In situ TEM imaging of CaCO3 nucleation reveals coexistence of direct and indirect pathways

See allHide authors and affiliations

Science  05 Sep 2014:
Vol. 345, Issue 6201, pp. 1158-1162
DOI: 10.1126/science.1254051

Watching nucleation pathways in calcite

The initial stage of crystallization, the formation of nuclei, is a critical process, but because of the length and time scales involved, is hard to observe. Nielsen et al. explored the crystallization of calcium carbonate, a well-studied material but one with multiple nucleation theories. Different calcium and carbonate solutions were mixed inside a fluid cell and imaged using a liquid cell inside a transmission electron microscope. Competing pathways operated during nucleation, with both the direct association of ions into nuclei from solution and the transformation of amorphous calcium carbonate into and between different crystalline polymorphs.

Science, this issue p. 1158


Mechanisms of nucleation from electrolyte solutions have been debated for more than a century. Recent discoveries of amorphous precursors and evidence for cluster aggregation and liquid-liquid separation contradict common assumptions of classical nucleation theory. Using in situ transmission electron microscopy (TEM) to explore calcium carbonate (CaCO3) nucleation in a cell that enables reagent mixing, we demonstrate that multiple nucleation pathways are simultaneously operative, including formation both directly from solution and indirectly through transformation of amorphous and crystalline precursors. However, an amorphous-to-calcite transformation is not observed. The behavior of amorphous calcium carbonate upon dissolution suggests that it encompasses a spectrum of structures, including liquids and solids. These observations of competing direct and indirect pathways are consistent with classical predictions, whereas the behavior of amorphous particles hints at an underlying commonality among recently proposed precursor-based mechanisms.

Nucleation is a key step in the crystallization process, representing the initial transformation of a disordered phase into an ordered one. It is also the most difficult part of the process to observe because it happens on very short time and length scales. In the case of electrolyte solutions, there is an open debate as to whether classical nucleation theory (CNT), as initially developed by Gibbs (1), is a suitable framework within which to describe the process, or whether nonclassical elements such as dense liquid phases (24) or (meta)stable clusters (5) play important roles. Furthermore, uncertainty exists as to whether a final, stable phase can nucleate directly from solution or whether it forms through a multistep, multiphase evolution (6, 7). In the case of multistep nucleation pathways, whether transformation from one phase to another occurs through nucleation of the more stable phase within the existing precursor or through dissolution of the original phase and reprecipitation of the secondary phase is unclear (8, 9). Although many studies have provided snapshots of the nucleation process (8) or followed the ensemble evolution of phases in solution (9), and simulations have produced predictions for certain solution conditions (4, 10), in situ observations that follow the process from start to finish have been lacking.

The advent of liquid cell transmission electron microscopy (TEM) (11) permits imaging with nanometer-scale spatial resolution in time increments of fractions of a second. We used a dual-inlet flow stage (fig. S1) for in situ observations of CaCO3 nucleation pathways over a range of solution conditions. The observations we report here elucidate the existence of a range of nucleation pathways occurring under identical or similar solution conditions, often simultaneously within a single experiment.

To introduce CaCl2 and NaHCO3 solutions of varying concentrations into the flow cell, we used a range of flow rates set independently for the two reagents (tables S1 and S2). In some experiments, supersaturations increased initially but eventually decreased until undersaturated so that dissolution was also observed. We recorded nucleation of both metastable and stable phases, including amorphous calcium carbonate (ACC), vaterite, aragonite, and calcite, typically exhibiting morphologies common for these phases (fig. S2). All nucleation events occurred on the top or bottom membrane of the fluid cell. The thickness of the fluid layer varied and sometimes thinned substantially during an experiment, facilitating collection of diffraction data for unambiguous phase identification. However, collection of diffraction data was not always possible, either because phase transformations took place while operating in imaging mode or because the solution layer thickness produced multiple scattering events that degraded the diffraction signal beyond use.

