-
Notifications
You must be signed in to change notification settings - Fork 3
Expand file tree
/
Copy pathmt242.tex
More file actions
1272 lines (1034 loc) · 46.6 KB
/
mt242.tex
File metadata and controls
1272 lines (1034 loc) · 46.6 KB
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
\frfilename{mt242.tex}
\versiondate{19.11.03}
\copyrightdate{1997}
\def\chaptername{Function spaces}
\def\sectionname{$L^1$}
\newsection{242}
While the space $L^0$ treated in the previous section is of very great
intrinsic interest, its chief use in the elementary theory is as a
space
in which some of the most important spaces of functional analysis are
embedded. In the next few sections I introduce these one at a time.
The first is the space $L^1$ of equivalence classes of integrable
functions. The importance of this space is not only that it offers a
language in which to express those many theorems about integrable
functions which do not depend on the differences between two functions
which are equal almost everywhere. It can also appear as the natural
space in which to seek solutions to a wide variety of integral
equations, and as the completion of a space of continuous functions.
\vleader{72pt}{242A}{The space $L^1$} Let $(X,\Sigma,\mu)$ be any
measure space.
\header{242Aa}{\bf (a)} Let $\eusm L^1=\eusm L^1(\mu)$ be the set of
real-valued functions, defined on subsets of $X$, which are integrable
over $X$. Then
$\eusm L^1\subseteq\eusm L^0=\eusm L^0(\mu)$\cmmnt{, as
defined in \S241}, and, for $f\in\eusm L^0$, we
have $f\in\eusm L^1$ iff there is a $g\in\eusm L^1$ such that
$|f|\leae g$; if $f\in\eusm L^1$, $g\in\eusm L^0$ and $f\eae g$,
then
$g\in\eusm L^1$. \prooflet{(See 122P-122R.)}
\header{242Ab}{\bf (b)} Let $L^1=L^1(\mu)\subseteq L^0=L^0(\mu)$ be
the
set of equivalence
classes of members of $\eusm L^1$. If $f$, $g\in \eusm L^1$ and
$f\eae g$ then $\int f=\int g$\cmmnt{ (122Rb)}. Accordingly we may
define a
functional $\int$ on $L^1$ by writing $\int f^{\ssbullet}=\int f$ for
every $f\in\eusm L^1$.
\header{242Ac}{\bf (c)}\cmmnt{ It will be convenient to be able to
write} $\int_Au$ for $u\in L^1$, $A\subseteq X$\dvro{ is}{; this may
be} defined by saying that
$\int_Af^{\ssbullet}=\int_Af$ for every $f\in\eusm L^1$\cmmnt{,
where
the integral is defined in 214D}. \prooflet{\Prf\ I
have only to check that if $f\eae g$ then $\int_Af=\int_Ag$; and this
is because $f\restr A=g\restr A$ almost everywhere in $A$.\ \Qed}
If $E\in\Sigma$ and $u\in L^1$ then
$\int_Eu=\int u\times(\chi E)^{\ssbullet}$\prooflet{; this is because
$\int_Ef=\int f\times\chi E$ for every integrable function $f$
(131Fa)}.
\spheader 242Ad If $u\in L^1$, there is a $\Sigma$-measurable,
$\mu$-integrable function $f:X\to\Bbb R$ such that $f^{\ssbullet}=u$.
\prooflet{\Prf\ As noted in 241Bk, there is a measurable $f:X\to\Bbb
R$
such that $f^{\ssbullet}=u$; but of course $f$ is integrable because
it
is equal almost everywhere to some integrable function.\ \Qed}
\leader{242B}{Theorem} Let $(X,\Sigma,\mu)$ be any measure space.
Then
$L^1(\mu)$ is a linear subspace of $L^0(\mu)$ and $\int:L^1\to\Bbb R$
is
a linear functional.
\proof{ If $u$, $v\in L^1=L^1(\mu)$ and $c\in\Bbb R$ let
$f$, $g$ be integrable functions such that $u=f^{\ssbullet}$ and
$v=g^{\ssbullet}$; then $f+g$ and $cf$ are integrable, so
$u+v=(f+g)^{\ssbullet}$ and $cu=(cf)^{\ssbullet}$ belong to $L^1$.
Also
\Centerline{$\int u+v=\int f+g=\int f+\int g=\int u+\int v$}
\noindent and
\Centerline{$\int cu=\int cf=c\int f=c\int u$.}
}
\leader{242C}{The order structure of $L^1$} Let $(X,\Sigma,\mu)$ be
any
measure space.
\header{242Ca}{\bf (a)} $L^1=L^1(\mu)$ has an order structure derived
from that of $L^0=L^0(\mu)$\cmmnt{ (241E); that is,
$f^{\ssbullet}\le g^{\ssbullet}$ iff $f\le g$ a.e}. \cmmnt{Being a
linear subspace of
$L^0$,} $L^1$ \dvro{is}{must be} a partially ordered linear
space\cmmnt{; the two
conditions of 241Ec are obviously inherited by linear subspaces}.
\dvro{If}{Note also that if} $u$,
$v\in L^1$ and $u\le v$ then $\int u\le\int v$\prooflet{, because if
$f$, $g$ are
integrable functions and $f\leae g$ then $\int f\le\int g$ (122Od)}.
\header{242Cb}{\bf (b)} If $u\in L^0$, $v\in L^1$ and $|u|\le |v|$
then
$u\in L^1$. \prooflet{\Prf\ Let $f\in\eusm L^0=\eusm L^0(\mu)$,
$g\in\eusm L^1=\eusm L^1(\mu)$ be such that
$u=f^{\ssbullet}$ and $v=g^{\ssbullet}$; then $g$ is integrable and
$|f|\leae |g|$, so $f$ is integrable and $u\in L^1$.\ \Qed}
\header{242Cc}{\bf (c)} In particular, $|u|\in L^1$ whenever $u\in
L^1$,
and
\Centerline{$|\int u|=\max(\int u,\int(-u))\le\int|u|$\dvro{.}{,}}
\cmmnt{\noindent because $u$, $-u\le|u|$.}
\header{242Cd}{\bf (d)} \cmmnt{Because $|u|\in L^1$ for every $u\in
L^1$,
\Centerline{$u\vee v=\Bover12(u+v+|u-v|)$,
\quad$u\wedge v=\Bover12(u+v-|u-v|)$}
\noindent belong to $L^1$ for all $u$, $v\in L^1$. But if $w\in L^1$
we surely have
\Centerline{$w\le u\ \&\ w\le v\iff w\le u\wedge v$,}
\Centerline{$w\ge u\ \&\ w\ge v\iff w\ge u\vee v$}
\noindent because these are true for all $w\in L^0$, so
$u\vee v=\sup\{u,v\}$ and $u\wedge v=\inf\{u,v\}$ in $L^1$.
