Accepted Manuscript Quantitative mineralogical mapping of hydrated low pH concrete Stéphane Gaboreau, Dimitri Prêt, Valérie Montouillout, Pierre Henocq, Jean-Charles Robinet, Christophe Tournassat PII:
S0958-9465(16)30747-8
DOI:
10.1016/j.cemconcomp.2017.08.003
Reference:
CECO 2881
To appear in:
Cement and Concrete Composites
Received Date: 18 November 2016 Revised Date:
7 July 2017
Accepted Date: 7 August 2017
Please cite this article as: Sté. Gaboreau, D. Prêt, Valé. Montouillout, P. Henocq, J.-C. Robinet, C. Tournassat, Quantitative mineralogical mapping of hydrated low pH concrete, Cement and Concrete Composites (2017), doi: 10.1016/j.cemconcomp.2017.08.003. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Quantitative Mineralogical mapping of hydrated low pH concrete
1 2 3
RI PT
6 7 8 9 10 11 12 13
Stéphane Gaboreaua,*, Dimitri Prêtb, Valérie Montouilloutc, Pierre Henocqd, JeanCharles Robinetd, Christophe Tournassata,e a
BRGM, Environment and Process Division, 3, avenue Claude Guillemin, F-45060 Orléans Cedex 2, France b UMR CNRS 7285 IC2MP, Université de Poitiers, Equipe HydrASA, rue Albert Turpain, Bat B8, 86022 Poitiers, France c CNRS-CEMHTI UPR 3079, 1 Avenue de la Recherche Scientifique, 45071 Orléans, cedex 2 France d Andra, 1/7 rue Jean Monnet, Parc de la Croix Blanche, 92298 Châtenay-Malabry Cedex, France * Corresponding author:
[email protected] e Earth and Environmental Sciences Division, Lawrence Berkeley National Laboratory, Berkeley, USA
SC
4 5
M AN U
14
Abstract
16
Concrete materials are made of various minerals and phases, whose spatial
17
heterogeneous distributions impact the overall physical and chemical properties of the
18
materials. We have investigated the heterogeneous distribution of minerals and phases
19
in two types of concrete using quantitative X-ray intensity maps coupled with an
20
innovative data treatment method based on image segmentation. This method provided
21
quantitative data on spatial distribution, modal content and associated calculated
22
formulas for each identified mineral and phase in the binder with micrometer resolution.
23
We also obtained quantitative information on the porosity associated with the phases,
24
making it possible to differentiate poorly hydrated cement phases (initial clinker
25
hydration reaction) from highly hydrated phases (final cement product) despite their
26
similar chemical composition, when expressed in terms of cationic formulas. We
27
quantified the mineralogical and phase contents, independent of crystal size or
AC C
EP
TE D
15
1
ACCEPTED MANUSCRIPT crystallinity considerations. We report spatial resolution in the pozzolan hydration
29
process over different observation scales for the two investigated concretes.
30
Keywords: concrete, quantitative mapping, mineralogy, phases, porosity
AC C
EP
TE D
M AN U
SC
RI PT
28
2
ACCEPTED MANUSCRIPT
1. Introduction
32
Concretes are complex finely divided materials made of components of various natures
33
(cement, aggregates, water, and more) and with variable sizes. They are dynamic
34
materials that pass from a liquid to a solid state when prepared. While in the solid state,
35
hydration processes with slow kinetics lead to changes in concrete microstructure and
36
phase composition. Hydration results in the formation of various calcium silicate and
37
calcium aluminate hydrates, among other hydrates. Amounts, composition and
38
distribution are sensitive to the water/cement (w/c) ratio, the relative humidity and the
39
reaction kinetics; the chemistry stays the same with a redistribution of elements through
40
transition and transformation reactions like dissolution/precipitation. Many different
41
formulations may be used depending on the industrial application. The cement
42
component of concretes contains mainly portland cement clinkers, but can also contain
43
supplementary cementitious materials (SCM) such as fly ash (FA), metakaolin (MK),
44
blast furnace slag (BFS) and silica fume (SF) in relative amounts that depend on the
45
cement formulation. Using these formulations mitigates detrimental physical and
46
chemical effects caused by the environment surrounding the concrete such as sulfate
47
attack [1, 2], chloride attack [3], alkali-silica reaction [4] or carbonation [5]. Obviously,
48
hydration of cementitious materials involves complex chemical reactions. Their service-
49
life depends on the initial formulation and interactions with the environment over time.
50
Concrete mineralogy and the associated pore solution composition evolve according to
51
the initial cement composition, the kind and amount of SCM (SF, MK, FA, BFS) and
52
their respective reactivity. To assess the locally uneven distribution and reactivity of the
53
materials and the associated complex hydrate distribution, imaging techniques are
54
necessary. To tackle this type of problem, the method proposed (e.g. chemical X-ray
55
maps obtained with electron probe microanalysis) allow computing mineral/phase maps
AC C
EP
TE D
M AN U
SC
RI PT
31
3
ACCEPTED MANUSCRIPT based on procedures of chemical segmentation based on ternary scatterplot projections.
57
This method, initially developed by [6, 7] for clay materials, was improved and adapted
58
to the case of concrete materials in order to map the mineralogy with µm resolution on
59
millimeter-scale areas, in which the hydration processes were resolved by quantifying
60
the anhydrous residual components (i.e., clinker phases and SCMs) and analyzing the
61
spatial chemical evolutions resulting from hydration. To illustrate the numerous
62
advantages of this mineralogical mapping method, we studied two concrete
63
formulations with various substitutions with SF, BFS and MK and curing conditions
64
(hydric conditions). The chemistry of the hydraulic binders was characterized and
65
quantified, considering the initial formulation.
66
2. Materials & Methods
67
2.1. Concrete formulations
68
We studied two different low-pH concretes (Concretes I and II), with the Portland
69
cement partially substituted by SCMs, as silica fume, blast furnace slag and metakaolin.
70
Both samples were provided by the French National Radioactive Waste Management
71
Agency (ANDRA) as part of the Cigéo Waste Disposal Centre project. The detailed
72
compositions of both concretes are given in table 1.
SC
M AN U
TE D
EP
Table 1 concrete compositions
AC C
73
RI PT
56
Components Particle size Cement CEM I 52.5N <100 µm Cement additive SF <10 µm BFS <100 µm MK <20 µm Aggregates Carbonates 4-12 mm sand 0-4 mm
Concrete I
wt%
wt% (normalized to cement)
kg m-3
Concrete II
wt%
wt% (normalized to cement)
kg m-3
79
3.8
22.0
450
21.4
86.1
128 188
4.7 8.6
27.6 50.4
50
1.6
6.5
50
1.8
7.4
736 900
75.1
1035 813
82.9
4
ACCEPTED MANUSCRIPT Water Organic additives (Plasticizer) 74
181 11
161 14
Low-pH concrete I was a ternary mixture of CEM I 52.5 N CE PM ES CP2 NF
76
(Lafarge, Val d’Azergues), Slag (Orcem) and SF (CONDENSIL® S95 DM) with 80%
77
of substitution of the clinker by the SCMs (Table 1). Concrete I was drilled in the
78
LSMHM Underground Research Laboratory (URL, Meuse/Haute Marne, France),
79
where it had been in contact with the Callovian-Oxfordian clay-rock formation in a bore
80
hole located three meters below the floor of an access gallery. The experiment was
81
dismantled one year after its implementation. No ingress of pore water from the
82
surrounding clay-rich rock occurred during this period and the concrete was exposed to
83
atmospheric conditions.
