ORDINARY DIFFERENTIAL EQUATION
CHAPTER 8
RIEMANN P FUNCTION
Riemann P-function :
We now find the second order differential equation of Fuchsian type having regular sin-
gularities at three points a, b, c (∞ is an ordinary point) . It being given that the exponents
at a, b, c are given by α, α′, β, β ′, γ, γ ′ . Let the required equation be
d2 w dw
+ p 1 (z) + p2 (z)w = 0 (A)
dz 2 dz
Since a, b, c are only singularities of differential equation all of which are regular, then
P1 (z) P2 (z)
p1 (z) = and p2 (z) =
(z − a)(z − b)(z − c) (z − a)2 (z
− b)2 (z − c)2
where P1 (z), P2 (z) are holomorphic in the entire complex plane.
Recall Chapter 1. Setting t = 1z , (A) transforms to
d2 w
2 1 1 dw 1 1
+ − p 1 ( ) + p 2 ( )w = 0 (B)
dt2 t t2 t dt t4 t
Since ∞ is an ordinary point of (A), ie., t = 0 ia an ordinary point of (B), it follows that
2 1 1
F (t) = t − t2 p1 t is also holomorphic in a neighbourhood of zero.
t3 P1 1t
or, [2 − tF (t)]t =
(1 − at)(1 − bt)(1 − ct)
1
or, (1 − at)(1 − bt)(1 − ct) [2 − tF (t)] = t2 P1
t
Since P1 (z) is an entire function , it follows that P1 (z) is a quadratic function in z. Also the
coefficient of z 2 is 2. This is due
to following reason:
Given that F (t) = 2t − t12 p1 1t is also holomorphic in a neighbourhood of zero. So
2 1 1 2 tP1 (1/t)
lim F (t) = lim − 2 p1 = lim −
t→0 t→0 t t t t→0 t (1 − at)(1 − bt)1 − ct)
Let P1 (z) = qz 2 + rz + s so that P1 (1/t) = (q + rt + st2 )/t2 . Hence
2 tP1 (1/t) 2 q
lim F (t) = lim − = lim − − r = −r (finite)
t→0 t→0 t (1 − at)(1 − bt)(1 − ct) t→0 t t
if q = 2.
Similarly we show that P2 (z) is also a quadratic function and we can write
A B C
p1 (z) = + + where A + B + C = 2
(z − a) (z − b) (z − c)
1
and
D E F
(z − a)(z − b)(z − c) p2 (z) = + +
(z − a) (z − b) (z − c)
where A, B, C, D, E, F are constants.
We write the required equation considering z = a as the regular singular point
d2 w B(z − a) C(z − a) dw
(z − a)2
+ (z − a) A + +
dz 2 z−b z−c dz
D E(z − a) F (z − a)
+ + + w=0
(z − b)(z − c) (z − b)2 (z − c) (z − b)(z − c)2
d2 w B(z − a) C(z − a) dw
(z − a)2
+ (z − a) A + +
dz 2 z−b z−c dz
D E(z − a) F (z − a)
+ + + w=0
(z − a + a − b)(z − a + a − c) (z − b)2 (z − c) (z − b)(z − c)2
d2 w B(z − a) C(z − a) dw
(z − a)2
+ (z − a) A + +
dz 2 z−b z−c dz
D (z − a) −1 (z − a) −1 E(z − a) F (z − a)
+ {1 + } {1 + } + + w=0
(a − b)(a − c) (a − b) (a − c) (z − b) (z − c) (z − b)(z − c)2
2
d2 w B(z − a) C(z − a) dw
(z − a)2
2
+ (z − a) A + +
dz z−b z−c dz
k j
D k (z − a) j (z − a)
+ {1 + Σ∞k=1 (−1) } {1 + Σ∞
j=1 (−1) }
(a − b)(a − c) (a − b)k (a − c)k
E(z − a) F (z − a)
+ + w=0
(z − b)2 (z − c) (z − b)(z − c)2
We assume the solution in some deleted neighbourhood of z = a is
w = (z − a)ρ Σ∞ n
n=0 dn (z − a) ,
ρ being the exponent with respect to the singularity z = a. Substituting this in the given
equation we get
ρ+n B(z − a) C(z − a) ∞
Σn=0 dn (ρ + n)(z − a)ρ+n
Σ∞
n=0 (ρ + n)(ρ + n − 1)d n (z − a) + A + +
z−b z−c
k j
D k (z − a) j (z − a)
+ {1 + Σ∞
k=1 (−1) } {1 + Σ∞
j=1 (−1) }
(a − b)(a − c) (a − b)k (a − c)k
E(z − a) F (z − a) ∞
+ + Σ dn (z − a)ρ+n = 0
(z − b)2 (z − c) (z − b)(z − c)2 n=0
The indicial equation for z = a is obtained by making the coefficient of z ρ equal to zero
which correspond to n = 0. This gives
2
D
ρ(ρ − 1) + Aρ + =0
(a − b)(a − c)
D
ρ2 + (A − 1)ρ + =0
(a − b)(a − c)
This is a quadratic equation in ρ where ρ = α, α′ so that
D
α + α′ = 1 − A, αα′ =
(a − b)(a − c)
Similar analysis can be made for z = b and z = c.
