-
Notifications
You must be signed in to change notification settings - Fork 18
/
gik.tex
3395 lines (3152 loc) · 181 KB
/
gik.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{Null geodesics and images in Kerr spacetime}
\label{s:gik}
\minitoc
\section{Introduction}
Having investigated the properties of generic causal geodesics
in Kerr spacetime in Chap.~\ref{s:gek}, we focus here on null
geodesics, with application to images of a Kerr black hole.
First we discuss the main properties of null geodesics in Sec.~\ref{s:gik:properties},
in great part by taking the $\mu=0$ limit of results obtained for generic causal
geodesics in Chap.~\ref{s:gek}. Then, in Sec.~\ref{s:gik:spherical_orbits}, we
focus on null geodesics evolving at a fixed value of the coordinate $r$ --- the so-called
\emph{spherical photon orbits}. These geodesics play a crucial role in the
formation of the images perceived by an observer.
In particular, they are related to the key concepts of \emph{critical curve} and \emph{shadow}
in the observer's screen, which are investigated in Sec.~\ref{s:gik:shadow}.
Finally, we discuss the images themselves in Sec.~\ref{s:gik:images}, first by
considering computed images from a simplified model of accretion disk
and then by analyzing the actual image of the surroundings of the black
hole M87*, as obtained recently by the Event Horizon Telescope \cite{EHT19a}.
\section{Main properties of null geodesics} \label{s:gik:properties}
We shall distinguish the null geodesics with $E=0$ (the so-called \emph{zero-energy} geodesics,
cf. Sec.~\ref{s:gek:sign_E})
from those having
$E \neq 0$. Indeed, in the latter case, we will rescale the angular momentum
$L$ and the Carter constant $Q$ by $E$, so that only two constants of motion become
pertinent for the study: $L/E$ and $Q/E^2$.
We thus treat first the particular case $E=0$.
\subsection{Zero-energy null geodesics} \label{s:gik:zero_energy}
First, we note that a geodesic $\Li$ with $E=0$ cannot exist outside the ergoregion
$\mathscr{G}$, by virtue of the result (\ref{e:gek:E_positive}). In particular,
it cannot exist far from the black hole.
Another property of $\Li$ is to have a non-negative Carter constant:
\be \label{e:gik:Q_nonneg_E_zero}
\encadre{ Q \geq 0 }_{E=0} .
\ee
This follows immediately Property~\ref{p:gek:latitudinal_motion} in Sec.~\ref{s:gek:th_motion},
which, among other things,
states that a necessary condition for $Q < 0$ is $a\neq 0$ and
$|E| > \sqrt{\mu^2 + L^2/a^2}$. Specializing this last inequality to $\mu=0$
and $E=0$, we get $0 > |L|$, which is impossible.
Besides, if $\Li$ has some part in $\M_{\rm I}$ (necessarily in the outer ergoregion)
or in $\M_{\rm III}$ (necessarily in the inner ergoregion),
the constraint (\ref{e:gek:future_directed_Carter})
reduces to $L<0$:
\be \label{e:gik:E_zero_L_neg_ergo}
\Li \cap (\M_{\rm I} \cup \M_{\rm III}) \neq \varnothing \quad \Longrightarrow \quad L < 0 .
\ee
We shall see below that actually $L \leq 0$ for all zero-energy null geodesics, as soon as $a\neq 0$.
The equations of geodesic motion expressed in terms of the Mino parameter $\lambda'$
[system (\ref{e:gek:eom_Mino})] simplify considerably for a geodesic $\Li$ with $\mu=0$ and $E=0$:
\begin{subequations}
\begin{align}
& \derd{t}{\lambda'} = - \frac{2 a m L r}{\Delta} \label{e:gik:eom_t_E_zero} \\
& \derd{r}{\lambda'} = \eps_r \sqrt{ R(r) } \label{e:gik:eom_r_E_zero}\\
& \derd{\th}{\lambda'} = \eps_\th \sqrt{\Theta(\th)} \\
& \derd{\ph}{\lambda'} = \frac{L}{\Delta\sin^2\th} \left( r^2 - 2 m r + a^2\cos^2\th \right) ,
\label{e:gik:eom_ph_E_zero}
\end{align}
\end{subequations}
with [cf. Eqs.~(\ref{e:gek:R_r_powers}) and (\ref{e:gek:Theta_Q})]:
\be \label{e:gik:R_E_zero}
R(r) = - (Q+L^2) r^2 + 2m (Q+L^2) r - a^2 Q
\ee
\be \label{e:gik:Theta_E_zero}
\Theta(\th) = Q - \frac{L^2}{\tan^2\th} .
\ee
By combining (\ref{e:gik:eom_t_E_zero}) and (\ref{e:gik:eom_ph_E_zero}), we get
\be
\encadre{ \left. \derd{\ph}{t} \right| _{\Li} = \frac{2mr - r^2 - a^2\cos^2\th}{2 a m r\sin^2\th} }_{E=0} .
\ee
It is remarkable that this expression does not depend on $L$ or $Q$; it is therefore the
same for all zero-energy null geodesics. Moreover, we note that the numerator
of the right-hand side is always positive or zero in the closure $\overline{\mathscr{G}}$ of the ergoregion,
which is precisely defined by $2mr - r^2 - a^2\cos^2\th \geq 0$ (cf. Sec.~\ref{s:ker:ergoregion})
and where $\Li$ is necessarily confined. Since moreover $r > 0$ in $\overline{\mathscr{G}}$,
we conclude that
\be
\left. \derd{\ph}{t} \right| _{\Li} \geq 0 .
\ee
To proceed, we shall distinguish the subcases $Q\neq 0$ and $Q=0$.
\subsubsection{Case $Q\neq 0$}
This case actually corresponds to $Q > 0$, since $Q<0$ is forbidden by
(\ref{e:gik:Q_nonneg_E_zero}). We set
\be \label{e:gik:def_L_bar}
\bar{L} := \frac{L}{\sqrt{Q}}
\ee
and rewrite expression (\ref{e:gik:R_E_zero}) for $R(r)$ as
\be \label{e:gik:R_E_zero_Lb}
R(r)/Q = - (1 + \bar{L}^2) r^2 + 2m (1 + \bar{L}^2) r - a^2 .
\ee
Since $1 + \bar{L}^2 \neq 0$, this is a second-order polynomial in $r$,
the two roots of which are
\be \label{e:gik:E_zero_rmin_rmax}
r_{\rm min} = m - \sqrt{m^2 - \frac{a^2}{1 + \bar{L}^2}}
\qand
r_{\rm max} = m + \sqrt{m^2 - \frac{a^2}{1 + \bar{L}^2}} .
\ee
Since $m^2 \geq a^2$, the two roots are real. They are distinct
except for $a=m$ and $L=0$.
The range of radial motion being determined by $R(r)\geq 0$
[Eq.~(\ref{e:gek:R_non_neg})], we get
\be
r_{\rm min} \leq r \leq r_{\rm max} ,
\ee
with a turning point at $r_{\rm min}$ and at $r_{\rm max}$.
Given that $r_- = m - \sqrt{m^2 - a^2}$ and $r_+ = m + \sqrt{m^2 - a^2}$
[Eq.~(\ref{e:ker:def_r_pm})], we note that
\be
0 \leq r_{\rm min} \leq r_- \leq m \leq r_+ \leq r_{\rm max} \leq 2 m ,
\ee
with $r_{\rm min} = 0$ for $a = 0$ or $\bar{L}^2 \to +\infty$,
$r_{\rm min} = r_-$ for $L=0$, $r_{\rm max} = 2m$ for $a=0$
or $\bar{L}^2 \to +\infty$ and $r_{\rm max} = r_+$ for $L=0$.
If $L\neq 0$ and $a\neq 0$, then $r_{\rm max} > r_+$, so that $\Li$ has a part in the outer
ergoregion and (\ref{e:gik:E_zero_L_neg_ergo}) implies that $L < 0$. Hence
\be \label{e:gik:E_zero_Qnz_L_neg}
a\neq 0 \implies L \leq 0 .
\ee
Let us consider a zero-energy null geodesic $\Li$ emitted outward (i.e.
with $\epsilon_r = +1$) from a point $A$ in the outer ergoregion $\mathscr{G}^+$.