84
Concrete II was prepared with the same CEM I and SF (CONDENSIL S95 DM) as
85
concrete I, with added MK (ARGICAL-M 1000) with 20 % of substitution of the
86
clinker by the SCMs (Table 1). Concrete II was cured for one year in a desiccator at
87
equilibrium with a CO2-free atmosphere having relative humidity close to 100 %, to
88
enable the hydration of the cement phases, while avoiding carbonation. The concrete’s
89
coarse and fine aggregates were calcareous aggregates and quartz (Boulonnais, France).
90
Considering the substitution of the clinker by the SCMs in both concretes (80 and 20 %
91
in concrete I and II, respectively), the Ca/Si ratio of the hydraulic binder should be
92
lower in the case of concrete I.
93
2.2. Bulk chemical and mineralogical characterization
94
Si, Al, Ti, Fe (total), Mn, Ca, Mg, K and Na were chemically analyzed using a PW2400
95
sequential X-ray fluorescence (XRF) spectrometer (Philips). The amount of total carbon
96
(TC) and sulfur were determined by infra-red spectroscopy after burning the samples at
AC C
EP
TE D
M AN U
SC
RI PT
75
5
ACCEPTED MANUSCRIPT 97
900°C in an oxygen atmosphere. Carbonate contents were also measured by dissolution
98
in an HCl solution and titrating the CO2 produced using a volumetric method. The bulk
99
chemical analyses of concrete I and II together with the composition of CEMI and the SCMs are given in table 2.
101 102
Table 2 Global chemical composition of the concretes for major elements (in weight percentage)
RI PT
100
M AN U
SC
Na2O K2O CaO MgO SiO2 Al2O3 Fe2O3 Total Organic MnO TiO2 S LOI C C total Concrete I 0.12 0.08 48.7 1.5 9.8 1.4 0.37 9.97 1.05 0.04 0.09 0.11 36.5 Concrete II 0.09 0.10 45.2 0.8 13.8 1.7 0.87 8.67 1.1 0.02 0.11 0.14 36.5 CEMI 52N 0.1 0.6 65.1 0.6 20.9 3.4 4.4 2.7 1.3 SF <0.2 0.3 0.5 <0.2 96.3 <0.2 0.1 0.02 <0.05 2.2 BFS <0.2 0.2 42.7 7.2 36.0 11.6 0.6 0.2 0.5 <0.1 MK 0.8* 0.3** 55.0 40.0 1.4 1.5 0.3 1.0 103 LOI: loss on ignition at 1000°C; * Na2O+K2O; **CaO+MgO Total bulk porosity were measured by kerosene porosity method [8] and calculated from
105
measured grain density (ρgr) (helium pycnometry) and apparent dry density (ρ)
106
(mercury intrusion porosimetry) according to the following equation [8] :
TE D
104
= 1−
The measured and calculated porosity are 16 ± 1 % and 13 ± 1 % and the grain density
108
at 2.61 g cm-3 and 2.64 g cm-3, for concretes I and II respectively. The total porosity
109
includes the capillary and gel porosity. The local silicon and aluminum environment in
110
both concretes was probed by solid-state NMR.
111
NMR spectra were acquired at 59 MHz on a Bruker AVANCE 7.4 T (300 MHz)
112
spectrometer equipped with a 4 mm double-bearing MAS probe-head spinning at 12
113
kHz. About 20,000 scans were accumulated after a 45° pulse, using 10 s recycling
114
delay. This delay was optimized to ensure complete magnetization relaxation.
29
115
chemical shifts were reported relative to tetramethylsilane (TMS) resonance. The
27
AC C
EP
107
29
Si Magic Angle Spinning (MAS)
Si
Al
6
ACCEPTED MANUSCRIPT NMR high-resolution NMR experiments were carried out on a Bruker AVANCE
117
instrument (magnetic field 17.6 T–750 MHz) equipped with high speed MAS
118
probeheads (spinning rates of 30 kHz in aluminum-free zirconia rotors, diameter
119
2.5 mm). The 1D MAS spectra were acquired after a single short pulse (π/10) ensuring
120
quantitative excitation and identification of the 27Al central transition [9]. 29Al chemical
121
shifts were reported relative to Al(NO3)3 1M resonance. All the spectra were
122
deconvoluted using the Dmfit program [10] into individual Gaussian-Lorentzian peaks,
123
whose integration corresponded to the relative amount of the differently coordinated
124
species.
125
2.3. Electron Probe MicroAnalysis (EPMA) and quantitative chemical maps
126
Samples were fully impregnated with methylmethacryle (MMA) using an impregnation
127
procedure already described in the literature [6-8, 11-13]. Before impregnation, a
128
parallelepiped of 8x4x3 (length, width, heigh) was cutted with a diamond wire saw.
129
This impregnation made it possible to prevent physical perturbation while preparing
130
polished section. The samples were the polished with different diamond suspensions (3,
131
1, ¼ µm) for 1 hour.
AC C
EP
TE D
M AN U
SC
RI PT
116
132 133 134 135 136 137
Figure 1 – Mosaic of BSE images of concrete I (A). The white square represents the size of the analyzed area, illustrated with the BSE image (B) and the associated quantitative chemical map of Ca Kα for a dwell time of 100 ms (C). The brightest parts of the maps represent the highest Ca wt%. On the BSE image, the anhydrous phases appear in light gray, the hydraulic binder in dark gray and the macropores in black.
7
ACCEPTED MANUSCRIPT Quantitative X-ray intensity maps and a BSE image (Figure 1) were acquired with a
139
Cameca SX Five EPMA equipped with five wavelength dispersive spectrometer (WDS)
140
and operating at 15 keV and 30 nA. To investigate the distribution of the ten elements
141
identified in the concrete (Table 2), the area was scanned twice, because the number of
142
the simultaneously detected elements was constrained by the number of available
143
spectrometers. We collected Kα peak intensities (for Si, Al, Fe, K, Na, Ca, Mg, Ti, Mn
144
and S) using large Large Thallium Acid Phtalate (LTAP) and Large Pentaerythritol
145
(LPET) and Pentaerythritol (PET), Thallium Acid Phtalate (TAP) and Lithium Fluoride
146
(LIF) monochromator crystals, allowing for a high counting rate and short dwell time
147
(~4 s) for a quantitative point analysis. This dwell time is too time-consuming for a
148
mapping mode, so we used a shorter counting time of 100 ms per pixel, as discussed in
149
[6]. After the acquisition, no evidence of beam damage was identified on the mapped
150
area. To reduce the total acquisition time (two days), we did not measure background
151
with subtraction from X-ray emission peaks as its contribution at short dwell times and
152
for major concentrations is low [14], and as recording it would have doubled the
153
acquisition time. A standard-based PHIRHOZ matrix correction [15] was then applied
154
to provide a weight percentage for each element per pixel. The 512 by 512 pixel
155
elemental maps were recorded by stage rastering using a stationary beam, with spatial
156
resolution of 2 µm per pixel.
157
2.4. Mineralogical mapping from quantitative chemical maps
158
Mineralogical maps were created for both concretes to display the spatial distribution of
159
the minerals over a surface area of 1 x 1 mm² with resolution of 2 µm (Figure 1)
160
following the methodology developed by [6] and using the µMAPphase software [6].
161
This methodology consists in identifying the mineral phases composing the analyzed
162
area from step by step projections of the scanned elemental composition points, and
AC C
EP
TE D
M AN U
SC
RI PT
138
8
ACCEPTED MANUSCRIPT converting them into ternary plots from which we can visualize contrasting chemical
164
compositions for all the mineral phases.
SC
RI PT
163
166
EP
TE D
M AN U
165
Figure 2 – BSE images and Si – Ca – Al3FeMg ternary projections of all the pixels of the mapped areas of concrete I and II. The main aggregates, pozzolan and hydrate end members of a hardened cement material are added on the concrete I and II scatterplots.
