The Distribution, Structure, and Chemical Composition of Alkali-Silica Gels in Calcined Clay Concretes

Abstract

This study investigates the effect of calcined clays (metakaolin, metasilt, metaclay) on the chemical composition, distribution, and structure of alkali–silica reaction (ASR) gels. Using 10 wt% of calcined clays reduced concrete expansion and minimized cracking but did not inhibit ASR gel formation. Micro X-ray fluorescence mapping revealed an average ASR gel content of 3 wt% in concrete, incorporating up to two-thirds of K₂O and nearly all Na₂O from the binder. Raman spectroscopy indicated structural similarities among gels in different concrete mixes, with an increased degree of polymerization in the metakaolin-containing concrete. Automated mineralogy identified four gel phases: Si gel, Ca-Si gel, Al-Ca-Si gel, and Al-Si gel. Ca-Si gels are formed at binder interfaces, while non-swellable Al-bearing gels are mainly formed in metakaolin-containing concrete located within aggregates. This study shows that aluminum can be incorporated into gels in calcined clay concretes, altering their structure and potentially affecting their expansion behavior in concrete.

1. Introduction

The use of supplementary cementitious materials (SCMs) is a common way to mitigate alkali–silica reaction (ASR) in concrete [1,2,3,4,5]. ASR is a chemical reaction between the reactive silica of the aggregate and the alkali metals and hydroxides of the pore solution in concrete, leading to the formation of ASR products [6,7,8,9]. The pressure exerted by the ASR products leads to cracking, first in the aggregates and further into the binder’s matrix. More products continue to form within the cracks, leading to further expansion and eventually to the deterioration of the concrete structure [7]. The yearly worldwide maintenance costs of concrete structures are substantial [10,11]. Addressing ASR in existing structures shows notable challenges and financial implications, highlighting the need for proactive ASR prevention strategies and a comprehensive understanding of its mechanisms.

SCMs such as fly ash or calcined clays are pozzolanic materials that can partially replace cement in concrete, reducing the total carbon footprint and enhancing its sustainability [12]. The demand for new types of SCMs is increasing due to further increases in concrete production and the declining availability of, e.g., fly ash [13,14,15]. Therefore, novel SCMs, such as calcined clays, are a promising alternative to meet the requirement of eco-efficient concrete [13]. However, calcined clays are more complex in processing, as they are more cost- and energy-intensive than fly ash, ground granulated blast furnace slag (GGBFS), natural pozzolans, and biomass ashes [16]. Nevertheless, the global availability and high reactivity of calcined clays, along with their potential to significantly enhance the durability of concrete, such as by mitigating ASR, make them a promising alternative for sustainable construction applications [2,17,18,19].

The properties of SCMs regarding the mitigation of ASR are commonly described as the following: a refinement of the concrete pore structure and thus a reduction in the capillary porosity [20,21], a decrease in the pore solution alkalinity and therefore its pH, and modifying the composition of C-(A)-S-H [3,22,23,24]. During the pozzolanic reaction, pozzolanic SCMs consume greater amounts of portlandite than pure Portland cement (PC) systems, reducing the available calcium in the pore solution [25,26].

The composition of the SCM also plays a decisive role in its mitigation potential. Al-rich SCMs (e.g., fly ash, GGBFS) were found to be more effective in mitigating ASR than Si-rich SCMs (e.g., silica fume) [2,27]. This effect is explained by Chappex and Scrivener [27,28], who showed that Al-rich SCMs decrease the pore solution alkalinity and therefore the pH and modify the composition of C-(A)-S-H, which is more significant than their alkali uptake. Others postulate that aluminum might increase the alkali fixation of C-(A)-S-H [22]. However, studies have also shown that ASR products can absorb alkalis and thus reduce the overall expansion in the concrete [2,7,29]. This raises the question of whether Al-rich SCMs can affect the distribution and quantity of ASR products and their ability to uptake alkalis.

Another key factor in the ASR mitigation potential for calcined clays is the reduced dissolution rate of the reactive silicate aggregate by the adsorption of aluminum on the silicate surface [30,31], which might lead to the formation of zeolite instead of ASR products [32]. The adsorption of aluminum on the silica surface was already postulated by Iler [33] and further shown by Chappex and Scrivener [27]. However, the phenomenon of the zeolite formation was only observed at temperatures around 80 °C [30,31]. Zeolites are naturally occurring alkaline aluminosilicate minerals with various compositions characterized by an ordered, microporous framework structure. The term “zeolite precursors” is often used for amorphous aluminosilicate structures that may eventually transform into crystalline zeolites [31,34].

Similarly, sodium-alumina-silicate-hydrate (N-A-S-H) gels form a system that closely resembles ASR gels, as both are amorphous networks formed under alkaline conditions. However, N-A-S-H gels are formed during the alkaline activation of aluminosilicate materials, such as fly ash or metakaolin [35,36,37]. These gels play a significant role in the formation and properties of geopolymers, with applications in construction, insulation, and other fields [37,38]. Furthermore, alongside N-A-S-H gels, the alkaline activation of aluminosilicates may lead to the formation of various zeolite (nano) crystalline phases [39].

For ASR products, depending on the composition and formation conditions, crystalline (e.g., K-shlykovite) and amorphous products (e.g., ASR-P1) can form [40,41,42,43,44]. For amorphous ASR products the term ASR “gels” is often used. The composition of ASR gels varies widely and depends on factors such as the mineralogical composition of the concrete mix (e.g., binder, aggregate), gel formation location (e.g., near the cement paste or within an aggregate), and gel maturity [45,46]. Regarding the mechanism of pressure build-up during the alkali–silica reaction, various approaches, such as solidification pressure [47], osmosis [48,49,50], or the electrical double-layer theory [51], are considered. For amorphous ASR products or ASR gels, water-induced swelling has long been named the leading cause of cracking in concrete [7,48,49,52,53,54]. However, recent studies suggest that the solidification pressure may also be a significant factor in ASR-related concrete deterioration [40,47,55,56]. In this study, the focus will be on the water-induced swelling mechanism.

In addition to the previously mentioned positive effects of calcined clays in mitigating ASR, the alteration of ASR products by aluminum was also hypothesized [17,18,57,58]. The authors describe that the presence of aluminum may alter the composition and structure of ASR products so that they are no longer capable of swelling [17,18,57]. Krüger et al. [59] demonstrated the incorporation of aluminum in the structure of synthetic ASR gels by various analytical techniques. They showed an increase in the polymerization of the gel by the presence of aluminum compared to Al-free ASR gels. In a follow-up study, Krüger et al. [60] showed that aluminum stabilizes silicon within the gel structure and binds water in smaller pores than the Al-free gels. Swelling tests revealed an expansion suppression for synthetic Al-Ca-Si gels by a decreased water uptake. However, for crystalline ASR products, the incorporation of aluminum is unfavorable, and instead, a zeolite precursor is formed [31,61].

The controversial discussion about the role of aluminum in ASR mitigation highlights the necessity of conducting further investigations to understand the positive effect of aluminum on ASR completely. In particular, the question remains whether aluminum can be incorporated into the structure of natural ASR products and how this affects their properties. In addition, the question of whether Al-rich SCMs could alter the composition of ASR gels is still open.

Calcium also plays an important role in the expansion potential of ASR gels [43,54,62,63,64]. Gholizadeh-Vayghan and Rajabipour [63] showed that calcium changes the stiffness and viscosity of ASR gels, which affects their yield stress and swelling properties [59]. Vertical swelling experiments by Krüger et al. [60] showed a linear correlation between the vertical swelling and the Ca/Si concentration of the investigated synthetic gels. According to Mansfeld [54], only gels with 5 to 30 wt.% CaO can generate swelling pressures, which are significant enough to lead to deterioration in concrete. This agrees with Shi and Lothenbach [43], who showed that ASR products destabilize at higher Ca/Si ratios (i.e., above 0.5). According to several authors [44,65,66], the structure of ASR gels becomes more similar to that of a C-S-H phase when even the calcium content in the gel is increased.

In summary, results on synthetic gels have shown that aluminum and calcium alter the structure and therefore expansion properties of synthetic ASR gels. Since calcined clays can provide significant amounts of aluminum and might change the available calcium for gel formation by consuming portlandite, it is essential to understand to what extent calcined clays can affect the properties of natural ASR gels. Therefore, this study aims to elucidate whether calcined clays could affect the morphology, structure, and composition of ASR gels in concrete and how this may affect their expansion properties in concrete.