The coordinate $r$ increases along $\Li$ from $r_A$ to $r_{\rm max}$, which
corresponds to a $r$-turning point. Then $r$ decreases to $r_+$, which means
that $\Li$
crosses the black hole event horizon $\Hor$ and enters the region $\M_{\rm II}$.
In all $\M_{\rm II}$, $r$ keeps decreasing and reaches $r_-$. There $\Li$
crosses the inner horizon $\Hor_{\rm in}$ and enters the region $\M_{\rm III}$,
where $r$ continues to decrease until it reaches $r_{\rm min}$. The latter
corresponding to a $r$-turning point, $r$ starts to increase and reaches
$r_-$ again. There one might think that $\Li$ crosses the inner horizon
$\Hor_{\rm in}$ and enters into $\M_{\rm II}$. But this is impossible since
$\Hor_{\rm in}$ is a 1-way membrane: it can be crossed by a causal curve
from $\M_{\rm II}$ to $\M_{\rm III}$ but not in the reverse way. Moreover, $r$
could not continue to increase into $\M_{\rm II}$ since $r$ must be decreasing
towards the future in all this region (this follows from the hypersurfaces
$r=\mathrm{const}$ being spacelike in $\M_{\rm II}$, cf. Sec.~\ref{s:ker:Killing_hor}).
The solution to this apparent puzzle is immediate as soon as one realizes
that the boundary $r = r_-$ of $\M_{\rm III}$ is not entirely constituted
by $\Hor_{\rm in}$: it also comprises a null hypersurface that separates
$\M_{\rm III}$ from a region distinct from $\M_{\rm II}$ in
the maximally extended Kerr spacetime, cf. Fig.~\ref{f:ker:max_ext}. This region
is a ``time-reversed'' copy of $\M_{\rm II}$ and is
denoted by ${\M_{\rm II}^*}'$ in Fig.~\ref{f:ker:max_ext} (cf. Sec.~\ref{s:ker:max_extension}
for details).
So actually, when it reaches $r=r_-$, the null geodesic $\Li$ enters
${\M_{\rm II}^*}'$. There $r$ necessarily increases towards the future, at
the opposite of $\M_{\rm II}$. It reaches then $r=r_+$, where $\Li$
crosses a white hole\index{white hole} horizon and emerges into
the asymptotically flat region $\M_{\rm I}''$, as illustrated in
Fig.~\ref{f:gik:zero_ener_traj}. The region $\M_{\rm I}''$ is similar
to $\M_{\rm I}$. In particular, $\Li$ is confined into the outer ergoregion
of $\M_{\rm I}''$, having a $r$-turning point at $r=r_{\rm max}$ (same value
(\ref{e:gik:E_zero_rmin_rmax}) as in $\M_{\rm I}$). Then a new cycle
begins, with $\Li$ entering the future event horizon of $\M_{\rm I}''$.
\begin{figure}
\centerline{\includegraphics[width=0.6\textwidth]{gik_zero_ener_traj.pdf}}
\caption[]{\label{f:gik:zero_ener_traj} \footnotesize
Trajectory in the extended Kerr spacetime of a null geodesic
with $E=0$, $Q>0$ and $L<0$, emitted from a point $A$ in the outer ergoregion.
}
\end{figure}
The $\theta$-motion of $\Li$ is constrained by $\Theta(\th) \geq 0$ [Eq.~(\ref{e:gek:Theta_non_neg})],
which, given expression~(\ref{e:gik:Theta_E_zero}) for $\Theta$, is
equivalent to
\be
\th_{\rm m} \leq \th \leq \pi - \th_{\rm m} \quad\mbox{with}\quad
\th_{\rm m} := \arctan (-\bar{L}) .
\ee
\begin{remark}
The general formula for $\th_{\rm m}$ in the case
$a^2(E^2 - \mu^2) = 0$, Eq.~(\ref{e:gek:th0_a2E2mu2_zero}), which holds here since $\mu=0$ and $E=0$,
yields $\th_{\rm m} = \arccos\sqrt{1/(1+ \bar{L}^2)} = \arctan | \bar{L} |$.
Hence we recover the above formula.
\end{remark}
For $L = 0$, one has $\th_{\rm m} = 0$, so that $\th$ takes all values in the
range $[0,\pi]$, which means that $\Li$
crosses repeatedly the rotation axis. For $L< 0$, one has $0 < \th_{\rm m} < \pi/2$ and
$\Li$ oscillates symmetrically about the equatorial plane,
having two $\th$-turning points, at $\th_{\rm m} $ and $\pi-\th_{\rm m}$. Of course, we
recover the general results for $Q>0$ of Sec.~\ref{s:gek:th_motion}.
We can obtain $r$ as a function of $\th$ along $\Li$ by evaluating the integrals
in the identity (\ref{e:gek:integr_Mino}):
\[
\dashint_{r_0}^r \frac{\eps_r \, \D \bar{r}}{\sqrt{R(\bar{r})}}
= \dashint_{\th_0}^\th \frac{\eps_\th \, \D \bar{\th}}{\sqrt{\Theta(\bar{\th})}}
\]
Using (\ref{e:gik:R_E_zero_Lb}) and (\ref{e:gik:Theta_E_zero}), we get on
any portion of $\Li$ where $\eps_r$ and $\eps_\th$ are constant,
\[
\eps_r \int_{r_0}^r
\frac{\D \bar{r}}{\sqrt{- (1 + \bar{L}^2) \bar{r}^2 + 2m (1 + \bar{L}^2) \bar{r} - a^2}}
= \eps_\th \int_{\th_0}^\th \frac{\D \bar{\th}}{\sqrt{1 - \bar{L}^2/\tan^2\bar{\th}}} .
\]
The changes of variables
\[
x = \frac{r/m - 1}{\sqrt{1 - \frac{a^2}{m^2(1 + \bar{L}^2)}}}
\qand
\mu = \cos\th
\]
lead to
\[
\frac{\eps_r}{\sqrt{1 + \bar{L}^2}} \int_{x_0}^x \frac{\D\bar{x}}{\sqrt{1 - \bar{x}^2}}
= - \eps_\th \int_{\cos\th_0}^{\cos\th} \frac{\D\mu}{\sqrt{1 - (1 + \bar{L}^2)\mu^2}} .
\]
The integration is then immediate:
$\arcsin x = - \eps_r \eps_\th \arcsin(\sqrt{1 + \bar{L}^2} \cos\th) + K$,
where $K$ is a constant, from which we get
\be \label{e:gik:E_zero_r_theta}
\encadre{ r = m + m\sqrt{1 - \frac{a^2}{m^2(1 + \bar{L}^2)}} \; \sin \left[
K - \eps_r \eps_\th \arcsin\left(\sqrt{1 + \bar{L}^2} \cos\th\right) \right] }.
\ee
Since $\sqrt{1 + \bar{L}^2} \cos\th_{\rm m} = 1$, we see that the constant
$K$ is related to the value of $r$ at $\th = \th_{\rm m}$ by
\be \label{e:gik:E_zero_K}
K = \arcsin \left( \frac{r(\th_{\rm m})/m - 1}{\sqrt{1 - \frac{a^2}{m^2(1 + \bar{L}^2)}}} \right)
+ \eps_r \eps_\th \frac{\pi}{2} .
\ee
Note that $K$ is not a constant all along $\Li$, but only on portions where
$\eps_r$ and $\eps_\th$ are constant.
Expression~\ref{e:gik:E_zero_r_theta} gives the trace of the zero-energy
null geodesic $\Li$ in a meridional plane. It depends on $Q$ and $L$ only
through the ratio $\bar{L} := L / \sqrt{Q}$. It depends as well on the value of
$r$ at $\th_{\rm m}$ via $K$, as it appears in Eq.~(\ref{e:gik:E_zero_K}).
An example is shown in Fig.~\ref{f:gik:zero_ener_merid} for $a/m = 0.9$, $L / \sqrt{Q} = -1$
and $r(\th_{\rm m}) = 1.5\, m$. It has $\th_{\rm m} = \pi/4$,
$r_{\rm min} \simeq 0.229\, m$ and $r_{\rm max} \simeq 1.771\, m$,
while for $a/m = 0.9$, one has $r_- \simeq 0.564\, m$ and $r_+ \simeq 1.436\, m$.
