Sodium silicate solutions from dissolution of glass wastes. Statistical analysis

M. Torres-Carrasco^a*, J.G. Palomo^b, F. Puertas^a

a. Eduardo Torroja Institute for Construction Sciences (IETcc-CSIC) (Madrid, Spain)

b. School of Building Engineering (EUATM-UPM) (Madrid, Spain)

*mtorres@ietcc.csic.es

1. INTRODUCTIONTOP

The amount of urban and industrial waste has increased the world over in the last few decades. In the late twentieth century, urban waste was treated primarily in incineration plants, a technology that raised serious environmental problems, primarily associated with the emission of carbon dioxide and toxic particles. This situation led to the implementation of waste management and sorting policies in most highly developed countries. Hence the importance attached to recycling and valorization of all manner of industrial waste and by-products, to enhance environmental protection.

Waste glass collection and management are increasingly common elements of environmental policy in the developed world. In 1994, approximately 9.2 million metric tons of postconsumer glass was discharged in the municipal waste stream in the United States. Approximately 8.1 million metric tons or 80% of this waste glass was container glass (1, 2). In Hong Kong, there are 44 000 t of waste glass generated from domestic sources per year. In addition, there are 20 000 t of waste glass annually generated from commercial sources. Of the total 64 000 t of waste glass, only about 8000 t are recycled and reused although about 50 000 t of waste glass are technically recoverable (3). In Spain, a total of 712 t of glass containers or 15.1 kg of glass waste per inhabitant were deposited in street-side bins in 2010 (4). This statistic is encouraging, bearing in mind that glass container consumption has declined by 5.7% in recent years. In that same year, a further 2 240 kg were taken from sorting plants, the second source of glass waste after street-side bins (4). Moreover, many of these waste glasses are currently disposed of at the landfills which are being filled much faster than expected. If the current trend of waste generation and disposal continues, our landfills will be exhausted by 2015.

Glass comes in many shapes: as bottles, flasks or glazing for windows; flat or round; coloured or clear; and with or without ceramic or metal coating. Since its service life is nearly always very short despite its shape, it is typically reused or recycled. Glass waste must meet a series of requirements for reuse in the manufacture of other glass articles, however. The tendency is to collect and sort urban and industrial glass waste by type. Even so, the wide variety of materials and chemical compositions involved renders its reuse by conventional technological processes highly complex. As a result, from 10 to 30% of glass waste is not recyclable for these purposes and alternative valorisation pathways must be sought. All the existing technologies for recycling mixed glass waste involve crushing. The fragments obtained (1–8 mm fraction), blended in the form of powder (scantly reusable in glass manufacture), can be reused in construction in the following applications:

1. as pozzolanic additions to prepare Portland cements (5)

2. in the preparation of vitroceramic composites together with other industrial waste or by-products, such as fly ash, slag and ceramic discards (6, 7)

3. in the preparation of composites with a polymer matrix (vehicle and pedestrian pavements) (8)

4. as the main component in the production of foam glass for thermal insulating materials (9)

5. as a prime material for synthesizing solid sodium silicates or purified silica.

The fifth application, addressed here, has been scantly studied. Very recent research by the authors (10, 11) has shown that sodium silicate-based urban glass waste may serve as an active component in the preparation of alkali-activated slag or fly ash cements. Moreover, the construction industry is presently very keen on developing new cements and construction materials, whose manufacture would be less energy-intensive and entail the emission of less polluting gas (primarily CO₂) than conventional Portland cement manufacture. These cements are obtained by mixing amorphous silico-aluminates such as blast furnace slag, fly ash, metakaolin or volcanic rock, or blends of two or three of these materials, with highly alkaline solutions (NaOH, Na₂CO₃ or alkaline silicate hydrates) (12, 13, 14, 15, 16, 17 and 18). Inorganic polymer binders provide an alternative to traditional cements up to 80% less CO₂ emissions, and are derived from industrial waste materials (19, 20).

Prior studies (21) showed that the predominant factor in the strength of these alkaline cements was the nature of the alkaline activator. Sodium silicate hydrate (waterglass) proved to be most effective. Bearing in mind that urban glass waste is an amorphous material whose chemical composition includes SiO₂ (65–75%), CaO (6–12%), Na₂O (12–15%), Al₂O₃ (0.5–5%) and Fe₂O₃ (0.1–3%), it may be regarded as a potential (waterglass family) alkaline activator for blast furnace slag, fly ash or other aluminosilicates.