Since the exponents of a, b, c are α, α′, β, β ′, γ, γ ′ , roots of indicial equation are given by
A = 1 − α − α′ ; B = 1 − β − β ′ ; C = 1 − γ − γ ′
D = αα′(a − b)(a − c) ; E = ββ ′ (b − a)(b − c) ; F = γγ ′ (c − a)(c − b)
Since A + B + C = 2, so
α + α′ + β + β ′ + γ + γ ′ = 1 (1)
Hence the required differential equation is
d2 w
1 − α − α′ 1 − β − β ′ 1 − γ − γ ′ dw
+ + + −
dz 2 z−a z−b z−c dz
αα′ ββ ′ γγ ′ (a − b)(b − c)(c − a)
+ + w = 0 (2)
(z − a)(b − c) (z − b)(c − a) (z − c)(a − b) (z − a)(z − b)(z − c)
(∗ Interchanging either α, α′ or ββ ′ or γ, γ ′ (2) is unchanged.)
Any solution of Riemann (2) is called Riemann’s P-function and is denoted by the scheme
a b c
P α β γ z
α′ β ′ γ ′
Equation (2) is called Riemann P-equation or sometimes Generalized Hypergeometric
equation. Making b → ∞, putting a = 0, c = 1, (2) implies
d2 w
1 − α − α′ 1 − γ − γ ′ dw αα′ γγ ′ ′ w
2
+ + + − + + ββ = 0
dz z z−1 dz z (z − 1) z(z − 1)
i.e.
d2 w (1 − α − α′ ) − (1 + β + β ′ )z dw αα′ − (αα′ + ββ ′ − γγ ′ )z + ββ ′z 2
+ + w=0 (3)
dz 2 z(1 − z) dz z 2 (1 − z)2
any solution of which is
0 ∞ 1
P α β γ z
α′ β ′ γ ′
3
Definition
0 ∞ 1
α β γ z
P = P α β γ z
α′ β ′ γ ′
α′ β ′ γ ′
Remark
0 α 0 z
The differential equation satisfied by the particular P-function P
1−γ β γ−β−α
is Hypergeometric equation.
2
Recall that equation is given by z(1 − z) ddzw2 + {γ − (α + β + 1)z} dw
dz
− αβw = 0
R1
a b c a b c
(z − a)p (z − c)q
P α β γ z = P α+p β−p−q γ+q z
(z − b)p+q
α′ β ′ γ ′ α′ + p β ′ − p − q γ ′ + q
Proof :
Set
(z − a)p (z − c)q
W = w
(z − b)p+q
under above transformation, only singularity of the equation continue to be a, b, c which are
regular and the exponents at a, b, c change to α + p, α′ + p; β − p − q, β ′ − p − q; γ + q, γ ′ + q
respectively, s.t. sum of the exponents remain unaltered. Hence the result.
R2
pz+q
The bilinear transformation t = rz+s
, ps 6= rq maps the points a, b, c to a′ , b′ , c′ , then
a b c a′ b′ c′
P α β γ z = P α β γ t
α′ β ′ γ ′ α′ β ′ γ ′
R3
(z−a)(b−c)
Setting t = (z−b)(a−c)
then,
a b c
α β γ t
P α β γ z = P
α′ β ′ γ ′
α β γ
′ ′ ′
This transformation changes a, b, c to 0, ∞, 1.
4
R4
p q α β γ z α+p β−p−q γ+q z
z (1 − z) P = P
α′ β ′ γ ′ α′ + p β ′ − p − q γ ′ + q
R5
By setting t = z1 ,
α β γ z β α γ t
P = P
α′ β ′ γ ′ β ′ α′ γ ′
1 z z−1
Similar changes are obtained by the transformation t = 1 − z , t = 1−z
; t= −1+z
; t= z
.
Solution of Riemann P-equation in terms of Hypergeometric function :
We shall now show that Riemann’s P-equation (2) can always be transformed to Hyper-
geometric equation.
