You are on page 1of 17

Home Search Collections Journals About Contact us My IOPscience

Geometric interpretation of the Frenet-Serret frame description of circular orbits in stationary

axisymmetric spacetimes

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1999 Class. Quantum Grav. 16 1333

(http://iopscience.iop.org/0264-9381/16/4/022)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 128.227.24.141
This content was downloaded on 09/05/2017 at 04:46

Please note that terms and conditions apply.

You may also be interested in:

Absolute andrelativeFrenet-Serret frames and Fermi-Walker transport


Donato Bini, Fernando de Felice and Robert T Jantzen

Circular holonomy and clock effects in stationary axisymmetric spacetimes


Donato Bini, Robert T Jantzen and Bahram Mashhoon

Gravitomagnetism and relative observer clockeffects


Donato Bini, Robert T Jantzen and Bahram Mashhoon

Circular orbits in Kerr spacetime: equatorial plane embedding diagrams


Donato Bini and Robert T Jantzen

Circular holonomy in the Taub--NUT spacetime


Donato Bini, Christian Cherubini and Robert T Jantzen

Maximal acceleration is non-rotating


Don N Page

Kerr metric, static observers and Fermi coordinates


Donato Bini, Andrea Geralico and Robert T Jantzen

Gyroscopes and gravitational waves


Donato Bini and Fernando
de Felice
Frenet--Serret formalism for null world lines
Donato Bini, Andrea Geralico and Robert T Jantzen
Class. Quantum Grav. 16 (1999) 1333–1348. Printed in the UK PII: S0264-9381(99)97330-6

Geometric interpretation of the Frenet–Serret frame


description of circular orbits in stationary axisymmetric
spacetimes

Donato Bini†‡, Robert T Jantzen‡§ and Andrea Merloni‡k


† Istituto per Applicazioni della Matematica CNR, I-80131 Naples, Italy
‡ International Center for Relativistic Astrophysics, University of Rome, I-00185 Rome, Italy
§ Department of Mathematical Sciences, Villanova University, Villanova, PA 19085, USA
k Physics Department, University of Rome, I-00185 Rome, Italy

Received 8 September 1998, in final form 13 November 1998

Abstract. A geometrical interpretation is given for the Frenet–Serret structure along constant
speed circular orbits in orthogonally transitive stationary axisymmetric spacetimes. This gives a
simple visualization of the acceleration of these orbits and of the Fermi–Walker angular velocity of
the usual symmetry-adapted frame vectors along them and provides an elegant description of the
various observers characterized by critical values of the variables parametrizing the acceleration.

PACS number: 0420C

1. Introduction

Iyer and Vishveshwara [1] have introduced the rather involved computational formulae
necessary for explicit evaluation of the Frenet–Serret frame {eα } and its associated curvature κ
and torsions (τ1 , τ2 ) [2, 3] for arbitrary constant speed circular orbits in orthogonally transitive
stationary axisymmetric spacetimes, in general, and in the Kerr spacetime in Boyer–Lindquist
coordinates (t, r, θ, φ), in particular. When studied from a geometrical perspective, the content
of these formulae is seen to admit an elegant visual interpretation of the dependence of this
structure on the speed through pseudo-polar and polar coordinates, respectively, on the mutually
orthogonal (t, φ) and (r, θ) subspaces of the tangent space, the 2 + 2 splitting associated with
the orthogonal transitivity condition.
The rapidity (hyperbolic) angle α of the (unit) 4-velocity U in the first plane directly
parametrizes both e0 = U and e2 , related to each other by the α-derivative, while the 4-
acceleration a = κe1 in the second (‘acceleration’) plane is parametrized by its polar coordinate
decomposition with respect to the radial direction in terms of its magnitude κ (the curvature) and
polar angle χ, each in turn parametrized by α. Along the latter α-parametrized ‘acceleration
curve’ in the acceleration plane for fixed (r, θ ), the two torsions modulo signs are half the
respective α-derivatives of the arc lengths in the κ and χ directions, while the χ-derivative of
e0 yields −e3 . Thus critical values of the polar acceleration parameters lead to zeros of the
corresponding torsions, which in turn aligns the Frenet–Serret angular velocity with one of
the two Frenet–Serret frame vectors in that plane (parallel or perpendicular to the acceleration
vector).
0264-9381/99/041333+16$19.50 © 1999 IOP Publishing Ltd 1333
1334 D Bini et al

The properties of these special orbits can be interpreted in terms of the relationships of the
normal, tangent and position vector of the α-parametrized acceleration curve in the acceleration
plane. For example, the Frenet–Serret angular velocity itself is always perpendicular to
the tangent to the acceleration curve within its plane, which is therefore either parallel or
perpendicular to the acceleration at the critical points of the polar acceleration parameters.
Thus ironically the Frenet–Serret properties of the curve in the tangent space traced out by the
acceleration vectors of this family of circular orbits are intimately connected to the Frenet–
Serret properties of the original orbits in spacetime.
These particularly simple relationships depend on the fact that these orbits undergo purely
transverse relative motion [4] with respect to the static Killing vector field observers and have
constant magnitude acceleration, leading to a stationary orthogonal 2 + 2 decomposition of the
one-parameter family of constant speed Frenet–Serret frame vectors for fixed (r, θ ), allowing a
simple parametrization in each 2-plane. Finding these relationships was made possible only by
considering the parametrization by the local physical speed measured by the static observers
rather than by the angular velocity variable and calculations are considerably simplified by
using the corresponding geometric rapidity variable.
The class of circular orbits is important because of the numerous distinct families of
stationary circularly rotating test observers whose motion is characterized by various geometric
properties. These include the usual shear-free static observers (nonrotating at infinity) and the
hypersurface-forming zero-angular-momentum observers (the locally nonrotating ZAMOs),
which together determine the most useful coordinate system for black holes, the Carter
observers [5–7], which are the key to separability of the geodesic and test field equations,
and those observers associated with extremal properties of the acceleration (see, for example
[8–11]). It is also the most natural family of curves to fully understand from the point of view
of the differential geometry of curves in general relativity. Considering the enormous body
of literature discussing properties associated with these orbits, it is surprising that this has not
been done sooner. Furthermore, the relative observer decomposition of the acceleration vector
provides very useful insight into its behaviour, which depends on the interplay of the centripetal
acceleration and space curvature and gravitoelectromagnetic effects. These are described here
explicitly for the black hole case.

