A&A 548, A81 (2012) Astronomy

DOI: 10.1051/0004-6361/201220198 &


c ESO 2012 Astrophysics

CONGO, model of cometary non-gravitational forces combining

astrometric and production rate data
Application to comet 19P/Borrelly
L. Maquet1 , F. Colas1 , L. Jorda2 , and J. Crovisier3

1
IMCCE, Observatoire de Paris, CNRS, UPMC, Université Lille 1, 77 avenue Denfert-Rochereau, 75014 Paris, France
e-mail: lucie.maquet@imcce.fr
2
LAM – Laboratoire d’Astrophysique de Marseille, Pôle de l’Étoile, Site de Château-Gombert, 38 rue Frédéric Joliot-Curie,
13388 Marseille Cedex 13, France,
3
LESIA – Observatoire de Paris, CNRS, UPMC, Université Paris-Diderot, 5 place Jules Janssen, 92195 Meudon, France
Received 10 August 2012 / Accepted 12 October 2012
ABSTRACT

The gravitational orbit of a comet is affected by the sublimation of water molecules by the nucleus when the comet approaches perihe-
lion. This outgassing triggers a non-gravitational force (NGF) that significantly modifies the orbit of the comet. Up to now, modelling
of this effect is mostly based on an empirical model defined in the early 70s that uses a simplified outgassing model. Attempts have
been made to use advanced anisotropic thermal models both to calculate the NGF taking several observational constraints into account
and to retrieve the nucleus’s mass and density, but (i) this approach is restricted to a handful of cometary nuclei that are sufficiently
well known to allow this type of modelling, and (ii) the authors usually do not fit the astrometric measurements directly but rather
non-gravitational parameters calculated with the above-mentioned empirical model. We present a new model for non-gravitational
forces with the aim of revisiting the problem of NGF calculation and nucleus density determination. Our model is closer to the nucleus
outgassing physics with only a few free parameters. The amplitude of the perturbation depends on several parameters describing the
comet activity that can be constrained by visible, infrared, and radio observations of the coma and the nucleus of the comet. It also
depends on the nucleus mass, which can in turn be determined by modelling the effect of the NGF on the orbit of a comet. The method
is based on the decomposition of the surface of the nucleus in several elements located at different latitudes. The contribution of each
surface element to the overall NGF is fitted from the astrometric measurements, together with the density of the nucleus. This method
is the only one available so far to estimate the density of cometary nuclei from ground-based observations. This method is tested on
the well-known comet 19P/Borrelly. The density found for these comet is between 150 and 600 kg m−3 .
Key words. celestial mechanics – comets: individual: 19P/Borrelly

1. Introduction represents the variation in the sublimation rate as a function of

the heliocentric distance rh of the comet:
Back in the 19th century, cometary observers noted that the
orbital period of comets was subject to significant changes  r −2.15   r 5.093 −4.6142
h h
from one orbit to another. Having predicted the return of the g(r) = 0.111262 1+ · (1)
2.808 2.808
comet 2P/Encke a few hours too late, Encke (1826) proposed
that the advance of the comet should have been due to a resisting The non-gravitational acceleration vector is decomposed into a
medium causing a deceleration of the comet and therefore a de- radial (A1 g(rh )), a transverse (A2 g(rh )), and a normal (A3 g(rh ))
crease in its orbital period. As this resisting medium does not af- component in the orbital reference frame. The three free pa-
fect the planet motion, Bessel (1836) was the first to suggest that rameters A1 , A2 , and A3 are solved in a least square fit of as-
the orbital period changes stemmed from the ejection of material trometrical observations, together with the six Keplerian oscu-
by a solid nucleus. When a comet approaches the Sun, the solar lating orbital elements. In this model, the parameter A2 models
radiation is used to sublimate ices anisotropically, which trig- the effect of the thermal inertia on a rotating nucleus, which in-
gers a time-dependent recoil force. Because observers noticed troduces a lag angle between the direction of the Sun and that
that the orbital period could also grow, Bessel (1836) proposed of the non-gravitational force (Marsden et al. 1973). This semi-
that the outgassing may act more on one side of perihelion. empirical model soon became the standard to describe the non-
To describe the trajectories of cometary nuclei more accu- gravitational acceleration. It has been used successfully since
rately, Marsden (1969) introduced the first model of the non- the seventies to significantly improve the accuracy of the comets
gravitational accelerations in the equations of motion of the ephemeris. This model has been used for an important study of
comet. This model was later on improved by Marsden et al. the comet 1P/Halley by Rickman & Froeschle (1983a,b) with
(1973), who assumed that the sublimation of water ice acts sym- respect to the need of accurate ephemerides for the observations
metrically with respect to perihelion on a fast-rotating nucleus. and the Giotto spacecraft flyby during the 1986 apparition. Then,
The nucleus is also assumed to be spherical, and the whole they extended this study in a general framework to all the short-
surface is assumed to be isotropically outgassing. Marsden period comets in adding a thermal model interpretation of the
et al. (1973) introduced the dimensionless function g(rh ), which cometary nuclei (Froeschle & Rickman 1986).
Article published by EDP Sciences A81, page 1 of 8
 A&A 548, A81 (2012)

