Earth stabilisation with MgO-based cement

This study compares, for the first time, MgO-based (MB) cement and Portland-based (PB) cement for stabilizing earth mortars. While MB and PB earth mortars reach similar strength, MB cement stabilization demonstrates superior early-age performance. Thermogravimetric analysis, X-ray diffraction, 29 Si and 31 P NMR spectroscopies show that the cement reacts in both systems and allow to establish the phase assemblages. The stabilized earth pastes contain less hydroxide phases, indicating a pozzolanic reaction in both cases. MB-stabilized clay mortars retain about 1/3 of the compressive strength of pure MB mortar, while with PB this proportion is only 1/5. This difference demonstrates that MB is more compatible with clay minerals and more suitable for stabilizing earth mortars. If MB cement could be produced with renewable energy from CO 2 -free sources (instead of from magnesite), stabilization of earth mortars with MB would be substantially more CO 2 efficient than with PB.


Introduction
Efforts to convert unused excavation materials into building materials are motivated by their local availability and the potential to decrease concrete consumption.Concrete consumes local resources of sand and gravel and causes about 5-9 % of all anthropogenic CO 2 emissions, mainly due to the production of cement [1].Stabilized raw earth, potentially combined with structural elements made of timber, remains currently one of the best options to build housings with minimal CO 2 emissions [2].While earth materials cannot replace concrete in all applications, when adequately designed and implemented, they may be used for non-load-bearing concrete building blocks, as load-bearing walls for housing, and in soft foundations [3].Poured earth, as opposed to pressed earth as in rammed compressed earth blocks, has the highest potential as a concrete substitute [2,3].For this application, the material must be initially fluid and workable to be poured into moulds.
Raw earth contains a range of mineral particles, ranging in size from gravel to sand, silt, and clay.Only the latter two types of particles ensure binding in raw earth.The silts correspond to particles in the 2-63 μm range, while clays correspond to the finest particles (Ø < 2 μm) in the colloidal range.The cohesion and integrity of raw earth materials rely on physical binding that derives from the complex combination of multiscale forces, predominantly capillary, adhesion, and colloidal forces acting between the clay and silt fractions.These forces depend on the nature of the clays and on the relative humidity (RH).Fluidity, a crucial characteristic of poured raw earth materials, depends on the same parameters as cohesion: the content and type of clay and the RH.Whereas clays and other fine particles are the key parameters for the physical binding, these particles can also be challenging for the workability of earth materials, e.g.due to the formation of lumps that hamper the homogeneity and the fluidity of the viscous material.
Pressed earth (either compressed blocks or cob wall) is the common technique to include natural materials in housing.It is common practice to increase the compressive strength of such material with the addition of lime (Ca(OH) 2 ) or Portland-based (i.e.CaO-based) cement.This socalled stabilization relies on the hydration of the cement, which results in the formation of new hydrated binding phases that immobilize part of the water and lower the porosity [3].For the implementation of pressed earth, the stabilization with CaO-based additives offers the advantage of the presence of additional calcium cations which act as a flocculant for clay minerals [4,5].However, earth stabilized with Portland-based cements or CaO has a high CO 2 footprint (i.e. a CO 2eq ~ 0.8 t/t of cement produced [6]).Hence, earth stabilized with CaO-based cement presents the disadvantages of limited workability and rather low compressive strengths compared to concrete, while at the same time, it has a rather high environmental footprint.
To produce poured materials, deflocculation agents have been introduced [7], since they also enable significant reduction of the initial amount of water.A low water content reduces the porosity, which may result in an increase of the bulk modulus of the hardened material, while at the same time, the workability and fluidity of the fresh mixture are improved [8,9].In the past, deflocculating effects have been observed upon the addition of sodium silicate and sodium hexametaphosphate (HMP).It is generally observed that sodium disperses the clay minerals [10,11], and that HMP decreases the van der Waals attraction forces, further increasing dispersion effects [12].Although the dispersion of the clays is necessary for processability, it reduces the physical bonding and the mechanical properties.For this reason, flocculants (e.g.MgO) are subsequently added to immobilize the deflocculants, which improves the physical bonding in the finished product [13].By optimizing the physical bonding of raw earth blocks, it is even possible to double their compressive strengths [9].
Currently, MgO-based cements are under development for specific purposes such as fast repair of concrete, precasting concrete element and even nuclear waste immobilization [14].In particular, MgO-based cements containing silicate sources have shown good workability in combination with NaHMP as a superplasticizer and the formation of magnesium silicate hydrate (M-S-H) [15].In the presence of additional aluminum sources (e.g., from calcined clays), M-(A-)S-H or hydrotalcite form, enhancing the mechanical properties of MgO-silicate cements [16][17][18][19][20][21][22].The addition of carbonated sources increases the degree of reaction and the formation of M-(A-)S-H and may stabilize the hydrotalcite phases [18,23].Hydrotalcite is a so-called anionic clay [24], consisting of a Mg-and Al-based layered double hydroxide phase, with a variable isomorphic substitution of Mg 2+ by Al 3+ in the main brucite layer, generating positive charges that are compensated by hydrated anions [24][25][26][27].M-(A-)S-H, on the other hand, is more similar to a classical cationic clay: it is a hydrated nano crystalline phyllosilicate [28,29], which, however, has smaller and rounder particles compared to the classic clay platelets [28,30].It is, therefore, conceivable that a small amount of precipitated, hydrated nano-clay mineral (M-(A-)S-H) and hydrotalcite could contribute to improve the mechanical properties of poured raw earth without significantly modifying its mineralogy and microstructure.
The reactive MgO to produce magnesium silicate cements is currently obtained from the calcination of magnesium carbonates, resulting in even higher CO 2 emissions than the production of PC (i.e.~1.5-1.7 t/t vs. ~0.8t/t) [31,32].Moreover, magnesium carbonate sources are available in limited amounts [33].In order to make progress towards environmental sustainability, MgO could theoretically be produced from chemically treated Mg-silicate rocks with an additional thermal process at temperatures of 800-900 • C, which are far below those of PC clinker (~1450 • C).This latter approach avoids the calcination of carbonated rocks, thereby significantly can reducing the CO 2 emissions during production.However, the process requires a complex and energy-consuming chemical treatment, e.g. with concentrated acid or basic solutions [34,35].Alternatively, MgO can be extracted from seawater with chemical processes, including the precipitation of Mgsalts, which are transformed into Mg(OH) 2 in basic solutions followed by heat treatment to produce reactive MgO [36][37][38][39].To date, the sustainable production of pure MgO is still under investigation, in particular for applications as MgO-based cements.However, this binder could have a very low carbon footprint if it contained carbonated phases [32,[40][41][42] or if it carbonated during hardening [43,44].
In this study, we investigated, for the first time, the compressive strength at constant workability poured clay mortars containing different clay minerals, which were stabilized with CaO-or MgO-based cements.The phase assemblage of the associated pastes was studied over time by thermogravimetric analysis (TGA), X-ray diffraction (XRD), and 29 Si, 31 P solid-state magic angle nuclear magnetic resonance (ss MAS NMR) spectroscopy.Finally, the CO 2 equivalent per tons, and the estimation of the eco-efficiency per m 3 were determined for these materials to quantify their ecological footprints.

