is there something wrong with the Fourier transform coefficients?

73 Views Asked by At

define the Fourier transform and its inverse $$\hat{u}(y)=\frac{1}{(2\pi)^{n/2}}\int_{\mathbb R^n}e^{-ixy}u(x)\,dx$$ $$\check{u}(y)=\frac{1}{(2\pi)^{n/2}}\int_{\mathbb R^n}e^{ixy}u(x)\,dx$$

then there are two relations:

$$\tag{1}(u*v)^{\hat{}}=(2\pi)^{-n/2}\hat{u}\hat{v}$$ $$\tag{2}(uv)^{\hat{}}=\hat{u}*\hat{v}$$

where $*$ means convolution.

but I think there is a contradiction!

apply (1) to $\check{u},\check{v}$: $$(\check{u}*\check{v})^{\hat{}}=(2\pi)^{-n/2}uv$$

hence $$(\check{u}*\check{v})=(2\pi)^{-n/2}(uv)^{\check{}}$$

since $\hat{u}(x)=\check{u}(-x)$

we get : $$(\hat{u}*\hat{v})=(2\pi)^{-n/2}(uv)^{\hat{}}$$

but in $(2)$, the coefficient disappeared. what's the problem?


oh, I found the secret.. it should be $$\tag{1}(u*v)^{\hat{}}=(2\pi)^{n/2}\hat{u}\hat{v}$$ $$\tag{2}(uv)^{\hat{}}=(2\pi)^{-n/2}\hat{u}*\hat{v}$$