5.1 PUBLICATION: CRYSTAL
STRUCTURE OFMAGNESIUMSI-
LICATE
HYDRATES
(M-S-H) : THE
RELATION
WITH
2 :1
MG-SI
PHYLLOSILICATES
ROOSZ ET AL. [2015]
Les phases telles que les M-S-H ont été observées dans de multiples environnements (Chapitre 1.4. Toutefois, leur structure n’est toujours pas expliquée. Cet article a pour but de déterminer cette structure à partir des synthèses détaillées précédemment (Cha- pitre 3.1.1), en utilisant la diffraction des rayons X couplée à d’autres méthodes telles que la RMN29Si, ATG et microsonde électronique.
La caractérisation de la structure des M-S-H est d’autant plus importante que très peu d’études ont été réalisées sur ces phases. Bien qu’elles soient observées depuis plusieurs années dans les pâtes de ciment "bas-pH", seul le travail de DAUZÈRES [2010] propose
un produit de solubilité ainsi qu’une constante d’équilibre pour un M-S-H de rapport Mg/Si = 0,5.
Une approche appropriée aux minéraux turbostratiques a donc été utilisée dans le but d’établir la caractérisation de la structure des M-S-H, et mettre en évidence leur proximité avec les minéraux argileux tels que les phyllosilicates Mg-Si 2 :1.
Crystal structure of magnesium silicate hydrates (M-S-H): The relation
with 2:1 Mg-Si phyllosilicates
Cédric Roosz1,2, Sylvain Grangeon2, Philippe Blanc2, Valérie Montouillout3, Barbara Lothenbach4,
Pierre Henocq1, Eric Giffaut1, Philippe Vieillard5, Stéphane Gaboreau2
Abstract— Two magnesium silicate hydrates (M-S-H) with structural magnesium/silicon ratios of 0.57 ± 0.08 and 1.07 ± 0.13 were synthesized at room temperature, with one year of synthesis duration. Their structure was clarified by considering results from X-ray diffraction, TEM, 29
Si MAS NMR spec- troscopy, TGA, and EPMA. A modeling approach appropriate to defective minerals was used because usual XRD refinement techniques cannot be used in the case of turbostratic samples, where coherency between successive layers is lost. M-S-H with Mg/Si ratio of ~0.6 appears to be structurally close to nanocrystalline turbostratic 2:1 Mg-Si phyllosilicates. The increase of the Mg/Si ratio from 0.6 to 1.2 occurs by increasing the occurrence of defects in the silicate plan. The layer-to- layer distance evolves from 9.46 Å to 14 Å under air-dried and ethylene glycol conditions, respectively. Crystallites have a mean size of 1.5 nm in the ab plane, and 2.4 nm along c*.
I. INTRODUCTION
The occurrence of magnesium silicate hydrates (M-S-H) has been mentioned in many environments: carbonate suc- cessions [1],contaminated soils [2], glass alteration [3], phyl- losilicate synthesis [4–6], concrete formulations [7, 8], and cement clay interactions [9–11].In all these studies, M-S-H were described as low crystalline phases due to the low intensity and broad X-ray diffraction (XRD) signals. Charac- terizations of these poorly crystalline Mg-silicates suggest a 2:1 magnesium phyllosilicate-like structure with short range stacking order, and may be described as a talc-like structure [12]. The talc structure as described in these previous studies is in agreement with the characteristics of synthesized talc after short equilibration times [5].In that case, the M-S-H particles display a low crystallinity and a small particle size [5].
