-
Notifications
You must be signed in to change notification settings - Fork 18
/
gek.tex
3938 lines (3672 loc) · 198 KB
/
gek.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
\chapter{Geodesics in Kerr spacetime: generic and timelike cases}
\label{s:gek}
\index{geodesic!in Kerr spacetime}
\minitoc
\section{Introduction}
In various occasions during our study of Kerr spacetime in Chap.~\ref{s:ker}, we have already
encountered some peculiar geodesics, namely
the principal null geodesics. In this chapter, we begin the systematic study of causal
(timelike or null) geodesics in Kerr spacetime.
First, taking advantage of an algebraic particularity of this spacetime, giving
birth to a non-trivial Killing tensor and the associated Carter constant,
we shall see in Sec.~\ref{s:ges:equat_geod_motion}
that Kerr geodesic motion is fully governed by a system of first order equations,
the solution of which can be obtained by quadrature.
The basic properties of geodesics that can be inferred from this system are derived
in Sec.~\ref{s:ges:general_prop}.
Then we focus on timelike geodesics (Sec.~\ref{s:gek:timelike}) and in particular
on those that are bound; they represent orbits of massive particles or bodies around a
Kerr black hole. In Sec.~\ref{s:gek:circ_equat}, we shall investigate the particular
case of circular orbits in the equatorial plane and discuss their stability.
The detailed study of null geodesics is deferred to the next chapter.
Of course, when the Kerr spin parameter $a$ tends to zero, all
results in this chapter reduce to those obtained for Schwarzschild geodesics
in Chap.~\ref{s:ges}. As already suggested in the introduction of Chap.~\ref{s:ges}, before moving on, the reader might want to have
a look at Appendix~\ref{s:geo}, which recaps the properties of geodesics in
pseudo-Riemannian manifolds.
\section{Equations of geodesic motion} \label{s:ges:equat_geod_motion}
\subsection{Introduction} \label{s:gek:geod_motion_intro}
In all this chapter, we are concerned with the motion of a particle
$\mathscr{P}$ in the Kerr spacetime $(\M,\w{g})$, under the hypothesis that $\mathscr{P}$
feels only gravity, as described by $\w{g}$ (freely falling particle).
The worldline $\Li$ of $\mathscr{P}$
is then necessarily a geodesic\footnote{The definition and basic properties of geodesics
are recalled in Appendix~\ref{s:geo}; see also Sec.~\ref{s:fra:geod_motion}.} of
$(\M,\w{g})$. It is a timelike geodesic if $\mathscr{P}$ is a massive particle
and a null geodesic if $\mathscr{P}$ is massless (e.g. a photon).
In a given coordinate system $(x^\alpha)$, the geodesic $\Li$ is described
by a system of the form\footnote{In Appendices~\ref{s:bas} and \ref{s:geo},
a different symbol is used for the function $X^\alpha(\lambda)$ defining
$\Li$ and the coordinate $x^\alpha$ (cf. Secs.~\ref{s:bas:vectors} and
\ref{s:geo:geodesic_eq}). Following standard usage in the physics literature,
we shall not do this in this chapter.}
$x^\alpha = x^\alpha(\lambda)$, where $\lambda$ is an
affine parameter\footnote{See Sec.~\ref{s:geo:def} for the definition
of an affine parameter along a geodesic.} along $\Li$. We choose $\lambda$ to be the affine parameter
associated with $\mathscr{P}$'s 4-momentum $\w{p}$,
i.e. such that $\D x^\alpha / \D\lambda = p^\alpha$ [Eq.~(\ref{e:fra:p_dxdl})].
In particular, $\lambda$ is dimensionless and is increasing towards the future\footnote{Let us recall that the Kerr spacetime is time-oriented, cf. Sec.~\ref{s:ker:time_orientation}.} along $\Li$.
The curve $\Li$ is a geodesic iff the functions $x^\alpha(\lambda)$ obey
the geodesic equation [Eq.~(\ref{e:geo:eq_geod}) in Appendix~\ref{s:geo}]:
\be \label{e:gek:eq_geod}
\frac{\D^2 x^\alpha}{\D\lambda^2} + \Gamma^\alpha_{\ \, \mu \nu}
\frac{\D x^\mu}{\D\lambda} \frac{\D x^\nu}{\D\lambda} = 0 ,\qquad 0 \leq \alpha \leq 3,
\ee
where the $\Gamma^\alpha_{\ \, \mu \nu}$'s are the Christoffel symbols\index{Christoffel symbols} of the metric $\w{g}$
with respect to the coordinates $(x^\alpha)$, as given by Eq.~(\ref{e:bas:Christoffel}).
Equation~(\ref{e:gek:eq_geod}) is the component expression of
$\wnab_{\w{p}} \, \w{p} = 0$ [Eq.~(\ref{e:fra:p_geodesic})]. It is
a system of four coupled second-order
differential equations, which are non-linear. We are going to see that it is actually not necessary to
solve this system to compute the geodesics of Kerr spacetime. Indeed,
as in the Schwarzschild case studied in Chap.~\ref{s:ges}, there exists
enough first integrals of Eq.~(\ref{e:gek:eq_geod}) to reduce
the problem to four first-order equations
of the type $\D x^\alpha/\D\lambda = F^\alpha(x^0, x^1, x^2, x^3)$.
Three integrals of motion are similar to those of Schwarzschild geodesics:
one is the particle's mass $\mu$ (Sec.~\ref{s:gek:mass_int_motion} below)
and the two other ones are
the conserved energy $E$ and the conserved angular momentum
$L$ (Sec.~\ref{s:gek:int_motion_sym}), which arise from the common symmetries
of Kerr and Schwarzschild spacetimes: stationarity (outside the ergoregion) and axisymmetry.
In the Schwarzschild case, the fourth integral of motion was provided by
spherical symmetry, which constrained all geodesics to be planar, so that
a suitable choice of
coordinates $(t,r,\th,\ph)$ makes a given geodesic confined to the
hyperplane $\th=\pi/2$, yielding the first integral $p^\th=0$ [Eq.~(\ref{e:ges:pth_zero})].
The Kerr spacetime with $a\not=0$ being not spherically symmetric, we loose
this property here. Fortunately there exists another
integral of motion, known as the \emph{Carter constant}; it arises from a remarkable property
of Kerr spacetime: the existence of a non-trivial Killing tensor (Sec.~\ref{s:gek:Carter_const}).
This makes a total of four integral of motions,
which makes the problem integrable (Secs.~\ref{s:gek:first_order_system} to
\ref{e:gek:integration}).
\subsection{Integrals of motion from spacetime symmetries} \label{s:gek:int_motion_sym}
As for Schwarzschild spacetime (cf. Sec.~\ref{s:ges:fiom}), we have the following
property:
\begin{prop}[conserved quantities along causal geodesics]
\label{p:gek:conserved_E_L}
The Killing vectors $\w{\xi}$ and $\w{\eta}$ of Kerr spacetime,
associated respectively with
stationarity and axisymmetry [cf. Eq.~(\ref{e:ker:def_xi_eta})],
give birth to two conserved quantities
along any (causal) geodesic $\Li$:
\begin{subequations}
\label{e:gek:def_E_L}
\begin{align}
& \encadre{E := - \w{\xi}\cdot \w{p} = - \w{g}(\w{\xi},\w{p}) } \label{e:gek:def_E} \\
& \encadre{L := \w{\eta}\cdot \w{p} = \w{g}(\w{\eta},\w{p}) } , \label{e:gek:def_L}
\end{align}
\end{subequations}
where $\w{p}$ is the 4-momentum of the particle $\mathscr{P}$
having $\Li$ as worldline (cf. Sec.~\ref{s:fra:worldlines}).
For the same reasons as in Sec.~\ref{s:ges:fiom}, $E$ is called
the \defin{conserved energy}\index{conserved!energy}\index{energy!conserved --}
or \defin{energy at infinity}\index{energy!at infinity} of $\mathscr{P}$,
while $L$ is called the \defin{conserved angular momentum}\index{conserved!angular momentum}\index{angular momentum!conserved --}
or \defin{angular momentum at infinity}\index{angular momentum!at infinity}
of $\mathscr{P}$.
\end{prop}
In particular, if $\Li$ reaches the asymptotic region $|r|\to+\infty$,
$E$ is nothing but $\mathscr{P}$'s energy as measured by the asymptotic inertial
observer introduced in Sec.~\ref{s:ker:asymp_inertial_obs}, given that
the 4-velocity of the latter coincides with the Killing vector
$\w{\xi}$
[compare Eq.~(\ref{e:fra:E_obs}) with Eq.~(\ref{e:gek:def_E})].
