But this is only a problem if you use i, j, and k as unit vectors. I found $\hat e_{x,y,z}$ or even $\hat{e}_{1,2,3…}$ etc. to be the usual convention in physics, as it generalises much nicer to more dimensions.