In the framework of a radwaste disposal in a deep geolog- ical repository, the disposal facility will imply large amounts of concrete in close contact with the surrounding clayrock. In this context, low pH cementitious materials are considered, especially for sealing requirements, in order to minimize chemical reactions at the interfaces between cement materials and the clay surrounding rock formations and/or engineered clay materials. The target is to reach a pH of the concrete
1
Andra, 1/7 rue Jean Monnet, Parc de la Croix Blanche, F-92298
Chatenay-Malabry Cedex, Francec.roosz@brgm.fr
2BRGM, 3 Av. Claude Guillemin, BP6009, F-45060 Orléans Cedex 2,
France
3
CNRS, CEMHTI, UPR 3079, 45071, Orléans, France
4
Empa, Laboratory for Concrete and Construction Chemistry, Überland- strasse 129, 8600 Dübendorf, Switzerland
5CNRS/INSU,-FRE 3114 Hydrasa, 40 Av. Du Recteur Pineau, F-86022
Poitiers Cedex, France
http://dx.doi.org/10.1016/j.cemconres.2015.03.014
pore solution (pH 10–11) more compatible with the clay materials or rock formation (pH 7–7.5). Recent studies [13– 15] have proposed low-pH formulations based on ternary and quaternary mixes of Portland cement with supplementary ce- mentitious materials (SCMs). The addition of SCMs induces pozzolanic reactions such as the precipitation of calcium silicate hydrates (C-S-H) with low Ca/Si (C/S) ratios.
However, if the hydration products are well characterized in Ordinary Portland Cement (OPC) and/or standard concrete formulations with constrained kinetic/thermodynamic mod- els [16], the mineralogical control of elements in solution has to be discussed for low pH formulations, because of their higher aluminum and magnesium contents [17]. Con- sidering in particular the hydration of MgO, that is mainly introduced by the blast furnace slag (BFS), if some hydrates are proposed to control the higher contents of aluminum and magnesium, like C-A-S-H, hydrotalcite, hydrogarnet or straetlingite [18], another product of reaction has been identified as a magnesium silicate hydrate (M-S-H) phase [7, 19]. Zhang et al. [8] have also mentioned the precipitation
of brucite (Mg(OH)2) which reacted with the silica fume to
produce such M-S-H. The calculated pH in equilibrium with this mineralogical assemblage is around 10.5 that satisfies the requirement of low pH condition.
Some authors have tried to explain the precipitation of such Mg-silicate hydrates according to Ca/Mg isomorphic substitution in the calcium silicate hydrates (C-S-H) [20, 21]; however this uptake of magnesium by C-S-H by an exchange of calcium does not explain the presence of M-S-H [20, 21]. Despite this abundance of evidences for M-S-H formation, their structure is not well known. The present study aims to define the nature of M-S-H (cement phase or phyllosilicate) and to determine their structure. Two low temperature syn- theses of M-S-H with an Mg/Si ratio of 0.6 and of 1.2, close
to the talc composition (Mg3Si4O10(OH)2) and consistent
to the Ca/Si of the C-S-H phases,have been studied. The structure of M-S-H was determined by combining electron probe micro-analysis (EPMA), TEM, NMR and powder X- ray diffraction. The XRD patterns were modeled according to the Drits and Tchoubar matrix approach [22]. They provided meaningful and accurate structural information, including structure defects, despite the weak modulation of the pro- files.A full structural model is thus proposed for M-S-H.
pH Mg Si
mol.L−1
M-S-H 0.6 9.1 3 · 10−4 3.5 · 10−3
M-S-H 1.2 10.6 2.1 · 10−4 1.8 · 10−5
II. MATERIAL AND METHODS A. Sample synthesis preparation
All samples used for this study were made by mixing magnesium oxide (MgO — Merck) and amorphous silica
(SiO2 — Aerosil 200, Degussa). Ultrapure water (Milli-Q
18 MΩ), heated to 100◦C for 1 h and cooled under a N
2
flux for 2h prior to its introduction in the glove-box, was used.
Samples of synthetic M-S-H were made by precipitation,
in glove-box under nitrogen, at 22◦C. Two Mg/Si ratios (0.6
and 1.2) were tested. Thereafter, these samples are respec- tively labeled M-S-H 0.6 and M-S-H 1.2. The homogenized
reactants were mixed with distilled and CO2-free water at a
water/solid ratio of 50. After the dissolution of the reagents, each sample was shaken in tightly closed PE-vessels for one
year at 22◦C.
