Effects of Alumina Nanofibers and Cellulose Nanocrystals on Durability and Self-Healing Capacity of Ultrahigh-Performance Fiber-Reinforced Concretes

The experimental program focused on three ultrahigh-performance fiber-reinforced concretes, the composition of which is detailed in Table 1. The mixes contain an elevated amount of cement ($600\mathrm{kg}/{\mathrm{m}}^3$) and a very reduced water/binder ratio (0.18) typical of ultrahigh-performance RCs. One of the aims of the paper was to analyze their mechanical and durability performance when exposed to aggressive environments, represented by geothermal water, containing chlorides ($\mathrm{Cl}=440.9\mathrm{ppm}$) and high contents of sulfates (${\mathrm{SO}}_4=\mathrm{2,678}\mathrm{ppm}$) that classify the aggressiveness as XA; the composition of this water is presented in Table 2. These materials are defined in the ReSHEALience project as ultra high durability concretes and, therefore this name is used throughout this work. Because of the intended applications (cooling water tanks of a geothermal power plant), the reference mix (UHDC-CA) incorporated a significant amount of ground granulated blast-furnace slag (GGBS), a well-known mineral addition in conventional concretes but less common in UHPFRCs. This mineral addition contributes to refining the pores in the concrete matrix and hence enhancing its durability in combination with its ability to reduce aggressive transport as a consequence of capture and combine chlorides and sulfates, both of which are aggressive ions that were considered in this work (Kayali et al. 2012; Chen et al. 2020; Luna et al. 2018; Ozbay et al. 2016). The composition of the employed portland cement (PC) CEM I 52.5R and GGBS, as determined by X-ray fluorescence (XRF), is presented in Table 3. The main components of the PC were CaO (59.7%) and ${\mathrm{SiO}}_2$ (19.5%), while the GGBS contained CaO (39.2%) and ${\mathrm{SiO}}_2$ (38.9%), as well as ${\mathrm{Al}}_2{\mathrm{O}}_3$ (10.2%) that contributed to combine sulfates and chlorides. All the mixes included steel fibers (SFs) ($l_f=20\mathrm{mm}$; $d_f=0.22\mathrm{mm}$) at a volume fraction of 1.5% to obtain a strain-hardening tensile response at the required level of strength and deformation capacity (Lo Monte and Ferrara 2020). Moreover, in all the investigated mixes, Penetron (Collegno, Torino, Italy) Admix crystalline admixture was used as a stimulator of the autogenous self-healing capacity, dosed at 0.8% by weight of cement (Cuenca et al. 2021b). Its morphological and chemical characteristics were described by Cuenca et al. (2018); it is a powder consisting of portland cement, quartz sand, and other reactive proprietary chemicals. In addition, two further mixes, based on the reference UHDC-CA, were produced: UHDC-CA+ANF, containing 0.25% by weight of cement alumina nanofibers (ANFs); and UHDC-CA+CNC, containing 0.15% by weight of cement cellulose nanocrystals (CNCs).

The interest in including the aforementioned nanoconstituents was explained in detail in the section “Introduction.” Alumina nanofibers (Nafen, Tallinn, Estonia) had an average diameter between 4 and 11 nm and a length of 100–900 nm, with a specific surface of $155{\mathrm{m}}^2/\mathrm{g}$. To facilitate the mixing process in concretes with low water/binder ratios, this product is supplied in a water suspension with a solid phase concentration equal to 10%. Cellulose nanocrystals (API Europe, Athens, Greece) had an average diameter of 5 nm and were 50–500 nm in length, and were supplied in water suspension at 10% solid content.

For this investigation, different specimens based on the test requirements were produced: prismatic specimens $40\times 40\times 160\mathrm{mm}$ for shrinkage, flexural, and compressive strength, self-healing, and water sorptivity tests; and cylinders, $Ø100\mathrm{mm}\times 300\mathrm{mm}$ high, from which 50-mm-thick disks were obtained for characterization of microstructure and intrinsic durability parameters, including chloride transport and water capillary suction.

After demolding, all specimens were cured in a moist room [$T=20°\mathrm{C}$, relative humidity $(\mathrm{RH})=95\%$] until the related tests were performed. A series of $40\times 40\times 160\mathrm{mm}$ samples also was cured immersed in geothermal water to analyze the mechanical and durability behavior of sound and cracked concrete exposed to aggressive acid environments (XA exposure conditions). The geothermal water came from the cooling water tanks of a geothermal power plant owned by Enel Green Power (a partner of the ReSHEALience project) in Tuscany (Italy) (Table 2).