2. Frenet–Serret formulae

The spacetime Frenet–Serret frame along a single timelike test particle worldline with 4-
velocity U = e0 parametrized by the proper time τ is described by the equations [2, 3]
D D
e0 = κe1 , e2 = −τ1 e1 + τ2 e3 ,
dτ dτ (2.1)
D D
e1 = κe0 + τ1 e2 , e3 = −τ2 e2 .
dτ dτ
The curvature κ is allowed to be of either sign in order that the frame may be extended
smoothly through isolated points where DU/dτ = 0. Its absolute value is the magnitude of
the acceleration a = κe1 and describes the rate of change of the rapidity parameter of the boost
of the Frenet–Serret frame relative to one parallel transported along the worldline. The first
and second torsions τ1 and τ2 are the components of the Frenet–Serret angular velocity vector
ω(FS) = τ2 e1 + τ1 e3 (2.2)
of the spatial frame {ea } with respect to one Fermi–Walker transported along the curve, provided
that {ea } is a right-handed frame [1]. Its sign-reversal (gyro) = −ω(FS) therefore describes
Geometric interpretation of Frenet–Serret frames for circular orbits 1335

the precession angular velocity with respect to the Frenet–Serret frame of the Fermi–Walker-
transported spin vector of a test gyroscope moving along the worldline of U [1].

3. Circular orbits

Since the magnitude of the acceleration is constant for constant speed circular orbits in
stationary axisymmetric spacetimes, the first torsion describes the Fermi–Walker angular
velocity of the acceleration vector itself, namely the projection of Da/dτ orthogonal to U
yielding the Fermi–Walker derivative of a along U
D(fw) a/dτ = κτ1 e2 . (3.1)
Consider the form of the Boyer–Lindquist line element for the Kerr spacetime
orthogonalized with respect to the Killing vector ∂t associated with its stationary symmetry
ds 2 = −M 2 (dt − Mφ dφ)2 + γφφ dφ 2 + grr dr 2 + gθ θ dθ 2 , (3.2)
corresponding to the orthonormal frame (m, ∂r̂ , ∂θ̂ , φ̂ ) obtained by normalizing three
orthogonal coordinate frame vectors (normalization is indicated by a hat) and orthogonalizing
−1/2
and normalizing the remaining one ∂φ to φ̂ = γφφ (∂φ + Mφ ∂t ), where m = ∂tˆ = M −1 ∂t
is the 4-velocity of the static observers. It is assumed that the black hole angular momentum
parameter a satisfies a > 0 so that the black hole is rotating in the positive φ-direction when
a 6= 0. The context should always clearly distinguish the symbol for the acceleration from the
traditional angular momentum parameter a of the Kerr metric. The dimensionless parameter
ā = a/M and radial variable r̄ = r/M are also useful, where M is the black hole mass
parameter.
Stationary axisymmetric quantities depend only on (r, θ ), which are simplest to visualize
by representing them graphically in the ρ–z coordinate plane of the corresponding flat
spacetime cylindrical coordinates (ρ, z) = (r sin θ, r cos θ ). The ±ρ coordinate directions
will be referred to as ‘horizontally outward’ and ‘horizontally inward’ respectively away from
and towards the ‘vertical’ symmetry axis ρ = 0. Similarly the z coordinate direction will be
referred to as ‘vertical’, either away from or towards the equatorial plane z = 0 or θ = π/2.
Worldlines which correspond to circular orbits with constant speed ν (with respect to the
static observers, when they are timelike) have 4-velocity
e0 = U = 0(∂t + ζ ∂φ ) = γ (m + νφ̂ ) = cosh αm + sinh αφ̂ , (3.3)
where the relative velocity νφ̂ has constant signed magnitude ν = tanh α ∈ (−1, 1)
parametrized by the rapidity α ∈ (−∞, ∞) (note the useful formula dα/dν = γ 2 ) and is
related to the constant angular velocity ζ by
1/2 1/2
γφφ ζ dν Mγφφ
ν= , γ2 = . (3.4)
M(1 − Mφ ζ ) dζ 1/2
M 2 (1 − Mφ ζ )2 − γφφ ζ 2
Their acceleration a = DU/dτ is given by
a = κ(cos χ ∂r̂ + sin χ ∂θ̂ ), (3.5)
where (κ, χ) are polar coordinates in the ∂r̂ –∂θ̂ (acceleration) plane in the tangent space, namely
the signed magnitude κ of the acceleration and its orientation angle χ satisfying
 2 2 1/2
|κ| = a θ̂ + a r̂ , tan χ = a θ̂ /a r̂ , (3.6)
both constant for a given orbit. If we choose the convention κ > 0, χ ∈ [−π, π], the direction
of the acceleration is
e1 = cos χ ∂r̂ + sin χ∂θ̂ . (3.7)
1336 D Bini et al

