Available online at www.sciencedirect.

com

ScienceDirect
Advances in Space Research 69 (2022) 538–553
www.elsevier.com/locate/asr

Analysis of the PPN two-Body Problem using non-osculating

orbital elements
Pini Gurfil a,⇑, Michael Efroimsky b
a
Faculty of Aerospace Engineering, Technion, Haifa 3200003, Israel
b
US Naval Observatory, Washington, DC 20392 USA

Received 30 July 2021; received in revised form 7 September 2021; accepted 11 September 2021
Available online 22 September 2021

Abstract

The parameterised post-Newtonian (PPN) formalism is a weak-field and slow-motion approximation for both General Relativity
(GR) and some of its viable generalisations. Within this formalism, the motion can be approached using various parameterisations,
among which are the Lagrange-type and Gauss-type orbital equations. Often, these equations are developed under the premise of the
Lagrange constraint. This constraint makes the evolving orbital elements parameterise the instantaneous conics always tangent to the
actual trajectory. Arbitrary mathematically, this choice of a constraint is convenient under perturbations dependent only on positions.
However, under perturbations dependent also on velocities (like in the relativistic celestial mechanics) the Lagrange constraint unneces-
sarily complicates solutions that can be simplified by relaxing the constraint and introducing a freedom in the orbit parameterisation,
which is analogous to the gauge freedom in electrodynamics and gauge field theories. Geometrically, this freedom is the freedom of
non-osculation, i.e., of the degree to which the instantaneous conics are permitted to be non-tangent to the actual orbit. Under the same
perturbation, all solutions with different degrees of non-osculation look mathematically different, though describe the same physical
orbit. While non-intuitive, the modeling of an orbit with a sequence of nontangent instantaneous conics can at times simplify calcula-
tions. The appropriately generalised (‘‘gauge-generalised”) Lagrange-type equations, and their applications, appeared in the literature
heretofore. We, in this paper, derive the gauge-generalised Gauss-type equations and apply them to the PPN two-body problem. Fixing
the gauge freedom in three different ways (i.e., modeling an orbit with non-osculating elements of three different types), we find three
parameterisations of the PPN two-body dynamics. These parameterisations render orbits with either a fixed non-osculating semimajor
axis, or with a fixed non-osculating eccentricity, or with a fixed non-osculating argument of periastron. We also develop a transformation
from non-osculating to classical osculating orbital elements, and illustrate the new solutions using numerical simulations.
Ó 2021 COSPAR. Published by Elsevier B.V. All rights reserved.

Keywords: PPN formalism; General relativity; Gauge freedom; Relativistic celestial mechanics

1. Introduction

The oral tradition attributes to John Wheeler the follow-

ing formulation of the essence of general relativity (GR): 1
‘‘Spacetime tells matter how to move; matter tells space-
⇑ Corresponding author. time how to curve.” A possible footnote to this proverbial
E-mail addresses: pgurfil@technion.ac.il (P. Gurfil), michael.efroims- quote could be that in the zero-curvature limit matter keeps
ky@navy.mil (M. Efroimsky).
1 moving — though not necessarily in a Newtonian manner,
For recent critical overviews of GR see, e.g., (Iorio, 2015; Debono and
Smoot, 2016; Vishwakarma, 2016; Beltrán Jiménez et al., 2019) and GR because motion can be fast (luminal, for photons). Hence
centennial jubilee volumes edited by Rovelli (2015) and Ashtekar (2005). the question: would a slightly curved spacetime tell matter

https://doi.org/10.1016/j.asr.2021.09.009
0273-1177/Ó 2021 COSPAR. Published by Elsevier B.V. All rights reserved.
P. Gurfil, M. Efroimsky
how to deviate slightly from Newtonian trajectories? A
short answer to this question is partially positive: the
zero-curvature slow-motion limit of GR is Newtonian;
and there exists an approximation of GR, known as the
post-Newtonian (PN) formalism, intended to adjust some
GR problems to the Newtonian framework (Will, 2011).
But then, it turns out that this approximation is fraught
with obstacles.
A natural attempt to fit GR into a purely Newtonian
framework is to consider the N-body problem in an asymp-
totically flat spacetime covered with a single global coordi-
nate map. This treatment bears a trace of Newton’s
absolute space and time. However, practical needs (like
the relativistic treatment of tidal forces) require introduc-
tion of N þ 1 coordinate patches: a comoving patch for
each body, and a global patch (Kopejkin, 1988). The treat-
ment then implies writing down the relations linking each
local coordinate system to the global one (Damour,
1987). This development was pioneered by Brumberg and
Kopejkin (1989) and was later extended by Damour et al.
(1991) who included all multipole moments in the expan-
sion of the gravitational field.
The second major difficulty of the PN approach is
describing post-Newtonian motion of extended bodies.
While in the Newtonian theory the motion of spherically-
symmetric bodies is identical to that of point particles, this
identity holds in PN only for the translational motion of
nonrotating spherically symmetrical bodies. So, for model-
ing the translational motion of such bodies, a knowledge of
their masses is enough — a so-called effacing principle
(Kopeikin and Vlasov, 2006). But, in the presence of rota-
tion or for curved trajectories, the description requires the
knowledge not only of their masses but also of their multi-
pole momenta (Kopeikin, 1985; Panhans and Soffel, Dec.
2014; Meichsner and Soffel, 2015; Soffel et al., Jan. 1988;
Iorio, 2015, 2019).
Despite these challenges, the PN formalism is applicable
to various physical settings and renders numerous valuable
results. According to Blanchet (2014), this formalism has
three principal applications:
1. The modeling of the N-body solar system dynamics (i.e.
of the motion of the planets’ centres of mass) incorpo-
rates the first PN approximation, that of the order
ðv=cÞ2 (Iorio, 2015).
2. The description of the gravitational radiation-reaction
force emerging in pulsar dynamics employs the 2.5PN,
i.e. order-ðv=cÞ5 equations (Grishchuk and Kopeikin,
1983; Damour, 1983; Iorio, 2019; Iorio, 2021). The
model has been experimentally verified via observations
of the secular acceleration of the orbital motion of the
Hulse-Taylor binary pulsar PSR 1913 + 16 (Taylor,
1992, 1993) and of the pulsar PSR J0737-3039
(Possenti et al., 2004).
3. The equations of motion and the expressions for the
radiation field, written down to higher PN orders, show
539
Advances in Space Research 69 (2022) 538–553
up in the analysis of gravitational waves emitted by
inspiralling compact binaries, cosmic pairs comprising
two black holes, or two neutron stars, or a black hole
and a neutron star (Abbott et al., 2021), driven into coa-
lescence by emission of gravitational radiation
(Cervantes-Cota et al., 2016; Miller and Yunes, 2019;
Bailes et al., 2021).
Addressing the first of these three applications, for point
masses, our paper deals with the first post-Newtonian
(1PN) approximation, the one taking care of the terms
up to ðv=cÞ2 inclusively.
The modern PN theory is fit to approximate the GR
along with some of its generalisations. This tool is called
the Parameterised Post-Newtonian formalism, or PPN
(Blanchet, 2003; Will, 2018; Asada and Futamase, 1997).
Within this parameterisation, the relativistic force can be
expressed through a variation of the Lagrangian. One such
force model was developed by Chazy (1928) and Brumberg
(1991), another by Damour and Deruelle (1985).
Our goal in this paper is to find solutions to the PPN
two-body problem, in non-osculating elements (the bodies
assumed spherically symmetric and non-rotating). In one
such solution, the non-osculation is picked up in such a
way that the non-osculating semimajor axis stays fixed.
In another solution, the non-osculation is such that the
non-osculating eccentricity stays constant. In the third
solution, the non-osculation is chosen in such a manner
that the non-osculating argument of periastron is fixed.
Under the so-arranged parameterisations, orbit integration
gets simplified.
Be mindful that in our parameterisations it is the total
rates (not just their secular parts) that are nullified.
2. Methods
When a perturbing force is introduced into the Newto-
nian two-body setting, the classical variation-of-
parameters method suggests to recast the orbital elements
as functions of time. The resulting differential equations,
describing the temporal change of the Keplerian orbital ele-
ments for a conservative, position-only dependent per-
turbing potential, are known as the orbital equations in
the form of Lagrange. On the other hand, the orbital equa-
tions in the form of Gauss, also known as the Gauss varia-
tional equations or GVE, model the time evolution of the
orbital elements due to an arbitrary perturbing force.
For details of the method, we refer the reader to Appen-
dix A, while a more comprehensive explanation can be
found in Efroimsky (2002) and Efroimsky (2005). In short,
the method works as follows. The perturbed vectorial
equation of motion has three projections, which are three
scalar second-order differential equations for the three
Cartesian coordinates. In the Cauchy form, they render
six first-order equations for the three coordinates xa and
three velocities va . These are the three projections of the
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

