You are on page 1of 9

IOP PUBLISHING

EUROPEAN JOURNAL OF PHYSICS

Eur. J. Phys. 28 (2007) 11891197

doi:10.1088/0143-0807/28/6/016

Presenting Newtonian gravitation


Martin Counihan
School of Physics and Astronomy, University of Southampton, Highfield,
Southampton SO17 1BJ, UK
E-mail: counihan@soton.ac.uk

Received 29 June 2007, in final form 21 September 2007


Published 24 October 2007
Online at stacks.iop.org/EJP/28/1189
Abstract

The basic principles of the Newtonian theory of gravitation are presented


in a way which students may find more logically coherent, mathematically
accessible and physically interesting than other approaches. After giving
relatively simple derivations of the circular hodograph and the elliptical orbit
from the inverse-square law, the concept of gravitational energy is developed
from vector calculus. It is argued that the energy density of a gravitational
field may reasonably be regarded as g2/8 G, and that the inverse-square law
may be replaced by a Schwarzschild-like force law without the need to invoke
non-Euclidean geometry.

1. Introduction

It would be difficult to exaggerate the influence of Isaac Newtons theory of gravitation on the
subsequent development of physics. As well as explaining Johannes Keplers laws of planetary
motion, Newtons theory was central to the successful mathematization of physics using the
newly-invented calculus and it served as a paradigm for the later theories of electrostatics and
magnetostatics. The theory of gravitation was a key aspect of the Newtonian philosophy of
science according to which natural phenomena were believed to be describable by universal
and unchanging mathematical laws, with the universe being bound together by the interplay
of forces acting at a distance between corpuscles of matter.
In view of the importance of the topic, one would expect Newtonian gravitation theory to be
presented to students in a coherent, logical and mathematically straightforward way. However,
introductory courses at first-year university level, and the textbooks which accompany them,
often treat the topic awkwardly or inadequately. This is partly because of conceptual
difficulties, and partly to do with the mathematical skills required. The standard calculation
provided online by Weinstein (2005), for example, involves calculus which is apparently
considered too difficult for most students. Geometry, too, raises difficulties: many of todays
students have had little or no exposure to traditional Euclidean geometry at school, so when
they meet the conic sections in gravitation theory they may be meeting them for the first time.
For whatever reasons, some leading textbooks today (for example, Tipler and Mosca (2004)
c 2007 IOP Publishing Ltd Printed in the UK
0143-0807/07/061189+09$30.00 

1189

1190

M Counihan

and Halliday et al (2001)) do not even attempt to present a full derivation of Keplers laws from
the inverse-square law (the Kepler problem) or vice versa (the inverse Kepler problem).
The central parts of this paper take the form of the outline of a presentation of classical
gravitation theory appropriate at the first-year university level. Rather than geometry, algebra
and vector calculus are used. The mathematics may be challenging for a first-year student,
but not excessively so, and it is the same sort of mathematics which the student will soon
need anyway to deal with electromagnetism. Starting from Newtons inverse-square law,
the beautifully simple circular form of a planetary orbit in velocity space (the hodograph)
is derived by direct integration. A second integration then shows that the orbit in ordinary
space is an ellipse. This is done simply in a few lines of calculation, but the method seems
to be practically unknown among teachers of physics. Then the concept of potential energy
is introduced and used to recast the theory in its more general and rigorous form, i.e. with a
potential field whose Laplacian is proportional to the matter density. The spacial distribution
of gravitational field energy is discussed. Finally, it is shown that, to take the mass-equivalence
of the gravitational field energy into account, the inverse-square law should be modified to
one which turns out to have Schwarzschild-like singularities.
2. Newtonian gravity as a central force