However, measurements of the dust and gas production rates the subsolar meridian, the obliquity of the orbital plane to the
around perihelion acquired during the last perihelion return of nucleus equator, and the solar longitude at perihelion. These pa-
comet 1P/Halley confirmed that the ice sublimation can be asym- rameters, together with a set of osculating orbital elements, are
metric with respect to perihelion (Schleicher et al. 1998). To directly fitted by minimizing a chi-square function that repre-
take the observed asymmetry into account, Yeomans & Chodas sents the distance between the modelled and the measured astro-
(1989) replaced the function g(rh ) by g(rh ), where rh = rh (t−Δt). metrical positions.
The free parameter Δt allows shifting the maximum of the gas The last and most sophisticated model was proposed by
production rate with respect to the perihelion passage. This Davidsson & Gutiérrez (2004). They modelled the nucleus with
model is very useful for improving the ephemeris in the case a prolate ellipsoid, and the surface is divided into thousands of
of comets presenting a large asymmetry in their outgassing. facets. Each facet has its own thermic equilibrium. They tested
For comet 6P/d’Arrest, the best rms residual (1.473 arcsec ver- thousands of models where each facet is chosen randomly to be
sus 1.857 arcsec with the Marsden et al. 1973 model) were ob- active or not. They fit their models on the observed water produc-
tained for Δt = 40 days (Yeomans & Chodas 1989). It should be tion rate. An important result of Davidsson & Gutiérrez (2004)
noted that this model is commonly used by the JPL to calculate is that they can reproduce the asymetric production rate curve.
cometary ephemeris. With the knowledge of the outgassing rate, they computed the
In parallel, electronic imaging data unambiguously revealed non-gravitational force and therefore the mass of the nucleus.
the presence of dust and gas structures in the coma of several In this article, we propose a new implementation of the
comets. Dust jets and filaments also showed up in close-up im- “spotty” model like Sekanina (1993) into the equations of mo-
ages of the inner coma of 1P/Halley by the imaging systems tion of a comet. Compared to the implementation of Sitarski
aboard the Giotto and Vega space probes (Keller et al. 2004). (1990), our model has the following specific characteristics.
Since then, a succession of space mission have repeated
these observations: – We consider latitudinal bands on the surface of an ellipsoidal
– Deep Space 1 on 19P/Borrelly (Soderblom et al. 2002); nucleus, and we fit one parameter per band, which is the per-
– Stardust on 81P/Wild 2 (Sekanina et al. 2004); centage of active surface. Our idea is to develop a model with
– Deep Impact on 9P/Tempel 1 (A’Hearn et al. 2005); a few parameters with a physical meaning.
– and EPOXI (Deep Impact extended mission) on – We model the water production rate from the energy balance
103P/Hartley 2 (A’Hearn et al. 2011). equation.

These structures were interpreted as products by “active areas” The very low thermal inertia of cometary nuclei (e.g. Julian et al.
(Sekanina 1993) or more recently also by concavities (Crifo et al. 2000; Groussin et al. 2007) points toward a significant influence
1999) on the surface of a rotating nucleus. This led to the hy- by the seasonal effects compared to that of the thermal inertia in
pothesis that “seasonal effects” of the successive illumination of the asymmetric behaviour of the production rates.
different areas with different properties on the nucleus’s surface In Sect. 2, we give a thorough description of the model and
can cause the production rate to vary during a perihelion pas- of the algorithms used to retrieve the parameters. In Sect. 3, we
sage, thus affecting the non-gravitational acceleration (Sekanina apply our model to the astrometrical measurements of comet
1993). This led Sekanina (1993) to propose a “spotty model” of 19P/Borrelly and we compare the results with those of the stan-
non-gravitational acceleration in which the perihelion asymme- dard model. We conclude in Sect. 4.
try of the gas production rate was explained by the outgassing
of discrete sources distributed at the surface of the nucleus. In
his model, the maximum outgassing of a given active area took
place when the local elevation of the Sun above the area was 2. Description of the model
at its maximum. The non-gravitational acceleration is expressed
The spotty models are based on determining the latitude and ac-
as the sum of rotation-averaged contributions from all the active
tive surface for each spot. Averaging the effects of this spot over
sources taking their location at the surface of a spherical nucleus
one rotation is equivalent to a model with latitudinal bands that
into account. The other parameters involved in his model are the
are more or less active. Therefore, we restrict our model to a few
direction of the spin axis, the mass of the nucleus, the surface
bands in order to simplify it.
of the active sources, and their sublimation rate as a function of
the heliocentric distance. The free parameters were then fitted
until the non-gravitational acceleration got close to what is pre- 2.1. Expression of the non-gravitational acceleration
dicted by the standard model, in which the parameters Ai best-fit
the astrometrical data. Sekanina (1993) noted a correlation be- The nucleus is modelled as a triaxial ellipsoid divided into latitu-
tween the sign of A2 , the locations of the active sources, and the dinal bands (Fig. 1). The latitude is defined as the angle between
asymmetry of the outgassing. He also pointed out that, unlike the normal to the surface of the ellipsoid and the plane perpen-
the standard model, a negative value of A1 can be produced by dicular to the nucleus rotational axis. The thermal inertia of the
an active area located close to a pole for a given orientation of nucleus is neglected and the gas velocity is considered to be pro-
the spin-axis. portional to the thermal gas velocity. These hypotheses allow
The first attempt to directly introduce the rotation-averaged seasonal effects.
components of the non-gravitational acceleration of the spotty The goal at the beginning is to compute the maximum non-
model into the equations of motion of a comet was made by gravitational force for each band if the whole band is active.
Sitarski (1990) for comet 22P/Kopff. The parameters involved After that it will be possible to fit a coefficient that represents
in the model are the cometocentric latitude β j of the jth active the percentage of the active zone. For each point of the surface,
regions, the constants A j proportional to the active area, and the we have to compute the local non-gravitational force in order
mass of the nucleus and the three angles η, I, and φ, which are to, at the end, integrate it on one nucleus rotation. We need to
respectively the lag angle of the outgassing maximum behind compute the gas production rate and the ejection velocity.
A81, page 2 of 8
 L. Maquet et al.: CONGO, model of cometary NGF combining astrometric and production rate data

ice is interpolated as the following three-degree polynomial from

the values of Washburn (1928)
 
