Although most reconstructions of the evolution of the Mg to Ca ratio in seawater conclude that it has increased during the course of the Cenozoic, they disagree widely regarding the magnitude of this change. On the basis of fluid inclusion and CaCO3 mineralogy observations, the increase was at least threefold. On the basis of Mg content of foraminifera shells it was only a factor of 1.7. A recently published reconstruction based on the Mg content of calcite fillings of voids in ridge flank basalts lends support to the conclusion that the change was severalfold. But as it is very difficult to come up with a plausible geologic scenario which could account for such a large change, we lean toward the smaller estimate based on the magnesium content of foraminifera shells.
 Reconstructing the history of the composition of sea salt has proved a daunting challenge. The difficulties involved are well illustrated by attempts to determine how the Mg to Ca ratio has evolved. We know that the major source of magnesium is the weathering of igneous rock and that the major source of calcium is the dissolution of limestone. We also know that the major sink for calcium is the deposition of CaCO3, and for magnesium it is the removal in ridge crest hydrothermal systems. We know that magnesium is also removed during the formation of dolomite. Finally, on the basis of the ratio of the amounts of Mg and Ca presently dissolved in the sea to the yearly amounts currently being added by rivers, we know that the apparent steady state turnover time of Mg is about 13 Myr and that of Ca is about 1 Myr.
 Several lines of evidence suggest that the Mg to Ca ratio in sea salt has increased over the past few tens of millions of years. Most convincing perhaps is the observation that 40 or so Myr ago, the organisms responsible for forming massive CaCO3 reefs switched from precipitating calcite to precipitating aragonite [Hardie, 1996]. Laboratory experiments suggest that this change could have been triggered by an increase in Mg to Ca ratio in seawater. If this is the explanation, then prior to 40 Myr ago, the Mg to Ca ratio must have been more than 3 times lower than it is today. Of course, other explanations exist which might explain these switches in mineralogy [Adabi, 2004]. Support for this large increase comes from Mg to Ca ratios in fluid inclusions trapped in marine halite [Lowenstein et al., 2001; Horita et al., 2002; Timofeeff et al., 2006]. However, as only a few such measurements have been made and as a large evaporational enrichment of the salt must have preceded halite precipitation, it is not clear how seriously these results should be taken.
 Support for this large change in Mg to Ca ratio is, however, not forthcoming from measurements of the Mg content of early Cenozoic shells of either planktic or benthic foraminifera. Rather than a severalfold increase during the course of the Cenozoic, these measurements suggest that the Mg to Ca ratio has increased by no more than a factor of 1.7. Measurements of Mg/Ca ratios in planktonic foraminiferal shells from tropics show a value of ∼5.5 mmol/mol at ∼55 Ma [Zachos et al., 2003]. Using today's seawater Mg/Ca and the thermometric equation (Mg/Ca (mmol/mol) = 0.38exp(0.1T)) from Elderfield and Ganssen , assuming no change in seawater chemistry, planktonic Mg/Ca at 55 Ma suggests a sea surface temperature of ∼27°C, similar to today's value. However, results from an independent temperature proxy (TEX86) suggest that the tropical Pacific was about 4°C–5°C warmer than it is today [Pearson et al., 2007]. If so, seawater Mg/Ca at 55 Ma was lower than today's seawater value of 5.1 mol/mol. If surface water was 4.5°C warmer at 55 Ma, seawater Mg/Ca at that time would be 3.02 mol/mol, a factor of 1.7 lower than today's ratio. On the other hand, if the temperatures were similar to today's, no change in the Mg to Ca ratio would be necessary. A similar result is obtained from the Mg to Ca ratio in benthic foraminifera, which suggests that over the last 49 Myr the ratio in seawater has remained in the range 3.3 to 4.6 mol/mol [Lear et al., 2002]. In this connection, it should be mentioned that measurements on echinoderms [Dickson, 2002] suggest a severalfold change. But unlike the results on foraminifera, these results are exceedingly noisy.
 Another indication that the increase cannot have been a factor of 3 or more comes from the consideration of how it might have been accomplished. There are four knobs which might be turned: the Mg to Ca ratio in river water, the rate of seafloor spreading, the rate of dolomite formation, and the residence time of calcium in the sea (see Figure 1). Of these four, we have evidence only that the rate of dolomite formation decreased over the course of the Cenozoic. Wilkinson and Algeo  point to the near absence of dolomite in the last 100 Myr. As most dolomites are found in platform sediments, this is consistent with the transition from a hothouse to an icehouse world. Presumably, the buildup of ice on Antarctica caused the sea level to drop, and as a result, the extent of shallow seas decreased.
 However, if this were the only change, then prior to its demise, dolomite formation must have dominated Mg removal from the ocean. In order to create a threefold rise in the Mg to Ca ratio, dolomite formation would have had to dominate the removal of magnesium prior to 40 Myr ago and then after that time would have dropped to negligible. Of course, to the extent that changes in the other three knobs may well have contributed, the required demise in uptake by dolomite would not have been so large. But it must be kept in mind that as no large change in the rate of seafloor spreading appears to have occurred [Rowley, 2002; Müller et al., 2008; Seton et al., 2009], at least one of these three can seemingly be crossed off the list as the dominant cause.