Soluble sodium silicates have been used in industry for a wide variety of applications. The raw materials for sodium silicate manufacture are soda ash and sand. Heated to 1100–1200°C, the materials fuse and a glass forms upon cooling. This glass sold as anhydrous powder or dissolved in water at the manufacturing plant. The resulting solution, also known as “waterglass”, may be sold as it is, or it may be causticized to form more alkaline solutions (22). Therefore, the process of synthesis of waterglass is costly and entails negative environmental effects because is necessary to reach high temperatures in the decarbonation of Na₂CO₃.

In regard to the chemistry of silica, there are numerous studies concerning the solubility of silico-aluminous materials, polymerization, colloidal and surface properties, etc (23). Key features of the glass, along with transparence, is the high chemical resistance. Due to its good qualities, always takes place more or less interaction between the glass and chemicals. Glasses are attacked in both aqueous acid and alkaline solutions, although the mechanisms of attack and the degree of corrosion are different in each case (24). Apparently, the solubility of these glass materials is high when the pH values are high (25, 26 and 27). From pH 9 to 10.7, there is an apparent increase in the solubility of amorphous silica, owing to the formation of silicate ion in addition to the monomer which is in equilibrium with the solid phase. Above pH 10.7, all the solid phase of amorphous silica dissolves to form soluble silicate, since at higher pH the concentration of Si (OH)₄ is greatly lowered by conversion to ionic species, so that no amorphous solid can remain in equilibrium. Moreover, the effect of temperature was also studied on the solubility of glass (27, 28), where an increase in the temperature favors the dissolution of the waste glass.

The objective of this paper is to study the possibility to generate solutions of sodium silicates (as potential waterglass solutions) by the solubility of different types of waste glass. Three for different dissolution processes have been applied: at room temperature (22°C ± 2°C), at 80° ± 2°C and by mechano-chemical ones. The final goal is to establish the optimal solubility conditions to see in further studies the feasibility of the use of waste glass as potential alkaline activators of materials such as slag or fly ash.

2. EXPERIMENTAL PROGRAMTOP

2.1. Materials and their characterisationTOP

Four types of pre-management glass waste were chosen: clear, green, brown and mixed. These materials were characterised with the following techniques:

• • X-ray fluorescence (XRF): PHILIPS PW 2400 spectrometer fitted with a PW 2540 VTC sample changer; XRF readings calibrated and validated with certified reference materials
• • X-ray diffraction (XRD): BRUKER AXS D8 Advance diffractometer fitted with a Lynxeye super speed RX detector, a 2.2-kW Cu anode and no monochromator. The scanning range, from 5 to 60°, was covered in a 24 minutes period. The instrument was set at 40 kW and 30 mA and the sample was not rotated during scanning.
• • Fourier transform infrared spectroscopy (FTIR): ATIMATTSON Genesis FTIR-TM spectrometer, set to scan frequencies across a range of 4000 to 400 cm⁻¹; solid samples pressed into KBr pellets, using 1.0 mg of sample and 300 mg of KBr.
• • ²⁹Si NMR: BRUKER MSL 400 spectrometer operating at 79.49 MHz. For liquid samples ²⁹Si spectra not applied magic angle spinning (MAS).

The glass samples were crushed, ground and sieved to obtain three particle size fractions: i) <45 μm; ii) 45–90 μm; iii) >125 μm.

Table 1 gives the chemical composition (by XRF) of the glass waste for each particle size chosen for the study (from under 45 to over 125 μm). The XRF values showed that the proportions of the majority oxides (SiO₂, Na₂O and CaO) were similar across the entire range of particle sizes and types of glass. Some minor differences were detected in the trace elements, and specifically Cr, which plays a role in glass hue (with higher levels in green, brown and mixed glass).

