Nano and Sub-nano Multiscale Porosity Formation and Other Features

Aug 25, 2014 - 953 West Bonita Avenue, Claremont California 91711-4193, United States. •S Supporting Information. ABSTRACT: Cement hydration occurs ...
0 downloads 0 Views 3MB Size
Download PDF
Article pubs.acs.org/Langmuir

Nano and Sub-nano Multiscale Porosity Formation and Other Features Revealed by 1H NMR Relaxometry during Cement Hydration Villiam Bortolotti,† Leonardo Brizi,‡,§ Robert J. S. Brown,∥ Paola Fantazzini,*,‡,§ and Manuel Mariani‡,§ †

DICAM and ‡Department of Physics and Astronomy, University of Bologna, 40100 Bologna, Italy § Centro Enrico Fermi, 00184 Rome, Italy ∥ 953 West Bonita Avenue, Claremont California 91711-4193, United States S Supporting Information *

ABSTRACT: Cement hydration occurs when water is added to cement powder, leading to the formation of crystalline products like Portlandite and the quasi-amorphous, poorly crystalline, calcium silicate hydrate (C−S−H) gel. Despite its importance in determining the final properties of the cement, many models exist for the nano and sub-nano level organization of this “liquid stone.” 1H NMR relaxometry in White Portland Cement paste during hydration allowed us to monitor the formation and evolution of the multiscale porosity of the cement, with the formation of structures at nano and sub-nano levels of C−S−H gel (calcium silicate interlayer water, water in small and large gel pores) along with three low-mobility 1H pools, identified as 1 H nuclei in C−S−H layers, likely belonging to OH groups, with 1H nuclei in Portlandite, and in crystal water of Ettringite. By assuming these assignments, our data allowed us to compute the distances of pairs of 1H nuclei in Portlandite and in crystal water ((1.9 ± 0.2) Å and (1.6 ± 0.1) Å, respectively), consistent with the known values of these distances. The picture of the porous structure at nano and sub-nano levels emerging from our results is consistent with the Jennings colloidal model for C−S−H gel. Moreover, the constant values observed during hydration of parameters extracted from our data analysis strongly support that model, being compatible with the picture of C−S−H gel developing in comparable-sized clumps of the same composition, but not easily interpretable by models proposing quasi continuous sheets of C−S−H layers.

1. INTRODUCTION Cement is a material exploited for generations, but notwithstanding its current impact on industrial activities and environment, a model for organization at nano and sub-nano levels of this “liquid stone”, that greatly influence its final properties, is not unanimously accepted. The hardened cement is formed through hydration, a complex process following the addition of water to cement powder, leading to the formation of reaction products. The most important are the calcium silicate hydrate (C−S−H) gel, a complex nanostructured material generally assumed to be made of calcium silicate layers, with interlayer water, and of pores of different sizes filled with water (the so-called gel pores), Portlandite (calcium hydroxide, Ca(OH)2), and Ettringite (3CaO.Al2O3.3CaSO4.32 H2O) crystals. Small−angle neutron scattering (SANS), X-ray scattering, quasi-elastic neutron scattering, sorption isotherms,1−4 and computer simulations5,6 have contributed to improving the understanding of C−S−H gel, but many models still exist for its nanostructure,7−11 even if many of them are similar and there is more consensus than might appear at first sight.7 Existing structural models are based substantially on two theories: one considers the C−S−H gel as formed by irregular interconnected © 2014 American Chemical Society

layers with adsorbed and interlayer water molecules; the other supposes the existence of basic C−S−H units, as a colloid made of small “bricks” and its associated gel pores.10 Many results, including from the grid indentation technique,12 SANS10 and Grand Canonical Monte Carlo simulations,13 calorimetry and other techniques,11 suggest that C−S−H exhibits a nanogranular behavior, and a colloidal model, combined with a layer-like structure of the colloidal units, appears to be the most effective approach to describe the characteristics of this material.11 Jennings4 has proposed a hybrid model of the nanostructure of C−S−H gel, based on the interpretation of water sorption isotherms, schematically described in ref 2. C−S−H is seen as a collection of clumps of sheet-like structures (nanometer size, referred to as “globules” in ref 4), made of calcium silicate layers containing OH groups, separated by interlayer spaces with physically bound water. A thin layer of water is adsorbed on the external surfaces of the clumps, and liquid water fills the nanopores among the particles. Because of the observation Received: January 24, 2014 Revised: August 10, 2014 Published: August 25, 2014 10871