2. Materials and Methods

2.1. Materials

Five different concrete mixes were produced with a binder content of 400 kg/m³ and a water-to-binder (w/b) ratio of 0.45 (

Table 1. Concrete mix designs. As SCMs LS, MK, MS, and MC were used. BG = borosilicate glass.

		90PC10SCM	100PC
Cement	[kg/m3]	360	400
SCM	[kg/m3]	40	0
0/2 sand	[vol.%]	30
	[kg/m3]	531
2/8 BG	[vol.%]	70
	[kg/m3]	1054
w/b ratio		0.45

). As cement, an ordinary Portland cement (OPC, PC) CEM I 32.5 R with a Na₂O_eq of 1.3 wt% was used. 10 wt% of OPC was replaced by an SCM (metakaolin (MK), metasilt (MS), and metaclay (MC)) or by a filler reference material (LS = limestone). Additionally, one concrete mix was tested with pure OPC as a binder. The aggregates used were 531 kg/m³ of 0/2 quartz sand and 1056 kg/m³ of 2/8 borosilicate glass (BG) as a homogenous, reactive material. The binder mixes, and the chemical composition of the binders and aggregates are listed in

Table 2. Binder mixes investigated.

	OPC	MK	MS	MC	LS
	[wt%]
90PC10MK	90	10	0	0	0
90PC10MS	90	0	10	0	0
90PC10MC	90	0	0	10	0
90PC10LS	90	0	0	0	10
100PC	100	0	0	0	0

and

Table 3. Chemical composition of PC, SCMs, and aggregates in [wt%] determined by inductively coupled plasma–optical emission spectroscopy (ICP-OES). Values that were not determined are indicated as [-].

	SiO2	Al2O3	CaO	Fe2O3	K2O	MgO	MnO	Na2O	P2O5	SO3	B2O3	TiO2	LOI	Na2Oeq
OPC	21.0	5.7	60.5	3.2	1.7	2.4	0.1	0.1	0.2	3.0	-	0.3	1.6	1.3
Sand	69.6	14.3	2.2	4.6	2.9	1.7	0.0	3.1	0.1	0.1	-	0.6	0.2	5.0
BG	80.0	2.3	0.1	0.1	1.1	0.0	0.0	3.6	0.0	0.3	12.3	0.0	0.1	4.3
MK	51.7	42.2	0.4	0.6	0.3	0.1	0.0	0.6	0.1	0.1	-	1.5	2.1	0.8
MC	52.6	20.5	6.3	8.0	3.0	2.9	0.1	0.5	0.3	1.5	-	1.0	2.8	2.5
MS	54.0	32.4	0.3	1.4	4.6	0.3	0.0	0.3	0.5	0.1		0.6	0.6	3.1
LS	2.1	0.6	52.5	0.3	0.3	0.9	0.0	0.4	0.1	0.3	-	0.0	42.1	0.6

.

Three different calcined clays were used in small concentrations (10 wt.%) to allow gel formation. The mineralogical composition of the SCMs determined by QXRD is listed in

Table A1. Mineralogical composition of the SCMs used as determined by quantitative X-ray diffraction (QXRD) [wt%].

Phase Name	Formula	MK	MS	MC
Amorphous content		94.4	61.8	49.4
Muscovite	KAl2(AlSi3O10)(F,OH)2	1.5	2.1	13.7
Kaolinite	Al2(Si2O5)(OH)4	2.7	0.9	-
Quartz	SiO2	-	4.7	18.0
Microcline	KAlSi3O8	-	10.5	4.8
Illite	(K,H3O)Al2(Si3Al)O10(H2O,OH)2	-	5.4	-
Orthoclase	KAlSi3O8	-	5.6	-
Sanidine	(K,Na)(Si,Al)4O8	-	9.0	-
Hematite	Fe2O3	-	-	1.9
Albite	NaAlSi3O8	-	-	1.9
Anorthite	CaAl2Si2O8	-	-	3.3
Anhydrite	CaSO4	-	-	1.5
Calcite	CaCO3	-	-	3.6
Dolomite	CaMgCO3	-	-	1.9

. The chemical composition of the amorphous content evaluated by ICP-OES is listed in

Table 4. Chemical composition of the amorphous content (Table A1) of the calcined clays used in [wt%] as determined by ICP-OES.

	MK	MS	MC
LOI	1.7	0.2	0.0
Al2O3	42.5	43.0	23.8
CaO	0.4	0.6	10.4
Fe2O3	0.7	2.5	11.4
K2O	0.1	0.0	1.0
MgO	0.1	0.5	5.4
MnO	0.0	0.0	0.3
Na2O	0.6	0.0	0.5
P2O5	0.1	0.8	0.5
SO3	0.1	0.1	2.7
SiO2	52.1	51.0	42.0
TiO2	1.6	1.1	1.9
Al2O3/SiO2 [wt%]	0.8	0.8	0.6
Al/Si [mol%]	0.5	0.5	0.3

. The metakaolin (MK) (BURGESS OPTIPOZZ) used in this study is a commercial product from Burgess Pigment Company. The silt for the metasilt (MS) was produced as a side product from a Kaolin production company called Amberger Kaolinwerke. For calcination, the silt was first crushed in the lab to a fineness smaller than 125 µm. Several ceramic beakers were then filled with 200 ± 1 g of the silt and calcined from room temperature to 900 °C in an oven at a 10 K/min heating rate. The maximum temperature of 900 °C was maintained for 1 h and then cooled back to room temperature. The metaclay (MC) is a calcined Amaltheen-mixed clay provided by the Institute for Construction Materials at the Bundeswehr University Munich [67]. The clay was calcined in a rotary kiln at 750 °C.

Three prisms (40 × 40 × 160 mm) of each concrete mixture were produced and prepared for ASR testing. The prisms underwent curing at 20 °C in a 100% R.H. environment for 28 days, followed by storage at 40 °C above water for 140 days, with regular expansion monitoring. After a 140-day storage period, the concrete samples were cut with a slight amount of water and stored in the desiccator until analysis (Raman spectroscopy, µ-XRF). The only exception was the concrete 100PC, which was examined after approx. 190 days, as this sample was part of a previous project and was also used as a reference here. For further microscopic investigations (light microscopy, scanning electron microscopy (SEM), electron probe microanalysis (EPMA)), thin sections (30 µm, 40 × 40 mm) at the same locations as the cut samples (Figure 1) were prepared in oil to minimize leaching by an external preparation lab (Dettmar dissection technology GmbH & Co. KG, Bochum, Germany). The particle size distribution of the PC and SCMs used is shown in Figure 2.

2.2. Analytical Methods

Automated mineralogy measurements were performed on thin sections of 90PC10MK, 90PC10MS, 90PC10MC, and 90PC10LS at the Department of Geoscience and Petroleum, NTNU Trondheim, Norway, to investigate the concrete microstructure and chemical composition of the ASR gels. The ZEISS Sigma 300VP SEM (The device was used at NTNU in Trondheim, Norway) operates with two Bruker Xflash 6|60 129 eV energy dispersive spectrometry (EDS) detectors and ZEISS Mineralogic as automated mineralogy software (SEM-AM). For SEM-AM measurements, high vacuum mode with an acceleration voltage of 15 keV, an aperture width of 120 µm, and a beam current of 100 nA at a working distance of 8.5 mm were applied with a minimum of 2000 counts and a dwell time of 0.008 s at a step size of 1 µm. This small step size ensures the detection of the chemical composition of the gels of interest in high resolution. Thresholds and other arithmetic image processing scripts were applied to avoid unnecessarily measuring epoxy. Before the measurement, the thin sections were coated with a carbon layer of 15–20 nm thickness. For the SEM-AM, a minimum of four areas per concrete with a multitude of data points, numbering several thousand points per gel phase, were recorded for each measurement area.

Additionally, data for the mineral/phase maps were reprocessed as elemental maps, visualizing the intensities and spatial distribution of the elements of interest. The classification list for these elemental maps (Si, K, Ca, Al, and Na) contains classifications for one element with incremental increases in concentrations. For the major elements Si, K, and Ca, an increment of 2 wt% at a range of 0–100 wt% in their concentration, indicated by different colors in the respective elemental maps, was set. Whereas for the minor elements Na and Al, an increment of 0.5 wt% at a range of 0–25 wt% was used to illustrate the differences in concentration. Mineral and backscattered electron (BSE) maps and data tables were generated simultaneously for each analyzed area.

Table 5. SEM-AM main classified phases and set concentration ranges in [wt%]; BG = borosilicate glass.