170
All the pre-processing steps that convert the initial quantitative X-ray maps given in
171
element weight percent (Ati wt%) into element molar percent (Ati mol%) have been
172
described in [6]. This conversion led to a chemical oxide composition for each pixel of
173
the map. All the pixels of the mapped area were plotted in ternary diagrams, where each
174
axis of the plot represented the concentration of an element or a combination of
175
elements. This method was not affected by porosity variations as normalized axis
AC C
167 168 169
9
ACCEPTED MANUSCRIPT weights were used to generate the ternary scatter plots. We identified several clusters of
177
pixels in the chemical ternary plots. These clusters represented the chemical
178
compositional fields of one mineral. Their stoichiometry was compared directly to the
179
different mineral end-members by adding their theoretical compositions to the
180
projection (Figure 2). For mixtures or solid solutions, clusters were stretched along lines
181
between the end-members. Pixels with similar chemical compositions (i.e. a cluster)
182
were selected with a polygon tool and back-projected on a mineral map using the same
183
color as the selected polygon. This method of segmentation based on ternary diagrams
184
was particularly efficient for locating small crystals and mineral mixtures [6]. For our
185
study, another advantage of this method was the possibility of comparing the chemistry
186
of the two hardened concretes based on the position of the clusters in the projections
187
(Figure 2). For most materials, the projection using only three elements was not
188
sufficient to distinguish all the mineral phases, so we used a succession of ternary
189
diagrams where the axes covered a large range of element combinations. The theoretical
190
compositions of the main clinker phases, pozzolans, hydrates and aggregates were
191
superimposed on each of these diagrams in order to facilitate reading and interpretation
192
of these chemical scatterplots (Figure 2).
193
3. Results
194
3.1. NMR results
195
The 29Si solid-state NMR spectrum of concrete I was complex as seen in Figure 3 (A),
196
but the majority extends over the typical C-S-H chemical shift region ranging from -70
197
ppm to -100 ppm [16]. The main signal was deconvoluted into at least four components
198
corresponding to the Q1, Q2b, Q2 and Q3 species [17]. Q1 refers to an end of chain silica
199
or dimers connected to only one neighbor; Q2b and Q2 refer a bridging tetrahedron and
200
to pairing silica respectively, connected to two neighbors; and Q3 refers to a crosslinked
AC C
EP
TE D
M AN U
SC
RI PT
176
10
ACCEPTED MANUSCRIPT 201
silica connected to three other tetrahedra. The signal between -104 and -120 ppm was
202
probably related to the presence of quartz and amorphous silica. In the range -70/-100
203
ppm,
204
population characterizing long silicon tetrahedron chains in the C-S-H structure. The
205
presence of long tetrahedron chains is related to low C/S ratio C-S-H. According to the
206
equation published by [18], the C/S ratio calculated from the Q1/Q2tot content is at ~ 0.6,
207
where Q2tot is the total population of Q2 and Q2b. We assigned a weak signal at -71.4
208
ppm to clinker phases (alite – belite) [17, 19]. The
209
concrete I (Figure 3A) had two areas: a broad peak between 40 and 80 ppm
210
corresponding to tetracoordinated aluminum associated with C-S-H, and a much
211
narrower peak centered at 6 ppm, corresponding to hexacoordinated aluminum. The
212
aluminum in tetrahedral sites was assigned to Al substituting for Si in the C-S-H
213
structure [20], whereas the octahedrally coordinated Al was assigned to a low
214
crystalline, disordered aluminate, the so-called “third aluminate hydrate” (TAH) [21-
215
23].
216
The 29Si and 27Al NMR spectra acquired for concrete II were quite different from that of
217
concrete I (Figure 3 B). In addition to C-S-H contributions between -75 and -95 ppm,
218
the
219
were assigned to anhydrous clinker phases (C3S – C2S) and quartz respectively [17, 19].
220
This difference is coherent as regards to the proportion of clinker in both concretes. In
221
concrete II, only 20 % of clinker was substituted by SCMs while in concrete I 80 % was
222
substituted. Consequently, the relative intensity of the C-S-H signal compared to the
223
clinker phase signal was weaker than for concrete I. In terms of Q1 and Q2 populations,
224
29
225
Figure 3 (B) clearly displays that Q1 and Q2 populations were quite similar traducing the
29
Al solid state NMR spectrum of
Si spectrum was composed of two narrow signals at -71.4 and -107.2 ppm that
AC C
29
EP
TE D
M AN U
SC
27
RI PT
Si NMR in Figure 3 A showed a predominant Q2 population compared to Q1
Si NMR showed also significant differences between concrete I and concrete II.
11
ACCEPTED MANUSCRIPT 226
greater presence of dimers in the tetrahedron chains in the C-S-H structure. The C/S
227
ratio of C-S-H calculated from the Q1/Q2tot content was ~ 0.9 [18]. The higher C/S ratio
228
of C-S-H in concrete II is consistent with the expected C/S regarding the concrete
229
composition and the chemical composition (Table 1 and 2).
230
exhibited a weak tetrahedral contribution between 50 and 80 ppm, corresponding to Al
231
substituting for Si in the C-S-H structure and a narrow signal in the range of
232
hexacoordinated aluminum. This resonance was composed of at least three
233
contributions obtained by deconvolution, centered at 6, 10 and 13 ppm, and was
234
assigned to many aluminum hydrates such as TAH [21-23], hydrogarnet [22, 24], AFm,
235
hydrotalcite and ettringite [23]. There was too much overlap to assign these signals to a
236
specific phase. The relative intensity of the C-S-H signal compared to that of phases
237
bearing hexacoordinated aluminum was weaker than for concrete I. Another broad
238
NMR shift ranging between 20 and 40 ppm was detected and possibly match the
239
metakaolin signal [25].
27
Al spectrum
AC C
EP
TE D
M AN U
SC
RI PT
The
240 241 242
Figure 3 – 29Si and 27Al NMR spectra of concrete I (A) and concrete II (B). The dotted lines are the result of the spectral decomposition. 12
ACCEPTED MANUSCRIPT 3.2. Image segmentation step 1: macroporosity and anhydrous phases
M AN U
SC
RI PT
243
244
Figure 4 – First processing step to threshold the macroporosity and the poorly hydrated cement phases of the hardened concrete I. The frequency histogram of the sum of atomic wt% was used to threshold the low hydrated phases of the whole map by limits. The thresholded pixels were plotted in a Si-Ca-Al3FeMg ternary diagram to discriminate the different poorly hydrated phases. The pixels selected with the polygon tool are displayed on a mineral map and superimposed on the BSE image.
251
The first step of mineralogical mapping was to segment the pore network, as described
252
in [6]. The macropores, corresponding to pore larger than the image pixel size (e.g., 2
253
µm) were segmented from the BSE images and the associated derivative of the
254
histogram of the gray level frequency (figure 4). All the pixels attributed to the
255
macropore network were back projected onto the BSE image and mapped. Only the
256
remaining pixels were used in the subsequent segmentation steps to identify minerals
257
(Figure 4).
258
One of the main issues with this method was to differentiate poorly hydrated initial
259
phases (anhydrous cement and additive phases) from highly hydrated phases with
AC C
EP
TE D
245 246 247 248 249 250
13
ACCEPTED MANUSCRIPT similar chemical composition. This issue was related to the kinetics of hydration of
261
clinker phases and slag grains, some of which remained unhydrated at the time of
262
sample analysis. As elements with low atomic numbers (H, C, O) were not analyzed by
263
EPMA, the sum of element weight concentration analyzed per pixel Σ (oxide) wt%
264
(including the stoichiometric O content) varied according to the proportion of H and C
265
composing the mineral (depending on the degree of hydration). For non-porous phases,
266
the theoretical Σ (oxide) wt% was calculated from the theoretical unit formula [7]. As
267
exemplified in Table 3, it was possible to discriminate the anhydrous clinker grains (Σ
268
(oxide) wt% = 100 %) from portlandite (Σ (oxide) wt%=76 %). According to the
269
polymodal distribution in the frequency histogram of Σ (oxide) wt% (Figure 4), the
270
transition between anhydrous phases and hydrates corresponded to a Σ (oxide) wt%
271
value of approximately 82 %. Considering the Gaussian distribution of Σ (oxide) wt% in
272
one phase, the minimum value of Σ (oxide) wt% equal to 82 % was considered for
273
thresholding all the pixels belonging to the unhydrated phases (Figure 4). Note that a
274
few pixels associated with a mix of hydrated and anhydrous phases were thus included
275
in this set of pixels.
