global-analysis-and-the-banach-manifold-of-class-h1-curvers

Riemannian Geometry course project on the manifold H¹(I, M) of class H¹ curves on a Riemannian manifold M and its applications to the geodesics problem

NameSizeMode
..
sections/applications.tex 26240B -rw-r--r--
001
002
003
004
005
006
007
008
009
010
011
012
013
014
015
016
017
018
019
020
021
022
023
024
025
026
027
028
029
030
031
032
033
034
035
036
037
038
039
040
041
042
043
044
045
046
047
048
049
050
051
052
053
054
055
056
057
058
059
060
061
062
063
064
065
066
067
068
069
070
071
072
073
074
075
076
077
078
079
080
081
082
083
084
085
086
087
088
089
090
091
092
093
094
095
096
097
098
099
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
\section{Applications to the Calculus of Variations}\label{sec:aplications}

As promised, in this section we will apply our understanding of the structure
of \(H^1(I, M)\) to the calculus of variations, and in particular to the
geodesics problem. We also describe some further applications, such as the
Morse index theorem and the Jacobi-Darboux theorem. We start by defining\dots

\begin{definition}\label{def:variation}
  Given \(\gamma \in H^1(I, M)\), a variation \(\{ \gamma_t \}_t\) of
  \(\gamma\) is a smooth curve \(\gamma_\cdot : (-\epsilon, \epsilon) \to
  H^1(I, M)\) with \(\gamma_0 = \gamma\). We call the vector
  \(\left.\frac\dd\dt\right|_{t = 0} \gamma_t \in H^1(\gamma^* TM)\) \emph{the
  variational vector field of \(\{ \gamma_t \}_t\)}.
\end{definition}

We should note that the previous definition encompasses the classical
definition of a variation of a curve, as defined in \cite[ch.~5]{gorodski} for
instance: any piece-wise smooth function \(H : I \times (-\epsilon, \epsilon)
\to M\) determines a variation \(\{ \gamma_t \}_t\) given by \(\gamma_t(s) =
H(s, t)\). This is representative of the theory that lies ahead, in the sense
that most of the results we'll discuss in the following are minor refinements
of the classical theory. Instead, the value of the theory we will develop in
here lies in its conceptual simplicity: instead of relying in ad-hoc methods we
can now use the standard tools of calculus to study the critical points of the
energy functional \(E\).

What we mean by this last statement is that by look at the energy functional as
a smooth function \(E \in C^\infty(H^1(I, M))\) we can study its classical
``critical points'' -- i.e. curves \(\gamma\) with a variation \(\{ \gamma_t
\}_t\) such that \(\left.\frac\dd\dt\right|_{t = 0} E(\gamma_t) = 0\) -- by
looking at its derivative. The first variation of energy thus becomes a
particular case of a formula for \(d E\), and the second variation of energy
becomes a particular case of a formula for the Hessian of \(E\) at a critical
point. Without further ado, we prove\dots

\begin{theorem}\label{thm:energy-is-smooth}
  The energy functional
  \begin{align*}
    E : H^1(I, M) & \to \mathbb{R} \\
    \gamma
    & \mapsto \frac{1}{2} \norm{\partial \gamma}_0^2
    = \frac{1}{2} \int_0^1 \norm{\dot\gamma(t)}^2 \; \dt
  \end{align*}
  is smooth and \(d E_\gamma X = \left\langle \partial \gamma, \frac\nabla\dt X
  \right\rangle_0\).
\end{theorem}

\begin{proof}
  The fact that \(E\) is smooth should be clear from the smoothness of
  \(\partial\) and \(\norm\cdot_0\). Furthermore, from the definition of
  \(\frac\nabla\dt\) we have
  \[
    \begin{split}
      \left\langle \partial \gamma, \frac\nabla\dt X \right\rangle_0
      & =
        \left\langle
        \partial \gamma, (\nabla_X^0 \partial)_\gamma
        \right\rangle_0 \\
      & = (\tilde X \langle \partial, \partial \rangle_0)(\gamma) -
        \left\langle
        (\nabla_X^0 \partial)_\gamma, \partial \gamma
        \right\rangle_0 \\
      & = 2 \tilde X E(\gamma) -
        \left\langle \frac\nabla\dt X, \partial \gamma \right\rangle_0 \\
      & = 2 d E_\gamma X -
        \left\langle \partial \gamma, \frac\nabla\dt X \right\rangle_0
    \end{split}
  \]
  where \(\tilde X \in \mathfrak{X}(H^1(I, M))\) is any vector field with
  \(\tilde X_\gamma = X\).
\end{proof}

