Study of Eco-Friendly Belite-Calcium Sulfoaluminate Cements Obtained from Special Wastes

: Belite-calcium sulfoaluminate (BCSA) cements are special binders obtained from non-Portland clinkers; they have become increasingly more important due to their environmental impact during the manufacturing process compared to Portland cements, such as lower energy consumption and CO 2 emissions. The aim of this paper was to assess the possible use of titanogypsum (T) and water potabilization sludge (W) to reduce the amount of natural raw materials (natural gypsum and clay, respectively) used in the production of BCSA cements. Three BCSA clinker generating raw mixes, containing T and / or W, and a reference mix based only on natural materials (limestone, clay, bauxite and natural gypsum) were heated in an electric furnace at temperatures ranging from 1200 to 1350 ◦ C. Quantitative X-ray di ﬀ raction (XRD) analysis of the burnt products showed high conversion of reactants towards the main hydraulically active BCSA clinkers components of C 2 S and C 4 A 3 $, particularly at temperatures of 1300 and 1350 ◦ C. Isothermal calorimetric measurements, di ﬀ erential thermal–thermogravimetric and XRD analyses as well as porosimetric measurements showed that all BCSA cements, from mixing the clinkers (at optimum temperatures) with commercial anhydrite, exhibited similar hydration behavior.


Introduction
The cement industry consumes huge amounts of natural raw materials and fuels, mainly fossil fuels and pet coke. It is one of the main contributors to climate change due to greenhouse gas emissions, mainly CO 2 . In 2018, global cement production was about 3.99 billion tonnes, accounting for about 8% of all anthropogenic carbon dioxide emissions [1]. Ordinary Portland cement (OPC), a blend of clinker (about 95%) and gypsum, is the most used binder in the world. OPC clinker is usually obtained by heating a mixture (raw meal) of limestone (L, 80% by mass) [2] and clay (C) in a rotary kiln at about 1500 • C. According to [3], about 0.83 kg of CO 2 is released from the production of 1 kg of OPC; it is generated from both L thermal decomposition (about 60% of the total CO 2 emissions) and fuel combustion [4,5].

Raw Materials and Proportioning of CSA Clinker-Generating Raw Mixtures
The chemical compositions of the natural materials (L, B, C, and G) and the special wastes (T and W) used in this investigation are listed in Table 1. The L, B, C and G samples were supplied by an Italian cement manufacturer together with B. T and W came from an international titanium dioxide pigment company and a water treatment plant of the "Camastra" artificial reservoir located in the Basilicata Region (Italy), respectively. All the samples were first dried in an electric oven at 105 • C until a constant mass value was reached and then, finely pulverized in a Fritsch Pulverisette 6 (FP6) laboratory planetary mill to pass through the 90 µm sieve.
Four BCSA clinker-generating raw mixtures (RM_R, RM_T, RM_W and RM_TW) were investigated: the first (RM_R) was based only on natural materials; the others contained T (RM_T) or W (RM_W) or both T and W together (RM_TW). The composition of raw mixtures was determined according to modified Bogue calculations (MBCs) [35]; it was assumed that C 2 S, C 4 A 3 $, C 4 AF and C$ amounts were, respectively, comprised in the ranges 45-60%, 20-30%, 8-20% and 4-10%. A concentration value of free-CaO (C f ) lower than 1.5% was also considered. In order to obtain reactive belite in the cement clinkers, 1% by mass of B 2 O 3 was added to the raw mixtures.
The proportion of raw meals and mineralogical compositions, estimated according to the MBCs, of the four clinkers (C_R, C_T, C_W and C_TW) are reported in Tables 2 and 3, respectively.

Burning of Raw Mixtures
Four 100 g batches of each clinker-generating raw mixture, placed in covered alumina crucibles to prevent sulfur emissions, were heated in an electric furnace at four different temperatures (1200, 1250, 1300 and 1350 • C, with an accuracy of ±2 • C) according the procedure reported in [8]. The resulting BCSA clinkers (65-70 g) were finely ground using a planetary mill to pass through the 90 µm sieve and then submitted to QXRD analysis in order to assess the best synthesis temperature to maximize the sum of C 2 S and C 4 A 3 $.

