ANALYTICAL EPHEMERIDES OF THE MOON

IN THE 20 TH CENTURY

Michelle Chapront-Touzé and Jean Chapront

2002

SYRTE – Observatoire de Paris

UMR 8630 – CNRS
Paris Observatory Lunar Analysis Center

61 avenue de l'Observatoire
75014 Paris, France
2
ANALYTICAL EPHEMERIDES OF THE MOON IN THE 20TH CENTURY

Michelle Chapront-Touzé and Jean Chapront

SYRTE – Observatoire de Paris
UMR 8630 - CNRS
Paris Observatory Lunar Analysis Center

1. INTRODUCTION
In the middle of the 18th century, the best lunar ephemerides - Mayer's tables - were derived from expressions of the
coordinates: longitude, latitude, and parallax, under the formulation :
sin
σ = ελ + ∑ Ai φi (1)
i cos

in which arguments φi were provided by lunar theory and numerical coefficients Ai fit to observations. λ is the lunar
mean mean longitude ; ε is 1 for longitude and 0 for other coordinates. Lunar tables of Mayer's type - Burckhardt's
tables (1812) - were still used in French and English national ephemerides, Connaissance des temps and Nautical
Almanac, in 1861 (Explanatory supplement, 1961 ; Bureau des longitudes, 1997). Tables based solely on a lunar theory,
except for a small number of necessary constants, were introduced for the first time in the above mentioned national
ephemerides in 1862, with Hansen's tables (1857). In the following, we shall denote by analytical ephemerides, tables
or other means of computing a large number of lunar positions essentially based on analytical expressions provided by
theory. Analytical ephemerides existed before Hansen's tables and may have been used in certain works, but early lunar
theories were not precise enough to make them competitive with tables of the Mayer's type.
Hansen's tables were still in use at the beginning of the 20th century, with corrections due to Newcomb, and the
general use of analytical ephemerides continued till the last decades, in national ephemerides and in studies using lunar
positions, but with many changes.
First, lunar theory was considerably improved by Brown (1899 to 1910), especially in the case of planetary
perturbations and Brown's theory became the basis of lunar ephemerides from 1923 till 1984.
Formulation also changed. At the beginning of the century, several years of hand computation were necessary
to transform the expressions provided by lunar theory to the tables which were used for the practical computation of
lunar positions. The advent and progresses of electronic computers made the expressions provided by lunar theory
usable without any transformation (Eckert et al., 1954).
In the second part of the 19th century, lunar theory, with the works of Adams and Delaunay, had contributed to
put into evidence the irregularity of the Earth's rotation. In 1952, the adoption of an 'uniform' time scale in lunar
ephemerides, instead of mean solar time, made empiric terms previously introduced in lunar tables to disappear. At the
same time the tidal secular acceleration was introduced in longitude, with the observed value derived from Spencer
Jones works (1939).
Brown's theory is the last 'hand computed' theory. From the 1950s, the powerful tools that are electronic
computers, have been used to build new theories of solar system bodies. Several valuable works concerning lunar
theory have been published. If the authors' theory, ELP, has been the only one to take into account the numerous
perturbations which contribute to a precise lunar theory, the other studies, in addition to their theoretical interest, have
been very helpful by providing useful comparisons. Lunar theory has been also dependent of the simultaneous
progresses in the theories of planetary motions and lunar librations and on a better knowledge of the physical
parameters of the Earth and Moon.

3
 Simultaneously, the development of electronic computers has allowed the use of purely numerical methods of
integrations. This approach is the most direct and the most precise one, on the time interval that the integration covers.
This is the solution that has been chosen by the Jet Propulsion Laboratory which produces regularly a set of
ephemerides for the Moon and planets such as DE102 (Newhall et al., 1983), DE200, or more recently DE405
(Standish, 1998a). In order to take into account the interactions between the various motions, one has to integrate the
whole problem (Moon, planets, lunar libration) including simultaneously all the forces (figures of the bodies, tides,
relativity,…).
The success of lunar ephemerides generated by numerical integrations (numerical ephemerides) is connected
with the advent of a new method of lunar observations: the measurements of the Earth-Moon distance with Lunar Laser
Ranging (LLR) afford today the most precise observations of the Moon. They provide at a given time, the distance
separating a station of observation located on the Earth's surface from a reflector located on the lunar surface. The
ranging accuracy, which was about 30 centimeters in 1970 is now below the centimeter level, representing a relative
accuracy better than 10-10. Such observations require, among numerous contributions, a very precise representation of
the lunar orbital motion and inversely contribute to the accuracy of lunar ephemerides by providing accurate values of
constants. Though ELP analytical solution is more precise than Brown's solution, it does not reach the level of precision
needed by LLR observations, contrary to numerical ephemerides.
Nevertheless, analytical and numerical methods should not be opposed but rather considered as complementary
approaches: numerical solutions afford the utmost precision while analytical solutions allow to isolate selected
components of the perturbations in order to analyze the effects of different models. Furthermore, since the lack of
precision of analytical theories is essentially due to planetary perturbations, it is possible to complete analytical theories
with numerical complements under the form of Chebychev polynomials. The so obtained solutions are usable for fits to
LLR observations and preserve the advantages of analytical solutions.
In this paper, Sect. 2 and Sect. 3 describe the analytical theories from which analytical ephemerides of the
Moon are derived, while Sect. 4 and Sect. 5 are devoted to the analytical ephemerides themselves.

2. LUNAR THEORY: THE MAIN PROBLEM

2.1 The Keplerian problem
If the Moon was governed by the Earth's attraction alone and if the two bodies were homogeneous spheres, the Moon M
would describe a Keplerian ellipse whose Earth's mass center E is a focus. This ellipse is characterized by its semi-
major axis a, its eccentricity e, the inclination i of its plane on a fixed reference plane defined by two axes Ex and Ey,
the longitude Ω of its ascending node on the reference plane, reckoned from Ex, and the angle ω (argument of perigee)
in the lunar orbit from the ascending node to the perigee. We denote by ϖ the longitude of perigee: ϖ = ω + Ω. All these
quantities are constants. The position of the Moon in its orbit can be computed by means of a development involving e
and the mean anomaly l = l 0 + nt , t being the time reckoned from a fixed epoch and n a constant tied to a by Kepler's
third law. Lunar longitude V and latitude U referred to the reference plane and Ex are expressed by :
1 5
V = λ + ( 2e − e 3 + ...) sin l + ( e 2 + ...) sin 2 l + ( −γ 2 + ...) sin 2 F + K
4 4 (2)
U = ( 2γ − 2γe 2 + ...) sin F + ( 2 γe + ...) sin(F + l) + ( −2γe + ...) sin(F − l ) + K

where λ = l + ϖ is the mean longitude and F =l + ω is the argument of latitude. We have noted γ = sin i/2. Lunar radius
vector (or distance) r is given by :
1 2 3 3 1 2
r = a[1 + e + ... + ( − e + e + ...) cos l + ( − e + ...) cos 2 l + K] .
2 8 2
It is sometimes replaced by the parallax p defined by the relation: sin p =a E /r, where aE is the equatorial Earth's radius.
Keplerian ellipse cannot be regarded as a starting approximation for the lunar orbit because the periods of
revolution of perigee and node are very short; hence, it is wise to replace it by an ellipse rotating with precessional
motions in ω and Ω, the periods being about 6.0 years and 18.6 years respectively.

2.2 Definition and formulations of the main problem

The main problem is a better approximation of the complete perturbed problem. It is restricted to the Earth E, the Moon
M, and the Sun S, regarded as point masses, in the absence of other disturbing forces. Moreover the apparent orbit of
the Sun around the Earth-Moon barycenter is supposed to be a Keplerian ellipse whose plane (the ecliptic) is parallel to

4
the reference plane. Note that the plane of the Sun's apparent orbit may be a fixed ecliptic (the mean ecliptic of a given
epoch) or the mean ecliptic of date. In the latter case, the reference plane is supposed to move in the same way, but the
induced corrections are not taken into account in the main problem and the results are the same whatever the ecliptic is.
The heliocentric orbit of the Earth-Moon barycenter is described by four parameters: a', e', ϖ', and l ' = l ' 0 + n ' t , n' being
tied to a' by Kepler's third law. The parameters of the Sun's orbit are the same except that the longitude of Sun's perigee
is ϖ'+180°. The main problem solutions are developed under the condition that they remain quasi periodic, i.e. without
Poisson terms which should grow rapidly with the time because of the relatively short periods of node and perigee. The
determination of the secular motions of node and perigee is derived from that condition.
Hence, the solutions σ of the main problem (e.g., coordinates V, U and r) are expressed as series (1) in which
φi are linear combinations of four fundamental arguments whose variations are proportional to time, and Ai constant
coefficients. Since Delaunay, the adopted formulation is :
sin
σ = ελ + ∑ A j1 j2 j3 j4 cos ( j1 D + j2 F + j3l + j4 l ' )
j1 j 2 j3 j 4 (3)

D, F, l being given by:

D = λ − l '−ϖ' +180 °
F = λ− Ω
l = λ− ϖ
where λ is now the mean mean longitude, ϖ the mean longitude of perigee, Ω the mean longitude of node, and l the
mean mean anomaly. ε is 1 or 0 according to coordinate σ. The variations of λ, ϖ, and Ω are proportional to time. The
main problem solution involves six integration constants which are traditionally: the values of λ, ϖ, and Ω at the epoch,
the mean motion n of λ and two constants tied to eccentricity and inclination. Delaunay (1867) introduced in his final
expressions the constants e and γ so that the coefficients of sinl in V and sinF in U retain the Keplerian form of eq. (2).
Brown used different constants e and k whose expansions in e and γ with numerical coefficients are given in (Eckert et
al., 1966). At the zero order, e is approximately equal to 2e and k to γ. The values of the coefficients of sinl and sinF
themselves or their half values E and Γ may also be used as eccentricity and inclination constants (Chapront-Touzé,
1974). Note that, at the beginning of the 20th century, theories introduced the constant term of the sine of parallax Π0 as
a complementary constant. Π0 depends on n, but also on the geocentric constant of gravitation, Earth's radius and ratio
of Earth's and Moon's masses which were not accurately known.

In completely analytical theories coefficients A j are sums of monomials expressed with the small literal
1 j 2 j 3 j4

parameters e, e', γ, α and m (plus eventually the mass ratio Moon/Earth). m stands for the ratio n'/n and α for the ratio
a/a', a being tied to n by Kepler's third law. We denote all these quantities as metric constants. The orders of magnitude
of the small quantities are 0.055 for e, 0.015 for e', 0.045 for γ, 0.0025 for α, and 0.076 for m. But the very slow
convergence of the developments in powers of m restricts the precision of the solutions if one wants to limit the sums in
coefficients A j j j j to a reasonable number of monomials. This is the case of Delaunay's theory (1860, 1867) which is
1 2 3 4

valid to the seventh order with respect to the whole set of parameters.

In the case of Hill-Brown solution all the parameters are kept literal except m that receives a nominal value m0
to get rid of the slow convergence. Such a theory can be considered as partially analytical.

In semi-analytical theories, the approach is slightly different. The values of the parameters are regarded as the
sum of nominal values (e0 , e'0 , γ0 , α0 , m0 ), and literal increments δe, δe', …, δm ; coefficients A j j j j are developed
1 2 3 4

with respect to these increments. Hence the solution remains analytical in a tube of trajectories around the nominal
solution computed for the set (e0 , e'0 , γ0 , α0 , m0 ).
Completely analytical and semi-analytical theories have two advantages. The first one is the ability of fitting
all the constants to observations and introducing the fitted values in the theory in order to get an ephemeris. In semi-
analytical theories, as far as the nominal values of the constants are close to the observed values, the first order
development of A j j j j with respect the analytical increments is sufficient. Nevertheless, in the case of partially
1 2 3 4

analytical theories, giving to m a numerical value in the coefficients of the series is not a disadvantage for that purpose

5
since, lunar and solar mean motions being very accurately known, it is sufficient to introduce the corrected value in the
mean mean longitude and in the arguments of the leading terms.
A second advantage of completely analytical and semi-analytical theories is to provide partial derivatives with
respect to constants. Indeed those quantities are necessary in the computation of perturbations to the main problem (due
to planets, Earth's figure, etc…), and Brown (1903) had to perform a special treatment for the derivatives with respect to
m which are not available in his theory.
An other apparent advantage of a completely analytical solution of the main problem is its 'universal character':
Once the effort has been done to construct a lunar solution it can then be used for many other satellites. In fact, this
consideration is quite optimistic. Because of the slow convergence of the series, in particular if m is not sufficiently
small, the solution cannot be used efficiently with a good accuracy in other satellite problems. Actually, such
applications have seldom been realized, as for example in the tables of the faint satellites of Jupiter JVI and JVII by
Bobone (1937a, 1937b) using Delaunay's lunar theory.
In series (3) for longitude, several types of arguments appear, related to small divisor problems in the
integration of the differential equations (see examples in Table 3) .
• Monthly arguments: j1 + j2 + j3 ≠ 0 ;

• Annual arguments: j1 + j2 + j3 = 0 and j1 − j4 ≠ 0 ;

• Long period arguments: j1 + j2 + j3 = j1 − j4 = 0 and j2 − j3 ≠ 0 ;

• Critical arguments: j1 + j2 + j3 = j1 − j4 = j2 − j3 = 0 .

