Dissolution–Reprecipitation of Amorphous and Crystalline (Ca)–Mg-Carbonates in Thin Water Films

Abstract

Ca–Mg-carbonate minerals crystallized from amorphous precursors have complex morphologies and unique material properties. The transformation of amorphous Ca-, Ca–Mg-, and Mg-carbonate (ACC, ACMC, and AMC) phases is sensitive to relative humidity (RH) of the surrounding air. Here, we measure the effective thicknesses of water films on amorphous carbonates from 10 to 98% RH using a dynamic vapor sorption analyzer and specific surface areas from N₂ adsorption. We find that water sorption increases exponentially starting at 76–84% RH for ACMC phases (21 ≤ mol % Mg ≤ 100). We supplement these data with static sorption experiments at 98% RH and constrain sorption rates and effective equilibrium water film thicknesses to 2–6 nm. In long-term experiments with fixed humidity, ACMC and its transformation products show evidence of continued recrystallization over a period of seven months at 98% RH. Diverse crystal morphologies suggest distinct microenvironments of recrystallization. Finally, recrystallization of mixed Ca–Mg-carbonate phases to discrete Ca- and Mg-carbonate minerals suggests a process of ion transport in the thin films. We discuss how Ostwald ripening, phase metastability, and ion diffusion may operate as drivers of recrystallization in thin film environments.

1. Introduction

Over the past few decades, amorphous carbonate nanoparticles have garnered attention due to their importance in the biotic precipitation of carbonate minerals (e.g., gastropod, mollusk, arthropod shells) and the role that they play in lowering the energetic demands for crystallization. − The precipitation of amorphous phases requires a lower activation energy than their crystalline equivalents; thus, amorphous phases can precipitate more readily at low-temperature conditions despite requiring higher ion activity products. , These transiently stable precursors then undergo stepwise transformation through metastable crystalline intermediaries, until a stable crystalline phase is reachedtypically following Ostwald’s step rule. Crystallization via amorphous nanoparticles may offer several benefits that can be exploited during the development of functional nanomaterials. First, the isotropic nature of the amorphous spherical nanoparticles results in increased mechanical strength compared to crystalline materials; second, their lack of long-range order may result in increased scavenging of trace metals during precipitation, and coupled with their greater solubility, they likely act as dense-storage phases for essential ions that can be readily dissolved. − Finally, they are highly moldable and are linked to the crystallization of diverse and complex hierarchical structures. These attributes have resulted in increasing interest in studying the material properties and transformation mechanisms of amorphous carbonates for use as functional nanomaterials across a wide array of industries (e.g., pharmaceutical, medical, generation of sorbents) and for creating low-energy or “green” production pathways in the synthesis of other nanomaterials.

Although reaction pathways and transformation mechanisms of amorphous carbonate nanoparticles are well documented in bulk aqueous solutions including in the presence of various additives (e.g., SO₄ ^2– and organic macromolecules), − there has been less research on amorphous-to-crystalline phase transformations in air or under more water-limited conditions. Given that soluble salts (including hydrated sulfates and calcite) are known to be hygroscopic and may undergo dissolution–reprecipitation in humid air, , it is likely that highly soluble amorphous carbonate phases do too. Previous work examining the interaction of sorbed water on calcite surfaces shows that calcite recrystallization is facilitated by increasing relative humidity (RH), and that cation impurities within the starting material affect the extent of humidity-driven topography changes. Konrad et al. showed that amorphous Ca-carbonate (ACC) crystallization is dependent on the RH of the atmosphere in which it is reacted or stored. Additional work by Asta et al. showed faster ACC transformation and increasing crystal size of calcite with increasing RH. Ion transport and carbonate mineral precipitation in thin films of adsorbed water have also been demonstrated during carbonation of forsterite (Mg₂SiO₄) using supercritical CO₂, , highlighting the industrial importance of understanding how thin water films facilitate geochemical reactions.

In our previous study, we incubated amorphous Ca–Mg-carbonates (ACMCs) and amorphous Mg-carbonate (AMC) with 16–100 mol % Mg at six different relative humidities between 5 and 98% for up to 210 days to understand the effect of RH on the rate and pathways of phase transformations. We found that conditions of <75% RH lead to enhanced stability of amorphous phases and that a threshold between 75 and 100% RH accelerates the amorphous-to-crystalline transformation in air. However, the thickness of the water film that controls this amorphous-to-crystalline phase transformation was not measured and therefore the mechanism behind humidity-driven phase transformations could not be determined. In the present study, we better constrain the extent of H₂O physisorption on ACMCs with 22–100 mol % Mg at humidity conditions of 10–98% using a dynamic vapor sorption (DVS) analyzer and specific surface areas from N₂ adsorption. We supplement our previous study with extensive characterization of the morphological changes undergone by ACMC and AMC phases (16–100 mol % Mg) incubated at 98% RH for 210 days (experiments previously described by Patel et al.) to better understand (re)­crystallization mechanisms realized by exposure to humid air. We determine whether the effective thickness of thin films of water at ACMC surfaces is dependent on RH and whether there is a threshold effective thickness of water required for crystallization. Based on our observations, we discuss the geochemical processes that may underpin recrystallization of complex assemblages of Ca–Mg-carbonates as facilitated by thin films of sorbed water.

2. Methods

2.1. Synthesis of Amorphous Ca–Mg-Carbonates

All amorphous Ca–Mg-carbonates were synthesized following a modified method from Purgstaller et al. Equimolar solutions of CaCl₂ and MgCl₂ were mixed to generate solutions with a [Ca + Mg] concentration of 0.25 M and with [Mg]/[Ca] + [Mg] mol % values of 32.0, 72.1, 89.1, and 100%. The rapid mixing of 80 mL of the Ca + Mg solution with 80 mL of 0.25 M Na₂CO₃ resulted in the immediate precipitation of a gel-like suspension. Solids were separated from the fluids by centrifuging at 7000 rpm. The solids were rehomogenized with deionized water and centrifuged for 30 s at 7000 rpm to prevent precipitation of halite (NaCl) as a drying artifact (due to its water adsorbing properties), and this process was repeated twice. The solids were then rehomogenized with 100% ethanol to drive off surface water. The ethanol–solid suspension was then separated by vacuum filtration using a 0.22 μm filter. The solids were left to dry at 22 °C and an RH of 40% overnight. A first set of ACMCs (the “ACMC-reactor” samples) were held at constant RH for up to 210 d in reactors that are described in detail by Patel et al. A duplicate set of ACMCs (“ACMC-film_quant” samples) was synthesized to quantify water film thickness on ACMCs and to compare their behavior to nontransformed ACMC-reactor materials used in the long-term humidity experiment. The film_quant amorphous Ca–Mg-carbonates were synthesized with the same method as the reactor ACMCs, but three cycles of rinsing with 4 °C deionized (DI) water and centrifugation were completed during separation of the solids to maximize the removal of NaCl. All synthesis products are reported as “ACMC” or “AMC” followed by a number that refers to the [Mg]/[Mg] + [Ca] mol % in the solids.

