Experimental Investigation on the Durability of Metakaolin-Based Geopolymer Concrete in Aggressive Environments

doi:10.21203/rs.3.rs-2247685/v1

The properties and performance of geopolymer concrete have been evaluated in numerous studies in recent decades. One of the fields of study is the evaluation of the durability of geopolymer concrete in aggressive environments. In this study, the influence of four experimental factors, including curing temperature (T), the weight ratio of sand to metakaolin (S/M), aggressive environment type (AE), and aggressive environment exposure time (Et) were investigated on the durability and mechanical properties of metakaolin-based geopolymer concrete (MGPC). Moreover, 5% sulfuric acid solution, 5% sodium chloride solution, and Persian Gulf seawater were used as aggressive environments. Response surface methodology (RSM) was implemented to statistically design and analyze the experiments. Mathematical models were derived for the prediction of residual compressive strength (RF_c), weight changes (∆W), and compressive strength changes (∆F_c) of MGPC. The analysis of variance showed that all four factors had a significant effect on the responses, while only AE-S/M and AE-ET factors had an interaction. Design-Expert software (DX7) was employed for the optimization of design factors. Results revealed that the use of the optimum weight ratios of S/M (2.52, 2.07, and 1.29) at a curing temperature of 46 °C and an exposure time of 56 days led to RF_cvalues of 34.12, 43.60, and 44.15 for sulfuric acid, sodium chloride, and Persian Gulf seawater aggressive environments, respectively. These results, which are higher than the minimum required compressive strength of concrete, along with the ∆F_cand ∆W values, confirm the acceptable durability of MGPC.

Geopolymer concrete

Durability

Residual compressive strength

Aggressive environments

Response surface methodology

Cement manufacturing produces environmental pollution and suffers from deficiencies in required resources, making an alkali-activated binder a promising alternative to cement. Geopolymers, named by Davidovits, are usually obtained by using an alkali solution to activate aluminosilicate powder (such as metakaolin or fly ash) [1]. The reaction mechanisms include three steps:(1) the aluminosilicate oxide of solid materials is dissolved in concentrated alkaline solutions to form free SiO₄ and AlO₄ tetrahedral units, followed by (2) the condensation reaction of the alumina/silica–hydroxyl species to form an inorganic polymer gel phase and finally, (3) the gel phase gradually hardens to form geopolymers [2]. Myers reported that the geopolymers microstructure varied with the type of the used precursor [3]. Based on polymerization reaction products, alkali-activated materials (AAMs) can be divided into two primary groups. AAMs polymerization reaction products (gels) are as follows: (a) CaO–Al₂O₃–SiO₂–H₂O (C–A–S–H gel) and (b) Na₂O–Al₂O₃–SiO₂–H₂O (N–A–S–H gel). Alkali-activated slag (AAS) belongs to the first group, while alkali-activated metakaolin (AAMK) and fly ash (AAFA) belong to the second group.

Although past studies show that geopolymer concrete (GPC) has excellent mechanical properties, such as high early strength, low creep, and low shrinkage, compared to ordinary Portland concrete (OPC) [4–6], the durability performance is a key concern for the application of GPC in civil infrastructure [7–9]. Okoyo defines concrete durability as the capability of concrete to resist weathering action, chemical attack, and abrasion [10]. Chemical attack is one of the most concerning causes of performance decrease, and it generally happens in environments with the presence of sulfate, chloride, and acids [7, 11]. External sulfate attack results from the impact of sulfate ions present in soils, underground waters, seawater, or industrial wastewaters on hardened concrete [12]. Sulfate attack on OPC concrete results from the chemical reaction of sulfate ions as the aggressive substance and the aluminate component of hardened cement paste. If enough water is present, the reaction between these substances produces ettringite and gypsum and causes the expansion of the OPC concrete, leading to cracking. At the same time as the sulfate attack, the attack of magnesium ions and sodium ions on C-S-H starts when CH is depleted. This attack leads to gypsum precipitation and decalcification of C-S-H. The decalcification of C-S-H leads to a loss of adhesion and strength in concrete [13].

In the marine environment, a high concentration of chloride ions can also result in corrosion and degradation of steel bars embedded in reinforced concrete. According to previous works in the chloride environment, GPC shows better performance over OPC concrete [8, 10]. Better performance of GPC over OPC is attributed to the lack of the C-S-H phase in the microstructure of GPC [14].

