Matrix Exponential and series idenfication

Click For Summary

Discussion Overview

The discussion revolves around finding the matrix exponential \(e^{\mathbf{A}t}\) for a specific matrix \(\mathbf{A}\). Participants explore various methods for computing the matrix exponential, including diagonalization and numerical approaches, while also addressing challenges related to the eigenvalues of the matrix.

Discussion Character

  • Exploratory, Technical explanation, Debate/contested

Main Points Raised

  • One participant presents the matrix \(\mathbf{A}\) and the series expansion for \(e^{\mathbf{A}t}\), noting difficulty in identifying the series beyond the 6th term.
  • Another participant suggests diagonalizing the matrix \(\mathbf{A}\) to simplify the computation of \(e^{At}\), but points out complications due to irrational and complex eigenvalues.
  • A later reply reiterates the diagonalization approach, mentioning the need for numerical methods due to the nature of the eigenvalues.
  • Some participants reference a document outlining various methods to compute matrix exponentials, labeling them as "dubious" and cautioning against their use.
  • One participant shares a Mathematica command that produces a result for the matrix exponential, but expresses uncertainty about the exactness of the solution.
  • Another participant corrects their earlier approach by noting a mistake in the matrix elements, indicating that this adjustment resolves their issue.

Areas of Agreement / Disagreement

Participants do not reach a consensus on the best method for computing the matrix exponential, with multiple competing views and approaches discussed throughout the thread.

Contextual Notes

Challenges include the irrational and complex nature of the eigenvalues, as well as the limitations of various computational methods discussed.

Dustinsfl
Messages
2,217
Reaction score
5
Let
\[
\mathbf{A} =
\begin{bmatrix}
0 & 1 & 0\\
0 & 0 & 1\\
-4 & -5 & -4
\end{bmatrix}
\]
Then I want to find \(e^{\mathbf{A}t}\).
\[
\mathbf{I} + \mathbf{A}t +\frac{\mathbf{A}^2t^2}{2!} + \frac{\mathbf{A}^3t^3}{3!} + \cdots
\]
I have up to the 6th term but I can't identify the series.
\[
\begin{bmatrix}
1 - \frac{2t^3}{3} + \frac{2t^4}{3} - \frac{11t^5}{30} + \frac{7t^6}{45} &
t - \frac{5t^3}{6} + \frac{2t^4}{3} - \frac{13t^5}{40} + \frac{2t^6}{15} &
\frac{t^2}{2} - \frac{2t^3}{3} + \frac{11t^4}{24} - \frac{7t^5}{30} + \frac{73t^6}{720}\\
-2t^2 + \frac{8t^3}{3} - \frac{11t^4}{6} + \frac{14t^5}{15} - \frac{73t^6}{180} &
1 - \frac{5t^2}{2} + \frac{8t^3}{3} - \frac{13t^4}{8} + \frac{4t^5}{5} - \frac{253t^6}{720} &
-\frac{49 t^6}{180} + \frac{73 t^5}{120} - \frac{7 t^4}{6} + \frac{11 t^3}{6} - 2t^2 + t\\
\frac{49 t^6}{45} - \frac{73 t^5}{30} + \frac{14 t^4}{3} - \frac{22 t^3}{3} + 8t^2 - 4t &
\frac{43 t^6}{45} - \frac{253 t^5}{120} + 4t^4 - \frac{13 t^3}{2} + 8t^2 - 5t &
\frac{59 t^6}{80} - \frac{49 t^5}{30} + \frac{73 t^4}{24} - \frac{14t^3}{3} + \frac{11 t^2}{2} - 4t + 1
\end{bmatrix}
\]
 
Physics news on Phys.org
In principle you can diagonalise $A$, $A = U^*DU$, where $U$ is unitary and $D$ is diagonal (with the eigenvalues of $A$ as its diagonal elements). Then $e^{At} = U^*D^{At}U$, and you can easily evaluate $D^{At}$ by taking the exponentials of the eigenvalues. Unfortunately in this case the eigenvalues are irrational (and two of them are complex). Maybe you can use some numerical method to get an approximate diagonalisation?
 
Opalg said:
In principle you can diagonalise $A$, $A = U^*DU$, where $U$ is unitary and $D$ is diagonal (with the eigenvalues of $A$ as its diagonal elements). Then $e^{At} = U^*D^{At}U$, and you can easily evaluate $D^{At}$ by taking the exponentials of the eigenvalues. Unfortunately in this case the eigenvalues are irrational (and two of them are complex). Maybe you can use some numerical method to get an approximate diagonalisation?

I tried that method first that is why I moved on to this form.
 
I can't pretend to be an expert, but http://www.cs.cornell.edu/cv/researchpdf/19ways+.pdf outlining 19 "dubious" ways to compute a matrix exponential. Use at your own risk!
 
Ackbach said:
I can't pretend to be an expert, but http://www.cs.cornell.edu/cv/researchpdf/19ways+.pdf outlining 19 "dubious" ways to compute a matrix exponential. Use at your own risk!

I have read that plus the numerous other links that come up from Googling the topic.
 
dwsmith said:
I have read that plus the numerous other links that come up from Googling the topic.

Well, then you know more than I do about matrix exponentials! I hope you find a do-able method for your problem. The command

Code:
N[MatrixExp[{{0, 1, 0}, {0, 0, 1}, {-4, -5, -4}}t]] // MatrixForm

gives a result in Mathematica. It's pretty long and ugly - I don't know how exact a solution you need.
 
Found the problem, I should have had a -6 not -5 and everything works.
 

Similar threads

  • · Replies 2 ·
Replies
2
Views
2K
  • · Replies 4 ·
Replies
4
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K
  • · Replies 5 ·
Replies
5
Views
992
  • · Replies 12 ·
Replies
12
Views
3K
  • · Replies 1 ·
Replies
1
Views
682
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 3 ·
Replies
3
Views
814
  • · Replies 14 ·
Replies
14
Views
2K
  • · Replies 1 ·
Replies
1
Views
4K