Physical, chemical and mineralogical characterization of Dutch fine recycled concrete aggregates A comparative study

(cid:1) The properties of individual and total fractions of fRCA were investigated. (cid:1) Based on combined experimental approaches, characteristic quality indicators were deﬁned for fRCA. (cid:1) The content and surface area of ﬁne fraction (0–0.250 mm) and particle size distribution of fRCA varied with the source. (cid:1) fRCA are made up by crystalline phases (>70 wt%), notably quartz (>60 wt%), and adhered cement paste (<30 wt%). (cid:1) Whether the unreacted binder particles in studied fRCA still represent any reactivity may be doubted. of individual and total fractions (0–0.250 mm, 0.250–4 mm and 0–4 mm). The physical properties such as grading, density, surface area, water absorption and cement paste content of fRCA were tested. The chemical analyses include quantiﬁcation of element com- position with X-ray ﬂuorescence spectrometry (XRF) and carbonate content with thermogravimetry and mass spectrometry (TG-MS). Potential contamination (chlorides and sulfates) and reactivity of selected fractions were evaluated. In addition, qualitative and quantitative phase analyses with X-ray diffraction (XRD) combined with Rietveld reﬁnement method were performed and supported by optical polarizing- and-ﬂuorescence microscopic (PFM) study. Based on combined experimental approaches, characteristic quality indicators were deﬁned for fRCA. These indicators showed that fRCA were uncontaminated and nonreactive. Despite fRCA were from different origins, they had similar chemical and mineralogical composition and contained comparative chlo- ride content. In contrast, the content and surface area of ﬁne fraction (0–0.250 mm) and particle size distribution of fRCA varied with the source. With this it can be assumed that fRCA will have different effect on the properties of the new concrete.


Introduction
In the coming decade, sustainable construction materials with low environmental impact will have a crucial role in achieving a Circular Economy (CE). The Dutch CE strategy aims to make the Netherlands 'fully circular' by 2050 with an interim target of a 50% reduction of primary materials (minerals, fossil fuels and https://doi.org/10.1016/j.conbuildmat.2020.121475 0950-0618/Ó 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). metals) by 2030 [1]. Therefore, in the Netherlands, beside cement replacement with more environmentally friendly cement types, there is also an urgent need for circular materials for sand. Despite the potential that industrial waste has for concrete production, mainly as replacement for cement, resources for aggregates are not sufficient and concrete rubble is necessary to be included in the stream of alternative raw materials. The largest source of aggregates is the construction and demolition waste (C&DW) and its quantity will be significantly increasing due to aging and deterioration of the concrete infrastructure [2].
The variations in chemical and physical properties of fine recycled concrete aggregates (fRCA) [3,4], lack of experience, lack of standards for their quality evaluation, lack of existence of regular supply/demand systems and environmental concerns are the main reasons for fRCA being mainly used for low grade applications such as road pavements, sub-basements, soil stabilization, improvement of sub-ground [5,6]. However, this is not the most sustainable way of recycling, because the production of new concrete still requires large amounts of primary aggregates.
In some studies, parts of fRCA were used in cement raw meal for cement production [7,8]. Recent laboratory studies have shown the development of mortars and concretes with different fRCA replacement levels and the results related to their performances were encouraging [9,10]. However, the reuse of fRCA as a fine aggregate in new concrete have not been employed yet in practice. This is caused by the fact that concrete rubble has not been sorted at the source and therefore, recycled concrete aggregates might have large variations in the properties. The key question for application of fRCA as sand replacement in concrete is which properties would hamper the performance of concrete and, related to this, which quality indicators should be used.
Material characterization is an important step in order to understand variations of recycled material and their key properties affecting concrete performance. For example, the type of aggregate, or strictly speaking its modulus of elasticity, influences shrinkage of concrete, the particle size and shape influence workability and the mechanical properties of concrete [11]. Aggregate packing directly affects the way of aggregate particles forming a skeleton to transmit and distribute loads, thus influencing the stability and mechanical performance of the concrete. One of the most critical aspects for durability of concrete, are soluble chlorides and sulfates contained in concrete components of which the amount is limited in concrete in order to prevent degradation, in this case initiation of corrosion of reinforcement or sulfate attack. Typically, chlorides can be present in concrete components when cast, i.e. in water, cement, aggregates and in admixtures, but also concrete structures are often exposed to chloride-contaminated environments in the Netherlands, causing chloride ions to penetrate through the concrete cover [12]. After recycling of such a concrete, chemical composition and phase analysis can be an indicator for contamination/deterioration level of a recycled concrete. X-ray diffraction analysis (XRD) is a powerful tool in this regard, but challenges can arise due to the complex and often varying composition of fRCA. For this reason it is necessary to combine XRD with other characterization techniques such as thermogravimetry and microscopy studies. Up to now, the water absorption and water demand of fRCA are the most extensively and comprehensively studied among all properties of fRCA [13][14][15][16]. Only a few studies on physical-chemical-mineralogical characterization of C&DW recycled concrete aggregates including bricks [4,17] and of laboratory crushed concrete [18] have been done.
In the present study, characterization of physical properties, chemical and mineralogical composition of various Dutch fRCA (from unknown concrete rubble) is performed. The aim is not only to investigate properties of fRCA, but also to focus on how these properties (quality indicators) can be correlated to the mortar and concrete mix design with respect to the specific property of the mix. The physical analyses were performed with the following techniques: particle size distribution using a sieve tower and laser diffraction analyzer; density with helium gas pycnometer; BET specific surface area using a nitrogen adsorption. The chemical analyses consisted of: chemical analysis of acid-soluble chlorides and water-soluble sulfates; element composition measured with X-ray fluorescence spectrometry (XRF); carbonate content with thermogravimetry and mass spectrometry (TG-MS), isothermal calorimetry, X-ray diffraction analysis to characterize the crystalline phases; optical polarizing-and-fluorescence microscopy (PFM) to characterize mineralogy and morphology of fRCA.

