Orthogonal Polynomials

1 Introduction
Mathematically orthogonal means perpendicular, that is, at right angles. For example, the set of vectors
fx y z in three-dimensional sppace are orthogonal. The concept of orthogonality can be formalised, with
g

respect to vectors, as
x y = 0 x z = 0 y z = 0
that is, the dot product of two orthogonal vectors is zero. Indeed the set of vectors x y z is an orthogonal
f g

set. If x y z are all unit vectors we say that x y z is an orthonormal set of vectors. Orthogonal vectors
f g f g

are linearly independent and x y z therefore spans three-dimensional space. Thus any point in space
f g

may be written in terms of its components in the x, y and z directions (or indeed in terms of components
in the directions of any three orthogonal vectors). In this way x y z forms a basis for three-dimensional
f g

space.

1.1 Orthogonal Functions

Two functions, fi (x) and fj (x), are orthogonal on an interval a b] if
Zb
fi (x)fj (x)dx = 0 (1)
a
Mathematically we say that the inner product of the functions fi (x) and fj (x) is zero. The functions are
orthonormal if Zb
fi (x)fj (x)dx =
ij (2)
a
where

ij = 1 i=j
= 0 i=j
6

is the Kronecker delta function.

2 Orthogonal Polynomials
A set of orthogonal polynomials is an in nite sequence of polynomials, p0 (x) p1 (x) p2 (x) : : :, where pn (x)
has degree n and any two polynomials in the set are orthogonal to each other, that is,
Zb
fi (x)fj (x)dx = 0 for i = j 6 (3)
a
The set of polynomials is orthonormal if
Zb
fi (x)fj (x)dx =
ij (4)
a
The interval a b] is called the interval of orthogonality and may be in nite at one or both ends.

1
Any in nite sequence of polynomials pn , with pn having degree n forms a basis for the in nite-dimensional
f g

vector space of all polynomials. Such a sequence can be turned into an orthogonal basis using the Gram-
Schmidt orthogonalisation process, by projecting out the components of each polynomial that are orthog-
onal to the polynomials already chosen.

2.1 Hilbert Space

Hilbert (David Hilbert 1862-1943) space generalises the idea of Euclidean space (that is, three-dimensional
vector space etc) to in nite-dimensional spaces. Mathematically, a Hilbert space is an inner product space
that is complete.
Hilbert spaces typically arise as in nite-dimensional function spaces. An element of a Hilbert space can
be expressed by its coordinates with respect to an orthonormal basis of the space (directly analogous to
fx y z coordinates in three-dimensional vector space).
g

Note:
Complete means that as a sequence of functions approaches some limit, then the limit is also part of the
Hilbert space. For example, if a molecular wave function is expanded in terms of orthogonal basis functions
of a Hilbert space, then the molecular wavefunction is also an element of the Hilbert space.

The concept of Hilbert space oers one of the best mathematical formulations of quantum mechanics.
Quantum mechanical states (wavefunctions) are described by vectors in a Hilbert space, quantum mechan-
ical observables can be expressed by linear operators and quantum measurement is related to orthogonal
projection.

2.2 Properties of Orthogonal Polynomials

All sets of orthogonal polynomials have a number of fascinating properties.

 Any polynomial f (x) of degree nPcan be expanded in terms of p0 p1 : : : pn , that is, there exist
coecients ai such that f (x) = ni=0 ai pi (x).
 Given an orthogonal set of polynomials fp0 (x) p1 (x) : : :g, each polynomial, pk (x) is orthogonal to
any polynomial of degree < k.
 Any orthogonal set of polynomials fp0 (x) p1 (x) : : :g has a recurrence formula that relates any three
consecutive polynomials in the sequence, that is, the relation pn+1 = (an x + bn )pn ; cn pn;1 exits,
where the coecients a, b and c depend on n. Such a recurrence formula is often used to generate
higher order members in the set.
 Each polynomial in fp0(x) p1 (x) : : :g has all n of its roots real, distinct and strictly within the
interval of orthogonality (ie not on its ends). This is an extremely unusual property! It is particularly
important when considering the classes of polynomials that arise as quantum mechanical solutions to
a given Hamiltonian (or other Hermitian) operator - see below.
 Furthermore the roots of the nth degree polynomial, pn lie strictly inside the roots of the (n + 1)th
degree polynomial pn+1 .

