Incorporation of strontium and calcium in geopolymer gels

Radioactive waste streams containing 90 Sr, from nuclear power generation and environmental cleanup opera- tions, are often immobilised in cements to limit radionuclide leaching. Due to poor compatibility of certain wastes with Portland cement, alternatives such as alkali aluminosilicate ‘geopolymers’ are being investigated. Here, we show that the disordered geopolymers ((N,K)-A-S-H gels) formed by alkali-activation of metakaolin can readily accommodate the alkaline earth cations Sr 2+ and Ca 2+ into their aluminosilicate framework structure. The main reaction product identiied in gels cured at both 20°C and 80°C is a fully polymerised Al-rich (N,K)-A-S-H gel comprising Al and Si in tetrahedral coordination, with Si in Q 4 (4Al) and Q 4 (3Al) sites, and Na + and K + balancing the negative charge resulting from Al 3+ in tetrahedral coordination. Faujasite-Na and partially Sr-substituted zeolite Na-A form within the gels cured at 80°C. Incorporation of Sr 2+ or Ca 2+ displaces some Na + and K + from the charge-balancing sites, with a slight decrease in the Si/Al ratio of the (N,K)-A-S-H gel. Ca 2+ and Sr 2+ induce essentially the same structural changes in the gels. This is important for understanding the me- chanism of incorporation of Sr 2+ and Ca 2+ in geopolymer cements, and suggests that geopolymer gels are excellent candidates for immobilisation of radioactive waste containing 90 Sr.


Introduction
Conditioning of radioactive waste arising from nuclear facility operation and decommissioning is essential to its safe disposal and environmental remediation. Cementation ofers signiicant advantages over other waste conditioning routes (such as thermal conversion of ceramics or glasses) including simplicity, relatively high throughput, low cost, and the absence of secondary waste generation (Aldridge et al., 1997). Immobilisation of radionuclides such as strontium-90 ( 90 Sr, T 1/2 = 28.80 yrs) by cementation has received particular attention due to its presence as a ission product, including in nuclear reactor cooling water (Qian et al., 2002;Ochs et al., 2015;Kuenzel et al., 2015), which necessitates the use of granular ion exchange resins which must then be conditioned prior to disposal (Tusa, 2014). Traditionally, strontium-containing waste has been encapsulated in Portland cement (PC), which has shown potential for uptake of Sr as well as lanthanides and actinides including Nd, U, and Np. Sr uptake is usually via sorption to the surface of calcium silicate hydrate (C-S-H) (Tits et al., 2015;Vespa et al., 2014;Gaona et al., 2012;Mandaliev et al., 2010;Wieland et al., 2008;Tits et al., 2006), but co-precipitation of the strontium silicate hydrate phase Sr 5 Si 6 O 16 (OH) 2 .5H 2 O exhibiting structural similarity to tobermorite (Ca 5 Si 6 O 16 (OH) 2 .xH 2 O, where x = 4 or 5) has been observed (Felmy et al., 2003). Furthermore, C-S-H is space-illing and hence restricts ion mobility. However, the high water content of PC and the prevalence of hydrate phases means that the mechanism of Sr 2+ uptake involves adsorption to partially hydrated phases such as calcium silicate hydrate through reversible ion-exchange interactions (Wieland et al., 2008;Tits et al., 2006), while retention of other radioisotopes such as Cs may be too low to prevent difusion into the environment (Goñi et al., 2006). Consequently, the applicability of encapsulation of these radionuclides in PC is limited.
Alkali aluminosilicate cements, often called geopolymers, comprise a fully polymerised, three-dimensionally cross-linked, structurally disordered alkali aluminosilicate network. Because of their cation-binding sites, mechanical properties and chemical resistance, these materials have received signiicant interest in recent years as suitable materials for the cementation of radioactive waste (Arbel Haddad et al., 2017;Williams et al., 2016;Provis et al., 2008;Ofer-Rozovsky et al., 2019). The alkali aluminosilicate hydrate gel framework is often abbreviated as (N,K)-A-S-H, as Na and K are the most common alkalis used in their production, and it has a pseudo-zeolitic or proto-zeolitic structure (Provis et al., 2015). Al and Si are both present in tetrahedral coordination, analogous to their roles in zeolites, with Si existing in gel), while Al is predominantly in q 4 (4Si) environments because of the energetic penalty associated with Al IV −O-Al IV bonding . The negative charge associated with Al substitution for Si is balanced by the alkali cations. The gel nanostructure can be signiicantly afected by kinetic limitations on silica and alumina release from solid precursors used in gel synthesis (Hajimohammadi et al., 2010(Hajimohammadi et al., , 2011 and it consequently evolves as the alkali activation reaction proceeds. Due to the pseudo-zeolitic gel nanostructure, crystalline zeolite phases with longer-range order have been observed to form in geopolymer gels when held at elevated temperatures (Rivera et al., 2016;Rickard et al., 2015Rickard et al., , 2011. The temperature-dependent phase stability of geopolymer gel-based wasteforms is of particular importance, as wasteforms which are stored in geological disposal facilities are expected to experience temperatures of 20°C during initial storage above ground (Hunter and Swift, 2013;Godfrey and Cann, 2015;Boden, 2002;Nuclear Decommissioning Authority, 2014) and between 35°C and 80°C across the irst 100 years of their storage in the geological disposal facility (Boden, 2002;Nuclear Decommissioning Authority, 2014;Prentice et al., 2019).