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877

Langmuir

Article

from “liquid-like” pools of higher-mobility nuclei with longer, exponential, FID decay. The T1 relaxation analysis of the solid signal allowed us to distinguish two solid components, one with T1 of a few milliseconds and another with T1 of the order of a second. One was ascribed to 1H of anisotropically bound groups in the C−S−H gel and the other to 1H nuclei in crystalline components. The liquid components were ascribed to interlayer or intergrain spaces of the order of sub-nano and nanometers. In ref 34, a better solid−liquid separation allowed us to identify an intermediate T2* liquid-like component with very short T2 (∼80 μs) and T1 of a few milliseconds, both quite stable during hydration, that was ascribed to water confined in the interlayer spaces of the gel. A component with T2 on the order of 100 μs has been reported,21−23 and more recently a signal with T2 of this order of magnitude has been ascribed to “water diffusing in planar C−S−H gel pores”35 and to interlayer water within C−S−H gel.9 In refs 32−34, the relaxation times of the liquid-like component with much longer T2* were seen to move during hydration to shorter and shorter values, with lower intensities, because of the surface relaxation effect for water in the intergrain spaces and in the C−S−H gel pores that were being formed. Liquid water molecules are confined in spaces that become smaller and smaller, varying over orders of magnitude during hydration, down to sizes ranging from nanometers to fractions of a nanometer. In this paper we have better described the behavior of the different components during hydration and found that the “solid” signal is due to three different separable components. The NMR results appear to be consistent with the colloidal model4 proposed for the C−S−H gel. To avoid misunderstanding when referring to layer spaces and gel pores, in the following we will adopt the definitions given in ref 4, where the pores are defined “as having a free surface when empty and the interlayer space as space with surfaces that collapse when water is removed”.

of very large local deformations on drying, C−S−H is described as behaving more like a granular material than a porous continuous material. According to this model, the globules are packed with statistically well-defined patterns, and different classes of water-filled spaces are considered. Nuclear magnetic resonance (NMR) relaxation of 1H nuclei techniques have contributed in recent years to advancing the knowledge of the nanoscale structure and water location during hydration,14 with the advantage that these techniques do not require removing water. Key properties of NMR relaxometry are (1) the proportionality of NMR signal to the number of 1H nuclei in the sensitive volume,15 (2) the dependence of longitudinal and transverse relaxation times on correlation times for molecular motion, allowing separation of lowermobility from higher-mobility 1H nuclei on the basis of their free induction decay (FID) signals,15 (3) the dependence of the relaxation times T1 and T2 of the nuclear magnetization of 1 H nuclei of water completely filling a pore on the surface-tovolume (S/V) ratio of the pore (1/T1,2 = ρ1,2 S/V, where ρ1,2 is the surface relaxivity), under the assumptions that the bulk relaxation rates are negligible and the molecular diffusion is fast enough to maintain the magnetization uniform.16−20 Under these conditions, the distribution of relaxation rates observed in a porous medium corresponds to a “pore size” distribution. On the basis of these properties, different pools of 1H nuclei can be distinguished and assigned to different components and/or compartments in the cement paste. In ref 21, by means of 1H NMR transverse relaxation, the spaces of the gel pores were investigated and appeared to be in the range 0.5−10 nm. In ref 22, the thickness of tightly bound gel water layers between the sheets was estimated to be in the range 1−2 nm, and pores were estimated to vary in size over orders of magnitude. Experiments to distinguish classes of signals by two-dimensional T1−T2 and T2−T2 relaxation time measurements, analyzed by algorithms for inverse Laplace transform (2D-ILT), showed a distribution of classes of pore sizes in the range 1−20 nm.23−29 In ref 9, five narrow 1H signal peaks were found in cement at 10 days of hydration, assigned to 1 H nuclei in Portlandite and Ettringite crystals and in water in C−S−H interlayers and in different classes of pores. It has been known for years21,29 that there are present in cement very low-mobility 1H nuclei, distinguishable from higher mobility ones by the effect of molecular mobility on T2, although this possibility has been only recently exploited.30−35 In ref 30, the relative specific area and the widths of intra−C−S−H sheets and inter− C−S−H gel pores were estimated from the ratio of the solid echo to FID signal amplitudes. In their experimental conditions, the widths were found to be 1.5 and 4.1 nm, respectively. The sensitivity of NMR measurements to the different mobilities of 1H nuclei has been used to study the kinetics of hydration and the formation of the hydration products of commercial endodontic cements.32−34 The analysis of the FID of 1H from White Portland Cement pastes, combined with quasi-continuous T1 distribution analysis, enabled the authors to distinguish different 1H pools, operating a sort of 2D T1−T2 correlation analysis on a somewhat more physical than mathematical basis. In all the samples, the FIDs showed the formation during hydration of two major 1 H signal components, distinguished in the FIDs by the different decay time constants (T2*). A “solid-like” 1H signal (S) had a quasiGaussian FID with transverse relaxation time on the order of 10-15 μs. This is typical of systems with rigid pairs of dipolar coupled spin 1/2 nuclei.36 The “solid-like” signal was separated

