Introduction

Clay-rich soils have poor geotechnical properties, such as high swell and shrinkage potential, high compressibility and low strength [31]. Treatment of clayey soils using lime is an ancient technique [13] and has more recently been supplemented by the use of a variety of cementitious compounds, such as Portland cement and fly ash, as well as enhancing methods including alkali-activation [15]. The fundamental mechanism of the treatment is that these compounds increase the pH of the system and lead to the formation of strength- inducing cementitious hydrates [25]. Despite this being a very old method, it is still largely empirical both because of the highly variable nature of clay soil properties [28] and the limited quantitative understanding of the underlying processes.

Extensive qualitative analysis of the chemical reactions underlying the stabilization process have established the reactions taking place upon lime addition to clay [5, 10, 28, 34] cation exchange and flocculation are the early reactions providing immediate improvement in soil plasticity, followed by carbonation of portlandite (Ca(OH)2) and pozzolanic reactions that form Calcium Alumina Hydrates (CAH) and Calcium Silica Hydrates (CSH) that are responsible for long term strength development, with properties similar to cementitious compounds formed in Portland Cement [13, 32]. In a kaolinite-lime mixture, lime dissolution occurs immediately increasing the pH of the system to 12 establishing alkaline condition in the system. This condition favors clay dissolution with release of alumina and silica in solution and formation of CSH and CAH in the system. The reactions can be written as [8]

$${\text{Ca}}\left( {{\text{OH}}} \right)_{{2}} \left( {\text{s}} \right) \to {\text{Ca}}^{{{2} + }} \left( {{\text{aq}}} \right) \, + {\text{ 2OH}}^{ - } \left( {{\text{aq}}} \right)$$
(1)
$${\text{Al}}_{{2}} {\text{Si}}_{{2}} {\text{O}}_{{5}} \left( {{\text{OH}}} \right)_{{2}} \left( {\text{s}} \right) \, + {\text{ 6OH}}^{ - } \left( {{\text{aq}}} \right) + {\text{3H}}_{{2}} {\text{O}} \to {\text{2Al}}\left( {{\text{OH}}} \right)_{{4}}^{ - } \left( {{\text{aq}}} \right) \, + {\text{ 2H}}_{{3}} {\text{SiO}}_{{4}}^{ - } \left( {{\text{aq}}} \right)$$
(2)
$${1}.{\text{67 Ca}}^{{{2} + }} \left( {{\text{aq}}} \right) + {2}.{\text{34 OH}}^{ - } \left( {{\text{aq}}} \right) + {\text{H}}_{{3}} {\text{SiO}}_{{4}}^{ - } \left( {{\text{aq}}} \right) \to {\text{Ca}}_{{{1}.{67}}} {\text{SiO}}_{{{3}.{67}}} \left( {{\text{H}}_{{2}} {\text{O}}} \right)_{{{2}.{1}}} \left( {\text{s}} \right) + 0.{\text{57 H}}_{{2}} {\text{O}}$$
(3)
$${\text{4Ca}}^{{{2} + }} \left( {{\text{aq}}} \right) + {\text{2Al}}\left( {{\text{OH}}} \right)_{{4}}^{ - } \left( {{\text{aq}}} \right) + {\text{12H}}_{{2}} {\text{O }} + {\text{ 6H}}^{ + } \left( {{\text{aq}}} \right) \to {\text{4CaO}}\cdot{\text{Al}}_{{2}} {\text{O}}_{{3}} \cdot{\text{13H}}_{{2}} {\text{O}}\left( {\text{s}} \right)$$
(4)

Formation of these reaction products induces further dissolution of kaolinite and lime, altering the pH level and shifting the stability to form other hydrates over time.

With the advancement of molecular analysis tools, recent studies have attempted to quantitatively analyze the chemical reactions between clays and chemical stabilizers, and link those to strength increase, both for single-mineral [2, 8, 30, 42] and more complex soils [11, 17, 18]. For example, the linear relationship between dissolution of kaolinite and amorphous content was observed by Chrysochoou [8] from quantitative XRD data up to 360 days of curing. Lime consumption of kaolinite over time was quantified by Maubec et al. [30] using thermogravimetric (TGA) data up to 98 days of curing. A long-term study conducted by Kavak and Baykal [20] over 10 years indicated that there are long term reactions occurring in kaolinite that are not captured by shorter term studies,thus, a full kinetic model of its reaction and the relationship to strength development is still lacking.

