Application of 1H NMR to hydration and porosity studies of lime–pozzolan mixtures

Microporous and Mesoporous Materials 139 (2011) 16–24

Contents lists available at ScienceDirect

Microporous and Mesoporous Materials journal homepage: www.elsevier.com/locate/micromeso

Application of 1H NMR to hydration and porosity studies of lime–pozzolan mixtures M. Tziotziou a,c, E. Karakosta a, I. Karatasios a, G. Diamantopoulos a, A. Sapalidis b, M. Fardis a, P. Maravelaki-Kalaitzaki c, G. Papavassiliou a, V. Kilikoglou a,⇑ a

Institute of Materials Science, National Centre for Scientific Research ‘‘Demokritos’’, Aghia Paraskevi, 15310 Athens, Greece Institute of Physical Chemistry, National Centre for Scientific Research ‘‘Demokritos’’, Aghia Paraskevi, 15310 Athens, Greece c Analytical and Environmental Chemistry Lab, Technical University of Crete, 73100 Chania, Greece b

a r t i c l e

i n f o

Article history: Received 18 July 2010 Received in revised form 9 October 2010 Accepted 10 October 2010 Available online 16 October 2010 Keywords: NMR relaxometry Mercury porosimetry Nitrogen adsorption Porosity Mortar hydration

a b s t r a c t 1 H nuclear magnetic resonance (NMR) relaxation is applied to lime–pozzolan mixtures to monitor in realtime the hydration and porosity evolution during setting. The hydrated products formed during the setting of these mixtures are similar to those formed in hydraulic binders (e.g. cement and natural hydraulic limes). In this work, we demonstrate that by using a portable, low field (0.29 T) Halbach magnet it is possible to study in detail, the evolution of the pore structure, modified through the formation of C–S–H and C–A–H, by means of the 1H NMR technique. Contrary to the standard porosimetry methods, which require drying procedures before analysis, NMR is a non-invasive straightforward technique, allowing the study of the hydration kinetics in real-time, during the setting and hardening of the lime–pozzolan binding system. Thus, by measuring the 1H NMR spin–lattice relaxation T1 of two lime–pozzolan mixtures, it was possible to distinguish between different pore populations within the system (at different setting periods), and to study the growth of the hydrated phases. The interpretation of the results proved that, similarly to mercury intrusion porosimetry (MIP) and nitrogen adsorption, 1H NMR probes the development of the pore structure within the lime–pozzolanmortar matrix, indicating key changes of pore size populations. Ó 2010 Elsevier Inc. All rights reserved.

1. Introduction The simultaneous monitoring of hydration and porosity evolution in freshly synthesized mortars is a key issue for assessing their performance and adjusting their compositional and curing parameters. In the case of conservation mortars for archaeological and architectural heritage, the performance requirements are variable and specific for each monument. At the same time, due to the nature of these structures, there is a need for continuous but noninvasive monitoring of the mortar hardening process, through the recording of parameters that control it, such as hydration and porosity. During the setting and hardening processes of lime–pozzolan mortars, which are strongly affected by the presence of water [1], various physical and chemical phenomena occur, which modify their microstructure. The setting reactions that take place during their hardening result in the formation of a variety of calcium sili⇑ Corresponding author. Tel.: +30 210 6503317; fax: +30 210 6519430. E-mail addresses: [email protected] (M. Tziotziou), elkarako@ims. demokritos.gr (E. Karakosta), [email protected] (I. Karatasios), gior15@ims. demokritos.gr (G. Diamantopoulos), [email protected] (A. Sapalidis), [email protected] (M. Fardis), [email protected] (P. MaravelakiKalaitzaki), [email protected] (G. Papavassiliou), [email protected]. gr (V. Kilikoglou). 1387-1811/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2010.10.010

cate (C–S–H) and calcium aluminate hydrated (C–A–H) phases [2], which grow within the pore space of the mixtures and modify their initial pore space characteristics (open porosity, pore-size distribution and specific surface area). It is, therefore, clear that hydration reactions and pore space characteristics are directly interrelated and are responsible for the gradual strengthening of the mortar mixtures [3,4]. Several methods are currently available for determining the pore structure of porous materials [5,6]. Most of them are invasive, as they involve sampling and an hydration – stop process, which affects the amount and the type of the hydrates present in the paste [7] and consequently the pore – space properties of the mixtures. In contrast, nuclear magnetic resonance (NMR) is a fast, potentially non-invasive, technique, for the characterization of the internal structure of a porous material, based on its mobile water molecule content [8]. Over the years, NMR has managed to successfully provide valuable information about porosity, pore-size distributions and the hydration kinetics, especially in cement based materials [9–14]. This technique has the advantage of the nuclear-spin selectivity, where only one nuclear – spin isotope is detected at a time, providing information in real-time, even the earliest few minutes. The successful application of 1H NMR relaxometry to the study of the microstructure development in cement pastes provides the

