Shrinkage of blended cement concrete with fly ash or limestone calcined clay

This study investigates the shrinkage of two sustainable aluminosilicate blends with fly ash or limestone-calcined clay (LC3). Paste and concrete were prepared using these SCMs for the highest possible replacement of binder without compromising the strength. The chemical and autogenous shrinkage were assessed for paste samples and further investigation were conducted on hydration by thermogravimetric analysis (TGA) and Fourier transform infrared spectroscopy (FTIR). Opting for an engineering approach, comparison among different segments of shrinkage i.e., autogenous, drying and total shrinkage of concrete having a specific compressive strength were considered. The initial investigation on paste samples highlighted the dissimilarities in shrinkage and hydration of fly ash and calcined clay. LC3 hydrated faster compared to fly ash leading to greater autogenous shrinkage. The high autogenous shrinkage in the LC3 blend was compensated by a low drying shrinkage for a specific compressive strength. Considering the replacement level of cement, shrinkage, and ecological impact, LC3 proved to be a more sustainable and eco-friendly concrete compared to fly ash.


Introduction
Though concrete is not mandatory for human survival, it is the second most-consumed substance in the world having a per capita consumption of two cubic meters annually [1,2]. It is the most popular building material due to its versatility and durability. Typically, concrete is composed of 10-15% of Portland cement (PC). Paste, which is the mixture of Portland cement (PC) and water, is the main component of concrete to impact the mechanical and durability performance including shrinkage. The production of clinker in cement is associated with CO 2 emissions due to the thermal decomposition of limestone (raw material) and associated combustion of fossil fuel to produce the required energy [3]. Even the highest maximum thermodynamically possible efficiency in the modern kiln was unable to reduce the emission and energy consumption [4]. In 2020 alone, the estimated worldwide cement production was 4.1 giga metric tons (Gt) [5] and this accounts for 5-8% anthropogenic CO 2 emissions and 90% of industrial CO 2 emission [3,[6][7][8][9]. One of the most promising ways to reduce clinker production and subsequent CO 2 emissions is to use Supplementary Cementitious Materials (SCMs) to replace Portland cement (PC) in concrete [10]. In addition, using SCMs as PC replacement leads to economical and durable concrete in the long run and further contributes to a sustainable environment by utilizing industrial waste or by-products [11].
Aluminosilicates such as fly ash have been used as SCMs in the concrete industry for decades. Fly ash is the by-product of pulverized coal combustion in the thermal power plant which is essentially the most polluting means for the generation of electricity. Moreover, fly ash itself can cause various environmental problems if handled improperly during construction [12,13]. Despite the environmental issues, fly ash has been used as a replacement for PC due to its pozzolanic reactivity [14][15][16][17]. However as reported in 2018, only 10% of the total cement demand for concrete around the world can be replaced by good quality fly ash [2,18]. Additionally in recent times, the coal-based power plants are being shut down and the production of fly ash is reducing. Other conventional SCMs like slag also has less supply not meeting the demand [18]. These factors acted as a driving force to consider the introduction of new materials and one of the promising cements is the ternary blend of cement with limestone and calcined clay (LC3).
Both limestone and calcined clay are abundantly available making the blend one of the most economical and desirable options [2,18]. Calcined clay is produced by igniting kaolinite rich clays at 700-850°C as opposed to the high temperature (1450°C) required for clinker production [18][19][20]. Moreover, the calcination of the clays does not involve the CO 2 emission due to the decarbonization of limestone in the cement kiln. Thus, LC3 is a sustainable and environmentally friendly cement that significantly reduce clinker usage and replace other SCMs such as fly ash and slag.
Both fly ash and calcined clay are aluminosilicates having similar chemical compositions and reaction mechanisms. Low calcium fly ash is mainly pozzolanic only, which is similar to calcined clay [21,22]. Fly ash shows a late pozzolanic effect and reacts with calcium hydroxide to form C-(A)-S-H [23,24]. Similarly, calcined clay is a pozzolan, but it is reported to react at an early age [18,25]. Moreover, in a LC3 mix, the limestone can also react and produce mono and/or hemicarboaluminate [26,27]. There is a synergy between calcined clay and limestone enhancing each other's reactivity [25,28]. Therefore, the commencement of reaction in the two aluminosilicates, i.e. fly ash and calcined clay is different and that can influence the subsequent degree of hydration and other related parameters.
Shrinkage is one of the mechanical behaviour of concrete closely related to the mineral admixture used, their proportion to cement, the fineness of the binders and curing [29][30][31]. The shrinkage (in other words, the total shrinkage) is a combination of chemical, autogenous and drying shrinkage each of which dominate different ages of hydration. While chemical shrinkage is entirely associated with the hydration reactions, the autogenous shrinkage is also influenced by hydration induced self-desiccation [32][33][34][35]. The drying shrinkage occurs due to the physical loss of water from the pores of hardened concrete [36]. While shrinkage is well correlated with the concrete mix design, it is often estimated based on the compressive strength. For practical purposes, the knowledge of the extent of shrinkage development for a particular strength grade of concrete is vital for the selection of appropriate materials. In this study, an engineering approach was considered to evaluate two different strength grades for fly ash and LC3.
The replacement level of the SCMs significantly influences the hydration and all related parameters of the cementitious system. On the other hand, to achieve the goal to produce sustainable concrete, the highest possible replacement level is desired. However, this highest feasible replacement level depends on the material type and comparable mechanical properties. Between fly ash and LC3 blends, fly ash blends show low early strength and LC3 blends can achieve satisfactory strength from an early age due to the synergy between the calcined clay and limestone [37][38][39][40]. As much as 50% fly ash has been reported in high-volume fly ash mixes [41]. However, Singh and Garg [42] used 25% fly ash replacement in ordinary concrete without sacrificing any mechanical strength. For LC3 concrete since 2012 following Antoni et al. [37], 45%-50% replacement by calcined clay and limestone at a ratio of 2:1 is the most popular mix design [25,28,[43][44][45][46][47][48][49]. In this study, the highest possible replacement levels were selected for respective SCMs based on literature. This paper aims to advance knowledge that facilitates the selection of suitable aluminosilicate as a cement replacement in sustainable concrete for a specific target strength in term of shrinkage. Shrinkage largely governs the risk of early age concrete cracking [49,50] which is one of the major barriers for new concrete technology adoption by the industry. Though the shrinkage of fly ash blends were extensively investigated [51][52][53][54][55][56], very little data on the shrinkage of LC3 mixes available [45,57]. Moreover, the effect of the early commencement of limestone-calcined clay reaction on shrinkage was not explored previously in comparison to the pozzolanic effect of fly ash. In this study, two aluminosilicate blends with high amounts of respective SCMs i.e., 30% replacement by fly ash and 44% replacement by calcined clay and limestone were investigated and their shrinkage was compared in light of the difference in their hydration. Firstly, the general trend of autogenous shrinkage was identified on paste samples. Secondly, the shrinkage was explained by analyzing the hydration. Finally, the shrinkage and ecological impact of concrete samples with 30% fly ash and 44% LC3 having 25 MPa and 32 MPa compressive strength were compared.