Materials
Three different fRCA were used in this study. They were chosen to be representative of major Dutch fRCA because of their geographical spread (B-fRCA is delivered from the South, C-fRCA is delivered from the West and D-fRCA from the East of the Netherlands) and recycling technique (jaw crusher combined with a cone crusher or rotor crusher). As reference material, Dutch river sand (A) with 93 wt.% of SiO 2 was used. All fRCA were first dried in an oven at 105°C. The appearance of fresh and dry materials is shown in Fig. 1.
The particle size distribution for the fraction (0-0.250 mm) was measured using a laser diffractometer. In the tests, 0.5 g of sample powder was mixed with 800 mL ethanol (rather than water since hydration may take place when mixing cementitious material with water) and stirred for 5 min. Meanwhile, an ultrasonic vibrator was turned on to prevent agglomeration of fine particles. After mixing for 5 min, particle size measurement (distribution in volume) of fine powders was measured by laser. Three replicates were made per sample.

Density and specific surface area
Density of individual and total fractions (0-0.250 mm, 0.250-4 mm and 0-4 mm) was determined with a Helium Pycnometer (Model: Ultrapycnometer 1000), employing the method of gas displacement to measure volume accurately. Before the density measurements, the samples were dried and degassed at 105°C until a constant mass in order to remove any trapped air and moisture. Ten consecutive volume measurements were performed to obtain an average density value. An average mass of the tested sample was 60 g.
Low temperature gas adsorption has been used to determine the surface area of river sand and fRCA. Next, the specific surface area is calculated by dividing absolute surface area of the sample by the sample mass. The specific surface area, S_BET, of fraction 0-0.250 mm was determined using nitrogen adsorption method based on Brunauer-Emmett-Teller (BET) theory [19]. The nitrogen adsorption tests were conducted by using Gemini VII 2390 with a relative pressure (P/P0) range from 0.05~0.99. The relative pressure is defined as equilibrium vapor pressure divided by the saturation vapor pressure. Approximately 1 g of sample was used for the analysis.

Moisture content and water absorption
Materials were first dried in a ventilated oven at 105°C until a constant mass to determine the moisture content according to NEN-EN 1097-5:2008 [20]. Constant mass is determined by successive weighings performed at least 1 h apart and not differing by>0.1%.
Water absorption of sands was determined according to standard NEN-EN 1097-6:2013 [21], by measuring the quantity of water present in the sands at saturated surface dry state (SSD).

Chemical-mineralogical properties testing
The procedure for extracting subsamples from fRCA (0-4 mm) was followed from NEN-EN 932-2:1999 [22] for chemical-mineralogical analysis. The original amount of samples was about 20 kg for each batch. Of this, a subsample of about 3 kg was taken according to NEN-EN 932-2. For the grain size fraction 0-4 mm, the standard requires a minimum of 0.5 kg. The amount of 3 kg is taken as a representative amount in relation to the possible inhomogeneous distribution of any fRCA particles, and inhomogeneous distribution of any relevant chemical elements in them. For XRF, TG-MS and XRD analyses, the fRCA samples were first ground with a grinding machine (a Fritsch pulverizer 5) to an average particle size of 100 mm. After grinding, milling with an Alpine e200 LS air jet sieve of fRCA fractions was done and particles with an average size of 25 mm were obtained. This powder was used for XRD and TG-MS analyses. Additional manual grinding down to an average size of 10 mm was done for XRF analysis. For analyses of acidsoluble chlorides, water-soluble sulfates, calorimetry and microscopic study, samples were not ground.

Acid-soluble chlorides and water-soluble sulfates
The acid-soluble chlorides were obtained by acid dissolution using 6 M nitric acid according to NEN-EN 1744-5:2006 [23]. As this standard is designed for samples with a much lower weight and much higher chloride concentration, representativity may be a problem for the fRCA samples. Therefore, quantities of test specimens were modified. Samples of ca. 20 g were used. Acid-soluble  chloride salts content was further determined following Volhard method cf. NEN-EN 1744-5:2006. The solid residue left on the filter paper was dried in an oven at 105°C for 24 h. This allowed determination of the cement paste content by subtracting the weight of insoluble substances from the initial weight of the oven-dried material. The water-soluble sulfate content in the fRCA was determined according to NEN-EN 1744-1:2009 +A1:2012 § 10.2 [24]. The sample consisting of 25 g of sand (crushed below 4 mm) was mixed with 1 L of water at 65°C under stirring to extract water-soluble sulfate ions. A neat sample is filtered using a Buchner filter with a medium grade filter paper 110, and the filtrate is analyzed by Metrohm 850 ion chromatography system. Sulfate ion was resolved on an anion exchange column and detected by a conductivity detector. Quantification was performed by linear regression analysis of peak areas from the standard sulfate calibration curve containing five points.