As promised, by applying the chain rule and using the compatibility of
\(\nabla\) with the metric we arrive at the classical formula for the first
variation of energy \(E\).

\begin{corollary}
  Given a piece-wise smooth curve \(\gamma : I \to M\) with
  \(\gamma\!\restriction_{[t_i, t_{i + 1}]}\) smooth and a variation \(\{
    \gamma_t \}_t\) of \(\gamma\) with variational vector field \(X\) we have
  \[
    \left.\frac\dd\dt\right|_{t = 0} E(\gamma_t)
    = \sum_i
      \left. \langle \dot\gamma(t), X_t \rangle\right|_{t = t_i}^{t_{i + 1}}
    - \int_0^1 \left\langle \frac\nabla\dt \dot\gamma(t), X_t \right\rangle
      \; \dt
  \]
\end{corollary}

Another interesting consequence of theorem~\ref{thm:energy-is-smooth} is\dots

\begin{corollary}
  The only critical points of \(E\) in \(H^1(I, M)\) are the constant curves.
\end{corollary}

\begin{proof}
  Clearly every constant curve is a critical point. On the other hand, if
  \(\gamma \in H^1(I, M)\) is such that \(\left\langle \partial \gamma,
  \frac\nabla\dt X \right\rangle_0 = d E_\gamma X = 0\) for all \(X \in
  H^1(\gamma^* TM)\) then \(\partial \gamma = 0\) and therefore \(\gamma\) is
  constant.
\end{proof}

Another way to put is to say that the problem of characterizing the critical
points of \(E\) in \(H^1(I, M)\) is not interesting at all. This shouldn't
really come as a surprise, as most interesting results from the classical
theory are concerned with particular classes of variations of a curves, such as
variations with fixed endpoints or variations through loops. In the next
section we introduce two submanifolds of \(H^1(I, M)\), corresponding to the
classes of variations previously described, and classify the critical points of
the restrictions of \(E\) to such submanifolds.

\subsection{The Critical Points of \(E\)}

We begin with a technical lemma.

\begin{lemma}
  The maps \(\sigma, \tau: H^1(I, M) \to M\) with \(\sigma(\gamma) =
  \gamma(0)\) and \(\tau(\gamma) = \gamma(1)\) are submersions.
\end{lemma}

\begin{proof}
  To see that \(\sigma\) and \(\tau\) are smooth it suffices to observe that
  their local representation in \(U_\gamma\) for \(\gamma \in {C'}^\infty(I,
  M)\) is given by the maps
  \begin{align*}
    U \subset H^1(W_\gamma) & \to     T_{\gamma(0)} M &
    U \subset H^1(W_\gamma) & \to     T_{\gamma(1)} M \\
                          X & \mapsto X_0 &
                          X & \mapsto X_1
  \end{align*}
  which are indeed smooth functions. This local representation also shows that
  \begin{align*}
    d\sigma_\gamma : H^1(\gamma^* TM) & \to     T_{\gamma(0)} M &
    d\tau_\gamma   : H^1(\gamma^* TM) & \to     T_{\gamma(1)} M \\
                                    X & \mapsto X_0 &
                                    X & \mapsto X_1
  \end{align*}
  are surjective maps for all \(\gamma \in H^1(I, M)\).
\end{proof}

We can now show\dots

\begin{theorem}
  The subspace \(\Omega_{p q} M \subset H^1(I, M)\) of curves joining \(p, q
  \in M\) is a submanifold whose tangent space \(T_\gamma \Omega_{p q} M\) is
  the subspace of \(H^1(\gamma^* TM)\) consisting of class \(H^1\) vector
  fields \(X\) along \(\gamma\) with \(X_0 = X_1 = 0\). Likewise, the space
  \(\Lambda M \subset H^1(I, M)\) of free loops is a submanifold whose tangent
  at \(\gamma\) is given by all \(X \in H^1(\gamma^* TM)\) with \(X_0 = X_1\).
\end{theorem}

\begin{proof}
  To see that these are submanifolds, it suffices to note that \(\Omega_{p q}
  M\) and \(\Lambda M\) are the inverse images of the closed submanifolds
  \(\{(p, q)\}, \{(p, p) : p \in M\} \subset M \times M\) under the submersion
  \((\sigma, \tau) : H^1(I, M) \to M \times M\).

