Researches in the Lunar Theory

Author(s): G. W. Hill
Source: American Journal of Mathematics , 1878, Vol. 1, No. 1 (1878), pp. 5-26
Published by: The Johns Hopkins University Press

Stable URL: https://www.jstor.org/stable/2369430

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms

The Johns Hopkins University Press is collaborating with JSTOR to digitize, preserve and
extend access to American Journal of Mathematics

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 RESEARCHES IN THE LUNAR THEORY,*

BY G. W. HILL, NlyaclI Turnpike, N. Y.

WHEN we consider how we may best contribute to the advancemnent of

this muclh-treated subject, we cannot fail to notice that the great majority of
writers on it have had before them, as their ultimate aim, the constructioni
of Tables: that is they have viewed the problem froin the stand-point of prac-
tical astronomy rather tl-an of mathematics. It is on this account that we
find such a restricted choice of variables to express the position of the moon,
and of parameters, in terms of which to express the coefficients of the peii-
odic terms. Again, their object compelling them to go over the whole field,
they have neglected to notice many minor points of great interest to the
mathematician, simply because the knowledge of them -was unnecessary for
the formnation of Tables. But the developments having now been carried
extremely far, without comnpletely satisfying all desires, one is led to ask
whether such modifications cannot be made in the processes of integration,
and such co6rdinates and parameters adopted, that a. much nearer approach
may be had to the law of the series, and, at the same tinme, their convergence
augmented.
Now, as to choice of codrdinates, it is known that, in the elliptic theory,
the rectangular codrdinates of a planet, relative to the central body, the axes
being parallel to the axes of the ellipse described, can be developed, in terms
of the time, in the following series,
=+0 1 (i-1)
x a C cosig,

i+ ? 1 (i-1)
b=t -.J ie sinig
i=-OO -/-2

a and b being the semi-axes of the ellipse, e the eccentricity, g the mean
anomnaly and, for positive values of i, the Besselian function (in Hansen's
notation)

A) 1.2.. _X_[_1._(i + 1) 1-.2. (i + 1) (i + 2)

* Co nunnunicated to the Natios al Academy of Sciences at the session of Apri

2 5

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
6 HILL, Researches in thze Lunar Theory.

while, for negative-values of i, we have

(_i) (i)

A -A

and, for the special case of i 0, we put the indeterminate

1 J-' = - -23e.
0 2
Here the law of tlle series is manifest, and the approximation can easily be
carried as far as we wish. But the longitude and latittude, variables em-
ployed by nearly all the lunar-theorists, are far from having such simple
expressions; in fact, tlleir coefficients cannot be expressed finitely in terms of
Besselian functions. And if this is true in the elliptic theory, how mliuch
more likely is a similar thing to be true when the complexity of the problem
is increased by the consideration of disturbing forces? We are then justified
in thinking that the coefficients of the periodic termns in the development of
rectangular or quasi-rectangular co6rdinates are less complex functions of
their parameters than those of polar co6rdinates. There is also another
advantage in employing co6rdinates of the former kind; the differential
equations are expressed in purely algebraic forms; while, with the latter,
circular functions inimediately present themselves. For these reasons I have
not hesitated to substitute rectangular for polar co6rdinates.
Again, as to parameters, all those who have given literal developments,
Laplace setting the example, have used the parameter m, the ratio of the
sidereal month to the sidereal year. But a slight examination, even, of the
results obtained, ought to convince any one that this is a inost unfortunate
selection in regard to convergence. Yet nothing seems to render this para-
meter at all desirable, indeed, the ratio of the synodic month to the sidereal
year would appear to be more naturally suggested as a parameter. Soine
instances of slow convergence with the parameter m may be given from
Delaunay's Lunar Theory: the development of the principal part of the
15
coefficient of the evection in longitude begins with the term -5 mne, and ends

with the term 41523108033 m8e; again, in the principal part of the co6fficient
of the inequality whose argument is the difference of the mean anomalies
of the sun and moon, we find. at the beginniing, the ternm 21 mzee', and, at

end the term 12074540264843

3538944
ml7ee'. It is probable that, by the adoption of so
fuinction of mi as a parameter in place of this quantity, whose numerical