2. EXPERIMENTAL SECTION The cement investigated is TECHBIOSEALER Standard (ISASAN s.r.l., Como, Italy), a White Portland Cement (WPC) of standard oxide composition (EN-197.1), with added calcium chloride (5% by weight, to accelerate hydration) and bismuth oxide (15% by weight, for radioopacity, United States Patent: US 8, 075,680 B2, by Mongiorgi R. et al.). Samples were prepared by adding 250 mg of water to 500 mg of the cement powder, mixed at the bottom of a 10 mm external diameter glass NMR tube, immediately sealed with Parafilm. More information about the cement composition is reported in the Supporting Information. NMR measurements were performed systematically at the following hydration times tH (1, 2.5, 4, 6, 7.5, 9, 25, 33, 50, 57, 73, 79, 99, 168, 240 h), at 20 MHz and 25 °C. Also, data at 2.5 years were acquired. Longitudinal relaxation data were acquired by inversion-recovery (IR) pulse sequences (πx − tIR − (π/2)x − FID acquisition), with 128 tIR values chosen in geometrical progression, starting at 50 μs and extending to maximum times varying from 0.4 to 5 seconds. The 128 FID signals acquired at a given tH were analyzed by best fit to eq 1.34 This fit allowed us to divide the total FID signal into three components: a longer-T2* liquid (exponential) component called liquid-long (LL), a shorter-T2* liquid (exponential) component called liquid-short (LS), and a quasi-Gaussian solid component (S).

G(t ) = SS

⎛ ⎞2 ⎡ −1/2Gfrac⎜ t ⎟ ⎢ ⎝ Tg ⎠ 1 e

⎢ ⎣

−

⎛ t ⎞2 1 (1 − Gfrac)⎜⎜ ⎟⎟ 2 ⎝ Tg ⎠

⎛ ⎞4 ⎤ 1 2⎜ t ⎟ ⎥ + Cc(1 − Gfrac) ⎜ ⎟ + SLL e−t / T2LL + SLSe−t / T2LS ⎥ 4 ⎝ Tg ⎠ ⎦ (1) 10872

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877

Langmuir

Article

where SS, SLL, and SLS are the solid, liquid-long, and liquid-short FID amplitudes extrapolated to FID time t = 0. Tg is the time constant of the Gaussian with the same initial curvature in a log plot as that of the solid component of the FID, Gfrac is the fraction of that curvature due to the Gaussian factor, and Cc affects recovery from the negative excursion due to the algebraic factor. T2LL and T2LS are FID T2* for LL and LS, respectively. For more information, see the Supporting Information, where the fit procedure is described in detail. The signals SS(tIR), SLL(tIR), and SLS(tIR) of these components as functions of tIR were determined at each tH. The inversion from tIR times to T1 relaxation times gave us the distributions of T1 relaxation times for the three components LL, LS and S at each tH. The inversion was performed by the algorithm UPEN,37−39 implemented in UpenWin software.40

3. RESULTS 3.1. Characterization of the Different 1H Pool Signals. Figure 1 shows, as an example, the experimental longitudinal

Figure 1. Experimental longitudinal relaxation data of the three FID components, LL, LS, and total solid (S = SS + SL), obtained as functions of inversion time (tIR) at hydration time tH = 10 days. The data points of these curves are the amplitudes extrapolated to FID time =0 of the signal of each component, obtained at each tIR by best fit to eq 1 (see list of the meanings of acronyms in the Results section).

relaxation data of the three components, LL, LS and S, at 10 days of hydration. The extrapolations to tIR = 0 of the curves give the values of the signals of the three components SS(tH), SLL(tH) and SLS(tH) at each tH. As observed in previous studies,32−34 and also shown by Figure S1 (Supporting Information), the total signal, the sum of these three extrapolated amplitudes, is nearly constant with tH. Both S and LS are growing at the expense of LL, which decreases substantially. It is clear from Figure 1 that all the curves are non-single-exponential. In particular, the curve for S shows a very fast initial decay that suggests that this signal is due to two components. All this is better revealed by the inversion of the data to quasi-continuous T1 relaxation time distributions. Such T1 distributions, obtained at each tH, show the evolution of the different components during hydration. Figure 2a−c shows the inversion of the experimental relaxation curves at tH = 6 h, tH = 25 h and tH = 10 days, showing the T1 distributions for LL, LS and S. Figure 2c is the inversion for the data of Figure 1. Over hydration time, the shift of LL to shorter relaxation times is clear, as is the gradual formation of two solid components (SS and SL) and of the liquid component with shorter T2 (LS). At 10 days (Figure 2c) most of the signal for LL is at T1 ≈ 5 ms, but a partially resolved peak at about a