2.2. Compositional Analysis of ACMC Materials

Thermogravimetric analysis (TGA) of the ACMCs used in the long-term experiments was conducted at the University of Alberta on a Discovery Thermogravimetric analyzer (TA Instruments). Approximately 3 mg of sample was loaded into a platinum-HT sample pan and was heated at a rate of 10 °C/min from an ambient temperature to 1000 °C under a N₂ gas atmosphere with a flow rate of 20 mL/min. The TGA analyses of the solids used for water film quantification were conducted at the University of Bern using a PerkinElmer TGA4000 connected to a Spectrum2 Fourier transform infrared spectrometer (FTIR) with a transfer line at 270 °C. Approximately 3 mg of material was added to a ceramic crucible and exposed to temperatures of 30–850 °C at a rate of 10 °C/min and a N₂ flow rate of 30 mL/min. The gases generated during decomposition were analyzed by FTIR spectroscopy. Spectra were collected continuously with a resolution of 4 cm^–1. Weight loss curves were analyzed using the Trios software package (TA Instruments) or the Pyris software package (PerkinElmer), and the molar fractions of H₂O and OH^– were calculated using the equation from Jensen et al. Additional FTIR spectra of the solid samples were collected on a PerkinElmer Spectrum2 FTIR by attenuated total reflection with a diamond crystal. Spectra were collected from 450 to 4000 cm^–1 with a resolution of 4 cm^–1 and 4 scans accumulated per analysis. Background spectra were collected regularly.

The concentrations of major cations present in the synthetic amorphous Ca–Mg-carbonates were obtained using a Thermo iCAP6300 Duo inductively coupled plasma-optical emission spectrometer (ICP-OES) at the Natural Resources Analytical Laboratory, University of Alberta. Solids were digested with trace metal grade HNO₃ at 25 °C and diluted 20× prior to analysis. Limits of detection and limits of quantification are reported in Tables S2 and S3 (Supporting Information).

2.3. Long-Term Humidity Reactor Experiment

Initially, amorphous carbonates of compositions ACMC16, ACMC32, ACMC45, ACMC80, and AMC100 were reacted for between 14 and 210 d at room temperature (∼22 °C) under RH conditions of 5–98% as described by Patel et al. Here, we focus only on samples that were exposed to 75 or 98% RH because they crystallized in humid air, whereas ACMCs exposed to RH < 75% remained amorphous. Stable headspace relative humidities in the reactors were generated by placing a saturated salt solution of NaCl (75% RH) or H₂O_(l) (98% RH) at the base of airtight plastic reactors. Aqueous salt solutions are known to be effective at stabilizing RH conditions on long time frames if the solutions remain saturated with respect to the salt, a constant temperature is used, and closed reactor conditions are maintained. , Importantly, in our experimental setup, the fluids were physically isolated from the amorphous Ca–Mg-carbonates by nesting sample cups in the base of the slurry or water, thereby ensuring that the only contact the solids had with the fluid was through the vapor phase and the resulting adsorbed water film. It should be noted that the RH in the water reactor was likely 100% rather than 98%; the hygrometer used in our experiments is only accurate within 10–95% RH with an accuracy of ±4% RH at elevated humidity. Further details of reaction conditions and setup, including other reaction conditions not included in this study, are described by Patel et al.

2.4. Mineralogical Identification

After samples were collected from the humidity reactors, they were mounted for imaging using scanning electron microscopy (SEM) on carbon tape and coated with ∼10 nm of gold. Imaging was done using a Zeiss Sigma 300 Field Emission Variable Pressure Scanning Electron Microscope (SEM) at a voltage of 3 kV at the University of Alberta. Energy-dispersive spectroscopy was done at a voltage of 10 kV for semiquantitative determination of elemental compositions.

Solids were homogenized with 100% ethanol using an agate pestle and mortar and mounted onto zero-diffraction quartz plates for X-ray diffraction (XRD) analysis. XRD data were collected at the University of Alberta on a Bruker D8 Advance θ–θ powder X-ray diffractometer equipped with a LYNXEYE XE-T 1D position sensitive detector and a Co X-ray source operating under conditions of 35 kV and 40 mA. Data were acquired across a range of 3–80° 2θ using a step size of 0.01°/step at a rate of 2 s/step. The DIFFRAC.EVA software (Bruker) with reference to the Powder Diffraction File Database (PDF 4+) from the International Center for Diffraction Data was used to qualitatively identify crystalline phases. Quantitative mineral abundances were determined by Rietveld refinement − using the Topas 5 software (Bruker) and following the fundamental parameters approach. Unit cell parameters were refined within prescribed minimum and maximum values to determine physically meaningful unit cell dimensions. Microstrain and crystallite size were determined using the Double Voight Approach. Poorly ordered or no known crystal structure (PONKCS) models were used for phases with unknown or variable unit cell parameters, such as dypingite [Mg₅(CO₃)₄(OH)₂·∼5H₂O]. A hydromagnesite-like phase [Mg₅(CO₃)₄(OH)₂·xH₂O], which had similar peak positions to hydromagnesite [Mg₅(CO₃)₄(OH)₂·4H₂O] but showed a shift in the d ₁₀₀ peak to d = 11.28–11.75 Å, was modeled by modifying the a-cell parameter of the hydromagnesite structure. Please refer to Patel et al. for a more detailed account of the refinement procedure and the complete quantitative data set.

2.5. Constraining Water Sorption onto ACMC Surfaces

Surface area analysis of the products synthesized for dynamic water vapor sorption analysis was measured on a MicroTrac Belsorp Mini X at the University of Bern following the BET N₂ technique. Seven adsorption points were used to define the sorption isotherm. Water vapor sorption analysis was completed on a Discovery SA DVS Analyzer (TA Instruments) at the University of Alberta using three ACMCs synthesized for the long-term reactor experiments and then stored in a closed container with drierite (∼10% RH) for 3 years and the fresh set of ACMCs synthesized specifically for water film quantification. Note that ACMC16-reactor and ACMC32-reactor are excluded from DVS analysis as these materials partially crystallized during storage. Approximately 3 mg of ACMC was placed into metallized quartz pans and loaded into the instrument operating with a 20 mL/min flow rate of a 100% N₂ gas. Isothermal humidity analyses were run at 98% RH and 25 °C for 24 h, with each run preceded by a 1 h isothermal equilibration period at 25 °C and 10% RH. Humidity ramp analyses were collected from 10 to 98% RH with a ramp rate of 0.2% RH/min after a 1 h isothermal equilibration period at 25 °C and 10% RH. Recorded data include time (min), moisture content (g water/g sample), weight (mg), RH (% RH), and weight % change (compared to the weight of material at t ₀). Data were processed using the Trios software analysis package, which includes sorption kinetics analysis and the application of derivative functions.

