“…A very characteristic example of that is the melting point of a very commonly studied IL, [C 2 C 1 im][BF 4 ], for which the available data vary from 5.8 to 16 C. 136 There are also many studies of reaction kinetics, which is another domain where ML methods could be useful, 137 that show that common impurities, such as moisture or unreacted starting materials can signicantly affect the results. 138 In order for results to be reproducible, the ILs have to be either ultrapure (<0.1% of impurity levels) 139 or the level of purity has to be clearly stated in each work. This has become more common in works published in the last few years, but there are minimal purity data from earlier IL research, which unfortunately makes their use very limited without re-testing the results.…”