The kaolinite content is a key criterion for the selection of suitable clay. The kaolinite content can be determined by thermogravimetric analysis (TGA) using the tangent method.7 The dehydroxylation of kaolinite AS₂H₂ occurs from about 400 to 650°C.8–10 This leads to the formation of the amorphous metakaolin phase AS₂ according to the equation

$${\text{AS}}_2{\text{H}}_2\to {\text{AS}}_2+2\text{H}$$

1

The water loss during kaolinite dehydroxylation, wt%_kaol-OH, allows the determination of the kaolinite content wt%_kaolinite according to Equation 2. M _kaolinite and M _water refer to the molecular weights of kaolinite and water, respectively.

$${\text{wt}\%}_{\text{kaolinite}}={\text{wt}\%}_{\text{kaol-OH}}\frac{M_{\text{kaolinite}}}{2M_{\text{water}}}$$

2

TGA is also carried out on calcined clays to detect any water loss wt%_{kaol-OH, calcined} corresponding to an incomplete calcination. The calcined kaolinite content is then obtained according to the equation

$${\text{wt}\%}_{\text{calcined kaolinite}}=\frac{M_{\text{kaolinite}}}{2M_{\text{water}}}\left({\text{wt}\%}_{\text{kaol-OH}}-{\text{wt}\%}_{\text{kaol-OH, calcined}}\times \frac{100-{\text{wt}\%}_{\text{kaol-OH, calcined}}}{100}\right)$$

3

The influence of secondary phases on the determination of the kaolinite content was investigated to ensure that they do not significantly impact the kaolinite content quantification. Various phases lose water or carbon dioxide during calcination, but most of them do not interfere with kaolinite dehydroxylation. Among the thermally decomposing impurities most frequently identified in kaolinitic clays, goethite and gibbsite dehydroxylate at a lower temperature than kaolinite, at about 250–350°C.11–13 Calcite decarbonation occurs at a higher temperature, about 700–800°C. The presence of 2:1 clays with two tetrahedral silicate sheets and a single octahedral aluminate sheet (illite, muscovite, montmorillonite) was studied more deeply since they partially dehydroxylate over the temperature range of kaolinite dehydroxylation,14,15 as shown in Figure 1.

Figure 1 TGA of clays rich in kaolinite, illite, muscovite and montmorillonite

Since illite is the main phase interfering with kaolinite, mixes of these two clays were prepared; 25, 50 and 75% of kaolinitic clay were substituted by illitic clay. The TGA and differential thermal gravimetric (DTG) curves of these mixes are shown in Figure 2. The loss of water increases with the kaolinite fraction of the mix because kaolinite has a higher water mass fraction than illite.

Figure 2 TGA and DTG curves of the mixes of kaolinitic and illitic clays

The kaolinite content was then determined on these binary mixes using the tangent method. The comparison of the kaolinite content measured by TGA and the theoretical content is shown in Figure 3. The 1:1 line is also plotted to simulate a perfect fit. Very close results are obtained. The maximum deviation between the measured and the theoretical values is obtained for the mix containing 75% illitic clay, with 2·7% difference. Thus, even in the presence of a significant amount of illite, the kaolinite content can still be accurately determined.

Figure 3 Correlation between the measured kaolinite content by TGA and the theoretical kaolinite content

Concerning the reproducibility that TGA provides, more than 50 different kaolinitic clays were tested in LC³. TGA was carried out three times for each clay. An average deviation of 1·0% was obtained for the kaolinite content. Therefore, it is concluded that TGA provides good reliability and reproducibility for the determination of the kaolinite content.

Another common way of quantifying the kaolinite content is the use of X-ray diffraction (XRD–Rietveld). However, the quantification of kaolinite by XRD is very challenging due to the structure of kaolinite, in which extensive layer disorder and preferred orientation are observed. XRD–Rietveld was carried out on 40 clays. The kaolinite content is compared with the values obtained by TGA in Figure 4. It shows that a general trend following the 1:1 line is obtained. However, some significant difference can also be observed. The maximal deviation obtained is 29·8%, and the average absolute difference is 6·1%. TGA is certainly preferable to XRD for such a purpose.

Figure 4 Comparison of kaolinite content by XRD–Rietveld and TGA

The dehydroxylation of kaolinite leads to the formation of metakaolin, and the highly disordered structure of metakaolin combined with the presence of penta-coordinated Al(V) are responsible for its reactivity. The calcination temperature of kaolinitic clays was investigated to optimize the reactivity of calcined clay. Batches of a clay with 50% kaolinite were calcined in an oven at five different calcination temperatures from 600 to 850°C. The TGA and DTG curves in Figure 5 show that kaolinite is partially dehydroxylated at 600°C and the dehydroxylation is complete from 700°C onward.