Materials
In this study, two types of aluminosilicate blends were investigated. The binders used were General Purpose (GP) cement, two aluminosilicates i) fly ash and ii) calcined clay and limestone. Table 1 presents the  oxide compositions of the binders obtained by wavelength-dispersive X-ray fluorescence (WD-XRF). General Purpose (GP) cement can contain up to 7.5% mineral additives including limestone in accordance with AS 3972 [58]. The fly ash (Blue Circle fly ash) was supplied by Boral Cement. This fly ash had a combined amount of SiO 2 , Al 2 O 3 and Fe 2 O 3 exceeding 50% and CaO less than 18% (see Table 1) and thus conformed with the chemical composition criteria of Class F fly ash in ASTM C618 [59]. The calcined clay was produced by the rotary kiln method and was readily available. The calcined kaolinite content of the calcined clay was estimated using R3 calorimetry test [28] and that content was approximately 47.5% (Afroz et al.,unpublished results [60]). Omya Australia supplied the limestone under the commercial name of Omyacarb. The mineralogical phases present in the raw materials were quantified by Rietveld analysis of X-ray diffraction (XRD) patterns and silicon was used as the external standard. To ensure adequate intensity to noise ratio for Rietveld analysis, the X-ray diffraction (XRD) patterns were collected with a PANalytical X'Pert diffractometer (CuKa radiation, k = 1.54 Å ) with the following settings: range of scanning of 5°to 70°2h, scan rate at 0.0839 deg/sec and step size of 0.0131°. From the phase composition in Table 2, the amorphous content of fly ash and calcined clay were 85.6% and 84.0% respectively. The amorphous content of calcined clay is higher than the estimated kaolinite content (47.5%) before the calcination. This may be due to: i) the error of the estimation using R3 calorimetry test [28], ii) the additional amorphous content in the calcined clay [22], iii) the formation of glassy silica during cooling after the calcination [61]. The particle size distribution of the raw binders was determined by a laser diffraction particle size analyzer (Malvern Mastersizer 2000) and is presented in Fig. 1.
The fly ash was coarser and both calcined clay and limestone were finer than the GP cement. For concrete, Sydney sand and 10 mm basalt were used as fine and coarse aggregate. The Sydney sand had specific gravity and water absorption of 2.65 and 3.50% respectively. The basalt aggregate had specific gravity and water absorption of 2.80 and 1.08% respectively. The particle size distribution of fine and coarse aggregate by sieve analysis [62] is presented in Fig. 1.

Sample preparation
Three paste mixes were prepared with a water to binder ratio of 0.4, i) control with no replacement of GP cement, ii) 30F with 30% replacement by fly ash and iii) 44L with 44% replacement by calcined clay and limestone. In 44L, the ratio of added 44% calcined clay and limestone was 2 to 1. However, the GP cement was assumed to have about 6% limestone (5.2% calcite from Table 2) which essentially resulted in about 50% clinker content in 44L. No gypsum optimization was carried out for any of the mixes to investigate the effect of direct replacement by SCMs on cementitious systems.
The paste was mixed according to ASTM C305 [63] and three types of samples were prepared for, i) chemical shrinkage, ii) autogenous shrinkage and iii) microstructural analysis. The details about the shrinkage sample preparation will be described in Sect. 3.2. For the preparation of microstructural analysis samples, the paste after mixing was poured into cylindrical plastic containers having 50 mm diameter and 100 mm height. The containers were sealed with plastic lids and covered with parafilm. The prepared samples were cured inside an environmental chamber at 23°C till testing.
Three mortar mixes were prepared to determine the compressive strengths of the blends. These companion mortar mixes had identical binder components (no replacement, 30% fly ash and 44% calcined clay and limestone) and water-to-binder ratio (0.4) of to the paste mixes and termed as i) control-M, ii) 30F-M and iii) 44L-M. The mortar samples were mixed according to ASTM C305 [63] with a binder to fine aggregate ratio of 1:1. Mortar cubes (50 9 50 9 50 mm) were casted, demolded at 24 h, and cured under water in a temperature-controlled room at 23 ± 2°C till testing.
A total of six concrete mixes were prepared with different binders having the same replacement levels as the paste samples corresponding to each binder. Two strength categories were considered, (i) 25 MPa and (ii) 32 MPa. Concrete mix designs are given in  MPa are the most used concrete grades in practice which means potentially to highest embodied carbon savings. As the water-to-binder ratios were high in this study for low grade concrete, adequate hydration reactions of the binders were ensured and thus could be reasonably compared with the paste results. Before mixing the aggregates, both fine and coarse, were brought to saturated surface dry (SSD) condition. The aggregates were dried for 24 h at 105°C and after that absorption water corresponding to absorption % was added before 24 h of casting. The binder and the aggregates were mixed in dried condition for 2 min, then water was added, and then the concrete was further mixed for 5 min. Superplasticizer (MasterGlenium SKY 8100; a polycarboxylic ether polymer) was added if necessary at a dose below 0.15% of binder content to ensure negligeable effect of superplasticizer on shrinkage [57]. Two types of concrete samples were prepared, i) cylinders (100 mm in diameter and 200 mm in height) and ii) prisms (75 9 75 9 280 mm). After mixing, the samples were covered and kept in a temperature-controlled room at 23 ± 2°C and relative humidity of 50 ± 3%. All of the samples were demolded after 1 day and cured. The curing conditions of the concrete samples are described in the experimental methods (Sect. 3.5).

Heat of hydration by isothermal calorimetry
The heat of hydration of paste samples was monitored by isothermal calorimetry. A TAM Air calorimeter from TA instruments was used to monitor the heat evolution according to ASTM C1702 [64] up to 7 days at 23°C. Approximately 7 g of the paste sample were mixed inside a glass ampule thoroughly. After mixing, the ampoule was sealed with an aluminum cap and plaed inside the calorimeter. The corresponding reference sample was prepared using quartz sand. Two replicates were prepared for each mix.

Shrinkage of paste
Chemical shrinkage specimen was prepared with paste according to a modified version of ASTM C1608 [65] as described by Zhang et al. [66]. Erlenmeyer flask was used instead of small glass vials and the thickness of paste in the flask was maintained at 7-8 mm for all samples [66]. Two replicates were prepared for each mix. The data acquisition was as per ASTM C1608 [65]. Chemical shrinkage was monitored for 170 days, and the temperature of the samples was maintained at 23 ± 2°C.
The paste was used to prepare the autogenous shrinkage samples in corrugated tubes and three replicates were prepared for each mix. The samples were prepared according to ASTM C1698 [67]. The plastic corrugated tubes supplied by Germann Instruments was filled with paste and the ends were sealed with caps according to the code. To mitigate the bleed water absorption and subsequent expansion, the samples were rotated at 4 rpm as suggested by Mohr and Hood [68] for 24 h. After 24 h, the samples were stored in an environmental chamber at 23°C. The first measurement was taken at the final setting time with a horizontal dilatometer; also supplied by Germann Instruments. The final setting time was determined by isothermal calorimetry for each mix according to the method described by a previous study [69]. Initially, the measurements were taken frequently, and the frequency was reduced with time as needed when the shrinkage stabilized.

Microstructural analysis of paste
At the specific age for the analysis, each paste sample was removed from the cylindrical plastic containers. Then, the cylinders were broken into fragments. Paste samples for attenuated total reflection (ATR) -Fourier transform infrared (FTIR) analysis were prepared using the solvent exchange method to stop the hydration. At predesignated days, some of the fragments were ground with a Fritsch Pulverissette 6 planetary mono mill and the powders were passed through an 850 microns sieve following Snoeck et al. [70]. The powders were immersed in isopropanol (IPA) for hydration stoppage by solvent exchange according to a modified version of RILEM technique [71]. Every 3 g powder were immersed in 100 ml IPA. Though the RILEM technique [71] advised 15 min as the duration of immersion, the powders were treated for 60 min and the solution was stirred with a magnetic stirrer. The powders were vacuum filtered by membrane filter using a Büchner funnel and rinsed twice with diethyl ether. In a recent study, Zhang and Scherer [72] also advised a longer exchange period for proper hydration stoppage. The filtered powder on the membrane filter was dried in a ventilated oven at 40°C for 30 min. This IPA treatment ensured negligible carbonation of the samples. The dried powder was collected inside plastic vials and stored inside an N 2 atmosphere glove box till testing. ATR-FTIR spectra were collected for the IPA treated powdered samples with a Perkin Elmer Frontier IR spectrophotometer with a 3 mm diameter diamond/ZnSe crystal. Each spectrum of collected for wavenumbers from 4000 to 500 cm -1 with a resolution and accumulation of 2 cm -1 and 32 respectively. Solvent exchange can cause additional weight loss at high temperatures. As Zhang and Scherer [73] reported, excess thermogravimetric analysis (TGA) weight loss was observed at temperatures higher than 600°C for samples treated with solvents. Therefore, TGA was carried out using the powdered oven-dried paste samples at 1, 3, 7, 14, 28, 56 and 90 days. The remaining fragments of paste samples were dried in an oven for 24 h and then ground with a Fritsch Pulverissette 6 planetary mono mill. The powders were passed through an 850 microns sieve. The ovendried powdered samples were also stored inside the N 2 atmosphere glove box. For each test, approximately 30 mg samples were analyzed in an alumina crucible using a Q600 SDT from TA instrument. The temperature was increased from room temperature to 900 oC at 10°C/min using a nitrogen purge of 50 ml/min.

