Is there a formula in radicals for the eigenvalues, or at least the largest eigenvalue, of an $n\times n$ symmetric matrix, in terms of the entries? For what $n$ is there a formula? There obviously is if $n\leq 4$, and one suspects not if $n\geq 5$. However, the characteristic polynomial of a symmetric matrix is rather special (e.g. all roots are real) and could possibly be solvable by radicals for larger $n$, although I would guess not.

  • 1
    $\begingroup$ Have you tried computing the Galois group of a random example? $\endgroup$ – Douglas Zare Apr 28 '17 at 3:20
  • 5
    $\begingroup$ Edward Bender, Characteristic ppolynomials of symmetric matrices, proved that any totally real monic polynomial of odd degree with rational coefficients can occur as the characteristic polynomial of a symmetric rational matrix. Some discussion at math.stackexchange.com/questions/574443/… $\endgroup$ – Gerry Myerson Apr 28 '17 at 3:40
  • 3
    $\begingroup$ Here us a formula literally "in radicals" for the largest eigenvalue, although probably not what you had in mind: $$ \lambda_1=\lim_{n\to\infty}\sqrt[n]{\|A^n\|}.$$ (Of course, if you want a purely algebraic expression, there is no way to distinguish the largest eigenvalue from the others.) $\endgroup$ – Victor Protsak Apr 28 '17 at 5:38
  • 3
    $\begingroup$ About Gerry's comment. Given a totally real monic polynomial $P$ of degree $n$, a simple procedure constructs a "companion" in the form of a real symmetric tridiagonal matrix $T$. If $P$ has rational coefficients (or in a number field $k$), then $T$ has rational entries (resp. in $k$). See Exercise #92 in my list perso.ens-lyon.fr/serre/DPF/exobis.pdf . $\endgroup$ – Denis Serre Apr 28 '17 at 6:16
  • $\begingroup$ If one takes your question more broadly as a sign that you need to know more about maps (class of numbers to take the entries of the matrix from)$\rightarrow$(class of all numbers that arise as ev of symmetric matrices), then the following might help: for any finite symmetric matrix, if all entries come from a fixed real-closed field $K$, then then all eigenvalues of the matrix are still in $K$. $\endgroup$ – Peter Heinig Apr 28 '17 at 7:41

If I calculated correctly, $$\begin{pmatrix}2 & 1 & 0 & 0 & 0\newline 1 & 3 & 1 & 0 & 0\newline 0 & 1 & 1 & 1 & 0 \newline 0 & 0 & 1 & 1 & 1 \newline 0 & 0 & 0 & 1 & 1 \end{pmatrix}$$ has characteristic polynomial $-x^5+8x^4-20x^3+15x^2+4x-5$ which is irreducible mod $3$ and has $3$ linear factors and a quadratic factor mod $79$, so its Galois group is $S_5$, so its roots can't be expressed with just radicals.


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.