equation of motion, reading as With the Lagrange constraint relaxed, the Lagrange-
dva =dt ¼ uðC 1 ; . . . ; C 6 ; C_ 1 ; . . . ; C_ 6 Þ, and the three obvious type or Delaunay-type orbital equations acquire the so-
equations dxa =dt ¼ va . Be mindful that the insertion of called gauge-generalised form. In this form, these equations
xa ¼ f a ðC 1 ðtÞ; . . . C 6 ðtÞ Þ in the equation of motion entails are presented in Efroimsky (2002, 2005) and Efroimsky
three (for a ¼ 1; 2; 3) scalar equations for the twelve vari- and Goldreich (2003, 2004). For an example of their
ables C i ðtÞ and C_ i ðtÞ. (We get C_ i because the ‘‘constants” employment in astrophysics, see the work by Dosopoulou
are time dependent). If we now treat C_ i as independent and Kalogera (2016). In this paper, we derive a gauge-
variables, and amend the equations of motion with the generalised form of the orbital equations in the form of
Gauss. To this end, we utilise the freedom endowed by
six trivial equations dC i =dt ¼ C_ i , we end up with nine
relaxing the Lagrange constraint, and arrive at new solu-
first-order differential equations for the twelve variables
tions to the relativistic PPN two-body problem of point
C i ðtÞ, C_ i . Therefore, three arbitrary scalar constraints can masses. Fixing the gauge in three different ways, we find
be imposed on the said twelve variables. three such new solutions written in terms of non-
Deriving his variational equations, Lagrange set these osculating orbital elements. These solutions render orbits
three constraints so that the actual physical velocity in each with either a fixed non-osculating semimajor axis, or a fixed
point became equal to the Keplerian velocity (see Appendix non-osculating eccentricity, or a fixed non-osculating argu-
A). Thereby he ensured that in each point of the orbit the ment of periastron. We also develop a transformation from
instantaneous Keplerian conic (the one defined by the non-osculating to osculating orbital elements, and illustrate
instantaneous values of the time-varying orbital elements) the new solutions using numerical simulations.
is tangent to the orbit — so that the perturbed physical tra-
jectory would coincide with the Keplerian orbit along 3. The Relativistic Force in the PPN Formalism
which the body would move if the perturbing force were
to cease instantaneously. Written in the vector form, these 3.1. The Lagrangian Perturbation and the Relativistic Force
constraints go under the name of osculation condition. The
corresponding instantaneous orbit is called an osculating When a Lagrangian L0 ðr; r_ ; tÞ ¼ m r_ 2 =2  U ðr; tÞ is
orbit, while the orbital elements satisfying the Lagrange modified by an additional perturbing term DLðr; r_ ; tÞ the
constraint are called osculating orbital elements (Kopeikin Euler–Lagrange equations written for the perturbed setting
et al., 2011; Gurfil and Seidelmann, 2016).
Mathematically, the imposition of any triple of scalar Lðr; r_ ; tÞ ¼ L0 þ DL
constraints (or of one vector constraint) confines the r_ 2
dynamics of the orbital state space to a 9-dimensional sub- ¼ m  U ðr; tÞ þ DLðr; r_ ; tÞ ð1Þ
2
manifold of the 12-dimensional manifold spanned by the
orbital elements and their time derivatives (Efroimsky, acquire the form of:
2002, 2005). The choice of the constraints, however, is arbi- @U
m €r ¼  þ F ; ð2Þ
trary. The employment of constraints different from @r
Lagrange gives birth to non-osculating orbital elements. the term F being the disturbing force:
Thus, while the physical orbit remains invariant to the par-

ticular selection of constraints, its description in the @ DL d @ DL
F   : ð3Þ
orbital-element space looks different for different constraint @r dt @ r_
choices. We shall apply this formalism to the classical reduced two-
Orbital elements, osculating or not, canonical or not, are body problem Lagrangian
useful mathematical variables that are not available in di-
rect astronomical measurements. Direct measurements ren- r_ 2 GM
L0 ðr; r_ Þ ¼ m þ ð4Þ
der us parallaxes, variations of brightness, etc. It is only a 2 r
posteriori that we translate data into elements. At the same disturbed by the lowest-order relativistic correction
time, while not being immediately observable, elements (Brumberg, 1991)
may have convenient mathematical sense. For example, " #
osculating elements parameterise a sequence of instanta- 1  2 2 1 r_ 2 ðr  r_ Þ
2
DL ¼ 2 B1 r_ þ B2 2 þ B3 þ B4 ;
neous conics sharing one focus and tangent to the physical c r r r3
orbit. Non-osculating elements parameterise instantaneous
ð5Þ
conics non-tangent to the physical orbit. However, non-
osculating elements of a certain type (the so-called contact B1 ; . . . ; B4 being constants given by Eq. (8) below.
elements) osculate the orbit (i.e., are tangent to it) in the Now m has the meaning of the reduced mass:
phase space (Efroimsky, 2005; Efroimsky and Goldreich, m1 m2
2003). The choice of elements to employ is dictated by cal- m¼ ; ð6Þ
m1 þ m2
culational convenience, analytical or numerical.
while M is the total mass of the system:

540
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

M ¼ m1 þ m2 ; ð7Þ (1985). who employed parameters a0 , a1 , a2 , b1 . The

Chazy–Brumberg and Damour–Deruelle formulations are
m1 and m2 being the partners’ masses.
equivalent. The relations among the parameters of each
Insertion of (5) into (3) renders the so-called Chazy
of these force models can be found by comparing the
force. Historically, it was pioneered by Chazy (1928) for
expressions for the radial, transverse, and normal force
General Relativity. We, however, shall be interested in a
components (i.e. Eqs. (13) and (36), appearing in the
generalisation of this force to a broader class of theories
sequel). This procedure yields the relations
falling within the scope of the PPN formalism.
a0 ðeÞ ¼ 2ðr  Þ  ð2   2 l  3 aÞ
3.2. The PPN Perturbation in the Chazy–Brumberg Form e2 ¼  2 a þ 2 b þ c þ ðc þ 2Þ e2 ; ð14aÞ
a1 ¼ r  4 Þ ¼ 2 b  4 a ð14bÞ
Within the PPN parameterisation, which embraces both
the GR and a class of its mathematical generalisations, the a2 ¼  ð2 l þ 3 aÞ ¼  a  2 c  2 ð14cÞ
Lagrangian variation retains the above form, with the con- b1 ¼ 2 l ¼ 2 c  2 a þ 2; ð14dÞ
stants expressed through the Newton gravity constant G,
the total mass M, and parameters a; ; l; r: the inverse expressions being
  1
B1  2a2þl ; B2  ðGMÞ2  þ 12 l  r ; a ¼  ða2 þ b1 Þ ; ð15aÞ
8
  ð8Þ 3
B3  GM  a þ  þ 12 l ; B4  GMa ; b1
l ¼ ; ð15bÞ
see Eq. (3.1.112) in (Brumberg, 1991). 2
An equivalent expression through G; M, and parameters a0  a1 þ a2 e 2
 ¼ ; ð15cÞ
a; b; c, as given by equation (3.1.45) from (Brumberg, 2 ð1  e2 Þ
1991), is in use also. Interrelation between the two param- 2 a0  a1 ð1 þ e2 Þ þ 2 a2 e2
eterisations reads: r ¼ ; ð15dÞ
2 ð1  e2 Þ
r¼bþca ; 2 ¼ c þ a ; l ¼ c  a þ 1 ; ð9aÞ 2 a0  a1 ð1 þ e2 Þ þ 2 a2 e2 1
b¼  b1 þ 1 ; ð15eÞ
the inverse formulae being 2 ð1  e Þ
2 2
1 1 a0  a1 þ a2 e 2 1 1
a¼ lþ ; b ¼ r  lþ1 ; c c ¼ þ b1  : ð15fÞ
2 2 2 ð1  e2 Þ 4 2
1 1
¼  þ l  : ð9bÞ
2 2 4. Gauge-invariant Gauss Variational Equations
Within the first parameterisation, the specific case of GR
corresponds to Our goal in this section is to introduce a gauge-
r ¼ 2  a ; 2 ¼ 1 þ a ; l ¼ 2  a : ð10Þ generalised form of the Gauss variational equations
(GVEs). In this form, these equations will be used to derive
Within the second, to new parameterisations of the relativistic two-body
b ¼ c ¼ 1 : ð11Þ problem.

Consequently, the Bi coefficients read, for GR, as 4.1. Generalised Form of the Gauss Equations
ðGRÞ ðGRÞ 2 
B1 ¼ 18 ; B2 ¼ ðGMÞ a  12 ;
ðGRÞ   ðGRÞ
ð12Þ We begin with a short reminder of how the gauge free-
B3 ¼ GM  a þ 32 ; B4 ¼ GM a : dom (freedom of non-osculation) shows itself in celestial
In GR, the PPN perturbation exerts on a point-like particle mechanics. After this, we shall derive the gauge-
the so-called Chazy–Brumberg force, which assumes (see generalised equations in the form of Gauss.
Appendix B) the form
" ! 4.1.1. Gauge freedom
GM 2 GM ðr  r_ Þ2 r Consider the equations of motion in the Newtonian
r  2  ð r_ Þ þ 3 a
2
F ¼ 2 2 two-body problem:
c r r r3
 GM
ðr  r_ Þ r_ €r þ r¼0 ; ð16Þ
þ2l : ð13Þ r3
r3
where M ¼ ðm1 þ m2 Þ is the total mass, with m1 and m2
being the masses of the partners, and r is the position vec-
3.3. Thse PPN Perturbation in the Damour-Deruelle Form tor in some inertial frame of reference. The solution for r is
known (Battin, 1999), and can be symbolically expressed as
Another PPN-type parameterisation often used in the
literature is the one suggested by Damour and Deruelle r ¼ fðt; C 1 ; . . . ; C 6 Þ ; ð17Þ
541
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

