-
Notifications
You must be signed in to change notification settings - Fork 0
/
IV-elliptic.tex
1058 lines (975 loc) · 45.9 KB
/
IV-elliptic.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{\texorpdfstring{$\ell$}{ℓ}-adic representations attached to elliptic
curves}\label{ch:iv}
\chaptermark{Elliptic curves}
Let $K$ be a number field and let $E$ be an elliptic curve over $K$.
\dpage
If $\ell$ is a prime number, let
\[
\rho_\ell \colon \Gal(\algcl K/K) \longrightarrow \Aut(V_\ell(E))
\]
be the corresponding $\ell$-adic representation of $K$, cf.\ chap.~\ref{ch:i},
\ref{sec:I_12}. The main result of this Chapter is the determination of the
Lie algebra of the $\ell$-adic Lie group $G_\ell = \Img(\rho_\ell)$. This is
based on a finiteness theorem of \v Safarevi\v c (\ref{sec:IV_14}) combined
with the properties of locally algebraic abelian representations
(chap.~\ref{ch:iii}) and Tate's local theory of elliptic curves with
non-integral modular invariant (Appendix, \ref{sec:IV_A1}). The variation of
$G_\ell$ with $\ell$ is studied in \S\ref{sec:IV_3}.
The Appendix gives analogous results in the local case (i.e.\ when $K$ is a
local field).
\section{Preliminaries}%
\dpage
\subsection[Elliptic curves]{Elliptic curves \textmd{(cf.\
\citeauthor{5}~\cite{5}, \citeauthor{9}~\cite{9}, \citeauthor{10}~\cite{10})}}
\label{sec:IV_11}
By an elliptic curve, we mean an abelian variety of dimension
1, i.e.\ a complete, non singular, connected curve of genus 1 with a
given rational point $P_0$, taken as an origin for the composition law
(and often written $o$).
Let $E$ be such a curve. It is well known that $E$ may be embedded, as a
non-singular cubic, in the projective plane $\PP^2_K$, in such a way that $P_0$
becomes a \textquote{flex} (one takes the projective embedding defined by the
complete linear series containing the divisor $3\cdot P_0$). In this embedding,
three points $P_1$, $P_2$, $P_3$ have sum $0$ if and only if the divisor $P_1 +
P_2 + P_3$ is the intersection of $E$ with a line. By choosing a suitable
coordinate system, the equation of $E$ can be written in Weierstrass form
\[
y^2 = 4x^3 - g_2x - g_3
\]
where $x$, $y$ are non-homogeneous coordinates and the origin $P_0$ is
the point at infinity on the $y$-axis. The discriminant
\[
\Delta = g_2^3 - 27g_3^2
\]
is non-zero.
The coefficients $g_2$, $g_3$ are determined up to the transformations $g_2
\mapsto u^4 g_2$, $g_3 \mapsto u^6 g_3$, $u \in K^\times$. The modular
invariant $j$ of $E$ is
\[
j = 2^6 \, 3^3 \, \frac{g_2^3}{g_2^3 - 27g_3^2}
= 2^6 \, 3^3 \, \frac{g_2^3}{\Delta}.
\]
\dpage
Two elliptic curves have the same $j$ invariant if and only if they become
isomorphic over the algebraic closure of $K$.
(All this remains valid over an arbitrary field, except that, when the
characteristic is 2 or 3, the equation of $E$ has to be written in the more
general form
\[
y^2 + a_1xy + a_3y + x^3 + a_2x^2 + a_4x + a_6 = 0.
\]
Here again, 0 is the point at infinity on the $y$-axis and the
corresponding tangent is the line at infinity. There are corresponding
definitions for $\Delta$ and $j$, for which we refer to \citeauthor{9}~\cite{9}
or \citeauthor{20}~\cite{20}; note, however, that there is a misprint in Ogg's
formula for $\Delta$: the coefficient of $\beta_4^3$ should be $-8$ instead of
$-1$.)
\subsection{Good reduction}
\label{sec:IV_12}
Let $v \in M_K^0$ be a finite place of the number field $K$. We denote by
$\mathcal{O}_v$ (resp.\ $\mathfrak{m}_v$, $k_v$) the corresponding local ring
in $K$ (resp.\ its maximal ideal, its residue field).
Let $E$ be an elliptic curve over $K$. One says that $E$ has \strong{good
reduction at $v$} if one can find a coordinate system in $\PP^2_K$ such that
the corresponding equation $f$ for $E$ has coefficient in $\mathcal{O}_v$ and
its reduction $\tilde f \mod{\mathfrak{m}_v}$ defines a non-singular cubic
$\widetilde{E}_v$ (hence an elliptic curve) over the residue field $k_v$ (in
other words, the discriminant $\Delta(f)$ of $f$ must be an invertible element
of $\mathcal{O}_v$).
The curve $\widetilde{E}_v$ is called the \strong{reduction} of $E$ at $v$;
\dpage
it does not depend on the choice of $f$, provided, of course, that $\Delta(f) \in
\mathcal{O}_v^\times$.
One can prove that the above definition is equivalent to the following one:
there is an abelian scheme $E_v$ over $\Spec(\mathcal{O}_v)$, in the sense of
\citeauthor{19}~\cite{19}, chap.\ VI, whose generic fiber is $E$; this scheme
is then unique, and its special fiber is $\widetilde{E}_v$. Note that
$\widetilde{E}_v$ is defined over the finite field $k_v$; we denote its
\strong{Frobenius endomorphism} by $F_v$.
On either definition, one sees that $E$ has \strong{good reduction for
almost all places of $K$}.
If $E$ has good reduction at a given place $v$, its $j$ invariant is
\strong{integral at $v$} (i.e.\ belongs to $\mathcal{O}_v$) and its reduction
$\tilde\jmath \mod{\mathfrak{m}_v}$ is the $j$ invariant of the reduced curve
$\widetilde{E}_v$.
The converse is almost true, but not quite: if $j$ belongs to $\mathcal{O}_v$,
there is a finite extension $L$ of $K$ such that $E \otimes_K L$ has good
reduction at all the places of $L$ dividing $v$ (this is the ``potential good
reduction'' of \citeauthor{32}~\cite{32}, \S 2). For the proof of this, see
\citeauthor{29}~\cite{29}, \S 4, n\textsuperscript{o}~3.
\begin{obs}
The definitions and results of this section have nothing to do with number
fields. They apply to every field with a discrete valuation.
\end{obs}
\subsection{Properties of \texorpdfstring{$V_\ell$}{Vℓ} related to good
reduction}\label{sec:IV_13}
Let $\ell$ be a prime number. We define, as in chap.~\ref{ch:i},
\ref{sec:I_12}, the Galois modules $T_\ell$ and $V_\ell$ by:
\[
V_\ell = T_\ell \otimes \Q_\ell, \qquad T_\ell = \invlim_n E_{\ell^n}
\]
where $E_{\ell^n}$ is the kernel of $\ell^n \colon E(\algcl K) \to E(\algcl K)$.
