Greens Functions and their applications in Physics

Erik M. Olsen
University of Tennessee Knoxville, TN 37996-1200 (Dated: October 1, 2008) Dierential equations appear frequently in various areas of mathematics and physics. In this paper the method of Greens Functions as solutions to these equations will be discussed in length. Also included will be various examples of physical problems where Greens Functions solutions are useful. 1. INTRODUCTION

In studying physics, it would be dicult to avoid the presence of dierential equations. These mathematical constructs appear quite often and can be extremely varied in type. For example, there is the equation of simple harmonic motion (a homogeneous ordinary dierential equation): d2 + k2 = 0 dx2

linear operator. We also assume that this equation is subjected to certain boundary conditions. Now consider the idea that we can nd a function G(r, r0 ) that solves this particular dierential equation with a delta function as a source (or inhomogeneity) instead of f (r): W G(r, r0 ) = (r r0 ) (4)

(1)

As an example, let W = 2 . Using Greens Theorem [1] on G(r, r0 ) and m(r): (m2 G G2 m)d =
V S

There is also Poissons equation (an inhomogeneous partial dierential equation): 2 =

0

(mG Gm) dA (5)

(2)

From the equations above: [m(r) (r r0 ) G(r, r0 )f (r)]d

V

Fortunately, there exist many ways, both analytical and numerical, of solving these equations and others. The method of Greens Functions (named for English mathematician and physicist George Green) is particularly useful for the latter type of equation shown here. In section 2 of this paper the general process of forming a Greens Function and the properties of Greens Functions will be discussed. In section 3 an example will be shown where Greens Function will be used to calculate the electrostatic potential of a specied charge density. In section 4 an example will be shown to illustrate the usefulness of Greens Functions in quantum scattering. Finally, in section 5 a conclusion of all things discussed will be given.

(6) (7)

=
S

(m(r)G(r, r0 ) G(r, r0 )m(r)) dA

If we interchange r and r0 (to be justied later): [m(r0 ) (r0 r) G(r0 , r)f (r0 )]d
V

(8) (9)

=
S

(m(r)G(r, r0 ) G(r, r0 )m(r)) dA

By performing the integral on the delta function and adding we nd [2]: m(r) =
V

G(r, r0 )f (r0 )d

(10) (11)

2.

GREENS FUNCTIONS SOLUTIONS

+
S

(m(r0 )G(r, r0 ) G(r, r0 )m(r0 ))

Suppose we have a dierential equation of the following form: W m(r ) = f ( r ) (3)

Where m(r) is the function to be determined, f (r) is a term that contains m and derivatives of m and W is a

Electronic

address: eolsen1@utk.edu

We call G(r) the Greens Function of this particular dierential equation. If we can nd a Greens Function G(r) that satises the equation above, then we can nd our desired function. The whole problem then is to manipulate this formula to give something useful, and such manipulations are specic to the particular equation to be solved and the geometry of the problem. The derivation above is not the only way to solve for the Greens Function; it just happened to be the most

2 convenient since the linear dierential operator was the Laplacian. Greens Functions are always the solution of a -like inhomogeneity. However, it is worthwhile to mention that since the Delta Function is a distribution and not a function, Greens Functions are not required to be functions. It is important to state that Greens Functions are unique for each geometry. However, you may add a factor G0 (r) to the Greens Function G(r) where G0 (r) satises the homogeneous dierential equation in question[3]: W G0 (r) = 0 (12) we can change some terms in our equation: [G(r, r2 )p(r)G(r, r1 )] [G(r, r1 )p(r)G(r, r2 )] = G(r, r2 ) (r r1 ) + G(r, r1 ) (r r2 ) (24) (25) (26)

Performing a volume integral on both sides and using the Divergence Theorem [1] to simplify the left side: [G(r, r2 )p(r)G(r, r1 )
S

(27) (28) (29)

G(r, r1 )p(r)G(r, r2 )] dA = G(r1 , r2 ) + G(r2 , r1 )

Where W is a linear dierential operator. They also rely heavily on the specied boundary conditions; a Greens Function for one boundary may not be dened on other. One of the most useful properties of Greens Functions is that they are always symmetric[4]: G(a, b) = G(b, a)a = b (13)

By the boundary conditions imposed at the beginning, the left hand side is zero. Hence: G(r1 , r2 ) = G(r2 , r1 ) Thus proving the symmetry of Greens Function.
3. EXAMPLE: ELECTROSTATIC POTENTIAL[5]

(30)

When calculating the Greens Function earlier, we depended on switching r with r0 ; if the variables of this Greens Function were not symmetric we could not do this. As such, it is worthwhile to prove the symmetry of the variables in any Greens Function. Suppose we have two Greens Functions G(r, r1 ) and G(r, r2 ). We require that these equations satisfy the following equations [5]: [p(r)G(r, r1 )] + q (r)G(r, r1 ) = (r r1 ) [p(r)G(r, r2 )] + q (r)G(r, r2 ) = (r r2 ) (14)

Let us start with Poissons equation: 2 =

0

(31)

Suppose we have a collection of point charges qk . From the study of electrostatics, we know the electric potential of such a conguration: = 1 4 0 qk rk (32)

(15)