276 277
Table 3 Example of Σ (oxide) wt% of unhydrated and hydrated cement phases used in the first segmentation step
AC C
EP
TE D
M AN U
SC
RI PT
260
Phases
C2 S
Formula Ca Si O*
2CaOSiO2 46 16
Portlandite Ca(OH)2 278
(Ati)wt%
54
0
38 22
H2O Σ (oxide) wt% (not analyzed) 100 24
76
*Total calculated stoichiometric oxygen
14
ACCEPTED MANUSCRIPT The pixels of the mapped area with a Σ (oxide) wt% above 82 % were plotted in a
280
ternary Si-Ca-Al3FeMg diagram for concrete I (Figure 4). Pixels related to clinker and
281
two types of slag were superimposed on the BSE images to verify that the segmented
282
pixels corresponded to grains in the BSE image. Some clusters corresponding to a part
283
of the pixels of weakly hydrated phases with Σ (oxide) wt% just above 82 % were
284
detected on the scatterplot but were not thresholded at this step.
AC C
285
EP
TE D
M AN U
SC
RI PT
279
286 287 288 289 290 291 292 293
Figure 5 – First processing step for concrete II to discriminate the hydrated and unhydrated MK. The frequency histogram of the sum of wt% (Σ (oxide) wt%) elements revealed a bimodal composition (pink polygon) in the pixel group belonging to the hydrated and unhydrated MK grains on the Si-Ca-Al3FeMg projection of the whole image (upper left and central figures). The segmentation by limits, considering the Σ (oxide) wt%, discriminated the hydrated MK (blue polygon) from the unreacted MK (purple polygon) with a threshold at 82 %. Their composition was then plotted on the Si-Ca- Al3FeMg diagram and the associated pixels back-projected onto the BSE image.
294
The process of segmentation based on the Σ (oxide) wt% was applied to concrete II as
295
shown in Figure 5, in which the Si-Ca-Al3FeMg ternary diagram illustrates the chemical
15
ACCEPTED MANUSCRIPT composition distribution of the whole pixels in the analyzed area of concrete II. The
297
polygon drawn in pink on the ternary diagram selected a cluster of pixels corresponding
298
to the pure end-members of MK grains and the associated mixture with C-(A-)S-H
299
(pink area on the BSE image). The Σ (oxide)wt% frequency histogram computed from
300
this selected cluster of pixels evinced the presence of a bimodal distribution with
301
Gaussian distributions centered on 75 and 90 wt%. This bimodal distribution for this
302
narrow chemical composition field implied that two phases with different hydration
303
states exist and could be segmented. Only pixels with Σ (oxide) wt% above 82 % were
304
segmented and plotted into a Si-Ca-Al3FeMg ternary diagram (Figure 5). The selected
305
pixels were segmented and displayed on the BSE images as MK (purple). The
306
remaining pixels corresponding to so-called “hydrated MK” were plotted again in the
307
same projection and segmented with the same polygon selection (as MK) in blue. The
308
corresponding pixels were distributed on an external ring of the anhydrous MK grains
309
(Figure 5). The illustration of the pixels on the BSE images confirmed that the selected
310
pixels correspond to grains and not to a mixture with the non-segmented binder pixels.
311
3.3. Image segmentation step 2: aggregates and portlandite
312
This step was identical for both concretes so we only describe it for concrete I (Figure
313
6). The pixels remaining after the first step were plotted into a sequence of ternary
314
diagrams (Figure 6) with different chemical axes to discriminate the pixels with
315
different chemical compositions. This use of sequential projection was well suited for
316
discriminating pixel clusters corresponding to pure end-members of carbonates or
317
quartz. Quartz grains were segmented by considering the pixels included in the yellow
318
polygon at the Si2 coordinate. The carbonates were segmented in different steps to
319
distinguish calcite, dolomite and siderite. This distinction was assessed by extracting all
320
pixels with high calcium content and devoid of Si (black polygon) in the Si2-CaFeMg-
AC C
EP
TE D
M AN U
SC
RI PT
296
16
ACCEPTED MANUSCRIPT Al3S3 plot and plotting them into another scatterplot to differentiate calcium, iron and
322
magnesium content (Figure 6). One of the main drawbacks linked with this aggregate
323
segmentation was the superimposition of theoretical compositional fields of calcite
324
(CaCO3) and portlandite (CaOH2) because among the analyzed elements, they were
325
only composed of Ca. As the C, O and H contents were not analyzed, we could not
326
distinguish then by only considering the measured chemical composition. This issue
327
was solved by considering the Σ (oxide) wt%, as an additional constraint: portlandite
328
has a CaO weight percentage of 75 % whereas calcite has a CaO weight percentage of
329
56 %. The pixels corresponding to the chemical composition of calcite and portlandite
330
were plotted in a new scatterplot Ca – Si – Σ (oxide) wt%. We could not detect any
331
portlandite in either concrete, as all the pixels were only stretched along a line towards
332
C-(A-)S-H composition that corresponds to calcite grains at the boundaries of
333
aggregates (Figure 6). This observation is consistent with the low-pH concrete
334
composition (Table 1).
AC C
EP
TE D
M AN U
SC
RI PT
321
17
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
335
Figure 6 – Segmentation methodology by polygon selection in ternary plots for the pixels remaining after step I for concrete I. The colored polygons outline the selected pixels for each mineral back-projected on the mineral maps.
339
3.4. Image segmentation step 3: S-bearing phases and hydrated phases
340
Due to the small number of pixels corresponding to S-bearing phases, all these phases
341
were differentiated according to their sulfur, calcium and aluminum content. The sulfur
342
frequency histogram (Figure 7) displayed progressive decrease in pixel frequency for
343
values between 0 and 0.11 sulfur atoms for every four oxygen atoms. This feature was
344
assigned to the background level at the position of the S Kα peak and corresponds to
345
pixels devoid of sulfur [6].
AC C
EP
TE D
336 337 338
18
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
346
Figure 7 – Thresholding methodology by polygon selection in ternary plots for the pixels remaining after step II for concrete I. The colored polygons outline the selected pixels for each mineral back-projected on the mineral maps.
350
The zone observed on the histogram having low pixel density with more than 0.1 sulfur
351
atom per 4 oxygen atoms, was assigned to the presence of S-bearing phases mixed with
352
different amounts of the surrounding hydraulic binder, although they could have been
353
related to sulfate in the C-S-H [26]. The selected pixels were then plotted on a S-Ca-Al
354
ternary plot for discriminating and mapping pyrite (FeS2), gypsum (CaSO4, not detected
355
here), monosulfate (Ca4(Al, Fe)2SO10 12 H2O) and AFt (Ca6(Al, Fe)2(SO4)3(OH)12 26
356
H2O) (Figure 7). Next, the remaining pixels were used spread out on a series of ternary
357
scatterplots in order to identify clusters of pixels related to hydrated grains.