\dpage
We denote by $\rho_\ell$ the corresponding homomorphism of $\Gal(\algcl K/K)$
into $\Aut(T_\ell)$. Recall that $E_{\ell^n}$, $T_\ell$ and $V_\ell$ are of
rank 2 over $\Z/\ell^n\Z$, $\Z_\ell$ and $\Q_\ell$, respectively.
Let now $v$ be a place of $K$, with $p_v \ne \ell$ and let $v$ be some
extension of $v$ to $\algcl K$; let $D$ (resp.\ $I$) be the corresponding
decomposition group (resp.\ inertia group), cf.\ chap.~\ref{ch:i},
\ref{sec:I_21}. If $E$ has good reduction at $v$, one easily sees that
reduction at $v$ defines an \emph{isomorphism} of $E_{\ell^n}$ onto the
corresponding module for the reduced curve $\widetilde{E}_v$. In particular,
$E_{\ell^n}$, $T_\ell$, $V_\ell$ are \emph{unramified at $v$}
(chap.~\ref{ch:i}, \ref{sec:I_21}) and the Frobenius automorphism $F_{v,
\rho_\ell}$ of $T_\ell$ corresponds to the Frobenius endomorphism $F_v$ of
$\widetilde{E}_v$. Hence:
\[
\det(F_{v, \rho_\ell}) = \det(F_v) = \numnorm v
\]
and
\[
\det(1 - F_{v, \rho_\ell}) = \det(1 - F_v) = 1 - \tr(F_v) + \numnorm v
\]
is equal to the number of $k_v$-points of $\widetilde{E}_v$.
Conversely:
\begin{thm}[Criterion of Néron-Ogg-\v Safarevi\v c]
If $V$ is unramified at $v$ for some $\ell \ne p_v$, then $E$ has good
reduction at $v$.
\end{thm}
For the proof, see \citeauthor{32}~\cite{32}, \S 1.
\begin{cor}
Let $E$ and $E'$ be two elliptic curves which are isogenous (over $K$). If one
of them has good reduction at a place $v$, the same is true for the other one.
\end{cor}
(Recall that $E$ and $E'$ are said to be \strong{isogenous} if there
\dpage
exists a non-trivial morphism $E \to E'$.)
This follows from the theorem, since the $\ell$-adic representations associated
with $E$ and $E'$ are isomorphic.
\begin{obs}
For a direct proof of this corollary, see \citeauthor{11}~\cite{11}.
\end{obs}
\subsubsection*{Exercise}
Let $S$ be the finite set of places where $E$ does not have good reduction. If
$v \in M_K^0 \setminus S$, we denote by $t_v$ the number of $k_v$-points of the
reduced curve $\widetilde{E}_v$.
\begin{enumerate}[(a)]
\item Let $\ell$ be a prime number and let $m$ be a positive integer.
Show that the following properties are equivalent:
\begin{enumerate}[(i)]
\item\label{exr:cebotarev_i}
$t_v \equiv 0 \mod{\ell^m}$ for all $v \in M_K^0 \setminus S$, $p_v \ne \ell$.
\item\label{exr:cebotarev_ii}
The set of $v \in M_K^0 \setminus S$ such that $t_v
\equiv 0 \mod{\ell^m}$ has density one (cf.\
chap.~\ref{ch:i}, \ref{sec:I_22}).
\item\label{exr:cebotarev_iii}
For all $s \in \Img(\rho)$, one has $\det(1-s) \equiv 0 \mod{\ell^m}$.
\end{enumerate}
(The equivalence of \ref{exr:cebotarev_ii} and \ref{exr:cebotarev_iii}
follows from \v Cebotarev's density theorem. The implications
\ref{exr:cebotarev_i} $\implies$ \ref{exr:cebotarev_ii} and
\ref{exr:cebotarev_iii} $\implies$ \ref{exr:cebotarev_i} are easy.)
\item We take now $m = 1$. Show that the properties
\ref{exr:cebotarev_i}, \ref{exr:cebotarev_ii} and
\ref{exr:cebotarev_iii} are equivalent to:
\begin{enumerate}[resume*]
\item\label{exr:cebotarev_iv}
There exists an elliptic curve $E'$ over $K$ such that:
\begin{itemize}
\item[$(\alpha)$]
Either $E'$ is isomorphic to $E$, or there
exist an isogeny $E' \to E$ of degree $\ell$.
\item[$(\beta)$]
The group $E'(K)$ contains an element of order
$\ell$.
\end{itemize}
\end{enumerate}
(The implication \ref{exr:cebotarev_iv} $\implies$
\ref{exr:cebotarev_iii} is easy. For the proof of the converse, use
Exer.~\ref{ex:I11_ex2} of chap.~\ref{ch:i}, \ref{sec:I_11}.)
[For $m > 2$, see \citeauthor{64}~\cite{64}.]
\end{enumerate}
\subsection{\v Safarevi\v c's theorem}%
\label{sec:IV_14}
\dpage
It is the following (cf.\ \cite{23}):
\begin{thm}
Let $S$ be a finite set of places of $K$. The set of isomorphism classes of
elliptic curves over $K$, with good reduction at all places not in $S$, is
finite.
\end{thm}
Since isogenous curves have the same bad reduction set (cf.\ \ref{sec:IV_13}),
this implies:
\begin{cor}
Let $E$ be an elliptic curve over $K$. Then, up to isomorphism, there are only
a finite number of elliptic curves which are $K$-isogenous to $E$.
\end{cor}
To prove the theorem, we use the following criterion for good reduction:
\begin{lem}
Let $S$ be a finite set of places of $K$ containing the divisors of 2 and 3,
and such that the ring $\mathcal{O}_S$ of $S$-integers is principal. Then, an
elliptic curve $E$ defined over $K$ has good reduction outside $S$ if and only
if its equation can be put in the Weierstrass form $y^2 = 4x^3 - g_2 x - g_3$
with $g_i \in \mathcal{O}_S$ and $\Delta = g_2^3 - 27 g_2^3 \in
\mathcal{O}_S^\times$ (the group of units of $\mathcal{O}_S$).
\end{lem}
\begin{proof}
The sufficiency is trivial. To prove necessity, we write the
curve $E$ in the form
\begin{equation}
y^2 = 4x^3 - g_2^\prime x - g_3^\prime
\tag{$*$}
\label{eq:safarevich_lemma}
\end{equation}
with $g_i^\prime \in K$. Let $v$ be a place of $K$ not in $S$. Then, since
there is good reduction at $v$, and since the divisors of 2 and 3 do not belong
\dpage
to $S$, the curve $E$ can be written in the form
\[
y^2 = 4x^3 - g_{2, v}^\prime x - g_{3, v}^\prime
\]
with $g_{i, v}$ in the local ring at $v$ and the discriminant $\Delta_v$ a unit
in this ring. Using the properties of the Weierstrass form, there is an element
$u_v \in K$ such that $g_{2, v} = u_v^4 g^\prime_2$, $g_{3, v} = u_v^6
g^\prime_3$, $\Delta_v = u_v^{12} \Delta'$;\label{errata:uv12} moreover, as we
can take $g_{i,v} = g_i^\prime$ for almost all $v$, we see that we can assume
that $u_v = 1$ for almost all $v \notin S$. Since the ring $\mathcal{O}_S$ is
principal, there is an element $u \in K^\times$ with $v(u) = v(u_v)$ for all $v
\notin S$. Then, if we replace $x$ by $u^{-2} x$ and $y$ by $u^{-3} y$ in
\eqref{eq:safarevich_lemma}, the curve $E$ takes the form
\[
y^2 = 4x^3 - g_2^\prime x - g_3^\prime
\]
with $g_2 = u^4 g^\prime_2$, $g_3 = u^6 g^\prime_3$ and $\Delta = u^{12}
\Delta'$. Since, by construction, $g_i \in \mathcal{O}_S$ and $\Delta \in
\mathcal{O}_S^\times$ the lemma is established.