Thus }$L^1$ is\cmmnt{, in itself,} a Riesz space.
\header{242Ce}{\bf (e)} Note that if $u\in L^1$,
then $u\ge 0$ iff $\int_Eu\ge 0$ for every $E\in\Sigma$\cmmnt{;
this
is because if $f$ is an integrable function on $X$ and $\int_Ef\ge 0$
for every $E\in\Sigma$,
then $f\ge 0$ a.e.\ (131Fb)}. \dvro{If}{More generally, if} $u$, $v\in
L^1$ and $\int_Eu\le\int_Ev$ for every $E\in\Sigma$, then $u\le v$.
\dvro{If}{It follows at once that
if} $u$, $v\in L^1$ and $\int_Eu=\int_Ev$ for every $E\in\Sigma$, then
$u=v$\cmmnt{ (cf.\ 131Fc)}.
\spheader 242Cf If $u\ge 0$ in $L^1$, there is a non-negative
$f\in\eusm L^1$ such that $f^{\ssbullet}=u$\cmmnt{ (compare
241Eg)}.
\leader{242D}{The norm of $L^1$} Let $(X,\Sigma,\mu)$ be any measure
space.
\header{242Da}{\bf (a)} For $f\in\eusm L^1=\eusm L^1(\mu)$ I write
$\|f\|_1=\int|f|\in\coint{0,\infty}$.
For $u\in L^1=L^1(\mu)$ set $\|u\|_1=\int|u|$, so that
$\|f^{\ssbullet}\|_1=\|f\|_1$ for every $f\in\eusm L^1$.
Then $\|\,\|_1$ is a norm on $L^1$. \prooflet{\Prf\ (i) If $u$,
$v\in L^1$ then $|u+v|\le|u|+|v|$, by 241Ee, so
\Centerline{$\|u+v\|_1=\int|u+v|
\le\int|u|+|v|=\int|u|+\int|v|=\|u\|_1+\|v\|_1$.}
\noindent (ii) If $u\in L^1$ and $c\in\Bbb R$ then
\Centerline{$\|cu\|_1=\int|cu|=\int|c||u|=|c|\int|u|=|c|\|u\|_1$.}
\noindent (iii) If $u\in L^1$ and
$\|u\|_1=0$, express $u$ as $f^{\ssbullet}$, where $f\in\eusm L^1$;
then $\int|f|=\int|u|=0$. Because $|f|$ is non-negative, it must be
zero almost everywhere (122Rc), so $f=0$ a.e.\ and $u=0$ in $L^1$.\
\Qed}
\header{242Db}{\bf (b)}\cmmnt{ Thus} $L^1$, with $\|\,\|_1$, is a
normed space and $\int:L^1\to\Bbb R$ is a linear operator;
\cmmnt{observe that} $\|\int\|\le 1$\prooflet{, because
\Centerline{$|\int u|\le\int|u|=\|u\|_1$}
\noindent for every $u\in L^1$}.
\header{242Dc}{\bf (c)} If $u$, $v\in L^1$ and $|u|\le|v|$, then
\Centerline{$\|u\|_1=\int|u|\le\int|v|=\|v\|_1$.}
\noindent In particular, $\|u\|_1=\||u|\|_1$ for every $u\in L^1$.
\header{242Dd}{\bf (d)} \dvro{If}{Note the following property of the
normed Riesz space $L^1$: if} $u$, $v\in L^1$ and $u$, $v\ge 0$, then
\Centerline{$\|u+v\|_1\prooflet{\mskip5mu=\intop
u+v
=\intop u+\intop v}=\|u\|_1+\|v\|_1$.}
\spheader 242De The set $(L^1)^+=\{u:u\ge 0\}$ is closed in $L^1$.
\prooflet{\Prf\ If $v\in L^1$, $u\in(L^1)^+$ then
$\|u-v\|_1\ge\|v\wedge 0\|_1$; this is because if $f$,
$g\in\eusm L^1$ and $f\ge 0$ a.e.,
$|f(x)-g(x)|\ge|\min(g(x),0)|$ whenever $f(x)$ and $g(x)$ are both
defined and $f(x)\ge 0$, which is almost everywhere, so
\Centerline{$\|u-v\|_1=\int|f-g|\ge\int|g\wedge\tbf{0}|
=\|v\wedge 0\|_1$.}
\noindent Now this means that if $v\in L^1$ and $v\not\ge 0$, the
ball
$\{w:\|w-v\|_1<\delta\}$ does not meet $(L^1)^+$, where
$\delta=\|v\wedge 0\|_1>0$ because $v\wedge 0\ne 0$. Thus
$L^1\setminus(L^1)^+$ is open and $(L^1)^+$ is closed.\ \Qed}
\leader{242E}{}\cmmnt{ For the next result we need a variant of
B.Levi's theorem.
\medskip
\noindent}{\bf Lemma} Let $(X,\Sigma,\mu)$ be a measure space and
$\sequencen{f_n}$ a sequence of $\mu$-integrable real-valued functions
such that $\sum_{n=0}^{\infty}\int|f_n|<\infty$. Then
$f=\sum_{n=0}^{\infty}f_n$ is integrable and
\Centerline{$\int f=\sum_{n=0}^{\infty}\int f_n$,
\quad$\int|f|\le\sum_{n=0}^{\infty}\int|f_n|$.}
\proof{{\bf (a)} Suppose first that every $f_n$ is non-negative.
