1
Microstructural and bulk property changes in hardened cement paste during the first drying process
2 3
I. Maruyama1,*, Y. Nishioka2, G. Igarashi2, and K. Matsui3
4 5
1
6
ES Building, No. 546, Furo-cho, Chikusa-ku, Nagoya 464-8603 Japan
7
*Corresponding author:
8
E-mail:
[email protected]
9
Tel: +81-52-789-3761, Fax: +81-52-789-3773
Assoc. Prof., Graduate School of Environmental Studies, Nagoya University,
10 11
2
12
ES Building, No. 539, Furo-cho, Chikusa-ku, Nagoya 464-8603 Japan
Graduate School of Environmental Studies, Nagoya University,
13 14
3
15
106 Someya, Sakai-machi, Sashima-gun, Ibaraki, 306-0493 Japan
16 17
Senior Researcher, Construction Materials Laboratory, Asahi-KASEI Construction Materials Corporation,
18
Abstract
19
This paper reports the microstructural changes and resultant bulk physical property changes in hardened cement paste
20
(hcp) during the first desorption process. The microstructural changes and solid-phase changes were evaluated by water
21
vapor sorption, nitrogen sorption, ultrasonic velocity, and
22
modulus, and drying shrinkage were also examined. The first drying process increased the volume of macropores and
23
decreased the volume of mesopores and interlayer spaces. Furthermore, in the first drying process globule clusters were
24
interconnected. During the first desorption, the strength increased for samples cured at 100% to 90% RH, decreased for
25
90% to 40% RH, and increased again for 40% to 11% RH. This behavior is explained by both microstructural changes in
26
hcp and C-S-H globule densification. The drying shrinkage strains during rapid drying and slow drying were compared
27
and the effects of the microstructural changes and evaporation were separated.
28
Keywords: Drying, Calcium Silicate Hydrate (C-S-H), Microstructure, Surface Area, Bending strength.
29
29
Si and
27
Al nuclear magnetic resonance. Strength, Young’s
30
1. Introduction
31
Among cement hydrates, calcium silicate hydrate (C-S-H) plays an important role in the physical properties of
32
concrete and cement paste, because it forms the matrix. Experimental studies of sorption properties, drying shrinkage,
33
strength, and Young’s modulus of hardened cement paste (hcp) confirm that C-S-H is a gel-like or colloidal system, and
34
the behavior of the system depends strongly on its surrounding water molecules [1-6]. Concrete is usually made with a
35
water-to-binder ratio of 0.50-0.60, and begins its service life with relatively high water content. Over time, concrete
36
releases water into the atmosphere, causing shrinkage and changes in physical properties. The performance of concrete is
37
strongly affected by the first desorption process that the hcp undergoes. Therefore, studies of C-S-H and cement paste
38
during the first desorption process are crucial for concrete engineering.
39 40
1.1 Microstructural changes
41
Mesoscale microstructural changes: N2 sorption is widely used for characterizing hcp. The Brunauer–Emmett–
42
Teller (BET) surface area of N2 (SN2) corresponds to the surface of the C-S-H gel particles in hcp. Hereafter, “globule” is
43
used according to the definition in ref [3]. N2 can penetrate into only the mesopores and can be adsorbed on only the
44
outer surface of globules; it does not reach the interlayers of the lamellar C-S-H structure [7].
45
Hunt et al. reported the difference in SN2 of hcp samples cured under different drying conditions [8]. A local minimum
46
value of SN2 was reached for curing conditions of around 50% RH under drying conditions from 80% to 5% RH. The
47
decrease in SN2 from saturated curing conditions to around 40% RH has also been reported by Parrott [9], and Litvan and
48
Myers [10]. Parrott concluded that the development of capillary tension stress in hcp is related to the closure of
49
nitrogen-accessible pores.
50
Small-angle X-ray scattering, which assumes that the globules are spherical, also shows a decrease in the estimated
51
surface area from saturated curing conditions to around 40% RH [11-13]. Small-angle neutron scattering (SANS) can
52
also detect mesoscale microstructural changes in hcp. Thomas et al. used SANS to study cement paste cured under
53
different RHs and resaturation conditions [14]. They found that the number of interparticle spaces (gel pores) decreased
54
and the bulk density of the C-S-H gel increased for curing conditions of 100% to 54% RH. Below 54% RH, a gas/C-S-H
55
gel interface was formed and the C-S-H gel lost the water adsorbed on the surface.
56
Environmental scanning electron microscopy gives a comprehensive view of the morphological changes in C-S-H
57
during drying. Fonseca and Jennings experimentally confirmed that fibrillar and amorphous morphologies of the outer
58
product C-S-H (Op C-S-H) are present under wet conditions, and the amorphous phase is dominant. This is consistent
59
with the reported microstructural changes mentioned above [15].
60
Interlayer spaces in C-S-H: Fledman and Sereda reported that the water adsorption process under low pressure is
61
interlayer rehydration [7]. Thus, the surface area derived from the water vapor isotherm (SH2O) corresponds to the
62
interlayer spaces in C-S-H globules. In addition, the decrease in SH2O caused by prolonged drying can be explained by the
63
difficulty in breaking the solid-solid contact in the interlayer space, due to its large connecting force.
64
Powers and Brownyard reported a linear relationship between chemically bound water and SH2O [16,17]. However,
65
Tomes et al. found that the SH2O of hcp decreases with repeated drying and wetting cycles, and that the total volume of
66
adsorbed nitrogen decreases [18]. Chiang et al. used SANS to determine the behavior of synthesized C-S-H globules
67
during drying. The globules were modeled as disks containing a lamellar structure. They experimentally confirmed that
68
the interlayer distance and the number of C-S-H layers in the globules decreased during drying [19].
69
Solid-phase changes: Parrott and Young carried out quantitative gel permeation chromatography of trimethylsilyl
70
silicate derivatives and found that the effect of prolonged drying on the silicate chain length was small [20]. However,
71
Bentur et al. observed polymerization in pure calcium silicate paste during the drying process [21]. Cong and Kirkpatrick
72
used nuclear magnetic resonance (NMR) to observe the polymerization of silicate chains in C-S-H. The polymerization
73
did not occur during drying at room temperature, whereas it did occur at elevated temperatures [22]. Aono et al. also
74
reported silicate polymerization in Portland cement paste during drying at 50 °C [23].
75 76
1.2 Physical properties of hcp during drying
77
The physical properties of dry hcp have been extensively studied by Feldman and Sereda [7,24,25]. Sereda et al. used
78
hcp and compacts of bottle hydrated cement. The strength of the hcp increased when it was strongly dried, and then
79
gradually decreased during the re-wetting process. Young’s modulus increased from 50% to 100% RH during the
80
rewetting process and exhibited a slight additional increase during the re-drying process with a maximum at around 30%
81
RH. However, these values were obtained after strong drying, and there were no data for the first desorption process.
82
Wittmann studied the properties of hcp during the first desorption process [26]. Young’s modulus decreased for
83
drying at 90% to 40% RH, reached a local minimum around 40% RH, and then increased for 40% to 11% RH. He
84
proposed that the increase in Young’s modulus for 40% to 11% RH was caused by the surface free energy of the C-S-H
85
gel and the decrease under 90% to 40% RH was caused by the disjoining pressure of the adsorbed water. The study did
86
not include data on the strength of the hcp between 100% and 40% RH, although the strength increased below 40% RH.
87
Wittman concluded that the surface free energy of the C-S-H gel also contributes to this behavior.
88
The behavior of the dynamic elastic modulus of hcp cured under different RHs is similar to that of Young’s modulus
89
[27]. Zech and Setzer reported a multicomponent model showing that the decrease in the dynamic elastic modulus under
90
100% to 40% RH can be explained by water content, although below 40% RH, the surface free energy should be taken
91
into account. Pihlajavaara measured many physical properties of mortar under different curing conditions [28]. The
92
behavior of the compressive strength is similar to that of Young’s modulus [26] and the dynamic modulus [27].
93
Studies of the shrinkage behavior of hcp during the first desorption process are limited. Helmuth and Turk showed
94
that irreversible shrinkage occurred at 100% to 47% RH and was strongly affected by the water-to-binder ratio. This
95
shrinkage was irreversible because of the collapse of the small pores, although after the hcp was dried at 11% RH, the
96
change in the hcp strain caused by changes in water content became reversible [29]. Recently, Setzer developed an
97
apparatus to measure the length-change isotherm. The length-change isotherms of the first desorption and the subsequent
98
re-wetting and re-drying processes had identical curves when the strain data were plotted as a function of the evaporable
99
water content [2]. Maruyama confirmed experimentally that the ratio of the amount of C-S-H to adsorbed water, which
100
can be represented by a statistical adsorption thickness, determined the change in the length of hcp during both the first
101
desorption process and the subsequent re-adsorption process [30].
102 103
1.3 C-S-H model and physical properties during the first drying
104
Feldman and Sereda proposed a model of hcp in which the layers or sheets in the C-S-H gel are rearranged during the
105
first desorption process, where the main driving force is the capillary tension force or the surface free energy acting on
106
the C-S-H gel. Consequently, the external surface area exposed to nitrogen decreases, producing irreversible shrinkage.
