Author(s): V. Rama Murthy (corresponding author) ; Wim van Westrenen [2, 3]; Yingwei Fei 
The hypothesis that 40 K may be a significant radioactive heat source in the Earth's core was proposed on theoretical grounds [1, 2] over three decades ago, but experiments [3, 4, 5, 6, 7, 8] have provided only ambiguous and contradictory evidence for the solubility of potassium in iron-rich alloys. The existence of such radioactive heat in the core would have important implications for our understanding of the thermal evolution of the Earth and global processes such as the generation of the geomagnetic field, the core-mantle boundary heat flux and the time of formation of the inner core [9, 10, 11, 12]. Here we provide experimental evidence to show that the ambiguous results obtained from earlier experiments are probably due to previously unrecognized experimental and analytical difficulties. The high-pressure, high-temperature data presented here show conclusively that potassium enters iron sulphide melts in a strongly temperature-dependent fashion and that 40 K can serve as a substantial heat source in the cores of the Earth and Mars.
We have found that polishing of a sample for electron microprobe analysis using oil lubricants results in substantial K loss from the Fe-S phase within a few hours and a near total loss in a few days (Fig. 1). A second problem is the loss of K in experiments when the charge is contained in a single capsule, as shown by mass-balance calculations after the experiment. The degree of loss depends on temperature (T ), run duration and pressure (P ). These unrecognized problems in previous studies could have contributed to the ambiguities concerning K entry into iron sulphide melts. The validity of other data on alkali distribution in terrestrial and planetary materials gathered by such experiments may also need to be examined.
We report here data from experiments specifically designed to avoid these analytical difficulties (see Methods). Our initial effort was to establish that reactions such as those suggested previously , for example:[eq. 1]K 2 SiO3 + FeS [right arrow] K2 S + FeSiO3 (1) [eq. 2]1/2K4 SiO4 + FeS [right arrow] K2 S + 1/2Fe2 SiO4 (2) are possible and that K2 S is soluble in a Fe-S liquid that would have segregated to form the core.
Synthetic mixtures of Fe-metal, FeS, a K-silicate glass and, in one instance, a natural peridotite (KLB-1) were used for experiments conducted at 1-3 GPa at temperatures above the liquidus of both metal and silicate phases (1,200-1,800 [degrees]C). The calculated oxygen fugacity, f O2 , in all runs was about 1.5 log units below the iron-wüstite buffer. This is in accord with core-formation models in which metallic liquid equilibrates with molten silicate under reducing conditions in a magma ocean in the early Earth . We determined the partition coefficient for potassium, D K (concentration of K in sulphide/concentration of K in silicate) as a function of pressure, temperature, and composition.
Data on D K at different temperatures at constant...