Heterogeneous Nucleation and Growth of CaCO 3 on Calcite (104) and Aragonite (110) Surfaces: Implications for the Formation of Abiogenic Carbonate Cements in the Ocean

: Although near-surface seawater is supersaturated with CaCO 3 , only a minor part of it is abiogenic (e.g., carbonate cements). The possible reason for such a phenomenon has attracted much attention in the past decades. Substrate e ﬀ ects on the heterogeneous nucleation and growth of CaCO 3 at various Mg 2 + / Ca 2 + ratios may contribute to the understanding of the origin of abiogenic CaCO 3 cements. Here, we used in situ atomic force microscopy (AFM), scanning electron microscopy (SEM), X-ray di ﬀ raction (XRD) and Raman spectroscopy to study the heterogeneous nucleation and growth of CaCO 3 on both calcite (104) and aragonite (110) surfaces. The results show that (1) calcite spiral growth occurs on calcite (104) surfaces by monomer-by-monomer addition; (2) the aggregative growth of aragonite appears on aragonite (110) surfaces through a substrate-controlled oriented attachment (OA) along the [001] direction, followed by the formation of elongated columnar aragonite; and (3) Mg 2 + inhibits the crystallization of both calcite and aragonite without impacting on crystallization pathways. These ﬁndings disclose that calcite and aragonite substrates determine the crystallization pathways, while the Mg 2 + / Ca 2 + ratios control the growth rate of CaCO 3 , indicating that both types of CaCO 3 substrate in shallow sediments and aqueous Mg 2 + / Ca 2 + ratios constrain the deposition of abiogenic CaCO 3 cements in the ocean. perspective the origin of abiogenic CaCO 3 cements in the