The force on a particle of mass m and position r due to another particle of mass M and position
R is given by
GMm
(r R).
(1)
F=
|r R|3
If M  m, then R may be considered to be fixed and to be the origin of coordinates, giving
an acceleration
GM
a= 3r
(2)
|r|
using F = ma. Denoting the particles velocity by v, the vector product v r is formed and
differentiated with respect to time:
d
(v r) = a r + v v.
(3)
dt
The second term on the right-hand side vanishes identically, and the first term also vanishes
if a and r are parallel, a condition which defines any central force including that described
by (1). So, v r is a constant, j, which is interpreted as the specific angular momentum of
the orbiting particle. It follows that the motion is restricted to a plane which is perpendicular
to j and which passes through the origin. In terms of circular coordinates r and describing
the position of the orbiting particle in that plane, j = v r is equivalent to
d
(4)
j = r2 ,
dt
which expresses Keplers second law. The next step is to rewrite (2) as
MG
(5)
a = 2 (cos , sin ).
r
Here, and in the vector equations below, the components of two-vectors in the plane of
the orbit are often displayed explicitly so as to make the procedure as clear as possible.
Using (4), (5) becomes
d
MG
(cos , sin ) .
(6)
a=
j
dt

Presenting Newtonian gravitation

This is integrated with respect to time:





MG
MG
d
a dt =
(cos , sin ) dt =
(cos , sin ) d ,
j
dt
j
giving
vu=

MG
(sin , cos ),
j

1191

(7)

(8)

where the velocity-like vector u is a constant of integration. Consequently,


|v u| =

MG
,
j

(9)

which is the equation of a circle in velocity space with its centre at v = u and with a radius
MG/j. What this means is that an orbiting particles path in velocity space, its hodograph,
is always a perfect circle. There is historical irony in this, because the early astronomers,
including Ptolemy and Copernicus, always insisted that planetary orbits should fundamentally
be circular, even though it was necessary for them to construct circles upon circlestowers
of epicyclesin order to model accurately the real movements of the planets. Even Kepler,
before he eventually realized that planets move in ellipses, was obsessed by the idea that the
heavens must be governed by spheres and circles for reasons of symmetry and simplicity. It
can now be seen that the ancient astronomers were, in a sense, right: but the circles are in v
space, not r space.
Represented in velocity space, then, the solar system is a set of nested but not quite
concentric circles. Each planets hodographic circle has its centre offset from the origin by
an amount u, and, as mentioned below, u is proportional to the eccentricity e of the Keplerian
ellipse.
The Sun, being stationary, is at the origin (v = 0). The innermost circles are those of
the most distant, and therefore slowest, planets. The outermost circles are those of Venus and
Mercury.
It may seem paradoxical that the planets hodographically closest to the Sun are those
which, in ordinary space, are furthest away. It comes about simply because the more distant
planets have lower speeds: v is proportional to r1/2. Another issue concerns the divergent
behaviour on the right-hand side of (9) as j 0, which arises from the divergence of the
gravitation force and potential energy as we approach the centre of attraction. This point is
commented on further below.
For each planet, the angular velocity around its hodographic circle (relative to the centre
of its circle, not to the Sun) varies, and at any moment is the same as the angular velocity that
it has around the Sun on its elliptical orbit in ordinary space (because the in (8) is the same
as the in (5)).
In the case of an object on an open hyperbolic orbit, the hodograph is still circular, but the
origin is outside the circle and only an arc of the circle is actually traversed before the object
slows down and asymptotically reaches a fixed final position on the circle. A parabolic orbit
corresponds to a hodographic circle which intersects the origin: the object goes once around
the circle, but takes forever to do so, with an angular velocity tending to zero at the origin.
The properties of the hodograph are still not widely appreciated, although they were
pointed out long ago by Mobius (1843) and shortly afterwards by Hamilton (1847). A
succession of later writers has pointed out the pedagogical value of the hodograph: for
example, Derbes (2001), whose paper included a calculation which comes close to that given
above. Apotolatos (2003) emphasized that the hodographic approach is useful not only in the
context of planetary orbits but for presenting other problems too, such as the movement of a