For concreteness, the arrows indicate some direction of motion, but depending upon some
initial conditions, the opposite direction is possible.
In particular, one may consider that the geodesic is the same as that shown
in Fig.~\ref{f:gik:zero_ener_traj}, being emitted outward in the outer ergoregion
from a point $A$ in the equatorial plane ($\th=\pi/2$).
\begin{figure}
\centerline{\includegraphics[width=0.6\textwidth]{gik_zero_ener_merid.pdf}}
\caption[]{\label{f:gik:zero_ener_merid} \footnotesize
Trajectory in a meridional plane, as given by Eq.~(\ref{e:gik:E_zero_r_theta}), of a null geodesic (green curve)
with $E=0$, $Q>0$, $L = - \sqrt{Q}$ and $r(\th_{\rm m}) = 1.5 \, m$
in the Kerr spacetime with $a/m = 0.9$.
The meridional plane is described in terms of
O'Neill exponential coordinates\index{O'Neill!coordinates} $x = \mathrm{e}^{r/m}\sin\th$ and $z = \mathrm{e}^{r/m}\cos\th$,
as in Figs.~\ref{f:ker:ergo_a90} -- \ref{f:ker:ergo_a99}.
The ergoregion is shown in grey.
The black (resp. light brown) half-circle at $r=r_+$ (resp. $r=r_-$)
is the trace of the outer (resp. inner) Killing horizon.
The dotted orange half-circle marks the locus of $r=0$, with the
red dot indicating the curvature singularity at $r=0$ and $\th=\pi/2$.
The area $r> r_+$
corresponds to the regions $\M_{\rm I}$ and $\M_{\rm I}''$ in Fig.~\ref{f:gik:zero_ener_traj},
the area $r_- < r < r_+$
corresponds to the regions $\M_{\rm II}$ and ${\M_{\rm II}^*}'$ in Fig.~\ref{f:gik:zero_ener_traj} and the area $r < r_-$
corresponds to the region $\M_{\rm III}$ in Fig.~\ref{f:gik:zero_ener_traj}.
\textsl{[Figure generated by the notebook \ref{s:sam:Kerr_null_geod_zero_ener}]}
}
\end{figure}
\subsubsection{Case $Q=0$}
If the zero-energy null geodesic $\Li$ has a vanishing Carter constant $Q$, Eq.~(\ref{e:gik:Theta_E_zero}) reduces to $\Theta(\th) = -L^2 / \tan^2\th$,
so that the constraint $\Theta(\th) \geq 0$ [Eq.~(\ref{e:gek:Theta_non_neg})]
implies $L=0$ or $\th = \pi/2$ .
In the first case, the four constants of motion $\mu$, $E$, $L$ and $Q$ are
zero. By virtue of the result (\ref{e:gek:all_const_zero}), $\Li$
is nothing but a null geodesic generator of the event horizon $\Hor$ or of the
inner horizon $\Hor_{\rm in}$.
In the second case ($\th=\pi/2$), $\Li$ is confined to the equatorial plane.
If $L=0$, we are back to the first case: $\Li$ is null geodesic generator of $\Hor$ or
$\Hor_{\rm in}$ lying in the equatorial plane.
If $L\neq 0$, the radial motion of $\Li$ is governed by Eq.~(\ref{e:gik:eom_r_E_zero}) with the
expression (\ref{e:gik:R_E_zero}) of $R(r)$ reduced to
\be
R(r) = L^2 r (2m - r) .
\ee
The constraint $R(r) \geq 0$ [Eq.~(\ref{e:gek:R_non_neg})] implies then
that the motion is within the range $0 \leq r \leq 2 m$, with $r=2m$
being a $r$-turning point, since it is a simple root of $R(r)$ (cf. Sec.~\ref{s:gek:turning_points}).
It corresponds to the outer edge of the ergoregion in the equatorial plane,
cf. Eq.~(\ref{e:ker:r_ergo_p_eq}). Hence we have necessarily $\Li\cap \M_{\rm I} \neq \varnothing$
and (\ref{e:gik:E_zero_L_neg_ergo}) applies: $L < 0$.
The inner boundary of the radial motion, $r=0$, is the ring singularity.
Accordingly, in the maximally extended Kerr spacetime, $\Li$ starts at the ring
singularity in a $\M_{\rm III}$-type region (cf. Fig.~\ref{f:gik:zero_ener_traj_q0}),
has $r$ increasing, enters
a $\M_{\rm II}^*$-type region (time reversed copy of $\M_{\rm II}$), emerges
in $\M_{\rm I}$ via the white hole horizon at $r=r_+$ and reaches a $r$-turning point at
$r=2m$, then $r$ decreases continuously until $\Li$ terminates at the ring
singularity of $\M_{\rm III}$, after having crossed the black hole horizon $\Hor$
and the inner horizon $\Hor_{\rm in}$.
This trajectory, depicted in
Fig.~\ref{f:gik:zero_ener_traj_q0}, is similar to that for $Q\neq 0$
shown in Fig.~\ref{f:gik:zero_ener_traj}, except that it is ``blocked''
by two ring singularities and cannot oscillate forever between distinct
$\M_{\rm I}$-type regions.
\begin{figure}
\centerline{\includegraphics[width=0.6\textwidth]{gik_zero_ener_traj_q0.pdf}}
\caption[]{\label{f:gik:zero_ener_traj_q0} \footnotesize
Trajectory in the extended Kerr spacetime of a null geodesic
with $E=0$, $Q=0$ and $L<0$.
}
\end{figure}
\begin{remark}
For $Q\neq 0$ and $L\neq 0$, the limit $Q\to 0$ corresponds to
$\bar{L}^2 \to +\infty$ [cf. Eq.~(\ref{e:gik:def_L_bar})], so that
Eq.~(\ref{e:gik:E_zero_rmin_rmax}) yields
$r_{\rm min}\to 0$ and $r_{\rm max}\to 2m$.
We recover then the range $[0, 2m]$ for $r$ obtained here for $Q=0$.
\end{remark}
We conclude:
\begin{prop}[null geodesic with $E=0$ and $Q=0$]
Any null geodesic with $E=0$ and $Q=0$
is either a null generator of one of the two Killing horizons
$\Hor$ or $\Hor_{\rm in}$ (in which case, it has $L=0$) or it has
$L < 0$, lies in the equatorial plane,
emanates from a ring singularity, reaches the outer ergosphere ($r=2m$),
where it has a $r$-turning point, and terminates at a ring singularity.
\end{prop}
Regarding the sign of $L$, we can combine the above result with that
obtained for $Q\neq 0$ [Eq.~(\ref{e:gik:E_zero_Qnz_L_neg})] to get:
\begin{prop}[negative angular momentum for zero-energy null geodesics]
For $a\neq 0$, any null geodesic with $E=0$ has necessarily
\be
\encadre{ L \leq 0 }_{{a\neq 0\atop E=0}}.
\ee
\end{prop}
\begin{hist}
The zero-energy null geodesics in Kerr spacetime appear to have been first
studied by Zden\v{e}k Stuchl\'{\i}k\index[pers]{Stuchl\'{\i}k, Z.} in 1981, in the appendix
of the article~\cite{Stuch81}; some corrections and refinement of his results have been
performed by George Contopoulos\index[pers]{Contopoulos, G.} in 1984 \cite{Conto84},
who studied zero-energy timelike geodesics as well.
\end{hist}
\subsection{Equations of geodesic motion for $E\neq 0$} \label{s:gik:eom_Enonzero}
For any null geodesic $\Li$ with $E\neq 0$, we introduce the reduced constants of motion
\be \label{e:gik:def_ell_q}
\encadre{\ell := \frac{L}{E}} \qand
\encadre{q := \frac{Q}{E^2}} .
\ee
Note that, in geometrized units ($G=1$ and $c=1$), $\ell$ has the dimension of
a length and $q$ that of a squared length.
\begin{remark}
In the literature, $\ell$ is sometimes denoted by $\lambda$ (e.g. Refs.~\cite{Barde73a,GrallL20b,DokucN20a})
or by $\xi$ (e.g. Ref.~\cite{Chand83}), while $q$ is sometimes denoted by
$\eta$ (e.g. Refs.~\cite{Barde73a,Chand83,GrallL20b}). Moreover, in studies restricted
to $q\geq 0$, it may happen that the notation $q$ stands for the square root of the quantity $q$ defined by
(\ref{e:gik:def_ell_q}) (e.g. Refs.~\cite{DexteA09,GrallLS18,DokucN20a}).
