-
Notifications
You must be signed in to change notification settings - Fork 0
/
comprehensive.tex
2101 lines (1966 loc) · 104 KB
/
comprehensive.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
\documentclass[12pt]{article}
\usepackage{amsmath}
\usepackage{amsthm}
\usepackage{amssymb}
\usepackage[top=1in,bottom=1.25in,left=1.25in, right=1.25in]{geometry}
\newtheorem{theorem}{Theorem}
\newtheorem{definition}{Definition}
\newtheorem{problem}{Problem}
\begin{document}
\section{Introduction} This document contains theorems and sample
problems for the Graph Theory Comprehensive exam at Arizona State
University. Most of the theorems are found in Diestel's Graph Theory
and when a theorem number is mentioned, that number refers to the 3rd
edition. Most of the proofs that are provided are generally similar to
proofs in Diestel's Graph Theory, or West's Introduction to Graph
Theory. This document was originally formed while studying for the ASU
graph theory comprehensive and as such is incomplete. Some proofs are
missing, some proofs are better described as proof summaries, and the
occasional proof may be incorrect. Any corrections/additions to this
document are welcome and can be submitted to this documents GitHub
page located at https://github.com/rjoursler/graph\_notes.
\section{Basic Theorems}
\begin{theorem} (1.4.2) If $G$ is non-trivial, then
$\kappa(G) \leq \lambda(G) \leq \delta(G)$.
\end{theorem}
\begin{proof} Let $G$ be a non-trivial graph. To show that
$\kappa(G) \leq \lambda(G)$, let $C$ be a a set of edges with
$|C| = \lambda(G)$ which separates $G$ into sets $A$ and $B$. Note
$|A|, |B| \geq \lambda(G)$. By deleting the ends of edges in $C$
from $G$ in such a way that $A$ and $B$ are still non-empty we get
$\kappa(G) \leq \lambda(G)$. If this cannot be done, then
$\lambda(G) = 1$ and $|A| = |B| = 1$, implying $G = K_2$ in which
the claim still holds. For the fact that
$\lambda(G) \leq \delta(G)$, note than any vertex $v$ can be
disconnected from the graph by deleting all edges which contain $v$.
\end{proof}
\begin{theorem} (1.5.6) Every connected graph contains a normal spanning tree, with any
specified vertex as its root.
\end{theorem}
\begin{proof} The depth first search algorithm will yield a normal
tree starting from any vertex. Since $G$ is connected, every vertex
will be visited, thus the spanning tree is normal.
\end{proof}
\begin{theorem} (1.6.1) A graph is bipartite iff it contains no odd cycle.
\end{theorem}
\begin{proof} The forward direction follows immediately since an odd
cycle is $3$-partite.\\
For the backward direction, assume $G$ is a graph which contains no
odd cycle. WLOG we may assume that $G$ is connected, otherwise we
can reduce the problem to independently partitioning all the
connected subgraphs of $G$. Let $v \in E(G)$. Define the set $A$ to
be the set of vertices $a$ such that there is an even path from $v$
to $a$ and $B$ to be vertices $b$ such that there is an odd path
from $v$ to $b$. Since $G$ is connected $A \cup B = V(G)$. Also, if
the edge $xy$ exits within $A$ or $B$, then $x,y \in A$ and
$x,y \in B$. Finally, note that $A \cap B = \emptyset$ otherwise an
odd cycles could be found. Thus $G$ is bipartite with bipartition
$A,B$.
\end{proof}
\begin{theorem} (1.4.3 Maders Theorem) Let $0 \neq k \in
\mathbb{N}$. Every graph $G$ with $d(G) \geq 4k$ has a $(k + 1)$
connected subgraph $H$ such that $\epsilon(H) \geq \epsilon(G) - k$.
\end{theorem}
\begin{proof} Consider all graphs $G' \subseteq G$ such that
$||G'|| \geq \epsilon(G) (|G'| - k)$. Such graphs exists since $G$
satisfies these conditions. Let $H$ be such a graph such that $|H|$
is minimum. By the minimality of $H$, $\delta(H) \geq \epsilon(G)$,
otherwise we could delete such vertex to find a smaller graph
$H'$. Thus $|H| > \epsilon(G)$. Using this fact gives
\[
\epsilon(H) = \dfrac{||H||}{|H|} > \epsilon(G)(1 - \dfrac{k}{|H|})
\geq \epsilon(G) - k
\]
as desired.
All that is left to show is that $H$ is $k + 1$-connected. Assume to
the contrary. Then $H$ has a separating set $X$ of size at most $k$
which separates sets $A$ and $B$. Consider the graph $H[A \cup
X]$. Because of the minimum degree in $H$, for any vertex $a \in A$,
the degree of $a$ in $H[A \cup X]$ is at least $\delta(H)$ implying
$|H[A \cup X]| \geq 2k$. The same applies for $H[B \cup X]$. Thus we
get that
\[
||H[A \cup X] || \leq \epsilon(G)(|H[A \cup X] - k)
\]
and
\[
||H[B \cup X] || \leq \epsilon(G)(|H[B \cup X] - k)
\]
This implies that
\[
\begin{array}{rcl} ||H|| & \leq & ||H[A \cup X]|| + ||H[B \cup X]||\\
& \leq & \epsilon(G) (|H[A \cup X] + |H[B
\cup X]| - 2k)\\
& \leq & \epsilon(G) (|H| - 2k + |X|)\\
& \leq & \epsilon(G) (|H| - k)\\
\end{array}
\]
a contradiction.
\end{proof}
\begin{theorem} (1.8.1) A connected graph is Eulerian iff every vertex
has even degree.
\end{theorem}
\begin{proof} To begin with, assume $G$ is an Eulerian graph with
Eulerian tour $W$. Note that every vertex must be entered as many
times as it is exited in the tour $W$ thus every vertex in $G$ must
have even degree.\\
For the reverse direction, let $G$ be a connected graph such that
every vertex has even degree. Let $W = v_0 v_1 \ldots v_\ell$ be a
maximum length walk such that each edge appears at most once. Since
the degree of $v_0$, $v_0 = v_\ell$. Thus $W$ is a closed walk. If
$G$ has an edge outside of the walk then we can extend the walk, a
contradiction.
\end{proof}
\begin{theorem} (Dirac's Theorem 10.1.1) Every graph with $n \geq 3$
vertices and a minimum degree at least $n/2$ has a Hamilton cycle.
\end{theorem}
\begin{proof} To begin with, note that $G$ must be connected, because
the maximum degree in the smallest component otherwise would be less
than $n/2$. Let $P = v_0, \ldots, v_k$ be a maximum path in $G$. By
the maximality, of $P$, $N(v_0), N(V_k) \in V(P)$. If there is an
edge from $v_0$ to $v_k$ then we get a cycle $C$ on the vertices of
$P$. Otherwise, since the size of $P$ is smaller than $n$, by pigeon
hole, there must exists vertices $x_i$ and $x_{i + 1}$ such that
$v_0 x_{i + 1}$ and $v_k x_i$ are edges creating a different cycle
$C$ on the vertices of $P$. If $C$ is not spanning, then by the
connectivity of $G$, $C$ is connected to a vertex not in $C$, which
forms a longer path contradicting maximality of $P$.
