1. Limited time only! Sign up for a free 30min personal tutor trial with Chegg Tutors
    Dismiss Notice
Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Homework Help: Isometries math problem

  1. Aug 14, 2009 #1
    1. The problem statement, all variables and given/known data
    Prove or give a counterexample: if S ∈ L(V) and there exists
    an orthonormal basis (e1, . . . , en) of V such that llSejll = 1 for
    each ej , then S is an isometry.

    2. Relevant equations

    3. The attempt at a solution
    Can't think of a counterexample. I am assuming that since ej is a member
    of an orthonormal basis, llejll HAS to be 1 by definition. And if llSejll=1,
    then llSejll=llejll. I mean, considering that llSejll=1 for ALL ej,
    all ej have 1 as an eigenvalue. Should I go ahead and do the proof, because
    I am having trouble seeing how S is not an isometry given the information I posted
  2. jcsd
  3. Aug 14, 2009 #2
    Re: isometries

    You could start by proving that every basis vector has the same length after transformation ( Hint: squared length ). And if every vector is a linear combination of these basis vectors, do the same proof has above, with a little more coefficients.
  4. Aug 14, 2009 #3


    User Avatar
    Science Advisor
    Homework Helper

    Re: isometries

    ||Sej||=||ej||=1 DOES NOT mean 1 is an eigenvalue with ej as an eigenvector. IF ej is an eigenvector of S with eigenvalue 1, THEN ||Sej||=||ej||=1. The converse is emphatically NOT TRUE. You make this mistake A LOT. Stop it! Construct a counterexample in R^2. Quick. DON'T try to prove it. I know you will throw a 'proof' together. That would be a bad thing.
    Last edited: Aug 14, 2009
  5. Aug 15, 2009 #4
    Re: isometries

    I digress, but to clear up any misconceptions I have about isometries,
    lets say we have a vector (2, 2). apply T (2x2 matrix of rank 1 with sqrt(8)/2 at
    the far upper left corner and 0 everywhere else) to (2, 2) to get (sqrt(8), 0).
    (2, 2) is not an eigenvector of T and T(2, 2)=/=1*(2, 2). But llT(2, 2)ll=ll(2, 2)ll.

    (sqrt(2), sqrt(-1)) Let T be a 2x2 matrix of rank 1 with sqrt(1/2)at the far upper left corner and 0 everywhere else. T(sqrt(2), sqrt(-1))=(1, 0). Now applying T again we get
    (sqrt(1/2), 0). llT^2(sqrt(2), sqrt(-1))ll=sqrt(1/2)=/=1. This shows that llTekll=llekll does not imply that
    T is an isometry since all isometries are such that llT^kekll=llekll where k is any number >=1.
    Last edited: Aug 15, 2009
  6. Aug 15, 2009 #5
    Re: isometries

    [tex]\left\|(\sqrt{2},i)\right\| = (\sqrt{2},i) \cdot (\sqrt{2},i)^* = 3[/tex]
  7. Aug 15, 2009 #6
    Re: isometries

    Oh right. :rofl: Ok here's plan B. (sqrt(2/3), sqrt(1/3) ). Now
    let T's 2x2 matrix be of rank 1 with sqrt(3/2) at the top left hand corner
    and 0 everywhere else. Now T(sqrt(2/3), sqrt(1/3))=(1, 0). Now applying T
    again, we get (sqrt(2/3), 0). ll(sqrt(2/3), 0)ll=sqrt(2/3).
  8. Aug 15, 2009 #7
    Re: isometries

    But you don't have to apply T twice to see that (1,0) do not preserves length. Write it in term of the basis vectors. And you will see that only one basis vector preserves its length.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook