Effect of using limestone fines on the chemical shrinkage of pastes and mortars

The main aim of this study is to examine the effect of incorporating limestone fine (LF) on chemical shrinkage of pastes and mortars. For this purpose, five paste and five mortar mixes were prepared with 0, 5, 10, 15, and 20% (by weight) LF as a replacement of cement. The water-to-binder (w/b) ratio was 0.45 for all mixes. The sand-to-binder (s/b) ratio in the mortar mixes was 2. Testing included chemical shrinkage, compressive strength, density, and ultrasonic pulse velocity (UPV). Chemical shrinkage was tested each hour for the first 24 h, and thereafter each 2 days until a total period of 90 days. Furthermore, compressive strength and UPV tests were conducted at 1 day, 7, 28, and 90 days of curing. The results show that the long-term chemical shrinkage of pastes was found to increase with the increase in LF content up to 15%. Beyond this level of replacement, the chemical shrinkage started to decrease. However, the chemical shrinkage for mortars increased with the increase in LF content up to 10% LF and a decrease was observed beyond this level. It was also noticed that compressive strength for pastes and mortars attained the highest value for mixes containing 10 and 15% LF. The trend in the UPV results is somewhat similar to those of strength. Density for pastes and mortars increased up to 15% LF followed by a decrease at 20% replacement level. Correlations between the various properties were conducted. It was found that an increase in chemical shrinkage led to an increase in compressive strength.


Introduction
Concrete is one of the dominant building materials in the construction industry. The main problem of concrete is the early-age cracking which is caused by chemical shrinkage (Safiuddin et al. 2018). Chemical shrinkage is defined as the internal volume change of cement paste resulting from the hydration of cementitious materials (Khatib et al. 2021). The volume of hydration products is smaller than the sum of both cement and water volumes (Zhutovsky and Kovler 2010). This phenomenon is also known as Le Chatelier contraction. Le Chatelier reported that total chemical shrinkage for normal Portland cement was about 4.6% by weight of cement (Le Chatelier 1900). Chemical shrinkage is considered an essential property of Portland cement and also assumed proportional to the degree of hydration (Zhutovsky and Kovler 2008). Geiker and Knudsen (1982) mentioned that chemical shrinkage is independent of waterto-cement ratio during the early age of hydration. However, other authors (Zhang et al. 2010;Zhang and Scherer 2020) reported that chemical shrinkage is dependent on the w/c ratio. Chemical shrinkage is also a function of the initial cement content (Zhutovsky and Kovler 2010).
Three main measurement methods for chemical shrinkage have been used: dilatometry, pyconometry, and gravimetry. Amid these test methods, dilatometry is the most popular method used by many authors (Justnes et al. 1994;Tazawa et al. 1995;Justnes et al. 1996;Persson 1997;Justnes et al. 1998;Justnes et al. 1999;Hammer and Heese 1999;Justnes et al. 2000;Beltzung and Wittmann 2000;Charron et al. 2001;Charron et al. 2002;Zhang et al. 2013;Liao and Wei 2014). The dilatometry method measures the drop in level of water in a hydrating cement paste system using a graduated pipette or a suitable graduated tube (Lura, 2003). This drop of water in the pipette represents the chemical shrinkage (volume change) of cement paste.
The addition of supplementary materials as partial replacement of cement or aggregate Ghanem et al, 2019Ghanem et al, , 2020Machaka et al, 2019aMachaka et al, , 2019b will mainly affect the hydration process which in turn will affect the chemical shrinkage. Several studies investigated the influence of incorporating supplementary materials on chemical shrinkage. Wild et al. (1998) found that incorporating up to 15% metakaolin (MK) enhanced the chemical shrinkage. This finding is also reported by Markandeya et al. (2019). Moreover, Lu et al. (2020) noted that using blast furnace slag (BFS) in concrete achieved higher chemical shrinkage than those containing fly ash (FA) and silica fume (SF). Furthermore, Wang et al. (2020) indicated that chemical shrinkage increased as the percentage of nanosilica increased.
Over the last decades, the incorporation of limestone as cement replacement has been investigated (Bonavetti et al. 2000;Tsivilis et al. 2000;Heikal et al. 2000;Matschei et al. 2007;Lothenbach et al. 2008). Limestone is a sedimentary rock consisting of calcium carbonate (CaCO 3 ). It is used for various purposes especially in concrete production. Limestone is usually considered an inert filler material that produces an acceleration in the chemical shrinkage and hydration process during the early ages (Bouasker et al. 2008;Weerdt et al. 2011). This filler played the role of nucleation sites for hydrates (Bouasker et al. 2008;Weerdt et al. 2011). Several studies reported the effect of replacing limestone fine (LF) on the chemical shrinkage of concrete. Using LF in cement-based mixes affected the chemical shrinkage (Menendez et al. 2003;Wang et al. 2010;Weerdt et al. 2011). Another study confirmed that incorporating LF with calcined clay (CC) increased rapidly the chemical shrinkage within the first 7 days (Du and Dai Pang 2020).
There has been some research on the chemical shrinkage of cement paste incorporating LF. However, there is hardly any research on the chemical shrinkage of mortar. Also, there do not seem studies on comparing the effect of incorporating limestone fines on the chemical shrinkage of cement paste and mortar. Moreover, the cement industry produces a large amount of CO 2 emission (7%), which result in greenhouse effect that can contribute towards the rise in the earth temperature (Thongsanitgarn et al. 2012;Meddah et al. 2014). Therefore, incorporating LF may reduce such problem. Therefore, the main objective of this study is to examine on the effect of using LF on the chemical shrinkage of pastes and mortars. Other properties such as density, compressive strength, and ultrasonic pulse velocity (UPV) were also conducted.

