Effect of carbon nanotubes on Portland cement matrices

abstract: This study evaluates the effects of three carbon nanotubes with different geometric characteristics on the rheological behaviour and mechanical performance, as well as on the microstructure of mortars and cement pastes. For nanotube content ranging from 0.025 to 0.2 wt%, the yield stress and viscosity were determined by rotational rheometry, and mechanical performance was evaluated by flexural strength and dynamic modulus of elasticity. The microstructural analysis was performed by X-ray diffraction, thermo-gravimetric analysis and scanning electron microscopy (SEM). The obtained results showed that the yield stress presents a considerable increase as the carbon nanotube content increases. The viscosity was also influenced by the presence of carbon nanotubes. The flexural strength of mortars increases for different amounts of carbon nanotubes, and depending on the geometric characteristics of the carbon nanotubes, the material behaves like a composite. The microstructural analysis showed the nucleation of hydration products on the surface of the carbon nanotubes, and that the better mechanical performance of matrices containing carbon nanotubes is not related to the increase in hydration products.


INTRODUCTION
The use of carbon nanotubes is a recent practice.The first studies were developed in the 20th century and published by Li et al. [1] and Yakolev et al. [2].Since this research, interest in this topic has increased considerably.The rheological behaviour of cement-based materials is directly affected by the incorporation of carbon nanotubes.An increase in viscosity of cement pastes due to the incorporation of the same type of carbon nanotubes was demonstrated by Konsta-Gdoutos et al. [3].A linear increase in yield stress was published by Andrade et al. [4], Reales et al. [5], [6], and Batiston et al. [7] for cement based-material pastes formulated with a carbon nanotube content of less than 0.2 wt%.For larger amounts of carbon nanotubes, Medeiros et al. [8] and Jiang et al. [9] observed an exponential increase in yield stress for mixtures containing up to 0.5 wt% of carbon nanotubes.
Many studies have evaluated the effect of the same type of carbon nanotubes with and without functionalization on mechanical performance of cement-based materials.However, only some of the available references evaluated the geometric characteristics of carbon nanotubes: the diameter was kept constant and the length varied by Konsta-Gdoutos et al. [10], Musso et al. [11], Al-Rub et al. [12] and Chen et al. [13].While the length was kept constant and the diameter of carbon nanotubes was changed in the studies developed by Collins et al. [14] and Gao et al. [15].The diameter and length were both varied in studies published by Sobolkina et al. [16], Souza et al. [17], Hawreen [18], [19] and Manzur et al. [20].
Considering that the effect of carbon nanotubes on mechanical performance of cement-based materials is a microstructure response, the mechanisms of these fibers in cementitious matrices were discussed by Fehervari et al. [21], Chen et al. [22] and Ramezani et al. [23].Chen et al. [22] showed that part the carbon nanotubes can adsorb on the surface of cement particles, and part of them can perform as a reinforcing material.The improvement in the strength of mortars formulated with a low content of carbon nanofibers (< 1 wt%) was not related to the increase in hydration products or the reduction of porosity as published by Fehervari et al. [21].Thus, this paper aims to study the effect of carbon nanotubes with different geometric characteristics on rheological behaviour, mechanical performance and cementitious matrix microstructure.

