The first study evaluates morphological properties and chemical composition of RHA via two combustion types including straight high temperature (one period) and first lower temperature then higher temperature combustion (two periods), which aims to explore a new burning method to decrease the K₂O content in RHA. The raw rice husk was burnt in a box furnace which could achieve the automatic function of controlling temperature and time. The combustion process contains two key factors, namely the burning temperature (300 ºC, 600 ºC) and residence time (30 min -180 min). In each combustion, rice husk with a mass of 1kg was embedded in the box furnace and then burned according to the set procedure. Nine different schemes were presented in this study as listed in Table 2 and Figure 1.

A second study was developed to reveal the influence of RHA particle size on the mechanical properties of mortar. RHA produced by L60H120 combustion type was selected and ground for 5 - 30 min in a planetary ball mill at a speed of 60 r/min. The cumulative distribution curve of RHA was measured by laser granulometry (Mode JL-6000) equipment as presented in Figure 2. Due to the chemical properties of RHA were significantly close to silica fume, which also consists of non-crystalline silicon dioxide and high specific surface area. Hence, substituting cement by RHA with four mean particle sizes into mortar was conducted to investigate the pozzolanic activity of RHA in accordance with GB/T 27690 (27). The mixture proportions of mortar are shown in Table 3. The sample names were derived from the mean particle size of RHA. For instance, the sample of RHA35.5 implies that RHA substituted 10% of cement (by mass) with a mean particle size of 35.5 μm. Whereafter, grey correlation analysis was employed to analyze the influencing degree of each group of RHA on the basis of the compressive and flexural strength of mortar.

In this paper, the interval grey correlation analysis based on the microscopic or macroscopic geometric proximity of the factor sequence was adopted to assess the correlation degree among subsequences and further find an optimum subsequence for impacting the parent sequence (28). In the grey correlation theory, the index column is expressed as x m ( n ) , where i is the natural number sets, and m is related to the influencing factors _. It should be noted that there are i assessment plans in the analysis, i.e., x i = x i 1 , x i 2 , … , x i ( n ) . Meanwhile, parent sequences and subsequences are represented as x 0 = x 0 ( n ) a n d x i = x i n ( i = 1,2 , … , m ) , respectively (29). Therefore, the correlation coefficient, ξ i ( j ) , between subsequence and parent sequence with respect to j ^th factor could be obtained by Equation [1] (30):

Where Δ i is equal to x 0 j - x i j ρ is the resolution coefficient, which is to enlarge the difference between the various coefficients within the interval [0, 1]; generally, it is fixed at 0.5 (31). Δ i   m i n     and Δ i   m a x denote the minimum and maximum difference in the first level using Equation [2]. Meanwhile, m a x i ( Δ i m a x ) and m i n i ( Δ i m i n ) are the maximum and minimum difference in the second level, respectively using Equation [3] (29).

Subsequently, the grey correlation degree, r _i , could be obtained from Equation [4] by manipulating the grey correlation coefficients, which reflect the correlation degree of each subsequence in the same parent sequence. It means that we can find the optimal subsequence by comparing the correlation degree of all subsequences.

Where ξ i j is the correlation coefficient from Equation [1], and j is within the interval [1, n].

On the other hand, using Δ i = x 0 j - x i j calculates the correlation coefficient and correlation polarity including the positive and negative correlation among the subsequence could not be estimated accurately. Hence, in this paper, the judging method of the correlation polarity was defined by a new mathematical way listed in Equation [5], which confirms the correlation including positive and negative correlation from Equation [6] and Equation [7]. The expressions for each of these formulas are shown as follows (32):

Where both of σ n and σ i are the new defining value, is the natural number set, and m represents the influencing parent sequence in the analysis. In Equation [6], the value of σ i overs than zero, while k is within the interval [1, n]. It was noted that subsequence has a positive correlation on enhancing parent sequence when σ k satisfies the requirement of Equation [6]. Otherwise, a negative correlation that subsequence plays a weakening role on the parent sequence is emerged according to Equation [7].

