Production and properties of ferrite-rich CSAB cement from metallurgical industry residues Science of the Total Environment

consid-erablyfromtheblastfurnaceslaginchemicalandphysicalproperties.Inthisstudy,calciumsulfoaluminatebelite (CSAB) clinkers unutilized metallurgical residues: stainless composition clinkers ﬁ ed using quantitative XRD analyses, and the chemical composition of the clinker phases produced was established using FESEM-EDS and mechanicalpropertiesweretestedthroughcompressivestrengthtest.Itisdemonstratedthatthesemetallurgical residuescanbeusedsuccessfullyasalternativerawmaterialsfortheproductionofCSABcementthatcanbeused for special applications. In addition, it isshown that the availablequantities ofthese side-streams are enoughfor signi ﬁ cant replacement of virgin raw materials used in cement production.


Introduction
Population growth and urbanization increases the demand to expand our built environment (Miller et al., 2016) in which concrete is the fundamental building material (WBCSD, 2009). Cement is the key binding component in concrete, and global cement production was estimated to be 4.1 billion tons in 2017 (Cembureau, 2018). The most commonly used binder is Portland cement (PC), in which the major and characteristic mineral phase is alite (C 3 S) 1 (Taylor, 1997). PC manufacture emits approximately 0.87 tons equivalent carbon dioxide (CO 2 e) per ton of cement (Hanein et al., 2018). CO 2 emissions arise primarily from the calcination of limestone, which is mainly CaCO 3 , during the firing process, and secondarily from the burning of fossil fuels necessary to achieve the required firing temperature, which is approx. 1450°C in conventional PC production (Gartner and Hirao, 2015). The growing demand for cement, coupled with increasing environmental awareness, has led to increased focus from the cement community on mitigation strategies. The UNEP SBCI-report concludes that the most promising ways to decrease CO 2 emissions related to cement production are 1) increased use of supplementary cementitious materials (SCM's) and 2) more efficient use of PC (UNEP, 2017). Alternative non-Portland clinkers were also proposed to provide substantial reduction of CO 2 emissions (Scrivener et al., 2018).
One of the alternative, lower-carbon clinkers is calcium sulfoaluminate (CSA) cement that is based on less calcium-containing cementitious phases such as ye'elimite (C 4 A 3 $), belite (C 2 S), and ferrite (C 2 (A,F)) (Álvarez-Pinazo et al., 2013;Gartner and Hirao, 2015;Miller et al., 2018). The variations of CSA cements can be classified according to the main crystalline phases of the clinkers that they consist of (Álvarez Pinazo, 2015). In this study, CSAB (calcium sulfoaluminatebelite) refers to the cements with ye'elimite (C 4 A 3 $) as the main phase and belite (C 2 S) as the secondary phase; compositions also contain ferrite (C 2 (A,F)) and anhydrite (C$).
CSA cements have been manufactured commercially in China since the 1970s under the name "third cement series" (Zhang et al., 1999). In the United States of America, belite calcium sulfoaluminate (BCSA) cement has been successfully used for pavement rehabilitation of interstate highways for the past 40 years (Bescher et al., 2018). Recently the interest to use these low CO 2 cements has been growing. As shown in the work by Hanein et al., 2018, the benefit of CSA cements is that they have less CaCO 3 -related CO 2 emissions through reduced usage of limestone and lower production energy required. Hanein et al., 2018 concluded that the overall, net emissions can be reduced 25-35% when CSA cements are produced as alternative to PC. The main drawback of these CSA cements is the requirement of aluminium-rich raw materials, which are typically more expensive, and their availability have locational limitations. The assemblage of clinker composed from ye'elimite, belite, and ferrite phases allows for the employment of industrial side-streams in their manufacture, which would otherwise go into landfills (Chen, 2009).
The use of industrial side-streams in CSA clinker production can decrease the costs and increase local availability of the raw materials. However, the side-streams must be available locally to avoid transportation costs. Different industrial processing plants use different raw materials and manufacturing methods for their products; thus, producing varied side-streams. When these industrial side-streams are to be used as raw materials for CSAB cement production, the usability of a specific raw material and clinkering parameters should be tested and proven for the unique impurities and mineralogy of the side-stream materials which can affect cement clinker production and the properties of the hydrated product.
