Professional Documents
Culture Documents
1
To see that this equation is of Sturm-Liouville form, divide through by :
µ ¶
2
+ 2 − =0 (1)
Now we have a Sturm-Liouville equation with () = () = 2 eigen-
value = 2 and weighting function () = Equation (1) is Bessel’s equation.
The solutions are orthogonal functions. Since (0) = 0 we do not need to spec-
ify any boundary condition at = 0 if our range is 0 ≤ ≤ as is frequently
the case. We do need a boundary condition at =
It is simpler and more elegant to solve Bessel’s equation if we change to the
dimensionless variable = Then:
µ ¶
2
+ 2 − = 0
µ ¶
2
+ − = 0
The equation has a singular point at = 0 So we look for a series solution of
the Frobenius type (cf Lea Chapter 3 §3.3.2):
∞
X
=
=0
∞
X
0 = ( + ) +−1
=0
µ ¶ X∞
2
= ( + ) +−1
=0
2
Let’s look first at the solution with = + We can step down to find each
If we start the series with 0 then will always be even, = 2 and
−1 −1
2 = 2−4
2 (2 + 2) (2 − 2) (2 − 2 + 2)
3
(−1)
= 2−6
23 ( − 1) ( − 2) 23 ( + ) ( + − 1) ( + − 2)
(−1) 1
= 0
2 ! 2 ( + ) ( + − 1) · · · ( + 1)
The usual convention is to take
1
0 = (2)
2 Γ ( + 1)
Then
1 (−1) 1
2 =
2 Γ ( + 1) 2 ! 2 ( + ) ( + − 1) · · · ( + 1)
(−1)
= (3)
!Γ ( + + 1) 22+
and the solution is the Bessel function:
X∞
(−1) ³ ´+2
() = (4)
=0
!Γ ( + + 1) 2
The function () has only even powers if is an even integer and only odd
powers if is an odd integer. The series converges for all values of
Let’s see what the second solution looks like. With = − the recursion
relation is:
−2
= (5)
( − 2)
where again = 2 Now if is an integer we will not be able to determine
2 because the recursion relation blows up. One solution to this dilemma is
to start the series with 2 Then we can find the succeeding 2 :
−2(−1)+2 −2(−2)+2 (−1) Γ ( + 1)
2(+) = = = 2
22 ( + ) 24 ( + ) ( + − 1) ( − 1) !Γ ( + + 1) 22
which is the same recursion relation we had before. (Compare the equation
above with equation (3). Thus we do not get a linearly independent solution
this way. (This dilemma does not arise if the separation constant is taken to
be − 2 with non-integer. In that case the second recursion relation provides
a series − () that is linearly independent of the first.) Indeed we find:
∞
X
(−1) Γ ( + 1)
− () = 2(+)−
2 2
=0
!Γ ( + + 1) 2
X∞
(−1) Γ ( + 1) 2 ³ ´+2
= 2
=0
!Γ ( + + 1) 2
3
and if we choose
(−1)
2 =
Γ ( + 1) 2
then
− () = (−1) () (6)
With this choice () is a continuous function of (Notice that we can also
express the series using equation (3) for the coefficients, with → − and
→ + and where 2 ≡ 0 for )
We still have to determine the second, linearly independent solution of the
Bessel equation. We can find it by taking the limit as → of a linear
combination of and − known as the Neumann function () :
() cos − − ()
() = lim () = lim
→ → sin
+ () cos ( + ) − −(+) ()
= lim
→0 sin ( + )
+ () (cos cos − sin sin ) − −(+) ()
= lim
→0 sin cos + cos sin
+ () (−1) cos − −(+) ()
= lim
→0 (−1) sin
Now we expand the functions to first order in We use a Taylor series for the
Bessel functions. Note that appears in the index, not the argument, so we
have to differentiate with respect to .
+ () (−1) − −(+) ()
() = lim
→0 (−1)
∙ µ ¯ ¶ µ ¯ ¶¸
(−1) ¯¯ − ¯¯
= lim (−1) + − − +
→0 ¯= ¯=
and
ln
= = ln ln = ln
4
and so has a term containing ln This term diverges as → 0
provided that (0) is not zero, i.e. for = 0 The function () also
diverges as → 0 for 6= 0 because it contains negative powers of (The
series for − starts with a term − ) is finite as → ∞ because goes
to zero sufficiently fast.