MATERIALS AND EXPERIMENTAL PROGRAM
Three different aqueous suspensions of multi-walled carbon nanotubes (MWCNT) were obtained from Nano Amorphous Materials.They were called CNT1, CNT 2 and CNT 3 , and their characteristics are listed in Table 1.According to the supplier procedures, the carbon nanotubes were dispersed using a combination of 90 wt% of aromatic modified polyethylene glycol ether (PEG) and 10 wt% of water.The amount of dispersing agent present in each product can be calculated (Nanoamor, 2022).According to information provided by the supplier, the diameter and length of different samples of carbon nanotubes was obtained from High Resolution Transmission Electron Microscopy and Raman spectroscopy.The specific surface area was measured by gas adsorption (B.E.T).Samples were dried in an oven at 105°C and kept under a vacuum of 68.9 pascal at 60°C for 24 hours.The density of carbon nanotubes was obtained from the supplier.Table 2 presents the physical-chemical characteristics of Portland cement type CPV-ARI, equivalent to type III cement [24].Chemical composition was determined by X-ray fluorescence in an unpressed sample, using the EDX 750 Shimadzu equipment.Lignosulfonate-based superplasticizer was used as a dispersing additive.The particle size distributions of carbon nanotubes and Portland cement CPV-ARI were determined with a Malvern 2200 laser granulometer, which was also used to evaluate the granulometry of carbon nanotubes by Yakovlev et al. [25], and Krause et al. [26] for carbon nanotubes in dry powder condition.Transmission Electron Microscopy (TEM) of carbon nanotube suspensions was carried out with Fei Tecnai G2 equipment, operating at 80 and 120 kV.The samples of nanotube suspensions were diluted to 1 mg/ml.Five milliliters (5μl) of these diluted suspensions were placed on a copper-carbon grid (300 mesh) and covered with Formvar.
Figure 1 shows the Portland cement diffractogram and the main phases were identified: Alite (C3S) -3CaO.SiO 2 , Belite (C 2 S) -2CaO.SiO 2 , Gypsum -Ca.SO 4 .1/2H 2 O and Calcium hydroxide -Ca(OH) 2 [27], as well as the aluminum oxide (Al 2 O 3 ) peaks at 2θ = 25.5, 35.15, 43.35 o .The presence of Arkanite -K 2 .SO 4 , and Langbeinite -K 2 Mg 2 (SO 4 ) 3 phosphates phases was identified considering their peaks at 2θ = 30.78o and 31.02o , respectively.The chemical analysis of cement (Table 2) did not reveal the presence of aluminum oxide, which does not allow the formation of Tricalcium aluminate (C 3 A) and was not identified in X-ray diffraction considering its main peaks at 2θ = 20.95o and 33.25 o .The presence of tetra-calcium ferro-aluminate (C 4 AF) -4CaO.Al 2 O 3 .Fe 2 O 3 can be replaced by tetra calcium ferrite, 4CaO.Fe 2 O 3 , as iron ions can take the place of aluminum ions in a crystalline structure of C 4 F [27].Chemical analysis also revealed a considerable percentage of potassium oxide, 5.87% of K 2 O.This is even considering that the amount of this element was combined in sulfate phase, for example, the Arkanite K 2 .SO 4 .There is only 0.693% of SO 3 according to the chemical analysis, which can represent an amount around 1.5% of K 2 .SO 4 .Thus, subtracting the percentage of 0.816% of K 2 O chemically bounded in the sulfate phase, there is 5.05% of K 2 O which must be chemically bounded with the clinker phases.Taylor [27] reported that Woermann et al. found that up to 1.4% of Na 2 O or K 2 O can be incorporated into C 3 S, where the sodium and potassium ions can replace calcium ions.Or even a type of belite phase compound (Glascrite) KC 23 S 12 can be used, which contains 3.5% of potassium A reference mixture without carbon nanotubes, and mixtures formulated with 0.025, 0.05, 0.1 and 0.2 wt% of CNT 1 , CNT 2 and CNT 3 were evaluated.The lignosulfonate-based dispersant content was fixed at 1 wt.% of solids, and the water/solids ratio of 0.375 was kept constant.For all mixtures, the water content of the dispersant and carbon nanotube suspensions was subtracted from the total amount.For mortars, the cement:sand ratio was fixed at 1:1.In order to consider the possible influence of the amount of polyethylene glycol on the evaluated properties, the concentration of this dispersing agent in each formulation was calculated, whose values are listed in Table 3.
The suspensions of carbon nanotubes and the dispersant were previously diluted in deionized water.The mixing was carried out in a laboratory planetary mixer, applying the following process: (i) the dry powder was mixed at 60 rpm for 60s; (ii) 2/3 of suspension (water + dispersant) was added and mixed at 60 rpm for 120s; (iii) 1/3 of suspension was added (water + carbon nanotubes) and mixed at 60 rpm for 120s.Contents ranging from 600 to 650g were mixed for each batch.A temperature-controlled rheometer with Vane geometry with a diameter of 15 mm and a height of 30 mm was used to measure the rheological properties of pastes.Flow curves were obtained using a shear rate control, ranging from 10 to 100 s -1 , in 10 s -1 intervals.The shear rate (γ) was increased from 10 to 100 s -1 (upper curve) and reduced from 100 to 10 s -1 (downs curve).The pastes were held for 30 seconds at each shear rate and the values were recorded in the last 3 seconds.The yield stress (τ 0 ) and the plastic viscosity (η) were calculated using the Bingham Equation (τ= τ 0 + η*γ), considering the upper-steps curve.All rotational rheometry tests were performed at 23 o C after 5 minutes of adding water to the powder.Six prismatic samples (1.6x4x16cm) of mortars were molded; compaction was applied to avoid molding defects.The samples were kept at 23°C for 24 hours and immersed in a saturated calcium hydroxide solution at 23°C for 7 days.The flexural strength of mixtures was measured on six prismatic samples, which had the upper face milled to remove surface irregularities.The flexural strength was measured using a universal testing machine applying a strain rate of 0.1 mm/s.The dynamic modulus of elasticity was measured according to standard procedure [28].
For the microstructural analysis, the inner part of the paste sample was extracted, which was maintained under the same curing conditions described above.For X-ray diffraction and thermogravimetric analysis, the samples were ground and sieved through 200 mesh (0.075 mm).The diffractogram of samples was obtained in a Buckler X-ray diffractometer operating at 30 kV and 10 mA, with a copper tube, from 10 to 70° at 0.01 o /s.Thermogravimetric analysis was performed from 100 to 950°C with a heating rate of 10°C/min in an alumina crucible, with a nitrogen flow of 20 ml/min.The microstructure of the samples was analyzed by scanning electron microscope (SEM) coupled to an energy dispersive spectrometer (EDS) using Quanta 600 FEI-Philips equipment, operating at 25 kV.The gold coating was applied to the surface of the sample for this analysis.

