Fig. 1. Schematic diagrams illustrating the concept of storage capacity (indicated by bold curves) for a hypothetical nominally anhydrous mineral (or assemblage of minerals) with a finite capacity to incorporate H₂O. (A) Shows relations at low pressure, where the mineral coexists with an H₂O-rich fluid at a low temperature and a hydrous silicate melt at a high temperature. The former condition would traditionally be considered the solubility of H₂O in the mineral and the latter would traditionally represent hydrous melting, but in both cases, the maximum H₂O sequestered in the mineral at a given temperature is the storage capacity. (B) Shows relations at a high pressure, where the fluid changes continuously from H₂O-rich at a low temperature to silicate-rich at a high temperature. Again, the storage capacity is the maximum H₂O that can be sequestered in the mineral. Although other scenarios are possible, the storage capacity may increase with temperature, reach a maximum, and then diminish with further temperature increases. The increasing trend at a low temperature would be owing to the common increase in mutual solubility with temperature for immiscible phases. The decreasing trend at a high temperature is owing to the reduced activity of H₂O in increasingly silicate-rich fluid and because the storage capacity must be zero at the melting point of the dry mineral (or at the dry liquidus, for an assemblage of minerals). These diagrams are schematic and for clarity exaggerate the H₂O storage capacity of nominally anhydrous minerals.

Fig. 2. Storage capacity in mantle olivine as a function of depth. FTIR data from annealing experiments at 1100 °C from Kohlstedt et al. [1], multiplied by a factor of 3 (see text). Also shown are SIMS data from hydrous experiments at 1200 °C from Chen et al. [41]. Curve is a least squares fit of Eq. (1) to the adjusted data of Kohlstedt et al. [1], with a = − 1.194, b = 2.263, and c = 0.128. The composition of olivine is Mg_0.90Fe_0.10SiO₄ in the study of Kohlstedt et al. [1], and Mg_0.91Fe_0.89SiO₄ in Chen et al. [41], except for the highest pressure datum, which is pure Mg₂SiO₄.

Fig. 3. H₂O concentrations of coexisting olivine and aluminous pyroxenes from high pressure experiments (1–2 GPa) [26], [43], [52] and [53]. H₂O concentrations of cpx and opx are between 5 and 20 times that in coexisting olivine and and cluster near 10. Also shown for reference are compositions of natural cpx, opx, and coexisting olivine from mantle xenoliths [64], though these may have been affected by post-eruption dehydrogenation.

Fig. 4. Experimental constraints on storage capacity of low Ca pyroxene as a function of depth and compared to storage capacity of olivine (from Fig. 2). MgSiO₃ enstatite [44] and [87] and clinoenstatite [56] and [57] have storage capacities similar to olivine. Data for aluminous orthopyroxene with up to 1 wt.% Al₂O₃ at 1.5 GPa [44] illustrates that aluminous orthopyroxene likely has a storage capacity much greater than MgSiO₃ or olivine. Data from natural Tanzanian enstatite (1.5 wt.% Al₂O₃) under oxidizing conditions [44] indicates storage capacities intermediate between MgSiO₃ and synthetic aluminous MgSiO₃. Also shown is the model storage capacity of mantle orthopyroxene from trend “C” in Fig. 6 (see text).

Fig. 5. Modal mineralogy of the upper mantle with depth, from the experiments of Irifune and Isshiki [72]. The large increase in garnet beginning at about 12 GPa corresponds to increasing stabilization of the majorite component.

