Valorisation of dune sand-tuff for Saharan pavement design

The current experimental research aims to valorise two Saharan soils (dune sand and tuff) for pavement purposes in arid zones. A mixture of 65 % tuff and 35 % dune sand (named TDSopt) was previously studied and exhibited optimum geotechnical characteristics that were suitable for low-traffic pavements. However, the rise of groundwater level with aggressive sulphate in the Ouargla region, Algeria, weakened the mechanical properties of this TDSopt. Further treatment with sulphate-resistant binders was therefore proposed to remedy this defect. In this research, laboratory testing was performed to define the optimum treatment for the TDSopt, followed by the hydromechanical characterisation of the treated TDSopt. Experimental results validated the following observations. First, the treatment of TDSopt with cement and/or lime resulted in decreased dry density, but an increased California Bearing Ratio (CBR) index. Also notably improved were unconfined compressive strength (UCS), secant modulus (E 50 ) and resistance to immersion (R imm ). More specifically, the defined optimum recipe – TDSopt plus 1.33 % cement and 2.66 % lime – increased the CBR index by 75 % and 35 % for the soaked and unsoaked specimens, respectively. UCS, E 50 and R imm increased by 65 %, 75 % and 65 %, respectively. The optimum treatment also increases the shrinkage limit void ratio of compacted TDSopt from 0.39 to 0.46. Finally, the strength parameters (UCS, cohesion and friction angle) of optimum treated specimens were larger than those of untreated material regardless of water content. At Modified Proctor Optimum, the UCS, cohesion and friction angle increase from 0.54 to 2.46 MPa, 167 to 800 kPa and 23 ◦ to 40 ◦ , respectively.


