 1. This final composition was determined by subtracting from the target glass composition (starting melt chemistry) the fraction of each element removed from the glass matrix to the crystalline phase upon cooling. These are considered a best estimate based on the analyzed composition of each crystalline phase and the corresponding measured mass% in the glass-ceramic. It was assumed that the relative partitioning of elements between the crystalline and glass phase remains consistent with different cooling rates. The waste loading in the final glass fraction drops from 47 mass% from the starting melt compositions upon cooling to 26 mass% as listed in Table 1. Thus, the crystalline phases contain just less than half of the representative waste components. A large fraction of the lanthanides, CaO, SrO, and MoO₃ report to the crystalline phases, while the Cs₂O and ZrO₂ remain in the glass phase. The concentration of additives Al₂O₃, B₂O₃, and SiO₂ remain similar or become more concentrated in the final glass compared to the starting melt.

The morphology of the 1× glass-ceramic is shown in Figure 3, where the oxyapatite crystals are shown in cross section through the short (Figure 3[a]) and long (Figure 3[b]) axes. The oxyapatite crystals are very long needle-like crystals with sharp, well-defined edges and glass-filled voids inside the crystals. The average size of the oxyapatite crystals measured across the short axis with SEM was 7.7±4.8 μm. A superstructure of powellite contained within oxyapatite, with all crystals having the same crystallographic orientation²²  (Figure 3[a]), was present in the glass-ceramics with a higher number occurring in the rapidly cooled sample, 4×. It is postulated that within these superstructures each phase has a unique anisotropic CTE, such that upon cooling stress is placed on the crystal to glass interface. This can lead to enhanced corrosion at these locations.

Periodic measurements of analyte concentrations were conducted to monitor the release of various glass-ceramic constituents as a function of time. A summary of SPFT tests, the measured flow rates, and flow rate to surface area (q/S) values are presented in Table 3. After the 3-week test duration, the analyte concentration was invariant (<15% relative standard deviation for three consecutive samples) with respect to time. From these steady-state analyte concentrations it is possible to compute the steady-state release rate from the following equation:

where C_i,s is the steady-state concentration of element i (g/m³) and is the average background concentration of element i (g/m³), q is the flow rate (m³/d), S is the surface area available for leaching (m²), and f_i is the mass fraction of element i in the pristine material (unit-less). The specific surface area, S, of the glass samples was estimated based on the following geometric formula:²⁰  

where m is the mass of glass particles (g), ρ is the glass density (g/m³), and r is the average particle radius (m). The use of the geometric surface area formula was recently shown to be more similar to the solution-derived dissolution rate than when the Brunauer-Emmett-Teller (BET) surface area was used in calculations.²¹  Because the glass-ceramic density is 3.3 g/cm³ and the average radius can be estimated from the powder size fraction, applying the above equation results in a specific surface area of 0.016 m²/g.

For the uncorrelated random errors in the experiment, the standard deviation of a function of the form f(x₁, x₂, x_n) was estimated using the Gaussian error propagation method:

where σ_f is standard deviation of the function f, x_i is parameter i, and σ_i is standard deviation of parameter i. By using the relative error, ⁠, the following equation is obtained:

Relative errors of 5%, 5%, 10%, 3%, and 15% were assigned to C_i, ⁠, q, f_i, and S, respectively. In combination of the error, an overall error of ∼20% on the rate values can be expected. The rate has a larger deviation the closer the solution concentration is to the background (blank) concentrations or DL of the instrument.

The steady-state release rates for various elements that are present in a variety of phases present in the glass-ceramic material are provided in Table 6. Two main trends are evident: (1) the release rate of the element increases with increasing values of q/S and (2) the release rate of the elements is higher in the pH(RT) 7 tests compared to the pH(RT) 9 tests. The impact of the different cooling rates of the glass-ceramic samples will be discussed later in more detail.

Changes in release rates are observed with changes in q/S because, at increasing values of q/S, the contacting solution becomes increasingly dilute. In these far-from-equilibrium conditions, the elements from the glass-ceramic material are released at or near their maximum rates. As rate-controlling constituent concentrations in the contacting solution are allowed to increase, i.e., at lower values of q/S, the system is nearer to equilibrium and the release rate slows. It should be noted that the glass phase of the glass-ceramic, being a thermodynamically unstable solid, never achieves thermodynamic equilibrium with solution.