358
Hydrogarnets (Ca3(Al, Fe)2(OH)12), zeolites (NaAlSi2O6 H2O) and hydrates formed by
359
the pozzolanic reaction of the silica fume with the calcium hydroxide (CaOH2) were
360
segmented in a Si2–CaFe–Al3 diagram. These minerals were segmented by successive
AC C
EP
TE D
347 348 349
19
ACCEPTED MANUSCRIPT polygons outlining the pixels that can be attributed to pure end-members or mixtures
362
between these pure end-members and C-(A-)S-H. The pixels attributed to the silica
363
fume hydrates spread out between the silica and C-(A-)S-H domain, while the pixels
364
assigned to the hydrogarnet domain spread out between the C-(A-)S-H domain and the
365
theoretical hydrogarnet solid solution domain. The selected pixels were plotted in a Ca-
366
AlSi-Fe diagram for discriminating the aluminous hydrogarnets from the ferrous
367
hydrogarnets. The aluminous hydrogarnets displayed compositions ranging from pure
368
Al-hydrogarnet (Ca3Al2(OH)12), devoid of Si, to katoite (Ca3Al2SiO4(OH)8) end-
369
members along the Ca and Al+Si solid solution line. The iron-rich hydrogarnet cluster
370
position in the Ca-AlSi-Fe diagram (Figure 7) indicated that this phase had an
371
intermediate composition in the composition range of the hydrogrossular –
372
hydroandradite solid solution [27]. The remaining pixels matched well with the
373
theoretical chemical composition domain of the C-(A-)S-H matrix [28] in the Si2–
374
CaFe–Al3 diagram (Figure 7). The associated cluster was isotropic without stretching
375
along a mixture line towards Al3 end-members of the scatterplot where the hydrotalcite
376
was located.
377
The pixels, which were previously identified as the C-(A-)S-H matrix, were plotted in a
378
Ca-Al-Mg6 diagram to inspect the variation of the chemical composition of the matrix
379
(Figure 7). In the Ca-Al-Mg6 plot, the pixels were distributed along a Ca-Mg line spread
380
between C-(A-)S-H and the hydrotalcite end-members. Two clusters were
381
distinguished. The upper gray polygon corresponded to pixels with the chemical
382
composition of C-(A-)S-H. The lower light blue polygon exhibited Mg-rich pixels
383
organized mainly around some of the unhydrated slag grains, where high fractions of
384
Mg were initially present. At first sight, we might interpret such spreading of
385
compositions along a line as a mixture including hydrotalcite and C-(A-)S-H
AC C
EP
TE D
M AN U
SC
RI PT
361
20
ACCEPTED MANUSCRIPT nanometric crystals. However, no mixture line between C-(A-)S-H and hydrotalcite
387
(Mg6Al2CO3OH16) end-member positions was detected in the previous Si2–CaFe–Al3
388
diagram. This attests to the presence of Mg-rich C-(A-)S-H instead of the presence of
389
hydrotalcite mixed with C-(A-)S-H.
390
4. Discussion
391
4.1. Quantitative spatial distribution of the minerals and phases
392
We
393
simultaneously using only one technique. The ternary projection tool (especially the Si-
394
Ca-Al3FeMg scatterplot) provided a direct overview of the chemistry and how it
395
evolved according to hydration state or the type of hardened cement materials (Figure
396
2). The different phases corresponded either to contrasted isotropic clusters or to
397
mixture/solid solution lines when plotted in these scatterplots. Since the pixel frequency
398
related to each cluster of pixels was encoded on a logarithmic color scale, we could also
399
qualitatively deduce the relative modal content of each cluster/mineral from visual
400
inspection of the scatterplot. For example, C2S and hydrogarnet (AlVI bearing mineral)
401
pixel clusters on the ternary projection were clearly less dense for concrete I than for
402
concrete II (Figure 2). Conversely, the cluster intensity corresponding to silica fume
403
hydrates was less intense for concrete II than for concrete I. This pixel distribution
404
related to the proportion of each phase over the ternary projection (Figure 2) is
405
consistent with the initial concrete compositions (Table 1). In the concrete II, only 20 %
406
of the clinker was substituted by SCMs, expecting to have more residual clinker phase.
407
As in the concrete I, 50 % of the clinker was substituted by slag, which are more
408
detected (e.g., higher wt%, table 4) in the hydraulic binder than in the concrete II. The
409
same observation could be done for the hydrated silica fume, the aluminous hydrate, for
mineral
types
and
associated
calculated
structural
formula
AC C
EP
TE D
M AN U
SC
quantified
RI PT
386
21
ACCEPTED MANUSCRIPT which the pixel distribution intensities (Figure 2) are consistent with the cement
411
compositions (Table 1).
412
4.2. Estimating hydrated phase compositions
413
The statistical composition of the C-(A-)S-H binder was obtained from a relatively large
414
sample volume including C/S and A/S ratios. In this study, the C/S of the two concretes
415
was 0.7 and 1.1 and the A/S was 0.06 and 0.2 for concretes I and II, respectively. These
416
C/S ratios were consistent with the information obtained by solid-state NMR data
417
(Figure 3) and the calculated Q2/Q1 ratio measured from the C-S-H chemical shift [16].
418
As expected, the calculated cationic formula showed that alkalis were present initially in
419
the cement and SCMs (slag, SF, MK). In the hardened concrete, alkalis were found in
420
the different hydrates, mainly in the hydrated silica fume and the C-(A-)S-H, which is
421
the main phase of the cement paste. As a consequence, local K/Na concentrations were
422
low and analyzing tiny changes in concentration for these elements was not relevant
423
since X-ray background intensities were not accounted for in the mapping mode (Table
424
4). Moreover, many of the hydrates (e.g. C-(A-)S-H or hydrogarnet) were in fact solid
425
solutions ranging between different end-members. The existence of such solid solutions
426
implied that it would have been difficult to quantify them by conventional methods such
427
as XRD or NMR because of lattice parameters evolving with chemical composition,
428
whereas mineral maps can discriminate the composition of such solid solutions. Two
429
types of C-(A-)S-H and solid solutions between hydrogarnet – katoite – stratlingite
430
domains were identified in concrete I. The chemical composition obtained for each
431
phase gave information on the composition of some poorly characterized phases such as
432
TAH, which was abundant in concrete II, with
433
mineral mapping results supported the idea that TAH could be considered as a solid
434
solution within the solid solution domain of aluminum-rich hydrogarnet – katoite, in
AC C
EP
TE D
M AN U
SC
RI PT
410
27
Al solid-state NMR (figure 3). Our
22
ACCEPTED MANUSCRIPT agreement with previous evidence based on chemical analyses of cement materials [21,
436
23].
437
4.3. Modal content of mineral and phases in concretes I and II
438
The wt(min) % for all the minerals in the hydrated cement paste obtained from the
439
mineral map was calculated from the spatial distribution of the minerals and their
440
measured modal composition (Table 4). This modal composition is only applied to the
441
map area and could be considered representative of the concretes’ compositions only if
442
this area was representative of the average concrete composition. The modal
443
composition is given according the whole composition (including the aggregates) and
444
normalized to the hydraulic binder composition (without considering the aggregates).
445
Obviously, considering the resolution of the map (1024 µm x 1024 µm) and considering
446
the size of the aggregates (> 4 000 µm), the mapped mineralogy (Table 4, mapped area)
447
is not representative of the whole concrete. The map was focused on the hydraulic
448
binder and the constitutive hydrates gel and the residual cement grains. The normalized
449
mass fraction was thus given (Table 4, normalized) without considering the aggregates.
450
15.4 wt% (14.5 wt% of slag) of residual cement grains were detected in the binder of
451
the concrete I, while only 6.4 wt% (3.3 wt% of C2S) were detected in the binder of the
452
concrete II. This proportion of anhydrous cement grains is coherent in view of the
453
percentage of clinker substitution and of the hydration kinetics of the slag.