Introduction
The development of Saharan regions requires the design of communication routes, especially roads, to connect urban areas to the most remote villages.The scarcity of standardised road materials and the arid climate have prompted engineers and researchers to look at various possible solutions to this challenge.One of these solutions is to reuse naturally available local materials, including dune sand, as a base material in the engineering of Saharan pavements, in accordance with current pavement design standards.In order to achieve the required performance in terms of hydromechanical properties and durability, it is necessary to mix the dune sand with fine soil particles and treat it with a low percentage of hydraulic binders.
Several research studies have been carried out on the techniques of amending local materials with missing granular fractions (fine or coarse) [1][2][3][4][5][6].Cabalar et al. [5] have studied the influence of sand addition on the behaviour of sand/clay mixtures.The authors observed that the addition of 30 % sand led to an improvement in compaction characteristics, CBR index and a reduction in swelling potential.However, UCS decreased with increasing sand content.Therefore, the mechanical performance based on the addition of sand alone is not sufficient [2,5,7].To increase the performance of these materials according to the standards, it is necessary to treat them with different binders.In such a context, some studies have combined fly ash (FA) with cement as a stabiliser [8][9][10][11][12][13], and on the other hand, they have used a fly ash-based geopolymer with a high calcium content [14,15].Chompoorat et al. [12] conducted studies on the stabilisation of dredged sediments mixed with different percentages of ordinary Portland cement (OPC) (3-10 %) and fly ash (FA) (5-20 %) as pavement material.The authors showed that the stabilisers greatly improved the strength characteristics of these materials, and the addition of 15 % FA generated the highest strengths and material stiffnesses in the stabilised sediments.Phummiphan et al. [14] investigated the effects of alkali activator and curing time on the UCS and microstructural characteristics of a marginal lateritic soil stabilised with a high calcium fly ash-based geopolymer.The authors observed that the stabilised material meets the standard requirements for heavy and light traffic-related base materials specified by local and national standards.
Other studies have favoured mixed cement-lime treatments [16][17][18][19][20][21][22][23].Depending on soil cohesion, the treatment with lime is often used in highly cohesive clay soils.Cement is mainly used to stabilise coarse and non-plastic soils.A mixed lime-cement treatment is used for slightly to moderately plastic soils.The effects of lime-only, cement-only and limecement stabilisation on compaction characteristics, bearing capacity, plasticity, compressive strength, modulus of elasticity, resilient modulus, shrinkage-swelling and shear strength parameters have been prospected by different authors [24][25]19,[26][27]22].Chompoorat et al. [22] showed that the cement treatment is more effective than the lime treatment.Stabilisation with cement leads to a significant increase in CBR indices, especially for soaked samples [17].Studies by Aytekin et al. [28] and Asgari et al. [21] showed that cement stabilisation significantly improves the compressive strength and bearing capacity of the soil.In addition, Croft [29] and Nelson et al. [30] also reported that lime or cement treatment reduces the liquid limit, plasticity index and maximum dry density of the soil and increases its optimum water content, shrinkage limit and strength.
In arid areas of the world, the valorization of dune sand has been addressed in numerous works in the literature, particularly for pavement design [31][32][33][34][35][36][37].Wahhab et al. [31] treated dune sand with liquid asphalt, cement and lime at different percentages.The authors summarised that the hydromechanical properties of dune sand stabilised with these additives were improved, especially for its shear strength and water sensitivity.This treated dune sand can be used in the base and subbase layers of pavements.Smaida et al. [33] treated dune sand with cement, lime and pozzolan for pavement subbase applications in arid areas.The authors concluded that the treatment of dune sand with 12 % pozzolan and 3 % lime gave better performance for pavement foundations.Almadwi et al. [34] used dune sand as an additive for the development of different asphalts.The results showed that the addition of 33 % dune sand allows the use of treated asphalts for the design of lowtraffic pavements.
In the Algerian Saharan regions, the dune sand has been extensively investigated for amended tuff to improve pavement in arid zones [36][37][38][39][40].These two local materials are abundant and cover a very large area in this region [36].Raw tuff was first used in Saharan pavement design because of its acceptable cohesive strength after compaction, especially on the dry side of the Proctor optimum condition [41,42].However, as the water content increases, this cohesion decreases considerably and the pavement starts to crack, due to insufficient bearing capacity [43,44].Raw tuff has low water resistance.In fact, it is not recommended to use raw tuff for the construction of modern highvolume and heavy-traffic Saharan pavements [45].Previous studies have shown that the addition of dune sand to the crusting tuffs led to a considerable increase in the Optimal Modified Proctor characteristics and CBR indices [4,7].Daheur et al. [7] studied the geotechnical properties of tuff/dune sand mixtures.The results showed that a 65 % tuff and 35 % dune sand mixture (named TDSopt hereafter) exhibited optimal geotechnical characteristics and mechanical performance.However, for applying TDSopt to medium and heavy traffic roads, and to make it less sensitive to humidity variations (capillary rise and/or rainwater), additional treatment with binding agents such as cement and/or lime is still required.
Groundwater in the Sahara, particularly in the Ouargla region (Algeria), is currently undergoing a significant increase to the surface.Indeed, the arid climate of this region induces a high concentration of mineral precipitates in the shallow waters.This leads to the paragenesis of saturated saline solutions and thus creates an aggressive environment [46].In addition, several experimental studies (e.g., [47,48]) have confirmed the excessive salinity of groundwater and soils in the Ouargla region (e.g., SO4 2-, Cl-, Mg 2+ , Ca 2+ , Na + ).In fact, groundwater levels rise in the winter and fall rapidly in the summer due to the extreme dryness of the climate, including temperatures typically above 50 • C [49].Understanding the effect of hydric conditions (e.g.moisture and suction) on local pavement materials is crucial to improve the strength of road layers and, ultimately, the long-term viability of Saharan pavements.In order to construct roads that can withstand both traffic and aggressive environmental loads, it is necessary to treat the pavement layers with salt-resistant cement.
The present research is a continuation of previous experimental works conducted by Daheur et al. [7,38].The purpose is to extend the sustainable use of two Saharan soils (dune sand and tuff) to include future pavement applications under medium and heavy traffic loads and wide water conditions.The novelty of this paper lies, on the one hand, in the treatment of this optimised mixture with a low percentage of hydraulic binders, so as to comply with the design parameters, with regard to hydro-mechanical properties, sensitivity to water and durability.On the other hand, a detailed analysis of the behaviour of this optimal mixture (in particular, its stiffness or modulus, its resistance to rupture, its soil water retention characteristic (SWRC) and its shrinkage-swelling paths) treated with binders versus the variation of the water content was performed.

