MHB Lim of Convolution: Fourier Analysis Solution

  • Thread starter Thread starter Sonifa
  • Start date Start date
  • Tags Tags
    Convolution Limit
Sonifa
Messages
4
Reaction score
0
Let f(x)=1+cos 2\pix and let fk=f*...*f (k-times convolution)

what is the value of lim fk(1/2) when k tends to infinity

Should use something about the Fourier Analysis, Could someone help me how to solve this problem?
 
Physics news on Phys.org
Sonifa said:
Let f(x)=1+cos 2\pix and let fk=f*...*f (k-times convolution)

what is the value of lim fk(1/2) when k tends to infinity

Should use something about the Fourier Analysis, Could someone help me how to solve this problem?
Hi Sonifa, and welcome to MHB!

If you want to get some replies to your post, I think you need to explain the problem a bit more clearly. In particular,
(1) the definition of $f(x)$ is unclear. Is it meant to be $f(x) = 1 + \cos\bigl(\frac2\pi x\bigr)$?
(2) what is the domain of the function? It could be the unit interval, the whole real line, or something in between. That will affect the definition of convolution, because $f*f$ is defined as $$(f*f)(x) = \int f(y)f(x-y)\,dy,$$ where the integral is taken over the domain of $f$.
 
Opalg said:
Hi Sonifa, and welcome to MHB!

If you want to get some replies to your post, I think you need to explain the problem a bit more clearly. In particular,
(1) the definition of $f(x)$ is unclear. Is it meant to be $f(x) = 1 + \cos\bigl(\frac2\pi x\bigr)$?
(2) what is the domain of the function? It could be the unit interval, the whole real line, or something in between. That will affect the definition of convolution, because $f*f$ is defined as $$(f*f)(x) = \int f(y)f(x-y)\,dy,$$ where the integral is taken over the domain of $f$.

The function should be $f(x) = 1 + \cos\bigl(2\pi x\bigr)$ and it is defined on the unict circle R/Z
 
If $f(x) = 1 + \cos(2\pi x) = 1 + \frac12e^{2\pi ix} + \frac12e^{-2\pi ix}$ then the (complex) Fourier coefficients are given by $$\hat{f}(n) = \begin{cases}\frac12&(n = \pm1), \\ 1&(n=0, \\ 0&(\text{otherwise}). \end{cases}$$ The Fourier transform converts convolution products to pointwise products, so it follows that $$\hat{f_k}(n) = \begin{cases}2^{-k}&(n = \pm1), \\ 1&(n=0, \\ 0&(\text{otherwise}). \end{cases}$$ Therefore $f_k(x) = 1 + 2^{-k}\bigl(e^{2\pi ix} + e^{-2\pi ix}\bigr) = 1 + 2^{1-k}\!\cos(2\pi x).$ From that, you should easily be able to find $$\lim_{k\to\infty}f_k(1/2).$$
 
Opalg said:
If $f(x) = 1 + \cos(2\pi x) = 1 + \frac12e^{2\pi ix} + \frac12e^{-2\pi ix}$ then the (complex) Fourier coefficients are given by $$\hat{f}(n) = \begin{cases}\frac12&(n = \pm1), \\ 1&(n=0, \\ 0&(\text{otherwise}). \end{cases}$$ The Fourier transform converts convolution products to pointwise products, so it follows that $$\hat{f_k}(n) = \begin{cases}2^{-k}&(n = \pm1), \\ 1&(n=0, \\ 0&(\text{otherwise}). \end{cases}$$ Therefore $f_k(x) = 1 + 2^{-k}\bigl(e^{2\pi ix} + e^{-2\pi ix}\bigr) = 1 + 2^{1-k}\!\cos(2\pi x).$ From that, you should easily be able to find $$\lim_{k\to\infty}f_k(1/2).$$

Finally, I got the same solution as yours. But still many thanks!
 
A sphere as topological manifold can be defined by gluing together the boundary of two disk. Basically one starts assigning each disk the subspace topology from ##\mathbb R^2## and then taking the quotient topology obtained by gluing their boundaries. Starting from the above definition of 2-sphere as topological manifold, shows that it is homeomorphic to the "embedded" sphere understood as subset of ##\mathbb R^3## in the subspace topology.
Back
Top