Strength of mortar
Compressive strengths of the mortar cubes were measured according to ASTM C109/C109M [74] after 7 and 28 days of moist curing.

Strength and shrinkage of concrete
Three cylindrical concrete samples were used to determine the compressive strength of each concrete mix according to ASTM C39 [75] at 28 days. Before the test, the samples were moist cured under 23 ± 2°C for 28 days.
The prism (75 9 75 9 280 mm) samples were used to measure the shrinkage of concrete. The shrinkage was segmented into two parts and the experimental procedure for each part was followed accordingly. For autogenous shrinkage determination, the prism samples were wrapped with an adhesive aluminium tape after demolding to prevent any moisture exchange and stored in a temperaturecontrolled room at 23 ± 2°C. The total shrinkage prisms were demolded and directly kept in a temperature-controlled room at 23 ± 2°C and relative humidity of 50 ± 3%. For all cases, the initial measurement was collected at 24 h on three replicates with a vertical length comparator. After that, subsequent measurements were taken with a short interval at an early age before decreasing the measurement frequency with stabilization of shrinkage with time up to 100 days.

Compressive strength
The compressive strengths of two sets of concrete mixes showed at least 25 MPa and 32 MPa respectively (Table 3) which were used as a basis of separation into two groups. The authors acknowledge the limitations of the mix design of the blended mixes.
Compressive strength of mortar mixes showed the effects of fly ash and limestone-calcined clay inclusion on strength development (Table 4) which is closely related to the hydration process. At 7 and 28 days, control-M exhibited the highest compressive strength followed by 44L-M and 30F-M respectively. Even with 56% replacement of GP cement, 44L-M performed better compared to 30F-M regarding strength development. At 28 days, the compressive strength of 44L-M was similar to control-M.

Chemical shrinkage and heat of hydration
Chemical shrinkage is the reduction in absolute volume of paste due to hydration as the volume of the products are less than that of the reactants [35]. Chemical shrinkage describes the earliest deformation in cement paste immediately after mixing. Calcined clay has been reported to react as early as 24 h [25]. Although the class F fly ash does not participate to the early age hydration and initially act as a filler [76], the chemical shrinkage was considered with respect to the binder content instead of cement content to account for any possible pozzolanic reaction of calcined clay. Similar to chemical shrinkage, the heat of hydration of cementitious materials are often used to estimate the rate of hydration at a very early age. This section presents the chemical shrinkage and heat of hydration of the different paste systems and their interrelation. Figure 2a presents the chemical shrinkage of paste with respect to the binder content up to 170 days. The control mix showed a plateau after about 50 days. The chemical shrinkage of the 30F sample was the lowest up to 20 days and after that, it continued to increase with a slow rate, indicating the late pozzolanic behaviour. The chemical shrinkage of the 30F seems to surpass that of control at 84 days. The chemical shrinkage of 44L increased faster than that of 30F for up to about 20 days (see Fig. 2a and b). However, it reached a plateau at about 20 days and then showed the lowest chemical shrinkage at the end of the measurement period. Figure 2b plots both chemical shrinkage (dashed lines) and heat of hydration (solid lines) at an early age and highlights that the chemical shrinkage correlates with the early age hydration for the three different paste mixes. As shown in Fig. 2b, the overall trends of chemical shrinkage development over 7 days were very similar to the corresponding trends of the cumulative heat of hydration evolution. However the order of heat of hydration (control [ 44L [ 30F) did not match the order of the chemical shrinkage (44L [ control [ 30F) for 7 days of hydration. At the early hydration period (up to 7 days), the chemical shrinkage of the 44L paste was comparable to or slightly higher than that of the control paste (Fig. 2b). Control and 44L pastes had higher chemical shrinkage compared to 30F paste although 44L pastes had less amount of GP (56% of binder) than 30F (70% of binder).
It is also interesting to note that for 12 h, the heat of hydration of the 44L paste was almost identical to that of the control despite 56% of GP cement replacement. The filler effect and the increased effective w/c due to the addition of limestone and calcined clay could play a significant role in the hydration rate and compensate for the dilution effect (due to less amount of GP cement in 44L than in the control paste). However, these two effects that accelerate the hydration of GP cement could also be valid for the case of the 30F paste but 30F exhibited less heat of hydration. This strongly indicated the participation of the early hydration reaction of limestone and calcined clay in 44L paste. For 12 h of the very early hydration, the chemical shrinkages of 30F and 44L were slightly higher than that of the control sample. The above observations confirmed that limestonecalcined clay blend in 44L paste caused faster initial hydration reactions than the fly ash blended cement did. The combined effects of early filler effect and early pozzolanic reaction in 44L mix led to the highest chemical shrinkage till 20 days hydration, which was also supported by the TGA results to be discussed in Sect. 4.3.2. However, 44L showed the lowest chemical shrinkage after 20 days compared to the other mixes indicating the absolute volume change due to hydration had a minor influence on the shrinkage of 44L paste at later ages. The rate of early age chemical shrinkage development has the potential to influence the early-age autogenous shrinkage of the paste and will be discussed in Sect. 4.2.2.
After the initial accelerated hydration (up to 0.2 days or about 5 h), the heat of hydration and chemical shrinkage varied almost linearly for all mixes up to 7 days (Fig. 2c). The slope of the linear part (shaded in Fig. 2c) were 8.96, 8.27 and 8.10 kJ/ml for control, 30F, and 44L respectively. A lower value of the slope indicates that less heat was produced to achieve a similar extent of chemical shrinkage for blended mixes. This may be the result of different hydration products or reactions involved in the hydration of pozzolanic materials.