Introduction
The supersaturation of near-surface seawater is about six and four times higher than that with respect to calcite and aragonite, respectively, whereas a much smaller fraction of the total sedimentary CaCO 3 minerals is abiogenic [1], which contradicts traditional thermodynamics [2,3]. These abiogenic CaCO 3 minerals, composed of aragonite and high-magnesium calcite in shoal-shallow marine environments [4][5][6], typically form either as heterogeneously precipitated primary marine cements or through post-depositional and diagenetic reactions [1]. In order to further understand the lack of abiogenic CaCO 3 in the modern ocean, many efforts have been conducted to ascertain factors which affect the precipitation of abiogenic CaCO 3 [7][8][9][10][11].
The precipitation of abiogenic CaCO 3 correlates with aqueous chemistry [12]. The pH, temperature, P CO 2 and a HCO − 3 affect the saturation index (SI, which is defined as SI = logΩ = log(IAP/ K sp ), where Ω denotes saturation state, and IAP and K sp represent activity product and solubility product, respectively) of solutions [13] and play key roles in determining the precipitate rate (r) of CaCO 3 polymorphs. The r is constrained by Ω, following the equation r = k (Ω − 1) n [2]. In addition, the formation of CaCO 3 minerals is partly kinetically driven, i.e., calcite growth favors a low reaction rate, while aragonite and vaterite prefer to form in relatively fast reactions [14]. In addition, the precipitation of CaCO 3 (e.g., calcite) depends on the a Ca 2+ /a CO 2− 3 of solutions. Due to the differences in the dehydration properties between Ca 2+ and CO 3 2− and the geometry of calcite step kink sites, the step velocities of the obtuse and acute angled edges can reverse with the variation of a Ca 2+ /a CO 2− 3 at a fixed SI [15][16][17]. On the other hand, impurity ions and molecules in solutions can also influence CaCO 3 precipitation. Inorganic anions (e.g., PO 4 3− and SO 4 2− ) exert the impacts by competing with CO 3 2− for Ca 2+ sites or incorporating in a crystal lattice [13]. Divalent metal cations, such as Mg 2+ , Ba 2+ and Sr 2+ , inhibit calcite growth by step pining, incorporation or kink blocking [18][19][20][21]. Similarly, soluble organic molecules (such as humate, fulvate, citrate and polyfunctional aromatic acids) inhibit the precipitation of aragonite by the modification of functional groups in these additives [11,22,23]. The surface properties of the growth substrate are another factor affecting the precipitation of abiogenic CaCO 3 [24][25][26][27]. The surface charge of an inorganic substrate relates to the growth of CaCO 3 polymorphs. Metastable CaCO 3 phases (e.g., aragonite or vaterite) are favored by negatively charged substrates, whereas positively charged substrates contribute to the formation of the stable CaCO 3 phase (i.e., calcite) [28]. In addition, the lattice mismatch between substrates and surface precipitates is also vital in the growth of abiogenic CaCO 3 [29]. Based on classical nucleation theory, the energy barrier (∆G C ) to form a critical nucleus is positively correlated with γ 3 (∆Gc ∝ γ 3 /(−RTlnσ) 2 , where γ denotes the interfacial free energy, σ the supersaturation, and R and T represent gas constant and Kelvin temperature, respectively [30]. The precipitation rate is subsequently controlled by the lattice mismatch, because a larger lattice mismatch leads to a higher ∆G C [31].
Abiogenic CaCO 3 cements tend to overgrow on shoal-shallow calcium carbonate-rich sediments [32]. Different types of CaCO 3 inorganic substrate should be taken into consideration, to study the precipitation of abiogenic CaCO 3 cements. The substrate effects of CaCO 3 have been found to control the mineralogical properties of carbonate cements [7,13], and plenty of works have also been conducted on the heterogeneous nucleation and growth of CaCO 3 on calcite and aragonite seeds [33][34][35]. However, little attention has been paid to the internal mechanism and crystallization pathways of CaCO 3 precipitates in these systems. Although the precipitation of CaCO 3 on calcite (104) surfaces has been well documented [3,23,36,37], it may not represent the formation of abiogenic aragonite cements, due to the distinct surface properties of the two polymorphs [38]. Additionally, there are some controversies regarding aragonite nucleation and growth in Mg 2+ -bearing solutions. Mg 2+ -induced inhibition was reported in both nucleation and growth [39], whereas either inhibition [33] or non-inhibition [40] was merely found in the growth processes. Therefore, the evolution and mechanism of heterogeneous nucleation and growth of both calcite and aragonite at the nanoscale on CaCO 3 substrates in Mg 2+ -bearing solutions with different concentrations should be investigated, to deeply understand the formation of abiogenic carbonate cements in the ocean.
In this study, we used AFM, SEM, XRD and Raman spectroscopy to investigate how calcite (104) and aragonite (110) substrates affect the heterogeneous nucleation and growth of CaCO 3 under different solution supersaturations at pH = 8.0 ± 0.1. Since Mg 2+ is a chief modifier for CaCO 3 precipitation in sedimentary environments, we altered the Mg 2+ /Ca 2+ ratios from 0 to 3 in the growth experiment. We observed that both the mineral phases and crystallization pathways of CaCO 3 precipitated on different types of CaCO 3 substrates are disparate. In addition, the precipitation rates of CaCO 3 generated on calcite and aragonite substrates are negatively correlated to Mg 2+ /Ca 2+ ratios. These findings reveal the different crystallization processes of CaCO 3 grown on calcite (104) and aragonite (110) surfaces, providing a new perspective on the origin of abiogenic CaCO 3 cements in the ocean.

