ORDINARY DIFFERENTIAL EQUATION
CHAPTER 3
SOLUTION IN THE NEIGHBOURHOOD OF
SINGULARITY
We shall investigate the existence of a fundamental set of solution of suitable form
in a neighbourhood of a singularity of a differential equation
dm w dm−1 w dm−2 w
= p 1 (z) + p 2 (z) + ..... + pm (z)w. (1)
dz m dz m−1 dz m−2
For simplicity we consider the case m = 2 , i.e. the second order DE
d2 w dw
2
= p1 (z) + p2 (z)w (2)
dz dz
Let z0 be a singularity of (2) and z1 be another singularity nearest to z0 and |z1 −z0 | =
r. We shall find a fundamental set of solution in deleted neighbourhiid N ′ (z0 , r)
of z0 . Let z be any point in N ′ (z0 , r) and D the domain of ordinary point z.
Take any positively oriented rectifiable closed Jordon curve Γ passing through z s.t.
Γ ⊆ N ′ (z0 , r) and z0 ∈ I(Γ). Let Γ : z(t), t ∈ [α, β] so that z(α) = z(β) = z. Then
we can find a partition {t0 , t1 , ...., tn } of [α, β] s.t. if ζi = z(ti ) (i = 0, 1, 2, ...., n) [ζ0 =
ζn = z] and Di denote the domain of ζi (i = 0, 1, ..., n) D0 = Dn = D . Then
D0 , D1 , ....., Dn form a closed chain of domains s.t. ζi ∈ Di−1 (i = 1, 2...., n). Let
w(z) be integral of (2) in D = D0 , w1 (z) is the analytic continuation of w(z) into D1 ,
w2 (z) the continuation of w1 (z) in D2 and so on and finally wn (z) is continuation
of wn−1 (z) into Dn = D . This wn (z) is the analytic continuation of w(z) by
means of chain of analytic functions w(z), w1 (z), ....., wn (z). Writing for convenience
wn (z) = W (z), we know that W (z) is in general different from w(z) and the above
procedure could be briefly expressed by saying that integral w(z) in D changes to
W (z) after describing the closed curve Γ.
Let w1 , w2 be a fundamental set of integrals (2) in D, the domain of z ∈ N ′ (z0 , r).
1
Let w1 , w2 become W1 , W2 after describing Γ . Since W1 , W2 are solution of (2) in
D, we have
W1 = a11 w1 + a12 w2
(3)
W2 = a21 w1 + a22 w2
in D where aij s are constants. Now W1 , W2 are also linearly independent for if
c1 W1 + c2 W2 = 0 in D , c1 , c2 being constants. Then since c1 W1 + c2 W2 becomes
c1 w1 + c2 w2 after describing −Γ , we get c1 w1 + c2 w2 = 0 in D which implies
c1 = c2 = 0. Thus W1 , W2 also form a fundamental set of integrals in D and since
a11 a12
∆(W1 , W2 ) = ∆(w1 , w2 ) (4)
a21 a22
a11 a12
6= 0
a21 a22
Let us seek a a non-zero solution u in D which change to U = λu (λ is constant)
after describing Γ. If u = αw1 + βw2 in D, α, β constants, then
λu = U = αW1 + βW2 (5)
=⇒ λ(αw1 + βw2 ) = α(a11 w1 + a12 w2 ) + β(a21 w1 + a22 w2 ) in D
=⇒ αw1(a11 − λ) + βa21 w1 + βw2 (a22 − λ) + a12 αw2 = 0 in D
α (a11 − λ) + β a21 = 0
=⇒ [since w1 , w2 linearly independent] (6)
α a12 + β (a22 − λ) = 0
Eliminating α, β , we get
a11 − λ a21
= 0 (7)
a12 a22 − λ
which is a quadratic equation in λ.