Autogenous shrinkage
The autogenous shrinkage is closely related to the hydration reactions because it is partly generated by chemical shrinkage. During the hydration process, the relative humidity in the capillary pores in pastes reduces causing self-desiccation. The self-desiccation produces stresses on the pore walls and subsequent  macro shrinkage. This shrinkage due to self-desiccation contributes significantly to the autogenous shrinkage [32][33][34]. In addition, the time-dependent refinement of the pore structure can further increase autogenous shrinkage [77][78][79][80].
The autogenous shrinkage of paste samples was monitored for up to 200 days and the results are presented in Fig. 3a. To compare the performance of 30F and 44L samples with the control sample, Fig. 3b shows the percentages of autogenous shrinkage of 30F and 44L based on that of the control sample at each day.
The autogenous shrinkage was higher for both 30F and 44L samples at 1 day compared to the control sample ( Fig. 3a and b). This may be associated with the accelerated initial hydration of 30F and 44L samples, as seen from chemical shrinkage results (Fig. 2b) because the high chemical shrinkage originating from rapid hydration can cause both high selfdesiccation and high autogenous shrinkage. In fact, the autogenous shrinkage of 44L at 1 day was about 3 times higher than that of control (5.4%). Due to the filler effect and the increased effective w/c in 44L explained in Sect. 4.2.1, the pore structure for LC3 mix (i.e. 44L) was expected to be more refined even at a very early age, which was also reported in several studies [37,81,82]. The combined effect of the faster hydration and finer pore structure seems to cause the higher autogenous shrinkage than those observed from both control and 30F at 1 day of autogenous shrinkage.
After 1 day, the autogenous shrinkage of 30F was significantly lower than that of the control sample up to 7 days. In previous studies, fly ash blends also showed a lower autogenous shrinkage compared to other binders [51,78]. However, Termkhajornkit et al. [51] stated that the reduction in autogenous shrinkage with fly ash was strongly dependent on the replacement level. After 7 days, their relative differences started to diminish (Fig. 3b) when compared with the autogenous shrinkage of the control sample. This indicates that the rate of autogenous shrinkage development of 30F after 7 days was greater than for control sample. A possible explanation for the increasing rate of autogenous shrinkage development in 30F after 7 days can be the onset of slow pozzolanic reaction of fly ash causing further hydration and subsequent self-desiccation.
Unlike 30F, the autogenous shrinkage of 44L sample was higher than that of the control sample during the entire monitoring period (Fig. 3a). After 7 days the rate of autogenous shrinkage development of 44L was similar to that of control (see Fig. 3b). Thus, the relative autogenous shrinkage of 44L relative to that of the control sample varied within the range of 143% to 155.5% after 7 days (see Fig. 3b). Previous works on LC3 autogenous shrinkage were contradictory reporting both decrease [46,83] and increase [84].
In this study, autogenous shrinkage was the highest for 44L and the lowest for 30F samples. The different hydration mechanisms of the two different aluminosilicate binders seemed to influence the autogenous shrinkage differently. The hydration of calcined clay and fly ash will be investigated in Sect. 4.3 and correlated with their shrinkage behaviour.

Evolution of chemical bonding structure (FTIR)
The presence of different raw materials and hydration products were observed by FTIR at 1, 3, 7, 14 and 28 days. Figure 4 presents several distinct peaks over a range of wavenumbers and the peaks indicate absorbance intensities. For solid samples, the absorbance intensity can be influenced by the pressure applied by the ATR unit and the particle size of the sample [85,86]. The ground samples were sieved thus the particle sizes of the samples can be considered similar for all samples. Identical pressure was applied each time during spectra collection. Therefore, the absorbances of different samples can be considered free from these limitations and comparable. For the identification of the peaks of interest, the spectra were considered between the wavenumbers of 3700-3600 cm -1 and 1050-850 cm -1 . Figure 4a1, a2 and a3 feature the peaks corresponding to calcium hydroxide (CH) at about 3644 cm -1 wavenumber due to stretching vibration of O-H [87][88][89] of control, 30F, and 44L respectively. This study is also interested in two other peaks shown in Fig. 4b1, b2, and b3. The peak located at around 950 cm -1 due to Si-O stretching indicates the main hydration product C-S-H [87] and the peak around 874 cm -1 wavenumber was due to the bending (t 2 ) of CO 3 2present in limestone [88,90,91].
The calcium hydroxide produced by the hydraulic reaction of the clinker in the blended mixes, 30F and 44L, was lower compared to control because of the dilution effect, also seen in previous studies [92,93]. Fly ash reacted at later ages producing hydration products and consuming calcium hydroxide [23,24]. The process starts as the amorphous silica and alumina from fly ash dissolve in the pore solution with high pH [24]. Therefore, the commencement and the extent of the pozzolanic reaction of fly ash depends on the reactivity of the fly ash and the availability of calcium hydroxide from the hydraulic reactions of clinker [93,94]. From Fig. 4a2  the reduction in the CH peak in Fig. 4a3. The pozzolanic reaction of fly ash and calcined clay was also supported by the TGA data in Sect. 4.3.2. The pozzolanic effect was so intense that the CH peak at 28 days for 44L had minute absorption intensity. Calcined clay has high early pozzolanic behaviour which is well supported by previous studies [18,20,28]. The limestone present in the LC3 mix i.e., 44L can further promote the consumption of CH as CH can participate in the formation of carboaluminates from limestone reaction [95]. The calcium carbonate in limestone reacts with alumina to produce mono and hemi carboaluminate [26,27]. As seen in Fig. 4b2 and b3, the calcium carbonate peak at around 874 cm -1 in both 30F and 44L samples showed a diminishing trend. The peak in 30F seems to start to decrease at around 28 days whereas this peak in 44L started to reduce significantly from 3 days, indicating the reaction of limestone. In 44L, the presence of excess alumina from calcined clay could accelerate the limestone reaction [37,83]. Though 30F mix also had limestone content from GP cement (5.2% as seen from Table 2), the lower amount of the reactive alumina content in fly ash and GP cement than that of calcined clay led to a comparatively very slow reaction of limestone.
The pozzolanic reaction from aluminosilicates produces hydration products refining the pore structure leading to dense microstructure. Moreover, the carboaluminates from limestone reaction can also refine the pore structure [37,81,82]. The FTIR results indicated higher pozzolanic and limestone reaction of 44L mix compared to 30F mixes which can lead to finer pore structure and denser microstructure from an early age. The observed increase in autogenous shrinkage in the LC3 mix could be attributed to the finer pore structure produced by the hydration products [96].

