A Distant Mirror:: My Joint Paper With Aad After 35 Years

A distant mirror:
my joint paper with Aad after 35 years

Tom Koornwinder
thk@science.uva.nl
Korteweg-de Vries Institute, University of Amsterdam
lecture on 24 February 2006

during the Dijksma Conference, Groningen,
23–24 February 2006
A distant mirror: – p.1/27

my 1971 paper with Aad
A. Dijksma & T. H. Koornwinder,
Spherical harmonics and the product of two Jacobi
polynomials,
Proc. KNAW, Series A, 74 = Indag. Math. 33
(1971), 191–196
(communicated by prof. A. C. Zaanen)

Adriaan Zaanen, 1913–2003

Jacobi polynomials
(α,β)
Jacobi polynomials Pn (x) are orthogonal polynomials
on [−1, 1] with respect to the measure
(1 − x)α (1 + x)β dx (α, β > −1),
and with normalization
(α,β) (α + 1)n
Pn (1) = ,
n!
where (a)k := a(a + 1) . . . (a + k − 1).

Jacobi polynomials
(α,β)
Jacobi polynomials Pn (x) are orthogonal polynomials
on [−1, 1] with respect to the measure
(1 − x)α (1 + x)β dx (α, β > −1),
and with normalization
(α,β) (α + 1)n
Pn (1) = ,
n!
where (a)k := a(a + 1) . . . (a + k − 1).
Gegenbauer polynomials are special Jacobi polynomials
(2λ)n (λ− 21 ,λ− 12 )
Cnλ (x) := 1 Pn (x).
(λ + 2 )n

Dijksma-Koornwinder product formula
(α,β) (α,β)
2
Pn (1 − 2x ) Pn (1 − 2y 2 )
Z 1 Z 1 p
α+β+1
p
= const. C2n xyu + 1 − x2 1 − y2 v
−1 −1
2 α− 21 2 β− 21
× (1 − u ) (1 − v ) du dv (α, β > − 12 ).

Dijksma-Koornwinder product formula
(α,β) (α,β)
2
Pn (1 − 2x ) Pn (1 − 2y 2 )
Z 1 Z 1 p
α+β+1
p
= const. C2n xyu + 1 − x2 1 − y2 v
−1 −1
2 α− 21 2 β− 21
× (1 − u ) (1 − v ) du dv (α, β > − 12 ).
For α, β = 0, 12 , 1, 32 , . . . interpretation on spherical harmonics.

Spherical harmonics
Unit sphere S d−1 in Rd .
O(d) is group of orthogonal transformations of Rd .
O(d) acts on functions on S d−1 : (T f )(x) := f (T −1 x).

Spherical harmonics
Spherical harmonic of degree n on S d−1 is restriction to S d−1
of harmonic homogeneous polynomial of degree n on Rd .
Hn is the space of spherical harmonics of degree n on S d−1 .

Spherical harmonics
f ∈ Hn =⇒ T f ∈ Hn .
The representation of O(d) on Hn is irreducible.

Spherical harmonics
f ∈ Hn =⇒ T f ∈ Hn .
The representation of O(d) on Hn is irreducible.
O(d − 1) is the subgroup of O(d) which leaves
e1 = (1, 0, . . . , 0) fixed.
S d−1 = O(d)/O(d − 1).

Zonal spherical harmonics
Theorem f ∈ Hn and O(d − 1)-invariant
1
d−1
iff f (x) = const. Cn
2
(hx, e1 i).

1
d−1
2
(hx, e1 i).
Z
Lemma If g ∈ Hn and f (x) := g(T x) dT
O(d−1)
then f ∈ Hn and O(d − 1)-invariant.

1
d−1
2
(hx, e1 i).
Z
O(d−1)
Haar measure

1
d−1
2
(hx, e1 i).
Z
O(d−1)
Corollary 1 Take g(x) := (x1 + ix2 )n .