Amorphous calcium carbonate particles nucleated (Fig. 1, A and B, and movie S1) and grew to diameters of up to hundreds of nanometers (Fig. 1, C and D). Diffraction data confirmed the amorphous nature of the particles (Fig. 1D, inset). Similarly for vaterite, nucleation (Fig. 1, E and F, and movie S2) was followed by extensive growth (Fig. 1G); some crystals merged to form larger crystals (Fig. 1H) with visible texture (Fig. 1I). In some cases, at later stages of growth the growing outer edge exhibited a layered structure while the interior of the growing platelet dissolved away (Fig. 1J). The diffraction pattern (Fig. 1J, inset) identified the phase as vaterite.

Fig. 1 Direct formation of ACC and vaterite.

Frames from movie S1 show the fluid cell before nucleation (A) and during nucleation and growth of ACC (B to D). Diffraction analysis [inset to (D)], performed when the fluid layer thinned (see supplementary materials), confirms the amorphous nature of the particles. Frames from movie S2 follow vaterite formation and growth (E to J). Gray spots already present (E) are salt deposits that formed on the outer surface of the liquid cell window during cell assembly. In (E) and (F), the nucleation site of a vaterite particle is circled for clarity. The particle grows (G), merges with a second particle (H), and exhibits layering at the growth front and dissolution in the center (I and J). Diffraction analysis [inset to (J)] identifies the material as vaterite. Scale bars are 500 nm in (A) to (J) and 2 nm–1 in the insets to (D) and (J). Solution conditions—designated in all figure legends by [CaCl2]:[NaHCO3]/R(CaCl2):R(NaHCO3) with concentrations in millimolar and flow rates R in microliters per minute—are 50:50/10:0.2 for (A) to (D) and 40:40/9:1 for (E) to (J). Times listed in all figures are relative to the beginning of the associated movie.

We also observed multistep nucleation pathways starting with ACC (Fig. 2, and movies S3 to S5). ACC particles formed and grew to sizes ranging from hundreds of nanometers to micrometers (Fig. 2, A and E) before suddenly transforming to the aragonite “sheaf-of-wheat” morphology (Fig. 2, C and D) or vaterite (Fig. 2, F to H). Typically, the ACC particle began to shrink just before the appearance of a secondary phase (Fig. 2, B and F) on, or possibly just below, the surface of the original particle. This shrinkage perhaps indicates either the expulsion of water from the amorphous particle or a sudden decrease in concentration leading to partial dissolution. This secondary phase grew rapidly, consuming the original amorphous particle (Fig. 2, C, D, G, and H). The two phases maintained constant physical contact throughout this transformation process. Because the surrounding medium is supersaturated with respect to the secondary phase, growth presumably also involves monomer addition from solution.

Fig. 2 Direct transformation of ACC to crystalline phases.

Frames from movie S3 show a previously nucleated ACC particle (A), with the secondarily nucleated crystalline phase forming on or in the amorphous particle (B). The secondary phase, exhibiting typical aragonitic sheaf-of-wheat morphology, grows at the expense of the ACC, with the two phases maintaining physical contact during the entire transformation (C and D). Frames from movie S5 show a previously nucleated ACC particle (E), with secondarily nucleated vaterite plates forming on or in the amorphous particle (F). These plates grow at the expense of the ACC (G and H), in the same manner as above (C and D). Diffraction from the resulting plates identifies them as vaterite [inset to (H)]. Scale bars are 500 nm in (A) to (H) and 2 nm–1 in the inset to (H). Solution conditions are 30:100/10:0.2 for (A) to (H).

These two examples demonstrate the occurrence of multistep pathways of CaCO3 crystal nucleation by which a metastable, amorphous precursor appears first and then transforms into a more energetically favorable crystalline phase through a direct, physical connection between the growing and shrinking phases. These multistep pathways contrast with direct pathways in which a crystalline phase nucleates from solution, either in the absence of ACC or independently of any amorphous particles that may have already formed. In this latter case, which is well documented (9, 1214), the nucleation of the crystalline phase is followed by dissolution of preexisting amorphous particles and reprecipitation onto the crystalline phase, as inferred from in situ optical (12, 13) and x-ray studies (9). However, among the hundreds of experiments conducted, transformation of ACC into calcite (the most stable phase of CaCO3) was never observed.