Set $g_n=\sum_{k=0}^nf_k$ for each $n$; then $\sequencen{g_n}$ is
increasing a.e.\ and
\Centerline{$\lim_{n\to\infty}\int g_n
=\sum_{k=0}^{\infty}\int f_k$}
\noindent is finite, so by B.Levi's theorem (123A)
$f=\lim_{n\to\infty}g_n$ is integrable and
\Centerline{$\int f=\lim_{n\to\infty}\int g_n
=\sum_{k=0}^{\infty}\int f_k$.}
\noindent In this case, of course,
\Centerline{$\int|f|=\int f
=\sum_{n=0}^{\infty}\int f_n
=\sum_{n=0}^{\infty}\int|f_n|$.}
\medskip
{\bf (b)} For the general case, set $f_n^+={1\over 2}(|f_n|+f_n)$,
$f_n^-={1\over2}(|f_n|-f_n)$, as in 241Ef; then $f_n^+$ and
$f_n^-$ are non-negative integrable functions, and
\Centerline{$\sum_{n=0}^{\infty}\int f_n^+
+\sum_{n=0}^{\infty}\int f_n^-
=\sum_{n=0}^{\infty}\int|f_n|<\infty$.}
\noindent So $h_1=\sum_{n=0}^{\infty}f_n^+$ and
$h_2=\sum_{n=0}^{\infty}f_n^-$ are both integrable. Now
$f\eae h_1-h_2$, so
\Centerline{$\int f=\int h_1-\int h_2
=\sum_{n=0}^{\infty}\int f_n^+
-\sum_{n=0}^{\infty}\int f_n^-
=\sum_{n=0}^{\infty}\int f_n$.}
\noindent Finally
\Centerline{$\int|f|\le\int|h_1|+\int|h_2|
=\sum_{n=0}^{\infty}\int f_n^+
+\sum_{n=0}^{\infty}\int f_n^-
=\sum_{n=0}^{\infty}\int|f_n|$.}
}
\leader{242F}{Theorem} For any measure space $(X,\Sigma,\mu)$,
$L^1(\mu)$ is
complete under its norm $\|\,\|_1$.
\proof{ Let $\sequencen{u_n}$ be a sequence in
$L^1$ such that $\|u_{n+1}-u_n\|_1\le 4^{-n}$ for every $n\in\Bbb N$.
Choose integrable functions $f_n$
such that $f_0^{\ssbullet}=u_0$,
$f_{n+1}^{\ssbullet}=u_{n+1}-u_n$ for each $n\in\Bbb N$. Then
\Centerline{$\sum_{n=0}^{\infty}\int|f_n|
=\|u_0\|_1+\sum_{n=0}^{\infty}\|u_{n+1}-u_n\|_1<\infty$.}
\noindent So $f=\sum_{n=0}^{\infty}f_n$ is integrable, by 242E, and
$u=f^{\ssbullet}\in L^1$. Set $g_n=\sum_{j=0}^nf_j$ for each $n$;
then $g_n^{\ssbullet}=u_n$, so
\Centerline{$\|u-u_n\|_1
=\int|f-g_n|
\le\int\sum_{j=n+1}^{\infty}|f_j|
\le\sum_{j=n+1}^{\infty}4^{-j}
=4^{-n}/3$}
\noindent for each $n$. Thus $u=\lim_{n\to\infty}u_n$ in $L^1$.
As $\sequencen{u_n}$ is arbitrary, $L^1$ is complete (2A4E).
}
\leader{242G}{Definition}\cmmnt{ It will be convenient, for later
reference, to
introduce the following phrase.} A {\bf Banach lattice} is a Riesz
space $U$ together with a norm $\|\,\|$ on $U$ such that (i)
$\|u\|\le\|v\|$ whenever $u$, $v\in U$ and $|u|\le|v|$\cmmnt{,
writing
$|u|$ for $u\vee(-u)$, as in 241Ee} (ii) $U$ is
complete under $\|\,\|$. \cmmnt{Thus 242Dc and 242F amount to
saying
that the
normed Riesz space $(L^1,\|\,\|_1)$ is a Banach lattice.}
\leader{242H}{$L^1$ as a Riesz \dvrocolon{space}}\cmmnt{ We can
discuss the ordered linear space $L^1$ in the language already used in
241E-241G %241E 241F 241G
for $L^0$.
\medskip
\noindent}{\bf Theorem} Let $(X,\Sigma,\mu)$ be any measure space.
Then $L^1=L^1(\mu)$ is Dedekind complete.
\proof{{\bf (a)} Let $A\subseteq L^1$ be any non-empty set which is
bounded above in $L^1$. Set
\Centerline{$A'=\{u_0\vee\ldots\vee u_n:u_0,\ldots, u_n\in A\}$.}
\noindent Then $A\subseteq A'$, $A'$ has the same upper bounds as $A$
and $u\vee v\in A'$ for all $u$, $v\in A'$. Taking $w_0$ to be any
upper bound of $A$ and $A'$, we have $\int u\le\int w_0$ for every
$u\in A'$, so $\gamma=\sup_{u\in A'}\int u$ is defined in $\Bbb R$.
For each
$n\in\Bbb N$, choose $u_n\in A'$ such that $\int u_n\ge\gamma-2^{-n}$.
Because $L^0=L^0(\mu)$ is Dedekind $\sigma$-complete (241Ga),
$u^*=\sup_{n\in\Bbb N}u_n$ is defined in $L^0$, and $u_0\le u^*\le
w_0$
in $L^0$. Consequently
\Centerline{$0\le u^*-u_0\le w_0-u_0$}
\noindent in $L^0$. But $w_0-u_0\in L^1$, so $u^*-u_0\in L^1$
(242Cb)
and $u^*\in L^1$.
\medskip
{\bf (b)} The point is that $u^*$ is an upper bound for $A$.
\Prf\ If $u\in A$, then $u\vee u_n\in A'$ for every $n$, so
$$\eqalignno{\|u-u\wedge u^*\|_1
&=\int u-u\wedge u^*
\le \int u-u\wedge u_n\cr
\noalign{\noindent (because $u\wedge u_n\le u_n\le u^*$, so $u\wedge
u_n\le u\wedge u^*$)}
&=\int u\vee u_n-u_n\cr
\noalign{\noindent (because $u\vee u_n+u\wedge u_n=u+u_n$ -- see the
formulae in 242Cd)}
&=\int u\vee u_n-\int u_n
\le\gamma-(\gamma-2^{-n})
=2^{-n}\cr}$$
\noindent for every $n$; so $\|u-u\wedge u^*\|_1=0$. But this
means that $u=u\wedge u^*$, that is, that $u\le u^*$. As $u$ is
arbitrary, $u^*$ is an upper bound for $A$.\ \Qed
\medskip
{\bf (c)} On the other hand, any upper bound for $A$ is surely an
upper
bound for $\{u_n:n\in\Bbb N\}$, so is greater than or equal to $u^*$.
Thus $u^*=\sup A$ in $L^1$. As $A$ is arbitrary, $L^1$ is Dedekind
complete.
}%end of proof of 242H
\cmmnt{
\medskip
\noindent{\bf Remark} Note that the order-completeness of $L^1$,
unlike
that of $L^0$, does not depend on any particular property of the
measure
space $(X,\Sigma,\mu)$.