Figure 2. T1 distributions for the three components, LL, LS, and total solid (S = SS + SL), at three hydration times: (a) tH = 6 h, (b) tH = 25 h, (c) tH = 10 days. A partially resolved peak at about a millisecond allows one to distinguish two LL features: the main peak (LLP) and a hump (LLH). The areas under the distributions are proportional to the signal intensities. The inset figures show the full LLP peaks.

millisecond allows one to distinguish two LL features, the main peak (LLP) and a hump (LLH). The solid component S has two fully resolved peaks. The peak near the maximum data time is solid-long (SL), and it will be shown later that this can be resolved into two separate FID components. The peak at a few milliseconds is solid-short (SS). The LS peak is broad, and, though not sharply defined, its position is constant during hydration at about the same position as for SS. 10873

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877

Langmuir

Article

To improve the text readability, a list of the acronyms used is reported: S

quasi-Gaussian solid FID component, called solid (with FID transverse relaxation time on the order of 10−15 μs) SL long T1 peak of the T1 distribution of the solid component, called solid-long SS short T1 peak of the T1 distribution of the solid component, called solid-short LL longer-T2* liquid (exponential) FID component, called liquid-long LS shorter-T2* liquid (exponential) FID component, called liquid-short LLH hump of the T1 distribution of the LL component, called liquid-long-hump LLP main peak of the T1 distribution of the LL component, called liquid-long-peak SXX NMR signal of the component XX, where XX stands for S or SL or SS or LL or LS or LLH or LLP 3.2. Time Evolution of the Ratio between Signal Intensities and a Surface Area Parameter. The amplitude ratios of LS/LL and SS/LL have been observed34 to go linearly with the relaxation rate of the peak of the T1 distribution of LLP, as shown in Figure 3a, where data for LLH/LL are also added. The linear behavior suggests a surface effect for the LLP component. Because the surface effect is given to the liquid component by an interaction with a surface area, it can be useful to consider at each tH the value of the LLP surface area parameter σ = SLLP/T1LLP, the signal of LLP divided by the T1 of its peak (T1LLP). If T1LLP is due to surface effect, σ = SLLP ρ ALLP/VLLP, where ρ is the surface relaxivity and ALLP is the total surface area of the pore volume VLLP filled by LLP. The ratio SLLP/VLLP is a constant, because SLLP is proportional to the amount of water filling VLLP, so σ is proportional to the surface of the pores confining LLP. The ratios SLS/σ, SSS/σ and SLLH/σ are plotted in Figure 3b as functions of tH. As clearly shown, SSS/σ and SLS/σ are nearly constant and go closely parallel during hydration, about a factor of 2 apart. After tH ≈ 6−8 h, also SLLH/σ is constant within experimental scatter. All this suggests that the amounts of 1H in both the SS and LS components, and of LLH after 6−8 h, grow proportionally to the surface of the pore space filled by LLP. 3.3. FID Analysis of 1H Solid Signals. Equation 1 is an empirical model, useful to obtain robust parameters to fit the data of FIDs with amplitudes and shapes changing not only with tH, but also with tIR. In the case of a system with rigid pairs of dipolar coupled spin 1/2 nuclei, a more physically founded model is obtained by considering that the FID signal G(t) is the inverse transform of the Pake doublet, given by refs 36, 41, and 42:

Figure 3. Parameters computed from SS, LS, LL, LLH and R1pk = 1/T1LLP. (a) Amplitudes of the signal ratios LS/LL, SS/LL and LLH/LL plotted against R1pk. (b) Ratios of the amplitudes of the signals of the components LLH, SS, and LS to the surface area parameter σ = SLLP/T1LLP, plotted against hydration time tH. The LLP surface area parameter σ is given by the signal of LLP divided by the T1 of its peak (T1LLP) (see section 3.2).

where r is the distance between the two interacting nuclei, γ is the gyromagnetic ratio, μ0 is the magnetic permeability of free space, and ℏ the Planck constant divided by 2π. Equation 3 allows one to compute the distance r between the pairs of interacting 1H nuclei. Equation 2 has been applied to the data of the same cement sample after 2.5 years of hydration. The Fresnel integrals were approximated by the functions given in ref 41. Surprisingly, it was found that the fit to the SL data by eq 2 was significantly improved by allowing two SL components instead of only one. In Figure 4, the data corresponding to tIR = 17 ms are presented along with the fitting functions. The values of the distances between pairs of interacting 1H nuclei obtained by eq 3 for the two SL components are (1.9 ± 0.2) Å and (1.6 ± 0.1) Å, with weights 85% and 15%, respectively. Substantially identical results were obtained from data corresponding to tIR = 43 ms. Figure 5 shows the part of the FID representing SS. The decay has the shape of a pure Gaussian, as shown by the inset plot against (FID Time)2, revealing that for SS there is a distribution of the distances between pairs of interacting nuclei and/or of shorter correlation times, consistent with the longer Tg of about 20 μs.

