-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathAA.tex
2294 lines (1423 loc) · 72.7 KB
/
AA.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
\subsection*{Completion by one point}
\begin{pr}
Let $S$ be a surface equipped with a Riemannian metric of curvature $\ge 1$;
denote by $\bar S$ the completion of $S$.
Suppose that $\bar S\setminus S$ is a one-point set.
Is it true that $\diam S\le \pi$?
\end{pr}
This problem was mentioned by Changfeng Gui and Qinfeng Li
\cite[right after Theorem 1.4]{gui-li}.
It is straightforward to show that all points in $\bar S$ lie on the distance at most $\pi$ from the only point in $p=\bar S\setminus S$.
In particular, $\diam S\le 2\cdot\pi$.
Note that Alexandrov comparison around $\bar S$ might fail;
an example can be found among surfaces with rotational symmetry.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{Curvature hollow}
\label{Curvature hollow}
\begin{pr}
Construct a Riemannian metric on $\RR^3$
which is Euclidean outside of an open bounded set $\Omega$
and with negative scalar curvature in $\Omega$.
\end{pr}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\parbf{Curvature hollow.}
An example can be found among the metrics with isometric $\mathbb{S}^1$ action which fix a line.
\medskip
First note that it is easy to construct such metric on the connected sum
$\RR^3\#\mathbb{S}^2\times\mathbb{S}^1$.
Indded make two holes in $\RR^3$, deform the obtained space conformally so that scalar curvature drops and the two holes fit and glue the holes to each other.
Further, note that such metric can be constructed in such a way that it has a closed geodesic $\gamma$ with trivial holonomy and with constant negative curvature in its a tubular neighborhood.
Cut tubular neighborhood $V\simeq\mathbb{D}^2\times \mathbb{S}^1$ of $\gamma$
and glue in the product $W\simeq\mathbb{S}^1\times \mathbb{D}^2$ with the swapped factors.
Note that after this surgery we get back $\RR^3$.
It remains to construct a metric $g$ on $W$ with negative scalar curvature which
is identical to the original metric on $V$ near its boundary.
The needed space $(W,g)$ can be found among wrap products $\mathbb{S}^1\times_f \mathbb{D}^2$ [see page~\pageref{page:warped product}].
\qeds
This construction was given by Joachim Lohkamp \cite[see][]{lohkamp},
he describes there yet an other equally simple construction.
In fact,
his constructions produce
$\mathbb{S}^1$-invariant hollows
with negative Ricci curvature.
On the other hand,
there are no hollows with
positive scalar curvature and negative sectional curvature.
The former is equivalent to the positive mass conjecture \cite[see][and the references therein]{witten}
and the latter is an easy exercise.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{Almost flat manifold\easy}
\label{almost-flat}
\emph{Nil-manifolds} form the minimal class of manifolds which includes a point, and has the following property:
the total space of any principle $\mathbb{S}^1$-bundle over a nil-manifold is a nil-manifold.
The nil-manifolds can be also defined as the quotients of a connected nilpotent Lie group by a lattice.
A compact Riemannian manifold $M$ is called $\eps$-flat if its sectional curvature at all points in all directions lie in the interval $[-\eps,\eps]$.
The main theorem of Gromov in \cite{gromov-almost-flat},
states that for any positive integer $n$ there is $\eps>0$ such that any $\eps$-flat compact $n$-dimensional manifold with diameter at most $1$ admits a finite cover by a nil-manifold.
A more detailed proof can be found in \cite{buser-karcher}
and a more precise statement can be found in \cite{ruh}.
\begin{pr}
Given $\eps>0$, construct a compact Riemannian manifold $M$ of sufficiently large dimension which admits a Riemannian metric with diameter $\le 1$ and sectional
curvature $|K|<\eps$,
but does not admit a finite covering by a nil-manifold.
\end{pr}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\parbf{Almost flat manifold.}
An example can be found among solve manifolds;
that is, quotients of solvable lie group by a lattice.
In fact, torus bundles over circle circles are sufficient.
\medskip
A torus bundle $\TT^m\to E\to\mathbb{S}^1$ is obtained by taking $\TT^m\times [0,1]$ and gluing $\TT^m\times 0$ to $\TT^m\times 1$ along the map given by a matrix $A\in \SL(\ZZ,n)$.
The matrix $A$ has to meet two conditions.
On one hand, we need to make sure that the fundamental group of $E$ does not contain a nilpotent group of finite index.
This can be achieved by making at least one of eigenvalues of $A$ different from 1 by absolute value.
On the other hand the space $E$ has to admit a metric with small curvature and diameter, this can be achieved by making all
eigenvalues of $A$ close enuf to 1 by absolute value.