Moreover, good resistance of GPC compared to OPC was observed in the presence of acids such as sulfuric acid and acetic acid [11]. Acid attack on OPC leads to a loss of strength and some damage due to its chemical composition, which is high in calcium compounds. High acid resistance was reported for a geopolymer as an alternative to OPC [1].

It is reported that the low degradation of GPC in all aggressive environments may be a result of low permeability and low calcium content as compared to conventional OPC [14].

Palomo subjected a metakaolin-based geopolymer to sulfuric acid and found little variation in flexural strength after the attack, which can be the result of low calcium content of metakaolin [15].

Rovnanik reported a relationship between curing temperature and porosity properties of GPCs, and curing at lower temperatures could reduce porosity [16]. However, it is still unknown whether or not temperature affects the durability of the geopolymer.

It is reported that GPC shows the best compressive strength in a certain ratio of sand to precursor, [17] but no investigation is accessible about the effect of the sand to precursor ratio on the durability of GPC.

This study aims to investigate the effects of curing temperature and the sand to metakaolin ratio on durability properties.

In this work, metekaoline was the precursor of geopolymer concrete. The X-ray fluorescence (XRF) test was carried out to determine the chemical composition of kaolin, as presented in Table 1. After making sure about the appropriate chemical composition of kaoline for use in the study, the unprocessed kaolin was ground with an industrial mill and then passed through a sieve with No. 325 (45µm) openings to obtain kaolin particles with a size of less than 45 microns. The micronized kaolin was then calcined in a furnace with a temperature gradient of 10°C per minute at 750°C for 5 h and then returned to ambient temperature with the same temperature gradient. In this work, the used alkaline activator is a composition of Na₂SiO₃ and NaOH. The characteristics of liquid Na₂SiO₃ and flakes of NaOH are represented in Table 2.

Table 1

The chemical composition of metakaolin based on the XRF analysis
Constituents	Chemical term	Weight percentage
Silicon Dioxide	(SiO₂)	45.17%
Aluminum Oxide	(Al₂O₃)	37.12%
Ferric Oxide	(Fe₂O₃)	0.79%
Titanium dioxide	(TiO₂)	0.31%
Calcium Oxide	(CaO)	0.28%
Magnesium oxide	(MgO)	0.46%
Potassium Oxide	(K₂O)	0.89%
Sodium Oxide	(Na₂O)	0.13%
Loss in ignition	-	14.15%

Table 2

The chemical composition of activators
Na₂SiO₃ )L)		NaOH
Chemical composition	Constituents	Chemical composition	Constituents
28.4%	SiO₂	98%	NaOH
8.8%	Na₂O	1%	Na₂CO₃
62.8%	water	200ppm	NaCl
3.3%	Modulus ratio (SiO₂/Na₂O)	6 ppm	Fe
3.3%	Modulus ratio (SiO₂/Na₂O)	15.6%	SiO₂

Since the present study investigates a metakaolin-based geopolymer mortar, only one type of fine grain, silica sand, was used here. In this study, control samples were made in the form of Portland cement mortar. The Portland cement chemical composition is presented in Table 3.

Table 3

The chemical composition of OPC control samples
Constituents	Chemical composition
Silicon Dioxide	19.01%
Calcium Oxide	66.89%
Magnesium Oxide	0.81%
Phosphate(P₂O₅)	0.08%
Sodium Oxide	0.09%
Potassium Oxide	1.17%
Manganese Oxide	0.19%
Aluminum Oxide	4.68%
Ferric Oxide	3.20%
Sulfur trioxide	3.0%
Loss on ignition	2.48%

Chloride and acidic environments and seawater media were used to evaluate the durability of MGPC and OPC samples. For this purpose, H₂SO₄ and NaCl solutions with a concentration of 5% by weight were used to simulate acidic and chloride media, respectively. A Persian Gulf water sample was also used to simulate the seawater environment. The chemical composition of the Persian Gulf Seawater used in this research is given in Table 4.