where C i are constants of motion known as orbital constraints are needed. This becomes especially clear when
elements. we write the equations of motion in the form of Cauchy.
Introducing the velocities r_ and writing the equation of First observed by Eforimsky (2002), this ambiguity is an
motion (16) in the Cauchy form, we observe that (17) can internal symmetry, in that a gauge transformation (a switch
be expressed as a mapping from one set of constraints to another) leaves the physical
trajectory unchanged. So, mathematically, it is analogous
ð C 1 ; . . . ; C 6 Þ $ ð xðtÞ; yðtÞ; zðtÞ; x_ ðtÞ; yðtÞ;
_ z_ ðtÞ Þ ;
to the gauge freedom in electrodynamics.
ð18Þ As was suggested in Ibid., we do not set U to be zero,
but permit it to be an arbitrary vector function of the orbi-
which is one-to-one over one orbit cycle, see Appendix A
tal parameters:
for details.
In the presence of a perturbing force F, equation (16) U ¼ UðC 1 ðtÞ; . . . ; C 6 ðtÞ Þ : ð26Þ
assumes the form
So we keep U both in (21) and in the subsequent deriva-
GM tions. Thereafter, we choose one or another particular
€r þ 3 r ¼ F ; ð19Þ
r functional form of UðC 1 ðtÞ; . . . ; C 6 ðtÞ Þ to nullify the rate
and its standard solution requires the ‘‘constants” of of one or another orbital element, as given by the orbital
motion C i to become time-dependent: equations. Thus, we employ the three projections of U to
set three arbitrary constraints, whose necessity was dis-
r ¼ fðt; C 1 ðtÞ; . . . ; C 6 ðtÞÞ : ð20Þ cussed above. Stated alternatively, we use U as a tool to
For time-dependent C i , the velocity reads: remove the gauge freedom. An important caveat is that
for U – 0 the elements C i come out non-osculating.
r_ ¼ g þ U ; ð21Þ Substituting expression (24) for €r into the equation of
where motion (19), we obtain the following variational equations
for the orbital elements:
@
gðt; C 1 ; . . . ; C 6 Þ  fðt; C 1 ; . . . ; C 6 Þ ; ð22Þ X
@t
6
@f _
Ci ¼ U ; ð27aÞ
while the ‘‘convective term” is i¼1
@C i

X6
@g _
X6
@f _ _ ;
Ci ¼ F  U ð27bÞ
U Ci : ð23Þ @C i
i¼1
@C i i¼1

The acceleration is then given by 2 where the first equation is a restatement of (23), while
gðt; C 1 ; . . . ; C 6 Þ is the Keplerian velocity given by expres-
@2f X 6
@2f _ _ : sion (22).
€r ¼ þ Ci þ U ð24Þ
@t2 i¼1
@C i @t To combine equations (27) into a single expression, we
take the dot product of equation (27a) with @C j =@f and
As explained in Appendix A, correspondence (18) should equation (27b) with @C j =@g and then add the two resulting
now be changed to expressions, arriving at
 
C 1 ðtÞ; . . . ; C 6 ðtÞ; C_ 1 ðtÞ; . . . ; C_ 6 ðtÞ ! X 6

ð25Þ @C j @f @C j @g _
 þ  Ci
ð xðtÞ; yðtÞ; zðtÞ; x_ ðtÞ; y_ ðtÞ; z_ ðtÞ Þ ; @f @C i @g @C i
i¼1

a time-dependent mapping between a 12-dimensional space @C j _ þ @C j  U

¼  ðF  UÞ ð28Þ
and a 6-dimensional space. This mapping evidently cannot @g @f
be one-to-one, and the resulting ambiguity of the parame- 3
or, in a vector form,
terisation with C i ðtÞ and C_ i ðtÞ can be removed by setting
arbitrary constraints on the functions C i ðtÞ and/or C_ i ðtÞ. @C _ þ @C U :
C_ ¼ ðF  UÞ ð29Þ
Since the insertion of ansatz (20) into the three projections @g @f
of the equation of motion (19) furnishes us with three con-
ditions on C i and C_ i (Eq. (27) below), only three arbitrary Here we employed the notation C ¼ ½C 1 ; . . . ; C 6 T and kept
in mind that C i bears no explicit dependence on time.
2
Pioneered in the preprint by Efroimsky (2002), equa-
Be mindful that dot denotes a full time derivative. So, in Eq. (24), the
tions (29) are the gauge-generalised variational equations
second term does not emerge as a partial time derivative of expression (23).
It shows up as a result of the differentiation of @f=@t: of orbital motion. The ensuing gauge-generalised orbital
equations in the forms of Lagrange and Delaunay were
d @f @2f X6
@2 f _
¼ 2þ Ci : derived by Efroimsky and Goldreich (2003, 2004), see also
dt @t @t @C i @t
i¼1 Efroimsky (2005). A comprehensive discussion on the topic
is provided in the book by Kopeikin et al. (2011). For an
542
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

example of practical use of this method, see Dosopoulou gauge-invariant GVEs. The resulting equations are valid
and Kalogera (2016). in an arbitrary gauge, and are hence a generalised form
^ þ FS
of the classical GVEs. Writing F ¼ F R R
4.1.2. Derivation of the generalised Gauss-type equations
The derivation of the gauge-generalised orbital equa- ^ þ FWW
S ^ ¼ ½F R ; F S ; F W T ; U ¼ UR R ^ þ UW W
^ þ US S ^
T
tions in the form of Gauss comprises two steps. First, it ¼ ½UR ; US ; UW 
is convenient to keep on the right-hand side of (29) only
the partial time derivative of the gauge function U. To this and U _ ¼ U_ R R ^ þ U_ W W
^ þ U_ S S ^ ¼ ½U_ R ; U_ S ; U_ W T provides
end, we substitute the following gauge-generalised GVEs (wherein we used
the notation U_  @U=@t):
_ ¼ @U þ @U C_
U ð30Þ
@t @C    
da F R  U_ R a2 e sin f F S  U_ S a2 p U R a2
into the right-hand side of (29), to obtain ¼2 þ2 þ2 2 ;


 dt h hr r
@C @U _ @C @U @C ð34aÞ
Iþ C¼ F þ U; ð31Þ    
@g @C @g @t @f de F R  U_ R p sin f F S  U_ S ððp þ rÞ cos f þ reÞ
¼ þ
where I is a 6  6 identity matrix. dt h h
The second step is to express f; g; @C=@g; F and @C=@f in UR ðcos f þ eÞð1 þ e cos f Þ US sin f
þ þ ; ð34bÞ
a coordinate system whose origin is set at the primary. In   p a
this coordinate system, usually referred to as the RSW di F W  U_ W r cos ðf þ xÞ UW ðsin ðf þ xÞ þ e sin xÞ
^ is directed along the radius vector ¼ þ ;
frame, the unit vector R dt h p
of the orbiter, radially outwards, while S ^ is perpendicular ð34cÞ
 
^ residing in the instantaneous orbital plane defined
to R, dX F W  U_ W r sin ðf þ xÞ
¼
by R^ and the instantaneous orbital velocity. As ever, W ^ dt h sin i
completes the right-hand triad, so that R ^ S ^ ¼ W.^ UW ½ðsin ðf þ xÞ þ e sin xÞr sin ðf þ xÞ  p
þ ;
_
Within the orthodox approach, U and U appearing in rp sin i cos ðf þ xÞ
   
Eq. (29) are set to be identically zero. This is why in all dx F R  U_ R p cos f F S  U_ S ðp þ rÞ sin f
¼ þ
the textbooks and papers, which develop the Gauss-type dt he he
 
orbital equations, only the expression for @C=@g are listed F W  U_ W r sin ðf þ xÞ cos i UR sin f ð1 þ e cos f Þ
 þ
(Battin, 1999). Thus, to obtain explicit expressions for the h sin i pe
rates of change of the orbital elements, we shall have to US ðcos f þ eÞ

first derive expressions for @C=@f, which are absent in the pe
h i
current literature. Presented in Appendix C, this derivation UW ðcos ðf þ xÞÞ2  sin ðf þ xÞe sin x þ e cos f cos i
is performed by choosing the classical orbital elements as þ ;
T p sin i cos ðf þ xÞ
the phase variables, i. e. C ¼ ½a; e; i; X; x; l0  , where a is
ð34eÞ
the semimajor axis, e is the eccentricity, i is the inclination,  
X is the right ascension of the ascending node, x is the dl0   ð2 e þ cos f þ e cos2 f Þð1  e2 Þ
l ¼ F R  U_ R
argument of periastron, and l0 is defined as 4 dt e ð1 þ e cos f Þna
 
Z t   ðe2  1Þðe cos f þ 2Þ sin f
þ F S  U_ S
l0 ¼ l  ndt ; ð32Þ e ð1 þ e cos f Þna
t0
" #
3  e þ 2 cos2 f ð1 þ e2 Þ  2 e cos f sin2 f
2
þUR pffiffiffiffiffiffiffiffiffiffiffiffiffi
where l is the mean anomaly, t0 is a reference time (which 2 a e 1  e2 sin f
may differ from the periastron passage time), and pffiffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi US 1  e2 cos f
n ¼ GM=a3 is the mean motion. þ ; ð34fÞ
ae
In addition, we write the position and velocity vectors in
the RSW frame, so that
where p ¼ að1  e2 Þ is the semi-latus rectum and
^ ½v ¼ @r R
^ þ r @f S
^ ; pffiffiffiffiffiffiffiffiffiffi
½rR ¼ rR; ð33Þ h ¼ GMp is the magnitude of the angular momentum vec-
R
@t @t
tor. Another useful relationship is the gauge-generalised
f being the true anomaly. We shall omit the subindex R variational equation for the true anomaly, obtained by
for the sake of brevity. using Eq. (125):
By collecting all the expressions derived in Appendix C

and substituting them into equation (29), we obtain the f_ ¼ rh2 þ eh1 p cos f ðF R  U_ R Þ  ðp þ rÞ sin f ðF S  U_ S Þ
 ep Ucos
R
f
sin2 f ð1 þ e cos f Þ þ ep UcosS f sin f ðe2 cos f þ 2e þ cos f Þ :
4
To avoid confusion with our notation for the mass, we have used
Brumberg’s notation (Brumberg, 1991) for the mean anomaly. ð35Þ
543
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