Accordingly, the objective of this study is to advance the quantitative understanding between the geochemical evolution of the clay-lime system and the resulting changes in strength, contributing towards the overall goal of developing a fully coupled chemical–mechanical time dependent model of the system. The pozzolanic reactions between kaolinite and lime were studied up to 2 years of curing, suing a combination of spectroscopic techniques, XRD, TGA and nuclear magnetic resonance (NMR). Quantitative analysis of the spectra obtained by each technique were conducted to determine kaolinite dissolution and lime consumption, and the relationship with the changes in Unconfined Compressive Strength (UCS) was explored.

Materials and methods

Materials

Kamin 90 (K90), obtained from KaMin Performance Minerals (average particle size 1.5 μm, as provided by the manufacturer), was used, as in Chrysochoou [8]. Basic characterization of the material including Atterberg limits, hydrometer test data, chemical composition and mineralogy determined by X-ray Diffraction (XRD) are shown in Additional file 1: Table S1, Figures S1 and S2. Briefly, the material is composed of pure crystalline kaolinite, has chemical composition consistent with the stoichiometric ratio of the chemical formula Al2Si2O5 (OH)4 and relatively low plasticity compared to other studies (Table 1). Slaked lime (SL) was used as the stabilizer, containing 95% Ca(OH)2 (ACS grade, Fisher Scientific).

Table 1 Comparison of kaolinite properties in other lime stabilization studies
Full size table

Methods

Standard Proctor compaction curves and 7-day UCS analyses of kaolinite mixed with 5% SL were performed according to ASTM methods [3, 4], respectively, and used to determine the moisture content for long term studies. As discussed in [1]), the results showed that the optimum moisture content (OMC) for compaction of K90 mixed with 5% lime was 40%, while the highest 7-day UCS was observed as a plateau between 28 and 33%. In addition, it was observed that the material at 40% exhibited substantial plasticity and was difficult to compact. A water content of 30% was ultimately chosen to conduct the long term studies to reflect the optimum strength condition.

Samples were prepared in quadruplicates and tested for ten curing times (0, 7, 28, 90, 120, 180, 270, 360, 540 and 720 days). Sample preparation involved standard Proctor compaction according to method [4], followed by extraction from the mold and storage in a plastic sleeve at 95% humidity. At the end of the curing time, the samples underwent unconfined compression strength (UCS) test [3].

A portion of the UCS breaks was used to conduct pH and spectroscopic analyses of the samples. The solvent exchange method [23] was used to dry the samples for spectroscopic analyses, using isopropyl alcohol (IPA) for TGA and XRD samples and acetone for NMR samples.

TGA analyses were performed using an SDT 600 equipment at the Institute of Materials Science at UConn. The sample was heated up to 1000 °C in an argon atmosphere, with a rate of 5 °C/min. Processing of the TGA spectra was performed using the Universal Analysis software.

The XRD analyses were performed with a Bruker D2 Phaser diffractometer, with slit size of 1 × 10 mm and Cukα radiation (λ = 1.54Å) in the two-theta range 5–65° with a step of 0.01° 2θ and a 0.5 s counting time. Corundum (α-Al2O3) was used as internal standard for two-theta calibration and quantitative analysis; 0.2 g of corundum was mixed with 0.8 g of the sample. Previous analysis of the corundum obtained from Sawyer LLC using a NIST ZnO standard indicated 93% crystallinity [12]. The Jade software v. 8.5 (Materials Data Inc.) was used for qualitative and quantitative analysis of the diffractograms. Quantitative analysis was performed using the Rietveld method [36], using structural files obtained from the MDI crystal structure database and American Mineralogist Crystal Structure Database [14]. Additional details on the quantitative analysis are provided in Supporting Information.

For NMR analysis, a Bruker Advance III 400 MHz solid state NMR spectrometer was used with the magic angle spinning (MAS) technique. Two elements- 29Si and 27Al were analyzed, with spectrometer frequency 79 and 107 MHz respectively, with a recycle delay of 30 s. The chemical shifts in the samples were referred to tri(trimethyl-silyl)silyl (TTMS) during experiment and were corrected later to compare the data with literature where trimethylsilane (TMS) is widely used as reference [19, 26]. The NMR spectra were analyzed using the OriginPro software.

Results

Long-term strength test results

The results of UCS tests up to 720 days of curing compared to other studies on pure kaolinite soils are shown in Fig. 1, while Table 1 summarizes conditions and results from the literature.