 Another way to look at this question is from the tectonic viewpoint. The collision of northward moving India and Africa with Eurasia certainly triggered one of the most pronounced disruptions of Earth surface processes experienced during the Phanerozoic. As pointed out by Edmond , it resulted in the largest Phanerozoic change in the isotopic composition of oceanic strontium. Plate motions reoriented. The CO2 content of the atmosphere began an order of magnitude decrease [Pearson and Palmer, 2000; Pagani et al., 2005]. It would be surprising if these pronounced changes did not lead to a significant change in the ocean's chemical composition. But how big a change? Did the Mg to Ca ratio increase by a factor of at least 3 as suggested by the switch from calcitic to aragonitic reefs or by a factor of 1.7 or less as indicated by the Mg content of foraminifera shells?
2. Evidence From Calcites From Ridge Flank Basalts
Coggon et al.  provide evidence from a previously untapped archive. They made measurements of the magnesium content in calcite veins formed in cracks and voids in ocean ridge flank basalts. On the basis of these results, they conclude that after remaining more or less constant from 170 to about 46 Myr ago, the Mg to Ca ratio in seawater underwent a threefold increase. On the basis of a reanalysis of their raw data, we conclude that their results suggest an even larger fivefold increase during the last 46 Myr. The difference between our analysis and that of Coggon et al.  stems from their failure to include two key data points for the Juan de Fuca Ridge (see below). As explaining even a threefold change in Mg to Ca ratio proves very difficult, explaining a fivefold increase is nigh unto impossible. Further, if the results for 15 Myr old calcites are included, then this large change appears to have occurred during the last 15 Myr (i.e., in a time period comparable to the calculated 13 Myr residence time of Mg in the ocean).
 On the basis of strontium isotope measurements, Coggon et al.  show that the composition of the interstitial water in which their calcites formed was modified by interaction with basalt. Further, they show that the extent of this modification is related to the degree to which the water was warmed by contact with the basalts. They estimate the magnitude of this warming on the basis of the oxygen isotope composition of the calcite and demonstrate through analysis of the 87Sr/86Sr ratio for the strontium contained in the calcites that warming increased the extent of interaction with the host basalt. For the Juan de Fuca Ridge samples, the warmer the temperature at which the calcite formed, the larger the isotope ratio shift toward that in basalt (see Figure 2). Hence, their strategy was to linearly extrapolate the magnesium content versus oxygen-isotope-based formation temperature relationship back to the ambient deep-sea temperature. However, in the case of the Juan de Fuca Ridge data set, they deviated from this procedure. Instead of using the magnesium contents of two calcites formed at temperatures closest to that of deep seawater, they substituted seawater analyses reported by Elderfield et al. . To our minds, there is a circularity in this substitution. By placing water data and calcite data together on the same graph, they imply that they somehow know the distribution coefficient of magnesium between seawater and calcite. But they present no justification for their choice of a coefficient. The scatter in the data for their warmer temperature calcites (r2 = 0.36) precludes any meaningful extrapolation to the bottom water temperature. As this extrapolation is key to their conclusions, one has to question why they substituted Elderfield et al.'s  water data for their own calcite measurements. They note only that the two excluded low-temperature calcite samples came from a diabase intrusion into the sediment overlying the basalt rather than from within the basalt itself.
 We have made similar plots of their raw data for the samples from sites where the seafloor sites have ages of 46, 110, 129, 132, and 170 Ma (Figure 2) and extrapolated the magnesium content–temperature relationships for each site to the bottom water temperature at the time (Tables 1a and 1b and Figure 3). We are aware that at all these times the deep sea temperature was likely warmer than today's, but as the slopes of magnesium content versus temperature relationships are quite small, the choice of deep ocean temperature is not critical to our argument. For five of these sites we get an extrapolated magnesium concentration close to 6 ± 1 parts per thousand (ppt). On the basis of the intercept of 31 ppt for the Juan de Fuca data set, these results suggest that the Mg to Ca ratio in seawater is 5 times higher today than it was between 170 and 46 Myr ago (Figure 2). At a sixth time horizon (15 Myr old crust), we get a value of 3.3 ppt. But as all the δ18O-based formation temperatures are 38°C or warmer, this result is suspect.
Table 1a. Correlations of Calcite 87/86Sr and Mg Content Versus Temperaturea
Basement Age (Ma)
A × 106 (°C−1)
A (ppt °C−1)
JdFR, Juan de Fuca Ridge; A and B, coefficients in the regressive equation.
Calcite formation temperatures are all greater than 38°C; hence, interpolation likely yields a minimum Mg content.