Effective water film thicknesses were calculated assuming a uniform, homogeneous water film thickness spread over a flat surface. Given the relationship between the mass of the water film, the density of the water, and the volume of the water, we calculated the effective water film thickness (f _eff) using the measured BET surface area (S), the mass of the sample (m), the density of water at 25 °C (ρ), and the weight of the water film (w _water film) based on eq from Lin et al.

$$f_{\mathrm{e}\mathrm{f}\mathrm{f}}=\frac{w_{\mathrm{w}\mathrm{a}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{f}\mathrm{i}\mathrm{l}\mathrm{m}}}{S\times m\times \rho }$$ 1

It should be noted that the use of this equation provides an effective water film thickness rather than a true measure of water film thickness. The underlying assumption made in using eq is that a continuous water film of uniform thickness completely covers the surfaces of mineral grains. This is an approximation of what is almost certainly a variably thick, discontinuous water film affected by properties such as mineral surface chemistry, surface roughness, internal porosity, particle aggregation, and capillary condensation in pores and at grain contacts. −

3. Results

3.1. Characterization of Initial Solids

XRD patterns of all ACMC samples showed no distinct Bragg peaks, which indicates a lack of long-range order and hence a lack of crystallinity. Scanning electron micrographs show amorphous nanoparticles in aggregates ranging from 50 to 200 nm (Figure S1), and no crystalline materials were present. The syntheses for the long-term experiments produced the following ACMC compositions: ACMC16, ACMC32, ACMC45, ACMC80, and AMC100, where the number refers to the amount of Mg present as a mol % of total Mg + Ca. TGA analyses of the mass loss between 25 and 250 °C showed that ACMCs contain between 1.7 and 2.2 pfu H₂O; the mass loss step between 250 and 800 °C indicated that ACMC45, ACMC80, and AMC100 comprised 0.02–0.78 pfu OH^– (Table S1; Figure S2). The syntheses for water film quantification produced the following ACMC compositions: ACMC21, ACMC23, ACMC47, ACMC80, and AMC100 (Table S1). Supporting Information includes ICP-OES data, TGA-FTIR data and interpretation, and scanning electron micrographs of the ACMC solids.

All synthesized ACMCs had BET surface areas between 30.4 and 45.0 m²/g, and there was no relationship evident between the surface area of the amorphous precursor materials and the Mg content in the materials (Table S1). The surface areas of ACMCs synthesized for the long-term humidity experiment show systematically smaller surface areas after three years of storage than ACMCs synthesized for thin film quantification.

3.2. Water Sorption from 10 to 98% Relative Humidity

A DVS analysis was conducted on ACMCs in a ramp from 10 to 98% RH to quantify the rapid sorption of water to ACMC. During the humidity ramp, ACMCs show similar sorption characteristics as a function of increasing humidity, with an initially slow increase in water sorption from 10% to 75% RH followed by a faster increase of water sorption up to 98% RH (Figure A). Compositions of ACMC with <50 mol % Mg and AMC100 (100% Mg) had a 2–4% weight increase during the initial period of slow water sorption, whereas ACMC80 showed less water sorption across the same range of humidities (1% increase). Water sorption at 75–98% RH resulted in weight increases between 5.5 and 9.3%. By the end of the humidity ramp, at 98% RH, ACMC78-film_quant and AMC100-film_quant had the largest increase in water mass. Notably, there was good agreement between the water sorption behavior of ACMC80 and AMC100 materials synthesized for long-term reactor experiments and film quantification, and overall, the total increase in water adsorbed onto ACMC47-film_quant and ACMC45-reactor was similar. This suggests that despite the use of different batches of synthetic solids, the samples used in the water sorption analysis represented the materials used in the long-term humidity reactor experiments well. There was no clear influence of the Mg content on the sorption behavior (Figures A and S9) or the total amount of water sorbed during the humidity ramp (Supporting Information Section 1.3.1).

1.

Data from DVS analysis of ACMC materials synthesized for humidity reactor experiments (reactor) and for film quantification (film_quant). (A,B) Materials were exposed to humidity conditions increasing from 10 to 98% RH at a rate of 0.2% RH/min. (A) The weight change (%) compared to the weight recorded immediately after the 1 h isothermal phase (t ₀). (B) The effective water film thickness across the humidity range calculated using eq ; the gray box present on this graph reflects previously reported thresholds of accelerated movement of ions in thin film environments using data from Tokunaga et al. (C, D) Weight change (%) corrected to t ₀ (60 min) for experimental runs conducted at 98% RH for 1500 min (approximately 24 h) on a DVS analyzer. Values reporting the exact weight change (wt %) for 13 representative timeframes are included in Table S4. The effective water film thickness (nm) is corrected to t ₀ (after 60 min equilibration period) for experimental runs conducted at 98% RH for 1500 min on a DVS analyzer. Note that data for AMC100-reactor show anomalous sorption behavior during fixed humidity tests compared to the equivalent AMC100-film_quant and ACMC phases and are considered outliers.

The effective thicknesses of water sorbed to ACMCs and AMC were calculated with eq using BET surface area data (m²/g), water mass (g water/g sample) from DVS analysis, and the density of water (assumed to be 1000 kg/m³) and are plotted against RH in Figure B. We observe that for all ACMC and AMC compositions, the effective water film thickness was between 0 and 1 nm at RH conditions of <60%. From 60 to 98% RH, the effective water film thickness increased to 1.0–1.6 nm. By 98% RH, the effective thicknesses of water films were between 2.0 and 3.0 nm for all compositions and there was good agreement between different sets of ACMC syntheses (Figure ).

3.3. Constant Humidity Sorption Analysis at 98% Relative Humidity

When held at an RH of 98% in a DVS analyzer, ACMCs of all compositions and AMC showed a sharp increase in weight due to sorption of water during the first 200 min of exposure (Figure C). For both synthesis batches, AMC showed the largest increase in weight as a result of water sorption with 16% for AMC100-film_quant and a 23% increase for AMC100-reactor after 200 min (Figure C, Table S4). The AMC100-reactor sample shows anomalously high water sorption when compared to AMC100-film_quant and other ACMC phases; thus, we consider this to be an outlier. ACMCs of all compositions showed a weight change of 9–15% within the first 200 min, and there was no clear relationship between the Mg content and water sorbed (Table S4). Further description, including a detailed comparison of sorption rates for ACMCs and AMC, can be found in the Supporting Information.