4.2. The Post-Newtonian Force Model which is correct to first-order in e. We now re-write (31)
into
The perturbing force in the first post-Newtonian  
~ 
approximation of general relativity can be written by using C_ ¼ I  eK~ @C eF ~  e @ U þ @C eU~ : ð44Þ
either the Chazy–Brumberg (Chazy, 1928; Brumberg, 1991) @g @t @f
or Damour – Deruelle (Damour and Deruelle, 1985) forms. Retaining terms up to first-order in e implies that Eq. (31)
These two force models can be rendered equivalent by a can be written as
proper choice of the force coefficients. We shall hence use

@C @U @C
the Damour – Deruelle form. C_ ¼ F þ U : ð45Þ
Since the post-Newtonian force model bears direct @g @t @f
dependance on the velocity, its parametrisations with oscu-
Consequently, the gauge velocity in the presence of the
lating and non-osculating elements will be different. The
post-Newtonian perturbations will be several orders of
expressions for the force components, written in the
magnitude smaller than the nominal Keplerian orbital
RSW frame, are given by Damour and Deruelle (1985),
" #
velocity. (in our example considered later on,
GM ðr  r_ Þ2 GM OðeÞ ¼ 108 ; for e.g. binary pulsars, it can be larger,
F R ¼ 2 2 ð1 þ 3mÞ_r  r_ þ ð4  0:5mÞ 2 þ ð4 þ 2mÞ
cr r r though still much smaller than unity.) Therefore, since we
are seeking first-order solutions, we can omit second-
ð36aÞ
order expressions of the gauge; this will result in the follow-
GM
F S ¼ 2 3 ð4  2mÞðr  r_ Þ_r ð36bÞ ing expressions for the force components, written in terms
cr
of non-osculating orbital elements:
F W ¼0 ; ð36cÞ
GMe

where c is the speed of light, M ¼ m1 þ m2 is the total mass, FR ¼ 2

ð1 þ e cos f Þ2 a0 ðeÞ þ a1 e cos f þ a2 e2 cos2 f
p
while m denotes the dimensionless reduced mass: sffiffiffiffiffiffiffiffi
m1 m2 GM eUR 2
m¼ : ð37Þ  ð1 þ e cos f Þ ð12 þ 14mÞe sin f
2 p3 2
ðm1 þ m2 Þ sffiffiffiffiffiffiffiffi
Based on Eq. (21), we have r_ ¼ v þ U, with v given by Eq. GM eUS
 ð1 þ e cos f Þ2 ½ð12m þ 4Þe cos f þ 4 þ 12m ;
(33). Hence the gauge U will show up in (36). p3 2
At this point, we shall perform an order-of-magnitude ð46aÞ
analysis in order to quantify the effect of the gauge on GMe 3
the force components, which is necessary for expressing F S ¼ 2 ð1 þ e cos f Þ b1 e sin f
p
the force components in terms of orbital elements. To that sffiffiffiffiffiffiffiffi
end, we define the dimensionless parameter GM
þ eUR b1 ð1 þ e cos f Þ3
p3
GM sffiffiffiffiffiffiffiffi
e¼ 1 : ð38Þ
c2 p GM 2
þ eUS b1 ð1 þ e cos f Þ e sin f ; ð46bÞ
We first note that the post-Newtonian force components p3
can be expressed as a function of the small parameter e, F W ¼0 ; ð46cÞ
~ :
F ¼ eF ð39Þ where

In addition, because the gauge U can be determined ad lib, a0 ðeÞ ¼ 3  m þ 3e2  3:5me2 ; a1 ¼ 2  4m; a2 ¼ 4 þ 0:5m;
we choose it to be proportional to e as well, so that b1 ¼ 4  2m ¼ 0:4ða1  2a2 Þ :
~ :
U ¼ eU ð40Þ ð47Þ

Looking at (31), let us denote To first order, neglecting the terms eUR and eUS , we obtain
the expressions
@C @U ~ ;
K ¼ eK ð41Þ G2 M 2 2

@g @C FR ¼ ð1 þ e cos f Þ a0 ðeÞ þ a1 e cos f þ a2 e2 cos2 f ;

c2 p 3
where, due to (40), ð48aÞ
~
~  @C @ U :
K ð42Þ FS ¼
G2 M 2
ð1 þ e cos f Þ3 b1 e sin f ; ð48bÞ
@g @C c2 p 3
We now utilise the approximation, FW ¼0 ; ð48cÞ
 1
~ ~ which are identical to the force components written in
I þ eK  I  eK; ð43Þ
terms of the osculating elements. Thus, to first order, the
544
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

gauge velocity does not affect the post-Newtonian ple closed-form solutions illustrating the utility of the pro-
perturbation. posed approach.
In the relativistic reduced two-body problem, we have
m ! 0, and the coefficients become 5.1. Constant Semimajor Axis
a0 ðeÞ ¼ 3 þ 3e ; a1 ¼ 2; a2 ¼ 4; b1 ¼ 4 :
2
ð49Þ
Looking at Eqs. (34a) and (34b), we note that there
could be UR and US that simultaneously nullify a_ and e_ .
5. Fixing the Gauge However, this would require solving two linear non-
autonomous, non-homogenous differential equations. We
Both Brumberg (1991) and Damour and Deruelle (1985) _
shall thus nullify aand e_ separately.
solved the GVEs (34) with osculating elements (i.e. U ¼ 0 From Eq. (34a) it is evident that if UR  0, then a_ ¼ 0 if
and U_ ¼ 0) and the post-Keplerian force components (48). 2
G2 M 2 e sin f ½a0 þ a1 e cos f þ a2 e2 cos2 f þ b1 ð1 þ e cos f Þ 
The solution was obtained under the following U0S ¼ pffiffiffiffiffiffiffiffiffiffiffiffi
assumptions: c2 GMp3 ð1 þ e cos f Þ
ð53Þ
1. The orbital elements appearing in the right-hand side of and, for some integration constant j1 ,
the GVEs are treated as constants. Z
2. The independent variable is transformed from time to US ¼ U0S df
true anomaly by approximating (cf. (Brumberg, 1991), G2 M 2 ½2ða0  a1 þ a2 Þ lnð1 þ ecf Þ þ ½ða2 þ b1 Þe2 c2f þ 2ða1  a2 þ b1 Þecf 
p. 7; in this approximation, we use only the Keplerian ¼ pffiffiffiffiffiffiffiffiffiffiffiffi
2c2 GMp3
rate of the true anomaly), þj1 ; ð54Þ
sffiffiffiffiffiffiffiffi
df @f h GM 2
 ¼ 2¼ ð1 þ e cos f Þ ; ð50Þ
dt @t r p3
where we have used the compact notation
0
so that, using the notation ðÞ to denote derivative with cx ¼ cos x; sx ¼ sin x. The resulting semimajor axis now sat-
respect to true anomaly, isfies a ¼ a0 , where ðÞ0 denotes an initial value, and the
remaining orbital elements can be evaluated by substituting
@f
C_ j  C 0j : ð51Þ Eqs. (53) and (54) into the GVEs (34).
@t Brumberg (1991, p. 7) introduced a modified semimajor
axis aH , expressed as a function of the semimajor axis and
Assumptions 1 and 2 imply that the solutions obtained eccentricity. In our presentation, the non-osculating semi-
for the osculating orbital elements under the post- major axis equals the initial semimajor axis.
Keplerian perturbation are first-order approximations; nev- A useful approximation for US given in Eq. (54) can be
ertheless, these solutions are of prime importance for both obtained for near-circular motion in the restricted two-
modeling and experimental validation of GR. To obtain body problem; in this case,
new parameterisations, we shall solve the GVEs (34) with
3
_ will not be identically 7e GM 2
non-osculating elements, i. e. U and U US ¼  2 cos f þ Oðe2 Þ ð55Þ
zero. In fact, we shall use the extra degrees-of-freedom c a
introduced by U to nullify the rates of change of some of and the approximate post-Newtonian parametrisation with
the orbital elements, thus providing simpler parameterisa- non-osculating elements is given by
tions of the relativistic two-body problem. To that end,
a ¼a0 ; ð56aÞ
we shall adopt Assumptions 1 and 2, and, in the spirit of  
GM  
Assumption 2, write that (cf. Eq. (31) e ¼e0 1 þ 2 11  12cf =e0 þ 11c2f þ Oðe2 Þ ; ð56bÞ
4c a
@U @U @f h GM

¼ ¼ U0 : ð52Þ x ¼x0 þ 2 2f  ð15e þ 12e1 Þsf þ 11s2f  7es3f þ Oðe2 Þ ;

@t @f @t r2 4c a
This will allow us to use f as an independent variable in a ð56cÞ
seamless manner. GM

l0 ¼ðl0 Þ0 þ 2 ð12e1 þ 45eÞsf  11ðs2f þ 2f Þ þ 7es3f þ Oðe2 Þ :

In addition, since the post-Newtonian effect manifests 4c a
itself as a perturbing force in the orbital plane (F W ¼ 0), ð56dÞ
X and i remain constant. To preserve the two-
dimensionality of the orbit parametrisation with non-
5.2. Constant Eccentricity
osculating elements, we must therefore set UW ¼ U_ W ¼ 0.
We are thus left with two degrees-of freedom – Requiring that e_ ¼ 0 and choosing US  0 yields the dif-
ðUR ; U_ R ; US ; U_ S Þ – that can be used to nullify, at most, ferential equation
two additional orbital elements. We shall now derive sam-
545
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553


sf ð1þecf Þ2
UR ðf Þ0 þ
ðcf þeÞð1þecf Þ
UR ðf Þ ¼
G2 M 2 ð1þecf Þ2 sf where j2 is a constant of integration. Choosing UR as in Eq.
p p hp2 c2
n o ð57Þ (61) will yield x ¼ x0 . The remaining elements can be cal-
 a0 þ a1 ecf þ a2 e2 c2f þ b1 e½ð2 þ ecf Þcf þ e ; culated using the GVEs (34).
We can again obtain a simple approximation for UR by
assuming near-circular motion in the restricted two-body
whose solution is given by setup:
 