The ‘horizontal’ directions ±∂ρ̂ correspond to χ = χ± = ±π/2 − θ , with (+) outward and
(−) inward with respect to the symmetry axis.
Since U is a one-parameter family of unit vectors, its α-derivative must be orthogonal to
it and in fact defines the next Frenet–Serret vector
e2 = dU/dα = sinh αm + cosh αφ̂ . (3.8)
Similarly e1 is a one-parameter family of unit vectors, so its χ-derivative must be orthogonal
to it and defines the last vector
e3 = −de1 /dχ = sin χ ∂r̂ − cos χ∂θ̂ , (3.9)

which is the clockwise rotation of e1 by 90 in the ∂r –∂θ plane, chosen so that the spatial frame
is right-handed.
As α varies from −∞ to ∞, or ν from −1 to 1, the 4-velocity U traces out one branch
of a hyperbola (pseudocircle) in its plane (the relative observer plane of m and U ), while
the acceleration vector a traces out one branch of a hyperbola in the orthogonal transverse
relative acceleration plane (transverse to the relative motion of m and U ) that depends on the
location of the tangent space, i.e. on r and θ . The second branch of this hyperbola corresponds
to superluminal motion (a spacelike circular orbit) with ν −1 ∈ (−1, 1), which includes the
closed φ-coordinate circles.
The line element in the acceleration plane
2
ds(a) = dκ 2 + κ 2 dχ 2 (3.10)
measures the arc length along this curve of acceleration vectors in the tangent space. With
the above choices for e1 and e3 , the two torsions must be allowed to change sign. They are
then simply equal to half the sign-reversed rate of change of the arc length along the two polar
coordinate directions with respect to α
τ1 = − 21 dκ/dα, τ2 = − 21 κ dχ/dα. (3.11)
The tangent to the acceleration hyperbola and consequently the Frenet–Serret angular
velocity (2.2) can therefore be expressed as
da dκ de1 1 da
= e1 + κ = 2(−τ1 e1 + τ2 e3 ), ω(FS) = e2 ×U . (3.12)
dα dα dα 2 dα
Thus ω(FS) is obtained from 21 da/dα by a counterclockwise rotation of 90◦ with respect to the
ordered axes (e3 , e1 ), which have the same orientation as (∂r̂ , ∂θ̂ ), placing it along the inward
normal to the hyperbola as will be seen below. Its magnitude
1/2  1/2
kω(FS) k = τ12 + τ22 = 21 (dκ/dα)2 + κ 2 (dχ/dα)2 = 21 |ds(a) /dα| (3.13)
is just half the speed with which the acceleration curve is traced out in its plane with respect to
the α parametrization. The Frenet–Serret angular velocity traces out a curve in the acceleration
plane which is also a hyperbola with its asymptotes rotated by 90◦ , but lying in a complementary
branch (the opening angles of the two hyperbolas sum to π).
The derivation of these simple torsion formulae (3.11) on which all of this beautiful
geometry rests starts from equations (49) and (54) of [1] which determine the polar variables
κ and χ. Using the resulting formulae to evaluate the derivatives (3.11) with the relation
B = 21 ∂ A/∂ω following from equations (46) and (47) in their notation (their ω is our ζ )
and the further relation for the parameter derivative dζ /dα = γ −2 dν/dζ , interchanging the
partial derivatives with the parameter derivative, one finds their formulae (50) and (51). These
formulae describe the entire orthogonally transitive stationary axisymmetric symmetry class,
not just the Kerr metric to which they are later applied.
Geometric interpretation of Frenet–Serret frames for circular orbits 1337

For a given orbit the Frenet–Serret triad is locked to the static observer frame in the sense
that they differ only by the fixed relative observer boost between m and U along the direction of
relative motion and a fixed rotation in the orthogonal 2-plane. The Frenet–Serret axes therefore
rotate with respect to the nonrotating observers at infinity with the angular velocity associated
with the orbital motion around the symmetry axis. 
By using the ZAMOs whose 4-velocity field n = N −1 ∂t − N φ ∂φ is the unit normal to
the Boyer–Lindquist time slicing of the Kerr spacetime, one can investigate the properties of
the acceleration all the way to the horizon, while the static observers become spacelike within
the ergosphere. The lapse and shift may be read from the Boyer–Lindquist line element for
Kerr orthogonalized with respect to the normal vector field n
2
ds 2 = −N 2 dt 2 + gφφ dφ + N φ dt + grr dr 2 + gθ θ dθ 2 , (3.14)
corresponding to the orthonormal frame (n, ∂r̂ , ∂θ̂ , ∂φ̂ ) obtained by normalizing the three
orthogonal spatial coordinate frame vectors and orthogonalizing and normalizing the time
coordinate frame vector. The ZAMO relative velocity (or equivalently the ZAMO rapidity)
parametrizes the 4-velocity and the angular velocity of the circular orbit
U = γ(n) (n + ν(n) ∂φ̂ ) = cosh α(n) (n + tanh α(n) ∂φ̂ ), ζ = −N φ + Ngφφ −1/2 ν(n) . (3.15)
Since m and n differ by a boost (outside the ergosphere where they are both timelike)
which depends only on r and θ which are constant along an orbit, the corresponding rapidity
parameters only differ by a constant α = α(n) + α(n,m) on a given orbit, so the two rapidity
derivatives are equal. The α(n) -derivative continues to be valid inside the ergosphere where
the static observer 4-velocity m is spacelike and so the timelike orbits are described by values
ν −1 ∈ (−1, 1) and the parametrization by α with hyperbolic sine and cosine interchanged
is valid. This latter α-derivative must be understood inside the ergosphere. However, in
the following sections all rapidities and relative velocities will be taken with respect to the
ZAMOs, using the unsubscripted symbols to represent these quantities. Furthermore, co-
rotating circular orbits (ν > 0) and counter-rotating orbits (ν < 0) in this context will refer to
the ZAMOs as nonrotating (ν = 0), corresponding to Wilkins’ terminology of co-revolving
and counter-revolving orbits [12], based on the sign of the conserved angular momentum U ·∂φ
associated with the axisymmetry.

