Fractal Structure Evolution during Cement Hydration by Differential Scanning Calorimetry: Effect of Organic Additives

Fractal Structure Evolution during Cement Hydration by DifferentialScanning Calorimetry: Effect of Organic AdditivesFrancesca Ridi, Emiliano Fratini, and Piero Baglioni*

Department of Chemistry “Ugo Schiff” and CSGI, University of Florence, via della Lastruccia 3-Sesto Fiorentino, I-50019 Florence,Italy

*S Supporting Information

ABSTRACT: Low-temperature differential scanning calorimetry (LT-DSC) is used toinvestigate the microstructure of tricalcium silicate pastes, hydrating in pure water and inthe presence of comb-shaped polycarboxylate ether superplasticizers. LT-DSC is shown tobe a powerful technique, able to provide important information on the porosity and on thefractality of the porous evolving matrices by means of rapid and nondestructivemeasurements. In particular, LT-DSC gives a semiquantitative estimation of the evolvingporosity (capillary, small gel, and large gel pores), the depercolation threshold of thecapillary pores, and the fractal dimension associated with the probed porosity. The resultsare in good agreement with those obtained by small-angle scattering methods ensuring thatthis approach, based on the well-established and easily accessible DSC technique, canprovide valuable information on the evolving porosity and the fractal nature of hydratingcement pastes.

■ INTRODUCTION

Cement hydration is a complex process that starts from apowder of heterogeneous composition and water and forms arigid solid network with high mechanical resistance. The mostimportant hydrated phase, acting as “glue” between the grains(unreacted phases, inert fillers, crystalline hydrated products), isthe amorphous calcium silicate hydrate (briefly named C−S−H), formed upon reaction of tricalcium silicate (C3S) and, to alesser extent, of dicalcium silicate (C2S) with water. A numberof macroscopic properties (e.g., elasticity, compressive strength,resistance to degradation, transport phenomena) depend onthe microstructural features of C−S−H, and their completecontrol would require the full understanding of both thehydrates microstructure and packing.1 For this reason, sincemany years, an important part of the literature is devoted to theinvestigation of the C−S−H microstructure2−15 and to theintrinsic relationship between the evolving microstructure andultimate macroscopic properties.16−19 Many efforts in this fieldare oriented on the difficult task of providing a comprehensivepicture of this complex material, whose constituent featureshave dimensions ranging over several orders of magnitude,from nanometers (C−S−H unit globules, interlamellar water-filled spaces) to tens of micrometers (largest capillary pores).The increasing use of organic polymers in the industrialpractice, with the aim of modifying the hydration kinetics andtuning the final properties of the material, compels to accountfor their possible effect on the forming C−S−H. In particular,in a recent work it was pointed out that the addition of lastgeneration superplasticizers to tricalcium silicate pastesmodifies both the structure of the C−S−H basic globulesand the overall microstructural arrangement.20

It is generally accepted that a realistic picture of disorderedporous systems could be provided by a fractal description. Inparticular, in the case of cement pastes, the hydration consistsin a “gelation” process, where the developing C−S−H phaselinks together the unreacted grains. This kind of process isefficiently described by the percolation theory.21,22 From aphysical point of view, the term percolation refers to theformation of a long-range connectivity in stochastic systems,and its occurrence produces random fractal structures. Theapplication of the percolation theory to an evolving poroussystem allows the measurement of the percolation threshold andof the f ractal dimension, both directly linked to the macroscopiccharacteristics of the material. For example, one of the mostrelevant industrial challenges in the cement chemistry field isthe reduction (or at least the control) of the degradationmechanisms due to the permeation of salt-rich water because ofatmospheric agents. This event is strictly related to transportphenomena taking place in the specimen. According to thepercolation model, the transport properties inside a paste canbe rationalized in terms of two percolation thresholds: (i) theset point and (ii) the capillary porosity (de)percolation.23−25 Theset point is the time when the solid grains become connectedtogether by means of hydration products and form a rigidnetwork. This point occurs at very low degree of hydration (α =0.02−0.08), roughly coinciding with the end of the inductionperiod. Even when the solid matrix is percolated, the hydrationreaction inside the sample keeps on changing the micro-structure of the sample. The volume of the hydrated phases

Received: June 25, 2013Revised: November 12, 2013Published: November 13, 2013

Article

pubs.acs.org/JPCC

© 2013 American Chemical Society 25478 dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−25487

increases until the pores (initially interconnected) depercolate.This time establishes the capillary porosity depercolationthreshold. After this point, the capillary porosity becomes“closed” toward the external surface and is progressivelyreduced by the developing hydrate phases. This implies thatthe “in situ” determination of the depercolation threshold couldbe important to understand and control the transportproperties that in most degradation mechanisms govern therate of damage of the specimen and hence the long-termdurability of the material.23,26−28

Cement pastes have been extensively studied by means ofscattering techniques10,20,29−33 because the length scale of thestructure (1−1000 nm) matches the dimension of the X-ray orneutron probe (i.e., the inverse of the scattering vector). Theeasiest information that can be achieved by the log−log plot ofa scattering curve is the mass f ractal dimension. Moreover, thesetechniques do not require drying procedures and can bedirectly performed on wet samples. Hence, SAS techniques areparticularly suitable to investigate the evolution of the fractalarrangement during the microstructure development.30,33