1
Z
d−1
Cn2 (cos θ) = const. (cos θ e1 + i sin θ (T e2 )2 )n dT
O(d−1)
Z 1
1
n 2 2 d−2
= const. (cos θ + i sin θ v) (1 − v ) dv.
−1

Zonal spherical harmonic: product formula
1
d−1
Corollary 2 Take g(x) := Cn 2
(hx, yi) (y ∈ S d−1 ).
1 1
Z 1
d−1 d−1 d−1
Cn
2
(hx, e1 i) Cn
2
(hy, e1 i) = const. Cn
2
(hT x, yi) dT,
O(d−1)
1 1
d−1 d−1
Cn2
(cos θ1 ) Cn 2
(cos θ2 ) = const.
Z 1 1
d−1 1
× Cn2 (cos θ1 cos θ2 + sin θ1 sin θ2 v) (1 − v ) 2 d−2 dv.
2
−1

Zonal spherical harmonic: product formula
1
d−1
Corollary 2 Take g(x) := Cn 2
(hx, yi) (y ∈ S d−1 ).
1 1
Z 1
d−1 d−1 d−1
Cn
2
(hx, e1 i) Cn
2
(hy, e1 i) = const. Cn
2
(hT x, yi) dT,
O(d−1)
1 1
d−1 d−1
Cn2
(cos θ1 ) Cn 2
(cos θ2 ) = const.
Z 1 1
d−1 1
× Cn2 (cos θ1 cos θ2 + sin θ1 sin θ2 v) (1 − v ) 2 d−2 dv.
2
−1
Analytic continuation w.r.t. d by Carlson’s theorem.

For λ > 0:
Z 1 p p
λ λ
Cn (x) Cn (y) = const. Cn (xy+ 1 − x2 1 − y 2 v) (1−v 2 )λ−1 dv.
λ
−1

Mathematisch Centrum
2e Boerhaavestraat, Amsterdam

O(d′ ) × O(d′′ )-invariant spherical harmonics

′ ′′
d= d′ + d′′ , Rd = Rd d
×R , x = (x′ , x′′ ).
!
T′ 0
(T ′ , T ′′ ) = ∈ O(d′ ) × O(d′′ ) ⊂ O(d).
0 T ′′

′ ′′
d= d′ + d′′ , Rd = Rd d
×R , x = (x′ , x′′ ).
!
T′ 0
(T ′ , T ′′ ) = ∈ O(d′ ) × O(d′′ ) ⊂ O(d).
0 T ′′
Theorem (Braaksma & Meulenbeld, 1968)

f ∈ H2n and O(d′ ) × O(d′′ )-invariant
( 21 d′′ −1, 21 d′ −1)
iff f (x) = const. Pn (2|x′ |2 − 1).

′ ′′
d= d′ + d′′ , Rd = Rd d
×R , x = (x′ , x′′ ).
!
T′ 0
(T ′ , T ′′ ) = ∈ O(d′ ) × O(d′′ ) ⊂ O(d).
0 T ′′
Theorem (Braaksma & Meulenbeld, 1968)

f ∈ H2n and O(d′ ) × O(d′′ )-invariant
( 21 d′′ −1, 21 d′ −1)
iff f (x) = const. Pn (2|x′ |2 − 1).
Z
Lemma If g ∈ H2n and f (x) := g(T x) dT
O(d′ )×O(d′′ )
then f ∈ H2n and O(d′ ) × O(d′′ )-invariant.

O(d′ ) × O(d′′ )-invariant spherical harmonics: integral representation
Corollary 1 (Braaksma & Meulenbeld, 1968)

Take g(x) := (x1 + ixd )2n .
( 21 d′′ −1, 21 d′ −1)
Pn (2|x′ |2 − 1)
Z Z
′ ′ ′′ ′′
2n
= const. (T x )1 + i(T x )d dT ′ dT ′′ ,
O(d′ ) O(d′′ )

O(d′ ) × O(d′′ )-invariant spherical harmonics: integral representation
Corollary 1 (Braaksma & Meulenbeld, 1968)

Take g(x) := (x1 + ixd )2n .
( 21 d′′ −1, 21 d′ −1)
Pn (2|x′ |2 − 1)
Z Z
′ ′ ′′ ′′
2n
= const. (T x )1 + i(T x )d dT ′ dT ′′ ,
O(d′ ) O(d′′ )
( 12 d′′ −1, 21 d′ −1)

Pn (cos 2θ)
Z 1 Z 1
= const. (cos θ u + i sin θ v)2n
−1 −1
1 1
2 2 d′ − 23 2 2 d′′ − 23
× (1 − u ) (1 − v ) du dv.

