Introduction

Aqueous sodium silicate solutions are frequent ingredients for the synthesis of zeolites and silica gel based catalysts. Despite many decades of commercial production experience and research, there is still much empirical knack in these syntheses since the transition of dissolved silicate molecules into desired solids is poorly understood. Research efforts for better understanding have intensified recently [1–11], but especially the initial assembly of molecules into nanoparticles is still quite an obscure process.

Raman and FTIR spectroscopy are widely used for distinguishing various siloxane bonds in both aqueous and solid phase hence in principle these techniques could be well-suited for in situ investigation of the silica-solidification processes. However, such studies are surprisingly rare in the literature, which is partly due to the far less developed molecular spectroscopic identification of dissolved silicate structures than that of solid silicates. There are several Raman [12–23] and IR [17,21–28] publications which focus specifically on aqueous silicate solutions and one can also find such spectra in some zeolite synthesis related papers [28–41]. However, only a small group from these authors has assigned certain vibrational bands to specific molecular structures like monomers or siloxane chains, rings, etc. [16–20,26,35,38,41] and even these contain much inconsistency [42–46]. One impediment to more reliable structural interpretations is the lack of relevant model compounds. A particularly important issue is separation of the vibrational bands of monomer silicate ions from those of larger molecules since most theories assume that the formation of crystalline and amorphous silicates starts with the assembly of dissolved monomers into precursor nanostructures [2,47–49].

We report here that 0.2–3 mol/L aqueous solutions of crystalline sodium metasilicate, Na2SiO3, contain almost exclusively monosilicate ions at pH > 10 unlike most other aqueous sodium silicate solutions typically manufactured from amorphous glasses [43]. Thus, these metasilicate solutions are excellent models for studying the IR and Raman spectra of sodium silicate monomers. It will be shown that in contrast to prior assumptions these vibrational spectra depend on the level of sodium ion dissociation, which alternates from about 30% to 80% in this practically important range of silicate concentrations.

Experimental

Anhydrous Na2SiO3 from PQ Corporation (Metso Beads® 2048) and Na2SiO3 × 9H2O from Sigma (>98% purity) were dissolved in deionized water or in 99.9% D2O from Aldrich. Powder XRD measurements indicated that both crystalline silicates had pure orthorhombic phases [50,51]. Every experiment was carried out at ambient conditions at temperatures alternating between 22 and 23 °C with solutions equilibrated for at least 24 h after making any alteration including further dilutions with water or addition of NaOH. Numerous measurements were also repeated after several weeks of storage in closed (but not sealed) containers. Repeated results deviated with less than <±5% from each other unless stated differently.

Vapor pressures were measured with a Wescor Type 5500 osmometer calibrated with both NaCl and NaOH standards. Electrical conductivity measurements were carried out in a YSI Model 32 conductometer calibrated with a 1,000 mS × cm−1 standard. A Thermo Orion Model 720Aplus pH/Ion Selective Meter equipped with matching ROSS electrodes was used for measuring the pH and Na+ concentrations. A Model 86–11 sodium selective electrode was filled with 2 M potassium acetate and calibrated with home made NaOH solutions based on their pH values. The conductivity of Na2SiO3 solutions was consistent with data reported by Kohlraush [52] more than a century ago. Applying the Kohlrausch law or Debye–Hückel–Onsager equation [53], Λ0 = Λ N  + A ( N 1/2, where Λ0 is the equivalent conductivity at infinite dilution, Λ N is the equivalent conductivity at N concentration, A is a constant for a given solute–solvent system, and N is the normality of solutions, we obtained Λ0 = 163 mS × cm−1 × mole−1 which fits well to that reported by Harman [54,55]. These data were used to calculate the α dissociation level using Ostwald’s dilution function [56]

$$ K_{{\text{d}}} \; = \;\frac{{\Lambda _{N} ^{2} \cdot a}} {{\Lambda _{0} (\Lambda _{0} - \Lambda _{N} )}}\; = \;\frac{{\alpha ^{2} \cdot a}} {{1 - \alpha }} $$

FTIR spectra were obtained on a Nicolet Magna 550 spectrometer using a single bounce diamond attenuated total reflectance (ATR) accessory from ASI. A Kaiser HoloProbe Raman spectrometer (200 mW frequency doubled Nd:YAG; −40 °C CCD detector) was used to measure the Raman scattering without polarization. Further details of our spectroscopic equipment and methods have been published elsewhere [57–59].