X-ray fluorescence spectrometry
The X-ray fluorescence spectra of studied materials were recorded with a Bruker S8 Tiger spectrometer on pressed pellets. As binding agent, Boreox (C 12 H 22 O 11 ) from Fluxana, was used. Samples were mixed in a stainless steel grinding cup. The loss on ignition (LOI) was measured by heating the fRCA to 1050°C. The LOI is expressed in weight percent of the dry matter [25]. The mass of the sample was 5 ± 0.1 g.

Thermogravimetric-mass spectrometric analysis
A TGA (Netzsch STA 449 F3 Jupiter) coupled with a Netzsch QMS 403C mass spectrometer was used to quantify H 2 O and CO 2 . Samples of circa 35 ± 0.5 mg were placed in an alumina crucible and exposed under an inert atmosphere of argon and increasing temperatures ranging from 40°C to 1050°C at a heating rate of 10°C/min. A blank curve, obtained under the same conditions with the same empty alumina crucible, was systematically subtracted.
The mass spectrometer was coupled to the TG to allow separation of concurrent mass loss from H 2 O and CO 2 releases, which were compared to a standard sample with known concentration (in this case calcite and gypsum) under identical analytical conditions [26]. The peak integral, i.e. the area under the MS H 2 O and CO 2 curves, was quantified using OriginPro 9 software for each sample.

Isothermal calorimetry
The fRCA may contain unreacted binder particles which in contact with water may dissolve and react. Isothermal calorimetry analysis was carried out to investigate possible reactivity of unreacted binder particles. Fines under 0.250 mm were used for experiments, since cement paste has a tendency to accumulate in the finest fractions. The control sample was cement paste, CEM I 42.5 N with water-to-cement ratio 0.5. All the raw materials were conditioned at the measurement temperature. Samples were prepared outside the calorimeter. Water was added to fines in order to study their reactivity (water-to-fines ratio 0.5). For a calorimetry test, the sample (mix of water and fines) is stored in a glass ampoule (125 mL) and placed in the calorimeter (3-channel TAM Air, Thermometric). The content of fines was 28.8 g. The calorimeter consists of 3 parallel twin type measurement channels: one from the sample, the other for the reference. A glass ampoule with water was used as a reference. Once the thermal equilibrium is reached, the heat evolution is recorded at the constant temperature (20°C). Three replicates were measured simultaneously.

X-ray powder diffraction and rietveld refinement analysis
Diffractograms of studied materials were recorded using a Bruker D8 Advance X-ray powder diffractometer, equipped with LynxEye detector. The machine was operated with an accelerating voltage of 40 kV and an X-ray beam current of 40 mA. The X-ray source was a Cu-tube working with characteristic Cu-Ka wavelength of 1.54060 Å. The samples were scanned varying the 2theta angles between 10°and 70°. For quantitative phase analysis, standard reference material (alumina, Al 2 O 3 , NIST Standard Reference Material 676a) was added and mixed for about 2 min. Phase identification was performed using Bruker Eva 4.2 software and appropriate databases. Rietveld quantitative phase analysis was performed using Bruker Topas 5.1 software.

Optical polarizing-and-fluorescence microscopy (PFM)
Mineralogical composition of fRCA is important since it can have both positive and negative impacts on the concrete properties if aggregates are altered/deteriorated as to unstable silica phases. Therefore, PFM study was performed to identify components present (clinker phases, supplementary cementitious materials (SCMs), reaction products like portlandite, ettringite), and any internal debonding, microcracking and deterioration processes (e.g. carbonation, ASR, sulfate attack), morphology of aggregates and the presence of any other materials (e.g. fibers), following the approach of Nijland and Larbi [27].
Twelve thin sections were made and examined in planepolarized and cross-polarized transmitted light at magnifications up to Â 400.
Sample preparation is based on Dutch CUR recommendation 89 [28]. The materials (0-4 mm) were sieved to prepare two groups of samples, respectively, sized 0.063-2 mm and 2-4 mm. The representative grains were chosen for each group (group I: 0.063-2 mm, group II: 2-4 mm).

Results
In this section, first, the characterization of the physical properties is presented, followed by the chemical characterization. Subsequently, mineralogy was determined by XRD and microscopy. Finally, fRCA particles are evaluated with optical polarizing-andfluorescence microscopy.

Appearances of fRCA
Aggregate shape and surface texture affect concrete mix workability and adhesion of the cement paste, respectively. An overview of different fRCA fractions are presented in Fig. 3. The smaller fractions show more homogeneous shapes. Visual observations of larger size particles show that the shape of the particles varies from round to angular. The key observations are: -the variation of particle morphology and colour within a particle population is low, -the variation of particle morphology and colour between river sand and fRCA is high because the fRCA has particles with high sphericity to angular or sub-angular with low sphericity cf. Power's scale of roundness [29]; and the presence of cement mortar in fRCA, -the variation of particle morphology between different fRCA (B, C, D) is small compared to that between fRCA and reference sand (Fig. 4), -the variation in colour of subfractions in the range 0-0.5 mm is high in Fig. 3: each batch is defined by a different colour.
Contaminants in fRCA are<5% and include wood, iron/steel, different plastics and polymers, glass and plant fibers as identified by stereomicroscopy.