Sample Preparation
An Iceland spar (from Guizhou, China) was cleaned using ethanol and deionized water (resistivity = 18.2 MΩ cm −1 ). Then, fresh calcite (104) surfaces (2 × 2 × 1 mm 3 ) were prepared by scalpel cutting along the cleavage plane. Additionally, a diamond wire cutter (STX-202A, Kejing Auto-instrument Co., Ltd., Shenyang, China) was used to slice aragonite (110) surfaces (2 × 2 × 1 mm 3 ) from single crystals (from Morocco). To obtain polished aragonite (110) surfaces, a Leica SP1600 (Leica, Wetzlar, Germany) saw microtome was employed. Since polished aragonite (110) surfaces would enhance the growth rate of CaCO 3 by the increase in surface defects (which will reduce the energy barrier (∆G C ) to form a critical nucleus, and the number of nanoparticles will increase after increasing surface defects, thus causing the increase in growth rate), we used them in the next experiments to elevate experimental efficiency. Next, 15 mL acetone was used to wash these polished (110) surfaces through ultrasonic bathing for 1 h, and then these (110) surfaces were taken out from acetone and dried with pure N 2 . Afterwards, these prepared calcite (104) and aragonite (110) samples were glued onto steel pucks with wax for AFM experiments.

Solution Preparation
High-purity CaCl 2 , MgCl 2 , NaHCO 3 , NaCl (purchased from Aldrich and Macklin) and deionized water were used to prepare solutions with different Mg 2+ /Ca 2+ ratios (0 and 3) and saturation indexes, with respect to calcite and aragonite (i.e., SI calcite and SI aragonite ), based on Visual MINTEQ [41] calculations (Table 1). Since the SI calcite is 0.13 greater than SI aragonite (SI calcite = SI aragonite + 0.13) in the same solution, we only used the former to label saturation states of solutions. Solution chemistry in the system is constrained by the amounts of [Mg 2+ ], [Ca 2+ ], CO 2 , pH, and ionic strength (IS). The pH was maintained at 8.0 ± 0.1, and the IS was controlled up to 0.10 M (to avoid the coverage of substrates by too much NaCl, we did not used the salinity of seawater). All solutions were freshly prepared before the growth experiment, to prevent the precipitation of supersaturated solutions. Before introducing growth solutions, the solution with SI calcite = 0.07 was injected into the fluid cell for at least one hour to make the surface of samples flatter. Based on solution chemistry, equilibriums in chemical reactions were taken into consideration as the following.
where (Mg 2+ ) represents the activity of Mg 2+ , and CO 2 is the total concentration of carbonate species. Equations (5) and (6)

Growth Experiments Measured by In Situ AFM
A Bruker Nanoscope IV Scanning Probe Microscope equipped with a flow-through fluid cell and Si 3 N 4 cantilevers was used to collect AFM images. The operational process was similar to that in our previous study [42]. After flatjaw pinchcocks were turned on, a flowing system formed between the O-ring installed on the fluid cell and sample. The flow rate of solutions was performed at 550 uL/min throughout the in situ growth experiment.

SEM, XRD and Raman Spectroscopy Analysis
SU8010 cold field emission SEM (FESEM, Hitachi, Japan), equipped with energy dispersive X-ray spectroscopy (EDS) (AMETEK-EDAX, Mahwah, NJ, USA), was utilized to observe surface morphologies and chemical compositions after growth under different conditions. In addition, mineral phases precipitated on the surfaces of substrates were analyzed by a Rigaku DMAX Rapid II X-ray diffraction system (Rigaku, Tokyo, Japan) (MoKα radiation), at 50 mV and 30 mA. A Micro-confocal Raman spectrometer (in Via, Renishaw-RM 2000 (Renishaw, Gloucestershire, UK)) at 532 nm laser extinction was also used to identify precipitates formed on these substrates.