This paper aims at showing that calorimetry, a conventionaland easily accessible technique, can effectively estimate thevolume of the meso/macroporosity, the depercolation thresh-old of the capillary porosity, and the fractal dimension ofhydrated cementitious samples, with results well in agreementwith those obtained by other techniques like small-angleneutron or X-ray scattering, which are more sophisticated butless accessible. Furthermore, even though SANS/SAXSmethods have to be considered by far the most rigorous wayto attain the mass fractal dimension, the possibility to obtainthis information by means of an easily accessible techniquebased on calorimetry is very appealing, also considering that itis possible to simultaneously couple the fractality evaluationwith information like the porosity evolution and hydrationkinetics. In this regard, low-temperature differential scanningcalorimetry (LT-DSC) enables to follow the evolution of themicrostructure throughout the hydration process and does notrequire special treatments of the samples. In particular, in thispaper we report the results of the investigation of saturatedtricalcium silicate (C3S) pastes hydrated in water and in thepresence of four structurally defined organic additives thatbelong to the last generation of superplasticizers (polycarbox-ylic backbone with grafted PEO chains) used in advancedcement formulations.

■ EXPERIMENTAL SECTIONSynthetic tricalcium silicate (C3S) with BET specific surfacearea of 0.65 ± 0.05 m2/g and particles median radius of 4.66μm was obtained from the CTG-Italcementi Group.PCEs were obtained from BASF. They are polymethacrylic

chains partially esterified with poly(ethylene oxide) lateralchains. The molecular formula is sketched in Figure 1. Eachpolymer is identified by the acronym PCEp-n, where p is thenumber of repeating PEO units in the lateral chains and nidentifies the number of free carboxylic groups per esterifiedgroup.The pastes were prepared by manually mixing 500 mg of C3S

with 200 mg of water or additive solutions (1% w/w). In thisway a water/cement ratio (w/c) of 0.4 and the 0.4% of PCEs byC3S mass was obtained. Each sample was maintained at 20 °Cin closed containers, to avoid water loss, until the end of theacceleration period, when the microstructure due to thegrowing hydrated phases has formed. This time (td, diffusional

time) was identified through the acquisition of the hydrationkinetics, as reported in a previous work.34 The td times of thepastes are 1 day for C3S/water, C3S/PCE102-2, and C3S/PCE102-6 pastes; 3 days for C3S/PCE23-2; and 12 days forC3S/PCE23-6. At these given times, some milligrams of eachpaste were put in aluminum pans and studied with LT-DSC.The remaining part of each paste was put under water andstored at 20 °C. After 3, 7, 14, and 28 days some milligramswere withdrawn, externally dried with paper, and analyzed. Asthe induction time of the C3S/PCE23-6 paste was very long,only for this sample the sampling times were 12, 21, and 28days. Differential scanning calorimetry measurements wereperformed using a DSC Q2000 (TA Instruments, Philadelphia,PA), and the obtained data were elaborated with the Q Seriessoftware, version 5.4.0. Each measurement was carried out withthe following temperature program: equilibrate to 5 °C; coolingramp from 5 to −80 °C at 0.5 °C/min; heating ramp from −80to +10 °C at 0.5 °C/min.Scanning electron microscopy was performed on uncoated

samples using a field emission scanning microscope ΣIGMA(Carl Zeiss Microscopy GmbH, Germany). The images wereacquired using the in-lens secondary electron detector with anacceleration potential of 5 kV.

■ RESULTS AND DISCUSSIONThe heating scans (from −80 °C to room temperature) of theDSC thermograms, reported in Figure 2A, show a single humpover the whole temperature range. As documented in theliterature,28,35−40 this particular shape is due to the peculiarcharacteristics of the cement microstructure: being thedistribution of the pore sizes continuous in a range spanningfrom nanometers to tens of micrometers, while heating, themelting of the water occurs progressively from the fractionsconfined in the smallest cavities to the fractions present in thelargest ones. This continuous melting and release of heatproduces the broad hump characteristic of the heating part ofthe thermogram. However, the freezing process occurs in adiscontinuous way, and the cooling scans present some definitepeaks, as shown in Figure 2B. This peculiar behavior has to beascribed to the combination of the two freezing mechanisms ofthe water: homogeneous and heterogeneous. The homoge-neous nucleation is an activated process because a free energybarrier must be surmounted to form a critical nucleus. Theheterogeneous nucleation occurs at preferential sites andrequires less energy than the homogeneous process. Accordingto a very recent paper by Sanz et al.,41 at low degree ofsupercooling (−20 °C < T < 0 °C), due to the high free energyrequired for the homogeneous formation of a critical cluster,only the heterogeneous nucleation is possible. For this reason,in a saturated system, the water confined in cavities whosedimension is large enough to host critical nuclei of at least

Figure 1. Sketch of the PCEs molecular formula.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725479