Grading curves and standard limits of sand
The aggregate grading is an important characteristic as it affects, amongst others, workability and packing density of the concrete mix. The particle size distribution of different fRCA with range 0-4 mm is shown in Fig. 5. In addition, lower and upper limits according to NEN-EN 12620:2002+A1:2008 [30] are shown for assessment of the suitability of fRCA for use as a sand for concrete mix design. The fRCA from different batches complies with the overall limits in NEN-EN 12620:2002+A1:2008 [30].
Variations in particle size distributions between three samples within the same batch is low, meaning that tested samples are representative for a single batch. On the other hand, the variations of particle size distributions (S-shape) between different fRCA (B, C and D) is high (Fig. 5).
The different percentages for grain size fractions of different fRCA, like the higher percentages of fraction 0-0.250 mm in C and D (15-18 wt%) compared to B (~6 wt%) (Fig. 6) may be due to the number of crushing steps applied during recycling. For instance, source B has one step of crushing in the process, while C and D have three steps.
The particle size distribution of sands of particles sizes 0-0.250 mm analyzed by laser diffraction, is shown in Fig. 7. The fRCA, with a dominantly angular shape, has finer particle size distributions compared to river sand (A). It is interesting to note that median size of fRCA is larger compared to some of industrial byproducts such as fly ash, ground granulated blast furnace slag or metakaolin, but similar to that of CEM I 42.5 N and limestone powder (Table 1). Fig. 8 shows density results. For every fraction of fRCA (B, C, D) the density is lower than for river sand fraction (A); the largest difference occurs in the finest fraction, i.e. 0-0.250 mm. Typically, density of quartz and hardened cement paste are 2.65 g/cm 3 and 1.5-2.2 g/cm 3 , respectively. The density of fRCA is reduced    compared to density of pure quartz, but it is higher than that of pure cement paste. The BET specific surface area was measured for the fraction 0-0.250 mm (Table 1). fRCA has much higher specific surface areas than river sand. The BET specific surface area of fRCA is higher than BET values of cement, slag, fly ash and fillers (Table 1), but similar to those found by [35]. This high BET specific surface area of fRCA is likely due to presence of calcium silicate hydrate (CSH) phase in mortar adhered to sand particles. CSH has high specific surface areas ranging from 30-112 m 2 /g [36]. Furthermore, particle size distribution also differs between different fRCA (B, C, D) (Fig. 7) and affects their specific surface areas. Fig. 9a shows moisture content of river sand and fRCA upon delivery. The fRCA B has the highest free water content. The moisture content of studied fRCA was similar to that of studied fRCA by Lotfi et al. [37], which was 12 wt.%. Fig. 9b shows the large difference in water absorption between river sand and fRCA, the latter similar to values found by Ulsen et al. [38]. Table 2 lists cement paste content, chlorides and sulfates in different fractions. Content of old cement paste adhered to the fRCA ranges from 14 wt.% to 29.8 wt.%, depending on the batch (B, C, D) and particle size (0-0.250 mm, 0.250-4 mm, 0-4 mm). These are comparable to cement paste contents 16.1 wt.% and 24.6 wt. % for recycled coarse and fine concrete aggregates, respectively [4]. The cement paste concentrates in the finer fraction (0-0.250 mm), in agreement with [38].

Cement paste, chlorides and sulfates
In contrast to river sand (A), fRCA contains chlorides and sulfates (Table 2). fRCA C has a higher Clcontent than batches B and D ( Table 2). The water soluble sulfates content in the fRCA is below standard limit value 0.2 wt.% (NEN-EN 12620:2002 +A1:2008) [30], except for B 0-0.250 mm and D 0-0.250 mm. The high water soluble sulfates of fRCA B 0-0.250 mm and D 0-0.250 mm might be due to various causes, including breakdown of ettringite due to carbonation of cement paste [39] or contamination by gypsum.

Bulk chemical composition
The bulk chemical composition of the different grain size fractions is given in Table 3, expressed as oxides. The river sand is composed mainly of >93 wt.% SiO 2 and minor CaO,  Fig. 10 shows the TG-DTG curves for different fRCA. In addition, the MS H 2 O and MS CO 2 curves of materials are plotted in Fig. 11. Using MS curves, the precise temperature ranges for H 2 O and CO 2 emission can be determined. Using both the MS and TG-DTG curves, enables the weight loss calculation. After distinction between H 2 O and CO 2 , content of Ca(OH) 2 and CaCO 3 could be determined based on the independent decomposition reactions for Ca(OH) 2 and CaCO 3 , according to the equations:

TG-MS
where w ðH 2 OÞ and w ðCO 2 Þ are the weight losses determined by TG-MS.   Fig. 10 shows a three-stage weight loss in the TGA-DTG curves of fRCA samples. The first stage is due to the evaporation of physically and chemically bound H 2 O from CSH, monosulfate and ettringite between 105 and 430°C. The second stage of weight loss from 410 to 480°C is due to dehydroxylation of Ca(OH) 2 . The third stage can be divided to: a substage 250-700°C due to further dehydration of CSH [40,41] and decomposition of amorphous carbonates and semi-crystalline carbonates, and a substage 700-925°C due to decomposition of crystalline carbonates. The major weight loss was due to CO 2 emission and therefore carbonates decomposition. All fractions (except fraction 0-0.250 mm in fRCA (B)) have the characteristic peak for CO 2 emission at 775°C, suggesting the same origin of CO 2 regardless the batch (B, C, D) being calcite. Less strongly held CO 2 molecules in fraction 0-0.250 mm in fRCA (B) suggest presence of less stable forms of CaCO 3 , such either amorphous or the polymorphs vaterite and aragonite. The temperature range for main H 2 O emission is the same for all fractions, with the exception of the peak shape which is different. For batch (B), the peak for H 2 O emission is more narrow than for batches (C) and (D), suggesting different type (structures) of reaction products. From TG-MS H 2 O, it is clear that portlandite is present still in some of fRCA, its amount not exceeding 2 wt.% as shown in Table 4. Table 4 gives the overview of temperature ranges and corresponding weight loss for H 2 O and CO 2 , but also the total weight loss and content of Ca(OH) 2 and CaCO 3 . Table 4 shows that there was more H 2 O released from fraction 0-0.250 mm, compared to other two fractions. This is due to larger cement paste content in fraction 0-0.250 mm (Table 2) and therefore, more hydration products containing physically and chemically bound water. The CO 2 content presents sum of physically and chemically bound CO 2 . The CaCO 3 content presents the sum of Ca-carbonates (amorphous, vaterite, aragonite, calcite), limestone filler and aggregate (and any other carbonate bearing aggregates), shown to be present by PFM study (section 3.3.2).
Reactivity of unreacted binder particles The comparison of normalized heat release rate and cumulative heat between the cement paste and different fRCA (0-0.250 mm) is shown in Fig. 12 and Fig. 13. The hydration of cement shows two reaction stages (two calorimetric peaks) that are controlled by different reaction mechanisms (Fig. 12a). The first calorimetric peak, occurring within the first few minutes, mainly reflected the initial wetting/dissolution of cement. The second calorimetric peak, occurring after around 6 to 24 h, reflected the formation of hydration products. Fig. 12b presents only a small initial peak for all types of fRCA (B, C, D) compared to cement hydration kinetics (Fig. 12a). Based on the total cumulative heat results (Fig. 13), the reactivity of fRCA (from sources B/C/D) is negligible (<0.7 J/g of sand) compared to reactivity of cement (91.1 J/g of cement).

Mineralogical characterization
3.3.1. XRD Fig. 14 shows the X-ray diffractograms of investigated river sand and fRCA, while Table 5 gives an overview of the phases combined with quantification using Rietveld refinement method. It is apparent that total fraction and subfractions within each batch contain the same phases (Fig. 14). The main phases of fRCA are quartz and calcite. Besides quartz and calcite, albite, orthoclase and microcline are present in all aggregates; quartz makes up 60 wt.% (Table 5). Besides dolomite which is also observed in batch C, there were no other phases present in the fRCA. The amount of quartz is significantly less in the fRCA (B, C, D) compared to river sand (A). Each of the phases, their origin and content, is presented next.
Quartz. Occurring in all three fRCA (B, C, D) in high amounts, deriving from the original aggregate which is usually siliceous river or marine sand.
Calcite. Potentially derived from one of the following sources: aggregates (limestone, carbonate cemented sandstone; also the source of dolomite) and carbonation of cement hydrates either during concrete service life or after demolition and crushing of concrete.
Furthermore, Table 5 gives the total content of amorphous phase that is likely to be composed of amorphous hydration products in cement paste (e.g. CSH), unreacted binder components (such as slag, fly ash, amorphous clinker) and amorphous CaCO 3 from carbonation of fRCA. Unhydrated cement clinkers were not identified within the XRD detection limit. Fig. 15, an overview is given of the grain size fractions 0.063-2 mm and 2-4 mm for river sand and different fRCA (B, C, D). Particles of river sand are generally (sub) rounded while fRCA are more angular and often composite (Fig. 15). In the fRCA, the visual bonding between cement paste and aggregates is variable, from good to lack of adhesion as is the thickness of cement paste around natural aggregate grains.

