RINDLER-ISHAK METHOD: LIGHT DEFLECTION IN WEYL

GRAVITY
Amrita Bhattacharya,
1,2,o
Ruslan Isaev,
3,b
Massimo Scalia,
2,c
Carlo
Cattani,
2,4,J
and Kamal K. Nandi
1.3,5,t
1
Department of Mathematics, University of North Bengal, Siliguri 734 013,
India
2
Dipartimento di Matematica, Istituto G.Castelnuovo, Universit La
Sapienza, P.le Aldo Moro, 2, Rome, Italy
3
Joint Research Laboratory, Bashkir State Pedagogical University, Ufa 450000,
Russia
4
DiFarma, Universit di Salerno, Via Ponte Don Melillo, 84084 Fisciano,
Salerno, Italy
5
Department of Theoretical Physics, Sterlitamak State Pedagogical Academy,
Sterlitamak 453103, Russia
o
Email: [email protected]
b
Email: [email protected]
c
Email: [email protected]
J
Email: [email protected]
t
E-mail: [email protected]
PACS number(s): 04.50.1h, 04.20.Cv
Abstract
The null geodesic equation in the Schwarzschild-de Sitter gravity does not
contain the cosmological constant /, yet the Rindler-Ishak method yields its
repulsive inuence on the deection of light. We shall implement the method in
the more general environment of Weyl gravity where / still cancels out of the
null geodesic equation. The resulting light deection (we do it up to one order
higher than originally considered) shows that the method leads to consistent
and new results when applied to the Mannheim-Kazanas-de Sitter solution of
Weyl gravity. Dierent cases are analyzed.
-
I. Introduction
There are two common methods for calculating the bending of light rays in
a static, spherically symmetric gravity eld. One method uses small perturba-
tions and the other uses direct integration of the null trajectory (Weinbergs
method). Recently, a third method has been proposed by Rindler and Ishak [1].
It combines the perturbative solution with an invariant denition of a bending
angle. The method has been developed targetting a certain goal: It has been
1
known for a long time that the cosmological constant / (= .,3) cancels out [2]
of null geodesic in the Schwarzschild-de Sitter (SdS) spacetime and so the pre-
vailing belief has been that / does not inuence the deection of light. Rindler
and Ishak have shown that this need not be the case. They argued that "the
dierential equation and its integral are only half the story. The other half is the
metric itself, which determines the actual observations that can be made on the
r, , orbit equation. When that is taken into account, a quite dierent picture
emerges: the cosmological constant does contribute to the observed bending
of light." They buttress the argument by Ricci and Weyl focusing mechanism
as well. Their method indeed yielded the expected repulsive inuence of / on
light deection.
On the other hand, such cancellation might occur in a more general space-
time too and it would be interesting to know what the method yields in the
general case. An excellent example for this purpose is the Mannheim-Kazanas-
de Sitter (MKdS) solution of Weyl conformal gravity. The theory is based on
the conformal invariance and it attempts to resolve the dark matter/dark en-
ergy problem without hypothesizing them. The theory has its own merits and
demerits, but their discussion is beyond the scope of this paper.
In the present paper, implementing the Rindler-Ishak method to one higher
order in ' than considered originally, we wish to calculate the bending of light
rays in the MKdS solution. It turns out that the method does deliver the
expected results in addition to revealing new terms otherwise unavailable. The
implication of one particular term is pointed out though this is not the essential
focus of the paper.
The contents are organized as follows: In Sec.II, we derive the geodesic
equation in the MKdS solution. In Sec.III, we work out the bending of light
rays following the Rindler-Ishak method. We discuss specic cases in Sec.IV
while Sec.V summarizes the paper. There are two appendices.
II. Geodesic equation
An interesting solution of the Weyl gravity eld equations is the MKdS
metric given by [3] (in units 8G = c
0
= 1):
dt
2
= 1(r)dt
2

1
1(r)
dr
2
r
2
(d0
2
+sin
2
0d,
2
), 1(r) = 1
2'
r
+r/r
2
, (1)
where ' is the central mass, / and are constants. The numerical value of
/ = .,3 - 10
56
cm
2
, and is of the order of inverse Hubble length. However,
there is a reported ambiguity in the sign of . Mannheim and Kazanas x it
from at rotation curve data to be positive, - +10
28
cm
1
[3] while Pireaux
[4] argues for - 10
33
cm
1
(see also [7]). We are free to assume any sign
for but for deniteness, we take 0 in the present work.
In the light orbit equation, / surprisingly cancels out. In general, using
n = 1,r, we derive the following path equation for a test particle of mass :
0
on the equatorial plane 0 = ,2 as follows:
d
2
n
d,
2
= n + 3'n
2

2
+
'
/
2
+
1
2/
2
n
2
_

2/
n
_
, (2)
2
where / =
I3
n0
, the angular momentum per unit test mass. For photon, :
0
=
0 =/ and one ends up with the null geodesic equation without / making
its appearance:
d
2
n
d,
2
= n + 3'n
2

2
. (3)
In the SdS metric ( = 0), such a disappearance has been noted for long. Here
we nd that similar cancellation of / occurs despite the presence of in the
metric. Exactly like in the SdS case, one would now tend to believe that the
bending of light in the MKdS case would be the uninuenced by / to any order.
We shall show below that it is not the case.
III. Rindler-Ishak method
Right at the outset, we wish to mention that we shall adhere to the three
principles from the method adopted by Rindler and Ishak. (i) The method is
originally applied in the SdS metric to show that, despite the non-appearance of
/ in the null geodesic, its eect still appears in light bending. Hence to preserve
the essence of the method, we shall try to retain / ,= 0 except in special cases.
(ii) With / ,= 0 in the metric, the limit r makes no sense since, in the
too far eld regime, the MKdS metric becomes predominantly dS providing
a horizon radius
_
1,/. The only intrinsically characterized r value replacing
r is the one at , = 0. The measurable quantities are the various c angles
that the photon orbit makes with successive coordinate planes , = const. We
shall qualitatively verify the results for , ,= 0 as well, say, at , = ,4. (iii) We
shall strictly use the Rindler-Ishak expression for the deection angle c.
Let us develop the basic equations now. Although the MKdS metric is
dierent from the SdS metric, we shall show that not only the inuence of /
but also the otherwise known factor of appear in the light bending. The light
path equation (3) in the zeroth order is
d
2
n
0
d,
2
+n
0
=

2
, (4)
The exact solution of Eq.(4) is
n
0
=
cos ,
1


2
(5)
where 1 is the distance of closest approach to the origin. This solution is to
be used as the zeroth order approximation [5]. Following the usual method of
small perturbations [6], we want to derive the solution as
n = n
0
+n
1
+n
2
(6)
where n
1
and n
2
respectively satisfy
d
2
n
1
d,
2
+n
1
= 3'n
2
0
(7)
d
2
n
2
d,
2
+n
2
= 6'n
0
n
1
. (8)
3
The exact solutions are
n
1
=
'
41
2
_
6 + 31
2

2
61 cos , 2 cos , 61,sin,

(9)
n
2
=
3'
2
161
3
[
_
10 + 31
2

2
_
5 2,
2
__
cos , 121(4 +1
2

2
)
+ cos 3, + 20,sin, + 161 cos 2, + 301
2

2
,sin, + 81,sin2,]. (10)
Formally changing ,
t
2
,, the nal solution up to second order in ' can
be written as
n =
1
r
=
sin,
1


2
+
'
41
2
_
6 + 31
2

2
31( 2,) cos , + 2 cos 2, 61 sin,

3'
2
321
3
[961 + 241
3

3
10(2 + 31
2

2
)( 2,) cos , + 321 cos 2,
20 sin, 301
2

2
sin, + 3
2
1
2

2
sin, 121
2

2
,sin,
+ 121
2

2
,
2
sin, 81 sin2, + 161,sin2, + 2 sin3,]. (11)
When = 0, it may be veried that one recovers the equation for light trajectory
up to '
2
order in the Schwarzschild metric [6].
The method of Rindler and Ishak is based on the invariant formula for the
cosine of the angle c between two coordinate directions d and c such that
cos c =
q
I
d
I
c