Table 4

The chemical composition of Persian Gulf water
Ca²⁺	675	*Si⁴⁺	0
K⁺	555	Cl⁻	21100
Mg²⁺	1340	Br⁻	73
Na⁺	8819	S²⁻	765
Ni²⁺	0.320	SO₄²⁻	2295
*Li⁺	0.310	CO₃²⁻	156
*Fe²⁺	0.010	pH	8.05
*Zn²⁺	0.013	Density	1.027 g/ml
*Mn²⁺	0.037	Density	1.027 g/ml

In this study, the number of tests and save costs were optimized using Design-Expert software. Among the different experimental design methods available in this software, the response surface method (RSM) was used to design the experiment. As shown in Table 5, the variables used in the RSM are classified into two categories, numerical and categorical. Numerical variables include the curing temperature and the ratio of sand to metakaolin. Categorical variables include the type of aggressive environment and the time of exposure to the aggressive environment. AE has three levels, Cen, Aen, and PGW. For Et, other categorical variables, 7, 28, and 56 days, are defined as three investigation levels.

Table 5

The Specifications of independent variables defined for DX software
Numerical variables				Categorical variables
Factor	Variable name	Minimum	Maximum	Factor	Variable name	Number of levels	Level 1	Level 2	Level 3
A	T (°C)	40.00	80.00	C	AE	3	Cen	PGW	Aen
B	S/M	1.00	3.00	D	Et(Day)	3	7	28	56

After defining the values for the independent variables, DX software presented its own suggested values for carrying out experiments, as shown in Table 6. Based on the software method used, each numerical variable is suggested in five levels. There were 39 suggested experiments in each aggressive medium. Considering three aggressive environments, 117 experiments were entirely suggested by DX software.

Table 6

Experimental designs suggested by DX software for all three levels of the AE variable
Variable Run	T	S/M	Et	Variable Run	T	S/M	Et	Variable Run	T	S/M	Et
1	74.14	2.70	7	14	74.14	2.70	28	27	74.14	2.70	56
2	60	3	7	15	60	3	28	28	60	3	56
3	74.14	1.29	7	16	74.14	1.29	28	29	74.14	1.29	56
4	80	2	7	17	80	2	28	30	80	2	56
5	45.85	2.70	7	18	45.85	2.70	28	31	45.85	2.70	56
6	60	2	7	19	60	2	28	32	60	2	56
7	60	2	7	20	60	2	28	33	60	2	56
8	60	2	7	21	60	2	28	34	60	2	56
9	60	2	7	22	60	2	28	35	60	2	56
10	60	2	7	23	60	2	28	36	60	2	56
11	60	1	7	24	60	1	28	37	60	1	56
12	45.85	1.29	7	25	45.85	1.29	28	38	45.85	1.29	56
13	40	2	7	26	40	2	28	39	40	2	56

To evaluate the durability of samples, Residual compressive strength (RF_c), weight changes (∆W), and changes in compressive strength (∆F_c) were selected as response variables. ∆F_c and ∆W are obtained according to Equations 1 and 2, respectively.

\(\Delta Fc=\frac{{RFc - IFc}}{{IFc}} \times 100\) Eq. (1)

In Eq. 1, (RF_c) is residual compressive strength and (IF_c) is the initial compressive strength, the value of which is obtained by performing a 28-day compressive strength test on the designs presented in Table 7.

\(\Delta W=\frac{{RW - IW}}{{IW}} \times 100\) Eq. (2)

In Eq. 2, (RW), the residual weight is the result of measuring the weight of the samples after exposure to aggressive environments, and (IW) is the initial weight obtained from measuring the weight of 28-day cured samples according to the presented design in Table 7.

Table 7

Experimental designs for 28-day compressive strength and weight
Number	T	S/M
1	74.14	2.70
2	60	3
3	74.14	1.29
4	80	2
5	45.85	2.70
6	60	2
7	60	1
8	45.85	1.29
9	40	2

The designs created by the DX software included five different S/M values. Variation in the values of S/M created different mix designs. All the mix designs, including the control sample mix design, are presented in Table 8. The mix design in this research is based on the investigation of Alanazi et al [18]. Accordingly, the SiO₂/Na₂O ratio, which is the molar ratio of alkaline solution, is equal to 1 in this study.

