THE LEGENDRE POLYNOMIALS

The Generating Function

The origin of
the Legendre polynomials lies in the treatment of potential problems. For example, the potential at the point **r** due to a point charge located at the point is given by

(1.1)

Assuming that , we write

(1.2)

or, introducing the notation and ,

(1.3)

It is often desirable to expand
this potential in powers of *h* for a given value of *x*. So we write

The expansion coefficients are a set of polynomials of order *n*,
known as the Legendre polynomials, and the function is known as the generating function for the
Legendre polynomials.

The Legendre polynomials of low order may be readily obtained by writing down the first few terms of the binomial expansion of the generating function:

(1.5)

(1.6)

whence, by comparing coefficients of *h*, we obtain

Incidentally, reverting to the notation and , we see that the potential due to a point charge may be written as

(1.8)

or (1.9)

The term in is known as the *monopole* term, and
represents the potential due to a point charge placed at the origin (note that is actually located at the point ).
The term involving the quantity is known as the dipole term, being the dipole moment of the point charge . The next term is known as the *quadrupole*
term, and so on. This *multipole
expansion* of the potential is not often employed in the case of a single
point charge, but provides a useful approximation in the case of an extended
charge distribution. The first term in
the expansion is obtained by replacing the whole charge by a point charge of
equal magnitude at the origin. If the
charge distribution is spherically symmetric, and the center of symmetry is
chosen as the origin, this first term actually gives the potential exactly,
despite the apparent crudity of the procedure.
In general, however, the monopole term alone gives a rather poor
approximation of the potential. For
example, if the total charge is equal to zero, the potential does not necessarily
vanish, as would be suggested by truncating the series after the monopole
term. The dipole term, at the very
least, must also be included. However,
for some charge distributions, this term is also zero, and we must resort to
the quadrupole term to obtain a meaningful approximation of the potential.

__Special Properties of the Legendre Polynomials__

If we put in Equation (1.4), we obtain

(2.1)

(2.2)

Note, further, that

(2.3)

Comparing coefficients of , we see that

(2.4)

or (2.5)

This result is known as the Parity property. It can be used to obtain an important property of the Legendre polynomials, which can be seen as follows:

Let (2.6)

By virtue of the parity property, we then have

(2.7)

whence, on comparing coefficients of , we have

(2.8)

This means that if *n* is even, only even values of *r*
are permitted, i.e. the Legendre polynomial contains only even powers of *x*. Similarly, if *n* is odd, contains only odd powers of *x*.

We have already obtained analytic expressions for some of the lower-order Legendre polynomials. In the general case, we use the binomial expansion of the generating function, as follows:

(2.9)

(2.10)

(2.11)

Using the binomial expansion once again, we get

(2.12)

For purposes of comparison, the
Right Hand Side of Equation (2.13) must be modified in such
a way that the exponent of *h* becomes equal to *n*. To do this, we change *n* to ,
which gives

where is equal to if *n* is even and if *n* is odd. The upper limit on *r* is dictated by the presence of the
term in the denominator of Equation (2.14),
which requires that . Comparing coefficients of in Equation (2.14),
we obtain the series representation of the Legendre polynomials:

The Legendre polynomials can be cast into a more useful form by writing Equation (2.15) as follows:

and using the result

It follows from Equations (2.16) and (2.17) that

(2.18)

where we have extended
the upper limit in the summation from to *n*.
The justification for this procedure is that if *r* >
,
then the exponent . The result

(2.20)

then shows that each of the additional terms is zero. Now the summation in Equation (2.19) is immediately recognized as the binomial expansion of , so we obtain

(2.21)

This result is known as
Rodrigues’ formula for the Legendre polynomials. It shows explicitly that is a polynomial of order *n*. For the highest power of *x* in the
expansion of is ,
and in view of Equation (2.17) it follows that the
highest power of *x* in the expression for is ,
i.e. is a polynomial of order *n*.