Previous work utilising high resolution transmission electron microscopy coupled with energy dispersive X-ray spectroscopy and selected area electron difraction has identiied incorporation of 1-2 mol. % Sr in amorphous geopolymer gels with molar Si/Al = 2.0 and Na/ Al = 1 (Blackford et al., 2007;Perera et al., 2006). The gels were loaded with 5 wt. % Sr during formulation, with the remaining Sr preferentially forming SrCO 3 rather than being incorporated into the geopolymer gel. Synthetic (calcium, alkali) aluminosilicate gels loaded with 0.1 wt. % Sr showed retention of up to 99.9% Sr 2+ when leaching for 7 days at 20°C in MilliQ water, with Sr retention enhanced by reduction of the Si/Al and Ca/(Si + Al) (independently of Si/Al) molar ratios in the (calcium, alkali) aluminosilicate gels (Vandevenne et al., 2018). This was attributed to increased formation of N-A-S-H, relative to the predominant sodium-and aluminium-substituted calcium silicate hydrate gel, for gels with lower Ca/(Si + Al) molar ratios. Together, these indings demonstrate the potential for geopolymer cements to yield suitable wasteforms for the cementation of radioactive waste.
Despite these recent advances, and the interest in immobilisation of radionuclides within non-Portland cements, the transport, solubility and incorporation processes which control the long-term performance of these wasteforms are still not well understood. In particular, a detailed analysis and understanding of the incorporation of Sr (either as 90 Sr or the non-radioactive isotopes of Sr) in geopolymer gels at the atomic scale is not yet available.
Here, we investigate the mechanism of incorporation of non-radioactive isotopes of Sr 2+ and Ca 2+ in a series of metakaolin-based geopolymer gels, to assess their potential as materials for cementation of radioactive waste. Ca 2+ was also investigated to compare the efects of the difering ionic radii of these two divalent cations (Shannon radii r = 1.18 Å and 1.00 Å for Sr 2+ and Ca 2+ , respectively) (Shannon, 1976). The trace element level radionuclide loading investigated is representative of that expected for wasteforms produced by cementation of radioactive ion exchange resins (Kirishima et al., 2015;Koma et al., 2017). Multinuclear magic angle spinning (MAS) and cross polarisation (CP) MAS nuclear magnetic resonance (NMR) spectroscopy probing the local environments of 29 Si, 27 Al and 23 Na, in conjunction with Fourier transform infrared (FTIR) spectroscopy and X-ray diffraction (XRD) measurements, are used to examine any nanostructural changes that may be induced by incorporation of the alkaline earth cations, and to investigate the inluence of the alkali cations Na + and K + on the incorporation mechanism of the alkaline earth cations. The indings presented here have signiicant implications for the long term stability and durability of these materials and suggest that metakaolinbased geopolymer gels are attractive candidate wasteforms for immobilisation of radioactive waste containing radioactive 90 Sr.

Sample preparation
Geopolymer gels were produced by reaction of metakaolin (MetaStar 501, Imerys, composition as determined by X-ray luorescence provided in Table 1, D 50 = 4.57 μm) with a sodium silicate, potassium silicate, or mixed alkali (containing both sodium and potassium silicate) activating solution.
The activating solutions were prepared by dissolving sodium hydroxide powder (AnalaR 99 wt.%), potassium hydroxide powder (AnalaR 99 wt.%), or a combination of both, in sodium silicate (PQ-NS, 44.1 wt.% sodium silicate, with a solution modulus of SiO 2 /M 2 O = 2.1, with the balance water, PQ UK) or potassium silicate solution (PQ-KS, 51.6 wt.% potassium silicate, with a solution modulus of SiO 2 / M 2 O = 2.2, with the balance water, PQ UK) and distilled water. Stoichiometry was designed to obtain an activating solution modulus of SiO 2 /M 2 O = 1 (where M represents Na and/or K), an activator dose such that M 2 O/Al 2 O 3 = 1 and SiO 2 /Al 2 O 3 = 3 (Si/Al = 1.5) in the inal reaction mixture, water content such that M 2 O/H 2 O = 11, and the nominal chemical composition and water/solids (w/s) ratios outlined in Table 2.
The activating solution was mixed with metakaolin to form a homogeneous paste which was subsequently cast in sealed containers cured for 3 months at either 20°C ± 2°C or 80°C ± 2°C. Reactant proportions were approximately 40 wt. % metakaolin, 44 wt. % alkaline activating solution and 16 wt. % additional water. Samples containing nominally natural abundance proportions of the isotopes of alkaline earth cations Ca 2+ and Sr 2+ were produced by mixing metakaolin, the activating solution (prepared as described above) and either Ca(OH) 2 (Sigma Aldrich) or Sr(OH) 2 •8H 2 O (Sigma Aldrich) to obtain ME/ Al = 0.00025 (where ME represents the alkali earth cations Sr 2+ or Ca 2+ ) in the inal reaction mixture. This concentration is representative of that expected for wasteforms produced by cementation of radioactive ion exchange resins (Kirishima et al., 2015;Koma et al., 2017).