TDSopt mixture
The two components of the studied TDSopt mixture, 65 % tuff and 35 % dune sand are from the Ouargla region located 800 km south of the capital city Algiers (Algeria).This region is considered a Saharan zone, which is characterised by very low annual precipitation, e.g., < 200 mm, as shown in Fig. 1a.The tuff is a powdery material of light yellow colour (Fig. 1b).It is mainly composed of fine particles, agglomerates and solid limestone and/or gypsum grains.Dune sand is a coarse material, to the naked eye, composed of discrete grains of rounded shape and uniform diameter (Fig. 1c).The sieve analysis of these two materials and the TDSopt mixture is shown in Fig. 1d.
The TDSopt mixture was previously studied by Daheur et al. [7,38], and its major geotechnical and mechanical properties are replicated in Table 1.American Association of State Highway and Transportation Officials (AASHTO) [50], Unified Soil Classification System (USCS) [51] and Technical guideline on the embankment and capping layer construction (GTR) [52,53] were used to classify the studied materials.The chemical composition of used materials, determined by X-ray fluorescence (XRF), is presented in Table 2.This analysis shows that the dune sand is classified as a siliceous material, with a SiO 2 content of about 97 %.The tuff is classified as gypsum-limestone material, with a CaCO 3 + CaSO 4 content of about 65 %.

Binders
Two hydraulic binders were used to stabilise the TDSopt mixture.One was sulphate-resistant cement (CEM I 42.5N-SR3), and the other was a natural hydraulic lime (NHL).The sulphate-resistant cement was chosen mainly due to its ability to resist chemical attacks, whereas the choice of the natural hydraulic lime was based on its low cost and abundance in the studied area.Table 3 summarizes the chemical composition of the two binders.The combination of these two binders, with varying percentages of each component, has rarely been applied to the stabilisation of Saharan soils.

Method
As schematically shown in Fig. 2, the experimental methods of the present study include two major stages.The first stage aims to E.G. Daheur et al. investigate the effect of treatment on the compaction characteristics (dry density, optimum water content), bearing capacity and unconfined compressive strength of the TDSopt mixture.The treatment consists of mixing TDSopt mixture with pure cement, pure lime and cement-lime mixture CLMix (mass ratio 1:2).The amount of the binder added was 2, 4 and 6 %.After mixing the dry TDSopt mixture with binders, a specific amount of water corresponding to MPO was added to the dry TDSopt-binders mixture [54].The wet mixture was stirred by hand and stored in a plastic bag for 24 h [5].One part of the mixture was compacted with modified Proctor effort (ASTM D1557) [55] and then tested by CBR tests (ASTM D1883) [56].The other part of the mixture was statically compacted in a double-piston mould (D = 50 mm, H = 100 mm) at 1.27 mm/min.A specific amount of mixture was weighed, added to the mould and then compacted to a single layer of 100 mm in height and density of the MPO condition [2,57].This method leads to a homogeneous distribution of the compaction stress [57,58] and also avoids trim of the specimen to preserve the initial structure.These statically compacted specimens were divided into two groups.Specimens of Group no. 1 were intended for assessing water sensitivity.The specimens were first cured in the air at T = 25 ± 2 • C for 21 days and then in natural aggressive water (taken from the field) for 1, 3 and 7 days [59,32].Specimens of Group no. 2 were cured in the air for 28 days under the same ambient condition [9].After curing, specimens from both groups were compressed by UCS tests following the ASTM D2166 [60].The results presented for the UCS and CBR tests are the averages of three readings.In most cases, the results obtained under the same test conditions were reproducible, with a small mean standard deviation (SD) ((SD/m) < 10 %, where m is the mean value of UCS or CBR).In the corresponding curves, the error bars for each point and the range of these bars were presented.Through this stage, the TDS mixture plus binder recipe that satisfies the resistance criteria was defined, while considering also the economic aspect.
The second stage aims to understand, forecast and control the behaviour of the chosen recipe in different hydric states by dryingwetting, UCS and direct shear tests (Fig. 2).The procedure of dryingwetting tests was described in detail by Fleureau et al. [61,62].It consists of imposing suctions on the specimens by using the osmotic technique for low suctions (10-1500 kPa) and the saturated salt solution technique for high suctions (2-20 MPa).The carefully trimmed specimens, approximately 2-3 cm 3 in dimension, were placed in contact with the imposed suction for about one week for the osmotic technique and one month for the salt solution technique [57].When equilibrium was reached, the specimens were taken out.Their volumes were measured by hydrostatic weighing in non-wetting oil, followed by oven-drying for water content determination [61].The void ratio and degree of saturation were then deduced from measurements.The tests were performed on specimens compacted at the optimum water content [63].The initial suction of the specimens was measured by filter paper (ASTM D5298) [64].The determination of initial suction allows the definition of the limit between the drying and wetting paths [62].The specimen follows the drying path if the imposed suctions are larger than the initial suction, otherwise, a wetting path will be followed.Direct shear tests are performed on parallelepipedic specimens of 60 × 60 × 30 mm 3 with normal stresses of 100, 200 and 300 kPa and a shearing rate of 0.5 mm/min (ASTM D3080-04) [65].The specimens were initially compacted at the MPO condition and then dried or wetted to different target water