(q
I
d
I
d

)
1/2
(q
I
c
I
c

)
1/2
. (12)
Dierentiating Eq.(11) with respect to ,, and denoting
J:
J,
= (r, ,), we get
(r, ,) = (r
2
)
[
cos ,
321
3
(321
2
3'
2
20 + 31
2

2
(10 + ( 2,)
2

+ 32'1(9' 2) sin,) 6'3' cos 3, 8'1( 2,) cos 2,

+ (10' 41
2
+ 9'1
2

2
)( 2,) sin,]. (13)
Eq.(12) then yields [1]
cos c =
[[
[
2
+1(r)r
2
]
1
2
. (14)
In a more convenient form, the nal Rindler-Ishak expression for c to be used
is [see (iii) above]
tanc =
1
1/2
r
[[
. (15)
The bending angle is dened by c = c ,. The basic ingredients are Eqs.(1),
(11), (13) to be plugged into Eq.(15). We shall now discuss specic cases.
4
IV. Specic cases
Case 1: When , = 0 [see (ii) above], and ' ,= 0, we get from Eq.(11)
r = 161
3
,[4'1(8 31 + 31
2

2
) 81
3

+ 3'
2
5(2 + 31
2

2
) 41(16 + 31
2

2
)], (16)
and putting this value in Eq.(13), we get
= 81
3
['
2
(78 481 + 901
2

2
+ 9
2
1
2

2
) 321
2
],[4'1(8 31
+ 31
2

2
) 81
3
+ 3'
2
5(2 + 31
2

2
) 41(16 + 31
2

2
)]
2
. (17)
If we ignore '
2
terms and put = 0 above, we recover the values for r and [[
as in the SdS spacetime. The one sided bending angle is given by c = c ,
and let us calculate c = c = c
0
when , = 0. Putting in Eq.(15) the values for
1(r) from Eq.(1), r from Eq.(16) and from Eq.(17), we get
tanc =
_
_
1
2'
r
+r /r
2
_

2[4'1(8 31 + 31
2

2
) 81
3
+ 3'
2
5(2 + 31
2

2
)
41(16 + 31
2

2
)],['
2
(78 481 + 901
2

2
+ 9
2
1
2

2
) 321
2
].
(18)
This is the exact formula for light deection but is unilluminating. Therefore,
successively expanding in the rst order of , and in the second order in ',
we nd, for a small angle c
0
(or, tanc
0
c
0
), the following expression (see
Appendix A for details):
c
0

2'
1
_
1 +
15'
161

/1
4
8'
2
_

_
1
2
+
3'
4
+
455'
2
321
+
4'
2
/1
3
_
(19)
where we have retained only the leading order terms in the coecient of
21
1
and assuming 1 '. We nd that all terms in the last third bracket are
positive, meaning that the eect of is to diminish the Schwarzschild bending
even up to second order in '. We also nd that Eq.(19) reproduces the term

~1
2
obtained by Edery and Paranjape [7] using Weinbergs method. Thus, it is
clear that the Rindler-Ishak method is not only correct but also tells us more in
the form of other new terms. Eq.(19) is the central result of the present paper.
Case 2: , = 0, ' = 0. This is a bit of a tricky case because when ' = 0,
the exact solution
1
:
=
sin ,
1

~
2
gives a negative r at , = 0. If we still brutally
proceed keeping / ,= 0 but ignoring the fact that r is negative, we end up with
c = i
1
2
_
1 +
4/

2
, (20)
which is clearly meaningless even at / = 0.
5
The best that one can do under the circumstances is to consider an angle ,
very close to zero [5]. Then we have from the exact solution
1
:
=
sin ,
1

