On crystallization at the inner core boundary