Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Would appreciate help with the following problem:

Given an invertible $n \times n$ real matrix $A$ with eigenvalue $t$ of geometric multiplicity at least $2$, determine whether the following statement is true: "There is a real polynomial $Q(x)$ of degree smaller or equal to $n-2$ such that the inverse of $A$ equals $Q(A)$".

It's easy to prove a similar statement, that the inverse of A can be represented by an $n-1$ degree polynomial (if I am correct, this is true for any invertible matrix), but that's as far as I got. I was thinking maybe the Jordan form of $A$ can help me out, but I haven't gotten very far with that. Would appreciate hints!

share|improve this question
    
If $\mu(x)$ is the characteristic polynomial, then $\mu(A)=0$; since $A$ is invertible, $\mu(0)\neq 0$, so you can write $1 = -\frac{1}{\mu(0)}xm(x)$, where $m(x) = (\mu(x)-\mu(0))/x$. Plugging $A$ tells you that you can always express $A^{-1}$ as a polynomial of degree at most $n-1$ in $A$. But you can play the same game with the minimal polynomial, and the fact that the geometric multiplicity of $t$ is at least $2$ tells you something about the largest degree of $(x-\mathbf{t})$ that can divide the minimal polynomial. Play with that and see if you can get the desired conclusion. –  Arturo Magidin Sep 11 '11 at 23:16

1 Answer 1

up vote 1 down vote accepted

To expand on my comment: A typical way of showing that if $A$ is invertible then there is a polynomial $q(x)$ such that $A^{-1}=q(A)$ is to use Cayley-Hamilton: writing $$\chi(x) = x^n + a_{n-1}x^{n-1}+\cdots+a_0$$ for the characteristic polynomial of $A$, we know that $a_0$ is the product of the eigenvalues, and so $A$ is invertible if and only if $a_0\neq 0$. The Cayley-Hamilton Theorem tells us that $\chi(A)=0$, so $$0 = A^n + a_{n-1}A^{n-1}+\cdots + a_1A + a_0I,$$ from which we get that $$-a_0I = A^n+a_{n-1}A^{n-1}+\cdots + a_1A,$$ hence $$I =-\frac{1}{a_0}A\left( A^{n-1} + a_{n-1}A^{n-2}+\cdots + a_1I\right).$$ Thus, putting $q(x) = -\frac{1}{a_0}\left(x^{n-1}+a_{n-1}x^{n-2}+\cdots + a_1\right)$ gives the desired polynomial.

Of course, you can play the exact same game with the minimal polynomial instead of the characteristic polynomial. The matrix $A$ is invertible if and only if $0$ is not an eigenvalue, if and only if $x$ does not divide the minimal polynomial. The geometric multiplicity of $\mathbf{t}$ tells you something about the degree of the minimal polynomial: since the highest degree of $x-\mathbf{t}$ that divides the minimal polynomial is the size of the largest Jordan block associated to $\mathbf{t}$, and since the geometric multiplicity equals the number of Jordan blocks, if the algebraic multiplicity of $\mathbf{t}$ is, say, $m$, having geometric multiplicity $2$ tells you that there are at least two Jordan blocks associated to $\mathbf{t}$, and so the largest block will have size at most $m-1$; which tells you that if the largest power of $x-\mathbf{t}$ that divides $\chi(x)$ is $m$, then the largest power of $x-\mathbf{t}$ that divides the minimal polynomial is at most $m-1$.

Can you leverage that into showing that a $Q(x)$ as you want must exist?

share|improve this answer
    
Very helpful, thank you! I was working the details from when you posted your comment. I think I have it worked out. Right now, I need to catch up on my sleep, but I'll post my proof (using your hints) here tomorrow, to see if I got it right. –  josh Sep 11 '11 at 23:35
1  
Well, though if I have it right, there is very little work to be done at this point (I think we get that the minimal polynomial is of degree at most n-1, which, using the same trick you've shown, yields a polynomial of degree <= n-2 as required). In any case, good night. –  josh Sep 11 '11 at 23:44

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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