O(d′ ) × O(d′′ )-invariant spherical harmonics: product formula
Corollary 2 (Dijksma & K, 1971)

1
d−1
Take g(x) := C2n 2
(hx, yi) (y ∈ S d−1 ).
( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)
Pn (2|x′ |2 − 1)Pn (2|y ′ |2 − 1)
Z Z 1
d−1
= const. C2n
2
(hT ′ x′ , y ′ i + hT ′′ x′′ , y ′′ i) dT ′ dT ′′ ,
O(d′ ) O(d′′ )


1
d−1
Take g(x) := C2n 2
(hx, yi) (y ∈ S d−1 ).
( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)
Pn (2|x′ |2 − 1)Pn (2|y ′ |2 − 1)
Z Z 1
d−1
= const. C2n
2
(hT ′ x′ , y ′ i + hT ′′ x′′ , y ′′ i) dT ′ dT ′′ ,
O(d′ ) O(d′′ )
( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)

Pn (cos 2θ1 )Pn (cos 2θ2 )
Z 1 Z 1 1
d−1
= const. C2n
2
(cos θ1 cos θ2 u + sin θ1 sin θ2 v)
−1 −1
1 1
2 2 d′ − 23 2 2 d′′ − 23
× (1 − u ) (1 − v ) du dv.


1
d−1
Take g(x) := C2n 2
(hx, yi) (y ∈ S d−1 ).
( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)
Pn (2|x′ |2 − 1)Pn (2|y ′ |2 − 1)
Z Z 1
d−1
= const. C2n
2
(hT ′ x′ , y ′ i + hT ′′ x′′ , y ′′ i) dT ′ dT ′′ ,
O(d′ ) O(d′′ )
( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)

Pn (cos 2θ1 )Pn (cos 2θ2 )
Z 1 Z 1 1
d−1
= const. C2n
2
(cos θ1 cos θ2 u + sin θ1 sin θ2 v)
−1 −1
1 1
2 2 d′ − 23 2 2 d′′ − 23
× (1 − u ) (1 − v ) du dv.
Analytic continuation w.r.t. d′ , d′′ by Carlson’s theorem.

Gasper’s product formula (1971)
(α,β) (α,β) 1 (α,β)

Pn (x) Pn (y) Pn (z)
Z
(α,β) (α,β)
= (α,β)
Kα,β (x, y, z) dz
Pn (1) Pn (1) −1 Pn (1)
for an explicit kernel Kα,β (x, y, z) ≥ 0 on [−1, 1]3 if

α ≥ β ≥ − 21 .

Gasper’s product formula (1971)
(α,β) (α,β) 1 (α,β)

Pn (x) Pn (y) Pn (z)
Z
(α,β) (α,β)
= (α,β)
Kα,β (x, y, z) dz
Pn (1) Pn (1) −1 Pn (1)
for an explicit kernel Kα,β (x, y, z) ≥ 0 on [−1, 1]3 if

α ≥ β ≥ − 21 .
For α = β this can be rewritten as the original product
formula for Gegenbauer polynomials.
Question by Richard Askey

Is a similar rewriting of Gasper’s product formula possible if
α 6= β ?

Dick Askey

Complex spherical harmonics
Answer to Askey’s question is yes.
One way is by using analogues of spherical harmonics on
U (d)/U (d − 1) (K, 1972, independently from earlier work by
Vilenkin and Šapiro).