107
They attributed the irreversible shrinkage to the rearrangement of layers in the C-S-H gel [7].
108
Jennings proposed a colloidal model based on existing cement paste data for water vapor adsorption, density, C-S-H
109
stoichiometry, and specific surface area [1,3]. The data were unified using the basic D-drying conditions for C-S-H. In
110
the colloidal model, the C-S-H system is composed of C-S-H flocculations with two different densities. The C-S-H
111
system depends on time, temperature, RH, and the history of the C-S-H. The aging process consists of the polymerization
112
of the silicates in the C-S-H, and the globules become more tightly packed as it ages. Thus creep and shrinkage are a
113
function of the C-S-H aging [31]. In the refined CM-II model, C-S-H is mainly composed of 4-nm-thick C-S-H globules,
114
30-60 nm globule clusters, and water-filled spaces, which consist of interlayer spaces, inter-globule spaces, small gel
115
pores, and large gel pores.
116 117
1.4 Objective
118
Our research focuses on the microstructural changes during the first desorption process and the resultant changes in
119
the physical properties of hcp, such as the strength, Young’s modulus, and drying shrinkage. The relationship between
120
microstructural changes and the physical properties of hcp during the first desorption process has not been thoroughly
121
investigated. The previous studies of microstructural changes were reported based on the measuring methods and did not
122
focus on the changes in the physical properties of the material. The issues relating to changes in physical properties due
123
to drying have been discussed in terms of thermodynamics in stable porous materials, even though there are many reports
124
of microstructural changes occurring during the first desorption. In the present paper, microstructural change is explicitly
125
treated as a key factor in the changes in the physical properties. Data are collected for providing mechanistic explanations
126
for the changes in physical properties.
127
The microstructural changes are interpreted using the water vapor and nitrogen isotherms, and the changes in the
128
interlayer spaces, mesopores, and macropores are determined quantitatively. The nitrogen sorption isotherm is also used
129
to confirm the microstructural changes, and N2 accessible and inaccessible mesopores are discussed. NMR techniques are
130
used for the solid-phase transition during drying. In addition, ultrasonic measurement, particularly using shear waves, is
131
used for detecting large solid-phase differences from saturated curing conditions to slightly drier conditions, such as 95%
132
RH, which have not been thoroughly investigated. On the basis of these experimental results, we describe C-S-H globule
133
cluster conditions, globule densification, and the resultant pore structures.
134
Changes in strength during the first drying process are explained by the microstructural changes. By comparing the
135
shrinkage strain during long-term drying and short-term drying, the effects of microstructural changes and water
136
evaporation on the shrinkage strain are experimentally separated. The shrinkage mechanism and effect of microstructural
137
changes on shrinkage strain are discussed.
138 139
140
2．Experimental procedure
141
2.1 Specimen preparation
142
We used cement provided by the Taiheiyo Cement Corporation. The chemical components, physical properties, and
143
mineral compositions of white cement are shown in Tables 1 and 2. The mineral composition was determined by X-ray
144
diffraction (XRD) measurements and Rietveld analysis.
145
The water-cement ratio was 0.55, and the paste (10 L) was mixed in a 20 L Hobart mixer for 3 min after the water
146
was added, and then for a further 3 min after the paste was scraped from inside the mixer. All the materials were stored in
147
a thermostatic room at 20 ± 1 °C for 1 day prior to mixing. The mixing was performed at room temperature and the
148
specimens were then immediately moved to a thermostatic room. To minimize segregation, the paste was remixed every
149
30 min for 6 h.
150 151
The specimens were placed in a thermostatic chamber at a temperature of 20 ± 1 °C. They were demolded after 4 days, immersed in lime-saturated water for 180 days, and then dried under different RHs.
152
The specimen size was 3 × 13 × 300 mm to allow the samples to dry efficiently [30]. Specimens of 50 × 100 mm
153
were also made using light-weight steel molds. The 50 × 100 mm specimens and some of the 3 × 13 × 300 mm
154
specimens were cured and stored under lime-saturated water for up to 18 months.
155
The majority of the 3 × 13 × 300 mm specimens were placed in a controlled humidity chamber for 12 months after
156
180-day cruing under lime saturated water. The RH in the chamber was controlled with a sodium hydroxide solution. The
157
target RHs were 95%, 90%, 80%, 70%, 60%, 50%, 40%, 30%, 20%, and 11% at 20 ± 1 °C [30,32]. The RH was
158
monitored using a humidity sensor with a precision of 1.8% RH (Sensirion SHT75), and controlled to within ±2% RH
159
after the first 56 days. During the first 56 days, the relative humidity was gradually decreased. This process is defined as
160
“slow drying process”. For the slow drying process, air circulation was not used. Sodium hydroxide was used to avoid
161
carbonation of hcp. The slowly dried sample at a target RH is named SDSXX, where XX is the RH value. In addition,
162
sample SDS100b was the sample just before drying, and SDS100 was immersed in lime-saturated water for 18 months.
163
The experimental method and the sample names are summarized in Fig. 1.
164
2.2 Characterization of hydrated cement paste
165
- XRD/Rietveld analysis
166
The hcp specimens were analyzed by XRD (D8 advance, Bruker AXS) after curing for 12 months. All the hcp
167
specimens were submerged in acetone for 6 h and dried under vacuum for several minutes with an aspirator. The
168
specimens were stored at 11% RH and 20 °C for 2 weeks. The specimens were ground in a ball mill, and corundum
169
powder (10 mass %) was added to the sample powder as a standard reference. The XRD conditions were as follows: tube
170
voltage, 50 kV; tube current, 250 mA; scan range of 2 5-65°; step width, 0.02°; scan speed, 2°/min. The software used
171
for Rietveld analysis was TOPAS Ver. 4.0 (Bruker AXS). In the Rietveld analysis, the quantified phases were gypsum,
172
bassanite, portlandite, ettringite (AFt), monosulfate (AFm), as well as typical cement minerals, such as alite (C3S), belite
173
(C2S), cubic-C3A, orthorhombic-C3A, and C4AF. Structural models for clinker minerals (alite, belite, cubic-C3A,
174
orthorhombic-C3A, and C4AF) were taken from an NIST Technical Report [33]. Those for calcite, gypsum, bassanite,
175
portlandite (CH), AFt, AFm, and corundum were taken from the ICSD database [34]. The halo pattern of the amorphous
176
phase was refined as a background function. The samples were measured 3 times for each set of curing conditions. After
177
the Rietveld analysis, the degrees of hydration of the clinker minerals were calculated.
178
- Mass change
179
Specimen mass was measured by using a balance with a precision of 0.1 mg before and after drying for 12 months.
180
- Water content
181
The free water content in the hcp samples slowly dried under various RHs was determined by the mass difference
182
before and after drying at 105 °C. The 3 × 13 × 30 mm specimens reached equilibrium at 105 °C with no CO2 within 5
183
days. To determine the total pores in the cement paste, hcp specimens were re-saturated by submerging them in water
184
under a vacuum of ~0.1 hPa for 24 h at room temperature and then dried at 105 °C. The mass difference between the
185
specimens before and after drying at 105 °C was determined with a balance.
186
- Water vapor sorption isotherm
187
Water vapor sorption measurements were conducted by the volume method using a water vapor sorption analyzer
188
(Hydrosorb 1000, Quantachrome). A sample (~20 mg) was used for each measurement at 20 °C, with a pressure
189
tolerance of 0.05 mmHg and a time tolerance of 120 s. The measurement points on the adsorption and desorption
190
branches were at p/p0 = 0.05 intervals up to 0.95, and ended at 0.98 (RH = 98%). The SH2O of the adsorption branch was
191
calculated using BET theory with a water molecule section value of 0.114 nm2 [35]. The samples were ground in a ball
192
mill, and powder with diameters of 25-75 μm was used for this analysis. For pre-treatment, the samples were dried using
193
a vacuum pump (theoretical minimum pressure of 6.7 × 10-2 Pa, observed maximum pressure of 50 Pa) and heated to
194
105 °C with a heating mantle for 30 min. The specimens measured were SDS11 to SDS95 dried for 6, 9, and 12 months,
195
and SDS100.
196
- Nitrogen sorption test
197
The nitrogen sorption of the SDS11 to SDS95 samples dried for 9 months was measured using the same sample
198
conditions as those for the water vapor isotherms at 77.4 K with a Micrometrics TriStar II 3020. The SN2 of the adsorption
199
branch was calculated using BET theory.
200
- NMR measurements
201
29
Si MAS NMR spectra were collected on a Bruker BioSpin DSX 400 spectrometer (magnetic field of 9.4T) using a
202
CPMAS probe with 7 mm o.d. ZrO2 rotors at a spinning frequency of 5 kHz. A recycle delay of 100 s was used and the
203
total accumulation time was 800 s.
204
(-22.4 ppm with respect to liquid tetramethylsilane).
