A. Morphological characterization of microscale reactions

To identify the chemical attributes of microscale reaction products formed when water is mixed with this model calcium silicate phase under ambient conditions, the monoclinic tricalcium silicate Ca₃SiO5 (C₃S), we conducted optical and microRaman spectroscopy (μRS) (see Methods in the SI) under two different conditions: high initial ratios of water-to-solid (w/s ∼ 100) and lower ratios typical of infrastructural cement pastes [w/s ∼ 0.5 (Ref. 22)]; see SI and Figs. S.1–S.4 for Methods. While calcium silicates used in cement clinkers can contain elemental impurities including Al and Fe, we chose C₃S (the predominant component) to minimize the range of chemically distinct reaction products. The findings from both microRaman and time-lapse optical studies were then compared to bulk macroscale cementitious systems in which the reaction interface between cement and water is not directly accessible. Hardened cement pastes with w/s ∼ 0.5 were hydrated for durations ranging between several hours to 1 wk.

Morphological changes and growth regimes upon exposure of tricalcium silicate to water were monitored via time-lapse optical microscopy [Fig. 1, see also Fig. S.3(a)], using an experimental technique developed by Grossier et al.^{Reference Grossier, Hammadi, Morin, Magnaldo and Vessler21} to visualize crystal nucleation, as described in Experimental Details and in SI. Note that this optical analysis provides a basis of visual and quantitative comparison with others' morphological characterization but does not chemically identify the observed products. Here, we tracked evolution of product reaction in picoliter-scale droplets of supersaturated, degassed aqueous solutions containing tricalcium silicate microparticles (details in SI). Reaction droplets were surrounded by paraffin oil to slow but not completely eliminate evaporation or gas exchange^{22,Reference Grossier and Van Vliet23} ; complete water evaporation occurred after 15 h. Image analysis of reaction product growth rates quantified interface motion away from the immersed particles [Fig. S.3(b); details in SI]. Figure 1(a) illustrates the growth of the hydration product interface, indicating projections extending radially such that growth fronts (red arrows) can be approximated by arcs. The measured growth rate of this interface, for C₃S particles within NaOH-containing droplets at pH 12, was ∼0.5 μm/h. In every droplet exhibiting byproduct growth, once the reacting particles were covered with this distinct phase, we then observed a transition in the growth front at later times. The stable, radial progression then became “finger-like,” dendritic projections of distinct optical contrast [Fig. 1(b) and Fig. S.8].

FIG. 1.

Evolution of the product growth front inside an aqueous microdroplet. (a) Evolution of the reaction growth front of a reaction byproduct during hydration of tricalcium silicate, by optical microscopy (scale bar: 5 μm). Calcium trisilicate particles, indicated by the red arrow, contribute to the growth of the product. Initially (in this case, from 1–5:20 h), the growth front can be approximated as a smooth arc. Eventually (in this case, after 5 h), the growth front breaks up to develop rough “finger-like” projections. (b) The increment in the radius of the byproduct is measured directly from the optical images, as described in the SI. The growth evolution follows a series of linear regimes (black lines: linear fits). The overall growth profile does not follow a power law, R ∝ t ^α with α ≤ 0.5 (shown as a red curve in the inset), which is typical of a diffusion-limited growth regime.^{22,Reference Grossier and Van Vliet23}

Particles exhibiting growth in Fig. 1 were covered within 6 h by the radially expanding reaction product; after this time, irregular projections initiated. This observation is consistent with a previous atomic force microscopy-based analysis of growth from lime-wetted tricalcium silicate pellet surfaces, in which Garrault et al. reported an initially flat precipitate (up to at least 4 h) that then transitioned to a less dense, porous topography; see Fig. S.5 and discussion.^{Reference Garrault, Finot, Lesniewska and Nonat8} Although those authors did not report the transition time, in the present experiments we found the initiation of the unstable growth front to vary among droplets, due ostensibly to particle geometry, interparticle spacing, and/or crystallographic orientation of the exposed facets. This morphological transition has been interpreted previously as a transition from interface-controlled to diffusion-driven kinetics.^{Reference Garrault, Finot, Lesniewska and Nonat8,Reference Severinghaus, Broecker, Dempster, MacCallum and Wahlen18}