In ACMCs synthesized for water film quantification, the effective water film thicknesses were between 2 and 4 nm for the entirety of the experiments but were thickest between 150 and 250 min of reaction (Figure D). By the end of the experiments, ACMC22-film_quant and AMC100-film_quant had the greatest effective water film thicknesses (>3 nm) and ACMC47-film_quant the least (1.74 nm). Thicker water films were sorbed onto the ACMC materials used for the long-term humidity reactor experiments, whereby the effective water film thicknesses reached 6.00 (±0.01) nm for ACMC45-reactor and ACMC80-reactor materials by the end of the 1500 min experimental runs (Figure D; Table S6). Differences in sorption behavior at a fixed 98% RH between batches synthesized for long-term humidity experiments versus for film quantification suggest that some fundamental physical or chemical differences between synthesis batches change the interaction of amorphous carbonates with humid air; however, an exact cause is unknown. One possible explanation may be that the slightly larger particle size of the ACMCs used in reactors permitted greater condensation of water at grain surfaces as observed by Yeşilbaş and Boily. Another explanation could pertain to the presence of trace amounts of NaCl in the ACMC-reactor materials compared to those in the ACMC-film_quant materials. The Na concentrations of ACMC- and AMC-reactor materials are uniformly higher (680–11,760 mg/kg) than those synthesized for film quantification (385–3000 mg/kg), thus suggesting the possibility of higher NaCl. The deliquescence of NaCl at 75% RH may be one contributing factor to higher water sorption for the materials used in the long-term experiments during our static humidity test (e.g., Wise et al.). However, no halite crystals were observed in SEM images or XRD patterns of the ACMC80-reactor or ACMC100-reactor materials and halite was only detected in one of eight ACMC45-reactor samples during the long-term humidity experiments (Figure ). Nebel et al. previously reported that trace amounts of Na may either substitute for Ca in ACC or be present as amorphous sodium carbonate, presenting another possibility for the deportment of Na. Nonetheless, the influence of NaCl and other salts produced during the synthesis of amorphous carbonate minerals should be considered in this and future studies as they are almost impossible to remove completely and may affect water sorption.

2.

Mineralogical compositions of ACMCs and AMC reacted at 75% RH and 98% RH for 14, 30, 130, and 210 days in the long-term humidity reactor experiments. Relative weight contributions of mineral phases were determined by Rietveld refinement with XRD data. Reproduced with modifications from Patel et al.

3.4. Mineralogy of Phases Reacted in Long-Term Humidity Experiments

As described by Patel et al., synthetic ACMCs (here labeled as ACMC-reactor) remained stable for >7 months at conditions of <53% RH when placed in reactors in which RH was controlled by saturated salt slurries. At 75% RH, ACMC16 had begun to crystallize before 130 d of reaction and the first phases observed to have formed were high Mg-calcite (HMC) and minor aragonite (Figure ). It is almost certain that other intermediate phases may have crystallized prior to the formation of the first minerals detected, but these were not always captured in our long-term humidity experiments owing to the much longer time frames required to monitor crystallization in humid air compared to those in bulk solution. Minor amounts of a hydromagnesite-like phase [Mg₅(CO₃)₄(OH)₂·nH₂O], described by Patel et al., were detectable by 130 d of reaction at 75% RH for AMC100, whereas ACMC32 and ACMC45 remained stable as amorphous phases under the same conditions. All ACMC and AMC phases (except for ACMC45) crystallized by 14 d in reactors held at 98% RH. The first phases observed after 14 d at 98% RH from ACMC16 were HMC, aragonite (CaCO₃), and minor amounts of monohydrocalcite (CaCO₃·H₂O). The first crystalline products observed in ACMC32 and ACMC45 were Ca-rich phases (HMC, monohydrocalcite, and aragonite) and Mg-carbonate phases, including a hydromagnesite-like phase and nesquehonite (MgCO₃·3H₂O). In ACMC80 and AMC100, the hydromagnesite-like phase and nesquehonite were the first crystalline phases to be detected. Monohydrocalcite transformed to aragonite by 130 days of reaction at 98% RH in the experiments for ACMC16, ACMC32, and ACMC45. In experiments for ACMC with >32 mol % Mg, nesquehonite and the hydromagnesite-like phase (which likely has more structural H₂O than dypingite) transformed to dypingite. It was necessary to surpass a threshold of between 75 and 98% RH to accelerate crystallization of ACMC phases in these long-term experiments.

3.5. Change in Crystal Morphologies through Time

Under high humidity conditions (98% RH), ACMC reacted in long-term humidity experiments produced multiphase assemblages comprising 2–5 coexisting crystalline phases with a diverse range of morphologies. By coupling SEM with EDS, we were able to identify most crystalline phases and link them to the corresponding Ca-, Ca–Mg-, or Mg-carbonate phases detected using XRD. For instance, Mg-bearing calcite was distinguished from aragonite because aragonite cannot accommodate Mg in its crystal structure. In samples where monohydrocalcite and aragonite were both present, the use of elemental compositions became more challenging because negligible Mg-substitution is expected in both phases. Where there were multiple phases with similar chemical compositions as identified with XRD within the same sample, and they could not be distinguished using their crystal habit or crystal system in SEM images, we used the more general terminology of “Ca-carbonate” and “Mg-carbonate” to label phases based on EDS data. The morphologies of some phases were similar for all ACMC compositions; for instance, we observed spheres of HMC with pitted or smooth surfaces for all starting compositions of ACMC (Figures S11A–D and S13–S15), whereas the habits of other phases differed greatly as a function of ACMC composition, and the forms taken by some minerals even varied across a single sample.

In ACMC16-reactor, ACMC32-reactor, and ACMC45-reactor transformation products, aragonite occurred as prismatic crystals with either steep or shallow dipyramidal terminations (Figure A,C) or as tabular crystals with visible twinning to form pseudohexagonal prisms (Figures D and S12B). In ACMC32 and ACMC45, we also observed aragonite with spear-shaped morphologies that gave rise to chevron-like growth patterns (Figures B and S12A,D) and the stacking of tabular, twinned crystals to form aggregates in elongated pyramidal shapes (Figures A and S12C). Interestingly, the morphology of aragonite changed over time in some samples, i.e., ACMC45 reacted at 98% RH for 130 days and 210 days (Figure ). We further note morphological changes over time for dypingite/hydromagnesite in ACMC80 and AMC100 samples (Figure E–H).

3.