\end{proof}
\begin{theorem} (Konig's Theorem 2.1.1) Let $G$ be a bipartite
graph. Then the size of the maximum matching in $G$ is equal to the
size of a minimum vertex covering in $G$
\end{theorem}
\begin{proof} Let $G$ be a bipartite graph with bipartition
$(A,B)$. To begin with note that the size of the minimum covering
has to be at least as large as the maximum matching in order to
cover every edge in a maximum matching. Now for equality, consider a
maximum matching $M$. We will create a covering $C$ by choosing a
one vertex from each edge of the matching, picking a vertex from $B$
if an alternating path with an unmatched vertex in $A$ ends on that
vertex, and picking the vertex from $A$ otherwise. Now we will show
that the set of vertices $C$ forms a covering.
Let $ab \in E$ be any edge in $G$. We will show that either $a$ or
$b$ lies in $C$. If $ab \in M$, then this holds trivially, so we may
assume that $ab \notin M$. To begin with, assume that $a$ is not
matched by $M$. Then there must exists an edge $a'b \in M$, and
since $b$ is the end of the alternating path $ab$, $ab$ is covered
by $C$. Otherwise there exists an edge $ab' \in M$. If $a \in C$,
then we are done. Otherwise $b' \in C$ by construction, so there is
an alternating path ending in $b'$. But this creates an alternating
path ending in $b$ as well. Since $M$ is maximum, that implies $b$
is matched with an edge $a'b \in M$ and thus $b \in C$. Therefore
$C$ is an edge cover proving the theorem.
\end{proof}
\begin{theorem} (Hall's Theorem 2.1.2) A bipartite graph $G$ with
bipartition $(A,B)$, contains a matching of $A$ iff
$|N(S)| \geq |S|$ for all $S \subseteq A$.
\end{theorem}
\begin{proof} To begin with, note that a matching of $A$ immediately
that $|N(S)| \geq |S|$ for all $S \subseteq A$. Now to show the
other direction, assume $G$ is a bipartite graph with bipartition
$(A,B)$ such that $N(S) \geq |S| $ for all $S \subseteq A$. Let $M$
be a maximum matching of $A$ and assume that $A$ is not
matched. Therefore there exists a vertex $a \in A$ not matched by
$M$. Let $A'$ be the set of vertices reachable from $a$ by an
alternating path, and let $B'$ be the penultimate vertices in all
such paths. Since very vertex of $A'$ corresponds to at least one
vertex of $B'$ we get that $|A'| \leq |B'|$. Similarly, every
neighbor of a vertex $a' \in A$, must appear in $B'$, otherwise we
could find either a new alternating path or an augmenting path. Thus
$|A'| = |B'|$ by the assumption. Thus by the assumption, if
$S = A' \cup \{ a\}$, there must be an edge $ab$ such that
$b \notin B'$. Note that $b$ must be matched otherwise we could add
$ab$ to the matching $M$ contradicting maximality. But then $b$ is
the penultimate vertex some some alternating path starting at $a$
and that $b \in B'$, a contradiction. Therefore $M$ must match all
of $A$.
\end{proof}
\begin{theorem} (2.1.3) If $G$ is a $k$-regular bipartite graph with
$k \geq 1$, then $G$ has a 1-factor
\end{theorem}
\begin{proof} Since $G$ is $k$-regular, then $|A| = |B|$. Thus if we
can find a perfect matching of $A$ onto $B$, then we are done. Let
$S \subseteq A$. Note that the degree sum of vertices in $N(S)$ is
at least as large as the degree sum of the vertices of $S$. Since
$G$ is $k$ regular we get that $k |S| \leq k N(S)$. Thus by Hall's
Marriage theorem we get a matching of $A$ onto $B$.
\end{proof}
\begin{theorem} (Petersen's Theorem 2.1.5) Every regular graph of
positive even degree has a 2-factor.
\end{theorem}
\begin{proof} Let $G$ be a regular graph of positive even
degree. Since every vertex has even degree, there exists an Eulerian
tour $W$. Use this tour to induce an orientation of $G$ and then
create to copies of $V(G)$, $V^-$ and $V^+$. Let $v_- v_+$ be an
edge between those two sets iff $v_- v_+$ is an edge in the
orientation of $G$. Note that by definition this new graph of $V^-$
and $V^+$ is $k$ regular, thus there exists a matching of $V^-$ onto
$V^+$. Projecting this matching back into $G$ then provides a
2-factor of $G$.
\end{proof}
\begin{theorem} (Stable Matching Theorem 2.1.4) Let $G$ be a bipartite
graph. For every set of preferences, $G$ has a stable matching.
\end{theorem}
\begin{proof} Construct a matching greedily as follows. Given a
matching $M$, for a given vertex, $b$ say that a vertex $a \in A$ is
acceptable if $a$ is unmatched or of $b$ prefers $a$ over its
current matched neighbor. Starting with the empty matching, pick a
vertex $a$ such that $a$ is unmatched but acceptable to some
$b$. Then add the edge $ab$ to the matching for the $b$ that $a$
most prefers, while deleting any edge $bc$ already existing in our
matching. Since this always increases the happiness of a vertex $b$,
this algorithm must end. Note at each step, no vertex $a$ desires to
pick a different vertex $b$ since it chose the best possible one
when the edge $ab$ was added and once a vertex is matched in $b$, it
will always be matched. Thus the matching is stable.
\end{proof}
\begin{theorem} (Tuttes's Theorem 2.2.3) Every graph $G$ contains a
subset $S$ such that $S$ is matchable to $C_{G - S}$ and every
component in $G-S$ is factor critical.
\end{theorem}
\begin{proof} The proof is by induction. For the base case, not that
when $|G| = 0$, the claim holds. For the induction, pick $G$ and
consider all sets $T$ such that $q(G-T) - |T|$ is maximized. Let $S$
be the the the set for which $|T|$ is also maximized. Now note that
there must be no even sized components in $G - S$, otherwise we
could get a larger set $S'$ by moving a vertex from the even
component into $S$. This new set $S'$ still maximizes
$q(G-S) - |T|$, thus contradicting the maximality of $S$.
Now assume that there exists a component $C$ in $G - S$ such that
$|C|$ is not factor critical. Let $C' = C-c$ where $c$ is a vertex
such that $C'$ has no $1$-factor. Then by the inductive hypotheses
there exists a set $S' \subseteq C$ such that $q(C' - S') >
|S'|$. Since $|C'|$ is even, the parity of $q(C'-S')$ and $|S'|$ are
the same, thus $q(C'-S') \geq |S'| + 2$. But then for
$T = S \cup \{c\} \cup S'$,
$q(G-T) - |T| \geq q(G-S) - |S| - 1 - 1 + q(G - S') - |S'| \geq
q(G-S) - |S|$, once again contradicting choice of $S$.
Finally, all that is left is to show that $S$ is matchable onto the
components of $G - S$. Assume to the contrary, then by Halls theorem
there exists a set $S' \subseteq S$ such the set $S'$ is connected
to fewer components than $|S'|$. But than deleting $S'$ from $S$
destroys fewer odd components from $G-S$ than it removes from $S$,
thus increasing the size of $q(G-T) - |T|$, contradicting the choice
of $S$ again.