L(T ) = 2740+2.01T −0.014T 2 +0.00002T 3 × 17.529929. (4)
The surfacic sublimation rate is given by
P(T )
Z(T ) = (1 − α)  , (5)
2πMH2 O kT
where MH2 O is the water molecule mass (kg), k the Boltzmann
constant (J K−1 ), α a coefficient for the recondensation of wa-
ter ice on the surface introduced by Crifo (1987) (we adopted
Crifo’s value: α = 0.25), and P(T ) is the water saturation vapor
pressure (Pa) given by Fanale & Salvail (1984) as
 T 
0
P(T ) = A exp − (6)
T
Fig. 1. Geometrical view of a nucleus modelled with seven latitudinal
strips. with A = 3.56 × 1012 Pa and T 0 = 6141.6607 K.
We consider a surfacic sublimation rate that is equal in each
latitudinal band point. Thus, the maximal water production rate
2.1.1. The surface temperature of a band is the product of the rotation-averaged surfacic subli-
mation rate (over one rotation period P) of a point situated at the
The surface temperature is needed to compute the water subli- mid-latitude θ j and the band total surface S j . The water produc-
mation rate for each surface element of the model. The follow- tion rate of the comet QH2 O (t) is given by
ing surface energy balance is solved by dichotomy to obtain the

1 t +P
Nb
temperature of a point on the nucleus surface QH2 O (t) = C jS j Z j (t , θ j )dt (7)
j=1
P t 

F L(T )
(1 − Ab ) cos z = ησT 4 + Z(T ) (2) where Nb is the number of bands, and C j are coefficients describ-
2
rh Na
ing the activity of each surface band. (C j = 0 if the band is not
outgassing, C j = 1 if the whole band is outgassing.)
with

– Ab : the Bond albedo (equal to the product of the geometric 2.1.3. Non-gravitational accelerations
albedo by the phase integral);
– F : the solar constant (W m−2 ); From the water production rate, we derive the non-gravitational
– rh : the heliocentric distance (AU); acceleration with the Newton second law of motion

  
1 1 t +P dF(t)
– z: the elevation of the Sun; Nb
– η: the beaming factor introduced by Lebofsky & Spencer ANG (t) = C jS j dt (8)
MC j=1 P t dS j
(1989);
– : the nucleus infrared emissivity;  
where dF(t) is the surfacic force of the jth band depending on
– σ: the Stefan-Boltzmann constant (J K−4 m−2 s−1 ); dS j
– T : the surface temperature (K) depending on the time and the sublimation rate Z j and the thermal gas velocity Vg j
the latitude of the point on the surface;  
dF(t)
– L(T ): the sublimation latent heat (J mol−1 ); = Z j (t)Vg j (t)MH2 O N j (9)
dS j
– Na : the Avogadro constant (mol−1 );
– Z(T ): the surfacic sublimation rate (molecules s−1 m−2 ). where MH2 O is the water molecular mass and Ni is the surface
normal.
The elevation of the Sun z is the angle between the sub-solar We used the expression of the gas ejection velocity proposed
point and a point on the nucleus surface. It is defined as by Crifo (1987):

cos z = cos θ cos θ cos(ϕ − ϕ) + sin θ sin θ (3) 8kT n (t)
Vg j (t) = η (10)
πmH2 O
with
with k the Boltzman constant (k = 1.380662 × 10−23 J K−1 ),
– (θ , ϕ ): the coordinates in the cometocentric frame of the and η a local parameter varying in relation with the position on
sub-solar point; the cometary nucleus. This parameter allows us to consider that
– (θ, ϕ): the considered point on the nucleus. the comet surface is not composed of pure ice and is not perfectly
smooth. Crifo (1987) suggests choosing this parameter in the in-
The axis of the cometocentric frame is along the pincipal axis of terval [0.39, 0.5]. We fix the value of this parameter to 0.45 for
inertia of the ellipsoidal nucleus. the application to comets. It should be noted that, from the ex-
pression of the non-gravitational acceleration, the determination
2.1.2. Water sublimation rate of the nucleus mass is inversely proportional to this parameter.
This expression of the non-gravitational acceleration can be
The water sublimation rate is the key value in computing non- introduced into the equation of the comet’s motion after de-
gravitational forces because with the ejection velocity, it gives composition into three components in the equatorial heliocentric
the quantity of motion. The sublimation latent heat of the water reference frame.
A81, page 3 of 8
 A&A 548, A81 (2012)

