Row rank = column rank over a PID

linear algebramatrix-rankmodules

Suppose I have some matrix $M$ with entries from a PID. Define the column rank and row rank as you would expect: the rank of the module generated by the columns (resp. rows).

Is it always true that the column rank is equal to the row rank?

Best Answer

Yes, due to the existence/uniqueness of Smith normal form:

For any PID $R$ and any $(n\times m)$-dimensional matrix $A$ with entries in $R$, it holds that $A$ is equivalent to a diagonal matrix $B$ with diagonal entries $\delta_1,\delta_2,\ldots,\delta_k,0,\ldots,0$, where the $\delta_i$ are nonzero elements of $R$ and $\delta_i|\delta_{i+1}$ for all $i<k$.

Equivalence of $A$ and $B$ here means there are invertible square matrices $P\in M_n(R)$ and $Q\in M_m(R)$ with $B=PAQ$. In particular the rank of the images of $A$ and $B$ agree, and similarly for the null spaces.