Intensiﬁed Pozzolanic Reaction on Kaolinite Clay-Based Mortar

: The objective of this study is to develop and characterize kaolinite clay-based structural mortar. The pozzolanic reaction induced from two mineral additives, i.e., calcium hydroxide and silica fume (SF), and the physical ﬁlling effect from SF, were found to be effective on the enhancement of structural properties. Based on several preliminary experiments, 7:3 ratio of kaolinite clay/calcium hydroxide was selected as a basic binder. Then, the amount of SF was chosen as 0%, 7.5%, and 15% of the total binder to consider both the chemical and physical effects. The results showed that compressive strengths of samples with 7.5% and 15% SF are signiﬁcantly increased by approximately 200% and 350%, respectively, at 28 days compared to the sample without SF. However, based on the results of the sample with 15% SF, it is found that excessive addition of SF causes long-term strength loss, possibly owing to micro cracks. With the careful consideration on this long-term behavior, this suggested new mix design can be further extended to develop sustainable structural materials using natural minerals or waste materials with nonbinding properties.


Introduction
Concrete is one of the most widely used construction materials in the world with a consumption of almost 10 billion tons per year [1]. It has numerous advantages as a structural material, such as convenience for making the shape of a structure, structural soundness, and good durability. Recently, however, the ordinary Portland cement (OPC) has been regarded as environmentally harmful because its manufacturing process generates a significant amount of carbon dioxide (CO 2 ) [2]. CO 2 is considered as one of greenhouse gases that is associated with global warming and climate change [3]. In addition, OPC production gives rise to other environmental concerns, such as large amounts of dust during handling and high energy consumption to attain high kiln temperatures [4].
As a result of extensive research activities focusing on the development of eco-friendly construction materials, several promising alternatives have been proposed for low CO 2 -embedded construction materials. These include high-volume fly ash [5,6] or slag [7] concrete, geopolymer concrete [8], and sustainable cement mixtures having CO 2 -sequestrated minerals [9], or fillers such as limestone or quartz powder [10][11][12][13]. In addition, materials having pozzolanic properties, such as fly ash (type F), slag, and SF are receiving attention as sustainable supplementary cementitious materials [14][15][16][17][18]. Along with the environmental and economic benefits as replacements for OPC, these materials also have other advantages, such as improving mechanical properties and durability [19].
As a sustainable building material, kaolinite clay (also known as hwangtoh clay in Korea) has been used for more than 1500 years in traditional houses in Korea, where there are enough reserves [23]. It consists primarily of kaolinite and quartz. Owing to its abundance and sustainable benefit, it has been studied by Korean researchers as a potential supplementary or main building material [23,24]. The well-known advantages as a building material include high heat-storing capacity and its purifying, deodorizing, and antibiotic properties [25]. Moreover, when heating through Ondol (traditional floor heating system in Korea), the infrared radiation emission from the clay was confirmed to be beneficial to the human body [25]. With these advantages, it has been used as plastering, flooring, and finishing materials for buildings up to the present time [23,26]. From a mix design perspective, the clay has been customarily mixed with sand and lime [27], as stated in Korean building standards [28]. However, the clay (the kaolinite mineral is the main reactive component in the clay) simply mixed with lime does not provide sufficient strength for construction applications. The main reason for the low compressive strength (<3 MPa) of kaolinite clay-based mortar (water-to-binder ratio of 0.5 to 0.6) is reported as the poor pozzolanic reaction between the kaolinite clay and lime [29]. This limits the application of this sustainable building material to mostly non-structural elements.
To improve the mechanical strength of kaolinite clay, several studies were performed. The most well-known method to increase its reactivity is the heating of the clay up to 600 • C and above [4,26]. OPC mixed with the heated kaolinite clay is also suggested [25,30]. However, the heated clay with or without the OPC has a shortcoming of high energy consumption. The application of the geopolymer reaction using alkali activators is another interesting approach [23]. However, this also has disadvantages, such as significant workability loss because of rapid reaction, the hazard of handling an alkali solution, and high cost [31,32]. In order to overcome these shortcomings, it is necessary to develop a new mixing formula for kaolinite clay-based mortar with improved material properties.
In this study, a new mix proportion of the kaolinite clay-based mortar was developed. To increase its reactivity, an intensified pozzolanic reaction was intended by adding SF and hydrated lime together on the kaolinite clay. SF, known as microsilica, is a highly effective pozzolanic material due to its high silica content and extreme fineness [33,34]. The addition of SF is expected to play a key role on the enhancement of material properties by providing the physical filling effect and the pozzolanic reaction with added hydrated lime. In this study, a series of experiments of compressive strength, heat of hydration, dimensional stability, and microstructural analysis was performed to evaluate the effects of the mineral admixtures on various material properties of kaolinite-based clay mortar.

