Eur. J. Phys. 21 (2000) L1L3.

Printed in the UK PII: S0143-0807(00)03980-5

LETTERS AND COMMENTS

On the Laplacian of 1/r

V Hnizdo
Department of Physics, Schonland Research Centre for Nuclear Sciences, and Centre for
Nonlinear Studies, University of the Witwatersrand, Johannesburg, 2050 South Africa

Received 5 May 1999

Abstract. It is pointed out that the distinction between standard and non-standard
representations of the radial delta function (r) emphasized by Menon in a recent paper on the
solving of the radial Poisson equation for a point charge is devoid of any significance. It is also
shown how the solution 1/|r r 0 | of the three-dimensional Poisson equation for a point charge
can be derived with no recourse to any representation of the delta function, thereby clarifying the
solutions precise meaning.

One way to show that ansatz like Jacksons regularization (r 2 +

1 (r) b2 )1/2 , and instead solves the radial Poisson
2 = 2 (1) equation
r r  
1 d 2 dV (r)
i.e. that the function 1/r is a solution of Pois- V (r) = 2
2
r = 2 (3)
sons equation for the spherically symmetric r dr dr r
point distribution (r) = (r)/4r 2 , is to by substituting a standard representation of
suitably regularize the function 1/r in terms the radial delta function into a formal solution
of a parameter b so that the regularized func- V (r) of equation (3):
tion is non-singular at r = 0 when the pa- Z Z
1 r 0 0 (r 0 )
rameter b 6= 0. The verification of equa- V (r) = dr (r ) + dr 0 0 . (4)
r 0 r r
tion (1) is then accomplished by showing that,
in the limit b 0, r 2 times the Laplacian For Menon a standard representation of the
of the regularized function is a representation delta function is such that it results in what
of the radial delta function (r). For exam- he believes to be a diverging value of the
ple, Jackson [1] regularizes the function 1/r delta function at r = 0, (0) = +, as, for
as (r 2 + b2 )1/2 ; it can then be shown that example, in the rational representation
2 b
1 3b2 (r) = lim (5)
lim 2 = lim b0+ r 2 + b2
b0 (r 2 + b2 )1/2 b0 (r 2 + b2 )5/2
and unlike the representation (2), which he
(r) takes as resulting in (0) = 0. Here, Menon
=
. (2) understands by the values of (r) at r =
r2
0 the direct limits b 0 of the various
In a recent paper in this journal, representations of (r) evaluated at r = 0.
Menon [2] regards such a procedure as When Menon uses the representation (5) of
ambiguous because it involves an ad hoc (r) in equation (4), he obtains the following
Present address: National Institute for Occupational expression for the solution V (r):
Safety and Health, M/S L-3030, 1095 Willowdale Road,

2 1 r
Morgantown, WV 26505, USA. V (r) = lim arctan
In fact, Jackson [1] shows the validity of a more general b0+ r b
statement, namely that of limb0 2 (r 2 + b2 )1/2 = 2 2
4 (3) (r ), where (3) (r ) is the three-dimensional delta 1 r +b
function.
+ ln (6)
2b r2
0143-0807/00/010001+03$30.00 2000 IOP Publishing Ltd L1
L2 Letters and Comments

which he considers as exact because it distribution rather than a classical function,