Preparation of BCSA Cements
BCSA cements were prepared by grinding the four clinkers heated at the best synthesis temperatures with commercial anhydrite (containing 96.6% by mass of C$) using a mill to pass through the 90 µm sieve. The specific surfaces of the clinkers (according to EN 196-6 (CEN, 2018)) were in the range 4600 ± 100 cm 2 /g. The amount of anhydrite to be added was determined by subtracting the quantity of C$ already present in the BCSA clinkers from the value necessary for the complete reaction of C 4 A 3 $ to ettringite according to the following reaction [15]: Therefore, the resulting BCSA cements were indicated using the same symbols used for their generating raw mixtures (CEM_R, CEM_T, CEM_W, CEM_TW); their mix design is summarized in Table 4.

Hydration Procedure
Cement powders were hydrated using a water/cement ratio equal to 0.5 by mass. The paste samples were cast into small cylindrical molds (15 mm high and 30 mm diameter) and put inside a thermostatic bath (at 20 • C and 95% relative humidity) for aging periods ranging from 4 h up to 90 days.
At the end of each aging period, the specimens were broken in half: one part was submitted to mercury intrusion porosimetry (MIP) characterizations, the other was gently pulverized for DT-TG and XRD measurements. Both components, namely hardened fragments and fine powder, were treated with acetone (to stop hydration) and diethyl ether (to remove water) and finally, stored in a desiccator containing silica gel and soda lime (to ensure protection against H 2 O and CO 2 , respectively) 2.5. Characterization Techniques 2.5.1. X-ray Fluorescence Analysis X-ray fluorescence (XRF) analysis was used to determine the chemical composition of the raw materials. A wave dispersive Bruker Explorer S4 apparatus (maximum power = 1 kW; LiF200, PET, OVO-55, OVO-B as analyzing crystals) was used. Loss on ignition (l.o.i.) was determined according to EN 196-2 [54]. One gram (1 ± 0.05 g) of each material was calcined at 950 ± 25 • C for 15 min; calcination was repeated until a constant weight was achieved.

DT-TG Analysis
A simultaneous DT-TG apparatus (NETZSCH-Tasc 414/3), operating between room temperature and 1000 • C with a heating rate of 10 • C min −1 in 150 µL alumina crucibles, was employed to evaluate BCSA cements' hydration products (at 4 h, 2, 28 and 90 d). TG analysis was also used to quantitatively determine chemically bound water [55] on samples hydrated at 4 h, 2, 28 and 90 d.

XRD Analysis
XRD analysis was performed to determine the mineralogical composition of both the CSA clinkers and the hydrated cement pastes (at 4 h, 7 and 28 d). A Bruker D4 Endeavor X-ray diffractometer using CuKα radiation and a Linx Eye dispersive detector were used. The samples were scanned at a step size of 0·02 • 2θ for 45 min between 8 • and 55 • 2θ. Rietveld refinement of the XRD patterns was performed using the TOPAS 2.0 package.

Isothermal Calorimetry
Isothermal calorimetry, using a Calmetrix n I-cal flex calorimeter, was applied to measure the heat release during hydration of cement pastes (w/c = 0.5) up to 14d. Measurements were taken at room temperature. In total, 20 g of each cement paste was externally mixed for 30 s using a Stuart SA 6 variable speed vortex mixer with a mixing speed of 2500 rpm.

Mercury Intrusion Porosimetry (MIP)
The porosity measurements were taken using a Thermo-Finnigan Pascal 240 Series mercury porosimeter (maximum pressure, 200 MPa) equipped with a low-pressure unit (140 Series) able to generate a high vacuum level (10 Pa) and operate between 100 and 400 kPa. MIP measurements were carried out on fragments hydrated for 7, 28 and 90 days. MIP was employed to better understand the properties of the hydrated binders through the evaluation of microstructure development over time.

