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.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm preparing for a test in linear algebra and I've come across a problem I'm having trouble with for some reason:

Given a square matrix A, $A^2=2I$, prove that $A-I$ is invertible.

I know this is pretty simple but I can't seem to play with the equations to get it so that for some $B$, $B(A-I)=I$

It's pretty easy to see that $A^{-1}=\frac{1}{2}A$, but beyond that I haven't been able to get very far.

Can anyone help with this?

share|cite|improve this question

From your original equation subtract $\;I\;$ from both sides, and you'll get at once what you want:

$$A^2=2I\implies A^2-I=2I-I=I\implies (A-I)(A+I)=I\;\;\color{green}\checkmark$$

share|cite|improve this answer
+1 for the green checkmark – Martijn Feb 20 at 18:32

No “guess and check” is needed. Set $B=A-I$, so $A=B+I$; then $$ (B+I)^2=2I $$ that becomes $$ B^2+2B+I=2I $$ or $$ I=B^2+2B=B(B+2I) $$ Thus $$ (A-I)^{-1}=B^{-1}=B+2I=A+I $$

share|cite|improve this answer
I would say that setting $B=A-I$ counts as a guess-and-check, how do you know it will help? – David Feb 20 at 8:15
@David No guess: it's the matrix we want to know about. By using $A=B+I$ and substituting, we get rid of $A$. Whether the method works is not known in advance, of course, but it's an attempt to be made. – egreg Feb 20 at 9:24
@David What on Earth would you consider not to be "guess-and-check" in this case? You have to think at least a tiny bit to do mathematics! – Najib Idrissi Feb 20 at 13:31
@David I'm not saying this method is better or worse than “guess-and-check”. Of course, observing that $(A-I)(A+I)=I$ (in this particular case) is much faster. A “safe-though-slower” method should always be available and this was the purpose of my answer. – egreg Feb 20 at 14:05

Note the if $A-I$ is not invertible then exist a vector $v$ such that $(A-I)v=0$ and $Av=v$. Therefore $v$ is an eigenvector with eigenvalue $1$ but the possible eigenvalues of $A$ are $\pm \sqrt 2$. Indeed if $v$ is an eigenvector for $A$ with eigenvalue $\lambda$ then $$\lambda ^2v-2v=0$$ $$v(\lambda ^2-2)=0$$ from here $\lambda =\pm \sqrt 2$

share|cite|improve this answer
This argument can be made somewhat more elementary by eschewing the concept of "eigenvalue" and simply noting that since $Av=v$ we have $2v=A^2v=Av=v$ so $v=0$, contradiction. – Meni Rosenfeld Feb 20 at 18:05
@MeniRosenfeld right – Domenico Vuono Feb 20 at 18:07
But the concept of eigenvalue is beautiful – Domenico Vuono Feb 20 at 18:08
Of course it is. But there's merit in not bringing into a solution heavier machinery than is necessary. – Meni Rosenfeld Feb 20 at 18:33
Not to mention, there's a reasonable chance the OP hasn't yet gotten as far as eigenvalues in their studies. – Meni Rosenfeld Feb 20 at 18:58

We have $$(A + I)(A - I) = A^2 - I^2 = 2I - I = I.$$ So $(A - I)^{-1} = A + I$.

share|cite|improve this answer
Cool. Was that kind of a "guess and check" situation or did you know that if you'd multiply $(A+I)(A-I)$ you'd get 0 as the coefficient of $A$ in the resultant polynomial and that that might solve the problem? – Or Bairey-Sehayek Feb 19 at 22:11
I knew that because of the difference of two squares formula. The answer below shows a more deductive way of finding the inverse. – Ethan Alwaise Feb 19 at 22:16
Difference-of-squares is useful in many areas of math. The generalized difference-of-powers is also useful in places you might not expect: – mephistolotl Feb 20 at 6:22

Suppose that $A-I$ is not invertible. Then $\lambda = 1$ is an eigenvalue of $A$, but then $1$ is an eigenvalue of $A^{2}$. So that $v = A^{2}v = 2v \implies 1 = 2$ a contradiction.

share|cite|improve this answer

Your Answer


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.