In completely analytical theories the orders of the divisors with respect to m (0 for monthly terms, 1 for annual terms, 2
for long period terms, and 3 for critical terms) must be taken into account in the integration since one wants to obtain a
given order with respect to small parameters in the coefficients of the final series. In semi-analytical theories,
integration increases considerably the amplitude of terms associated with small divisors. That forces to keep very
lengthy series in the right hand members of the equations before integration, to ensure a good precision for the
coefficients associated with small divisors in the final series. Nevertheless, the quantity to be considered in semi-
analytical theories is the actual value of the divisor instead of its order with respect to m, and arguments denoted above
as "critical" are critical only for completely analytical solutions since their periods amount to a few ten years similarly
to common long period arguments of the main problem. The actual critical argument for semi-analytical theories is
Laplace argument: 3D – 2F – l + 3l', with a period about 2 centuries. The effects of small divisors appear also in latitude
and distance, but not directly.

2.3 The solutions of the main problem

It is worth noticing that Brown (1899) considers Euler's works as the foundations of the most recent theories of the
lunar main problem at his epoch. He links Hansen's theory to Euler's first theory (Theoria Motus Lunae, 1753) by the
formulation adopted, Delaunay's theory to the Additamentum of Theoria Motus Lunae, in which Euler introduces
variations of constants, and Hill's theory to Euler's second theory (Theoria Motuum Lunae, 1772) because of the
rotating axes.
In Hansen's lunar theory, as it appears through the introduction of his tables (1857), longitude, latitude, and
parallax are expressed by means of three quantities nδ z , w , s . nz defined by :
nz = l + nδ z

is a disturbed mean anomaly, l (g in Hansen's notations) being the mean mean anomaly. nz allows to compute a kind of
true anomaly f by means of the elliptic expansion:

1 5 5
f = nz + ( 2 e − e3 + e ) sin nz +L
4 96
where e is an eccentricity constant which is different from Delaunay constant. w stands for perturbations to the
logarithm of a radius vector derived from f by means of the Keplerian expression. Then lunar parallax p is given by :

aE 1− e 2
log sin p = log − log −w
a 1 + e cos f

6
where a E (D in Hansen's notations) is the terrestrial equatorial radius and a a constant lunar semi-major axis. s stands for
perturbations to a latitude derived from f by means of the Keplerian expression, in such a way that lunar latitude U (B in
Hansen's notations) is given by :
sin U = sin I sin( f + ω) + s

where I is a inclination constant and ωthe mean argument of perigee. Finally lunar longitude V (L in Hansen's
notations) is given by :
V = f + ϖ + R + R'

where ϖ (Π in Hansen's notations) is the mean longitude of perigee, R is the reduction to ecliptic derived from f by
means of the Keplerian expression :
I
sin 2( f + ω)
tan2
tan R = − 2
I
1 + tan2 cos 2 ( f + ω)
2
and R' perturbations to R given by :
s tan I cos( f + ω)
R' = − + ∆R'
1 − sin 2 I sin 2 ( f + ω)

∆R' being a supplementary expression of a few periodic terms which do not concern the main problem.
The arguments of the main problem terms in nδ z , w , s are linear combinations of literal arguments l, l', ω,
and ω', l' (g' in Hansen's notations) being the solar mean anomaly and ω' the mean angle from the lunar ascending node
to the solar perigee. Coefficients are numerical. Considering the precision of the solution, it is sufficient to introduce
fitted corrections to metric constants through e and I and through the mean mean longitude. Hansen's variables will no
longer be used in the subsequent theories.
Brown's theory, for the main problem, is based on Hill's concept of the variational orbit. Hill (1877) considers
a simplified main problem in which α is neglected in the development of the disturbing function, the motion is planar
(γ = 0), and the Sun's orbit around the Earth-Moon barycenter is circular (e' = 0). Using rectangular coordinates in a
frame rotating with angular motion n' (instead of n used in Euler's second theory), the system of differential equations is
drastically simplified. All the terms of the simplified main problem depend on the two parameters m and e, only. Hill
shows that such a system has periodic solutions called 'variational orbits' or 'variation orbits', corresponding to e = 0. In
the variation orbits, coordinates can be developed in Fourier series of the unique argument D, and the coefficients
depend only on m which Hill replaces by m = m /(1 − m) to improve the convergence. The name 'variation orbit' comes
from the fact that it gives the leading part of the term with argument 2D named 'variation'.
The complete problem is then regarded as a perturbed variation orbit. Brown (1904c) mentions that the
original idea of the variation orbit belongs to Euler in his memoir : 'Réflexions sur la variation de la Lune' (1768).
Variation orbit provides Brown with an orbit of order 0 with respect to constants e, e', k, α. The value of m
being accurately known, the coefficients of the variation orbit can be computed under a numerical form, which avoids
poorly convergent developments. This orbit is used to obtain the actual orbit of the main problem in which constants e,
e', k, αare kept under a literal form. The computations are performed order by order, partially up to the sixth order.
Brown's main problem was initially formulated in rectangular coordinates, and then transformed to polar
coordinates: longitude, latitude, and parallax, while substituting numerical values for the constants involved in the
coefficients. The aim was to obtain all the coefficients greater than 0".01 in longitude and latitude, and 0".001 in
parallax, but Brown estimated that a large number of smaller terms had also been obtained. The actual construction of
the series begun in 1896 and ended in 1904. Brown was helped by one computer, Mr. Sterner, who spent about three
thousand hours, while Brown himself estimated his time to about five or six thousand hours (Brown, 1904c). The
development of the solution is published in volumes 53, 54, and 57 of the Memoirs of the Royal Astronomical Society
(1899, 1904a, 1908a), with references to several papers in the Monthly Notices of the Royal Astronomical Society.
Brown (1905a) has compared his solution to Hansen's, using the transformation of the latter into longitude, latitude, and
parallax by Newcomb, and corrections which take into account the differences between the constants used. Table 1
gives the differences larger than 0".10 in longitude and latitude.

7
Table 1. Differences Brown - Hansen for main problem longitude and latitude. Unit: arcsecond

Longitude Latitude

-0.32 sinD 0.15sin(F-D)

0.23sin(l-l'-D) -0.15sin(F+D)
-0.21sin(2l -2F) 0.13sin(l+F)
0.16sin2D -0.13sin(-l+F)

A similar method to Brown's has been used later by Gutzwiller and Schmidt (1986), with previous works by
Eckert and Bellesheim (Gutzwiller, 1979), and by Schmidt (1980). Trajectories of the main problem near the variational
orbit are constructed partially up to the degree 11 in the sum of the exponents of e, e', k, and b = αm − 2 / 3 , the criterion
for disregarding a monomial being based on its numerical value. The coefficients of the variational orbit are computed
numerically.
Brown, in the last years of his life (he died in 1938), suggested to Eckert to undertake the construction of a new
theory of the main problem by the method of Airy. It consists in computing numerical corrections to the coefficients of
a departure solution by solving variations equations. The resulting solution has completely numerical coefficients and
does not depend on the departure solution, provided the latter is precise enough. The departure solution was Brown's
main problem. The project was interrupted by the war and started again in the 1950s with the help of Smith. Eckert –
Smith solution (1966) is the first lunar theory (main problem only) of the electronic computer era. The coordinates used
are rectangular coordinates in a frame rotating with the lunar mean motion around the polar axis of the ecliptic, as in
Euler's second theory. In the corrections to Brown's rectangular coordinates obtained, 1 is larger than 0".01, 8 are as
large as 0".005, and 51 as large as 0".002. From the largest correction results a correction of 0".07 to the coefficient of
argument 2F – 2l in Brown's longitude.
In all the above described theories, the solution consists in a set of three coordinates, solutions of three
differential equations of the second order. Since in a rough approximation the Moon describes a rotating ellipse, a
natural choice of variables leads to the osculating elements : a, e, i, l, ω, and Ω, which are solutions of six equations
of the first order (Lagrange's equations for the osculating elements of a celestial body). After completion, the solution
has to be transformed to longitude, latitude, and distance or parallax. The following theories use sets of six variables
derived from osculating elements in order to avoid singularities or to take advantage of the Hamiltonian formulation.
Delaunay (1860, 1867) used canonical variables L = (µa )1/2 , G =L(1-e2 )1/2 , H = G cosi , l , ω ,and Ω (g and h
in Delaunay's notations), µ being the geocentric constant of gravitation. He built a completely analytical solution valid
to the seventh order with respect to the whole set of constants m, e, e', γ, α, some coefficients being calculated at an
upper order. In Delaunay's method, the disturbing function R is expressed in cosines of linear combinations of l, ω, Ω
and l', the coefficient of each trigonometric term being expanded explicitly in powers of e, e', γ, α. Trigonometric terms
of R are eliminated one after another by special kinds of canonical transformations. Delaunay worked alone, verifying
his computations by himself, and he spent about twenty years on the main problem.
At the time of electronic computers, Deprit, Henrard, and Rom (1971a, 1971b) have built, with a different
algorithm – the Lie transforms - a completely analytical solution up to the degree 20 in the sum of the exponents of e,
e', γ, α, m. Their solution is named ALE (Analytical Lunar Ephemeris). They use canonical variables L, P = L − G ,
Q = G − H , λ, p = −ω− Ω , and q = −Ω . In the disturbing function R expressed in those variables, a cascade of
canonical transformations are made, in order to eliminate successively the monthly terms - or short periodic terms - ,
next, the annual terms and finally, the terms with long period (node and perigee).
A semi-analytical solution, named SALE (Semi Analytical Lunar Ephemeris), has been built by Henrard
(1979), considering as an intermediate orbit a completely analytical solution of the non planar Hill's problem obtained
by the method of Lie transforms. SALE involves developments with respect to analytical increments δe, δe', δγ, δα and
δm up to the order 5.
In the main problem of ELP (Éphéméride Lunaire Parisienne) (Chapront-Touzé, 1980), an iterative process is
used to solve the Lagrange's equations in the variables n , λ , h = e cosϖ , k = e sinϖ , p = γ cosΩ , and q = γ sinΩ, n
being the osculating mean motion tied to a by Kepler's third law. Nevertheless, it was found that a pure iterative process
was divergent in particular in case of small divisors and the procedure was modified with a technique inspired by a
method of identification, for a differential system linearized in the vicinity of an approximate solution. ELP main

8
problem is a semi-analytical solution, the developments with respect to analytical increments δe, δe', δγ, δα, δm, and
δ(M/E+M) (E and M are the Earth's and Moon's masses) being provided at the order 1. Developments up to the order 2
have also been obtained, but never published.

Table 2. Comparisons between several solutions of the main problem longitude and latitude. Unit: arcsecond

Coordinate ALE - ILE SALE - ALE ELP - SALE

Longitude -0.0144 sin(D+l'-l) 0.0066 sin(D+l'-l) 0.00086 sin (D+l')

0.0081 sin(2D-4l'-l) 0.0033 sin(D+l') 0.00074 sin(D+l'-2l )
0.0070 sin(l'+l ) 0.0016 sin(D+l'+l) -0.00035 sin(D+l'-l)
0.0061 sin(2D-2F) -0.0015 sin(D-l') + Terms ≤ 0.00020
-0.0055 sin(2D-l'-l) + Terms ≤ 0.0010
+ Terms ≤ 0.0050

Latitude 0.0047 sin(l'+F) 0.0015 sin(D+l'-l+F) -0.00019 sin(D+l'-l-F)

0.0039 sin(l'-F) 0.0012 sin(D+l'-l-F) -0.00017 sin(D+l'-l+F)
-0.0035 sin(2D+l'-2l-F) + Terms ≤ 0.0010 + Terms ≤ 0.00005
+ Terms ≤ 0.0030

Several comparisons term by term have been performed between the new solutions of the main problem and
with Brown's solution. The differences between ELP and the solution of Gutzwiller and Schmidt are at the level of 1.
10-5 arcsecond for longitude and latitude (Gutzwiller and Schmidt, 1986). Table 2 gives the leading differences between
ALE and the series of ILE j=2 obtained from Brown's series by Eckert (see Sect. 5.1), the differences between SALE
and ALE, and the differences between SALE and ELP from (Henrard, 1973 and 1979 ; Chapront-Touzé and Henrard,
1980), for longitude and latitude. The differences have been computed with a consistent set of constants, not necessarily
the same in the three comparisons.
A global comparison of ELP and SALE main problems with numerical integrations has also been performed
by Kinoshita (1982). For a time span of one year, the root mean squares of the residuals between numerical integration
and theory are, for SALE and ELP respectively: 0".00037 and 0".00009 in longitude, 0".00012 and 0".00008 in latitude,
104 cm and 10 cm in distance. The results concerning ELP are much decreased (0".00001 in longitude and latitude, 1.2
cm in distance) by using series truncated to 1. 10-6 arcsecond (0.2 cm for distance) instead of the current ones truncated
to 1. 10-5 arcsecond (2 cm for distance).