Phase	Na	K	Ca	Si	Al
Si gel	2–20	4–24	0	31–44	0
Ca-Si gel	0–8	4–20	5–20	27–50	0
Al-Ca-Si gel	0–4	4–20	0.1–16	31–44	0.1–6
Al-Si gel	0–4	4–20	0	28–44	0.1–6
BG	0–6	0–3	0	44–57	0

illustrates the predefined concentration limits for each element within the classified phases. Although the text frequently mentions ASR gels, the phases classified here are referred to simply as “gels”. This terminology was used because the study did not examine whether the different gels contribute to concrete deterioration. The alkali metals (Na, K) contained in the gel phases were not listed separately in the phase labels, as they were contained in all phases.

Both µ-XRF and SEM-EDS are techniques for visualizing the distribution of elements in a sample, but they differ significantly in resolution and application. SEM-EDS offers much higher spatial resolution, allowing for the detailed analysis of the microstructure and chemical composition (Ca/Si, Na/Si, K/Si, and Al/Si ratios) of ASR products at the microscale. In contrast, µ-XRF provides lower spatial resolution but is advantageous for quickly scanning larger samples to determine the elemental distribution, map the gel distribution across larger sample areas, estimate the amount of gel present, and assess its potential for alkali uptake.

The Bruker M4 Tornado µ-XRF spectrometer of the Department of Structural Engineering at NTNU Trondheim, Norway, was used to determine the spatial distribution and alkali uptake of ASR products within the concrete samples. The µ-XRF is equipped with a silicon drift detector energy dispersive spectrometer (SDD-EDS), a silver X-ray tube, and polycapillary lenses. For the measurements, a current of 600 µA, a voltage of 50 kV, a chamber pressure of 20 mbar, and a spot size of 20 µm with a pixel time of 2 ms/pixel was set. It is important to note that the resolution set may not be sufficient to fully capture small pockets, crystallites, or products in thin aggregate veins. However, the goal was to use µ-XRF as a rapid screening method, complementing other microscopy techniques such as light microscopy and SEM-EDS. The intensity distribution contrast map for K and the superimposed K+S maps were generated by the M4 Tornado software. The K and S element maps were used as areas where ASR gels are located, with potassium being most pronounced while sulfur was absent.

The µ-XRF maps were processed further using a Python script to calculate the number of pixels per unit area. The script was executed in Jupyter Notebook (web-based 6.4.12), provided through Anaconda Navigator (Anaconda 3), and is included as Supplementary Data. First, the elemental maps (K, K+S) of the five different concrete mixes were cropped to a specific pixel size (approx. 550 × 550 pixels). Then, the respective image was converted to RGB mode for data processing to determine the number of “red” pixels in the intensity distribution contrast map for K. A counter for R > 100, G < 100, and B < 100 for the “red” pixels was set. In the second step, a loop through each image pixel was generated, and the RGB values in each image were determined. In the last step, the number of “red” pixels was counted in relation to the total number of pixels and displayed as a percentage.

To determine the proportions of the binder, gel, and aggregate in the respective concretes, the superimposed K+S maps were used. The same steps described above were followed, except this time, three different pixel colors were counted: yellow for the binder, pink for the gel, and black for the aggregate. The following RGB values were set for the pixel calculation of the image: pink (R > 120, G < 100, B > 100), yellow (R > 150, G > 150, B < 100), and black (R < 50, G < 50, B < 50). The proportions for the binder (yellow), the gel (pink), and the aggregate (black) pixels were calculated and normalized to 100. These area % were used to further determine the proportion of gel per concrete. It should be noted that only one section per concrete sample was analyzed; therefore, conclusions drawn from these analyses are prone to a certain error due to potential heterogeneity. Nevertheless, these results provide an initial assessment of the gel contents and serve as a promising screening method.

The proportion of gel and its alkali uptake in the investigated sections was calculated from the µ-XRF results and concrete recipe data (Table 1). For this, a volume model was calculated using the densities of the starting materials and the determined area % of the respective gel, aggregate, and paste. A density of 1300 kg/m³ was assumed for the gel based on results from a previous study [60]. For the aggregate fractions, a density of 2275 kg/m³ and for the paste of 2108 kg/m³ were used from the concrete recipes (Table 1).

To determine the alkali uptake of the gels in the investigated concrete sections, the proportion of alkalis (Na₂O, K₂O) provided by the binders was calculated from the ICP-OES results (Table 3) assuming that all alkalis present in the binder (PC and SCM) reacted. It should be noted that the aggregates can also provide alkalis to the system (Table 3), which were not included in the calculation.

First, we determined the amount of gel present in the concrete by dividing the determined area % by the gel density. Knowing the average composition of the gel, we calculated the quantities of Na₂O and K₂O based on the known volumes and composition. For this, an average composition of the Ca-Si gel with ratios of Ca/Si = 0.18, K/Si = 0.30, and Na/Si = 0.08 was used based on the SEM-AM results. By combining the proportion of gel in the concrete with the chosen gel composition, we calculated the amount of alkali metals in the gel. This alkali content was then compared to the total alkali content of the paste for each individual section. Leaching effects during the exposure were not considered in the calculation. Moreover, it cannot be excluded that alkali-absorbed C-(A)-S-H was measured as an intermixing with the gels and therefore inadvertently included in the volume calculation, leading to an overestimation of the gels.

EPMA-wavelength dispersive spectra (WDS) measurements were performed to verify the ASR gels composition in the respective sections determined by SEM-EDS, especially for the elements of low concentrations, e.g., aluminum and sodium. The concrete thin sections of 100PC, 90PC10MK, 90PC10MC, and 90PC10MS were analyzed with a Cameca SX 100 electron microprobe of the Department of Geosciences at LMU Munich, Germany, with five wavelength dispersive spectrometers. Before the measurements, a calibration was performed using the standards listed in

Table 6. EPMA crystals and standards used for calibration, and defined WDS element lines.

Element Line	Crystal	Standard
Na Ka	TAP	Albite
Al Ka	TAP	Orthoclase
Si Ka	TAP	Albite
K Ka	PET	Orthoclase
Ca Ka	PET	Wollastonite

, and the electron beam focus was adjusted. For the measurements, a 15 keV accelerating voltage, a 5–10 nA beam current, and a 5 μm beam diameter were used. The counting time was 10 s for all elements using the integral method and a working distance of 300–320 µm. To assess the effect of the interaction volume, an intact area of the borosilicate glass was measured before each analysis, aligning with the chemical data from ICP-OES measurements. The thin sections were coated with a carbon layer of 15–20 nm thickness.

The Raman spectrometer Witec alph2800 R at the Department of Structural Engineering at NTNU Trondheim, Norway, was used to determine the structure of the ASR gels formed in the different concrete samples (100PC, 90PC10MK, 90PC10MC, and 90PC10MS). The samples were measured using a frequency-doubled Nd: YAG laser with a wavelength of 532 nm, a maximum laser power of 20 mW, and a 50 × long working distance objective lens (laser diameter 2 μm). The measuring process was carried out with the software Control Five 5.2. The measurements were performed in a spectral range of 50–4000 cm⁻¹ with a spectral resolution of 2 cm⁻¹. For the measurements, a grit of 600 gr/mm, an integration time of 20 s, and an accumulation of 6 sec were used. At least five measurements were carried out at different positions in each sample area. During each measurement, a background correction and spike suppression were applied. The measured spectra were processed using a baseline correction and a smoothing function. For the deconvolution of the enlarged sections of the Raman spectra, the program PeakFit (version 4.06, Systat Software, Inc., Chicago, IL, USA) was used. For this, the first derivation of each spectrum was baseline corrected and smoothed; afterward, the minimum number of peaks was fitted with a Gauss+Lor Amp function.

3. Results and Discussion

• Concrete expansion measurements

Figure 3 shows the expansion curves of the investigated concretes. After 140 days, the concretes 100PC and 90PC10LS show the greatest expansions of 1.70 mm/m and 1.51 mm/m, respectively. The mixes containing calcined clay show a maximum expansion of 0.96 mm/m (90PC10MC), 0.77 mm/m (90PC10MS), and 0.51 mm/m (90PC10MK). Hence, the use of 10 wt% calcined clays significantly lowers the ASR-related expansion of the concretes. This positive effect is most pronounced for the concrete 90PC10MK, where the expansion after 140 days is more than three times lower than the concrete 100PC.