  The characterization of their tangent bundles should also be clear: any curve
  \((-\epsilon, \epsilon) \to H^1(I, M)\) passing through \(\gamma \in
  \Omega_{p q} M\) whose image is contained in \(\Omega_{p q} M\) is a
  variation of \(\gamma\) with fixed endpoints, so its variational vector field
  \(X\) satisfies \(X_0 = X_1 = 0\). Likewise, any variation of a loop \(\gamma
  \in \Lambda M\) trough loops -- i.e. a curve \((-\epsilon, \epsilon) \to
  \Lambda M\) passing through \(\gamma\) -- satisfies \(X_0 = X_1\).
\end{proof}

Finally, as promised we will provide a characterization of the critical points
of \(E\!\restriction_{\Omega_{p q} M}\) and \(E\!\restriction_{\Lambda M}\).

\begin{theorem}\label{thm:critical-points-char-in-submanifolds}
  The critical points of \(E\!\restriction_{\Omega_{p q} M}\) are precisely the
  geodesics of \(M\) joining \(p\) and \(q\). The critical points of
  \(E\!\restriction_{\Lambda M}\) are the closed geodesics of \(M\) --
  including the constant maps.
\end{theorem}

\begin{proof}
  We start by supposing that \(\gamma\) is a geodesic. Since \(\gamma\) is
  smooth,
  \[
    d E_\gamma X
    = \int_0^1
      \left\langle \dot\gamma(t), \frac\nabla\dt X \right\rangle \; \dt
    = \int_0^1
      \frac\dd\dt \langle \dot\gamma(t), X_t \rangle -
      \left\langle \frac\nabla\dt \dot\gamma(t), X \right\rangle \; \dt
    = \langle \dot\gamma(1), X_1 \rangle - \langle \dot\gamma(0), X_0 \rangle
  \]

  Now if \(\gamma \in \Omega_{p q} M\) and \(X \in T_\gamma \Omega_{p q} M\)
  then \(d E_\gamma X = \langle \dot\gamma(1), 0 \rangle - \langle
  \dot\gamma(0), 0 \rangle = 0\). Likewise, if \(\gamma\) is a closed geodesic
  and \(X \in T_\gamma \Lambda M\) we find \(d E_\gamma X = 0\) since
  \(\dot\gamma(0) = \dot\gamma(1)\) and \(X_0 = X_1\). This establishes that
  the geodesics are indeed critical points of the restrictions of \(E\).

  Suppose \(\gamma \in \Omega_{p q} M\) is a critical point and let \(Y, Z \in
  H^1(\gamma^* TM)\) be such that
  \begin{align*}
    \frac\nabla\dt Y & = \partial \gamma &
                 Y_0 & = 0 &
    \frac\nabla\dt Z & = 0 &
                 Z_1 & = Y_1
  \end{align*}

  Let \(X_t = Y_t - t Z_t\). Then \(X_0 = X_1 = 0\) and \(\frac\nabla\dt X =
  \partial \gamma - Z\). Furthermore,
  \[
    \langle Z, \partial \gamma - Z \rangle_0
    = \left\langle Z, \frac\nabla\dt X \right\rangle_0
    = \int_0^1 \frac\dd\dt \langle Z_t, X_t \rangle \; \dt
    = \langle Z_1, X_1 \rangle - \langle Z_0, X_0 \rangle
    = 0
  \]
  and
  \[
    \langle \partial \gamma, \partial \gamma - Z \rangle_0
    = \left\langle \partial \gamma, \frac\nabla\dt X \right\rangle_0
    = d E_\gamma X
    = 0,
  \]
  which implies \(\norm{\partial \gamma - Z}_0^2 = 0\). In other words,
  \(\partial \gamma = Z \in H^1(\gamma^* TM)\) and therefore \(\frac\nabla\dt
  \dot\gamma(t) = \frac\nabla\dt Z = 0\) -- i.e. \(\gamma\) is a geodesic.

