MgO‐based binders

To limit global temperature rise to below 2°C, we need to radically and rapidly change the way we build and use materials, since construction is responsible for 20‐40% of industrial CO2 emissions. Magnesium carbonate‐based cements have the potential to become a major carbon sink in construction industry, as CO2 will not be emitted during their production, but CO2 will rather be bound during hardening. Two different reaction mechanisms lead to setting and hardening of such HMC cements: i) hydration of MgO‐Mg‐carbonate, also in blends with silica and other mineral additions, in the presence of water or salt solutions (such as sodium bicarbonate solution) at ambient conditions or ii) carbonation hardening of MgO‐based systems at increased CO2 partial pressure and/or at increased temperatures.


Introduction
Concrete and other cement-based materials are, with the exception of water, the most consumed materials globally.The production of cements with Portland cement clinker as the primary active component requires heating of limestone and clays to 1450 °C.Limestone, CaCO3, used as main calcium source, decarbonizes during heating releasing CO2 such that CO2 emissions related to concrete are 'hard to abate' due to the nature of the raw materials.As cement based binders are employed in vast quantities, 4.1 Gt in 2019 [1], cement manufacture accounts for as much as ≈8% of the global anthropogenic CO2 emissions [2,3].
In the last decades, cement industry significantly reduced CO2 emissions by the use of alternative fuels, increased energy efficiency, and by substituting Portland cement clinker with supplementary cementitious materials (SCMs).The most suitable SCMs are latent-hydraulic or pozzolanic materials such as granulated blast-furnace slags (herein "slag") and coal combustion fly ashes [2,4].However, the availability of slag and fly ash with adequate quality is limited to only ≈ 20 wt.-% of global cement production [2] and this fraction will decrease further in the future.The recently explored calcined clays are abundant, but only capable of replacing in combination with limestone max.approx.50% of the clinker in the so-called LC3-cements [5].The sole replacement of limestone is limited to 5 to 20 wt.-% because of its restricted contribution to the cement performance [6,7].Thus, "alternative clinkers" with fundamentally different chemistry are needed to replace partially Portland cementbased cements [8,9].However, most of the proposed alternative materials such as e.g.calcium sulfoaluminate ce-ments [10,11] or belitic cements [8], are based on calcium silicates and aluminates, which strongly limits the maximum possible CO2 reduction, as limestone remains the main calcium source.
Although magnesium is abundant in the earth crust, magnesium is little used up to now for cement production.Magnesium oxychloride cements, also known as "Sorel cements", magnesium oxysulfate and magnesium phosphate cements are used in niche applications.In recent years, magnesia (MgO)-based cements containing silica and/or carbonates have gained increasing attention.Such cements based on magnesium carbonate and silicate hydrate have low pH values, can have high strength and their production is potentially associated with low CO2 emissions.In contrast to CaO-based cements, where CaCO3 is decarbonated during production, the raw materials for MgObased cements can be gained from desalination brines or magnesium silicates, without direct emissions of CO2 from the raw materials.However, compared to Portland cements, relatively little is known about these cements and their durability.This paper summarises the present understanding of the factors affecting the hardening of such cements, points out important gaps in our knowledge and potential future research routes.This reaction can lead to compressive strengths of up to 50 to 70 MPa after 28 days, comparable to Ca-based cements [12][13][14][15].Among the different strategies to reduce the CO2 in cement industry, carbon capture and storage (CCS), and particularly the cementitious binders based on carbon capture and utilisation (CCU) are of interest.Cements based on hydrated magnesium cements, HMC, could be an attractive alternative due to the potentially negative CO2 emissions.Such HMC binders harden and gain mechanical strength from the reaction of reactive MgO to brucite Mg(OH)2(s) and HMC: This reaction does not generate any CO2 but consumes CO2, leading to a permanent sequestration of CO2 and to strength gain in HMC cements.