Table 8

Mix Designs
Cement	Metakaoline	Quartz Sand	S/M	Silicate sodium	Sodium hydroxide	Alkaline activator	Added water	Modulus ratio of alkaline activator
	600	600	1	496	130	626	266	1
	562	725	1.29	462	122	584	270	1
	480	960	2	397	104	501	275	1
	418	1130	2.71	346	91	437	284	1
	400	1200	3	330	87.5	417	300	1
520		1560					600

To make geopolymer samples, metakaolin and quartz sand were first added into a mixer, and then mixed for 3 min. The pre-prepared alkaline solution was added to the mixer and mixed for 5 min in the next step. Water was added to the concrete to increase its workability. After finishing the work with the concrete mixer, the geopolymer was poured into 5*5*5 mm (125 mm³) molds. Following molding, MGPC samples were placed in an oven at different temperatures according to the experimental designs. Then, samples were cured at the defined temperatures in the oven for 24 h. Next, all samples were taken out of the oven and stored at 25°C for 28 days. When all the samples were cured for 28 days, a cement control sample and nine MGPC samples were subjected to compressive strength and weighing tests. The mentioned samples were made to obtain the initial compressive strength and initial weight, as represented in Table 7. Afterward, the rest of the MGPC samples and control samples were placed in aggressive environments, and samples were evaluated by measuring RF_c, ∆F_c, and ∆W after 7, 28, and 56 days of exposure to aggressive environments. Then, the values of RF_c, ∆F_c, and ∆W were entered into the software, which created a model using the entered data.

3.1 Model Generation by DX software

After entering the results into the software, the software suggested models based on the results of completed experimental designs. By default, Design-Expert software automatically shifts to the “Suggested” polynomial model that best fits the criteria presented in the Fit Summary section. Design-Expert calculates the Whitcomb Score (a heuristic scoring system) to choose a default model. In this study, the model suggested by the software for RF_c, ∆F_c, and ∆W is a quadratic model, as presented in Tables 9-11.

Table 9 Suggested models for the prediction of RF_c

Model	Sum of squares	df	Mean squares	R²	F-value	P-value
Linear	550.05	74	7.43	0.79	123.43	< 0.0001
2FI	406.43	61	6.66	0.84	110.64	< 0.0001
Quadratic	173.16	59	2.93	0.93	48.73	< 0.0001
Cubic	91.96	37	2.49	0.96	41.27	< 0.0001

Table 10 Suggested models for the prediction of ∆F_c

Model	Sum of squares	df	Mean squares	R²	F-value	P-value
Linear	1987.71	74	26.86	0.80	79.90	< 0.0001
2FI	1279.00	61	20.97	0.87	62.37	< 0.0001
Quadratic	680.68	59	11.54	0.93	34.32	< 0.0001
Cubic	218.15	37	5.90	0.97	17.54	< 0.0001

Table 11 Suggested models for the prediction of ∆W

Model	Sum of squares	df	Mean squares	R²	F-value	P-value
Linear	187.86	74	2.54	0.83	22.40	< 0.0001
2FI	56.94	61	0.93	0.94	8.24	< 0.0001
Quadratic	49.65	59	0.84	0.95	7.43	< 0.0001
Cubic	36.34	37	0.98	0.96	8.67	< 0.0001

RF_c, ∆F_c, and ∆W are predicted by a second-order model in the form of a quadratic polynomial equation:

Equation (3)

where Y is the predicted response variable, β₀, β_i, β_ii, and β_ij are constant regression coefﬁcients of the model, and x_i, x_j (i, j= 1, 2, 3, 4; i≠j) represent the coded values of independent variables used in statistical calculations. The relationship between the coded and actual values is as the following equation:

Equation (4)

where X_i is the actual value of the variable, X₀ is the actual value of X_i at the center point, and ∆X is the step change of the variable.

Quadratic models for all three response variables, RF_c, ∆F_c, and ∆W, are presented in Tables 12-14. For each of the response variables (RF_c, ∆F_c, and ∆W), nine different models are presented according to the values of AE and Et.