Similarly, $L$ is the component along the rotation axis of $\mathscr{P}$'s (total) angular momentum vector $\w{L}_{\rm tot}$
as measured by the asymptotic inertial observer [cf. Eq.~(\ref{e:ges:L_tot_asympt})].
\begin{remark}
Because it represents only a component of the total angular momentum,
$L$ is sometimes denoted by $L_z$.
\end{remark}
In coordinates $(t,r,\th,\ph)$ adapted to spacetime symmetries,
i.e. coordinates such that $\w{\xi} = \wpar_t$ and $\w{\eta}=\wpar_\ph$,
for instance Boyer-Lindquist coordinates (Sec.~\ref{s:ker:expr_BL}),
advanced Kerr coordinates (Sec.~\ref{s:ker:Kerr_coord}) or Kerr coordinates
(Sec.~\ref{s:ker:3p1_Kerr_coord}), one can rewrite
(\ref{e:gek:def_E_L})
in terms of the components $p_t = g_{t\mu} \, p^\mu$ and $p_\ph = g_{\ph\mu} \, p^\mu$
of the 1-form $\uu{p}$ associated to $\w{p}$ by metric duality:
\begin{subequations}
\label{e:gek:E_pt_L_pph}
\begin{align}
& E = - p_t \\
& L = p_\ph
\end{align}
\end{subequations}
Indeed, in such a coordinate system, $\xi^\mu = \delta^\mu_{\ \, t}$
and $\eta^\mu = \delta^\mu_{\ \, \ph}$, so that $E = -g_{\mu\nu} \, \xi^\mu p^\nu = -g_{t\nu} \, p^\nu = -p_t$
and $L = g_{\mu\nu} \, \eta^\mu p^\nu = g_{\ph\nu} \, p^\nu = p_\ph$.
\begin{example}[Generators of the event and inner horizons] \label{x:gek:null_generator_hor}
Let us choose for $\Li$ a null geodesic generator of the event horizon $\Hor$
(cf. Sec.~\ref{s:gek:null_gen_hor}). $\Li$ is then an outgoing principal null
geodesic $\Li^{{\rm out},\Hor}_{(\th,\psi)}$
and the 4-momentum vector $\w{p}$ is proportional to the null vector
$\wl \equalH \w{\chi} = \w{\xi} + \Omega_{\Hor} \w{\eta}$ [Eq.~(\ref{e:ker:l_eqH_chi})
and (\ref{e:ker:def_chi})].
By definition of a
generator of a null hypersurface (cf. Sec.~\ref{s:def:null_geod_gen}),
$\w{p}$ is normal to $\Hor$. Since the Killing vector fields $\w{\xi}$ and $\w{\eta}$
are tangent to $\Hor$ (for $\Hor$ is globally preserved by the
spacetime symmetries), we get immediately from
the definitions (\ref{e:gek:def_E_L}) $E=0$ and $L=0$.
Similarly, if $\Li$ is a null geodesic generator of the
inner horizon $\Hor_{\rm in}$, $\Li$ is an outgoing principal null
geodesic $\Li^{{\rm out},\Hor_{\rm in}}_{(\th,\psi)}$, with $\w{p}$ proportional to
the null normal
$\wl \stackrel{\Hor_{\rm in}}{=} \w{\chi}_{\rm in} = \w{\xi} + \Omega_{\rm in} \w{\eta}$
[Eq.~(\ref{e:ker:chi_in_ell}) and (\ref{e:ker:def_chi_in})], so that we
have $E=0$ and $L=0$ as well. Hence we conclude:
\be \label{e:gek:generator_hor_E_L_zero}
\Li \ \mbox{null geodesic generator of}\ \Hor\ \mbox{or}\ \Hor_{\rm in}
\ \Longrightarrow\ E = 0 \quad\mbox{and}\quad L = 0 .
\ee
\end{example}
\begin{example}[Ingoing principal null geodesics] \label{x:gek:ingoing_null_E_L}
For the ingoing principal null geodesics $\Li^{\rm in}_{(v,\th,\tph)}$
introduced in Sec.~\ref{s:ker:principal_geod}, the 4-momentum vector
is
\be \label{e:gek:p_alpha_k}
\w{p} = \alpha \w{k} ,
\ee
where $\alpha$ is a (positive) constant, since $\w{k}$ is a geodesic vector
[Eq.~(\ref{e:ker:nab_k_k})], as $\w{p}$ [Eq.~(\ref{e:fra:p_geodesic})].
Equations~(\ref{e:gek:E_pt_L_pph}) with the components relative to Kerr coordinates
$(\ti, r, \th, \tph)$ lead then to
\[
E = - \alpha k_{\ti} = - \alpha(-1) = \alpha \qand
L = \alpha k_{\tph} = \alpha a \sin^2\th,
\]
where the components $k_{\ti}$ and $k_{\tph}$
have been read on Eq.~(\ref{e:ker:k_form_Kerr}).
Hence for any ingoing principal null geodesic $\Li^{\rm in}_{(v,\th,\tph)}$,
\be \label{e:gek:ingoing_null_E_L}
E > 0 \qand L = a E \sin^2 \th .
\ee
Recall that $\th$ is constant along $\Li^{\rm in}_{(v,\th,\tph)}$, so that
the above formula does yield a constant value for $L$. Moreover, it fulfills
$L\geq 0$ with $L=0$ only for $a=0$ or $\th\in\{0,\pi\}$ (rotation axis).
\end{example}
\begin{example}[Outgoing principal null geodesics] \label{x:gek:outgoing_null_E_L}
For the outgoing principal null geodesics $\Li^{\rm out}_{(u,\th,\tilde{\tph})}$,
$\Li^{{\rm out},\Hor}_{(\th,\psi)}$ and $\Li^{{\rm out},\Hor_{\rm in}}_{(\th,\psi)}$
introduced in Sec.~\ref{s:ker:principal_geod}, the 4-momentum vector
is
\be \label{e:gek:p_beta_l}
\w{p} = \beta(\lambda) \wl .
\ee
where $\beta(\lambda)$ is a function of the affine parameter $\lambda$
associated to $\w{p}$ that obeys $\beta(\lambda) > 0$, since both $\w{p}$
and $\wl$ are future-directed (cf. Sec.~\ref{s:ker:principal_geod}).
Contrary to the coefficient $\alpha$ in Eq.~(\ref{e:gek:p_alpha_k}), $\beta(\lambda)$
is not constant because $\wl$ is not a geodesic vector, but only
a pregeodesic one: it fulfills $\wnab_{\wl}\, \wl = \kappa_{\wl} \, \wl$
with $\kappa_{\wl}\neq 0$
[Eq.~(\ref{e:ker:pregeod_ell})] (cf. Remark~\ref{r:ker:k_versus_l}
on p.~\pageref{r:ker:k_versus_l}).
The geodesic equation $\wnab_{\w{p}}\, \w{p} = 0$ implies that
$\beta$ obeys the differential equation
$\wnab_{\w{p}} \beta + \kappa_{\wl} \beta^2 = 0$, or equivalently
\be \label{e:gek:ode_beta}
\derd{\beta}{\lambda} + \kappa_{\wl} \beta^2 = 0 .
\ee
The outgoing principal null geodesics either (i) are confined to one of the
horizons $\Hor$ and $\Hor_{\rm in}$, generating them (case of $\Li^{{\rm out},\Hor}_{(\th,\psi)}$ and $\Li^{{\rm out},\Hor_{\rm in}}_{(\th,\psi)}$), or (ii) never intersect them
(case of $\Li^{\rm out}_{(u,\th,\tilde{\tph})}$, cf. the solid curves in Figs.~\ref{f:ker:princ_null_geod_a90} -- \ref{f:ker:princ_null_geod_a50}).
In the first case, $\kappa_{\wl}$ is a constant on $\Hor$ and $\Hor_{\rm in}$,
given by Eq.~(\ref{e:ker:pregeod_ell}) with $r=r_+$ for $\Hor$ and $r=r_-$ for $\Hor_{\rm in}$.
The solution of Eq.~(\ref{e:gek:ode_beta}) is then
\[
\beta(\lambda) = \frac{1}{\kappa_{\wl}(\lambda - \lambda_0)},
\]
where $\lambda_0$ is a constant. Outside the horizons, $\kappa_{\wl}$ is the function
of $r$ given by Eq.~(\ref{e:ker:pregeod_ell}). We then search for a solution
of Eq.~(\ref{e:gek:ode_beta}) in the form $\beta(\lambda) = B(r(\lambda))$,
where $r(\lambda)$ is the function giving the coordinate $r$ along the geodesic.
