Ordinary Portland cement (OPC) was used in this study. OPC had equivalent Na₂O of 0.6%; thus it could be classified as a low alkali cement according to ASTM C150 (15). The physical properties of OPC are shown in Table 1 and the chemical compositions are given in Table 2.

Bagasse ash (BGA) was obtained from a sugar industry in central part of Thailand. The received bagasse ash directly from the industry or original BGA had large particles with high moisture content of 50%. The high moisture content of BGA was due to the spraying of water on the bagasse ash to reduce the spreading of dust. Thus, the BGA was placed in an oven at a temperature of 110ºC for a period of 24 h and then was screened using a No. 16 sieve to remove large particles and incomplete burning of sugarcane bagasse. Next, the BGA was ground by a ball mill until the particles retained on a No. 325 (opening 45 μm) sieve were 33±1% and 5±1% by weight in accordance with ASTM C430 (16). The specific gravity of OPC and BGA were determined according to ASTM C188 (17). The abbreviations of BGA with the particles retained on a No. 325 sieve of 33±1% and 5±1% by weight were defined as 33BGA and 5BGA, respectively. The physical properties of ground BGA are also presented in Table 1.

The different BGA particles were used to investigate the effect of finenesses on ASR. The BGA with particles retained on a No. 325 sieve of 32–34% was the lowest fineness of a pozzolanic material for use in concrete as specified by ASTM C618 (18) while BGA with particles retained on a No. 325 sieve of 4–6% were used as a high fineness pozzolan. BGA particles with different finenesses were used to replace OPC at rates of 10, 20, 30, and 40% by weight of binder for studying the effect of BGA replacement on ASR. Moreover, the fineness of BGA was also presented as the particle size distribution (see Figure 1). The median particle size (d₅₀) was the particle with 50% of cumulative passing from graph of particle size distribution and can be found in Figure 1 which 5BGA and 33BGA had median particle size of 20.05 and 16.11 μm, respectively.

Figure 2 shows the X-ray diffraction (XRD) analysis of the high fineness of ground BGA (5BGA), plotted using the Lin (count) and 2q. The mineral compositions of BGA were identified by X-ray diffractometer; Bruker AXS model D8 Discover. The conditions of test were targeted Cu, voltage 40 kV, current 40 mA and scanning from 5 to 70 º2q with increment 0.02 degree/step and scan speed 0.2 second/step. The results presented the clear peak of quartz SiO₂ of hexagonal type and calcite (CaCO₃) in the form of Rhombo H axes. Scanning electron microscope (SEM) of 5BGA and 33BGA samples (from JEOL model JSM-6610) using accelerating voltage 15–20kV, magnification 1000–2000 times, and other details are presented in Figure 3 and Figure 4, respectively. The SEM images of the ground BGA showed that 5BGA had smaller particles and was crushed into irregular shapes while 33BGA had large particles and high porosity.

For the chemical compositions shown in Table 2, 5BGA sample was only used to determine the chemical compositions because the finenesses of pozzolans did not affect the chemical compositions (19). Chemical compositions of the materials were identified by wavelength-dispersive X-ray fluorescence spectrometer (WD-XRF) model Bruker S4 Explorer and the conditions test were 50kV of voltage, current 250 μA, beam filter Cu 250 μm, live time 100s, and limit 50 KCps. In this study, 5BGA had SiO₂ + Al₂O₃ + Fe₂O₃ of 66.5% and less than 4% SO₃ by weight. The oxides of Na₂O and K₂O were 0.1% and 2.1%, respectively. Thus the Na₂O equivalent content was calculated to be 1.5%, which was lower than the limited values recommended by Glauz et al. (20), who found that alkali content lower than 3% of fly ash class F or pozzolan class N was effective against ASR. The major oxide of BGA was SiO₂, representing 52.3%, suggesting that a high content of SiO₂ and a low alkali content of BGA was a suitable for use as a pozzolanic material.

The loss on ignition of the ground bagasse ash was 16.2%, which was higher than 6% and did not meet the requirement of ASTM C618 (18) for class F fly ash. The high LOI of BGA was due to the incomplete combustion process which did not remove totally the organic phase included the calcite in BGA also contributed to LOI. In addition, LOI of BGA was investigated at the temperature of 750±50ºC according to ASTM C311 (21), both organic phase and calcite were burnt caused to weight loss.

However, many researchers (4, 22) confirmed that the high LOI of bagasse ash at 19.36 to 20.4% had little effect on the compressive strength and sulfate resistance of concrete or mortar. Moreover, the use of bagasse ash with high LOI to investigate the expansion due to ASR of mortar was rarely found.