Next we assert that this equation is independent of fundamental set of integrals
w1 , w2 which we started with. Let v1 , v2 be another fundamental set of integral in
D which change to V1 , V2 after describing Γ and
V1 = b11 v1 + b12 v2
2
V2 = b21 v1 + b22 v2
Then corresponding to (7) we have
b11 − λ b21
= 0
b12 b22 − λ
Let
v1 = c11 w1 + c12 w2
v2 = c21 w1 + c22 w2
Since w1 , w2 and v1 , v2 are both l.i., we have
c11 c12
6= 0
c21 c22
Then
V1 = c11 W1 + c12 W2
V2 = c21 W1 + c22 W2
or,
b11 v1 + b12 v2 = c11 (a11 w1 + a12 w2 ) + c12 (a21 w1 + a22 w2 )
b21 v1 + b22 v2 = c21 (a11 w1 + a12 w2 ) + c22 (a21 w1 + a22 w2 )
In D, and hence
b11 (c11 w1 + c12 w2 ) + b12 (c21 w1 + c22 w2 ) = c11 (a11 w1 + a12 w2 ) + c12 (a21 w1 + a22 w2 )
b21 (c11 w1 + c12 w2 ) + b22 (c21 w1 + c22 w2 ) = c21 (a11 w1 + a12 w2 ) + c22 (a21 w1 + a22 w2 )
so that
w1 (b11 c11 + b12 c21 − c11 a11 − c12 a21 ) + w2 (b11 c21 + b12 c22 − c11 a12 − c12 a22 ) = 0
w1 (b21 c11 + b22 c21 − c21 a11 − c22 a21 ) + w2 (b21 c12 + b22 c22 − c21 a12 − c22 a22 ) = 0.
Since w1 , w2 are l.i, so
b11 c11 + b12 c21 = c11 a11 + c12 a21
3
b11 c21 + b12 c22 = c11 a12 + c12 a22
b21 c11 + b22 c21 = c21 a11 + c22 a21
b21 c12 + b22 c22 = c21 a12 + c22 a22
Now
c11 c12 a11 − λ a21 c11 (a11 − λ) + c12 a21 c11 a12 + c12 (a21 − λ)
=
c21 c22 a12 a22 − λ c21 (a11 − λ) + c22 a21 c21 a12 + c2 2(a22 − λ)
(b11 − λ)c11 + b12 c21 (b11 − λ)c12 + b12 c22 b11 − λ b12 c11 c12
= =
(b22 − λ)c21 + b21 c11 (b22 − λ)c22 + b21 c12 b21 b22 − λ c21 c22
So
a11 − λ a12 b11 − λ b12
=
a21 a22 − λ b21 b22 − λ
This proves our assertion. Thus equation (7) depends only on singularity z0 and
is called the fundamental equation or characteristic equation to singularity z0 . The
characteristic equation has two roots λ1 , λ2 which may however coincide. By virtue
of (4) λ1 , λ2 6= 0 and corresponding to each value to λ we can determine a set of
values α, β both not zero by (6) and thereby obtain an integral u of desired type by
(5).
Case I : The roots of the characteristic equation are unequal i.e. λ1 6= λ2 . For
λ = λi , let αi , βi both not zero be a set of values of α, β respectively and ui =
αi w1 + βi w2 (i = 1, 2) then u1 , u2 are solutions in D which respectively change
to λ1 u1 and λ2 u2 after description of Γ and u1 , u2 are linearly independent, for if
c1 u1 +c2 u2 ≡ 0 in D where c1 , c2 are constants, then we also have c1 λ1 u1 +c2 λ2 u2 ≡ 0
in D and it follows that λ1 = λ2 , which is untrue.
