Orbits in Axisymmetric Potentials I

Axisymmetric potentials (oblate or prolate) are far more realistic examples to

consider in astronomy. Elliptical galaxies might well be spheroidal (but could
also be ellipsoidal), while disk galaxies almost certainly are axisymmetric
(though highly flattened).
For axisymmetric systems the coordinate system of choise are the
cylindrical coordinates (R, θ, z), and Φ = Φ(R, z).
Solving Newton’s equation of motion in cylindrical coordinates yields:

2 ∂Φ
R̈ −
 R θ̇  = − ∂R
d
dt
R2 θ̇ = 0
z̈ = − ∂Φ
∂z

The second of these expresses conservation of the component of angular

momentum about the z -axis; Lz = R2 θ̇ , while the other two equations
describe the coupled oscillations in the R and z -directions.
NOTE: For stars confined to equatorial plane z = 0, equations of motion are
identical to that of motion in spherical density distribution (not surprising,
since in this case the motion is once again central). Therefore, orbits
confined to equatorial plane are rosette orbits.
 Orbits in Axisymmetric Potentials II
As for the spherical case, we can reduce the equations of motion to

R̈ = − ∂Φ
∂R
eff
z̈ = − ∂Φ
∂z
eff

L2
with Φeff (R, z) = Φ(R, z) + z
2R2
the effective potential. The
L2z /R2 -term serves as a centrifugal barrier, only allowing orbits with
Lz = 0 near the symmetry-axis.
This allows us to reduce the 3D motion to 2D motion in Meridional Plane
(R, z), which rotates non-uniformly around the symmetry axis according to
θ̇ = Lz /R2 .
In addition to simplifying the problem, it also allows the use of
surfaces-of-section to investigate the orbital properties.
For the energy we can write
h i  
1 2 2 2 1 2 2
E= 2
Ṙ + (Rθ̇) + ż +Φ= 2
Ṙ + ż + Φeff
so that the orbit is restricted to the area in the meridional plane satisfying
E ≥ Φeff . The curve bounding this area is called the zero-velocity curve
(ZVC) (since for a point on it ~v = 0).
 Epicycle Approximation I
L2
We have defined the effective potential Φeff = Φ + z
2R2
. This has a
minimum at (R, z) = (Rg , 0), where
∂Φeff ∂Φ L2
∂R
= ∂R
− z
R3
=0
The radius R = Rg corresponds to the radius of a circular orbit with energy
1 2 L2
E = Φ(Rg , 0) + v
2 c
= Φ(Rg , 0) + z
2R2
= Φeff .
g

If we define x = R − Rg and expand Φeff around the point

(x, y) = (0, 0) in a Taylor series we obtain

Φeff = Φeff (Rg , 0) + (Φx )x + (Φy )y + (Φxy )xy + 12 (Φxx )x2 +

1 2 2 2
2
(Φ yy )y + O(xz ) + O(x z) + etc

where
 2
  2

∂Φeff ∂ Φeff ∂ Φeff


Φx = ∂x (Rg ,0)
Φxx = ∂x2
Φxy = ∂x∂y
(Rg ,0) (Rg ,0)

By definition of Rg , and by symmetry considerations, we have that

Φx = Φy = Φxy = 0
 Epicycle Approximation II
In the epicycle approximation only terms up to second order are considered:
all terms of order xz 2 , x2 z or higher are considered negligble. Defining

κ2 ≡ Φxx ν 2 ≡ Φyy
we thus have that, in the epicycle approximation,

Φeff = Φeff (Rg , 0) + 12 κ2 x2 + 12 ν 2 y 2

so that the equations of motion in the meridional plane become

ẍ = −κ2 x ÿ = −ν 2 y
Thus, the x- and y -motions are simple harmonic oscillations with the
epicycle frequency κ and the vertical frequency ν .
In addition, we have the circular frequency
q
vc (R) 1 ∂Φ

Lz
Ω(R) = R
= R ∂R (R,0)
= R2

which allows us to write

   
∂ 2 Φeff 2
2
κ = ∂R2
= R dΩ
dR
+ 4Ω 2
(Rg ,0) Rg
 Epicycle Approximation III
As we have seen before, for a realistic galactic potential Ω < κ < 2Ω,
where the limits correspond to the homogeneous mass distribution
(κ = 2Ω) and the Kepler potential (κ = Ω)
In the epicycle approximation the motion is very simple:

R(t) = A cos(κt + a) + Rg
z(t) = B cos(νt + b)
2Ω A
φ(t) = Ωg t + φ0 − κRgg sin(κt + a)

with A, B , a , b, and φ0 all constants. The φ-motion follows from

 −2  
Lz Lz x 2x
φ̇ = R2
= R2
1+ Rg
' Ωg 1 − Rg
g

Note that there are three frequencies (Ω, κ, ν) and also three isolating
integrals of motion in involution: (ER , Ez , Lz ) with ER = 1
2
(ẋ 2
+ κ 2 2
x )
and Ez = 1 2
(ż 2
+ ν 2 2
z ) . all orbits are regular.
The motion in (R, φ) can be described as retrogate motion on an ellipse (the
epicycle), whose guiding center (or epicenter) is in prograde motion around
the center of the system.
 Epicycle Approximation IV
An important question is: “When is the epicycle approximation valid?”
First consider the z -motion: The equation of motion, z̈ = −ν 2 z implies a
constant density in the z -direction. Hence, the epicycle approximation is
valid as long as ρ(z) is roughly constant. This is only approximately true
very close to equatorial plane. In general, however, epicycle approx. is poor
for motion in z -direction.
In the radial direction, we have to realize that the Taylor expansion is only
accurate sufficiently close to R = Rg . Hence, the epicycle approximation is
only valid for small librations around the guiding center; i.e., for orbits with
an angular momentum that is close to that of the corresponding circular
orbit.

Epicyclic Motion


 Orbits in Axisymmetric Potentials III
thin tube
parent ZVC




















































































































































































Typical orbit in axisymmetric potential. If orbit admits two isolating integrals

of motion, it would (ultimately) fill entire area within ZVC. Rather, orbit is
restricted to sub-area within ZVC, indicating that orbit admits a third isolating
integral of motion.
Since this is not a classical integral of motion, and we don’t know how to
express it in terms of the phase-space coordinates, it is simply called I3 .
Note that the point where orbit touches ZVC can be used to ‘label’ I3 : The
set (E, Lz , I3 ) uniquely defines an orbit.
 Orbits in Axisymmetric Potentials IV

The orbit shown on the previous page is a so-called short-axis tube orbit.
This is the main orbit family in oblate potentials, and is associated with
(parented by) the circular orbits in equatorial plane.
Orbits (c), (e) and (f) above are from the same orbit family. Orbits (a), (b) and
(d) are special in that Lz = 0.
 Orbits in Axisymmetric Potentials V
Because of the centrifugal barrier only orbits with Lz = 0 will be able to
come arbitrarily close to the center.
However, not all orbits with Lz = 0 are box orbits. There is another family of
zero-angular momentum orbits, namely the two-dimensional loop orbits (e.g.,
orbit (d) on previous page). Their meriodional plane is stationary (i.e., θ̇ = 0)
and their angular momentum vector is perpendicular to the z -axis. Hence,
I3 = L; note that [L, Lz ] = 0.
Numerous authors have investigated orbits in axisymmetric potentials using
numerical techniques. The main conclusions are:
• Most orbits in axisymmetric potentials designed to model elliptical
galaxies are regular and appear to respect an effective third integral I3 .
• The principal orbit family in oblate potentials is the short-axis tube family,
while two families of inner and outer long-axis tube orbits dominate in
prolate potentials.
• In scale-free or cusped potentials several minor orbits families become
important. These are the (boxlets) associated with resonant parents.
• The fraction of phase-space occupied by stochastic, irregular orbits is
generally (surprisingly) small.
 Orbits in Triaxial Potentials I
Consider a triaxial density distribution with the major, intermediate, and
minor axes aligned with the x, y , and z axes, respectively.
In general, triaxial galaxies have four main orbit families: box orbits, and
three tube orbits: short axis tubes, inner long-axis tubes, and outer long-axis
tubes.
Orbit structure different in cusp, core, main body, and outer part (halo).
In central core, potential is harmonic, and motion is that of a 3D harmonic
oscillator. . all orbits are box orbits, parented by stable long-axial orbit
Outside of core region, frequencies become strongly radius (energy)
dependent. There comes an energy where ωx = ωy . At this
1 : 1-resonance the y -axial orbit becomes unstable and bifurcates into
short-axis tube family (two subfamilies with opposite sense of rotation).
At even higher E the ωy : ωz = 1 : 1 resonance makes z -axial orbit
unstable → inner and outer long-axis tube families (each with two
subfamilies with opposite sense of rotation).
At even larger radii (in ‘halo’ of triaxial system) the x-axial orbit becomes
unstable . box orbits are replaced by boxlets and stochastic orbits. The
three families of tube orbits are also present
Orbits in Triaxial Potentials II
box orbit short−axis tube orbit