\end{proof}
\begin{proof}[ of the theorem]
After possibly adding a finite number of places of $K$ to $S$, we may assume
that $S$ contains all the divisors of 2 and 3, and that the ring
$\mathcal{O}_S$ is principal. If $E$ is an elliptic curve defined over $K$
having good reduction outside $S$, the above lemma tells us that we can write
$E$ in the form
\begin{equation}
y^2 = 4x^3 - g_2^\prime x - g_3^\prime
\tag{$*$}
\label{eq:wei_form_star2}
\end{equation}
with $g_i \in \mathcal{O}_S$ and $\Delta = g_2^3 - 27 g_2^3 \in \mathcal{O}_S$.
But, since we are free to multiply $\Delta$ by any $u \in
{(\mathcal{O}_S^\times)}^{12}$, and since
$\mathcal{O}_S^\times/{(\mathcal{O}_S^\times)}^{12}$ is a finite group, we see
that there is a finite set $X \subset \mathcal{O}_S^\times$ such that any
elliptic
\dpage
curve of the above type can be written in the form \eqref{eq:wei_form_star2}
with $g_i \in \mathcal{O}_S$ and $\Delta \in X$. But, for a given $\Delta$, the
equation
\[
U^3 - 27V^2 = \Delta
\]
represents an affine elliptic curve. Using a theorem of Siegel (generalized by
Mahler and Lang, cf.\ \citeauthor{14}~\cite{14}, chap.~VII), one sees that this
equation has only a \emph{finite} number of solutions in $\mathcal{O}_S$. This
finishes the proof of the theorem.
\end{proof}
\begin{obs}
There are many ways in which one can deduce \v Safarevi\v c's theorem from
Siegel's. The one we followed has been shown to us by Tate.
\end{obs}
\section{The Galois module attached to $E$}
In this section, $E$ denotes an elliptic curve over $K$. We are
interested in the structure of the Galois modules $E_{\ell^n}$, $T_\ell$, $V_\ell$
defined in \ref{sec:IV_13}.
\subsection{The irreducibility theorem}
\label{sec:IV_21}
Recall first that the ring $\End_K(E)$ of $K$-endomorphisms of $E$
is either $\Z$ or of rank 2 over $\Z$. In the first case, we say that $E$
has ``no complex multiplication over $K$.'' If the same is true for any
finite extension of $K$, we say that $E$ has ``no complex
multiplication.''
\begin{thm}
Assume that $E$ has no complex multiplication over $K$.
\dpage
Then:
\begin{enumerate}[(a)]
\item\label{thm:IV_21_a} $V_\ell$ is irreducible for all primes $\ell$;
\item $E_\ell$ is irreducible for almost all primes $\ell$.
\end{enumerate}
\end{thm}
We need the following elementary result:
\begin{lem}
Let $E$ be an elliptic curve defined over $K$ with $\End_K(E) = \Z$. Then, if
$E' \to E$, $E'' \to E$ are $K$-isogenies with non-isomorphic cyclic kernels,
the curves $E'$ and $E''$ are non-isomorphic over $K$.
\end{lem}
\begin{proof}
Let $n'$ and $n''$ be respectively the orders of the kernels of
$E' \to E$ and $E'' \to E$. Suppose that $E'$ and $E''$ are isomorphic
over $K$, and let $E' \to E''$ be an isomorphism. If $E \to E'$ is the
transpose of the isogeny $E' \to E$, it has a cyclic kernel of order
$n'$, and hence the isogeny $E \to E$, obtained by composition of
$E \to E'$, $E' \to E''$, $E'' \to E$, has for kernel an extension of
$\Z/n''\Z$ by $\Z/n'\Z$. But, since $\End_K(E) = \Z$, this isogeny must be
multiplication by an integer $a$, and its kernel must therefore be of
the form $\Z/a\Z \times \Z/a\Z$. Hence $n'$ and $n''$ divide $a$. Since
$a^2 = n'n''$, we obtain $a = n' = n''$, a contradiction.
\end{proof}
\begin{proof}[ of the theorem]
\begin{enumerate}[(a)]
\item It suffices to show that, if $\End_K(E) = \Z$, there is no
one-dimensional $\Q_\ell$-subspace of $V_\ell$ stable under
$\Gal(\algcl K/K)$. Suppose there were one; its intersection $X$ with
$T_\ell$ would be a submodule of $T_\ell$ with $X$ and $T_\ell/X$ free
$Z_\ell$-modules of rank 1. For $n \ge 0$, consider the image $X(n)$ of
$X$ in $E_{\ell^n} = T/\ell^n T$. This is a submodule of $E_\ell$ which
is cyclic of order $\ell^n$ and stable by $\Gal(\algcl K/K)$. Hence it
corresponds to a finite $K$-algebraic subgroup of
\dpage
$E$ and one can define the quotient curve $E(n) = E/X(n)$. The kernel
of the isogeny $E \to E(n)$ is cyclic of order $\ell^n$. The above
lemma then shows that the curves $E(n)$, $n \ge 0$, are pairwise
non-isomorphic, contradicting the corollary to \v Safarevi\v c's
theorem (\ref{sec:IV_14}).
\item If $E$ is not irreducible, there exists a Galois submodule $X$ of $E$
which is one-dimensional over $\F_\ell$. In the same way as above, this
defines an isogeny $E \to E/X_\ell$ whose kernel is cyclic of order
$\ell$. The above lemma shows that the curves which correspond to
different values of $\ell$ are\break non-isomorphic, and one again applies
the corollary to \v Safarevi\v c's\break theorem.
\qedhere
\end{enumerate}
\end{proof}
\begin{obs}
One can prove part \ref{thm:IV_21_a} of the above theorem by a quite different
method (cf.\ \cite{25}, \S 3.4); instead of the \v Safarevi\v c's theorem, one
uses the properties of the decomposition and inertia subgroups of
$\Img(\rho_\ell)$, cf.\ Appendix.
\end{obs}
\subsection{Determination of the Lie algebra of \texorpdfstring{$G_\ell$}{Gℓ}}
\label{sec:IV_22}
Let $G_\ell = \Img(\rho_\ell)$\index{Gl@$G_\ell$} denote the image of $\Gal(\algcl K/K)$ in
$\Aut(T_\ell)$, and let $\mathfrak{g}_\ell \subset \End(V_\ell)$ be the Lie
algebra of $G_\ell$.
\begin{thm}
If $E$ has no complex multiplication (cf.\ \ref{sec:IV_21}), then
$\mathfrak{g}_\ell = \End(V_\ell)$, i.e.\ $G_\ell$ is open in
$\Aut(T_\ell)$.