MgO sources
At present, magnesia is mainly produced by calcination of magnesite, MgCO3, in a process similar to the production of CaO from limestone.As the calcination of MgCO3 releases 1.1 kg CO2 per kg MgO (or 0.5 g CO2 per kg MgCO3), the production of MgO from magnesite or dolomite, CaMg(CO3)2 is not sustainable.Magnesium, however, is not only present as carbonate rocks (magnesite, dolomite) at the earth's surface but also as magnesium silicate rocks.
The largest natural deposits available are olivine, (Mg,Fe)2SiO4, and serpentine, Mg3Si2O5(OH)4.Serpentines are more abundant than olivine and originate from the partial hydration of olivine near the Earth's surface.Mgsilicates are available worldwide [16,17] as illustrated by Figure 1.However, magnesium silicates react only very slowly with water and CO2.Therefore, in a first step they need to be decomposed to MgO, SiO2, Fe2O3 and minor quantities of Al2O3, depending on the composition of the raw material.
Recent breakthroughs in hydrometallurgy make economical extraction of high-purity MgO from Mg-rich minerals such as olivine and serpentine possible [17][18][19].Silica and iron compounds are the main by-products of the extraction of MgO from olivine or serpentine, while Al2O3 as well many other elements are present as minor component [17,18], as illustrated in Figure 2. Both, the MgO as well as the SiO2-rich by-product, will contain in addition either SiO2 or MgO, together with Al, Fe, K, Na, Ca and many other elements [17].The (partial) re-use of the SiO2-rich by-product in the production of HMC cements would contribute to close the material cycle and avoid waste deposition.SiO2-rich by-products can also be used as supplementary cementitious materials in PC blends, where they show comparable properties to silica fume [20].Similarly, other, little used industrial by-products such as pyro-metallurgical slags [21] or bauxite residues [22] could possibly be used in HMC.The use of industrial by-products however, will also increase the amount of minor elements present, such as Zn 2+ , Co 2+ or Pb 2+ .
MgO can also be produced from seawater, resp.desalination brines, via precipitation of brucite at controlled pH.In the first step of the process, brucite is precipitated by adding alkaline bases such as sodium hydroxide, ammonia or lime [23][24][25][26][27].The obtained brucite can on the one hand be calcined at rather low temperatures of 450-550°C to obtain reactive magnesia for further use in construction products [24,25], as discussed in more detail in the next section.
On the other hand, brucite can also be carbonated directly to obtain hydrated magnesium carbonates (HMC) such as nesquehonite, dypingite and hydromagnesite [23,26].In another approach, sodium carbonate/bicarbonate solution can be used to directly precipitate HMC from seawater [28,29] resulting in the formation of nesquehonite and/or hydromagnesite like-products depending on the experimental conditions.Calcium, which is part of the brines as well, precipitates as calcium carbonate monohydrate and/or dolomite.The obtained HMCs could be used in blends with MgO to make hydrated magnesium carbonate cements (as discussed in the next section).The obtained nesquehonite phase also can be partially dehydrated at moderate temperatures (≈ 100-250°C) [30].Upon rehydration, the dehydrated nesquehonite develops a compressive strength of up to 5 MPa, which could be used to manufacture construction materials similar to plasterboards [31].Both approaches do not release chemically bound CO2, as they do not start from carbonates.Thus, magnesium-based binding systems show a high potential to become a major binder and to enable a leap closer to the carbon-negative cements.

Hydrated magnesium carbonate cements
Two different reaction mechanisms lead to setting and hardening of such hydrated magnesium carbonate (HMC) cements: i) hydration of MgO-Mg-carbonate blends with water or salt solutions (such as sodium bicarbonate solution) at ambient conditions [32] or ii) carbonation hardening of MgO-based systems at increased CO2 partial pressure and possibly at increased temperatures [33], see Figure 2. Carbonation hardening is of potential interest for pre-fabricated products (precast), and has the added advantage that it consumes large amounts of CO2 in the hardening process (CO2 negative binder).The potential value of a HMC binder based on the hydration of MgO-Mgcarbonate is much greater, as it can be used to make ready-mix concrete for on-site applications.
The hydration of magnesium oxide in the presence of hydromagnesite (Mg5(CO3)4(OH)2・4H2O) results in a poorlycrystalline form of brucite, which is thought to be the cohesive phase in MgO-hydromagnesite blends [32,34].Carbonation hardening at higher CO2 pressures can lead to faster reaction and the formation of different HMCs such as hydromagnesite, dypingite, nesquehonite or lansfordite, depending on CO2 pressure, relative humidity, temperature and time [33,35,36], although the exact conditions are not yet clarified.
Despite the acceptable performance of HMC cements, unhydrated MgO and uncarbonated brucite have been observed in HMC binders [32,36,37], which leads to a low efficiency of carbon utilization as well as to the risk of late expansion due to the continued reaction of MgO and brucite.In addition, the stability of the different HMC phases depends on CO2 pressure, relative humidity, and temperature, which could lead to serious long-term instabilities as a conversion of water-rich HMC such as nesquehonite or lansfordite to HMC with less water per Mg such as hydromagnesite would lower the volume of such a binder by more than 40% [35], leading to cracking and destabilization in the long-term.Furthermore, CO2 is released during this conversion [19], which also can cause volume changes and thus internal cracking.
HMC cements suffer from poor a workability due to a high water demand such that a high water/cement ratio is needed resulting in limited strength, slow reaction kinetics and slow strength gain, which makes their use in construc- tion not yet practical.The long-term development of the hydrate assemblage and their mechanical properties and the expected changes upon interaction with the environment are not yet investigated.

