The Green's function in rectangular coordinates was given by
Eq. (13.4) as
However, problems are often encountered when using this expression because the space variables are all enclosed in a square-root sign and therefore cannot be separated. This makes finding analytical solutions very difficult, and one often has to resort to using moveable-origin coordinate systems. This limits its use to numerical integration in the Kirchhoff–Helmholtz surface integral. Unfortunately, the Green's function of
Eq. (13.31) is not particularly amenable to numerical integration because it is singular at the origin and leads to oscillatory integrands at high frequencies. The dipole integral is even more problematic because the Green's function normal gradient has a 1/
R
2 term, which leads to diverging numerical and analytical integrals. Furthermore, where the surface of integration encloses one or more sources, we can only calculate the field on the side of the surface where there are no sources and the waves are diverging. In other words we cannot solve the
reverse problem and calculate the field in which there are sources and the waves converge toward them. A more powerful formula is given by
where
k
x
,
k
y
, and
k
z
represent the spatial frequency components in the
x,
y, and
z directions respectively of a plane wave of spatial frequency
k traveling in an arbitrary direction. For example, if the direction of travel subtends an angle
θ with the
z-axis, then the trace velocity seen along the
z-axis is
c/cos
θ and the wave number is
k
z
=
k cos
θ.
Hence the wavelength will appear to be longer along the
z-axis. To gain a better understanding of
Eq. (13.33) we may compare it with
Eq. (7.113) for the pressure field inside an enclosure by letting
k
x
=
mπ/
l
x
,
k
y
=
nπ/
l
y
, and
k
z
=
k
mn
. We also replace the infinite integrals with summations. In other words,
Eq. (13.33) may be thought of as the spatial distribution of an infinite enclosure in which traveling plane waves of any wavelength may exist as opposed to standing ones of particular wavelengths that correspond to the dimensions of the finite enclosure. The fact that a point source can be represented as integral over all spatial frequencies is not so surprising when we consider that an infinite impulse contains all frequencies.
It may seem counterintuitive to introduce two extra integrals, but the troublesome 1/
R term has vanished along with the square-root sign in the exponent. When used in the Kirchhoff–Helmholtz integrals, we will show in
Section 13.19 that this integral form of the Green's function is an inverse Fourier transform. This leads to an important theorem that forms the basis of near-field acoustical holography in which the dipole Kirchhoff–Helmholtz integral evaluated over one plane is the Fourier transform of the pressure distribution in that plane. The sound field spectra is then propagated in
k-space to another parallel plane in which the Green's function is the inverse Fourier transform that gives the pressure in that plane. Furthermore, we can solve the so-called reverse problem where there are one or more sources in the field of interest. This method of calculation is particularly amenable to the digital processing of sound fields captured by planar microphone arrays to calculate the entire sound field of interest. In other words, if there are sources on one side of the array, we can plot the pressure field on
both sides of the array. This is not possible using the Euclidean form of the Green's function of
Eq. (13.31).