Materials
The scanning electron microscope (SEM, JSM-7800F Prime from JEOL Ltd., Tokyo, Japan) images of all used materials are shown in Figure 1. The chemical compositions determined by X-ray fluorescence (XRF-1700, Shimadzu, Tokyo, Japan) are summarized in Table 1. Kaolinite clay was obtained from Gochang-goon, Chonnam, Korea. As shown in Table 1, the main oxide components of the clay are SiO 2 (55.34%), Al 2 O 3 (22.33%), and Fe 2 O 3 (8.43%), while the hydrated lime is mainly composed of CaO (74.51%). The particle size distributions of kaolinite clay, hydrated lime, and SF are presented in Figure 2, which were determined by laser diffraction method using Mastersizer 3000 (Malvern Instruments, Malvern, UK). Used SF (SiO 2 > 96.9%) has a density of 2.2 g/cm 3 and specific surface area of 200,000 cm 2 /g. In this study, the particle size distribution of SF was measured again using image-processing technique to correct its agglomeration effect during the laser diffraction measurement [35]. To obtain more accurate particle size distribution of SF, 20 SEM images containing total 3720 SF particles were analyzed under various magnifications (30,000×-100,000×). The obtained new distribution curve of SF is added in Figure 2. The laser diffraction method could not detect the particles smaller than 100 nm (except for very fine particles of 5-11 nm) due to its agglomeration effect. Thus, the highest volume density is formed at 265 nm ( Figure 2a); while image-processing method detected the particles between 40-100 nm and yielded the highest density peak at 172 nm. This comparison reveals that a significant portion of small SF particles cannot be detected and the particle size can be overestimated from the laser diffraction method.
The fine aggregate satisfies the international organization for standardization (ISO) standards [36] which regulates that SiO 2 contents are no less than 98% and the particle size is ranged between 80 µm and 2 mm.
Appl. Sci. 2017, 7, 522 3 of 12 The particle size distributions of kaolinite clay, hydrated lime, and SF are presented in Figure 2, which were determined by laser diffraction method using Mastersizer 3000 (Malvern Instruments, Malvern, UK). Used SF (SiO2 > 96.9%) has a density of 2.2 g/cm 3 and specific surface area of 200,000 cm 2 /g. In this study, the particle size distribution of SF was measured again using image-processing technique to correct its agglomeration effect during the laser diffraction measurement [35]. To obtain more accurate particle size distribution of SF, 20 SEM images containing total 3720 SF particles were analyzed under various magnifications (30,000×-100,000×). The obtained new distribution curve of SF is added in Figure 2. The laser diffraction method could not detect the particles smaller than 100 nm (except for very fine particles of 5-11 nm) due to its agglomeration effect. Thus, the highest volume density is formed at 265 nm ( Figure 2a); while image-processing method detected the particles between 40-100 nm and yielded the highest density peak at 172 nm. This comparison reveals that a significant portion of small SF particles cannot be detected and the particle size can be overestimated from the laser diffraction method.
The fine aggregate satisfies the international organization for standardization (ISO) standards [36] which regulates that SiO2 contents are no less than 98% and the particle size is ranged between 80 μm and 2 mm.