The limestone concrete (90PC10LS) exhibits only a slight reduction in expansion compared to the pure Portland cement concrete (100PC) (Figure 3). In contrast, the concrete mixes with calcined clays (90PC10MK/MS/MC) demonstrate a significant reduction in expansion. This indicates that the reduction in expansion by the calcined clays is not solely due to dilution but involves more reactive mechanisms. The substantial mitigation potential of the calcined clays cannot be attributed to a single mechanism; it results from a complex combination of various processes.

Looking at the expansion curves, one can see that all curves show a steep increase up to day 28, which continues and reaches its maximum after approx. 84 to 140 days. The ASR mechanism can be divided into four petrographic stages: initial (I), development (II), acceleration (III), and deterioration stage (IV) as defined by [7]. The areas marked in Figure 3 were determined based on the expansion curve progression compared to [7]. The following description is not derived from the results of this study but rather serves as a general overview of ASR development, as outlined in [7]. During the initial stage (I), ASR leads to the formation of reaction rims and sol/gel migration through the microstructure of the concrete, where the gel first deposits within voids. The length of the initiation phase and the subsequent pace of growth of the ASR gels relies on the alkali content in the concrete mix, the types of aggregates, and environmental factors (e.g., temperature, relative humidity). During the development stage (II), the crack formation in aggregates propagates into the cement paste. The expansion process and ASR gels within cracks can be found in the acceleration stage (III). Within the deterioration stage (IV), the expansion rate decreases, and a maximum crack width is reached.

Under consistent conditions, the initial growth rate might diminish, and expansion may eventually halt. This halt in expansion can be explained by a reduction in available alkalis due to leaching and/or chemical binding of alkalis by ASR products formed [2,7,29]. Lindgård et al. [29] reported that a significant amount of alkalis can be leached out of concrete prisms during the exposure condition of accelerated ASR testing. Upon examining the concrete samples after 140 days of storage or later, it is very likely that the composition had already undergone alterations prior to the mineralogical investigation of the gel.

The positive effect of SCMs on mitigating ASR has already been extensively investigated in recent decades [2,58,68,69]. As mentioned above, various mechanisms are reported to explain the ASR mitigation effect of Al-rich SCMs, such as the lowering of the pH in the pore solution [27,70,71] by alkali binding or alkali dilution, mass transport reduction [20], the consumption of portlandite [25,72,73], and reducing the dissolution rate of the reactive aggregates [27,30,72].

It was found that Al-rich SCMs (e.g., fly ash, GGBFS) are more effective in mitigating ASR than Si-rich SCMs (e.g., silica fume) [2,27,74]. This enhanced effectiveness is attributed to their ability to reduce the pH of the pore solution and modify the composition of C-A-S-H, as previously discussed. Since calcined clays contain large quantities of aluminum [73], a positive effect of MK, MC, and MS on ASR mitigation was expected. In this study, MK had the highest amorphous content while having, in total, the highest Al₂O₃ content and therefore the highest reactivity [75], followed by MS and MC (Table 3, Table A1).

The pozzolanic reaction of the SCMs can lead to a lower Ca/Si ratio and therefore increased alkali uptake in C-(A)-S-H compared to pure PC mixes [27]. However, Kunther et al. [76] showed with thermodynamic modeling that a replacement of only 10 wt% PC by MK has no significant effect on the total Ca/Si of C-(A)-S-H. Moreover, Leemann et al. [18] discovered that the existence of aluminum in the pore solution derived from metakaolin and calcium aluminate clinker results in a delayed dissolution of SiO₂ and therefore a slower formation of reaction products, but not necessarily in a lower overall quantity of SiO₂ dissolution. According to Hemstad et al. [77], the soluble aluminum in SCMs is primarily incorporated into C-(A)-S-H, whereas a significant portion is likely to remain in unreacted phases.

As a next step, the concrete microstructure was investigated to link the ASR-related expansion in concrete to the gel formation. First, µ-XRF measurements were performed to determine the ASR gel distribution and calculate their potential alkali uptake.

3.1. Effect of Calcined Clays on the Distribution of ASR Gels and Their Alkali Uptake

• µ-XRF

Figure 4 shows the µ-XRF maps for the elements Ca, Si, K, S, and Al, together with an overview image of the investigated concrete 100PC section. The Ca, Si, S, and Al concentrations appear relatively homogenous within the investigated section. For the areas where ASR gels are located, sulfur seems absent, whereas the contrast for potassium is most significant. Therefore, these element maps (S, K) were further processed to identify the ASR gel distribution in the concrete (Figure 5, Figure 6 and Figure 7).

Figure 5 shows the overview image of the sample 100PC (a), the superimposed K+S element map (b), and the intensity distribution map for K (c). From the overview image, large areas of ASR gel are visible, which appear whitish and predominantly porous. A large proportion of air voids and some cracks within aggregates are partly filled with ASR gel (Figure 5a).

The K+S map shows ASR gels in pink, whereas areas in the cement paste with no ASR gel present appear yellow (Figure 5b). The concentration of K is higher in the ASR gel than in the surrounding cement paste or the aggregates; therefore, the intensity distribution contrast map of K is best suited to investigate the distribution of ASR gels (Figure 5c). The areas with the highest potassium concentration appear red, whereas lower concentrations appear yellowish-greenish and lowest blue. However, this method is quite threshold sensitive when determining the area percentage. Depending on the threshold set, this affects the area percentage of gel determined. If the areas with a low K concentration (yellowish-greenish) are also taken into account, the percentage of gel is almost twice as much. To counteract the threshold problem, the amount of gel was additionally determined by the superposition of the K+S map.

Figure 6 shows the intensity distribution map of K for the investigated concretes. The red areas indicate the presence of ASR gels in the investigated concrete section. The number at the right corner of each figure indicates the relative area % of the gels determined by red pixel counting in the respective concretes.The comparison of the respective concrete sections shows the highest area percentage of ASR gels for the concrete 90PC10MC (Figure 6c, approx. 5.0%), followed by 100PC (Figure 6d) and 90PC10MS (Figure 6b) with approx. 4.0%. The lowest amount of red pixels are shown in 90PC10MK (Figure 6a, 2.6%), followed by 90OPC10LS (Figure 6e, 3.2%).

Taking the expansion results into account (Figure 3), the ASR gel distribution results (Figure 6) indicate that the concrete expansion does not necessarily correlate with the amount of ASR gel in the concrete. However, as only one section per concrete was investigated within this study, the potential heterogeneity of the concrete was not focused. Furthermore, additional mechanisms, including mechanical and physical aspects such as strengths and porosity, significantly affect the transport processes and thus the expansion behavior of concrete, which were not investigated in the present study.

It is known that the location of the gel (within the aggregate or paste) in the concrete plays a crucial role [45]. If the gel propagates from the aggregate further into the binder matrix, this is an indicator of a deleterious ASR gel [7]. If ASR gel is found in air voids, this is not necessarily related to cracking in the concrete and therefore is a non-deleterious product. While the refinement of concrete pore structure through the reduction in capillary porosity is commonly cited as a beneficial effect of SCMs in mitigating ASR, it is essential to distinguish between capillary pores and larger air voids. Capillary pores facilitate the movement of moisture and alkalis, contributing to ASR expansion [78].

In contrast, larger air voids may serve as ‘relief zones’ for ASR gel formation, preventing the development of internal pressures that would otherwise lead to cracking, similar to internal frost damage [79].

In addition to the proportion of gel, the amount of aggregate and cement paste in the respective concrete were determined using the K+S maps of the respective concretes (Figure 7a). After evaluating the pixel per area unit (Figure 7b), the cement paste content for the concrete 100PC (yellow pixels) was approximately 34.5%, and the proportion of aggregate was approximately 60% (black pixels). The pink pixels represent the gel area %, contributing to 5.5% of the total. The percentages differ between the various concrete mixes in the range of 2–6% for the gel, 28–43% for the paste, and 54–66% for the aggregate. These values were used to determine the approximate amount of gel in the concrete and thus the potential alkali uptake of the gels.