\end{thm}
\begin{proof}
The irreducibility theorem of \ref{sec:IV_21} shows that, for any open
subgroup $U$ of $G_\ell$, $V_\ell$ is an irreducible $U$-module. Hence,
$V_\ell$ is an irreducible $\mathfrak{g}_\ell$-module. By Schur's
lemma, it follows that the commuting algebra $\mathfrak{g}^\prime_\ell$
of $\mathfrak{g}_\ell$ in $\End(V_\ell)$ is a field; since $\dim V_\ell
= 2$, this field is either $\Q_\ell$ or a quadratic extension of
$\Q_\ell$. If $\mathfrak{g}^\prime_\ell = \Q_\ell$, then
$\mathfrak{g}_\ell$ is equal to either $\End(V_\ell)$, or the
subalgebra $\Sl(V_\ell)$
\dpage
% \todo[pinktask]{Revisar si $\Sl(V)$ es la notación adecuada.}
of $\End(V_\ell)$ consisting of the endomorphisms with trace 0; but, in
the second case, the action of $\mathfrak{g}_\ell$ on $\bigwedge^2
V_\ell$ would be trivial, and this would contradict the fact that the
Galois modules $\bigwedge^2 V_\ell$ and $V_\ell(\mu)$ are isomorphic
(chap.~\ref{ch:i}, \ref{sec:I_12}). Hence $\mathfrak{g}_\ell =
\Sl(V_\ell)$ is impossible.
Suppose now that $\mathfrak{g}_\ell^\prime$ is a quadratic extension of
$\Q_p$. Then $V_\ell$ is a one-dimensional
$\mathfrak{g}_\ell^\prime$-vector space and the commuting algebra of
$\mathfrak{g}_\ell^\prime$ in $\End(V_\ell)$ is
$\mathfrak{g}_\ell^\prime$ itself. Hence $\mathfrak{g}_\ell$ is
contained in $\mathfrak{g}_\ell^\prime$, and is \emph{abelian}
($\mathfrak{g}_\ell^\prime$ is a ``non-split Cartan algebra'' of
$\End(V_\ell)$). After replacing $K$ by a finite extension (this does
not affect $\mathfrak{g}_\ell$, cf.\ chap.~\ref{ch:i}, \ref{sec:I_11}),
we may then suppose that $G_\ell$ itself is abelian. The $\ell$-adic
representation $V_\ell$ is then semi-simple, abelian and rational. It
is, moreover, \emph{locally algebraic}. To see this, we first remark
that, at a place $v$ dividing $\ell$, we have $v(j) \ge 0$ since
otherwise the decomposition group of $v$ in $G_\ell$ would be
non-abelian by Tate's theory (cf.\ Appendix, \ref{sec:IV_A13}); hence,
after a finite extension of $K$, we can assume that $E$ has good
reduction at all places $v$ dividing $\ell$ (cf.\ \ref{sec:IV_12}). Let
$E(\ell)$ be the $\ell$-divisible group attached to $E$ at $v$ (cf.\
\citeauthor{39}~\cite{39}, 2.1, example~(a)). We have $V_\ell \cong
V_\ell(E(\ell))$ and this module is known to be of Hodge-Tate type
(\emph{loc. cit.}, \S 4). Using another result of Tate
(chap.~\ref{ch:iii}, \ref{sec:III_12}), this implies that the
representation $V_\ell$ is locally algebraic, as claimed above. (This
could also be seen by using, instead of the theory of Hodge-Tate
modules, the local results of the Appendix, \ref{sec:IV_A2}.)
We may now apply to $V_\ell$ the results of chap.~\ref{ch:iii},
\ref{sec:III_23}. Hence, there is, for each prime $\ell'$, a rational,
abelian, semi-simple $\ell'$-adic representation $W_{\ell'}$ compatible
with $V_\ell$. But $V_{\ell'}$ is compatible with $V_\ell$, and
$V_{\ell'}$ is semi-simple. Hence $V_{\ell'}$, is isomorphic to
$W_{\ell'}$ (cf.\ chap.~\ref{ch:i}, \ref{sec:I_23}). But we know
(chap.~\ref{ch:iii}, \ref{sec:III_23}) that we may choose $\ell'$ such
\dpage
that $W_{\ell'}$ is the direct sum of one-dimensional subspaces stable
under $\Gal(\algcl K/K)$. This contradicts the irreducibility of
$V_\ell$. Hence, we must have $\mathfrak{g}_\ell^\prime = \Q_p$ and
$\mathfrak{g}_\ell = \End(V_\ell)$.
\end{proof}
\begin{obs}
If $E$ has complex multiplication, and $L = \Q \otimes \End(E \otimes_K
\algcl K)$ is the corresponding imaginary quadratic field, one shows
easily that $\mathfrak{g}_\ell$ is the Cartan subalgebra of
$\End(V_\ell)$ defined by $L_\ell = \Q_\ell \otimes L$. It splits if and
only if $\ell$ decomposes in $L$.
\end{obs}
\subsubsection*{Exercises}
(In these exercises, we assume $E$ has no complex multiplication. Let $S$ be
the set of places $v \in M_K^0$ where $E$ has bad reduction. If $v \in M_K^0
\setminus S$, we denote by $F_v$ the Frobenius endomorphism of the reduced
curve $\widetilde{E}_v$; if $\ell \ne p_v$, we identify $F_v$ to the
corresponding automorphism of $T_\ell$.)
\begin{enumerate}
\item Let $H(X, Y)$ be a polynomial in two indeterminates $X$, $Y$ with
coefficients in a field of characteristic zero. Let $V_H$ be the set of
those $v \in M_K^0 \setminus S$ for which $H(\Tr(F_v), \numnorm v) =
0$. If $H$ is not the zero polynomial, show that $V_H$ has density 0.
(Show that the set of $g \in \GL(2, \Z_\ell)$ with $H(\Tr(g), \det(g)) =
0$ has Haar measure zero.)
\item The eigenvalues of $F_v$ may be identified with complex numbers of the
form
\[
(\numnorm v)^{1/2} e^{\pm i \varphi_v},
\qquad 0 \le \varphi_v \le \pi,
\]
cf.\ chap.~\ref{ch:i}, Appendix~\ref{sec:I_A2}. Show that the set of
$v$ for which $\varphi_v$ is a given angle $\varphi$ has density zero.
(Show that $\Tr(F_v)^2 = 4(\numnorm v)\cos^2 \varphi$ and then use the
preceding exercise.)