4. Extremely accelerated observers

‘Extremely accelerated observers’ [8–11] are those whose worldlines are constant speed
circular orbits for which the magnitude of the acceleration is extremized by the value of
the speed. These occur when dκ/dα = 0, which implies τ1 = 0 and hence D(fw) a/dτ = 0
by equation (3.1), so that the acceleration is Fermi–Walker transported along the worldline.
In other words ‘extremal acceleration is nonrotating’, first shown for Kerr by de Felice [13]
for the equatorial plane case and extended to general circular orbits and arbitrary stationary
axisymmetry by Page [9] with an alternative derivation later by Semerák [10]. The same
remarks apply to critical points of κ which are not extremal.
Alternatively, the vanishing of the first torsion aligns the Frenet–Serret angular velocity
with the acceleration direction
τ1 = 0 → ω(FS) = τ2 e1 , (4.1)
so this direction naturally remains fixed with respect to Fermi–Walker transport. When the
second torsion is also zero, as occurs identically in the equatorial plane of Kerr and for all
circular orbits in the case of additional cylindrical symmetry, the Frenet–Serret frame is Fermi–
Walker transported along the worldline of U , so the spin vector of a test gyroscope moving
1338 D Bini et al

along this worldline is locked to the static observer frame, or ‘phase-locked’ in the terminology
of de Felice [13], thus rotating with exactly the angular velocity of the orbital motion.
The second torsion vanishes at a critical point of the acceleration angle χ where
τ2 = 0 → ω(FS) = τ1 e3 , (4.2)
so the Frenet–Serret angular velocity is orthogonal to the acceleration in the acceleration plane,
i.e. along the normal to the acceleration hyperbola at that value of the acceleration.
Inflection points of κ with respect to the speed are critical points of the first torsion
0 = d2 κ/dα = −2dτ1 /dα. (4.3)
In the equatorial plane of Kerr such inflection points occur in the zone between the outer
counter-rotating circular geodesic photon orbit and inner co-rotating one (region B in the
terminology of [8]). These turn out to be minima of the absolute value of the first torsion,
described as the spin-critical orbits in [8]. Such inflection points also occur off the equatorial
plane.

5. Geometry of the transverse relative acceleration plane

The orthonormal components of the parametrized acceleration hyperbola in the ∂r̂ –∂θ̂ plane
have the fractional quadratic form
 
a r̂ = F ν; k r̂ , ν+(r) , ν−(r) , a θ̂ = F ν; k θ̂ , ν+(θ ) , ν−(θ ) , (5.1)
where ν = tanh α ∈ (−1, 1) and
(ν − ν− )(ν − ν+ )
F (ν; k, ν+ , ν− ) = k
1 − ν2
 
= k sinh α − (ν− + ν+ ) sinh α cosh α + ν− ν+ cosh2 α
2

= 21 k[(ν− ν+ + 1) cosh 2α − (ν− + ν+ ) sinh 2α + (ν− ν+ − 1)]. (5.2)


With the natural rapidity parametrization, the acceleration curve takes the matrix form
 r̂      
a A B cosh 2α E
= + . (5.3)
a θ̂ C D sinh 2α F
Solving this for cosh 2α, sinh 2α and using the identity cosh2 2α − sinh2 2α = 1, one
immediately finds the explicit form for the unparametrized equation of a general hyperbola
with centre [E, F ] displaced from the origin
    −1  r̂ 
  A B −1 T 1 0 A B a −E
a −E a −F
r̂ θ̂ = 1. (5.4)
C D 0 −1 C D a θ̂ − F
A change of observer corresponds to a translation in α which in turn leads to an SL(2, R)
transformation of the matrix of coefficients.
The Frenet–Serret angular velocity vector ω(FS) is obtained from the acceleration vector by
half the α-derivative, which interchanges cosh 2α and sinh 2α, and a counterclockwise rotation
by π/2
 ωr̂      
(FS) 0 −1 A B 0 1 cosh 2α
= . (5.5)
ωθ̂ 1 0 C D 1 0 sinh 2α
(FS)

This is a hyperbola centred at the origin with complementary asymptotes to the acceleration
hyperbola rotated by 90◦ . The minimum value of the magnitude kω(FS) k clearly occurs at
Geometric interpretation of Frenet–Serret frames for circular orbits 1339

its vertex, which corresponds to the same relative velocity ν(vert) describing the vertex of the
acceleration hyperbola.
The acceleration hyperbola is characterized by the zeros and critical points of the

components a r̂ , a θ̂ individually and by the critical points of the polar acceleration variables.
The critical points of the function F alone are the zeros of a function of the same type as F
but with different parameters, also leading to a quadratic condition as discussed in [8]. This
discussion may be extended to the polar acceleration variables.
This general structure applies to the timelike circular orbits in any orthogonally transitive
stationary axisymmetric spacetime. For a specific spacetime, one may study the particular
functions of r and θ which determine the behaviour of the acceleration and Frenet–Serret
hyperbolas as a function of position. The remaining discussion of this section holds for the
Kerr spacetime.
The two roots ν±(r) of a r̂ are always real and of opposite sign and describe the circular orbits
with vanishing radial acceleration, reducing to geodesics in the equatorial plane where a θ̂ is
identically zero. Both velocities are subluminal at sufficiently large radius, leading to a pair of
orbits with vanishing radial acceleration, but as one approaches the black hole first the counter-
rotating velocity ν−(r) and then also the co-rotating velocity ν+(r) become superluminal, defining
three regions A (both subluminal, far from the hole), B (one subluminal, one superluminal), and
C (both superluminal, close to the hole) exactly as discussed for the equatorial plane geodesics
in [8] to which this discussion reduces in the equatorial plane case. In region C, the acceleration
is always radially outward (a r̂ > 0), while in region B the ultrarelativistic counter-rotating
acceleration is radially outward but the ultrarelativistic co-rotating acceleration is inward.
The interface radii rAB (θ) and rBC (θ ) separating these three regions occur, respectively,
where the counter-rotating and co-rotating photon orbits have zero radial acceleration (namely,
ν−(r) = −1 and ν+(r) = 1). The middle region B collapses to the sphere r̄ = 3 in the
Schwarzschild limit ā = 0. Figure 1 illustrates these three regions in the ρ–z coordinate
plane for ā = 0, 0.5, 1. In the extreme case ā = 1, the curve separating region C and B
bifurcates on the horizon. The details are discussed elsewhere [14].
The outward (+) and inward (−) horizontal directions along ±∂φ̂ in the ∂r̂ –∂θ̂ plane
correspond to the values χ± = ±π/2 − θ , while χ ∈ (−π/2, π/2) corresponds to outward
acceleration a r̂ > 0 with respect to the radial direction, and the complementary half-plane to
inward such acceleration. Our convention that κ > 0 means that as r → ∞, then χ ≈ χ− for
sufficiently large ν in region A, i.e. the acceleration is horizontally inward toward the symmetry
axis far from the black hole at a high enough speed, corresponding to the flat spacetime inward