To show this we have from R1
a b c a b c
(z − a)p (z − c)q
P α β γ z = P α+p β−p−q γ+q z
(z − b)p+q
α β γ
′ ′ ′
α′ + p β ′ − p − q γ ′ + q
We know that the differentialequation satisfied by the particular P-function
0 α 0 z
P is Hypergeometric equation.
1−γ β γ−β−α
So in the above equation p = −α and q = −γ. Hence
a b c a b c
(z − a)−α (z − c)−γ
P α β γ z = P 0 β+α+γ 0 z
(z − b)−(α+γ)
α β γ
′ ′ ′
α −α β +α+γ γ −γ
′ ′ ′
a b c a b c
(z − a)α (z − c)γ
=⇒ P α β γ z = P 0 α+β+γ 0 z
(z − b)α+γ
α β γ
′ ′ ′
α −α α+β +γ γ −γ
′ ′ ′
(z−a)(b−c)
Now by R3 using transformation t = (z−b)(a−c)
, the triplet (a, b, c) transform to (0, ∞, 1)
so that
a b c
0 α+β+γ 0 t
P 0 α+β+γ 0 z =P
α −α α+β +γ γ −γ
′ ′ ′
α′ − α α + β ′ + γ γ ′ − γ
5
Hence we have
a b c
(z − a)α (z − c)γ
0 α+β+γ 0 t
P α β γ z = P
(z − b)α+γ α′ − α α + β ′ + γ γ ′ − γ
α β γ
′ ′ ′
satisfies the hypergeometric equation having parameter α + β + γ, α + β ′ + γ, 1 + α − α′ .
Now, we assume that none of the exponent differences α − α′ , β − β ′ , γ − γ ′ is zero
or an integer. Since α − α′ is neither zero, nor an integer a solution of (2) can be written as
a b c
(z − a)α (z − c)γ
0 α+β+γ 0 t
w1 = P α β γ z = P ,
(z − b)α+γ α′ − α α + β ′ + γ γ ′ − γ
α β γ
′ ′ ′
(z−a)(b−c)
where t = (z−b)(a−c)
.
(z − a)α (z − c)γ
′ ′ (z − a)(b − c)
w1 = F α + β + γ, α + β + γ, 1 + α − α ,
(z − b)α+γ (z − b)(a − c)
Interchanging α, α′ or γ, γ ′ or both we obtain 3 other solutions
′
(z − a)α (z − c)γ
′ ′ ′ ′ (z − a)(b − c)
w2 = F α + β + γ, α + β + γ, 1 + α − α,
(z − b)α′ +γ (z − b)(a − c)
′
(z − a)α (z − c)γ
′ ′ ′ ′ (z − a)(b − c)
w3 = F α + β + γ ,α + β + γ ,1 + α −α ,
(z − b)α+γ ′ (z − b)(a − c)
Interchanging both
′ ′
(z − a)α (z − c)γ
′ ′ ′ ′ ′ ′ (z − a)(b − c)
w4 = F α + β + γ , α + β + γ , 1 + α − α,
(z − b)α′ +γ ′ (z − b)(a − c)
Now P-equation is symmetric w.r.t columns of P-function (R5). Since 3 columns can
permutate in 6 ways and corresponding to each permutation of columns we get 4 solutions
of above type , we get all together 24 solutions . If any 3 of the solutions are valid in common
region, then they must be linearly dependent.
6
APPENDIX
FIND A SECOND ORDER DIFFERENTIAL EQUATION FUCHSIAN
TYPE HAVING ZERO, INFINITY AND ONE AS REGULAR SINGULAR
POINTS,
Let us consider the second order differential equation Fuchsian type having regular singu-
larities at three points, 0, ∞, 1. Let the exponents at 0, ∞, 1 be given by α, α′, β, β ′, γ, γ ′
respectively.
Let the required equation be
d2 w dw
2
+ p1 (z) + p2 (z)w = 0 (A1)
dz dz
Since z = 0 and z = 1 are regular singular points, we have
P1 (z) P2 (z)
p1 (z) = and p2 (z) = 2 (A2)
z(z − 1) z (z − 1)2
where P1 (z) and P2 (z) are holomorphic in the entire complex plane.
