On the nucleation of martensite