LECTURE 8

During the last le ture we studied the riti al points of the fun tion:
k;n(t; z; m) =

1i<j k

(ti

Y Y
(ti
tj )2
l=1 i=1

zl )

ml

We studied the a tion on the set of the riti al points of the symmetri group and
we dis ussed the following:
Theorem 0.1.

2k = k +n n 2 2 :
!

dimSingM
m [jmj

and also:

Theorem 0.2. Let mi 2 Z>0 for i = 1; :::; n, k 2 Z>0 and let ! (m = (m1 ; :::; mn ); k)
be the multipli ity of Ljmj 2k in Lm1
:::
Lmn . Then:
1. If jmj +1 k > k, then for generi z all riti al points of k;n(t) are non degenerate
and the riti al set onsists of ! (m; k) orbits.
2. If jmj + 1 k = k, then for any z the fun tion k;n (t) has no riti al points.
3. If 0  jmj + 1 k < k then for generi z the fun tion k;n(t) may have only
non isolated riti al points. Written in symmetri oordinates 1 ; :::; k , the riti al set
onsists of ! (m; jmj + 1 k) straight lines in the spa e C k .
4. If jmj + 1 k < 0, then for any z the fun tion k;n(t) has no riti al points.

Today we will dis uss the relation of the previous theorem (in parti ular the statements 2., 3., and 4.) with the theory of the Fu hsian dierential equations.
1.

Differential Equations with Regular Singular Points.

We will fo us to the ase of se ond order equations:

u00 + p(z )u0 + q (z )u = 0

Let's start with the following:

(1)

Denition 1.1. A point zo 2 C is alled ordinary for the dierential equation if the
fun tions p(z ) and q (z ) are holomorphi at a neighborood of this point.

To study the ase z = 1 put z = 1 :

du
du d du
=
=
2
dz dz d
d
1

and

d2 u
du
d2 u
4
=
 2 + 2 3 :
2
d
d
d

With this hange of variables the equation turns into:

d2 u 2 1 1 du 1 1
+   2 p  d +  4 q  u:
d 2




(2)

So we have the following:

Denition 1.2. The point z = 1 is an ordinary point of the equation if  = 0 is an

ordinary point of the transformed equation, i.e if:


1p 1

2 

and



1q 1


4 

are holomorphi fun tions at  = 0.

Denition 1.3. If zo is alled singular point for the equation if it is not ordinary.
Denition 1.4. The point zo is alled regular singular if the following onditions are
veried:
1) zo is singular;
2) at zo the fun tion p(z ) at has a pole of order  1;
3) at zo the fun tion q (z ) has a pole of order  2.
Remark 1.1. We have analogous denitions for the point z = 1 and lo al oordinate
z = 1 .

In the following two se tion we will dis uss the solutions of the equation at an
oridinary point and at a singular point.
2.

Solutions at a neighborood of an ordinary point.

Let zo an oridinary point, so we an write:

p( z ) =

and

q (z ) =

n=0

n=0

b n (z

zo )n

dn (z

zo ) n

for jz zo j < R. We look for solutions of the form:

u(z ) =

n=0

a n (z

Under these hypothesis we have the following:

zo )n :

Theorem 2.1. For any hoi e of a0 ; a1 there exist a unique series

1

u(z ) =

n=0

an (z

zo )

with a non zero radius of onvergen e, satisfying the dierential equation:

u00 + p(z )u0 + q (z )u = 0:
00
0
Prof. Plugging u(z) = 1
n=0 an (z zo ) into the equation u + p(z )u + q (z )u = 0 we
get:
for n = 0, 0a0 = 0, for n = 1, 0a1 + 0a0b0 = 0, while for n  2,
P

n(n 1)an +

3.

nX1
k=0

kak bn

1+

nX2
k=0

ak bn

2:

Solutions at a neighborood of a regular singular point.