~
2
, the
following approximation for very large r,
, sin,
1
2
=tan, = tan
_
1
2
_
(21)
Then, according to the present method [see (iii)], we have
tanc =
1
2
sec
_
1
2
_
_
2
2
1
2
4/1
2
2 cos (1). (22)
These give to rst order in the deection
c
1
2
i
_
/1. (23)
(See Appendix B for an alternative derivation). The only way to extract phys-
ically meaningful information from this is to set / = 0. This then implies we
are considering pure conformal gravity with metric potential 1(r) = 1 +r for
which the deection is c =
~1
2
.
Case 3: When the , value is other than zero, say, , = ,4 [see (ii) above],
and ' ,= 0, the Rindler-Ishak formula gives the bending angle c as
c = c , - tan(c ,) =
tanc tan,
1 + tanc tan,
=
tanc 1
1 + tanc
. (24)
Proceeding in the similar manner as above, we can nd the value of tanc and
thence of c. Surprisingly the coecient of 1 becomes identically zero. We do
not give detailed expressions here but only state the nal result to leading order:
c =
_
2'
1

3'
8


2
1
2
8

/
2
1
4
4

/1
2
2
. (25)
Once again the deection due to is negative. Incidentally, the rst term is
slightly dierent from that of Rindler and Ishak [their Eq.(19)]. Also, for = 0,
their expression [Eq.(18)] for tanc slightly modies to
tanc = 1 +
2
_
2'
1
/1
2
. (26)
These modications by no means alter their demonstration of the repulsive
eect of / (in fact it is exactly the same). Even the pattern (loosely speaking)
of division by 2 in the last two pieces of tanc results in the corresponding
pieces in c, just as it is in Ref.[1]. The minor changes in the coecient of '
have come about because we have used the path equation at , = ,4 to obtain
r =
_
21 3' +O('
2
) rather than r =
_
21.
Case 4: , = ,4, ' = 0. Then
r =
21
_
2 1
, =
2
_
21
(
_
2 1)
2
, tan, = 1, tanc =
_

2
1
2
+ 4/1
2
2
4
_
21 2
2
1
2
4
_
1/2
.
(27)
6
For small deections, expanding successively in powers of and rst power of
/, we get
c =
tanc tan,
1 + tanc tan,

2
1
2
8

/
2
1
4
4

/1
2
2
. (28)
This result can also be obtained directly from Eq.(25) by putting ' = 0.
V. Summary and conclusion
As shown in Eq.(3), in the MKdS gravity too, which is more general than
the SdS gravity, the constant / cancels out of the light orbit equation even
though ,= 0, the latter fact distinguishing the MKdS metric from the SdS
metric. We should also remember that the corresponding parent theories are
generically very dierent; one is fourth order and the other is the second order
gravity. Nevertheless, in view of similar cancellation in the two metrics, we
investigated the applicability of the Rindler-Ishak method taking the MKdS
solution as an example. This is a nice example because Weyl conformal gravity
accommodates the successes of Schwarzschild gravity and has been the subject
of active research for several years.
We rst derived the exact null geodesic equation in the MKdS gravity. Next,
we perturbatively solved the equation up to the order '
2
though the zeroth or-
der equation is dierent from that in standard Schwarzschild gravity. Whatever
follows below are the results obtained after the detailed solution is plugged into
the Rindler-Ishak procedure we have implemented without any deviation.
We should note that, generally speaking, none of the quantities /, and '
is zero. (We can nonetheless set one or the other to zero as limiting cases). Then
Eq.(19) at once shows the inuence of / both in the Schwarzschild and conformal
sector. The equation nicely reproduces the attractive Schwarzschild bending
due to ' 0 as well as the repulsive bending due to cosmological constant
/ 0 and the Weyl conformal parameter 0. Eq.(19) combines in one place
deections at various orders obtained by Bodenner & Will [6], by Rindler &
Ishak [1] and by Edery & Paranjape [7]. Because of '
2
order, there appeared
new terms in the conformal sector. In particular, the term
455~1
2

