the evaluation of some definite integrals involving bessel ... · pressed in terms of complete...
TRANSCRIPT
The Evaluation of Some Definite IntegralsInvolving Bessel Functions Which Occur
in Hydrodynamics and Elasticity
By A. H. Van Tuyl
1. Introduction. Integrals of the form
(1.1) h(m.,v)= f e~xtJv(yt)JÁbt)tm dtJo
and
(1.2) h(m,v)= [ e~xtJ„-i(yt)Jv(bt)tm+idtJo
occur in various problems of potential theory, hydrodynamics, and elasticity (A.
Weinstein [1, 2], H. Bateman [3, p. 417], and A. G. Webster [4, p. 367-375]). In [1]-[4], v is an integer, x, y, and b are real, and m is a positive or negative integer, or
zero, such that (1.1) and (1.2) converge. When v is an integer, special cases of (1.1)
and (1.2) have often been expressed in the literature in terms of elliptic integrals,
the modulus of which is an elementary algebraic function of x, y, and b. The object
of the present paper is to obtain closed expressions for h(m, v) in terms of known
functions, assuming that v is complex when m ^ 0, and an integer when m < 0,
and to find recurrence relations involving I\(m, v) and /2(to, v). It is assumed
throughout that x, y, and b are complex. When m ^ 0, I\(m, v) is evaluated in
closed form in terms of associated Legendre functions of the second kind of the form
QTv-w(z); when to < 0 and v is an integer, it is expressed in terms of associated
Legendre functions of the form P~^!m(z), and Appell's generalized hypergeometric
function F i. By means of either the recurrence relations in Section 4, or certain
differentiation formulas, explicit expressions of the same form as in Sections 2 and 3
can also be obtained for 72(to, v). When v is an integer, these explicit expressions
for h(m, v) and h(m, v) can be obtained in terms of standard elliptic integrals by
means of well-known reduction formulas.
The more general integral
(1.3) I(ß, *;X)= f e'ctJß(at)JXbt)tx dt•>o
has been treated by G. Eason, B. Noble, and I. N. Sneddon [5] when p, v, and X are
integers such that (1.3) converges, and a, b, and c are real. A summary of [5] is given
by Y. L. Luke [6, p. 314]. The results of [5] can be used for the calculation of (1.3)
in terms of elliptic integrals for all integers ß, v, and X, X ^ 1, for which it converges.
However, explicit expressions are given in [5] only when (1.3) is of the form of
certain special cases of (1.1) and (1.2). The calculation of (1.3) for other values of
/x, v, and X can be carried out by use of recurrence relations obtained in [5]. The three
recurrence relations obtained in Section 4 of the present paper are seen not to be
special cases of the recurrence relations in [5].
Received August 22, 1963. Revised January 31, 1964.
421
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
422 A. H. VAN TUYL
When v is an integer, and either m S: 0 or b = y, (1.1) and (1.2) can be ex-
pressed in terms of complete elliptic integrals of the first and second kinds. When
m < 0 and b = y, complete elliptic integrals of the third kind occur also. Examples
of the latter can be obtained from results of several authors (W. M. Hicks [7, p. 628],
G. M. Minchin [8, p. 354], A. Van Tuyl [9], M. R. Shura-Bura [10], M. A. Sadowskyand E. Stemberg [11], G. E. Pringle [12, p. 385 and 392], and H. E. Fettis [13]).In [7], [8], and [11], integrals of the form of (1.1) or (1.2) are not considered ex-
plicitly. A quantity known to be proportional to 72( — 1, 1) is evaluated in terms of
elliptic integrals in [7] and [8], and quantities proportional to 7i ( — 1, 1 ), 72 ( —T, 1 ),
and 72(—2, 1) are evaluated in [11]. The evaluation of 72( —1, 1) and 72(—2, 1)
by numerical integration has been discussed by G. P. Weeg [14].