outer long−axis tube orbit inner long−axis tube orbit

 Orbits in Triaxial Potentials III
If center is cusped rather than cored, resonant orbits families (boxlets) and
stochastic orbits take over part of phase-space formerly held by box orbits.
The extent to which this happens depends on cusp slope.
Short-axis tubes contribute angular momentum in z -direction; Long-axis
tubes contribute angular momentum in x-direction . total angular
momentum vector may point anywhere in plane containing long and short
axes. NOTE: this can serve as kinematic signature of triaxiality.
The closed loop orbit around the intermediate y -axis is unstable . no family
of intermediate-axis tubes.
Gas moves on closed, non-intersecting orbits. The only orbits with these
properties are the stable loop orbits around x- and z -axes. Consequently,
gas and/or dust disks in triaxial galaxies can exist in xy -plane and yz -plane,
but not in xz -plane. NOTE: these disks must be ellipsoidal rather than
circular, and the velocity varies along ellipsoids.
 Stäckel Potentials I
Useful insight may be obtained from separable Stäckel models. These are the
only known triaxial potentials that are completely integrable.
In Stäckel potentials all orbits are regular and part of one of the four main
families.
Stäckel potentials are separable in ellipsoidal coordinates (λ, µ, ν)

(λ, µ, ν) are the roots for τ of

x2 y2 z2
τ +α
+ τ +β
+ τ +γ
=1

Here α < β < γ are constants and −γ ≤ ν ≤ −β ≤ µ ≤ −α ≤ λ

Surfaces of constant λ are ellipsoids
Surfaces of constant µ are hyperboloids of one sheet
Surfaces of constant ν are hyperboloids of two sheets

Stäckel potentials are of the form:

F1 (λ) F2 (µ) F3 (ν)
Φ(~
r) = Φ(λ, µ, ν) = − (λ−µ)(λ−ν) − (µ−ν)(µ−λ)
− (ν−λ)(ν−µ)

with F1 , F2 and F3 arbitrary functions

 Stäckel Potentials II
The figure below shows contours of constant (λ, µ, ν) plotted in the three
planes (from left to right) xy , xz and yz
At large distances, the ellipsoidal coordinates become close to spherical.
Near the origin they are close to cartesian. For more details, see de Zeeuw,
1985, MNRAS, 216, 273

In triaxial Stäckel potentials all three integrals (E, I2 , I3 ) are analytical, and
the orbits are confined by contours of constant ellipsoidal coordinates (see
next page).
Although Stäckel are a very special class, the fact that they are separable
makes them ideally suited to get insight. Most triaxial potentials that do not
have a Stäckel form have orbital structures that are similar to that of Stäckel
potentials.
Stäckel Potentials III
box orbits

inner long axis tube

outer long axis tube

short axis tube

 Orbits in Ellipsoid Land; Summary

System Dim Orbit Families

Oblate 3D S
Prolate 3D I +O
Triaxial 3D S+I+O+B
Elliptic Disk 2D S+B

B = box orbits
S = short-axis tubes
I = inner long-axis tubes
O = outer long-axis tubes
 Libration versus Rotation

Three-dimensional orbits
All tube orbits are build up from 2 librations and 1 rotation.
All box orbits are build up from 3 librations.
All boxlets are build up from 2 librations and 1 rotation.

Two-dimensional orbits
All loop orbits are build up from 1 libration and 1 rotation.
All box orbits are build up from 2 librations.
All boxlets are build up from 2 librations.
 Rotating Potentials I
The figures of non-axisymmetric potentials may rotate with respect to inertial
space.
The example of interest for astronomy are barred potentials, which are
rotating with a certain pattern speed.
~ p = Ωp ~
We express the pattern speed in angular velocity Ω ez
In what follows we denote by d~
a/dt the rate of change of a vector ~
a as
measured by an inertial observer, and by ~a˙ the rate of change as measured
by an observer corotating with the figure.
It is straightforward to show that
d~
a
dt
a˙ + Ω
=~ ~p ×~
a