The Heterogeneous Nucleation and Growth of CaCO 3 on Calcite (104) Surfaces
Under Mg 2+ /Ca 2+ = 0 conditions, spiral growth of calcite, with the monomolecular of growth hillock around 3.1 Å, is observed as the solution system reaches a steady-state (Figure S1a-c and Figure 2a). The morphology of growth hillock is a rhombus which consists of two acute and obtuse angles ( Figure S1d). The terrace width (λ) of steps is negatively correlated to the increase in solution SI calcite (Figure S1a-c) [3]. When the SI calcite of the solutions is equal to 0.50, 0.83 and 1.05, the λ values of obtuse steps are 175.43, 171.25 and 123.60 nm, respectively, while those of acute steps are 77.33, 74.25 and 62.80 nm, respectively. In addition, the calcite (104) surface after growing in the solution at SI calcite = 1.05 and Mg 2+ /Ca 2+ = 0 was characterized by SEM and EDS. We observed that these growth hillocks are covered by a layer of CaCO 3 crust, which is smooth and evenly distributed (Figure S1e-g).  (Figure 2e). Except for the needlelike NaCl crystals produced on the substrate, the other precipitates are Ca (1-x) Mg x CO 3 with segmented surfaces. We also used EDS to semi-quantify the Mg 2+ contents in these overgrowths precipitated on calcite (104) substrates in the solution at Mg 2+ /Ca 2+ =3 and SI calcite = 1.05, discovering that it ranges from 24.68 at% (P2) to 43.20 at% (P1).  Figure 4). Increasing SI calcite to 0.83, the aggregation rate of nanoparticles is enhanced ( Figure 5). Crusts composed of nanoparticles cover the substrate after 5 min (Figure 5c), and these crusts rapidly evolve to elongated columnar crystals (Figure 5c-h). Further increasing SI calcite to 1.05, the phenomena are generally similar to what is observed in the solution at SI calcite = 0.83, except that crusts only sporadically cover the substrate ( Figure S2).   SEM was used to observe aragonite (110) surfaces after the heterogeneous nucleation and growth of CaCO 3 in solutions with different SI calcite . The size and morphology of these crystals are identical, while the orientation of crystals in the lower and upper layers is disparate. Crystals in the lower layer have their c axes parallel to aragonite (110) substrates, whereas those in the upper layer have their c axes perpendicular to the substrates ( Figure 6). The surfaces of these crystals are generally smooth, only with local cracks and holes (as indexed by the red arrows in Figure 6f,h). In addition, nanoparticles are observed from the broken parts and a few crystal surfaces (as indexed by the red arrows in Figure 6e,f). The heterogeneous nucleation and growth of CaCO 3 on polished aragonite (110) substrates were investigated in solutions under Mg 2+ /Ca 2+ = 3. The aggregation process of nanoparticles, which is similar to that without Mg 2+ addition, can be captured by in situ AFM (Figure 7). Meanwhile, the interfaces among these nanoparticles are extremely obvious, even after 180 min (as indexed by the red arrows in Figure 7h). Increasing SI calcite to 0.83 and eventually to 1.05, the aggregation rate of nanoparticles soars, while their solid-solid interfaces are still preserved (Figure 8 and Figure S3).  The surfaces after growing in supersaturated solutions with Mg 2+ /Ca 2+ = 3 were analyzed by SEM. Although some rhombohedral particles form on these substrates, tower-like and elongated columnar crystals are major precipitates (Figure 9a,c,e). Additionally, some crystals with smooth surfaces generate around these tower-like crystals (Figure 9c,f). The surface roughness of these tower-like crystals, positively correlated with the increase in SI calcite , is greater than that of crystals precipitated in Mg 2+ -free solutions (Figure 9b,d,f).

Mineral Phases of CaCO 3 Dependent on Calcite (104) and Aragonite (110) Substrates
Mineral phases of CaCO 3 precipitates have been identified by XRD and Raman analyses. The XRD results show that calcite and aragonite tend to precipitate on calcite (104) and aragonite (110) substrates, respectively (Figure 10a). Under our experimental conditions, the addition of Mg 2+ does not induce the formation of protodolomite, dolomite, hydromagnesite (e.g., nesquehonite) or magnesite (Figure 10b). In addition, the relative intensity of XRD reflections is distinct after CaCO 3 grows on calcite (104) and aragonite (110) substrates in different solutions. The Raman results demonstrate that the mineral phase of aragonite formed on aragonite (110) substrates will not be affected by solution SI calcite (Figure 11a). In addition, we cannot observe the obvious change in mineral phase under Mg 2+ /Ca 2+ = 3 compared with that under Mg 2+ /Ca 2+ = 0 (Figure 11b).

