Experimental determination of C, F, and H partitioning between mantle minerals and carbonated basalt, CO2/Ba and CO2/Nb systematics of partial melting, and the CO2 contents of basaltic source regions

A. Rosenthal, E. H. Hauri, M. M. Hirschmann

Research output: Contribution to journalArticlepeer-review

135 Scopus citations


To determine partitioning of C between upper mantle silicate minerals and basaltic melts, we executed 26 experiments between 0.8 and 3 GPa and 1250-1500°C which yielded 37 mineral/glass pairs suitable for C analysis by secondary ion mass spectrometry (SIMS). To enhance detection limits, experiments were conducted with 13C-enriched bulk compositions. Independent measurements of 13C and 12C in coexisting phases produced two C partition coefficients for each mineral pair and allowed assessment of the approach to equilibrium during each experiment. Concentrations of C in olivine (ol), orthopyroxene (opx), clinopyroxene (cpx) and garnet (gt) range from 0.2 to 3.5 ppm, and resulting C partition coefficients for ol/melt, opx/melt, cpx/melt and gt/melt are, respectively, 0.0007±0.0004 (n=2), 0.0003±0.0002 (n=45), 0.0005±0.0004 (n=17) and 0.0001±0.00007 (n=5). The effective partition coefficient of C during partial melting of peridotite is 0.00055±0.00025, and therefore C is significantly more incompatible than Nb, slightly more compatible than Ba, and, among refractory trace elements, most similar in behavior to U or Th. Experiments also yielded partition coefficients for F and H between minerals and melts. Combining new and previous values of DFmineral/melt yields bulk DFperidotite/melt=0.011±0.002, which suggests that F behaves similarly to La during partial melting of peridotite. Values of DHpyx/melt correlate with tetrahedral Al along a trend consistent with previously published determinations.Small-degree partial melting of the mantle results in considerable CO2/Nb fractionation, which is likely the cause of high CO2/Nb evident in some Nb-rich oceanic basalts. CO2/Ba is much less easily fractionated, with incompatible-element-enriched partial melts having lower CO2/Ba than less enriched basalts. Comparison of calculated behavior of CO2, Nb, and Ba to systematics of oceanic basalts suggests that depleted (DMM-like) sources have 75±25 ppm CO2 (CO2/Nb=505±168, CO2/Ba=133±44), whereas enriched sources of intraplate basalts similar in concentrations to primitive mantle have 600±200 ppm CO2. If all mantle reservoirs are expressed in the current inventory of oceanic basalts for which nearly undegassed CO2 concentrations are available, then we estimate the likely range of mantle C concentrations to be 1.4-4.8×1023 grams of C, or 1.5-5.2 times the mass of the current C surface reservoir. Depending on the assumed Ba and Nb contents of average oceanic crust, resulting ridge fluxes of C range from 7.2×1013 to 2.9×1014 g/yr.

Original languageEnglish (US)
Pages (from-to)77-87
Number of pages11
JournalEarth and Planetary Science Letters
StatePublished - Feb 5 2015

Bibliographical note

Funding Information:
We thank Jianhua Wang, Frances Jenner, Paola Ardia, Fred Davis, Ben D. Stanley, Hongluo Zhang, and Tony Withers for help in the laboratory. M. Hartley kindly provided a spreadsheet with the compiled CO 2 –Ba–Nb data from Laki, Iceland, and A. Saal allowed us to use unpublished data from Siquieros melt inclusions. We thank D. Chavrit, an anonymous referee, and T. Elliott for their comments on the original draft of this manuscript. Financial support was provided by the Sloan Foundation through the Deep Carbon Observatory initiative and NSF grant EAR 1161023 . A.R. also received support from a Marie Curie International Incoming Fellowship ( 302637 ).

Publisher Copyright:
© 2014.


  • C partitioning
  • CO/Ba
  • CO/Nb
  • Deep Earth carbon cycle
  • Mantle carbon


Dive into the research topics of 'Experimental determination of C, F, and H partitioning between mantle minerals and carbonated basalt, CO2/Ba and CO2/Nb systematics of partial melting, and the CO2 contents of basaltic source regions'. Together they form a unique fingerprint.

Cite this