Magnesium silicate cements
The mixing of MgO with silica and water results in the rather formation of magnesium silicate hydrates (M-S-H).
The formation of M-S-H has also been observed at the surface of cements in contact with solutions with high magnesium concentrations such as seawater or the interstitial solution of clayey rocks [38,39].Magnesium from the seawater or clay water reacts with the amorphous silica from the degradation of C-S-H in cement leading to the formation of M-S-H [40].M-S-H has also been observed in Roman cements rich in chert and dolostone [41] indicating a long-term stability of M-S-H.Electron microprobe, SEM/EDS and TEM analyses indicate that M-S-H contains not only magnesium, silica, hydroxides and water but that aluminium, calcium and iron can be incorporated in its structure [42,43].
M-S-H prepared from MgO and silica in water or carbonate solutions shows Mg/Si ratios from ~0.8 to ~1.5 after long equilibration time [12,44,45].The presence of carbonate accelerates the reaction of the intermediate brucite to M-S-H [12].The Mg/Si ratio in M-S-H is limited by brucite formation at high Mg/Si and by the presence of amorphous SiO2 at low Mg/Si ratio.
Detailed studies of synthetic M-S-H indicated a sheet silica structure similar to phyllosilicates [44,45]. 29Si NMR data indicate that silica is arranged in tetrahedral sheets next to magnesium oxide in octahedral sheets (Figure 3), comparable to the structure of phyllosilicates such as talc or antigorite.M-S-H, however, is less ordered, has very small coherent areas as visible by the high fraction of Q 2 signals, contains significant amounts of water and hydroxide groups and has a high specific surface area [44][45][46].
Figure 3 29 Si MAS NMR spectra of the synthesized M-S-H samples with total Mg/Si ratios ranging from 0.4 to 1.7.Effective Mg/Si in C-S-H ranged from 0.8 to 1.4 due to the formation of brucite at high Mg/Si and the persistence of silica at low Mg/Si.Reproduced with permission from [45].
The formation of M-S-H from MgO and silica fume in water is rather slow such that significant amounts of brucite form as intermediate product.Brucite reacts only very slowly further to M-S-H as the dissolution of brucite is kinetically hindered by silicates [44].The reaction of brucite can greatly be accelerated in the presence of hexametaphosphate or carbonates [12,14,47].
M-S-H has, such as low Ca/Si C-S-H, a negative surface charge which decreases further with increasing pH.The negative charge is partially compensated by the presence of exchangeable Mg 2+ on the surface; the effect, however, is relatively weak as the Mg concentrations in equilibrium with M-S-H are generally low [46].In the presence of calcium, Mg 2+ can be replaced by Ca 2+ .Also aluminium has been observed to be taken up in M-S-H [48].Aluminium is present in both the tetrahedral silica and octahedral magnesium oxide sheets [48][49][50][51].Aluminium distributes relatively evenly between tetrahedral and octahedral sites, such that no significant difference in charge can be observed independent on the amount of aluminium present.
In the presence of aluminium and carbonate, nitrate or higher hydroxide concentrations also the formation of hydrotalcite has been observed [49,51,52].The formation of hydrotalcite increases the volume of solid phase formed, lowers the porosity and thus improves the mechanical properties [13,[53][54][55].
Long-term and durability studies of M-S-H cements are not reported in literature.However, M-S-H cements show a high resistance against carbonation [12,13] as also underlined by the observation of M-S-H in some Roman cements [41].
Similarly to the HMC cements, the M-S-H cement have a high water demand resulting in limited compressive strength.The use of hexametaphosphate does not only accelerate the brucite reaction, but also lowers the water demand resulting in compressive strengths of M-S-H pastes and concretes similar to Portland cements after 28 days [14,15].

