The Burrows & Sharp 1999 flexchem scheme results in NaNs / Infs if the temperature is too low and if LFRic is compiled with single precision. Ian put limits on the input temperature and exponents as a temporary patch but this needs to be investigated.