Figure 1. The profiles of the regions A, B and C for a black hole with (a) ā = 0, (b) ā = 0.5 and
(c) ā = 1 represented in the ρ–z coordinate plane (horizontal axis ρ, vertical axis z). In the ā = 0
limit (see figure 3(a)), region B collapses to the sphere r̄ = 3, while the ergosphere collapses to
the horizon r̄ = 2. In the ā = 1 limit (see figure 3(c)), region C collapses to the horizon r̄ = 1 for
angles greater than the angle θ∗ ' 47◦ , but continues to exist at smaller angles.
1340 D Bini et al

centripetal acceleration. As one approaches the black hole, the ultrarelativistic acceleration
vectors tilt further up, rotating away from the symmetry axis towards the outward radial
direction to resist the increasingly strong attraction toward the hole.
The roots ν±(θ) of the a θ̂ are always complex, so a θ̂ is always of the same sign and never
zero (except in the equatorial plane where it is identically zero). Above the equatorial plane
where θ ∈ (0, π/2), a θ̂ and hence χ are always negative, leading to a single relative extremum
χ(ext) ∈ (0, −π/2) of the acceleration angle χ which is a minimum of its absolute value (the
minimum tilt upward from the outward radial direction), where the second torsion is zero and
the Fermi–Walker rotation is orthogonal to the acceleration vector. This extremum occurs
where the position vector in the acceleration plane is tangent to the acceleration hyperbola,
so the normal to the curve gives the direction of the Frenet–Serret angular velocity. Above
the equatorial plane the angle χ satisfies χ− < χ < 0 along the acceleration curve, so the
acceleration vector is always tilted upward with respect to the equatorial plane.
Figure 2 shows the acceleration hyperbola for regions A, B and C, and its geometry in the
acceleration plane at θ = π/3, a typical angle above the equatorial plane. The figures must
be reflected across the broken line and rotated by 30◦ to make that line horizontal to orient
them correctly in the ρ–z plane above the equatorial plane. The centre of the acceleration
hyperbola always lies in the first quadrant in the acceleration plane, while the hyperbola itself
lies below the radial acceleration axis. The asymptote of this lower branch which is to the
right corresponds to ν → −1, while the one to the left corresponds to ν → 1, which reflects
the fact that counter-rotating orbits require a stronger radial acceleration to resist the increased
gravitational attraction due to the contribution from the gravitomagnetic force, in the relative
observer description of [15] (see figure 2 therein). The directions of the two asymptotes
themselves are the directions of the acceleration vectors of the counter and co-rotating photon
orbits.
Figure 3 illustrates the behaviour of κ as a function of ν for selected radii in the three
regions A, B, C. In most of region A extending out to infinity, there are three local extrema
of κ where the position vector in the 2-plane of the acceleration hyperbola is orthogonal
to its tangent line (see figure 2(a)): two are local minima, one closest to the origin (the
global minimum) and one behind it on the hyperbola, and one is the local maximum in
(κ) (κ)
between them. These occur at the ZAMO relative velocities ν−1 < ν−2 < 0 < ν+(κ) (local
minimum, local maximum, global minimum, respectively). In each case the Frenet–Serret
angular velocity is along the position vector direction (the acceleration vector). There is a
single local maximum of the negative angle χ where the position vector of the acceleration
hyperbola is tangent to the hyperbola, and the Frenet–Serret angular velocity is along the
normal to the hyperbola. This angle in absolute value is the minimum acute angle that the
acceleration vector is tilted upward from the radial direction above the equatorial plane. This
extremum occurs at ν (χ) satisfying ν (χ) < 0 < ν+(κ) , an inequality which shows that the ZAMOs
at ν = 0 must lie somewhere between these two extrema on the acceleration curve. The
horizontal tangent line to this curve in the acceleration
plane occurs at the absolute minimum
of the absolute value acceleration component a θ̂ , which always occurs at a small positive
velocity. Thus the ZAMOs must lie between the location of the horizontal tangent line and
the location of the tangent line which is aligned with the position vector on the acceleration
hyperbola.
As one moves closer to the black hole at a fixed angle θ above the equatorial plane
(figures 2(b) and (c)), the acceleration hyperbola symmetry semi-axis rotates towards the
outward radial direction as more and more radial acceleration is required to resist falling
into the hole. At the interface with region B, one of the asymptotes of the acceleration
Geometric interpretation of Frenet–Serret frames for circular orbits 1341

Figure 2. The acceleration hyperbolas in the tangent space (horizontal axis along ∂r̂ , vertical axis
along ∂θ̂ ) for regions A, B and C for ā = 0.5, shown at a typical angle θ = π/3 at radii (a)
r̄ = 4 (region A), (b) r̄ = 3 (region B), (c) r̄ = 2.25 (region C). The broken line represents the
‘horizontal’ direction in the ρ–z plane, with the symmetry axis to the left along it at the polar angle
χ− = −5π/6, and the equatorial plane above it. Figure 5 shows the orientation of these diagrams
in the ρ–z plane for the same angle.
1342 D Bini et al

Figure 3. The magnitude κ of the acceleration versus ZAMO relative velocity ν ∈ (−1, 1) for
selected radii in regions A (lower curves), B (intermediate curves) and C (upper curves) at a typical
angle θ = π/3 for ā = 0.5.