RESULTS AND DISCUSSIONS
Figure 2 presents the transmission electron micrographies of carbon nanotubes CNT 1 , which shows in Figure 2a the presence of large agglomerates of nanotubes, with dimensions of 14 and 23 μm.Near this large particle, it is possible to observe a smaller particle of nanotube agglomerates with a diameter around 2 μm.A tangle of carbon nanotubes with a length around 6.5 μm can also be seen, as well as some individual fibers of carbon nanotubes with a size of 3.4 μm. Figure 2b and c show carbon nanotubes with lengths of 250 and 263 nm and diameters of 14 and 16 nm, respectively.Figure 3 presents the TEM micrographies of carbon nanotubes CNT 2 , which shows a similar agglomeration of carbon nanotubes.Here the carbon nanotubes have a length of around 3.26 and 1,06 μm and diameters varying from 47 to 69 nm. Figure 4 presents the TEM micrographies of carbon nanotubes CNT 3 , which in Figure 4a shows micrometric agglomerates of carbon nanotubes, as well as entangled fibers and individual particles of carbon nanotubes.The diameter of the carbon nanotubes varies from 37 to 39 nm.Considering the values of diameter and length declared by the supplier, there is a clear difference from the values observed in the TEM micrographies for CNT 1 and CNT 2 carbon nanotubes.For carbon nanotubes CNT 3 , the diameter values observed in the TEM images are in interval range declared by the supplier.According to the supplier, for values described in the Table 1, the length was obtained from high resolution transmission electron microscopy and the diameter was found from Raman spectroscopy.The particle sizes of three carbon nanotubes, namely diameter and length, were also observed in a previous publication [29].As seen in the TEM images, the sonication process used for preparation of samples of carbon nanotubes for microscopy, or even possibly used by the supplier for dispersing them, resulted in breaking carbon nanotubes into small parts.Carbon nanotubes have micrometric particles ranging from 1 to 100 μm and another population of particles smaller than 1 μm.CNT 1 is coarser than CNT 2 and CNT 3 .Thus, considering the specific surface area of carbon nanotubes, it is verified in Table 1 that CNT 3 presented the highest SSA, while CNT 1 and CNT 2 presented similar values.The granulometry of carbon nanotubes before and after the ultrasonic dispersion was evaluated by Yakovlev et al. [25].When in an agglomerated condition, nanotubes present the main volume of micrometric particles (> 1 μm), while the use of ultrasonic dispersion leads to the predominance of nanoparticles.The ultrasonic dispersion leads not only to a disentanglement of agglomerates of carbon nanotubes, but also to their rupture or breakage as reported by Hawreen et al. [18], [19] and Yakovlev et al. [25].These two populations of particles, micrometric agglomerates or tangles of carbon nanotubes, are also observable in the TEM micrographies between 1 and 100 μm, as well as particles smaller than 1 μm, which are probably individual fibers of carbon nanotubes or pieces of them, resulting from the sonication process.The laser granulometry test cannot assess the morphology of fibrous particles, such as diameter and length, even with equipment that combines laser granulometry and image analysis techniques.This technique provides information of the sizes of agglomerates or clamped particles, as well as dispersed carbon nanotubes and their small pieces, without a classification of their dispersion conditions [25], [26].When combined with other techniques such as transmission electron microscopy and specific surface area, the laser granulometry test can help to understand how these particles affect the rheological behaviour of cementitious matrices [29].Other techniques are available, such as UV-VIS spectroscopy [24], electro-acoustic attenuation spectroscopy [29], and LUMiSizer centrifugal methods [26].