⎛ 6 ⎞ 2 2 ⎡ cos(αt ) G(t ) = αA 8π e−1/2β t ⎢ C⎜ αt ⎟ αt ⎝ π ⎠ ⎣ +

sin(αt ) ⎛ 6 ⎞⎤ S⎜ αt ⎟ ⎥ αt ⎝ π ⎠ ⎦

(2)

where C(x) and S(x) are the Fresnel integrals, β is a parameter that accounts for dipolar interactions between non-nearestneighbor 1H nuclei, and

α=

3 μ0 γ 2ℏ 4 4π r 3

(3) 10874

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877

Langmuir

Article

WPC endodontic cement MTA (ProRoot MTA, Mineral Trioxide Aggregate, Dentsply Tulsa Products, Tulsa, OK) with no CaCl2 added. The results are reported in the Figure S2 of Supporting Information. Also in this sample, which does not contain Friedel salt, the part of the FID representing SL was significantly improved by allowing two SL components instead of only one. Again, the values of the distances between pairs of interacting 1H nuclei obtained by eq 3 for the two fitting functions are (1.9 ± 0.2) Å and (1.6 ± 0.1) Å, with weights 84% and 16%, respectively. Besides the solid components assigned to Portlandite and to crystal water, a third solid component, the solid with the shorter T1 (SS), is clearly identified. SS also has very stable T1 and T2. Likely, this component, which appears to have 1H more mobile than in SL, as revealed by the pure Gaussian behavior and somewhat longer Tg (Figure 5), is ascribable to Ca−OH and/or anisotropically oriented water molecules inside the layers of C−S−H gel. On the other hand, it is known that OH groups in these layers may be associated with Ca ions.4,47 The common T1 shared by SS and LS (Figure 2), could be due to an exchange of 1H nuclei between these two close components. In the Jennings model, different classes of water-filled spaces are considered: interlayer spaces, intraglobular pores (IGP), small gel pores (SGP), and large gel pores (LGP).4 All our results can be interpreted in the light of this model, giving a contribution to clarifying the nano and sub-nano structure and state of water in cement. Our LLP and LLH are due to liquid water in compartments of different sizes, with relaxation times determined by surface effects. The component LLH is water in small and LLP in large gel pores, likely the compartments SGP and LGP, of the order of fractions of a nanometer and nanometers. Also LS is liquid water, as revealed by its exponential decay, and has very stable T1 and T2 during hydration. All this strongly suggests that the LS signal is due to the interlayer water in the C−S−H gel, described in ref 4, corresponding to the “confined water in the planar C−S−H gel pores” detected also in ref 35 and described as interlayer water in ref 9. The wide T1 distributions, as shown in Figure 2 for LS, could reflect the presence inside the interlayers of some irregular spaces filled with water molecules (IGP, ref 4). The schematic diagram in Figure 6, drawn in the style of Figure 1 of ref 2, should help in understanding the correspondence between the colloidal Jennings model components and our compartments and explain why the simplest interpretation of the striking behavior of the data in Figure 3a,b is just the growth of the gel in comparable-sized clumps. If the clumps contain substantially the same relative amounts of the compartments LS and SS, the signals of LS, SS and the total surface area (ALLP) in contact with the major part of the liquid water (LLP) at each hydration time tH are proportional to the number of clumps at that tH, in such a way that the ratios represented in Figure 3b are constant with tH. The results in Figure 3b cannot be easily explained by models proposing quasi continuous sheets of C−S−H layers.

Figure 4. Part of the FID representing the component SL, obtained at the average tIR= 17 ms after 2.5 years of hydration. The solid lines are eq 2 computed with two solid components (A) and with only one (B). The inset shows the portion of the FID where the two models differ most, with the corresponding values of the reduced χ2.

Figure 5. Part of the FID representing the component SS at tH = 2.5 years. The solid line is obtained with the assumption of a pure Gaussian form. In the inset, the same data are plotted as a function of (FID Time)2 in order to show the pure Gaussian nature of the decay.