Figure 5 TGA and DTG curves of a kaolinitic clay calcined at 600, 700, 750, 800 and 850°C

TGA indicates the efficiency of calcination but does not provide any information on the reactivity of calcined clay. The R³ pozzolanic test was developed as reactivity indicator of the calcined clay.2 This test also gives a good prediction of mortar strength, even at late ages. The R³ pozzolanic test consists of assessing the reactivity of calcined clays in model systems composed of portlandite, calcined clay and limestone with levels of sulfate and alkali adjusted to simulate the conditions in real blended cements. Two methods are possible to assess the reactivity. The monitoring of the heat release by isothermal calorimetry at 40°C is the most rapid way (24 h of testing), also giving the most relevant prediction of mortar strength. The determination of the bound water takes slightly more time but does not require any specific equipment other than an oven. Complete details of the test are given in the paper by Avet et al. 2

The R³ test was applied on the batches calcined at different temperatures. The cumulative heat release is shown in Figure 6. It shows that the highest heat after 24 h of testing is obtained for the clay calcined at 800°C, with a slightly lower reactivity for the clay calcined at 700 and 750°C. A similar ranking was obtained using the bound water alternative and mortar strength results, even for strength variations of only 10% between 700 and 850°C. The optimal pozzolanic reactivity is thus obtained 100°C higher than the completion of dehydroxylation. The reason for the lower reactivity at 600°C is the incomplete dehydroxylation of kaolinite. At 850°C, the decrease in reactivity is due to the physical coarsening and agglomeration of calcined clay particles.16

Figure 6 R³ cumulative heat release for the raw clay and the clay calcined at 600, 700, 750, 800 and 850°C

The sulfate content in cement is adjusted based on the aluminate content from the clinker. In LC³ systems, part of clinker is substituted by calcined clay containing an aluminate-rich metakaolin phase. Thus, the sulfate content used for a properly sulfated PC cannot be simply diluted when using LC³. A new sulfate optimization is required. An example is shown in Figure 8, where different gypsum additions (in addition to the amount in the cement) from 0 to 3% are used for an LC³-50 containing a calcined clay with 50% calcined kaolinite. A 2:1 calcined clay-to-limestone mass ratio is used in this study (30 parts of calcined clay for 15 parts of limestone). It shows that without adjustment, silicate and aluminate peaks occur close to each other, leading to a lower total heat release and reactivity, whereas with a proper gypsum addition of 1%, it is possible to differentiate the silicate from the aluminate peak. This ensures the optimal properties of the blended cement.23 In this case, the total sulfur trioxide (SO₃) content of the LC³ system is actually lower than that of PC (1·9% for LC³-50 against 2·6% for reference PC).

Figure 8 Heat flow of LC³-50 with different gypsum additions

Alkali content is known to affect hydration reactions, mechanical properties and durability properties of concrete significantly. It is accepted that alkalis accelerate the early hydration of cement,24 while resulting in reduced hydration and strength at later ages.25 A certain level of alkalinity is also necessary to favor pozzolanic reaction.26,27 Figure 9 shows the heat flow for an LC³-50 system with a clay containing 50% kaolinite and varying alkali contents from 0·3 to 0·6% sodium oxide equivalent (Na₂O_eq). It is clearly seen that the increase in alkali content accelerates the reaction of silicate and aluminate phases. The optimum for LC³ consists of providing enough alkalinity to enhance the pozzolanic reaction without significantly impairing the properties at late ages. The optimum in this case is 0·48% Na₂O_eq.

Figure 9 Heat flow of LC³-50 with various Na₂O_eq contents

During hydration of LC³, additional reactions occur compared with PC. Metakaolin in calcined clay reacts as pozzolanic material, consuming portlandite and forming mainly calcium aluminum silicate hydrate (C-A-S-H).31 Limestone also reacts with the C₃A from clinker to form carboaluminate hydrates.32 In LC³, the formation of carboaluminate hydrates is also enhanced thanks to the reaction of aluminate from metakaolin.23 In order to characterize the phase assemblage evolution in LC³, clinker hydration degree can be easily quantified using XRD, as well as the amount of reacted limestone. The determination of the reaction degree of metakaolin is more challenging. A comparative study was carried out to determine the best method.33 For the various clays tested, the mass balance approach is the most suitable method. It consists of calculating the phase assemblage forming based on the consumption of anhydrous phases. The amount of reacted metakaolin is unknown and considered as the free variable of the system. It is varied from 0 to 100%, and the actual amount is determined from the best agreement with experimental data. The results obtained for the amount of reacted metakaolin for LC³-50 blends is shown in Figure 11 for various calcined kaolinitic clays. The metakaolin content of each calcined clay is indicated for each system. It shows that the amount of reacted metakaolin increases with time and globally increases at late ages with the grade of the calcined clay. However, this increase becomes less important with the metakaolin content of calcined clay.

