1. Introduction
A geopolymer (GP) is a type of alkali-activated material (AAM) crafted from fine powders rich in silicon and aluminum. These powders are dissolved in alkali solutions, forming a sodium alumino-silicate hydrate (N-A-S-H) gel [
1,
2], which gives GPs their unique properties. The structure of a GP is akin to zeolite, existing in amorphous or semi-crystalline forms [
3,
4]. When calcium-rich materials like ground granulated blast furnace slag (GGBFS), CaO, or Ca(OH)
2 are incorporated, a calcium (alumino-)silicate hydrate (C-S-H) gel forms, enhancing the mechanical properties and accelerating hardening [
5,
6,
7,
8]. This has led to some debate in academic circles, with certain studies suggesting that high-calcium-containing raw materials produce an AAM rather than a traditional GP, which is an alkali-activated material containing silicon and aluminum without calcium [
9,
10].
For practical applications, GP is often mixed with calcium sources from industrial by-products such as GGBFS, class-C fly ash (FA), waste ceramic powder, and oyster shell powder. The purpose is to reuse these by-products or waste from construction, industry, agriculture, and breeding as building materials to avoid an excessive accumulation and impact on the environment. This blending blurs the lines between GPs and AAMs [
11,
12,
13]. Compared to ordinary Portland cement (OPC), GP variants like GGBFS and GGBFS/FA-based GPs not only demonstrate higher compressive strength but also significantly reduce carbon emissions—between 50 to 90% [
14,
15]. It is worth noting that the evaluation conditions and carbon emissions factors varied in each case and country [
16]. Consequently, a GP is regarded as a low-carbon-emission material with the potential to replace OPC [
17].
While GPs exhibit a higher compressive strength compared to OPC, their practical application is limited due to excessive shrinkage issues [
18,
19,
20,
21,
22]. This shrinkage is primarily attributed to the consumption of nano-pore water in the GP [
23]. In particular, a GGBFS-based GP activated by sodium silicate and sodium hydroxide can experience shrinkage up to three and six times greater than that of OPC, respectably. Although the activation of a GGBFS-based GP using sodium carbonate can have a similar shrinkage to that of OPC, the compressive strength will be too low [
24]. Alternatives like FA and MK-based GPs offer improved volume stability compared to GGBFS-based GPs but tend to have lower strength [
25]. Adding silica fume (SF) to FA-based GPs can enhance strength, yet it may also increase shrinkage [
26,
27,
28]. Furthermore, FA’s composition can vary significantly depending on the coal source and power station types [
29], leading to varying properties in GPs. Consequently, a blend of FA, GGBFS, MK, and SF has been researched extensively to strike a balance between strength and shrinkage in GP formulations [
20,
21,
22,
23,
24,
25,
26,
27,
28,
29,
30,
31].
MK as a raw material for GPs demonstrates superior volume stability compared to GGBFS-based GPs [
32]. MK, similar in properties to FA and GGBFS, possesses cementitious properties and is commonly used in OPC concrete [
33,
34]. It is derived from calcining kaolin clay at temperatures between 600 to 850 °C for durations ranging from 1 to 12 h, depending on the kaolin’s chemical composition [
35,
36]. This calcination process eradicates the hydroxyl groups in the kaolin, transforming it into a more reactive, amorphous MK, ideal for effective geopolymerization [
37,
38,
39]. Additionally, an MK-based GP is particularly suitable for repair applications, given its mechanical properties closely resembling those of OPC and its superior tensile strength [
40,
41,
42].
Acknowledging MK’s more consistent source and composition compared to FA, this study focuses on substituting GGBFS with varying proportions of MK in a GP. It also involves adjusting the concentration of NaOH in the alkaline activator to examine the resulting changes in compressive strength and shrinkage. Given the multitude of factors influencing GPs, a comprehensive statistical analysis is utilized, including regression models and analysis of variance (ANOVA). These analytical methods are employed to assess the impacts of different parameters on GPs and to identify the optimal combinations [
43,
44,
45,
46]. The research methodology incorporates a two-variable, three-level experimental design, with the findings subjected to ANOVA and regression modeling.
2. Material
This study aimed to enhance the shrinkage properties of GPs by incorporating MK into a GGBFS-based GP formulation. To achieve this, a blend of GGBFS and MK was activated using a sodium-based alkaline activator. This activator comprised sodium hydroxide (NaOH), sodium silicate (Na2SiO3), and sodium aluminate (NaAlO2). In order to focus solely on the effects of the binder and activator, no aggregates were added to the GP mix, ensuring a more controlled study of the material’s intrinsic properties.
2.1. Binder
The GGBFS powder used in this research was sourced from CHC Resources Corp. (Kaohsiung City, Taiwan), Taiwan. It had a mean particle size (D
50) of 12.33 μm, a specific surface area of 4000 cm
2/g, and a specific gravity of 2.9. The MK, chosen for its fine particle size and chemical composition, was BURGESS No. 30 (Burgess Pigment Company, Sandersville, GA, USA). It featured an average particle size of 1.4 μm and a specific gravity of 2.63. The chemical compositions of both GGBFS and MK were meticulously analyzed using X-ray fluorescence (XRF), with the results detailed in
Table 1.