Compaction characteristics
Fig. 3 presents the modified Proctor compaction curves of the TDSopt mixture treated with different binder types and amounts.Stabilising the TDS mixture with cement and/or lime leads to a flatter Proctor curve.Fig. 4 presents the optimum parameters, showing a linear increase in optimum water content (OWC) and a linear decrease in maximum dry density (MDD) with binder amount.OWC increases mainly due to the abundance of fine particles in the studied cement and lime, which exhibit a larger specific surface area and thus a higher absorption ability for water.Besides, the addition of binders requires additional water allocated for hydration and pozzolanic reactions [68,69,27,11].The reduction in maximum dry density can be explained as follows.As mentioned in the Methods section, the wet specimens were mixed by hand, stored in a large plastic bag for 24 h, and then compacted.During the storage phase in the bag, cationic and pozzolanic exchange reactions take place, and cementation bonds can take place between the soil particles.During Proctor compaction, these bonds prevent the particles from freely rearranging, which ultimately results in a lower dry density [70,71].The variations in the optimum dry density and water content due to the addition of binders can be summarized as follows:: (i) MDD is larger for the untreated soil than for the treated soil; (ii) MDD is larger for the soil-cement mixture than for the soil-lime mixture; and (iii) for the same binder content, OWC is smaller for the soil-cement mixture than for the soil-lime mixture.These results are consistent with the literature results [72,73,27,23].For the TDSopt stabilised with lime, the dry density is less sensitive to variations in water content, resulting in a Proctor curve remarkably flattered.This observationthe flattering of the Proctor curve of lime-treated soilswas also noticed by Sweeney et al. [74] and Sivapullaiah et al. [71].

Bearing capacity
CBR values of both soaked and unsoaked samples after compaction are illustrated in Fig. 5. Adding binders enhances the CBR index of the specimens, especially for the soaked ones (Fig. 5a).As an example, after immersion, the CBR index of the TDSopt treated with 6 % cement increased by two times, i.e., from 65 % to 194 %.The improvement of the CBR index makes the soil suitable for the movement of various construction machinery and facilitates the compaction of different pavement layers.These findings are in agreement with those of Ojuri et al. [26] and Mahmood et al. [23].
The increase in CBR is different when the TDSopt is treated with cement or lime alone.In the former case, pozzolanic reactions take place upon hydration, resulting in cementing bonds between the granular soil particles and consequently a remarkable increase in soil bearing capacity [75,76,17,77].In the case of treatment with pure lime, the shortterm reactions are dominated by the cationic exchange [78,79].The exchange of alkaline cations with calcium cations enables the agglomeration of fine particles and forms into aggregates, which leads to less increase in the CBR index in comparison with cement treatment alone.Regarding the TDSopt treated with the CLMix, the reactions are a mix of pozzolanic reactions and cationic exchanges, and the increase in the CBR index falls between that of lime alone and that of cement alone (continuous red line in Fig. 5a and 5b).