Composition of clay and clay mixes
Clay samples, identified as "kao", "ill" and "smec" and metakaolin were commercially obtained.These binders were characterized by XRF, XRD, and TGA (Table 1, Fig. 1a and b).The "kao" sample was mostly composed of kaolinite (>95 wt%) containing illite and quartz impurities.The "ill" contained ¼ of quartz and ½ of clay minerals; in this 50 wt% of clay minerals, mainly illite (~70 wt%) is present, the rest being mostly muscovite and feldspar.The "smec" sample mainly consisted of montmorillonite (i.e.dioctahedral smectite, see Supporting Information and Fig. S1) with traces of carbonates.The large amount of MgO identified by XRF in the "smec" sample indicates the presence of Mg together with Ca at the exchangeable sites. 29Si MAS NMR data (Fig. 1c) showed resonances in the chemical shift region of Q 3 species characteristic for the layered structure of clays with a rather sharp signal observed for the "kao" sample.Considerably broader signals assigned to illite and montmorillonite were observed for samples "ill" and "smec".Additionally, the signal at − 108 ppm assigned to Q 4 shows the quartz in the "ill" sample and confirmed the presence of traces of quartz in the "kao" sample.In the 29 Si NMR spectrum of metakaolin, only poorly resolved and very broad resonances covering the chemical shift region from − 80 to − 120 ppm are found, as already described in [45,46].
To mimic a typical composition of natural earths found in Europe [47] that can be used for construction, the clay minerals were mixed in a first mix (Clay 1) with proportions 35 wt% "kao", 50 wt% "ill", 10 wt% "smec" and 5 wt% of metakaolin.The metakaolin is added to mimic the presence of reactive clay that is usually present in actual excavation materials but usually not in mined clay deposits [48].In a second mix (Clay 2), we focused on a model clay system with mostly kaolinite (70 wt % "kao", 20 wt% metakaolin and 10 wt% of "ill").For more details on the clay mixes, see XRD, TGA and 29 Si MAS NMR spectra in Fig. S2.
In this study, crystalline sodium hexametaphosphate (NaHMP) was used as superplasticizer.Its content was kept constant in the stabilized earth mortars and the MB cement mortar at 4 wt%.The crystalline NaHMP (96 % purity, +200mesh) was commercially obtained from Sigma-Aldrich.The workability of the clay-PB systems decreases fast, making it necessary to increase the amount of water compared to the clay-MB systems to keep comparable workability.See section 2.4.

