The ASR gels were synthesized from the following starting materials: Ca(OH)₂, colloidal SiO₂ (40 % suspension in H₂O), deionized water and NaOH or KOH pellets. The NaOH or KOH was dissolved in the deionized water and the colloidal SiO₂ and the Ca(OH)₂ were then added. Six blends were prepared at room temperature in high-density polyethylene bottles and stored for three weeks in an overhead rotating mixer at a mixing speed of twelve revolutions per minute. After mixing, a portion of the sediment was removed and dried in a vacuum desiccator for seven days. It was not possible to prevent a small degree of carbonation. The sample designations and compositions are listed in Table 1. The exact molar ratios of the synthetic and dried ASR gels were determined by ICP-OES.

The measurements were carried out with a Raman spectrometer Senterra I (Bruker) at the Fraunhofer IBP, Valley, Germany. The Raman spectrometer is equipped with an Olympus BX 51 confocal polarizing microscope. The samples were measured using frequency doubled Nd: YAG laser with a wavelength of 532 nm, a maximum laser power of 10 - 20 mW and 20x objective lens (laser diameter 5 μm). The measuring process was carried out with the software OPUS 7.5. The measurements were performed in a spectral range of 50 - 4000 cm^-1 with a spectral resolution of 2 cm^-1. At least five measurements were carried out at different positions on each sample. During each measurement, a background correction and spike suppression were applied. The measured spectra were processed with using a baseline correction and a smoothing function. The phase assignments of the Raman spectra were performed by comparison with recorded spectra of the databases specID and MSC-RD (14). The Raman spectra were evaluated using the program PeakFit (version 4.06, Systat Software, Inc.). The R² values of the fitted Raman data ranged between 0.99 - 0.999. The settings for the evaluation are listed in Table 2.

The ²⁹Si NMR experiments were performed with a Bruker Avance 300 spectrometer (magnetic field strength 7.0455 T, resonance frequency for ²⁹Si: 59.63 MHz) in MAS mode (magic angle spinning) using the single pulse technique (90° pulse). The samples were placed in 4 mm zirconia rotors and spun at 8 kHz. About 5000 scans were recorded for each spectrum with a repetition time of 45 s. The chemical shifts were set relative to the external standard tetramethylsilane. The signal patterns of the spectra were deconvoluted with Bruker WINNMR software.

Two characteristic Raman peak pattern of the sample KS0.3 are illustrated in Figure 1. The vibrational bands in the ranges 400 - 650 cm^-1 and 850 - 1200 cm^-1 correspond to Si-O-Si and Si-O bonds, respectively. The evaluation of Si-O-Si group reveals one main peak at 533 cm^-1 for Q³ sites (17). The small peak at 431 cm^-1 is characteristic for SiO₄ tetrahedra vibration of Q⁴ sites and the peak at 462 cm¹ corresponds to symmetric bending (SB) vibration of O_NBO-Si-O_NBO (15, 16). At 597 cm^-1, there is one SB vibration band, which is attributed to Q² sites (17). The dominant peak for Q³ silicate sites at 1040 cm^-1 corresponds to symmetric stretching (SS) vibration. The SS vibrational peak at 990 cm^-1 is linked to Q² sites. The flat and broad band at 897 cm^-1 is related to the symmetric stretching vibration of Q¹ sites (12, 17). The peak position at 1065 cm^-1 can be linked to CO₃ ^2- vibration (12, 20). The peak position at 1108 cm^-1 can be assigned to Q⁴ sites (21).

The ²⁹Si NMR spectrum of the potassium silica gel (KS0.3) is shown in Figure 2. Four main peaks are apparent and attributed to different tetrahedral Qⁿ environments for Si in silicates. The main fitted peaks are centred at -80 ppm, -88 ppm, -97 ppm and -107 ppm and assigned to Q¹, Q², Q³ and Q⁴ sites, respectively (8, 22, 23). The two smaller resonances at -89 ppm and -99 ppm can be assigned to Q² and Q³ sites (24). The fitted relative integral intensities are about 49 % for Q³, 19 % for Q², 16 % for Q⁴ and 4 % for Q¹.

The evaluation of the ²⁹Si NMR spectrum of KS0.3 agrees with the band assignments of the Raman spectrum. The four Q units can be determined by both measurement techniques.

