Results 1 to 7 of 7
Like Tree6Thanks
  • 3 Post By HallsofIvy
  • 2 Post By Deveno
  • 1 Post By Deveno

Math Help - Preserving the inner product

  1. #1
    Senior Member Educated's Avatar
    Joined
    Aug 2010
    From
    New Zealand
    Posts
    433
    Thanks
    12

    Preserving the inner product

    Show that any map f: R^n \rightarrow R^n which preserves the inner product of R^n, i.e., for which \langle f(\mathbf{\em x}), f( \mathbf{\em y}) \rangle = \langle \mathbf{\em x},  \mathbf{\em y} \rangle for all \mathbf{\em x},  \mathbf{\em y} \in R^n, is linear:

    f( \alpha \mathbf{\em x} + \beta  \mathbf{\em y}) = \alpha f(\mathbf{\em x}) + \beta f( \mathbf{\em y}) for all \alpha , \beta \in R, and \mathbf{\em x}, \mathbf{\em y} \in R^n

    Hint: Show that the vector f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) - \alpha f(\mathbf{\em x}) - \beta f( \mathbf{\em y}) has vanishing (squared) length

    So first of course I used the hint, and the length squared is:

    =[f(\alpha \mathbf{\em x} + \beta \mathbf{\em y})]^2 + [\alpha f(\mathbf{\em x})]^2 + [\beta f( \mathbf{\em y})]^2

    = [f(\alpha \mathbf{\em x} + \beta \mathbf{\em y})\cdot f(\alpha \mathbf{\em x} + \beta \mathbf{\em y})] + [\alpha^2 f(\mathbf{\em x}) \cdot f(\mathbf{\em x})] + [\beta^2 f( \mathbf{\em y}) \cdot f( \mathbf{\em y})]


    = [(\alpha \mathbf{\em x} + \beta \mathbf{\em y})\cdot (\alpha \mathbf{\em x} + \beta \mathbf{\em y})] + [\alpha^2 \mathbf{\em x} \cdot \mathbf{\em x}] + [\beta^2 \mathbf{\em y} \cdot \mathbf{\em y}]

    = 2 \alpha \beta \mathbf{\em x} \mathbf{\em y} (because \mathbf{\em x} \cdot \mathbf{\em x} = 0)


    Now I want that to be zero, but I can't see how...
    Follow Math Help Forum on Facebook and Google+

  2. #2
    MHF Contributor

    Joined
    Apr 2005
    Posts
    16,232
    Thanks
    1795

    Re: Preserving the inner product

    Your very first step is incorrect. |a- b- c|^2 is NOT equal to |a|^2+ |b|^2+ |c|^2.
    Thanks from Deveno, Educated and topsquark
    Follow Math Help Forum on Facebook and Google+

  3. #3
    MHF Contributor

    Joined
    Mar 2011
    From
    Tejas
    Posts
    3,401
    Thanks
    762

    Re: Preserving the inner product

    Hint, compare:

    $\langle f(\alpha\mathbf{x} + \beta\mathbf{y}) - \alpha f(\mathbf{x}) - \beta f(\mathbf{y}), f(\alpha\mathbf{x} + \beta\mathbf{y}) - \alpha f(\mathbf{x}) - \beta f(\mathbf{y})\rangle$

    with:

    $\langle \alpha\mathbf{x} + \beta\mathbf{y} - \alpha\mathbf{x} - \beta\mathbf{y}, \alpha\mathbf{x} + \beta\mathbf{y} - \alpha\mathbf{x} - \beta\mathbf{y}\rangle = \langle \mathbf{0}, \mathbf{0}\rangle = 0$

    using the fact that $\langle f(\mathbf{u}),f(\mathbf{v})\rangle = \langle \mathbf{u},\mathbf{v}\rangle$ for ANY two vectors $\mathbf{u},\mathbf{v}$.

    You'll need to expand the first inner product using bilinearity, you should wind up with SIX terms.
    Thanks from Educated and topsquark
    Follow Math Help Forum on Facebook and Google+

  4. #4
    Senior Member Educated's Avatar
    Joined
    Aug 2010
    From
    New Zealand
    Posts
    433
    Thanks
    12

    Re: Preserving the inner product

    Okay, I see what I did wrong,

    =|| f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) - \alpha f(\mathbf{\em x}) - \beta f( \mathbf{\em y}) ||^2

    =f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) - f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot \alpha f(\mathbf{\em x}) - f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot \beta f( \mathbf{\em y}) - \alpha f(\mathbf{\em x}) \cdot f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) + \alpha f(\mathbf{\em x}) \cdot \alpha f(\mathbf{\em x}) + \alpha f(\mathbf{\em x}) \cdot \beta f( \mathbf{\em y}) - \beta f( \mathbf{\em y}) \cdot f(\alpha \mathbf{\em x} + \beta f( \mathbf{\em y}) \cdot \alpha f(\mathbf{\em x})  + \beta f( \mathbf{\em y}) \cdot \beta f( \mathbf{\em y})