Thermodynamic stability of hydrated magnesium cements
Thermodynamic modelling can be used to predict the kind and volume of the solid phases in hydrated cements as a function of MgO reaction degree, SiO2, Al2O3 and iron content, temperature and CO2 pressure, which helps to optimise solid volume, long-term volume stability and durability and to identify composition with maximum CO2 binding.
For reliable prediction of stable mixes and compositions, a good and adequate thermodynamic database will be of uttermost importance.

HMC
Thermodynamic data for several hydrated magnesium carbonates are known as illustrated by the data summarised in Table 1.The thermodynamic data of brucite and periclase are well known [56,57].Accurate thermodynamic data for hydromagnesite, dypingite and nesquehonite have been published in last decade [58,59].Data for other magnesium carbonate are less reliable, as they are either estimated [60] or indirectly calculated from calorimetric measurements [61]: artinite, MgCO3•H2O, barringtonite, lansfordite and MgCO3•6H2O.In addition, the data for many other magnesium carbonates such pokrovskite:  [32,34] are not known at all.
Thermodynamic modelling can be used to calculate the relative stability of the different HMCs at different temperatures or CO2 partial pressures [60] and as illustrated in Figure 4 and Figure 5.Under all conditions magnesite (MgCO3) is most stable, which is only observed at temperatures well above ambient temperatures [62][63][64]: Magnesite formation is kinetically hindered by the strong hydration shell around the small Mg2+ ion resulting in a slow kinetics of Mg 2+ •6H2O dehydration as a relatively high free energy is needed to dehydrate Mg 2+ •6H2O to the anhydrous Mg 2+ [63,65].Calculated based on the thermodynamic data compiled in Table 1.
While the formation of magnesite is not expected due to its extremely slow formation kinetics, other HMC such as hydromagnesite, artinite, and dypingite together with (metastable) brucite could precipitate at above ambient temperatures, while at lower temperature rather nesquehonite or lansfordite can be expected to form (see Figure 4) depending on the exact temperature conditions, time, pH values and CO2 partial pressure.
Table 1 Thermodynamic data of selected inorganic phases in the MgO-CO2-H2O system.Values are given relative to 25°C and 1 bar, adapted from [12].

Phase
Chemical formula log K° a ΔfG° [kJ/mol]    The different water, CO2 and hydroxide contents of HMC lead to largely different molar volumes of these solids (expressed per 1 Mg in their chemical formula), and thus to large differences in solid volume, which can potentially lead to serious long-term instabilities in some HMC cements [35].Both the late destabilisation of MgO or brucite to water rich HMC such as nesquehonite or lansfordite as well as their potential destabilization to hydromagnesite would change the solid volume of such binders by more than 40% ( [35], Figure 6), possibly leading to cracking and destabilization in the long-term.It should be noted that the potential conversion of lansfordite (MgCO3•5H2O) or nesquehonite (MgCO3•3H2O) to hydromagnesite (Mg5(CO3)4(OH)2•4H2O) or dypingite (Mg5(CO3)4(OH)2 •5H2O) would also liberate some CO2.
Such destabilisation problems may be tackled either by directly stabilising some HMC, e.g.enabling the direct formation of hydromagnesite (instead of forming initially nesquehonite) based on the inclusion of other cations or anions to enable the formation of a solid solution [66], which can stabilise solid phases.Alternatively also the presence of other magnesium or carbonate containing hydrates such as M-S-H or LDH phases, can shift the relative stability of HMC by changing pH values, magnesium and carbonate concentrations [12,13].1.