Synthesis of BCSA Clinkers
The main BCSA clinkers mineralogical phases (taken at 1200, 1250, 1300 and 1350 • C) are displayed in Table 5. From an overall examination of the QXRD clinker data, it could be argued that, at all investigated temperatures, I) C 2 S and C 4 A 3 $ are the main burning products and II) the conversion of reactants is complete; in addition, the best results were obtained at the 1300 • C for RM_W and at 1350 • C for the other mixtures. Furthermore, tricalcium aluminate (C 3 A) brownmillerite (C 4 AF), and periclase (MgO) were observed in all the clinkers as secondary components; anhydrite (C$), and CaO f were identified only as minor components. Table 5. Mineralogical composition of BCSA clinkers determined according to Rietveld, wt%. QXRD results showed that Al 2 O 3 , not involved in the formation of C 4 A 3 $, was more inclined to give C 3 A instead of C 4 AF. In addition, the amounts of sulfates (namely those present in both C 4 A 3 $ and C$) were lower than those estimated using the Bogue equations; this difference was presumably related to the volatilization phenomena involving part of the sulfur oxides. On the whole, these results emphasize the active role played by the sources of silicate, alumina and sulfate supplied by the wastes analyzed (T and W), which reacted as the corresponding natural materials. Figure 1 reports the XRD patterns for the four mixes heated at their best synthesis temperature.

β-C 2 S C 4 A 3 $ * C 3 A C 4 AF Others
Appl. Sci. 2020, 10, x FOR PEER REVIEW 6 of 14 the wastes analyzed (T and W), which reacted as the corresponding natural materials. Figure 1 reports the XRD patterns for the four mixes heated at their best synthesis temperature.  Figure 2 shows the DT-TG thermograms for the four BCSA cements hydrated for periods between 4 h and 90 days. With DT-TG temperature increase, three different endothermal effects were identified at 93 ± 9 °C, 149 ± 9 °C and 277 ± 5 °C, and were attributed, in this order, to the following hydration products [55]: CSH, C6A$3H32 and AH3, respectively. Moreover, no significant effects were found above 500 °C.  Figure 2 shows the DT-TG thermograms for the four BCSA cements hydrated for periods between 4 h and 90 days. With DT-TG temperature increase, three different endothermal effects were identified at 93 ± 9 • C, 149 ± 9 • C and 277 ± 5 • C, and were attributed, in this order, to the following hydration products [55]: CSH, C 6 A$ 3 H 32 and AH 3 , respectively. Moreover, no significant effects were found above 500 • C.

Hydration of BCSA Cements
Appl. Sci. 2020, 10, x FOR PEER REVIEW 6 of 14 the wastes analyzed (T and W), which reacted as the corresponding natural materials. Figure 1 reports the XRD patterns for the four mixes heated at their best synthesis temperature.  Figure 2 shows the DT-TG thermograms for the four BCSA cements hydrated for periods between 4 h and 90 days. With DT-TG temperature increase, three different endothermal effects were identified at 93 ± 9 °C, 149 ± 9 °C and 277 ± 5 °C, and were attributed, in this order, to the following hydration products [55]: CSH, C6A$3H32 and AH3, respectively. Moreover, no significant effects were found above 500 °C.  On the whole, the DT results reveal the presence of C6A$3H32 and AH3 for all the curing periods, apart from the pastes based on C_R and C_TW, where aluminum hydroxide materializes after 2 days of hydration; furthermore, both C6A$3H32 and AH3 concentrations increased with aging time. Calcium silicate hydrates were not always distinguished, as their DT effect was overlapped by the ettringite (E) peak. Monosulfate could not be clearly detected owing to the overlap of the E signal.