Figure 8a shows the calculated volumes of gel, paste, and aggregate per concrete based on the results from Figure 7b and the calculated densities for paste and aggregate from the concrete recipe (Table 1) and gel (according to [60]). The calculations showed an average gel content per concrete of approximately 3 wt%. The amount of gel was lowest for 90PC10MK (1.4 wt%) and highest for 90PC10MC (4.4 wt%). Figure 8b,c show the calculated amount of K₂O and Na₂O in the binder and the ASR gel. The average amount of K₂O was approx. 28 kg/m³ and of Na₂O approx. 2 kg/m³ in the binder. For the gel, the average of K₂O was approx. 15 kg/m³ and of Na₂O approx. 2.5 kg/m³. Figure 8d shows the alkali uptake of the ASR gel relative to the binder (PC+SCM). The alkali uptake is 100% when the gel has absorbed all the alkalis from the binder. The results indicate that the gel can absorb nearly all the Na₂O, or even more, and up to two-thirds of the K₂O in the paste. This suggests that the gel incorporates more Na₂O than the binder alone can provide. A possible explanation for this phenomenon is the dissolution of aggregates, which also contain Na₂O and are present in the concrete mix in significant volumes (Table 1 and Table 3).

An exception is observed in the 90PC10MK mixture, where the smallest amount of gel is formed, resulting in the lowest alkali uptake by the gel. The higher Na₂O uptake is therefore not due to the gel inherently absorbing more Na₂O per unit mass but rather due to the greater amount of gel present in the system.

As mentioned above, the chemical absorption of alkalis by the ASR gel could have led to a halt in expansion in concrete in addition to leaching (Figure 3). The results show that the gels formed in the concrete have a high potential to absorb large quantities of alkali metals.

The µ-XRF investigations have shown that this method is well suited to determine the distribution of ASR gel within concrete samples and even allows for a rough estimation of the alkali uptake in the gels. In the next step, thin sections from the same areas were examined using light microscopy to observe the various concrete samples’ crack patterns and microstructures.

• Light microscopy

The microscopic examination in polarized and fluorescent light showed significant amounts of ASR gels in all investigated concrete samples. Hence, all investigated samples have overcome the initial stage of ASR deterioration (Figure 3). During the initial stage, the cracks that form remain void. However, as the reaction consistently develops (development stage), the voids within the aggregates gradually become filled. Subsequently, the reaction products extrude into the fissured cement paste [7]. The weighing factor of petrographic features starts with cracks in coarse aggregates as the lowest factor, followed by open cracks or filled cracks with ASR products within aggregates [7]. These features can be observed for all investigated concretes.

As there have been significant differences in expansion between the concretes 100PC and 90PC10MK, these two concretes are directly compared in the following with respect to their microscopic structures (Figure 9). Concrete 100PC exhibits numerous cracks in the coarse aggregates. Cracks within the aggregate narrow toward the aggregate edges and continue out into the surrounding paste. The cracks in the cement paste are partially filled with clear alkali–silica gel, displaying a typical texture [80]. Alkali–silica gel is observed at the mouth of the cracks and within the cracks through the paste in several spots (Figure 9, left 100PC). ASR gel also frequently appears within air voids as a rim. The cement matrix shows a network-like crack structure in various locations, typical for ASR (acceleration stage). However, intact areas without cracks are also present.

The direct comparison of 100PC and 90PC10MK significantly shows fewer pronounced cracks in the coarse aggregate for the 90PC10MK section. Nevertheless, ASR gel fillings are found in several places within the coarse aggregate of 90PC10MK. The binder matrix is mainly intact in many of those areas.

The investigation of the other concretes (90PC10LS/MS/MC) reveals a greater crack pattern than for 90PC10MK (Figure A1). The thin section of the concrete 90PC10LS shows heavily cracked coarse aggregate grains and air voids with rims of ASR gels, although the cement matrix appears intact in main areas. In some parts, ASR gel occurs as a crack filling or is within the borosilicate aggregates. It should be noted that the glass particles used in the concretes underwent crushing and may retain intra-particle microcracks. The concrete 90PC10MC shows a more pronounced crack pattern in the binder’s matrix than concrete 90PC10LS. However, in both concretes, ASR gels are found primarily in air voids and borosilicate glass grains, not as crack fillings within the paste like for 100PC. Leemann et al. [81] found a significant number of cracks in aggregates without ASR products, indicating a substantial degree of product mobility by low gel viscosity.

The overall analysis of the crack structure of the investigated concretes reveals the most substantial crack pattern for concrete 100PC, followed by 90PC10MC and 90PC10LS. However, concrete 90PC10MC shows fewer cracks in the coarse aggregate than 90PC10LS and 100PC. Concrete 90PC10MK shows the least cracking, with only a few fine net-like cracks in the binder matrix. The same applies to sample 90PC10MS, although more gel occurs in air voids in the investigated sections.

In order to further investigate the properties of the ASR gels formed in concrete, Raman spectroscopic investigations were performed to determine the effect of different concrete mixes on the structure of ASR gels.

3.2. Effect of Calcined Clays on the Structure of ASR Gels

• Raman spectroscopy

In Figure A2, the Raman spectrum of the borosilicate glass (BG) is plotted together with the ASR gels present in 100PC and 90PC10MK. The spectral comparison shows that the prominent peak position of BG at approx. 445 cm⁻¹ does not interfere with the prominent peak positions of the gels.

Figure 10 shows the two characteristic Raman peak patterns of the ASR gels in the investigated concrete samples. The vibrational bands are in the range of 400–700 cm⁻¹ and 850–1200 cm⁻¹, corresponding to a Si-O-Si and Si-O bond, respectively (Table 7). The small peak around 465 cm⁻¹ can be attributed to the symmetric bending of either O_NBO-Si-O_NBO [82,83] or Q⁴ sites [84,85]. The major Raman band for the low-frequency region is around 650 cm⁻¹, which can be attributed to the symmetric bending of Q² sites [84,86]. The peak at 900 cm⁻¹ is attributable to the symmetric stretching of Q¹ sites [87]. The peak around 1020 cm⁻¹ can be assigned to the symmetric stretching of Q² sites [86,88].

The prominent peak for the high-frequency region is around 1080 cm⁻¹, which can be attributed to the symmetric stretching of Q³ sites [84]. It should be noted that the peak position at 1080 cm⁻¹ overlaps with the prominent peak position for calcite (v_s [CO₃²⁻) [83]. Since the carbonation of the gel surface cannot be avoided entirely, this attribution would also be conceivable. Overall, the ASR gels formed in the investigated concrete samples show strong similarities to the Raman spectra measured in other concrete samples [80,85] and synthetic ASR products with similar compositions [89,90].

Table 7. Raman bands with peak assignments of the spectra shown in Figure 10; T = Al, Si.

Raman Shift [cm−1]	Assignment	Reference
465	Si-O-Si (Q4)	[84,91]
650–661	SB Si-O-T (Q2/Q3)	[84,86]
900	SS Si-O (Q1)	[87,89]
1012–1020	SS Si-O (Q2)	[86,88]
1077–1088	SS Si-O (Q3)/C-O (CO32−)	[84,87]

Table 7. Raman bands with peak assignments of the spectra shown in Figure 10; T = Al, Si.

Raman Shift [cm−1]	Assignment	Reference
465	Si-O-Si (Q4)	[84,91]
650–661	SB Si-O-T (Q2/Q3)	[84,86]
900	SS Si-O (Q1)	[87,89]
1012–1020	SS Si-O (Q2)	[86,88]
1077–1088	SS Si-O (Q3)/C-O (CO32−)	[84,87]

The direct comparison of the spectra shows no significant difference between the investigated gels in the respective concrete compositions (90PC10MK/MS/MC and 100PC). In order to see detailed differences in the structure of the measured gels, a deconvolution of the spectra was performed. However, it should be noted that Raman spectroscopy is not a quantitative method; therefore, the different Q sites can only be compared qualitatively.

Figure 11 shows the deconvoluted Raman spectra of the investigated gels in the various concretes. The corresponding peak positions of the deconvoluted Raman bands are listed in

Table A2. Peak assignment of Raman bands from deconvolution results of the investigated concretes, bold numbers indicate the prominent peak positions.

	Ramanshift [cm−1]
90OPC10MC	496	574	611	662	897	951	1010	1081	-	1139
90OPC10MS	477	558	600	652	890	974	1010	1079	-	1149
90OPC10MK	498	522	603	660	885	993	1024	1086	1113	1144
100PC	466	537	597	653	902	967	1017	1074	1112	1151
Assignment	SB Si-O-Si Q4	SB Si-O-Si Q3	SB Si-O-Si Q3	SB Si-O-T Q2 or Q3	SS Si-O Q1	SS Si-O Q2	SS Si-O Q2	SS Si-O Q3/SS C-O	SS Si-O Q3	SS Si-O Q4
Ref.	[84,86]	[87,90]	[40,87,105]	[40,106]	[89,90]	[84,87]	[86,88]	[84,87,107]	[40,86]	[108]

. The comparison of the four spectra shows different intensity ratios between the Q² (1010–1024 cm⁻¹) and the Q³ band (1075–1086 cm⁻¹). The intensity difference for Q²:Q³ is more pronounced for the gels measured in the calcined clay mixed concretes (90PC10MK/MS/MC) compared to 100PC. The intensity difference could indicate a higher proportion of Q³ and thus a higher degree of Si polymerization (DP) for the gels formed in the concretes 90PC10MK/MS/MC than in 100PC.

