A&A 372, 784-792 (2001)
DOI: 10.1051/0004-6361:20010571
Y. Revaz - D. Pfenniger
Geneva Observatory, University of Geneva, 1290 Sauverny, Switzerland
Received 22 November 2000 / Accepted 9 April 2001
Abstract
It is often assumed that a warped
galaxy can be modeled by a set of rings.
This paper verifies numerically
the validity of this assumption by the study of periodic orbits populating a
heavy self-gravitating warped disk.
The phase space structure of a warped model reveals that the circular periodic
orbits of a flat disk are transformed in quasi annular periodic orbits which conserve their stability. This lets us also explore the problem
of the persistence of a large outer warp. In particular,
the consistency of its orbits with the density distribution is checked as a
function of the pattern speed.
Key words: galaxies: kinematics and dynamics of galaxies - galaxies: spiral
In the conventional understanding of galaxies embedded in a kinematically hot spheroidal or triaxial halos of dark matter, the frequently observed pronounced outer warps of neutral hydrogen, interstellar dust and, to a lesser noticeable extent, optical disks set a challenge for explaining their presumably long-lived existence, and particularly, their frequent quasi-straight line of nodes (LON), or pair of quasi-straight LON (Briggs 1990). At least half the spirals possess a detectable warped HI and stellar disk(e.g., Briggs 1990, and more should remain undetected due to projection confusion.
Hypotheses like cosmic infall (Jiang & Binney 1999), gravitational interactions (Hernquist 1991; Weinberg 1995, 1998), normal modes (Sparke 1984; Sparke & Casertano 1988), misaligned dark halos (Kuijken 1991; Dubinski & Kuijken 1995; Debattista & Sellwood 1999) or magnetic fields (Battaner et al. 1990) have been proposed, without clearly satisfying all the constraints provided by the observations. See Binney (1992) or Kuijken (2000) for reviews.
But, as argued by Arnaboldi et al. (1997), Becquaert & Combes (1998) and Reshetnikov & Combes (1998), also in the context of polar rings, the evidence for non-self-gravitating HI disks is actually weak. Instead, the opposite assumption of coexistence of warps containing most of the dynamical mass is not in contradiction with the available observations. Such an assumption of thick and flaring disks of visible and dark matter in the form of cold gas approximately proportional to neutral hydrogen has been proposed by Pfenniger et al. (1994) to account for many more known facts about spirals. For the warp problem, the self-gravitating disk assumption is attractive for several reasons. HI observations reveal that most of the galaxies possess an inner flat disk, while the warp develops beyond a specific radius (Briggs 1990; Burton 1992). Such a comportment, which is the signature of a radical change of the dynamic along the radius, can hardly be explained by a hot halo which would impose a uniform dynamic over a larger scale. In contrast, a strong rotational support in the dark matter component offers a natural explanation for the straight lines of nodes. They would result as a natural consequence of the weak diffusive properties of angular momentum in a self-gravitating disk: if, following for example an accretion event, angular momentum is deposited in a self-gravitating disk, bending its outer parts, the angular momentum will slowly diffuse across the whole disk in a monotonous way, because angular momentum is a quasi conserved quantity. As its diffusion is slow, the line of nodes must be quasi-straight and persist for many rotational periods. Finally, self-gravitating optical disks are in agreement with the often stated maximum disk interpretation of observations of the Milky Way and other spirals (see for example Gerhard 2000). In numerous models of warped galaxies, and also polar ring models, stars are assumed to move along circular rings which are tilted as a function of radius. Because stars are driven by periodic orbits, which are the backbone of galaxies, such models implicitly presuppose the existence of stable circular tilted periodic orbits. Yet, to our knowledge they have never been verified, and do not appear obvious in non-spherical geometries. It is far from obvious that a perturbation in the form of a warp will conserve the well known circular orbits of a flat disk. In order to verify the latter assumptions, it is useful to find the exact periodic orbits existing in a warped disk. Moreover, understanding the stable periodic orbits allows us to grasp the basic properties of the other quasi-periodic orbits in an efficient way, particularly when no analytical tools exist. Such an approach has been very useful in the context of barred galaxies for which the complexity of motion is substantial (e.g., Contopoulos & Papayannopoulos 1980; Pfenniger 1984). For studying the principal periodic orbits in a galactic potential, even rough representations of the potential are sufficient to yield the basic properties of the main periodic orbits, as that shown in numerous situations. In this work, the disk autogravitation will play a dominant role. Therefore, the following results will be applicable as long as the local density is dominated by the disk.
The alternative case would be the existence of a hot dark matter halo. In this case, one must admit a center dominated by the halo density, because the density of a warm gravitating system increases at the center. The following implications should be explored:
In order to compute periodic orbits, we have based our model on a superposition
of three simple Miyamoto-Nagai potentials (Nagai & Miyamoto 1975).
