Suppose $T\in L(V)$. Prove that $\dim(\operatorname{null}(T))$ is equal to the number of zero singular values of T.

Proof. Suppose $T\in L(V)$. By Singular-Value Decomposition, $T$ has singular values $s_1,...,s_2$. And orthonormal bases $(e_1,...,e_n)$ and $(f_1,...,f_n)$ of V such that $$Tv=s_1 \langle v, e_1 \rangle f_1 + ... + s_n \langle v, e_n \rangle f_n $$ for every $v\in V$. For each $j$, we have $ Tv_j = s_j f_j$. Hence, each $f_j$ corresponding to $s_j$ are in range $T$. Hence, nonzero $s_j$ span the range T. Hence, $\dim (\operatorname{range} (T))$ equals the number of nonzero singular values of $T$. Note that $T$ has $\dim V$ singular values. By rank-nullity theorem, we have $\dim (V) = \dim (\operatorname{null} T) + \dim (\operatorname{range} T) $.

We have $ \dim V = n $ and $ \operatorname{range} T = n$ since number of nonzero singular values is equal to $\dim \operatorname{range} T$ . Hence $ \dim (\operatorname{null} T) = \dim V - \dim (\operatorname{range} T) = n - n = 0$. Also we have 0 number of singular values. Hence, $ \dim \operatorname{null} T = 0 = $ number of zero singular values.

Is my proof right? Thanks in advance!

  • 2
    $\begingroup$ You don't need the final paragraph (in fact, it's incorrect, because there's no reason the range has to be $n$ dimensional). Once you have noted that $\textrm{dim(range}\,T)$ is equal to the number of nonzero singular values, you have: $\textrm{dim(null}\,T) = \textrm{dim}(V) - \textrm{dim(range}\,T) = n - \textrm{number of nonzero singular values} = \textrm{number of zero singular values}$. $\endgroup$ – BaronVT Dec 16 '14 at 3:40
  • 1
    $\begingroup$ Thanks, so i guess my proof is fine other than that? $\endgroup$ – needhelp Dec 16 '14 at 3:43

You're correct, but needlessly elaborate. Here's an alternative:

In terms of matrices, we can write $$ A = U\Sigma V^T $$ Where $\Sigma$ is the diagonal matrix of singular values, and $U,V$ are orthogonal. Clearly, the rank of $\Sigma$ is the number of nonzero singular values.

Multiplying a matrix by an invertible matrix maintains its rank. So, $A=U \Sigma V^T$ has the same rank as $\Sigma$.

  • $\begingroup$ Thanks for providing me the alternative! $\endgroup$ – needhelp Dec 16 '14 at 4:04
  • $\begingroup$ Is the number of zero eigenvalues equal to the number of zero singular values ? $\endgroup$ – Fareed Abi Farraj Jul 15 '19 at 7:53
  • 1
    $\begingroup$ @FareedAF If you're referring to geometric multiplicity, then no. If you're referring to algebraic multiplicity, then yes. $\endgroup$ – Omnomnomnom Jul 16 '19 at 14:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.