MATH 597

MATH 597: Measure Theory

Typst-first course notes and worked homeworks

Qiulin Fan ยท Winter 2025

1 sigma-algebra ไธŽ measure

1.1 ๐œŽ-algebra [Fol 1.2]

ๆˆ‘ไปฌ (่ง my Math 395 notes) ๅทฒ็ป่ฏๆ˜Ž: ๅœจ โ„ ไธŠไธๅญ˜ๅœจไธ€ไธช measure function ๐œ‡:๐’ซ๏ธ€(โ„)โ†’[0,โˆž] satisfying:

  1. ๐œ‡(โˆ…)=0;

  2. translate invariant

  3. countably additivite

ๅ› ่€Œ, ๅฏนไบŽๆฏ”ๅฆ‚ โ„ ็š„่ฟ™็งๆ— ๆณ•ๅœจๅ…ถๅน‚้›†ไธŠๅฎšไน‰่‰ฏๅฅฝ็š„ measure function ็š„้›†ๅˆ, ๆˆ‘ไปฌ่ฆๅฎšไน‰ไธ€ไธช ๐’œ๏ธ€โІ๐’ซ๏ธ€(๐‘‹), ไฝฟๅพ—ๆˆ‘ไปฌ่ƒฝๅœจ่ฟ™ไธช power set ็š„ๅญ้›†ไธŠ, ๅฎšไน‰ไธ€ไธช make sense ็š„ measure.

้ฆ–ๅ…ˆ, ไธบไบ†ๅฏนไบŽไธ€ไธชไปปๆ„็š„้›†ๅˆ ๐‘‹ ้ƒฝ่ƒฝๅœจๅ…ถไธŠๅฎšไน‰ measure, ๆˆ‘ไปฌ่ฆ่€ƒ่™‘ๅœจ ๐‘‹ ็š„ไธ€ไธชไป€ไนˆๆ ท็š„ๅญ้›†็ฐ‡ไธŠๆœ‰ๅธŒๆœ›ๅฎšไน‰่ฟ™ๆ ท็š„ measure.

Definition 1.1 : ๐œŽ-algebra

ๅฏนไบŽ set ๐‘‹, ๐‘†โІ๐’ซ๏ธ€(๐‘‹) ่ขซ็งฐไธบ ๐‘‹ ไธŠ็š„ไธ€ไธช ๐œŽ-algebra, if ๅ…ถๆปก่ถณ:

  1. โˆ…โˆˆ๐‘‹;

  2. closed under complement: if ๐ธโˆˆ๐‘† then ๐‘‹\๐ธโˆˆ๐‘†;

  3. closed under countable union: if ๐ธ1,๐ธ2,โ‹ฏโˆˆ๐‘† then โ‹ƒ๐‘˜=1โˆž๐ธ๐‘˜โˆˆ๐‘†.

ๅฆ‚ๆžœ็ฌฌไธ‰ๆกๅนถไธๆปก่ถณ, ่€Œๆ˜ฏๅชๆปก่ถณ closed under finite union, ๅˆ™็งฐ ๐‘† ๆ˜ฏ ๐‘‹ ไธŠ็š„ไธ€ไธช algebra of sets . ๅฝ“็„ถ, ๐œŽ-algebra ๆ˜ฏๆฏ” algebra ไธฅๆ ผๆ›ดๅผบ็š„ๆกไปถ.

ๆˆ‘ไปฌๅฎšไน‰ ๐‘‹ ็š„ไธ€ไธชๅญ้›†็ฐ‡ไธบไธ€ไธช ๐œŽ-algebra ๅฆ‚ๆžœๅฎƒๅŒ…ๅซ็ฉบ้›†ๅนถ closed under complement and countable union. ไฝ†่ฟ™ๅนถไธๆ˜ฏ ๐œŽ-algebra ็š„ๅ…จ้ƒจๆ€ง่ดจ. ่ฟ™ไธ‰ไธชๆ€ง่ดจ่ฟ˜่•ดๆถตไบ†: ๐œŽ-algebra ไนŸไธ€ๅฎšๅŒ…ๅซ ๐‘‹, ไธ” closed under set difference, symmetric difference ไปฅๅŠ countable intersection.
ๅฏนไบŽ algebra, ๅฎƒไนŸๆœ‰ไปฅไธŠ็š„ๆ‰€ๆœ‰ๆ€ง่ดจ็š„ finite version.

Theorem 1.1 : ๐œŽ-algebra closure properties: set difference, symmetric difference and countable intersection

Let ๐‘† be a ๐œŽ-algebra on set ๐‘‹.
Claim:

  1. ๐‘‹โˆˆ๐‘†

    Proof

    Directly from def.

    โ–ก

  2. ๐ท,๐ธโˆˆ๐‘†โŸน๐ทโˆช๐ธ,๐ทโˆฉ๐ธ,๐ท\๐ธโˆˆ๐‘†

    Proof

    union: from def by leaving others as โˆ…;
    intersection:

    (๐ทโˆฉ๐ธ)๐ถ=๐ท๐ถโˆช๐ธ๐ถโˆˆ๐‘†

    setminus:

    ๐ท\๐ธ=๐ทโˆฉ(๐‘‹\๐ธ)โˆˆ๐‘†

    โ–ก

  3. ๐ท,๐ธโˆˆ๐‘†โŸน๐ทฮ”๐‘†โˆˆ๐‘†

    Proof
    ๐ทฮ”๐ธ=(๐ท\๐ธ)โ‹ƒ(๐ธ\๐ท)

    โ–ก

  4. ๐ด1,๐ด2,โ‹ฏโˆˆ๐‘†โŸนโ‹‚๐‘–=1โˆž๐ด๐‘–โˆˆ๐‘†

    Proof
    (โ‹‚๐‘›=1โˆž)๐ถ=โ‹ƒ๐‘›=1โˆž๐ธ๐‘›๐ถโˆˆ๐‘†

    โ–ก

Lemma 1.1 : ไปปๆ„ ๐œŽ-algebra ็š„ intersection ไปๆ˜ฏ ๐œŽ-algebra

Let {๐‘†๐›ผ}๐›ผโˆˆ๐ด be a collection of ๐œŽ-algebra on ๐‘‹, then โ‹‚๐›ผโˆˆ๐ด๐‘†๐›ผ is a ๐œŽ-algebra on ๐‘‹.

Proof

่ฟ™ๆ˜ฏไธช trivial proof. ไฝ†ๆ˜ฏๅฎƒๅ…ทๆœ‰ไธ€ๅฎš็†่งฃไธŠ็š„ๅฏๅ‘.
ๆˆ‘ไปฌๅฏน ๐œŽ-algebra ๆœ‰ไธ€ไธช็›ด่ง‚็†่งฃ: ๅฆ‚ๆžœๆˆ‘ไปฌๆƒณๆŠŠไธ€ไบ›้›†ๅˆๅšๆˆไธ€ไธช ๐œŽ-algebra, ้‚ฃไนˆ้ฆ–ๅ…ˆๆˆ‘ไปฌๆŠŠๅฎƒไปฌ็š„่กฅ้›†ๆ”พ่ฟ›่ฟ™ไธช ๐œŽ-algebra ้‡Œ, ๅ…ถๆฌกๆˆ‘ไปฌๆŠŠ่ฟ™ไบ›้›†ๅˆ็š„ up to countable ็š„ไปปๆ„็ป„ๅˆ็š„ๅนถ้›†ไนŸๆ”พ่ฟ›่ฟ™ไธช ๐œŽ-algebra ้‡Œ.
ๅ› ่€Œๅณไพฟๆˆ‘ไปฌๆŠŠไธ€ไบ› ๐œŽ-algebra ็ป™ intersect ่ตทๆฅ, ๅ…ถไธญๆฏไธช้›†ๅˆ็š„่กฅ้›†ๅ’Œ่ฟ™ไบ›้›†ๅˆ็š„ up to ctbl ็š„ไปปๆ„็ป„ๅˆ็š„ๅนถ้›†ไนŸๅœจ่ฟ™ไธช intersection ้‡Œ.
่ฟ™ๆ˜ฏไธช้‡่ฆ็š„็›ด่ง‚็†่งฃ. ๆˆ‘ไปฌๆƒณๅˆฐ, ๅฆ‚ๆžœๆˆ‘ไปฌ่ฆๆŠŠไธ€ไธช sigma-algebra ้‡Œ็š„ไธ€้ƒจๅˆ†ๅŽปๆމ๏ผŒๅนถไฟๆŒๅฎƒไป็„ถๆ˜ฏไธ€ไธช sigma-algebra๏ผŒ้‚ฃไนˆๆˆ‘ไปฌๅพ—ๆŠŠ่ฟ™ไบ›้›†ๅˆ็š„่กฅ้›†, ไปฅๅŠ่ƒฝๅคŸ ctbly union ๆˆ่ฟ™ไบ›้›†ๅˆ็š„ๅฐ้›†ๅˆไนŸๅŽปๆމ, ๅนถๅฏน่ฟ™ไบ›ๅฐ้›†ๅˆไนŸ recursively ่ฟ›่กŒ่ฟ™ไธชๆ“ไฝœ.

โ–ก

Corollary 1.1 : unique smallest ๐œŽ-algebra containing a collection of subsets

Given ๐œ€โІ๐’ซ๏ธ€(๐‘‹)

<๐œ€>โ‰”โ‹‚๐œ€โІ๐‘†โІ๐’ซ๏ธ€(๐‘‹),๐‘† is ๐œŽ -algebra on ๐‘‹๐‘†
Definition 1.2 : ๐œŽ-algebra generated by a subset

We call

<๐œ€>โ‰”โ‹‚๐œ€โІ๐‘†โІ๐’ซ๏ธ€(๐‘‹),๐‘† is ๐œŽ -algebra on ๐‘‹๐‘†

the ๐œŽ-algebra generated by ๐œ€

1.2 Borel ๐œŽ-algebra on โ„ and measure [Fol 1.2, finished; 1.3]

Recall: the ๐œŽ-algebra generated by ๐œ€

<๐œ€>โ‰”โ‹‚๐œ€โІ๐‘†โІ๐’ซ๏ธ€(๐‘‹),๐‘† is ๐œŽ -algebra on ๐‘‹๐‘†

is the smallsest ๐œŽ-algebra containing ๐œ€.

Example 1.1
<{๐ธ}>={โˆ…,๐ธ,๐ธ๐‘,๐‘‹}
Lemma 1.2 : inclusion properties of generated ๐œŽ-algebra
  1. if โ„ฐ๏ธ€โІ๐’œ๏ธ€ where ๐’œ๏ธ€ is a ๐œŽ-algebra, then <โ„ฐ๏ธ€>โІ๐’œ๏ธ€.

  2. if โ„ฐ๏ธ€โІโ„ฑ๏ธ€, then <โ„ฐ๏ธ€>โІ<โ„ฑ๏ธ€>.

  3. if โ„ฐ๏ธ€โІ<โ„ฑ๏ธ€>, then <โ„ฐ๏ธ€>โІ<โ„ฑ๏ธ€>.

Proof

trivial.

โ–ก

Definition 1.3 : Borel ๐œŽ-algebra defined on a topological space

For topological space (๐‘‹,๐’ฏ๏ธ€), we define:

โ„ฌ๏ธ€๐‘‹โ‰”<๐’ฏ๏ธ€>

Borel ๐œŽ-algebra on a topological space ๅฐฑๆ˜ฏ ๐œŽ-algebra generated by the topology. Its members are called Borel sets. ๅฝ“็„ถ, ๆ‰€ๆœ‰็š„ open sets ๅ’Œ closed sets ้ƒฝๆ˜ฏ Borel sets.

1.2.1 generating Borel ๐œŽ-algebra on โ„

Example 1.2

Let โ„ฐ๏ธ€1: โ„ ไธŠๆ‰€ๆœ‰็š„ open intervals;
โ„ฐ๏ธ€2: โ„ ไธŠๆ‰€ๆœ‰็š„ closed intervals;
โ„ฐ๏ธ€3: โ„ ไธŠๆ‰€ๆœ‰็š„ๅทฆๅผ€ๅณ้—ญ intervals;
โ„ฐ๏ธ€4: โ„ ไธŠๆ‰€ๆœ‰็š„ๅทฆ้—ญๅณๅผ€ intervals;
โ„ฐ๏ธ€5: โ„ ไธŠๆ‰€ๆœ‰็š„ๅทฆๅผ€ๅณๆ— ็•Œ intervals;
โ„ฐ๏ธ€6: โ„ ไธŠๆ‰€ๆœ‰็š„ๅทฆ้—ญๅณๆ— ็•Œ intervals;
โ„ฐ๏ธ€7: โ„ ไธŠๆ‰€ๆœ‰็š„ๅทฆๆ— ็•Œๅณๅผ€ intervals;
โ„ฐ๏ธ€8: โ„ ไธŠๆ‰€ๆœ‰็š„ๅทฆๆ— ็•Œๅณ้—ญ intervals;
โ‹ƒ๐‘–=1,โ‹ฏ,8โ„ฐ๏ธ€๐‘– ๅณ โ„ ไธŠ็š„ๆ‰€ๆœ‰ๅฝขๅผ็š„ interals.

Lemma 1.3

ไปปๆ„ไปฅไธŠ โ„ฐ๏ธ€๐‘–,๐‘–=1,โ‹ฏ,8 ้ƒฝๅฏไปฅ generate โ„ฌ๏ธ€โ„

Proof

ๆˆ‘ไปฌ recall: ๆ‰€ๆœ‰็š„ countable ไปฅๅŠ second countable ็š„ topological space ้ƒฝๅ…ทๆœ‰ Lindelรถf property: ไปปๆ„ open covering ้ƒฝๅญ˜ๅœจไธ€ไธช countable ็š„ subcovering.
Lindelรถf property ็š„ไธ€ไธชๆŽจ่ฎบๅฐฑๆ˜ฏ, ๅœจๅ…ทๆœ‰ Lindelรถf property ็š„ metric space ๆˆ–่€… second countable ็š„ space ไธญ, ไปปๆ„ open set ้ƒฝๅฏไปฅๅ†™ๆˆ countable ไธช open balls ็š„ union.
ๆˆ‘ไปฌๅœจ elementary ็š„ real analysis ไธญๅทฒ็ปๅญฆ่ฟ‡, [๐‘Ž,๐‘)=โˆฉ๐‘›โ‰ฅ1(๐‘Žโˆ’1/๐‘›,๐‘), ไปฅๅ…ถไฝœไธบไพ‹ๅญ, ่ฟ™ไบ› intervals ๅฝผๆญคไน‹้—ด้ƒฝๅฏไปฅ็›ธไบ’่ฝฌๆข.

โ–ก

1.2.2 measure

Definition 1.4 : measurable space and measure space

Let ๐‘‹ be a set, โ„ณ๏ธ€ be a ๐œŽ-algebra on ๐‘‹. We call (๐‘‹,โ„ณ๏ธ€) a measurable space.
A measure on this measurable space is a function ๐œ‡:โ„ณ๏ธ€โ†’[0,โˆž) satisfying:

  1. ๐œ‡(โˆ…)=0

  2. countable additive:

    ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ธ๐‘–)=โˆ‘๐‘–=1โˆž๐œ‡(๐ธ๐‘–)

    for disjoint seq of ๐ธ๐‘–โˆˆโ„ณ๏ธ€.

ๆˆ‘ไปฌ็งฐ (๐‘‹,โ„ณ๏ธ€,๐œ‡) ไธบไธ€ไธช measure space.

Example 1.3
  1. ๅฏนไบŽไปปๆ„็š„ (๐‘‹,โ„ณ๏ธ€), ๆˆ‘ไปฌๅฏไปฅๅฎšไน‰:

    ๐œ‡(๐ด)โ‰”#๐ด(โˆˆโ„คโ‰ฅ0โˆช{โˆž})

    ่ฟ™ไธช measure ๅซๅš counting measure.

  2. Fix ๐‘ฅ0โˆˆ๐‘€, ๅฏไปฅ define

    ๐œ‡(๐ด)โ‰”๐›ฟ๐‘ฅโ‰”{1, if ๐‘ฅ0โˆˆ๐ด0, if ๐‘ฅ0โˆ‰๐ด

    ่ฟ™ไธช measure ๅซๅš the Dirac measure at ๐‘ฅ0.

  3. ็ป™ๅฎšไธ€ไธช ๐‘‹ ไธŠ็š„ๅ‡ฝๆ•ฐ ๐‘“:๐‘‹โ†’[0,โˆž), ๆˆ‘ไปฌๅฏไปฅ้€š่ฟ‡่ฟ™ไธชๅ‡ฝๆ•ฐๆฅๅฎšไน‰:

    ๐œ‡(๐ด)โ‰”โˆ‘๐‘ฅโˆˆ๐ด๐‘“(๐‘ฅ)

    ่ฟ™ไธชๆต‹ๅบฆไพ่ต–ไบŽๅ‡ฝๆ•ฐๅ€ผๆฅ่กจ็คบๆฏไธช็‚น็š„ๅ•็‚น้›†็š„ measure, ๅนถ้€š่ฟ‡ไธ€ไธช้›†ๅˆไธŠๆ‰€ๆœ‰็‚น็š„ๅ•็‚น้›† measure ็›ธๅŠ ๅพ—ๅˆฐ่ฟ™ไธช้›†ๅˆๅœจ่ฟ™ไธชๅ‡ฝๆ•ฐไธ‹็š„ measure. (็ผบ็‚น: ๆˆ‘ไปฌๅทฒ็ป็Ÿฅ้“, ๅฆ‚ๆžœไธ€ไธชๅ‡ฝๆ•ฐๅœจไธ€ไธช้›†ๅˆไธŠ็š„ๆญฃ้›†ๆ˜ฏ uncountable ็š„, ้‚ฃไนˆ่ฟ™ไธช้›†ๅˆไธŠ็š„่ฟ™ไธชๆต‹ๅบฆไธ€ๅฎšๆ˜ฏ โˆž.)

ไปฅไธ‹ๆ˜ฏ measure function ็”ฑๅฎƒ็š„ๅฎšไน‰็š„ไธคๆกๆ€ง่ดจ(็ฉบ้›†ไธบ0ไปฅๅŠ ctbl additivity)ๆŽจๅฏผๅ‡บ็š„ไธ€ไบ›ๅŸบๆœฌๆ€ง่ดจ:

Lemma 1.4 : measure is finitely additive

Measure is finitely additive.

Proof

ๆ˜พ็„ถ, ctbl additive implies finite additive.

โ–ก

Lemma 1.5

๐ด,๐ตโˆˆโ„ณ๏ธ€โŸน

๐œ‡(๐ด)+๐œ‡(๐ต)=๐œ‡(๐ดโˆฉ๐ต)+๐œ‡(๐ดโˆช๐ต)
Proof
๐ดโˆช๐ต=(๐ด\๐ต)โŠ”(๐ดโˆฉ๐ต)โŠ”(๐ต\๐ด)

่€ŒๅŽไฝฟ็”จ finite additive ๅฏๅพ—. ่ฟ™ๆ˜ฏไธ€ไธช direct corollary of countable additivity.

โ–ก

Corollary 1.2

๐ด,๐ตโˆˆโ„ณ๏ธ€,๐ดโІ๐ต,๐œ‡(๐ด)<โˆžโŸน

๐œ‡(๐ต\๐ด)=๐œ‡(๐ต)โˆ’๐œ‡(๐ด)
Theorem 1.2 : properties of measure

ๅฏนไบŽไปปไฝ• measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡):

  1. monotonicity: ๐ดโІ๐ตโˆˆโ„ณ๏ธ€โŸน๐œ‡(๐ด)โ‰ค๐œ‡(๐ต)

    Proof

    trivial.

    โ–ก

  2. countable subadditivity:

    ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)โ‰คโˆ‘๐‘–=1โˆž๐œ‡(๐ด๐‘–)
    Proof

    By setting ๐ต๐‘–=๐ด๐‘–\โ‹ƒ๐‘—=1๐‘–โˆ’1๐ด๐‘—, ่€ŒๅŽ้€š่ฟ‡ ctbl disjoint additivity ไธŽ monotonicity ๅฏๅพ—

    โ–ก

  3. continuous from above: ๅฆ‚ๆžœ ๐ด๐‘–โІ๐ด๐‘–+1โˆ€๐‘–โ‰ฅ2โŸน

    ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=lim๐‘–โ†’โˆž๐œ‡(๐ด๐‘–)
    Proof

    ไฝฟ็”จ same trick as 2.

    โ–ก

  4. countinuous from below: ๅฆ‚ๆžœ ๐ด๐‘–โЇ๐ด๐‘–+1โˆ€๐‘– ไธ”ๅญ˜ๅœจๆŸไธช ๐‘— ไฝฟๅพ— ๐œ‡(๐ด๐‘–)<โˆž, ๅˆ™

    ๐œ‡(โ‹‚๐‘–=1โˆž๐ด๐‘–)=lim๐‘›โ†’โˆž๐œ‡(๐ด๐‘›)
    Proof

    ๅ‰้ข็š„้ƒฝๆ— ่ง†, ็›ดๅˆฐ็ฌฌไธ€ไธช measure <โˆž ็š„้›†ๅˆ, ๆ˜ฏๅฏ่ƒฝๅ‡บ็Žฐๅœจๆœ€ๅŽ็š„ intersection ้‡Œ็š„ๆœ€ๅคง้›†ๅˆ. ๆˆ‘ไปฌ Fix ่ฟ™ไธช ๐ด๐‘—. ้€š่ฟ‡ๆž„้€ ่กฅ้›†็š„ๆ–นๅผ, ๆŠŠไบค่ฝฌไธบๅนถ, ไปŽ่€Œ็”จ (3) ๅพ—่ฏ. Define: ๐ธ๐‘–โ‰”๐ด๐‘—\๐ด๐‘–โˆ€๐‘–โ‰ฅ๐‘— ไปŽ่€Œ

    โ‹ƒ๐‘–=๐‘—โˆž๐ธ๐‘–=๐ด๐‘—\(โ‹‚๐‘–=๐‘—โˆž๐ด๐‘–)

    ่ฟ›่€Œ

    ๐œ‡(โ‹ƒ๐‘–=๐‘—โˆž๐ธ๐‘–)=๐œ‡(๐ด๐‘—)โˆ’๐œ‡(โ‹‚๐‘–=๐‘—โˆž๐ด๐‘–)

    ่ฟ›่€Œ by (3)

    ๐œ‡(โ‹‚๐‘–=1โˆž๐ด๐‘–)=๐œ‡(โ‹‚๐‘–=๐‘—โˆž๐ด๐‘–)=๐œ‡(๐ด๐‘—)โˆ’lim๐‘–โ†’โˆž๐œ‡(๐ธ๐‘–)=๐œ‡(๐ด๐‘—)โˆ’lim๐‘–โ†’โˆž(๐œ‡(๐ด๐‘—)โˆ’๐œ‡(๐ด๐‘–))=lim๐‘–โ†’โˆž๐ด๐‘–

    โ–ก

Recall: the ๐œŽ-algebra generated by ๐œ€

<๐œ€>โ‰”โ‹‚๐œ€โІ๐‘†โІ๐’ซ๏ธ€(๐‘‹),๐‘† is ๐œŽ -algebra on ๐‘‹๐‘†

is the smallsest ๐œŽ-algebra containing ๐œ€.

ไปฅไธ‹ๆ˜ฏ measure function ็”ฑๅฎƒ็š„ๅฎšไน‰็š„ไธคๆกๆ€ง่ดจ(็ฉบ้›†ไธบ0ไปฅๅŠ ctbl additivity)ๆŽจๅฏผๅ‡บ็š„ไธ€ไบ›ๅŸบๆœฌๆ€ง่ดจ:

Measure is finitely additive.

Proof

ๆ˜พ็„ถ, ctbl additive implies finite additive.

โ–ก

Lemma 1.7

๐ด,๐ตโˆˆโ„ณ๏ธ€โŸน

๐œ‡(๐ด)+๐œ‡(๐ต)=๐œ‡(๐ดโˆฉ๐ต)+๐œ‡(๐ดโˆช๐ต)
Proof
๐ดโˆช๐ต=(๐ด\๐ต)โŠ”(๐ดโˆฉ๐ต)โŠ”(๐ต\๐ด)

่€ŒๅŽไฝฟ็”จ finite additive ๅฏๅพ—. ่ฟ™ๆ˜ฏไธ€ไธช direct corollary of countable additivity.

โ–ก

Corollary 1.3

๐ด,๐ตโˆˆโ„ณ๏ธ€,๐ดโІ๐ต,๐œ‡(๐ด)<โˆžโŸน

๐œ‡(๐ต\๐ด)=๐œ‡(๐ต)โˆ’๐œ‡(๐ด)
Theorem 1.3 : properties of measure

ๅฏนไบŽไปปไฝ• measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡):

  1. monotonicity: ๐ดโІ๐ตโˆˆโ„ณ๏ธ€โŸน๐œ‡(๐ด)โ‰ค๐œ‡(๐ต)

    Proof

    trivial.

    โ–ก

  2. countable subadditivity:

    ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)โ‰คโˆ‘๐‘–=1โˆž๐œ‡(๐ด๐‘–)
    Proof

    By setting ๐ต๐‘–=๐ด๐‘–\โ‹ƒ๐‘—=1๐‘–โˆ’1๐ด๐‘—, ่€ŒๅŽ้€š่ฟ‡ ctbl disjoint additivity ไธŽ monotonicity ๅฏๅพ—

    โ–ก

  3. continuous from above: ๅฆ‚ๆžœ ๐ด๐‘–โІ๐ด๐‘–+1โˆ€๐‘–โ‰ฅ2โŸน

    ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=lim๐‘–โ†’โˆž๐œ‡(๐ด๐‘–)
    Proof

    ไฝฟ็”จ same trick as 2.

    โ–ก

  4. countinuous from below: ๅฆ‚ๆžœ ๐ด๐‘–โЇ๐ด๐‘–+1โˆ€๐‘– ไธ”ๅญ˜ๅœจๆŸไธช ๐‘— ไฝฟๅพ— ๐œ‡(๐ด๐‘–)<โˆž, ๅˆ™

    ๐œ‡(โ‹‚๐‘–=1โˆž๐ด๐‘–)=lim๐‘›โ†’โˆž๐œ‡(๐ด๐‘›)
    Proof

    ๅ‰้ข็š„้ƒฝๆ— ่ง†, ็›ดๅˆฐ็ฌฌไธ€ไธช measure <โˆž ็š„้›†ๅˆ, ๆ˜ฏๅฏ่ƒฝๅ‡บ็Žฐๅœจๆœ€ๅŽ็š„ intersection ้‡Œ็š„ๆœ€ๅคง้›†ๅˆ. ๆˆ‘ไปฌ Fix ่ฟ™ไธช ๐ด๐‘—. ้€š่ฟ‡ๆž„้€ ่กฅ้›†็š„ๆ–นๅผ, ๆŠŠไบค่ฝฌไธบๅนถ, ไปŽ่€Œ็”จ (3) ๅพ—่ฏ. Define: ๐ธ๐‘–โ‰”๐ด๐‘—\๐ด๐‘–โˆ€๐‘–โ‰ฅ๐‘— ไปŽ่€Œ

    โ‹ƒ๐‘–=๐‘—โˆž๐ธ๐‘–=๐ด๐‘—\(โ‹‚๐‘–=๐‘—โˆž๐ด๐‘–)

    ่ฟ›่€Œ

    ๐œ‡(โ‹ƒ๐‘–=๐‘—โˆž๐ธ๐‘–)=๐œ‡(๐ด๐‘—)โˆ’๐œ‡(โ‹‚๐‘–=๐‘—โˆž๐ด๐‘–)

    ่ฟ›่€Œ by (3)

    ๐œ‡(โ‹‚๐‘–=1โˆž๐ด๐‘–)=๐œ‡(โ‹‚๐‘–=๐‘—โˆž๐ด๐‘–)=๐œ‡(๐ด๐‘—)โˆ’lim๐‘–โ†’โˆž๐œ‡(๐ธ๐‘–)=๐œ‡(๐ด๐‘—)โˆ’lim๐‘–โ†’โˆž(๐œ‡(๐ด๐‘—)โˆ’๐œ‡(๐ด๐‘–))=lim๐‘–โ†’โˆž๐ด๐‘–

    โ–ก

Homework 1: on ๐œŽ-algebra (39/40)

Borel vs Open

Let ๐‘‹ be a metric space such that every subset of ๐‘‹ is Borel set. Does it follow that every subset of ๐‘‹ is open? Give a proof or a counterexample.

Solution

It is not true.
Every subset of ๐‘‹ is Borel set โ‡”๐’ซ๏ธ€(๐‘‹)โŠ‚โ„ฌ๏ธ€๐‘‹. And We know โ„ฌ๏ธ€๐‘‹โŠ‚๐’ซ๏ธ€(๐‘‹), so it is equivalent to saying that โ„ฌ๏ธ€๐‘‹=๐’ซ๏ธ€(๐‘‹).
So consider this counterexample: โ„š with the Euclidean metric.
Claim: every singleton set in โ„š is closed, thus in โ„ฌ๏ธ€โ„š. This is because this only sequence in a singleton set is the point itself repeating, thus converging to itself, in the singleton set. This proves the claim.
And since โ„š is countable, every subset of โ„š is a countable union of singleton sets, thus by property of ๐œŽ-algebra, every subset of โ„š is in โ„ฌ๏ธ€โ„š. Thus:

โ„ฌ๏ธ€โ„š=๐’ซ๏ธ€(โ„š)

But clearly, not every subset in โ„š is open. Consider any singleton set, {1} as an example. Any open ball centered at 1 is not contained in {1}, thus contradicting the statement.

Restriction of a ๐œŽ-algebra to a Subset

Let ๐‘‹ be a set, and ๐‘ŒโŠ‚๐‘‹ a subset.

  • Given a ๐œŽ-algebra ๐’œ๏ธ€ on ๐‘‹, prove that

    ๐’œ๏ธ€|๐‘Œโ‰”{๐ธโˆฉ๐‘Œโˆฃ๐ธโˆˆ๐’œ๏ธ€}

    is a ๐œŽ-algebra on ๐‘Œ.

  • Given a ๐œŽ-algebra โ„ฌ๏ธ€ on ๐‘Œ, prove that there exists a ๐œŽ-algebra ๐’œ๏ธ€ on ๐‘‹ such that ๐’œ๏ธ€|๐‘Œ=โ„ฌ๏ธ€.

  • Is the ๐œŽ-algebra ๐’œ๏ธ€ in (b) unique? Give a proof or a counterexample.

Proof
    1. Since โˆ…โˆˆ๐’œ๏ธ€, โˆ…โˆฉ๐‘Œ=โˆ…, we have โˆ…โˆˆ๐’œ๏ธ€|๐‘Œ

    2. Let ๐นโˆˆ๐’œ๏ธ€|๐‘Œ, we must have ๐ธโˆˆ๐’œ๏ธ€ s.t. ๐ธโˆฉ๐‘Œ=๐น. Since ๐ธโˆˆ๐’œ๏ธ€, we have ๐‘‹\๐ธโˆˆ๐’œ๏ธ€, so ๐‘‹\๐ธโˆฉ๐‘Œโˆˆ๐’œ๏ธ€|๐‘Œ. Since ๐ธโˆฉ๐‘Œ=๐น and ๐‘Œ=(๐ธโˆฉ๐‘Œ)โŠ”((๐‘‹\๐ธ)โˆฉ๐‘Œ), it implies (๐‘‹\๐ธ)โˆฉ๐‘Œ=๐‘Œ\๐น, therefore ๐‘Œ\๐นโˆˆ๐’œ๏ธ€|๐‘Œ.

    3. Let ๐น1,๐น2,โ‹ฏ be a sequence of subsets in ๐’œ๏ธ€|๐‘Œ. Then for each ๐‘–โˆˆโ„•, we have ๐น๐‘–=๐ธ๐‘–โˆฉ๐‘Œ for some ๐ธ๐‘–โˆˆ๐’œ๏ธ€. Then โ‹ƒ๐‘–=1โˆž๐น๐‘–=โ‹ƒ๐‘–=1โˆž(๐ธ๐‘–โˆฉ๐‘Œ)=(โ‹ƒ๐‘–=1โˆž๐ธ๐‘–)โˆฉ๐‘Œโˆˆ๐’œ๏ธ€|๐‘Œ since โ‹ƒ๐‘–=1โˆž๐ธ๐‘–โˆˆ๐’œ๏ธ€.

  • Let โ„ฌ๏ธ€ be a ๐œŽ-algebra on ๐‘Œ.

    prove that there exists a ๐œŽ-algebra ๐’œ๏ธ€ on ๐‘‹ such that ๐’œ๏ธ€|๐‘Œ=โ„ฌ๏ธ€. Consider let

    ๐’œ๏ธ€โ‰”{๐ธโŠ‚๐‘‹โˆฃ๐ธโˆฉ๐‘Œโˆˆโ„ฌ๏ธ€}

    Then

    ๐’œ๏ธ€|๐‘Œ={๐ธโˆฉ๐‘Œโˆฃ๐ธ,๐‘ŒโŠ‚๐‘‹,๐ธโˆฉ๐‘Œโˆˆโ„ฌ๏ธ€}=โ„ฌ๏ธ€

    We then prove that this is a ๐œŽ-algebra on ๐‘‹.

    1. โˆ…โˆฉ๐‘Œ=โˆ… so โˆ…โˆˆ๐’œ๏ธ€.

    2. Closed under complement: Let ๐ธโˆˆ๐’œ๏ธ€, we have ๐ธโˆฉ๐‘Œโˆˆโ„ฌ๏ธ€, so ๐‘Œ\(๐ธโˆฉ๐‘Œ)=๐‘Œ\๐ธโˆˆโ„ฌ๏ธ€.
      Then (๐‘‹\๐ธ)โˆฉ๐‘Œ=๐‘Œ\๐ธโˆˆโ„ฌ๏ธ€, so ๐‘‹\๐ธโˆˆ๐’œ๏ธ€.

    3. Closed under countable union: Let ๐ธ1,๐ธ2,โ‹ฏ be a sequence in ๐’œ๏ธ€, then ๐ธ๐‘›โˆฉ๐‘Œโˆˆโ„ฌ๏ธ€. for each ๐‘›. Hence

      (โ‹ƒ๐‘›=1โˆž๐ธ๐‘›)โˆฉ๐‘Œ=โ‹ƒ๐‘›=1โˆž(๐ธ๐‘›โˆฉ๐‘Œ)โˆˆโ„ฌ๏ธ€,

      since โ„ฌ๏ธ€ is a ๐œŽ-algebra on ๐‘Œ. Therefore, โ‹ƒ๐‘›=1โˆž๐ธ๐‘›โˆˆ๐’œ๏ธ€.

  • This is not unique.
    Counterexample:

    ๐‘‹={0,1,2},๐‘Œ={0}โŠ‚๐‘‹

    Consider

    ๐ด1โ‰”๐’ซ๏ธ€(๐‘‹),๐ด2โ‰”{โˆ…,{0},{1,2},๐‘‹}

    are valid ๐œŽ-algebra on ๐‘‹.
    Then we have ๐ด1|๐‘Œ=๐ด2|๐‘Œ={โˆ…,{0}}, while ๐ด1 is different from ๐ด2.

โ–ก

Invariance Properties of the Borel ๐œŽ-algebra on โ„๐‘›

  • Prove that โ„ฌ๏ธ€(โ„๐‘›) is translation invariant, i.e., if ๐ดโŠ‚โ„๐‘› is a Borel measurable set, then

    ๐‘ก+๐ดโ‰”{๐‘ก+๐‘ฅโˆฃ๐‘ฅโˆˆ๐ด}

    is a Borel measurable set for every ๐‘กโˆˆโ„๐‘›. (Hint: For any fixed ๐‘ก, show that ๐ด={๐ตโŠ‚โ„๐‘›:๐‘ก+๐ตโˆˆโ„ฌ๏ธ€(โ„๐‘›)} is a ๐œŽ-algebra.)

  • Prove that โ„ฌ๏ธ€(โ„๐‘›) is scaling invariant, i.e., if ๐ดโŠ‚โ„๐‘› is a Borel measurable set, then

    ๐œ†๐ด={๐œ†๐‘ฅโˆฃ๐‘ฅโˆˆ๐ด}

    is a Borel measurable set for every ๐œ†โˆˆโ„.

(1)

Proof

Fix ๐‘กโˆˆโ„๐‘›. Define

๐’œ๏ธ€โ‰”{๐ตโІโ„๐‘›:๐‘ก+๐ตโˆˆโ„ฌ๏ธ€(โ„๐‘›)}.

We want to show that ๐’œ๏ธ€=โ„ฌ๏ธ€(โ„๐‘›). We first show that ๐’œ๏ธ€ is a ๐œŽ-algebra.

1. โˆ…โˆˆ๐’œ๏ธ€ since ๐‘ก+โˆ…=โˆ…โˆˆโ„ฌ๏ธ€(โ„๐‘›).

2. ๐’œ๏ธ€ is closed under complement: Let ๐ตโˆˆ๐’œ๏ธ€, then ๐‘ก+๐ตโˆˆโ„ฌ๏ธ€(โ„๐‘›). The complement (๐‘ก+๐ต)๐‘ is also in โ„ฌ๏ธ€(โ„๐‘›). Observe

๐‘ก+๐ต๐‘=๐‘ก+โ„๐‘›\๐ต=(๐‘ก+โ„๐‘›)\(๐‘ก+๐ต)=โ„๐‘›\(๐‘ก+๐ต)=(๐‘ก+๐ต)๐‘

Since ๐‘ก+๐ต is Borel, its complement is Borel, hence ๐‘ก+๐ต๐‘ is Borel, so ๐ต๐‘โˆˆ๐’œ๏ธ€.

3. ๐’œ๏ธ€ is closed under countable unions: Let ๐ต๐‘˜โˆˆ๐’œ๏ธ€ for ๐‘˜=1,2,โ€ฆ, then ๐‘ก+๐ต๐‘˜โˆˆโ„ฌ๏ธ€(โ„๐‘›). Thus

๐‘ก+โ‹ƒ๐‘˜=1โˆž๐ต๐‘˜=โ‹ƒ๐‘˜=1โˆž(๐‘ก+๐ต๐‘˜)โˆˆโ„ฌ๏ธ€(โ„๐‘›).

Hence โ‹ƒ๐‘˜=1โˆž๐ต๐‘˜โˆˆ๐’œ๏ธ€. These three properties show that ๐’œ๏ธ€ is a ๐œŽ-algebra.
Since ๐‘ก+๐‘ˆ is open if ๐‘ˆ is open in โ„๐‘›, ๐’œ๏ธ€ contains all open sets. Since โ„ฌ๏ธ€(โ„๐‘›) is the smallest ๐œŽ-algebra containing all open sets in โ„๐‘›, we have:โ„ฌ๏ธ€(โ„๐‘›)โІ๐’œ๏ธ€ Hence suppose ๐ดโˆˆโ„ฌ๏ธ€(โ„๐‘›), then ๐ดโˆˆ๐’œ๏ธ€, so ๐‘ก+๐ดโˆˆโ„ฌ๏ธ€(โ„๐‘›). This completes the proof of translation invariance.

โ–ก

(2)

Proof

Fix ๐œ†โˆˆโ„. Case 1: ๐œ†=0, then ๐œ†๐ด={0} if ๐ดโ‰ โˆ…, and ๐œ†๐ด=โˆ… otherwise. Both {0}(closed set) and โˆ… is Borel set.

Case 2: ๐œ†โ‰ 0. We define

๐’œ๏ธ€โ‰”{๐ตโІโ„๐‘›:๐œ†๐ตโˆˆโ„ฌ๏ธ€(โ„๐‘›)}.

We want to show that ๐’œ๏ธ€=โ„ฌ๏ธ€(โ„๐‘›). We first show that ๐’œ๏ธ€ is a ๐œŽ-algebra.

1. โˆ…โˆˆ๐’œ๏ธ€ since ๐œ†โˆ…=โˆ….

2. ๐’œ๏ธ€ is closed under complement: Let ๐ตโˆˆ๐’œ๏ธ€, then ๐œ†๐ตโˆˆโ„ฌ๏ธ€(โ„๐‘›), then (๐œ†๐ต)๐‘ is also in โ„ฌ๏ธ€(โ„๐‘›). Observe (๐œ†๐ต)๐‘=๐œ†๐ต๐‘, so ๐œ†๐ต๐‘โˆˆโ„ฌ๏ธ€(โ„๐‘›), therefore ๐ต๐‘โˆˆ๐’œ๏ธ€. 3. ๐’œ๏ธ€ is closed under countable unions: Let ๐ต๐‘˜โˆˆ๐’œ๏ธ€ for ๐‘˜=1,2,โ€ฆ, then ๐œ†๐ต๐‘˜โˆˆโ„ฌ๏ธ€(โ„๐‘›). Thus

๐œ†โ‹ƒ๐‘˜=1โˆž๐ต๐‘˜=โ‹ƒ๐‘˜=1โˆž(๐œ†๐ต๐‘˜)โˆˆโ„ฌ๏ธ€(โ„๐‘›).

Hence โ‹ƒ๐‘˜=1โˆž๐ต๐‘˜โˆˆ๐’œ๏ธ€. These three properties show that ๐’œ๏ธ€ is a ๐œŽ-algebra.
Since ๐œ†โ‰ 0, ๐œ†๐‘ˆ is open iff ๐‘ˆ is open in โ„๐‘›, thus ๐’œ๏ธ€ contains all open sets, so โ„ฌ๏ธ€(โ„๐‘›)โІ๐’œ๏ธ€,

Hence if ๐ดโˆˆโ„ฌ๏ธ€(โ„๐‘›), we have ๐ดโˆˆ๐’œ๏ธ€, therefore ๐œ†๐ดโˆˆโ„ฌ๏ธ€(โ„๐‘›). This completes the proof of translation invariance.

โ–ก

Hex and Such

Let ๐ดโŠ‚[0,1] be the set of real numbers in [0,1] having a hexadecimal expansion with the digit 5 appearing infinitely many times, and the โ€˜digitโ€™ E appearing at most finitely many times. Prove that ๐ด is a Borel set. (Hint: see p. 2 of Follandโ€™s book.)

Proof

Define๏ผš

๐ตโ‰”{๐‘ฅโˆˆ[0,1]โˆฃthe digit โ€™5โ€™ appears infinitely many times in the hex expansion of ๐‘ฅ}.๐ถโ‰”{๐‘ฅโˆˆ[0,1]โˆฃthe digit โ€™Eโ€™ appears at most finitely many times in the hex expansion of ๐‘ฅ}.

Then clearly

๐ด=๐ตโˆฉ๐ถ.

Hence it suffices to show that ๐ต and ๐ถ are Borel sets, since intersection of two Borel sets is a Borel set. And thus it suffices to show that ๐ต๐‘ and ๐ถ are Borel sets. Note

๐ต๐‘={๐‘ฅโˆˆ[0,1]โˆฃthe digit โ€™5โ€™ appears at most finitely many times in the hex expansion of ๐‘ฅ}

, so the proof for ๐ต๐‘ and ๐ถ are about the same. We now show ๐ต๐‘ is a Borel set: We define

๐ถ๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›โ‰”{๐‘ฅโˆˆ[0,1]:the first ๐‘› hexadecimal digits of ๐‘ฅ are ๐‘‘1,๐‘‘2,โ€ฆ,๐‘‘๐‘›},

where each ๐‘‘๐‘– is one of the 16 hexadecimal digits {0,1,2,โ€ฆ,9,๐ด,๐ต,๐ถ,๐ท,๐ธ,๐น}. Then the set contains all real numbers between ๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›16๐‘› and ๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›+116๐‘›, so actually it is an interval:

๐ถ๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›=[๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›16๐‘›,๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›+116๐‘›)

Since it is an interval, it is a Borel set on [0,1]. And we define:

๐ท๐‘={๐‘ฅ:from digit ๐‘ onward, there are no โ€™5โ€™s}.

Then we have

๐ต๐‘=โ‹ƒ๐‘=1โˆž๐ท๐‘,

So it suffices to prove that each ๐ท๐‘ is Borel set, since a countable union of Borel sets is Borel set.

Claim : any ๐ท๐‘ is a Borel set. To prove this, we fix an ๐‘ and define for each ๐‘›โ‰ฅ๐‘

๐ธ๐‘›={๐‘ฅโˆˆ[0,1]:๐‘‘๐‘›(๐‘ฅ)โ‰ 5}.

Then we have

๐ธ๐‘›=โ‹ƒ๐‘‘๐‘–โˆˆ{1,โ‹ฏ,๐น}โˆ€1โ‰ค๐‘–โ‰ค๐‘›,๐‘‘๐‘›โ‰ 5๐ถ๐‘‘1๐‘‘2โ‹ฏ๐‘‘๐‘›

Thus each ๐ธ๐‘› is a Borel set since it is a finite union of Borel set, which shows that ๐ท๐‘ is Borel set, since

๐ท๐‘=โ‹‚๐‘˜=๐‘โˆž๐ธ๐‘˜.

This finishes the proof that ๐ต๐‘ is a Borel set, and by a similar argument, ๐ถ is a Borel set, and thus ๐ด=๐ตโˆฉ๐ถ is a Borel set.

โ–ก

Admissible Annuli generating โ„ฌ๏ธ€(โ„๐‘›)

Define an admissible annulus in โ„2 to be a set of the form

{(๐‘ฅ,๐‘ฆ)โˆˆโ„2โˆฃ๐‘Ÿ2<(๐‘ฅโˆ’๐‘Ž)2+(๐‘ฆโˆ’๐‘)2<๐‘…2},

where ๐‘Ž,๐‘โˆˆโ„š, ๐‘Ÿ,๐‘…โˆˆโ„š>0, and ๐‘Ÿ<๐‘….

  • Prove that there are only countably many admissible annuli.

  • Prove that every open subset of โ„2 is a countable union of (not necessarily disjoint) admissible annuli.

  • Prove that the Borel ๐œŽ-algebra on โ„2 is generated by the collection of admissible annuli.

(1)

Proof

Let

๐ดโ‰”{all admissible annulis in โ„2}

And we define

๐‘“:โ„š4โ†’๐ด(๐‘Ž,๐‘,๐‘Ÿ,๐‘…)โ†ฆ{(๐‘ฅ,๐‘ฆ)โˆˆโ„2โˆฃ๐‘Ÿ2<(๐‘ฅโˆ’๐‘Ž)2+(๐‘ฆโˆ’๐‘)2<๐‘…2}

Since a Annuli defined by this (๐‘Ž,๐‘,๐‘Ÿ,๐‘…) is unique, this is a well-defined function; and since every admissible annulis can be defined by an element of โ„š4, this map is surjective. Therefore card(๐ด)โ‰คcard(โ„š4), so ๐ด is countable.

โ–ก

(2)

Proof

Claim 1: every open set in โ„2 is a countable union of open balls, each centered at some ๐‘žโˆˆโ„š2.
Proof for Claim 1:
Let ๐‘ˆ be an open set in โ„2. Define

โ„š๐‘ˆโ‰”๐‘ˆโˆฉโ„š2

By definition, every point in ๐‘ˆ have an open ball centered at it that is completely contained in ๐‘ˆ, so we pick such ball ๐ต๐‘Ÿ๐‘ฅ(๐‘ฅ) for each ๐‘ฅโˆˆ๐‘ˆ. Since โ„š2 is dense in โ„2, for each ๐‘ฅโˆˆ๐‘ˆ and each corresponding ๐‘Ÿ๐‘ฅ, we can find a rational point ๐‘ž๐‘ฅโˆˆโ„š2 such that |๐‘ž๐‘ฅโˆ’๐‘ฅ|<๐‘Ÿ๐‘ฅ3. (Or more generally, as small as we wish.)

Let ๐‘Ÿ๐‘ž๐‘ฅ>0 be chosen so that ๐‘Ÿ๐‘ž๐‘ฅ=๐‘Ÿ๐‘ฅ3, Then observe that ๐‘ฅโˆˆ๐ต(๐‘ž๐‘ฅ,๐‘Ÿ๐‘ž๐‘ฅ)

๐ต(๐‘ž๐‘ฅ,๐‘Ÿ๐‘ž๐‘ฅ)โŠŠ๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ)โŠ‚๐‘ˆ

which follows from the triangle inequality.

Figureย 1:

For each ๐‘žโˆˆโ„š๐‘ˆ, we define:

๐‘Ÿ๐‘ž,๐‘ ๐‘ข๐‘โ‰”sup{๐‘Ÿ๐‘ž๐‘ฅโˆฃ๐‘ž is chosen by ๐‘ฅ}

Now we have:

๐‘ˆโŠ‚โ‹ƒ๐‘žโˆˆ๐‘ˆ๐‘ž๐ต๐‘Ÿ๐‘ž,๐‘ ๐‘ข๐‘(๐‘ž)

This is because for each each ๐‘ฅโˆˆ๐‘ˆ, ๐‘ฅโˆˆ๐ต๐‘Ÿ๐‘ž๐‘ฅ(๐‘ž๐‘ฅ)โŠ‚๐ต๐‘Ÿ๐‘ž๐‘ฅ,๐‘ ๐‘ข๐‘(๐‘ž๐‘ฅ)

And we also have the other direction:

โ‹ƒ๐‘žโˆˆ๐‘ˆ๐‘ž๐ต๐‘Ÿ๐‘ž,๐‘ ๐‘ข๐‘(๐‘ž)โІ๐‘ˆ

since every ๐ต๐‘Ÿ๐‘ž(๐‘ž) is guaranteed to be the subset of some ball around some ๐‘ฅโˆˆ๐‘ˆ. All togethe we have

๐‘ˆ=โ‹ƒ๐‘žโˆˆ๐‘ˆ๐‘ž๐ต๐‘Ÿ๐‘ž,๐‘ ๐‘ข๐‘(๐‘ž)

This finishes the proof of claim 1.

Claim 2: every open ball centered at some ๐‘žโˆˆโ„š2 is a countable union of admissible annulises with the same center, together with another admissible annulis whose center is also rational. Proof for Claim 2: Let ๐‘ž=(๐‘Ž,๐‘)โˆˆโ„š2.
We have

๐ต(๐‘ž,๐‘…)\{๐‘ž}=โ‹ƒ๐‘›=1โˆž{(๐‘ฅ,๐‘ฆ):(๐‘…โˆ’1๐‘›)2<(๐‘ฅโˆ’๐‘Ž)2+(๐‘ฆโˆ’๐‘)2<๐‘…2}

-1, ่ฟ™้‡Œๅ†™็š„็•ฅๆœ‰้—ฎ้ข˜, ๅ› ไธบ ๐‘… ไธไธ€ๅฎšๆ˜ฏ rational ็š„, ไธ่ฟ‡ๆˆ‘ไปฌๅฏไปฅ็”จ density of โ„š in โ„ ๆฅๅ†™. It remains to cover the center. Let ๐‘žโ€ฒโ‰”(๐‘Žโ€ฒ,๐‘โ€ฒ)โˆˆโ„š2 such that ๐‘…/6<|๐‘žโ€ฒโˆ’๐‘ž|<๐‘…/3, ๐‘Ÿโ€ฒโ‰”๐‘…/6 and ๐‘…โ€ฒโ‰”๐‘…/2 . Then the annuli ๐ด(๐‘Žโ€ฒ,๐‘โ€ฒ,๐‘Ÿโ€ฒ,๐‘…โ€ฒ) defined by the four parameters is contained in the ๐ต(๐‘ž,๐‘…) and it covers {๐‘ž}. Therefore

๐ต(๐‘ž,๐‘…)=(โ‹ƒ๐‘›=1โˆž{(๐‘ฅ,๐‘ฆ):(๐‘…โˆ’1๐‘›)2<(๐‘ฅโˆ’๐‘Ž)2+(๐‘ฆโˆ’๐‘)2<๐‘…2})โˆช๐ด(๐‘Žโ€ฒ,๐‘โ€ฒ,๐‘Ÿโ€ฒ,๐‘…โ€ฒ)
Figureย 2:

This finishes the proof of Claim 2.
Combining Claim 1 and Claim 2, we can conclude that every open subset of โ„2 is a countable union of admissible annuli.

โ–ก

(3)

Proof

As defined,

โ„ฌ๏ธ€(โ„2)=<๐’ฏ๏ธ€๐‘š๐‘’๐‘ก๐‘Ÿ๐‘–๐‘>=<{all open sets in โ„2}>

Let

๐ดโ‰”{all admissible annulis in โ„2}

Every admissible annuli is open in โ„2, so

๐ดโŠ‚{all open sets in โ„2}

and since โ„ฌ๏ธ€(โ„2) is a ๐œŽ-algebra, we have

<๐ด>โŠ‚<{all open sets in โ„2}>=โ„ฌ๏ธ€(โ„2)

by the proposition proved in class. And by (2), any open set is a countable union of admissible annulis, therefore every open set is in <๐ด> since any countable union of sets in a ๐œŽ-algebra is still in the set. So

{all open sets in โ„2}โŠ‚<๐ด>

This finishes the proof that

<๐ด>=<{all open sets in โ„2}>=โ„ฌ๏ธ€(โ„2)

โ–ก

Nur fรผr Verrรผckte

(Itโ€™s really not necessary to attempt these problems. Do not hand them in!)

  • Let ๐‘‹ be a set, and define two operations on ๐’ซ๏ธ€(๐‘‹):

    • The โ€œproductโ€ of two subsets ๐ธ,๐นโŠ‚๐‘‹ is the intersection ๐ธโˆฉ๐น.

    • The โ€œsumโ€ of two sets ๐ธ,๐นโŠ‚๐‘‹ is the symmetric difference ๐ธฮ”๐น.

    • Prove that these operations endow ๐’ซ๏ธ€(๐‘‹) with the structure of a commutative ring. What are the additive and multiplicative units? Prove that this ring is idempotent.

    • Let us say that a nonempty subset ๐ดโŠ‚๐’ซ๏ธ€(๐‘‹) is a ring if it is closed under differences and finite unions. In other words, if ๐ธ,๐นโˆˆ๐ด, then ๐ธ\๐นโˆˆ๐ด and ๐ธโˆช๐นโˆˆ๐ด. Prove that a subset ๐ดโŠ‚๐’ซ๏ธ€(๐‘‹) is an algebra iff it is a ring containing ๐‘‹.

    • Prove that a nonempty subset ๐ดโŠ‚๐’ซ๏ธ€(๐‘‹) is a ring iff it is a subring of ๐’ซ๏ธ€(๐‘‹). Also prove that it is an algebra iff it is a subring containing the multiplicative identity.

  • Let (๐‘‹,๐’œ๏ธ€) and (๐‘Œ,โ„ฌ๏ธ€) be measurable spaces. Say that a map ๐‘“:๐‘‹โ†’๐‘Œ is measurable (with respect to the ๐œŽ-algebras ๐’œ๏ธ€ and โ„ฌ๏ธ€) if ๐‘“โˆ’1(๐ธ)โˆˆ๐’œ๏ธ€ for every ๐ธโˆˆโ„ฌ๏ธ€.

    • Prove that measurable spaces with measurable maps as morphisms form a category.

    • Try convincing an analyst that (a) is useful.

2 outer measure ไธŽ completion of a measurable space

2.1 complete measure space and outer measure [Fol 1.3, finished; 1.4]

Definition 2.5 : null set , subnull set , almost everywhere

ๅฏนไบŽ measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡), ๅ…ถไธญ ๐œ‡ ๆ˜ฏ็›ธๅบ”็š„ measure,

  1. ๆˆ‘ไปฌ็งฐ ๐ดโˆˆโ„ณ๏ธ€ ไธบไธ€ไธช null set, ๅฆ‚ๆžœ ๐œ‡(๐ด)=0;

  2. ๆˆ‘ไปฌ็งฐ ๐ตโІ๐‘‹ ไธบไธ€ไธช subnull set, ๅฆ‚ๆžœๅญ˜ๅœจๆŸไธช null set ๐ด containing it.

  3. ๆˆ‘ไปฌ็งฐไธ€ไธช statement about ๐‘‹ ๆ˜ฏ almost everywhere (a.e.) ็š„, ๅฆ‚ๆžœ่ฟ™ไธช statement ้™คไบ†ๅœจๆŸไธช null set ไธŠไน‹ๅค–, ๅœจ ๐‘‹ ไธŠๅค„ๅค„ๆˆ็ซ‹.

Definition 2.6 : complete measure space

ๆˆ‘ไปฌ็งฐ (๐‘‹,โ„ณ๏ธ€,๐œ‡) ๆ˜ฏไธ€ไธช complete measure space, ๅฆ‚ๆžœๅฎƒๅ…ถไธญ็š„ไปปๆ„ subnull set ้ƒฝๆ˜ฏ null set. (ๅณๅฎƒ measurable)

Example 2.4

ไธ€ไธช not complete ็š„ measure space ็š„ไพ‹ๅญ:

๐‘‹={1,2},โ„ณ๏ธ€=โˆ…,๐‘‹,๐œ‡(โˆ€)=0.

่ฟ™ไธชไพ‹ๅญไธญ, {1},{2} ่ฟ™ไธคไธช้›†ๅˆไธๆ˜ฏ measurable ็š„, ไฝ†ๆ˜ฏๅดๆ˜ฏ nullset (ๅ…จ้›†) ็š„ๅญ้›†.

Theorem 2.4 : every measure space can be completed

Suppose (๐‘‹,โ„ณ๏ธ€,๐œ‡) is a measure space.
Let

๐’ฉ๏ธ€โ‰”{all null sets in โ„ณ๏ธ€}

Claim:

๐‘€ฬ„โ‰”{๐ธโˆช๐นโˆฃ๐ธโˆˆโ„ณ๏ธ€,๐นโІ๐‘ for some ๐‘โˆˆ๐’ฉ๏ธ€}

is a ๐œŽ-algebra, ๅนถไธ”ๅœจ โ„ณ๏ธ€ฬ„ ไธŠๅญ˜ๅœจไธ€ไธช unique ็š„ extension ๐œ‡ฬ„ of ๐œ‡.

Proof

่ฟ™ไธ€้ƒจๅˆ†็š„ proof ไปฅๅŠ remark ๅœจ hw2. ่ฟ™้‡Œ, ๐‘€ฬ„ ็งฐไธบ completion of โ„ณ๏ธ€ with respect to ๐œ‡, ไปฅๅŠ ๐œ‡ฬ„ ็งฐไธบ completion of ๐œ‡.

โ–ก

2.1.1 outer measure

Definition 2.7 : outer measure

An outer measure on ๐‘‹ is a function ๐œ‡โˆ—:๐’ซ๏ธ€(๐‘‹)โ†’[0,โˆž) such that

  1. ๐œ‡(โŒ€)=0

  2. monotone (๐ดโŠ‚๐ตโŸน๐œ‡โˆ—(๐ด)โ‰ค๐œ‡โˆ—(๐ต))

  3. countable subadditive (๐œ‡โˆ—(โ‹ƒ๐‘–=1โˆž๐ธ๐‘–)โ‰คโˆ‘๐‘–=1โˆž๐œ‡โˆ—(๐ธ๐‘–))

ๅœจ่ฟ™ไธคไธชๆกไปถ็š„็ผฉๅ‡ไธ‹, ๆˆ‘ไปฌ่ง„ๅฎš outer measure ๅ…ทๆœ‰ monotonicity ๅ’Œ countable subadditivity. ๆณจๆ„: measure ๆœฌ่บซไนŸๆœ‰่ฟ™ไธชๆ€ง่ดจ, ่ฟ™ๆ˜ฏ measure ็š„ countable additivity ็š„ๆŽจ่ฎบ.
outer measure ็š„ๆ„ไน‰ๅœจไบŽ, ๆˆ‘ไปฌ็š„ measure ๅชๅฎšไน‰ๅœจ ๐œŽ-algebra ไธŠ, ่€Œๆˆ‘ไปฌๆƒณ่ฆ็ป™ๆฏไธชๅญ้›†้ƒฝ่ต‹ไบˆไธ€ไธช่ฟ‘ไผผไบŽๆต‹ๅบฆ็š„ไธœ่ฅฟ.

2.1.2 induce outer measure out of a "elementary length function"

Theorem 2.5 : construct outer measure out of an "elementary length function"

ๅฆ โ„ฐ๏ธ€โІ๐’ซ๏ธ€(๐‘‹) ไธบไธ€ไธชๅŒ…ๅซ โŒ€,๐‘‹ ็š„้›†ๅˆ, ๅนถๅฎšไน‰ ๐œŒ:โ„ฐ๏ธ€โ†’[0,โˆž) ไธบไธ€ไธชๆปก่ถณ ๐œŒ(โŒ€)=0 ็š„ๅ‡ฝๆ•ฐ, ๅˆ™

๐œ‡โˆ—(๐ด)=inf{โˆ‘๐‘–=1โˆž๐œŒ(๐ธ๐‘–)โˆฃ๐ธ๐‘–โˆˆโ„ฐ๏ธ€ for each i and ๐ดโІโ‹ƒ๐‘–=1โˆž๐ธ๐‘–}

is an outer measure.

Proof
  1. ๅ–ๆ‰€ๆœ‰ ๐ธ๐‘—=โŒ€, ๅพ—ๅˆฐ ๐œ‡โˆ—(โŒ€)=0

  2. monotonicity ๆ˜พ็„ถ, ๅ› ไธบๅฆ‚ๆžœ ๐ดโІ๐ต, ้‚ฃไนˆ ๐ด ๅ– inf ็š„่ฟ™ไธช้›†ๅˆๆ˜ฏๅŒ…ๅซไบŽ ๐ต ็š„, ๅ› ่€Œๅ–ๅˆฐ็š„ inf ๆ˜ฏๅฐไบŽ็ญ‰ไบŽ็š„.

  3. ่ฏๆ˜Ž ctbl subadditivity, ๆˆ‘ไปฌไฝฟ็”จ็ปๅ…ธ็š„ ๐œ–/2๐‘– argument. ่ฟ™ไธช statement ็›ด่ง‚ไธŠๆ˜ฏๆ˜พ็„ถ็š„, ๅ› ไธบๅฏนไธ€ไธช seq of sets, ๆฏไธ€ไธช้‡Œ้ข้ƒฝๆœ‰ไธ€ไธช seq of covering, ้‚ฃไนˆ่ฟ™ไธช seq of seq of covering ๆ€ปไฝ“ไนŸๆ˜ฏ่ฟ™ไธช seq union ็š„ไธ€ไธช covering. ไธ่ฟ‡ๆˆ‘ไปฌไธ่ƒฝ่ฟ™ไนˆ่ฏด, ๅ› ไธบ่ฟ™้‡Œๆœ‰ไธ€ไธช inf ๆ“ไฝœ็š„ๆขๅบ. ๆ‰€ไปฅๆˆ‘ไปฌไปค ๐œ–>0, ๅฏนไบŽๆฏไธช ๐ด๐‘– ็š„ covering (๐ธ๐‘–,๐‘˜)๐‘˜โˆˆโ„•, ๆˆ‘ไปฌไปค โˆ‘๐‘˜๐œŒ(๐ธ๐‘–,๐‘˜)โ‰ค๐œ‡โˆ—(๐ด๐‘–)+๐œ–/2๐‘–, ๆœ€ๅŽๅฏไปฅๅพ—ๅˆฐ ๐œ‡โˆ—(โ‹ƒ๐‘–๐ด๐‘–)โ‰คโˆ‘๐‘–๐œ‡โˆ—(๐ด๐‘–). ็”ฑไบŽ ๐œ– arbitrary, ๅพ—่ฏ.

โ–ก

Example 2.5

ๆˆ‘ไปฌๅ– โ„ฐ๏ธ€ ไธบ โ„ ไธŠๆ‰€ๆœ‰็š„ intervals, ๅนถๅ– ๐œŒ ไธบ interval ็š„ length, ๅฐฑๅพ—ๅˆฐไบ†ไธ€ไธชๅค–ๆต‹ๅบฆ. (ไนŸๅฐฑๆ˜ฏ Lebesgue outer measure)

2.2 ๐œ‡โˆ—-measurability and Carathรฉodoryโ€™s Theorem [Fol 1.4]

2.2.1 ๐œ‡โˆ—-measurable

Definition 2.8 : ๐œ‡โˆ—-measurable

Given outer measure ๐œ‡โˆ—, ๆˆ‘ไปฌ็งฐ ๐ดโІ๐‘‹ ๆ˜ฏ ๐œ‡โˆ—-measurable ็š„, if:

๐œ‡โˆ—(๐ธ)=๐œ‡โˆ—(๐ธโˆฉ๐ด)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘)

2.2.2 Carathรฉodoryโ€™s Theorem

Theorem 2.6 : Carathรฉodory theorem

ๅฏนไบŽไปปๆ„็š„ outer measure ๐œ‡โˆ—,

โ„ณ๏ธ€โ‰”{all ๐œ‡โˆ—-measurable sets}

is a ๐œŽ-algebra.
ๅนถไธ”, ๐œ‡โˆ—|โ„ณ๏ธ€ is a complete measure.

Proof

ๆˆ‘ไปฌ้ฆ–ๅ…ˆ่ฏๆ˜Ž่ฟ™ไธช โ„ณ๏ธ€ ๆ˜ฏไธ€ไธช ๐œŽ-algebra

  1. โŒ€โˆˆโ„ณ๏ธ€ by def.

  2. โ„ณ๏ธ€ closed under complement, by def of ๐œ‡โˆ—-measurablity. (ๅฎƒๅฏนไบŽ complement ๆ˜ฏๅฏน็งฐ็š„.)

  3. ไธบ่ฏๆ˜Ž โ„ณ๏ธ€ closed under countable union, ๆˆ‘ไปฌ้ฆ–ๅ…ˆ prove it for two sets. ๅ‡่ฎพ ๐ด,๐ตโˆˆโ„ณ๏ธ€, ไธ” disjoint. Let ๐ธโІ๐‘‹. ๆˆ‘ไปฌๅทฒ็Ÿฅ

    ๐œ‡โˆ—(๐ธ)=๐œ‡โˆ—(๐ธโˆฉ๐ด)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘)

    ๆˆ‘ไปฌ WTS: ๐œ‡โˆ—(๐ธ)=๐œ‡โˆ—(๐ธโˆฉ(๐ดโˆช๐ต))+๐œ‡โˆ—(๐ธโˆฉ(๐ดโˆช๐ต)๐‘)
    ๆˆ‘ไปฌๅฏนไบŽ ๐ธโˆฉ๐ด, ๐ธโˆฉ๐ด๐‘ ๅฏไปฅๅพ—ๅˆฐ:

    ๐œ‡โˆ—(๐ธโˆฉ๐ด)=๐œ‡โˆ—(๐ธโˆฉ๐ดโˆฉ๐ต)+๐œ‡โˆ—(๐ธโˆฉ๐ดโˆฉ๐ต๐‘)
๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘)=๐œ‡โˆ—(๐ธโˆฉ๐ดโˆฉ๐ต)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘โˆฉ๐ต๐‘)

By ๐ดโˆช๐ต=(๐ด\๐ต)โŠ”(๐ดโˆฉ๐ต)โŠ”(๐ต\๐ด), ๅฏไปฅๅพ—ๅˆฐ:

๐œ‡โˆ—(๐ธโˆฉ(๐ดโˆช๐ต))โ‰ฅ๐œ‡โˆ—(๐ธโˆฉ๐ดโˆฉ๐ต)+๐œ‡โˆ—(๐ธโˆฉ๐ดโˆฉ๐ต๐‘)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘โˆฉ๐ต)

็ป“ๅˆไปฅไธŠๅ››ไธช equations ๅฏไปฅๅพ—ๅˆฐ

๐œ‡โˆ—(๐ธ)โ‰ฅ๐œ‡โˆ—(๐ธโˆฉ(๐ดโˆช๐ต))+๐œ‡โˆ—(๐ธโˆฉ(๐ดโˆช๐ต๐‘))

ๅˆ โ‰ค by countable subadditivity ๆˆ็ซ‹, ๆˆ‘ไปฌๅพ—่ฏ closed under two union (ไปŽ่€Œ inductively closed under any finite union, โ„ณ๏ธ€ ๅ› ่€Œๆ˜ฏไธ€ไธช algebra).

(Continuing the proof:) ็Žฐๅœจๆˆ‘ไปฌๅ†ๆŠŠ่ฟ™ไธช closed under finite union ๆŽจๅนฟๅˆฐ closed under countable union, ไปฅๆ˜ ่ฏ โ„ณ๏ธ€ ๆ˜ฏไธ€ไธช ๐œŽ-algebra. ๆณจๆ„ๅˆฐ STS (suffices to show): โ„ณ๏ธ€ closed under countable disjoint union. ๅ› ไธบไปปๆ„ไธ disjoint ็š„ไธคไธช้›†ๅˆ้ƒฝๅฏไปฅๆ‹†ๅˆ†ๆˆไธ‰ไธช disjoint ็š„้›†ๅˆ.
ๆˆ‘ไปฌไปค (๐ด๐‘–) ไธบไธ€ไธช โ„ณ๏ธ€ ไธญ็š„ disjoint sequence, ๅนถๅฎšไน‰ ๐ต๐‘›โ‰”โ‹ƒ๐‘–=1๐‘›๐ด๐‘–, ๆˆ‘ไปฌ็”ฑไธŠไธ€ๆญฅ็š„็ป“่ฎบ็Ÿฅ้“, ๐ต๐‘›โˆˆโ„ณ๏ธ€ for all ๐‘›. Define ๐ตโ‰”โ‹ƒ๐‘–=1โˆž๐ด๐‘–, Let ๐ธโІ๐‘‹, WTS: ๐œ‡โˆ—(๐ธ)=๐œ‡โˆ—(๐ธโˆฉ๐ต)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘).
่€ƒ่™‘ ๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›)=๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›โˆฉ๐ด๐‘›)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›โˆฉ๐ด๐‘›๐‘)=๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘›)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›โˆ’1), ๅ› ไธบ inductively ๅฏๅพ—ๅˆฐ:

๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›)=โˆ‘๐‘–=1๐‘›๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘–)

ไปŽ่€Œ๏ผš

๐œ‡โˆ—(๐ธ)=๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›๐‘)โ‰ฅโˆ‘๐‘–=1๐‘›๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘–)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘)

by monotonicity (๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘›๐‘)โ‰ฅ๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘)), ่ฟ™้‡Œๆ˜ฏไธ€ไธช infinite sum, ๅนถไธ” true for every ๐‘›, ๅ› ่€ŒๅฏไปฅๆŽจๅนฟๅˆฐ infinity, ๅพ—ๅˆฐ

๐œ‡โˆ—(๐ธ)โ‰ฅโˆ‘๐‘–=1โˆž๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘–)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘)โ‰ฅ๐œ‡โˆ—(โ‹ƒ๐‘–=1โˆž(๐ธโˆฉ๐ด๐‘–))+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘)=๐œ‡โˆ—(๐ธโˆฉ๐ต)+๐œ‡โˆ—(๐ธโˆฉ๐ต๐‘)โ‰ฅ๐œ‡โˆ—(๐ธ)

โ–ก

This finishes the proof of โ„ณ๏ธ€ being a ๐œŽ-algebra. ๆˆ‘ไปฌๅŒๆ—ถๅ‘็Žฐ, ๐œ‡โˆ—|โ„ณ๏ธ€ ๆ˜ฏไธ€ไธช complete measure on โ„ณ๏ธ€ ๆ˜ฏไธ€ไธช trivial fact after the proof, ๅ› ไธบ taking ๐ต=๐ธ, ๅฏไปฅๅพ—ๅˆฐ

๐œ‡โˆ—(๐ต)=โˆ‘๐‘–=1โˆž๐œ‡โˆ—(๐ด๐‘–)

ๅนถไธ” by monotonicity, ๅฏนไบŽไปปๆ„็š„ ๐œ‡โˆ—(๐ด)=0, ไปปๅ– ๐ธโІ๐‘‹, ้ƒฝๆœ‰

๐œ‡โˆ—(๐ธ)โ‰ค๐œ‡โˆ—(๐ธโˆฉ๐ด)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘)=๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘)โ‰ค๐œ‡โˆ—(๐ธ)

ๅ› ่€Œ

๐œ‡โˆ—(๐ธ)=๐œ‡โˆ—(๐ธโˆฉ๐ด)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘)

ๅพ—ๅˆฐ ๐ดโˆˆโ„ณ๏ธ€. ไปŽ่€Œๅพ—่ฏ่ฟ™ๆ˜ฏไธ€ไธช complete measure.

2.3 premeasure and Hahn-Kolmogrov extension Theorem [Fol 1.4, finished]

ๆˆ‘ไปฌๅ‘็Žฐ: ๆœ‰ไบ›ๅญ้›†็ฐ‡ไธŠ็š„ "length" ๅพˆๆ˜Žๆ˜พ, ๅนถไธ”ไนŸ็ฌฆๅˆ measure ็š„ๅฎšไน‰, ไฝ†ๆ˜ฏ่ฟ™ไธชๅญ้›†็ฐ‡ๅดๅนถไธๆž„ๆˆไธ€ไธช ๐œŽ-algebra. ๆฏ”ๅฆ‚:

Example 2.6

{all half-open, half-closed intervals}โІโ„ ไธŠ, ไปฅ interval ็š„ length ไฝœไธบ measure, ๅพˆๆ˜พ็„ถ็ฌฆๅˆ measure function ็š„ๅฎšไน‰, ไฝ†ๆ˜ฏ {all half-open, half-closed intervals}โІโ„ ๅนถไธๆ˜ฏไธ€ไธช ๐œŽ-algebra, ๅ› ไธบๅฎƒๅฏไปฅ้€š่ฟ‡ ctbl union ๅ‡บ open interval, ๅนถไธๅœจ่ฟ™ไธชๅญ้›†็ฐ‡ไธญ. ไธ่ฟ‡, ่ฟ™ๆ˜ฏไธ€ไธช algebra.

ๅ› ๆญค, ๆˆ‘ไปฌๆƒณ่ฆไธ€ไธชๆ–นๆณ•ๆฅ extend a "measure" function on an algebra, to a measure on a ๐œŽ-algebra.

Definition 2.9 : premeasure

็ป™ๅฎš ๐’ซ๏ธ€(๐‘‹) ไธŠ็š„ไธ€ไธช algebra of sets ๐’œ๏ธ€0, ๆˆ‘ไปฌ็งฐ ๐œ‡0:๐’œ๏ธ€0โ†’[0,+โˆž] ไธบไธ€ไธช premeasure, if

  1. ๐œ‡0(โŒ€)=0

  2. ๐œ‡0 ctbl disjoint additive in ๐’œ๏ธ€0

2.3.1 induce outer measure out of a premeasure: preserving ๐œ‡0 on ๐’œ๏ธ€0

Proposition 2.1 : premeasure extension via induced outer measure

Any premeasure can induce an outer measure:

๐œ‡โˆ—(๐ธ)=inf{โˆ‘๐‘–=1โˆž๐œ‡0(๐ด๐‘–)โˆฃ๐ด๐‘–โˆˆ๐’œ๏ธ€0,๐ธโІโ‹ƒ๐‘–=1โˆž๐ด๐‘–}

ๅนถไธ”, we have:

๐œ‡โˆ—|๐’œ๏ธ€0=๐œ‡0

ๅนถไธ” every set in ๐’œ๏ธ€0 is ๐œ‡โˆ—-measurable.

Proof

่ฟ™ไธช outer measure ็š„ construction directly follows from Theoremย 2.5.
Proof that ๐œ‡โˆ— restricted to ๐’œ๏ธ€0 is ๐œ‡0: ไปค ๐ธโˆˆ๐’œ๏ธ€0, ๅ‡่ฎพ ๐ธโІโ‹ƒ๐‘–=1โˆž๐ด๐‘–, ๆˆ‘ไปฌไปค ๐ต๐‘›โ‰”๐ธโˆฉ(๐ด๐‘›\โ‹ƒ๐‘–=1๐‘›โˆ’1๐ด๐‘–), ๅณๆŠŠ covering intersecting ๐ธ ๅ˜ๆˆ disjoint covering (๐ต๐‘›), ไปŽ่€Œ็”ฑ ๐œ‡0 ็š„ ctbl disjoint additivity ๅฏๅพ—, ่ฟ™ไธ€ไธชๆ–ฐ covering ็š„ measure sum โˆ‘๐‘–=1โˆž๐œ‡0(๐ต๐‘–)โ‰”๐œ‡0(๐ธ). ๅนถไธ”็”ฑไบŽ ๐’œ๏ธ€0 ๆ˜ฏไธ€ไธช algebra, ่ฟ™ไบ› ๐ต๐‘› ไนŸๅœจ ๐’œ๏ธ€0 ้‡Œ้ข, ไปŽ่€Œๅฎƒๆปก่ถณ monotonicty, then ๐œ‡0(๐ธ)=โˆ‘๐‘–=1โˆž๐œ‡0(๐ต๐‘–)โ‰คโˆ‘๐‘–=1โˆž๐œ‡0(๐ด๐‘–)
Proof that every set in ๐’œ๏ธ€0 is ๐œ‡โˆ—-measurable: Fix ๐ดโˆˆ๐’œ๏ธ€0, ๆˆ‘ไปฌๅ–ไปปๆ„ ๐ธโІ๐‘‹. Let ๐œ–>0, by def of the outer measure, ๅญ˜ๅœจไธ€ไธช seq {๐ต๐‘–}๐‘–=1โˆžโІ๐’œ๏ธ€0, ไฝฟๅพ— ๐ธโІโ‹ƒ๐‘–=1โˆž๐ต๐‘– ๅนถไธ” โˆ‘๐‘–=1โˆž๐œ‡0(๐ต๐‘–)โ‰ค๐œ‡โˆ—(๐ธ)+๐œ–. ๆœ‰ disjoint additivity of ๐œ‡0 ๅฏๅพ—, โˆ‘๐‘–=1โˆž๐œ‡0(๐ต๐‘–)=โˆ‘๐‘–=1โˆž๐œ‡0(๐ต๐‘–โˆฉ๐ด)+โˆ‘๐‘–=1โˆž๐œ‡0(๐ต๐‘–โˆฉ๐ด๐‘). ไปŽ่€Œ ๐œ‡โˆ—(๐ธ)โ‰ฅ๐œ‡โˆ—(๐ธโˆฉ๐ด)+๐œ‡โˆ—(๐ธโˆฉ๐ด๐‘), ๅพ—่ฏ. (ๅฎž้™…ไธŠ่ฟ™ๆ˜ฏไธช trivial argument, ้€š่ฟ‡๐œ– argument ๆฅไธฅๆ ผ่ฏๆ˜Ž.)

โ–ก

2.3.2 Hahn-Kolmogrov Theorem

Definition 2.10 : ๐œŽ-finite measure

Let (๐‘‹,โ„ณ๏ธ€,๐œ‡) be a measure space.
ๅฆ‚ๆžœ ๐œ‡(๐‘‹)<โˆž, ๅˆ™็งฐ ๐œ‡ ๆ˜ฏ finite ็š„.
ๅฆ‚ๆžœๅญ˜ๅœจไธ€ไธช sequence (๐ธ๐‘–) in โ„ณ๏ธ€ ไฝฟๅพ— โ‹ƒ๐‘–๐ธ๐‘–=๐‘‹ ๅนถไธ”ๆฏไธช ๐œ‡(๐ธ๐‘–)<โˆž, ๅˆ™็งฐ ๐œ‡ ๆ˜ฏ ๐œŽ-finite ็š„.

Theorem 2.7 : Hahnโ€“Kolmogorov theorem

็ป™ๅฎšไธ€ไธช premeasure ๐œ‡0 on algebra โ„ณ๏ธ€0 of ๐‘‹, ไปฅๅŠๅ…ถ induced outer measure ๐œ‡โˆ—, ๆˆ‘ไปฌไปค ๆŒ‰ ๐œŽ-algebra generated by a subset ็š„ๅฎšไน‰,

โ„ณ๏ธ€โ‰”<โ„ณ๏ธ€0>

่กจ็คบ ๐œŽ-algebra generated by the algebra โ„ณ๏ธ€0.
ๅนถไปค

๐œ‡โ‰”๐œ‡โˆ—|โ„ณ๏ธ€

then we have:

  1. (๐‘‹,โ„ณ๏ธ€0,๐œ‡0) extends to (๐‘‹,โ„ณ๏ธ€,๐œ‡)
    ๅณ: ๐œ‡|โ„ณ๏ธ€0=๐œ‡0

  2. ๐œ‡|โ„ณ๏ธ€ ๆ˜ฏ the largest extension of ๐œ‡0 to โ„ณ๏ธ€ (ๅณ: ๅฏนไบŽไปปๆ„ๅ…ถไป–็š„ โ„ณ๏ธ€ ไธŠ็š„ measure ๐œˆ that extends ๐œ‡0 to โ„ณ๏ธ€, ้ƒฝๆœ‰ ๐œˆ(๐ธ)โ‰ค๐œ‡(๐ธ) for all ๐ธโˆˆโ„ณ๏ธ€);
    ๅนถไธ” if ๐œ‡0 is ๐œŽ-finite, ๅˆ™ ๐œ‡ ๆ˜ฏ the unique extension of ๐œ‡0 to โ„ณ๏ธ€.

Proof

Proof of (๐‘‹,๐’œ๏ธ€0,๐œ‡0) extends to (๐‘‹,โ„ณ๏ธ€,๐œ‡):
่ฟ™ไธช Statement directly follows from Theoremย 2.6(Carathรฉodoryโ€™s Theorem) ไปฅๅŠไธŠไธ€ไธช proposition Propositionย 2.1.
. ๆˆ‘ไปฌ้ฆ–ๅ…ˆ็”จ ๐œ‡0 induce ๅ‡บ ๐œ‡โˆ—, ๅ† restrict ๐œ‡โˆ— to โ„ณ๏ธ€โˆ—โ‰”{all ๐œ‡โˆ—-measurable sets}, ๅพ—ๅˆฐไธ€ไธช ๐œŽ-algebra โ„ณ๏ธ€โˆ—.
ๆณจๆ„ๆญคๆ—ถ: ็”ฑไธŠไธ€ไธช proposition Propositionย 2.1 ๅฏๅพ— โ„ณ๏ธ€0 ไธญๆ‰€ๆœ‰้›†ๅˆ้ƒฝๆ˜ฏ ๐œ‡โˆ—-measurable ็š„, thus ๐‘€0โІโ„ณ๏ธ€โˆ—, ็”ฑไบŽ โ„ณ๏ธ€โˆ— ๆ˜ฏไธ€ไธช ๐œŽ-algebra, ็”ฑ Lemmaย 2.2 ๅฏๅพ—: โ„ณ๏ธ€โ‰”<โ„ณ๏ธ€0>โІโ„ณ๏ธ€โˆ—.
. ็”ฑ Carathรฉodoryโ€™s Theorem ๅฏไปฅๅพ—ๅˆฐ: ๐œ‡โˆ—|โ„ณ๏ธ€โˆ— ๆ˜ฏไธ€ไธช measure, ไปŽ่€Œ ๐œ‡โ‰”๐œ‡โˆ—|โ„ณ๏ธ€ ไนŸๆ˜ฏไธ€ไธช measure(็ญ‰ไบŽๆŠŠ ๐œ‡โˆ—|โ„ณ๏ธ€โˆ— ้™ๅˆถๅœจไบ†ไธ€ไธชๆ›ดๅฐ็š„ sub-๐œŽ-algebra ไธŠ).
(Note: this is a trivial fact that if ๐‘€โˆ— is a ๐œŽ-algebra and ๐‘€โŠ‚๐‘€โˆ—is also a ๐œŽ-algebra, then ๐œ‡|๐‘€ is a measure if given that ๐œ‡ is a ๐œŽ-algebra on ๐‘€โˆ—)

Proof of ๐œ‡ being the largest extension of ๐œ‡0 to โ„ณ๏ธ€: ๅ‡่ฎพ ๐œˆ ๆ˜ฏไธ€ไธช โ„ณ๏ธ€ ไธŠ็š„ ๐œŽ-algebra s.t. ๐œˆ|โ„ณ๏ธ€0=๐œ‡0.
Let ๐ธโІโ„ณ๏ธ€. (WTS: ๐œˆ(๐ธ)โ‰ค๐œ‡(๐ธ), ๅณ๐œˆ(๐ธ)โ‰ค๐œ‡โˆ—(๐ธ) .)
็”ฑๅค–ๆต‹ๅบฆ ๐œ‡โˆ— ็š„ๅฎšไน‰, ๅฏนไบŽไปปๆ„ ๐œ–>0, ๅญ˜ๅœจไธ€ๅˆ—้›†ๅˆ {๐ด๐‘–}๐‘–=1โˆžโŠ‚๐’œ๏ธ€0 ๆปก่ถณ

๐ธโŠ‚โ‹ƒ๐‘–=1โˆž๐ด๐‘–ไธ”โˆ‘๐‘–=1โˆž๐œ‡0(๐ด๐‘–)โ‰ค๐œ‡โˆ—(๐ธ)+๐œ–.

