Jane Selverstone and Zachary Sharp University of New Mexico Chlorine isotope evidence for syn-subduction modification of serpentinites by interaction with sediment-derived fluid during exhumation Jane Selverstone and Zachary Sharp University of New Mexico NSF grant EAR 0911669

3 2 GPa 1 Zermatt-Saas ophiolite, Aosta, Italy 300 400 500 600°C Lago di Cignana: dia+coe Zermatt-Saas ophiolite, Aosta, Italy 3 2 Z-S ophiolite GPa ? 1 Combin P-T data from Cartwright & Barnicoat 2002, Li et al. 2004, Angiboust et al. 2009, Groppo et al. 2009 300 400 500 600°C

Chlorine Isotope Data A simple story? Zermatt-Saas data serpentinization via seawater serpentinization via sedimentary pore fluids Zermatt-Saas data HP serp and rodingite data are confined to ranges defined by Jaime Barnes’ work for modern seafloor serpentinites hydrated via seawater and sed pore fluids. At first glance, then, it looks as though there is no change in chlorine isotope composition during subduction, in accord with John et al. 2011. But, there is more to the story. A simple story?

Textural Types (Li et al. 2004) Type A seafloor pseudomorphic replacement Type B/C Type B Ol+TiClh mylonitization ± folding during subduction Type C recrystallization during initial exhumation Very detailed petrologic work by Li et al. defined five different stages of serpentinization and recrystallization in the Z-S ophiolite, and correlated these stages with different parts of the PTt path constructed from associated rock types. Our samples fall into 4 of these categories, with a few samples that show partial overprinting of younger on older. We made our initial correlations solely on the basis of textural features, but we see mineral and bulk chemical differences between the types as well: Type D greenschist overprint F.O.V. = 4 mm for all photos

A B C D Type A Serpentine analyses Type B Type C Type D For example, microprobe data show clear differences between the textural types, with a progression towards nearly endmember Mg contents and low Al in the samples with the texturally youngest serpentine. Type C Type D Serpentine compositions correlate with textural types Texturally youngest serp in each sample has lowest Al and highest Mg#

A Chlorine Isotopes Revisited 3 2 GPa 1 300 400 500 600°C LdC Serpentinite stages: Li et al. 2004, JMG LdC 3 2 GPa 1 Type A: few samples, but d37Cl is exactly what we expect for serpentinization via seawater. A 300 400 500 600°C

B A Chlorine Isotopes Revisited 3 2 GPa 1 300 400 500 600°C LdC Serpentinite stages: Li et al. 2004, JMG LdC 3 B A 2 GPa 1 Type B samples were mylonitized and recrystallized during subduction to depths of 75-80 km. They have isotopic compositions that are around 0‰ – the same as or somewhat lower than the Type A samples. 300 400 500 600°C

C B A Chlorine Isotopes Revisited 3 2 GPa 1 300 400 500 600°C LdC Serpentinite stages: Li et al. 2004, JMG LdC 3 B C A 2 GPa 1 Type C serpentinites were completely recrystallized during the early stages of exhumation. Their Cl isotope compositions are tightly clustered and are 1-2‰ lower than Type A and B values. 300 400 500 600°C

C B D A E Chlorine Isotopes Revisited 3 2 GPa 1 300 400 500 600°C LdC Serpentinite stages: Li et al. 2004, JMG LdC 3 B C E D A 2 GPa 1 1σ error in δ37Cl And finally, type D samples record recrystallization and veining associated with a late greenschist-facies overprint of the ophiolite and surrounding units. There is more scatter in this group, but d37Cl values of all samples are negative and extend to very low values (-2.7‰ in a vein sample). Now that we’ve seen that the Cl isotopes differ for samples that equilibrated during subduction vs. those that recrystallized during exhumation, let’s also look at bulk chemical data for the different groups. 300 400 500 600°C

Bulk Composition Constraints on Fluid History Multiple fluids? sed-derived? ??? Here I’ve lumped Type A and B together, but C and D are shown separately. For some elements there is a strong correlation between chemical and isotopic composition. There is also a good correlation between LOI values and d37Cl; note here that LOI>10-12% indicates a significant percentage of carbonate in the samples (confirmed in thin section). We also see that all of the Type C and D samples have very low concentrations of Cl – see possible explanations. In contrast to the other data shown here, Ca and Cr suggest that Type C and D serpentinites record interaction with different fluids. Hi-Cr samples may have seen fluid from other serps, lo-Cr, hi-Ca samples clearly record a different fluid source, likely from nearby metasedimentary rocks. Explore these possibilities in greater detail. High LOI = carbonate-bearing Single fluid?