\item Let $L_v = \Q(F_v)$ be the field generated by $F_v$. By the preceding
\dpage
exercise, $L_v$ is quadratic imaginary except for a set of $v$ of
density 0.
\begin{enumerate}
\item\label{ex:IV_22_3a} Let $\ell$ be a fixed prime. Let $C$ be a semi-simple commutative
$\Q_\ell$-algebra of rank 2. Let $X_C$ be the set of elements
$s \in \Aut(V_\ell)$ such that the subalgebra $\Q_\ell[s]$ of
$\End(V_\ell)$ generated by $s$ is isomorphic to $C$. Show that
$X_C$ is open in $\Aut(V_\ell)$, and show that it has a
non-empty intersection with every open subgroup of
$Aut(V_\ell)$, in particular, with $G_\ell$.
\item Show that $F_v \in X_C$ if and only if the field $L_v$ is
quadratic and $L_v \otimes \Q_\ell$ is isomorphic to $C$.
\item Let $\ell_1, \dots, \ell_n$ be distinct prime numbers, and choose
for each an algebra $C_i$ of the type considered in
\ref{ex:IV_22_3a}. Show that the set of $v$ for which $F_v \in
X_C$ for $i = 1, \dots, n$ has density $> 0$.
(Use the fact that the image of $\Gal(\algcl K/K)$ in any
finite product of the $\Aut(V_\ell)$ is open; this is an easy
consequence of the theorem proved above.)
\item Deduce that, for any finite set $P$ of prime numbers, there exist
an infinity of $v$ such that $L_v$ is ramified at all $\ell \in
P$. In particular, there are an infinite number of distinct
fields $L_v$.
\end{enumerate}
\end{enumerate}
\subsection{The isogeny theorem}
\label{sec:IV_23}
\todo[section]{Belen.}
\section{Variation of \texorpdfstring{$G_\ell$}{Gℓ} and
\texorpdfstring{$\widetilde{G}_\ell$}{Ḡℓ} with \texorpdfstring{$\ell$}{ℓ}}
\subsection{Preliminaries}
\label{sec:IV_31}
\todo[section]{Belen.}
\subsection{The case of a non integral $j$}
\label{sec:IV_32}
\dpage
\begin{thm}
Assume that the modular invariant $j$ of $E$ is not an integer of $K$.
Then $E$ enjoys the equivalent properties (i), (ii), (iii), (iv) of
\ref{sec:IV_31}.
\end{thm}
Since $j$ is not integral, we can choose a place $v$ of $K$ such
that $v(j) < 0$. Let $q$ be the element of the local field $K$ which
corresponds to $j$ by Tate's theory (cf.\ Appendix, \ref{sec:IV_A11}) and let $E$
be the corresponding elliptic curve over $K$. There is a finite
extension $K'$ of $K_v$ over which $E$ and $E_q$ are isomorphic; one
can even take for $K'$ either $K_v$ or a quadratic extension of $K_v$.
Let $v'$ be the valuation of $K'$ which extends $v$; assume $v'$ is
normalized so that $v'({K'}^\times) = \Z$, and let
\[
n = v'(q) = - v'(j)
\]
We have $n > 1$.
\begin{lem}\label{lem:IV_32_1}
Assume $\ell$ does not divide $n$, and let $I_{v, \ell}$ be the inertia
subgroup of $\widetilde{G}_\ell$ corresponding to some extension of $v$
to $\algcl K$. Then $I_{v, \ell}$ contains a transvection, i.e.\ an
element whose matrix form is $
\begin{psmallmatrix}
1 & 1 \\
0 & 1
\end{psmallmatrix}
$ for a suitable $\F_\ell$-basis of $E_\ell$.
\end{lem}
This is true for the curve $E_q$ over $K'$, cf.\ Appendix, \ref{sec:IV_A15}.
The result for $E$ follows from the isomorphism $E_{/K'} \cong {E_q}_{/K'}$.
\begin{lem}\label{lem:IV_32_2}
Let $H$ be a subgroup of $\GL(2, \F_\ell)$ which acts irreducibly on
$\F_\ell\times\F_\ell$ and which contains a transvection. Then $H$
contains $\SL(2, \F_\ell)$.
\end{lem}
For any transvection $s \in H$, let $D$ be the unique one dimensional subspace
of $\F_\ell\times\F_\ell$ which is fixed by $s$. If all such lines were the
same, the line so defined would be stable by $H$, and $H$ would not be
irreducible. Hence there are transvections $s, s' \in H$ such that $D_s \ne
D_{s'}$. If we choose a suitable basis $(e,e')$ of $\F_\ell\times\F_\ell$, this
means that the matrix forms of $s$, $s'$ are
\[
s =
\begin{pmatrix}
1 & 1 \\
0 & 1
\end{pmatrix}, \qquad s' =
\begin{pmatrix}
1 & 0 \\
1 & 1
\end{pmatrix}
\]
The lemma follows then from the well known fact that these two matrices
generate $\SL(2, \F_\ell)$.
\begin{proof}[ of the theorem]
Lemma~\ref{lem:IV_32_1} shows that, for almost all $\ell$, $I_{v,
\ell}$. and \emph{a fortiori} $\widetilde{G}_\ell$, contains a
transvection. On the other hand, we know (cf.\ \ref{sec:IV_21}) that
$\widetilde{G}_\ell$ is irreducible for almost all $\ell$. Applying
lemma~\ref{lem:IV_32_2} to $\widetilde{G}_\ell$ we then see that
$\widetilde{G}_\ell$ contains $\SL(E_\ell)$ for almost all $\ell$;
hence we have (iv).
\end{proof}
\begin{obs}
It seems likely that the condition ``$j$ is not integral'' can be
replaced by the weaker one ``$E$ has no complex multiplication.''
$\to$ [yes: see \cite{76}.]
\end{obs}
\subsection{Numerical example}
\label{sec:IV_33}
\todo[section]{Belen.}
\subsection{Proof of the main lemma of 3.1}
\label{sec:IV_34}
\todo[section]{José.}
\begin{subappendices}
\section{Local results}
\dpage
In what follows, $K$ denotes a field which is complete with respect to a
discrete valuation $v$; we denote by $\mathcal{O}_K$ (resp.\ by $k$) the ring
of integers (resp.\ the residue field) of $K$; we assume that $k$ is perfect
and of characteristic $p \ne 0$.
Let $E$ be an elliptic curve over $K$ and let $\ell$ be a prime number
different from the characteristic of $K$. Let $T_\ell$ and $V_\ell$ be the
corresponding Galois modules; we denote by $G_\ell$ the image of $\Gal(\sepcl
K/K)$ in $\Aut(T_\ell)$, and by $I_\ell$ the inertia subgroup of $G_\ell$. The
Lie algebras $\mathfrak{g}_\ell = \Lie(G_\ell)$, $\mathfrak{i}_\ell =
\Lie(I_\ell)$ are subalgebras of $\End(V_\ell)$ and we will determine them
under suitable assumptions on $K$ and $v$; note that, since $I_\ell$ is an
invariant subgroup of $G_\ell$, its Lie algebra $\mathfrak{i}_\ell$ is an
\emph{ideal} of $\mathfrak{g}_\ell$.
If $j = j(E)$ is the modular invariant of $E$ (cf.\ \ref{sec:IV_11}), we
consider the cases $v(j) < 0$ and $v(j) \ge 0$ separately.