็”ฑไบŽ ๐œˆ ๅœจ ๐’œ๏ธ€0 ไธŠๅ’Œ ๐œ‡0 ไธ€่‡ด๏ผŒๅณ

๐œˆ(๐ด๐‘–)=๐œ‡0(๐ด๐‘–)โˆ€๐‘–,

ๅ› ๆญค๏ผŒ

โˆ‘๐‘–=1โˆž๐œˆ(๐ด๐‘–)=โˆ‘๐‘–=1โˆž๐œ‡0(๐ด๐‘–)โ‰ค๐œ‡โˆ—(๐ธ)+๐œ–

ๅˆฉ็”จ ๐œˆ ็š„ additivity ๅ’Œ monotoncity ๅพ—

๐œˆ(๐ธ)โ‰ค๐œˆ(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)โ‰คโˆ‘๐‘–=1โˆž๐œˆ(๐ด๐‘–)=โˆ‘๐‘–=1โˆž๐œ‡0(๐ด๐‘–)โ‰ค๐œ‡โˆ—(๐ธ)+๐œ–

็”ฑไบŽ ๐œ– arbitrary, ๅพ—ๅˆฐ

๐œˆ(๐ธ)โ‰ค๐œ‡โˆ—(๐ธ)

(่ฏๆ˜Žๆ€่ทฏ: ๅœจ โ„ณ๏ธ€ ไธŠ ๐œ‡ ๅฐฑ็ญ‰ไบŽ ๐œ‡0 induce ็š„ๅค–ๆต‹ๅบฆ, ๅฏนไบŽๅ…ถไป–็š„ extended measure, ๅ…ถไฝœ็”จๅœจไธ€ไธช้›†ๅˆไธŠ็š„ๆต‹ๅบฆไธ€ๅฎšๅฐไบŽ็ญ‰ไบŽไปปๆ„็š„ โ„ณ๏ธ€0 covering ็š„ premeasure ๅ’Œ, ่€Œๆˆ‘ไปฌๅฏไปฅ้€š่ฟ‡ๆŽงๅˆถ่ฟ™ไธช covering ็š„ๆต‹ๅบฆๅ’ŒไธŽๅฎƒ็š„ๅค–ๆต‹ๅบฆ็š„ๅทฎ่ท(since inf), ไปŽ่€Œไฝฟๅพ—่ฟ™ไธชๆต‹ๅบฆๅฐไบŽ็ญ‰ๅฎƒ็š„ๅค–ๆต‹ๅบฆๅŠ ไธ€ไธชๆ— ้™ๅฐ็š„ ๐œ–, ไปŽ่€Œๅพ—่ฏ.)

Proof of ๐œ‡ being the unique extension of ๐œ‡0 to โ„ณ๏ธ€, provided that ๐œ‡0 is ๐œŽ-finite:
(recall ๐œ‡0 is ๐œŽ-finite ๅณ ๐œ‡0(๐‘‹)<โˆž) It remains to show that ๐œˆ(๐ธ)โ‰ฅ๐œ‡โˆ—(๐ธ).

Continuing ไธŠไธ€ไธช proof, we have:

๐œ‡โˆ—(๐ธ)โ‰ค๐œ‡โˆ—(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=๐œˆ(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=๐œˆ(๐ธ)+๐œˆ(โ‹ƒ๐‘–=1โˆž๐ด๐‘–\๐ธ)โ‰ค๐œˆ(๐ธ)+๐œ‡โˆ—(โ‹ƒ๐‘–=1โˆž๐ด๐‘–\๐ธ)

ๆˆ‘ไปฌๅช่ฆ controling ๐œ‡โˆ—(โ‹ƒ๐‘–=1โˆž๐ด๐‘–\๐ธ)=๐œ‡โˆ—(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)โˆ’๐œ‡โˆ—(๐ธ)=๐œ– ้€ผ่ฟ‘ 0, ๅณๅฏๅพ—ๅˆฐๅๅ‘็š„ไธ็ญ‰ๅผๅ…ณ็ณป.
(่ฏๆ˜Žๆ€่ทฏ: ๆˆ‘ไปฌ่ฏๆ˜Žไบ† ๐œˆ(๐ธ)โ‰ค๐œ‡โˆ—(๐ธ) ไน‹ๅŽ, ๆณจๆ„ๅˆฐ covering set ๅ’Œ ๐ธ ไน‹้—ด็š„ๅทฎ้›†็š„ ๐œˆ-measure ่‡ช็„ถไนŸๅฐไบŽ็ญ‰ไบŽ่ฟ™ไธชๅทฎ้›†็š„ ๐œ‡โˆ—-measure, which can approximate 0.)

โ–ก

Homework 2: on Carathรฉodoryโ€™s and Hahn-Holmogrov Thm(40/40)

None of the following questions will be graded. Do them, but do not hand them in.

The Borelโ€“Cantelli Lemma

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space. Let ๐ด๐‘–โˆˆ๐’œ๏ธ€ for ๐‘–โˆˆโ„•, and suppose that

โˆ‘๐‘–=1โˆž๐œ‡(๐ด๐‘–)<โˆž.

(a) Prove that ๐œ‡(limโ€‰sup๐‘–๐ด๐‘–)=0, where

limโ€‰sup๐‘–๐ด๐‘–={๐‘ฅโˆˆ๐‘‹โˆฃ๐‘ฅโˆˆ๐ด๐‘– for infinitely many ๐‘–}.

(By the way, why is limโ€‰sup๐‘–๐ด๐‘– measurable?)

(b) Conversely, is it true that if ๐ด๐‘–โˆˆ๐’œ๏ธ€ for ๐‘–โˆˆโ„•, and ๐œ‡(limโ€‰sup๐‘–๐ด๐‘–)=0, then โˆ‘๐‘–๐œ‡(๐ด๐‘–)<โˆž? Provide a proof or a counterexample. (Wrong)

The Completion of a Measure Space

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space, and set

๐’œ๏ธ€ฬ„โ‰”{๐ธโˆช๐นโˆฃ๐ธโˆˆ๐’œ๏ธ€ and ๐น is a ๐œ‡-subnull set}.

(a) Prove that ๐’œ๏ธ€ฬ„ is a ๐œŽ-algebra. (b) Define ๐œ‡ฬ„(๐ด)โ‰”๐œ‡(๐ธ) if ๐ด=๐ธโˆช๐นโˆˆ๐’œ๏ธ€ฬ„. Prove that ๐œ‡ฬ„ is a well-defined measure on ๐’œ๏ธ€ฬ„. (c) Prove that ๐œ‡ฬ„ extends ๐œ‡ (i.e., ๐œ‡ฬ„(๐ด)=๐œ‡(๐ด) if ๐ดโˆˆ๐’œ๏ธ€). (d) Prove that ๐œ‡ฬ„ is the unique extension of ๐œ‡ to (๐‘‹,๐’œ๏ธ€ฬ„). In other words, prove that if ๐œ‡โ€ฒ is another measure on (๐‘‹,๐’œ๏ธ€ฬ„) that extends ๐œ‡, then ๐œ‡โ€ฒ=๐œ‡ฬ„. (e) Prove that ๐œ‡ฬ„ is complete. (f) Suppose (๐‘‹,๐’œ๏ธ€โ€ฒ,๐œ‡โ€ฒ) is another complete measure space that extends (๐‘‹,๐’œ๏ธ€,๐œ‡) (i.e., ๐’œ๏ธ€โŠ‚๐’œ๏ธ€โ€ฒ and ๐œ‡โ€ฒ|๐’œ๏ธ€=๐œ‡). Show that ๐’œ๏ธ€ฬ„โŠ‚๐’œ๏ธ€โ€ฒ and ๐œ‡โ€ฒ|๐’œ๏ธ€ฬ„=๐œ‡ฬ„. Hint: Start by reading Theorem 1.9 in Folland.

Proof

็•ฅ.(ๅ˜ปๅ˜ป)

โ–ก

The Hahnโ€“Kolmogorov Extension as a Completion

Let (๐‘‹,๐’œ๏ธ€0,๐œ‡0) be a ๐œŽ-finite measure pre-measure space, and (๐‘‹,๐’œ๏ธ€,๐œ‡) its Hahnโ€“Kolmogorov extension. Prove that (๐‘‹,๐’œ๏ธ€,๐œ‡) is the completion of its restriction to the ๐œŽ-algebra โŸจ๐’œ๏ธ€0โŸฉ generated by ๐’œ๏ธ€0.

Proof

Proved in lec notes.

โ–ก

Some of the following questions will be graded. Do them, and do hand them in.

๐œ‡(โˆ…)=0 ็š„ๅฎšไน‰ๅนถ้ž redundant

Let (๐‘‹,๐’œ๏ธ€) be a measurable space. Is the condition ๐œ‡(โˆ…)=0 in the definition of a measure on (๐‘‹,๐’œ๏ธ€) redundant? In other words, if ๐œ‡:๐’œ๏ธ€โ†’[0,โˆž] is a function such that

๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=โˆ‘๐‘–=1โˆž๐œ‡(๐ด๐‘–),

for any disjoint subsets ๐ด๐‘–โˆˆ๐’œ๏ธ€, ๐‘–โˆˆโ„•, does it follow that ๐œ‡(โˆ…)=0? If not, what can you say?

Proof

It does not follow.
Counterexample: Consider ๐œ‡(๐ธ)=โˆžโˆ€๐ธโˆˆ๐’œ๏ธ€.
This measure satisfies the countably disjoint additivity condition, since for every disjoint sequence of sets in ๐’œ๏ธ€, ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=โˆ‘๐‘–=1โˆž๐œ‡(๐ด๐‘–)=โˆž has infinite measure.

โ–ก

measurable set seq ็š„ limit ไนŸ measurable (ไธ”ๅฆ‚ๆžœ seq tail ๐œŽ-finite โŸน limit commute )

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space, and let ๐ด๐‘–โˆˆ๐’œ๏ธ€,๐‘–โˆˆโ„•. Assume that the sets ๐ด๐‘– converge to the set ๐ดโŠ‚๐‘‹ in the sense that: - If ๐‘ฅโˆˆ๐ด, then ๐‘ฅโˆˆ๐ด๐‘– for all but finitely many ๐‘–; - If ๐‘ฅโˆ‰๐ด, then ๐‘ฅโˆ‰๐ด๐‘– for all but finitely many ๐‘–.

(a) Prove that ๐ด is measurable, that is, ๐ดโˆˆ๐’œ๏ธ€.

Proof

Deduing from the conditions: If ๐‘ฅโˆˆ๐ด, then ๐‘ฅโˆˆ๐ด๐‘– for all but finitely many ๐‘–; โŸน ๐ดโŠ‚limโ€‰inf๐ด๐‘– If ๐‘ฅโˆ‰๐ด, then ๐‘ฅโˆ‰๐ด๐‘– for all but finitely many ๐‘–. โŸน if ๐‘ฅโˆˆ๐ด๐‘– for all but finitely many ๐‘– then ๐‘ฅโˆˆ๐ด โŸน limโ€‰sup๐ด๐‘–โŠ‚๐ด Thus

limโ€‰sup๐ด๐‘–โŠ‚๐ดโŠ‚limโ€‰inf๐ด๐‘–

Claim1: For any sequence of sets (๐ด๐‘–)๐‘–โˆˆโ„•, we have

limโ€‰inf๐ด๐‘–โŠ‚limโ€‰sup๐ด๐‘– Proof of Claim 1: Follows trivially from the definition, since ๐‘ฅโˆ‰๐ด๐‘– for all but finitely many ๐‘– โŸน ๐‘ฅโˆ‰๐ด๐‘– for infinitely many ๐‘–.

Combining claim (1) with (2.1) we have

limโ€‰sup๐ด๐‘–=๐ด=limโ€‰inf๐ด๐‘–

Claim 2: For any sequence of sets (๐ด๐‘–)๐‘–โˆˆโ„• in a ๐œŽ-algebra, limโ€‰inf๐‘–๐ด๐‘– and limโ€‰sup๐‘–๐ด๐‘– is also in the ๐œŽ-algebra.
Proof of Claim 2: This follows from the def and fact that union and intersection of a countable sequence sets in a ๐œŽ-algebra is also in this ๐œŽ-algebra. We have

Figureย 3:

This finishes the proof of Claim 2.
Combining claim 2 with (2.2), ๐ดโˆˆ๐’œ๏ธ€, this finishes the proof.

โ–ก

(b) Prove that if there exists ๐‘›โ‰ฅ1 such that ๐œ‡(โ‹ƒ๐‘–=๐‘›โˆž๐ด๐‘–)<โˆž, then ๐œ‡(๐ด)=lim๐‘–๐œ‡(๐ด๐‘–).

Proof
Figureย 4:
Figureย 5:

โ–ก

(c) Give an example showing that the condition in (b) is necessary.

Solution

Let ๐œ‡ be the Lebesgue measure defined on โ„ฌ๏ธ€(โ„). We set for each ๐‘–โˆˆโ„• that

๐ด๐‘–=(๐‘–,๐‘–+1)

Since this is an interval, it is Lebesgue measurable. Note that no element of any ๐ด๐‘– show up infinitely many times in the sequence. So

limโ€‰inf๐ด๐‘–=limโ€‰sup๐ด๐‘–=โŒ€

So ๐ด=โŒ€, we have lim๐‘–๐œ‡(๐ด๐‘–)=0. But we have lim๐‘–๐œ‡(๐ด๐‘–)=1 since it is true for every ๐‘–.
In this case, ๐œ‡(โ‹ƒ๐‘–=1โˆž๐ด๐‘–)=โˆž, which causes (b) to fail.

Hint: In analysis, it is often fruitful to use limโ€‰sup and limโ€‰inf to study limits.

measure space of two elements

Let ๐‘‹ be a set with two elements, for example, ๐‘‹={๐‘‚,๐‘„}.
(a) Find all ๐œŽ-algebras on ๐‘‹.

Solution
  1. trivial ๐œŽ-algebra:

    ๐’œ๏ธ€1โ‰”{โŒ€,๐‘‹}
  2. power set:

    ๐’œ๏ธ€2โ‰”๐’ซ๏ธ€(๐‘‹)={โŒ€,{๐‘‚},{๐‘„},๐‘‹}

These are the only ๐œŽ-algebras on ๐‘‹.

(b) Let ๐’œ๏ธ€ be a ๐œŽ-algebra on ๐‘‹, and ๐œ‡ a measure on (๐‘‹,๐’œ๏ธ€). Is ๐œ‡ necessarily complete? Provide a proof or a counterexample.

Solution

It is not necessarily complete.
Cosider the trivial ๐œŽ-algebra:๐’œ๏ธ€1โ‰”{โŒ€,๐‘‹}, and set ๐œ‡ as that ๐œ‡(โŒ€)=๐œ‡(๐‘‹)=0. This makes ๐‘‹ a null set, so {๐‘‚},{๐‘„} are subnull sets, but they are not measurable by ๐œ‡.

(c) Find all outer measures ๐œ‡โˆ— on ๐‘‹. For each outer measure on ๐‘‹, find the ๐œŽ-algebra of ๐œ‡โˆ—-measurable sets (see Carathรฉodoryโ€™s theorem).

Solution

Suppose ๐œ‡โˆ— is an outer measure on ๐‘‹. Since ๐’ซ๏ธ€(๐‘‹) only has four elements: โŒ€, {๐‘‚}, {๐‘„}, ๐‘‹; and the outer measure of โŒ€ is 0, so we first parametrize ๐œ‡โˆ— by:

๐‘Žโ‰”๐œ‡โˆ—({๐‘‚}),๐‘โ‰”๐œ‡โˆ—({๐‘„}),๐‘โ‰”๐œ‡โˆ—(๐‘‹).

Then ๐œ‡โˆ— is well-defined iff it satisfies:

  1. ๐‘Ž,๐‘โ‰ค๐‘

  2. ๐‘=๐œ‡โˆ—({๐‘‚}โˆช{๐‘„})โ‰ค๐œ‡โˆ—({๐‘‚})+๐œ‡โˆ—({๐‘„})=๐‘Ž+๐‘.

Any (๐‘Ž,๐‘,๐‘)โˆˆ[0,โˆž]3 satisfying

max(๐‘Ž,๐‘)โ‰ค๐‘โ‰ค๐‘Ž+๐‘,

can make ๐œ‡โˆ— a well-defined outer measure on ๐‘‹.
Therefore

๐‘†โ‰”{all ๐œŽ-algebra on ๐‘‹}={๐œ‡โˆ—:๐’ซ๏ธ€(๐‘‹)โ†’[0,โˆž]โˆฃmax(๐œ‡โˆ—({๐‘‚}),๐œ‡โˆ—({๐‘„}))โ‰ค๐œ‡โˆ—(๐‘‹)โ‰ค๐œ‡โˆ—({๐‘‚})+๐œ‡โˆ—({๐‘„})}


Now we specify the ๐œŽ-algebra of ๐œ‡โˆ—-measurable sets for each ๐œ‡โˆ—โˆˆ๐‘†.
By Carathรฉodoryโ€™s criterion, a set ๐ธโŠ‚๐‘‹ is ๐œ‡โˆ—-measurable iff for all ๐ดโŠ‚๐‘‹,

๐œ‡โˆ—(๐ด)=๐œ‡โˆ—(๐ดโˆฉ๐ธ)+๐œ‡โˆ—(๐ดโˆฉ๐ธ๐‘).

Note that โŒ€, ๐‘‹ are always measurable since for any ๐ดโŠ‚๐‘‹, ๐ดโˆฉโŒ€=โŒ€,๐ดโˆฉ(โŒ€)๐‘=๐ด; and ๐ดโˆฉ๐‘‹=๐ด,๐ดโˆฉ(๐‘‹)๐‘=โŒ€. So it suffices to check for {๐‘‚},{๐‘„}. We first check for {๐‘‚}. {๐‘‚} is ๐œ‡โˆ—-measurable iff ๐œ‡โˆ—(๐ด)=๐œ‡โˆ—(๐ดโˆฉ{๐‘‚})+๐œ‡โˆ—(๐ดโˆฉ{๐‘‚}๐‘) for any choice of ๐ด. There are only four possibilities for ๐ด: โŒ€, {๐‘‚}, {๐‘„}, ๐‘‹.

  1. If ๐ด=โŒ€, both sides are 0, always stands.

  2. If ๐ด={๐‘‚}, then ๐œ‡โˆ—({๐‘‚})+๐œ‡โˆ—(โŒ€)=๐‘Ž+0=๐‘Ž, always stands.

  3. If ๐ด={๐‘„}, then ๐œ‡โˆ—(โŒ€)+๐œ‡โˆ—({๐‘„})=0+๐‘=๐‘, always stands.

  4. If ๐ด=๐‘‹, then ๐œ‡โˆ—(๐‘‹)=๐‘=๐œ‡โˆ—({๐‘‚})+๐œ‡โˆ—({๐‘„})=๐‘Ž+๐‘.

Therefore {๐‘‚} is ๐œ‡โˆ—-measurable iff ๐‘=๐‘Ž+๐‘. For the same reasoning, {๐‘„} is ๐œ‡โˆ—-measurable iff ๐‘=๐‘Ž+๐‘.
Thus we can conclude that:

  1. If ๐‘=๐‘Ž+๐‘, {๐œ‡โˆ—-measurable sets}=๐’ซ๏ธ€(๐‘‹).

  2. otherwise, {๐œ‡โˆ—-measurable sets}={โŒ€,๐‘‹}.

(d) Find an example of a collection โ„ฐ๏ธ€ of subsets of ๐‘‹ with โˆ…,๐‘‹โˆˆโ„ฐ๏ธ€ and a function ๐œŒ:โ„ฐ๏ธ€โ†’[0,โˆž] with ๐œŒ(โˆ…)=0 such that โ„ฐ๏ธ€โŠ„๐’œ๏ธ€, where ๐’œ๏ธ€ is the Carathรฉodory ๐œŽ-algebra for the outer measure ๐œ‡โˆ— induced by (โ„ฐ๏ธ€,๐œŒ).

Solution

Consider โ„ฐ๏ธ€={โŒ€,๐‘‹,{๐‘‚}}, with ๐œŒ such that ๐œŒ(โˆ…)=0, ๐œŒ(๐‘‹)=1, ๐œŒ({๐‘‚})=1.
The outer measure ๐œ‡โˆ— induced by ๐œ‡โˆ— is: ๐œ‡โˆ—(๐‘‹)=1, ๐œ‡โˆ—({๐‘‚})=1, ๐œ‡โˆ—({๐‘„})=1. (the inf of length sum of sets covering {๐‘„} is 1, by taking {๐‘‹} as the covering.)
Since ๐‘โ‰ ๐‘Ž+๐‘, by (4), the Carathรฉodory ๐œŽ-algebra by ๐œ‡โˆ— by โ„ฐ๏ธ€ is {โŒ€,๐‘‹}, so โ„ฐ๏ธ€โŠ„๐’œ๏ธ€.

Remark: The Hahnโ€“Kolmogorov theorem states that if โ„ฐ๏ธ€=๐’œ๏ธ€0 is an algebra and ๐œŒ=๐œ‡0 is a pre-measure, then ๐’œ๏ธ€0โŠ‚๐’œ๏ธ€. This exercise provides a counterexample when โ„ฐ๏ธ€ and ๐œŒ are general. ใ€

Hahnโ€“Kolmogorov Collapse (when ๐œ‡0 not ๐œŽ-finite)

Let ๐‘‹โŠ‚โ„ be the set of dyadic rational numbers, that is, the set of numbers of the form ๐‘Ÿ2๐‘›, where ๐‘Ÿ and ๐‘› are integers. Let ๐’œ๏ธ€0โŠ‚๐’ซ๏ธ€(๐‘‹) be the collection of finite unions of intervals of the form (๐‘Ž,๐‘]โˆฉ๐‘‹, where โˆ’โˆžโ‰ค๐‘Ž<๐‘โ‰คโˆž.

(a) Prove that ๐’œ๏ธ€0 is an algebra.

Proof
  1. โŒ€โˆˆ๐’œ๏ธ€0, since it is the empty union of intervals of the given form.

  2. Closed under complements: Let ๐ดโˆˆ๐’œ๏ธ€0. Then ๐ด is a finite union of intervals of the form (๐‘Ž๐‘–,๐‘๐‘–]โˆฉ๐‘‹. So

    ๐ด๐‘โˆฉ๐‘‹=๐‘‹\๐ด=๐‘‹\โ‹ƒ๐‘–=1๐‘›((๐‘Ž๐‘–,๐‘๐‘–]โˆฉ๐‘‹))=๐‘‹โˆฉ(โ‹‚๐‘–=1๐‘›((๐‘Ž๐‘–,๐‘๐‘–]โˆฉ๐‘‹)๐‘)=๐‘‹โˆฉ(โ‹‚๐‘–=1๐‘›((โˆ’โˆž,๐‘Ž๐‘–]โˆช(๐‘๐‘–,โˆž]))

    Note that finite intersection of intervals of the form (โˆ’โˆž,๐‘Ž๐‘–], (๐‘๐‘–,โˆž] is still of this form. Hence ๐ด๐‘โˆฉ๐‘‹โˆˆ๐’œ๏ธ€0.

  3. Closed under finite unions: Suppose ๐ด1 and ๐ด2 are finite unions of intervals ((๐‘Ž๐‘–,๐‘๐‘–]โˆฉ๐‘‹), then ๐ด1โˆช๐ด2 is still a finite union of intervals of that form. (They either merge into one such interval, so are disjoint.) Hence ๐ด1โˆช๐ด2โˆˆ๐’œ๏ธ€0. The same reasoning extends to any finite union.

This finishes the proof that ๐’œ๏ธ€0 is an algebra on ๐‘‹.

โ–ก

(b) Prove that the ๐œŽ-algebra on ๐‘‹ generated by ๐’œ๏ธ€0 equals ๐’ซ๏ธ€(๐‘‹).

Proof

Since <๐’œ๏ธ€0>โŠ‚๐’ซ๏ธ€(๐‘‹), it suffices to show that ๐’ซ๏ธ€(๐‘‹)โŠ‚<๐’œ๏ธ€0>. Note that ๐‘‹ is countable, so any set in ๐’ซ๏ธ€(๐‘‹) is a countable union of singleton sets. Thus it suffices to show that any singleton set {๐‘ฅ} where ๐‘ฅโˆˆ๐‘‹ is in <๐’œ๏ธ€0>, since if so, then any countable union of singleton sets from ๐’ซ๏ธ€(๐‘‹) is also in <๐’œ๏ธ€0>, with implies that ๐’ซ๏ธ€(๐‘‹)โŠ‚<๐’œ๏ธ€0>
Let ๐‘ฅโˆˆ๐‘‹. Then we have:

{๐‘ฅ}=โ‹‚๐‘›=1โˆž((๐‘ฅโˆ’12๐‘›,๐‘ฅ]โˆฉ๐‘‹),

since ๐‘ฅ is in the RHS set, and for any ๐‘ฆ<๐‘ฅ, we can find a ๐‘›โˆˆโ„ฌ๏ธ€ such that ๐‘ฅโˆ’12๐‘›>๐‘ฆ.
This finishes the proof that <๐’œ๏ธ€0>=๐’ซ๏ธ€(๐‘‹).

โ–ก

(c) Define ๐œ‡0:๐’œ๏ธ€0โ†’[0,โˆž] by ๐œ‡0(โˆ…)=0 and ๐œ‡0(๐ด)=โˆž for ๐ดโ‰ โˆ…. Prove that ๐œ‡0 is a pre-measure on ๐’œ๏ธ€0

Proof

It suffices to show the countable disjoint additivity.
Let (๐ด๐‘–)๐‘–โˆˆ๐’ฉ๏ธ€ be a sequence of disjoint sets in ๐’œ๏ธ€0.
Case 1: all ๐ด๐‘–=โŒ€, then โŠ”๐‘–โˆˆ๐’ฉ๏ธ€๐ด๐‘–=โŒ€, so ๐œ‡0(โŠ”๐‘–โˆˆ๐’ฉ๏ธ€๐ด๐‘–)=โˆ‘๐‘–โˆˆ๐’ฉ๏ธ€๐œ‡0(๐ด๐‘–)=0.
Case 2: ๐ด๐‘˜โ‰ โŒ€ for some ๐‘˜, then ๐œ‡0(๐ด๐‘˜)=โˆž and โŠ”๐‘–โˆˆ๐’ฉ๏ธ€๐ด๐‘–โ‰ โŒ€. Thus โˆ‘๐‘–โˆˆ๐’ฉ๏ธ€๐œ‡0(๐ด๐‘–)โ‰ฅ๐œ‡0(๐ด๐‘˜)=โˆž=๐œ‡0(โŠ”๐‘–โˆˆ๐’ฉ๏ธ€๐ด๐‘–).
The two cases cover all circumstances, finishing the proof.

โ–ก

(d) Prove that there exist infinitely many different measures ๐œ‡ on ๐’ซ๏ธ€(๐‘‹) whose restriction to ๐’œ๏ธ€0 equals ๐œ‡0.

Proof

Given ๐‘›โˆˆโ„•, We define the "n-timed counting measure" on a ๐œŽ-algebra ๐‘† as:

๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘›(๐ธ)โ‰”{๐‘›ร—#(๐ธ), if ๐ธ is finite โˆž, if ๐ธ is infinite

Claim 1: For any set ๐‘‹ and any ๐œŽ-algebra ๐‘† on ๐‘‹, the "n-timed counting measure" is a well-defined measure on ๐‘†, for all ๐‘›โˆˆโ„•.
Proof of claim 1: ๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘›(โŒ€)=0 since card(โŒ€)=0, and countable disjoint additivity trivially follows from the rule of counting.
Claim 2: for any ๐‘›โˆˆโ„•, ๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘›(๐ธ) on ๐’ซ๏ธ€(๐‘‹) restricted to ๐’œ๏ธ€0 equals ๐œ‡0. Proof of claim 2: Let ๐ธโˆˆ๐’œ๏ธ€0\โŒ€, then ๐ธ contains at least one interval of the form (๐‘Ž,๐‘]โˆฉ๐‘‹, where โˆ’โˆžโ‰ค๐‘Ž<๐‘โ‰คโˆž. Sicne ๐‘Ž<๐‘, there are infinitely many elements in (๐‘Ž,๐‘]โˆฉ๐‘‹, so ๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘›(๐ธ)=โˆž.
This finishes the proof of the original statement.

โ–ก

(e) Explain why (d) does not contradict the uniqueness part of the Hahnโ€“Kolmogorov theorem (see Theorem 1.14 in Folland).

Solution

This is because Hahnโ€“Kolmogorov theorem requires ๐œ‡0 to be ๐œŽ-finite to extend uniquely on <๐’œ๏ธ€0>. But ๐œ‡0 here is not ๐œŽ-finite.

Nur fรผr Verrรผckte (Only for nuts)

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!)

1. Let (๐‘‹,๐’œ๏ธ€,๐œ‡) and (๐‘Œ,โ„ฌ๏ธ€,๐œˆ) be measure spaces. Define a morphism from (๐‘‹,๐’œ๏ธ€,๐œ‡) to (๐‘Œ,โ„ฌ๏ธ€,๐œˆ) to be a map ๐‘“:๐‘‹โ†’๐‘Œ that is measurable, that is, ๐‘“โˆ’1(๐ต)โˆˆ๐’œ๏ธ€ for all ๐ตโˆˆโ„ฌ๏ธ€, and moreover measure preserving, in the sense that ๐œ‡(๐‘“โˆ’1(๐ต))=๐œˆ(๐ต) for all ๐ตโˆˆโ„ฌ๏ธ€.

(a) Prove that measure spaces with measure-preserving maps as morphisms form a category. Denote this category by ๐ถ3.

(b) Denote by ๐ถ1 the category of sets, and by ๐ถ2 the category of measurable spaces (see HW1). Consider the evident forgetful functors ๐ถ3โ†’๐ถ2 and ๐ถ2โ†’๐ถ1. Are these functors faithful? Are they full? Are they essentially surjective?

3 distribution function ไธŽ Lebesgue-Stieltjes measures

3.1 distribution function and Borel measures on โ„ฌ๏ธ€(โ„) [Fol 1.5]

This lecture: 1. distribution function ๆ˜ฏ increasing ไธ” right continuous ็š„, 2. ไปปๆ„ increasing ไธ” right continuous ็š„ๅ‡ฝๆ•ฐๅฏไปฅไฝœไธบ distribution function, ็”จๅฎƒๆฅๆž„้€ ๅฎƒๅฏนๅบ”็š„ measure.

3.1.1 distribution function of a locally finite (i.e. regular) Borel measure

Definition 3.11 : distribution function of ๐œ‡

็ป™ๅฎšไธ€ไธช locally finite (finite on all compact sets) ็š„ Borel measure on โ„ (ๅณ (โ„,โ„ฌ๏ธ€(๐‘๐‘…),๐œ‡)), ๆˆ‘ไปฌๅฎšไน‰:

๐น๐œ‡(๐‘ฅ)โ‰”{๐œ‡((0,๐‘ฅ]),๐‘ฅโ‰ฅ0โˆ’๐œ‡((๐‘ฅ,0]),๐‘ฅ<0

่ฟ™ไธชๅ‡ฝๆ•ฐ่ขซ็งฐไธบ ๐œ‡ ็š„ distribution function.

Proposition 3.2

ๅฎนๆ˜“ๅ‘็Žฐ: ๐น ๆ˜ฏ ๐œ‡ ็š„ distribution function, ๅฝ“ไธ”ไป…ๅฝ“ ๐œ‡((๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Ž), ไปปๅ–่ฟ™ๆ ท็š„ interval.

่ฟ™ไธคไธชๅฎšไน‰ๆ˜ฏ็ญ‰ไปท็š„.

Theorem 3.9 : distribution function is increasing and right ctn

ๅฏนไบŽ โ„ ไธŠ็š„ไปปๆ„ locally finite Borel measure ๐œ‡, ๅ…ถ distribution function ๐น๐œ‡ ้ƒฝๆ˜ฏ increasing ไธ” right continuous ็š„. (right ctn:

๐น๐œ‡(๐‘Ž)=lim๐‘ฅโ†’๐‘Ž+๐‘“(๐‘ฅ)
Proof

increasing: trivially by monotonicity of measure.
right continuous: follows from measure ็š„ ctnity. ๆญฃ่ฝดไธŠ: ๐œ‡((0,๐‘ฅ+1/๐‘›]) ็š„ sequence ๆž้™ไธบ ๐œ‡(0,๐‘ฅ]), by ctn from above; ่ดŸ่ฝดไธŠ, ๐œ‡((๐‘ฅ+1/๐‘›,0]) ็š„ sequence ๆž้™ไธบ ๐œ‡((๐‘ฅ,0]), by ctn from below.

โ–ก

3.1.2 any increasing and right ctn function is a unique distribution function

Definition 3.12 : h-interval

ๆˆ‘ไนˆๅฎšไน‰ๅฝขๅฆ‚ (๐‘Ž,๐‘], (โˆ’โˆž,๐‘] ็š„ โ„ ็š„ๅญ้›†, ไปฅๅŠ โŒ€, โ„, ไธบ h-intervals.

h-intervals ๅณๆ‰€ๆœ‰็š„ๅทฆๅผ€ๅณ้—ญๅŒบ้—ด.

Figureย 6:
Lemma 3.8 : h-intervals form an algebra and generate borel set
๐’œ๏ธ€0โ‰”{finite (disjoint) unions of h-intervals}

ๆ˜ฏไธ€ไธช algebra, ๅนถไธ”

<๐’œ๏ธ€0>=โ„ฌ๏ธ€(โ„)
Proof

trivial. follows from lec 2 ็š„ generating set of borel set on โ„.

โ–ก

Theorem 3.10 : ไปปๆ„ increasing ไธ” right ctn ๅ‡ฝๆ•ฐ้ƒฝๆ˜ฏๆŸไธช regular Borel measure ็š„ distribution ๅ‡ฝๆ•ฐ

ๅ– lemma ไธญ็š„ ๐’œ๏ธ€0. ๅฏนไบŽไปปๆ„็š„ increasing ไธ” right ctn ็š„ ๐น:โ„โ†’โ„, ๆˆ‘ไปฌ define ๐œ‡0:๐’œ๏ธ€0โ†’[0,โˆž], by:

๐œ‡0(โ‹ƒ๐‘–=1๐‘›(๐‘Ž๐‘–,๐‘๐‘–])=โˆ‘๐‘–=1๐‘›(๐น(๐‘๐‘–)โˆ’๐น(๐‘Ž๐‘–))

ๅนถ่ง„ๅฎš ๐œ‡0(0)=0, ไปฅๅŠ ๐น(โˆž)=lim๐‘ฅโ†’โˆž๐น(๐‘ฅ)
, Claim 1: ๐œ‡0 ๆ˜ฏไธ€ไธช ๐’œ๏ธ€0 ไธŠ็š„ ๐œŽ-finite premeasure.
Claim 2: (by Hahn-Kolmogrov) ๐œ‡0 extend to a locally finite Borel measure ๐œ‡๐น, ๅนถไธ” ๐œ‡๐น((๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Ž) for any h-interval, i.e. ๐น ๆ˜ฏ ๐œ‡๐น ็š„ distribution function.
Claim 3: ๐น ๆ˜ฏ ๐œ‡๐น ็š„ๅ”ฏไธ€ distribution function up to constant term, in the sense that ไปปๆ„ๅ…ถไป–็š„ such function ๐บ ๅฆ‚ๆžœไนŸๆ˜ฏ๐œ‡๐น ็š„ distribition function, ๅˆ™ๅฟ…็„ถๆœ‰ ๐นโˆ’๐บ ไธบ const.

Proof

Claim1

  1. well-definedness of ๐œ‡0: ๅฏนไบŽไธคไธช็ป“ๆžœไธ€ๆ ท็š„ union, finding common refinement ๅณๅฏ.

  2. ๐œ‡0(โŒ€)=0: ๅ› ไธบ โŒ€ ๅฐฑๆ˜ฏ (๐‘Ž,๐‘Ž].

  3. finite additivity: trivial.

  4. ๐œŽ-finiteness: each ๐œ‡0((๐‘›,๐‘›+1])<โˆž

  5. ctbl additivity: nontrivial, ไธ‹้ข่ฏฆ็ป†ๅฑ•ๅผ€.

Suppose ๐ด1,๐ด2,โ‹ฏ ๆ˜ฏ seq of disjoint h-intervals in ๐’œ๏ธ€0. Let ๐ดโ‰”โจ†๐‘–๐ด๐‘–.
WTS: ๐œ‡0(๐ด)=โˆ‘๐‘–๐œ‡0(๐ด๐‘–).
(1) WTS ๐œ‡0(๐ด)โ‰ฅโˆ‘๐‘–๐œ‡0(๐ด๐‘–) ่ฟ™ไธช direction easy. We define ๐ต๐‘›โ‰”โจ†1๐‘›๐ด๐‘–, ็”ฑ finite additivity ๅพ—ๅˆฐ: ๐œ‡0(๐ต๐‘›)=โˆ‘๐‘–๐‘›๐œ‡0(๐ด๐‘–), ไปŽ่€Œ

๐œ‡0(๐ด)=๐œ‡0(๐ต๐‘›)+๐œ‡0(๐ด\๐ต๐‘›)โ‰ฅ๐œ‡0(๐ต๐‘›)

for each ๐‘›, ็”ฑไบŽ่ฟ™ๆ˜ฏไธ€ไธช numerical seq, ๅฏไปฅ conclude ๐œ‡0(๐ด)โ‰ฅโˆ‘๐‘–๐œ‡0(๐ด๐‘–). (2) WTS โˆ‘๐‘–๐œ‡0(๐ด๐‘–)โ‰ฅ๐œ‡0(๐ด).
่ฟ™ไธช direction ่พƒ้šพ, ้œ€่ฆ็”จๅˆฐ ๐œ–/2๐‘› ็š„ argument.
For simplicity, ๆˆ‘ไปฌๅช้œ€่ฆ่€ƒ่™‘ ๐ด๐‘–=(๐‘Ž๐‘–,๐‘๐‘–] ็š„ interval ๅฝขๅผ, ๅ…ถไป–ๅฝขๅผ can trivially prove. ๅนถไธ”, ็”ฑไบŽ ๐’œ๏ธ€0 ไธญไปปไฝ•ไธ€ไธชๅ…ƒ็ด ่‡ณๅคšๅชๆœ‰ finite ไธช็ฆปๆ•ฃ็š„ h-intervals, ๆˆ‘ไปฌ suffice to assume ๐ด ๆ˜ฏไธ€ไธช h-interval.
ไปŽ่€Œ, ๆˆ‘ไปฌไนŸๅฏไปฅ denote ๐ด=(๐‘Ž,๐‘].
Let ๐œ–>0.
By ๐น ็š„ increasing ๅ’Œ right ctn, ๅญ˜ๅœจ ๐›ฟ,๐›ฟ๐‘– s.t.

๐น(๐‘Ž+๐›ฟ)โˆ’๐น(๐‘Ž)โ‰ค๐œ–

ๅŒๆ ทๅœฐ, ๅฏนไบŽๆฏไธช ๐ด๐‘–. ๆˆ‘ไปฌ้ƒฝๅฏไปฅๆ‰พๅˆฐ ๐›ฟ๐‘– ไฝฟๅพ—

๐น(๐‘๐‘–+๐›ฟ๐‘–)โˆ’๐น(๐‘๐‘–)โ‰ค๐œ–2๐‘–

ไบŽๆ˜ฏ (๐‘Ž๐‘–,๐‘๐‘–+๐›ฟ๐‘–)๐‘–โˆˆโ„• ๅฐฑๅฝขๆˆไบ†ไธ€ไธช open covering for [๐‘Ž+๐›ฟ,๐‘]. By cptness, ๅญ˜ๅœจไธ€ไธช finite subcovering (๐‘Ž๐‘–,๐‘๐‘–+๐›ฟ๐‘–)1โ‰ค๐‘–โ‰ค๐‘.
By relabelling, ๆˆ‘ไปฌ suppose ๐ด๐‘– ๆ˜ฏไปŽๅทฆๅˆฐๅณๆŽ’ๅบ็š„. ไบŽๆ˜ฏๆฏไธช ๐‘๐‘–+๐›ฟ๐‘– ้ƒฝๅค„ไบŽไธ‹ไธ€ไธช ๐ด๐‘–+1 ไน‹ๅ†….

Figureย 7:

ไปŽ่€Œ:

๐œ‡0(๐ด)โ‰ค๐น(๐‘)โˆ’๐น(๐‘Ž+๐›ฟ)โˆ’๐œ–โ‰ค๐น(๐‘๐‘+๐›ฟ๐‘)โˆ’๐น(๐‘Ž1)+๐œ–=๐น(๐‘๐‘+๐›ฟ๐‘)โˆ’๐น(๐‘Ž๐‘)+โˆ‘1๐‘โˆ’1(๐น(๐‘Ž๐‘–+1)โˆ’๐น(๐‘Ž๐‘–))+๐œ–โ‰ค๐น(๐‘๐‘+๐›ฟ๐‘)โˆ’๐น(๐‘Ž๐‘)+โˆ‘1๐‘โˆ’1(๐น(๐‘๐‘–+๐›ฟ๐‘–)โˆ’๐น(๐‘Ž๐‘–))+๐œ–<โˆ‘1๐‘(๐น(๐‘๐‘–)โˆ’๐ด(๐‘Ž๐‘–)+๐œ–2๐‘–)+๐œ–<โˆ‘1โˆž๐œ‡0(๐ด๐‘–)+2๐œ–

Claim 2, 3 ้ƒฝ directly follows from Hahn-Komogrov Thm.

โ–ก

Example 3.7

ๆˆ‘ไปฌๅทฒ็ป่ฏๆ˜Ž, ไปŽไปปๆ„็š„ increasing ไธ” right ctn ็š„ๅ‡ฝๆ•ฐ้ƒฝๅฏไปฅๆž„้€ ๅ‡บไธ€ไธชไปฅๅ…ถไธบ distribution function ็š„ locally finite Borel measure on โ„, ๅ› ่€Œๆˆ‘ไปฌ็ฎ€็งฐ่ฟ™ๆ ท็š„ๅ‡ฝๆ•ฐ้ƒฝๅซๅš distribution function.
ไปฅไธ‹ไธบไธคไธช distribution function ็š„ไพ‹ๅญ:
1. Heaviside function

๐ป(๐‘ฅ)={1,๐‘ฅโ‰ฅ00,๐‘ฅ<0

2. ๆˆ‘ไปฌๅฐ† โ„š ไปฅๆŸ็งๅฝขๅผๅˆ—ๅ‡บ: โ„š={๐‘ž1,๐‘ž2,โ‹ฏ} ่€ŒๅŽๅฎšไน‰:

๐น(๐‘ฅ)โ‰”โˆ‘๐‘–=1โˆž2โˆ’๐‘›๐ป(๐‘ฅโˆ’๐‘Ÿ๐‘›)โˆˆ(0,1)

่ฟ™ไธชๅ‡ฝๆ•ฐ้€š่ฟ‡ๆœ‰็†ๆ•ฐ็š„ๆฌกๅบ็ป™ๆฏไธชๆœ‰็†ๆ•ฐ่ต‹ไบ†ไธ€ไธช"weight", ๅนถๅฏนไบŽๆฏไธช๐‘ฅ, ๆŠŠๆ‰€ๆœ‰ๆœ‰็†ๆ•ฐๅˆ†ไธบ >๐‘ฅ ๅ’Œ โ‰ค๐‘ฅ ็š„ไธค้ƒจๅˆ†, ๅชๆŠŠ โ‰ค๐‘ฅ ็š„้‚ฃ้ƒจๅˆ†ๆœ‰็†ๆ•ฐ็š„ๆƒ้‡็ฎ—่ฟ› ๐น(๐‘ฅ). ไบŽๆ˜ฏ ๐‘ฅ ่ถŠๅคง, ่ขซ็ฎ—่ฟ› ๐น(๐‘ฅ) ็š„ๆœ‰็†ๆ•ฐ่ถŠๅคš, ๐น(๐‘ฅ) ๅฐฑ่ถŠๅคง. (่™ฝ็„ถๆฏไธชๆœ‰็†ๆ•ฐ็š„ๆƒ้‡ๆ˜ฏไนฑ็š„). ่ฟ™ไธชๅ‡ฝๆ•ฐๅœจๆฏไธ€็‚นไธŠ้ƒฝ discrete.
่ฟ™ไธช่ฟ‡็จ‹ๅฏๆŽจๅนฟ, ไธๅ– โ„š ่€Œๅ–ไปปๆ„็š„ countable sets in โ„ ไฝœไธบๅ‚็…ง.

ๆœฌ lec ๆ€ป็ป“: ้€š่ฟ‡็›ดๆŽฅๅฎšไน‰ distribution function ๆฅๅพ—ๅˆฐ็š„ measure, ๅฎžๅˆ™ๅฐฑๆ˜ฏไธๅŒไบŽ็›ดๆŽฅๅ– interval ้•ฟๅบฆ, ๆˆ‘ไปฌ้šๆ€งๅœฐ็ป™ๆฏไธช็‚นไธ€ไธช mass (็ฑปไผผๆฆ‚็އๅฏ†ๅบฆ), ไปŽ่€ŒๆŠŠๅŒบ้—ด็š„้•ฟๅบฆไธญๆฏไธ€ไธช็‚นๅŠ ไธŠไธ€ไธชๆƒ้‡. ๆœ€ๅŽๅฝขๆˆไธ€ไธชไธไธ€ๅฎšๅ‡ๅŒ€็š„ measure. ่ฟ™ไธช distribution ็š„ๅˆ†ๅธƒๆ›ฒ็บฟๅ†ณๅฎšไบ†่ฟ™ไธช measure.

3.2 Lebesgue-Stieltjes measure [Fol 1.5, finished]

็ป™ๅฎšไธ€ไธช increasing ไธ” right ctn ็š„ๅ‡ฝๆ•ฐ ๐น, ๆˆ‘ไปฌๅทฒ็ปๅฑ•็คบไบ†็”จๅฎƒไฝœไธบ distribution function ๆฅ induce ๅ‡บไธ€ไธช regular Borel measure ๐œ‡๐น on โ„ฌ๏ธ€(โ„).
ๅœจๆž„้€ ่ฟ™ไธชๅ‡ฝๆ•ฐๆ—ถ, ๆˆ‘ไปฌไฝฟ็”จ็š„ๆ˜ฏ็”จ premeasure ๐’œ๏ธ€0 (of all finite unions of h-intervals), ไฝฟ็”จ Hahn-Kolmogrov ๆฅ induce outer measure ๐œ‡๐นโˆ—, ๅ†ๆŠŠ restrict ๅฎƒๅˆฐ <๐’œ๏ธ€0>, ๅณ โ„ฌ๏ธ€(โ„) ไธŠ, ่Žทๅพ—็š„ measure. ่ฟ™ไธ€ไธช measure ๆ˜ฏไธ€ไธช Borel measure, ไฝ†ๆ˜ฏๅฎƒๅนถไธ complete.
recall in lec 6: ๆˆ‘ไปฌๅ…ถๅฎžๅฏไปฅ complete ่ฟ™ไธช measure, ๅช้œ€่ฆๅœจ็ฌฌไบŒๆญฅ, ็”จ premeasure ๐’œ๏ธ€0 induce ๅ‡บ outer measure ๅŽ, ไธ่ฆ restrict ๅฎƒๅˆฐ โ„ฌ๏ธ€(โ„) ไธŠ, ่€Œๆ˜ฏ restrict ๅˆฐๅ– โ„ณ๏ธ€๐œ‡โ‰”{all ๐œ‡๐นโˆ—-measurable set} ไธŠ, ๅพ—ๅˆฐ็š„ๅฐฑๆ˜ฏ completion of ๐œ‡๐น, ๅณ

(โ„,โ„ณ๏ธ€๐œ‡,๐œ‡๐นฬ„)

ๅ…ถไธญ, ๐’œ๏ธ€โˆ— ๆ˜ฏ <๐’œ๏ธ€0> ๅณ โ„ฌ๏ธ€(โ„) ็š„ proper super set. ๆˆ‘ไปฌๆŠŠ่ฟ™ไธช completed measure ๅซๅš Lebesgue Stieltjes measure associated with ๐น, ๅนถ็”จ ๐œ‡๐น ๆฅๆŒ‡ไปฃๅฎƒ. (ๅˆšๆ‰, ๆˆ‘ไปฌๆŠŠๆœชๅฎŒๅค‡็š„ measure ๅซๅš ๐œ‡๐น, ไฝ†็Žฐๅœจๆˆ‘ไปฌไธๅ†ไฝฟ็”จ่ฟ™ไธช measure, ่€Œๆ˜ฏไฝฟ็”จๅฎƒ็š„ completion, ๅนถ่ฝฌ่€Œ็งฐๅฎƒ็š„ completion (๐œ‡๐นฬ„) ไธบ ๐œ‡๐น.)

Regular Borel measure โ†’ completion LS measure
Definition 3.13 : Lebesgue-Stieltjes measure associated with ๐น

็ป™ๅฎšไธ€ไธช distribution function ๐น, ๆˆ‘ไปฌไฝฟ็”จๅฎƒๆฅๅฎšไน‰ h-intervals ็š„ premeasure ๐œ‡0, ๅนถๆŠŠ่ฟ™ไธช premeasure induce ๅ‡บ็š„ outer measure ๐œ‡โˆ— ้™ๅˆถๅœจ

โ„ณ๏ธ€๐œ‡โ‰”{all ๐œ‡โˆ—-measurable set}

ไธŠ, ็”ฑ Carathรฉodory Thm ๅพ—ๅฎƒๆ˜ฏ complete ็š„. ็งฐ่ฟ™ไธช complete ็š„ measure

๐œ‡๐นโ‰”๐œ‡โˆ—|โ„ณ๏ธ€๐œ‡

ไธบ Lebesgue Stieltjes measure associated with ๐น.

3.2.1 inner and outer regularity of LS measure

่™ฝ็„ถๆˆ‘ไปฌไฝฟ็”จ h-intervals ๆฅ induce ไบ†่ฟ™ไธช measure, ไฝ†ๆ˜ฏๅฎž้™…ไธŠๆˆ‘ไปฌๅœจ่กจ็คบ measure ๆ—ถ,ๅฏไปฅ็”จ open intervals ๆฅไปฃๆ›ฟ h-intervals:

Lemma 3.9 : open-interval covers for Lebesgueโ€“Stieltjes measure

ๅ›บๅฎšไธ€ไธช Lebesgue-Stieltjes measure associated with ๐น, ไปปๆ„ ๐ธโˆˆโ„ณ๏ธ€๐œ‡, ๅฎƒ็š„ measure ็ญ‰ไบŽ:

๐œ‡๐น(๐ธ)=inf{โˆ‘1โˆž(๐น(๐‘๐‘–)โˆ’๐น(๐‘Ž๐‘–))โˆฃ๐ธโІโ‹ƒ1โˆž(๐‘Ž๐‘–,๐‘๐‘–)}
Proof

ๆฏไธช open interval ้ƒฝ็ญ‰ไบŽ a ctbl disjoint union of h-intervals, ไปŽ่€Œๆ˜ฏๅœจ่ฟ™ไธช่ขซๅ– inf ้›†ๅˆๅ†…็š„; ๆ‰€ไปฅๅช้œ€่ฆ่ฏๆ˜Ž่ƒฝๅ–ๅˆฐ่ฟ™ไธช inf ๅณๅฏ. Fix ๐œ–>0, ๆˆ‘ไปฌๆ นๆฎๅฎšไน‰ๅฏไปฅๅ–ๅˆฐไธ€ไธช seq (๐‘Ž๐‘–,๐‘๐‘–] ไฝฟๅพ—ๅฎƒ measure sum โ‰ค๐œ‡(๐ธ)+๐œ–/2, ่€Œๆˆ‘ไปฌๅฏนไบŽๆฏไธช ๐‘–, ๅœจ interval ็š„ๅณ่พนๅ†ๅ–ไธ€ไธช <๐œ–/2๐‘–+1 ็š„ ๐›ฟ๐‘–, ๅฐฑๅ˜ๆˆไบ†ไธ€ไธช open interval, ๅนถไธ”ๆœ€ๅŽ่ท็ฆป่ฟ™ไธช h-interval seq ็š„ measure sum ๅทฎ่ท่‡ณๅคš ๐œ–/2. ไปŽ่€Œๅพ—่ฏ.

โ–ก

Theorem 3.11 : outer regularity

ๅฏนไบŽไธ€ไธช Lebesgue-Stieltjes measure associated with ๐น, ไปปๆ„ ๐ธโˆˆโ„ณ๏ธ€๐œ‡, ๅฎƒ็š„ measure ็ญ‰ไบŽ:

๐œ‡๐น(๐ธ)=inf{๐œ‡๐น(๐‘ˆ)โˆฃ๐‘ˆ open , and ๐ธโІ๐‘ˆ}
Proof

Directly follows from lemma. ้ฆ–ๅ…ˆ, by monotonicity, ไธ€ไธชๅŒ…ๅซ ๐ธ ็š„ๅผ€้›† ๐‘ˆ ็š„ ๐œ‡๐น ไธ€ๅฎšๆฏ” ๐ธ ็š„ๅคง. ๅนถไธ”, ๅฏนไบŽไปปๆ„็š„ ๐œ–>0, ้ƒฝๅฏไปฅๆ‰พๅˆฐไธ€ไธช open covering ไฝฟๅพ— measure sum <๐œ‡๐น(๐ธ)+๐œ–, by def.

โ–ก

Theorem 3.12 : inner regularity

ๅฏนไบŽไธ€ไธช Lebesgue-Stieltjes measure associated with ๐น, ไปปๆ„ ๐ธโˆˆโ„ณ๏ธ€๐œ‡, ๅฎƒ็š„ measure ็ญ‰ไบŽ:

๐œ‡๐น(๐ธ)=sup{๐œ‡๐น(๐พ)โˆฃ๐พ compact , and ๐พโІ๐ธ}
Proof

้ฆ–ๅ…ˆ่ฏๆ˜Ž ๐ธ bounded ็š„ case. ๅ‡่ฎพ ๐ธ bdd.
ๅฆ‚ๆžœ ๐ธ closed, ๅˆ™ ๐ธ cpt, trivially true.
ๅฆ‚ๆžœ ๐ธ open, ้‚ฃไนˆ ๐ธ ็š„ bounadry ๆ˜ฏ closed (cpt) ็š„, ไปŽ่€Œ ๐œ•๐ธโˆˆโ„ณ๏ธ€๐œ‡ ๆˆ‘ไปฌ let ๐œ–>0. ๆˆ‘ไปฌๅฏน ๐œ•๐ธ ไฝฟ็”จ outer regularity, ๅฏไปฅๅ–ไธ€ไธช open set ๐‘ˆ covering ๐œ•๐ธ, ๅนถไธ”ไฝฟๅพ— ๐œ‡๐น(๐‘ˆ)โ‰ค๐œ‡๐น(๐ธ)+๐œ–
ๆญคๆ—ถๅ– ๐พโ‰”๐ธ\๐‘ˆ, ๆˆ‘ไปฌๅ‘็Žฐ่ฟ™ๆ˜ฏไธ€ไธช approximating ๐ธ ็š„ compact set, ๅนถไธ”ๆœ‰:

๐ธ=๐พโŠ”(๐‘ˆโˆฉ๐ธ)

ไปŽ่€Œ:

Figureย 8:

่€ŒๅฏนไบŽ unbounded ็š„ case, ็›ดๆŽฅ็”ฑ

๐ธ=โจ†๐‘—(๐ธโˆฉ(๐‘—,๐‘—+1])

ๅพ—ๅˆฐ.

โ–ก

3.2.2 Lebesgue-Stieltjes measurable ็š„็ญ‰ไปทๆกไปถ

Definition 3.14 : ๐บ๐›ฟ,๐น๐œŽ sets

Topological space ไธญ, ไธ€ไธช coutable intersection of open sets ่ขซ็งฐไธบไธ€ไธช ๐บ๐›ฟ set, ไธ€ไธช countable union of closed sets ่ขซ็งฐไธบไธ€ไธช ๐น๐œŽ set.

Theorem 3.13 : Lebesgue-Stieltjes measurable ็š„็ญ‰ไปทๆกไปถ

TFAE:

  1. ๐ธโˆˆโ„ณ๏ธ€๐œ‡
  2. ๅญ˜ๅœจไธ€ไธช ๐บ๐›ฟ set ๐‘‰ ไปฅๅŠไธ€ไธช measure zero set ๐‘1 (๐œ‡๐น(๐‘1)=0) ไฝฟๅพ—

    ๐ธ=๐‘‰\๐‘1
  3. ๅญ˜ๅœจไธ€ไธช ๐น๐œŽ set ๐ป ไปฅๅŠไธ€ไธช measure zero set ๐‘2 (๐œ‡๐น(๐‘2)=0) ไฝฟๅพ—

    ๐ธ=๐ปโˆช๐‘2
  4. ๅญ˜ๅœจไธ€ไธช open set ๐‘ˆ ไฝฟๅพ—ๅฏนไบŽไปปๆ„็š„ ๐œ–>0, ้ƒฝๆœ‰

    ๐œ‡โˆ—(๐‘ˆ\๐ธ)<๐œ–
Proof

็”ฑ (ii) ๅ’Œ (iii) ๆŽจๅพ— (i) ๆ˜ฏ trivial ็š„. ่ฟ™ๆ˜ฏๅ› ไธบ LS measure ๆ˜ฏ complete measure, ไปปๆ„ null set ้ƒฝๆ˜ฏ measurable ็š„. ็”ฑ (i) ๆŽจ (ii) ๅ’Œ (iii): follows from outer ไธŽ inner regularity. ๅ‡่ฎพ ๐ธ ๆ˜ฏ LS-measurable ็š„, ๆˆ‘ไปฌ็›ดๆŽฅๅ–ไธ€ไธช inner seq of cpt subsets ไปฅๅŠไธ€ไธช outer seq of open super sets, ไฝฟๅพ—

๐œ‡๐น(๐‘ˆ๐‘—)โˆ’12๐‘–โ‰ค๐œ‡๐น(๐ธ)โ‰ค๐œ‡๐น(๐พ๐‘—)+12๐‘–

ไบŽๆ˜ฏๅฐฑๅพ—ๅˆฐ: ๐‘‰โ‰”โ‹‚๐‘–๐‘ˆ๐‘–, ๐ปโ‰”โ‹ƒ๐‘–๐พ๐‘–, ไธŽ ๐ธ ็š„ๅทฎ้›†้ƒฝๆ˜ฏไธ€ไธช null set. ๅนถไธ”ๅฎƒไปฌๅˆ†ๅˆซไธบ ๐บ๐›ฟ ๅ’Œ ๐น๐œŽ sets.

โ–ก

3.2.3 Lebesgue measure and its invariance properties

Definition 3.15 : Lebesgue measure

Lebesgue measure ๅณ Lebesgue-Stieltjes measure associated with ๐น(๐‘ฅ)=๐‘ฅ. ๆˆ‘ไปฌ็”จ ๐‘šโ‰”๐œ‡๐น ๆฅ่กจ็คบๅฎƒ, ๅนถ็”จ โ„’๏ธ€โ‰”โ„ณ๏ธ€๐‘š ๆฅ่กจ็คบๆ‰€ๆœ‰็š„ Lebesgue measurable sets.
ไปŽ่€Œ โ„ ไธŠ็š„ Lebesgue measure space ่กจ็คบไธบ:

(โ„,โ„’๏ธ€,๐‘š)
Theorem 3.14 : โ„’๏ธ€ preserves translation and scaling

if ๐ธโˆˆโ„’๏ธ€ โ‡’ ๐ธ+๐‘ ,๐‘Ÿ๐ธโˆˆโ„’๏ธ€ โˆ€๐‘ ,๐‘Ÿโˆˆโ„.
ๅนถไธ”, ๐‘š(๐ธ+๐‘ )=๐‘š(๐ธ),๐‘š(๐‘Ÿ๐ธ)=|๐‘Ÿ|๐‘š(๐ธ)

Proof

้ฆ–ๅ…ˆ, ๅฆ‚ๆžœ ๐ธโˆˆโ„ฌ๏ธ€(โ„), ้‚ฃไนˆ by hw 1, ๆˆ‘ไปฌ่ฏๆ˜Žไบ† โ„ฌ๏ธ€(โ„) ๆ˜ฏ closed under translation ๅ’Œ scaling ็š„, ๅ› ่€Œ ๐‘Ÿ๐ธ,๐ธ+๐‘ โˆˆโ„ฌ๏ธ€(โ„).
ๆˆ‘ไปฌ define on ๐’œ๏ธ€0โ‰” {finite union of h-intervals}:

๐‘š๐‘ (๐ธ)โ‰”๐‘š(๐ธ+๐‘ )๐‘š๐‘Ÿ(๐ธ)โ‰”๐‘š(๐‘Ÿ๐ธ)

ๆ˜พ็„ถ, ่ฟ™ไธคไธชๅ‡ฝๆ•ฐ agree with ๐‘š,|๐‘Ÿ|๐‘š. ็”ฑไบŽ ๐‘š ๆ˜ฏ ๐œŽ-finite ็š„, ไปŽ่€Œ by Hahn-Kolmogrov, ๅฎƒ uniquely extend to โ„ฌ๏ธ€(โ„). ๅ› ่€Œ, ๐‘š๐‘  ๅœจ โ„ฌ๏ธ€(โ„) ไธŠๅ’Œ ๐‘š ็›ธ็ญ‰, ๐‘š๐‘Ÿ ๅœจ โ„ฌ๏ธ€(โ„) ไธŠๅ’Œ |๐‘Ÿ|๐‘š ็›ธ็ญ‰. ๅนถไธ”, ๆˆ‘ไปฌ็Ÿฅ้“ (โ„,โ„’๏ธ€,๐‘š) ๆ˜ฏ completion of (โ„,โ„ฌ๏ธ€(โ„),๐‘š), ๅ› ่€Œ ๐‘š๐‘  ไนŸๅŒๆ ท complete to ๐‘š on โ„’๏ธ€. (ๅŒ็†, ๐‘š๐‘Ÿ ไนŸๅŒๆ ท complete to |๐‘Ÿ|๐‘š on โ„’๏ธ€)

โ–ก

Homework 3: on Lebesgue-Stieljes measures(30/40)

None of the following questions will be graded. Do them, but do not hand them in.

Fun facts about increasing functions

Let ๐น:โ„โ†’โ„ be an increasing function, that is, ๐น(๐‘ฅ)โ‰ค๐น(๐‘ฆ) whenever ๐‘ฅโ‰ค๐‘ฆ.

  • Prove that the following limits exist (and make sure you understand the definitions):

    • ๐น(๐‘Žโˆ’)โ‰”lim๐‘ฅโ†’๐‘Žโˆ’๐น(๐‘ฅ)โˆˆโ„ and ๐น(๐‘Ž+)โ‰”lim๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ)โˆˆโ„ for ๐‘Žโˆˆโ„;

    • ๐น(โˆž)โ‰”lim๐‘ฅโ†’โˆž๐น(๐‘ฅ)โˆˆ(โˆ’โˆž,โˆž];

    • ๐น(โˆ’โˆž)โ‰”lim๐‘ฅโ†’โˆ’โˆž๐น(๐‘ฅ)โˆˆ[โˆ’โˆž,โˆž).

  • Fix any ๐‘Žโˆˆโ„.

    • Prove that ๐น(๐‘Žโˆ’)โ‰ค๐น(๐‘Ž)โ‰ค๐น(๐‘Ž+);

    • Prove that ๐น is continuous at ๐‘Ž iff ๐น(๐‘Žโˆ’)=๐น(๐‘Ž+).

    We say that a function is left continuous if ๐น(๐‘Žโˆ’)=๐น(๐‘Ž) for every ๐‘Žโˆˆโ„. It is right-continuous if instead ๐น(๐‘Ž+)=๐น(๐‘Ž) for every ๐‘Žโˆˆโ„.

  • If ๐‘‹ is a metric space (or, more generally, a topological space), then a function ๐‘“:๐‘‹โ†’โ„ is upper semicontinuous if the set {๐‘ฅโˆˆ๐‘‹โˆฃ๐‘“(๐‘ฅ)<๐‘Ž} is open for every ๐‘Žโˆˆโ„. It is lower semicontinuous if instead the set {๐‘ฅโˆˆ๐‘‹โˆฃ๐‘“(๐‘ฅ)>๐‘Ž} is open for every ๐‘Žโˆˆโ„. Prove that our function ๐น:โ„โ†’โ„ is right-continuous (resp. left continuous) iff it is upper semicontinuous (resp.ย lower semicontinuous). Give an example showing that this is no longer true if ๐น is not assumed increasing.

  • Prove that the following are equivalent:

    • ๐น is surjective;

    • ๐น is continuous, ๐น(โˆž)=โˆž, and ๐น(โˆ’โˆž)=โˆ’โˆž.

  • Let ๐ดโŠ‚โ„ be the set of points where ๐น fails to be continuous. Prove that ๐ด is a countable (i.e. empty, finite, or countably infinite) set. Hint: prove that for any integers ๐‘š,๐‘›โ‰ฅ1, the set of points ๐‘ฅโˆˆ[โˆ’๐‘š,๐‘š] where ๐น(๐‘ฅ+)โˆ’๐น(๐‘ฅโˆ’)โ‰ฅ1/๐‘› is finite.

Locally finite measures

If ๐‘‹ is a metric space (or, more generally, a topological space), then a Borel measure ๐œ‡ on ๐‘‹ is said to be locally finite if ๐œ‡(๐พ)<โˆž for every compact set ๐พโŠ‚๐‘‹. Now let ๐œ‡ be a Borel measure on โ„, that is ๐œ‡:โ„ฌ๏ธ€(โ„)โ†’[0,โˆž] satisfies ๐œ‡(โˆ…)=0 and is countably additive.

  • Prove that the following are equivalent:

    • ๐œ‡ is locally finite;

    • ๐œ‡([โˆ’๐‘,๐‘])<โˆž for every ๐‘โ‰ฅ0;

    • ๐œ‡(๐ผ)<โˆž for every bounded interval ๐ผ.

  • Prove that if ๐œ‡ is locally finite, then ๐œ‡ is ๐œŽ-finite. Is the converse true? Give a proof or a counterexample.

Basic formulas for LS measures

Let ๐น:โ„โ†’โ„ be a distribution function, and ๐œ‡=๐œ‡๐น the associated Lebesgueโ€“Stieltjes measure. From its definition using h-intervals, it follows that ๐œ‡((๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Ž) for โˆ’โˆž<๐‘Ž<๐‘<โˆž. Using this property together with basic general properties of (๐œŽ-finite) measures, we proved in class that ๐œ‡((๐‘Ž,๐‘))=๐น(๐‘โˆ’)โˆ’๐น(๐‘Ž) for โˆž<๐‘Žโ‰ค๐‘<โˆž. Using a similar strategy, prove the following:

  • ๐œ‡([๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Žโˆ’) for โˆ’โˆž<๐‘Žโ‰ค๐‘<โˆž;

  • ๐œ‡([๐‘Ž,๐‘))=๐น(๐‘โˆ’)โˆ’๐น(๐‘Žโˆ’) for โˆ’โˆž<๐‘Žโ‰ค๐‘<โˆž;

  • ๐œ‡({๐‘Ž}))=๐น(๐‘Ž)โˆ’๐น(๐‘Žโˆ’) for โˆ’โˆž<๐‘Ž<โˆž;

  • ๐œ‡((โˆ’โˆž,๐‘])=๐น(๐‘)โˆ’๐น(โˆ’โˆž) for โˆ’โˆž<๐‘<โˆž;

  • ๐œ‡((โˆ’โˆž,๐‘))=๐น(๐‘โˆ’)โˆ’๐น(โˆ’โˆž) for โˆ’โˆž<๐‘<โˆž;

  • ๐œ‡((๐‘Ž,โˆž))=๐น(โˆž)โˆ’๐น(๐‘Ž) for โˆ’โˆž<๐‘Ž<โˆž;

  • ๐œ‡([๐‘Ž,โˆž))=๐น(โˆž)โˆ’๐น(๐‘Žโˆ’) for โˆ’โˆž<๐‘Ž<โˆž;

  • ๐œ‡([โˆ’โˆž,โˆž))=๐น(โˆž)โˆ’๐น(โˆ’โˆž).

Vitali sets

For ๐‘ฅ,๐‘ฆโˆˆ[โˆ’1,1], write ๐‘ฅโˆผ๐‘ฆ iff ๐‘ฅโˆ’๐‘ฆโˆˆโ„š.

  • Show that โˆผ is an equivalence relation, i.e. show that (i) ๐‘ฅโˆผ๐‘ฅ, (ii) ๐‘ฅโˆผ๐‘ฆ implies ๐‘ฆโˆผ๐‘ฅ, (iii) if ๐‘ฅโˆผ๐‘ฆ and ๐‘ฆโˆผ๐‘ง, then ๐‘ฅโˆผ๐‘ง.

  • The set [โˆ’1,1] is partitioned into equivalence classes. Let ๐‘‰โŠ‚[โˆ’1,1] be a set containing exactly one element from each equivalence class. (Here, we use the Axiom of choice.) We call ๐‘‰ a Vitali set. Let {๐‘Ÿ1,๐‘Ÿ2,โ€ฆ}=[โˆ’2,2]โˆฉโ„š. Define ๐‘‰๐‘–=๐‘Ÿ๐‘–+๐‘‰={๐‘Ÿ๐‘–+๐‘ฅโˆฃ๐‘ฅโˆˆ๐‘‰}. Prove that the sets ๐‘‰1,๐‘‰2,โ€ฆ are mutually disjoint, and that

    [โˆ’1,1]โŠ‚โ‹ƒ๐‘–=1โˆž๐‘‰๐‘–โŠ‚[โˆ’3,3].

Vitali sets, season 2

Let ๐‘‰โŠ‚[โˆ’1,1] be a Vitali set (see above).

  • Using the translation invariance of Lebesgue measure, prove ๐‘‰ is not Lebesgue measurable.

  • Prove that if ๐ธ is a Lebesgue measurable set and satisfies ๐ธโŠ‚๐‘‰, then ๐‘š(๐ธ)=0.

  • Using the technique inย (a), prove the following statement: if ๐ดโŠ‚โ„ is any Lebesgue measurable set with ๐‘š(๐ด)>0, then ๐ด contains a set which is not Lebesgue measurable.

The middle-thirds Cantor set

Let ๐ถ be the middle-thirds Cantor set, defined as

๐ถโ‰”โ‹‚๐‘›=1โˆž๐ถ๐‘›,

where

๐ถ๐‘›โ‰”โ‹ƒ๐‘Ž1,โ€ฆ,๐‘Ž๐‘›โˆˆ{0,2}[โˆ‘๐‘–=1๐‘›๐‘Ž๐‘–3๐‘–,โˆ‘๐‘–=1๐‘›๐‘Ž๐‘–3๐‘–+13๐‘›]
  • Set ๐ถ0=[0,1]. Show that ๐ถ๐‘›โŠ‚๐ถ๐‘›โˆ’1 for all ๐‘›โ‰ฅ1. Also prove that ๐ถ๐‘› is the union of 2๐‘› disjoint closed intervals, that the set ๐‘ˆ๐‘›โ‰”๐ถ๐‘›โˆ’1\๐ถ๐‘› is the union of the middle thirds open intervals of the disjoint closed intervals of ๐ถ๐‘›โˆ’1, and that

    ๐‘ˆ๐‘›=โ‹ƒ๐‘Ž1,โ‹ฏ,๐‘Ž๐‘›โˆ’1โˆˆ{0,2}(โˆ‘๐‘–=1๐‘›โˆ’1๐‘Ž๐‘–3๐‘–+13๐‘›,โˆ‘๐‘–=1๐‘›โˆ’1๐‘Ž๐‘–3๐‘–+23๐‘›).

    (We interpret this as the interval (1/3,2/3) when ๐‘›=1.) Thus, ๐ถ is the set obtained by removing successive middle thirds of the remaining disjoint closed intervals starting with [0,1]. Sketch the first few sets ๐ถ๐‘› and ๐‘ˆ๐‘›.

  • Show that ๐ถ is a compact set, and that ๐‘š(๐ถ)=0, where ๐‘š denotes Lebesgue measure. Also show that ๐ถ does not contain any non-empty open interval (๐‘Ž,๐‘).

  • Show that ๐ถ equals the set of numbers ๐‘ฅโˆˆ[0,1] which have a base-3 expansion of the form ๐‘ฅ=0.๐‘Ž1๐‘Ž2๐‘Ž3โ‹ฏ where ๐‘Ž๐‘– is either 0 or 2, i.e.ย 

    ๐ถ={โˆ‘๐‘–=1โˆž๐‘Ž๐‘–3๐‘–โˆฃ ๐‘Ž๐‘–โˆˆ{0,2} for all ๐‘–โˆˆโ„•}.

    (Note: A point may have two base-3 expansions such as 1/3=0.1000โ€ฆ=0.0222โ€ฆ; this number is in ๐ถ since one of the expansions is of the desired form.)

  • Show that 14,913โˆˆ๐ถ but 536โˆ‰๐ถ.

The Devilโ€™s Staircase: an increasing function build on Cantor set

Let ๐ถ be the middle-thirds Cantor set, and define ๐น:๐ถโ†’[0,1] by

๐น(๐‘ฅ)=โˆ‘๐‘–=1โˆž๐‘Ž๐‘–/22๐‘–

for ๐‘ฅ=โˆ‘๐‘–=1โˆž๐‘Ž๐‘–3๐‘–, ๐‘Ž๐‘–โˆˆ{0,2}.

  • Prove that ๐น is an increasing function, and that ๐น(๐ถ)=[0,1].

  • Suppose that ๐‘ฅ,๐‘ฆโˆˆ๐ถ and ๐‘ฅ<๐‘ฆ. Prove that ๐น(๐‘ฅ)=๐น(๐‘ฆ) iff ๐‘ฅ and ๐‘ฆ are the endpoints of a removed open interval, that is, one of the 2๐‘›โˆ’1 disjoint open intervals whose union equals ๐‘ˆ๐‘›=๐ถ๐‘›โˆ’1\๐ถ๐‘› for some ๐‘›โ‰ฅ1.

  • Prove that ๐น:๐ถโ†’[0,1] extends uniquely to a continuous function which is constant on all the intervals in ๐‘ˆ๐‘›, ๐‘›โ‰ฅ1. Sketch the graph of ๐น. Hint: to prove continuity, it suffices to show that ๐น([0,1])=[0,1] (Why?)

  • Prove that ๐นโ€ฒ(๐‘ฅ)=0 for a.e.ย ๐‘ฅ. In other words, there exists a set ๐ธโŠ‚[0,1] such that ๐‘š(๐ธ)=0, and such that limโ„Žโ†’0(๐น(๐‘ฅ+โ„Ž)โˆ’๐น(๐‘ฅ))/โ„Ž=0 for ๐‘ฅโˆˆ[0,1]\๐ธ.

(Remark 1: because ofย (c) andย (d), the graph of ๐น is called the Devilโ€™s Staircase; it is horizontal almost everywhere, and has no vertical jumps, but nevertheless climbs upwards.)

(Remark 2: the fact that ๐น(๐ถ)=[0,1] implies that ๐ถ has the same cardinality as [0,1], in particular the Cantor set is uncountable.)

Some of the following questions will be graded. Do them, and do hand them in.

fun facts about distribution functions

  • Let ๐ดโŠ‚โ„ be a countable set. Exhibit a distribution function ๐น that is discontinuous at every point in ๐ด, but continuous everywhere else. Justify your answer. Hint: play around with the Heaviside function.

  • Let ๐น:โ„โ†’โ„ be an increasing function. Prove that there exists a unique distribution function ๐บ such that ๐บ(๐‘ฅ)=๐น(๐‘ฅ) for all points ๐‘ฅ where ๐น is continuous. Hint: there is a simple formula for ๐บ in terms of ๐น.

Solution

of (a):
We list ๐ด={๐‘Ž๐‘›}๐‘›=1โˆž as a sequence to label its elements. Define:

๐น(๐‘ฅ)=โˆ‘๐‘›=1โˆž12๐‘›๐ป(๐‘ฅโˆ’๐‘Ž๐‘›)

where ๐ป(๐‘ฅ) is the Heaviside function: ๐ป(๐‘ฅ)={0,๐‘ฅ<0,1,๐‘ฅโ‰ฅ0..
Claim 1.1 ๐น is non-decreasing.
Proof: Suppose ๐‘ฆ>๐‘ฅโˆˆโ„, then ๐ป(๐‘ฆโˆ’๐‘Ž๐‘›)โ‰ฅ๐ป(๐‘ฅโˆ’๐‘Ž๐‘›) for each ๐‘›โˆˆโ„•, so we have ๐น(๐‘ฆ)โ‰ฅ๐น(๐‘ฅ).

Claim 1.2 ๐น is right continuous but not left continuous (thus discontinuous) at every ๐‘Ž๐‘›.
Proof: Let ๐œ–>0.
We take ๐‘โˆˆโ„• s.t. โˆ‘๐‘˜โ‰ฅ๐‘,๐‘›โˆˆโ„•12๐‘˜<๐œ–.
Then we take ๐›ฟ>0 such that ๐‘Ž1,๐‘Ž2,โ‹ฏ,๐‘Ž๐‘โˆ‰(๐‘Ž๐‘›,๐‘Ž๐‘›+๐›ฟ).(This can be done since there are only finite points here)
Thus โˆ€๐‘ฆโˆˆ(๐‘Ž๐‘›,๐‘Ž๐‘›+๐›ฟ), we have |๐น(๐‘ฆ)โˆ’๐น(๐‘Ž๐‘›)|<๐œ–, since ๐น(๐‘ฆ)<๐น(๐‘Ž๐‘›)+โˆ‘๐‘˜โ‰ฅ๐‘,๐‘›โˆˆโ„•12๐‘˜. Since ๐œ– is arbitrary, this finishes the proof that ๐น is right continuous at ๐‘Ž๐‘›.
Also, โˆ€๐‘ฆ<๐‘Ž๐‘›, we have ๐น(๐‘ฆ)<๐น(๐‘Ž๐‘›)โˆ’12๐‘›, which means that |๐น(๐‘ฆ)โˆ’๐น(๐‘Ž๐‘›)|>12๐‘› for any ๐‘ฆ on the left, so ๐น is not left continuous at ๐‘Ž๐‘›.

Claim 1.3: ๐น is continuous at every ๐‘ฅโˆˆโ„\๐ด.
Proof: This is similar to the proof in Claim 1.2.
Fix ๐‘ฅโˆˆโ„\๐ด. Let ๐œ–>0.
We take ๐‘โˆˆโ„• s.t. โˆ‘๐‘˜โ‰ฅ๐‘,๐‘›โˆˆโ„•12๐‘˜<๐œ–.
Then we take ๐›ฟ>0 such that ๐‘Ž1,๐‘Ž2,โ‹ฏ,๐‘Ž๐‘โˆ‰(๐‘Ž๐‘›โˆ’๐›ฟ,๐‘Ž๐‘›+๐›ฟ). This can be done since there are only finite points here.
Thus โˆ€๐‘ฆโˆˆ(๐‘Ž๐‘›โˆ’๐›ฟ,๐‘Ž๐‘›+๐›ฟ), we have |๐น(๐‘ฆ)โˆ’๐น(๐‘Ž๐‘›)|<โˆ‘๐‘˜โ‰ฅ๐‘,๐‘›โˆˆโ„•12๐‘˜<๐œ–.
ย Since ๐œ– is arbitrary, this finishes the proof that ๐น is continuous at ๐‘ฅ.

By claim 1.1, 1.2, 1.3, we have proved that ๐น is a distribution function that is discontinuous at every point of ๐ด but continuous elsewhere.

Proof

of (b):
Given an increasing function ๐น:โ„โ†’โ„, we define ๐บ:โ„โ†’โ„ by:

๐บ(๐‘ฅ)=lim๐‘ฆโ†’๐‘ฅ+๐น(๐‘ฆ).

We will show that this is the unique distribution function ๐บ such that ๐บ(๐‘ฅ)=๐น(๐‘ฅ) for all points ๐‘ฅ where ๐น is continuous.
Incresing: Since ๐น is increasing, for any ๐‘ฅ<๐‘ฆ we have ๐น(๐‘ฅ)โ‰ค๐น(๐‘ฆ). Thus, for any ๐‘ฅ<๐‘ฆ we have:

๐บ(๐‘ฅ)=lim๐‘งโ†’๐‘ฅ+๐น(๐‘ง)โ‰คlim๐‘งโ†’๐‘ฆ+๐น(๐‘ง)=๐บ(๐‘ฆ).

Thus, ๐บ is increasing.
Right-continuity: Since ๐น is an increasing function, it can only have jump discontinuities, and the right limit exists for all ๐‘‹. By construction, ๐บ is right-ctn.
Above finishes the proof that ๐บ is a distribution function.
Agree with ๐น at ctn point: ๐บ(๐‘ฅ)=๐น(๐‘ฅ) where ๐น is continuous at ๐‘ฅ, since ๐บ(๐‘ฅ)=lim๐‘ฆโ†’๐‘ฅ+๐น(๐‘ฆ)=๐น(๐‘ฅ) there.
It remains to show that it is unique.
Suppose ๐‘… is another such function. It suffices to show: ๐‘… agrees with ๐บ on discontinuous points of ๐น.
Since ๐‘…,๐บ are right continuous, their right limit must exist at each point. Therefore, let ๐‘ฅ be an arbitrary point where ๐น is discontinuous at ๐‘ฅ, it suffices to show that there is a sequence {๐‘ฅ๐‘›} approaching ๐‘ฅ, such that lim๐‘›๐บ(๐‘ฅ๐‘›)=lim๐‘›๐‘…(๐‘ฅ๐‘›).
Since ๐น is increasing, the points where ๐น is disctn is at most countable. Therefore the points where ๐น is ctn, denote it as ๐ถ, is dense in โ„. Thus we can pick a sequence {๐‘ฅ๐‘›} in ๐ถ approaching ๐‘ฅ, then ๐บ(๐‘ฅ๐‘›)=๐‘…(๐‘ฅ๐‘›)=๐น(๐‘ฅ๐‘›) for each ๐‘›, impling that lim๐‘›๐บ(๐‘ฅ๐‘›)=lim๐‘›๐‘…(๐‘ฅ๐‘›). This finishes the proof of uniqueness.

โ–ก

Finding intervals

Let ๐ธโŠ‚โ„ be a Lebesgue measurable subset with ๐‘š(๐ธ)>0. Prove that for every ๐›ผโˆˆ(0,1) there exists an (nonempty) bounded open interval ๐ผ such that ๐‘š(๐ธโˆฉ๐ผ)โ‰ฅ๐›ผ๐‘š(๐ผ). Hint: first reduce to the case when ๐ธ is bounded, then use outer regularity.

Proof

Let ๐›ผโˆˆ(0,1) be arbitrary and fix it.
We first consider the case that ๐ธ is bounded. By outer regularity of Lebesgue measure, there exists an open set ๐บ such that

๐ธโŠ‚๐บand๐‘š(๐บ)โˆ’๐‘š(๐ธ)โ‰ค(1/๐›ผโˆ’1)๐‘š(๐ธ)

since ๐›ผโˆˆ(0,1). Then we have:

๐‘š(๐บ)โ‰ค1/๐›ผ๐‘š(๐ธ)

Note that in โ„, an open set is just a countable disjoint union of open intervals. We write:

๐บ=โจ†๐‘–โˆˆโ„•๐ผ๐‘–

Since ๐ธโŠ‚๐บ, we have:

๐ธ=โจ†๐‘–โˆˆโ„•(๐ผ๐‘–โˆฉ๐ธ)

Thus

๐‘š(๐ธ)=โˆ‘๐‘–โˆˆโ„•๐‘š(๐ผ๐‘–โˆฉ๐ธ)โ‰ฅ๐›ผโˆ‘๐‘–โˆˆโ„•๐‘š(๐ผ๐‘–)

So there must exist some ๐‘– such that ๐‘š(๐ผ๐‘–โˆฉ๐ธ)โ‰ฅ๐›ผ๐‘š(๐ผ๐‘–), otherwise contradicting with the ineq above.
This finishes the proof of the bounded case.
The we consider the case when ๐ธ is unbounded. We can write

๐ธ=โจ†๐‘›โˆˆโ„ค(๐ธโˆฉ(๐‘›,๐‘›+1])

where each ๐ธ๐‘›โ‰”๐ธโˆฉ(๐‘›,๐‘›+1] is bounded.
We apply the case where ๐ธ is bounded, confirming that there is some interval ๐ผ such that ๐‘š(๐ธ1โˆฉ๐ผ)โ‰ฅ๐›ผ๐‘š(๐ผ). By monotonicity of measure, we have ๐‘š(๐ธโˆฉ๐ผ)โ‰ฅ๐‘š(๐ธ1โˆฉ๐ผ)โ‰ฅ๐›ผ๐‘š(๐ผ).

โ–ก

So many differences

Let ๐ธโŠ‚โ„ be a Lebesgue measurable subset with ๐‘š(๐ธ)>0.

  • Prove that the set

    ๐ธโˆ’๐ธโ‰”{๐‘ฅโˆ’๐‘ฆโˆฃ๐‘ฅ,๐‘ฆโˆˆ๐ธ}โŠ‚โ„

    contains a nonempty open interval centered at the origin. Hint: use the previous exercise with ๐›ผ large enough, together with the translation invariance of Lebesgue measure.

  • Prove that there exists ๐œ–>0 such that ๐ธร—๐ธโŠ‚โ„2 intersects every line ๐‘ฆ=๐‘ฅ+๐‘ก with |๐‘ก|<๐œ–.