Cement characteristics
Portland-based (PB) cement was composed of 85 wt% Portland cement (PC) [CEM I 42.5 R] and 15 wt% semi-reactive filler limestone (LS), which were separately obtained (both materials from Heidelberg cement).After grinding of the limestone and homogenisation of the PB (see SI and Fig. S 3), the chemical composition of the cements was analysed by XRF (Table 1).
15 wt% metakaolin, 10 wt% silica fume, and 5 wt% hydromagnesite to have a similar MO/SiO 2 (M = Mg or Ca) in both PB and MB cements.Additionally, in both cements, the amount of Fe 2 O 3 + Al 2 O 3 and the carbonate contents were similar.

Preparation of the samples 2.3.1. Mortars
Stabilized clay mortar specimens were mixed according to 25 wt% of binder and 75 wt% of standardized sand.The binders consisted of 67 wt % of clays and 33 wt% of MB or PB cement.The water-to-solid ratios of the mortars were adjusted to the lowest possible water amount for sufficient workability to cast mortar samples, resulting in water/solid ratios between 0.6 and 0.8, as detailed in Table 2 and Table S 1. Mini-slump tests were conducted for mortars clay-MB and clay-PB to achieve a slump flow of about 4 cm diameter.However, it is essential to note that the clay-PB mixes became unworkable after 10 min and had to be vibrated to regain workability.
References PB and MB mortars were also cast with 25 wt% of cement (100 wt% of MB or PB cement) and 75 wt% of standardized sand.Note that the PB mortar was much more fluid due to coarser particles; the highest w/c without segregation was used for this mortar to allow a better comparison with the MB mortar.Interestingly, the use of phosphates as a superplasticizer proved to be less effective for the clay-PB systems.These seem to require a different type of superplasticizer, such as e.g.polycarboxylate ether (PCE), to adjust workability.
The mortars were cast and unmolded after 24 h, then cured in a climate chamber at 20 • C and 98 % relative humidity until the compressive strength was tested.

Pastes
To study the phase assemblage by XRD, TGA and NMR, paste samples were prepared, a mixture of 67 wt% clay and 33 wt% cement was used.For each sample, 6 g of the binder was used, consisting of 4.08 g of clay and 1.92 g of cement.3 g of a solution containing 87.5 g/L NaHMP was added.Additional MilliQ water was added (between 1 and 2 g) depending on the sample (see details in Table S 1).Cement pastes consisted of 100 wt% cement in the amount of solution as indicated in Table S 1.

Compressive strength measurement
Compressive strength was tested on 50 × 50 × 50 mm 3 stabilized clay mortar specimens with an Instron 5982 100 kN setup applying a 0.3 MPa/s loading rate.

TGA and XRD
Thermogravimetric analyses (TGA) were carried out on ground paste powder (~40 mg) under N 2 flow gas with a Mettler Toledo TGA 2 instrument using a heating rate of 20 • C/min from 30 to 980 • C.
X-ray diffraction (XRD) measurements were performed using a Panalytical diffractometer in the Bragg-Brentano (θ-θ) geometry equipped with a rotating sample stage with a CoKα radiation (45 kV, 40 mA).The samples were scanned between 5 and 90 • 2θ and reflection recorded with a X'Celerator detector.

Multinuclear solid-state NMR studies
The 29 Si solid-state magic angle spinning (ss MAS) NMR single pulse experiments were conducted on a Bruker Avance III 400 NMR spectrometer using a 7 mm CP/MAS probe at 79.5 MHz.Ground paste samples (200-300 mg) were packed into 7 mm zirconia rotors, and Teflon inserts (3 mm thickness) were used to allow smoother sample rotation.The NMR data were recorded under the following conditions: 4′500 Hz sample rotation rate, 10 k-20 k scans, 30 • 29 Si pulses of 2.5 μs, RF field strength of 33.3 kHz during SPINAL64 proton decoupling.The 29 Si NMR chemical shifts were referenced to the most intense resonance at − 2.3 ppm of an external sample of octamethylsilsesquioxane (Aldrich No. 52,683-5), which was referenced to tetramethylsilane (TMS, δ 29 Si = 0.0 ppm): for more details see [17]. 31P MAS-NMR data were acquired at 162.0 MHz on a 4 mm CP/MAS probe at rotation rates of 10′000 Hz as single pulse experiments with 2.0 μs (40 • ) excitation pulses, 58 kHz SPINAL64 proton decoupling was applied during acquisition, and the recycle times of 10 s ensured quantitative acquisition of the spectra. 31P NMR chemical shifts were referenced to the external standard of NH 4 H 2 PO 4 (solid) at 0.0 ppm.
The deconvolutions of the 29 Si and 31 P NMR spectra were carried out with the software "DMFIT" [49] as described in our previous work [22].

