『Abstract
　During chemical weathering, magnesium (Mg) is released by the dissolution of both carbonate and silicate sources. The degree to which solute concentrations and isotopic compositions of rivers reflect the relative proportions of these two inputs, or cycling by a series of processes associated with weathering is poorly constrained. In the river waters of the Mackenzie Basin (Canada), the Mg content is high and Mg isotope ratios (²⁶Mg/²⁴Mg expressed as δ²⁶Mg) show in excess of one per mil variability. Part of this variability is attributed to the 3‰ range in the carbonate and silicate rocks drained. Despite this inherent lithological control on river water δ²⁶Mg values, there is also evidence for a fractionation control. A linear positive covariation between lithium (⁷Li/⁶Li, expressed as δ⁷Li) and Mg isotope ratios in the river waters of the Mackenzie Basin is reported. This covariation is not expected because previously reported fractionation related to physicochemical processes associated with clays or oxides should induce a negative covariation with Mg isotope ratios.
　This continental-scale covariation can be resolved by either process-related fractionation or mixing. Evidence for fractionation associated with clays is provided firstly by comparing Mg and Li isotopes in both the waters and sediments carried in suspension. Secondly a linear covariation between the sediment concentrations of large ion lithophile elements caesium and rubidium (a proxy for clay content of the sediment) and δ²⁶Mg values of the water suggests that processes linked to clay, such as neoformation of clay, cation exchange or adsorption may be important. Simple models illustrate that if the covariation is induced by fractionation, there is either more than one process acting, or a single process is kinetically limited. Alternatively, the data can be reconciled by mixtures between at least three different water bodies, two of which have similar isotopic compositions but differing Li/Mg ratios. This intriguing data set highlights the challenges associated with distinguishing mixing from process with stable isotope data. Despite the complexity, the data question to what extent and by what mechanism clays mediate river water chemistry, at least in terms of the stable isotope compositions of Mg and Li. These questions are fundamental to the quantification of carbon dioxide consumption by silicate weathering and its role in climatic feedback.

Keywords: river geochemistry; magnesium; Mg isotopes; lithium; Li isotopes; chemical weathering』

1. Introduction
2. Materials and methods
　2.1. Study area and sample description
　2.2. Mg isotope analysis
3. Results
　3.1. δ²⁶Mg in river suspended and bed sediments
　3.2. δ²⁶Mg in river waters
4. Trends in river sediment and water chemistry
　4.1. Mineralogical control on δ²⁶Mg sediment composition
　4.2. Lithological control of riverine Mg isotope ratios?
　4.3. Unexpected covariation between Li and Mg isotopes
5. Process versus mixing conundrum; reconciling Mg and Li isotope data
　5.1. Binary or degenerate ternary mixing?
　5.2. Coupled Li and Mg isotope fractionation linked to clay?
　5.3. A Rayleigh model of coupled Li and Mg isotope fractionation
　5.4. Reconciling the counter intuitive positive gradient in δ⁷Li-δ²⁶Mg space in the context of a reservoir effect
6. Implications and the significance of the data
7. Conclusions
Acknowledgements
References