    Which all cancels out and then equals 0
    Follow Math Help Forum on Facebook and Google+

  5. #5
    MHF Contributor

    Joined
    Mar 2011
    From
    Tejas
    Posts
    3,401
    Thanks
    762

    Re: Preserving the inner product

    Quote Originally Posted by Educated View Post
    Okay, I see what I did wrong,

    =|| f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) - \alpha f(\mathbf{\em x}) - \beta f( \mathbf{\em y}) ||^2

    =f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) - f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot \alpha f(\mathbf{\em x}) - f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot \beta f( \mathbf{\em y}) - \alpha f(\mathbf{\em x}) \cdot f(\alpha \mathbf{\em x} + \beta \mathbf{\em y}) + \alpha f(\mathbf{\em x}) \cdot \alpha f(\mathbf{\em x}) + \alpha f(\mathbf{\em x}) \cdot \beta f( \mathbf{\em y}) - \beta f( \mathbf{\em y}) \cdot f(\alpha \mathbf{\em x} + \beta f( \mathbf{\em y}) \cdot \alpha f(\mathbf{\em x})  + \beta f( \mathbf{\em y}) \cdot \beta f( \mathbf{\em y})

    Which all cancels out and then equals 0
    No, it doesn't "all cancel out".

    Continuing from what you have, we have:

    $\langle f(\alpha\mathbf{x}+\beta\mathbf{y}),f(\alpha \mathbf{x}+\beta\mathbf{y})\rangle - \alpha\langle f(\alpha\mathbf{x}+\beta\mathbf{y}),f(\mathbf{x}) \rangle - \beta\langle f(\alpha\mathbf{x}+\beta\mathbf{y}),f(\mathbf{y}) \rangle

    -\alpha\langle f(\mathbf{x}),f(\alpha\mathbf{x}+\beta\mathbf{y}) \rangle + \alpha^2\langle f(\mathbf{x}),f(\mathbf{x}) \rangle $

    $+ \alpha\beta \langle f(\mathbf{x}),f(\mathbf{y}) \rangle - \beta \langle f(\mathbf{y}), f(\alpha\mathbf{x}+\beta\mathbf{y}) \rangle + \alpha\beta\langle f(\mathbf{y}),f(\mathbf{x}) \rangle + \beta^2\langle f(\mathbf{y}),f(\mathbf{y}) \rangle$

    Which upon combining terms, and using the symmetry of the inner product becomes:

    $ = \langle f(\alpha\mathbf{x}+\beta\mathbf{y}),f(\alpha \mathbf{x}+\beta\mathbf{y})\rangle + \alpha^2\langle f(\mathbf{x}),f(\mathbf{x}) \rangle + \beta^2\langle f(\mathbf{y}),f(\mathbf{y}) \rangle$

    $- 2\alpha\langle f(\alpha\mathbf{x}+\beta\mathbf{y}),f(\mathbf{x}) \rangle - 2\beta\langle f(\alpha\mathbf{x}+\beta\mathbf{y}),f(\mathbf{y}) \rangle + 2\alpha\beta \langle f(\mathbf{x}),f(\mathbf{y}) \rangle$

    These are the six terms I was referring to. Note how I have all the "f's" inside the inner product brackets. By the inner-product preserving property, we can now get rid of these, so this is equal to:

    $ = \langle \alpha\mathbf{x}+\beta\mathbf{y},\alpha \mathbf{x}+\beta\mathbf{y}\rangle + \alpha^2\langle \mathbf{x},\mathbf{x} \rangle + \beta^2\langle \mathbf{y},\mathbf{y} \rangle$

    $- 2\alpha\langle \alpha\mathbf{x}+\beta\mathbf{y},\mathbf{x} \rangle - 2\beta\langle \alpha\mathbf{x}+\beta\mathbf{y},\mathbf{y} \rangle + 2\alpha\beta \langle \mathbf{x},\mathbf{y} \rangle$

    $= \langle \alpha\mathbf{x} + \beta\mathbf{y} - \alpha\mathbf{x} - \beta\mathbf{y}, \alpha\mathbf{x} + \beta\mathbf{y} - \alpha\mathbf{x} - \beta\mathbf{y} \rangle$

    (here I just did the reverse of the expansion we did first)

    $=\langle \mathbf{0},\mathbf{0} \rangle = 0$.