2
2.3 Examples of Orthogonal Polynomials
The eld of orthogonal polynomials was developed in the late 19th century and many of the sets of orthog-
onal polynomials described arose from descriptions of speci c physical problems.
Legendre Polynomials
These polynomials are orthogonal on the interval  1 1]. They arise as solutions to Legendre's dierential
;

equation (Adrien-Marie Legendre 1752-1833),

(1 x2 ) @ @x
2 P (x)
;
2 2x @P@x(x) + P (x) = 0
; with  = n(n + 1) (5)
This equation is frequently encountered in Physics, particularly when spherical polar coordinates are used
and where the problem has cylindrical symmetry. Legendre polynomials form part of the quantum me-
chanical solution for rotational motion.
P0 (x) = 1
P1 (x) = x
P2 (x) = 21 (3x2 1) ;

P3 (x) = 12 (5x3 3x) ;

P4 (x) = 18 (35x4 30x2 + 3)

;

P5 (x) = 18 (63x5 70x3 + 15x)

;

P6 (x) = 161 (231x6 315x4 + 105x2 5) : : :

; ;

Laguerre Polynomials
The Laguerre polynomials, Ln are solutions to Laguerre's dierential equation (Edmond Laguerre 1834-
1886),
x @@L x) @L
2
n n
2 L + (1
n @x + Ln = 0 with  = n
; (6)
This equation arises in the radial part of the Hamiltonian for one-electron atoms.
L0 (x) = 1
L1 (x) = x + 1
;

L2 (x) = 21 (x2 4x + 2)
;

L3 (x) = 16 ( x3 + 9x2 18x + 6)

; ;

1 (x4 16x3 + 72x2 96x + 24)

L4 (x) = 24 ; ;

1 ( x5 + 25x4 200x3 + 600x2 600x + 120x)

L5 (x) = 120 ; ; ;

1 (x6 36x5 + 450x4 2400x3 + 5400x2 4320x + 720) : : :

L6 (x) = 720 ; ; ;

The Laguerre polynomials are orthogonal over the interval 0 + ) with respect to the weight function e;x.
1

Indeed the radial component of the wavefunction of the one-electron atom is obtained by multiplying the
Laguerre polynomials by e;x

3
Hermite Polynomials
Hermite polynomials, Hn (x), arise as part of the solution to the quantum harmonic oscillator Hamiltonian.
They are solutions to Hermite's equation (Charles Hermite 1822-1901):
@Hn2 2xH + H = 0 with  = n (7)
@ 2 Hn n; n

The2 Hermite polynomials are orthogonal over the interval ( + ) with respect to the weight function
;1 1

e ; x = 2 . Indeed the quantum harmonic oscillator wavefunctions are obtained by multiplying the Hermite
polynomials by e;x2 =2 .
H0 (x) = 1
H1 (x) = 2x
H2 (x) = 4x2 2x + 2
;

H3 (x) = 8x3 12x

;

H4 (x) = 16x4 48x2 + 12

;

H5 (x) = 32x5 160x3 + 120x)

;

H6 (x) = 64x6 + 480x4 + 720x2 120 : : :

;

3 Operators
Every observable physical quantity can be characterised by an operator.
An operator is a mathematical device that converts one function into another. It is one step up from a
function, which is a device that converts one number (or collection of numbers) into another.
In general a caret or hat is used to denote an operator.
The simplest form of operator arises for physical observables that are just functions of position coordinates.
These just multiply the functions on which they operate.
Example: The classical potential energy of a harmonic oscillator is 12 kx2. The associated quantum
mechanical operator V^ just multiplies any function by 12 kx2 , that is
V^ (x) = 12 kx2 (x)
A more interesting type of operator is a dierential operator.
For instance, the operator dxd changes the function sin x to cos x:
d
dx sin x = cos x
If a particle's position is described by coordinates (x y z ), then its momentum in the x direction is given
by the operator
@
p^x = ih @x
;