So the $n{\times}n$ matrix
\[A=
\left(
\begin{matrix}
0&0&\dots&0&1&1
\\
1&0&\dots&0&0&0
\\
\vdots&\vdots&\ddots&\vdots&\vdots&\vdots
\\
0&0&\dots&0&0&0
\\
0&0&\dots&1&0&0
\\
0&0&\dots&0&1&0
\end{matrix}
\right).
\]
for large enuf $n$ does the job;
its characteristic polynomial is
\[\lambda^n+\lambda+1.\qedsin\]
\medskip
\label{page-sol:almost-flat}
This example was constructed
by Galina Guzhvina \cite[see][]{guzhvina}.
It is expected that for small enuf $\eps>0$,
a Riemannian manifold $(M,g)$ of any dimension
with $\diam(M,g)\le 1$ and $|K_g|\le \eps$ cannot be simply connected,
here $K_g$ denotes the sectional curvature of $g$.
The latter does not hold with the condition $K_g\le \eps$ instead.
In fact,
for any $\eps>0$,
there is a metric $g$ on $\mathbb{S}^3$
with $K_g\le \eps$ and $\diam(\mathbb{S}^3,g)\le 1$.
This example was originally constructed by Mikhael Gromov in \cite{gromov-almost-flat};
a simplified proof was given by
Peter Buser
and Detlef Gromoll in \cite{buser-gromoll}.
The space $P$ can be found among the spherical polyhedral spaces which admit
an isometric circle action $\mathbb S^1\acts P$ with geodesic orbits.
One of these orbits is the singular set.
\medskip
\begin{wrapfigure}{r}{19 mm}
\begin{lpic}[t(-0 mm),b(-2 mm),r(0 mm),l(0 mm)]{pics/deltoid(1)}
\lbl[b]{9,36;$n$}
\lbl[t]{9,1;$s$}
\lbl[r]{1.5,25;$w$}
\lbl[l]{17.5,25;$e$}
\lbl[t]{9.5,29.5;{\small $\tfrac{2{\cdot}\pi}p$}}
\lbl[b]{9.5,6.5;{\small $\tfrac{2{\cdot}\pi}q$}}
\lbl{5,24;{\small $\theta$}}
\lbl{14,24;{\small $\theta$}}
\end{lpic}
\end{wrapfigure}
First let us construct the quotient space $P'=P/\mathbb S^1$.
Consider the Hopf action $\mathbb S^1\acts\mathbb S^3$.
Let
\[h\:\mathbb S^3\to\CP^1=\mathbb S^3/\mathbb S^1\]
be the quotient map.
The complex projective line $\CP^1$ is isometric to the 2-sphere of radius $\tfrac12$.
Fix large relatively prime integers $p$ and $q$.
Consider a solid deltoid $\Delta=[nesw]$ in $\CP^1$ with angles
\begin{align*}
\measuredangle n&=2 \cdot\tfrac\pi p,
\\
\measuredangle s&=2\cdot\tfrac\pi q,
\\
\measuredangle e=\measuredangle w&=\theta<\pi.
\end{align*}
The quotient space $P'$ is obtained from $\Delta$ by gluing its boundary to itself along the reflection symmetry.
Note that $P'$ is homeomorphic to $\mathbb S^2$ and
it has 3 singular points with total angles $2\cdot\tfrac\pi p$,
$2\cdot\tfrac\pi q$ and $2\cdot\theta$.
The required $\mathbb S^1_{p,q}$-action on $\mathbb{S}^3\subset\CC^2$ is given by complex diagonal matrices $\left(\begin{smallmatrix}z^p&0\\0&z^q\end{smallmatrix}\right)$.
The orbits with isotropy groups $\ZZ_p$ and $\ZZ_q$ correspond to $n$ and $s$ in $P'$.
There is a metric $\rho$ on $\mathbb{S}^3$ which is invariant with respect to the $\mathbb S^1_{p,q}$-action, which has $P'$ as the quotient space and geodesic orbits.
(In fact, such metric $\rho$ is unique up to isometry.)
This space $P=(\mathbb{S}^3,\rho)$ can be constructed the following way.
Set $W=h^{-1}(\Delta)$;
let $\tilde W$ be the universal metric cover of $W$.
The $\mathbb{S}^1$-action on $W$ lifts to $\RR$-action on $\tilde W$,
so the subgroup $2\cdot\pi\cdot\ZZ<\RR$ is the group of deck transformations of the covering $\tilde W\z\to W$.
Consider the quotient $W'=\tilde W/a\cdot\ZZ$;
note that $W'$ is a solid torus with $\mathbb{S}^1$-action and the quotient isometric to $\Delta$.
For the right choice of the constant $a$, the gluing map $\Delta\to P'$ can be lifted to a gluing map boundary of $W'\to P$.
In the constructed example
the singular points with total angles $2\cdot\tfrac\pi p$ and
$2\cdot\tfrac\pi q$
correspond to the points with isotropy groups $\ZZ_p$ and $\ZZ_q$ of the action.