Materials
One commercial Portland cement type I was used in this study (Portland cement PA-L 42.5 N). The LF was extracted from Lebanon quarries. Figure 1 shows the particle size distribution of limestone and cement. The limestone was firstly ground before sieving through the 300 µm sieve. The density of LF is found to be 2.7 g/cm 3 . The chemical compositions for cement and limestone are shown in Table 1. The natural sand used had a fineness modulus of 2.8.

Mix proportions
Five paste mixes and five mortar mixes were employed. The cement in the paste and mortar mixes are partially replaced with 0, 5, 10, 15, and 20% LF (by weight). The water-tobinder (w/b) ratio for all mixes was kept constant at 0.45. In the mortar mixes, the sand-to-binder (s/b) ratio was 2. The binder consists of cement and LF. The total number of samples tested was 100. The details of paste and mortar mix proportions are presented in Tables 2 and 3 respectively.

Sample preparation and test procedure
Chemical shrinkage was tested according to ASTM C1608 (2007). The cement and LF were dry mixed for 2 min. In the case of mortar, sand was added to the dry materials and also mixed for 2 min. After adding the prescribed water to the blended materials, the mixing continued until homogeneity was attained. The mixing was conducted by a spatula due to the small volume of specimens. After mixing, the paste or mortar was placed in a 250-ml glass bottles by using a funnel to a depth of around 1.8 cm equivalent to approximately 30 g of paste. The bottle was weighed before and after the addition of paste or mortar sample. The bottles were slightly shaken in order to maintain the same level and to eliminate air voids. After that, distilled water was added slowly on the top surface of the sample using a straw so that the surface layer of the sample is not disturbed. For each bottle, a 2-ml graduated pipette was inserted through a hole in a rubber stopper. The stoppers were immersed in water for a few hours to avoid any water absorbed by the rubber. Then, water was slowly added using a syringe until the top of the pipette. This level of water was recorded as the initial value. In order to prevent water evaporation, a drop of oil was placed on the top level of water in each pipette. The bottles were then placed in a room at constant temperature (25 °C) as shown in Fig. 2. The level of water in the pipette was observed each hour for the first 24 h, and thereafter each 2 days until a period of 90 days. The drop of water in the pipette represents the volume change (i.e., chemical shrinkage) which was converted in milliliters per gram of binder (PC + LF). Each chemical shrinkage value determined is the average of two values from two replicate specimens. The chemical shrinkage measurement setup is shown in Fig. 3. For compressive strength and UPV test, standards steel cubic molds of dimensions 50 mm × 50 mm × 50 mm were used. Compressive strength was measured according to ASTM C109 (2016) using a universal testing machine with a rate of loading of 3 kN/s. UPV test was applied according to ASTM C597 (2016) and was calculated as follows: (1) V = D∕T  where D is the distance separating the transmitter and the receiver (km), and T is the propagation time for the pulse to get throughout the specimen (s). Moreover, the density was calculated according to ASTM C188 (2014) and was calculated as follows: where D is the density (g/cm 3 ); M and V are the mass and volume of the cube respectively.

