Proving the polarization formula for complex inner products

3 replies · 3K views · last reply
Math Amateur
Gold Member
MHB
Messages
3,920
Reaction score
48
I am reading D. J. H. Garling's book: "A Course in Mathematical Analysis: Volume II: Metric and Topological Spaces, Functions of a Vector Variable" ... ...

I am focused on Chapter 11: Metric Spaces and Normed Spaces ... ...

I need some help with the polarization formula for the complex case ...

Garling's statement of the polarization formulae reads as follows:
Text page with the polarization formulae: in the real case <x,y> = (1/2)(||x+y||^2 - ||x||^2 - ||y||^2) = (1/4)(||x+y||^2 - ||x-y||^2); in the complex case <x,y> = (1/4) times the sum over j from 0 to 3 of i^j ||x + i^j y||^2.

In the above text from Garling we read the following:" ... ... in the complex case we have the polarization formula## \langle x,y \rangle = \frac{1}{4} \left( \sum_{ j = 0 }^3 i^j \| x + i^j y \|^2 \right) ## ... ... "
Can someone please demonstrate how to prove that ##\langle x,y \rangle = \frac{1}{4} \left( \sum_{ j = 0 }^3 i^j \| x + i^j y \|^2 \right)## ...?Help will be appreciated ...

Peter
==========================================================================================***NOTE***

It may help readers of the above post to know Garling's notation and approach to inner-product spaces ... ... so I am providing the same ... as follows:
Page from a textbook, section 11.3 Inner-product spaces, with the definition of an inner product on a real vector space and the bilinearity conditions written out as equations.

Printed textbook page defining symmetric bilinear forms, inner products on real and complex vector spaces, sesquilinearity, the Hermitian condition and positive definiteness, with the complex polarization formula.

Printed textbook page stating and proving the Cauchy-Schwarz inequality for inner product spaces, with its proof, equality condition, and Corollary 11.3.2 (triangle inequality).

Typeset proof page with equations: parallelogram law and polarization formulae for real and complex inner products, including the sum over j from 0 to 3 of i^j times the norm of x + i^j y squared.

Hope that helps ...

Peter
 
Attachments
  • Text page with the polarization formulae: in the real case <x,y> = (1/2)(||x+y||^2 - ||x||^2 - ||y||^2) = (1/4)(||x+y||^2 - ||x-y||^2); in the complex case <x,y> = (1/4) times the sum over j from 0 to 3 of i^j ||x + i^j y||^2.
    Garling - Polarization Formulae for Inner Product Spaces ,,, ,,,.png
    15.5 KB · Views: 1,503
  • Page from a textbook, section 11.3 Inner-product spaces, with the definition of an inner product on a real vector space and the bilinearity conditions written out as equations.
    Garling - 1 - Start of Section on Inner-Product Spaces ... PART 1 ... .png
    29.1 KB · Views: 782
  • Printed textbook page defining symmetric bilinear forms, inner products on real and complex vector spaces, sesquilinearity, the Hermitian condition and positive definiteness, with the complex polarization formula.
    Garling - 2 - Start of Section on Inner-Product Spaces ... PART 2 ... .png
    25.3 KB · Views: 731
  • Printed textbook page stating and proving the Cauchy-Schwarz inequality for inner product spaces, with its proof, equality condition, and Corollary 11.3.2 (triangle inequality).
    Garling - 3 - Start of Section on Inner-Product Spaces ... PART 3 ... .png
    23.3 KB · Views: 697
  • Typeset proof page with equations: parallelogram law and polarization formulae for real and complex inner products, including the sum over j from 0 to 3 of i^j times the norm of x + i^j y squared.
    Garling - 4 - Start of Section on Inner-Product Spaces ... PART 4 ... .png
    23.8 KB · Views: 1,061
We don't need any of the attached support material. The result follows directly from the axioms of a complex inner product:

\begin{align*}
\sum_{j=0}^3 i^j\|x+i^j y\|^2
% 1
&= \sum_{j=0}^3 i^j \langle x+i^j y ,x+i^j y\rangle
\\ % 2
&= \sum_{j=0}^3 i^j
\left(
\langle x ,x\rangle +
i^j \langle y,x\rangle +
\left( i^j\right)^*\langle x , y \rangle +
i^j \left( i^j\right)^* \langle y , y\rangle
\right)
\\ % 3
&= \sum_{j=0}^3 i^j
\left(
\langle x ,x\rangle +
i^j \langle y,x\rangle +
\left(i^j \langle y,x\rangle \right)^* +
\left( i\ i^*\right)^j \langle y , y\rangle
\right)
\\ % 4
&= \left(\langle x,x\rangle + \langle y,y\rangle\right)
\sum_{j=0}^3 i^j
+
\sum_{j=0}^3 i^j\cdot
2\ \Re \left( i^j \langle y,x\rangle\right)
\\ % 5
&=0
+
2
\left(
\Re \langle y,x\rangle +
i\cdot \Re \left( i \langle y,x\rangle\right)
- \Re \left( - \langle y,x\rangle\right)
-i\cdot \Re \left(-i \langle y,x\rangle \right)
\right)
\\ % 6
&=
2
\left(
\Re \langle y,x\rangle
- i\cdot \Im \langle y,x\rangle
+\Re \langle y,x\rangle
-i\cdot \Im \langle y,x\rangle
\right)
\\ % 7
&=
2
\left(
\langle y,x\rangle^*
+
\langle y,x\rangle^*
\right)
\\ % 8
&=
4
\langle x,y\rangle
\end{align*}
 
  • Like
Likes   Reactions: Keith_McClary and Math Amateur
A lot of times greater generality isn't necessarily helpful but because ##m## evenly spaced points on the unit circle comes up so often, it seems worth pointing out that your complex polarization identity works for any natural number ##m \geq 3##, not just the ##m = 4## case that the book stated.

- - - -
where we recall that ##\sum_{k=0}^{m-1} \omega^k = 0##, and again with the restriction that natural number ##m \geq 3## we recall ##\sum_{k=0}^{m-1} \omega^{2k} = 0##
(why?)
- - - -
start by expanding the squared 2 norm
##\langle \mathbf x, \mathbf y \rangle = \frac{1}{m} \sum_{k=0}^{m-1} \big \Vert \mathbf x + \omega^k \mathbf y \big \Vert_2^2 \omega^k = \frac{1}{m} \sum_{k=0}^{m-1} \Big(\big \Vert \mathbf x \big \Vert_2^2 \omega^k + \big \Vert \mathbf y \big \Vert_2^2 \omega^k + \langle \mathbf x, \omega^k \mathbf y \rangle \omega^k + \langle \omega^k \mathbf y, \mathbf x \rangle \omega^k \Big)##

split the summation, and make use of conjugate linearity by removing the ##\omega^k## from the ##\mathbf y## terms of the inner product

##\langle \mathbf x, \mathbf y \rangle = \frac{1}{m} \sum_{k=0}^{m-1} \Big(\big(\big \Vert \mathbf x \big \Vert_2^2 + \big \Vert \mathbf y \big \Vert_2^2\big)\omega^k \Big) + \frac{1}{m} \sum_{k=0}^{m-1} \Big(\langle \mathbf x, \mathbf y \rangle \bar{\omega}^k\omega^k\Big) + \frac{1}{m} \sum_{k=0}^{m-1} \Big(\langle \mathbf y, \mathbf x \rangle \omega^{k} \omega^k\Big)##

finally:

##\langle \mathbf x, \mathbf y \rangle = \Big(\big( \big \Vert \mathbf x \big \Vert_2^2 + \big \Vert \mathbf y \big \Vert_2^2 \big) \frac{1}{m} \big(\sum_{k=0}^{m-1} \omega^k\big) \Big) + \langle \mathbf x, \mathbf y \rangle \Big(\frac{1}{m} \sum_{k=0}^{m-1} 1 \big)+ \langle \mathbf y, \mathbf x \rangle\Big(\frac{1}{m} \sum_{k=0}^{m-1} \omega^{2k} \Big) = \big(0\big) + \langle \mathbf x, \mathbf y \rangle \big(1\big)+ \big(0\big) = \langle \mathbf x, \mathbf y \rangle ##
 
Last edited:
  • Like
Likes   Reactions: Math Amateur
THanks Andrew, StoneTemplePython ...

Appreciate the help and insights ...

Just working through your posts now ...

Thanks again ...

Peter