\end{remark}
\begin{remark} \label{r:gik:ell_q_intrinsic}
Contrary to $L$ and $Q$,
the quantities $\ell$ and $q$ are independent from the affine parametrization of the geodesic $\Li$.
Indeed, if instead of the affine parameter $\lambda$ associated with the particle's 4-momentum $\w{p}$,
one considers the affine parameter $\lambda' = \alpha \lambda$, where $\alpha$ is a positive constant,
the tangent vector field becomes
$\w{p}' = \alpha^{-1} \w{p}$, so that the associated conserved
quantities are $E' = - \w{\xi}\cdot\w{p}' = \alpha^{-1} E$ [cf. Eq.~(\ref{e:gek:def_E})], $L' = \w{\eta}\cdot\w{p}' = \alpha^{-1} L$ [cf. Eq.~(\ref{e:gek:def_L})]
and $Q' = \w{\tilde K}(\w{p}',\w{p}') = \alpha^{-2} Q$ [cf. Eq. (\ref{e:gek:Q_tK_pp})],
so that $\ell' := L'/E' = \ell$ and $q' := Q'/{E'}^2 = q$.
\end{remark}
The non-negative property of Carter constant $\mathscr{K}$ [Eq.~(\ref{e:ges:K_non_negative})],
along with the identity (\ref{e:gek:def_Q}) leads to the following constraint on
the parameters $\ell$ and $q$:
\be \label{e:gik:q_ell_a_constraint}
\encadre{ q + (\ell - a)^2 \geq 0} .
\ee
\begin{example}[Principal null geodesic]
A principal null geodesic has $E\neq 0$ if, and only if, it does not belong to the outgoing
family generating the horizon $\Hor$ or $\Hor_{\rm in}$. This follows immediately from
Eqs.~(\ref{e:gek:ingoing_null_E_L}) and (\ref{e:gek:outgoing_null_E_L}).
One has then
$L = a E\sin^2\th_0$
[Eqs.~(\ref{e:gek:ingoing_null_E_L}) and (\ref{e:gek:outgoing_null_E_L})]
and $Q = - a^2 E^2 \cos^2\th_0$ [Eq.~(\ref{e:gek:Q_principal_null})], where
$\th_0$ is the constant value of $\th$ along the geodesic. We have thus
\be \label{e:gik:principal_null_l_q}
\ell = a \sin^2\th_0 \qand q = - a^2 \cos^4\th_0 .
\ee
This yields
\be
q + (\ell - a)^2 = 0 ,
\ee
so that the inequality (\ref{e:gik:q_ell_a_constraint}) is saturated for principal null
geodesics.
\end{example}
The equations of motion in terms of the Mino parameter $\lambda'$,
Eqs.~(\ref{e:gek:eom_Mino}) specialized to $\mu=0$, can be rewritten as
\begin{subequations}
\label{e:gik:eom_Mino}
\begin{align}
& \encadre{ \frac{1}{E} \derd{t}{\lambda'} = \frac{1}{\Delta} \left[ (r^2 + a^2)^2 - 2 a m \ell r \right] - a^2 \sin^2\th } \label{e:gik:dtdl_Mino} \\
& \encadre{ \frac{1}{|E|} \derd{r}{\lambda'} = \eps_r \sqrt{ \mathcal{R}(r) } } \label{e:gik:drdl_Mino}\\
& \encadre{ \frac{1}{|E|} \derd{\th}{\lambda'} = \eps_\th \sqrt{\tilde\Theta(\th)} } \label{e:gik:dthdl_Mino}\\
& \encadre{ \frac{1}{E} \derd{\ph}{\lambda'} = \frac{\ell}{\sin^2\th}
+ \frac{a}{\Delta}(2m r - a \ell) } , \label{e:gik:dphdl_Mino}
\end{align}
\end{subequations}
with
\be
\mathcal{R}(r) := \frac{R(r)}{E^2}
\qand
\tilde\Theta(\th) := \frac{\Theta(\th)}{E^2} .
\ee
From the general expressions (\ref{e:gek:def_R_Q}), (\ref{e:gek:R_r_powers}) and (\ref{e:gek:Theta_Q}) specialized to $\mu=0$, we get
\begin{subequations}
\label{e:gik:mcR}
\begin{align}
&\encadre{\mathcal{R}(r) = (r^2 + a^2 - a \ell)^2 - \Delta \left[ q + (\ell -a)^2 \right] }
\label{e:gik:mcR_Delta}\\
&\encadre{\mathcal{R}(r) = r^4 + (a^2 - \ell^2 - q) r^2 + 2m\left[ q + (\ell -a)^2 \right] r
- a^2 q } \label{e:gik:mcR_powers}
\end{align}
\end{subequations}
and
\be \label{e:gik:tTheta}
\encadre{ \tilde\Theta(\th) = q + \cos^2\th \left( a^2
- \frac{\ell^2}{\sin^2\th} \right) } .
\ee
It suffices to use the parameter $\lambda'' := |E| \lambda'$ to make $E$
disappear from the system~(\ref{e:gik:eom_Mino}). We therefore conclude:
\begin{prop}
In Kerr spacetime, a null geodesic with $E\neq 0$ is entirely determined
by the two constants $(\ell,q)$, by the sign of $E$ and by the values of the two
signs $\eps_r=\pm 1$ and $\eps_\th=\pm 1$ at a given point.
\end{prop}
\begin{example}[Principal null geodesic]
Given the values (\ref{e:gik:principal_null_l_q}) of $\ell$ and $q$
for a principal null geodesic, Eq.~(\ref{e:gik:mcR_powers}) reduces to
a simple expression for the quartic polynomial $\mathcal{R}$:
\be \label{e:gik:mR_PNG}
\mathcal{R}(r) = \left( r^2 + a^2 \cos^2 \th_0 \right)^2 .
\ee
We note that $\mathcal{R}(r) = \rho^4$, which makes sense because $\th = \th_0$
is constant along such a geodesic. For any principal null geodesic lying
in the equatorial plane, the polynomial simplifies even further:
\be
\mathcal{R}(r) = r^4 \qquad \left(\th_0 = \frac{\pi}{2} \right) .
\ee
\end{example}
\subsection{Position on a remote observer's screen} \label{s:gik:remote_screen}
The constants $(\ell,q)$ are closely related to the impact coordinates $(\alpha,\beta)$
on the screen of an asymptotic inertial observer (cf. Sec.~\ref{s:ker:asymp_inertial_obs})
in case the null geodesic $\Li$ reaches the asymptotic region $r\to +\infty$.
Indeed, let us consider an asymptotic inertial observer $\Obs$ located at (fixed) Boyer-Lindquist
coordinates $(r_{\Obs}, \th_{\Obs}, \ph_{\Obs})$, with $r_{\Obs} \gg m$ (cf. Fig.~\ref{f:gik:obs_screen}).
In order to reach $\Obs$, the geodesic $\Li$ must be such that the constraints
constraints $\mathcal{R}(r_{\Obs}) \geq 0$
[Eq.~(\ref{e:mcR_non_neg})]
and $\tilde\Theta(\th_{\Obs}) \geq 0$ [Eq.~(\ref{e:gek:Theta_non_neg})]
are fulfilled.
The first one is always satisfied due to the
assumption that $\Obs$ is an asymptotic observer, since
$\mathcal{R}(r) \sim r^4$ for $r\to +\infty$ [cf. Eq.~(\ref{e:gik:mcR_powers})].
In view of expression~(\ref{e:gik:tTheta}) for $\tilde\Theta$, the second
constraint is equivalent to
\be \label{e:gik:constraint_theta_obs}
\left( q + a^2 \cos^2\th_{\Obs} \right) \sin^2\th_{\Obs}
- \ell^2 \cos^2\th_{\Obs} \geq 0 .
\ee
If $\sin\th_{\Obs}$ is small,
this constraint limits significantly the amplitude of $\ell$.
\begin{figure}
\centerline{\includegraphics[width=0.6\textwidth]{gik_obs_screen.pdf}}
\caption[]{\label{f:gik:obs_screen} \footnotesize
Impact of a null geodesic $\Li$ onto the screen of a remote observer. $(x,y,z)$ are the Cartesian Boyer-Lindquist coordinates
defined by Eq.~(\ref{e:gek:Cartesian_BL}).