1192

M Counihan

projectile in a constant gravitational field and the brachistochrone problem. Fano and Fano
(1959) applied the idea to Rutherford scattering.
Continuing from (8), it is not difficult to find the shape of an orbit in ordinary space:
MG
(sin , cos )
(10)
v=u+
j
can be written as
r2
d MGr 2
d
v = (0, u)
+
(11)
(sin , cos )
j
dt
j2
dt
where (4) has once again been used and the directions of the coordinates have been chosen so
that u is in the negative y-direction. Integrating with respect to time,

v dt = r = r(cos , sin )


MG
u
d
d
2
=
r 2 (sin , cos ) dt
r (0, 1) dt 2
j
dt
j
dt


u
MG
=
(12)
r 2 (sin , cos ) d.
r 2 (0, 1) d 2
j
j
Now one does the trick of picking out the first component of this vector equation and
differentiating it with respect to , giving the separable differential equation
MG
dr
cos r sin = 2 r 2 sin
(13)
d
j
which can be solved straightforwardly, giving
1
j2
,
(14)
MG 1 e cos
where e is a positive constant of integration. Of course, (14) is the equation of a conic section,
and when e < 1 the orbit is an ellipse with eccentricity e. It is therefore demonstrated that
Newtons law of gravitation implies Keplers first law of planetary motion.
The second component of (12) may be used to show that e = uj/MG. In other words, u =
eMG/j, meaning that the constant u (by which the planets hodographic circle is offset from
the origin) is proportional to the orbital eccentricity e.
Incidentally, many of the above equations are divergent at j = 0 (zero specific angular
momentum). This occurs either for a closed orbit infinitesimally near to the centre of attraction,
or for motion limited to one dimension (the extreme case of a tight parabolic orbit). In any
case of j 0, velocities, accelerations and gravitational potential energies are divergent. All
this means is that Newtons inverse-square law diverges mathematically as the distance goes
to zero, and this will happen if the orbiting particle has no orbital angular momentum.
The manipulation of vector equations is not at first easy for students unfamiliar with them,
but basically the calculation above does not involve any calculus more complicated than the
integration of a sine, cosine, or reciprocal, and it should be understandable by any universitylevel student of physics. Moreover, it is a strictly Newtonian calculation in the sense that it
does not involve the concept of energy, which, of course, was not properly understood until
over a century after Newtons death, if then.
r=

3. Potential energy

Newtons inverse-square law of gravity is in some respects rather illogical and students may
well feel uncomfortable with it. One problem is that the Sun and the planets are represented as

Presenting Newtonian gravitation

1193

point objects when in fact they are small spheres, and at this stage it has not been proved that
a sphere behaves, gravitationally, as if its mass were all concentrated at its centre. Moreover,
students will soon discover that a point mass would have infinite negative gravitational potential
energy. Another problem lies with the central force approximation (M  m, leading to (2)).
If one considers the JupiterSun system, ignoring the other planets, then Jupiter and the Sun
must have linear momenta of equal magnitudes, regardless of how heavy the Sun might be;
so it is far from obvious at the outset that sensible results can be obtained by assuming that
the Sun is immobile. If one can pretend that linear momentum is not conserved, students may
wonder why they should take seriously the conservation of angular momentum represented
by (4).
It is advisable, therefore, to present the law of gravitation in a more rigorous and
general form. This is best approached in the conventional way by developing the concept
of the potential field. By a familiar argument, (1) leads to the potential energy function, or
work, U:
GM m
(15)
U =
|r R|
plus a constant of integration which is conventionally set to zero so that the potential energy
will tend to zero as the distance between the particles tends to infinity. The gravitational force
F is the negative of the gradient of U. When there are several particles pulling on the mass m,
then the potential energy attributable to the interaction between the mass m and all the other
masses mi is
  mi 
.
(16)
U (r) = G m
|r ri |
i
The specific potential energy (or just potential) , such that U = m, is now introduced:
  mi 
(r) = G
.
(17)
|r ri |
i
It should be emphasized to students that this represents a significant conceptual development.
U (r) is energy associated with a specific particle and it is defined only for the location r where
that particle happens to be at the time, but is defined at every point whether there is anything
there or not. U is an energy, measured in joules, but is a new kind of quantity whose units
(a squared velocity, but negative) have no name.
One now takes the crucial step of replacing the unphysical point masses mi by a continuous
mass distribution (r), so that