The three components can
be viewed as representing respectively a bulge, a visible disk and a gas disk
containing a large amount of dark matter.
In cylindrical coordinates, the total potential can be written as:
Replacing z by
in the potential (1) we get an analytical
warped disk potential model:
The parameters a_{i}, b_{i} and GM_{i} have been chosen
to provide a typical
rotation curve, increasing linearly below
and staying flat up to
,
beyond which it decreases smoothly.
For the bulge,
,
and
,
which is equivalent to a Plummer sphere.
The parameters for the visible and gas disks are respectively:
,
,
,
,
,
.
For convenience the length unit is chosen to be the kpc and time unit the Myr. Fixing
the gravitational constant G to a value of 1, the mass unit corresponds to
.
With this set, the velocity unit is about
.
Figure 1 shows the isodensity
curves corresponding to the model. Well above the galaxy disk, negative density regions do
exist, however our results are not
spoiled by negative densities as long as orbits do not cross such regions.
Figure 1: 3D isodensity surface of the warped model at and (w=0.005 kpc^{-1}). The surface is sliced with respect to the y=0 plane. | |
Open with DEXTER |
In the present model, we allow the possibility for a global rotation of the
warped potential about the z-axis. This rotation is parametrised by the angular
pattern speed
or, equivalently, by the corotating radius
,
taking a negative value for retrograde rotation. Thus, the Hamiltonian
in the rotating frame of reference with angular speed
is:
The periodic orbits are found by numerically determining the fixed points of the 4D Poincaré map (T) at y=0, generated by the equations of motion (see, e.g., Pfenniger & Friedli 1993). The algorithm uses in particular the method proposed by Hénon (1982) giving the intersection between an orbit and a surface, and a least squares stabilized Newton-Raphson root-finding procedure.
The stability of the orbits are determined by the eigenvalues of
the Jacobian of the Poincaré map ().
Since the system is Hamiltonian and the motion is described by
real numbers, the four eigenvalues of
occur by conjugate and inverse pairs
and it is possible to condense the information with two stability indexes b_{1}, b_{2}:
In this work, we will focus our investigations on orbits following the potential disk. Hence, we choose initial conditions with a starting position on the line of nodes, here equivalent to the y-axis. We concentrate on the main orbit families keeping the symmetry of the potential, thus the initial velocity is perpendicular to it (p_{y}(0)=0) and points to arbitrary negative values of x (p_{x}(0)<0). In this way, the free parameters in the initial conditions are the Jacobi constant (energy, H), the position y(0) along the y axis, and the velocity component p_{z}(0) along the z direction. This latter parameter will be small compared to p_{x}(0), which avoids a situation where orbits stray too far from the disk and penetrate negative density regions. This point has been systematically checked because, a priori, orbits may explore regions far from the expected ones.
In order to understand the influence of the warp, we first look at the model without deformation (w=0) which is completely axisymmetric. The rotation of the potential is set to zero ( ).
Below the orbit, families are shown with their initial starting point in the H-p_{z}(0) diagram. In Fig. 2, bottom, we recognize the circular orbit family (horizontal line at p_{z}(0)=0). The stability indices for this family are traced at the top of Fig. 2. In this particular case, b_{1} indicates the stability in the galaxy plane while b_{2} corresponds to the stability transverse to it. Both indices remain in the interval [-2,+2], insuring the stability for this family.
At H=-0.0701, b_{1}=+2, is also degenerate with two eigenvalues equal to -1. A bifurcation occurs through period doubling. This bifurcation coincides with the resonance between the radial frequency and the circular frequency ( ).
At H=-0.0748, -0.0566, -0.0451, -0.0345, and -0.0236, b_{2}=+2 generating transverse bifurcations through period doubling. These bifurcations coincide with the resonances between the transverse frequency and the circular frequency ( , ). Orbit families from period doubling bifurcations will not be discussed further in this paper.
Transverse bifurcations keeping the same period occur at H=-0.0632, -0.0508, -0.0396, -0.0294, and -0.0158. In this case b_{2}=-2 and , . The two latter families have a non zero p_{z}(0)initial velocity and also oscillate round the z=0 plane crossing it respectively 2k and 2(k+1) times per period. All these sub-families are marginally stable with b_{2}=-2 due to the axisymmetry of the potential.
Figure 2: H-p_{z}(0) phase space for w=0. | |
Open with DEXTER |
The break in the z symmetry transforms the family of circular orbits in the unperturbed model into a set of families pk (main families) with . The significance of the indice k will be explained further. Except at bifurcations, the pk families are stable.
The shape of the corresponding orbits can be approximated
with the following parametric function:
Another influence of the warp is to move the bifurcations
toward slightly lower energies. The corresponding transverse bifurcations
arise at H=-0.0633, -0.0509, -0.0398, -0.0298, and -0.0176.
Moreover, a new bifurcation takes place within the limits of the diagram at H=-0.0108,
where the subfamily starting at H=-0.0176 rejoins the main family. The other
families would also rejoin the main family at higher energies, yet they would reach
radii larger than
,
which was chosen as a limit for our study of a
warped disk.
Figure 3: H-p_{z}(0) phase space (w=0.005). | |
Open with DEXTER |
Figure 4: Projections over the y=0, z=0 planes and the evolution of z as a function of azimuthal angle of orbits of the families s4+ (H=-0.03), s5+ (H=-0.02) and s6- (H=-0.015). The dashed line represents the potential minimum of the warped disk at a fixed radius which is similar to the main family. | |
Open with DEXTER |
If the warp has only a slight influence on the shape of the orbits, it is important to look carefully at the behavior of the bifurcations in presence of the perturbation. Figures 5 and 8 present a magnification of the H-p_{z}(0) phase space around the first (r1) and second (r2) bifurcations. Here the periodic orbits are computed with w=0.01 in order to increase the effect of the warp. The physical correspondent would be a galaxy with a warp twice as high as that observed in the Milky Way at a radius of 30 kpc.
Except for the weak decreasing of p_{z}(0) with respect to H, the type of the first bifurcation (around H=-0.0633) remains similar to the one without perturbation (pitchfork). The two new generated families (s2+ and s2-) are both stable. They appear at the same energy as two other unstable families, namely z2+ and z2-. These are not seen in Fig. 5 because their projection merges with the p2 family. They are distinguished from it by a non zero z(0) and p_{y}(0). Projections of families s2+ and s2- are plotted in Fig. 7.
In order to estimate the importance of orbits associated with a stable periodic orbit, we have computed the surface of the section (z,p_{z}) near the bifurcation (H=-0.06). Moreover, it allows us to know the number of effective integrals of motion in this specific region of phase space. In Fig. 6 the stable p2 family is represented by the cross near the center (z=0, p_{z}=-0.0058) and is surrounded by quasi periodic families. The thinness of these curves indicates that the motion is well decoupled in z and a third integral exists. At p_{z}=0.031,z=0 and p_{z}=-0.043,z=0 we find respectively the s2+ and s2- families. The upper and lower panels show details of the section around these two families. Very thin stability islands exist. For comparison, the width in p_{z} of the separatrix (squares points) is about 1/100 times smaller than the width between families s2+ and s2-. This gives an indication of the phase space occupied by orbits associated with these families. The unstable z2+ and z2- families appear on the separatrix, at p_{z}=-0.0054,z=0.427 and p_{z}=-0.0054,z=-0.427. Outside the separatrix, the invariant curves are regular until p_{z}= 0.06 where the motion in z begins to be strongly coupled to the motion in x and y.
Figure 5: H-p_{z}(0) and H-y(0) phase space around the firstbifurcation (w=0.01). | |
Open with DEXTER |
Figure 6: Section (z,p_{z}) at H=-0.06, near the bifurcation r1. The crosses represent the three stable periodic families s2+ (top), p2 (middle) and s2- (bottom) and the two unstable sz+ (right) and sz- (left). The upper and lower panels show a zoom of the section around s2+ and s2-. In this panels the separatrix is marked with squares while in the middle it is marked with a bold line. The crosses indicate the position of the periodic families. | |
Open with DEXTER |
Figure 7: Shape of three orbits linked to the r1 bifurcation (H=-0.06). The dot indicates the starting point of the computation and the arrow shows the direction of the corresponding initial velocity. | |
Open with DEXTER |
The behavior of the second bifurcation r2 around H=-0.0509 is quite different.
Figure 8 reveals that the family s3+ is in fact the extension of the main family
p2. This arises because the perturbation favors a higher p_{z}.
The s3+ family is also stable. At H=-0.0508 the families p3 and s3- are created simultaneously. They are respectively stable and unstable. The whole bifurcation forms a pitchfork
with a symmetry breaking. The shapes of three of the four families involved in this bifurcation are presented in Fig. 10.
The section after the bifurcations is shown in Fig. 9 (H=-0.05).
The stable p3 family appears at
p_{z}=-0.009 and is surrounded by a set of quasi
periodic orbits embedded in the separatrix.
The unstable s3- family belongs to the separatrix at
p_{z}=-0.023.
The s3+ family is located at p_{z}=0.006 and its associated
orbits occupy the crescent-shaped region between the separatrix.