Since the latter obeys $\D r/\D\lambda = p^r = \beta \ell^r$ with $\ell^r$ read
on Eq.~(\ref{e:ker:def_ell_outgoing}), we get the
following linear differential equation for $B$:
\[
(r^2 - 2 m r + a^2) \derd{B}{r} + 2 m \frac{r^2 - a^2}{r^2 + a^2} B = 0 .
\]
The solution is
\[
B(r) = 2\frac{r^2 + a^2}{r^2 - 2 m r + a^2} B_0 = \beta(\lambda),
\]
where $B_0$ is a positive constant in $\M_{\rm I}$ and $\M_{\rm III}$
and a negative constant in $\M_{\rm II}$, to ensure that $\beta(\lambda) > 0$.
Plugging (\ref{e:gek:p_beta_l}) into Eqs.~(\ref{e:gek:E_pt_L_pph}) with the components
relative to Kerr coordinates yields
\[
E = - \beta(\lambda) \ell_{\ti} = \beta(\lambda) \frac{\Delta}{2(r^2 + a^2)} \qand
L = \beta(\lambda) \ell_{\tph} = \beta(\lambda) \frac{a\Delta\sin^2\th}{2(r^2 + a^2)} ,
\]
where the components $\ell_{\ti}$ and $\ell_{\tph}$ have been
read on Eq.~(\ref{e:ker:ell_form_Kerr}). Given the above results for $\beta(\lambda)$,
we get $E=0$ and $L=0$ on $\Hor$ and $\Hor_{\rm in}$, since $\Delta = 0$ there,
and $E=B_0$ and $L = a B_0 \sin^2\th$ elsewhere. We conclude that
for the outgoing principal null geodesics,
\begin{subequations}
\label{e:gek:outgoing_null_E_L}
\begin{eqnarray}
& \mbox{in}\ \Hor\cup\Hor_{\rm in},\quad & E = 0 \qand L = 0 \label{e:gek:E_L_outgoing_hor} \\
& \mbox{in}\ \M_{\rm I}\cup\M_{\rm III},\quad & E > 0 \qand L = a E \sin^2 \th
\label{e:gek:E_L_outgoing_MI_III} \\
& \mbox{in}\ \M_{\rm II},\quad & E < 0 \qand L = a E \sin^2 \th . \label{e:gek:E_L_outgoing_MII}
\end{eqnarray}
\end{subequations}
In particular, (\ref{e:gek:E_L_outgoing_hor}) is nothing but the
result (\ref{e:gek:generator_hor_E_L_zero}) already obtained in Example~\ref{x:gek:null_generator_hor}.
We note also that the relation between $L$ and $E$ is identical to that obtained
in Example~\ref{x:gek:ingoing_null_E_L} for
the ingoing principal null geodesics [Eq.~(\ref{e:gek:ingoing_null_E_L})].
\end{example}
In what follows, we will use Boyer-Lindquist coordinates
$(x^\alpha)=(t,r,\th,\ph)$
as introduced in Sec.~\ref{s:ker:expr_BL}.
Given the components (\ref{e:ker:metric_BL}) of the metric tensor $\w{g}$
in these coordinates, evaluating $E$ and $L$
via $E = - g_{t\mu} \, p^\mu$ and $L = g_{\ph\mu} \, p^\mu$ yields
\be \label{e:gek:E_first_int}
E = \left( 1 - \frac{2 m r}{\rho^2} \right)\, p^t
+ \frac{2 a m r \sin^2\th}{\rho^2}\, p^\ph .
\ee
\be \label{e:gek:L_first_int}
L = - \frac{2 a m r \sin^2\th}{\rho^2} \, p^t
+ \left( r^2 + a^2 + \frac{2 a^2 m r \sin^2\th}{\rho^2} \right)
\sin^2\th \, p^\ph ,
\ee
where $\rho^2 := r^2 + a^2 \cos^2\th$ [Eq.~(\ref{e:ker:def_rho2})].
Let us recall that the components $(p^\alpha)$ of the 4-momentum are
related to the parametric equation $x^\alpha = x^\alpha(\lambda)$ of the geodesic $\Li$
in terms of the affine parameter $\lambda$ by $p^\alpha = \D x^\alpha / \D\lambda$
[cf. Eq.~(\ref{e:fra:p_dxdl})], i.e.
\be \label{e:gek:pa_der_xa}
p^t = \derd{t}{\lambda},\quad
p^r = \derd{r}{\lambda},\quad
p^\th = \derd{\th}{\lambda},\quad
p^\ph = \derd{\ph}{\lambda} .
\ee
\subsection{Mass as an integral of motion} \label{s:gek:mass_int_motion}
The mass $\mu$ of particle $\mathscr{P}$ is related to the scalar square of
the 4-momentum vector $\w{p}$ via Eq.~(\ref{e:fra:def_mass}):
\be \label{e:gek:mu_gpp}
\mu^2 = - \w{g}(\w{p}, \w{p}) .
\ee
The fact that $\Li$ is a geodesic implies that $\mu$ is constant along
$\Li$ (cf. Eq.~(\ref{e:geo:vv_const}) in Appendix~\ref{s:geo}). It therefore
provides a third integral of motion, after $E$ and $L$.
It is convenient to express (\ref{e:gek:mu_gpp}) in terms of the inverse metric
in order to let appear $p_t = -E$ and $p_\ph = L$:
\[
\mu^2 = - g^{\mu\nu} p_\mu p_\nu .
\]
Given the components (\ref{e:ker:inv_met_BL}) of the inverse metric
in Boyer-Lindquist coordinates, we get
\bea
\mu^2 & = & \frac{1}{\Delta}
\left( r^2 + a^2 + \frac{2 a^2 m r \sin^2\th}{\rho^2} \right) E^2
-\frac{4 a m r}{\rho^2 \Delta} E L
- \frac{1}{\Delta}\left(1 - \frac{2 m r}{\rho^2} \right) \frac{L^2}{\sin^2\th}
\nonumber \\
& & - \frac{\rho^2}{\Delta} \left( p^r \right) ^2
- \rho^2 \left( p^\theta \right) ^2 , \label{e:gek:mu2_first_int}
\eea
where $\Delta := r^2 - 2 m r + a^2$ [Eq.~(\ref{e:ker:def_Delta})].
Note that we have expressed $p_r$ and $p_\th$ in terms of $p^r$ and $p^\th$
thanks to the relations $p_r = g_{r\mu} p^\mu$ and $p_\th = g_{\th\mu} p^\mu$,
which are very simple for the Boyer-Lindquist components (\ref{e:ker:metric_BL})
of $\w{g}$:
\be \label{e:gek:p_r_p_th_cov_con}
p_r = \frac{\rho^2}{\Delta} \, p^r
\qquad\mbox{and}\qquad
p_\th = \rho^2 \, p^\th .
\ee
\subsection{The fourth integral of motion: Carter constant} \label{s:gek:Carter_const}
It turns out that the Kerr spacetime is endowed with a non-trivial
Killing tensor of valence 2: the
\defin{Walker-Penrose Killing tensor}\index{Killing!tensor!Walker-Penrose --}\index{Walker-Penrose Killing tensor} $\w{K}$, which is the symmetric tensor of type
$(0,2)$ defined by
\be \label{e:gek:def_K}
\encadre{ \w{K} := (r^2 + a^2) \left( \uu{k} \otimes \uu{\el} + \uu{\el} \otimes \uu{k} \right) + r^2 \w{g} } ,
\ee
where $\uu{\el}$ and $\uu{k}$ are the 1-forms associated by metric duality
to the null vector fields $\w{k}$ and $\wl$ tangent to the principal null
geodesics introduced in Sec.~\ref{s:ker:principal_geod}. In index notation,
Eq.~(\ref{e:gek:def_K}) writes
\be
K_{\alpha\beta} = (r^2 + a^2) \left(k_\alpha \el_\beta + \el_\alpha k_\beta \right)
+ r^2 g_{\alpha\beta} .
\ee
$\w{K}$ is called a \defin{Killing tensor}\index{Killing!tensor} because its symmetrized covariant derivative vanishes identically:
\be \label{e:gek:Killing_eq_K}
\encadre{ \nabla_{(\alpha} K_{\beta\gamma)} = 0 }.
\ee
This property can be seen as a generalization of the Killing equation\index{Killing!equation}
(\ref{e:neh:Killing_equation}) to tensors of valence 2.
That the tensor $\w{K}$ defined by (\ref{e:gek:def_K}) obeys the Killing
identity (\ref{e:gek:Killing_eq_K}) is established in the
SageMath notebook~\ref{s:sam:Kerr_Killing_tensor}.
Killing tensors are discussed in Sec.~\ref{e:geo:Killing_tensor} of Appendix~\ref{s:geo}.