205 206
29
Si chemical shifts were referenced using silicon rubber as the external reference
SDS11, 30, 40, 70, and 90 after 12-month drying were measured. During measuring, the samples were as they were, and they contained the evaporable water. The measured signals were normalized to the mass of sample dried at 105 °C.
207
The chemical shift values of Q0, Q1, Q2(1Al), Q2, and Q3 were assumed to be -69.6 to -73.5, -75.6 to -79.0, -82.0,
208
-85.0, and -91.5 to -92.3 ppm, respectively, based on the shifts reported by Brunet et al. [36] and Rawal et al. [37]. The
209
peaks were decomposed assuming a Gaussian distribution. The mean chain length (MCL) of C-S-H and the Al/Si atomic
210
ratio in C-S-H were calculated with the following equations [38,39]
211
MCL =
212
Al/Si =
213
27
2 ( Q1 + Q2 + 1.5Q2 (1Al ) ) Q1 0.5Q2 (1Al )
Q1 + Q2 + Q2 (1Al )
,
.
eq. 1 eq. 2
Al MAS NMR spectra were collected on an ECA700 spectrometer (JEOL; magnetic field of 16.4 T) using a
214
CPMAS probe with an o.d. of 3.2 mm and ZrO2 rotors at a spinning frequency of 18 kHz. A recycle delay of 0.5 s was
215
used and the total accumulation time was 6400 s.
216
external reference (0 ppm). SDS11, 40, and 90 after 12-month drying were measured.
Al chemical shifts were referenced using AlK(SO4)212H2O as the
27
217
The chemical shift values of 4-coordinate Al in C-S-H are 71-76 ppm (bridging site) and 63-65 ppm (pairing site)
218
[37,40-42], and 6-coordinate Al in cement hydrates was confirmed by a synthesized sample other than the third aluminate
219
hydrate. The chemical shift for the third aluminate hydrate was based on literature values [41,43,44].
220
- Ultrasonic pulse velocity test
221
The ultrasonic pulse velocity of the P-wave (longitudinal elastic wave) and the S-wave (transverse elastic wave) of the
222
hcp samples were measured using an ultrasonic probe (V103-RM and V153-RM, Panametrics-NDT), and a pulser
223
receiver (5077PR, Panametrics-NDT). The voltage of the pulse oscillator was -400 V, the frequency was 1.0 MHz, and
224
the pulse repetition frequency was 100 Hz for the transmission method. The width of the samples was measured across
225
the 13 mm width with a digital micrometer caliper with an accuracy of 0.02 mm. The reference curves were obtained by
226
direct contact, and the period of the pulse peak in the reference curve was subtracted from the period of the pulse peak in
227
the sample record to determine the propagation time. The pulse velocities of the P-wave (vp) and S-wave (vs) were
228
calculated from the sample width and propagation time. Specimens after 12-month drying were measured.
229
2.3 Physical properties of hardened cement paste
230
- Loading experiment
231
The bending strength of hcp was obtained with a three-point bending test. The span was 60 mm and the center of the span
232
was loaded. The 3 × 13 × 300 mm specimens were used for the bending test, and the loading test was conducted four
233
times for each sample, and an average value was obtained from more than eight experiments. SDS100b and specimens
234
after 12-month drying were measured.
235
- Short-term length change isotherm (SLCI)
236
The shrinkage of hcp during the first desorption process is normally accompanied by microstructural changes. To
237
determine the shrinkage strain without the effects of long-term microstructural changes, a short experiment using small
238
specimens controlled by a forcible drying system, which is using RH-controlled air flow, was conducted. The short
239
drying time should minimize the microstructural changes in hcp. The length change isotherms for samples dried under
240
different RHs were obtained with a thermomechanical analyzer (TMA) coupled with an RH generator (AXS
241
TMA4000SA and HC9700, Bruker). Each experiment was completed within 50 h. Samples 3 × 3 × 1 mm in size were
242
cut from the dried specimens. Before the experiment, the samples were re-saturated with water under vacuum conditions
243
for 30 min and then placed in the apparatus. The experiment was started after stable conditions were achieved. The
244
SDS80, 60, 40, 11 after 12-month drying and SDS100b samples were measured. The target p/p0 values for the samples
245
other than SDS100b were 0.90, 0.75, 0.60, 0.50, 0.40, 0.30, 0.20, and 0.10, and the target p/p0 values for the SDS100b
246
samples were 0.90, 0.70, 0.60, 0.40, 0.30, 0.20, and 0.10. In each step, the p/p0 of the surrounding atmosphere was kept
247
constant for about 5 h.
248
The length change was measured by using a linear variable differential transformer (LVDT) with a precision of 0.5 m
249
and a contact load of 0.098 N.
250
During some steps, the shrinkage did not reach an equilibrium state, and the value was extrapolated by using the
251
following equation [45,46],
252
(t ) − RH
1 − b exp(−ct ) − 0.25b exp(−4ct ) −0.1b exp(−9ct ) = a , −0.0625b exp(−16ct ) −0.04b exp(−25ct )
eq. 3
253
where (t ) is the strain change (μ), RH is the equilibrium strain value of the last step, t is the measuring time for the
254
step (h), and a , b , and c are fitting parameters.
255
- Long-term length change isotherm (LLCI)
256
The length change of the 3 × 13 × 300 mm specimens before and after drying for 12 months under different RHs was
257
measured by LVDT with a precision of 0.001 mm [30]. Each specimen shrank with long-term microstructural changes.
258
Because the air in chamber was not circulated, the chambers reached the target RH after about 2 months drying. A
259
300-mm-long Fe-36Ni alloy sample was used as a reference and the length difference between the samples and the
260
reference was determined by LVDT. The length of the sample was calculated based on the length of reference. The
261
sample deformation was divided by the original sample length to obtain the shrinkage strain. For each set of drying
262
conditions, measurements were taken from more than five specimens. The curing conditions, curing period, and
263
experiments are summarized in Table 3.
264
265
3. Results and discussions
266
3.1 Sample condition
267
A typical phase composition of the SDS11 after 12-month drying calculated from the Rietveld analysis is shown in
268
Table 4. The monocarbonate phase was observed due to the original calcite content. The degrees of hydration of the
269
clinker minerals in hcp are summarized in Fig. 2. The degrees of hydration of the alite and aluminate phases were almost
270
independent of the RH of drying, except for the saturated samples (SDS100). In contrast, the degree of hydration of belite
271
from SDS70 to SDS95 increased with the RH of drying. The maximum difference in the degree of belite hydration for
272
the SDS11 to SDS95 samples was about 4%, and saturated sample showed a 7% increase in the degree of belite
273
hydration from SDS100b (6 months after casting) to SDS100 (18 months after casting).
274
3.2 Changes in hardened cement paste
275
(1) Water vapor and nitrogen sorption
276
The ratios of the mass difference caused by water evaporation for the samples dried under different RHs for 12
277
months (Fig. 3), and the water content of the samples (Fig. 4) were consistent. The re-saturated water content of the
278
samples is also shown in Fig. 3. Parrott et al. reported that the re-saturated water content reached a local minimum at
279
40% RH [11], which is similar to the local minimum observed at SDS30 in the present study. Parrott et al. concluded that
280
the decrease in the water capacity was caused by the spatial rearrangement of C-S-H gel and the closure of mesopores
281
arising from capillary tension stress.
282
The experimental water vapor sorption isotherms of the samples dried under different RHs for 6 months are shown in
283
Fig. 5. The sorption isotherms of samples from SDS95 to SDS70 showed no significant difference from p/p0 = 0 to 0.4.
284
However, the sorption behavior above p/p0 = 0.4 showed a clear decrease as the drying RH decreased. The water sorption
285
capacity above p/p0 = 0.4 is thought to result from the closure of mesopores. To confirm this, the incremental sorption
286
from p/p0 = 0.4 to 0.95, dW95-40, was plotted as a function of the drying RH (Fig. 6). It should be noted that dW95-40,
287
reflects a change in mesopores with Kelvin radii from 1.2 to 20 nm, and in CM-II, dW95-40 includes the size of large and
288
small gel pores among the C-S-H globules [3]. A decrease in dW95-40 from SDS95 to SDS40 and an increase from SDS40
289
to SDS11 were observed. It has been reported that the decrease in the number of mesopores from saturated conditions to
290
around 40% RH is caused by capillary tension force based on the results of N2 sorption [8-10] and low temperature
291
differential scanning calorimetry [47]. Figure 6 shows that dW95-40 decreased from 6 months to 12 months for all the
292
drying RH conditions, and that it was larger for samples dried at lower RHs. Therefore, we conclude that the force
293
driving the closure of the mesopores is surface free energy acting on the C-S-H globule clusters, because capillary
294
tension force does act on the globule clusters only above 40% RH.
295
Figure 7 shows the SN2 data for the samples dried for 9 months plotted as a function of the drying RH. The value of
296
SN2 decreased from SDS95 to SDS40 and there was no change from SDS40 to SDS11. This trend is similar to that
297
reported by Parrott et al. [9]. Feldman and Sereda suggested that nitrogen adsorption mainly measures the external
298
surface of the C-S-H gels [7], and this was experimentally confirmed by Tennis and Jennings [48].