Results and discussion

Molweight, conductivity, dissociation

We used the osmolality method as reported by Bass et al. [27] to determine the average molecular weight (AMW) of silicates in variously diluted solutions. Since this measurement actually gives the number of total ions per unit solution, it is imperative to know how many free Na+ ions entered the solution from the silicate, i.e., the degree of dissociation, to calculate AMW. Therefore, we determined the Na+ concentrations in every solution both with using a Na+ selective electrode and, since the interaction of silicate with the glass electrode made this method unreliable even after appropriate modification of the salt filling [60], also by electrical conductivity or EC measurements. These two differently measured dissociations were found to correlate well but their numerical values are somewhat different from each other. We could not decide which α is more correct hence use their mathematical average to represent the correct dissociation level of Na2SiO3.

The calculation of AMW along with some characteristics of the sodium silicate solutions can be followed in table 1. Monomers have been identified in such solutions by various authors before [12,24,61], but many studies have suggested dimer or polymer silicon units to be present [24,61–64]. Since silicon is always tetrahedrally bonded to the oxygen [24,65], stoichiometry and charge balance dictates that the composition of silicate ions in the solution of Na2SiO3 must be Na2H2SiO4 when it is not dissociated and H2(SiO4)2− ions must exist in fully dissociated solutions (<0.2 M). Based on data in table 1, other structures like Na2SiO3, Na4H2(Si2O7), H3(SiO4)−, H(SiO4)3−, etc. that have all been assigned to the vibrational spectra of Na2SiO3 [17,18,24] can be neglected especially at our major pH, concentration, and temperature ranges. Note that the bulk of the above-described experiments was done by dissolving Na2SiO3 × 9H2O in water. Several control experiments with anhydrous Na2SiO3 showed identical results within experimental error.

Table 1 Concentration, dissociation, ion size and other parameters of aqueous Na2SiO3 solutions
Full size table

FTIR spectra

Figure 1 compares the FTIR spectra of solid crystalline Na2SiO3 × 9H2O and Na2SiO3 in the Si–O vibrational range. It is known from X-ray structure analysis [51] that the anhydrous sodium metasilicate is built up from chains of [–O–(SiO2)2−–O–] n tetrahedra, each sharing two oxygen atoms with their neighbors (Q2 type siloxane bonds according to the widely accepted NMR nomenclature [66]). Sodium ions compensate the negative charges along the chains. The Na2SiO3 × 9H2O crystals on the other hand contain Na+ ions surrounded by edge-shared, distorted octahedra of H2O molecules. Some vacant coordination positions of these octahedra are occupied by isolated Q0 type HO–(SiO2)2−–OH tetrahedral that compensate the positive charges of sodium ions [50].

Figure 1
figure 1

FTIR ATR spectra of crystalline Na2SiO3 and Na2SiO3 × 9H2O

Full size image

The Na2SiO3 spectrum in figure 1 is quite similar to that reported by Gaskell [67], but we found only one, extremely low resolution IR spectrum for the hydrated metasilicate in the literature [17]. In table 2, we assign the observed IR vibrations of these two crystalline metasilicates to various Si–O vibrations based on their molecular geometry and published calculations for sodium silicate glasses [79–81] and gels [14,68,69,82–84]. Pertinent data of certain crystalline silicates like quartz [14,70,71,85,94,95] and siliceous zeolites [72,73,96] are also referred in the table. A vast number of silica related spectroscopic study demonstrates that the stretching Si–O vibrations (ν)appear at higher wavenumbers than the less energetic deformation vibrations (δ) and asymmetric vibrations (νas, δas) have usually also higher energies than their symmetric pairs (νs, δs). The frequencies (wave numbers) of oxygen motions are also higher when nothing (Si–O−) or only a proton (Si–OH) is attached to the oxygen than the frequencies of oxygens connected to Na+ ions (Si–ONa) or that of the bridging oxygens that connect [SiO4] tetrahedra with each other.

Table 2 Assignments of IR and Raman vibrations of solid, crystalline sodium metasilicates
Full size table

The distinct 1165 cm−1 band of the Q0 type Na2SiO3 × 9H2O silicate (figure 1) is rather surprising. Many researchers associate this band position with νas Si–O–Si bridging vibrations in Q3 or Q4 type silicates [18,25,36,71]. In contrast, a band near this position has been observed in various zeolites and assigned to “intra-tetrahedral” or “localized” νas Si–O vibration as opposed to “inter-tetrahedral” (bridging) or “delocalized” Si–O vibration [68,72,73,96,97]. A longitudinal-optic-transverse-optic (LO-TO) frequency splitting of the vibrational modes of Si–O with one component appearing near 1200 cm−1 can also be considered [72,74,98], but one would rather expect this in the spectrum of the Q2 type dehydrated metasilicate in which such band does not show up (figure 1).

