Lin alg final coming up, getting the study sesh in

Fun fact:just learned latex so yay

The full problem comes in two parts. Part A is simple/

Given $A$= \begin{pmatrix} 2 & -3/2 \\ 1 & -1/2 \\ \end{pmatrix}


$B$= \begin{pmatrix} 3 & -3 \\ 2 & -2 \\ \end{pmatrix}

a. Find a matrix $Q$ such that $Q^{-1}AQ= D$ where $D$ is a diagonal matrix.

Which is simple enough, found the eigen-vectors that came out to be $Q$: \begin{pmatrix} 1 & 3/2 \\ 1 & 1 \\ \end{pmatrix} (which could very well be wrong).

Part B is where the difficulty arises:

b. Show that $\lim_{n\rightarrow\infty} A ^{n} = B$

I'm assuming this has to be solved using some eigen-vectors or some orthogonal basis transformation, however I cannot seem to wrap my head around this problem.

Thank you in advance!

PS. Thanks to John Moeller for the formatting tip.

  • $\begingroup$ What is $D{}{}$? $\endgroup$ – Euler....IS_ALIVE May 6 '14 at 1:09
  • 2
    $\begingroup$ Are $A$ and $B$ supposed to be equal to each other? $\endgroup$ – Braindead May 6 '14 at 1:09
  • $\begingroup$ Re: formatting: you want pmatrix, not array. $\endgroup$ – John Moeller May 6 '14 at 1:14
  • $\begingroup$ Thanks, Ill change that now. I updated the question. $\endgroup$ – Aric May 6 '14 at 1:15
  • 1
    $\begingroup$ What's confusing everyone is that $A$ is in fact equal to $B$. $\endgroup$ – John Moeller May 6 '14 at 1:16

If $A = Q^{-1}DQ$, then $$ A^n = \underbrace{(Q^{-1}DQ)(Q^{-1}DQ) \ldots (Q^{-1}DQ)}_{n\text{ times}} = Q^{-1}D^nQ \text{,} $$ and if $D = \textrm{diag}(d_1,\ldots,d_k)$ then $D^k = \textrm{diag }(d_1^n,\ldots,d_k^n)$.

It follows that for $A = Q^{-1}DQ$ and $D = \textrm{diag }(d_1,\ldots,d_k)$ $$\begin{eqnarray} (1) && \lim_{n\to\infty} A^n \text{ exists if } d_i \in (-1,1] \text{ for all } i, &\text{ and in this case} \\ (2) && \lim_{n\to\infty} A^n = Q^{-1}\hat{D}Q \quad \text{where} & \hat D = \textrm{diag }(\hat d_1,\ldots,\hat d_k),\, \\ &&& \hat d_i = \begin{cases} 1 &\text{if $d_i=1$} \\ 0 &\text{if $-1 < d_i < 1$.} \end{cases} \end{eqnarray}$$

Note that the conditions on the $d_i$ simply guarantee that $\lim_{n\to\infty} d_i$ exists, and $\hat d_i$ is then defined so that $\hat d_i = \lim_{n\to\infty} d_i$. Also note that the $d_i$ are exactly the eigenvalues of $A$.

It follows that, at least for a diagonalizable $A$, $\lim_{n\to\infty} A^n$ exists exactly if all the eigenvalues of $A$ lie within $(-1,1]$.

  • $\begingroup$ I wish I thought of this. Thanks, I focused too much on the eigenvalues and completely missed this. $\endgroup$ – Aric May 6 '14 at 1:23
  • 1
    $\begingroup$ @Klick Well, you need the eigenvalues of $A$, because those are exactly the $d_i$. $\endgroup$ – fgp May 6 '14 at 1:25

Hint: Since you know that $Q$ is invertible, you know that you can set $A = QDQ^{-1}$, right? So what's $A^2$? $A^3$? $A^n$?


The matrix $A$ will have eigenvalues $1$ and $\frac{1}{2}$, so $$A^n=Q\pmatrix{1&0\cr0&({\textstyle\frac{1}{2}})^n\cr}Q^{-1}\ .$$ Can you see what you get for the bit in the middle as $n\to\infty$? If you can do this, then since you have already found $Q$, the problem is solved.


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.