Different Crystallization Pathways of Heterogeneous Nucleation and Growth of CaCO 3 on Calcite (104) and Aragonite (110) Surfaces
Classical and nonclassical nucleation and growth theories play critical roles in understanding the formation of minerals. The former stresses the whole processes, including the diffusion of monomers from solution to solid surfaces, the transformation of monomers from surfaces to active sites, and eventually the growth of crystals [43]. The latter emphasizes the formation and aggregation of precursors in the solution [44]. Based on our results, CaCO 3 crystals formed on calcite (104) surfaces crystallize by spiral growth, and these CaCO 3 crystals are not typical rhombohedral calcite particles. In contrast, nanoparticles are initially generated on aragonite (110) surfaces, and then followed by OA and finally form elongated columnar crystals. These crystals precipitated on aragonite (110) surfaces are divided into two layers. The crystals in the first layer are arranged with their c axes parallel to the substrate, while those in the second layer are perpendicular to the substrate. These two different orientations of crystals can be ascribed to the limitation of growth space [45]. The growth of crystals in the second layer is constrained by gaps between crystals formed in the first layer, leading to vertical extension. The SEM results show that the crystals are composed of nanoparticles, which is evidence of particle aggregation. Therefore, we proposed that the heterogeneous nucleation and growth of CaCO 3 on calcite (104) surfaces are in accordance with classical nucleation and growth theory (i.e., monomer-by-monomer addition), while those on aragonite (110) surfaces conform to nonclassical nucleation and growth theory (i.e., precursor attachment).
The results of XRD and Raman analyses demonstrate that the CaCO 3 precipitated on calcite (104) surfaces is calcite, whereas that formed on aragonite (110) surfaces is aragonite, which is attributed to the complete lattice match between minerals with identical phases [31,46]. The relative intensity of XRD reflections, as well as Raman frequencies, of the same type of substrate grown in different solutions is distinct, which is ascribed to different orientations of CaCO 3 precipitates on local surfaces. Additionally, the surface roughness of aragonite (110) is greater than that of the calcite (104) surfaces in this study, causing the drop in local supersaturation on aragonite (110) surfaces and the eventual formation of sporadic rhombohedral CaCO 3 (probably calcite) (Figure 9).
The crystallization pathways of calcite and aragonite on calcite (104) and aragonite (110) surfaces in this study are similar to those precipitated in homogeneous systems [47][48][49]. Hence, the differences in crystallization pathways between calcite and aragonite may be decided by their own crystalline behaviors. A previous study proposed that the binding force of rhombic crystals in 3D directions is identical, while that of prismatic crystals along the c axes is greater than that in both the a and b directions [50], which perfectly explains our AFM observations that aggregates of nanoparticles formed on aragonite (110) surfaces extend along the c axes.