∼8000 molecules (diameter ≈ 8 nm, that is, the size of thestable critical cluster at T ≈ −15 °C)41 freezes viaheterogeneous nucleation, even if it is isolated from thesurface. Once the water in the capillary pores is frozen, theformed ice remains in contact with the liquid water present inthe smaller cavities. By lowering the temperature below T <−20 °C, the size of the critical cluster sensibly decreases (at T≈ −35 °C, diameter ≈ 3.5 nm corresponding to ∼600molecules).41 Moreover, the nucleation free energy barrierdecreases, making the homogeneous process possible to occur.In these conditions both the homogeneous and theheterogeneous nucleation mechanisms become accessible,with comparable rates. The freezing will then occur in thepores whose dimensions can host the critical cluster stable atthat temperature. The appearance of peaks in the cooling curveindicates, however, that the process starts at the pore entrance,where the liquid water is in contact with the surrounding ice,which constitutes a preferential site for the nucleation. Thedimensional range of these freezing pores must be comparableto the size of their entrance because, for the reasons givenbefore, it is very unlikely that water in large pores freezes at thisstage and the smaller pores cannot host the critical nucleus. Inthe case of cement pastes, the process of formation of themicrostructure originates two classes of nanometric porosities,named small gel pores (SGP) and large gel pores (LGP),according to the Jennings’ colloidal model II. The former class(SGP) derives from the packing of individual ∼5 nm C−S−Hglobules: the nanometric dimension of these “building block”implies that both the pore size and the entrance size arenecessarily in the range of a few nanometers (namely, 1−3nm).11,42 This dimensional range is compatible with the peakaround −40 °C in the thermogram.43,44 The latter pore class(LGP) originates from the fractal aggregation of a “super-structure” of globules, resulting in a more “open” porosity, interms of both pore size (the literature indicates these cavities torange between 4 and 12 nm)11,39,42 and entrance size. In thecooling part of the thermogram, the freezing of the waterconfined in this pore class occurs in the −20/−35 °Ctemperature range.43,44

Analysis of Fractal Dimension. As the heating scandirectly depends on the distribution of the pore sizes, it can beanalyzed to obtain information on the fractality of the samples.Several studies in the literature evidence the fractal nature ofC−S−H and extract the mass fractal dimension, Dm, as afunction of the degree of saturation,31 the hydration time,30,33

and the degradation degree.45 As already mentioned, SAS

techniques are generally the methods of choice to investigatethese properties.Previous works report the possibility to extract these

quantities in the case of wet sonogels from the DSCincremental volume distribution, in very good agreement withSAXS patterns measured on the same system.46 The rationalebehind the extrapolation of fractal properties from DSC relieson the fact that the melting temperature of an ice crystalconfined in a pore is depressed of a quantity ΔT = Tm

0 − Tmrelated to the radius R, being R = r − l, where r is the radius ofthe pore and l is the thickness of the nonfreezable layer of waterat the solid interface. Previous estimations of l from NMRmeasurements report a value of 0.5 ± 0.1 nm for porousglasses.47 For this reason, no peak is observed in the DSCcooling scan for the water confined into the calcium silicatelayers constituting the C−S−H primary units (“interlamellargel pores” or IGP), the size of these pores being about 1 nm.The Gibbs−Thomson equation states that the meltingtemperature Tm and the radius R are inversely related asfollows:

γ=

−Δ

⎜ ⎟⎛⎝

⎞⎠T T

VHR

1 2m m

0 s

(1)

where Tm0 is the melting temperature of an ice crystal of infinite

dimension, γ is the solid−liquid interfacial tension, ΔH is thespecific melting heat, and Vs is the specific volume of the solid.For the water case (assuming Tm

0 = 273.15 K, γ = 40 × 10−3 Nm−1, ΔH = 334 J g−1, and Vs = 1.02 cm3 g−1), eq 1 becomessimply

Δ =TR

68.29(2)

where R is given in nanometers.Heating from −80 °C to room temperature a porous sample

with pore size distribution P(r) and saturated with waterproduces the melting of the liquid confined in pores ofprogressively increasing dimension. In other words, theregistered heat flux is proportional to the incremental volumedV of the ice that melts at each temperature Tm. To provide aquantitative estimation, the detected heat flow must beindependent of the heating rate. In fact, in a DSC experimentthe heat flow depends on the mass of the sample, on its thermalconductivity, on the thermal contact and on the heating/cooling rate. According to Neffati and Rault,48 a DSCthermogram registered at rates lower than 2 °C/min maintains

Figure 2. Heating part (A) and cooling part (B) of a typical LT-DSC thermogram recorded on cementitious saturated samples (C3S/PCE102-2).

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725480

the equilibrium conditions, yielding real information on theaccessible porosity.From the definition given above, the incremental pore

volume per solid mass can be written as dV = P(r) dr, whereP(r) is the pore size distribution. To obtain dV, the heatingDSC signal has been normalized with the total pore volume Vp

(see the table reported in the Supporting Information),obtained by integration of the whole melting peak, scaled bythe bulk water density value at 0 °C (0.9998 g/cm3).49 As anexample, Figures 3A and 3B respectively report the heating

DSC signals and the plots of the incremental pore volume vsthe melting temperature, obtained for the paste C3S/PCE102-2.Neffati and Rault48 assumed that the heat flow, Jq, measured

by DSC on porous glasses is related to ΔT by a scaling lawconsistent with the fractal nature of the systems. Furthermore,in some papers46,50 the mass fractal of wet gels was extractedfrom DSC data and compared with the results obtained fromSAXS. To compare the DSC and the SAXS approaches, Volletet al.51 used a method originally proposed to link SAXS andnitrogen adsorption data on porous systems. In their approach,the porous system was regarded as a homogeneous solid of

Figure 3. (A) DSC heating scans on the paste C3S/PCE102-2. (B) Incremental pore volume per solid mass vs temperature obtained from the scansshown in panel A. The curves are shifted for the sake of clarity.