Chlorine Isotopes Revisited A B C D Back to same box plot as before, showing Cl isotope compositions of the four serpentinite types.

Chlorine Isotopes Revisited A B C D Type C&D Ti-clh FIs All of the data I have shown have been bulk data. We also have data on the water-soluble chloride fraction from 3 samples with fluid inclusions and 1 vein sample, and they record fluids with d37Cl between -1 and -3‰. WS chlorides are 1-2‰ lower than silicate-bound chloride from same samples (not shown), consistent with derivation from Type A/B serpentinites during interaction with an isotopically light fluid. oliv

Chlorine Isotopes Revisited A B C D Type C&D Now, let’s add in isotopic data from HP metasedimentary rocks within the Z-S unit, and from UHP rocks from the associated Lago di Cignana unit. The data are from calcmica schists and horizons rich in Mn nodules in both cases. The isotopic values nicely overlap those of the WS chlorides. The low d37Cl values of the metasedimentary rocks are a bit of a mystery, but once generated, devolatilization at temperatures in excess of 500°C should produce fluids with nearly the same isotopic composition. (i.e., no fractionation). Calcmica schist with diamond-bearing Mn layers

+ Chlorine Isotopes Revisited A B C D mixing between reservoirs One possible way to generate the low d37Cl values in the metaseds is via significant interaction with sedimentary pore fluids from the overlying accretionary wedge during the initial stages of subduction. Pore fluids, along with oceanic basement fluids, have measured isotopic compositions that extend down to -8‰, and are the only materials yet analyzed with such low values. They have low-Cl contents, and are enriched in Ca. So, we do have direct evidence for the existence of metasedimentary fluids that are consistent with those we infer to have been involved in interactions with the serpentinites.

A Physical Scenario Consistent with the Chemical Data Faccenda et al. 2009 Sedimentary pore fluid Seawater-dominated 1 3 2 Bending stresses induce mantle serpentinization via seawater δ37Clserp ≥ 0‰ Finally, let’s try to place the chemical data into a physical model that takes into account the different depths at which fluid-rock interaction occurred as well as the different isotopic reservoirs. ± Local mantle serpentinization via sedimentary pore fluids δ37Clserp ≤ 0‰ Uppermost sedimentary layer modified by extensive interaction with pore fluids from accretionary prism δ37Clmetased << 0‰

A Physical Scenario Consistent with the Chemical Data Faccenda et al. 2012 5 4 4) Unbending stresses create regime of slab-parallel fluid flow in serpentinites during subduction δ37Clserp≅ constant 5) Onset of exhumation moves rocks into regime of cross-slab fluid flow: fluids from metasediments locally modify serpentinites δ37Clserp<< 0‰

Conclusions LdC Serpentinites preserved seafloor δ37Cl values during subduction phase Fluid-rock interaction during exhumation caused localized shifts to lower δ37Cl values Fluids likely derived from nearby metasedimentary rocks in subduction channel Switch from burial to exhumation expanded length scales of fluid-rock interaction and facilitated flow across lithologic boundaries Cl allows us to see chemical interaction between downgoing slab and accretionary wedge fluids 3 B C E D A ≈ 0‰ 2 ≤ -1‰ GPa 1 Our original conclusion that there was no chance in chlorine isotope composition during the subduction history was overly simplistic. When we parsed the data into different textural types and different stages in the PT history, we saw that in fact there were significant change – though localized – in d37Cl values. Most of the isotopic shift occurred during the exhumation stage, however, and not during the subduction phase. Transfer from downgoing plate to upwardly mobile part of subduction channel clearly involved mechanical changes that facilitated communication between what had previously been isotopically isolated reservoirs. Cl provides confirmation of chemical interaction between accretionary prism and downgoing slab. δ37Cl ≥0‰ ≤ -2‰ 300 400 500 600°C