\end{proof}
\[
s_k := 4k(\log k + \log \log k + 4) if k \geq 2 1 if k \leq 1
\]
\begin{theorem} (2.3.1) Let $k \in \mathbb{N}$, and let $H$ be a cubic
multigraph. If $|H| \geq s_k$, then $H$ contains $k$ disjoint
cycles.
\end{theorem}
\begin{proof} The proof will be by induction. Note the claim is
trivial if $k \leq 1$. Let $C$ be a shortest cycle in $H$. If $C$ is
a loop $xx$, then the only vertex meeting $C$ is $x$. Since
$d(x) = 3$, the graph $H - C$ contains $|C|$ vertices with degree
exactly one. Thus none of these vertices are in a cycle in $H -
C$. Thus taking $H - V(C)$ only contains vertices of degree $2$ or
degree $3$. Suppressing the vertices of degree $2$ results in
another cubic multigraph. In this process, at most $2|C|$ vertices
were deleted from $H$ and at least $|C|$ vertices were
deleted. Since $|C| \leq 2 \log |H|$, we get that
$|H'| \geq s_k - 4 \log s_k = s_{k -1 }$ for $k \geq 2$.
\end{proof}
\begin{theorem} (Erdos, Posa 2.3.2) There is a function
$f: \mathbb{N} \to \mathbb{R}$ such that, given any
$k \in \mathbb{N}$, every graph contains either $k$ disjoint cycles
or a set of at most $f(k)$ vertices meeting all of its cycles.
\end{theorem}
\begin{proof} Let $G$ be any graph with fewer than $k$ disjoint
cycles. We may assume it has a cycle. Let $H$ be the maximal
subgraph of $G$ for which ever vertex has degree either $2$ or
$3$. Let $U$ be the vertices of $H$ with degree at least $3$. Let
$C$ be the set of all cycles that avoid $U$ and meet $H$. Since $H$
is maximal, each cycle must meet at exactly one point. Thus the set
$Z$ disconnects all such cycles. Let $Y$ be a set of vertices with
exactly one vertex from the two regular components of $H$ that avoid
$Z$. Note each vertex in $Z$ and $Y$ corresponds to a disjoint
cycle, thus $|Y \cup Z| \leq k - 1$. Now consider any cycle $C$
which avoids $X$. Then $C$ is not a component, so $C$ must meet
$U$. From $2.3.1$, if $|U| \geq s_k$, $H$ contains $k$ disjoint
cycles. Otherwise $U \cup Y \cup X$ meets every cycle in $G$ and
this set is smaller than $s_k + k - 1$.
\end{proof}
\begin{theorem}{2.4.3} Let $G$ be a graph and $F = (F_1, \cdots F_k)$
be a $k$-tuple of edge-disjoint forests such that the number of
edges contained in the forests of $F$ is maximum. Then for every
edge $e \notin F'_i$ for all $i$, for some $F'$ obtained by a series
of edge replacements, there exists a set $U$ with $e \subseteq U$
such that $U$ is connected in all $F_i$.
\end{theorem}
\begin{proof} Let $E^0$ be the set of all edges which can be removed
from all $F_i$ by series of edge replacement and define
$G^0 = (V, E^0)$. Let $C^0$ be a component in $G^0$. We claim that
$U = V(C^0)$ is connected in all $F_i$.
Consider a family of trees $F'=(F'_1, \ldots, F'_k)$ obtained by
replacing an edge of $F_i$. We claim that if $x,y$ are the end of a
path in $F'_i \cap C^0$, then $x F_i y \subseteq C^0$. To see why,
let $e = vw$ be the new edge in $F'$. We may assume that
$e \in x F'_i y$, because otherwise the claim is obviously. Thus it
suffices to show that $v F_i w \subseteq C^0$. Let $e'$ be any edge
in $v F_i w$. Since we could replace $e'$ by $e$ we have that
$e' \in E^0$. Since this applies to all $e'$ we get that
$vF_i w \subseteq C^0$.
To finish the proof of this theorem, we can obtain a new forest
tuple $F'_i$ by replacing $F_i$ using $e$. Since $e$ can be viewed
as an $xF'_i y$ path in $C_0$, we get that $x F_i y \subseteq
C_0$. As this applies for all $i$ and edges in $C_0$, $U$ must be
connected in every $F_i$.
\end{proof}
\begin{theorem} (Nash-Williams 2.4.1) A multigraph contains $k$
edge-disjoint spanning trees iff for every partition $P$ of its
vertex set it has at least $k(|P|-1)$ cross-edges.
\end{theorem}
\begin{proof} For the forward direction, note that the graph obtained
by contracting the parts in a partition $P$ into single vertices
results in a connected graph. Thus each edge-disjoint spanning tree
has at least $|P| - 1$ vertices crossing between parts, resulting in
at least $k(|P| - 1)$ crossing edges.
For the reverse direction will be proved by induction on $|G|$. Note
for $|G| = 2$, the assertion holds. Now pick a $k$-tuple of edge
disjoint forests $F = (F_{1},\cdots F_{k})$ such that the number of
edges is maximized. If each $F_{i}$ is a tree, then we are
done. Otherwise we have
\[
||F|| = \sum_{i = 1}^k ||F_{i}|| < k(|G| - 1)
\]
On the other hand we know that $||G|| \geq k (|G| - 1)$ by the
assumption applied when $G$ is partitioned into $|G|$ parts. Thus
there exists an edge $e \notin F^0$. Thus by the previous lemma,
there exists a set $U$ with $|U| \geq 2$ that is connected in all
$F_i$. By the inductive hypothesis, $G\backslash U$ contains
$k$-edge disjoint trees $T_1 \cdots T_k$. Replacing the vertex $v_U$
in each $T_i$ with the corresponding graph $F_i[U]$ in $G$ then
gives $k$ edge-disjoint spanning trees in $G$.
\end{proof}
\begin{theorem} (Nash-Williams 2.4.4) A multigraph $G = (V, E)$ can be
partitioned into at most $k$ forests iff $||G[U]|| \leq k(|U| - 1)$
for every non-empty set $U \subseteq V$.
\end{theorem}
\begin{proof} For the forward direction, note that on any set $U$,
there are at most $|U| - 1$ edges associated with any forest. As
such, $||G[U]|| \leq k (|U| - 1)$ for all sets $U$.
The other direction will be proved by contrapositive. Assume that
$G$ cannot be partitioned into at most $k$ forests. Then for any
$k$-tuple $F = (F_1, \cdots F_k)$ with the number of edges
maximized, $F$ does not form a partition. But then there exists an
edge $e$ not in any $F_i$, so by a previous lemma there exists a set
$U$ such that $|U| \geq 2$ and such that $U$ is connected in every
$F_i$. But then $||G[U]|| \geq k(|U| - 1) + 1$.
\end{proof}
\begin{theorem} (Gallai \& Milgram 2.5.1) Every directed graph $G$ has
a path cover $\mathcal{P}$ and independent set
$\{ v_P | P \in \mathcal{P} \}$ of vertices such that $v_P \in P$
for every $P \in \mathcal{P}$.