Figure 4. Incremental pore volume per solid mass as a function of the melting depression, ΔT = Tm0 − Tm for the differently formulated paste

investigated in this study: (A) C3S/H2O, (B) C3S/PCE102-2, (C) C3S/PCE102-6, (D) C3S/PCE23-2, and (E) C3S/PCE23-6.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725481

density ρS where an incremental pore volume per solid unitmass dV = P(r) dr was used to account for the change of thebulk density of the porous sample, ρ(r), as a function of thepore filling steps. The resulting process can be mathematicallydescribed as

∫ρ ρ= +

rP r r

1( )

1( ) d

r

S 0 (3)

As a matter of fact, ρ(r) will then scale with r, in the fractalrange a ≤ r ≤ ξ, as

ρ ρ= −r r a( ) ( / )DS

3m (4)

where a is the characteristic dimension of the smallest repeatingunit and ξ is the maximum correlation length of the fractalaggregate. Combining eq 2 and eq 4, the following relationholds:

= Δ −V A Td ( )D 3m (5)

The plots dV vs ΔT on a log−log scale for the investigatedsamples are reported in Figure 4.Table 1 (as well as Figure SI2 in the Supporting Information)

reports the Dm coefficients extracted from the fitting of the

log−log plots in Figure 4 from ΔTξ ≅ 1 K to ΔTa ≅ 10 K,corresponding to pores with radius between ξ ≅ 70 nm and a ≅7 nm, as calculated by means of eq 2. This range is directlycomparable with that investigated by small-angle scatteringtechniques.30,52

In all the samples, immediately after the nucleation andgrowth period (that is after 1 day of curing for C3S/water, C3S/PCE102-2, and C3S/PCE102-6 systems; after 3 days for C3S/PCE23-2; and after 12 days for C3S/PCE23-6)

34 the fractaldimension is very low, as expected for poorly packed systems,where the three-dimensional microstructure is still beginning todevelop. As the process goes on, the microstructure of thepastes becomes increasingly packed as a consequence of thehydrated phases growth and Dm values rise. In the case of C3S/water paste, the fractal dimension increases from 2.0 to 2.6 inthe first 28 days. In a previous SAXS investigation, the massfractal dimension in a similar C3S/water paste was shown tovary from 1.9 to 2.8 in the very same time interval, from 1 to 28days.30 During the first month of hydration the pastescontaining superplasticizers exhibit Dm values lower thanC3S/water. This means that a more “open” nanoscale structureis formed in PCE-containing pastes with high w/c values (w/c= 0.4). The need of reducing the water content in realapplications involving superplasticizers is well-known, as theexcess water is recognized to increase the capillary porosity.The present investigation shows that high w/c values inducealso the formation of a nanostructure less packed than that ofthe C3S/water sample, especially in the first part of thehydration process. However, after 6 months Dm reaches values

around 2.4−2.6 even in the C3S/SPs cases. Similarly, a recentSANS/SAXS investigation20 on pure C−S−H phases, synthe-sized from C3S in excess of water or in the presence of a watersolution at the same concentration of the PCEs used in thisstudy, reported a Dm = 2.8 for the pure C−S−H phase andvalues around 2.6 in presence of PCEs. Dm extracted for verymature samples (6 months) by LT-DSC results in two- orthree-tenths lower than what is reported by the SANS/SAXSexperiment performed on the correspondent C−S−H purephase. This small discrepancy can be attributed to the differentnature of the probes and to the small differences in the samples.In the first case (i.e., different nature of the probe) while LT-DSC is sensitive to the water phase, SAXS detects both thescattering contributions of the hydrated phases and theanhydrous phases, which is not always obvious to decouple.In the second case (i.e., small differences in the samples), theDm values obtained on pure C−S−H cases are not affected bythe presence of Ca(OH)2, which is always contained in thesamples investigated by LT-DSC. Even a different w/c ratiocould lead to small differences in the mass fractal values. Allthese comparisons show that the LT-DSC technique is areliable method to access, within a certain extent, the evolutionof the mass fractal dimension in cement-based materials.

