Coarsely crystalline cryogenic cave carbonates (CCCcoarse) are widely used indicators of past permafrost conditions. Nevertheless, details of their formation with respect to macroscopic morphology, stable isotope evolution and potential metastable precursor phases are poorly understood.
CCCcoarse were found in 5 caves located along a north-south transect of the Ural Mountains, Russia. A comprehensive data set was generated including results of carbonate stable isotope composition, stable isotopic composition of fluid inclusion water, stable oxygen isotope thermometry, trace element composition, X-ray diffraction and transmission properties and U/Th disequilibrium dating.
Detailed petrographic characterization of the samples allows for the proposal of a morphological classification scheme for CCCcoarse. Non-crystallographic branching of crystallites, also referred to as crystal splitting, is identified as the mechanism enabling morphological variety in CCCcoarse. Splitting propensity is likely related to physico-chemical properties of the mineral forming solution such as Mg++ concentration and supersaturation with respect to carbonate minerals.
Numerical modelling of C and O stable isotope evolution shows that open system style degassing of CO2 during CCCcoarse formation accounts for 10 20% of the observed isotopic trends. These results support a model of slow and continuous degassing of CO2 via nucleation of gas bubbles in co-precipitating ice in compliance to freezing experiments.
Fluid inclusion analyses suggest that early stages of CCCcoarse formation take place at isotopic equilibrium between carbonate minerals and the parent solution. However, oxygen isotope fractionation (()18=1.0318 0.0005) appears to be smaller than expected from literature values extrapolated to 0C.
Diffraction properties of a CCCcoarse specimen indicate that it formed via non-classical crystallisation pathways. Crystallisation by particle attachment (CPA) of poorly crystalline, or even amorphous precursor phases can explain diffraction data.