3 
 
G2 M 2 sf 2sf 1 GM 2 1 þ sf
UR ðf Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffi ln ½ð4b1 þ 2a0 UR  2 3cf ln ð1  ecf Þ þ eð13f cf  8sf Þ :
cf ð1 þ cf Þ c a cf
2c2 GMp3 ð1 þ ecf Þ
ð62Þ
þ2a1 Þe þ ð2a2 þ 4b1 Þe3 
h i
þ ðb1 þ a2 Þeð1  c2f Þ  ð6b1 þ 2a2 þ 2a1 Þðcf þ 1Þ
   5.4. The Relation Between the Osculating and Non-
1  cf
osculating Elements
ln ð2a2 þ 8b1 þ 2a1 Þe2 þ 2a0 : ð58Þ
sf
An important aspect of the ongoing discussion is finding
a relationship between the osculating elements and the
The resulting eccentricity satisfies e ¼ e0 ; the remaining ele- non-osculating elements. It is important to keep in mind
ments can be calculated using the GVEs (34). that the non-osculating elements do not possess the same
Similarly to the previous procedure, a simple approxi- physical meaning of their osculating counterparts. Thus,
mation for UR (58), linear in the eccentricity, can be for analytical studies as well as for numerical exploitation
obtained for near-circular motion in the restricted relativis- of the new parameterisations offered above, expressions
tic two-body problem: relating the osculating and non-osculating elements must

3
 be provided. These relations will be later used for a numer-
sf GM 2 1  cf ical illustration of the new formalism.
UR  2 3 ln ðecf  1Þ þ 10ðec2f  1Þ
c a sf To begin, let us denote an osculating element by C H i . For
 
2sf the purpose of this discussion, it would be convenient to use
þ13e ln : ð59Þ the true anomaly as one of the orbital element; thus, a set of
cf ð1 þ cf Þ
osculating elements that uniquely determine the position
and velocity in inertial space are
5.3. Constant Argument of Periastron CH ¼ faH ; eH ; iH ; XH ; xH ; f H g, and the corresponding set
of non-osculating elements is C ¼ fa; e; i; X; x; f g. Because
The secular growth of x, which is linearly proportional the position remains invariant to a selection of a gauge,
to f using the osculating elements (this is consistent in all
the representations listed by Klioner and Kopeikin rðCH Þ ¼ rðCÞ : ð63Þ
(1994)), can be nullified using non-osculating elements The relationship between the velocities is given by
and a particular gauge choice. Requiring that x_ ¼ 0 and
choosing US  0 yields the following differential equation: gðCH Þ ¼ gðCÞ þ UðCÞ : ð64Þ

cos f ð1 þ e cos f Þ2 0 sin f ð1 þ e cos f Þ Equations (63) and (64) define six algebraic equations, that
UR ðf Þ þ UR ðf Þ can be solved for CH .
pe pe
For the post-Newtonian perturbation, the motion is two
G2 M 2 ð1 þ e cos f Þ2
¼ dimensional, since the node and inclination remain unaf-
hp2 c2 fected by the presence of the relativistic correction. Thus,
 
ða0 þ a1 e cos f þ a2 e2 cos2 f Þ cos f without loss of generality, we can choose
  þ ð2 þ e cos f Þb1 sin2 f
e i ¼ iH ¼ X ¼ XH  0. In this case, the position vector
ð60Þ assumes the form
að1  e2 Þ T
rðCÞ ¼ ½cosðf þ xÞ; sinðf þ xÞ; 0 ð65Þ
1 þ e cos f
Eq. (60) can be solved for UR ðf Þ, yielding
and the Keplerian velocity is
1 G2 M 2  sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
UR ðf Þ ¼ ð6b1 e2 cf þ 2a0 cf Þ lnðsec f þ tan f Þ
2 hc pð1 þ ecf Þ
2 GM
gðCÞ ¼ ½e sin x  sinðf þ xÞ; e cos x
þ½ða2  b1 Þe2 þ ð2a1 þ 4b1 þ 2a0 Þecf f að1  e2 Þ
o
þsf ½ðb1 þ a2 Þe3 c2f þ ð2a1 þ 2a2 þ 6b1 Þe2 cf  4b1 e þ cosðf þ xÞ; 0T : ð66Þ
j2 cf The algebraic equations ensuing from (63) and (64) are
þ : ð61Þ
ð1 þ ecf Þ quite complex and require numerics.
546
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

6. Numerical Example The problem parameters are

Consider, for illustration purposes, the relativistic cor- GM ¼ 1:327120220308192  1011 km3 = sec2 ; m
rection of Mercury’s orbit about the Sun. In osculating ele- ¼ 1:660046706402425  107 : ð68Þ
ments, using the subscript 0 to denote initial values, the
orbit is given by the initial conditions (in a J2000 system The non-osculating initial elements, obtained by solving
of reference) Eqs. (63) and (64), are
a0 ¼ 57910001:278597308147 km; e0 ¼ 0:20560001753959287403
aH H H
0 ¼ 57910000 km; e0 ¼ 0:2056; x0
x0 ¼ 1:351869887693436 rad; f 0 ¼ 0:19171292543848736670  106 rad :
¼ 1:351870079406362 rad; f H
0 ¼ 0 : ð67Þ ð69Þ

Fig. 1. A comparison between the osculating and non-osculating elements. The non-osculating semimajor axis remains fixed due to the effect of the gauge
velocity. The drift of the argument of periastron is much smaller.

Fig. 2. The time history of the gauge velocity. This velocity is about 7 orders of magnitude smaller than the nominal orbital velocity.
547
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

Equations (34), (35) can be integrated with the initial con- Declaration of Competing Interest
ditions (68), to yield the time history of the osculating ele-
ments, and initial conditions (69), with a given gauge The authors declare that they have no known competing
function, to yield the time history of the non-osculating financial interests or personal relationships that could have
elements. appeared to influence the work reported in this paper.
We shall provide a comparison of the time histories of
the osculating and non-osculating elements for the gauge Acknowledgments
(54), yielding a constant semimajor axis, with j1 arbitrarily
chosen as zero. This comparison is depicted in Fig. 1, The authors are grateful to Joseph O’Leary for a stimu-
showing the osculating and non-osculating semimajor axis, lating conversation, which moved the authors to complete
eccentricity, argument of periastron and true anomaly. The an unfinished draft . One of the authors (ME) would like to
initial values of the semimajor axis and eccentricity were thank Sergei M. Kopeikin for a useful consultation on the
subtracted from the time histories of these elements to PN formalism.
accentuate the post-Newtonian effect. As expected, the
non-osculating semimajor axis remains constant, while Appendix A. Transition from Cartesian to orbital
the osculating semimajor axis oscillates, reaching a maxi- coordinates
mum of about 9.5 km. The eccentricity in both cases
behaves similarly. The argument of periastron drifts lin- Referring the reader for detailed treatment to the preprint
early (on average) in both cases, although the rotation of by Efroimsky (2002) and review (Efroimsky, 2005), we pro-
the non-osculating apsidal line is retrograde. The differ- vide here a squeezed explanation of why the switch from
ences between the anomalies are two small to show on Cartesian to orbital coordinates produces internal freedom.
the given scale. The reduced two-body problem,
Fig. 2 shows the time history of the gauge function,
Gðm1 þ m2 Þ r
which is zero in the osculating case. In the non- €r þ ¼ 0 ; r  r2  r1 ;
osculating case, the gauge velocity oscillates between the r2 r
values of 60 km/yr. This value is about 7 orders of mag- has a generic solution, a Keplerian conic, which can be
nitude smaller than the nominal orbital velocity. expressed in some fixed Cartesian frame as
x ¼ f 1 ðC 1 ; . . . ; C 6 ; t Þ ; x_ ¼ g1 ðC 1 ; . . . ; C 6 ; tÞ ;
7. Conclusions
y ¼ f 2 ðC 1 ; . . . ; C 6 ; t Þ ; y_ ¼ g2 ðC 1 ; . . . ; C 6 ; tÞ ; ð71Þ
Utilising the generalised Lagrange constraint for deriv- z ¼ f 3 ðC 1 ; . . . ; C 6 ; t Þ ; z_ ¼ g3 ðC 1 ; . . . ; C 6 ; tÞ ;
ing the Gauss variational equations introduces gauge free-
or, shortly:
dom, which can be employed to write down new solutions
for orbital motion affected by a post-Netwonian perturba- r ¼ f ðC 1 ; . . . ; C 6 ; t Þ ; r_
tion. The gauge velocity required to derive the new solu- ¼ gð C 1 ; . . . ; C 6 ; t Þ : ð72Þ
tions is a few orders of magnitude smaller than the
nominal orbital velocity. While the classical PPN solutions Here C i are six adjustable constants, while gi are the partial
entail variation of all six classical osculating orbital ele- derivatives of f i over the last argument:
ments, the transformation to non-osculating elements por-