hyperbola goes vertical (i.e. parallel to ∂θ̂ ), reducing the two zeros of the radial acceleration
to one. At the interface with region C, the remaining asymptote of the hyperbola goes
vertical, so that no zeros of the radial acceleration occur within the region C as it continues
rotating away from the axis of zero radial acceleration. Far from the black hole in region A
there are two local minima separated by a local maximum. As one approaches the hole
within this region, the two local extrema of κ which are not the global minimum (counter-
rotating with respect to n) coalesce at a point of inflection of κ at a point near where
the acceleration hyperbola has a vertical tangent and then disappear, leaving only a single
global minimum of κ and a single global minimum of |χ| which continue to exist up to the
horizon.
The angular opening of the pair of asymptotes of the acceleration hyperbola increases
with ā and decreases to zero when ā = 0. In the limiting case ā = 0 of the Schwarzschild
spacetime, the branch of the acceleration hyperbola degenerates to a half-line terminating on
the axis of purely radial acceleration at ν = 0, traced once for each sign of ν. This half-line
intersects the zero radial acceleration axis in a single point in region A but as one moves towards
the black hole, this half-line rotates away from the symmetry axis towards the outward radial
direction, aligning itself parallel to the zero radial acceleration axis at the interface of region A
with region C (region B shrinks to a surface), beyond which it is tilted away from that axis and
thus has no intersection. The angle of tilt reveals the competition between the outward radial
acceleration necessary to resist falling into the hole and the centripetal acceleration towards
Geometric interpretation of Frenet–Serret frames for circular orbits 1343

the symmetry axis perpendicular to the circular orbit.


Figure 4 illustrates the velocities of the various special circular orbits with respect to
the ZAMOs at θ = π/3, π/2.1, π/2.01 for ā = 0.5, r̄ = 4. (Compare with the equatorial
plane limit shown in figure 7(b) of [8], where the spin-critical velocity curve in region B
should be corrected to connect the extremely accelerated observers in regions A and C.) As
one approaches the equatorial plane θ = π/2 from above in regions A and B, the co-rotating
extremal acceleration velocity ν+(κ) moves upward to approach the zero radial acceleration value
ν+(r) , which in turn converges to the co-rotating geodesic velocity ν(geo)+ in the equatorial plane.
(κ)
Similarly in region A alone, of the two counter-rotating extremal acceleration velocities ν−1 ,
(κ)
ν−2 shown only in figure 4(a), the lower one converges to the zero radial acceleration value
ν−(r) which in turn converges to the counter-rotating geodesic velocity ν(geo)− in the equatorial
plane, while the upper one converges to the counter-rotating extremal acceleration velocity
ν(ext) .

6. The relative observer decomposition of the acceleration

The qualitative behaviour of the acceleration hyperbola as a function of position in the r–


θ plane is easily understood in terms of the geometry of the relative motion as seen by the
ZAMOs [16, 17]. In the nonrotating Schwarzschild case, it reflects the competition between the
(unique) relative centripetal acceleration of the circular motion, tilted away from the equatorial
plane by the spatial geometry, and the outward acceleration necessary to resist falling into the
hole, scaled by a proper time dilation factor. Turning on the rotation of the hole then breaks
the symmetry between co-rotating and counter-rotating orbits, spreading apart the asymptotes
from the symmetry axis of the degenerate acceleration half-line hyperbola branch, tilting the
counter-rotating orbit half further from the black hole symmetry axis and the co-rotating half
closer towards it due to an increased/decreased radial acceleration term, while moving the
vertex slightly away from the equatorial plane off the ∂r̂ axis.
Since the circular orbit motion is a case of purely transverse relative acceleration,
the acceleration vector already lies in the observer local rest space and so may itself be
directly represented by its relative observer decomposition (given by a(U ) = γ A(U, u) in
equation (9.3) of [17] with u = n). Since the acceleration lies in the ∂r̂ –∂θ̂ plane, it is
convenient to introduce the orthonormal component 2-vector notation X E = (Xr̂ , Xθ̂ ). For the
ZAMO decomposition, the two nonzero acceleration components have the form
aE = kE sinh2 α + 2θE sinh α cosh α + AE cosh2 α
φ̂

= (kE + A)
E sinh α + 2θE sinh α cosh α + AE
2
φ̂

= 21 (kE + A)
E cosh 2α + θE sinh 2α + 1 (−kE + A).
φ̂ 2
E (6.1)
The first line is its relative observer decomposition into a relative centripetal acceleration term
involving the relative Lie curvature vector kE of the φ coordinate circles in the slicing geometry
of the ZAMOs (see equation (13.3) of [16]), a gravitomagnetic term involving the observer
expansion tensor θ α β (let θEφ̂ = (θ r̂ φ̂ , θ θ̂ φ̂ )), and a gravitoelectric term involving the observer
acceleration A. E The second line is the corresponding optical metric decomposition, where

kE + AE = N k is along the optical Lie relative curvature vector of the φ-coordinate circles in
−1

the optical slicing geometry of the ZAMOs (see equation (13.4) of [16] and also [18]) and the
remaining two terms are the optical gravitomagnetic and gravitoelectric terms. The Lie relative
Frenet–Serret frame is {∂φ̂ , k̂, ∂φ̂ ×n k̂}, where k̂ is the unit vector direction of kE [19]. For the
static observer decomposition, the ZAMO gravitomagnetic expansion vector θEφ̂ ∼ θ α β ν̂ β term
1344 D Bini et al

Figure 4. The ZAMO relative velocities ν of the various special circular orbits plotted versus the
radius r for the angles (a) θ = π/3, (b) θ = π/2.1, (c) θ = π/2.01. The upper/lower thick broken
curves are the velocities ν(car) , ν(m) of the Carter and static observers, while the horizontal axis
(r)
ν = 0 is the ZAMO relative velocity. The thick full outer curves ν± correspond to zero radial
(κ) (κ)
acceleration. The thin full curves are the extremal acceleration velocities ν−1 , ν−2 , ν+(κ) . The
thin broken line crossing the horizontal axis in region B is the hyperbola vertex velocity ν(vert)
minimizing kω(FS) k. The nearly indistinguishable thin broken line slightly under the horizontal
axis is the extremal acceleration angle velocity ν (χ ) . The extremal acceleration curves are not
shown in (b) and (c) due to a numerical root finding instability.
Geometric interpretation of Frenet–Serret frames for circular orbits 1345

is instead replaced by the gravitomagnetic cross product ν̂ ×m HE , where HE = 2ω is twice the


vorticity of the static observer 4-velocity m and ν̂ is the unit relative velocity with respect to
the relevant observer.
The diagonals of the two parallelograms formed by the vectors ±θEφ̂ and (kE + A)/2E thought
of as tangent to the hyperbola centre at (−kE + A)/2
E determine the directions of the asymptotes
of the hyperbola, namely (kE + A)/2
E ± θE . By extremizing the distance between the acceleration
φ̂
hyperbola and its centre, one easily finds its vertex rapidity α(vert) to satisfy the condition
  