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

necessary background for the application of the technique to similar materials. In this context, the present work deals with the application of 1H NMR relaxometry, to continuously follow and interpret the hydration and porosity evolution in a lime – natural pozzolan system. The method was applied to two mixtures, consisting of two different amounts of pozzolan powder with the same kneading water. The selection of the specific type of mortar was based on the fact that lime–pozzolan mortars have been widely used, since antiquity, as a multi-purpose building material, exhibiting remarkable performance over time [15,16]. In addition, traditional building materials and techniques are seeing in revival lately aimed at reducing energy consumption, CO2 and particulate matter emissions [17,18] for the construction of buildings and the restoration of architectural monuments, leading to renewed interest in the use of lime–pozzolan binders [19,20]. In the work presented here, 1H NMR T1 distribution profiles, attributed to mobile water molecules in lime – pozzolan pastes, as a function of hydration time, are obtained and modeled to describe the pore distribution. The results were compared and correlated to pore-size distributions obtained by conventional MIP and nitrogen adsorption. The T1 relaxation measurements were carried out in a portable lightweight permanent Halbach magnet, with a low static magnetic field Bo = 0.29 T. Data analysis was performed by using an inverse Laplace transform, which leads to T1 distribution profiles, clearly demonstrating the development of porosity and the formation of two distinct pore populations. In conjunction with 1H NMR, MIP and nitrogen adsorption, microstructural parameters were studied by scanning electron microscopy (SEM). 1.1. NMR relaxation in porous materials Lime – pozzolan mixtures are characterized as hydraulic binders hardening through a hydration mechanism. Compared to cements, they have a considerably slower hydration rate and develop a higher porosity matrix [21,22]. The pore-size distributions in cement systems have been successfully determined using NMR spin–lattice relaxation measurements [23–27], where the relationship between T1 and the pore size is based on the fast-exchange relaxation theory [28]. In principle, the T1 of a fluid (typically water) confined within a pore, can be used to determine the pore diameter (actually, the ratio of pore volume to surface area). The presence of the pore wall increases the relaxation rate and thus, for a given material, T1 will decrease as the pore size decreases [24,29]. In particular, it is well known that the relaxation rate, 1/T1, of mobile water molecules is enhanced near a liquid–solid interface. This is due to the exchange between the free and bonded water and the presence of paramagnetic sites on the solid surface [30]. In the fast-exchange limit the overall spin–lattice rate within a given pore is expressed as:

1 1g g ¼ þ T 1 T 1ðbulkÞ T 1ðsurf Þ

ð1Þ

where 1/T1(surf) is the surface relaxation rate of the water adsorbed at the pore surface, 1/T1(bulk) is the free water relaxation rate and g is the fraction of adsorbed water molecules at the pore surface [31,32]. As, 1/T1(bulk) is negligible compared to the rate of relaxation at the pore surface 1/T1(surf), the overall spin–lattice (1/T1) rate within a spherical pore filled with water, with volume V and surface S is expressed as [31,32] 1=T 1 ¼ ð1=T 1ðsurf Þ ÞðeS=VÞ, where e is the thickness of the water layer at the pore surface and S/V is the pore surface to volume ratio. Thus, the overall relaxation rate depends linearly on the S/V ratio of the pores. As a consequence, in complex porous materials

17

such as lime–pozzolan systems, with pore sizes extending from nanometers to micrometers, T1 is spread over a wide distribution of relaxation times, which directly reflect the complex pore microstructure and the interaction strength between the adsorbed water molecules and the pore surface [32]. Specifically, during mortar hydration where a pore network has been developed, the nuclear magnetization in a T1 experiment, using a saturation recovery pulse sequence, cannot be described by a single exponential and is expressed as [32]:

RðtÞ ¼

M 0  MðtÞ ¼ M0

Z

1

gðT 1 Þ expðt=T 1 ÞdT 1

ð2Þ

0

where R(t) is the proton magnetization recovery function, M0 is the magnitude of the magnetization at equilibrium and M(t) is the observed magnetization as a function of time t. Here, g(T1) is the spin– lattice relaxation time T1 distribution function, which according to Eq. (1) is related to the pore-size distribution function [23]. On the basis of Eq. (2), g(T1) can be resolved by means of an inverse Laplace transform, [32] revealing important information about the porous microstructure in the hardened material. 2. Experimental 2.1. Sample preparation Two different lime–pozzolan pastes were prepared to study the evolution of the pore structure during setting. The system consisted of hydrated lime powder (Ca(OH)2), (Merck, Germany) and a natural pozzolan powder of volcanic origin. The natural pozzolan used was a commercial product available in Greece, namely Ifestiaki Gaia (IG), a finely ground product with particle size <20 lm supplied by Dalkafoukis, Greece. The mineral phases present in the hydrated lime powder are mainly portlandite and traces of calcite, whereas the mineral phases present in the pozzolan powder consisted of quartz, illite/muscovite, kaolinite and Na-feldspars (albite, anorthoclase). The iron content (Fe2O3) of the pozzolan was 2.6% (w/w). The pastes were prepared by slightly modifying the EN 196-1 [33] standard for cement paste production. In particular, de-ionized water and lime were mixed thoroughly using a planetary mixer. Once water was added to the lime powder, the resultant paste was mixed for 90 s at 140 rpm and another 90 s at 285 rpm. Then, natural pozzolan was added to the hydrated paste and mixed for 90 s at 140 rpm. The mixing process was interrupted for 90 s and then mixed again for 90 s at 285 rpm. The lime to pozzolan ratios were 50:50 and 30:70 (w/w) in the mixtures CaIG5 and CaIG7, respectively. In both mixtures, the water added was 90% (w/w) of the solid constituents. After mixing, the samples were sealed into NMR glass tubes using ParafilmÒ membrane to avoid moisture loss and then, immediately placed into the spectrometer. The sample dimensions were 9 mm in diameter and 30 mm in height. The water content of each mixture satisfies the requirement for optimum flow characteristics [34], while the mixtures present a flow value of 180 mm. A second batch of the above mixtures was moulded in prismatic moulds, with the dimensions 20  20  80 mm, and then placed in a curing chamber for setting at RH = 95 ± 3% and T = 25 ± 2 °C. These specimens were used for monitoring the evolution of the microstructure and phase formation during setting by SEM and MIP. 2.2. NMR relaxometry 1