An increase in release rates at pH(RT) 7 compared to pH(RT) 9 can be attributed to the effect of proton- and hydroxyl-promoted corrosion. The relative effect of proton and hydronium activity in solution is manifested by a commonly-observed v-shaped dependence of corrosion rate on pH for glasses, where a minimum in corrosion rate is observed at near-neutral pH and the corrosion rate increases as the solution becomes either more acidic or more basic.

In Figure 4, the steady-state release results of B at varying q/S ratios and the two different pH values used in this experiment are presented. Boron is mainly found in the glass portion of the glass-ceramic where it acts as a network-modifying element. As stated previously, the release rates at the pH(RT) 7 are roughly three to four times greater than the release at pH(RT) 9 at similarly large q/S values. From the figure, it is also observed that the B release rates increase with increasing q/S values as the system becomes increasingly dilute. Though conditions were never unambiguously observed where the B release rate is invariant with respect to changes in q/S, i.e., forward rate corrosion conditions, the data demonstrate that near forward rate conditions are achieved as the release rates asymptotically approach a constant value with increased q/S values. The release rates of B are near those calculated for the alkali elements, Li, Na, and Cs, the alkaline earth elements, Ca, Sr, and Ba, and the network-forming element, Si. Within the alkali elements, the relative release is systematically Li > Na > Cs, though the release rates of the three elements can be considered equal within error. This can be attributed to slower releases of the larger alkali ions through the ion exchange process and a minor portion of the Na reports to the powellite phase as well. No trend in release rates is observed for the alkaline earth elements.

Figure 4 also demonstrates the effect of cooling rate on the relative durability of the glass-ceramic. Independent of pH and q/S, the release rate from the 4× material is systematically lower than that of the ¼× and 1× materials.

In addition to B, in Figure 5, the release rates of Mo are provided, an element that is found primarily in the powellite phase of the material and can be used as an indicator of the relative durability of that phase. While, theoretically, some trace Mo may be present in the remainder glass or oxyapatite phase, none was measured in the analysis of these phases. Thus, assuming all of the Mo is present in the powellite phase is possible and the dissolution rate values can be viewed as conservative. Results are provided for experiments conducted at pH(RT) 7 and 9 at varying q/S values. Similarly to B, the release at pH(RT) 7 is greater than at pH(RT) 9 and the releases at higher q/S values are greater than those at lower q/S values. However, two major differences exist for this dataset compared with the dataset provided in Figure 4. First, the rates of Mo release are generally two times greater than those of B. Additionally, contrary to the results presented for B, where the 4× cooled material was more durable than the ¼× and 1× materials, the Mo results show that the 1× material has a higher release rate than both the ¼× and 4× material. As will be discussed in subsequent sections, this phenomenon may be attributed to the mechanical properties of the material that result from the different cooling rates.

Just as Mo can be used as a marker for the corrosion of the powellite phase, La and Nd can be used as a marker for the corrosion of the oxyapatite phase. From Table 1, it can be seen that the inventories of La and Nd do not fully segregate to the oxyapatite and both can also be found in the remainder glass. However, La and Nd are the most unique components of the oxyapatite phase and an assumption of their release resulting from the oxyapatite phase will, at best, result in a conservative corrosion rate of the oxyapatite phase. The release of La at both pH(RT) 7 and 9 at a variety of q/S is presented in Figure 6. Compared with Mo and B, the relative release of La is nearly two times lower at both pH values examined in this set of experiments. This result suggests that the oxyapatite phase is the most resistant to corrosion within the glass-ceramic. As well, this result suggests that the La release is resulting from the oxyapatite degradation and not driven by the inventory of La in the glass matrix. If the La was being released congruently from the glass phase, the rates measured would be comparable to B. However, the measured normalized rates for La are much lower than B, suggesting congruent dissolution from the glass is not occurring and the measured La release can be used as a conservative measurement of the dissolution of the oxyapatite phase. Analogous to the B data (Figure 4), the 4× material consistently shows a lowered release of La compared with the 1× and ¼× specimens.