Greenhouse gas emission calculations
The carbon footprints in kg of CO 2 equivalent of MB and PB cements and mortars were calculated based on the mass composition and the carbon intensity of each raw material given in Table 3.The uncertainties of the values are estimated from the pedigree matrix developed by the Ecoinvent team [50,51].The kg CO 2 eq for 1 MJ is estimated from the market for 'heat, district or industrial, natural gas'-CH given in Ecoinvent v3.10.
The greenhouse gas emissions associated to the PB cement were calculated from the kg CO 2 eq/kg of clinker at approximately 0.74 kg CO 2 eq/kg, including gypsum (market for 'cement production, Portland/ CEM I' -CH) and limestone at about 0.006 kg CO 2 eq/kg ('limestone, crushed, washed to generic market for supplementary cementitious materials -Europe) according to Ecoinvent v3. 10 [52,53].
Metakaolin, a component of MB cement, exhibits variable CO 2 eq/kg depending on the fuel used [54].Recent publications [55][56][57] suggest that with energy with lower range of emissions, the production of 1 kg of metakaolin or calcined clay would result in 0.2-0.35kg CO 2 eq/kg.The data for silica fume was found in Ecoinvent v3.10 [52,53]: 0.004 kg CO 2 eq/kg with in uncertainty of 0.084.The calculations are based on economic allocation [58].
The MB cement is not industrially produced due the scarcity of MgCO 3 (critical raw material in Europe [59]) from which MgO is currently produced [32,60].However, in this calculation we will assume MgO production based on the calcination of MgCO 3 , which releases approximately 1.7 kg CO 2 eq/kg [31].Note that the uncertainty on this figure is very high: with rather clean energy (i.e. market for 'heat, district or industrial, natural gas'-CH), this number could be reduced to ~1.3 kg CO 2 eq/kg, mostly from the direct release of CO 2 from magnesite, while in Ecoinvent v3. 10 [52,53] it is estimated as 2.05 kg CO 2 eq/kg.New routes, such as those starting from magnesium silicate rocks [35,40,61,62] or magnesium-rich brines [37,38,63,64], show high potential for sustainable MgO production, with estimated emissions of 0.26 and 0.20 kg CO 2 eq/kg, respectively.As very little data are available in the literature for such processes, we applied the pedigree matrix to define the errors [65].The production of hydromagnesite strongly depends on the initial production of MgO.Scott et al. [40] estimated the The CO 2 eq/kg for the production of NaHMP was estimated to 0.44 kg CO 2 eq/kg from the chemical reaction and the heating of Na 2 H 2 P 2 O 7 (with the value given in Ecoinvent 3.10).
The carbon footprint (kg CO 2eq /t) of hydrated pastes was translated into kg CO 2eq for concrete per cubic meter (kg CO 2eq /m 3 ), estimating a composition of 0.25 volume of hydrated binders in 1 m 3 concrete and considering the kg of CO 2eq /m 3 of the water and of the aggregates in the concrete negligible compared to the kg of CO 2eq /m 3 of the cements.This transformation was necessary to obtain the CO 2 intensity [66], defined as the amount of CO 2 emitted to deliver one unit of performance, i.e. (kg/m 3 /MPa).

Workability and compressive strengths
The amount of water used during hydration of cement systems plays a crucial role in their workability.A proper workability is essential to obtain homogenous samples and to avoid large entrapped air voids.However, increasing the amount of water in the mixture also leads to a decrease in mechanical properties (e.g.compressive strength, elastic modulus) in the hardened state due to higher porosity.The MB mortar contained more water (w/c = 0.68) than the PB mortar (w/c = 0.48), even with 4 wt% of NaHMP.
Fig. 2 displays the compressive strengths measured on the different mortars.The PB mortar had higher compressive strength than the MB mortar after 3 and 42 days: 36 and 50 MPa compared to 20 and 30 MPa, respectively.As already detailed, the higher water content in the MB mortar was likely the main reason for its lower compressive strength.
The stabilized earth mortars had compressive strengths between 2 and 13 MPa.After 3 days, the compressive strength of the clay-MB systems was significantly higher than that of the clay-PB systems, likely due to the faster reaction of the MB cement and/or again to the lower amount of water needed for the same workability.Phosphates play an essential role in both systems, as they are needed for the workability of the clays.However, in the clay-PB system, they retarded PB hydration [67] and did not act as efficient superplasticizers, as confirmed by the higher water requirement.
After 42 days, the compressive strength range was 8-13 MPa for the stabilized earth mortars (Fig. 2b and c).While the MB mortars alone had lower compressive strength than PB mortars (Fig. 2a), stabilized earth mortars with MB and PB show similar compressive strength at later ages, while stabilization with MB was advantageous for early strength (Fig. 2c).
Interestingly, the strength of the MB stabilized mortars was roughly 1/3 compared to the pure MB mortar, while for the PB stabilized mortars this figure was 1/5.This observation confirms some incompatibility of the PB cement with clays (as detailed in [3]) and indicates that the MB cement seems to be a better and more efficient stabilizer.
Note that the water resistance tests on the mortars after 1 week (see Fig. S4) showed a good water resistance: <0.2 % in the mass of the samples after 7 days.