Evolution of Porosity. While the heating scan contains theinformation on the fractality of the samples, the cooling scanallows the quantification of the water confined in the porosityduring each stage of the cement hydration. Figure 5 shows thecooling part of the DSC thermograms registered on C3S/waterpaste and on C3S/PCEs pastes during the hydration.The thermogram registered on each sample at the end of the

acceleration period does not show the sharp peak between −10and −20 °C due to the crystallization of bulk water. This meansthat at this time the main part of the initial water has beenconsumed by the formation of hydrates. Part of this waterremains unreacted, constrained in the just formed micro-structure. Because of this confinement, this water freezesaround −40 °C.After the acceleration, the pastes have been dipped in water

to ensure the whole porosity to be saturated. This procedureallows extracting quantitative information on pore volume andon its evolution during the hydration. From this point on, theDSC thermograms show three main features: an intense sharppeak in the −10/−20 °C range, due to the crystallization ofbulk water contained in capillary pores; a peak in the −20/−35°C region, corresponding to the crystallization of water in LGPpores; a peak at −40 °C due to the solidification of the waterconstrained in the SGP porosity.According to a procedure reported elsewhere,42 each peak

was integrated, and the areas were used to calculate the amountof water involved. As the standard enthalpy of fusion varies withthe temperature, to quantify the water, we used the estimationof ΔH0 by Hansen et al.53 at the mean temperature within eachintegrated region. Figure 4 shows the results of this calculation.The histograms display the pore volume of capillary, LGP, andSGP pores as a function of the hydration time. To estimate thepore volume, we used the density of bulk liquid water at 25 °C,0.9998 g/cm3. This calculation does not take into account anypossible change of the water density due to the confinement.Further investigations could be planned to give a more preciseestimation of this assumption, which is not the purpose of thepresent work.49

In all the samples the capillary pore volume evolvesaccording to a depercolation process, showing a decreasing

Table 1. Mass Fractal Dimension Dm (±0.1) for theAnalyzed Pastes during the Hydration

3 days 7 days 14 days 21 days 28 days 6 min

water 2.0 2.3 2.5 2.6 2.5PCE102-2 2.0 2.2 2.1 2.1 2.5PCE102-6 2.1 2.2 2.2 2.4 2.7PCE23-2 1.9 2.4 2.2 2.3PCE23-6 2.6 2.5 2.5

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725482

Figure 5. Low-temperature DSC thermograms recorded on (A) C3S/water, (B) C3S/PCE102-2, (C) C3S/PCE102-6, (D) C3S/PCE23-2, and (E)C3S/PCE23-6 during hydration.

Figure 6. Histograms showing the evolution of the pore volume (cm3 per g of paste) during hydration: capillary porosity (red), LGP (yellow), andSGP (blue). Error bars indicate uncertainties of ±25% for capillary porosity and of ±40% for LGP and SGP, estimated analyzing different portions ofthe same sample.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725483

behavior during the hydration process due to the increase ofthe solid volume fraction, as the hydrated phases have molarvolume higher than the anhydrous ones. In the C3S/water case(Figure 6A, red bars) the volume is almost constant during thefirst 14 days and starts to decrease at 28 days, showing a drasticdrop only after 6 months. In this sample the depercolationthreshold can be then estimated to occur after 28 days from themixing. The addition of PCEs sensibly decreases the capillaryvolume with respect to the C3S/water sample and, in somecases, alters the percolation threshold. The sample with

PCE102-2 (having the lowest adsorption propensity, plasticiz-ing efficiency54 and retarding power34 among these polymers)shows a behavior very similar to the paste without additives, thecapillary pore space remaining percolated throughout the first28 days of hydration, and showing a reduction only after 6months. When C3S is hydrated in presence of PCE102-6(Figure 6C) the depercolation of capillary porosity isanticipated (with respect to C3S/water paste) at 7 days, whilePCE23-2 and PCE23-6 (Figures 6D and 6E) maintain thedepercolation threshold at 28 days after mixing, despite their

Figure 7. SEM images of fracture surfaces of pastes cured 28 days: (A, B) C3S/water; (C, D) C3S/PCE102-2; (E, F) C3S/PCE102-6; (G, H) C3S/PCE23-2; and (I, J) C3S/PCE23-6. The bar in parts A, C, E, G, and I is 2 μm. The bar in parts B, D, F, H, and J is 200 nm.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725484

very effective retarding action, which cause the initial dormantperiod to stop after 42 and 260 h, respectively (as reportedpreviously34).The LT-DSC technique also provides the evolution of the

finest microstructure in the samples. Blue and yellow bars inFigure 6 show respectively SGP and LGP pores. It is evidentthat the SGP volume is always higher than LGP, meaning thatapart from the capillary pores (acting as internal waterreservoirs), the microstructure due to the growth of C−S−Hgel mainly consists of a network of SGP pores. The analysis ofthe LT-DSC data shows that the volume of the nanometricpores (both SGP and LGP) in all the saturated samples doesnot change much, maintaining almost constant valuesthroughout the hydration process.Scanning Electron Microscopy. The microscopic analysis

performed on the samples by scanning electron microscopyevidence their multiscale porosity, ranging from the micro-meters to tenths of a nanometer. The SEM micrographsreported in Figure 7 show that the superplasticizers additioninfluences the aggregation of the C−S−H fibers and thus theporosity. The lowest magnification evidence the effect of PCEsin reducing the mean dimension of the C3S grains: this effect isevident in all samples except C3S/PCE102-2 (Figure 7C), inline with its lower efficiency. The other PCEs produce theseparation of the initial anhydrous C3S into grains withdimensions of the order of the micrometer, as shown in theFigures 7E,G,I. The best performing PCE, in terms ofdisaggregating the anhydrous grains, is PCE23-6 (Figure 7I),in line with its high adsorption propensity and retardationefficiency with respect to the other additives of the series.34