Hydration of BCSA Cements
TG curves show the mass loss related to dehydration and dehydroxylation phenomena involving CSH, ettringite and monosulfate (from room temperature to ~200 °C) as well as aluminum hydroxide (from 200 to ~300 °C), respectively. The higher weight loss values up to about 200 °C are On the whole, the DT results reveal the presence of C 6 A$ 3 H 32 and AH 3 for all the curing periods, apart from the pastes based on C_R and C_TW, where aluminum hydroxide materializes after 2 days of hydration; furthermore, both C 6 A$ 3 H 32 and AH 3 concentrations increased with aging time. Calcium silicate hydrates were not always distinguished, as their DT effect was overlapped by the ettringite (E) peak. Monosulfate could not be clearly detected owing to the overlap of the E signal.
TG curves show the mass loss related to dehydration and dehydroxylation phenomena involving CSH, ettringite and monosulfate (from room temperature to~200 • C) as well as aluminum hydroxide (from 200 to~300 • C), respectively. The higher weight loss values up to about 200 • C are related to the formation of the main hydration products, namely CSH and ettringite; moreover, as curing time increases, their amount increases.
The BCSA cements' hydration rates were also evaluated using the amount of chemically bound water (Figure 3) calculated from the mass loss values from TG analyses (Figure 2).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 8 of 14 The BCSA cements' hydration rates were also evaluated using the amount of chemically bound water (Figure 3) calculated from the mass loss values from TG analyses (Figure 2). Figure 3 shows that the four systems followed a similar trend in terms of the evolution of the hydration process. With the exception of the reference system, which displayed the worst hydration behavior, the other binders reached comparable bound water content after 90 days of curing. The hydration process was also investigated using XRD analysis. XRD patterns (for samples hydrated at 4 h, 7 and 28 days) are shown in Figure 4, where only the most significant phases are indicated; they allowed for the detection of: (a) the consumption of reactants, (b) the presence of inert phases, and (c) the development of hydration products (mainly ettringite). The absence of CSH and AH3 among the crystalline phases detected using XRD analysis is associated with their amorphous nature. As for the DT thermograms, it has been found that as curing time increases, the ettringite peaks increase at the expense of both the ye'elimite and anhydrite peaks. Weak peaks for monosulfate were detectable in the systems CEM_R, CEM_W and CEM_TW at 7 days of curing, and at 28 days for the CEM_T systems. Furthermore, CEM_TW hydrated more slowly than the other systems. XRD patterns at 28 days still revealed the presence of some unconsumed reactants such as ye'elimite and anhydrite.  Figure 3 shows that the four systems followed a similar trend in terms of the evolution of the hydration process. With the exception of the reference system, which displayed the worst hydration behavior, the other binders reached comparable bound water content after 90 days of curing.
The hydration process was also investigated using XRD analysis. XRD patterns (for samples hydrated at 4 h, 7 and 28 days) are shown in Figure 4, where only the most significant phases are indicated; they allowed for the detection of: (a) the consumption of reactants, (b) the presence of inert phases, and (c) the development of hydration products (mainly ettringite). The absence of CSH and AH 3 among the crystalline phases detected using XRD analysis is associated with their amorphous nature. As for the DT thermograms, it has been found that as curing time increases, the ettringite peaks increase at the expense of both the ye'elimite and anhydrite peaks. Weak peaks for monosulfate were detectable in the systems CEM_R, CEM_W and CEM_TW at 7 days of curing, and at 28 days for the CEM_T systems. Furthermore, CEM_TW hydrated more slowly than the other systems. XRD patterns at 28 days still revealed the presence of some unconsumed reactants such as ye'elimite and anhydrite.
Appl. Sci. 2020, 10, x FOR PEER REVIEW 8 of 14 The BCSA cements' hydration rates were also evaluated using the amount of chemically bound water (Figure 3) calculated from the mass loss values from TG analyses (Figure 2). Figure 3 shows that the four systems followed a similar trend in terms of the evolution of the hydration process. With the exception of the reference system, which displayed the worst hydration behavior, the other binders reached comparable bound water content after 90 days of curing. The hydration process was also investigated using XRD analysis. XRD patterns (for samples hydrated at 4 h, 7 and 28 days) are shown in Figure 4, where only the most significant phases are indicated; they allowed for the detection of: (a) the consumption of reactants, (b) the presence of inert phases, and (c) the development of hydration products (mainly ettringite). The absence of CSH and AH3 among the crystalline phases detected using XRD analysis is associated with their amorphous nature. As for the DT thermograms, it has been found that as curing time increases, the ettringite peaks increase at the expense of both the ye'elimite and anhydrite peaks. Weak peaks for monosulfate were detectable in the systems CEM_R, CEM_W and CEM_TW at 7 days of curing, and at 28 days for the CEM_T systems. Furthermore, CEM_TW hydrated more slowly than the other systems. XRD patterns at 28 days still revealed the presence of some unconsumed reactants such as ye'elimite and anhydrite.   