Here p(r) and q (r) are random functions of r. We shall specify the Greens Functions further by imposing Dirichlet boundary conditions [5] on them where G(r, r1 ) and G(r, r2 ) will yield the same values over the surface S of some volume. Should this condition not be met, the Greens Functions will disappear on S . Multiplying the equation for G(r, r1 ) by G(r, r2 ) and vice versa we nd: G(r, r2 ) [p(r)G(r, r1 )] + G(r, r2 )q (r)G(r, r1 ) = G(r, r2 ) (r r1 ) G(r, r1 ) [p(r)G(r, r2 )] + G(r, r1 )q (r)G(r, r2 ) = G(r, r1 ) (r r2 ) Subtracting the second equation from the rst: G(r, r2 ) [p(r)G(r, r1 )] G(r, r1 ) [p(r)G(r, r2 )] = G(r, r2 ) (r r1 ) + G(r, r1 ) (r r2 ) Using the following product rule [1] (f A) = f ( A) + A f (23) (20) (21) (22) (16) (17) (18) (19)

If instead of a discrete number of point charges we had a continuous distribution of charge (let be such a charge density) then we have: 1 4 0 (r ) |r r |

( r ) =

dV

(33)

Where r is the vector pointing from the origin to the eld point, r is the vector pointing from the origin to the source point and dV is a volume element. Observing Poissons Equation and the equation directly above, we are now in a position to try and nd a Greens Function G(r, r ). Using the formalism from section 2: 2 G(r, r ) = (r r ) (34)

We will now solve for G(r, r ). Employing Greens Identity [1]: ((r )2 G(r, r ) G(r, r )2 (r ))dV
V

(35) (36)

=
S

((r )G(r, r ) G(r, r )(r )) dA

3 Where dA is the surface element for the surface that surrounds the volume V . We shall consider the case where the surface term is zero. This will simplify the problem: ((r )2 G(r, r ) G(r, r )2 (r ))dV = 0
V

2 (r) + k 2 (r) = [

2mV (r)

(r)]

(45)

(37)

Ultimately, we will want a solution that contains the inikr cident wave eik0 r , the scattered wave e r and the amplitude of the scattered wave fk (, ). Thus, (r) will have the asymptotic form: eikr r

(r )2 G(r, r )dV =
V V

G(r, r )2 (r )dV

(38) (r) eik0 r + fk (, ) (46)

Plugging in for the values of 2 G(r, r ) and nabla2 (r): (r ) (r r )dV = G(r, r )(r )
0

dV

(39)

So we nd: ( r ) = 1
0

Here k0 is the vector pointing in the direction of the incident wave and k is the vector pointing in the direction of the scattered wave. By the general form of the Greens Function found in section 2: (r1 ) =
V

G(r, r )(r )dV

(40) 2m
2

V (r2 ) (r2 )G(r1 , r2 )d3 r2

(47)

We are now in a position to nd our Greens Function. From studies of Poissons Equation and Coulombs Law: 1 2 = 4 (r) r (41)

Because this is meant to describe an outgoing wave that approaches innity, we set the surface term equal to zero. We add a term to make (r) asymptotic: (r1 ) = eik0 r1
V

Looking back at the original equation for 2 G, we can conclude: G(r, r ) = This yields for : ( r ) = 1 4 0 (r ) |r r | 1 4 | r r | (42)

2m
2

V (r2 ) (r2 )G(r1 , r2 )d3 r2 (48)

Since we are working with the Helmholtz operator, our Greens Function will have the form [5]: G(r1 , r2 ) = eik|r1 r2 | 4 |r1 r2 | (49)

dV

(43)

Hence we get the exact equation: (r1 ) = eik0 r1

V

Even though we knew the answer from the very beginning, it is nice to see that the Greens Function formalism gives the correct results. Of course, there are other examples where we dont know the answer from hindsight and must rely on Greens Functions to solve the problem.

2m
2

V (r2 ) (r2 )

eik|r1 r2 | 3 d r2 (50) 4 |r1 r2 |

4.

QUANTUM SCATTERING

By making approximations on , we can get extremely useful information out of this equation. For example: if we assume the incident wave is not considerably changed by the potential:

Suppose we have a beam of particles incident on a target (represented by a potential V (r)). The particles that hit the target scatter o of it as spherical waves; we represent those waves with the wave function (r). Obviously (r) obeys the Schrodinger Equation [5]:
2

(r2 ) = eik0 r2 Plugging this into the integral:

(51)

2m

2 (r) + V (r) (r) = E (r)

(44)

This will be more useful in the form of the Helmholtz Equation. As such, we dene k 2 = 2mE and write: 2

eik|r1 r2 | 3 d r2 2 4 |r1 r2 V (52) The above equation is known as the Born Approximation; it is very useful for scattering problems where the potential is weak compared to the given potential. 1 (r1 ) = eik0 r1 2m V (r2 )eik0 r2

4
5. CONCLUSION

There are numerous methods available to solve dierential equations. At rst glance, it may seem as though the method of Greens Functions is rather limited since

it can only be used on equations of a particular form and by the fact that not all linear operators admit a Greens Function. Nevertheless, these kinds of equations do appear frequently in physics, so Greens Functions prove to be invaluable in the understanding of physical systems.

[1] D. J. Griths, Introduction to Electrodynamics (PrenticeHall, Inc, 1999), 3rd ed. [2] L. Challis and F. Sheard, The Green of Greens Functions, vol. 56 (2003). [3] D. J. Griths, Introduction to Quantum Mechanics (Pearson Prentice Hall, 1995), 2nd ed.

[4] W. Strauss, Partial Dierential Equations (John Wiley and Sons, 1992), 1st ed. [5] G. Arfken and H. Weber, Mathematical Methods for Physicists (Elsevier Adademic Press, 2005), sixth ed.