log λ1 log λ2
Set ρ1 = 2πi
; ρ2 = 2πi
, so that ρ1 , ρ2 are determinate except of additive
integers. Now we know that the function (z −z0 )ρ1 in D changes to e2πiρ1 (z −z0 )ρ1 =
u1
λ1 (z − z0 )ρ1 after description of Γ and hence φ1 = (z−z0 )ρ1
returns to its initial value
after description of Γ, i.e. φ1 is single valued in N ′ (z0 , r) and can be expanded in
Laurent series about z0 in N ′ (z0 , r)
∞
X
φ1 = an (z − z0 )n in N ′ (z0 , r)
n=−∞
4
∞
X
ρ1
=⇒ u1 = (z − z0 ) an (z − z0 )n in N ′ (z0 , r)
n=−∞
Parallelly,
∞
X
ρ2
u2 = (z − z0 ) bn (z − z0 )n in N ′ (z0 , r)
n=−∞
These form a fundamental set of solution of (2) in N ′ (z0 , r).
Case II : The characteristic equation has equal roots, i.e. λ1 = λ2 . In this case
we get only one integral u1 in D which changes to λ1 u1 after describing Γ.
Let u2 be another integral s.t. u1 , u2 form a fundamental set of integrals in D. If
u1 , u2 become U1 , U2 after describing Γ , then
U1 = λ1 u1
U2 = d1 u1 + d2 u2
in D where d1 , d2 are constants. The characteristic equation can be written as
λ1 − λ 0
= 0
d1 d2 − λ
which gives λ = λ1 , d2 and since roots are equal, λ1 = d2 .
Thus we can write
U2 d1 u2
= +
U1 λ1 u 1
We know that the function Log(z − z0 ) changes to Log(z − z0 ) + 2πi after
description of Γ so that the function
u2 d1 Log(z − z0 )
ψ= −
u 1 λ1 2πi
remains unchanged after description of Γ and ψ is single values in N ′ (z0 , r).
Hence we have
∞
X
ρ1
u1 = (z − z0 ) φ1 ; φ1 = an (z − z0 )n as in case-I
n=−∞
ρ1 d1
u2 = (z − z0 ) Log(z − z0 )φ1 + φ1 ψ
2πiλ1
5
∞ ∞
ρ1 d1 X
n
X
n
= (z − z0 ) Log(z − z0 ) an (z − z0 ) + bn (z − z0 )
2πiλ1 n=−∞ n=−∞
where d1 may be zero . Thus without any loss of generality , fundamental set of
solution in N ′ (z0 , r) may be written as
∞
X
ρ1
u1 = (z − z0 ) an (z − z0 )n
n=−∞
∞
X ∞
X
ρ1 n n
u2 = (z − z0 ) A Log(z − z0 ) an (z − z0 ) + bn (z − z0 )
n=−∞ n=−∞
where A = 0 or 1. Thus we have proved the following result .
Theorem 1.11 :
Let z0 be a singularity of (2) and r be the distance of another singularity of the
equation nearest to z0 . Then the fundamental set of solution in N ′ (z0 , r) may be
written in the forms
∞
X
ρ1
w1 = (z − z0 ) an (z − z0 )n
n=−∞
∞
X
w2 = (z − z0 )ρ2 bn (z − z0 )n
n=−∞
or ∞
X ∞
X
ρ1 n n
w2 = (z − z0 ) A Log(z − z0 ) an (z − z0 ) + bn (z − z0 )
n=−∞ n=−∞
where A = 0 or 1.
Any integral of (2) in a deleted neighbourhood of a singularity z0 of the equation is
of the form
∞
X ∞
X
ρ n n
w = (z − z0 ) A Log(z − z0 ) an (z − z0 ) + bn (z − z0 ) where A = 0 or 1
n=−∞ n=−∞
and the number of ρ is determinate except for additive integers.
If now the Laurent expansion contain only finite number of negative powers of (z−z0 )
then the above integral will be reduced to the form
′
w = (z − z0 )ρ φ + ψLog(z − z0 )
6
where φ and ψ are holomorphic in the nbd. of z0 s.t. φ(z0 ), ψ(z0 ) donot vanish
simultaneously and hence ρ′ is exactly determined.
This leads to the following definition .
