Ion micro-probe imaging of the aragonite skeleton of Pavona clavus, a massive reef-building coral, shows that magnesium and strontium are distributed very differently. In contrast to strontium, the distribution of magnesium is strongly correlated with the fine-scale structure of the skeleton and corresponds to the layered organization of aragonite fibers surrounding the centers of calcification, which have up to ten times higher magnesium concentration. This indicates a strong biological control over the magnesium composition of all structural components within the skeleton. Magnesium may be used by the coral to actively control the growth of the different skeletal crystal components.
 Scleractinian corals are among the most prolific biomineralizing organisms on Earth [Lowenstam and Weiner, 1989] and massive, reef-building corals are used extensively as proxies for past variations in the global climate. It is therefore of wide interest to understand the degree to which biological versus inorganic processes control the chemistry of the coral skeleton. Early workers considered aragonitic (CaCO3) coral skeleton formation as a purely physiochemical process; essentially an inorganic precipitation of aragonite from a supersaturated solution [Bryan and Hill, 1941]. More recent studies have increasingly emphasized the role of a skeletal organic matrix, or intercalated organic macro-molecules that control the macroscopic shape and size of the growing crystals. It is now well established that organic compounds play a key role in controlling the morphology of crystals in a wide variety of calcium carbonate biomineralization processes by binding to specific sites and thereby causing direction-specific binding energies on the crystal surfaces [Aizenberg et al., 2002; Berman et al., 1993; Cuif et al., 2003; Gotliv et al., 2003; Johnston, 1980; Lowenstam and Weiner, 1989; Orme et al., 2001; Weiner et al., 2003]. Macro-molecules, such as aspartic acid-rich or glutamic proteins [Constantz and Weiner, 1988; Mitterer, 1978; Young, 1971] and sulfated polysaccharides [e.g., Cuif et al., 2003] are known to be embedded within the aragonitic skeletal components of coral. In addition, endosymbiotic algae and the layer of cells adjacent to the mineralizing surface, the calicoblastic ectoderm, are believed to play important roles in driving and controlling hermatypic coral skeletogenesis [Barnes, 1970; Clode and Marshall, 2002; Johnston, 1980]. However, until recently, it was not possible to obtain trace-elemental analyses of a coral skeleton with sufficiently high spatial resolution and sensitivity to correlate chemical variations with the micrometer (μm) scale organization of its different structural components.
 In coral skeletons, individual corallites are composed of two primary structural components (Figure 1). Centers of calcification (COC) are μm-sized, rounded aggregates of randomly oriented, nanometer-sized calcium carbonate crystals embedded in an organic matrix of sulfated polysaccharides and acidic proteins [Constantz and Weiner, 1988; Cuif and Dauphin, 1998; Cuif et al., 2003; Gladfelter, 1982]. The spatial organization of individual COC determines the macroscopic morphology of the skeleton. The COC are overgrown by the second skeleton component, a composite of fibrous aragonite crystals and organic macro-molecules organized into growth layers [Constantz and Weiner, 1988; Cuif et al., 2003] (Figure 1c). This fibrous aragonite composite, of which organic macromolecules constitute on the order of 1 wt% [Cohen and McConnaughey, 2003; Cuif et al., 2004, 2003], constitutes the bulk of the skeleton and gives it mechanical strength. The processes involved in the formation of COC and fibrous aragonite are not well understood in detail.
2. Materials and Methods
 Using a NanoSIMS NS50 ion microprobe (SIMS: Secondary Ion Mass Spectrometry) [Hillion et al., 1993], we produced images of the Mg and Sr distributions in the COC and adjacent fibrous aragonite in a specimen of the massive, hermatypic shallow water coral Pavona clavus, which is used frequently for paleo-climate reconstructions. Our coral specimen was collected live at Bartolomé, Galápagos Islands in January 1995 at a water depth of 3 meters [Wellington et al., 1996]. A thick-section of this coral was cut perpendicular to the growth direction, mounted in epoxy and polished to a 0.25 micrometer surface finish using diamond and silica colloidal suspensions. During analyses a 16 keV O− beam of about 10 pA, focused to a diameter of about 0.4 μm, was rastered across a pre-sputtered area on the gold coated sample. Secondary ions of 24Mg+, 40Ca+ and 88Sr+ were collected simultaneously in electron multipliers. The mass resolving power (M/ΔM) of the mass spectrometer was about 3000 during all analyses; no mass-interferences on the measured peaks were observed. Magnesium and Sr data were calibrated against electron microprobe analyses of the average Mg and Sr concentration in the fibrous aragonite. We used the JEOL Superprobe 733 at Stanford University operated at 15kV with a beam current of 15nA; dolomite, calcite, and strontianite crystals of known composition were used as standards.
