home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!europa.eng.gtefsd.com!gatech!paladin.american.edu!howland.reston.ans.net!sol.ctr.columbia.edu!The-Star.honeywell.com!umn.edu!gaia.ucs.orst.edu!pmontgom
- From: pmontgom@math.orst.edu (Peter Montgomery)
- Newsgroups: sci.math
- Subject: Re: Matrix Problem
- Date: 28 Jan 1993 19:15:34 GMT
- Organization: Oregon State University Math Department
- Lines: 53
- Message-ID: <1k9bcmINN34t@gaia.ucs.orst.edu>
- References: <C1Knzo.54s@ecf.toronto.edu>
- NNTP-Posting-Host: ilanga.math.orst.edu
-
- In article <C1Knzo.54s@ecf.toronto.edu> rairan@ecf.toronto.edu (RAI Ranjan)
- writes:
- >I remember a result from first year linear algebra that said that the
- >eigenvectors of AB are the same as the eigenvectors of BA where A and B
- >are nxn real matrices. I don't know how to prove this. Any pointers or
- >references to the proof?
-
- AB and BA have the same eigenvalues, but not the same eigenvectors.
-
- Consider two matrices A and B with symbolic entries.
- The characteristic polynomials det(lambda*I - AB) and
- det(lambda*I - BA) are polynomials in these entries and lambda.
- We claim that these are the same polynomial.
-
- Since A has symbolic entries, it is invertible in
- the field generated by its coefficients. Then
-
- det(lambda*I - AB) = det(lambda*A*A^(-1) - AB)
- = det(A) * det(lambda*A^(-1) - B)
- = det(lambda*A^(-1) - B) * det(A)
- = det(lambda*A^(-1)*A - BA)
- = det(lambda*I - BA)
-
- in the field of rational functions generated by the symbolic coefficients
- of A and B. Since both det(lambda*I - AB) and det(lambda*I - BA) are
- polynomials, they must also agree as polynomials.
-
- If two polynomials are the same, then they generate the same
- polynomial function. Upon substituting numerical values into the coefficients
- of A and B, the two characteristic polynomials will agree.
- Hence they have the same roots, which are the eigenvalues of AB and BA.
-
- The reader may feel uncomfortable if his numerical A is
- singular (i.e. non-invertible). As an analogy, consider the derivation
-
- x^6 - 1 (x^3 - 1)(x^3 + 1)
- x^4 + x^2 + 1 = ------- = ------------------ = (x^2 + x + 1)(x^2 - x + 1)
- x^2 - 1 (x - 1)(x + 1)
-
- of the polynomial identity x^4 + x^2 + 1 = (x^2 + x + 1)(x^2 - x + 1).
- The left and right sides agree as rational functions (by the derivation)
- and hence must be the same polynomial. Therefore the two agree
- at all numerical values of x, including x = +- 1, even though we
- divided by x - 1 and x + 1 during the proof.
-
- EXERCISE: Find two 2 x 2 matrices A, B such that AB and BA
- are not similar (i.e., there is no 2 x 2 invertible matrix C such that
- (AB)C = C(BA).
- --
- Peter L. Montgomery Internet: pmontgom@math.orst.edu
- Dept. of Mathematics, Oregon State Univ, Corvallis, OR 97331-4605 USA
- My visiting professorship ends in June. Interested in program optimization,
- compilers, computer arithmetic, number theory. 17 yrs industrial experience.
-