Let z = 1t .
d d dt 1 d d
= . = − 2 = −t2
dz dt dz z dt dt
and
d2 d 2 d
2
2 d dt d dt 2 d 2 d
2
= (−t ) = −t − 2t = t (2t + t )
dz 2 dz dt dt2 dz dt dz dt dt2
Substituting in (A1) we get,
d2 w dw 1 dw 1
t4 2
+ 2t3 + p1 ( )(−t2 ) + p2 ( )w = 0
dt dt t dt t
4 ′′ 2 1 ′ 1
=⇒ t w + t [2t − p1 ( )]w + p2 ( )w = 0
t t
1
P 1 ( ) P2 ( 1 )
=⇒ t4 w ′′ + t2 [2t − 1 1 t ]w ′ + 1 1 t 2 w = 0
( − 1)
t t
( − 1)
t2 t
t2 P1 ( 1t ) ′ t4 P2 ( 1t )
=⇒ t4 w ′′ + t2 [2t − ]w + w=0
1−t (1 − t)2
′′ 2 P1 ( 1t ) ′ P2 ( 1t )
=⇒ w + [ − ]w + w=0
t 1−t (1 − t)2
2 − 2t − tP1 ( 1t ) P2 ( 1t )
=⇒ w ′′ + w ′ [ ]+ w=0 (A3)
t(1 − t) (1 − t)2
Since z = ∞ is a regular singular point of the differential equation, so, t = 0 is a regular
singular point of (A3).
2−2t−tP ( 1 ) P2 ( 1t )
Hence, t[ t(1−t)1 t ] and t2 (1−t) 2 are holomorphic in a neighbourhood of t = 0.
7
Thus, P1 ( 1t ) must be of order 1
t
and P2 ( 1t ) must be of order 1
t2
.
If P1 ( 1t ) is of order 1
t2
, then let
1 p q p + qt + rt2
P1 ( ) = 2 + + r =
t t t t2
Hence
2 − 2t − tP1 ( 1t )
t
t(1 − t)
2
2(1 − t) − t p+qt+rt
t2
=t
t(1 − t)
2 p q r
=t − 2 − −
t t (1 − t) t(1 − t) (1 − t)
p q rt
=2− − −
t(1 − t) (1 − t) 1 − t
is not holomorphic in a neighbourhood of t = 0.
Hence, P1 ( 1t ) cannot be of order t12 .
If P1 ( 1t ) is of order 1t , then let
1 A A + Bt
P1 ( ) = + B =
t t t
Hence,
2 − 2t − tP1 ( 1t )
t
t(1 − t)
2 − 2t − t A+Bt
t
=t
t(1 − t)
2 A B
=t − −
t t(1 − t) 1 − t
A Bt
=2− −
1−t 1−t
is holomorphic in a nbd. of t = 0. Therefore
P1 ( 1t ) must be of order 1t .
Similarly , P2 ( 1t ) must be of order t12 .
Thus, P1 (z) must be of order z and P2 (z) must be of order z 2 .
Let
P1 (z) A B
p1 (z) = = +
z(z − 1) z z−1
and
P2 (z) Dz + E1 F E F
z(z − 1)p2 (z) = = + =D+ + where E = D + E1
z(z − 1) z−1 z z−1 z
8
Substituting the values in equation (A1),
d2 w
A B dw E F w
2
+ + + D+ + =0 (A4)
dz z z − 1 dz z−1 z z(z − 1)
d2 w 1
Bz dw 1 Ez w
=⇒ 2
+ A+ + 2 Dz + +F =0
dz z z − 1 dz z z−1 z−1
2
2d w Bz dw Ez w
=⇒ z 2
+z A+ + Dz + +F =0
dz z − 1 dz z−1 z−1
2
2d w Bz dw Dz Ez F
=⇒ z +z A+ + + + w=0 (A5)
dz 2 z − 1 dz z − 1 (z − 1)2 z − 1
Let w = ∞ ρ+n
P
n=0 an z be a solution of equation (A5).
Thus we have ,
2
X
ρ+n−2 Bz X
z an (ρ + n)(ρ + n − 1)z +z A+ an (ρ + n)z ρ+n−1
z−1
X
Dz Ez F
+ + + an z ρ+n = 0
z − 1 (z − 1)2 z − 1
∞ ∞
X
ρ+n Bz X
=⇒ an (ρ + n)(ρ + n − 1)z + A+ an (ρ + n)z ρ+n
n=0
z − 1 n=0
X∞
Dz Ez F
+ + + an z ρ+n = 0
z − 1 (z − 1)2 z − 1 n=0
X∞ X ∞
ρ+n
=⇒ an (ρ + n)(ρ + n − 1)z + A − Bz(1 − z) −1
an (ρ + n)z ρ+n
n=0 n=0
∞
X
+ − Dz(1 − z) −1
+ Ez(1 − z) −2
− F (1 − z) −1
an (ρ + n)z ρ+n = 0
n=0
Using (1 − z)−1 = 1 + z + z 2 + ......... and (1 − z)−2 = 1 + 2z + 3z 2 + ....