Results and discussion
The results of chemical shrinkage, compressive strength, density and UPV tests, data analysis, and the correlations are reported in the following sections.

Chemical shrinkage
The results of chemical shrinkage of paste and mortar specimens during the first 24 h are presented in Figs. 4 and 5 respectively. The zero time was considered when water was added to the cement. At 24 h, the chemical shrinkage of paste with 0% LF reached a value of 0.026 ml/g. This value increases to a maximum value of 0.047 ml/g for pastes with 15% LF. As deduced from Fig. 5, the chemical shrinkage of mortars at 0% LF is 0.029 ml/g. This value goes up to 0.037 ml/g at 5% LF then subsequently drops with the incorporation of 10, 15, and 20% LF. This can be interpreted by the fact that the existence of LF causes an acceleration in the chemical shrinkage for the first few hours. This finding is consistent with that of Bouasker et al. (2008). The results of chemical shrinkage of pastes and mortars for 90 days are shown in Figs. 6 and 7 respectively. Those figures show similar characteristics. For the first 3 days, the chemical shrinkage values are approximately similar for all mixes in paste and mortar specimens. At 90 days, the chemical shrinkage of pastes with 0% LF attains a value of 0.107 ml/g. This value increases to 0.133 ml/g at 5% LF then slightly drops with the addition of 10% LF (0.13 ml/g). After this slight decrease, the chemical shrinkage increases to the optimum value of 0.194 ml/g at 15% LF. For a replacement of 20%, the chemical shrinkage exhibits a sharp decline and reaches a value of 0.12 ml/g. During the 1st day, the inclusion of LF with the same percentages in mortars has mostly the same effect as pastes but with a smaller value. As shown in Fig. 7, the chemical shrinkage achieves a value of 0.121 ml/g at 0% LF. This value slightly decreases to about 0.117 ml/g for an amount of 5% LF then goes up to the maximum value of 0.132 ml/g at 10% LF. After that, the chemical shrinkage displays a sharp drop at 15 and 20% LF. The sharp decline occurring after the addition of 10 and 15% LF in mortar and paste samples respectively may be explained by the fact that with high amount of LF (above 10%), the formation of calcium-carboaluminate hydrate could consume CH with the increase in curing time. Therefore, the chemical effect of LF leads to the stabilization of the ettringite and could increase the solid volume of hydration products (external expansion) which decrease the total system volume (chemical shrinkage) (Wild et al. 1998;Menendez et al. 2003;Wang et al. 2010;Weerdt et al. 2011;Wang et al. 2018a, b). Besides, with the same w/b ratio, since LF has minimal cementitious or pozzolanic properties, the substitution of cement with different percentages of LF increases the free water to react with cement particles. That is known as the dilution effect (Wang et al. 2018a).
The inclusion of different percentages of LF enhances the chemical shrinkage of pastes and mortars. In order to clarify this finding, three explanations can be made. First of all, as chemical shrinkage is a direct result of the hydration reactions, the acceleration of the cement hydration in the presence of LF which acts as a filler refines the pore structure, reduces the porosity of cement (Wang et al. 2018b), and makes an increase in chemical shrinkage. Additionally, LF can also interact slowly with the monosulfoaluminate (Afm) to create monocarboaluminate (Afmc), which has a higher density than the Afm phase (Weerdt et al. 2011). Last point, LF with fine particle size provides nucleation sites for hydration products to precipitate, accelerates the hydration process, and enhances the hydration degree of cement (Wang et al. 2018a, b). It was mentioned that the advanced precipitation of C-S-H on the surface of LF was due to the propinquity between configuration of Ca and O atoms in calcite and CaO layers in C-S-H. With the increase of LF amount, additional nucleation sites would be available, and more hydration products could be absorbed (Wang et al. 2018a).
By comparing Figs. 6 and 7, it can be deducted that chemical shrinkage results for paste specimens are greater than those for mortars. For example, the addition of 5% LF in pastes exhibits a value of 0.133 ml/g while shows a value of 0.117 ml/g in mortars. This is mainly due to the presence of sand which automatically reduces the amount of cement and LF and slows down the hydration process.
To further illustrate the influence of LF on the chemical shrinkage of paste and mortar specimens, Figs. 8 and 9 are displayed. The plots exhibit the change in volume against different percentages of LF at various curing ages (1, 7, 28, 45, and 90 days) for paste and mortar samples respectively. For pastes (Fig. 8), the extreme drop in volume occurs at 15% LF and then subsequently decreases. On the other side, the extreme drop in volume for mortars occurs at 10% LF then successively declines (Fig. 9). Proportionally, as curing time increases, the chemical shrinkage of pastes and mortars goes up as well. For example, the volume changes for 5% LF in mortar and paste samples increases about 8 and 9% from 1 to 90 days respectively. The chemical shrinkage becomes much steeper as curing time increases in pastes and mortars (e.g., at 7 days). The comparative proportions of phases present at specific curing time depend on the LF content. For a specific LF content, as the curing time increases, CH is consumed (Wild et al. 1998;Wang et al. 2018a). Accordingly, the CH content of system drops, and as a result, the CH/LF ratio will decline. Similarly, this last finding is also present when the LF content increases (Wild et al. 1998;Wang et al. 2018a).

