Magma Chamber Evolution as Revealed by 226Ra-230Th Disequilibrium at Longonot Volcano, Kenya

Abstract

The rates at which magmatic processes operate are seldom determined, yet they offer powerful tests of competing physical models of magmatic evolution. U-series disequilibria have the potential to place constraints on the rates of geological processes and highlight significant differences in magmatic timescales. Recent advances in thermal ionisation mass spectrometry have made it possible to detect small variations in 226Ra-23~ disequilibria and here we apply these techniques to a suite of peralkaline trachytes from Longonot volcano in Kenya to evaluate the timescales of closed system magma chamber fractionation. Longonot volcano is situated in central Kenya in the vicinity of Lake Naivasha and has erupted both pyroclastic material and lava flows of almost exclusively trachyte composition for about 10,000 years. The latest phase of activity started around 5650 _+ 120 y BP with the cessation of pyroclastic eruptions and the production of a sequence of trachyte lava flows. The earliest eruptions in this lava pile are mixed peralkaline trachytes and hawaiites and were followed by successive lava flows of peralkaline trachyte which terminated at 3280 _+ 120 y with the eruption of a large (2 km 3) ash horizon. The lava pile stage is estimated to have a total volume of 18-20 km 3 (Scott 1980) of which the main trachyte sequence accounts for 16 km 3. The lava pile shows minimal variation in major elements (61.4-62.2% SiO2) and radiogenic isotopes (average 2~176 19.65 + 0.02, n = 17; 143Nd/144Nd: 0.51262 _+ 1.8 x 10 -5, n = 21). By contrast, incompatible elements increase dramatically (e.g. Th: 12.5-23.6 ppm; Zr: 576-1126 ppm) whereas Ba, which is compatible in sanidine, the major fractionating phase, decreases rapidly from 198-4.7 ppm (Fig. 1). It is well established that Zr behaves as an incompatible element in F-rich silicic peralkaline systems (Leat, 1984), and the strong linear trend shown by Th and Zr can be modeled as the result of-50% closed system sanidine fractionation (DTh ~ Dzr Dna and that it has a value of~12. Fig. 2 shows how the initial 226Ra/Ba ratio is reduced by fractionation if DRa = 12 and consequently how the ages calculated from the measured 226Ra/Ba ratios are likely to be greater than that calculated assuming the linear array represents an isochron. Applying this model to the data from Longonot yields a number of results. Firstly, all the data points now give distinct model ages that are greater then 3 000 y and are therefore consistent with the 14C ages. Secondly, the calculated ages of Ra/Th fractionation decrease as the amount of fractionation increases. Thus the two least evolved samples have ages of -9000 y because they lie close to the equiline,