@f
trays a different picture, wherein some of the elements can gð C 1 ; . . . ; C 6 ; t Þ  : ð73Þ
@t C¼const
be kept constant. This is achieved by relaxing the tradi-
tional requirement to represent a perturbed two-body orbit Now consider a perturbed problem, where the disturbing
by a series of Keplerian conics instantaneously osculating force DF is an arbitrary vector-valued function of the posi-
the perturbed trajectory. As opposed to the known solu- tion and velocity:
tions to the post-Newtonian perturbation, in which solu- l r
tions the mean eccentricity and mean semimajor axis €r þ 2 ¼ F : ð74Þ
r r
remain constant, and the argument of periastron exhibits
secular growth, – the solutions developed in this paper pro- Following Lagrange (1808a,b, 1809), we utilise solution
vide exact nullification of the time-varying eccentricity, or (71)–(73) as an ansatz for a solution to the perturbed prob-
the semimajor axis, or the argument of periastron, without lem (74), the ‘‘constants” now being time dependent:
requiring averaging, and without utilising a canonical r ¼ f ðC 1 ðtÞ; . . . ; C 6 ðtÞ; tÞ ; ð75Þ
transformation to switch between osculating and mean
elements. and the functional form of f being the same as in (72). The
velocity

548
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

dr @f X @f dC i X @f dC i Keplerian ellipses. Mappings between different representa-

¼ þ ¼ g þ ð76Þ
dt @t i
@C i dt i
@C i dt tions reveal an internal symmetry (and a symmetry group)
P underlying this formalism.
now contains a ‘‘convective” term ð@f=@C i ÞðdC i =dtÞ, Substitution of (77) with (80) leaves the physical motion
while the first term, g, has the same functional form as unchanged, but alters its mathematical description. Specif-
in (73). ically, it entails different solutions for the orbital elements.
The insertion of f ðC 1 ðtÞ; . . . ; C 6 ðtÞ; tÞ into the perturbed This invariance of a physical theory under a change of
equation of motion (74) renders three scalar differential parametrisation goes under the names of gauge symmetry.
equations of the second order. They contain an indepen- The gauge transformations make a group, which is isomor-
dent parameter, time, and six time-dependent quantities phic to the group of all real-valued functions on R3 , with
C i ðtÞ whose evolution is to be found. This cannot be car- the group operation being addition (Gurfil, 2007).
ried out in a unique way because the number of variables Just like in electrodynamics, a right choice of gauge
exceeds, by three, the number of equations. So, while the often simplifies the solution of the equations of motion.
actual physical orbit (comprising the locus of points in A deliberate choice of non-osculating orbital elements —
the Cartesian coordinates and the values of velocity in each i.e., of a set C i obeying some condition (80) different from
point) is unique, its parametrisation through the orbital (77) — can simplify the expressions for these elements’
variables C i ðtÞ is ambiguous. Lagrange chose to remove rates.
this freedom by setting three independent constraints on
C_ i ðtÞ: Appendix B. Derivation of the Generalised Chazy–Brumberg
X @f dC i Force
¼ 0 ; ð77Þ
i
@C i dt
Through the medium of the formulae
These constraints nullify the ‘‘convective” term in expres-
@ 1 r @
sion (76) and thereby make the velocity equal to the Kep- ¼  n nþ2 and ðr  r_ Þ 2
lerian velocity g. Stated alternatively, these constraints @r r n r @r
make the instantaneous conic tangent to the physical orbit. ¼ 2 ðr  r_ Þ r_ ; ð81Þ
Within the unperturbed two-body problem (70), its solu-
we derive from (5) the following:
tion (72) defines a time-dependent one-to-one (within one " #
orbital period) mapping @DL 1 B2 r_ 2 ðr  r_ Þ 2 r_
¼ 2  2 4 r  B3 3 r  3 B4 r þ 2 B4 ðr  _
r Þ : ð82Þ
@r c r r r5 r3
ð C 1 ; . . . ; C 6 Þ $ ð xðtÞ; yðtÞ; zðtÞ; x_ ðtÞ; yðtÞ;
_ z_ ðtÞ Þ :
Also, using
ð78Þ
2
@ r_ 2 @ ðr_ 2 Þ
Under disturbance, ansatz (75) entails the emergence of the ¼ 2 r_ ;
@ r_ @ r_
time derivatives C_ i in (76). So, instead of (78), we obtain a
@ðr  r_ Þ
2
time-dependent mapping between a 12-dimensional and a ¼ 4 ð r_ Þ 2 r_ ; and ¼ 2 ðr  r_ Þ r ; ð83Þ
6-dimensional spaces: @ r_
 
C 1 ðtÞ; . . . ; C 6 ðtÞ; C_ 1 ðtÞ; . . . ; C_ 6 ðtÞ ! we deduce from (5) that
ð79Þ  
ð xðtÞ; yðtÞ; zðtÞ; x_ ðtÞ; y_ ðtÞ; z_ ðtÞ Þ : @DL 1 2 B3 2 B4
_ 2
_
¼ 2 4 B1 ð r Þ r þ _ _
r þ 3 ðr  rÞ r : ð84Þ
@ r_ c r r
This mapping cannot be one-to-one. Its ambiguity is
another expression of the obvious fact that the three equa- The subsequent calculations, though elementary, need care.
tions of motion (74) are insufficient to determine the six For r_ @ 2 DL=@r @ r_ , we have:
functions C 1 ; . . . ; C 6 , for which reason one can impose  
@ 2 DL 1 X i @ 2 B3 2 B4
three arbitrary constraints on these functions and their r_ ¼ 2 r_ 4 B1 ð r_ Þ 2 r_ þ r_ þ 3 ðr  r_ Þ r ¼
@r @ r_ c i @ri r r
derivatives. 5 Together with the three equations of motion, 

 
1 X r_ i ri r_ i ri 2 B4 2 B4
the three constraints remove the freedom and make map- 2 B3  3 r_ þ 2 B4  3 5 ð r  r_ Þ r þ 3 ð r_ i r_ i Þ r þ 3 r_ i ð r_  r Þ ^ei
c2 i r r r r
ping (79) unambiguous.
From the mathematical point of view, the Lagrange " #
constraint (77) is completely arbitrary. We could as well 1 r  r_ ð r  r_ Þ2 r_ 2 2 B4
¼  2 B3 r_  6 B4 r þ 2 B4 3 r þ 3 ð r  r_ Þ r_ : ð85Þ
c2 r3 r5 r r
have chosen some different supplementary condition
X @f dC i By similar means, for €r @ 2 DL=@ r_ @ r_ we obtain:
¼ UðC 1;...; 6 ; tÞ ; ð80Þ  
i
@C i dt @ 2 DL 1 X i @ 2B3 2B4
€r ¼ 2 €r 4B1 ð r_ Þ r_ þ
2
r_ þ 3 ðr  r_ Þr
with U an arbitrary function of time and the parameters @ r_ @ r_ c i @ r_ i r r
C i . 6 The arbitrariness of these conditions reveals the ambi- ¼
guity of the representation of an orbit by instantaneous
549
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553
  "
1 X i i 2 B3 i 2 B1 i r ð r  r_ Þ2 r_ 2 GM 2 B4
€
r 4 B 1 2 _
r i
_
r þ €
r i
4 B 1 ð _
r Þ
2
^
e i þ €
r ^
e i þ €
r r_ r  2  6 B4 þ 2 B  3 ðr  rÞ 3
c2 i r r3 c r 5 4
r 3 r r


  GM 2 B3
1 2 B3 2 B4  3 4 B1 r_ 2 þ
8 B 1 ð €
r  _
r Þ _
r þ 4 B 1 €
r ð _
r Þ
2
þ €
r þ ð €
r  r Þ r : ð86Þ r r
c2 r r3
1 2
Together, (85) and (86) yield: ¼ 2 r ½ GM ðB3 þ B4 Þ  B2 
c r4
ð r_ Þ
2
@ 2 DL @ 2 DL 3 B4
 r_  €r ¼ þ ½  B3 þ 4 GM B1  2 B4  r þ 5 ðr  r_ Þ2 r
@r "@ r_ @ r_ @ r_ r3 r
1 r  r_ ð r  r_ Þ
2
r_ 2 ðr  r_ Þ r_
 2  2 B3 r_  6 B r þ 2 B r þ ½ 8 B1 G M þ 2 B3  ð90Þ
c r3
4
r5
4
r3 r3
 A further substitution of (8) results in
2 B4
þ 3 ð r  r_ Þ r_ h 2
i
r F ¼ 1
c2
2r
r4
ðGMÞ2 r þ GM ð  2  Þ ð r_r3Þ r þ 3 GM r5
a
ðr  r_ Þ2 r þ 2 l ðr_rr3Þ r_
h 2
 i ð91Þ
¼ GM
c2
2 GM
r
r  2  ð r_ Þ 2 þ 3 a ðr_rr2Þ rr3 þ 2 l ðr_rr3Þ r_ :
 
1 2 B3 2B
 2 8 B1 ð €r  r_ Þ r_ þ 4 B1 €r ð r_ Þ þ
2
€r þ ð €r  r Þ 3 4 r
c r r Appendix C. Calculation of ›C=›r

In this Appendix, all orbital elements are non-

  osculating. The known Keplerian relations are evaluated
r_ r  r_ 2 B4
¼  8 B1 ð €r  r_ Þ  2 B3 þ 3 ð r  r_ Þ so that the Keplerian velocity (equation (93) below) is a
c2 r3 r
function of the non-osculating elements. In accordance
" #
r ð r  r_ Þ
2
r_ 2 2 B4 with equation (20), we set
  6 B4 þ 2 B4 3 þ ð €r  r Þ 3 r  fðt; C 1 ; . . . ; C 6 Þ : ð92Þ
c2 r5 r r
 