It is shown there that the Killing identity (\ref{e:gek:Killing_eq_K}) implies that
the following quantity is constant along any geodesic $\Li$ [cf. Eq.~(\ref{e:geo:Kvv_const})]:
\be
\encadre{ \mathscr{K} := \w{K}(\w{p}, \w{p}) = K_{\mu\nu} p^\mu p^\nu } .
\ee
$\mathscr{K}$ is named \defin{Carter constant}\index{Carter!constant}
(cf. the historical note on p.~\pageref{h:gek:eom_Carter}).
From the definition (\ref{e:gek:def_K}) of $\w{K}$, we have
\be \label{e:gek:Kcarter_prov}
\mathscr{K} = 2(r^2 + a^2) \langle \uu{k}, \w{p} \rangle
\langle \uu{\el}, \w{p} \rangle + r^2 \w{g}(\w{p}, \w{p}).
\ee
Now by Eq.~(\ref{e:gek:mu_gpp}), $\w{g}(\w{p}, \w{p}) = - \mu^2$. Besides, we
have $\langle \uu{k}, \w{p} \rangle = k_\mu p^\mu = k^\mu p_\mu$.
The last form lets appear the constants of motion $p_t = -E$ and
$p_\ph = L$ [Eq.~(\ref{e:gek:E_pt_L_pph})]. Using it with the components
of $\w{k}$ as given by Eq.~(\ref{e:ker:k_BL}), we get
\[
\langle \uu{k}, \w{p} \rangle = - \frac{r^2 + a^2}{\Delta} E
- p_r + \frac{a}{\Delta} L .
\]
Similarly, from the components (\ref{e:ker:ell_BL}) of $\wl$, we obtain
\[
\langle \uu{\el}, \w{p} \rangle = - \frac{1}{2} E + \frac{\Delta}{2(r^2 + a^2)} p_r
+ \frac{a}{2(r^2 + a^2)} L .
\]
Accordingly, Eq.~(\ref{e:gek:Kcarter_prov}) becomes
\[
\mathscr{K} = \frac{1}{\Delta} \left[ (r^2 + a^2) E + \Delta p_r - a L \right]
\left[ (r^2 + a^2) E - \Delta p_r - a L \right] - r^2 \mu^2 ,
\]
which can be rewritten as
\be \label{e:gek:Kcarter_first_int}
\encadre{ \mathscr{K} = \frac{1}{\Delta} \left[ \left( (r^2 + a^2) E - a L \right)^2
- \rho^4 (p^r)^2 \right] - r^2 \mu^2 }.
\ee
Note that we have expressed $p_r$ in terms of $p^r$ via Eq.~(\ref{e:gek:p_r_p_th_cov_con}).
Some physical interpretation of the Carter constant can be inferred from the above
expression, in the case where the particle $\mathscr{P}$ following the
geodesic $\Li$ visits the asymptotic region $r\to+\infty$. Indeed, given
that $\Delta := r^2 - 2m r + a^2 \sim r^2$ and $\rho^4 := (r^2 + a^2\cos^2\th)^2 \sim r^4$
as $r\to+\infty$, we deduce from Eq.~(\ref{e:gek:Kcarter_first_int})
that
\be \label{e:gek:Kcarter_r_inf}
\mathscr{K} \underset{r\to + \infty}{\sim} r^2 \left[ E^2 - \mu^2 - (p^r)^2 \right] .
\ee
As discussed above, $E$ is $\mathscr{P}$'s energy as measured by
the asymptotic inertial observer $\Obs$. Then, according to
Einstein's formula (\ref{e:fra:E2_m2_P2}), $E^2 - \mu^2 = \w{P}\cdot\w{P}$, where
$\w{P}$ is $\mathscr{P}$'s linear momentum as measured by $\Obs$. Given
that asymptotically, $p^r\sim P^r$ [cf. Eq.~(\ref{e:fra:p_E_P})],
Eq.~(\ref{e:gek:Kcarter_r_inf}) becomes
\[
\mathscr{K} \underset{r\to + \infty}{\sim} r^2 \left[ \w{P}\cdot\w{P} - (P^r)^2 \right]
= r^2 \left[ (P^{(\th)})^2 + (P^{(\ph)})^2 \right],
\]
where $P^{(\th)} = r P^\th \sim r p^\th$ and $P^{(\ph)} = r\sin\th\, P^\ph \sim r\sin\th\, p^\ph$
are the angular components of $\w{P}$ in the
orthonormal basis $(\w{e}_{(r)}, \w{e}_{(\th)}, \w{e}_{(\ph)}) := (\wpar_r, r^{-1} \wpar_\th,
(r\sin\th)^{-1}\wpar_\ph)$.
Now the total angular momentum of $\mathscr{P}$ measured by $\Obs$ is
\[
\w{L}_{\rm tot} := \w{r} \times \w{P} = - r P^{(\ph)} \w{e}_{(\th)}
+ r P^{(\th)} \w{e}_{(\ph)} .
\]
Hence we may conclude that asymptotically, the Carter constant coincides with
the square of $\mathscr{P}$'s angular momentum as measured by the inertial observer $\Obs$:
\be \label{e:gek:Kcarter_asymptot}
\mathscr{K} \underset{r\to + \infty}{\sim} \w{L}_{\rm tot} \cdot \w{L}_{\rm tot} .
\ee
We shall see later that one has always $\mathscr{K}\geq 0$. For now, let us
establish the following characterization of null geodesics with $\mathscr{K} = 0$:
\begin{prop}[null geodesics with vanishing Carter constant]
\label{p:gek:null_geod_K_0}
A null geodesic $\Li$ has a vanishing Carter constant $\mathscr{K}$ if, and
only if, $\Li$ is a principal null geodesic:
\be \label{e:gek:K_zero_null}
\mathscr{K} = 0 \iff \Li = \Li^{\rm in}_{(v,\th,\tph)} \quad\mbox{or}\quad
\Li = \Li^{\rm out},
\ee
where $\Li^{\rm in}_{(v,\th,\tph)}$ (resp. $\Li^{\rm out}$)
is one of the ingoing (resp. outgoing) principal null geodesic
introduced in Sec.~\ref{s:ker:principal_geod}, with $\Li^{\rm out}$ standing
for $\Li^{\rm out}_{(u,\th,\tilde{\tph})}$, $\Li^{{\rm out},\Hor}_{(\th,\psi)}$
or $\Li^{{\rm out},\Hor_{\rm in}}_{(\th,\psi)}$.
\end{prop}
\begin{proof}
Consider expression (\ref{e:gek:Kcarter_prov}) for $\mathscr{K}$. If $\Li$
is a null geodesic, the term $\w{g}(\w{p},\w{p})$ vanishes identically, so that
one is left with
\[
\mathscr{K} = 2 (r^2 + a^2) (\w{k}\cdot\w{p}) (\wl\cdot\w{p}) .
\]
Given that $r^2 + a^2\neq 0$ (this is clear for $a\neq 0$, while if $a=0$
(Schwarzschild case),
$r=0$ is excluded from the spacetime manifold), we have then
\[
\mathscr{K} = 0 \iff \w{k}\cdot\w{p} = 0 \quad\mbox{or}\quad \wl\cdot\w{p} = 0 .
\]
The vectors $\w{k}$, $\wl$ and $\w{p}$ are all null.
Now, according to Property~\ref{p:fra:corol2} in Sec.~\ref{s:fra:time_orientation},
two null vectors are orthogonal iff they are collinear. Hence $\mathscr{K} = 0$
is equivalent to $\w{p} = \alpha \w{k}$ or $ \w{p} = \alpha \wl$ with $\alpha > 0$.
Since $\w{k}$ (resp. $\wl$) is tangent to $\Li^{\rm in}_{(v,\th,\tph)}$ (resp.
$\Li^{\rm out}$), this completes the proof.
\end{proof}
\begin{remark}
Property~\ref{p:gek:null_geod_K_0} is consistent with the
interpretation (\ref{e:gek:Kcarter_asymptot}) of $\mathscr{K}$, since asymptotically
the principal null geodesics are purely radial\footnote{Indeed,
we deduce from Eqs.~(\ref{e:ker:k_BL}) and (\ref{e:ker:ell_BL}) that, for $r\to+\infty$,
$\w{k} \sim \wpar_t - \wpar_r$ and $\wl \sim (\wpar_t + \wpar_r)/2$.},
and hence have $\w{L}_{\rm tot} = 0$.
\end{remark}
\begin{example}[Generators of the event and inner horizons]
We have seen in Example~\ref{x:gek:null_generator_hor} in Sec.~\ref{s:gek:int_motion_sym}
that the null geodesic generators of the event horizon $\Hor$ and the
inner horizon $\Hor_{\rm in}$ have $E=0$ and $L=0$.