Mix Proportion
Mix proportions of the kaolinite clay-based mortar are shown in Table 2. Both kaolinite clay and hydrated lime are considered as binder materials. The main variable is SF to binder ratio, which was set as 0%, 7.5%, and 15% by weight of the binder. The reference sample (Ref) does not include SF, while SF_7.5 and SF_15 samples contain 7.5% and 15% of SF, respectively. The water to binder ratio (W/B) for all samples was maintained at 40%. Polycarboxylate-ether type superplasticizer was added The particle size distributions of kaolinite clay, hydrated lime, and SF are presented in Figure 2, which were determined by laser diffraction method using Mastersizer 3000 (Malvern Instruments, Malvern, UK). Used SF (SiO2 > 96.9%) has a density of 2.2 g/cm 3 and specific surface area of 200,000 cm 2 /g. In this study, the particle size distribution of SF was measured again using image-processing technique to correct its agglomeration effect during the laser diffraction measurement [35]. To obtain more accurate particle size distribution of SF, 20 SEM images containing total 3720 SF particles were analyzed under various magnifications (30,000×-100,000×). The obtained new distribution curve of SF is added in Figure 2. The laser diffraction method could not detect the particles smaller than 100 nm (except for very fine particles of 5-11 nm) due to its agglomeration effect. Thus, the highest volume density is formed at 265 nm ( Figure 2a); while image-processing method detected the particles between 40-100 nm and yielded the highest density peak at 172 nm. This comparison reveals that a significant portion of small SF particles cannot be detected and the particle size can be overestimated from the laser diffraction method.
The fine aggregate satisfies the international organization for standardization (ISO) standards [36] which regulates that SiO2 contents are no less than 98% and the particle size is ranged between 80 μm and 2 mm.

Mix Proportion
Mix proportions of the kaolinite clay-based mortar are shown in Table 2. Both kaolinite clay and hydrated lime are considered as binder materials. The main variable is SF to binder ratio, which was set as 0%, 7.5%, and 15% by weight of the binder. The reference sample (Ref) does not include SF, while SF_7.5 and SF_15 samples contain 7.5% and 15% of SF, respectively. The water to binder ratio (W/B) for all samples was maintained at 40%. Polycarboxylate-ether type superplasticizer was added

Mix Proportion
Mix proportions of the kaolinite clay-based mortar are shown in Table 2. Both kaolinite clay and hydrated lime are considered as binder materials. The main variable is SF to binder ratio, which was set as 0%, 7.5%, and 15% by weight of the binder. The reference sample (Ref) does not include SF, while SF_7.5 and SF_15 samples contain 7.5% and 15% of SF, respectively. The water to binder ratio (W/B) for all samples was maintained at 40%. Polycarboxylate-ether type superplasticizer was added to the mixing water to obtain workability. The ratio of fine aggregate to binder was determined as 3 by several preliminary tests.

Specimen Preparation
The mortar was mixed according to ASTM C305 [37] using a five-liter Hobart mixer. Dry kaolinite clay, hydrated lime, and fine aggregate were blended for 30 s. When preparing SF_7.5 and SF_15 samples, SF was mixed with the dry materials together. Then, water and superplasticizer were added to the dry mixture and mixed for 3 min at low speed (140 ± 5 rpm) and another 1 min at high speed (285 ± 10 rpm). After mixing, the mortar was placed and compacted in the prepared molds. All specimens were wrapped with vinyl sheets to prevent moisture loss and were cured in a steam chamber (60 ± 2 • C, RH 95 ± 5%) for 72 h. This curing condition was chosen considering the typical precast concrete production. After demolding at 72 h, all samples were cured in a constant temperature and humidity chamber (20 ± 2 • C, RH 60 ± 5%) until the test.