M-S-H and hydrotalcite
Preliminary thermodynamic models to describe the solubility of M-S-H and the uptake of aluminium in the main layers and of calcium as an exchangeable cation have been published [13,38,45,48,67], although a robust thermodynamic solid solution model taking into account the structural knowledge is still missing.Thermodynamic models for the uptake of other elements such as Fe 2+ , Fe 3+ , Na + , K + , and Zn 2+ in M-S-H are not available due to the lack of systematic experimental data.
LDHs have a layered structure consisting of a positively charged brucite-like main layer, with variable Mg/(Al+Fe) ratio [68][69][70][71]: [Mgl-x(Al,Fe)x(OH)2] x+ [Ax/n n-mH2O] x-, with 0 < x < 0.33 and A n-representing an anion.Mg 2+ can be replaced by Zn 2+ , Co 2+ , Fe 2+ and other bivalent cations [72][73][74].The isomorphic substitution of Mg 2+ by Al 3+ or Fe 3+ in the main layer generates positive charges, which are compensated by anions in the interlayer region [68,75].Generally, divalent anions such as CO3 2-, SO4 2-are preferred over monovalent anions such as OH − , Cl − , or NO3 − , with a clear preference for carbonates [74][75][76].While crystalline LDHs are frequently studied and used as adsorbents, catalysts and anion exchangers [77], their solubility is not well-known.The variable composition in terms of Mg/(Al+Fe) ratio in the main layer and in interlayer anions, make the development of adequate thermodynamic models for LDH phases very challenging as also visible in the large difference between the few reported solubility products [76,78].Even less data are available for Fe(II)-, Fe(III)-and Zn-containing LDH [70,71,79,80].These large variations and missing data limit the potential to use thermodynamic modelling to predict conditions where LDH might be stabilised.
For only recently described or still unknown phases, such as amorphous magnesium silicate [81], amorphous magnesium carbonate hydrate [32,34] or magnesium-based zeolitic precursors, neither the composition, nor structure and solubility data are available and will need to be measured or estimated as far as relevant.

6
Blended HMC and M-S-H cements

Effect of SiO2 on HMC cements
The (partial) re-use of the SiO2-rich by-product (see Figure 2), of pyro-metallurgical slag or bauxite residue in the production of a HMC cement would contribute to close the material cycle and avoid waste.The formation of M-S-H in pure systems has been intensely investigated in the last years, e.g.[45,48,49], while much less is known about the effect of magnesium silicate hydrates (M-S-H) and LDH on different HMCs or on the potential stabilising effect of iron (II/III), aluminium and alkali ions on HMC at ambient conditions.
The presence of SiO2 will lead to the formation of magnesium silicate hydrates (M-S-H).Compressive strengths up to 55-70 MPa have been reported for M-S-H mortar and concrete specimens [15,53], i.e. well above the strength required for most construction applications.Additional carbonation can increase compressive strength up to 90 MPa after 180 days.Such magnesium silicate hydrate cements are stable between pH 9 -13 [45,82] making them suitable for non-structural and potentially structural applications.The addition of alkali carbonate solutions seems to accelerate to their setting and leads in addition to M-S-H also to the formation of yet unidentified HMCs [12,83].

Effect of Al2O3 and SiO2
The performance of HMC cement may also be improved by adding Al, which has received little attention so far.The combination of MgO with metakaolin leads to comparable compressive strength as MgO plus silica fume [13,53].For sodium-carbonate activated Mg-cements a significant increase in compressive strength is observed in the presence of Al (Figure 7, [13]) confirming that Al in fact plays an important role in HMC cements.The presence of aluminium leads to the formation of M-S-H with Al (M-A-S-H gels) [49,53,84], where Al 3+ can replace Mg 2+ in the octahedral sites and Si 4+ in the tetrahedral position [48].In addition, layer-double hydroxides (LDH) can form [13], in many cases stabilised by CO3 2-in the interlayer, which is strongly preferred over SO4 2-or monovalent anions such as OH − , Cl − , or NO3 − [74][75][76].The presence of Al and high alkali concentrations can lead in addition to the formation of zeolitic phases [49,51,52].At very high alkali concentrations also the formation of an amorphous magnesium silicate phase has been reported [81].Experimentally, the formation of M(-A)-S-H and CO3hydrotalcite (LDH) has been observed in hydrated cements based on MgO, hydromagnesite and metakaolin [13].
Thermodynamic modelling predicts also formation of M-A-S-H and CO3-LDH at a high fraction of metakaolin (see Figure 8) in agreement with the experimental observations.However, modelling also predicts only a minor consumption of hydromagnesite, while experimentally a more important reaction of hydromagnesite was observed [13].
The over-prediction of hydromagnesite by thermodynamic modelling is due to the absence of thermodynamic models describing the potential uptake of carbonates by M-S-H and/or amorphous magnesium carbonates due to the lack of systematic experimental data.
The blending of MgO with industrial by-products such as fly ash, blast furnace or pyro-metallurgical slags [85] or bauxite residues will also lead to the presence of SiO2 and Al2O3, albeit released at a much slower rate than from metakaolin or calcined clays.Little research is available, although it has been shown that a high fraction of MgO is beneficial for alkali activated cements due to the formation of LDH phases [86].  1 and in [13].
The formation of HMC such as nesquehonite depends on pH value and on the availability and concentration of magnesium, carbonate, pH and temperature [87].The presence of silica, iron, or aluminium in addition to MgO will change the hydrates formed, the concentrations in solution and thus stabilize and destabilize individual HMC.These changes can be drastic as illustrated in Figure 8, where the presence of metakaolin (Al2O3•2SiO2) led to the formation of carbonate-LDH and M-A-S-H, while brucite was destabilised in agreement with experimental observations, while the reaction of hydromagnesite was underestimated due to lack of thermodynamic data to consider the potential uptake of carbonates by M-S-H and/or amorphous magnesium carbonates.