Two additional functions called Hankel functions are defined as linear com-
binations of and :
(1)
() = () + () (7)
and
(2)
() = () − () (8)
Compare the relation between sine, cosine, and exponential:
± = cos ± sin
1.0
J
0.8
0.6
0.4
0.2
0.0
2 4 6 8 10 12 14
-0.2 x
-0.4
5
For large values of the argument, both and oscillate: they are like damped
cosine or sine functions:
r ³
2 ´
() ≈ cos − − for À 1 (11)
2 4
r ³
2 ´
() ≈ sin − − for À 1 (12)
2 4
and thus the Hankel functions are like complex exponentials:
r h ³
(12) 2 ´i
≈ exp ± − − for À 1 (13)
2 4
Notice that if 1 the large argument expansions apply for À rather
than the usual À 1
6
boundary condition at = 0 because the function () = is zero there. Then
the eigenvalues are
=
where is the th zero of (The zeros are tabulated in standard references
such as Abramowitz and Stegun. Also programs such as Mathematica and
Maple can compute them.) Then
Z
2 2 0 2
[ ( )] = [ ( )] (19)
0 2
Next we make use of the orthogonality of the Bessel functions. Multiply both
sides by 0 (0 )and integrate from 0 to (Note here that the weight function
() = This is the first time we have seen a weight function that is not 1.)
Only one term in the sum, with = survives the integration.
Z Z ∞
X
0 (0 ) = 0 (0 ) 0 (0 ) sinh (0 )
0 0 =1
Z
2
= [0 (0 )] sinh (0 )
0
2 2
= [00 (0 )] sinh (0 )
2
To evaluate the left hand side, we use equation (16) with = 1 :
Z Z
1
0 () = (1 ())
0 0
1
= 1 ()|0 = 1 ()
7
So
2
= 1 (0 ) 2
0 2 [00 (0 )] sinh (0 )
2
=
0 1 (0 ) sinh (0 )
where we used the result from equation (15) that 00 = −1 Finally our solution
is:
X∞
0 (0 ) sinh (0 )
Φ = 2
=1 0
1 (0 ) sinh (0 )
The first two zeros of 0 are: 01 = 2.4048, 02 = 5.5201, and thus the first
two terms in the potential are:
à ¡ ¢ ¡ ¢ ¡ ¢ ¡ ¢ !
0 24048 sinh 24048 0 55201 sinh 55201
Φ = 2 ¡ ¢+ ¡ ¢ + ···
240481 (24048) sinh 24048 552011 (55201) sinh 55201
2
= −2 ⇒ = sin + cos
2
At = 0 () = 0 so we need the sine, and therefore set = 0 We also need
() = 0 so we choose the eigenvalue = .
This change in sign of the separation constant also affects the equation for
the function ()because the sign of the 2 term changes.
µ ¶
2
− 2 − =0
which is called the modified Bessel equation. The solutions to this equation
are () It is usual to define the modified Bessel function () by the
relation:
1
() = () (21)
8
so that the function is always real (whether or not is an integer). Using
equation (4) we can write a series expansion for :
∞ µ ¶+2
1 X
(−1)
() =
=0 !Γ ( + + 1) 2
∞
X 1 ³ ´+2
= (22)
=0
!Γ ( + + 1) 2
and
0 () ≈ −05772 − ln for ¿ 1 (26)
µ ¶ 2
Γ () 2
() ≈ 0 for ¿ 1 (27)
2
9
10
I,K
8
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
x
Modified Bessel functions. solid dashed Black = 0Blue 1 Red 2
The recursion relations satisifed by the modified Bessel functions are similar
to, but not identical to, the relations satisfied by the s. For the s, again we
can start with the series:
µ ¶ ∞ ³ ´2
1 X 1
=
2 =0 !Γ ( + + 1) 2
1 X
∞
³ ´2−1
=
2 =1 !Γ ( + + 1) 2
Now let = − 1 :
µ ¶ ∞ ³ ´2+1
1 X 1
=
2 !Γ ( + + 2) 2
=0
1 X∞
1 ³ ´2++1
=
!Γ ( + + 2) 2
=0
µ ¶
+1 ()
= (30)
and similarly
( ) = −1 (31)
Expanding out and combining, we get:
0
2 = +1 + −1
2
= −1 − +1 (32)
For the s, the relations are:
µ ¶
() +1 ()
( ) = − −1 ; =−
10
and consequently:
2
−1 − +1 = −
0
−1 + +1 = −2 (33)
11
We make use of orthogonality by mutltiplying by sin and integrating from
0 to
+∞ X
X ∞ Z ³ ´ ³ ´ Z
sin sin = sin sin
=−∞ =1 0
0
³ ´ ¯¯
+∞
X
= − sin cos ¯
=−∞
2 0
= sin [1 − (−1) ]
The result is non-zero only for odd
Next we make use of the orthogonality of the
+∞
X Z ³ ´ Z 2
2 −0 0
= 2 sin − odd
=−∞
2 0 0
= 0 even
4
∞
X ³ ´ sin ¡ ¢
1
Φ ( ) = sin ¡ ¢
1
=1 odd
The dimensions are correct, and the 1 gives us reasonable convergence. The
ratio ¡ ¢
1
¢
¡ ≤ 1 for ≤
1
The potential at = 0 is zero, as expected. The = 1 "source" (the potential
on the surface) generates an = 1 response. Again we can look at the first few
12
terms (here 10) . Let’s take = 2 and = 4 (red) and Φ = 2 (black)
phi/V 1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
rho/a
becomes Z ∞
1 0
− = ( − 0 )
2 −∞
That is, the Kronecker delta becomes a delta function, and the countable set
of eigenvalues becomes a continuous set of values −∞ ∞.
