when you have a lagrangian multiplier between a constraint and a function. is it possible that there are no points of tangency between the constraint and the function?