Table 12 The prediction of RF_c based on the suggested model

AE	Et	C	α	γ	δ	ε	ζ
Cen	7	35.91	0.13	10.73	0.032	-0.003	-3.79
	28	35.42	0.09	11.44	0.032	-0.003	-3.79
	56	35.52	0.08	11.43	0.032	-0.003	-3.79
PGW	7	38.36	0.14	8.74	0.032	-0.003	-3.79
	28	38.12	0.10	9.46	0.032	-0.003	-3.79
	56	38.31	0.09	9.45	0.032	-0.003	-3.79
Aen	7	26.94	0.13	12.34	0.032	-0.003	-3.79
	28	23.59	0.09	13.05	0.032	-0.003	-3.79
	56	23.75	0.09	13.04	0.032	-0.003	-3.79

Table 13 The prediction of ∆F_c based on the suggested model

AE	Et	C	α	γ	δ	ε	ζ
Cen	7	63.87	-1.28	-24.99	0.031	0.01	5.13
	28	62.98	-1.16	-26.23	0.031	0.01	5.13
	56	61.96	-1.14	-26.06	0.031	0.01	5.13
PGW	7	56.06	-1.28	-19.87	0.031	0.01	5.13
	28	54.66	-1.17	-21.11	0.031	0.01	5.13
	56	53.42	-1.14	-20.94	0.031	0.01	5.13
Aen	7	77.24	-1.21	-27.23	0.031	0.01	5.13
	28	83.12	-1.10	-28.47	0.031	0.01	5.13
	56	81.94	-1.07	-28.30	0.031	0.01	5.13

Table 14 The prediction of ∆W based on the suggested model

AE	Et	C	α	γ	δ	ε	ζ
Cen	7	-7.28	0.27	1.95	-0.02	-0.001	-0.36
	28	-4.04	0.25	1.66	-0.02	-0.001	-0.36
	56	-3.38	0.25	1.98	-0.02	-0.001	-0.36
PGW	7	-12.03	0.28	4.21	-0.02	-0.001	-0.36
	28	-7.69	0.26	3.92	-0.02	-0.001	-0.36
	56	-5.42	0.26	4.24	-0.02	-0.001	-0.36
Aen	7	-0.85	0.24	0.97	-0.02	-0.001	-0.36
	28	5.65	0.22	0.68	-0.02	-0.001	-0.36
	56	6.19	0.22	1.00	-0.02	-0.001	-0.36

In Tables 12-14, the terms A and B represent the first (T) and the second (S/M) variables, respectively.

Response variable = C+ α.A + γ.B + δ.AB + ε.A² + ζ.B²

According to Figure 2, the low distribution of the tested points compared to the predicted line can be observed for the model. One of the main parts of studies of this type is to test the validity of the models presented by the software. For this purpose, an optimal sample with T = 45.86 and S/M = 2.28 was made to be placed in Aen for 7 days. After obtaining the results of the sample made in the laboratory, a very good agreement was observed between the result of the proposed model by the software and the tested sample. According to the model, the results were RF_c= 38.74 MPa, ∆W = 5.04%, and ∆F_c = 10.91%, and the results of the tested sample were RF_c = 40.01 Mpa, ∆W = 5.24%, and ∆F_c = 11.45%. Therefore, the differences in the results of the tested sample from the modeled sample were 3.17%, 3.96%, and 4.94%, respectively.

After confirming the validity of the models provided by the software, it is possible to evaluate the effect of independent variables on response variables. In this study, the graphs are presented in three categories based on the aggressive environment. Three graphs are presented with different Ets for each response variable, and the effect of two numerical variables is illustrated on that response variable. For convenience in comparing modeled samples with different (S/M) and different (T) samples, it is necessary to introduce the abbreviations of the modeled samples, which are given in Table 15.

Table 15 Abbreviations for all model-based samples

T (°C) S/M	40	60	80
1	MGPC1-40	MGPC1-60	MGPC1-80
2	MGPC2-40	MGPC2-60	MGPC2-80
3	MGPC3-40	MGPC3-60	MGPC3-80

3.2 Residual compressive strength of control samples

As mentioned before, it is essential to make control samples (OPC) to have criteria for evaluating the performance of MGPC samples. The RF_c values of OPC in intervals of 0, 7, 28, and 56 are presented in Figure 3 to represent the criteria for MGPC and OPC comparisons. The compressive strength in time 0 is the 28-day compressive strength of mortar sample that hasn’t been exposed to aggressive environment yet.

According to the output model of DX software, all graphs are categorized based on the aggressive environment in which samples are placed.

3.3 Model-based MGPC samples exposed to the chloride environment

Figure 4 shows the effect of numerical variables on the RF_c value of MGPC samples in the chloride environment.