Fig. 1
figure 1

UCS of kaolinite treated with 5% SL, compared to studies summarized in Table 1

Full size image

Kaolinite strength shows steady increase from 250 to 685 kPa up to 360 days, followed by a small decrease to 590 kPa at 720 days. The observed trend exhibited slower increase and a lower plateau compared to the 2014 study at 36% water content; the only other difference between the two studies was the use of slaked lime versus quicklime. Kavak and Baykal [20] reported an even faster increase in UCS with time, with similar experimental conditions and baseline UCS of the untreated material; the source material had a higher dry unit weight and was slightly coarser. The observed trends in this study were very similar to the results of [30] at 20 °C up to 98 days of curing. [20] showed substantial increases at 8, 9 and 10 years of curing, with no data points between 300 and 2920 days.

Comparing the material properties and experimental conditions of Table 1, it appears that dry unit weight is the only parameter that correlates with the observed strength trends; [20] had the highest γd at 14 kN/m3, followed by [8] at 12.4 kN/m3 and the two studies with the same behavior at 11.4 kN/m3 ([30] and this study). Note that the water contents of compaction in the different studies do not follow the dry unit weight trend, so that the observed trend is not related to the amount of water present in the system. The dry unit weight of the cured samples did not change as a function of time; water contents were analyzed post-curing and were generally found to be between 29 and 31% with minimal loss during storage; dry unit weight fluctuated between 11 and 11.6 kN/m3. To further investigate the underlying mechanisms for the observed strength behavior, solid phase analyses were utilized, which were also partially available in [8, 30].

TGA results

Evaluation of the TGA data was performed using the mass loss and first derivative spectra as shown in Fig. 2. The observed peaks were compared to the spectrum of pure kaolinite (Additional file 1: Figure S3), pure hydrated lime and hydrated cement (Additional file 1: Figure S4), along with literature references. The kaolinite spectrum shows a single dehydroxylation peak between 400 and 560 °C, similar to the range 460–660 °C reported by [30] for kaolinite and muscovite combined. The spectrum of hydrated lime showed peaks in three regions: the initial mass loss of 5% up to 100 °C is due to free water; the peak in the temperature range from 350 to 410 °C corresponds to portlandite (Ca(OH)2) dehydroxylation [44] and mass loss from 500 to 760 °C is the range where the different polymorphs of calcium carbonate (CaCO3) release CO2 [9]. Similar ranges are observed for these transitions in the hydrated cement spectrum, with additional peaks in the 100–200 °C range corresponding to CSH and ettringite. Klimesch and Ray [21] reported CSH peaks in TGA spectra centered around 150 °C, and [37] reported a range of 150–200 °C for release of highly-bound water in CSH. Collier [9] gave a range of 100 to 125 °C for CSH in most reported studies. Steiner et al. [39] reported a peak at 140 °C for synthesized ettringite, which decomposed to hemihydrate (150 °C) and anhydrite (165 °C) upon carbonation.

Fig. 2
figure 2

TGA of as cured K90 mixed with 5% SL

Full size image

The TGA spectrum of kaolinite mixed with 5% SL (Fig. 2) shows that there is an overlap between the portlandite and kaolinite dehydroxylation peaks, which is resolved in the derivative spectrum. Based on the pure and combined spectra, four regions were chosen for quantitative analysis of different phases: 100–350 °C corresponding to CSH, ettringite and other hydrates such as CAH, CSAH and monocarboaluminates [40]; 350–410 °C corresponding to portlandite; 410–560 °C for kaolinite and 560–750 °C for CaCO3. The mass loss for these four temperature intervals over time is shown in Fig. 3, and representative spectra for each time frame are shown in Additional file 1: Figure S5.

Fig. 3
figure 3

Mass loss of TGA spectra in four temperature regions as a function of time—dashed lines indicate value in untreated kaolinite spectrum

Full size image

The mass loss in the kaolinite temperature region (410–560 °C) showed little change up to 180 days of curing, followed by a 12% progressive decrease up to 720 days. More distinctive changes were observed in the temperature ranges corresponding to portlandite and cementitious hydrates. The portlandite region (350–410 °C) had clear peaks in the derivative spectra (Additional file 1: Figure S5) up to 90 days curing, which turned into a shoulder at 120 and 180 days, and was integrated into the kaolinite peak at later times. The mass losses above the “background” value of untreated kaolinite at times after 270 days of curing cannot thus be solely attributed to portlandite dehydroxylation.

