The fundamental theorem of Gaussian quadrature states that an -point Gaussian quadrature
rule integrates polynomials through polynomial
degree
exactly, and its abscissas
are the roots of the orthogonal
polynomial of polynomial degree
for the same measure. Assume a
finite positive measure
on the real line with infinite
support and finite moments of all orders.
Let be the monic polynomial
of polynomial degree
orthogonal to all
lower-degree polynomials, with roots
,
...,
. The Lagrange
interpolating polynomials are
|
(1)
|
with , where
is the Kronecker delta.
Define the weights by
. Then
|
(2)
|
for every polynomial of polynomial degree
at most
.
To prove this, polynomial division gives
|
(3)
|
where and
have polynomial degree
at most
. The first term has zero integral
by orthogonality and vanishes at every
.
The Lagrange interpolating polynomial
is integrated exactly by the definition of the weights.
Conversely, exactness through polynomial degree
forces the monic polynomial with the quadrature
nodes as its roots to be orthogonal
to every polynomial of polynomial
degree at most
, so it is
.
Exactness applied to gives
|
(4)
|
No rule using real nodes can also integrate every polynomial
of polynomial degree
exactly. The square of its nodal polynomial
vanishes at every node but has positive integral.
For Gaussian nodes, the Christoffel function
also gives
.
For example, with on
, the two-point Legendre-Gauss
quadrature rule has nodes
and both weights equal
to 1. It is exact through polynomial degree
3, but gives
for
, whose integral is
.