Characterisation
The hardened binders were hand ground using a pestle and mortar, immersed in acetone for 15 min to remove loosely bound water (thereby halting the alkali-activation reaction without signiicantly altering the geopolymer gel structure (Ismail et al., 2013)), and subsequently iltered prior to storing in a sealed desiccator under vacuum.

Table 2
Nominal chemical composition (molar basis) of the alkali activating solution. 2.2.1. X-ray diffraction X-ray difraction (XRD) data were obtained to examine the longrange ordering and crystallinity of the reaction products in each sample. Data were obtained across a 2θ range of 5°-70°using a Panalytical X'Pert³ Powder X-ray difractometer with Cu Kα radiation (1.54 Å), a nickel ilter, a step size of 0.020°and a count time of 1 s/ step. Difracted background intensity at low angles was reduced using an anti-scatter blade, and an incident beam divergence of 1.0 mm and a 2.5°Soller slit in the difracted beam were used. Phase identiication was performed using Difrac.EVA V4.1 software with the ICDD PDF4 + 2015 database.

Solid state nuclear magnetic resonance spectroscopy
Solid state single pulse 29 Si, 27 Al and 23 Na magic angle spinning (MAS) NMR data were obtained to examine the local structure of the reaction products in each sample. All spectra were acquired on a Bruker Avance III HD 500 spectrometer at 11.7 T (B 0 ) using a 4.0 mm dual resonance CP/MAS probe, yielding a Larmor frequency of 99.35 MHz for 29 Si, 130.32 MHz for 27 Al and 132.29 MHz for 23 Na. 29 Si MAS NMR spectra were acquired using a 5.5 μs non-selective (π/2) excitation pulse, a measured 90 s relaxation delay, a total of 256 scans and spinning at 12.5 kHz. 1 H-29 Si cross-polarisation (CP) MAS NMR experiments were performed using the same instrument with a spinning frequency of 12.5 kHz, a 29 Si non-selective (π/2) pulse width of 1.7 μs, an initial 1 H non-selective (π/2) pulse width of 2.5 μs, a recycle delay of 1.5 s and Hartmann-Hahn contact periods of 1.7 ms. A nominal 1 H decoupling ield strength of 80 kHz was employed during acquisition and 10,240 scans were collected per experiment. 27 Al MAS NMR spectra were acquired using a 1.7 μs non-selective (π/2) excitation pulse, a measured 10 s relaxation delay, a total of 128 scans and spinning at 12.5 kHz. 23 Na MAS NMR spectra were acquired using a 3 μs non-selective (π/2) excitation pulse, a measured 10 s relaxation delay, a total of 128 scans and spinning at 12.5 kHz. All 29 Si, 27 Al and 23 Na spectra were referenced to pure tetramethylsilane (TMS), 1.0 M aqueous Al(NO 3 ) 3 and 1.0 M aqueous NaCl (aq) , respectively, at 0 ppm.
Gaussian peak proiles were used to deconvolute the 29 Si MAS and 1 H-29 Si CPMAS NMR spectra (Massiot et al., 2002). The minimum number of peaks possible were itted, and the isotropic chemical shift (δ iso ) and peak full width at half maximum (FWHM) of each resonance were required to be consistent in both the 29 Si MAS NMR and 1 H-29 Si CP MAS NMR spectral deconvolutions. Peak intensities were required to be consistent with the structural constraints described by the thermodynamics of a statistical distribution of Si and Al sites within a Q 4 aluminosilicate network for (N,K)-A-S-H gel products . Previous work utilising 29 Si and 27 Al MAS NMR and Fourier transform infrared spectroscopy has shown that these models best describe the environments in which Si and Al exist in (N,K)-A-S-H gels Fernández-Jiménez et al., 2006;Lecomte et al., 2006;Criado et al., 2008;Walkley et al., 2016). In all spectral deconvolutions congruent dissolution was assumed, and the contribution from unreacted precursor particles was taken into account by itting a scaled component spectrum calculated from the 29 Si and 27 Al MAS NMR spectrum of unreacted metakaolin.

Fourier transform infrared spectroscopy
Fourier transform infrared (FTIR) spectroscopy data were obtained to examine the local structure of the reaction products in each sample. Samples for FTIR spectroscopy analysis were prepared by mixing 2 mg of sample with 200 mg KBr and pressing the mixture into a pellet. FTIR spectra were measured using a Perkin Elmer Frontier Mid FT-IR spectrometer equipped with a deuterated triglycine sulfate (DTGS) detector and KBr beam splitter optical system, scanning 16 times at a resolution of 4 cm −1 .