The decrease in the portlandite peaks over time was accompanied by an increase in the region of the cementitious hydrates. In addition to the overall mass loss, distinctive peaks emerged in the first derivative spectra at 130 °C and 190 °C at all curing times past 180 days. The presence of two peaks may imply the formation of two distinctive hydration products, and there are multiple phases that can account for the observed peaks. Comparison of the temperatures with the literature points to CSH, CAH and ettringite for 130 °C and certain CAH and stratlingite for 190 °C [9]. Ettringite is a typical sulfate-bearing compound of hydrated cement, while stratlingite (Ca2Al2SiO7∙8H2O) is a siliceous AFm-type phase that is associated with blended cements with kaolin, silica fume and fly ash [33]. The two peaks were further separated by evaluating the mass loss in two regions 100 to 160 °C and 160 to 350 °C (Additional file 1: Figure S6). The peak at 190 °C follows the overall increase trend up to 720 days; the peak at 130 °C appears to stabilize at 270 days, indicating that there are likely two phases accounting for each of the peaks.

Finally, the CaCO3 region exhibits an oscillating behavior, increasing up to 180 days, then decreasing to 360; both 540 and 720 data points exhibited too highly variability to determine a trend. The initial increase coincides with the decrease in the portlandite mass loss, while the decrease then coincides with the increase in the CAH/CSH compounds.

Comparing the average mass loss in the 100 to 350 °C region with the average UCS values of Fig. 1, there is a linear correlation with R2 of 0.81 (Additional file 1: Figure S7). Thus, strength increase appears to be linked to the formation of the two cementitious phases observed in the TGA spectrum. These trends and the nature of the newly formed phases were further investigated with XRD.

XRD results

The qualitative analysis results of the XRD patterns over time showed few changes in the crystalline phases (Fig. 4): portlandite was observed up with diminishing intensity up to 180 days of curing. Calcite and stratlingite were identified as new crystalline phases formed. No sulfate-bearing or CAH phases were observed in any of the patterns. Based on the SO3 concentration of 0.18% (Additional file 1: Table S1), the maximum concentration of ettringite in the solid would be 0.94%, which is approximately the XRD detection limit. It is, therefore, possible, that small amounts of ettringite could not be detected with this method. In addition, monosulfualuminate is likely thermodynamically stable over ettringite at this low sulfate concentration as shown by stability diagrams in various clay soils [27].

Fig. 4
figure 4

XRD patterns of K90 with 5SL at 0, 180, 270 and 540 days of curing

Full size image

QXRD analysis of long-term K90 samples are shown in Fig. 5. Given the quality of the fit that we were able to achieve using published structures, which was consistent across all patterns, the results are interpreted in a semi-quantitative fashion, i.e. as relative trends over time.

Fig. 5
figure 5

QXRD analysis results for all identified phases

Full size image

Similar to TGA, the quantitative results for kaolinite show a slow, limited (7% relative) decrease in content from 360 to 720 days. Accordingly, there is little change in the amorphous content of the sample. Portlandite decreases with time at the same rate as the TGA results. Also similar to the TGA, the calcite content shows a fluctuation with increase up to 180 days, followed by decrease to 270 and increase again at 540 days; interestingly, no calcite was detected in the XRD patterns of 720 days. Stratlingite could only be quantified in the last four time frames and the observed trend generally followed the TGA results (increase from 270 to 360, small decrease at 540, increase at 720 days). The fact that the quantitative trends in all phases are similar between TGA and XRD, which rely on different principles and methods of analysis, indicates that they reflect real phase changes in the samples.

NMR analysis

Additional file 1: Figure S7a shows the 29Si NMR spectrum of pure kaolinite, with the Q3 peak of tetrahedrally coordinated Si observed at -91 ppm. A shoulder at − 88 ppm is attributed to distortion from Al substitution in the tetrahedral site as previously observed for kaolinite [6]. An additional shoulder at − 86 to − 87 ppm emerges in the treated spectra after 90 days of curing (Fig. 6), which corresponds to the Q2 peak representing middle group Si chains that form calcium silicate hydrates including stratlingite [24]. The evolution of the Q2 and Q3 phases was evaluated by integrating the peak areas with OriginPro using a Gaussian–Lorentzian function that provides better fits for NMR spectra [29]. As seen in Fig. 7, the peak area of the Q3 peak decreases slightly at 540 days of curing, while the Q2 peak increases in the same time frame. These results are consistent with both the TGA and XRD semi-quantitative analysis results