}
\end{figure}
\subsubsection{Observer not located on the rotation axis}
Here we treat the generic case of an observer $\Obs$ that is not located on the rotation axis,
i.e. we assume $\th_{\Obs}\not\in\{0,\pi\}$.
The orthonormal frame of $\Obs$ is $\w{e}_{(0)} =\w{\xi}$, $\w{e}_{(r)} = \wpar_r$,
$\w{e}_{(\th)} = r_{\Obs}^{-1} \wpar_\th$ and $\w{e}_{(\ph)} = (r_{\Obs}\sin\th_{\Obs})^{-1} \wpar_\ph$
[Eq.~(\ref{e:ker:def_ZAMO_frame}) with $r = r_{\Obs}\to +\infty$].
Let us assume that the observer set up a screen (via a telescope) centered on the direction to the
black hole, i.e. such that $\w{e}_{(r)}$ is the normal to the screen. The 4-momentum of the photon
having $\Li$ as worldline at the location of $\Obs$ writes
\[
\w{p} = p^t \, \w{\xi} + p^r \, \w{e}_{(r)} + p^\th r_{\Obs} \, \w{e}_{(\th)}
+ p^\ph r_{\Obs} \sin\th_{\Obs} \, \w{e}_{(\ph)} = E (\w{\xi} + \w{n}) ,
\]
where the second equality follows from the orthogonal decomposition (\ref{e:fra:p_E_V}) with respect to
$\Obs$, $\w{\xi}$ being the 4-velocity of $\Obs$ and the unit spacelike vector $\w{n}$ being the photon's velocity\footnote{$\w{n}$ is denoted by $\w{V}$ in Eq.~(\ref{e:fra:p_E_V}).} relative to $\Obs$; it
obeys $\w{\xi}\cdot\w{n} = 0$ and $\w{n}\cdot\w{n} = 1$ [Eq.~(\ref{e:fra:photon_V_one})].
That the conserved energy $E$ appears in the above equation is a direct consequence of its
definition as $E := - \w{\xi}\cdot \w{p}$ [Eq.~(\ref{e:gek:def_E})] with $\w{\xi}\cdot\w{\xi} = -1$
for $r_{\Obs}\to +\infty$. The above identity implies
$p^t = E$ and
\[
\w{n} = \frac{p^r}{E} \, \w{e}_{(r)} + \frac{p^\th}{E} r_{\Obs} \, \w{e}_{(\th)}
+ \frac{p^\ph}{E} r_{\Obs} \sin\th_{\Obs} \, \w{e}_{(\ph)} .
\]
With respect to $\Obs$, the incoming direction of the photon is given by
the vector $-\w{n}$ and the trace on the screen is indicated by the component of
$-\w{n}$ that is tangent to the screen, namely
\be \label{e:gik:m_screen_p}
\w{m} = - \w{n}_{\parallel} = - \frac{p^\th}{E} r_{\Obs} \, \w{e}_{(\th)}
- \frac{p^\ph}{E} r_{\Obs} \sin\th_{\Obs} \, \w{e}_{(\ph)}
\ee
since the screen plane is spanned by $\w{e}_{(\th)}$ and $\w{e}_{(\ph)}$
(cf. Fig.~\ref{f:gik:obs_screen}). Let us choose the screen dimensionless
Cartesian coordinates $(\alpha,\beta)$ so that the black hole's rotation axis
appears as the $\beta$-axis (cf. Fig.~\ref{f:gik:obs_screen}), then
\be \label{e:gik:m_screen_ab}
\w{m} = \alpha \w{e}_{(\alpha)} + \beta \w{e}_{(\beta)}, \quad\mbox{with}\quad
\w{e}_{(\alpha)} = \w{e}_{(\ph)} \quad\mbox{and}\quad
\w{e}_{(\beta)} = - \w{e}_{(\th)} .
\ee
By comparing with (\ref{e:gik:m_screen_p}), we get
\[
\alpha = - \frac{p^\ph}{E} r_{\Obs} \sin\th_{\Obs}
\qand
\beta = \frac{p^\th}{E} r_{\Obs} .
\]
Now, for $r_{\Obs}\to +\infty$,
\[
\frac{p^\th}{E} = \frac{1}{E} \derd{\th}{\lambda} = \frac{1}{E r_{\Obs}^2} \derd{\th}{\lambda'} =
\frac{\eps_\th}{r_{\Obs}^2} \sqrt{\tilde\Theta(\th_{\Obs})}
\]
\[
\frac{p^\ph}{E} = \frac{1}{E} \derd{\ph}{\lambda} = \frac{1}{E r_{\Obs}^2} \derd{\ph}{\lambda'}
= \frac{\ell}{r_{\Obs}^2 \sin^2\th_{\Obs}} ,
\]
where we have used Eqs.~(\ref{e:gik:dthdl_Mino}) and (\ref{e:gik:dphdl_Mino}), with the
term involving $a/\Delta$ neglected since $\Delta \sim r^2$ for $r\to +\infty$.
By inserting these formula into the above expressions of $\alpha$ and $\beta$,
and using Eq.~(\ref{e:gik:tTheta}) for $\tilde\Theta(\th_{\Obs})$,
we get the sought relation between the constants of motion $(\ell, q)$ and
the screen coordinates:
\begin{subequations}
\label{e:gik:screen_alpha_beta}
\begin{align}
& \encadre{ \alpha = - \frac{\ell}{r_{\Obs}\sin\th_{\Obs}} } \\
& \encadre{ \beta = \frac{\eps_\th}{r_{\Obs}} \sqrt{ q + \cos^2\th_{\Obs} \left( a^2
- \frac{\ell^2}{\sin^2\th_{\Obs}} \right) } } .
\end{align}
\end{subequations}
We have defined $(\alpha,\beta)$ as dimensionless Cartesian coordinates on
the screen, cf. Eq.~(\ref{e:gik:m_screen_ab}), where $\w{m}$ is
dimensionless and $(\w{e}_{(\alpha)}, \w{e}_{(\beta)})$
is the screen's orthonormal basis. In practice, their values are tiny,
being exactly zero at the limit $r_{\Obs}\to +\infty$. $(\alpha,\beta)$ can thus be interpreted as
\emph{angular} coordinates, measuring the departure from the direction
of the black hole ``center'' on the celestial sphere of observer $\Obs$.
We shall then call $(\alpha,\beta)$ the
\defin{screen angular coordinates}\index{screen!angular coordinates}\index{angular!screen -- coordinates}.
\begin{remark}
When studying null geodesics in Schwarzschild spacetime in Chap.~\ref{s:gis}, we
introduced the impact parameter $b$ as $b := |L|/E$ [Eq.~(\ref{e:ges:def_b})], hence
$b$ is related to $\ell$ by
\be
b = |\ell| .
\ee
Moreover, thanks to spherical symmetry, we could reduce the study to the case
where both the observer
and the geodesic lie in the equatorial plane, which imply $\th_{\Obs} = \pi/2$
and $q=0$. Equations~(\ref{e:gik:screen_alpha_beta}) yield then
\be
|\alpha| = \frac{b}{r_{\Obs}} = \hat{b} \qand \beta = 0 ,
\ee
where $\hat{b}$ is the angle introduced by formula (\ref{e:gis:hat_b}).
\end{remark}
\begin{remark}
The angular impact parameters $(\alpha, \beta)$ depend on the geodesic $\Li$ as a curve
in spacetime and not on any affine parametrization of $\Li$. Their direct
connection with $(\ell, q)$ expressed by (\ref{e:gik:screen_alpha_beta}) is thus in
perfect agreement with the invariance of $(\ell, q)$ in any affine reparametrization
of $\Li$, as noticed in Remark~\ref{r:gik:ell_q_intrinsic} p.~\pageref{r:gik:ell_q_intrinsic}.
\end{remark}
We deduce from Eqs.~(\ref{e:gik:screen_alpha_beta}) a
simple relation between the squared angular distance to the screen center, $\alpha^2 + \beta^2$,
and the constants of motion $(\ell, q)$ of the incoming geodesic:
\be \label{e:gik:alpha2_beta2}
\alpha^2 + \beta^2 = \frac{1}{r_{\Obs}^2} \left(
\ell^2 + q + a^2 \cos^2\th_{\Obs} \right) .