\end{theorem}
\begin{proof} Let
$ter(\mathcal{P}) = \{ v_0, \cdots v_{|\mathcal{P}|} \}$ denote the
set of vertices ending the directed paths in $\mathcal{P}$. We will
prove by induction on $|G|$ that if $|ter(\mathcal{P})|$ is minimal
then there exists such an independent set as is claimed. Clearly
this is satisfied when $|G| = 0$. Let $\mathcal{P}$ be a path cover
over $G$ with $|ter(\mathcal{P})|$ minimal. If $ter(\mathcal{P})$ is
independent, then we are done. Otherwise we may assume that
$v_1 v_0$ is an edges. Let $v$ be the vertex preceding $v_0$ in the
path ending in $v_0$. Define $G' = G - v_0$. Now consider another
path family $\mathcal{P}'$ covering $G'$ for which
$ter(\mathcal{P'}) = \{ v, v_1, \ldots, v_{|\mathcal{P}|} \}$ (which
obviously exists). Now if this path family in $G'$ contains an
independent set then that set is obviously independent in $G$ as
well so we are done. Thus it suffices to show that
$|ter(\mathcal{P}')|$ is minimal.
For contradiction, assume there exists a family $\mathcal{P}''$ such
that $ter(\mathcal{P}'') \subset ter(\mathcal{P}')$. If $v$ or $v_1$
is in $ter(\mathcal{P}'')$, then this path cover can be extended to
a path cover in $G$ which contradicts the minimality of
$ter(\mathcal{P})$. Hence
$ter(\mathcal{P}'') \subseteq \{ v_3, \ldots, v_{|\mathcal{P}|}
\}$. But then $\mathcal{P}''$ and $v_1$ form a path cover of $G$
contradicting the minimality of $|ter(\mathcal{P})|$ again.
\end{proof}
\begin{theorem} (3.1.3) A graph is 2-connected iff it can be
constructed from cycles by successively adding $H$-paths to graphs
$H$ already connected.
\end{theorem}
\begin{proof} Note that any graph constructed as above must clearly be
$2$-connected. For the converse, let $G$ be a $2$-connected
graph. Since $G$ is $2$-connected it contains a cycle and thus $G$
has a maximum sized subgraph $H$ such that $H$ can be constructed by
successively adding $H$-paths to a cycle. Note that $H$ must be an
induced graph, because any edge not in $H$, but between vertices of
$H$ is an $H$-path. Finally, form connectivity there must exists an
edge $vw$ such that $v \in V(H)$ and $w \notin V(H)$. Since $G$ is
$2$-connected there must be a path $P$ from $w$ to $H$ in $G-v$, but
then $vwP$ is an $H$ path, contradicting the maximality of $H$.
\end{proof}
\begin{theorem} (3.2.1) If $G$ is $3$-connected and $|G| > 4$, then
$G$ has an edge $e$ such that $G/e$ is again $3$-connected.
\end{theorem}
\begin{proof} Assume there exists no edge $e$ such that $G/e$ is $3$
connected. Then for any edge $xy$, there must exists a vertex $z$
such that $\{v_{xy}, z\}$ is a separator in $G/e$, and therefore
${x,y,z}$ must be a separator in $G$ since $G$ is
$3$-connected. Pick an edge $xy$ such that the separator $\{x,y,z\}$
produces a component $C$ of minimum size. Since $\{x,y,z\}$ is a
minimum separator, there must exists a vertex
$v \in N(z) \cap V(C)$, because otherwise $\{x,y\}$ would be a
separator for $C$. Let $w$ be the vertex associated with the edge
$zv$ so that $\{z, v, w\}$ is a minimum separator in $G$. Since $xy$
is an edge, there must exists a component $D$ of $G - \{z,v,w\}$
such that $V(D) \cap \{x,y\} = \emptyset$, since $x$ and $y$ are not
separated by $\{z, v, w\}$. Also, since
$N(v) \cap V(D) \subseteq V(C)$, we get that $D$ is contained in
$C$, contradicting the minimality of $C$.
\end{proof}
\begin{theorem} (3.2.2) A graph $G$ is $3$-connected iff there exists
a sequence $G_0, \ldots, G_n$ of graph with the following
properties:
\begin{itemize}
\item[(i)] $G_0 = K^4$ and $G_n = G$
\item[(ii)] There exists an edge $xy \in G_{i+1}$ with
$d(x), d(y) \geq e$ and $G_i = G_{i + 1} /xy$ for every $i < n$.
\end{itemize}
\end{theorem}
\begin{proof} Proof follows immediately from $3.2.1$.
\end{proof}
\begin{theorem} (3.2.3) The cycle space of a $3$-connected graph is
generated by its non-separating induced cycles.
\end{theorem}
\begin{proof} Let $G$ be a $3$-connected graph with $|G| = n$. We will
show that each cycle $C$ is a sum of non-separating induced cycles
by induction on $k(C) = n - b$ where $b$ is the order of the largest
component in $G-C$, and $b = 0$ if $V(G) = V(C)$. Since there are no
cycles with $k(C) = 0$, the base case trivially holds. Now let $C$
be a cycle assume the inductive hypothesis holds for all cycles $C'$
with $k(C') < k(C)$.
To begin with, if $C$ is a spanning cycles, then it contains a chord
$e$ because $G$ is $3$-connected. But then $C$ is the sum of two
non-spanning cycles $C_1,C_2 \subseteq C + e$ with $e \in C_1,
C_2$. Since $k(C_1),k(C_2) < n = k(C)$ due to the largest component
of $G -C_i$ having order at most $n - |C_i|$, we are done by the
inductive hypothesis.
Assume that $G-C$ is non-empty and let $B$ be the largest component
of $G-C$. Suppose there a $C$-path $P$ from $u$ to $v$ in $G-B$ such
that both $uCv$ paths contain an internal vertex in $N(B)$. Then
there are two cycles $C_1,C_2 \subset C \cup P$ with
$P \subset C_1, C_2$ such that $C$ is the sum of $C_1$ and
$C_2$. Also, there is a component in $G - C_i$ properly containing
$B$, so we are done by induction.
Assume there is no such path $P$. Then all vertices $v$ have a
neighbor in $B$. Assume to that contrary and let $v$ be a vertex
with $v \notin N(B)$. Let $x,y$ be $v$'s two closest neighbors with
$x,y \in N(B)$ and let $Q$ be the path from $x$ to $y$ on $C$
passing through $v$. Since $G$ is $3$-connected, $C-Q$ is nonempty
and there must be a path $P$ from $Q-x-y$ to $C-Q$ in $G - x -
y$. But this path has the property described above, a
contradiction. Thus all vertices in $C$ contain a neighbor in
$B$. Finally note that $C$ cannot contain a chord since that chord
forms one of the banned paths.
Thus $C$ is an induced cycles. If $C$ is non-separating then we are
trivially done. Otherwise there exits another component $B' \neq B$
in $G-C$. Let $P = u, \cdots v$ be a $C$-path through $B'$ and let
$Q$ be a $C-P$ path. Then there exists cycles
$C_1, C_2, C_3 \subset C \cup P \cup Q$ which sum to $C$. As before
$G - C_i$ has a component properly containing $B$, so we are done by
the inductive hypothesis.
\end{proof}
\begin{theorem} (3.3.1 Menger's Theorem) Let $G=(V,E)$ be a graph and
$A,B \subseteq V$. Then the minimum number of vertices separating
$A$ form $B$ in $G$ is equal to the maximum number of disjoint
$A$-$B$ paths in $G$.