€r 2 B3 while by v we understand the Keplerian velocity along an
4 B1 ð r_ Þ þ
2
 : ð87Þ instantaneous (non-osculating) Keplerian ellipse, i.e. the
c2 r
same as gðt; C 1 ; . . . ; C 6 Þ in equations (21,22):
Making use of v  gðt; C 1 ; . . . ; C 6 Þ : ð93Þ
GM GM 1 GM
€r ¼  3 r þ 3 2 ½. . .    3 r ; ð88Þ Under this premise, we can use all known expressions for
r r c r Keplerian conics. E.g., to calculate @a=@r, we write the
we then arrive at vis-viva equation
v2 GM GM
 ¼ : ð94Þ
2 r 2a
@ DL 2
@ DL 2
 r_  €r Differentiation of this expression with respect to r yields
@r @ r_ @ r_ @ r_
" @a r
ð r  r_ Þ
2
r_ 2 ¼ 2a2 3 ; ð95Þ
r
 2  6 B4 þ 2 B
GM 2 B4
 3 ðr  rÞ 3 @r r
4
c r5 r3 r r which in the polar representation reads as

  2 
GM 2 B3 @a 2a
 3 4 B1 r_ 2 þ : ð89Þ ¼ ; 0; 0 : ð96Þ
r r @r r2
Combined, formulae (3), (82), and (89) entail: The next step is to evaluate the partial derivative of h, the
magnitude of the Keplerian angular momentum vector h.
@DL @ 2 DL @ 2 DL
F ¼  r_  €r  To that end, we use the definition
@r @r @ r_ @ 2 r_
" # h ¼ r  v ¼ S v r ; ð97Þ
1 B2 r_ ðr  r_ Þ
2 2
r_ where S v is the matrix cross-product equivalent,
 2 4 r  B 3 3 r  3 B4 r þ 2 B4 ð r  r_ Þ
c2 r r r5 r3 2 3
0 0 r @f
@t
6 0  @r 7
Sv ¼ 4 0 @t 5: ð98Þ
  r @f @r
0
r_ GM r  r_ r  r_ @t @t
  8 B1 ð r  r_ Þ  2 B3 3 þ 2 B4 3
c2 r3 r r Using h2 ¼ hT h, we deduce that
550
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

@h @h the preceding procedure; or, using the known results for

2h ¼ 2hT ¼ 2hT S v ; ð99Þ
@r @r the derivatives of the elements with respect to v and their
wherefrom Poisson brackets, ðC i ; C j Þ, solve for the remaining
  unknown derivatives of the elements with respect to r.
@h ^ v ¼ r @f ;  @r ; 0 :
¼ ^
hS v ¼ WS ð100Þ We shall adopt the second option. Thus, to obtain e.g.
@r @t @t @X=@r, one can solve the equations
We now proceed to obtain @e=@r. This is performed by dif-
T
T
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi @i @X @i @X
ferentiating both sides of the relation h ¼ GMað1  e2 Þ:  ¼ði; XÞ : ð110aÞ
@r @v @v @r
 
T
T
@h GM 2 @a @e @a @X @a @X
¼ ð1  e Þ  2ae : ð101Þ  ¼ða; XÞ : ð110bÞ
@r 2h @r @r @r @v @v @r

T
T
At this point, it is useful to introduce the Keplerian @e @X @e @X
formulae  ¼ðe; XÞ : ð110cÞ
@r @v @v @r
p
r¼ ð102Þ where (Battin, 1999)
1 þ e cos f
1
and ða; XÞ ¼ ðe; XÞ ¼ 0; ði; XÞ ¼ ð111Þ
sffiffiffiffiffiffiffiffi h sin i
@f h GM 2 and
¼ 2¼ ð1 þ e cos f Þ ð103Þ
@t r p3 @a 2

¼ 2ah e sin f ; pr ; 0 ; @e ¼ 1h ½p sin f ; ðp þ rÞ cosðf Þ þ re; 0

@v @v
@i
ð112Þ
¼ 1h ½0; 0; r cosðf þ xÞ :
where p ¼ að1  e2 Þ ¼ h2 =ðGMÞ is the semilatus rectum.
@v

Equation (103) can be used to transform the partial time

derivatives to partial derivatives with respect to the true Solving equations (110a) for @X=@r, we obtain:
anomaly, a procedure essential for facilitating the underly-  
ing algebra. We have: @X ½sinðf þ xÞ þ e sin xr sinðf þ xÞ  p
¼ 0; 0; :
@r @r @f eh sin f @r rp sin i cosðf þ xÞ
¼ ¼ : ð104Þ
@t @f @t rð1 þ e cos f Þ Similarly, the expression for @x=@r can be derived by solv-
The employment of equation (96), (100), and (104) fur- ing the equations

T
T
nishes us with @X @x @X @x
   ¼ðX; xÞ ; ð114aÞ
@e ðcos f þ eÞð1 þ e cos f Þ sin f @r @v @v @r
¼ ; ;0 : ð105Þ
T
T
@r p a @a @x @a @x
 ¼ða; xÞ ; ð114bÞ
A similar procedure may be exploited to find @i=@r. We @r @v @v @r

T
T
start out with the definition @e @x @e @x
 ¼ðe; xÞ ; ð114cÞ
^
hT k @r @v @v @r
cos i ¼ ; ð106Þ
h where (Battin, 1999)
^ is a unit vector normal to the fundamental plane of
where k h
an inertial reference system, whose components in the ða; xÞ ¼ ðX; xÞ ¼ 0; ðe; xÞ ¼ ð115Þ
aeGM
RSW frame are given by (Battin, 1999)
and
^ ¼ ½sin i sinðf þ xÞ; sin i cosðf þ xÞ; cos iT :
k ð107Þ  
@X r sinðf þ xÞ
¼ 0; 0;
Differentiation of equation (106) with respect to r entails @v h sin i

  
@i 1 @h ^ @x 1 p cos f ðp þ rÞ sin f r sinðf þ xÞ cos i
sin i ¼ cos i  S v k : ð108Þ ¼  ; ; :
@r h @r @v h e e sin i
^ given by equa-
By using the expressions for S v ; @h=@r and k ð116Þ
tions (98), (100) and (107), respectively, while keeping in
mind Eqs. (103) and (104), one arrives at
  This procedure provides the expression
@i e sinðf þ xÞ sin x 2 3
¼ 0; 0; : ð109Þ sin f ðe cos f þ1Þ
@r p
T 6
@x
pe
7
6  peeþcos f 7
At this point, one may proceed in one of the following ¼6 7 : ð117Þ
@r 4 2 5
ways: using the definitions of the remaining orbital ele- ½cos ðf þxÞsinðf þxÞe sinðxÞþe cos f  cot i
ments, take the partial derivatives with respect to r as in p cosðf þxÞ

551
P. Gurfil, M. Efroimsky Advances in Space Research 69 (2022) 538–553

The expression for @l0 =@r can be derived by solving the Battin, R. H., 1999. An Introduction to the Mathematics and Methods of
equations Astrodynamics. AIAA, Reston, VA.
Beltrán Jiménez, J., Heisenberg, L., Koivisto, T.S., 2019. The geometrical

T
T
@X @l0 @X @l0 trinity of gravity. Universe 5 (7), URL: https://www.mdpi.com/2218-
 ¼ðX; l0 Þ ; ð118aÞ 1997/5/7/173.
@r @v @v @r Blanchet, L., Apr. 2003. Post-Newtonian Theory and its Application.

T
T
@a @l0 @a @l0 arXiv e-prints, gr–qc/0304014.
 ¼ða; l0 Þ ; ð118bÞ Blanchet, L., 2014. Gravitational Radiation from Post-Newtonian Sources
@r @v @v @r

T
T and Inspiralling Compact Binaries. Living Rev. Relativ., URL:
@e @l0 @e @l0 https://doiorg/10.12942/lrr-2014-2.
 ¼ðe; l0 Þ ; ð118cÞ Brumberg, V.A., 1991. Essential Relativistic Celestial Mechanics. Taylor
@r @v @v @r
and Francis, Florida, US.
where (Battin, 1999) Brumberg, V. A., Kopejkin, S. M., 1989. Relativistic Theory of Celestial
Reference Frames. In: Reference Frames. J. Kovalevsky, I.I. Mueller,
2 1 þ e2 B. Kolaczek (Editors). Kluwer, Dordrecht 1989. Part of the Astro-
ðX; l0 Þ ¼ 0; ða; l0 Þ ¼ ; ðe; l0 Þ ¼ pffiffiffiffiffiffiffiffiffiffiffi ð119Þ physics and Space Science Library book series (ASSL), Volume 154,
na e aGM
115–141. URL: doi: 10.1007/978-94-009-0933-5_6.
and Cervantes-Cota, J.L., Galindo-Uribarri, S., Smoot, G.F., 2016. A brief
2 3 history of gravitational waves. Universe 2 (3), URL: https://www.
ð2 eþcos f þe cos2 f Þð1e2 Þ mdpi.com/2218-1997/2/3/22.