Additionally, we detected concurrent nucleation of multiple phases. For example, direct nucleation of calcite rhombohedra (Fig. 3, A to D, and movie S6) was observed alongside the formation of a (hemi-)spherical particle that, based on morphology, was either ACC or vaterite. Numerous optical studies have shown that ACC dissolves in the presence of calcite (1214). Consequently, although the particle lacked any visible internal structure, as was typically seen in vaterite (Figs. 2, H to J, and 3, H and I), it is unlikely to be ACC.

Fig. 3 Concurrent formation of multiple phases.

Frames from movie S6 show simultaneous nucleation and growth of calcite crystals and either ACC or vaterite (A to D). The nitride window edge is visible at the bottom right corner of each panel. Frames from movie S7 show direct nucleation of vaterite (E) and aragonite (F) and subsequent growth (G and H). The formation of calcite occurs on aragonite (I), followed by calcite growth and concomitant dissolution of the aragonitic bundle (J to L). Scale bars are 500 nm in all panels. Solution conditions are 30:30/8:2 for (A) to (D) and 50:50/10:0.2 for (E) to (H).

Nucleation of multiple phases, followed by transformation to secondary phases, was also observed in a single experiment (Fig. 3, E to L, and movie S7). For example, a crystal with vateritic morphology first formed (Fig. 3, E to I) and continued to grow as bundles with aragonitic morphology formed in the vicinity and merged into larger aggregates (Fig. 3, F to J). A calcite rhombohedron then nucleated in apparent contact with the aragonitic bundle (Fig. 3I) and grew throughout the rest of the experiment (Fig. 3, J to L) as the aragonitic bundle dissolved (Fig. 3, K and L). We were unable to collect diffraction information to unequivocally assign phases in this experiment, leading to some ambiguity. Furthermore, because the image is a two-dimensional (2D) projection of a 3D volume, there is some uncertainty as to the precise location of the calcite nucleus relative to the surface of the bundle.

Contrary to our expectations, on the time scale of our experiments, ACC nucleation occurred only when the solution was exposed to the electron beam under sufficiently high solution concentrations (movie S8). Moreover, varying the electron dose accelerated, delayed, or prevented its formation (movies S9 and S10). By contrast, none of the crystalline phases showed such a relation to the electron beam; crystals were regularly found far from areas exposed to the beam.

We measured growth rates for ACC (fig. S6A), calcite (fig. S6B), and vaterite (fig. S6C) in 14 experiments, tracking observations of both single and multiple particles. All phases grew linearly, indicating that postnucleation growth occurred under steady-state solution conditions and was controlled by surface kinetics rather than diffusive transport (15). Although some growth curves exhibited single linear trends, others showed two distinct linear regions. This change in slope marked a drop in the growth rate and, therefore, supersaturation, suggesting a reduced rate of solute input to the cell, perhaps due to CaCO3 nucleation near the cell inlet.

Finally, we observed multiple distinct dissolution behaviors for ACC particles under continued illumination by the electron beam, following thinning of the liquid layer. Some ACC particles underwent uniform shrinking to the point of complete disappearance, behaving as if they were liquid droplets evaporating into the surrounding medium (Fig. 4, A to F, and movie S11). By contrast, other ACC particles at a later time in the same experiment exhibited behavior indicative of a dissolving solid, becoming rough and pitted over time (Figs. 4, G to L, and movie S12) before finally disappearing. Others exhibited behavior combining or intermediate to these two end points.

Fig. 4 Two dissolution behaviors of ACC.

Frames from movie S11 show that some ACC particles undergo liquid-like shrinking and disappearance (A to F). At a later time in the same experiment, frames from movie S12 depict nearby ACC particles exhibiting solid-like behavior, becoming pitted and developing roughened edges (G to L) while still amorphous [inset to (L)]. Scale bars are 500 nm in (A) to (L) and 2 nm–1 for the inset to (L). Solution conditions are 100:100/10:0.2 for all panels.