(r) = G ((r )/|r r |) d3 r .
(18)
V

This, again, is a significant conceptual development, and ironically it is one of which Newton
himself would probably have disapproved, since he was convinced (correctly) that matter is
corpuscular, not smeared out like butter across space. Paradoxically, one fiction (infinitesimal
point masses) is being replaced by another (non-atomic continuously-distributed matter) in
order for progress to be made. Taking the Laplacian of (18) with respect to r,



1
2

2
d3 r .
(19)
(r) = G (r )
|r r |
V
The identity

1
|r r |

= 4 3 (r r )

(20)

1194

M Counihan

is then used to give


2 (r) = 4 G

(r ) 3 (r r ) d3 r

(21)

and so
2 = 4 G.

(22)

This is the general expression of the Newtonian law of gravity, in a form which avoids the
mathematical and physical difficulties associated with infinitesimal point masses. Where there
is no matter,
2 = 0.

(23)

If there is spherical symmetry, then this is equivalent to




1
2
r
=0
(24)
r 2 r
r
or, writing g = /r,
2
(25)
(r g) = 0
r
which immediately gives the inverse-square law for the gravitational centripetal acceleration g.
One may write g = MG/r2, with M being a constant of integration which must be identified
as the mass of all the material which is distributed (spherically symmetrically, by assumption)
closer to the origin than the point in question. Note that there is no need to prove the
spherical shell theorem by long-winded direct integration: (25) immediately implies that any
spherically-symmetric distribution of matter will give rise to an inverse-square gravitational
field in an empty region outside it. Likewise, the interior spherical-shell theorem (that
there is no gravitational field inside a hollow spherical shell of matter) also follows directly
from (25).
What is the total potential energy of a distribution of gravitating matter? Note that the
potential (r) is the potential energy per unit of mass present at r as a consequence of its
interaction with the mass present at all other points. So, the total potential energy due to a
density distribution (r) is

1
(r) (r) d3 r.
(26)
UTOT =
2 V
The factor 1/2 is necessary to avoid double counting. Without the halving, the integral would
include both the potential energy of the material at position r due to its interaction with all the
matter elsewhere, and the potential energy of all the matter elsewhere due to its interaction
with what is at r, and the two are the same thing. Now, (26) suggests that the integrand might
perhaps represent the distribution of gravitational potential energy across space. In other
words, perhaps there is a gravitational potential energy density
Eg = 12 (r) (r).

(27)

However, there would be profound difficulties with this interpretation. It would mean that
gravitational potential energy is only located where there is matter, and is not present in empty
space. It would follow that in some situations (for example, as a comet plunges towards the
Sun) energy would have to be transmitted instantaneously between distant points without ever
occupying the intervening space. But there is an alternative interpretation: using (22), (26)
can be replaced by

1
(r) 2 (r) d3 r.
(28)
UTOT =
8 G V

Presenting Newtonian gravitation

1195

One may then make use of the case of Greens theorem which is usually written (e.g. Korn
and Korn (1967))



2
2
dV +
|| dV = ( ) dA.
(29)
V

Subject to certain differentiability conditions on the field , and when the system boundary is
chosen (at infinity, say) so that the surface integral may be neglected, this gives

1
||2 d3 r,
(30)
UTOT =
8 G V
suggesting that the spacial distribution of the gravitational potential energy may be
||2
8 G
or, defining g to be the gradient of ,
Eg =