Thermogravimetric analysis (TGA)
TGA weight losses of control, 30F and 44L mixes between 105°C and 900°C at 1, 3, 7, 14, 28, 56 and 90 days are presented in Fig. 5a1, b1, c1 and the derivative weight loss within a certain temperature range is shown in Fig. 5a2, b2, c2. The percentages of TGA weight loss were based on the total weight of the binder of each mix at 105 oC to avoid the effect of any residual free water. The weight loss from 105°C to 400°C accounted for C-S-H, alumina-ferric oxidemonosulfate (AFm) phases and other minor phases [70,97,98]. This provides an indication of the amount of chemically bound water associated with hydration products present in the samples and an indirect measurement of the degree of hydration (DOH) as presented in Fig. 6a. At 3 days, the TGA weight loss % of 30F and 44L between 105°C to 400°C were 4.9% and 5.6% respectively. It indicated high reactivity of the LC3 mix (44L) despite having a higher replacement level. At 7 days, the TGA weight loss % of 30F and 44L between 105°C to 400°C were 5.8% and 6.5% respectively and both values were higher than that of control (5.4%). The presence of calcined clay and limestone accelerates the hydration at an early age, as early as 24 h, due to the superimposition of filler effect of calcined clay and limestone and pozzolanic effect of calcined clay [25]. Fly ash also promoted the degree of reaction at an early age as reported by Haha et al. [24]. After 7 days, the hydration in 30F mix was the lowest among the three mixes up to 56 days (Fig. 6a). Between 56 and 90 days, the amount of hydration products increased due to the late pozzolanic reaction of fly ash. On the other hand, 44L continued to have a high hydration rate from 7 to 14 days and slowed down after that to almost a constant value between 28 to 90 days.
The weight loss between 400°C and 500°C (a slight shift of temperature was observed for control samples at 28, 56 and 90 days) in Fig. 5 were due to the decomposition of calcium hydroxide (CH). The derivative weight losses due to CH decomposition are also illustrated in Fig. 5 a2, b2, c2. The amount of CH per 100 g of anhydrous cement was quantified according to the method described by Kim and Olek [99] and the results are presented in Fig. 6b. For the control samples, the amount of CH gradually increased with time from 1 to 90 days as hydration progressed (Figs. 5 a2 and 6b). The pozzolanic effect was observed for both 30F and 44L samples (Fig. 5b2,  c2 and also in Fig. 4). For 30F samples, the amount of CH increased up to 14 days and started to diminish after that (Figs. 5b2 and 6b) indicating pozzolanic action. In the case of 44L, the commencement of the pozzolanic effect was observed earlier. As seen from Figs. 5c2 and 6b, the amount of CH started to decrease after 7 days. Moreover, the reduction of the amount of CH up to 90 days was considerably high for 44L compared to the fly ash mix. The calcined clay showed very high pozzolanic reactivity as reported previously [18,20,28].
The weight loss between 700 and 810°C in Fig. 5a1, b1, c1 indicated the presence of calcium carbonate [99] from limestone in the cement in the case of control and 30F mixes and the additional limestone for the 44L mix. The calcium carbonate in limestone can react with the excess alumina to produce mono and hemicarboaluminate [26,27].
The high reactivity of calcined clay promoted the early age hydration and pozzolanic action in 44L mix whereas the pozzolanic effect of fly ash manifested at a later age. Though both of the materials were aluminosilicates in a broader sense, their hydration kinetics were very different. The high hydration rate and high amount of hydration products in 44L compared to 30F influenced the early-age autogenous shrinkage (Fig. 3) by enhancing the extent of pore refinement and selfdesiccation.