To further characterize the growth kinetics and possible controlling growth regimes, we measured the radial evolution of the expanding byproduct fronts as a function of time (as described in SI and Fig. S.6). Diffusion-limited, circularly spreading growth fronts are typically well described by power laws R ∝ t ^α (α ≤ 0.5), whereas interface-controlled reaction kinetics are linear with time t.^{Reference Reuter, Gerth and Kirschner24,Reference Ferralis, El Gabaly, Schmid, Maboudian and Carraro25} Our measured growth evolution was initially linear [Fig. 1(b), inset], highlighting the interface-controlled kinetics of the early stage reaction.^{Reference Ferralis, El Gabaly, Schmid, Maboudian and Carraro25} The differing transitions from linear regimes among regions also emphasize the competition between the growth mechanisms at the reaction front with the variable diffusion rate of reagents at the source (tricalcium silicate) particle (Fig. S.7). The subsequent irregular projections resulted from partial dissolution of the tricalcium silicate and water evaporation [Fig. 1(a), Figs. S.7 and S.8].

B. Chemical characterization of microscale reactions indicates predominant carbonation

Detailed interpretations of these growth mechanisms should not rely on an a priori assumption that C-S-H is the major reaction byproduct. We thus tested the validity of the assumption that C-S-H is the predominant reaction product within these microscale reaction volumes (which would support mechanisms proposed by Juenger et al.^{Reference Juenger, Monteiro, Gartner and Denbeaux10}; see SI) by explicitly identifying the chemical composition via μRS. RS is a critical approach for the characterization of chemical and mineral species in complex multiscale composites such as cementitious materials. In the context of industrial composites, RS is usually used at advanced stages of hydration for bulk cement pastes,^{Reference Black26–Reference Potgieter-Vermaak, Potgieter and Van Grieken28} rather than the early stages of interest herein. (See SI for details and methods of RS experiments.) To our knowledge, this study is the first to leverage μRS to characterize the early stages of hydration and carbonation of calcium trisilicate for microscale reaction volumes. The key benefit of microRaman is comparable length scales of the probed volume and that of either the droplet (for microscale reaction volumes) or the near-surface region of a hydrating paste (for macroscale reaction volumes).

Figure 2 shows Raman spectra for tricalcium silicate particles hydrated for 2.5 h under the same conditions used to quantify morphological growth rates. The microscale interaction volume afforded spectral acquisition at several distinct locations. Based on the presence of very prominent peaks at 281 and 1085 cm⁻¹ (corresponding to calcite) and the absence of any spectral signature corresponding to C-S-H, we identified the main product to be crystalline calcium carbonate (CaCO₃), rather than C-S-H (Fig. 2). It is expected that the Raman spectral sensitivity to calcium carbonates is approximately 30 times greater than that of C-S-H [defined by a broad spectral shoulder originating from Q1 silicates at 600–700 cm⁻¹ (Ref. Reference Black26–Reference Potgieter-Vermaak, Potgieter and Van Grieken28)], so the co-existence of both carbonate and hydrate phases cannot be excluded entirely from this observation alone. Alternatively, C-S-H formation can be assessed through reduction in peak intensity corresponding to the silicates on the anhydrous calcium trisilicate (∼841 and 993 cm⁻¹), and the formation of other intermediary by-products (such as calcium hydroxide or portlandite). However, in the present experiments, no intensity reduction of the C₃S peak was observed, and a strong carbonate peak was present; this also supports the interpretation of these spectra as carbonate-dominated. We note that the observed calcium carbonate was found consistently and homogeneously throughout any point of the extending front that was observable via optical microscopy, and no by-products were detectable via microRaman signatures between the leading edges of the optically visible carbonate reaction fronts. Furthermore, within the time scale of early hydration prior to the acceleration period (2.5 h), significant formation of dense C-S-H by-products is not expected^{Reference Bullard, Jennings, Livingston, Nonat, Scherer, Schweitzer, Scrivener and Thomas5} and has not been observed previously.^{Reference Gallucci, Mathur and Scrivener29} Thus, despite the similar morphologies and growth rates of this product to previous reports of putative C-S-H,^{Reference Juenger, Monteiro, Gartner and Denbeaux10,Reference Grossier and Van Vliet23} we find that the early stage microscale product obtained in dilute, degassed aqueous suspensions under ambient conditions is not a hydration product but rather a carbonation product. We highlight that the observed calcium carbonate was formed within the hydrated volume, not ex situ upon exposure to higher ambient CO₂ concentrations, and the acquisition conditions of the Raman spectra were optimized to maintain the w/s ratio and minimally affect hydration (detailed in SI).