T 6 eð1þe cos f Þna 7
@l0 6 7 Chazy, J., 1928. La théorie de la relativité et la mécanique céleste. Paris,
¼6 ðe2 1Þð2þe cos f Þ sin f 7 : ð120Þ Gauthier-Villars et cie, 1928-.
@v 4 eð1þe cos f Þna 5 Damour, T., 1983. Gravitational radiation reaction in the binary pulsar
0 and the quadrupole-formula controversy. Phys. Rev. Lett. 51, 1019–
1021, URL: https://link.aps.org/doi/10.1103/PhysRevLett.51.1019.
This results in Damour, T., 1987. The problem of motion in Newtonian and Einsteinian
2 3e2 þ2 cos2 f ð1þe2 Þ2 e cos f sin2 f 3 gravity. pp. 128–198.
pffiffiffiffiffiffiffi

T Damour, T., Deruelle, N., 1985. General relativistic celestial mechanics of
@l0 6 2 a e 1e2 sin f
pffiffiffiffiffiffiffi 7
¼6 4 1e2 cos f 7 :
5 ð121Þ binary systems. I. The post-Newtonian motion. Annales de l’institut
@r ae Henri Poincaré (A) Physique théorique 43 (1), 107–132.
Damour, T., Soffel, M., Xu, C., 1991. General-relativistic celestial
0
mechanics. I. Method and definition of reference systems. Physical
Finally, we shall derive the gauge-generalised variation of Review D 43, 3273–3307, URL: https://link.aps.org/doi/10.1103/
the true anomaly. The relation PhysRevD.43.3273.
Debono, I., Smoot, G.F., 2016. General relativity and cosmology:
h2 Unsolved questions and future directions. Universe 2 (4), URL:
rð1 þ e cos f Þ ¼ ð122Þ https://www.mdpi.com/2218-1997/2/4/23.
GM Dosopoulou, F., Kalogera, V., 2016. Orbital Evolution of Mass-transfer-
entails, upon differentiation with respect to r, ring Eccentric Binary Systems. I. Phase-dependent Evolution. Astro-

  phys J 825, 70.
@f 1 1 2h @h @r @e
¼  ð1 þ e cos f Þ  cos f : Efroimsky, M., 2002. Equations for the Orbital Elements. Hidden
@r e sin f r GM @r @r @r Symmetry. Preprint No 1844 of the Institute of Mathematics and its
Applications, University of Minnesota https://www.ima.umn.edu/
ð123Þ sites/default/files/1844.pdf.
Utilising the identity Efroimsky, M., 2005. Gauge Freedom in Orbital Mechanics. Ann. N. Y.
Acad. Sci. 1065 (1), 346.
@r Efroimsky, M., Goldreich, P., 2003. Gauge symmetry of the n-body
¼ ½1; 0; 0 ð124Þ
@r problem in the hamilton-jacobi approach. Journal of Mathematical
Physics 44, 5958–5977.
and plugging in equations (101), (105), we find that Efroimsky, M., Goldreich, P., 2004. Gauge freedom in the n-body
2 3 problem of celestial mechanics. Astronomy & Astrophysics 415, 1187–

T  sin f ð1 þ e cos f Þ
@f sin f 6 2 7 1199.
¼ 4 ðe cos f þ 2e þ cos f Þ 5 : ð125Þ Grishchuk, L.P., Kopeikin, S.M., 1983. The Motion of a Pair of
@r ep cos f Gravitating Bodies Including the Radiation Reaction Force. Soviet
0 Astronomy Letters 9, 230–232.
Gurfil, P., 2007. Rescriptive and Descriptive Gauge Symmetry in Finite-
References Dimensional Dynamical Systems. AIP Conf. Proc. 886, 42.
Gurfil, P., Seidelmann, P.K., 2016. Celestial Mechanics and Astrodynam-
Abbott, R. et al., 2021. Observation of gravitational waves from two ics: Theory and Practice. Springer-Verlag.
neutron star–black hole coalescences. The Astrophysical Journal Iorio, L., 2015. Editorial for the Special Issue 100 Years of Chrono-
Letters 915, L5. https://doi.org/10.3847/2041-8213/ac082e. geometrodynamics: The Status of the Einstein’s Theory of Gravitation
Asada, H., Futamase, T., 1997. Post-Newtonian approximation: Its in Its Centennial Year. Universe 1 (1), 38–81.
Foundation and applications. Prog. Theor. Phys. Suppl. 128, 123–181. Iorio, L., 2015. Post-Newtonian direct and mixed orbital effects due to the
Ashtekar, A., 2005. 100 Years of Relativity: Space-Time Structure – oblateness of the central body. International Journal of Modern
Einstein and Beyond. World Scientific. Physics D 24 (8), 1550067–59.
Bailes, M., Berger, B.K., Brady, P.R., 2021. Gravitational-wave physics Iorio, L., 2019. A Post-Newtonian Gravitomagnetic Effect on the Orbital
and astronomy in the 2020s and 2030s. Nature Rev. Phys. 3 (5), 344– Motion of a Test Particle around Its Primary Induced by the Spin of a
366. Distant Third Body. Universe 5. URL: doi: 10.3390/universe5040087.

552
P. Gurfil, M. Efroimsky
Iorio, L., 2019. A hero for general relativity. Universe 5 (7), URL: https://
www.mdpi.com/2218-1997/5/7/165.
Iorio, L., 2021. On the 2PN Pericentre Precession in the General Theory of
Relativity and the Recently Discovered Fast-Orbiting S-Stars in Sgr
A*. Universe 7. URL: doi: 10.3390/universe7020037.
Klioner, S.A., Kopeikin, S.M., 1994. The post-Keplerian orbital repre-
sentations of the relativistic two-body problem. Astrophys J 427, 951–
955.
Kopeikin, S.M., 1985. General Relativistic Equations of Binary Motion
for Extended Bodies with Conservative Corrections and Radiation
Damping. Soviet Astronomy 29, 516–524.
Kopeikin, S.M., Efroimsky, M., Kaplan, G., 2011. Relativistic Celestial
Mechanics of the Solar System. Wiley-VCH.
Kopeikin, S. M., Vlasov, I. Y., 2006. The Effacing Principle in the Post-
Newtonian Celestial Mechanics. In: Proceedings of the 11th Marcel
Grassmann Meeting. Berlin, 23 - 30 July 2006, pages 2475 - 2477
https://arxiv.org/abs/gr-qc/0612017. URL: https://doi.org/10.1142/
9789812834300_0437
Kopejkin, S.M., 1988. Celestial coordinate reference systems in curved
space-time. Celestial Mechanics and Dynamical Astronomy 44, 87–
115.
Lagrange, J.-L., 1808a. Mémoire sur la théorie des variations des éléments
des planétes, et en particulier des variations des grands axes de leurs
orbites. Mémoires de la premiére classe de l’Institut de France, année
1808Lu, le 22 août 1808 à l’Institut de France. Later edition: Œuvres
de Lagrange. Tome VI, pp. 713–768. Paris: Gauthier-Villars, 1873.
URL: http://math-doc.ujf-grenoble.fr/cgi-bin/oetoc?
id=OE_LAGRANGE__6.
Lagrange, J.-L., 1808b. Mémoire sur la théorie générale de la variation des
constantes arbitraires dans tous les problèmes de la mécanique.
Mémoires de la premiére classe de l’Institut de France, année 1808Lu,
le 13 mars 1809 à l’Institut de France. Later edition: Œuvres de
Lagrange. Tome VI, pp. 771–805. Paris: Gauthier-Villars, 1873. URL:
http://math-doc.ujf-grenoble.fr/cgi-bin/oetoc?
id=OE_LAGRANGE__6.
Lagrange, J.-L., 1809. Second mémoire sur la théorie générale de la
variation des constantes arbitraires dans tous les problémes de la
mécanique, dans lequel on simplifie l’application des formules
générales á ces problémes. Mémoires de la premiére classe de l’Institut
553
Advances in Space Research 69 (2022) 538–553
de France, année 1809Lu, le 19 février 1810 à l’Institut de France.
Later edition: Œuvres de Lagrange. Tome VI, pp. 809–816. Paris:
Gauthier-Villars, 1873. URL: http://math-doc.ujf-grenoble.fr/cgi-bin/
oetoc?id=OE_LAGRANGE__6.
Meichsner, J., Soffel, M.H., 2015. Effects on satellite orbits in the
gravitational field of an axisymmetric central body with a mass
monopole and arbitrary spin multipole moments. Celestial Mech. Dyn.
Astron. 123, 1.
Miller, M.C., Yunes, N., 2019. The new frontier of gravitational waves.
Nature 568 (7753), 469–476.
Panhans, M., Soffel, M.H., Dec. 2014. Gravito-magnetism of an extended
celestial body. Class. Quantum Gravity 31 (24), 245012.
Possenti, A., Burgay, M., D’Amico, N., Lyne, A.G., Kramer, M.,
Manchester, R.N., Camilo, F., McLaughlin, M.A., Lorimer, D., Joshi,
B.C., Sarkissian, J.M., Freire, P.C.C., 2004. The double-pulsar PSR
J0737–3039A/B. Memorie della Societa Astronomica Italiana Supple-
menti 5, 142.
Rovelli, C. (Ed.), 2015. General Relativity: The most beautiful of theories:
Applications and trends after 100 years. De Gruyter. URL: https://doi.
org/10.1515/9783110343304
Soffel, M., Wirrer, R., Schastok, J., Ruder, H., Schneider, M., Jan. 1988.
Relativistic Effects in the Motion of Artificial Satellites - the Oblateness
of the Central Body I. Celestial Mechanics 42 (1–4), 81–89.
Taylor, J.H., 1992. Pulsar timing and relativistic gravity. Philosophical
Transactions of the Royal Society of London. Series A: Physical and
Engineering Sciences 341, 117–134. https://doi.org/10.1098/
rsta.1992.0088.
Taylor, J.H., 1993. Pulsar timing and relativistic gravity. Class. Quantum
Gravity 10, 167–174, URL: https://iopscience.iop.org/article/10.1088/
0264-9381/10/S/017.
Vishwakarma, R.G., 2016. Einstein and beyond: A critical perspective on
general relativity. Universe 2 (2), URL: https://www.mdpi.com/2218-
1997/2/2/11.
Will, C.M., 2011. On the unreasonable effectiveness of the post-newtonian
approximation in gravitational physics. Proc. Nat. Acad. Sci. 108 (15),
5938–5945, URL: https://www.pnas.org/content/108/15/5938.
Will, C.M., 2018. Theory and Experiment in Gravitational Physics.
Cambridge University Press.