Microscopic investigation 3.3.2.1. Mineralogy of aggregates. In
For fRCA 2-4 mm, four types of representative particles are distinguished (Fig. 15i):  (Fig. 16). In general, two subtypes can be distinguished: -particles of smaller (below 2 mm) and bigger (2-4 mm) sizes and different shapes in a confined space linked by cement paste, -particles of smaller sizes than 2 mm and different shapes in a confined space linked by cement paste; type IV: Loose cluster of particles, i.e. with a high aggregate-tobinder ratio, large amount of voids and poor internal coherence (Fig. 16). Their microstructure resembles that of a bedding or pointing mortar or plaster rather than concrete. In general, type IV particles contain two groups of particles: -particles of smaller sizes (below 2 mm) and bigger (2-4 mm), -particles of smaller sizes than 2 mm and different shapes in a confined space linked by cement paste.
For fRCA 0.063-2 mm, three types of representative particles are distinguished (Fig. 15ii): type I: Single aggregate particles, as in the 2-4 fraction, but in the range of 0.063-2 mm. type II: Several aggregate particles linked with cement paste, usually in a dense cluster of particles (as type III in the 2-4 fraction); in the fRCA 0.063-2 mm fraction, nearly all voids in these particles are clean, without any phases such as portlandite or ettringite. type III: Mortar particles, with a low content of aggregates and high content of cement paste. Fig. 17i shows fRCA with a weak cohesion between aggregate particles and cement paste, an example of poorly compacted concrete. In contrast, Fig. 17ii shows fRCA with strong bond between the hardened cement paste and sand particles.
The microphotographs in Fig. 18 show that the limestone is present in variable amount in each thin section, which may be up to about 10 wt.% of the aggregate. It originates from the aggregate source of the parent concrete. A small number of fired clay brick particles were also observed in fRCA.
The interfacial zone between aggregate and cement paste is often considered the weakest part of normal concrete, being the most feasible part for crack initiation and propagation. An example of an open aggregate-cement paste interface is shown in Fig. 19.
However, such open interfaces are negligible in fRCA; the interfacial zone between cement paste and original aggregate is visually good (Fig. 20). Nevertheless, different ITZs will be present in microstructure of hardened mortars and concretes with fRCA due to compositional differences among fRCA. In addition to compositional differences, differences in porosity of the adhered cement paste will also cause different ITZs. Some (entrapped) air voids and irregular voids due to lack of compaction are also present.

3.3.2.2.
Mineralogical composition of old cement paste. As might be expected, unreacted binder components common in the Netherlands such as Portland clinker, slag, fly ash and limestone filler are encountered in the old cement paste. Fig. 21 shows an example of calcium silicates (C 3 S and C 2 S). Similarly, Fig. 21 shows the unreacted slag in old cement paste typically 20-100 lm in size with well-developed microstructure of the cement paste. Unreacted fly ash spheres were also observed. The cement paste also contains occasional secondary reaction products. Minor amounts of undeleterious secondary ettringite were found in some air voids of fRCA (D batch) in the fraction 2-4 mm (Fig. 21). Fig. 22 shows the presence of limestone filler particles (B batch) with irregular shapes in the cement paste.
More important than the minor ettringite is carbonation of the cement paste, either during the service life of the parent concrete (Fig. 23), or, in case of the very thin outermost rim on the particles, during the storage after recycling of the parent concrete (Fig. 24).