H NMR spin–lattice T1 relaxation experiments were conducted using a home-built circular Halbach array magnet,

18

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

suitable for low-field NMR measurements [31]. The field at the magnet centre was 0.29 T, corresponding to a proton resonance frequency of 12.1718 MHz. The Halbach magnet was coupled to a broadband spectrometer operating in the frequency range 5–800 MHz. A magnetic field gradient equal to G = 1.03 T/m is present at the centre of the magnet. The value of the gradient was determined by the cubic exponential dependence of the spin echo decay of a reference sample (distilled water) by applying

Mercury porosimetry measurements are based on the gradual intrusion of mercury to the pore system of the mixture, during the application of an external pressure. The pore radius is then calculated as function of pressure using the Washburn Eq. (3). The open pore volume is expressed in diagrams representing the pore-size distribution (dV/dlog(r)).

the well-known expression expðð2=3Þc2 DG2 s3 Þ [8,31], where c is the proton gyromagnetic ratio and D is the self-diffusion coefficient. Using this relationship and the known value of water selfdiffusion coefficient at room temperature (D = 2.3  109 m2/s), the linear part of the gradient was determined. The probe head had a dead time of 20 ls at the Larmor frequency employed. The spin–lattice relaxation time, T1 of both samples wasmeasured by employing a saturation recovery sequence [(p/2)  t  (p/2)  s  (p)] using 30 data points for the interpulse delay, t, ranging from 100 ls to 6 s in a logarithmic scale. The signal was detected by the common Hahn echo pulse sequence with a s value of 60 ls. All experiments were performed at room temperature and the hydration process for each sample was monitored for up to 8 months with time intervals between successive experiments ranging from minutes and several hours in the initial hydration stage up to full days at the later hydration stages. Initially, the spin–lattice relaxation mechanism was characterized by a single exponential function. However, with progressive hydration a multi-exponential behavior developed, which was resolved by means of an inverse Laplace transform (Eq. (2)). Although the proposed mathematic procedures for the problem of separating exponentials lead to unique solutions for the relaxation times, enormous practical problems arise when they are applied to experimental data from physical experiments. The principal reason is the exceedingly non-orthogonal behavior of the exponential functions. Therefore, many ‘‘numerically equivalent’’ solutions can be obtained, which all give satisfactory fit of the experimental data points [32]. In these experiments, the numerical Laplace inversion of the 1H NMR saturation recovery curves was obtained by using a modified CONTIN algorithm, introduced by Provencher [35], adapted on a Matlab application. The CONTIN algorithm has been used successfully to solve many ill-posed problems [36]. CONTIN usually produces 10–20 physically permissible solutions and selects the most appropriate one based on statistical evaluations. In general, NMR studies in adsorbed water systems have shown that the CONTIN method is able to produce broad distributions (as those encountered in cement systems), which agree well with the true distributions of relaxation times [37]. In this study, CONTIN was constrained to a positive output for 30 logarithmically distributed points between the T1min = 0.1 ms and T1max = 1000 ms.

ð3Þ

r¼

2cm cos h P

where r is the pore radius, cm is the surface tension between mercury and the pore wall (N/m), h is the contact angle between the mercury and the pore wall (degrees) and P is the pressure applied on mercury to intrude into the pore (N/m2). In addition, representative samples of both mixtures cured for 180 days were measured using the nitrogen adsorption technique, aiming to determine the pores in the size range from 0.35 to about 500 nm. The specific surface area (SSA) was also determined in these samples by the Brunauer–Emmet–Teller (BET) method [39], using am (N2) = 0.162 nm2, where am is the molecular area of nitrogen at 196 °C, i.e. the area occupied by a nitrogen molecule in a complete mono-layer. Adsorption isotherms of N2 at 196 °C were performed in an automated volumetric system (AUTOSORB 1 – Krypton version, by Quantachrome). The samples were crashed to small pieces prior to testing and were outgassed under high vacuum (<1.3  106 Pa) for 72 h at 70 °C. Although outgas of the specimens under these conditions may slightly affect the water in the hydrated phases and consequently the pore – space properties of the mixtures, this process is necessary for ensuring the adsorption of nitrogen. The exact effect of the drying process on the measurable nitrogen surface area has been studied in detail in other papers [7,40]. The nitrogen desorption isotherms were further analyzed using the Dubinin–Astakov (DA) and the Barrett–Joyner–Halenda (BJH) methods [41], in order to obtain complementary data on the mesopore and micropore populations. The BJH method is used to describe the pore distribution of mesopores in the range of 1.7– 300 nm size in diameter [42] and is based on a model of the adsorbent as a collection of cylindrical pores. The theory accounts for capillary condensation in the pores using the classical Kelvin equation, which in turn assumes a hemispherical liquid–vapor meniscus and a well-defined surface tension. According to the BJH theory, the pore diameter (Dp) is equal to the sum of the Kelvin diameter (Dk) and the thickness of the adsorbed nitrogen film on the pore walls:

Dp ¼ Dk þ 2  t

ð4Þ

Accordingly, to describe the pores in the micropore range the Dubinin–Astakhov equation [43] was used:

2.3. Pore space characteristics Open porosity (po), pore-size distribution and total surface area (A) were used for the description of the pore space characteristics of mortar specimens during setting [38]. Measurements were carried out on specimens cured for 3, 7, 28, and 180 days, by means of MIP. Interruption of the setting process of the mixtures at the above time intervals was affected by immersion in two stop-bath solutions (acetone and diethyl-ether) for 45 min each, followed by drying at 70 °C for 24 h. Porosity measurements were recorded using a Quantachrome Autoscan 60 porosimeter, in the 2–4000 nm range. Before measurements, each sample was placed in a special cell and was vacuum pumped to 7.99 Pa. The vacuum was applied for approximately 1 h to ensure that all the air and residual moisture were removed from the pores.

n    RT ln P=P 0 W ¼ W 0 exp  E

ð5Þ

where W is the weight of nitrogen adsorbed at P/P0 and T, W0 is the total weight adsorbed, E is the characteristic energy and, n is the non-integer value (typically between 1 and 3). The DA equation is based on a Weibull distribution of pore sizes and is a generalized form of the Dubinin–Radushkevich equation (n = 2), which has been found to better fit the adsorption data for heterogeneous micropores [44,45]. Finally, the microstructure of the mixtures (after hydration stop) was examined in polished sections and fractured surfaces under a FEI/Quanta Inspect D8334 scanning electron microscope, coupled with an energy dispersive X-ray analyzer (SEM/EDX).

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

19

3. Results and discussion 3.1. Porosity analysis The evolution of the porosity of lime–pozzolan mortars was studied by MIP, at different hydration periods. These periods (3, 7, 28 and 180 days) were selected according to the evolution of hydrated phases and the resultant microstructural changes determined by SEM. At the early stages of setting (e.g. 3–7 days), where the hydration process is not advanced, the pore spaces observed are attributed to the free space formed between lime and pozzolan particles. These pores lie in the lm scale and are filled with water. Here, the water added to the mixtures corresponds to the amount required for optimum flow/plasticity and workability characteristics. After molding of the mixtures, part of this water participates in the hydration process and serves as a solvent medium for the dissolution of the lime binder to Ca2+ and OH ions, the reaction of Ca2+ with pozzolan particles and later, the formation of calcium silicon and aluminum hydrates (C–S–H, C–A–H) [46,47]. At this early stage, pore-size distribution plots of the mixtures present a single sharp population distribution of macropores in the region of 300–400 nm (Fig. 1). This peak corresponds to a pore fraction that is very little affected by the formation of new setting products and therefore dominates up to the 7th day of setting period. The slight differences observed between the two mixtures, as regards the exact position and the area of the respective peaks, are attributed to the different lime to pozzolan ratio used in the mixtures (1 and 0.43 for mixtures CaIG5 and CaIG7, respectively). After the first week of setting, the newly hydrated phases partially fill the empty pore space (Fig. 2) and thus, after 28 days of setting the network of pores creates a smooth distribution of smaller pores, between 10 and 200 nm (Fig. 1). This distribution is modified according to the amount of the pozzolan used. In the CaIG7 mixture a clear peak is formed at 50 nm, along with a smooth shoulder at 20 nm. In contrast, CaIG5 presents a single peak at 100 nm and a large tail towards smaller pore sizes. Finally, after 180 days of setting, the majority of C–S–H and C– A–H hydrates have been formed, filling the initial macropores and creating a new pore space structure, significantly different from the previous one (Fig. 3). This structure is mainly characterized by the presence of mesopores and micropores, as it can be seen in the distribution of Fig. 1. At this age, the CaIG5 mixture presents

Fig. 2. Calcium silicon and aluminum hydrates after 28 days of setting partially fill and modify the initial pore space of the lime – pozzolan mixture (CaIG7, SEM photomicrograph).

Fig. 3. Calcium silicon and aluminum hydrates after 180 days of setting fill the initial inter-grain pore space of CaIG7 and create distributions at smaller pore radii (small capillaries).

Fig. 1. Differential distribution curves of lime–pozzolan mixtures at different time periods.

a pore distribution maximum at 20–30 nm (medium capillaries) and a tail down to 3 nm. The distribution plot of the CaIG7 mixture shows two additional pore populations that are distinguished at 3 and 10 nm. To further study the meso- and micro-pores’ structure of the lime – pozzolan mixtures, a set of nitrogen adsorption measurements were conducted, in mixtures cured for 180 days. The adsorption isotherms (Fig. 4) are classified as type IV [48], being characteristic of porous materials containing both mesopores and macropores. The two mixtures present similar isotherms, with a hysteresis loop of type H3 [41]. This type of hysteresis is attributed to the plate – like pozzolan particles and to the shape of the pores within the hydrated setting products. The inception point of the hysteresis loop in the two mixtures (Fig. 4) indicates capillary condensation beginning at a lower vapor pressure (P/Po) and the presence of more pores with smaller radii in CaIG7 mixture.