Phase assemblage of pure cements
Fig. 3 displays XRD and TGA data of the pure hydrated cement pastes (between 3 days and 1 year of age) and of unhydrated cements.The hydration of the PB system (Fig. 3a) showed the classical reactions of the clinker phases with gypsum, forming various minerals with XRD reflections assignable to ettringite, portlandite, C-S-H, and SO 4 -AFm (monosulphates).The calcite initially present in the PB system was easily observed with the characteristic TGA mass loss at 700-800 • C (Fig. 3c).Over time, the SO 4 -AFm phase (monosulphate) converted into CO 3 -AFm (hemi and monocarbonates), as indicated by the decrease in the calcite content (Fig. 3a).
In the case of MB, the hydrates observed by XRD in Fig. 3a were brucite -Mg(OH) 2 (main reflections at 21 and 44 • 2θ, resulting from MgO hydration), and hydrotalcite (resulting from the reaction of metakaolin).After 3 days, no more reflections of hydromagnesite were observed, confirming that this phase is only partially stable [68] and reacts rapidly.However, MgO (with the main reflection at 50.2 • 2θ) was still present in the sample after 1 year, probably due to the saturation of the NaHMP (see mix composition in 2.4.1.)in the samples, which limits hydration [69].The amount of hydrotalcite (003 reflection at 13.3 • 2θ) increased and/or its crystallinity increased over time, while the brucite reflections slightly decrease.This potentially indicated the transformation of brucite into hydrotalcite over time.TGA data (Fig. 3d) confirmed the presence of brucite and/or hydrotalcite with the characteristic dihydroxylation water loss around 400-450 • C [17].Finally, the water losses below 200 • C in the TGA indicate loosely bound water, probably related to M-(A-)S-H in the system [16] as no other phases containing poorly bound water were present.
The 29 Si ss MAS NMR spectra of the pure cement systems after 42 days and 1 year of hydration are shown in Fig. 4. The 29 Si ss MAS NMR spectra of the hydrated PB cement pastes showed the presence of Q 1 , and Q 2 , which are characteristic for C-(A-)S-H [70][71][72][73][74][75], indicating a significant amount of this hydrate.After 42 days, the presence of Q 0 , assignable to C 2 S and C 3 S (belite and alite phases) suggested that the hydration process was still incomplete.After 1 year, however, the Q 0 resonance was negligible.
The unreacted MB cement contained metakaolin and silica, with characteristic broad signals of Q 3 and Q 4 at − 100 and − 110 ppm, respectively.After 42 days, the 29 Si ss MAS NMR spectra revealed that about 1/3 of the Si species were still present as unreacted silica and metakaolin, certainly due to the overdosage of NaHMP that slows down the kinetics of reaction [15,22].However, the signals Q 1 , Q 2 , and Q 3 characteristic for M-(A-)S-H can be observed very clearly [16,18].Moreover, since they were rather narrow, only minor Al incorporation in the silicate layers of M-(A-)S-H may be inferred [18].
In summary, TGA, XRD, and NMR data showed that both types of pure cement systems (PB and MB) exhibited incomplete hydration after 42 days.After 1 year, they had undergone significant hydration, with the formation of C-(A-)-S-H/Ettrigite/AFm as the main hydrate phases in the PB-based systems and of M-(A-)S-H/hydrotalcite, in the MB-based systems.