Scanning electron micrographs of different aragonite morphologies formed from ACMC45-reactor at 98% RH. (A) Tabular aragonite crystals with shallow-angle pyramidal terminations on both ends stacking and growing into larger conical formations. (B) Spear-headed aragonite growing to form a chevron morphology. (C) Acicular or tabular aragonite with steep dipyramidal terminations clustered with no apparent oriented growth. (D) Blocky aragonite showing some pseudohexagonal twinning.

4.

Scanning electron micrographs showing morphologies of Mg-carbonate minerals (hydromagnesite, dypingite, and a hydromagnesite-like phase) produced by the reaction of amorphous Ca–Mg-carbonates (ACMC32-reactor, ACMC45-reactor, ACMC80-reactor) and amorphous Mg-carbonate (AMC-reactor) at 98% RH. (A,B) Rosettes formed during earlier stages of crystallization for ACMC with >30 mol % Mg. (C,D) By >130 days of reaction, ACMC with <50 mol % Mg showed similar, but much larger (∼20–60 μm diameter) rosettes composed of crosscutting, platy crystals. (E–H) Larger agglomerates (20–100 μm across) of stacked, subparallel (and occasionally curving) plates of dypingite/hydromagnesite formed from ACMC80 and AMC100 after 130 days of reaction.

Dypingite, hydromagnesite, and a hydromagnesite-like phase in ACMC32 and ACMC45 always formed spheroidal rosettes of crosscutting, planar crystals (Figure A–D), whereas in ACMC80 and AMC100, a hydromagnesite-like phase formed rosettes in earlier stages of crystallization only to produce 5–20 μm stacks or curving fan-like structures composed of platy crystals between 130 and 210 days (Figure E,F). Alongside these books of platy crystals, there are also agglomerates of spherical rosettes; however, the individual crystals that comprise these rosettes are much smaller and subhedral.

4. Discussion

During exposure to high RH in our long-term reactor experiments, we show that transformations of amorphous carbonates and metastable crystalline carbonates are facilitated by the physisorption of nanometer-scale thin water films at mineral surfaces. Subsequent interactions at the mineral–thin film interface drive continued recrystallization over time, which is evidenced by changing phase morphologies and mineral dehydration (transformation) reactions. The main controls on crystal transformation and growth in thin water films are likely (1) Gibbs free energy, (2) mineral saturation state, (3) action of impurities, (4) changes in surface energy across the substrate (i.e., controlled by the underlying structure, surface defects, and crystal size and morphologies), (5) solvent properties (activity, hydrogen bonding), (6) transport of solutes, and (7) the space available for growth (governed by the architecture of the substrate and the thickness of the water film). − Here, we discuss the possible underlying controls on carbonate (re)­crystallization in these thin film environments. First, we highlight evidence attributing (re)­crystallization to dissolution–precipitation and present hypotheses regarding ion mobility. The roles of phase metastability, surface energy minimization, and ion diffusion as possible factors facilitating recrystallization in thin films are then considered, as is the impact of water film thickness on ion diffusion. Finally, we discuss the possible effects of changing substrate properties with time (phase solubility, surface energy, and morphology) on recrystallization processes in thin films.

4.1. Heterogeneity of Water Films

The bulk analytical tools used in this study provide a measure of the effective water film thickness at the surface of ACMCs and AMC, and the corresponding RH, required to drive crystallization. For the purpose of simplicity, we assumed the formation of continuous, uniformly thick water films; however, previous work shows that such films are commonly discontinuous, of variable thickness, and affected by the evolving physical and chemical properties of mineral surfaces. For instance, preferential sorption of water may occur at surface defects, corners, kink sites, or ledges, and water may condense within pores and at grain boundaries. Indeed, the heterogeneous nature of crystal growth on flat calcite surfaces has led previous authors to conclude that water film thickness is variable across relatively smooth mineral surfaces even at >70% RH. , As the aggregates of ACMC and AMC nanoparticles used in our experiments are more topographically heterogeneous than freshly cleaved calcite surfaces, it is highly likely that the water films we formed were of variable thickness and noncontiguous (i.e., patchy or discontinuous).

When held at 98% RH, ACMCs and AMC in our static DVS experiments sorbed a considerable amount of water (∼8–20 wt %) across a 24 h period; however, these films are unlikely to be of constant thickness due to the local architecture of the substrate. Nonetheless, given that phase transformation is accelerated between 75% and 98% RH in our long-term humidity reactor experiments, and that the effective water film thickness increased to ∼1–2 nm, we have shown that increased water sorption strongly controls crystallization of amorphous carbonates. It may be inferred from our results that crystallization is strongly inhibited for effective water film thicknesses below ∼1 nm.

4.2. Interface-Coupled Dissolution–Reprecipitation in Nanometric Water Films

Exposure to high humidity conditions facilitates transformation of amorphous and crystalline metastable carbonate phases, and previous studies have proposed that this transformation occurs through interface-coupled dissolution–reprecipitation within physisorbed water films at carbonate surfaces. , The SEM images of ACMC phases reacted at 98% RH capture the intermediate stages of these interface-coupled processes. First, development of etch pits (black arrows, Figure A) suggests that mineral surfaces are in the process of dissolving. The surfaces of HMC spheres show remnants of an original nanoparticulate texture that is being progressively overgrown by smooth exteriors (Figure A–D). For example, in Figure D, the outermost layer of the HMC sphere is being replaced by the formation of pyramidal crystals (blue arrows). Surface restructuring and loss of the original, rough (nanoparticulate) texture at the surface characterizes a progressive dissolution–reprecipitation front that appears to begin at the surface and progress deeper into the solids over time. The coalescence of nanoparticles, indicated by the decreasing definition between grain boundaries and the flattening of nanospheres, suggests partial dissolution and recrystallization driven by Ostwald ripening (green arrows in Figure A,C). These features confirm that crystallization of amorphous material, and subsequent recrystallization, occurs in the presence of adsorbed water (or possibly condensed water) at surfaces, which induces dissolution of the solids due to disequilibrium between the water film and the solid phases. Dissolution could generate pathways for the fluid to access areas beneath the particle surface, allowing the dissolution–reprecipitation front to progress deeper into the solids as is common in mineral replacement reactions.

5.

Scanning electron micrographs taken after reaction of ACMC-reactor precursors under 98% RH showing interface-coupled dissolution–precipitation reactions to form HMC. In (A,B), yellow arrows show the restructuring of roughly pitted materials into smooth exterior surfaces. In (A,C), dissolution features include increasingly pitted surfaces (black arrows) and green arrows show areas where nanoparticles appear to be coalescing. In (C,D), recrystallization appears to begin at the exteriors of amorphous and crystalline aggregates resulting in restructuring of the surfaces into different morphologies (blue arrows).