Figure 6a shows the rheograms of reference paste and mixtures formulated with 0.1 and 0.2 wt% of CNT1 carbon nanotubes.The Bingham Model was considered to fit a linear equation.The rheological behaviour was considerably altered due to the incorporation of carbon nanotubes in cementitious matrix.Table 4 lists the calculated volumetric concentration (V S ), volumetric surface area (VSA), porosity (P 0 ) and interparticle separation (IPS) calculations of the studied mixtures according to Funk and Dinger [30], and the yield stress and viscosity of cement pastes obtained from the Bingham model.Figure 6b shows the correlation between the interparticle separation (IPS) and yield stress of mixtures, in which a good correlation can be obtained.The most probable effect of carbon nanotubes is to reduce the interparticle separation of these suspensions, resulting in an exponential increase in yield stress, which is inversely proportional to interparticle separation IPS (1/IPS 2 ) [31].Regarding the yield stress, the volumetric surface area (VSA) can be pointed as the main variable responsible for this rheological property.While the viscosity of suspensions is mainly affected by the increase of the solid concentration (V S ), which changed from 47.14% to 47.16%, viscosity does not present a good correlation with the Interparticle Separation (IPS).The type and content of surfactant was also cited as a variable responsible to the rheological change in cement formulated with carbon nanotubes [5], [6].However, a similar increase of the rheological properties was observed by Batiston et al. [7] and Medeiros et al. [8] for mixtures containing carbon nanotubes, which used samples of carbon nanotubes in a powdered form without a surfactant agent for dispersing the carbon nanotubes.A polycarboxylate acid was used as a dispersing agent in both cases.Thus, considering the high amount of dispersing lignosulphonate employed here (1 wt%), which also acts as an entrapped air agent, and the lower dosage of surfactant in a mmol fraction, these same rheological properties, yield stress and viscosity, do not present good correlation with the content of surfactant (Table 2).
The mechanical properties of mixtures formulated with carbon nanotubes CNT1, CNT 2 and CNT 3 were also listed in Table 5.All mixtures have flexural strength values higher than reference.For mixtures containing carbon nanotubes CNT 1 and CNT 3 , the maximum value was reached at 0.025 wt%.For CNT 2 carbon nanotubes, the optimal content, or the smallest quantity of nanotubes that led to the highest mechanical performance, was 0.05 wt%.Unlike the others, the flexural strength of CNT 2 carbon nanotubes increases after reaching the optimal content.Recently, a statistical study was published by Ramezani et al. [23], evaluating the large number of published studies on the use of carbon nanotubes in cement-based materials.Considering the flexural or tensile strength, the authors suggested an optimal content of 0.15 wt% of carbon nanotubes.For an experimental approach in the same study, they reached an optimal content of 0.1 wt% of carbon nanotubes, considering the flexural strength.
Figure 7a presents the stress (f tm ) x deflection (δ) curves of all evaluated samples.For the reference mixture, the typical behaviour of brittleness was identified.A maximum deflection (δ max ) of 0.2287 mm was achieved for this cementitious matrix.When 0.025 wt% of CNT 1 was added to this cementitious matrix, one of the six samples presented a higher deformation capacity.A maximum deflection (δ max ) of 0.4612 mm was observed, and the inclination of initial part of stress x deflection curve showed a reduction compared to the other samples.These two differences suggest a typical composite behaviour for this sample.The tenacity of this sample reached 1.88 J, considerably higher than the average value of 0.79 for mixtures containing 0.025 wt% of CNT 1 and listed in Table 5.A similar change in deformability of the same samples was observed for other mixtures formulated with CNT 1 , as seen in Figure 7c, d and e. Figure 7f presents the typical behaviour of the reference mixtures and all samples containing CNT 1 , which present the behaviour of a composite.A type of "plasticity" was verified in some stress x deflection curves, which can be attributed to the sample cracking and the performance of carbon nanotubes as a reinforcing material.
The same approach was considered for the mixtures formulated with carbon nanotubes CNT 2 and CNT 3 , which are showed in Figures 8 and 9.There are many samples for different contents of carbon nanotubes, which present a typical composite behaviour.The behaviour of a composite type was mainly observed in Figure 8f, Figure 9f and Figure 7f.Thus, there is a higher probability of obtaining a composite type behaviour for CNT 2 (12/24), followed by a smaller probability for CNT 3 (7/24) and the least probability for CNT1 (5/24).According to Callister [32], a minimum critical length was required when fibers are added into a matrix (Equation A).Considering that the length (l c ), the diameter (d) and the tensile strength of carbon nanotubes (σ f ) are chosen, this last one can be considered equal to 850 MPa [33] or kept constant.The matrix