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researches in thie Lunar Theory. 7

value, in the case of our mnoon, sh-ould not greatly exceed that of in, tlhe
foregoing large numerical co6fflcieuits nmight be very much diminished. Andl
nothing coinpels us to use the same parameter throughout; one may be used
in one class of inequalities, another in another, as may prove most advan-
tageous. It is known what rapid convergence has been obtained in the series
giving the values of logarithms, circular and elliptic functions, by simply
adopting new parameters. Similar transformations, with like effects, are,
perhaps, possible iin the coefficients of the lunar inequalities. However, a
far as my experience goes, no useful results are obtained by experitnenting
with the present knowvn developments; in every case it seems the proper
parameter must be deduced from a priori considerations furniished in the
course of the inte'ration.
With regard to the form of the differential equations to be employed,
althoughl DelauLnay's method is Avery elegant, and, perhaps, as short as any,
when one wishes to go over the whole ground of the lunar theory, it is subject
to some disadvantages when the attentioni is restrieted to a certain class of
lunar inequalites. Thus, do we wish to get only the inequalities whose coeffi-
cients depend solely on ni, we are yet cornpelled to develop.the disturbing
functiorL 1 to all powers of e. Again, the miethod of integrating by unde-
termined coefficients is most likely to give us the nearest approach- to the
law of the series; and, in this method, it is as easy to integrate a differential
equation of the second order as one of the first, while the labor is increased
by augmnenting the num-ber of variables and equations. But Delaunay's
method doubles the number of variables in order that the differential
equations may be all of the first order. Hence, in this disquisition, I have
preferred to use the equations expressed in terms of the co6rdinates, rather
than those in terms of the elements; and, in general, always to diminish the
number of unknown quantities and equations by augmenting the order of the
latter. In this way the labor of making a preliminary developnent of R in
terms of the elliptic elements is avoided.
In the present memnoir I propose, dividing the periodic developments of
the lunar co6rdinates into classes of terms, after the manner of Euler in his
last Lunar Theory,* to treat the following five classes of inequalities: -
1. Those which depend only on the ratio of the mean motions of the
sun and moon.

* Thzeoria Motiumn Lunce, nova methlo(lo pertractata. Petropoli, 1772.

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
8 HILL, Researckes in the Lutnar Tleory.

2. Those which are proportional to the lunar eccentricity.

3. Those which are proportional to the sine of the lunar inclination.
4. Those which are proportional to the solar eccentricity.
5. Those which are proportional to the solar parallax.
A general method will also be given by which these investigations imay be
extended so as to cover the whole ground of the lunar theory. My methods
have the advantage, which is not possessed by Delaunay's, that they adapt
themselves equally to a special numerical computation of the coefficients,
or to a general literal development. The application of both procedures
will be given in each of the just mientiolned five classes of inequalities, so
that it will be possible to compare them.
I regret that, on account of the difficulty of the subject and the length
of the investigations it seems to require, I have been obliged to pass over
the important questions of the limits between which the series are con-
vergent, and .of the determination of superior linits to the errors committed
in stopping short at definite poinits. There cannot be a reasonable doubt
that, in all cases, where we are compelled to employ infinite series in the
solution of a problem, analysis is capable of being perfected to the point of
showing us within what limits our solution is legitimate, and also of givTing
us a limiyit which its error cannot surpass. When the co6rdinates are devel-
oped in ascending powers of the time, or in ascending powers of a parameter
attached as a multiplier to the disturbing lorces, certain investigations of
Cauchy afford us the means of replying to these questions. But when, for
powers of the time, are substituted circular functions of it, and the coefficients
of these are expanded in powers and products of certain parameters pro-
duced from the comiibination of the masses with certain of the- arbitrary
constants introduced by integration, it does not appear that aniythi-ng in the
writings of Cauchy will help us to the conditions of convergence.

CHAPTER I.

_Diferential Egucations.-Properties of motion derived front Jacobi's integral.

We set aside the action of the planets anid the influence of the figures
of the sun, earth and inoon, together with the action of the last upon the

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researches in the Lutnar Theory. 9

sun, as also the product of the solar disturbing force on the mioon by the
small fraction obtained from dividing the mass of the earth by the. mass of
the sun. These are the same restrictions as those which Delauinay has
imposed on his Lunar Theory contained in Vols. XXVIII and XXIX of the
Memoirs of the Paris Academy of Sciences. The motion of the sun, about
the earth, is then in accordance with the elliptic theory, and the ecliptic is
a fixed plane.
Let us take a system of rectangular axes, having its origin at the centre
Df gravity of the earth, the axis of x being constanitly directed toward the
centre of the sun, the axis of y toward a point in the ecliptic 900 in advance
of the sun, and the axis of z perpendicular to the ecliptic. In addi.tion, we
adopt the following notation:
r' the distance of the sun from the earth;
- the sun's lonigitude;
- the sum of the masses of the earth and moon, measured by the
velocity these masses produce by their action, in a unit of time,
and at the unit of distance;
nt - the mass of the sun, measured in the same way;
- the mean angular velocity of the sun about the earth;
a' the sun's mean distance from the earth.
In accordance with one of the above-mentioned restrictions we have the
equation:

The axes of x and y having a velocity of rotation in their plane, equal

to dd' it is evident that the square of the velocity of the moon, relative to the

earth's centre, has for expression, in terms -of the adopted co6rdinates,

rx d'-t d dz'2
2 d YTdt + [dYt + x t + dz
- L dt +d2
dx2 + dy2+dz2 dXI xdy - ydx d2J2
- dt'2 dt~ dt + dt'
Th.e potential function, in terms of th

Q- - _) + X.na na
tt 9f~~~~~~2 '3 #2 aI3

,/ (X2 +y Z) V [(r_'X)2 +y2 + z-2] 2

If the second radical in this expression is expandled in a series
according to descending powers of r', and the first term n2a om
3 ri

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
10 HILL, Researches in the Lunar Theory.

disa,ppears in all differentiations witlh respect to the inoon's codrdinates, the

following expression is obtained

V(X2-+-zf?z2) + y +[Z22) 3 (y2+ z)]

+2 4 3 X

n2 an
+ n a [b 4 X2 (y/2 + z2) + 8(y2 2+ 92]

+ ?3 aG [i -a 5x3 (y2 + z2) + 1 X (!i2 + z2)21

a + .

Since the differential equations of motion are of the form

d dT dT dQ

dt d. d dq'
qp denoting, in succession, each of the variables which
the moon, it is plain that the term

2-d~ (X2 + y)
may be rernoved from T and added to Ql; and these modified quantities may
be denoted by the synmbols T' and n.'. Then these equations may be written
thus:

(12x t12JX dy d2 dQ'

-dt2 - dt -dt -dt2y -di,
d2y Ad dx d2W cQ
-- 2 T2 - ---c -

djt2 dtTt+-i -di

When we wish to restrict our attention to the lunar inequalities which
are independent of the solar parallax, all the terms, in the last expression of
Ql, which are divided by r'4, r'5, r'6, &c., may be omitted. In this case it w
be seen that all the terms, introduced into the differential equationis by the
solar action, are linear in forin, with variable, but known coefficients, since
dW d2Xad a'3
dt-,- yand 2 are known functions of 1.

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researches in thte Lw6nar Theory. 11

When all the inequalities, involving the solar eccentricity, are neglected,
the equations admit an integral in finite term-s. For, in this case, we have
d220
dt ='-dt' )r='

and El, does not explicitly contain t; hence, multiplying the equiations sever-
ally by the factors dx, dy, and dz, and adding the products, both members
of the resulting equation are exact differentials. Integrating this equation,
we have
dx2+ dy2 + dz2
2dt2- - =?1' + a constant.
This integral equation appears to have been first obtained by Jacobi.*
As it will be frequently referred to in what follows, I shall take the liberty
of calling it Jacobi's integral.
If we take two imaginary variables

- X + V(-+-Z2) V-i,.

s - x _ &y + Z2) i s//

Ql has the following simple expression. being a function of two variables only,
d2at3 If2a!3

V/6 V(r'-u) ( S) 2-2 -

If this is expanded in deseending powers of r', and, as before, the term

omitted,

i-^ t6 + if 2-jV3 [08 U2 + I US + ]

I a' ? [2 3+?13 U2S+1 382+ 5 S3

+
+ na
n2 '5[@ ,3 5 qU3S + 94 U2S2+ B5 US83 + -3 5E
. 2 2 3 1 . . . . . .

The additional variable, necessary to complete the definition of the moon's

position, may be so taken that the expression of T inay be simplified as miuch
as possible. This expression may be written

2 T- duds
dt2- 4(uti-s)2dtd2
( tydz-zdy)2 + 2dt
dAdt
xdy-ydx
dt2 + t2 (us - z2

* Compntes Rendus de l'Acuedmie des Sciences de Paris. T

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
12 HILL, Researches in thie Lutnar Tlheory.