  • Let ๐ถโŠ‚โ„ be the middle-third Cantor set (so ๐‘š(๐ถ)=0). Does ๐ถโˆ’๐ถ contain a nonempty open interval centered at the origin?

Proof

of (a):

Figureย 9:
Figureย 10:

โ–ก

Proof

of (b):
Consider taking ๐œ– as the one in (a) where the interval contained in ๐ธโˆ’๐ธ is (โˆ’๐œ–,๐œ–), then the box (โˆ’๐œ–,๐œ–)ร—(โˆ’๐œ–,๐œ–) is contained in ๐ธร—๐ธ. It trivially follows that ๐ธร—๐ธโŠ‚โ„2 intersects every line ๐‘ฆ=๐‘ฅ+๐‘ก with |๐‘ก|<๐œ–, since the intercept of this line with ๐‘ฆ-axis is below ๐œ– and above โˆ’๐œ–.

Figureย 11:

โ–ก

Solution

of (c): ๐ถโˆ’๐ถ contain a nonempty open interval centered at the origin, and we will prove that one such interval is (โˆ’1,1).

Proof

Recall the balanced ternary representation of [โˆ’1/2,1/2]: โˆ€๐‘ฅโˆˆ[โˆ’1/2,1/2], there is a seq of (๐‘Ž๐‘›)๐‘›โˆˆโ„• in {โˆ’1,0,1} s,t,

๐‘ฅ=โˆ‘๐‘›=1โˆž๐‘Ž๐‘›3๐‘›,๐‘Ž๐‘›โˆˆ{โˆ’1,0,1},

Thus every ๐‘ฅโˆˆ[โˆ’1,1] can be halved, ternary expanded and then doubled to recover:

๐‘ฅ=2โˆ‘๐‘›=1โˆž๐‘Ž๐‘›3๐‘›=โˆ‘๐‘›=1โˆž2๐‘Ž๐‘›3๐‘›,๐‘Ž๐‘›โˆˆ{โˆ’1,0,1}=โˆ‘๐‘›=1โˆž๐‘๐‘›3๐‘›,๐‘๐‘›โˆˆ{โˆ’2,0,2}

And by the problem "The middle-thirds Cantor set", we learned that

๐ถ={โˆ‘๐‘–=1โˆž๐‘Ž๐‘–3๐‘–โˆฃ ๐‘Ž๐‘–โˆˆ{0,2} for all ๐‘–โˆˆโ„•}.

Therefore we can write every number ๐‘ฅโˆˆ[โˆ’1,1] into a difference of two ๐‘ฅ,๐‘ฆโˆˆ๐ถ, i.e. an element of ๐ถโˆ’๐ถ:

๐‘ฅ=โˆ‘๐‘›=1โˆž๐‘๐‘›3๐‘›,๐‘๐‘›โˆˆ{โˆ’2,0,2}=โˆ‘๐‘›=1โˆž๐‘๐‘›โˆ’๐‘ž๐‘›3๐‘›,๐‘๐‘›,๐‘ž๐‘›โˆˆ{โˆ’2,0,2}=โˆ‘๐‘›=1โˆž๐‘๐‘›3๐‘›โˆ’โˆ‘๐‘›=1โˆž๐‘ž๐‘›3๐‘›,๐‘๐‘›,๐‘ž๐‘›โˆˆ{โˆ’2,0,2}

since each series converges independently. Here we let ๐‘๐‘›=2,๐‘ž๐‘›=0 if ๐‘๐‘›=2; ๐‘๐‘›=0,๐‘ž๐‘›=2 if ๐‘๐‘›=โˆ’2, ๐‘๐‘›=0,๐‘ž๐‘›=0 if ๐‘๐‘›=0.
Thus ๐‘ฅโˆˆ๐ถโˆ’๐ถ, so [โˆ’1,1]โŠ‚๐ถโˆ’๐ถ.

โ–ก

a holey set

Let (๐‘ฅ๐‘›)1โˆž be a countable dense sequence in (0,1). For each ๐‘ก>0, consider the set

๐ด๐‘กโ‰”[0,1]\โ‹ƒ๐‘›=1โˆž(๐‘ฅ๐‘›โˆ’2โˆ’๐‘›๐‘ก,๐‘ฅ๐‘›+2โˆ’๐‘›๐‘ก).
  • Prove that ๐ด๐‘ก is a compact (possibly empty) subset of โ„. Also prove that ๐ด๐‘ก has empty interior, that is, ๐ด๐‘ก contains no nonempty open set.

  • Prove that ๐‘กโ†ฆ๐‘š(๐ด๐‘ก) is continuous.

  • Prove that there exists ๐‘ก>0 such that ๐‘š(๐ด๐‘ก)=597/2025.

Proof

of a:

Figureย 12:

โ–ก

Proof

of b: Define for each ๐‘›โˆˆโ„•

๐ผ๐‘›(๐‘ก)=(๐‘ฅ๐‘›โˆ’2โˆ’๐‘›๐‘ก,๐‘ฅ๐‘›+2โˆ’๐‘›๐‘ก)

Then

๐ด๐‘ก=[0,1]\โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก)

So

๐‘š(๐ด๐‘ก)=๐‘š([0,1])โˆ’๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))=1โˆ’๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))

Thus it suffices to show ๐‘กโ†ฆ๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก)) is continuous.

Let ๐œ–>0. Let ๐‘ก>0.
We consider ๐‘โˆˆ(๐‘ก,๐‘ก+๐œ–/2):

By set inclusion relation and measureโ€™s property, we have:

๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘))โˆ’๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))=๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘)\โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))

Since

(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘))\(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))โŠ‚โ‹ƒ๐‘›=1โˆž(๐ผ๐‘›(๐‘)\๐ผ๐‘›(๐‘ก))

We have:

๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘))โˆ’๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))โ‰ค๐‘š(โ‹ƒ๐‘›=1โˆž(๐ผ๐‘›(๐‘)\๐ผ๐‘›(๐‘ก)))โ‰คโˆ‘๐‘›=1โˆž(๐‘š(๐ผ๐‘›(๐‘))โˆ’๐‘š(๐ผ๐‘›(๐‘ก)))=โˆ‘๐‘›=1โˆž2โ‹…2โˆ’๐‘›(๐‘โˆ’๐‘ก)=2(๐‘โˆ’๐‘ก)โ‰ค๐œ–

Similarly for ๐‘โˆˆ(๐‘กโˆ’๐œ–/2,๐‘ก), we get the same bound. This finishes the proof pf (b).

โ–ก

Proof

of c:
We use the same notation of ๐ผ๐‘›(๐‘ก) as in (b). We have:

๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))โ‰คโˆ‘๐‘›=1โˆž๐‘š(๐ผ๐‘›(๐‘ก))=โˆ‘๐‘›=1โˆž๐‘š(๐ผ๐‘›(๐‘ก))=โˆ‘๐‘›=1โˆž2โ‹…2โˆ’๐‘›๐‘ก=2๐‘ก.

So by choosing ๐‘กโ‰”1/6, we have ๐‘š(๐ด๐‘ก)=1โˆ’๐‘š(โ‹ƒ๐‘›=1โˆž๐ผ๐‘›(๐‘ก))โ‰ฅ2/3. And by choosing ๐‘กโ‰”4, ๐ผ1(๐‘ก) covers an interval of length 4, so ๐ด๐‘ก=โŒ€, ๐‘š(๐ด๐‘ก)=0. By intermediate value theorem, there exists some ๐‘กโˆˆ(1/6,4) such that ๐‘š(๐ด๐‘ก)=597/2025.

โ–ก

a Cantor measure

(A Cantor measure.) Let ๐ธโŠ‚โ„ be a nonempty compact set with the following property: for every ๐‘ฅโˆˆ๐ธ and every ๐œ–>0, the set (๐‘ฅโˆ’๐œ–,๐‘ฅ)โˆช(๐‘ฅ,๐‘ฅ+๐œ–) has nonempty intersection with both ๐ธ and ๐ธ๐‘. Prove that there exists a Borel measure ๐œ‡ on โ„ with the following properties:

  • if ๐ผโŠ‚โ„ is a nonempty open interval, then ๐œ‡(๐ผ)>0 iff ๐ผโˆฉ๐ธโ‰ โˆ….

  • ๐œ‡({๐‘ฅ})=0 for all ๐‘ฅโˆˆโ„;

  • ๐œ‡(โ„)=0.597597597โ€ฆ.

Hint: set ๐œ‡=๐œ‡๐น, where ๐น is a distribution function whose graph is similar to the Devilโ€™s staircase above.

Proof

Write ๐‘‡โ‰”0.597597597โ€ฆ
Since ๐ธ is compact, ๐ธ๐‘ is open. Also, since ๐ธ is compact, it takes min and max element.
Thus we consider ๐ดโ‰”๐ธ๐‘โˆฉ(min๐ธ,max๐ธ), this is an open set. We know any open set in โ„ is a countable disjoint union of open intervals, so ๐ดโ‰”๐ธ๐‘โˆฉ(min๐ธ,max๐ธ)=โจ†๐‘›=1โˆž๐ผ๐‘› for some disjoint intervals ๐ผ1=(๐‘Ž1,๐‘1),๐ผ2=(๐‘Ž2,๐‘2),โ‹ฏ .

Now we construct a function ๐บ:๐ดโ†’[0,๐‘‡) by sending ๐บ(๐‘ฅ)=โˆ‘๐‘๐‘–โ‰ค๐‘Ž๐‘๐‘‡2๐‘›, for ๐‘ฅโˆˆ๐ผ๐‘.
This is an increasing step function since, each ๐ผ๐‘› is disjoint and on a fixed interval ๐ผ๐‘, the number of ๐‘๐‘– that its ๐‘Ž๐‘ surpasses is constant. And suppose ๐‘ฆ>๐‘ฅ is on ๐ผ๐‘€, we have must ๐บ(๐‘ฆ)โ‰ฅ๐บ(๐‘ฅ) because he number of ๐‘๐‘– that ๐‘Ž๐‘€ surpasses is at least at many as that ๐‘Ž๐‘ surpasses.
And for each ๐‘ฅโˆˆ๐ด, we have ๐บ(๐‘ฅ)<๐‘‡, by geometric series
Then we construct ๐น out of ๐บ, define:

๐นโ‰”{0,๐‘ฅโ‰คmin๐ธinf{๐บ(๐‘ฆ)โˆฃ๐‘ฆโ‰ฅ๐‘ฅ,๐‘ฆโˆˆ๐ด},๐‘ฅโˆˆ(min๐ธ,max๐ธ)๐‘‡,๐‘ฅโ‰ฅmax๐ธ

๐น is increasing: It is constant on (โˆ’โˆž,min๐ธ)โˆช(max๐ธ,โˆž) and is the infimum of ๐บ(๐‘ฆ) with ๐‘ฆโ‰ฅ๐‘ฅ on (min๐ธ,max๐ธ). Since ๐บ is increasing, ๐น is also increasing.
๐น is right continuous: It suffices to prove the right-continuity of ๐น on ๐‘ฅโˆˆ๐ธ๐‘โˆฉ(min๐ธ,max๐ธ).
Fix ๐‘ฅ0โˆˆ๐ธ๐‘โˆฉ(min๐ธ,max๐ธ).
Let ๐œ–>0.
Let ๐‘˜โˆˆโ„• such that ๐œ–>๐‘‡2๐‘˜+1.
We define for each ๐‘ฆ, ๐‘†๐‘ฆโ‰”{๐‘๐‘–โˆฃ๐‘ฅ0โ‰ค๐‘๐‘–โ‰ค๐‘ฆ} as the set of all ๐‘๐‘– (right endpoint of ๐ผ๐‘˜) that is witin ๐‘ฅ0 and ๐‘ฆ. Note that ๐ผ๐‘งโŠ‚๐ผ๐‘ฆ for all ๐‘ฆ>๐‘ง.
Consider ๐‘ฆ1โ‰”min({๐‘1,โ‹ฏ,๐‘๐‘˜}\๐‘†๐‘ฅ).
Then for all ๐‘ฆโˆˆ(๐‘ฅ0,๐‘ฆ1), we have:

๐น(๐‘ฆ)โ‰ค๐น(๐‘ฅ0)+โˆ‘๐‘–=๐‘˜โˆž๐‘‡2๐‘–โ‰ค๐น(๐‘ฅ0)+๐œ–

By defining ๐›ฟ:=๐‘ฆ1โˆ’๐‘ฅ0, we have shown the right continuity of ๐น.
(By dual reason, we can prove that ๐น is left continuous. So ๐น is actually continuous.) Above, we have shown that ๐น is a distribution function.

Now let ๐œ‡๐น be the Lebesgue-Stieljes measure associated with ๐น. We will prove for the three properties above:

  • Let {๐‘ฅ} be a singleton set in โ„, for each ๐‘›โˆˆโ„•, we can construct an h-intervals seq of covering of {๐‘ฅ} by (๐‘ฅโˆ’1/๐‘›,๐‘ฅ] as the first covering set and โŒ€ as all other covering sets.
    Then by the definition of ๐œ‡๐น, we have:

    ๐œ‡๐น({๐‘ฅ})=inf๐‘›โˆˆโ„•(๐น(๐‘ฅ)โˆ’๐น(๐‘ฅโˆ’1/๐‘›))

    By continuity, it shows that ๐œ‡๐น({๐‘ฅ})=0.

  • ๐œ‡๐น(โ„)=lim๐‘ฅโ†’โˆž๐น(๐‘ฅ)โˆ’lim๐‘ฅโ†’โˆ’โˆž๐น(๐‘ฅ)=๐‘‡โˆ’0=๐‘‡=0.597597597โ‹ฏ
  • Let ๐ผ=(๐‘Ž,๐‘) be a nonempty open interval.
    Suppose ๐œ‡(๐ผ)>0, then ๐น(๐‘)โˆ’๐น(๐‘Ž)>0, so by definition of ๐บ, some must be at least two different intervals ๐ผ๐‘›1, ๐ผ๐‘›2 in ๐ด such that for some ๐‘ฅ,๐‘ฆโˆˆ(๐‘Ž,๐‘), ๐‘ฅโˆˆ๐ผ๐‘›1 and ๐‘ฆโˆˆ๐ผ๐‘›2, thus โˆƒ some ๐‘’โˆˆ๐ธ such that ๐‘’โˆˆ(๐‘ฅ,๐‘ฆ). Thus ๐ผโˆฉ๐ธโ‰ โˆ….
    Suppose ๐ผโˆฉ๐ธโ‰ โˆ…. Let ๐‘’โˆˆ๐ธโˆฉ๐ผ. Since โˆ€๐œ–>0, (๐‘ฅโˆ’๐œ–,๐‘ฅ)โˆช(๐‘ฅ,๐‘ฅ+๐œ–) has nonempty intersection with both ๐ธ and ๐ธ๐‘, ๐‘’ has some open neighborhood ๐ต๐œ–(๐‘’)โŠ‚๐ผ, intersecting two different ๐ผ๐‘, ๐ผ๐‘€โŠ‚๐ด. Take ๐‘›โˆˆ๐ผ๐‘, ๐‘šโˆˆ๐ผ๐‘š. Then ๐น(๐‘š)โˆ’๐น(๐‘›)=๐บ(๐‘š)โˆ’๐บ(๐‘›)>0, so ๐œ‡๐น(๐ธ)โ‰ฅ๐น(๐‘š)โˆ’๐น(๐‘›)>0 by monotinicity of measure.
    This finishes the proof.

โ–ก

โ€˜

4 measurable functions and integration on ๐ฟ+(๐œ‡)

4.1 measurable function [Fol 2.1]

4.1.1 general measurable function

Definition 4.16 : (โ„ณ๏ธ€,๐’ฉ๏ธ€)-measurable function

Let (๐‘‹,โ„ณ๏ธ€), (๐‘Œ,๐’ฉ๏ธ€) be measurable spaces, ๅฆ‚ๆžœ ๐‘“:๐‘‹โ†’๐‘Œ ๆปก่ถณ:

๐ตโˆˆ๐’ฉ๏ธ€โŸน๐‘“โˆ’1(๐ต)โˆˆโ„ณ๏ธ€

, ๅˆ™็งฐ ๐‘“ ไธบไธ€ไธช (โ„ณ๏ธ€,๐’ฉ๏ธ€)-measurable function.

ไปŽไธ€ไธช measurable space ๅˆฐๅฆไธ€ไธช measurable space ็š„ function ่ขซ็งฐไธบ measurable ็š„ๆกไปถๆ˜ฏ: ่ขซๆ˜ ๅฐ„ๅˆฐๅฏๆต‹้›†็š„้›†ๅˆๅช่ƒฝๆ˜ฏๅฏๆต‹้›†.

่ฟ™ไธชๅฎšไน‰ๅ’Œ topological space ไธŠ continuous ็š„ๅฎšไน‰: ่ขซๆ˜ ๅฐ„ๅˆฐๅผ€้›†็š„ๅช่ƒฝๆ˜ฏๅผ€้›†, ๅฝขๅผๆ˜ฏๅฎŒๅ…จไธ€ๆ ท็š„. ๅนถไธ”ๆˆ‘ไปฌ็Ÿฅ้“, topological space ๅ’Œ measure space ไนŸๆœ‰ๅพˆๅคš็›ธไผผไน‹ๅค„. ๅ› ่€Œ่ฟž็ปญๆ€งๅ’Œๅฏๆต‹ๆ€งๆœ‰ไธ€ๅฎš็š„ๅ…ณ็ณป.

ๅ‡ฝๆ•ฐ็š„ๅฏๆต‹ๆ€ง็š„ๅฎšไน‰ๆ˜ฏ with respect to ๅฎƒไปฌๆ‰€ๅœจๅฏๆต‹็ฉบ้—ด้€‰ๅฎš็š„ ๐œŽ-algebra ็š„, ๅฐฑๅƒ topologica spaces ไน‹้—ดๅ‡ฝๆ•ฐ็š„่ฟž็ปญๆ€ง็š„ๅฎšไน‰ๆ˜ฏ with respect to ๅฎƒไปฌๆ‰€ๅœจ็š„ topological spaces ้€‰ๅฎš็š„ topology.

่ฟ™ไธคไธชๅฎšไน‰้ƒฝ่กจ็คบ็š„ๆ˜ฏ: ๆ€ง่ดจไธๅฅฝ็š„้›†ๅˆไธไผš่ขซๆ˜ ๅฐ„ๅˆฐๆ€ง่ดจ่‰ฏๅฅฝ็š„้›†ๅˆ. (ไฝ†ๆ˜ฏๆ€ง่ดจ่‰ฏๅฅฝ็š„้›†ๅˆๆœ‰ๅฏ่ƒฝ่ขซๆ˜ ๅฐ„ๅˆฐๆ€ง่ดจไธๅฅฝ็š„้›†ๅˆ.)

Proposition 4.3 : composition preserves measurability

ๅฆ‚ๆžœ ๐‘“ ๆ˜ฏ (๐’œ๏ธ€,โ„ฌ๏ธ€)-measurable ็š„, ๐‘” ๆ˜ฏ (โ„ฌ๏ธ€,๐’ž๏ธ€)-measurable ็š„, ้‚ฃไนˆ ๐‘”โˆ˜๐‘“ ๆ˜ฏ (๐’œ๏ธ€,๐’ž๏ธ€)-measurable ็š„.

Proof

Trivial.

โ–ก

Lemma 4.10

Let (๐‘‹,โ„ณ๏ธ€), (๐‘Œ,๐’ฉ๏ธ€) be measurable spaces, ๅฆ‚ๆžœ ๐’ฉ๏ธ€=<๐œ€> for some ๐œ€โІ๐‘Œ, ้‚ฃไนˆ

๐‘“:๐‘‹โ†’๐‘Œ (โ„ณ๏ธ€,๐’ฉ๏ธ€)-measurable โ‡” ๐‘“โˆ’1(๐ธ)โˆˆโ„ณ๏ธ€โˆ€๐ธโІ๐œ€

Proof

foward direction: trivial.
backward direction: Let

๐ท:={๐ธโІ๐‘Œโˆฃ๐‘“โˆ’1(๐ธ)โˆˆโ„ณ๏ธ€}

ๅฎนๆ˜“่ฏๆ˜Ž: ๐ทโЇ๐œ€, ๅนถไธ” ๐ท ๆ˜ฏไธ€ไธช ๐œŽ-algebra.
ๅ› ่€Œ ๐ทโЇ<๐œ€>=๐’ฉ๏ธ€

โ–ก

Proposition 4.4

ๅฏนไบŽ topological space ๐‘‹,๐‘Œ, let ๐‘“:๐‘‹โ†’๐‘Œ

๐‘“ continuous โŸน๐‘“ ๆ˜ฏ (โ„ฌ๏ธ€(๐‘‹),โ„ฌ๏ธ€(๐‘Œ)) measurable ็š„.

4.1.2 real and complex-valued measurable function

Definition 4.17 : (real-valued) measurable functions

Let (๐‘‹,๐’œ๏ธ€) be a measurable space, ๅฏนไบŽ ๐‘“:๐‘‹โ†’โ„ฬ„ ๅฆ‚ๆžœๅฎƒๆ˜ฏ (๐’œ๏ธ€,โ„ฌ๏ธ€(โ„ฬ„))-measurable ็š„, ๆˆ‘ไปฌ็›ดๆŽฅ็ฎ€็งฐๅฎƒๆ˜ฏ ๐’œ๏ธ€-measurable ็š„, ๆˆ–่€…็ฎ€็งฐไธบ measurable ็š„.

Definition 4.18 : (complex-valued) measurable functions

ๅฆ‚ๆžœ ๐‘“:๐‘‹โ†’โ„‚ ๆปก่ถณ: Re๐‘“,Im๐‘“ ้ƒฝๆ˜ฏ (real-valued) ๐‘‹-measurable ็š„, ้‚ฃไนˆไนŸ็งฐ ๐‘“ ๆ˜ฏ ๐‘‹-measurable ็š„, ๆˆ–่€…็›ดๆŽฅ่ฏดๆ˜ฏ measurable ็š„.

Definition 4.19 : Lebesgue measurable functions, Borel measurable functions

Naturally, ๅฆ‚ๆžœ ๐‘“:โ„โ†’โ„‚ ๆ˜ฏไธ€ไธช ๐”-measurable ็š„ๅ‡ฝๆ•ฐ, ้‚ฃไนˆๆˆ‘ไปฌ็งฐ ๐‘“ ๆ˜ฏ Lebesgue measurable ็š„.

ๅŒๆ ทๅœฐ, ๅฆ‚ๆžœๅฎƒๆ˜ฏไธ€ไธช โ„ฌ๏ธ€(โ„)-measurable ็š„ๅ‡ฝๆ•ฐ, ็งฐ ๐‘“ ๆ˜ฏ Borel measurable ็š„.

Proposition 4.5

ๅœจไปปไฝ• โ„ณ๏ธ€-measurable function ๐‘“ ๅ‰ compose ไธ€ไธช Borel measurable ็š„ function, ็ป“ๆžœไป็„ถๆ˜ฏ โ„ณ๏ธ€-measurable ็š„, follows from composition preserves measurability.

Proof

Follows from def.

โ–ก

Example 4.8

๐‘“2, โˆ’3๐‘“, 1|๐‘“| (๐‘“โ‰ 0) ้ƒฝไป็„ถๆ˜ฏ โ„ณ๏ธ€-measuble ็š„.

4.1.3 arithmetic and sequential preservation of measurable functions

Proposition 4.6 : addition and multiplication preserve measurability

ๅฆ‚ๆžœ ๐‘“,๐‘” ๆ˜ฏ โ„ณ๏ธ€-measurable function, ้‚ฃไนˆ ๐‘“+๐‘”,๐‘“๐‘” ไนŸๆ˜ฏ.

Proof

Suffices to assume ๐‘“,๐‘” is (extended) real-valued. Complex case follows trivially.

Suppose ๐‘“,๐‘” ๆ˜ฏ โ„ณ๏ธ€-measurable ็š„, ๆˆ‘ไปฌๆƒณ่ฆ่ฏๆ˜Ž: ๐‘“+๐‘” ๆ˜ฏ โ„ณ๏ธ€-measurable ็š„, suffices to show: (๐‘“+๐‘”)โˆ’1(๐‘Ž,โˆž]โˆˆโ„ณ๏ธ€ for any ๐‘Žโˆˆโ„.

ๆˆ‘ไปฌ notice:

{๐‘ฅโˆˆ๐‘‹โˆฃ๐‘“(๐‘ฅ)+๐‘”(๐‘ฅ)>๐‘Ž}=โ‹ƒ๐‘Ÿโˆˆโ„š{๐‘ฅโˆฃ๐‘“(๐‘ฅ)>๐‘Ÿ}โˆฉ{๐‘ฅโˆฃ๐‘”(๐‘ฅ)>๐‘Žโˆ’๐‘Ÿ}

ไบŽๆ˜ฏ finishes the proof.

ๅฏนไบŽ ๐‘“๐‘”, ๆˆ‘ไปฌๅ‘็Žฐๆœ‰

๐‘“๐‘”=12((๐‘“+๐‘”)2โˆ’๐‘“2โˆ’๐‘”2)

ไบŽๆ˜ฏไนŸ finishes the proof, following ๅ‰ไธ€ไธช proposition.

โ–ก

Lemma 4.11 : sequential behavior of real-valued measurable function

ๅฆ‚ๆžœ {๐‘“๐‘›:๐‘‹โ†’โ„ฬ„}๐‘›โˆˆโ„• ๆ˜ฏไธ€ไธช seq of โ„ณ๏ธ€-measurable functions, ้‚ฃไนˆ

  • ๐‘”1(๐‘ฅ):=sup๐‘—๐‘“๐‘—(๐‘ฅ)
  • ๐‘”2(๐‘ฅ):=inf๐‘—๐‘“๐‘—(๐‘ฅ)
  • ๐‘”3(๐‘ฅ):=limโ€‰sup๐‘—โ†’โˆž๐‘“๐‘—(๐‘ฅ)
  • ๐‘”4(๐‘ฅ):=limโ€‰inf๐‘—โ†’โˆž๐‘“๐‘—(๐‘ฅ)

้ƒฝๆ˜ฏ โ„ณ๏ธ€-measurable ็š„.

Proof
๐‘”1(๐‘ฅ)=sup๐‘—โˆˆโ„•๐‘“๐‘—(๐‘ฅ).

็”ฑไธŠ็กฎ็•Œ็š„ๅฎšไน‰๏ผš

๐‘”1(๐‘ฅ)>๐‘Žโ‡”โˆƒ๐‘—โˆˆโ„•, such that ๐‘“๐‘—(๐‘ฅ)>๐‘Ž.

ๅ› ๆญค๏ผŒ

{๐‘ฅโˆฃ๐‘”1(๐‘ฅ)>๐‘Ž}=โ‹ƒ๐‘—โˆˆโ„•{๐‘ฅโˆฃ๐‘“๐‘—(๐‘ฅ)>๐‘Ž}.

ๅ› ่€Œ:

๐‘”1โˆ’1((๐‘Ž,โˆž])=โ‹ƒ1โˆž๐‘“๐‘—โˆ’1((๐‘Ž,โˆž])

็”ฑไบŽ ๐‘“๐‘— ๅฏๆต‹๏ผŒ้›†ๅˆ {๐‘ฅโˆฃ๐‘“๐‘—(๐‘ฅ)>๐‘Ž} ๆ˜ฏ โ„ณ๏ธ€-measurable ๏ผŒ่€Œๅฏๆต‹้›†ๅˆ็š„ๅฏๆ•ฐๅนถไป็„ถๆ˜ฏๅฏๆต‹็š„๏ผŒๅ› ๆญค ๐‘”1 ๅฏๆต‹ใ€‚

inf: dually.

limsup: ็ญ‰ไบŽ inf of sup (๐‘˜โ‰ฅ๐‘›)

liminf: ็ญ‰ไบŽ sup of inf (๐‘˜โ‰ฅ๐‘›)

โ–ก

Corollary 4.4

ๅฆ‚ๆžœ {๐‘“๐‘›:๐‘‹โ†’โ„ฬ„}๐‘›โˆˆโ„• ๆ˜ฏไธ€ไธช seq of โ„ณ๏ธ€-measurable functions, ไธ”ๅœจไปปๆ„ ๐‘ฅ ๅค„ๆž้™้ƒฝๅญ˜ๅœจ, ้‚ฃไนˆ

๐‘“(๐‘ฅ)โ‰”lim๐‘—โ†’โˆž๐‘“๐‘—(๐‘ฅ)

ๆ˜ฏ โ„ณ๏ธ€-measurable ็š„.

Proof

directly follows from lemma. ๅ› ไธบ ๐‘ฅ ๅค„ๆž้™ๅฆ‚ๆžœๅญ˜ๅœจ, ้‚ฃไนˆ sup๐‘“๐‘“๐‘—(๐‘ฅ)=inf๐‘—๐‘“๐‘—(๐‘ฅ)

โ–ก

Corollary 4.5

๐‘“,๐‘” โ„ณ๏ธ€-measurable โŸน max(๐‘“,๐‘”),min(๐‘“,๐‘”)โ„ณ๏ธ€- measurable

Proof

two element sequence, ๅ‰ฉไฝ™็š„็”จ็ฉบ้›†, ไบŽๆ˜ฏ follows form above.

โ–ก

4.2 simple function and integration of nonnegative functions [Fol 2.1, finished; 2.2]

4.2.1 indicator and simple function

Definition 4.20 : characteristic (indicator) function

Given ๐ธโІ๐‘‹, ๆˆ‘ไปฌๅฎšไน‰:

๐œ’๐ธ(๐‘ฅ)โ‰”{1,๐‘ฅโˆˆ๐ธ0,๐‘ฅโˆ‰๐ธ
Lemma 4.12

ๅฆ‚ๆžœ (๐‘‹,โ„ณ๏ธ€) ๆ˜ฏไธ€ไธช measurable space, ้‚ฃไนˆไธ€ไธช indicator function

๐œ’๐ธ on ๐‘‹ ๆ˜ฏ measurable ็š„ โ‡” ๐ธโˆˆโ„ณ๏ธ€

indicator function measurable ๅฝ“ไธ”ไป…ๅฝ“ๅฎƒ indicate ็š„้›†ๅˆๆ˜ฏ measurable ็š„.

Definition 4.21 : simple function

ไธ€ไธช simple function on measurable space (๐‘‹,๐’œ๏ธ€) ๆ˜ฏไธ€ไธช ๐’œ๏ธ€-measurable function ๐œ™:๐‘‹โ†’โ„‚, taking only finitely many values.

ๅณ: ๐œ™(๐‘‹)={๐‘1,โ‹ฏ,๐‘๐‘˜}

Proposition 4.7 : ไฝฟ็”จ a sum of indicator functions of measurable sets ๆฅๅฎšไน‰ simple function

ๅฏนไบŽ simple function ๐œ™:๐‘‹โ†’โ„‚ s.t. ๐œ™(๐‘‹)={๐‘1,โ‹ฏ,๐‘๐‘›}, ๆˆ‘ไปฌไนŸๅฏไปฅๅฎšไน‰ๅฎƒไธบ:

๐œ™(๐‘ฅ)=โˆ‘๐‘—=1๐‘›๐‘๐‘—๐œ’๐ธ๐‘—

ๅ…ถไธญ, ๐ธ๐‘—=๐œ™โˆ’1({๐‘๐‘—}). ๆˆ‘ไปฌ็งฐไน‹ไธบ: the standard representation of simple ๐œ™.

่ฟ™ๆ˜ฏๅ› ไธบ, ๅ•็‚น้›†ๅœจ โ„ฌ๏ธ€(โ„‚) ไธŠๆ˜ฏ measurable ็š„, ็”ฑไบŽ ๐œ™ measurable, ๆˆ‘ไปฌๅพ—ๅˆฐ ๐ธ๐‘—โˆˆโ„ณ๏ธ€.

Lemma 4.13

ๅฆ‚ๆžœ ๐œ™,๐œ“:๐‘‹โ†’โ„‚ ๆ˜ฏ simple functions, ้‚ฃไนˆ

  • ๐œ™+๐œ“

  • ๐œ™๐œ“

  • |๐œ™|

  • ๐‘˜๐œ™ โˆ€๐‘˜โˆˆโ„‚

้ƒฝๆ˜ฏ simple functions.

็‰นๅˆซๅœฐ, ๅฆ‚ๆžœ ๐œ™,๐œ“:๐‘‹โ†’โ„, ้‚ฃไนˆ max(๐œ™,๐œ“),min(๐œ™,๐œ“) ไนŸๆ˜ฏ simple functions.

Proof

trivial.

โ–ก

4.2.2 measurable function is a limit of simple functions

Theorem 4.15 : approximating a nonneg measurable function by simple function

ไปปๆ„็š„ measurable ๐‘“:๐‘‹โ†’[0,โˆž] ้ƒฝๆ˜ฏ pointwise limit of an increasing sequence of simple functions {๐œ™๐‘›:๐‘‹โ†’[0,โˆž]}๐‘›โˆˆโ„•.

Proof

่ฟ™ไธชๆž„้€ ็œ‹่ตทๆฅๆœ‰็‚นๅคๆ‚ไฝ†ๆ˜ฏๅ…ถๅฎž้žๅธธ็›ด่ง‚.

ๅฏนไบŽ ๐‘›โˆˆโ„•, ๆˆ‘ไปฌ้ƒฝ index 0โ‰ค๐‘˜โ‰ค22๐‘›โˆ’1

็„ถๅŽๅฏนๆฏไธช ๐‘˜ ๅ–:

๐ธ๐‘›๐‘˜โ‰”๐‘“โˆ’1((๐‘˜2๐‘›,๐‘˜+12๐‘›+1])

ไปฅๅŠ:

๐น๐‘›โ‰”๐‘“โˆ’1((2๐‘›,โˆž])

ๅณ, ๆˆ‘ไปฌๆŠŠ (0,2๐‘›] ่ฟ™ไธ€้ƒจๅˆ†ๅ€ผๅŸŸๅˆ‡ๆˆไบ† 22๐‘› ไปฝ, ๅ†ๆŠŠ (2๐‘›,โˆž] ่ฟ™ไธ€้ƒจๅˆ†ๅ€ผๅŸŸๅ•็‹ฌๅˆ—ๆˆไธ€ไปฝ.

่ฟ™ 22๐‘›+1 ไปฝๅ€ผๅŸŸ็š„ๅˆ‡็‰‡, ๆˆ‘ไปฌๅฏนๆฏไธ€ไปฝๆ‰€ๅฏนๅบ”็š„ function graph, ้ƒฝๅ–ๅฎƒๅฏนๅบ”็š„ Preimage ไธŠ็š„ indicator function ไน˜ไปฅ ๐‘˜2๐‘›, ่ฟ™ๆฎตๅ€ผๅŸŸ็š„ๆœ€ๅฐๅ€ผ็š„ constant ๅ‡ฝๆ•ฐ, ไบŽๆ˜ฏไธ€ๅฎšไผšๅพ—ๅˆฐไธ€ไธช well approximation:

๐œ™๐‘›โ‰”โˆ‘๐‘˜=022๐‘›โˆ’1๐‘˜๐‘˜2๐‘›๐œ’๐ธ๐‘›๐‘˜+2๐‘›๐œ’๐น๐‘›

ๆ˜“ๅพ—,

๐œ™๐‘›โ‰ค๐œ™๐‘›+1โ‰ค๐‘“

for all ๐‘›. ๅนถไธ”ๅœจ ๐‘‹\๐น๐‘›={๐‘ฅโˆฃ๐‘“(๐‘ฅ)โ‰ค2๐‘›} ไธŠๆˆ‘ไปฌๆœ‰:

0โ‰ค๐‘“โˆ’๐œ™๐‘›โ‰ค12๐‘›

้š็€ ๐‘› ๅขžๅคง, ๆœ€็ปˆ่ฟ™ไธช่ฟ‘ไผผไผš่ฆ†็›–ๆ•ดไธช image, (้™ค้žๅ…ทๆœ‰้ž้›ถๆต‹ๆ•ฐ้‡็š„ๆ— ็ฉท้—ดๆ–ญ็‚น, ้‚ฃๆ ท็š„่ฏๆœ€ๅŽ็ป“ๆžœไนŸๆ˜ฏๆ— ็ฉท), ๅนถไธ”ๅ€ผๅŸŸ็š„ๅˆ’ๅˆ†่ถŠๆฅ่ถŠ็ฒพ็ป†, ๆœ€ๅŽไผšๅพ—ๅˆฐ:

  • ๐œ™๐‘›โ†’๐‘“ pointwisely

  • ๅœจ ๐‘“ bounded ็š„ๅฎšไน‰ๅŸŸ {๐‘ฅโˆฃ๐‘“(๐‘ฅ)<โˆž} ไธŠ, ๐œ™๐‘›โ†’๐‘“ uniformly.

Figureย 13:

โ–ก

Corollary 4.6 : simple approximation of complex measurable functions

ๅฏนไบŽไปปๆ„็š„ measurable ๐‘“:๐‘‹โ†’โ„‚, ้ƒฝๅญ˜ๅœจ a seq of simple functions

0โ‰ค|๐œ™1|โ‰ค|๐œ™2|โ‰คโ‹ฏโ‰ค|๐‘“|

ไฝฟๅพ—

  • ๐œ™๐‘›โ†’๐‘“ pointwisely

  • ๐œ™๐‘›โ†’๐‘“ uniformly on {๐‘ฅโˆฃ|๐‘“(๐‘ฅ)|<โˆž}

Proof

ๆˆ‘ไปฌๅฏไปฅๆŠŠ ๐‘“ ๆ‹†ไธบ Im๐‘“,Re๐‘“, ็„ถๅŽๅ†ๆŠŠๅฎƒไปฌๅˆ†ๅˆซๆ‹†ไธบ Im๐‘“+โˆ’Im๐‘“โˆ’, ไปฅๅŠ Re๐‘“+โˆ’Re๐‘“โˆ’. ๅพ—ๅˆฐๅ››ไธช real-valued nonng functions.

โ–ก

4.2.3 integration of non-neg functions

Definition 4.22 : ๐ฟ+ space and integration on it

็ป™ๅฎšไธ€ไธช measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡) ๆˆ‘ไปฌๅฎšไน‰:

๐ฟ+(๐œ‡)โ‰”{๐ฆ๐ž๐š๐ฌ๐ฎ๐ซ๐š๐›๐ฅ๐ž ๐Ÿ๐ฎ๐ง๐œ๐ญ๐ข๐จ๐ง๐ฌ ๐‘“:๐‘‹โ†’[0,โˆž]}

ๅฏนไบŽๆ‰€ๆœ‰็š„ simple functions ๐œ™=โˆ‘๐‘—=1๐‘›๐‘Ž๐‘—๐œ’๐ธ๐‘—โˆˆ๐ฟ+(๐œ‡), ๅณๆ‰€ๆœ‰้ž่ดŸ็š„ simple functions, ๆˆ‘ไปฌๅฎšไน‰ the integral of ๐œ™ with respect to ๐œ‡ by:

โˆซ๐œ™๐‘‘๐œ‡(=โˆซ๐‘‹๐œ™๐‘‘๐œ‡)โ‰”โˆ‘๐‘–=1๐‘›๐‘Ž๐‘—๐œ‡(๐ธ๐‘—)

ๅฏนไบŽไปปๆ„็š„ ๐‘“โˆˆ๐ฟ+(๐œ‡), ๆˆ‘ไปฌๅฎšไน‰ the integral of ๐‘“ with respect to ๐œ‡ by:

โˆซ๐‘“๐‘‘๐œ‡(=โˆซ๐‘‹๐‘“๐‘‘๐œ‡)โ‰”sup{โˆซ๐œ™๐‘‘๐œ‡โˆฃ0โ‰ค๐œ™โ‰ค๐‘“,๐œ™ simple}
Definition 4.23 : integration on a subset

ๅฏน้ž่ดŸ simple functions ๐œ™=โˆ‘๐‘—=1๐‘›๐‘Ž๐‘—๐œ’๐ธ๐‘—โˆˆ๐ฟ+(๐œ‡), ๆˆ‘ไปฌๅฎšไน‰ the integral of ๐œ™ on ๐ดโˆˆโ„ณ๏ธ€ with respect to ๐œ‡ by:

โˆซ๐ด๐œ™๐‘‘๐œ‡โ‰”โˆซ๐œ™๐œ’๐ด๐‘‘๐œ‡

ๅฏนไบŽ general ็š„ ๐‘“โˆˆ๐ฟ+(๐œ‡), ๆˆ‘ไปฌไนŸไปŽ่€Œๅฎšไน‰:

โˆซ๐ด๐‘“๐‘‘๐œ‡โ‰”sup{โˆซ๐ด๐œ™๐‘‘๐œ‡โˆฃ0โ‰ค๐œ™โ‰ค๐‘“,๐œ™ simple}
Proposition 4.8 : integral of simple functions ็š„ๆ€ง่ดจ

Let ๐œ™,๐œ“ be simple functions in ๐ฟ+(๐œ‡), ๆœ‰:

  • homogeneity: ๅฏนไบŽไปปๆ„้ž่ดŸ ๐‘, ๆœ‰ โˆซ๐‘๐œ™=๐‘โˆซ๐œ™

  • linearity: โˆซ(๐œ™+๐œ“)=โˆซ๐œ™+โˆซ๐œ“

  • monotonicity: ๐œ™โ‰ค๐œ“โŸนโˆซ๐œ™โ‰คโˆซ๐œ“

  • induced measure: ๐ดโ†ฆโˆซ๐ด๐œ™๐‘‘๐œ‡ ๆ˜ฏไธ€ไธช โ„ณ๏ธ€ ไธŠ็š„ measure.

Proof

homogeneity trivial .

linearity: Let

๐œ™=โˆ‘๐‘–=1๐‘›๐‘Ž๐‘–๐œ’๐ธ๐‘–,๐œ“=โˆ‘๐‘—=1๐‘›๐‘๐‘—๐œ’๐น๐‘—

ๅˆ™ๆœ‰:

๐ธ๐‘—=โจ†๐‘˜(๐ธ๐‘—โˆฉ๐น๐‘˜),๐น๐‘˜=โจ†๐‘—(๐ธ๐‘—โˆฉ๐น๐‘˜)

for each ๐‘—,๐‘˜. ไปŽ่€Œๆœ‰

โˆซ๐œ™+โˆซ๐œ“=โˆ‘๐‘—,๐‘˜(๐‘Ž๐‘—+๐‘๐‘˜)๐œ‡(๐ธ๐‘—โˆฉ๐น๐‘˜)

Monotonicity: trivial.

induced measure: ๅช้œ€่ฆ่ฏๆ˜Ž countable additivity, ไบŽๆ˜ฏๆˆ‘ไปฌ่ฎฉ ๐ด be the union of a disjoint seq in โ„ณ๏ธ€,ๆœ‰:

โˆซ๐ด๐œ™=โˆ‘๐‘—๐‘Ž๐‘—๐œ‡(๐ดโˆฉ๐ธ๐‘—)=โˆ‘๐‘—,๐‘˜๐‘Ž๐‘—๐œ‡(๐ด๐‘˜โˆฉ๐ธ๐‘—)=โˆ‘๐‘˜โˆซ๐ด๐‘˜๐œ™

โ–ก

้‚ฃไนˆๅฏนไบŽ general ็š„ ๐‘“โˆˆ๐ฟ+(๐œ‡), ๆœ‰ๅˆšๆ‰็š„ๅ››ๆกๆ€ง่ดจๆˆ็ซ‹ๅ—? ๆ˜พ็„ถ, monotonicity ๅ’Œ homogeinity ๆ˜ฏๆˆ็ซ‹็š„, ไฝ†ๆ˜ฏๆˆ‘ไปฌไผšๅ‘็Žฐ, ๅพˆ้šพ่ฏๆ˜Ž

โˆซ๐‘“๐‘‘๐œ‡+โˆซ๐‘”๐‘‘๐œ‡=โˆซ(๐‘“+๐‘”)๐‘‘๐œ‡

โ‰ค ๆ˜ฏๅฎนๆ˜“่ฏๆ˜Ž็š„, ไฝ†ๆ˜ฏ โ‰ฅ ๆœ‰็‚นๅ›ฐ้šพ. ไธบไบ†่ฏๆ˜Ž โ‰ฅ ่ฟ™ไธชๆ–นๅ‘, ๆˆ‘ไปฌ้œ€่ฆไธ‹้ข่ฟ™ไธช้‡่ฆๅฎš็†:

4.2.4 MCT

Theorem 4.16 : monotone convergence theorem

Let {๐‘“๐‘›}๐‘›โˆˆโ„• be a seq in ๐ฟ+(๐œ‡), ๅนถไธ”ๆœ‰ ๐‘“๐‘›โ‰ค๐‘“๐‘›+1 for each ๐‘›.
ๆˆ‘ไปฌ define:

๐‘“โ‰”lim๐‘›๐‘“๐‘›(=sup๐‘›๐‘“๐‘›)

, ๅˆ™ไธ€ๅฎšๆœ‰

โˆซ๐‘“=lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›
Proof

้ฆ–ๅ…ˆ Note ๅ‡ ไธชไบ‹ๆƒ…: 1. ่ฟ™ไธชๆž้™ๅ‡ฝๆ•ฐ ๐‘“ ๆ˜ฏ well-defined ็š„ (ๅฏ่ƒฝ โˆž), by numerical sequence ็š„ monotone bounded convergence theorem.

2. ๅŒๆ ทๅœฐ, ็”ฑไบŽ โˆซ๐‘“๐‘›โ‰คโˆซ๐‘“๐‘›+1โ‰คโˆซ๐‘“, ่ฟ™ไธช limโˆซ๐‘“๐‘› ไนŸๆ˜ฏๅญ˜ๅœจ็š„.

3. ๅนถไธ”, ๐‘“ ไนŸๆ˜ฏไธ€ไธชๅฏๆต‹ๅ‡ฝๆ•ฐ, ๅ› ไธบ by ไธŠไธช lecture ็š„ๅฎš็†: ๅฏๆต‹ๅ‡ฝๆ•ฐๅบๅˆ—็š„ๆž้™ไนŸๆ˜ฏๅฏๆต‹ๅ‡ฝๆ•ฐ.

็Žฐๅœจ่ฟ›่กŒ่ฏๆ˜Ž: By monotonicity of integral,

limโˆซ๐‘“๐‘›โ‰คโˆซ๐‘“

ๆ˜ฏ natural ็š„. ๅ› ่€Œๅช้œ€่ฆ่ฏๆ˜Žๅฆไธ€ๆ–นๅ‘.

By def, โˆซ๐‘“=sup{โˆซ๐œ™โˆฃ๐œ™โ‰ค๐‘“} where ๐œ™ is simple. ๅ› ่€Œ it suffices to show: ๅฏนไบŽไปปๆ„ simple ๐œ™โ‰ค๐‘“, ้ƒฝๆœ‰ limโˆซ๐‘“๐‘›โ‰ฅโˆซ๐œ™.

ๆˆ‘ไปฌ fix ไธ€ไธช 0โ‰ค๐œ™โ‰ค๐‘“. WTS:

lim๐‘›โˆซ๐‘“๐‘›โ‰ฅโˆซ๐œ™

่ฆ่ฏๆ˜Ž limโˆซ๐‘“๐‘›โ‰ฅโˆซ๐œ™, ๆˆ‘ไปฌๅ†ๆŠŠๅฎƒ่ฝฌๅŒ–ๆˆ่ฏๆ˜Ž:

โˆ€๐›ผโˆˆ(0,1)lim๐‘›โˆซ๐‘“๐‘›โ‰ฅ๐›ผโˆซ๐œ™

ๆˆ‘ไปฌๅ–

๐ธ๐‘›โ‰”{๐‘ฅโˆฃ๐‘“๐‘›(๐‘ฅ)โ‰ฅ๐›ผ๐œ™}=๐‘“โˆ’1([๐›ผ๐œ™,โˆž])โˆˆโ„ณ๏ธ€

ๅฎนๆ˜“ๅ‘็Žฐ, ๐ธ๐‘›โІ๐ธ๐‘›+1 for each ๐‘›. ๅนถไธ” Claim: โ‹ƒ๐‘›๐ธ๐‘›=๐‘‹. (่ฟ™ๅฐฑๆ˜ฏไธบไป€ไนˆ่ฆๅšๅ– ๐›ผ ่ฟ™ไธชๆ„ไน‰ไธๆ˜Ž็š„่กŒไธบ) ่ฟ™ๆ˜ฏๅ› ไธบ ๐›ผ<1, ๅนถไธ” ๐‘“๐‘› converge pointwisely to ๐‘“, by measurable function ็š„ limit behavior. ่€Œ็”ฑไบŽ simple function ๐œ™ ๆ˜ฏ bounded ็š„, ไปŽ่€Œ ๐‘“๐‘› ไผš uniformly ๅ‘ไธŠๆŽฅ่ฟ‘(ไปฅ่‡ณไบŽ่ถ…่ฟ‡) ๐œ™. ๅ– ๐›ผ ๆ˜ฏไธบไบ†ไฟ่ฏ, ไธ€ๅฎšๅญ˜ๅœจไธ€ไธช ๐‘› ไฝฟๅพ— ๐ธ๐‘›=๐‘‹

ไบŽๆ˜ฏๆˆ‘ไปฌๆœ‰:

โˆซ๐‘“๐‘›โ‰ฅโˆซ๐‘“๐‘›๐œ’๐ธ๐‘›โ‰ฅโˆซ๐›ผ๐œ™๐œ’๐ธ๐‘›=๐›ผโˆซ๐ธ๐‘›๐œ™

ๆˆ‘ไปฌๆญคๅค„ๅˆๅฏไปฅ็”จๅˆฐไธ€ๆกๅ†ท้—จ็š„ๆ€ง่ดจ: ็”ฑไบŽ ๐ธโ†ฆโˆซ๐ธ๐œ™ ๆ˜ฏไธ€ไธช measure on (๐‘‹,๐’œ๏ธ€), by continuous from below, ๆœ‰:

lim๐‘›โˆซ๐ธ๐‘›๐œ™=โˆซ๐œ™

ไปŽ่€Œๆœ‰

lim๐‘›โˆซ๐‘“๐‘›โ‰ฅ๐›ผโˆซ๐œ™

finishing the proof.

โ–ก

ไปฅไธ‹ไธบไธ€ไธชๅบ”็”จ MCT ๅพ—ๅˆฐ็š„็ป“่ฎบ.

Example 4.9

ๅ–

(โ„•,๐’ซ๏ธ€(โ„•),๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘–๐‘›๐‘”)

ไบŽๆ˜ฏ

๐ฟ+(๐œ‡)={๐‘“:โ„•โ†’[0,โˆž]}

ๆ˜ฏๆ‰€ๆœ‰็š„ไปŽ่‡ช็„ถๆ•ฐๅˆฐ reals ็š„ๅ‡ฝๆ•ฐ. (ๅ› ไธบๆˆ‘ไปฌๅ–ไบ† power set ไฝœไธบ ๐œŽ-algebra)

ๆณจๆ„ๅˆฐไปปไฝ•ไธ€ไธช่ฟ™ๆ ท็š„ๅ‡ฝๆ•ฐ้ƒฝๅฏไปฅ่ขซ

๐œ™๐‘›โ‰”โˆ‘๐‘—=1๐‘›๐‘“(๐‘—)๐œ‡({๐‘—})=โˆ‘๐‘—=1๐‘›๐‘“(๐‘—)

ๆฅ้€ผ่ฟ‘. ไปŽ่€Œ

โˆซ๐‘“=โˆ‘๐‘—=1โˆž๐‘“(๐‘—)โˆˆ[0,โˆž]

ๅฆ‚ๆžœๅ–ไธ€ไธชไปŽไธ‹้€ผ่ฟ‘ ๐‘“ ็š„ๅฏๆต‹ๅ‡ฝๆ•ฐๅบๅˆ— {๐‘“๐‘›}๐‘›โˆˆโ„•, ้‚ฃไนˆ by MCT, ๆˆ‘ไปฌๆ€ปๆœ‰:

โˆ‘1โˆž๐‘“๐‘›(๐‘—)โ†—๏ธŽโˆ‘1โˆž๐‘“(๐‘—)

4.2.5 (countable) linearity of integral

Corollary 4.7
๐‘“,๐‘”โˆˆ๐ฟ+(๐œ‡)โ‡’โˆซ(๐‘“+๐‘”)=โˆซ๐‘“+โˆซ๐‘”
Proof

ไฝฟ็”จ approximation by simple functions ไปฅๅŠ MCT. ๅ–

๐œ™๐‘›โ†—๏ธŽ๐‘“,๐œ“๐‘›โ†—๏ธŽ๐‘”

, ไปŽ่€Œ

๐œ™๐‘›+๐œ“๐‘›โ†—๏ธŽ๐‘“+๐‘”

, ไปŽ่€Œๆˆ‘ไปฌๆœ‰

โˆซ(๐‘“+๐‘”)=๐‘€๐ถ๐‘‡lim๐‘›โˆซ(๐œ™๐‘›+๐œ“๐‘›)

ไปŽ่€Œ็”ฑ simple function ็š„ Linearity ๅพ—ๅˆฐ:

โˆซ(๐‘“+๐‘”)=lim๐‘›โˆซ๐œ™๐‘›+lim๐‘›โˆซ๐œ“๐‘›

ๅนถไธ”็”ฑไบŽ

โˆซ๐œ™๐‘›โ†—๏ธŽโˆซ๐‘“,โˆซ๐œ“๐‘›โ†—๏ธŽโˆซ๐‘”

ๆˆ‘ไปฌๅพ—ๅˆฐ:

โˆซ(๐‘“+๐‘”)โ‰ฅโˆซ๐‘“+โˆซ๐‘”

ๅฆไธ€ๆ–นๅ‘ trivial.

โ–ก

4.2.6 Tonelli for sum and integrals

Corollary 4.8 : Tonelli for sum and integrals

for {๐‘“๐‘–}๐‘–โˆˆโ„• in ๐ฟ+(๐œ‡), ๆœ‰:

โˆซโˆ‘๐‘–=1โˆž๐‘“๐‘–=โˆ‘๐‘–=1โˆžโˆซ๐‘“๐‘–
Proof

Apply MCT to

๐‘”๐‘›=โˆ‘๐‘–=1๐‘›๐‘“๐‘–

ๅฏๅพ—่ฏ.

โ–ก

4.3 properties of integration on ๐ฟ+(๐œ‡) [Fol 2.2, finished]

4.3.1 Fatouโ€™s Lemma

Theorem 4.17 : Fatouโ€™s Lemma

ไปค (๐‘“๐‘›) be a seq of functions in ๐ฟ+(๐œ‡), then

limโ€‰inf๐‘›โˆซ๐‘“๐‘›โ‰ฅโˆซlimโ€‰inf๐‘›๐‘“๐‘›
Proof

Set

๐‘”๐‘›โ‰”inf๐‘šโ‰ฅ๐‘›๐‘“๐‘›

ไบŽๆ˜ฏ

๐‘”๐‘›โ†—๏ธŽlimโ€‰inf๐‘›๐‘“๐‘›

ไบŽๆ˜ฏ by MCT, we have:

lim๐‘›โˆซ๐‘”๐‘›=โˆซlim๐‘›๐‘”๐‘›=โˆซlimโ€‰inf๐‘›๐‘“๐‘›

By def, ๆˆ‘ไปฌๆœ‰ ๐‘”๐‘›โ‰ค๐‘“๐‘›โˆ€๐‘›, ไบŽๆ˜ฏ by monotonicity, โˆซ๐‘”๐‘›โ‰คโˆซ๐‘“๐‘›. ๅ› ่€Œ

limโ€‰inf๐‘›โˆซ๐‘“๐‘›โ‰ฅlimโ€‰inf๐‘›โˆซ๐‘”๐‘›=lim๐‘›โˆซ๐‘”๐‘›=โˆซlimโ€‰inf๐‘›๐‘“๐‘›

โ–ก

Example 4.10

ๅ– (โ„,๐”,๐‘š), ่€ƒ่™‘ ๐ฟ+(๐‘š) ไธŠ็š„ๅ‡ฝๆ•ฐ, ๅณ้ž่ดŸ Lebesgue ๅฏๆต‹ๅ‡ฝๆ•ฐ.

ไธ‹้ขๆœ‰ๅ‡ ไธช้žๅธธ็ปๅ…ธ็š„ Fatouโ€™s Lemma ็š„ไพ‹ๅญ:

. escape to hat:

๐‘“๐‘›=๐œ’(๐‘›,๐‘›+1)

๐‘“๐‘› ๅœจ โ„ ไธŠๅนณ็งป

. escape to width:

๐‘“๐‘›=1๐‘›๐œ’(0,๐‘›)

๐‘“๐‘›้€ๆธๅ˜ๅพ—ๅนณๅฆ

. escape to height:

๐‘“๐‘›=๐‘›๐œ’(0,1๐‘›)

๐‘“๐‘› ้€ๆธๅ˜ๆˆไธ€ๆ น้’ˆ.

่ฟ™ไธ‰ไธชไพ‹ๅญไธญ้ƒฝๆœ‰ ๐‘“๐‘›โ†’0 pointwisely. ๅ› ่€Œ

โˆซlim๐‘“๐‘›=0

, ่€Œ

limโˆซ๐‘“๐‘›=1

, ๅ› ไธบๅฏนไบŽๆ‰€ๆœ‰ ๐‘“๐‘› ้ƒฝๆœ‰ โˆซ๐‘“๐‘›=1

4.3.2 Chebyshevโ€™s inequality with corollaries

Lemma 4.14 : Chebyshevโ€™s inequality

ๅฏนไบŽ measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡), ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ+(๐œ‡) ๅนถไธ” ๐‘>0, ้‚ฃไนˆ

๐œ‡{๐‘“โ‰ฅ๐‘}โ‰ค1๐‘โˆซ๐‘“
Proof

Let ๐ธโ‰”๐œ‡{๐‘“โ‰ฅ๐‘}

โˆซ๐‘“โ‰ฅโˆซ๐‘“๐œ’๐ธโ‰ฅโˆซ๐‘๐œ’๐ธ=๐‘โˆซ๐œ’๐ธ=๐‘๐œ‡(๐ธ)

โ–ก

Proposition 4.9 : ้ž่ดŸๅ‡ฝๆ•ฐ็งฏๅˆ†ไธบ 0 ็ญ‰ไปทไบŽๅ‡ ไนŽๅค„ๅค„ไธบ 0

ไปค ๐‘“โˆˆ๐ฟ+(๐œ‡), ๆœ‰:

โˆซ๐‘“=0 โ‡” ๐‘“=0 a.e. (ๅณๅชๅœจไธ€ไธช้›ถๆต‹้›†ไธŠ้ž 0)

Proof

forward direction: directly follows from Chebyshev: set ๐ด๐‘›โ‰”{๐‘“โ‰ฅ1๐‘›}, ๅฏนไบŽไปปๆ„ ๐‘› ้ƒฝๆœ‰ ๐œ‡(๐ด๐‘›)โ‰ค๐‘›โˆซ๐‘“=0. ไปŽ่€Œ by ctn from below, >0 ๅค„ๆž„ๆˆ้›ถๆต‹้›†.

backward direction: ๅฏนไบŽ simple function, trivial by ็งฏๅˆ†็š„ๅฎšไน‰; ๅฏนไบŽ general ๐‘“, ้€š่ฟ‡ limit ๅพ—ๅˆฐ (ๅฎƒไธ‹ๆ–น็š„ๆ‰€ๆœ‰ simple functions ไนŸ a.e. ไธบ 0 ไปŽ่€Œ็งฏๅˆ†ไธบ 0).

โ–ก

Corollary 4.9 : ๅ‡ ไนŽๅค„ๅค„็›ธ็ญ‰็š„้ž่ดŸๅ‡ฝๆ•ฐ็งฏๅˆ†็›ธ็ญ‰

Let ๐‘“,๐‘”โˆˆ๐ฟ+(๐œ‡) ไธ” ๐‘“=๐‘” a.e., ๅˆ™ๆœ‰

โˆซ๐‘“=โˆซ๐‘”
Proof

Set ๐ท:={๐‘ฅโˆฃ๐‘“(๐‘ฅ)โ‰ ๐‘”(๐‘ฅ)}, ๅˆ™ ๐œ‡(๐ท)=0 by def

โˆซ๐‘“=โˆซ๐ท๐‘“+โˆซ๐ท๐‘๐‘“=0+โˆซ๐ท๐‘๐‘”=โˆซ๐‘”

โ–ก

Corollary 4.10 : liminf version of MCT

suppose (๐‘“๐‘›)๐‘›โˆˆโ„• ๆ˜ฏไธ€ไธช seq of functions in ๐ฟ+(๐œ‡), ไธ” ๐‘“๐‘›โ†’๐‘“โˆˆ๐ฟ+(๐œ‡), ๅˆ™:

limโ€‰inf๐‘›โˆซ๐‘“๐‘›โ‰ฅโˆซ๐‘“
Proof

่ฟ™ๆ˜ฏไธ€ไธชๆกไปถ็จๅพฎๅผฑๅŒ–็š„ MCT: ๆŠŠ ๐‘“๐‘›โ†—๏ธŽ๐‘“ ็š„ๆกไปถๆ”นๆˆไบ† ๐‘“๐‘›โ†’๐‘“ a.e., ๅพ—ๅˆฐ็š„็ป“่ฎบไนŸ็จๅผฑๅŒ–.
modify ๐‘“๐‘› and ๐‘“ on a null set (thus without chaning the integral) ๅŽ, follows directly from Fatouโ€™s lemma,

โ–ก

Theorem 4.18 : ็งฏๅˆ†ๆ”ถๆ•› โŸน ๅ‘ๆ•ฃ็‚น้›†้›ถๆต‹, ไปฅๅŠ support ๐œŽ-finite

ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ+(๐œ‡) ไธ” |โˆซ๐‘“|<โˆž, ๅˆ™ๆœ‰:

๐œ‡({๐‘ฅโˆˆ๐‘‹โˆฃ๐‘“(๐‘ฅ)=โˆž})=0

ๅนถไธ”

{๐‘ฅโˆฃ๐‘“(๐‘ฅ)>0}

is ๐œŽ-finite

Proof

็›ดๆŽฅ follows from Chebyshev. ๅ–

๐ด๐‘กโ‰”{๐‘ฅโˆฃ๐‘“(๐‘ฅ)โ‰ฅ๐‘ก}

for ๐‘ก>0.

ไบŽๆ˜ฏ:

{๐‘ฅโˆˆ๐‘‹โˆฃ๐‘“(๐‘ฅ)=โˆž}=โ‹‚๐‘›=1โˆž๐ด๐‘›

By Chebyshev, each ๐ด๐‘› ้ƒฝๆœ‰: ๐œ‡(๐ด๐‘›)โ‰ค1๐‘›โˆซ๐‘ก, ไปŽ่€Œ by continuous from above ๅฏๅพ—่ฟ™ไธชไบค้›†็š„ measure ไธบ 0.

ๅˆๆœ‰:

{๐‘ฅโˆˆ๐‘‹โˆฃ๐‘“(๐‘ฅ)>0}=โ‹ƒ๐‘›=1โˆž๐ด1๐‘›

ๅ…ถไธญ, each set has measure โ‰ค๐‘›โˆซ๐‘“โ‰คโˆž. By def, ่ฟ™ไธช้›†ๅˆ ๐œŽ-finite.

โ–ก

Homework 4: on measurable functions(36/40)

None of the following questions will be graded. Do them, but do not hand them in.

One with Vitali.

Let (๐‘‹,๐’œ๏ธ€) be a measurable space, and ๐ธโŠ‚๐‘‹ a subset. Prove that ๐ธโˆˆ๐’œ๏ธ€ iff the function ๐œ’๐ธ is measurable. Use this to construct a function ๐‘“:โ„โ†’โ„ that is not Lebesgue measurable.

Truncations in ๐ฟ+: ้€š่ฟ‡ โˆซ๐‘“๐‘› ๆˆ–่€… โˆซ๐‘‹๐‘›๐‘“ ็š„ๆž้™ (bounded function / subset) ๅพ—ๅˆฐ โˆซ๐‘‹๐‘“

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space and ๐‘“:๐‘‹โ†’[0,โˆž] a measurable function.

  • (Horizontal truncation) Suppose that ๐‘‹=โ‹ƒ๐‘›=1โˆž๐‘‹๐‘› for some ๐‘‹1โŠ‚๐‘‹2โŠ‚โ‹ฏ with ๐‘‹๐‘›โˆˆ๐’œ๏ธ€. Prove that

    โˆซ๐‘‹๐‘“๐‘‘๐œ‡=lim๐‘›โ†’โˆžโˆซ๐‘‹๐‘›๐‘“๐‘‘๐œ‡
  • (Vertical truncation) Prove that

    โˆซ๐‘“๐‘‘๐œ‡=lim๐‘›โ†’โˆžโˆซmin{๐‘“,๐‘›}๐‘‘๐œ‡.
  • Explain the terminology โ€œhorizontal truncationโ€ and โ€œvertical truncationโ€.

Disregarding null sets.

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a complete measure space.

  • Let ๐‘“:๐‘‹โ†’โ„ฬ„ and ๐‘”:๐‘‹โ†’โ„ฬ„ be functions such that ๐‘“=๐‘” ๐œ‡-a.e.

    • Prove that ๐‘“ is measurable (i.e.ย ๐’œ๏ธ€-measurable) iff ๐‘” is measurable.

    • Prove the same statement when ๐‘“ and ๐‘” are โ„‚-valued, rather than โ„ฬ„-valued.

    • Give examples showing that the condition that ๐œ‡ be complete is necessary.

  • Let ๐‘“๐‘›:๐‘‹โ†’โ„ฬ„, ๐‘›โˆˆโ„•, and ๐‘“:๐‘‹โ†’โ„ฬ„ be functions such that lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ) for a.e. ๐‘ฅโˆˆ๐‘‹.

    • Prove that if ๐‘“๐‘› is measurable for all ๐‘›, then so is ๐‘“.

    • Prove the same statement when ๐‘“๐‘› and ๐‘“ are โ„‚-valued, rather than โ„ฬ„-valued.

    • Give examples showing that the condition that ๐œ‡ be complete is necessary.

Hint: this is Proposition 2.11 of [Folland].

Measurable functions and completions.

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space and let (๐‘‹,๐’œ๏ธ€ฬ„,๐œ‡ฬ„) be its completion. Suppose that ๐‘“:๐‘‹โ†’โ„ฬ„ is ๐’œ๏ธ€ฬ„-measurable. Prove that there is an ๐’œ๏ธ€-measurable function ๐‘”:๐‘‹โ†’โ„ฬ„ such that ๐‘”=๐‘“ ๐œ‡ฬ„-a.e., and hence โˆซ๐‘”๐‘‘๐œ‡=โˆซ๐‘“๐‘‘๐œ‡ฬ„. Hint: this is Proposition 2.12 of [Folland].

Measurability on subsets.

Let (๐‘‹,๐’œ๏ธ€) be a measurable space, and ๐‘ŒโŠ‚๐‘‹ a nonempty subset. We say that a function ๐‘”:๐‘Œโ†’โ„ฬ„ is ๐’œ๏ธ€-measurable on ๐‘Œ if ๐‘” is ๐’œ๏ธ€|๐‘Œ-measurable, where the ๐œŽ-algebra ๐’œ๏ธ€|๐‘Œ on ๐‘Œ is defined as in HW1.

  • Prove that if ๐‘“:๐‘‹โ†’โ„ฬ„ is measurable and ๐‘ŒโŠ‚๐‘‹, then ๐‘”=๐‘“|๐‘Œ is ๐’œ๏ธ€-measurable on ๐‘Œ.

  • Prove that if ๐‘” is ๐’œ๏ธ€-measurable on ๐‘Œ and ๐‘Œโˆˆ๐’œ๏ธ€, then ๐‘” can be extended to an ๐’œ๏ธ€-measurable function ๐‘“ on ๐‘‹. Is the extension unique?

  • Let ๐‘“:๐‘‹โ†’โ„ฬ„ be any function, and set ๐‘Œ=๐‘“โˆ’1(โ„). Prove that ๐‘“ is measurable iff ๐‘“โˆ’1({โˆž})โˆˆ๐’œ๏ธ€, ๐‘“โˆ’1({โˆ’โˆž})โˆˆ๐’œ๏ธ€, and ๐‘“|๐‘Œ:๐‘Œโ†’โ„ is ๐’œ๏ธ€-measurable on ๐‘Œ.

Suprema of uncountable families.

Construct (using the Axiom of Choice, if needed) an uncountable family (๐‘“๐›ผ)๐›ผ of real-valued Borel measurable functions on โ„ such that the function sup๐›ผ๐‘“๐›ผ is not Lebesgue measurable, let alone Borel measurable.

Increasing functions again.

Let ๐‘“:โ„โ†’โ„ be an increasing function. Prove that ๐‘“ is Borel measurable. Use this to give an example of a function ๐‘“:โ„โ†’โ„ that cannot be written as a difference between increasing functions.

Lebesgue but not Borel.

Let ๐น:[0,1]โ†’[0,1] be the function from HW3, whose graph is the Devilโ€™s Staircase. Define ๐บ(๐‘ฅ)=๐น(๐‘ฅ)+๐‘ฅ.

  • Prove that ๐บ:[0,1]โ†’[0,2] is an increasing homeomorphism. In other words, ๐บ is increasing, bijective, and both ๐บ and ๐บโˆ’1 are continuous.

  • Let ๐ถ be the middle-thirds Cantor set, and set ๐พโ‰”๐บ(๐ถ). Prove that ๐‘š(๐พ)=1.

  • Since ๐‘š(๐พ)>0, we know from HW3 that there is a set ๐ดโŠ‚๐พ that is not Lebesgue measurable. Prove that ๐ต=๐บโˆ’1(๐ด) is Lebesgue measurable but not Borel measurable.

Measurability and absolute values.

Let (๐‘‹,๐’œ๏ธ€) be a measure space. Suppose that ๐‘“:๐‘‹โ†’โ„‚ is a measurable function. Prove that the function |๐‘“|:๐‘‹โ†’โ„ is also measurable. Is the converse true?

Some of the following questions will be graded. Do them, and do hand them in. You may use the results from the exercises above.

Measurability of limit loci.

Let (๐‘‹,๐’œ๏ธ€) be a measurable space. For each ๐‘›โˆˆโ„•, let ๐‘“๐‘›:๐‘‹โ†’โ„ be a measurable function. Consider the set

๐ธโ‰”{๐‘ฅโˆˆ๐‘‹โˆฃlim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)converges to a real number}.

Prove that ๐ธ is a measurable set in two ways:

  • by expressing ๐ธ in terms of the functions ๐‘”(๐‘ฅ)=limโ€‰sup๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ) and โ„Ž(๐‘ฅ)=limโ€‰inf๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ);

  • by expressing ๐ธ in terms of the sets

    ๐ธ๐‘–,๐‘—,๐‘˜={๐‘ฅโˆฃ|๐‘“๐‘—(๐‘ฅ)โˆ’๐‘“๐‘˜(๐‘ฅ)|<1๐‘–},

    where ๐‘–,๐‘—,๐‘˜โˆˆโ„•. Hint: a sequence (๐‘Ž๐‘›)๐‘› of real numbers converges iff it is a Cauchy sequence, i.e. for every ๐œ–>0 there is ๐‘› such that for every ๐‘—,๐‘˜โ‰ฅ๐‘›, |๐‘Ž๐‘—โˆ’๐‘Ž๐‘˜|<๐œ–.

Hint: note that ยฑโˆž are not real numbers, and please avoid considering โˆžโˆ’โˆž; you may want to prove a lemma to the effect that if ๐‘”,โ„Ž:๐‘‹โ†’โ„ฬ„ are measurable functions, then the set

{๐‘ฅโˆˆ๐‘‹โˆฃ๐‘”(๐‘ฅ)=โ„Ž(๐‘ฅ)โˆˆโ„ฬ„}

is measurable; to do this, you may want to consider functions like max{๐‘”,๐œ…}, min{โ„Ž,๐œ…} and min{๐‘”,โˆ’๐œ…}, min{โ„Ž,โˆ’๐œ…} for large real constants ๐œ…>0.

Proof

of method (i):
Define:

๐‘”(๐‘ฅ)โ‰”limโ€‰sup๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)andโ„Ž(๐‘ฅ)โ‰”limโ€‰inf๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)

Since each ๐‘“๐‘› is measurable function, by proposition in lecture (sequential preservation of measurability), ๐‘”,โ„Ž are measurable.

And as we know, for any real sequence (๐‘Ž๐‘›),

lim๐‘›โ†’โˆž๐‘Ž๐‘› exists (as a real number)โ‡”limโ€‰sup๐‘›โ†’โˆž๐‘Ž๐‘›=limโ€‰inf๐‘›โ†’โˆž๐‘Ž๐‘›โˆˆโ„

Thus, for each ๐‘ฅโˆˆ๐‘‹ we have:

๐‘ฅโˆˆ๐ธโ‡”limโ€‰sup๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=limโ€‰inf๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)โˆˆโ„

Thus, we can write ๐ธ as:

๐ธ={๐‘ฅโˆˆ๐‘‹โˆฃ๐‘”(๐‘ฅ)=โ„Ž(๐‘ฅ)โˆˆโ„}

Note: here we want to have a difference function of the two functions, but it is undefined on โˆžโˆ’โˆž type of points. So actually it is not valid to take the difference for functions mapping to โ„ฬ„. This is why we use the following method instead:

For each ๐‘›โˆˆโ„•, we define:

๐‘”๐‘›(๐‘ฅ)โ‰”min{max{๐‘”(๐‘ฅ),โˆ’๐‘›},๐‘›}andโ„Ž๐‘›(๐‘ฅ)โ‰”min{max{โ„Ž(๐‘ฅ),โˆ’๐‘›},๐‘›}

Notice that, each ๐‘”๐‘›,โ„Ž๐‘› is measurable, since ๐‘”,โ„Ž are measurable and constant function is measurable and we have proved in lecture that taking the max, min of two measurable functions is measurable.

Claim 1.1:

๐‘”(๐‘ฅ)=โ„Ž(๐‘ฅ)โˆˆโ„โ‡”โˆƒ๐‘0>0,โˆ€๐‘›โ‰ฅ๐‘0,๐‘”๐‘›(๐‘ฅ)=โ„Ž๐‘›(๐‘ฅ)

proof of claim 1.1: Suppose ๐‘”(๐‘ฅ)=โ„Ž(๐‘ฅ)โˆˆโ„. Let ๐‘€โ‰”max{|๐‘”(๐‘ฅ)|,|โ„Ž(๐‘ฅ)|}<โˆž, then for any ๐‘›>๐‘€, we have ๐‘”๐‘›(๐‘ฅ)=๐‘”(๐‘ฅ),โ„Ž๐‘›(๐‘ฅ)=โ„Ž(๐‘ฅ), so ๐‘”๐‘›(๐‘ฅ)=โ„Ž๐‘›(๐‘ฅ).

Suppose โˆƒ๐‘0>0,โˆ€๐‘›โ‰ฅ๐‘0,๐‘”๐‘›(๐‘ฅ)=โ„Ž๐‘›(๐‘ฅ), Then it is clear that

๐‘”(๐‘ฅ)=๐‘”๐‘0(๐‘ฅ)=โ„Ž๐‘0(๐‘ฅ)=โ„Ž(๐‘ฅ)<โˆž
Figureย 14:

proof of remaining: Therefore we have:

๐ธ=โ‹ƒ๐‘=1โˆžโ‹‚๐‘›โ‰ฅ๐‘{๐‘ฅโˆˆ๐‘‹โˆฃ๐‘”๐‘›(๐‘ฅ)=โ„Ž๐‘›(๐‘ฅ)}

Foe each ๐‘›โˆˆโ„•, we define

๐ธ๐‘›โ‰”{๐‘ฅโˆˆ๐‘‹โˆฃ๐‘”๐‘›(๐‘ฅ)=โ„Ž๐‘›(๐‘ฅ)}

Since each ๐‘”๐‘›,โ„Ž๐‘› is measurable and real-valued (finite), ๐‘”๐‘›โˆ’โ„Ž๐‘› is measurable and |๐‘”๐‘›โˆ’โ„Ž๐‘›| is measurable, so we have for each ๐‘šโˆˆโ„•,

{๐‘ฅโˆˆ๐‘‹:|๐‘”๐œ…๐‘›(๐‘ฅ)โˆ’โ„Ž๐œ…๐‘›(๐‘ฅ)|<1/๐‘š}=|๐‘”๐‘›โˆ’โ„Ž๐‘›|โˆ’1([0,1/๐‘š))โˆˆ๐’œ๏ธ€

Thus

๐ธ๐‘›=โ‹‚๐‘šโˆˆโ„•|๐‘”๐‘›โˆ’โ„Ž๐‘›|โˆ’1([0,1/๐‘š))โˆˆ๐’œ๏ธ€

is a measurable set. Thus ๐ธ is a countable union of countable intersections of mea surable sets, then measurable.

โ–ก

Proof

of method (ii):
Recall: a seq of real numbers converges iff it is a Cauchy. Now we fix an arbitrary ๐‘–โˆˆโ„• and let ๐œ–=1/๐‘–. Define:

๐ธ๐‘–,๐‘—,๐‘˜={๐‘ฅโˆˆ๐‘‹:|๐‘“๐‘—(๐‘ฅ)โˆ’๐‘“๐‘˜(๐‘ฅ)|<1/๐‘–}

Since each ๐‘“๐‘— is measurable, the function ๐‘ฅโ†ฆ|๐‘“๐‘—(๐‘ฅ)โˆ’๐‘“๐‘˜(๐‘ฅ)| is measurable (since each term in the sequence maps to โ„ but not โ„ฬ„), and hence each ๐ธ๐‘–,๐‘—,๐‘˜=|๐‘“๐‘—(๐‘ฅ)โˆ’๐‘“๐‘˜(๐‘ฅ)|โˆ’1([0,1/๐‘–)) is measurable.

For each ๐‘–, consider the set of ๐‘ฅโˆˆ๐‘‹ for which the sequence (๐‘“๐‘›(๐‘ฅ)) satisfies the Cauchy condition with respect to ๐œ–=1/๐‘–. That is,

๐ธ๐‘–={๐‘ฅโˆˆ๐‘‹:โˆƒ๐‘โˆˆโ„• s.t. โˆ€๐‘—,๐‘˜โ‰ฅ๐‘,|๐‘“๐‘—(๐‘ฅ)โˆ’๐‘“๐‘˜(๐‘ฅ)|<1๐‘–}

We can write ๐ธ๐‘– as

๐ธ๐‘–=โ‹ƒ๐‘=1โˆžโ‹‚๐‘—,๐‘˜โ‰ฅ๐‘๐ธ๐‘–,๐‘—,๐‘˜

Since countable unions and intersections of measurable sets are measurable, ๐ธ๐‘– is measurable.

Now, since (๐‘“๐‘›(๐‘ฅ)) converges in โ„ iff it is Cauchy, i.e. it is in ๐ธ๐‘– for each ๐‘–โˆˆโ„•, we have:

๐ธ=โ‹‚๐‘–=1โˆž๐ธ๐‘–=โ‹‚๐‘–=1โˆž(โ‹ƒ๐‘=1โˆžโ‹‚๐‘—,๐‘˜โ‰ฅ๐‘๐ธ๐‘–,๐‘—,๐‘˜)

This is a countable intersection of measurable sets, and therefore ๐ธ is measurable.

โ–ก

Measurability of continuity loci.

Let (๐‘‹,๐‘‘) be a metric space, and ๐‘“:๐‘‹โ†’โ„‚ any function. Prove that the set of points ๐‘ฅโˆˆ๐‘‹ such that ๐‘“ is continuous at ๐‘ฅ is a ๐บ๐›ฟ-set, and in particular a Borel set. Hint: consider sets of the form

{๐‘ฅโˆˆ๐‘‹โˆฃ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)|โ‰ค1๐‘›whenever max{๐‘‘(๐‘ฆ,๐‘ฅ),๐‘‘(๐‘ง,๐‘ฅ)}โ‰ค๐›ฟ}

and show off your skills with quantifiers.

Proof

Recall: ๐‘“:๐‘‹โ†’โ„‚ from a metric space is continuous at ๐‘ฅโˆˆ๐‘‹ iff for every ๐œ€>0 there exists a ๐›ฟ>0 such that|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|<๐œ€ whenever ๐‘‘(๐‘ฆ,๐‘ฅ)<๐›ฟ. We can easily check that, this condition is equivalent to: for every ๐œ€>0 there exists a ๐›ฟ>0 such that |๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)|<๐œ€โˆ€๐‘ฆ,๐‘งโˆˆ๐ต๐›ฟ(๐‘ฅ), by the relation of diameter and radius of the open ball).

Thus we have:

๐‘ฅโˆˆ๐ถโ‡”โˆ€๐‘›โˆˆโ„•,โˆƒ๐‘šโˆˆโ„• s.t.๐‘ฆ,๐‘ง with ๐‘‘(๐‘ฆ,๐‘ฅ)<1๐‘š and ๐‘‘(๐‘ง,๐‘ฅ)<1๐‘š,|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)|<1๐‘›

In other words, ๐‘ฅ is a continuity point iff it belongs to:

๐ถ=โ‹‚๐‘›=1โˆžโ‹ƒ๐‘š=1โˆž๐‘ˆ๐‘›,๐‘š.

where

๐‘ˆ๐‘›,๐‘š={๐‘ฅโˆˆ๐‘‹โˆฃ๐‘ฆ,๐‘งโˆˆ๐ต1๐‘š(๐‘ฅ)โŸน|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)|<1๐‘›}

Claim: ๐‘ˆ๐‘›,๐‘š is open.
Proof of Claim:
Let ๐‘ฅโˆˆ๐‘ˆ๐‘›,๐‘š. WTS: โˆƒ an ๐œ€>0 such that ๐ต๐œ€(๐‘ฅ)โŠ‚๐‘ˆ๐‘›,๐‘š.
Consider: ๐œ€=12๐‘š.
Let ๐‘ฆโˆˆ๐ต๐œ€(๐‘ฅ). Take any two points ๐‘ง,๐‘คโˆˆ๐‘‹ satisfying

๐‘‘(๐‘ง,๐‘ฆ)<12๐‘šand๐‘‘(๐‘ค,๐‘ฆ)<12๐‘š

Then by the triangle inequality, we have:

๐‘‘(๐‘ง,๐‘ฅ)โ‰ค๐‘‘(๐‘ง,๐‘ฆ)+๐‘‘(๐‘ฆ,๐‘ฅ)<12๐‘š+12๐‘š=1๐‘š

Similarly, ๐‘‘(๐‘ค,๐‘ฅ)<1๐‘š. Since ๐‘ฅโˆˆ๐‘ˆ๐‘›,๐‘š, it follows that

|๐‘“(๐‘ง)โˆ’๐‘“(๐‘ค)|<1๐‘›

Thus, the condition defining ๐‘ˆ๐‘›,๐‘š holds for ๐‘ฆ, meaning ๐‘ฆโˆˆ๐‘ˆ๐‘›,๐‘š. This proves that ๐ต๐œ€(๐‘ฅ)โŠ‚๐‘ˆ๐‘›,๐‘š, thus ๐‘ˆ๐‘›,๐‘š is open since ๐‘ฅ is arbitrary.

Therefore:

๐ถ=โ‹‚๐‘›=1โˆžโ‹ƒ๐‘š=1โˆž๐‘ˆ๐‘›,๐‘š

is ๐บ๐›ฟ since each โ‹ƒ๐‘š=1โˆž๐‘ˆ๐‘›,๐‘š is a union of open sets, thus open; and ๐ถ is thus a countable intersection of open sets, namely a ๐บ๐›ฟ-set. (thus Borel).

โ–ก

Measurability of differentiability loci.

Let ๐‘“:โ„โ†’โ„ be any function. Let us say (as usual) that ๐‘“ is differentiable at ๐‘ฅ if there exists ๐œ†โˆˆโ„ such that lim๐‘ฆโ†’๐‘ฅ๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)๐‘ฆโˆ’๐‘ฅ=๐œ†.

We also declare ๐‘“ to be strongly differentiable at ๐‘ฅ if there exists ๐œ†โˆˆโ„ with the following property: for each ๐œ–>0 there exists ๐›ฟ>0 such that if |๐‘ฆโˆ’๐‘ฅ|โ‰ค๐›ฟ and |๐‘งโˆ’๐‘ฅ|โ‰ค๐›ฟ, then |๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)โˆ’๐œ†(๐‘ฆโˆ’๐‘ง)|โ‰ค๐œ–|๐‘ฆโˆ’๐‘ง|.

  • Does ๐‘“ being differentiable at ๐‘ฅ imply that ๐‘“ is strongly differentiable at ๐‘ฅ? Give a proof or a counterexample.

  • Prove that the set of points ๐‘ฅโˆˆโ„ at which ๐‘“ is strongly differentiable is a Borel set. Hint: consider sets of the form

    ๐ธ๐œ†,๐‘š,๐‘›โ‰”{๐‘ฅโˆˆโ„โˆฃ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)โˆ’๐œ†(๐‘ฆโˆ’๐‘ง)|โ‰ค1๐‘›|๐‘ฆโˆ’๐‘ง|whenever max{|๐‘ฆโˆ’๐‘ฅ|,|๐‘งโˆ’๐‘ฅ|โ‰ค1๐‘š}.
  • Extra credit: is the set of points ๐‘ฅโˆˆโ„ at which ๐‘“ is differentiable a Borel set?