The indicial equation for z = 0 is obtained by making the coefficient of z ρ equal to zero
which correspond to n = 0.
This gives
ρ(ρ − 1) + Aρ − F = 0 =⇒ ρ2 + (A − 1)ρ − F = 0
This is a quadratic equation in ρ where ρ = α, α′. Therefore
α + α′ = 1 − A and αα′ = −F.
Therefore
F = −αα′ and A = 1 − α − α′ (A6)
By making similar analysis for z = 1, we obtain
E = γγ ′ and B = 1 − γ − γ ′
9
Since from (A4),
d2 w
A B dw E F w
2
+ + + D+ + =0
dz z z − 1 dz z−1 z z(z − 1)
d2 w
1 A(z − 1) dw 1 F (z − 1) w
=⇒ + B+ + D(z − 1) + E + =0
dz 2 z−1 z dz (z − 1)2 z z
2
2d w A(z − 1) dw D(z − 1) E F (z − 1)
=⇒ (z − 1) + (z − 1) B + + + + w=0
dz 2 z dz z z z2
Let w = ∞ ρ+n
P
n=0 bn (z − 1) be a solution. Therefore
∞
2
X
ρ+n−2 A(z − 1)
(z − 1) bn (ρ + n)(ρ + n − 1)(z − 1) + (z − 1) B + ×
n=0
z
∞ ∞
X
ρ+n−1 D(z − 1) E F (z − 1) X
bn (ρ + n)(z − 1) + + + 2
bn (z − 1)ρ+n = 0
n=0
z z z n=0
∞ X∞
X A(z − 1)
=⇒ bn (ρ + n)(ρ + n − 1)(z − 1)ρ+n + B + bn (ρ + n)(z − 1)ρ+n
n=0
z n=0
∞
D(z − 1) E F (z − 1) X
+ + + bn (z − 1)ρ+n = 0
z z z2 n=0
The indicial equation for z = 1 is obtained by making the coefficients of z ρ equal to zero
which correspond to n = 0.
This gives,
ρ(ρ − 1) + Bρ + E = 0 =⇒ ρ2 + (B − 1)ρ + E = 0
This is a quadratic equation in ρ where ρ = γ, γ ′ . Therefore
γ + γ ′ = 1 − B and γγ ′ = E
=⇒ B = 1 − γ − γ ′ and E = γγ ′ (A7)
Now from (A4),
d2 w
A B dw E F w
2
+ + + D+ + =0
dz z z − 1 dz z−1 z z(z − 1)
Let z = 1t . Therefore
d d d2 d d2
= −t2 and 2 = 2t3 + t4
dz dt dz dt dt2
So we have,
2
2
4d w 3 dw Bt 2 dw Et wt
t 2
+ 2t + At + −t + D+ + Ft =0
dt dt 1−t dt 1−t 1−t
2
Bt3 dw
4d w 3 3 2 Et w
=⇒ t + 2t − At − + t D + + F t =0
dt2 1 − t dt 1−t 1−t
2
2d w B dw D Et Ft
=⇒ t +t 2−A− + + + w=0
dt2 1 − t dt 1 − t (1 − t)2 1 − t
10
P∞ ρ+n
Let w = n=0 cn t be a solution.
X
ρ+n B X
cn (ρ + n)(ρ + n − 1)t + 2−A− cn (ρ + n)tρ+n
1−t
D Et F t X ρ+n
+ + + cn t =0
1 − t (1 − t)2 1 − t
The indicial equation for t = 0 is obtained by making the coefficient of tρ equal to zero
which correspond to n = 0.
This gives,
ρ(ρ − 1) + ρ(2 − A − B) + D = 0 =⇒ ρ2 + ρ(1 − A − B) + D = 0
This is a quadratic equation in ρ where ρ = β, β ′.
β + β ′ = A + B − 1 and ββ ′ = D
=⇒ β + β ′ = (1 − α − α′ ) + (1 − γ − γ ′ ) − 1.
Therefore
α + α′ + β + β ′ + γ + γ ′ = 1
and ββ ′ = D (A8)
Substituting the values of A, B, D, E, F from (A6), (A7), (A8) in equation (A4), we get,
d2 w
1 − α − α′ 1 − γ − γ ′ dw ′ γγ ′ αα′ w
2
+ + + ββ + − =0
dz z z−1 dz z−1 z z(z − 1)
any solution of which is
0 ∞ 1
P α β γ z
α′ β ′ γ ′
11