Figure 4 Values of RF_c for MGPC samples exposed to the chloride environment at Et = 7(a), 28 (b), and 56 (c) based on the proposed model

According to Figure 4, the best RF_c for all three values of Et belongs to the MGPC2-40 sample. The RF_c value for the MGPC2-40 sample reached 44.4 MPa after 56 days of exposure to Cen. RF_c of OPC in Et = 56 is equal to 47Mpa, and the superiority of OPC over MGPC is valid for all Et’s. Following the RF_c values, the values of ∆F_c and ∆W for model-based MGPC samples and actual control samples are represented in Figures 5 and 6.

The lowest decrease in compressive strength after 56 days of exposure to Cen is related to MGPC2-60 sample. Figure 5 shows MGPC2-40, MGPC2-60, and MGPC3-60 samples with less compressive strength reduction than OPC samples, which is related to the low permeability of these samples. At Et =7, the value of ∆F_c is -1/91 and this indicates an increase in compressive strength. According to a study by Albitar, all types of concrete may experience an initial increase in compressive strength when exposed to an aggressive environment [7]. Albitar attributes this initial increase to the hydration process of calcium silicate and other pozzolanic reactions. Following these reactions and the pressure caused by the expansion of the elements, internal blockage occurs in the concrete structure, which has a positive effect on the increase of compressive strength.

Among all MGPC’s, in 56 days, the lowest ∆W value is related to MGPC3-40. In terms of weight change, MGPC shows more weight reduction and thus poorer performance than OPC concrete after 28 and 56 days exposure to Cen.

3.4 Model-based MGPC samples exposed to the acidic environment

According to Figure 7, the highest RF_c is related to MGPC2-40 for all Et values. MGPCs at S/M = 2 show the highest value. The RF_c value for the MGPC2-40 sample decreased to 36.8 MPa after 56 days of exposure to Aen, while RF_c decreased to 39.17 MPa for OPC. In Aen, the superiority of OPC over MGPC about the RF_c value is valid for all Et values.

In Figure 8, after 56 days of exposure, the lowest decrease in compressive strength among MGPC’s is related to the MGPC3-40 sample. In these graphs, all MGPC samples show a greater reduction in compressive strength than OPC samples. Among MGPC samples, MGPC3-40 shows a very close compressive strength reduction to OPC samples.

The lowest ∆W among all MGPCs is related to MGPC3-80 after 56 days of exposure. Figure9 (a) shows the better performance of MGPC concretes featuring S/M = 3. By comparing these graphs with those related to ∆F_c, it can be concluded that the better performance of MGPC in ∆W graphs can be due to weight gain resulting from the entry of the solution material into the concrete, although the result of weight change shows weight loss. In fact, the more compressive strength reduction of MGPC than OPC at Aen and Et =7 indicates that the low weight loss of MGPC specimens with S/M = 3 compared to OPC occurred only due to weight gain.

3.5 Model-based MGPC samples exposed to PGW

As shown in Figure 10, the maximum RF_c value belongs to MGPC1-40. The value of RF_c is maximum at S/M = 1 or 2 depending on Et and T. the RF_c value for the MGPC1-40 sample decreased to 44.26 MPa after 56 days of exposure to PGW while the RF_c for OPC is 45.17 Mpa after 56 days of exposure. Regarding RF_c, the superiority of OPC over MGPC in PGW is valid for all Et values.

After 56 days, the lowest ∆F_c in PGW is related to the MGPC2-60 sample. In Figure11, the MGPC2-60 and MGPC2-40 samples show a lower reduction in compressive strength than that of OPC samples.

The value of ∆W = -0.33% for MGPC1-40 indicates an increase in sample weight after 7 days, and this finding is in accordance with Albitar's research in which a slight increase occurred in weight followed by a decrease [7]. This initial weight gain may be due to the entry of chemical particles existing in the aggressive solution into the MGPC pores.

After observing the graphs, it can be generally argued that with an increase in T a decrease happens in RF_c in all AEs and for all values of Et. An increase in T affects both the initial compressive strength (28 days) and the change in compressive strength. According to studies by Rovnanik, large porosities are formed in MGPC samples and the volume of cumulative porosity is increased by increasing the T value above 60 °C, which has a negative effect on the mechanical properties of MGPC at the age of 28 days [16]. Okoye concluded that this increase in porosity affects the permeability of the specimens, and an increase in porosity is causes an increase in permeability [14]. In this study, increasing the temperature reduced the initial compressive strength(28 days) and increased the porosity. thus, the increased porosity of concrete increased its permeability in exposure to aggressive environment, and the increase in the permeability of concrete causes more decreases its compressive strength in exposure to aggressive environments. According to Albitar's research, the reduction of GPC resistance can be due to the leakage of alkaline-activating substances, such as NaOH, in the corrosive solution [7].