Effect of Fe(II) and Fe(III)
Iron present in magnesium silicates such as olivine will be present as impurities containing Fe(II) or Fe(III) or mixtures thereof, in the extracted MgO and SiO2-rich fraction.The presence of Fe(III) can affect the stability of the HMC and magnesium silicates formed, as i) its uptake can stabilise one of the hydrates formed via the formation of a solid solution, or ii) new phases form, which change the kind and amount of HMC and magnesium silicates formed.The presence of iron with different redox states could lead to the formation of Fe(II)-Fe(III) containing LDH phases as observed in iron rich slag cements and in corroding steel under marine environments [21,72,88].In addition, as Fe(II) is common in natural magnesium silicates [89], an uptake of Fe(II) in M-S-H phases seems probable.Also Fe(III) might be taken up by M-S-H [42], similarly as Al(III) [43,48].Fe(II) and Fe(III) uptake in M-S-H has not yet been systematically investigated, and for most Fe-LDH solubility measurements are missing.

7
Roadmap towards the establishment of durable MgO-based cements

Reaction kinetics
Promising methods to accelerate the reaction of MgO and brucite could either be based on the use of HMC precursors as nucleation sites for the formation of HMC [32,90], hydration agents such as magnesium acetate [33] or with the use of inorganic accelerators such as sodium phosphates [47], which strongly accelerates the reaction of M-S-H cements.The use of nucleation agents such as calcite or calcium silicate hydrates (C-S-H) in Portland cement based systems has recently received increasing attention as they are very efficient to accelerate the nucleation of C-S-H leading to faster hydration and strength gain for Portland cements [91,92].In HMC cements the presence of hydromagnesite, artinite, or dypingite as nucleation agent has been reported to increase the early reaction of MgO from slightly to considerably depending on the experimental conditions [32,90], but seems to have little effect on the long-term reaction degree.An alternative, potentially more promising route is the use of organic hydration agents such as magnesium acetate, which increases the carbonation degree and the strength development considerably [33,66].The principal mechanism of interaction between the organic ligand, dissolution and precipitation is unknown, but is presently studied.
As detailed previously, sodium phosphates have been successfully used to accelerate the reaction of brucite with amorphous silica to form magnesium silicate cements [14,47,93].Sodium or potassium phosphates are also used to strongly accelerate the MgO reaction in magnesium phosphate cements, where they lead to the rapid precipitation of magnesium phosphates [94].It can be hypothesized that phosphates would also have a strongly accelerating effect on HMC cements leading to faster reaction.
Also small organic molecules might play an important role.Nguyen et al. [66] observed the formation of a giorgiositelike phase instead of hydromagnesite in the presence of acetate, indicating that the use of organic and inorganic accelerators could lead to different HMC, either due to kinetic reason or because they are stabilized due to the incorporation of the accelerator (solid solution formation).
The presence of alkalis, ammonium, sulfates, carbonates, nitrates and chlorides are known to strongly affect the hydration kinetics of Portland cements [95], of calcium sulfoaluminate [96,97] and calcium aluminate cements [98][99][100] as well as to affect the calcite nucleation rate [101].Also the reaction kinetics of MgO and brucite in silicate bearing systems is affected by the presence of different salts such as e.g.NaCl, NaOH, NaNO3, Na2SO4, and by temperature [49], pointing into possible directions to modify the reaction kinetics, which have not yet systematically been addressed in literature.The presence of a small amount of inorganic salts in the system can lead to disproportionately large effects on the dissolution of brucite [87] as well as the stability of HMC [102].However, little is known about the driving forces of these phenomena, and obviously, such systematic studies are missing for HMC cements as well as any knowledge of potential effects on the long-term stability of HMC cements.
Once the fundamental mechanisms of enhanced reaction kinetics and stabilized Mg-based cements are uncovered, one could tune the binding systems to maximize the carbon utilization and selectively stabilize wanted HMCs in the binders.