Figure 5 displays the results of the isothermal calorimetry measurements of the four cements. C_TW, namely the system with the highest amount of C4A3$, showed the highest cumulative heat release over 14 days of hydration. It is worth noting that, compared to the other systems, the main hydration reactions were significantly slower and only started after ~24 h, which also agrees closely with the DT-TG data. Another interesting feature was observed for C_R, which exhibited a second acceleration of the hydration reactions after about 9 days. The reason for this behavior is most likely related to a change in the resulting solid phase equilibria, which may have caused changes in the pore solution chemistry. The porosimetric characteristics of OPC pastes have been widely studied and extensively reported in the scientific and technical literature [56][57][58]; it is well known that total porosity, threshold pore width and cumulative pore volume decrease as curing times increase [18]. At early ages, the differential curves show an initial sharply defined peak, indicating a unimodal pore size distribution, centered on the lowest width of pore necks connecting a continuous system. At longer curing times, a second peak appears, corresponding to the pressure required to break through the blockages formed by the hydration products which isolate the interior pore space. Figure 6 reports both cumulative and derivative plots for intruded Hg volume vs. pore radius for C_R at various curing times. For periods between 7 and 90 days, the pore size distribution was unimodal, with a threshold pore width ranging from about 300 to 15 nm. During the same period, the cumulative pore volume significantly reduces (from about 215 to 130 mm 3 /g).  C_TW, namely the system with the highest amount of C 4 A 3 $, showed the highest cumulative heat release over 14 days of hydration. It is worth noting that, compared to the other systems, the main hydration reactions were significantly slower and only started after~24 h, which also agrees closely with the DT-TG data. Another interesting feature was observed for C_R, which exhibited a second acceleration of the hydration reactions after about 9 days. The reason for this behavior is most likely related to a change in the resulting solid phase equilibria, which may have caused changes in the pore solution chemistry.   Figure 5 displays the results of the isothermal calorimetry measurements of the four cements. C_TW, namely the system with the highest amount of C4A3$, showed the highest cumulative heat release over 14 days of hydration. It is worth noting that, compared to the other systems, the main hydration reactions were significantly slower and only started after ~24 h, which also agrees closely with the DT-TG data. Another interesting feature was observed for C_R, which exhibited a second acceleration of the hydration reactions after about 9 days. The reason for this behavior is most likely related to a change in the resulting solid phase equilibria, which may have caused changes in the pore solution chemistry. The porosimetric characteristics of OPC pastes have been widely studied and extensively reported in the scientific and technical literature [56][57][58]; it is well known that total porosity, threshold pore width and cumulative pore volume decrease as curing times increase [18]. At early ages, the differential curves show an initial sharply defined peak, indicating a unimodal pore size distribution, centered on the lowest width of pore necks connecting a continuous system. At longer curing times, a second peak appears, corresponding to the pressure required to break through the blockages formed by the hydration products which isolate the interior pore space. Figure 6 reports both cumulative and derivative plots for intruded Hg volume vs. pore radius for C_R at various curing times. For periods between 7 and 90 days, the pore size distribution was unimodal, with a threshold pore width ranging from about 300 to 15 nm. During the same period, the cumulative pore volume significantly reduces (from about 215 to 130 mm 3 /g). The porosimetric characteristics of OPC pastes have been widely studied and extensively reported in the scientific and technical literature [56][57][58]; it is well known that total porosity, threshold pore width and cumulative pore volume decrease as curing times increase [18]. At early ages, the differential curves show an initial sharply defined peak, indicating a unimodal pore size distribution, centered on the lowest width of pore necks connecting a continuous system. At longer curing times, a second peak appears, corresponding to the pressure required to break through the blockages formed by the hydration products which isolate the interior pore space. Figure 6 reports both cumulative and derivative plots for intruded Hg volume vs. pore radius for C_R at various curing times. For periods between 7 and 90 days, the pore size distribution was unimodal, with a threshold pore width ranging from about 300 to 15 nm. During the same period, the cumulative pore volume significantly reduces (from about 215 to 130 mm 3 /g). Compared to C_R, C_T always showed a bimodal pore size distribution. The peak oriented toward higher radii values varied in a small range, from 200 at 7 days to about 180 nm at 90 days; however, the second peak ranged from about 100 at 7 days to 35 nm at 90 days. The cumulative Hg volume for C_T comprised the small range of 180-165 mm 3 /g. Figure 6 displays both cumulative and derivative curves for C_W. At 7 days of aging, the pore size distribution was trimodal, with a pore radius equal to about 650, 50 and 4 nm. After longer curing periods, the pore size distribution became bimodal, with pore radii equal to about 550 and 45 nm at 28 days as well as 150 and 20 nm at 90 days; this was due to the shift of the smallest peak beyond the minimum detectable pore value by the porosimeter (3.6 nm). During the same period, the cumulative pore volume decreased from about 180 to 140 mm 3 /g. At 7 days of curing, the differential curve for C_TW exhibited two quite wide peaks centered at about 500 and 25 nm, respectively, indicating a bimodal pore size distribution. A unimodal pore size distribution was established over longer aging periods (peaks centered at about 25 and 20 nm, respectively, at 28 and 90 days). The cumulative pore volume appreciably reduced between 7 and 90 days (about 175 to 100 mm 3 /g).