Applying this twice to the position vector ~

r , we obtain
 
d2 ~
dt
r
2 = d
dt
r˙ + Ω
~ ~p ×~ r
r¨ + Ω
= ~ ~p ×~ r˙ + Ω~ p × d~r
dt 
r¨ + Ω
= ~ ~p ×~ r˙ + Ω~p × ~ r˙ + Ω~p ×~
r
   
r¨ + 2 Ω
= ~ ~p ×~ r˙ + Ω~p × Ω ~p ×~
r
 Rotating Potentials II
Since Newton’s laws apply to inertial frames we have that
   
r¨ = −∇Φ
~ ~ −2 Ω r˙ − Ω
~p ×~ ~p × Ω
~p ×~
r
 
r˙ represents the Coriolis force and
~p ×~
Note the two extra terms: −2 Ω
 
~p × Ω
−Ω ~p ×~
r the centrifugal force.
The energy is given by
 2
1 d~
r
E = 2 dt
+ Φ(~r)
 2
= 1 ˙+Ω
~
r ~p ×~r + Φ(~ r)
2
   2
1 ˙2 ˙· Ω ~p ×~ 1 ~
= 2
~
r + ~
r r + 2
Ωp × ~ r + Φ(~ r)
 
1 ˙2 ˙· Ω ~p ×~ 1 ~
= 2
~
r + ~
r r + 2
|Ωp × ~ r|2 + Φ(~
r)
 
= EJ + ~ r˙ · Ω~p ×~ r + |Ω ~p ×~ r |2
Where we have defined Jacobi’s Integral

r˙ 2 + Φ(~
EJ ≡ 12 ~ ~p ×~
r ) − 12 |Ω r |2
 Rotating Potentials III
The importance of EJ becomes apparent from the following:
 
dEJ
dt
r˙ dt
= ~ d
r˙ +
~ dΦ
dt
− ( ~p ×~
Ω r ) · d ~
dt
(Ω p × ~ r)
h i
r˙ ~
= ~ r¨ + (Ω~p ×~ r˙ ) + ∇Φ~ ·~ r˙ − (Ω ~p ×~ r˙ )
~p ×~
r ) · (Ω
r˙ · ~
= ~ r¨ + ~
r˙ · ∇Φ
~ − (Ω ~p ×~
r ) · (Ω~p ×~ r˙ )
~ · (A
Here we have used that A ~ × B) ~ p /dt = 0. If we
~ = 0 and that dΩ
r˙ we obtain that
multiply the equation of motion with ~
h i
r˙ · ~
~ r¨ + ~
r˙ · ∇Φ r˙ · (Ω
~ + 2~ ~p ×~ r˙ ) + ~
r˙ · Ω ~p ×~
~ p × (Ω r) = 0
r˙ · ~
⇔ ~ r¨ + ~
r˙ · ∇Φ
~ + (Ω
~p ×~ r˙ × Ω
r ) · (~ ~ p) = 0

~ · (B
Where we have used that A ~ × C)
~ =C
~ · (A
~ × B)
~ . Since
~×B
A ~ = −B
~ ×A
~ we have that
dEJ
dt
=0
The Jacobi Integral is a conserved quantity, i.e. an integral of motion.
 Rotating Potentials IV
For comparison, since Φ = Φ(t) the Hamiltonian is explictely
time-dependent; consequently, the total energy E is not a conserved
quantity (i.e., is not an integral of motion).
The angular momentum is given by
~ =~
L r× d~
r
dt
=~ r˙ + ~
r×~ ~p ×~
r × (Ω r)
This allows us to write
h i
Ω ~ = Ω
~p ·L ~ p · (~
r×~r˙ ) + Ω
~p · ~
r × (Ω~p ×~
r)
r˙ · (Ω
= ~ ~p ×~ ~p ×~
r ) + |Ω r |2
from which we obtain that

~p ·L
EJ = E − Ω ~

Thus, in a rotating, non-axisymmetric potential neither energy E not angular

~p ·L
momentum L are conserved, but the Jacobi integral EJ = E − Ω ~ is.