454 455
Table 4 Weight (mineral) % of the whole analyzed area for concretes I and II and associated calculated cationic formula obtained from the mineral maps
AC C
EP
TE D
M AN U
SC
RI PT
435
Mineral
bind er
Aggregates
Concrete I Calcite Dolomite Siderite Quartz Belite Slag
Wt (min)%
Wt (min)%
Mapped area
Normalized
34.9 2.0 0.1 1.0 0.5 2.5
0.9 4.1
Calculated cationic formula CaCO3 Ca0.6Mg0.4CO3 FeCO3 SiO2 Ca2SiAl0.05O4 Na0.01K0.02Mg0.43Ca1.28Si1.1Al0.52O5
23
Na0.01K0.02Mg0.25Ca1.48Si1.18Al0.44O5 FeS2 Ca4Al2SO10:12H2O Ca3Al1Fe0.5Si0.5O6 :6H2O Ca3Al1.2Mg0.2Si0.5O8 :4H2O CaAl2Si4O12: 5.8H2O K0.01Ca0.34Si1.0Al0.02O2.4:1.4H2O Na0.01K0.02Ca0.56Mg0.15Si0.86Al0.16O2.8:2H2O
46.0
Na0.01K0.02Ca0.65Mg0.03Si0.94Al0.06O2.8:2.1H2O
100.0
CaCO3 Ca0.6Mg0.4CO3 FeCO3 SiO2 Ca2Si0.9Al0.05O4 Na0.01K0.02Mg0.25Ca1.5Si1.12Al0.36O5 K0.06Ca0.36Al1.5Si2.1O7 Ca6Al1.5S0.6O18:31H2O Ca3Al1.2Fe0.2Si0.2O6 :6H2O K0.03Ca0.58Si0.8Al0.2Fe0.03O2.8:1.8H2O K0.01Ca0.44Si0.95Al0.05O2.6:1.3H2O K0.02Ca0.94 Mg0.01Si0.84Al0.16O3.1:2.3H2O
SC
Normalized
RI PT
10.4 0.1 0.2 2.2 1.5 0.6 16.8 17.2
3.3 0.3 2.8 0.1 17.6 5.4 3.6 66.9
M AN U
Slag 6.4 Pyrite 0.1 AFm/Aft 0.1 Fe-Hydrogarnet 1.4 Al-Hydrogarnet 0.9 Ca zeolite 0.4 Hydrated silica fume 10.4 Mg-rich C-(A)-S-H 10.7 (light blue) C-(A)-S-H (gray) 28.6 Sum 100.0 Sum hydraulic binder 62 Mapped area Concrete II Calcite 40.4 Dolomite 2.6 Siderite 0.5 Quartz 5.2 Belite 1.7 Slag 0.2 Metakaolin 1.4 AFm/Aft 0.07 Al-Hydrogarnet 9.0 Hydrated metakaolin 2.8 Hydrated silica fume 1.8 C-(A)-S-H 34.3 Sum 100.0 Sum hydraulic binder 51.3
TE D
binder
Aggregates
ACCEPTED MANUSCRIPT
100.0
In particular, the C/S ratio of C-(A-)S-H determined for both concretes by quantitative
457
mapping in Table 4 is remarkably in accordance with the C/S ratio determined by 29Si
458
NMR as seen in section 3.1. That confirms the expected C/S ratio of C-(A-)S-H for
459
concrete
460
Moreover the relative proportion of hydrated silica fume and aluminous hydrates is also
461
coherent with the clinker and SCMs proportion given in the table 1 and the chemical
462
composition given in the table 2.
463
4.4. Spatial distribution of porosity
464
The nanometer scale microstructure of cement materials depends on the nanometer
465
scale porosity of the C-(A-)S-H gel phases [29]. Spatial resolution of BSE and X-ray
466
maps (2 µm in the present work) are not sufficient to resolve the nanometer pore
and II regarding their formulation and binder chemical composition.
AC C
I
EP
456
24
ACCEPTED MANUSCRIPT domain of this gel porosity. Nonetheless, the C-(A-)S-H gel volume should be corrected
468
for the presence of a porosity with a range of gel pore size. As the capillary porosity
469
(2.6 % and 5.9 % for concretes I and II, respectively), detected on the mapped area, was
470
smaller than the total bulk porosity of samples (16 ± 1 % and 13 ± 1 % for concrete I
471
and II, respectively); the undetected pore volume was probably associated to the
472
hydration products gel [29].
473
The gel porosity, ε, associated to the C-(A-)S-H gel phase was thus calculated from the
474
method developed by [7], from the difference between the theoretical sum of oxides
475
weight concentration Σtheo (oxides)wt% constitutive of C-(A-)S-H gel phases (~95 %,
476
table 5) and the value measured by EPMA that was actually measured (~70 %) by
477
applying the following formula for each of the pixels corresponding to the C-(A-)S-H
478
matrix:
SC
(
%) × ( %)
M AN U
=
with
(
%) =
TE D
479
RI PT
467
∑ "#$( ∑ '"((
!)%&% !)%&%
× 100
where ρr is the density of the resin that had polymerized in the pores, and ρm is the
481
mineral grain density (table 5). This method allows calculating the phase weight
482
proportion m(wt%) in the X-ray emission volume and to calculate the mean total
483
porosity (ε) within a X-ray volume and according to the spatial resolution (2 µm). This
484
method was thus adapted for cement materials to calculate the gel porosity.
485 486
Table 5 – Example of grain density and theoretical sum of oxide weight percentages Σtheo (oxides)wt% used in the porosity calculation.
AC C
EP
480
Mineral/phase Calcite Dolomite Quartz/Am. Si C-S-H0.6
Chemical composition CaCO3 CaMg(CO3)2 SiO2 Ca0.6SiO2.53(OH)0.55*
Σtheo (oxides) wt% 56 52 100 95
Grain density (g/cm3) 2.71 2.84 2.65 2.13* 25
ACCEPTED MANUSCRIPT Ca0.8SiO2.58(OH)0.57* Ca1.2SiO3.0(OH)0.73* Ca6Al2(SO4)3(OH)12:26H2O 3CaOSiO2 2CaOSiO2 Ca(OH)2
95 95 87 100 100 75
2.22* 2.42* 1.78 3.28 3.15 2.26
RI PT
487
C-S-H0.8 C-S-H1.2 Ettringite Alite Belite Portlandite * data from [30]
Using this method, we estimated the mean total porosity associated with each pixel, and
489
thus the intrinsic porosity of each phase, providing useful information on the porosity of
490
the domains prone to host diffusive processes [31]. The approach made it possible to
491
show that the largest part of the porosity was associated to C-(A-)S-H gel phases with
492
an intrinsic porosity of ~45 % (Figure 8 and Table 5). This gel porosity of 45 %, needs
493
to be linked to the average total porosity of the concrete (16 and 13 % ± 1), considering
494
the mass fraction of C-(A-)S-H gel phase into the concrete. In the present study, 75-80
495
% of the binder (which represent only 17-25% of the whole concrete composition, table
496
1) are considered to be C-(A-)S-H gel phases, whether 12-20% of C-(A-)S-H in a
497
concrete. Thus, porosity could be calculated considering the gel porosity weighted to
498
the mass fraction of gel phases, highlighting the need to consider the presence of
499
heterogeneous domains for the simulation of transport and reactivity processes in these
500
materials.
AC C
EP
TE D
M AN U
SC
488
501
26
ACCEPTED MANUSCRIPT Figure 8 – BSE images of the mapped area (A) and porosity map computed from the sum of oxide of each pixel (B).
504
4.5. Visualizing and identifying the hydration process
M AN U
SC
RI PT
502 503
505
Figure 9 – Mineral map of the analyzed area of concrete I. Zoom at the bottom of the maps illustrate the hydration process (in green and light blue) of the cement component (slag, SF)
509
From the mineralogical mapping we could discriminate the anhydrous phases and the
510
different hydrates and display the hydration process of some pozzolanic reactions
511
(Figures 9 and 10 for concretes I and II respectively).