Figure 3 shows the Raman spectra for sodium silica gels with different Na/Si molar ratios. All recorded Raman spectra exhibit a strong vibrational band in the range of 2500 - 3750 cm^-1. The band position around 3420 cm^-1 corresponds to OH stretching vibration of H₂O (18). The results for KS0.3 and NS0.3 are almost identical. Therefore, the results for NS0.3 will not be discussed further. The main Raman peak of the sample NS0.6 in the Si-O-Si range (400 - 750 cm^-1) appears at 600 cm^-1 and is linked to SB vibration of Q² sites. There is a peak shoulder at 548 cm^-1 which is related to SB vibration of Q³ sites. In the higher frequency region, there is a pronounced peak at 1042 cm^-1 which is attributed to SS vibration (Q³) (25, 26). The peak position at 1065 cm^-1, dominant for samples NS0.6 and NS1.2 which is linked to CO₃ ^2- vibration, indicates the presence of trona [Na₃H(CO₃)₂·2H₂O] (12, 20). The Raman spectrum of the sample NS1.2 contains two main peaks in the lower frequency region: one peak at 600 cm^-1 related to SB vibration of Q² units and one peak at 450 cm^-1 linked to SB of O_NBO-Si-O_NBO (15). In the higher frequency region, a pronounced peak occurs at 776 cm^-1 that is attributed to an internal deformation vibration of Si-O. Another peak is found at 928 cm^-1 (SS of Q¹) (12, 17).

On comparing the Raman spectra of the sodium silica gels the following trends were determined. The dominant peak in the low frequency region changes from 530 cm^-1 (Q³) to 600 cm^-1 (Q²) with increasing Na/Si molar ratio. The dominant vibrational band at 530 cm^-1 for the sample NS0.3 disappears in the case of the sample NS0.6 and a shoulder at 548 cm^-1 (Q³) remains. This shoulder linked to Q³ sites then disappears in the spectrum of NS1.2.

It is also apparent that the Raman peak around 450 cm^-1 becomes more prominent with increasing Na/Si molar ratio. This indicates an increasing amount of non-bridging oxygen (NBO) and hence a decreasing degree of cross-linking in dependence of the sodium content. The internal deformation vibration band of Si-O at approximately 780 cm^-1 increases in intensity. Another indication for the decrease in silicate linkage with increasing Na/Si ratio is the intensity increase of the Raman band around 930 cm^-1 (Q¹). The vibrational band at 430 cm^-1 (Q⁴), similar to that of KS0.3 in Figure 1, is not present in the Raman spectra of the samples NS0.6 and NS1.2.

The main vibrational band in the high frequency region in the spectrum of sample NS0.3 at 1040 cm^-1 (Q³) disappears with increasing Na/Si molar ratio; only a sharp Raman peak remains at 1065 cm^-1, indicating the presence of trona, in the case of the sample NS1.2. In the Raman spectrum of the gel sample NS0.6, a shoulder, corresponding to SS vibration of Q³ sites, is still apparent. This shoulder disappears in the spectrum of NS1.2 and only a shallow peak shoulder at 994 cm^-1 (Q²) remains. The peak at 928 cm^-1 (Q¹) emerges as new prominent Raman band. Accordingly, the high-frequency prominent Raman band shifts with increasing Na/Si molar ratio to lower wavenumbers. The carbonate band around 1065 cm^-1 is sharper at higher Na/Si molar ratios.

In Figure 4, the position of the prominent peak of the Raman spectra with varying (Na,K)/Si molar ratio is shown including results of synthetic ASR gels of Balachandran et al. (12). A linear trend between the Q³ peak position in Raman spectra and the (Na,K)/Si molar ratio is apparent (Figure 4, left). With increasing (Na,K)/Si molar ratio from 0.2 - 0.8 the peak position of Q³ shifts from 520 cm^-1 to 565 cm^-1. In Figure 4, the relationship between the prominent peak positions and the (Na,K)/Si molar ratios is shown. The prominent peak position of the sodium and potassium silica gels changes from Q³ to Q² with increasing (Na,K)/Si molar ratio.

The Raman spectra of the sodium-calcium silica gels are compared with a Raman spectrum of a C-S-H phase in Figure 5. The C-S-H phase was synthesized in a previous study with a Ca/Si molar ratio of 0.8 (C-S-H_0.8) (27). Both synthetic ASR gels exhibit a vibrational band at 780 cm^-1 which corresponds to an internal deformation vibration of Si-O (12). The sharp vibrational band at 1065 cm^-1 indicates the presence of trona (20).