Definition
Let z0 be a singularity of (2) , then an integral w1 of the equation in a deleted
neighbourhood. N ′ (z0 ) is said to be regular if it can be written in the form
w = (z − z0 )ρ φ + ψLog(z − z0 )
where φ and ψ are holomorphic in N(z0 ) s.t. φ(z0 ), ψ(z0 ) donot vanish simultane-
ously and the number ρ is called the index or exponent at z0 . An integral is said
to be irregular if it is not regular.
Remark: It is possible that one of a fundamental set of integrals in a deleted
neighbourhood of a singularity of equation is regular and the other irregular.
Let us now find the necessary condition that both integrals of fundamental set are
regular.
In this case one integral can be written as
∞
X
ρ1
w1 = (z − z0 ) an (z − z0 )n a0 6= 0
n=0
∞
X
= an (z − z0 )ρ1 +n
n=0
∞
X
w1′ = (ρ1 + n)an (z − z0 )ρ1 +n−1
n=0
∞
X
ρ1 −1
= (z − z0 ) (ρ1 + n)an (z − z0 )n
n=0
∞
X
w1′′ = (ρ1 + n)(ρ1 + n − 1)an (z − z0 )ρ1 +n−2
n=0
∞
X
ρ1 −2
= (z − z0 ) (ρ1 + n)(ρ1 + n − 1)an (z − z0 )n
n=0
w1′ φ w1′′ ψ
= ; =
w1 z − z0 w1 (z − z0 )2
7
where φ and ψ are holomorphic in any neighbourhood of z0 since a0 6= 0, but may
w1′ w1′′
vanish at z0 , i.e. w1
has a pole of order 1 atmost and w1
has a pole of order 2 atmost
at z0 .
R R
qdz
Another integral of a fundamental set can be taken to be w2 = w1 e dz where
2w1′
q = p1 − w1
. If w2 is also regular integral q can have a pole of order one atmost at
R
qdz
z0 otherwise e will have essential singularity at z0 , so that w2 cannot be regular.
2w1′
Hence p1 = q + w1
can have a pole of order 1 atmost at z0 .
w1′′ w′
Since p2 = w1
− p1 w11 , p2 can have a pole of order 2 atmost at z0 .
Theorem 1.12 : A necessary condition for all integrals of differential equation
(14) in a deleted nbd of a singularity z0 of the equation to be regular is that p1 (z)
has a pole of order 1 atmost at z0 and p2 (z) has a pole of order 2 atmost at z0 i.e.
(14) may be written in the form
d2 w P1 (z) dw P2 (z)
2
= + w
dz z − z0 dz (z − z0 )2
where P1 (z), P2 (z) are holomorphic in a nbd of z0 .
Definition A singularity z0 of the differential equation (14) is said to be regular
if z0 is a pole of order 1 atmost of p1 (z) and pole of order 2 atmost of p2 (z), i.e.
(z − z0 )p1 (z) and (z − z0 )2 p2 (z) are holomorphic in a nbd of z0 .
Let us now consider the solution of (14) at infinity. Setting t = z1 , (1.14) trans-
forms to
d2 w
2 1 1 dw 1 1
2
= − + 2 p1 ( ) + 4 p2 ( )w
dt t t t dt t t
Theorem 1.13 :
(i) ∞ is an ordinary point of (14) if 2t + p1 ( 1t ) has a zero of order 2 atleast at zero
and p2 ( 1t ) has a zero of order 4 atleast at zero i.e. 2
z
+ p1 (z) has a zero of order 2
atleast at ∞ and p2 (z) has a zero of order 4 atleast at ∞.
(ii) ∞ is a regular singular point of (14) if 1t p1 ( 1t ) and 1
p (1)
t2 2 t
are holomorphic in a
nbd of zero, i.e. zp1 (z), z 2 p2 (z) are holomorphic at ∞.
8
Definition Equation (14) is said to be of Fuchsian type if all the singularities
of the equation (including ∞ ) are regular.