For an arbitrary function defined on the interval , consider the integral

(3.1)

Using the Rodrigues’ formula, this may be written as

(3.2)

Integrating by parts, we obtain

(3.3)

The first term vanishes at both limits, so we obtain

(3.4)

and it is clear that *n* partial integrations yields
the final result

__Orthogonality of the Legendre Polynomials__

__ __

The result we have just derived can be used to obtain a very important property of the Legendre polynomials. Let us suppose that . Equation (3.5) then becomes

If *m* and *n* are
different, one of them must be less than the other, and we lose nothing in
generality by assuming that *m* < *n*. The highest power in the expansion of is . The result

(4.2)

then shows that

Equation (4.3) is known as the orthogonality relation, and the Legendre polynomials are said to form an orthogonal sequence on the interval .

The case can be investigated by inserting the result

(4.4)

into Equation (4.1), to give

The integral on the Right Hand Side of Equation (4.5) may be evaluated by recourse to the Gamma Function

(4.6)

the Beta Function

(4.7)

the relations

(4.8)

(4.9)

and the *duplication formula*

Making the substitution , we may write

(4.11)

Replacing *n* by in the duplication formula (4.10),
we get

(4.13)

Substituting Equation (4.14) and Equation (4.12) into Equation (4.5), we get

(4.15)

which may be combined with Equation (4.3) in the form

(4.16)

Alternatively, the orthogonality condition above may be obtained as follows:

(4.17)

or (4.18)

The integral on the Right may be evaluated by making the substitution

(4.19)

Putting in Equation (4.18), we get

(4.20)

Since we have made the assumption that , we may expand the logarithm to give

(4.21)

(4.22)

and, comparing coefficients of , we obtain the result

(4.23)

It is sometimes desirable to express a given function , defined on the interval in the form of a series of Legendre polynomials:

A set of functions for which such an expansion can be made
is known as a *complete set*. The
expansion coefficients can be found by multiplying both sides of
Equation (5.1)
by ,
integrating with respect to *x* from and using the orthogonality relation. This gives

(5.2)

(5.3)

It is assumed, of course, that the function is sufficiently well behaved that the integral on the Right Hand Side of Equation (5.4) can be evaluated.

A recurrence relation is a relation between functions and/or their derivatives of different order. To obtain such relations for the Legendre polynomials, we start with the generating function

and differentiate with respect to *h*:

Multiplying both sides by and using Equation (6.1), we get

(6.3)

and comparing coefficients of , we obtain the three-term recurrence relation

(6.4)

or (6.5)

Since the low-order Legendre polynomials are already known [see Equation (1.7)], this recurrence relation allows us to obtain expressions for Legendre polynomials of any order. For example, if we put in Equation (6.5), we get

(6.6)

(6.7)

Higher order Legendre polynomials may be obtained by putting , etc. in Equation (6.5). While this procedure may appear to be rather cumbersome, it provides a more efficient method of computer evaluation than the series formula, Equation (2.15).

If we
differentiate Equation (1.4) with respect to *x*,
we get

(6.8)

and, combining this with Equation (6.2), we obtain

(6.9)

from which a comparison of the coefficients of yields the three-term recurrence relation

Differentiation of Equation (6.5) gives

(6.11)

which, when combined with Equation (6.10) to eliminate , gives

(6.12)

Upon simplification, this yields the recurrence relation

Subtracting Equation (6.10) from Equation (6.13), we get

If we
multiply Equation (6.10)
by *x*, we get

and replacing *n* by Equation (6.14)
gives

Subtracting Equations (6.15) and (6.16), we get

If we now differentiate Equation (6.17)
with respect to *x*, we obtain the relation

(6.18)

The quantity in square brackets is obtained from Equation (6.10), to give

(6.19)

(6.20)

Thus we see that, if *n* is an integer, the Legendre
polynomial satisfies the differential equation

Equation (6.21) is known as Legendre’s differential equation.