Several studies examine the use of alternative raw materials such as industrial side-streams or side-streams to manufacture CSA cement. Hanein et al. (2016) utilized sulfur, a side-stream of the Claus process, both as a fuel and source of clinker SO 3 to produce CSAB cements. Other studies focused on substituting CaCO 3 and relatively expensive bauxite respectively with CaO-bearing materials. This was done to reduce CO 2 emissions and Al 2 O 3− containing materials to lower raw material costs. El-Alfi and Gado (2016) used marble sludge with high CaO content to prepare CSAB cement at 1250°C. Chen and Juenger (2012) successfully synthesized side-stream material based CSAB cement using flue gas desulfurization (FGD) sludge, class C fly ash, and fluidized bed ash; however, they noted that impurities in raw materials could significantly affect CSAB cement formation, hydration, and properties. da Costa et al. (2016) used aluminium anodizing sludge (AAS) to replace bauxite and achieved good mechanical properties with their clinkers; however, they underlined that the use of alternative aluminium sources can affect the clinkerisation process through the incorporation of foreign ions such as Na, K, and Mg. Iacobescu et al. (2013) produced high ferrite containing CSAB cement using electric arc furnace slag (EAFS) at 1320°C; they found that their cements had good mechanical properties when they used 10-17 wt% EAFS (C 4 AF 33.0-38.9 wt%) together with FGD gypsum; but, a higher amount of EAFS led to poor mechanical properties. Adolfsson et al. (2007aAdolfsson et al. ( , 2007b produced CSAB cement by making four different mixtures from basic oxygen furnace (BOF) converter slag, EAFS, argon oxygen decarburisation (AOD) slag, ladle slag, gypsum, alumina, and limestone; they showed that there is potential in using metallurgical slags as a raw material for CSAB cement manufacture. When 14% of AOD slag was used to raw material mixture the major phases of CSAB could be synthesized but with 55% of AOD slag the major phases were ye'elimite and bredigite (Ca 7 Mg(SiO 4 ) 4 ); the latter having slow hydration and thus undesirable.
The use of AOD slag as raw material for CSAB cement production showed that the major phases of CSA could be synthesized from the material (Adolfsson et al., 2007a); however, their study did not contain information on the effect of foreign elements e.g. fluorine on the clinkering properties of CSAB cement. The literature is lacking in studies of using Fayalitic slags and Fe slags as raw materials for CSAB cement manufacture; although, some studies assessed the use of fayalitic slags as SCM's to replace PC (Rahman et al., 2017) and as raw materials for geopolymers (Maragkos et al., 2009). In this work, these three varied metallurgical residues are tested separately with reagent grade chemicals (RGC) to understand how they affect CSAB clinker properties individually. The aim of this study is to assess how the properties of clinkers changes when certain impurities are added to the system through using these industrial side-stream materials.
The CSAB clinkers manufactured in this study are intended for special applications e.g. mine backfilling, waste handling and industrial infrastructure. The hydration product of CSAB cement, ettringite, is known to encapsulate/stabilize heavy metals and other hazardous components (Chrysochoou and Dermatas, 2006;Kumarathasan et al., 1989;Peysson et al., 2005) and can enable the control of hazardous components in raw materials and also to stabilize hazardous components in other added materials such as mine tailings (Kiventerä et al., 2019). The target phase composition of the CSAB clinker is 40% (C 4 A 3 $), 35% (C 2 S), 20% (C 2 F), and 5% C$. Three different slags along with reagent grade chemicals are used to prepare 1.5 kg of four different clinkers. The clinkers are then characterized via XRF, XRD (coupled with Rietveld calculations), and FESEM-EDS. The mechanical properties of cement mortars were tested via compressive strength tests and selected cured cement pastes were analysed through FESEM-EDS analysis.

Materials
Four CSAB cement clinkers with the same target phase composition were manufactured from reagent-grade chemicals and Finnish 1 The cement oxide chemistry notation used in this work: C = CaO, S = SiO 2 , A = Al 2 O 3 , F = Fe 2 O 3 , and $ = SO 3 . metallurgical industrial side-streams. Details of the reagent grade chemicals used are provided in Table 1.
The Finish industrial side-streams used were: Argon Oxygen Decarburisation (AOD) slag resulting from the steel refining process at Outokumpu Stainless Oy (Tornio), fayalitic slag from the nickel flash furnace process at Boliden Harjavalta, and Fe slag from the pyrometallurgically treated jarosite at Kokkola Boliden. AOD is the generally employed process to refine stainless steel melt. The mineralogical composition of general AOD slag consist of C 2 S, C 3 S, calcium-aluminium silicates, CaCO 3 , free lime, periclase, and cuspidine (Ca 4 (Si 2 O 7 )(F,OH) 2 ) (Adegoloye et al., 2015;Durinck et al., 2008;Shen et al., 2004). The metallic oxide components (Fe, Cr, and Ni) in AOD slag are present as magnetite and chromite, or chromium spinel. AOD slag also contains metal droplets in forms of Fe\ \Ni alloy, Fe-Cr-Ni alloy, and stainless steel (Adegoloye et al., 2015;Durinck et al., 2008;Shen et al., 2004). Even after successful reduction, the slag still contains small amounts of Cr 2 O 3 , which limits its application because of the risk of leaching (Di Maria et al., 2018;Huaiwei and Xin, 2011;Shen et al., 2004;Zhao et al., 2013). Another problem in direct utilization of AOD slag is dusting of C 2 S during cooling of the slag (Seki et al., 1986). For this study, metal particles were removed from AOD slag and slag was ground by a contractor of the steel plant without describing. The fayalitic slag originates from a flash-smelting furnace process where nickel concentrate is smelted to produce nickel matte. After flash smelting, the slag is cleaned within a reductive electric furnace