There does not seem to be any function of x, y and z, which, adopted as a

new variable to accompany u and s, would reduce this to a very simple form.
However, when we are engaged in determining the inequalities independent
of the inclination of the lunar orbit, this transformation will be useful to us.
For, in this case, z - 0, and the values of u and s become
= x + y V-1i
sx -y V-1,
and T is given by the equation

2 T duds d2J uds-sdu dXu

d t2 d t dt dt2
Although Q is expressed most simply by the systems of coord'inates we
-have just employed, the integration of the differential equations will be
easier, if we suppose that the axes of x and y have a constant instead of a
variable velocity of rotation, the axis of x being made to pass through the
mean position of the sun instead of the true. To obtain the expression for

T correspondent to this supposition, it is necessary only to write n' for d-% in

the former values. As for .Q., it can be written thus
2 ,a3 ,2 J S
r+ [r2 2r2S +r2]' - a
wlhere
r X2 + y2 ? Z2 the square of the moon's radius vector;
S. x cos v +y sin v;
V = the solar equation of the centre.
This function being expanded in a series of descending powers of r', as before,
we have

= L +ln'2(x+9)
Y+ 1t 3 2 8(2. 2

e2 34 S2
+ le- -j
n/ 8 a' [ 33 r

f-,2-,L 55 145 r253 + s r4]

nt2 a'6

+..t3 C 385 35 r283 +15 1.48

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researchles in thte Lunar Theory. 13

Anid the corresponding differential equations are

d 2x dy (in,
dt2 (It dx

d Y + 2n' dx -In,
(It2 dit - y
d2z dfQ
d t2 dz'
Thus much in reference to the equations under as general a form as we
shall have occasion for in the present disquisition. We shall now suppose
that they are reduced to as restricted a form as is possible without their
becoming the equatio.ns of the elliptic theory; that is, we shall assume that
the solar parallax and eccentricity and the lunar inclination vanish. With
these simplifications, in the first system of coordinates,
T - dx2 + dgy2 + xwdy-ydx
2djt2 dt

Q'
_ ?_X2g
2 ;_ 2+ y2) + 2 7
and, in the second,
= duds n' uds - sdu
2dt2 dt

Qs - + 3W2 (q6+ S)2

And the differential equations, correspondent, are, in the first case,

dt2 2n dYt + rt-3n,2 x 0

I2y + 2n'd+4y 0
dt2 dit *r6
and, in the second,
d2u dt -

d2ts d 2 s - + U 3 3 2 ( + S) _O-

The Jacobian integral has severally the expressions

'A, ty2 _ + 3 n2X2 _C
2dt2 r n
duds --it 2(U +S)2 C
2dtV +/ R

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
14 HILL, Researcles in the Lunar Theory.

The terms -2n' dy 2n' dX &c., have been introduced i

dt 'dt
by making the axes of co6rdinates movable; but since the putting of n' = 0
makes the solar disturbing force vanish, there is no inconsistency in attribu-
ting themn to the solar action. Then, in the case of the vanishing of this
action, we have the equations of ordinary elliptic motion
d2x ~1
dt2+ r3

dtY +_ rt Y = ?.

Thus, in the restricted case we consider, all the terms, added to the differ-
ential equations of motion by the solar action, are linear in form and have
constant coefficients. This, and the circumstance that t does not explicitly
appear in the equations, are two advantages which are due to the particular
selection of the variables x and y. If 4 were constant, the equations would
r
be linear with constant coefficients and easily integrable.
The constants y and n' can be made to disappear fronm the differenltial
equations, if, instead of leaving the units of length and time arbitrary, we
assume them so that u = 1, and n' -1. The new unit of length, will then
be equal to ,34 units of the previous measurement. The equations, thus
simplifieed, are

dt22dt 3]x=o
d2y +2 dx 1 -O
with thelr integral
dX2 +dy2 2
=- + 3X2 --2C.
dt' r
It will be perceived that, in this way, we make the differential equations
absolutely the samlie for all cases of the satellite problem.
dlx
Let us put p = , then
dY 2 r 3x2-2C P ,2]-

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 H3ILL, Researches in tlhe Lunar Tlheory. 15

Or, by making.y the independent vTariable,

dxp

dy 2 + 3X2 -2C_ p

dp2 [3 ]x
dy [2+3x2 2C p212

The problem is then reduced to the integration of two differential equations

of the first order. Were this accomplished, and p eliminated fromn the two
integral equations, we should have the equation of the orbit. If we put

W 2x+ [r+3X2 S2C,p2jj2

the differential equations can be written in the canonical form,
dx dW
dy dp'
dp dW
dy dx
It may be worth while to notice also the single partial differential
equation, to the integration of which our problem can be reduced. Returnin
to the arbitrary linear and temporal u-nits, and, for convenience, reversing
the sign of C, if a function of x and y can be found satisfying the partial
differential equation
[dV+nfy2?['-nx2~
[d V 2
- +dnly
V o2=
+ n x2~t _+ 3'2X+
(x+y2)+ 2 ,20,
3n'2x2+-
and involving, a single arbitrary constant h, distinct from tllat which can be
joined to it by addition, the intermediate integrals of the problem -will' be
dx d TT dyd dVfn,
dt- dx dt dy
and the final integrals
dV dV
dA a ,K t + c
a an-d c being two additional arbitrary constants.
evident if we differentiate the four integral equations with respect to t and
compare severally the results with the partial differential coefficients of the
partial differential equation with respect to x, y, h and C.

