It is easy to notice that
k
n
(
|
t
|
, y
)
and
l
n
(
|
t
|
, y
)
are infinitely differentiable
and rapidly decreasing functions
t
∈
R
n
, i.e. they belong to the space
S
(
R
n
)
[7]. These functions are also spherically symmetric, so while
calculating the inverse Fourier transform it is possible to pass to the
spherical coordinates in the space
R
n
. Let us denote
|
x
|
=
r
,
|
t
|
=
ρ, σ
n
−
1
as an area of the unit sphere in the space
R
n
.
For any spherically symmetric
function
h
n
(
|
t
|
) =
h
n
(
ρ
)
∈ S
(
R
n
)
we have
H
n
(
|
x
|
) =
H
n
(
r
) =
F
−
1
t
[
h
n
] (
x
) =
1
(2
π
)
n
Z
R
n
h
n
(
|
t
|
)
e
−
i
h
x,t
i
dt
=
=
σ
n
−
1
(2
π
)
n
∞
Z
0
h
n
(
ρ
)
ρ
n
2
dρ
π
Z
0
e
−
irρ
cos
θ
sin
n
−
2
θdθ
=
=
1
(2
π
)
n
2
r
n
2
−
1
∞
Z
0
h
n
(
ρ
)
ρ
n
2
J
n
2
−
1
(
rρ
)
dρ,
where
J
n
2
−
1
(
rρ
)
is the Bessel function of the first kind of order
ν
=
n/
2
−
1
[7, 8]. The above formula is valid and it is easy to be checked.
By differentiating the above equation and by considering the formula
of the theory of Bessel functions [4]
νJ
ν
(
rρ
)
rρ
−
J
0
ν
(
rρ
) =
J
ν
+1
(
rρ
)
,
we obtain the recurrence formula
H
n
+2
(
r
) =
−
1
2
πr
∂
∂r
H
n
(
r
)
.
For this formula it is necessary to know the kernels
K
n
(
x, y
)
and
L
n
(
x, y
)
only for
n
= 1
and
n
= 2
.
Since the Fourier transform moves the space
S
(
R
n
)
into itself,
the kernels
K
n
(
x, y
) =
K
∗
n
(
|
x
|
, y
)
and
L
n
(
x, y
) =
L
∗
n
(
|
x
|
, y
)
when
∀
y
∈
(0
, a
)
, are spherically symmetric functions of
x
in the space
S
(
R
n
)
.
Therefore, the convolution (1) exists for any tempered distribution
ϕ
(
x
)
∈
∈ S
0
(
R
n
)
,
ψ
(
x
)
∈ S
0
(
R
n
)
, and can be written as
ϕ
(
x
)
∗
K
n
(
x, y
) +
ψ
(
x
)
∗
L
n
(
x, y
) =
= (
ϕ
(
t
)
, K
n
(
x
−
t, y
)) + (
ψ
(
t
)
, L
n
(
x
−
t, y
))
.
When
ϕ
(
x
)
and
ψ
(
x
)
are normal functions of polynomial growth,
convolution (1) can be written as a sum of integrals
Z
R
n
ϕ
(
t
)
K
n
(
x
−
t, y
)
dt
+
Z
R
n
ψ
(
t
)
L
n
(
x
−
t, y
)
dt .
6
ISSN 1812-3368. Herald of the BMSTU. Series “Natural Sciences”. 2015. No. 1