Let's assume zo be a regular singular point so that the oe ients of the equation
an be written in the following way:
p( z ) =

and

q (z ) =

b n (z

zo )n 1

dn(z

zo )n 2 ;

n=0

n=0

for jz zo j < R. In this ase we look for solutions of the form:

u(z ) =

n=0

an (z

zo )n+

where a0 6= 0 and is a ostant that has to be determined. Plugging the fun tion
u(z ) = 1
zo )n+ into the equation, we get the following equation:
n=0 an (z
P

n=0

(n + )(n + 1)an(z z0 )n+ 2 +



1
X

n=0


pn (z z0 )n 1
1
X



n=0


an (z z0 )n+ = 0:
+ q n ( z z0 ) n 2
n=0
n=0
From this equation we get the following re urren e relations:
for n = 0:
[ ( 1) + p0 + q0a0 = 0;
while for n  1:

[(n + )(n + 1) + (n + )p0 + q0 an =

Sin e a0 6= 0, we have:

2 + (p0

(n + )an(z z0 )n+ 1 +

nX1
k=0

1) + q0 = 0:

ak [(k + )pn k + qnk :

(3)
(4)
(5)

Denition 3.1. The equation 2 + (p0 1) + q0 = 0 is alled indi ial equation, while
the roots 1 and 2 are alled the exponents of the equation at the point z = z0 .

Let's write F ( ) = 2 + (p0 1) + q0 so that we have:

nX1

F (n + )an =

k=0

ak [(k + )pn k + qnk

for n  1, and F ( ) = 0, for the indi ial equation. Let's order the roots 1; 2 of the
polynomial F ( ) in su h a way <e 1  <e 2 and set s = 1 2. Now if we start with
= 1 , we get:
F (n + 1 ) = n(n + 1 2 ) = n(n + s)
from whi h we see that the re urren e equations are satised; so we an state the
following:
Theorem 3.1. If z0 is a regular singular point with exponentials 1 and 2 su h that
<e 1  <e 2 , then the series:
u 1 (z ) = (z

z0 ) 1 +

n=1

an (z

z0 )n+ 1

(6)

satises the equation and it is onvergent in a region of the form 0 < jz

Question 3.1. How to nd the se ond solution?

z0 j < r.

The answer to this question depends on s = 1 2 . Put = 2, then the equation

for n = 0 is satised; for n  1 we get:
n(n s)an =

nX1
k=0

qk (k + 2 )pn k + qn

Theorem 3.2. If s 2= Z then there exists a series:

u 2 (z ) = (z

z0 ) 2 +

n=1

an (z

z0 )n+ 1

that satises the equation and onverges in a region of the form 0 < jz

(7)
z0 j < r:

In the ase s = 0 we don't get any new solution, while if s 2 Z>0, for n = s, we get:
0=

sX1

k=0

qk (k + 2 )ps k + qs

This gives a non trivial onstraint: given arbitrary a0 and as we an get:

and

u 1 (z ) = (z

z0 ) 1 +

u 2 (z ) = (z

z0 ) +
2

n=1

n=1

an (z

z0 )n+ 1

an (z

z0 )n+ 2

if a0 and as authomati ally satisfy the re urren e relation or:

u2 (z ) = u1 ln(z

z0 ) +

if not.
4.

n=0

d n (z

z0 )n+ 2

Fu hsian Equations.

Let's start with the following:

Denition 4.1. [Lazarus Fu hs (1833-1902) A linear dierential equation is alled
fu hsian if all its singular points are regular singular.

Remark 4.1. During this se tion we will fo us only on se ond order dierential equations.
Remark 4.2. If the equation:
u00 + p(z )u0 + q (z )u = 0

is fu hsian, then the oe ients q (z ) and p(z ) are rational fun tions.

We also have the following:

Proposition 4.1. A fu hsian equation has at least a singular point.

In what follows we will give a des ription of the se ond order fu hsian dierential
equations with one two and three singular points. Let's start with the rst ase. Let's
assume that u00 + p(z)u0 + q(z)u = 0 has only one singular point at z = z1 , then:
Proposition 4.2. The equation has the form:

2 u 0 = 0:
z z1
and its solutions are of the form u(z ) = a + z dz1 for given a and b.
Prof. From the hypothesis we have that:
p (z )
p( z ) = 1
u00 +

and

z1

q1 (z )
(z z1 )2
where p1 (z) and q1(z) are polynomials. Sin e z = 1 is an ordinary point we have:
2
2
2
2z z2 p(z) = 2z zz p1(zz) = 2z zz p1z 2z1 z
1
1

and

q (z ) =

z 4 q (z ) =

z 4 q1 (z )
(z z1 )2

(8)

are holomorphi at z = 1. From the last equation we dedu e that q1  0 (otherwise

degz4 q1 (z)  4) while from the rst one follows that p1 = 2. Clearly u1(z) = 1 and
u2 (z ) = z 1z1 are solutions.
We also have the following result:
Theorem 4.1. If z = 1 is the only singular point of a fu hsian dierential equation,
then the equation is:

u00 = 0:

(9)

Prof. In this ase the oe ients p(z) and q(z) are two polynomials. So we have
that the fun tion 2z z2 p(z) has poles of order  1 while the fun tion z4 q(z) has poles
of order  2. From this observation follows that p(z) = q(z) = 0.
Now let's dis uss the ase of fu hsian equations with two singular points z1 and z2 .
In this ase we have two dierent ases:
1) z2 = 1, or:
2) z1 and z2 2 C .
In the rst ase we have:

Proposition 4.3. The fu hsian equation an be written as:

k2
k
u00 + 1 u0 +
z z1
(z z1 )2 u = 0;
with k1 ; k2 2 C . Moreover the solutions are:
u(z ) = A(z z1 ) 1 + B (z z1 ) 2
if 1 6= 2 , or:


u(z ) = (z z1 ) A + B ln(z z1 )
if 1 = 2 = .

(10)
(11)
(12)

For the se ond ase, when both z1 and z2 belong to C , it is possible to prove the
following:
Proposition 4.4. The fu hsian equation has the form:
2z + k3 u0 +
k4
u00 +
(z z1 )(z z2)
(z z1)2 (z z2 )2 u = 0:

Let's now start to dis uss the ase of three singular point.

(13)

Remark 4.3. We have the following general:

Theorem 4.2. A se ond order fu hsian equation with n +1 singular points z1 ; :::; zn ; 1
has the form:
T 2n 2
Tn 1
u0 +
(14)
u00 +
(z z1):::(z zn)
(z z1)2 :::(z zn )2 u = 0;
where Tn 1 and T2n 2 are polynomials of degree less or equal to n 1 and 2n 2.

If we all 1;k and 2;k the exponents of the equation at the point z = zk and 1;1
and 2;1 the exponents at the point z = 1, then:
Theorem 4.3. [Fu hsian Invariants.
1;1 + 2;1 +

n
X

(1;k + 2;k ) = n 1:

(15)

k=1

Let's fo us on the ase of three singular points z1 , z2 and 1. In this ase the
exponents are: (1;1; 2;1) (1;1; 2;1) (1;2; 2;2) and we have:
1;1 + 2;1 +

(1;k + 2;k ) = 1:

(16)

k=1

Claim 4.1. the fu hsian equation an be written in the following way:


1
1;1 2;1 1 1;2 2;2  0
00
+ z z
u+
u +
z z1
2

1;1 2;1
2;1 2;2 1;1 2;1 1;1 2;1 1;2 2;2 
u=0
(z z1 )2 + (z z2 )2 +
(z z1)(z z2)
Remark 4.4. The previous equation is the form of the general Fu hsian equation of the
se ond order and with three regular singular points written in terms of the exponents.

Using the notation z = a; b; for the singular points (all of them dierent by 1) and
orrespondents exponents, we have:

(a0; a00), (b0 ; b00), ( 0; 00) for the

Theorem 4.4. The previous equation an be written as:


1
a0 a00 1 b0 b00 1 0 00  0
00
u +
+ z b + z u+
z a
 0 00
u
a a (a b)(a ) b0 b00 (b a)(b ) 0 00 ( a)( b) 
+
+
z
z b
z
(z a)(z b)(z
Remark 4.5. This equation is alled Riemann-Papperitz equation.
Denition 4.2. [Riemann P-symbol.
Let's write:
0

u=PB


a b
a0 b0 0 z
a00 b00 00

= 0:

1
C
A

(17)

This is alled the Riemann P-symbols. This denotes a solution of a Fu hsian dierntial equation with singular points at a; b; (distin t but an be 1) and exponentials
(a0; a00), (b0 ; b00), ( 0; 00).
Remark 4.6. Under these hypothesis we have:
a0 + a00 + b0 + b00 + 0 + 00 = 1:
Remark 4.7. The informations given in the P-symbol determines the equation uniquely.