\end{theorem}
\begin{proof} The proof is by induction. If $||G|| = 0$, then the
number of $A$-$B$ paths must equal $|A \cap B|$ since there are
$|A \cap B|$ trivial $A$, $B$ paths. Otherwise we have an edge
$e = xy$. Let $k$ be the size of the minimum separator separating
$A$ and $B$. Consider the graph $G / e$. If $v_c$ is the vertex
created by contraction, let $v_c \in A$ if $x$ or $y$ was in $A$
(and similarly for $B$). Then $G/e$ contains a minimum separator
$Y$. If $|Y| \geq k$, then we are done by induction. Thus
$|Y| \geq k$. Note that $v_c \in Y$, otherwise $Y$ would be a
separator of $G$, a contradiction. Define $X = Y - v_c + x +
y$. Then $X$ is an $A,B$ separator in $G$ on exactly $k$
vertices. Now consider $G-e$. By the inductive hypothesis there must
be $k$ disjoint $A-X$ paths and $k$ disjoint $B-X$ paths. These sets
of paths must be disjoint as well since $X$ is a
separator. Combining this paths produces $k$ disjoint $A$-$B$ paths.
\end{proof}
\begin{theorem} (3.3.6 Menger's Theorem (global version))
\begin{itemize}
\item A graph is $k$-connected iff it contains $k$-independent paths
between any two vertices.
\item A graph is $k$-edge connected iff it contains $k$ edge
disjoint paths between any two vertices.
\end{itemize}
\end{theorem}
\begin{proof} Follows immediately from 3.3.1.
\end{proof}
\begin{theorem} (3.4.1 Maders' Theorem) Given a graph $G$ with an
induced subgraph $H$, there are always $M_G(H)$ independent
$H$-paths in $G$.
\end{theorem}
\begin{proof} No proof given in Diestel
\end{proof}
\begin{theorem} (3.4.2) Given a graph $G$ with an induced subgraph
$H$, there are at least $1/2 \kappa_G (H)$ independent $H$-paths and
at least $1/2 \lambda_G(G)$ edge disjoint $H$-paths in $G$.
\end{theorem}
\begin{proof}
\end{proof}
\begin{theorem} (3.5.1) There exists a function
$h: \mathbb{N} \to \mathbb{N}$ such that every graph of average
degree at least $h(r)$ contains $K^r$ as a topological minor for
every $r \in \mathbb{N}$.
\end{theorem}
\begin{proof} The proof will be by induction. To begin with, note that
$h(r) = 1$ holds for $r \leq 2$. We will show by induction on
$m = r \cdots {r \choose 2}$ that if $G$ has average degree $2 ^ m$,
then $G$ has a topological copy of $X$ where $X$ has $r$ vertices
and $m$ edges. When $m = {r \choose 2}$ this will immediately imply
the claim. If $m = r$, then $G$ contains a cycle of length at least
$\epsilon(G) + 1 \geq 2^{r -1} + 1 \geq r + 1$. Letting $X = C^r$
suffices. Now assume $m > r$ and the claim holds for any smaller
value. If $G$ is not connected, then there exists a component of $G$
with at least the same average degree, and we are done by
induction. Thus $G$ is connected. Let $U$ be a maximum set of
vertices such that $U$ is connected and
$\epsilon(G \backslash U) \geq 2 ^ {m - 1}$. Note such a set exists
because deleting a vertex cannot decrease the average degree of $G$
to below $2 ^ {m - 1}$. Also, since $G$ is connected, $N(U)$ is
non-empty. Let $H = G[N(U)]$. Not $\delta(H) \geq 2 ^ {m -1}$,
otherwise we could add another vertex to $U$, contradicting $U$'s
maximality. Thus by the induction hypothesis, $H$ contains a $TY$,
with $|Y| = r$, and $||Y|| = m - 1$. Let $x,y$ be two non-adjacent
branch vertices. Since $G$ is connected there is a $x-y$ path in $U$
creating a $TX$ on $m$ vertices.
\end{proof}
\begin{theorem} (3.5.2) There is a function
$f: \mathbb{N} \to \mathbb{N}$ such that every $f(k)$-connected
graph is $k$-linked, for all $k \in \mathbb{N}$.
\end{theorem}
\begin{proof} Let $G$ be $h(3k) + 2k$ connected, where $h$ is as
defined in the previous Lemma. Let $X$ be the set of vertices
$s_1 \cdots s_k$ and $t_1 \cdots t_k$ we would like to connect. Then
$G - X$ contains $K$ a copy of $TK^{3k}$ since
$\kappa(G) \geq h(3k) \geq \delta(G)$. Let $U$ be the branch
vertices of a $K$. Since $G$ is $2k$ connected there are $2k$
independent paths from $s_1 \cdots s_k$ and $t_1 \cdots t_k$ to
$U$. Let $P$ be a set of paths minimizing the edges outside of
$K$. Label the non-ending vertices as $u_i$. Let $s'_i$ denote the
end the path $P_i$ starting at $s_i$ in $U$ and $t'_i$ denote the
end of $t_i$ in $U$. Note that $s'_i u_i$ in $K$ can only contain
vertices on $P_i$, by minimality of $P$. Thus there exists a path
from $s_i$ to $u_i$ that is disjoint from all other paths in
$P$. Similarly there is a path from $t_i$ to $u_i$, show that $G$ is
$k$-linked.
\end{proof}
\begin{theorem} (4.2.6) In a two connected plane graph, every face is
bounded by a cycle.
\end{theorem}
\begin{proof} Let $G$ be a two connected plane graph of minimum order
which contradicts the proposition. Since $G$ is 2-connected, it can
be constructed by adding $H$-paths, thus there exists graph $H$ and
an $H$-path $P$ such that $G = H \cup P$. By the minimality of $G$,
every face in $H$ is bounded by a cycles, so $V(P) \subseteq
F$. Since $P$ only effects $1$ face in $G$, we may assume that
$H = K_3$. There is obviously no way connect an $H$-path to $K_3$
invalidating the theorem. Thus theorem is true.
\end{proof}
\begin{theorem} (4.2.7) The face boundaries in a $3$-connected plane
graph are precisely its non-separating induced cycles.
\end{theorem}
\begin{proof} Let $C$ be a non-separating induced cycle, then $C$ must
obviously bound a face of $G$. For the converse. Let $f$ be a face
in $G$ and let $C$ be the cycle bounding $f$. If $C$ has a chord
$xy$, then the components of $C-\{x, y\}$ are linked by a $C$ path
in $G$, because $G$ is $3$ connected. This path must run through
$f$, a contradiction. Thus $C$ is induced. Now it remains to show
that $C$ does not separate any two vertices in $G - C$. By Menger's
theorem, there exists $3$ vertex independent paths between any two
vertices $x$ and $y$. Clearly $f$ lies inside a face of the union of
these paths. This path is bounded by only two of the paths, thus the
third path must avoid $C$ entirely.
\end{proof}
\begin{theorem} (4.2.8) A plane graph of order at least 3 is maximally
planar iff it s a plane triangulation.