FIG. 2.

Raman spectra of system comprising C₃S particles in NaOH pH 12 after 2.5 h of hydration. MicroRaman characterization of the chemical nature of the reaction by-products of hydration of tricalcium silicate C₃S, for which growth rates were computed from time-lapse image analysis. Raman spectra of pristine C₃S exhibited typical peaks at ∼841 and 993 cm⁻¹. Raman spectra of the reaction byproduct of C₃S in NaOH pH 12 after 2.5 h of hydration were acquired at locations marked by a white cross in the inset optical image (scale bar: 5 μm). Spectral bands of the reaction by-products at 281 and 1085 cm⁻¹ were consistent of those of calcium carbonate, which is formed by CO₂ diffusing through the paraffin oil. Inset: Optical micrograph showing the radially symmetric by-products (white arrows), extending from irregularly shaped C₃S microparticles (yellow arrows). These products share the same growth morphology as to those in Fig. 1(a).

The rapid emergence of a stable crystalline form of calcium carbonate for microscale chemical reaction volumes (as probed with μRS in the pL droplet experiments as opposed to macroscale pastes) is in contrast with the formation of various surface carbonate phases in macroscopic pastes, from amorphous to vaterite and calcite.^{Reference Garbev, Stemmermann, Black, Breen, Yarwood and Gasharova15–Reference Dubina, Korat, Black, Strupi-Suput and Plank17} This points to a rapid carbonate formation at the microscale, induced by a rapidly changing solvent composition within the microscale chemical reaction volume and enhanced by water evaporation, that is not common or expected within the calcium silicate hydration that proceeds in typical macroscopic bulk pastes.

C. Time and length scale-dependent susceptibility to carbonation

While not the chief product of calcium silicate hydration in typical bulk synthesis of cement pastes or concrete used for structural applications, the formation of surface carbonate by-products is usually artificially induced under CO₂ rich hydration conditions, rather than normal ambient conditions. Therefore, one may question the relevance of hydration and carbonation in microreaction volumes in relation to standard bulk processing conditions. We note that these experiments in pL-scale confined environments differ from typical bulk processing of cement pastes in two key characteristics: (i) water/solid ratio, w/s; and (ii) surface area-to-volume ratio, s/v (i.e., microscopic scale reaction volumes). Given that carbonation occurs through dissolved CO₂ in the aqueous media over several hours of observation, one would expect the high w/s ratio in the droplets to be the principal cause of carbonation, and that larger scale experiments at similarly high w/s ratios should also promote carbonation. However, this was not the case: Fig. 3(a) shows no evidence of calcium carbonate from the Raman spectra of a cement suspension with 10⁸-fold greater mass of tricalcium silicate particles, as compared to pL-volume droplet experiments. Clearly, high w/s ratio is not the main driver toward extensive microscale carbonation.

FIG. 3.

Role of length scales of reaction volumes in carbonate formation at the early stages of tricalcium siliate (C₃S) hydration. (a) Raman spectra of a suspension of 500 mg of C₃S in 50 mL of water, with high water-to-solid ratio (w/s = 100) after 6 h of hydration: This paste had a 10⁸-fold increase in tricalcium silicate mass as compared to picoliter-volume droplet experiments. The absence of the band at 1085 cm⁻¹ indicates the absence of carbonates. (b) Raman spectra of the same amount of C₃S as in (a) after 6 h of hydration with uncontrolled evaporation of the aqueous media, but with a much lower water-to-solid ratio (w/s = 0.5) than that used in the picoliter-volume droplet experiment. Carbonation is clearly identified by the band at 1085 cm⁻¹. (c) The significantly higher intensity of the carbonate peak in the Raman spectra of the same hydrated C₃S in (b), but after laser-induced enhanced water evaporation, shows that enhanced evaporation strongly favor carbonate formation, even at larger reaction volumes.