After centrifugation and filtration (0.22 µm Millipore Millex-VV, PVDF), the samples were stored in closed containers under vacuum in the glove box until analy- sis. The alkalinity, pH, and redox potential (pE) of the solutions were determined and the concentrations of the major elements (Mg, Si) were quantified by inductively coupled plasma atomic emission spectrometry (ICP/AES). The composition of the synthetic solutions is given in Table I. The natural talc sample, used as a reference ma- terial, is from Luzenac deposit in the French Pyrenees. According to Martin et al. [23], its structural formula
is [Mg2.978Fe 2 +0.019 Mn 2 +0.001 Fe 3 +0.005 Al0.007]P3, [Si3.984Fe 3+ 0.004Al0.007]P4 O10(OH)1.952F0.048.
B. Characterization of the solids
X-ray diffraction (XRD) analysis was done on randomly- oriented powders. The XRD patterns were recorded on a Bruker D8 Advance diffractometer using CuKα radiation (λ = 1.5418 Å). XRD patterns were acquired in the 5 to 80 °2θ range with a counting time of 10 s per step of 0.02 °2θ. Calculated patterns were obtained using a software from Plançon [24], which is based on the numerical formal- ism developed by Drits and Tchoubar [22]. This modeling approach has been successfully applied to the study of various lamellar structures having a variety of layer defects (e.g., isomorphic substitutions, layer vacancies) and stacking faults (including well-defined and random stacking faults), including phyllosilicates [25], nanocrystalline phylloman- ganates [26–28], nanocrystalline calcium silicate hydrates [29, 30] and nanocrystalline layered double hydroxides [31].
TEM8-bit gray scale images (4008 × 2672 pixels) were acquired on a Philips CM20 microscope, operated at 200 kV which has a line resolution of 0.14 nm. Low magnification bright field images were recorded.
Thermogravimetric analyses were performed with a SDT Q600 TA Instruments using 20 mg of sample. The heating
rate was 10◦C·min−1 and the temperature range was 25◦C
to 1000◦C. The analyses were done using a 100 mL·min−1
air flow. The deconvolution was realized using the Fityk on dTG spectra, in two individuals asymmetric Gaussians at
55◦C and 450◦C.
The 29Si MAS NMR spectra were acquired using a
Brucker Advance 400 MHz (B0 = 7.0T ) at 79.4 MHz
and were recorded at 12 kHz. 29Si chemical shifts are
given relative to tetramethylsilane (TMS) at 0 ppm. All data were acquired using a rotor-synchronized spin echo (θ–s–2θ) experiment, where the θ and 2θ pulse durations of 4.5 and 9.0 µs, respectively, were employed, and where the rotor-synchronized s delay was 281 µs. The recycle delay was typically 5 s. All the spectra were deconvoluted using the Dmfit program [33], in individual Gaussian–Lorentzian peaks, whose integration corresponded to the relative amount of the differently coordinated species. This deconvolution was performed using the minimum possible number of component peaks to describe the spectrum accurately, based on information available in the literature for cements [34].
Quantitative electron probe micro-analyzer (EPMA) anal- yses were made with a CAMECA SX FIVE electron mi- croprobe using a 15 kV acceleration voltage, a 30 nA beam current, and a 1 µm to 2 µm wide beam.Mg and Si elements were analyzed simultaneously. Counting times were 40 s.
