From a comment on this question:

@draks, there is a connection between the Chebyshev polynomials and the Faber polynomials (a.k.a. Shur polynomials), which 'invert" the cyclic partition polynomials (a.k.a. the refined partition polynomials of the first kind--signed OEIS A036039) to isolate the indeterminates in their definition.

Interesting. Do you have something where I can read on this?

@Draks, see oeis.org/A263916 and oeis.org/A127672 along with the Damianou and Damianou and Evripidou links.

The links don't help. Can anyone explain the connection?


The characteristic polynomial and the traces of the powers of a square matrix $A$ are related by

$$\operatorname{det}(I-x \; A)=\exp\left(-\sum_{m\geq 1} \frac{\operatorname{tr}(A^m) \; x^m}{m} \right)$$

$$=1 + \sum_{k \ge 1} P_k(b_1,..,b_k) \; x^k / k! = 1 + \sum_{k \geq 1} d_k \; x^k$$

where $b_k=-\operatorname{tr}(A^k)$ for $k>0$ and $P_m(b_1,..,b_m)$ are the cycle index partition polynomials of OEIS A036039 for the symmetric groups (see the Lang link there for a compilation). Presented umbrally,

$$\operatorname{det}(I - x \; A)=\exp[\ln(1 - b.\;x)]=e^{P. \; x}=1/(1-d.\;x) \; .$$

The Faber polynomials $F_m$ of A263916 can, in general, invert the the determinant polynomial, or the partition polynomials $P_n$, to generate the power sums, or traces:

$$F_m[P_1(b_1),P_2(b_1,b_2)/2!,..,P_m(b_1,..,b_m)/m!] = F_m(d_1,d_2,..,d_m) = -b_m = tr(A^m) \; .$$

Damianou points out that the adjacency matrix $A_n$ for the simple Lie algebra $B_n$ has the characteristic polynomial

$$det(xI - A_n) = 2\; T_n(x/2)= a_n(x)\; ,$$

the Chebyshev polynomials given by A127672, so with

$$ det(I - x\;A_n) = x^n \; a_n(1/x) = 2 x^n \; T(1/(2x)) = 1 + \sum_{k = 1}^{n} d_k \; x^k \; ,$$

$$F_m(d_1,d_2,..,d_m) = -b_m = tr(A^m) \;,$$

unwrapping the determinant relation.

An example is afforded by the adjacency matrix for for the simple Lie algebra $B_3$, given by Damianou:

$$A_3=\begin{bmatrix} 0 & 1 & 0\\ 1 & 0 & 2\\ 0 & 1 & 0 \end{bmatrix} \; .$$

The characteristic polynomial $\operatorname{det}(xI-A_3)=a_3(x) = 2T_3(x/2)= x^3 -3x=$ has the zeroes $x_1=\sqrt{3},x_2=-\sqrt{3},x_3=0$, so $\operatorname{tr}(A_3^k)= x_1^k + x_2^k +x_3^k=3^{k/2} + (-1)^k 3^{k/2}$ (aerated A025192), which vanishes for odd $k$ and equals $2\; 3^{k/2}$ for even $k$. Then $P_k(b_1,..,b_k)$ vanishes except for $P_2(0,-6) = -6$, and $\operatorname{det}(I-uA_3)= 1-3u^2=u^3a_3(1/u) = 2x^3T_3(1/2x)) $, so only $d_2 = -3$ is non-vanishing and $F_k[0,-3,0,..,0]= -b_k = \operatorname{tr}(A_3^k)$ for $k>0$.

| cite | improve this answer | |
  • $\begingroup$ See A263916 for other connections of the Faber polynomials to the Chebyshev polynomials. $\endgroup$ – Tom Copeland Dec 4 '15 at 4:59
  • $\begingroup$ Follow the OEIS links for relations to walks on paths of P_n and K_n graphs. $\endgroup$ – Tom Copeland Dec 8 '15 at 22:47

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.