321
diminishes
the Schwarzschild rst order one way bending
21
1
by a radius independent
factor
_
1
455~1
64
_
, but the deviation from unity is so minute (10
23
) even in
the light-grazing-sun experiment that the precision required for its detection is
technologically unattainable even in the far future.
We nd that the Rindler-Ishak method at , = 0 strictly requires that / ,= 0,
' ,= 0. If any of the condition is violated, one or the other term in Eq.(19)
blows o. We expect that these conditions might be relaxed at large nonzero
values of ,. We have worked out only the , = ,4 case for an illustration, which
shows that such relaxation is indeed possible. There is absolutely no problem in
the case , = ,4, / = 0, ' = 0 [Eq.(25)] because the expression for c does not
blow o. In this case, the term
~1
2
of course does not appear, but one notices
that the eect is
~
2
1
2
8
, which remains unaltered by the choice of sign for . In
either case (, = 0 or ,4), the conformal parameter can always be set to zero
at will, but then one ends up with the already discussed SdS case. Only the
7
case , = 0, / = 0, ' = 0 is a bit troublesome, which is understandable because
the reason lies in the negative r value leading to the occurrence of imaginary
quantities. (Note that imaginary quantities arise also in Weinbergs integration
method, see Appendix B.) A plausible way out is to take a very small , and do
the calculation with / ,= 0. Eventually, one can put / = 0 to obtain c =
~1
2
.
This then is the deection in the pure conformal gravity dened by ' = 0,
/ = 0 so that 1(r) = 1 + r (see Sec.IV, Case 2). Overall, the conclusion is
that the Rindler-Ishak method can be applied to more general situation than
SdS and that it leads to new terms unavailable otherwise.
The implication of the last term in Eq.(19), viz.,
4~1
2
|1
3
, seems to provide
the background for a certain postulate. To have some idea how, let us estimate
the terms in Eq.(19) for a light ray grazing a stellar sized massive object, say, the
Sun itself, and that - 10
28
cm
1
, / - 10
56
cm
2
, the values being already
independently estimated. We then have the following numerology:
'

= 1.48 10
5
cm, 1

= 6.96 10
10
cm, (29)
so that
2'

= 4.24 10
6
,
30'
2

161
2

= 2.65 10
11
,
/1
3

4'

- 10
30
(30)

2
- 10
18
,
3'

4
- 10
23
,
455'
2

321

- 10
28
,
4'
2

/1
3

- 10
6
.
(31)
We nd that, while all the other terms above are quite small compared to
21
1
, the spoiling term is the very last one, which could be discovered only
by using the Rindler-Ishak approach. If we had limited ourselves to the direct
integration of the null trajectory in which / does not appear, we would have
missed this term. This term dominates giving rise, at the limb of the Sun,
to a total bending, c
0
- 10
6
, which is negative and means repulsion 10
12
times bigger than the Schwarzschild attraction! Certainly this is contrary to
our experience. Moreover, with ' = '

, for any 1 _ 1

, one nds that

the values of c
0
continue to remain only negative. Without that spoiling term,
however, the eect of would indeed be negligible near the Sun so that c
0
would not appreciably dier from the positive Schwarzschild value, which does
not appear to be the case here.
One meaningful resolution probably is to postulate that = 0 on the local
scale so that deection there is dictated only by the Schwarzschild terms as
in (30), while deection in the galactic halo region is likely to be determined
by some conspiracy between and /dependent terms. We cannot say what
the conspiracy is until we are wiser with concrete magnitude and sign of .
Historically, / has had a chequered career as well, being employed at convenience
from time to time (current cosmic acceleration model noted). With all these
uncertainties around, it would indeed be too premature to draw any reliable
picture. This is not the focus of the present article either.
Acknowledgment
8
One of us (AB) wishes to thank the authorities of the Universit La Sapienza,
Roma, for nancial support during the project work. KKN and RI wish to thank
Guzel N. Kutdusova for technical assistance.
References
[1] W. Rindler and M. Ishak, Phys. Rev. D 76, 043006 (2007).
[2] J.N. Islam, Phys. Lett. A 97, 239 (1983).
[3] P. D. Mannheim and D. Kazanas, Astrophys. J. 342, 635 (1989).
[4] S. Pireaux, Class. Quant. Grav. 21, 4317 (2004).
[5] We thank an anonymous referee for pointing this out.
[6] J. Bodenner and C.M. Will, Am. J. Phys. 71, 770 (2003).
[7] A. Edery and M. B. Paranjape, Phys. Rev. D 58, 024011 (1998).
Appendix A
For the benet of the readers we give below the exact steps leading to the
expression (19). After putting the value of r and from Eqs.(16), (17) into
Eq.(18) and simplifying, we obtain
tanc
0
=
1
o
_
_
2(1 +Q)
_
1 +'
1
2
(83t~1+3~
2
1
2
)
21
2