Theorem 4.5. Let's suppose a; 2 C and:

0
a b
0 b0 0 z
u=PB
 a
00
a b00 00
If

z a k
u=
w
z
then:
0

u=PB

Theorem 4.6. If:

a)k then:
u=PB


Theorem 4.7. If

a b 1
a0 b0 0 z
a00 b00 00

u=PB


C
A

(19)

(20)

C
A

b 1
a0 k b0 0 + k z
a00 k b00 00 + k
a b
a0 b0 0 z
a00 b00 00

(18)

b

a0 k b0 0 + k z
a00 k b00 00 + k
0

C
A

u=PB

and u = (z

1
C
A

(21)

1
C
A

the fra tional transformation:

(22)

Az + B
v=
Cz + D
with AD CB 6= 0 maps z = a; b; into v = a1 ; b1 ; 1 and the solution u an be witten
in terms of v in the following way:
1
0
a1 b1 1
0 b0 0 v C
(23)
u=PB
A
 a
00
00
00
a b

We also have the following general:

Lemma 4.1. Any linear fra tional transformation:

Az + B
v=
Cz + D
with AD BC 6= 0 is the omposition of elementary transformations:
v = z , v = + z and v = z1 .

We have the following important:

Theorem 4.8. Any P-symbols an be redu ed to a P-symbols of the form:

0
0

y=PB


1
0

 xC
A
1   
Remark 4.8. The orrisponding dierential equation is the following:
d2 y  1 (1 ) 1 (   )  dy

+
+
+
y = 0:
2
dx
x
x 1
dx x(x 1)
Multiplying both sides by x(x 1) we get:
x(x 1)y 00 (x) + [

( +  + 1)xy0 y = 0;

(24)

(25)
(26)

i.e. the Gauss dierential equation.

Let's start the dis ussion the onne tione of the theory of the Fu hsian dierential
equations with the riti al points of the master fun tion. Let's onsider the following:
F (x)u00 (x) + G(x)u0 (x) + H (x)u(x) = 0
(27)
where F , G and H are polynomials of degree n, n 1 and n 2 respe tively.
Claim 4.2. If the polynomial F has no multiple roots then the previous equation is
Fu hsian.

In this ase let's write:

F (x) =

n
Y

(x zj )

j =1
n
X

G(x)
mj
=
F (x) j =1 x zj
for suitable omplex numbers mj and zj .
Claim 4.3. It is possible to prove that 0 and mj +1 are exponents of the above equation
and if k is one of the exponents at z = 1 then the other is k l(m) 1.
Problem 4.1. Given polynomials F (x) and G(x) as above:
1. nd a polynomial H (x) of degree at most n 2 su h that the equation has a polynomial
solution of a preassigned degree k;
2. nd the number of solution of 1.

we have the following lassi al results (referen e: Szego G., "Orthogonal Polynomials"):
Theorem 4.9. i. Let u(x) a polynomial solution of degree k of the equation, with roots
t01 ; :::; t0k of multipli ity one. Then t0 = (t01 ; :::; t0k ) is a riti al point of the fun tion
k;n(t; z; m), where z = (z1; :::; zn) and m = (m1 ; :::; mn).

10

ii. Let t0 be a riti al point of the fun tion k;n (t; z; m), then the polynomial u(x) =
(x t01):::(x t0k ) of degree k is a solution of the equation with


F (x)u00 (x) G(x)u0 (x)

H (x) =
u(x)
being a polynomial of degree at most n 2.
Remark 4.9. Re all that the riti al points of the fun tion k;n(t; z; m) are given by:
n
X
mj X 2
+
; i = 1; :::; k:
(28)
zj j 6=i ti tj
j =1 ti
Remark 4.10. This an be seen as an equivalent version of the Bethe ansazt problem.

