The reported abundances of the argon isotopes in the terrestrial atmosphere4 at present are 40Ar, 99.600%; 38Ar, 0.063%; and 36Ar, 0.337%, giving an atmospheric 40Ar/36Ar ratio of 295.5. This anomalous situation is clearly related to the production of radiogenic 40Ar derived from the natural decay of 40K in the Earth's crust and mantle. An increase of more than four orders of magnitude over the initial terrestial 40Ar/36Ar ratio is indicated. Table 1 Measured variation in (40Ar( /36Ar)i ratiosSource (x!06yr) Age (4oAr/36Ar)i 1 Historic 282.9-^294.82 Historic 292.1->320.4 3 0.075 Â0.061 295.2Â11.1 4 0.120Â0.011 289.8Â3.8 5 0.194Â0.056 293.2Â6.0 6 0.259 Â0.035 289.5Â7.3 7 0.534Â0.145 282.6i 14.6 8 12.4 Â0.3 302 Â6 9 15.3 Â0.2 295 Â3 10 15.8 Â1.3 293 Â19 11 53.1 Â2.5 289 db 21 12 64.8 Â3.1 263 Â21 13 55.4 Â1.2 287 Â9 14 119 Â4 275 db 8 15 410 Â4 183 Â70 16 415 Â4 224 Â37 17 411 Â3 196 Â61 18 770 Â14 131 Â88 (1) Basalts8; (2) single basalt flow26; (3) ? (7) New Zealand volcanics27; (8) East Iceland volcanics28; (9) Continental volcanics26; (10) North-west Iceland volcanics28; (11) ? (13) Faeroes basalts29; (14) Plutonic minerals7; (15) Continental margin, volcanics30; (16) Plutonic biotites30; (17) Plutonic biotites30,31; (18) Continental dolerite dykes32.To demonstrate the isotopic evolution of argon in the solid Earth, we calculated the initial 40Ar/36Ar ratios retained in analysed suites of mantle-derived igneous rocks, both volcanic and plutonic, of different ages (Table 1). The initial 40Ar/36Ar ratios were calculated in a manner exactly analogous to the determination of initial 87Sr/86Sr ratios, the intercept of an argon isochron following the linear equation: (40Ar/36Ar)total^[Xe/(Xe+XB)](40K/36ArXexp(XO-l) + (40Ar/36Ar)initiaiA plot of the initial 40Ar/36Ar ratios obtained from eighteen suites of mantle-derived samples against their time of emplacement or extrusion (Fig. 1) strongly suggests that there has been an increase in the initial 40Ar/36Ar ratio with time. This we interpret as evidence for a significant increase in the 40Ar/ 36Ar ratio in the mantle over the past 800 Myr. Implicit in this interpretation is the assumption that a significant proportion of the 36Ar observed in the mass spectrum of argon extracted from terrestial rocks and minerals is derived from the source region of the analysed samples and does not represent an atmospheric contaminant. We are supported in this contention by reported analyses of lava flows5?8 which have shown that most of the so-called 36Ar 'contamination' comes not from the extraction line but is contained within the rocks themselves and is released along with 40Ar during fusion. The initial 40Ar/36Ar ratios preserved in igneous rocks seem therefore to be representative of the 40Ar/36Ar ratios in the source region of the parent magmas. The 40Ar/36Ar ratio in any source region will increase with time at a rate dependent on the concentrations of potassium and 36Ar, and the rate of argon release to the atmosphere.Analyses of historically erupted lava flows5,6,8, indicate initial 40Ar/36Ar ratios ranging from 282.9 to 350.0. Measured ratios greater than the present-day atmospheric ratio have been explained by the inclusion of xenoliths and xenocrysts in the analysed whole-rock samples but a satisfactory explanation has not been given for the lower ratios. The possibility of diffusion and fractionation of atmospheric argon into the cooling lava8 does not seem realistic to us. This variation of the observed 40Ar/36Ar ratios in modern volcanic rocks may be due to inhomogeneities in the upper mantle or to remelting of older gabbroic rocks in layer 3 as suggested for some Icelandic rocks9. We therefore conclude, from the reported analyses of modern basalts, that the average 40Ar/36Ar ratio in the upper mantle is at present not greater than the atmospheric ratio (295.5) and possibly a little lower. A similarity between the 40Ar/36Ar ratio in the upper mantle and the terrestial atmosphere leads us to the conclusion that it is the Earth's mantle, not the crust, that is the major source of atmospheric argon. Despite the obvious generation of radiogenic 40Ar in the potassium-rich crust of the Earth, there is little evidence that significant outgassing of crustal argon is occurring