Phases assemblage of the MB-clay and PB-clay binders
XRD and TGA data of the unhydrated and the hydrated clay 1cement systems from 3 days to 1 year are presented in Fig. 5 (data of the clay 2-cement systems are shown in Fig. S 3).The XRD analysis revealed the presence of remaining MgO (reflection at 50.2 • 2θ) in the clay-MB systems (Fig. 5a), which may be partly due to the high concentration of NaHMP that covers the MgO surface and limits its reaction [22].TGA and XRD data further confirmed the presence of hydrotalcite in all samples, as evidenced by the prominent reflection at 13 • 2θ and a water loss observed around 420 • C in the TGA.However, it should be noted that the water loss in this temperature range could also be attributed to brucite dehydroxylation, since brucite was also present in the samples, as indicated by XRD peaks observed at 22 • and 44 • 2θ.It can be noted that the amount of brucite is much lower than in the pure paste potentially attributed i) a pozzolanic reaction in the case of the clay minerals mixed with MB cement, and/or ii) the precipitation of Mg in other phases.
In the clay-PB system, there was clear evidence that hydration of the PB cement was also occurring: XRD reflections of C-S-H, ettringite, AFmphases were observed, even though the water loss observed at 180 • C characteristic of the AFm-phases was not really observable after 3 days.This lower early reaction is potentially due to the presence of phosphates [67].Portlandite, which has a characteristic water loss at 490 • C and reflection at ~21 • 2θ (Fig. 3), was not observed in the clay -PB systems.This could be related to i) the pozzolanic reaction of the portlandite with metakaolin or, more generally, with the reactive clay minerals, or ii) the presence of the phosphate, which inhibits its formation, similarly to the inhibition of brucite formation in the MgO systems [69].In both systems, the XRD analysis showed no significant changes in the structure of kaolinite and illite, as there were no specific variations in their reflection peaks within the error of the Rietveld quantification.
Interestingly, in the clay-MB system, the first reflection of the montmorillonite (occurring below 10 • 2θ) was not observed, while it was slightly visible in the mixes clay-PB system.The hydration stoppage treatment cannot be responsible for the complete absence or the strong reduction of the montmorillonite phase, as the observable reflection of the interlayer distance did not change when the hydration stoppage of the montmorillonite was carried out without the cement (see Fig. S 5).Partial dissolution of the smectite phase could occur due to the high pH induced by the cement; this has been observed, for example, at the interface bentonite-cement in the context of waste encapsulation [76,77].Here, the pH value of the clay-MB system is controlled by the presence of brucite and buffered at 10.5, while the pH of the clay-PB system is buffered by the presence of portlandite at 12.5.If the reflection of the interlayer distance in the smectite were solely related to the pH, the reflection would be more prominent in the clay-MB system and less visible in the clay-PB system.The rather surprising observations reported here suggest that the MB cement affects the smectite, which will require further investigations.
The 29 Si ss MAS NMR data (Fig. 6 [16,18,[70][71][72][73][74][75].In the PB-clay systems, the ratio of Q 1 to Q 2 decreased with time, which may indicate a lower Ca/Si ratio in the C-A-S-H formed and a larger amount of silicate in the C-A-S-H phases.This indicated the presence of a higher amount of silicate in the hydrated products, possibly due to a partial reaction of the clay component.It is difficult to draw firm conclusions based on the structure of M-(A-)S-H in the MB-clay systems, since their Q 3 signal in 29 Si NMR strongly overlaps with the signal of the kaolinite.
Overall, the 29 Si ss MAS NMR data complemented the TGA and XRD information, proving the formation of M-(A-)S-H and C-(A-)S-H in addition to the other hydrates and the clay minerals.

The fate of phosphate
The 31 P ss MAS NMR spectra of stabilized clays together with the raw clays and pure PB and MB hydrated cements are shown in Fig. 7.The 31 P NMR spectra of PB hydrated cement showed as expected a very limited amount of phosphate, mostly present as orthophosphate with characteristic signals at 2 to 3 ppm.
The 31 P MAS NMR spectra of the MB hydrated cement showed high similarities with a MgO-MK sample from [22], in which the phosphate seemed to be mainly present as orthophosphate with signals at 2 to 3 ppm and pyrophosphate (or hydrogen phosphate) − 8 to − 9 ppm, with some remaining Q 2 signals (< 8 %) observed at around − 24 ppm, indicating the presence of some remaining phosphate starting material.The raw clays in presence of NaHMP were also analysed and the 31 P ss MAS NMR data showed that >50 % of the added phosphate stayed in the ring configuration, while the rest was modified to ortho-or pyrophosphate.The comparison of the 31 P NMR data of stabilized and nonstabilized clays indicated that in presence of the cements, there was some opening of the phosphate rings and that the orthophosphate sites are related to their binding with the cement phases.The 31 P NMR signals of ortho and/or pyro-phosphates from the clays stabilized with MB cement appear at very similar chemical shifts as observed for hydrated pure MB cements (small at ca. 4-0 ppm and main at ca. -4 ppm), while for clays stabilized with PB cement, only one main resonance centred at 2 ppm is observed.This could be related to the Ca-phosphate complexation at the surface of Ca-rich phases [78] or even to a precipitation of poorly-or nano-crystalline hydroxyapatite (Ca 5 (PO 4 ) 3 OH) [79] or simply to the formation of amorphous calcium phosphate [80].
The chemical and mineral investigations on the pastes from 3 days to 1 year clearly showed the hydration of both cement types, with M-(A-)S-H, brucite and hydrotalcite as well as C-(A-)S-H, AFm and AFt phases being observed in addition to clay minerals.At early ages, the PB system showed a delayed hydration rate due to the presence of phosphate known to slow down hydration [67].In all clay-MB systems, MgO was still detectable due to the overdosage of phosphate, but the phosphate presence was rather beneficial compared to the PB system, as it accelerated the formation of the Mg-hydrates [22,81].
The analysis also showed no significant alterations in the structure of the kaolinite and illite, as no variations were observed in XRD reflection peaks of the two minerals.The absence of the first reflection of smectite in the MB-stabilized mixes, in contrast to its faint visibility in PBstabilized mixes, is not yet understood.