512
In Figure 9, areas were magnified as indicated by the black squares to illustrate the
513
hydration process of the pozzolanic reactions. In concrete I, the light blue pixels (i.e.
514
Mg-rich C-(A-)S-H) revealed the heterogeneous distribution of the hydration reaction of
515
slag. The dark green pixels were associated with the pozzolanic reaction of silica fume.
516
The spatial distribution of the minerals detected showed a limited dispersion of hydrated
517
silica fume, which was mainly present in the form of clusters with size ranging from 10
518
µm to 100 µm. In concrete II, metakaolin particles (dark pink clusters in Figure 10),
519
which had the same grey level as the others hydrated phases (Figure 10, BSE image),
520
were identified with their contrasted chemical compositions (Figure 10). The grains
AC C
EP
TE D
506 507 508
27
ACCEPTED MANUSCRIPT were still present, with chemical compositions varying according to their hydration state
522
(dark blue ring in Figure 10).
523
In addition to the advantage of displaying spatial distribution of hydration reactions, the
524
segmentation by limits, based on the Σ (oxide) wt%, offered the possibility to quantify
525
the different anhydrous phases (cement components and SCMs) with their associated
526
chemical composition. The reaction pathways and changes in the chemistry of the
527
different anhydrous phases were identified, such as the hydration of the SF, slag and
528
MK.
529
EP
TE D
M AN U
SC
RI PT
521
Figure 10 – Mineral map of the analyzed area of concrete II. Zoom at the bottom of the maps illustrate the hydration process (in dark blue) of the cement component (MK)
532
In the case of concrete I, Mg-rich C-(A-)S-H identified in light blue in the map,
533
originated from the slow hydration of the slag. This phase with a C/S ratio of 0.65 had
534
an intermediate composition between the Mg-rich slag and the C-(A-)S-H identified in
535
gray with a C/S ratio of 0.7 (Figure 9). This evolution occurred with a release of Mg
536
(0.12 mol%) and Al (0.08 mol%), and Ca enrichment (0.11 mol%). For concrete II,
537
hydration of MK towards C-S-H was a stepwise process characterized by an
AC C
530 531
28
ACCEPTED MANUSCRIPT intermediate so-called hydrated MK phase whose composition domain was located in
539
the same region as the Mg-rich C-(A-)S-H of concrete I. Basically, the hydrated MK
540
and Mg-rich C-(A-)S-H compositions differed only by their contrasted Al, Fe and Mg
541
contents (table 4, (Al0.2Fe0.05)/O2.8 and (Al0.15Mg0.15)/O2.8, respectively).
542
The mineral maps indicated that the remaining SF was present in the hydraulic binder in
543
the form of clusters with sizes ranging from 10 to 100 µm (Figure 9). This observation
544
was also verified for the MK in concrete II and illustrated the low SCM dispersions
545
during concrete mixing. Local external layer of hydrated products were detected in the
546
boundaries of the coarser, and still unhydrated, SF and MK grains.
547
5. Conclusions
548
We proposed a method, adapted from [6], that displays quantitative spatial phase
549
distribution and the porosity constituting cement materials. While the spatial resolution
550
of the method (2 µm) could be considered a problem in view of the nanometric size of
551
the hydrated cement phases, we showed that the method offered could provide a full
552
mineralogy with associated phase compositions to supply geochemical modelling and to
553
provide better understanding of the hydration process governing physical and chemical
554
properties in cement materials. It yielded a straightforward discrimination of the
555
anhydrous phases and the various hydrated phases and gave a spatial resolution of the
556
hydration process in pozzolans over different observation scales. Two concretes with
557
different SCMs and environmental conditions (hydration condition) were investigated.
558
The hydration of concrete I, based originally on PC, SF and slag, resulted in the
559
formation of C-(A-)S-H mainly with low C/S and A/S, while for concrete II, based on
560
PC, SF and MK, the C-(A-)S-H exhibited a higher C/S and A/S. C-S-H characterization
561
by the quantitative mineralogical mapping was successfully correlated with
562
measurements in case of both concretes. In addition to the quantitative mineralogy, our
AC C
EP
TE D
M AN U
SC
RI PT
538
29
Si NMR
29
ACCEPTED MANUSCRIPT characterization method made it possible to visualize the spatial phase distribution. In
564
the case of concrete I, we highlighted the presence of clusters of hydrated silica fume.
565
For the pozzolan reactions, we further addressed the hydration rate by quantifying the
566
area corresponding to residual anhydrous grains compared to hydrated grains with the
567
same chemical composition. In a further step, by estimating the reactive surface areas
568
for each phase it should be possible to obtain realistic reactive surface areas for each
569
phase and mineral, following procedures established for other materials [32, 33]. This
570
could provide the kinetic parameters for reactive transport modeling approaches [34-
571
38].
M AN U
SC
RI PT
563
Acknowledgements
573
The results presented in this article were collected during the GL-ESC, MLH and UP-
574
Transfert projects granted by Andra as part of the Andra/BRGM scientific partnership.
575
We gratefully acknowledge contributions from Yannick Linard, Xavier Bourbon and
576
Guillaume Wille.
[1] A. Neville, The confused world of sulfate attack on concrete, Cement and Concrete Research 34(8) (2004) 1275-1296. [2] T. Schmidt, B. Lothenbach, M. Romer, J. Neuenschwander, K. Scrivener, Physical and microstructural aspects of sulfate attack on ordinary and limestone blended Portland cements, Cement and Concrete Research 39(12) (2009) 1111-1121. [3] S. Ahmad, Reinforcement corrosion in concrete structures, its monitoring and service life prediction––a review, Cement and Concrete Composites 25(4–5) (2003) 459-471. [4] T. Chappex, K. Scrivener, Alkali fixation of C–S–H in blended cement pastes and its relation to alkali silica reaction, Cement and Concrete Research 42(8) (2012) 1049-1054. [5] V.G. Papadakis, Effect of supplementary cementing materials on concrete resistance against carbonation and chloride ingress, Cement and Concrete Research 30(2) (2000) 291299. [6] D. Pret, S. Sammartino, D. Beaufort, A. Meunier, M. Fialin, L.J. Michot, A new method for quantitative petrography based on image processing of chemical element maps: Part I. Mineral mapping applied to compacted bentonites, American Mineralogist 95(10) (2010) 1379-1388. [7] D. Pret, S. Sammartino, D. Beaufort, M. Fialin, P. Sardini, P. Cosenza, A. Meunier, A new method for quantitative petrography based on image processing of chemical element maps: Part II. Semi-quantitative porosity maps superimposed on mineral maps., American Mineralogist 95 (2010) 1389-1398.