It is known that carbonation is enhanced if the w/s ratio was reduced to approach levels typical of cement paste synthesis in excess of CO₂. Figure 3(b) shows that a cement paste of w/s ratio reduced to 0.5, mixed and then cured for 6 h in air with uncontrolled evaporation of aqueous media leads to formation of calcium carbonates. We infer that the change in surface area/volume ratio leads to a substantially different and dominating surface reaction dynamics, which cannot necessarily be assumed to remain the same as in bulk. Furthermore, the extreme sensitivity to carbonation for microscale reaction volumes, and the low levels of CO₂ needed (<0.038 vol%) to promote this, underscores the essential requirement of assessing in situ and in real time the dynamical evolution of the reaction product composition under realistic environmental conditions that include such species. Whenever such conditions cannot be satisfied (e.g., ex situ or in vacuo), only a static snapshot of the reaction conditions and by-products can be expected; such conditions can also modify the by-products via exposure to air and/or dehydration.

To further demonstrate the crucial role of water evaporation in the preferential formation of calcium carbonate, we artificially enhanced the formation of calcium carbonates by accelerating evaporation at ambient, constant (CO₂) via heating induced by the Raman excitation laser-probe at full power [Fig. 3(c), with details in SI]. We note that while a nonadiabatic environment within the micro reaction volume limits the quantification of heat transferred by the laser, optimization of the laser power levels between 1 and 50 mW/m² was sufficient to guarantee from none to rapid evaporation. Further, based on the kinetic analysis of growth mechanisms above, we posit that the mobile species within the aqueous media are the Ca²⁺ ions supplied from the C₃S particle. During interface-controlled growth, evaporation increases the local concentration of Ca²⁺ ions, due to both continued ion leaching from the tricalcium silicate particle and reduced water volume. Once a critical concentration of Ca²⁺ ions is reached at the microscale, cations are available for two competing reactions: hydration and/or carbonation. Our findings show that the thermodynamically preferred reaction is carbonation. Even assuming that hydration by-products could form under these conditions but that their spectral signature was to be below detection limits (due to the low Raman spectral sensitivity of C-S-H compared to carbonates), the very rapid reaction times (below 5 s) would not be favorable toward the much longer the formation time of C-S-H.^{Reference Bullard, Jennings, Livingston, Nonat, Scherer, Schweitzer, Scrivener and Thomas5,Reference Gallucci, Mathur and Scrivener29} From these results, we conclude that susceptibility of cementitious systems to carbonation is modulated strongly by the evaporation rate of aqueous media—and therefore by the locally high concentration of Ca²⁺ ions that can depend on reaction volume or calcium silicate volume—but not by the initial ratio of water to calcium silicates.

D. Mechanisms of carbonation, hydration acceleration, and suppression

Microscale understanding of the evolution of Ca²⁺ supersaturation induced by evaporation is essential in optimizing carbonate formation, whether enhancement or suppression is the goal. Such insight applies to both applications and mechanistic studies of interfacial growth of hydration products in cementitious materials. Furthermore, while it is typically desirable to promote hydration over carbonation in macroscale structures, in specific applications the capacity to control and therefore tune early stage carbonation at the microscale may facilitate certain hydrate/carbonate mixtures.^{Reference Rostami, Shao, Boyd and He14} Thus, we investigated potential chemical approaches to control and reduce surface carbonation during early stage hydration induced by enhanced evaporation, benefiting from ability to probe microscale volumes with μRS. We note that such approaches to modulate carbonation can also be explored empirically, for example in enabling chemical reactions that prevent Ca²⁺ and CO²⁻ ions from reacting.^{Reference Morse and Luttge30} However, μRS can elucidate the reasons why such processing conditions work.