Figure 11 Amount of reacted metakaolin in calcined clay in LC³-50 systems33

From these data, the whole phase assemblage of LC³-50 can be characterized, as shown in Figure 12 at 3 and 28 d of hydration. Globally, the phase assemblage does not show significant differences from PC. Similar hydration products form. Carboaluminate hydrates are the main AFm phases in LC³; this is similar to most modern PCs, which contain small amounts of limestone (<5%). Monosulfoaluminate is only formed in PCs with no limestone addition.34 The highest amount of C-A-S-H is found in PC, due to the higher initial clinker content. The pH of the pore solution is very similar to that of PC, from 13·2 for LC³-50 (95·0%) with fairly pure metakaolin to 13·8 for LC³-50 (0%) containing quartz.

Figure 12 Phase assemblage of PC and LC³-50 (50·3%) at 3 and 28 d of hydration. CH, portlandite; Ett, ettringite; Hc, hemicarboaluminate; MK, metakaolin; Ms, monosulfoaluminate

A benchmark test on compressive strength was carried out on standard mortar, using LC³-50. More than 50 clays were tested, with their calcined kaolinite content ranging from 0 up to 95% at ages of 1, 3, 7, 28 and 90 d.35 Results are presented in Figure 13, showing the dependency of compressive strength on calcined kaolinite content. At 1 d, mechanical strength is only slightly affected by the clay grade. This is expected, as the pozzolanic reaction of metakaolin in calcined clay is just starting. Between 3 and 28 d of hydration, the positive effect of clay grade on the compressive strength is clearly visible, as strength is almost linearly correlated with the calcined kaolinite content. This is the key finding of this benchmark test, as it demonstrates that – for the LC³-50 design – strength differences are a first-order function of the calcined kaolinite content, independent of the secondary clay phases. Those phases can be anything from metastable clay minerals, such as metaillite or inert phases, such as quartz or hematite. In addition, the dashed horizontal lines represent the PC strength at different ages – from 3 to 90 d. Interestingly, LC³ using very high-grade clays can catch up with PC after only 3 d and clays with a calcined kaolinite content as low as 40% are also able to reach PC strength at 28 d. The blends used in this test were produced by simple mixing of the calcined clay and limestone with PC. It is highly likely the early-age properties could be further improved by optimizing the grinding and blending process.

Figure 13 Compressive strength evolution over time for LC³-50 mixes with clays of varying calcined kaolinite contents. The solid lines are merely guides for the reader, illustrating the low spread of the experimental data. The dashed horizontal lines show the PC strength at age varying from 3 to 90 d35

It is also worth noting that clays with a calcined kaolinite content from about 40 to 75% exhibit the highest strength gain between 7 and 28 d. This is also valid from 28 to 90 d, as the highest-grade clays (>80%) already reached their maximal strength after 1 month and the lowest-grade clays (<30%) have probably depleted their metakaolin. Thus, there is no major benefit of using clays with more than 50–60% calcined kaolinite since the strength gain is not significant at late ages.

Some clays were also selected for a testing campaign on flexural strength, on standard mortar as well. The trends in flexural strength evolution follow those of compressive strength; see Figure 14. The ratio of compressive strength to flexural strength is very similar between PC and LC³-50 blends.

Figure 14 Flexural strength as a function of compressive strength for five LC³-50 mixes compared to a reference PC

The evolution of Young’s modulus against compressive strength was also monitored on mortar samples, through non-destructive compressive measurements; these are shown in Figure 15. A similar trend is observed for PC and LC³-50 blends.

Figure 15 Young’s modulus as a function of compressive strength for five LC³-50 mixes compared to a reference PC

Uniaxial compressive creep was assessed on mature paste samples in sealed conditions, for LC³-50 binders and a PC reference. Samples were loaded after 28 d (t ₀) of sealed curing at a constant stress corresponding to 15% of their strength, for a duration of 28 d. This level of loading ensures being in the linear regime between strain and applied stress.36 The results are reported in Figure 16, shown as the basic creep compliance over loading time. Creep compliance is obtained by subtracting the elastic strain at loading and the autogenous shrinkage strain (see the following section for details) from the total measured strain and then dividing this strain by the applied load. The most striking feature is the sharp reduction in the creep compliance of LC³ binders compared to that of the PC reference. This is consistent with previous results from the study of Brooks and Johari,37 although that study focused on concrete and clinker substitution by metakaolin only. This decrease seems independent of the clay purity – at least within the studied range – and could be attributed to both a refined pore structure and a lower amount of the viscous C-A-S-H phase in LC³.