Scanning electron microscopy (SEM, Helios G4 UC, Japan) with energy-dispersive X-ray (EDX) provided by the analytical and testing center of Sichuan University was used to obtain the microstructure of RHA. Moreover, rice husk ash was compacted to a flake layer; thereafter, the surface scan of energy-dispersive X-ray was employed to obtain the accurate chemical elements of RHA samples, as utilized in some previous studies (33, 34).

All of the mortar mixtures were bathed into JJ-5 cement and mortar mixer with a 5L capacity. Initially, water was poured into the JJ-5 mixer, then cement and RHA were blended for 120 s throughout the mixing procedure. Then, sand was incorporated slowly and thoroughly mixed for 60 s. Finally, the fresh paste was shaped in the molds and stored with plastic films at the air temperature. After 24 hours, all mortar samples were demolded. The flexural and compressive test complying with GB/T 27690 (27) was conducted on three samples for each mortar mixture. Paste bars having the dimension of 40 mm×40 mm×160 mm were cast. After curing at 65 ± 2 ºC for six days and then 21 days in a standard room, all specimens were subject to the compressive and flexural test. Prior to the test, a piece of an unforced segment was cut from the specimens by a saw bar with a thickness of about 0.6 mm and polished by sanding paper, whiling the cement hydration process of stores pieces was ceased by isopropyl alcohol and dried in an oven at 60 ºC for two hours.

The crystal structure of RHA was measured using an X-ray diffractometer (XRD, EMPYREAN, Holland) operating at an accelerating voltage of 60 kV and 60 mA. Under continuous scanning pattern, Cu Kα radiation worked as the source of X-rays, and crystalline phases were scanned from 5◦ to 80◦ with a 2θ increment of 0.02◦/step. Although the scanning was performed up to 85° (2θ), the range of 10 - 50° was analyzed. This is because essential peaks were satisfactorily detected in this range, and meaningful data could not be obtained due to the overlap of peaks at a high angle of > 50° (2θ).

To obtain the crystal structure of paste samples blended RHA, the hydration of paste samples cured for 28 days was stopped by isopropyl alcohol and dried in an oven at 60 ºC for two hours, thereafter, X-ray diffraction was applied after grinding the pieces and passing the sieve of 240 meshes.

Figure 3 displays the macro-morphology and SEM micrographs of RHA using one period burning method. It can be seen that, as the burning time increases at 600 ºC, the black color of RHA tends to change to gray-white. As shown in the SEM micrographs, the scattered particles on the outer surface of RHA present a decreasing development with the increment of high temperature at 600 ºC. This phenomenon demonstrates that rice husk biomass such as lignin, cellulose, and spent carbon polymer can be burnt with the increase in the total burning time, which in turn presents a whiter color.

Figure 4 shows the macro-morphology and SEM micrographs of RHA by two periods burning methods. Similarly, the white color of RHA tends to enhance as the total burning time increases; however, RHAs produced in L60H90, L90H90, and L60H120 appear remarkably similar gray-white color, which is possible due to the reducing biomass which could be wholly oxidized and then burnt when the burning time was up to 90 min at 600 ºC. This perspective could be certified by the low carbon content in Table 5. From the SEM micrographs, massive granular impurities are attached on the outer surface of RHA in L30H30 due to the insufficient combustion time. The scattered particles on the outer surface of RHA present a rapidly decreasing tendency as the total burning time increases.

The XRD patterns of RHA by nine burning methods are presented in Figure 5. It could be seen that all samples emerge no noticeable sharp peak from the diffraction curve. Besides, a hump at 2θ=near 15°-25° could be observed, which confirms the presence of amorphous silica in terms of the qualitative assessment of RHA. Previous researchers, who explored the varied temperature to the influence on the crystallinity of silica in RHA from XRD patterns, also found that amorphous silica could be obtained and untransformed to crystalline silica such as cristobalite and tridymite using temperature lower than 700 ºC, which was is responsible for the critical pozzolanic activity of RHA (35, 36).

