How I can compute a matrix given the characteristic equation?

All I found are references and functions that do the exact opposite, but I know the characteristic equation and I need the corresponding matrix / linear system.

For example, the equation is

$$ ( \lambda - \alpha ) ( \lambda - \beta ) = 0 $$

I have to find the matrix $A$ such that $A$ itself maps to this polynomial.


migrated from mathematica.stackexchange.com Jul 29 '16 at 16:03

This question came from our site for users of Wolfram Mathematica.

  • 1
    $\begingroup$ You have a uniqueness problem here. The matrices $$\begin{pmatrix}1&0\\0&1\end{pmatrix}$$ and $$\begin{pmatrix}1&1\\0&1\end{pmatrix}$$ have the same characteristic polynomial, but drastically different eigensystems. $\endgroup$ – J. M. is a poor mathematician Jul 28 '16 at 18:45
  • $\begingroup$ @J.M. I think that my specific problem maps to the second matrix since it cannot be diagonalized, infact I did some math before and that matrix is basically the system that I got from my pen and paper work $\endgroup$ – user356815 Jul 28 '16 at 18:49
  • $\begingroup$ @MichaelE2 since I'm a beginner with both, I'll say both, I'm trying to see if this can help me advance in my studies . $\endgroup$ – user356815 Jul 28 '16 at 18:50
  • 1
    $\begingroup$ So, tell me what you know about the Jordan form… $\endgroup$ – J. M. is a poor mathematician Jul 28 '16 at 18:51
  • $\begingroup$ This might be related to the given problem Finding the characteristic polynomial of a matrix modulus n $\endgroup$ – Artes Jul 28 '16 at 18:56

This gives you the entries for such a matrix:

mat = {{a, b}, {c, d}};
SolveAlways[CharacteristicPolynomial[mat, λ] == (λ - α) (λ - β), λ]

However, it is probably preferable to choose mat such that it has only as many unknowns as the matrix dimension.

  • $\begingroup$ I don't get the notation on the output, it's the first time I see this, I get $\begin{pmatrix} \alpha \rightarrow 1 & \beta \rightarrow 1 \\ \alpha \rightarrow 1 & \beta \rightarrow 1 \\ \end{pmatrix} $ for mat = {{1,1},{0,1}}, which means that the matrix elements are always 1 ? $\endgroup$ – user356815 Jul 28 '16 at 19:02

If you are working over an algebraically closed field: Any matrix $\bf A$ you can write on this form:

$${\bf A = T}^{-1}{\bf DT}$$

If there exists such a relation it is said that $\bf A$ and $\bf D$ are similar to each other.

Where $\bf D$ is diagonal or block diagonal with elements / blocks $\lambda_k$ or $\left[\begin{array}{cccc} \lambda_k&1&0&0\\ 0&\lambda_k&\ddots&0\\ 0&0&\lambda_k&1\\ 0&0&0&\lambda_k \end{array}\right]$

In this case it will be either a diagonalization ( if there are only $\lambda_k$ and $0$ in $\bf D$ matrix ) and a Jordan canonical form if there are blocks like the one to the right with $1$ values on the first superdiagonal.

$\bf T$ is free to design as you will - as long as it has determinant $\neq 0$.

The values on the diagonal of $\bf D$ are fixed to $\lambda_k$, but not how large blocks or contents of $\bf T$.

  • $\begingroup$ what is the technical name for the math process that can lead you to rewrite$${\bf A}$$ as $${\bf A = T}^{-1}{\bf DT}$$ ? $\endgroup$ – user356815 Jul 29 '16 at 16:30
  • $\begingroup$ It is a matrix similarity. en.wikipedia.org/wiki/Matrix_similarity $\endgroup$ – mathreadler Jul 29 '16 at 16:37
  • 1
    $\begingroup$ perhaps you could have mentioned diagonalization $\endgroup$ – windircurse Jul 29 '16 at 16:38
  • $\begingroup$ Yes added. Also Jordan form. $\endgroup$ – mathreadler Jul 29 '16 at 16:41

Given the characteristic polynomial

$$(s - \alpha) (s - \beta)$$

we know that the eigenvalues are $\alpha$ and $\beta$. Hence, one matrix that yields the characteristic polynomial above is the diagonal matrix

$$\begin{bmatrix} \alpha & 0\\ 0 & \beta\end{bmatrix}$$

Another matrix is the upper triangular matrix

$$\begin{bmatrix} \alpha & 1\\ 0 & \beta\end{bmatrix}$$

Yet another matrix is the lower triangular matrix

$$\begin{bmatrix} \alpha & 0\\ 1 & \beta\end{bmatrix}$$

There are infinitely many more, though.


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.