X-ray diffraction
XRD data for unreacted metakaolin and for each geopolymer gel cured at 20°C are shown in Fig. 1. Data for unreacted metakaolin displays a dominant broad feature due to difuse scattering centred at approximately 22°2θ, consistent with its amorphous nature. Relections due to the crystalline phases anatase (TiO 2 , Powder Difraction File (PDF) # 01-084-1286), quartz (SiO 2 , PDF # 01-078-2315) and hydroxylated muscovite (illite-2, (K,H 3 O)(Al,Mg,Fe) 2 (Si,Al) 4 O 10 [(OH) 2 , (H 2 O)], PDF # 00-026-0911) are observed in the X-ray difraction pattern for unreacted metakaolin. The X-ray difraction patterns for each geopolymer gel cured at 20°C exhibit a dominant broad feature due to difuse scattering, centred at approximately 29°2θ, and indicating the formation of a crystallographically disordered reaction product (Duxson et al., 2006;Provis et al., 2005b). Relections assigned to anatase, quartz and hydroxylated muscovite are also visible in the XRD data for each aluminosilicate gel and indicate that while these phases remain largely intact throughout the reaction process under these conditions, small changes in the intensity of the relections assigned to hydroxylated muscovite suggest partial reaction and/or changes in crystallinity due to partial reaction of hydroxylated muscovite has occurred. A small amount of unreacted metakaolin is also observed by solid state NMR in all geopolymer gel samples, and is discussed further below.
The X-ray difraction patterns for each geopolymer gel cured at 80°C (Fig. 2) also exhibit a dominant broad feature due to difuse scattering from (N,K)-A-S-H gel centred at approximately 29°2θ, as at 20°C, and the peaks assigned to anatase, quartz and hydroxylated muscovite are again observed. The XRD data for geopolymer gels produced by reaction of metakaolin with potassium silicate, or with a combination of sodium silicate and potassium silicate, do not exhibit any other observable relections of crystalline phases. No diferences observable by XRD are induced by the presence of alkaline earth cations at the concentrations used here. The X-ray difraction patterns for the geopolymer produced by reaction of metakaolin with sodium silicate at 80°C exhibit, in addition to those identiied above, relections due to the zeolite faujasite-Na (Na 7 [Al 7 Si 17 O 48 ]·32(H 2 O), PDF 48-0731). The formation of crystalline zeolite phases from the geopolymer gels cured at 80°C is consistent with the proto-zeolitic structure that has been identiied in alkali aluminosilicate gels synthesised directly under similar conditions . Hydrothermal synthesis of faujasite-Na at 80°C has been previously reported to be favourable at Si/Al ratios close to unity in the presence of Na + cations (Barrer and Mainwaring, 1972a;Davis and Lobo, 1992). In the presence of K + (either in Na + -free or mixed alkali systems containing both Na + and K + ) formation of chabazite-K is favoured when 2 ≤ Si/Al ≤ 4.5, and faujasite formation has not been previously observed (Barrer and Mainwaring, 1972a;Barrer et al., 1968;Barrer and Mainwaring, 1972b). Furthermore, previous work has shown that small alkali metal cations produce the highest rates of zeolite nucleation (McCormick and Bell, 1989), and chabazite-type zeolite crystallite yield decreases with increasing K + (in the form of KOH) concentration (Barrer and Mainwaring, 1972b). Geopolymers produced by alkali-activation of ly ash with potassium silicate also display a lower extent of crystallisation than those produced by alkaliactivation of ly ash with sodium silicate (van Jaarsveld and van Deventer, 1999). Together, these previous observations are consistent with the formation of faujasite-Na in the geopolymer cement produced by reaction of metakaolin with sodium silicate at 80°C, and the absence of any zeolite phases in the geopolymer cements produced by reaction of metakaolin with potassium silicate or a mixture of sodium and potassium silicate at 80°C (where Si/Al < 2).
The presence of a low content of Ca 2+ (r = 1.00 Å) does not alter the nature of the reaction products formed due to its similarity in size with Na + (r = 1.02 Å), however the intensity of the relections due to crystalline phases is decreased, suggesting a lower extent of formation of these phases. Suppression of zeolite crystallisation in the presence of Ca 2+ has been previously observed, including in geopolymers Catalfamo et al., 1997;Phair et al., 2003). This results from the larger number of potential nucleation sites and therefore smaller mean crystallite size, as well as from the reduction in Si supersaturation levels due to formation of calcium silicate hydrates (Catalfamo et al., 1997;Jacobs, 1992). The presence of Sr 2+ (r = 1.18 Å) in the geopolymer gel produced by reaction of metakaolin with sodium silicate results in formation of a zeolite A-type phase, the structure of which may contain some Sr although there is not suicient change in lattice parameter to enable us to distinguish between the LTA structures of both zeolite A-Na (Na 12 [(AlO 2 ) 12 (SiO 2 ) 12 ]⋅27H 2 O, PDF # 39-0222) and a partially Sr-substituted zeolite A-Na ((Na,Sr) 12 [(AlO 2 ) 12 (SiO 2 ) 12 ] ⋅27H 2 O, PDF # 38-0243). Previous work has shown strong adsorption selectivity for Sr 2+ in zeolite A (when compared with faujasite-Na) in the presence of Na + and K + due to the high cation exchange capacity of the LTA structure (Munthali et al., 2015), aligning with the observation of relections due to zeolite A in the XRD data for geopolymer cements produced by reaction of metakaolin with sodium silicate at 80°C and in the presence of Sr. The extent of formation of zeolite A is much lower than that of faujasite-Na formed in the Sr 2+ -free geopolymers produced by alkali-activation of metakaolin with sodium silicate. This is consistent with the larger ionic radius of Sr 2+ (r = 1.18 Å) compared with Na + (r = 1.02 Å), which results in a decrease in zeolite nucleation and formation (McCormick and Bell, 1989).