Unconfined compressive strength
UCS tests results are included in Fig. 6, showing the axial stress against axial strain for the TDSopt and treated TDSopt.Axial stress grows with axial strain until the peak at 0.5-0.65 % followed by a gradual decrease.The post-peak behaviour tends towards residual stress that depends on the binder amount and type.
The peak strain of each test is extracted and presented in Fig. 7a, from which one may notice a linear decrease trend of peak strain against   binder amount.This implies that the ductility of the treated TDSopt decreases when more binder amount is added.In a similar experimental study on a cement-treated poorly graded sand-gravel mixture, Amini and Hamidi [80] found the occurrence of peak strength at small strains of 0.2-0.7 %.
The values of UCS and E 50 (secant modulus defined at 50 % of the maximum axial stress) are calculated and plotted against the amount of binder.As shown in Fig. 7b and 7c.Both UCS and E 50 increase linearly with the percentage of binders.The results are consistent with those of previous studies, such as [71,81,21,82,22].However, we can observe in Fig. 7 that the slopes are larger for the treatment with cement than for the treatment with lime [21].For the UCS and E 50 plots (Fig. 7b and 7c), the slopes are nearly-four times larger for treatment with cement than with lime.
To check the efficiency of the treatment, the gain in UCS and E 50 is defined using the following formula: where V BT is the value before the treatment, and V AT is the value after the treatment.The calculated values are presented in Fig. 8. Similar to the results in Fig. 7, the gains are the largest for the cement-treated specimens.For example, the gain in UCS is approximately 190 % (UCS from 1.71 to 4.98 MPa) when the cement dose increases from 0 to 6 %, compared to the corresponding gains of 90 % and 60 % for the CLMix and lime treated specimens, respectively.Similarly, the gain in E 50 of 230 % (E 50 from 298 to 987 MPa) for the cement treatment is more than doubled and four times larger than that of CLMix (104 %) and lime (58 %) treatments, respectively.These results are in agreement with those of [21,83,23].MPa and from 3.8 to 2.82 MPa (i.e. 9 % and 26 % loss of resistance) for the soil treated with 6 % and 4 % cement, respectively, and then stabilises regardless of the immersion time.For the TDSopt treated with 2 % cement, the loss of resistance is 46 % after 24 h of immersion.As regards the treatment with CLMix, the loss of resistance in the first 24 h was 20 % and 26 % for the binder amount of 6 and 4 %, respectively.After that, the UCS keeps nearly constant for CLMix of 6 and 4 %, but declines gradually and becomes null for 2 %CLMix.The treatment with lime alone is detrimental to the resistance of the material after immersion.UCS decreases continuously with the immersion time whatever the lime amount and completely disappears after 3-and 1-day immersion for the treatment with 4 and 2 % lime, respectively.
According to [59,32,84], the resistance to immersion ratio R imm is defined as follows: where UCS i is the unconfined compressive strength of specimens stored in air for 21 days followed by an immersion period of 7 days, and UCS 28 is the compressive strength curing in the air for 28 days and without immersion.The calculated R imm was plotted against the binder.As shown in Fig. 10, R imm of the specimens treated with 4 % and 6 % cement and 4 % and 6 % CLMix are above the 60 % threshold required by the GTS standard (French acronym for 'Soil Treatment Guide' [85]).The results, in good agreement with the literature [32,84], suggest that the dosage of cement or CLMix greater than or equal to 4 % satisfies the resistance to immersion.

Discussion about the optimal binder treatment
The results concerning the binder treatment of TDSopt highlighted the effectiveness of certain recipes over others.However, to select the optimal treatment, several criteria must be comprehensively considered in accordance with the design standards of Algerian Saharan roads and the AASHTO pavement design guide.These criteria are UCS at 28 days of curing time (UCS 28 ) (as reported by Horpibulsuk et al. [81] and Jamsawang et al. [10]), resistance to immersion (R imm ) and the cost of the treatment.More specifically, the TRS (French acronym for 'Saharan Road Technique') requires the UCS 28 larger than 2 MPa [40,41], AASHTO requires UCS 28 to be larger than 1.7 MPa for a stabilised subbase and 5.2 MPa for stabilised base [10] and the GTS recommends a threshold of 60 % for R imm [85].As highlighted in Fig. 11, the mechanical parameters satisfying all conditions require a binder treatment by cement or by cement-lime mixture with an amount larger than 4 %, for stabilised subbase layer pavement.AASHTO's Pavement Design Guide does not allow treated TDSopt to be used as the base layer.Since 4 % CLMix is approximately 35 % more economic than 4 % pure cement, the treatment of TDSopt with 4 % CLMix is considered the optimum recipe with respect to all three criteria set above and can be used as subbase layer pavement.
In Sahara, the drought climate and high temperature in summer make the pavement surface extremely dry and lead to desiccation cracks.In addition, this region also suffers from problems caused by the groundwater increase.The Saharan pavement in this case has been subjected to significant drying-wetting during the servicing time.The effect of the drying-wetting on the mechanical behaviour is investigated in detail hereafter with the 4 % CLMix (optimum recipe) treated TDSopt.