SEM images acquired at the highest magnification providethe visual evidence that PCEs modify the C−S−H micro-structure. In all cases the formed C−S−H is fibrillar. Thepresence of PCEs induces the growth of fibrils thinner thanthose observed in the C3S/water case. Moreover, PCEs affectthe spatial arrangement of these fibrils: PCE102-2 (Figure 7D)induces the formation of “suprafibrils” which are bundles ofaligned primary fibrils. PCE102-6 does not generate thesesuprastructures, while the primary fibrils are radially orientedfrom the anhydrous core. PCE23-X series results in a C−S−Harrangement that is intermediate among the two previouslydescribed, alternating regions with well-separated primary fibrilsto regions where the bundles are present. In these cases thebundles are less compact than those formed with PCE102-2.

■ CONCLUSIONSIt is known that some macroscopic properties of cement(degradation phenomena, elasticity, compressive strength, etc.)are influenced by porosity. For this reason the elaboration ofmethods able to easily access the characteristics of the porestructure during the hydration process is a task of primaryimportance. In this paper we showed for the first time that avery common and accessible technique like differential scanningcalorimetry can be used to extract most of the relevantinformation on the samples porosity. We investigated thedevelopment of C−S−H microstructure during the hydrationof C3S in the presence of PCEs estimating the volume of pores(capillary, large gel, and small gel pores), monitoring itthroughout the hydration. We also estimated the depercolationthreshold of the capillary porosity and how this value isinfluenced by commonly used cement additives (polymetha-crylic chains partially esterified with poly(ethylene oxide) lateralchains). Furthermore, we showed that an accurate analysis of

low-temperature calorimetric measurements performed inequilibrium conditions could disclose the fractal dimension ofthe pastes, and these values are in good accordance with thoseobtained from small-angle scattering measurements. It wasfound that the mass fractal dimension, Dm, grows during thehydration from 2.0 after 3 days to about 2.6 after months,because of the packing of the matrix due to the formation ofhydrated phases. For pastes prepared with w/c = 0.4, thepresence of PCEs decreases the packing of the matrix duringthe first month of curing with respect to the C3S/water paste, asdemonstrated by the lower Dm, while the fractal dimension isroughly the same in all the samples after 6 months. Theseresults, apart the significant information on the effect of additiveon the curing process of cement pastes, show that the analysisof simple DSC measurements can provide a wealth ofinformation on the fractal properties of this importantconstruction material.

■ ASSOCIATED CONTENT*S Supporting InformationHistogram of the total pore volume vs time for all theinvestigated samples; evolution of the fractal dimension (Dm) asa function of the hydration time for all the investigated samples.This material is available free of charge via the Internet athttp://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*Phone +39 055 457-3033; Fax +39 055 457-3032; [email protected] (P.B.).NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSAuthors thank CTG-Italcementi for providing the C3S purephase and Dr. S. Becker and Dr. J. Pakusch (BASF AG,Ludwigshafen, Germany) for providing the PCE super-plasticizers. Consorzio Interuniversitario per lo Sviluppo deiSistemi a Grande Interfase, CSGI, Ministero per la Istruzione,Universita e Ricerca (MIUR), and CTG-Italcementi aregratefully acknowledged for partial financial support.

■ REFERENCES(1) Masoero, E.; Del Gado, E.; Pellenq, R. J.-M.; Ulm, F.-J.; Yip, S.Nanostructure and Nanomechanics of Cement: Polydisperse ColloidalPacking. Phys. Rev. Lett. 2012, 109, 155503.(2) Richardson, I. G. The Nature of CSH in Hardened Cements.Cem. Concr. Res. 1999, 29, 1131−1147.(3) Jennings, H. M. A Model for the Microstructure of CalciumSilicate Hydrate in Cement Paste. Cem. Concr. Res. 2000, 30, 101−116.(4) Tennis, P.; Jennings, H. A Model for Two Types of CalciumSilicate Hydrate in the Microstructure of Portland Cement \ldots.Cem. Concr. Res. 2000, 30, 855−863.(5) Plassard, C.; Lesniewska, E.; Pochard, I.; Nonat, A. Investigationof the Surface Structure and Elastic Properties of Calcium SilicateHydrates at the Nanoscale. Ultramicroscopy 2004, 100, 331−338.(6) Nonat, A. The Structure and Stoichiometry of C-S-H. Cem.Concr. Res. 2004, 34, 1521−1528.(7) Fratini, E.; Ridi, F.; Chen, S.-H.; Baglioni, P. Hydration Waterand Microstructure in Calcium Silicate and Aluminate Hydrates. J.Phys.: Condens. Matter 2006, 18, S2467−S2483.(8) Garrault, S.; Behr, T.; Nonat, A. Formation of the C-S-H LayerDuring Early Hydration of Tricalcium Silicate Grains with DifferentSizes. J. Phys. Chem. B 2006, 110, 270−275.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725485