Since these generators are principal null geodesics (cf.
Secs.~\ref{s:gek:null_gen_hor} and \ref{s:ker:Cauchy_hor}), the above results shows
that in addition $\mathscr{K}=0$. As $\mu=0$ by definition of a null geodesic,
we see that the four integrals of motion $\mu$, $E$, $L$ and $\mathscr{K}$ all
vanish for these geodesics.
\end{example}
The converse is true:
\begin{prop}[characterization of the null generators of the two horizons]
The null geodesic generators of the event horizon $\Hor$ and of the
inner horizon $\Hor_{\rm in}$ are the only geodesics of Kerr spacetime
having their four integrals of motion vanishing:
\be \label{e:gek:all_const_zero}
\Li \ \mbox{null geodesic generator of}\ \Hor\ \mbox{or}\ \Hor_{\rm in} \iff
(\mu, E, L, \mathscr{K}) = (0, 0, 0, 0) .
\ee
\end{prop}
\begin{proof}
If a geodesic $\Li$ has $(\mu,\mathscr{K}) = (0,0)$,
$\Li$ is necessary null and (\ref{e:gek:K_zero_null}) shows that
$\Li$ is a principal null geodesic. Moreover
Eq.~(\ref{e:gek:Kcarter_first_int}) with $(\mu, E, L, \mathscr{K}) = (0, 0, 0, 0)$
implies $p^r = 0$, i.e. $\Li$ lies at a constant value of $r$.
If $\Li$ is ingoing, then $\w{p} \propto \w{k}$, with the Boyer-Lindquist components
of $\w{k}$ given by Eq.~(\ref{e:ker:k_BL}). Since $k^r = -1$, this precludes
$p^r = 0$. Hence $\Li$ is an outgoing principal null geodesic
and one has $\w{p} \propto \wl$, with the Boyer-Lindquist components
of $\wl$ given by Eq.~(\ref{e:ker:ell_BL}). We read $\ell^r = \Delta / (2(r^2+a^2))$
with $\Delta\neq 0$, except precisely on $\Hor$ and
$\Hor_{\rm in}$\footnote{This is graphically confirmed by
Figs.~\ref{f:ker:princ_null_geod_a90} and
\ref{f:ker:princ_null_geod_a50}, which show that $\Hor$ and $\Hor_{\rm in}$ are the only
locations where a principal null geodesic can have $r=\mathrm{const}$.}. Hences
$p^r = 0$ is possible only on $\Hor$ and $\Hor_{\rm in}$.
\end{proof}
\begin{hist}
That $\mathscr{K} = 0$ for principal null geodesics has been pointed out
by Ji\v{r}\'{i} Bi\v{c}\'{a}k\index[pers]{Bi\v{c}\'{a}k, J.} and
Zden\v{e}k Stuchl\'{\i}k\index[pers]{Stuchl\'{\i}k, Z.}
in 1976 \cite{BicakS76}.
\end{hist}
\subsection{First order equations of motion} \label{s:gek:first_order_system}
We have thus four first integrals of the geodesic equation
(\ref{e:gek:eq_geod}) at disposal:
$E$ [Eq.~(\ref{e:gek:E_first_int})], $L$ [Eq.~(\ref{e:gek:L_first_int})],
$\mu^2$ [Eq.~(\ref{e:gek:mu2_first_int})] and $\mathscr{K}$
[Eq.~(\ref{e:gek:Kcarter_first_int})]. In the expressions of each of these integrals,
$p^\alpha$ has to be thought of as the first order derivative $\D x^\alpha/\D\lambda$
[Eq.~(\ref{e:gek:pa_der_xa})].
Two first integrals, namely $E$ and $L$, are linear in the $p^\alpha$'s, while
the two others, namely $\mu^2$ and $\mathscr{K}$, are quadratic.
Furthermore, Eqs.~(\ref{e:gek:E_first_int}) and (\ref{e:gek:L_first_int})
constitute a decoupled subsystem for $(p^t, p^\ph)$, which can easily be solved\footnote{An
intermediate step is combining Eqs.~(\ref{e:gek:E_first_int}) and (\ref{e:gek:L_first_int})
to get $a E - L /\sin^2\th = a p^t - (r^2 + a^2) p^\ph$
and $(r^2 + a^2) E - a L = \Delta(p^t - a\sin^2\th p^\ph)$.},
yielding
\be \label{e:gek:rho2_pt}
\rho^2 p^t = \frac{1}{\Delta} [(r^2 + a^2)^2 E - 2a m r L] - a^2 E \sin^2\th
\ee
\be \label{e:gek:rho2_pph}
\rho^2 p^\ph = \frac{L}{\sin^2\th}
+ \frac{a}{\Delta} ( 2 m r E - a L).
\ee
Besides, Eq.~(\ref{e:gek:Kcarter_first_int}) involves only $p^r$ and can
be recast as
\be \label{e:gek:p_r_R}
\rho^4 (p^r)^2 = R(r) ,
\ee
with
\be \label{e:gek:def_R}
\encadre{ R(r) := \left[ (r^2 + a^2) E - a L \right]^ 2 - \Delta (r^2 \mu^2 + \mathscr{K}) }.
\ee
Recall that $\Delta$ is the function of $r$
given by Eq.~(\ref{e:ker:def_Delta}): $\Delta := r^2 - 2 m r + a^2$. All other
quantities appearing in Eq.~(\ref{e:gek:def_R}) are constant. Accordingly, $R(r)$ is a 4th order polynomial in $r$.
Equation~(\ref{e:gek:p_r_R}) implies that this polynomial must be non-negative along the geodesic $\Li$:
\be \label{e:gek:R_non_neg}
\encadre{R(r) \geq 0 } .
\ee
Finally, if we substitute $p^r$ by the value given by Eqs.~(\ref{e:gek:p_r_R})-(\ref{e:gek:def_R}) in the mass first integral (\ref{e:gek:mu2_first_int}), we get, after simplification,
\be \label{e:gek:p_th_Th}
\rho^4 (p^\th)^2 = \Theta(\th) ,
\ee
with
\be \label{e:gek:def_Theta}
\encadre{ \Theta(\th) := \mathscr{K} - \left(\frac{L}{\sin\th} - a E \sin\th \right)^2
- \mu^2 a^2 \cos^2\th } .
\ee
Equation.~(\ref{e:gek:p_th_Th}) imposes that $\Theta(\th)$ is non-negative along
the geodesic $\Li$:
\be \label{e:gek:Theta_non_neg}
\encadre{\Theta(\th) \geq 0 } .
\ee
The following constant is often used instead of $\mathscr{K}$:
\be \label{e:gek:def_Q}
\encadre{Q := \mathscr{K} - (L - a E)^2 }
\ee
Thanks to it, we may rewrite (\ref{e:gek:def_Theta}) as
\be \label{e:gek:Theta_Q}
\encadre{ \Theta(\th) = Q + \cos^2\th \left[ a^2 (E^2 - \mu^2)
- \frac{L^2}{\sin^2\th} \right] } .
\ee
Following the standard usage, we call $Q$ the \defin{Carter constant}\index{Carter!constant}
as well. To distinguish between the two Carter constants, we shall specify \emph{Carter constant $Q$}
or \emph{Carter constant $\mathscr{K}$}. As we shall see in Sec.~\ref{s:gek:th_motion},
$Q$ is well adapted to the description of the $\th$-motion of geodesics. On the other
hand, a nice property of $\mathscr{K}$, which is not shared by $Q$, is to be always non-negative,
as Eqs.~(\ref{e:gek:def_Theta}) and (\ref{e:gek:Theta_non_neg}) show:
\be \label{e:ges:K_non_negative}
\encadre{ \mathscr{K} \geq 0 }.
\ee
If the particle $\mathscr{P}$ reaches the asymptotic region $r\gg m$, we deduce from
Eqs.~(\ref{e:gek:def_Q}) and (\ref{e:gek:Kcarter_asymptot}) the following behavior of $Q$:
\be \label{e:gek:Q_Ltot2_L2}
Q \underset{r\to + \infty}{\sim} \w{L}_{\rm tot} \cdot \w{L}_{\rm tot}
- L^2 + a E (2 L - a E) .