2.2. Alkaline Activator
For the alkaline activation process, the study employed a sodium-based activator. This activator was prepared by blending tap water with NaOH, Na2SiO3, and NaAlO2. The concentration of the NaOH was varied among 6 M, 10 M, and 14 M, determined by the ratio of NaOH to tap water. Furthermore, the molar ratios of SiO2/Na2O and Al2O3/SiO2 in the activator were maintained at 1.28 and 0.02, respectively, calculated based on their total content in the mixture. These specific concentrations and ratios were critical for understanding the activation process and the resulting properties of the GP.
2.3. Ordinary Portland Cement
This study used Portland Type I cement (Taiwan Cement Corp., Taipei City, Taiwan) as a benchmark for evaluating GP specimens. The length change ratio of the OPC specimen at 28 days was established as a standard for assessing the adequacy of the GP specimens. Concurrently, the criterion for compressive strength at 28 days was set at a minimum of 28 MPa, in accordance with the specifications for rapid-hardening cementitious materials outlined in ASTM C928 [
47].
3. Experiment
The test method, parameters, and preparation of the specimen are described as follows.
3.1. Test Method
This research adhered to the ASTM C109/C109M-20 [
48] standard for compressive strength testing (specimen size is 50 mm × 50 mm × 50 mm), which necessitates the slurry’s fluidity to be within 110 ± 5%. Consequently, adjustments in the activator-to-binder ratio (A/B) were required for all specimens, ensuring compliance with the specified fluidity criteria, as verified by fluidity tests in accordance with ASTM C230/C230M-20 [
49]. The length change ratio measurements and calculations were conducted following the ASTM C157-75 [
50] standard (specimen size is 25 mm × 25 mm × 285 mm). Importantly, the preparation of the specimens for length change measurements maintained the same A/B ratio as those used for compressive strength testing. After a 24-h setting period, all specimens were demolded. The initial length of the length change rate specimens was recorded immediately post-demolding, then these specimens were placed in a controlled curing chamber. Length measurements for these specimens were taken at various ages, while the compressive strength tests were conducted at 28 days.
3.2. Experimental Design Parameter
The experimental design incorporated two variables across three levels to assess the impact of MK substitution and NaOH concentration. This resulted in nine distinct parameter sets, as detailed in
Table 2. For each parameter, a minimum of three specimens were prepared to conduct both compressive strength and shrinkage tests. The A/B ratio for each parameter was determined based on achieving the targeted fluidity range of 110 ± 5%, ascertained through fluidity tests. Additionally, an OPC specimen was included for comparative analysis.
3.3. Preparation of Specimen
The preparation process involved mixing MK, GGBFS, and the activator for 2 to 5 min to achieve a homogeneous slurry. This slurry was then poured into molds and sealed, followed by a resting period in a room-temperature environment for 24 h. After demolding, the specimens were transferred to the curing chamber with the specified conditions of 52 ± 3% humidity and 23.2 ± 2 °C temperature.
3.4. ANOVA
In this study, F-tests were employed to analyze the data, focusing on comparing variances across multiple groups. This approach was chosen to ascertain the significance of the NaOH concentration and MK content in alkaline activators as critical influencing factors. The essence of the F-test involves contrasting the ratio of variation between groups against the variation within groups. This comparison aims to determine if the variation observed between groups significantly exceeds what might be expected from random fluctuations. A significant result is indicated either when the F-value surpasses the critical value, or when the p-value falls below the established significance level (α). In such cases, the null hypothesis, which posits no difference in population variation among the groups, is rejected. This suggests that the variation in at least one group is distinct from others. When significant differences between groups are identified, additional post-hoc analyses are conducted. These subsequent comparisons are crucial for pinpointing specific differences among the groups.
3.5. Second-Order Response Surface Methodology (RSM)
The second-order response surface model was used to create a predictive mathematical model. This model aimed to draw the curvature response surface of the compressive strength and length change ratio at 28 days based on the NaOH concentration and MK content to identify the qualification range. The model is established using the following formula:
where
is a variable of MK content, which is 0, 50 (wt.%), and 100 (wt.%) in this study;
is a variable of NaOH concentration of the alkaline activator, which is 6 (M), 10 (M), and 14 (M) in this study; and
,
are coefficients.
The equation can be expressed in matrix form as:
where,
In the above equation, n is the number of the experimental parameters and the number of observations obtained for the corresponding experimental parameters; in this study n is 9. Note that n needs to be at least 6 or above to solve for 6 coefficients.
In this representation,
is the matrix of observed responses,
is the matrix containing the values of the predictors for each observation, and
is the matrix of coefficients to be estimated. To find the coefficients, the least squares method is used, and the solution is given by:
This method aims to minimize the sum of squared errors between the observed responses and the predicted values of the model.