Outside, the invariant curves are regular until p_{z}=-0.03 where the motion in z is
not longer decoupled.
Figure 8: H-p_{z}(0) and H-y(0) phase space around the second bifurcation (w=0.01). | |
Open with DEXTER |
Figure 9: Section (z,p_{z}) at H=-0.05, after the bifurcation r2. The separatrix is marked with a bold line. The crosses indicate the position of the three periodic families s3+ stable (top), p3 stable (middle) and s3- unstable (bottom). | |
Open with DEXTER |
Our study of the bifurcation is limited to the two first bifurcations, r1 and r2. The behavior of the following bifurcations are similar to the first two. The rk bifurcation with an odd (resp. even) k is of the same type as r1 (resp. r2). This difference is related to the symmetry of the families which depends on the parity of k.
In summary, the warp does not destroy the strength of the main family, which at any H possesses a comfortable surrounding stable region. For an odd k, the effect of the warp is: (i) to change the stability of sk- subfamilies and (ii) to increase the phase space allowed to the associated sk+families. For an even k, the subfamilies are quasi unchanged, preserving a very thin phase space around the stable subfamilies .
To study the influence of the rotation, we first look for
Lagrangian points.
The latter are found by setting the right terms of Eq. (6) to zero.
Substituting p_{x}, p_{y} and p_{z},
the Lagrangian points are solutions
of the following equations:
Figure 10: Shape of three among the four orbits linked to the r2 bifurcation (H=-0.0509). The dot indicates the starting point of the computation and the arrow shows the direction of the corresponding initial velocity. | |
Open with DEXTER |
Despite the fact that a direct and retrograde rotations respectively add and remove resonances, the type of bifurcations as well as the stability and the shape of orbits are not affected by a global pattern speed. This has been tested in the range and , where is the corotation radius and a negative value corresponds to a retrograde rotation.
For a radius less than 40 kpc, both prograde and retrograde 1:1 resonances are not observable in this range of pattern speed. For a corotation of 7 kpc they appear only beyond 40 kpc (resp. 30 kpc). Thus they have not been studied.
The main influence appears when we look at the consistency of orbits with the mass density. If the warp is mostly self-gravitating and made of thin and distinct tube orbits, one can check the self-consistency constraint by noting that the spatial occupation of a periodic orbit is locally inversely proportional to its local speed ( ). The reason is explained by the fact that at a given point of the orbit, the local speed remains constant in time. This argument is not available at exceptional points, for example the points where an orbit crosses itself. Since the density is proportional to the spatial occupation time, it must also be proportional to the inverse speed along the orbit ( ). Strictly this condition is only fulfilled by a structure entirely made of distinct exactly periodic orbits, such as a disk made of circular orbits. Nevertheless the check is useful in this problem because few hot orbits far from periodic round orbits are expected to exist.
In practice, the consistency has been calculated using the indice I defined
by the norm:
We have tested the consistency of families pk with respect to the energy (Fig. 11) and radius (Fig. 12), which is more convenient in galactic dynamics. This has been calculated for different (direct and retrograde) pattern speeds between and corresponding to . The white regions correspond to I=0 (consistency) while the darker gray correspond to a value of (inconsistency). The shaded parts give the limit of the computation, either because of the corotation or because of the positive energy regions. The black pattern points out the missing data due to computational difficulty arising because of the proximity of the forbidden regions.
Figure 11: Consistency of orbit occupation with the mass density as a function of the energy and the global rotation. The dark gray corresponds to inconsistent regions while the white are consistent regions. The black parts are missing data. The shading part represents the limit of the corotation. | |
Open with DEXTER |
Figure 12: Same graph than in Fig. 11 but as a function of radius R. The upper shading part represents the limit of the corotation. While the lower is the region where H>0. The vertical resonances ( , ) are drawn in solid lines. | |
Open with DEXTER |
For a direct rotation, except in small regions near the corotation, the inverse local speed along an orbit varies exactly in opposition to the density ( ). This causes it to depopulate the higher density regions to the advantage of the lower. The density distribution is also slowly modified. However, for a slowly retrograde rotation a zone appears around 3 kpc, where the inverse local speed varies exactly as the density. This latter is also reinforced and the potential is self-consistent. This zone grows with increasing rotation and a second zone appears for larger radii. For a corotation smaller than the zones of consistency cover the whole disk under the curve of zero energy.
The same work applied to the subfamilies reveals that they are clearly inconsistent with the density distribution, but these families are also less relevant with the assumption of almost circular rotation.
The previous description of periodic orbits in a warped analytical potential allows us to draw the following conclusions.
Acknowledgements
This work as been supported by the Swiss National Science Foundation.