2.2. Equations of motion Table 1. Distribution of the astrometric observations from MPC.
We develop the equations of motion in an equatorial heliocen-
tric reference frame (P , x, y, z). The variables are the Cartesian Perihelion time Observation interval Number of obs.
coordinates of the position (rh ) of the comet. The differential 1994 Nov. 1 1994 Jun. 12–1997 Jun. 24 557
equations of the motion of the comet can be written as 2001 Sep. 14 1999 Jun. 13–2003 Apr. 28 853
2008 Jul. 22 2007 Aug. 19–2009 Jun. 16 796

9  
d2 rh rh ri − rh ri
= −GM + Gm i − + R + ANG
dt2 rh 3 i=1 ri − rh 3 ri 3
(11)
3.1. Observational data and parameters of the model
with
This present investigation is based on two kinds of observa-
– rh : heliocentric position vector of the comet; tions: astrometric positions and water production rates obser-
– G: Gauss gravitation constant; vations. Astrometric measurements were taken from the IAU
– M : solar mass; Minor Planet Center. The observational material contains 2206
– mi : mass of the ith perturbating body; observations from June 12, 1994 to June 16, 2009. These obser-
– ri : heliocentric position vector of the ith body; vations cover three orbits of the comet but are not evenly dis-
– R: acceleration due to relativistic effects; tributed (Table 1). The number of observations was increased
– ANG : non-gravitational acceleration (Sect. 2.1). since 2001 perihelion because of the support to the space mis-
The acceleration due to relativistic effects is given by Beutler sion Deep Space 1.
(2005) To determine the mass of the nucleus, we used a compilation
⎡⎛  2 ⎞   ⎤ of 219 observations of the water production rates covering the
GM ⎢⎢⎢⎢⎜⎜⎜⎜ GM drh ⎟⎟⎟⎟ drh drh ⎥⎥⎥⎥ same time period as astrometrical positions (see Maquet 2012,
R= 2 ⎢⎜4 − ⎟ rh + 4 rh ⎥ (12)
c rh 3 ⎣⎝ rh  dt ⎠ dt dt ⎦ for a detailed table of these observations). This compilation con-
sists of data sets coming from different methods of observations
where c is the light velocity. It is important to introduce this ac- of water or of its photolysis products:
celeration in the equation of motion because, for a comet passing
close to the Sun, it can modify the value of the non-gravitational
– direct observations of water from its submillimetric line at
accelerations.
557 GHz with the Odin satellite (Lecacheux et al. 2003)
We fit the orbital motion on astrometric measurements and
Ci (2001 passage);
derive the M parameters.
c – ground-based observations of OH from its 18-cm lines, with
the Nançay radio telescope (Bockelée-Morvan et al. 2004)
2.3. The determination of the nucleus mass (1994 and 2001 passages);
– ground-based observations of the OH radical in the near UV
From Eq. (7), we can see that the water production rate is propor- (Schleicher et al. 2003) (2001 passage);
tional to the coefficients Ci . Nevertheless, thanks to astrometric – space observations of the Lyman-α line of hydrogen with
measurements, we can determine the ratio Ci /Mc . Thus, we can the SOHO/SWAN instrument (Combi et al. 2011) (2001
introduce these ratios into the water production rate in order to passage);
deduce the nucleus mass: – observations of the OI forbidden line in the visible (Fink
Nsub  Nb  Ci (i)  2009) (1994 passage).
j=1 Qobs (t j ) i=1 Mc Qmax (t j )
Mc = N N  C (i) 2 (13)
sub b
j=1 j=1 Mc
i
Q max (t j ) The OH production rates were converted to water production
rates by assuming QH2 O = 1.1QOH (Huebner et al. 1992).
where Qobs (t j ) is the water production rate observed, Q(i)
max (t j ) The water production rate is observed to peak around 6 ×
the maximal water production rate calculated for the ith band at 1028 molecules/s about 20 days before perihelion (Fig. 5). We
the same time t j , and Nsub and Nb are the number of observations see that Lyman-α observations give significantly different results
and the number of bands, respectively. than the other methods.
The nucleus mass and the corresponding error are deter- In this work, we consider the nucleus of comet 19P/Borrelly
mined as the mean value and the standard deviation of a as a triaxial prolate ellipsoid (a = 4.6 km, b = 1.8 km, c =
Monte-Carlo distribution taking the precision on the observa- 1.6 km). rotating in 26 h (Lamy et al. 1998).
tions and errors on the ratio Ci /Mc into account.

3.2. The fit to astrometrical data

3. Application to comet 19P/Borrelly
The comet 19P/Borrelly was discovered in 1904 by the French The goal of the fit is to optimize the parameters of the model
astronomer Alphonse Borrelly at the Marseilles observatory (such as the initial conditions and the non-gravitational param-
(Borrelly 1905). With its short period (6.75 years) and low in- eters divided by the mass of the nucleus) in order to minimize
clination (30◦), it belongs to the Jupiter family. It is a well- the differences between the observed positions of the comet
documented comet observed at many of its returns. It was the (α(o) (o) (c) (c)
n , δn ) and those delivered by our model (αn , δn ) at the
target of Deep Space 1 mission in 2001 (Soderblom et al. 2002). same time tn . These differences are noted (O–C)n .
We also chose this comet for its stable axis (Schleicher et al. We first fit independently the non gravitational parameters
2003) to avoid problem due to a moving axis. and the initial position and velocity of the comet in a iterative
A81, page 4 of 8
 L. Maquet et al.: CONGO, model of cometary NGF combining astrometric and production rate data