to recover Ni and Co from the slag, and the slag is granulated with water to d = 80 mm particles. The mineralogical composition of Fayalitic slag consists of N60% of amorphous glass; the rest is crystalline fayalite and a few percent of magnetite. The Fe slag used for this study is a modified jarosite from zinc production. The chemical formula of jarosite is MFe 3 (SO 4 ) 2 (OH) 6 where M is Ag + , H3O + , K + , Li + , Na + , NH 4+ ,or 1/2Pb 2+ (Dutrizac, 1984). For this study, the pyrometallurgically treated jarosite residue (Fe slag) is synthesized with the method introduced in the study of Rämä et al. (2018). The jarosite was treated with laboratory-scale pyrometallurgical oxidation-reduction at 1400°C. The aim of the pyrometallurgical treatment is to recover valuable metals from jarosite residue and to obtain clean inert slag, which could be used for construction purposes (Rämä et al., 2018). The mineralogical composition of Fe slag consists of Ca\ \Fe silicate and magnetite. The main chemical components and impurities of the side-stream materials are shown in Table 2. The main phases of Fayalitic and Fe slags have olivine structure where iron is present as FeO, during clinkering FeO is assumed to oxidize to Fe 2 O 3 . Before grinding, AOD slag was sieved through 500 μm sieve; metal droplets were not detected. AOD slag and fayalite slag were ground with stainless steel jar mill (10 L) for 1 h using 120 stainless steel balls (d = 30 mm) to make sure that the particle size of the slags are homogenous, estimation for particle size D 50 b 10 μm. Fe slag was crushed/ground using vibratory disc mill, the estimation for particle size D 50 b 200 μm.

Analysis of materials
Prior to preparation of melt-fused beads, loss on ignition (LOI) was measured using thermogravimetric analysis (TGA) using Prepash Precisa Gravimetrics AG: "prepASH automatic drying and ashing system". The heating cycle in TGA was 30 min ramp to 105°C/4 h hold, 2 h 30 min ramp to 525°C/3 h hold and 1 h ramp to 950°C/3 h hold. For melt-fused beads, 1.5 g of each LOI treated side-stream material and clinker were mixed separately in a platinum crucible with 7.5 g of X-ray flux Type 66:34 (66% LiB 4 O 7 and 34% LiBO 2 ). The melt fused beads were prepared using a Pananalytics Eagon 2 fluxer. The following sequences were used during fluxing: melting 1150°C/120 s, mixing for 480 s with rocking angle 45°, injecting of the releasing agent for 240 s, pouring duration was 15 s with angle 135°, and the fan cooling for 120 s. Before XRF analysis, the samples were ground with a planetary mill with 300 rpm for 1 min. The X-ray fluorescence (XRF) from a melt-fused bead was operated using an Omnian Pananalytics Axiosmax 4 kV to determine the main chemical components of side-stream materials and clinkers.
Before XRD analysis, each clinker was ground to a fine powder by hand using an agate mortar for 20 min and then loaded into powder sample holders. The XRD patterns were collected using a Rigaku SmartLab 9 kW diffractometer in the Bragg-Brentano geometry. The diffractometer was operated at 40 kV and 135 mA and is equipped with Co K α X-ray source (K α1 = 1.78892, K α2 = 1.79278) and D/teX Ultra 250 detector Co Kα radiation source was used because of iron content in the samples. Kβ radiation was reduced with a filter. The XRD-patterns were measured from 5.0°to 120.0°2-theta with a step size of 0.02°a nd scanning speed 5°/min; the total scan time was 24 min. A quantitative analysis to identify crystalline phases was performed using Rigaku PDXL 2 software that compares the diffraction patterns on the ICDD (PDF-4+ 2019 RDB) database using WPPF (Whole Powder Pattern Fit) analysis with Rietveld method; the lattice parameters and profile were refined, all peaks are manually checked and the crystallographic information files are selected based on literature, elemental composition, and knowledge from supporting analyses (FESEM and XRF). The WPPF analysis finds the best agreement between measured and calculated diffraction pattern through simultaneous refinement of models for diffraction optics, instrumental parameters, background, phase specific parameters (cell parameters and crystal structure) and sample specific parameters (crystallite size, stress and preferred orientation). The model functions during the analysis were as follows: background = bspline; peak-shift = shift axial displacement; peak shape function = split pseudo-Voigt and preferred orientation = March-Dollase. The WPPF analysis gives the phase composition normalized to 100% and the quality of analysis is determined from the weighted pattern Rfactors (Rwp) and goodness of fit indicator (S-value). The sources to  Polished sections for field emission scanning electron microscope (FESEM) analysis were prepared by using optical-grade epoxy resin (Struers Epofix). First, a piece of cement clinker was submerged in epoxy resin in a cylindrical 40 mm mould under vacuum and the cast was placed in a 40°C oven for 24 h to allow the epoxy to cure. After curing, a thin section was cut from the middle of the 40 mm cast. The cut thin section was placed to a 25 mm cylindrical mould, impregnated with epoxy resin in a vacuum and cured with same parameters. After curing, extra epoxy from the top of the cast was cut down, using a lathe and the bottom of the sample was polished with a wheel polisher (Struers LaboPol-6). Instead of water, 98% pure ethanol was used as lubricant/coolant. The polished sections were coated with carbon to achieve electrical conductivity. A FESEM (Zeiss Ultra Plus) equipped with an energy-dispersive X-ray spectroscopy (EDS) analysis system and back-scattered electron detector (BSE) was used to analyse phase distribution in CSAB clinkers. Configuration was operated with acceleration voltage of 15 kV, current 3.2-4.8 nA, and working distance between 7.3 and 8.0 mm. AZtec software (Oxford Instrument) was used to analyse the BSE pictures and the elemental chemical composition of the selected EDS points. The amount of each chemical element was received as weight percent.