Table 3. Various kinds of contributions in lunar orbital motion referred either to the mean ecliptic of date or to the
mean ecliptic of J2000.0: longitude (periodic and Poisson terms). Units: arsecond for longitude, century for time.
Theory: ELP 2000-82 (constants of the theory)

Main origin Longitude Comment

Keplerian motion + 22639.55000 sin l

+ 769.02326 sin 2l
Main problem + 4586.43061 sin (2D-l) Short period term: 'Evection'
+ 2369.91227 sin 2D Short period term: 'Variation'
- 666.44186 sin l' Annual term: 'Annual equation'
- 1.37259 sin(2l-2F) Long period term (3.0 yr)
+ 1.07773 sin (D+l'-l) Long period term (8.9 yr)

9
Planetary + 14.24883 sin (18V-16T-l+26°.54) Lunar quasi resonant term (273 yr)
perturbations - 1.14307 sin (2T-2J+2D-l+0°.12) Short period term
+ 0.90114 sin (4T-8M+3J-74°.01) Planetary quasi resonant term (1783 yr)
+ 1.67680 t sin l' Poisson term (solar eccentricity)
- 0.25425 t sin(18V-16T-l-65°.43) Poisson term, lunar quasi resonant arg.
2
+ 0.00487 t sin l' Poisson term (solar eccentricity)
Ecliptic plane motion + 0.28938 sin(D-F+T+95°.14) Long period term (18.6 yr)
- 0.00247 t sin(D-F+T+14°.81) Poisson term, long period argument
Ecliptic J2000.0 only + 2.11222 t sin(D-F+T +5°.13) Poisson term, long period argument
- 2.08930 t sin(D+F+T +5°.13) Poisson term
2
- 0.00901 t sin(D-F+T -75°.26) Poisson term, long period argument
+ 0.00892 t2 sin(D+F+T -75°.26) Poisson term
Earth's figure + 7.06304 sin (ζ-F) Long period term (18.6 yr)
- 0.00300 t sin (ζ-F) Poisson term, long period argument
Lunar figure + 0.00223 sin l' Annual term
Tidal perturbations(1) - 0.00082 sin (D+l'-F+12°94) Long period term (18.6 yr)
+ 0.00058 t sin l Poisson term
Tidal perturbations(2) - 0.00023 sin (D+l'-F+15°62) Long period term (18.6 yr)
+ 0.00044 t sin l Poisson term
General Relativity - 0.00081 sin l' Annual term
(1) Williams' model, constants from DE200 – (2) Chapront's model, constants from DE245

3. LUNAR THEORY: THE PERTURBATIONS

3.1 General presentation
In the 18th century, lunar theory was restricted to the main problem only. Perturbations to the main problem were
introduced for the first time in Laplace's theory (1802), and were developed at the same time as the precision of lunar
theory increased. They now involve :
• Direct planetary perturbations due to the direct action of the planets (except the Earth) on the Moon ;
• Indirect planetary perturbations due to the perturbations in the apparent motion of the Sun around the Earth-Moon
barycenter, superimposed to the Keplerian motion adopted in the main Problem ;
• Perturbations due to the Earth's figure ;
• Perturbations due to the Moon's figure ;
• Tidal effects due to the deformations of the Earth and Moon ;
• Relativistic effects.
Furthermore, if the lunar motion is referred to the mean ecliptic of date, instead of the mean ecliptic of a given
epoch, corrections due to the moving reference plane must be taken into account. The former option is usually adopted
because it avoids large Poisson terms in the solution, the mean ecliptic of date appearing as the natural reference plane
of the lunar motion. Brown introduces the 'departure point' of a fixed epoch as the direction of the Ex axis in the mean
ecliptic of date. It is defined by (EN,Ex) = (EN,Eγ), N being the ascending node of the mean ecliptic of date on the
mean ecliptic of the epoch and γ the mean equinox of the epoch. (EN,Ex) is measured in the mean ecliptic of date and
(EN,Eγ) in the fixed ecliptic. This feature avoids to introduce the equator motion in the solution. The lunar mean motion
referred to the mean ecliptic of date and departure point of a fixed epoch is the sidereal mean motion of the epoch, and

10
the transformation from the lunar longitude and latitude in this frame to coordinates referred to the mean ecliptic and
mean equinox of date is limited to the addition of general precession in longitude.
Besides the main problem series, perturbations induce in the expressions of the lunar coordinates, periodic and
Poisson terms. They also change the fundamental arguments of the main problem by contributing to the mean motions
of perigee and node, and by introducing quadratic, cubic … terms with respect to time in the mean mean longitude and
in the mean longitudes of perigee and node. Tables 3 and 4 give the leading periodic and Poisson terms for the main
problem and the various perturbations, from (Chapront-Touzé and Chapront, 1983 and 1988). Tables 5 and 6 give the
contributions to the mean motions of perigee and node, and to the quadratic terms of the mean elements from the second
paper and, for comparison, from Brown's theory.
The integration constants of the whole problem are defined in the same way as those of the main problem (see Sect.
2.2). Note that, in perturbations, series have usually numerical coefficients except in the case of the Earth's figure
perturbations obtained by Hill and later by Deprit, Henrard, and Rom.
Details about the various kinds of perturbations are given in the following sections.

3.1.1 Planetary perturbations

They include direct and indirect planetary perturbations as mentioned above. In analytical theories, only the effects of
the eight planets, from Mercury to Neptune, are taken into account, in contrast to recent numerical integrations which
involve also the effects of Pluto and several minor planets. In the formers, computation requires a planetary theory in
which heliocentric coordinates or osculating elements usually consist in expressions like:
K
s= ∑ ∑ Ai(1k,i)2 ,Ki 8 t k sin(i1λ1 + i2 λ2 + K + i8 λ8 + φi(1k,i )2 ,Ki 8 )
i1 , i2 ,Ki8 k = 0

where λi are the mean mean longitudes of the eight planets referred to a fixed ecliptic and equinox, φi( k, i) ,Ki is a
1 2 8

numerical phase and Ai( k,i ) ,Ki a numerical coefficient. Constant and secular terms are included in that formulation
1 2 8

( i1 = i2 = K = i8 = 0, φi(1 k,i )2 ,Ki8 = 90° ). The upper limit K of the powers of time in Poisson terms depends on planetary
theories. Those Poisson terms are the developments of periodic terms with very long periods amounting to several ten
thousand years. Note that in planetary theories, 'planet' stands for the barycenter of the planet and its satellites ; hence
the Earth's heliocentric elements are actually the Earth-Moon barycenter's ones.
In the planetary perturbations of the Moon, the planetary arguments are combined with the lunar arguments. Such
combinations are divided in three categories (examples are given in Table 3):
• Lunar-type arguments – Short period arguments associated with an ordinary divisor ;
• Planetary quasi-resonant arguments – Small divisor arguments in the planetary motions, combinations of the
planetary longitudes alone ;
• Lunar quasi-resonant arguments – Small divisor combinations of lunar and planetary arguments.
For the quasi-resonant arguments appearing in longitude, the amplitudes of the corresponding terms in the disturbing
function are considerably increased in the integration process. Special techniques for the manipulation of the series have
to be settled. The separation method which consists in writing, before integration, the right-hand members of the
differential equations under the form of products of lunar terms and planetary terms is very often used (e.g. Brown,
1910 ; Standaert, 1980). At the epoch of hand computing it allowed to select the terms contributing to each quasi-
resonant argument. Now the multiplications are performed with a weighting factor which depends on the resulting
divisor, in order to keep only the terms that will contribute significantly to the solution. Very special combinations may
appear, with very long periods. Such terms are not retained because their amplitudes are poorly determined and their
contributions are absorbed in the integration constants (longitude at the epoch and mean motion). Practically the periods
retained are limited to 5000 years and less. In fact, as far as the validity of planetary series is itself limited to a few
thousand years because of Poisson expansions, it is quite natural to keep the same kind of limitation in the elaboration
of the lunar series.
Secular and Poisson terms in the developments of the planetary coordinates or osculating elements generate the
same kind of contributions in the lunar ones. In particular the secular terms in the eccentricity of the Earth-Moon
barycenter give rise to quadratic, cubic …. terms in the mean mean longitude and in the mean longitudes of perigee and
node (Table 6). They also give rise to Poisson terms in the coordinates.

11
 At this level, the choice of the reference plane adopted in the lunar theory - mean ecliptic of date or fixed ecliptic -
must be taken into account by a rotation of the reference plane of the planetary theory if necessary. As mentioned above
the choice of the mean ecliptic of date avoids large Poisson terms in the lunar coordinates. Then corrections due to the
moving ecliptic are computed separately, yielding periodic terms and small Poisson terms. When these periodic terms
are joined to those from the planetary perturbations, the resulting series are the same as in the case of planetary
perturbations referred to a fixed ecliptic. In the latter case, Poisson terms obtained are the sum of the small Poisson
terms resulting from the ecliptic plane motion and of the large ones resulting from the transformation of the longitude
and latitude referred to mean ecliptic of date, to the corresponding coordinates referred to the ecliptic of epoch (see
'Ecliptic plane motion' and 'Ecliptic J2000.0 only' in Tables 3 and 4).
A coupling effect exists between the lunar and the Earth-Moon barycenter motions. The coordinates of the
Earth-Moon barycenter contain terms whose arguments are linear combinations of D, F, l and l', due to the direct
perturbations of the Moon on the Earth. Conversely, in the lunar indirect planetary perturbations, such terms combined
with lunar arguments give rise to periodic terms and to corrections the mean motions of node and perigee.
Finally, the contributions due to the planets in lunar coordinates have the form :
K
σ = ∑ ∑ A(j k,i )t k sin( j1 D + j2 F + j3 l + j4 l' + i1Q + i2V + i3T + i4 M + i5 J + i6 S + i7U + i8 N + φ(j ,ki ) ) (4)
j ,i k =0

where j,i stands for j1 , j2 , j3 , j4 , i1 , i2 , i3 , i4 , i5 , i6 , i7 , i8 . We have used Brown's notations for the planetary mean
longitudes: Q, Mercury ; V, Venus ; T, Earth ; M, Mars ; J, Jupiter ; S, Saturn ; U, Uranus ; and N, Neptune. K is the
upper limit of the powers of time in the Poisson terms retained in the lunar theory ; it may be smaller than that of the
planetary theory used.
From Hansen's theory to ELP, planetary perturbations are the main cause which limits the precision of analytical
solutions because of the large number of small divisors and of the numerous contributions.

3.1.2 Perturbations due to the Earth's figure and to the Moon's figure
Mass distribution in the Earth and in the Moon does not proceed rigorously from a spherical symmetry as it is supposed
in the main problem hypotheses, and that gives rise to perturbations. Early works on this subject took into account
estimates of the principal moments of inertia, and conversely, the perturbations of the Moon's motion by the Earth's
figure were a mean to determine the Earth's ellipticity. Nowadays, space research provides determinations of the
coefficients of the developments of Earth's and Moon's potentials in spherical harmonics, which allow a better
knowledge of that kind of perturbations. Their computations depend also on the motions of the bodies around their
centers of mass.
Table 4. Various kinds of contributions in lunar orbital motion referred either to the mean ecliptic of date or to the
mean ecliptic of J2000.0: latitude and distance. Units: arsecond for latitude, km for distance, century for time. Theory:
ELP 2000-82 (constants of the theory)

Main origin Latitude Distance

Keplerian motion + 18461.40000 sin F +385000.52719

+ 1010.17430 sin (l+F) - 20905.32206 cos l
+ 999.70079 sin (l-F)
Main problem + 623.65783 sin (2D-F) - 3699.10468 cos (2D-l)
- 2955.96651 cos 2D
Planetary + 0.63037 sin (18V-16T-l+F+26°.54) + 1.05870 cos(2T-2J+2D-l+0°.12)
perturbations + 0.63014 sin (18V-16T-l-F+26°.54)
- 0.07430 t sin (2D-l'-F) + 0.51395 t cos(2D-l')
- 0.01126 t sin (18V-16T-l-F-65°.49) - 0.01302 t cos(18V-16T-2l-65°.45)
- 0.01123 t sin (18V-16T-l+F-65°.38) + 0.01225 t cos(18V-16T-65°.20)
2
- 0.00022 t sin (2D-l'-F) + 0.00149 t2 cos(2D-l')
Ecliptic plane + 1.37497 sin (T+D-84°.87) + 0.05765 cos(T+D+l-F-84°.87)

12
motion - 0.05703 cos(T+D-l -F-84°.87)
+ 0.01172 t sin(T+D+14°.81) + 0.00049 t cos(T+D+l-F+14°.81)
- 0.00049 t cos(T+D-l -F+14°.81)
Ecliptic J2000.0 + 46.84817 t sin(T+D+5°.13)
only - 0.19994 t sin(T+D-75°.26)
Earth figure - 8.04508 sin ζ - 0.45648 cos (ζ+l-F)
+ 0.45276 cos (ζ-l -F)
+ 0.00342 t sin ζ + 0.00019 t cos ( ζ+l-F)
- 0.00019 t cos ( ζ-l-F)
Lunar figure + 0.00010 sin (l'-F) + 0.00130
+ 0.00010 sin (l'+F)
Tidal - 0.00004 sin (D+l'-2F+12°94) - 0.00004 cos(D+l'-l-F+12°.94)
Perturbations(1) - 0.00004 sin (D+l'+12°94) + 0.00004 cos(D+l'+l-F+12°.94)
- 0.00005 t sin F + 0.00356 t
Tidal - 0.00001 sin (D+l'-2F+17°01) + 0.00000
Perturbations(2) - 0.00002 sin (D+l'-9°41) + 0.00000
- 0.00010 t sin F + 0.00381 t
General Relativity - 0.00004 sin (l'-F) - 0.00828
- 0.00004 sin (l'+F)
(1) Williams' model, constants from DE200 – (2) Chapront's model, constants from DE245

For the Earth, since a symmetry of revolution around the polar axis may be assumed, sidereal time is not
involved, but precession and nutation are. Precession makes argument ζ(the mean mean longitude of the Moon referred
to the mean equinox of date) appear in the series and yields Poisson terms. The general form of the perturbations in
lunar coordinates is:
K
σ= ∑ ∑ A(j1k, ), j 2 , j3 , j4 , j5 tk sin( j1 D + j2 F + j3l + j4 l '+ j5ζ + φj1, , j2 , j 3 , j 4 , j 5 ) . (5)
j1, , j 2 , j3 , j 4 , j5 k = 0

Sometimes ζ is replaced by the sum of the mean longitude of node referred to the same equinox and F.
For the Moon, a libration theory is necessary. Much attention must be paid to the symmetry between the
contributions involved in the libration theory and in the computation of the perturbations, in order to avoid spurious
terms (Chapront-Touzé, 1983). Up to now, only the main problem of the libration theory has been used and the general
form of the perturbations in the lunar coordinates is:
σ= ∑ A j1 ,, j2 , j3 , j4 sin( j1 D + j2 F + j3 l + j4 l' +φj1, , j 2 , j 3 , j 4 ) . (6)
j1, , j 2 , j3 , j4

Table 5. Various kinds of contributions in the sidereal mean motions of perigee and node ("/cy). Brown's values are
referred to 1850 and computed with the constants of the memoirs. ELP 2000-85 values are referred to J2000.0 and
computed with the constants of the theory.