Fig. 6. Model storage capacity of the mantle for different assumptions discussed in the text. Trends A, B and C use the variations in mantle mode with depth from Fig. 5 and assume that the olivine storage capacity is given by Eq. (1) and the trend in Fig. 2. Trend “A” is realized if olivine and pyroxene have the same storage capacity and if the storage capacity of garnet is the same as olivine up to a limit of 2000 ppm H₂O and then stay constant. This trend is believed to be a realistic minimum for the upper mantle storage capacity. Trend “B” represents a plausible maximum storage capacity trend and assumes , . Trend “C” is the preferred model, as described in the text. Local maxima in trends “B” and “C” result from increasing garnet mode at the expense of pyroxene (Fig. 5), and the maxima is more pronounced in “C” owing to assumed diminished preference of H₂O for pyroxene with increasing depth. Also shown is a storage capacity limit calculated by extrapolation of the cryoscopic approximation shown along a ridge geotherm in Fig. 8.

Fig. 7. Storage capacity of wadsleyite as a function of temperature for (A) Mg₂SiO₄ [36], [56], [74], [75], [78] and [88] and (B) (Mg,Fe)₂SiO₄ [1], [31], [41], [50] and [89]. Mg/(Mg + Fe) of wadsleyite in (B) ranges from 0.875 to 0.93. Concentrations determined by SIMS, except for those from Gasparik [36] and data below 1300 °C from Litasov and Ohtani [75], which were estimated from Mg / Si ratios; those of Kohlstedt et al. [1], Jacobsen et al. [88] and the data of Litasov and Ohtani [75] above 1300 °C, which were determined by FTIR, and the datum of Kohn [78], which derives from NMR. The FTIR data from [1] represent minimum estimated concentrations. Data of Smyth et al. [50] and Smyth and Kawamoto [89] are for monoclinic wadsleyite II. H₂O storage capacities of Mg₂SiO₄ correlate negatively with temperature and are therefore small at temperatures appropriate for the transition zone > 1500 °C. However, storage capacities of Fe-bearing wadsleyite are systematically greater than for Mg₂SiO₄, and remain substantial to high temperature.

Fig. 8. Effect of initial melting owing to intersection of mantle adiabats with the H₂O storage capacity (“dehydration solidus”) along adiabats appropriate for the oceanic mantle, including potential temperatures (T_p) of 1350, 1450, and 1550 °C. Calculations are done by application of the cryoscopic equation according to the partition coefficients and methods of Aubaud et al. [26] and with adiabats of 15 °C/GPa. Heavy curves show plausible depths of initial melting beneath ridges (1350 °C, 50–200 ppm H₂O) and oceanic islands (1550 °C, 300–1000 ppm H₂O).

Fig. 9. Cartoon showing melting of wet peridotite after advection from the transition zone into the upper mantle. Shaded regions show ranges of likely storage capacities of the upper mantle (from Fig. 6) and the uppermost transition zone (see text). Gray curve is the preferred model storage capacity of the upper mantle (trend “C” from Fig. 6). Peridotite advected from the transition zone into the upper mantle will partially melt if its H₂O content exceeds the storage capacity of the upper mantle at 410 km. Melting will extract H₂O from the residual peridotite until the storage capacity for those conditions is reached; then no further melting can occur at that temperature and pressure. Note that this condition occurs at a minimum of 0.4 wt.% H₂O, meaning the partially dehydrated peridotite will retain substantial H₂O in the deeper parts of the upper mantle. Further upwelling results in continued melting unless there is a local minimum in the storage capacity, in which case melting resumes when the storage capacity is again exceeded (at 275 km in the example in the figure). Then, further upwelling is accompanied by small amounts of dehydration melting. The dashed line labeled “MORB source” illustrates the average H₂O composition of the upper mantle (50–200 ppm) as inferred from MORB basalts [7], [8], [9] and [10]. The large difference between the likely storage capacity at 410 km and the average upper mantle concentration indicates that residues of dehydration melting cannot be a principal source for upper mantle material.

Fig. 10. Thickness of the 410 km discontinuity as a function of bulk H₂O content, calculated according to the method of Wood [20], but with values of equal to 1, 5, and 10. Wood [20] assumed , but if partitioning of H₂O between wadsleyite and olivine is less extreme, then H₂O has a much more modest effect on the thickness of the transition.