3.6 Optimization

The optimization module of DX software searches for a combination of factor levels that simultaneously satisfy the criteria defined for each response and factor. To include a response in the optimization criteria, it must have a model fit through analysis or be supplied via an equation-only simulation. Moreover, factors are automatically included “in range. Numerical optimization is a hill-climbing technique. In addition to the design points, a set of random points are examined to see if there is a more desirable solution. Finding an initial feasible region can be difficult. We start with a small value of a penalty function in a downhill simplex (Nelder-Mead) multi-dimensional pattern search, which converges at either a fixed point or a design space boundary. In this study, it is desirable that RF_c would be the maximum, and ∆F_c and ∆W would be the minimum. According to the suggested models and the defined desirability for each response variable, the optimal values for the independent variables in each of AE and Et are represented in Table 16.

Table 16 Suggested values for producing optimal samples and predicted values for the response variables of each sample

Categorical variable		Numerical variable		Response variable
AE	Et	T	S/M	RF_c	∆W	∆F_c
Cen	7	45.86	1.95	45.04	2.34	0.18
	28	45.86	2.11	43.70	4.08	1.51
	56	45.86	2.07	43.60	5.38	2.12
PGW	7	45.97	1.42	45.42	1.39	2.80
	28	45.86	1.39	44.35	4.47	5.09
	56	45.86	1.29	44.15	6.90	5.78
Aen	7	45.86	2.28	38.74	5.04	10.91
	28	45.86	2.49	34.40	9.65	18.68
	56	45.86	2.52	34.12	10.88	19.03

As shown in Figure13, there is a slight difference between the RF_c values of OPC control samples and model-based optimized MGPC samples, and no difference of more than 5 Mpa is observable. Since all samples conform to the criteria accepted by the world civil society, such as ACI318 and ASHTO, it is feasible to use optimized MGPC samples as an alternative to OPC samples. Following the RF_c graphs, ∆F_c graphs are presented as the criteria to evaluate the durability performance of samples. As shown in Figure 11 (c), the model-based optimized MGPCs have better performance over OPC control samples. According to Figure 11 (b), ∆W values for MGPC samples are higher than those for OPC samples, As explained before, the lower ∆W of OPC compared to MGPC samples is attributed to the more porous microstructure and, consequently, more pore solution content of OPC control samples.

This investigation on the durability of metakaoline-based geopolymer concrete led to the following results:

The optimum value of operating temperature causes increase in compressive strength and decrease in porosity and it is an important factor in the performance of MGPC samples in the corrosive environment. As the curing temperature increases, the 28-day compressive strength decreases and the porosity increases, but as the operating temperature decreases, the 28-day compressive strength increases. The decrease in compressive strength is due to the increased porosity of MGPC samples, and this porosity affects the permeability and durability of MGPC.

The ratio of sand to metakaolin is one of the factors affecting the durability of MGPC. With the best ratio of 2, increasing the sand to metakaolin ratio to more than 2 reduces the compressive strength, and decreasing this ratio to less than 2 also reduces the compressive strength of concrete. The decrease in resistance at ratios of more than 2 was due to the lack of an increase in the amount of the alkaline activator compared to the increase in the sand. In the low ratio of sand to metakaolin, low compressive strength occurs due to the porous structure, and even this porous structure has a negative effect on the durability of geopolymer concrete. This porous structure is related to the lack of a sufficient O-Si factor resulting from the addition of silica sand to form the N-A-S-H bond.

By changing the type of the aggressive environment, the results related to MGPC and OPC change compared to each other, thus the decision to choose the type of concrete is related to the type of the aggressive environment. According to the graphs presented in this study, OPC shows better performance than MGPC in an acidic environment. In the chloride medium, some MGPC samples can perform better than OPC by observing the sand to metakaolin ratio and the curing temperature. In the Persian Gulf water environment, some MGPC samples show better performance than OPC by observing the ratio of sand to metakaolin and curing temperature. After using optimization, however, MGPCs showed better performance than OPC.