4.3. Mobility of Aqueous Species during Dissolution

The earliest observations of crystallization in long-term humidity reactor experiments are characterized by the presence of either smooth or pitted spheres of HMC (Figures A and S11A). The spheres of HMC in samples reacted for 30 d at 98% RH had jagged or sinuous grain boundaries, with cracks extending from the boundaries toward the interior (Figure S16A,B). The grain boundaries of the smooth spheres of HMC appeared to be dissolving and being replaced by sheet-like Mg-carbonate phases (Figures B and S16A,B) or pyramidal tabular crystallites of Ca-carbonate (Figure S16D). These textures suggest a spatially linked dissolution–reprecipitation process in which crystallization of Mg- or Ca-carbonate phases occurs at the expense of HMC. Additionally, in ACMC45 after 30 d at 98% RH, there appears to be the concurrent precipitation of platy Mg-carbonates interspersed with 100–200 nm pyramidal crystals of a Ca-carbonate phase (Figure S16C). These images show that distinct Mg-rich and Ca-rich phases can recrystallize via dissolution of a single Ca–Mg-rich precursor.

6.

Scanning electron micrographs of reaction products of ACMC32 held at 98% RH in the long-term humidity experiments. Dashed yellow lines highlight the boundaries between carbonate phases with distinct compositions. (A,B) The recrystallization of HMC spheres (“Mg-cal”) into flaky Mg-carbonate rosettes after 14 days (Mg ion transport shown by the blue arrows). (C) Grain boundaries forming between crystalline Ca- and Mg-carbonates growing from ACMC after 30 d. (D–F) The segregation of Ca- and Mg-rich phases becomes more pronounced and occurs on scales >20 μm due to increasing crystallite size. (E) Continued replacement of HMC spheres by Mg-carbonate phases is observed at 7 months of reaction and is now colocated adjacent to aragonite (arg).

Through time, we observed increasing delineation (or segregation) between Ca- and Mg-rich phases formed from Ca–Mg-carbonate phases such as ACMC or HMC. In ACMC32 reacted at 98% RH, the first crystalline products observed contained smooth spheres of HMC, from which platy aggregates of hydrated Mg-carbonate minerals crystallized (Figure A,B). The crystallization of these Mg-carbonates is initially constrained to edges between two adjacent spheres, where pendular rings of water may have formed and become intermittently interconnected, and occurs only in discrete regions of the outer surfaces of the spheres (Figure A,B). At 30 d, we begin to observe the co-occurrence of flaky, crosscutting plates of Mg-carbonate adjacent to tabular crystals of a Ca-carbonate phase (Figure C). The boundary between the two phases is punctuated by the occurrence of an amorphous nanoparticulate phase of ACMC, suggesting that ACMC may be transforming into both of these materials. As time progresses, the aggregates of Ca- and Mg-carbonates coarsen in size (10–20 μm) with more pronounced distinction between Ca-carbonate and Mg-carbonate phases to form spatially distinct agglomerates of 20–40 μm in size (Figure D–F). By 7 months of reaction, we see the apparent differentiation of the Ca and Mg into pure Ca-carbonate and Mg-carbonate phases (Figure E,F), with aragonite (present as either the chevron morphology or as stubby prisms with pyramidal terminations) and dypingite rosettes having replaced much of the HMC. The observation of the separation and growth of Ca- versus Mg-carbonates demonstrates that Ca and Mg ions are mobilized by dissolution–precipitation and effectively transported within thin water films.

4.4. Possible Mechanisms Facilitating Recrystallization in Thin Water Films

Evidence from long-term humidity experiments shows extensive recrystallization within thin water films over time scales of greater than a month. Our SEM observations suggest that (re)­crystallization may have occurred in distinct microenvironments with variable water film thickness including the possibility that thicker water films may have formed at grain boundaries and become intermittently connected (after Tokunaka). Recrystallization may be facilitated by several processes including (1) reduction of bulk Gibbs free energy of the system, commonly described by Ostwald’s step rule, (2) reduction of interfacial energy (Ostwald ripening), and (3) ion diffusion facilitating dissolution–precipitation.

The first of these processes, exemplified by Ostwald’s step rule, pertains to recrystallization coupled to phase transformations. Typically, the kinetic barriers for precipitating metastable amorphous or crystalline carbonates are lower than for the most thermodynamically stable phase. , The initial precipitation of metastable phases is thus governed by kinetics, but importantly these phases do not represent the lowest free energy state of the system and so phase transformations proceed to reach increasingly lower states of Gibbs free energy. The second mechanism: Ostwald ripening, describes recrystallization that proceeds to minimize interfacial energy of a crystal or a group of crystals. Commonly, reduction of surface energy is achieved by decreasing surface area of the crystal or group of crystals, which involves the sacrificial dissolution of small crystals (with high surface energy) and concurrent growth of larger crystals (with lower surface energy). Alternatively, surface area may be minimized by the preferential dissolution of areas with high surface energy (defects, kink sites) and mass transfer to areas with lower surface energy (fewer defects, homogeneous crystal surfaces, larger crystals). Finally, recrystallization in thin-film environments must be reconciled with dissolution–precipitation, and thus ion transport mechanismswhich control local saturation state. In our experiments, advection-driven transport is assumed to be negligible and instead ion transport is likely controlled by diffusion driven by concentration gradients (Fickian diffusion). In Fickian diffusion, movement of a solute occurs to even out an overall solute concentration gradient. The transport of ions by Fickian diffusion may then impact mineral dissolution or precipitation by changing the local saturation state with respect to the reacting phases.

XRD data from the long-term humidity experiments show amorphous-to-crystalline phase transformations of ACMC and AMC, where crystallization of these materials is initiated by sorption of water at carbonate surfaces (at humidity conditions ≥75%). We observed the formation of HMC, monohydrocalcite, nesquehonite, aragonite, and a hydromagnesite-like phase, all of which are metastable crystalline materials under the considered experimental conditions. Subsequent phase transformations of monohydrocalcite-to-aragonite and nesquehonite-to-hydromagnesite/dypingite offer important evidence to support recrystallization driven by the architecture of the free energy landscape, i.e., Ostwald’s step rule. Although this may comprise an important driving force for recrystallization, it is sometimes unable to fully account for the crystallization pathways taken by metastable intermediates during ACMC transformation. Studies of ACMC transformation in bulk aqueous solution report variable phase transformation pathways, where ACMCs (of equivalent composition) may crystallize to either a high- or very-high Mg-calcite, ,− or crystallize to biphasic assemblages comprising a Ca-carbonate (monohydrocalcite or aragonite) and a hydrated Mg-carbonate (nesquehonite or hydromagnesite) at temperatures of 20–30 °C. , Differences in crystallization pathways are even observed for an ACMC phase reacted in duplicate in the study conducted by Montes-Hernandez et al. We also observe variable transformation pathways, in which one or more of HMC, monohydrocalcite, nesquehonite, and dypingite form early during the crystallization process; however, it should be recalled that our experiments could not be monitored in situ owing to the long timeframes required for crystallization of ACMC in humid air; as such, the identities of the very first crystallization products of ACMC in the long-term reactor experiments cannot be known with certainty. The variability of ACMC transformation products in solution and in thin water films suggests that recrystallization is not easily predictable based on Ostwald’s rule of phases alone. Fascinatingly, we also observe the crystallization of aragonite and hydromagnesite/dypingite at the expense of HMC, and this, to the best of our knowledge, has never been reported in ACMC transformation studies conducted in solution. Typically, during dissolution–precipitation of HMC in solution, it will recrystallize without phase transformation by Ostwald ripening.