\ee
Hence, if $a=0$, $Q$ can be interpreted as the square of
the part of $\mathscr{P}$'s angular momentum (measured by the asymptotic inertial
observer) that is not in $L$.
\begin{example}[Carter constant $Q$ of the principal null geodesics]
As (\ref{e:gek:K_zero_null}) shows, for a principal null geodesic, be it ingoing
or outgoing, the Carter constant $\mathscr{K}$ vanishes identically. According
to Eq.~(\ref{e:gek:def_Q}), the
Carter constant $Q$ is then $Q = - (L - aE)^2$. In view of the relation
$L = a E \sin^2\th$ for these geodesics [Eqs.~(\ref{e:gek:ingoing_null_E_L}) and
(\ref{e:gek:outgoing_null_E_L})], we get
\be \label{e:gek:Q_principal_null}
Q = - a^2 E^2 \cos^4\th ,
\ee
where $\th$ is the constant value of the $\th$-coordinate along the principal
null geodesic. Note that the above relation holds in all Kerr spacetime,
including on the horizons $\Hor$ and $\Hor_{\rm in}$, where it reduces to $Q=0$, for
$E=0$ there [Eq.~(\ref{e:gek:E_L_outgoing_hor})]. Equation~(\ref{e:gek:Q_principal_null})
implies
\be
Q \leq 0 ,
\ee
with $Q=0$ only for principal null geodesics lying in the equatorial plane or
for the outgoing principal null geodesics $\Li^{{\rm out},\Hor}_{(\th,\psi)}$ and $\Li^{{\rm out},\Hor_{\rm in}}_{(\th,\psi)}$ generating the horizons $\Hor$ and $\Hor_{\rm in}$.
\end{example}
\begin{remark}
As for $\mathscr{K}$,
one may derive the Carter constant $Q$ from a Killing tensor.
Indeed, from the Walker-Penrose Killing tensor $\w{K}$ [Eq.~(\ref{e:gek:def_K})], let us form
the tensor field
\be \label{e:gek:def_tilde_K}
\w{\tilde{K}} := \w{K} - \uu{\tilde{\eta}}\otimes\uu{\tilde{\eta}} ,\quad
\mbox{where}\quad
\w{\tilde{\eta}} := \w{\eta} + a \w{\xi} .
\ee
Being a linear combination with constant coefficients of the Killing vectors
$\w{\eta}$ and $\w{\xi}$, $\w{\tilde{\eta}}$ is itself a Killing vector.
It follows that $\uu{\tilde{\eta}}\otimes\uu{\tilde{\eta}}$ is a Killing tensor
(cf. Example~\ref{x:geo:trivial_Killing_tensors} in Sec.~\ref{e:geo:Killing_tensor}),
so that $\w{\tilde{K}}$ a Killing tensor, the Killing equation (\ref{e:gek:Killing_eq_K})
being linear. Applying $\w{\tilde{K}}$ to the
4-momentum $\w{p}$, we get
\[
\w{\tilde{K}}(\w{p},\w{p}) = \underbrace{\w{K}(\w{p},\w{p})}_{\mathscr{K}}
- (\underbrace{\langle\uu{\eta},\w{p}\rangle}_{L}
+ a \underbrace{\langle\uu{\xi},\w{p}\rangle}_{-E})^2 .
\]
Comparing with the definition (\ref{e:gek:def_Q}), we conclude that
\be \label{e:gek:Q_tK_pp}
Q = \w{\tilde{K}}(\w{p},\w{p}) .
\ee
The Boyer-Lindquist components of the contravariant tensor associated to $\w{\tilde{K}}$ by
metric duality have an expression particularly simple in terms of those of
the inverse metric (cf. the notebook~\ref{s:sam:Kerr_Killing_tensor}):
\be
\tilde{K}^{\alpha\beta} = - a^2 \cos^2 \theta \, g^{\alpha\beta}
+ \operatorname{diag}(-a^2\cos^2\theta, 0, 1, \tan^{-2}\theta)^{\alpha\beta} .
\ee
\end{remark}
In view of the relation (\ref{e:gek:pa_der_xa}) between the $p^\alpha$'s
and the derivatives of the functions $x^\alpha(\lambda)$, we may
collect Eqs.~(\ref{e:gek:rho2_pt}), (\ref{e:gek:rho2_pph}), (\ref{e:gek:p_r_R})
and (\ref{e:gek:def_Theta}) as the first-order system
\begin{subequations}
\label{e:gek:eom_first_order}
\begin{align}
& \encadre{ \rho^2 \derd{t}{\lambda} = \frac{1}{\Delta} [(r^2 + a^2)^2 E - 2a m r L] - a^2 E \sin^2\th } \label{e:gek:dtdl} \\
& \encadre{ \rho^2 \derd{r}{\lambda} = \eps_r \sqrt{ R(r) } } \label{e:gek:drdl}\\
& \encadre{ \rho^2 \derd{\th}{\lambda} = \eps_\th \sqrt{\Theta(\th)} } \label{e:gek:dthdl}\\
& \encadre{ \rho^2 \derd{\ph}{\lambda} = \frac{L}{\sin^2\th}
+ \frac{a}{\Delta}(2 m r E - a L) } , \label{e:gek:dphdl}
\end{align}
\end{subequations}
where $\eps_r := \operatorname{sgn} p^r = \pm 1$, $\eps_\th := \operatorname{sgn} p^\th = \pm 1$
and the functions $R(r)$ and $\Theta(\th)$ are defined by Eq.~(\ref{e:gek:def_R})
and Eq.~(\ref{e:gek:def_Theta}) or (\ref{e:gek:Theta_Q}).
Since $p^r = \D r/\D\lambda$, $\eps_r$
is $+1$ (resp. $-1$) in the parts of the geodesic $\Li$ where $r$ increases
(resp. decreases) with $\lambda$. Similarly, $\eps_\th$
is $+1$ (resp. $-1$) in the parts of $\Li$ where $\th$ increases
(resp. decreases) with $\lambda$.
We may rewrite Eq.~(\ref{e:gek:def_R}) for $R(r)$ in terms of $Q$
instead of $\mathscr{K}$, via Eq.~(\ref{e:gek:def_Q}):
\be \label{e:gek:def_R_Q}
\encadre{ R(r) := \left[ (r^2 + a^2) E - a L \right]^ 2 - \Delta \left[ r^2 \mu^2 + Q +(L-a E)^2) \right] }.
\ee
Note that, beside the constants of motion $E$, $L$, $\mu$ and $Q$, $R(r)$ depends on both Kerr parameters $a$ and $m$
(via $\Delta = r^2 - 2 m r + a^2$), while $\Theta(\th)$
depends on $a$ only [cf. Eq.~(\ref{e:gek:Theta_Q})].
Along $\Li$, these functions must obey
$R(r) \geq 0$ [Eq.~(\ref{e:gek:R_non_neg})] and $\Theta(\th) \geq 0$
[Eq.~(\ref{e:gek:Theta_non_neg})].
\subsection{Turning points} \label{s:gek:turning_points}
Let $\Li$ be a geodesic that is not stuck at some constant value of the
coordinate $r$.
We define a \defin{$r$-turning point}\index{turning point!$r$-turning point} of $\Li$ as a point $p_0\in\Li$, the $r$-coordinate $r_0$ of which obeys
\be \label{e:gek:def_r_turning}
R(r_0) = 0 \quad\mbox{and}\quad R'(r_0) \neq 0 ,
\ee
i.e. $r_0$ is a simple root of the polynomial $R$.
We have then
\be \label{e:gek:der_r_turning}
\left. \derd{r}{\lambda} \right| _{\lambda_0} = 0
\quad\mbox{and}\quad
\left. \frac{\D^2 r}{\D\lambda^2} \right| _{\lambda_0} = \frac{R'(r_0)}{2\rho_0^4} \neq 0 ,
\ee
where $\lambda_0$ is the value of the affine parameter $\lambda$ at $p_0$
and $\rho_0 := \rho(p_0)$.
\begin{proof}
The vanishing of $\D r/\D\lambda$ at $\lambda_0$ follows immediately
from Eq.~(\ref{e:gek:drdl}) with $R(r_0)=0$, since $\rho^2$ never vanishes
on $\M$. Besides, by
taking the derivative of Eq.~(\ref{e:gek:drdl}) with respect to $\lambda$, we
get
\[
2 \left(r \derd{r}{\lambda} + a^2 \cos\th\sin\th \derd{\th}{\lambda} \right)
\derd{r}{\lambda} + \rho^2 \frac{\D^2 r}{\D\lambda^2} =
\eps_r \frac{R'(r)}{2\sqrt{R(r)}} \derd{r}{\lambda} = \frac{R'(r)}{2\rho^2} .
\]
At $\lambda=\lambda_0$, the first term in left-hand side vanishes identically,
due to the $\D r/\D\lambda$ factor,
and we get the second part of (\ref{e:gek:der_r_turning}).