\subsection{The case $v(j) < 0$}
\label{sec:IV_A1}
In this section we assume that the modular invariant $j$ of the elliptic curve
$E$ has a pole, i.e.\ that $v(j) < 0$.
\subsubsection{The elliptic curves of Tate}
\label{sec:IV_A11}
Let $q$ be an element of $K$ with $v(q) > 0$, and let $\Gamma_q$ be the
discrete subgroup of $K^\times$ generated by $q$. Then, by Tate's theory of
ultrametric theta functions (unpublished, but see Morikawa, \emph{Nagoya Math.
Journ.}, 1962),
\todo[bluetask]{Añadir referencia}
\dpage
there is an elliptic curve $E$ defined over $K$ with the property that, for any
finite extension $K'$ of $K$, the analytic group ${K'}^\times / \Gamma_q$ is
isomorphic to the group $E_q(K')$ of points of $E_q$ with values in $K'$. The
equation defining $E_q$ can he written in the form
\[
y^2 + xy = x^3 - b_2x - b_3,
\]
with
\[
b_2 = 5 \sum_{n\ge 1} n^3 \, \frac{q^n}{1-q^n}, \qquad \text{and}
\qquad b_3 = \sum_{n\ge 1} (7n^5 + 5n^3) \frac{q^n}{12(1-q^n)},
\]
these series converging in $K$. The modular invariant $j(q)$ of $E_q$ is
given by the usual formula
\[
j(q) = \frac{(1 + 48b_2)^3}{q \prod_{n\ge 1} (1 - q^n)^{24}}
= \frac{1}{q} + 744 + 196\,884 q + \cdots
\]
a series with integral coefficients. The function field of $E_q$ consists
of the fractions $F/G$, where $F$ and $G$ are Laurent series
\[
F = \sum_{n=-\infty}^{+\infty} a_n z^n, \qquad
G = \sum_{n=-\infty}^{+\infty} b_n z^n
\]
with coefficients in $K$, converging for all values of $z \ne 0, \infty$, and
such that $F(qz)/G(qz) = F(z)/G(z)$.
Since the modular invariant $j$ of the given elliptic curve $E$
is such that $v(j) < 0$, and since the series for $j(q)$ has integral
coefficients, one can choose $q$ so that $j = j(q)$. The elliptic curves
$E$ and $E_q$ become then isomorphic over a finite extension of $K$
(which can be taken to be of degree 2). Hence, after possibly
replacing $K$ by a finite extension, \emph{we may assume that $E = E_q$}.
\subsubsection{An exact sequence}
\label{sec:IV_A12}
We conserve the notation of \ref{sec:IV_A11}.
\dpage
Let $E_n$ be the kernel of multiplication by $\ell^n$ in $\sepcl
K^\times/\Gamma_q$. If $\mu_n$ is the group of $\ell^n$-th roots of unity in
$\sepcl K$, we have an injection $\mu_n \to E_n$. On the other hand, if $z \in
E_n$, we have $z^{\ell^n} \in \Gamma_q$, and hence there exists an integer $c$
such that $z^{\ell^n} = q^c$. If we associate to $z$ the image of $c$ in
$\Z/\ell^n\Z$, we obtain a homomorphism of $E_n$ into $\Z/\ell^n\Z$, and the
resulting sequence
\begin{equation}
\begin{tikzcd}
0 \rar & \mu_n \rar & E_n \rar & \Z/\ell^n\Z \rar & 0
\end{tikzcd}
\label{cd:IV_A2_1}
\end{equation}
is an exact sequence of $\Gal(\sepcl K/K)$-modules, $\Gal(\sepcl K/K)$ acting
trivially on $\Z/\ell^n\Z$. Passing to the limit, we obtain an exact sequence
of Galois modules
\begin{equation}
\begin{tikzcd}
0 \rar & T_\ell(\mu) \rar & T_\ell(E_n) \rar & \Z_\ell \rar & 0
\end{tikzcd}
\end{equation}
where $\Gal(\sepcl K/K)$ acts trivially on $\Z_\ell$. Tensoring with $\Q_\ell$, we
obtain the exact sequence
\begin{equation}
\begin{tikzcd}
0 \rar & V_\ell(\mu) \rar & V_\ell(E_n) \rar & \Q_\ell \rar & 0.
\end{tikzcd}
\label{cd:IV_A2_3}
\end{equation}
We now show that this sequence of $\Gal(\sepcl K/K)$-modules does not split. To
do this we introduce an invariant $x$ which belongs to the group $\invlim_n
H^1(G, \mu_n)$, where $G = \Gal(\sepcl K/K)$. Let $d$ be the coboundary
homomorphism:
\[
H^0(G, \Z/\ell^n\Z) \longrightarrow H^1(G, \mu_n)
\]
with respect to the exact sequence \eqref{cd:IV_A2_1} and let $x_n = d(1)$. The
invariant $x$ is the element of $\invlim_n H^1(G, \mu_n)$ defined by the family
$(x_n)_{n \ge 1}$.
\begin{prop}
\begin{enumerate}[(a)]
\item\label{prop:IV_A2a}
The isomorphism $\delta \colon K^\times/{K^\times}^{\ell^n} \to
H^1(G, \mu_n)$ of Kummer theory transforms the class of $q
\mod{ {K^\times}^{\ell^n} }$ into $x_n$.
\item\label{prop:IV_A2b}
The element $x$ is of infinite order.
\end{enumerate}
\end{prop}
(Recall that $\delta$ is induced by the coboundary map relative to
the exact sequence
\[\begin{tikzcd}
1 \rar & \mu_n \rar & {\algcl K}^\times \rar["()^{\ell^n}"] &
{\algcl K}^\times \rar & 1.)
\end{tikzcd}\]
\begin{proof}
Assertion \ref{prop:IV_A2a} is proved by an easy computation. To prove
\ref{prop:IV_A2b}, note that the valuation $v$ defines a homomorphism
\[
f_n \colon K^\times/{K^\times}^{\ell^n} \longrightarrow
\Z/\ell^n\Z,
\]
and hence a homomorphism
\[
f \colon \invlim_n K^\times/{K^\times}^{\ell^n} \longrightarrow
\Z_\ell.
\]
If we identify $x$ with the corresponding element of $\invlim
K^\times/{K^\times}^{\ell^n}$, as in \ref{prop:IV_A2a}, we have $f(x) =
v(q)$, hence $x$ is of infinite order.
\end{proof}
\begin{corp}
The sequence \eqref{cd:IV_A2_3} does not split.
\end{corp}
\begin{proof}
Assume it does, i.e.\ there is a $G$-subspace $X$ of $V_\ell(E_q)$
which is mapped isomorphically onto $\Q_\ell$. Let $X_T = T_\ell(E_q)
\cap X$. The image of $X_T$ in $\Z_\ell$ is $\ell^N\Z_\ell$, for some
$N \ge 0$. It is then easy to see that $\ell^N x = 0$, and this
contradicts the fact that $x$ is of
\dpage
infinite order.