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
16 HILL, Researchles tn the Ltunar Theory.

Although, in this mnanner, the problem seems reduced. to its

terms, yet, when we essay to solve it, setting out withl this partial differential
equation, we are led to more conmplex expressions than would be expected. It
would be advisable, in this method of proceeding, to substitute polar for
rectangular co6rdinates, or to put
x - r cos p, y r sin p.
The partial differential equation, thus transformed, is
dd
T72
2 + 1 d'lV k
[tf q,]2 2~ +S 2
_n'2r + q2r2
2C+2+ C
~ 3r2
Wt2r2t
cos
dr r d r

This would have to be integrated by successive approximations, and it is

found that this method, which, at first sight, seems likely to afford a briefer
solution of the problem, because but one unknown quantity was to be deter-
mined, and this free from the variable t, and involving only half of the nunmber
of arbitrary constants introduced by integration, when developed, leads to as
complex operations as the older methods, and, moreover, has the disadvantage
of giving results which neecl prolix transformations before the co6rdinates can
be exhibited in terms of the time.
Althouoh we shall make no use of the equations in terms of polar
co6rdinates, they may be given here, for the sake of some special propertie
they possess in this form. They are

rrd 2r /1L 2nWr2dO+ W3n2r2COS2p O,

dt2 dt2 dt r

dj [r2 (dp+ )] + n'2r2 sin 2q -O

with their integral

dr2 + r2dcp2 2y + U2r2 COS2 (p - 2C,

dt2 -r 3'rcs-t
By dividing the second of the differential equations by r2
separated, andc X denoting the longitude of the mnoon, we
-3 n 2 fsin 2 (A-A) dt
r _ E ~~~~~ dt

IdX
wdt

K being a constant. Thus, after the longitude is determiined in terms of t,

the radius -vector is obtained by a quadrature. But it can also be found,
without the necessity of an integration, by dividing the' integral by r2 and

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 IIILL, Researckes in thte Lunar Thteory. 17

1 dr'
then elimninating the term -2 dt2 by mieans of its value derivedl from the

second differential equation; in this way we get

it = L (I td 2 d(pt2 eo2 P
r3 r2 __ j -- n'2 cos%2
dt

As we desire to malke constant numerical application of the general

theory, established in what follows, to the particular case of the inoon, we
delay here, for a moment,i to obtain the numerical values of the three
constants ~t, n' and C. The value of yi may be derived either from thLe
observed value of the constant of lunar parallax combined with the mnean
angular motion of the moon, or from the intensity of gravity at the earth's
surface and the ratio of the moon's mass to that of the eartlh. WXTe will adopt
the latter procedure. The value of gravity at the equator, g = 9.779741
metres, the unit of timne being, the mean solar second. We propose, however,
to take the mean solar day as the unit of time, and the equatorial radius of
the earth as the linear unit. This nlumber must then be multiplied by

637473 15' (6377397.15 metres is Bessel's value of the equatorial radius

Moreover, the theory of the earth's figure shows that, in order to obtain the
attracti ve force of the earth's mass, considered as concentrated at its centre
of gravity, a second multiplication must be made by the factor 1.001818356.
WTitlh our units then this force is represented by the niumber 11468.338: and
the moon's mass beinr taken at 1 of the earth's, her attractive force is
n ~~~81.52277

represented by the number 140.676. Consequently

=U 11609.014.
The sidereal mnean motion of the sun in a Julian year is 1295977".41516,
whence
n' ~ 0.017202124.