  Finally, if \(\gamma \in \Lambda M\) with \(\gamma(0) = \gamma(1) = p\) we
  may apply the argument above to conclude that \(\gamma\) is a geodesic
  joining \(p\) to \(q = p\). To see that \(\gamma\) is a closed geodesic apply
  the same argument again for \(\eta(t) = \gamma(1 + \sfrac{1}{2})\) to
  conclude that \(\dot\gamma(0) = \dot\eta(\sfrac{1}{2}) = \dot\gamma(1)\).
\end{proof}

We should point out that the first part of
theorem~\ref{thm:critical-points-char-in-submanifolds} is a particular case of
a result regarding critical points of the restriction of \(E\) to the
submanifold \(H_{N_0, N_1}^1(I, M) \subset H^1(I, M)\) of curves joining
submanifolds \(N_0, N_1 \subset M\): the critical points of
\(E\!\restriction_{H_{N_0, N_1}^1(I, M)}\) are the geodesics \(\gamma\) joining
\(N_0\) to \(N_1\) with \(\dot\gamma(0) \in T_{\gamma(0)} N_0^\perp\) and
\(\dot\gamma(1) \in T_{\gamma(1)} N_1^\perp\). The proof of this result is
essentially the same as that of
theorem~\ref{thm:critical-points-char-in-submanifolds}, given that \(T_\gamma
H_{N_0, N_1}^1(I, M)\) is subspace of \(H^1\) vector fields \(X\) along
\(\gamma\) with \(X_0 \in T_{\gamma(0)} N_0\) and \(X_1 \in T_{\gamma(1)}
N_1\).

\subsection{Second Order Derivatives of \(E\)}

Having establish a clear connection between geodesics and critical points of
\(E\), the only thing we're missing to complete our goal of providing a modern
account of the classical theory is a refurnishing of the formula for second
variation of energy. Intuitively speaking, the second variation of energy
should be a particular case of a formula for the second derivative of \(E\).
The issue we face is, of course, that in general there is no such thing as
``the second derivative'' of a smooth function between manifolds.

Nevertheless, the metric of \(H^1(I, M)\) allow us to discuss ``the second
derivative'' of \(E\) in a meaningful sense by looking at the Hessian form,
which we define in the following.

\begin{definition}
  Given a -- possibly infinite-dimensional -- Riemannian manifold \(N\) and a
  smooth functional \(f : N \to \mathbb{R}\), we call the symmetric tensor
  \[
    d^2 f(X, Y) = \nabla d f (X, Y) =  X Y f - df \nabla_X Y
  \]
  \emph{the Hessian of \(f\)}.
\end{definition}

We can now apply the classical formula for the second variation of energy to
compute the Hessian of \(E\) at a critical point.

\begin{theorem}
  If \(\gamma\) is a critical point of \(E\!\restriction_{\Omega_{p q} M}\)
  then
  \begin{equation}\label{eq:second-variation-general}
    (d^2 E\!\restriction_{\Omega_{p q} M})_\gamma(X, Y)
    = \left\langle \frac\nabla\dt X, \frac\nabla\dt Y \right\rangle_0
    - \langle R_\gamma X, Y \rangle_0,
  \end{equation}
  where \(R_\gamma : H^1(\gamma^* TM) \to H^1(\gamma^* TM)\) is given by
  \((R_\gamma X)_t = R(X_t, \dot\gamma(t)) \dot\gamma(t)\). Formula
  (\ref{eq:second-variation-general}) also holds for critical points of \(E\)
  in \(\Lambda M\).
\end{theorem}

\begin{proof}
  Given the symmetry of \(d^2 E\), it suffices to take \(X \in T_\gamma
  \Omega_{p q} M\) and show
  \[
    d^2 E_\gamma(X, X)
    = \norm{\frac\nabla\dt X}_0^2 - \langle R_\gamma X, X \rangle_0
  \]