3. Results and Discussion
 Results are shown in Figures 2 and 3. Several observations arise from these data. First, there is a striking μm-scale zoning in Mg concentration in the fibrous aragonite composite (Figure 2) that corresponds closely with the layered structure of this material (Figure 1c). Within the regular zoning pattern the Mg concentration varies by as much as a factor of two on a characteristic length scale of a few μm (Figure 3b). This distinct Mg zoning pattern is ubiquitous in the skeleton. Second, the COC (Figure 2a) have up to one order of magnitude higher Mg concentration than the surrounding fibrous aragonite composite (Figure 3a) and the very high Mg concentrations are strongly localized to the COC (Figure 2a). Such high Mg concentrations in COC have not previously been observed, most likely because of the limited spatial resolution of the analytical techniques applied, which may not have allowed the COC to be completely resolved. Third, in contrast to Mg, the distribution of Sr is more smooth and does not display the same degree of zoning; the distributions of Mg and Sr are not correlated on shorter length scales. Consistent with previous observations [Cohen and McConnaughey, 2003], the Sr concentration increases smoothly towards the COC, forming a “halo” of relatively high Sr surrounding the COC (Figure 3b). In contrast with Mg, the higher Sr concentrations are not strongly localized to the COC (Figure 3b).
 It is not likely that the observed variations in Mg concentration in the Pavona coral are simply due to localized organic matter in the skeleton; i.e., a strong matrix effect affecting the yield of secondary ions. If this was the case, variations in the Mg and Ca signals would be correlated. This is not observed (Figure 3a). Figure 3b demonstrates that normalizing Mg to Ca preserves the clear oscillatory banding in the Mg signal evident in Figure 2. Instead, the banded Mg micro-distribution might reflect either systematic differences in the amount of Mg in solid solution within the aragonite crystal structure and/or the preferential concentration of Mg at the surfaces of individual aragonite fibers at the initiation or termination of each fibrous growth layer (Figure 1c), as it has been observed in mollusks [Rosenberg et al., 2001] and in foraminifera [Eggins et al., 2004; Erez, 2003]. The presence of surface adsorbed Mg in the COC would be consistent with their high Mg concentration, compared with the surrounding aragonite fibers (Figure 3), due to the larger surface to volume ratio of their constituent nanometer-sized crystals. Another possibility is that the small crystals making up the COC in Pavona clavus are calcitic, as observed in Mussa angulosa [Constantz and Meike, 1988]. However a recent Raman spectroscopy study of 15 coral species, including Mussa angulosa, did not find evidence for calcite [Cuif and Dauphin, 1998].
 The time-span involved in formation of the skeletal structures mapped in Figure 2 is on the order of hours to one day [Barnes, 1972; Risk and Pearce, 1992], too short for any external environmental parameters to change significantly. Diurnal temperature changes of up to 1–2°C are possible at Bartolomé. Such temperature variations would correspond to changes in the Mg concentration of less than 5% in thermodynamic equilibrium between aragonite and sea-water [Mitshuguchi et al., 1996]; even smaller variations would be expected for Sr. Therefore the Mg and Sr concentration variations documented in Figures 2 and 3 can not be explained by changes in the thermodynamic equilibrium between aragonite and sea-water. These variations are more likely to result from a strong biological control over the flux of both Mg and Sr to the surface of the growing aragonite skeleton.