Figure 16 Creep compliance plotted against loading time on mature paste samples (28 d of sealed curing). The use of calcined clay causes an important reduction in compressive creep compliance

The autogenous shrinkage of paste specimens was measured according to the ASTM C 1698-0938 standard in an oil bath kept at 20 ± 0·1°C. LC³-50 mixes containing calcined clays with a metakaolin content ranging from 25 to 60% were compared to a PC reference. Results in Figure 17 show the autogenous shrinkage over the first month of hydration. Curves are zeroed after the initial expansion to display only the pure shrinkage part. Note that this initial expansion was more important for the PC reference than the LC³ samples. All ternary blends show a delay of the shrinkage onset, postponing it to around 4 d for the studied systems. This is a valuable feature as shrinkage strains seem to be limited at young age, when the strength of the material is still low and therefore subject to cracking. After 28 d of hydration, the total autogenous shrinkage amplitude is similar between PC and LC³ mixes.

Figure 17 Autogenous shrinkage over the first month of hydration for a set of LC³-50 with clay of varying purities. Curves are zeroed after the initial expansion. All ternary blends exhibit a delay of about 4 d on the initial shrinkage onset

Corrosion of steel reinforcement in concrete due to chemical attack of chloride ions (from seawater or deicing salts) is the most important durability concern of reinforced-concrete structures worldwide. Diffusion of chloride ions through concrete is mainly governed by the porosity characteristics of concrete,39 as well as the phase assemblage of the binder.40,41 Figure 18 shows the profile of chloride ions in a one-directional ponding experiment in 3 wt.% sodium chloride (NaCl) solution over 2 years for PC and three LC³ blends, all prepared with a same clay of 50% kaolinite content. In LC³-65, the substitution level is lower with a clinker content of 65%. The calcined clay-to-limestone ratio was also changed from 2:1 to 1:1. The chloride profile in LC³-50 blends indicates significant improvement with respect to chloride ion diffusion compared to the systems with higher clinker content. A higher calcined clay-to-limestone ratio tends to reduce further the penetration depth of chloride. Furthermore, the authors’ recent analysis has shown that the performance of the LC³-50 blend is better than or similar to that of ternary blends prepared with fly ash or slag (instead of calcined clay) at the same clinker replacement levels. Thus, significant improvement in the service life of LC³-50 concrete structures in saline environments is expected.

Figure 18 Total chloride profiles in LC³ and PC mortars, exposed to 3 wt.% sodium chloride solution for 2 years

The main reason for such excellent chloride resistance is explained in Figure 19 by the significant refinement of pore connectivity. Even for a calcined kaolinite content of 38·9%, the critical pore entry radius is three times smaller than that for PC at 28 d of hydration.

Figure 19 Porosity distribution of PC and LC³-50 pastes at 28 d of hydration, prepared with clays of various kaolinite contents

The carbonation rate of concrete with blended cements is generally known to be higher than that of PC, and this is also the case for LC³ systems.42 The rate of carbonation is mainly controlled by the calcium oxide (CaO) content of concrete, which is the main component binding carbon dioxide from the atmosphere in concrete. Figure 20 shows how the initial calcium oxide content of concrete determines the rate of carbonation. Portland pozzolanic cement blends were composed of binary PC–calcined clay, PC–fly ash or PC–slag systems. Calcined clays with higher purity have slightly better carbonation resistance than low-grade clays and comparable to other conventional SCMs, such as fly ash and slag.

Figure 20 Carbonation coefficients for PC, LC³-50 and binary PC–SCM mortars, measured in the natural environment

The carbonation resistance of LC³, and other blended systems, can be improved by good curing of concrete prior to exposure, as seen in Figure 21. Also, it should be noted that carbonation is highly related to the percentage of relative humidity (RH%) of the environment. At very low or very high humidity, the carbonation rate is slow. As such, at locations where LC³ has the highest potential of use, such as tropical countries, or in marine structures where the RH is generally high, carbonation will not be a concern.

Figure 21 LC³-50 and PC mortars carbonated in an outdoor sheltered environment for 18 months

For the study of the alkali–silica reaction (ASR), the reactive aggregate (Jobe) was used. The clay tested in LC³-65 and LC³-50 contains 50% calcined kaolinite. The mortar bars after curing in fog condition for 28 d were soaked in 0·32 M sodium hydroxide (NaOH) solutions at 38°C. Generally, the use of SCMs has an effective preventive effect against ASR in concrete43,44 due to the lower alkalinity and the presence of aluminum (Al) in pore solution.45 Figure 22 shows that LC³ is extremely promising for mitigating ASR. As previously observed for chloride resistance, even better results are obtained with increasing the clinker substitution level.

Figure 22 Expansion measured for PC and LC³-65 and LC³-50 with a calcined kaolinite content of 50% in the calcined clay