Hydromechanical properties of the TDSopt treated with 4 %CLMix 4.2.1. Drying-wetting behaviour
Fig. 12 shows the global representation of the drying-wetting paths of the TDSopt treated with 4 % CLMix and the untreated TDSopt mixture.Fig. 12a and 12b present the variation of void ratio versus water content and suction, respectively.These figures are also referred to as shrinkage-swelling curves [86], reflecting the volumetric changes of the soil when its hydric condition changes.The shrinkage limit of water content (w SL ) is defined as the intersection of two asymptotes of the curves (dashed lines in Fig. 12a).The values of w SL are 11.5 % and 10 % for the TDSopt + 4 %CLMix and TDSopt, respectively; whereas the corresponding void ratios (e SL ) are 0.46 and 0.39, accordingly.The shrinkage limit of suction (s SL ) is defined as the intersection of asymptotes of the two parts of Fig. 12b one is characterised by a large deformation and the other represents the quasi-constant volumetric change.The values of s SL are about 300 and 400 kPa for the TDSopt + 4 %CLMix and TDSopt, respectively.The position of the shrinkage limit of TDSopt + 4 %CLMix, compared to that of the untreated TDSopt, can be explained by the fact that the binders create cementing bridges between the soil particles which minimises the overall shrinkage of the treated material.The position of this shrinkage limit is thus higher than that of the untreated material.These cementation bridges in the treated soil also justify the smaller swelling slope, observed on the wetting path [20,87].
Fig. 12c and 12d show the saturation-desaturation curves.In (w, S r ) plane, there are two quasi-linear parts.The first part (i.e., wetting path) has smaller slope values, which indicate the quasi-saturation of the soil.The pores fill with water without completely saturating the sample due to the effect of the ink bottle phenomenon.The second part (i.e., drying path) exhibits a larger decrease in S r , which corresponds to the rapid drainage of water when w < w SL on the drying path.In (s, S r ) plane, the degree of saturation decreases rapidly to approximately 10 % for suctions greater than 3000 kPa.On wetting path, the degree of saturation slightly increases to reach approximately 80 % for suction of about 100 kPa, then tends towards a plateau of S r = 85 % for very low suction values.Full soil saturation can not be achieved due to the ink bottle effect, as explained above.
Fig. 12e shows the soil-water retention curve (SWRC).The water content varies almost linearly in the semi-logarithmic plane when suction varies from 10 to 300 kPa (wetting path) for TDSopt + 4 %CLMix and TDSopt specimens.For both soils, the residual water content is about 1.5 % when the suction is higher than 3 MPa.
In the literature, Bourabah [88] and Djelloul et al. [87] have investigated the effect of cement treatment on drying-wetting paths.Their results show that the treatment decreases the values of w SL and s SL but increases e SL of the material.In this study, the influence of treatment by 4 %CLMix on the drying-wetting paths is compared and summarised in Table 4.
Fleureau et al. [62] correlated the optimum characteristics of compacted soils (w MPO , γ dMPO and s MPO ) to liquid limit (w L ) by the following equations: In addition, Fleureau et al. [62] also correlated the swelling indices     C ms and D ms of the wetting paths in the (logs, e) and (logs, w) planes by the following formulae: Table 5 summarizes the measured and correlated values of various parameters.The observed difference is between 0.5 % and 20 %.To further generalise this concept, Fleureau et al. [62] defined a normalised water content (w/D ms ).Fig. 13a presents the relationship between the normalised water content and suction on the wetting path for 10 compacted soils from the literature and the TDSopt + 4 %CLMix of this study.The wetting path of TDSopt + 4 %CLMix is well superimposed to the mean results, leading to a single generalised wetting path for the 11 obtained soils.In Fig. 13b, wetting paths of the 11 compacted soils are presented in the (logs, e) plane.The swelling index C ms of the TDSopt + 4 %CLMix (solid diamond with red colour) is well positioned in the correlated (logs, e) system.For the 11 studied compacted soils, the liquid limit appears a relevant, intrinsic and easy-to-determine parameter that predicts the behaviour of compacted soils on wetting paths.