20

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

Fig. 4. Nitrogen adsorption isotherms at 196 °C for mixtures CaIG5 and CaIG7. The inset presents the end of the part of the isotherm (0.05 < P/Po < 0.1), which is indicative of the completion of a continuous monomolecular layer of gas on the pores’ surface.

The nitrogen desorption isotherms were further analyzed using the Dubinin–Astakov (DA) and Barrett–Joyner–Halenda (BJH) methods [41] to obtain complementary data on the amount of mesopore and micropore populations of the mixtures. The estimated values (pGas) are in Table 1, along with specific surface area (SSA) values. 3.2. 1H NMR T1 relaxation experiments The evolution of T1 as a function of hydration time for the two lime–pozzolan mixtures is presented in Fig. 5. More specifically, the data are presented as T1 contour plots and T1 distribution profiles, obtained using an Inverse Laplace Transform. As it can be seen, during the early hours of hydration the relaxation time T1 is nearly constant, having similar values for both samples. During this stage, all the protons’ moments relax with a common T1, due to the fast exchange between water spins in the various environments [10]. At the same time, the lime (Ca(OH)2) particles are dissolved in pore-water and create a saturated calcium hydroxide solution [49]. As hydration proceeds a gradual decrease of T1 is observed. The pozzolan particles are corroded, resulting in the leaching of sodium and potassium ions into the pore solution. Consequently, Ca2+ ions take their place on the pozzolan surface and react with the free OH-, Si–O- and Al–O-radicals [50]. This dissolution-precipitation mechanism in-

side the lime–pozzolan matrix leads to the formation of calcium–aluminum and calcium–silicon hydrates (C–A–H, C–S–H) and the gradual modification of the porosity and specific surface area (Table 1). It should be noted here that the small volume of micropores (pGas) measured by nitrogen adsorption minimizes the potential negative effect of the outgassing process on the surface area values (SSA) obtained, as mentioned in Section 2.3. The progress of hydration is mainly controlled by the amount of Ca(OH)2 available and its solubility, which depend on particle size and crystal shape [51], as well as on the particle size and mineralogy of the pozzolan. The increase of the surface area within the sample is responsible for the reduction of T1, (Fig. 5), as seen from Eq. (1), which illustrates the hydration evolution and the modification of the pore structure. Through the above mechanism, the initial hydration products are formed, which create an amorphous matrix around the pozzolan grains (Figs. 2 and 3). From the decrease of the T1 values, it is observed that the hydration evolution is a very slow process, compared to hydrated cement pastes, due to the low rate of pozzolanic reaction described above. In general, the reactivity of pozzolans is slower than that of Portland cement with a consequent effect on the setting time. During the initial hydration stages the NMR data are characterized by a single relaxation time. However, after a certain hydration period the T1 relaxation mechanism deviates from a single exponential function and a multi-exponential behavior is observed. This crossover

Table 1 Pore space properties of mixtures determined by mercury intrusion and nitrogen adsorption porosimetries. Method

Nitrogen adsorption

Sample

pGas (%)

SSA (m2/g)

pHg (%)

Mercury intrusion SSA (m2/g)

Centre of gravity of pore volume distribution (nm)

CaIG5_3d CaIG5_7d CaIG5_28d CaIG5_180d CaIG7_3d CaIG7_7d CaIG7_28d CaIG7_180d

nd nd nd 4.7 nd nd nd 5.3

nd nd nd 51.43 nd nd nd 64.71

93.5 76.8 70.1 68.0 91.8 73.7 72.0 60.9

10.0 18.7 46.6 96.8 9.1 15.8 31.8 124.0

161.7 91.6 28.4 15.5 161.7 97.3 34.9 12.6

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

21

Fig. 5. T1 contour plots and T1 distribution profiles vs. hydration time for CaIG5 (a and b) and CaIG7 (c and d).

is accompanied by a massive growth of the hydrated phases (Fig. 3) forming a dense network of amorphous hydrated products resulting in the reduction of open porosity (Table 1). The further reduction of T1 is readily explained by the growing surface area of the hydration products (Table 1). At this point the T1 distribution splits in two components, which remain almost invariant with hydration time. The point where the two pore-reservoirs, i.e. small and large pores, become distinguishable by NMR is an indication that the small pores have become isolated from the large pores. At this point the total porosity for both samples is reduced to approximately 70% (as observed from Table 1). This porosity level appears to be a limit, below which, a second pore population emerges on both samples. For the CaIG5 sample this occurs at 24 days and for CaIG7 at 7 days, as is seen from the T1 distribution profiles in Fig. 5. Therefore, two different pore-size distributions peaks are observed; one with short T1  2.0 ms attributed to small pore sizes and a second with longer T1  10 ms corresponding to larger pores. For the CaIG7 mixture, the second pore size population appears much earlier, due to the higher amount of pozzolan powder present, which increases the reaction and formation rate of the C–A–H and C–S–H. This effect in the CaIG7 sample is also clearly demonstrated by MIP (Fig. 1) where, besides the main peak, a second pore population, with pore radius around 20 nm, is developed at 7 days of hydration. The fast-exchange relaxation model allows one to convert the NMR relaxation time to a mean pore size. Thus, according to Eq. (1), the observed relaxation time of the pore water is expected to be given by:

1 1 ¼ T 1 T 1ðsurf Þ

    eS 1 3e ¼ V T 1ðsurf Þ r

ð6Þ

where e the thickness of the layer of water molecules, T1(surf) is the relaxation time of the water molecules in that layer and r is the pore radius. The 1/T1(surf) relaxation rate is primarily induced by the fluctuating local fields due to the paramagnetic impurities within the grains of the samples. For cement pastes, T1(surf) lies within the range 0.1–1 ms [28]. The thickness of the water surface layer (assuming a mono-layer of molecules), is given by e = 0.35 nm [52]. In Eq. (6), the T1(surf) relaxation term is generally an unknown parameter. Therefore, we directly compare the distribution profiles obtained by the two different techniques by correlating the centers of gravities for MIP (Table 1) and NMR distributions (Fig. 6) and consequently substituting the r value from MIP in Eq. (6). From the comparison of the T1 values of the NMR data and the pore radius of the MIP experiments and by using Eq. (6) it was found that the value of T1(surf) varied as a function of hydration time for the two samples studied. In particular, the average value of T1(surf) between the 7th and the 180th days of hydration was found in the range 0.1–0.3 ms, suggesting that the distribution of paramagnetic impurities on the surface varies with hydration time. The values of T1(surf) were within the range quoted in the literature for hydrated cement pastes [28]. Fig. 6 shows the comparison of the two methods for the selected hydration times where the two distributions appear to overlap at late curing periods after which, dual pore distributions appear. However, the latter does not apply in early periods where large capillaries (>50 nm) are not directly detectable with NMR T1 measurements

22

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

Fig. 6. Comparison of NMR and MIP distribution profiles at selected hydration times for (a) CaIG5 and (b) CaIG7.

(Fig. 6). A similar conclusion was deduced by Monteilhet et al. [53] using NMR T2  T2 two-dimensional correlation spectroscopy. In the mixtures studied, the T1 values of large and medium size capillaries are attributed to the fast water exchange between these two pore size populations, rendering them unresolved. It must be mentioned that NMR surveys the pore microstructure by water molecules within the interconnected and isolated pores whereas MIP surveys the pore throat distribution, providing an advantage to small pore radii. Although comparison between the two methods is quite complicated, Fig. 6 indicates that the pore sizes obtained by NMR lies within the same range of those obtained by MIP. Finally, when comparing the distribution profiles determined by NMR with the normalized differential pore volume distributions (cm3/g nm) obtained by MIP and nitrogen adsorption in mixtures cured for six months, where the amount of micropores can be safely determined (Table 1, pGas values), the relative contribution of each group of pore sizes can be evaluated (Fig. 7). The nitrogen desorption isotherms are analyzed using the Dubinin– Astakov (DA) and the Barrett–Joyner–Halenda (BJH) methods, to obtain complementary data on the amount of mesopore and

micropore populations of the mixtures (Table 1). Contrasting the porosity values of micropores (pGas) and mesopores (pHg) it is observed that the amount of micropores is very small in relation to that of the mesopores, as determined by MIP, and therefore has no actual impact on the total effective porosity of the mixtures. Consequently, the microstructure of both lime–pozzolan mixtures is described almost exclusively by the formation of mesopores, which can be safely monitored by NMR and MIP methods. It should be noted here, however, that MIP and nitrogen adsorption results are referring to open or interconnected pores, while NMR can additionally detect water molecules trapped in isolated pores and therefore, it increases the sensitivity of the measurement. Thus, the two distinguished pore-size distributions in the NMR plots indicate the presence of two populations within the mesopore range [54], which according to Mindess [55] can be assigned to small capillaries, with pore radius between 2.5 and 10 nm, and to medium capillaries, with pore radius between 10 and 50 nm. It is also apparent that the room temperature NMR relaxometry cannot measure the micropore population in porous systems [11]. Furthermore, the extremely low amount of

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

23

Fig. 7. The pore-size distributions (cm3/g nm) of the mixtures CaIG5 (a) and CaIG7 (b) at 180 days, as determined by NMR, MIP and nitrogen adsorption (BJH and DA plots).

micropores (pGas) of the lime–pozzolan mixtures (4.7–5.3%) (Table 1) produces a very weak NMR signal, which is hard to detect in comparison with that of the mesopores. The present results provide strong evidence that, similar to MIP, 1 H NMR effectively probes the development of the pore structure within the lime–pozzolan matrix, indicating key changes of pore size populations. In the systems studied, mesopores have the most important role in the performance of lime–pozzolan mixtures, since they form the main pore volume of the specimens and are involved in all processes related to strength development, moisture diffusion and transportation of water solutions. Consequently, the NMR method is capable of following the evolution of different pore sizes within the mesopore range formed in lime–pozzolan mixtures as a function of hydration time. Moreover, the 1H NMR method has the advantage that it is a fast, non-invasive technique, capable of monitoring the hydration process in real time. The latter may be very valuable in field applications where a portable NMR configuration can be used [8,56].