299
The relationship between SH2O and the drying RH is summarized in Fig. 8. The samples dried at lower RHs had
300
smaller SH2O values; in particular, a sharp decrease was observed from SDS40 to SDS11. The decrease in SH2O from
301
SDS40 to SDS11 indicates irreversible behavior of the interlayer hydrate water, which is a structural and chemical
302
component [7]. Removing the interlayer water and the partial closure of the C-S-H interlayers produced cohesion
303
between C-S-H layers [3,7], which may be Si-O-Si or Si-O-Ca-O-Si bonding in the interlayers [49]. Note that the
304
difference in SH2O between SDS95 and SDS40 was almost the same value as that of SN2, therefore, the difference of SH2O
305
in this range is explained by the decrease of the nitrogen accessible mesopores, although comparing SN2 and SH2O may be
306
contentious [50]. And this fact indirectly indicates that in this RH range from 95% to 40% the layered structure of C-S-H
307
was not affected by the drying.
308
Assuming that both SN2 and dW95-40 reflect the surface area and volume of the mesopores, respectively, the decrease in
309
SN2 from SDS95 to SDS50 is consistent with the trend in dW95-40 from SDS95 to SDS40. The decrease probably resulted
310
from the morphological change in the low density Op C-S-H [15] and the consolidation of the C-S-H globule clusters
311
[3,7] during drying.
312
In contrast, from SDS40 to SDS11, the trends for SN2 and dW95-40 are different. dW95-40 increased as the sample was
313
dried at lower RHs, whereas SN2 was constant. Because H2O is polar, it can penetrate and move in very narrow pores [51].
314
We propose that at RHs of less than 40%, nitrogen inaccessible mesopores are created.
315
On the globule scale, the removal of interlayer water causes densification and a decrease in the volume of the C-S-H
316
globules. Therefore, under drying conditions of less than 40% RH, there are two different morphological change
317
mechanisms. One is at the meso scale, where the driving force is the surface free energy acting on the C-S-H globule
318
clusters, and it reduces the volume of the mesopores. We refer to this as “globule cluster consolidation”. The second is on
319
an atomic scale, and the removal of interlayer water by drying reduces the C-S-H globule volume. We refer to this as
320
“globule densification”. If the rate of globule densification is slower than the rate of globule cluster consolidation and the
321
globule cluster skeleton is fixed when the globule contains a large amount of water, then nitrogen-inaccessible internal
322
mesopores can be created after the globule cluster skeleton is formed. Figures 5 and 6 show that the 90% difference in
323
dW95-40 between SDS40 and SDS11 is explained by the decrease in adsorbed water at p/p0 = 0.4 resulting from the
324
decrease in SH2O.
325
Because of the globule cluster consolidation and globule densification during drying, the volume of the macropores
326
increased, which can be evaluated by the incremental water sorption from p/p0 = 0.95 to re-saturated conditions (dW100-95).
327
Figure 9 shows the water sorption apportionments of samples dried at different RHs. dW40-0 was calculated from the
328
adsorption branch and is incremental water sorption from dry conditions at 105 °C to p/p0 = 0.40. dW40-0 corresponds to
329
micropores, including interlayer spaces and small gel pores [3], with Kelvin radii of less than 2.1 nm. Figure 9 shows the
330
increase in macropores in samples dried at the lower RHs.
331
(2) NMR measurements
332
Figure 10 shows the 29Si-MAS NMR results. The Q0 peak increased for samples dried at lower RHs compared with
333
that for sample SDS90. The 29Si signal at 70.9 ppm is from -C2S [37]. This was explained by the slight increase in the
334
degree of hydration of belite at higher RHs. The Q1 and Q2 peaks were similar, whereas the Q2(1Al) peak increased as the
335
samples were dried at lower RHs. It has been reported that there are 4 types of Q 2 peak: the Q2p connecting proton peak
336
at -82.1 ppm; the Q2i connecting calcium ion peak at -83.9 ppm; the Q2v connecting Q3 site peak at -87.3 ppm; and the
337
Q2Ca peak at -84.3 ppm which is assigned to the silicate tetrahedra coordinated to the calcium planes [36,37]. In the
338
present study, the peak at -85.0 ppm is considered representative of the Q2 sites, because peak deconvolution was not
339
necessary for our purposes. The deconvolution of the
340
ratios of the peaks are summarized in Fig. 12. The Q3 peak was clearly visible in samples SDS30 and SDS11. Alizadeh
341
[50] reported that the Q3 peak appeared in synthesized C-S-H during the drying process at less than 11% RH, because of
342
the removal of water from the interlayers pulled the C-S-H sheets closer together and resulted in interlayer bonding. This
29
Si signals is presented in Fig. 11, and the relative composition
343
is indirect evidence of the mechanism responsible for the decrease in SH2O.
344
The MCL and Al/Si atomic ratio of C-S-H were calculated based on Fig. 12, and the results are shown in Fig. 13. The
345
MCL increased as the Al/Si ratio increased. The difference in MCL between SDS90 and SDS11 was about 15%;
346
however, this difference cannot fully explain the 45% decrease in SH2O. The mechanism of the decrease in SH2O is
347
probably the irreversible cohesion between C-S-H layers due to the removal of interlayer water.
348
Figure 14 shows the
27
Al-MAS NMR results for the SDS90, SDS40, and SDS11. The largest peak was the
349
6-coordinate Al signal from the monocarbonate, and the second largest peak was the 6-coordinate Al signal from
350
ettringite. The 6-coordinate peak from the third aluminate hydrate, which corresponds to Al in the interlayer of C-S-H or
351
Al(OH)3 [52], increased as the sample was dried at lower RHs. In the XRD chart, there was no gibbsite signal. Al atoms
352
probably migrate into the C-S-H interlayers during drying. The 4-coordinate Al peak, which was attributed to the Al in
353
pairing or bridging site in C-S-H [37,40-42], increased as the hcp was dried at lower RHs. This was inextricably linked to
354
the increase in Q2(1Al) in the 29Si NMR results.
355
The source of Al in hcp was investigated. The
27
Al NMR signal of ettringite decreased as the sample was dried at
356
lower RHs. This suggests that the Al is from ettringite. To confirm this, the ettringite quantification by XRD/Rietveld
357
analysis was plotted as a function of the drying RH (Fig. 15). Figure 15 clearly shows that ettringite decomposed at lower
358
RHs, whereas single-phase ettringite would not decompose at 11% RH and 20 °C [53]. The trend for the amount of
359
ettringite is similar to that for SN2. The surface free energy of the cement hydrates causes globule cluster consolidation
360
and this force may also affect the decomposition of ettringite surrounded by C-S-H globule clusters through “pressure
361
solution”.
362
(3) Ultrasonic pulse velocity
363
Figure 16 shows the P-wave pulse velocity, vp, as a function of the drying RH. Compared with SDS100, SDS95 had
364
slightly larger vp; vp decreased from SDS95 to SDS50 and increased from SDS50 to SDS11. Figure 17 shows the S-wave
365
pulse velocity, vs, which exhibited a trend similar to that of vp, although a large difference was observed between SDS95
366
and SDS100.
367
P-waves propagate in both the solid and liquid phases, whereas S-waves propagate in only the solid phase; therefore,
368
the large difference in vs between SDS95 and SDS100 reflected the number of solid-phase spatial connections.
369
Solid-phase connections are formed by Op C-S-H, thus, the characteristics of C-S-H at 100% RH were examined.
370
Usually, SEM or transmission electron microscopy (TEM) measurements are made under dry conditions, which means
371
the characteristics of C-S-H has not been clarified at 100% RH. Environmental SEM observations of C-S-H [15] suggest
372
that part of the C-S-H at 100% RH is amorphous. The ultrasonic pulse velocity measurements indicate that in the
373
amorphous C-S-H, the globule clusters were not strongly connected. Many weakly connected globule clusters could be
374
dispersed in the capillary water space at 100% RH. Although the change from 100% to 95% RH appears small, the water
375
content in hcp changed dramatically (Fig. 9). During the drying process, the capillary water located in pores with a
376
Kelvin radius greater than 21 nm evaporated. Furthermore, ions dissolved in the water precipitated as hydrates. The
377
capillary condensation localized the floating hydrates. Some of the precipitated and localized hydrates should cause
378
inter-globule bonding and orientate the C-S-H globules.
379
SEM or TEM observations of dry Op C-S-H show that it has fibrillar directional morphology, and the width of the
380
fibrils is about 100 nm [54]. This means that the fibrillar C-S-H cannot be submerged in capillary water at RHs of less
381
than 95%. Long-term drying at 95% RH may produce fibrillar C-S-H which connects the globules through inter-globule
382
bonding and orientation of the C-S-H globules.
383
Decreases in vp and vs were observed from SDS95 to SDS50. Setzer et al. showed that the decrease in vp can be
384
explained by the decrease in the water content [27], whereas the decrease in vs cannot. The decrease in SH2O from SDS95
385
to SDS50 was comparable with that of SN2 and this meant that the change in SH2O was caused by the change in the
386
globule cluster consolidation and not by changes in the globules. Therefore, the changes in the properties of the C-S-H
387
globule may not be the most important factor. Thus, the increase in the number of macropores and the decrease in the
388
number of mesopores (Fig. 9) delayed ultrasonic propagation.