Solution

of (a): No. Consider the following counterexample:

๐‘“(๐‘ฅ)={๐‘ฅ2sin(1๐‘ฅ),๐‘ฅโ‰ 00,๐‘ฅ=0

We know that

๐‘“(๐‘ฅ)โˆ’๐‘“(0)๐‘ฅโˆ’0=๐‘ฅ2sin(1/๐‘ฅ)๐‘ฅ=๐‘ฅsin(1/๐‘ฅ)

Note |๐‘ฅsin(1/๐‘ฅ)|โ‰ค|๐‘ฅ|, so when ๐‘ฅโ†’0 we have:

lim๐‘ฅโ†’0๐‘ฅsin(1/๐‘ฅ)=0

Thus ๐‘“ is differentiable at 0 and ๐‘“โ€ฒ(0)=0.

Lemma 4.15

๐‘“:โ„โ†’โ„ is strongly differentiable at ๐‘ฅ โŸน it is differentiable at ๐‘ฅ, and ๐œ† is uniquely equal to the derivative at ๐‘ฅ.

Proof

of lemma 4.1:
Suppose ๐‘“:โ„โ†’โ„ is strongly differentiable at ๐‘ฅ, so for any ๐œ–>0, there exists ๐›ฟ>0 s.t. for all ๐‘ฆ,๐‘งโˆˆ๐ต๐›ฟ(๐‘ฅ), we have:

|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)โˆ’๐œ†(๐‘ฆโˆ’๐‘ง)|โ‰ค๐œ–|๐‘ฆโˆ’๐‘ง|.

Suppose ๐‘ฆโ‰ ๐‘ง, then dividing by |๐‘ฆโˆ’๐‘ง| on both sides, we have

|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)๐‘ฆโˆ’๐‘ฅโˆ’๐œ†|โ‰ค๐œ–

Since ๐œ– is arbitrary, this proves that

๐‘“โ€ฒ(๐‘ฅ)=lim๐‘ฆโ†’๐‘ฅ๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)๐‘ฆโˆ’๐‘ฅ=๐œ†

โ–ก

Now we go back to the counterexample. Suppose for contradiction that ๐‘“ is strongly differentiable at 0, then ๐œ†=0, so for all ๐œ–>0, there exist ๐›ฟ>0 s.t. for all ๐‘ฆ,๐‘งโˆˆ๐ต๐›ฟ(0), we have

|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)|โ‰ค๐œ–|๐‘ฆโˆ’๐‘ง|

Consider ๐œ–=14. Let ๐›ฟ>0. Take ๐‘›โˆˆโ„• s.t.

1(2๐‘›+32)๐œ‹<๐›ฟ

and then take

๐‘ฆ๐‘›โ‰”1(2๐‘›+12)๐œ‹,๐‘ง๐‘›โ‰”1(2๐‘›+32)๐œ‹

Note that each |๐‘ฆ๐‘›|,|๐‘ง๐‘›|<๐›ฟ. And we have

sin[(2๐‘›+12)๐œ‹]=(โˆ’1)๐‘›,sin[(2๐‘›+32)๐œ‹]=โˆ’(โˆ’1)๐‘›

Thus

๐‘“(๐‘ฆ๐‘›)โˆ’๐‘“(๐‘ง๐‘›)=(โˆ’1)๐‘›[๐‘ฆ๐‘›2+๐‘ง๐‘›2]

while

๐‘ฆ๐‘›โˆ’๐‘ง๐‘›=1(2๐‘›+12)๐œ‹โˆ’1(2๐‘›+32)๐œ‹=1๐œ‹(2๐‘›+12)(2๐‘›+32)

Taking limit of this behavior (increasing ๐‘›), we get the sequential limit of |๐‘“(๐‘ฆ๐‘›)โˆ’๐‘“(๐‘ง๐‘›)||๐‘ฆ๐‘›โˆ’๐‘ง๐‘›| indexing over ๐‘› is 12๐œ‹2๐‘›214๐œ‹๐‘›2=2๐œ‹. By taking large enough ๐‘›, we can alwasy get |๐‘“(๐‘ฆ๐‘›)โˆ’๐‘“(๐‘ง๐‘›)||๐‘ฆ๐‘›โˆ’๐‘ง๐‘›| to be arbitrarily close to 2๐œ‹>14. This shows that ๐‘“ is not strongly differentiable at 0.

Proof

of (b):
Let ๐‘“:โ„โ†’โ„ be any a function.Denote

๐ธ:={๐‘ฅโˆˆโ„โˆฃ๐‘“ is strongly differentiable at ๐‘ฅ}

WTS: ๐ธ is a Borel set.

Set for each ๐œ†โˆˆโ„,๐‘š,๐‘›โˆˆโ„•:

๐ธ๐œ†,๐‘š,๐‘›โ‰”{๐‘ฅโˆˆโ„โˆฃ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)โˆ’๐œ†(๐‘ฆโˆ’๐‘ง)|โ‰ค1๐‘›|๐‘ฆโˆ’๐‘ง|โˆ€๐‘ฆ,๐‘งโˆˆ๐ต1๐‘š(๐‘ฅ)}

where ๐ต1๐‘š(๐‘ฅ) denote the open ball centered at ๐‘ฅ with radius 1๐‘š.

Then by the definition of strongly differentiable, we have:

๐ธ=โ‹ƒ๐œ†โˆˆโ„โ‹‚๐‘›โˆˆโ„•โ‹ƒ๐‘šโˆˆโ„•๐ธ๐œ†,๐‘š,๐‘›

Claim 3.1: Each ๐ธ๐œ†,๐‘š,๐‘› is open.
Proof of Claim 3.1: Let ๐‘ฅโˆˆ๐ธ๐œ†,๐‘š,๐‘›. Then

โˆ€๐‘ฆ,๐‘งโˆˆ๐ต1/๐‘š(๐‘ฅ),|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)โˆ’๐œ†(๐‘ฆโˆ’๐‘ง)|โ‰ค1๐‘›|๐‘ฆโˆ’๐‘ง|

In particular, the inequality holds for all ๐‘ฆ,๐‘งโˆˆ๐ต1/(2๐‘š)(๐‘ฅ). Now consider ๐ต1/(2๐‘š)(๐‘ฅ), let ๐‘ฅโ€ฒโˆˆ๐ต1/(2๐‘š)(๐‘ฅ), then for every ๐‘ฆโˆˆ๐ต1/(2๐‘š)(๐‘ฅโ€ฒ), we have

|๐‘ฆโˆ’๐‘ฅ|โ‰ค|๐‘ฆโˆ’๐‘ฅโ€ฒ|+|๐‘ฅโ€ฒโˆ’๐‘ฅ|<12๐‘š+12๐‘š=1๐‘š

so ๐ต1/(2๐‘š)(๐‘ฅโ€ฒ)โŠ‚๐ต1/๐‘š(๐‘ฅ). Hence the inequality holds for all ๐‘ฆ,๐‘งโˆˆ๐ต1/(2๐‘š)(๐‘ฅโ€ฒ). This confirms that every ๐‘ฅโˆˆ๐ธ๐œ†,๐‘š,๐‘› has a neighborhood contained in ๐ธ๐œ†,๐‘š,๐‘›, proving that ๐ธ๐œ†,๐‘š,๐‘› is open.

Now that each ๐ธ๐œ†,๐‘š,๐‘› is open, we have โ‹ƒ๐‘šโˆˆโ„•๐ธ๐œ†,๐‘š,๐‘› is each for each ๐œ†,๐‘›; thus each for each ๐œ†, ๐บ๐œ†โ‰”โ‹‚๐‘›โˆˆโ„•โ‹ƒ๐‘šโˆˆโ„•๐ธ๐œ†,๐‘š,๐‘› is a ๐บ๐›ฟ set.

๐ธ=โ‹ƒ๐œ†โˆˆโ„๐บ๐œ†

is a union of ๐บ๐›ฟ sets.

(I do not now how to deal with it then, it might be that we somehow reduce it to countable union of ๐บ๐›ฟ sets, getting something like ๐ธ=โ‹ƒ๐œ†โˆˆโ„š๐บ๐œ† using the density of โ„š in โ„, thus confirming that it is Borel.) -2. ่ฟ™้‡Œ็š„ๆญฃ่งฃๆ˜ฏ: ่ฆๅˆฉ็”จ density of โ„š in โ„ ็š„่ฏ, ๅช้œ€่ฆ่€ƒ่™‘ไบคๆข set operation ็š„้กบๅบๅฐฑๅฅฝไบ†. ๆˆ‘ไปฌไผšๅ‘็Žฐๅ…ถๅฎž:

๐ธ=โ‹‚๐‘›โˆˆโ„•โ‹ƒ๐œ†โˆˆโ„šโ‹ƒ๐‘šโˆˆโ„•๐ธ๐œ†,๐‘š,๐‘›

ๅฐฑ่ฟ™ไนˆ็ฎ€ๅ•ใ€‚ใ€‚

โ–ก

Proof

of extra credit: yes. ่ฟ™ไธช่งฃๆณ•้žๅธธ้บป็ƒฆ. ้œ€่ฆๅ†ๅคš่€ƒ่™‘ไธคๅฑ‚. ไปค ๐ธ๐œ†,๐‘˜,๐‘™,๐‘š,๐‘› ่กจ็คบ the set of points ๐‘ฅ s.t.

|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ง)โˆ’๐œ†(๐‘ฆโˆ’๐‘ง)|โ‰ค1๐‘›|๐‘ฆโˆ’๐‘ง|

whenever

12๐‘™+1(1+12๐‘˜)โ‰ค|๐‘ฆโˆ’๐‘ฅ|โ‰ค12๐‘™โˆ’1(1โˆ’12๐‘˜)and|๐‘งโˆ’๐‘ฅ|โ‰ค12๐‘š(1โˆ’12๐‘˜)

Claim:

๐‘“ is differentiable at x iff ๐‘ฅโˆˆ๐ธโ‰”โ‹‚๐‘›โˆˆโ„•โ‹ƒ๐œ†โˆˆโ„šโ‹ƒ๐‘™โˆˆโ„•โ‹‚๐‘Ÿโ‰ฅ๐‘™โ‹ƒ๐‘šโ‰ฅ1โ‹ƒ๐‘˜โˆˆโ„•๐ธ๐œ†,๐‘˜,๐‘Ÿ,๐‘š,๐‘›

โ–ก

decreasing MCT: ๆˆ็ซ‹ๅฝ“ไธ”ไป…ๅฝ“ integral ็š„ limit ๆ˜ฏ finite ็š„

Let (๐‘“๐‘›)1โˆž be a decreasing sequence of non-negative measurable functions on a measure space.

  • Prove that if lim๐‘›โˆซ๐‘“๐‘›<โˆž, then lim๐‘›โˆซ๐‘“๐‘›=โˆซlim๐‘›๐‘“๐‘›.

  • Give an example of a decreasing sequence (๐‘“๐‘›)๐‘› of nonnegative measurable functions such that lim๐‘›โˆซ๐‘“๐‘›โ‰ โˆซlim๐‘›๐‘“๐‘›.

Hint: use MCT correctly.

Proof

of (a):
Since (๐‘“๐‘›) is a decreasing sequence, i.e. for every ๐‘ฅโˆˆ๐‘‹ we have

๐‘“1(๐‘ฅ)โ‰ฅ๐‘“2(๐‘ฅ)โ‰ฅ๐‘“3(๐‘ฅ)โ‰ฅโ‹ฏ

We can define the function

๐‘”๐‘›(๐‘ฅ)=๐‘“1(๐‘ฅ)โˆ’๐‘“๐‘›(๐‘ฅ)

for each ๐‘›โˆˆโ„•. Then for the seq (๐‘”๐‘›(๐‘ฅ)) we have:

  • non-negatice: ๐‘”๐‘›(๐‘ฅ)โ‰ฅ0โˆ€๐‘ฅ because ๐‘“1(๐‘ฅ)โ‰ฅ๐‘“๐‘›(๐‘ฅ).

  • increasing in ๐‘›:

    ๐‘”๐‘›(๐‘ฅ)=๐‘“1(๐‘ฅ)โˆ’๐‘“๐‘›(๐‘ฅ)โ‰ค๐‘“1(๐‘ฅ)โˆ’๐‘“๐‘š(๐‘ฅ)=๐‘”๐‘š(๐‘ฅ)โˆ€๐‘šโ‰ฅ๐‘›,โˆ€๐‘ฅ

    since (๐‘“๐‘›) is decreasing.

Define ๐‘“(๐‘ฅ)โ‰”lim๐‘›๐‘“๐‘›(๐‘ฅ)โˆˆโ„ฬ„ for each ๐‘ฅโˆˆ๐‘‹.

Since ๐‘“๐‘›(๐‘ฅ) decreases to ๐‘“(๐‘ฅ)โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ), we have

lim๐‘›โ†’โˆž๐‘”๐‘›(๐‘ฅ)=๐‘“1(๐‘ฅ)โˆ’lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“1(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)

Now we apply MCT to the increasing sequence (๐‘”๐‘›). We have:

lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐œ‡=โˆซ(lim๐‘›โ†’โˆž๐‘”๐‘›)๐‘‘๐œ‡=โˆซ(๐‘“1โˆ’๐‘“)๐‘‘๐œ‡

And since lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡<โˆž, we have

lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐œ‡=โˆซ๐‘“1๐‘‘๐œ‡โˆ’โˆซ๐‘“๐‘‘๐œ‡

Also, because of lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡<โˆž, โˆซ๐‘“๐‘› is eventually finite. Say, it is finite after ๐‘›โ‰ฅ๐‘โˆˆโ„•. We only need to consider ๐‘›โ‰ฅ๐‘ when considering the limit behavior.
Then for each ๐‘›โ‰ฅ๐‘,

โˆซ๐‘”๐‘›๐‘‘๐œ‡=โˆซ(๐‘“1โˆ’๐‘“๐‘›)๐‘‘๐œ‡=โˆซ๐‘“1๐‘‘๐œ‡โˆ’โˆซ๐‘“๐‘›๐‘‘๐œ‡

-2. ่ฟ™้‡Œๆณจๆ„, ๆˆ‘ไปฌๆ—ข็„ถ็Ÿฅ้“ ๐‘“1 ็š„ integral ๆœชๅฟ… finite, ๅฐฑไธ่ƒฝ่ฟ™ไนˆๅฎšไน‰ ๐‘”๐‘›. ๆญฃ่งฃๆ˜ฏๅ– ๐‘ s.t. โˆซ๐‘“๐‘ finite, ็„ถๅŽๅฎšไน‰ ๐‘”๐‘›โ‰”๐‘“๐‘โˆ’๐‘“๐‘›. Taking the limit as ๐‘›โ†’โˆž, have

lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐œ‡=lim๐‘›โ†’โˆž(โˆซ๐‘“1๐‘‘๐œ‡โˆ’โˆซ๐‘“๐‘›๐‘‘๐œ‡)=โˆซ๐‘“1๐‘‘๐œ‡โˆ’lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡

by linearity of numerical sequence.
Thus, combining with the result from MCT we have:

โˆซ๐‘“1๐‘‘๐œ‡โˆ’lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡=โˆซ๐‘“1๐‘‘๐œ‡โˆ’โˆซ๐‘“๐‘‘๐œ‡

Rearrange to get:

lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡=โˆซ๐‘“๐‘‘๐œ‡,

which is exactly what we wanted to prove.

โ–ก

Solution

of (b):
Consider defining (๐‘“๐‘›:โ„โ†’โ„)๐‘›โˆˆโ„• with

๐‘“๐‘›(๐‘ฅ)=๐œ’[๐‘›,โˆž)(๐‘ฅ)

Note that:

  • ๐‘“๐‘› is a decreasing seq: For each ๐‘› and every ๐‘ฅโˆˆโ„,

    ๐‘“๐‘›+1(๐‘ฅ)=๐œ’[๐‘›+1,โˆž)(๐‘ฅ)โ‰ค๐œ’[๐‘›,โˆž)(๐‘ฅ)=๐‘“๐‘›(๐‘ฅ)

    since [๐‘›+1,โˆž)โŠ‚[๐‘›,โˆž).

  • (๐‘“๐‘›) the pointwise limit:

    lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=0โˆ€๐‘ฅโˆˆโ„

    since for each ๐‘ฅ there exists an ๐‘ (any integer greater than ๐‘ฅ) such that for all ๐‘›โ‰ฅ๐‘, ๐‘ฅ<๐‘› and hence ๐‘“๐‘›(๐‘ฅ)=0.

  • For each ๐‘›,

    โˆซโ„๐‘“๐‘›๐‘‘๐œ†=โˆซ๐‘›โˆž1๐‘‘๐‘ฅ=โˆž

    But on the other hand

    โˆซโ„(lim๐‘›โ†’โˆž๐‘“๐‘›)๐‘‘๐œ†=โˆซโ„0๐‘‘๐œ†=0

Then we have the decreasing seq of function with

lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ†=โˆžwhileโˆซ(lim๐‘›โ†’โˆž๐‘“๐‘›)๐‘‘๐œ†=0

This shows that in the absence of the finiteness assumption, the limit and integration need not commute.

Vitali meet Cantor.

Construct a function ๐‘“:[0,1]โ†’[0,1] such that:

  • ๐‘“ fails to be Lebesgue measurable;

  • there exists a compact subset ๐พโŠ‚(0,1) of positive Lebesgue measure such that ๐‘“ is differentiable at every point ๐‘ฅโˆˆ๐พ.

Hint: use the function ๐‘”(๐‘ฅ)=inf{|๐‘ฅโˆ’๐‘ฆ|โˆฃ๐‘ฆโˆˆ๐พ}; then square this with the title of the problem.

Solution

Let ๐‘‰ be a Vitali set on [0,1], ๐ถ be the fat Cantor set on [0,1] by recursively taking away the middle open subinterval of length 14๐‘› on the ๐‘›th recursion. We consider the function:

๐‘“(๐‘ฅ)=๐œ’๐‘‰โ‹…๐‘‘(๐‘ฅ,๐ถ)2

where

๐‘‘(๐‘ฅ,๐ถ)โ‰”{inf{|๐‘ฅโˆ’๐‘ฆ|โˆฃ๐‘ฆโˆˆ๐ถ}

By Hw3, we know ๐‘‰ is not Lebesgue measurable, and ๐ถ is compact with positive Lebesgue measure 12.

And since ๐‘“โˆ’1({1})=๐‘‰, mapping a not measurable set to a measurable set, ๐œ’๐‘‰ is not measurable function.

And since the distance function ๐‘‘(๐‘ฅ,๐ถ) is a continuous function of [0,1], it is measurable, by the result proved in class that a continuous funciton on a topological space is measurable.

Lemma 4.16

The product of a measurable ๐‘“:โ„โ†’โ„>0 and a not measurable ๐‘”:โ„โ†’โ„ is not measurable.

Proof of Lemma 4.2: ๐‘“ measurable โŸน 1/๐‘“ measurable. Suppose for contradiction that ๐‘“๐‘” is measurable, then ๐‘”=1๐‘“(๐‘“๐‘”) is the product of two measurable functions, thus measurable, contradicting the fact that ๐‘” is not measurable. Thus ๐‘“๐‘” is not measurable.

Claim 5.1: ๐‘“ is not measurable. Proof of claim 5.1: Thus on the open set ๐ด=[0,1]\๐ถ, ๐‘‘(๐‘ฅ,๐ถ)2 is positive, so ๐œ’๐‘‰|๐ด๐‘‘(๐‘ฅ,๐ถ)2|๐ด is not measurable since it is a product of measurable and not measurable function by lemma 4.2. Thus ๐‘“ is not measurable, otherwise its restriction on ๐ด should also be measurable.

Claim 5.2: ๐‘“ is differentiable on ๐ถ. Proof of claim 5.2: Fix ๐‘ฅโˆˆ๐ถ, then ๐‘“(๐‘ฅ)=0. We want to show:๐‘“โ€ฒ(๐‘ฅ)=limโ„Žโ†’0๐‘“(๐‘ฅ+โ„Ž)โˆ’๐‘“(๐‘ฅ)โ„Ž=limโ„Žโ†’0๐‘“(๐‘ฅ+โ„Ž)โ„Ž exists Let โ„Ž>0. Case 1: ๐‘ฅ+โ„Žโˆ‰๐‘‰, then ๐œ’๐‘‰(๐‘ฅ+โ„Ž)=0, so we have ๐‘“(๐‘ฅ+โ„Ž)=๐œ’๐‘‰(๐‘ฅ+โ„Ž)๐‘‘(๐‘ฅ+โ„Ž,๐ถ)2=0, then ๐‘“(๐‘ฅ+โ„Ž)โ„Ž=0. Case 2: ๐‘ฅ+โ„Žโˆˆ๐‘‰, we have:

๐‘‘(๐‘ฅ+โ„Ž,๐ถ)=inf๐‘ฆโˆˆ๐ถ|(๐‘ฅ+โ„Ž)โˆ’๐‘ฆ|โ‰ค|(๐‘ฅ+โ„Ž)โˆ’๐‘ฅ|=|โ„Ž|

So

|๐‘“(๐‘ฅ+โ„Ž)โ„Ž|=๐‘‘(๐‘ฅ+โ„Ž,๐ถ)2|โ„Ž|โ‰ค|โ„Ž|2|โ„Ž|=|โ„Ž|

Therefore for all cases we have:

|๐‘“(๐‘ฅ+โ„Ž)โˆ’๐‘“(๐‘ฅ)โ„Ž|=|๐‘“(๐‘ฅ+โ„Ž)โ„Ž|โ‰ค|โ„Ž|

This confirms that

๐‘“โ€ฒ(๐‘ฅ)=limโ„Žโ†’0๐‘“(๐‘ฅ+โ„Ž)โˆ’๐‘“(๐‘ฅ)โ„Ž=0

This finishes the proof of required properties of ๐‘“.

4.3.3 harder Vitali meet Cantor (extra credit)

We change the requirement of (a) to be: "the restriction of ๐‘“ to any open interval ๐ผโŠ‚[0,1] fails to be Lebesgue measurable". Then how can we make the construction?

Solution

I donโ€™t know.
ๅฎ˜ๆ–น็ญ”ๆกˆ: ๆˆ‘ๅœจๅ‰ไธ€้—ฎ็ป™ๅ‡บ็š„

๐‘“(๐‘ฅ)=๐œ’๐‘‰โ‹…๐‘‘(๐‘ฅ,๐ถ)2

่ฟ™ไธชๅ‡ฝๆ•ฐ, ๅŒๆ ทไนŸๆ˜ฏๆปก่ถณ่ฟ™ไธ€้—ฎ็š„็ญ”ๆกˆ. (ๅฏนไบŽ ๐ถ, ไธไป…ๅฏไปฅ้€‰ๆ‹ฉ fat Cantor set, ๅฎž้™…ไธŠไปปไฝ• choice of compact nowhere dense set ้ƒฝๅฏไปฅ.)

5 integration of real and complex functions

5.1 integration of real and complex functions-I [Fol 2.3]

ๆˆ‘ไปฌ็›ฎๅ‰ๅชๅฎšไน‰ไบ† non-negative โ„ฬ„-valued measurable function ็š„็งฏๅˆ†, ่€Œๆˆ‘ไปฌๆƒณ่ฆๅฎŒๆ•ดๅœฐๅฎšไน‰: โ„ฬ„-valued measurable function ็š„็งฏๅˆ† โˆซ๐‘“โˆˆโ„ฬ„, ไปฅๅŠ โ„‚-valued measurable function ็š„็งฏๅˆ† โˆซ๐‘“โˆˆโ„‚.

recall: ๅฏนไบŽไปปๆ„ โ„ฬ„-valued ๐‘“,

๐‘“=๐‘“+โˆ’๐‘“โˆ’

ๅ› ่€Œๆˆ‘ไปฌๅธŒๆœ› define:

โˆซ๐‘“=โˆซ๐‘“+โˆ’โˆซ๐‘“โˆ’

ไฝ†ๆ˜ฏๅ…ถไธญๆœ‰ไธ€ไธช undefined ็š„้—ฎ้ข˜: ๆˆ‘ไปฌ่ฆ้ฟๅ… โˆžโˆ’โˆž ่ฟ™ไธ€็ฑป็š„้—ฎ้ข˜. ๅ› ่€Œๆˆ‘ไปฌๆ— ๆณ•ๅฏนๆ‰€ๆœ‰็š„ๅฏๆต‹ๅ‡ฝๆ•ฐ่ฟ›่กŒ็งฏๅˆ†, ่€Œๆ˜ฏๅฎšไน‰ "integrable" ็š„ๅฏๆต‹ๅ‡ฝๆ•ฐ.

Lemma 5.17
{โˆซ๐‘“+<โˆžโˆซ๐‘“โˆ’<โˆžโ‡”โˆซ|๐‘“|<โˆž
Proof

trivial.

โ–ก

ๆญฃ่ดŸ้ƒจๅˆ†้ƒฝๅฏๆŽง, ่‚ฏๅฎšๆ˜ฏๅฝ“ไธ”ไป…ๅฝ“็ปๅฏนๅ€ผๅ‡ฝๆ•ฐๅฏๆŽง.

ๆˆ‘ไปฌๆŽฅไธ‹ๆฅๅฐ†ๅฎšไน‰ๅฏ็งฏๅ‡ฝๆ•ฐ็š„็ฉบ้—ดๆ˜ฏ: ๆ‰€ๆœ‰็ปๅฏนๅ€ผ็งฏๅˆ†้žๆ— ็ฉท็š„ๅ‡ฝๆ•ฐ. (ๆ€Žไนˆๅ’Œ้ข„ๆœŸไธไธ€ๆ ทโ€ฆ่ฟ™ๆ ท็š„่ฏ่ฟ™ไธช็ฉบ้—ดๅœจ็งฏๅˆ†่ฟ็ฎ—ไธ‹็š„ๅ€ผๅŸŸๅฐฑๆ˜ฏ โ„ ่€Œไธๆ˜ฏ โ„ฬ„ ไบ†. ๆˆ‘ๆœŸๅพ…็š„ๆ˜ฏไธบไบ†้ฟๅ…ๆ— ็ฉทไน‹้—ด็›ธๅ‡็š„ undefined behavior ๅช้œ€่ฆๆญฃ่ดŸ้ƒจๅˆ†ๆœ‰ไธ€ไธช็งฏๅˆ†้žๆ— ็ฉทๅฐฑ่กŒไบ†. ไฝ†ๆ˜ฏๆˆ‘ไปฌ่ฆๆฑ‚็š„ๆ˜ฏ้ƒฝไธๆ˜ฏๆ— ็ฉท. ไธ่ฟ‡ๆ—ข็„ถ่ฟ™ไนˆๅฎšไน‰ไบ†่‚ฏๅฎšๆœ‰ๅ…ถ้“็†.)

5.1.1 ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚) and ๐ฟ1(๐‘‹,๐œ‡,โ„‚)

Definition 5.24 : real-valued integrable function

Given measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡), measurable ๐‘“:๐‘‹โ†’โ„ฬ„ ่ขซ็งฐไธบ integrable ็š„, ๅฆ‚ๆžœๅฎƒๆปก่ถณ

โˆซ|๐‘“|<โˆž

ๅนถๅฎšไน‰ๅ…ถ integral ไธบ:

โˆซ๐‘“=โˆซ๐‘“+โˆ’โˆซ๐‘“โˆ’
Definition 5.25 : complex-valued integrable function

Further, ๆˆ‘ไปฌๅฎšไน‰ measurable ๐‘“:๐‘‹โ†’โ„‚ ๆ˜ฏ integrable ็š„, ๅฆ‚ๆžœๅฎƒๅŒๆ ทๆปก่ถณ:

โˆซ|๐‘“|<โˆž

ๆณจๆ„ๅˆฐ่ฟ™ไธชๆกไปถ็ญ‰ไปทไบŽ Re๐‘“,Im๐‘“ integrable, ๅ› ไธบ

|๐‘“|โ‰ค|Re๐‘“|+|Im๐‘“|โ‰ค2|๐‘“|

ๆˆ‘ไปฌๅฎšไน‰ๅ…ถ integral ไธบ:

โˆซ๐‘“=โˆซRe๐‘“+๐‘–โˆซIm๐‘“
Proposition 5.10

ๆ‰€ๆœ‰็š„ real-valued integrable functions ๆž„ๆˆไธ€ไธช โ„-vector space, ๅนถไธ” integral ๆ˜ฏไธ€ไธช linear functional on it.

ๆ‰€ๆœ‰็š„ complex-valued integrable functions ๆž„ๆˆไธ€ไธช โ„‚-vector space, ๅนถไธ” integral ๆ˜ฏไธ€ไธช linear functional on it.

Proof

trivial.

โ–ก

ไธ‹้ขๆˆ‘ไปฌๅฏไปฅๅฎšไน‰่ฟ™ไธช vector space ๅนถๅœจไธŠ้ข่ฟ›่กŒไธ€ๅฎš็ ”็ฉถ. ๆญคๅค„ไธบไธ€ไธช temporary ็š„่ฎฐๅท:

Definition 5.26 : ๐ฟฬƒ(๐‘‹,๐œ‡,โ„) ไปฅๅŠ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚) space

็ป™ๅฎš measure space (๐‘‹,โ„ณ๏ธ€,๐œ‡) ๆˆ‘ไปฌๅฎšไน‰

๐ฟฬƒ(๐‘‹,๐œ‡,โ„)โ‰”{all (extended) real-valued integrable functions on ๐‘‹}

ไปฅๅŠ

๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚)โ‰”{all complex-valued integrable functions on ๐‘‹}
Proposition 5.11

๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚) ไธŠ, ๐‘“โ†ฆโˆซ๐‘“ ไธบไธ€ไธช linear functional.

ๅ› ไธบ็งฏๅˆ†ๆ˜ฏ linear ็š„, as we have proved.

Proposition 5.12
๐‘“โˆˆ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚)โŸน|โˆซ๐‘“|โ‰คโˆซ|๐‘“|
Proof

For real-valued case,

|โˆซ๐‘“|=|โˆซ๐‘“+โˆ’โˆซ๐‘“โˆ’|โ‰ค|โˆซ๐‘“+|+|โˆซ๐‘“โˆ’|=โˆซ๐‘“++โˆซ๐‘“โˆ’=โˆซ|๐‘“|

For complex-valued case, Set

๐›ผ=โˆซ๐‘“|โˆซ๐‘“|

ไบŽๆ˜ฏๆœ‰ ๐›ผโˆˆโ„‚ ไธ” |๐›ผ|=1. Note: ไธ€ไธช็ปๅฏนๅ€ผไธบ 1 ็š„ complex number ็š„ๅ€’ๆ•ฐๆ˜ฏๅฎƒ็š„ conjuate.
ๅ› ่€Œ:

|โˆซ๐‘“|=๐›ผฬ„โˆซ๐‘“=โˆซ๐›ผฬ„๐‘“โˆˆโ„

ไปŽ่€Œ

|โˆซ๐‘“|=โˆซ๐›ผฬ„๐‘“=โˆซRe(๐›ผฬ„๐‘“)โ‰คโˆซ|Re(๐›ผฬ„๐‘“)|โ‰คโˆซ|๐›ผฬ„๐‘“|=โˆซ|๐‘“|

โ–ก

Definition 5.27 : integral restricted to a measurable set

if ๐‘“โˆˆ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚), ๐ธโˆˆ๐’œ๏ธ€ (๐œ‡ ็š„ ๐œŽ-algebra), ๆˆ‘ไปฌ define:

โˆซ๐ธ๐‘“๐‘‘๐œ‡โ‰”โˆซ๐‘“๐œ’๐ธ๐‘‘๐œ‡
Proposition 5.13 : ๅฏ็งฏๅ‡ฝๆ•ฐๅ‡ ไนŽๅค„ๅค„็›ธ็ญ‰็š„็ญ‰ไปทๆกไปถ

if ๐‘“,๐‘”โˆˆ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚), ๅˆ™ TFAE:

  • ๐‘“=๐‘” a.e.

  • โˆซ|๐‘“โˆ’๐‘”|=0

  • โˆซ๐ธ๐‘“=โˆซ๐ธ๐‘” for all ๐ธโˆˆ๐’œ๏ธ€

Proof

(๐‘–)โ‡”(๐‘–๐‘–): by last time proposition.
(๐‘–๐‘–)โŸน(๐‘–๐‘–๐‘–): ๅ› ไธบ

|โˆซ๐ธ๐‘“โˆ’โˆซ๐ธ๐‘”|=|โˆซ(๐‘“โˆ’๐‘”)๐œ’๐ธ|โ‰คโˆซ|๐‘“โˆ’๐‘”|๐œ’๐ธโ‰คโˆซ|๐‘“โˆ’๐‘”|=0

(๐‘–๐‘–๐‘–)โŸน(๐‘–๐‘–): ไปค ๐‘ขโ‰”โ„œ(๐‘“โˆ’๐‘”), ๐‘ฃโ‰”โ„‘(๐‘“โˆ’๐‘”), ๅˆ™

โˆซ|๐‘“โˆ’๐‘”|=โˆซ๐‘ข++โˆซ๐‘ขโˆ’+๐‘–โˆซ๐‘ฃ++๐‘–โˆซ๐‘ฃโˆ’

่ฟ™ๅ››ไธช็งฏๅˆ†้ƒฝๆ˜ฏๆญฃๅ€ผ. ๅฎนๆ˜“ๅ‘็Žฐๅฆ‚ๆžœ ๐‘ข+ ๅœจไธ€ไธช positive measure set ๐ธ ไธŠ้ž 0, ้‚ฃไนˆ โˆซ๐ธ๐‘ข+>0 , ้‚ฃไนˆ โˆซ|๐‘“โˆ’๐‘”|>0. (ๅ…ถไป–ไธ‰ไธช็งฏๅˆ†ๅŒ็†.)

โ–ก

ๅนถไธ”ๆˆ‘ไปฌๅ‘็Žฐ, a.e. ็›ธ็ญ‰็š„ไธคไธชๅฏ็งฏๅ‡ฝๆ•ฐ ๐‘“,๐‘”โˆˆ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚) ๅœจไปปๆ„ๅฏๆต‹้›†ไธŠ็š„็งฏๅˆ†้ƒฝ็›ธ็ญ‰. ไบŽๆ˜ฏ่ฟ™ไธคไธชๅ‡ฝๆ•ฐๅœจ ๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚) ไธญ็š„่กจ็Žฐๆ˜ฏ็›ธ็ญ‰็š„. ๅ› ่€Œๆˆ‘ไปฌๅฏไปฅๆŠŠ a.e. ็›ธ็ญ‰็š„่ฟ™็งๅ…ณ็ณป quotient ๆމ, ็ฎ€ๅŒ–่ฟ™ไธช็ฉบ้—ด:

Definition 5.28 : ๐ฟ1(๐œ‡) space

ๆˆ‘ไปฌๅฎšไน‰ ๐ฟ1(๐‘‹,๐œ‡,โ„‚), ๆˆ–็ฎ€็งฐไธบ ๐ฟ1(๐œ‡), ไธบ:

๐ฟฬƒ(๐‘‹,๐œ‡,โ„‚)/โˆผ

ๅ…ถไธญ โˆผ ่กจ็คบไธ€ไธช equivalent class: ๐‘“โˆผ๐‘” if ๐‘“=๐‘” a.e. (็ญ‰ไปทไบŽ โˆซ|๐‘“โˆ’๐‘”|=0)

๐ฟ1(๐œ‡) ไธญ็š„ๆฏไธชๅ‡ฝๆ•ฐไน‹้—ดๅฝผๆญค่‡ณๅฐ‘้ƒฝๅœจไธ€ไธชๆญฃๆต‹ๅบฆ้›†ไธŠ็›ธไบ’ไธๅŒ. ่ฟ™ๅ‡ๅŽปไบ†ๅˆ†ๆžไธŠ่€ƒ่™‘ๅ‡ ไนŽๅค„ๅค„็›ธ็ญ‰็š„้›†ๅˆ็š„้กพ่™‘, ๅฏนไบŽๅค„ๅค„็›ธ็ญ‰็š„ๅ‡ฝๆ•ฐ, ๆˆ‘ไปฌ่ฎคไธบๅฎƒไปฌๅœจ ๐ฟ1(๐œ‡) ไธŠ็›ดๆŽฅ็›ธ็ญ‰. ๅนถไธ”, ๆˆ‘ไปฌๆœ‰:

๐‘“โ†ฆโˆซ๐‘“

ๅœจ ๐ฟ1(๐œ‡) ไธŠๆ˜ฏไธ€ไธช well-defined function.

5.1.2 DCT

Lemma 5.18

ไปค (๐‘“๐‘›) ไธบ a seq of a.e. defined measurable functions on ๐‘‹., s.t.

๐‘“(๐‘ฅ)โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)

exists a.e.
Claim: ๐‘“ is measurable.

Theorem 5.19 : dominated convergence theorem

Let (๐‘“๐‘›) be a seq of functions in ๐ฟ1(๐œ‡), s.t.

  • ๐‘“๐‘›โ†’๐‘“ a.e.

  • ๅญ˜ๅœจ ๐‘”โˆˆ๐ฟ1(๐œ‡) s.t. |๐‘“๐‘›|โ‰ค๐‘” a.e. for all ๐‘›.

Claim: ๐‘“โˆˆ๐ฟ1(๐œ‡) ๅนถไธ”

โˆซ๐‘“=lim๐‘›โˆซ๐‘“๐‘›
Proof

้ฆ–ๅ…ˆ็”ฑไบŽ ๐‘“๐‘›โ†’๐‘“ a.e., by lemma ๅฏไปฅๅพ—ๅˆฐ ๐‘“ ๆ˜ฏ measurable ็š„.
ๅนถไธ”

|๐‘“๐‘›|โ‰ค|๐‘”| a.e. โŸน|๐‘“|โ‰ค|๐‘”| a.e.

ไบŽๆ˜ฏ

โˆซ|๐‘“|โ‰คโˆซ|๐‘”|<โˆž

ๅณ ๐‘“โˆˆ๐ฟ1. (ไปŽ่€Œ |๐‘“| ่‡ณๅคšๅœจไธ€ไธช measure zero set ไธŠๆ— ็ฉท).
ๅนถไธ” ๐‘”(๐‘ฅ)ยฑ๐‘“๐‘›(๐‘ฅ)โ‰ฅ0 a.e. ่ฟ™ไธ€็‚นๅพˆ้‡่ฆ, ๅ› ไธบไปŽ่€Œๆˆ‘ไปฌๅฏไปฅๅฏน ๐‘”+๐‘“๐‘›, ๐‘”โˆ’๐‘“๐‘› ไฝฟ็”จ Fatouโ€™s Lemma:

โˆซ๐‘”+โˆซ๐‘“=โˆซ(๐‘”+๐‘“)=โˆซ(๐‘”+lim๐‘›โ†’โˆž๐‘“๐‘›)=โˆซlim๐‘›โ†’โˆž(๐‘”+๐‘“๐‘›)โ‰คby Fatoulimโ€‰inf๐‘›โˆซ(๐‘”+๐‘“๐‘›)=โˆซ๐‘”+limโ€‰inf๐‘›โˆซ๐‘“๐‘›

ไปŽ่€Œ (็”ฑไบŽ โˆซ๐‘”<โˆž)

โˆซ๐‘“โ‰คlimโ€‰inf๐‘›โˆซ๐‘“๐‘›

ไปฅๅŠ similarly get:

โˆซ๐‘”โˆ’โˆซ๐‘“โ‰คby Fatoulimโ€‰inf๐‘›โˆซ(๐‘”โˆ’๐‘“๐‘›)=โˆซ๐‘”โˆ’limโ€‰sup๐‘›โˆซ๐‘“๐‘›

ไปŽ่€Œ:

โˆซ๐‘“โ‰ฅlimโ€‰sup๐‘›โˆซ๐‘“๐‘›

(่ฟ™้‡Œๆณจๆ„, negate ไธ€ไธช numerical seq ๅŽ liminf ๅ˜ limsup. ็”ฑๆญคๅฏ่ง Fatouโ€™e Lemma ๅ…ถๅฎžๆ˜ฏๅพˆๅผบๅคง็š„, ๅช้œ€่ฆๅฏน โˆซ๐‘”+โˆซ๐‘“ ๅ’Œ โˆซ๐‘”โˆ’โˆซ๐‘“ ๅ„็”จไธ€ๆฌกๅฐฑๅฏไปฅๅพ—ๅˆฐ: )

โˆซ๐‘“=lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›

โ–ก

Example 5.11

Suppose ๐‘ข:[0,1]โ†’[0,1] is Lebesgue measurable.
่€ƒ่™‘่ฟ™ไธ€ seq of function: (๐‘ข๐‘›).
ๅฎนๆ˜“ๅ‘็Žฐ ๐‘ข๐‘›โ†’๐œ’{๐‘ข=1} p.w. ๆˆ‘ไปฌๅฏไปฅ็”จ ๐‘”=1 ไฝœไธบ bound function. ไปŽ่€Œๅพ—ๅˆฐ:

โˆซ๐‘“=lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›=โˆซ{๐‘ข=1}1=๐‘š({๐œ‡=1})
Example 5.12

compute

๐ผ=lim๐‘›โ†’โˆžโˆซ[0,1]1+๐‘›๐‘ฅ2(1+๐‘ฅ2)๐‘›

ไปค ๐‘“๐‘›(๐‘ฅ):=1+๐‘›๐‘ฅ2(1+๐‘ฅ2)๐‘›, ๆœ‰: ๐‘“๐‘›(๐‘ฅ)โ†’0 as ๐‘›โ†’โˆž for ๐‘ฅโˆˆ(0,1];
ๅนถไธ”่€ƒ่™‘ ๐‘”=1, ไฝœไธบ bound.
ๅ› ่€Œๆœ‰ ๐ผ=0

5.2 integration of real and complex functions-II [Fol 2.3]

5.2.1 corollaries of DCT

ไปฅไธ‹ไธบ DCT ็š„ corollaries:

5.2.2 Fubini for series and integral

Corollary 5.11 : Fubini for series and integral

ๅฏนไบŽ ๐ฟ1(๐œ‡) ไธญ็š„ sequence (๐‘“๐‘›), ๅฆ‚ๆžœ โˆ‘๐‘›=1โˆžโˆซ|๐‘“๐‘›|<โˆž, ๅˆ™

โˆ‘๐‘›=1โˆž๐‘“๐‘›โ†’๐‘Ž.๐‘’.๐นโˆˆ๐ฟ1(๐œ‡)

ๅนถไธ”

โˆซโˆ‘๐‘›=1โˆž๐‘“๐‘›=โˆซ๐น=โˆ‘๐‘›=1โˆžโˆซ๐‘“๐‘›
Proof

Recall Tonelli for sum and integrals: ๅฏนไบŽ {๐‘“๐‘›}๐‘›โˆˆโ„• in ๐ฟ+(๐œ‡), ๆœ‰:

โˆซโˆ‘๐‘›=1โˆž๐‘“๐‘›=โˆ‘๐‘›=1โˆžโˆซ๐‘“๐‘›

(ๅˆๆ˜ฏ็ปๅ…ธ Fubini ่กฅๅ…… Tonelli) ่ฟ™ไธชๅฎš็†ๆ˜ฏ Tonelli for sum and integrals ๅœจ ๐ฟ1 ไธŠ็š„ๆŽจๅนฟ.
ๆˆ‘ไปฌ set

๐น๐‘›:=โˆ‘๐‘–=1๐‘›๐‘“๐‘—๐บโ‰”โˆ‘๐‘›=1โˆž|๐‘“๐‘›|

By Tonelli for sum and integrals, ๆœ‰:

โˆซ๐บ=โˆซโˆ‘๐‘›=1โˆž|๐‘“๐‘›|=โˆ‘๐‘›=1โˆžโˆซ|๐‘“๐‘›|

็”ฑๆกไปถ็Ÿฅ้“, โˆซ๐บ<โˆž, ๅ› ่€Œ ๐บโˆˆ๐ฟ1(๐œ‡). ๆ‰€ไปฅ ๐บ ๅฏไปฅไฝœไธบ ๐น๐‘› ็š„ DCT bound:

โˆซ|๐น|โ‰คโˆซ๐บ=โˆ‘๐‘›=1โˆžโˆซ|๐‘“๐‘›|

ๅ› ่€Œ by DCT::

โˆซ๐น=lim๐‘›โ†’โˆžโˆ‘๐‘–=1๐‘›โˆซ๐‘“๐‘–=โˆ‘๐‘›=1โˆžโˆซ๐‘“๐‘›

โ–ก

5.2.3 a function that is measurable in one var and ctn/diffble in another

Corollary 5.12

ไปค (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space.
ๅฆ‚ๆžœ ๐‘“:๐‘‹ร—[๐‘Ž,๐‘]โ†’โ„‚ ๆปก่ถณ ๐‘“(โ‹…,๐‘ก)โˆˆ๐ฟ1(๐œ‡) for all ๐‘กโˆˆ[๐‘Ž,๐‘], ไปค

๐น(๐‘ก)โ‰”โˆซ๐‘“(๐‘ฅ,๐‘ก)๐‘‘๐œ‡(๐‘ฅ)

ๅˆ™ๆœ‰:

  1. ๅฆ‚ๆžœ ๐‘กโ†ฆ๐‘“(๐‘ฅ,๐‘ก) ๅฏนไบŽไปปๆ„ ๐‘ฅ ้ƒฝ่ฟž็ปญ, ๅนถไธ”ๅญ˜ๅœจไธ€ไธช ๐‘”โˆˆ๐ฟ1(๐œ‡) ไฝฟๅพ— |๐‘“(๐‘ก,๐‘ฅ)|โ‰ค๐‘”(๐‘ฅ) for all ๐‘ก,๐‘ฅ, ้‚ฃไนˆ ๐น ไนŸๆ˜ฏ ctn ็š„.

  2. ๅฆ‚ๆžœ ๐œ•๐‘“๐œ•๐‘ก(๐‘ฅ,๐‘ก) ๅฏนไบŽไปปๆ„ ๐‘ฅ,๐‘ก ้ƒฝๅญ˜ๅœจ, ๅนถไธ”ๅญ˜ๅœจไธ€ไธช ๐‘”โˆˆ๐ฟ1(๐œ‡) ไฝฟๅพ— |๐œ•๐‘“๐œ•๐‘ก(๐‘ฅ,๐‘ก)|โ‰ค๐‘”(๐‘ฅ) for all ๐‘ก,๐‘ฅ, ้‚ฃไนˆ ๐น ๆ˜ฏ differentiable ็š„, ๅนถไธ”

    ๐นโ€ฒ(๐‘ก)=โˆซ๐œ•๐‘“๐œ•๐‘ก(๐‘ฅ,๐‘ก)๐‘‘๐œ‡(๐‘ฅ)
Proof

่ฟ™ไธ€่ฏๆ˜Žๅนถไธๅ›ฐ้šพ.
For part(1), STS: ๐‘ก๐‘›โ†’๐‘กโŸน๐น(๐‘ก๐‘›)โ†’๐น(๐‘ก)
Apply DCT with ๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ,๐‘ก๐‘›), ๐‘“(๐‘ฅ)=๐‘“(๐‘ฅ,๐‘ก).
For part(2), Suppose ๐‘ก๐‘›โ†’๐‘ก.
Apply DCT to

โ„Ž๐‘›(๐‘ฅ)โ‰”๐‘“(๐‘ฅ,๐‘ก๐‘›)โˆ’๐‘“(๐‘ฅ,๐‘ก)๐‘ก๐‘›โˆ’๐‘ก

็”ฑๅฏๅฏผๅพ—่ฟž็ปญๅพ— ๐‘ฅโ†ฆ๐œ•๐‘“๐œ•๐‘ก(๐‘ฅ,๐‘ก) measurable.
ๅนถไธ” by MVT,

|โ„Ž๐‘›(๐‘ฅ)|โ‰คsup๐‘กโˆˆ[๐‘Ž,๐‘]|๐œ•๐‘“๐œ•๐‘ก(๐‘ฅ,๐‘ก)|โ‰ค๐‘”(๐‘ฅ)

ไปŽ่€Œๆˆ‘ไปฌไนŸ็”จ ๐‘” bound ไฝไบ† โ„Ž๐‘›(๐‘ฅ). Apply DCT:

๐นโ€ฒ(๐‘ก)=lim๐‘›โ†’โˆž๐น(๐‘ก๐‘›)โˆ’๐น(๐‘ก)๐‘ก๐‘›โˆ’๐‘ก=lim๐‘›โ†’โˆžโˆซ๐‘“(๐‘ฅ,๐‘ก๐‘›)โˆ’๐‘“(๐‘ฅ,๐‘ก)๐‘ก๐‘›โˆ’๐‘ก=lim๐‘›โ†’โˆžโˆซโ„Ž๐‘›=โˆซ๐œ•๐‘“๐œ•๐‘ก(๐‘ฅ,๐‘ก)๐‘‘๐œ‡(๐‘ฅ)

โ–ก

Example 5.13

ๆ˜ฏๅฆๆœ‰:

๐œ•๐œ•๐‘กโˆซโ„>0๐‘’โˆ’๐‘ก๐‘ฅ๐‘‘๐‘š(๐‘ฅ)=???โˆซโ„>0โˆ’๐‘ฅ๐‘’โˆ’๐‘ก๐‘ฅ๐‘‘๐‘š(๐‘ฅ)=โˆ’1๐‘ก2

Here

๐‘“(๐‘ก,๐‘ฅ)=๐‘’โˆ’๐‘ก๐‘ฅ,๐‘ก>0,๐‘ฅ>0

ๅ› ่€Œ

|๐œ•๐œ•๐‘ก๐‘“(๐‘ก,๐‘ฅ)|=๐‘ฅ๐‘’โˆ’๐‘ก๐‘ฅ,๐‘ก>0,๐‘ฅ>0

ๅฐ่ฏ•ๆ‰พๅˆฐๅฎƒ็š„ dominating ๐‘”(๐‘ฅ): ่ฟ™ไธชๅ‡ฝๆ•ฐๅœจ ๐‘กโ†’0 ๅค„็š„ไธŠๆž้™ๆ˜ฏ ๐‘”(๐‘ฅ,๐‘ก)=๐‘ฅ, ไฝ†ๆ˜ฏ่ฟ™ไธช ๐‘” ๅดไธๆ˜ฏไธ€ไธช ๐ฟ1 ๅ‡ฝๆ•ฐ (ๅœจๅŠ่ฝดไธŠ็งฏๅˆ†ไธบ โˆž). ไปŽ่€Œๅฎƒไธๅฏไปฅ่ฟ™ไนˆไบคๆข็งฏๅˆ†ๅ’Œๆฑ‚ๅฏผ้กบๅบ. ไฝ†ๆ˜ฏๅฆ‚ๆžœๆŠŠ ๐‘ก ็š„่Œƒๅ›ด้™ๅˆถๅœจ ๐‘กโ‰ฅ๐‘Žโˆˆโ„+ ่€Œไธๆ˜ฏ ๐‘ก>0, ๆˆ‘ไปฌๅฐฑๅฏไปฅไบคๆข่ฟ™ไธช็งฏๅˆ†ๅ’Œๆฑ‚ๅฏผ้กบๅบ, ๅ› ไธบๆญคๆ—ถๅฏไปฅ่ฎพๅฎš

๐‘”(๐‘ฅ,๐‘ก)=๐‘ฅ๐‘’โˆ’๐‘Ž๐‘ฅ

5.2.4 ๐ฟ1 as a Banach space

Theorem 5.20 : ๐ฟ1(๐œ‡) ไปฅ integral w.r.t. ๐œ‡ ไฝœไธบ norm ๆ˜ฏไธ€ไธช normed VS

ๅœจ ๐ฟ1(๐œ‡) ไธŠ, ๆˆ‘ไปฌ set

||๐‘“||โ‰”โˆซ|๐‘“|

ๅˆ™ (๐ฟ1(๐œ‡),||โ‹…||) ไธบไธ€ไธช normed โ„‚-vector space. ๅณ, ่ฟ™ๆ˜ฏไธ€ไธช well-defined norm.

Proof

recall norm ็š„ๅฎšไน‰, ้œ€่ฆ็ฌฆๅˆ:

  • Homogeneity:

    ||๐‘Ž๐‘“||=|๐‘Ž|โ‹…||๐‘“||
  • triangle ineq:

    ||๐‘“+๐‘”||โ‰ค||๐‘“||+||๐‘”||
  • nonnegativity:

    ||๐‘“||โ‰ฅ0,= iff ๐‘“=0โˆˆ๐ฟ1 (i.e. ๐‘“(๐‘ฅ)=0 a.e.)

ๅ‰ไธคๆกๆ˜ฏ็งฏๅˆ†็š„ linearity ็š„ไธ‹ไฝๆŽจ่ฎบ. ๅŽไธ€ๆก by def.

โ–ก

Corollary 5.13 : (๐ฟ1(๐œ‡),||โ‹…||) ๆ˜ฏไธ€ไธช Banach space

(๐ฟ1(๐œ‡),||โ‹…||) ็š„ induced metric space ๆ˜ฏ complete ็š„. ๅณ, every Cauchy seq converges.
(ไปŽ่€Œ่ฟ™ๆ˜ฏไธ€ไธช Banach space. )

Proof

ๅ–ไธ€ไธช Cauchy seq (๐‘“๐‘›) in ๐ฟ1.
่ฟ™้‡Œๆœ‰ไธ€ไธชๅ€ผๅพ— recall ็š„ proposition:

Proposition 5.14

ๅœจไธ€ไธช metric space ไธญ, ไธ€ไธช Cauchy seq converges ๅฝ“ไธ”ไป…ๅฝ“ๅฎƒๅญ˜ๅœจไธ€ไธช convergent ็š„ subsequence.

่ฏๆ˜Žๅพˆ็ฎ€ๅ•. ๅฏนไบŽไปปๆ„็š„ ๐œ–, ๅฏไปฅๅ– max(๐‘,๐‘€), ๅ…ถไธญ N ไธบไฝฟๅพ—่ฟ™ไธชๅญๅบๅˆ—ๆ‰€ๆœ‰ๅ…ƒ็ด ่ท็ฆป ๐‘ฅโˆ—<๐œ–/2 ็š„ไธ‹ๆ ‡๏ผŒM ไธบไฝฟๅพ—ไธปๅบๅˆ—ๆ‰€ๆœ‰ๅ…ƒ็ด ไธคไธคไน‹้—ด่ท็ฆป <๐œ–/2 ็š„ไธ‹ๆ ‡.
ๅ› ่€Œๆˆ‘ไปฌๅช้œ€่ฆ่ฏๆ˜Žๅญ˜ๅœจไธ€ไธช subseq (๐‘“๐‘›๐‘—) s.t. ๐‘“๐‘›๐‘—โ†’๐‘—โ†’โˆž๐‘“โˆˆ๐ฟ1 ๅณๅฏ.
ๅทฒ็Ÿฅ Cauchy, WTS: ๐‘“๐‘› ๆ”ถๆ•›ไธ”ๆž้™ๅœจ ๐ฟ1 ไธญ. ๆˆ‘ไปฌ็›ด่ง‰: ็”จ Cachy ๆกไปถๆž„้€  1/๐œ–2 argument.
ๆˆ‘ไปฌ pick ๅญไธ‹ๆ ‡ (๐‘›๐‘—)๐‘—โˆˆโ„• ไฝฟๅพ—ๅฏนไบŽๆฏไธช ๐‘— ้ƒฝๆœ‰

๐‘š,๐‘›โ‰ฅ๐‘›๐‘—โŸน||๐‘“๐‘šโˆ’๐‘“๐‘›||1โ‰ค12๐‘—

ๅนถ set

๐‘”๐‘—โ‰”๐‘“๐‘›๐‘—โˆ’๐‘“๐‘›๐‘—โˆ’1,๐‘”1=๐‘“๐‘›1

ๅˆ™ๆœ‰

โˆ‘๐‘—=1โˆžโˆซ|๐‘”๐‘—|โ‰ค1<โˆž

ไปŽ่€Œ by Fubiniโ€™s Thm for series and seqs, ๅญ˜ๅœจ:

๐‘“:=lim๐‘—โ†’โˆžโˆ‘๐‘–=1๐‘—๐‘”๐‘—=lim๐‘—โ†’โˆž๐‘“๐‘›๐‘—โˆˆ๐ฟ1โˆƒ๐‘Ž.๐‘’.

ๅŒๆ—ถๆœ‰

โˆซ|๐‘“โˆ’๐‘“๐‘›๐‘—|โ‰คโˆ‘๐‘—+1โˆžโˆซ|๐‘”๐‘—|โ‰ค12๐‘—โ†’๐‘—โ†’โˆž0

โ–ก

5.2.5 density of simple function of ๐ฟ1(๐œ‡)

Theorem 5.21 : density of simple functions in ๐ฟ1(๐œ‡)

ไปค (๐‘‹,๐’œ๏ธ€,๐œ‡) ไธบไธ€ไธช measure space, ไปค ๐‘“โˆˆ๐ฟ1(๐œ‡),
ๅฏนไบŽไปปๆ„ ๐œ–>0, ้ƒฝๅญ˜ๅœจ simple ๐œ™:๐‘‹โ†’โ„‚ in ๐ฟ1(๐œ‡), ไฝฟๅพ—

โˆซ|๐‘“โˆ’๐œ™|<๐œ–
Proof

่ฟ™ๆ˜ฏๆ˜พ็„ถ็š„, by ็งฏๅˆ†็š„ๅฎšไน‰. ๆˆ‘ไนˆ้ฆ–ๅ…ˆๆŠŠ ๐‘“ divide ไธบ

๐‘“=๐‘ข+๐‘–๐‘ฃ,๐‘ข=๐‘ข+โˆ’๐‘ขโˆ’,๐‘ฃ=๐‘ฃ+โˆ’๐‘ฃโˆ’

่€ŒๅŽๅฏน่ฟ™ๅ››ไธช้ž่ดŸๅ‡ฝๆ•ฐ ๐‘ข+,๐‘ขโˆ’,๐‘ฃ+,๐‘ฃโˆ’ๅˆ†ๅˆซไฝฟ็”จ simple function seq approximation, ๅ†ไฝฟ็”จ DCT:

โˆซlim๐œ™๐‘›=โˆซ๐‘ข+=limโˆซ๐œ™๐‘›

ๆฏ”ๆ–น่ฏด (๐œ™๐‘›) ไธบไปŽไธ‹้€ผ่ฟ‘ ๐‘ข+ ็š„ simple function seq, ้‚ฃไนˆ ๐‘ข+ ๆ˜ฏๅฎƒ็š„ dominating function, ๅŒๆ—ถไนŸๆ˜ฏๆž้™. ้‚ฃไนˆๅฏนไบŽไปปๆ„็š„ ๐œ–>0 ้ƒฝๅญ˜ๅœจไธ€ไธช ๐‘› ไฝฟๅพ—

||๐‘ข+โˆ’๐œ™๐‘›||1โ‰คโˆซ๐‘ข+โˆ’โˆซ๐œ™๐‘›<๐œ–

โ–ก

ๅฐคๅ…ถๆ˜ฏ่ฟ™ไธ€็‰นๆฎŠๆƒ…ๅ†ต:

5.2.6 density of step functions in ๐ฟ1(๐‘š)

Theorem 5.22 : LS measure space ็š„ ๐ฟ1 space ไธŠ็š„ density of step functions

่€ƒ่™‘ (โ„,โ„’๏ธ€,๐‘š๐‘ ) where ๐‘š๐‘  ไธบไธ€ไธช Lebesgue-Stieljes measure on โ„, let ๐‘“โˆˆ๐ฟ1(๐œ‡),
ๅฏนไบŽไปปๆ„ ๐œ–>0, ้ƒฝๅญ˜ๅœจ step function ๐œ™=โˆ‘๐‘—=1๐‘๐‘๐‘—๐œ’๐ผ๐‘—, ไฝฟๅพ—

โˆซ(๐‘“โˆ’๐œ™)<๐œ–

where each ๐ผ๐‘— ้ƒฝๆ˜ฏ open intervals.

Proof

ๅ’Œ general case ็›ธไผผ. ๅˆฉ็”จ the fact that ไปปๆ„ไธ€ไธช Lebesgue mble function ้ƒฝๅฏไปฅ็”จ step function ๆฅ approximate.

โ–ก

5.3 integration of real and complex functions-III [Fol 2.3, finished]

5.3.1 another dense subspace of ๐ฟ1(๐‘š๐‘ ): ๐ถ๐‘(โ„)

ไธŠไธ€่Š‚่ฏพๆˆ‘ไปฌ็Ÿฅ้“ไบ†: ๆ‰€ๆœ‰็š„ simple functions ๅœจ ๐ฟ1(๐œ‡) ไธญๆž„ๆˆไบ†ไธ€ไธช dense subspace. ๅฐคๅ…ถๆ˜ฏ็‰นๆฎŠๆƒ…ๅ†ต: ๅฏนไบŽ (โ„,โ„’๏ธ€,๐‘š๐‘ ), ๆ‰€ๆœ‰็š„ step functions ๆž„ๆˆไบ†ไธ€ไธช dense subspace of ๐ฟ1(๐‘š๐‘ ).

ไปŠๅคฉๆˆ‘ไปฌๅ…ˆไป‹็ปๅฆไธ€ไธช็‰นๆฎŠๆƒ…ๅ†ต (โ„,โ„’๏ธ€,๐‘š๐‘ ) ็š„ ๐ฟ1(๐‘š๐‘ ) ็š„ ๅฆไธ€ไธช dense subspace: ๆ‰€ๆœ‰็š„ cpt supported continuous function.

ไนŸๅฐฑๆ˜ฏ่ฏด, ไปปๆ„็š„ Lebesgue intble function ้ƒฝๅฏไปฅ็”จ ctn function with compact supp ๆฅ่ฟ‘ไผผ. ไธ€ไธชๅฏ็งฏๅ‡ฝๆ•ฐๅฏไปฅๆ˜ฏ supp ้žๅธธๆ€ชๅผ‚็š„ไปฅๅŠ้žๅธธ unctn ็š„, ไฝ†ๆ˜ฏๅดๅฏไปฅ็”จ ctn and cpt supp functions ๆฅ้€ผ่ฟ‘, in ๐ฟ1 sense. ๅฝ“็„ถ่ฟ™ๆ˜ฏไธ€็งๅผฑ้€ผ่ฟ‘. ๅ‡ฝๆ•ฐๅฏไปฅๅทฎๅผ‚ๅพˆๅคง.

Definition 5.29 : ๐ถ๐‘(๐‘‹)

ไปค ๐‘‹ be a metric space, ๆˆ‘ไปฌๅฎšไน‰:

๐ถ๐‘(๐‘‹)โ‰”{all ctn functions ๐‘“:๐‘‹โ†’โ„‚ with cpt supp}
Theorem 5.23 : ๐ถ๐‘(๐‘‹)โŠ‚๐ฟ1(๐œ‡) ๆ˜ฏไธ€ไธช dense linear subspace

๐ถ๐‘(โ„)โŠ‚๐ฟ1(๐œ‡๐‘š) ไธบไธ€ไธช dense linear subspace.

Proof

ๅฏนไบŽ ๐‘“โˆˆ๐ฟ1(๐‘š๐‘ ), let ๐œ–>0.ๆˆ‘ไปฌ้ฆ–ๅ…ˆ pick ไธ€ไธช step function ๆฅapproximate ๐‘“:

๐œ™=โˆ‘๐‘—=1๐‘›๐‘๐‘—๐œ’๐ผ๐‘—,๐‘ .๐‘ก.||๐‘“โˆ’๐œ™||1<๐œ–2

็ฉบๅ‡บๆฅ็š„ ๐œ–2, ๆˆ‘ไปฌไฝฟ็”จ ctn and cpt supp function ๐‘“๐‘—ๅฏนๆฏไธช ๐œ’๐ผ๐‘— ่ฟ›่กŒ้€ผ่ฟ‘, by:

Figureย 15:

ไปŽ่€Œ ||โˆ‘๐‘—๐‘“๐‘—โˆ’๐œ™||<๐œ–2, ๅ› ๆญค ||โˆ‘๐‘—๐‘“๐‘—โˆ’๐‘“||<๐œ–2 by tri ineq. ๅพ—่ฏ.

โ–ก

5.3.2 Riemann v.s. Lebesgue integral

ๆˆ‘ไปฌๅทฒ็ปๅฎŒๆˆไบ†ไธ€ไธชไปปๆ„็š„ measure space ไธŠ็š„ Lebesgue ็งฏๅˆ†็š„ๅฎšไน‰, ไปฅๅŠๅฏ็งฏ็ฉบ้—ด็š„ๅฎšไน‰.
Recall: Riemann integral ๆ˜ฏๅฏนไบŽ โ„๐‘›โ†’โ„ ็š„ๅ‡ฝๆ•ฐๅฎšไน‰็š„, ็ปๅ…ธๅฎšไน‰ไธบ โ„โ†’โ„ ็š„ๅ‡ฝๆ•ฐ.
็Žฐๅœจๆˆ‘ไปฌๆฏ”่พƒๅฏนไบŽ โ„โ†’โ„ ็š„ๅ‡ฝๆ•ฐ็š„ Riemann ๅ’Œ Lebesgue ็งฏๅˆ†. ๆˆ‘ไปฌๅฐ†ไผšๅพ—ๅ‡บ็ป“่ฎบ: Riemann ็งฏๅˆ†ๆ˜ฏ Lebesgue ็งฏๅˆ†็š„็‰นๆฎŠๆƒ…ๅ†ต, ๅณ, Riemann ๅฏ็งฏ็š„ๅ‡ฝๆ•ฐไธ€ๅฎšไนŸ Lebesgue ๅฏ็งฏ, ๅนถไธ”็งฏๅˆ†ๅ€ผ็›ธๅŒ. (ๅฏนไบŽ โ„๐‘›โ†’โ„ ็š„ๅ‡ฝๆ•ฐไนŸไธ€ๆ ท, ไน‹ๅŽๅฐ†ๅฑ•ๅผ€.)
Recall Riemann integral ็š„ๅฎšไน‰:

Definition 5.30

ๅฏนไบŽ ๐‘“:[๐‘Ž,๐‘]โ†’โ„ bdd, ไธ€ไธช partition ๐’ซ๏ธ€={๐‘ก๐‘—}๐‘—=0๐‘› on [๐‘Ž,๐‘] ๆปก่ถณ

๐‘Ž=๐‘ก0<๐‘ก1<โ‹ฏ<๐‘ก๐‘›=๐‘

Define:

๐‘†๐’ซ๏ธ€(๐‘“):=โˆ‘๐‘—=1๐‘›sup[๐‘ก๐‘—โˆ’1,๐‘ก๐‘—]๐‘“(๐‘ก๐‘—โˆ’๐‘ก๐‘—โˆ’1)๐‘ ๐’ซ๏ธ€(๐‘“):=โˆ‘๐‘—=1๐‘›inf[๐‘ก๐‘—โˆ’1,๐‘ก๐‘—]๐‘“(๐‘ก๐‘—โˆ’๐‘ก๐‘—โˆ’1)

Define over all possible partition on [๐‘Ž,๐‘]: lower integral and upper integral

๐ผฬ„(๐‘“):=inf๐’ซ๏ธ€ partition๐‘†๐’ซ๏ธ€(๐‘“)๐ผยฏ(๐‘“):=sup๐’ซ๏ธ€ partition๐‘ ๐’ซ๏ธ€(๐‘“)

ๆณจๆ„ๅˆฐ, ๅฏนไบŽไปปๆ„็š„ ๐‘“, ๆ€ปๆ˜ฏๆœ‰

๐ผยฏ(๐‘“)โ‰ค๐ผฬ„(๐‘“)

ๆˆ‘ไปฌ็งฐ ๐‘“ ๆ˜ฏ Riemann integrable ็š„, if

๐ผยฏ(๐‘“)=๐ผฬ„(๐‘“)โ‰”๐ผ(๐‘“)

่ฟ™ไธช ๐ผ(๐‘“) ็งฐไธบ ๐‘“ ๅœจ [๐‘Ž,๐‘] ไธŠ็š„ Riemann integral.

5.3.3 Riemann intble โŸน Lebesgue intble

Theorem 5.24 : Riemann integral ๆ˜ฏ Lebesgue integral ็š„็‰นๆฎŠๆƒ…ๅ†ต
๐‘“ Riemann integrableโŸน{๐‘“โˆˆ๐ฟ1([๐‘Ž,๐‘],โ„’๏ธ€.๐‘š)๐ผ(๐‘“)=โˆซ[๐‘Ž,๐‘]๐‘“๐‘‘๐‘š
Proof

for (a): ๅฏนไบŽ็ป™ๅฎš partition ๐’ซ๏ธ€, ๆˆ‘ไปฌ set:

๐บ๐’ซ๏ธ€:=โˆ‘๐‘—๐‘€๐‘—๐œ’[๐‘ก๐‘—โˆ’1,๐‘ก๐‘—],๐‘”๐’ซ๏ธ€:=โˆ‘๐‘—๐‘š๐‘—๐œ’[๐‘ก๐‘—โˆ’1,๐‘ก๐‘—]

ไปŽ่€Œๆœ‰:

๐‘†๐’ซ๏ธ€(๐‘“)=โˆซ๐บ๐’ซ๏ธ€๐‘‘๐‘š,๐‘ ๐’ซ๏ธ€(๐‘“)=โˆซ๐‘”๐’ซ๏ธ€๐‘‘๐‘š

ๆˆ‘ไปฌ็Ÿฅ้“, refinement ่ƒฝๅขžๅŠ  ๐‘ ๐’ซ๏ธ€, ๅ‡ๅฐ ๐‘†๐’ซ๏ธ€ ไปŽ่€ŒๅขžๅŠ ้€ผ่ฟ‘็ฒพๅบฆ, ่ฟ™ไธ€็‚นๅœจ Lebesgue integral ไธญๆ›ดๅŠ ๆ˜Žๆ˜พ:

๐’ซ๏ธ€โŠ‚๐’ซ๏ธ€โ€ฒโŸน๐‘”๐’ซ๏ธ€โ‰ค๐‘”๐’ซ๏ธ€โ€ฒโ‰ค๐‘“โ‰ค๐บ๐’ซ๏ธ€โ€ฒโ‰ค๐บ๐’ซ๏ธ€โŸน๐‘ ๐’ซ๏ธ€โ‰ค๐‘ ๐’ซ๏ธ€โ€ฒโ‰ค๐ผ(๐‘“)โ‰ค๐‘†๐’ซ๏ธ€โ€ฒโ‰ค๐‘†๐’ซ๏ธ€

็”ฑไบŽ๐‘“ Riem integrable, ๅญ˜ๅœจไธ€ไธช seq of partitions (๐’ซ๏ธ€๐‘›) ไฝฟๅพ— ๐’ซ๏ธ€๐“ƒ๏ธ€โŠ‚๐’ซ๏ธ€๐‘›+1, ||๐’ซ๏ธ€||โ†’0 (mesh), ๅนถไธ”

๐‘ ๐’ซ๏ธ€๐“ƒ๏ธ€,๐‘†๐’ซ๏ธ€๐“ƒ๏ธ€โ†’๐‘›โ†’โˆž๐ผ(๐‘“)

ๅ› ่€Œ settiing

๐‘”:=lim๐‘›โ†’โˆž๐‘”๐’ซ๏ธ€๐‘›

ไธบไธ€ไธช increasing limit;

๐บ:=lim๐‘›โ†’โˆž๐บ๐’ซ๏ธ€๐‘›

ไธบไธ€ไธช decreasing limit; ็”ฑ mble seq ็š„ limit behvior ๅพ— ๐‘”,๐บโˆˆ๐ฟ1(๐‘š) ไธ” ๐‘”โ‰ค๐‘“โ‰ค๐บ ๅนถไธ” by DCT:

โˆซ๐‘”๐‘‘๐‘š=lim๐‘›โˆซ๐‘”๐’ซ๏ธ€๐‘›=๐ผ(๐‘“)โˆซ๐บ๐‘‘๐‘š=lim๐‘›โˆซ๐บ๐’ซ๏ธ€๐‘›=๐ผ(๐‘“)

ไปŽ่€Œ

๐‘”โ‰ค๐‘“โ‰ค๐บ,and โˆซ(๐บโˆ’๐‘”)๐‘‘๐‘š=0

ๅ› ่€Œ

๐‘”=๐บ๐‘Ž.๐‘’.(โŸน=๐‘“๐‘Ž.๐‘’.)

ๅ› ่€Œ

๐ผ(๐‘“)=โˆซ๐‘“๐‘‘๐‘š

(็”ฑไบŽ ๐‘š complete, ๐‘“ ๆ˜ฏ Lebesgue mble ็š„.)

โ–ก

5.3.4 Lebesgueโ€™s criterion for Riemann integrability

Theorem 5.25 : Lebesgueโ€™s characterization of Riemann integrability

ๅฎšไน‰

๐ท๐‘“={๐‘ฅ where ๐‘“ is not ctn at}

ๅˆ™ๆœ‰

๐‘“ Riemann intble โ‡”๐‘š(๐ท๐‘“)=0
Proof

ๅœจ 395 ไธญๅทฒ็ป่ฏๆ˜Žไธ€ๆฌก. ่ฟ™้‡Œๅ†ๅ›ž้กพไธ€ๆฌก.
Backward direction: trivial.
Forward direction: assume ๐‘“ Riemann intble .
ๅฏนไบŽ ๐‘“:[๐‘Ž,๐‘]โ†’โ„, ๆˆ‘ไปฌ define:

๐ป(๐‘ฅ)โ‰”lim๐›ฟโ†’0sup|๐‘ฆโˆ’๐‘ฅ|โ‰ค๐›ฟ๐‘“(๐‘ฆ),โ„Ž(๐‘ฅ)โ‰”lim๐›ฟโ†’0inf|๐‘ฆโˆ’๐‘ฅ|โ‰ค๐›ฟ๐‘“(๐‘ฆ)

ๅณ ๐‘“ ๅœจ ๐‘ฅ ๅค„็š„ไธŠไธ‹ๆž้™. ไปŽ่€Œ:

๐‘“ ctn at ๐‘ฅโ‡”lim๐‘ฆโ†’๐‘ฅ๐‘“(๐‘ฆ)=๐‘“(๐‘ฅ)โ‡”๐ป(๐‘ฅ)=โ„Ž(๐‘ฅ)

ๅ› ่€Œ่ฆ่ฏๆ˜Ž ๐‘š(๐ท๐‘“)=0, STS: ๐ป(๐‘ฅ)=โ„Ž(๐‘ฅ) a.e.
To prove this: ่ง 395.

โ–ก

5.4 modes of convergence [Fol 2.4, finished]

5.4.1 convergence family

ๅฏนไบŽ ๐‘“๐‘›,๐‘“:๐‘‹โ†’โ„‚, ๆˆ‘ไปฌ็›ฎๅ‰ๆœ‰ 4 ็งไธๅŒ็š„ convergence.
2 general ones:

  • pointwise convergence: ๅญ—้ขๆ„ๆ€.

  • uniform convergence (on a subset): ๅฏนไบŽไปปๆ„ error bound ๐œ–, ๅญ˜ๅœจๅŒไธ€ไธชๅบๅท ๐‘ ๅฏไปฅ ๐œ–-bound ไฝ่ฟ™ไธช้›†ๅˆ้‡Œๆ‰€ๆœ‰็š„ ๐‘ฅ ็š„ๅ‡ฝๆ•ฐๅ€ผๅ’Œ limit ๅ‡ฝๆ•ฐๅ€ผ็š„ error.

2 in a measure space:

  • a.e. convergence: ptwise convergence for a.e. ๐‘ฅ, ๅณ outside a null ๐ธ.

  • convergence in ๐ฟ1: โˆซ|๐‘“๐‘›โˆ’๐‘“|โ†’0

ๆˆ‘ไปฌ recall trivial relation:

uni. convโŸนptwise. convโŸนconv. a.e.

ไฝ†ๆ˜ฏๆˆ‘ไปฌไธๆธ…ๆฅš ๐ฟ1-convergence ๅ’Œๅฎƒไปฌไน‹้—ด็š„ๅ…ณ็ณป.
ๆˆ‘ไปฌ็œ‹ไปฅไธ‹็š„ examples:

5.4.2 examples showing a.e. ptwise conv ๅ’Œ ๐ฟ1 conv ไธ่ƒฝไบ’ๆŽจ

Example 5.14

on (โ„,๐”,๐‘š), ไปฅไธ‹ (๐‘“๐‘›):

  • escape to width

    ๐‘“๐‘›=1๐‘›๐œ’(0,๐‘›)

    ๐‘“๐‘›โ†’0 uniformly ไฝ† โ†’ฬธ0 in ๐ฟ1

  • escape to hat:

    ๐‘“๐‘›=๐œ’(๐‘›,๐‘›+1)

    ๐‘“๐‘›โ†’0 ptwisely ไฝ†ๅนถไธ uniformly, ๅนถไธ” โ†’ฬธ0 in ๐ฟ1

  • escape to height:

    ๐‘“๐‘›=๐‘›๐œ’[0,1๐‘›)

    ๐‘“๐‘›โ†’0 a.e., ไฝ†ๆ˜ฏๅนถไธ ptwisely, ๅฝ“็„ถไนŸๅนถไธ uniformly, ๅนถไธ”โ†’ฬธ0 in ๐ฟ1

  • typewriter: ๆˆ‘ไปฌๆŠŠๅŒบ้—ด[0,1]ๅˆ’ๅˆ†ๆˆ2๐‘˜ไธช็ญ‰้•ฟๅญๅŒบ้—ด, ๅฏนไบŽ 1โ‰ค๐‘›โ‰ค2๐‘˜ ไปค ๐‘“๐‘˜,๐‘›(๐‘ฅ) ไบคๆ›ฟๅ– 1, ๅ…ถไป–ๅ– 0.

    ๐‘“๐‘›,๐‘˜(๐‘ฅ)={1,๐‘ฅโˆˆ[๐‘›โˆ’12๐‘˜,๐‘›2๐‘˜]0,otherwise

    ๅณ, for given ๐‘˜, ๐‘“๐‘› is the indicator function of the ๐‘›-th dyadic interval.

    โˆฅ๐‘“๐‘›,๐‘˜โˆฅ1=12๐‘˜โ†’0

    ๅ› ่€Œ ๐‘“๐‘›,๐‘˜โ†’0 in ๐ฟ1, ไฝ†ๆ˜ฏ โˆ€๐‘ฅโˆˆ[0,1], ๐‘“๐‘›,๐‘˜(๐‘ฅ)โ†’ฬธ0 ptwisely. (ไนŸไธ a.e.) (่ฟ™ไธชไพ‹ๅญ, ๅœจๆŽจๅนฟ่‡ณ ๐ฟ๐‘ ็ฉบ้—ด็š„ๆ—ถๅ€™, ไนŸๆœ‰ โˆฅ๐‘“๐‘›,๐‘˜โˆฅ๐‘โ†’0, ไนŸๅฏไปฅ่ฏดๆ˜Ž ๐ฟ๐‘ convergence ๅนถไธ่ƒฝๆŽจๅฏผ a.e. convergence, ้™คไบ† ๐ฟโˆž ็š„ไพ‹ๅค–.)

ๅœจ่ฟ™ไบ›ไพ‹ๅญไธญ, ๆˆ‘ไปฌๅ‘็Žฐ, ๐ฟ1-convergence ๅ’Œ uniform, ptwise, a.e. ่ฟ™ไธ‰ไธช modes of covergence ้ƒฝไบ’ไธๆŽจๅฏผ. ๅฏนไบŽ uniform convergence ๅ’Œ ptwise convergence, ่ฟ™ๆ˜ฏๅพˆๅˆ็†็š„, ๅ› ไธบๅฏไปฅๅ‡ฝๆ•ฐ่ถŠๆฅ่ถŠๅฎฝๅ’Œๆ‰ไฝฟๅพ—็งฏๅˆ†ไธๅ˜ไฝ†ๆ˜ฏๅด uni conv; ไนŸๅฏไปฅๅ‡ฝๆ•ฐ็งฏๅˆ†ๆ”ถๆ•›ไฝ†ๆ˜ฏๅœจไธ€ไธช้›ถๆต‹้›†ไธŠๅๅค่ทณ่ทƒ.
ๅนถไธ”ๆˆ‘ไปฌ่ฟ›ไธ€ๆญฅๅ‘็Žฐ, ๅฐฑ็ฎ—ๆ˜ฏ a.e. ๆ”ถๆ•›, ไนŸๅ’Œ ๐ฟ1 ๆ”ถๆ•›ๆฒกๆœ‰ไบ’ๆŽจๅ…ณ็ณป. ๆฏ”ๅฆ‚ ex (3), ่ฟ™ไธชๅ‡ฝๆ•ฐๅชๅœจ 0 ๅค„ไธๆ”ถๆ•›่‡ณ 0, ไฝ†ๆ˜ฏๆ•ดไฝ“็š„็งฏๅˆ†ๅดๆ˜ฏ const 1.
ๆˆ‘ไปฌ recall: ไธคไธชๅ‡ฝๆ•ฐ a.e. ็›ธ็ญ‰, ็ญ‰ไปทไบŽๅฎƒไปฌ็š„ ๐ฟ1 distance ไธบ 0. ไฝ†ๆ˜ฏๅฎƒไปฌไฝœไธบๅ‡ฝๆ•ฐๅˆ—ๆž้™่กŒไธบ, ๅนถไธ็›ธๅนฒ.
ๅ…ณไบŽ ๐ฟ1-convergence ๅ’Œ uniform, ptwise, a.e. convergence ็š„ๅ…ณ็ณปๆˆ‘ไปฌๅทฒ็ป่ฎจ่ฎบๅฎŒไบ†.
ๆŽฅไธ‹ๆฅๆˆ‘ไปฌๅฐ†ๅ…ณไบŽ ๐ฟ1-convergence ่ฟ™ไธ€ๆก็บฟ, ๅผ•ๅ…ฅไธ€ไบ›ๆ–ฐ็š„ convergence modes, ๅœจๆ›ดๅคง็š„ convergence family ไธญ่ฎจ่ฎบ่ฟ™ไบ› convergence ็š„ๅ…ณ็ณป.

5.4.3 3 new modes of convergence: fast ๐ฟ1-conv, conv measure and subseq a.e. conv

Definition 5.31 : modes of convergence for measurable functions

ๅฏนไบŽ ๐‘“๐‘›,๐‘“:๐‘‹โ†’โ„‚, ๆˆ‘ไปฌๅฎšไน‰ไปฅไธ‹ไธ‰็ง convergence:

  • fast ๐ฟ1-convergence: if

    โˆ‘๐‘›=1โˆžโˆซ|๐‘“๐‘›โˆ’๐‘“|<โˆž
  • convergence in measure: if

    ๐œ‡(๐‘ฅ:|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|>๐œ–)โ†’๐‘›โ†’โˆž0
  • subseq a.e. convergence: if ๅญ˜ๅœจไธ€ไธช subseq (๐‘“๐‘›๐‘—) ไฝฟๅพ—

    ๐‘“๐‘›๐‘—โ†’๐‘—โ†’โˆž๐‘“๐‘Ž.๐‘’.

ๆ˜พ็„ถ, fast ๐ฟ1-convergence โŸน ๐ฟ1-convergence;
ๆˆ‘ไปฌๆŽฅไธ‹ๆฅๅฐ†่ฏดๆ˜Ž, fast ๐ฟ1-convergence ไนŸ โŸน a.e. convergence (ไบŽๆ˜ฏๅฎƒๅŒๆ—ถไฝœไธบ a.e. convergence ๅ’Œ ๐ฟ1-convergence ็š„ไธŠไฝๆ”ถๆ•›, ไฝœไธบ่ฟ™ไธคๆก็บฟ่ทฏ็š„ไธŠไฝไบคๆฑ‡.)
่€Œๆˆ‘ไปฌไนŸๅฐ†่ฏดๆ˜Ž: ๐ฟ1-convergence ๅ’Œ a.e. convergence ้ƒฝ โŸน subseq a.e. convergence, ไฝœไธบ่ฟ™ไธคๆก็บฟ่ทฏ็š„ไธ‹ไฝไบคๆฑ‡.
ไปฅๅŠ, ๐ฟ1-convergence โŸน convergence in measure.