    Since the square of the length of $f(\alpha\mathbf{x}+\beta\mathbf{y}) - \alpha f(\mathbf{x}) - \beta f(\mathbf{y})$ is zero, we conclude from the positive-definiteness of the inner product, that:

    $f(\alpha\mathbf{x}+\beta\mathbf{y}) - \alpha f(\mathbf{x}) - \beta f(\mathbf{y}) = \mathbf{0}$, that is:

    $f(\alpha\mathbf{x}+\beta\mathbf{y}) = \alpha f(\mathbf{x}) + \beta f(\mathbf{y})$, in other words: $f$ is linear.
    Thanks from Educated
    Follow Math Help Forum on Facebook and Google+

  6. #6
    Senior Member Educated's Avatar
    Joined
    Aug 2010
    From
    New Zealand
    Posts
    433
    Thanks
    12

    Re: Preserving the inner product

    Yea that's what I meant when I said they cancel out.
    I'm used to using the 'dot product' instead of the inner product (but I guess they're the same thing). They never taught me the inner product (at university), but they expect me to just know it, or they just said it's like the dot product.

    I expanded everything which gave me a long list of terms which just cancelled out,
    ie. for the first term:
    f( \alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot f( \alpha \mathbf{\em x} + \beta \mathbf{\em y}) = (\alpha \mathbf{\em x} + \beta \mathbf{\em y}) \cdot ( \alpha \mathbf{\em x} + \beta \mathbf{\em y}) = \alpha^2 \mathbf{\em x} \cdot \mathbf{\em x} + 2 \alpha \beta \mathbf{\em x} \cdot \mathbf{\em y} + \beta^2 \mathbf{\em y} \cdot \mathbf{\em x}
    and so on with all the terms.

    I think your way is better though, I should stop using this dot product since the question uses inner product. Thanks for the great explanation!
    Follow Math Help Forum on Facebook and Google+

  7. #7
    MHF Contributor

    Joined
    Mar 2011
    From
    Tejas
    Posts
    3,401
    Thanks
    762

    Re: Preserving the inner product

    The "dot product" and an "inner product" are not quite the same thing. "Inner product" is the more general concept, and is characterized by these properties:

    Given a vector space $V$ over a field $F$, where $F$ is a quadratically closed subfield of the complex numbers (typically either $\Bbb R$ or $\Bbb C$) we say a function:

    $B:V \times V \to F$ is an inner product if:

    1. $B(x,y) = \overline{B(y,x)}$ This is called conjugate-symmetry, and if $F \subseteq \Bbb R$, just symmetry, as then $\overline{z} = z$.

    2. $B(-,y): V \to F$ for every $y \in V$ is linear (some presentations change this to linearity in the second argument).

    Such a function is said to be sesquilinear, which in the real case ($F \subseteq \Bbb R$), means "bilinear".

    3. $B(x,x) > 0$ and $B(x,x) = 0 \iff x = 0$ (this is called "positive-definiteness", and requires that $B(x,x)$ take its values in an ordered subfield of $F$).

    The "dot-product" or "Euclidean inner product" is a particular KIND of inner product (one of many possible) on $\Bbb R^n$, given by:

    $\displaystyle (x_1,x_2,\dots,x_n)\cdot (y_1,y_2,\dots,y_n) = \sum_{i = 1}^n x_iy_i$

    It is relatively easy to check that the dot product satisfies conditions 1-3.

    In the real case, it is common to associate a bilinear form on $\Bbb R^n$ with an $n \times n$ matrix $A$, in which case the associated bilinear function is:

    $B(x,y) = x^TAy$

    We say the bilinear form $B$ is symmetric if the matrix $A$ is, and likewise with positive-definite. The "usual dot product" is thus the bilinear form associated with the identity matrix.

    Similar considerations hold in the complex case, but we talk of a sesquilinear form, and use the complex-conjugate transpose (or hermetian) instead of the transpose.
    Follow Math Help Forum on Facebook and Google+

Similar Math Help Forum Discussions

  1. Orientation-preserving isometry of R^n
    Posted in the Advanced Algebra Forum
    Replies: 0
    Last Post: November 10th 2013, 07:59 AM
  2. Proof that unitary operator is inner product preserving.
    Posted in the Advanced Algebra Forum
    Replies: 5
    Last Post: December 21st 2012, 06:14 AM
  3. Preserving inclusions, and unions.
    Posted in the Differential Geometry Forum
    Replies: 1
    Last Post: September 15th 2010, 12:04 PM
  4. Preserving Volume
    Posted in the Advanced Algebra Forum
    Replies: 1
    Last Post: April 22nd 2010, 12:08 AM
  5. Preserving structural properties
    Posted in the Advanced Algebra Forum
    Replies: 1
    Last Post: October 22nd 2009, 05:36 PM

Search Tags


/mathhelpforum @mathhelpforum