EP
578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
References
AC C
577
TE D
572
30
ACCEPTED MANUSCRIPT
EP
TE D
M AN U
SC
RI PT
[8] S. Gaboreau, D. Pret, E. Tinseau, F. Claret, D. Pellegrini, D. Stammose, 15 years of in situ cement-argillite interaction from Tournemire URL: Characterisation of the multi-scale spatial heterogeneities of pore space evolution, Applied Geochemistry 26(12) (2011) 2159-2171. [9] D. Massiot, B. Touzo, D. Trumeau, J.P. Coutures, J. Virlet, P. Florian, P.J. Grandinetti, Twodimensional magic-angle spinning isotropic reconstruction sequences for quadrupolar nuclei, Solid State Nuclear Magnetic Resonance 6(1) (1996) 73-83. [10] D. Massiot, F. Fayon, M. Capron, I. King, S. Le Calve, B. Alonso, J.O. Durand, B. Bujoli, Z.H. Gan, G. Hoatson, Modelling one- and two-dimensional solid-state NMR spectra, Magnetic Resonance in Chemistry 40(1) (2002) 70-76. [11] S. Gaboreau, C. Lerouge, S. Dewonck, Y. Linard, X. Bourbon, C.I. Fialips, A. Mazurier, D. Pret, D. Borschneck, V. Montouillout, E.C. Gaucher, F. Claret, In-Situ Interaction of Cement Paste and Shotcrete with Claystones in a Deep Disposal Context, American Journal of Science 312(3) (2012) 314-356. [12] P. Sardini, A. El Albani, D. Pret, S. Gaboreau, M. Siitari-Kauppi, D. Beaufort, Mapping and Quantifying the Clay Aggregate Microporosity in Medium- to Coarse-Grained Sandstones Using the C-14-Pmma Method, Journal of Sedimentary Research 79(7-8) (2009) 584-592. [13] D. Pret, P. Sardini, D. Beaufort, R. Zellagui, S. Sammartino, Porosity distribution in a clay gouge by image processing of C-14-PolyMethylMethAcrylate (C-14-PMMA) autoradiographs: Case study of the fault of St. Julien (Basin of Lodeve, France), Applied Clay Science 27(1-2) (2004) 107-118. [14] J.I. Goldstein, D.E. Newbury, P. Echlin, D.C. Joy, A.D.J. Romig, C.E. Lyman, C. Fiori, E. Lifshin, Scanning Electron Microscopy and X-ray Microanalysis, 2nd Edition Plenum Press, New York, NY, 1992. [15] C. Merlet, An Accurate Computer Correction Program for Quantitative Electron-Probe Microanalysis, Mikrochim Acta 114 (1994) 363-376. [16] X. Cong, R.J. Kirkpatrick, 29Si MAS NMR study of the structure of calcium silicate hydrate, Advanced Cement Based Materials 3(3–4) (1996) 144-156. [17] P. Colombet, A.-R. Grimmer, H. Zanni, P. Sozzani, Nuclear Magnetic Resonance Spectroscopy of Cement Based Materials, Springer-Verlag Berlin Heidelberg1998. [18] A. Nonat, The structure and stoichiometry of C-S-H, Cement and Concrete Research 34(9) (2004) 1521-1528. [19] X. Ke, S.A. Bernal, J.L. Provis, Controlling the reaction kinetics of sodium carbonateactivated slag cements using calcined layered double hydroxides, Cement and Concrete Research 81 (2016) 24-37. [20] I.G. Richardson, A.R. Brough, R. Brydson, G.W. Groves, C.M. Dobson, Location of Aluminum in Substituted Calcium Silicate Hydrate (C-S-H) Gels as Determined by 29Si and 27Al NMR and EELS, Journal of the American Ceramic Society 76(9) (1993) 2285-2288. [21] M.D. Andersen, H.J. Jakobsen, J. Skibsted, A new aluminium-hydrate species in hydrated Portland cements characterized by 27Al and 29Si MAS NMR spectroscopy, Cement and Concrete Research 36(1) (2006) 3-17. [22] B. Lothenbach, G. Le Saout, E. Gallucci, K. Scrivener, Influence of limestone on the hydration of Portland cements, Cement and Concrete Research 38(6) (2008) 848-860. [23] R.J. Myers, S.A. Bernal, R. San Nicolas, J.L. Provis, Generalized Structural Description of Calcium-Sodium Aluminosilicate Hydrate Gels: The Cross-Linked Substituted Tobermorite Model, Langmuir 29(17) (2013) 5294-5306. [24] M.D. Andersen, H.J. Jakobsen, J. Skibsted, Incorporation of aluminium in the calcium silicate hydrate (C-S-H) of hydrated portland cements: a high field 27Al and 29Si MAS NMR investigation, Inorganic Chemistry 42 (2003) 2280-2287. [25] B. Fabbri, S. Gualtieri, C. Leonardi, Modifications induced by the thermal treatment of kaolin and determination of reactivity of metakaolin, Applied Clay Science 73 (2013) 2-10. [26] R. Barbarulo, H. Peycelon, S. Leclercq, Chemical equilibria between C-S-H and ettringite, at 20 and 85 degrees C, Cement and Concrete Research 37(8) (2007) 1176-1181.
AC C
597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
31
ACCEPTED MANUSCRIPT
RI PT
SC
M AN U
TE D
EP
684 685
[27] B.Z. Dilnesa, B. Lothenbach, G. Renaudin, A. Wichser, D. Kulik, Synthesis and characterization of hydrogarnet Ca3(AlxFe1 − x)2(SiO4)y(OH)4(3 − y), Cement and Concrete Research 59 (2014) 96-111. [28] I.G. Richardson, The calcium silicate hydrates, Cement and Concrete Research 38(2) (2008) 137-158. [29] H.M. Jennings, J.J. Thomas, J.S. Gevrenov, G. Constantinides, F.J. Ulm, A multi-technique investigation of the nanoporosity of cement paste, Cement and Concrete Research 37(3) (2007) 329-336. [30] C. Roosz, S. Gaboreau, S. Grangeon, D. Prêt, V. Montouillout, N. Maubec, S. Ory, P. Blanc, P. Vieillard, P. Henocq, Distribution of water in synthetic calcium silicate hydrates, Langmuir (2016). [31] Y. Elakneswaran, A. Iwasa, T. Nawa, T. Sato, K. Kurumisawa, Ion-cement hydrate interactions govern multi-ionic transport model for cementitious materials, Cement and Concrete Research 40(12) (2010) 1756-1765. [32] C.I. Steefel, L.E. Beckingham, G. Landrot, Micro-Continuum Approaches for Modeling PoreScale Geochemical Processes, Reviews in Mineralogy and Geochemistry 80(1) (2015) 217-246. [33] G. Landrot, J.B. Ajo-Franklin, L. Yang, S. Cabrini, C.I. Steefel, Measurement of accessible reactive surface area in a sandstone, with application to CO2 mineralization, Chemical Geology 318–319 (2012) 113-125. [34] N.C.M. Marty, C. Tournassat, A. Burnol, E. Giffaut, E.C. Gaucher, Influence of reaction kinetics and mesh refinement on the numerical modelling of concrete/clay interactions, Journal of Hydrology 364(1-2) (2009) 58-72. [35] S. Molins, D. Trebotich, L. Yang, J.B. Ajo-Franklin, T.J. Ligocki, C. Shen, C.I. Steefel, PoreScale Controls on Calcite Dissolution Rates from Flow-through Laboratory and Numerical Experiments, Environmental Science & Technology 48(13) (2014) 7453-7460. [36] L.E. Beckingham, E.H. Mitnick, C.I. Steefel, S. Zhang, M. Voltolini, A.M. Swift, L. Yang, D.R. Cole, J.M. Sheets, J.B. Ajo-Franklin, D.J. DePaolo, S. Mito, Z. Xue, Evaluation of mineral reactive surface area estimates for prediction of reactivity of a multi-mineral sediment, Geochimica et Cosmochimica Acta 188 (2016) 310-329. [37] N.C.M. Marty, I. Munier, E.C. Gaucher, C. Tournassat, S. Gaboreau, C.Q. Vong, E. Giffaut, B. Cochepin, F. Claret, Simulation of Cement/Clay Interactions: Feedback on the Increasing Complexity of Modelling Strategies, Transport Porous Med 104(2) (2014) 385-405. [38] N.C.M. Marty, F. Claret, A. Lassin, J. Tremosa, P. Blanc, B. Madé, E. Giffaut, B. Cochepin, C. Tournassat, A database of dissolution and precipitation rates for clay-rocks minerals, Applied Geochemistry 55 (2015) 108-118.
AC C
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683
32