29 Si MAS and 1 H-29 Si CP MAS NMR
The 29 Si MAS and 1 H-29 Si CP MAS NMR spectra for unreacted metakaolin and the geopolymer gels cured at 20°C are shown in Fig. 3. 29 Si MAS NMR data for unreacted metakaolin exhibits a single broad resonance at δ iso between -80 to -125 ppm with maximum intensity at δ iso = -107 ppm. This indicates an extensive distribution of Si environments and chemical shifts, consistent with the broad amorphous feature observed in the XRD data for unreacted metakaolin and with previous 29 Si MAS NMR observations for similar materials . This resonance contains contributions from a distribution of Q 4 (mAl) environments (where 0 ≤ m ≤ 4), with the maximum intensity at δ iso = -107 ppm indicating that this distribution is dominated by species with lower Al substitution (Q 4 (0Al) and Q 4 (1Al) sites). Quartz (identiied by XRD, Fig. 1) will also contribute to the 29 Si MAS NMR spectrum of metakaolin at approximately δ iso = -110 ( ± 3) ppm due to resonance of Q 4 (0Al) sites (Kowalczyk and Roberts, 1994;Myers et al., 1998;Bernal et al., 2013), however due to the very long 29 Si relaxation times (on the order of 1 h) (Kowalczyk and Roberts, 1994;Myers et al., 1998) and the extensive disorder of Si sites in metakaolin, the Q 4 (0Al) site in quartz cannot be resolved unambiguously in the spectra presented here. 29 Si MAS NMR data for illite exhibit resonances within the region δ iso = -85 to -95 ppm, (Lausen et al., 1999) and therefore this phase also contributes to the signal in this region in the spectra presented here.
The 29 Si MAS NMR spectra of the geopolymer gels each exhibit a broad resonance spanning from δ iso = -75 to -110 ppm and centred at δ iso = -87.8 ppm, with a consistent lineshape of the distribution of δ iso across all samples. This resonance is attributed to a distribution of  (Walkley et al., 2016;Duxson et al., 2005b), and the shift in the resonance compared with that of metakaolin indicates that alkali activation of metakaolin has resulted in formation of a reaction product dominated by higher Al substituted Q 4 (4Al) and Q 4 (3Al) sites. The 1 H-29 Si CP MAS NMR spectra of the geopolymer gels each exhibit a broad resonance spanning from δ iso = -75 to -95 ppm and centred at δ iso = -83.8 ppm, again with a consistent lineshape of the distribution of δ iso across all samples. The 1 H-29 Si CP MAS NMR signal is sensitive to the internuclear distance between the Si atoms and nearby protons, such that 1 H-29 Si CP MAS NMR signal of Si atoms in closest proximity to protons is preferentially enhanced (Kolodziejski and Klinowski, 2002) and this data therefore indicates those Si species within the hydrated reaction product. Consequently, it is possible to diferentiate between resonances resulting from hydrated Si species in the (N,K)-A-S-H gel and unreacted Si species in remnant unreacted metakaolin particles. The maximum intensity at δ iso = -83.8 ppm in the 1 H-29 Si CP MAS NMR spectra of the geopolymer gels indicates that the hydrated (N,K)-A-S-H gel contains primarily Q 4 (4Al) Si sites, independent of the activator used.
Deconvolution and quantiication of the 29 Si MAS and 1 H-29 Si CPMAS NMR data ( Fig. 4 and Table 3) identify the presence of two new Si environments which form in each geopolymer gel upon alkali activation of metakaolin. These environments resonate at δ iso = -84.7 and -89.0 ppm, respectively, and are attributed to Q 4 (4Al) and Q 4 (3Al) sites within a fully polymerised, Al-rich (Si/Al ≤ 1.2) (N,K)-A-S-H gel . The molar Si/Al ratio of the (N,K)-A-S-H gel may be calculated using Engelhardt's formula (Engelhardt et al., 1981)  (1) where I AQ 4 (mAl) refers to the normalised relative integral areas of the resonances in the 29 Si MAS NMR spectral deconvolutions of each Q 4 (mAl) site within the (N,K)-A-S-H gel (i.e. excluding resonances due to Q 4 (mAl) sites within remnant unreacted metakaolin).
