Use of series and integral (Fourier or not) to write solution of three dimensional wave equation in different coordinates systems

58 Views Asked by At

Consider wave equation in three dimensions $$\square f=0$$ Consider first cartesian coordinates. With separation of variables one can find an infinite number of indipendent solutions

$$f_{\omega,\alpha,\beta,\gamma}(x,y,z,t)=f^{0}_{\omega,\alpha,\beta,\gamma}e^{i(\alpha x+\beta y+ \gamma z-\omega t)}\tag{1}$$

To get the general solution it is necessary to write the linear combination

$$f(x,y,z,t)=\sum_{\omega,\alpha,\beta,\gamma}f^{0}_{\omega,\alpha,\beta,\gamma}e^{i(\alpha x+\beta y+ \gamma z-\omega t)}\tag{2}$$

On textbook is it claimed that $(2)$ is a Fourier series, as long as $f$ is periodic. It is not explicit but I suppose that values of $\omega,\alpha,\beta,\gamma$ considered in $(2)$ must be a multiple of a foundamental value $\omega_0,\alpha_0,\beta_0,\gamma_0$. Is that correct?


If $f$ is not periodic, then it is necessary to use an integral. What is the form of the integral rapresentation of $f$ in this case and what is its Fourier trasform? My guess is:

$$f(x,y,z,t)=\frac{1}{(2\pi)^{\frac{4}{2}}}\int_{-\infty}^{\infty} \int_{-\infty}^{\infty}\int_{-\infty}^{\infty}\int_{-\infty}^{\infty}F(\omega,\alpha,\beta,\gamma)e^{i(\alpha x+\beta y+ \gamma z-\omega t)} d\omega \,\,d\alpha \,\, d\beta \,\,d \gamma$$

$$F(\omega,\alpha,\beta,\gamma)=\frac{1}{(2\pi)^{\frac{4}{2}}}\int_{-\infty}^{\infty} \int_{-\infty}^{\infty}\int_{-\infty}^{\infty}\int_{-\infty}^{\infty}f(x,y,z,t)e^{-i(\alpha x+\beta y+ \gamma z-\omega t)} dx\,\,dy \,\, dz \,\,d t$$

Are these the correct expressions?



If I go in cylindrical coordinates for istance, the solution is

$$f(r,\phi , z,t) =\sum_{\omega,n,h} R^{0}_{\omega, n, h } H_n\Bigg(r \sqrt{\frac{\omega^2}{c^2}-h^2}\Bigg) e^{i(n\phi +hz-\omega t)} \tag{3}$$

Is $(3)$ still a Fourier series? There is an Hankel function $H$ that contains $r$ but also $\omega$ and $h$ (and its order is $n$), so it should not be a Fourier series but then what "kind" of series is $(3)$?

In particular supposing $f(r,\phi , z,t) =R(r) \Phi(\phi) Z(z) \chi (t)$ can I factorize $(2)$ as a Fourier series for $\Phi(\phi) Z(z) \chi (t)$ and a Fourier-Bessel series for $R(r)$?

Finally if $f$ is not periodic and I'm using cylindrical coordinates, is $(3)$ subsituted by something else (on textbook there are no explicit assumption for $(3)$, i.e. periodicity)?

(I have similar doubts for solution in spherical coordinates)