The points in $P=(\mathbb{S}^3,\rho)$ on the orbits over these points will be regular;
that is, they admit a neighborhoods isometric to open sets in the unit 3-sphere.
The singular locus $P_s$
of $P$ corresponds to the remaining singular point of $P'$.
Note that by construction,
\begin{itemize}
\item $P_s$ is a closed geodesic with angle $2\cdot\theta$ around it and $\theta$ can take arbitrary value close to $\pi$.
\item $P_s$ forms a $(p,q)$-torus knot in the ambient $\mathbb{S}^3$.
\end{itemize}
\qedsf
Fix large relatively prime integers $p>q$.
Choose a regular $p$-gon $K_p$ and $q$-gon $K_q$ in $\CP^1$ with equal sides.
Consider their inverse images $W_p=h^{-1}K_p$ and $W_q=h^{-1}K_q$;
let $\tilde W_p$ and $\tilde W_q$ be their universal metric covers.
The spaces $\tilde W_p$ and $\tilde W_q$ admit natural isometric actions by groups $\RR\times\ZZ_p$ and $\RR\times\ZZ_q$ respectively.
It remains to find $\ZZ$ subactions on $\tilde W_p$ and $\tilde W_q$ such that the solid toruses $\tilde W_p/\ZZ$ and $\tilde W_q/\ZZ$ can be glued together long an isometry of their boundaries which sends orbits to orbits.
Consider the quotient space $\CP^1=\mathbb S^3/\mathbb S^1$;
it is isometric to the 2-sphere of radius $\tfrac12$.
Fix large relatively prime integers $p>q$.
Consider the solid triangle $\Delta=[nsw]$ with angles $\measuredangle n=\tfrac\pi p$, $\measuredangle s=\tfrac\pi q$ and $\measuredangle w=\pi\cdot(1-\tfrac1 p)$ in $\CP^1$.
Denote by $\hat \Delta$ the doubling of $\Delta$ along its boundary.
Note that $\hat \Delta$ is homeomorphic to $\mathbb S^2$ and
it has 3 singular points with total angles $2\cdot\tfrac\pi p$,
$2\cdot\tfrac\pi q$ and $2\cdot\pi\cdot(1-\tfrac1 p)$.
Consider the quotent map $\mathbb S^3\to \mathbb S^3/\mathbb S^1$ for the $\mathbb S^1$-action on $\mathbb{S}^3\subset\CC^2$ by the diagonal matrices $\left(\begin{smallmatrix}z^p&0\\0&z^q\end{smallmatrix}\right)$, $z\in\mathbb S^1\subset\CC$.
Identify the quotient space $\mathbb S^3/\mathbb S^1$ with $\hat \Delta$ such that the singular orbits mapped to $v$ and $w$ respectively.
Let us equip $\mathbb S^3$ with a metric such that the
Note that there is a spherical polyhedral metric $\rho$ on $\mathbb S^3$
such that the $\mathbb S^1$-orbits become geodesics
and the quotient space $(\mathbb S^3,\rho)/\mathbb S^1$
is isometric to $\hat \Delta$.
To construct such metric, take the inverse image $W$ of $\Delta$ under the Hopf fibration $\mathbb S^3 \to \CP^1$.
The universal cover $\tilde W$ of $W$ admits a natural $\RR$-action with geodesic orbits.
For the right choice of $\ZZ$ subaction of $\RR$,
the quotient space
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{Function with no critical points}
\begin{pr}{}{Function with no critical points}\label{Function with no critical points}
Given an integer $m\ge 2$,
construct a smooth function $f\:\RR^m\to \RR$
with no critical points in the unit ball $B^m$
such that the restriction $f|_{B^m}$ does not factor thru a linear function;
that is,
$f|_{B^m}$ cannot be presented as a composition
$\ell\circ\phi$,
where $\ell\:\RR^m\to\RR$ is a linear function
and $\phi\:B^m\to\RR^m$ is a smooth embedding.
\end{pr}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{wrapfigure}{o}{24 mm}
\begin{lpic}[t(-0 mm),b(0 mm),r(0 mm),l(0 mm)]{pics/no-critical-points()}
\end{lpic}
\end{wrapfigure}
\parbf{Function with no critical points.}
Construct an immersion
$\psi\:B^m\z\looparrowright\RR^m$ such that
\[\ell\circ\phi\ne\ell\circ\psi\]
for any embedding $\phi\:B^m\to\RR^m$.
The two-dimensional case can be guessed from the picture.
It remains to note that the composition $f=\ell\circ\psi$ has no critical points.\qeds
The problem was suggested by Petr Pushkar.
\parbf{A family of sets with no section.}
Identify $\mathbb{S}^1$ with $[0,1]/(0\sim 1)$.