4. Conclusions The hydration process of the lime – natural pozzolan mixtures was successfully monitored by measuring the proton spin–lattice relaxation time T1 in a similar manner as in hydrated cement pastes. For the systems studied, two distinct spin groups, characterized by different spin–lattice relaxation times T1, were resolved. These correspond to water in small and medium capillary pores. For the sample containing a higher amount of pozzolan powder, the split of T1 appeared at a much earlier hydration time due to the acceleration of the hydration reactions. The distributions of pore populations determined by MIP and nitrogen adsorption proved to be in a good agreement with those obtained by NMR relaxometry. In conclusion, we have shown that the T1 relaxation technique at low external magnetic field is a sensitive tool to probe the capillary pores in lime–pozzolan mixtures and follow the time evolution of their pore structure in a consistent way.

24

M. Tziotziou et al. / Microporous and Mesoporous Materials 139 (2011) 16–24

Consequently, the possibility for continuous monitoring of the setting process and the evolution of the microstructure of these mixtures by NMR relaxometry provides a promising tool for scholars involved in conservation of architectural heritage to accurately evaluate in the field the durability and service life of lime– pozzolan mortars. Acknowledgment This work has been partially funded by the European Program EUROSTARS, 4291 (2009). References [1] N. Nestle, A. Kühn, K. Friedemann, C. Horch, F. Stallmach, G. Herth, Microporous Mesoporous Mater. 125 (2009) 51–57. [2] H.F.W. Taylor, Cement Chemistry, second ed., Academic Press Limited, London, 1990. [3] N.S. Martys, C.F. Ferraris, Cem. Concr. Res. 27 (1997) 747–760. [4] M. Arandigoyen, Appl. Surf. Sci. 252 (2005) 1449–1459. [5] J. Kaufmann, R. Loser, A. Leemann, J. Colloid Interface Sci. 336 (2009) 730–737. [6] M. Krus, K.K. Hansen, H.M. Kiinzel, Mater. Struct. 30 (1997) 394–398. [7] L.D. Mitchell, J.C. Margeson, J. Therm. Anal. Calorim. 86 (2006) 591–594. [8] B. Blumich, J. Perlo, F. Casanova, Prog. Nucl. Magn. Reson. Spectrosc. 52 (2008) 197–269. [9] J.-P. Korb, Curr. Opin. Colloid Interface Sci. 14 (2009) 192–202. [10] N. Nestle, Cem. Concr. Res. 34 (2004) 447–454. [11] R.M.E. Valckenborg, L. Pel, K. Hazrati, K. Kopinga, J. Marchand, Mater. Struct. 34 (2001) 599–604. [12] K. Friedemann, F. Stallmach, J. Karger, Cem. Concr. Res. 36 (2006) 817–826. [13] J. Greener, H. Peemoeller, C. Choi, R. Holly, E.J. Reardon, C.M. Hansson, M.M. Pintar, J. Am. Ceram. Soc. 83 (2000) 623–627. [14] P.J. Prado, B.J. Balcom, S.D. Beyea, T.W. Bremner, R.L. Armstrong, R. Pishe, P.E. Gratten-Bellew, J. Phys. D: Appl. Phys. 31 (1998) 2040–2050. [15] K. Callebaut, J. Elsen, K. Van Balen, W. Viaene, Cem. Concr. Res. 31 (2001) 397– 403. [16] C. Sabbioni, G. Zappia, C. Riontino, M.T. Blanco-Varela, J. Aguilera, F. Puertas, K. Van Balen, E.E. Toumbakari, Atmos. Environ. 35 (2001) 539–548. [17] K. Gäbel, A.J. Tillman, Cleaner Prod. 13 (2005) 1246–1257. [18] R. Rehan, M. Nehdi, Environ. Sci. Policy 8 (2005) 105–114. [19] R. Cerny, A. Kunca, V. Tydlitat, J. Drchalova, P. Rovnanıkova, Constr. Build. Mater. 20 (2006) 849–857. [20] A. Velosa, P. Cachim, in: L. Braganca, M. Pinheiro, S. Jalali, R. Mateus, R. Amoeda, M. Correia Guedes (Eds.), Portugal SB07. Sustainable Construction, Materials and Practices. Challenge of the Industry for the New Millenium, Ios Press, Amsterdam, 2007, pp. 856–860. [21] J. Lanas, Cem. Concr. Res. 34 (2004) 2191–2201. [22] M. Arandigoyen, J.I. Alvarez, Cem. Concr. Res. 37 (2007) 767–775. [23] R. Blinc, G. Lahajnar, S. Zumer, M.M. Pintar, Phys. Rev. B 38 (1988) 2873–2875. [24] L.J. Schreiner, J.C. Mactavish, L. Miljkovic, M.M. Pintar, R. Blinc, G. Lahajnar, D. Lasic, L. Reeves, J. Am. Ceram. Soc. 68 (1985) 10–16. [25] A. Plassais, M.-P. Pomies, N. Lequeux, P. Boch, J.-P. Korb, D. Petit, F. Barberon, Magn. Reson. Imaging 21 (2003) 369–371.