}%end of comment
\leader{242I}{The Radon-Nikod\'ym \dvro{Theorem}{theorem}}\cmmnt{ I
think it is worth re-writing the Radon-Nikod\'ym theorem (232E) in the
language of this chapter.
\medskip
\noindent{\bf Theorem}} Let $(X,\Sigma,\mu)$ be a measure space.
Then
there is a canonical bijection between $L^1=L^1(\mu)$ and the set of
truly continuous additive functionals $\nu:\Sigma\to\Bbb R$, given by
the formula
\Centerline{$\nu F=\int_Fu$ for $F\in\Sigma$, $u\in L^1$.}
\cmmnt{\medskip
\noindent{\bf Remark} Recall that if $\mu$ is $\sigma$-finite, then
the
truly continuous additive functionals are just the absolutely
continuous
countably additive functionals; and that if $\mu$ is totally finite,
then all absolutely continuous (finitely) additive functionals are
truly
continuous (232Bd).
}%end of comment
\proof{ For $u\in L^1$, $F\in\Sigma$ set $\nu_uF=\int_Fu$.
If $u\in L^1$, there is an integrable function $f$ such that
$f^{\ssbullet}=u$, in which case
\Centerline{$F\mapsto\nu_uF=\int_Ff:\Sigma\to\Bbb R$}
\noindent is additive and truly continuous, by 232D. If
$\nu:\Sigma\to\Bbb R$ is additive and truly continuous, then by 232E
there is an integrable function $f$ such that $\nu F=\int_Ff$ for
every
$F\in\Sigma$; setting $u=f^{\ssbullet}$ in $L^1$, $\nu=\nu_u$.
Finally, if $u$, $v$ are distinct members of $L^1$, there is an
$F\in\Sigma$ such
that $\int_Fu\ne\int_Fv$ (242Ce), so that $\nu_u\ne\nu_v$; thus
$u\mapsto\nu_u$ is injective as well as surjective.
}%end of proof of 242I
\leader{242J}{Conditional expectations revisited}\cmmnt{ We now have
the machinery necessary for a new interpretation of some of the ideas
of \S233.
\medskip
}{\bf (a)} Let $(X,\Sigma,\mu)$ be
a measure space, and $\Tau$ a $\sigma$-subalgebra of
$\Sigma$\cmmnt{,
as in 233A}. Then $(X,\Tau,\mu\restrp\Tau)$ is a measure space, and
$\eusm L^0(\mu\restrp\Tau)\subseteq\eusm L^0(\mu)$;
\cmmnt{moreover,}
if $f$, $g\in\eusm L^0(\mu\restrp\Tau)$, then
$f=g\,\,(\mu\restrp\Tau)$-a.e.\ iff
$f=g\,\,\mu$-a.e. \prooflet{\Prf\ There are
$\mu\restrp\Tau$-conegligible sets
$F$, $G\in\Tau$ such that $f\restr F$ and $g\restr G$ are
$\Tau$-measurable; set
\Centerline{$E=\{x:x\in F\cap G,\,f(x)\ne g(x)\}\in\Tau$;}
\noindent then
\Centerline{$f=g\,\,(\mu\restrp\Tau)$-a.e.\ $\iff
(\mu\restrp\Tau)(E)=0
\iff \mu E=0 \iff f=g\,\,\mu$-a.e. \Qed}
}
Accordingly we have a canonical map $S:L^0(\mu\restrp\Tau)\to L^0(\mu)$
defined by saying that if $u\in L^0(\mu\restrp\Tau)$ is the
equivalence
class of $f\in\eusm L^0(\mu\restrp\Tau)$, then $Su$ is the equivalence
class of $f$ in $L^0(\mu)$. \cmmnt{It is easy to check, working
through the
operations described in 241D, 241E and 241H, that} $S$ is linear,
injective and order-preserving, and\cmmnt{ that} $|Su|=S|u|$,
$S(u\vee v)=Su\vee Sv$ and $S(u\times v)=Su\times Sv$ for $u$,
$v\in L^0(\mu\restrp\Tau)$.
\header{242Jb}{\bf (b)} Next, if $f\in\eusm L^1(\mu\restrp\Tau)$, then
$f\in\eusm L^1(\mu)$ and
$\int fd\mu=\int fd(\mu\restrp\Tau)$\prooflet{ (233B)}; so
$Su\in L^1(\mu)$ and $\|Su\|_1=\|u\|_1$ for every
$u\in L^1(\mu\restrp\Tau)$.
Observe also that every member of $L^1(\mu)\cap
S[L^0(\mu\restrp\Tau)]$
is actually in $S[L^1(\mu\restrp\Tau)]$. \prooflet{\Prf\ Take
$u\in L^1(\mu)\cap S[L^0(\mu\restrp\Tau)]$. Then $u$ is expressible
both as
$f^{\ssbullet}$ where $f\in\eusm L^1(\mu)$, and as $g^{\ssbullet}$
where
$g\in \eusm L^0(\mu\restrp\Tau)$. So $g\eae f$, and $g$ is
$\mu$-integrable, therefore $(\mu\restrp\Tau)$-integrable (233B
again).\
\Qed}
This means that
$S:L^1(\mu\restrp\Tau)\to L^1(\mu)\cap S[L^0(\mu\restrp\Tau)]$ is a
bijection.
\header{242Jc}{\bf (c)} Now suppose that $\mu X=1$, so that
$(X,\Sigma,\mu)$ is a
probability space. \cmmnt{Recall that $g$ is a conditional
expectation of $f$
on $\Tau$ if $g$ is $\mu\restrp\Tau$-integrable, $f$ is
$\mu$-integrable
and $\int_Fg=\int_Ff$ for every $F\in\Tau$; and that every
$\mu$-integrable function has such a conditional expectation
(233D).} If
$g$ is a conditional expectation of $f$ and $f_1=f\,\,\mu$-a.e.\ then
$g$ is a conditional expectation of $f_1$\cmmnt{, because
$\int_Ff_1=\int_Ff$
for every $F$}; and \cmmnt{I have already remarked in 233Dc that}
if
$g$, $g_1$ are conditional expectations of $f$ on $\Tau$ then
$g=g_1\,\,\mu\restrp\Tau$-a.e.