The XRD mineralogical characterisation (see Figure 1) showed the glass to be an amorphous material, with low range structural order. Crystalline phases were not detected via XRD, consistent with the highly amorphous nature of the materials used in this study at 2θ values of around 20–40°, an indication of the vitrification of silicates, sodium silicate (Na₂SiO₃·nH₂O) and calcium silicate type phases, along with potassium and barium. The IR spectra for these glass waste (Figure 2) contained a wide and intense band at 900–1100 cm⁻¹ associated with the Si-O (ν₃ (Si-O)) asymmetric stretching vibrations in the SiO₄ tetrahedral groups present in the silicates, and another at 450–470 cm⁻¹ due to the ν₄ (O-Si-O) bending vibrations in the SiO₄ groups, likewise in the silicates. The assignment of the spectrum bands is listed in Table 2. The XRD and FTIR findings were practically identical in the four types of glass waste.

2.2. Glass waste solubility trials in highly alkaline mediaTOP

The variables studied to determine the solubility of the four types of glass in highly alkaline media were: type of process, nature of the activating solution, particle size of the glass waste, temperature and process/solution time.

Glass solubility was determined in the following three processes:

Chemical process, which consisted of magnetic stirring at ambient (22 ± 2°C) temperature and a constant speed, was studied in the four types of glass specified. The solid: liquid ratio was 1 g of glass material per 100 mL of solution (1:100). These conditions were constant for the four types of glass and three particle sizes (<45, 45–90 and >125 μm). Stirring times of 10 min and 2, 4 and 6 h were studied to determine the effect on the time on glass dissolution. The three dissolutions used were deionised water (pH=7), the reference, a NaOH solution (pH=13.8) and a 1:1 NaOH/Na₂CO₃ solution (pH=13.6). The concentration of the two alkaline solutions was 4 M. A total of 144 trials were conducted to test all these variables in the four types of glass.

Dissolution trials for chemical process at 80°C were limited to the brown and mixed glass, based on the findings for chemical process at ambient temperature. All the other conditions remained unchanged: stirring times, glass particle size and the activating solutions (NaOH and NaOH/Na₂CO₃ at the same concentration as in chemical process at ambient temperature). Solution temperature was held steady at 80 ± 2°C. A total of 48 trials were conducted under these conditions.

Lastly, in the mechano-chemical process trials, the glass was ground in a steel ball grinder. In such procedures, the relationship between the size and number of balls and the solid: liquid ratio had an important effect on the solubility process. Several trials were run to establish the optimal conditions. The best grinding performance was found for 5 g of solid in 500 ml of solution and 1 kg of balls (with a size distribution of approximately 2.5% of balls 20 mm in diameter; and 12.3% measuring 15 mm, 29.3%, 9 mm and 56%, 5 mm in diameter). As in the case of chemical stirring at 80°C, only brown and green glass were studied, using the same particle size fractions and stirring times as established for the ambient temperature trials, and the same concentration of NaOH and NaOH/Na₂CO₃ solutions as for chemical process at ambient temperature. A total of 48 trials were conducted.

All the solutions obtained were gravity filtered and brought to a known volume for subsequent ICP-AES laboratory analysis of elements Si, Al, Ca and Mg. The analyses were conducted in a VARIAN 725-ES inductively coupled plasma atomic emission spectrometer.

The objective of studying all these variables was to determine the type of process, type of glass (clear, green, brown or mixed), particle size of the glass waste and type of activating solution that yielded the highest glass solubility. This information is essential to establishing the optimal conditions for obtaining sodium silicates, possible waterglass as family alkaline activators, from glass waste.

2.3. Statistical analysis of the solubility valuesTOP

The experimental findings for each dissolution process were statistically corroborated. Variables and their respective levels were established to obtain a mathematical model able to determine the solubility of all the species present in the glass (Si, Al, Ca and Mg) in a single equation accommodating all the variables. STATGRAPHICS Plus 5.1 software was used in this analysis.

The data were obtained, i.e., the response variables were determined, by running a full factorial model on the experimental findings. Model parameter significance was analysed with the F-test (Fischer’s test). The general criterion applied was to accord significance to factors as well as binary and ternary interactions whose p-values at 95% confidence were lower than <0.05.