Unconfined compressive strength
Fig. 14a shows the variation of USC and gains as a function of water content.The treatment by 4 %CLMix increases significantly the UCS of specimens.At dry state (w = 0 %), the UCS of TDSopt + 4 %CLMix is 2.84 MPa, which yields a gain of 65 % compared to the UCS of untreated TDSopt, i.e., 1.71 MPa.As water content increases to quasi-saturation (w = 18 %), UCS of TDSopt + 4 %CLMix is 1.95 MPa, which is greater than that of TDSopt at w = 0 %.The gain increases and reaches 1047 %.Fig. 14b presents the relationship between E 50 and w.At w = 0 %, the values of E 50 are 523 MPa and 298 MPa for TDSopt + 4 %CLMix and TDSopt, respectively.This represents a gain of 75 %.At w = 18 %, E 50 are 338 MPa and 25 MPa for the treated and untreated specimens, suggesting a gain of 1252 %.The gain in both mechanical parameters increases exponentially with the increase of water content and is more significant on the wet side of MPO.The treatment of TDSopt with 4 % CLMix reduces the sensitivity of the material to water and enhances its mechanical properties.Abu-Farsakh et al. [77] observed the same shape of UCS variation with water contents.
The relationships between USC and E 50 and suction, in the linear coordinate system, are presented in Fig. 15a and 16b.Both UCS and E 50 increase parabolically with suction.The curves tend towards a plateau when the suction approaches 4 MPa.In the bi-logarithmic plane (Fig. 15c and 15d), UCS and E 50 increase quasi-linearly with suction.In the literature [89][90][91]38], E is often estimated with the following formula: In this study, Eq. ( 8) is extended to correlate UCS and E 50 with suction.The new correlations are as follows: where α and β correspond to the E 50 and UCS at the suction value of 1 kPa, n and m correspond to the slope of (logs, logE 50 ) and (logs, logUCS), respectively.Table 6 summarizes the values of these parameters.
Compared to those of TDSopt, the slopes n and m decrease for TDSopt + 4 %CLMix, suggesting that the treatment leads to the material being less sensitive to the suction variation.
A quasi-linear relationship exists between E 50 and USC, as demonstrated in Fig. 16.Such a relationship was noticed in many previous experimental studies on cement or lime-treated soils [92,93,84,10,22,12,13].However, no generalised correlation equation can be obtained due to the different soil components and binders used.Although the slope of this study is located between previous studies, it is over 3.27 times greater than that of the study by Chompoorat et al. [22], who showed that E 50 = 56.5 UCS.In their study, UCS tests were performed on dredged clay treated with cement or lime.In another similar study, Chompoorat et al. [12] carried out the UCS tests on dredged sediments treated with ordinary Portland cement and fly ash.The results showed that E 50 = 330 UCS, with the corresponding slope 1.78 times larger than that in this study.

Shear strength
Fig. 17 presents the variation of the apparent cohesion C and friction angle φ versus water content and suction.In Fig. 17a and 17b, C app decreases rapidly on the dry side and then slightly on the wet side of the optimum water content.φ app decreases linearly with water content, with a larger slope value for the TDSopt.At the wettest state (w = 18 %), C app and φ app of the TDSopt + 4 %CLMix are 677 kPa and 35.9 • , respectively.These values are larger than those of the TDSopt even at the driest state (w = 0 %), i.e., 625 kPa and 32.4 • , respectively.The gain in both C app and φ app increases with water content.The observed gains show that the cohesion is considerably improved, e.g., 120 % for w = 0 % and 1400 % for w = 18 %.Similarly, the improvement in friction angle is also remarkable but with fewer gains, e.g., 40 % for w = 0 % and 140 % for w = 18 %.This suggests that the improvement in mechanical properties (UCS and E 50 ) could be mainly due to the enhanced cohesion during treatment.These observations are in agreement with the work of Wang et al. [20] and Amini et al. [80].In Fig. 17c and 17d, the variation of C app and φ app versus suction in a semi-logarithmic plane.C app and φ app increase linearly with suction.The slopes of C app and φ app for TDSopt + 4 %CLMix are smaller than for TDSopt, implying that the cohesion and the friction angle become less sensitive to the variations in suction after treatment.This has already been observed in paragraph 4.2.2,where the parameters n and m of Eqs. ( 9) and (10) decrease after the treatment.These results, concerning the increase of the apparent cohesion and the friction angle following the treatment with hydraulic binders as well as its variation with the water content and the suction, are in agreement with those found in [94][95][96].
To evaluate the mechanical parameters of the optimal TDSopt + 4 % CLMix mix design, we compared them to some literature results using similar base materials.Table 7 presents the cohesion and friction angle values of these materials from Saharan regions intended for the design of low-traffic pavements.For the tuff compacted at MPO, the friction angle varies from 30 to 40 • and the cohesion changes from 90 to 160 kPa.For  the treated tuff mixtures, the variation of friction angle and cohesion is in the range of 35-44 • and 50-100 kPa, respectively.In this study, the treated material (TDSopt + 4 %CLMix) features a cohesion of 677-1375 kPa and a friction angle of 36-46 • .The values of both mechanical parameters are larger than those of untreated tuff and treated tuff mixtures in the literature.The comparison suggests that the optimum recipe used in this study is superior to previous recipes and may better serve the pavement constructions in the Saharan zone for medium to heavy traffic loads.