E · θE 1 (kE + A)
tanh 4α(vert) = −2 21 (kE + A) E 2 + kθE k2 , (6.2)
φ̂ 2 φ̂

which also determines the vertex of the closely related Frenet–Serret angular velocity hyperbola
at which kω(FS) k is minimized. The corresponding relative velocity ν(vert) = tanh α(vert) is
shown in figure 4, counter-rotating until very near the horizon where it becomes co-rotating.
As one approaches the equatorial plane, this single velocity curve converges to the spin-
critical velocity in region B (zero-derivative critical value for the spin precession magnitude,
see [8]) and to the extremal acceleration velocity curves in regions A and C (nondifferentiable
critical value for the spin precession magnitude, which vanishes for the extremal acceleration
observers). These three limiting curves are connected at cusps at the photon geodesics
determining the interfaces between the regions A, B and B, C. The radial components of
the centripetal acceleration and the total spatial gravitational force exactly balance at these
interfaces for the counter-rotating and co-rotating photon orbits, respectively.
To describe accelerated photon circular orbits, one can choose the limits ν → ±1 of
γ −1 U and γ −2 a (corresponding to a unit energy affine parameter [16]), yielding, respectively
the 4-momenta P± = n ± ∂φ̂ and the corresponding acceleration vectors

aE± = kE ± 2θEφ̂ + A,
E (6.3)
where the +/− sign refers to the co-rotating/counter-rotating photon orbits. These two vectors
give the directions of the asymptotes of the timelike circular orbit acceleration hyperbola,
already found directly above apart from a factor of 2. It is worth noting that these vectors can
be expressed in the form aE± = ∓(g̃φφ )1/2 ∇ζ± , where ζ± = −N φ ± (g̃φφ )−1/2 are the photon
angular velocities (see equation (3.15)), g̃φφ = N −2 gφφ is the corresponding optical metric
component, and ∇ = (∂r̂ , ∂θ̂ ). Thus their lines of force can be represented as orthogonal
trajectories to these potential functions. One can also express the vertex value of the rapidity
in the equivalent form α(vert) = 41 ln(kE
a− k/kE
a+ k).
For the Schwarzschild case, the expansion tensor is zero and the nonzero orthonormal
components of the acceleration vector are
E sinh2 α + AE = 1 (kE + A)
aE = (kE + A) 2
E cosh 2α + 1 (−kE + A),
2
E (6.4)
where

kE = −(r sin θ)−1 (1 − 2M/r)1/2 sin θ, cos θ ,
 (6.5)
AE = (1 − 2M/r)−1/2 M/r 2 , 0 .

The relative normal kE is tilted away from the equatorial plane compared to the horizontal inward
√ −1
flat spacetime relative normal −ρ −1 (sin θ, cos θ ) by the change ρ −1 → ρ/ 1 − 2M/r
reflecting the dilation of radial arc length compared to coordinate radius near the hole. This
contracts the radial component of the relative normal, tilting the normal away from its horizontal
direction as shown in figure 5. As one approaches the horizon r̄ = 2, the radial component of
kE goes to zero and kE becomes tangent to the horizon, pointing away from the equatorial plane.
1346 D Bini et al

Figure 5. In the Schwarzschild case, the geometry of the ZAMO Lie relative curvature kE tilted by
E the photon acceleration kE + A,
the radial contraction 1k r̂ , the ZAMO acceleration A, E the timelike
circular orbit degenerate acceleration hyperbola half-line a(α) and its centre at 21 (−kE + A) E in
region A for the typical angle θ = π/3, shown with the correct orientation of the acceleration plane
in the ρ–z coordinate plane. The additional tilt of the half-line a(α) above the broken line is due
E
to the proper time dilation factor multiplying A.

The circular photon acceleration is the vector kE + AE shown in figure 5 parallel to the
acceleration half-line. This vector has the same direction as the optical Lie curvature vector
[16]

kẼ = N(kE + A)
E = −ρ −1 (1 − 3M/r) sin θ, (1 − 2M/r)1/2 cos θ . (6.6)

Its radial component goes to zero at the interface r̄ = 3 between regions A and C where kẼ
becomes tangent to the interface circle, inside of which the vector points radially outward
rather than inward as in region A, becoming purely radial at the horizon. This is the famous
‘reversal of the optical centrifugal force’ effect which occurs in region C in the Schwarzschild
case [20].
Continuing the Schwarzschild discussion, at zero ZAMO relative velocity ν = 0 = α, the
acceleration starts out at AE on the radial axis, but the resultant of the centripetal acceleration
term kE sinh2 α and AE scaled by the factor γ 2 = 1 + sinh2 α (reflecting the change in the
proper time parametrization of the second time derivative) leads to an additional tilting of the
acceleration half-line (due to the term sinh2 α) compared to the resultant of the centripetal
acceleration and AE alone. Far from the hole the decrease in the radial component of kE
  
 r→∞ M M
   −→ 2 1 +
1k r̂ = 1 − (1 − 2M/r)1/2 ρ −1 sin θ r 2r (6.7)

