The specimens were developed according to the type of test: modified Proctor, CBR, simple compression or tension by splitting. Thus, after sampling, heaps of 6 kg of lateritic clay are mixed with quantities of water with different mass percentages. Homogenize before compacting. For the CBR test, a clay-water mixture is made taking into account the optimum moisture content and the maximum dry density given by the modified Proctor test. The prepared specimen is kept in the mould, immersed directly in a water tank for four days or cured in the open air for three days before immersion. For the specimens to be subjected to simple compression and to tension by splitting, the confection is of the modified Proctor type in a split Proctor mould. A compaction of 56 strokes on 5 layers is made, immediately unmolded and then dried for at least 28 days. The mixture to be compacted is identical to that of the CBR type. For composites (laterite-cement-metakaolin), the manufacturing techniques remain the same. Only the preparation of the different mixtures changes. For the modified Proctor tests, the mixtures are made taking into account the percentages of cement and metakaolin given in Table 4 [12] [13] .

The chemical composition of ALK lateritic clay is given in Table 5. Analysis of the table shows that the sample contains a large amount of silica (49.88 wt.%)

and alumina (25.95 wt.%) as well as an appreciable amount of iron oxide (9.69 wt.%). These results suggest that quartz, aluminosilicates and iron minerals are predominant in the sample studied.

Analysis of the powder diffractogram of the lateritic clay shown in Figure 2 shows that it is essentially composed of kaolinite (Si₂O₅Al₂(OH)₄), quartz (SiO₂) and goethite (FeO(OH)). The presence of these minerals is also confirmed by the infrared spectrum in Figure 3. The spectrum shows bands of quartz (Si-O-Si vibration at 786 cm⁻¹); of kaolinite: 3620and 3690 cm⁻¹ (O-H stretching vibrations); 1113, 1000 cm⁻¹ (Si-O vibrations); 913 cm⁻¹ (Al-OH bending vibration) [10] [20] [21] .

The DSC/TG thermogram in Figure 4 shows three endothermic peaks and one exothermic peak. The first endothermic peak at 140˚C corresponds to the departure of hydration or hygroscopic water. This thermal phenomenon is associated with a mass loss of 0.63 wt.%. The second endothermic peak at 310˚C corresponds to the dehydroxylation of goethite and its transformation into hematite [22] , with a mass loss of 1.93 wt.%. This dehydroxylation of the iron oxyhyoxides present in the sample takes place following reactions 1 and 2:

2 γ -FeOOH → 255 ˚ C γ Fe 2 O 3 + H 2 O (1)

2 α -FeOOH → 279 ˚ C α Fe 2 O 3 + H 2 O (2)

As for the last endothermic peak around 515˚C, it corresponds to the dehydroxylation of kaolinite and its transformation into metakaolinite followed by a mass loss of 9.23 wt.% [23] .

The overall dihydroxylation reaction can be schematized according to reaction 3:

Si₂O₅Al₂(OH)₄(kaolinite) → 2SiO₂-Al₂O₃(métakaolinite) + 2H₂O (3)

The dehydroxylation temperature of an ordered kaolinite is around 600˚C while that of a disordered kaolinite is lower. Thus, the ALK sample would therefore be composed of disordered kaolinite. The only exothermic peak at 960˚C corresponds to the recrystallization or the structural reorganization of the metakaolinite [9] . This is the reorganization of mekaolinite into more stable structures (spinel, mullite) in addition to amorphous silica following reaction 4:

2(2SiO₂-Al₂O₃)(métakaolinite) → Si₃Al₄O₁₂(Al-Si-spinelle) + SiO₂(amorphe) (4)

The results of the elementary chemical analysis and the mineralogy given by X-ray diffraction from relation 5 of Yvon et al. [24] made it possible to draw up the mineralogy composition of the ALK sample.

T ( a ) = ∑ M i P i ( a ) (relation 5)

with: T(a): content (oxide%) of the chemical element “a”; M_i: content (%) of mineral “i” in the material studied and containing the element “i”; P_i(a): proportion of the element “a” in the mineral “i”.

Table 6 gives the semi-quantitative composition of the mineral phases of the sample studied. It consists of a high proportion of kaolinite (42 wt.%). The quantities of quartz (19.3 wt.%) and goethite (10.8 wt.%) are significant. The balance (4.2 wt.%) consists mainly of organic matter and amorphous mineral phases.

The geotechnical properties of lateritic clay are recorded in Table 7. The Atterberg limits indicate that the raw material is moderately plastic clay. This is probably due to the presence of clay minerals such as kaolinite which makes it plastic [25] . The methylene blue value of 1.57 g/100 g attests that the sample belongs to the category of sandy clay soils with low plasticity. This result corroborates well with those of the Atterberg limits. However, the values of the plasticity index and the methylene blue value remain low compared to those reported by Millogo et al. [26] in 2008 on lateritic gravel from Sapouy (Ip = 10.5% and VBS = 0.17 g/100 g). This is justified by the clayey nature of the sample studied compared to that of Sapouy which is very gravelly and therefore poor in clay minerals.