Fig. 6.
figure 6

29Si (a) and 27Al (b) NMR analysis of K90 with 5SL

Full size image
Fig. 7
figure 7

Integrated peak areas of Q3 (kaolinite) and Q2 (silica hydrates) peaks in 29Si

Full size image

The nature of the reaction products was further investigated using 27Al NMR analysis. The pure K90 spectrum is shown in Additional file 1: Figure S7b, showing the main octahedral Al peak at − 7.9 ppm and Al substituted in the Si tetrahedral sites at 70 ppm. There is a shoulder in between the peaks at 67 ppm which indicates tetrahedral Si. Starting from the 180-day-cured sample, a new peak emerges at 50 ppm which is tetrahedrally coordinated Al, corresponding to the stratlingite structure [24]. The peak starts to increase at 270 days and decreases at 540 days, similar to XRD and TGA and the 29Si NMR observations. Similar to XRD, NMR also does not show any peak for ettringite (supposed peak location 13–14 ppm [38]) or monosulfoaluminate (11.5 ppm, [35]), possibly due to their smaller concentrations.

Discussion

Similar to [8, 20], compressive strength initially increased, doubling between 0 and 360 days of curing, this was followed by a 14% loss by 720 days, a result that has not been previously observed. The dry unit weight and post-curing water content of the 540- and 720-day samples were similar to all previous samples, so that loss in strength was not related to the compaction process or sample storage. Chemical reactions and changes in mineralogy are therefore examined to explain this behavior.

First, the evolution of calcium-bearing phases in the system was assessed. A quantitative evaluation of portlandite consumption using the TGA data can be performed using the following equation:

$${\mathrm{\%Ca}({\text{OH}})}_{2}=\mathrm{\%}{{\text{ML}}}_{({\text{Ca}}\left({\text{OH}}\right)2)}\times (\frac{{M}_{Ca\left(OH\right)2}}{{M}_{H2O}})$$

Where MCa(OH)2 and MH2O are the molar masses of Ca(OH)2 and H2O respectively, %MLCa(OH)2 represents the mass loss due to water as measured by TGA and %Ca(OH)2 represents the mass loss due to Ca(OH)2. For this calculation, the average net TGA signal was taken from Fig. 3, i.e. the baseline mass loss value from the pure kaolinite curve was subtracted from the cured spectrum curve. Figure 8 shows the calculated results compared to the portlandite concentration determined by XRD quantification. The two methods are in perfect agreement up to 180 days of curing; after that, portlandite was no longer detected in the XRD patterns. For TGA, the increase in the signal at 360 and 540 days causes the two methods to deviate in terms of the overall fit. A logarithmic fit to the XRD data yields 300 days of curing for complete portlandite consumption, while the TGA curve yields 500 days if all data points are included. As discussed in the respective section, these points are unlikely to be due to actual portlandite presence and more likely due to the kaolinite peak broadening. The solid pH data (Additional file 1: Figure S10) showed that the pH dropped below 12.4 by 270 days, which is the equilibrium pH of a saturated portlandite solution. This analysis, along with the strength data that show a peak at 360 days, would converge towards the conclusion that upon addition of 5% hydrated lime, reaction of portlandite with kaolinite to produce cementitious hydrates and increase strength takes approximately one year. The only other study that produced similar data was [30], these authors added 10% quicklime to pure kaolinite and detected no decreasing trend of portlandite at 20 °C up to 98 days of curing, while consumption was faster than observed in this study at 50 °C; based on a logarithmic fit, all 10% would be consumed within 120 days at that temperature.

Fig. 8
figure 8

Lime consumption curves based on TGA and XRD data

Full size image

Both TGA and XRD only yielded 2% of portlandite as the initial concentrations observed in the as-cured sample, compared to 5% added. While both methods have limitations in terms of absolute accuracy, the agreement between the two raises a question on the fate of the remaining calcium in the system. There are three potential phases for Ca to be distributed: dissolved, sorbed on the kaolinite surface and precipitated as solid. In addition to the spectroscopic analyses, pore water was extracted from the cured samples and analyzed for all dissolved elements (data not shown); the Ca concentration was very low throughout and could only account for 0.05% of the total amount added. Similarly, a Visual Minteq analysis using Ca adsorption data on kaolinite from previous studies [7, 22] showed that the sorbed amount is also very low, about 1.45% of the total added. Thus, precipitated phases account for most of the Ca added. A mass balance on the Ca added and the Ca contributed by the different phases identified through XRD and TGA was conducted, shown in Fig. 9. Details on the calculations and figures for individual phases are provided in Additional file 1.