strength (σ m ) or the adherence strength between carbon nanotubes and cementitious matrix is the only variable considered, which is proportional to its tensile strength.Peyvandi et al. [34] evaluated the effect of cementitious paste density, and consequently strength, on the performance efficiency of carbon nanofibers.The authors demonstrated that the efficiency of carbon nanofiber reinforcement was improved when the mechanical strength of cementitious matrix was increased.To assess how the adherence strength influenced the reinforcement efficiency of carbon nanotubes in cement-based materials, an efficiency factor called reinforced efficiency (RE) can be estimated by Equation B. This is given by the ratio between the adherence strength (σ m ) required from Equation A, and the tensile strength of refence mixture without carbon nanotubes (σ REF = 5.89 MPa). Figure 10a presents the probability of occurrence of similar composite behaviour for different Reinforced efficiency (R E ) values, which are indirectly related to the aspect ratio of carbon nanotubes.As the adherence mechanism is improved due to the geometric characteristics of carbon nanotubes, the probability of occurrence of behaviour similar to composite increases.In the mixtures formulated with CNT 2 , which presented the best adherence mechanism between fiber and cementitious matrix, the probability of occurrence of behaviour similar to composite also increases according to CNT 2 content, as shown in Figure 10b.For the other carbon nanotubes, this relationship between fiber volume and composite like behaviour was not identified.Figure 11 shows the effects of carbon nanotubes on the hydration of Portland cement pastes containing 0.2 wt% of carbon nanotubes CNT 1 , CNT 2 and CNT 3 , named C01, C02 and C03.The thermogravimetric analysis of these mixtures showed that the composition C01 presents the highest mass loss, followed by compositions C03 and C02.Considering the mass loss between 100 and 400 o C, which represents the calcium silicate and aluminate hydrates (C-S-H/C-A-H) [27], the mixtures C01, C03 and C02 showed values of 8.4, 7.8 and 5.2%, respectively.As for calcium hydroxide content, for temperatures ranging from 400 to 600 o C [27], these formulations showed a mass loss of 4.7, 3 and 2.3%, respectively.For temperatures above 600 o C [27], the decarbonation of calcium carbonate showed similar values ranging from 11.1 to 12.1%.
Figure 12 shows the diffractogram of composites C01, C02 and C03, containing 0.2 wt% of carbon nanotubes CNT 1 , CNT 2 and CNT 3 .Considering the C 3 S peaks at 2θ = 29.5 o , the highest consumption of this phase was observed for the composite C01, followed by C02 and C03, respectively.Consequently, an inverse trend was observed for the main calcium hydroxide peaks at 2θ = 34.1 o .These results revealed that the increase in flexural strength observed for these formulations was not related to hydration products, since the C02 mixture, which presented the highest flexural strength, had the lowest mass loss for the hydration products.Likewise, the C02 mixture, which presents the highest flexural strength, did not present the maximum consumption of C 2 S/C 3 S phases, or the lowest peak for these phases.Recently, Fehervari et al. [21] revealed that mechanisms for improved strength of mortars formulated with carbon nanofibers were not related to the hydration products.For higher dosages of carbon nanofibers (1 wt%), as compared with the values considered here, authors reported no difference on the porosity of mixtures formulated with or without these nanoparticles.The obtained results indicate that the geometric characteristic of carbon nanotubes can be pointed as an important issue.This approach of considering the aspect ratio and mechanical performance was also used by Hawreen et al. [19] and Batiston et al. [7].For mixtures containing 0.1 wt% of carbon nanotubes with different aspect ratios, they showed that flexural strength was improved as the aspect ratio of carbon nanotubes increases.Figure 13a shows a clear increase of diameter of the carbon nanotubes due to the growth of hydration products on the surface of nanoparticles. Figure 13b shows that the carbon nanotubes are entrapped into the hydration products.Figure 13c shows one carbon nanotube adhered on the surface of a cement particle, with the longest length covered by hydration products.Figure 13d shows many carbon nanotubes in a pore of the cementitious matrix, and the indication of the growth of a hydration product due to the nucleation effect of carbon nanotubes.