\end{theorem}
\begin{proof} Let $G$ be a plane graph of order at least 3. Obviously
any plane triangulation must be maximally planar, since any edge
added must be added within a face. To show the other direction, let
$G$ by maximally planar and pick a face $F$. Note that $G[F]$ must
be complete, otherwise we could add an edge between the two
unconnected elements in $F$ through $F$ to find a larger planar
graph. Since $F$ is planar, this implies that $|F| \leq 3$. If
$|F| < 3$, then there exists only one face in $G[F]$, if there is an
edge from $F$ to another vertex, that vertex must also be in $F$
because there is now way for that edge to close off $F$ from a new
space. Thus if $|F| < 3$ there must be no other edges in the graph
contradicting $|G| \geq 3$. Thus $G[F] = K_3$, proving the claim.
\end{proof}
\begin{theorem} (4.2.9) Let $G$ be a connected planar graph with $n$
vertices, $m$ edges, and $f$ faces. Then $n - m + f = 2$.
\end{theorem}
\begin{proof} The proof is by induction on $|m|$. For the base case,
consider $n = m + 1$, in which case $G$ must be a tree since $G$ is
connected. Thus there is exactly one face yielding the desired
inequality. Otherwise $m > n - 1$. Then $G$ must contain a
cycle. Let $C$ be a cycle of minimum length and delete an edge from
the cycle, to obtain a graph $G'$. $G'$ contains exactly 1 fewer
edges and 1 fewer faces thus $n - (m - 1) + (f - 1) = 2$ which
immediately implies out conclusion.
\end{proof}
\begin{theorem} (4.4.2) A graph contains $K^5$ or $K_{3,3}$ as a minor
iff it contains $K^5$ or $K_{3,3}$ as a topological minor.
\end{theorem}
\begin{proof} The reverse direction is obvious since all topological
minors are also minors. To show that forward direction, if $G$
contains a $K_{3,3}$, it obviously contains a $TK_{3,3}$. So assume
$G$ contains a $K^5$ minor. Let $K$ be a minimal $K^5$ minor in
$G$. Then every branch set of $K$ induces a tree in $K$, and between
any two branch sets, $K$ has exactly one edge. Let $T_x$ be the tree
on vertex set $V_x$, including the edges connecting $V_x$ to the
other branch sets. From minimality, there are exactly $4$ such
edges, and $4$ leafs in $T_x$. If each of the $5$ treas from a
$TK{1,4}$, then we have found a $TK_5$. If $T_x$ is not a
$TK_{1,4}$, then it has exactly of vertices of degree
$3$. Contracting $V_x$ into these vertices and the other branch sets
into a vertex yields a graph on $6$ vertices containing a
$TK_{3,3}$, Thus $G$ contains a $TK_{3,3}$.
\end{proof}
\begin{theorem} (4.4.3) Every $3$-connected graph without a $K_5$ or
$K_{3,3}$ topological minor is planar.
\end{theorem}
\begin{proof} Proof is by induction. Note this applies when
$|G| \leq 4$. Otherwise, let $G$ be a planar graph on at least $5$
vertices. Since $G$ is $3$-connected, there exists an edge $xy$ such
that $G'=G/xy$ is $3$-connected. By the inductive hypothesis this
graph has a planar drawing. Consider the face $f$ containing
$v_{xy}$ in $G'-v_{xy}$, and let $C$ be the boundary of $f$. Note
that $C$ must be a cycle since $G'$ must be $3$-connected. Let
$X = \{ x_1, \cdots x_k\}$ be $N(x) \cap V(C)$ where the vertices
are enumerated in order around $C$. Define $Y$ to be
$N(y) \cap V(C)$. Let $P_i$ be the path on $C$ from $x_i$ to
$x_{i + 1}$.
Note that $N(y) \subseteq P_i$ for some $i$. Assume to the contrary,
then there exists vertices $y'$ and $y''$ not contained in the same
$P_i$, and these two vertices are separated by two vertices $x'$ and
$x''$ on $C$. Pick $y'$, $y''$, $x'$, $x''$ such that the number of
distinct vertices in maximized. If every vertex is distinct, then we
get a $TK_{3,3}$ with partition $(\{x, y', y''\}, \{y, x',
x''\})$. Because $G$ is $3$-connected, the fewest distinct vertices
possible is $3$, and this is only achieved if $N(X) = N(Y)$ with
$|N(X)| = 3$. In this case though, $N(X)$ along with $x$ and $y$
form a copy of $TK^5$. Thus $N(Y) \subseteq V(P_i)$ for some
$i$. But then we can construct a planar drawing of $G$ by starting
with a planar drawing of $G-y$ and placing $y$ and its edges inside
the face containing $x$ and $P_i$.
\end{proof}
\begin{theorem} (4.4.6) If $G$ is a graph, then the following are
equivalent:
\begin{itemize}
% \item $G$ is planar
\item $G$ contains no $K_5$ or $K_{3,3}$ minor.
\item $G$ contains not topological $K_5$ or $K_{3,3}$ minor.
\end{itemize}
\end{theorem}
\begin{proof} The proof immediately follows from 4.4.3 and 4.4.2, and
the observation that all maximally planar graphs on at least $4$
vertices are $3$-connected.
\end{proof}
\begin{theorem} (5.1.2) Every planar graph is $5$-colorable.
\end{theorem}
\begin{proof} The proof is the proof that every planar graph is 5
chooseable (5.4.2).
\end{proof}
\begin{theorem} (5.2.4 Brooks Theorem) Let $G$ be connected graph. If
$G$ is neither complete, nor an odd cycle, then
$\chi(G) \leq \Delta(G)$.
\end{theorem}
\begin{proof} Note that the statement is obviously true if
$\Delta(G) \leq 2$ since in that case $G$ is either a path or a
cycle. Let $G$ be a graph minimizing $|G|$ which is a counter
example to the theorem. Note that $G$ is not a complete graph. Pick
any vertex $x \in G$. Now consider the graph $G-x$ and note that
$\chi(G-x) \leq \Delta(G)$. This follows since if $G - x$ is a
complete graph or odd cycle, then $\Delta(G -x) < \Delta(G)$ since
$G$ is connected and otherwise $\Delta(G - x) \leq \Delta(G)$. Color
$G - x$ with $\Delta(G)$ colors. Since $G-x$ cannot be colored with
$\Delta(G)$ colors, the $N(x)$ must contain at least $\Delta(G)$
colors. Let $N(x) = \{v_1, v_2, \cdots v_{\Delta(G)} \}$ where $v_i$
has color $i$ in $f$.
Define $H_{i,j}$ be component of the graph induced by vertices with
colors $i$ and $j$ in the coloring of $G - x$. Note that $v_i$ and
$v_j$ must appear in the same component in $C_{i,j}$, otherwise, in
one of the components we could swap the colors to obtain a coloring
$f'$ in which $v$ and $w$ have the same color, allowing us to color
$x$ with either color $i$ or $j$.
We will show that $C_{i,j}$ must be a path. Otherwise there exists a
vertex $u$ which is as close to $v$ as possible which has degree at
least $3$. Then in $G-x$, $N(u)$ can contain at most $\Delta(G) - 1$
colors since 3 of the vertices have at most $2$ colors. Thus we can
swap the color of $u$ to the missing color to create a coloring
$f'$. Since $u$ was the color closest to $v$, in $f'$ $C_{i,j}$ $v$
and $w$ must now be in different components (if both $u$ and $v$ are
even in $C_{i,j}$ anymore), which will allow us to color $x$ once
again. Thus $C_{i,j}$ must be a path.