The 1165 cm−1 vibration never appears in the FTIR spectra of the aqueous solutions of Q0 type silicates (figures 2–4) regardless of Na2SiO3 or Na2SiO3 × 9H2O origin. The characteristic changes in figure 2 must be related to changes in the dissociation of dissolved Na2H2SiO4 since there is no other significant molecular difference between the solutions. Specifically the shift of the 989 cm−1 band to 1022 cm−1, the disappearance of 931 cm−1 and the appearance of 883 cm−1 bands, as well as the increase of the 420 cm−1 band should reflect changes in the dissociation level.

Figure 2
figure 2

FTIR ATR spectra of aqueous solutions of Na2SiO3 normalized to bands near 1000 cm−1. Spectra vary with level of dissociation

Full size image

Based on these observations and considerations described in connection with the solid silicates, we assigned the major IR bands of dissolved metasilicates as shown in table 3. The ν1 → ν2 shift of Si–O(H) vibration near 1000 cm−1 caused by the dissociation of Na2H2SiO4 can be estimated using the ν1/ν2 = (μ2/μ1)1/2 formula [99] in which μ1 and μ2 are the reduced masses of the OH group and the rest of the molecule, respectively. Assuming that 1020 cm−1 corresponds to the totally dissociated H2SiO4 2−, the calculated values for NaH2SiO4 − and Na2H2SiO4 are shown in table 3. Considering the dissociation results from table 1, these data match nicely the measured values in figure 2. There is no measurable difference if Si–O− group is considered instead of the Si–O(H).

Table 3 Assignments of IR and Raman vibrations of dissolved Na2SiO3
Full size table

We attempted to fabricate Na4SiO4 orthosilicates by adding two mole NaOH to each mole Na2SiO3 in the variously diluted sodium metasilicate solutions. This attempt resulted in precipitation of solid silicate within a few hours from the 3 mol/L solution made from dehydrated Na2SiO3 and within about 1–2 days from the 3 mol/L solutions made from Na2SiO3 × 9H2O. We believe that this effect is due to the over-saturation of solvent with solutes (salting out effect), which occurs slower in the latter solutions with more H2O delivered with the crystals. One can reason that about 10–12H2O molecules form a hydrate sphere around each Na2H2SiO4 molecule and roughly two spheres are needed to keep the silicate in solution. The amount of water (∼55 moles/L) is “just enough” to do this job in a 3 mol/L solution of Na2SiO3 and the extra water from Na2SiO3 × 9H2O helps a lot. Water supply becomes a premium when the ions of added NaOH have to be also hydrated hence the silicate starts to polymerize. There was no NaOH induced precipitation from the 2 mol/L or more diluted solutions.

As figure 3 illustrates, the added NaOH did not generate any new IR bands. The suppression of 887 cm−1 band, increase of 934 cm−1 band, and red-shift of the 1018 cm−1 band suggest that NaOH shifts the dissociation of Na2H2SiO4 backward, from approximately H2SiO4 2− to NaH2SiO4 − in this case.

Figure 3
figure 3

Added NaOH retards the dissociation of Na2H2SiO4 monomers thus νas ∼ 1018 cm−1 of H2SiO4 2− shifts to lower wave numbers and νas ∼ 934 cm−1 reappears (see text and Table 3 )

Full size image

To see which IR bands are associated with OH vibrations we replaced H2O with D2O as solvent for Na2SiO3. Our attempts to make stable 3 and 2 mol/L solutions in D2O were not successful but the 1, 0.5, and 0.2 mol/L solutions proved to be stable for at least 8 weeks. Figure 4 compares the FTIR spectra of the 1.0 mol/L Na2SiO3 solutions in D2O and H2O. The increased number of relatively sharp bands and the appearance of 1170 and 1230 cm−1 bands in the optically clear D2O solution hint on the presence of nanoscale crystalline silicates [72–74]. Only the 438 cm−1 band shifted to 426 cm−1 (calculated 427 cm−1) and the extremely weak 769 cm−1 band to 737 cm−1 (calculated 750 cm−1) that might be associated with the OH→OD exchange. One would expect for the 1001, 934, and 881 cm−1 bands shifts to about 977, 911, and 865 cm−1, respectively, if they were sensitive to the OH→OD exchange. The lack of such shifts is probably due to the fact that these stretching vibrations are mostly related to motions of the small Si atom.

