ELSEVIER Earth and Planetary Science Letters 160 (1998) 505±519 Uranium and thorium diffusion in diopside James A. Van Orman a,Ł, Timothy L. Grove a, Nobumichi Shimizu b a Department of Earth, Atmospheric, and Planetary Sciences, Massachusetts Institute of Technology, Cambridge, MA 02139, USA b Department of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA 02543, USA Received 25 August 1997; revised version received 26 March 1998; accepted 6 April 1998 Abstract This paper presents new experimental data on the tracer diffusion rates of U and Th in diopside at 1 atm and 1150± 1300ëC. Diffusion couples were prepared by depositing a thin layer of U±Th oxide onto the polished surface of a natural diopside single crystal, and diffusion pro®les were measured by ion microprobe depth pro®ling. For diffusion parallel . : : / . = /= : to [001] the following Arrhenius relations were obtained: log10 DU D 5 75 š 0 98 418 š 28 kJ mol 2 303RT . : : / . = /= : log10 DTh D 7 77 š 0 92 356 š 26 kJ mol 2 303RT. The diffusion data are used to assess the extent to which equilibrium is obtained during near fractional melting of a high-Ca pyroxene bearing mantle peridotite. We ®nd that the diffusion rates for both elements are slow and that disequilibrium between solid and melt will occur under certain melting conditions. For near-fractional adiabatic decompression melting at ascent rates >3cm=yr, high-Ca pyroxene will exhibit disequilibrium effects. High-Ca pyroxene will become zoned in U and Th and the melts extracted will be depleted in these incompatible elements relative to melts produced by equilibrium fractional melting. U and Th diffusivities in high-Ca pyroxene are similar, and diffusive fractionation of these elements will be limited. Numerical solutions to a dynamic melting model with diffusion-controlled chemical equilibration indicate that the activity ratio [230Th=238U] in a partial melt of spinel peridotite will be slightly less than 1 for a broad range of melting parameters. This result reinforces the already widely accepted conclusion that melting of spinel peridotite cannot account for 230Th excesses in mid-ocean ridge and ocean island basalts, and that garnet must therefore be present over part of the melting column. 1998 Elsevier Science B.V. All rights reserved. Keywords: diffusion; diopside; pyroxene group; uranium disequilibrium; U-238=Th-230; igneous processes 1. Introduction the mantle U and Th are fractionated from each other. Clinopyroxene (cpx) and garnet are the prin- Radioactive disequilibrium between 238Uand cipal hosts of U and Th in upper mantle rocks and 230Th is widely used to infer the rate and style control the ¯uxes of these elements during melting. of melting and melt transport in the mantle [1±8]. Equilibrium mineral=melt partition coef®cients for Making full use of the 238U±230Th system requires U and Th between clinopyroxene and basaltic melt knowing in which minerals and at what depths in are small (¾103 to 102) and differ by only about a factor of 2, with U being the more incompatible Ł Corresponding author. Fax: C1 (617) 253-7102; E-mail: element [9±13]. The similarity between U and Th [email protected] partition coef®cients implies that clinopyroxene has 0012-821X/98/$19.00 1998 Elsevier Science B.V. All rights reserved. PII S0012-821X(98)00107-1 506 J.A. Van Orman et al. / Earth and Planetary Science Letters 160 (1998) 505±519 little ability to fractionate these elements. Even if Table 1 signi®cant fractionation were possible it would be Kunlun Mts. diopside composition 230 in the wrong sense to explain the Th excesses Oxide Weight (%) observed in nearly all recently erupted mid-ocean ridge basalts (MORB) [14,15]. On the other hand, SiO2 55.5 TiO2 0.06 U and Th partition coef®cients between garnet and Al2O3 0.88 basaltic melt differ by nearly an order of magnitude FeOtot 0.55 [11,16,17], with Th being the more incompatible MgO 18.1 element. The equilibrium partitioning data, then, ap- CaO 24.7 pear to require melting in the presence of garnet Na2O0.53 Total 100.3 to explain the excess 230Th in MORB. This is an important conclusion because it requires that melting begins at depths greater than ¾75 km, where garnet becomes the stable aluminous phase in peridotite near pure diopside (Table 1). Each crystal was sec- [18,19] or, alternatively, that garnet pyroxenite is an tioned perpendicular to the c crystallographic axis important component of the MORB source [20]. into slabs ¾0.5 mm thick. One side of each slab The conclusion that clinopyroxene cannot deliver was polished to 0.06 mm alumina grit and then cut excess 230Th to the melt is valid if chemical equi- into pieces 1±2 mm on a side. These pieces were librium between cpx and melt is maintained during