For compressive strength test, a river sand was used as a fine aggregate. The fine aggregate had a specific gravity of 2.64, a water absorption of 1.21%, and a fineness modulus of 1.52.

For ASR test, a crushed coarse aggregate with highly reactive was used to investigate the effect of finenesses and cement replacement of BGA on ASR. This highly reactive aggregate from central of Thailand was used as a fine aggregate and was processed to have a grading requirement according to ASTM C1567 (23), as shown in Table 3. The fine aggregate had a specific gravity of 2.63 with water absorption of 0.83%. The high reactive aggregate was investigated for the mineral properties and found that the high reactive aggregate was siliceous limestone with major mineral composition of calcite included minors of quartz and dolomite (see Figure 5). Chemical compositions of the high reactive aggregate showed CaO of 91.3%, SiO₂ of 5.5%, and other compositions in small percentages (see Table 2).

Mortar cubes of 50 × 50 × 50 mm³ were cast to investigate the compressive strength of mortar containing ground BGA with different finenesses and replacements. River sand was used as a fine aggregate. The ratio of binder to fine aggregate was set as a constant 1:2.75 by weight, and all mortars were prepared to have flow at the range of 105 to 115%, as shown in Table 4. Note that the use of ground BGA to replace OPC required higher water to maintain the flow between 105–115%. For example, the use of 5BGA to replace OPC of 10, 20, 30, and 40% required W/B ratios of 0.69, 0.71, 0.73, and 0.76, respectively, while the W/B ratio of control mortar (CT mortar) was 0.67. Moreover, use of coarser BGA (33BGA) to replace OPC required a higher water content than the use of finer BGA (5BGA) and the higher replacement of 33BGA resulted in the higher water requirement to maintain the specified flow of mortar. For example, 33BGA20 mortar had a W/B ratio of 0.76 while that of 5BGA20 mortar had W/B ratio of 0.71. This was due to the irregular shapes and high porosity of the large particles of BGA, which caused more friction and absorb more water than the small particles of BGA. The result was in agreement with the findings of other researches (24), who used ground rice husk ash and limestone filler to replace cement.

Mortar prisms with 25 × 25 mm² in cross section and 285 mm in length were cast. In addition, ground BGA samples with different finenesses were also used separately to replace Portland cement at rate of 10, 20, 30, and 40% by weight of binder. The binder to fine aggregate (highly reactive aggregate) ratio was set as a constant of 1:2.25 by weight to cast mortar. The water-to-binder (W/B) ratio of 0.47 was used for ASR test, as specified by ASTM C1567 (23).

After casting for 24 h, the mortar cubes of 5 × 5 × 5 cm³ were removed from the molds and then cured in saturated lime water. The compressive strengths of mortars were determined at the ages of 7, 28, 60 and 90 days. At each testing age, five mortar specimens were tested, and the average value was used as suggested by ASTM C109 (25).

The alkali silica reactivity potential of the aggregate was assessed following the procedure described in ASTM C1567 (23). The accelerated mortar bar test method was used to determine ASR by increasing temperature and using sodium hydroxide (NaOH) solution for increasing the concentration of external alkali for activating the silica in aggregate. The mortar bars were removed from the molds after casting for 24 h, cured in water at temperature of 80 °C for 24 h, and then immersed in 1N sodium hydroxide (NaOH) solution (which had volume 4±0.5 times of the sample). Mortar bars were measured the length until 28 days and the length change was calculated using equation (1). According to ASTM C1567 (23), the mortar with expansion below 0.10% is considered the aggregate as innocuous, that between 0.1–0.2% is considered the aggregate as slow reactive, and that higher than 0.2% is considered the aggregate as high reactive [1].

where L = change in length at x age, %; Lx = comparator reading of the specimen at x age minus comparator reading of reference bar at x age, mm; Li = initial comparator reading of the specimen minus comparator reading of reference bar at that same time, mm; G = nominal gage length, 250 mm.

Figure 6 and Figure 7 present the relationship between the compressive strength of mortar and replacement of ordinary Portland cement by 5BGA and 33BGA, respectively. The control mortar (CT) had compressive strengths of 22.8 and 29.4 MPa at the ages of 7 and 28 days, respectively, and increased to 32.2 and 33.5 MPa at the ages of 60 and 90 days, respectively. At 28 days, mortars containing 5BGA at rates of 10, 20, 30, and 40% by weight of binder had compressive strengths of 27.5, 26.6, 24.1, and 23.2 MPa or 94, 90, 82, and 79% of CT mortar, respectively. 5BGA mortars had compressive strength more than 75% of CT mortar at 28 days. Moreover, the compressive strengths of 5BGA10, 5BGA20, 5BGA30, and 5BGA40 mortars could increase to 102, 92, 86, and 81% of CT mortar, respectively at 90 days. For mortars containing 33BGA at rates of 10 to 40% by weight of binder had compressive strengths of 87 to 52% of CT mortar at 28 days and increased to 90 to 54% of CT mortar at 90 days. The results showed that the compressive strengths of mortars containing small particles of BGA (5BGA) were higher than the mortars containing large particles of BGA (33BGA), consistent with many researchers studied the use of concrete or mortar containing bagasse ash (3) or other pozzolans, such as fly ash and palm oil fuel ash (26, 27).