Trace elements of industrial side-stream materials were measured with inductively coupled plasma atomic emission spectroscopy (ICP-OES). The microwave-assisted wet digestion was performed using a 3:1 ratio of HNO 3 and HCl acid mixture for 0.5 g of AOD slag, Fe Slag, and fayalitic slag at 175°C according to EPA3051A (United States Environmental Protection Agency, 2007).
For compressive strength tests, cements mortars were prepared according to cement standard (EN 196-1, 2005). Due to limited amount of material availability, 2 × 2 × 8 cm prisms were used instead of 4 × 4 × 16 cm prisms. Cement mixtures for mortars were mixed by using 85% clinker (AOD_C, Fay_C, Fe_C, RG_C) and 15% gypsum. Mortars were prepared using lab scale cement mixer (65-L0006/AM AUTOMIX) using prepared cement mixes, standard sand and tap water with cement to water ratio of 0.5. Mortars were cured in 98% air humidity at room temperature for 1, 7 and 28 days prior to being tested. Three parallel 2 × 2 × 8 cm mortar prisms were cut to two halves using precision cutting machine (Struers Secotom 10). Each half was then used for the unconfined compressive strength tests, totalling six measurements for each type of mortar. Tests were operated using Zwick testing machine with a maximum load of 100 kN, here employing a loading force of 2.4 kN/s. The reference samples from commercial PC (CEM II/A-LL 52.5 N) was mixed according to standard EN 196-1 and also tested with 2 × 2 × 8 cm prisms after 7 and 28 days of curing.

Clinker preparation
Stoichiometric calculations were used to estimate the amount of raw material oxides used to prepare the CSAB cement clinker. The following assumptions were carried out for the calculations: • C 2 F is the only phase that contains iron (Fe 2 O 3 = C 2 F (wt%) / (C 2 F (g/ mol) / Fe 2 O 3 (g/mol)) • C 2 S is the only phase that contains SiO 2 (SiO 2 = C 2 S (wt%) / (C 2 S (g/ mol) / SiO 2 (g/mol)) • C 4 A 3 $ is the only phase that contains alumina (Al 2 O 3 = C 4 A 3 $ (wt%) / (C 4 A 3 $ (g/mol)/3 * Al 2 O 3 (g/mol)) • SO 3 in excess remains as C$ (SO 3 = C 4 A 3 $ (wt%) / (C 4 A 3 $ (mol)g / SO 3 (g/mol)) + C$ (wt%) / (C$(g/mol) / SO 3 (g/mol)) • CaO required is the sum of CaO in empirical formula of single phases multiplied with the target amount of each phase The target composition is presented in Table 3. Based on prior knowledge, excess free lime should be added to prevent gehlenite formation and extra SO 3 to compensate possible SO 3 loss during firing and to maximize ye'elimite formation over mayenite. In cement chemistry, ferrite generally refers to tetracalcium aluminoferrite (C 4 AF), which represents one point of the actual solid solution series (Redhammer et al., 2004). The common mineral name for Ferrite is brownmillerite Ca 2 Al x Fe 2−x O 5 . Ferrite appears with various Al/Fe ratios and ionic substitutions (Taylor, 1997). At atmospheric pressure, a ternary continuous solid-solution series Ca 2 Al x Fe 2−x O 5 can exist in the compositional range 0.0 ≤ x ≤ 1.4 (Geller et al., 1971;Redhammer et al., 2004;Smith, 1962;Taylor, 1997). Due to iron content of the residues (Table 2), x = 0 was selected to maximize the iron content in the solid-solution series and allow the Al to be fully exploited in a more cementitious phase, ye'elimite. Our target ferrite phase in this study is Ca 2 Fe 2 O 5 (C 2 F).
Four different CSAB clinkers were prepared. Three clinkers were produced using industrial side-streams (AOD_C, Fay_C, and Fe_C) and one control sample was produced from reagent grade chemicals (RGC) alone. Each starting mixture was prepared based on the corresponding industrial slag mixed with the reagent grade chemicals necessary to achieve the target clinker composition; see Table 4.