Where partial derivatives have been used because there are often several coordinates to consider. The form
of the quantum operator for momentum, p^x is unusual, it does not appear to have any correspondance with

4
classical momentum, mv. This form represents another of the basic axioms on which quantum mechanics
has been developed.
If we can write some physical quantity in terms of position and momentum variables, then the rules for
constructing operators are:
 All position variables remain unchanged
 A momentum in the direction of a coordinate q is replaced by the operator ih @q@ . ;

Notes:
 We almost always work with momentum in quantum mechanics rather than velocity.
 This second rule applies to angular coordinates as well as Cartesian ones. For example, rotation
about the z axis is often described using a polar coordinate . Angular momentum about the z axis
is then described by the operator ih @
;
@.

3.1 Expectation Values

In quantum mechanics every physical observable, Q, can be characterised by an operator, Q^ , and the
expectation value of Q^ is calculated as R
^
Q = R
Qdd
h i (8)
R
Here : : : d is a conventional notation meaning that we integrate over all the variables and over all space.
The order of the factors in the numerator is important: operators operate to the right only, so in this
case Q^ operates only on , not on 
. This does not matter for position operators, which operate by
multiplication, but is crucial for dierential operators.

3.2 Dirac Notation

The notation  is often used for the wavefunction . Often we abbreviate, writing n as n . n or
j i j i j i j i

n is called a ket vector.

j i

h is called a bra vector. The bra notation implies the complex conjugate 
.
j

a complete bra-ket expression like   or  Q^  implies integration over all space.

h j i h j j i

Thus
Z
h  j i = 
d
Z
= j  2 d
j

Z
h  Q^  =
j j i 
Q^ d
and
^
h Q = Q
i
h

h
j

j
j

i
i

5
3.3 Eigenfunctions and Eigenvalues
Suppose that we have an operator Q^ and a wavefunction q that satisfy the equation:
Q^ q = qq (9)
where q is just a number (with the appropriate dimensions). That is, if q depends on some set of variables
(x1 x2 : : :) then q is independent of all these variables.
Equation (9) is an eigenvalue equation. The wavefunction q is an eigenfunction and q is the corre-
sponding eigenvalue.
Much of the eort in practical quantum chemistry involves nding eigenfunctions and eigenvalues of inter-
esting operators.
The operator for energy is called the Hamiltonian and its symbol is . Its eigenvalue equation is H

Schrodinger's time-independent equation:

 = E H (10)

In general this equation has many solutions

H n = En n n = 1 2 3 :::
and we have a set of eigenfunctions n for Hamiltonian. In such cases the symbol  does not carry any
information, and the label n is important, so it is common to use the Dirac notation n , which gives the
j i

label more prominence.

A general property of the set of eigenfunctions of an operator like the Hamiltonian is that they are or-
thogonal Za
hmn j m
n dx = 0 if m = n
i  6
0
Another important property is that the eigenfunctions form a complete set, provided that En as ! 1

n ! 1 . That is, any function of the same variables with the same boundary conditions can be expressed
as a linear combination of the n : X
 = cn  n
n

3.4 Hermiticity
This is an important technicality. Hermiticity is a generalisation of complex conjugation: under complex
conjugation an operator Q^ becomes its Hermitian conjugate, Q^ y .
Consider the complex conjugation:
Z


h 1 Q^ 2
=
j j i 1
(Q
^ 2 )d
Z
= ^ 2 )
d
1 (Q
^ 2 1
= Qh j i

However, we can consider the bra vector, h^ 2 in terms of a new operator, Q^ y . Where we dene
Q j

Q^ 2
f j ig

= jQ
^ 2 i
= hQ
^ 2 j = h2 jQ^ y

6
Therefore we can write
1 Q^ 2
= Q
h j j ^ 2 1 = 2 Q^ y 1
i h j (11)
i h j j i

Thus every operator Q^ has a Hermitian conjugate, Q^ , which can be de ned using eq. (11), or its
y
integral equivalent: Z Z
1
Q
^ 2 d = (Q^ y 1 )
2 d

An operator that is equal to its Hermitian conjugate is said to be Hermitian.