Given $t\in[0,\tfrac12]$,
Consider the set $K_t\subset \mathbb{S}^1$
formed by all possible sums $\sum_{n=1}^\infty a_n\cdot t^n$,
where $a_n$ is $0$ or $1$.
Note that $K_{t}$ is a Cantor set and $K_{\frac12}=\mathbb{S}^1$.
Denote by $\rho_t\:\mathbb{S}^1\to\mathbb{S}^1$
the counterclockwise rotation by angle $\frac1{1-2\cdot t}$.
Set
\[Z_t=\left[\begin{aligned}
Z_t&=\rho_t(K_t)&&\text{if}\ t\in[0,\tfrac12),
\\
Z_{t}&=\mathbb{S}^1&&\text{if}\ t=\tfrac12.
\end{aligned}
\right.
\]
Note that the Hausdorff distance from $K_t$ to $K_{\frac12}=\SS^1$ converges to $0$ as $t\to\tfrac12$.
Therefore $Z_t$ is a continuous family.
Any continuous section $z(t)\in Z_t$ for $t\in[0,\tfrac12)$ can be described as
\[z(t)=\rho_t\left(\sum_{n=1}^\infty a_n\cdot t^n\right)\]
for a fixed sequence $(a_n)$.
None of these can be extended continuousely to $[0,\tfrac12]$.\qeds
Let us denote by $\mathrm{N} M$ and $\T M$ the normal and tangent bundle of $M$ in the Euclidean space $\RR^n$.
Assume $m=\dim M$.
Consider the normal exponential map $\exp_M\:\mathrm{N} M\to\RR^n$ [defined on page~\pageref{page:Normal exponential map}]. Denote by $J_v$ the Jacobian of $\exp_M$ at $v\in \mathrm{N}M$.
Note that for all small $r>0$, we have
\[\vol B(M,r)
=
\int\limits_M d_p\vol_m\cdot\int\limits_{B(0,r)_{\mathrm{N}_p}}J_v\cdot d_v\vol_{n-m},\]
where $B(0,r)_{\mathrm{N}_p}$ denotes the ball in the fiber $\mathrm{N}_p\subset \mathrm{N} M$ at $p\in M$.
Fix $p\in M$.
Note that it is sufficient to show that following.
\begin{itemize}
\item[$({*})$] The integral
\[V(r)=\int\limits_{B(0,r)_{\mathrm{N}_p}}J_v\cdot d_v\vol_{n-m}\]
can be written in terms of $m$, $n$, $r$ and the Riemannian curvature tensor $R$ of $M$ at $p$.
\end{itemize}
In fact, $V(r)$ is a polynomial in $r$ and its coefficient are up to a coefficient depending on $m$ and $n$ equal to the so-called Lipschitz--Killing curvatures at $p$.
Denote by $s\:\T_p\times \T_p\to \mathrm{N}_p$
the second fundamental form of $M$ [defined on page~\pageref{page:second fundamental form}].
Recall that the curvature tensor $R$ of $M$ at $p$ can be expressed the following way
\[\langle R(X, Y) V, W\rangle
=\langle s(X,W), s(Y,V)\rangle-\langle s(X,V), s(Y,W)\rangle.\]
Fix a orthonormal basis $e_1,\dots, e_{n-m}$ in $\mathrm{N}_p$.
Consider the real-valued quadratic forms $s_i(X,Y)=\<s_p(X,Y),e_i\>$.
Consider the tensors $R_i$ defined as
\[\langle R_i(X, Y) V, W\rangle
=s_i(X,W)\cdot s_i(Y,V)-s_i(X,V)\cdot s_i(Y,W).\]
Note that
\[R=R_1+\dots+R_{n-m}.\leqno{({*}{*})}\]
Assume $v=r\cdot e_i$ for some $r\in \RR$.
Denote by $k_{1},\dots, k_{m}$ the eigenvalues of the real-valued quadratic form $s_i(X,Y)$.
By straightforward calculations we get
\begin{align*}
J_{v}&=r^{n-m-1}\cdot(1+k_1\cdot r)\cdots(1+k_n\cdot r),
\\
J_{-v}&=r^{n-m-1}\cdot(1-k_1\cdot r)\cdots(1-k_n\cdot r).
\end{align*}
It follows that the sum $J_{v}+J_{-v}$ is a polynomial in $r$ and its coefficient
symmetric polynomials of $Q_j(k_1,\dots, k_n)$; more over each monomial of $Q_j$ is a product of even number of distinct $k_i$'s.
Since each product $k_a\cdot k_b$ is the sectional curvature in the tensor $R_i$,
we get that the coefficient $Q_j(k_1,\dots, k_n)$ can be expressed in terms of the tensor $R_i$.
Using $({*}{*})$, we can sum up by $i$, we get that
\[\sum_i(J_{r\cdot e_i}+J_{-r\cdot e_i})\]
can be expressed in terms of $r$, $m$, $n$ and $R$.