  To that end, we fix a variation \(\{ \gamma_t \}_t\) of \(\gamma\) with fixed
  endpoints and variational field \(X\) and compute
  \[
    \begin{split}
      d^2 E_\gamma(X, X)
      & = \left.\frac{\dd^2}{\dt^2}\right|_{t = 0} E(\gamma_t) \\
      \text{(second variation of energy)}
      & = \int_0^1 \norm{\frac\nabla\dt X}^2
        - \langle R(X_t, \dot\gamma(t)) \dot\gamma(t), X_t \rangle \; \dt \\
      & = \norm{\frac\nabla\dt X}_0^2 - \langle R_\gamma X, X \rangle_0
    \end{split}
  \]
\end{proof}

Next we discuss some further applications of the theory we've developed so far.
In particular, we will work towards Morse's index theorem and and describe how
one can apply it to establish the Jacobi-Darboux theorem. We begin with a
technical lemma, whose proof amounts to an uninspiring exercise in analysis --
see lemma 2.4.6 of \cite{klingenberg}.

\begin{lemma}\label{thm:inclusion-submnfds-is-compact}
  Let \(\Omega_{p q}^0 M \subset C^0(I, M)\) be the space of continuous curves
  joining \(p\) to \(q\). Then the inclusion \(\Omega_{p q} M
  \longhookrightarrow \Omega_{p q}^0 M\) is continuous and compact. Likewise,
  if \(M\) is compact and \(\Lambda^0 M \subset C^0(I, M)\) is the space of
  continuous free loops then the inclusion \(\Lambda M \longhookrightarrow
  \Lambda^0 M\) is continuous and compact.
\end{lemma}

As a first consequence, we prove\dots

\begin{proposition}\label{thm:morse-index-e-is-finite}
  Given a critical point \(\gamma\) of \(E\) in \(\Omega_{p q} M\), the
  self-adjoint operator \(A_\gamma : T_\gamma \Omega_{p q} M \to T_\gamma
  \Omega_{p q} M\) given by
  \[
    \langle A_\gamma X, Y \rangle_1
    = \langle X, A_\gamma Y \rangle_1
    = d^2 E_\gamma(X, Y)
  \]
  has the form \(A_\gamma = \operatorname{Id} + K_\gamma\) where \(K_\gamma :
  T_\gamma \Omega_{p q} M \to T_\gamma \Omega_{p q} M\) is a compact operator.
  The same holds for \(\Lambda M\) if \(M\) is compact.
\end{proposition}

\begin{proof}
  Consider \(K_\gamma = - \left( \operatorname{Id} - \frac{\nabla^2}{\dt^2}
  \right)^{-1} \circ (\operatorname{Id} + R_\gamma)\). We will show that
  \(K_\gamma\) is compact and that \(A_\gamma = \operatorname{Id} + K_\gamma\)
  for \(\gamma\) in both \(\Omega_{p q} M\) and \(\Lambda M\) -- in which case
  assume \(M\) is compact.

  Let \(\gamma \in \Omega_{p q} M\) be a critical point. By
  theorem~\ref{thm:critical-points-char-in-submanifolds} we know that
  \(\gamma\) is a geodesic. Let \(X, Y \in \Gamma(\gamma^* TM)\) with \(X_0 =
  X_1 = Y_0 = Y_1 = 0\). Then
  \begin{equation}\label{eq:compact-partial-result}
    \begin{split}
      \langle X, Y \rangle_1
      & = \langle X, Y \rangle_0
        + \left\langle \frac\nabla\dt X, \frac\nabla\dt Y \right\rangle \\
      & = \langle X, Y \rangle_0
        + \int_0^1
          \left\langle \frac\nabla\dt X, \frac\nabla\dt Y \right\rangle
          \; \dt \\
      & = \langle X, Y \rangle_0
        + \int_0^1
          \frac\dd\dt \langle X_t, Y_t \rangle -
          \left\langle \frac{\nabla^2}{\dt^2} X, Y \right\rangle \; \dt \\
      & = \langle X, Y \rangle_0
        - \left\langle \frac{\nabla^2}{\dt^2} X, Y \right\rangle_0
        + \left.\langle X_t, Y_t \rangle\right|_{t = 0}^1 \\
      & = \left\langle
          \left(\operatorname{Id} - \frac{\nabla^2}{\dt^2}\right) X, Y
          \right\rangle_0
    \end{split}
  \end{equation}