today. It has been variously suggested (see Rankama10, for a summary) that the 40Ar present in the Earth's atmosphere has been gradually released from crustal rocks throughout geological time as a result of weathering processes11 or volcanism12. Alternative models involving major degassing of the crust and mantle during the first 109 years of the Earth's history have been suggested by Nicolet13 and Damon and Kulp14, while Turekian15,16 and Schwartzman17 advocated continuous transfer of argon from both the crust and mantle. Modern plate tectonic theory provides a viable mechanism for the continual release of argon from the mantle, by way of volcanism on active plate boundaries, but there is no simple mechanism available for the release of argon accumulating in the Earth's crust. Furthermore, a large component of crustal argon in the atmosphere would result in a present-day atmospheric 40Ar/36Ar ratio much greater than 295.5.Fig. 1 Variation of the initial 40Ar/36Ar ratio in the upper mantle-derived igneous rocks over the past 800 Myr. This variation is considered to reflect the variation in argon isotopic ratio of the upper mantle region with time. It is generally assumed that minerals melting and recrystallising in the Earth's crust under igneous and metamorphic conditions will lose argon to the atmosphere as a result of temperature-controlled diffusion processes. In reality, what is more likely to occur under these conditions is lattice diffusion and isotopic homogenisation of argon, with effective retention in the whole-rock system, a situation analagous to strontium isotopic homomogenisation under similar conditions. Very little argon would escape from the crust directly to the atmosphere, except perhaps along major fracture zones. Homogenisation of the argon isotopes would have the effect of equalising the 40Ar/36Ar ratio in the whole-rock system, so that initial 40Ar/36Ar ratios measured in crustal rocks should be greater than the present-day atmospheric ratio. Excessively high initial ratios have been recorded in metamorphic rocks, expressed in terms of an 'excess' 40Ar concentration18?21. To test our conclusions we have made a semiquantitative analysis of the available data. Any relevant calculations suffer, of course, from uncertainties in the estimation of potassium abundances22,23 in the Earth (Table 2). The calculated mass of radiogenic 40Ar produced in the Earth from the decay of 40K consequently has a large error which is propogated throughout the calculations.The mass of 40Ar produced in the Earth is calculated assuming the Earth to be 4,500 Myr old. The observed mass of 40Ar in the atmosphere (0.065´1015 tons), compared to the calculated mass of radiogenic 40Ar produced in the Earth (0.11´1015 tons), suggests that only 60% of all the 40Ar produced in the Earth has been released to the atmosphere. Table 2 Estimat e of potass ium and ra diogenic i irgon in the EarthMas: s of shell24 Potassium i22'23K% Radiogenic 40Ar (x: LO21 tons) (xlO11 tons) (x!015tons) Crust 0.024 320Â55 1.350 0.0423 Â0.0073 Mantle and core 5.941 500 Â200 0.008 0.0668 Â0.0267 Total 5.965 820 Â207 0.014 0.1091 Â0.0275 Pyrolite Atmosphere 0.441 500 Â200 0.108 (ref. 25) 0.0668 Â0.0267 0.0653 (ref. 11)40Ar calculated from the potassium abundances with an age of 4,500 Myr for the Earth. In calculating the mass of 40Ar in the mantle we make the following assumptions: (1) All 40Ar in the atmosphere is derived from the solid Earth. (2) The upper mantle and crust are separate systems with respect to argon isotopic evolution.(3) Sediments make up approximately 5 % of all crustal material24, therefore not more than 5% of crustal argon has been released to the atmosphere and perhaps as little as 1 % (ref. 10). From assumption (e maximum released to the atmosphere, the balance of the 40Ar in the atmosphere is assumed to have come from the mantle, so (in units of 1015 tons):40Ar in atmosphere = (0.0653 Â0.0033) 40Ar released from crust (5%) = (0.0021 Â0.0004) 40Ar released from mantle = (0.0632Â0.0034) 40Ar remaining in mantle = (0.0036 Â0.0269) The large error in the estimation of 40Ar remaining in the mantle arises because the abundance is the difference between two nearly equal and poorly known numbers.For purposes of calculation, we now assume that: (4) All the potassium, and thus radiogenic 40Ar, is concentrated in an upper mantle region with a pyrolite25 composition.