For convergence of (1.1) and (1.2) at the lower limit, it is necessary to have
Re (2v + to + 1) > 0, and for convergence at the upper limit, all four quantities
Re (x ± ib ± iy) must be either positive or zero. In the latter case, we must also
have either to < 1 or to < 0, depending on the values of x, b, and y. When
Re (2v + m + 1) > 0 and Re (x ± ib ± iy) > 0, it is easily shown that (1.1) and
(1.2) are analytic functions of each of the variables x, y, and b, and that differen-
tiations of (1.1) and (1.2) with respect to x, y, and b can be carried out under the
integral sign.
Defining
(1.4) 4>(x, y) = y'-'him - 1, v),
(1.5) >p(x, y) = -yh(rn, v),
it is seen that <t>(x, y) and \p(x, y) satisfy the differential equations
(1.6) y2-ld^=f,dx dy
(1.7) y^f=-f.dy ox
Despite the fact that v may be complex, it is convenient, following A. Weinstein
[15], to call <t>(x, y) a generalized axially symmetric potential in 2v -4- 1 dimensions,
and \p(x, y) its associated stream function. The analogy with the case when v is an
integer has proven fruitful. Similarly, writing
(1.8) 4>(x, b) = b-yh(m, v),
(1.9) Hx, b) = -b'+lh(m,v + 1),
we have
(1.10) ^+id* = djP
dx dy
(1.11) ^+'^=_^.dy dx
In [l]-f4], integrals 7i(m, n) and 72(to, n) arise from the solution of special cases of
these equations.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
SOME DEFINITE INTEGRALS INVOLVING BESSEL FUNCTIONS 423
We note that 72(»i, v) can be calculated in terms of 7i(m, v) by means of the
relations
(1.12) him, v) = y-'^- [yh(m, v)]dy
(1.13) = -b-'-lb'-'him^- 1)].
Thus, explicit expressions for 72(to, v), to ^ 0, and 72(to, n), m < 0, can be ob-
tained from the results of Sections 2 and 3 respectively. Alternatively, we can calcu-
late 72(to, v) in terms of 7i(to, v) by means of one of the recurrence relations in
Section 4.
2. Evaluation of 7i(to, v) for to ^ 0. From Watson [16, p. 389], equation (1), we
have
. / x (byyr(2v + to + l)him' v) = *x^^T(2v + 1)
f 2v + to + 1 2, + to + 2 p2\ . 2r
where F (a, b; c; z) is the hypergeometric function, and p2 = b2 + y2 — 2by cos <£.
Equation (2.1) holds for the same values of the parameters for which (1.1) and
(1.2) converge. From [6, p. 75], it follows that it is sufficient to assume that
| arg b | ^ 7r/2 and | arg y | g t/2. It is assumed in (2.1) that | arg (1 + p2/x2) \ <
■K, and that arg (sin <t>) = 0.
Setting b = c + n+ein [17, p. 109], equation (3), where n ^ 0 is an integer,
and where c and a — c — n are not negative integers, and letting e —» 0, we obtain
EV , \ r(c)r(c - a + n) , s_aF(a,c + n;c;z)= _ (I - z)
(2.2)
F ia, —n; a — c — n + 1;-)
when | arg (1 — z) \ < x. Hence, noting that the preceding conditions are satisfied
in each case, we have
„ (2v + to + 1 2v + to + 2*A-2-'-2-;" + 1;2
r9oN r(y + pr(j) n x-,-r-i,2(2-3) - r(„ + r + i)r(| ^7) (1 - z)
(2.4)Y(v + r + l)r(-i - r)
■F (v + r + |, -r; §; j-i-A to = 2r,
(i -^rr(*+ l)r(-§) „ ^-r-r-I/1
•FU + r + i'~r;I;r^>
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
424 A. H. VAN TUYL
r = 0,1, • • • , | arg (1 — z) | < r. From the preceding, using Legendre's duplica-
tion formula for T(z), we obtain
(-Dr2—lj2r(r + |)
, N l{ 'V) ~ T(byY^Y(v + h)
(2.5) , . . ,.^ . y M T(v + r + \ + ¿) i!\., , . .