The vTalue of C might be obtained fromn the observed values of the moon's
co6rdinates and their rates of variation at any tirne. However, as the eccen-
tricity of the earth's orbit is not zero, C obtained in this manner would be
fouind to undergo slight variations. The mean of all the values obtained
in a long series of observations might be adopted as the proper value of this
5

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
18 IHILL, Researches in the Lunar Thteory.

quanitity when regarded as constant. But it is much easier to derive it

approximately from the series

2C= (tn)'3[1 +2m m 5 i n2 m3, 1319 n4 7m 5_ 287rn6_ 1321 mi7],

which will be established in the following, chapter. Here n deenotes the
mooni's sidereal mean motion, and in is put for ,. In this fo'rmula the

terms which involve the squares of the lunar eccentricity and inclination and'
of the solar parallax are neglected; this, however, is not of great momnent,
as, being multipled by at least in2, they are of the fourth order with respect
to smallness. The. observations give n 0.22997085, hence
C - 111.18883.
If it is proposed to assume the units of time ancl length so that yt and
n' may both be unity, it will be found that the first is equal to 58.13236
mnean solar days, and the secondl to 339.7898 equatorial radii of the ear
The corresponding value of C is 3.254440.
Let Ius now notice some of the properties of motion which can be derived
fromn Jacobi's integral. This integ,ral gives the square of the velocity relative
to the moving axes of co6rdinates; and, as this square is necessarily positive,
the putting it equal to zero g,ives the equation of the surface which separates
those portions of space, in which the velocity is real, from those in which
it is imaginary. This equation is, in its most general formn,

+
+/(2+Y2 + Z2) naf
s/(tX)2 + y2 + z2] 2
which is seen to be- of the 16th degree. As y and z enter it only in even
powers, the surface is symmetrically situated with respect to the planes of xy
and xz. The left member is necessarily positive, (the folds of the surface, for
which either or both the radicals receive negative values, are excluded from
consideration), hence the surface is inclosed withini the. cylinder whose axis
passes througlh the centre of the sun perpendicularly to the ecliptic, and
whose trace on this plane is a circle of the radius

a'V/(3 + 2C,
As, in general, the second term of the quantity, under the radical sign, is
much smaller than the first, this radius is, quite approximnately V3 a'. Thus,

in the case of our moon, assuming , = sin 8".848, we have this radius =
a

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researches in the Lunar Theory. 19

V/3.001383a'. It is evident that, for all points without this cylinder, the
velocity is real; and as, for large values of z, whether positive or negative,
the left muenmber of the equation becomes very small, it is plain that the
cylinder is asymptotic to the surface. Every right line, perpendicuiar to the
ecliptic, intersects the surface not mnore than twice, at equal distances from
this plane, once above and once below.
Let us now find the trace of the surface on the plane of xy. Putting
p for the distanice of a point on this trace from the ceintre of the sun,
p2 (a'- X)2 + y2
and it is evident that the cubic equation,

p3 af2 (3+4 2C)p 2a'3,

will give the limits between which the values of p can oscillate. If C is
negative, this equation has but one real root which is negative; consequently,
in this case, the surface has no intersection with the plane of xy. But, in all
the satellite systems we know, C is positive, and this conditioni is probably
necessary to insure stability. Hence we shall restrict our attention to the
case where C is positive. Then all the roots of the equation are real, and
two are positive. It is between the latter roots that p mylust always be found.
To compute them, we derive the auxillary angle 0 from the formula

sill [13--at]

or, since 0 differs but little from 900, w

C.~~~~~~ 2I qf2at2[ + /q2at2 n 2 xt4a

Cnos 20 C
[1+ 3 22a]

or, as '2 '2 iS a small quantity, with sufficient a.pproxiination from

2 2~ ~ ~ ~~~~q
I2- -_c
n a ~ ~ ~ ~ ~ ~~
Cos 0 ,2,
+2 c
+ 3 - -a,

The two roots are then

v1=: 2a\U(1 + 3 2- sin-,

p2 2aV(1+ X 2) sin 0

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
20 HILL, Researches in the Lunar Tkeory.

The trace of the surface on the plane of xy is theni wholly comprised in the
annular space between the two circles described from the centre of the sun as
centre with the radii pi and p2. Moreover, as in mnost satellite systemns we
have , equal to a very small fraction, (for our moon = 1
a, ~~~~~~~n a 322930 2)
it is plain that, for points whose
their distance from the su-n, the trace is approximately coineident witlh these

circles. For the term ' in the equation, may theni be neglected in coi11-
r
parison with the other termus.
In the case of our mioon there is found
0 870 52' 11"L.53.
and hence
p, = 22815.15. p2 = 23816.09,
and, if r and p are regarded as the variables defining the po
in the plane xy, the following table gives some correspond
(juantities, for each of the two branches of the trace appro
to the two circles.
r. p. r. p.
433.3257 22878 69 439.7922 23751.81
450 22876.17 450 23753.37
500 22869.68 500 23760.04
600 22860.13 600 23769.85
1000 22841.59 1000 23788.87
10000 22817.70 10000 23813.43
46127.70 22815.68 47127.55 23815.53

rThe first anid last values correspond to the four points wlhere the curves
intersect the axis of x on the hither and thither side of the sun. It will be
seen that the approximation of the branches to th7e circles is quite close,
except in the vicinity of the earth, where there is a slight protruding away
froin themn.
In addition to these two braniches of the trace, there is, in the case where
C exceeds a certain limit a third closed one about the origin nmuch smaller
than the forimer. As the co6rdinates of points in this branch aire small
fractions of a', its equatioi may be written, quite approximately,