For evaporation-influenced carbonation, aqueous media is the source of both Ca²⁺ and CO²⁻ ions. Use of an acidic aqueous medium could suppress carbonation by effectively increasing the solubility limit of calcium carbonate^{Reference Morse and Luttge30} without solvating the Ca²⁺ ions required of hydration products, such as C-S-H and Ca(OH)₂. We explored this possibility by comparing early-stage cement pastes hydrated with neutral deionized water and acidic HCl (pH 2) solution. Note that this strategy is most effective within the first ∼10 h of hydration when the paste is moist (i.e., damp but not immersed). Figure 4(a) compares Raman spectra of cement pastes of w/s = 0.5 using water or this acidic media and hydrated in a high relative humidity environment to minimize evaporation of aqueous media. Hydration spectra at 3 h showed early signs of carbonate formation for pastes initiated with water.

FIG. 4.

Carbonation dynamics through hydration acceleration and suppression. (a) Calcium carbonate formation induced by water evaporation is mitigated through the use of acidic aqueous medium, as indicated by the low intensity of the calcium carbonate band in the Raman spectra on the surface of C₃S pastes mixed an aqueous solution of HCl pH 2 after 3, 6, and 10 h of hydration. (b) The crucial role played by evaporation in carbonate formation is evident in the absence of CaCO₃ Raman band in the bulk of C₃S pastes mixed with deionized water and HCl pH 2 after 10 h of hydration. The intensity of the CaCO₃ band is instead rather significant at the surface of the same paste. (c) Reduction in surface carbonation and increase in paste hydration homogeneity are also achieved by extending the initial mixing time (5 and 30 min, respectively, with hydration times up to 75 h). (d) Schematic of surface carbonation mechanisms induced by the gradient in CO₂ concentration from the environment into a suspension or paste. Scale bar is approximate in this schematic.

We note that the carbonate signature at ∼1085 cm⁻¹ (Ref. Reference Black26) is broad compared to standard crystalline calcite, suggesting the structure is not fully crystalline or includes other carbonate polytypes such as vaterite [Fig. S.9, Refs. Reference Garbev, Stemmermann, Black, Breen, Yarwood and Gasharova15–Reference Dubina, Korat, Black, Strupi-Suput and Plank17]. The principal hydration product, low density C-S-H,^{Reference Gallucci, Mathur and Scrivener29} at very early stages of hydration may have a similar morphology as carbonation, underscoring the importance of chemical characterization of early stage reaction products. In contrast, pastes hydrated with acidic media exhibited a spectral band at ∼1085 cm⁻¹ of lower intensity after 3 h, indicating the presence of little or no carbonates; but spectra at later times up to 10 h in these samples showed comparatively sharper carbonation signatures. At these later time points, it is likely that acidity of the aqueous medium was not maintained sufficiently controllable to prevent the precipitation of calcium carbonate: continuous dissolution of tri-calcium silicate in the aqueous media increases the concentration of Ca²⁺ ions, eventually rendering acidity ineffective against carbonation. Thus, for aqueous media at initial pH 2, we find this critical time to be between 3 and 6 h.

Carbonation during early-stage hydration can also be suppressed, perhaps counter intuitively, via mechanical mixing that minimizes the exposure of given surface to air and thus to evaporation. These conditions have long been met empirically in macroscale processing, which typically includes copious addition of water at early stages and mechanical mixing of reacting suspensions. Figure 4(c) provides microscale analysis that elucidates one reason that this practice is beneficial: longer mixing times (30 min) reduced carbonate formation and promoted phase heterogeneity throughout the paste volume (probed through at least 20 μRS spectral scans), whereas brief mixing (5 min) in air promoted phase heterogeneity at only the paste surface. This observation supports the concept that mechanical mixing can overcome poor hydration conditions that would otherwise fail to homogenize the Ca²⁺ concentration throughout the macroscale volume itself and avoids imbalances due to longer exposures of a particular portion of the macroscale volume to air evaporation. Figure 4(d) summarizes this dependence of carbonation on CO₂ gradients such as those extending from an air-exposed surface of the reacting aqueous suspension.