\header{242Jd}{\bf (d)} This means that we have an operator
$P:L^1(\mu)\to L^1(\mu\restrp\Tau)$
defined by saying that $P(f^{\ssbullet})=g^{\ssbullet}$ whenever
$g\in\eusm L^1(\mu\restrp\Tau)$ is a conditional expectation of
$f\in\eusm L^1(\mu)$
on $\Tau$\cmmnt{; that is, that $\int_FPu=\int_Fu$ whenever
$u\in L^1(\mu)$ and $F\in\Tau$}. \cmmnt{If we identify $L^1(\mu)$,
$L^1(\mu\restrp\Tau)$ with the sets of absolutely continuous additive
functionals defined on $\Sigma$ and $\Tau$, as in 242I, then $P$
corresponds to the operation $\nu\mapsto\nu\restrp\Tau$.}
\header{242Je}{\bf (e)} \dvro{$P$ is linear and
order-preserving.}{Because $Pu$ is uniquely defined in
$L^1(\mu\restrp\Tau)$ by the requirement $\int_FPu=\int_Fu$ for every
$F\in\Tau$ (242Ce), we see
that $P$ must be linear.} \prooflet{\Prf\ If $u$, $v\in L^1(\mu)$
and
$c\in\Bbb R$, then
\Centerline{$\int_FPu+Pv=\int_FPu+\int_FPv=\int_Fu+\int_Fv
=\int_Fu+v=\int_FP(u+v)$,}
\Centerline{$\int_FP(cu)=\int_Fcu=c\int_Fu=c\int_FPu=\int_FcPu$}
\noindent for every $F\in\Tau$.\ \QeD\ Also, if $u\ge 0$, then
$\int_FPu=\int_Fu\ge 0$ for every $F\in\Tau$, so $Pu\ge 0$ (242Ce
again).
It follows at once that $P$ is order-preserving, that is, that
$Pu\le Pv$ whenever $u\le v$.} Consequently
\Centerline{$|Pu|=Pu\vee(-Pu)=Pu\vee P(-u)\le P|u|$}
\noindent for every $u\in L^1(\mu)$\prooflet{, because
$u\le|u|$ and $-u\le|u|$}. Finally, $P$ is a bounded linear operator,
with norm $1$. \prooflet{\Prf\ The last formula tells us that
\Centerline{$\|Pu\|_1\le\|P|u|\|_1=\int P|u|=\int |u|=\|u\|_1$}
\noindent for every $u\in L^1(\mu)$, so $\|P\|\le 1$. On the other hand,
$P(\chi X^{\ssbullet})=\chi X^{\ssbullet}\ne 0$, so $\|P\|=1$.\ \Qed}
\header{242Jf}{\bf (f)} We may legitimately regard
$Pu\in L^1(\mu\restrp\Tau)$ as
`the' conditional expectation of $u\in L^1(\mu)$ on $\Tau$; $P$ is
the {\bf conditional expectation operator}.
\header{242Jg}{\bf (g)} If $u\in L^1(\mu\restrp\Tau)$, then\cmmnt{
we have a
corresponding} $Su\in L^1(\mu)$, as in (b); now $PSu=u$.
\prooflet{\Prf\
$\int_FPSu=\int_FSu=\int_Fu$ for every $F\in\Tau$.\ \QeD}
Consequently $SPSP=SP:L^1(\mu)\to L^1(\mu)$.
\cmmnt{\header{242Jh}{\bf (h)} The distinction drawn above between
$u=f^{\ssbullet}\in L^0(\mu\restrp\Tau)$ and $Su=f^{\ssbullet}\in L^0(\mu)$
is of course pedantic. I believe it is necessary to be aware of such
distinctions, even though for nearly all purposes it is safe as well
as convenient to regard $L^0(\mu\restrp\Tau)$ as actually a subset of
$L^0(\mu)$. If we do so, then (b) tells us that we can identify
$L^1(\mu\restrp\Tau)$ with $L^1(\mu)\cap L^0(\mu\restrp\Tau)$, while
(g) becomes `$P^2=P$'.
}
\leader{242K}{}\cmmnt{ The language just introduced allows the
following re-formulations of 233J-233K.
\medskip
\noindent}{\bf Theorem} Let $(X,\Sigma,\mu)$ be a probability space
and
$\Tau$ a $\sigma$-subalgebra of $\Sigma$. Let $\phi:\Bbb R\to\Bbb R$
be a convex function and $\bar\phi:L^0(\mu)\to L^0(\mu)$ the
corresponding operator defined by setting
$\bar\phi(f^{\ssbullet})=(\phi f)^{\ssbullet}$\cmmnt{ (241I)}. If
$P:L^1(\mu)\to L^1(\mu\restrp\Tau)$ is the conditional expectation
operator, then $\bar\phi(Pu)\le P(\bar\phi u)$ whenever
$u\in L^1(\mu)$ is such that $\bar\phi(u)\in L^1(\mu)$.
\proof{ This is just a restatement of 233J.}
\leader{242L}{Proposition} Let $(X,\Sigma,\mu)$ be a probability
space,
and $\Tau$ a $\sigma$-subalgebra of $\Sigma$. Let
$P:L^1(\mu)\to L^1(\mu\restrp\Tau)$
be the corresponding conditional expectation
operator. If $u\in L^1=L^1(\mu)$ and $v\in L^0(\mu\restrp\Tau)$,
then
$u\times v\in L^1$ iff $P|u|\times v\in L^1$, and in this case
$P(u\times v)=Pu\times v$; in particular, $\int u\times v=\int
Pu\times
v$.
\proof{ (I am here using the identification of $L^0(\mu\restrp\Tau)$
as
a subspace of $L^0(\mu)$, as suggested in 242Jh.)
Express $u$ as $f^{\ssbullet}$ and $v$ as $h^{\ssbullet}$, where
$f\in\eusm L^1=\eusm L^1(\mu)$ and $h\in\eusm L^0(\mu\restrp\Tau)$.
Let $g$, $g_0\in\eusm L^1(\mu\restrp\Tau)$ be conditional expectations
of $f$, $|f|$ respectively, so that $Pu=g^{\ssbullet}$ and
$P|u|=g_0^{\ssbullet}$. Then, using 233K,
\Centerline{$u\times v\in L^1\iff f\times h\in\eusm L^1
\iff g_0\times h\in\eusm L^1\iff P|u|\times v\in L^1$,}
\noindent and in this case $g\times h$ is a conditional expectation of
$f\times h$, that is, $Pu\times v=P(u\times v)$.
}%end of proof of 242L
\leader{242M}{$L^1$ as a \dvro{completion:}{completion}}\cmmnt{ I
mentioned in the introduction to
this section that $L^1$ appears in functional analysis as a completion
of some important spaces; put another way, some dense subspaces of
$L^1$ are significant. The first is elementary.