(31)

g2
.
(32)
8 G
This has many advantages over (27). It conforms more closely to the spirit of general relativity,
with corresponding to an aspect of spacetime curvature. It is analogous to the energy density
of an electric or magnetic field, which is proportional to the square of the field strength at each
point. Above all (although this point will not be pursued further in this paper) it opens up the
possibility that, although gravitational forces between points A and B classically involve the
instantaneous propagation of information between the two points (action at a distance), they
need not necessarily involve the propagation of energy at infinite speed if gravitational energy
may be distributed throughout the intervening space between A and B according to a formula
along the lines of (32).
However, (32) has a curious status in physics. It is generally ignored in the literature, and
many physicists seem unaware of it. Most physicists do not appear to believe either (27) or
(32), but prefer to regard gravitational potential energy as having a mysteriously indefinable
locationbeing, as it were, a property of the system as a whole. It is difficult to understand
how this situation has arisen. Possibly the neglect of (32) has come about simply because the
energy density which it predicts is negative, and negative energies (alongside negative masses,
tachyons, temperatures below absolute zero, and what have you) have long been regarded as
part of the disreputable fringe of physics. Possibly there has been resistance to the idea that
energy may be located in completely empty space, unattached either to matter or to radiation.
Today, however, cosmologists are very familiar with the concept of dark energy or vacuum
energy.
Of course, (32) has not been proved above, and no experiments have directly detected
the negative energy density to which it refers. However, from a pedagogical point of view,
there is a strong case for accepting the energy distribution in (32) as the most physicallyreasonable of the possibilities which are open to us in order to present Newtonian gravitation
in a logically satisfying way. The idea that gravitational potential energy is in some sense
unreal, or inherently non-localized, seems to be an unnecessary and confusing alternative.
Eg =

4. A Schwarzschild-like solution

It is arguable that a completely consistent theory of gravitation should allow for the gravitational
effect of the field energy itself. In other words, the mass density on the right-hand side of

1196

M Counihan

(22) should be incremented with the mass-equivalent of the energy density in (31). Using the
massenergy correspondence
E = mc2 ,
we should modify (22) to read


Eg
2 = 4 G + 2
c

(33)

(34)

which, using (31), becomes


1
||2 .
(35)
2c2
It is important to understand that this is not an attempt to produce a hybrid semi-relativistic
version of the theory. Although historically (33) was revealed by the special theory of relativity,
the idea that energy is proportional to mass is not in itself necessarily a relativistic notion, and
it could in principle have been suggested in the context of 19th-century physicsfor example,
by arguing that a container of hot gas might weigh more than one of cold gas because of the
kinetic energy of the gas molecules. As long as it is admitted that energy might be equivalent
to mass for gravitational purposes, then something like (33) can be written down whatever the
value of the constant c might be.
In a vacuum, (35) gives
2 = 4 G

| |2
.
(36)
2c2
Looking for a spherically-symmetric solution, to represent the gravitational field outside a star
or planet, this reduces to
 2


1

1
2
r
=

.
(37)
r 2 r
r
2c2 r
2 =

Writing g = /r, this has the general solution


1
,
(38)
r(r r0 )
where r0 is a constant of integration. For large r, (38) approaches the standard Newtonian
inverse-square law for the field surrounding a mass M with
g = 2c2 r0

MG
.
(39)
2c2
So, the gravitational field is found to have a singularity not only at the origin r = 0 but
also at a radius r0 proportional to the mass M. This value of r0 is apparently not correct,
because it is only a quarter of the Schwarzschild radius 2MG/c2. However, it is remarkable
and instructive that a classical approach gives a Schwarzschild-like singularity at all, whether
or not the critical radius is different from that revealed by the general relativity. It is amusing
to think that in principle (38) could have been used to obtain the value of the massenergy
equivalence coefficient c2 by measuring the advance of the perihelion of Mercury long before
Einstein devised the theory of relativity.
r0 =