Conclusion
An experimental study is performed to valorise Saharan local soils for the design of medium-to heavy-traffic pavements.A mixture composed of tuff and dune sand was treated with hydraulic binders to meet the requirements of the specifications and standards in terms of strength and durability for pavement materials in Saharan areas.The treatment of the tuff and dune sand mixture leads to the following conclusions: • The treatment with the binders, particularly the lime, reduces the sensitivity of the dry density to water content variation.This reduces the risk of fluctuation of compaction water content in situ.• The binders, in particular the cement, improve considerably the unsoaked and soaked CBR indices and therefore the compatibility and trafficability, especially during the phases of the pavement construction.• The study of the optimisation of treatment with binders has concluded that the soil mixture called TDSopt + 4 %CLMix, composed of 65 % tuff, 35 % dune sand and 4 % binders (2/3 of lime and 1/3 of cement) is the recipe that meets both the technical criteria (i.e.mechanical properties and durability in Saharan areas) and the economic optimisation.• The wetting-drying paths carried out on the TDSopt + 4 %CLMix material compacted at MPO, reveal that the shrinkage limit void ratio increases from 0.39 for untreated TDSopt to 0.46 for treated TDSopt.On the wetting path, its swelling is lower than that of the untreated material.Liquid limit is a simple intrinsic parameter to predict the optimal characteristics and swelling indices of compacted soils on wetting paths.These correlations are valid for treated compacted soils.In accordance with the design standards, these results highlight the possibility of using local materials based on tuff and dune sand for the design of Saharan pavements for the subbase layer.However, complementary studies on the adopted formula (TDSopt + 4 %CLMix) must be carried out in the long term and in fatigue.In addition, monotonic and cyclic triaxial tests, with large and small deformations, must be performed.Furthermore, an in-depth study of the microstructures of the samples treated by a combination of the scanning electron microscope (SEM) and energy dispersive X-ray spectroscopy (EDX) will provide a better understanding and interpretation of the results obtained in the current study.
The Academy of Finland project Geomeasure (decision number 347602) provided financial support for the second author.

Fig. 1 .
Fig. 1.(a) Location, (b) photograph of tuff, (c) photographs of dune sand and (d) sieve analysis of the studied materials.

Fig. 2 .
Fig. 2. Experimental steps and flow in this study.

Fig. 4 .
Fig. 4. Variation of the compaction characteristics: (a) optimum water content and (b) maximum dry unit weight as a function of binder content.

Fig. 5 .
Fig. 5. Variation of (a) soaked and (b) unsoaked CBR index versus the amount of binder.

Fig. 6 .
Fig. 6.Variation of axial stress versus axial strain of the different mixtures of TDSopt + binders.

Fig. 7 .
Fig. 7. Parameters derived from the axial stress-strain curve: (a) axial strain at the peak; (b) unconfined compressive strength and (c) secant modulus versus the binder amount.

Fig. 10 .
Fig. 10.Resistance to immersion ratio versus the amount of binders.

•
Compared to untreated material (TDSopt), the optimised TDSopt + 4 %CLMix exhibits mechanical properties (e.g.UCS and E 50 ) that are less sensitive to the variation in water content.For high water content, the values of UCS = 1.95 MPa and E 50 = 338 MPa, these values are larger for TDSopt + 4 %CLMix than for dried untreated TDSopt (UCS = 1.71MPa and E 50 = 298 MPa).The gain in UCS and E 50 of TDSopt + 4 %CLMix increases with the increase in water content.Depending on the suction, the E 50 and UCS can be modelled by a linear law in a bi-logarithmic plane, i.e., E 50 = αs n and UCS = βs m .•The direct shear tests show the same trend in terms of mechanical parameters at failure.The values of the apparent cohesion and the friction angle are larger for the TDSopt + 4 %CLMix than for the untreated TDSopt.At high water content (w = 18 %), the cohesion and the friction angle increase, respectively, from 44 to 677 kPa and from 15 • to 36 • .

Fig. 14 .
Fig. 14.Variation of (a) UCS and gain versus water content; (b) E 50 and gain versus water content.

Fig. 15 .
Fig. 15.Variation of (a) UCS and gain versus suction; (b) E 50 and gain versus suction in the linear coordinate system; (c) UCS versus suction; and (d) E 50 versus suction in the bi-logarithmic plane.

Table 1
Geotechnical characteristics and physico-mechanical properties of used materials.

Table 2
Chemical composition of used materials.

Table 3
Chemical composition of used binders.

Table 5
Comparison of the measured and correlated values of the characteristics of MPO compacted TDSopt + 4 %CLMix (w L = 22 %).

Table 6
Correlation parameters of the UCS and E 50 versus suction.

Table 7
Cohesion and friction angle values of tuff-dominant materials from Saharan regions intended for the construction of pavements.