ไปฅไธ‹็š„ๆ ‡่ฎฐๅฐ†ๅœจไน‹ๅŽๅ‡ ไธชๅฎš็†็š„่ฏๆ˜Žไธญ็”จๅˆฐ: ๆˆ‘ไปฌ็Žฐๅœจ define:

๐ต๐‘›,๐‘˜โ‰”{๐‘ฅโˆˆ๐‘‹:|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|โ‰ค1๐‘˜}

่ฟ™ไธช้›†ๅˆ่กจ็คบๅฏน็ฌฌ ๐‘›th term, error ๆŽงๅˆถๅœจ 1๐‘˜ ไปฅๅ†…็š„็‚น.
ไปŽ่€Œๆˆ‘ไปฌๅฏไปฅ็”จไบคๅนถ็š„ๅฝขๅผๆฅ่กจ็คบ ptwise ๆ”ถๆ•›็‚น็š„้›†ๅˆ:

{๐‘ฅโˆฃ๐‘“๐‘›(๐‘ฅ)โ†’๐‘“(๐‘ฅ)}=โ‹‚๐‘˜=1โˆžโ‹ƒ๐‘=1โˆžโ‹‚๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜

Recall Chebyshev:

๐‘”โˆˆ๐ฟ1โŸน๐œ‡({|๐‘”|โ‰ฅ๐‘})โ‰ค1๐‘โˆซ|๐‘”|
Proposition 5.15 : fast ๐ฟ1-conv โŸน a.e. conv.
โˆ‘๐‘—=1โˆžโˆซ|๐‘“๐‘›โˆ’๐‘“|<โˆžโŸน๐‘“๐‘›โ†’๐‘“๐‘Ž.๐‘’.
Proof

ๆˆ‘ไปฌๅ–

{๐‘ฅโˆฃ๐‘“๐‘›(๐‘ฅ)โ†’๐‘“(๐‘ฅ)}=โ‹‚๐‘˜=1โˆžโ‹ƒ๐‘=1โˆžโ‹‚๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜

็š„ complement

๐ธโ‰”โ‹ƒ๐‘˜=1โˆžโ‹‚๐‘=1โˆžโ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜๐‘={๐‘“๐‘›โ†’ฬธ๐‘“}

By Cheb, for each ๐‘›,๐‘˜ we have:

๐œ‡(๐ต๐‘›,๐‘˜๐‘)โ‰ค๐‘˜โˆซ|๐‘“๐‘›โˆ’๐‘“|

ๅ› ่€Œ็”ฑ fast ๐ฟ1-convergence ็š„ๆกไปถๅฏๅพ—

โˆ€๐‘˜โˆ€๐‘,๐œ‡(โ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜๐‘)โ‰ค๐‘˜โˆ‘๐‘›=๐‘โˆžโˆซ|๐‘“๐‘›โˆ’๐‘“|(โ†’0 as ๐‘โ†’โˆž)

ๅ› ่€Œ by ctn from above,

๐œ‡(โ‹‚๐‘=1โˆžโ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜๐‘)=0

ๅ› ่€Œ

๐œ‡(๐ธ)=0

โ–ก

Corollary 5.14 : ๐ฟ1-convergence (โŸนconv. in measure) โŸน subseq a.e. conv.

if ๐‘“๐‘›โ†’๐‘“ in ๐ฟ1, then there exists subseq (๐‘“๐‘›๐‘—)๐‘—โˆˆโ„• s.t. ๐‘“๐‘›๐‘—โ†’๐‘“ a.e.
(ๅณ ๐ฟ1 convergence implies subseq a.e. convergence)

Proof

ๆณจๆ„: ๅฏนไบŽ ๐ฟ1-convergent ็š„ seq, ๆˆ‘ไปฌๅฏไปฅ pick ๅ‡บไธ€ไธช fast ๐ฟ1-convergent ็š„ subseq.
Pick (๐‘›๐‘—)๐‘—โˆˆโ„• s.t.

โˆซ|๐‘“๐‘›๐‘—โˆ’๐‘“|โ‰ค1๐‘—๐‘›

Then

โˆ‘๐‘—=1โˆžโˆซ|๐‘“๐‘›๐‘—โˆ’๐‘“|<โˆž

็”ฑๅˆšๆ‰็š„ prop ๅพ—, ๐‘“๐‘›๐‘—โ†’๐‘“ a.e.

โ–ก

5.4.4 a.u. conv.(ๅนถ้ž uni. conv. a.e.) ๅ’Œ Egoroffโ€™s Theorem

Definition 5.32

ๆˆ‘ไปฌ็งฐ ๐‘“๐‘›โ†’๐‘“ almost uniformly (a.u.), ๅฆ‚ๆžœ โˆ€๐œ€>0, ้ƒฝๅญ˜ๅœจ ๐ธโІ๐ด s.t. ๐œ‡(๐ธ)<๐œ€ ๅนถไธ” ๐‘“๐‘›โ†’๐‘“ uniformly on ๐ธ๐ถ

Theorem 5.26 : Egoroffโ€™s Theorem

ๅฆ‚ๆžœ ๐œ‡ ๆ˜ฏไธช finite measure (๐œ‡(๐‘‹)<โˆž), ้‚ฃไนˆ

๐‘“๐‘›โ†’๐‘“๐‘Ž.๐‘’.โ‡”๐‘“๐‘›โ†’๐‘“๐‘Ž.๐‘ข.
Proof

a.u. โŸน a.e.: DIY (ๆ˜พ็„ถ)
a.e. โŸน a.u.: Fix ๐œ€>0, ๆˆ‘ไปฌๆœ‰

๐‘“๐‘›โ†’๐‘“๐‘Ž.๐‘’.โ‡”๐œ‡(โ‹ƒ๐‘˜=1โˆžโ‹‚๐‘=1โˆžโ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜๐‘)=0

ๅ› ่€Œ

โˆ€๐‘˜,๐œ‡(โ‹ƒ๐‘˜=1โˆžโ‹‚๐‘=1โˆžโ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜๐‘)=0

By Ctn from Above:

โˆ€๐‘˜,lim๐‘โ†’โˆž๐œ‡(โ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜)=0

Then:

โˆ€๐‘˜,โˆƒ๐‘๐‘˜๐‘ .๐‘ก.๐œ‡(โ‹ƒ๐‘›โ‰ฅ๐‘๐ต๐‘›,๐‘˜)<๐œ€2๐‘˜

Set

๐ธโ‰”โ‹ƒ๐พ=1โˆžโ‹ƒ๐‘›โ‰ฅ๐‘๐‘˜๐ต๐‘›,๐‘˜๐‘

Then we have:

{๐œ‡(๐ธ)<โˆ‘1โˆž๐œ€2๐‘˜=๐œ€๐‘“๐‘›โ†’๐‘“unif. on ๐ธ๐‘=โ‹‚๐‘˜=1โˆžโ‹‚๐‘›โ‰ฅ๐‘๐‘˜๐ต๐‘›,๐‘˜

โ–ก

Example 5.15

๐œ‡=โˆž ๆ—ถ็š„ๅไพ‹: ่€ƒ่™‘ escape to hat function ๐‘“๐‘›โ‰”๐œ’(๐‘›,๐‘›+1) on (โ„,๐”,๐‘š).
๐‘“๐‘›โ†’0 a.e. ไฝ†ๆ˜ฏๅนถไธ a.u., ๅ› ไธบ ๐œ‡(๐‘‹)=โˆž.

Theorem 5.27 : Lusinโ€™s Theorem

If ๐‘“:[๐‘Ž,๐‘]โ†’โ„‚ ๆ˜ฏ Leb. mble ็š„, ้‚ฃไนˆ โˆ€๐œ€>0, ้ƒฝๅญ˜ๅœจ compact ๐พโІ[๐‘Ž,๐‘] s.t. ๐‘š(๐พ๐‘)<๐œ€ ๅนถไธ” ๐‘“|๐พ ctn.

Proof

่ฟ™้‡Œๆˆ‘ไปฌ restrict (โ„,๐”,๐‘š) to [๐‘Ž,๐‘], ๅพ—ๅˆฐ่ฟ™ไธช subspace ๆ˜ฏไธ€ไธช finite (=๐‘โˆ’๐‘Ž) ็š„ measure space. ๆˆ‘ไปฌ็Ÿฅ้“ ๐ถ๐‘([๐‘Ž,๐‘])โІ๐ฟ1(๐‘š) ๆ˜ฏ dense subset.
First assume ๐‘“ bounded, then ๐‘“โˆˆ๐ฟ1(๐‘š), โˆซ|๐‘“|<โˆž.
Then:

โˆƒ(๐‘“๐‘›)โІ๐ถ๐‘([๐‘Ž,๐‘])๐‘ .๐‘ก.๐‘“๐‘›โ†’๐‘“ in ๐ฟ1

Pass to subseq: (๐‘“๐‘›๐‘—)โ†’๐‘“ a.e.
Then by Egorov:

โˆƒ๐นโІ[๐‘Ž,๐‘] mble ๐‘ .๐‘ก.๐œ‡(๐น)<๐œ€2

ๅนถไธ” (๐‘“๐‘›๐‘—)โ†’๐‘“ uniformly on ๐น๐‘.
By inner regu: ๅญ˜ๅœจ ๐พโІ[๐‘Ž,๐‘] cpt s.t. ๐พโІ๐น๐‘ ๅนถไธ” ๐‘š(๐น๐‘\๐พ)<๐œ€2, ไปŽ่€Œ ๐‘š(๐พ๐‘)<๐œ€ ๅนถไธ” ๐‘“๐‘› conv unif. on ๐พ, so ๐‘“ ctn on ๐พ.

โ–ก

5.4.5 summary: convergence mode relations

Figureย 16:

ไธ€ๆก็บฟๆ˜ฏๅ‡ฝๆ•ฐๅ€ผๆ–น้ข็š„ๆ”ถๆ•›, ไธ€ๆก็บฟๆ˜ฏๆต‹ๅบฆๅ’Œ็งฏๅˆ†ๆ–น้ข็š„ๆ”ถๆ•›, ็ฌฌไธ€ๆฌกไบคๆฑ‡ๆ˜ฏ fast ๐ฟ1 conv, ๆฑ‡่šๅœจ subseq a.e. conv.
subseq a.e. conv. ๆ˜ฏๆœ€ๅผฑ็š„ convergence, ่ฟ™้‡Œๆ‰€ๆœ‰็š„ convergence ้ƒฝๅฏไปฅๆŽจๅˆฐๅฎƒ.
่ฟ™้‡Œๅฏ่ƒฝ่ฟ˜ๆœ‰ๅ…ถไป–็š„ convergence ๅ…ณ็ณป. ไฝ†ๆ˜ฏๆˆ‘ไปฌไธๅ…ณๅฟƒ. ๅ› ไธบไธๅคชไผš็”จๅˆฐๅฎƒไปฌ็š„ๅ…ณ็ณป.

Homework 5: on integration(50/50)

None of the following questions will be graded. Do them, but do not hand them in.

Dirac measure: โˆซ๐‘“๐‘‘๐›ฟ๐‘ฅ0=๐‘“(๐‘ฅ0)

Let (๐‘‹,๐’œ๏ธ€) be a measurable space, and ๐‘ฅ0โˆˆ๐‘‹ a point. Let ๐›ฟ๐‘ฅ0 be the Dirac measure at ๐‘ฅ0, i.e. for ๐ธโˆˆ๐’œ๏ธ€, ๐›ฟ๐‘ฅ0(๐ธ)=1 if ๐‘ฅ0โˆˆ๐ธ and ๐›ฟ๐‘ฅ0(๐ธ)=0 if ๐‘ฅ0โˆ‰๐ธ. Show that every measurable function ๐‘“:๐‘‹โ†’โ„ is integrable and

โˆซ๐‘“๐‘‘๐›ฟ๐‘ฅ0=๐‘“(๐‘ฅ0)

Remark: what is often called a Dirac delta function is actually this Dirac measure.

measure space ็š„ extension ไฟ็•™ measurable function ็š„ๅฏๆต‹ๆ€งๅ’Œ็งฏๅˆ†

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) and (๐‘‹,โ„ฌ๏ธ€,๐œˆ) be measure spaces on the same set ๐‘‹. Suppose that (๐‘‹,โ„ฌ๏ธ€,๐œˆ) is an extension of (๐‘‹,๐’œ๏ธ€,๐œ‡).

  • Show that if a function ๐‘“ on ๐‘‹ is ๐’œ๏ธ€-measurable, then it is โ„ฌ๏ธ€-measurable.

  • Show that if a function ๐‘“ on ๐‘‹ is ๐’œ๏ธ€-measurable and ๐‘“โˆˆ๐ฟ1(๐’œ๏ธ€,๐œ‡), then ๐‘“โˆˆ๐ฟ1(โ„ฌ๏ธ€,๐œˆ) and โˆซ๐‘“๐‘‘๐œ‡=โˆซ๐‘“๐‘‘๐œˆ.

almost everywhere defined measurable function

Carefully think through the notion of an โ€œalmost everywhere definedโ€ measurable (or integrable) function. How can we deduce the โ€œalmost everywhereโ€ versions of the main convergence theorems (MCT, FL, DCT) from their โ€œeverywhereโ€ counterparts? Propositions 2.11 and 2.12 inย [Folland] are useful here (these appeared on HW4).

new measure from old: ๐œˆ(๐ด)โ‰”โˆซ๐ด๐‘“๐‘‘๐œ‡โŸนโˆซ๐‘”๐‘‘๐œˆ=โˆซ๐‘”๐‘“๐‘‘๐œ‡

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space. Let ๐‘“:๐‘‹โ†’[0,โˆž] be an ๐’œ๏ธ€-measurable function. Define ๐œˆ:๐’œ๏ธ€โ†’[0,โˆž] by ๐œˆ(๐ด)=โˆซ๐ด๐‘“๐‘‘๐œ‡=โˆซ๐‘“๐œ’๐ด๐‘‘๐œ‡ for ๐ดโˆˆ๐’œ๏ธ€.

  • Prove that ๐œˆ is a measure on (๐‘‹,๐’œ๏ธ€).

  • Prove that โˆซ๐‘”๐‘‘๐œˆ=โˆซ๐‘”๐‘“๐‘‘๐œ‡ for every ๐’œ๏ธ€-measurable function ๐‘”:๐‘‹โ†’[0,โˆž]. Hint: Start with the case when ๐‘”=๐œ’๐ธ; then treat the case when ๐‘” is a simple function; finally consider the case when ๐‘” is a general nonnegative function.

  • Now consider the case (๐‘‹,๐’œ๏ธ€,๐œ‡)=(โ„,โ„ฌ๏ธ€(โ„),๐‘š), where ๐‘š is Lebesgue measure. Each nonnegative function ๐‘“:โ„โ†’[0,โˆž] induces a Borel measure ๐œˆ๐‘“(๐ด)=โˆซ๐ด๐‘“๐‘‘๐‘š by (a).

    • Which functions ๐‘“ induce a locally finite Borel measure? In that case, what is the distribution function for ๐œˆ๐‘“?

    • Do all locally finite Borel measures arise from some ๐‘“?

    • Can you interpret (b) as a change of variables formula?

Truncations in ๐ฟ1: ้€š่ฟ‡ โˆซ๐‘“๐‘› ๆˆ–่€… โˆซ๐‘‹๐‘›๐‘“ ็š„ๆž้™ (bounded function / subset) ๅพ—ๅˆฐ โˆซ๐‘‹๐‘“

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space and ๐‘“:๐‘‹โ†’โ„‚ an integrable function.

  • (Horizontal truncation) Suppose that ๐‘‹=โ‹ƒ๐‘›=1โˆž๐‘‹๐‘› for some ๐‘‹1โŠ‚๐‘‹2โŠ‚โ‹ฏ with ๐‘‹๐‘›โˆˆ๐’œ๏ธ€. Prove that

    โˆซ๐‘‹๐‘“๐‘‘๐œ‡=lim๐‘›โ†’โˆžโˆซ๐‘‹๐‘›๐‘“๐‘‘๐œ‡
  • (Vertical truncation) Prove that

    โˆซ๐‘“๐‘‘๐œ‡=lim๐‘›โ†’โˆžโˆซ๐‘“๐œ’{|๐‘“|โ‰ค๐‘›}๐‘‘๐œ‡

Remark: a similar question for nonnegative measurable functions appeared in HW4.

๐ฟ1-convergence from dominated convergence

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space, and ๐‘“๐‘›,๐‘“, measurable functions on ๐‘‹, ๐‘›โˆˆโ„•. Suppose that ๐‘“๐‘›โ†’๐‘“ a.e.ย and there is an integrable nonnegative function ๐‘” such that |๐‘“๐‘›(๐‘ฅ)|โ‰ค๐‘”(๐‘ฅ) a.e.ย for all ๐‘›. Prove that ๐‘“๐‘›โ†’๐‘“ in ๐ฟ1, i.e.ย 

lim๐‘›โ†’โˆžโˆซ|๐‘“๐‘›โˆ’๐‘“|=0.

Hint: use DCT.

Lebesgue integrals and affine transformations

Let ๐‘“ be a Lebesgue integrable function on โ„. Prove that

โˆซ๐‘“(๐‘Ÿ๐‘ฅ+๐‘ )๐‘‘๐‘š(๐‘ฅ)=1|๐‘Ÿ|โˆซ๐‘“(๐‘ฅ)๐‘‘๐‘š(๐‘ฅ)

for all real numbers ๐‘Ÿ,๐‘  with ๐‘Ÿโ‰ 0.

Hint: approximate using simple functions ๐‘“.

even moments of Gaussian distribution

Using Multivariable Calculus (and the fact that Riemann integrals coincide with Lebesgue integrals) one can show that

12๐œ‹โˆซโˆ’โˆžโˆž๐‘’โˆ’๐‘ก๐‘ฅ22๐‘‘๐‘ฅ=1๐‘ก

for every ๐‘ก>0. Prove, by (justified!) differentiating with respect to ๐‘ก, that

12๐œ‹โˆซโˆ’โˆžโˆž๐‘ฅ2๐‘›๐‘’โˆ’๐‘ฅ22=(2๐‘›โˆ’1)!!โ‰”(2๐‘›)!2๐‘›๐‘›!

for ๐‘›โˆˆโ„•.

Remark: here the integrals are as defined in this course. Remark: in probability theory, these are the even moments of the standard normal distribution.

Generalized DCT

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space, and ๐‘“๐‘›,๐‘”๐‘›,๐‘“,๐‘”โˆˆ๐ฟ1, ๐‘›โˆˆโ„•. Suppose that

  • lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ) and lim๐‘›โ†’โˆž๐‘”๐‘›(๐‘ฅ)=๐‘”(๐‘ฅ) for a.e. ๐‘ฅ;

  • |๐‘“๐‘›(๐‘ฅ)|โ‰ค๐‘”๐‘›(๐‘ฅ) a.e. for every ๐‘›โˆˆโ„•;

  • ๐‘”๐‘›:๐‘‹โ†’[0,โˆž] and lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐œ‡=โˆซ๐‘”๐‘‘๐œ‡.

Prove that

lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡=โˆซ๐‘“๐‘‘๐œ‡.

Hint: Follow the proof of the DCT, based on FL.

Criterion for ๐ฟ1-convergence

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space. Let ๐‘“๐‘›,๐‘“ be integrable functions on ๐‘‹, ๐‘›โˆˆโ„•. Suppose that lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ) a.e. Prove that

lim๐‘›โ†’โˆžโˆซ|๐‘“๐‘›โˆ’๐‘“|๐‘‘๐œ‡=0ifflim๐‘›โ†’โˆžโˆซ|๐‘“๐‘›|๐‘‘๐œ‡=โˆซ|๐‘“|๐‘‘๐œ‡

Hint: use the generalized DCT.

Some of the following questions will be graded. Do them, and do hand them in.

Formal equivalence between MCT and FL

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space and ๐ฟ+=๐ฟ+(๐‘‹,๐’œ๏ธ€) the space of measurable functions ๐‘“:๐‘‹โ†’[0,โˆž].
Let ๐ผ:๐ฟ+โ†’[0,โˆž] be a function that is increasing in the sense that ๐‘“โ‰ค๐‘” implies ๐ผ(๐‘“)โ‰ค๐ผ(๐‘”). Prove that the following properties are equivalent:

  • ๐ผ is continuous along increasing sequences: if ๐‘“๐‘›โˆˆ๐ฟ+, and ๐‘“๐‘›โ‰ค๐‘“๐‘›+1 for ๐‘›โˆˆโ„•, then lim๐ผ(๐‘“๐‘›)=๐ผ(lim๐‘“๐‘›).

  • if ๐‘“๐‘›โˆˆ๐ฟ+, ๐‘›โˆˆโ„•, then limโ€‰inf๐‘›๐ผ(๐‘“๐‘›)โ‰ฅ๐ผ(limโ€‰inf๐‘›๐‘“๐‘›).

  • ๐ผ is lower semicontinuous: if ๐‘“๐‘›,๐‘“โˆˆ๐ฟ+, and lim๐‘›๐‘“๐‘›=๐‘“, then ๐ผ(๐‘“)โ‰คlimโ€‰inf๐‘›๐ผ(๐‘“๐‘›).

Here lim๐‘›๐‘“๐‘›=๐‘“ means that lim๐‘›๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ) for all ๐‘ฅโˆˆ๐‘‹, and similarly for limโ€‰inf๐‘“๐‘›. Remark: the equivalence betweenย (a) andย (b) shows that the Monotone Convergence Theorem and Fatouโ€™s Lemma are equivalent.

Proof

of (๐šโŸน๐›):
Suppose ๐ผ is continuous along increasing sequences. WTS:

limโ€‰inf๐‘›๐ผ(๐‘“๐‘›)โ‰ฅ๐ผ(limโ€‰inf๐‘›๐‘“๐‘›)

for any sequence (๐‘“๐‘›) in ๐ฟ+.
Define for each ๐‘˜โˆˆโ„•

๐‘”๐‘˜โ‰”inf๐‘›โ‰ฅ๐‘˜๐‘“๐‘›

Then for all ๐‘˜โˆˆโ„•, ๐‘”๐‘˜ is a measurable function. Also notice that by definition, {๐‘”๐‘˜} is an increasing sequence, and

lim๐‘˜โ†’โˆž๐‘”๐‘˜(๐‘ฅ)=limโ€‰inf๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)

for each ๐‘ฅโˆˆ๐‘‹.
Applying (๐š) to ๐‘”๐‘˜: since ๐‘”๐‘˜โ†‘lim๐‘˜๐‘”๐‘˜, we get

lim๐‘˜โ†’โˆž๐ผ(๐‘”๐‘˜)=๐ผ(lim๐‘˜โ†’โˆž๐‘”๐‘˜)=๐ผ(limโ€‰inf๐‘›โ†’โˆž๐‘“๐‘›)

By def of ๐‘”๐‘˜, we have:

๐‘”๐‘˜โ‰ค๐‘“๐‘›for all ๐‘›โ‰ฅ๐‘˜

Since ๐‘”๐‘˜โ‰ค๐‘“๐‘› implies ๐ผ(๐‘”๐‘˜)โ‰ค๐ผ(๐‘“๐‘›), we also have:

๐ผ(๐‘”๐‘˜)โ‰คinf๐‘›โ‰ฅ๐‘˜๐ผ(๐‘“๐‘›)

Taking the limit as ๐‘˜โ†’โˆž, we get

lim๐‘˜โ†’โˆž๐ผ(๐‘”๐‘˜)โ‰คlim๐‘˜โ†’โˆžinf๐‘›โ‰ฅ๐‘˜๐ผ(๐‘“๐‘›)=limโ€‰inf๐‘›โ†’โˆž๐ผ(๐‘“๐‘›)

Combining (5.1) and (5.2), we obtain:

๐ผ(limโ€‰inf๐‘›๐‘“๐‘›)=lim๐‘˜๐ผ(๐‘”๐‘˜)โ‰คlimโ€‰inf๐‘›๐ผ(๐‘“๐‘›).

which is exactly what we want.

โ–ก

Proof

(๐›โŸน๐œ): We now assume (๐›) and prove that ๐ผ is lower semicontinuous, i.e. WTS:

๐‘“๐‘›โ†’๐‘“pointwiselyโ‡’๐ผ(๐‘“)โ‰คlimโ€‰inf๐‘›๐ผ(๐‘“๐‘›).

Given ๐‘“๐‘›โ†’๐‘“ pointwise, we have

๐‘“(๐‘ฅ)=lim๐‘›๐‘“๐‘›(๐‘ฅ)=limโ€‰inf๐‘›๐‘“๐‘›(๐‘ฅ)โˆ€๐‘ฅ

Hence for the sequence {๐‘“๐‘›}, the pointwise limit of ๐‘“๐‘› is exactly limโ€‰inf๐‘›๐‘“๐‘›. (๐›) gives:

lim๐‘›๐‘“๐‘›(๐‘ฅ)=limโ€‰inf๐‘›๐ผ(๐‘“๐‘›)โ‰ฅ๐ผ(limโ€‰inf๐‘›๐‘“๐‘›)=๐ผ(๐‘“)

This is precisely the definition of lower semicontinuity, proving (๐›)โŸน(๐œ).

โ–ก

Proof

of (๐œโŸน๐š):
Assume ๐ผ is lower semi-continuous, i.e. If ๐‘“๐‘›โ†’๐‘“ pointwise, then

๐ผ(๐‘“)โ‰คlimโ€‰inf๐‘›๐ผ(๐‘“๐‘›)

Let (๐‘“๐‘›) be a sequence in ๐ฟ+ such that ๐‘“๐‘›โ†‘๐‘“, i.e.

๐‘“1โ‰ค๐‘“2โ‰คโ‹ฏandlim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ)ptwisely for all ๐‘ฅ

WTS (a): lim๐‘›๐ผ(๐‘“๐‘›)=๐ผ(๐‘“).
Since ๐‘“๐‘› is an increasing seq, ๐‘“๐‘›โ‰ค๐‘“ for each ๐‘›, and since ๐ผ is monotone, we have

๐ผ(๐‘“๐‘›)โ‰ค๐ผ(๐‘“)โˆ€๐‘›

Hence

limโ€‰sup๐‘›๐ผ(๐‘“๐‘›)โ‰ค๐ผ(๐‘“)

And by (๐œ), since ๐‘“๐‘›โ†’๐‘“ pointwisely, we have

๐ผ(๐‘“)โ‰คlimโ€‰inf๐‘›๐ผ(๐‘“๐‘›)

Combining (1) and (2), we get

limโ€‰inf๐‘›๐ผ(๐‘“๐‘›)โ‰ฅ๐ผ(๐‘“)โ‰ฅlimโ€‰sup๐‘›๐ผ(๐‘“๐‘›)

This we also has limโ€‰inf๐‘›๐ผ(๐‘“๐‘›)โ‰คlimโ€‰sup๐‘›๐ผ(๐‘“๐‘›), this shows that lim๐‘›๐ผ(๐‘“๐‘›) exists and equals ๐ผ(๐‘“). This is exactly the statement of (a). Thus (๐œ)โŸน(๐š).

โ–ก

Here we finished the proof that the three properties are equivalent. In particular, the equivalence of (a), (b) shows the equivalence of Fatouโ€™s Lemma and MCT.

Convergence on subsets

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space. Let ๐‘“๐‘›:๐‘‹โ†’[0,โˆž] be a measurable function for each ๐‘›โˆˆโ„•. Suppose that there is a function ๐‘“:๐‘‹โ†’[0,โˆž] such that

lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ) for every ๐‘ฅโˆˆ๐‘‹ and lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›=โˆซ๐‘“
  • Assume that โˆซ๐‘“<โˆž. Show that lim๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=โˆซ๐ธ๐‘“ for every ๐ธโˆˆ๐’œ๏ธ€. Hint: Use Fatou twice. It may be useful to note that even though limโ€‰inf(๐›ผ๐‘›+๐›ฝ๐‘›)โ‰ฅlimโ€‰inf๐›ผ๐‘›+limโ€‰inf๐›ฝ๐‘› in general, if lim๐›ผ๐‘› exists, then limโ€‰inf(๐›ผ๐‘›+๐›ฝ๐‘›)=lim๐›ผ๐‘›+limโ€‰inf๐›ฝ๐‘› for sequences of extended real numbers ๐›ผ๐‘›,๐›ฝ๐‘›.

  • Find an example of ๐‘“๐‘›:โ„โ†’[0,โˆž] on the measure space (โ„,โ„ฌ๏ธ€(โ„),๐‘š) showing that (a) does not necessarily hold if โˆซ๐‘“=โˆž.

Proof

of (a):
By Fatouโ€™s Lemma, since ๐‘“๐‘›โ†’๐‘“ pointwise and all ๐‘“๐‘› are nonnegative,

limโ€‰inf๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=limโ€‰inf๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐œ’๐ธโ‰ฅโˆซ๐‘“๐œ’๐ธ=โˆซ๐ธ๐‘“

For the same reason,

limโ€‰inf๐‘›โ†’โˆžโˆซ๐ธ๐‘๐‘“๐‘›โ‰ฅโˆซ๐ธ๐‘๐‘“

Since

โˆซ๐‘“๐‘‘๐œ‡=โˆซ๐‘‹๐‘“๐‘‘๐œ‡=โˆซ๐ธ๐‘“๐‘‘๐œ‡+โˆซ๐ธ๐‘๐‘“๐‘‘๐œ‡

, we have:

โˆซ๐‘“๐‘‘๐œ‡โˆ’โˆซ๐ธ๐‘“๐‘‘๐œ‡=โˆซ๐ธ๐‘๐‘“๐‘‘๐œ‡โ‰คlimโ€‰inf๐‘›โˆซ๐ธ๐‘๐‘“๐‘›๐‘‘๐œ‡=limโ€‰inf๐‘›(โˆซ๐‘“๐‘›๐‘‘๐œ‡โˆ’โˆซ๐ธ๐‘“๐‘›๐‘‘๐œ‡)=lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡+limโ€‰inf๐‘›(โˆ’โˆซ๐ธ๐‘“๐‘›๐‘‘๐œ‡)=lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡โˆ’limโ€‰sup๐‘›โˆซ๐ธ๐‘“๐‘›๐‘‘๐œ‡=โˆซ๐‘“๐‘‘๐œ‡โˆ’limโ€‰sup๐‘›โˆซ๐ธ๐‘“๐‘›๐‘‘๐œ‡

Rearranging the terms, gives:

โˆซ๐ธ๐‘“โ‰ฅlimโ€‰sup๐‘›โˆซ๐ธ๐‘“๐‘›๐‘‘๐œ‡

Combining with the statement given by Fatouโ€™s Lemma:

limโ€‰inf๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›โ‰ฅโˆซ๐ธ๐‘“

We then have:

limโ€‰inf๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=โˆซ๐ธ๐‘“โ‰ฅlimโ€‰sup๐‘›โˆซ๐ธ๐‘“๐‘›

Since also by definition of limsup and liminf we have:

limโ€‰inf๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›โ‰คlimโ€‰sup๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›

We have:

limโ€‰inf๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=limโ€‰sup๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=lim๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=โˆซ๐ธ๐‘“

This completes the proof.

โ–ก

Solution

of (b): Define for each ๐‘›โˆˆโ„•

๐‘“๐‘›(๐‘ฅ)โ‰”๐œ’[๐‘›,๐‘›+1]+๐œ’(โˆ’โˆž,0]

Then we have:

โˆซ๐‘“๐‘›(๐‘ฅ)=1+โˆž=โˆž

for each ๐‘›. So

lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›(๐‘ฅ)=โˆž

And the pointwise limit of ๐‘“๐‘› is

๐‘“(๐‘ฅ):=lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐œ’(โˆ’โˆž,0]

So the integral of ๐‘“ is also:

โˆซlim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=โˆซ๐‘“(๐‘ฅ)=โˆž

But consider the subset ๐ธ=[0,โˆž), we have:

โˆซ๐ธ๐‘“๐‘›=โˆซ๐œ’[๐‘›,๐‘›+1]=1for all ๐‘›

So

lim๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›=1

while

โˆซ๐ธ๐‘“=0โ‰ lim๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›

This completes the counterexample.

Some integrals

Use the DCT to evaluate the following limits:

  • lim๐‘›โ†’โˆžโˆซ0โˆž๐‘›sin(๐‘ฅ๐‘›)๐‘ฅ(1+๐‘ฅ2)๐‘‘๐‘ฅ
  • lim๐‘›โ†’โˆžโˆซ0๐‘›๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›๐‘‘๐‘ฅ,

    where ๐‘š is a non-negative integer. (The integrals are Lebesgue integrals.)

Solution

of (a):
Define

๐‘“๐‘›โ‰”{๐‘›sin(๐‘ฅ๐‘›)๐‘ฅ(1+๐‘ฅ2),๐‘ฅ>00,๐‘ฅโ‰ค0

Recall that for all ๐‘ฅโˆˆโ„, we have:

|sin(๐‘ฅ)|โ‰ค|๐‘ฅ|

So for all ๐‘›, and for all ๐‘ฅ>0, we have:

|๐‘“๐‘›(๐‘ฅ)|=|๐‘›sin(๐‘ฅ๐‘›)๐‘ฅ(1+๐‘ฅ2)|=๐‘›sin(๐‘ฅ๐‘›)๐‘ฅ(1+๐‘ฅ2)โ‰ค๐‘›๐‘ฅ๐‘›๐‘ฅ(1+๐‘ฅ2)=11+๐‘ฅ2

So by taking:

๐‘”(๐‘ฅ)โ‰”{11+๐‘ฅ2,๐‘ฅ>00,๐‘ฅโ‰ค0

We have:

๐‘”(๐‘ฅ)โ‰ฅ|๐‘“๐‘›(๐‘ฅ)|โˆ€๐‘ฅโˆˆโ„,โˆ€๐‘›

Since ๐‘” is continuous a.e. (except on ๐‘ฅ=0), it is a measurable function. And it is Riemann integrable. We can do Riemann integration of ๐‘”:

โˆซ0โˆž11+๐‘ฅ2๐‘‘๐‘ฅ=[arctan(๐‘ฅ)]0โˆž=๐œ‹2<โˆž

Also, for each ๐‘ฅ>0, since

lim๐‘›โ†’โˆžsin(๐‘ฅ๐‘›)๐‘ฅ๐‘›=1

We have for each ๐‘ฅ>0:

lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=11+๐‘ฅ2lim๐‘›โ†’โˆžsin(๐‘ฅ๐‘›)๐‘ฅ๐‘›=11+๐‘ฅ2

Thus the pointwise limit of ๐‘“๐‘› is:

๐‘“(๐‘ฅ)โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)={11+๐‘ฅ2,๐‘ฅ>00,๐‘ฅโ‰ค0

(Notice it coincides with the ๐‘” that we chose as bound.) We also have:

โˆซ0โˆž๐‘“(๐‘ฅ)๐‘‘๐‘ฅ=๐œ‹2

Then by DCT,

lim๐‘›โ†’โˆžโˆซ0โˆž๐‘›sin(๐‘ฅ๐‘›)๐‘ฅ(1+๐‘ฅ2)๐‘‘๐‘ฅ=lim๐‘›โ†’โˆžโˆซ0โˆž๐‘“๐‘›(๐‘ฅ)๐‘‘๐‘ฅ=โˆซ0โˆžlim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)๐‘‘๐‘ฅ=โˆซ0โˆž๐‘“(๐‘ฅ)๐‘‘๐‘ฅ=๐œ‹2

This finishes the calculation.

Solution

of (b):
Define for each ๐‘›โˆˆโ„•

๐‘“๐‘›(๐‘ฅ)=๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›for0โ‰ค๐‘ฅโ‰ค๐‘›

and ๐‘“๐‘›(๐‘ฅ)=0 for ๐‘ฅ>๐‘›.
Then the integral we wish to evaluate can be written as

lim๐‘›โ†’โˆžโˆซ0๐‘›๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›๐‘‘๐‘ฅ=lim๐‘›โ†’โˆžโˆซ0โˆž๐‘“๐‘›(๐‘ฅ)๐‘‘๐‘ฅ

We first evaluate the ptwise limit function ๐‘“โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ).
For ๐‘ฅ=0:

๐‘“๐‘›(0)=0๐‘š(1โˆ’0๐‘›)๐‘›=0๐‘šโ‹…1=0๐‘š๐‘’โˆ’๐‘ฅโˆ€๐‘›

For 0<๐‘ฅ<โˆž:

๐‘“๐‘›(๐‘ฅ)=๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›

for all large enough ๐‘›.
Recall the standard limit lim๐‘›โ†’โˆž(1โˆ’๐‘ฅ๐‘›)๐‘›=๐‘’โˆ’๐‘ฅ, hence

๐‘“(๐‘ฅ)โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=lim๐‘›โ†’โˆž๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›=๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ

Thus

๐‘“(๐‘ฅ)={0,๐‘ฅ<0๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ,๐‘ฅโ‰ฅ0

Now we determine the dominating function ๐‘”.
Consider the same function as ๐‘“:

๐‘”(๐‘ฅ)โ‰”{0,๐‘ฅ<0๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ,๐‘ฅโ‰ฅ0

We now prove this same function ๐‘” works.
Let ๐‘›โˆˆโ„•.
It is sure that for ๐‘ฅ>๐‘›, ๐‘”(๐‘ฅ)โ‰ฅ|๐‘“๐‘›(๐‘ฅ)| since ๐‘“๐‘›(๐‘ฅ)=0.
So consider ๐‘ฅโˆˆ[0,๐‘›].
Recall the inequality:

ln(1โˆ’๐‘ก)โ‰คโˆ’๐‘กโˆ€๐‘กโˆˆ[0,1]

Thus we have:

(1โˆ’๐‘ฅ๐‘›)๐‘›โ‰ค๐‘’โˆ’๐‘ฅ๐‘›๐‘›=๐‘’โˆ’๐‘ฅ

Therefore,

0โ‰ค๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›โ‰ค๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅfor all 0โ‰ค๐‘ฅโ‰ค๐‘›

Thus in all cases,

|๐‘“๐‘›(๐‘ฅ)|=๐‘“๐‘›(๐‘ฅ)โ‰ค๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ=๐‘”(๐‘ฅ)

Recall:

โˆซ0โˆž๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ๐‘‘๐‘ฅ=ฮ“(๐‘š+1)=๐‘š!

is finite for all nonnegative integers ๐‘š. Thus ๐‘” is integrable. Then ๐‘” is indeed a dominating function for (๐‘“๐‘›).
Applying the DCT, we exchange the limit and the integral:

lim๐‘›โ†’โˆžโˆซ0โˆž๐‘“๐‘›(๐‘ฅ)๐‘‘๐‘ฅ=โˆซ0โˆžlim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)๐‘‘๐‘ฅ=โˆซ0โˆž๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ๐‘‘๐‘ฅ

thus

lim๐‘›โ†’โˆžโˆซ0๐‘›๐‘ฅ๐‘š(1โˆ’๐‘ฅ๐‘›)๐‘›๐‘‘๐‘ฅ=โˆซ0โˆž๐‘ฅ๐‘š๐‘’โˆ’๐‘ฅ๐‘‘๐‘ฅ=ฮ“(๐‘š+1)=๐‘š!

This finishes the evalutation of this integral.

Continuity of translations

Let ๐‘“โˆˆ๐ฟ1(โ„,โ„’๏ธ€,๐‘š). For ๐‘ฅโˆˆโ„, set ๐‘“๐‘ (๐‘ฅ)=๐‘“(๐‘ฅโˆ’๐‘ ). Prove that ๐‘ โ†ฆ๐‘“๐‘  is a continuous map from โ„ to ๐ฟ1. In other words, prove that if ๐‘กโˆˆโ„, then

lim๐‘ โ†’๐‘กโˆซ|๐‘“๐‘ โˆ’๐‘“๐‘ก|๐‘‘๐‘š=0

Hint: approximate ๐‘“.

Proof

We write:

||๐‘“โˆ’๐‘”||1โ‰”โˆซ|๐‘“โˆ’๐‘”|๐‘‘๐‘š

for ๐‘“,๐‘”โˆˆ๐ฟ1(โ„,โ„’๏ธ€,๐‘š). Let ๐œ–>0.
Recall that ๐ถ๐‘(โ„) is dense in ๐ฟ1(โ„). So there exists a function ๐‘”โˆˆ๐ถ๐‘(โ„) such that

โˆฅ๐‘“โˆ’๐‘”โˆฅ1<๐œ–3

Since ๐‘” is continuous and compactly supported, it is uniformly continuous. Denote ๐พโ‰”supp(๐‘”). There exists ๐›ฟ>0 such that for all ๐‘ฅโˆˆโ„,

|๐‘ โˆ’๐‘ก|<๐›ฟโŸน|๐‘”(๐‘ฅโˆ’๐‘ )โˆ’๐‘”(๐‘ฅโˆ’๐‘ก)|<๐œ–3โ‹…๐‘š(๐พ)

Integrating the difference over this support gives:

โˆฅ๐‘”๐‘ โˆ’๐‘”๐‘กโˆฅ1โ‰ค๐œ–3โ‹…๐‘š(๐พ)โ‹…๐‘š(๐พ)=๐œ–3

Recall that ๐ฟ1(โ„,โ„’๏ธ€,๐‘š) is a normed vector space with ||โ‹…||1 as the norm. So by the triangle inequality of a norm, we have:

โˆฅ๐‘“๐‘ โˆ’๐‘“๐‘กโˆฅ1โ‰คโˆฅ๐‘“๐‘ โˆ’๐‘”๐‘ โˆฅ1+โˆฅ๐‘”๐‘ โˆ’๐‘”๐‘กโˆฅ1+โˆฅ๐‘”๐‘กโˆ’๐‘“๐‘กโˆฅ1

By the translation invariance of Lebesgue measure, we have:

โˆฅ๐‘“๐‘ โˆ’๐‘”๐‘ โˆฅ1=โˆฅ๐‘“โˆ’๐‘”โˆฅ1<๐œ–3andโˆฅ๐‘”๐‘กโˆ’๐‘“๐‘กโˆฅ1=โˆฅ๐‘”โˆ’๐‘“โˆฅ1<๐œ–3

By choosing ๐›ฟ such that โˆฅ๐‘”๐‘ โˆ’๐‘”๐‘กโˆฅ1<๐œ–3, we get

โˆฅ๐‘“๐‘ โˆ’๐‘“๐‘กโˆฅ1<๐œ–3+๐œ–3+๐œ–3=๐œ–

Since ๐œ– is arbitrary, this proves that for any ๐‘กโˆˆโ„,

lim๐‘ โ†’๐‘กโˆซ|๐‘“๐‘ โˆ’๐‘“๐‘ก|๐‘‘๐‘š=||๐‘“๐‘ โˆ’๐‘“๐‘ก||1=0

finishing the proof of continuity of the map ๐‘ โ†ฆ๐‘“๐‘ .

โ–ก

An interesting integrable function

For ๐›ผโˆˆ(0,1), define ๐‘”๐›ผ:โ„โ†’โ„ by ๐‘”๐›ผ(๐‘ฅ)=(1โˆ’๐›ผ)๐‘ฅโˆ’๐›ผ for 0<๐‘ฅ<1 and ๐‘”๐›ผ(๐‘ฅ)=0 otherwise. Let (๐‘ฅ๐‘›)๐‘› be an enumeration of the rational numbers, and define ๐‘“:โ„โ†’[0,โˆž] by

๐‘“(๐‘ฅ)=โˆ‘๐‘›=1โˆž2โˆ’๐‘›๐‘”1โˆ’๐‘›โˆ’๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)

Prove that ๐‘“ has the following properties:

  • ๐‘“ is Borel (and hence Lebesgue) measurable;

  • ๐‘“ is Lebesgue integrable, that is โˆซโ„๐‘“๐‘‘๐‘š<โˆž;

  • there exist uncountably many ๐‘ฅโˆˆโ„ such that ๐‘“(๐‘ฅ)<โˆž;

  • ๐‘“ is discontinuous at every point ๐‘ฅโˆˆโ„ where ๐‘“(๐‘ฅ)<โˆž;

  • ๐‘“ is unbounded on any nonempty open interval ๐ผ=(๐‘Ž,๐‘), that is sup๐ผ๐‘“=โˆž;

  • the statements inย (d) andย (e) remain true even if we redefine ๐‘“ on a set of (Lebesgue) measure zero.

  • โˆซ๐ผ๐‘“๐‘๐‘‘๐‘š=โˆž for all ๐‘>1 and all intervals ๐ผ=(๐‘Ž,๐‘).

Proof

of (a):
We define

๐›ผ๐‘›:=1โˆ’๐‘›โˆ’๐‘›

and

โ„Ž๐‘›(๐‘ฅ)โ‰”2โˆ’๐‘›๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)

and

๐‘“๐‘˜(๐‘ฅ)โ‰”โˆ‘๐‘›=1๐‘˜2โˆ’๐‘›๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)=โˆ‘๐‘›=1๐‘˜โ„Ž๐‘›(๐‘ฅ)

to simplify the expression.
Then we have:

๐‘“(๐‘ฅ)=lim๐‘˜โ†’โˆž๐‘“๐‘˜(๐‘ฅ)

Notice that, since each ๐‘”๐›ผ๐‘› is nonnegative, ๐‘“๐‘˜(๐‘ฅ) is a increasing sequence of functions, so for any ๐‘ฅโˆˆโ„, lim๐‘˜โ†’โˆž๐‘“๐‘˜(๐‘ฅ) exists in โ„ฬ„. This shows the well-definedness of ๐‘“=lim๐‘˜โ†’โˆž๐‘“๐‘˜.
Now we claim: each โ„Ž๐‘›(๐‘ฅ) is Borel measurable.
By translate invariance and scaling invariance of Borel measurability, to prove the claim, it suffices to prove that each ๐‘”๐›ผ is Borel measurable for any ๐›ผโˆˆ(0,1).

Figureย 17:

If ๐‘Ž<0, we have:

๐‘”๐›ผโˆ’1((๐‘Ž,โˆž))=โ„

if 0โ‰ค๐‘Žโ‰ค1โˆ’๐›ผ, then we have

๐‘”๐›ผโˆ’1((๐‘Ž,โˆž))=(0,1)

if ๐‘Ž>1โˆ’๐›ผ, then we have

๐‘”๐›ผโˆ’1((๐‘Ž,โˆž))=(0,(1โˆ’๐›ผ๐‘Ž)1/๐›ผ)

This proves that ๐‘”๐›ผ is Borel measurable for any ๐›ผโˆˆ(0,1).
Thus each ๐‘“๐‘˜ being a finite sum of Borel measurable functions, is Borel measurable.
Then ๐‘“ as the limit of Borel measurable function sequence (๐‘“๐‘˜), is Borel measurable.

โ–ก

Proof

of (b):
We define:

โ„Ž๐‘›(๐‘ฅ)โ‰”2โˆ’๐‘›๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)

in order to simplify the expression.
By translation invariance of Lebesgue measure, we have for any ๐›ผ๐‘›, :

โˆซโ„๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)๐‘‘๐‘š=โˆซโ„๐‘”๐›ผ๐‘›(๐‘ฅ)๐‘‘๐‘š๐‘ก=(1โˆ’๐›ผ)โ‹…1โˆ’01โˆ’๐›ผ=1

So by homogeneity of integral,

โˆซโ„โ„Ž๐‘›(๐‘ฅ)๐‘‘๐‘š=โˆซโ„2โˆ’๐‘›๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)๐‘‘๐‘š=2โˆ’๐‘›โˆซโ„๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)๐‘‘๐‘š=12๐‘›

Thus we have:

โˆ‘๐‘›=1โˆžโˆซโ„|โ„Ž๐‘›(๐‘ฅ)|=โˆ‘๐‘›=1โˆžโˆซโ„โ„Ž๐‘›(๐‘ฅ)=1/21โˆ’1/2=1<โˆž

by sum of geometric series. Since this sum of integral of the sequence is finite, we can apply theorem 2.25 on Folland, to exachange the order of limit and integral, and have:

โˆซโ„โˆ‘๐‘›=1โˆžโ„Ž๐‘›(๐‘ฅ)=โˆ‘๐‘›=1โˆžโˆซโ„โ„Ž๐‘›(๐‘ฅ)=1

Hence,

โˆซโ„๐‘“๐‘‘๐‘š=โˆซโ„โˆ‘๐‘›=1โˆžโ„Ž๐‘›(๐‘ฅ)๐‘‘๐‘š=โˆ‘๐‘›=1โˆžโˆซโ„โ„Ž๐‘›(๐‘ฅ)๐‘‘๐‘š=1

So โˆซโ„๐‘“<โˆž. This proves ๐‘“โˆˆ๐ฟ1(โ„).

โ–ก

Proof

of (c):

Lemma 5.19

For ๐‘“โˆˆ๐ฟ+(๐œ‡), if ๐‘“(๐‘ฅ)=+โˆž on a set ๐‘† where ๐œ‡(๐‘†)>0, then โˆซ๐‘“=โˆž

Proof for Lemma: trivially follows from definition. We can pick make a sequence of simple functions (๐œ™๐‘›), setting ๐œ™๐‘›|๐‘†=๐‘› (doable since ๐‘“|๐‘†={โˆž}) then we have:

โˆซ๐œ™๐‘›๐‘‘๐œ‡โ‰ฅโˆซ๐‘›๐œ’๐‘†=๐‘›

So the limit of integral of this simple function sequence is โˆž.

Then (c) follows from the lemma: suppose for contradiction that there exist only countably many ๐‘ฅโˆˆโ„ such that ๐‘“(๐‘ฅ)<โˆž, we denote this this by ๐ถ, then on โ„\๐ถ which has positive measure (since ๐ถ has measure 0), ๐‘“(๐‘ฅ)=โˆž. So by lemma, โˆซ๐‘“=โˆž, contradicting with the fact that โˆซ๐‘“=1 proven in (b). So there exist uncountably many ๐‘ฅโˆˆโ„ such that ๐‘“(๐‘ฅ)<โˆž.

โ–ก

Proof

of (e): Fix an interval ๐ผ. By the density of rational numbers in any interval, there exists some rational ๐‘ฅ๐‘โˆˆ๐ผ. Note that though ๐‘”๐›ผ๐‘(๐‘ฅ๐‘)=0, ๐‘”๐›ผ๐‘(๐‘ฅ) can be arbitrarily large near ๐‘ฅ๐‘.
Fix ๐‘€>0.
It suffices to pick some ๐‘ฅ s.t.

2โˆ’๐‘๐‘”๐›ผ๐‘(๐‘ฅโˆ’๐‘ฅ๐‘)=1โˆ’๐›ผ๐‘2๐‘(๐‘ฅโˆ’๐‘ฅ๐‘)โˆ’๐›ผ๐‘>๐‘€

So by taking any

๐‘ฅโˆˆ(๐‘ฅ๐‘,๐‘ฅ๐‘+(2๐‘๐‘€1โˆ’๐›ผ๐‘)๐›ผ๐‘)โˆฉ๐ผ

then it is done.
Since we already have 2โˆ’๐‘๐‘”๐›ผ๐‘(๐‘ฅโˆ’๐‘ฅ๐‘)>๐‘€, we have

๐‘“(๐‘ฅ)>2โˆ’๐‘๐‘”๐›ผ๐‘(๐‘ฅโˆ’๐‘ฅ๐‘)>๐‘€

Since ๐‘€ is arbitrary, this proves that the value of ๐‘“ on ๐ผ can be unboundedly large, finishing the proof that

sup๐ผ๐‘“=โˆž

โ–ก

Proof

of (d): Notice that we first proved (e) and then letโ€™s prove (d) using the conclusion of (e).
Let ๐‘ฅโˆˆโ„ s.t. ๐‘“(๐‘ฅ)<โˆž.
Suppose ๐‘“ is continuous at ๐‘ฅ, then by definition, there exists an open neighborhood ๐ต๐›ฟ(๐‘ฅ)=(๐‘ฅโˆ’๐›ฟ,๐‘ฅ+๐›ฟ) s.t. |๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|<183 for all ๐‘ฆโˆˆ๐ต๐›ฟ(๐‘ฅ).
But since the neighborhood is an interval, we have:

sup(๐‘ฅโˆ’๐›ฟ,๐‘ฅ+๐›ฟ)๐‘“=โˆž

by (e). This two facts contradicts. So by contradiction we have proved that ๐‘“ is discontinuous at ๐‘ฅ.
So we can conclude that ๐‘“ is discontinuous at any point ๐‘ฅ s.t. ๐‘“(๐‘ฅ)<โˆž.

โ–ก

Proof

of (f): Let ๐ผ be an interval.
Suppose we have redefined ๐‘“ on a measure 0 set. We pick a rational ๐‘ฅ๐‘โˆˆ๐ผ (It does not matter whether the new ๐‘“ is defined there.)
For arbitrary ๐‘€>0, we can still always find an ๐‘ฅ s.t. ๐‘ฅโˆˆ(๐‘ฅ๐‘,๐‘ฅ๐‘+(2๐‘๐‘€1โˆ’๐›ผ๐‘)๐›ผ๐‘)โˆฉ๐ผ that keeps its original ๐‘“(๐‘ฅ), which guarantees that ๐‘“(๐‘ฅ)>๐‘€, implying sup๐ผ๐‘“=โˆž. This is because, if not so, then it means that we have modified the whole interval (๐‘ฅ๐‘,๐‘ฅ๐‘+(2๐‘๐‘€1โˆ’๐›ผ๐‘)๐›ผ๐‘)โˆฉ๐ผ, which is not a measure zero set, conflicting with the statement "redefining ๐‘“ on a measure zero set". So (e) must still hold true.
For (d), we apply the same trick as original, getting an open interval around ๐‘ฅ s.t. |๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|<183 for all ๐‘ฆโˆˆ๐ต๐›ฟ(๐‘ฅ)=(๐‘ฅโˆ’๐›ฟ,๐‘ฅ+๐›ฟ). And by the restated (d), even if we modified a set of measure zero on (๐‘ฅโˆ’๐›ฟ,๐‘ฅ+๐›ฟ), we still reaches the the same conclusion that sup(๐‘ฅโˆ’๐›ฟ,๐‘ฅ+๐›ฟ)๐‘“=โˆž, thus causing the same contradiction.
This finishes the proof.

โ–ก

Proof

of (g): WTS: โˆซ๐ผ๐‘“๐‘๐‘‘๐‘š=โˆž for all ๐‘>1 and every interval ๐ผ Claim: for each ๐‘›, ๐‘”๐›ผ๐‘›๐‘ fails to be in ๐ฟ1 when ๐‘>1, i.e its integral is โˆž. Fix ๐‘>1.
Since by translation invariance of Lebesgue integral,:

โˆซโ„(2โˆ’๐‘›๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›))๐‘๐‘‘๐‘š=2โˆ’๐‘›๐‘โˆซโ„๐‘”๐›ผ๐‘›(๐‘ฅ)๐‘๐‘‘๐‘š

where

๐‘”๐›ผ๐‘›(๐‘ก)๐‘=(๐‘›โˆ’๐‘›๐‘กโˆ’๐›ผ๐‘›)๐‘=๐‘›โˆ’๐‘›๐‘๐‘กโˆ’๐‘๐›ผ๐‘›=๐‘›โˆ’๐‘›๐‘๐‘กโˆ’๐‘(1โˆ’๐‘›โˆ’๐‘›)

Since ๐‘>1, there eixst ๐‘ such that for all ๐‘โ‰ฅ๐‘›, the exponent โˆ’๐‘(1โˆ’๐‘›โˆ’๐‘›) is less than โˆ’1, causing โˆซ01๐‘กโˆ’๐‘+๐‘๐‘›โˆ’๐‘›๐‘‘๐‘ก=+โˆž for sufficiently large ๐‘›. Multiplying by the constant ๐‘›โˆ’๐‘›๐‘ does not remove the infinity.
Hence for large enough ๐‘›, each individual summand has an infinite integral, then by monotonicity of integral,

๐‘“๐‘(๐‘ฅ)=(โˆ‘๐‘›2โˆ’๐‘›๐‘”๐›ผ๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›))๐‘โ‰ฅ2โˆ’๐‘๐‘”๐›ผ๐‘๐‘(๐‘ฅโˆ’๐‘ฅ๐‘)

also has an infinite integral, finishing the proof.

โ–ก

Nur fรผr Verrรผckte

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!)

  1. Make an accurate sketch of the graph of the function in the last problem.

6 product measure and Fubini-Tonelli theorem

6.1 product space and product measure [Fol 1.2, finished; 2.5]

Goal: Given (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–), construct ๐‘‹=โˆ๐‘‹๐‘–, s.t.

๐œ‡(๐ธ1ร—๐ธ2)=๐œ‡1(๐ธ1)๐œ‡2(๐ธ2)

So that we can do Fubini (iterated integration) like that in Riemann integral.

6.1.1 product ๐œŽ-algebra

Definition 6.33 : product ๐œŽ-algebra

Suppose (๐‘‹๐‘–,๐’œ๏ธ€๐‘–) mble, 1โ‰ค๐‘–โ‰ค๐‘›, the product ๐œŽ-algebra ๐ด1โŠ—โ‹ฏโŠ—๐ด๐‘› on ๐‘‹1ร—โ‹ฏร—๐‘‹๐‘› is the smallest ๐œŽ-algebra s.t. the coordinate map

๐œ‹๐‘—:๐‘‹1ร—โ‹ฏร—๐‘‹๐‘›โ†’๐‘‹๐‘—

is measurable.
ๅณ the ๐œŽ-algebra generated by: {๐œ‹๐›ผ(๐ธ๐›ผ):๐ธ๐›ผโˆˆ๐’œ๏ธ€๐›ผ}.

๐ด1โŠ—โ‹ฏโŠ—๐ด๐‘›:=<{๐œ‹๐›ผ(๐ธ๐›ผ):๐ธ๐›ผโˆˆ๐’œ๏ธ€๐›ผ}>

ๆˆ‘ไปฌๅฎนๆ˜“ๅ‘็Žฐ:

Proposition 6.16
๐ด1โŠ—โ‹ฏโŠ—๐ด๐‘›=<{๐ธ1ร—โ‹ฏร—๐ธ๐‘›โˆˆ๐’œ๏ธ€๐‘–ร—โ‹ฏร—๐’œ๏ธ€๐‘›}>
Proof

By def ๆ˜“ๅพ—. (Prop 1.14 in book).

โ–ก

6.1.2 product Borel algebra โŠ‚ Borel algebra of the product space

Proposition 6.17

If ๐‘‹1,โ‹ฏ,๐‘‹๐‘› are metric spaces. Let ๐‘‹โ‰”๐‘‹1ร—โ‹ฏร—๐‘‹๐‘› (with product metric), then:

โจ‚๐‘–=1๐‘›โ„ฌ๏ธ€๐‘‹๐‘–โІโ„ฌ๏ธ€๐‘‹

and the equality holds if ๐‘‹๐‘– separable โˆ€๐‘–.

Proof
โจ‚๐‘–=1๐‘›โ„ฌ๏ธ€๐‘‹๐‘–=by prop<{๐‘ˆ1ร—โ‹ฏร—๐‘ˆ๐‘› each open}>โІโ„ฌ๏ธ€๐‘‹

Now let ๐ถ๐‘–โІ๐‘‹๐‘– dense, ctbl.
Set

โ„ฐ๏ธ€๐‘–โ‰”{๐ต๐‘Ÿ(๐‘ฅ)โˆฃ๐‘ฅโˆˆ๐ถ๐‘–,๐‘Ÿโˆˆโ„š>0}โІโ„ฌ๏ธ€๐‘‹๐‘–

Then: every open set in ๐‘‹1ร—โ‹ฏร—๐‘‹๐‘› is a ctbl union of products ๐ต1ร—โ‹ฏร—๐ต๐‘›, each ๐ต1โˆˆโ„ฐ๏ธ€๐‘–. Then we have:

๐ต๐‘‹=<{๐ต1ร—...ร—๐ต๐‘›}>โŠ‚โจ‚1๐‘›๐ต๐‘‹๐‘–

โ–ก

Example 6.16
โ„ฌ๏ธ€โ„๐‘›=โ„ฌ๏ธ€โ„โŠ—โ‹ฏโŠ—โ„ฌ๏ธ€โ„
Corollary 6.15

if (๐‘‹,๐’œ๏ธ€) is a mble space, then

๐‘“:๐‘‹โ†’โ„‚(๐’œ๏ธ€,โ„ฌ๏ธ€โ„‚)-mble โ‡”โ„œ๐‘“,โ„‘๐‘“๐’œ๏ธ€-mble
Proof

็•ฅ.

โ–ก

6.1.3 construction of product measure

ไธ‹้ขๆˆ‘ไปฌๆž„ๅปบ product measure: Let (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–), 1โ‰ค๐‘–โ‰ค๐‘› be mble spaces.
And let ๐‘‹โ‰”๐‘‹1ร—...ร—๐‘‹๐‘›, ๐’œ๏ธ€โ‰”๐’œ๏ธ€๐‘–โŠ—โ‹ฏโŠ—๐’œ๏ธ€๐‘› Goal: ้€š่ฟ‡ Hahn-Kromolgrov ๆฅๆž„ๅปบ product measure on product mble space. Idea: Let

๐’œ๏ธ€โ€ฒโ‰”{all finite disjoint unions of rectangles (each measurable)๐ด1ร—โ‹ฏร—๐ด๐‘›}

Step 1:

6.1.4 all finite disjoint unions of rectangles as an algebra

Proposition 6.18

๐’œ๏ธ€โ€ฒ is an algebra.

Proof

The set โ„ฐ๏ธ€โ‰”{rectangles}โІ๐’ซ๏ธ€(๐‘‹) satisfies:

  • โŒ€โˆˆโ„ฐ๏ธ€

  • ๐ธ,๐นโˆˆโ„ฐ๏ธ€โŸน๐ธโˆฉ๐นโˆˆโ„ฐ๏ธ€

  • ๐ธโˆˆโ„ฐ๏ธ€โŸน๐ธ๐‘ is a finite disjoint union of recs (็”ปๅ›พๅฏ็Ÿฅ).

โ–ก

Step 2:

6.1.5 ๅ„็ปดๅบฆ measure ็š„ product ไฝœไธบ rectangle ็š„ measure, ไปŽ่€Œๅฎšไน‰ premeasure

Now define

๐œ‡โ€ฒ:๐’œ๏ธ€โ€ฒโ†’[0,โˆž]

as follows:

๐œ‡โ€ฒ(โจ†๐‘˜=1๐‘๐ธ1(๐‘˜)ร—โ‹ฏร—๐ธ๐‘›(๐‘˜))=โˆ‘๐‘˜=1๐‘โˆ๐‘–=1๐‘›๐œ‡๐‘–(๐ธ๐‘–(๐‘˜))

Claim 2:

Proposition 6.19

(1) ๐œ‡โ€ฒ is a well-defined premeasure on ๐’œ๏ธ€โ€ฒ.
(2) If each ๐œ‡๐‘– is ๐œŽ-finite, so is ๐œ‡โ€ฒ.

Proof

Sketch: (2) DIY. (1) STS(check): if ๐ธ=๐ธ1ร—โ‹ฏ๐ธ๐‘› is a finite or ctbl union of rects ๐ธ(๐‘˜)=๐ธ1(๐‘˜)ร—โ‹ฏร—๐ธ๐‘›(๐‘˜), then

โˆ1๐‘›๐œ‡๐‘–(๐ธ๐‘–)=โˆ‘๐‘˜โˆ1๐‘›๐œ‡๐‘–(๐ธ๐‘–(๐‘˜))

Use Tonelli for sums and integrals:

โˆ1๐‘›๐œ’๐ธ๐‘–(๐‘ฅ๐‘–)=๐œ’๐ธ(๐‘ฅ1,โ‹ฏ,๐‘ฅ๐‘›)=โˆ‘๐‘˜๐œ’๐ธ(๐‘˜)(๐‘ฅ1,โ‹ฏ,๐‘ฅ๐‘›)=โˆ‘๐‘˜โˆ1๐‘›๐œ’๐ธ๐‘–(๐‘˜)(๐‘ฅ๐‘–)

Integrate w.r.t. ๐‘ฅ1:

๐œ‡1(๐ธ1)โˆ๐‘–=1๐‘›๐œ’๐ธ๐‘–(๐‘ฅ๐‘–)=โˆซโˆ‘๐‘˜โˆ1๐‘›๐œ’๐ธ๐‘–(๐‘˜)(๐‘ฅ๐‘–)=Tonelliโˆ‘๐‘˜โˆซ(โˆ1๐‘›๐œ’๐ธ๐‘–(๐‘˜)(๐‘ฅ๐‘–))๐‘‘๐œ‡1(๐‘ฅ1)=โˆ‘๐‘˜๐œ‡1(๐ธ1(๐‘˜))โˆ๐‘–=1๐‘›๐œ’๐ธ๐‘–(๐‘˜)(๐‘ฅ๐‘–)

And repeat for ๐‘–=2,โ‹ฏ,๐‘›.

โ–ก

6.1.6 HK extension of the premeasure as definition of product measure

Step 3: ็Žฐๅœจๅทฒ็ปๆœ‰ไบ† ๐œŽ-finite ็š„ premeasure, ๆˆ‘ไปฌๅฏไปฅๅบ”็”จ HK Thm ๆž„ๅปบๅ‡บๅฎŒๆ•ด็š„ measure. Now use HK:

Corollary 6.16

โˆƒ measure ๐œ‡โ‰”๐œ‡1ร—โ‹ฏร—๐œ‡๐‘› on ๐’œ๏ธ€=๐’œ๏ธ€1โŠ—โ‹ฏ๐’œ๏ธ€๐‘› extending ๐œ‡โ€ฒ.
(And if each ๐œ‡๐‘– are ๐œŽ-finite, then product measure ไนŸ ๐œŽ-finite, ไปŽ่€Œ ๐œ‡ ๆ˜ฏ unique extension.)

็”ฑๆญค, ๆˆ‘ไปฌไปŽ ๐’œ๏ธ€1,โ‹ฏ,๐’œ๏ธ€๐‘› ็š„ measure ไธญๆž„ๅปบๅ‡บไบ†ๅฎƒไปฌ็š„ product measure.

6.1.7 associativity of product ๐œŽ-algebra and ๐œŽ-finite product measure

Corollary 6.17 : associativity of product measure

ๆ€ปๆœ‰

๐’œ๏ธ€1โŠ—๐’œ๏ธ€2โŠ—๐’œ๏ธ€3=(๐’œ๏ธ€1โŠ—๐’œ๏ธ€2)โŠ—๐’œ๏ธ€3=๐’œ๏ธ€1โŠ—(๐’œ๏ธ€2โŠ—๐’œ๏ธ€3)

ๅนถไธ”, if ๐œ‡1,๐œ‡2,๐œ‡3 are ๐œŽ-finite, then:

๐œ‡1ร—๐œ‡2ร—๐œ‡3=(๐œ‡1ร—๐œ‡2)ร—๐œ‡3=๐œ‡1ร—(๐œ‡2ร—๐œ‡3)
Proof

DIY. ๅ‰่€… play with def, ๅŽ่€…็›ดๆŽฅ็”ฑ ๐œŽ-finite ็š„ premeasure ็š„ HK extension unique ๅพ—ๅˆฐ.

โ–ก

6.1.8 ๅฆ‚ไฝ•่ฏๆ˜Žไธ€ไธชๅ‡ฝๆ•ฐ product measurable

่ฆ่ฏๆ˜Žไธ€ไธชๅ‡ฝๆ•ฐๆ˜ฏ product measurable ็š„, ๅช้œ€่ฆ่ฏๆ˜ŽๅฎƒๅฏนไบŽๆฏไธช measurable rectangle ็š„ preimage ้ƒฝๆ˜ฏ measurable ็š„ๅณๅฏ.

Lemma 6.20

Suppose ๐‘“:๐‘‹โ†’๐‘Œร—๐‘ is a function from a measurable space (๐‘‹,๐’œ๏ธ€) to a product measure space (๐‘Œร—๐‘,โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2).
Claim: If ๐‘“โˆ’1(๐ต1ร—๐ต2)โˆˆ๐’œ๏ธ€ for each measurable rectangle ๐ต1ร—๐ต2โˆˆโ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2, then ๐‘“ is an (๐’œ๏ธ€,โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2)-measurable function.

Proof

ๅ› ไธบ product ๐œŽ-algebra ็”ฑๆ‰€ๆœ‰็š„ measurable rectangles ็”Ÿๆˆ.
Hw7 ไธญๆœ‰ๅฆไธ€็‰ˆ็š„่ฏๆ˜Žไฝœไธบ lemma.

โ–ก

ๅœจ Hw7 ไธญๆˆ‘ไปฌๅฏไปฅ้€š่ฟ‡่ฟ™ไธช lemma ๅพ—ๅˆฐไธ€ไธช็ป“่ฎบ: ๅฆ‚ๆžœ ๐ธโˆˆ๐’œ๏ธ€โŠ—๐’œ๏ธ€, ้‚ฃไนˆ ๐ธ ็š„ diagonal ไธ€ๅฎš โˆˆ๐’œ๏ธ€.

ๅฏนไบŽ็‰นๆฎŠ็š„ๅ‡ฝๆ•ฐ, ๆฏ”ๅฆ‚ไธคไธช measurable function ็š„ไน˜็งฏ, ๅ…ถไธ€ๅฎšๆ˜ฏ product measurable ็š„.

Lemma 6.21 : easier Fubini

ๆกไปถ: (๐‘‹,๐’œ๏ธ€,๐œ‡), (๐‘Œ,โ„ฌ๏ธ€,๐œˆ) ไธบ arbitrary measure space (ไธ้œ€่ฆ ๐œŽ-finite.), ๐‘“:๐‘‹โ†’โ„‚, ๐‘”:๐‘Œโ†’โ„‚ ไธบ measurable functions.
็ป“่ฎบ:

โ„Žโ‰”๐‘“๐‘”is (๐’œ๏ธ€โŠ—โ„ฌ๏ธ€)-measurable

ๅนถไธ”ๅฆ‚ๆžœ ๐‘“,๐‘” ๆ˜ฏ ๐ฟ1 ็š„, ้‚ฃไนˆ โ„Žโˆˆ๐ฟ1(๐œ‡ร—๐œˆ) ๅนถไธ”

โˆซโ„Ž๐‘‘(๐œ‡ร—๐œˆ)=(โˆซ๐‘“๐‘‘๐œ‡)(โˆซ๐‘”๐‘‘๐œˆ)
Proof

in hw 8. ่ฟ™ไธช statement ่กจ็คบไธ€ไธช ๐’œ๏ธ€-measurable ็š„ๅ‡ฝๆ•ฐๅ’Œไธ€ไธช โ„ฌ๏ธ€-measurable ็š„ๅ‡ฝๆ•ฐ็š„ไน˜็งฏๆ˜ฏ (๐’œ๏ธ€โŠ—โ„ฌ๏ธ€)-measurable ็š„.

โ–ก

6.2 Tonelliโ€™s Thm [Fol 2.5]

ๆˆ‘ไปฌๅฐ† focus on the case ๐‘›=2: (๐‘‹,๐’œ๏ธ€,๐œ‡), (๐‘Œ,โ„ฌ๏ธ€,๐œˆ), ่€ƒ่™‘

(๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€,๐œ‡ร—๐œˆ)

ไปŽ่€Œ, ๅฎƒๅฏไปฅๆŽจๅนฟๅˆฐไปปไฝ• finite ไธช measure space ็š„ product ไธŠ.

6.2.1 ๐ธโŠ‚๐‘‹ร—๐‘Œ ็š„ section

Definition 6.34 : ๐‘ฅ-section, ๐‘ฆ-section

็ป™ๅฎš product space ไธŠ็š„้›†ๅˆ ๐ธโŠ‚๐‘‹ร—๐‘Œ, ๅฏนไบŽ ๐‘ฅโˆˆ๐‘‹, ๐‘ฆโˆˆ๐‘Œ, ๆˆ‘ไปฌๅฎšไน‰:

๐ธ๐‘ฅ:={๐‘ฆโˆˆ๐‘Œโˆฃ(๐‘ฅ,๐‘ฆ)โˆˆ๐ธ}๐ธ๐‘ฆโ‰”{๐‘ฅโˆˆ๐‘‹โˆฃ(๐‘ฅ,๐‘ฆ)โˆˆ๐‘Œ}
Figureย 18:

็ป™ๅฎšไปŽ product space ๅ‡บๅ‘็š„ๅ‡ฝๆ•ฐ ๐‘“:๐‘‹ร—๐‘Œโ†’โ„‚, ๅฏนไบŽ ๐‘ฅโˆˆ๐‘‹, ๐‘ฆโˆˆ๐‘Œ, ๆˆ‘ไปฌๅฎšไน‰:

๐‘“๐‘ฅ(๐‘ฆ)โ‰”๐‘“๐‘ฆ(๐‘ฅ)โ‰”๐‘“(๐‘ฅ,๐‘ฆ)

่กจ็คบๅ›บๅฎšไฝไธ€ไธชๅ˜้‡, ๅฆไธ€ไธชๅ˜้‡็š„ๅ˜ๅŒ–.

Example 6.17

ๅฏนไบŽไปปๆ„็š„ ๐ธโŠ‚๐‘‹ร—๐‘Œๅฆ‚ๆžœๅฎšไน‰:

๐‘“โ‰”๐œ’๐ธ

้‚ฃไนˆๆœ‰:

๐‘“๐‘ฅ=๐œ’๐ธ๐‘ฅ,๐‘“๐‘ฆ=๐œ’๐ธ๐‘ฆ

ๅฏนไบŽ rectangle: ๐ธ=๐ดร—๐ต, ๐ดโˆˆ๐’œ๏ธ€,๐ตโˆˆโ„ฌ๏ธ€, ๆœ‰

๐ธ๐‘ฅ={โŒ€,๐‘ฅโˆ‰๐ด๐ต,๐‘ฅโˆˆ๐ด
Proposition 6.20

(a)

๐ธโˆˆ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€โŸน{๐ธ๐‘ฅโˆˆโ„ฌ๏ธ€,โˆ€๐‘ฅโˆˆ๐‘‹๐ธ๐‘ฆโˆˆ๐’œ๏ธ€,โˆ€๐‘ฆโˆˆ๐‘Œ

(b)

๐‘“ is ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€-measurableโŸน{๐‘“๐‘ฅ is โ„ฌ๏ธ€-measurableโˆ€๐‘ฅ๐‘“๐‘ฆ is ๐’œ๏ธ€-measurableโˆ€๐‘ฆ
Proof

(a) Let

โ„ฐ๏ธ€โ‰”{๐ธโŠ‚๐‘‹ร—๐‘Œโˆฃ๐ธ๐‘ฅโˆˆโ„ฌ๏ธ€โˆ€๐‘ฅโˆˆ๐‘‹ and ๐ธ๐‘ฆโˆˆ๐’œ๏ธ€โˆ€๐‘ฆโˆˆ๐‘Œ}

Claim: โ„ฐ๏ธ€ ๅŒ…ๅซไบ†ๆ‰€ๆœ‰็š„ rectangles, ๅนถไธ” โ„ฐ๏ธ€ a ๐œŽ-algebra.
ๅฎนๆ˜“่ฏๆ˜Ž่ฟ™ไธ€็‚น. ไปŽ่€Œ, ็”ฑ ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€ ็š„ๅฎšไน‰ (ไธบๅŒ…ๅซๆ‰€ๆœ‰ rectangles ็š„ๆœ€ๅฐ ๐œŽ-algebra) ๅพ— ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€โŠ‚โ„ฐ๏ธ€, ไปŽ่€Œ (a) ๆˆ็ซ‹
ๅนถไธ”็”ฑไบŽ (check) ๐‘“๐‘ฅโˆ’1(๐‘‰)=(๐‘“โˆ’1(๐‘‰))๐‘ฅ (Similar for ๐‘“๐‘ฆ), (a)โŸน(b).

โ–ก

Definition 6.35 : monotone class

Given a set ๐‘‹, a collection ๐ถโŠ‚๐’ซ๏ธ€(๐‘‹) is called a monotone class, if it is closed under countable increasing unions and countable decreasing intersections

ๅฝ“็„ถ, ไธ€ไธช ๐œŽ-algebra ๆ˜ฏไธ€ไธช monotone class. monotone class ๆ˜ฏไธ€ไธชๆฏ” ๐œŽ-algebra ๆ›ดๅผฑ็š„ๅฎšไน‰.

6.2.2 tool needed to show Tonelli: Monotone Class Lemma

Lemma 6.22 : monotone class lemma

Let ๐’œ๏ธ€โŠ‚๐’ซ๏ธ€(๐‘‹) be an algebra.
Define ๐’ž๏ธ€ ไธบๅŒ…ๅซ ๐’œ๏ธ€ ็š„ๆœ€ๅฐ็š„ montone class.
Claim:

<๐’œ๏ธ€>=๐’ž๏ธ€
Proof

๐’ž๏ธ€โŠ‚<๐’œ๏ธ€> is trivial.
<๐’œ๏ธ€>โŠ‚๐’ž๏ธ€: STS ๐’ž๏ธ€ ๆ˜ฏไธ€ไธช ๐œŽ-algebra. see p.66. ๅ…ทไฝ“ๅšๆณ•ๆฏ”่พƒ tricky, ไฝ†ๆ˜ฏๆ€่ทฏๆ˜ฏๅ…ˆ่ฏๆ˜Ž ๐’ž๏ธ€ ๆ˜ฏไธ€ไธช algebra (่ฟ™ไธ€้ƒจๅˆ†่พƒ้šพ. ๆˆ‘ไปฌๅฏนไบŽ ๐ธโˆˆ๐’ž๏ธ€, define ๐’ž๏ธ€(๐ธ) ไธบ ๐’ž๏ธ€ ไธญๆ‰€ๆœ‰ๅ’Œๅฎƒ็š„ไบคๅ’ŒๅทฎไนŸไป็„ถๅœจ ๐’ž๏ธ€ ไธญ็š„ ๐น ๆž„ๆˆ็š„้›†ๅˆ, ๅนถๅ‘็Žฐ่ฟ™ไธชๅญ้›† ๐’ž๏ธ€(๐ธ) ไนŸๅŒๆ ทๆ˜ฏไธ€ไธช monotone class. ไปŽ่€Œ ๐’ž๏ธ€=๐’ž๏ธ€(๐ธ) );

็„ถๅŽๅฏนไบŽไปปๆ„็š„ seq, ๅ…ถ ๅ‰ ๐‘› ้กน็š„ finite union (โˆช๐‘–=1๐‘›๐ธ๐‘–)seq ๆ˜ฏไธ€ไธช increasing seq, ๅ…ถ limit ็ญ‰ไบŽๅŽŸ seq limit, ๆ˜ฏๅฑžไบŽ ๐’ž๏ธ€ ็š„.

โ–ก

ไฝ†ๆ˜ฏๅฆ‚ๆžœๆˆ‘ไปฌๅช็Ÿฅ้“ ๐’ž๏ธ€โŠƒ๐’œ๏ธ€, ๆฒกๆœ‰ "ๅŒ…ๅซ ๐’œ๏ธ€ ็š„ๆœ€ๅฐ็š„ monotone class" ่ฟ™ไธชๆกไปถๆ€ŽไนˆๅŠž? ้‚ฃไนŸๆฒกๅ…ณ็ณป, ๆˆ‘ไปฌๅพˆ่‡ช็„ถๅพ—ๅ‡บ

Corollary 6.18

Let ๐’œ๏ธ€โŠ‚๐’ซ๏ธ€(๐‘‹) be an algebra, ๐’ž๏ธ€โŠƒ๐’œ๏ธ€ be a monotone class, ้‚ฃไนˆไธ€ๅฎšๆœ‰

๐’ž๏ธ€โŠƒโŸจ๐’œ๏ธ€โŸฉ

ๅ› ไธบ โŸจ๐’œ๏ธ€โŸฉ= "ๅŒ…ๅซ ๐’œ๏ธ€ ็š„ๆœ€ๅฐ็š„ monotone class" โŠ‚๐’ž๏ธ€.

6.2.3 Tonelli for sets: integrating a section to get product measure

Theorem 6.28 : Tonelli for sets

Let (๐‘‹,๐’œ๏ธ€,๐œ‡), (๐‘Œ,โ„ฌ๏ธ€,๐œˆ) be ๐œŽ-finite measure spaces.
Take ๐ธโˆˆ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€. Then:

๐‘ฅโ†ฆ๐œˆ(๐ธ๐‘ฅ),๐‘ฆโ†ฆ๐œ‡(๐ธ๐‘ฆ) are measurable

ๅนถไธ”

(๐œ‡ร—๐œˆ)(๐ธ)=โˆซ๐œˆ(๐ธ๐‘ฅ)๐‘‘๐œ‡(๐‘ฅ)=โˆซ๐œ‡(๐ธ๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)
Proof

Define:

๐’ž๏ธ€={๐ธโŠ‚๐‘‹ร—๐‘Œโˆฃ๐‘ฅโ†ฆ๐œˆ(๐ธ๐‘ฅ),๐‘ฆโ†ฆ๐œ‡(๐ธ๐‘ฆ) are measurable โˆ€(๐‘ฅ,๐‘ฆ)โˆˆ๐ธ andโ‹ฏ(2)}

Claim 1: ๐’ž๏ธ€ contains ๆ‰€ๆœ‰็š„ rectangles. Proof of Claim 1: ่€ƒ่™‘ ๐ธ=๐ดร—๐ตโˆˆ๐’œ๏ธ€ร—โ„ฌ๏ธ€, ๅณไธบไธ€ไธช rectangle. ไธŠไธ€ lec ไธญ, ๆˆ‘ไปฌ by def confirm: ๐’œ๏ธ€ร—โ„ฌ๏ธ€โŠ‚๐’œ๏ธ€โŠ—โ„ฌ๏ธ€.
้‚ฃไนˆๅฏนไบŽไปปๆ„็š„ (๐‘ฅ,๐‘ฆ)โˆˆ๐ธ, ๆˆ‘ไปฌๆœ‰: ๐œˆ(๐ธ๐‘ฅ)=๐œˆ(๐ต), ๅฏนไบŽๆ‰€ๆœ‰็š„ (๐‘ฅ,๐‘ฆ)โˆ‰๐ธ, ๅˆ™ๆœ‰ ๐œˆ(๐ธ๐‘ฅ)=โŒ€.
ๆ‰€ไปฅๅฏนไบŽไปปๆ„็š„ ๐‘ฅ, ๐œˆ(๐ธ๐‘ฅ)=๐œ’๐ด(๐‘ฅ)๐œˆ(๐ต), ๅŒ็† ๐œ‡(๐ธ๐‘ฆ)=๐œ’๐ด(๐‘ฅ)๐œ‡(๐ด).
็”ฑๆญคๅพ—ๅˆฐ ๐‘ฅโ†ฆ๐œˆ(๐ธ๐‘ฅ),๐‘ฆโ†ฆ๐œ‡(๐ธ๐‘ฆ) ๆ˜ฏ measurable ็š„, ๅนถไธ”

๐œ‡ร—๐œˆ(๐ธ)=๐œ‡(๐ด)ร—๐œˆ(๐ต)=๐œ‡(๐ด)โˆซ๐œ’๐ต(๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)=โˆซ๐œ‡(๐ธ๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)

ๅŒ็†, ๐œ‡ร—๐œˆ(๐ธ)=โˆซ๐œˆ(๐ธ๐‘ฅ)๐‘‘๐œ‡(๐‘ฅ). ไปŽ่€Œๅพ—่ฏ. ไปŽ่€Œ, ๅฏนไบŽไปปๆ„ union of finite disjoint rectangles, ่ฟ™ไธช็ป“่ฎบไนŸๆˆ็ซ‹, by additivity in definition. ๅ› ่€Œ ๐’ž๏ธ€ ไธบไธ€ไธช algebra.

Note: ็”ฑไบŽ ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€ ไธบๅŒ…ๅซๆ‰€ๆœ‰ rectangles ็š„ๆœ€ๅฐ ๐œŽ-algebra, ๆˆ‘ไปฌๅช้œ€่ฆ่ฏๆ˜Ž ๐’ž๏ธ€ ไธบไธ€ไธช ๐œŽ-algebra, ้‚ฃไนˆๅฎƒไธ€ๅฎšๅŒ…ๅซ ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€. ๅนถไธ” by Monotone Class Lemma, STS: ๐’ž๏ธ€ ไธบไธ€ไธช monotone class.

Claim 2: ๐’ž๏ธ€ ไธบไธ€ไธช monotone class. ไปค {๐ธ๐‘›} ไธบไธ€ไธช increasing seq in ๐’ž๏ธ€, ๅฎšไน‰ๅ…ถ union ไธบ ๐ธโ‰”โ‹ƒ๐‘›=1โˆž๐ธ๐‘›. ๅนถๅฎšไน‰:

๐‘“๐‘›(๐‘ฆ)โ‰”๐œ‡((๐ธ๐‘›)๐‘ฆ)

ๆ นๆฎ ๐’ž๏ธ€ ็š„ definition, ๆฏไธช ๐‘“๐‘› ้ƒฝๆ˜ฏ measurable ็š„, ๅนถไธ”ๆˆ‘ไปฌๅฎนๆ˜“่ฏๆ˜Ž:

๐‘“๐‘›โ†—๏ธŽ๐‘“(๐‘ฆ)โ‰”๐œ‡((๐ธ)๐‘ฆ) ptwise.