Definition A harmonic of type (m, n) on S 2d−1 ⊂ Cd is

the restriction to S 2d−1 of a polynomial homogeneous of
degree m in z1 , . . . , zd and homogeneous of degree n in
∂2 ∂2
z1 , . . . , zd which is annihilated by ∂z1 ∂z1 + · · · + ∂zd ∂zd .
Let Hm,n be the space of all such harmonics.

Definition A harmonic of type (m, n) on S 2d−1 ⊂ Cd is

the restriction to S 2d−1 of a polynomial homogeneous of
degree m in z1 , . . . , zd and homogeneous of degree n in
∂2 ∂2
z1 , . . . , zd which is annihilated by ∂z1 ∂z1 + · · · + ∂zd ∂zd .
Let Hm,n be the space of all such harmonics.
U (d) is the group of unitary d × d matrices. U (d − 1) is the
subgroup which leaves the vector e1 = (1, 0, . . . , 0) in Cd
fixed. Then S 2d−1 = U (d)/U (d − 1).
U (d) acts irreducibly on Hm,n .
Zonal complex spherical harmonics
Theorem f ∈ Hm,n and U (d − 1)-invariant iff
|m−n| (d−2,|m−n|)
f (reiθ , z2 , . . . , zd ) = r e i(m−n)θ Pm∧n (cos 2θ).

|m−n| (d−2,|m−n|)
Disk polynomials:
(α) (α,|m−n|)
Rm,n (reiθ ) := const. r|m−n| ei(m−n)θ Pm∧n (cos 2θ).
(α)
The polynomials Rm,n (z) are orthogonal on the unit disk in C
with respect to the weight function (1 − x2 − y 2 )α
(z = x + iy ).

|m−n| (d−2,|m−n|)
Disk polynomials:
(α) (α,|m−n|)
Rm,n (reiθ ) := const. r|m−n| ei(m−n)θ Pm∧n (cos 2θ).
(α)
The polynomials Rm,n (z) are orthogonal on the unit disk in C
with respect to the weight function (1 − x2 − y 2 )α
(z = x + iy ).
Essentially Zernike polynomials.

Frits Zernike (1888–1966)

Zernike polynomials
http://bifano.bu.edu/tgbifano/Web/Images/zernike_bar.jpg
Product formula for Jacobi polynomials
From interpretation on U (d)/U (d − 1) a product and addition
formula for disk polynomials.
Hence a product and addition formula for Jacobi polynomials
(α,0)
Pn (x).
Hence by differentiation and analytic continuation an addition
(α,β)
formula for Jacobi polynomials Pn (x).

Product formula for Jacobi polynomials
From interpretation on U (d)/U (d − 1) a product and addition
formula for disk polynomials.
Hence a product and addition formula for Jacobi polynomials
(α,0)
Pn (x).
Hence by differentiation and analytic continuation an addition
(α,β)
formula for Jacobi polynomials Pn (x).
(α,β)
Hence product formula for Jacobi polynomials Pn (x)
(α > β > − 12 ):
(α,β) (α,β)
Pn (cos 2θ1 ) Pn (cos 2θ2 ) = const.
Z 1 Z π
(α,β)
× Pn 2| cos θ1 cos θ2 + sin θ1 sin θ2 reiφ |2 − 1
r=0 φ=0
× (1 − r2 )α−β−1 r2β+1 (sin φ)2β dr dφ.

General theory: spherical functions
G compact group, K closed subgroup.
π irreducible unitary representation of G on Hilbert space H
of finite dimension dπ .

Suppose there is a K -fixed unit vector eK in H, unique up to
a constant factor.
Then φ(x) := hπ(x)eK , eK i is called a spherical function.