The same thing happens with Bessel functions. With a finite domain in
say 0 ≤ ≤ we can determine a countable set of eigenvalues from the set of
zeros of the Bessel functions If our domain in becomes infinite, then we
cannot determine the eigenvalues, and instead we have a continuous set. The
orthogonality relation:
Z ³ ´ ³ ´ 2 0 2
= [ ( )]
0 2
13
becomes: Z ∞
( − 0 )
() (0 ) = (34)
0
(The proof of this relation is in Lea Appendix 7.) Then the solution to the
physics problem is determined as an integral over For example, a solution of
Laplace’s equation may be written as:
X Z ∞
Φ ( ) = () () ()
0
Now we can make use of the orthogonality of the Multiply both sides by
0
− and integrate over the range 0 to 2 On the LHS, only the term with
= 0 survives the integration, and on the RHS only the term with = 0
survives.
Z 2 µ ¶ Z 2 X Z ∞
0 0
0 sin − = − 0 () 0 ()
0 0 0
µ ¶ Z ∞
20 sin 0 0 = 2 0 () 0 ()
0
—a Fourier Bessel transform. Next1 multiply both sides by (0 ) integrate
from 0 to ∞ in and use equation (34) to get:
Z ∞µ ¶ Z ∞Z ∞
0
0 sin ( ) 0 0 = () () (0 )
0 0 0
Z ∞
( − 0 ) (0 )
= () =
0 0
which determines the coefficient (0 ) in terms of the known potential on the
plane. Only 0 is non-zero, as expected from the azimuthal symmetry.
1 We also drop the primes on the 0 for convenience.
14
On the RHS, let = We get (dropping the primes on )
Z ∞
0 ()
0 2 sin 0 () =
0
This integral is GR 6.671#7. So
(
0 0 1
() = 0 2 √ 1
0 1
1−()2
So finally we have
Z 1
() −
Φ ( ) = 0 q 0
2
0 2
1 − ()
Z 1 ¡ ¢
0 −
= 0 √
0 1 − 2
For = 0 we have
Z ¡ ¢
1
0 1
Φ ( 0) = 0 √ = 0 sin GR 6.554#2
0 1− 2
which gives us back the potential on the plane.
The charge density on the plane is given by
¯
· ̂ ¯¯
=
=0 0
So we need
¯ Z ¡ ¢
Φ ¯¯ 0 2 0
1
= − √
¯=0 0 1 − 2
Z 1
(−1) ³ ´2
∞
X
0 2
= − √
0 1 − 2 =0 [!]2 2
Z
(−1) ³ ´2 1 2(+1)
X∞
0
= − 2
√
=0 [!]
2 0 1 − 2
Let = sin = cos Then the integral is
Z 2 Z 2
sin2(+1) (2 + 1)!!
= cos = sin2(+1) = GR 3.621#3, Lea Ch2 P29d
0 cos 0 (2 + 2)!! 2
Thus
¯
0 X (−1) ³ ´2
∞
Φ ¯¯ (2 + 1)!!
= −
¯=0 2 =0 [!]2 2 2+1 ( + 1)!
0 X (−1) ³ ´2
∞
(2 + 1)!!
= −
4 =0 (!)2 2 2 ( + 1)!
15
and so the charge density on the plane is
∞ µ 2 ¶
0 X (−1) (2 + 1)!!
= 0 = 0
4 =0 (!)2 82 ( + 1)!
1.0
sigma a/epsV
0.8
0.6
0.4
0.2
0.0
1 2 3 4 5 6 7 8 9 10
rho/a
-0.2
Convince yourself that this agrees with the field line diagram. (The dashed line
is the potential.)
16