In Table 4, the amount of raw material in industrial side-stream clinkers is over 100% because the impurities of raw materials were not taken account in the calculation of main oxides of target compositions. The amounts of main impurities in 100% of clinkers (Table 5) were calculated from XRF and ICP analysis results of raw materials (Table 2). MgO was the main "impurity" in all raw materials. In AOD slag, MgO came from a slag-forming agent (dolomitic lime) that is added to process and from the wearing of refractory materials. In fayalitic and Fe slag, MgO originates from the raw material feed. In addition, all the raw materials include minor quantities of various impurities as shown in Table 5. Before weighing the components, CaO was calcined at 800°C for 4 h to remove possible moisture and/or CO 2 that may have interacted with it. Calcium sulphate was dried at 250°C for 2 h, and all the side-stream raw materials were ground with a ball mill and dried at 105°C for 24 h. Raw materials were weighed to produce 1.5 kg of each clinker. The materials for each clinker were mixed in a stainless-steel bowl and sieved several times. After mixing, raw material mixtures of each type of clinker were compressed in three cylinder-shaped corundum crucibles by hand.
Clinkering (firing) was conducted in a Nabertherm chamber furnace. The crucibles were placed within a furnace pre-heated to 800°C, then heated 10°C/min to 1300°C, and then held for 4 h. After, the crucibles were taken out. To achieve faster cooling, the crucibles were placed on a copper table with water circulation.

Clinker visual characterization
After firing, it was observed that AOD_C clinker differed from the other clinkers produced. The indurated AOD_C clinker "cake" was remarkably hard to grind, where other clinkers were friable and easy to grind.
The visually observed melting or more advanced sintering behaviour in AOD_C (Fig. 1) can be explained with presence of fluorine, which is known to work as flux or mineralizer in Portland cement and modified Portland cements (Blanco-Varela et al., 1995;Odler and Zhang, 1996;Shame and Glasser, 1987). Flux lowers the initial liquid formation temperature and mineralizer can lower the formation temperature of minerals. (Blanco-Varela et al., 1995;Duvallet, 2014).

XRD analysis
After visual analysis, clinkers were ground using a tumbling ball mill and the phase contents of each produced clinker were determined using XRD and quantified with the Rietveld method. The XRD-patterns with the six main phases are shown in Fig. 2, and the results of Rietveld analysis with major and minor phases are presented in Table 6. The presented diffraction patterns are limited to 10-55°in order to make the data discernible.
The weighted pattern R-factors (R wp ) are in the range 5.0-5.3%, and the goodness of fit indicator, S-value, was between 2.7 and 3-9. The Rietveld analysis was validated by comparing the back calculation of oxide content from Rietveld measurements with that from XRF analysis of the clinkers and with initial composition, which is the sum of target composition of the clinkers and impurities (MgO and CaF 2 ) based on the XRF data of the raw materials.
The phase compositions of RG_C, Fay_C, and Fe_C clinkers were close to the target composition (see Table 3); the main phases identified by XRD were ye'elimite (C 4 A 3 $), belite (C 2 S), and Ferrite (C 2 F). Both cubic and orthorhombic polymorphs of ye'elimite are present in all clinkers while belite is mostly present as β-C 2 S (i.e. larnite). For the analysis, the reference for ferrite phase was selected to be Ca 2 (Fe 2 )O 5 with space group Pnma, because according to Redhammer et al. (2004), pure Ca 2 Fe 2−x Al x O 5 has space group Pnma up to x = 0.56 and I2mb when x N 0.56. The minor phases present in clinkers were anhydrite (C$), mayenite (C 12 A 7 ), periclase (MgO), and unreacted SiO 2 .
On the other hand, in AOD_C, part of the belite and ferrite was replaced with unwanted fluorine containing fluorellestadite (Ca 10 (SiO 4 ) 3 (SO 4 ) 3 F 2 ), with apatite structure; see Table 6. Due to the lower ferrite formation, iron was detected, through XRD it was analysed to be unreacted iron oxide, but later in this study at least some metallic iron particles originating from AOD slag were detected with FESEM-EDS from hydrated AOD_C clinker. The formation of fluorellestadite can be attributed to the fluorine content of AOD slag. In the study of Adolfsson et al., 2007a fluorellestadite was not detected even though the slag used may have contained fluorine as fluorspar is normally added to AOD converter during the AOD process. When fluorine is present in raw mix, care must be taken for mix design and clinkering conditions in order to avoid melting and fluorellestadite formation. Fluorellestadite should be avoided in the final product because it ties SO 3 , SiO 2 , and CaO with F and is not cementitious (Blanco-Varela et al., 1995;Hewlett, 2003). According to the literature concerning PC and modified PC, fluorellestadite should decompose at 1240-1250°C (Blanco-Varela et al., 1995;Hewlett, 2003); however, it seems that in our sample, the fluorellestadite stays stable in AOD_C. AOD_C also had significant amounts (4 wt%) of periclase (MgO) which is an unwanted component but cannot be avoided when using side-stream based raw materials. The Rietveld analysis presented in Table 6 is validated in Table 7 by comparing the back calculation of oxide content from Rietveld measurements with that from XRF analysis (Table 8) of the clinkers and with initial composition, which is the sum of target composition of the clinkers, including MgO and CaF 2 , based on the XRF data of the raw materials (Table 2). For back calculation, the phases considered were: C 4 A 3 $, C 2 S, C 2 F, C$, C 12 A 7 , Ca 10 (SiO 4 ) 3 (SO 4 ) 3 F 2 , MgO, SiO 2 , and Fe 2 O 3 . XRF analysis of the clinkers validates the Rietveld analysis (WPPF) results with desired accuracy. Even for typical phase assemblages the reproducibility of XRD analysis for cement phase is challenging and general uncertainties are in the range 4.1-6.5% (León-Reina et al., 2009).