According to Neuville et al. [92], with increasing Al₂O₃ content, the prominent Raman bands of the glass in the 900–1300 cm⁻¹ region become narrower and shift to lower frequencies. However, no shift to lower frequencies is observed for the gels (Figure 11 and Table A2). This may be due to the aluminum concentration in the gels, as measured by Raman spectroscopy, not being high enough to induce a shift. Krüger et al. [59] demonstrated a shift in the maximum vibrational band using FTIR spectroscopy, but only for gels with an Al/Si ratio greater than 0.03. This absence of a frequency shift makes it challenging to directly determine the presence of an aluminosilicate structure from the Raman spectra, as the band positions for aluminosilicates and alkali–silicate structures overlap [93]. According to Leemann et al. [18], the presence of aluminum results in a gel structure primarily characterized by Q² sites, whereas in the absence of aluminum, the structure is dominated by Q³ sites. However, Leemann et al. [18] could not find significant alterations of the ASR product structure, morphology, or composition with an increasing amount of aluminum in the pore solution.

Although the intensities can only be compared relatively amongst each other, investigations by Krüger et al. [90] have shown that the main band position in the deconvoluted Raman spectrum also reflected the main structural units in the ²⁹Si NMR spectrum. In a previous study by Krüger et al. [93], ASR gels formed in concrete (Ca-ASR gel C_0.2S(N,K)_0.40 and Al-ASR gel C_0.1S(N,K)_0.40.A_0.10) were synthesized based on their average composition and structurally characterized using ²⁹Si MAS NMR to connect the effect of the chemical composition with the structure of natural ASR gels. A comparison of the ²⁹Si NMR spectra after deconvolution revealed consistent resonances at approximately −80 ppm, −85 ppm, −88 ppm, −92 ppm, −97 ppm, and −107 ppm, corresponding to Q¹, Q²(I), Q², Q³(I)/(1Al), Q³, and Q⁴ units in ASR gels (Figure A3). The calculated degree of polymerization was 0.56 for the C_0.2S(N,K)_0.40 gel and 0.64 for the C_0.1S(N,K)_0.40.A_0.10 gel, indicating increased connectivity in the Al-ASR gel compared to the Al-free Ca-ASR gel [93]. A higher DP could indicate lower expansion properties of the gels, as found by Krüger et al. [60] on synthetic gels.

Moreover, the gel formed in the sample 90PC10MK shows narrower bands compared to the gels in the other concrete mixes. The Raman spectrum of crystalline quartz shows primarily more narrow bands, in contrast to fused quartz, which exhibits very broad peaks, showing the effect of the long-range translational symmetry on the peak broadness [94]. This points out that a higher polymerized structure leads to a sharper peak, whereas a less cross-linked or more disordered structure shows broader peaks [95]. Based on this theory, the gels formed in 90PC10MK are expected to show the highest cross-linking compared to the other concrete mixes. It can thus be concluded that the gel in the concrete 90PC10MK exhibited the highest aluminum content.

Raman spectroscopy was employed to analyze the structure of the bulk compositions of the investigated sections; however, it did not extend to examining various gel types. Consequently, SEM-AM was used to distinguish between different gel phases and further elucidate the effect of calcined clays on the characteristics (composition and morphology) of ASR gels.

3.3. Effect of Calcined Clays on the Chemical Composition and Morphology of ASR Gels

• SEM-AM

Figure 12 shows the BSE micrograph image and elemental map montages of a section from the concrete 90PC10 MK determined by SEM-EDS. In this section, ASR gel is distributed as a rim around an air void and as crack fillings within an aggregate. The gel in the displayed section exhibits a relatively uniform Si, K, and Na concentration. As shown by µ-XRF, the potassium concentration stands out in the gel compared to the surrounding binder matrix and the aggregates (Figure 5). Figure 12 shows a reaction front for Ca in the gel. The outer edge of the gel, in direct contact with the paste, contains Ca, while the inner gel, in direct contact with the air void, is free of calcium. A similar pattern is observed for the gel formed within the aggregate. At direct contact with the binder matrix, the gel contains calcium, whereas areas inside the aggregate are entirely devoid of calcium. The situation is reversed for aluminum. While aluminum is present in the binder matrix, it could only be detected in gels formed within aggregates (Figure 12).

The absence of aluminum in the ASR gels found around air voids and cracks in the cement paste suggests that the aluminum-rich gel formed within the aggregate is more stable and remains confined within the aggregate. This confinement, likely due to the higher viscosity of the gel, may prevent it from migrating into the surrounding cement paste matrix. The incorporation of aluminum into the ASR gel structure was already shown in several studies by Krüger et al. [59,60,96]. Studies on synthetic alkali–silica gels have shown that aluminum stabilizes silicon in the gel and reduces its solubility [60]. At the same time, aluminum in the ASR gel leads to free swelling suppression and reduced water uptake [60]. Therefore, as already discussed in [60], Al-bearing gels are not expected to lead to free swelling-induced expansion. However, these experiments only tested the free swelling expansions of the synthetic gels and did not represent the natural concrete system, as the gel expansion in concrete is restrained inside the aggregate particles and in the cement paste.

Several authors highlight calcium as the critical factor for whether ASR products are deleterious in concrete [43,54,62,63]. The Ca/Si ratio primarily controls the stiffness and viscosity of the ASR gel, affecting its swelling properties [53,54]. As shown in Figure 12, only the gel in direct contact with the cement paste exhibits calcium concentrations. After Wang and Gillott [97], Ca²⁺ ions can replace alkali metal ions in the silica gel network. Therefore, it is assumed that a calcium-free gel forms initially and absorbs calcium in the subsequent step.

The SEM-EDS results indicate that gels of various compositions are present within the investigated concretes. Due to this observation, a phase analysis was conducted by SEM-AM. Different gel compositions were classified as phases. The compositional ranges of the classified phases are listed in Table 5. The alkali metals (Na, K) present in the gel phases were not listed separately in the phase labels, as they were included in all phases. Figure 13 shows the phase-classified sections for 90PC10LS, 90PC10MS, 90PCMC, and 90PC10MK determined by SEM-AM. Three main phases could be identified for the investigated sections: Si gel, Ca-Si gel, and Al-Ca-Si gel. The average composition of the identified phases is listed in

Table 8. Average composition of the SEM-AM classified gel phases of the investigated concretes in [at%].

Phase	Na	K	Ca	Si	Al
Si gel	3.8	8.8	0.0	30.0	0.0
Ca-Si gel	2.1	8.5	5.0	28.1	0.0
Al-Ca-Si gel	1.0	8.6	4.1	29.1	1.9
Al-Si gel	1.5	7.6	0.0	30.8	2.3

. The Si gel is mainly found in the cement paste, surrounded by a Ca-Si gel, as indicated in Figure 12. The Si gel is expected to form first due to its low viscosity; it exudates as a fluid into the cement paste and is not expected to lead to any damage [7]. The formation of ASR product initiates near the aggregate and cement paste interface. As the reaction progresses, the formation of ASR products gradually extends toward the interior of the aggregate [7]. The composition of ASR products may evolve with the stage of reaction. The Ca-Si gel falls within the compositional range of reported ASR gels and can be found within aggregates, in the paste, or as a rim in air voids [54,98,99] (Table 8).

Due to its composition, it is expected that this gel will lead to deterioration in the concrete. The Al-Ca-Si gel is present in aggregates and can be found mainly in the concretes containing calcined clays (90PC10MK/MS/MC) and in small amounts in the limestone concrete (90PC10LS). As shown by SEM-EDS measurements in Figure 12, aluminum is present in the gel within the aggregate. The Al-Ca-Si gel exhibits a similar composition to the Ca-Si gel (Table 8). Mainly in the 90PC10MK concrete, an additional phase of an Al-Si gel, which is free of calcium, could be identified. The depletion in calcium can be attributed to the high reactivity of the metakaolin, which consumes the greatest quantity of Ca(OH)₂ in comparison to the other SCMs [75]. Since calcium is essential for building up sufficient solidification [47] or swelling pressure [11,54], it is not expected that the Al-Si gel will cause damage to the concrete.

