$1$If you include interactions with : or ⋆, R will include new predictors that are products of the original variables. Technically, there is nothing wrong with this, but it rarely has a straightforward interpretation.