\end{proof}
We deduce from the result (\ref{e:gek:der_r_turning}) that at $\lambda=\lambda_0$,
$\D r/\D\lambda$ moves from positive to negative values or vice-versa
(depending on the sign of $R'(r_0)$), which means that the function $r(\lambda)$
switches from increasing to decreasing or vice-versa, hence the name
\emph{$r$-turning point}. The factor $\eps_r=\pm 1$ in Eq.~(\ref{e:gek:drdl}) necessarily
changes sign at $\lambda=\lambda_0$.
\begin{remark} \label{r:gek:der_r_analytic}
For a generic smooth function $r(\lambda)$ with $\D r/\D \lambda=0$ at
$\lambda_0$, the condition $\D^2 r/\D\lambda^2\neq 0$ at $\lambda_0$
is sufficient but not necessary for $r$ to change its direction of variation there.
Indeed, the same property holds with $\D^2 r/\D\lambda^2= 0$ and higher order derivatives vanishing up to some even order $k$ for which
$\D^k r/\D\lambda^k\neq 0$. However, in the present case,
$\D^2 r/\D\lambda^2=0$ would imply $R'(r_0)=0$ and we shall see
Sec.~\ref{s:gek:asymptotic_values} that a geodesic can reach such a point only asymptotically,
i.e. for $\lambda\to +\infty$. Hence it cannot be a turning point.
\end{remark}
Similarly, if $\Li$ is a geodesic that is not stuck at some constant value of the
coordinate $\th$, we define a \defin{$\th$-turning point}\index{turning point!$\th$-turning point} of $\Li$ as a point $p_0\in\Li$, the
$\th$-coordinate $\th_0$ of which obeys
\be \label{e:gek:def_th_turning}
\Theta(\th_0) = 0 \quad\mbox{and}\quad \Theta'(\theta_0) \neq 0 .
\ee
We deduce then from the equation of motion (\ref{e:gek:dthdl}):
\be \label{e:gek:der_th_turning}
\left. \derd{\th}{\lambda} \right| _{\lambda_0} = 0
\quad\mbox{and}\quad
\left. \frac{\D^2 \th}{\D\lambda^2} \right| _{\lambda_0} = \frac{\Theta'(\th_0)}{2\rho_0^4} \neq 0 .
\ee
This implies that at a $\th$-turning point, the function $\th(\lambda)$
switches from increasing to decreasing or vice-versa. The factor $\eps_\th=\pm 1$ in Eq.~(\ref{e:gek:dthdl}) necessarily
changes sign at such a point.
\begin{remark}
A comment similar to Remark~\ref{r:gek:der_r_analytic} can be made:
a geodesic with varying $\th$ that has $\D\th/\D\lambda = 0$ for some finite
value of $\lambda$
cannot have $\D^2\th/\D\lambda^2 = 0$ at the same point, since we shall see
in Sec.~\ref{s:gek:asymptotic_values}
that a value of $\th$ with both $\Theta(\th)=0$
and $\Theta'(\th)=0$ can only be reached asymptotically along a geodesic.
Hence (\ref{e:gek:def_th_turning})
is a necessary and sufficient condition for a $\th$-turning point.
\end{remark}
\subsection{Equations of motion in terms of Mino parameter} \label{s:gek:Mino_time}
In view of the right-hand sides of the system (\ref{e:gek:eom_first_order}),
it is quite natural to introduce a new
parameter $\lambda'$
along the geodesic $\Li$ such that
\be \label{e:gek:def_Mino_time}
\D\lambda' = \frac{\D\lambda}{\rho^2}
= \frac{\D\lambda}{r(\lambda)^2 + a^2 \cos^2 \th(\lambda)} .
\ee
Since $\rho^2$ never vanishes on the spacetime manifold $\M$ (by construction
of the latter, cf. Eq.~(\ref{e:ker:def_M_Kerr_spacetime})), the above relation
leads to a well-defined parameter\footnote{Note however that $\lambda'$ may blow up
if $\Li$ comes arbitrarily close to the ring singularity, i.e. if $\rho\to 0$.} along $\Li$.
Moreover, since $\rho^2>0$, $\lambda'$
increases towards the future, as $\lambda$. A difference between
the two parametrizations is that $\lambda'$ is not an affine parameter
of $\Li$ in general\footnote{The only
exception is for a circurlar orbit at $\theta=\pi/2$, since then $\rho^2$
is constant and Eq.~(\ref{e:gek:def_Mino_time}) reduces to an affine relation
between $\lambda$ and $\lambda'$.}, contrary to $\lambda$.
The parameter $\lambda'$ is called \defin{Mino parameter}\index{Mino!parameter}\index{parameter!Mino --} \cite{Mino03}.
In terms of Mino parameter, the system (\ref{e:gek:eom_first_order}) becomes
\begin{subequations}
\label{e:gek:eom_Mino}
\begin{align}
& \encadre{ \derd{t}{\lambda'} = \frac{1}{\Delta} [(r^2 + a^2)^2 E - 2a m r L] - a^2 E \sin^2\th } \label{e:gek:dtdl_Mino} \\
& \encadre{ \derd{r}{\lambda'} = \eps_r \sqrt{ R(r) } } \label{e:gek:drdl_Mino}\\
& \encadre{ \derd{\th}{\lambda'} = \eps_\th \sqrt{\Theta(\th)} } \label{e:gek:dthdl_Mino}\\
& \encadre{ \derd{\ph}{\lambda'} = \frac{L}{\sin^2\th}
+ \frac{a}{\Delta}(2 m r E - a L) } , \label{e:gek:dphdl_Mino}
\end{align}
\end{subequations}
where $R$ is the quartic polynomial defined by Eq.~(\ref{e:gek:def_R_Q})
and $\Theta$ is the function defined by Eq.~(\ref{e:gek:Theta_Q}).
It is remarkable that Eqs.~(\ref{e:gek:drdl_Mino}) and (\ref{e:gek:dthdl_Mino})
are two fully decoupled equations: Eq.~(\ref{e:gek:drdl_Mino}) is an ordinary
differential equation (ODE) for the function $r(\lambda')$, while Eq.~(\ref{e:gek:dthdl_Mino})
is an ordinary differential equation for the function $\th(\lambda')$. This was
not the case for Eqs.~(\ref{e:gek:drdl}) and (\ref{e:gek:dthdl}) since $\rho^2$ involves
both $r$ and $\th$.
\subsection{Integration of the geodesic equations} \label{e:gek:integration}
The ODE (\ref{e:gek:drdl_Mino}) can be integrated by the method of separation
of variables. On a part of $\Li$ where $R(r)\neq 0$, this yields
\be \label{e:gek:lp_int_dr_ov_R}
\lambda' - \lambda'_0 = \int_{r_0}^r \frac{\eps_r\, \D \bar{r}}{\sqrt{R(\bar{r})}} ,
\ee
with $r_0 := r(\lambda'_0)$.
The hypothesis $R(r)\neq 0$
excludes any $r$-turning point between $\lambda'_0$ and $\lambda'$, so that
$\eps_r=\pm 1$ is constant along
the considered part of $\Li$.
Actually, the solution (\ref{e:gek:lp_int_dr_ov_R})
can be extended to include a turning point at one or two of its boundaries,
despite $R(r)=0$ there.
Indeed, let us assume that $r=r_1$ corresponds to a $r$-turning point of $\Li$.
Due to $R'(r_1)\neq 0$ [Eq.~(\ref{e:gek:def_r_turning})], the
integral in the right-hand side of (\ref{e:gek:lp_int_dr_ov_R}) with $r=r_1$
is finite. Indeed,
the Taylor expansion $R(\bar{r}) = R'(r_1) (\bar{r} - r_1) + O((\bar{r} - r_1)^2)$
makes the integral behave near $r_1$
as\footnote{We assume here $r_0<r_1$, so that the constraint
$R(\bar{r})\geq 0$ on the interval $[r_0,r_1]$ implies $R'(r_1)<0$.}
\[
\frac{1}{\sqrt{-R'(r_1)}} \int_{r_0}^{r_1} \frac{\D \bar{r}}{\sqrt{r_1 - \bar{r}}} ,
\]
which is a convergent improper integral.
\begin{remark}
This is the same reasoning as in Sec.~\ref{s:gis:deflect_winding}.
\end{remark}
Let us assume that there are $M\geq 1$ $r$-turning points $p_1,\ldots, p_M$ between
$\lambda'_0$ and $\lambda'$. Their radial coordinates take at most two distinct values, $r_1$ and $r_2$, such
that $r(p_1) = r_1$, $r(p_2) = r_2$, $r(p_3) = r_1$, $r(p_4) = r_2$, etc.