The mechanical and durability behavior of concrete exposed to aggressive chemical environments was analyzed in uncracked and cracked states. Tests in the uncracked state (section “Mercury Intrusion Porosimetry” (MIP) to section “Mechanical Properties”) were required to characterize the cementitious matrix of the investigated UHDCs from the microstructural and intrinsic durability points of view, whereas tests of cracked samples (sections “Visual Crack Analysis to Quantify Self-Sealing Capacity” and “Sorptivity Tests for Evaluation of Self-Healing Efficiency”) were performed to identify the durability enhancement in expected service conditions (Fig. 1).

The X-ray diffraction patterns of the UHDC-CA, UHDC-CA+ANF, and UHDC-CA+CNC after 2 months of curing are reported in Fig. 10(a). Similar crystalline phases were detected in the three UHDCs which were typical of hydrated cement paste, including calcium silicate hydrate. The presence of portlandite also was indicated, but only residual ettringite, although monocarboaluminate, hemicarbonate or hydrotalcite cannot be disregarded. As suggested by Fernández et al. (2018) in cement pastes with 30% GGBS content, or higher, all these hydrated phases will contribute later to interact with aggressive water ions as chloride or sulfates. The peak of diffraction associated with C─ S─ H ($29.5°\Theta$) overlaps with the most intense peak of the calcite. The diffractograms also indicate the presence of quartz and albite, which were associated with the aggregate and slag, as well as a significant content of unreacted cement particles as alite and ferrite phases. The presence of the nanoconstituents (alumina nanofibers or cellulose nanocrystals) did not affect the stability of the hydrated phases because analogous phases were detected in the related diffractograms, compared with those recorded for the UHDC-CA concrete. The diffraction peak at $18.0°\Theta$ appears in all concretes and is associated with portlandite and related to the progression of cement hydration. The peak at $29.5°\Theta$ is related to the calcite phase, partly as a consequence of CA addition, also identified by Ferrara et al. (2014). Thermogravimetric analysis of the three concretes, Fig. 10(b), shows the derivate weight loss that allows identifying the temperature stability of hydrated cement paste phases. A mass loss in the 110°C–170°C region is more evident for the ANF and CNC mixes, likely as a consequence of stimulation of hydration induced by the nanoconstituents. This peak is associated with the dehydration of the hemi- or monocarboaluminate, as suggested by Fernández et al. (2018), and also can be related to a certain reaction between the calcium carbonate, the production of which also may be stimulated by the CA, and the aluminate hydrates of the binder. Monocarboaluminate causes the stabilization of ettringite, thus increasing the volume of hydration products, which can improve the compressive strength measured in the concretes with ANF and CNC. A shoulder appears just before the portlandite peak that can be attributed to hydrotalcite and which also increases in intensity with BFS hydration (Fernández et al. 2018). Three main peaks are common in portland base concretes: (1) a peak up to 350°C is associated mainly with C─ S─ H gel dehydration, but also with other aluminate phases (Kumar et al. 2008; Soin et al. 2013), (2) a peak at 400°C–500°C is associated with portlandite dehydration (Kumar et al. 2008; Soin et al. 2013), and (3) a peak at 650°C–750°C is associated with the decomposition of carbonates (Kumar et al. 2008; Soin et al. 2013).

The mass loss by weight of cement was estimated at different temperature ranges for bound water, portlandite, and calcite (Table 5). The bound water contents were quite similar for the three cementitious composites, and ranged between 2.13 and 2.25. Some higher portlandite content in the concretes with alumina nanofibers or cellulose nanocrystals was a consequence of higher production of hydration products, probably due to two reasons: the nanoparticles acted as nucleation sites and, in the case of CNC, their ability to cumulate water can favor the hydration of the cement, and in consequence, it can cause the increase of compressive strength, the lower porosity and the lower shrinkage observed in this concrete (Kawashima and Shah 2011).

The chemical composition of the cement paste also was studied by scanning electron microscopy (SEM)/energy dispersive X-Ray spectroscopy (EDX). The main element distribution in cement paste is highlighted in Fig. 11 and expressed as a percentage of the main oxides in Fig. 12 for the three UHDC concretes. Intensive Ca content is indicated. In addition, Al was detected which was distributed homogenously in the cement paste. This also occurred in the case of sulfates or sulfoaluminate phases, detected from XRD and thermogravimetry (TG)/derivative thermogravimetry (DTG) as characteristic due to the presence of GGBS. Anhydrous grains of cement and GGBS were observed clearly. The most characteristic aspect was the high content of CaO, which likely was related to the composition of the type of cement used (Table 3) and the incorporation of the CA.

The percentage distribution of the different oxides in the cement paste shows that the concretes with the CA and ANF and the CA and CNC had a higher CaO content, 60%, than that observed in UHDC-CA (50%). This can be due to the presence of the nanoadditions of ANF and CNC, which stimulated the hydration and growth of C─ S─ H and the formation of hemicarbonate or monocarboaluminate phases detected in the DTG curves. In these two concretes, a greater percentage of portlandite also was formed. In addition, there was hardly any variation of the other main two oxides in cement paste in the three UHDCs studied; the ${\mathrm{SiO}}_2$ content was 25% and the ${\mathrm{Al}}_2{\mathrm{O}}_3$ content was 6%–7%. This is $\mathrm{CaO}/{\mathrm{SiO}}_2>1.8$, which is common in ordinary portland cement paste and higher than that found by Fernández et al. (2018) in a paste also containing 30% GGBS. T Al was detected in the cement paste.