Once the significant factors were determined, the equation parameters were estimated and the model initially proposed was validated. Further to the model, the variability found in a response variable is attributed to the sum of a series of terms assigned to each factor studied and its binary and ternary interactions. In addition to the variability attributed to the factors studied, the model envisages a term representing the variability due to experimental error. In the present study, where interactions >3 were assumed to be nil, the equation [1] developed to define the model, a priori, was as follows:

where y_ijkl is the estimated value of the response variable at levels i, j, k and l for the factors studied; μ is a scale factor indicating the mean of the estimated value; A_i, B_j, C_k, and D_l represent the effects of the change in factor values on the response variable; (AB)_ij, (AC)_ik, (AD)_jl, (BC)_jk, (BD)_il and (CD)_kl represent the effect of the binary interactions between factors on the response variable; eijkl, which represents the random measuring error, is independent of the factors and levels studied, subject to the assumption that ε_ijkl ≈ N (0, σ²); and σ² is the variance in the experimental error estimated from the available degrees of freedom, given that the interactions of an order higher than three are regarded as nil.

Table 3 lists the four factors studied in the factorial design for the experiments conducted, along with the associated levels for chemical activation at ambient temperature (for a total of 144 trials). Fewer levels of some of these factors were defined for chemical at 80°C and mechano-chemical processes (see Table 4).

The model was validated by plotting the experimental against the expected values. The model is valid when the points on the graph obtained are randomly distributed, i.e., where no pattern emerges.

3. RESULTS AND DISCUSSIONTOP

3.1. Glass waste solubility under different conditionsTOP

The formation of sodium silicate solutions from the dissolution of glass waste is followed by the content of SiO₂ dissolved from these wastes by the different methods applied. The Na ions content cannot be used because the alkaline solutions used were NaOH or NaOH/Na₂CO₃. The main analytical and statistical results obtained according to the methods used are presented below.

3.1.1. Ambient temperature (22 ± 2°C) processTOP

In most cases where glass is exposed to chemical attack, the medium is aqueous. The mechanism governing glass-water interaction must consequently be the first question addressed. Contact with the aqueous medium induces an exchange of sodium and hydrogen ions. Since the latter are present in water in equilibrium with OH⁻ ions, this exchange governs dissolution (24). As Figure 3 shows, the solubility of the silicon oxide present in the glass rose with stirring time, irrespective of particle size. Nonetheless, solubility was highest when the glass particle size was under 45 microns, for the smaller the particle size the greater the specific surface and consequently the more intense the contact between the solution and the glass. The result is higher solubility.

Highly alkaline attacks are governed by a different mechanism, in which the OH⁻ groups are predominant, further to the following reactions (24) [2]:

Unsaturated ≡Si-O⁻ groups may react with water molecules and form new silanol groups and more OH⁻ groups [3]:

The reaction between the glass and the OH⁻groups always hydrolyses oxygen bridges, partially destroying the network. For that reason, glass is much less resistant to alkaline than acidic media. In alkaline media, glass may be said to be depolymerised, resulting in total destruction of its network and gradual solubilisation.

Figure 4 shows the solubility of the four types of glass in the two alkaline media (NaOH and NaOH/Na₂CO₃) for glass particle sizes of under 45 μm. These findings confirmed the greater solubility of SiO₂ and Al₂O₃ (forming oxides) in alkaline media than in water. In NaOH, solubility was 16 to 43% higher than in water, depending on the type of glass. When the solution pH was raised (using NaOH/Na₂CO₃ or NaOH), the amount of SiO₂ and Al₂O₃ extracted rose substantially, due to network destruction after the Si-O-Si and Al-O-Al bonds were broken. The differences between the two solutions were significant: SiO₂ solubility was 12 to 53% higher and Al₂O₃ solubility 17 to 54% higher with NaOH/Na₂CO₃ than with NaOH (Table 5). However, the difference between both pH values was not very remarkable but in this case; it is possible that other determining factors, such as the presence of a common ion (Na⁺) or the presence of carbonates contributed to dissolve a higher amount of glass (33).

As Figure 4 shows, the Mg and Ca oxides, regarded as network modifiers on the grounds of their electronegative potential values (29), failed to exhibit very high solubility over time. Because of their higher solubility (lower Ca-O and Mg-O bond energy than in Al-O and Si-O), they dissolved from the time of initial contact between the alkaline solution and the glass.