Suppose there is a K -fixed unit vector eK in H, unique up to
a constant factor.
Then φ(x) := hπ(x)eK , eK i is called a spherical function.
Properties
φ(k1 xk2 ) = φ(x) (k1 , k2 ∈ K ), φ(e) = 1, |φ(x)| ≤ 1.
Theorem (Gelfand)
Let φ be continuous, not identically zero. Then φ is a
spherical function for some irreducible representation iff
Z
φ(x)φ(y) = φ(xky) dk (x, y ∈ G).
K
General theory: intertwining functions
G compact group with closed subgroups K and H .
π irreducible unitary representation of G on H.
K -fixed resp. H -fixed unit vectors eK , eH in H, unique up to
constant factors.
φKK (x) := hπ(x)eK , eK i,

φHH (x) := hπ(x)eH , eH i,
φHK (x) := hπ(x)eK , eH i.
φHK is called an intertwining function.

φHK (e) = heK , eH i, |φHK (x)| ≤ 1, |φHK (e)| ≤ 1.
φKH (x) = φHK (x−1 ).

Intertwining property of φHK
Put
fv,K (x) := hπ(x)eK , vi,

fv,H (x) := hπ(x)eH , vi.
Then Z
fv,H (y) φHK (y −1 x) dy = d−1
π fv,K (x),
G
so
fv,H ∗ φHK = d−1
π fv,K .
φH,K is a convolution kernel which maps the representation

space of π in L2 (G/H) onto the representation space of π in
L2 (G/K), and which intertwines with the G-action.
Projection and product formulas
Z
π(h)eK dh = heK , eH i eH ,
ZH
φKK (h) dh = |heK , eH i|2 ,
H
Z
φKK (xh) dh = φHK (e) φKH (x),
H
Z
φKK (xhy) dh = φKH (x) φHK (y)
H
= φHK (x−1 ) φHK (y).

The case of the Dijksma-K product formula
G = O(d), K = O(d − 1), H = O(d′ ) × O(d′′ ) (d = d′ + d′′ ).
Assume d′ ≤ d′′ . Let A be the one-parameter subgroup of G
consisting of:
 
cos θ Id′ − sin θ Id′ 0
aθ :=  sin θ Id′ cos θ Id′ 0  . Then G = KAH .
 
0 0 Id′′ −d′

The case of the Dijksma-K product formula
G = O(d), K = O(d − 1), H = O(d′ ) × O(d′′ ) (d = d′ + d′′ ).
Assume d′ ≤ d′′ . Let A be the one-parameter subgroup of G
consisting of:
 
aθ :=  sin θ Id′ cos θ Id′ 0  . Then G = KAH .
 
0 0 Id′′ −d′
( 21 d′′ −1, 21 d′ −1)

Pn (cos 2θ)
φHK (Kaθ H) = heK , eH i .
( 12 d′′ −1, 21 d′ −1)
Pn (1)
Hence |φHK (x)| ≤ |φHK (e)|.
( 21 d′′ )n ( 12 )n
|heK , eH i|2 = .
( 12 d′ )n ( 21 (d − 1))n
The Flensted-Jensen condition (1973)
G = O(d), K = O(d − 1), H = O(d′ ) × O(d′′ ), d′ ≤ d′′ ,
 
aθ =  sin θ Id′ cos θ Id′ 0  ∈ A.
 
0 0 Id′′ −d′

The Flensted-Jensen condition (1973)
G = O(d), K = O(d − 1), H = O(d′ ) × O(d′′ ), d′ ≤ d′′ ,
 
aθ =  sin θ Id′ cos θ Id′ 0  ∈ A.
 
0 0 Id′′ −d′
Let (T ′ , T ′′ ) ∈ H . Observe:
 
! T ′ 0 0
T ′ 0
aθ ⊂ aθ  0 T
 ′ 0 K

0 T ′′
0 0 Id′′ −d′
 
′
T 0 0
=0 T
 ′ 0  aθ K ⊂ Haθ K.