Eq. 1 requires that Loewenstein's rule is obeyed such that there are negligible Al-O-Al bonds present. This assumption has been shown to be valid by application of 17 O 3QMAS NMR spectroscopy to synthetic geopolymer gels with Si/Al > 1 (Walkley et al., 2018), and can therefore be safely assumed here where the nominal Si/Al = 1.5 for each geopolymer gel. Each (N,K)-A-S-H gel exhibits a Si/Al = 1.20 to 1.26. This value is lower than that of the initial reaction mixtures, indicating preferential precipitation of an Al-rich gel. Incorporation of alkaline earth cations Ca 2+ and Sr 2+ results in a slight increase in Q 4 (4Al) sites and a slight decrease in Q 4 (3Al) sites (Table 3), and consequently a slight decrease in the Si/Al ratio of the (N,K)-A-S-H gel (Table 3). This is likely to be due to the increased charge balancing capacity resulting from substitution of a divalent charge alkaline earth cation for a monovalent alkali cation. The magnitude of the decrease in Si/Al ratio of the gel appears to be inluenced by the alkali cations present, with K + resulting in a greater increase in the Si/Al ratio of the gel, compared with Na + , upon incorporation of Ca 2+ or Sr 2+ . There is no variation (within the estimated error) in the Si/Al ratio of the geopolymer gels when comparing between the incorporation of Ca 2+ or Sr 2+ , indicating that both alkaline earth cations induce comparable structural changes.
The 27 Al MAS NMR spectrum for each geopolymer gel cured at 20°C exhibits a broad tetrahedral Al resonance spanning from δ obs = 70 to 50 ppm and centred at δ obs =60.7 ppm. This resonance is attributed to Al within a full polymerised tetrahedral (q 4 ) site in a (N,K)-A-S-H type gel (Walkley et al., 2018;Duxson et al., 2005c), with the negative charge arising from the substitution of Al 3+ for Si 4+ balanced by alkali cations (Na + or K + ). The presence of Al in exclusively tetrahedral coordination is expected due to the presence of excess alkali cations (Stebbins et al., 2000) and indicates that all Al within metakaolin has reacted, consistent with the 29 Si MAS and 1 H-29 Si CPMAS NMR observations of preferential dissolution of Al during alkali activation discussed above. The lineshape and width of the distribution of δ obs for this resonance are identical across all samples, regardless of the presence of, or variation in, diferent alkali (Na + or K + ) or alkaline earth (Ca 2+ , Sr 2+ ) cations. Slight shielding of the 27 Al nuclei is observed when K + is present, resulting in δ obs =60.9 ppm for geopolymer gels activated with sodium silicate compared with δ obs =60.6 ppm for geopolymer gels activated with potassium silicate or a combination of sodium and potassium silicates. This is consistent with the presence of a larger transverse magnetic ield within the local environment surrounding the Al atoms, induced by replacement of some of the charge balancing Na + ions within the irst coordination sphere of Al 3+ with larger charge balancing K + ions. A slight narrowing of the distribution of δ obs and an increase in intensity (between 1 and 3%) of the maximum of this distribution with addition of Sr 2+ and Ca 2+ is observed for all samples, consistent with the slight reduction in the Al/Si ratio observed by 29 Si MAS NMR discussed above. The absence of any other observable variation in the local environment of Al within the (N,K)-A-S-H type gel Fig. 3. 29 Si MAS (B 0 = 11.7 T, ν R =12.5 kHz, shown in black) NMR and 1 H-29 Si CPMAS (B 0 = 11.7 T, ν R =12.5 kHz and Hartmann-Hahn contact period t =1.7 ms, shown in red) NMR spectra for each geopolymer gel and for metakaolin. (For interpretation of the references to colour in this igure legend, the reader is referred to the web version of this article).
after incorporation of alkaline earth cations indicates that this does not alter the local structure of the gel framework to a signiicant extent and suggests that the physicochemical properties (e.g. phase stability, durability, ion transport properties) which are controlled, in part, by gel nanostructure are likely to be unafected.

23 Na MAS NMR
The 23 Na MAS NMR spectrum for each geopolymer gel cured at 20°C (Fig. 4) exhibits a broad resonance spanning from δ obs = 10 to -20 ppm. In geopolymer gels activated with Na 2 SiO 3 or a combination of Na 2 SiO 3 and K 2 SiO 3 this resonance is centred at δ obs = -2.5 ppm and is attributed to Na + ions within a (N,K)-A-S-H type gel, charge balancing Al 3+ ions in fully polymerised q 4 sites (Walkley et al., 2016(Walkley et al., , 2018Duxson et al., 2005c). There is no variation in the lineshape and width of the distribution of δ obs between geopolymer gels activated with sodium silicate or with a combination of sodium and potassium silicates, indicating that K + is not present the local region surrounding Na + ions and therefore suggesting that when present K + displaces some of the Na + ions and independently charge balances q 4 Al sites without disturbing the electric ield gradient surrounding nearby Na + ions (likely due to suicient distance between charge balancing Na + and K + ions). After incorporation of alkaline earth cations, the 23 Na resonance shifts toward lower δ obs values (see inset of Fig. 6), exhibiting a maximum at δ obs = -3.2 ppm in geopolymer gels activated with sodium silicate and δ obs = -3.6 ppm in geopolymer gels activated with a combination of sodium and potassium silicates. This indicates increased shielding of Fig. 4. 29 Si MAS (B 0 = 11.7 T, ν R =12.5 kHz) NMR and b) 1 H-29 Si CPMAS (B 0 = 11.7 T, ν R =12.5 kHz and Hartmann-Hahn contact period t =1.7 ms) NMR spectra and associated deconvolutions for geopolymer gels produced by reacting metakaolin with a) sodium silicate, b) a combination of sodium silicate and potassium silicate and c) potassium silicate. In each case, the data are shown in black and the it (shown in red) is the sum of the deconvoluted peaks. Peaks attributed to Si sites in (N,K)-A-S-H are shown in blue, while those attributed to sites within remnant unreacted metakaolin are shown shaded in grey. (For interpretation of the references to colour in this igure legend, the reader is referred to the web version of this article). Na + , consistent with displacement of Na + and/or K + ions with Sr 2+ or Ca 2+ ions, which are expected to fulil an equivalent charge balancing role. As the charge and ionic size of the alkaline earth and alkali cations difers (Shannon ionic radii 1.02, 1.38, 1.00, and 1.18 Å for octahedrally coordinated Na + , K + , Ca 2+ and Sr 2+ , respectively), substitution of one double valence alkali earth cation for two single valence alkali cations is expected to result in distortion of the local structure of the (N,K)-A-S-H gel framework. However, the absence of any observable variation in the distribution of δ obs in geopolymer gels after incorporation of Sr 2+ or Ca 2+ , regardless of the alkali source, suggests that at the concentrations investigated here both Sr 2+ and Ca 2+ are easily accommodated within the (N,K)-A-S-H gel framework. This can be attributed to the high degree of local disorder inherent in these gels which enables signiicant local variation and distortion in bond angles and lengths (Walkley et al., 2018). Consequently, the physicochemical properties of (N,K)-A-S-H gels that have sorbed diferent proportions of these alkaline earth cations are likely to be equivalent.

