The pastes studied within the scope of the present research are composed of different wastes – ceramic residue (CR), fly ash (FA) or ladle furnace slag (LFS); and alkaline activators in solution form – sodium hydroxide (NaOH) or sodium silicate (Na₂SiO₃). The CR (Figure 1a) was provided by a Portuguese licensed waste management operator. Based on the technical sheet, it is part of a batch from demolition operations and, specifically, from the selected floor and wall tiles and sanitaryware furniture. After dried, it was mechanically milled (MCR, Figure 1b), for 32 hours, in a purposely built ball mill, with an increased capacity of 5 kg. The FA (Figure 1c) was obtained from a Portuguese thermo-electric power plant, owned by the company PEGOP, and the LFS (Figure 1d) was supplied by the Portuguese ironwork company Megasa, located in Maia, in the north of Portugal.

The particle size distribution (PSD) of all precursors (CR, MCR, FA, and LFS), presented in Figure 2, were determined by laser granulometry, on a Sympatec Helos BF analyzer, able to measure particles from 0.9 to 175 μm and using Fraunhofer diffraction theory. The analyses were conducted after a 5-min ultrasonic bath, on a water-ethanol solution, for increased particle dispersion. The PSD of the original CR revealed a poorly graded material, with all particles above 0.075 mm (fine/course reference line). This scenario was significantly modified after the milling process, which produced a more well-graded material, with a fine fraction of approximately 35%.

The chemical composition of the original wastes was determined by X-ray fluorescence (Table 1). The main constituents of the CR are silica (70.98%) and alumina (17.87%), with a calcium content of 3.72%. The FA showed a similar composition, with significant silica (56.11%) and alumina (21.44%) contents, although with a lower calcium content of 1.30%. The composition of the LFS contains an important calcium content (49.48%), while its significant silica content (18.79%) is, nonetheless, the lowest of the three wastes.

The mineralogy of the original materials precursors and the subsequent pastes was assessed on a PANalytical X’Pert Pro diffractometer, fitted with an X’Celerator detector and secondary monochromator, using CuKα radiation at 40kV and 30mA. The scans were acquired with Bragg-Brentano geometry, covering a 2θ range between 7 and 85º, with a nominal step size of 0.017º and 100 s/step. In the case of the CR, LFS and resulting pastes, phase quantification was performed on diffraction patterns applying Rietveld (19), with the code Highscore Plus 4.8. The values are presented in weight percentage, in Table 2. For the FA refinement, the code PowderCell 2.4 (20) was applied instead. The amorphous phase was simulated through the Le Bail model (21). Volume percentages were converted into weight percentages using the respective density. Figure 3 presents the X-ray diffractograms, showing that the CR is essentially quartz, although several other minor phases were also detected, as sillimanite, oligoclase, mullite, and anorthite. The main crystalline phases identified in the FA were quartz, mullite, hematite, and graphite. The last one shares the main peak in the same location as quartz, halo between 25 and 27°(2θ). While the LFS presents several crystalline phases, namely calcio, olivine, andradite, calcite, cuspidine, titanite, and larnite, it presented mostly crystalline structures. The slag presents a slight halo, between 27 and 40°(2θ), characteristic of vitreous/amorphous materials (22), while the fly ash and ceramic residue showed a similar halo, although between 2θ ≈ [25 - 30]º.

Sodium hydroxide (SH) and sodium silicate (SS), in solution form, were used as activators throughout this study. The SH was prepared 4 to 6 h before the fabrication of the pastes, with a concentration of 8 molal, using deionized water. The sodium silicate had a unit weight of 1.464 g/cm³ at 20ºC, a SiO₂/Na₂O weight ratio of 2.0 (molar oxide ratio of 2.063) and a Na₂O concentration in the solution of 13.0%.

The composition of the mixtures was defined based on the results of a preliminary test where the variation of the solid/liquid ratio between the precursor and the activator determined the appropriate ratio to obtain better workability of the mixtures. The performance of sodium hydroxide (SH) or sodium silicate (SS) were studied in separate experiments. Each experimental setup included a parallel treatment of both precursors and covered the same alkali activator concentrations (SH or SS). Table 3 presents the studied pastes, including its respective activator/precursor ratio. CR was tested without the addition of an additional precursor (FA or LFS), to assess the possibility of direct activation, which would translate into more economical and less logistical demanding pastes.

The sodium hydroxide was prepared 4 to 6 hours before use because it generates an exothermic reaction which needs is needs to cool down. The activator temperature at the time of the mixing was 20ºC ± 1ºC. A mechanic mixing process followed, for a period of 3 minutes, complemented by manual homogenization, with the help of a spatula (whenever required, the mechanical homogenization period was extended for 2 minutes).