Test Method
The heat of hydration was measured for 72 h at 60 • C using an isothermal calorimeter (TAM AIR, TA Instruments, New Castle, DE, USA). For each measurement, 15 g of fresh paste without fine aggregate was prepared by assuming that the aggregate is inert. The total heat of hydration was determined by integrating the heat flow curve. To compare the hydration reaction quantitatively, the measured hydration heat was normalized by weight of the binder in each paste.
X-ray diffraction (XRD) analysis was performed to investigate the mineralogical characteristics of all raw materials and hydrated samples (without fine aggregate). After 28 days of curing, the crushed and grounded pastes were placed in a holder to perform the analysis. Each sample was scanned from 5 • to 80 • (2θ) with a step size of 0.0033 • using Miniflex (Rigaku, Tokyo, Japan). The crystalline phases were identified by comparing Bragg peak positions and intensities from the inorganic crystal structure database (ICSD) [38].
To investigate the porosity and pore size distribution, mercury intrusion porosimetry was used. Three hydration stages of 14, 28, and 91 days were selected, and mercury parameters were set to values of 485 erg/cm 2 for the surface tension and 130 • for the contact angle.

Heat of Hydration
The heat flow evolution of kaolinite clay-based mortar is shown in Figure 3a. All samples have two main peaks regardless of SF addition. Abrupt increases of heat flow within 2 h were measured for all samples almost identically. This infers that experiments were reliably conducted, and the addition of SF does not alter the start of first hydration reaction. The second main peak initiated at 12 h and reached its maximum at 20 h in the Ref sample, whereas it starts quickly from 5 h in samples with SF.
This observation supports the acceleration of the start of the main hydration reaction and increases in the duration of its period due to the addition of SF. As a result, the main hydration reaction was significantly intensified. This eventually results in a larger amount of cumulative heat of hydration as shown in Figure 3b. As will be explained later, this intensified hydration reaction is the intended pozzolanic reaction between amorphous silica and hydrated lime. of hydration as shown in Figure 3b. As will be explained later, this intensified hydration reaction is the intended pozzolanic reaction between amorphous silica and hydrated lime. Although the samples with SF emitted more cumulative heats compared to Ref (Figure 3b), there is no noticeable difference in the heats between two samples with SF (the heats are 42.53 J/g and 44.67 J/g in SF_7.5 and SF_15, respectively). Thus, it can be concluded that an excessive addition of SF (e.g., more than 7.5 wt % of SF) scarcely contributes to the enhancement of hydration reaction during heat treatment period. However, it may improve the strength because of the physical filling effect by unreacted SF particles, which will be discussed later.

XRD Analysis
The results of XRD analysis of raw materials and hydrated kaolinite clay-based pastes at 28 days are shown in Figure 4. The kaolinite clay is composed of quartz, kaolinite, and illite. Hydrated lime consists of mostly calcium hydroxide and calcium carbonate. SF contains only amorphous silica. In all hydrated samples, an alumino-silicate type of mineral (clinoptilolite) is found. However, while calcium hydroxide is strongly detected in Ref (see red-dotted boxes in Figure 4), it is almost not detected in samples with SF (SF_7.5 and SF_15). It is obvious that calcium hydroxide was consumed by the pozzolanic reaction induced from the addition of SF [40,41].
Quartz from kaolinite clay is detected in all hydrated samples. The intensity of quartz peak (at 26.5° theta) increases with addition of SF (see blue-dotted box in Figure 4). This can be explained by calcium hydroxide in hydrated lime preferentially reacting with amorphous silica first rather than with quartz. Thus, the intensity of unreacted quartz peak is the highest in SF_15. This observation also shows that crystalline quartz in kaolinite clay can react with calcium hydroxide to some extent as a pozzolanic agent. The kaolinite peak (at 12.2° theta) is slightly weaker in samples with SF. Along with the fact that kaolinite has a weak pozzolanic reactivity [42], the weakness of peak intensity can be also due to the smaller absolute amount of the clay in the paste by the SF addition (Table 2).
Meanwhile, the calcium carbonate peak (at 29.4° theta) of the hydrated sample increases with SF addition (see purple-dotted box in Figure 4). As can be confirmed by the XRD result, calcium carbonate was initially included in raw hydrated lime. However, the same amount of hydrated lime was added in all samples during the mixing process. In this sense, it is interesting to note that the intensity of calcium carbonate is higher in SF_7.5 or SF_15 than that in Ref. In our system, SF addition increased the carbonation reaction in kaolinite clay-based mortar that contains calcium hydroxide. Similarly, it has been previously reported that the addition of SF increases the carbonation depth of cementitious materials [43,44]. Although the samples with SF emitted more cumulative heats compared to Ref (Figure 3b), there is no noticeable difference in the heats between two samples with SF (the heats are 42.53 J/g and 44.67 J/g in SF_7.5 and SF_15, respectively). Thus, it can be concluded that an excessive addition of SF (e.g., more than 7.5 wt % of SF) scarcely contributes to the enhancement of hydration reaction during heat treatment period. However, it may improve the strength because of the physical filling effect by unreacted SF particles, which will be discussed later.