Taking average for all orthonormal frames $e_1,\dots, e_{n-m}$ in $\mathrm{N}_p$
we get the needed claim.
Show that for small $r$ the integral
\[v(r)=\int\limits_{B(0,r)_{\mathrm{N}_pM}}J_V\cdot d_V\vol_{n-m}\]
is a polynomial
of $r$ and its coefficients can be expressed in terms of the curvature tensor $R_p$.
It follows that the right hand side in $({*})$ can be expressed in terms of curvature tensor of $M$.
The problem follows since the curvature tensor can be expressed in terms of metric tensor of $M$.\qeds
The formula for volume of tubular neighborhood
was given by Hermann Weyl in \cite{weyl}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{Bounded geometry}
Denote by $\mathcal{R}$ the space of
all Riemannian metrics on $\mathbb S^5$
with absolute value of sectional curvature $\le1$,
and injectivity radius $\ge1$.
It is easy to see that any metric $g_1\in \mathcal{R}$
can be connected to the canonical metric $g_0$ on $\mathbb S^5$
by a continuous family of metrics $g_t\in \mathcal{R}$ where $t\in[0,1]$.
In fact, the one parameter family of metrics $g_t$
can be found among the metrics of the type
\[g_t=a(t)\cdot g_0+b(t)\cdot g_1,\]
where $a,b\:[0,1]\to\RR$
are smooth functions such that $a(0)=1=b(1)$ and $a(1-s)=0=b(s)$ for $s\le \tfrac13$.
In order to keep the bounds on the curvature and injectivity radius,
the functions $a$ and $b$
have to take huge values in the middle of interval.
\begin{pr}{\thm}{Bounded curvature}\label{Bounded curvature}
Fix a fast growing function, say
\[\Phi(x)=1000^{1000\cdot (x+1000)}.\]
Show that there is a metric $g_1$
such that
for any family $g_t$ as above
\[\max_{t\in[0,1]}\{\vol(g_t)\}
>
\Phi(\vol(g_1)).\]
\end{pr}
The expected solution requires Novikov theorem on the algorithmic undecidability of the problem of recognition of the sphere $\mathbb S^m$ for
$m\ge 5$.
A detailed proof of this theorem can be found in \cite{nabutovsky-NovThm}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\parbf{Bounded geometry.}
Show that there is an algorithm to estimate the Gromov--Hausdorff distance between two Riemannian manifolds given in any reasonable way.
Show that if two manifolds with bounded curvature are sufficiently close to each other then they are diffeomorphic.
Now assume contrary;
that is, for any metric $g_1\in \mathcal{R}$ there is a path $g_t$ in $\mathcal{R}$
connecting the canonical metric $g_0$ to $g_1$ such that
\[\vol g_t\le \Phi(\vol g_1).\]
If a 5-dimensional Riemannian manifold $M$ with curvature between $\pm1$ is diffeomorphic to $\mathbb{S}^5$
then it can be described by a metric $g_1$ in $\mathcal{R}$.
Let $g_t$ be the path as above.
Construct a finite set $F\subset \mathcal{R}$
which is sufficiently dense in the set of metrics in $\mathcal{R}$ with volume at most $\Phi(\vol g_1)$.
The path $g_t$ as above can be approximated by a sequence of metrics from $F$.
Using the algorithms above one can list all such sequences.
It implies existence of algorithm
which recognize $\mathbb S^5$ among $5$-dimensional manifolds.
The later contradicts Novikov theorem.
\qeds
Instead of function $\Phi$ one can take any Turing computable function.
The problem and number of its generalizations
are due to Alexander Nabutovsky \cite[see][]{nabutovsky-Disconnectedness}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
{
\begin{wrapfigure}{r}{21 mm}
\begin{lpic}[t(-0 mm),b(-4 mm),r(0 mm),l(0 mm)]{pics/long-geodesic(1)}
\end{lpic}
\end{wrapfigure}
\subsection*{Long geodesic}
Recall that a closed curve
called \index{simple closed curve}\emph{simple} if it has no self-intersections.
\begin{pr}{}{Long geodesic}\label{Long geodesic}
Assume that the surface $\Sigma$ of a convex body $B$ in $\RR^3$
admits an arbitrary long simple closed geodesic.
Show that $B$ is a tetrahedron with equal opposite sides.
\end{pr}
}
A tetrahedron with equal opposite edges is called \emph{isosceles} it has many interesting properties.
Let us mention couple of theorems about intrinsic metric on convex surfaces which should help to solve this problem.
These theorems can be proved easily for smooth or polyhedral surfaces
and then the general case can be done by approximation.
A very short but comprehensive introduction to the subject was written by Alexander Alexandrov in \cite{alexandrov1941}.
On a convex surface (not necessary smooth) one could define
so
called \index{curvature measure}\emph{curvature measure} which we denote further by $\kappa$.