 For the 6.9 Myr old site, there is a calcite which formed quite close to the temperature of the bottom water at the time. It has a magnesium content close to 15 ppt. Taken together with the Juan de Fuca result of 31 ppt, this suggests a doubling of Mg to Ca ratio in 7 Myr. This would be remarkable as the steady state residence time for Mg based on the ratio of the seawater inventory to the river input is about 13 Myr.
 We conclude that any interpretation of Coggon et al.'s  data set must rest heavily on the results for three calcite samples which formed at temperatures close enough to that of deep seawater to yield a meaningful calibration. Coggon et al.  reject the two from Juan de Fuca (their youngest locale). So, in a sense, their calibration rests on the result from a single calcite at the 6.9 Myr old locale. Unfortunately, the only other site with an age falling within the last 45 Myr (i.e., Site 1256 with basement age of 15 Ma) has no calcite samples formed at low enough temperatures to yield reliable information.
 It must be mentioned that Coggon et al.  point to support provided by the mineralogical record contained in CaCO3 cements. Prior to 40 Ma, reefs formed of CaCO3 were largely calcitic, and after 40 Ma they were largely aragonitic [Hardie, 1996]. Laboratory studies show that the crystal form of calcium carbonate depends on the Mg to Ca ratio in the artificial seawater from which they were precipitated [e.g., Oomori et al., 1987]. For ratios below 1.5, calcite forms. For ratios above 1.5, aragonite forms. The CaCO3 veins at the 6.9 Ma and younger sites are predominantly aragonitic, and with one exception, all those in the 46 Ma and older crust are calcitic. For the 15 Myr old site, there are roughly equal numbers of calcites and aragonites. So the record in ridge crest CaCO3 is more or less consistent with the Hardie  reconstruction. But as already mentioned, other explanations for the mineralogy changes have been proposed [Adabi, 2004].
 A possible alternate explanation for the Coggon et al.  results is to call on postformation diagenesis to lower the magnesium contents of the calcites gradually toward equilibrium with the surrounding basalts. If this were the case, it would be expected that the strontium isotope compositions should be shifted toward that of basalt (0.703). But as shown in Figure 4, the extrapolated isotope ratios are close to those for seawater at the time the basalts formed. It could, of course, be postulated that all the CaCO3 reefs were initially aragonite and that the older ones have recrystallized to calcite. But it would have to be assumed that although this recrystallization led to a loss of Mg, it produced no change in the isotopic compositions of strontium or oxygen. Although a stretch, it could explain why the Mg to Ca ratio in the older calcites is so uniform [see Fantle and DePaolo, 2006]. Were several million years required for the recrystallization, then the time history of Coggon et al.'s  results could be explained in this way [see Baker et al., 1991].
 Our preferred explanation involves what Watson  refers to as surface entrapment. He argues that in the absence of diffusion into the interior of growing crystals, trace elements do not achieve their thermodynamic equilibrium concentration. By coincidence, W. S. Broecker was the first to provide evidence that this process occurs. As reproduced in Figure 5, Hamza and Broecker  demonstrated that when CO2 gas is exchanged with the surface layer of optical calcite at 200°C, the 18O to 16O ratio in this layer is 5.4‰ greater than the thermodynamic equilibrium value calculated by Bottinga .
Carpenter and Lohmann  summarize the Mg and Sr contents of a series of Holocene calcite samples from the Bahamas, Jamaica, and Eniwetok (see Figure 6). As can be seen, the Mg contents of these calcites range from 50 ppt down to 10 ppt. As the Holocene calcites all formed at similar temperatures and in water with today's Mg to Ca ratio, this range must reflect some other factor. The most obvious one is growth rate. In this connection, it would help if the driving force for crystallization were known. Could it be warming of the host waters, or could it be increased alkalinity due to dissolution of the host basalt? On the same plot, we have added the Mg and Sr contents (extrapolated to the temperature of deep seawater) for the Juan de Fuca site, for the 6.9 Myr old seafloor site, and for the average for the older sites. As can be seen, these three points form a linear array and cover a range of a factor of 5 in Mg content. But if this is the explanation, it raises several questions. Why is the slope of the trend for the ridge crest calcites lower than that for the warm-water calcites? Why does the trend of these points correlate with seafloor age? Why do the Mg and Sr concentrations at any given site not scatter over the range seen for warm-water Holocene calcites? Why are the results at the older sites so similar to each other? In answer to the first question, the seafloor calcites formed at lower temperatures and much higher pressures. We have no answers to the other questions.
 The quest to determine how the Mg to Ca ratio in seawater has evolved over the course of the Cenozoic remains unfulfilled. The general consensus is that it has increased, but the magnitude of this increase is not agreed upon. Our interpretation of the Coggon et al.  reconstruction based on inorganic calcites formed in ridge flank basalts suggests that it was fivefold, and those based on foraminifera suggest that it was less than twofold. The switch from calcitic to aragonitic marine reefs and the limited set of fluid inclusion data lend support to the larger change. Considerations of the mechanisms by which the change could have been accomplished lend support to the smaller change. Clearly, if a definitive answer to this important question is to be obtained, more work must be done.