  Since \(\Gamma(\gamma^* TM) \subset H^1(\gamma^* TM)\) is dense,
  (\ref{eq:compact-partial-result}) extends to all of \(T_\gamma \Omega_{p q}
  M\). Hence given \(X, Y \in T_\gamma \Omega_{p q} M\) we have
  \[
    \begin{split}
      \langle A_\gamma X, Y \rangle_1
      & = \left\langle \frac\nabla\dt X, \frac\nabla\dt Y \right\rangle_0
        - \langle R_\gamma X, Y \rangle_0 \\
      & = \langle X, Y \rangle_1 - \langle X, Y \rangle_0
        - \langle R_\gamma X, Y \rangle_0 \\
      & = \langle X, Y \rangle_1
        - \langle (\operatorname{Id} + R_\gamma) X, Y \rangle_0 \\
      & = \langle X, Y \rangle_1
        - \left\langle
          \left( \operatorname{Id} - \frac{\nabla^2}{\dt^2} \right)^{-1}
          \circ (\operatorname{Id} + R_\gamma) X, Y
          \right\rangle_1 \\
      & = \langle X, Y \rangle_1 + \langle K_\gamma X, Y \rangle_1 \\
    \end{split}
  \]

  Now consider a critical point \(\gamma \in \Lambda M\) -- i.e. a closed
  geodesic. Equation (\ref{eq:compact-partial-result}) also holds for \(X, Y
  \in \Gamma(\gamma^* TM)\) with \(X_0 = X_1\) and \(Y_0 = Y_1\), so it holds
  for all \(X, Y \in T_\gamma \Lambda M\). Hence by applying the same argument
  we get \(\langle A_\gamma X, Y \rangle_1 = \langle (\operatorname{Id} +
  K_\gamma) X, Y \rangle_1\).

  As for the compactness of \(K_\gamma\) in the case of \(\Omega_{p q} M\),
  from (\ref{eq:compact-partial-result}) we get \(\norm{K_\gamma X}_1^2 = -
  \langle (\operatorname{Id} + R_\gamma) X, K_\gamma X \rangle_0\), so that
  proposition~\ref{thm:continuous-inclusions-sections} implies
  \begin{equation}\label{eq:compact-operator-quota}
    \norm{K_\gamma X}_1^2
    \le \norm{\operatorname{Id} + R_\gamma}
      \cdot \norm{K_\gamma X}_\infty \cdot \norm{X}_0
    \le \sqrt{2} \norm{\operatorname{Id} + R_\gamma} \cdot \norm{K_\gamma X}_1
      \cdot \norm{X}_0
  \end{equation}

  Given a bounded sequence \((X_n)_n \subset T_\gamma \Omega_{p q} M\), it
  follow from lemma~\ref{thm:inclusion-submnfds-is-compact} that \((X_n)_n\) is
  relatively compact as a \(C^0\)-sequence. From
  (\ref{eq:compact-operator-quota}) we then get that \((K_\gamma X_n)_n\) is
  relatively compact as an \(H^1\)-sequence, as desired. The same argument
  holds for \(\Lambda M\) if \(M\) is compact -- so that we can once more apply
  lemma~\ref{thm:inclusion-submnfds-is-compact}.
\end{proof}

Once again, the first part of this proposition is a particular case of a
broader result regarding the space of curves joining submanifolds of \(M\): if
\(N \subset M\) is a totally geodesic manifold of codimension \(1\) and
\(\gamma \in H_{N, \{q\}}^1(I, M)\) is a critical point of the restriction of
\(E\) then \(A_\gamma = \operatorname{Id} + K_\gamma\). These results aren't
that appealing on their own, but they allow us to establish the following
result, which is essential for stating Morse's index theorem.