Fourier transform infrared spectroscopy
FTIR data provide information regarding the energy of bond vibrations that are occurring within the sample (compared with the measurement of atomic nuclei by solid state NMR), and as such can distinguish between Si-O-Si and Si-O-Al bonds and allow observation (albeit indirectly) of the location of non-bridging oxygen sites (Provis and Bernal, 2014) and the presence of ring structures or carbonate species which may not be readily apparent in the solid state NMR data.
The FTIR spectrum for unreacted metakaolin (Fig. 7) exhibits a high intensity band at 1090 cm −1 that is assigned to asymmetric stretching of Si-O-T bonds (Hajimohammadi et al., 2011;Lee and van Deventer, 2003;Rees et al., 2007) (where T = Al or Si in tetrahedral coordination), and a low intensity band at 810 cm −1 that is assigned to symmetric stretching of Si-O-T bonds (Lee and van Deventer, 2002). The FTIR spectrum for each geopolymer gel cured at 20°C (Fig. 7) exhibits a Table 3 Relative integral areas for Q 4 (4Al) and Q 4 (3Al) environments within (N,K)-A-S-H in each sample, calculated from the deconvoluted 29 Si MAS NMR spectra.

Relative integral area (%) a Site
Na Na-Sr Na-Ca NaK NaK-Sr NaK-Ca K K-Sr K-Ca a The relative integrated intensity for each resonance is normalised to the sum of all sites within the reaction product and is obtained by simulating the 29 Si MAS NMR spectra. Residual unreacted metakaolin component contributions were observed in all geopolymer gel spectra and are shown in Fig. 4 but are excluded from the quantiication of (N,K)-A-S-H gel constituents. Estimated error in the relative integral area is 1%. The isotropic chemical shift (δ iso ) and the peak full width at half maximum (FWHM) were obtained by simulating the 29 Si MAS and 1 H-29 Si MAS NMR spectra and are δ iso = -84.7 and -89.0 and FWHM = 12.7 and 13.0 for Q 4 (4Al) and Q 4 (3Al) environments, respectively.   6. 23 Na MAS NMR spectra (B 0 = 11.7 T, ν R =12.5 kHz) for the geopolymer gels produced using solutions of a) sodium silicate and b) mixed sodium-potassium silicate. The inset in each case shows a magniied region of the spectra.
high intensity band at approximately 1000 cm −1 that is also assigned to asymmetric stretching of Si-O-T bonds (Hajimohammadi et al., 2011;Lee and van Deventer, 2003;Rees et al., 2007), with the shift towards lower wavenumbers (compared with the spectrum for unreacted metakaolin) consistent with the formation of an (N,K)-A-S-H gel. This band is relatively broad, and will contain overlapping contributions from both the aforementioned asymmetric Si-O-T stretching band and Si-O-Si bridge vibrations (Criado et al., 2007). The shoulder at approximately 870 cm −1 is assigned to the asymmetric stretching of Al-O-Si bonds linking AlO 4 and SiO 4 groups (Farmer, 1974;Gadsden, 1975), while the band at 720 cm -1 is assigned to pseudolattice vibrations of small aluminosilicate rings (Handke et al., 1998;Mozgawa, 2001), which have been previously observed in synthetic geopolymer gels (Walkley et al., 2016). A low intensity band at 1650 cm −1 is due to the presence of free and chemically bound water, while the small band at 1380 cm −1 indicates the presence of a small amount of sodium carbonate in some of the samples (Huang and Kerr, 1960) (not identiiable by XRD). The band at 585 cm -1 is assigned to vibrations within 5membered single rings and 6-membered double rings comprised of TO 4 tetrahedral units (Mozgawa, 2001).