It is the (necessary unique) non-negative measure such that for any triangle $\triangle$, we have
\[\kappa(\triangle)=\alpha+\beta+\gamma-\pi,\]
where $\alpha$, $\beta$ and $\gamma$ are the angles of $\triangle$, measured in the intrinsic metric of the surface.
For curvature measure, an analog of Gauss--Bonnet formula holds;
in particular for any closed convex surface $\Sigma$ in $\RR^3$
\[\kappa(\Sigma)=4\cdot\pi\]
and
if a closed geodesic cuts a disc $D$ from $\Sigma$ then
\[\kappa(D)=2\cdot\pi.\]
%???+PIC
Further, given a triangle $\triangle$ in a metric space,
its model triangle $\tilde \triangle$ is defined as a triangle in the plane with the same side lengths.
The angles $\tilde\alpha$, $\tilde\beta$ and $\tilde\gamma$ of the model triangle are called \index{model angle}\emph{model angle}
of triangle.
The comparison theorem states that for any triangle in a surface with non-negative curvature measure its model angles do not exceed the actual angles; that is,
\[\tilde\alpha\le \alpha,\quad
\tilde\beta\le\beta,\quad
\tilde\gamma\le\gamma.\]
The same holds for the area; that is,
\[\area\tilde\triangle\le \area\triangle.\]
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\parbf{Long geodesic.}
By cutting the surface $\Sigma$ along a sufficiently long closed simple geodesic,
we get two discs.
The key step is to show that each of these discs
is long and thin.
\medskip
Let $D$ be one of these discs.
Equip it with the intrinsic metric further denoted by $|{*}-{*}|_D$.
Since $\Sigma$ has non-negative curvature in the sense of Alexandrov,
so does $D$.
Choose a pair of points $p,q\in\partial D$ which maximize the distance $|p-q|_D$.
Clearly,
\[|x-p|_D,|x-q|_D\le |p-q|_D\]
for any other point $x\in\partial D$.
By comparison,
\[\measuredangle[x\,^p_q]\ge \tfrac\pi3.\leqno({*})\]
\begin{wrapfigure}{r}{28 mm}
\begin{lpic}[t(-0 mm),b(-0 mm),r(0 mm),l(0 mm)]{pics/long-geodesic-diam(1)}
\lbl[r]{0,8;$p$}
\lbl[l]{28,8;$q$}
\lbl[b]{10,16;$\gamma_1(t)$}
\lbl[l]{12,9,-60;$\ge\tfrac\pi3$}
\end{lpic}
\end{wrapfigure}
The points $p$ and $q$ divide $\partial D$ into two arcs,
say $\gamma_1$ and $\gamma_2$;
let us parametrize them by arclength from $p$ to $q$.
Then by $({*})$
\[\tfrac{d}{dt}\left(|x-\gamma_i(t)|_D-|x-\gamma_i(t)|_D\right)
\ge
\tfrac12.\]
In particular
\[|p-q|_D\ge \tfrac18{\cdot}\length[\partial D].\]
In particular, if $\partial D$ is long
then $D$ has large diameter.
Choose two points $x\in \gamma_1$ and $y\in\gamma_2$ sufficiently close to $p$ such that $|x-q|_D=|y-q|_D$.
By comparison
\[\area \tilde \triangle qxy\le \area \triangle qxy\le \area \Sigma.\]
It follows that
\[\begin{aligned}|x-y|_D&\le2\cdot\frac{ \area[\tilde\triangle(xyq)]}{|q-x|_D}
\le
\\
&\le
100\cdot\frac{ \area\Sigma}{\length[\partial D]}.
\end{aligned}
\leqno({*}{*})\]
Cut $D$ along a minimizing geodesic $[xy]$ into two lenses.
Denote by $L_p$ the lens with the point $p$ in it.
Note that the curvature of $L_p$ is $\alpha+\beta$, where $\alpha$ and $\beta$ the angles as on the diagram.
By comparison $\alpha\ge \tilde\measuredangle(x\,^p_y)$
and $\beta\ge \tilde\measuredangle(y\,^p_x)$.
Therefore curvature of $L_p$ is at least $\pi-\tilde\measuredangle(p\,^x_y)$.
In particular, if $|x-y|_D$ much less then $|p-x|_D+|p-y|_D$ then the curvature of $L_p$ is $\pi$ or just a little less.
Fix $\eps>0$.
If $\length[\partial D]$ is long enuf,
by $({*})$,
we can find a lens $L_p$ with diameter at most $\eps$,
such that curvature $L_p$ is at least $\pi-\eps$.