ไบŽๆ˜ฏไฝฟ็”จ MCT, ๅฎนๆ˜“ๅพ—ๅˆฐ

โˆซ๐œ‡(๐ธ๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)=limโˆซ๐œ‡((๐ธ๐‘›)๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)=lim๐œ‡ร—๐œˆ(๐ธ๐‘›)=CFB๐œ‡ร—๐œˆ(๐ธ)

ไปŽ่€Œ ๐ธโˆˆ๐’ž๏ธ€.
It remains to show: ๐’ž๏ธ€ closed under ctbl decreasing intersection. ไธ่ฟ‡่ฟ™้‡Œๆˆ‘ไปฌๆถ‰ๅŠๅˆฐไธ€ไธช decreasing sequence ไธญ้—ด็ช็„ถไปŽ infinite measure ๅ˜ไธบ finite measure ็š„้—ฎ้ข˜, ๆ‰€ไปฅๆˆ‘ไปฌไปŽ่ฟ™้‡Œๅผ€ๅง‹่ฆๅˆ† ๐œ‡,๐œˆ finite ๅ’Œ not finite (but still ๐œŽ-finite) ็š„ไธค็งๆƒ…ๅ†ต่ฎจ่ฎบ. finite measure ไธ็”จๆ‹…ๅฟƒไธŠ่ฟฐ่ฟ™ไธ€้—ฎ้ข˜.
Case 1: ๐œ‡,๐œˆ finite, ไบŽๆ˜ฏไปค {๐ธ๐‘›} ไธบไธ€ไธช decreasing seq in ๐’ž๏ธ€, ๅ’Œ increasing ็š„ๆƒ…ๅ†ต similar, ๅพ—ๅˆฐ ๐œ‡((๐ธ๐‘›)๐‘ฆ)=:๐‘“๐‘›โ†˜๏ธŽ๐‘“(๐‘ฆ)โ‰”๐œ‡((๐ธ)๐‘ฆ) ptwise., ไปŽ่€Œ by DCT (ๅ– ๐œ‡(๐‘‹) ไธบ donimating function), ๅพ—ๅˆฐ

โˆซ๐œ‡(๐ธ๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)=limโˆซ๐œ‡((๐ธ๐‘›)๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)=lim๐œ‡ร—๐œˆ(๐ธ๐‘›)=CFA๐œ‡ร—๐œˆ(๐ธ)

ไปŽ่€Œๆˆ‘ไปฌ่ฏๆ˜Žไบ†ๅœจ ๐œ‡,๐œˆ ไธบ finite measure ็š„ๆƒ…ๅ†ตไธ‹, ๐’ž๏ธ€ ไธบไธ€ไธช monotone class, ไปŽ่€Œไธบไธ€ไธช ๐œŽ-algebra, ไปŽ่€Œ โ„ณ๏ธ€โŠ—๐’ฉ๏ธ€โŠ‚๐’ž๏ธ€.

Case 2: ๐œ‡,๐œˆ ๐œŽ-finite measure: ๆˆ‘ไปฌๅฏไปฅๆŠŠ ๐‘‹ร—๐‘Œ ๅ†™ไฝœ union of a seq of finite measure sets {๐‘‹๐‘–ร—๐‘Œ๐‘–}๐‘–โˆˆโ„•, ไปŽ่€ŒไนŸๆ˜ฏ a union of increaasing seq of finite measure sets {๐‘‹๐‘—ร—๐‘Œ๐‘—}๐‘—โˆˆโ„•. (ๅ– ๐‘‹๐‘—ร—๐‘Œ๐‘—=โ‹ƒ๐‘–=1๐‘—๐‘‹๐‘—ร—๐‘Œ๐‘—) ไปŽ่€ŒๅฏนไบŽไปปๆ„็š„ ๐ธโˆˆโ„ณ๏ธ€โŠ—๐’ฉ๏ธ€,

๐ธ=lim๐‘—โ†’โˆž(๐ธโˆฉ(๐‘‹๐‘—ร—๐‘Œ๐‘—))

ๅฏนไบŽๆฏไธช ๐ธโˆฉ(๐‘‹๐‘—ร—๐‘Œ๐‘—), ๆˆ‘ไปฌๅฏไปฅๅบ”็”จๅ‰ไธ€็ป“่ฎบ, ๅพ—ๅˆฐ

๐œ‡ร—๐œˆ(๐ธโˆฉ(๐‘‹๐‘—ร—๐‘Œ๐‘—))=โˆซ๐œ’๐‘Œ๐‘—(๐‘ฆ)๐œ‡(๐ธ๐‘ฆโˆฉ๐‘‹๐‘—)๐‘‘๐œˆ(๐‘ฆ)

ไปŽ่€Œๅบ”็”จ MCT, ๅพ—ๅˆฐ

๐œ‡ร—๐œˆ(๐ธ)=โˆซ๐œ‡(๐ธ๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)

ๅŒ็† ๐œ‡ร—๐œˆ(๐ธ)=โˆซ๐œ‡(๐ธ๐‘ฅ)๐‘‘๐œ‡(๐‘ฅ), ไปŽ่€Œ ๐ธโˆˆ๐’ž๏ธ€, ๅพ—่ฏ.

โ–ก

6.2.4 Tonelliโ€™s Theorem

Theorem 6.29 : Tonelli

Let (๐‘‹,๐’œ๏ธ€,๐œ‡), (๐‘Œ,โ„ฌ๏ธ€,๐œˆ) be ๐œŽ-finite measure spaces.
ๆกไปถ: ไปค ๐‘“โˆˆ๐ฟ+(๐‘‹ร—๐‘Œ),
็ป“่ฎบ:

๐‘”(๐‘ฅ)โ‰”โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐œˆโˆˆ๐ฟ+(๐‘‹)โ„Ž(๐‘ฆ)โ‰”โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐œ‡โˆˆ๐ฟ+(๐‘Œ)

(ๆ˜พ็„ถ) ๅนถไธ”

โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)=โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)]๐‘‘๐œ‡(๐‘ฅ)=โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐œ‡(๐‘ฅ)]๐‘‘๐œˆ(๐‘ฆ)
Proof

้ฆ–ๅ…ˆ, ๅฏนไบŽ ๐‘“ ๆ˜ฏ simple function ็š„ case, ็›ดๆŽฅ follows from Tonelli for sets. (mentioned in remark.)
ๅฏนไบŽ general case: ๐‘“โˆˆ๐ฟ+(๐‘‹ร—๐‘Œ), ไปค {๐‘“๐‘›} ไธบไธ€ไธช seq of simple functions ptwisely converging to ๐‘“.
ไบŽๆ˜ฏ

โˆซ๐‘”๐‘‘๐œ‡=limโˆซ๐‘”๐‘›๐‘‘๐œ‡=limโˆซ๐‘“๐‘›๐‘‘(๐œ‡ร—๐œˆ)=โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)โˆซโ„Ž๐‘‘๐œ‡=limโˆซโ„Ž๐‘›๐‘‘๐œ‡=limโˆซ๐‘“๐‘›๐‘‘(๐œ‡ร—๐œˆ)=โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)

by MCT.

โ–ก

6.3 Fubiniโ€™s Theorem and Lebesgue integral in โ„๐‘› [Fol 2.5, finished; 2.6]

recall Tonelliโ€™s Theorem: Given ๐‘“โˆˆ๐ฟ+(๐‘‹ร—๐‘Œ), set ๐‘”(๐‘ฅ)โ‰”โˆซ๐‘“๐‘ฅ๐‘‘๐œˆ, โ„Ž(๐‘ฆ)โ‰”โˆซ๐‘“๐‘ฆ๐‘‘๐œ‡. Then ๐‘”โˆˆ๐ฟ+(๐‘‹), โ„Žโˆˆ๐ฟ+(๐‘Œ), ไปฅๅŠๆœ‰:

โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)=โˆซ๐‘”๐‘‘๐œ‡=โˆซโ„Ž๐‘‘๐œˆ

ๅฑ•ๅผ€ๅŽๅฏๅ†™ไฝœ:

โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)=โˆฌ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐œˆ(๐‘ฆ)๐‘‘๐œ‡(๐‘ฅ)=โˆฌ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐œ‡(๐‘ฅ)๐‘‘๐œˆ(๐‘ฆ)

ๆ›ดๅŠ ็ฎ€ๆดๅฏๅ†™ไฝœ:

โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)=โˆฌ๐‘“๐‘‘๐œˆ๐‘‘๐œ‡=โˆฌ๐‘“๐‘‘๐œ‡๐‘‘๐œˆ
Corollary 6.19

if ๐‘“โˆˆ๐ฟ1(๐‘‹ร—๐‘Œ) and ๐‘“โ‰ฅ0 then

  • ๐‘”(๐‘ฅ)<โˆž for a.e. ๐‘ฅ

  • โ„Ž(๐‘ฆ)<โˆž for a.e. ๐‘ฆ

Next: Fubiniโ€™s Theorem.
Fubiniโ€™s Theorem ๆ˜ฏ Tonelliโ€™s Theorem ๅฏน โ„‚-valued ๅ‡ฝๆ•ฐ (instead of โ„โ‰ฅ0-valued) ็š„ๆŽจๅนฟ. ไฝ†ๆ˜ฏๅ…ถๅฎž่ฏๆ˜Žๅพˆ trivial.

6.3.1 Fubiniโ€™s Theorem

Theorem 6.30 : Fubiniโ€™s Theorem

ๆกไปถ: ๐‘“โˆˆ๐ฟ1(๐œ‡ร—๐œˆ),
็ป“่ฎบ:

  • ๐‘“๐‘ฅโˆˆ๐ฟ1(๐œˆ) for a.e. ๐‘ฅ, ๐‘“๐‘ฆโˆˆ๐ฟ1(๐œ‡) for a.e.

  • The a.e. defined functions:

    ๐‘”(๐‘ฅ)โ‰”โˆซ๐‘“๐‘ฅ๐‘‘๐œˆโˆˆ๐ฟ1(๐œ‡),โ„Ž(๐‘ฅ)โ‰”โˆซ๐‘“๐‘ฆ๐‘‘๐œˆโˆˆ๐ฟ1(๐œˆ)
  • โˆซ๐‘“๐‘‘(๐œ‡ร—๐œˆ)=โˆซ๐‘”๐‘‘๐œ‡=โˆซโ„Ž๐‘‘๐œˆ(=โˆฌ๐‘“๐‘‘๐œ‡๐‘‘๐œˆ)
Proof

๐‘“=โ„œ๐‘“+๐‘–โ„‘๐‘“, so WLOG can assume ๐‘“ is โ„-valued.
ๅˆ ๐‘“=๐‘“+โˆ’๐‘“โˆ’, ็›ดๆŽฅ apply Tonellisโ€™s Thm ๅฏๅพ—.

โ–ก

Example 6.18

ๆฑ‚ๅ’Œๆขๅบ็š„ๅˆ็†ๆ€ง:
่€ƒ่™‘

(๐‘‹,๐’œ๏ธ€,๐œ‡)=(๐‘Œ,โ„ฌ๏ธ€,๐œˆ)=(โ„•,๐’ซ๏ธ€(โ„•),๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘–๐‘›๐‘”)

if ๐‘Ž๐‘š๐‘›โˆˆโ„‚ for (๐‘š,๐‘›)โˆˆโ„•2 and

โˆž>โˆ‘๐‘š,๐‘›|๐‘Ž๐‘š๐‘›|=:sup๐นโŠ‚โ„•2finiteโˆ‘(๐‘š,๐‘›)โˆˆ๐น|๐‘Ž๐‘š,๐‘›|

Thm: ๅฏนไบŽไปปๆ„ ๐‘›โˆˆโ„•, โˆ‘๐‘š๐‘Ž๐‘š๐‘› conv absly to some ๐‘๐‘›โˆˆโ„‚;
ๅŒๆ ท, ๅฏนไบŽไปปๆ„ ๐‘šโˆˆโ„•, โˆ‘๐‘›๐‘Ž๐‘š๐‘› conv absly to ๐‘๐‘šโˆˆโ„‚. ไปฅๅŠ โˆ‘๐‘›๐‘๐‘›,โˆ‘๐‘š๐‘๐‘š conv absly to โˆ‘๐‘š,๐‘›๐‘Ž๐‘š๐‘›.
ๅณ:

โˆ‘๐‘›=1โˆžโˆ‘๐‘š=1โˆž|๐‘Ž๐‘š๐‘›|=โˆ‘๐‘š=1โˆžโˆ‘๐‘›=1โˆž|๐‘Ž๐‘š๐‘›|=โˆ‘(๐‘š,๐‘›)โˆˆโ„•2|๐‘Ž๐‘š๐‘›|

6.3.2 complete Fubiniโ€™s Theorem

Example 6.19

่€ƒ่™‘ (๐‘‹,๐’œ๏ธ€,๐œ‡)=(๐‘Œ,โ„ฌ๏ธ€,๐œˆ)=(โ„,โ„’๏ธ€,๐‘š) ่€ƒ่™‘ไธ€ไธช Vitali set.

๐‘‰ร—{0}โŠ‚โ„ร—{0} is a subnull set, not measurable

ไฝ†ๆ˜ฏๅฆ‚ๆžœๆˆ‘ไปฌ consider completion:

(๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€ฬ„,๐œ‡ร—๐œˆฬ„)
Theorem 6.31 : complete Fubini-Tonelli

ๅฏนไบŽ complete measure space (๐‘‹,๐’œ๏ธ€,๐œ‡), (๐‘Œ,โ„ฌ๏ธ€,๐œˆ), ๅ–ๅฎƒไปฌ็š„ product measure space ็š„ completion:

(๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€ฬ„,๐œ‡ร—๐œˆฬ„)

ๆˆ‘ไปฌๅฐ† ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€ฬ„ ็ฎ€ๆ˜“ๅ†™ไฝœ โ„’๏ธ€, ๐œ‡ร—๐œˆฬ„ ็ฎ€ๆ˜“ๅ†™ไฝœ ๐œ†.
Suppose ๐‘“:๐‘‹ร—๐‘Œโ†’โ„‚ is โ„’๏ธ€-measurable ๅนถไธ” ๐‘“โˆˆ๐ฟ+(๐œ†) or ๐‘“โˆˆ๐ฟ1(๐œ†), ๅˆ™ๆœ‰

  • ๐‘“๐‘ฅ ๆ˜ฏ โ„ฌ๏ธ€-measurable ็š„ for a.e. ๐‘ฅ ไธ” ๐‘ฅโ†ฆโˆซ๐‘“๐‘ฅ๐‘‘๐œˆ ๆ˜ฏ measurable ็š„

  • ๐‘“๐‘ฆ ๆ˜ฏ ๐’œ๏ธ€-measurable ็š„ for a.e. ๐‘ฆ ไธ” ๐‘ฆโ†ฆโˆซ๐‘“๐‘ฆ๐‘‘๐œ‡ ๆ˜ฏ measurable ็š„

ๅนถไธ”, ๅœจ ๐‘“โˆˆ๐ฟ1(๐œ†) ็š„ๆƒ…ๅ†ตไธ‹, ๐‘“๐‘ฅ,๐‘“๐‘ฆ, ๐‘ฅโ†ฆโˆซ๐‘“๐‘ฅ๐‘‘๐œˆ, ๐‘ฆโ†ฆโˆซ๐‘“๐‘ฆ๐‘‘๐œ‡ ไนŸๆ˜ฏ integrable ็š„, ๅณ โˆˆ๐ฟ1(๐œ†), ๅนถไธ”

โˆซ๐‘“๐‘‘๐œ†=โˆฌ๐‘“๐‘‘๐œ‡๐‘‘๐œˆ=โˆฌ๐‘“๐‘‘๐œˆ๐‘‘๐œ‡
Proof

exercise. ๆฏ”่พƒ็ฎ€ๅ•.

โ–ก

6.3.3 remark: integral of ้ž่ดŸๅ‡ฝๆ•ฐ็ญ‰ไบŽ area under graph

Theorem 6.32

ไปค (๐‘‹,๐’œ๏ธ€,๐œ‡) ไธบไธ€ไธช arbitrary measure space, ๐‘“โˆˆ๐ฟ+(๐œ‡) ไธบ arbitrary ๅฏๆต‹้ž่ดŸๅ‡ฝๆ•ฐ, ๆˆ‘ไปฌๅฎšไน‰:

๐บ๐‘“โ‰”{(๐‘ฅ,๐‘ฆ)โˆˆ๐‘‹ร—[0,โˆž]:0โ‰ค๐‘ฆโ‰ค๐‘“(๐‘ฅ)}

Claim: ๐บ๐‘“ ๆ˜ฏ (๐’œ๏ธ€ร—โ„ฌ๏ธ€(โ„))-measurable ็š„, ๅนถไธ”

(๐œ‡ร—๐‘š)(๐บ๐‘“)=โˆซ๐‘“๐‘‘๐œ‡
Proof

In hw 6.

โ–ก

Homework 6: on product measure and mode of convergence (49/50)

Some of the following questions will be graded. Do them, and do hand them in.

Order of integration: โˆซ0โˆžโˆซ๐‘ฅโˆž๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฆ๐‘‘๐‘ฅ=1

Use Tonelliโ€™s Theorem and 1-variable calculus to give a rigorous proof for the equality

โˆซ0โˆžโˆซ๐‘ฅโˆž๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฆ๐‘‘๐‘ฅ=1
Proof

Define

๐‘“(๐‘ฅ,๐‘ฆ)โ‰”{๐‘’โˆ’๐‘ฆ2/2,if 0โ‰ค๐‘ฅโ‰ค๐‘ฆ,0,otherwise.

Then we have

โˆซ0โˆžโˆซ๐‘ฅโˆž๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฆ๐‘‘๐‘ฅ=โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)]๐‘‘๐‘š(๐‘ฅ)

Since ๐‘“(๐‘ฅ,๐‘ฆ)=๐‘’โˆ’๐‘ฆ2/2 is nonnegative and continuous, it is measurable and thus in ๐ฟ+(๐‘‹ร—๐‘Œ), where ๐‘‹=๐‘Œ=(โ„,โ„’๏ธ€,๐‘š) is ๐œŽ-finite.
Thus we can apply Tonelliโ€™s theorem:

โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)]๐‘‘๐‘š(๐‘ฅ)=โˆซ๐‘“๐‘‘(๐‘š(๐‘ฅ)ร—๐‘š(๐‘ฆ))=โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฅ)]๐‘‘๐‘š(๐‘ฆ)=โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฅ)]๐‘‘๐‘š(๐‘ฆ)=โˆซ[โˆซ๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘š(๐‘ฅ)]๐‘‘๐‘š(๐‘ฆ)

Where

โˆซ๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘š(๐‘ฅ)=โˆซ[0,๐‘ฆ]๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฅ=๐‘ฆ๐‘’โˆ’๐‘ฆ2/2

Thus

โˆซ[โˆซ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)]๐‘‘๐‘š(๐‘ฅ)=โˆซ[โˆซ๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘š(๐‘ฅ)]๐‘‘๐‘š(๐‘ฆ)=โˆซ๐‘ฆ๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘š(๐‘ฆ)=โˆซ[0,โˆž)๐‘ฆ๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฆ

Make the substitution ๐‘ก=๐‘ฆ22, then we have

โˆซ0โˆž๐‘ฆ๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฆ=โˆซ0โˆž๐‘’โˆ’๐‘ก๐‘‘๐‘ก=[โˆ’๐‘’โˆ’๐‘ก]0โˆž=1

This finishes the proof that

โˆซ0โˆžโˆซ๐‘ฅโˆž๐‘’โˆ’๐‘ฆ2/2๐‘‘๐‘ฆ๐‘‘๐‘ฅ=1

โ–ก

integration of a function = Area under the curve

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a ๐œŽ-finite measure space, and let ๐‘“โˆˆ๐ฟ+(๐‘‹). Consider the subset ๐บ๐‘“โŠ‚๐‘‹ร—[0,โˆž) consisting of all points (๐‘ฅ,๐‘ฆ) with ๐‘ฆ<๐‘“(๐‘ฅ).

  • Prove that ๐บ๐‘“ is ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€โ„-measurable.

  • Prove that (๐œ‡โŠ—๐‘š)(๐บ๐‘“)=โˆซ๐‘“๐‘‘๐œ‡.

Proof

of 2(a):

๐‘ฆ<๐‘“(๐‘ฅ)โ‡”โˆƒ๐‘žโˆˆโ„š,๐‘ฆ<๐‘ž<๐‘“(๐‘ฅ)

Hence

๐บ๐‘“=โ‹ƒ๐‘žโˆˆโ„š,๐‘ž>0({๐‘ฅ:๐‘“(๐‘ฅ)>๐‘ž}ร—{๐‘ฆ:๐‘ฆ<๐‘ž})

Since {๐‘ฅ:๐‘“(๐‘ฅ)>๐‘ž}โˆˆ๐’œ๏ธ€ (by the measurability of ๐‘“) and {๐‘ฆ:๐‘ฆ<๐‘ž}โˆˆโ„ฌ๏ธ€โ„, each set in the union is a measurable rectangle, thus measurable in the product measurable space ๐‘‹ร—โ„. Since a countable union of measurable sets is measurable in the product ๐œŽ-algebra, We have

๐บ๐‘“โˆˆ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€โ„

โ–ก

Proof

of 2(b):
Since ๐‘“โ‰ฅ0, and ๐œŽ-finiteness of ๐‘‹ is assumed, ๐œŽ-finiteness of ๐‘Œ is known,
we can apply Tonelliโ€™s theorem to compute:

(๐œ‡โŠ—๐‘š)(๐บ๐‘“)=โˆซ๐‘‹ร—[0,โˆž)๐œ’๐บ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘(๐œ‡โŠ—๐‘š)=โˆซ๐‘‹[โˆซ[0,โˆž)๐œ’๐บ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)]๐‘‘๐œ‡(๐‘ฅ)

By definition of ๐บ๐‘“, ๐œ’๐บ๐‘“(๐‘ฅ,๐‘ฆ)=1 if and only if ๐‘ฆ<๐‘“(๐‘ฅ), and 0 otherwise. Hence, for each fixed ๐‘ฅ,

โˆซ[0,โˆž)๐œ’๐บ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)=โˆซ[0,โˆž)๐œ’{๐‘ฆ<๐‘“(๐‘ฅ)}๐‘‘๐‘š(๐‘ฆ)={๐‘“(๐‘ฅ),if ๐‘“(๐‘ฅ)<โˆž,โˆž,if ๐‘“(๐‘ฅ)=โˆž

Therefore

โˆซ[0,โˆž)๐œ’๐บ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)=๐‘“(๐‘ฅ)๐‘Ž.๐‘’.

Applying Tonelliโ€™s theorem again yields

(๐œ‡โŠ—๐‘š)(๐บ๐‘“)=โˆซ๐‘‹[โˆซ[0,โˆž)๐œ’๐บ๐‘“(๐‘ฅ,๐‘ฆ)๐‘‘๐‘š(๐‘ฆ)]๐‘‘๐œ‡(๐‘ฅ)=โˆซ๐‘‹๐‘“(๐‘ฅ)๐‘‘๐œ‡(๐‘ฅ)

Thus we conclude that

(๐œ‡โŠ—๐‘š)(๐บ๐‘“)=โˆซ๐‘‹๐‘“๐‘‘๐œ‡

โ–ก

Oscillations: ๐‘“๐‘›(๐‘ฅ)=(sin(๐œ‹๐‘›๐‘ฅ))๐‘›โ†’๐‘“=0 in measure

Consider the sequence ๐‘“๐‘›(๐‘ฅ)=(sin(๐œ‹๐‘›๐‘ฅ))๐‘›, ๐‘›=1,2,โ€ฆ, on the interval [0,1]. Prove that there exists a set ๐ธโŠ‚[0,1] such that ๐‘š(๐ธ๐‘)โ‰ค2โˆ’597 and a sequence 1โ‰ค๐‘›1<๐‘›2<โ€ฆ such that |๐‘“๐‘›๐‘—(๐‘ฅ)|โ‰ค๐‘—โˆ’597 for all ๐‘ฅโˆˆ๐ธ and all ๐‘—โ‰ฅ1. Hint: use E. Consider convergence in measure

Proof

Claim 1: It suffices to show that ๐‘“๐‘› converges in measure.
Proof of Claim 1: Suppose ๐‘“๐‘› converges in measure to ๐‘“=0, then by Folland 2.30, there exists a subseq (๐‘“๐‘›๐‘˜)โ†’๐‘˜โ†’โˆž๐‘“=0 a.e. .And since [0,1] has finite measure 1, by Egoroffโ€™s Theorem, for any ๐œ–>0 there exists ๐ธโŠ‚[0,1] s.t. ๐œ‡(๐ธ๐‘)<๐œ– and (๐‘“๐‘›๐‘˜)โ†’๐‘˜โ†’โˆž๐‘“=0 uniformly on ๐ธ.
Then we take ๐œ–:=2โˆ’597and coresponding ๐ธ.
And for each ๐‘—โˆˆโ„•, we let ๐›ฟ๐‘—=๐‘—โˆ’597. By the uniform convergence property of (๐‘“๐‘›๐‘˜), we can take ๐‘๐‘— s.t. |๐‘“๐‘›๐‘˜(๐‘ฅ)|<๐›ฟ๐‘— for all ๐‘ฅโˆˆ๐ธ whenever ๐‘›๐‘˜โ‰ฅ๐‘๐‘—.
Therefore, ๐ธ and the sequence (๐‘“๐‘๐‘—) satisfty the requirements in the context.
This shows that, as long as we can show (๐‘“๐‘›) converges in measure to ๐‘“=0, the statement is proved.

Let ๐‘“๐‘›(๐‘ฅ):=sin(๐‘›๐œ‹๐‘ฅ)๐‘› for ๐‘›โˆˆโ„•.
Claim 2: ๐‘“๐‘› converges in measure.
Proof of Claim 2: The idea is that the exponent ๐‘› makes the sequence converge faster than the linear growth of ๐‘›๐‘ฅ that shortens a period and messes up the sin values.
Fix ๐œ–>0. (WLOG ๐œ–<1.) WTS:

๐‘š({๐‘ฅ:|sin(๐‘›๐œ‹๐‘ฅ)โ‰ฅ๐œ–1/๐‘›})โ†’0 as ๐‘›โ†’โˆž

We know that sin(๐‘›๐œ‹๐‘ฅ)=1 iff ๐‘ฅ=2๐‘˜โˆ’12๐‘› for some ๐‘˜=0,โ‹ฏ,2๐‘›โˆ’1. Consider ๐‘ฅโˆˆ[0,12๐‘›), let |sin(๐‘›๐œ‹๐‘ฅ0)|:=๐œ–1/๐‘›.
Denote

๐›ฟ๐‘›โ‰”|12๐‘›โˆ’๐‘ฅ0|

Then we can express the measure as:

๐‘š({๐‘ฅ:|sin(๐‘›๐œ‹๐‘ฅ)โ‰ฅ๐œ–1/๐‘›})=2๐‘›๐›ฟ๐‘›

Notice that by the monotonicity of arcsin function, we can solve for ๐‘ฅ0 as:

๐‘ฅ0=1๐‘›๐œ‹arcsin(๐œ–1๐‘›)

Thus

๐›ฟ๐‘›=12๐‘›=1๐‘›๐œ‹arcsin(๐œ–1๐‘›)

Thus

lim๐‘›โ†’โˆž๐‘š({๐‘ฅ:|sin(๐‘›๐œ‹๐‘ฅ)โ‰ฅ๐œ–1/๐‘›})=lim๐‘›โ†’โˆž2๐‘›๐›ฟ๐‘›=1โˆ’lim๐‘›โ†’โˆž2๐œ‹arcsin(๐œ–1๐‘›)=1โˆ’2๐œ‹โ‹…๐œ‹2=0

Since ๐œ– is arbitrary, this finishes the proof that ๐‘“๐‘›โ†’๐‘“=0 in measure.
Thus combining Claim 1, the whole statement is proved.

โ–ก

Indicator functions ๆ˜ฏ ๐ฟ+ ็š„ไธ€ไธช closed subset

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be any measure space. Let ๐‘€โŠ‚๐ฟ+ be the set of indicator functions ๐œ’๐ธ, where ๐ธโˆˆ๐’œ๏ธ€ and ๐œ‡(๐ธ)<โˆž. Prove that ๐‘€ is a closed subset of ๐ฟ1. In other words, prove that ๐‘€โŠ‚๐ฟ1, and that if ๐‘“๐‘›โˆˆ๐‘€, ๐‘“โˆˆ๐ฟ1, and โˆซ|๐‘“๐‘›โˆ’๐‘“|โ†’0, then ๐‘“โˆˆ๐‘€.

Proof

Let (๐‘“๐‘›โ‰”๐œ’๐ธ๐‘›)๐‘›โˆˆโ„• be a seq of indicator functions in ๐ฟ+ s.t. โˆซ|๐‘“๐‘›โˆ’๐‘“|โ†’0 for some ๐‘“โˆˆ๐ฟ1.
Define for all ๐‘˜โˆˆโ„•

๐ด๐‘˜โ‰”{๐‘ฅ:|๐‘“(๐‘ฅ)|>1๐‘˜,|๐‘“(๐‘ฅ)โˆ’1|>1๐‘˜}

Fix one ๐‘˜โˆˆโ„•, bt monotonicity of integration in ๐ฟ1, we have

โˆซ|๐‘“โˆ’๐œ’๐ธ๐‘›|โ‰ฅโˆซ๐ด๐‘˜|๐‘“โˆ’๐œ’๐ธ๐‘›|โ‰ฅโˆซ๐ด๐‘˜1๐‘˜โ‰ฅ๐œ‡(๐ด๐‘˜)๐‘˜

Thus

๐œ‡(๐ด๐‘˜)โ‰ค๐‘˜โˆซ|๐‘“โˆ’๐œ’๐ธ๐‘›|

Since ๐œ’๐ธ๐‘›โ†’๐‘“ in ๐ฟ1, it follows that ๐œ‡(๐ด๐‘˜)=0.
Since ๐ด๐‘˜ is arbitrary, by ctbl sub additivity,

๐œ‡(โ‹ƒ๐‘˜=1โˆž๐ด๐‘˜)โ‰คโˆ‘๐‘˜=1โˆž๐œ‡(๐ด๐‘˜)=0

Define

๐ดโ‰”{๐‘ฅ:๐‘“(๐‘ฅ)โ‰ 0,1}

By the definition of ๐ด๐‘˜, we have the equality:

๐ด=โ‹ƒ๐‘˜=1โˆž๐ด๐‘˜

Thus ๐œ‡(๐ด)=0, which means that ๐‘“(๐‘ฅ)โˆˆ{0,1} a.e., showing that ๐‘“ is a.e. an indicator function, in the same equivalence class of some indicator function in ๐ฟ1, thus we have ๐‘“โˆˆ๐‘€โŠ‚๐ฟ1. This finishes the proof that ๐‘€ is a closed subset of ๐ฟ1.

โ–ก

a complete metric space of measurable functions (other then ๐ฟ1(๐œ‡))

Suppose that (๐‘‹,๐’œ๏ธ€,๐œ‡) is a measure space such that ๐œ‡(๐‘‹)<โˆž. Set ๐œ’(๐‘ก)=๐‘ก1+๐‘ก for ๐‘กโ‰ฅ0.
Given measurable functions ๐‘“,๐‘”:๐‘‹โ†’โ„‚, set

๐œŒ(๐‘“,๐‘”)โ‰”โˆซ๐œ’(|๐‘“โˆ’๐‘”|)๐‘‘๐œ‡
  • Prove that ๐œŒ induces a metric, also denoted ๐œŒ, on the space

    ๐ฟโ‰”{๐‘“:๐‘‹โ†’โ„‚measurable}/โˆผ,

    where ๐‘“โˆผ๐‘” iff ๐‘“=๐‘” a.e. Hint: prove that ๐œ’(๐‘ +๐‘ก)โ‰ค๐œ’(๐‘ )+๐œ’(๐‘ก) for ๐‘ ,๐‘กโ‰ฅ0.

  • Prove that if ๐‘“๐‘›,๐‘“โˆˆ๐ฟ, then ๐œŒ(๐‘“๐‘›,๐‘“)โ†’0 iff ๐‘“๐‘›โ†’๐‘“ in measure.

  • Prove that (๐ฟ,๐œŒ) is a complete metric space.

Proof

of 5(a): ๐œ’(๐‘ก)=๐‘ก1+๐‘ก=1โˆ’11+๐‘ก is an increasing function on ๐‘กโ‰ฅ0.
Claim: for all ๐‘ ,๐‘กโ‰ฅ0, we have ๐œ’(๐‘ )+๐œ’(๐‘ก)โ‰ค๐œ’(๐‘ +๐‘ก).
Proof of claim:
Let ๐‘ ,๐‘กโ‰ฅ0, we have

๐œ’(๐‘ )+๐œ’(๐‘ก)=๐‘ 1+๐‘ +๐‘ก1+๐‘ก=๐‘ (1+๐‘ก)+๐‘ก(1+๐‘ )(1+๐‘ )(1+๐‘ก)=๐‘ +๐‘ ๐‘ก+๐‘ก+๐‘ก๐‘ (1+๐‘ )(1+๐‘ก)=๐‘ +๐‘ก+2๐‘ ๐‘ก(1+๐‘ )(1+๐‘ก)

while

๐œ’(๐‘ +๐‘ก)=๐‘ +๐‘ก1+๐‘ +๐‘ก

Note

(๐‘ +๐‘ก)(1+๐‘ )(1+๐‘ก)=(๐‘ +๐‘ก)(1+๐‘ +๐‘ก+๐‘ ๐‘ก)=๐‘ +๐‘ก+๐‘ 2+2๐‘ ๐‘ก+๐‘ก2+๐‘ 2๐‘ก+๐‘ ๐‘ก2(๐‘ +๐‘ก+2๐‘ ๐‘ก)(1+๐‘ +๐‘ก)=๐‘ +๐‘ก+๐‘ 2+4๐‘ ๐‘ก+๐‘ก2+2๐‘ 2๐‘ก+2๐‘ ๐‘ก2

We have:

(๐‘ +๐‘ก)(1+๐‘ )(1+๐‘ก)โ‰ค(๐‘ +๐‘ก+2๐‘ ๐‘ก)(1+๐‘ +๐‘ก)

Since (1+๐‘ +๐‘ก) and (1+๐‘ )(1+๐‘ก) are positive, we can rearrange the ineq to be

๐‘ +๐‘ก1+๐‘ +๐‘กโ‰ค๐‘ +๐‘ก+2๐‘ ๐‘ก(1+๐‘ )(1+๐‘ก)

which is exactly

๐œ’(๐‘ )+๐œ’(๐‘ก)โ‰ค๐œ’(๐‘ +๐‘ก)

as needed.

First, ๐œŒ is a well-defined function on the quotient set, since if ๐‘“โˆผ๐‘” and ๐‘“โ€ฒโˆผ๐‘”โ€ฒ then |๐‘“โˆ’๐‘”|=|๐‘“โ€ฒโˆ’๐‘”โ€ฒ| a.e. Consequently,

๐œ’(|๐‘“โˆ’๐‘”|)=๐œ’(|๐‘“โ€ฒโˆ’๐‘”โ€ฒ|)a.e.

and hence

โˆซ๐‘‹๐œ’(|๐‘“โˆ’๐‘”|)๐‘‘๐œ‡=โˆซ๐‘‹๐œ’(|๐‘“โ€ฒโˆ’๐‘”โ€ฒ|)๐‘‘๐œ‡

Now we prove that ๐œŒ is a metric:

  • Nonnegativity: ๐œŒ(๐‘“,๐‘”)โ‰ฅ0 is immediate since ๐œ’(โ‹…)โ‰ฅ0 and ๐œ‡ is a measure; and since ๐œ’(โ„Ž)=0 iff โ„Ž=0 a.e., we have ๐œŒ(๐‘“,๐‘”)=0 iff ๐‘“=๐‘” a.e., that is, ๐‘“=๐‘”โˆˆ๐ฟ1(๐œ‡)

  • Symmetry: ๐œŒ(๐‘“,๐‘”)=๐œŒ(๐‘”,๐‘“) follows immediately from ๐œ’(|๐‘“โˆ’๐‘”|)=๐œ’(|๐‘”โˆ’๐‘“|).

  • Triangle inequality: For any three functions ๐‘“,๐‘”,โ„Ž, we have pointwise

    |๐‘“(๐‘ฅ)โˆ’โ„Ž(๐‘ฅ)|โ‰ค|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|+|๐‘”(๐‘ฅ)โˆ’โ„Ž(๐‘ฅ)|.

    Then applying the subadditivity of ๐œ’ proved above, we have:

    ๐œ’(|๐‘“(๐‘ฅ)โˆ’โ„Ž(๐‘ฅ)|)โ‰ค๐œ’(|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|+|๐‘”(๐‘ฅ)โˆ’โ„Ž(๐‘ฅ)|)โ‰ค๐œ’(|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|)+๐œ’(|๐‘”(๐‘ฅ)โˆ’โ„Ž(๐‘ฅ)|)

    Integrating both sides over ๐‘‹ gives

    ๐œŒ(๐‘“,โ„Ž)=โˆซ๐‘‹๐œ’(|๐‘“โˆ’โ„Ž|)๐‘‘๐œ‡โ‰คโˆซ๐‘‹๐œ’(|๐‘“โˆ’๐‘”|)๐‘‘๐œ‡+โˆซ๐‘‹๐œ’(|๐‘”โˆ’โ„Ž|)๐‘‘๐œ‡=๐œŒ(๐‘“,๐‘”)+๐œŒ(๐‘”,โ„Ž)

Therefore, ๐œŒ is a metric on ๐ฟ={๐‘“:๐‘‹โ†’โ„‚ measurable}/โˆผ as desired.

โ–ก

Proof

of 5(b):
Claim 1: ๐œŒ(๐‘“๐‘›,๐‘“)โ†’0 โŸน ๐‘“๐‘›โ†’๐‘“ in measure
Suppose ๐œŒ(๐‘“๐‘›,๐‘“)โ†’0. Let ๐œ–>0.
Since ๐œ’(๐‘ก)=๐‘ก1+๐‘ก is strictly increasing in ๐‘ก:

|๐‘“๐‘›โˆ’๐‘“|>๐œ–โ‡”๐œ’(|๐‘“๐‘›โˆ’๐‘“|)>๐œ’(๐œ–)=๐œ–1+๐œ–

Hence

{|๐‘“๐‘›โˆ’๐‘“|>๐œ–}={๐œ’(|๐‘“๐‘›โˆ’๐‘“|)>๐œ–1+๐œ–}

Since the function is nonnegative, by Chebyshev:

๐œ‡({|๐‘“๐‘›โˆ’๐‘“|>๐œ–})=๐œ‡({๐œ’(|๐‘“๐‘›โˆ’๐‘“|)>๐œ–1+๐œ–})โ‰ค1๐œ–1+๐œ–โˆซ๐œ’(|๐‘“๐‘›โˆ’๐‘“|)๐‘‘๐œ‡=๐œŒ(๐‘“๐‘›,๐‘“)๐œ’(๐œ–)

By assumption, ๐œŒ(๐‘“๐‘›,๐‘“)โ†’0, thus

๐œ‡({|๐‘“๐‘›โˆ’๐‘“|>๐œ–})โ‰ค๐œŒ(๐‘“๐‘›,๐‘“)๐œ’(๐œ–)โ†’0

Since ๐œ– is arbitrary, it proves that ๐‘“๐‘›โ†’๐‘“ in measure.

Claim 2: ๐‘“๐‘›โ†’๐‘“ in measure โŸน ๐œŒ(๐‘“๐‘›,๐‘“)โ†’0
Now assume ๐‘“๐‘›โ†’๐‘“ in measure.
Let ๐›ฟ>0.
Observe that for any ๐œ–>0:

  • |๐‘“๐‘›โˆ’๐‘“|โ‰ค๐œ–โŸน|๐‘“๐‘›โˆ’๐‘“|1+|๐‘“๐‘›โˆ’๐‘“|โ‰ค๐œ–1+๐œ–.

  • |๐‘“๐‘›โˆ’๐‘“|โ‰ฅ๐œ–โŸน|๐‘“๐‘›โˆ’๐‘“|1+|๐‘“๐‘›โˆ’๐‘“|โ‰ค1

Hence by choosing any arbitrary ๐œ–, we can bound the integral by:

0โ‰คโˆซ๐‘‹|๐‘“๐‘›โˆ’๐‘“|1+|๐‘“๐‘›โˆ’๐‘“|๐‘‘๐œ‡โ‰คโˆซ{|๐‘“๐‘›โˆ’๐‘“|โ‰ค๐œ–}๐œ–1+๐œ–๐‘‘๐œ‡+โˆซ{|๐‘“๐‘›โˆ’๐‘“|>๐œ–}1๐‘‘๐œ‡

For the first term:

โˆซ{|๐‘“๐‘›โˆ’๐‘“|โ‰ค๐œ–}๐œ–1+๐œ–๐‘‘๐œ‡=๐œ–1+๐œ–๐œ‡({|๐‘“๐‘›โˆ’๐‘“|โ‰ค๐œ–})โ‰ค๐œ–1+๐œ–๐œ‡(๐‘‹)

Because ๐œ‡(๐‘‹) is finite, we can choose ๐œ– s.t. ๐œ–1+๐œ–๐œ‡(๐‘‹)<๐›ฟ/2.
Once ๐œ– is fixed, by convergence in measure there exists ๐‘ such that for all ๐‘›โ‰ฅ๐‘,

๐œ‡({|๐‘“๐‘›โˆ’๐‘“|>๐œ–})<๐›ฟ/2

Then for any ๐‘›โ‰ฅ๐‘, we have:

๐œŒ(๐‘“๐‘›,๐‘“)=โˆซ๐‘‹๐œ’(|๐‘“๐‘›โˆ’๐‘“|)๐‘‘๐œ‡โ‰ค๐œ‡(๐‘‹)๐œ–1+๐œ–+๐œ‡({|๐‘“๐‘›โˆ’๐‘“|>๐œ–})<๐›ฟ

Hence

๐œŒ(๐‘“๐‘›,๐‘“)โ†’๐‘›โ†’โˆž0

โ–ก

Proof

of 5(c):
Suppose (๐‘“๐‘›) is a Cauchy seq in (๐ฟ,๐œŒ), i.e. for any ๐œ–>0, exists some ๐‘>0 s.t. ๐œŒ(๐‘“๐‘š,๐‘“๐‘›)<๐œ– whenever ๐‘›,๐‘šโ‰ฅ๐‘.
WTS: (๐‘“๐‘›) converges, i.e. ๐œŒ(๐‘“๐‘›,๐‘“)โ†’0.
By (b) we know it suffices to show that ๐‘“๐‘›โ†’๐‘“ in measure.
And by Folland 2.30, STS: (๐‘“๐‘›) is Cachy in measure.
Let ๐œ–>0. Let ๐›ฟ>0.
by Chebyshev:

๐œ‡({|๐‘“๐‘›โˆ’๐‘“๐‘š|>๐œ–})=๐œ‡({๐œ’(|๐‘“๐‘›โˆ’๐‘“๐‘š|)>๐œ–1+๐œ–})โ‰ค1๐œ–1+๐œ–โˆซ๐œ’(|๐‘“๐‘›โˆ’๐‘“๐‘š|)๐‘‘๐œ‡=๐œŒ(๐‘“๐‘›,๐‘“๐‘š)๐œ’(๐œ–)

So since (๐‘“๐‘›) is a Cauchy, there exists ๐‘>0 s.t. ๐œŒ(๐‘“๐‘›,๐‘“๐‘š)<๐œ’(๐œ–)๐›ฟ whenever ๐‘›,๐‘šโ‰ฅ๐‘, thus ๐œ‡({|๐‘“๐‘›โˆ’๐‘“๐‘š|>๐œ–})โ‰ค๐›ฟ whenever ๐‘š,๐‘›โ‰ฅ๐‘.
This proves that (๐‘“๐‘›) is Cachy in measure, thus ๐‘“๐‘›โ†’๐‘“ in measure, and thus (๐‘“๐‘›) converges, showing that every Cachy seq converges in (๐ฟ,๐œŒ). Therefore (๐ฟ,๐œŒ) is a complete metric space.

โ–ก

Nur fรผr Verrรผckte (Only for nuts).

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!)

  1. Prove that the category of measurable spaces (see HW1) admits finite products, and that the product of (๐‘‹,๐’œ๏ธ€) and (๐‘Œ,โ„ฌ๏ธ€) equals (๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€).

  2. Now consider the category of measure spaces (see HW2). Consider two measure spaces (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–), ๐‘–=1,2, and set ๐‘‹=๐‘‹1ร—๐‘‹2, ๐’œ๏ธ€=๐’œ๏ธ€1โŠ—๐’œ๏ธ€2, and ๐œ‡=๐œ‡1ร—๐œ‡2.

    • Prove that the projection maps ๐‘‹โ†’๐‘‹๐‘– are measurable, and that they are measure preserving iff ๐œ‡๐‘—(๐‘‹๐‘—)=1 for ๐‘—=1,2. Thus (๐‘‹,๐’œ๏ธ€,๐œ‡) is not the categorical product of (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–) in general.

    • Prove that even if ๐œ‡๐‘–(๐‘‹๐‘–)=1, the measure space (๐‘‹,๐’œ๏ธ€,๐œ‡) is not the categorical product of (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–) in general. Hint: consider the case when the ๐‘‹๐‘– consist of two elements, for example ๐‘‹๐‘–={๐”ฌ๐‘–,๐”ณ๐‘–}.

7 Lebesgue measure on โ„๐‘›

7.1 Lebesgue measure in โ„๐‘› [Fol 2.6]

ไปŠๆ—ฅ: Lebesgue measure in โ„๐‘› ็š„

  • regularity

  • behavior under affine transformation

  • behavior under diffeomorphism

7.1.1 Lebesgue measure in โ„๐‘›

่ฟ™ๆ˜ฏ product measure ๆœ€ๅธธ่ง็š„ๅบ”็”จๅ’Œไพ‹ๅญ.

Definition 7.36

(โ„๐‘›,โ„’๏ธ€๐‘›,๐‘š) Lebesgue measure is completion of (โ„๐‘›,โ„ฌ๏ธ€โ„๐‘›,๐‘š|๐‘๐‘œ๐‘Ÿ๐‘’๐‘™).

where โ„ฌ๏ธ€โ„๐‘›=โ„ฌ๏ธ€โ„โŠ—โ‹ฏโŠ—โ„ฌ๏ธ€โ„ โ„’๏ธ€๐“ƒ๏ธ€={Leb meas sets}โŠƒโ„ฌ๏ธ€โ„๐‘› Write:

โˆซ๐‘“๐‘‘๐‘š๐‘›
Theorem 7.33 : Fubini-Tonelli for ๐‘š๐‘›

Suppose ๐‘“โˆˆ๐ฟ+(โ„๐‘›) or ๐ฟ1(โ„๐‘›)

โˆซ๐‘“๐‘‘๐‘š๐‘›=โˆซโ‹ฏโˆซ๐‘“(๐‘ฅ1,โ‹ฏ,๐‘ฅ๐‘›)๐‘‘๐‘ฅ1โ‹ฏ๐‘‘๐‘ฅ๐‘›=โˆซโ‹ฏโˆซ๐‘“(๐‘ฅ1,โ‹ฏ,๐‘ฅ๐‘›)๐‘‘๐‘ฅ๐‘›โ‹ฏ๐‘‘๐‘ฅ1
Example 7.20

Show:

โˆซ0โˆž๐‘’โˆ’๐‘ ๐‘ฅsin2(๐‘ฅ)๐‘ฅ๐‘‘๐‘ฅ=14log(1+4๐‘ โˆ’2)

for ๐‘ >0, by integrating ๐‘’โˆ’๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฆ=๐‘“(๐‘ฅ,๐‘ฆ) over the rectangle ๐‘ฅโˆˆ(0,โˆž),๐‘ฆโˆˆ(0,1).
Sketch: ๐‘“โˆˆ๐ฟ1 (since it is ctn on โ„) ไปฅๅŠ

|๐‘“|โ‰ค๐‘’โˆ’๐‘ ๐‘ฅ,โˆซโ„๐‘’โˆ’๐‘ ๐‘ฅ<โˆž

ๅฏ่ฎก็ฎ—ๅพ—

โˆซ01sin2๐‘ฅ๐‘ฆ๐‘‘๐‘ฆ=12๐‘ฅsin2๐‘ฅ

่€ŒๅŽ compute

โˆซ01๐‘’โˆ’๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฆ๐‘‘๐‘ฆ

by integration by part for twice.

7.1.2 regularities of Lebesgue measure in โ„๐‘›

Theorem 7.34 : regularities of โ„’๏ธ€๐‘›

If ๐ธโŠ‚โ„’๏ธ€๐‘›, ๅˆ™ๆœ‰:

  • outer regularity:

    ๐‘š(๐ธ)=inf{๐‘š(๐‘ˆ)โˆฃ๐‘ˆ openโŠƒ๐ธ}
  • inner regularity:

    ๐‘š(๐ธ)=sup{๐‘š(๐พ)โˆฃ๐พ compactโŠ‚๐ธ}
  • if ๐‘š(๐ธ)<โˆž, ๅˆ™ๅฏนไบŽไปปๆ„ ๐œ–>0, ้ƒฝๅญ˜ๅœจ disjoint rectangles ๐‘…1,โ‹ฏ๐‘…๐‘ with sides that are open intervals (literally rectangles) s.t.

    ๐‘š(๐ธฮ”โ‹ƒ๐‘—๐‘…๐‘—)<๐œ–
Proof

for (a,b) i.e. regularities:
Fix ๐œ–>0. By construction, ๅญ˜ๅœจ finite disjoint union of rectangle ๐‘‡๐‘— for each ๐‘—, ไฝฟๅพ—

๐ธโŠ‚โ‹ƒ๐‘—=1โˆž๐‘‡๐‘— and โˆ‘๐‘—=1โˆž๐‘š(๐‘‡๐‘—)โ‰ค๐‘š(๐ธ)+๐œ–

By outer regularity of ๐‘š1, ๅญ˜ๅœจ ๐‘ˆ๐‘—โŠƒ๐‘‡๐‘— open rect s.t. ๐‘š(๐‘ˆ๐‘—)โ‰ค๐‘š(๐‘‡๐‘—)+๐œ–/2๐‘— Then:

๐ธโŠ‚๐‘ˆโ‰”โ‹ƒ๐‘—=1โˆž๐‘ˆ๐‘—and๐‘š(๐‘ˆ)โ‰คโˆ‘๐‘—=1โˆž๐‘š(๐‘ˆ๐‘—)

Construct ๐พ as in dim 1 (DIY) โ‰ค๐‘š(๐ธ)+2๐œ–.
(ๅฎŒๆ•ด Pf ๅฏ่ง 395 ็ฌ”่ฎฐ, ๆญค็•ฅ)

โ–ก

Proof

for (c):
Notation as above.

๐‘š(๐ธ)<โˆžโŸน๐‘š(๐‘ˆ)<โˆžโŸน๐‘š(๐‘ˆ๐‘—)<โˆžโˆ€๐‘—

Sides of ๐‘ˆ๐‘— are disjoint union of ctbly many open finite intervals.
ๅ› ่€Œๅญ˜ๅœจ open rectangle ๐‘‰๐‘—โŠ‚๐‘ˆ๐‘— for each ๐‘— that are finite disjoint union of finite open intervals s.t.

๐‘š(๐‘ˆ๐‘—\๐‘‰๐‘—)<๐œ–/2๐‘—

Now pick ๐‘…1,โ‹ฏ,๐‘…๐‘ from honest rectangles (ๅณ sides ้ƒฝๆ˜ฏ intervals ็š„ rectangle) insides ๐‘‰๐‘— (DIY). (ๅฎŒๆ•ด Pf ๅฏ่ง 395 ็ฌ”่ฎฐ, ๆญค็•ฅ)

โ–ก

Corollary 7.20

For ๐‘“โˆˆ๐ฟ1(๐‘š), if ๐‘“โˆˆ๐ฟ1(๐‘š) and ๐œ–>0 then

  • ๅฏนไบŽไปปๆ„ ๐œ–>0, ้ƒฝๅญ˜ๅœจ ๐œ™=โˆ‘๐‘—=1๐‘๐‘๐‘—๐œ’๐‘…๐‘— s.t.

    โˆซ|๐œ™โˆ’๐‘“|๐‘‘๐‘š<๐œ–

    ๅ…ถไธญ each ๐‘๐‘—โˆˆโ„‚, ๐‘…๐‘— ๆ˜ฏ rectangles with sides as finite open intervals.

  • ๅญ˜ๅœจ ๐œ™โˆˆ๐ถ๐‘0(โ„๐‘›) s.t.

    โˆซ|๐‘“โˆ’๐œ™|๐‘‘๐‘š<๐œ–
Proof

Similar to 1 dim case, ๅฏไปฅ่ฏๆ˜Ž {all step functions}, ๐ถ๐‘0(โ„๐‘›) ๆ˜ฏ dense subspace of ๐ฟ1(๐‘š).

โ–ก

7.1.3 approximating an open set ๐ธโŠ‚โ„๐‘› by countable disjoint interior cubes

ๅฏนไบŽ ๐‘˜โˆˆโ„ค, ไปค ๐’ฌ๏ธ€๐‘˜ be the collection of cubes whose side length is 12๐‘˜ ไธ” vertices ๅœจ lattice (2โˆ’๐‘˜โ„ค)๐‘› ไธญ, ๅณ็ฒพ็ป†ๅบฆไธบ 12๐‘˜ ็š„็ฝ‘ๆ ผไธญ็š„ๆ‰€ๆœ‰ cubes.

Figureย 20:

ๅฏนไบŽ ๐ธโŠ‚โ„๐‘›, ๆˆ‘ไปฌๅฎšไน‰:

๐ดยฏ(๐ธ,๐‘˜)โ‰”โ‹ƒ{๐‘„โˆˆ๐’ฌ๏ธ€๐“€๏ธ€:๐‘„โŠ‚๐ธ},๐ดฬ„(๐ธ,๐‘˜)โ‰”โ‹ƒ{๐‘„โˆˆ๐’ฌ๏ธ€๐“€๏ธ€:๐‘„โˆฉ๐ธโ‰ โŒ€}

ๅณ, ไธ€ไธชๆ˜ฏ่ขซๅŒ…ๅซๅœจ ๐ธ ไธญ็š„ๆ‰€ๆœ‰ๆ ผๅญ, ไธ€ไธชๆ˜ฏๆœ€ๅฐ็š„่ฆ†็›– ๐ธ ็š„ๆ‰€ๆœ‰ๆ ผๅญ. ๅนถๅฎšไน‰:

๐ดยฏ(๐ธ):=โ‹ƒ๐‘˜=1โˆž๐ดยฏ(๐ธ,๐‘˜),๐ดฬ„(๐ธ):=โ‹ƒ๐‘˜=1โˆž๐ดฬ„(๐ธ,๐‘˜)

ไปฅๅŠ

๐œ…ฬ„(๐ธ)โ‰”lim๐‘˜โ†’โˆž๐‘š(๐ดยฏ(๐ธ,๐‘˜)),๐œ…ยฏ(๐ธ)โ‰”lim๐‘˜โ†’โˆž๐‘š(๐ดฬ„(๐ธ,๐‘˜))

By CFB, CFA ๅฎนๆ˜“ๅพ—ๅˆฐ:

๐œ…ฬ„(๐ธ)=๐‘š(๐ดฬ„(๐ธ)),๐œ…ยฏ(๐ธ)=๐‘š(๐ดยฏ(๐ธ))

Note: ่ฟ™้‡Œ็š„ ๐ดยฏ(๐ธ,๐‘˜),๐ดฬ„(๐ธ,๐‘˜),๐ดยฏ(๐ธ),๐ดฬ„(๐ธ) ้ƒฝๆ˜ฏ union of cubes with disjoint interiors.

Lemma 7.23 : approximate an open set by disjoint interior cubes

Let ๐ธโŠ‚โ„๐‘› be open.
Claim: ๐ธ=๐ดยฏ(๐ธ)

Proof

Folland 2.43.

โ–ก

Corollary 7.21

๐ธโŠ‚โ„๐‘› ๆ˜ฏ Lebesuge measurable ็š„ โ‡” ๐œ…ฬ„(๐ธ)=๐œ…ยฏ(๐ธ)

7.1.4 behavior under affine transformation

Affine transformation ๅณ linear transformation + translation.

7.1.5 Lebesgue measure and integral is invariant under translation

ๅฏนไบŽ ๐‘Žโˆˆโ„๐‘›, ไธ€ไธช translation ๐‘ก:โ„๐‘›โ†’โ„๐‘›,๐‘ฅโ†ฆ๐‘ฅ+๐‘Ž ๆ˜ฏ ctn ็š„ๅนถไธ”

๐‘ก๐‘Žโˆ’1=๐‘กโˆ’๐‘Ž
Theorem 7.35 : Lebesgue measure and integral is invariant under translation

(a) ไปปๅ– ๐‘Žโˆˆโ„๐‘›,

๐ธโˆˆโ„’๏ธ€๐‘›โŸน๐‘ก๐‘Ž(๐ธ)โˆˆโ„’๏ธ€๐‘› and ๐‘š(๐‘ก๐‘Ž(๐ธ))=๐‘š(๐ธ)

(b) if ๐‘“:โ„๐‘›โ†’โ„‚ is Leb measurable, then so is ๐‘“โˆ˜๐‘ก๐‘Ž.
More, if ๐‘“โˆˆ๐ฟ+ or ๐‘“โˆˆ๐ฟ1, then ๐‘“โˆ˜๐‘ก๐‘Žโˆˆ๐ฟ1 ๅนถไธ”

โˆซ(๐‘“โˆ˜๐‘ก๐‘Ž)๐‘‘๐‘š=โˆซ๐‘“๐‘‘๐‘š
Proof

(Folland 2.42)
(a) ๐‘ก๐‘Ž ctn โŸน ๐‘ก๐‘Ž(โ„ฌ๏ธ€โ„๐‘›)โŠ‚โ„ฌ๏ธ€โ„๐‘›, ๅ› ่€Œ ๐‘ก๐‘Ž(โ„ฌ๏ธ€โ„๐‘›)=โ„ฌ๏ธ€โ„๐‘› ๐ธ rectangle, so ๐ธ=๐ธ1ร—โ‹ฏร—๐ธ๐‘›, each in โ„ฌ๏ธ€โ„ ๐‘š(๐ธ)=โˆ1๐‘›๐‘š(๐ธ๐‘–), ๐‘ก๐‘Ž(๐ธ)=โˆ๐‘ก๐‘Ž๐‘–(๐ธ๐‘–) ๅ› ่€Œ

๐‘š(๐‘ก๐‘Ž(๐ธ))=โˆ๐‘š(๐‘ก๐‘Ž๐‘–(๐ธ๐‘–))=โˆ๐‘š(๐ธ๐‘–)โŠ‚๐‘š(๐ธ)

BY HK uniqueness, get

๐‘š(๐‘ก๐‘Ž(๐ธ))=๐‘š(๐ธ)โˆ€๐ธโˆˆโ„ฌ๏ธ€โ„๐‘›

if ๐‘โŠ‚โ„๐‘› subnull set, so is ๐‘ก๐‘Ž(๐‘). ๅ› ่€Œ

๐‘š(๐‘ก๐‘Ž(๐ธ))=๐‘š(๐ธ)โˆ€๐ธโˆˆโ„’๏ธ€๐‘›

(b) Pick ๐ตโˆˆโ„ฌ๏ธ€โ„‚โŸน๐‘“โˆ’1(๐ต)โˆˆโ„’๏ธ€. ๅ› ่€Œ ๐‘“โˆ’1(๐ต)=๐ธโˆช๐‘, ๐ธโˆˆโ„ฌ๏ธ€โ„๐‘›, ๐‘ null set ๅ› ่€Œ

(๐‘“โˆ˜๐‘ก๐‘Ž)โˆ’1(๐ต)=๐‘ก๐‘Žโˆ’1(๐‘“โˆ’1(๐ต))=๐‘ก๐‘Žโˆ’1(๐ธ)โˆช๐‘ก๐‘Žโˆ’1(๐‘) (one Borel, one null)=๐‘กโˆ’๐‘Ž(๐‘“โˆ’1(๐ต))

ๅฝ“ ๐‘“=๐œ’๐ธ ๆ—ถ, ็งฏๅˆ† reduce to measure, ๅณ (a); ๅ› ่€Œ

โˆซ(๐‘“โˆ˜๐‘ก๐‘Ž)๐‘‘๐‘š=โˆซ๐‘“๐‘‘๐‘š

also holds for simple ๐‘“, by linearity.
ไปŽ่€Œ by def, ไนŸ hold for ๐‘“โˆˆ๐ฟ+ ๅ’Œ ๐‘“โˆˆ๐ฟ1.

โ–ก

7.1.6 Lebesgue measure and integration is scaled |det๐‘‡| under linear map

Theorem 7.36 : Lebesgue measure and integration is scaled |det๐‘‡| by linear map

For ๐‘‡โˆˆ๐บ๐ฟ(๐‘›,โ„) (ๅณ linear map ๐‘‡:โ„๐‘›โ†’โ„๐‘› ไธ”ๅฏ้€†) (a) ๅฆ‚ๆžœ ๐‘“:โ„๐‘›โ†’โ„‚ is Lebesgue measurable, then so is ๐‘“โˆ˜๐‘‡.
Moreover if ๐‘“โˆˆ๐ฟ+ or ๐‘“โˆˆ๐ฟ1, then ๐‘“โˆ˜๐‘‡โˆˆ๐ฟ+, ๐‘“โˆ˜๐‘‡โˆˆ๐ฟ1 respectively. And

โˆซ๐‘“๐‘‘๐‘š=|det๐‘‡|โˆซ๐‘“โˆ˜๐‘‡๐‘‘๐‘š

(b)

๐ธโˆˆโ„’๏ธ€๐‘›โŸน๐‘‡(๐ธ)โˆˆโ„’๏ธ€๐‘›and๐‘š(๐‘‡(๐ธ))=|det๐‘‡|๐‘š(๐ธ)
Proof

Note: ๅฏนไบŽ ๐‘‡,๐‘†โˆˆ๐บ๐ฟ(๐‘›,โ„), ๅฆ‚ๆžœ

โˆซ๐‘“=|det๐‘‡|โˆซ๐‘“โˆ˜๐‘‡ and โˆซ๐‘“=|det๐‘†|โˆซ๐‘“โˆ˜๐‘†

, ้‚ฃไนˆๅˆ™ๆœ‰

โˆซ๐‘“=|det(๐‘‡โˆ˜๐‘†)|โˆซ๐‘“โˆ˜(๐‘‡โˆ˜๐‘†)(๐‘ฅ)

which trivially follows from computation. (and det(๐‘†โˆ˜๐‘‡)=det๐‘†ร—det๐‘‡ for any linear map ๐‘†,๐‘‡.)
recall that:

Lemma 7.24 : row reduction

ไปปๆ„ invertible linear map ๅฏไปฅ่ขซๆ‹†ๅˆ†ไธบ finite ไธช elementary linear maps. ( ๐‘‡1: scale ไธ€่กŒ; ๐‘‡2: ไบคๆขไธค่กŒ; ๐‘‡3: ไธ€่กŒๅŠ ไธŠๅฆไธ€่กŒ็š„ๅ€ๆ•ฐ).

ไบŽๆ˜ฏ, ๆˆ‘ไปฌๅช้œ€่ฆ prove the theorem for elementary linear maps ๅฐฑๅฏไปฅไบ†. ่€Œ elementary linear maps ็š„ cases ๅˆ™ easily follows from Fubini-Toneilli.
Let ๐‘“ be Borel measurable.
ๅฏนไบŽ ๐‘‡2: ไบคๆขไธค่กŒ (ๅ…ถ det ไธบ โˆ’1), ๆˆ‘ไปฌๆ”นๅ˜ the order of integration for two coordinates, ๅ› ่€Œ integration ไธๅ˜;
ๅฏนไบŽ ๐‘‡1: scale ไธ€่กŒ by const ๐‘ (ๅ…ถ det ไธบ ๐‘), ๆˆ‘ไปฌๅœจไธ€ไธช coordinate ไธŠ็งฏๅˆ†ๅ€ผ็ฟป ๐‘ ๅ€, ๅ› ่€Œๆ•ดไฝ“็งฏๅˆ†ๅ€ผ็ฟป ๐‘ ๅ€. ่ฟ™้‡Œ็”จๅˆฐไบ† โ„โ†’โ„ ็š„ Lebesgue integral ็š„ๅทฒ่ฏๆ˜Ž็ป“่ฎบ:

โˆซ๐‘“(๐‘ก)๐‘‘๐‘ก=|๐‘|โˆซ๐‘“(๐‘๐‘ก)๐‘‘๐‘ก

ๅฏนไบŽ ๐‘‡3: ไธ€่กŒๅŠ ไธŠๅฆไธ€่กŒ็š„ๅ€ๆ•ฐ (ๅ…ถ detไธบ 1), ๆˆ‘ไปฌ recall โ„โ†’โ„ ็š„ Lebesgue integral ็š„ translation invariance:

โˆซ๐‘“(๐‘ก+๐‘Ž)๐‘‘๐‘ก=โˆซ๐‘“(๐‘ก)๐‘‘๐‘ก

ๅ› ่€Œๆ•ดไฝ“็งฏๅˆ†ๅ€ผไธๅ˜.
ไปŽ่€Œๆˆ‘ไปฌ่ฏๆ˜Žไบ† (a) for Borel measurable ๐‘“.
ไปŽ่€Œ, (b) for Borel set ๐ธ trivially follows from (a), by taking indicator function.
่€ŒๅฏนไบŽ (b) ็š„ ๐ธ Lebesgue measurable case, ๐ธ=๐ตโˆช๐‘ for some Borel set ๐ต ไปฅๅŠ subnull set ๐‘, ไปŽ่€Œ ๐‘š(๐ธ)=๐‘š(๐ต).
ไปŽ่€Œ (b) proved.
่€Œ (a) ็š„ ๐‘“ Lebesugue measurable ็š„ case, by def reduces to ๐‘“=๐œ’๐ธ where ๐ธ is Lebesgue measurable set, ไบŽๆ˜ฏ follows from the (b).

โ–ก

7.1.7 Lebesgue measure is invariant under rotation (and reflection)

Corollary 7.22 : Lebesgue measure is invariant under rotation

ๅฏนไบŽ rotation ๅ’Œ reflection (ๅณ orthogonal transformation), ๅณ ๐‘‡๐‘‡โˆ—=๐ผ๐‘› ็š„ linear map ๐‘‡, ๆœ‰ ๐‘š(๐‘‡(๐ธ))=๐‘š(๐ธ).

Proof

๐‘‡๐‘‡โˆ—=๐ผ๐‘›โŸน|det(๐‘‡)|=1.

โ–ก

7.2 Change of Variable Thm on โ„๐‘›[Fol 2.6, finished]

7.2.1 COV

Theorem 7.37 : general change of variable theorem

Suppose ฮฉโŠ‚โ„๐‘› open, ๐บ:ฮฉโ†’โ„๐‘› ไธบไธ€ไธช ๐ถ1 diffeomorphism.
Claim:

  • ๅฆ‚ๆžœ ๐‘“:๐บ(ฮฉ)โ†’โ„‚ ไธŠๆ˜ฏ Lebesgue measurable ็š„, ๅˆ™ ๐‘“โˆ˜๐บ:ฮฉโ†’โ„‚ ไนŸๆ˜ฏ Lebesgue measurable ็š„. ๅนถไธ”, ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ+(๐บ(ฮฉ),๐‘š) ๆˆ–่€… ๐‘“โˆˆ๐‘“โˆˆ๐ฟ1(๐บ(ฮฉ),๐‘š), ๅˆ™ๆœ‰

    โˆซ๐บ(ฮฉ)๐‘“๐‘‘๐‘š=โˆซฮฉ(๐‘“โˆ˜๐บ)|det๐ท๐บ|๐‘‘๐‘š
  • ๅฆ‚ๆžœ ๐ธโŠ‚ฮฉ ๆ˜ฏ Lebesgue measurable set, ๅˆ™ ๐บ(๐ธ) ไนŸๆ˜ฏ Lebesgue measurable set, ๅนถไธ”

    ๐‘š(๐บ(๐ธ))=โˆซ๐ธ|det๐ท๐บ|๐‘‘๐‘š
Proof

้ฆ–ๅ…ˆ, ็ฑปไผผไบŽไธŠไธ€ไธช lecture ไธญ็š„ๅ„ไธช่ฏๆ˜Ž, ๅช้œ€่ฆ prove for Borel measurable functions ๅ’Œ Borel sets ๅฐฑๅฏไปฅไบ†. ๆˆ‘ไปฌๅˆ†ไธบไบ”ๆญฅ่ฏๆ˜Ž.
Step 1: ๆˆ‘ไปฌ้ฆ–ๅ…ˆ่ฏๆ˜Ž, ๅœจ ๐ธ ไธบไธ€ไธช closed cube ็š„ๆƒ…ๅ†ตไธ‹ (ๆˆ‘ไปฌ่ฝฌ่€Œ็”จ ๐‘„ ๆฅ่กจ็คบๅฎƒ), ๆœ‰

๐‘š(๐บ(๐‘„))โ‰คโˆซ๐‘„|det๐ท๐บ(๐‘ฅ)|๐‘‘๐‘ฅ

Proof of Step 1:

๐‘„={๐‘ฅ:||๐‘ฅโˆ’๐‘Ž||supโ‰คโ„Ž}

By MVT ๅฎนๆ˜“ๅพ—ๅˆฐ, ๅฏนไบŽไปปๆ„็š„ ๐‘ฅโˆˆ๐‘„, ๆœ‰:

||๐บ(๐‘ฅ)โˆ’๐บ(๐‘Ž)||supโ‰คโ„Žโ‹…(sup๐‘ฆโˆˆ๐‘„||๐ท๐บ(๐‘ฆ)||sup)

(by bounding each entry.)
ไปŽ่€Œ, ๆˆ‘ไปฌๅ‘็Žฐ ๐บ(๐‘„) ๆ˜ฏ contained in ไธ€ไธช่พน้•ฟๆ˜ฏ โ„Žโ‹…sup๐‘ฆโˆˆ๐‘„||๐ท๐บ(๐‘ฆ)||sup ็š„ cube ็š„.
ไปŽ่€Œๆœ‰:

๐‘š(๐บ(๐‘„))โ‰ค(sup๐‘ฆโˆˆ๐‘„||๐ท๐บ(๐‘ฆ)||)๐‘›๐‘š(๐‘„)

ๅœจ invertible ๐‘‡ ็š„ไฝœ็”จไธ‹, ๐‘‡โˆ’1โˆ˜๐บ ไป็„ถๆ˜ฏไธ€ไธช diffeomorphism, ไปŽ่€Œ

๐‘š(๐บ(๐‘„))=|det๐‘‡|๐‘š(๐‘‡โˆ’1(๐บ(๐‘„)))โ‰ค|det๐‘‡|(sup๐‘ฆโˆˆ๐‘„||๐‘‡โˆ’1๐ท๐บ(๐‘ฆ)||)๐‘›๐‘š(๐‘„)

Let ๐œ–>0.
็”ฑไบŽ ๐ท๐บ ๆ˜ฏ continuous ็š„, ๐ท๐บ(๐‘ฅ)โˆ’1๐ท๐บ(๐‘ฆ) ไนŸๆ˜ฏ ctn ็š„ (ไปŽ่€Œ uni.ctn. in the compact cube), ๆˆ‘ไปฌๅฏนไบŽไปปๆ„ ๐œ–>0 ้ƒฝๅฏไปฅๆ‰พๅˆฐไธ€ไธช ๐›ฟ>0 ไฝฟๅพ— ๅฏนไบŽไปปๆ„็š„ ๐‘ฆ,๐‘งโˆˆ๐‘„ s.t. ||๐‘ฆโˆ’๐‘ง||supโ‰ค๐›ฟ, ้ƒฝๆœ‰

||๐ท๐บ(๐‘ฅ)โˆ’1๐ท๐บ(๐‘ฆ)||โ‰ค1+๐œ–

ไบŽๆ˜ฏๆˆ‘ไปฌๅฏไปฅๆŠŠ ๐‘„ ๅˆ‡ๅˆ†ๆˆ interior disjoint ็š„ closed subcubes ๐‘„1,โ‹ฏ,๐‘„๐‘, ๆ ‡่ฎฐๅ…ถๅ„ไธชไธญๅฟƒไธบ ๐‘ฅ1,โ‹ฏ๐‘ฅ๐‘, ๅ…ถๆฏไธช็š„ side length ้ƒฝ่‡ณๅคšไธบ ๐›ฟ, ไปŽ่€Œๆœ‰ ๐บ(๐‘„)โŠ‚โ‹ƒ๐‘—=1๐‘๐‘š(๐บ(๐‘„๐‘—)). ไบŽๆ˜ฏ

๐‘š(๐บ(๐‘„))โ‰คโˆ‘๐‘—=1๐‘๐‘š(๐บ(๐‘„๐‘—))โ‰คโˆ‘๐‘—=1๐‘|det๐ท๐บ(๐‘ฅ๐‘—)|(sup๐‘ฆโˆˆ๐‘„๐‘—||๐ท๐บ(๐‘ฅ๐‘—)โˆ’1๐ท๐บ(๐‘ฆ)||sup)๐‘›๐‘š(๐‘„๐‘—)โ‰ค(1+๐œ–)โˆ‘๐‘—=1๐‘|det๐ท๐บ(๐‘ฅ๐‘—)|๐‘š(๐‘„๐‘—)โ†’(1+๐œ–)|det๐ท๐บ(๐‘ฅ)|๐‘š(๐‘„) as ๐›ฟโ†’0โ†’|det๐ท๐บ(๐‘ฅ)|๐‘š(๐‘„)=โˆซ๐‘„|det๐ท๐บ(๐‘ฅ)|๐‘‘๐‘š as ๐œ–โ†’0

่ฏๆ˜Žไบ†่ฟ™ไธ€็ป“่ฎบ, ๆˆ‘ไปฌๅฐฑๅฎŒๆˆไบ†่ฟ™ไธช proof ็š„ไธ€ๅคงๅŠ.

Step 2: Prove

๐‘š(๐บ(๐‘ˆ))โ‰คโˆซ๐‘ˆ|det๐ท๐บ(๐‘ฅ)|๐‘‘๐‘š

for open ๐‘ˆ ็š„ case.
Proof of Step 2: Directly follows from ไธŠไธ€ lecture ็š„่ฟ™ไธช statement: ไปปๆ„ open ๐ธโŠ‚โ„๐‘› ้ƒฝๆ˜ฏ countable disjoint interior cubes ็š„ union.

Step 3: Prove

๐‘š(๐บ(๐ธ))โ‰คโˆซ๐ธ|det๐ท๐บ(๐‘ฅ)|๐‘‘๐‘š

for ๐ธ Borel ็š„ case.
Proof of Step 3: Apply step 2 ็š„็ป“่ฎบ, ไฝฟ็”จ MCT for ๐ฟ+ case, ไฝฟ็”จ DCT for ๐ฟ1 case. ่‡ณๆญค, ๆˆ‘ไปฌๅฎŒๆˆไบ† (b) ็š„่ฏๆ˜Ž็š„ไธ€ไธชๆ–นๅ‘, ็”ฑๆญคๅฏไปฅๅฎŒๆˆ (a) ็š„ไธ็ญ‰ๅผ็š„ไธ€ไธชๆ–นๅ‘:

Step 4: ่ฏๆ˜Ž

โˆซ๐บ(ฮฉ)๐‘“๐‘‘๐‘šโ‰คโˆซฮฉ๐‘“โˆ˜๐บ|det๐ท๐บ(๐‘ฅ)|๐‘‘๐‘š

simple function ็š„ case reduces to measure, ่€Œ ๐ฟ+ ็š„ case follows from MCT.

Step 5: ไธ็ญ‰ๅผ็š„ๅฆไธ€ๆ–นๅ‘: ๅ…ถๅฎžๅพˆ็ฎ€ๅ•, ๅ› ไธบ diffeomorphism ็š„ inverse ไป็„ถๆ˜ฏ diffeomorphism, ๆ‰€ไปฅ apply inverse ๅฏๅพ—.
ๆณจๆ„, ่ฟ™ๅชๆ˜ฏ for Borel ๐ธ ๅ’Œ ๐ฟ+ Borel measurable ๐‘“, ไธ่ฟ‡ๆˆ‘ไปฌๅฎนๆ˜“ๆŽฅ็€ๆŽจๅฏผๅ‡บ Lebesgue measurable ๐ธ ็š„ๆƒ…ๅ†ตๅ’Œ ๐‘“โˆˆ๐ฟ+(๐‘š) ็š„ๆƒ…ๅ†ต; ไปŽ่€Œๅ†ๆŽฅ็€ๆŽจๅฏผๅ‡บ ๐‘“โˆˆ๐ฟ1(๐‘š) ็š„ๆƒ…ๅ†ต.

โ–ก

7.2.2 application of COV: polar coordinate

Definition 7.37 : mapping from Euclidean coord to polar coord

ๆˆ‘ไปฌๅฎšไน‰:

ฮฆ:โ„๐‘›\{0}โ†’(0,โˆž)ร—๐‘†๐‘›โˆ’1

by:

๐‘ฅโ†ฆ(๐‘Ÿโˆˆโ„,๐œƒโˆˆ๐•Š๐•Ÿโˆ’๐Ÿ™)

ๅ…ถไธญ,

๐‘Ÿ=|๐‘ฅ|,๐œƒ=๐‘ฅ|๐‘ฅ|โˆˆ๐‘†๐‘›โˆ’1

่ฟ™ๆ˜ฏไธ€ไธชๅพˆ็›ด่ง‚็š„ๅๆ ‡ๅ˜ๆข, ๅณไธ€ไธช diffeomorphism.

Definition 7.38 : a Borel measure on (0,โˆž)ร—๐‘†๐‘›โˆ’1

ๆˆ‘ไปฌๅฎšไน‰

๐‘šโˆ—(๐ธ)โ‰”๐‘š(ฮฆโˆ’1(๐ธ))

่ฟ™ๆ˜ฏไธ€ไธช้€š่ฟ‡ๅๆ ‡ๅ˜ๆข็š„ preimage ็š„ Borel measure ๅฎšไน‰็š„ๆ–ฐ็š„ Borel measure.

Theorem 7.38

Define Borel measure ๐œŒ on (0,โˆž) by:

๐œŒ(๐ธ)=โˆซ๐ธ๐‘Ÿ๐‘›โˆ’1๐‘‘๐‘Ÿ

ๅญ˜ๅœจ unique ็š„ Borel measure ๐œŽ๐‘›โˆ’1 on ๐‘†๐‘›โˆ’1, ไฝฟๅพ— for Borel measurable ๐‘“:โ„๐‘›โ†’โ„‚ ไธ” ๐‘“โ‰ฅ0 or ๐‘“โˆˆ๐ฟ1(๐‘š), ๆœ‰

โˆซโ„๐‘›๐‘“(๐‘ฅ)๐‘‘๐‘š=๐ถ๐‘‚๐‘‰โˆซ(0,โˆž)ร—๐‘†๐‘›โˆ’1๐‘“(๐‘Ÿ๐œƒ)๐‘‘๐‘šโˆ—=๐น๐‘ข๐‘๐‘–๐‘›๐‘–โˆซ0โˆžโˆซ๐‘†๐‘›โˆ’1๐‘“(๐‘Ÿ๐œƒ)๐‘‘๐œŽ๐‘‘๐œŒ=โˆซ0โˆž๐‘Ÿ๐‘›โˆ’1โˆซ๐‘†๐‘›โˆ’1๐‘“(๐‘Ÿ๐œƒ)๐‘‘๐œŽ๐‘‘๐‘Ÿ
Proof

่ง Folland 2.49.

โ–ก

Example 7.21

๐œŽ(๐‘†1)=2๐œ‹, ๐œŽ(๐‘†2)=4๐œ‹.

Example 7.22

ไฝฟ็”จ polar coordinate ่ฎก็ฎ—็งฏๅˆ†:

โˆซโ„๐‘›๐‘’โˆ’๐‘Ž|๐‘ฅ|2๐‘‘๐‘ฅ=(๐œ‹๐‘Ž)๐‘›2

่ฟ™ๆ˜ฏๅ› ไธบ:

๐ผ2=2๐œ‹โˆซ0โˆž๐‘Ÿ๐‘’โˆ’๐‘Ž๐‘Ÿ2๐‘‘๐‘Ÿ=๐œ‹๐‘Ž

่€Œ็”ฑไบŽ

๐‘’โˆ’๐‘Ž|๐‘ฅ|2=โˆ๐‘—=1๐‘›๐‘’โˆ’๐‘Ž๐‘ฅ๐‘—2

ๆˆ‘ไปฌๅพ—ๅˆฐ

๐ผ๐‘›=(๐ผ1)๐‘›

็‰นๅˆซๅœฐ,

๐ผ2=๐ผ12,thus ๐ผ1=(๐œ‹๐‘Ž)12

8 Hardy-Littlewood maximal function and Lebesgue differentiation theorem

8.1 Hardy-Littlewood max function and max theorem [Fol 3.4]

็›ฎๅ‰ๆˆ‘ไปฌ finish ไบ† Folland ็š„ Ch1, Ch2.
ๆˆ‘ไปฌๅฐ†ๅ…ˆ่ทณ่ฟ‡ Radon-Nikodym differentiation theory, ๅœจ่ฎฒๅฎŒ ๐ฟ๐‘ space theory ๅŽๅ†ๅ›žๅˆฐ Radon-Nikodym differentiation theory. ไฝ†ๆ˜ฏๆˆ‘ไปฌๅฐ†ไผšๅ…ˆๅฐ† differentiation theory ไธญ็š„ไธ€ไธช็‰นๆฎŠ้ƒจๅˆ†: HL max theorem ๅ’Œ Lebesgue differentiation theorem, ๅ› ไธบๅฎƒไปฌๅœจ ๐ฟ๐‘ space theory ไธญ้œ€่ฆ่ขซ็”จๅˆฐ.
ๆญค lec ๅฏนๅบ”: Folland 3.4( 1)
Differentiation theorey ็š„ overview:
Radon-Nikodym derivative ๅนถไธๆ˜ฏ classical calculus ็š„ๆ‰ฉๅฑ• (classical calculus ่กจ็คบๅ˜้‡ไน‹้—ด็š„็›ธๅฏนๅ˜ๅŒ–), ่€Œๆ˜ฏไธ“้—จ้’ˆๅฏน: ๅŒไธ€ไธช measure space ไธŠ, ไธ€ไธชๆต‹ๅบฆๅฏนไบŽๅฆไธ€ไธชๆต‹ๅบฆ (่ฆๆฑ‚ๅฎƒไปฌไน‹้—ด็ปๅฏน่ฟž็ปญ) ็š„ๅ˜ๅŒ–็އ.

๐‘‘๐œˆ=๐‘“๐‘‘๐œ‡

ไปŽ่€Œ

๐œˆ(๐ด)=โˆซ๐ด๐‘“๐‘‘๐œ‡

่ฟ™ไฝฟๅพ—ๆˆ‘ไปฌๅฏไปฅๆ›ดๆ”นไธ€ไธช็งฏๅˆ† with respect to ็š„ๆต‹ๅบฆ.

ๅ…ถๆ ธๅฟƒๅฎš็†: Radon-Nikodym Theorem, ่กจ็คบไบ†ๅœจไธ€่ˆฌ็š„ measure space ไธŠ, ่ฟ™ไธคไธชๆต‹ๅบฆๆปก่ถณไธ€ๅฎšๆกไปถไธ‹, ่ฟ™ไธช Radon-Nikodym derivative ็š„ๅญ˜ๅœจๆ€ง; ่€Œ LDT ๆๅ‡บไธ€็งๅœจ Euclidean space ไธŠ, ๆฑ‚ Radon Dikodym derivative ็š„ๆ–นๆณ•.
LDT ๆœฌ่บซๆ˜ฏไธ€็งๅฐ†็งฏๅˆ†ไฟกๆฏ่ฝฌๆขไธบ็‚นๆ€ไฟกๆฏ็š„ๆ‰‹ๆฎต. ๅฎƒ่กจ็คบๅฏนไบŽ locally integrable ็š„ๅ‡ฝๆ•ฐ, ๅฑ€้ƒจ็งฏๅˆ†ๅนณๅ‡ๅ€ผๅฏไปฅๆ”ถๆ•›ๅˆฐๅ‡ฝๆ•ฐๆœฌ่บซ, a.e.
ๅ› ่€Œๅœจ็Ÿฅ้“ ๐œˆ ๅ’Œ ๐œ‡ ็š„ๆƒ…ๅ†ตไธ‹, LDT ๆไพ›ไบ†็ฑปๆฏ”็ปๅ…ธๅพฎ็งฏๅˆ†ไธญ "ๅฏผๆ•ฐๆ˜ฏๅฑ€้ƒจๅ˜ๅŒ–็އ็š„ๆž้™" ็š„่ง‚็‚น: ๅœจ Euclidean space ไธŠ, Radon Nikodym derivative ็ญ‰ไบŽๅฑ€้ƒจๅ‡ๅ€ผ:

๐‘“(๐‘ฅ)=lim๐‘Ÿโ†’0๐œˆ(๐ต(๐‘Ÿ,๐‘ฅ))๐œ‡(๐ต(๐‘Ÿ,๐‘ฅ)) for a.e. ๐‘ฅ

Radon Nikodym derivative ็š„ๅบ”็”จ: ๆฏ”ๅฆ‚ๅœจๆฆ‚็އ่ฎบไธญ, pdf/pmf ้ƒฝๆ˜ฏ cdf ๅฏนไบŽ Lebesgue measure ็š„ Radon Nikodym derivative; ๅœจ่ดๅถๆ–ฏๆŽจ็†ไธญ๏ผŒ็ป™ๅฎšๅ…ˆ้ชŒ prior ๅ’Œ่ง‚ๆต‹ๆ•ฐๆฎ็š„ๅˆ†ๅธƒ, ๅŽ้ชŒๅˆ†ๅธƒ posterior ็š„ๅฏ†ๅบฆๅฏไปฅ้€š่ฟ‡ Radon-Nikodym ๅฏผๆ•ฐ่ฎก็ฎ—. ไธ‹้ขไป‹็ปๆฆ‚ๅฟต:

8.1.1 ๐ฟ๐‘™๐‘œ๐‘1 and local average

Definition 8.39 : locally integrable

ๅฆ‚ๆžœ measurable ๐‘“:โ„๐‘›โ†’โ„‚ ๅœจไปปๆ„ bounded subset of โ„๐‘› ไธŠ็š„ integral ้ƒฝ <โˆž, ๅˆ™็งฐ function ๐‘“:โ„๐‘›โ†’โ„‚ ๆ˜ฏ locally integrable ็š„, ๅ†™ไฝœ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(๐‘š).