As for the particle size composition, it is deduced from the particle size curve of Figure 5 determined by sieving, it indicates that ALK would be made up of 70% fines fraction. This confirms the clayey nature of the sample.

The results of the modified Proctor test presented graphically in Figure 6 showed that the optimum moisture content and maximum dry density are 15.9 wt.% and 1.76, respectively. The optimal high moisture content of the ALK sample and its low density compared to those obtained by Millogo et al. on the lateritic gravel of Sapouy can be explained by the clayey nature of the ALK sample. These values are in agreement with the values of the Atterberg limits and that of methylene blue.

The value of the CBR lift index at 95% after four days of immersion is 40% for the natural sample. This value between 30% and 80% shows that the sample can be used as a sub-base course raw material in road construction [27] . Its use as a base course raw material in road construction requires a chemical improvement (by mineral binders) or mechanical improvement (by adding crushed stones).

After the geotechnical parameters, the various specimens were subjected to mechanical tests by simple compression and tension by splitting. Simple compressive strength and splitting tensile strength are Rc = 1.59 MPa and ft = 0.149 MPa respectively. This compressive strength value is greater than 1.25 MPa and that of the tensile strength by splitting is greater than 0.125 MPa, values obtained by Messou on the stabilization of laterites in Ivory Coast [28] . These results are justified by the fact that the laterites studied by Messou are richer in clay minerals (therefore more plastic) than the laterite clay ALK studied. In order to improve the geotechnical and mechanical properties of the raw material and at the same time reduce cement consumption, cement-metakaolin-clay composites have been produced.

The pieces of the specimens prepared (lateritic clay-cement-metakaolin) according to the modified Proctor test process then subjected to the simple compression and tensile tests by splitting after 28 days of curing in the open air were finely ground and analyzed by X-ray diffraction, infrared spectrometry and differential scanning calorimetry.

The powder diffractograms of these composites are shown in Figure 7. Analysis of these diffractograms shows the presence of laterite raw material minerals such as kaolinite, quartz and goethite. Also we note the presence of new phases such as calcite and portlandite.

The low presence of portlandite is probably due to its involvement in the pozzolanic reaction with amorphous silica of metakaolin. To complete the mineralogy of the composites and particularly the demonstration of any amorphous phases formed, samples of ground clay-cement-metakaolin mixtures were analyzed by infrared spectrometry and then by differential scanning calorimetry

(DSC). The analysis of Figure 8 shows that in the high frequencies, we have the characteristic bands of the vibration of the OH bond of kaolinite between 3692 and 3620 cm⁻¹ [9] [29] . It can be seen that the intensity of these bands decreases with the addition of cement. This shows the reaction between the cement and some minerals of the lateritic raw material. Indeed, the kaolinite contained in the clay matrix is weakly crystallized, thus the cement can react with the non-crystalline fraction then causing a reduction in the quantity of kaolinite in the material. Also, the band at 913 cm⁻¹ corresponds to the vibration of the Al-OH bond while those at 1000 and 1113 cm⁻¹ would be due to the vibration of the Si-O bond of the kaolinite. At 748 cm⁻¹ we have the vibration of the Si-O-Al bond also of the kaolinite. The band at 786 cm⁻¹ is attributable to quartz. The spectra also show new vibration bands around 1450 and 3400 cm⁻¹. The discreet band around 1450 cm⁻¹ is due to the combined effect of calcite [9] . On the other hand, that around 3400 cm⁻¹ is attributable to the vibrations of the CSH hydroxyls [30] [31] . On the DSC thermograms of the composites presented in Figure 9, we can identify thermal accidents attributable both to the minerals of the clay sample and those due to the formation of calcite and CSH. Composites containing cement-metakaolin exhibit a large endothermic phenomenon around 100˚C. This is the phenomenon of dehydration corresponding to the loss of hygroscopic water and/or the dehydration of hydrated calcium silicates (CSH) [32] . The thermal accidents around 310˚C, 515˚C and 960˚C correspond respectively to the loss of water from crystallization of goethite, to the dehydroxylation of kaolinite and to the recrystallization of metakaolinite from the lateritic raw material. The area of the kaolinite dehydroxylation peak decreases with the addition of cement. This would be due to the involvement of kaolinite in the pozzolanic

reaction [33] . The broad peak around 700˚C corresponds to the decomposition of calcite into lime (CaO) and carbon dioxide [9] .

The newly formed phases are essentially calcite (CaCO₃), portlandite (CH) and hydrated calcium silicate (CSH). Portlandite is formed as a result of the hydration of alite (C₃S) and belite (C₂S) according to the following reactions 5 and 6 [9] [34] :

2C₃S + 6H → C₃S₂H₃(CSH) + 3CH (5)

2C₂S + 4H → C₃S₂H₃(CSH) + CH (6)

with: CaO = C; SiO₂ = S et H₂O = H.

Calcite is formed as a result of a carbonation reaction involving portlandite produced by the hydration of cement and carbon dioxide from the atmosphere according to reaction 7 [29] :

Ca(OH)₂+CO₂ → CaCO₃+ H₂O (7)

Hydrated calcium silicate (CSH) is formed according to two synthetic pathways, including the direct hydration of alite (reaction 5) and belite (reaction 6) of cement and according to the pozzolanic reaction involving the silicate of metakaolinite and hydrated lime (portlandite).