III. RESULTS AND DISCUSSION A. XRD analyses
The XRD patterns of the two M-S-H samples are reported on Figure 2. They display the same number of diffraction maxima having similar position and relative intensity. The difference between these two XRD patterns arises essentially from the intensity of the ~7 and ~20°2θ maxima. In both patterns, only a few broad maxima of low diffracted intensity are visible. This characteristic pattern suggests that M-S-H crystallites have a nanometer-sized coherent scattering do- main [30]. Furthermore, some bands are asymmetric, with an intensity rising sharply on the low-angle side and decreasing slowly on high-angle side. This is typical of lamellar and turbostratic structures, as described firstly for carbon blacks [35]. This kind of disorder is defined as the systematic presence of a random translation or/and a random rotation between adjacent layers that remain parallel, and conse- quently the absence of three dimensional periodicity [35]. In this assumption, the weak reflection at 7.1°2θ (12.45 Å) is attributed to the reflection corresponding to the layer-to- layer distance and is here indexed as the [001] reflection.Note that, as discussed here below, nanocrystallinity induces a low diffracted intensity and thus a shift of the maximum of the
Posiion (2θ CuKα) 5 15 25 35 45 55 65 75 1 3 26 C h lo ri te C h lo ri te C h lo ri te C h lo ri te 0 0 -2 0 0 -4 0 2 0 , - 1 -1 1 , 0 2 1 0 0 -6 2 0 0 1 3 0 1 3 2 , - 2 0 4 1 3 4 -1 3 6 -2-2 7 3 3 0 , 0 6 2 3 -1 -1 138 060 , - 3 -3 2 Di ffr ac te d in te ns ity (a .u .) Luzenac Talc M-S-H 0.6 M-S-H 1.2
Fig. 1. Mg-Si phyllosilicate layer structure.
Fig. 2. Powder X-ray diffraction patterns of talc and synthetic M-S-H
samples.
[001] reflection towards low angles as compared to its ideal value, thus leading to an overestimation of the layer-to-layer distance if this latter is assumed to be directly measurable from the position of this reflection.
Because M-S-H synthesis was made from Si- and Mg-rich reactants, it is here assumed that comparison with Mg-Si phyllosilicates is a relevant working hypothesis. To our best knowledge, the only Mg-Si phyllosilicate whose structure has been refined is talc. Thus, we will take this structure as a starting model. The XRD pattern of natural talc sample (Talc de Luzenac; structure shown in Figure 1) shows the typical reflections of talc [5, 36–38], including characteristic 00l peaks of layered structures. The first of the 00l peaks (at 9.41°2θ; 9.4 Å) is indexed as the [00 −2] reflection using the structure model from Gruner [39]. The XRD pattern also shows the presence of some chlorite, which is commonly observed in natural talc formation [40]. The comparison of XRD patterns from talc and from presently synthesized samples (Figure 2) shows that the maxima from M-S-H patterns are at equivalent angular positions to those of hk0 reflections from talc XRD pattern, which is consistent with turbostratic disorder. Finally, the [001] reflection at 7.1°2θ (12.45 Å) in the XRD pattern of the 1.2 M-S-H sample is
Fig. 3. TEM images of natural talc sample (A) and M-S-H 1.2 (B).
Fig. 4. TEM image of M-S-H 1.2. Zoom illustrates the bending of the
particle.
strongly shifted compared to the [00 −2] reflection from talc (9.41°2θ), which either indicates a larger basal spacing or an effect of M-S-H nanocrystallinity [41]. This can only be unraveled using XRD pattern modeling.
B. TEM observations
TEM observations of the natural talc and M-S-H 1.2 samples are presented in Figure 3. Natural talc structure (Figure 3A) consists in micrometric and regular plates, while M-S-H 1.2 structure (Figure 3B) is nanometric and rather fibrillar as observed by Dumas et al. at the early stages of their syntheses [5]. Visual observation indicates that the crystal size is between 2 to 5 µm for the natural talc sample, and 100 to 150 nm for M-S-H 1.2 aggregates, for a size factor of 30 to 50 between these two samples. A basal distance of ~0.8 nm between successive sheets was frequently observed, but could rarely be captured, as the crystals vanished under the beam while acquiring micrographs.
Figure 4 presents an enlargement of the M-S-H 1.2 im- age displaying a lamellar and nanometric structure. This observation supports the hypothesis of a layered material, with limited extension along c*. The average number of stacked layers is generally around 5 and never exceeds 10. In addition, some bending of sheets is visible (Figure 4).