To investigate accelerated chloride transport, Fig. 13(a) shows the non-steady-state chloride migration coefficient, $D_{\mathrm{nssm}}$ after 2 months of curing. The UHDC-CA concrete had $D_{\mathrm{nssm}}$ values of about $0.99\times {10}^{-12}{\mathrm{m}}^2/\mathrm{s}$. The incorporation of alumina nanofibers and cellulose nanocrystals slightly modified the chloride resistance of concrete, because values of the same order were obtained. However, all these $D_{\mathrm{nssm}}$ values, indicating a low permeability of chloride ions for all the investigated mixes, correspond to expected very high durability according to the classification given by Baroghel-Bouny (2004). This good resistance to the accelerated Cl transport in these concretes mainly is a consequence of the low porosity and the compact and dense microstructure of the three investigated concretes.

The natural chloride transport is represented in Fig. 13(b), taking into account the chloride penetration depth and percentage of total and bound chloride (by weight of binder) in 1 cm for the three concretes after 6 and 24 months of the exposure. At low exposure ages (6 months) no significant differences in chloride penetration were observed, mainly due to the low porosity of the concretes, although some penetration values were lower in UHDC containing ANF or CNC. The difference in total Cl penetrated was more evident at longer exposure ages (24 months). The bound chlorides were lower with ANF and CNC, in agreement with some higher cement hydration identified by XRD and TG. In the three concretes, the percentage of bound chlorides was very high (between 40% and 50%), which can contribute to retard the aggressive ion transport, and highlights the beneficial contribution of GGBS. These concretes contained $500\mathrm{kg}/{\mathrm{m}}^3$ of blast furnace slag in their formulations, and the presence of this raw material favored the formation of hydrotalcite with excellent efficiency in binding chloride ions and monocarboaluminates, phases which were detected by XRD [Fig. 10(b)], which also adsorb a large amount of chloride ions. These results corroborate those obtained in previous studies, such as those carried out by Kayali et al. (2012), Luna et al. (2018), and Chen et al. (2020), which found an increase in the content of Friedel’s salts and the formation of hydrotalcite, enhancing the total capacity of chloride binding of the blended slag cementitious matrixes and retardation of chloride transport.

These data obtained from the accelerated and natural tests confirm the impermeability of these UHDCs to the penetration of aggressive substances such as chlorides; the incorporation of nanoalumina or nanocellulose led to some improvement, and no negative response was observed, indicating an overall high durability as a consequence of the refined pore network and high Cl binding ability.

Water transport was analyzed through the capillary suction coefficient ($k$). The water mass gained per unit surface is shown in Fig. 14. Similar water uptake was detected at early exposure time for all concretes with both types of nanoconstituents; the incorporation of neither alumina nanofibers nor nanocellulose crystals resulted in significant changes. Upon longer exposure times after reaching equilibrium, some higher water absorption was measured in the case of the cellulose nanocrystal addition mix, which is likely to confirm that CNC seems to absorb more water in the long term. The capillary suction coefficients showed the clear similarities between the concretes [Fig. 14(b)], with a slight reduction in the case of UHDC with nanoconstituents CNC and ANF, which most likely was correlated with the refined porosity, as was discussed with reference to Fig. 4. Electrical resistivity (${\rho }_e$) also was evaluated (Fig. 15). The UHDC-CA, UHDC-CA+ANF, and UHDC-CA+CNC mixes had resistivity values in the range $20–30\mathrm{\Omega m}$ at 2 months.

After 6 months of curing, an increase of the electrical resistivity occurred, which was related to the progress of hydration, with values ranging between 65 and 140 Ωm. Although limiting values of electrical resistivity between 50–100 and 100–250 Ωm indicate low and average durability, respectively (Wang et al. 2014), the obtained results should not be misleading about the actual level of durability of the investigated mixes with respect to the electrical parameter and compared with the total porosity (Fig. 4) and chloride transport properties (Fig. 13). A reasonable explanation may be the incorporation of steel fibers into the concrete. In addition, a small difference was observed at 6 months, where the concrete containing only the crystalline additive (UHDC-CA) had slightly lower resistivity, while the combination with alumina nanofibers and nanocellulose increased this electrical resistance, the refinement of pore structure at long hydration ages, shown in Fig. 4 can explain these results and the beneficial contribution of the nanoadditives that also contribute to reducing the microcracks, as shown in Fig. 9.