already satisfies equation (3) before the limit is defined solely by its sifting property
Z
b 0+ is taken. However, Menon
cautions that, when used in equation (4), f (r)(r) dr = f (0). (10)
0
different representations (presumably both
standard and non-standard) of the radial As the delta function has a proper meaning
only in an integral, the expressions of
delta function (r) will lead to solutions V (r) equations (1)(3), (5) and (7) are meaningful
that differ in the interior region r . b while only when used inside an integral and when
agreeing in the exterior region r  b. they involve a limit, it must be taken outside
We would like to point out that the the integral. Thus all the different solutions
differences seemingly arising from the use of V (r) Menon would obtain in his procedure
different representations of the delta function by using different representations of (r) are
are devoid of any significance in fact, in fact the same function 1/r when the limit
they are non-existent when a representation b 0 is performed properly.
of the delta function is understood properly. To derive relation (7), and to clarify its
Also, we use this opportunity to present meaning, we shall use the divergence theorem
a derivation of the well known three- and Gausss law. Consider the Laplacian of
an integral over all space:
dimensional generalization of relation (1): Z
1
1 2 (r 0 ) d3 r 0
2 = 4 (3) (r r 0 ) (7) |r r 0 |
|r r 0 | Z
1
which does not need any representation of the = (r 0 ) d3 r 0
three-dimensional delta function (3) (r r 0 ) |r r 0 |
Z
and instead uses its defining sifting property. r r0 3 0
Such a derivation turns out to be instructive as = (r 0 ) d r. (11)
|r r 0 |3
it clarifies the precise meaning of relation (7).
Here, (r 0 ) is a well-behaved localized
Different representations of the radial function. One is permitted to take one gradient
delta function, like the non-standard one operator inside the integral as the resulting
in equation (2) or the standard one in integrand (r 0 )(r r 0 )|r r 0 |3 is perfectly
equation (5), involve a limit that is understood integrable because its 1/r 2 -type singularity
to be taken only after a test function is is neutralized by the volume element that is
integrated together with the representation proportional to r 2 . However, taking the whole
function. Thus, to use the representation (2), Laplacian 2 = inside the integral
Z would not be a legitimate operation here this
3b2 r 2 can be seen simply from the fact that
lim f (r) 2 dr = f (0) (8)
b0 0 (r + b2 )5/2 1
2 =0 for all r 0 6= r (12)
where f (r) is a test function that is well |r r 0 |
behaved at r = 0. When the sifting and so, as Rlimr0 r 2 |r r 0 |1 = 0,
property (8) can be shown to hold for any well- the integral (r 0 ) 2 |r r 0 |1 d3 r 0 , taken
behaved test function f (r), the representation literally, would yield zero for any given
in question properly defines the delta function point r . Using suggestive electromagnetic
(r), irrespective of its actual functional form symbols, we write the right-hand side of
or what value it takes at r = 0 in fact, the equation (11) as E (r ), where
Z
latter characteristic is meaningless, as the limit r r0 3 0
b 0 must be taken outside the integral. In E (r ) = (r 0 ) d r (13)
|r r 0 |3
other words, all functions g(b, r) that satisfy
Z integrate over a volume V , apply the
divergence theorem, and then Gausss law:
lim f (r)g(b, r) dr = f (0) (9) Z I
b0 0
E (r ) d r = E (r ) dS
3
for any well-behaved function f (r) are V
Z S
entirely equivalent as representations of the
radial delta function (r), which, as a = 4 3
(r ) d r. (14)
V
 Letters and Comments L3

We can use the divergence theorem here (r 0 ) as a test function. Thus, we can write
because E (r ) is a well-behaved function Z
1
of r ; note that one cannot use it legitimately ( ) d3 r 0
with the singular function (r r 0 )|r r 0 |3 |r r 0 |
Z
itself this would yield a non-zero value
for an integral whose integrand, according to = 4 ( ) (3) (r r 0 ) d3 r 0 (17)
equation (12), vanishes everywhere except at
a single point where it is not defined which where ( ) stands for a well-behaved
means that derivations of equation (7) that do function. Usually, this result is put formally
so are not correct (see, e.g., [3]). Gausss as relation (7), but equation (17) shows that,
law, which is applied to the surface integral in strictly speaking, relation (7) has a meaning
equation (14), holds with the function E (r ) only when integration is implied and at least
on account of the inverse-square nature of one gradient operator is outside the integral.
(r r 0 )|r r 0 |3 (see, e.g., [4]). Putting the The derivation of relation (7) given here is
explicit expression (13) for E (r ) back into a rigorous alternative to a regularization pro-
equation (14), but again writing (r r 0 ) cedure, such as that due to Jackson [1]; unfor-
|r r 0 |3 as |r r 0 |1 , we obtain tunately, both the procedures of regularization
Z Z  and proof along the lines of the present deriva-
0 1 3 0 tion have to be somewhat less straightforward
(r ) d r d3 r
V |r r 0 | than the frequently given non-regularization
Z proof [3] which attempts to use the diver-
= 4 (r ) d3 r. (15) gence theorem directly with a function that
V has a (1/r 2 )r /r-type singularity.
As equation (15) holds for any volume V , it
follows that
Z References
0 1 3 0
(r ) d r
|r r 0 | [1] Jackson J D 1975 Classical Electrodynamics
2nd edn (New York: Wiley) p 39
= 4(r ) (16) [2] Menon V J 1999 On solving the Poisson equation
for a point charge Eur. J. Phys. 20 813
which is, of course, Maxwells equa- [3] Jackson J D 1962 Classical Electrodynamics 1st edn
tion E (r ) = 4(r ), but writ- (New York: Wiley) p 13
ten in such a way that it exhibits the [4] Purcell E M 1965 Electricity and Magnetism
(Berkeley Physics Course 2) (New York:
defining sifting property of the three- McGraw-Hill) ch 1.10
dimensional delta function (3) (r r 0 ), with