Compressive strength
The results of the compressive strength for pastes and mortars are presented in Figs. 10 and 11 respectively. Compressive strength measurements were carried out at 1, 7, 28, and 90 days. As shown in these plots, the compressive strength of paste and mortar specimens for 15% and 10% LF respectively displays the highest strength among all percentages. For example, at 7 days, the compressive strength of pastes is 31.5 MPa for the control mix. This value increases slightly with the addition of LF until achieving a maximum value of 35.8 MPa for 15% LF followed by a decrease in strength at 20%. This strength loss of mixes with higher amount of LF (˃15% LF) may be due to the significant decrease of the potential cementitious material content, which is known as   (Wang et al. 2018a). The pastes and mortars become more compacted, hence increasing the compressive strength. Moreover, the presence of chemical reaction between LF and tricalcium aluminate (C 3 A) to form calcium-carboaluminate leads to decrease the porosity of pastes and mortars and consequently increase their compressive strength (Thongsanitgarn et al. 2012).
The results of compressive strength for pastes and mortars are mainly consistent with those of density. The density increases up to 15% LF then drops for a higher amount of LF as shown in Table 4. For example, at 90 days, the density of pastes is 2.13 g/cm 3 at 0% LF. This value increases to the maximum value of 2.14 g/cm 3 with the addition of 15% LF followed by a decline at 20%. This increase is almost negligible (only 0.5%). For mortars, the density at 0% LF is 2.43 g/cm 3 . This value goes up to the optimum value of 2.47 g/cm 3 at 15% LF succeeded by a reduction at 20%. This is due to cement hydration resulting from the reaction between LF and cement. The reduction occurring in the external dimensions of cement particles generates voids, a perfect host for water. This phenomenon reduces durability as well as the density of paste. LF acts as a filler that fills the voids around cement particles up to the optimum. For higher filler content, voids are already totally filled. The extra amount of LF occupies the place of sand particles, thus decreasing sand proportion and subsequently the density of mortars.
Besides, it is noted that the compressive strength values for pastes are greater than those for mortars. For example, the incorporation of 10% LF in pastes shows a value of 56.4 MPa at 28 days. However, it displays a value of 27.9 MPa in mortars. This is well expected because the amount of cement in paste samples is greater than that for mortars. Accordingly, the amount of cement and LF exists in small amount in mortar mixtures. Thus, the reaction between cement and LF as well as the hydration process will be reduced and negatively affects the compressive strength.

Ultrasonic pulse velocity
The UPV test is used to check the concrete quality. Figures 12 and 13 show the UPV for pastes and mortars versus curing ages for different percentages of LF. For paste samples, mix 4 with 15% LF exhibits a higher UPV value with a good quality of paste (3.88 km/s) among other mixes. Adding more than 15% LF reduces the quality of pastes. This means that whenever the percentage of LF increases beyond this limit, the LF may act as a filler and does not contribute to an increase in UPV.
For mortar samples (Fig. 13), mix 3 with 10% LF displays higher UPV values than the other LF mixes. For example, at 90 days, mixes 2, 4, and 5 with 5, 15, and 20%  Fig. 12 Variation of UPV for pastes at different curing ages LF replacement levels display a value of 3.81, 3.7, and 3.73 respectively. However, mix 3 with 10% LF achieves a value of 3.91 km/s. It is well noticed that UPV results are consistent with those of compressive strength and chemical shrinkage. Adding 10 and 15% LF in mortar and paste samples respectively achieves maximum UPV, compressive strength, and chemical shrinkage values. Besides, it is shown that UPV values for mortars are mainly greater than those of pastes. For example, the UPV value for a replacement of 10% LF in pastes shows a value of 3.78 km/s and a value of 3.91 km/s for mortar specimens. One possible explanation is that whenever the gel pores between cement and LF increase due to hydration, these voids are replaced by sand leading to a decrease in the time needed for the pulse to get throughout the mortar specimen. Figures 14 and 15 show the correlation between compressive strength and UPV of pastes and mortars respectively for different percentages of LF at 1, 7, 28, and 90 days. This correlation appears to be a normal exponential relationship for both pastes and mortars with a high coefficient of determination as reported in the literature (Khatib 2005(Khatib , 2008(Khatib , 2009. For example, for pastes, the coefficient is 0.91 and for mortars is 0.93.