0 0 Id′′ −d′
Hence aH ⊂ HaK for all a ∈ A

Flensted-Jensen condition: product formula
Theorem (Flensted-Jensen, 1973) If aH ⊂ HaK then
φHK (ahy) φHK (a) φHK (y)

Z
dh = (y ∈ G).
H φHK (e) φHK (e) φHK (e)

Flensted-Jensen condition: product formula
Theorem (Flensted-Jensen, 1973) If aH ⊂ HaK then
φHK (ahy) φHK (a) φHK (y)

Z
dh = (y ∈ G).
H φHK (e) φHK (e) φHK (e)
Corollary (K, 1973)

An alternative derivation of the product formula for Jacobi
polynomials.

Citations of the Dijksma-K paper
1. Yuan Xu, Orthogonal polynomials for a family of product
weight functions on the spheres, Canad. J. Math. 49
(1997), 175–192.
2. Yuan Xu, A product formula for Jacobi polynomials, in:
Special functions (Hong Kong, 1999), World Scientific,
2000, pp. 423–430.
3. Yuan Xu, Constructing cubature formulae by the method
of reproducing kernel, Numer. Math. 85 (2000),
155–173.
4. Boris Rubin, Weighted spherical harmonics and
generalized spherical convolutions, preprint, The Hebrew
University of Jerusalem, 1999.

A Distant Mirror:: My Joint Paper With Aad After 35 Years

Uploaded by

Copyright:

Available Formats

A Distant Mirror:: My Joint Paper With Aad After 35 Years

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Distant Mirror:: My Joint Paper With Aad After 35 Years

Uploaded by

Copyright:

Available Formats

A distant mirror:

my joint paper with Aad after 35 years

Korteweg-de Vries Institute, University of Amsterdam

lecture on 24 February 2006

A distant mirror: – p.1/27

A distant mirror: – p.2/27

A distant mirror: – p.3/27

A distant mirror: – p.4/27

A distant mirror: – p.4/27

A distant mirror: – p.5/27

For α, β = 0, 12 , 1, 32 , . . . interpretation on spherical harmonics.

A distant mirror: – p.5/27

A distant mirror: – p.6/27

A distant mirror: – p.6/27

A distant mirror: – p.6/27

A distant mirror: – p.6/27

A distant mirror: – p.7/27

A distant mirror: – p.7/27

A distant mirror: – p.7/27

Corollary 1 Take g(x) := (x1 + ix2 )n .

A distant mirror: – p.7/27

A distant mirror: – p.8/27

Analytic continuation w.r.t. d by Carlson’s theorem.

A distant mirror: – p.8/27

A distant mirror: – p.9/27

A distant mirror: – p.10/27

A distant mirror: – p.10/27

Theorem (Braaksma & Meulenbeld, 1968)

A distant mirror: – p.10/27

Theorem (Braaksma & Meulenbeld, 1968)

A distant mirror: – p.10/27

Corollary 1 (Braaksma & Meulenbeld, 1968)

A distant mirror: – p.11/27

Corollary 1 (Braaksma & Meulenbeld, 1968)

( 12 d′′ −1, 21 d′ −1)

A distant mirror: – p.11/27

Corollary 2 (Dijksma & K, 1971)

A distant mirror: – p.12/27

Corollary 2 (Dijksma & K, 1971)

( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)

A distant mirror: – p.12/27

Corollary 2 (Dijksma & K, 1971)

( 12 d′′ −1, 21 d′ −1) ( 21 d′′ −1, 21 d′ −1)

Analytic continuation w.r.t. d′ , d′′ by Carlson’s theorem.

(α,β) (α,β) 1 (α,β)

for an explicit kernel Kα,β (x, y, z) ≥ 0 on [−1, 1]3 if

A distant mirror: – p.13/27

(α,β) (α,β) 1 (α,β)

for an explicit kernel Kα,β (x, y, z) ≥ 0 on [−1, 1]3 if

Question by Richard Askey

A distant mirror: – p.13/27

A distant mirror: – p.14/27

A distant mirror: – p.15/27

Definition A harmonic of type (m, n) on S 2d−1 ⊂ Cd is

A distant mirror: – p.15/27

Definition A harmonic of type (m, n) on S 2d−1 ⊂ Cd is

A distant mirror: – p.16/27

A distant mirror: – p.16/27