The results indicate that the type of SCMs present in the concrete mix can affect the composition of the gels formed. The concrete 90PC10MK shows the highest amount of Al-Ca-Si gel, followed by 90PC10MC and 90PC10MS, while 90PC10LS exhibits the lowest (Figure 14). Reactive calcined clays are expected to provide mainly aluminum to the pore solution. The most significant amount of aluminum is expected from MK, followed by MS and MC (Table 4). However, the SEM-AM investigations indicate more Al-Ca-Si gel areas in the concrete sample 90PC10MC than 90PC10MS. This could be explained by the overall higher concentration of the gel for 90PC10MC (Figure 6) compared to 90PC10MS. Moreover, the time of gel formation and therefore the maturity of the gel might play a role in the composition and aluminum concentration of the gel. The results shown in Figure 14 also indicate that concretes with calcined clays form less Si gels than the mix with limestone.

However, the binder and the aggregate in which the ASR product is formed can affect its composition. Using four different reactive aggregates, de Souza and Sanchez [100] showed that the gel composition can vary depending on the type of aggregate (Ca/Si = 0.17–0.56, (Na+K)/Si = 0.19–0.46). Up to 3.6 wt% of aluminum could be measured for ASR products formed within an Al-rich reactive sand and 100 wt% of PC. Since the aggregates used in this concrete also contain aluminum (Table 3), it is challenging to distinguish between the effects of the binder and the aggregate on the gel composition, especially with respect to aluminum. However, there is a clear difference in the amount of aluminum in the gel between the concretes with calcined clays compared to the mix with limestone, so it can be assumed that using SCMs also affects the composition of the gels (Figure 14).

As particularly low concentrations were measured, especially for aluminum and sodium, additional EPMA-WDS measurements were conducted on the concretes 100PC, 90PC10MK, 90PC10MS, and 90PC10MC as an additional method to verify the SEM-EDS results. The results show that SEM-AM is a well-suited method to determine variations in the chemical composition of the gels.

• EMPA-WDS

Figure 15a shows a BSE micrograph image of the concrete section 90PC10MK. Large quantities of ASR gel are found within the aggregate and as a rim of an air void. The ASR gel shows different gray levels, from light to darker gray. The point measurements in Figure 15b reveal higher Ca/Si ratios for gel close to the paste (up to 0.18). The Na/Si concentration varies in the range of 0.05–0.10, and the K/Si ratio is in the range of 0.30–0.36. All investigated areas contained aluminum; however, higher concentrations can be measured within the aggregate as described previously for the SEM-EDS results of this study. The Al/Si ratio ranges from 0.02 to 0.16. Despite differences in grayscale, for example, between Point 7 and Point 8, no significant difference in composition can be identified.

Figure 16 shows the ternary diagram of the EMPA-WDS point analysis and the SEM-AM classified phases of the ASR products formed in the respective concretes 100PC, 90PC10MK, 90PC10MS, and 90PC10MC. The EPMA-WDS data ranges from 0.31 to 0.42 for (Na+K)/Si, 0.02 to 0.55 for Ca/Si, and 0 to 0.16 for Al/Si molar ratios (Figure 16). The SEM-AM determined gel phases (stars) are within the distribution of the EPMA-WDS point measurements, conforming to the measured results.

Overall, within one measurement area, the gel exhibits concentration differences in composition, especially for calcium (

Table A3. EPMA-WDS results for the aggregate reference and the investigated ASR gels measured in aggregates [at%]; SD = standard deviation.

[at%]	Na	Al	Si	K	Ca	O	Al/Si	Ca/Si	Na/Si	K/Si
Borosilicate glass	2.43	1.04	31.07	0.32	0.02	65.11	0.03	0.00	0.08	0.01
SD	0.02	0.02	0.02	0.02	0.02	0.01	0.00	0.00	0.00	0.00
90PC10MK	1.77	2.14	25.13	8.08	2.25	60.64	0.09	0.09	0.07	0.32
SD	0.30	0.81	0.67	0.64	1.67	0.21	0.03	0.07	0.01	0.07
90PC10MC	1.94	1.24	25.18	8.03	3.20	60.41	0.05	0.13	0.08	0.32
SD	0.27	0.17	0.65	0.61	1.41	0.27	0.01	0.06	0.01	0.07
90PC10MS	2.69	1.02	26.26	8.39	1.02	60.62	0.04	0.04	0.10	0.32
SD	0.63	0.25	1.09	1.97	1.15	1.04	0.01	0.04	0.03	0.13
100PC	1.73	1.46	25.95	7.10	2.64	61.13	0.06	0.10	0.07	0.27
SD	0.37	0.76	0.91	0.76	1.34	0.47	0.03	0.05	0.02	0.07

). Similarly to the results of the SEM-AM measurements, the majority of gels can be attributed to the composition of the Ca-Si gel. However, there are mainly Ca-poor and Al-Ca-Si gels within the aggregates. The comparison of the bulk average composition of the ASR gels measured by EPMA-WDS formed in the concrete with different calcined clays confirms the observations of SEM-AM (Table A3, Figure 16). The highest Al concentration could be measured for the gels formed in the concrete sample 90PC10MK. Gels from the concrete 90PC10MS show an overall lower Ca/Si ratio than 90PC10MC, which is consistent with the results from the SEM-AM analysis. The gels in the 90PC10MS exhibited a greater overall presence of calcium-deficient gel than 90PC10MC (Figure 13).

Figure 17a–c show the ternary diagrams of the EPMA-WDS data points from this study and gel compositions from the literature [18,54,58,61,84,99,100,101]. In comparison to the literature [18,54,58,61,84,99,100,101], the gels from this study are, on average, (Na+K) richer and lower in calcium but are in the range of typical gel compositions for less mature gels [18] (Figure 17a). The low concentration of calcium in the gels of the present study could be due to the age of the concrete and therefore less mature gel composition.

Moreover, as a highly reactive aggregate, borosilicate glass exhibits a different dissolution behavior compared to low- or late-reactive aggregates [7,8], which may lead to increased gel formation. Furthermore, it is known that calcined clays consume Ca(OH)₂ during their pozzolanic reaction, which could lead to a lower calcium concentration in the gel in this study compared to results from the literature obtained on PC concrete [72].

The effect of different SCMs on the composition of ASR gels, especially for aluminum, is controversial, as mentioned above. Some authors have hypothesized that the aluminum provided by the SCMs may alter the chemical composition of the gels [17,57,58,100]. In contrast, others have concluded that the presence of aluminum in ASR products comes only from contamination during sample preparation or could be attributed to intermixing with Al-containing phases [61,102].

Nguyen et al. [58] investigated the effect of limestone calcined clay cement (LC³) on the formation of ASR. They found that by replacing parts of the PC with LC³, the composition of ASR products changed in their Al/Si and Ca/Si ratios. Nguyen et al. [58] measured the highest Al/Si ratio for the ASR gels formed in concrete with the highest replacement level of LC³. It is important to note that these measurements were conducted close to the ITZ rather than within the aggregates [58]. De Souza and Sanchez [100] investigated the gel compositions of different concrete mixes with blast furnace slag, metakaolin, and pure PC samples. All ASR gels investigated in [100] contained aluminum. The mix with 50 wt% of blast furnace slag and 50 wt% of PC contained ASR products with up to 25 at% Al (Al/Si ratio of 0.13) measured in the binders’ paste. Shi et al. [103] demonstrated that the properties of ASR products vary significantly depending on the composition of the pore solution and the stage of ASR. This agrees with the observations of this study.

Figure 17b illustrates the Al-Si-Ca ternary diagram for the gels formed in the investigated concretes together with the literature data [58,84,100,101]. The referred authors [58,84,100,101] used SCMs in some cases, while for other concretes, it was pure PC, but the aggregates used potentially contained large amounts of aluminum. In conjunction with the existing literature, the data suggest a maximum threshold for aluminum incorporation in the gels with an Al/Si ratio around 0.16–0.18 (Figure 17b, Table A3). However, it is essential to note that these results likely represent mixed analyses due to the interaction volume of SEM-EDS or EPMA-WDS. This mixing effect could contribute to additional signals from the aggregate or cement paste, indicating that the aluminum concentration in the gels may be lower than the measured values.