' 5(_1) W-r(¥TÏ)-\2by) /(' + «> "}
and
(-l)'2r-'-1/2r(r + |)a;ii(,z/- -r x, j// =
(2.6)
/i(2r + 1, „) -
5 (_1) U-ñíTO-{¿by) I{,+1+ h V)>
r = 0, 1, • • • , where
sin2" </> d<t>(2.7) 7(to, ») = f
with
(2.8) /
(ß - cos4>)"<+''+1'2'
x + b2 + i/2
2by
We see that (2.7) is an analytic function of ß when the /3-plane is cut along the real
axis from 1 to — ». In order for (2.5) and (2.6) to be real when all parameters are
real and positive, we must then have | arg (ß — cos <£) | < x in (2.7). The cut
/3-plane is more extensive than the region in which (2.1) converges, hence, (2.5)
and (2.6) give an analytic continuation of (2.1).
Substituting cos0 = t in (2.7) and comparing with Hobson [18, p. 195], equation
(20), we obtain
\m 0M-l/2 Iap + |) , 2 Hs-mtfr.m(2.9) Km, v) = (-D"2'+1" xV^»'M0r - irml2QT-vAß)V(v + TO + %)
throughout the cut /3-plane, where Qa (z) is the associated Legendre function of the
second kind, and where [ arg (/3 — 1) | < x, | arg (ß + 1) | < x. We note that the
conditions arg (sin2 <¿>) = 0 and | arg (ß — cos 0) | < x which hold in (2.7) are
required in the preceding reference. Finally, from (2.5), (2.6), and (2.9), we have
<2J0) '■»■ '> " 2^m^ S 0) fâVî) (0 " - »"**•"«■<«and
t io„ _l i "\ 2 r(r + 1)War+ 1,0- - v(b y+w x
(2.11)
■emh)(Q,(ß2-ir^+1)K^^^^where ß satisfies the preceding conditions, and r = 0, 1, • • • .
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
SOME DEFINITE INTEGRALS INVOLVING BESSEL FUNCTIONS 425
The result
(by)" f" sin 0" d<f>(2.12) 7x(0, V)=^ML[
X Jo70 (z2 + b2 + y2 - 2by cos <t>Y+in
was obtained by A. Weinstein ([15], equations (9) and (12)), and the expression
(2.13) 7x(0, v) = ~^ Q,-i/2(/3)
was obtained by Watson ([16, p. 389], equation (2)).
When v is an integer n, n 3ï 0, we can express 7(to, v), and hence, Q7-w(ß), in
terms of complete elliptic integrals of the first and second kinds. Substituting
cos (¡> = 2sn u — 1 in (2.7) when v = n, where the modulus, k, of sn u is given by
(2.14) k2 = 2 áby
we obtain
1 + /3 x2 + (b + yy~'
„K in ïn jn-m+1/2, 2n+2m+l ¡ STl UCn U dU(2.15) 7(to, n) = 2n-m+ll2k2n+2m+l [
Jo dn2m+2nu
From (2.14), it follows that the cut in the ß-plane from 1 to — » corresponds to
cuts in the fc2-plane from 1 to =° and from 0 to — =° along the real axis. Concerning
the path of integration in (2.15), it is known that the residues of the integrand are
zero at all of its poles. Hence, it follows that (2.15) is independent of the path of
integration joining the points 0 and K. In particular, we can always take the latter
to be the straight line between 0 and K.