Shrinkage of concrete
The investigation on the paste samples confirmed that LC3 mix developed higher autogenous shrinkage compared to the fly ash blended mix though both of the mineral additives were mainly composed of aluminosilicates. From the current study on pastes, the high reactivity of calcined clay can cause this high autogenous shrinkage. 300 500 700 900 TGA weight loss (%)  Though concrete shrinkage is impacted by aggregate size, types and their relative proportions compared to paste, it is generally expected to show a similar trend in shrinkage development to that observed in the paste. The high autogenous shrinkage in LC3 paste could lead to high total shrinkage depending on drying shrinkage. In this study, both autogenous shrinkage and total shrinkage of concrete were investigated. The total shrinkage is defined as the overall shrinkage of a specimen exposed to a dry environment.
A grade-based approach is better suited for the practical knowledge about concrete shrinkage. Though shrinkage depends on a number of variables, the overall effect of adding any particular mineral additives is generally assumed to be comparable for mixes with similar compressive strength. However, the grade-based approach rendered the current data incomparable with the previous studies on shrinkage of fly ash based on mix design [51,[53][54][55]. No data was available for comparison of shrinkage of LC3 with other binders in literature except the work by Dhandapani et al. [45] who followed different curing schemes. The compressive strength of each mix is presented in Table 3.

Autogenous shrinkage of concrete
The autogenous shrinkage of concrete including the two different strength grades considered in this study are presented in Fig. 7. Similar to the autogenous shrinkage of pastes samples (Fig. 3), the autogenous shrinkage of LC3 mixes i.e., N25-LC44 and N32-LC44 was the greatest in their respective strength category as shown in Fig. 7a and b. To facilitate further understanding, the autogenous shrinkage of concrete was expressed with respect to the autogenous shrinkage of the corresponding control specimen at 28 days (bars with a red outline in Fig. 7c). It is important to mention that autogenous shrinkage is significantly influenced by the water to binder ratio [100]. LC3 mixes can provide comparable compressive strength as control mixes formulated with similar water to binder ratio [18] However, fly ash blended mixes exhibit a lower compressive strength, especially during the early age [101,102]. Therefore, fly ash blended mixes had a lower water to binder ratio (see Table 3) to achieve the target compressive strength in this study, which influenced the autogenous shrinkage.
As seen in Fig. 7c, autogenous shrinkage of N25-FA30 was higher than the control mix (N25-0) but N32-FA30 was similar to the control mixes (N32-0) though the fly ash concretes was prepared with about 18% less water to binder ratio. On the contrary, the LC3 mixes i.e., N25-LC44 and N32-LC44 showed the highest autogenous shrinkage despite having similar water to binder ratio compared to control (N25-0 and N32-0 respectively). The LC3 concrete result of autogenous shrinkage was also consistent with the observation on the paste samples. For the 25 MPa concrete, the autogenous shrinkage of N25-0, N25-FA30 and N25-LC44 at 100 days was 1.43, 2.87 and 5.06 times higher than that of the benchmark value of 28-day autogenous shrinkage of N25-0 (Fig. 7c). For the 32 MPa concrete, the autogenous shrinkage of N32-0, N32-FA30 and N32-LC44 at 100 days was 1.47, 1.66 and 2.98 times higher than that of the benchmark value of 28-day autogenous shrinkage of N32-0 (Fig. 7c).
The autogenous shrinkage results showed that fly ash blended concrete can be used to minimize the autogenous shrinkage development even with a low water to binder ratio required to achieve the desired compressive strength. In contrast, LC3 mixes produced high autogenous shrinkage despite having the same water to binder ratio as the control mixes. Therefore, depending on the aluminosilicate used, the autogenous shrinkage can vary to a great extent for concretes having similar compressive strength.

Total and drying shrinkage of concrete
Total shrinkage is considered as the superimposition of autogenous shrinkage and drying shrinkage of concrete in this study. The total shrinkage is the ultimate manifestation of shrinkage in practice, used for design purpose. As seen from Fig. 8, the total shrinkage of the aluminosilicate blends was similar to that of the control mixes despite having a high autogenous shrinkage in the case of the LC3 mixes (Fig. 7). For the 25 MPa concretes, the total shrinkage of N25-0, N25-FA30 and N25-LC44 at 100 days was 1.38, 1.44 and 1.34 times higher than that of the benchmark value of 28-day total shrinkage of N25 0 (Fig. 8c). For the 32 MPa concrete, the total shrinkage of N32-0, N32-FA30 and N32-LC44 at 100 days was 1.44, 0.98 and 1.51 times higher than that of the benchmark value of 28-day total shrinkage of N32-0 (Fig. 8c).
Separation of drying shrinkage from the total shrinkage can provide a possible explanation of the similar total shrinkage for aluminosilicate mixes. In this study, the drying shrinkage was determined by the simplified approach of taking the difference between total and autogenous shrinkage [57,[103][104][105]. The results are presented in Fig. 9a and b. However, this simplified method has its limitations. In a total shrinkage specimen, the autogenous shrinkage may reduce as opposed to a completely sealed condition due to the lack of available water for hydration. Therefore, in this paper, this simplified estimation can only qualitatively indicate the extent of drying shrinkage development. The drying shrinkage of the aluminosilicate blends was similar or lower than that of the control sample. For the 25 MPa concrete, the drying shrinkage of N25-0, N25-FA30 and N25-LC44 at 100 days was 1.38, 1.22 and 0.77 times than that of the benchmark value of 28-day drying shrinkage of N25-0 (Fig. 9c). For the 32 MPa concrete, the drying shrinkage of N32-0, N32-FA30 and N32-LC44 at 100 days was 1.44, 0.79 and 1.08 times higher than that of the benchmark value of 28-day drying shrinkage of N32-0 (Fig. 9c). The drying shrinkage was significantly lower in the case of the LC3 mixes (N25-LC44 and N32-LC44) which compensated for the high autogenous shrinkage resulting in a comparable total shrinkage with the control mixes. The high autogenous shrinkage of the LC3 mixes was compensated by a low drying shrinkage but if the autogenous shrinkage development is substantially high, for example in the case of a very low water to binder ratio, the mitigation effect of the drying shrinkage might be inadequate. For N32-FA30, the water to binder ratio was smaller (0.4) than other concrete mixes and thus less water was available for evaporation, which can explain the low drying shrinkage. Moreover, unreacted fly ash particles act as micro aggregates [15,51]. This effect could also decrease the extent of shrinkage, especially at an early age till the late pozzolanic reactions commenced.