\ee
%The asymptotic formula (\ref{e:gek:Q_Ltot2_L2}) relates $\ell^2 + q = E^{-2} (L^2 + Q)$
%to the square of the total angular momentum $\w{L}_{\rm tot}$ of the incoming
%photon as measured by observer $\Obs$. Using it, we can turn (\ref{e:gik:alpha2_beta2})
%into
%\be
%r_{\Obs}^2 (\alpha^2 + \beta^2) = \frac{\w{L}_{\rm tot}\cdot\w{L}_{\rm tot}}{E^2}
%+ 2 a \frac{L}{E} - a^2 \sin^2\th_{\Obs} .
%\ee
\subsubsection{Observer on the rotation axis}
If the asymptotic inertial observer $\Obs$ is located on the rotation axis,
i.e. if $\th_{\Obs}=0$ or $\th_{\Obs} = \pi$,
the only value of $\ell$ compatible with
the constraint (\ref{e:gik:constraint_theta_obs}) is
\be \label{e:gik:obs_axis_ell_zero}
\ell = 0 \qquad (\th_{\Obs} = 0\ \mbox{or}\ \th_{\Obs} = \pi).
\ee
Given that $\ell = L/E$, we
recover one of the properties listed in Sec.~\ref{s:gek:th_motion}:
a geodesic cannot encounter the rotation axis unless it has $L = 0$.
Moreover, on the rotation axis,
the vectors $\w{e}_{(\th)}$
and $\w{e}_{(\ph)}$ are not defined, due to the singularity of spherical
coordinates there. Consequently, the screen coordinates $(\alpha,\beta)$
cannot be defined by (\ref{e:gik:m_screen_ab}). In particular, the rotation axis appears as
single point on the screen, which forbids to use it to define the $\beta$-axis.
One has then to pick an arbitrary orthonormal frame
$(\w{e}_{(\alpha)}, \w{e}_{(\beta)})$ in the screen plane to define
$(\alpha,\beta)$. Formulas (\ref{e:gik:shadow_param_eq}) do no longer
hold, but formula (\ref{e:gik:alpha2_beta2}) is still valid,
since the distance from the screen's center is a quantity independent from
the orientation of the frame $(\w{e}_{(\alpha)}, \w{e}_{(\beta)})$. Another
way to see that (\ref{e:gik:alpha2_beta2}) is still valid is to notice that
it admits a well-defined limit for $\th_{\Obs}\to 0$ or $\pi$. Taking into
account $\ell=0$, we obtain
\be \label{e:gik:alpha2_beta2_axis}
\alpha^2 + \beta^2 = \frac{1}{r_{\Obs}^2} \left( q + a^2\right)
\qquad (\th_{\Obs} = 0\ \mbox{or}\ \th_{\Obs} = \pi).
\ee
\subsection{Latitudinal motion} \label{s:gik:th_motion}
Specializing the general results expressed in Property~\ref{p:gek:latitudinal_motion} to
$\mu=0$ and $E\neq 0$, we get
\begin{prop}[latitudinal motion of null geodesics]
\begin{itemize}
\item A null geodesic $\Li$ of Kerr spacetime cannot encounter the rotation axis unless it has $\ell=0$.
\item If $|\ell|\geq a$,
the reduced Carter constant $q$ is necessarily non-negative:
\be \label{e:gik:q_non-negative}
q \geq 0 .
\ee
\item The reduced Carter constant $q$ can take negative values only if $|\ell|<a$
(which implies $a\neq 0$); its range is then
limited from below:
\be \label{e:gik:q_min}
q \geq q_{\rm min} = - \left( a - |\ell| \right) ^2.
\ee
If $q<0$, $\Li$ is called a \defin{vortical null geodesic}\index{vortical!geodesic}; it
never encounters the equatorial plane.
\item If $q>0$ and $\ell\not=0$, $\Li$ oscillates symmetrically about the equatorial plane,
between two $\th$-turning points, at $\th=\th_{\rm m}$ and $\th=\pi-\th_{\rm m}$,
where $\th_{\rm m}\in(0,\pi/2)$
is given by Eqs.~(\ref{e:gek:th0_a2E2mu2_zero}) and (\ref{e:gek:th0_E_gt_mu}):
\begin{align}
& \th_{\rm m} = \arccos\sqrt{\frac{q}{\ell^2 + q}} \quad\mbox{for}\quad a=0\\
& \th_{\rm m} = \arccos \sqrt{ \frac{1}{2} \left[ 1 - \frac{\ell^2 + q}{a^2}
+ \sqrt{ \left(1 - \frac{\ell^2 + q}{a^2}\right) ^2
+ \frac{4q}{a^2} } \right] } \quad\mbox{for}\quad a\neq 0 . \label{e:gik:th0}
\end{align}
If $q>0$ and $\ell=0$, $\Li$
crosses repeatedly the rotation axis, with $\th$ taking all values in the
range $[0,\pi]$.
\item If $q=0$, $\Li$ is stably confined to the equatorial plane
for $|\ell| > a$ or $|\ell| = a\neq 0$;
for $|\ell| < a$, $\Li$ either lies unstably in the equatorial
plane or approaches it asymptotically from one side, while for $\ell=0$ and $a=0$,
$\Li$ lies at a constant value $\th=\th_0\in[0,\pi]$.
\item If $q_{\rm min} < q < 0$, $\Li$ never encounters the equatorial plane,
having a $\th$-motion entirely confined either to the Northern hemisphere
($0<\th<\pi/2$) or to
the Southern one ($\pi/2<\th<\pi$); if $\ell\neq 0$, $\Li$ oscillates between
two $\th$-turning points, at $\th=\th_{\rm m}$ and $\th=\th_{\rm v}$ (Northern hemisphere)
or at $\th=\pi-\th_{\rm v}$ and $\th=\pi-\th_{\rm m}$ (Southern hemisphere), where
$\th_{\rm m}$ is given by Eq.~(\ref{e:gik:th0}) above and $\th_{\rm v}$ is given
by Eq.~(\ref{e:gek:th1}):
\be \label{e:gik:th1_general}
\th_{\rm v} = \arccos \sqrt{ \frac{1}{2} \left[ 1 - \frac{\ell^2 + q}{a^2}
- \sqrt{ \left(1 - \frac{\ell^2 + q}{a^2}\right) ^2
+ \frac{4q}{a^2}} \right] } ;
\ee
if $\ell=0$,
$\Li$ oscillates about the rotation axis, with a $\th$-turning point at
$\th=\th_{\rm v}$ or $\th = \pi - \th_{\rm v}$, where $\th_{\rm v}$
is given by Eq.~(\ref{e:gek:th0_L_zero}), or equivalently by the $\ell\to 0$
limit of Eq.~(\ref{e:gik:th1_general}):
\be \label{e:gik:thv_ell_zero}
\th_{\rm v} = \arccos \left( \frac{\sqrt{-q}}{a} \right) \qquad (\ell=0) .
\ee
\item If $q = q_{\rm min}$, $\Li$ lies stably at a constant value
$\th=\th_*$ or $\th = \pi - \th_*$, with $\th_*\in [0, \pi/2)$%]$
\ given by
\be \label{e:gik:th_s_ell_over_a}
\th_* := \arcsin\sqrt{\frac{|\ell|}{a}} .
\ee
\end{itemize}
\end{prop}
\begin{remark}
For $\ell < 0$, the constraint (\ref{e:gik:q_min}) is tighter than
(\ref{e:gik:q_ell_a_constraint}).
\end{remark}
\begin{example}[Principal null geodesic]
A principal null geodesic moves at a constant angle
$\th=\th_0$ and has
$\ell = a\sin^2\th_0$ [Eq.~(\ref{e:gik:principal_null_l_q})].
For $\th_0\neq \pi/2$, we have $|\ell| < a$ and
Eq.~(\ref{e:gik:q_min}) yields $q_{\rm min} = - a^2 \cos^4\th_0$.