Fig. 2. Reduced-χ2 map in relation with the spin axis orientation. 102 points have been calculated on one hemisphere only because the problem
is symmetric. There are 25 points on the celestial equator and the number decreases like cos δ because of the degeneration in the pole. The other
points have been interpolated.

process. Denoting ϕ j as any Cartesian coordinates of the comet, with χ2 the sum of (O–C)2n , Nast the number of observations, Npar
the (O–C)n depend on the constant of the model pi as the number of parameters, and σi j the elements of the variance-
⎧ covariance matrix.
⎪
⎪
⎪ 3 N
∂α ∂ϕ j
⎪
⎪
⎪ Δα = Δpi
⎪
⎪
⎪
n
∂ϕ j ∂pi
⎨ j=1 i=1 3.3. Spin axis orientation determination
(O−C)n = ⎪⎪ (14)
⎪
⎪ 3 N
∂δ ∂ϕ j
⎪
⎪
⎪
⎪
⎪ Δδ = Δpi Up to now, the spin axis orientation has mainly been determined
⎩ n ∂ϕ j ∂pi
j=1 i=1 by observations of the coma morphology, which is dominated
or, in matrix notation (O–C)n = HφX, as by a primary jet (Samarasinha et al. 2004). The orientation of
⎛ ∂x the axis seems to be stable since Deep Space 1 measurements
⎜ ∂x ⎞⎟
⎛ ∂α ∂α ∂α ⎞ ⎜⎜⎜⎜ ∂p , . . . , ∂p ⎟⎟⎟⎟ ⎛ are compatible with the previous determination. This primary
  ⎜⎜⎜⎜ , , ⎟⎟⎟ ⎜⎜⎜ 1
⎜⎜⎜ ∂y
N ⎟⎟⎟⎟ ⎜⎜ Δp1 ⎞⎟⎟ jet, shown as stable in orientation and morphology by Deep
Δαn ⎜
⎜ ∂x ∂y ∂z ⎟⎟⎟ ⎜⎜⎟
⎟ ∂y ⎟⎟⎟ ⎜⎜⎜ . ⎟⎟⎟ Space 1 mission, seems to be closely aligned with the rotation
= ⎜⎜⎜⎜ ⎟ ⎜⎜⎜ ,..., ⎟⎟ ⎜⎜ . ⎟⎟ (15)
Δδn ⎟
⎜⎜⎝ ∂δ ∂δ ∂δ ⎟⎟⎠ ⎜⎜ ∂p1 ∂pN ⎟⎟⎟⎟ ⎜⎜⎝ . ⎟⎟⎠ axis (Soderblom et al. 2002). Several authors measure the loca-
, , ⎜⎜⎜ ⎟⎟⎟ ΔpN tion of this primary jet from (α = 214◦ ± 4◦ , δ = −5◦ ± 4◦ ) by
∂x ∂y ∂z ⎜⎜⎝ ∂z ∂z ⎟⎟⎠
,..., Farnham & Cochran (2002) to (α = 233◦, δ = −18◦) by Thomas
∂p1 ∂pN
et al. (2001).
where
⎛ ⎞ In this investigation, we try to determine the rotational axis
⎜⎜⎜ − sin α cos α
0 ⎟⎟⎟⎟
⎜⎜⎜ rh cos δ rh cos δ ⎟⎟⎟ through the dynamics of the body. Indeed, as seen in Sect. 2.1,
H = ⎜⎜⎜⎜ ⎟.
⎜⎝⎜ sin δ cos α sin δ cos α cos α ⎟⎟⎟⎠⎟
(16) the non-gravitational forces are directly related to the insolation
− − of the nucleus, hence to its axis of rotation position. In our work,
rh cos δ rh cos δ rh we consider a fixed axis position. We explored all the space with
To solve this kind of system by a least square method, we have a 15◦ grid. This hypothesis was based on Samarasinha et al.
to compute the partial derivates matrix φ. Each element of the (2004). Moreover, as noted in Samarasinha et al. (2004), the
matrix φ is obtained by numerical derivation as comet 19P/Borrelly is in an unexcited spin mode, and its axis
slowly precesses by 5◦ −10◦ per century, which is much longer
∂ϕ j ϕ j (p1 , . . . , pi + , . . . , pN ) − ϕ j (p1 , . . . , pi − , . . . , pN ) than a revolution period (about 6.8 years). This last consideration
=
∂pi 2 allows us to perform this work now. To this end, we performed a
(17) astrometrical fit reduced-χ2 map with different axis orientations
(Fig. 2).
with , a small variation on one of the model parameters pi .
We fit the different parameters in an iterative process. In one This map shows two large zones where the reduced-χ2 is
iteration, the optimization is divided into two sub-fits because of very low (about 2.06), which correspond to the optimal axis
the difference of order between magnitude of the parameters. We of rotation for a prograde or a retrograde nucleus. As noted in
begin by the fit of the ratios Ci /Mc and then optimize the initial Sect. 2.1, the forces are averaged over one rotation of the nu-
position and velocity of the comet. The errors on the parameters cleus and, as our model does not contain thermal inertia, it is
are determined through the variance-covariance matrix of all the impossible to distinguish between the two solutions for the spin
parameters as axis orientation. These zones, projected onto a sphere, are de-
scribing two caps and the bad reduced-χ2 are situated on the
χ2 great circle of the celestial sphere. To find the optimal solu-
δpi = − σii (18)
Nast − Npar tions, we fit the poles to correspond to this great circle. We
A81, page 5 of 8
 A&A 548, A81 (2012)

Table 2. Parameters of the model found after fitting to the astrometrical observations and corresponding osculating elements for comet
19P/Borrelly.