The mechanism of CSH synthesis by the pozzolanic reaction is as follows [9]

• dissociation of portlandite according to reaction 8:

Ca(OH)₂ ⇌ Ca²⁺ + 2OH⁻ (8)

• dissolution in basic medium of the silica of the metakaolinite according to the chemical reaction 9:

SiO₂(silica of the metakaolinite) + 2OH⁻ → H 2 SiO 4 2 − (9)

• reaction in aqueous medium of Ca²⁺ ions released with solubilized silica ( H 2 SiO 4 2 − ) to form CSH according to reaction 10:

Ca²⁺+ H 2 SiO 4 2 − → CaO∙SiO₂∙H₂O(CSH) (10)

The variations of the optimal moisture content and the maximum dry density as a function of the quantity of cement-metakaolin are illustrated in Figure 10. The OMC increases while the MDD decreases with the addition of cement and metakaolin. The increase in OMC is due to the strong affinity of the cement-metakaolin mixture with water to initiate the hydration reactions of their anhydrous components. In addition, the synthesis of hydrated calcium silicate (CSH) also needs water for the dissociation of the portlandite produced in order to produce Ca²⁺ ions. The OMC increases more when the metakaolin content in the mixture increases. This is linked to the high demand for water to initiate the process of the pozzolanic reaction, especially in the portlandite dissociation stage [35] . Also this significant water demand would be due to the fineness of the metakaolin.

The reduction of the MDD would be due to the fact that the addition of cement-metakaolin creates the aggregation of the particles. As a result, the volume of the material increases, leading to a change in the particle size distribution. This result is in agreement with that reported by Osulaon cement-amended laterites in Nigeria [33] . The decrease in MDD is more felt for composites rich in metakaolin and this is justified by the low density of metakaolin compared to that of cement.

The results of the California Bearing Ratio (CBR) test of the composites are presented in Figure 11. All the CBR index values are greater than 40%, which is

the natural sample. Also, the CBR for the formulations at three (3) days of air cure and after four (4) days of immersion are better than the CBR after 4 days of immersion. Finally, formulations containing more cement than metakaolin have the best CBR.

The increase in CBR would be essentially due to the formation of CSH following the pozzolanic reaction involving the portlandite of the cement and the amorphous silica of the metakaolin. The lower values of the CBR for the specimens immediately immersed result from the uninterrupted process of the hydration of the cement and the lack of consolidation of the CSH. For specimens having undergone a 3-day cure in the air before being immersed, the cement had time to set at a young age. Specimens containing more cement have the best CBR due to the high content of CSH formed by the hydration pathway of alite and belite compared to CSH formed by the pozzolanic reaction [34] . The CBR obtained are lower than those reported by Millogo et al. [26] on lateritic gravels improved with cement and lime. This difference is explained by the nature of the two samples. Indeed, the sample studied is more plastic than the sample studied by Millogo et al. The 2 wt.% metakaolin + 6 wt.% cement composite is suitable for a base course in road construction regardless of the type of traffic. The composite (4 wt.% metakaolin + 4 wt.%) cement is better suited for the sub-base course in road construction but could be used as a base course for low traffic [26] . As for the composite low in cement (6 wt.% metakaolin + 2 wt.% cement), it can only be used for the sub-base course like the natural sample. Nevertheless, the natural sample was improved in terms of CBR. These results prove that with the contribution of the cement-metakaolin mixture, the material reacts favorably both to the possible poor hygrometric conditions observable in practice and to the constraints imposed on a pavement. The addition of cement and metakaolin markedly improved the geotechnical properties of the base lateritic clay. The maximum observed (81%) for 2 wt.% metakaolin + 6 wt.% cement is the most favorable case with a CBR index greater than 80%.

The results of the simple compressive strength and tensile splitting strength tests for the different composites are presented in Figure 12. Overall, the simple compressive and tensile splitting strengths of the clay-cement-metakaolin composites are better than those of the raw sample. As with CBR, composites rich in cement compared to metakaolin have the best strengths. Thus, the resistances increase until reaching the maximum value of 4.5 MPa in simple compression against a value of 0.5 MPa in tension then decrease for the other formulations. The improvement in resistance is explained, as for CBR, by the formation of CSH, a cementitious compound responsible for the mechanical strength. These CSH then play the role of glue that will bind the isolated particles of the clay matrix while reducing the porosity [34] . The drop in resistance is related to the formation of an excessive amount of calcite in the mixture.

Calcite is formed in large quantities because the portlandite formed is not fully used for the pozzolanic reaction, so the rest will undergo a carbonation reaction [36] .

The simple compressive strengths and the splitting tensile strengths for the engineered composites are higher than those reported by Messou, Bahar and Kolias [28] [37] [38] [39] on lateritic soils improved with cement. This could be due to the mineralogical composition of the soils studied by these authors which would be richer in clay minerals compared to our basic raw material ALK.