I know in general that if a matrix $A$ is positive definite, then there exists a (unique?) square root matrix $B$, which is also positive definite, such that $BB=A$.

Therefore, suppose $A$ is positive definite. It is invertible, and its inverse is also positive definite. Therefore I know there exists $C$ (possibly unique?) so that $CC=A^{-1}$. For simplicity, I will call $C=A^{-1/2}$, so that in this notation $A^{-1/2}A^{-1/2}=A^{-1}$.

My question is, is it also true that $A^{-1/2}AA^{-1/2}=I$? Or is this not necessarily true?

I have posted an attempted solution below -- please let me know what you think.


Edit It occurs to me that at the essence of this question is whether or not $B=C^{-1}$; is this necessarily true? That is, is square root of the inverse of a matrix equal to the inverse of the square root of the matrix? (This also partly depends on whether or not $B$ and $C$ are themselves invertible; would this be true?)

Edit 2 There's now a related topic which asks what kind of matrix decomposition we're using here (and the differences between these different approaches to matrix decomposition. That question can be found here.

  • $\begingroup$ There is a risk of misunderstanding in referring to "the" square root of a matrix, as matrix square roots are typically even less unique than square roots of scalars (where we have a choice of sign in picking a square root). $\endgroup$ – hardmath Oct 23 '14 at 3:32
  • $\begingroup$ Hi Hardmath! Thanks for your comment. I got the uniqueness fact from here, see #7 under properties: en.wikipedia.org/wiki/Positive-definite_matrix $\endgroup$ – Mathemanic Oct 23 '14 at 4:39
  • 2
    $\begingroup$ Note that in that article "positive definite" assumes (at least for real matrices) that the matrix is also symmetric. Thus a real symmetric positive definite matrix has a unique real symmetric positive definite matrix square root (which is the analog of taking the positive square root of a positive real number). $\endgroup$ – hardmath Oct 23 '14 at 11:25

It is generally true that if $A$ is an $n\times n$ invertible and if $A^{-1}$ has a "square root" $C$, also $n\times n$, such that:

$$ A^{-1} = C^2 $$

then $C^{-1} A^{-1} C^{-1} = I$ holds.

The first fact we need is that since $A$ is invertible, $A^{-1}$ is invertible, and this implies $C$ is invertible. For if not, then there would exist a nonzero vector $x$ in the nullspace of $C$, and $Cx=0$ would imply $A^{-1}x=0$, contradicting the invertibility of $A^{-1}$. Thus $A = (C^2)^{-1} = (C^{-1})^2$.

The second fact we need is that a one-sided inverse of a matrix is a two-sided inverse, so that:

$$ A C^2 = I \; \implies \; C A C = I $$

That is, using associativity of matrix multiplication, the left hand side tells us $(AC)C = I$, so that $C$ is a (right) inverse of $AC$. Thus it must also be a (left) inverse of $AC$, which is what the right hand equation states.

Finally by the same reasoning:

$$ C A C = I \; \implies \; C^{-1} A^{-1} C^{-1} = I $$

In this discussion/proof we have not invoked the symmetry of $A$ nor the uniqueness of a symmetric positive definite square root $C$ for $A^{-1}$, which also would be symmetric positive definite. The reasoning above is correct even if $A$ is not symmetric, and even if $C$ is not positive definite, and relies only on $A^{-1} = C^2$.

  • $\begingroup$ Thank you for the thoughtful answer! What are the conditions under which $A^{-1}$ has a square root $C$ so that $A^{-1}=C^2$? $\endgroup$ – Mathemanic Oct 24 '14 at 0:09
  • $\begingroup$ For real symmetric matrices $A$ you have articulated the necessary and sufficient condition: $A$ must be positive definite. For nonsymmetric matrices we can analyze the conditions in terms of the Jordan normal form and the spectrum (eigenvalues) of $A$. It's worth a new Question if you'd like to pursue that topic. $\endgroup$ – hardmath Oct 24 '14 at 0:59
  • $\begingroup$ @hardmath towards your "necessary and sufficient condition:" what if $$ C = \pmatrix{0&-1\\1&0}? $$ $\endgroup$ – Ben Grossmann Oct 24 '14 at 4:15
  • $\begingroup$ @Mathemanic: You have defined $C=A^{-1/2}$, not $C^{-1} = A^{-1/2}$. It is easy to get lost in the exponents, but I believe $CAC = I = C^{-1} A^{-1} C^{-1}$ is correct. $\endgroup$ – hardmath Oct 24 '14 at 12:18
  • $\begingroup$ @Omnomnomnom: Excellent! If your $C^2$ is computed, then $A = C^{-2} = -I$, which is symmetric but negative definite. I stand corrected. $\endgroup$ – hardmath Oct 24 '14 at 12:23

We know \begin{equation} A^{-1}=CC \end{equation} Inverting both sides (Is this allowed? Depends on whether $C$ is invertible, which I do not know...), we would have \begin{equation} A=C^{-1}C^{-1} \end{equation} Multiply both sides on the left by $C$, and then on the right by $C$, so we get \begin{equation} CAC=I \end{equation} or, in the "$1/2$" notation, \begin{equation} A^{-1/2}AA^{-1/2}=I \end{equation}


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.