Example 8.23

่€ƒ่™‘

๐‘“(๐‘ฅ)โ‰”|๐‘ฅ|๐‘,๐‘ฅโˆˆโ„๐‘›

(ไฝฟ็”จ polar coord) ๅฏ้ชŒ่ฏ:

๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(๐‘š)โ‡”๐‘>โˆ’๐‘›
Definition 8.40 : average

ๅฏนไบŽ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(๐‘š), ไปฅๅŠ bounded and Lebesgue measurable ๐ธโŠ‚โ„๐‘› with ๐‘š(๐ธ)>0, ๆˆ‘ไปฌๅฎšไน‰:

avg๐ธ๐‘“:=1๐‘š(๐ธ)โˆซ๐ธ๐‘“๐‘‘๐‘š

ไธบ ๐‘“ ๅœจ ๐ธ ไธŠ็š„ average value.
็‰นๅˆซๅœฐ, ๅฝ“ ๐ธ ไธบไธ€ไธช ball ๐ต(๐‘Ÿ,๐‘ฅ) ๆ—ถ, ๆˆ‘ไปฌๅฏไปฅๅ†™ไฝœ:

๐ด๐‘Ÿ๐‘“(๐‘ฅ)โ‰”avg๐ต(๐‘Ÿ,๐‘ฅ)๐‘“

่กจ็คบๅฎƒๅœจ ๐‘ฅ ไธบไธญๅฟƒ็š„ ๐‘Ÿ ไธบๅŠๅพ„็š„ ball ไธŠ็š„ average.

Lemma 8.25

ๅฏนไบŽไปปๆ„ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1, ๐ด๐‘Ÿ๐‘“(๐‘ฅ) ้ƒฝๆ˜ฏ jointly continuous in ๐‘Ÿ and ๐‘ฅ ็š„. (๐‘Ÿ>0,๐‘ฅโˆˆโ„๐‘›)

Proof

Suppose (๐‘ฅ๐‘—,๐‘Ÿ๐‘—)โ†’(๐‘ฅ,๐‘Ÿ) in โ„๐‘›ร—โ„>0.
ไบŽๆ˜ฏ for sure:

๐‘š(๐ต(๐‘ฅ๐‘—,๐‘Ÿ๐‘—))โ†’๐‘—โ†’โˆž๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

ๅนถไธ” by DCT (ๅ– ๐œ’๐ต(๐‘Ÿ0+1,๐‘ฅ0) ไฝœไธบ bound) ๅฏไปฅๅพ—ๅˆฐ:

โˆซ๐‘“๐œ’๐ต(๐‘ฅ๐‘—,๐‘Ÿ๐‘—)โ†’โˆซ๐œ’๐ต(๐‘ฅ,๐‘Ÿ)

โ–ก

8.1.2 Hardy-Littlewood maximal function

Definition 8.41 : Hardy-Littlewood maximal function

ๅฏนไบŽ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1, ๆˆ‘ไปฌๅฎšไน‰ๅฎƒ็š„ HL maximal function ไธบ:

๐ป๐‘“(๐‘ฅ):=sup๐‘Ÿ>0๐ด๐‘Ÿ|๐‘“|(๐‘ฅ)

HF maximal ๅ‡ฝๆ•ฐ ๐ป๐‘“(๐‘ฅ)่กจ็คบ ๐‘“ ็š„็ปๅฏนๅ€ผๅ‡ฝๆ•ฐๅœจ ๐‘ฅ ๅค„่ƒฝๅ–ๅˆฐๆœ€ๅคง็š„ local average.

Theorem 8.39 : HL maximal function ๆ˜ฏ measurable ็š„

ๅฏนไบŽไปปๆ„ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1, ๐ป๐‘“ ้ƒฝๆ˜ฏ measurable ็š„.

Proof

Follows from lemma.

(๐ป๐‘“)โˆ’1((๐‘Ž,โˆž))=โ‹ƒ๐‘Ÿ>0(๐ด๐‘Ÿ|๐‘“|)โˆ’1((๐‘Ž,โˆž))

ๆ˜ฏ open ็š„, ๅ› ไธบ ๐ด๐‘Ÿ|๐‘“| ctn, ctn function ไธ‹ open set ็š„ preimage ไนŸ open.

โ–ก

Corollary 8.23

ๅฆ‚ๆžœ ๐‘“:โ„๐‘›โ†’[0,โˆž) ๆ˜ฏ lower semictn ็š„, ้‚ฃไนˆๅฎƒไธ€ๅฎš Borel (thus Lebesgue) measurable.

8.1.3 Vitali-type convering lemma

ๅฏนไบŽไธ€ไธช ball ๐ต=๐ต(๐‘ฅ,๐‘Ÿ) ไปฅๅŠไธ€ไธช constant ๐‘, ๆˆ‘ไปฌๅฎšไน‰:

๐‘๐ตโ‰”๐ต(๐‘ฅ,๐‘๐‘Ÿ)
Lemma 8.26 : Vitali-type covering lemma

For given collection of balls {๐ต๐‘—โŠ‚โ„๐‘›}๐‘—=1๐‘˜, ๅญ˜ๅœจ disjoint subcollection {๐ต๐‘—1,โ‹ฏ,๐ต๐‘—๐‘š} ไฝฟๅพ—

โ‹ƒ๐‘—=1๐‘˜๐ต๐‘—โŠ‚โ‹ƒ๐‘–=1๐‘š(3๐ต๐‘—๐‘–)

(ไบŽๆ˜ฏ,

๐‘š(โ‹ƒ๐‘—=1๐‘˜๐ต๐‘—)โ‰ค3๐‘›๐‘š(โ‹ƒ๐‘–=1๐‘š(3๐ต๐‘—๐‘–))

)

Proof

Greedy Algrithm: ็›ดๆŽฅๆŒ‰็…งๅŠๅพ„ๅคงๅฐๆŽ’ๅบ, ๅ–ๅ‡บๆœ€ๅคง็š„ disjoint subcollection.
Prove without words:

Figureย 22:

(ๆฏๆฌก้ƒฝ้€‰ๆ‹ฉไธ‹ไธ€ไธชๅ’Œๅ‰้ขๆ‰€ๆœ‰ๆ›ดๅคง็š„็ƒไธ intersect ็š„ๆœ€ๅคง็ƒ; ๅœจ่ฟ™ไธช่ฟ‡็จ‹ไธญ, ๆ‰€ๆœ‰ๅ’Œๅ‰้ขๆ›ดๅคง็š„็ƒๆœ‰ intersection ็š„็ƒ้ƒฝ่ขซ่ขซๅŒ…ๆ‹ฌๅœจ่ฏฅ็ƒ็š„ไธ‰ๅ€็ƒ้‡Œ.)

โ–ก

8.1.4 Hardy-Littlewood maximal theorem

Theorem 8.40 : HL maximal theorem

For ๐ฟ1(๐‘š๐‘›), take constant ๐ถโ‰”3๐‘›, ๅˆ™ๅฏนไบŽไปปๆ„ ๐‘“โˆˆ๐ฟ1(๐‘š๐‘›), ้ƒฝๆœ‰:

๐‘š({๐‘ฅ:๐ป๐‘“(๐‘ฅ)>๐›ผ})โ‰ค๐ถ๐›ผโˆซ|๐‘“|
Proof

Set

๐ธ๐›ผ:={๐‘ฅ:๐ป๐‘“(๐‘ฅ)>๐›ผ}โŠ‚โ„๐‘›

ๅ› ่€Œ by def of ๐ป๐‘“, ๅฏนไบŽไปปๆ„็š„ ๐‘ฅโˆˆ๐ธ๐›ผ ้ƒฝๅญ˜ๅœจ ๐‘Ÿ๐‘ฅ ไฝฟๅพ—

๐‘š(๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ))<1๐›ผโˆซ๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ)|๐‘“|

ๅฏนไบŽ compact ๐พโˆˆ๐ธ๐›ผ, ไธ€ๅฎšๅญ˜ๅœจ finite subcovering {๐ต(๐‘ฅ๐‘–,๐‘Ÿ๐‘–)} covers ๐พ.
ไบŽๆ˜ฏ Apply Vitali-type covering Lemma:

๐‘š(๐พ)โ‰คโˆ‘๐‘–๐‘š(3๐ต(๐‘ฅ๐‘–,๐‘Ÿ๐‘–))=3๐‘›โˆ‘๐‘–๐‘š(๐ต(๐‘ฅ๐‘–,๐‘Ÿ๐‘–))โ‰ค3๐‘›๐›ผโˆซ|๐‘“|

ไบŽๆ˜ฏ by inner regularity, taking sup over all compact subsets ๅพ—่ฏ.

โ–ก

8.2 Lebesgue differentiation Theorem [Fol 3.4]

ๅฏนๅบ”: Folland 3.4(2)

Definition 8.42 : Lebesgue set

ๆˆ‘ไปฌๅฎšไน‰ไธ€ไธชๅ‡ฝๆ•ฐ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(โ„๐‘›) ็š„ Lebesgue set ไธบ:

๐ฟ๐‘“โ‰”{๐‘ฅโˆˆโ„๐‘›โˆฃlim๐‘Ÿโ†’0+avg๐ต(๐‘ฅ,๐‘Ÿ)|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฆ=0}

ๅ…ถไธญๆฏไธช point ่ขซ็งฐไธบไธ€ไธชLebesgue point.

8.2.1 original LDT: locally ๐ฟ1 ๅ‡ฝๆ•ฐๅ‡ ไนŽๆฏไธ€็‚น้™„่ฟ‘็š„ๅ‡ฝๆ•ฐๅ‡ๅ€ผ้ƒฝ็ญ‰ไบŽ่ฟ™ไธ€็‚นไธŠ็š„ๅ€ผ

Theorem 8.41 : Lebesgue differentiation theorem

ๅฏนไบŽไปปๆ„็š„ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(โ„๐‘›), ๐ฟ๐‘“ is Leb mble and ๐‘š(๐ฟ๐‘“๐‘)=0.

Proof

็”ฑไบŽๅฏนไบŽไปปๆ„ ๐‘ฅ ้ƒฝๆœ‰:

๐‘“(๐‘ฅ)=lim๐‘โ†’โˆž๐‘“๐œ’๐ต(0,๐‘)(๐‘ฅ)

ๆ‰€ไปฅ it suffices to prove the statement for ๐‘“๐œ’๐ต(0,๐‘) for ไปปๆ„ ๐‘.
ๆณจๆ„, ๐‘“๐œ’๐ต(0,๐‘) ๆ˜ฏไธ€ไธช ๐ฟ1 function. ๅ› ่€Œๅช้œ€่ฆ prove the statement for ๐‘“โˆˆ๐ฟ1(โ„๐‘›) ๅฐฑๅฏไปฅ generalize it to ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(โ„๐‘›). ๅ› ่€Œ WLOG suppose ๐‘“โˆˆ๐ฟ1(โ„๐‘›).
้ฆ–ๅ…ˆ, it is true for ๐‘“โˆˆ๐ถ๐‘0(โ„๐‘›). (ๅฎนๆ˜“ check: ๅœจๆŸ็‚น่ฟž็ปญๆ€ง็š„ๅฎšไน‰่ƒฝๅคŸ imply ๅœจ่ฟ™ไธ€็‚น็š„ๅ‡ๅ€ผ็ญ‰ไบŽ่ฟ™ไธ€็‚น็š„ๅ€ผ).
In other cases, ๆˆ‘ไปฌ้œ€่ฆๅˆฉ็”จ ๐ถ๐‘0(โ„๐‘›) ๅœจ ๐ฟ1(โ„๐‘›) ไธญ็š„ density.
Let

๐‘„(๐‘ฅ,๐‘Ÿ):=๐ด๐‘Ÿ|๐‘“โˆ’๐‘“(๐‘ฅ)|(๐‘ฅ)

่ฟ™ๆ˜ฏไธ€ไธช nonnegative function. ๆณจๆ„, ไธŠไธช lec ไธญๆˆ‘ไปฌ่ฏๆ˜Žไบ† ๐ด๐‘Ÿ๐‘“(๐‘ฅ) ๆ˜ฏ jointly continuous in ๐‘Ÿ and ๐‘ฅ ็š„. ๅ› ่€Œ ๐‘„(๐‘ฅ,๐‘Ÿ) ไนŸๆ˜ฏ jointly continuous in ๐‘Ÿ and ๐‘ฅ ็š„.
ๆˆ‘ไปฌ้šๅŽๅฎšไน‰:

๐‘„(๐‘ฅ):=limsup๐‘Ÿโ†’0+๐‘„(๐‘ฅ,๐‘Ÿ)

ไบŽๆ˜ฏ ๐‘ฅโ†ฆ๐‘„ ็›ธๅฝ“ไบŽ maximal function ็š„ไธ€ไธชๅ˜ไฝ“. ๅฎนๆ˜“้ชŒ่ฏๅฎƒไนŸๆ˜ฏ measurable ็š„. ๆˆ‘ไปฌ WTS:

๐‘š({๐‘ฅ:๐‘„(๐‘ฅ)>0})=0

็ญ‰ไปทไบŽ show:

๐‘š({๐‘„>๐›ผ})=0for all ๐›ผ=1๐‘›,๐‘›โˆˆโ„•

Let ๐œ–>0. By density of ๐ถ๐‘0(โ„๐‘›) in ๐ฟ1(โ„๐‘›), ๆˆ‘ไปฌๅฏไปฅ pick ๐‘”โˆˆ๐ถ๐‘0(โ„๐‘›) s.t.

โˆซ|๐‘“โˆ’๐‘”|<๐œ–

By triangular ineq in ๐ฟ1, for a.e. ๐‘ฅโˆˆโ„๐‘› we have:

๐ด๐‘Ÿ|๐‘“โˆ’๐‘“(๐‘ฅ)|(๐‘ฅ)โ‰ค๐ด๐‘Ÿ|๐‘“โˆ’๐‘”|(๐‘ฅ)+๐ด๐‘Ÿ|๐‘”โˆ’๐‘”(๐‘ฅ)|(๐‘ฅ)+๐ด๐‘Ÿ|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|(๐‘ฅ)

where

(constant)๐ด๐‘Ÿ|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|(๐‘ฅ)=|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|

ๅ› ่€Œ putting ๐‘Ÿโ†’0+, we have ๐ด๐‘Ÿ|๐‘“โˆ’๐‘“(๐‘ฅ)|(๐‘ฅ)โ†’๐‘„(๐‘ฅ), ไปฅๅŠ ๐ด๐‘Ÿ|๐‘“โˆ’๐‘”|(๐‘ฅ)โ†’๐ป(๐‘“โˆ’๐‘”)(๐‘ฅ); ไปŽ่€ŒไธŠ่ฟฐไธ็ญ‰ๅผๅ˜ไธบ:

๐‘„(๐‘ฅ)โ‰ค๐ป(๐‘“โˆ’๐‘”)(๐‘ฅ)+0+|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|

ไปŽ่€Œไธ€ๅฎšๆœ‰:

๐‘š({๐‘„>๐›ผ})โ‰ค๐‘š({๐ป(๐‘“โˆ’๐‘”)>๐›ผ2})+๐‘š({|๐‘“โˆ’๐‘”|>๐›ผ2})

ๆˆ‘ไปฌๅˆ†ๅˆซไปฅ HL max Thm ๅ’Œ Chebyshevโ€™s Thm bound ไฝๅณ่พน่ฟ™ไธคไธชๅผๅญ, ๅพ—ๅˆฐ:

๐‘š({๐‘„>๐›ผ})โ‰ค2โ‹…3๐‘›๐›ผโˆซ|๐‘“โˆ’๐‘”|+2๐›ผโˆซ|๐‘“โˆ’๐‘”|โ‰ค2๐›ผ(3๐‘›+1)๐œ–โ†’๐œ–โ†’00

ไปŽ่€Œๅพ—่ฏ.

โ–ก

Corollary 8.24
๐‘ฅโˆˆ๐ฟ๐‘™๐‘œ๐‘1(โ„๐‘›))โŸนlim๐‘Ÿโ†’0+๐ด๐‘Ÿ๐‘“(๐‘ฅ)=๐‘“(๐‘ฅ)๐‘Ž.๐‘’.
Proof

If ๐‘ฅโˆˆ๐ฟ๐‘“ then

|avg๐ต(๐‘ฅ,๐‘Ÿ)(๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ))๐‘‘๐‘ฆ|โ‰คavg๐ต(๐‘ฅ,๐‘Ÿ)|(๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฆโ†’๐‘Ÿโ†’00

โ–ก

8.2.2 density of a set at a point

Definition 8.43 : density of a set at a point

ๅฏนไบŽ ๐ธโŠ‚โ„๐‘› Lebesgue measurable (which implies: ๐œ’๐ธโˆˆ๐ฟ๐‘™๐‘œ๐‘1), ๆˆ‘ไปฌๅฎšไน‰:

๐ท๐ธ(๐‘ฅ):=lim๐‘Ÿโ†’0+๐‘š(๐ธโˆฉ๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))
Corollary 8.25

ๅฏนไบŽ๐ธโŠ‚โ„๐‘› Lebesgue measurable (which implies: ๐œ’๐ธโˆˆ๐ฟ๐‘™๐‘œ๐‘1), ไธ€ๅฎšๆœ‰:

๐ท๐ธ(๐‘ฅ)={1,for a.e. ๐‘ฅโˆˆ๐ธ0,for a.e. ๐‘ฅโˆˆ๐ธ๐‘
Proof

ๅ› ไธบ่ฟ™ไธช indicator function ๆ˜ฏ measurable ็š„, ไปฅๅŠ locally ๐ฟ1 ็š„. ๆ‰€ไปฅๅฎƒๅœจ ๐‘ฅ ๅค„็š„ density ๅฐฑๅ˜ๆˆไบ†ๅฎƒๅœจ ๐‘ฅ ๅค„็š„ๅ‡ๅ€ผ, ไปŽ่€Œๅœจ ๐ธ ไธŠ a.e. ไธบ 1, ๅœจ ๐ธ๐‘ ไธŠ a.e. ไธบ 0 (ๅ‡ฝๆ•ฐๅ€ผ).

โ–ก

Example 8.24

ๆˆ‘ไปฌ่ฟ™้‡Œไป‹็ปไธ€ไบ› behavior ๆฏ”่พƒ็‰นๆฎŠ็š„้›†ๅˆ, ็ฉบ้—ดๆฏ็‚นไธŠ่ฟ™ไธช้›†ๅˆ็š„ density.
่€ƒ่™‘

๐ธ:={0}โˆชโ‹ƒ๐‘—=0โˆž[23โ‹…12๐‘—,12๐‘—]

่ฟ™ๆ˜ฏไธ€ไธช closed set.

Figureย 24:

ๅฏนไบŽ ๐‘ฅโˆˆ๐ธโˆ˜, ๐ท๐ธ(๐‘ฅ)=1.
ๅฏนไบŽ ๐‘ฅโˆ‰๐ธ, ๐ท๐ธ(๐‘ฅ)=0.
ๅฏนไบŽ ๐‘ฅโˆˆ๐œ•๐ธ, ๐ท๐ธ(๐‘ฅ)=12. (ๅŒบ้—ด็š„ไธ€่พนๅœจ ๐ธ ้‡Œ, ไธ€่พนไธๅœจ ๐ธ ้‡Œ)
ๅฏนไบŽ ๐‘ฅโˆˆ0, ๐ท๐ธ(0) undefined. ๅ› ไธบๆฏๆฎต็ฉบๅฟƒๅ’Œๅฎžๅฟƒ็š„ๅœฐๆ–น, ่ฟ™ไธชๆฏ”ไพ‹็š„่ฝๅทฎ้ƒฝ้žๅธธๅคง. (ๅฎนๆ˜“่ฏๆ˜Ž่ฟ™ไธชๆž้™ไธๅญ˜ๅœจ.)
่€Œๅ่ง‚, ไปปๅ– ๐›ผโˆˆ(0,1), ้‚ฃไนˆๅ–

๐ธโ‰”โ‹ƒ๐‘›=1โˆž(1๐‘›,1๐‘›+๐›ผ๐‘›(๐‘›โˆ’1))

ๅˆ™ๆœ‰:

๐ท๐ธ(0)=๐›ผ2
Figureย 25:

่ฟ™้‡Œ็š„ๅ…ณ้”ฎๅœจไบŽ๏ผŒharmonic seq ้š็€ ๐‘› ็š„ๅขž้•ฟ่€Œ็ผฉๅฐ็š„้€Ÿๅบฆ้žๅธธๆ…ข. ๅœจ ๐‘› ่พƒๅคง็š„ๆƒ…ๅ†ตไธ‹, ่ƒŒๆ™ฏๅŒบ้—ด ๐ฝ๐‘› ๅ‡ ไนŽไธŽ ๐ฝ๐‘›+1 ๅ…ทๆœ‰็›ธๅŒ็š„้•ฟๅบฆ, ๅ› ๆญคๆญฃๅฆ‚ๆˆ‘ไปฌๆ‰€็Ÿฅ๏ผŒ๐‘š(๐ฝ๐‘›)/๐‘š(โˆช๐‘˜>๐‘๐ฝ๐‘˜)=0. ๆ‰€ไปฅ, ๆ— ่ฎบ ๐‘Ÿ ไฝไบŽๅฎžๅฟƒ้ƒจๅˆ† ๐ผ๐‘› ่ฟ˜ๆ˜ฏ็ฉบๅฟƒ้ƒจๅˆ† ๐ฝ๐‘›\๐ผ๐‘› ้ƒฝไธๅคช้‡่ฆ.
ๅฆไธ€ๆ–น้ข, ๆˆ‘ไปฌๅˆšๆ‰็š„ไพ‹ๅญไฝฟ็”จ geometric seq ไฝœไธบ่ƒŒๆ™ฏๅŒบ้—ด ๐ฝ๐‘› ็š„ๆž„ๅปบๅ—, ๅˆ™ fail, ๅ› ไธบ ๐ฝ๐‘› ็š„้•ฟๅบฆไธŽ โˆช๐‘˜โ‰ฅ๐‘›๐ฝ๐‘˜ ็›ธๆฏ”ๅคชๅคงไบ†, ๆœ‰ ๐‘š(๐ฝ๐‘›)=๐‘š(โˆช๐‘˜>๐‘›๐ฝ๐‘˜), ๅ› ๆญคๆ— ่ฎบ ๐‘Ÿ ไฝไบŽๅฎžๅฟƒๅŒบ้—ด่ฟ˜ๆ˜ฏ็ฉบๅฟƒๅŒบ้—ด ่ฟ™ไฝฟๅพ— 0 ๅค„็š„ๅฏ†ๅบฆๆ— ๆณ•ๅฎšไน‰.

8.2.3 generalized LDT

genralized LDT ่กจ็คบๅฏนๅฝข็Šถไธ่ง„ๅˆ™ (ๆœชๅฟ…ๆ˜ฏ ball) ็š„ๆ”ถๆ•›่กŒไธบ, LDT ็š„ statement ไป็„ถ stay true. ๅณ, ๅช่ฆ a family of Lebesgue mble sets ๐ธ๐‘Ÿ shrink nicely to ๐‘ฅ, LDT ๅฐฑๆปก่ถณ.

Theorem 8.42 : generalized LDT

ๅฏนไบŽ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(โ„๐‘›), ไปปๆ„็š„ ๐‘ฅโˆˆ๐ฟ๐‘“, ไปค {๐ธ๐‘Ÿ(๐‘ฅ)} ไธบ a family of Lebesgue measurable sets, ๅ…ถไธญๅฏนไบŽๆฏไธช ๐ธ๐‘Ÿ(๐‘ฅ) ้ƒฝๆœ‰:

๐ธ๐‘Ÿ(๐‘ฅ)โŠ‚๐ต(๐‘ฅ,๐‘Ÿ)

ๅนถไธ”

๐‘š(๐ธ๐‘Ÿ(๐‘ฅ))>๐›ผ๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

for some 0<๐›ผ<1.
ๅˆ™ๆœ‰:

lim๐‘Ÿโ†’0+โˆซ๐ธ๐‘Ÿ(๐‘ฅ)|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฆ๐‘š(๐ธ๐‘Ÿ(๐‘ฅ)=0for a.e.๐‘ฅ

่ฏๆ˜Žๅพˆ็ฎ€ๅ•, ๅ› ไธบ

lim๐‘Ÿโ†’0+โˆซ๐ธ๐‘Ÿ(๐‘ฅ)|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฆ๐‘š(๐ธ๐‘Ÿ(๐‘ฅ))โ‰คlim๐‘Ÿโ†’0+โˆซ๐ต(๐‘ฅ,๐‘Ÿ)|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฆ๐‘š(๐ธ๐‘Ÿ(๐‘ฅ))โ‰คlim๐‘Ÿโ†’0+โˆซ๐ต(๐‘ฅ,๐‘Ÿ)|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฆ๐›ผ๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=0

Homework 7: on differentiaion (50/50)

None of the following questions will be graded. Do them, but do not hand them in.

Completion of (๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€,๐œ‡ร—๐œˆ) = Completion of (๐‘‹ร—๐‘Œ,๐’œ๏ธ€ฬ„โŠ—โ„ฌ๏ธ€ฬ„,๐œ‡ฬ„ร—๐œˆฬ„)

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) and (๐‘Œ,โ„ฌ๏ธ€,๐œˆ) be measure spaces. Let (๐‘‹,๐’œ๏ธ€ฬ„,๐œ‡ฬ„) and (๐‘Œ,โ„ฌ๏ธ€ฬ„,๐œˆฬ„) be their completions, respectively. Then, the completion of (๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€,๐œ‡ร—๐œˆ) is same as the completion of (๐‘‹ร—๐‘Œ,๐’œ๏ธ€ฬ„โŠ—โ„ฌ๏ธ€ฬ„,๐œ‡ฬ„ร—๐œˆฬ„).

Modified HL maximal inequality (โ‰ฅ instead of >)

Prove that there is a constant ๐ถ๐‘›>0 that only depends on ๐‘› such that for every ๐‘“โˆˆ๐ฟ1(โ„๐‘›) and ๐›ผ>0,

๐‘š({๐‘ฅโˆˆโ„๐‘›โˆฃ๐ป๐‘“(๐‘ฅ)โ‰ฅ๐›ผ})โ‰ค๐ถ๐‘›๐›ผโˆซโ„๐‘›|๐‘“(๐‘ฅ)|๐‘‘๐‘ฅ

(Remark: We had ๐ป๐‘“(๐‘ฅ)>๐›ผ for the HL maximal inequality. Here we have ๐ป๐‘“(๐‘ฅ)โ‰ฅ๐›ผ.)

density of a mble set at a point: ๐ท๐ธ(๐‘ฅ)=1 for a.e. ๐‘ฅโˆˆ๐ธ, 0 for a.e. ๐‘ฅโˆˆ๐ธ๐‘

For a Lebesgue measurable subset ๐ธ of โ„๐‘›, the density of ๐ธ at ๐‘ฅ is defined as

๐ท๐ธ(๐‘ฅ)=lim๐‘Ÿโ†’0๐‘š(๐ธโˆฉ๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ)

provided that the limit exists. Prove that ๐ท๐ธ(๐‘ฅ)=1 for a.e. ๐‘ฅโˆˆ๐ธ and ๐ท๐ธ(๐‘ฅ)=0 for a.e. ๐‘ฅโˆˆ๐ธ๐‘. Hint: ask Lebesgue.

Some of the following questions will be graded. Do them, and do hand them in.

An identity: โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฅ๐‘‘๐‘ฅ=14log(1+๐‘ โˆ’2)

Prove that โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฅ๐‘‘๐‘ฅ=14log(1+๐‘ โˆ’2) for ๐‘ >0 by integrating the function ๐‘’โˆ’2๐‘ ๐‘ฅsin(2๐‘ฅ๐‘ฆ) with respect to ๐‘ฅ and ๐‘ฆ over suitable regions.

Proof

For fixed ๐‘ฅ>0, by FTC we have:

sin2(๐‘ฅ)=โˆซ0๐‘ฅsin(2๐‘ก)๐‘‘๐‘ก

We do change of variable ๐‘ก=๐‘ฅ๐‘ฆ. This is a valid diffeomorphism mapping ๐‘ฆโˆˆ(0,1) to ๐‘กโˆˆ(0,๐‘ฅ).
Then by change of variable theorem we have:

โˆซ(0,๐‘ฅ)sin(2๐‘ก)๐‘‘๐‘ก=โˆซ(0,1)๐‘ฅsin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฆ

Thus

sin2๐‘ฅ๐‘ฅ=โˆซ01sin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฆ

Then we get:

โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฅ๐‘‘๐‘ฅ=โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅ[โˆซ01sin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฆ]๐‘‘๐‘ฅ

Consider the function

๐‘“(๐‘ฅ,๐‘ฆ)โ‰”๐‘’โˆ’2๐‘ ๐‘ฅsin(2๐‘ฅ๐‘ฆ),(๐‘ฅ,๐‘ฆ)โˆˆ(0,โˆž)ร—(0,1)

๐‘“ is a composition of continuous functions, thus continuous. Note that it is also in ๐ฟ1((0,โˆž)ร—(0,1)) since |๐‘“(๐‘ฅ,๐‘ฆ)| is bounded by ๐‘”(๐‘ฅ,๐‘ฆ)โ‰”๐‘’โˆ’2๐‘ ๐‘ฅ, which is ๐ฟ1 on the same domain (its integral is 12๐‘ ), then by DCT, ๐‘“โˆˆ๐ฟ1((0,โˆž)ร—(0,1)).
Thus we can apply Fubiniโ€™s theorem to switch the order of integration:

โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅ[โˆซ01sin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฆ]๐‘‘๐‘ฅ=โˆซ(0,โˆž)ร—(0,1)๐‘’โˆ’2๐‘ ๐‘ฅsin(2๐‘ฅ๐‘ฆ)๐‘‘(๐‘ฅร—๐‘ฆ)=โˆซ01(โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฅ)๐‘‘๐‘ฆ

Recall back in Calculus we use integration by part to get:

โˆซ0โˆž๐‘’โˆ’๐‘Ž๐‘ฅsin(๐‘๐‘ฅ)๐‘‘๐‘ฅ=๐‘๐‘Ž2+๐‘2

for ๐‘Ž>0. In our case, ๐‘Ž=2๐‘  and ๐‘=2๐‘ฆ. Thus

โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฅ=2๐‘ฆ(2๐‘ )2+(2๐‘ฆ)2=๐‘ฆ2(๐‘ 2+๐‘ฆ2)

Therefore we here get

โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฅ๐‘‘๐‘ฅ=โˆซ01(โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin(2๐‘ฅ๐‘ฆ)๐‘‘๐‘ฅ)๐‘‘๐‘ฆ=โˆซ01๐‘ฆ2(๐‘ 2+๐‘ฆ2)๐‘‘๐‘ฆ=12โˆซ01๐‘ฆ๐‘ 2+๐‘ฆ2๐‘‘๐‘ฆ

By Calculus we have (by chain rule):

โˆซ01๐‘ฆ๐‘ 2+๐‘ฆ2๐‘‘๐‘ฆ=[12log(๐‘ 2+๐‘ฆ2)]01=12log(๐‘ 2+1๐‘ 2)=12log(1+1๐‘ 2)

Thus we conclude:

โˆซ0โˆž๐‘’โˆ’2๐‘ ๐‘ฅsin2๐‘ฅ๐‘ฅ๐‘‘๐‘ฅ=12โˆซ01๐‘ฆ๐‘ 2+๐‘ฆ2๐‘‘๐‘ฆ=12โ‹…12log(1+1๐‘ 2)=14log(1+1๐‘ 2)

as desired.

โ–ก

๐ธโˆˆ๐’œ๏ธ€โŠ—๐’œ๏ธ€โŸนdiagonal of ๐ธโˆˆ๐’œ๏ธ€

  • Prove that if ๐ธโˆˆ๐’œ๏ธ€โŠ—๐’œ๏ธ€, then

    {๐‘ฅโˆˆ๐‘‹:(๐‘ฅ,๐‘ฅ)โˆˆ๐ธ}โˆˆ๐’œ๏ธ€
  • Using this fact, find an example of a subset ๐ธโŠ‚โ„ร—โ„ such that ๐ธ๐‘ฅโˆˆโ„’๏ธ€(โ„) for all ๐‘ฅโˆˆโ„ and ๐ธ๐‘ฆโˆˆโ„’๏ธ€(โ„) for all ๐‘ฆโˆˆโ„, but ๐ธโˆ‰โ„’๏ธ€(โ„)โŠ—โ„’๏ธ€(โ„). Hint: ask Vitali.

Proof

of (a):
We consider the map:

๐œ™:๐‘‹โ†’๐‘‹ร—๐‘‹๐‘ฅโ†ฆ(๐‘ฅ,๐‘ฅ)

Then it suffices to show that ๐œ™ is (๐’œ๏ธ€,๐’œ๏ธ€โŠ—๐’œ๏ธ€)-measurable. Since if so, then for each ๐ธโˆˆ๐’œ๏ธ€โŠ—๐’œ๏ธ€, ๐œ™โˆ’1(๐ธ)={๐‘ฅโˆˆ๐‘‹:(๐‘ฅ,๐‘ฅ)โˆˆ๐ธ}โˆˆ๐’œ๏ธ€, which is exactly what we want.
Let ๐ดร—๐ตโˆˆ๐’œ๏ธ€โŠ—๐’œ๏ธ€ be a measurable rectangle, we discover that:

๐œ™โˆ’1(๐ดร—๐ต)={๐‘ฅโˆˆ๐‘‹:๐‘ฅโˆˆ๐ด,๐‘ฅโˆˆ๐ต}=๐ดโˆฉ๐ตโˆˆ๐’œ๏ธ€
Figureย 26:

We first prove a lemma:

Lemma 8.27

Suppose ๐‘“:๐‘‹โ†’๐‘Œร—๐‘ is a function from a measurable space (๐‘‹,๐’œ๏ธ€) to a product measure space (๐‘Œร—๐‘,โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2).
Claim: If ๐‘“โˆ’1(๐ต1ร—๐ต2)โˆˆ๐’œ๏ธ€ for each measurable rectangle ๐ต1ร—๐ต2โˆˆโ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2, then ๐‘“ is an (๐’œ๏ธ€,โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2)-measurable function.

Proof

of Lemma:
Since ๐‘“โˆ’1(๐ตร—๐ถ)โˆˆ๐’œ๏ธ€ for each measurable rectangle ๐ต1ร—๐ต2โˆˆโ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2, the preimage of any countable disjoint unions of measurable rectangles, is also in ๐’œ๏ธ€, since ๐’œ๏ธ€ is an ๐œŽ-algebra.
We want to show: ๐‘“โˆ’1(๐ธ)โˆˆ๐’œ๏ธ€ for any ๐ธโˆˆโ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2. It is equivalent to show that

โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2โŠ‚๐’ž๏ธ€:={๐ธโˆˆ๐‘Œร—๐‘:๐œ™โˆ’1(๐ธ)โˆˆ๐’œ๏ธ€}

Note that, it suffices to show that: ๐’ž๏ธ€ is an ๐œŽ-algebra. This is because we have shown

{all disjoint unions of measurable rectangles in ๐‘Œร—๐‘}โŠ‚๐’ž๏ธ€

, and this is an algebra generating โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2. Thus, if ๐’ž๏ธ€ is an ๐œŽ-algebra, we must have โ„ฌ๏ธ€1โŠ—โ„ฌ๏ธ€2โŠ‚๐’ž๏ธ€.
And since {all disjoint unions of measurable rectangles in ๐‘Œร—๐‘} is an algebra, it suffices to show that ๐’ž๏ธ€ is a monotone class, by the monotone class lemma.
Suppose ๐ธ1โІ๐ธ2โІโ‹ฏ with each ๐ธ๐‘›โˆˆ๐’ž๏ธ€, i.e. ๐œ™โˆ’1(๐ธ๐‘›)โˆˆ๐’œ๏ธ€. Since {๐ธ๐‘›} is increasing, we hve

๐œ™โˆ’1(๐ธ1)โІ๐œ™โˆ’1(๐ธ2)โІโ‹ฏโІ๐œ™โˆ’1(๐ธ๐‘›)โІโ‹ฏ

Since ๐’œ๏ธ€ is an ๐œŽ-algebra, we have

๐œ™โˆ’1(โ‹ƒ๐‘›=1โˆž๐ธ๐‘›)=โ‹ƒ๐‘›=1โˆž๐œ™โˆ’1(๐ธ๐‘›)โˆˆ๐’œ๏ธ€

Thus

โ‹ƒ๐‘›=1โˆž๐ธ๐‘›โˆˆ๐’ž๏ธ€

This is dually true for decreasing intersection, finishing the proof that ๐’ž๏ธ€ is a monotone class thus ๐œŽ-algebra, thus proving the lemma.

โ–ก

After we proved the Lemma, we return to the original statement, concluding that ๐œ™ is (๐’œ๏ธ€,๐’œ๏ธ€โŠ—๐’œ๏ธ€)-measurable, thus finishing the proof: if ๐ธโˆˆ๐’œ๏ธ€โŠ—๐’œ๏ธ€, then

{๐‘ฅโˆˆ๐‘‹:(๐‘ฅ,๐‘ฅ)โˆˆ๐ธ}โˆˆ๐’œ๏ธ€

โ–ก

Solution

of (b):
Take a Vitali set ๐‘‰โŠ‚โ„, and consider:

๐ธโ‰”{(๐‘ฅ,๐‘ฆ)โˆˆโ„2:๐‘ฅโ‰ ๐‘ฆ}โˆช{(๐‘ฅ,๐‘ฅ):๐‘ฅโˆˆ๐‘‰}.
Figureย 27:

Then for any fixed ๐‘ฅโˆˆโ„, we have:

๐ธ๐‘ฅ={๐‘ฆ:(๐‘ฅ,๐‘ฆ)โˆˆ๐ธ}={โ„,๐‘ฅโˆˆ๐‘‰โ„\{๐‘ฅ},๐‘ฅโˆ‰๐‘‰

And for any fixed ๐‘ฆโˆˆโ„, we have:

๐ธ๐‘ฆ={๐‘ฅ:(๐‘ฅ,๐‘ฆ)โˆˆ๐ธ}={โ„,๐‘ฆโˆˆ๐‘‰โ„\{๐‘ฆ},๐‘ฆโˆ‰๐‘‰

Thus ๐ธ๐‘ฅโˆˆโ„’๏ธ€(โ„) for all ๐‘ฅโˆˆโ„ and ๐ธ๐‘ฆโˆˆโ„’๏ธ€(โ„) for all ๐‘ฆโˆˆโ„.
However, we have ๐ธโˆ‰โ„’๏ธ€(โ„)โŠ—โ„’๏ธ€(โ„), since by (a) we have proved that if ๐ธโˆˆโ„’๏ธ€(โ„)โŠ—โ„’๏ธ€(โ„), then

๐‘‰={๐‘ฅโˆˆโ„:(๐‘ฅ,๐‘ฅ)โˆˆ๐ธ}โˆˆโ„’๏ธ€(โ„)

But it contradicts with the fact that ๐‘‰ is not Lebesgue measurable.
Thus ๐ธ satisfies our requirements.
(This happends since, as shown in class, the product measure space of two complete measure space is not necesarily complete. Here, the diagonal is a null set in โ„2 and thus our Vitali portion is a subnull set, but โ„’๏ธ€(โ„)โŠ—โ„’๏ธ€(โ„) is not complete (its completion is โ„’๏ธ€(โ„2).)

Too dense: ๐‘š(๐ธโˆฉ๐ผ)โ‰ค๐›ผ๐‘š(๐ผ) for all ๐ผ โŸน๐‘š(๐ธ)=0 for mble ๐ธ

Prove that if ๐ธโŠ‚โ„’๏ธ€(โ„) is a Lebesgue measurable subset such that

๐‘š(๐ธโˆฉ๐ผ)โ‰ค0.123๐‘š(๐ผ)

for all open intervals ๐ผโŠ‚โ„’๏ธ€(โ„), then ๐‘š(๐ธ)=0.

Proof

Since ๐ธ is Lebesgue measurable, ๐‘š(๐ธ)=๐‘šโˆ—(๐ธ).
Let ๐œ–>0.
Then by definition of outer mesure, we can pick open intervals seq {๐ผ๐‘˜}๐‘˜=1โˆž covering ๐ธ s.t.

๐‘š(๐ธ)>โˆ‘๐‘˜=1โˆž๐‘š(๐ผ๐‘˜)โˆ’๐œ–

Since ๐ธโŠ‚โ‹ƒ๐‘˜๐ผ๐‘˜, we have

๐ธ=(โ‹ƒ๐‘˜๐ผ๐‘˜)โˆฉ๐ธ=โ‹ƒ๐‘˜(๐ผ๐‘˜โˆฉ๐ธ)

Thus

๐‘š(๐ธ)=๐‘š(โ‹ƒ๐‘˜(๐ผ๐‘˜โˆฉ๐ธ))โ‰คโˆ‘๐‘˜๐‘š(๐ผ๐‘˜โˆฉ๐ธ)by ctbl subadditivity โ‰ค0.123โˆ‘๐‘˜๐‘š(๐ผ๐‘˜)by our requirement

Thus we have:

โˆ‘๐‘˜๐‘š(๐ผ๐‘˜)โˆ’๐œ–<0.123โˆ‘๐‘˜๐‘š(๐ผ๐‘˜)0.877โˆ‘๐‘˜๐‘š(๐ผ๐‘˜)<๐œ–โˆ‘๐‘˜๐‘š(๐ผ๐‘˜)<๐œ–0.877

Thus

๐‘š(๐ธ)โ‰คโˆ‘๐‘˜๐‘š(๐ผ๐‘˜)<๐œ–0.877

Since ๐œ–>0 is arbitrary, this proves that

๐‘š(๐ธ)=0

โ–ก

็ป™ๅฎšไปปๆ„ 0<๐›ผ<1, prescribe ๅ‡บไธ€ไธชๅœจ 0 ๅค„ density ไธบ ๐›ผ/2 ็š„้›†ๅˆ

Let 0<๐›ผ<1. Find an example of a Lebesgue measurable subset ๐ธ of [0,โˆž)โŠ‚โ„’๏ธ€(โ„) whose density at 0 is ๐›ผ/2. Hint: Consider ๐ธ=โ‹ƒ๐‘›=1โˆž๐ผ๐‘›. where ๐ผ๐‘›=(๐‘ฅ๐‘›,๐‘ฅ๐‘›+๐›ฟ๐‘›) are disjoint small intervals accumulating at 0.

Proof

Consider take

๐ธโ‰”โ‹ƒ๐‘›=1โˆž(1๐‘›,1๐‘›+๐›ผ๐‘›(๐‘›โˆ’1))

as the union of a countable sequence of intervals drawing near 0.
Notice: There intervals are mutually disjoint, since

1๐‘›โˆ’1โˆ’1๐‘›=1๐‘›(๐‘›โˆ’1)>๐›ผ๐‘›(๐‘›โˆ’1)

we thus have for ๐‘›โ‰ฅ2,

1๐‘›+๐›ผ๐‘›(๐‘›โˆ’1)<1๐‘›โˆ’1

We use ๐‘ฅ๐‘›:=1๐‘›; ๐ผ๐‘›โ‰”(๐‘ฅ๐‘›,๐‘ฅ๐‘›+๐›ฟ๐‘›) to denote each component interval; ๐ฝ๐‘›:=(๐‘ฅ๐‘›,๐‘ฅ๐‘›โˆ’1) to denote the open interval where ๐ผ๐‘› is located at; and ๐›ฟ๐‘›โ‰”๐›ผ๐‘›(๐‘›โˆ’1) to denote the length of each interval. Note that for each ๐‘›,

๐›ฟ๐‘›=๐›ผ(1๐‘›โˆ’1โˆ’1๐‘›)=๐›ผ(๐‘ฅ๐‘›โˆ’1โˆ’๐‘ฅ๐‘›)=๐›ผ๐ฝ๐‘›
Figureย 28:

Now we show that this set has Lebesgue density ๐›ผ2 at 0 below.
Let ๐‘Ÿ>0 (WLOG ๐‘Ÿ<1), then we have

1๐‘›+1<๐‘Ÿโ‰ค1๐‘› for some ๐‘›โˆˆโ„•

Then for each ๐‘˜โ‰ฅ๐‘›+2, we have 1๐‘˜<1๐‘›+1<๐‘Ÿ. Hence ๐ผ๐‘˜ is entirely contained in (0,๐‘Ÿ):

โ‹ƒ๐‘˜=๐‘›+2โˆž๐ผ๐‘˜โІ๐ธโˆฉ(โˆ’๐‘Ÿ,๐‘Ÿ)

We know that by telescoping,

โˆ‘๐‘˜=๐‘›+2โˆž1๐‘˜(๐‘˜โˆ’1)=(1๐‘›+1โˆ’1๐‘›+2)+(1๐‘›+2โˆ’1๐‘›+3)+โ‹ฏ=1๐‘›+1

Multiplying this by ๐›ผ2 gives:

โˆ‘๐‘˜=๐‘›+2โˆž๐›ผ๐‘˜(๐‘˜โˆ’1)=๐›ผ๐‘›+1

Thus by monotonicity of measure:

๐‘š(๐ธโˆฉ(โˆ’๐‘Ÿ,๐‘Ÿ))โ‰ฅ๐›ผ๐‘›+1

And for each ๐‘˜โ‰ค๐‘›, ๐ผ๐‘˜ exceeds (0,๐‘Ÿ) on the right, thus we get dually:

๐‘š(๐ธโˆฉ(โˆ’๐‘Ÿ,๐‘Ÿ))โ‰ค๐›ผ๐‘›โˆ’1

And we have:

2๐‘›+1โ‰ค๐‘š(โˆ’๐‘Ÿ,๐‘Ÿ)โ‰ค2๐‘›

since 1๐‘›+1โ‰ค๐‘Ÿโ‰ค1๐‘›.
Therefore we get:

๐›ผ๐‘›+12๐‘›โ‰ค๐‘š(๐ธโˆฉ(โˆ’๐‘Ÿ,๐‘Ÿ))๐‘š((โˆ’๐‘Ÿ,๐‘Ÿ))โ‰ค๐›ผ๐‘›โˆ’12๐‘›+1

Further simplify:

๐‘›๐‘›+1โ‹…๐›ผ2โ‰ค๐‘š(๐ธโˆฉ(โˆ’๐‘Ÿ,๐‘Ÿ))๐‘š((โˆ’๐‘Ÿ,๐‘Ÿ))โ‰ค๐‘›+1๐‘›โˆ’1โ‹…๐›ผ2

As ๐‘Ÿโ†’0+, we must have ๐‘›โ†’โˆž, and we know

lim๐‘›โ†’โˆž๐‘›๐‘›+1โ‹…๐›ผ2=lim๐‘›โ†’โˆž๐‘›+1๐‘›โˆ’1โ‹…๐›ผ2=๐›ผ2

Thus by Squeeze Theorem, we have:

lim๐‘Ÿโ†’0+๐‘š(๐ธโˆฉ(โˆ’๐‘Ÿ,๐‘Ÿ))๐‘š((โˆ’๐‘Ÿ,๐‘Ÿ))=๐›ผ2

Hence by def, ๐ธ indeed has Lebesgue density ๐›ผ/2 at 0.
(My note: The key point here is that, the harmonic seq shrinks very slowly in proportion as ๐‘› grows, ๐ฝ๐‘› almost have same length as ๐ฝ๐‘›+1 for large ๐‘›, thus ๐‘š(๐ฝ๐‘›)/๐‘š(โˆช๐‘˜>๐‘๐ฝ๐‘˜)=0 as we knows, so that whether ๐‘Ÿ lies in ๐ผ๐‘› or ๐ฝ๐‘›\๐ผ๐‘› does not quite matter.
On the other hand, the counterexample in class, using the geometric sequence as build block of ๐ฝ๐‘›, fails since the length of ๐ฝ๐‘› is too much compared to โˆช๐‘˜โ‰ฅ๐‘›๐ฝ๐‘˜, actually ๐‘š(๐ฝ๐‘›)=๐‘š(โˆช๐‘˜>๐‘›๐ฝ๐‘˜), thus whether ๐‘Ÿ lies in ๐ผ๐‘› or ๐ฝ๐‘›\๐ผ๐‘› makes a lot difference, making the density at 0 undefined.)

โ–ก

Seqs of complex numbers: โ„“1โŠŠโ‹‚1<๐‘<โˆžโ„“๐‘ and โ‹ƒ1<๐‘<โˆžโ„“๐‘โŠŠโ„“โˆž

  • Prove that โ„“1โŠŠโ‹‚1<๐‘<โˆžโ„“๐‘.

  • Prove that โ‹ƒ1<๐‘<โˆžโ„“๐‘โŠŠโ„“โˆž.

Proof

of (a):
We first want to show: for any 1<๐‘<โˆž, we have:

โ„“1โІโ„“๐‘

Fix ๐‘>1.
Let (๐‘ฅ๐‘›)โˆˆโ„“1. By definition,

โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|<โˆž

We need to show that โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|๐‘<โˆž.
Claim: There are at most finitely many ๐‘›โˆˆโ„• s.t. |๐‘ฅ๐‘›|โ‰ฅ1.
Proof of Claim: Suppose for contradiction that there are inifinitely many ๐‘›โˆˆโ„• s.t. |๐‘ฅ๐‘›|โ‰ฅ1, say, all terms in the subseqence {๐‘ฅ๐‘›๐‘—}๐‘—=1โˆž has |๐‘ฅ๐‘›๐‘—|โ‰ฅ1. Then

โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|โ‰ฅโˆ‘๐‘—=1โˆž|๐‘ฅ๐‘›๐‘—|โ‰ฅโˆ‘๐‘—=1โˆž1=โˆž

which contradicts with (๐‘ฅ๐‘›)โˆˆโ„“1.
Thus, suppose only on the finite terms {๐‘ฅ๐‘›๐‘—}๐‘—=1๐‘ we have |๐‘ฅ๐‘›๐‘—|โ‰ฅ1 (WLOG ๐‘โ‰ฅ1). Then

โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|=โˆ‘๐‘—=1๐‘|๐‘ฅ๐‘›๐‘—|+โˆ‘๐‘›โ‰ ๐‘›๐‘— for any ๐‘—|๐‘ฅ๐‘›|

Since for ๐‘› s.t. n โ‰ ๐‘›๐‘— for any subseq index ๐‘—, we have |๐‘ฅ๐‘›|<1, for these indexes we have:

|๐‘ฅ๐‘›|๐‘<|๐‘ฅ๐‘›|for any ๐‘>1

Thus we have

โˆ‘๐‘›โ‰ ๐‘›๐‘— for any ๐‘—|๐‘ฅ๐‘›|๐‘<โˆ‘๐‘›โ‰ ๐‘›๐‘— for any ๐‘—|๐‘ฅ๐‘›|<โˆž

And also,

โˆ‘๐‘—=1๐‘|๐‘ฅ๐‘›๐‘—|๐‘<โˆž since only have finite terms

Thus

โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|๐‘=โˆ‘๐‘—=1๐‘|๐‘ฅ๐‘›๐‘—|๐‘+โˆ‘๐‘›โ‰ ๐‘›๐‘— for any ๐‘—|๐‘ฅ๐‘›|๐‘<โˆž

Thus

โ„“1โІโ„“๐‘

Since ๐‘>1 is arbitrary, this proves that

โ„“1โІโ‹‚1<๐‘<โˆžโ„“๐‘

To show the strictness of the inclusion, we consider the harmonic series โˆ‘๐‘›=1โˆž1๐‘›. We know that it diverges and for any ๐‘>1, the ๐‘-series โˆ‘๐‘›=1โˆž1๐‘›๐‘ (absolutely for sure) converges, thus (1๐‘›)โˆ‰โ„“1 but (1๐‘›)โˆˆโ„“๐‘ for every ๐‘>1, showing that

โ„“1โ‰ โ‹‚1<๐‘<โˆžโ„“๐‘

This finishes the proof that

โ„“1โŠŠโ‹‚1<๐‘<โˆžโ„“๐‘

โ–ก

Proof

of (b):
Fix ๐‘>1.
Suppose sequence (๐‘ฅ๐‘›) belongs โ„“๐‘, then

โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|๐‘<โˆž

This implies that ๐‘ฅ๐‘›โ†’0 as ๐‘›โ†’โˆž, because if it did not, there would be infinitely many terms where |๐‘ฅ๐‘›| is bounded away from zero, leading to divergence of the sum.
Suppose for contradiction that

sup๐‘›|๐‘ฅ๐‘›|=โˆž

Then there are infinitely many terms ๐‘› s.t. |๐‘ฅ๐‘›|>1, since otherwise, exists some ๐‘ s.t. all |๐‘ฅ๐‘›|โ‰ค1 for ๐‘›โ‰ฅ๐‘, then sup|๐‘ฅ๐‘›|โ‰คmax(1,max1โ‰ค๐‘›โ‰ค๐‘โˆ’1|๐‘ฅ๐‘›|)<โˆž.
Suppose for the subseq {๐‘ฅ๐‘›๐‘—}๐‘—=1โˆž we have |๐‘ฅ๐‘›๐‘—|>1. Thus

โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|๐‘โ‰ฅโˆ‘๐‘—=1โˆž|๐‘ฅ๐‘›๐‘—|๐‘>โˆ‘๐‘—=1โˆž1๐‘=โˆž

which contradicts with โˆ‘๐‘›=1โˆž|๐‘ฅ๐‘›|๐‘<โˆž. Therefore we have:

sup๐‘›|๐‘ฅ๐‘›|<โˆž

This shows that

โ„“๐‘โІโ„“โˆž

Since ๐‘>1 is arbitrary, this proves that

โ‹ƒ1<๐‘<โˆžโ„“๐‘โІโ„“โˆž

Now we show the inclusion is strict. Consider the sequence ๐‘ฅ๐‘›=1 for all ๐‘›. Clearly, (๐‘ฅ๐‘›)โˆˆโ„“โˆž because it is bounded. However, ๐‘ฅ๐‘›โˆ‰โ„“๐‘ for any ๐‘>1:

โˆ‘๐‘›=1โˆž|1|๐‘=โˆ‘๐‘›=1โˆž1=โˆž

This shows

โ‹ƒ1<๐‘<โˆžโ„“๐‘โ‰ โ„“โˆž

Thus we have

โ‹ƒ1<๐‘<โˆžโ„“๐‘โŠŠโ„“โˆž

โ–ก

Nur fรผr Verrรผckte

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!)

Prescribing a Lebesgue density, Season 2

Let 0<๐›ผ<1 and ๐‘›โ‰ฅ1. Find an example of a Lebesgue measurable subset ๐ธ of โ„’๏ธ€(โ„)๐‘› whose density at 0 is ๐›ผ. Hint: think spherically.

9 ๐ฟ๐‘ space and inequalities

9.1 Banach Space and ๐ฟ๐‘ space [Fol 5.1; 6.1]

ๅฏนๅบ” Folland 5.1(1), 6.1(1).

9.1.1 norm and completeness

Recall:

Definition 9.44 : semi-norm, norm

ไธ€ไธชsemi norm ๆ˜ฏไธ€ไธชๅ‡ฝๆ•ฐ ||โ‹…||:๐‘‰โ†’[0,โˆž) starting from a vector space ๐‘‰. ๅ…ถๆปก่ถณ (1): tri eq ๅ’Œ (2): homogeneity.
ๅฆ‚ๆžœไธ€ไธช semi-norm ๆปก่ถณ (3): ||๐‘ฃ||=0 iff ๐‘ฃ=0, ๅˆ™็งฐๅฎƒไธบไธ€ไธช norm.

Definition 9.45 : Banach space

ไธ€ไธช normed vector space (๐‘‰,||โ‹…||) ็š„ induced metric space ๅฆ‚ๆžœๆ˜ฏ complete ็š„, ๅฎƒๅฐฑ่ขซ็งฐไธบไธ€ไธช Banach space.

Example 9.25

โ„๐‘›,โ„‚๐‘› with Euclidean norm is a Banach space.
๐ถ0([0,1]): space of ctn functions on [0,1] equipped with sup norm is Banach.

||๐‘“โˆ’๐‘”||โ‰”sup๐‘ฅโˆˆ[0,1]|๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|

๐ถ๐‘0(โ„): space of ctn functions with cpt supp on โ„ equipped with sup norm is not Banach! ่ฟ™ๆ˜ฏๅ› ไธบ, ไธ€ไธชๆœ‰ cpt supp ็š„ function seq ็š„ๆž้™ๆœชๅฟ…ๆœ‰ cpt supp. ๆฏ”ๅฆ‚ (๐œ’[โˆ’๐‘›,๐‘›])๐‘›โˆˆโ„•.

Lemma 9.28

A metric space (๐‘‹,๐œŒ) is complete iff every Cauchy seq has a subseq that converges.

Proof

Trivial.
โŸน: Clear.
โŸธ: subseq conv dist bound + Cauchy dist bound can bound the whole tail with arbitrary ๐œ–.

โ–ก

่ฟ™ไธช statement, ็›ดๆŽฅๆŠŠ complete ็š„ๅฎšไน‰ไปŽๆฏไธช Cauchy seq ้ƒฝๆ”ถๆ•›, ไผ˜ๅŒ–ไธบๆฏไธช Cauchy seq ้ƒฝๆœ‰ไธ€ไธชๆ”ถๆ•› subseq.

9.1.2 every Cachy seq conv (complete) โ‡” every abs conv series convs

Definition 9.46 : series: convergence ๅ’Œ absolute convergence

ๅฏนไบŽไธ€ไธช normed VS (๐‘‰,||โ‹…||) ไธญ็š„ seq (๐‘ฃ๐‘›), ๆˆ‘ไปฌ็งฐ โˆ‘๐‘›=1โˆž๐‘ฃ๐‘› converges, ๅฆ‚ๆžœๅญ˜ๅœจ ๐‘ฃโˆˆ๐‘‰ s.t.

lim๐‘โ†’โˆžโˆ‘๐‘›=1๐‘๐‘ฃ๐‘›=๐‘ฃ

ๅณ

lim๐‘โ†’โˆžโˆฅ๐‘ฃโˆ’โˆ‘๐‘›=1๐‘๐‘ฃ๐‘›โˆฅ=0

ๆˆ‘ไปฌ็งฐ โˆ‘๐‘›=1โˆž๐‘ฃ๐‘› absolutely converges, ๅฆ‚ๆžœ

โˆ‘๐‘›=1โˆž||๐‘ฃ๐‘›||<โˆž

ๅณ่ฟ™ไธช series ๅฏนๅบ”็š„ norm series converges to some real number.

Theorem 9.44 : another criterion for Banach space

A normed VS (๐‘‰,||โ‹…||) is a Banach space iff every absolutely convergent series converges.

Proof

โ€œโŸน": ๅฆ‚ๆžœ (๐‘‰,||โ‹…||) is a Banach space, Suppose โˆ‘๐‘›=1โˆž||๐‘ฃ๐‘›||<โˆž, ๅ–้ƒจๅˆ†ๅ’Œๅบๅˆ—

๐‘†๐‘โ‰”โˆ‘๐‘›=1๐‘๐‘ฃ๐‘›

ๆœ‰

โˆฅ๐‘†๐‘šโˆ’๐‘†๐‘›โˆฅ=โˆฅโˆ‘๐‘˜=๐‘›+1๐‘š๐‘ฃ๐‘˜โˆฅโ‰คโˆ‘๐‘˜=๐‘›+1๐‘šโˆฅ๐‘ฃ๐‘˜โˆฅ

For large enough ๐‘š,๐‘› ่ฟ™ไธช bound ๅฏไปฅๆ— ้™ๅฐ, ๅ› ่€Œ (๐‘†๐‘) is Cauchy. โ€œโŸธ": ๅฆ‚ๆžœ (๐‘‰,||โ‹…||) ไธญ every absolutely convergent series converges.
Suppose (๐‘ฃ๐‘›) is Cauchy. WTS it converges.
By Cauchy, ๅญ˜ๅœจ subseq, say labeled ๐‘›1<๐‘›2<โ‹ฏ, s.t. ||๐‘ฃ๐‘šโˆ’๐‘ฃ๐‘›||<13๐‘—for all ๐‘š,๐‘›โ‰ฅ๐‘›๐‘— Then

โˆ‘๐‘—=1โˆž||๐‘ฃ๐‘›๐‘—+1โˆ’๐‘ฃ๐‘›๐‘—||<โˆž

Let (๐‘ฆ๐‘—) be s.t. ๐‘ฆ1=๐‘ฃ๐‘›1, ๐‘ฆ๐‘—=๐‘ฃ๐‘›๐‘—+1โˆ’๐‘ฃ๐‘›๐‘—, then

โˆ‘๐‘—=1โˆžโˆฅ๐‘ฆ๐‘—โˆฅโ‰คโˆฅ๐‘ฆ1โˆฅ+โˆ‘๐‘—12๐‘—=โˆฅ๐‘ฆ1โˆฅ+1<โˆž

ๅนถไธ”ๆœ‰:

๐‘ฃ๐‘›๐‘—=โˆ‘๐‘˜=1๐‘—๐‘ฆ๐‘˜

็”ฑไบŽ โˆ‘๐‘—=1โˆžโˆฅ๐‘ฆ๐‘—โˆฅ<โˆž, by our assumption ๅพ—ๅˆฐ, ่ฟ™ไธชๆž้™ lim๐‘—โ†’โˆž๐‘ฃ๐‘›๐‘—=โˆ‘๐‘˜=1โˆž๐‘ฆ๐‘˜ ๆ˜ฏๅญ˜ๅœจ็š„.

โ–ก

9.1.3 ไปปไฝ• finite dim normed VS ไธ€ๅฎš Banach, infinite dim ๅˆ™ไธไธ€ๅฎš Banach

ไธ‹้ขๆˆ‘ไปฌๅฐ†ไป‹็ปไธ€็ฑป infinite dimension ไฝ†ๆ˜ฏ Banach ็š„ normed VS: ๐ฟ๐‘ spaces.

9.1.4 ๐ฟ๐‘ spaces

Definition 9.48 : ๐ฟ๐‘ spaces

Consider ๐‘โˆˆ(0,โˆž).
Let (๐‘‹,๐’œ๏ธ€,๐œ‡) ไธบไธ€ไธช measure space.
Define for ๐‘“:๐‘‹โ†’โ„ measurable:

||๐‘“||๐‘:=(โˆซ|๐‘“|๐‘๐‘‘๐œ‡)1๐‘โˆˆ[0,โˆž]

Define

๐ฟ๐‘(๐œ‡):={๐‘“:||๐‘“||๐‘<โˆž}/โˆผ

where ๐‘“โˆผ๐‘” if ๐‘“=๐‘” a.e.

ๅ›บๅฎšไธ€ไธช measure space (๐‘‹,๐’œ๏ธ€,๐œ‡), ๆˆ‘ไปฌๅฐ†็”จ ๐ฟ๐‘ ๆฅ็ฎ€ๆ˜“ๆŒ‡ไปฃ ๐ฟ๐‘(๐œ‡).

Example 9.26

(๐‘‹,๐’œ๏ธ€,๐œ‡)โ‰”(โ„,โ„’๏ธ€,๐‘š),

๐‘“(๐‘ฅ):=1๐‘ฅ๐›ผ๐œ’(0,1),๐‘“โˆˆ๐ฟ๐‘(๐‘š)โ‡”๐›ผ๐‘<1๐‘“(๐‘ฅ):=1๐‘ฅ๐›ผ๐œ’(1,โˆž),๐‘“โˆˆ๐ฟ๐‘(๐‘š)โ‡”๐›ผ๐‘>1

(๐‘‹,๐’œ๏ธ€,๐œ‡)โ‰”(โ„•,๐’ซ๏ธ€(โ„•),๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘–๐‘›๐‘”),

๐ฟ๐‘(๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘–๐‘›๐‘”)={(๐‘Ž๐‘›)๐‘›โˆˆโ„•:โˆ‘๐‘›=1โˆž|๐‘Ž๐‘›|๐‘<โˆž}
Lemma 9.29 : ๐ฟ๐‘ space is a vector space

๐ฟ๐‘ space is a โ„‚-vector space.

Proof

Suppose ๐‘“,๐‘”โˆˆ๐ฟ๐‘.
็”ฑไบŽ

|๐‘“+๐‘”|๐‘โ‰ค(|๐‘“|+|๐‘”|)๐‘โ‰ค(2max{|๐‘“|,|๐‘”|})๐‘โ‰ค2๐‘(|๐‘“|๐‘+|๐‘”|๐‘)

ไบŽๆ˜ฏ by linearity of integral, ๅพ—ๅˆฐ:

๐‘“,๐‘”โˆˆ๐ฟ๐‘โŸน๐‘“+๐‘”โˆˆ๐ฟ๐‘

(Note: ๐‘>1 ๆ—ถไนŸๅฏไปฅ by |๐‘ฅ|๐‘ ่ฟ™ไธ€ๅ‡ฝๆ•ฐ็š„ convexity ๅพ—ๅˆฐ่ฟ™ไธช bound, ไฝ†ๆ˜ฏ่ฟ™ไธชๆ–นๆณ•ๅชๆœ‰ๆ•ˆไบŽ ๐‘>1)

โ–ก

ไฝ†ๆ˜ฏ Question 1: Is ๐ฟ๐‘ a normed VS? ๅณ, โˆฅโ‹…โˆฅ๐‘ ๆ€ปๆ˜ฏไธ€ไธช valid norm ๅ—? A: True for ๐‘โˆˆ[1,โˆž), false for ๐‘โˆˆ(0,1). Homogeneity ๅ’Œ โˆฅ๐‘“โˆฅ๐‘=0 iff ๐‘“=0 (a.e.) ๆ˜ฏๆ˜พ็„ถ็š„, ไฝ†ๆ˜ฏๆˆ‘ไปฌๅ‘็Žฐ, tri ineq ๆฒกๆœ‰ๆ˜พ็„ถ็š„่ฏๆ˜Ž.
Next lecture, we will show the Minkowskiโ€™s ineq, ๅณ ๐ฟ๐‘ space ไธŠ็š„ไธ‰่ง’ไธ็ญ‰ๅผ:

||๐‘“+๐‘”||๐‘โ‰ค||๐‘“||๐‘+||๐‘”||๐‘

ไฝ†ๆ˜ฏ่ฟ™ไธชไธ็ญ‰ๅผๅช hold for ๐‘โˆˆ[1,โˆž), ๅนถไธ” fail otherwise.
(ๅ› ่€ŒๅฏนไบŽ ๐ฟ๐‘ space ็š„็ ”็ฉถ, ๆˆ‘ไปฌๅฐ† focus on ๐‘โˆˆ[1,โˆž) ็š„ๆƒ…ๅ†ต.)
Question 2: Is ๐ฟ๐‘ space, ๐‘โˆˆ[1,โˆž), Banach? Answer: Yes.
ๆˆ‘ไปฌไนŸๅฐ†ๅœจ next lecture ่ฏๆ˜Žๅฎƒ.

9.2 inequilities on ๐ฟ๐‘ spaces [Fol 6.1]

ๅฏนๅบ” Folland 6.1(2).
ๆˆ‘ไปฌๅฐ†่ฏๆ˜Ž Hรถlderโ€™s ineq ไปฅๅŠๅฎƒ็š„ corollary Minkowskiโ€™s ineq, ไปŽ่€Œ่ฏๆ˜Ž: ๐ฟ๐‘ ๆ˜ฏไธ€ไธช normed VS, ๅนถไธ”ๆ˜ฏไธ€ไธช Banach space (่ฟ™้‡Œ 1โ‰ค๐‘<โˆž, ไฝ†ๆ˜ฏ later we will also prove ๐ฟโˆž ไนŸๆ˜ฏ Banach space).
่ฟ™ไธคไธชไธ็ญ‰ๅผ้žๅธธ้‡่ฆ.

9.2.1 Hรถlderโ€™s ineq

Theorem 9.46 : Hรถlderโ€™s ineq

Consider conjugate pair: ๐‘,๐‘žโˆˆ[1,โˆž) s.t.

1๐‘+1๐‘ž=1

ๅˆ™ๅฏนไบŽไปปๆ„ไธคไธช measurable function ๐‘“,๐‘”:๐‘‹โ†’โ„‚, ไธ€ๅฎšๆœ‰:

โˆฅ๐‘“๐‘”โˆฅ1โ‰คโˆฅ๐‘“โˆฅ๐‘โ‹…โˆฅ๐‘”โˆฅ๐‘ž

็‰นๅˆซๅœฐ, ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ๐‘(๐œ‡), ๐‘”โˆˆ๐ฟ๐‘ž(๐œ‡), ๅˆ™ ๐‘“๐‘”โˆˆ๐ฟ1(๐œ‡), ๅนถไธ” equality holds iff

โˆฅ๐‘”โˆฅ๐‘ž๐‘ž|๐‘“|๐‘=โˆฅ๐‘“โˆฅ๐‘๐‘|๐‘”|๐‘ž๐œ‡-a.e.
Proof

Trivial Case 1: ๅฆ‚ๆžœ โˆฅ๐‘“โˆฅ๐‘=0 (ๆˆ–่€…โˆฅ๐‘”โˆฅ๐‘ž=0 ), then ๐‘“ is zero ๐œ‡-almost everywhere, and the product ๐‘“๐‘” is zero ๐œ‡-almost everywhere, ไบŽๆ˜ฏไธค่พน้ƒฝๆ˜ฏ 0, ineq trivially true.
Trivial Case 2: ๅฆ‚ๆžœ โˆฅ๐‘“โˆฅ๐‘=โˆž or โˆฅ๐‘”โˆฅ๐‘ž=โˆž, ๅˆ™ๅณ่พน infinite, ineq trivially true. ๅ› ่€Œๆˆ‘ไปฌๅช้œ€่ฆ่€ƒ่™‘ โˆฅ๐‘“โˆฅ๐‘ and โˆฅ๐‘”โˆฅ๐‘ž are in (0,โˆž) ็š„ๆƒ…ๅ†ตๅฐฑๅฅฝไบ†.
Main case: ๆˆ‘ไปฌ้œ€่ฆไธ€ไธช Lemma:

Lemma 9.30 : Youngโ€™s inequality for products

Whenever ๐‘,๐‘žโˆˆ(1,โˆž) with 1๐‘+1๐‘ž=1, ้ƒฝๆœ‰

๐‘Ž๐‘โ‰ค๐‘Ž๐‘๐‘+๐‘๐‘ž๐‘ž,โˆ€๐‘Ž,๐‘โ‰ฅ0

where equality is achieved if and only if ๐‘Ž๐‘=๐‘๐‘ž.
ๅฆไธ€ไธช็ญ‰ไปทๅฝขๅผๆ˜ฏ:

๐‘Ž๐œ†๐‘1โˆ’๐œ†โ‰ค๐œ†๐‘Ž+(1โˆ’๐œ†)๐‘,โˆ€๐‘Ž,๐‘โ‰ฅ0
Proof

of Lemma:
๐‘=0 ๅˆ™ trivial case. ๅ› ่€Œ setting ๐‘ก:=๐‘Ž๐‘, reduced to show:

๐‘ก๐œ†โ‰ค๐œ†๐‘ก+(1โˆ’๐œ†)

with eq iff ๐‘ก=1. ่ฟ™ๆ˜ฏๆ˜พ็„ถ็š„, ๅ› ไธบ by Calculus, ๐‘ก๐œ†โˆ’๐œ†๐‘ก ๆ˜ฏ strictly increasing for ๐‘ก<1, strictly decreasing for ๐‘ก>1 ็š„, max ๅœจ ๐‘ก=1, ๆญฃๅฅฝๆ˜ฏ 1โˆ’๐œ†.

โ–ก

ไฝฟ็”จ Youngโ€™s inequality for products ๅพ—ๅˆฐ:

|๐‘“(๐‘ฅ)|โˆฅ๐‘“โˆฅ๐‘|๐‘”(๐‘ฅ)|โˆฅ๐‘”โˆฅ๐‘žโ‰ค|๐‘“(๐‘ฅ)|๐‘๐‘โˆฅ๐‘“โˆฅ๐‘๐‘+|๐‘”(๐‘ฅ)|๐‘ž๐‘žโˆฅ๐‘”โˆฅ๐‘ž๐‘ž,๐‘ฅโˆˆ๐‘‹

Integrating both sides gives

โˆฅ๐‘“๐‘”โˆฅ1โˆฅ๐‘“โˆฅ๐‘โˆฅ๐‘”โˆฅ๐‘žโ‰คโˆฅ๐‘“โˆฅ๐‘๐‘๐‘โˆฅ๐‘“โˆฅ๐‘๐‘+โˆฅ๐‘”โˆฅ๐‘ž๐‘ž๐‘žโˆฅ๐‘”โˆฅ๐‘ž๐‘ž=1๐‘+1๐‘ž=1,

which proves the claim.
Integration ็š„ equality holds iff point equality holds a.e., ๅนถไธ”, by Youngโ€™s inequality for products, ไธŠ้ข็š„ equality holds iff

โˆฅ๐‘”โˆฅ๐‘ž๐‘ž|๐‘“|๐‘=โˆฅ๐‘“โˆฅ๐‘๐‘|๐‘”|๐‘ž๐œ‡-a.e.

โ–ก

9.2.2 Minkowskiโ€™s ineq: tri ineq on ๐ฟ๐‘, ็กฎ่ฎค โˆฅโ‹…โˆฅ๐‘-norm ๆ˜ฏ ๐ฟ๐‘ ไธŠ็š„ valid norm

Minkowskiโ€™s ineq ๅณ ๐ฟ๐‘ space ไธŠ็š„ tri ineq.

Corollary 9.26 : Minkowski inequality

ๅฏนไบŽไปปๆ„ 1โ‰ค๐‘<โˆž, ้ƒฝๆœ‰:

โˆฅ๐‘“+๐‘”โˆฅโ‰คโˆฅ๐‘“โˆฅ๐‘+โˆฅ๐‘”โˆฅ๐‘
Proof

ๆ˜พ็„ถ, ๅฏนไบŽไปปๆ„ ๐‘ฅ ้ƒฝๆœ‰:

|๐‘“+๐‘”|๐‘โ‰ค(|๐‘“|+|๐‘”|)|๐‘“+๐‘”|๐‘โˆ’1

ๅ› ่€Œ:

โˆซ|๐‘“+๐‘”|๐‘โ‰คโˆซ|๐‘“|โ‹…|๐‘“+๐‘”|๐‘โˆ’1+โˆซ|๐‘”|โ‹…|๐‘“+๐‘”|๐‘โˆ’1

ๆˆ‘ไปฌๅฎšไน‰

โ„Ž(๐‘ฅ)โ‰”|๐‘“(๐‘ฅ)+๐‘”(๐‘ฅ)|๐‘โˆ’1

ไบŽๆ˜ฏ

โˆซ|๐‘“+๐‘”|๐‘โ‰คโˆซ|๐‘“โ„Ž|+โˆซ|๐‘”โ„Ž|โ‰คโˆฅ๐‘“โˆฅ๐‘โˆฅโ„Žโˆฅ๐‘ž+โˆฅ๐‘”โˆฅ๐‘โˆฅโ„Žโˆฅ๐‘ž=(โˆฅ๐‘“โˆฅ๐‘+โˆฅ๐‘”โˆฅ๐‘)(โˆซ|๐‘“+๐‘”|(๐‘โˆ’1)๐‘ž)1/๐‘ž

ๅ…ถไธญ ๐‘ž ๆ˜ฏ ๐‘ ็š„ Hรถlder conjugate. ่ฟ™้‡Œ็š„ punchline is actually: ็”ฑไบŽ

๐‘ž:=๐‘๐‘โˆ’1

actually,

(๐‘โˆ’1)๐‘ž=๐‘

ๅ› ่€Œ:

โˆซ|๐‘“+๐‘”|๐‘โ‰ค(โˆฅ๐‘“โˆฅ๐‘+โˆฅ๐‘”โˆฅ๐‘)(โˆซ|๐‘“+๐‘”|(๐‘โˆ’1)๐‘ž)1/๐‘ž=(โˆฅ๐‘“โˆฅ๐‘+โˆฅ๐‘”โˆฅ๐‘)(โˆซ|๐‘“+๐‘”|๐‘)1/๐‘ž=(โˆฅ๐‘“โˆฅ๐‘+โˆฅ๐‘”โˆฅ๐‘)(โˆซ|๐‘“+๐‘”|๐‘)1โˆ’1/๐‘

ไธค่พนๅŒๆ—ถ้™คไปฅ (โˆซ|๐‘“+๐‘”|๐‘)1โˆ’1/๐‘ ๅพ—ๅˆฐ:

(โˆซ|๐‘“+๐‘”|๐‘)1/๐‘=:โˆฅ๐‘“+๐‘”โˆฅ๐‘โ‰คโˆฅ๐‘“โˆฅ๐‘+โˆฅ๐‘”โˆฅ๐‘

ไปŽ่€Œๅพ—่ฏ.

โ–ก

9.2.3 properties of ๐ฟ๐‘ spaces (1โ‰ค๐‘<โˆž)

9.2.4 ๐ฟ๐‘ (1โ‰ค๐‘<โˆž) is Banach

Theorem 9.47 : ๐ฟ๐‘ space (1โ‰ค๐‘<โˆž) is Banach

๐ฟ๐‘ (1โ‰ค๐‘<โˆž) is Banach.

Proof

By last lec ็š„ๅฎš็†: ไธ€ไธช NVS ๆ˜ฏ Banach ็š„็ญ‰ไปทๆกไปถๆ˜ฏไปปๆ„ abs conv series ้ƒฝ conv. ๅ› ่€Œๆˆ‘ไปฌ่ฏๆ˜Ž่ฟ™ไธ€็‚นๅณๅฏ.
Suppose ๐‘“๐‘›โˆˆ๐ฟ๐‘ for each ๐‘›, ๅนถไธ”่ฟ™ไธช series abs conv, ๅณ:

๐ตโ‰”โˆ‘๐‘˜=1โˆžโˆฅ๐‘“๐‘˜โˆฅ๐‘<โˆž

ๆˆ‘ไปฌ define:

๐‘”(๐‘ฅ):=โˆ‘๐‘˜=1โˆž๐‘“๐‘˜(๐‘ฅ),๐‘”๐‘›(๐‘ฅ):=โˆ‘๐‘˜=1๐‘›๐‘“๐‘˜(๐‘ฅ)

ๆˆ‘ไปฌ WTS:

lim๐‘›โ†’โˆž๐‘”๐‘›=๐‘”

in ๐‘-norm induced metric sense, ๅณ, for some ๐‘“โˆˆ๐ฟ๐‘, ๆœ‰

lim๐‘›โ†’โˆžโˆฅ๐‘”โˆ’๐‘”๐‘›โˆฅ๐‘=0

ๆˆ‘ไปฌ Set:

๐บ๐‘›โ‰”โˆ‘๐‘˜=1๐‘›|๐‘“๐‘˜|,๐บโ‰”โˆ‘๐‘˜=1โˆž|๐‘“๐‘˜|

่ฟ™ไธชๅ‡ฝๆ•ฐไปฅๅŠๅ‡ฝๆ•ฐๅˆ—็š„ๅฎšไน‰ๆ˜ฏไธบไบ†ไฝฟ็”จ DCT, ไฝœ donimating function ็”จ.
By measurable function ็š„ limit behavior, ๆœ‰

๐บ๐‘›,๐บโˆˆ๐ฟ+

ๅนถไธ”

โˆฅ๐บ๐‘›โˆฅ๐‘โ‰คโˆ‘๐‘˜=1๐‘›โˆฅ๐‘“๐‘˜โˆฅ๐‘โ‰ค๐ต

็”ฑไบŽ ๐บ๐‘›โ†—๏ธŽ๐บ, by MCT ๆœ‰

โˆซ๐บ๐‘=lim๐‘›โ†’โˆžโˆซ๐บ๐‘›๐‘โ‰ค๐ต๐‘<โˆž

็”ฑไบŽ ๐บโˆˆ๐ฟ๐‘, ๆœ‰

๐บ(๐‘ฅ)<โˆž๐‘Ž.๐‘’.

ไบŽๆ˜ฏ:

๐‘”(๐‘ฅ):=โˆ‘๐‘˜=1โˆž๐‘“๐‘˜(๐‘ฅ)<โˆž๐‘Ž.๐‘’.

ๅˆ |๐‘”๐‘›|,|๐‘”|โ‰ค๐บ,๐‘”๐‘›โ†’๐‘”, ๅฏๅพ—ๅˆฐ:

|๐‘”๐‘›โˆ’๐‘”|๐‘โ‰ค2๐‘๐บ๐‘โˆˆ๐ฟ1

ๅ› ่€Œ by DCT ๅฏไปฅๅพ—ๅˆฐ:

lim๐‘›โˆซ|๐‘”๐‘›โˆ’๐‘”|๐‘=0

ไปŽ่€Œ

lim๐‘›โ†’โˆžโˆฅ๐‘”โˆ’๐‘”๐‘›โˆฅ๐‘=(lim๐‘›โˆซ|๐‘”๐‘›โˆ’๐‘”|๐‘)1/๐‘=0

โ–ก

9.2.5 Criterion for ๐ฟ๐‘ convergence: ้€็‚น a.e. conv + ๐ฟ๐‘ ็งฏๅˆ†ๅ€ผ conv

ๆˆ‘ไปฌๅˆšๆ‰ mention: DCT ๅฏนไบŽ function seq ๐ฟ๐‘ convergence ็š„่ฏๆ˜Žๆœ‰ๅพˆๅคงไฝœ็”จ. ่ฟ™้‡Œๆˆ‘ไปฌๅฐฑๆไพ›ไธ€ไธช DCT ๆŽจๅ‡บ็š„ ๐ฟ๐‘ convergence ็š„ๅˆคๆ–ญๅ‡†ๅˆ™:

Theorem 9.48 : Criterion for ๐ฟ๐‘ convergence

if ๐‘“๐‘›โ†’๐‘“ a.e. and โˆฅ๐‘“๐‘›โˆฅ๐‘โ†’โˆฅ๐‘“โˆฅ๐‘, then โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐‘โ†’0.

ๅณ

a.e. conv +๐ฟ๐‘ norm convโŸน๐ฟ๐‘๐‘๐‘œ๐‘›๐‘ฃ

ไฝ†ๆ˜ฏ converse ๅนถไธๆˆ็ซ‹. ๅไพ‹ๆ˜ฏ typewriter function.

Proof

In Hw 8.

โ–ก

9.2.6 dense subsets of ๐ฟ๐‘, and specially ๐ฟ๐‘(โ„,๐‘š)

Proposition 9.21

ๅฏนไบŽไปปๆ„ 1โ‰ค๐‘<โˆž, the set of {simple functions}, is dense in ๐ฟ๐‘.
ๅณ:

{๐‘“:๐‘‹โ†’โ„‚โˆฃ๐‘“=โˆ‘1๐‘›๐‘Ž๐‘—๐œ’๐ธ๐‘—,๐œ‡(๐ธ๐‘—)<โˆž}

ๆ˜ฏ ๐ฟ๐‘ ็š„ dense subset.

Proof

ๅฏน ๐‘“ ไฝฟ็”จ simple function seq ้€ผ่ฟ‘, ไฝฟ็”จ 2๐‘|๐‘“|๐‘ ไฝœไธบ dominating function of |๐‘“๐‘˜โˆ’๐‘“|๐‘; ่€ŒๅŽไฝฟ็”จ DCT ๅพ—่ฏ.

โ–ก

Theorem 9.49 : ๐ถ๐‘0(โ„๐‘›) is dense in ๐ฟ๐‘(โ„,๐‘š) for 1โ‰ค๐‘<โˆž

๐ถ๐‘0(โ„๐‘›) is dense in ๐ฟ๐‘(โ„,๐‘š) for 1โ‰ค๐‘<โˆž

Proof

exercise. Similar to the proof for ๐ฟ1, ๅช้œ€่ฆไฝฟ็”จๅŠ ๅ…ฅ ๐‘ power ็š„ function ไฝœไธบ dominating function ๅณๅฏ.