Greenhouse gas emissions
The calculated carbon footprints in kg of CO 2 equivalent of MB and PB cements are presented in Table 4.The PB cement (composed of 85 wt % CEM I (including 4.25 wt% gypsum) 15 wt% limestone) exhibited a carbon footprint of 631 kg CO 2 eq/t.this number is slightly higher than the one from the CEM II/A-LL (composed of 70 wt% CEM I, 5 wt% gypsum, 25 wt% limestone) which is about 585 CO 2 eq/t given by the GCCA [82], this is related to the relatively high amount of clinker used in our study.For the MB cement, this figure was about twice as much, 1290 kg CO 2 eq/t, due to the substantial carbon footprint associated with MgO obtained from MgCO 3 (1700 kg CO 2 eq/t [31]).However, in the case sustainable MgO would be obtained from magnesium silicate rocks or magnesium-rich brines with processes currently under development [35,37,38,40,[61][62][63][64], the carbon footprint of MB could be reduced to about 240 kg CO 2 eq/t.This would be 1/5 of the current footprint and slightly less than half of the footprint of the PB cement.
Fig. 8a illustrates the compressive strength of each material plotted against its carbon footprint, while Fig. 8b translates this data into CO 2 efficiency.By itself, the PB cement showed a good CO 2 efficiency (~2.6 kg CO 2 eq/m 3 /MPa) due to its high compressive strength.This data was within the range reported in literature [66].For the commonly used CEM II/A-LL [82], both its compressive strength and the greenhouse gas emissions are reduced compared to CEM I by the addition of limestone, which results to similar CO 2 efficiency to CEM I [66].
In contrast, the MB cement, currently hindered by sourcing MgO from MgCO 3 and by its lower compressive strength (30 MPa), exhibited a poor eco-performance (~10.8 kg CO 2 eq/m 3 /MPa).However, the potential shift in MB cement production methods, relying on MgO from Mg-silicate rocks or Mg-rich brines, could significantly reduce its CO 2 efficiency to approximately ~2 kg CO 2 eq/m 3 /MPa, which would be even lower than for PB.Note that the uncertainty of these estimations is very high, and based on the uncertainties, it might vary from the 1 to 3.5 kg CO 2 eq/m 3 /MPa.Additionally, the transport was not taken into account in this estimation, and this will depend on where the materials are produced, compared to Portland cement which is mostly locally available, e.g. in Europe.On a short distance (100-200 km) and with rail (i.e.excluding heavy goods vehicle) this transport should have a minimal impact compared to emission of the cement production.
The compressive strength of MB systems was proportional to the amount of MB cement added, resulting in a reduction by the same factor as the carbon footprint.This implies that the CO 2 efficiency of MB stabilized mortars remains consistent with the CO 2 efficiency of the MB cement alone, irrespective of the MgO source.This indicates that MB cement or MB stabilized mortars have a CO 2 efficiency comprised within 8.8 and 12.7 kg CO 2 eq/m 3 /MPa.However, if MgO could be produced from Mg-silicate rocks, the clay-MB mortars studied in this paper would have CO 2 efficiencies lower or equal than PB (1.8 and 2.6 kg CO 2 eq/m 3 / MPa).
Clay-PB mortars, on the other hand, with strengths reduced by a factor of 5 compared to PB mortar, demonstrate an almost twofold (depending on the type of the clay) eco-performance compared to the PB cement alone (about 3.9 and 6.3 kg CO 2 eq/m 3 /MPa).This highlights the incompatibility of PB with clays and confirms the limited appeal of PB cements when clays are involved.