It is also important to consider Ostwald ripening as a mechanism of carbonate recrystallization in thin water films. Scanning electron micrographs of solids following reaction at 98% RH show increasing nanoparticle coalescence in ACMC phases (Figure A,C), an indication of Ostwald ripening. In addition, following crystallization of HMC spheres, there is also evidence of surface restructuring from pitted and rough surfaces to those that are smooth (Figure A–C). A decrease in surface heterogeneity and defects at crystal surfaces reflects surface energy minimization by Ostwald ripening. Finally, coarsening of hydromagnesite and dypingite crystals with time (as shown in Figure ) from rosettes to fanning or stacked sheet-like structures and the coarsening of aragonite crystals (Figure ) further suggest that Ostwald ripening plays an important role in driving recrystallization and crystal growth in thin-film environments, similar to that observed in bulk solution.

Finally, our SEM data provide evidence to suggest recrystallization by dissolution–precipitation is facilitated by ion transport in thin water films. Notably, the sacrificial dissolution of HMC, the subsequent precipitation of Ca- and Mg-carbonate phases, and the increasing delineation between Ca and Mg phases over time suggest that ion transport is facilitated despite the absence of bulk water, at least on a local scale. Fickian diffusion following phase dissolution is a possible ion transport mechanism. During dissolution of a metastable phase, the site of dissolution would likely be characterized by high concentrations of divalent cations (Ca²⁺ and/or Mg²⁺) and (bi)­carbonate ions; thus, concentration gradients of these solutes could potentially arise between the dissolution site and more distal areas of a thin film permitting formation of discrete microenvironments of local undersaturation (dissolution) and supersaturation (precipitation). Diffusion processes in thin films may have been affected by crystallization kinetics, whereby preferential precipitation of Ca-carbonate minerals over Mg-carbonate minerals may have temporarily perpetuated concentration gradients of ions such as Ca and bicarbonate/carbonate. Subtle differences in precipitation rates could then have resulted in variable Ca/Mg ratios in thin films and thus precipitation of Ca-rich and Mg-rich phases in different locations. Although cation segregation might be facilitated by precipitation-driven diffusion (arising from preferential Ca-carbonate crystallization), it would first necessitate the precipitation of Ca-carbonate minerals. As such, during dissolution of HMC, where both Ca and Mg ions are released into the thin film contemporaneously, the primary formation of Ca-carbonate phases would imply their precipitation closer to the site of HMC dissolution than the precipitation of Mg-carbonate phases. In many cases, SEM images show the sacrificial dissolution of HMC colocated with precipitation of a hydrated Mg-carbonate (e.g., Figures A,B and S16A,B) and the notable absence of Ca-carbonate along the surface. This spatial correlation of HMC dissolution proximal to the precipitation site of hydrated Mg-carbonates is inconsistent with prior, local Ca-carbonate crystallization, thus suggesting that this ion transport mechanism may not fully account for these recrystallization features.

Dissolution of carbonate phases can also generate ionic gradients that drive ion transport through diffusiophoresis. In the case of calcium carbonate dissolution, Ca²⁺, HCO₃ ^–, and OH^– ions are released from the solid into solution and they diffuse at different rates away from the carbonate surface. The difference in diffusive rates causes small concentration gradients to arise between positive ions and negative ions, which induces spontaneous electric fields reported to be 1–10 V/cm. Charged ions then move in response to the electric fields and surface charges of the substrate based on their polarity in a mechanism of electrophoresis. , Diffusiophoretic gradients have been observed during mineral dissolution and in mineral replacement reactions, and the development of electric fields relies on the rate of dissolution and mineral solubility. Interestingly, Velegol et al. have suggested that generation of ionic gradients should be possible even in the absence of a free liquid and they speculate that condensation of water on a hygroscopic mineral could induce transient ionic gradients. If diffusiophoresis were to cause different rates of transport of Ca versus Mg, this could provide another potential explanation for the discrete precipitation of Ca- versus Mg-carbonate phases in different locations over time within thin water films.

Importantly, although diffusiophoretic electrophoresis may explain initial separation of Ca and Mg ions, the continued separation of Ca-rich and Mg-rich phases over time does not require this mechanism. That is, if initial separation of Ca-carbonate and Mg-carbonate nuclei occurs, then continued crystal growth may be controlled by crystal structure. For instance, Mg is preferentially excluded during aragonite growth, because its structure cannot accommodate Mg ions. Similarly, Ca is excluded from hydrated Mg-carbonates during their growth. Consequently, these ions would remain dissolved in the thin water film until they could be transported and incorporated into the structures of a compatible phase. As previously suggested, the local depletion of Ca or Mg ions during precipitation may also drive diffusion of these species to precipitation sites, thus contributing to crystal growth and the observed segregation of Ca and Mg phases with time. Finally, it is also possible that dynamic and intermittent connection and disconnection of pendular rings of water between mineral grains might influence water film chemistry in unpredictable ways, potentially giving rise to unexpected mineral assemblages such as the dypingite crystals that nucleated at the surface of high-Mg calcite spheres.