The right side of (2.15) can be evaluated in terms of elliptic integrals by means
of well-known reduction formulas (P. F. Byrd and M. D. Friedman [19, p. 191-
198]). Alternatively, we can compute Qn-i/2(/3) for to ^ 0 and n ^ 0 by use of the
recurrence relations for Q,(z) ([18, p. 290], equations (164) and (166)), starting
from the values of Q-m(ß), Qidß), Q-w(ß), and QU(ß). From (2.9) and (2.15),referring to [19], we have
(2.16) Q-i/»(0) - kK,
(2.17) Qm(ß) = I (K - E) - kK,
kk'E(2.18) Qlmiß) =
and
(2.19) QU(ß) = ¿-3 [2k'2K - (2 - k2)E],
where K and E are the complete elliptic integrals of the first and second kinds,
respectively, and k — 1 — k . Both k and k are single valued when the /c2-plane
is cut from 0 to — <x> and from 1 to ». Another method for the computation of
QZ-w(ß) in terms of elliptic integrals can be obtained from [7], Sections II and III.
Finally, we note that 7i(0, n) and 7i(l, n) satisfy simple recurrence relations.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
426 A. H. VAN TUYL
From (2.11), we have
(2.20) h(l, n) = —7^ (ß2 - irmQn-uAß).ir(by)m
From (2.13), (2.22), and (2.16), we obtain the recurrence relations
(2.21) (n + \)h(fi, n + 1) - 2nßh(0, n) + (n - %)h(0, n - 1) = 0
and
(2.22) (n - i)7x(l, n + 1) - 2n/tfi(l, n) + (n + J)/i(l, n - 1) - 0.
Expressions for 7i(0, 0), 7i(0, 1), 7i(l, 0), and 7i(l, 1) in terms of elliptic integrals
ollow from (2.13), (2.20), and equations (2.16) through (2.19).
3. Evaluation of 7x(to, n) for to < 0. From [17, p. 105 and 107], equations (1),
(10), (18), and (37), we have the relation
(-z)"~c(l - z)c-a-hF(l - b,c - b;a + 1 - b; l/z)
rtlv _ r(l - c)T(a+ 1 - b) w(n h „ ^ r(c)r(i - c)T(a + 1 - b)[aÄ) - T(l - b)T(a + 1 - c) Ha'b>c'z) r(2 - c)T(c - b)T(a)
■e^-'V-'d - z)c-a-"F(l - a, 1 - b; 2 - c; z)
when Im z > 0. Substituting a = r — n -\- \, b = r — n, and c = 1 — n + e,
n ^ r 2: 1, e > 0, and letting e —> 0, we obtain
Fir - r + -,n — r + \;n + 1; z
T(n + l)T(r - n+J) ^r;|;l/.)
, s r(r)r(|)(3.2)
, iyT(n + l)r(r -n + è)r(n)
r(w - r + l)T(r)r(r + §)
•2~"(1 - 2)2r~"~1/2Wr - n + -, r - n; 1 - n; zj ,
Im z > 0. The second hypergeometric function on the right side of (3.2) is of the
form F(a, —to; —to — I; z), where m è 0,1 ^ 0, and is defined as in [15, p. 101],
equation (3). We see that it remains a polynomial throughout the limiting process
e —> 0. Noting that both sides of (3.2) are analytic in the z-plane cut along the real
axis from 1 to «>, it follows by analytic continuation that (3.2) is valid for
| arg (1 — z) | < x. Similarly, setting a = r — n — \,b = r — n, and c = 1 —
n + «, n ¡à r è 1, « > 0, and letting e —> 0, we have
F ( n - r + 1, w + r + - ; n + 1; 2 j
_ r(n + l)r(r-n-j) / 1. \-f(or(ï)V ' r'2'VV
(3.3)
+ (-Dr r(w + i)r(r - n - è)r(n)
r(w - r + i)r(r)r(r - i)
-n/-, \2r-n-3/2j-, / 1 , \•2 (1—2) 7*M ?• — n — -, r — n; 1 — w; z 1
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
SOME DEFINITE INTEGRALS INVOLVING BESSEL FUNCTIONS 427
when I arg (1 — z) \ < x. Finally, from (3.2), (3.3), and (2.1), using Legendre's
duplication formula for T(z) together with the functional equations, we find that
T/ o -\ 2r?i! ,, v-ii 0„_2r ^ (n — r + i)[h( -2r,n) = (by) < -2 x ¿_
x(2ra)! "' \ i^>(n - I - i)l(2i+ 1)1
(3.4) • gj J(n - r + i + 1, ») + 22'—2 (-~^JT ^^
•E^^'-;/' ](p}K(2r-n,r +i_o (n — r — !)'.( .)l \byj
i,n)
and
(3.5)
7,(1 - 2r,n) = ** <5„r /*"* g (b " '' + ¿)!x(2n) ! J \ f=ó (r - 1 - i) !(2i) !
