Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

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

If I wanted to calculate the matrix exponential of $\left ( \begin{array}{cc} -3 & 4 \\ -4 & -3 \end{array} \right )$, I could do this by calculating the eigenvectors and values and then use $e^{At}=P\ \mbox{diag}(e^{\lambda t}) P^{-1}$ where $P$ is the matrix with the eigenvectors of $A$ as its columns and I get $$e^{At}= e^{-3t}\left ( \begin{array}{cc} \cos(4t) & \sin(4t) \\ -\sin(4t) & \cos(4t) \end{array} \right )$$ but I was wondering how I do this using the definition, I'm sure it shouldn't be tough but I'm getting a bit lost; I have: $e^{At}=\sum\limits^{\infty}_{m=0} \frac{1}{m!}(At)^m$ but then looking at the powers of this matrix I just get rubbish i think? Thanks very much for any help.

share|cite|improve this question
Cayley-Hamilton theorem gives a quadratic polynomial $\phi(x)$ for which the matrix vanish. So the problem reduces to the calculation of the remainder of $x^n$ divided by $\phi(x)$. But this method also requires the knowledge about eigenvalues, so it's not so different from the method you know. There are several methods that circumvents this problems, which I know not so well. – Sangchul Lee Mar 31 '12 at 12:03
Your $e^{At}$ does not correspond to your $A$. – Did Mar 31 '12 at 12:54
@Didier sorry I edited it, should be -3 in the top left, however my tex was right just the - didnt show up. I have had to put 2 in now so that one shows up which is a bit strange, sorry about that should have checked my post but as I say it something to do with the tex. – hmmmm Mar 31 '12 at 13:10
Funny bug... A somewhat surer way to encode this is to use \begin{array}{cc}, see the code of the revised version of your post. // The new version of your question has an even simpler solution than the previous one, see the second part of my post. – Did Mar 31 '12 at 14:52
@Didier Thanks for the answer and again sorry for that bug, i'll keep the {cc} in mind next time thanks – hmmmm Mar 31 '12 at 15:01
up vote 7 down vote accepted

$\color{blue}{\text{For a solution to the second version of the question, see below}}$.

This applies to the first version of the question, where $\color{red}{A=\begin{pmatrix}3 & 4 \\ -4 & -3\end{pmatrix}}$. Since $\text{tr}(A)=0$ and $\det(A)=7$, Cayley-Hamilton theorem yields $A^2=-7I$. Hence, for every $n\geqslant0$, $A^{2n}=(-7)^n\cdot I$ and $A^{2n+1}=(-7)^n\cdot A$. This yields $$ \mathrm e^{tA}=\sum\limits_{n=0}^{+\infty}\frac{t^n}{n!}\cdot A^n=\sum\limits_{n=0}^{+\infty}(-1)^n7^n\frac{t^{2n}}{(2n)!}\cdot I+\sum\limits_{n=0}^{+\infty}(-1)^n7^n\frac{t^{2n+1}}{(2n+1)!}\cdot A, $$ that is, $$ \color{red}{\mathrm e^{tA}=\cos(\sqrt7t)I+\frac{\sin(\sqrt7t)}{\sqrt7}A}. $$ The second version of the question, where $\color{blue}{A=\begin{pmatrix}-3 & 4 \\ -4 & -3\end{pmatrix}}$, can be solved even more easily. Since $A+3I=4J$ with $J=\begin{pmatrix}0 & 1 \\ -1 & 0\end{pmatrix}$, one has $\mathrm e^{tA}=\mathrm e^{-3t}\cdot\mathrm e^{4tJ}$. Note that $J^2=-I$ (and/or that $J$ is nothing but the matrix of the clockwise rotation by $90°$), hence, for every $n\geqslant0$, $J^{2n}=(-1)^nI$ and $J^{2n+1}=(-1)^nJ$. This yields $$ \mathrm e^{tJ}=\sum\limits_{n=0}^{+\infty}\frac{t^n}{n!}\cdot J^n=\sum\limits_{n=0}^{+\infty}(-1)^n\frac{t^{2n}}{(2n)!}\cdot I+\sum\limits_{n=0}^{+\infty}(-1)^n\frac{t^{2n+1}}{(2n+1)!}\cdot J, $$ that is, $$ \mathrm e^{tJ}=\cos(t)I+\sin(t)J. $$ Coming back to $A$ and using the formula for $\mathrm e^{4tJ}$, one gets $$ \color{blue}{\mathrm e^{tA}=\mathrm e^{-3t}\cdot\left(\cos(4t)I+\sin(4t)J\right)}. $$

share|cite|improve this answer
Thanks for the response, sorry about the question being wrong -I think that changes the answer somewhat. Don't know what is going on with the tex – hmmmm Mar 31 '12 at 13:11

Post problem change; i.e $A=\begin{bmatrix}-3&4\\-4&-3\end{bmatrix}$. $A^2+6A+25I=0$ therefore $A^n=a_nA+b_nI$ for some $a_n$ and $b_n$. Note that $$ \begin{align} a_nA+b_n &=(a_{n-1}A+b_{n-1}I)A\\ &=a_{n-1}(-6A-25I)+b_{n-1}A\tag{1} \end{align} $$ Therefore, $a_n=-6a_{n-1}+b_{n-1}$ and $b_n=-25a_{n-1}$; that is, $$ A^n=a_nA-25a_{n-1}I\tag{2} $$ where $$ a_n=-6a_{n-1}-25a_{n-2}\tag{3} $$ Solving $(3)$ with $a_1=1$ and $a_0=0$ yields $$ a_n=\frac{(-3+4i)^n-(-3-4i)^n}{8i}\tag{4} $$ Now $$ \begin{align} e^{tA} &=\sum_{n=0}^\infty\frac{1}{n!}t^nA^n\\ &=\sum_{n=0}^\infty\frac{t^n}{n!}\left(\frac{(-3+4i)^n-(-3-4i)^n}{8i}A-25\frac{(-3+4i)^{n-1}-(-3-4i)^{n-1}}{8i}I\right)\\ &=\frac14e^{-3t}\sin(4t)A-\frac{1}{4}e^{-3t}(-3\sin(4t)-4\cos(4t))I\\ &=e^{-3t}\begin{bmatrix}\cos(4t)&\sin(4t)\\-\sin(4t)&\cos(4t)\end{bmatrix}\tag{5} \end{align} $$

share|cite|improve this answer

I know you explicitly wanted to use the definition, but another useful way to get $e^{tA}$ would be to use the Laplace transform. $$ e^{tA}=\mathcal{L}^{-1}\left((sI-A)^{-1}\right)= \mathcal{L}^{-1}\left(\begin{bmatrix}s+3 & -4 \\ 4 & s+3\end{bmatrix}^{-1}\right) =\mathcal{L}^{-1}\left(\frac{1}{(s+3)^2+16}\begin{bmatrix}s+3 & 4 \\ -4 & s+3\end{bmatrix}\right)= e^{-3t}\mathcal{L}^{-1}\left(\frac{1}{s^2+16}\begin{bmatrix}s & 4 \\ -4 & s\end{bmatrix}\right) =e^{-3t}\begin{bmatrix}\cos{4t} & \sin{4t} \\ -\sin{4t} & \cos{4t}\end{bmatrix}. $$

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.