Proving Inverse Matrices: Simplifying the Last Step

5 replies · 2K views · last reply
swuster
Messages
40
Reaction score
0

Homework Statement


I'm trying to show that a given matrix is the inverse of the other, by showing that multiplying them together generates the identity matrix. I can't see a way to simplify the last step and I feel like I'm missing something..? Any input on this would be helpful. Thanks!

Homework Equations


see below
j = sqrt(-1)

The Attempt at a Solution


Handwritten-style typeset derivation proving P inverse is the inverse of P: a sum over i from 0 to M-1 of P times P inverse equals the identity matrix, with P_ki = exp(-j(2π/M)ki) and P inverse = (1/M)exp(j(2π/M)iλ), leading to a sum of (1/M)e^{j2π(λ-κ)i/M}, equal to 1 when λ = κ.

How do I prove that the rest of the terms are 0?
 
Attachments
  • Handwritten-style typeset derivation proving P inverse is the inverse of P: sum over i from 0 to M-1 of B times A gives C, then the identity I_M as a sum of P times P inverse, with P elements exp(-j(2pi/M)ki) and inverse elements (1/M)exp(j(2pi/M)il), leading to the diagonal case where the sum of 1/M equals 1.
    ece2200.jpg
    26.4 KB · Views: 497
It looks to me like the last sentence, which you cut off at the comma, might give you a hint. What does it say?
 
It's not cut off. It's as far as I have gotten with the proof, haha. There is nothing obvious that jumps out at me when I consider the case where the exponential argument isn't 0.
 
swuster said:
Thanks so much! That did the trick!

Handwritten derivation: for lambda not equal to kappa, the geometric series sum of z^m equals (1 - z^M)/(1 - z) is applied, giving C_kappa,lambda = 0 with e^(2 pi (lambda - kappa) i / M) terms.

Looks good, nice job.

It's also insightful to consider what is going on geometrically.

This sum:

[tex]\sum_{i=0}^{M-1}\frac{1}{M}e^{j\left(\frac{2\pi(\lambda-\kappa)i}{M}\right)[/tex]

is calculating the average of [itex]M[/itex] complex numbers. Furthermore, since [itex]\lambda - \kappa[/itex] is a nonzero integer, these complex numbers are evenly spaced samples around the unit circle, so their average is zero.