2A' r/„ . , „• , , „ï o2—2(2w - 2r + 1)1(I)' J(n _ r + i + i, n) _ 2—2 (2"(-; + 1)! (26,)-,
y (r-l+t)! /2,2V' ,„ , . ■ ■• 2-, 7-rrr.-. . I r— 1 K(2r - ÎI - 1, r + i, n.
,=o (n. — r — i)\(2i + 1)! \6///
r = 1, 2, - • • , where
/"" '„2n . i ,sin 0 d4>
(a - COS 0)"1
and
1 (0 — cos 4>)'l~ sin " 4> d<j>(3.7) 7i(9,TO,n) = [
Jo
with
(3.8) a =
(a — cos </>)"
& + y
2by '
and with ß defined as in (2.8). It follows from (3.4) and (3.5) that q may be a
positive or negative integer, or zero, and that to ^ n in (3.6) and (3.7). We see
that (3.6) and (3.7) are analytic functions of a when the a-plane is cut along the
real axis from — 1 to 1, and that (3.7) is an analytic function of ß when the /3-plane
is cut from 1 to — ». In order for K(q,m,n) to be positive when a > 1 and ß > 1,
we choose | arg (ß — cos t¡>) | < x in (3.7).
Substituting cos <t> = t in (3.6) and comparing with [18, p. 195], equation (20),
we find that
271 + 1/2-p/ i i \Í \n -\- 2 ) A"-m+W)ri/ 2 , \ (7i-m)/2+l/4nra-n-l/2/ \
(,3.9) J(m,n) = -v-^-e (a — 1) v»-i/2 (a)(to)
when the a-plane is cut from 1 to — « along the real axis, with | arg (a — 1 ) [ < x
and | arg (a + 1) | < x. Using Whipple's relation in (3.9) [18, p. 245], equation
(92), we obtain
rotrw tí \ (2n)!x, 2 .(„_m)/2D_„ i a \(3.10) J(m>n) = ___(a _i) pn_m ^-—=j
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
428 A. H. VAN TTJYL
in the cut a-plane whenRe a > 0. With | arg (« — 1) | < x and | arg (a +1) | < x
as before, we see that Re (a/y/o? — 1) > 0 both for Re a < 0 and Re a > 0, and
that the imaginary axis is mapped twice onto the segment 0 ^ a/\/a? -1^1.
Writing
(3.11) J(m,n) =^-~J(m,n),
we can verify that
(3.12) j(m,n) = (a2 - I)1""'?;!, i-^==j , Re« > 0,
(3.13) = e^ml2(ai2 + \Yn^),2PT-m (y==) , *=±iai,*i> 0,
(3.14) = (-l)V - lY—^PT-mly^—^j , Re« < 0
when the a-plane is cut along the real axis from —1 to 1, where Pñ-m(x) is defined
in the usual way when 0 < x < 1 [18, p. 99]. From (3.10) and [18, p. 100], equation
(28), we find that
/ 2 t \{n—m)/2 n—m / . -\, / 2 -, \~i/2(3 15) Jim n) = (a ~ 1} Y (n-w + i)! _ (« - 1)KÖ.LO) JKm,n} {a + ^^—¡y {^ {n _ m _ i)[(n +i)Hl2i (a + ^/¿rTjy
in the cut plane when Re « > 0, and with n 2ï m and n ^ 0. Noting that both
sides of (3.15) are analytic when the a-plane is cut along the real axis from —1 to
1, it follows by analytic continuation that (3.15) holds throughout the cut a-plane.