Another interesting finding at ambient temperature was that glass solubility was very low and similar for the four types of glass studied (Figure 4). For that reason, the subsequent high temperature and mechano-chemical trials were conducted using only two types (brown and mixed) of glass.

• Chemical process at ambient temperature: statistical analysis.

The statistical significance for oxide solubility in the four factors studied (see Table 3) and their binary and ternary interactions was found by analysing the F-test results on variance in the experimental findings. Table 6 shows the 95% confidence level p-values obtained for Si, Al, Ca and Mg oxide solubility in the chemical process at ambient temperature trials. Significance was attributed to the factors and binary and ternary interactions with p-values of under 0.05.

The findings revealed that not all four factors studied were statistically significant for all the oxides. For SiO₂, the four factors were found to be statistically significant. The statistical analysis identified the conditions for each variable that optimised oxide solubility (6 h of activation, waste particle size <45 μm, NaOH/Na₂CO₃ solution and clear-mixed glass: see Figure 5). All the binary interactions proved to be statistically significant. The explanation lies in the fact that even the slightest change in the levels of any of the main factors altered the final result. Indeed, in chemical process at ambient temperature the solubility values were so low that any change would be highly significant. This explanation is equally applicable to the ternary interactions (ABC, ABD and BCD).

Glass type, stirring time and the type of activating solution (A, C and D, respectively) were the significant factors for aluminium and magnesium solubility but particle size (factor B) was not significant for either of these compounds (see Table 6). In all three of these oxides, the energy in the bond between the oxide and the respective ion (Al³⁺, Ca²⁺ or Mg²⁺) was lower than between oxygen and Si. As a result, their greater solubility was specific surface-independent, lessening the role of particle size. One of the statistically significant binary interactions, stirring time-type of solution (interaction CD), was particularly prominent in the three cases. The inference is that regardless of the oxide used, its solubility always depended on these two factors, although the values varied with the respective levels (10 minutes, 2, 4 or 6 hours for stirring time and water, NaOH or NaOH/Na₂CO₃ for the type of solution). Interaction AD (glass type-activator type) also proved to be important, corroborating the above finding to the effect that NaOH/Na₂CO₃ induced the highest solubility in all the oxides studied.

The ternary interactions for Al, Ca and Mg were less significant than for silicon oxide, although interaction ACD (glass type-stirring time-type of activating solution) exhibited fairly high significance values.

3.1.2. Chemical process at 80 ± 2°CTOP

Figure 6 shows the percentage of solubilized Si, Al, Ca and Mg oxides in brown and mixed glass after chemical process at 80°C. As with chemical process at ambient temperature, glass particle size was a key factor in the solubility of silicon oxide, where the higher solubility was reached for particle size lower than 45 μm. Moreover, unlike what occurred at ambient temperature, the effect of temperature causes the particle size was significant for Al₂O₃ and CaO. By contrast, the type of alkaline activator appeared to have a smaller impact at 80°C than at ambient temperature. In the high temperature trial, the solubilisation percentages were on the order of 60% for SiO₂ and 55% for Al₂O₃ in all four types of glass (see Figure 6) and the amount of SiO₂ dissolved similarly in the two types of glass for both alkaline solutions. Nonetheless, the NaOH/Na₂CO₃ solution exhibited slightly better performance, dissolving 5 and 14% more of the SiO₂ present in the brown and mixed glass, respectively. Given the higher solubility of the Ca and Mg oxides, the percentages dissolved were similar to the values found at ambient temperature.

The findings for this type of process showed that temperature is a key variable in glass solubilisation.

• Chemical process at 80°C: statistical analysis.

The four factors studied in this experimental design are listed in Table 4. According to the p-values obtained for process at 80°C (see Table 7), all the main factors were significant for SiO₂ solubility. As above, the binary interactions were observed to follow two patterns, one in the network forming oxides (SiO₂ y Al₂O₃) and the other in the modifiers (CaO y MgO). In the former, the most significant binary interaction was particle size-stirring time (interaction BC). This finding was an indication that the smaller the particle size and the longer the stirring time, the higher was solubility (see Figure 7). As noted earlier, the use of NaOH or NaOH/Na₂CO₃ had a small effect as the type of dissolution was significant when studied separately, but played a less important role in the binary interactions (CD=0.0531 for SiO₂) than observed in chemical process at ambient temperature. Moreover, binary and ternary interactions were less significant here because with silicon and aluminium oxide solubility rates on the order of 60%, a change in any of the variables studied had a lower impact on the end result. This confirmed the key role of temperature in oxide solubility.

