11

This question aims at creating an "abstract duplicate" of various questions that can be reduced to the following:

Let $A$ be an $n\times n$ Hermitian matrix and $B$ be an $r\times r$ principal submatrix of $A$. How are the eigenvalues of $A$ and $B$ related?

Here are some questions on this site that can be viewed as duplicates of this question:

user1551
  • 139,064

1 Answers1

14

Proposition. Let $\lambda_k(\cdot)$ denotes the $k$-th smallest eigenvalue of a Hermitian matrix. Then $$ \lambda_k(A)\le\lambda_k(B)\le\lambda_{k+n-r}(A),\quad 1\le k\le r. $$

This is a well-known result in linear algebra. Since the usual proof is just a straightforward application of the celebrated Courant-Fischer minimax principle, we shall not repeat it here. See, e.g. theorem 4.3.15 (p.189) of Horn and Johnson, Matrix Analysis, 1/e, Cambridge University Press, 1985.

user1551
  • 139,064
  • 6
    They gave the Matrix Analysis book the volume number $\exp(-1)\approx0.3679$? How cute! – Marc van Leeuwen Feb 24 '16 at 11:01
  • 2
    Just thought of adding a link to the answer, in case someone is looking at editiopns 2e or 3e:

    http://www.cse.zju.edu.cn/eclass/attachments/2015-10/01-1446086008-145421.pdf, see P. 4.3.28.

    – Learning Math Dec 31 '19 at 21:30
  • 1
    I am looking for a similar result for not necessarily symetric or Hermitian matrices. I tried to follow the cited proof (as well as thm. 4.3.28 in the second edition) but could not find a point in the proof where it strictly needs the Hermitian property. Where does it fail? Are there similar results? – BadAtLaTeX May 19 '21 at 08:56