\end{proof}
\subsubsection{Determination of \texorpdfstring{$\mathfrak{g}_\ell$}{gℓ} and
\texorpdfstring{$\mathfrak{i}_\ell$}{iℓ}}
\label{sec:IV_A13}
We keep the notation of \ref{sec:IV_A11} and \ref{sec:IV_A12}. If $X$ is a
one-dimensional subspace of $V_\ell = V_\ell(E)$, let $\mathfrak{r}_X$ denote
the subalgebra of $\End(V_\ell)$ consisting of those endomorphisms $u$ for
which $u(V_\ell) \subset X$, and let $\mathfrak{n}_X$ be the subalgebra of
$\mathfrak{r}_X$ formed by those $u \in \mathfrak{r}_X$ with $u(X) = 0$.
\begin{thm}
\begin{enumerate}[(a)]
\item\label{thm:IV_A13a}
If $k$ is algebraically closed and $\ell \ne p$, then there is a
one-dimensional subspace $X$ of $V_\ell$ such that
$\mathfrak{g}_\ell = \mathfrak{n}_X$.
\item\label{thm:IV_A13b}
If $k$ is algebraically closed and $\ell = p$, then there is a
one-dimensional subspace $X$ of $V_\ell$ such that
$\mathfrak{g}_\ell = \mathfrak{r}_X$.
\item If $k$ is finite, then $\mathfrak{g}_\ell = \mathfrak{r}_X$ for
some one-dimensional subspace $X$ of $V_\ell$, and
$\mathfrak{i}_\ell = \mathfrak{n}_X$ (resp.\ $\mathfrak{i}_\ell
= \mathfrak{r}_X$) if $\ell \ne p$ (resp.\ $\ell = p$).
\end{enumerate}
\end{thm}
\begin{proof}
Note first that, since $\mathfrak{g}_\ell$ and $\mathfrak{i}_\ell$ are invariant under finite
extension of $K$, we may assume that $E = E_q$.
\begin{enumerate}
\item In this case, $K$ contains the $\ell^n$-th roots of unity, hence
$\Gal(\sepcl K/K)$ acts trivially on $T_\ell(\mu)$.
Consequently, there is a basis $e_1, e_2$ of $T_\ell(E)$ such
that, for all $\sigma \in \Gal(\sepcl K/K)$, we have
$\sigma(e_1) = e_1$, $\sigma(e_2) = a(\sigma)e_1 + e_2$ with
$a(\sigma) \in \Z_\ell$. Moreover, the homomorphism $\sigma
\mapsto a(\sigma)$ cannot be trivial since the sequence
\eqref{cd:IV_A2_3} does not split. It follows that $\Img(a)$ is
an open subgroup of $\Z_\ell$, and hence that
$\mathfrak{g}_\ell = \mathfrak{n}_X$ with $X = V_\ell(\mu)$.
\item Since $\ell = p$, we must have $\char(K) = 0$ as $\ell \ne
\char(K)$. In this case, the action of $\Gal(\algcl K/K)$ on
$V_\ell(\mu)$ is by means of the character $\chi_\ell$ (cf.\
chap.~\ref{ch:i}, \ref{sec:I_12}) which is of infinite order.
It follows
\dpage
that $\mathfrak{g}_\ell = \mathfrak{r}_X$ where $X =
V_\ell(\mu)$; in fact, $\mathfrak{g}_\ell \supset
\mathfrak{n}_X$ since the sequence \eqref{cd:IV_A2_3} does not
split, and we cannot have $\mathfrak{g}_\ell = \mathfrak{n}_X$.
\item Since $k$ is finite, the action of $\Gal(\sepcl K/K)$ on
$T_\ell(\mu)$ is not trivial nor even of finite order. Hence,
the argument used in \ref{thm:IV_A13b} shows that
$\mathfrak{g}_\ell = \mathfrak{r}_X$. where $X = V_\ell(\mu)$.
Applying (a) to the completion of the maximal unramified
extension of $K$, we see that $\mathfrak{i}_\ell =
\mathfrak{n}_X$ if $\ell \ne p$, and that $\mathfrak{i}_\ell =
\mathfrak{r}_X$ if $\ell = p$. \qedhere
\end{enumerate}
\end{proof}
\subsubsection*{Exercise}
In case \ref{thm:IV_A13a}, show that $\Img(a) = \ell^n \Z_\ell$, where $\ell^n$
is the highest power of $\ell$ which divides $v(q) = -v(j)$.
\subsubsection{Application to isogenies}
\label{sec:IV_A14}
\todo[section]{Belén.}
\subsubsection{Existence of transvections in the inertia group}
\label{sec:IV_A15}
\todo[section]{Belén.}
\subsection{The case \texorpdfstring{$v(j) \ge 0$}{v(j) ≥ 0}}
\label{sec:IV_A2}
In this section we assume that the modular invariant $j$ of the elliptic curve
$E$ is integral, i.e.\ that $v(j) \ge 0$. Hence, after possibly replacing $K$
by a finite extension, we may assume that $E$ has \emph{good reduction} (cf.\
\ref{sec:IV_12}). We also assume that \emph{$K$ is of characteristic zero.}
\subsubsection{The case \texorpdfstring{$\ell \ne p$}{ℓ ̸/= p}}
\label{sec:IV_A21}
\todo[section]{Belén.}
\subsubsection{The case \texorpdfstring{$\ell = p$}{ℓ = p} with good
reduction of height 2}
\label{sec:IV_A22}
Here we assume that the reduced curve $\widetilde{E}$ is of height 2; recall
that, if $A$ is an abelian variety defined over a field of characteristic $p$,
its height can be defined as the integer $h$ for which $p$ is the inseparable
part of the degree of the homothety ``multiplication by $p$.'' An elliptic
curve is of height 2 if and only if its \emph{Hasse invariant} (cf.\
\citeauthor{9}~\cite{9}) is 0. Since $E$ has good reduction, it defines an
\emph{abelian scheme} $E(p)$ over $\mathcal{O}_{K'}$, hence a
\emph{$p$-divisible group} $E(p)$ over $\mathcal{O}_K$ (cf.\
\citeauthor{39}~\cite{39}, 2.1 -- see also \cite{26}, \S1, Ex.~2). The Tate
module of $E(p)$ can be identified with $T_p$. The connected component
$E(p)^\circ$ of $E(p)$ coincides with the \emph{formal group} (over
$\mathcal{O}_K$) attached to $E_v$; the height of $\widetilde{E}$ is precisely
the height of this formal group (in the usual sense). In our case, we have
$E(p) = E(p)^\circ$ since the height is assumed to be 2.
\begin{thm}
One has $\mathfrak{g}_p = \mathfrak{i}_p$. This Lie algebra is either
$\End(V_p)$ or a non-split Cartan subalgebra of $\End(V_p)$.