Now consider two such paths $P_{i,j}$ and $P_{j,k}$ between vertices
$v_i$, $v_j$ and $v_j$ and $v_k$. Then
$V(P_{i, j}) \cap V(P_{j,k}) = v_j$. Otherwise there exists a vertex
$u \in V(P_{i, j}) \cap V(P_{j,k})$. The vertex $u$ satisfies that
$4$ of its neighbors get $2$ colors. Thus $N(u)$ contains at most
$\Delta(u) - 2$ colors. Thus $u$ could be recolored creating the
same situation as described previously.
Now finally we will use the previous properties to construct a
coloring of $G$. Note that $G$ must not be a complete graph, thus
there exists a vertex $x$ who's neighborhood is not complete. Let
$f$ be a coloring of $G-x$ such that $v_1v_2$ is not an edge in
$G$. Now construct a new coloring $f'$ by swapping the colors $1$
and $3$ on $P_{1,3}$. Consider the vertex $u$ which is the neighbor
of $v_1$ on $P_{1,2}$. Since no vertices of color $2$ were changed,
$u$ must be the neighbor of $v_1$ in $P'_{2, 3}$ or else we are
finished by the above properties. Also, since
$P_{1,3} \cap P_{1,2} = v_1$, $u$ must be in $P'_{1,2}$ as well or
we are finished by the above properties again. Finally, note that
$u$ contradicts the final property allowing us to construct a proper
coloring anyways.
\end{proof}
\begin{theorem} (Mycelskis construction) There exists a triangle free
graph with chromatic number $k$ for any integer $k$.
\end{theorem}
\begin{proof} The proof is by induction. For $k \leq 2$, the problem
is trivial. Let $G$ be a $k$ chromatic triangle free
graph. Construct a new graph $G'$ on $V(G)$, $U$ and $\{w\}$, where
$G'[V(G)] = G$, $v_i u_j$ is an edge iff $v_i v_j$ is an edge in
$G$, and where $N(w) = U$. Note that after doing this, $G'$ is still
triangle free. Note that a coloring of $V(G)$ can be extended to $U$
to obtain a proper coloring of $G'$. Thus
$\chi(G') \leq \chi(G) + 1$. Now it suffices to show that
$\chi(G) < \chi(G')$. Let $G$ be a proper color of $G'$ on fewer
than $\chi(G) + 1$ colors. We may assume that $G(w) = k$. Thus there
are $k-1$ colors on $U$. Let $A$ be the set of vertices on $G$ with
color $k$. For each vertex in $v_i \in A$, change the color of $v_i$
to $u_i$. Since $v_i$ and $u_i$ the same neighbors (excluding $w$),
this gives a proper coloring of $G$ on fewer than $\chi(G)$ colors,
a contradiction.
\end{proof}
\begin{definition} A graph $G$ is $k$-constructible if it is
isomorphic to one of the following:
\begin{itemize}
\item[(i)] $K^k$.
\item[(ii)] $(G' + xy)/xy$ where $G'$ is a $k$-constructible graph
with vertices $x$ and $y$ that are non-adjacent.
\item[(iii)] $(G_1 \cup G_2)-xy_1 - xy_2 + y_1y_2$ where $G_1$ and
$G_2$ are $k$-constructible with $V(G_1) \cap V(G_2) = x$ and
$xy_i \in G_i$.
\end{itemize}
\end{definition}
\begin{theorem} (5.2.6) Let $G$ be a graph $k \in \mathbb{N}$. Then
$\chi(G) \geq k$ iff $G$ has a $k$-constructible subgraph.
\end{theorem}
\begin{proof} The proof of the forward direction is done by induction
on the minimum number of steps to construct $G$ using the
$k$-constructible constructions. Let $G$ be a $k$-constructible
graph. If $G = K^k$, then obviously $\chi(G) \geq k$. Otherwise if
the last step to construct $G$ was such that $G = G' + xy /xy$, then
a coloring of $G'$ immediately induces a color on $G$ by setting the
color of $x$ and $y$ to be the same as the contracted vertex and
leaving all other vertices the same color. Finally, if the last step
to construct $G$ results form $G = G_1 + G_2 -xy_1 - xy_2 + y_1y_2$,
then a coloring of $G$ forces the color of $x$ to be 1different from
one of $y_1$ or $y_2$. Thus the coloring of $G$ induces a coloring
on $G_1$ or $G_2$, thus $\chi(G) \geq k$.
For the reverse direction, let $G$ be an edge maximal graph with
$\chi(G) = k$ without a $k$-constructible subgraph. Note $G$ is not
complete $r$-partite, because otherwise $G$ would have to contain
$K^k$. Thus there exists an edge $y_1 y_2$ and vertex $x$ such that
$xy_1$ and $xy_2$ is not an edges. Note $G + x y_i$ must contain a
$k$-constructible subgraph $H_i$ because $G$ is edge maximal. Let
$H'_2$ be an isomorphic copy of $H_2$ such that
$H'_2 \cap H_1 = \{x\}$, $|V(H'_2) \cap V(H_2)|$ is maximized, and
all vertices in $H'_2 - H_2$ are not in $G$. Then
$H_1 \cup H'_2 - xy_1 - xy'_2 + y_1y'_2$ is $k$-constructible. Now
for every vertex $v' \in H'_2 - G$, let $v$ be the corresponding
vertex in $H_2$ construct another $k$-constructible graph by adding
and then contracting the edge $v'v$ to create another
$k$-constructible graph. This new graph $H$ is then a
$k$-constructible subgraph of $G$, a contradiction.
\end{proof}
\begin{theorem} (5.3.2 Vizing's Theorem) Every graph $G$ satisfies
$\Delta(G) \leq \chi'(G) \leq \Delta(G) + 1$.
\end{theorem}
\begin{proof} Note that obviously $\Delta(G) \leq \chi'(G)$. So
consider the inequality of $\chi'(G) \leq \Delta(G) + 1$. This
inequality is true when $G$ is the empty graph. Let $G$ be the
counter example minimizing $||G||$. Note that for any proper edge
coloring of $G-xy'$ with $xy' \in G$, there must be some color
$\alpha$ missing among edges ending at $x$ and some color $\beta$
missing among edges ending at $y'$. Now consider a maximal
$\alpha \beta$ path $P$ starting at $y'$. Then this path must end at
$x$, otherwise we could swap the coloring on the path $P$ and color
the edge $xy'$ with $\alpha$ to get a proper coloring of $G$. Let
$x$ be a vertex in $G$, and $xy_0$ an edge. Then $G-xy_0$ has an
edge coloring $f_0$ using at most $\Delta(G) + 1$ colors. Let
$y_0, y_1, \cdots y_k$ be a maximal sequence of the neighbors of $x$
such that the color of the edge $x y_{i + 1}$ is missing among edges
ending at $y_i$. Note that for any edge $x y_j$ we can create a
coloring $f_j$ for $G-x y_j$ based on the coloring $f_0$ by
assigning $x y_i$ the color of $x y_{i + 1}$ in $f$ for $i < j$ and
keeping all other colors the same. Let $\beta$ be the coloring
missing at $y_k$. Then the maximal $\alpha$-$\beta$ path must start
at $y_k$ and end at $x$. Let $y$ be the vertex preceding $x$ on this
path. From the maximality of the sequence $y_0, y_1, \cdots y_k$,
$y = y_j$ for $j < k$ since $x y$ has color $\beta$. Now consider
the coloring $f_j$ of $G - xy$. Then the $\alpha-\beta$ path
starting at $y_k$ must end at $y$. Thus the colors on this path
could be swapped so that $xy_j$ can be colored with $\alpha$ to
obtain a proper edge coloring of $G$.