Post-reaction characterization was performed on glass-ceramics coupons corroded at pH(RT) 9. These conditions resulted in a lower corrosion rate of the material and they are more representative of expected disposal conditions. A cross section of the sample was made and imaged with SEM (Figure 7). Figure 7(a) presents the cross section of the 1× sample subjected to the fastest SPFT flow rate (157 mL/d, q/S of 4.76 × 10⁻⁶ m/s). For the magnification used in the image, no altered glass layer can be observed at the glass-ceramic surface (top of the image). On the other hand, for the glass corroded in slightly less dilute conditions (52 mL/d, q/S of 1.57 × 10⁻⁶ m/s), an altered glass layer can be observed (Figure 7[b]). This change confirms that at the highest SPFT flow rates, a near forward corrosion rate relative to the glass matrix was achieved. This altered layer grows in depth and thickness at the lower flow rates of 17 mL/d (q/S of 1.06 × 10⁻⁷ m/s, Figure 7[c]) and 6 mL/d (q/S of 1.13 × 10⁻⁵ m/s, Figure 7[d]). A similar trend is expected for the other cooling rate glass-ceramics based on the solutions data in the Single-Pass Flow-Through Results section. In general, the images in Figure 7 demonstrate that the corrosion damage was primarily done to the glass phase. The images also show the appearance of cracked dark gray regions extending into the samples in Figure 7 that appear to track around the oxyapatite phases. The powellite phase in the altered region is less prevalent compared to the bulk glass. This is possibly a result of the powellite dissolving or being spalled from the glass-ceramic upon contact with the bulk solution. This supports the findings of the SPFT results in that the oxyapatite phase is more durable in the test conditions compared to the powellite and bulk glass phases.

AFM images and the topographical maps were also utilized to observe the height difference between the corroded and masked regions of the glass-ceramic surface following pH(RT) 9 testing (Figure 8). Figure 8(a) shows the ¼× glass-ceramic following SPFT testing at pH(RT) 9 and a 156.0 mL/d flow rate. The right side of the image shows the protected (masked) area which did not undergo any corrosion and remains relatively flat. Lines scans across the step between the corroded and protected areas of the AFM image give a step height of 3.2 μm. The step height is the distance from the top of the “protected” surface to the corroded surface. Figure 8(b) shows the 1× glass-ceramic following SPFT testing in pH 9 at a 157.5 mL/d flow rate. Again, the masked portion is on the right side of the image. Within the unmasked region of this sample, several remaining “islands” are present. The averaged step height was 3.9 μm, which was larger than the ¼×. However, the islands present in the AFM map of the 1× material suggest less of the surface was corroded on that sample. The corroded 4× glass-ceramic can be seen in Figure 8(c). In the topographical AFM map within the corroded region, a large portion of the initial surface was retained compared with the ¼× and 1× samples present as retained islands. Performing depth profile line scans with the AFM from the corroded to uncorroded regions was challenging because of the mixture of high and low points on the surface; however, an average step height of 3.7 μm was measured.

Because of the observed localized nature of the corrosion damage and apparent preferential corrosion of the glass and powellite phases, the AFM may have too fine of a resolution to observe the nature of the damage along the sample surface. For this reason, a stylus profilometer was used to produce line scans across a large section of sample surface, crossing the protected portion of the sample coupon. The resulting line scans are presented in Figure 9. The protected region can be seen centered in all of the line scans as the region at height 0 μm. In all cases, the corroded region consists of numerous islands which have maintained the height of the original glass surface and are surrounded by recessed areas where corrosion has occurred. The ¼× sample has the deepest surface recessions and the observed recessions have the most girth. The 1× sample has more shallow features in the corroded regions while the 4× sample was measured to have the shallowest step height. For comparison, a line scan from a lower q/S sample (18 mL/d flow rate) on the 4× glass-ceramic is also shown. Because of the lower flow rate, the solution in the reactor is more concentrated and this would slow the corrosion of the phases. This is clearly shown in the line scan of the 4× sample at 18 mL/d which displays the least corrosion in the line scan.