Comparing with the value of $q$ given by Eq.~(\ref{e:gik:principal_null_l_q}),
we note that
\be
q = q_{\rm min} ,
\ee
which agrees with the last case listed above, i.e. motion at constant
$\th$, with $\th_0 = \th_*$ or $\th_0 = \pi - \th_*$ according
to Eq.~(\ref{e:gik:th_s_ell_over_a}), since $\sqrt{|\ell|/a} = \sin\th_0$.
\end{example}
\subsection{Radial motion} \label{s:gik:radial_motion}
As for any geodesic, the radial motion of a null geodesic $\Li$ is ruled by
$R(r)\geq 0$ [Eq.~(\ref{e:gek:R_non_neg})], which, in terms of $\mathcal{R}(r)$
[Eq.~(\ref{e:gik:mcR})],
writes
\be \label{e:mcR_non_neg}
\encadre{\mathcal{R}(r) \geq 0 }.
\ee
\begin{example}[Principal null geodesic]
Given the value (\ref{e:gik:mR_PNG}) of $\mathcal{R}(r)$ for
a principal null geodesic, we note that
$\mathcal{R}(r) > 0$ in all Kerr spacetime. This is
consistent with the fact that, for $E\neq 0$ and $\th_0\neq\pi/2$, ingoing principal null geodesics
travel from $r=+\infty$ to $r=-\infty$ (cf. the dashed green curve
in Fig.~\ref{f:ker:3blocks_in}) and outgoing principal null geodesics
travel from $r=-\infty$ to $r=+\infty$ (cf. the solid green curve
in Fig.~\ref{f:ker:3blocks_out}).
\end{example}
\begin{figure}
\begin{tabular}{c@{\hspace{-0.2ex}}c@{\hspace{-0.2ex}}c@{\hspace{-0.2ex}}c}
\includegraphics[height=0.15\textheight]{gik_R_in_M1_1.pdf}
& \includegraphics[height=0.15\textheight]{gik_R_in_M1_2.pdf}
&\includegraphics[height=0.15\textheight]{gik_R_in_M1_3.pdf}
&\includegraphics[height=0.15\textheight]{gik_R_in_M1_4.pdf}\\[-1.5ex]
(a) & (b) & (c)& (d)
\end{tabular}
\caption[]{\label{f:gik:R_in_M1} \footnotesize
Quartic polynomial $\mathcal{R}(r)$ in the region $\M_{\rm I}$ for
four values of $(\ell, q)$ and for $a=0.95\, m$.
The grey area marks the black hole region, with the vertical
black line at the event horizon, at $r=r_+ \simeq 1.312\, m$.
The green part of the $r$ axis corresponds to $\mathcal{R}(r) \geq 0$,
i.e. to regions where the geodesic motion is allowed.
\textsl{[Figure generated by the notebook \ref{s:sam:Kerr_null_geod_plots}]}
}
\end{figure}
Since $\mathcal{R}(r)$ is a polynomial of degree $4$ in $r$, its
behavior can be relatively complicated. However, it turns out that in the
black hole exterior, its behavior is quite simple, according to the
following lemma:
\begin{lemma}[behavior of $\mathcal{R}(r)$ in $\M_{\rm I}$]
\label{p:gik:lem_R_r}
In region $\M_{\rm I}$ of Kerr spacetime, i.e. for $r> r_+$, the
quartic polynomial $\mathcal{R}(r)$ associated to a given geodesic
[Eq.~(\ref{e:gik:mcR})] has one of the following behaviors, depending
on the value of $(\ell, q)$:
\begin{enumerate}
\item $\mathcal{R}(r)$ has no root in $(r_+,+\infty)$ and $\mathcal{R}(r) > 0$
there (Fig.~\ref{f:gik:R_in_M1}a);
\item $\mathcal{R}(r)$ has a double root in $(r_+,+\infty)$, at $r=r_0$ say, and
$\mathcal{R}(r) > 0$ iff $r\in(r_+, r_0)$ or $r\in(r_0, +\infty)$
(Fig.~\ref{f:gik:R_in_M1}b);
\item $\mathcal{R}(r)$ has two simple roots in $(r_+,+\infty)$, at $r=r_1$ and $r=r_2$ say, and
$\mathcal{R}(r) > 0$ iff $r\in(r_+, r_1)$ or $r\in(r_2, +\infty)$
(Fig.~\ref{f:gik:R_in_M1}c);
\item $\mathcal{R}(r)$ has a unique simple root in $(r_+,+\infty)$, at $r=r_0$ say; then
necessarily $\ell = 2 m r_+ /a$, $\mathcal{R}(r_+) = 0$ and
$\mathcal{R}(r) > 0$ iff $r\in(r_0, +\infty)$
(Fig.~\ref{f:gik:R_in_M1}d).
\end{enumerate}
\end{lemma}
\begin{proof}
First, we observe that $\mathcal{R}(r)$ is positive or zero at both
ends of $\M_{\rm I}$. This is clear
for the asymptotic flat end since, according to expression~(\ref{e:gik:mcR_powers}),
$\mathcal{R}(r) \sim r^4 > 0$ when $r\to +\infty$. At the inner end,
namely for $r \to r_+$,
we have $\Delta \to 0$ and expression~(\ref{e:gik:mcR_Delta}) yields the limit value
$\mathcal{R}(r_+) = (r_+^2 + a^2 - a \ell)^2 \geq 0$.
Now, as a polynomial of degree four, $\mathcal{R}(r)$ has at most four real roots.
Given the above boundary conditions, $\mathcal{R}(r)$ can have
a unique simple root in $(r_+,+\infty)$ only if $\mathcal{R}(r_+) = 0$ (cf. Fig.~\ref{f:gik:R_in_M1}d),
which occurs for a very specific value of $\ell$, namely $\ell = (r_+^2 + a^2)/a = 2 m r_+/a$.
This is case~4 above.
Similarly, the case of three roots in $(r_+,+\infty)$, either three simple roots or one double and one simple root,
is compatible with the boundary conditions only if $\mathcal{R}(r_+) = 0$, i.e.
if $r_+$ is a fourth root of $\mathcal{R}(r)$.
But then the four roots of $\mathcal{R}(r)$ would be positive (since $r_+ > 0$).
Now, since there is no $r^3$ term in the expression~(\ref{e:gik:mcR_powers}) of $\mathcal{R}(r)$,
the sum of the roots of $\mathcal{R}(r)$ is zero, which is impossible with all
the roots being positive. Hence there cannot be three roots in $(r_+,+\infty)$.
The same argument about the zero sum of the roots excludes as well the case
of four roots of $\mathcal{R}(r)$ in $(r_+,+\infty)$.
There remains then the cases of no root at all in $(r_+,+\infty)$ (case 1 of the lemma) and that of two
roots (cases 2 and 3). These three cases are compatible
with the boundary conditions and the vanishing of the sum of the roots.
\end{proof}
\begin{remark}
Case~4 of Lemma~\ref{p:gik:lem_R_r} can be seen as the limit $r_1\to r_+$ of case~3.
\end{remark}
\begin{remark}
In case~4, $\mathcal{R}(r_+) = 0$ occurs for $\Omega_{\Hor} \ell = 1$, where $\Omega_{\Hor}$ is the black hole
rotation velocity\index{black!hole!rotation velocity}\index{rotation!velocity},
as given by Eq.~(\ref{e:ker:def_OmegaH}). Similarly, $\mathcal{R}(r_-) = 0$ for $\Omega_{\rm in} \ell = 1$,
where $\Omega_{\rm in}$ is the rotation velocity of the inner horizon, cf. Eq.~(\ref {e:ker:def_Omega_in}).
\end{remark}
\begin{remark} \label{r:gik:R_zero_M_III}
The region $\M_{\rm III}$ shares with $\M_{\rm I}$ that $\mathcal{R}(r)$ is
non-negative at each of its ends: $\mathcal{R}(r) \sim r^4 > 0$ when $r\to -\infty$
and $\mathcal{R}(r_-) = (r_-^2 + a^2 - a \ell)^2 \geq 0$. However, the argument
used to limit the number of roots in $\M_{\rm I}$
cannot be applied to $\M_{\rm III}$ because the latter can accommodate for both positive and negative
values of $r$ and thus four roots of $\mathcal{R}(r)$ can be located in $\M_{\rm III}$.
\end{remark}
According to the definition given in Sec.~\ref{s:gek:turning_points},
a $r$-turning of a null geodesic $\Li$ is a point $p_0\in \Li$
such that $r_0 := r(p)$ is a simple root of $\mathcal{R}(r)$.