Possible errors in the analyses are caused by non-detected minor phases, overlapping peaks, atomic impurities, several different polymorphs of the phases and possible analysis errors (León-Reina et al., 2009). From XRF analysis, iron content in Rietveld back calculation is over estimated. The content of Al 2 O 3 and Fe 2 O 3 differs in back calculation because stoichiometric C 2 F was used instead of C 4 AF for calculations, and the ferrite phase present contains Al 2 O 3 . Actual A/F ratios in ferrite are determined with FESEM-EDS later in this study. In back calculation of Fe_C there is less SO 3 when compared to XRF measurements and initial chemical composition. From FESEM-EDS analysis, it was detected that in Fe_C there were barium-and sulphate-rich grains, probably baryte, which were not detected in XRD data; this must be validated in future studies.
To further confirm results presented in Table 9, the chemical composition of clinkers is calculated through FESEM-EDS to validate the Rietveld and XRF analysis presented in Tables 6, 7 and 8. The calculation is done by converting detected and analysed weight percent of main phases (C 4 A3$, C 2 S, C 2 F, Ca 10 (SiO 4 ) 3 (SO 4 ) 3 F 2 and MgO) from FESEM-EDS point analysis to oxides and then, the oxide contents are multiplied with the phase contents obtained from Rietveld analysis presented in Table 6. The advantage using FESEM-EDS point when compared to XRD is that the real chemical composition of single phases is used, which also takes foreign ions associated to phases in account. The disadvantage is that undetected/non-analysed minor phases are excluded from analysis. SO 3 content in Table 9 is low in RG_C, Fay_C and Fe_C because anhydrite phase could not be analysed with FESEM-EDS due to insufficient number of EDS-points. In AOD_C, SO 3 is found as fluorellestadite rather Table 6 XRD-WPPF (Whole Powder Pattern Fit) results of produced clinkers, the values are given in %. The quality of analysis is given as weighted profile R-factor (R wp ) and the goodness of fit S.  (Morikawa et al., 1975) Ca(SO 4 ) 04-007-6682 6.7 1.1 5.3 3.9 Mayenite (Palacios et al., 2008) C a 12 Al 14 O 33 04-015-5592 0.2 2.9 0 1.6 Fluorellestadite (Pajares et al., 2002) C a 5 (SiO 4 ) 1.5 (SO 4 ) 1.5 F 04-016-5550 0 14 0 0 Periclase (Sasaki et al., 1979) MgO 01-071-1176 0.2 3.7 1 0.7 Silicon dioxide (Ohsumi et al., 1982) SiO 2 04-008-8250 0.3 0.4 0.5 0.2 Iron (III) oxide (Bykova et al., 2016) F than anhydrite allowing the calculated SO 3 content to be closer to target. Iron content in XRF analysis of clinkers is the targeted 10%. The iron content of AOD_C in Table 9 is low because it is not all found as C 2 (A x F 1−x ).

Characterization with FESEM-EDS
The main phases of RG_C clinker (ye'elimite, belite, and ferrite) detected from a FESEM-BSE image (see Fig. 3) are existing mainly as mono-mineral clusters composed of at least ten of individual mineral grains. The grinding and re-firing could have led to a more homogenous microstructure. The microstructure of Fay_C (not shown) was very similar to RG_C; the only difference was the presence of periclase as a minor phase. FESEM-BSE images of AOD_C and Fe_C are presented in Figs. 4 and 5, respectively. The similar main phases, as in RG_C, could be detected from these clinkers.
Fluorine containing fluorellestadite (Ca 10 (SiO 4 ) 3 (SO 4 ) 3 F 2 ) in AOD slag clinker can be identified in Fig. 4A by the elemental mapping. Fluorellestadite has lower Si content than belite, higher S content than ye'elimite, and does not contain any Al. From Fig. 4B, it can be seen that fluorellestadite (dark grey) is located next to ye'elimite (black), and there is also some fluorellestadite inclusion in ye'elimite grains. The ferrite phase is observed as lightest colour in Fig. 4A and B, and it appears to have filled the empty spaces after formation of larnite/belite, ye'elimite, and fluorellestadite grains.
The identified phases in Fe_C were ye'elimite, belite (larnite), ferrite, and periclase (Fig. 5A). Barium and sulphate containing white grains were detected with EDS (Fig. 5B). Because of the small size of the grains, the EDS analysis was mixed with other surrounding minerals. Although the grains could not be identified accurately, this is probably barium sulphate (baryte).