5. Discussion

However one connects Keplers laws of planetary motion with Newtons inverse-square law, a
certain amount of mathematics is unavoidable. It is possible to rely on calculus alone. It is also

Presenting Newtonian gravitation

1197

possible to use geometrical methods alone, with practically no calculus at all (e.g. Abelson et al
(1975)). Sivardi`ere (1985) advocated a combination of algebraic and geometrical arguments
in an approach to the inverse Kepler problem.
This paper has demonstrated a novel approach which starts out by using relatively simple
vector calculus to solve the Kepler problem by a method which has the advantage that the
circularity of the hodograph is proved at an early stage. In spite of its pedagogical usefulness,
the hodograph has remained obscure probably because it has generally been presented in the
form of an elegant but rather involved geometrical construction. Such an approach fitted the
mathematical culture of the 19th century, but todays physicists are mostly unfamiliar with
ruler-and-compasses geometrical proofs of the classical kind, and the graphical construction
of the hodograph therefore seems like very heavy weather. Even Richard Feynman, in the
Lost Lecture where he dealt with it (Goodstein and Goodstein 1996), did not make it much
easier.
Only after showing that the inverse-square law leads to a circular hodograph and to an
elliptical orbit do we introduce the thorny concept of gravitational potential energy. This
involves somewhat more advanced vector calculus, the main points of difficulty for students
being the Laplacian of r1 (20) and Greens theorem (29). It is also necessary for students
to understand the expression of the Laplacian in spherical polar coordinates which reduces,
when there is spherical symmetry, to (24). This is not simple mathematics, but if the effort
is made to introduce it to students at this stage there are worthwhile trade-offs: in particular,
the Newtonian theory of gravitation can be expressed with proper rigour, as in (22); the nature
and distribution of gravitational potential energy can be discussed sensibly; and added interest
can be provided by extending the theory to yield a Schwarzschild-like black hole solution
when the gravitational effect of energy itself is allowed for.
This subject has been thoroughly picked over by generations of physicists, so no claim
is made for the originality of any of the calculations presented here. It is likely that they
have been forgotten and independently rediscovered several times, as Derbes (2001) wrote
referring to the circularity of the hodograph. The purpose of this paper, therefore, has merely
been to show how the principles of elementary gravitation theory might perhaps be better
presented to todays university students.
References
Abelson H, diSessa A and Rudolph L 1975 Velocity space and the geometry of planetary orbits Am. J. Phys. 43 57989
Apotolatos T A 2003 Hodograph: a useful tool for solving some difficult problems in dynamics Am. J. Phys. 71 2616
Derbes D 2001 Reinventing the wheel: hodographic solutions to the Kepler problems Am. J. Phys. 69 4819
Fano U and Fano L 1959 Basic Physics of Atoms and Molecules (New York: Wiley)
Goodstein D L and Goodstein J R 1996 Feynmans Lost Lecture: The Motion of Planets Around the Sun (New York:
Norton)
Halliday D, Resnick R and Walker J 2001 Fundamentals of Physics (Extended) 6th edn (New York: Wiley)
Hamilton W R 1847 The hodograph, or a new method of expressing in symbolical language the Newtonian law of
attraction Proc. R. Irish Acad. 3 34453
Korn G A and Korn T M 1967 Manual of Mathematics (New York: McGraw-Hill)
Mobius A F 1843 Die Elemente der Mechanik des Himmels, auf Neuem Wege Ohne Hulfe Hoherer Rechnungsarten
Dargestellt (Leipzig: Weidmannsche Buchhandlung) sec. 22 p. 47
Sivardi`ere J 1985 On the inverse Kepler problem Eur. J. Phys. 6 2458
Tipler P A and Mosca G 2004 Physics for Scientists and Engineers (Extended) 5th edn (New York: Freeman)
Weinstein E R 2005 Central orbit Eric Weinsteins World of Physics Wolfram Research (scienceworld.wolfram.com)

You might also like