The presence of a shoulder at approximately 1080 cm −1 on the main Si-O-T band in these spectra indicates that the geopolymer gels contain some unreacted metakaolin, as identiied by 29 Si and 27 Al MAS NMR and discussed above. There is no signiicant variation in the main asymmetric Si-O-T stretching band when comparing geopolymer gels produced by reaction of metakaolin with sodium silicate, a combination of sodium silicate and potassium silicate, and potassium silicate. The band assigned to asymmetric stretching of Si-O-T bond becomes broader and shifts towards slightly higher wavenumbers after incorporation of alkaline earth cations. This shift indicates an increase in the fraction of Si-O-Si linkages relative to Si-O-Al linkages in the (N,K)-A-S-H gel, indicating a slight decrease in the gel Al/Si ratio (consistent with 29 Si and 27 Al MAS NMR data discussed above).

Implications for immobilisation of alkaline earth radionuclides
The absence of any (in the case of Al) or minimal (in the case of Si and Na) observable variation in the local environments within the (N,K)-A-S-H type gel after incorporation of non-radioactive isotopes of the alkali earth cations Sr 2+ (which may be considered as a structural analogue of 90 Sr) and Ca 2+ indicates that this does not signiicantly alter the local gel structure and suggests that any physicochemical properties (e.g. phase stability, chemical durability, ion sorption properties) that are dictated primarily by the local gel structure are likely to be unafected. This suggests that metakaolin-based geopolymer gels are excellent candidates for production of wasteforms for immobilisation of radioactive waste containing 90 Sr. It also follows that incorporation of alkali or alkaline earth radionuclides of similar ionic size to Sr 2+ (r = 1.18 Å) such as Ba 2+ (r = 1.35 Å) and Na + (r = 1.02 Å) can be expected to result in minimal alteration to the local (N,K)-A-S-H gel structure. Even incorporation of radionuclides exhibiting signiicantly larger ionic sizes such as Cs + (r = 1.67 Å) are unlikely to alter the local gel structure to such an extent as to afect the physicochemical properties of the wasteform (and indeed these may even replace Na + or K + completely (Bell et al., 2009)), as the high extent of local disorder means that the structure can be expected to accommodate signiicant local variation and distortion in bond angles and lengths. Furthermore, the presence of the partially Sr-substituted zeolite A-Na in these sodium silicate activated metakaolin cements cured at 80°C will further increase the binding and immobilisation of Sr (Harnett et al., 2019), and suggests that the proto-zeolitic nature of these cements may be exploited to increase the incorporation of alkali or alkaline earth radionuclides of similar ionic size. Future research focusing on leaching of Sr 2+ and Ca 2+ from geopolymers, long term phase and structural stability, macroscale engineering properties (e.g. compressive and tensile strength) of geopolymers incorporating Sr 2+ and Ca 2+ , and assessment of incorporation of Sr 2+ and Ca 2+ in geopolymers above the trace element levels investigated here will be crucial to further assess the viability of these materials for cementation of radioactive waste containing these radionuclides.

Conclusions
Multinuclear spectroscopic and difractometric analysis show that the disordered (N,K)-A-S-H type gel in metakaolin-based geopolymer gels readily accommodates the alkaline earth cations Sr 2+ and Ca 2+ into their framework structure. Single pulse 29 Si, 27 Al and 23 Na MAS NMR, and 1 H-29 Si CP MAS NMR, FTIR spectroscopy and XRD measurements were used to probe nanostructural changes that may result from incorporation of the alkaline earth cations Sr 2+ and Ca 2+ in metakaolin-based geopolymer gels, and to investigate the inluence of the alkali cations Na + and K + on the alkaline earth incorporation mechanism.
The main reaction product in all geopolymer gels cured at 20°C was a fully polymerised, X-ray amorphous Al-rich (N,K)-A-S-H type gel, while the zeolitic phases faujasite-Na, and partially Sr-substituted zeolite Na-A were observed as additional phases for geopolymer gels cured at 80°C. The (N,K)-A-S-H type gel comprises Al and Si in tetrahedral coordination, with Si in Q 4 (4Al) and Q 4 (3Al) sites, and Na + and K + located within extra-framework sites charge balancing the net negative charge resulting from Al 3+ in tetrahedral coordination.
Upon incorporation of Sr and Ca, the alkaline earth cations displace some of the alkali cations Na + and K + from the extra-framework sites and fulil the charge balancing role. The remaining alkali cations are unafected. Incorporation of alkaline earth cations results in a slight decrease in the Si/Al ratio of the (N,K)-A-S-H gel due to an increased charge balancing capacity resulting from substitution of divalent alkaline earth cations for monovalent alkali cations. Both Ca and Sr induce the same structural changes and cause a slight decrease in Si/Al ratio in the geopolymer gels. No other changes were observed for geopolymer gels cured at 20°C, however in those cured at 80°C incorporation of Sr appeared to promote the formation of LTA (zeolite A) over FAU (faujasite) zeolite phases. The indings presented here have signiicant implications for the long term stability and durability of these materials and suggests that metakaolin-based geopolymer gels are excellent candidates for production of wasteforms for immobilisation of radioactive waste containing 90 Sr.