Note that EJ is the sum of 1 ˙ 2 + Φ, which would be the energy if the frame
~
r
2
were not rotating, and the quantity − 1 | ~p ×~
Ω r | 2
= − 1 2 2
Ω R , which can
2 2 p
be thought of as the potential energy corresponding to the centrifugal force.
 Rotating Potentials V
If we now define the effective potential

Φeff = Φ − 12 Ω2p R2

the equation of motion becomes

r¨ = −∇Φ
~ r˙ )
~b ×~
~ eff − 2(Ω

and the Jacobi integral is EJ = 1 |~

r˙ |2 + Φeff
2
An orbit with a given value for it’s Jacobi Integral is restricted in its motion to
regions in which EJ ≤ Φeff . The surface Φeff = EJ is therefore often
called the zero-velocity surface.
The effective potential has five points at which both ∂Φeff /∂x and
∂Φeff /∂y vanish. These points, L1 to L5 , are called the Lagrange Points
(cf. restricted three-body problem).
• Motion around L3 (minimum of Φeff ) always stable.
• Motion around L1 and L2 (saddle points of Φeff ) always unstable.
• Motion around L4 and L5 (maxima of Φeff ) can be stable or unstable
• depending on potential.
NOTE: stable/unstable refers to whether orbits remain close to Lagrange
points or not.
 Lagrange Points

Illustration of Lagrange points (L1 to L5 ) in Sun-Earth-Moon system.

                                    
                                    
                                    
L2
                                    
                                      
                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                    

circle) is called the region of corotation.

                      

              
L4
  


L3
L5

                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                    
                                      
Lagrange Points

L1

                                    
                                    
                                    
                                    
Illustration of Lagrange points (L1 to L5 ) in logarithmic potential. The
annulus bounded by circles through L1 , L2 and L3 , L4 (depicted as red
 Lindblad Resonances I
Let (R, θ) be the polar coordinates that are corotating with the planar
potential Φ(R, θ). If the non-axisymmetric distortions of the potential,
which has a pattern speed Ωp , is sufficiently small then we may write
Φ(R, θ) = Φ0 (R) + Φ1 (R, θ) |Φ1 /Φ0 |  1
It is useful to consider the following form for Φ1
Φ1 (R, θ) = Φp (R) cos(mθ)
where m = 2 corresponds to a (weak) bar.
In the epicycle approximation the motion in Φ0 (R) is that of an epicycle,
with frequency κ(R), around a guiding center which rotates with frequency
q
1 dΦ0
Ω(R) = R dR
.

In presence of Φ1 (R, θ), movement of guiding center is

θ0 (t) = [Ω(R) − Ωp ] t. In addition to natural frequencies Ω(R) and
κ(R) there is new frequency Ωp . Because Φ1 (R, θ) has m-fold symmetry,
guiding center at R finds itself at effectively same location in (R, θ)-plane
with frequency m [Ω(R) − Ωp ].
 Lindblad Resonances II
Motion in R-direction becomes that of harmonic oscillator of natural
frequency κ(R) that is driven by frequency m [Ω(R) − Ωp ].
At several R the natural and driving frequencies are in resonance.
(1) Corotation: Ω(R) = Ωp
(1) (Guiding center corotates with potential).
(2) Lindblad Resonances: m [Ω(R) − Ωp ] = ±κ(R)
(2) Most important of these are:
κ
Ω(R) − 2
= Ωp : Inner Lindblad Resonance
κ
Ω(R) + 2
= Ωp : Outer Lindblad Resonance
κ
Ω(R) − 4
= Ωp : Ultra Harmonic Resonance
Depending on Φ(R, θ) and Ωp one can have 0, 1, or 2 ILRs. If there are
two, we distinguish between Inner Inner Lindblad Resonance (IILR) and Outer
Inner Lindblad Resonance (OILR).
If cusp (or BH) is present there is always 1 ILR, because Ω(R) − κ(R)/2
increases monotically with decreasing R.
 Lindblad Resonances III
Frequency

Ωp
Ω+κ/2
Ω
Ω−κ/2
IILR OILR CR OLR Radius
Lindblad Resonances play important role for orbits in barred potentials.
 Lindblad Resonances IV
As an example, we discus the orbital families in a planar, rotating,
logarithmic potential

(a) Long-axial orbit → stable, oval, prograde, and oriented k to Φeff .

(x1 -family).
(b) Short-axial orbit → stable, oval, retrograde, and oriented ⊥ to Φeff .
At E > E1 (at IILR), family (b) becomes unstable and bifurcates into two
prograde loop families that are oriented perpendicular to Φeff . The stable
(unstable) family is called the x2 (x3 ) family. At the same energy the
x1 -orbits develop self-intersecting loops.
At E > E2 (at OILR) the x2 and x3 families dissapear. The x1 family looses
its self-intersecting loops.
In vicinity of corotation annulus there are families of orbits around L4 and
L5 (if these are stable).
At large radii beyond CR Ωp  Ω(R). Consequently, the orbits effectively
see a circular potential and the orbits become close to circular rosettes.