Cuif et al.  observed a layered distribution of sulfur associated with sulfated polysaccharides embedded within the fibrous aragonite layers and corresponding to the structural growth layers; i.e., similar to the distribution of Mg (Figure 2). It was inferred that cyclic secretion of polysaccharides is directly involved in the mineralization process [Cuif et al., 2003]. Qualitatively similar observations of banded or layered S and Mg distributions have been observed in bivalves [Rosenberg et al., 2001] which are forming their shells under strict biological control [Lowenstam and Weiner, 1989] and in the absence of symbiotic algae. In the bivalve shell, Mg and S are preferentially concentrated at the termination of elongated calcite crystals. Furthermore, these elongated calcite crystals are aligned in a way similar to the aragonite fibers shown in Figure 1c, which give rise to distinct Mg and S banding [Rosenberg et al., 2001], very similar to that observed in corals ([Cuif et al., 2003] and this study). For the bivalve it was inferred that Mg and S-rich organic components are used actively by the organism to control crystal size and morphology [Rosenberg et al., 2001].
 Magnesium is known to play a crucial role in regulating the formation of different calcium carbonate phases [Davis et al., 2000; Raz et al., 2000]. Recent advances in the study of calcium carbonate skeletons formed by a variety of organisms have revealed the widespread presence of amorphous calcium carbonate (ACC), either as a stable structural component in the skeleton, or as a transient precursor phase to crystalline polymorphs of calcium carbonate [Addadi et al., 2003; Aizenberg et al., 2002; Raz et al., 2000; Weiner et al., 2003]. Magnesium is an important component of ACC, which can have different degrees of short range order [Raz et al., 2000; Weiner et al., 2003]. In conjunction with organic macro-molecules, Mg is known to stabilize this otherwise inherently unstable phase [Addadi et al., 2003]. ACC can therefore either act as nucleation sites for calcium carbonate crystals, or cause their formation by subsequent transformation into crystalline polymorphs [Addadi et al., 2003; Aizenberg et al., 2002; Weiner et al., 2003]. The unstable and amorphous nature of ACC makes its identification in crystalline calcium carbonate skeletons difficult [Weiner et al., 2003]. However, it is entirely possible that Mg-rich ACC is transiently present in the coral skeleton, where it provides a controlled environment for the nucleation and growth of individual aragonite crystals in the fibrous aragonite layers as well as in COC [Cohen and McConnaughey, 2003]. The Mg-rich bands in the layered fibrous aragonite composite and the high magnesium concentration in COC (Figure 2) might in part be the trace element signature of transient ACC in both skeleton components. Isotopic labeling and wash-out (efflux) experiments using 45Ca have provided strong evidence for the presence of an unstable calcium carbonate phase in newly formed coral skeleton [Tambutté et al., 1996]. Thus, we propose that temporally and spatially controlled release of Mg to the mineralizing surfaces plays a key role in the formation of the coral skeleton.
 The main feature of the Sr profile shown in Figure 3b is the gradual increase in Sr concentration from the fibrous aragonite crystals into the domain of the COC, forming a “halo” of relatively high Sr concentrations around the COC, qualitatively consistent with previous observations [Cohen and McConnaughey, 2003]. Isotopic labeling experiments have demonstrated that Sr2+ and Ca2+ are delivered to the mineralizing surface through the same voltage-sensitive, transcellular pathway located in the calicoblastic ectoderm [Ferrier-Pages et al., 2002; Tambutté et al., 1996]. Furthermore, incorporation of Sr into the skeleton is inversely correlated with the rate of calcification [Ferrier-Pages et al., 2002]. As a consequence of the enhanced calcification rate, aragonite precipitated during the daytime is characterized by lower Sr/Ca ratios than aragonite precipitated during the night. The COC are believed to be precipitated at night and expected to be characterized by relatively high Sr/Ca ratios, as observed [Cohen and McConnaughey, 2003].
 More importantly, the observations presented here indicate that Mg is delivered to the surface of the growing skeleton by a different and more actively controlled mechanism. Magnesium can either be actively pumped through the calicoblastic ectoderm to the mineralizing surface, or be introduced by allowing sea-water to enter the mineralizing zone directly, such as it has been observed in foraminifera [Erez, 2003]. Either way provides a mechanism for the coral to dynamically control the Mg concentration in the mineralizing zone.
 We thank Bob Jones for help with electron microprobe analyses. AM is grateful for inspiring discussions with Gary Rosenberg, Anne Cohen, Denis Allemand, Christine Ferrier-Pages, and Dan Sinclair. Steve Weiner and Patricia Dove are thanked for thorough and highly constructive reviews. This work was supported by a grant from the Stanford Institute for the Environment, through the Environmental Interdisciplinary Initiative Program.