389
Increases in vp and vs from SDS50 to SDS11 were observed. The low water content and the increase in the number of
390
macropores suggest the existence of another governing factor. Because the decrease corresponded to the decrease in SH2O,
391
the C-S-H globule densification and the resultant increase in stiffness probably caused the increase in vp and vs.
392
(4) Microstructural changes in hcp
393
Figure 18 summarizes the microstructural changes on the cement particle scale, the globule cluster scale, and the
394
globule scale. This figure is based on the model of C-S-H globules proposed by Jennings [1,3] and the changes in the
395
mesopores and the densification of C-S-H gel during the drying process proposed by Feldman and Sereda [7]. On the
396
basis of our experimental results, we have made the following additions to the model: a) inter-globule bonding from
397
saturated condition to 95% RH, b) consolidation of globule clusters by the surface free energy at low RHs and the closure
398
of mesopores, and c) internal mesopore production from the densification of globules below 40% RH.
399
These additional ideas rely on the mesoscale structure of C-S-H globules. This means our research partially supports
400
the CM-II model. The shape of globule and the network system require further investigation. The densification of C-S-H
401
globules does not explain the reappearance of nitrogen-inaccessible internal mesopores if the globule shape is assumed to
402
be a particle [1], a brick [3], or a disk [19]. The layered structure proposed by Feldman and Sereda [7] seems to be
403
plausible because piles of sheet type units easily make closed pores. The growth of C-S-H in alite paste or Portland
404
cement paste is thought to start from a CaO-monolayer structure [55,56], and the resultant fibrillar or foil-like C-S-H
405
structure also supports the presence of the CaO-layer network [54,57]. Therefore, we assumed that, under the condition
406
of less than 40% RH, the re-formed nitrogen-inaccessible mesopores existed in a layered “mille-feuille” structure
407
composed of CaO-layer sheets of different thicknesses, resulting from by partial lamination with water, silicate and Ca
408
ions.
409 410
411
3.3 Physical properties of hcp after the first slow drying
412
(a) Bending strength and Young’s modulus
413
The bending strength and Young’s modulus of the hcp dried at different RHs are shown in Figs. 19 and 20
414
respectively. The results before drying (SDS100b) are shown for comparison. The bending strength of the samples
415
increased from SDS100b to SDS90, decreased from SDS90 to SDS40, and increased again from SDS40 to SDS11. A
416
similar trend was observed for the Young’s modulus. These results were also similar to the results for the compressive
417
strength of mortar reported by Pihlajavaara [28].
418
It is generally accepted that strength of porous material is determined by the porosity and the strength of solid phase.
419
A large number of pores or large pores produce localized concentrated stress around the pore and decrease the bulk
420
strength. Additionally, increasing the strength of the solid phase directly affects the bulk strength. Therefore, if the hcp is
421
assumed to be a porous material, the pores and the strength of the solid phase should also be examined.
422
The increase from SDS100b to SDS90 can be explained by the inter-globule bonding, which results from the
423
orientation and rearrangement of the globules forming the fibrillar C-S-H. The decrease from SDS90 to SDS40 can be
424
explained by the increase in the number of macropores caused by the consolidation of globule clusters according to the
425
surface free energy. The globules structure and strength of globule do not change over the range of this drying condition.
426
The final increase from SDS40 to SDS11 can be explained by the change in globule strength caused by the densification
427
of the globules by the removal of water from the interlayers. The decrease in SH2O shown in Fig. 8 is evidence of
428
densification, and the ultrasonic velocity measurements shown in Fig. 16 and 17 indicate an increase in the stiffness and
429
strength. And this is the behavior of colloids.
430 431
Figure 21 shows the relationship between the bending strength and the macropore volume (dW100-95), and between the bending strength and SH2O. The data clearly show two different governing mechanisms.
432
We then obtained a qualitative comprehensive explanation of the behavior of the hcp strength during the first
433
desorption, in which C-S-H showed both colloidal and porous material properties and C-S-H was strongly affected by the
434
surrounding water molecules (Fig. 22).
435 436
(b) Short-term length change isotherms Figure 23(a) shows the strain change over time for SDS100b in a short-term length change isotherm (SLCI)
437
experiment, and Fig. 23(b) shows the changes in strain for SDS80, SDS60, SDS40, and SDS11. They showed a large
438
incremental shrinkage strain at 70% and 40% RH. The total shrinkage strain was on the order of that for SDS100b,
439
SDS80, SDS60, SDS11, and SDS40. These results clearly showed that the shrinkage strain was affected by the
440
microstructural change in hcp.
441
The SLCI results for SDS80, SDS60, and SDS11 shown in Fig. 24 have an inflection point at p/p0 = 0.40. Figure 23
442
indicates that expansive behavior was observed at p/p0 = 0.30. This RH corresponded to the conditions where the
443
meniscus disappears. Moreover, the expansive behavior has also been observed in the length-change isotherm of Vycor
444
glass in the desorption branch and was attributed to the diminishing capillary tension force [58]. The driving force behind
445
the shrinkage of hcp has been extensively discussed and several driving forces, such as surface free energy, capillary
446
tension force, disjoining pressure, and various combinations of these forces, have been proposed [2, 30, 31, 59-62].
447
However, our experimental evidence indicates that the capillary tension force is acting from 100% to 40% RH.
448
The inflection point, which reflects a change in the mechanism governing the shrinkage, suggests that the shrinkage
449
strain can be divided into two regions: from 100% to 40% RH (100-40), and from 40% to 10% RH (40-10) (Fig. 25). The
450
minimum value of 100-40 was observed in SC40, and the trend in 100-40 was similar to that for dW95-40, which was
451
calculated from the water vapor adsorption data. If the dominant mechanism in this range is capillary tension force, the
452
similarity between 100-40 and dW95-40 is reasonable because the shrinkage is a function of the mesopore volume [59,60,63].
453
In contrast, because 40-10 and SH2O are in the same order of magnitude, the shrinkage at less than 40% RH is governed by
454
surface free energy [7,59-62]. The strain for SDS11 is smaller than the strain estimated by using the values of dW95-40 or
455
SH2O; this can be explained by the high Young’s modulus.
456
(c) Drying shrinkage
457
The LLCI and SLCI are compared in Fig. 26. The strains for the SLCI results are shifted so that the SLCI and LLCI
458
strains overlap at the drying RHs. The shrinkage strain of SLCI is less affected by the microstructural changes caused by
459
a short drying time, whereas that of LLCI is strongly affected by microstructural changes. In the higher RH range, e.g.
460
from 100% RH to 40% RH, the shrinkage under the first drying depends on the volume of the mesopores, and this
461
volume of the mesopores can be crushed and decreased by the long-term drying at more than 80% RH as it is confirmed
462
by Fig. 6 and 25. Therefore, the drying at these RHs affects on the incremental shrinkage from the drying RH to 40% RH.
463
Shrinkage strain is drying history dependent and the rate of drying has an impact on the final drying shrinkage of hcp. On
464
the other hand, surface area and densification of C-S-H are also the function of drying process, and it also has the impact
465
on the shrinkage strain less than 40% RH.
466
It has been reported that samples with larger volumes show smaller ultimate drying shrinkage strain [64], The reason
467
of this phenomenon is explained by the hypothesis mentioned above. To predict the drying shrinkage strain of hcp during
468
the first desorption process, the rate of C-S-H globule densification and globule cluster consolidation play an important
469
role.
470
The LLCI results share a tangent line with the SLCIs at each target RH. Therefore, the LLCI line simulates the drying
471
process with a very slow water evaporation rate, because the C-S-H globule cluster changes occurred at similar rates of
472
water evaporation. These experimental results show the effects of microstructural changes in hcp on the drying shrinkage
473
strain.
474 475
4．Conclusion
476
White Portland cement paste samples were dried under different RHs for 1 year and their microstructure and physical
477
properties were investigated in order to understand the changes in the properties of hcp during the first desorption
478
process.
479
We report the following new findings:
480
1)
Water vapor sorption isotherms showed clear gradation of adsorption isotherms of hardened cement pastes
481
dried under different RHs. Based on the incremental sorption from p/p0 = 0.40 to 0.95, the mesopore volume
482
decreased from saturated condition to 40% RH. This was consistent with the nitrogen sorption results. However,
483
the increase in the mesopore volume from 40% to 11% RH was observed using the water vapor sorption
484
isotherms, whereas the nitrogen sorption isotherm did not indicate this. Nitrogen inaccessible internal
485
mesopores are generated below 40% RH.
486
2)
The water vapor sorption isotherms and the water content of the re-saturated samples were used to
487
quantitatively determined microstructural changes in macropores, mesopores, and micropores, including
488
interlayer spaces.
489
3)
490 491
The driving force of the consolidation of globule clusters is the surface free energy, because continuous closure of mesopores is observed across all RH conditions from 6 months to 12 months curing.