The glass-ceramic waste form investigated in this work is part of an ongoing effort to reduce costs, volume, and time required for solidification and safe disposal of nuclear wastes. The influence of the microstructure size of the glass-ceramic was investigated by testing glass-ceramics of identical composition, but cooled at different rates, in SPFT testing at two pH values at 90°C. Faster cooling rates led to a smaller powellite and oxyapatite secondary phases. SPFT testing allowed for comparison of the corrosion rates of the individual phases compared with the glass matrix and other waste form materials. The release rate of B, a primary component of the glass matrix in the glass-ceramic, can be used to compare to the release rate of the glass-ceramics samples to simulated nuclear waste glasses subjected to similar experiments buffered to pH 9 with Tris in dilute conditions. Using the SPFT method, Icenhower and Steefel²⁹  found a dissolution/corrosion rate of 3.0 × 10⁻¹ g·m⁻²·d⁻¹ for SON68, a nuclear waste reference glass, while Pierce, et al.,³⁰  found a dissolution rates between 1.6 × 10⁻¹ g·m⁻²·d⁻¹ and 2.5 × 10⁻¹ g·m⁻²·d⁻¹ for three low-activity waste (LAW) glasses (LAWA44, LAWB45, LAWC22) and one high-level waste simulant glass (SRL202). The LAW glasses were developed and tested as candidate glasses for the immobilization of LAW waste at the DOE Hanford Site. At the highest q/S value used in this experiment, which is shown to be near the forward corrosion rate, the B release rates for the ¼×, 1×, and 4× glass-ceramics were found to be 4.03 × 10⁰ g·m⁻²·d⁻¹, 4.17 × 10⁰ g·m⁻²·d⁻¹, and 2.66 × 10⁰ g·m⁻²·d⁻¹, respectively, i.e., one order of magnitude greater than the rates measured by the other authors. However, forward rate values at pH(RT) 9 and 90°C for SON68 have been reported to be 4.5 × 10⁰ g·m⁻²·d^{−1 31}  and 5.5 × 10⁰ g·m⁻²·d^{−1 32}  and for ISG, a simplified version of SON68, to be 3.66 × 10⁰ g·m⁻²·d⁻¹³³  using different methods to obtain the forward rate. Therefore, though the B release values for the C1 glass-ceramic are larger than the forward rate calculated from other glasses using the SPFT method, the values are near forward rate values presented using different test methods. Additionally, it is shown that the glass and secondary phases dissolve at lower rates when the contacting solution becomes more concentrated. Because solution renewal conditions in a repository are expected to be low, the lower corrosion rate in more concentrated conditions assures that the dilute-condition reaction rate is conservative. The rate-controlling species for the overall corrosion rate of the glass-ceramic is most likely a combination of several solution species as each individual phase would be suspected to respond differently to different solution chemistries. Lastly, to the authors’ knowledge, the present paper is the first effort to measure the forward rate of a glass-ceramic and thus comparison to other values in the literature for a glass-ceramic is not possible.

While the behavior of the glass-ceramic matrix phase is comparable to simulated nuclear waste glasses, the crystalline phases will contain approximately half of the contaminants of concern and radionuclides contained in waste form, while others, such as Cs in the glass-ceramics studied within, will partition to the glass matrix. It is imperative to determine the corrosion rate of all components of the glass, including the individual crystalline phases, which can be accomplished using species unique to the phases. Mo can be found exclusively in the powellite phases present in the glass-ceramic and thus can be used as a tracer for the corrosion of the powellite from the surface. Mo has an approximate order-of-magnitude higher release rate than B. This result indicates that the relative durability of the powellite phase is lower than that of the bulk glass phase. As a result, the interface of the bulk glass phase and the powellite may be increasingly susceptible to corrosion attack as the boundary between the two most active phases may lead to enhanced corrosion in that area.

However, this finding is not detrimental to the overall performance of the glass-ceramic in terms of radionuclide release, as the oxyapatite phase will retain a significant portion of the radionuclides in the waste form. Specifically, the La inventory is predominantly present in the oxyapatite phase, making it a tracer for the specific corrosion of the phase. Comparing the release rate of La to B, for the glass matrix, and to Mo for the powellite phase, a near order-of-magnitude decrease in release was observed at pH(RT) 7 and an approximately three order-of-magnitude decrease was observed at pH(RT) 9. Thus, the oxyapatite phase appears to be more durable as the alkalinity of the solution increases. This result suggests that the oxyapatite is the most stable phase present in the glass-ceramic and, as a result, its constituents will be released at a much slower rate. This is also evidenced by the propagation of damage observed on the glass-ceramics following the SPFT testing, where the oxyapatite phase remained relatively untouched as corrosion progressed. Generation of the oxyapatite phase in the waste form containing radionuclides can increase retention over long periods.