4.5. Impacts of Water Film Thickness on Ion Diffusion

Previous work suggests that the lower limit for diffusion of solutes in thin films is approached when water film thicknesses approach the dimensions of hydrated ion diameters, which typically range from 0.25 to 0.5 nm, suggesting that diffusion within thin water films becomes feasible once they exceed an effective thickness of ∼1 nm. An earlier study of ACC transformation in thin films of water also suggests a threshold for the effective water film thickness of ∼1.2 nm required for transformation. In the materials used in this study, this 1 nm lower limit (gray box in Figure B) is exceeded for low Mg ACMC (ACMC22-film_quant and ACMC23-film_quant) at 67.7% RH and 65.2% RH, respectively. Yet, ACMCs with intermediate Ca/Mg ratios surpass this 1 nm water film threshold at higher humidities: ACMC48-film_quant at 69.8% and ACMC45-reactor at 75.8%. Both AMC samples (AMC100) first exceed this threshold at 75–76% RH, while ACMC80-reactor and ACMC78-film_quant only generated an effective water film thickness of 1 nm at 87–88% RH. The RH at which this threshold is surpassed may be linked to ACMC crystallization kinetics. For example, an effective water film thickness of 1 nm is exceeded at <68% RH in ACMC22-film_quant and ACMC23-film_quant, which is consistent with findings of the long-term humidity experiment where XRD data showed extensive crystallization of an ACMC with similar composition (ACMC16-reactor) at 75% RH after 130 d. Conversely, in materials for which the 1 nm threshold is exceeded at >75% RH, carbonate phases remained amorphous after 7 months of reaction at 75% RH (i.e., ACMC45-reactor, ACMC80-reactor, and AMC100-reactor). Most notably, limited crystallization of amorphous phases at <53% RH, as described by Patel et al., can be attributed to an effective water film thicknesses below this ion diffusion threshold for all ACMC and AMC materials. Therefore, water film thickness may strongly control amorphous-to-crystalline phase transformation by limiting ion diffusion.

4.6. Factors That May Affect Continued Recrystallization with Time

Previous studies of mineral reaction in unsaturated porous media show that continued dissolution–precipitation over time can be inhibited by surface passivation. , Surprisingly, here we do not observe surface passivation despite recrystallization proceeding through nanometer-scale sorbed water films with an effective thickness of 2–6 nm. Instead, sorption of these water films permits widescale dissolution–precipitation, where water must be transported into solid phase interiors to facilitate complete recrystallization. The pervasive and ongoing carbonate recrystallization shown by our long-term experiments at 98% RH may be driven by multiple factors and mechanisms. First, all crystalline phases detected after 7 months of reaction are metastable, thus rendering them susceptible to recrystallization as the system approaches a free energy minimum. This thermodynamic driving force is amplified by continued Ostwald ripening, which is expressed in our SEM data by crystal size coarsening, resurfacing of minerals, and decreasing surface roughness. These findings suggest that Ostwald ripening may be a key factor driving recrystallization even in water-limited environments. Ostwald ripening may further impact ion transport processes: as time progresses and bigger crystals form, absolute water sorption capacity may decrease due to lower surface area, and this may impact recrystallization processes driven by ion transport. The combined effects of (1) limiting the thickness of water available for ion transport, (2) increasing variability in surface roughness and surface chemistry (compared to the existence of solely one amorphous material), and (3) increasing the diversity of crystalline surfaces on which new crystals can form may also contribute to a wider range of growth mechanisms and the resultant diverse morphologies observed.

Over time, we expect mineral dissolution to decrease as a result of decreasing interfacial energy at the mineral surface and a possible decrease in water sorption and thus the “motor” propelling transport by diffusion or diffusiophoresis to slow. However, continuation of ion transport over these time scales may be made possible by several factors including variable water film thickness and differences in the solubility and metastability of adjacent phases. Differences in water film thicknesses may be produced by (1) increasingly complex and variable crystalline morphologies, which may produce topographical depressions in which water can accumulate, (2) formation of crystal aggregates comprising several intergrown phases with different water sorption capacities, and (3) changing surface energy and charge density across phases, all of which may lead to heterogeneous thicknesses of sorbed water. Variability in water film thickness across the surface of multiphase aggregates may permit continued dissolution, particularly if phases are highly soluble. If phases with differing solubilities and reactivities are adjacent to one anotheras seen in SEM data where multiphase assemblages occur across aggregatesit may be possible that different areas are susceptible to dissolution at different time points and this could potentially maintain concentration gradients that can drive ion transport. Ion transport processes over time may further be influenced by the changing architecture of the substrate, for example, changing surface charge may impact the transport of ions due to interactions with the charged mineral surfaces. In addition, changing morphology may change diffusion pathways. Based on our observations of mineral growth from our long-term experiments, it is possible that multiple processes including phase transformation, Ostwald ripening, ion diffusion, and electrophoretic diffusiophoresis may contribute to recrystallization of carbonate minerals in thin water films.

5. Conclusions

The amount of water sorbed to amorphous Ca–Mg-carbonates (ACMC) was constrained over humidity conditions of 10–98% RH, and the effective thicknesses of thin water films were calculated. Observations from long-term humidity experiments, reacting ACMC and AMC phases at 75% and 98% RH, were used to describe crystallization within comparable thin water films. Water sorption data show an increase in water sorption between 75 and 84% RH for all ACMC and AMC materials, which corresponds to faster crystallization in long-term experiments conducted at 98% RH compared to those at 75% RH. We find that accelerated amorphous-to-crystalline transformation is linked to humidity conditions (and ACMC compositions) for which the effective thickness of sorbed water films exceeds a threshold of 1 nm, which is required for ion diffusion, thus suggesting that diffusion in thin water films strongly controls the amorphous-to-crystalline phase transformation. Physisorbed water films generated at 98% RH are estimated to have an effective thickness of 2–6 nm for ACMC materials (22 ≤ mol % Mg ≤ 100). Crystallization of ACMC, and recrystallization of metastable crystalline phases, is realized via interface-coupled dissolution–precipitation reactions through long-term exposure to thin water films at high humidity. Notably, we provide the first report of crystallization of discrete Ca-carbonate and Mg-carbonate phases at the expense of a mixed Ca–Mg precursor (i.e., HMC), which has never been reported for ACMC transformation in aqueous solution.

These results provide the first experimental constraints on the minimum effective thin film thickness required to initiate crystallization of ACMCs and a discussion of the possible mechanisms through which crystallization and recrystallization may operate in thin water films. This work offers new insights and methods for understanding the behavior of other hygroscopic minerals in water-limited environments such as Martian landscapes, cement infrastructure, and mineral carbonation environments, and on the generation of diverse morphologies that could be exploited in material synthesis and crystal design.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.cgd.6c00090.

• Detailed DVS data and compositional information for ACMCs and AMC (PDF)

§.

Department of Physical Sciences, MacEwan University, Edmonton, AB T5J 4S2, Canada

This manuscript has been written with contributions from all authors. Sample synthesis and characterization of solids (DVS, XRD, TGA, SEM) were completed by ASP. Rietveld refinement was completed by ASP and SW. BET surface area and TGA-FTIR data were collected by JMB. Data interpretation was conducted by ASP, MJR, SW, and ALH.

The authors declare no competing financial interest.