Q
81
3
+
16~1
3
1+Q

256|1
6
(1+Q)
2
_
1/2
_
_
(A1)
where
1 = 4'1(8 31 + 3
2
1
2
) 81
3
(A2)
Q = 3'
2
5(2 + 3
2
1
2
) 41(16 + 3
2
1
2
) (A3)
o = '
2
(78 481 + 901
2

2
+ 9
2
1
2

2
) 321
2
. (A4)
In order to extract the contribution due purely to , we expand the right hand
side in the rst power of , and get
tanc
0

2Tl
78'
2
321
2
+
2T
\
_
' +
3'
2
21
+
24'
3
1
2
+
161
3
T

265/1
7
\
'
3
(15' + 161)
3
_
(A5)
where
T = 30'
2
+ 32'1
l =
_
1
4'
2
1
2

15'
3
41
3

256/1
6
T
2
_
1/2
(A6)
\ = l
_
156'
2
641
2

21\
39'
2
161
2
+
241'
3
(15' + 161)
(39'
2
161
2
)
2
_
(A7)
\ = 48'
2
+ 3'1 + 21
2
. (A8)
9
We rst expand the second term in (A5) in powers of ' obtaining
c
MKdS
0

_
1
2
+
3'
4
+
455'
2
321
+
4'
2
/1
3
_
. (A9)
Next let us put = 0 in (A5) in order to obtain pure Schwarzschild terms from
tanc
SdS
0

2Tl
78'
2
321
2
. (A10)
The deection is already known to be positive. To ensure it, Rindler and Ishak
in their SdS treatment prescribed [[ instead of just . This prescription is the
same as changing the denominator of (A10) into 321
2
78'
2
because 1 '
and Tl 0. For small c
0
, and with l 1
21
2
1
2

15t1
3
81
3

128|1
6
T
2
, the rst
term in (A5) then results in
c
SdS
0
=
2(30'
2
+ 32'1)
_
1
21
2
1
2

15t1
3
81
3

128|1
6
T
2
_
321
2
78'
2
, (A11)
which, when expanded in powers of ', yields
c
SdS
0

2'
1
_
1 +
15'
161

/1
4
8'
2
_
. (A12)
The total deection is of course
c
0
= c
SdS
0
+c
MKdS
0
(A13)
which is just the Eq.(19).
Appendix B
Consider the null geodesic equation when ' = 0:
d
2
n
d,
2
= n

2
(B1)
which has an exact solution
n =
1
r
=
sin,
1


2
. (B2)
The metric for ' = 0 is
1(r) = 1 +r /r
2
. (B3)
From (B2) it is clear that the angle , = 0 is meaningless as it gives a negative r
for 0. Nevertheless, Weinbergs method allows an exact integration giving
the deection as
, =
_
1
1
_
r
4
(1 +1)
1
2
r
2
r
3
_
1/2
dr (B4)
= i
_
lni(2 +1)
_
1 +1ln2
_
1 +1 i1
_
1 +1
_
.
10
As noted by Edery and Paranjape [7], / has cancelled out of the integrand. It
is to be expected since the path equation does not contain / either. Therefore,
even for pure conformal gravity, when 1(r) = 1 + r, the above result holds
true. Although , in general is complex, fortunately its rst order expansion
in is real and yields
,

2

1
2
+ complex terms. (B5)
The real term is of course consistent and supports the fact that pure conformal
gravity is repulsive. A somewhat similar situation arises also in the Rindler-
Ishak method as explained in Sec.IV, Case 2, Eq.(23).
11