Correlation between different properties
The correlation between compressive strength of pastes and compressive strength of mortars is presented in Fig. 16. This correlation looks linear with a high coefficient of determination (R 2 ). For example, the incorporation of 0% LF in paste and mortar samples shows a high correlation with a coefficient value of 0.92. The maximum coefficient value (R 2 = 0.98) is observed for 5% LF addition. It is also noted that the results of this correlation are consistent with previous studies (Al-Swaidani 2017;Al-Swaidani et al. 2017).
Additional correlations are shown between the compressive strength and chemical shrinkage at each curing age ( Fig. 17a-d). As chemical shrinkage increases, the compressive strength goes up. A positive correlation with a coefficient of determination 0.8 ˂ R 2 < 0.99 and 0.75 ˂ R 2 < 0.96 is shown for paste and mortar samples respectively. For pastes, the highest coefficient of determination is attained at 28 days. However, the highest R 2 is realized at 90 days for mortars. It is also noted that the linear correlation for paste  specimens is stronger than that of mortar specimens. One possible explanation is that the presence of sand in mortar samples (by delaying the hydration process) decreases the chemical shrinkage and accelerates the self-desiccation of samples. Therefore, compressive strength will be reduced (Soroka and Stern 1976;Voglis et al. 2005).
To illustrate the effect of LF on the chemical shrinkage of mortar and pastes, the slope of the regression line between the compressive strength and chemical shrinkage for each paste and mortar mixes versus % LF is plotted in Fig. 18. It can be observed that the peaks are attained at 10 and 15% LF for mortars and pastes respectively. Those results are consistent with the chemical shrinkage results obtained previously. Therefore, the optimal percentage of LF as a replacement of cement for pastes and mortars is between 10 and 15% respectively.

Conclusions
This study is about the influence of incorporating LF on chemical shrinkage of pastes and mortars. Based on the results of this experimental study, the following conclusions can be made: • Incorporating 5 and 15% LF enhances the chemical shrinkage for the first 24 h for mortars and pastes respectively. Chemical shrinkage for paste samples increases for replacements between 0 and 15% LF where it achieves the highest value of 0.194 ml/g of binder followed by a sharp reduction. The chemical shrinkage for mortar increases for replacements between 0 and 10% LF where it records the highest value of 0.132 ml/g of binder followed by decline beyond 10% LF. The chemical shrinkage for pastes samples shows higher values comparing to those for mortar samples. Therefore, the existence of sand affects the hydration process of cement. • The compressive strength appears to be related to the LF content. The incorporation of 10% and 15% LF in mortar and paste samples respectively enhances the compressive strength where it achieves the maximum value of 70.4 MPa in paste specimens and 31.6 MPa in mortar specimens. The values of compressive strength in paste specimens are greater than those for mortar specimens, and this relationship is linear with a high coefficient of correlation (R 2 above 0.91). • The variation of UPV of pastes and mortars depends on the LF content. The velocity values for paste increase between 0 and 15% LF and attain the highest value of 3.88 km/s at 90 days. However, for mortar samples, the UPV values exhibit an increase between 0 and 10% LF and display the highest value of 3.91 km/s at 90 days. This could be due to the interfacial transition zone in the mortar. • The correlation between compressive strength and chemical shrinkage at different curing ages shows a positive correlation with R 2 values above 0.75 for pastes and mortars. Similarly, correlation between compressive strength and UPV shows a positive correlation (0.82 ˂ R 2 ˂ 1). The higher the strength, the higher the chemical shrinkage and UPV.