\medskip
\noindent}{\bf Proposition} Let $(X,\Sigma,\mu)$ be any measure space,
and write $\eusm S$ for the space of $\mu$-simple functions on $X$.
Then
(a) whenever $f$ is a $\mu$-integrable real-valued function and
$\epsilon>0$, there is an $h\in\eusm S$ such that
$\int|f-h|\le\epsilon$;
(b) $S=\{f^{\ssbullet}:f\in\eusm S\}$ is a dense linear subspace of
$L^1=L^1(\mu)$.
\proof{{\bf (a)}(i) If $f$ is non-negative, then there is a simple
function $h$ such that $h\leae f$ and
$\int h\ge\int f-\bover12\epsilon$ (122K), in which case
\Centerline{$\int|f-h|=\int f-h
=\int f-\int h\le\Bover12\epsilon$.}
\noindent (ii) In the general case, $f$ is expressible as a
difference $f_1-f_2$ of non-negative integrable functions. Now there
are $h_1$, $h_2\in \eusm S$ such that
$\int|f_j-h_j|\le\bover12\epsilon$
for both $j$ and
\Centerline{$\int|f-h|\le\int|f_1-h_1|+\int|f_2-h_2|\le\epsilon$.}
\medskip
{\bf (b)} Because $\eusm S$ is a
linear subspace of $\Bbb R^X$ included in $\eusm L^1=\eusm L^1(\mu)$,
$S$ is a linear subspace of $L^1$. If
$u\in L^1$ and $\epsilon>0$, there are an $f\in\eusm L^1$ such that
$f^{\ssbullet}=u$ and an $h\in\eusm S$ such that
$\int|f-h|\le\epsilon$; now $v=h^{\ssbullet}\in S$ and
\Centerline{$\|u-v\|_1=\int|f-h|\le\epsilon$.}
\noindent As $u$ and $\epsilon$ are arbitrary, $S$ is dense in $L^1$.
}
\leader{242N}{}\cmmnt{ As always, Lebesgue measure on $\BbbR^r$ and
its subsets is by far the most important example; and in this case we
have further classes of dense subspace of $L^1$. If you have reached
this point without yet troubling to master multi-dimensional Lebesgue
measure, just take $r=1$. If you feel uncomfortable with general
subspace measures, take $X$ to be $\BbbR^r$ or $[0,1]\subseteq\Bbb R$
or some other particular subset which you find interesting. The
following term will be useful.
\medskip
\noindent}{\bf Definition} If $f$ is a real- or complex-valued
function
defined on a subset of $\BbbR^r$, say that the {\bf support} of $f$ is
$\overline{\{x:x\in\dom f,\,f(x)\ne 0\}}$.
\leader{242O}{Theorem} Let $X$ be any subset of $\BbbR^r$, where
$r\ge 1$, and let $\mu$ be Lebesgue measure on $X$\cmmnt{, that is,
the subspace measure on $X$ induced by Lebesgue measure on $\BbbR^r$}.
Write $C_k$ for the space of bounded continuous functions
$f:\BbbR^r\to\Bbb R$ which have bounded support, and $S_0$ for
the space of linear combinations of functions of the form $\chi I$
where
$I\subseteq\BbbR^r$ is a bounded half-open interval. Then
(a) whenever $f\in\eusm L^1=\eusm L^1(\mu)$ and
$\epsilon>0$, there are $g\in C_k$, $h\in S_0$ such that
$\int_X|f-g|\le\epsilon$ and $\int_X|f-h|\le\epsilon$;
(b) $\{(g\restr X)^{\ssbullet}:g\in C_k\}$ and
$\{(h\restr X)^{\ssbullet}:h\in S_0\}$ are dense
linear subspaces of $L^1=L^1(\mu)$.
\cmmnt{\medskip
\noindent{\bf Remark} Of course there is a redundant `bounded' in the
description of $C_k$; see 242Xh.}
\proof{{\bf (a)} I argue in turn that the result is valid for each of
an
increasing number of members $f$ of $\eusm L^1=\eusm L^1(\mu)$.
Write
$\mu_r$ for Lebesgue measure on $\BbbR^r$, so
that $\mu$ is the subspace measure $(\mu_r)_X$.
\medskip
\quad{\bf (i)} Suppose first that $f=\chi I\restr X$ where
$I\subseteq\BbbR^r$ is a bounded half-open interval.
Of course $\chi I$ is already in $S_0$, so I have only to show
that it is approximated by members of $C_k$. If $I=\emptyset$
the result is trivial; we can take $g=0$.
Otherwise, express $I$ as $\coint{a-b,a+b}$ where
$a=(\alpha_1,\ldots,\alpha_r)$, $b=(\beta_1,\ldots,\beta_r)$ and
$\beta_j>0$ for each $j$. Let $\delta>0$ be such that
\Centerline{$2^r\prod_{j=1}^r(\beta_j+\delta)
\le\epsilon+2^r\prod_{j=1}^r\beta_j$.}
\noindent For $\xi\in\Bbb R$ set
$$\eqalign{g_j(\xi)
&=1\text{ if }|\xi-\alpha_j|\le\beta_j,\cr
&=(\beta_j+\delta-|\xi-\alpha_j|)/\delta
\text{ if }\beta_j\le|\xi-\alpha_j|\le\beta_j+\delta,\cr
&=0\text{ if }|\xi-\alpha_j|\ge\beta_j+\delta.\cr}$$
\def\Caption{The function $g_j$}
\picture{mt242m}{100pt}
\noindent For $x=(\xi_1,\ldots,\xi_r)\in \BbbR^r$ set
\Centerline{$g(x)=\prod_{j=1}^rg_j(\xi_j)$.}
\noindent Then $g\in C_k$ and $\chi I\le g\le\chi J$, where
$J=[a-b-\delta \tbf{1},a+b+\delta\tbf{1}]$ (writing
$\tbf{1}=(1,\ldots,1)$), so that (by the choice of $\delta$) $\mu_r
J\le\mu_r I+\epsilon$, and
$$\eqalign{\int_X|g-f|
&\le\int(\chi(J\cap X)-\chi(I\cap X))d\mu
=\mu((J\setminus I)\cap X)\cr
&\le\mu_r(J\setminus I)
=\mu_rJ-\mu_rI
\le\epsilon,\cr}$$
\noindent as required.