Ecological evaluation
The ecological impact of the concrete mixes was evaluated by two environmental sustainability indicators, i) CO 2 emission and ii) energy consumption. The results showed that using SCMs to replace the cement content significantly reduced the CO 2 emission and energy consumption (Fig. 10). The estimation of the ecological impact was calculated based on the data presented in Table 5. The embodied carbon represents the total CO 2 associated with the material production and the embodied energy which consists of the total energy required for manufacturing only assuming all aggregates, cement, calcined clay, limestone etc. are available at concrete batching plant. Due to unavailability of data, the embodied carbon and energy per unit of GP cement was assumed to be equal to Portland cement from literature. However, these values should be lower in GP cement than Portland cement as GP cement can contain up to 7.5% mineral additions. Though the binder contents ranged between 12.87% to 15.17% for the six concrete mixes (Fig. 10a), majority of the CO 2 emission and energy consumption are originated from the binders, especially cement. The CO 2 emission reduced by 29.47% and 51.46% in the case of N25-FA30 and N25-LC44 respectively compared to N25-0 (Fig. 10b). These values were 29.5% and 51.73% for 32 grade concrete. The reduction of energy consumption was about 27% and 34% for 30% fly ash concrete and 44% limestone-calcined clay concrete respectively for both grades (Fig. 10c). The reduction in energy consumption for LC3 was less than that of CO 2 emission due to the calcination process of the clay. LC3 concrete exhibited excellent performance in terms of these two indicators for environmental sustainability. Limestone and calcined clay are sustainable considering the supply versus the demand. Moreover, a high percentage of cement can be replaced without compromising the early age mechanical strength unlike fly ash. This resulted in the significantly better performance of LC3 compared to fly ash concrete making it a more sustainable and ecofriendly concreting option.

Conclusions
This study compared the shrinkage of two aluminosilicate blends i.e., 30% fly ash and 44% limestonecalcined clay (LC3) with a reference mix without any Supplementary Cementitious Materials (SCM) addition. The replacement levels used were very high for the respective SCMs. The investigation was initiated with paste samples to assess shrinkage and microstructural parameters. Finally, the shrinkage of concrete mixes with specified compressive strength i.e., 25 MPa and 32 MPa was examined.
Chemical shrinkage of paste indicated accelerated early age hydration of the LC3 blend compared to the fly ash blend. The autogenous shrinkage was significantly higher for LC3 mixes. While the autogenous shrinkage of the fly ash blend was three-fourths of the control sample at 200 days. The autogenous shrinkage of the LC3 blend was 1.5 times that of control samples. TGA showed that the amount of hydration products (decomposed between 105 to 400°C) in LC3 mixes was the highest among the three mixes which can be correlated with its highest autogenous shrinkage. The fly ash blend showed a low amount of hydration products till the late pozzolanic reactions commenced. FTIR also corroborated the late pozzolanic reactions in fly ash blend compared to the very fast pozzolanic reactivity of LC3 blends. Moreover, the FTIR analysis confirmed the significant reactions involving limestone in LC3 blends, unlike fly ash blends. The early pozzolanic reaction and limestone reaction in LC3 blends are expected to produce a very fine pore structure. The fine pore structure coupled with a high degree of hydration in LC3 mixes produced a high autogenous shrinkage compared to the fly ash blend.
LC3 concrete autogenous shrinkage was also the highest compared to other mixes, which was  Table 5 consistent with the results obtained from the paste samples. Due to the compensating effect of low drying shrinkage, the total shrinkage of the control and LC3 mixes were similar for the two compressive strength grades. The fly ash mixes having lower water to binder ratio to achieve the target compressive strength, showed similar or lower total shrinkage than other mixes. Considering the replacement level of cement, shrinkage, and ecological impact, LC3 proved to be a more sustainable and eco-friendly concrete compared to fly ash. This study revealed the effect of different hydration processes in aluminosilicate blends on shrinkage of low strength concrete. Further investigation is required of blended concrete having a wider range of compressive strengths. The variability of the reactivity of the SCMs can also be correlated with the shrinkage development as it is heavily relied on the source of the SCMs.
Acknowledgements This research was funded by CCAA and ARC Linkage project LP170100912 titled ''Shrinkage, cracking, self-healing and corrosion in blended cement concrete''. The authors would like to thank the Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering and Mark Wainwright Analytical Centre of University of New South Wales (UNSW), UTS tech lab of University of Technology Sydney for providing technical support.
Funding Open Access funding enabled and organized by CAUL and its Member Institutions.

Declarations
Conflict of interest We wish to confirm that there is no conflict of interest associated with this publication and there has been no significant financial support for this work that could have influenced its outcome.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.