Figure 4
figure 4

FTIR ATR spectra of Na2SiO3 dissolved in H2O and D2O. Only 769 and 438 cm−1 bands shift in D2O in which band sharpening and 1170 and 1232 cm−1 vibrations hint on possible polymerization

Full size image

Raman spectra

Numerous papers report Raman spectra of anhydrous crystalline, amorphous, or melted Na2SiO3 [67,75,86,89,91–93,100]. These spectra are quite similar to each other and strongly resemble the spectrum of Metso Beads® 2048 in figure 5 indicating that these materials contain the same major structural units (silica chains, charge compensated with Na+ ions) regardless of their crystallinity. We have not found Raman spectrum for hydrated metasilicate in the literature. Its broadened Raman bands hint on the vibration-retarding and bond-distorting effect of hydrogen bonded H2O molecules, similar to the line broadening effect of silanol defects on zeolites [59].

Figure 5
figure 5

Raman spectra of crystalline Na2SiO3 and Na2SiO3 × 9H2O

Full size image

Proposed chemical bond assignments of the Raman vibrations of these solid silicates are shown in table 2. Most Raman study in the literature is directed toward the understanding of three dimensional (3D) silicate networks with Q3 and Q4 type [SiO4] tetrahedra and the published band assignments vary even more widely than the assignments of IR bands. For example vibrations near 1145 cm−1 are typically associated with 3D silica networks [18,36,69,72,74–78,93] which cannot be the case in our metasilicates. In contrast, most investigators identify a Raman band near 975 cm−1 with a chain like Q2 silicate structure [14,69,75–78,86–89,91,92] like that of Na2SiO3 (figure 5, table 2). Likewise, bands near 764 and 706 cm−1 have been associated with Si–O stretching [88] though a δ Si–O–Si vibration has also been assigned to 769 cm−1 Raman shift in certain aluminosilicates [90]. A band in the 910–940 cm−1 region that also appears in various calculations of spectra of a fully dissociated, Q0 type orthogonal [SiO4]4− monomer ion [14,64] has been assigned to νas Si–O [14,64,88], δ O–Si–O [88], and Q3 type ν Si–O–Si [75,79,87] vibrations. A Raman band near 600 cm−1 has been associated with νs Si–O–Si vibrations [14,20,29,32,72,79,88,101] and it is a known benchmark of planar 3-fold Si–O rings [14,72,77,102–105] in various silicates. This band was also assigned to O–Si–O bending in defect sites of molten and glassy alkaline silicates [77,93] and Si–O stretching vibrations in solid sodium metasilicate [88]. Raman shifts at around 460 and 270 cm−1 are also characteristic calculation results for δO–Si–O vibrations in [SiO4]4−; [14,72,89] but only the latter one has been reported in the experimental Raman spectra of solid or molten Na2SiO3 [77,88,92]. Except for the 347 cm−1 band, the other low wavelength vibrations in table 2 appear in various sodium silicate related papers and have been assigned to different deformation motions of the O–Si–O bonds [86,88].

Figure 6 indicates dramatic changes in the normalized Raman spectra of variously dissociated Na2H2SiO4. The 606, 931, and 1010 cm−1 bands must reflect sodium related vibrations in the 3 M solution. Published SiO4 4− related calculations [14,72] predict a νs Si–O band at 775 cm−1 and a ν O–Si–O band at 460 cm−1. The 935 cm−1 disappears from spectra of highly dissociated Na2H2SiO4 hence can be assigned to νas (H)O–Si–O(Na) vibration [we will show later why not (Na)O–Si–O(Na)]. The νas (X)O–Si–O(X) vibrations (X can be negative charge or hydrogen atom) in the 1000–1030 cm−1 range are less intense than those in the FTIR spectra which is a well-known common phenomenon in silicate spectroscopy. Assuming that 1032 cm−1 corresponds the fully dissociated H2SiO4 2−, we calculated the expected shifts in a similar manner as described in the previous chapter and show in table 3. Agreement with the experimental data in figure 6 is quite good.