4. DISCUSSION Our data allowed us to identify three solid components of WPC by relaxometry. It is clear that the solid signal component with the longer T1 (SL) is the signal from rigid pairs of 1 H nuclei. Our data reveal that SL is due to two classes of molecules whose 1H nuclei have very low chance to interact with liquid water, acquiring the typical relaxation times of “solids” (T2 of the order of tens of μs and T1 of the order of seconds). The computed distance (1.9 ± 0.2) Å for the first component of the signal is not far from the known value for Portlandite, reported as (2.202 ± 0.002) Å by neutron diffraction at 20 °C in ref 43 and as 2.186 Å in ref 44, obtained in both cases on pure bulk crystals of Ca(OH)2. A discrepancy between this distance in Portlandite inside cement and in bulk Portlandite crystals has been observed also in ref 29. For the second component, r = (1.6 ± 0.1) Å, consistent with the value of 1.54 Å reported for water molecules,45 is very likely due to crystal water in Ettringite and/or in Friedel salt (3CaO·Al2O3· CaCl2·10H2O) that can form in the presence of CaCl2, depending on the percentage of tricalcium aluminate.46 In order to test the possibility that Friedel salt could contribute, the same FID analysis was performed for a sample of another

5. CONCLUSIONS The analysis performed during the hydration of WPC allowed us to follow the evolution of different 1H pool signals, separated by different FID decay times and shapes and different quasicontinuous T1 relaxation time distributions. The NMR characteristics allowed us to assign the pools to different components of the hydration products, namely the C−S−H gel and the Portlandite and Ettringite crystals. For these last two products, 10875

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877

Langmuir

Article

Figure 6. Schematic diagram of the nanoscale C−S−H particles, drawn in the style of Figure 1 in ref 2. The C−S−H develops in clumps of approximately the same size, each clump containing about the same relative amounts of the compartments LS and SS. Between clumps there is the LLP water. Over time another compartment develops, representing LLH, liquid water in small pores formed between closest clumps.

■

realistic values for the 1H−1H distances are obtained by best-fit of FIDs to the equation derived by the Pake-doublet theory for the interaction of low-mobility spin 1/2 nuclei. A third lowmobility component is likely due to Ca−OH groups in C−S−H, where water molecules constitute a liquid-like component with which the OH groups are likely interacting. The other 1H pools are water molecules inside different size pores of the gel. The emerging structure of the gel is in perfect agreement with the classes of liquid water and pores described in the model of Jennings.4 However, our results go beyond, giving even stronger support to this model. If during hydration the C−S−H gel develops in comparable-sized clumps of the same composition, the external surface areas of the clumps confining water in pores among clumps should be proportional to the amount of material in the gel giving NMR signal, namely, to interlayer water (LS) and the low-mobility component with short T2 and short T1 (SS). This can explain the striking behavior of the parameters shown in Figure 3a,b. In particular, in Figure 3b the amount of 1H in SS, 2 times that in LS, suggests that for each water molecule in LS, four 1H nuclei in SS are formed. Also, the amount of water in the small gel pores (LLH) appears to grow in parallel with the number of clumps after 6−8 h of hydration, which is about the so-called “acceleration” period of cement hydration. This kind of analysis can be used to monitor the evolution of cement pastes during hydration and to compare their kinetics after addition of additives, all in a nondestructive way not requiring either radio-protection or invasive procedures like drying of the samples.

■

AUTHOR INFORMATION

Corresponding Author

*Mailing address: Department of Physics and Astronomy, University of Bologna, Viale Berti Pichat 6/2, 40127, Bologna, Italy; tel:+390512095119; e-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Allen, A. J.; Thomas, J. J. Analysis of C-S-H gel and cement paste by small-angle neutron scattering. Cem. Concr. Res. 2007, 37, 319−324. (2) Allen, A. J.; Thomas, J. J.; Jennings, H.M. Composition and density of nanoscale-calcium-silicate-hydrate in cement. Nat. Mater. 2007, 6, 311−316. (3) Thomas, J. J.; FitzGerald, S. A.; Neumann, D. A.; Livingston, R. A. State of water in hydrating tricalcium silicate and Portland cement pastes as measured by Quasi-Elastic Neutron Scattering. J. Am. Ceram. Soc. 200, 84, 1811−1816. (4) Jennings, H. J. Refinements to colloid model of C-S-H in cement: CM-II. Cem. Concr. Res. 2008, 38, 275−289. (5) Pellenq, R. J.-M.; Kushima, A.; Shahsavari, R.; Van Vliet, K. J.; Buehler, M. J.; Yip, S.; Ulm, F.-J. A realistic molecular model of cement hydrates. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 16102−16107. (6) Ji, Q.; Pellenq, R. J.-M.; Van Vliet, K. J. Comparison of computational water models for simulation of calcium-silicate-hydrate. Comput. Mater. Sci. 2012, 53, 234−240. (7) Richardson, I.G. The calcium silicate hydrate. Cem. Concr. Res. 2008, 38, 137−158. (8) Thomas, J. J.; Biernacki, J. J.; Bullard, J. W.; Bishnoi, S.; Dolado, J. S.; Scherer, G. W.; Luttge, A. Modeling and simulation of cement hydration kinetics and microstructure development. Cem. Concr. Res. 2011, 41, 1257−1278. (9) Muller, A. C. A.; Scrivener, K. L.; Gajewicz, A. M.; McDonald, P. J. Densification of C-S-H measured by 1H NMR relaxometry. J. Phys. Chem. C 2013, 117, 403−412.