4)
Shear ultrasonic pulse velocity tests indicated a dramatic structural change from saturated conditions to 95%
492
RH. This is explained by the formation of inter-globule bonds by hydrates newly precipitated from evaporated
493
capillary water and the localization of floating hydrates caused by capillary condensation. 29
Si MAS NMR showed increases in the Q2(1Al) peak as hcp dried at lower RHs. This was confirmed by the
495
27
Al MAS NMR results. Consequently, the mean chain length of C-S-H was increased by drying. The Al source
496
was ettringite, which can be decomposed by “pressure solution”. In addition, the third aluminate hydrate peak
497
in the 27Al MAS NMR increased as the sample was dried at lower RHs. The Al atoms probably migrate into the
498
C-S-H interlayers during drying.
494
499
5)
6)
The increase in the Q3 peak in the 29Si MAS NMR as sample dried at low RHs was confirmed. This is indirect
500
evidence that the specific surface area of water vapor sorption was decreased by cohesion between the C-S-H
501
layers caused by the removal of interlayer water.
502
7)
503
By using the short-term length change isotherm, a slight expansion was observed from p/p0 = 0.40 to 0.30. This result proves that the capillary tension force acts above p/p0 = 0.40.
504
Based on these findings, we can conclude the following:
505
8)
During drying, hcp showed an increase in macropore volume and a decrease in mesopore volume and the
506
interlayer space. The increase in macropore volume and the decrease in mesopore volume were caused by
507
globule cluster consolidation due to the surface free energy. The decrease in interlayer space resulted from the
508
removal of water from the interlayers and subsequent cohesion between the C-S-H sheets. The removal of water
509
from globules leads to the densification of C-S-H, which was mainly seen at RHs below 40% and this produced
510
nitrogen inaccessible mesopores. These trends are consistent with the model proposed by Feldman and Sereda,
511
in which “major rearrangement of the layers or sheets of tobermorite gel” occurs during the first desorption, and
512
the experimental observations summarized by Jennings. We have proposed the following additions to these
513
models: a) inter-globule bonding from saturated conditions to 95% RH, b) globule cluster consolidation caused
514
by surface free energy in the dried state and the closure of mesopores, and c) internal mesopores production
515
caused by globule densification below 40% RH. We have constructed a schematic of the microstructural
516
changes.
517
9)
The strength and Young’s modulus of hcp was determined by the balance between the changes in the pores and
518
the solid-phase strength. From 100% to 90% RH, the strength increased. This was explained by the
519
inter-globule bonding. The decrease in strength from 90% to 40% RH was explained by the increase in the
520
number of macropores because of the consolidation of globule clusters. The increase in strength from 40% to
521
11% RH was explained by the densification of the C-S-H globules and the resulting increase in the strength of
522
the solid phase. Therefore, the changes in strength and Young’s modulus of hcp during the first desorption
523
process is affected by the colloidal and porous properties of the material.
524
10)
The drying shrinkage of hcp during the first desorption process is accompanied by microstructural changes. The
525
effects of microstructural changes and shrinkage caused by water evaporation were successfully separated by
526
short-term length change isotherm experiments. The incremental shrinkage strain from p/p0 = 1.0 to 0.40 was
527
related to the volume of the mesopores and the incremental shrinkage strain from p/p0 = 0.40 to 0.10 was related
528
to the specific water vapor BET surface area. Therefore, the driving force for drying shrinkage is a combination
529
of capillary tension force and surface free energy. Our results support Powers’ theory and Feldman’s theory.
530
The drying rate has the greatest effect on the shrinkage strain of hcp during the first desorption process, and the
531
ultimate shrinkage strain is not constant. It is necessary to investigate the rates of C-S-H globule densification
532
and the consolidation of globule clusters further.
533 534 535 536
11)
All our results can be explained using the mesoscale structure of C-S-H globules. However, the shape of C-S-H globules and globule linking require further investigation.
537
Acknowledgements
538
The work at Nagoya University was performed within the framework of the “Japan Ageing Management Program on
539
System Safety” project sponsored by the Nuclear and Industrial Safety Agency (NISA) and the Nuclear Regulation
540
Authority (NRA). We thank NISA and NRA for their sponsorship.
541
The authors also thank Mr. H. Nakayama (Mitsubishi Materials Corporation) for the nitrogen sorption measurements,
542
Ms. M. Nayuki and Mr. S.Yamazaki (Asahi-KASEI Corporation) for the NMR measurements and discussion of the
543
results, and Dr. T. Sugiyama (Bruker AXS) for help with the humidity-controlled TMA measurements.
544 545 546 547
The authors gratefully acknowledge helpful discussions with Dr. K. Yamada (National Institute for Environmental Studies), Dr. E. Gartner (Lafarge), and Dr. S. Tada (Text Inc.).
548
References
549
1
550 551
H.M. Jennings, A model for the microstructure of calcium silicate hydrate in cement paste, Cem. Concr. Res. 30 (2000) 101-106.
2
552
M. J. Setzer, The solid-liquid gel-system of hardened cement paste, Proceedings of Essen Workshop on Transport in Concrete: Nano- to Macrostructure, Setzer M. J. ed., Aedificatio Publisher, (2007) 3-23.
553
3
H. M. Jennings, Refinements to colloid model of C-S-H in cement: CM-II, Cem. Concr. Res., 38 (2008) 275-289.
554
4
G. Scherer, Structure and properties of gels, Cem. Concr. Res. 29 (1999) 1149-1157.
555
5
J. Thomas, H. Jennings, A colloidal interpretation of chemical aging of the C-S-H gel and its effects on the
556 557
properties of cement paste, Cem. Concr. Res. 36 (2006) 30-38. 6
H. M. Jennings, J. W. Bullard, J. J. Thomas, J. E. Andratde, J. J. Chen, G. W. Scherer, Characterization and
558
modeling of pores and surfaces in cement paste: Correlations to processing and properties, J. Adv. Conc. Tech., 6(1)
559
(2008) 5-29.
560
7
561 562
change and mechanical properties, Mater. Struct. 1(6) (1968) 509-520. 8
563 564
9
10
G. G. Litvan and R. E. Myers, Surface area of cement paste conditioned at various relative humidities, Cem. Concr. Res. 13 (1983) 49-60
11
569 570
L. J. Parrott, W. Hansen, R. L. Berger, Effect of first drying on the pore structure of hydrated alite paste, Cem. Concr. Res. 10 (1980) 647-655.
567 568
C. M. Hunt, L. A. Tomes, R. L. Blaine, Some effects of aging on the surface area of Portland cement paste, J. Res. Natl. Bur. Stand. 64(2) (1960) 163-169.
565 566
R. F. Feldman and P. J. Sereda, A model for hydrated Portland cement paste as deduced from sorption-length
D. N. Winslow and S. Diamond, Specific surface of hardened Portland cement paste as determined by small-angle X-ray scattering, J. Am. Ceram. Soc., 57(5) (1974) 193-197.
12
J. Krop, T. Grafenecker, H. K. Hilsdorf, Characterization of microstructure of hydrated cement paste by small angle
571
X-ray scattering, Principles and Applications of Pore Structural Characterization, RILEM/CNR, Milan, (1983)
572
339-363.
573
13
Völkl, J. J., Beddoe, R. E. and Setzer, M. J., The specific surface area of hardened cement paste by small-angle
574 575
X-ray scattering-Effect of moisture content and chlorides, Cem. Concr. Res., 17 (1987) 81-88. 14
576 577
cement with age, drying, and resaturation, J. Am. Ceram. Soc., 91(10) (2008) 3362-3369. 15
578 579
16
17
18
19
20
21
22
23
24
J. Sereda, R. F. Feldman, E. G. Swenson, Effect of sorbed water on some mechanical properties of hydrated Portland cement pastes and compacts, Highw. Res. Bd., Spec. Rep., 90 (1968) 58-73.
25
598 599
Y. Aono, F. Matsushita, S. Shibata, Y. Hama, Nano-structural changes of C-S-H in hardened cement paste during drying at 50oC, J. Adv. Conc. Tech., 5(3), (2007) 313-323.
596 597
X. Cong, J. Kirkpatrick, Effects of the temperature and relative humidity on the structure of C-S-H gel, Cem. Conc. Res. 25(6) (1995) 1237-1245.
594 595
A. Bentur, et al. Creep and drying shrinkage of calcium silicate pastes III. A Hypothesis of irreversible strains, Cem. Concr. Res., 9 (1979) 83-96.
592 593
L. J. Parrott and J. F. Young, Effect of prolonged drying upon the silicate structure of hydrated alite paste, Cem. Conc. Res., 11 (1981) 11-17.
590 591
W-S, Chiang, E. Fratini, P. Baglioni, D. Liu, S-H Chen, Microstructure determination of calcium-silicate-hydrate globules by small-angle neutron scattering, J. Phys. Chem. C, 116 (2012) 5055-5061.
588 589
L. A. Tomes, C. M. Hunt, and R. L. Blain, Some factors affecting the surface area of hydrated Portland cement as determined by water-vapor and nitrogen adsorption, J. Res. Natl. Bur. Stand. 59(6) (1957) 357-364.