Parameter Value and error

Position x 0.481390947 ± 3.72 × 10−7 AU
Position y 1.196797597 ± 2.72 × 10−7 AU
Position z 0.425142676 ± 3.18 × 10−7 AU
Velocity V x −1.502141214 × 10−2 ± 3.91 × 10−9 AU days−1
Velocity Vy 2.001264858 × 10−3 ± 4.36 × 10−9 AU days−1
Velocity Vz 1.114184800 × 10−2 ± 4.13 × 10−9 AU days−1
C1 /MC 1.84 × 10−15 ± 1.06 × 10−15 kg−1
C2 /MC 1.26 × 10−14 ± 4.11 × 10−16 kg−1
C3 /MC 9.42 × 10−15 ± 9.18 × 10−16 kg−1
Time of perihelion passage September, 14.72972 2001 TT ± 1.0 × 10−4 h
Argument of periapsis ω 353.3749882 ± 2.54 × 10−5◦
Longitude of the ascending node Ω 75.4249176 ± 3.39 × 10−5◦
Inclination i 30.3247261 ± 1.61 × 10−5◦
Perihelion distance q 1.3582105 ± 3.0 × 10−7 AU
Eccentricity e 0.6238974 ± 1.1 × 10−6
Semi-major axis a 3.6112765 ± 1.13 × 10−5 AU
Mean motion n 0.14361935 ± 6.7 × 10−5◦ /days
Period P 6.863 ± 3.21 × 10−5 years

Notes. The position and velocity are given at the Julian epoch 2 452 166.50 (14/09/2001, 0h00 TT).

found two possible poles (α = 190◦ , δ = 30◦ ) and (α = 10◦ ,

δ = −30◦ ). Error bars were impossible to compute with our
“great circle” fit on bad χ2 . But, as shown in Fig. 2, the axis
position is within a flat χ2 zone, so the peak to valley error has
an order of 40◦ . The previous determinations are far from our
central determination but still within error bars. A possible ex-
planation of these discrepanscies is that our model is still too
simple. The thermal inertia must play a large role in polar axis
determination. Nevertheless, our method works, and we hope an
improved thermal model can be used after the Rosetta explo-
ration of 67P/Churyumov-Gerasimenko.

3.4. The CONGO model (COmetary Non-Gravitational

Orbits)
Fig. 3. Post-fit residuals in arcseconds in right ascension and
In the following, we adopted the spin axis orientation (192◦ , declination.
29◦ ). The nucleus is divided into three latitudinal bands: two po-
lar caps situated between the latitudes −90◦ , −45◦ and 90◦ , 45◦ where Δobs is the accuracy of the observations. We performed
and an equatorial band between −45◦ , 45◦. The ratio C1 /MC rep- a similar fit with the Marsden et al. (1973) model in the same
resents the southern cap, C2 /MC the equatorial band, and C3 /MC conditions as for our model. We can compare the reduced χ2
the northern cap. of the two models: 2.20 with the Marsden et al. (1973) model
and 2.06 with the CONGO model.
3.4.1. Astrometrical data fitting
3.4.2. Mass determination
As explained in Sect. 2, we first fit astrometrical data in order
to determine MCCi parameters. This fit is sufficient to produce To get the mass from Eq. (13), we used a Monte Carlo method
Ci
ephemeris. We carried out the fit on the astrometrical data in on M c
and Qobs From this distribution (see Fig. 4), we deduce
equatorial J2000 Cartesian coordinates. The equation of motion that the optimal mass of the comet is 2.21×1013 ± 0.27×1013 kg.
used are those presented in Sect. 2.2. We performed the integra- The error on the mass may seem very small, but it should be
tion using the Radau-integretor (Everhart 1985). The results of noted that the errors on the water production rate observations
the fit are given in Table 2. just take the S /N into account and not the uncertainty due to the
For our best fit (see Fig. 3), we obtain an rms of 1.43 . To Haser model and any other bias. In reality, we expect an error
compute the reduced χ2 , we assumed an arbitrary error of 1 for close to 50%. This estimation of the nucleus mass corresponds
all the astrometric data: to a density of about 400 kg m−3. This result is quite compara-
ble to the estimate 150 kg m−3 ≤ ρ ≤ 450 kg m−3 obtained by
Nast Davidsson & Gutiérrez (2004) (considering a nucleus volume
(O−C)n 2 equal to 5.5 × 1010 m3 , see Lamy et al. 1998) and supports the
χ2(R) = n=1
(19) hypothesis of a low-density object (see Table 3).
Nast Δobs
A81, page 6 of 8
 L. Maquet et al.: CONGO, model of cometary NGF combining astrometric and production rate data