It can be verified that (3.9) through (3.15) remain valid when to < 0, and that
(3.15) simplifies to 2~n when m = 0.
Similarly, substituting cos 4> = 2t — 1 in (3.7) and comparing with [17, p. 231],
equation (5), we have
„, , (2n)!T (ß + ir112Mq'm'n) = 2^W " («+D-
(3.16)
• F1(n + -,m,- - q,2n + 1;2' '2 " 'a+l'/3+l,
where Fi(a, b,b ,c; x, y) is the first of Appell's generalized hypergeometric functions
of two variables. From (3.4), (3.5), (3.11) and (3.16), we find that
7i(-2r,n) = -2-r(by)r^x
g (n-r + Ql (2x\jk(n-\- i)\(2i + l)l\by) J{n r + t + 1> n>
^r—Zn—312
(3.17) ' (2r- l)!n!
2"-3n-w(2n - 2r)l 1/2 ̂ r (r-l+i)! (tofV
,«„ in,„, ^V) ¿j (n-r-i)\(2i)l\by)
Fl(n + 5 r + i,n - 2r + -,2n + 1;(a + 1)^
2 2 \
«+ l'/3 + 1/*
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
SOME DEFINITE INTEGRALS INVOLVING BESSEL FUNCTIONS 429
7,(1 ~2r,n)
= 2-(%r £ ?(n;rt%, (^7 J(n - r + i + 1, n)i-o (r - 1 - z)!(2t)! \&?//
_ 23r-3n-"2(2n - 2r + 1)1 . _s/2 yT (r-l+i){ /2x2V'
(3.18) (2r-2)1*1 k J) X h (n - r - i)\ (2* + 1)! \by)
(ß + 1)2—3'2-, /,!.,,„ 0. , 3Fi(n + -,r + i,n-2r + ^,2n + l;
(a + l)r+i \ 2 ' '2
+ 1 ' ß + 1/ ' =
Finally, we consider the evaluation of K(q, to, n) in terms of elliptic integrals.
Let k be given by (2.8), and let
(3.19) Z2= 2 4by
a + 1 (6 + î/)2 "
We see that the cut in the a-plane from — 1 to 1 corresponds to a cut in the I -plane2
along the real axis from 1 to «j. Substituting cos <j> = 2sn u — 1 in (3.7), we have
22n-m+g+l,2z2ro K m2nu ^ ^2^ ^
(3.20) £(«,m,n)- ^ jJo (1 - Z2sn2w)m
When A; and I" lie in their cut planes, the path of integration in the w-plane corre-
sponding to 0 í= 4> á x is a simple curve which does not pass through any singu-
larities of the integrand. For values of k2 and t in the cut planes such that k2 ^ t,
k ?¿ 0, we see that (3.20) can be expressed in terms of complete elliptic integrals of
the first, second, and third kinds. When k = I, and hence, x = 0, only complete
elliptic integrals of the first and second kinds occur. Finally, when either b or y
vanishes, the integrand reduces to sin nu cos "u, and the upper limit becomes x/2.
Let
„K in in 7,__, \ fn \ C sn ucn udu(3'21) K(m,n) = I (1 _ lWu)m i
„K in in j/ooo\ ti \ r sn ucn udu(3-22) L(m,n) = l dntmu .