Conclusions
Titanogypsum (T) and water potabilization sludges (W) can, respectively, be used as suitable substitutes for natural gypsum and clay in belite-calcium sulfoaluminate (BCSA) clinker-generating raw mixes. The best results were obtained at 1300 • C (for the raw meal containing W) and at 1350 • C (for mixes containing only natural raw materials and for those containing T or both T and W) for selectivity of reactants towards the main hydraulically active components of BCSA cement (C 2 S and C 4 A 3 $).
Isothermal calorimetry (up to 14 days) and differential thermal-thermogravimetric analyses indicated some differences in the hydration kinetics, associated with the kind of sulfate and clayish sources used for the synthesis of BCSA clinkers. In addition, BCSA cements containing T (C_T) and/or W (C_TW, CW) displayed faster hydration kinetics than the reference system, while C_TW and CEM_T exhibited the highest hydration heat for the systems studied. As far as the ability of generating hydration products is concerned, the experimental results revealed an almost analogous tendency; however, at the longest curing periods, the best results were from the waste-based cements.
The behavior of BCSA cements can be improved using industrial wastes in place of raw materials, reducing the need for natural materials and landfill waste. The use of W in the clinker generating raw mix, averting the utilization of bauxite as the alumina source, could reduce BCSA manufacturing costs. Nonetheless, additional studies are necessary to understand the impact of the use of alternative raw materials and their minor components on hydration behavior.
Author Contributions: A.T.: conceptualization; methodology; formal analysis; investigation; data curation; writing-original draft preparation; writing-review and editing. T.M.: methodology; formal analysis; investigation; writing-review and editing. M.M.: conceptualization; methodology; formal analysis; investigation; data curation; writing-original draft preparation; writing-review and editing. All authors have read and agreed to the published version of the manuscript.