In this study, a semi-quantitative method, energy-dispersive X-ray, was used to determine the chemical composition of RHA, as listed in Table 4. From Table 4, in one period, the carbon content of RHA could be noticeably reduced when the total burning time at 600 ºC increased from 60 min to 120 min; meanwhile, the silica content of RHA also could noticeably increase. In two periods, the carbon content of RHA in L30H60 could be noticeably decreased by 86.5% compared with that of RHA in L30H30. Moreover, the silica content of RHA increased rapidly by prolonging the high temperature stage in L30H30, L30H60, and L30H90. However, the silica content of RHA by extending the low temperature stage in L30H90, L60H90, and L90H90 has an insignificant variation, indicating that it was hard to increase when the burning time at 600 ºC was up to 90 min.

Figure 6 shows the potassium and silica content in RHA with different combustion types. It is observed in Figure 6 that the potassium content in RHA generated by one period decreased by 58.2%, 40.9%, and 12.3% (average value), respectively, compared to the potassium of RHA produced by two periods that the total burning time was 60 min, 120 min, and 180 min, respectively. This phenomenon indicates that increasing combustion time at a low temperature of 300 ºC could reduce potassium content when total combustion time was constant. Two patterns of potassium were presented in RHA: potassium oxide and potassium complex compound. Potassium oxide in RHA could be dissociated to potassium when the temperature exceeds 620 K (347ºC) (23), thereafter the potassium reacts with other elements to form a potassium complex compound that could fix carbon to form black particles in RHA, thus resulting in a high potassium content in RHA. Hence, prolonging the low temperature at 300 ºC under 620 K could reduce the formation of black particles in RHA; meanwhile, potassium whose melting point is 336.8 K could fully evaporate accompanied by carbon in RHA as the combustion time increases at 300 ºC. On the other hand, RHA straightly burnt from room temperature to 600 ºC contributes to a longer combustion time above 620 K (347 ºC), thus forming a more potassium complex compound and maintaining a high potassium content in RHA. However, the decreasing potassium content of RHA in L90H90 and L60H120 was not noticeable compared to that of RHA in L0H180. Moreover, the potassium content of RHA in L30H90, L60H90, and L90H90 was 1.47%, 0.93%, and 1.61%, respectively, which implied that a longer time at 300 ºC than 60 min could not always motivate the evaporation of potassium accompanied by carbon due to the restricted oxygen in the furnace which was supplied from the outside air in this study. Hence, a reasonable low-temperature stage should be applied to effectively reduce the potassium content in RHA.

On the other hand, considering the amorphous silica and potassium content as well as the energy consumption among nine burning types, L60H90 presents a lower potassium content and electric energy consumption than L90H90 and L60H120. However, L60H120 has a greater silica content; meanwhile, the equivalent alkalis content of RHA in L60H120 was 1.33% by transforming the potassium to corresponding oxides, which was lower than 2.5% of fly ash and 1.5% of silica ash per GB/T 50733-2011 (16). To ensure the high pozzolanic activity of RHA, L60H120 with a superior silica content was selected in this paper.

The pozzolanic activity of RHA produced by L60H120 was investigated according to GB/T 27690 (27). The flexural and compressive strength of mortar containing RHA with four varying particle sizes at 7 and 28 days are plotted in Figure 7 and Figure 8, respectively. Also, the pozzolanic activity index of RHA is presented in Figure 8.

As can be seen, as RHA particle size decreases, substituting cement by RHA noticeably increases the compressive strength of mortar, while the flexural strength of mortar was slightly improved. Furthermore, the pozzolanic activity index of RHA increases as its particle size decreases since finer RHA opposes higher specific surface area, which can provide more nucleation sites for cement hydration, thus facilitating the pozzolanic reaction of RHA to form more C-S-H having a low Ca/Si and fill the pores of mortar (37, 38).