\end{thm}
(Recall that a non-split Cartan subalgebra of $\End(V_p)$ is a commutative
subalgebra of rank 2 with respect to which $V_p$ is irreducible. It is given by
a quadratic subfield of $\End(V_p)$.)
\begin{proof}
The Lie algebra $\mathfrak{g}_p$ has the property that $\mathfrak{g}_p
z = V_p$ for any non zero element $z$ of $V_p$ (cf.\ \cite[128]{27},
Prop.~8). In particular, $V_p$ is an irreducible
$\mathfrak{g}_p$-module; its commuting algebra is either a field of
degree 2 (which is then necessarily equal to $\mathfrak{g}_p$) or the
\dpage
field $\Q_p$, in which case $\mathfrak{g}_p$ is \emph{a priori} $\Sl_2$
or $\Gl_2$. But $\mathfrak{g}_p \ne \Sl_2$ since $\bigwedge^2 V_p$ is
canonically isomorphic to $V_p(\mu)$, and the action of $\Gal(\algcl
K/K)$ on $V_p(\mu)$ is by means of the character $\chi_p$, which is of
infinite order (indeed, no finite extension of $K$ can contain all
$p^n$-th roots of unity, $n = 1, 2, \dots$). Hence the Lie algebra
$\mathfrak{g}_p$ is either $\End(V_p)$ or a non split Cartan subalgebra
of $\End(V_p)$. Since the above applies to the completion of the
maximal unramified extension of $K$, we have the same alternative for
$\mathfrak{i}_p$. Moreover, $\mathfrak{i}_p$ is contained in
$\mathfrak{g}_p$. We have \emph{a priori} three possibilities:
\begin{enumerate}[(a)]
\item $\mathfrak{i}_p = \mathfrak{g}_p = \End(V_p)$.
\item $\mathfrak{i}_p = \mathfrak{g}_p$ is a non split Cartan
subalgebra of $\End(V_p)$.
\item $\mathfrak{i}_p$ is a Cartan subalgebra and
$\mathfrak{g}_p = \End(V_p)$.
\end{enumerate}
However, $\mathfrak{i}_p$ is an ideal of $\mathfrak{g}_p$. Hence, (c)
is impossible, and this proves the theorem.
\end{proof}
\begin{obs}
\begin{enumerate}
\item By a theorem of Tate (\cite{39}, \S4, cor.~1 to th.~4), the
algebra $\mathfrak{g}_p$ is a Cartan subalgebra of $\End(V_p)$
if and only if $E(p)$ has ``formal complex multiplication'',
i.e.\ if and only if the ring of endomorphisms of $E(p)$, over
a suitable extension of $K$, is of rank 2 over $\Z_p$. There
exist elliptic curves without complex multiplication (in the
algebraic sense) whose $p$-completion $E(p)$ have formal
complex multiplication.
\item Suppose that $\mathfrak{g}_p$ is a Cartan subalgebra of
$\End(V_p)$, and let $H = \mathfrak{g}_p \cap \Aut(V_p)$ be the
corresponding Cartan subgroup of $\Aut(V_p)$. If $N$ is the
normalizer of $H$ in $\Aut(V_p)$, then one knows that $N/H$ is
cyclic of order 2. Since $G_p \subset N$, it follows that $G_p$
is commutative if and only if $G_p \subset H$. The case $G_p
\subset H$ corresponds to the case where the formal complex
multiplication of $E(p)$ is
\dpage
defined over $K$, and the case $G_p \not\subset H$ corresponds to
the case where this formal multiplication is defined over a
quadratic extension of $K$.
\item Suppose that $G_p$ is commutative, and that the residue field $k$
is \emph{finite}. Let $F$ be the quadratic field of formal
complex multiplication (i.e.\ $\mathfrak{g}_p$ itself, viewed
as an associative subalgebra of $\End(V_p)$). If $U_F$ denotes
the group of units of $F$, the action of $\Gal(\algcl K/K)$ on
$V_p$ is given by a homomorphism
\[
\varphi_I \colon \Gal(\algcl K/K) \longrightarrow U_F.
\]
By local class field theory, we may identify the inertia group
of $\abcl{\Gal(\algcl K/K)}$ with the group $U_K$ of units of
$K$. Hence the restriction $\varphi_I$ of $\varphi$ to the
inertia group is a \emph{homomorphism of $U_K$ into $U_F$}. To
determine $\varphi_I$, we first remark that the action of
$\End(E(p))$ on the tangent space to $E(p)$ defines an
\emph{embedding} of $F$ into $K$. For that embedding, one has
(compare with chap.~\ref{ch:iii}, \ref{sec:III_A4})
\[
\varphi_I(x) = \Nm_{K/F}(x^{-1}), \qquad \text{for
all } x\in U_K.
\]
Indeed, by a result of Lubin (\emph{Ann. of Math.} 85, 1967),
\todo[bluetask]{Añadir referencia.}
there is a formal group $E'$ which is $K$-isogenous to $E(p)$,
and has for ring of endomorphisms the ring of integers of $F$.
But then, if $E''$ is a Lubin-Tate group over $K$ (cf.\
\citeauthor{17}~\cite{17}), the formal groups $E'$ and $E''$
are isomorphic over the completion of the maximal unramified
extension of $K$ (cf.\ \citeauthor{16}~\cite{16}, th.~4.3.2).
Hence to prove the formula (*),
\todo[bluetask]{¿A qué fórmula se refiere?}
we may assume that $E(p)$ is a
Lubin-Tate group, in which case the formula (*) follows from
the main result of \cite{17}.
\end{enumerate}
\end{obs}
\subsubsection{Auxiliary results on abelian varieties}
\label{sec:IV_A23}
Let $A$ and $B$ be two abelian varieties over $K$,
\dpage
with good reduction, so that the associated $p$-divisible groups $A(p)$ and
$B(p)$ are defined (these are $p$-divisible groups \emph{over the ring
$\mathcal{O}_{K'}$}, cf.\ \citeauthor{39}~\cite{39}). Let $\widetilde{A}$ and
$\widetilde{B}$ (resp.\ $\widetilde{A(p)}$ and $\widetilde{B(p)}$) be the
reductions of $A$ and $B$ (resp.\ of $A(p)$ and $B(p)$).
\begin{thm}\label{thm:IV_A23_1}
Let $\tilde f \colon \widetilde A \to \widetilde B$ be a morphism of
abelian varieties, and let $\widetilde{f(p)}$ be the corresponding
morphism of $\widetilde{A(p)}$ into $\widetilde{B(p)}$. Assume there
is a morphism $f(p) \colon A(p) \to B(p)$ whose reduction is
$\widetilde{f(p)}$. Then, there is a morphism $f \colon A \to B$ whose
reduction is $f$.
\end{thm}
A proof of this ``lifting'' theorem has been given by Tate in a Seminar (Woods
Hole, 1964), but has not yet been published;
\todo[bluetask]{Añadir referencia}
a different proof has been given by W. Messing (\emph{L. N.} 264, 1972).
\begin{thm}\label{thm:IV_A23_2}
Assume $T_p(A)$ is a direct sum of $\Z_p$-modules of rank 1 invariant
under the action of $\Gal(\algcl K/K)$. Then every endomorphism of
$\widetilde{A}$ lifts to an endomorphism of $A$, i.e., the reduction
homomorphism $\End(A) \to \End(\widetilde A)$ is surjective (and hence
bijective, since it is known to be injective).
\end{thm}
Using theorem~\ref{thm:IV_A23_1}, one sees that it is enough to show that any