The homogenized paste was then poured into cubic stainless-steel molds and mechanically compacted for 2 minutes. Three cubic specimens of 40 x 40 x 40 mm³ (nominal dimensions) were fabricated for each paste, and for each curing period. The initial curing period lasted 1 day at 70°C, with the pastes still inside the molds. It was followed by 14, 28 and 90 days curing at ambient temperature and humidity 20-23 ºC and around 60% RH, respectively. After measured and weighed, each specimen was then subjected to a uniaxial compression strength test (UCS), on a servo-hydraulic testing machine, using three different actuators, with load capacities of 40, 100 and 120 kN. These tests were carried out under monotonic displacement control, at a rate of 0.6 mm/min, and stress-strain curves were plotted for all tests performed.

The present research was complemented by microstructural analysis, which included Scanning Electron Microscopy (SEM), X-ray Energy Dispersive Spectroscopy (EDX), X-ray diffraction (XRD) and Fourier Transform Infrared Spectroscopy (FTIR). SEM analyses were developed on a FEI Quanta 400 scanning electron microscope, at 30 kV, in low vacuum mode, with partial pressure inside the chamber (1.3 mbar), avoiding the deposition of a conductive layer. An EDS analyzer, from EDAX, allowed a chemical analysis of the elements present in each sample, using the same spectrum acquisition time and a ZAF correction model.

The FTIR spectra were acquired using a Thermo Scientific Nicolet iS50 FTIR spectrometer, resorting to an ATR (Attenuated Total Reflectance) accessory with a diamond crystal. The instrument was controlled by the Omnic software package, version 9.2.28, from Thermo Fisher Scientific Inc. A spectrum range of 4000 to 400 cm-1 was defined, with a resolution of 4 cm^-1 and 64 number of scans by 90 seconds each sample.

The uniaxial compressive strength (UCS) results are presented in Figure 4, showing the data for the pastes prepared with CR and FA, and Figure 5, showing the data for the pastes with CR and LFS. Although the aim of the UCS tests was to assess the effectiveness of the CR, pastes with an increasing amount of FA and LFS (up to, and including, a 100% content for each case) are also presented, so that a threshold could be established, and direct comparison could easily be made. Curing periods of 1, 14, 28 and 90 days are presented, yet it is important to remember, at this stage, that the 1-day curing specimens were exclusively cured at 70ºC, while the remaining specimens were cured for 1 day, at 70ºC, while the remaining curing was developed at ambient temperature (20 - 23ºC).

In general, curing time increased the UCS, regardless of the type of precursor (CR+FA or CR+LFS) or activator (SH or SS). Also evident are the higher strength values obtained with the SS-based pastes, comparatively to the SH-based specimens, irrespective of whether the precursor combination. Finally, the third and most relevant observation is the higher UCS achieved by the CR+LFS precursor combination, when compared with the CR+FA blend.

For the CR+FA combination, activated with SH, the most effective solution was the 50% CR and 50% FA. When SS was used, the proportion 75% CR and 25% FA achieved the highest UCS value after 90 days. However, it is significant, considering that the purpose of the present study is the disposal of the ceramic residue, that the 100% CR paste (which represents the highest possible volume of CR incorporated) showed the highest UCS for the CR+FA combination after 1, 14 and 28 days, and showed only a marginally lower UCS after 90 days. This suggests that the use of FA is not absolutely required, which would represent a significant gain in terms of economic and logistical efforts.

Regarding the CR+LFS combination, the scenario was slightly different for the SH-based pastes, with the 75% CR + 25% LFS representing the higher UCS. Another significant difference was the extent of such strength increase, at the 90-day mark: the addition of 50% FA produced an UCS improvement of 1.9x, relatively to the 100% CR, while the addition of 25% LFS produced an UCS improvement of 4.6x, relatively to the same 100% CR paste. For the SS-based pastes, the CR+LFS scenario was similar to the one described for the CR+FA pastes, i.e. the addition of 25% of the alternative precursor represented the most performing combination for both cases. However, the 75%CR + 25%LFS paste now showed higher UCS values for all curing periods, and not just for 90 days. Not only that, but the difference, at the 90-day mark, between the 100%CR and the 75%CR + 25%LFS was clearly more significant (3.15x) than the difference between the 100%CR and the 75%CR + 25%FA (1.1x).

The pastes with the highest strength values from each set (M1, M2, M3, and M4), corresponding in most cases to the composition 75% CR + 25% (FA or LFS), were selected for microstructural analyses, using XRD, SEM/EDS and FTIR to characterize the cementitious materials developed during their respective reactions throughout the initial 90 days of curing.