Fig. 4. Distribution of the mass estimations from the Monte-Carlo

simulation. Fig. 5. The gas prodution rates obtained for ellipsoidal nucleus with a
spin axis orientation (α, δ) = (192◦ , 29◦ ) compared with observations
for the 2001 perihelion passage.
Table 3. Masses and densities (for a volume of 55 km3 ) estimated in
other studies compared to the present work.
Table 4. Percentage of active areas in the three bands.
12 −3
Paper Mass (×10 kg) Density (kg m )
(1) 2.7 ± 2.1 50 ± 40 Band Surface on the Percentage Percentage
(2) 8–24 150–450 nucleus (km2 ) of the band of the nucleus
(3) 33 600 South cap 19.74 3–6 1–2
(4) <8.4 <150 Equator band 38.54 7–19 9–21
(5) 15–33 270–600 North cap 19.74 14–31 4–8

References. (1) Sosa & Fernández (2009), (2) Davidsson & Gutiérrez
(2004), (3) Farnham & Cochran (2002), (4) Rickman et al. (1987),
(5) this study. 3.5.1. Active zones
Knowing the value of the mass and the Ci /Mc parameters, we
can now deduce the fraction of active area on the three bands of
the considered nucleus. As for the mass and the density deter-
3.5. Discussion of results mination, we can calcule the two extreme values of the active
surface interval by considering the Combi et al. (2011) obser-
To analyse our results, we can draw our theoretical water produc- vations or not. These results are reported in Table 4. The total
tion rate curve on data. We can see that the fit is mainly made active area percentage is between 14% and 31%. This estima-
on the big amount of data coming from Combi et al. (2011) tion is compatible with the previous determination (8–18%) of
(Fig. 5). Nevertheless, there is a systematic offset by a factor Davidsson & Gutiérrez (2004).
of 2.5 between Combi et al. (2011) data and the other mea-
We can note that the southern hemisphere is less active
surements (Bockelée-Morvan et al. 2004; Fink 2009; Lecacheux
than the northern one. This result corresponds to the Deep
et al. 2003; Schleicher et al. 2003). This was also viewed on the
Space 1 obervations of the northern hemisphere and to the post-
comets Hale-Bopp (Combi et al. 2000) and Hyakutake (Combi
perihelion observations by Farnham & Cochran (2002) and by
et al. 2005) data. As this offset is quite constant, we decided to
Schleicher et al. (2003) when the southern hemisphere was in
compute two separate fits, one on the Combi et al. (2011) data
the direction of the Sun (Davidsson & Gutiérrez 2004).
and the other on all the other data.
Figure 5 shows the gas production rate compared to obser-
The Combi et al. (2011) data are based on direct observa- vations. We note that the peak of the curve is shifted by 20 days
tion of H. The problem is that H come from H2 O but also from before perihelion. This is due to the ellipsoidal shape and the
many other molecules. This is taken into account in the model, specific axis of rotation of the nucleus (Davidsson & Gutiérrez
but we can say that it can be considered as the maximum value 2004). Indeed, before perihelion, the northern hemisphere is the
of H2 O production rate.The mass and the density found with most insolated part of the comet. Considering the locations of
this set of data are 33 × 1012 kg and 600 kg m−3 . In another active areas on the nucleus presented above, it is easy to under-
band, the estimations with OH are indirect measurements of the stand that the maximum outgassing occurs before perihelion.
H2 O production rate, and they can be considered as the mini-
mum estimation. The mass and the density found with these data
are 15 × 1012 kg and 270 kg m−3 .
4. Conclusions
We computed the error on the mass, with the astrometrical
fit and the errors given by authors of the H2 O production rate. This work shows that it is possible to use a simple realistic model
We found a 15% error, which seems rather optimistic. In fact, based on the physics of H2 O outgassing for a comet ephemeris.
the errors on the mass estimation are mainly due to H2 O produc- The model will evolve in the future with a better thermal model.
tion rates. We note a factor of 2.5 difference between the method But even now, it can produce more accurate ephemerides than
measuring H and the others measuring OH. previous models (Marsden et al. 1973).
A81, page 7 of 8
 A&A 548, A81 (2012)