Implications of characterization
Prior work has documented the use of fRCA, as sustainable alternative to natural sand, in the concrete. This concrete may exhibit similar or distinctive behavior from mortars/concretes with natural sand depending on the water absorption capacity of fRCA [15,16,[42][43][44][45]. However, these studies have not focused on other properties which might also influence the performance of concrete and should be considered for appropriate concrete mix designs with fRCA. In this study, comprehensive characterization of the three types of fRCA has been done. The results have important implications for future concrete mix designs with fRCA. Table 6 gives an overview of several physical, chemical and mineralogical quality indicators for fRCA in comparison to river sand. The indicators are: particle shape, particle size distribution, content of fraction 0-0.250 mm, density, specific surface area (BET), water absorption, Cland SO 4 2content, mineralogy of fRCA, cement paste content, milling time to complete crushing. Each of these indicators can be correlated to the concrete mix design with respect to the specific property of the mix. Particle shape. The fRCA particles are angular and elongated compared to round river sand. The angular shape and rough surface of fRCA particles result in more frictional resistance to the free flow and decreased packing of angular shapes vs. spherical shapes [46][47][48]. A maximization of the packing density of the particles can be achieved by adjusting the grading of the aggregates, consequently improving the overall performance of the concrete mix [11,[49][50][51].
Particle size distribution. The fRCA (B, C, D) show different particle size distribution compared to river sand, mainly due to differ- Note: Sum of individual weight losses (second, third, fourth column) is equal to the total weight loss (fifth column). The individual weight losses can be only calculated with combined TG-MS analysis. The total weight loss can be then compared to the total loss which can be determined from the TG curves.  ent content of the fine material (<0.250 mm). The impact of the fine material on the workability can be significant, depending on fRCA incorporation percentage. Therefore, it is important to control the particle size distribution of the fRCA within the grading limits for reference sand. Content of fraction 0-0.250 mm. Increased amount of fines (particles < 0.250 mm) increases the required amount of water to wet the particle surfaces adequately and to maintain a specified workability. The content of fine material (<0.250 mm) also affects the aggregate particle packing and significantly diminish the effect of superplasticizers on normal concrete rheology (flow characteristics). Therefore, NEN-EN 206 in combination with NEN 8005:2017 [52], limits the content and prescribes minimum required quantity of fine material (< 0.250 mm) per m 3 of concrete depending on the largest grain size (D max ), more fines being required in concrete with lower D max . This suggests that fRCA, with higher amounts of fine material (< 0.250 mm), is more suitable for concrete with lower D max . As showed in Fig. 2, the amount of fine material (< 0.250 mm) in fRCA (C and D) is about two times higher than in river sand (A), except for B sand. The content of fine material (< 0.250 mm) for B sand is 6 wt.%, while for C and D sand these values are 15 wt.% and 18 wt.% respectively. When the fraction 0-0.250 mm is removed from fraction 0-4 mm, chemical and mineralogical differences between both fractions are marginal. This is   due to the small part of fraction 0-0.250 mm which makes up of the total original fRCA fraction 0-4 mm (Fig. 2). The critical allowed content for fraction 0-0.250 mm in total fRCA fraction 0-4 mm should be established in the future. Specific surface area (BET). The increased BET specific surface area of fRCA compared to river sand is caused by the presence of porous adhered cement paste. In general, the porosity of cement paste can differ significantly, depending on the amount of paste, type of used cement, content of remaining reaction products [53], and carbonation of paste. For instance, carbonation makes CEM I-based binder less porous, while CEM III/B-based binder more porous [54]. The values of surface areas of studied fRCA are comparable to results of Bendimerad et al. [35], who found BET surfaces of 5.3 m 2 /g and 9.9 m 2 /g for the sand (0-4 mm) and fine (ii) fRCA D fRCA D ) i (    (<63 lm) fractions, respectively. An increase of surface area and greater significance of short-range, weak forces between the smaller particles below approximately 100 lm [55] will result in a decrease of packing density. The increased surface area of fRCA would also require more water to wet and more paste to coat and lubricate the aggregate particles. This has been observed previously by others [35] and it has been suggested that the effective water intended for the cement paste might be attracted and kept in the menisci created by the recycled fines. Therefore, the bleeding water might be lower than in the control mix and the plastic shrinkage related to insufficient curing may increase [35]. In that case relatively more water is necessary to overcome the curing.
Water absorption of fRCA. Results also show a considerably higher water absorption of fRCA than river sand. This is the sum of water absorption of original aggregate particles and water absorption of residual cement paste or mortar around them. The densities from helium pycnometry (Fig. 8) and water absorption ( Fig. 9) are comparable amongst different fRCA, although content of old cement paste and surface areas differ. This is contrary to what could be expected. Results in literature are differing. Whereas Zhao et al. [56] reported that the water absorption of fRCA is linearly proportional to the content of the old cement paste, Belin et al. [53] showed that two recycled concrete aggregate types with similar 24 h absorption values can have very different absorption kinetics. Rather than the amount of old cement paste present, its properties, like size, morphology and porosity, may explain the differences in water absorption [56]. PFM showed clear differences among these properties, which are to some extent determined by the original mix design(s) like water-to-cement ratio. The high water absorption of fRCA may have a detrimental effect since it affects concrete workability while it can have a beneficial effect when the fRCA is considered as internal water source for shrinkage  prevention. The high water absorption of the fRCA requires that the water absorbed by these aggregates is compensated by extra water during the mixing process [57].
Cland SO 4 2- [30] for recycled concrete aggregates, all investigated samples with particle range 0-4 mm, can be used in structural concrete.
Mineralogical composition of fRCA. Dutch concrete aggregates are typically river or marine sand and gravel, with a dominantly siliceous composition, i.e. monomineralic quartz and (impure) sandstone with minor (in river aggregate) to significant (in marine aggregate) chert. The minor constituents may be limestone or some volcanic rocks such as basalt. The fRCA is typically a blend of aggregates from different sources, possibly reflecting regional differences in aggregate and filler sources and cement types used. The original aggregate in the studied fRCA is also dominantly siliceous (Tables 3, 5 and Fig. 15). Other components, such as CaO, Al 2 O 3 , Fe 2 O 3 and MgO, reflect either differences in original aggregate composition, such as the presence of (dolomitic) limestone in the aggregate, and/or relative amounts of binders present in the adhering cement paste (e.g. CEM I vs. CEM III/B). The chemical composition of the 0-4 mm fraction of studied fRCA D is similar to that of the fraction 0-4 mm studied by Lotfi and Rem [37], which was obtained from Dutch recycled concrete based on the use of CEM III/B cement. Different results from other studies [3,58] illustrate the mentioned role of parent concrete origin. PFM showed minor amounts of clinkers, SCMs (slag, fly ash) and some secondary ettringite in old cement paste. The mineral composition and surface microstructure of the fRCA will affect the growth of new hydration products and the bonding between the new and old cement matrices, which might affect the mechanical properties of new mortars/concretes with fRCA. Substantial content of carbonates was measured for fRCA by TG-MS (Table 4) and XRD-Rietveld refinement (Table 5), especially for the specimens with particle range 0-0.250 mm. Microscopy shows that Ca-carbonates are present as original limestone filler and aggregate as well as carbonated cement paste (Section 3.3). Differences between TG-MS (Table 4) and XRD Rietveld refinement ( Table 5) in terms of Ca-carbonates amount, may be due to the presence of amorphous or poorly crystalline CaCO 3 and Rietveld fitting parameters (possibly not suitable for all types of carbonates present).
Reactivity of unreacted binder particles. In general, the content and composition of adhered cement paste in fRCA will depend on the parent concrete (type of cement used in concrete, concrete water-to-cement ratio), recycling technique and storage of fRCA. PFM shows that unreacted binder particles such as clinkers or SCMs (slag, fly ash) are present in the adhered cement paste. The common presence of SCMs, in particular slag, is in accordance with type of cement commonly used in the Netherlands. In the Netherlands, slag cement represents over 70 % of the cement market share [59]. Whether the unreacted binder particles in studied fRCA still represent any reactivity may be doubted. Based on the calorimetry results (Fig. 13), the contribution of studied fRCA to heat development, was negligible. This might be due to low amount of unreacted binder particles, their fineness and their chemical and mineralogical composition. Although not identified within the XRD detection limit, unhydrated Portland clinkers (such as C 2 S, C 3 S) were found in minor amounts in the thin sections by PFM. Part of the clinker and SCMs particles were relatively coarse (e.g. diameter > 63 mm), which may both be due to the fact that the smaller ones have reacted and Portland clinker and slag were usually more coarse in the past. Cementitious reaction products (e.g. CSH) are considered to be nonreactive, unless thermally reactivated [60,61].
Milling time to complete crushing of different fractions 0-4 mm has shown that the milling of fRCA compared to the river sand is much faster. This parameter may help to determine the  final application of the fRCA, such as type of substitution (sand, filler, binder). As milling time has indicated, the hardness of fRCA is low compared to river sand. It is believed that the high content of cement paste and angular particle shape of fRCA, enabled faster milling of fRCA compared to river sand. Microcracking. A good interface between adhered cement paste and original parent concrete aggregates is essential to resist forces in the bond plane and it is important in determining mechanical and durability properties. Based on PFM observations, the presence of microcracks in the fRCA was negligible. Nevertheless, different ITZs will be present in microstructure of hardened mortars and concretes with fRCA due to compositional and porosity differences among fRCA.
The results implied that physical, chemical and mineralogical properties of fRCA are dependent on the parent concrete, recycling technique and storage of fRCA. Mixing of different types of construction materials at the recycling plant (e.g. traditional, ultra-high strength concrete, light-and heavyweight concretes, fiber reinforced concrete, etc.) which may have been exposed to different environmental conditions (under chloride ingress, carbonation, chemical attack, alkali-silica reaction, etc.), caused the compositional variations (SiO 2 and CaO) and levels of contamination (Cl-and SO 4 2content) among investigated fRCA (Table 2 and Table 3). The way the fRCA are generated, will have a direct impact on the particle size distribution and quantity of the fines 0-0.250 mm, as demonstrated in Fig. 2 and Fig. 5. For instance, the increased number of crushing steps will increase the content of fine material (sand C and D have larger content of fraction 0-0.250 mm compared to sand B). The outdoor storage of fRCA causes carbonation of outer layers of sand piles, and agglomeration of the material. As shown in Fig. 1, agglomerates remain in sands C and D after drying.
The fRCA characterized by large mineral content, low contamination, negligible reactivity and low density reduction implies the potential of Dutch fRCA from different parent concretes to be used in the production of blended sands with similar performances to the natural sand. The findings provide guideline for selecting fRCA for concrete mix and support the broader utilization of fRCA in concrete in practical applications with reduced material cost and environmental impact. In the further studies, it will be investigated which parameters (quality indicators) govern workability and strength of mortars/concretes.