The presence of aluminum in the analyzed gels was confined to the borosilicate grain (Figure 12), indicating that the aggregate is a source of aluminum. In addition, while the gels formed in the metakaolin concrete contain the highest concentrations of aluminum (represented by the blue dots in Figure 17b), aluminum is also detected in the gels of pure PC concrete (gray dots, Figure 17b). This indicates that calcined clay is not the sole source of aluminum, as aluminum contributions can also arise from the pure PC matrix or the aggregate. Nevertheless, the EPMA-WDS analysis of the gels (Table A3) revealed that the 90PC10MK concrete exhibited the highest aluminum concentration in the gel, consistent with the SEM-AM findings. The SEM-AM data further demonstrate that the formation of Al-bearing gels is more prevalent in concrete with metakaolin compared to other calcined clays and the limestone mix (Figure 14) despite all concretes having the same aggregate content (Table 1). This indicates the contribution of metakaolin to the gel’s chemical alteration. It is therefore anticipated that studies investigating the effect of higher quantities of calcined clays in the binder will show an increase in the aluminum content within the gels. However, increased metakaolin content may potentially inhibit the formation of ASR gels.

Additionally, all calcined clays contain significantly more aluminum than the borosilicate glass, where the greatest gel quantities were measured (Table 3, Figure 13). The effect of the binder on the gel composition is further supported by the presence of calcium in the Al-Ca gel, which does not originate from the aggregate, implying a reaction with the pore solution. The results indicate that the raw materials, aggregate, cement, and calcined clay, can potentially affect the gel composition not only in terms of sodium, potassium, and calcium but also for aluminum.

Aluminum can replace silicon in a silicate gel network, requiring additional cations like Na⁺ or K⁺ for charge balance when SiO₄⁴⁻ is replaced by AlO₄⁵⁻. Incorporating aluminum into the gel structure may increase the alkali uptake capacity in the gel, as observed in C-(A)-S-H by Hong and Glasser [22]. The results of this study show no significant effect of alkali fixation in the gel, as the aluminum concentration can vary with the same amount of sodium and potassium (Figure 17c). For some gels, higher aluminum content was associated with higher calcium content in the literature [58,100,101]. For charge balancing, the gel can also incorporate Ca²⁺ or H⁺ and not necessarily Na⁺ and K⁺. The Ca/Si ratios were in the C-(A)-S-H range, indicating an intermixing with other phases or that the gels became more and more like a C-(A)-S-H phase. This trend cannot be inferred from the samples of this study. The investigated gels showed maximum Ca/Si ratios of 0.2. However, the different observations may be due to the maturing of the gel and that more and more calcium is incorporated into the gel over time.

• The effect of the ASR gel composition on the concrete expansion

As discussed earlier, the concretes with calcined clays (90PC10MK/MS/MC) showed a significant reduction in expansion compared to the pure PC (100PC) mix. The reduction in ASR-related expansion is due to several mechanisms. Previous studies have demonstrated that the solubility of silicate-rich aggregates is reduced in the presence of aluminum [30,31]. Additionally, factors such as the decrease in pH in the pore solution and changes in the C-(A)-S-H composition are known to impact the ASR-induced expansion in concrete [27], which were not the focus of this study. It is also important to consider that mechanical and physical factors, including concrete strength, porosity, and pore size distribution, significantly affect the ion transport of the concrete and expansion behavior. Thus, while the mechanisms mentioned above are likely the primary driving forces behind the observed reduction in ASR-related expansion (Figure 3), the results of this study indicate a beneficial effect of varying ASR gel chemical compositions, consistent with findings from [17,18,57,58].

The SEM-AM data showed four different gel compositions in the analyzed concretes (Figure 14). No expansion potential is expected for the Si gel, as some authors have reported calcium as a crucial factor in building up considerable swelling pressures in concrete [54,104]. On the other hand, the Ca-Si gel (Figure 14) is expected to lead to deterioration in concrete, as it is within the swellable region, as determined by Mansfeld [54]. All the concretes investigated contained the highest amount of Ca-Si gel (Figure 14). However, as discussed previously, the increased consumption of portlandite by reactive SCMs could result in Ca-poor or non-Ca-bearing gels, which are likely less swellable [104].

Comparing the three calcined clays, the concrete with metasilt exhibited the highest amount of Si gel, compared to metaclay and metakaolin (Figure 14), resulting in less expansive gels. However, the concretes with metaclay and metakaolin exhibited a greater amount of Al-Ca-Si gels compared to the metasilt concrete.

The Al-bearing gels (Al-Ca, Al-Si gel) presumably have a low swelling capacity and do not contribute to concrete expansion. This assumption is supported by earlier studies in which the incorporation of aluminum in synthetic gel led to the suppressed expansion and reduced water uptake of the gel [60,96]. As outlined at the beginning of this chapter, aluminum appears to stabilize within the gel inside the aggregates. The aluminum-bearing gel remains confined within the aggregate and does not migrate into the surrounding cement paste matrix. Consequently, this stabilization is not expected to contribute to any deleterious effects in concrete. In contrast, the Si gel is anticipated to migrate into the cement matrix upon extrusion, where it subsequently incorporates calcium ions and transforms into a Ca-Si gel.

De Souza and Sanchez [100] postulate that SCMs can reduce the formation of cracks in the cement paste and alter the gel chemo-mechanical properties of the ASR products. These findings align with other research studies, which point out that aluminum can alter the composition and structure of ASR products, rendering them non-swellable [17,18,57,93]. The concrete with metakaolin showed the highest amount of Al-Ca-Si and Al-Si gel and the lowest expansion.

The extent to which the Al-bearing gels contribute to ASR mitigation compared to the other mitigation mechanisms discussed needs further investigation. Moreover, the timing of gel characterization is likely a key factor, as gel properties evolve during maturation. Therefore, the potential for Al-bearing gels to change over time in terms of their expansion potential will be further explored in a follow-up study.

4. Conclusions

The effects of various SCMs on the composition, morphology, and structure of ASR gels in concrete was investigated. This study aimed to investigate whether calcined clays can alter the composition of ASR gels and how this affects their expansion properties in concrete.

• The use of 10 wt% calcined clays resulted in a reduction in expansion and minimized cracking in the concrete but did not prevent the formation of ASR gels.
• Investigations by µ-XRF provide insights into the distribution of ASR gels in concrete by examining the K element map. It was observed that the quantity of gel, which was on average approx. 3 wt%, does not directly correlate with the magnitude of expansion. However, as the gel can uptake alkalis, it is relevant to determine the amount of gel per area to generate a volume model. The calculation shows that the gel can take up to 70 wt% of K₂O and nearly all Na₂O from the binder. The present study highlights the potential of µ-XRF as a scanning technique for determining the spatial distribution and alkali uptake of alkali–silica reaction (ASR) gels in concrete. This method enables the monitoring of ASR gel formation and facilitates the scanning of large concrete areas to map gel distribution and alkali uptake across the sample. Furthermore, µ-XRF can be used as a preparatory tool to identify the regions of interest for more detailed analysis using methods such as SEM-EDS or EPMA-WDS.
• Investigations by Raman spectroscopy on the structure of ASR gels showed strong structural similarities for the gels formed in the various concrete mixes. However, varying intensity ratios and the narrowing of the Raman bands might indicate a higher degree of polymerization for the gels in 90PC10MK.
• With SEM-AM analyses, we were able to differentiate the gel into four gel phases according to its composition: Si gel, Ca-Si gel, Al-Ca-Si gel, and Al-Si gel. In contrast, Ca-Si gels formed at the interface with the binder matrix or within the aggregate, and Si gels surrounded by Ca-Si gels were identified in air voids or aggregates. The predominant portion of Al-bearing gels was formed in the concrete with metakaolin (90PC10MK). The Al-containing gel phases were only distributed within the aggregates. The results not only indicate the presence of Al gels in concrete but also show that the aggregate and paste (cement and calcined clay) can affect the composition of the ASR gels not only in terms of Na, K, and Ca but also in the Al concentration. We were the first to show that SEM-AM is an effective method for differentiating compositional variations in amorphous ASR gel phases within concrete.
• The findings of this study suggest that aluminum can be incorporated into gels formed in concrete, leading to changes in the gel structure and potentially reducing their ability to swell, thereby affecting their expansion properties. Therefore, adding aluminum into the system could be beneficial in terms of gel alteration, proving that in practical applications, aluminum-rich SCMs are more efficient in mitigating ASR than siliceous SCMs. The extent of this beneficial mechanism in terms of ASR mitigation needs to be further investigated.