3.1.3. Mechano-chemical process at ambient temperature (22 ± 2°C)TOP

The solubility values obtained for the four oxides in this process were very similar to those reported in the chemical process at room temperature. Figure 8 shows a comparative manner the three methods, confirming that chemical process at room temperature and mechanical-chemical dissolve the glass very similarly, while chemical method at 80 ± 2°C, is the most effective from the viewpoint of solubility of the glass.

• Mechano-chemical process at ambient temperature (22 ± 2°C): statistical analysis.

Of the three processes studied, the mechano-chemical procedure yielded the lowest levels of statistical significance, with no binary or ternary interactions in some cases (see Table 8). Contrary to the findings for type of activating solution in the two preceding methods, here factor D was not significant for silicon oxide dissolution, or in any binary interaction. No simple explanation can be given for this development, which may be associated with the grinding conditions and grinder characteristics. The statistical analysis of the other three oxides yielded findings that were not very different from the statistical results for the chemical procedures, although here calcium oxide was the sole compound to exhibit a third order interaction (interaction BCD).

3.2. Statistical analysis of the three different methods appliedTOP

The statistical study performed to take the solubilisation methods used (Table 9) into consideration yielded significant values for the all main factors as well as the binary and ternary interactions (Table 10). As noted earlier, the effect of the process used played an important role in glass waste solubilisation and sodium silicates formation, for both network forming and modifying oxides.

According to the p-values, the solubilisation method used (factor E) was significant in all the oxides studied, both separately and in the binary and ternary interactions. While all the binary interactions were significant for SiO₂ solubility, the interactions involving factor E were the most prominent. All the other main factors studied (A, B, C and D) were significant in binary and ternary interactions with factor E: in other words, regardless of the solubilisation process used, these factors would be altered. This reinforces the importance of temperature in the dissolution of SiO₂ oxide in glass. Aluminium oxide behaved in much the same way as silicon oxide. The solubilisation process was likewise significant for the modifying oxides, although the type of solution used and stirring time prevailed over the other factors.

The graphs in Figure 9 show that the highest solubility values for all the oxides in glass were attained with chemical process at 80°C. Only minor differences were observed between the other two methods, except for aluminium and magnesium oxide solubility. Mechano-chemical process proved to be more effective for the former and chemical process at ambient temperature for the latter.

3.3. Formation of sodium silicate solution from glass waste dissolution. Study by ²⁹Si NMR MASTOP

From the previous analytical findings in this work has been proved and quantified the SiO₂ contents dissolved from glass wastes. The main objective of this work was to generate solutions of sodium silicates (as potential waterglass solutions) from the different solubilisation process. Figure 10, in turn, reproduces the ²⁹Si NMR spectrum for the original glass (mixed glass) and the spectrum for the liquid obtained after treating the glass waste with NaOH/Na₂CO₃ at 80 ± 2°C for 6 hours. The spectrum for the waste glass contained a single signal at around −93 ppm, indicative of the presence of the Q⁴ Si units that characterize silica glass. The spectrum for the post treatment liquid, even at short times, exhibited a single signal at around −71ppm, associated with the presence of Q⁰ units, i.e., dissolved Si monomers. This is important because according to the literature [30–32], the effectiveness of Si in waterglass systems rises with declining condensation and polymerization of the molecule in the medium.

Next step is to demonstrate that these sodium silicate solutions formed from glass wastes can be used as alkaline activator of waterglass family in the preparation of Alkali-Activated Materials (AAMs).

CONCLUSIONSTOP

The inferences drawn from the analytical and statistical data are listed below.

ACKNOWLEDGEMENTSTOP

This research was funded by the Spanish Ministry of the Economy and Competitiveness under project BIA2010-15516. Thanks to C. Carrillo for her assistance with the ICP analyses and P. Rivilla for her aid with the laboratory trials.

REFERENCESTOP