Figure 6
figure 6

Raman spectra of aqueous solutions of Na2SiO3 and Na2SiO3 × 9H2O normalized to their 780 cm−1 band which splits in 1.0 and 0.5 M solutions of the hydrated silicate

Full size image

Although Alvarez and Sparks [11] assign the 1068 cm−1 band to Si–O–Si stretching vibration, we found similar to Twu et al. [28] that this band coincides with a Raman scattering of Na2CO3 that can be present as contamination in these solutions. Therefore, we ignore the assignment of this band.

As mentioned before the 1236 cm−1 band might indicate some crystallites in the 1 and 0.2 M solutions, but its intensity is surprising because more than 90% of the dissolved silicate is monomer (table 1) and the relative intensity of this band is usually low even in fully crystalline silicates [59]. Another anomaly is the split of νs ∼ 780 cm−1 band in the 1 and 0.5 mol/L solutions of Na2SiO3 × 9H2O (figure 6) which might also be related to LO–TO splitting as Galeener’s [113,119] and van Santen’s [53] calculations for vitreous silica predict. However, such splits never appeared in solutions made from the dehydrated Na2SiO3 and the Raman spectra of the two types of solutes were virtually identical with each other at other concentrations. The spectrum of 0.05 M solution of fully dissociated H2SiO4 2− starts to resemble the spectrum of crystalline Na2SiO3 × 9H2O (figure 5), a possible indication for ordering as a prelude to solidification. Note however that the 1226 cm−1 band is broad and relatively small in this spectrum.

Most authors assign the 606 cm−1 band to Q2 type linear or 3-fold ring structures in aqueous silicate solutions [14,18,22,36,72], but it cannot be the case with our monomer solutions. This band tends to disappear in more diluted solutions hence must be associated with the dissociation of Na+ ions. An example in figure 7 illustrates that, unlike the 930 cm−1 band, the 606 cm−1 band could not be recovered by adding NaOH to the dilute solutions. Therefore we assign this latter Raman shift to (Na)O–Si–O(Na) and the former one to (Na)O–Si–O(H) vibrations. This assignment agrees well with the fact that a band near 600 cm−1 appears only in the Raman spectrum of dehydrated crystalline Na2SiO3 in which direct Si–O(Na) connections exist while a band near 930 cm−1 appears only in the spectrum of the hydrated crystalline Na2SiO3 × 9H2O in which Si–OH also exists.

Figure 7
figure 7

Added NaOH retards the dissociation of Na2H2SiO4 monomers thus the νs ∼ 930 cm−1 (H)O–Si–O(Na) Raman band reappears but the δ s ∼ 606 cm−1 (Na)O–Si–O(Na) band does not

Full size image

Figure 8 shows that D2O also had some unexpected effects on the Raman scattering of these solutions. The increased 606 and 936 cm−1 bands hint on restricted dissociation of Na2H2SiO4. Only the 780 cm−1 band shows the expected shift to 761 cm−1. Freund [16] reported similar result with H2SiO4 2− solutions. The most unusual, but repeatable result is the extreme large 1210 cm−1 band in the 1 M solution. We do not have an explanation for this phenomenon. D2O itself does not give any Raman signal in the 400–1400 cm−1 spectral range.

Figure 8
figure 8

Raman spectra of Na2SiO3 dissolved in H2O and D2O. Large 607 and 936 cm−1 bands indicate retarded dissociation in D2O which promotes extreme 1209 cm−1 band with dilution

Full size image

Conclusions

  1. (1)

    Aqueous solutions of Na2SiO3 and Na2SiO3 × 9H2O contain mostly Na2H2SiO4 monomers at concentrations from 0.2 to 3 mol/L. The dissociation of these monomers changes from about 30% to 80% with dilution in this concentration range.

  2. (2)

    The Si–O vibrations in the FTIR and Raman spectra of these silicate solutions depend strongly on the dissociation level of Na2H2SiO4. Thus the common practice of neglecting dissociation can lead to serious misinterpretation of the spectra of silicate solutions.

  3. (3)

    Based on known structural data and studies on the effect of dilution, NaOH addition, and substitution of H2O for D2O as solvent, we propose consistent assignments for the observed IR and Raman vibrations of the above-mentioned two crystalline sodium metasilicates and their solutions as summarized in tables 2 and 3.

  4. (4)

    D2O is a significantly worse solvent for these metasilicates than H2O and retards the dissociation of Na2H2SiO4. Some unique and unexpected effects were observed in the Raman spectra of solutions made with D2O.