With the observed heterogeneous corrosion of the phases, their size and distribution may directly influence the overall corrosion behavior of the glass-ceramic. The faster cooled glass-ceramic (4×) produced smaller secondary phases that were spread more evenly across the surface, while slow cooling (¼×) facilitated larger secondary phase crystal growth. In general, the finer microstructure in the 4× ceramic leads to the lowest measured release rates for the bulk glass and oxyapatite elements. While literature on glass-ceramics corrosion is minimal, a similar observation has been made for highly corrosive metal alloys, such as Mg. In Al-containing Mg alloys, reducing microstructure size leads to a decrease in corrosion rate.³⁴  The reason for the increased corrosion resistance is that the Al, which is heterogeneously distributed within secondary phases, provides corrosion resistance to the alloy. As corrosion propagates along the surface through the Mg matrix, it is slowed when the corrosion front contacts an Al-rich region. As corrosion penetrates inward into the alloy a similar process occurs. In alloys where the secondary phases are large and spread sparsely, corrosion attack can find a clear, Al-free, path to widely propagate. However, with decreased secondary phase size and increased distribution of Al, corrosion attack is more likely to encounter an Al-rich region and be stopped. This process leads to increased corrosion resistance for the smaller microstructure alloys. A similar process is likely occurring in the glass-ceramics, with the oxyapatite acting as a corrosion resistance barrier that can limit corrosion propagation.

However, the powellite phase did not follow the trend of decreasing corrosion rate with decreasing microstructure size. Mo had the highest releases in the 1× glass-ceramic, along with a similar trend for the Sr and Ba, which are also powellite components. The size and distribution of the powellite phase in the 1× glass-ceramic microstructure may be a perfect storm that can lead to increased release of elements from the powellite phase. The powellite phase is the most susceptible to corrosion. In the ¼× glass-ceramic, large powellite structures are present. In the ¼× glass-ceramic, the initially exposed powellite can corrode, exposing the underlying glass. The corrosion attack must then expose a second, subsurface powellite phase before additional components can be released. As these phases are widely spaced in the ¼× glass-ceramic, it is less likely that the corrosion propagation will reach a new powellite phase. The oxyapatite superstructure may also be protecting the powellite. In the 4× glass-ceramic, the powellite phases are more evenly spread through the ceramic; however, as discussed previously, the oxyapatite will slow corrosion propagation and in turn prevent the exposure of new powellite phases. This can be assisted by the superstructures of oxyapatite surrounding the powellite. The 1× glass-ceramic may then have a microstructure whose size and distribution places the corrosion process in between these two scenarios, where powellite phases are more easily exposed and the oxyapatite distribution is not wide enough to slow the attack.

• •

  Glass-ceramic waste forms are a potentially viable technology for increasing waste loading during vitrification of nuclear wastes. These materials contain crystalline phases, commonly avoided to a large degree in glass waste form fabrication, which can hold a portion of the radionuclide inventory. The bulk glass matrix of the ceramic waste form had comparable corrosion rates to other simulated nuclear waste glasses. Additionally, this study showed that changing the rate of cooling of the glass-ceramic waste form leads to a change in the glass-ceramic microstructure and in turn the corrosion performance of the waste form. Through examination of the post-reaction solids, a smaller, more widely distributed microstructure produced with rapid cooling was found to provide a lower overall corrosion rate. This study suggests that by controlling the cooling rate of the material, in this case at a rate that is 4× longer than the expected “natural” cooling rate of the glass-ceramic in the container, a more durable product can be fabricated. However, this is not necessarily an optimized cooling rate and should be examined for additional glass-ceramic compositions. Lastly, the corrosion of the individual phases was also compared using phase specific elements and the rate of corrosion increased in the order of oxyapatite < bulk glass < powellite.

The Pacific Northwest National Laboratory is operated by Battelle under Contract No. DE-AC05-76RL01830. This work was funded by the U.S. Department of Energy (DOE) Office of Nuclear Energy under the Fuel Cycle Research and Development Program. The authors wish to thank Yingge Du of PNNL for his assistance in initial studies of the glass-ceramic. A portion of the research was performed at the Environmental Molecular Sciences Laboratory (EMSL), a DOE Office of Science user facility sponsored by the Office of Biological and Environmental Research and located at Pacific Northwest National Laboratory.