The open questions about nucleation from electrolyte solutions are especially germane to the CaCO3 system. Numerous studies have concluded that the final crystalline state often emerges long after the first appearance of ACC (5, 6, 9, 14, 1618). Physical chemical analyses (5) and cryogenic TEM (18) have given rise to a model of nucleation in which stable multi-ion clusters aggregate to form this first amorphous phase, which then transforms directly to the crystalline phases. Based on x-ray diffraction and optical microscopy, other studies have concluded that nucleation is described well by CNT, and crystalline phases that appear after ACC do so through dissolution and reprecipitation (9, 12). Ex situ TEM (19) and nuclear magnetic resonance data (3) indicate the existence of a dense liquid phase. Molecular dynamics simulations predict polymeric clusters (10), and dense liquid phases (4) form through spinodal decomposition with ACC then forming through partial dehydration. However, the existence of these phenomena remains unproven. Moreover, spectroscopic analyses have shown that many CaCO3-based biominerals form through aggregation and crystallization of ACC exhibiting multiple hydrated states (20, 21). The spatial resolution of the experiments described here does not allow us insight into the question of whether nucleation occurs via ion-by-ion attachment, as per CNT, or whether stable or metastable clusters serve as the primary species of addition. Our observations do, however, demonstrate that multiple nucleation pathways exist for the crystalline phases of CaCO3, including both direct formation from solution and direct transformation from more disordered phases. Moreover, our findings show that these multiple pathways and phases are simultaneously available to the system at moderate to high supersaturations, as fully expected from classical considerations.

Our results also shed light on the process by which the disordered phases transform to the more ordered phase. In all cases for which we can definitively identify the starting point of the secondary nucleus, it lies approximately at the surface of the parent particle. This is consistent with previous in situ TEM observations of solidification in liquid Au72Ge28 droplets where the first ordered domain appeared at the surface (22). Presumably, the higher mobility of surface ions and, in the case of solutions, their ability to rapidly exchange with the solution lead to this phenomenon.

Because the experimental cells used in these experiments have fluid layer thicknesses ranging from hundreds of nanometers to micrometers, effects of confinement might be expected. However, over the range of sizes observed here, the lateral growth rates remain constant and thus do not appear to be affected by the cell dimensions. Effects of confinement on nucleation could be evident in two ways: The first is through a similarity in cell dimensions to critical nucleus size (23), and the second is by structuring of the liquid layer through proximity to the cell membranes (24, 25). However, the cell dimensions are orders of magnitude above the ~1- to 10-nm critical sizes of the crystalline phases (12), as well as the ~10 Å thickness of the hydration layers (24, 25).

The findings reported here also bear upon the controversy concerning the nature of ACC. Initially, a single amorphous phase was reported (26). Later experiments demonstrated the existence of both hydrous ACC and anhydrous ACC (27). Other research suggested the existence of two forms of hydrous ACC (28), as well as proto-vateritic ACC and proto-calcitic ACC (29), with each serving as a precursor to the respective crystalline phase. The dense liquid phase referred to above was recently proposed as yet another amorphous form. Our results call into question whether these are fundamentally distinct phases or whether they exist as points on a continuum. Though certainly not conclusive, the disparate modes of ACC dissolution observed in our study suggest that the term ACC refers to a spectrum of structures ranging from the dense liquid phase to the anhydrous form, rather than a single or even a few closely related structures. Finally, whereas our results clearly show that direct transformation of ACC to the crystalline phase of CaCO3 readily occurs, we confirm previous suggestions from low-resolution optical measurements, macroscopic x-ray diffraction data (1214), and x-ray microscopy (30) that direct transformation from ACC to calcite is unlikely. Indeed, this formation pathway has never been directly observed.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S2

References (3133)

Movies S1 to S12

References and Notes

  1. Acknowledgments: We thank V. Altoe for the use of and assistance with the JEOL-2100F transmission electron microscope. This research was supported by the U.S. Department of Energy (DOE), Office of Basic Energy Sciences, Division of Chemical Sciences, Geosciences, and Biosciences at Pacific Northwest National Laboratory and Lawrence Berkeley National Laboratory. TEM was performed at the Molecular Foundry, Lawrence Berkeley National Laboratory, which is supported by the Office of Basic Energy Sciences, Scientific User Facilities Division. M.H.N. acknowledges support awarded by the U.S. Department of Defense, Air Force Office of Scientific Research, National Defense Science and Engineering Graduate Fellowship 32 CFR 168a, and the NSF under grant DMR-1312697. Pacific Northwest National Laboratory is operated by Battelle Memorial Institute for the DOE under contract DE-AC05-76RL01830.
View Abstract

Navigate This Article