Figure 8 illustrates that the pozzolanic activity index of RHA milled from 5 min to 30 min are 85%, 95%, 99%, and 101% at 7 days, respectively. The curing days have a significant influence on the hydration degree of cement, which found that the pozzolanic activity indexes of RHA are up to 92%, 98%, 107%, and 114% at 28 days, respectively. It can be obtained that all pozzolanic activity indexes of RHA at 7 days are lower than 105%, which does not conform to the prescriptive value of silica fume in accordance with GB/T 27690 (27). However, the activity index of RHA at 28 days meets the requirement that the activity index of silica fume should be greater than 85% complying with GB/T 51003 (39). Therefore, it can be inferred that the particle size of RHA should reasonably vary from 15 min to 30 min to meet its requirement of pozzolanic activity, which contributes the particle packing effect and pozzolanic effect to enhance the mortar strength.

As listed in Table 5, the flexural and compressive strength of mortar at 7 and 28 days are presumed as parent sequences by X 01 n 、 X 02 n 、 X 03 n   a n d   X 04 n , respectively. Moreover, the particle size system of RHA is cut apart into six subsequences including <5 μmˌ5-10 μmˌ10-15 μmˌ15-20 μmˌ20-45 μm and >45 μm, which were defined by X 1 n 、 X 2 n   , … , X 6 n , respectively (as listed in Table 6).

The correlation degree and polarity between the subsequences and parent sequences were analyzed by grey correlation analysis as summarized in Table 7. It can be observed that, overall, during the six subsequences, the correlation degree of four subsequences emerges as a consistent developing rule that contains the same arranging order. Meanwhile, the particle size of 15 μm is a critical factor in determining the correlation polarity in different subsequences. From Table 7, in terms of correlation polarity, the particle size below 15 μm in RHA can have a positive influence on promoting the strength development of mortar. According to the correlation degree of six subsequences, the particle size of 5-10 μm in RHA contributes mostly to the strength development, followed by 10-15 μm and <5 μm. Analyzed by the above value of correlation degree, a smaller particle size of < 5 μm is inconsistent with the conclusion that the finer particle of RHA contributes to the higher pozzolanic reaction. That is possible because of the increase of grinding degree, which destroys some pore structures of RHA particles, thus inducing the porous microstructure to collapse and partially break into finer ones (40).

On the other hand, using RHA with a particle size exceeding 15 μm could result in a negative correlation on the strength of mortar complying with correlation polarity analysis. The largest negative effect of particle sizes will be reflected when the particle size of RHA is greater than 45 μm, followed by 0-45 μm and 15 μm-20 μm in accordance with the correlation degree of six subsequences. Hereby, the grinding time of raw RHA should be reasonably controlled in 15-30 min to improve the particle content under 15 μm in RHA and significantly reduce the particle content above 45 μm.

The microstructure of Ref and RHA5.9 samples analyzed by scanning electron microscope (SEM) and energy-dispersive X-ray (EDX) are illustrated in Figure 9 and Figure 10. The chemical composition of hydrates products is listed in Table 8. The atom ratio of Ca/Si in hydrates products decreased from 5.88 to 1.03 when cement was replaced by RHA with a mean particle size of 5.9 μm, indicating that using RHA substitute cement contributed to the pozzolanic reaction of RHA, thus forming a new C-S-H gel that its Ca/Si is lower than 1.5 and presents a fibrous gel of fine C-S-H (Ⅰ) (41). Furthermore, from Figure 10. It could be observed that the peak intensity of portlandite (2θ=near 18°) in Ref was higher than that of portlandite in the RHA5.9, indicating that RHA with a mean particle size of 5.9 μm presented a superior pozzolanic activity. Additionally, incorporating RHA weakened the peak intensity of ettringite (2θ=near 9°), which also indicates that RHA accurate the hydration degree of cement.