Lemma~\ref{p:gik:lem_R_r} leads then to:
\begin{prop}[$r$-turning points of null geodesics]
A null geodesic in Kerr spacetime has
\begin{itemize}
\item at most one $r$-turning point in region $\M_{\rm I}$;
\item no $r$-turning point in region $\M_{\rm II}$.
\end{itemize}
\end{prop}
\begin{proof}
If a geodesic would have two turning points in $\M_{\rm I}$, this would mean
that there exist two simple roots in $\M_{\rm I}$ and that
$\mathcal{R}(r)$ is positive between them (so that the motion is possible
there). But this is excluded by Lemma~\ref{p:gik:lem_R_r}.
In region $\M_{\rm II}$, we have $\Delta < 0$ and
Eqs.~(\ref{e:gik:mcR_Delta}) and (\ref{e:gik:q_ell_a_constraint}) show that
$\mathcal{R}(r)$ is the sum of two non-negative terms. The only possibility
to have $\mathcal{R}(r)=0$ is then that each term vanishes separately:
$r^2 + a^2 - a\ell = 0$ and $q + (\ell -a)^2 = 0$, i.e.
\[
r^2 = a (\ell - a) \qand q = - (\ell - a)^2 .
\]
The second equation implies $q\leq 0$. The case $q=0$ would lead to $\ell = a$ and $r^2 = 0$,
which is impossible in $\M_{\rm II}$. There remains $q < 0$, but
according to the results of Sec.~\ref{s:gik:th_motion}, this implies $|\ell| < a$, so
that $\ell - a < 0$ and the first equation above would yield $r^2 < 0$, which is impossible.
There is thus no $r$-turning point in $\M_{\rm II}$.
\end{proof}
\begin{remark}
That no $r$-turning point can exist in $\M_{\rm II}$ has been established
above by some considerations on $\mathcal{R}(r)$. This property can also be deduced as
an immediate consequence of the result (\ref{e:gek:r_decay_MII}), namely
that $r$ must be a strictly decreasing function of $\lambda$ at any point
in $\M_{\rm II}$.
\end{remark}
Let us consider case~2 of Lemma~\ref{p:gik:lem_R_r} (double root of $\mathcal{R}(r)$ in $\M_{\rm I}$);
it corresponds to two distinct situations regarding the null geodesic $\Li$ having
$\mathcal{R}(r)$ as radial polynomial. First, $\Li$ can lie at a constant value
of $r$, which is necessarily the double root $r_0$ of $\mathcal{R}(r)$
by virtue of Eq.~(\ref{e:gik:drdl_Mino}); $\Li$ belongs then to the category
of the \emph{spherical photon orbits}, which will be
studied in Sec.~\ref{s:gik:spherical_orbits}. If $r$ is not constant along $\Li$,
then according to the definition in Sec.~\ref{s:gek:asymptotic_values},
$\Li$ has as an asymptotic $r$-value, which is the double root $r_0$.
Given that $r_0$ is the only root of $\mathcal{R}(r)$ in $(r_+,+\infty)$,
Eq.~(\ref{e:gik:drdl_Mino}) implies
$\D r/\D\lambda\neq 0$ all along $\Li$, so that $r(\lambda)\to r_0$ for
$\lambda\to+\infty$ (future asymptotic value)
or $\lambda\to-\infty$ (past asymptotic value). Such a geodesic belong to the
category of the \emph{critical null geodesics}, which will be studied in
Sec.~\ref{s:gik:critical_geod}.
%\begin{remark}
%Since their sum is zero, when the four roots of $\mathcal{R}(r)$ coincide,
%they must be equal to zero and we have $\mathcal{R}(r) = r^4$. This occurs for $\ell=a$ and $q=0$, i.e.
%for a principal null geodesics in the equatorial plane, cf. Eq.~(\ref{e:gik:principal_null_l_q})
%with $\th_0=\pi/2$.
%\end{remark}
In view of the above results, we can state:
\begin{prop}[radial behavior of null geodesics]
In the region $\M_{\rm I}$ of Kerr spacetime,
any null geodesic $\Li$ has one of the following behaviors.
For \emph{generic} values of the constant of motions $(\ell, q)$, the possibilities are:
\begin{enumerate}
\item $\Li$ arises from the past null infinity of $\M_{\rm I}$, $\scri^-$, has a $r$-turning point,
which we may call the \defin{periastron}\index{periastron}, and terminates at the future null infinity
of $\M_{\rm I}$, $\scri^+$;
\item $\Li$ arises from the past null infinity $\scri^-$, has $r$ decreasing
monotonically and crosses the black hole event horizon $\Hor$;
\item $\Li$ arises from the past event horizon $\Hor^-$ separating $\M_{\rm I}$
from the white hole region $\M_{\rm II}^*$, has $r$ increasing monotonically and
terminates at the future null infinity $\scri^+$;
\item $\Li$ arises from the past event horizon $\Hor^-$, has a $r$-turning point,
which we may call the \defin{apoastron}\index{apoastron}, and crosses the black hole event horizon $\Hor$;
\end{enumerate}
For some \emph{specific} values of the constant of motions $(\ell, q)$,
forming a 1-dimensional (hence zero-measure) subset of the set of all possible values\footnote{This
subset is given in parametric form by Eqs.~(\ref{e:gik:spher_ell_r0})-(\ref{e:gik:spher_q_r0}) below.},
the possibilities are:
\begin{enumerate}
\setcounter{enumi}{4}
\item $\Li$ evolves at a fixed value of $r$;
\item $\Li$ arises from the past null infinity $\scri^-$
and has $r$ decreasing monotonically to
a future asymptotic $r$-value at $r_0 > r_+$;
\item $\Li$ arises from a past asymptotic $r$-value at $r_0 > r_+$, has
$r$ increasing monotonically and terminates at the future null infinity
$\scri^+$;
\item $\Li$ arises from a past asymptotic $r$-value at $r_0 > r_+$, has
$r$ decreasing monotonically and crosses the black hole event horizon $\Hor$;
\item $\Li$ arises from the past event horizon $\Hor^-$, has
$r$ increasing monotonically to
a future asymptotic $r$-value at $r_0 > r_+$.
\end{enumerate}
\end{prop}
Case 1 corresponds to a scattering\index{scattering} trajectory, leading to the standard phenomenon of deflection of light\index{deflection!of light}. The polynomial $\mathcal{R}(r)$
belongs then to case~3 or 4 of Lemma~\ref{p:gik:lem_R_r}.
Ingoing principal null geodesics belong to case 2, while the outgoing ones with $E\neq 0$ belong to case 3
(cf. Example~\ref{x:gik:png_R_positive} below).
Both cases 2 and 3 correspond to case 1 of Lemma~\ref{p:gik:lem_R_r} (no root of $\mathcal{R}(r)$ in $\M_{\rm I}$).
Cases~5--9 correspond to case 2 of Lemma~\ref{p:gik:lem_R_r} (double root of $\mathcal{R}(r)$ in $\M_{\rm I}$).
Case 5 is that of \emph{spherical photon orbits} and will be discussed in Sec.~\ref{s:gik:spherical_orbits}, while cases~6--9 are those of \emph{critical null geodesics},
to be discussed in Sec.~\ref{s:gik:critical_geod}.
\begin{remark}
The terminology \emph{(periastron, apoastron)} employed here agrees with that
introduced for the Schwarzschild case in Sec.~\ref{s:gis:null_radial_behav}.
\end{remark}
\begin{example}[principal null geodesics] \label{x:gik:png_R_positive}
That the principal null geodesics with $E\neq 0$ belong to cases 2 and 3 above
is clear from their value (\ref{e:gik:mR_PNG}) for $\mathcal{R}(r)$:
$\mathcal{R}(r) = \rho^4 > 0$ in all $\M$, which precludes the existence
of any $r$-turning point nor any $r$-asymptotic value along these geodesics.
\end{example}
\begin{remark}
As a sequel of Remark~\ref{r:gik:R_zero_M_III} above, a null geodesic can have
two turnings points in region $\M_{\rm III}$. There can thus exist null geodesics that are trapped
between two distinct values of $r$ in $\M_{\rm III}$.
\end{remark}
%\section{Principal null geodesics}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%