586 587
T.C. Powers and T. L. Brownyard, Studies of the physical properties of hardened Portland cement paste: Part 3, Theoretical interpretation of adsorption data, Proc. Am. Conc. Inst. 43 (1946) 469-504.
584 585
T.C. Powers and T. L. Brownyard, Studies of the physical properties of hardened Portland cement paste: Part 2, Studies of water fixation, Proc. Am. Conc. Inst. 43 (1946) 249-336.
582 583
P. C. Fonseca, H. M. Jennnings, The effect of drying on early-age morphology of C-S-H as observed in environmental SEM, Cem. Concr. Res. 40 (2010) 1673-1680.
580 581
J. J. Thomas, A. J. Allen, H. M. Jennings, Structural changes to the calcium-silicate-hydrate gel phase of hydrated
R. F. Feldman, Factors affecting Young’s modulus - porosity relation of hydrated Portland cement compacts, Cem. Conc. Res., 1(2) (1972) 375-386.
26
F. H. Wittmann, Interaction of hardened cement paste and water, J. Am. Ceram. Soc., 56(8) (1973) 409-415.
600
27
601 602
pore size distribution, Mater. Struc. 22 (1989) 125-132. 28
603 604
29
30
31
32
33
P. Stutzman, S. Leigh: NIST Technical Note 1441-Phase Composition Analysis of the NIST Reference Clinkers by Optical Microscopy and X-ray Powder Diffraction, (2002) 34-43.
34
615 616
Stokes, R. H. and Robinson, R. A., Standard solutions for humidity control at 25 oC, Industrial and Engineering Chemistry, 41(9) (1949) 2013.
613 614
H. M. Jennings, Colloid model of C-S-H and implications to the problem of creep and shrinkage, Mater. Struct. 37 (2004) 59-70.
611 612
I. Maruyama, Origin of Drying Shrinkage of Hardened Cement Paste: Hydration pressure, Journal of Advanced Concrete Technology, 8(2) (2010) 187-200.
609 610
R. A. Helmuth and D. H. Turk, The reversible and irreversible drying shrinkage of hardened Portland cement and tricalcium silicate pastes, J. Polt. Cem. Assoc., Res. Dev. Lab., 9(2) 8-21 1967.
607 608
S. E. Pihlajavaara, A review of some of the main results of a research on the ageing phenomena of concrete: effect of moisture conditions on strength, shrinkage and creep of mature concrete, Cem. Concr. Res., 4 (1974) 761-771.
605 606
B. Zech, M. J. Setzer, The dynamic elastic modulus of hardened cement paste, Part II: Ice formation, drying and
Fachinformationzentrum Karlsruhe and National Institute of Standards and Technology, Inorganic Crystal Structure Database, ICSD., (2006) http://icsd.ill.fr/icsd/index.html
35
R. S. Mikhail, S. A. Selim, “ Adsorption of organic vapors in relation to the pore structure of hardened Portland
617
cement pastes”, Proc. Symp. Struct. Portland Cem. Paste Concr., Highw. Res. Board Spec. Rep., Washington D. C.
618
90 (1966) 123-134.
619
36
F. Brunet, Ph. Bertani, TH. Charpentier, A. Nonato, J. Virlet, Application of
29
Si Homonuclear and 1H-29Si
620
Heteronuclear NMR correlation to structural studies of calcium silicate hydrate, J. Phys. Chem. B, 108 (2004)
621
15494-15502.
622
37
623 624 625
A. Rawal, B. J. Smith, G. L. Athens, C. L. Edwards, L. Roberts, V. Gupta, B. F. Chmelka, Molecular silicate and aluminate species in anhydrous and hydrated cements, J. Am. Chem. Soc., 132 (2010) 7321-7337.
38
I. G. Richardson, G. W. Groves, The structure of the calcium silicate hydrate phases present in hardened pastes of white Portland cement/blast-furnace slag blends, J. Mater. Sci. 32 (1997) 4793-4802.
626
39
J. Skibsted, M. D. Andersen, The effect of alkali ions on the incorporation of aluminium in the calcium silicate
627
hydrate (C-S-H) phase resulting from Portland cement hydration studied by 29Si MAS NMR, J. Am. Ceram. Soc.,
628
96 (2013) 651-656.
629
40
630 631
Hydrates and Related Hydrates of Cement Pastes by 27Al MQ-MAS NMR, Inorg.Chem., 37 (1998) 3726-3733. 41
632 633
42
43
44
45
46
47
48
49
R. Alizadeh, Nanostructure and Engineering Properties of Basic and Modified Calcium-Silicate Hydrate Systems, Ph.D thesis, University of Ottawa (2009).
50
650 651
P. D. Tennis, H. M. Jennings, A model for two types of calcium silicate hydrate in the microstructure of Portland cement paste, Cem. Conc. Res. 30 (2000) 855-863.
648 649
D. H. Bager, E. J. Sellevold, Ice formation in hardened cement paste, Part II - Drying and resaturation on room temperature cured pastes, Cem. Conc. Res. 16 (1986) 835-844
646 647
S. Tada, S. Xishan, K. Watanabe, A Dynamic Method Determining Sorption Isotherms for Cement-based Materials, Proc. JCI, 22(2) (2000) 775-780.
644 645
J-F Daian, Condensation and isothermal water transfer in cement mortar, Transport in Porous Media, 3 (1988) 563-589.
642 643
I. G. Richardson, J. Skibsted, L. Black, R. J. Kirkpatrick, Characterisation of cement hydrate phase by TEM, NMR and Raman spectroscopy, Adv. Cem. Res. 20 (2010) 233-248.
640 641
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, Cem. Conc. Res., 36 (2006) 3-17.
638 639
X. Pardal, F.Brunet, T. Charpentier, I. Pochard, A. Nonat, 27Al and 29Si solid-state NMR characterization of calcium-aluminosilicate-hydrate, Inorg. Chem.,51 (2012) 1827-1836.
636 637
P. Faucon, A. Delagrave, J. C. Petit, Aluminum incorporation in calcium silicate hydrates (C-S-H) depending on their Ca/Si ratio, J. Phys Chem. B, 103 (1999) 7796-7802.
634 635
P. Faucon, T. Charpentier, D. Bertrandie, A. Nonat, J. Virlet, J. C. Petit, Characterization of Calcium Aluminate
K. K. Aligizaki, Pore structure of cement-based materials, Testing, interpretation and requirements, Taylor & Francis, London, (2005) 153-167.
51
I. Odler, J. Hagymassy, E.E. Bodor, M. Yudenfreund, S. Brunauer, Hardened Portland cement pastes of low
652 653
porosity IV. Surface area and pore structure, Cem. Conc. Res. 2 (1972) 577-589 52
P. Faucon, T. Charpentier, A. Nonat, J. C. Petit, Triple-quantum two-dimensional
27
Al magic angle nuclear
654
magnetic resonance study of the aluminum incorporation in calcium silicate hydrates, J. Am. Chem. Soc. 120
655
(1998) 12075-12082.
656
53
657 658
I. Baauerizo, T. Matschei, K. Scrivener, Impact of water activity on the water content of cement hydrates, Proc. Int. Conf. Build. Mater. "18. ibausil", Vol. 1, Weimar, Germany, September 12-15 (2012) 249-260.
54
I. G. Richardson, Tobermorite/jennite- and tobermorite/calcium hydroxide-based models for the structure of C-S-H:
659
applicability to hardened pastes of tricalcium silicate, beta-dicalcium silicate, Portland cement, and blends of
660
Portland cement with blast-fume, Cem Concr Res. 34 (2004) 1733-1777.
661
55
662 663
microscopy, Cem. Conc. Res., 30 (2000) 817-822 56
664 665
57
K. E. Kurtis, P.J.M. Monteiro, J. T. Brown, W. Meyer-Ilse, High resolution transmission soft X-ray microscopy of deterioration products developed in large concrete dams, J. Microsc. 196 (1999) 288-298
58
668 669
S. M. Clark, G. R. Omrrison, W. D. Shi, The use of scanning transmission X-ray microscopy for the real-time study of cement hydration, Cem Concr, Res. 29 (1999) 1099-1102
666 667
E. M. Gartner, K. E. Kruits, P. J. M. Monteiro, Proposed mechanism of C-S-H growth tested by soft X-ray
C. H. Amberg, R. A. McIntosh, A study of adsorption hysteresis by means of length changes of a rod of porous glass, Can. J. Chem., 30 (1952) 1012-1031.
59
T. C. Powers, The mechanisms of shrinkage and reversible creep of hardened cement paste, Proc. Int. Symp. of “Structure of Concrete and its behavior under load”, London, (1965) 319-344.
670 671
60
W. Hansen, Drying shrinkage mechanisms in Portland cement paste, J. Am. Ceram. Soc., 20(5) (1987) 323-328.
672
61
F. H. Wittmann, Shrinkage mechanisms of hardened cement paste, Cem. Concr. Res., 17(3) (1978) 453-464.
673
62
F. Beltzung, F. H.Wittmann, Role of disjoining pressure in cement based materials, Cem. Concr. Res., 35(12)
674 675
(2005) 2364-2370. 63
676 677
P.Schiller, Th. Bier, M. Wahab, H. -J. Mogel, Mesoscopic model of volume changes due to moisture variations in porous materials, Colloids Surf., A, 327 (2008) 34-43.