XRD Analysis
The results of XRD analysis of raw materials and hydrated kaolinite clay-based pastes at 28 days are shown in Figure 4. The kaolinite clay is composed of quartz, kaolinite, and illite. Hydrated lime consists of mostly calcium hydroxide and calcium carbonate. SF contains only amorphous silica. In all hydrated samples, an alumino-silicate type of mineral (clinoptilolite) is found. However, while calcium hydroxide is strongly detected in Ref (see red-dotted boxes in Figure 4), it is almost not detected in samples with SF (SF_7.5 and SF_15). It is obvious that calcium hydroxide was consumed by the pozzolanic reaction induced from the addition of SF [40,41].
Quartz from kaolinite clay is detected in all hydrated samples. The intensity of quartz peak (at 26.5 • theta) increases with addition of SF (see blue-dotted box in Figure 4). This can be explained by calcium hydroxide in hydrated lime preferentially reacting with amorphous silica first rather than with quartz. Thus, the intensity of unreacted quartz peak is the highest in SF_15. This observation also shows that crystalline quartz in kaolinite clay can react with calcium hydroxide to some extent as a pozzolanic agent. The kaolinite peak (at 12.2 • theta) is slightly weaker in samples with SF. Along with the fact that kaolinite has a weak pozzolanic reactivity [42], the weakness of peak intensity can be also due to the smaller absolute amount of the clay in the paste by the SF addition (Table 2).
Meanwhile, the calcium carbonate peak (at 29.4 • theta) of the hydrated sample increases with SF addition (see purple-dotted box in Figure 4). As can be confirmed by the XRD result, calcium carbonate was initially included in raw hydrated lime. However, the same amount of hydrated lime was added in all samples during the mixing process. In this sense, it is interesting to note that the intensity of calcium carbonate is higher in SF_7.5 or SF_15 than that in Ref. In our system, SF addition increased the carbonation reaction in kaolinite clay-based mortar that contains calcium hydroxide. Similarly, it has been previously reported that the addition of SF increases the carbonation depth of cementitious materials [43,44].

Porosity Analysis
The results of pore structure analysis are shown in Figure 5. The cumulative pore volume of all samples decreases with curing ages (Figure 5d-f). Figure 5a-c compare pore size distributions of the samples with different curing days. It is observed that there are differences in the first peaks of pore diameters depending on SF contents. The first peaks of both Ref and SF_7.5 are formed nearly at 100 nm and are not changed with the number of curing days (Figure 5a,b). On the other hand, the peak of SF_15 is formed at 9 nm, which is approximately 1/10 times compared to the others (Figure 5c). This difference is owing to the physical filling effect by the surplus SF, and it contributed to fine pore structure of SF_15, which will be discussed in detail.

Porosity Analysis
The results of pore structure analysis are shown in Figure 5. The cumulative pore volume of all samples decreases with curing ages (Figure 5d-f). Figure 5a-c compare pore size distributions of the samples with different curing days. It is observed that there are differences in the first peaks of pore diameters depending on SF contents. The first peaks of both Ref and SF_7.5 are formed nearly at 100 nm and are not changed with the number of curing days (Figure 5a,b). On the other hand, the peak of SF_15 is formed at 9 nm, which is approximately 1/10 times compared to the others (Figure 5c). This difference is owing to the physical filling effect by the surplus SF, and it contributed to fine pore structure of SF_15, which will be discussed in detail.