Moreover, just with astrometrical data, it is possible to es- Bessel, F. W. 1836, Astron. Nachr., 13, 345
timate the spin axis orientation. This determination is still raw, Beutler, G. 2005, Methods of celestial mechanics, Vol. I: Physical, mathematical,
mainly because of our simplistic thermal model. But, it can be and numerical principles (Berlin: Springer)
Bockelée-Morvan, D., Biver, N., Colom, P., et al. 2004, Icarus, 167, 113
applied to any comet, even a faint one. Borrelly, A. 1905, CRAS, 140, 104
The most important product of our model is determining Combi, M. R., Reinard, A. A., Bertaux, J.-L., Quemerais, E., & Mäkinen, T.
of the comet nucleus mass by fitting the H2 O production rate. 2000, Icarus, 144, 191
Using Deep Space 1 volume determination of the nucleus of Combi, M. R., Mäkinen, J. T. T., Bertaux, J.-L., & Quemérais, E. 2005, Icarus,
177, 228
19P/Borrelly, we evaluated its density to be between 270 and Combi, M. R., Lee, Y., Patel, T. S., et al. 2011, AJ, 141, 128
600 kg m−3 . It confirms the hypothesis of a low-density, porous Coradini, A., Capaccioni, F., Drossart, P., et al. 2007, Space Sci. Rev., 128, 529
object. Most of the uncertainty comes from the H2 O production Crifo, J. F. 1987, A&A, 187, 438
rate estimation. An additional error is due to the rather arbitrary Crifo, J. F., Rodionov, A. V., & Bockelee-Morvan, D. 1999, Icarus, 138, 85
choice of the η parameter that governs the gas expansion velocity Davidsson, B. J. R., & Gutiérrez, P. J. 2004, Icarus, 168, 392
Encke, J. F. 1826, Berliner Astronomische jahrbuch fur das jahr, 1826, 124
in Eq. (10). Everhart, E. 1985, in Dynamics of comets: Their Origin and Evolution, 185
Our estimation is compatible with previous determinations, Fanale, F. P., & Salvail, J. R. 1984, Icarus, 60, 476
especially with Davidsson & Gutiérrez (2004) who are using a Farnham, T. L., & Cochran, A. L. 2002, Icarus, 160, 398
more sophisticated thermic model. Fink, U. 2009, Icarus, 201, 311
Froeschle, C., & Rickman, H. 1986, A&A, 170, 145
Up to now, it has not been possible to measure cometary Groussin, O., A’Hearn, M. F., Li, J.-Y., et al. 2007, Icarus, 191, 63
masses from spacecraft flybys, because these masses are too Gulkis, S., Frerking, M., Crovisier, J., et al. 2007, Space Sci. Rev., 128, 561
low to significantly affect the spacecraft trajectory. It is an- Huebner, W. F., Keady, J. J., & Lyon, S. P. 1992, Ap&SS, 195, 1
ticipated, however, that this will be possible with the Rosetta Julian, W. H., Samarasinha, N. H., & Belton, M. J. S. 2000, Icarus, 144, 160
spacecraft orbiting 67P/Churyumov-Gerasimenko (Pätzold et al. Keller, H. U., Britt, D., Buratti, B. J., & Thomas, N. 2004, in Comets II, eds.
M. C. Festou, H. U. Keller, & H. A. Weaver, 211
2007). It will be a good test for our model. Lamy, P. L., Toth, I., & Weaver, H. A. 1998, A&A, 337, 945
We can also measure the fraction of active surfaces. Our Lebofsky, L. A., & Spencer, J. R. 1989, in Asteroids II, eds. R. P. Binzel,
model is quite robust since it uses the same number of param- T. Gehrels, & M. S. Matthews, 128
eters as do Marsden et al. (1973), but it is closer to the phys- Lecacheux, A., Biver, N., Crovisier, J., et al. 2003, A&A, 402, L55
Maquet, L. 2012, Ph.D. Thesis, Institut de Mécanique Céleste et de Calculs
ical reality. We can apply this method to all comets for which d’Éphémérides
the size is known. If the size is unknown, we can just compute Marsden, B. G. 1969, AJ, 74, 720
ephemerides of the comet. Marsden, B. G., Sekanina, Z., & Yeomans, D. K. 1973, AJ, 78, 211
Our model is a first attempt to bring together celestial me- Pätzold, M., Häusler, B., Aksnes, K., et al. 2007, Space Sci. Rev., 128, 599
Rickman, H., & Froeschle, C. 1983a, in Cometary Exploration, ed.T. I. Gombosi,
chanics and a realistic comet nucleus model. It can be used 3, 109
right now to compute ephemerides. Its physical parameters still Rickman, H., & Froeschle, C. 1983b, in Cometary Exploration, ed. T. I.
need to be improved. In this resspect, the future in situ explo- Gombosi, 1, 75
ration of 67P/Churyumov-Gerasimenko by Rosetta will be help- Rickman, H., Kamel, L., Festou, M. C., & Froeschle, C. 1987, in Diversity and
ful. Infrared sounding of the nucleus (Coradini et al. 2007) will Similarity of Comets, eds. E. J. Rolfe, & B. Battrick, ESA SP, 278, 471
Samarasinha, N. H., Mueller, B. E. A., Belton, M. J. S., & Jorda, L. 2004,
lead to an accurate thermal model, and the radio spectroscopic Comets II, eds. Festou, M. C., Keller, H. U., & Weaver, H. A., 281
observations (Gulkis et al. 2007) will directly measure the gas Schleicher, D. G., Millis, R. L., & Birch, P. V. 1998, Icarus, 132, 397
jet velocities. Schleicher, D. G., Woodney, L. M., & Millis, R. L. 2003, Icarus, 162, 415
The next step will also be to take the moving axis of rotation Sekanina, Z. 1993, AJ, 105, 702
Sekanina, Z., Brownlee, D. E., Economou, T. E., Tuzzolino, A. J., & Green, S. F.
into account. With this, we really hope to extend the domain of 2004, Science, 304, 1769
validity of the orbits, and perhaps to have a better idea of the past Sitarski, G. 1990, Acta Astron., 40, 405
of comets. Soderblom, L. A., Becker, T. L., Bennett, G., et al. 2002, Science, 296, 1087
Sosa, A., & Fernández, J. A. 2009, MNRAS, 393, 192
Thomas, N., A’Hearn, M. F., Boice, D. C., et al. 2001, in BAAS 33,
References AAS/Division for Planetary Sciences Meeting Abstracts, 1074
Washburn, E. 1928, International Critical Tables, Vol. III (New York:
A’Hearn, M. F., Belton, M. J. S., Delamere, W. A., et al. 2005, Science, 310, 258 McGraw-Hill)
A’Hearn, M. F., Belton, M. J. S., Delamere, W. A., et al. 2011, Science, 332, Yeomans, D. K., & Chodas, P. W. 1989, AJ, 98, 1083
1396

A81, page 8 of 8