64
A.D. Ross, Shape, size and shrinkage, Conc. Const. Eng., 39(8) (1944) 193-199.
678
65
Table 1 Chemical composition of white cement by X-ray fluorescence elemental analysis.*
Ig. loss
SiO2
Al2O3
Fe2O3
CaO
MgO
SO3
Na2O
K2O
TiO2
P2O5
MnO
(%)
(%)
(%)
(%)
(%)
(%)
(%)
(%)
(%)
(%)
(%)
(%)
2.93 679
66
680
67
681
68
22.43
4.67
0.16
65.69
0.98
53.1 ± 1.41
33.2 ± 1.50
C3A (%)
Bassanite
Gypsum
Calcite
(%)
(%)
(%)
(%)
2.01 ±
1.84 ±
0.23
0.11
0.5 ± 0.27 0.37
70
684
0.17
Periclase
5.46 ±
683
0.07
0.03
Table 2 Mineral composition of white cement determined by powder XRD and Rietveld analysis.
C2S (%)
69
0.00
*Conducted by Taiheiyo Cement Corporation.
C3S (%)
682
2.51
C: CaO, S: SiO2, A: Al2O3, M: MgO.
3.96 ± 0.7
Cl (%) 0.00
0.00
○
30
○
○
40
○
○
50
○
○
60
○
○
70
○
○
80
○
○
90
○
○
95
○
○
6 Months
○
SLCI
○
○
Bending test
20
○
UPVT
○
○
Al NMR
○
27
○
11
Si NMR
Saturated conditions
29
0
XRD
Curing conditions (% RH)
Density
Curing period
N2 sorption
H2O sorption
Water content
Table 3 Summary of experiments conducted on samples prepared under different curing conditions and periods.
Length
72
Mass
685
○
○
○
○
○
○
○
○
○
○
○
Saturated conditions
9 Months
11
○
○
20
○
○
30
○
○
40
○
○
50
○
○
60
○
○
70
○
○
80
○
○
90
○
○
95
○
○
Saturated conditions
12 Months
11
○
○
○
○
○
○
20
○
○
○
○
○
○
30
○
○
○
○
○
○
○
40
○
○
○
○
○
○
○
50
○
○
○
○
○
○
○
○
60
○
○
○
○
○
○
○
○
70
○
○
○
○
○
○
○
○
80
○
○
○
○
○
○
○
○
90
○
○
○
○
○
○
○
○
95
○
○
○
○
○
○
○
○
○
○
○
○
○
Saturated conditions 686
73
687
74
688
75
689
76
○
○
○
○ ○
○
UPVT: Ultrasonic pulse velocity test. SLCI: Short-term length change isotherm
○ ○ ○
690
77
691
78
692
Table 4 Phase composition of white cement paste determined by powder XRD and Rietveld analysis. The sample was cured at 11% RH.
C3S (%) 1.31 ±
2.83 ±
0.40
0.41
693
79
694
80
695
C2S (%)
AFt (%)*2
Portlandite
Calcite
C3AH6
(%)
(%)
(%)*1
13.3 ± 0.93
3.47 ±
1.18 ±
2.63 ±
0.74
0.13
0.24
C: CaO, S: SiO2, *1H: H2O, *2ettringite, *3monocarbonate.
AFm(C)
Amorphous
(%)*3
(%)
7.80 ± 0.53
67.3 ± 2.99
Total (%)
99.82
696 697 698 699
Fig. 1 Experimental flow and sample names.
700
701 702
Fig. 2 Degree of hydration of the alite, belite, and aluminate phases. Samples were dried and stored in various RHs for 12
703
months, except for SC100b which is the saturated sample at 6 months.
704
705 706
Fig. 3 Mass ratio of hcp samples after dried for 12 months at various RHs. The mass ratio is defined as the ratio of mass
707
after drying to the mass before drying.
708
709 710
711
Fig. 4 Water content of samples dried for 12 months at various RHs, and the water capacity of the re-saturated samples.
712
Fig. 5 (a) Water vapor adsorption isotherms of samples dried at various RHs for 6 months. (b) Water vapor sorption full
713
isotherms of SDS95 and SDS70. (c) Water vapor sorption full isotherms of SDS70, 40, and 11. Water vapor sorption
714
isotherms of samples are obtained at 293 K by volumetric method.
715 716
717 718
Fig. 6 The dW95-40 of samples dried at various RHs for 6, 9, and 12 months. The dW95-40 is defined as the incremental
719
adsorption from p/p0 = 0.40 to p/p0 = 0.95.
720
721
722
Fig. 7 Specific surface area (SN2) of specimens dried at various RHs for 9 months. SN2 was calculated by the BET theory
723
from nitrogen adsorption isotherms. The nitrogen isotherm of samples were obtained at 77 K.
724
725 726
Fig. 8 Comparison of SH2O with SN2. SH2O was calculated by the BET theory from water vapor sorption isotherms at 293 K
727
and samples for water vapor sorption isotherms were dried at various RHs for 6, 9, and 12 months. SDS100 is the
728
saturated sample at the age of 18 months.
729 730
731
732 733
Fig. 9 Water sorption apportionments of samples dried at various RHs for 12 months.
734
735 736 737
Fig.10 29Si-MAS NMR measurements for SDS90, 70, 40, 30, and 11. The samples were dried for 12 months.
738 739
Fig. 11 Deconvolution of the NMR signals.
740
(Q0: 69.6, 70.9, 73.5 ppm. Q1: 75.6, 79.0 ppm. Q2(1Al): 82.0 ppm. Q2: 85.0 ppm. Q3: 92.0 ppm.)
741
742 743 744
Fig. 12 Relative composition ratios of the 29Si-MAS NMR peak intensity for C-S-H of SDS90, 70, 40, 30, and 11.
745 746
Fig. 13 Mean chain length (MCL) and Al/Si ratio of C-S-H in hcp samples dried for 12 months at 90, 70, 40, 30, and 11
747
RH.
748 749
Fig.14 27Al-MAS NMR measurement result of SDS90, 40, and 11. The samples were dried for 12 months at 90 40 and
750
11% RH.
751
752 753 754
Fig. 15 Amount of ettringite in hcp samples dried at various RHs for 12 months.
755 756
Fig. 16 P-wave ultrasonic velocity in hcp samples dried at various RHs for 12 months. The saturated sample SDS100 was
757
at the age of 18 months.
758
759 760
Fig.17 S-wave ultrasonic velocity in hcp samples dried at various RHs for 12 months. The saturated sample SDS100 was
761
at the age of 18 months.
762 763
764 765
Fig. 18 Microstructural changes in hcp on the scale of the cement particles, globular flocs, and globules (LGP: large gel
766
pore; SGP, small gel pore).
767 768
769
770 771
Fig. 19 Bending strength of hcp samples dried at various RHs for 12 months (SDS95 ~ SDS11) and sample before drying
772
(SDS100b).
773
774 775
Fig. 20 Young’s modulus of hcp hcp samples dried at various RHs for 12 months (SDS95 ~ SDS11) and sample before
776
drying (SDS100b).
777 778
779 780
Fig. 21 Bending strength of hcp as a function of (a) dW100-95, indicating changes in porosity and (b) SH2O, indicating
781
changes in the solid phase. The dW100-95 is defined as incremental adsorption from p/p0 = 0.95 to saturation and indicates
782
the volume of macropores in hcp sample. The SH2O is water vapor BET surface area of hcp.
783
784 785 786 787
Fig. 22 Schematic of relationship between the microstructural changes caused by drying at various RHs and strength.
788
(a) SDS100b
789
790
(b) SDS80, 60, 40 and 11.
791 792
Fig. 23 Shrinkage strain as a function of elapsed time for (a) SDS100b (before drying) and (b) SDS80, 60, 40, and 11.
793
Samples are dried for 12 months.
794 795
796 797
Fig. 24 Short term length change isotherm (SLCI) of the samples SDS100b, 80, 60, 40, and 11. SDS100b is the sample
798
before drying, and SDS80 ~ SDS11 were the samples dried for 12 months.
799
800 801
Fig. 25 Incremental shrinkage strain from 100 to 40% RH (100-40) and from 40 to 10% RH (40-10). The trends for 100-40
802
and 40-10 are compared with those for dW95-40 and SH2O, respectively.
803
804 805
Fig. 26 Comparison of the Long-term length change isotherm (LLCI) with the short term length change isotherms
806
(SLCIs) of SDS100b, 80, 60, and 40. LLCI results were measured shrinkage strains of the samples dried at various RHs
807
for 12 months without air circulation. SLCI results were the length change isotherms of re-saturated dried samples under
808
compulsory RH-controlled air flow. SDS100b is the sample before drying, and SDS80, 60, and 40 were the samples
809
dried for 12 months at 80, 60, and 40% RH without air circulation.
810 811