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information contains additional information about cement composition and NMR data analysis. This material is available free of charge via the Internet at http://pubs.acs.org. 10876

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877

Langmuir

Article

of lower and higher mobility 1H. J. Phys. Chem. B 2010, 114, 1767− 1774. (33) Bortolotti, V.; Fantazzini, P.; Mongiorgi, R.; Sauro, S.; Zanna, S. Hydration kinetics of cements by Time-Domain Nuclear Magnetic Resonance: application to Portland-cement-derived endodontic pastes. Cem. Concr. Res. 2012, 42, 577−582. (34) Bortolotti, V.; Brown, R. J. S.; Fantazzini, P.; Mariani, M. Evolution of a short-T2 liquid-like 1H signal during the hydration of White Portland Cement. Microporous Mesoporous Mater. 2013, 178, 108−112. (35) Rodin, V.; Valori, A.; McDonald, P. J. A 1H double-quantumfiltered NMR study of water in cement pastes. New J. Phys. 2011, 13, 035017−1-14. (36) Pake, G. E. Nuclear resonance absorption in hydrate crystals: Fine structure of the proton line. J. Chem. Phys. 1948, 16, 327−336. (37) Fantazzini, P.; Galassi, F.; Bortolotti, V.; Brown, R. J. S.; Vittur, F. The search for negative amplitude components in quasi-continuous distributions of relaxation times: The example of magnetization exchange in articular cartilage and hydrated collagen. New J. Phys. 2011, 13, 065007−1-15. (38) Borgia, G. C.; Brown, R. J. S.; Fantazzini, P. Examples of marginal resolution of NMR relaxation peaks using UPEN and diagnostics. Magn. Reson. Imaging 2001, 19, 473−475. (39) Fantazzini, P.; Brown, R. J. S. Units in distributions of relaxation times. Concepts Magn. Reson. 2005, 27 A, 122−123. (40) Bortolotti, V.; Brown, R. J. S.; Fantazzini, P. UpenWin: A software to invert multi-exponential decay data, http://www.unibo.it/ PortaleEn/Research/Services+for+companies/UpenWin.htm. (41) Look, D. C.; Lowe, I. J.; Northby, J. A. Nuclear magnetic resonance study of molecular motions in solid hydrogen sulfide. J. Chem. Phys. 1966, 44, 3441−3452. (42) Hansen, E. W.; Kristiansen, P. E.; Pedersen, B. Crystallinity of polyethylene derived from solid-state proton NMR Free Induction. J. Phys. Chem. B 1998, 102, 5444−5450. (43) Busing, W. R.; Levy, H. A. Neutron diffraction study of calcium hydroxide. J. Chem. Phys. 1957, 26, 563−568. (44) van Velzen, G. D. F.; Wenckebach, W. Th. Nuclear magnetic ordering in Ca(OH)2. I. Molecular field approximation. J. Phys. (Paris) 1990, 51, 1445−1461. (45) Cook, R. L.; De Lucia, F. C.; Helminger, P. Molecular force field and structure of water: Recent microwave results. J. Mol. Spectrosc. 1974, 53, 62−76. (46) Suryavanshi, A. K.; Scantlebury, J. D.; Lyon, S. B. Mechanism of Friedel’s salt formation in cements rich in tri-calcium aluminate. Cem. Concr. Res. 1996, 26, 717−727. (47) Nonat, A. The structure and stoichiometry of C-S-H. Cem. Concr. Res. 2004, 34, 1521−1528.