Prof.(of part i.). Let u(x) = (x t01 ):::(x t0k ) be a solution of the equation:
F u00 + Gu0 + Hu = 0;
and set x = t0i ; then:
F (t0i )u00 (t0i ) + G(t0i )u0 (t0i ) + H (t0i )u(t0i ) = 0:
From this we an dedu e:
u00 (t0i ) G(t0i )
+
= 0;
u0(t0i ) F (t0i )
but by denition we have:
G(t0i )
mi
=
:
0
0
F (ti )
ti zi
To on lude we need the following:
Lemma 4.2. If u(x) = (x t01 ):::(x t0k ), then:
2 :
u00 (t0i )
=
(29)
0
0
0
u (ti ) j 6=i ti t0j
X

Example 4.1. If:

then:

u(x) = (x t01 )(x t02 )

u(x) = (x t01 ) + (x t02 );
u00 = 2

so:

u00 0
2
(
t1 ) = 0 0
0
u
t1 t2

This give us part i. of the previous theorem.

11

Remark 4.11. The riti al points label solutions of the problem 2., and a riti al point
denes a dierential equation and a polynomial solution of that equation.
Remark 4.12. In 19-th entury Heine and Stieltjes showed that for xed real z1 ; :::; zn
and negative m1 ; :::; mn the problem 2. has:
!
k+n 2
n 2
solutions (referen e: Szego G., "Orthogonal Polynomials").

Let m1 ; :::; mn 2 Z>0 and z 2 Z>0, su h that jmj 2k  0, then k < jmj k + 1.
Assume that t0 is a riti al point of k;n(t; z; m), then it is possible to onstru t a line
of riti al points of the fun tion jmj k+1;n(t; z; m).
Example 4.2. Let m = (1; 1; 1) and
1;3 (t; z; m) = (t z1) 1 (t z2) 1 (t z3 ) 1:

it is possible to onstru t a line of riti al points of the fun tion: Given the two riti al
points of the fun tion 1;3 (t; z; m) we get two lines of riti al points of the fun ton.

3;3 (t; z; m) =

3 3

Y Y

(ti zl )

i=1 l=1

(ti tj )2 :

1i<j 3

Moreover it is possible to show that for generi z1 ; z2 ; z3 the points on those two lines
are the only riti al points.

5.

Enumerative Algebrai Geometry.

It is possible to reformulate the main theorem in terms of Enumerative Algebrai

Geometry. Let V a two dimensional spa e of polynomials of one variable with oe ients in C . Let k1 the degree of a typi al polynomial and k2 the degree of a spe ial
polynomial (k1 < k2)
Denition 5.1. Given two fun tions f (x) and g (x) let's denote with:



W (f; g ) =  ff0 gg0

(30)

their Wronskian.
~ g~) are two basis of V then there exits a ostant su h
Remark 5.1. If (f; g ) and (f;
that:
~ g~):
W (f; g ) = W (f;
Lemma 5.1. If degf = k1 and degg = k2 then degW (f; g ) = k1 + k2 1.

Prof. This follows from dire t al ulation:

k1
::
xk2 + ::
k1 +k2 1
W (f; g ) = k xxk1 +1 +
+ :::
:: k xk2 1 + :: = (k2 k1 )x





(31)

12

Let WV = xk1 +k2 1 + ::: the Wronskian of V and let's suppose that WV (x) =
zl )ml , where ml  1.

Denition 5.2. We will say that V is not singular if 8x0

that f (x0 ) 6= 0.

Now we an state the following:

2C

there is f

Qn

l=1

2V

(x

su h

Problem 5.1. Assume that k1 < k2 and WV (x) = nl=1 (x zl )ml are xed. What is
the number of non singular ve tor spa es V of polynomials with su h data?
Q

It is possible to get an answer to question in terms of representation of sl(2):

Theorem 5.1. For generi z1 ; :::; zn the numeber of non singular ve tor spa es with
su h data is equal to:
multLk1 k1 1  Lm1
:::
Lmn :