Effects of Mg 2+ and Saturation States on the Heterogeneous Nucleation and Growth of CaCO 3 on Calcite (104) and Aragonite (110) Surfaces
Both Mg 2+ /Ca 2+ ratios and solution supersaturations impact CaCO 3 growth. On the one hand, when Mg 2+ /Ca 2+ ratios increase from 0 to 3, the spiral growth of calcite is inhibited. Sethmann et al. (2010) observed the segmentation of Mg-calcite thin films on calcite (104) surfaces, and they discovered that the premise of ridge formation is to break through the critical thickness (about 12 nm), which signifies the release of compress stress when the thickness exceeds the critical value [51]. Due to the ionic radius of Mg 2+ (0.86 Å) being about 30% smaller than that of Ca 2+ (1.14 Å) [52], the Mg 2+ substitution of Ca 2+ will increase the compress stress. This stress will be released through layer buckling and breaking along the [  [37]. On the other hand, Mg 2+ also decreases the crystallization rate of aragonite. The magnitude of the kinetic coefficient (β) is controlled by the density of kink sites along the step (n k ) and the net probability of attachment to a site ((exp(E k /kT), where k and T represent the Boltzmann constant and Kelvin temperature, respectively, and E k denotes an effective barriers to attachment at a kink). Namely, β~n k exp(-E k /kT) [53]. As many nanoparticles form on aragonite (110) surfaces, the n k reaches the maximum value. Therefore, the exp(E k /kT), instead of the n k , is the major factor affecting β. Since the exp(E k /kT) is related to the desolvation of cations, which is the rate-limiting step during CaCO 3 growth, the dynamics of particle aggregation would be restricted when solvation layers are strongly bounded [44]. The desolvation of Mg 2+ is more difficult than that of Ca 2+ , resulting in a slower crystallization rate of aragonite in Mg 2+ -bearing solutions. As a result, the OA of nanoparticles aggregates along the [001] direction is much easier to be observed.
As supersaturations of solutions are positively correlated with growth rates of CaCO 3 [2], the aggregation rate of nanoparticles is obviously accelerated in solutions with higher supersaturation. However, the solid-solid interfaces among nanoparticles are still obvious under these conditions ( Figure 8 and Figure S3), indicating that it is difficult to eliminate the inhibition of Mg 2+ on the crystalline rate of CaCO 3 , even in highly supersaturated solutions (from 0.50 to 1.05).

Comparison with Previous Studies
The IS and Mg 2+ /Ca 2+ ratios in our study are different with those in seawater and some previous studies; nevertheless, our results can be reasonably inferred to CaCO 3 precipitation in the ocean. Although the IS in our study (0.1 M) is much lower than that in seawater (IS = 0.7 M), Zhong and Mucci (1989) discovered that salinity variations alone will not significantly affect both the precipitation rates and overgrowth compositions of calcite and aragonite [35]. In addition, the critical Mg 2+ /Ca 2+ ratios favoring calcite or aragonite ranges from 1.5 to 2 [30,54], and when the Mg 2+ /Ca 2+ ratios in solutions are less than 7.5, the distribution coefficients of Mg 2+ , both in mineral overgrowths and on their adsorbed layers, are positively correlated with Mg 2+ /Ca 2+ ratios of solutions [34]. Therefore, the experimental consequences obtained at Mg 2+ /Ca 2+ = 3 in this study can represent those in real seawater at Mg 2+ /Ca 2+ = 5.