\begin{corollary}
  Given a critical point \(\gamma\) of \(E\!\restriction_{\Omega_{p q} M}\),
  there is an orthogonal decomposition
  \[
    T_\gamma \Omega_{p q} M
    = T_\gamma^- \Omega_{p q} M
    \oplus T_\gamma^0 \Omega_{p q} M
    \oplus T_\gamma^+ \Omega_{p q} M,
  \]
  where \(T_\gamma^- \Omega_{p q} M\) is the finite-dimensional subspace
  spanned by eigenvectors with negative eigenvalues, \(T_\gamma^0 \Omega_{p q}
  M = \ker A_\gamma\) and \(T_\gamma^+ \Omega_{p q} M\) is the proper Hilbert
  subspace given by the closure of the subspace spanned by eigenvectors with
  positive eigenvalues. The same holds for critical points \(\gamma\) of
  \(E\!\restriction_{\Lambda M}\) and \(T_\gamma \Lambda M\) if \(M\) is
  compact.
\end{corollary}

\begin{definition}\label{def:morse-index}
  Given a critical point \(\gamma\) of \(E\!\restriction_{\Omega_{p q} M}\) we
  call the number \(\dim T_\gamma^- \Omega_{p q} M\) \emph{the \(\Omega\)-index
  of \(\gamma\)}. Likewise, we call \(\dim T_\gamma^- \Lambda M\) for a
  critical point \(\gamma\) of \(E\!\restriction_{\Lambda M}\) \emph{the
  \(\Lambda\)-index of \(\gamma\)}. Whenever the submanifold \(\gamma\) lies in
  is clear from context we refer to the \(\Omega\)-index or the
  \(\Lambda\)-index of \(\gamma\) simply by \emph{the index of \(\gamma\)}.
\end{definition}

This definition highlights one of the greatest strengths of our approach: while
the index of a geodesic \(\gamma\) can be defined without the aid of the tools
developed in here, by using of the Hessian form \(d^2 E_\gamma\) we can place
definition~\ref{def:morse-index} in the broader context of Morse theory. In
fact, the geodesics problems and the energy functional where among Morse's
original proposed applications. Proposition~\ref{thm:morse-index-e-is-finite}
and definition~\ref{def:morse-index} amount to a proof that the Morse index of
\(E\) at a critical point \(\gamma\) is finite.

We are now ready to state Morse's index theorem.

\begin{theorem}[Morse]
  Let \(\gamma \in \Omega_{p q} M\) be a critical point of \(E\). Then the
  index of \(\gamma\) is given of the sum of the multiplicities of the proper
  conjugate points\footnote{By ``conjugate points of a geodesic $\gamma$'' we
  of course mean points conjugate to $\gamma(0) = p$ along $\gamma$.} of
  \(\gamma\) in the interior of \(I\).
\end{theorem}

Unfortunately we do not have the space to include the proof of Morse's theorem
in here, but see theorem 2.5.9 of \cite{klingenberg}. The index theorem can be
generalized for \(H_{N, \{q\}}^1(I, M)\) by replacing the notion of conjugate
point with the notion of focal points of \(N\) -- see theorem 7.5.4 of
\cite{gorodski} for the classical approach. What we are really interested in,
however, is the following consequence of Morse's theorem.

\begin{theorem}[Jacobi-Darboux]\label{thm:jacobi-darboux}
  Let \(\gamma \in \Omega_{p q} M\) be a critical point of \(E\).
  \begin{enumerate}
    \item If there are no conjugate points of \(\gamma\) then there exists a
      neighborhood \(U \subset \Omega_{p q} M\) of \(\gamma\) such that
      \(E(\eta) > E(\gamma)\) for all \(\eta \in U\) with \(\eta \ne \gamma\).