Porosity Analysis
The results of pore structure analysis are shown in Figure 5. The cumulative pore volume of all samples decreases with curing ages (Figure 5d-f). Figure 5a-c compare pore size distributions of the samples with different curing days. It is observed that there are differences in the first peaks of pore diameters depending on SF contents. The first peaks of both Ref and SF_7.5 are formed nearly at 100 nm and are not changed with the number of curing days (Figure 5a,b). On the other hand, the peak of SF_15 is formed at 9 nm, which is approximately 1/10 times compared to the others (Figure 5c). This difference is owing to the physical filling effect by the surplus SF, and it contributed to fine pore structure of SF_15, which will be discussed in detail.  As can be seen in Figure 6 that compares cumulative pore volumes of three samples at 28 days, the extra SF (over 7.5 wt %) is expected to play a role as a filler. Pores of the mortar between 40 and 500 nm in their diameter can be filled with the extra (unreacted) SF, because its size distribution (see Figure 2) coincides with the pore range. Thus, a part of macropores (25 nm-5 μm) [45] could be effectively filled with the filler, while increasing the volume of the smaller pores (<25 nm). In short, this filling effect led SF_15 to have fine pore structure (Figure 6b). The second peak position of the two samples with SF (SF_7.5 and SF_15) are formed at approximately 10 μm, which is 10 times larger than that in Ref (Figure 5a-c). The pore range of this peak (5-50 μm) is related to preexisting microcracks and entrained or entrapped air [45]. These pores, which did not change with curing ages, were possibly formed with the change of rheology of the mortar by SF addition [46]. As can be seen in Figure 6 that compares cumulative pore volumes of three samples at 28 days, the extra SF (over 7.5 wt %) is expected to play a role as a filler. Pores of the mortar between 40 and 500 nm in their diameter can be filled with the extra (unreacted) SF, because its size distribution (see Figure 2) coincides with the pore range. Thus, a part of macropores (25 nm-5 µm) [45] could be effectively filled with the filler, while increasing the volume of the smaller pores (<25 nm). In short, this filling effect led SF_15 to have fine pore structure (Figure 6b).  As can be seen in Figure 6 that compares cumulative pore volumes of three samples at 28 days, the extra SF (over 7.5 wt %) is expected to play a role as a filler. Pores of the mortar between 40 and 500 nm in their diameter can be filled with the extra (unreacted) SF, because its size distribution (see Figure 2) coincides with the pore range. Thus, a part of macropores (25 nm-5 μm) [45] could be effectively filled with the filler, while increasing the volume of the smaller pores (<25 nm). In short, this filling effect led SF_15 to have fine pore structure (Figure 6b). The second peak position of the two samples with SF (SF_7.5 and SF_15) are formed at approximately 10 μm, which is 10 times larger than that in Ref (Figure 5a-c). The pore range of this peak (5-50 μm) is related to preexisting microcracks and entrained or entrapped air [45]. These pores, which did not change with curing ages, were possibly formed with the change of rheology of the mortar by SF addition [46]. The second peak position of the two samples with SF (SF_7.5 and SF_15) are formed at approximately 10 µm, which is 10 times larger than that in Ref (Figure 5a-c). The pore range of this peak (5-50 µm) is related to preexisting microcracks and entrained or entrapped air [45]. These pores, which did not change with curing ages, were possibly formed with the change of rheology of the mortar by SF addition [46]. The measured pore volume was divided into four size ranges in Figure 7, on the basis of the international union of pure and applied chemistry system [45]. However, the micropore range was modified from <1.25 nm to <4.5 nm based on the previous studies which have shown that the enhanced pozzolanic reaction considerably changes the pore structure especially below 4.5 nm range [15,47]. Thus, this modification can help to effectively explain the change of pore structure by the intensified pozzolanic reaction.
The measured pore volume was divided into four size ranges in Figure 7, on the basis of the international union of pure and applied chemistry system [45]. However, the micropore range was modified from <1.25 nm to <4.5 nm based on the previous studies which have shown that the enhanced pozzolanic reaction considerably changes the pore structure especially below 4.5 nm range [15,47]. Thus, this modification can help to effectively explain the change of pore structure by the intensified pozzolanic reaction. As expected by Figure 5, the proportions of micro and mesopores of SF_15 are significantly higher than those of the other samples, due to the physical filling effect. However, the proportion of micropores of SF_15 steadily decreased with curing age. In particular, unlike the other samples, the first peak of SF_15 was changed to larger size (from 9 to 17 nm) between 28 and 91 days (see Figure 5c). This indicates that the volume of mesopores (4.5-25 nm) of SF_15 were increased due to expansion with curing age. It has been reported that the pores finer than 10 nm mainly affect shrinkage and creep of hardened cementitious materials [46], and also the degree of capillary tension and stress for drying shrinkage is increased with a finer pore system (below 80 nm range in pore diameter) [15,48]. Thus, this distinct expansion might be attributed to the accelerated drying shrinkage of matrix and resultant micro cracks, as shown in the low W/B (0.25) cement paste with SF (20% of cement) [49].