Calcite Precipitation
The growth features of calcite in Mg 2+ -free solutions (Figure S1a-c) are used for comparison with previous studies. We discovered that the differences in terrace width between obtuse and acute steps (∆λ) are 98.10, 97.00 and 60.80 nm at SI calcite = 0.50, 0.83 and 1.05, respectively. Since the λ of steps is negatively related to the step velocity [3], the difference in step velocity between obtuse and acute steps is positively correlated with solution SI calcite at a Ca 2+ /a CO 2− 3 = 10, which is consistent with previous experimental and fitting data [16,17]. Additionally, when we convert SI calcite = 0.50 to supersaturation σ (SI calcite = log(exp(σ))), the σ value is equal to 1.15, and we found that the λ values of obtuse and acute steps under our experiment (175.43 and 77.33 nm) are slightly higher than those (152.67 and 75.00 nm) reported by Wasylenki et al. (2005), which may be caused by the minor differences of a Ca 2+ /a CO 2− 3 [55].
The inhibiting effect of Mg 2+ on the spreading rate of the calcite growth hillock in this study is similar to that in previous studies. Two basic impurity models for the inhibiting effect of Mg 2+ on the growth of calcite (104) surfaces are widely accepted. One is step pinning [56,57], while the other is incorporation [57,58]. In the former model, the adsorption of Mg 2+ on step sites is a reversible process, while Mg 2+ incorporated into the calcite lattice cannot be reversed in the latter. Based on the inhibiting effect of a newly formed monolayer on the growth of subsequent monolayers, Astilleros et al. (2010) proposed a new solid solution-aqueous solution model, providing an explanation for the formation of "dead zones" in the step pinning model [59]. According to our AFM results (Figure 2b-d), these three models probably coexist. The Mg 2+ ions can pin at non-specific sites, while only incorporated at specific sites. Additionally, the surface properties of the first monolayer formed on the calcite (104) surface would be altered upon Mg 2+ incorporation, causing the growth inhibition of subsequent monolayers. Some researchers have declared that the incorporation of Mg 2+ into calcite leads to the increase in mineral solubility, which contributes to aragonite precipitation [33]. However, based on the thermodynamic simulation, the increase of surface energy by incorporating Mg 2+ into calcite is the main reason for the rise of the energy barrier (∆G C ) to form a critical nucleus of calcite [30]. Similarly, the spreading rates of calcite steps are negatively correlated with the free energy (∆g ± ) (∆g ± = -(L/b)∆µ + 2c<γ> ± , where L, b, ∆µ, c and <γ> ± denote step length, 6.4 Å intermolecular distance along the steps, chemical potential, 3.1 Å distance between rows and step edge free energies along the + and − steps, respectively.) [36]. When the ∆g ± increases with the incorporation of Mg 2+ into the steps, the compress stress increases and the spreading rates are inhibited.
The overgrowth compositions of calcite in Mg 2+ -bearing solutions are also compared. Since the concentration of Mg 2+ in calcite overgrowth is independent of precipitation rates [34], the SI calcite of solutions will not change Mg 2+ contents in overgrowth. The contents of Mg 2+ semi-quantified in our study (24.68-43.20 at%) are in accordance with Hong et al. (2016)'s calculation that the maximal Mg 2+ content which a calcite lattice can sustain before plastic deformation is around 40 at% [37]. However, our value is greater than that obtained in a previous study using the constant disequilibrium technique (crystal-seed systems) (7-10 at%), which is attributed to the effects of the surface or adsorbed layer. As the Mg 2+ /Ca 2+ ratios in the surface or adsorbed layer are higher than those in the overgrowth layer, Mg 2+ contents detected in precipitates are greater when the thickness of the overgrowth layer is extremely thin [34].

Aragonite Precipitation
The influence of Mg 2+ on aragonite precipitation is a controversial topic [33,39,40], which might be partly ascribed to the limitation of research techniques in the late 20th century. These studies determined the precipitation rate by measuring weight change per unit of time, neglecting the reaction pathways. Fortunately, the in situ AFM has been successfully employed to monitor chemical reaction pathways in the past decades [3,9,16,17,20,36,37,53,55]. Our in situ AFM study discovered that the interfaces among nanoparticles formed on aragonite (110) substrates in solutions with Mg 2+ /Ca 2+ = 3 are more obvious than those in Mg 2+ -free solutions, demonstrating that the precipitation rate of aragonite is actually inhibited by Mg 2+ through retarding the dehydration rate of precursors. However, since these previous experiments were carried out in extremely high supersaturated solutions for several hours, the inhibition of Mg 2+ on the precipitation rate of aragonite is not easily detected on the macro-scale. According to our investigation, the aragonite precipitated on aragonite (110) substrates undergoes the oriented attachment of nanoparticles, and thus impacts of Mg 2+ on these nanoparticles should not be ignored.
The stability of Mg 2+ in aragonite structure directly determines overgrowth compositions. Since the coordination number of Ca 2+ in aragonite structure is nine, Mg 2+ , commonly occurring in six and eight coordination, cannot be preserved in aragonite for a long time. Nevertheless, a small amount of Mg 2+ can be conserved in aragonite precursors (e.g., ACC and monoclinic aragonite, i.e., mAra), and the contents of Mg 2+ incorporated in structures are gradually released with the transformation of mineral phases [53,60]. This prolonged precipitation provides suitable conditions for releasing Mg 2+ from the particle precursors back into the solution, leading to low Mg 2+ contents in the ultimate aragonite crystals (<10 at%, which is the maximum value for mAra).