CONCLUSIONS
Carbon nanotubes affect the rheological properties of the Portland cement matrix.The high specific surface area of nanoparticles led to a reduction of the interparticle separation, and, consequently, to increasing the yield stress and viscosity of suspensions.
The mechanical performance is improved due to the incorporation of carbon nanotubes into the Portland cement matrix.A composite like behaviour was observed, which is directly related to the geometric characteristics of these nanoparticles.
The microstructural analysis showed that the geometric characteristics of carbon nanotubes do not change the hydration of these composite materials.The nucleation of hydration products takes place on the surface of carbon nanotubes.

Figure 4 .
Figure 4. Transmission electron micrographies of carbon nanotubes CNT3 Figures 5 presents the particle size distribution of carbon nanotubes and Portland cement obtained by laser granulometry.Carbon nanotubes have micrometric particles ranging from 1 to 100 μm and another population of particles smaller than 1 μm.CNT 1 is coarser than CNT 2 and CNT 3 .Thus, considering the specific surface area of carbon nanotubes, it is verified in Table1that CNT 3 presented the highest SSA, while CNT 1 and CNT 2 presented similar values.The granulometry of carbon nanotubes before and after the ultrasonic dispersion was evaluated by Yakovlev et al.[25].When in an agglomerated condition, nanotubes present the main volume of micrometric particles (> 1 μm), while the use of ultrasonic dispersion leads to the predominance of nanoparticles.The ultrasonic dispersion leads not only to a disentanglement of agglomerates of carbon nanotubes, but also to their rupture or breakage as reported by Hawreen et al.[18],[19] and Yakovlev et al.[25].These two populations of particles, micrometric agglomerates or tangles of carbon nanotubes, are also observable in the TEM micrographies between 1 and 100 μm, as well as particles smaller than 1 μm, which are probably individual fibers of carbon nanotubes or pieces of them, resulting from the sonication process.

Figure 5 .
Figure 5. Particle size distribution Portland cement and carbon nanotubes

Figure 6 .
Figure 6.(a) Rheograms of mixtures (b) Effect of Interparticle separation (IPS) on the yield stress

Figure 10 .
Figure 10.Probability of composite like behaviour as function of (a) Reinforced efficiency (b) volume of carbon nanotubes CNT2

Figure 13 .
Figure 13.Scanning electron micrographies of carbon nanotubes in Portland cement matrix

Table 5 -
Mechanical properties of mixtures

Table 6 -
Table 6 presents the calculated values for CNT 1 , CNT 2 and CNT 3 , considering their geometric characteristics.Geometric and adherence strength properties of carbon nanotubes