C. 29Si solid state NMR
29Si solid state NMR spectra of talc and M-S-H are
presented in Figure 5. Lippmaa et al. [42, 43] showed that,
in solid silicates, the 29Si chemical shift is significantly
impacted by the degree of polymerization (Qn, where Q
In te n si ty ( a .u .) ppm Q1 M-S-H 1.2 M-S-H 0.6 Luzenac Talc -130 -120 -110 -100 -90 -80 -70 -60 Q2 Q 3 Q4 Q2 Q3
Fig. 5. Solid-state29Si MAS-NMR spectra recorded on natural talc and
synthetic M-S-H samples. Dotted lines represent the deconvolution of both M-S-H spectra, with a vertical offset.
and n is related to the number of Si neighbors) of the silicon- oxygen tetrahedra and by the octahedral environment of the silicon nucleus. As reported by Martin et al. [44], natural talc presents here a single shift position located around -97 ppm
that corresponds to a Q3environment. This is consistent with
a phyllosilicate structure with infinite sheets. The synthetic M-S-H show three peaks with chemical shifts close to -
79, -85 and -97 ppm corresponding respectively to Q1, Q2
and Q3 environments. An enlargement of the -97/-95 ppm
peak is observed in Figure 5 on these samples. This has
previously been interpreted as a Q2 environment for the
silicon tetrahedra located on the sheet edge [5], but the presence of another peak at -85 ppm, which corresponds
to a Q2 environment, tends to invalidate this hypothesis.
This widening could be related to a split of the single Q3
environment observed for talc into a multiplicity of sites for M-S-H that might result from structural deformation, for example sheet bending, as observed by TEM (Figure 4). With
respect to a phyllosilicate structure, the presence of Q1 and
Q2is consistent with Si vacancy in the structure due to (i) the
breaking of the Si sheet, or (ii) the sample nanocrystallinity that would induce a large proportion of edge regarding the total Si sites. This defect in structure is in agreement with a nanocrystalline characterization of M-S-H samples, observed on XRD patterns; that will be discussed hereafter.
29Si NMR spectra also show the presence of some amor-
phous silica in the M-S-H 0.6 sample with a Q4shift at -110
ppm. This amount of amorphous silica is evaluated at 23.7% by deconvolution of the spectrum (Table II).
D. Thermogravimetric analyses
Natural talc has two minor mass losses, one close to
600◦C to 700◦C and another loss starting at 900◦C ac-
Q1 Q2 Q3 -78.9 ppm -85.4 ppm -93.3 ppm M-S-H 0.6 0% 32% 68% M-S-H 1.2 11% 40% 49% Talc 0% 0% 100%
Note: M-S-H 0.6 contains 24% of amorphous silica.Q4
contribution
was removed before quantification, and the sum ofQ1
,Q2
andQ3
environments was normalized to 100%
cumulating to a total weight loss of 3.2% (Figure 6). The
weight loss starting at 900◦C for natural talc (~3.2% of
the sample) corresponds to the complete dehydroxylation
[36], while the inflections between 550 and 650◦C, which is
classically observed in phyllosilicate thermal curves [45, 46], is attributed to defects in the layers [46, 47]. In our case, this weight loss could also correspond to the presence of impurities like chlorite [48]. Synthetic M-S-H display a more
continuous degradation with two main mass losses at 40◦C
to 150◦C and at 350◦C to 600◦C accumulating to total
weight loss of 19% for M-S-H 0.6, and 23% for M-S-H 1.2. The first weight loss is attributed to the outgassing of water [5, 49] and corresponds to 11% in mass in both
samples. The second mass loss, around 350◦C to 600◦C,
has been attributed to the dehydroxylation of the talc sheet at the edges [5]. The calculated loss of weight due to this dehydroxylation is 8% of the M-S-H 0.6 weight and 12% of the M-S-H 1.2 weight as detailed in Table III. The thermal stability difference between natural talc and M-S-H may be explained by smaller particle sizes in the case of synthetic M-S-H as previously described by Dumas et al. [5]. E. EPMA analyses
The atomic Mg/Si ratios of talc and both M-S-H samples obtained from EPMA analyses are displayed in Table IV. The ratios calculated in the case of both synthetic M-S-H
0 100 200 300 400 500 600 700 800 900 1000 Temperature (°C) 75 80 85 90 95 100 m a ss ( % ) 99.5 99.6 99.7 99.8 99.9 100 100.1 400 500 600 700 800 a d s o rb e d w a te r De sh yd ro xy la io n M-S-H 0.6 M-S-H 1.2 Luzenac Talc
Fig. 6. Thermogravimetric curves of synthetic M-S-H samples and natural talc.