Then when q ^ 0, we have
^271-171+3+1/2, 2m-2g+l72m-25 t» / \ / ,2 ,2\t
(3-23) K{^n) = (P-V« £.U - *)t-¿(í>)'
and when g < 0, we find on expanding du2qu(\ — tsnu)~m in partial fractions that
r,2n-m+5+l/2, 27n-29+l727íi-2(¡
7í(g, to, n)(Z2 - fc2)m~q
(3.24) • {g (-ir+«+< (_;!\) £^£ l(z»
+ E( * .)^^7C(,»lírl \to — %) k2'
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
430 A. H. VAN TUYL
We see that (3.21) and (3.22) can be expressed in terms of integrals of the forms
»X nK
(3.25) Un= [ (l-l2snu)ndu, Vn = / (1 - l2snu)-n duJo ¿o
and
(3.26) un = I dn"udu, v» = / dn~2nudu,Jo Jo
respectively, where n Sï 0, and the latter can be evaluated by use of well-known
recurrence relations [19, equations (331.03), (336.03), (314.05) and (315.05)].
Substituting q = \ in (3.20) and letting k tend to zero with I fixed, we obtain
(3.27) J(m, n) = 22^l2m C ¿f^™^^ .Jo (1 — I sin2 u)m
Thus, an alternative method for the evaluation of J(m, n) is to express (3.27) in
terms of the integrals
-.W2 W2
(3.28) gn = / (1 — I2 sin u)n du, hn = / (1 — Z2sin2 uY" du,Jo Jo
and to evaluate the latter by means of recurrence relations obtained from those for
Un and Vn , respectively.
4. Recurrence Relations for h(m, v) and 72(to, v). Using the identities
(4.1) ^MZ) =V-J,(Z) -Jr+xU)
and
(4.2) £jM =J,-i(z)-VJ,(z),dz z
and assuming the conditions on x, y, and b mentioned earlier, we have
y f e-'JUbt) ¡JUyt) - (y~1)f-lWl tm+1 dt(4.3) Ja L u Vt J
= f e-*T+1JUbt) 1 JUyt) dt,Jo dt
b re-juyt)\(>--l)'rm - um(4.4) J° L m U
= f e-xttm+1J^(yt) | J^bt) dt,Jo at
and
b f e~xtJr(yt) \j,-Ábt) - "4+^1 tm+1dt
(4.5) J° L M J= f e-xttm+lJv(yt)d-Jv(bt)dt.
Jo dt
r+1 <ii
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
SOME DEFINITE INTEGRALS INVOLVING BESSEL FUNCTIONS 431
We note that any other limits of integration can be used in the preceding as well
as 0 and <». Integrating the right-hand sides of (4.3), (4.4) and (4.5) by parts,
and using (4.1) in the first two cases and (4.2) in the last, we obtain the relations
bli(m, v) = yh(m, v — 1) + (to + 1)7i(to, v — 1) — xl\(m + 1, v — 1)
0, Re (2v + m - 1) > 0
+(4.6)
(by)'2v + TO - 1 = 0,
(4.7)
and
(2v + to - 1)7!(to, v - 1)
= y [ e-xtJv(yt)J„-i(bt)tm+ï dtJo
22"-2[r(y)]2'
bh(m, v) — xh(m + 1, v — 1)
0, Re (2v + m - 1) > 0
(by)-1
22-2[r(v)] 2 '2v + m - 1 = 0,
(4.8)
(2v — to — 1)7i(to, v) — yh(m, v) + xlx(m + 1, v)
= b [ e-xtJ„(yt)J„-i(bt)tm+1 dt, Re(2,/ + to + 1) > 0,Jo
respectively. Multiplying (4.7) by b and (4.8) by y and subtracting, we obtain
(if — b2)h(m, v) = (2v — to - l)yh(m, v) 4- xyh(m + 1, v)
-(2v + to - l)bh(m, v - 1) + zWiXto + 1, v - 1)
(4.9) 0, Re (2v + to - 1) > 0
2v + to - 1 = 0..22"-2[r(iv)]2'
Finally, from (4.6) and (4.9), we find the recurrence relation
(2v — to + l)byh(m, v + 1) = 2v(y + b2)h(m, v)
— bxyli(m + 1, v + 1) — (2v + to — \)byl\(m, v — 1)
(4.10)
+ bxyl\(m -\- 1, v — 1)
0, Re (2)v + to - 1) > 0
(&2/r+122"[r(,.+ i)]2'
2¡> + to - 1 = 0
for 7i(to, v) alone.