Compressive Strength
The compressive strength of prepared kaolinite clay-based mortar increased with SF contents ( Figure 8  As expected by Figure 5, the proportions of micro and mesopores of SF_15 are significantly higher than those of the other samples, due to the physical filling effect. However, the proportion of micropores of SF_15 steadily decreased with curing age. In particular, unlike the other samples, the first peak of SF_15 was changed to larger size (from 9 to 17 nm) between 28 and 91 days (see Figure 5c). This indicates that the volume of mesopores (4.5-25 nm) of SF_15 were increased due to expansion with curing age. It has been reported that the pores finer than 10 nm mainly affect shrinkage and creep of hardened cementitious materials [46], and also the degree of capillary tension and stress for drying shrinkage is increased with a finer pore system (below 80 nm range in pore diameter) [15,48]. Thus, this distinct expansion might be attributed to the accelerated drying shrinkage of matrix and resultant micro cracks, as shown in the low W/B (0.25) cement paste with SF (20% of cement) [49].

Compressive Strength
The compressive strength of prepared kaolinite clay-based mortar increased with SF contents ( Figure 8) because of the intended pozzolanic reaction and filling effect. The 91 days compressive strengths of SF_7. 5  The measured pore volume was divided into four size ranges in Figure 7, on the basis of the international union of pure and applied chemistry system [45]. However, the micropore range was modified from <1.25 nm to <4.5 nm based on the previous studies which have shown that the enhanced pozzolanic reaction considerably changes the pore structure especially below 4.5 nm range [15,47]. Thus, this modification can help to effectively explain the change of pore structure by the intensified pozzolanic reaction. As expected by Figure 5, the proportions of micro and mesopores of SF_15 are significantly higher than those of the other samples, due to the physical filling effect. However, the proportion of micropores of SF_15 steadily decreased with curing age. In particular, unlike the other samples, the first peak of SF_15 was changed to larger size (from 9 to 17 nm) between 28 and 91 days (see Figure 5c). This indicates that the volume of mesopores (4.5-25 nm) of SF_15 were increased due to expansion with curing age. It has been reported that the pores finer than 10 nm mainly affect shrinkage and creep of hardened cementitious materials [46], and also the degree of capillary tension and stress for drying shrinkage is increased with a finer pore system (below 80 nm range in pore diameter) [15,48]. Thus, this distinct expansion might be attributed to the accelerated drying shrinkage of matrix and resultant micro cracks, as shown in the low W/B (0.25) cement paste with SF (20% of cement) [49].