(1 _C a _2X2
r

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 IIILL, Researches in the Lunar Theory. 21

It intersects the axis of y at a distance from the origin

YoYc
=

and the axis of x at points whose co6rdinates are the smallest (without regard
to sign) roots of the equations

x a'-
+ n x c+ 3 n, 2a! 2 _ i// (at,_ X)2
i2at3
LI+ fl'2 ,- X
- a-

For the moon these quantities have the values

YO 104.408, x - 109.655, x2 +109.694.
This branch then does not differ nmuch froin a circle having its centre at the
origin, inore closely it approximnates to the ellipse whose major axis x2-XI,
and minor axis 2yo.
The value of the co6rdinate z, for the single intersectioni of the surface
with the axis of z above the plane of xy, is given by the single positive root
of the equation
__ a3 = n',2a'2.
z V (a'2 + z2)
For the moon the nutmerical value of this root is
zo 102.956.
The intersection of the surface with the perpendicular to the plane
passing through the centre of the sun is, in like manner, given by the eq
/C~~ g t2at3
+~~~C naf a
V(a' 2 +z2) z
having but a single positive root, which is nearly
2 a'
zo

1 + !T __ a
or, with less exactitude,
o a'.
From these investigations it is possible to get a tolerably clear idea of
the form of this surface. When C exceeds a certain limit, it consists of three
separate folds. The first being quite small, relatively to the other two, is
closed, surrounds the earth and somewhat resembles an ellipsoid whose axes
have been given above. The second is also closed, but surrounds the sun,
6

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
22 HILL, Researches in the Lunar Theory.

and has approxinmately the form of an ellipsoid of. revolution, the semiaxis ini
the plane of the ecliptic being somewhat less than a', and the semiaxis of
revolution perpendicular to the ecliptic and passing through the sun being
about two-thirds of this. This fold has a protuberance in the portion neigh-
boring the earth. The third fold is not closed, but is asymptotic to the
cylinder mentioned at the beginning of the investigation of the surface. Like'
the second, it also is nearly of revolution about an axis passing through the
cenitre of the sun and perpen-dicular to the ecliptic. The radius of its trace
on the ecliptic is about as much greater than a', as the radius of the trace
of the second fold falls short of that quantity. The fold has a protuberance
in the portion neighboring the earth, and which projects towards this body.
The whole fold resembles a cylinder bent inwards in a zone neighboring the
ecliptic.
'What modifications take place in these folds when the constants involved
in the equation of- the surface are inade to vary, will be clearly seen from
the following exposition. Let us, for brevity, put

h = 3 + 2 2,2,

and, for the moment, adopt a', the distance of the earth from the sun, as the
linear unit, and transfer the origin to the centre of the sun, and moreover put
7 2 '3
na'
Then the intersections of the surface, with the axis of x, will be giveni by th
two roots of the equation
x4 _' - hx2 + (h + 2-27) x--2 - O
which lie between the limits 0 and 1; by the two roots of
x4 - TI - hX2 + (h + 2 + 27) x - 2 - 0
which lie between I and V/h; and by the two roots of
X4- C3 - hx2 + (hA- 2 - 2y) x + 2 0,
which lie between 0 and-Vh.
Hence, if C diminishes so much that the first of these three equati
has the two roots, lying between the mentioiied limits, equal, the first fold
will have a contact with the second fold; and if C fall below this limit, the
roots become imaginary, -and the two folds become one. Again, if C is
dimyinished to the limit where the second equation has the mentioned pa
roots equal, the first fold will have a contact with the third; and when Cis