\medskip
\quad{\bf (ii)} Now suppose that $f=\chi(X\cap E)$ where $E\subseteq
\BbbR^r$ is a
set of finite measure. Then there is a disjoint family
$I_0,\ldots,I_n$ of half-open intervals such that
$\mu_r(E\symmdiff\bigcup_{j\le n}I_j)\le\bover12\epsilon$. \Prf\
There
is an open set $G\supseteq E$ such that
$\mu_r(G\setminus E)\le\bover14\epsilon$ (134Fa). For each
$m\in\Bbb N$, let $\Cal I_m$
be the family of half-open intervals in $\BbbR^r$ of the form
$\coint{a,b}$ where $a=(2^{-m}k_1,\ldots,2^{-m}k_r)$, $k_1,\ldots,k_r$
being integers, and $b=a+2^{-m}\tbf{1}$; then $\Cal I_m$ is a
disjoint
family. Set $H_m=\bigcup\{I:I\in\Cal I_m,\,I\subseteq G\}$; then
$\sequence{m}{H_m}$ is a non-decreasing family with union $G$, so that
there is an $m$ such that $\mu_r(G\setminus H_m)\le\bover14\epsilon$
and
$\mu_r(E\symmdiff H_m)\le\bover12\epsilon$. But now $H_m$ is
expressible as a disjoint union $\bigcup_{j\le n}I_j$ where
$I_0,\ldots,I_n$ enumerate the members of $\Cal I_m$ included in
$H_m$.
(The last sentence derails if $H_m$ is empty. But if $H_m=\emptyset$
then
we can take $n=0$, $I_0=\emptyset$.) \Qed
Accordingly $h=\sum_{j=0}^n\chi I_j\in S_0$ and
\Centerline{$\int_X|f-h|
=\mu(X\cap(E\symmdiff\bigcup_{j\le n}I_j))\le\Bover12\epsilon$.}
\noindent As for $C_k$, (i) tells us that there is for each
$j\le n$ a $g_j\in C_k$ such that
$\int_X|g_j-\chi I_j|\le\epsilon/2(n+1)$, so that
$g=\sum_{j=0}^ng_j\in C_k$ and
\Centerline{$\int_X|f-g|\le\int_X|f-h|+\int_X|h-g|
\le\Bover{\epsilon}2+\sum_{j=0}^n\int_X|g_j-\chi I_j|\le\epsilon$.}
\medskip
\quad{\bf (iii)} If $f$ is a simple function, express $f$ as
$\sum_{k=0}^na_k\chi E_k$ where each $E_k$ is of finite measure for
$\mu$.
Each $E_k$ is expressible as $X\cap F_k$ where $\mu_rF_k=\mu E_k$
(214Ca). By (ii), we can find $g_k\in C_k$, $h_k\in S_0$ such that
\Centerline{$|a_k|\int_X|g_k-\chi F_k|\le\Bover{\epsilon}{n+1}$,
\quad$|a_k|\int_X|h_k-\chi F_k|\le\Bover{\epsilon}{n+1}$}
\noindent for each $k$. Set
$g=\sum_{k=0}^na_kg_k$ and $h=\sum_{k=0}^na_kh_k$; then $g\in C_k$,
$h\in S_0$ and
\Centerline{$\int_X|f-g|\le\int_X\sum_{k=0}^n|a_k||\chi F_k-g_k|
=\sum_{k=0}^n|a_k|\int_X|\chi F_k-g_k|\le\epsilon$,}
\Centerline{$\int_X|f-h|
\le\sum_{k=0}^n|a_k|\int_X|\chi F_k-h_k|\le\epsilon$,}
\noindent as required.
\medskip
\quad{\bf (iv)} If $f$ is any integrable function on $X$, then by
242Ma
we can find a simple function $f_0$ such that
$\int|f-f_0|\le\bover12\epsilon$, and now by (iii) there are
$g\in C_k$, $h\in S_0$ such that
$\int_X|f_0-g|\le\bover12\epsilon$,
$\int_X|f_0-h|\le\bover12\epsilon$; so that
\Centerline{$\int_X|f-g|
\le\int_X|f-f_0|+\int_X|f_0-g|\le\epsilon$,}
\Centerline{$\int_X|f-h|
\le\int_X|f-f_0|+\int_X|f_0-h|\le\epsilon$.}
\medskip
{\bf (b)(i)} We must check first that if $g\in C_k$
then $g\restr X$ is
actually $\mu$-integrable. The point here is that if $g\in C_k$ and
$a\in\Bbb R$ then
\Centerline{$\{x:x\in X,\,g(x)>a\}$}
\noindent is the intersection of $X$ with an open subset of $\BbbR^r$,
and is therefore measured by $\mu$, because all open sets are measured
by $\mu_r$ (115G). Next, $g$ is bounded and the set
$E=\{x:x\in X,\,g(x)\ne 0\}$ is bounded in $\BbbR^r$, therefore of
finite outer
measure for $\mu_r$ and of finite measure for $\mu$. Thus there is
an
$M\ge 0$ such that $|g|\le M\chi E$, which is $\mu$-integrable.
Accordingly $g$ is $\mu$-integrable.
Of course $h\restr X$ is $\mu$-integrable for every $h\in S_0$
because (by the
definition of subspace measure) $\mu(I\cap X)$ is defined and finite
for
every bounded half-open interval $I$.
\medskip
\quad{\bf (ii)} Now the rest follows by just the same arguments as in
242Mb. Because $\{g\restr X:g\in C_k\}$ and
$\{h\restr X:h\in S_0\}$ are linear subspaces of $\Bbb R^X$
included in $\eusm L^1(\mu)$, their images $C_k^{\#}$ and $S_0^{\#}$
are
linear subspaces of $L^1$.
If $u\in L^1$ and $\epsilon>0$, there are an $f\in\eusm L^1$ such that
$f^{\ssbullet}=u$, and $g\in C_k$, $h\in S_0$ such that
$\int_X|f-g|$, $\int_X|f-h|\le\epsilon$; now
$v=(g\restr X)^{\ssbullet}\in C_k^{\#}$
and $w=(h\restr X)^{\ssbullet}\in S_0^{\#}$ and
\Centerline{$\|u-v\|_1=\int_X|f-g|\le\epsilon$,
\quad$\|u-w\|_1=\int_X|f-h|\le\epsilon$.}
\noindent As $u$ and $\epsilon$ are arbitrary, $C_k^{\#}$ and
$S_0^{\#}$ are dense in $L^1$.
}%end of proof of 242O
\leader{242P}{Complex $L^1$}\cmmnt{ As you would, I hope, expect, we
can repeat the work above with
$\eusm L^1_{\Bbb C}$, the space of complex-valued integrable
functions,
in place of $\eusm L^1$, to construct a complex Banach space
$L^1_{\Bbb C}$. The required changes, based on the ideas of 241J,
are
minor.