In the clinkers based on side-streams, impurities were detected in both major and minor phases with EDS point analysis. The impurities present in main phases are presented in Table 10. Due to the high iron content of the original chemical composition of all of the synthesized clinkers, ye'elimite and larnite contained iron as an impurity. According to FESEM-EDS analysis in larnite and ye'elimite, Fe 2 O 3 content is between 0.6 and 1.1 wt% and 2.7-3.6 wt%, respectively. Ye'elimite contained Mg, K, and Na; belite/larnite contained Mg and Na; Ferrite contained Na, Mg, Na, and K, and metals Ti, Cr, and Mn; periclase contained Fe, Ni, Zn and Co.
The impurities from industrial residues can stabilize different polymorphs of belite (for example α-C 2 S, α' H -C 2 S or β-C 2 S). In this study, most of the C 2 S in the produced clinkers was β-C 2 S (i.e. larnite) and only small amounts of α' H -C 2 S were detected with XRD. This means that foreign ions present in the systems studied here are stabilizing larnite which is well in accordance with current literature. The usage Table 9 The total chemical composition of clinkers based on average chemical composition of main phases (ye'elimite, larnite, ferrite and fluorellestadite) detected and analysed with FESEM-EDS point analysis and the phase composition determined with Rietveld analysis (  of foreign ions to chemically stabilize belite, especially α′, α, and β polymorphs, have been widely studied (Benarchid et al., 2004;Chen, 2009;Cuberos et al., 2010;Morsli et al., 2007). According to the review paper prepared by Aranda and De la Torre (2013), oxide components such as SO 3 , B 2 O 3 , Cr 2 O 3 , Na 2 O, MgO, P 2 O 5 , K 2 O, BaO, and MnO 2 can be used to stabilize or to cause distortion to different forms of belite. The stabilization of belite is based on defects in the crystal structure brought about by silicate anion replacements with phosphates and other anions and/ or calcium cation substitutions (Fukuda et al., 1995;Morsli et al., 2007).
In the produced clinkers, the ferrite phase contained several different impurities. The average chemical composition of ferrite was analysed by using FESEM-EDS analysis, converted to oxides, and is presented in Table 11. From Table 11, the simplified empirical formula estimated for the ferrite phase is C 2 (A x F 1−x ) with 0.2 ≤ x ≤ 0.3 (X RG_C = 0.24, X AOD_C = 0.2, X Fay_C = 0.28 and X Fe_C = 0.3). The content of foreign ions in ferrite might also promote stability/unreactivity during hydration. According to Table 11, the amount of MgO content in ferrite varied between 0.1 and 1.7 wt%.
According to Jupe et al. (2001), doping with Mg might lead to less hydration of ferrite. They concluded that doping might lead to a stabilized, ordered structure with fewer defects. Defects work as sites where hydration can be initiated. In AOD_C and Fe_C, ferrite contained 0.9 wt% of TiO 2 . Duvallet et al. (2012) detected that even a low TiO 2 (~0.5%) addition in ferrite slows the setting time slightly when compared to pure ferrite without TiO 2 . The low Al 2 O 3 content in ferrite might also lead to slow hydration because the main hydration product of ferrite is the aluminium containing mineral ettringite (Meller et al., 2004). In addition, the effect of iron, fluorine, and sulfur contents on phase formation, firing time, and clinkering temperature should be investigated. Further studies on optimised compositions and processing parameters for AOD slag-based clinker are needed and the fluorellestadite formation needs to be minimised/avoided. A promising alternative to utilize AOD slag is to take advantage of the fluorine content to manufacture alite-ye'elimite-ferrite clinker. Hanein et al. (2019) demonstrated that the presence of fluorine, sulphate and iron enhances the mineralization in the C 3 S-C 4 A 3 $-C 4 AF system, promoting alite formation at 1250°C. Duvallet et al. (2014b) successfully produced C 3 S-C 2 S-C 4 A 3 $-C 4 AF using red mud as raw material and CaF 2 and CaSO 4 as fluxes and mineralizers. In current study alite was not formed because there is not enough CaO to form alite and also the fluorine content is not adjusted to be optimal for alite formation.

Mechanical testing
The reported compressive strength results in Fig. 6 are average values of six tests. After 1 day of curing, Fay_C, FE_C and RG_C mortars reached approximately 20-30 MPa strength, then the strength kept developing to reach N30 MPa at 28 days of curing. The lower compressive strength in AOD_C can be attributed to the presence of fluorellestadite which is non-hydraulic. The results of CSAB clinkers are compared to commercial fast setting PC clinker (7 and 28 days of curing) and show that compressive strength of the produced CSAB clinkers is comparable with commercial PC clinker.  The strength development of prepared CSAB clinkers is associated to hydration reaction of ye'elimite together with anhydrite and gypsum to ettringite (Ca 6 Al 2 (SO 4 ) 3 (OH) 12 ·26H 2 O). In Fig. 7 the formation of ettringite was confirmed with FESEM-EDS from hydrated RGC_C cement sample after 7 days of curing. The cured samples were prepared by mixing 15% of gypsum with RG_C clinker, water addition with water to cement ratio 0.5 and cured in humidity chamber. After curing the hydration was stopped by immersing crushed (d b 4 mm) sample to acetone and the polished section was prepared and analysed with FESEM-EDS with methods mentioned in Section 2.2.1. Ettringite appears to be the main phase in cement matrix after 7 days of hydration, but also unreacted cement phases (Belite and Ferrite) could be identified from EDS-analysis. In EDS-analysis the total sum of ettringite phase is~90 wt% because hydrogen H cannot be detected with FESEM-EDS.