In this study, some OPC samples with a calcium content of 66% had a better performance than metakaolin with a calcium content of less than 1%, which indicates C-S-H and N-A-S-H gels are both susceptible to corrosion in an aggressive environment. On the other hand, the impermeability of concrete causes good durability in aggressive environments, and it was observed that MGPCs with lower porosity and, consequently, lower permeability show better performance than OPC samples.

The porous structure is more permeable, and permeability allows aggressive agents to penetrate the concrete structure. Aggressive solvents enter the pores of geopolymer concrete, causing it to be washed from alkaline activators, and the concrete is weakened in terms of compressive strength with the leakage of alkaline activators.

An initial increase in strength and weight occurs in all types of concrete and is not specific to a particular type of concrete, but the amount of these changes depends on the type of the aggressive medium and the type of concrete.

Gao, X., et al., Behavior of metakaolin-based potassium geopolymers in acidic solutions. Journal of Non Crystalline Solids, 2013. 380: p. 95–102.
Ren, D., et al., Durability performances of wollastonite, tremolite and basalt fiber-reinforced metakaolin geopolymer composites under sulfate and chloride attack. Construction and Building Materials, 2017. 134: p. 56–66.
Myers, R.J., et al., Generalized Structural Description of Calcium–Sodium Aluminosilicate Hydrate Gels: The Cross-Linked Substituted Tobermorite Model. Langmuir, 2013. 29(17): p. 5294–5306.
Assi, L.N., et al., Investigation of early compressive strength of fly ash-based geopolymer concrete. Construction and Building Materials, 2016. 112: p. 807–815.
Pouhet, R. and M. Cyr, Formulation and performance of flash metakaolin geopolymer concretes. Construction and Building Materials, 2016. 120: p. 150–160.
Bernal, S.A., et al., Effect of binder content on the performance of alkali-activated slag concretes. Cement and Concrete Research, 2011. 41(1): p. 1–8.
Albitar, M., et al., Durability evaluation of geopolymer and conventional concretes. Construction and Building Materials, 2017. 136: p. 374–385.
Ismail, I., et al., Influence of fly ash on the water and chloride permeability of alkali-activated slag mortars and concretes. Construction and Building Materials, 2013. 48: p. 1187–1201.
Pasupathy, K., et al., Durability of low–calcium fly ash based geopolymer concrete culvert in a saline environment. Cement and Concrete Research, 2017. 100: p. 297–310.
Okoye, F.N., J. Durgaprasad, and N.B. Singh, Effect of silica fume on the mechanical properties of fly ash based-geopolymer concrete. Ceramics International, 2016. 42(2, Part B): p. 3000–3006.
Sata, V., A. Sathonsaowaphak, and P. Chindaprasirt, Resistance of lignite bottom ash geopolymer mortar to sulfate and sulfuric acid attack. Cement and Concrete Composites, 2012. 34(5): p. 700–708.
Komljenović, M., et al., External sulfate attack on alkali-activated slag. Construction and Building Materials, 2013. 49: p. 31–39.
Bakharev, T., J. Sanjayan, and Y.B. Cheng, Sulfate Attack on Alkali-Activated Slag Concrete. Cement and Concrete Research, 2002. 32: p. 211–216.
Okoye, F.N., S. Prakash, and N.B. Singh, Durability of fly ash based geopolymer concrete in the presence of silica fume. Journal of Cleaner Production, 2017. 149: p. 1062–1067.
Palomo, A., et al., Chemical stability of cementitious materials based on metakaolin. Cement and Concrete research, 1999. 29(7): p. 997–1004.
Rovnaník, P., Effect of curing temperature on the development of hard structure of metakaolin-based geopolymer. Construction and building materials, 2010. 24(7): p. 1176–1183.
Wazien, A.Z.W., et al., Strength and Density of Geopolymer Mortar Cured at Ambient Temperature for Use as Repair Material. IOP Conference Series: Materials Science and Engineering, 2016. 133: p. 012042.
Alanazi, H., et al., Early strength and durability of metakaolin-based geopolymer concrete. Magazine of Concrete Research, 2017. 69(1): p. 46–54.

No competing interests reported.

Experimental Investigation on the Durability of Metakaolin-Based Geopolymer Concrete in Aggressive Environments

Status:

Version 1

Abstract

Figures

1. Introduction

2. Materials And Methods

3. Results And Discussion