Using the same construction for $p$ and $q$ in the disc $D$,
and for the two points in the other disc,
we get four lenses in $\Sigma$ each of diameter at most $\eps$ and each with curvature at least $\pi-\eps$.
\begin{center}
\begin{lpic}[t(-0 mm),b(-0 mm),r(0 mm),l(0 mm)]{pics/long-geodesic-D(1)}
\lbl{46,6;$D$}
\lbl{6,6;$L_p$}
\lbl[r]{0,6;$p$}
\lbl[l]{92.5,6;$q$}
\lbl[b]{19,11.5;$x$}
\lbl[t]{12.5,.5;$y$}
\lbl[b]{46,12;$\gamma_1$}
\lbl[t]{46,-.5;$\gamma_2$}
\lbl[tr]{16.5,9;{\small $\alpha$}}
\lbl[br]{12.5,4.5;{\small $\beta$}}
\end{lpic}
\end{center}
By Gauss--Bonnet formula, the total curvature of $\Sigma$ is $4\cdot\pi$.
Since $\eps>0$ is arbitrary, we get that there are four singular points in $\Sigma$, each with curvature $\pi$
and the remaining part of $\Sigma$ is flat.
\begin{wrapfigure}{o}{21 mm}
\begin{lpic}[t(-4 mm),b(-3 mm),r(0 mm),l(0 mm)]{pics/akopyan(1)}
\end{lpic}
\end{wrapfigure}
It remains to show that any surface with this property
is isometric to the surface of a tetrahedron with equal opposite edges.
To do this cut $\Sigma$ along three geodesics which connect one singular point to the remaining three,
develop the obtained flat surface on the plane and think;
also look at the diagram.\qeds
I learned the problem from Arseniy Akopyan,
the case of polyhedra was solved by Vladimir Protasov in \cite{protasov}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{The last problem of Poincar\'e\hard}
\begin{pr}{\hard}{The last problem of Poincar\'e}\label{The last problem of Poincare}
Let $f\:\CC\to \CC$ be an area preserving homeomorphism
such that
\[f(z)
=
\left[
\begin{aligned}
&z-i&&\text{if}&&\Re(z)\le -1,
\\
&z+i&&\text{if}&&\Re(z)\ge 1.
\end{aligned}
\right.
\]
and $f(z+i)=f(z)+i$ for any $z\in\CC$.
Show that $f$ has a fixed point.
\end{pr}
\parbf{The last problem of Poincar\'e.}
Set
\begin{align*}
H_+&=\set{z\in\CC}{\Re(z)\ge 1},
\\
H_-&=\set{z\in\CC}{\Re(z)\le -1}.
\end{align*}
Assume $f$ has no fixed points;
in other words the image of the map
\[\phi\:z\mapsto f(z)-z\]
lies in $\CC^*=\CC\setminus\{0\}$.
\begin{wrapfigure}{r}{44 mm}
\begin{lpic}[t(-0 mm),b(0 mm),r(0 mm),l(0 mm)]{pics/birkhoff(1)}
\lbl[r]{33,44.5;{\small $\hat\gamma(0)$}}
\lbl[r]{31,57.5;{\small $\hat\gamma(1)$}}
\lbl[rb]{6.5,63;{\small $\hat\gamma(n)$}}
\lbl[l]{10.3,28.3;{\small $\check\gamma(0)$}}
\lbl[l]{12.4,18.6;{\small $\check\gamma(1)$}}
\lbl[l]{37,19;{\small $\check\gamma(n)$}}
\lbl[b]{21.5,35.5;$0$}
\lbl[br]{33.5,35.5;$1$}
\lbl{40,3;$H_+$}
\lbl{3,3;$H_-$}
\end{lpic}
\end{wrapfigure}
Fix $\eps>0$ such that $|f(z)-z|>\eps$ for any $z\in\CC$.
Note that the map
\[\check f\:z\mapsto f(z)+\eps\]
is area preserving and has no fixed points.
Prove that for some positive integer $n$,
there is a curve
\[\check \gamma\:[0,n]\to \CC\]
which starts in $H_-$, ends in $H_+$
and
$\check f\circ\check\gamma(t)=\check\gamma(t+1)$
for any $t\z\in [0,n-1]$.
Repeat the same construction for the function
\[\hat f(z)=f(z)-\eps\]
and obtain a curve
\[\hat \gamma\:[0,m]\to \CC\] starting in $H_+$ and ending in $H_-$.
Connect $\check\gamma(n)$ to $\hat \gamma(0)$ by a curve in $H_+$
and
$\hat\gamma(m)$ to $\check\gamma(0)$ by a curve in $H_-$.
Denote by $\sigma$ the obtained loop.
Prove that
\begin{itemize}
\item The loop $\phi\circ\sigma$ has to be null-homotopic in $\CC^*$.
\item The loop $\phi\circ\sigma$ is a generator of $\pi_1\CC^*$.