(10) Chiang, W.-S.; Fratini, E.; Baglioni, P.; Liu, D.; Chen, S.-H. Microstructure determination of C-S-H globules by Small-Angle Neutron Scattering. J. Phys. Chem. C 2012, 116, 5055−5061. (11) Ridi, F.; Fratini, E.; Baglioni, P. Cement: A two thousand year old nano-colloid. J. Colloid Interface Sci. 2011, 357, 255−264. (12) Constantinides, G.; Ulm, F.-J. The granular nature of C-S-H. J. Mech. Phys. Solids 2007, 55, 64−90. (13) Bonnaud, P. A.; Ji, Q.; Coasne, B.; Pellenq, R. J.-M.; Van Vliet, K. J. Thermodynamics of water confined in porous Calcium-SilicateHydrates. Langmuir 2012, 28, 11422−11432. (14) Korb, J.-P. NMR and nuclear spin relaxation of cement and concrete materials. Curr. Opin. Colloid Interface Sci. 2009, 14, 192−202. (15) Cowan, B. Nuclear Magnetic Resonance and Relaxation; Cambridge University Press: Oxford, U.K., 1997. (16) Brown, R. J. S. Measurements of nuclear spin relaxation of fluids in bulk for large surface-to-volume ratios. Bull. Am. Phys. Soc., Ser. II 1956, 1, 216. (17) Brown, R. J. S. Nuclear magnetism logging at the Coyote institute. Magn. Reson. Imaging 1996, 14, 811−817. (18) Brownstein, K. R.; Tarr, C. E. Importance of classical diffusion in NMR studies of water in biological cells. Phys. Rev. A 1979, 19, 2446−2453. (19) D’Orazio, F.; Tarczon, J. C.; Halperin, W. P.; Eguchi, K.; Mizusaki, T. Application of nuclear magnetic resonance pore structure analysis to porous silica glass. J. Appl. Phys. 1989, 65, 742−751. (20) Borgia, G. C.; Brown, R. J. S.; Fantazzini, P. Nuclear magnetic resonance relaxivity and surface-to-volume ratio in porous media with a wide distribution of pore sizes. J. Appl. Phys. 1996, 79, 3656−3664. (21) Greener, J.; Peemoeller, H.; Choi, C.; Holly, R.; Reardon, E. J.; Hansson, C. M.; Pintar, M. M. Monitoring of hydration of white cement paste with proton NMR spin−spin relaxation. J. Am. Ceram. Soc. 2000, 83, 623−627. (22) Leventis, A.; Verganelakis, D. A.; Halse, M. R.; Webber, J. B.; Strange, J. H. Capillary imbibition and pore characterization in cement pastes. Transp. Porous Media 2000, 39, 143−157. (23) McDonald, P. J.; Korb, J.-P.; Mitchell, J.; Monteilhet, L. Surface relaxation and chemical exchange in hydrating cement pastes: A twodimensional NMR relaxation study. Phys. Rev. E 2005, 72, 011409−9. (24) Plassais, A.; Pomiès, M.-P.; Lequeux, N.; Korb, J.-P.; Petit, D.; Barberon, F.; Bresson, B. Microstructure evolution of hydrated cement pastes. Phys. Rev. E 2005, 72, 041401−8. (25) Monteilhet, L.; Korb, J.-P.; Mitchell, J.; McDonald, P. J. Observation of exchange of micropore water in cement pastes by twodimensional T2−T2 nuclear magnetic resonance relaxometry. Phys. Rev. E 2006, 74, 061404−9. (26) Korb, J.-P. Microstructure and texture of cementitious porous materials. Magn. Reson. Imaging 2007, 25, 466−469. (27) McDonald, P. J.; Mitchell, J.; Mulheron, M.; Monteilhet, L.; Korb, J.-P. Two-dimensional correlation relaxation studies of cement pastes. Magn. Reson. Imaging 2007, 25, 470−473. (28) Valori, A.; Rodin, V.; McDonald, P.J. On the interpretation of 1 H 2-dimensional NMR relaxation exchange spectra in cements: Is there exchange between pores with two characteristic sizes or Fe+ concentrations? Cem. Concr. Res. 2010, 40, 1375−1377. (29) Dolinšek, J.; Apih, T.; Lahajnar, G.; Blinc, R.; Papavassiliou, G.; Pintar, M.M. Two-dimensional NMR study of a hydrated porous medium: An application to white cement. J. Appl. Phys. 1998, 83, 3535−3540. (30) McDonald, P. J.; Rodin, V.; Valori, A. Characterisation of intraand inter-C-S-H gel pore water in white cement based on an analysis of NMR signal amplitudes as a function of water content. Cem. Concr. Res. 2010, 40, 1656−1663. (31) Barberon, F.; Korb, J.-P.; Petit, D.; Morin, V.; Bermejo, E. Probing the surface area of a cement-based material by Nuclear Magnetic Resonance dispersion. Phys. Rev. Lett. 2003, 90, 116103−1-4. (32) Gombia, M.; Bortolotti, V.; De Carlo, B.; Mongiorgi, R.; Zanna, S.; Fantazzini, P. Nanopore structure buildup during endodontic cement hydration studied by time-domain nuclear magnetic resonance 10877

dx.doi.org/10.1021/la501677k | Langmuir 2014, 30, 10871−10877