E. Relating reaction mechanisms across time and length scales

These results for microscale reaction volumes, contrasted with macroscale volumes, shape the understanding of “surface” and “bulk” regions within macroscale, hydrating calcium silicate pastes. For a microscale particle, hydration conditions can deviate substantially from the expected “bulk” behavior, similarly to the surface of a macroscale sample. Hence, the microscale reaction volume can be considered a realistic representation of the paste surface, and more generally of microscale regions exposed to even ambient CO₂ concentrations. For example, the microscale reactions may favor carbonation at the bulk level at early-stage hydration of concrete with high degrees of air entrainment (i.e., many internal micropores, where the internal pore wall is exposed to the entrapped air during early-stage hydration).^{Reference Cebeci31,Reference Cultrone, Sebastian and Ortega Huertas32} In this way, it is not unexpected that the predominant reaction products differ as a function of length scale, as well as with time scale. For example, we did not observe significant spectral signature of hydration products such as Ca(OH)₂ and C-S-H in the Raman spectra of these paste surfaces 10 h after initial mixing, for addition of either neutral or acidified (HCl, pH 2) water. (See SI and Fig. S.9 for discussion of longer time scales.) Referencing previous studies on pastes with similar hydration conditions (particle size ∼ 50 μm and w/s ∼ 0.5), this hydration time point lies toward the end and rate-maximum of the hydration kinetics acceleration period.^{Reference Bullard, Jennings, Livingston, Nonat, Scherer, Schweitzer, Scrivener and Thomas5,Reference Berliner, Popovici, Herwig, Berliner, Jennings and Thomas33} Given that the acceleration period is associated with the nucleation and growth of the C-S-H phase, the observed lack of C-S-H signature can be attributed directly to a retardation of C-S-H formation and to surface carbonation suppressing the formation of hydration products.

At the macroscale, preferential surface carbonation is appreciated in practice. Here, direct chemical analysis of microscale regions within such macroscale samples also facilitates comparison through the thickness of macroscale cementitious pastes (details in SI), to elucidate the role of initial mix chemistry in speed and extent of carbonation at the surface versus the interior. Figure 4(b) compares the Raman spectra for the surface and bulk, in pastes hydrated for 10 h upon mixing with pure H₂O or HCl, pH 2. In spectra from bulk paste mixed with H₂O, we observed a sharp peak at 350 cm⁻¹ attributable to Ca(OH)₂, as well as two broad crests at 600–750 cm⁻¹ attributable to presence of both Ca(OH)₂ and C-S-H. In contrast, spectra from the bulk paste mixed with HCl exhibited a single broad crest centered approximately at 680 cm⁻¹. While this feature can be attributed to contributions from both C-S-H and Ca(OH)₂, given the absence of a Ca(OH)₂ peak at 350 cm⁻¹, we assign this broad feature to the C-S-H phase. Additionally, we note that the acidic medium can deter the precipitation of Ca(OH)₂. These results are consistent with the concept that the C-S-H phase nucleates and grows in the acceleration period.^{Reference Thomas, Jennings and Chen34,Reference Thomas, Biernacki, Bullard, Bishnoi, Dolado, Scherer and Luttge35} In sharp contrast to the bulk spectra, hydration products were not prominent spectral features of the paste surface [Figs. 4(a) and 4(b)] and therefore were substantially delayed compared to the bulk. This illustrates that surface carbonate formation in early stages of hydration for macroscale reaction volumes delays the formation of hydration products, such as Ca(OH)₂ and C-S-H, and supports an evaporation-based model of carbonate formation [Fig. 4(d)] in pastes hydrated for less than one day. In short, the rapid evaporation leads to predominant and fast carbonation reactions that compete with and limit the hydration reaction over long time scales, in particular for microscale reaction volumes comprising Ca₃SiO₅ and H₂O. This predominant carbonation in microscale reaction volumes is attributed to the fact that even subambient levels of CO₂ act as an important reactant in concert with dissolved Ca²⁺ ions in such volumes. That predominance of carbonation within small reaction volumes is of note, particularly in design of experiments that aim to probe the hydration processes relevant to the dominant reaction product in macroscale volumes.