Fig. 9
figure 9

Total Ca in the solid calculated from TGA and XRD

Full size image

TGA yields a higher measurable content of calcium at all times, which is due to the quantification of the hydrates (attributed to the 130 °C peak) which are amorphous to XRD. As shown in Additional file 1: Figure S9, the amount of Ca contributed by portlandite, calcite and stratlingite is mostly in agreement between the two methods. Both TGA and XRD show a decrease in the total Ca in the first 180 days, driven by the decrease in portlandite content. TGA showed an increase in the amorphous hydrate content that appeared as distinct peaks in the first derivative (Additional file 1: Figure S5) and continued to increase up to 360 days, along with the strength and the total Ca content, which was 91% of the added Ca at that time. The Ca content tapered off, with a slight decrease by 720 days, as the additional hydrate formation does not offset the loss of portlandite.

The XRD trends are driven by the decrease in portlandite up to 270 days, followed by an increase in stratlingite until 720 days; the total quantifiable amount is below the TGA and the total amount of Ca at all times. This clearly indicates the presence of Ca-bearing phases that are amorphous to XRD. Taken these results in combination, the following potential explanation is put forward: there are two phases of kaolinite dissolution, the first one being incongruent dissolution with preferential release of either Si or Al in the first 360 days when portlandite is consumed, and a second one with congruent dissolution releasing Al or Si respectively thereafter. Preferentially release of Si results in amorphous CSH formation, while preferential release of Al results in AFm (monosulfate or monocarboaluminate) phase formation, the primary phase that is responsible for strength increase. After free lime has been consumed, the formation of stratlingite upon congruent dissolution of kaolinite occurs at the expense of other hydration products, disturbing the cementitious matrix and resulting in strength decrease.

The survey of the literature in terms of kaolinite dissolution did not yield conclusive results in terms of which element is preferentially released at early times. Xie and Walther [43] reported preferential Al release at pH < 4, preferential Si release at 4 < pH < 11 and stoichiometric Al and Si release above pH 11. Garg and Skibsted [16] also reported congruent dissolution at high pH for metakaolin calcined at 500 °C, which retained a structure closer to the original kaolinite compared to higher temperatures. However, [7] showed a preferential release of Al over Si in kaolinite in the first 24 h of reaction with lime. A similar study [41] on kaolinite containing small amounts of muscovite and stabilized with 5% quick lime observed CAH formation in the system after 28 days and C4ACH11 after 270 days. These authors also suggested preferential release of Al over Si in the system. An ongoing study on the time-dependent release of ions in the pore solution of the system will shed further light into the exact mechanism of kaolinite dissolution.

Conclusions

The long-term (up to 2 years) relationship between the pozzolanic reactions and Unconfined Compressive Strength development in lime-treated kaolinite was investigated in this study using three spectroscopic techniques, XRD, TGA and NMR. The key findings of the study are:

  1. 1.

    To understand pozzolanic reaction in kaolinite the curing time must be longer than 28 days and even studies up to 90 days are also not sufficient to observe the strength development pattern in kaolinite.

  2. 2.

    Strength increase with 5% hydrated lime was continuous and doubled up to 360 days, followed by a 14% decrease by 720 days.

  3. 3.

    Portlandite consumption was complete by 270 days of curing and is accompanied by an increase in the hydration product (CAH/CSH) region of the TGA spectra. No reaction products were observed by either XRD or NMR at that time.

  4. 4.

    Stratlingite (CASH) is observed in both XRD and NMR after 270 days of curing and continued to form up to 720 days.

The working hypothesis to explain these observations is that kaolinite dissolution follows two phases, largely related to the presence of free lime: the first phase, lasting approximately one year, is characterized by incongruent dissolution of kaolinite with preferential release of one of the main components (Al or Si), leading to the formation of CAH or CSH and strength increase. After portlandite consumption is complete, kaolinite dissolution becomes congruent, resulting in the formation of stratlingite, which uses the originally formed CAH/CSH as Ca source. This conversion has an adverse effect on the strength of the stabilized material. Literature data point to a preferential release of Al over Si in the early stages of reaction. Ongoing studies of the pore solution will shed further light into this hypothesis.