Conclusions
For the first time, two systems based on clay minerals with MgObased (MB) cement and clay minerals with CaO-based cement have been compared.The mortars examined in this study revealed similar compressive strength, with the MgO-based cement demonstrating even better performance at early ages (before 7 days).This improvement was primarily attributed to the large amounts of superplasticizer (NaHMP) required for the PB cement, which delayed the hydration reactions.In addition, also the higher degree of fineness of the MB cement compared Fig. 4. 29 Si ss MAS NMR spectra of the PB (upper part in red) and MB (lower part in green) cement pastes after 42 days (dashed lines) and 1 year (plain lines).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)to the PB cement might have played a role in the early strength.The development of the mineralogy over time showed that both cements in the presence of the clays reacted with the metakaolin and potentially with some of the reactive clays.
At later ages, the compressive strength of the two types of stabilized mortars was similar.It is important to note that MB appeared to be a more efficient stabilizer for the clay mortars.In fact, the strength of the clay-MB mortars was roughly 1/3 that of the pure MB mortar, while the PB stabilized mortars had 1/5 of the strength of the pure PB mortars.Additionally, stabilization of clays with PB is only possible with a large amount of superplasticizer to compensate for the loss of workabilitythis further increases the carbon footprint.With the loss of strength, the CO 2 efficiency also largely increases.Hence, the PB cements should not be used as stabilizers in clay-based mortars.
The sourcing of MB cement from magnesite is associated with a high carbon footprint and there is therefore currently no advantage over PB cement.However, a significant opportunity will emerge if the production of MB shifts to the use of Mg-silicate rocks or Mg-brines as sources of magnesium.In such a scenario, high purity kaolinite stabilized with MB cement showed the best CO 2 efficiency, which is about 1.8 kg CO 2 eq/ m 3 /MPa.More realistically, an average European clay, composed of ~40 wt% kaolinite, ~50 wt% of illite and ~ 10 wt% swelling clay showed the same CO 2 efficiency as PB cement, about 2.6 kg CO 2 eq/m 3 / MPa, indicating a potential environmental advantage that could be harnessed with a shift in production methods of the MB cement.In addition, the further use of excavation materials, currently considered as wastes, as a source of the clays would contribute to reducing the environmental footprint of these binders.Additionally, this study proves the good compatibility of the MB with clay minerals; based on this evidence, it may become possible to employ MB with unwashed sand and gravel containing clays, which is currently impossible in PB concrete [83].
Finally, dispersants are usually used for casting poured earth to obtain the required workability.Currently, research is mainly focused on the use of NaHMP.This study showed the good compatibility and synergy of NaHMP with clays in combination with MB cements.The phosphate improves the rheology, but also enhances the hydration kinetics of the MB cement with the formation of M-(A-)S-H and hydrotalcite and a later precipitation of phosphate [13,22].However, to avoid the competition with the global food production [60], the use of phosphate-based dispersants should be restricted and dispersants based on sustainable raw materials should be developed.The overall results sound promising and further studies with actual earth materials should be extended to both mortar and concrete applications.

Table 4
carbon footprint of the cements, calculated from the mass composition of each cement and the carbon footprints given in Table 3.
and Fig. S 6) revealed the presence of a small amount of M-(A-)S-H in the MB-clay systems and a small

Fig. 2 .
Fig. 2. Compressive strengths of the different mortar specimens after 3, 7, and 42 days of a) the pure binders and b) clay-cement systems.
The 31 P ss MAS NMR data of the clay 1 -MB system showed orthophosphates and probably pyrophosphates with signals from 10 to − 15 ppm, and again the Q 2 signal intensity observed at − 24 ppm indicates

Fig. 6 .
Fig.6.29 Si ss MAS NMR spectra of the a) Clay 1 stabilized with MB cement compared to the pure MB cement after 1 year, b) Clay 1 stabilized with PB cement compared to the pure PB cement after 1 year; *LC = linear combination of the clay and the hydrated cement after 1 year.

Fig. 7 .
Fig. 7. 31 P ss MAS NMR spectra of the Clay 1 stabilized with MB or PB cement compared to the pure MB or PB cement after 1 year, and the pure clays mixed with NaHMP.

Fig. 8 .
Fig. 8. a) Compressive strength versus carbon footprint, b) CO 2 intensity of each binder.The error bar for the CO 2 intensity is the max of the low carbon footprint*x2/(MPa + standard deviation) and low carbon footprint*x2/(MPa-standard deviation); * = calculated from the uncertainty factor.

Table 1
Chemical composition determined by XRF of the clay samples and cements used (wt%).n.d.= not determined.
a MB = MgO-based b PB = Portland-based.c TC = total carbon.

Table 2
Mixture composition of the different mortars.Cement and Concrete Research 186 (2024) 107655energy for the production of hydromagnesite from Mg(OH) 2 to be 8.17 MJ per kg of CO 2 stored.Hence, two values for hydromagnesite are given, either 0.617 or − 0.002 kg CO 2 eq/kg for MgO produced from MgCO 3 or Mg-silicate rocks, respectively.
* Details given in TableS 1.E.Bernard et al.

Table 3
Carbon footprint of the raw materials, per Global Warming Potential (GWP 100a).