Ettringite provides the early strength to hydrated CSAB cement but unreacted ferrite and belite may potentially further increase the mechanical strength upon their hydration. There was a slight increase in mechanical strength between 7 and 28 days with indicates some progress in hydration of these phases. The hydrationand mechanical properties can be further improved by optimization of gypsum and water content and using retarders i.e. tartaric acid and citric acid etc.
According to FESEM studies, the hydration behaviour of Fe_C and Fay_C were very similar to RG_C. High amount of unreacted cement components explains the lower mechanical strength of AOD_C. The elemental mapping of AOD_C is shown in Fig. 8. In AOD_C there was much more unreacted cement minerals, especially belite and ferrite. The phases presented in Fig. 8, descending from darkest to lightest are ettringite matrix (S, Al, and Ca), larnite grains (Ca and Si), ferrite grains (Fe, Si and Ca), and white particle in the bottom right corner was detected to be metallic iron originating from AOD slag. The presence of unreacted fluorellestadite in AOD_C after 7 days of curing was confirmed with XRD tests.

Estimation of raw material amounts to replace Portland cement
The availability of alternative raw materials in certain parts of the world is enough to fulfil a significant amount of virgin, raw cement material demand. These materials are a source of already de-carbonised CaO that can allow for a notable reduction in the carbon footprint of cement manufacture while also promoting a circular economy. It is revealed that N10%, perhaps up to 16%, of produced cement in Finland can be replaced through using these alternative raw materials, as estimated in in Table 12. Data for other countries is not directly available to the authors; however, similar positive impacts may be achieved in countries with a prospering metallurgical industry. Nevertheless, a localized solution for the use of these side-streams and for potentially reducing the cost and carbon footprint of the cement industry is presented. Using single industrial raw material to manufacture clinker is impractical in real applications, because a lot of RGC's or natural materials such as bauxite and calcite are required. It would be much more practical to manufacture clinker from a mix of a few industrial raw materials suggested in this study or alternative materials, which will be studied in the future. Examples of alternative industrial materials to replace RGC's in CSAB clinkers include: phosphogypsum (CaSO 4 ), ladle slag (Al 2 O 3 , CaO, SiO 2 ), red mud (Al 2 O 3 , Fe 2 O 3 , CaO, SiO 2 ), metakaolin (Al 2 O 3 , SiO 2 ) and fly ash (SiO 2 , CaO).

Conclusions
The title study reveals how the impurities in metallurgical industry residue materials affect the properties of CSAB cement clinker. CSAB cement prepared from reagent grade chemicals were compared with those prepared separately from certain industrial side-streams. Metallurgical industry residues can be used as alternative raw materials for the manufacture of CSAB cement which may be used in various applications depending on local regulations and standards; however, minor elements in the material will need to be considered.
The use of Fe slag from zinc production and fayalitic slag together with reagent grade chemicals led to the targeted clinker phase assemblage: ye'elimite-belite-ferrite. With blended gypsum, ettringite formation is observed upon hydration and the produced clinkers achieved N30 MPa compressive strength after 7 days of curing. Impurities originating from the raw materials were detected in minor amounts in all major phases of industrial residue-based CSAB clinkers, and those clinkers contain some minor phases such as periclase.
The utilization of stainless-steel refining (AOD) slag containing fluorine led to the formation of a significant amount of a non-hydraulic phase, fluorellestadite; the clinker phase assemblage was: ye'elimitebelite-fluorellestadite-ferrite instead of the targeted composition. The formation of fluorellestadite affected negatively to the mechanical properties and the achieved strength was only 22 MPa after 28 days of curing for the system studied. The unexpected formation of fluorellestadite demonstrates that, when using industrial side-streams, the raw mix design will need more consideration than is conventional and traditional stoichiometric/Bogue-type calculations may not be sufficient for clinkering calculations.
The different material manufacturing/processing methods worldwide can produce side-stream materials with compositions that are varied enough to have a major effect on properties of any produced    Fig. 8. FESEM-BSE-picture (×3000) of AOD_C cement sample after 7 days curing, where dark grey matrix is ettringite and light grey particles are unreacted clinker phases. White particle in Southeast corner is a metallic particle originating from AOD slag.
cement. Care must be taken when adopting procedures and designs created elsewhere; experimental testing may be needed locally to evaluate and confirm the suitability of these materials for use as alternative raw materials in cement manufacture.