Stabilisation of selected HMCs by solid solutions
The term solid solution describes the incorporation of foreign anions or cations in a mineral.Solid solutions are common in nature, are preferably formed with ions of similar charge, size and structural arrangement and can stabilise those solids [103].The best-known solid solution of magnesium carbonate is dolomite (CaMg(CO3)2), consisting of alternating arrangements of calcium and magnesium [104,105].The possibility of solid solution formation for other HMC is little investigated although a few indications exist that such solid solution could play an important role in the stabilisation of different HMC.Other cations with a comparable charge and ionic radius as Mg 2+ , e.g.Zn 2+ , Co 2+ or Cu 2+ , have been observed to be incorporated in magnesium carbonates [106,107], leading to a change in kinetics or to a stabilization of HMC due to solid solution formation.Similarly the formation of a giorgiosite-like phase instead of hydromagnesite was observed in the presence of acetate [66], which might indicate a stabilization due to a solid solution formation between carbonate and acetate.The effect different cations with comparable ionic radius to Mg 2+ such as Zn 2+ , Co 2+ , Cu 2+ or Li + on the kind of HMC formed has not yet been systematically investigated.

Durability
The long-term durability of construction materials is of outmost importance for the end-users.For Portland cement-based concrete, 150 years of experience are available, the underlying degradation mechanisms are well explored in science, and reasonable test methods for durability parameters such as sulfate attack, chloride ingress, carbonation and frost action (w/wo de-icing salts) are available.Also for potential low-CO2 alternatives such as alkali-activated binders or calcium sulfoaluminate cements, research has made significant progress in the recent years.For construction materials based on HMC and magnesium silicates, however, information on durability is extremely scarce: carbonated materials based on MgO with added clay showed a good performance against freeze-thaw [108] and sulfate attack [109], while they showed a decrease of strength and E-modulus during accelerated ageing (drying/wetting cycles).Mild steel reinforcement, as generally used in reinforced concrete structures, seems to be subjected to corrosion in carbonated MgO-systems, as the steel is not passivated due to the low pH of 10-10.5 [110].Weathering resistance seems to be similar to Portland cement-based products, and bio-based fibres are not degraded due to the low pH [111][112][113].
Most degradation processes in concrete structures are related to the moisture content.Both chemical reactions involving solid phases and the transport of ions from the environment are strongly dependent on the moisture content of the material [114], while drying and rewetting is strongly dependent on pore size distribution and pore space connectivity as well as moisture content spatial gradients.Porosity and permeability properties in M-S-H cements were found to be similar or even more favourable compared to Portland cement products [115,116].Pore sizes somewhat smaller, and chloride diffusion coefficients much lower compared to Portland cement have been reported [117].Towards ingress of calcium ions, M-S-H seems to be quite stable as well, at least at low pH, only a very slow ingress of Ca ions has been observed [38].
Based on the limited information available it can be concluded that systems with HMC plus M-S-H might be more durable than systems based on HMC only, but a huge research gap exists, and poor durability performance will strongly hamper a wide application of Mg-based construction materials.Systematic and fundamental studies and modelling approaches as available in the field of Portland cement regarding sulfate attack [118], chloride and seawater ingress [119,120] or carbonation [121] are missing for Mg-carbonate/-silicate cements.