The mortars containing smaller particles of ground BGA produced higher compressive strength than the mortars containing larger BGA particles at all ages. This was due to two factors, the first factor was 5BGA had smaller particles and higher surface area; thus, 5BGA had faster reactive and higher area for reaction than 33BGA particles, and the second factor was that the filler effect of smaller particles of 5BGA filled voids in the matrix and made the matrix denser (24). Similar results were found by other researchers (24, 28), they used other pozzolans as a replacement of cement.

The replacement of 5BGA at 10% by weight of binder (5BGA10 mortar) gave the highest compressive strength of 101 and 102% of CT mortar at 60 and 90 days, respectively. Note that 5BGA10 mortar had W/B ratio higher than CT mortar of 3%. In general the higher W/B ratio the lower compressive strength; however, 10% of BGA as a cement replacement could produce compressive strength from pozzolanic reaction higher than the loss of 10% of hydration reaction by cement. The results also suggested that the high LOI of BGA (16.2%) had little effect on the compressive strength of mortar and was similar to the other results (22, 29).

In the case of coarse BGA (33BGA), the mortar containing 33BGA at 10% replacement of cement gave lower compressive strength than CT mortar but the strength was still higher than 80% of CT mortar at all ages and the mortar 33BGA20 gave the compressive strength higher than 75% of CT mortar at 28 days. When 33BGA was used to replace cement at 30 and 40%, 33BGA30 and 33BGA40 mortars gave the compressive strengths lower than 75% of CT mortar or at the ranges of 47–59% of CT mortar. The replacement of OPC by 33BGA affected the reduction of CaO from OPC which was the main oxide for producing Ca(OH)₂, it was the substrate to produce the pozzolanic reaction. In addition, 33BGA mortar had high water to binder ratio included 33BGA had large particles caused the low compressive strength of mortar. This result was consistent with many studies, from which the use of smaller particles of pozzolanic materials provided higher compressive strength than the coarser one in concrete (30). However, 5BGA mortars utilized lower water content than 33BGA mortars to control flow at range of 110±5% and the low W/B ratio was a factor to provide more compressive strength of mortar.

Figure 8 shows the expansion of CT and 5BGA mortars, which are immersed in 1N NaOH solution at a period of 28 days. CT mortar had an expansion due to ASR more than 0.20% at 14 days and the expansion increased to 0.38% at 28 days, suggesting that the aggregate used in the mixture was highly reactive for ASR because the expansion limitation for innocuous aggregate in mortar due to ASR was not more than 0.10% at 14 days.

When 5BGA was used to replace Portland cement at rates of 10, 20, 30, and 40% by weight of binder, the expansions of mortars due to ASR were reduced. For example, mortars 5BGA10, 5BGA20, 5BGA30, and 5BGA40 had the expansions due to ASR of 0.1269, 0.0532, 0.0436, and 0.0435%, respectively, at 14 days and increased to 0.2661, 0.1274, 0.1019, and 0.1195%, respectively, at 28 days. The reduction of the expansion due to ASR depended on the replacement of BGA in mortar. Moreover, the effective replacement of 5BGA was 20% up to 40% by weight of binder because the mortars with these replacement rates could provide low expansions due to ASR at 14 days in the ranges of 0.043–0.053%, which were lower than 0.10%. The mechanism of reducing the potential ASR by the use of supplementary cementitious materials (SCMs) or pozzolan cement was reported by many researchers (31–36) since pozzolanic materials could react with Ca(OH)₂ and led to decrease concentration of hydroxyl ion and pH in pore solutions. Moreover, the use of pozzolan with low alkali content replaced OPC caused reduction of the total alkali contribution (37).

Figure 9 presents the results of ASR expansions of mortars containing 33BGA at rates of 10 to 40% by weight of binder. 33BGA10, 33BGA20, 33BGA30, and 33BGA40 mortars had the expansions due to ASR at 14 days of 0.1299, 0.2315, 0.3376, and 0.1922%, respectively, and the expansions were 0.2831, 0.3144, 0.5379, and 0.2093%, respectively at 28 days. Note that the mortars containing 33BGA had the expansions due to ASR higher than 0.10%.