Compressive Strength
The compressive strength of prepared kaolinite clay-based mortar increased with SF contents ( Figure 8) because of the intended pozzolanic reaction and filling effect. The 91 days compressive strengths of SF_7. 5   However, the compressive strength dropped between 28 and 56 days in all specimens. Ref reached the maximum strength (7.35 MPa) at 28 days, then the strength decreased by 3.05% at 56 days and increased again by 15.6% at 91 days. The strength of SF_7.5 decreased by 18.1% at 56 days and increased by 15.9% again. On the other hand, a significant strength loss occurred in SF_15 sample without recovery, i.e., the strength steadily decreased after three days until 91 days. In particular, significant strength loss (30%) occurred between 28 and 56 days.

Investigation on Strength Reduction
Unlike the other samples, the compressive strength of SF_15 steadily decreased after three days (Figure 8). To investigate possible reasons of the strength reduction, pore structures of the samples are additionally discussed in this section. As can be observed in Figure 5, the pore size distributions of Ref and SF_7.5 were almost unchanged between 14 and 91 days, which means that their pore structures stabilized at 14 days. On the other hand, in the case of SF_15, the first peak of pore diameters moved from 9 to 17 nm between 28 and 91 days (Figure 5c). This movement means that the pores around 10 nm expanded in size (by approximately double), possibly due to the accelerated drying shrinkage with a fine pore structure by excessive SF addition [15,49]. Similarly, Rao has reported that the addition of SF has a significant influence on the drying shrinkage of cement mortar due to the strong pozzolanic reaction and pore size refinement mechanism, especially with 15% (by wt % of cement) SF addition. In his study, the shrinkages of cement mortars with 0%, 10%, and 15% of SF showed 75.7, 303, and 812 µm/m at 28 days, and 114, 756, and 1030 µm/m at 60 days [50].
In summary, there is an optimum SF to binder ratio in the developed kaolinite clay-based mortar considering its long-term strength. Greater SF addition above the optimum ratio can have a negative effect on the strength. Thus, based on the conducted experiments herein, it can be concluded that 7.5% of SF to binder ratio is more suitable than 15%, considering the long-term performance and economic point of view.

Conclusions
The results of this study are summarized as follows.
(1) In the heat of hydration test, all samples showed two main peaks regardless of SF addition.
SF addition increased the hydration heat emission. The interesting result is that the cumulative hydration heat of SF_15 was only 5% higher than that of SF_7.5, although SF content in SF_15 was almost twice than that in SF_7.5. This result reveals that the excessive addition of SF has no chemical effect on the hydration reaction. (2) The XRD analysis showed that the mortar cured for 28 days contains quartz, kaolinite, illite, remaining calcium hydroxide, calcium carbonate, and clinoptilolite. Unlike Ref, calcium hydroxide was barely detected in two SF samples (SF_7.5 and SF_15), i.e., it was mostly consumed by the intended pozzolanic reaction between hydrated lime and SF. Thus, the intensified hydration reaction that was confirmed by the hydration heat test, was attributed to the intended pozzolanic reaction by adding two mineral admixtures. (3) The total pore volume of the two samples with SF was lower than that of Ref. Between the two samples with SF, the proportion of the pores, whose size range is similar with that of SF particles, was significantly lower in SF_15 than SF_7.5. This verifies the physical filling effect provided by excessive (unreacted) SF particles. However, in the pore size distribution of SF_15, the first peak was changed from 9 to 17 nm between 28 and 91 days. This change can be related to the long-term strength reduction in a fine pore structure of SF_15. (4) The compressive strength of the mortar was increased with SF addition. The main contribution was the intended pozzolanic reaction, but the filling effect also contributed to the strength gain as well. Based on the compressive strength, it was confirmed that the developed kaolinite clay-based mortar with SF has sufficient mechanical properties as a structural material.
(5) However, there is an optimal SF to binder ratio in the developed mortar. Excessive SF addition resulted in the degradation of long-term compressive strength. Considering price competitiveness and long-term strength, a SF to binder ratio of 7.5%, rather than 15%, is suitable. However, it should be noted that this optimum amount of SF can be different depending on the binder selection (i.e., the ratio of kaolinite clay and calcium hydroxide). The investigated mix design herein can be further expanded to develop other sustainable structural materials, such as natural minerals with low reactivity or waste materials with nonbinding properties.