Contribution Perigee Node

Brown ELP Brown ELP

Main problem 14 642 624 14 642 537.9368 -6 967 185 -6 967 167.2643

13
Planets 253 247.1052 -137 -144.0733
Earth's figure 641 (3) 633.4034 -600 (3) -592.5357
657 (4) -615 (4)
Lunar figure 3 -1.7472 -14 -16.9443
Tidal perturbations 0.0670 (1) 0.0000 (1)
0.1377 (2) -0.0030 (2)
General Relativity 1.7971 1.9019
(1) Williams' model, constants from DE200 – (2) Chapront's model, constants from DE245
(3) Hansen's value for the Earth's ellipticity - (4) Faye's value for the Earth's ellipticity

Table 6. Various kinds of contributions in the quadratic terms of the lunar mean elements ("/cy2 ). Brown's values are
referred to 1850, and ELP 2000-85 values are referred to J2000.0

Contribution Longitude Perigee Node

Brown ELP Brown ELP Brown ELP

Planets 5.82 5.8665 -38.32 -38.5540 6.46 6.5044
Earth's figure 0.20 0.1925 0.11 0.1003 -0.10 -0.0958
Tidal effects -11.9473 (1) 0.1761 (1) -0.0464 (1)
-12.8125 (2) 0.1898 (2) -0.0493 (2)
(1) Williams' model, constants from DE200 – (2) Chapront's model, constants from DE245

3.1.3 Tidal perturbations

In the perturbations mentioned in Sect. 3.1.2 , the Earth and the Moon were supposed to be rigid bodies. Tidal
perturbations are complementary perturbations due to their deformations under the effect of the attractions by the Moon,
the Sun and the Earth.
Attraction of disturbing bodies on an anelastic body produces time dependent corrections to the constant
harmonic coefficients of its potential when considered as a rigid body (Lambeck, 1980). These corrections depend on
constant factors named Love numbers, and time is involved by means of coordinates of the disturbing bodies evaluated
with a time delay. For the Earth, the disturbing bodies to be considered are the Moon and the Sun ; for the Moon, they
are the Earth and the Sun. Up to now, only corrections to second degree harmonic coefficients have been taken into
account.
Rotation of the anelastic body produces similar effects, but for the Earth, the rotation is regular enough to only
yield constant corrections which are supposed to be involved in the fitted values of the constant harmonic coefficients
of the rigid body. For the Moon, the corrections are shared between constant parts which are processed in a way similar
to the Earth, and periodic parts which are added to tidal corrections.
The leading effect is the tidal secular acceleration of the lunar longitude (see Sect. 4.4). Periodic terms and
Poisson terms are also produced with a formulation similar to (6).

3.1.4 Relativistic perturbations

Writing the differential equations of the lunar motion using a post-Newtonian N-body metric, the relativistic effects are
regarded as corrections or 'perturbations' to the lunar trajectory in the Newtonian framework. As for planetary
perturbations, two effects are distinguished: the direct effect produced by the non-Newtonian part of the acceleration,
and the indirect effect produced by the relativistic terms of the coordinates of the Earth and other planets.
Relativistic perturbations depend on the relativistic reference system to which the theory is referred. A natural
reference system for the lunar motion should be a (relativistic) geocentric reference system, but since relativistic terms

14
of the planetary coordinates are involved in the lunar solution, it is preferable to choose a (relativistic) barycentric
reference system (centered at the solar system barycenter) which is convenient for all the bodies of the solar system. In
this system, the rectangular coordinates of the Moon provided by lunar theory appear as differences between
(relativistic) barycentric coordinates of the Moon's mass center and (relativistic) barycentric coordinates of the Earth's
mass center. Note that the relativistic perturbations of the Moon are larger when computed in a barycentric reference
system than in a geocentric one (Soffel, 1989).

3.2 The solutions

3.2.1 Planetary perturbations
Hansen's theory involved only a rough solution for planetary perturbations. Nine short period terms were computed in
nδz, and one quasi-resonant term, with argument 18V −16 T − l . An other quasi-resonant argument was put into
evidence: 8V − 13T , but the coefficient was not computed. The effects of the linear secular term of the Earth's
eccentricity, e'1 t, were taken into account, on one hand through the quadratic terms of the lunar longitude and of the
mean longitudes of perigee and nodes, on the other hand through 19 Poisson terms in n δz. These Poisson terms can be
j4
obtained very easily by multiplying the coefficients of the main problem with formulation (3) by ( e'1 / e ' ) .

Delaunay computed the two quasi-resonant terms considered by Hansen and the quadratic contributions to the
lunar longitude and to the mean longitudes of perigee and node, but he died before performing a full solution. This work
was undertaken by Radau (1893) who obtained the best solution at his epoch. Radau separated lunar terms from
planetary ones in the disturbing function for selecting the useful contributions (see Sect 3.1.1), and each resulting
argument was eliminated following Delaunay's method with Hill's improvements (Hill, 1891).
Brown's planetary perturbations are ready in April 1908 and published in volume 59 of the Memoirs of the
Royal Astronomical Society (Brown, 1910). Direct planetary perturbations have also been published in Adams prize
Essay (1908b). The computations have been performed almost solely by Brown himself. The method used is inspired
from Jacobi's method. It is based on the variation of mean arguments: sidereal mean mean longitude λ, sidereal mean
longitude of perigee ϖ, and sidereal mean longitude of node Ω (respectively w1 , w2 , w3 in Brown's notations), and of
constants c1 , c2 , c3 which form with λ, ϖ, Ω a canonical system. c1 , c2 , c3 are tied to the usual integration constants:
sidereal mean motion n, eccentricity constant e , inclination constant γ . For periodic terms, the results are converted to
longitude, latitude, and parallax, except for some long period terms which are kept in constantγ , and arguments λ, ϖ, Ω
(or D, F, l) of the main problem series, in the same way as secular terms (Tables 5 and 6). A special process is applied
to the indirect perturbations due to the secular terms in the Earth's eccentricity and longitude of perihelia, and to long
period terms in the Earth's longitude (Brown, 1905b). These terms are substituted in the main problem series through
the 'constant' e' , and arguments l ' = T − ϖ' and D = λ − T + 180 ° . In addition, the process gives rise to quadratic and
cubic terms in λ, ϖ, Ω, and to small supplementary short period terms converted to polar coordinates. The planetary
theory used was Newcomb's, with corrections derived from Le Verrier's theory.
Nowadays, direct and indirect planetary perturbations have been computed by Chapront-Touzé and Chapront,
(1980, 1982, 1983), in the frame of the ELP solution. The method used is Brown's method including the special process
for the indirect perturbations due to the secular terms in the Earth's eccentricity and longitude of perihelia (Chapront-
Touzé, 1982), but not the ones due to the long period terms of the Earth's longitude. In solving the differential
equations, only one iteration has been performed, so the second order with respect to planetary masses is not complete
in the lunar coordinates though second order terms in the planetary coordinates have been taken into account. The series
in longitude and latitude are truncated to 10-5 arcsecond (2 cm for the distance series). Though all the terms greater than
this limit are probably not present, the total number of terms is about 19000 terms in V, 6500 in U and 8500 in r.
Furthermore, the number of terms in the series grows considerably when the truncation limit decreases. That was a
difficulty in the construction of the series since it was not always possible to give the truncation level of intermediate
series a value small enough to insure the precision of a few 10-5 arcsecond for each coefficient of the final series. These
defects joined to the accumulation of round off errors in the summation of such a large number of terms, make the
global precision of the series about 10-2 arcsecond. This estimate arises from the comparison of ELP to JPL numerical
integration DE200 (Chapront-Touzé and Chapront, 1983). The upper limit K of the powers of time in Poisson terms is 2
; the terms with power 2 come solely from the substitution of the quadratic term of the Earth's eccentricity in the main
problem series. The upper limit K of the powers of time in the mean elements λ, ϖ, and Ω, which was originally 2
(Chapront-Touzé and Chapront, 1983), has been increased to 4 in (Chapront-Touzé and Chapront, 1988). The planetary
theory used is VSOP82 (Bretagnon, 1982) except for secular terms which are taken from (Laskar, 1986) in the second
paper. The difference between the reference planes of VSOP82 and ELP (ELP is referred to the mean ecliptic of date
while VSOP82 is referred to the mean ecliptic of J2000.0) is taken into account by disregarding the secular terms in the
variables p and q for the Earth. The values of the planetary masses involved in the coefficients of series are those of the

15
UAI 1976 system and the planetary orbital elements are the VSOP82 values fitted to DE200. The values of the lunar
orbital elements involved in the coefficients of series are the nominal values of the main problem series.
Direct planetary perturbations have also been computed by Standaert (1980 and 1983) in the frame of the
SALE solution, and indirect planetary perturbations have been computed to a lower precision by Kubo (1982).
Standaert uses Lie transforms, the main problem solution being that of ELP with the derivatives from SALE.
For the coordinates of planets, he takes into account the elliptic motion only but he solves the differential equations of
the lunar perturbations to the second order with respect to planetary masses.
Kubo uses Hori's Hamiltonian method with Delaunay's variables, the main problem solution being Delaunay's
and the planetary solution being Newcomb's.
The comparisons of these new solutions to Brown's planetary perturbations, made in the papers mentioned
above, have put into evidence several defects in Brown's solution. The leading ones are:
• An error of 180° in the phases of the arguments with odd multiples of D in the direct perturbations. The
amplitudes of the corresponding terms can amount to 0".04 in longitude ;
• Missing indirect perturbations, related to the motion of the instantaneous node of the Earth's orbit. The
coefficients of the corresponding terms can amount to 0".075 in latitude ;
• A missing term whose the argument contains the longitude of Mercury and whose coefficient is 0".066 in
indirect perturbations.
Thanks to progresses in the domain of electronic computers, a new computation of the planetary perturbations
has been undertaken at Paris Observatory by Bidart (Bidart and Chapront, 1999). The method used is still Brown's
method but the secular terms in the Earth's eccentricity and longitude of perihelia are processed similarly to other terms.
The upper limit K of the powers of time in Poisson terms is 5. The planetary coordinates are derived from a new theory
undertaken by Moisson (1999). The values of planetary masses, planetary elements and lunar elements involved in the
coefficients of series are improved ones. The results are expected to appear before the end of the 20th century.

3.2.2 Perturbations due to the Earth's figure and to the Moon's figure
Hansen's theory included four periodic terms due to the Earth's figure in n δz, as well as one periodic term and one
Poisson term in s, but the first precise solution for these perturbations is Hill's completely analytical solution (1891).
Hill expresses the disturbing function with respect to the principal moments of inertia A, B, C (the latter referred to the
polar axis), and next he considers A and B as equal, which corresponds to a symmetry of revolution. In the integration
of the differential equations, Hill uses Delaunay's method with a few improvements.
Brown's solutions for the perturbations due to the Earth's and Moon's figures are published in the same volume
of the Memoirs of the Royal Astronomical Society as his planetary perturbations, but the contribution of the Earth's
figure to the mean motions of perigee and node has yet been introduced in a previous paper (Brown, 1904b). The
method used is the same as for planetary perturbations.
For the Earth, A and B are supposed to be equal and C - A is expressed in function of the Earth's ellipticity. The
value adopted in the memoirs (Brown, 1909a) is Faye's value (1/292.9) but Hansen's value (1/296.3) is also considered
in the previous paper. Similar to long period terms in planetary perturbations, the terms with the node period are kept in
the arguments and in γ. Later, Brown (1909b) computed quadratic terms in the lunar longitude and mean longitudes of
perigee and node, due to the motion of the Earth's equatorial plane (Table 6).

