Evidence of 109 Ag/ 107 Ag variability in ancient silver coins led us to calculate the reduced partition functions for 107 Ag and 109 Ag in various dissolved Ag species by ab initio methods in order to evaluate the extent of Ag fractionation in fluids and the potential of Ag isotopes to discriminate between different metal sources. We used a hybrid density functional implemented by the Gaussian 09 code and consisting of Bickley's three-parameter non-local hybrid exchange potential with Lee-Yang-Parr non-local functionals. The ratios ln b of reduced partition functions were for the free Ag + ion with various degrees of hydra-tion, hydrates, chloride complexes, sulfides, sulfates, Sb-As sulfosalts, and Ag-ammines. At 0 °C, the magnitude of the nuclear field shift effect between metal and dissolved sulfide is À1 Â 10 À4. Using literature stability fields at different temperatures, we conclude that only weak Ag isotope fractionation is expected in the Ag-Cl-S system regardless of the pH of hydrothermal solutions at 300 °C. Stronger effects are predicted when Sb and As are added to the solutions. Bonding with SbS 3 and AsS 3 reduces ln b values by $2 Â 10 À4. Under the more oxic conditions of the subsurface and at the temperatures of groundwater , Ag should be present as Ag + and, at higher chlorinity, as AgCl 0. The latter component is isotopically heavier than Ag +. In groundwater underneath forests and grasslands, ammonia resulting from nitrogen fixation produces the particularly strong complex known as di-ammine silver Ag(NH 3) 2 + (Tollen's reagent). Upon reduction by aldehydes and melanin, Ag(NH 3) 2 + precipitates metallic Ag(0). Biological oxidation of NH 3 to NO 2 À and NO 3 À (nitrification) also is expected to destroy Ag(NH 3) 2 + and precipitate metallic Ag(0). Both chloragyrite (AgCl) and native Ag are found among the weathering products (gossan) of bedrock ore deposits. The high end of the 109 Ag/ 107 Ag histogram of silver coinage from around the Mediterranean and from Spanish Americas, particularly well-represented in Mexican silver ore, therefore attests to the presence of gossan silver, while the variability at the low end is more likely to represent temperature effects and variable abundances of S, Sb, and As in hydrothermal fluids. Should the yield of Ag during separation chemistry be less than 100%, or samples be altered by meta-morphism, low-temperature Ag isotope fractionation becomes an issue seriously affecting (i) so-far published 109 Ag/ 107 Ag data on chondrites, (ii) ages derived from the extinct 107 Pd-107 Ag chronometer (T 1/2 = 6.5 Ma), and (iii) inferences about the volatile content of the Earth. It is argued that the NIST SRM 978a value should be retained to represent the Bulk Silicate Earth and not the literature values on basalts, which clearly have been affected by incomplete Ag separation.