    \item Let \(k > 0\) be the sum of the multiplicities of the conjugate
      points of \(\gamma\) in the interior of \(I\). Then there exists an
      immersion
      \[
        i : B^k \to \Omega_{p q} M
      \]
      of the unit ball \(B^k = \{v \in \mathbb{R}^k : \norm{v} < 1\}\) with
      \(i(0) = \gamma\), \(E(i(v)) < E(\gamma)\) and \(L(i(v)) < L(\gamma)\)
      for all nonzero \(v \in B^k\).
  \end{enumerate}
\end{theorem}

\begin{proof}
  First of all notice that given \(\eta \in U_\gamma\) with \(\eta =
  \exp_\gamma(X)\), \(X \in H^1(W_\gamma)\), the Taylor series for \(E(\eta)\)
  is given by \(E(\eta) = E(\gamma) + \frac{1}{2} d^2 E_\gamma(X, X) +
  \cdots\). More precisely,
  \begin{equation}\label{eq:energy-taylor-series}
    \frac
      {\abs{E(\exp_\gamma(X)) - E(\gamma) - \frac{1}{2} d^2 E_\gamma(X, X)}}
      {\norm{X}_1^2}
    \to 0
  \end{equation}
  as \(X \to 0\).

  Let \(\gamma\) be as in \textbf{(i)}. Since \(\gamma\) has no conjugate
  points, it follows from Morse's index theorem that \(T_\gamma^- \Omega_{p q}
  M = 0\). Furthermore, by noticing that any piece-wise smooth \(X \in \ker
  A_\gamma\) is Jacobi field vanishing at \(p\) and \(q\) one can also show
  \(T_\gamma^0 \Omega_{p q} M = 0\). Hence \(T_\gamma \Omega_{p q} M =
  T_\gamma^+ \Omega_{p q} M\) and therefore \(d^2 E\!\restriction_{\Omega_{p q}
  M}\) is positive-definite. Now given \(\eta = \exp_\gamma(X)\) as before,
  (\ref{eq:energy-taylor-series}) implies that \(E(\eta) > E(\gamma)\),
  provided \(\norm{X}_1\) is sufficiently small.

  As for part \textbf{(ii)}, fix an orthonormal basis
  \(\{X_j : 1 \le j \le k\}\) of \(T_\gamma^- \Omega_{p q} M\) consisting of
  eigenvectors of \(A_\gamma\) with negative eigenvalues \(- \lambda_i\). Let
  \(\delta > 0\) and define
  \begin{align*}
    i : B^k & \to \Omega_{p q} M \\
    v & \mapsto \exp_\gamma(\delta (v_1 \cdot X_1 + \cdots + v_k \cdot X_k))
  \end{align*}

  Clearly \(i\) is an immersion for small enough \(\delta\). Moreover, from
  (\ref{eq:energy-taylor-series}) and
  \[
    E(i(v))
    = E(\gamma) - \frac{1}{2} \delta^2 \sum_j \lambda_j \cdot v_j + \cdots
  \]
  we find that \(E(i(v)) < E(\gamma)\) for sufficiently small \(\delta\). In
  particular, \(L(i(v))^2 \le E(i(v)) < E(\gamma) = L(\gamma)^2\).
\end{proof}

We should point out that part \textbf{(i)} of theorem~\ref{thm:jacobi-darboux}
is weaker than the classical formulation of the Jacobi-Darboux theorem -- such
as in theorem 5.5.3 of \cite{gorodski} for example -- in two aspects. First, we
do not compare the length of curves \(\gamma\) and \(\eta \in U\). This could
be amended by showing that the length functional \(L : H^1(I, M) \to
\mathbb{R}\) is smooth and that its Hessian \(d^2 L_\gamma\) is given by \(C
\cdot d^2 E_\gamma\) for some \(C > 0\). Secondly, unlike the classical
formulation we only consider curves in an \(H^1\)-neighborhood of \(\gamma\) --
instead of a neighborhood of \(\gamma\) in \(\Omega_{p q} M\) in the uniform
topology. On the other hand, part \textbf{(ii)} is definitively an improvement
of the classical formulation: we can find curves \(\eta = i(v)\) shorter than
\(\gamma\) already in an \(H^1\)-neighborhood of \(\gamma\).

This concludes our discussion of the applications of our theory to the
geodesics problem. We hope that these short notes could provide the reader with
a glimpse of the rich theory of the calculus of variations and global analysis
at large. We once again refer the reader to \cite[ch.~2]{klingenberg},
\cite[ch.~11]{palais} and \cite[sec.~6]{eells} for further insight on modern
variational methods.