(5) This pyrolite layer is approximately 200 km thick and has a mass of 0.441´1021 tons (Table 1). (6) The argon isotopic composition is homogeneous throughout the pyrolite layer. The maximum concentration of 40Ar at present in this pyrolite layer can now be calculated knowing the mass of the layer and the mass of 40Ar remaining in the upper mantle. Thus the maximum 40Ar concentration in the total pyrolite layer is (8.2Â61.0) ´10?3 p.p.m. 40Ar. The present-day 36Ar concentration in the upper mantle can be estimated from the observed 36Ar concentrations in modern volcanic rocks derived from the mantle. The concentration of 36Ar in young volcanic rocks, measured in our laboratory, after correcting for line blank, is estimated to be (2.4 Â1.1)´10?5 p.p.m. 36Ar. Because these rocks have been at least partially outgassed on extrusion, this value serves at least as a minimum concentration for the 36Ar in the upper mantle source region. Assuming that the present-day 40Ar/36Ar atomic ratio in the upper mantle source region is 295 Â10 (or 327 weight ratio), then the present-day 40Ar concentration in the upper mantle is at least (7.8Â3.6)´10?3 p.p.m. 40Ar. This value is close to that estimated in the pyrolite layer by the difference method. This near coincidence could be due to the cancelling out of the different assumptions (1) to (6), or it might indicate that Hurley's potassium abundances22,23 are more precise than indicated in Table 2 and that most of the 40Ar at present in the atmosphere has been released from the mantle. According to our model, approximately 95 % of all the 40Ar produced in the mantle has been lost to the atmosphere. At least that percentage of 36Ar must also have been lost, assuming no isotopic fractionation. From assumptions (5) and (6), using the observed concentration of 36Ar in mantle-derived volcanic rocks, the mass of 36Ar in the pyrolite shell (200 km thick) is estimated to be (1.05 Â0.50)´1010 tons at present. This must represent at most 5 % of the 36Ar originally present in the mantle. Therefore, 4,500 Myr ago, the mass of 36Ar in the mantle was at least (21.0Â9.0)´1010 tons. There is at present 20.0x1010 tons of 36Ar in the atmosphere, so it is quite reasonable to postulate that the majority of 36Ar in the atmosphere has also come from the mantle. Protocrustal material, with an average potassium concentration of 1.350% 22,23, would yield in 4,500 Myr a concentration of 40Ar ten times greater than that in the upper mantle. The net result of this would be to give 40Ar/36Ar ratios in the deep crust that are an order of magnitude greater than the 40Ar/36Ar ratio in the upper mantle.Initial 40Ar/36Ar ratios on igneous rocks which have not undergone isotopic homogenisation since crystallisation should yield information about the source region. It is thus possible to distinguish clearly between partial melting in a low potassium environment (such as the upper mantle) and partial melting of old sialic material. Initial 40Ar/36Ar ratios should also be very sensitive to any geological contamination (assimilation) of a basaltic magma by older potassium-rich material. In summary: (1) Initial 40Ar/36Ar ratios of mantle-derived igneous rocks have increased steadily over the past 800 Myr.(2) The present-day 40Ar/36Ar ratios in the upper mantle and atmosphere are very similar. (3) The majority of argon at present in the atmosphere has been derived from the mantle rather than the crust. (4) Initial 40Ar/36Ar ratios on igneous rocks are potentially more sensitive petrogenetic indicators than strontium or lead isotopes. Calculation of argon isochron parameters from the literature suffers from several omissions in the presentation of K-Ar age data for publication and we suggest the following minimum list of items that need to be reported: (1) The measured argon ratio in air. (2) The system line blank and isotopic ratio. (3) The weight of sample fused. (4) The concentration of 36Ar measured or the percentage of 'atmospheric' to 3 significant figures.We acknowledge early discussions with Drs Pankhurst and O'Nions at the University of Oxford, and Dr John Mitchell at the University o