For the effects due to the lunar figure, Brown admits that ratio C /(MR2 ) is the same for the Earth and the
Moon, M being the mass of the body, and R its equatorial radius ; he obtains the perturbations to the mean motions of
perigee and node as :
A+ B B−A
δb2 = 191" (1 − ) − 503 "
2C C
A
δb3 = − 230 " (1 − ).
C
Brown uses Hayn's values :
B − A = 0 . 000 157 C
C − A = 0. 000 629 C

16
The results are given in Table 5. The solution involved for the motion of the Moon around its center of gravity is
restricted to Cassini's laws.
Nowadays, perturbations due to the Earth's figure have been computed under a completely analytical form by
Deprit, Henrard and Rom in the frame of ALE (Henrard, 1971), by using Lie transforms. Similar perturbations, but with
numerical coefficients, have been obtained by Henrard (1981) in the frame of SALE, by Kubo (1982), and by Chapront-
Touzé (1982) in the frame of ELP. Henrard uses Lie transforms, the main problem solution being that of SALE. Kubo
and Chapront-Touzé use the methods described above for planetary perturbations. In addition to perturbations due to the
zonal harmonic coefficient J2 of the Earth's potential, SALE and ELP take into account coefficient J3 and SALE
coefficient J4 , but the effects of J4 are not larger than 10-5 arcsecond.
Henrard has compared his solution to Hill's, to Brown's and to ALE, term by term. The largest differences that he
has put into evidence amount to –0".015, 0".012, and 0".038 respectively, and concern the coefficient of the leading
term in longitude (see Table 3). The large discrepancy with ALE has been attributed to the poor convergence of
analytical expansions in ALE. The agreement with Kubo's coefficients is at the level of 0".004 (Kubo, 1982). With ELP
coefficients, the largest differences are of 0".0004, due to second order terms with respect to J2 missing in ELP, except
for corrections due to the Earth's nutation, missing in SALE, which amount to 0".002 on the leading term of longitude.
Perturbations due to the lunar figure has been computed with similar methods by Henrard (1980) in the frame of
SALE, Chapront-Touzé (1983) in the frame of ELP, and Kubo (1983). In ELP, the harmonics of the lunar potential up
to degree 4 are taken into account, and the solution for the libration involved is the main problem of Moons' solution
(1982). Henrard takes into account all the harmonics up to degree 2 and two harmonics of degree 3, but disregards
physical libration. Kubo takes into account harmonics up to degree 3 and computes a rough estimate of physical
libration effects. Chapront-Touzé's and Kubo's results are in good agreement but unexplained discrepancies with
Henrard's results exist, concerning periodic terms. Note that the quadratic term introduced in the mean longitude by
harmonics of degree 4 in (Chapront-Touzé, 1983) was due to a program error discovered later on.

3.2.3 Tidal perturbations

Series of the tidal perturbations has been computed by the authors with the model proposed by Williams, Sinclair, and
Yoder (1978) for the Earth's tidal potential. That model supposes Love number and time delay independent of the
harmonic (except for zonal harmonics in which time delay is zero). The method used for integration is the same as in
other perturbations and the values adopted for Love number and time delay are those of the JPL numerical integration
DE200.
More recently, the authors have computed new series with a more refined model of tidal potential in which
Love numbers and time delays depend on the harmonics. This model is probably similar to the one used in the recent
JPL numerical integrations, e.g. DE245 and DE403. Perturbations due to the Earth's and Moon's tidal potentials are
considered with two sets of constants: the first one from DE245 and the second one from DE403. The values of the
coefficients strongly depend on the model (Tables 3 and 4); the differences between DE245 and DE403 constants have
smaller effects.

3.2.4 Relativistic perturbations

Direct and indirect relativistic perturbations of the lunar motion have been computed by Brumberg and Ivanova (1981)
in the PPN formalism, with two systems of coordinates in a barycentric reference system.
In the frame of the ELP solution, the indirect relativistic perturbations have been computed by Lestrade,
Chapront and Chapront-Touzé (1982) using the relativistic terms of the Earth's motion from (Lestrade and Bretagnon,
1982). The direct perturbations have been computed and added to the indirect perturbations by Lestrade and Chapront-
Touzé (1982). This solution is based on the General Relativity theory and uses isotropic coordinates in a barycentric
reference system. The differences with Brumberg's and Ivanova's solution, limited to the General Relativity theory with
the same coordinates, are less than 0".00006 in longitude.

4. LUNAR TABLES AT THE BEGINNING OF THE 20TH CENTURY

4.1 Hansen's Tables
At the beginning of the 20th century lunar ephemerides were computed from Hansen's tables (1857), sometimes with the
corrections introduced by Newcomb (see Sect. 4.4).

17
 Hansen's lunar tables are derived from Hansen's lunar theory (see Sect. 2.3) and allow to compute longitude
and latitude referred to the true ecliptic and true equinox of date, and parallax. The expressions of nδ z , w , s , given in
the introduction of the tables, contain about 510 periodic terms and 20 Poisson terms with numerical coefficients.
Among them, about 280 periodic terms and 7 Poisson terms are actually used in the tables, but those numbers cannot be
compared easily with the corresponding numbers in longitude, latitude, and parallax series, because of the intricate
formulation of Hansen's theory.
Hansen fit to observations the six integration constants, the coefficient of the parallactic term (to fix the ratio
of the solar mean distance to the lunar one), the coefficients of leading terms due to the Earth's figure in longitude and
latitude (to avoid the uncertainty on the Earth's ellipticity value), the constant term of parallax, the mean motions of
lunar perigee and node, and the coefficient of the planetary term in longitude whose argument is 8V − 13T (period 239
years), which appears as an empirical term (Sect. 4.4). The coefficients of the other periodic terms and of the quadratic
terms of the lunar longitude and of the mean longitudes of perigee and node are derived from the theory.

4.2 Radau's tables

The first new lunar tables of the 20th century were Radau's tables (1911) based on Delaunay's theory. The construction
of tables from his analytical theory (1860 and 1867) had been started by Delaunay himself and he had received proofs
of some parts when he died in august 1872. In 1873, after a report by Loewy and Puiseux, the Bureau des longitudes
decided to continue Delaunay's work. In the 1880s, under the direction of Tisserand, many changes were made to
Delaunay's initial plans in order to render the tables easier to use. Finally, the tables were achieved by Radau, about 40
years after Delaunay's death.
For the main problem, Delaunay's theory (Sect. 2.3) consisted of completely analytical series for longitude,
latitude and parallax, but because of the poor convergence of the coefficients with respect to m, the ratio of the solar and
lunar mean motions, he completed them by small numerical contributions denoted as 'compléments probables', which
are introduced in the tables.
For planetary perturbations, the series are taken from (Radau, 1893) with improvements derived from
Newcomb's and Brown's works published in the meantime. The perturbations by the Earth's figure are taken from (Hill,
1891).
The constants appearing in the main problem coefficients have received values from Newcomb (e and the
parallax constant) and Airy (γ). For the mean longitudes of perigee and node, the values at the epoch 1850 and the mean
motions have been taken from Hansen's tables, and the quadratic terms from Brown's theory. Radau has only fit to
observations the complete expression of the lunar mean mean longitude and two empirical terms (see Sect. 4.4). The
observations, gathered by Newcomb and Airy, covered the time span [688 B.C. ; 1910].
Radau's tables were used in the Connaissance des temps from 1915 till 1925.

4.3 Brown's tables

Brown's tables (1919) were the most accurate source of lunar ephemerides before the introduction of electronic
computers. They are based on Brown's theory (Sect. 2.3). After the completion of the theory, they needed about nine
years of computation, from autumn 1909 till summer 1918, by one to four (human) computers. The chief computer,
H.B. Hedrick, is mentioned on the title page of the book and Brown acknowledges the help of three other computers:
Mr. Murray, Miss Gundersen and Mrs. Hedrick.
The tables have been built so as to facilitate the computation of lunar positions for a large number of
continuous dates (e.g. annual ephemerides), rather than the computation for a few separate dates, and with the care of
optimizing the number of individual tables and their entries. It is for these reasons that long period terms from
planetary perturbations are kept in 'constant' γ and arguments λ, ϖ, Ω, T, and the secular term is kept in e' (see Sect.
3.2.1). When computing lunar position, all these secular and long period terms must be taken into account in the main
j j j j
problem series through the main characteristic α 1 γ 2 e 3 e' 4 of each term of (3), and in the computation of
fundamental arguments D , F , l , l'. Seeking tables optimization, Brown (1911) set his latitude series for the main
problem to the form :
U = (k + C )[sin(F + S ) + µ sin 3(F + S ) + ν sin 5(F + S ) + N ]

where k, µ, ν are constants, S, N, C are series without constant terms, and C is independent of F . With that
transformation, Brown's series for longitude, latitude, and parallax contain about 1500 terms.

18
 For the constants involved in the main problem coefficients, Brown has used three sets of values (Table 7). The
first set, from Newcomb, has been used in the memoirs and in (Brown, 1905a). The second set, denoted here as
'Tables/1', has been used in the construction of Brown's tables for computing the individual tables which correspond to
periodic terms. Earth's eccentricity e' and constant term Π0 of the sine of parallax are still from Newcomb, but the
coefficients of the terms with arguments l and F, in longitude and latitude respectively, are now from Cowell. The last
set, denoted here as 'Tables/2', is introduced in the tables as small corrections by means of the principal characteristics,
similar to secular and long period terms. Except for e', it has been fit by Brown to Greenwich meridian observations, as
well as the constant terms of the mean longitudes of perigee and node (Brown, 1915). In these fits, Brown was helped
by several comparisons of the same observations to Hansen's tables, made previously by Cowell, from which the values
of E and Γ of the second set are derived. The angular mean elements of the Earth's orbit are taken from Newcomb. The
mean motions and coefficients of quadratic and cubic terms of the lunar perigee and node are the computed values
(Tables 5 and 6, with small corrections due to the new values of constants). Nevertheless the value 1/294.0 of the
Earth's ellipticity is adopted in order to get a better agreement of the computed values of the mean motions with the
observed ones. This value of the Earth's ellipticity stands between Faye's value (1/292.9) and Hansen's value (1/296.3).
For comparison, the present value is 1/298.26.

Table 7. Constants adopted in the main problem coefficients (see definitions in Sect. 2.2)

Solution 2E (arcseconds) 2Γ (arcseconds) Π0 e' Epoch

(arcseconds)
Radau 22 639.580 18 461.260 3 422.520 0.016 771 06 1850
Brown (memoirs) 22 639.580 18 461.480 3 422.700 0.016 771 91 1850
Brown (Tables/1) 22 639.500 18 461.350 3 422.700 0.016 751 04 1900
Brown (Tables/2) 22 639.550 18 461.400 3 422.540 0.016 751 04 1900
& ILE j = 0
ILE j = 1 & j = 2 22 639.550 18 461.400 3 422.452 0.016 751 04 1900

n ("/cy)
1732 559 343 +
ELP 2000-82 & 96 22 639.550 00 18 461.400 00 0.180 00 0.016 709 2400 2000
ELP 2000 22 639.585 78 18 461.238 68 0.736 04 0.016 708 6156 2000
ELP 2000-96 (DE403) 22 639.586 06 18 461.241 30 0.356 24 0.016 708 6156 2000
ELP 2000-96 (LLR) 22 639.586 14 18 461.239 90 0.344 39 0.016 708 6159 2000

4.4 The secular acceleration of the longitude and the empirical terms
The expression of the tropic mean mean longitude resulting from the mean elements adopted in Hansen's tables was :
LHansen = 335°43'26".70 + 1 732 564 419".61 T0 + 13".301 T0 2 + 0".013 473 T0 3 ,
T0 being the time elapsed from 1800.0 (1800 January 0, Greenwich mean noon) in Julian centuries. The expression of
the general precession in longitude involved in Hansen's tables was :
ψ= 5022".30 T0 + 1".121 T0 2 .
Hence the coefficient of the quadratic term of the sidereal longitude was 12".18 /cy2 . It resulted from theory, i.e. from
the secular terms of the Earth's eccentricity, but Hansen had verified that it was in agreement with observations of
ancient eclipses.
Hansen's tables contained also the long period term (239 years):
∆ LHansen = 21".47 sin(8V − 13T + 274 °14 ' )

19
whose coefficient was empirical. Later Delaunay emphasized that his analytical calculation of this term yielded a
smaller value, by a factor of a hundredth. Note that Burckhardt's tables contained also a long period term (179 years)
with an empirical coefficient whose theoretical amplitude was much smaller:
∆LBurckhardt = 15" cos(2 Ω + ϖ + 3ψ) .