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researches in the Lunar Theory. 23

less than. this, these two folds form but one. And when C is less -than both
these limits, there will be but one fold to the surface.
In the spaces inclosed by the first and second folds the velocity, relative
to the moving axes of co6rdinates, is real; but, in the space lying between
these folds and the third fold, it is imnaginary; without the third fold it is
again real. Thus, in those cases, where C and y have such values that the
three folds exist,. if the body, whose motion is considered, is found at any
tim.i-e within the first fold, it must forever remain within it, and its radius
vector will have a superior limit. If it be found within the second fold, the
same thing is true, but the radius vector will have an inferior as well as a
superior limit. And if it be found without the third fold, it must forever
remain without, and its radius vector will have an inferior limit.
Applying this theory to our satellite, we see that it is actually within the
first fold, and consequently must always remain there, and its distance fromn
the earth can never exceed 109.694 equatorial radii. Thus, the eccentricity
of the earth's orbit being neglected, we have a rigorous demonstration of a
superior limit to the radius vector of the moon.
In the cases, where C and y have such values that the surface forms but
one fold, Jacobi's integral does not afford any linmits to the radius vector.
When we neglect the solar parallax and the lunar inclination, the pre-
ceding investigation is reduced to much simpler terms. The surface then
degenerates into a plane curve, whose equation, of the sixth degree, is

- C- n2X2
r

It is evidently symmetrical with respect to both axes of coordinates, and is

,contained betweenr the two right lines, whose equations are

_ + 12C

ancl which are asymptotic to it. It intersects the axis

whose coordinates are

y +~M

The cubic equation,

r3- ,2C r + 2t -_ 0
3,2 2

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 24 IIILL, Researches in the Lunar Theory.

gives tlle values of r, for whichl the curves intersect the ax

(2 C) -2 > 9yiz, _
this equation lias two real roots between the limits 0 and + - If

(2C)2- 9yn=
these roots become equal. And if'

(2C') -!-< 9yn',

there are no real roots between these limits, and the curve lhas no intersection
with the axis of x. The figures belowv exhibit the three varieties of this curve.

FA. I

Fig. 1 represents the form of the curve in the case of our mnoon. In Fig. 2
we see that the smrrall oval of Fig. 1 has enlaroed and elong,ated it-self so as to
touclh the tvo infinite branches; while, in Fig. 3, it has disappeared, the
portions of the curve, lying on eitlher side of the axis of x, having lifted
themselves away from it, and the angles having become rounded off. In Fig.
1, the velocity is real within the oval, and also without thee inflnite branches,
but it is imiaginary in the portion of the plane lying between the oval and
these branches. Hence, if the body be found, at any timne, within the oval,
it cannot escape thence, and its radius vector will have a superior limnit; and,
if it be found in one of the spaces on the concave side of the infinite branches,
it cannot remove to the other, and its radius vector will have an inferior
lirriit.

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
 HILL, Researches in the Lunar Theory. 25

In the case represented in Fig. 2, the same things are true, but it seems
as if the body might escape from the oval to the infinite spaces, or vice versa,
at the points where the curve intersects the axis of x. However, at these
points, the force, no less than the velocity,- is reduced to zero. For the
distance of these points from the origin is the positive root of the equation

2 2-C 0,
3r--
3n'2
or

A2C f 9ln'
.3nf 3n"

and this value is the same as that given by the equation

4,,,'3n/2 _0.
r

In consequence the forces vanish at these two points, and thus we have two
particular solutions of our differenitial equations.*
In the case represented in Fig. 3, the integral does not afford any
superior or inferior limit to the radius vector.
The surface, or, in the more simple case, the plane curve, we have
discussed, is the locus of zero velocity; and the surface or plane curve, upon
which the velocity has a definite value, is precisely of the same character and
has a similar equation. It is only necessary to suppose that the C of the
preceding formulae is augmented by half the square of the value attributed
to the velocity. Thus, in the case of our moon, it is plain the curves of equal
velocity will form a series of ovals surrounding the origin, and approaching
it, and becoming more nearly circular as the velocity increases.
Applying the simple formulao, where the solar parallax is neglected, to
the moon, we find that the distance of the asymptotic lines, from the origin, is

2C = 500.4992.

The distance of the points on the axis of x, at which the inoon would remain
stationary with respect to the sun, is

\'3ny?2 =235.5971.

* The corresponding solution, in the more general problem of three bodies, may be seen in the

Mecanique 6eleste, Tom. IV; p. 310.

7

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms
26 HILL, Researches in the Lunar Theory.

If the auxillary angle 0 is derived from the equation

sin - (2_) a

(2(J) 12
we get
0 -320 49' 6"1.63;
and the distances from the origin, at wlich the curve of zero velocity inter-
~ects the axis of x, -are given bv the two expressions
2 V2C 0
3 Sn si
3ie 3

2,/C i (co - 0)
and the numbers are 109.6772 anId 435 5623. These values differ but little
fromi the previous more general determinations.

(To be continued.)

This content downloaded from

49.36.100.169 on Tue, 14 Sep 2021 14:48:48 UTC
All use subject to https://about.jstor.org/terms