Implications for the Formation of Abiogenic Carbonate Cements
Our experimental results are consistent with previous inhibition models of Mg 2+ on the growth of calcite (104) surfaces. We discovered that the modification mechanisms of Mg 2+ on the growth of calcite (104) and aragonite (110) substrates act on the step sites and nanoparticles, respectively, providing a new perspective on the precipitation of abiogenic CaCO 3 in the ocean.
It was discovered that the ACC preferentially deposits at the edges of nacreous tablets in the early stage of precipitation, and then transforms into tower-like aragonite with time [61]. This precipitation behavior of CaCO 3 is similar to the template effect of soluble macromolecules (SM) on CaCO 3 aggregation [62], which is ascribed to the change in interfacial free energy by organic templates. The regulation of organic templates on nucleation widely exists in the formation of CaCO 3 in vivo or on organic membranes [63,64].
In shallow marine sedimentary environments, the precipitation of CaCO 3 can be divided into three stages. In the first stage, phytoplankton and zooplankton produce biogenic CaCO 3 shells through photosynthesis and metabolism [63], respectively, belonging to direct biological regulation. When these plankton die, the precipitation of CaCO 3 proceeds into the second stage, in which organic membranes on these shells chiefly control the formation of CaCO 3 cements. After these CaCO 3 cements cover the majority of organic membranes, the precipitation of CaCO 3 enters the third stage, in which abiogenic CaCO 3 cements directly overgrow on the CaCO 3 substrate.
The findings presented in this study mainly contribute to the CaCO 3 precipitation in the third stage, without direct or indirect biological regulation. We observed that elongated columnar aragonite in two layers with different orientations forms on aragonite (110) surfaces, while smooth layered calcite precipitates form on calcite (104) surfaces. That is to say, the heterogeneous nucleation and growth of CaCO 3 on aragonite are an important origin of acicular and fibrous aragonite cements, whereas those on calcite are conducive to the formation of the secondary enlargement of calcite cements. Additionally, Mg 2+ will inhibit the step growth of calcite and the dehydration rate of aragonite precursors, without changing the crystallization pathways of CaCO 3 on these two substrates, suggesting that the precipitation rate of CaCO 3 cements is subject to Mg 2+ /Ca 2+ ratios in the ocean. These discussions indicate that the distribution of different types of CaCO 3 in shallow-water sediments and Mg 2+ /Ca 2+ ratios in seawater chiefly control the formation and subsequent input flux of abiogenic CaCO 3 cements.

Conclusions
AFM, SEM, XRD and Raman analyses were utilized to investigate the heterogeneous nucleation and growth of CaCO 3 on calcite (104) and aragonite (110) surfaces in abiotic environments, leading to the following conclusions: (1) smooth layered calcite forms on calcite (104) surfaces; (2) elongated columnar aragonite generated by OA of nanoparticles, aggregating along the [001] directions, precipitates on aragonite (110) surfaces; (3) Mg 2+ inhibits the growth of aragonite and calcite formed on aragonite (110) and calcite (104) surfaces, by retarding the dehydration of precursors and blocking step growth, respectively, without affecting the crystallization pathways of CaCO 3 on these two substrates. The aforementioned conclusions suggest that different types of CaCO 3 in shallow-water sediments determine the mineralogy and morphology of abiotic CaCO 3 cements, and the lack of abiogenic CaCO 3 cements can be partly ascribed to the retardation of Mg 2+ on the crystallization rates of both calcite and aragonite.