(9) Allen, A. J.; Thomas, J. J.; Jennings, H. M. Composition andDensity of Nanoscale Calcium-Silicate-Hydrate in Cement. Nat. Mater.2007, 6, 311−316.(10) Allen, A. J.; Thomas, J. J. Analysis of C−S−H Gel and CementPaste by Small-Angle Neutron Scattering. Cem. Concr. Res. 2007, 37,319−324.(11) Jennings, H. M. Refinements to Colloid Model of CSH inCement: CM-II. Cem. Concr. Res. 2008, 38, 275−289.(12) Pellenq, R. J.-M.; Kushima, A.; Shahsavari, R.; Vliet, K. J. V.;Buehler, M. J.; Yip, S.; Ulm, F.-J. A Realistic Molecular Model ofCement Hydrates. Proc. Natl. Acad. Sci. U. S. A. 2009, 106, 16102.(13) Alizadeh, R.; Beaudoin, J. J.; Raki, L. Mechanical Properties ofCalcium Silicate Hydrates. Mater. Struct. 2011, 44, 13−28.(14) Chiang, W.; Fratini, E.; Baglioni, P.; Liu, D.; Chen, S. H.Microstructure Determination of Calcium-Silicate-Hydrate Globulesby Small-Angle Neutron Scattering. J. Phys. Chem. C 2012, 116, 1−8.(15) Ridi, F.; Fratini, E.; Milani, S.; Baglioni, P. Near-infraredSpectroscopy Investigation of the Water Confined in TricalciumSilicate Pastes. J. Phys. Chem. B 2006, 110, 16326−16331.(16) Jennings, H. Colloid Model of C-S-H and Implications to theProblem of Creep and Shrinkage. Mater. Struct. 2004, 37, 59−70.(17) Thomas, J. J.; Jennings, H. M. A Colloidal Interpretation ofChemical Aging of the CSH Gel and Its Effects on the Properties ofCement Paste. Cem. Concr. Res. 2006, 36, 30−38.(18) Thomas, J. J.; Jennings, H. M.; Allen, A. J. RelationshipsBetween Composition and Density of Tobermorite, Jennite, andNanoscale CaO−SiO2−H2O. J. Phys. Chem. C 2010, 114, 7594−7601.(19) Jones, C. A.; Grasley, Z. C.; Ohlhausen, J. A. Measurement ofElastic Properties of Calcium Silicate Hydrate with Atomic ForceMicroscopy. Cem. Concr. Comp. 2012, 34, 468−477.(20) Chiang, W. C.; Fratini, E.; Ridi, F.; Lim, S. H.; Yeh, Y. Q.;Baglioni, P.; Choi, S. M.; Jeng, U. S.; Chen, S. H. MicrostructuralChanges of Globules in Calcium-Silicate-Hydrate Gels with andWithout Additives Determined by Small-Angle Neutron and X-RayScattering. J. Colloid Interface Sci. 2013, 398, 67−73.(21) Stauffer, D.; Aharony, A. Introduction to Percolation Theory, 2nded.; Taylor & Francis: London, 1994.(22) Scherer, G. W. Structure and Properties of Gels. Cem. Concr. Res.1999, 29, 1149−1157.(23) Bentz, D.; Garboczi, E. J. Percolation of Phases in a Three-dimensional Cement Paste Microstructural Model. Cem. Concr. Res.1991, 21, 325−344.(24) Bentz, D. P. Three-dimensional Computer Simulation ofPortland Cement Hydration and Microstructure Development. J. Am.Ceram. Soc. 1997, 80, 3−21.(25) Garboczi, E. J.; Bentz, D. Percolation Aspects of Cement Pasteand Concrete-properties and Durability. ACI Spec. Publ. 1999, 189,147−164.(26) Garboczi, E. J.; Bentz, D. The Effect of Statistical Fluctuation,Finite Size Error, and Digital Resolution on the Phase Percolation andTransport Properties of the NIST Cement Hydration Model. Cem.Concr. Res. 2001, 31, 1501−1514.(27) Bentz, D. P. Influence of Alkalis on Porosity Percolation inHydrating Cement Pastes. Cem. Concr. Comp. 2006, 28, 427−431.(28) Bentz, D. P. Capillary Porosity Depercolation/Repercolation inHydrating Cement Pastes Via Low-Temperature CalorimetryMeasurements and CEMHYD3D Modeling. J. Am. Ceram. Soc.2006, 89, 2606−2611.(29) Allen, A.; Oberthur, R.; Pearson, D. Development of the FinePorosity and Gel Structure of Hydrating Cement Systems. Philos. Mag.B 1987, 56, 263−288.(30) Kriechbaum, M.; Degovics, G.; Tritthart, J.; Laggner, P. FractalStructure of Portland Cement Paste During Age Hardening Analyzedby Small-Angle X-ray Scattering. Prog. Colloid Polym. Sci. 1989, 79,101−105.(31) Winslow, D.; Bukowski, J. M.; Young, J. F. The FractalArrangement of Hydrated Cement Paste. Cem. Concr. Res. 1995, 25,147−156.