From the above convergence property, the solution of Eq.~(\ref{e:gek:drdl_Mino})
is then
\be \label{e:gek:Minotime_r_sum}
\lambda' - \lambda'_0 = \int_{r_0}^{r_1} \frac{\eps_r\, \D \bar{r}}{\sqrt{R(\bar{r})}}
+ (M-1) \int_{r_1}^{r_2} \frac{\eps_r\, \D \bar{r}}{\sqrt{R(\bar{r})}}
+ \int_{r_{1,2}}^r \frac{\eps_r\, \D \bar{r}}{\sqrt{R(\bar{r})}} ,
\ee
where $r_{1,2} = r_1$ for $M$ odd and $r_{1,2} = r_2$ for $M$ even.
Note that each term in the above sum
is positive, the sign of $\eps_r$ compensating the order of the integral
boundaries.
The right-hand side of Eq.~(\ref{e:gek:Minotime_r_sum}) is actually a path
integral and is often abriged by means of a slash notation:
\be \label{e:gek:Minotime_r_slash_int}
\lambda' - \lambda'_0 = \dashint_{r_0}^r \frac{\eps_r \, \D \bar{r}}{\sqrt{R(\bar{r})}} .
\ee
Similarly, if $\Theta(\th)\neq 0$ along $\Li$, except possibly
at some $\th$-turning points,
Eq.~(\ref{e:gek:dthdl_Mino}) can be integrated as
\bea
\lambda' - \lambda'_0 & = & \dashint_{\th_0}^\th \frac{\eps_\th \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}} \nonumber \\
& = & \begin{cases}
\displaystyle
\int_{\th_0}^\th \frac{\eps_\th\, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}} & \ \mbox{if}\ N=0 \\[3ex]
\displaystyle
\int_{\th_0}^{\th_1} \frac{\eps_\th\, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
+ (N-1)
\int_{\th_1}^{\th_2} \frac{\eps_\th\, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
+ \int_{\th_{1,2}}^{\th} \frac{\eps_\th\, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
& \ \mbox{if}\ N\geq 1 ,
\end{cases} \nonumber \\
& & \label{e:gek:Minotime_th_slash_int}
\eea
where $\th_0 := \th(\lambda'_0)$,
$N$ is the number of $\th$-turning points between $\lambda'_0$ and $\lambda'$,
$\theta_1$ (resp. $\theta_2$) is the value of $\th$ at the first (resp. second)
turning point, if any, $\th_{1,2} = \th_1$ for $N$ odd and $\th_{1,2} = \th_2$ for $N$ even.
We are now in position to state the full expression of the general
solution for geodesic motion:
\begin{prop}[solution for geodesic motion]
Let $\Li$ be a null or timelike geodesic in Kerr spacetime,
with conserved energy $E$, conserved angular momentum $L$, mass $\mu$ and
Carter constant $Q$. We assume that the associated quartic polynomial $R(r)$,
as defined by Eq.~(\ref{e:gek:def_R_Q}) (see also Eq.~(\ref{e:gek:R_r_powers}) below),
and the associated function $\Theta(\th)$, as defined by Eq.~(\ref{e:gek:Theta_Q}),
do not vanish along $\Li$ except possibly at some $r$-turning points or
$\th$-turning points.
Let $\lambda$ be the affine parameter of $\Li$
associated with the 4-momentum $\w{p}$ and $\lambda'$ the
Mino parameter. If for $\lambda=\lambda_0$
$\Li$ lies at the point of Boyer-Lindquist coordinates $(t_0,r_0,\th_0,\ph_0)$,
then at any value of $\lambda$, the Boyer-Lindquist coordinates $(t,r,\th,\ph)$ along $\Li$
obey
\begin{subequations}
\label{e:gek:integrated_eqs}
\begin{align}
& \lambda' - \lambda'_0 = \dashint_{r_0}^r \frac{\eps_r \, \D \bar{r}}{\sqrt{R(\bar{r})}}
= \dashint_{\th_0}^\th \frac{\eps_\th \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
\label{e:gek:integr_Mino}\\
& \lambda - \lambda_0 = \dashint_{r_0}^r \frac{\eps_r \bar{r}^2 \D \bar{r}}{\sqrt{R(\bar{r})}}
+ a^2 \dashint_{\th_0}^\th \frac{\eps_\th \cos^2\bar{\th} \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
\label{e:gek:integr_lambda} \\
& t - t_0 = \dashint_{r_0}^r \frac{(\bar{r}^2 + a^2)^2 E - 2 a m \bar{r} L}{\bar{r}^2 - 2m\bar{r} + a^2} \frac{\eps_r\D \bar{r}}{\sqrt{R(\bar{r})}}
- a^2 E \dashint_{\th_0}^\th \sin^2\bar{\th} \frac{\eps_\th \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}} \label{e:gek:integr_t} \\
& \ph - \ph_0 = a \dashint_{r_0}^r \frac{2m\bar{r} E - a L}{\bar{r}^2 - 2m \bar{r} + a^2}
\frac{\eps_r \, \D \bar{r}}{\sqrt{R(\bar{r})}}
+ L \dashint_{\th_0}^\th \frac{1}{\sin^2\bar{\th}} \frac{\eps_\th \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}} . \label{e:gek:integr_ph}
\end{align}
\end{subequations}
\end{prop}
\begin{proof}
Equation~(\ref{e:gek:integr_Mino}) is nothing but the gathering of
Eqs.~(\ref{e:gek:Minotime_r_slash_int}) and (\ref{e:gek:Minotime_th_slash_int}).
For Eq.~(\ref{e:gek:integr_lambda}), it suffices to rewrite Eq.~(\ref{e:gek:def_Mino_time})
as
\[
\D\lambda = r^2 \, \D\lambda' + a^2\cos^2\th \, \D\lambda'
\]
and to substitute $\D\lambda'$ by $\eps_r \D r / \sqrt{R(r)}$
in the first term [cf. Eq.~(\ref{e:gek:drdl_Mino})] and by
$\eps_\th \D\th / \sqrt{\Theta(\th)}$ in the second term [cf. Eq.~(\ref{e:gek:dthdl_Mino})].
Similarly, by rewriting Eq.~(\ref{e:gek:dtdl_Mino}) as
\[
\D t = \frac{(r^2 + a^2)^2 E - 2 a m r L}{r^2 - 2mr + a^2} \, \D\lambda'
- a^2 E \sin^2\th \, \D\lambda'
\]
and performing the same substitutions for $\D\lambda'$ as above, we get
Eq.~(\ref{e:gek:integr_t}). Finally, Eq.~(\ref{e:gek:integr_ph}) is deduced
in the same fashion from Eq.~(\ref{e:gek:dphdl_Mino}).
\end{proof}
The system (\ref{e:gek:integrated_eqs}) shows that the geodesic motion can be
fully solved in terms of the Mino parameter $\lambda'$. Indeed,
the integral on $r$ in Eq.~(\ref{e:gek:integr_Mino}) can be evaluated by
means of elliptic integrals since $R$ is a polynomial of degree 4. This provides
$\lambda' = \lambda'(r)$. Inverting this relation via Jacobi elliptic functions
(cf. Remark~\ref{r:gis:Jacobi_elliptic_sine} on p.~\pageref{r:gis:Jacobi_elliptic_sine})
yields $r = r(\lambda')$. The integral on $\th$ in Eq.~(\ref{e:gek:integr_Mino})
can be evaluated by means of elliptic integrals as well since the change of variable
$\zeta := \cos\th$ along with expression (\ref{e:gek:Theta_Q}) for $\Theta(\th)$ results in
\[
\dashint_{\th_0}^\th \frac{\eps_\th \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
= - \dashint_{\zeta_0}^\zeta
\frac{\eps_\th \, \D \bar{\zeta}}{\sqrt{Q(1-\bar{\zeta}^2) + \bar{\zeta}^2[a^2(E^2-\mu^2)(1-\bar{\zeta}^2) - L^2]}} ,
\]
with the term under the square root in the right-hand side being a polynomial of degree 4
in $\bar{\zeta}$.
The use of Jacobi elliptic functions leads then to $\th = \th(\lambda')$.
From $r(\lambda')$ and $\th(\lambda')$ one can get the functions $\lambda(\lambda')$,
$t(\lambda')$ and $\ph(\lambda')$
by evaluating the integrals in the right-hand sides of
Eqs.~(\ref{e:gek:integr_lambda})--(\ref{e:gek:integr_ph}). Again, these integrals are
reducible to elliptic integrals. We shall not give the details of all the
elliptic integrals computations, referring the reader to Ref.~\cite{FujitH09}
for bound timelike geodesics and to Ref.~\cite{GrallL20b} for null geodesics.