cleaned ultrasonically in deionized water and then melting. If, however, diffusion in clinopyroxene is pre-annealed for 2 days at 1200ëC, with oxygen fu- slow relative to the melting rate, then equilibrium gacity controlled near the quartz±fayalite±magnetite may not be achieved. In this case fractionation be- (QFM) buffer. tween U and Th depends strongly on the relative The diffusion source material was deposited as diffusion rates of these elements [21±23]. A reversal an aqueous solution onto the polished surface of in the effective compatibility order of two elements the diopside. A dilute (0.05 M) nitric acid solution is possible if the more compatible element diffuses that contained dissolved U, Th, and Al in 1 : 1 : 4 signi®cantly faster than the incompatible element molar proportions was prepared from 10,000 mg=ml [21]. To produce excess 230Th in the melt by a ICP standard solutions and diluted with puri®ed disequilibrium mechanism would require that (1) U water to 270 mgU=ml. Approximately 1 mlofthis diffuses slowly enough in clinopyroxene that equi- solution was deposited onto the polished surface of librium between cpx and melt is not achieved, and the diopside, along with a small amount of methanol (2) Th diffuses signi®cantly faster than U. To deter- to reduce surface tension and allow the solution mine whether disequilibrium melting can in¯uence to spread uniformly over the sample surface. The U and Th partitioning, we performed a series of ex- solution was evaporated in air at 120ëC, which left a periments to measure the diffusion rates of U and Th thin layer of nitrates with concentration ¾ 5 ð 1010 in diopside in the temperature range 1150±1300ëC. mol=mm2. At the conditions of the diffusion anneal, with fO2 at the QFM buffer, the nitrates decompose to Al2O3 and nearly stoichiometric UThO4 [24,25]. 2. Experimental and analytical techniques Because U and Th have the same formal charge in the oxide (C4) that they presumably have in 2.1. Sample preparation diopside, no further redox reactions are necessary to introduce these elements into the diopside lattice. Diffusion experiments were performed on gem There was no evidence in any of the experiments quality natural diopside crystals from the Kunlun that the tracer layer reacted with the diopside. We Mts., China (American Museum of Natural History calculate that less than 0.01% of the U and Th samples #100242 and #104501). The crystals were diffused into the diopside crystal during any of the light green, transparent, free of cracks and visible in- experiments, and thus the tracer layer provided an clusions, and had major element compositions very effectively in®nite reservoir of U and Th. J.A. Van Orman et al. / Earth and Planetary Science Letters 160 (1998) 505±519 507 Al was added to the tracer solution as a possi- 2.3. Analyses ble charge-balancing species. However, as discussed below in the Analyses section, we were unable to Concentration pro®les in the annealed samples detect Al diffusion pro®les in any of our samples, were measured using the Cameca IMS 3f ion micro- and it is unclear whether Al was actually transported probe at the Woods Hole Oceanographic Institution. into the crystal during any of the diffusion anneals. A primary beam of O ions was focused onto the surface of the sample, producing secondary ions that 2.2. Diffusion anneals were continuously analyzed in a mass spectrome- ter. In this procedure successively deeper layers are Diffusion anneals were performed at constant sampled as the primary beam sputters through the temperature in open Pt crucibles placed in the sample, allowing concentrations to be measured as hotspot of a Deltech DT31VT vertical gas mixing functions of depth. The primary beam was acceler- furnace, with anneal times ranging from 2 to 24 ated under a potential of ¾8.2 kV and was focused to days. Run temperature was attained within approx- a diameter of ¾20±30 mm, with beam currents usu- imately 5 minutes after introducing the charge into ally in the range 10±20 nA. By analyzing `zero-time' the furnace and was continuously monitored using experiments, we found that optimum resolution was aPt±Pt90Rh10 thermocouple calibrated against the achieved by rastering the primary beam over a square melting points of NaCl, Au, and Pd. Fluctuation in area 150±200 mm on a side and inserting a circu- temperature over the course of each anneal was gen- lar mechanical aperture 68 mm in diameter into the erally within š2ëC, with the exception of a single secondary optics. The aperture was centered over the experiment (UTh1200b) in which the thermocouple sputtered area so that secondary ions were collected degraded during the anneal; temperature for this ex- only from the central, ¯at portion of the pit.