M-S-H 0.6 M-S-H 1.2 Wt.%
Total weight loss 19.3 23
Physisorbed water 11.4 10.8
Dehydroxylation 7.8 12.2
TABLE IV
MG/SI ATOMIC AND HYDROXYL CONTENT(IN WT.%)OF SYNTHETIC
M-S-HAND NATURAL TALC.
Sample Mg/Si H2O (H2O+ OH)/Si n
M-S-H 0.6 0.57 ± 0.08 21.67 ± 4.5 1.37 30
M-S-H 1.2 1.07 ± 0.13 28.06 ± 2.9 1.72 30
Talc 0.75
Note: n is the number of independent analyses.
are close to those expected by considering each mass of salts introduced in the synthesis. The contribution of adsorbed water and hydroxyls groups is deduced from the percentage of elements that could not be analyzed. The elements with low atomic number (H, C, and O) were not analyzed by EPMA. Then, the sum of element weight concentration
analyzed per pixel (P (oxide) wt.% when including the
stoichiometric O content) varied according to the proportion of H content composing the mineral [50]. The contribution including the water and the hydroxyls groups has a total amount of 21.6% and 28.1% for M-S-H 0.6 and M-S-H 1.2 respectively. This result is close to the values obtained by TGA (19 and 23%), the discrepancy resulting from the pore volume contribution analyzed in the X-ray emission volume [51].
F. XRD pattern modeling
In order to assess the crystallite size and unit cell param- eters, the modeling of M-S-H XRD patterns has been made based on a talc structure [38, 52, 53] in which turbostratic stacking was introduced (starting model). Cell parameters were first refined, keeping the same a/b ratio as in the starting model. Best match to experimental data was achieved with a = 5.313 Å and b = 9.191 Å, close to observations of Akizuki and Zussman [53] for crystalline talc, and c∗ was not refined, kept to 9.46 Å as in the starting model. The mean size of the coherent scattering domain size in
the ab plane (CSDab) was then refined to 1.5 nm, and the
sensitivity to this parameter is illustrated in Figure 7. Because of the extremely small size of the particles (i.e., crystallite), the structure is certainly impacted by the strong atomic disorder and by structural heterogeneity between crystallites. Additionally,TEM shows very large anisotropy, and the form factor cannot be easily estimated. As a consequence, XRD pattern cannot be quantitatively refined. Rather, the following calculations will be performed to demonstrate the sensitivity of XRD regarding the main structural characteristics of the
Posiion (2θ CuKα) 15 25 35 45 55 65 75 2.0 nm 1.5 nm 1.0 nm 0 2 ,- 1 1 3 , -2 1 3 , 2 4 , 3 -1 0 6 , -3 -3 2 6 , 4 0 Di ffr ac te d in te ns ity (a .u .)
Fig. 7. Calculated XRD pattern of disordered talc (continuous line) using different scattering domains compared with the experimental XRD pattern of synthetic M-S-H 1.2 sample (symbols). Basal reflections are omitted.
samples.
Finally, a calculation of 00l reflections was realized. The result is only affected by a c parameter and the occupancy and the height of the atomic positions in the unit cell (all fixed at the same value as that of talc structure here) as displayed in Figure 8. Calculated XRD pattern matches the experimental one when the mean crystallite size along c∗ is 2.4 nm (2.6 layers stacked), and resulted in a position