 −→
2M 1
2M
is exactly compensated for by the observer acceleration
  
 −1/2 
 r→∞ M M
M 2M −→ 1 +
Ar̂ = 2 1 − r2 r (6.8)
r r 
 −→ ∞
2 M
Geometric interpretation of Frenet–Serret frames for circular orbits 1347

in the vector sum (−kE + A)/2 E defining the centre of the degenerate acceleration hyperbola,
putting the centre on the horizontal axis χ = π/2 − θ (see figure 5), but moving it away from
the equatorial plane on the opposite side of this axis as the observer acceleration grows without
bound approaching the horizon while the radial shortening factor is bounded. Both 1k r̂ and
Ar̂ are decreasing functions of r satisfying 1k r̂ < Ar̂ .
Since k r̂ + Ar̂ = −r −1 (1 − 2M/r)−1/2 (1 − 3M/r) changes sign (negative in region A,
positive in region C) while Ar̂ > 0, the component a r̂ can change sign (so that the acceleration
curve crosses the ∂θ̂ -axis) only in region A. For r̄ > 3 (region A) the degenerate acceleration
hyperbola half-line always intersects the ∂θ̂ -axis, but at r̄ = 3 it is parallel to this axis with
a constant radial acceleration component, and for 2 < r̄ < 3 (region C) this half-line tilts
away from the ∂θ̂ -axis, corresponding to the fact that the acceleration must always be radially
outward to resist the strong attraction to the hole no matter what value the velocity has.
The acceleration is symmetric under the transformation ν → −ν (or α → −α) which
interchanges the co-rotating and counter-rotating orbits in the Schwarzschild case. To leading
order in the black hole angular momentum parameter a, the nonzero observer expansion tensor
components and additional acceleration component are
 
 Ma a2 Ma 2
θ r̂ φ̂ , θ θ̂ φ̂ → 3 sin θ −3, 2 sin 2θ , Aθ̂ → − 3 sin 2θ. (6.9)
r r r
As one turns on the angular momentum parameter a from 0 to a small positive value, the
leading-order  effect is the introduction of a small radially inward symmetry breaking term
− 3Ma/r 2 sin θ sinh 2α which separates the co-rotating and counter-rotating acceleration
vectors in the radial direction by equal amounts about the ā = 0 acceleration half-line, with
the counter-rotating/co-rotating orbits experiencing an increased/decreased radial acceleration
due to the gravitomagnetic field. The small nonzero component Aθ̂ shifts the vertex of the
acceleration hyperbola slightly off the ∂r̂ -axis, tilting AE away from the equatorial plane. These
effects qualitatively describe the situation for all physical values 0 < ā 6 1 since the vector θEφ̂
is always radial to within less than about 1% at its most extreme value not far from the horizon

( θ θ̂ φ̂ θ r̂ φ̂ . 0.01), while AE is always radial to within less than about 5% at its most extreme
value not far from the horizon. Thus the acceleration half-line parallel to (kE + A) E cosh 2α
E
terminating at A (combined spatial geometry and gravitoelectric terms) is split into the two
halves of the one branch of the acceleration hyperbola by the addition of the nearly radial vector
fields ±2θEφ̂ | sinh 2α| (the gravitomagnetic term) which pulls the counter-rotating acceleration
further away from the symmetry axis ρ = 0.

7. Concluding remarks

The discovery of the geometric content of the complicated formulae of Iyer and Vishveshwara
has made possible a simple and very visual description of the complete behaviour of the
acceleration vector and Fermi–Walker angular velocity for circular orbits off the equatorial
plane in the Kerr spacetime, in particular, and in the orthogonally transitive stationary
axisymmetric case in general. Until now attention has been limited either to the equatorial
plane or to the discussion of the alignment of the acceleration and Fermi–Walker angular
velocity, special cases which now follow easily from the general picture. The properties of
the equatorial plane case in Kerr are also easier to understand in the context of the general
nonequatorial case. The Schwarzschild limit reveals the essential roles of the spatial geometry,
relative centripetal acceleration and gravitoelectric field in shaping the qualitative behaviour
of the acceleration vector in the absence of rotation, while the gravitomagnetic field nicely
explains the asymmetry between the co-rotating and counter-rotating orbits in the rotating
1348 D Bini et al

case. The next step in the program of understanding how the properties of orbits in black hole
spacetimes are related to the properties of the gravitational field is the spherical orbit case,
where the motion is no longer along Killing vector directions. This will be discussed in future
work.

References

[1] Iyer B R and Vishveshwara C V 1993 Phys. Rev. D 48 5706


[2] Hönig E, Schücking E L and Vishveshwara C W 1974 J. Math. Phys. 15 744
[3] Synge J L 1960 Relativity: the General Theory (Amsterdam: North-Holland)
[4] Bini D, Carini P and Jantzen R T 1995 Class. Quantum Grav. 12 2549
[5] Carter B 1968 Phys. Rev. 174 1559
[6] Damour T and Ruffini R 1975 Phys. Rev. Lett. 35 463
[7] Znajek R L 1977 Mon. Not. R. Astron. Soc. 179 457
[8] Bini D, Carini P and Jantzen R T 1997 Int. J. Mod. Phys. D 6 143
[9] Page D 1998 Class. Quantum Grav. 15 1669
[10] Semerák O 1996 Gen. Rel. Grav. 28 1151
[11] Semerák O 1998 Gen. Rel. Grav. 30 1203
[12] Wilkins D 1972 Phys. Rev. 5 814
[13] de Felice F 1995 Class. Quantum Grav. 12 1119
de Felice F and Usseglio-Tomasset S 1991 Class. Quantum Grav. 8 1871
[14] Bini D, Jantzen R T and Merloni A 1999 Proc. 3rd William Fairbank Meeting (1998) ed C Sigismondi (Singapore:
World Scientific)
[15] Bini D, Carini P and Jantzen R T 1992 Proc. 3rd Italian–Korean Astrophysics Meeting ed S Kim, H Lee and
K T Kim (J. Korean Phys. Soc. 25 S190)
[16] Bini D, Carini P and Jantzen R T 1997 Int. J. Mod. Phys. D 6 1
[17] Jantzen R T, Carini P and Bini D 1992 Ann. Phys., NY 215 1
[18] Abramowicz M A, Nurowski P Wex N 1995 Class. Quantum Grav. 12 1467
[19] Bini D, de Felice F and Jantzen R T 1998 Preprint
[20] Abramowicz M A and Prasanna A R 1990 Mon. Not. R. Astron. Soc. 245 720

You might also like