Figure 6 shows the XRD diffractograms of the different pastes, 75CR+25FA (SH or SS) (M1 and M2) and 75CR+25LFS (SH or SS) (M3 and M4). The products resulting from the activation reaction present similar peaks which were also found in the initial materials, suggesting that the equivalent phases are essentially inert to the alkaline activation process. In cases where a slight reduction of their intensity is detected, it is due to the material dilution (induced by the blend with the activator). The presence of mullite in M1 (25.1 ± 1.5) and M2 (23.2 ± 1.5) is somewhat higher in comparison to the other two mixtures (see Table 4), fundamentally influenced by the outset constituents, CR and FA with significant Si content (Table 1). Crystalline phases of Mu in the pastes were also displaced in range, from 28 – 30°(2θ) to 30 to 35°(2θ) in contrast to CR and FA XDR diffractograms (Figure 3). These variations are a consequence of the formation of amorphous phases in the aluminosilicate gel, which will partially crystalize, in time (zeolitic precursor) (22–24). The anorthite (A) was identified in the four mixtures since the CR is the major component. However, higher A content was evidenced in M3 (16.5 ± 1.1) and M4 (15.0 ± 1.1) (Table 4) as a result of having a precursor (LFS) with high Ca content. In all blends, the main peak was registered at approximately 26º (2θ), corresponding to quartz, which makes it the most relevant crystalline phase. It also worth to mention that for neither case the Na₂CO₃.H2O (thermonatrite) phase was found, denoting that the Na was retained in the gel, thus not migrating to the surface.

The same four pastes, cured at ambient temperature for 90 days, were analyzed with SEM. The chemical composition was also obtained, with 10 points per image. Figure 7 shows SEM images of selected samples, enabling an understanding of the morphology and general distribution of the different constituents in the mixtures. Base on all the images exhibited, it can be said that the resulting paste is generally compacted, homogeneous and with low porosity. In M1(75CR-25FA/SH) is possible to observe several FA (points A) or CR (points C) unreacted particles covered with reaction products. However, in M2 (75CR-25FA/SS), even after 90 days, an unreacted particle without any signs of Na ‘attack’, can still be detected (points A). It was also observed a slightly higher Ca content in the M2 paste, compared with the M1, which might result from a higher dissolution degree. Regarding the type of gel developed, a N-A-S-H type was detected (point *) when using FA as a complementary precursor, regardless of the activator solution used (SH or SS).

A similar situation can be seen for the 75CR-25FA/SH (M3) and 75CR-25LFS/SH (M4) pastes. Some small LFS (points B) and CR (points C) unreacted particles are distinguished. The most significant aspect of the M4 paste was the Ca heterogeneous distribution throughout the sample, which can be due to the incidence of particles with an important presence of Ca. Concerning the type of gel developed, and when LFS in combination with CR were used, its calcium content naturally affected the composition of the gel, which can be classified as a N, C-A-S-H, if SH acted as activator (M3), or C-A-S-H (points *) (25, 26), when the precursors were instead activated with SS (M4). The general chemical composition of the four studied mixtures is summarized in Table 5, where is presented the average value of the analyzed points and its standard deviation (SD).

Figure 8, shows the FTIR spectra for the starting materials (LFS, CR, and FA) as well as the M1 (75CR25FA), M2 (75CR25FA), M3 (75CR25FA), and M4 (75CR25FA) pastes activated with SH or SS, after 90 days of curing, which present a set of bands centered in the frequency range between [400 - 2000] cm^-1. Different behavior in the spectra of each of the materials is observed, for LFS two main bands are detected at 840 cm^-1 and 1040 cm^-1, which are generally associated at deformation vibration of carbonates (CO₃^2-), C-O, while the 560 cm^-1 band is linked with the Al-O-Al bond and 490 cm^-1 and 440 cm^-1 are associated with O-Si-O connections. The fly ash (FA) evidences a wide main band at 1015 cm^-1, this is associated with Si-O-Si or Si-O-Al bonds, while the 790 cm^-1 and 420 cm^-1 bands are related with Si-O bonds (quartz). In addition, the presence of another band that records at 550 cm^-1 is associated with the octahedral aluminum (Al-O-Al bonds) of the mullite (27), previously distinguished in this material by XRD, (see Figure 3). The CR sample shows a deep and wide main band that appears around 1040 cm^-1. This is typical of asymmetric tension vibration, related to Si-O-Si or Si-O-Al, confirming some crystallinity of the material. The bands at 795 cm^-1, 775 cm^-1, and 420 cm^-1 correspond to O-Si-O bonds (quartz).

The FTIR spectra of the four different mixtures are also presented in Figure 8, where basically the characteristic vibration bands of the main components of the starting materials are shown. At the end of the 90 days of curing, the most significant change was the displacement of the main CR band (1040 cm^-1), for shorter wavelengths, which is attributed to a structural reorganization caused by the formation of reaction products (28). But, in mixtures of the studied proportions of M1 and M2, the displacement was between 995 cm^-1 and 1010 cm^-1, which is related to the formation of the typical N-A-S-H gel, while M3 and M4 shifted around 960 cm^-1, this translation is connected to the presence of LFS in the composition, promoting the formation of N, C-A-S-H type gel.