Soon after the publication of Hansen's tables, Adams (1853) and Delaunay (1859) demonstrated that, since
Laplace, an error existed in the lunar theories concerning the computation of the quadratic term of the longitude, and
that the actual theoretical value was smaller than Hansen's by about 6"/cy2 . Since Hansen's value seemed to agree with
observations, Adams's and Delaunay's results raised the question of the invariability of the Earth's rotation rate,
observations being dated in time scales tied to the Earth's rotation and theory involving an uniform time scale. The time
argument of the lunar tables was in mean solar time, tied to the Earth's rotation, and it was necessary to fit the
coefficient of the quadratic term of the longitude. Nevertheless, the possibility of an explanation of the 'empirical terms'
by Newtonian gravitation was not eliminated since a precise investigation of the perturbations of the lunar motion had
not yet been performed.
The corrections to Hansen's tables by Newcomb (1878) mainly concern the mean mean longitude and
empirical term ∆L. They result from comparisons of Hansen's tables to modern and ancient observations, the coefficient
of the quadratic term and the empirical term (coefficient and argument) being fit. Hansen's tropic mean mean longitude
and empirical term are replaced by:
LNewcomb/1 = 335°43'25".56 + 1 732 564 390".44 T0 + 9".44 T0 2 + 0".0135 T0 3 ,
∆ LNewcomb/1 = − 15".49 cos(18V − 16 T − l ) − 0".09 sin(18V − 16 T − l )

T0 being the mean solar time elapsed from 1800.0 in Julian centuries.
Later, new corrections were fit by Newcomb (1912), yielding:
LNewcomb/2 = 335°43'26".39 + 1 732 564 393".04 T0 + 9".07 T0 2 + 0".0068 T0 3 ,
∆LNewcomb/2 = 12".95 sin(131° T0 + 100°.6).
In Radau's tables, the fitted expression of the tropic lunar mean mean longitude was:
LRadau = 335°43'26".70 + 1 732 564 392".61 T0 + 8".821 T0 2 + 0".013473 T0 3 ,
and two empirical terms were introduced:
∆LRadau = 11".5 cos(1° .37θ1 ) + 3".3 cos(5 °. 55θ2 )

where θ1 and θ2 are the numbers of Julian years reckoned from 1790.5 and 1856.5 respectively.
In the second part of the 19th century and at the beginning of the 20th , theoretical studies about tides are
performed by several authors. The contribution of tides to the quadratic term of the lunar longitude becomes an
evidence, as Brown writes (1910): The reaction of tidal friction produces a real secular retardation of the Moon's
motion, as well as the apparent acceleration due to the slowing down of the Earth's rotation. The former is nearly equal
to the latter, and the real retardation would be between two and three times the observed acceleration. There being no
data on which to base any exact numerical estimates of either of these quantities, the secular acceleration will be
considered as an observed quantity, the magnitude of the apparent value being not very different from that (5".8) found
by the attraction of the planets. He is nevertheless wrong when he states that the real and apparent tidal effects on the
Moon's motion (dated in mean solar time) are nearly opposite as we shall see below.
However, tides do not explain the empirical terms. Radau (1911) believes that they could arise from the
perturbations by small planets located between the Sun and Mercury, but it is not the opinion of Brown (1910) who
writes: If these inequalities have a real existence, it would seem that the cause must be sought in some action not purely
gravitational.
Brown prefers to introduce in his tables the theoretical values he has obtained for the coefficients of quadratic
and cubic terms of the longitude (Table 6). He probably considers that fitted values of the quadratic term are not
precise enough and must be improved. On the other hand he believes that the 'observed value' (in mean solar time scale)
is close to the theoretical one (in uniform time scale). He is nevertheless obliged to include an empirical term in his
tables in order to satisfy to observations ; its coefficient, mean motion, and phase, as well as the mean motion and the
constant term of the mean mean longitude, are fit to the expression LNewcomb/2 + ∆LNewcomb/2 slightly modified in order
to take into account long period terms from planetary perturbations which were not known by Newcomb (Brown,
1915). The five unknowns are obtained by solving the equations of condition for five equidistant dates from 1724 till
1928. The results are:

20
 LBrown = 335°43'27".81 + 1 732 564 391".80 T0 + 7".12 T0 2 + 0".0068 T0 3 ,
∆LBrown = 10".71 sin(140°.0 T0 + 100°.7).
With Brown's tables an important gain of precision had been obtained in the 'Newtonian position' of the Moon.
New comparisons to all the observations available could be undertaken, leading to new determinations of the 'observed'
coefficient of the quadratic term of the lunar longitude. In particular, De Sitter found that Brown's value was too small
by 5".22 /cy2 . Simultaneously, fluctuations with respect to Newcomb's theory were detected in the longitudes of the Sun
and planets.
Spencer Jones (1939) cleared the problem of the discrepancies between observed and theoretical motions of the
Moon and planets by distinguishing two causes to the variations of the Earth's rotation: tidal friction and redistribution
of matter inside the Earth. Tidal friction causes a retardation of the Earth's rotation coupled with a true deceleration of
the Moon's orbital motion, but it has no true effect on planets. Redistribution of matter produces fluctuations in the
Earth's rotation, but no true effect on all the orbital motions. Variations in the Earth's rotation, whatever their causes are,
give through the mean solar time scale, apparent effects on the longitudes of Moon and planets, proportional to their
mean motions. Spencer Jones rediscussed previous analysis of a large number of observations of the Moon, Sun,
Mercury, and Venus from the 17th century till the 20th one. Setting the quadratic term of the tropic lunar longitude to the
value obtained by De Sitter, he obtained the following corrections to Brown's lunar longitude and to Newcomb's solar
longitude in the mean solar time scale, respectively:
∆LSpencer Jones = - ∆LBrown + 4".65 + 12".96 T2 + 5".22 T2 2 + B (7)
∆L'Spencer Jones = 1".00 + 2".97 T2 + 1".23 T2 2 + 0.0747B (8)
T2 is the mean solar time reckoned from 1900.0 (1900 January 0, Greenwich mean noon) in Julian centuries, and B the
apparent fluctuation of the Moon's longitude due to redistribution of matter inside the Earth. The fitted coefficient of the
quadratic term of ∆ LSpencer Jones (i.e. 5".22 /cy2 ) results from the apparent effect of the tidal friction

n
× 1".23 = 16 ".44 /cy2
n'
and of the true effect which, by difference is found to be –11".22 /cy2 . Contrary to the conjecture of Brown quoted
above, the apparent tidal acceleration differs from the true retardation by about 50%.
Following Spencer Jones's paper, Clemence (1948) proposes to adopt a 'Newtonian time' as time argument in
the ephemerides, Universal time (derived from mean solar time) being still used for dating observations and for current
use. The Ephemeris time is adopted by the IAU in 1952, its unit and origin being defined by the solar mean mean
longitude in 1900 from Newcomb's theory. Then expression (8) gives (IAU, 1954):
ET – UT = 24s .349 + 72 s .318 T2 + 29 s .950 T2 2 + 1.82144B. (9)
In (9) the term including B corresponds to fluctuations of UT due to non tidal effects. Ephemeris time can be introduced
as time argument in Brown's tables, as well as Spencer Jones's correction (7), by means of the unique correction to be
added to the longitude :
∆LET= - 8".72 – 26".74 tc – 11".22 tc2 - ∆LBrown (10)
where tc is Ephemeris time in Julian centuries reckoned from 1900.0.
With the introduction of a Newtonian time, empirical terms have disappeared from lunar ephemerides and the
global coefficient of the quadratic term of the sidereal longitude is the sum of the perturbations by planets and Earth's
figure (6".03 /cy2 ), and of the tidal effect (-11".22 /cy2 ).

5. ANALYTICAL LUNAR EPHEMERIDES AT THE ERA OF ELECTRONIC COMPUTERS

5.1 Ephemerides based on Brown's solution
The 8th IAU general assembly, held in 1952, recommended the introduction of the Ephemeris time in lunar ephemeris as
well as the direct use of Brown's series instead of his tables. This was made possible by the development of electronic
computers, and a previous test had been performed by Eckert, Jones, and Clark on a IBM SSEC.
The list of terms in Brown's series, and the method for computing the apparent lunar polar coordinates form the
Improved Lunar Ephemeris (ILE) j = 0 (Eckert et al., 1954). The series used in ILE j = 0 are those of longitude,
transformed latitude (Brown, 1911), and sine of parallax. The coefficients involve the set of constants labeled 'Tables/1'
in this paper (Table 7). The set 'Tables/2' and the secular terms of e' are introduced in the computation of the lunar
positions by means of the method described in Sect. 4.3 plus small additional corrections. Two other terms are added in

21
order to get a better evaluation of aberration. The empirical term ∆LBrown is disregarded and the lunar mean arguments
are those of Brown's tables except for the tropic mean mean longitude which is corrected by means of the polynomial
part of Eq. (10). So the time argument is Ephemeris time, and Spencer Jones' correction is taken into account. The
tropic mean mean longitude is:
LILE = 270°26'2".99 + 1 732 564 379".31 tc - 4".08 tc2 + 0".0068 tc3 ,
the general precession in longitude, from Newcomb, being (Eckert et al., 1966):
ψ = 5 025".75 tc + 1".11 tc2 .
Hence the sidereal mean mean longitude involved in ILE is:
λILE = 270°26'2".99 + 1 732 559 353".56 tc - 5".19 tc2 + 0".0068 tc3 . (11)
Planetary arguments are referred to the fixed equinox of 1850 instead of mean equinox of date (an error in Brown's
tables). ILE j = 0 was introduced in national ephemerides from 1960 till 1967.
Ephemeris ILE j = 1 differs from ILE j = 0 by the introduction of the new value for the constant term of sine of
parallax recommended by the IAU in 1964 (Table 7), and by the correction – 0".072sin(2l - 2F) in longitude, from an
error in Brown's series put into evidence by the solution of Eckert and Smith (see Sect. 2.3). ILE j = 1 replaced ILE j =
0 in national ephemerides from 1968 till 1971.
In Ephemeris ILE j = 2, Brown's main problem series for longitude and sine of parallax are completed by
additional series, and the series of transformed latitude are replaced by a new unique series for latitude. New series and
additional series have been obtained by Eckert, Walker, and Eckert from Brown's original series in rectangular
coordinates, more precise than his series in polar coordinates. ILE j = 2 was used in national ephemerides from 1972 till
1983.

5.2 New analytical ephemerides

Among the recent contributions to analytical lunar theory, only ELP theory has given rise to ephemerides. Several
versions of ELP are used.
ELP 2000-82 is the basic theory. It includes the main problem under a semi-analytical form (see Sect. 2.2 and
2.3) and all the perturbations described in Sect. 3, the tidal perturbations being computed with Williams's model. The
perturbations to the mean mean longitude and to the mean longitudes of perigee and node are restricted to the powers 1
and 2 of time. The nominal values of the constants involved in the coefficients of the main problem (Table 7) are those
of the ILE (Eckert et al, 1966), n n', and e' being converted to J2000.0. Note that Π0 is not introduced in ELP which
gives the distance instead of the sine of parallax. The constant term of the distance involves n (which replaces Π0 in
Table 7), the geocentric constant of gravitation and the ratio of the Moon's and Earth's masses. The values adopted for
the last two parameters are those recommended by the IAU in 1976, as well as the values of planetary masses involved
in the perturbations. The values of the other physical constants are close to the adopted or fitted values in DE200.
In order to get an ephemeris, ELP 2000-82 has been compared to DE200 on the time span [1900 ; 2000], the
differences between the values adopted for the geocentric constant of gravitation and lunar mass being taken into
account. The following constants have been fit in the comparison: E, Γ, n, e', n', the values of λ, ϖ, Ω, T, ϖ' in J2000.0,
the inclination ε of the mean ecliptic of J2000.0 on the reference plane (equator ) of DE200, the angle φ separating the
origin of right ascensions in DE200 from the 'DE200 inertial mean equinox of J2000.0' (i.e. the ascending node of the
mean ecliptic of J2000.0, involved in ELP, on the equator of DE200), and three bias parameters consisting in
corrections to the theoretical values of the mean motions of perigee and node (once the corrections to orbital elements
have been included), and tidal secular acceleration. The first four values are given in Table 7 ; the other ones can be
found in (Chapront-Touzé and Chapront, 1983). The new values of the constants involved in the main problem
coefficients, have to be substituted in the series of ELP 2000-82 and in the mean motions of perigee and node by means
of the derivatives provided in ELP 2000-82. Fitted or corrected mean motions and angular constant terms are used for
computing the values of the arguments for each value of time. The series yield longitude, latitude, and distance. The
first two coordinates are referred to the mean ecliptic of date. The origin of the longitude is a J2000.0 departure point
derived from the 'DE200 inertial mean equinox of J2000.0', as described in Sect. 3.1. The 'DE200 inertial mean equinox
of J2000.0' differs from the inertial dynamical mean equinox of J2000.0 as far as the equator of DE200 differs from the
mean equator of J2000.0. The separation between the equinoxes can be taken into account by adding a small constant
correction to the longitude. Mean longitude of date is obtained by adding the general precession in longitude ψ. The so
obtained ephemeris is labeled as ELP 2000. It is used in the Connaissance des temps since 1984. It has also replaced
Brown's solution in the construction of tables requiring precise lunar coordinates (e.g. for occultations) such as the
second edition of Meeus's Astronomical tables of Sun, Moon and planets (1995). Note that the fitted bias parameters are