(32) Heinemann, A.; Hermann, H.; Haußler, F. SANS Analysis ofFractal Microstructures in Hydrating Cement Paste. Physica B:Condens. Matter 2000, 276, 892−893.(33) Livingston, R. A. Fractal Nucleation and Growth Model for theHydration of Tricalcium Silicate. Cem. Concr. Res. 2000, 30, 1853−1860.(34) Ridi, F.; Fratini, E.; Luciani, P.; Winnefeld, F.; Baglioni, P.Tricalcium Silicate Hydration Reaction in the Presence of CombSuperplasticizers: Boundary Nucleation and Growth Model Applied toPolymer-modified Pastes. J. Phys. Chem. C 2012, 116, 10887−10895.(35) Sellevold, E. J.; Bager, D. H. Low Temperature Calorimetry as aPore Structure Probe. In 7th International Congress on the Chemistryof Cement, Paris, 1981; Vol. 4, pp 394−399.(36) Bager, D. H.; Sellevold, E. J. Ice Formation in HardenedCement Paste, Part I: Room Temperature Cured Pastes with VariableMoisture Contents. Cem. Concr. Res. 1986, 16, 709−720.(37) Bager, D. H.; Sellevold, E. J. Ice Formation in HardenedCement Paste, Part II: Drying and Resaturation on Room Temper-ature Cured Pastes. Cem. Concr. Res. 1986, 16, 835−844.(38) Bager, D. H.; Sellevold, E. J. Ice Formation in HardenedCement Paste, Part III: Slow Resaturation of Room TemperatureCured Pastes. Cem. Concr. Res. 1987, 17, 1−11.(39) Snyder, K. A.; Bentz, D. P. Suspended Hydration and Loss ofFreezable Water in Cement Pastes Exposed to 90% Relative Humidity.Cem. Concr. Res. 2004, 34, 2045−2056.(40) Bentz D. P.; Stutzman P. E. Curing, Hydration, andMicrostructure of Cement Paste. ACI Mater. J. 103, 348.(41) Sanz, E.; Vega De Las Heras, C.; Espinosa, J. R.; Caballero-Bernal, R.; Abascal, J. L. F.; Valeriani, C. Homogeneous Ice Nucleationat Moderate Supercooling from Molecular Simulation. J. Am. Chem.Soc. 2013, 135, 15008−15017.(42) Ridi, F.; Luciani, P.; Fratini, E.; Baglioni, P. Water Confined inCement Pastes as a Probe of Cement Microstructure Evolution. J.Phys. Chem. B 2009, 113, 3080−3087.(43) Brun, M.; Lallemand, A.; Quinson, J. F.; Eyraud, C. A NewMethod for the Simultaneous Determination of the Size and Shape ofPores: The Thermoporometry. Thermochim. Acta 1977, 21, 59−88.(44) Morishige, K.; Yasunaga, H.; Denoyel, R.; Wernert, V. Pore-Blocking-Controlled Freezing of Water in Cagelike Pores of KIT-5. J.Phys. Chem. C 2007, 111, 9488−9495.(45) Thomas, J. J.; Chen, J. J.; Allen, A. J.; Jennings, H. M. Effects ofDecalcification on the Microstructure and Surface Area of Cement andTricalcium Silicate Pastes. Cem. Concr. Res. 2004, 34, 2297−2307.(46) Vollet, D.; Donatti, D.; Ruiz, A. I.; Gatto, F. Mass FractalCharacteristics of Wet Sonogels as Determined by Small-Angle X-rayScattering and Differential Scanning Calorimetry. Phys. Rev. B 2006,74, 024208.(47) Rault, J.; Neffati, R.; Judeinstein, P. Melting of Ice in PorousGlass: Why Water and Solvents Confined in Small Pores Do NotCrystallize? Eur. Phys. J. B 2003, 36, 627−637.(48) Neffati, R.; Rault, J. Pore Size Distribution in Porous Glass:Fractal Dimension Obtained by Calorimetry. Eur. Phys. J. B 2001, 21,205−210.(49) As the crystalline ice density at 0 °C is 0.9167 g/cm3, the errorinvolved in this approximation is expected to be lower than 5%.(50) Vollet, D. R.; Scalari, J. P.; Donatti, D. A.; Ruiz, A. I. AThermoporometry and Small-Angle X-ray Scattering Study of WetSilica Sonogels as the Pore Volume Fraction Is Varied. J. Phys.:Condens. Matter 2007, 20, 025225.(51) Vollet, D.; Donatti, D.; Ruiz, A. I. Comparative Study UsingSmall-Angle X-ray Scattering and Nitrogen Adsorption in theCharacterization of Silica Xerogels and Aerogels. Phys. Rev. B 2004,69, 064202.(52) Teixeira, J. Small-Angle Scattering by Fractal Systems. J. App.Crystallogr. 1988, 21, 781−785.(53) Hansen, E. W.; Gran, H. C.; Sellevold, E. J. Heat of Fusion andSurface Tension of Solids Confined in Porous Materials Derived froma Combined Use of NMR and Calorimetry. J. Phys. Chem. B 1997, 101,7027−7032.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725486

(54) Zingg, A.; Winnefeld, F.; Holzer, L.; Pakusch, J.; Becker, S.;Gauckler, L. Adsorption of Polyelectrolytes and Its Influence on theRheology, Zeta Potential, and Microstructure of Various Cement andHydrate Phases. J. Colloid Interface Sci. 2008, 323, 301−312.

The Journal of Physical Chemistry C Article

dx.doi.org/10.1021/jp406268p | J. Phys. Chem. C 2013, 117, 25478−2548725487