According to ASTM C618 (18), the expansion due to ASR of mortar at 14 days should be less than that of CT mortar. This finding suggested that the mortars containing 5BGA at all replacement rates and the mortars containing 33BGA at rates of 10, 20, and 40% resulted in the expansion due to ASR being less than that of CT mortar.

Figure 10 presents the surfaces of mortars containing high and low finenesses of BGA. The figures demonstrated that the use of 5BGA was much more effective than 33BGA in reducing the expansion of mortar due to ASR. The replacing 33BGA at 10 and 20% in mortar provided expansions lower than CT mortar; however, the mortars with 30 to 40% replacement had higher expansions due to ASR and cracks were observed in the mortar bars, as shown in Figure 10. Mortar 33BGA40 had crumbly surfaces, although it exhibited a low expansion and also a low compressive strength. Moreover, 20 and 30% replacements of 33BGA resulted in many cracks in the mortars.

Replacing OPC by 5BGA caused a reduction of CaO and alkali in the binder, which was a substrate of ASR; thus this reaction was lower. This finding was similar to Aydin et al. (38) who used original fly ash and ground fly ash to replace cement at rates of 5–60% and caused a reduction of CaO in the binder. They found that the ground fly ash was more effective to reduce ASR expansion than original fly ash and the effective replacement of both original and ground fly ashes were 40 to 60% by weight of binder.

In this study, OPC had high amount of CaO while BGA was primarily composed of SiO₂. When OPC was replaced by BGA, the amount of CaO was decreased, whereas SiO₂ was increased from which the lower CaO and the higher SiO₂ had affected to lower expansion due to ASR because low CaO produced low Ca(OH)₂ content (which was a part of substrate), while a higher SiO₂ in BGA could react with Ca(OH)₂ to form to C-S-H; thus, a denser paste was obtained. According to the report of Forster, et al. (35), the pozzolan cement had a CaO/SiO₂ ratio lower than OPC from which more C-S-H gel could entrap more alkalis and reduce the pH in pore fluid in the concrete. Figure 11 presents the pore size distributions of 5BGA pastes which were investigated by using mercury intrusion porosimetry (MIP) method. Pastes for this investigation were removed from the mold after 24 h for casting and cured in boil water at 80 ºC for period of 24 h before immersion in NaOH solution. The results showed that 5BGA10, 5BGA20, 5BGA30, and CT pastes had cumulative pore volume in the same ranges while 5BGA40 paste had high cumulative pore volume. However, pore volume and pore size of paste reduced with the increased paste ages (39).

However, the above explanation could not be used for mortars containing 33BGA because the reduction of CaO/SiO₂ ratio in 33BGA mortars resulted in a higher ASR expansion compared to mortar containing 5BGA. 33BGA had many large particles and less surface area than 5BGA; thus, silica in 33BGA was low reactive and caused more alkali to remain in cement, resulting in higher ASR. Moreover, 33BGA could not fill voids in mortar (due to large particles) and produced low compressive strength; thus, a sodium hydroxide solution could easily penetrate into mortar to activate silica, both in aggregate and 33BGA, causing more alkali-silica reactions. In addition, high fineness of BGA could better react with Ca(OH)₂ from hydration reaction and reduced the amount of CaO and SiO₂ for producing ASR lower than the BGA with low fineness. The result was similar to the use of rice husk ash with different finenesses studied by Zerbino et al. (14) who reported that the unground rice husk ash affected high expansion due to ASR of mortar while the use of ground rice husk ash could reduce expansion due to ASR of mortar to be lower than 0.10%. The pore size distribution of 33BGA pastes are given in Figure 12, 33BGA40 paste had high cumulative pore volume. The results confirmed that the higher replacement of OPC by 33BGA resulted in higher amount of pores and voids.

In contrast, the replacement of OPC by 5BGA in mortar reduced Ca(OH)₂ due to the pozzolanic reaction, thus, produced the lower ASR gel and resulting in a low expansion due to ASR. Moreover, the reduction of Ca(OH)₂ and the production of more C-S-H gel due to the pozzolanic reaction resulted in a denser paste and caused the low permeability and slowly ASR to be occurred.

Note that high LOI of BGA did not affect the expansion due to ASR because 5BGA, which had a high LOI of 16.2%, could be used effectively to reduce the expansion due to ASR. Likewise, the other pozzolan such as palm oil fuel ash (POFA) from research of Awal and Hussin (13), which had high LOI of 18% could prevent the expansion of mortars while the results of the use of high LOI biomass fly ash (10.4%) on ASR could reduce the expansion compared with CT mortar similarly to this research (40).