Conclusions and outlook
The massive use of M-S-H cements or of HMC cements, where CO2 is no longer an emitter but a precursor, to produce cement needed to build infrastructure would lower the amount of Portland cement produced (and thus lower CO2 emissions).At the same time HMC cements could act as a major carbon sink in construction industry.This would allow to make a large step towards the necessary global approach to decrease the CO2 emissions related to concrete industry.The impact of the development and implementation of such very low CO2 or carbon-negative cements will go beyond a purely scientific one, due to the socio-economic relevance of cement and concrete, and is expected to lead to technological breakthroughs to the benefit of society and environment.
While the use of the HMC cements has been suggested already in 2001 [16], HMC cements started to receive increasing attention in the last 5 years only.In the papers exploring HMC cements, e.g.[32-35, 37, 85, 108-110], only selected aspects of HMC cements were investigated; many of the investigations focussed mainly on strength development.Together they point towards a potential of HMC cements, although how to increase workability, steer the kinetics, their long-term volume stability, the extent of their CO2 binding and many other aspects are still highly uncertain.
The structure of M-S-H, the poorly crystalline main phase in magnesium silicate cements has been relatively well characterized in the last decade [14, 38, 44-46, 48, 49, 67, 82, 122-124].M-S-H cement prepared with water reducing agents such as hexametaphosphate, can reach compressive strengths comparable to Portland cements [14,15,116] and M-S-H cements show a high resistance against carbonation [12,13].The combination of M-S-H with HMC cements could potentially result in good mechanical and durability properties.However, there is a lack of experimental studies.
There is an urgent need to not only address many new, up to now not explored fundamental aspects such as e.g.steering the kinetics and stabilisation of specific HMC by solid solution formation from the nano-to the macroscale, but to tackle at the same time in a combined effort the most important aspects related to the development and implementation of such carbon binding magnesium cements.Pioneering, high-risk/high gain work in the field of HMC and M-S-H cements is needed, to create a fundamental understanding of the factors influencing carbon storage, reaction kinetics, phase assemblages, porosity/permeability, mechanical properties and durability of HMC and M-S-H binders.A profound understanding is urgently needed to enable industry to upscale this technology and use it in a wide range of construction materials, potential utilisations range from precast, and ready mix concrete.Fundamental understanding is urgent to ensure its sustainability and durability and as well as its possible use in new construction technologies such as e.g.digital fabrication (see www.dfab.ch).Fundamental work on the hardening process, on the impacts of the residues in the raw materials and additives, on the volume stability, and porosity changes will directly influence the optimization of the properties of HMCs in terms of enhanced CO2 mineralisation, mechanical properties and durability.
In addition, thermodynamic data for magnesium-based solids, which are fundamental to predict phase assemblages and their potential changes induced by temperature, relative humidity and CO2 partial pressure or due to interaction with the environment, are urgently needed.Such thermodynamic data for M-S-H, HMC and LDH phases, will not be only of importance to understand Mgbased cements, but also for the efficient use of industrial by-products containing Mg, Fe, or Zn as well as for life cycle assessment and modelling carbon storage.We are convinced that a combination of systematic experiments with thermodynamic and or molecular modelling will contribute to the development of well-understood and well-designed durable HMC concretes within the next decade(s).
Magnesia silicate cements have a low carbon footprint, when the MgO sources are non-carbonate Mg-based minerals.Reactive magnesia silicate cements are an option to partially substitute Portland cements.The hydration of MgO-SiO2 cements results in the precipitation of a magnesium silicate hydrate phase (M-S-H) as primary reaction product: (a+b)•MgO + SiO2 + (a+c)•H2O aq → a•Mg(OH)2 (brucite) + (MgO)b SiO2 (H2O)c (M-S-H).

Figure 1
Figure 1 Worldwide availability of ultramafic rocks (Mg-silicate-rich rocks) illustrating a high potential of Mg-based cement as a major binding system.Reproduced from Scott et al.[17].

Figure 2
Figure 2 Scheme of MgO extraction and production of HMC cements.

a
solubility products referring to Mg 2+ , CO3 2-, OH -and H2O 0 ; b Thermodynamic data of artinite are based on calorimetric measurements resulting in a high (potential) error of ΔfG° and log K°.

Figure 4 and
Figure 4 and Figure 5 also illustrate the diversity of phases that may form at near ambient conditions, i.e. at temperatures between 5 and 40°C.

Figure 6
Figure 6Comparison of the molar volume of different magnesium phases per cm 3 /mol Mg.Data as compiled in Table1.