\end{itemize}
\noindent
These two statements contradict each other. \qeds
The question was asked by Henri Poincar\'e \cite[see][]{poincare}
and answered by George Birkhoff in \cite{birkhoff}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{wrapfigure}{r}{40 mm}
\begin{lpic}[t(-0 mm),b(-0 mm),r(0 mm),l(0 mm)]{pics/serpinski-cemetery(1)}
\end{lpic}
\end{wrapfigure}
\subsection*{Fat curve}
\label{Fat curve}
\begin{pr}{\easy}{Fat curve}
Construct a simple plane curve with positive Lebesgue measure.
\end{pr}
\parbf{Fat curve.}
Modify your favorite space filling curve
keeping its area nearly the same
and removing the self-intersections.
It works for the \index{Sierpi\'nski curve}\emph{Sierpi\'nski curve}
which can be constructed as a limit of
recursively defined sequence of curves;
the 8th iteration is on the diagram.\qeds
This example was constructed
by William Osgood \cite[see][]{osgood}.
%Knopp, K. (1917), "Einheitliche Erzeugung und Darstellung der Kurven von Peano, Osgood und von Koch", Archiv der Mathematik und Physik, 26: 103--115.
%Lebesgue, H. (1903), "Sur le probl\`eme des aires", Bulletin de la Soci\'et\'e Math\'ematique de France , 31: 197--203
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{Characterization of polytope}
\begin{pr}{}{Characterization of polytope}
\label{conic neighborhoods}
Let $P$ be a compact subset of the Euclidean space.
Assume for every point $x\in P$
there is a cone $K_x$ with a tip at $x$ and $\eps>0$
such that
$$B(x,\eps)\cap P
=
B(x,\eps)\cap K_x.$$
Show that $P$ is a polytope;
that is, $P$ is a union of finite collection of simplices.
\end{pr}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\parbf{Characterization of polytope.}
Arguing by contradiction, let us assume that $P\subset \RR^m$
is a counterexample and $m$ takes minimal possible value.
Choose a finite cover $B_1,B_2,\dots B_n$ of $K$,
where $B_i=B(z_i,\eps_i)$
and $B_i\cap P=B_i\cap K_i$,
where $K_i$ is a cone with the tip at $z_i$.
For each $i$, consider function $f_i(x)=|z_i-x|^2-\eps_i^2$.
Note that
\[W_{i,j}=\set{x\in\RR^m}{f_i(x)=f_j(x)}\]
is a hyperplane for any pair $i\ne j$.
The subset $P_{i,j}=P\cap W_{i,j}$ satisfies the same assumptions as $P$, but lies in a hyperplane.
Since $m$ is minimal, we get that $P_{i,j}$ is a polytope for any pair $i,j$.
Consider Voronoi domains
\[V_{i}=\set{x\in\RR^m}{f_i(x)\ge f_j(x) \ \text{for any}\ j}.\]
Note that $P\cap V_i$ is formed by the points which lie on the segments from $z_i$ to a point in $P\cap \partial V_i$.
The statement follows since $\partial V_i$ is covered by the hyperplanes $W_{i,j}$.
\qeds
This problem is mentioned
by Nina Lebedeva and me in \cite{lebedeva-petrunin}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection*{Thin triangles\easy}
A triangle in a metric space is defined as three minimizing geodesics connecting three distinct points.
A model triangle is defined as a flat triangle with the same side lengths.
If the distance between any two points in the triangle is less or equal to the distance between corresponding points in
A triangle is called thin if the distance between any pair of
\begin{pr}{\easy}{Unique geodesics imply $\mathrm{CAT}(0)$}\label{Unique geodesics imply CAT}
Let $P$ be a polyhedral space.
Assume that any two points in $P$
are connected by unique geodesic.
Show that $P$ is a $\mathrm{CAT}(0)$ space.
\end{pr}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\parbf{Unique geodesics imply $\mathrm{CAT}(0)$.}
Uniqueness of geodesics implies that $P$ is contractible.
In particular, $P$ is simply connected.
It remains to prove that $P$ is locally $\mathrm{CAT}(0)$;
equivalently, the space of directions $\Sigma_p$
at any point $p\in P$ is a $\mathrm{CAT}(1)$ space.
We can assume that the statement holds in all dimensions less than $\dim P$.
In particular, $\Sigma_p$ is locally $\mathrm{CAT}(1)$.
If $\Sigma_p$ is not $\mathrm{CAT}(1)$,
then it contains a periodic geodesic $\gamma$ of length $\ell<2\cdot\pi$,
such that any arc of $\gamma$ of length $\tfrac\ell2$ is length minimizing.
Consider two points $x$ and $y$
in the tangent cone of $p$
in directions $\gamma(0)$ and $\gamma(\tfrac\ell2)$.
Show that there are two distinct minimizing geodesics between $x$ and $y$.
The latter leads to a contradiction.
\qeds
The existence of geodesic $\gamma$ was proved by Brian Bowditch in \cite{bowditch};
a simpler proof can be found in the book
by Stephanie Alexander, Vitali Kapovitch and me \cite[see][]{akp}.