\medskip
\noindent}{\bf (a)}\dvro{ For}{ In 242Aa, it is perhaps helpful to
remark that, for} $f\in\eusm L^0_{\Bbb C}$,
\Centerline{$f\in\eusm L^1_{\Bbb C}\iff |f|\in\eusm L^1\iff
\Real(f),\,\Imag(f)\in\eusm L^1$.}
\noindent Consequently, for $u\in L^0_{\Bbb C}$,
\Centerline{$u\in L^1_{\Bbb C}\iff |u|\in L^1\iff
\Real(u),\,\Imag(u)\in
L^1$.}
\header{242Pb}{\bf (b)}\cmmnt{ To prove a complex version of 242E,
observe that if $\sequencen{f_n}$ is a sequence in $\eusm L^1_{\Bbb
C}$
such that
$\sum_{n=0}^{\infty}\int|f_n|<\infty$, then
$\sum_{n=0}^{\infty}\int|\Real(f_n)|$ and
$\sum_{n=0}^{\infty}\int|\Imag(f_n)|$ are both finite, so we may apply
242E twice and see that
\Centerline{$\int(\sum_{n=0}^{\infty}f_n)
=\int(\sum_{n=0}^{\infty}\Real(f_n))
+\int(\sum_{n=0}^{\infty}\Imag(f_n))
=\sum_{n=0}^{\infty}\int f_n$.}
\noindent Accordingly we can prove that} $L^1_{\Bbb C}$ is complete
under $\|\,\|_1$\cmmnt{ by the argument of 242F}.
\cmmnt{\header{242Pc}{\bf (c)} Similarly, little change is needed to
adapt 242J to give a description of a
conditional expectation operator $P:L^1_{\Bbb C}(\mu)\to L^1_{\Bbb
C}(\mu\restrp\Tau)$ when $(X,\Sigma,\mu)$ is a probability space and
$\Tau$ is a $\sigma$-subalgebra of $\Sigma$.
In the formula
\Centerline{$|Pu|\le P|u|$}
\noindent of 242Je, we need to know that
\Centerline{$|Pu|=\sup_{|\zeta|=1}\Real(\zeta Pu)$}
\noindent in $L^0(\mu\restrp\Tau)$ (241Jc), while
\Centerline{$\Real(\zeta Pu)=\Real(P(\zeta u))
=P(\Real(\zeta u))\le P|u|$}
\noindent whenever $|\zeta|=1$.
}
\cmmnt{
\header{242Pd}{\bf (d)} In 242M, we need to replace $\eusm S$ by
$\eusm S_{\Bbb C}$, the space
of `complex-valued simple functions' of the form
$\sum_{k=0}^na_k\chi E_k$ where each $a_k$ is a complex number and
each $E_k$ is a measurable set
of finite measure; then we get a dense linear subspace $S_{\Bbb
C}=\{f^{\ssbullet}:f\in\eusm S_{\Bbb C}\}$ of $L^1_{\Bbb C}$. In
242O, we must replace $C_k$ by $C_k(\BbbR^r;\Bbb C)$, the space of
bounded continuous complex-valued functions of bounded support, and
$S_0$ by the linear span over
$\Bbb C$ of $\{\chi I:I$ is a bounded half-open interval$\}$.
}
\exercises{
\leader{242X}{Basic exercises $\pmb{>}$(a)} Let $X$ be a set, and let
$\mu$ be
counting measure on $X$. Show that $L^1(\mu)$ can be identified with
the space $\ell^1(X)$ of absolutely summable real-valued functions on
$X$ (see 226A). In particular, the space $\ell^1=\ell^1(\Bbb N)$ of
absolutely summable real-valued sequences is an $L^1$ space.
Write out proofs of 242F adapted to these special cases.
%242F, but put at start
\sqheader 242Xb Let $(X,\Sigma,\mu)$ be any measure space, and
$\hat\mu$ the completion of $\mu$. Show that
$\eusm L^1(\hat\mu)=\eusm L^1(\mu)$ and $L^1(\hat\mu)=L^1(\mu)$ (cf.\
241Xb).
%242A
\spheader 242Xc Let
$\langle(X_i,\Sigma_i,\mu_i)\rangle_{i\in I}$ be a family of measure
spaces, and $(X,\Sigma,\mu)$ their direct
sum. Show that the isomorphism between $L^0(\mu)$ and
$\prod_{i\in I}L^0(\mu_i)$ (241Xd) induces an identification between
$L^1(\mu)$ and
\Centerline{$\{u:u\in\prod_{i\in I}L^1(\mu_i),\,
\|u\|=\sum_{i\in I}\|u(i)\|_1<\infty\}
\subseteq\prod_{i\in I}L^1(\mu_i)$.}
%242D
\spheader 242Xd Let $(X,\Sigma,\mu)$ and $(Y,\Tau,\nu)$ be measure
spaces, and $\phi:X\to Y$ an \imp\ function. Show that
$g\mapsto g\phi:\eusm L^1(\nu)\to\eusm L^1(\mu)$ (235G) induces a
linear operator
$T:L^1(\nu)\to L^1(\mu)$ such that $\|Tv\|_1=\|v\|_1$ for every
$v\in L^1(\nu)$.
%242D
\spheader 242Xe Let $U$ be a Riesz space (definition: 241Ed). A
{\bf Riesz norm} on $U$ is a norm $\|\,\|$ such that $\|u\|\le\|v\|$
whenever $|u|\le|v|$. Show that if $U$ is given its norm topology
(2A4Bb) for such a norm, then (i)
$u\mapsto|u|:U\to U$, $(u,v)\mapsto u\vee v:U\times U\to U$ are
continuous (ii) $\{u:u\ge 0\}$ is closed.
%242D
\spheader 242Xf Show that any Banach lattice must be an
Archimedean Riesz space (241Fa).
%242G
\spheader 242Xg Let $(X,\Sigma,\mu)$ be a probability space, and
$\Tau$ a $\sigma$-subalgebra of $\Sigma$, $\Upsilon$ a
$\sigma$-subalgebra of $\Tau$. Let
$P_1:L^1(\mu)\to L^1(\mu\restrp\Tau)$,
$P_2:L_1(\mu\restrp\Tau)\to L^1(\mu\restrp\Upsilon)$ and
$P:L^1(\mu)\to L^1(\mu\restrp\Upsilon)$ be the corresponding