[26] G. Papavassiliou, M. Fardis, E. Laganas, A. Leventis, A. Hassanien, F. Milia, A. Papageorgiou, E. Chaniotakis, J. Appl. Phys. 82 (1997) 449–452. [27] M. Fardis, G. Papavassiliou, L. Abulnasr, L. Miljkovic, R.J. Rumm, F. Milia, E. Chaniotakis, D. Frangoulis, Advn. Cem. Bas. Mat. 1 (1994) 243–247. [28] R. Blinc, J. Dolinsek, G. Lahajnar, A. Sepe, I. Zupancic, S. Zumer, F. Milia, M.M Pintar, Z. Naturforsch 43a (1988) 1026–1038. [29] R. Holly, E.J. Reardon, C.M. Hansson, H. Peemoeller, J. Am. Ceram. Soc. 90 (2007) 570–578. [30] G. Papavassiliou, F. Milia, M. Fardis, R. Rumm, E. Laganas, J. Am. Ceram. Soc. 76 (1993) 2109–2111. [31] E. Karakosta, G. Diamantopoulos, M.S. Katsiotis, M. Fardis, G. Papavassiliou, P. Pipilikaki, M. Protopapas, D. Panagiotaras, Ind. Eng. Chem. Res. 49 (2010) 613– 622. [32] E. Laganas, G. Papavassiliou, M. Fardis, A. Leventis, F. Milia, E. Chaniotakis, C. Meletiou, J. Appl. Phys. 77 (1995) 3343–3348. [33] European Committee for Standardization (CEN), Methods of Testing Cement: Part 1. Determination of Strength, European Standard EN 196-1. [34] European Committee for Standardization (CEN), Methods of Test for Mortar for Masonry – Part 3. Determination of Consistence of Fresh Mortar (by Flow Table). European Standard EN 1015-3, 1999. [35] S.W. Provencher, Comput. Phys. Comm. 27 (1982) 213–227. [36] R.M. Kroeker, R.M. Henkelman, J. Magn. Reson. 69 (1986) 218–235. [37] K. Overloop, L. Van Gerven, J. Magn. Reson. 100 (1992) 303–315. [38] J.L. Pérez Bernal, M.A. Bello López, Surf. Sci. 161 (2000) 47–53. [39] S. Brunauer, P.H. Emmet, E. Teller, J. Am. Chem. Soc. 60 (1938) 309–319. [40] M.C.G. Juenger, H.M. Jennings, Cem Concr Res 31 (2001) 883–892. [41] S.J. Gregg, K.S.W. Sing, Adsorption, Surface Area and Porosity, second ed., Academic Press, London, 1982. [42] E.P. Barrett, L.G. Joyner, P.P. Halenda, J. Am. Chem. Soc. 73 (1951) 373–380. [43] M.M. Dubinin, Chem. Rev. 60 (1960) 235–241. [44] H.F. Stoeckli, Carbon 28 (1990) 1–6. [45] A. Gil, P. Grange, Colloid Surf. A 113 (1996) 39–50. [46] S. Song, H.M. Jennings, Cem. Concr. Res. 29 (1999) 159–170. [47] R. Snellings, G. Mertens, S. Herstens, J. Elsen, Microporous Messoporous Mater. 126 (2009) 40–49. [48] K.S.W. Sing, D.H. Everett, R.A.W. Haul, L. Moscou, R.A. Pierotti, J. Rouquirol, J.T. Siemieniewska, Pure Appl. Chem. 57 (1985) 603–619. [49] P. De Silva, L. Bucea, D.R. Moorehead, V. Sirivivatnanon, Cem. Concr. Compos. 28 (2006) 613–620. [50] F. Massazza, in: P.C. Hewlett (Ed.), Lea’s Chemistry of Cement and Concrete, Elsevier Butterworth – Heinemann, Oxford, 1998, pp. 471–602. [51] O. Cazalla, C. Rodriguez-Navarro, E. Sebastian, G. Cultrone, J. Am. Ceram. Soc. 83 (2000) 1070–1076. [52] A.J. Bohris, U. Goerke, P.J. McDonald, M. Mulheron, B. Newling, B. Le Page, Magn. Reson. Imaging 16 (1998) 455–461. [53] L. Monteilhet, J.-P. Korb, J. Mitchell, P.J. McDonald, Phys. Rev. E 74, (2006) 061404-1–061404-9. [54] L.K. Koopal, Manual of Symbols and Terminology for Physicochemical Quantities and Units – Appendix II Definitions, Terminology and Symbols in Colloid and Surface Chemistry, Part I, International Union of Pure and Applied Chemistry, Division of Physical Chemistry Washington, DC, USA, 2001. Available from: ), last accessed 05.10. [55] S. Mindess, J.F. Young, D. Darwin, Concrete, second ed., Prentice Hall, Englewood Cliffs, New Jersey, 2002. [56] S. Anferova, V. Anferov, J. Arnold, E. Talnishnikh, M.A. Voda, K. Kupferschlager, P. Blumler, C. Clauser, B. Blumich, Magn. Reson. Imaging 25 (2007) 474–480.