The characteristic vibrational bands of the sodium-calcium silica gel NSC0.3_0.22 in the low frequency region are at around 600 cm^-1 and 660 cm^-1 (12, 19). The exact assignment of the peak at 600 cm^-1 is not possible. This Raman peak can be linked to SB of Si-O-Si in Q³ or Q² sites. The present ²⁹Si NMR spectroscopic studies indicate that Q³ sites dominate the structure of the sample NSC0.3_0.22. Based on the relative integral intensities, the NMR Q³:Q² ratio is roughly 5:3. According to studies (1, 16) on C-S-H phases, the Q³ peak lies within a range of 594 - 628 cm^-1. The peak around 660 cm^-1 is at a characteristic position for chain silicates as in the case of C-S-H phases (18). Evaluation of the vibrational band reveals a peak around 584 cm^-1, which is presumably related to Q³ sites. The high frequency region contains a peak shoulder at 983 cm^-1, linked to SS of Q² sites and a peak at 1027 cm^-1, which probably linked to Q³ sites. Due to overlap of the prominent trona vibrational band, band assignment is difficult here. The prominent peak of the sodium-calcium silica gel NSC0.7_0.46 is at 670 cm^-1. This peak position is linked to SB of Q² sites. A shoulder around 600 cm^-1 is linked to SB vibration of Si-O-Si and presumably due to Q³ sites. A broad Raman peak around 772 cm^-1 is explained by an internal deformation vibration of Si-O. In the higher frequency region, two shallow peaks occur at 932 cm^-1 and 996 cm^-1, which are attributed to SS vibration of Q¹ and Q² sites, respectively (12, 16).

As mentioned above, a synthetic C-S-H phase with a Ca/Si molar ratio of 0.8 was measured for comparison purposes. The dominant vibration bands are at 669 cm^-1 and 1006 cm^-1 corresponding to SB of Si-O-Si and SS of Si-O, respectively, in Q² sites (18). The origin of the vibrational Raman band at 798 cm^-1 is not certain. According to Black et al. (19), this peak is due to the translation motion of Si compared with O in Q⁴ units. The Raman peak at 890cm^-1 is related to SS of Si-O in Q¹ units. The vibrational band at 1081 cm^-1 is linked to SS of CO₃ ^2- and indicates the presence of calcium carbonate (15).

The comparison of the two different sodium-calcium silica gels shows a shift of the dominant Raman peak in the low frequency region from 600 cm^-1 to 670 cm^-1. In the Raman spectrum of NSC0.7_0.46 the vibration band at 600 cm^-1 almost disappears and a dominate peak at 670 cm^-1 emerges. In the high frequency region, the dominant peak around 1030 cm^-1 shifts to lower wavenumbers around 1000 cm^-1 with increasing Ca/Si molar ratio (from 0.22 to 0.46). It is concluded that the main vibrational band position of the sodium-calcium silica gels in the Raman spectra shifts from a Q³ position with increasing calcium content towards a Q² position. The shifts of dominant Raman bands with changing Ca/Si molar ratio point to a structural change in the ASR gel. In particular, the structure of the ASR gel becomes more like that of C-S-H phases, a change which has also been suggested by other authors (28-30).

²⁹Si NMR spectroscopic investigations on synthetic ASR products by Hou et al. (30) and Leemann et al. (2) have shown that in the presence of calcium new Raman bands occur corresponding to Q¹ and Q² sites as in C-S-H phases. The Raman spectra of sodium-calcium silica gels and the C-S-H phase both possess bands around 670 cm^-1 and 1000 cm^-1. Differences in the spectra are for the band position at 800 cm ^-1 (in C-S-H gel) and 780 cm^-1 (in N-S-C gel). The Q¹ position of the C-S-H phase is at around 890 cm^-1, whereas this band occurs at 930 cm^-1 in the case of N-S-C gels.

Figure 6 shows the ²⁹Si NMR spectrum of the sample NSC0.7_0.46. The characteristic resonances at -79 ppm, -84 ppm, -88 ppm and -95 ppm correspond to Q¹, Q² _(I), Q² _(II), and Q³_(I) sites of silicate tetrahedra (22, 31). The resonances at -99 ppm and -107 ppm can be assigned to Q³ _(II) and Q⁴ sites (24). The assignments of the two different resonances for Q² and Q³ can be explained by different chemical environments. Depending on the =Si-O counter ion (e.g. Ca²⁺, Na⁺), a chemical shift occurs due to different shielding effects (32). The resonance at -84 ppm (Q² _(I)) can also indicate that a C-S-H phase is present (33). However, in this study there is only one peak for pairing or bridging positions, the second peak for a C-S-H dreierkette is missing here. This suggests that the ASR gel is not in coexistence with a C-S-H gel, but becoming as a whole structurally more like that of a C-S-H phase.