Search the corpus

    Trace as a weighted sum of characteristic values

    E-HK-63-9

    Area
    Algebra
    Topics
    Characteristic Values, Linear Algebra
    Status
    draft

    Let \(A\) be an \(n \times n\) matrix with characteristic polynomial \[ f = (x - c_1)^{d_1} \cdots (x - c_k)^{d_k}. \] Show that \[ c_1 d_1 + \cdots + c_k d_k = \operatorname{trace}(A). \]

    Solution

    1. Coefficient of \(x^{n-1}\) from the determinant definition of the characteristic polynomial:

    1.1. By definition, the characteristic polynomial of \(A = [a_{ij}] \in M_n(F)\) is \(f(x) = \det(xI - A)\).

    definition of characteristic polynomial.

    1.2. By the Leibniz formula for the determinant:\[ \det(xI - A) = \sum_{\sigma \in S_n} \operatorname{sgn}(\sigma) \prod_{i=1}^n (x \delta_{i, \sigma(i)} - a_{i, \sigma(i)}). \]

    Leibniz formula for determinants.

    1.3. For any non-identity permutation \(\sigma \neq \operatorname{id}\), \(\sigma\) fixes at most \(n - 2\) elements, so the corresponding product has degree at most \(n - 2\) in \(x\).

    permutations with at least \(n - 1\) fixed points must be the identity.

    1.4. Thus the terms of degree \(n\) and \(n - 1\) arise solely from the diagonal product (\(\sigma = \operatorname{id}\)):\[ \prod_{i=1}^n (x - a_{ii}) = x^n - \left(\sum_{i=1}^n a_{ii}\right) x^{n-1} + O(x^{n-2}) = x^n - \operatorname{trace}(A) x^{n-1} + O(x^{n-2}). \]

    expanding \(\prod_{i=1}^n (x - a_{ii})\) and definition of trace.

    1.5. Thus the coefficient of \(x^{n-1}\) in \(f(x)\) is \(-\operatorname{trace}(A)\).

    3 and 4.

    2. Coefficient of \(x^{n-1}\) from the factored form of \(f(x)\):

    2.1. The factored characteristic polynomial is \(f(x) = (x - c_1)^{d_1} \cdots (x - c_k)^{d_k}\), with \(\sum_{j=1}^k d_j = n\).

    given.

    2.2. By Vieta’s formulas, the coefficient of \(x^{n-1}\) in a monic polynomial with roots (counted with multiplicity) is the negative of the sum of the roots:\[ f(x) = x^n - \left(\sum_{j=1}^k d_j c_j\right) x^{n-1} + \cdots + (-1)^n \prod_{j=1}^k c_j^{d_j}. \]

    expanding \((x - \lambda_1) \cdots (x - \lambda_n)\) where \(\lambda_1, \dots, \lambda_n\) lists each \(c_j\) repeated \(d_j\) times.

    2.3. Thus the coefficient of \(x^{n-1}\) in \(f(x)\) is \(-(c_1 d_1 + \cdots + c_k d_k)\).

    2.

    3. Equating the coefficients:

    3.1. Equating the coefficients of \(x^{n-1}\) obtained in 1 and 2 gives:\[ -\operatorname{trace}(A) = -(c_1 d_1 + \cdots + c_k d_k). \]

    uniqueness of polynomial coefficients.

    3.2. Multiplying both sides by \(-1\) yields:\[ \operatorname{trace}(A) = c_1 d_1 + \cdots + c_k d_k. \]

    scalar multiplication by \(-1\).

    4. Conclusion:\(c_1 d_1 + \cdots + c_k d_k = \operatorname{trace}(A)\). Q.E.D.

    3.