Example: Find the Hermitian conjugates of the operators dxd and i dxd
Z1 Z 1 d

1
d 2
dx =  
2 ]1
;1 ;1 dx 2 dx
1
;1 dx
1 ;

Z 1 d
= ( 1 )
2 dx
;1 dx
;

therefore ( dx d )y = d
dx ;

Z 1 d Z 1 d

1
i dx2 dx = 1
i2 ]1;1 ;1 ( i dx )2 dx
+ 1
;
;1 Z 1 d
= (i 1 )
2 dx
;1 dx
therefore (i dx d )y = i d
dx
Thus i dxd is Hermitian (cf the momentum operator) but dxd is not.

3.5 Properties of Hermitian operators:

 Their eigenvalues are always real
 Eigenfunctions corresponding to dierent eigenvalues are orthogonal

Proof: Show the eigenfunctions of a Hermitian operator are real and that eigenfunctions with dierent
eigenvalues are orthogonal
Suppose 1 and 2 are eigenfunctions of an Hermitian operator B^ with eigenvalues b1 and b2 . Then, using
Dirac notation:
B^ 1 = b1 1
j i (a) j i

B^ 2 = b2 2
j i (b) j i

Premultiply eq. (a) by 2
and integrate over all space (this is a standard trick and is used in most proofs
in quantum mechanics):
h 2 B^ 1 = b1 2 1
j j i h j i

Now take the complex conjugate:

h 2 B^ 1
= b
1 2 1

j j i h j i

7
Now use Hermiticity:
h 1 B^ 2 = b
1 1 2
j j i h j i

Premultiply eq. (b) by 1
and integrate over all space:

h1 B^ 2 = b2 1 2
j j i h j i

Combining the last two results:

(b
1 b2 ) 1 2 = 0
; h j i

Now, if we had started with 2 = 1 and b2 = b1 then we would have

(b
1 b1 ) 1 1 = 0
; h j i

But the integral of 1 2 over all space is non-zero, (the basic postulate of quantum mechanics that  2 is
j j j j

proportional to probability) so for the equation to be true we must have b
1 = b1 , that is, the eigenvalues
must be real.
Alternately, if we had started with b1 = b2 then we must conclude that the integral 1 2 = 0. Hence the
6 h j i

eigenfunctions of a Hermitian operator with dierent eigenvalues are orthogonal. This is one side of an if
and only if proof, it can also be shown that if an operator has all its eigenvalues real and its eigenfunctions
orthogonal it must be Hermitian.
Other useful properties of Hermitian operators are, if A^ and B^ are Hermitian:

 If A^ is invertible so is A^y and (A^y );1 = (A^;1 )y

 (A^ + B^ )y = A^y + B^ y
 (A^)y = 
A^y where 
is the complex conjugate of the scalar 
 (A^B^ )y = B^ y A^y

All observable (measurable) properties of a system must be real and hence the operators associated with
observable properties, like the Hamiltonian, momentum, dipole moment etc, are Hermitian. Thus, when
we deal with such operators, we can make use of all the useful properties of Hermitian operators.
Another property of operators is that, if two observables are to have simultaneously precisely de ned values,
their corresponding operators must commute. If you would like more details on this please let me know (if
nothing else it will tell me if anyone has actually read this far!!)
This (non-examinable) description of some of the mathematics behind quantum mechanics touches on the
20th century advances in mathematical theory. (Most of the maths we use is 19th century or earlier).
The most elegant derivation of quantum mechanics (to me anyway) comes from the recognition that the
quantum mechanical commutator, identi ed as the Lie bracket, generates a Lie Algebra. The eigenvalues
of the operators are the roots of the Lie Algebra. As such the recurrence relations between the various
eigenfunctions (and step-up and step-down operators) arise as the operators that step from one root of the
Lie algebra to the next.