22
not included in ELP 2000. In a comparison of the right ascensions and declinations computed from ELP 2000 and from
ILE j = 2 over the time span [1978 ; 1984] (Francou et al., 1983), the differences can reach 0".75.
In order to make the solution suitable for the analysis of ancient observations, improvements have been
brought later to the theoretical part of the angular mean elements of the Moon. The secular perturbations to the mean
mean longitude and to the mean longitudes of perigee and node have been expanded up to the power 4 of time, and
better values of the secular terms of the Earth's eccentricity have been used. On the other hand, ancient observations do
not require a highly precise solution and the series of ELP 2000-82 are very bulky. So the series have been truncated.
The new mean elements joined to the truncated series form the solution labeled ELP 2000-85 (Chapront-Touzé and
Chapront, 1988). The values of constants proposed for ELP 2000-85 were the same as in ELP 2000. The resulting
ephemeris is published in Lunar tables and programs from 4000 B.C. to A.D. 8000 (Chapront-Touzé and Chapront,
1991), where the listed series involve the corrected constants. The new mean elements can also be joined to the
complete series ; the solution is then labeled ELP 2000-82B. With ELP 2000 constants, the expression of the sidereal
mean mean longitude is:
λELP 2000 = 218°18'59".955 71 + 1 732 559 343".736 04 t - 5".8883 t2 + 0".006 604 t3 – 0".000 03169 t4 (12)
where t is time in TDB* scale (see Sect. 5.4) reckoned from J2000.0 (2000 January 1, 12h TDB) in Julian centuries. It is
referred to the 'DE200 inertial mean equinox of J2000.0'.
The post fit residuals of the comparison of ELP 2000-82 to DE200 were of 0".011 at maximum in V , 0".006
in U, and 15 meters in r. They represent over one century the global maximum errors of the series. If this precision is
sufficient for most astronomical applications, it is insufficient when compared to the accuracy of LLR observations.
Since a few years, the precision of ELP has been considerably increased by introducing numerical
complements derived from the comparison to a numerical integration (Chapront and Chapront-Touzé, 1997). The new
version is labeled ELP 2000-96 ; the numerical complements are derived from the JPL numerical integration DE245 as
described below. The series and the analytic expressions of the mean elements introduced in ELP 2000-96 are those of
ELP 2000-82B, except that :
• The values of the geocentric constant of gravitation and ratio of the Moon's and Earth's masses are those of DE245
(it has been sufficient to change the constant term of r) ;
• The perturbations due to the Earth's and Moon's figures have been computed again with the values of the physical
constants of DE245 (the differences with the previous results are very small) ;
• The tidal perturbations computed with the new model and the physical constants of DE245 (see Sect. 3.2.3) have
been introduced instead of the previous ones.
The analytic part of ELP 2000-96 has been compared to DE245 over the time span [1900 ; 2010] and the same
parameters have been fit as in the comparison of ELP 2000-82 to DE200. The numerical complements are given by the
differences ρ = DE245 – ELP, after the substitution of all the fitted values, including bias parameters, in ELP (here
ELP stands for the analytic part of ELP 2000-96). Practically the numerical ρ have been build as a sequence of
Chebychev polynomials as in the source DE245. The analytical series of the main problem and of perturbations other
than the planetary ones, have a good precision and involve the same values as DE245 for the non-fitted parameters, with
close models of forces. So the numerical complements consist for the most part of corrections to the planetary
perturbations series, due to missing contributions, round off errors, and inaccurate values of masses. As far as they are
sufficiently small, they do not depend on small changes in the orbital parameters and can be considered as independent
of the numerical integration, if the same values of masses are kept.
Via its analytical part, ELP 2000-96 keeps all the advantages of a semi-analytical theory and its internal
precision is at the level of a few centimeters. These properties have been verified by a comparison to the JPL numerical
integration DE403 while fitting the same parameters as previously (Chapront and Chapront-Touzé, 1997). Note that for
the comparison, the tidal perturbations series of ELP 2000-96 have been changed to the series involving the same values
of the physical parameters as DE403, and the difference between the two values of the geocentric constant of
gravitation has been taken into account. The results are partially given in Table 7. The expression of the sidereal mean
mean longitude becomes:
λELP/DE403 = 218°18'59".874 84 + 1 732 559 343".356 24 t - 6".7772 t2 + 0".006 604 t3 – 0".000 03169 t4 (13)
It is referred to the 'DE403 inertial mean equinox of J2000.0' separated from the 'DE200 inertial mean equinox of
J2000.0' by 0".04085.
LLR measurements are the observational basis of the lunar part of JPL numerical integrations, e.g. DE200,
DE245, DE403. Recently, new analytical ephemerides based on ELP 2000-96 and constants directly fit to LLR

23
observations have been obtained by Chapront, Chapront-Touzé, and Francou (1999). The comparison of ELP 2000-96
to LLR observations is shortly described in the next paragraph. The results are partially given in Table 7.

5.3 Lunar laser ranging, the observational basis of modern ephemerides

Lunar laser ranging consists in measuring the round-trip time of light between a LLR station on the Earth and a lunar
reflector. Using the velocity of light, these time measurements are equivalent to distance measurements. The first
retroreflector array was placed on the Moon on 21st July 1969. The period of observations covers now about 30 years.
Four efficient reflectors have been landed on the Moon (Apollo 11, Apollo 14, Apollo 15 and Lunakhod 2). Three
stations are (or were) involved in LLR campaigns of observation: McDonald Observatory (Texas, USA) from 1969,
Haleakala (Maui, Hawaii) from 1987 till 1990, CERGA (Grasse, France) from 1984. The accuracy of the measurements
has continuously decreased since the first echoes, from a few ten centimeters on the one-way distance in the seventies to
the centimeter level presently. The very high accuracy of the LLR measurements provides with the Moon an
exceptional 'laboratory of metrology' (Dickey et al., 1994). Precision and duration of measurements have brought during
the past decades a lot of physical knowledge on the Moon and have contributed to a great improvement of the
ephemerides.
LLR stations provide normal points i.e. an average of several tens measurements, covering about 10 minutes of
observations, in order to diminish individual errors. They are used as observed values of the light time ∆t from LLR
station transmitter O at time t1 to a lunar reflector R at time t2 , and then from R at t2 to the station receiver O at t3 . This
'observed value' ∆tO is compared to the corresponding 'computed value', ∆tC. The propagation time has to be corrected
of several effects (relativistic curvature of the light beams, influence of the troposphere) and relativistic changes of time
scales have to be performed in order to express the various components in the same time scale.
In the frame of the General Relativity theory, the computed ∆tC depends on the barycentric positions of T
(Terrestrial mass center), L (Lunar mass center), and S (Solar mass center). Hence the computation involves:
• The coordinates of TL (Earth-Moon) in a celestial barycentric reference system provided by lunar theory, and the
barycentric coordinates of the Earth-Moon barycenter provided by planetary theory ;
• The coordinates of TO (observing station) in the same celestial reference system. It requires a precise knowledge of
the Earth rotation (precession, nutation, polar motion, and Universal time through the sidereal time), coordinates of
the station in a terrestrial reference system, and relativistic corrections ;
• The coordinates of LR in the celestial reference system. It requires a theory of the lunar librations (forced and free
libration) and selenocentric coordinates of the reflector.
Actually, the complete model to represent at best the observations is extraordinary complex and one has to know or
force a great variety of models: plate motions, atmospheric delay, solid tides of Earth and Moon.
In the comparison of ELP 2000-96 to LLR observations, the fitted parameters are similar to those mentioned
above in the comparison of ELP 2000-82 to DE200 (see Sect. 5.2), with in addition free libration parameters, reflector
coordinates, and precession constant. The equator and origin of right ascensions of DE200 are replaced by the celestial
equator and origin of right ascensions to which the Earth's rotation is referred by means of precession, nutation, polar
motion, and Universal time. So parameters ε and φ fix the position of the mean ecliptic of J2000.0 with respect to that
system. Various systems can be considered according to the models and data used for precession, nutation, polar
motion, and Universal time. Physical parameters such as harmonic coefficients of the lunar potential, Love numbers and
time delays, could also be fit, but the leading corrections induced on the analytic ephemeris should concern the tidal
secular acceleration and the mean motions of perigee and node. Hence we have preferred to keep in the set of fitted
parameters the three bias parameters mentioned above, which stand as the results of those corrections. Consequently,
the bias parameters are added to the theoretical values in the angular mean elements for the derived ephemerides. The
resulting expression of the sidereal mean mean longitude (Chapront et al., 1999) is:
λELP/LLR = 218°18'59".835 36 + 1 732 559 343".344 39 t - 6".8417 t2 + 0".006 604 t3 – 0".000 03169 t4 (14)
It is referred to the inertial dynamical mean equinox of J2000.0, which is separated from the 'ICRS inertial mean
equinox of J2000.0' by 0".0460. We denote by 'ICRS inertial mean equinox of J2000.0' the ascending node of the mean
ecliptic of J2000.0 on the reference (equatorial) plane of the International Celestial Reference System recently adopted
by IAU.

24
5.4 Time argument of new ephemerides, mean motions, and secular acceleration
Since ELP is referred to a relativistic barycentric reference system, the time argument of the ELP theory is a barycentric
relativistic time. Two scales can be used : TDB (Temps dynamique barycentrique) and TCB (Temps coordonnée
barycentrique). We have chosen to use a time scale labeled here as TDB*, which differs from TCB by linear terms only
and which, at the present state of accuracy, does not differs from TDB. TDB* is similar to the time scale involved in
JPL numerical integrations (Standish, 1998b). It has the advantage of continuity with the Ephemeris time (as far as
small periodic terms are disregarded) and does not introduce large differences with the ILE lunar sidereal mean motion
(referred to J2000.0) as TCB should do. TDB* is the time scale of the ephemerides based on the ELP theory. At a
precision not better than 0".01, as it is required for common lunar ephemerides, TDB* can be replaced by TT
(Terrestrial time) without any change in the lunar ephemeris.
The first two bias parameters obtained in the various fits are the differences between the values of the
theoretical mean motions of perigee and node, after substitution of the fitted values of the other parameters, and the
implicit values involved in the numerical integration (fits to DE200 and DE403) or the observed values (fit to LLR
observations). From the results given in (Chapront-Touzé and Chapront, 1983 ; Chapront and Chapront-Touzé, 1997 ;
Chapront et al., 1999), one may believe that the precision of the theoretical mean motions in the most recent analytical
lunar ephemerides is about 0".05/cy.
Expressions (12) to (14) show the evolution with respect to the time span covered by LLR observations, in the
value of the tidal part of the coefficient of the longitude quadratic term (half tidal secular acceleration). The value has
decreased by about 1" from DE200, which involved about 10 years of LLR observations, to the recent determinations,
which involve about 30 years. The most recent fitted value (eq. 14) is –12".901 /cy2 , smaller than Spencer Jones's value
(-11".22 /cy2 ) by 1".68/ cy2 . It is close to the value –13"± 1" /cy2 obtained by Morrison and Ward (1975) from optical
observations covering a time span of three centuries, but the incertitude is better, probably about 0".05 /cy2 .
6. CONCLUSION
Many progresses have been done by analytical ephemerides during the 20th century. They are tied to progresses in lunar
theory, knowledge of the Earth's rotation, technology, and computer science.
In the last field, specific tools were necessary to manipulate Fourier series either in an analytical form or in a semi-
numerical form. In the seventies various Poisson series manipulators have been developed in several institutions to
perform the necessary operations on formal series:
• elementary operations: addition, multiplication, integration, derivation…;
• substitution of numerical values for a list of formal parameters, and time substitution ;
• operations with Poisson brackets and Lie transforms ;
• management: ordering, sorting, compression-decompression of formal arguments or monomials,…
A crucial problem in all that business was the handling of bulky objects, and the necessity to develop fast procedures to
arrange terms in very large arrays and monitor the precision that can be governed for example by the integer powers
associated to the 'small quantities' (e, e', γ, α, m). It is worth noticing that the general tools and formal languages that are
commonly used for algebraic manipulations like FORMAC or MATHEMATICA were not much useful in lunar theory.
Because of the bulk of the objects and the necessary optimization to save computer time, most of the astronomers have
developed their own manipulators suited to their project.
The development of electronic computers has also given rise to a new type of lunar ephemerides based on
numerical integration. The advantages and disadvantages of analytical lunar ephemerides with respect to numerical ones
are discussed hereafter.
If we ignore the mathematical problems of convergence beyond the scope of this contribution, the length of
validity of the main problem is unlimited in the sense that the solution can be expressed in a form of purely periodic
terms. Practically it is valid over several million years if we ignore the uncertainty on the integration constants. The
length of validity is considerably reduced by planetary perturbations, or more precisely because of the planetary
developments in Poisson's form, valid only on a few thousand years. In planetary solutions such as those of Le Verrier,
Bretagnon (1982), or Bretagnon and Francou (1988), Poisson's terms are due to the expansion in time of the secular
motions of planetary nodes and perihelions (long period inequalities of ten thousand years or more). Hence, a lunar
solution is also valid only on a few thousand years.
Within such a time span of a few thousand years, an analytical ephemeris allows to compute lunar position for
any date, but the accuracy decreases as the separation from the epoch of the theory grows. The leading cause of that loss
of accuracy is the incertitude on the value of the tidal secular acceleration, but good progress has been realized over the
last ten years, and that loss of accuracy is probably less than 0".05/ cy2 in longitude. Note that numerical integrations
suffer from the same limitation.