\end{proof}
\begin{theorem} (5.4.2) Every planar graph is $5$-chooseable.
\end{theorem}
\begin{proof} Proof by induction, but with the extra hypothesis that
every connected planar graph whose external vertices form a cycle
can be list colored if the list on all interior vertices has size at
least $5$ and all exterior vertices has size at least $3$ and two
adjacent exterior vertices $v_1$ and $v_2$ are already colored. This
is sufficient because one can always add edges to a planer graph $G$
to form a cycle, unless $|G| < 3$ in which case the claim is
trivially true.
For the base case, note the new claim trivially holds when
$|G| \leq 3$. Now let $G$ be a connected planar graph with exterior
vertices in a cycle $C = v_1 \cdots v_k$ and assign $G$ a lists such
that it satisfies the inductive hypothesis. To begin with, assume
that $G$ has a cord $vw$. Then we can split $G$ into two planar
graphs, one of which contains $v_1$ and $v_2$ and the other which
doesn't. By induction we can color the graph containing $v_1$ and
$v_2$. Now setting the colors of $v$ and $w$ as found in the last
coloring we can again by our inductive hypothesis color the other
cycle. Otherwise for the vertex $v_k$, reserve two colors which are
not the color of $v_1$ and delete these colors from its interior
neighbors. Then by the inductive hypothesis we can color
$G-v_k$. Since $v_{k-1}$ is the only neighbor of $v_k$ which could
have one of the colors reserved for $v_k$ we can still color $v_k$.
\end{proof}
\begin{theorem} (5.4.3) Let $H$ be a graph and ${(S_v)}_{v \in V(H)}$
a family of lists. If $H$ has an orientation $D$ with
$d^+(v) < |S_v|$ for every $v$, and such that every induced subgraph
of $D$ has a kernel, then $H$ can be colored from the lists $S_v$.
\end{theorem}
\begin{proof} The proof is by induction on $|G|$. If $|G| = 0$, then
the claim is obviously true. Otherwise let $G$ be a graph satisfying
the hypothesis. Let $\alpha$ be a color and let $T$ be the set of
vertices such that the color $\alpha$ is a possible color. Then the
graph $G[T]$ has a kernel $U$. Coloring all the vertices of $U$ with
the color $\alpha$ and deleting $U$ from $G$ and $\alpha$ from the
list of all other vertices in $T$ then creates a smaller graph still
satisfying the inductive hypothesis, thus completing the proof by
induction.
\end{proof}
\begin{theorem} (5.4.4) Every bipartite graph $G$ satisfies
$ch'(G) = \chi'(G)$.
\end{theorem}
\begin{proof} Let $G$ be a bipartite graph with partition $X,Y$ and
let $c$ be an edge coloring of $G$ in $\chi'(G)$ colors. Place an
ordering on the colors and let $H$ be the line graph of $G$ with a
direction applied to the edges as follows: Let $e_1$ and $e_2$ be
intersecting edges with $c(e_1) < c(e_2)$. If $e_1e_2$ meet in $x$
direct the edge from $e_1$ to $e_2$ and if $e_1e_2$ meet in $Y$,
direct the edge from $e_2$ to $e_1$. Now for a given edge $e$, the
out edges of $e$ that meet in $y$ must have colors strictly larger
than $c(e)$, and the color for edges meeting in $X$ must have color
strictly less than $c(e)$. Since any two neighbors of $e$ meeting in
$X$ or $Y$ are also neighbors, we get that $d^+(e) < k$. Now it
suffices to show that every induced subgraph of $H$ contains a
kernel. Let $Z$ be a subset of $X \cup Y$ and consider $G[Z]$. Let
the edge colors represent preferences. Then we can find a maximum
stable matching given these preferences. This stable matching is an
independent set $S$ in $H$. Since it is maximum, all out edges of a
vertex in $H$ must enter $S$, otherwise we would get a larger stable
matching. Thus $S$ is the desired kernel, and by the previous lemma,
$ch'(G) = \chi'(G)$.
\end{proof}
\begin{theorem} (5.5.1) A chordal graph can be constructed recursively
by pasting along complete subgraphs.
\end{theorem}
\begin{proof} Consider a chordal graph $G$. If $G$ is complete, then
we are done, otherwise let $a$ and $b$ two unconnected vertices. Let
$X$ be a minimal $a$-$b$ separator. Note that for any two vertices
$u, v \in X$, there must be a path through the component of $G - X$
containing $a$ and similarly in the component containing $b$. This
creates a cycle of length at least $4$, thus $u,v$ must be an edge
since $G$ is chordal.
\end{proof}
\begin{theorem} Every chordal graph contains a simplicial vertex
\end{theorem}
\begin{proof} The proof is by induction. Note this is true if
$G = K^1$. Now let $x \in V(G)$ such that $G-x$ is connected. Such a
vertex must as exist, otherwise consider a vertex $x$ such that
$G-x$ contains a component $C$ of minimum size. Then for $x' \in C$,
$G-x'$ must contain a smaller component as $x'$ is a cut vertex by
assumption. If $N(x) = V(G) - x$, then $G-x$ has a simplicial vertex
and adding $x$ to $G$ does not change this, so we are done
again. Otherwise $N(x)$ is a minimal vertex cut separating $x$ from
the rest of the graph. Since $G - x$ is connected, there is exactly
one component $C$ in $G - x - N(x)$. Also, by minimality there must
be a $u,v$ $C$-path for all vertex pairs $u,v \in N(x)$. Thus $u,v$
must be an edge since $G$ is chordal. Therefore $x$ is a simplicial
vertex and we are done.
\end{proof}
\begin{theorem} (5.5.2) All chordal graphs are perfect
\end{theorem}
\begin{proof} This proof is by induction. If $|G| = 1$. Otherwise,
every induced subgraph of $G$ is chordal, so we are done by
induction. Thus it just suffices to consider if
$\chi(G) \leq \omega(G)$. Since $G$ is chordal, $G$ can be created
by pasting along a complete subgraph. We can independently color
these two chordal subgraphs, permuting colors if necessary to create
a coloring of $G$. Since the subgraphs are chordal we get that
$\chi(G) \leq \omega(G)$ by induction.
\end{proof}
\begin{theorem} (5.5.4 Lovasz) A graph $G$ is perfect iff $\bar{G}$ is
perfect.
\end{theorem}
\begin{proof} This proof is by induction. Note that $|G| = 1$, then
this is true. Now let $G$ be a graph with $|G| > 1$. By inductions,
for any strict subgraph of $G$, the complement of that subgraph is
perfect, thus from symmetry, all that we need to show is that if $G$
is perfect, then $\chi(\bar{G}) \leq \omega(\bar{G})$. Let
$\mathcal{A}$ be the set of independent sets of size
$\alpha(G) := \alpha$. If $G$ contains a complete subgraph $X$ such
that $X \cap A \neq \emptyset$ for all $A \in \mathcal{A}$, then