Conclusions
This study investigated the physical, chemical and mineralogical properties of Dutch fRCA from unknown concrete by means of multiple characterization techniques combined.
The following conclusions can be drawn based on this study: (1) The amount of fine material (<0.250 mm) in fRCA (C and D) is about two times higher than in river sand (A), except for B sand. Similarly, their surface area was greatly increased compared to river sand. Even so, it was demonstrated that when fraction 0-0.250 mm is cut from the sand fraction 0-4 mm, the chemical and mineralogical differences between fractions 0.250-4 mm and 0-4 mm are marginal. (2) The water absorption of fRCA was high and reduction of density was low, and their values were similar for different fRCA. Although the water absorption was similar, the water absorption kinetics of different fRCA is believed to be different due to the different morphologies of the cement patches at the surface of the grains. This assumption is based on the PFM observations, where the cement patches varied, from (i) fRCA C batch (ii) fRCA C batch  smaller and thinner to larger and thicker. Microcracking was absent at the interface between adhered cement paste and original parent concrete aggregates. (3) Although the content of the cement paste in fRCA was large (on average 22 wt.%), the content of chlorides was minor. In addition, the content of sulfates was below the critical values adopted for aggregates for use in reinforced concrete, with exception for fine fraction 0-0.250 mm. Therefore, from the aspect of contamination, the fRCA (0-4 mm) has large potential for sand replacement in new concrete. (4) Based on XRD characterization of different fRCA, it was found that they are mainly made up by crystalline phases (>70 wt.%), notably quartz (>60 wt.%). PFM study supported XRD results and revealed minor minerals such as clinkers and secondary ettringite or distinguished between limestone aggregates and CaCO 3 formed by carbonation of cement paste. The fRCA (B, C, D) fractions (0-0.250 mm, 0.250-4 mm, 0-4 mm) show some differences in chemical composition, notably for SiO 2 and CaO contents. TG-MS analysis revealed minor amount of portlandite. (5) Whether the unreacted binder particles in studied fRCA still represent any reactivity may be doubted. Based on the calorimetry results, the contribution of studied fRCA to heat development, was negligible. Cementitious reaction products (e.g. CSH) are considered to be nonreactive, unless thermally reactivated. (6) This study suggests that the determination of the chemical composition can provide first line control regarding composition and potential contamination of fRCA. In addition, grading (particle size distribution) and water absorption (kinetics) should be measured. After that, it can be decided which additional tests are necessary (if necessary) to be done in order to evaluate the suitability of fRCA for replacement of natural fine aggregates in new concrete.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.