25
 Within a time span of a few centuries, analytical ephemerides without numerical complements give the lunar
position with an accuracy of a few ten meters, which is far under the accuracy of numerical ephemerides. Nevertheless
with numerical complements, the accuracy of analytical ephemerides can be improved by a factor of 1000. Note that the
large number of terms in analytical ephemerides at their full precision is no longer a difficulty with the appearance of
very performing computers.
The principal advantage of using analytical theories for the construction of ephemerides is their modularity.
The various components can be easily changed as for example the series of tidal perturbations from ELP 2000-82 to
ELP 2000-96, and new constants can be easily introduced. Furthermore analytical ephemerides can be easily shortened
in function of the needs of users, by disregarding the smallest coefficients. For example, Meeus, in his Astronomical
algorithms (1991), uses ELP 2000-85 for giving lunar tables reduced to less than 70 terms for each coordinate:
longitude, latitude and distance, and whose precision is sufficient for many users.

REFERENCES

Adams, J.C.: 1853, On the secular variation of the Moon's mean motion, Monthly Notices Roy. Astron. Soc., 14, 59
Bidart, P., Chapront, J.: 1999, Moon's planetary perturbations, in Proceedings of the IAU colloquium 172, J. Henrard
and S. Ferraz-Mello eds, Kluwer, p. 413
Bobone, J.: 1937a, Tablas del VI satelite de Jupiter, Astr. Nach., 262, 321
Bobone, J.:1937b, Tablas del VII satelite de Jupiter, Astr. Nach., 263, 401
Bretagnon, P.: 1982, Théorie du mouvement de l'ensemble des planètes. Solution VSOP82, Astron. Astrophys., 114,
278
Bretagnon, P., Francou, G.: 1988, Planetary theories in rectangular and spherical variables. VSOP87 solutions, Astron.
Astrophys., 202, 309
Brown, E. W.: 1896, An introductory treatise on the lunar theory, Cambridge University Press. Republication: Dover
(1960), New York
Brown, E.W.: 1899, Theory of the motion of the Moon, …, part I and part II, Mem. Roy. Astron. Soc., 53, 39 and 163
Brown, E.W.: 1903, On the formation of the derivatives of the lunar coordinates with respect to the elements, Trans. of
Amer. Math. Soc., 64, 524
Brown, E.W.: 1904a, Theory of the motion of the Moon, …, part III, Mem. Roy. Astron. Soc., 54, 1
Brown, E.W.: 1904b, On the degree of accuracy of the new lunar theory and on the final values of the mean motions of
perigee and node, Monthly Notices Roy. Astron. Soc., 64, 524
Brown, E.W.: 1904c, On the completion of the solution of the main problem in the new lunar theory, Monthly Notices
Roy. Astron. Soc., 65, 104
Brown, E.W.: 1905a, The final values of the coefficients in the new lunar theory, Monthly Notices Roy. Astron. Soc., 65,
276
Brown, E.W.: 1905b, On a general method for treating transmitted motions and its applications to indirect perturbations,
Trans. Amer. Math. Soc., 6, 332
Brown, E.W.: 1908a, Theory of the motion of the Moon, …, part IV, Mem. Roy. Astron. Soc., 57, 51
Brown, E.W.: 1908b, Adams Prize Essay, Cambrigde Univ. Press
Brown, E.W.: 1909a, On an error in the new lunar theory, Monthly Notices Roy. Astron. Soc., 70, 3
Brown, E.W.: 1909b, On an addition to the theoretical secular acceleration of the Moon's mean motion, Monthly
Notices Roy. Astron. Soc., 70, 143
Brown, E.W.: 1910, Theory of the motion of the Moon, …, part V, Mem. Roy. Astron. Soc., 59, 1
Brown, E.W.: 1911, The transformation of the Moon's latitude, Monthly Notices Roy. Astron. Soc., 71, 651
Brown, E.W.: 1915, The elements of the Moon's orbit, Monthly Notices Roy. Astron. Soc., 75, 508
Brown, E.W.: 1919, Tables of the motion of the Moon, with the assistance of H.B. Hedrick chief computer, Yale Univ.
Press

26
Brumberg, V.A., Ivanova, T.V.: 1981, Proceedings of colloquium VI ERMA in Dubrovnik
Burckhardt, J.K.: 1812, Tables de la Lune, Tables astronomiques publiées par le Bureau des longitudes, Paris
Bureau des Longitudes: 1997, Introduction aux éphémérides astronomiques, J.L. Simon, M. Chapront-Touzé,
B. Morando and W. Thuillot eds, Les Éditions de Physique, p. 235
Chapront, J., Chapront-Touzé, M.: 1982, Planetary perturbations of the Moon in ELP 2000, Celest. Mech., 26, 83
Chapront, J., Chapront-Touzé, M.: 1997, Lunar motion: theory and observations, Celest. Mech., 66, 31
Chapront, J., Chapront-Touzé, M., Francou, G.: 1999, Determination of the lunar orbital and rotational parameters and
of the ecliptic reference frame orientation from LLR measurements and IERS data, Astron. Astrophys., 343, 624
Chapront-Touzé, M.: 1974, Construction itérative d'une solution du problème central de la Lune. Influence des petits
diviseurs, Astron. Astrophys., 36, 5
Chapront-Touzé, M.: 1980, La solution ELP du problème central de la Lune, Astron. Astrophys., 83, 86
Chapront-Touzé, M.: 1982, The ELP Solution for the main problem of the Moon and some applications, Celest. Mech.,
26, 63
Chapront-Touzé, M.: 1983, Perturbations due to the shape of the Moon in the lunar theory ELP2000, Astron.
Astrophys., 119, 256
Chapront-Touzé, M., Chapront, J.: 1980, Les perturbations planétaires de la Lune, Astron. Astrophys., 91, 233
Chapront-Touzé, M., Chapront, J.: 1983, The lunar ephemeris ELP2000, Astron. Astrophys., 124, 50
Chapront-Touzé, M., Chapront, J.: 1988, ELP2000-85: a semi-analytical lunar ephemeris adequate for historical times,
Astron. Astrophys., 190, 342
Chapront-Touzé, M., Chapront, J.: 1991, Lunar tables and programs from 4000 B.C. to A.D. 8000, Willmann- Bell
Chapront-Touzé, M., Henrard, J.: 1980, Le problème principal du mouvement de la Lune. Comparaison de deux
théories, Astron. Astrophys., 86, 221
Clemence, G.M.: 1948, On the system of astronomical constants, Astron. J., 84, 889
Delaunay, C.: 1859, Sur l'accélération séculaire du moyen mouvement de la Lune, Comptes rendus Acad. Sciences, 48,
137
Delaunay, C.: 1860, Théorie du mouvement de la Lune, volume 1, Mém. Acad. Sciences, 28, Paris
Delaunay, C., 1867: Théorie du mouvement de la Lune, volume 2, Mém. Acad. Sciences, 29, Paris
Deprit, A., Henrard, J., Rom, A.: 1971a, Analytical Lunar Ephemeris: Delaunay's theory, Astron. J., 76, 269
Deprit, A., Henrard, J., Rom, A.: 1971b, Analytical Lunar Ephemeris: definition of the main problem, Astron.
Astrophys., 10, 257
Dickey, J.O., Bender, P.L., Faller J.E., Newhall XX, Ricklefs, R.L., Ries, J.G., Shelus, P.J., Veillet, C., Wipple, A.L.,
Wiant, J.R., Williams, J.G., Yoder, C.F.: 1994, Lunar laser ranging: a continuing legacy of the Appolo program,
Science, 265, 482
Eckert, W.J., Jones, R., Clark, H.K.: 1954, Improved Lunar Ephemeris 1952-1959, US Government Printing Office,
Washington
Eckert, W.J., Smith, H.F.: 1966, The solution of the main problem of the lunar theory by the method of Airy, Astron.
Pap. Amer. Ephem. and Naut. Almanac, 19, 187
Eckert, W.J., Walker, M.J., Eckert, D.: 1966, Transformation of the lunar coordinates and orbital parameters, Astron. J.,
71, 314
Euler, L.: 1768, Réflexions sur la variation de la Lune, Hist. et Mem. Acad. Sc. Berlin, 1766, 334
Explanatory supplement to the Astronomical Ephemeris and the American Ephemeris and Nautical Almanac: 1961, Her
Majesty Stationery Office, p. 177
Francou, G., Bergeal, L., Chapront, J., Morando, B.: 1983, Nouvelles éphémérides du Soleil, de la Lune et des planètes,
Astron. Astrophys., 128, 124
Gutzwiller, M. C.: 1979, The numerical evaluation of Eckert's lunar ephemeris, Astron. J., 84, 889

27
Gutzwiller, M. C., Schmidt, D., S.: 1986, The motion of the Moon as computed by the method of Hill, Brown and
Eckert, Astron. Pap. Amer. Ephem. and Naut. Almanac, 23 part. I
Hansen, P.A.: 1857, Tables de la Lune construites d'après le principe newtonien de la gravitation universelle, London.
Henrard, J.: 1971, Analytical Lunar Ephemeris: a report, Facultés universitaires de Namur
Henrard, J.: 1973, L'éphéméride analytique lunaire - ALE, Ciel et Terre, 89, 1
Henrard, J.: 1979, A new solution to the main problem of lunar theory, Celest. Mech., 19, 337
Henrard, J.: 1980, Perturbations due to the shape of the Moon in lunar theory, Celest. Mech., 22, 335
Henrard, J.: 1981, The Earth figure perturbations in the lunar theory, Celest. Mech., 25, 417
Hill, G.W.: 1877, Researches in the lunar theory, Amer. J. of Math., 5, 129 and 245
Hill, G.W.: 1891, Determination of the inequalities of the Moon's motion which are produced by the figure of the Earth,
Astron. Pap. Amer. Ephem., 3
IAU: 1954, Transactions of the international Astronomical Union, vol. VIII
Kinoshita, H.: 1982, Comparison of lunar ephemerides (SALE and ELP) with numerical integration, in High precision
Earth rotation and Earth-Moon dynamics, Reidel Publ. Co., Dordrecht, 245
Kubo, Y.: 1982, Perturbations by the oblateness of the Earth and by the planets in the motion of the Moon, Celest.
Mech., 26, 97.
Kubo, Y.: 1983, Perturbations by the figure of the Moon in the motion of the Moon, Report of Hydrographic
Researches n° 18
Lambeck, K.:1980, The Earth's variable rotation, Cambridge University Press
Laplace, P.S. de: 1802, Mécanique céleste, tome 3, livre VII
Laskar, J.: 1986, Secular terms of classical planetary theories using the results of general theory, Astron. Astrophys.,
157, 59
Lestrade, J.F., Bretagnon, P.: 1982, Perturbations relativistes pour l'ensemble des planètes, Astron. Astrophys., 105, 42
Lestrade, J.F., Chapront, J., Chapront-Touzé, M.: 1982, The relativistic planetary perturbations and the orbital motion of
the Moon, in High precision Earth rotation and Earth-Moon dynamics, Reidel Publ. Co., Dordrecht, 217
Lestrade, J.F., Chapront-Touzé, M.: 1982, Relativistic perturbations of the Moon in ELP2000, Astron. Astrophys., 116,
75
Meeus, J.: 1991, Astronomical algorithms, Willmann- Bell
Meeus, J.: 1995, Astronomical tables of the Sun, Moon, and planets, 2nd edition, Willmann- Bell
Moisson, X.: 1999, Solar system planetary motion to third order of the masses, Astron. Astrophys., 341, 318
Moons, M.: 1982, Analytical theory of the libration of the Moon, The Moon and the planets, 27, 257
Morrison, L.V., Ward, C.G.: 1975, An analysis of the transits of Mercury 1677 – 1973, Monthly Notices Roy. Astron.
Soc., 173, 183
Newcomb, S.: 1878, Researches on the motion of the Moon, I, Government Printing Office, Washington
Newcomb, S.: 1912, Researches on the motion of the Moon, II, Astron. Pap. Amer. Ephem., 9, part I
Newhall, X.X., Standish, E.M., Williams, J.G.: 1983, DE102: A numerically integrated ephemeris of the Moon and the
planets spanning forty-four centuries, Astron. Astrophys., 125, 150
Radau, R.: 1893, Recherches concernant les inégalités planétaires du mouvement de la Lune, Annales Obs. Paris, 21,
B1
Radau, R.: 1911, Tables de la Lune fondées sur la théorie de Delaunay, Annales Bureau des longitudes, 7
Schmidt, D., S.: 1980, The main problem of the lunar theory solved by the method of Brown, The Moon and the
planets, 23, 135
Soffel, M.: 1989, Relativity in astronomy, celestial mechanics, and geodesy, Springer, Berlin

28
Spencer Jones, H.: 1939, The rotation of the Earth, and the secular acceleration of the Sun, Moon, and planets, Monthly
Notices Roy. Astron. Soc., 99, 541
Standaert, D.: 1980, Direct perturbations of the planets on the Moon's motion, Celest. Mech., 22, 357
Standaert, D.: 1983, Direct perturbations of the planets on the Moon's motion: results and comparisons, Celest. Mech.,
30, 21
Standish, E.M.: 1998a, JPL planetary and lunar ephemerides, DE405/LE405, Jet Propulsion Laboratory Interoffice
Memorandum 312.F-98-048
Standish, E.M.: 1998b, Time scales in the JPL and CfA ephemerides, Astron. Astrophys., 336, 381
Williams, J.G., Sinclair, W.S., Yoder, C.F.: 1978, Tidal acceleration of the Moon, Geophys. Res. Letters, 5, 943

29