It is easily verified that when 7j.(to, n) is known for 2n + to = 0, 1, and 2, to ^
— 1, and when 7j.(0, n) is known for n ^ 0, we can obtain I\(m, n) for all other
values of to < 0 and n ^ 0 for which it converges by means of (4.10). We can then
obtain 72(to, n) from (4.9). Also, when 7j.(to, 0) and 7i(to, 1) are known for m^0,
and when 7i(0, n) is known for n ^ 0, we can calculate 7i(to, n) for all other values
of to > 0 and n > 1 by solving (4.10) for 7i(to + 1, n + 1). As before, we can
then obtain 72(to, n) from (4.9).
U. S. Naval Ordnance Laboratory
White Oak, Silver Spring, Maryland
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use
432 A. H. VAN TTJYL
1. A. Weinstein, "On axially symmetric flows," Quart. Appl. Math., v. 5, 1948, p. 429-444.2. A. Weinstein, "On the torsion of shafts of revolution," Proc. Vllth Int. Congr. Appl.
Mech., v. 1, 1948, p. 108.3. H. Bateman, Partial Differential Equations of Mathematical Physics, Dover, New York,
1944.4. A. G. Webster, Partial Differential Equations of Mathematical Physics, Hafner, New
York, 1947.5. G. Eason, B. Noble & I. N. Sneddon, "On certain integrals of Lipschitz-Hankel type
involving products of Bessel functions," Philos. Trans. Roy. Soc, v. 247, 1955, p. 529-551.6. Y. L. Luke, Integrals of Bessel Functions, McGraw-Hill, New York, 1962.7. W. M. Hicks, "On toroidal functions," Philos. Trans. Roy. Soc, v. 171, 1881, p. 609-
652.8. G. M. Minchin, "The magnetic field of a circular current," Philos. Mag., v. 35, 1893,
p. 354-365.9. A. Van Tutl, "On the axially symmetric flow around a new family of half-bodies,"
Quart. Appl. Math., v. 7, 1950, p. 399-409.10. M. R. Shura-Bura, "Calculation of an integral containing a product of Bessel func-
tions," Dokl. Akad. Nauk. SSSR., v. 73, 1950, p. 901-903.11. M. A. Sadowsky & E. Sternberg, "Elliptic integral representations of axially sym-
metric flows," Quart. Appl. Math., v. 8, 1950, p. 113-126.12. G. E. Pringle, "The potential of a circular current," Proc. Cambridge Philos. Soc, v.
57, 1961, p. 385-392.13. H. E. Fettis, "Note on /¡¡° e-xJo(f,x/í)Ji(x/í)x-'í dx," Math. Comp., v. 14, 1960, p.
372-374.14. G. P. Weeg, "Numerical Integration of /0°° e~V0(W£).A(:c/£)x~* dx," MTAC, v.
13, 1959, p. 312-313.15. A. Weinstein, "Discontinuous integrals and generalized potential theory," Trans.
Amer. Math. Soc, v. 63, 1948, p. 342-354.16. G. N. Watson, A Treatise on the Theory of Bessel Functions, Second Edition, Cam-
bridge University Press, 1944.17. A. Erdélyi, W. Magnus, F. Oberhettinger & F. G. Tricomi, Higher Transcendental
Functions, Vol. 1, McGraw-Hill, New York, 1953.18. E. W. Hobson, The Theory of Spherical and Ellipsoidal Harmonics, Cambridge Uni-
versity Press, 1931.19. P. F. Byrd & M. D. Friedman, Handbook of Elliptic Integrals for Engineers and Physi-
cists, Springer, Berlin, 1954.
License or copyright restrictions may apply to redistribution; see https://www.ams.org/journal-terms-of-use