โ–ก

9.3 ๐ฟโˆž space, and relationship between ๐ฟ๐‘ spaces (0โ‰ค๐‘โ‰คโˆž) [Fol 6.1, finished]

ๅฏนๅบ” Folland 6.1(3), finishing 6.1.
ๆˆ‘ไปฌๅทฒ็ปๅฎŒๆˆไบ†ๅฏน 1โ‰ค๐‘<โˆž ็š„ ๐ฟ๐‘ space ็š„ๆž„ๅปบ. ็Žฐๅœจ, ๆˆ‘ไปฌๆฅๆž„ๅปบๆœ€ๅŽไธ€ๅ—ๆ‹ผๅ›พ: ๐ฟโˆž space.

9.3.1 ๐ฟโˆž space

ๆˆ‘ไปฌ่€ƒ่™‘่ฟ™ไธชๅฏๅ‘ๅผ็š„ไพ‹ๅญ:

๐‘‹โ‰”{1,2,โ‹ฏ,๐‘›},๐’œ๏ธ€โ‰”๐’ซ๏ธ€(๐‘‹),๐œ‡=๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘–๐‘›๐‘”

ไบŽๆ˜ฏ:

๐ฟ๐‘(๐œ‡)={(๐‘Ž1,โ‹ฏ,๐‘Ž๐‘›):โˆฅ(๐‘Ž1,โ‹ฏ,๐‘Ž๐‘›)โˆฅ๐‘=(โˆ‘|๐‘Ž๐‘–|๐‘)1/๐‘<โˆž}=โ„‚๐‘›

ๆˆ‘ไปฌๅ‘็Žฐ:

โˆฅ(๐‘Ž1,โ‹ฏ,๐‘Ž๐‘›)โˆฅ๐‘โ†’max๐‘—|๐‘Ž๐‘—|as๐‘โ†’โˆž

ๅ› ไธบ ๐‘ ๅ–ๅพ—่ถŠๅคง, ๆœ€ๅคง็š„ entry ็š„ contribution ๅ ๆฏ”ๅฐฑ่ถŠ็ชๅ‡บ.
ๅฏนไบŽ่ฟ™ๆ ท็š„ ๐ฟ๐‘ space, ๆˆ‘ไปฌๅฏไปฅๅฎšไน‰ sup norm, ๅฎšไน‰ไธบๆœ€ๅคง็š„ entry.
ๅณไพฟ ๐‘‹ ๆ˜ฏ countable ็š„, ่ฟ™ไธชๅฎšไน‰ไนŸๅฏไปฅๅฎšไน‰ไธบ sup๐‘—|๐‘Ž๐‘—|, make sense.
้‚ฃไนˆๅฆ‚ๆžœๆˆ‘ไปฌๆƒณ่ฆ็ป™ไปปๆ„็š„ measure space ๅฎšไน‰ sup norm ๅ‘ข? ๆˆ‘ไปฌๅฏไปฅ่€ƒ่™‘

โˆฅ๐‘“โˆฅโˆž:=sup๐‘ฅโˆˆ๐‘‹|๐‘“(๐‘ฅ)|?

ๅฎž้™…ไธŠๆˆ‘ไปฌๆœ‰ๆ›ดๅฅฝ็š„ๅฎšไน‰ๆ–นๅผ:

Definition 9.49 : essential supremum
โˆฅ๐‘“โˆฅโˆž:=inf{๐‘Žโ‰ฅ0:๐œ‡{๐‘ฅ:|๐‘“(๐‘ฅ)|>๐‘Ž}=0}

ไนŸๅฏไปฅๅ†™ไฝœ:

esssup๐‘ฅโˆˆ๐‘‹|๐‘“(๐‘ฅ)|
Definition 9.50 : ๐ฟโˆž space
๐ฟโˆž(๐œ‡):={๐‘“:๐‘‹โ†’โ„‚ measurable:โˆฅ๐‘“โˆฅโˆž<โˆž}/โˆผ

where โˆผ ่กจ็คบ a.e. ็›ธ็ญ‰็š„ๅ‡ฝๆ•ฐ็š„ equiv class.

ไธ‹้ขๆ˜ฏไธ€ไธชๆฏ”่พƒๅ…ธๅž‹็š„ไพ‹ๅญ:

9.3.2 โ„“โˆž space

Definition 9.51 : โ„“โˆž
โ„“โˆž:={(๐‘Ž๐‘—)1โˆž:โˆฅ(๐‘Ž๐‘—)โˆฅโˆž:=sup๐‘—|๐‘Ž๐‘—|<โˆž}
Example 9.27
๐‘“=๐‘ฅ๐œ’โ„šโˆˆ๐ฟโˆž(๐‘š)

with

โˆฅ๐‘“โˆฅโˆž=0

ๅ› ไธบๆ•ดไธช โ„š ้ƒฝๆ˜ฏ้›ถๆต‹็š„.

9.3.3 ๐ฟโˆž ็š„ๅŸบๆœฌๆ€ง่ดจ: as a NVS; Hรถlderโ€™s ineq on it; dense subsets

Lemma 9.31

ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟโˆž(๐œ‡) ๅˆ™:

  • ไธ€ๅฎšๆœ‰ |๐‘“(๐‘ฅ)|โ‰คโˆฅ๐‘“โˆฅโˆž for a.e. ๐‘ฅ.

  • ๅญ˜ๅœจไธ€ไธช bounded ๅ‡ฝๆ•ฐ ๐‘”, ไฝฟๅพ— ๐‘“=๐‘” a.e.

Proof

ๆ˜พ็„ถ.

โ–ก

Theorem 9.50
  • โˆฅ๐‘“๐‘”โˆฅโˆžโ‰คโˆฅ๐‘“โˆฅ1โˆฅ๐‘”โˆฅโˆž

    ๅฏไปฅๆŠŠๅฎƒ็œ‹ไฝœ Hรถlder ็š„ไธ€้ƒจๅˆ†็‰นๆฎŠๆƒ…ๅ†ต, ๅ› ไธบๅฏไปฅ็œ‹ไฝœ

    11+1โˆž=1

    ไปŽ่€Œ่กฅๅ……ๅฎŒๆ•ดไบ† Hรถlder ineq for 1โ‰ค๐‘,๐‘žโ‰คโˆž

  • ๐ฟโˆž ๆ˜ฏไธ€ไธช normed vector space, equipped with โˆฅโ‹…โˆฅโˆž

  • simple functions are dense in ๐ฟโˆž

Proof

ๅฎนๆ˜“่ฏๆ˜Ž.

โ–ก

9.3.4 ๐ฟโˆž-convergence ไฝœไธบ (finite measure space ไธ‹) ๆœ€ๅผบ็š„ ๐ฟ๐‘ convergence: ็ญ‰ไปทไบŽ uni. conv a.e.

Theorem 9.51 : convergence in ๐ฟโˆž โ‡”uniform convergence a.e.
๐‘“๐‘›โ†’๐‘“ in ๐ฟโˆžโ‡”exists null set ๐ธโŠ‚๐‘‹๐‘ .๐‘ก.๐‘“๐‘›โ†’๐‘“ uniformly on ๐ธ๐‘

(ๆณจๆ„, ่ฟ™ไธๆ˜ฏ conv almost uniformly, ่€Œๆ˜ฏไธ€ไธชๆฏ” almost uniformly ๆ›ดๅผบ็š„ๆกไปถ: conv uniformly almost everywhere, ๅ› ไธบ almost uniformly ๅช่ฆๆฑ‚ๅฏนไบŽไปปๆ„็š„ ๐œ–, ้ƒฝๅญ˜ๅœจไธ€ไธช measure ๅฐไบŽ ๐œ– ็š„ ๐ธ, ไฝฟๅพ—ๅœจ ๐ธ๐‘ ไธŠ uni conv ๅณๅฏ.)

Proof

โ‡: Suppose ๐‘“๐‘›โ†’๐‘“ uni. a.e; WTS: ๐‘“๐‘›โ†’๐‘“ in ๐ฟโˆž ๐‘“๐‘›โ†’๐‘“ uni. a.e ๅณ: ๅญ˜ๅœจ้›ถๆต‹้›† ๐ธโŠ‚๐‘‹, ๐‘“๐‘›โ†’๐‘“ on ๐ธ๐‘.
Let ๐œ–>0.
๐‘“๐‘›โ†’๐‘“ uni. a.e ่กจๆ˜Ž, ๅญ˜ๅœจ ๐‘ ไฝฟๅพ— for all ๐‘›โ‰ฅ๐‘ ๆœ‰

โˆ€๐‘ฅโˆˆ๐ธ๐‘,|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|<๐œ–

by def, exactly is:

โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐ฟโˆž=essโ€†sup๐‘ฅโˆˆ๐‘‹|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|โ‰ค๐œ–

This shows that โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐ฟโˆžโ†’0, ๅณ ๐‘“๐‘›โ†’๐‘“ in ๐ฟโˆž.
โ‡: Suppose ๐‘“๐‘›โ†’๐‘“ in ๐ฟโˆž; WTS: ๐‘“๐‘›โ†’๐‘“ uni. a.e.Denote:

๐œ–๐‘›โ‰”โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐ฟโˆž

By assumption, ๐œ–๐‘›โ†’0. Define for each ๐‘›:

๐ด๐‘›โ‰”{๐‘ฅโˆˆ๐‘‹:|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|>๐œ–๐‘›}

By def โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐ฟโˆž=ess sup๐‘ฅ|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|โ‰ค๐œ–๐‘›, ไบŽๆ˜ฏ ๐œ‡(๐ด๐‘›)=0 ้‚ฃไนˆไปค:

๐ธโ‰”โ‹ƒ๐‘›=1โˆž๐ด๐‘›

by subadditivity of measure ๆœ‰ ๐œ‡(๐ธ)=0. ไบŽๆ˜ฏๅฏนไบŽไปปๆ„ ๐œ–๐‘›, ้ƒฝๆœ‰

|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|โ‰ค๐œ–๐‘›โ†’0,for all ๐‘ฅโˆˆ๐ธ๐‘

็”ฑไบŽ ๐œ–๐‘›โ†’0, showing that outside ๐ธ, ๆœ‰ โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐ฟโˆžโ†’0.

โ–ก

9.3.5 ๐ฟโˆž as Banach space

Theorem 9.52 : ๐ฟ๐‘ (1โ‰ค๐‘โ‰คโˆž) is Banach

For any measure space (๐‘‹,๐’œ๏ธ€,๐œ‡), ๐ฟ๐‘(๐œ‡) is Banach for all 1โ‰ค๐‘โ‰คโˆž

Proof

ๆˆ‘ไปฌๅทฒ็ป proved ไบ† 1โ‰ค๐‘<โˆž ็š„ case, ็Žฐๅœจ prove ๐‘=โˆž ็š„ case.
By Theoremย 9.44, ๆˆ‘ไปฌ็Ÿฅ้“ STS: every abs conv series conv in ๐ฟโˆž.
ๆˆ‘ไปฌ suppose ๐‘“๐‘˜โˆˆ๐ฟโˆž ๆœ‰

โˆ‘๐‘˜=1โˆžโˆฅ๐‘“๐‘˜โˆฅโˆž<โˆž

WTS: โˆ‘๐‘˜=1โˆž๐‘“๐‘˜ converges.
Set:

๐ธ๐‘˜โ‰”{๐‘ฅ:|๐‘“๐‘˜(๐‘ฅ)|>โˆฅ๐‘“๐‘˜||โˆž}

ไบŽๆ˜ฏๆœ‰

๐œ‡(๐ธ๐‘˜)=0for each ๐‘˜

ๅ› ่€Œ setting

๐ธ:=โ‹ƒ๐‘˜=1โˆž๐ธ๐‘˜

ๆœ‰

๐œ‡(๐ธ)=0

note:

๐‘ฅโˆˆ๐ธ๐‘โŸนโˆ‘๐‘˜=1โˆž|๐‘“๐‘˜(๐‘ฅ)|โ‰คโˆ‘๐‘˜=1โˆžโˆฅ๐‘“๐‘˜โˆฅโˆž<โˆž

ไปŽ่€Œ,

๐‘”:=โˆ‘๐‘˜=1โˆž๐‘“๐‘˜

ๅœจ ๐ธ๐‘ ไธŠๆ˜ฏ well-defined ็š„, ไธ” bounded by โˆ‘๐‘˜=1โˆžโˆฅ๐‘“๐‘˜โˆฅโˆž.
ๅฏนไบŽ ๐‘ฅโˆˆ๐ธ, ๆˆ‘ไปฌๅฏไปฅ้šไพฟ่ฎพ็ฝฎๅ€ผ, ๆฏ”ๅฆ‚ ๐œ‹, ็„ถๅŽ define ๐‘”(๐‘ฅ)=๐œ‹ on ๐‘ฅโˆˆ๐ธ. ็„ถๅŽๅฏนไบŽ each ๐‘›, ๆˆ‘ไปฌ set:

๐‘”๐‘›(๐‘ฅ):={โˆ‘๐‘˜=1๐‘›๐‘“๐‘˜(๐‘ฅ),๐‘ฅโˆˆ๐ธ๐‘1๐œ‹,๐‘ฅโˆˆ๐ธ

ไปŽ่€Œ

โˆฅ๐‘”๐‘›โˆ’๐‘”โˆฅโˆžโ‰คsup๐‘ฅโˆˆ๐ธ๐‘|๐‘”๐‘›(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)|โ‰คsup๐‘ฅโˆˆ๐ธ๐‘|โˆ‘๐‘›+1โˆž๐‘“๐‘˜(๐‘ฅ)|โ‰คsup๐‘ฅโˆˆ๐ธ๐‘โˆ‘๐‘›+1โˆž|๐‘“๐‘˜(๐‘ฅ)|โ‰คโˆ‘๐‘›+1โˆžโˆฅ๐‘“๐‘˜โˆฅโˆžโ†’0

โ–ก

9.3.6 relationship between ๐ฟ๐‘ spaces

9.3.7 ๐ฟ๐‘š(๐œ‡)โŠ‚๐ฟ๐‘›(๐œ‡),0<๐‘›โ‰ค๐‘šโ‰คโˆž, for measure finite space)

ๅˆšๆ‰ๆˆ‘ไปฌๅทฒ็ป state ไบ†, ไฝ†่ฟ˜ๆฒกๆœ‰่ฏๆ˜Ž:

Theorem 9.53 : inclusion relation between ๐ฟ๐‘ spaces (when base space is finite measure)

ๅฆ‚ๆžœ measure space ๐‘‹ has finite measure, ้‚ฃไนˆๆœ‰

๐ฟโˆž(๐‘‹)โŠ‚โ‹ฏโŠ‚๐ฟ๐‘š(๐‘‹)โŠ‚โ‹ฏโŠ‚๐ฟ๐‘›(๐‘‹)โŠ‚โ‹ฏ

for ไปปๆ„็š„ ๐‘šโ‰ฅ๐‘›.

่ฟ™ๆ˜ฏๆˆ‘ไปฌ้ฆ–ๆฌกๆŠŠ ๐‘<1 ไนŸ include ่ฟ›ๆˆ‘ไปฌ็š„่ฎจ่ฎบ.

่ฟ™ไธช statement ๅณ: ๅฏนไบŽ from finite measure space to โ„‚ ็š„ function ๐‘“, ๅฎƒ็š„ โˆฅ๐‘“โˆฅ๐‘š<โˆž ๆ˜ฏๆฏ” โˆฅ๐‘“โˆฅ๐‘›<โˆž ๆ›ดๅผบ็š„ๆกไปถ.
ๅฐคๅ…ถ, ้™คๅŽป ๐ฟโˆž ็š„ๆƒ…ๅ†ต, ๅฎƒๆ›ด็›ดๆŽฅ็š„ๆ„ๆ€ๆ˜ฏ: ๅฏนไบŽ 0<๐‘›โ‰ค๐‘š<โˆž ่€Œ่จ€, ๐‘“ ็š„็ปๅฏนๅ€ผ็š„ ๐‘š ๆฌกๆ–น็š„็งฏๅˆ† <โˆž ๆ˜ฏๆฏ” ๐‘“ ็š„็ปๅฏนๅ€ผ็š„ ๐‘› ๆฌกๆ–น็š„็งฏๅˆ† <โˆž ่ฆๆ›ดๅผบ็š„ๆกไปถ.

่ฟ™ๅ…ถๅฎžๆ˜ฏไธ€ไปถๆฏ”่พƒ็›ด่ง‚็š„ไบ‹ๆƒ…. ๅ› ไธบๅฏนไบŽ |๐‘“|โ‰ฅ1 ็š„้ƒจๅˆ†,

โˆซ|๐‘“|โ‰ฅ1|๐‘“|๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’โ‰ฅโˆซ|๐‘“|โ‰ฅ1|๐‘“|๐‘ ๐‘š๐‘Ž๐‘™๐‘™

่€ŒๅฏนไบŽ |๐‘“|<1 ็š„้ƒจๅˆ†,

โˆซ|๐‘“|<1|๐‘“|๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’โ‰คโˆซ|๐‘“|<1|๐‘“|๐‘ ๐‘š๐‘Ž๐‘™๐‘™

็„ถ่€Œ็”ฑไบŽๆ•ดไธช space ็š„ measure ๆ˜ฏ finite ็š„, |๐‘“|<1 ็š„้ƒจๅˆ†ๅนถไธๅฝฑๅ“. ๅ› ไธบ

โˆซ|๐‘“|<1|๐‘“|๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’โ‰คโˆซ|๐‘“|<1|๐‘“|๐‘ ๐‘š๐‘Ž๐‘™๐‘™โ‰คโˆซ|๐‘“|<11โ‰ค๐œ‡(๐‘‹)

ๅ› ่€Œ, ๅฏนไบŽ ๐œ‡(๐‘‹)<โˆž ็š„ๆƒ…ๅ†ต, ๆ˜พ็„ถๆœ‰ โˆฅ๐‘“โˆฅ๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’<โˆž ๆ˜ฏๆฏ” โˆฅ๐‘“โˆฅ๐‘ ๐‘š๐‘Ž๐‘™๐‘™<โˆž ๆ›ดๅผบ็š„ๆกไปถ.
(ๅฎž้™…ไธŠ, ๅฆ‚ๆžœๅชๆœ‰ measure finite ็š„ ๐‘ฅ ไธŠ |๐‘“(๐‘ฅ)|<1, ้‚ฃไนˆๅณไพฟ ๐œ‡(๐‘‹)=โˆž, โˆฅ๐‘“โˆฅ๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’<โˆž ไนŸๆ˜ฏๆฏ” โˆฅ๐‘“โˆฅ๐‘ ๐‘š๐‘Ž๐‘™๐‘™<โˆž ๆ›ดๅผบ็š„ๆกไปถ; ่€Œๅฆ‚ๆžœๆœ‰ measure infinite ็š„ ๐‘ฅ ไธŠ |๐‘“(๐‘ฅ)|<1, ้‚ฃไนˆๆœ‰ๅฏ่ƒฝ โˆฅ๐‘“โˆฅ๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’<โˆž ๆ˜ฏๆฏ” โˆฅ๐‘“โˆฅ๐‘ ๐‘š๐‘Ž๐‘™๐‘™<โˆž ๆ›ดๅผฑ็š„ๆกไปถ)
My point: ่™ฝ็„ถ่ฏด |๐‘“(๐‘ฅ)|๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’ ๆฏ”่ตท |๐‘“(๐‘ฅ)|๐‘ ๐‘š๐‘Ž๐‘™๐‘™ ๆ˜ฏๆ›ดๅคง่ฟ˜ๆ˜ฏๆ›ดๅฐๅ–ๅ†ณไบŽ |๐‘“(๐‘ฅ)| ๆ˜ฏๅฆ โ‰ฅ1 or <1, ไฝ†ๆ˜ฏ โ‰ฅ1 ็š„ๅ€ผๆ˜ฏๅฏไปฅ unbounded ็š„, ่€Œ <1 ็š„ๅ€ผๅ†ๆ€Žไนˆ้€š่ฟ‡ๅฐๆฌกๆ–นๅ˜ๅพ—ๆ›ดๅคง, ไนŸ่ถ…ไธ่ฟ‡ 1. ๅ› ่€Œ |๐‘“(๐‘ฅ)|โ‰ฅ1 ็š„้ƒจๅˆ†้€šๅธธๆ›ด่ƒฝๅ‡ฝๆ•ฐ็งฏๅˆ†ๅ€ผ็š„ๆœ‰้™ๆ€ง, ้™ค้žๅœจไธ€ไธช measure infinite ็š„้›†ๅˆไธŠ |๐‘“(๐‘ฅ)|<1.

่ฟ™้‡Œๆœ‰ไธ€ไธชๆ›ดๅŠ ไธฅๆ ผ็š„่ฏๆ˜Ž:

Proof

้ฆ–ๅ…ˆ, ๅฏนไบŽ ๐‘š=โˆž ็š„ case, ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ๐‘š=๐ฟโˆž, ้‚ฃไนˆๅ–ไปปๆ„ 1โ‰ค๐‘›<โˆž ้ƒฝๆœ‰:

โˆซ|๐‘“|๐‘›โ‰คโˆซโˆฅ๐‘“โˆฅโˆž๐‘›=โˆฅ๐‘“โˆฅโˆž๐‘›๐œ‡(๐‘‹)<โˆž

ๅ…ถๆฌก, ๅฏนไบŽๆญฃๅธธ็š„ ๐‘š<โˆž ็š„ case, ๆˆ‘ไปฌไฝฟ็”จ Hรถlder: ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ๐‘š, ้‚ฃไนˆๅฏนไบŽไปปๆ„ ๐‘›<๐‘š, ๆˆ‘ไปฌๅฏไปฅๆž„้€ ๅ‡บ Hรถlder conjugate ๐‘š๐‘› ๅ’Œ ๐‘š๐‘šโˆ’๐‘›,ไปŽ่€Œ:

โˆซ|๐‘“|๐‘›=โˆซ|๐‘“|๐‘›โ‹…1โ‰ค(โˆซ(|๐‘“|๐‘›)๐‘š๐‘›)๐‘›๐‘š(โˆซ1๐‘š๐‘šโˆ’๐‘›)๐‘šโˆ’๐‘›๐‘š=โˆฅ๐‘“โˆฅ๐‘š๐‘›๐œ‡(๐‘‹)๐‘šโˆ’๐‘›๐‘š<โˆž

ไปŽ่€Œ

โˆฅ๐‘“โˆฅ๐‘š<โˆžโŸนโˆฅ๐‘“โˆฅ๐‘›<โˆž

่ฟ™ไธ€ proof ๅˆฉ็”จ Hรถlder conjuate, ้€š่ฟ‡ๆž„้€ ๅŒ…ๅซ ๐‘š๐‘› ็š„ Hรถlder conjugate, ๆŠŠ โˆซ|๐‘“|๐‘› ๆ”นๆˆไบ† โˆฅ๐‘“โˆฅ๐‘š ็š„ expression.

โ–ก

ไปฅไธ‹ๆ˜ฏไธ€ไธช็ปๅ…ธ็š„ไพ‹ๅญ:

Example 9.28

่€ƒ่™‘ measure finite ็š„ measure space (0,1): ้€š่ฟ‡็ปๅ…ธ็š„ Calculus ๆˆ‘ไปฌ็Ÿฅ้“:

๐‘“(๐‘ฅ)=1๐‘ฅ๐‘šโˆˆ๐ฟ๐‘(0,1) for all ๐‘<1๐‘š

ไฝ†ๆ˜ฏๅฏนไบŽไปปๆ„็š„ ๐‘š, ้ƒฝๆœ‰:

๐‘“(๐‘ฅ)=1๐‘ฅ๐‘šโˆ‰๐ฟ๐‘(0,1)

่€Œๆˆ‘ไปฌๅ†็œ‹ไธ€ไธช measure infinite ็š„ measure space (1,โˆž) ไธŠ็š„ๅไพ‹, ้‡‡็”จๅŒไธ€ไธชๅ‡ฝๆ•ฐ:

๐‘“(๐‘ฅ)=1๐‘ฅ๐‘š,๐‘ฅโˆˆ(1,โˆž)

่ฟ™ไธชๆ—ถๅ€™, ๐‘ ่ถŠๅคง, โˆซ|๐‘“|๐‘=โˆฅ๐‘“โˆฅ๐‘๐‘ ๅ่€Œ่ถŠๅฐ, ้€š่ฟ‡็ปๅ…ธ็š„ Calculus ๆˆ‘ไปฌ็Ÿฅ้“:ๆˆ‘ไปฌ็Ÿฅ้“ ่€ŒๅฏนไบŽ

๐‘“(๐‘ฅ)=1๐‘ฅ๐‘šโˆˆ๐ฟ๐‘(1,โˆž)for all ๐‘>1๐‘š

ๅนถไธ” ๐‘“โˆˆ๐ฟโˆž(1,โˆž), ๅ› ไธบ โˆฅ๐‘“โˆฅโˆž=1.
่ฟ™ไธช็ฉบ้—ดไธŠ็š„่ฟ™ไธชๅ‡ฝๆ•ฐๆญฃๅฏนๅบ”ไบ†ๆˆ‘ไปฌๅˆšๆ‰่ฎจ่ฎบ็š„, ๅฆ‚ๆžœๆœ‰ infinite measure ๆ•ฐ้‡็š„ ๐‘ฅ ไธŠ |๐‘“(๐‘ฅ)|<1, ้‚ฃไนˆๅพˆๅฏ่ƒฝ โˆฅ๐‘“โˆฅ๐‘™๐‘Ž๐‘Ÿ๐‘”๐‘’<โˆž ๆ˜ฏๆฏ” โˆฅ๐‘“โˆฅ๐‘ ๐‘š๐‘Ž๐‘™๐‘™<โˆž ๆ›ดๅผฑ็š„ๆกไปถ

9.3.8 control arbitrary โˆฅ๐‘“โˆฅ๐‘š ๅ’Œ โˆฅ๐‘“โˆฅ๐‘› ็š„ๅคงๅฐๆฏ”ไพ‹, in measure finite space

9.3.9 (๐ฟ๐‘›โˆฉ๐ฟ๐‘Ÿ)โŠ‚๐ฟ๐‘šโŠ‚(๐ฟ๐‘›+๐ฟ๐‘Ÿ), ๅฏนไปปๆ„ 0<๐‘›<๐‘š<๐‘Ÿโ‰คโˆž

Proposition 9.22

ๅฏนไบŽ measurable ๐‘“:๐‘‹โ†’โ„‚,

๐‘กโ†ฆโˆฅ๐‘“โˆฅ1๐‘ก

is log-convex.
equivalently ๅณ: ๅฏนไบŽไปปๆ„็š„ 0<๐‘›<๐‘š<๐‘Ÿโ‰คโˆž, ้ƒฝๆœ‰

โˆฅ๐‘“โˆฅ๐‘šโ‰คโˆฅ๐‘“โˆฅ๐‘›๐œ†โ‹…โˆฅ๐‘“โˆฅ๐‘Ÿ1โˆ’๐œ†

where

๐œ†โ‰”1๐‘šโˆ’1๐‘Ÿ1๐‘›โˆ’1๐‘Ÿโˆˆ(0,1),๐‘–.๐‘’.(1๐‘š)=๐œ†(1๐‘›)+(1โˆ’๐œ†)(1๐‘Ÿ)
Proof

For ๐‘Ÿ=โˆž, then ๐œ†=๐‘›๐‘š.
Since

|๐‘“|๐‘š=|๐‘“|๐‘›โ‹…|๐‘“|๐‘šโˆ’๐‘›โ‰ค|๐‘“|๐‘›โ‹…โˆฅ๐‘“โˆฅโˆž๐‘šโˆ’๐‘›๐‘Ž.๐‘’.

ๅฏไปฅๅพ—ๅˆฐ

โˆซ|๐‘“|๐‘šโ‰ค(โˆซ|๐‘“|๐‘›)โ‹…โˆฅ๐‘“โˆฅโˆž๐‘šโˆ’๐‘›=โˆฅ๐‘“||๐‘›๐‘›โ‹…โˆฅ๐‘“โˆฅโˆž๐‘šโˆ’๐‘›

ไปŽ่€Œ Taking ๐‘žth root ๅพ—ๅˆฐ็ป“ๆžœ:

โˆฅ๐‘“โˆฅ๐‘šโ‰คโˆฅ๐‘“โˆฅ๐‘›๐‘›/๐‘šโˆฅ๐‘“โˆฅโˆž1โˆ’๐‘›/๐‘š

For ๐‘Ÿ<โˆž: ๆˆ‘ไปฌ้‡‡็”จ conjugate exponents:

๐‘›๐œ†๐‘š,๐‘Ÿ(1โˆ’๐œ†)๐‘š

่ฟ™ๆ˜ฏๅ› ไธบ:

(1๐‘š)=๐œ†(1๐‘›)+(1โˆ’๐œ†)(1๐‘Ÿ)โŸน1=๐œ†(๐‘š๐‘›)+(1โˆ’๐œ†)(๐‘š๐‘Ÿ)

ไปŽ่€Œ Applying Hรถlder:

โˆซ|๐‘“|๐‘š=โˆซ|๐‘“|๐œ†๐‘š|๐‘“|(1โˆ’๐œ†)๐‘šโ‰ค(โˆซ|๐‘“|๐‘›)๐œ†๐‘š๐‘›(โˆซ|๐‘“|๐‘Ÿ)(1โˆ’๐œ†)๐‘š๐‘Ÿ=โˆฅ๐‘“โˆฅ๐‘›๐œ†๐‘šโ‹…โˆฅ๐‘“โˆฅ๐‘Ÿ(1โˆ’๐‘Ÿ)๐‘š

Taking ๐‘ž th root ๅพ—ๅˆฐ็ป“ๆžœ.

โ–ก

Example 9.29

ไปค ๐ด ไธบไปปๆ„้›†ๅˆ, 0โ‰ค๐‘<๐‘žโ‰คโˆž, ๆœ‰:

โˆฅ๐‘“โˆฅ๐‘žโ‰คโˆฅ๐‘“โˆฅ๐‘and thusโ„“๐‘(๐ด)โŠ‚โ„“๐‘ž(๐ด)

่ฟ™ๆ˜ฏๅ› ไธบ

โˆฅ๐‘“โˆฅโˆž๐‘=sup๐›ผ|๐‘“(๐›ผ)|๐‘โ‰คโˆ‘๐›ผ|๐‘“(๐›ผ)|๐‘=โˆฅ๐‘“โˆฅ๐‘๐‘

ไบŽๆ˜ฏ for ๐‘žโ‰ โˆž case

โˆฅ๐‘“โˆฅ๐‘žโ‰คโˆฅ๐‘“โˆฅ๐‘๐œ†โˆฅ๐‘“โˆฅโˆž1โˆ’๐œ†โ‰คโˆฅ๐‘“โˆฅ๐‘

(ๅฆไธ€ case, trivial.)
ๆˆ‘ไปฌๅ‘็Žฐ โ„“๐‘ ็ฉบ้—ด, ๐‘ ่ถŠๅฐ่ฆๆฑ‚ๅ่€Œ่ถŠไธฅๆ ผ.
่ฟ™ๆ˜ฏๅ› ไธบ โ„“๐‘ ็ฉบ้—ดไธญไธ€ไธชๅ‡ฝๆ•ฐๅฐฑๆ˜ฏไธ€ไธช seq, ๅ…ถ ๐‘-norm ๅฐฑๆ˜ฏๅ„้กน็š„ ๐‘ ๆฌกๆ–นๅ’Œ, ๅ†ๅผ€ ๐‘ ๆฌกๆ–นๆ น.
ๅฏนไบŽไธ€ไธช seq, ๅฆ‚ๆžœๅฎƒ็š„็ดฏๅ’Œ series ๆ”ถๆ•›, ๅฎƒ็š„ๅ„้กน่‚ฏๅฎšๆ˜ฏ eventually ๆ”ถๆ•›็š„, ้‚ฃไนˆ่ฟ™ไบ›้™คไบ†ๆœ‰้™้กนๅค–็š„่ฟ™ไบ›้กน็š„็ปๅฏนๅ€ผ้ƒฝๆ˜ฏ <1 ็š„, ้‚ฃไนˆ ๐‘ ่ถŠๅคง, ๅฎƒไปฌ ๐‘ ๆฌกๆ–นๅ’Œๅชไผš่ถŠๅฐ. ่ฟ™ๆญฃๅฏนๅบ”ไบ†ๆˆ‘ไปฌไน‹ๅ‰่ฏด็š„ "|๐‘“(๐‘ฅ)|<1 ็š„็‚นไธปๅฏผๅ‡ฝๆ•ฐ" ็š„ๆƒ…ๅ†ต.

็›ธๅฏนไบŽ่ฟ™ไธชinclusion ๅ…ณ็ณป, ๆˆ‘ไปฌ่ฟ˜ๆœ‰ๅฆๅค–ไธ€ไธช inclusion ๅ…ณ็ณป:

Proposition 9.23 : ๆฏไธช ๐ฟ๐‘š ๅ‡ฝๆ•ฐ้ƒฝๆ˜ฏไธ€ไธช ๐ฟ๐‘› ๅ‡ฝๆ•ฐๅ’Œไธ€ไธช ๐ฟ๐‘Ÿ ๅ‡ฝๆ•ฐ็š„ๅ’Œ (0<๐‘›<๐‘š<๐‘Ÿโ‰คโˆž)

ๅฏนไบŽไปปๆ„็š„ 0<๐‘›<๐‘š<๐‘Ÿโ‰คโˆž, ้ƒฝๆœ‰

๐ฟ๐‘šโŠ‚(๐ฟ๐‘›+๐ฟ๐‘Ÿ)

่ฟ™ไธช inclusion ๅ…ณ็ณปๆœ‰ไธ€็ง่ฐƒๅ’Œ็š„ๆ„Ÿ่ง‰ๅœจ้‡Œ้ข. ๅฎƒ roughly mean ็ป™ๅฎšไธ€ไธชๅ‡ฝๆ•ฐ, ๅฎƒๅฏไปฅๆ‹†ๆˆไธ€ไธชๆ›ดๅŠ ๅฎนๆ˜“็งฏ็š„ๅ‡ฝๆ•ฐๅ’Œไธ€ไธชๆ›ดๅŠ ไธๅฎนๆ˜“็งฏ็š„ๅ‡ฝๆ•ฐ, ๅนถไธ”ๆˆ‘ไปฌๅพˆๅคง็จ‹ๅบฆไธŠๅฏไปฅๆŽงๅˆถ่ฟ™ไธคไธชๅ‡ฝๆ•ฐ็š„ๅฏ็งฏๆ€ง.
ไฝ†ๅ…ถๅฎžๅพˆ็ฎ€ๅ•, ๅฐฑๆ˜ฏ็”จๆˆ‘ไปฌไน‹ๅ‰็š„ |๐‘“(๐‘ฅ)|<1 ๅ’Œ โ‰ฅ1 ็š„็‚นไฝœไธบๅŒบๅˆ†, ๆŠŠๅ‡ฝๆ•ฐ็š„ๅฎšไน‰ๅŸŸๅˆ†ๆˆไธค้ƒจๅˆ†. ๅฆ‚ๆžœ |๐‘“| ็š„ m ๆฌกๆ–นๆ˜ฏๅฏ็งฏ็š„, ้‚ฃไนˆๆ›ดๅฐ็š„ ๐‘› ๆฌกๆ–น, ๅฏนไบŽ |๐‘“(๐‘ฅ)|โ‰ฅ1 ็š„้ƒจๅˆ†่‚ฏๅฎšไนŸๆ˜ฏๅฏ็งฏ็š„; ๆ›ดๅคง็š„ ๐‘Ÿ ๆฌกๆ–น, ๅฏนไบŽ |๐‘“(๐‘ฅ)|<1 ็š„้ƒจๅˆ†่‚ฏๅฎšไนŸๆ˜ฏๅฏ็งฏ็š„;

Proof

Suppose ๐‘“โˆˆ๐ฟ๐‘š. Let

๐ธโ‰”{๐‘ฅ:|๐‘“(๐‘ฅ)|>1}

let

๐‘”:=๐‘“๐œ’๐ธ,โ„Žโ‰”๐‘“๐œ’๐ธ๐‘

ไบŽๆ˜ฏ ๐‘”โˆˆ๐ฟ๐‘› for all 0<๐‘›โ‰ค๐‘š, โ„Žโˆˆ๐ฟ๐‘Ÿ for all ๐‘Ÿโ‰ฅ๐‘š and ๐‘Ÿ=โˆž.

โ–ก

Homework 8: on ๐ฟ๐‘ spacecs (50/50)

Some of the following questions will be graded. Do them, and do hand them in.

ไธ€ไธช Barely in ๐ฟ1 ็š„ๅ‡ฝๆ•ฐ

Find a function ๐‘“โˆˆ๐ฟ1(โ„2025) such that ๐‘“โˆ‰๐ฟ๐‘(๐‘ˆ) for any ๐‘>1 and any nonempty open subset ๐‘ˆโŠ‚โ„2025. Hint: see HW5(g).

Solution

Recall Hw 5(g):For ๐›ผโˆˆ(0,1), define ๐‘”๐›ผ:โ„โ†’โ„ by ๐‘”๐›ผ(๐‘ฅ)=(1โˆ’๐›ผ)๐‘ฅโˆ’๐›ผ for 0<๐‘ฅ<1 and ๐‘”๐›ผ(๐‘ฅ)=0 otherwise. Let (๐‘ฅ๐‘›)๐‘› be an enumeration of the rational numbers, and define ๐‘“:โ„โ†’[0,โˆž] by

๐‘“(๐‘ฅ)=โˆ‘๐‘›=1โˆž2โˆ’๐‘›๐‘”1โˆ’๐‘›โˆ’๐‘›(๐‘ฅโˆ’๐‘ฅ๐‘›)

We have proved ๐‘“ has the following properties:

  • ๐‘“ is Lebesgue integrable and โˆซโ„|๐‘“|๐‘‘๐‘š=โˆซโ„๐‘“๐‘‘๐‘š<โˆž;

  • โˆซ๐ผ๐‘“๐‘๐‘‘๐‘š=โˆžfor all ๐‘>1,for all open interval ๐ผ

Now we continuing this definition of ๐‘“, and further define:

๐น:โ„2025โ†’โ„(๐‘ฅ1,โ‹ฏ,๐‘ฅ2025)โ†ฆโˆ๐‘—=12025๐‘“(๐‘ฅ๐‘—)

Claim 1: ๐นโˆˆ๐ฟ1(โ„2025).
To prove this, we just need this lemma.

Lemma 9.32 : (Folland 2.5 exercise 51)

If ๐‘“ is โ„ณ๏ธ€-measurable, ๐‘” is ๐’ฉ๏ธ€-measurable, then ๐‘“๐‘” is (โ„ณ๏ธ€โŠ—๐’ฉ๏ธ€)-measurable.
Particularly, if ๐‘“โˆˆ๐ฟ1(๐œ‡), ๐‘”โˆˆ๐ฟ1(๐œˆ), then ๐‘“๐‘”โˆˆ๐ฟ1(๐œ‡ร—๐œˆ) and

โˆซ๐‘“๐‘”๐‘‘(๐œ‡ร—๐œˆ)=(๐‘“๐‘‘๐œ‡)(๐‘”๐‘‘๐œˆ)

It seems like we have not proved this yet so here letโ€™s prove it.

Proof

of Lemma: Define

โ„Ž:=๐‘“๐‘”

Note

๐‘:(๐‘ข,๐‘ฃ)โ†ฆ๐‘ข๐‘ฃ

from โ„‚2โ†’โ„‚ is a product of two coordinate maps, thus is measurable since coordinate map is measurable, and product of two measurable functions is measurable.
And

๐œ‹:(๐‘ฅ,๐‘ฆ)โ†ฆ(๐‘“(๐‘ฅ),๐‘”(๐‘ฆ))

from ๐‘‹ร—๐‘Œโ†’โ„‚2 is (โ„ณ๏ธ€โŠ—๐’ฉ๏ธ€,โ„‚2)-measurable, since for any measurable rectangle ๐ต1ร—๐ต2โˆˆโ„‚2, we have

๐œ‹โˆ’1(๐ต1ร—๐ต2)=๐‘“โˆ’1(๐ต1)ร—๐‘”โˆ’1(๐ต2)โˆˆ๐’œ๏ธ€โŠ—โ„ฌ๏ธ€as a measurable rect

Thus โ„Ž=๐œ‹โˆ˜๐‘ is (โ„ณ๏ธ€โŠ—๐’ฉ๏ธ€)-measurable, as a composition of two measurable functions.
To show the second statement, it suffices to assume ๐‘“,๐‘” takes positive real values, since otherwise we can decompose ๐‘“,๐‘” into their real and imaginary parts, and for each part decompose them into positive part minus negative part.
Take two seq of simple functions approximating ๐‘“,๐‘” respectively from below, say:

๐‘ ๐‘›(๐‘ฅ)โ‰”โˆ‘๐‘˜=1๐พ๐‘Ž๐‘˜๐œ’๐ด๐‘˜(๐‘ฅ),๐‘ก๐‘›(๐‘ฆ)=โˆ‘โ„“=1๐ฟ๐‘๐‘™๐œ’๐ต๐‘™(๐‘ฆ)

their product on ๐‘‹ร—๐‘Œ is

๐‘ ๐‘›(๐‘ฅ)๐‘ก๐‘›(๐‘ฆ)=โˆ‘๐‘˜=1๐พโˆ‘๐‘™=1๐ฟ๐‘Ž๐‘˜๐‘๐‘™๐œ’๐ด๐‘˜ร—๐ต๐‘™(๐‘ฅ,๐‘ฆ)

By definition of the product measure ๐œ‡ร—๐œˆ, we have

(๐œ‡ร—๐œˆ)(๐ด๐‘˜ร—๐ต๐‘™)=๐œ‡(๐ด๐‘˜)๐œˆ(๐ต๐‘™)

Hence

โˆซ๐‘‹ร—๐‘Œ๐‘ ๐‘›(๐‘ฅ)๐‘ก๐‘›(๐‘ฆ)๐‘‘(๐œ‡ร—๐œˆ)=โˆ‘๐‘˜,๐‘™๐‘Ž๐‘˜๐‘๐‘™๐œ‡(๐ด๐‘˜)๐œˆ(๐ต๐‘™)=(โˆ‘๐‘˜๐‘Ž๐‘˜๐œ‡(๐ด๐‘˜))(โˆ‘๐‘™๐‘๐‘™๐œˆ(๐ต๐‘™))=(โˆซ๐‘‹๐‘ ๐‘›๐‘‘๐œ‡)(โˆซ๐‘Œ๐‘ก๐‘›๐‘‘๐œˆ)

Since ๐‘ ๐‘›(๐‘ฅ)โ†—๏ธŽ๐‘“(๐‘ฅ) and ๐‘ก๐‘›(๐‘ฆ)โ†—๏ธŽ๐‘”(๐‘ฆ), we also have ๐‘ ๐‘›๐‘ก๐‘›โ†—๏ธŽ๐‘“๐‘”, thus by MCT we have:

lim๐‘›โˆซ๐‘‹๐‘ ๐‘›๐‘‘๐œ‡=โˆซ๐‘‹๐‘“,lim๐‘›โˆซ๐‘Œ๐‘ก๐‘›๐‘‘๐œˆ=โˆซ๐‘Œ๐‘”

and

lim๐‘›โˆซ๐‘‹ร—๐‘Œ๐‘ ๐‘›(๐‘ฅ)๐‘ก๐‘›(๐‘ฆ)๐‘‘(๐œ‡ร—๐œˆ)=โˆซ๐‘‹ร—๐‘Œ๐‘“๐‘”๐‘‘(๐œ‡ร—๐œˆ)

Then, since the right side are two finite positive reals, we have:

โˆซ๐‘‹ร—๐‘Œ๐‘“(๐‘ฅ)๐‘”(๐‘ฆ)๐‘‘(๐œ‡ร—๐œˆ)=(โˆซ๐‘‹๐‘“๐‘‘๐œ‡)(โˆซ๐‘Œ๐‘”๐‘‘๐œˆ)<โˆž

Thus โ„Ž=๐‘“๐‘”โˆˆ๐ฟ1(๐œ‡ร—๐œˆ)

โ–ก

After proving the Lemma, we can extend it to the product of any finite number of functions. Applying it, we get

๐นโˆˆ๐ฟ1(โ„2025)

Then, we take arbitrary open set ๐‘ˆโŠ‚โ„2025 and arbitrary ๐‘>1, and fix it.
Claim 2: ๐นโˆ‰๐ฟ๐‘(๐‘ˆ). Sine ๐‘ˆ is open in โ„2025, it must contain an open ball, thus must contain an open box (e.g., the one internally connected in the open ball), say ๐ผ1ร—โ‹ฏร—๐ผ2025.
Suppose for contradiction that ๐นโˆˆ๐ฟ๐‘(๐‘ˆ).
Then by monotonicity of integration:

โˆซ๐ผ1ร—โ‹ฏร—๐ผ2025|๐น|๐‘๐‘‘(๐‘ฅ1,โ€ฆ,๐‘ฅ2025)โ‰คโˆซ๐‘ˆ|๐น|๐‘๐‘‘(๐‘ฅ1,โ€ฆ,๐‘ฅ2025)<โˆž

Then by Fubiniโ€™s Thm we have:

โˆซ๐ผ1ร—โ‹ฏร—๐ผ2025โˆ๐‘—=12025|๐‘“(๐‘ฅ๐‘—)|๐‘๐‘‘(๐‘ฅ1,โ€ฆ,๐‘ฅ2025)=โˆ๐‘—=12025โˆซ๐ผ๐‘—|๐‘“(๐‘ฅ๐‘—)|๐‘๐‘‘๐‘ฅ๐‘—<โˆž

Since for each ๐ผ๐‘—, we in hw 5 proved that:

โˆซ๐ผ๐‘—|๐‘“(๐‘ฅ๐‘—)|๐‘๐‘‘๐‘ฅ๐‘—=โˆž

This contradicts with what we got. Thus we must have ๐นโˆ‰๐ฟ๐‘(๐‘ˆ).
This finishes the proof.

๐ฟ๐‘ norm version of LDT

Let 1โ‰ค๐‘<โˆž. Suppose that ๐‘“โˆˆ๐ฟ๐‘(โ„). Prove that

lim๐‘Ÿโ†’012๐‘Ÿโˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆ=0

for a.e. ๐‘ฅ.
(Hint: Follow the proof of the Lebegue Differentiation Theorem when ๐‘=1, i.e. approximate ๐‘“ by ๐‘”โˆˆ๐ถ๐‘(โ„) satisfying โˆฅ๐‘“โˆ’๐‘”โˆฅ๐‘<๐œ–. At some point, use Minkowskiโ€™s inequality; note that we have |๐‘Ž+๐‘|โ‰ค|๐‘Ž|+|๐‘|, but we donโ€™t have |๐‘Ž+๐‘|๐‘โ‰ค|๐‘Ž|๐‘+|๐‘|๐‘ for ๐‘>1.)

Proof

Claim 1: The statement is true for ๐‘“โˆˆ๐ถ๐‘0(โ„๐‘›).
Proof of Claim 1:Let ๐‘“โˆˆ๐ถ๐‘0(โ„), then it is uniformly continuous on any compact set, thus uniformly continuous on an open ball, since its closure is compact.
Therefore, let ๐œ–>0, then there exists ๐›ฟ>0 such that

|๐‘ฆโˆ’๐‘ฅ|<๐›ฟโŸน|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|<๐œ–

Thus

|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘<๐œ–๐‘whenever |๐‘ฆโˆ’๐‘ฅ|<๐›ฟ

Now fix ๐‘ฅโˆˆโ„, and take ๐‘Ÿ<๐›ฟ. Then,

12๐‘Ÿโˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆ<12๐‘Ÿโˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ๐œ–๐‘๐‘‘๐‘ฆ=๐œ–๐‘

Since this holds for all ๐‘Ÿ<๐›ฟ, we get:

limโ€‰sup๐‘Ÿโ†’012๐‘Ÿโˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆโ‰ค๐œ–๐‘

Since ๐œ–>0 was arbitrary, this proves claim 1:

lim๐‘Ÿโ†’012๐‘Ÿโˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆ=0

Next we will prove the general case.
Step 1: Translate the problem into proving the measure of disqualified points is zero, for which we can use arbitrary error bound.
Define for each ๐‘ฅโˆˆโ„,๐‘Ÿ>0:

๐‘„(๐‘ฅ,๐‘Ÿ):=โˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆ=โˆฅ๐‘“๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘๐‘

And then we define for each ๐‘ฅโˆˆโ„:

๐‘„(๐‘ฅ):=limโ€‰sup๐‘Ÿโ†’0+๐‘„(๐‘ฅ,๐‘Ÿ)1/๐‘(2๐‘Ÿ)1/๐‘

Then what we want to show is just:

๐‘š({๐‘ฅ:๐‘„(๐‘ฅ)>0})=0

which is equivalent to show:

๐‘š({๐‘ฅ:๐‘„(๐‘ฅ)โ‰ฅ๐›ผ})=0for all ๐›ผ>0

Fix ๐›ผ>0. It suffices to show: for any ๐œ–>0, we have:

๐‘š({๐‘ฅ:๐‘„(๐‘ฅ)โ‰ฅ๐›ผ})<๐œ–

Now fix ๐œ–>0. Take ๐‘”โˆˆ๐ถ๐‘0(โ„) s.t. โˆฅ๐‘“โˆ’๐‘”โˆฅ๐‘<๐œ–. This can be done, by the density of ๐ถ๐‘0(โ„) in ๐ฟ๐‘(๐‘š).
Step 2: Bound the lim๐‘Ÿโ†’012๐‘Ÿโˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆ by ๐œ–-controllable expressions, using Minkowskiโ€™s ineq; thus bound the measure of disqualified points by two ๐œ–-controllable sets
Define for each ๐‘ฅโˆˆโ„,๐‘Ÿ>0:

๐‘„(๐‘ฅ,๐‘Ÿ):=โˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ|๐‘“(๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐‘ฆ=โˆฅ๐‘“๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘๐‘

This is nonnegative. And since |๐‘“โˆ’๐‘“(๐‘ฅ)| is measurable and ๐ฟ๐‘ (since |๐‘“| is ๐ฟ๐‘), |๐‘“โˆ’๐‘“(๐‘ฅ)|๐‘ is ๐ฟ1, and thus, recall we proved in lecture that ๐‘„(๐‘ฅ,๐‘Ÿ) is jointly continuous in ๐‘Ÿ and ๐‘ฅ.
By triangular ineq

๐‘„(๐‘ฅ,๐‘Ÿ)1/๐‘โ‰ค(โˆซ๐‘ฅโˆ’๐‘Ÿ๐‘ฅ+๐‘Ÿ(|๐‘“(๐‘ฆ)โˆ’๐‘”(๐‘ฆ)|+|๐‘”(๐‘ฆ)โˆ’๐‘”(๐‘ฅ)|+|๐‘”(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|)๐‘๐‘‘๐‘ฆ)1/๐‘

Then by Minkowskiโ€™s ineq:

๐‘„(๐‘ฅ,๐‘Ÿ)1/๐‘โ‰คโˆฅ๐‘“๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘”๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘+โˆฅ๐‘”๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘”(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘+โˆฅ๐‘”(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘

Thus

limโ€‰sup๐‘Ÿโ†’0+๐‘„(๐‘ฅ,๐‘Ÿ)1/๐‘(2๐‘Ÿ)1/๐‘โ‰คlimโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ตโˆฅ๐‘(2๐‘Ÿ)1/๐‘+limโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘”๐œ’๐ตโˆ’๐‘”(๐‘ฅ)๐œ’๐ตโˆฅ๐‘(2๐‘Ÿ)1/๐‘+limโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘”(๐‘ฅ)๐œ’๐ตโˆ’๐‘“(๐‘ฅ)๐œ’๐ตโˆฅ๐‘(2๐‘Ÿ)1/๐‘=limโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ตโˆฅ๐‘(2๐‘Ÿ)1/๐‘+limโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘”(๐‘ฅ)๐œ’๐ตโˆ’๐‘“(๐‘ฅ)๐œ’๐ตโˆฅ๐‘(2๐‘Ÿ)1/๐‘

Since we already proved the middle one of the three norms is zero, as continuous funciton with cpt supp.
Step 2: Reduce the statement to For simplication of notation, we also define for each ๐‘ฅโˆˆโ„:

๐‘€1(๐‘ฅ)โ‰”limโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘“๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘”๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘(2๐‘Ÿ)1/๐‘,๐‘€2(๐‘ฅ):=limโ€‰sup๐‘Ÿโ†’0+โˆฅ๐‘”(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘(2๐‘Ÿ)1/๐‘

By the ineq we obtained, we have:

{๐‘ฅ:๐‘„(๐‘ฅ)โ‰ฅ๐›ผ}โŠ‚{๐‘ฅ:๐‘€1(๐‘ฅ)โ‰ฅ๐›ผ2}โˆช{๐‘ฅ:๐‘€2(๐‘ฅ)โ‰ฅ๐›ผ2}

Since if we have both ๐‘€1(๐‘ฅ)<๐›ผ2 and ๐‘€2(๐‘ฅ)<๐›ผ2, we cannot have ๐‘„(๐‘ฅ)โ‰ฅ๐›ผ.
Thus

๐‘š{๐‘ฅ:๐‘„(๐‘ฅ)โ‰ฅ๐›ผ}โ‰ค๐‘š{๐‘ฅ:๐‘€1(๐‘ฅ)โ‰ฅ๐›ผ2}+๐‘š{๐‘ฅ:๐‘€2(๐‘ฅ)โ‰ฅ๐›ผ2}

Step 3: Bound ๐‘š{๐‘ฅ:๐‘€1(๐‘ฅ)โ‰ฅ๐›ผ2} using HL max Thm.
Note

โˆฅ๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ตโˆฅ๐‘(2๐‘Ÿ)1/๐‘=(12๐‘Ÿโˆซ|๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ต|๐‘)1๐‘

And we can express it as HL max function of

sup๐‘Ÿ12๐‘Ÿโˆซ|๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ต|๐‘=๐ป(๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ต)๐‘(๐‘ฅ)

We want

๐‘š{๐‘ฅ:(๐ป(๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ต)๐‘(๐‘ฅ))1/๐‘>๐›ผ2}=๐‘š{๐‘ฅ:๐ป(๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ต)๐‘(๐‘ฅ)>(๐›ผ2)๐‘}

And by HL max Thm:

๐‘š{๐‘ฅ:๐ป(๐‘“๐œ’๐ตโˆ’๐‘”๐œ’๐ต)๐‘(๐‘ฅ)>(๐›ผ2)๐‘}โ‰ค2๐‘3๐‘›๐›ผ๐‘โˆซ(|๐‘“โˆ’๐‘”|๐œ’๐ต)๐‘โ‰ค2๐‘3๐‘›๐›ผ๐‘โˆซ|๐‘“โˆ’๐‘”|๐‘โ‰ค2๐‘3๐‘›๐›ผ๐‘๐œ–๐‘

Step 4: Bound ๐‘š{๐‘ฅ:๐‘€2(๐‘ฅ)โ‰ฅ๐›ผ2} using Markovโ€™s ineq.
Notice that ๐‘€2(๐‘ฅ) is independent with ๐‘Ÿ:

โˆฅ๐‘”(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)๐œ’๐ต๐‘Ÿ(๐‘ฅ)โˆฅ๐‘(2๐‘Ÿ)1/๐‘=((๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ))๐‘2๐‘Ÿ)1/๐‘(2๐‘Ÿ)1/๐‘=(๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ))๐‘

Thus

๐‘š{๐‘ฅ:๐‘€2(๐‘ฅ)โ‰ฅ๐›ผ2}=๐‘š{๐‘ฅ:(๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ))๐‘โ‰ฅ๐›ผ2}

Therefore by Markovโ€™s ineq:

๐‘š{๐‘ฅ:๐‘€2(๐‘ฅ)โ‰ฅ๐›ผ2}=๐‘š{๐‘ฅ:(๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ))๐‘โ‰ฅ๐›ผ2}โ‰ค2๐›ผโˆซ(๐‘“(๐‘ฅ)โˆ’๐‘”(๐‘ฅ))๐‘=2๐›ผ๐œ–๐‘

Put it all together we have:

๐‘š{๐‘ฅ:๐‘„(๐‘ฅ)โ‰ฅ๐›ผ}โ‰ค(2๐‘3๐‘›๐›ผ๐‘+2๐›ผ)๐œ–๐‘

Since ๐œ– is arbitrary, we finally proved that

๐‘š{๐‘ฅ:๐‘„(๐‘ฅ)โ‰ฅ๐›ผ}=0for any ๐›ผ

finishing the proof.

โ–ก

generalization of Hรถlder: bootstrapped Hรถlder

Prove the following generalization of Hรถlderโ€™s inequality. Let 0<๐‘ <โˆž and 0<๐‘1,โ€ฆ,๐‘๐‘›<โˆž be such that

1๐‘1+1๐‘2+โ€ฆ+1๐‘๐‘›=1๐‘ ;

then

โˆฅ๐‘“1๐‘“2โ‹ฏ๐‘“๐‘›โˆฅ๐‘ โ‰คโˆฅ๐‘“1โˆฅ๐‘1โˆฅ๐‘“2โˆฅ๐‘2โ‹ฏโˆฅ๐‘“๐‘›โˆฅ๐‘๐‘›.
Proof

We prove by induction, applying Hรถlderโ€™s inequality each time.
base case: If ๐‘›=1 then the result is Hรถlderโ€™s inequality, as proved.
Inductive step: Suppose the inequality holds for all ๐‘ ,๐‘1,โ‹ฏ,๐‘๐‘›โˆ’1 such that the equality holds, then we assume there are ๐‘› positive reals ๐‘1,โ‹ฏ,๐‘๐‘› and some ๐‘ >0 s.t.

1๐‘1+1๐‘2+โ€ฆ+1๐‘๐‘›=1๐‘ 

WTS the ineq also hold.
We set:

1๐‘Ÿโ‰”1๐‘1+1๐‘2+โ‹ฏ+1๐‘๐‘›โˆ’1

Then we have

1๐‘Ÿ+1๐‘๐‘›=1๐‘ 

By the induction hypothesis applying to the ๐‘›โˆ’1 functions ๐‘“1,โ€ฆ,๐‘“๐‘›โˆ’1, we have

โˆฅ๐‘“1๐‘“2โ‹ฏ๐‘“๐‘›โˆ’1โˆฅ๐‘Ÿโ‰คโˆฅ๐‘“1โˆฅ๐‘1โˆฅ๐‘“2โˆฅ๐‘2โ‹ฏโˆฅ๐‘“๐‘›โˆ’1โˆฅ๐‘๐‘›โˆ’1

Now we define:

๐‘”(๐‘ฅ)โ‰”๐‘“1(๐‘ฅ)๐‘“2(๐‘ฅ)โ‹ฏ๐‘“๐‘›โˆ’1(๐‘ฅ),โ„Ž(๐‘ฅ)=:๐‘“๐‘›(๐‘ฅ)

Applying the classical Hรถlder inequality with conjugate exponents ๐‘Ÿ and ๐‘๐‘›, we have:

โˆฅ๐‘”โ„Žโˆฅ๐‘ =โˆฅ๐‘“1๐‘“2โ‹ฏ๐‘“๐‘›โˆ’1โ‹…๐‘“๐‘›โˆฅ๐‘ โ‰คโˆฅ๐‘“1๐‘“2โ‹ฏ๐‘“๐‘›โˆ’1โˆฅ๐‘Ÿโ‹…โˆฅ๐‘“๐‘›โˆฅ๐‘๐‘›.

Putting it all together, we obtain:

โˆฅ๐‘”โ„Žโˆฅ๐‘ =โˆฅ๐‘“1๐‘“2โ‹ฏ๐‘“๐‘›โˆ’1โ‹…๐‘“๐‘›โˆฅ๐‘ โ‰ค|๐‘“1๐‘“2โ‹ฏ๐‘“๐‘›โˆ’1โˆฅ๐‘Ÿโˆฅ๐‘“๐‘›โˆฅ๐‘๐‘›โ‰ค(โˆฅ๐‘“1โˆฅ๐‘1โ‹ฏโˆฅ๐‘“๐‘›โˆ’1โˆฅ๐‘๐‘›โˆ’1)โˆฅ๐‘“๐‘›โˆฅ๐‘๐‘›=โˆฅ๐‘“1โˆฅ๐‘1โ‹ฏโˆฅ๐‘“๐‘›โˆฅ๐‘๐‘›

This completes the inductive step, and thus the proof of the generalized Hรถlder inequality.

โ–ก

Translated a function by ๐‘ก: ๐‘“๐‘กโ†’๐‘“ in ๐ฟ๐‘ (1โ‰ค๐‘<โˆž), but not in ๐ฟโˆž

For any measurable function ๐‘“:โ„โ†’โ„, set

๐‘“๐‘ฆ(๐‘ฅ)โ‰”๐‘“(๐‘ฅโˆ’๐‘ฆ),๐‘ฅโˆˆโ„
  • Suppose that ๐‘“ is continuous with compact support. Prove that lim๐‘ฆโ†’0โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅโˆž=0.

  • Suppose that ๐‘“โˆˆ๐ฟ๐‘(โ„) for some ๐‘โˆˆ[1,โˆž). Prove that lim๐‘ฆโ†’0โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅ๐‘=0.

  • Prove by example thatย (ii) is false for ๐‘=โˆž.

Proof

of (a):
Suppose ๐‘“ is continuous with compact support ๐พโŠ‚โ„, then it is uniformly continuous.
Let ๐œ–>0 and fix it. By uniform continuity, there exists ๐›ฟ>0 such that

|๐‘ฅโˆ’๐‘ง|<๐›ฟโŸน|๐‘“(๐‘ฅ)โˆ’๐‘“(๐‘ง)|<๐œ–

For given ๐‘ฆ, we have:

โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅโˆž=esssup๐‘ฅโˆˆโ„|๐‘“๐‘ฆ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|โ‰คsup๐‘ฅโˆˆโ„|๐‘“๐‘ฆ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|=sup๐‘ฅโˆˆโ„|๐‘“(๐‘ฅโˆ’๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|

Then for |๐‘ฆ|<๐›ฟ: for any ๐‘ฅ, |๐‘ฅโˆ’๐‘ฆโˆ’๐‘ฅ|=|๐‘ฆ|<๐›ฟ. Thus by uniform continuity, must have |๐‘“(๐‘ฅโˆ’๐‘ฆ)โˆ’๐‘“(๐‘ฅ)|<๐œ–. Thus we got:

โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅโˆžโ‰ค๐œ–โˆ€|๐‘ฆ|<๐›ฟ

Since ๐œ– is arbitrary, this proves that

lim๐‘ฆโ†’0โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅโˆž=0

โ–ก

Proof

of (b):
Since ๐ถ๐‘(โ„) is dense in ๐ฟ๐‘(โ„) for 1โ‰ค๐‘<โˆž, we can take a seq of continuous functions with compact support, say (๐œ‘๐‘›), s.t. ๐œ‘๐‘›โ†’๐‘“ in ๐ฟ๐‘.
Then for each ๐‘ฆโˆˆโ„, we can define

๐œ‘๐‘›๐‘ฆ(๐‘ฅ)โ‰”๐œ‘๐‘›(๐‘ฅโˆ’๐‘ฆ)

From (a) we have, for each ๐‘›:

lim๐‘ฆโ†’0โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅโˆž=0

Note that since each ๐œ‘๐‘› have compact ๐พ whose measure is finite, we have:

โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅ๐‘=โˆซ|๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›|๐‘๐‘‘๐‘šโ‰คโˆซsup๐‘ฅ|๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›|๐‘๐‘‘๐‘š=โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅโˆž๐‘๐‘š(๐พ)

Thus,

lim๐‘ฆโ†’0โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅโˆž=0โŸนlim๐‘ฆโ†’0โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅ๐‘=0

Also, by translation invariance of Lebesgue measure, for each ๐‘ฆ we have:

โˆฅ๐‘“๐‘ฆโˆ’๐œ‘๐‘›๐‘ฆโˆฅ๐‘=โˆฅ๐‘“โˆ’๐œ‘๐‘›โˆฅ๐‘

Therefore for each ๐‘ฆ, we can bound

โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅ๐‘โ‰คโˆฅ๐‘“๐‘ฆโˆ’๐œ‘๐‘›๐‘ฆโˆฅ๐‘+โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅ๐‘+โˆฅ๐œ‘๐‘›โˆ’๐‘“โˆฅ๐‘=2โˆฅ๐œ‘๐‘›โˆ’๐‘“โˆฅ๐‘+โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅ๐‘

The construction of bound has finished. Now Let ๐œ–>0 and fix it. We first choose ๐‘› large enough so that

|๐œ‘๐‘›โˆ’๐‘“โˆฅ๐‘<๐œ–3

and for the fixed ๐‘›, we choose ๐›ฟ s.t. for all |๐‘ฆ|<๐›ฟ we have

โˆฅ๐œ‘๐‘›๐‘ฆโˆ’๐œ‘๐‘›โˆฅ๐‘<๐œ–3

Then we have:

โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅ๐‘โ‰ค๐œ–โˆ€|๐‘ฆ|<๐›ฟ

Since ๐œ– is arbitrary, this proves that

lim๐‘ฆโ†’0โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅ๐‘=0

โ–ก

Proof

of (c):

We consider

๐‘“(๐‘ฅ)โ‰”๐œ’(0,1)

We have

โˆฅ๐‘“โˆฅโˆž=1

and the sup is taken on ๐‘ฅโˆˆ(0,1).
Then for any ๐‘ฆ, we have: We have

|๐‘“๐‘ฆ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|=|๐œ’(0,1)(๐‘ฅโˆ’๐‘ฆ)โˆ’๐œ’(0,1)(๐‘ฅ)|=|๐œ’(๐‘ฆ,๐‘ฆ+1)(๐‘ฅ)โˆ’๐œ’(0,1)(๐‘ฅ)|

Thus for all ๐‘ฆ>0, on the open set (1,๐‘ฆ+1) which has positive measure, we have |๐‘“๐‘ฆ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|=1;
For all ๐‘ฆ<0, on the open set (๐‘ฆ,0) which has positive measure, we have |๐‘“๐‘ฆ(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|=1; Thus the function โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅโˆž with respect to ๐‘ฆ actually has a jump discontinuity at 0, since it is 0 at ๐‘ฆ=1 and 1 elsewhere.
This serves as an counterexample that we do not necessarily have lim๐‘ฆโ†’0โˆฅ๐‘“๐‘ฆโˆ’๐‘“โˆฅโˆž=0.

โ–ก

Criterion for ๐ฟ๐‘-convergence: a.e. conv + ็งฏๅˆ†ๅ€ผ conv

Suppose that 1โ‰ค๐‘<โˆž and that ๐‘“๐‘›,๐‘“โˆˆ๐ฟ๐‘ for some measure space (๐‘‹,๐’œ๏ธ€,๐œ‡). Prove that if ๐‘“๐‘›โ†’๐‘“ a.e. and โˆฅ๐‘“๐‘›โˆฅ๐‘โ†’โˆฅ๐‘“โˆฅ๐‘, then โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐‘โ†’0. Is the converse true? Hint: revisit the โ€œGeneralized DCTโ€ problem on HW5.

Proof

Recall we have proved

Theorem 9.55 : Generalized DCT

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a measure space, and ๐‘“๐‘›,๐‘”๐‘›,๐‘“,๐‘”โˆˆ๐ฟ1, ๐‘›โˆˆโ„•. Suppose that

  • lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=๐‘“(๐‘ฅ) and lim๐‘›โ†’โˆž๐‘”๐‘›(๐‘ฅ)=๐‘”(๐‘ฅ) for a.e. ๐‘ฅ;

  • |๐‘“๐‘›(๐‘ฅ)|โ‰ค๐‘”๐‘›(๐‘ฅ) a.e. for every ๐‘›โˆˆโ„•;

  • ๐‘”๐‘›:๐‘‹โ†’[0,โˆž] and lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐œ‡=โˆซ๐‘”๐‘‘๐œ‡.

Then we have:

lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡=โˆซ๐‘“๐‘‘๐œ‡

which is the case ๐‘=1. Now we prove the general case with the help of the case ๐‘=1. We notice that ๐‘“๐‘›โ†’๐‘“ in ๐ฟ๐‘, is just to prove the function |๐‘“๐‘›โˆ’๐‘“|๐‘โ†’0 in ๐ฟ1, thatโ€™s how we can use the generalized DCT.
Assume the hypothesis. Since ๐‘ฅ๐‘ is convex for ๐‘โ‰ฅ1, we have for any ๐‘ฅ,๐‘ฆ:

(๐‘ฅ+๐‘ฆ2)๐‘โ‰ค๐‘ฅ๐‘+๐‘ฆ๐‘2

Thus

(๐‘ฅ+๐‘ฆ)๐‘โ‰ค2๐‘โˆ’1(๐‘ฅ๐‘+๐‘ฆ๐‘)

Therefore for each ๐‘› and almost every ๐‘ฅ, we have:

|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|๐‘โ‰ค(|๐‘“๐‘›(๐‘ฅ)|+|๐‘“(๐‘ฅ)|)๐‘โ‰ค2๐‘โˆ’1(|๐‘“๐‘›(๐‘ฅ)|๐‘+|๐‘“(๐‘ฅ)|๐‘)

Hence

|๐‘“๐‘›โˆ’๐‘“|๐‘โ‰ค2๐‘โˆ’1(|๐‘“๐‘›|๐‘+|๐‘“|๐‘)

We define for each ๐‘›:

๐‘”๐‘›โ‰”2๐‘โˆ’1(|๐‘“๐‘›|๐‘+|๐‘“|๐‘)

Since ๐‘“๐‘›โ†’๐‘“ a.e., we have |๐‘“๐‘›|๐‘โ†’|๐‘“|๐‘ a.e. Thus

๐‘”๐‘›(๐‘ฅ)=2๐‘โˆ’1(|๐‘“๐‘›(๐‘ฅ)|๐‘+|๐‘“(๐‘ฅ)|๐‘)โ†’๐‘›โ†’โˆž2๐‘โˆ’1(|๐‘“(๐‘ฅ)|๐‘+|๐‘“(๐‘ฅ)|๐‘)=2๐‘|๐‘“(๐‘ฅ)|๐‘=:๐‘”(๐‘ฅ)

Note that

โˆซ๐‘”๐‘›๐‘‘๐œ‡=2๐‘โˆ’1(โˆฅ๐‘“๐‘›โˆฅ๐‘๐‘+โˆฅ๐‘“โˆฅ๐‘๐‘)

Since โ€–๐‘“๐‘›โ€–๐‘โ†’โ€–๐‘“โ€–๐‘, we have

lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐œ‡=2๐‘โˆ’1(โˆฅ๐‘“โˆฅ๐‘๐‘+โˆฅ๐‘“โˆฅ๐‘๐‘)=2๐‘โˆฅ๐‘“โˆฅ๐‘๐‘=โˆซ๐‘”๐‘‘๐œ‡

Now we have (1) ๐‘”๐‘›โ†’๐‘”, (2) โˆซ๐‘”๐‘›โ†’โˆซ๐‘”, and (3) ๐‘”๐‘› is an upper bound for |๐‘“๐‘›โˆ’๐‘“|๐‘. Then we can apply generalized DCT to the function seq |๐‘“๐‘›โˆ’๐‘“|๐‘:

lim๐‘›โ†’โˆžโˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐‘๐‘=lim๐‘›โ†’โˆžโˆซ|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|๐‘๐‘‘๐œ‡=โˆซ0๐‘‘๐œ‡=0

Thus

lim๐‘›โ†’โˆžโˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ๐‘=01/๐‘=0

This finishes the proof that ๐‘“๐‘›โ†’๐‘“ in ๐ฟ๐‘.

โ–ก

Solution

The converse does not hold.
We recall the typewriter function on [0,1]:

๐‘“๐‘›,๐‘˜(๐‘ฅ)={1,๐‘ฅโˆˆ[๐‘›โˆ’12๐‘˜,๐‘›2๐‘˜]0,otherwise

We index over ๐‘˜โˆˆโ„•, and for each ๐‘˜ we index over ๐‘›=1 to 2๐‘˜. That is, for given ๐‘˜, ๐‘“๐‘› is the indicator function of the ๐‘›-th dyadic interval.
Then

โˆฅ๐‘“๐‘›โˆฅ๐‘=(โˆซ[0,1]|๐‘“๐‘›(๐‘ฅ)|๐‘๐‘‘๐‘ฅ)1/๐‘=(length of the dyadic interval)1/๐‘โ‰ค2โˆ’๐‘˜/๐‘

Therefore, since each ๐‘“๐‘› has support of shrinking length, we get:

โˆฅ๐‘“๐‘›,๐‘˜โˆฅ๐‘โ†’0as ๐‘˜โ†’โˆž

but for each ๐‘ฅ, ๐‘“๐‘›,๐‘˜(๐‘ฅ)=1 for infinitely many (๐‘›,๐‘˜). so ๐‘“๐‘›(๐‘ฅ) does not converge to 0 for any ๐‘ฅโˆˆ[0,1].

Nur fรผr Verrรผckte

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!)

  1. Prove that the category of measurable spaces (see HW1) admits finite products, and that the product of (๐‘‹,๐’œ๏ธ€) and (๐‘Œ,โ„ฌ๏ธ€) equals (๐‘‹ร—๐‘Œ,๐’œ๏ธ€โŠ—โ„ฌ๏ธ€).

  2. Now consider the category of measure spaces (see HW2). Consider two measure spaces (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–), ๐‘–=1,2, and set ๐‘‹=๐‘‹1ร—๐‘‹2, ๐’œ๏ธ€=๐’œ๏ธ€1โŠ—๐’œ๏ธ€2, and ๐œ‡=๐œ‡1ร—๐œ‡2.

    • Prove that the projection maps ๐‘‹โ†’๐‘‹๐‘– are measurable, and that they are measure preserving iff ๐œ‡๐‘—(๐‘‹๐‘—)=1 for ๐‘—=1,2. Thus (๐‘‹,๐’œ๏ธ€,๐œ‡) is not the categorical product of (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–) in general.

    • Prove that even if ๐œ‡๐‘–(๐‘‹๐‘–)=1, the measure space (๐‘‹,๐’œ๏ธ€,๐œ‡) is not the categorical product of (๐‘‹๐‘–,๐’œ๏ธ€๐‘–,๐œ‡๐‘–) in general. Hint: consider the case when the ๐‘‹๐‘– consist of two elements, for example ๐‘‹๐‘–={๐”ฌ๐‘–,๐”ณ๐‘–}.

10 signed measure and Jordan decomposition

10.1 signed measure [Fol 3.1]

10.1.1 remainder: ๅฝ“ๅ‰ Folland ่ฟ›ๅบฆ

ๆˆ‘ไปฌ็›ฎๅ‰ finish ไบ† Folland ็š„ Ch1, Ch2, 6.1 ็š„ๅ…จ้ƒจ, 3.4 ็š„ๅคง้ƒจๅˆ†,
ๅ…ถๅฎžๅœจ่ฟ™ไธช lec ๅ‰่ฟ˜ๆœ‰ไธ€ไธช lec ่ฎฒไบ† Folland 5.1, 6.2 ็š„ไธ€้ƒจๅˆ†, ๅœจ่ฟ™้‡Œๆˆ‘ๅŽปๆމไบ†่ฟ™ไธช lec, ๆŠŠๅฎƒๆ”พๅœจไบ† 3.1-3.3 ็ป“ๆŸไน‹ๅŽ. ่ฟ™ๆ˜ฏๅ› ไธบ็”ฑไบŽ็›ฎๅ‰ๆฒกๆœ‰่ฏๆ˜Ž Radon-Nikodym Thm, ๆฒกๆœ‰่ถณๅคŸ็š„ๅทฅๅ…ทๅŽปๅฎŒๆˆ

(๐ฟ๐‘)โˆ—=๐ฟ๐‘ž

็š„่ฏๆ˜Ž (ๅทฎไบ†ไธ€ไธช proof surjectivity of the isometry ๐‘”โ†’โ„“๐‘”). ไธๆธ…ๆฅš่€ๅธˆไธบไป€ไนˆ่ฆๆŠŠๅฎƒๆ”พๅœจ่ฟ™้‡Œ่ฎฒ.
็Žฐๅœจๆˆ‘ไปฌๅฐ†ๅ›žๅˆฐ Ch3 ็š„ signed measure and differentiation between measures ็š„ theory, ๅœจๆŽฅไธ‹ๆฅ็š„ a few lectures ไธญ finish ๆމ Ch3.
ๅœจ finish ๆމ Ch3 ๅŽ, ๆˆ‘ไปฌๅฐ†ๆŽŒๆก่ถณๅคŸ็š„็Ÿฅ่ฏ†็ปง็ปญๆŽจ่ฟ› ๐ฟ๐‘ space ็š„็†่ฎบ, ไปŽ่€Œ finish ๅฎŒ 6.2, ็„ถๅŽๅฎŒๆˆ 6.3, 6.4 ็š„ไธ€้ƒจๅˆ†, ไปฅๅŠ a bit Hilbert space theory ๅ’Œ Fourier Analysis.
What will not be covered: Ch4 on point set topology (assume we have learned part of it, and the rest is not needed to be learned systematically) ไปฅๅŠๅ‰ฉไฝ™็š„ๆณ›ๅ‡ฝๅˆ†ๆžๅ†…ๅฎน (should be covered in functional analysis course next semester).

10.1.2 signed measure

Motivation: ๆˆ‘ไปฌ้ƒฝ็Ÿฅ้“, ๅฏนไบŽ nonnegative measurable ๐‘“ ๅณ ๐‘“โˆˆ๐ฟ+,

๐œˆ(๐ธ):=โˆซ๐ธ๐‘“๐‘‘๐œ‡

้€š่ฟ‡ integration of the function with respect to some measure ๐œ‡ ๅฎšไน‰ๅ‡บไบ†ๅฆไธ€ไธช measure ๐œˆ.
But what about ๐‘“โˆˆ๐ฟ1?

Definition 10.52 : signed measure

ไธ€ไธช signed measure on a measurable space (๐‘‹,๐’œ๏ธ€) ๆ˜ฏไธ€ไธช function ๐œˆ:๐’œ๏ธ€โ†’[โˆ’โˆž,โˆž) ๆˆ–่€… ๐œˆ:๐’œ๏ธ€โ†’(โˆ’โˆž,โˆž], ๅ’Œๆ™ฎ้€š meausre ไธ€ๆ ทๆปก่ถณ ๐œˆ(โŒ€)=0 ไปฅๅŠ ctbl disjoint additivity.
Note: signed measure ๅช admit +โˆž ๅ’Œ โˆ’โˆž ไธญ็š„ไธ€ไธชๅ€ผ (ไธๅฏไปฅๅŒๆ—ถๅญ˜ๅœจไธคไธช้›†ๅˆ ๐œˆ(๐ด)=โˆž, ๐œˆ(๐ต)=โˆ’โˆž)

Example 10.30

ๅฎนๆ˜“้ชŒ่ฏ:

Proposition 10.24

ๅฏนไบŽ positive measure ๐œ‡1,๐œ‡2, ๅฆ‚ๆžœๅ…ถไธญๆœ‰่‡ณๅฐ‘ไธ€ไธชๆ˜ฏ finite ็š„, ้‚ฃไนˆ

๐œˆ:=๐œ‡1โˆ’๐œ‡2

ๆ˜ฏไธ€ไธช signed measure.

This follows from ctbl disjoint additivity. (ไธคไธช ctbl sum ๅŠ ่ตทๆฅ)

Example 10.31
Proposition 10.25

ๅฏนไบŽ measurable function ๐‘“, ๅฆ‚ๆžœ ๐‘“+ ๅ’Œ ๐‘“โˆ’ ไธญ่‡ณๅฐ‘ๆœ‰ไธ€ไธชๆ˜ฏ ๐ฟ1 ็š„ (่ฟ™ไธชๆกไปถๅผฑไบŽ ๐‘“โˆˆ๐ฟ1, ่ขซ็งฐไธบ ๐‘“ is extended ๐œ‡-integrable), ้‚ฃไนˆ

๐œˆ(๐ธ)=โˆซ๐ธ๐‘“๐‘‘๐œ‡

ๅฐฑๆ˜ฏไธ€ไธช well-defined ็š„ signed measure.

This follows from that (1) ๅฏนไบŽ ๐‘“โˆˆ๐ฟ+, ๐œˆ(๐ธ):=โˆซ๐ธ๐‘“๐‘‘๐œ‡ ๅฎšไน‰ไบ†ไธ€ไธช measure; (2) ไธŠไธ€ไธช proposition.

10.1.3 signed measure ็š„ CFB, CFA

Proposition 10.26 : continuity from below and above for signed measures

็ป™ๅฎš signed measure ๐œˆ, ๅฏนไบŽ increasing seq ๐ธ๐‘—, ๆœ‰

๐œˆ(โ‹ƒ๐‘—=1โˆž๐ธ๐‘—)=lim๐‘—โ†’โˆž๐œˆ(๐ธ๐‘—)

ๅฏนไบŽ decreasing seq ๐น๐‘—, ๆœ‰:

๐œˆ(โ‹‚๐‘—=1โˆž๐น๐‘—)=lim๐‘—โ†’โˆž๐œˆ(๐น๐‘—)
Proof

ๅ’Œ positive measure ็š„ CFB, CFA ไธ€่‡ด.

โ–ก

10.1.4 positive / negative / null set

Elementary fact: ๅฏนไบŽ signed measure ่€Œ่จ€,

๐ดโŠ‚๐ตโ‡’ฬธ๐œˆ(๐ด)โ‰ค๐œˆ(๐ต)

ไฝ†ๆ˜ฏ

Lemma 10.33

ๅฏนไบŽ signed measure ๐œˆ, ๅ’Œ measurable ๐ดโŠ‚๐ต,

๐œˆ(๐ด)=โˆžโŸน๐œˆ(๐ต)=โˆž

ไปฅๅŠๅŒ็†

๐œˆ(๐ด)=โˆ’โˆžโŸน๐œˆ(๐ต)=โˆ’โˆž
Proof

่ฟ™ๆ˜ฏๅ› ไธบ

๐ต=๐ดโŠ”(๐ต\๐ด)

็”ฑไบŽๆˆ‘ไปฌๅฏนไบŽ signed measure, ๅชๅ…่ฎธ โˆž,โˆ’โˆž ไธญ็š„ไธ€็งๆƒ…ๅ†ต, ๅ› ่€Œไธ่ฎบ ๐œˆ(๐ต\๐ด) ็š„ measure ไนŸๅŒๅ‘ๆ— ็ฉทๆˆ–่€…ๆœ‰็ฉท, ้ƒฝ่ƒฝๅคŸๆŽจๅ‡บ ๐ต ็š„ measure ไนŸๅŒๅ‘ๆ— ็ฉท.

โ–ก

Definition 10.53 : positive set, negative set, null set

็ป™ๅฎš signed measure ๐œˆ, ๅฏนไบŽ ๐ธโˆˆ๐’œ๏ธ€, ๆˆ‘ไปฌ็งฐ ๐ธ ๆ˜ฏไธ€ไธช postive set, ๅฆ‚ๆžœๅฏนไบŽๅฏนไบŽไปปๆ„็š„ ๐นโŠ‚๐ธ, ้ƒฝๆœ‰

๐œˆ(๐น)โ‰ฅ0

negative set ๅ’Œ null set ๅŒ็†.
Note: For signed measure, ไธ€ไธช้›†ๅˆ็š„ signed measure ไธบ 0 ๅนถไธไปฃ่กจๅฎƒ็š„ไปปไฝ•ๅญ้›†็š„ measure ไนŸๆ˜ฏ 0, ๅฎƒๅฏไปฅๆ˜ฏไธคไธชๆญฃ่ดŸ measure ็›ธๆŠต็š„้›†ๅˆ็š„ union. ๅ› ่€Œๆˆ‘ไปฌ่ฆ่ฟ™ๆ ท้ขๅค–ๅฎšไน‰ null set.

Lemma 10.34 : measurable subset preserves sign

measurable subset of a measurable set ๐นโŠ‚๐ธ preserves the sign of ๐ธ.
ๅณ: ๐ธ ๆ˜ฏไธ€ไธช positive / null / negative โŸนไปปๆ„ ๐นโŠ‚๐ธ ๆ˜ฏไธ€ไธช positive / null / negative.

Proof

By def, ๅฏไปฅ by contradiction ๅพ—ๅˆฐ.

โ–ก

Lemma 10.35 : positive, negative, null set ๅ†…็š„ๅฑ€้ƒจๆ€ง่ดจๅ’Œๆ™ฎ้€š็š„ measure space ไธ€ๆ ท

ๅฆ‚ๆžœ ๐ธ ๆ˜ฏไธ€ไธช positive set for signed measure ๐œˆ, ้‚ฃไนˆ

๐นโŠ‚๐ธโŸน๐œˆ(๐น)โ‰ค๐œˆ(๐ธ)

้€š่ฟ‡ไธŠไธ€ไธช lemma, ๐ธ ็š„ไปปไฝ•ๅญ้›†ไนŸๆœ‰่ฟ™ไธชๆ€ง่ดจ. ๅ› ่€Œ ๐ธ ๅฑ€้ƒจๆ˜ฏไธ€ไธชๆ™ฎ้€š็š„ measure space.
ๅŒ็†, ๅฆ‚ๆžœ ๐ธ ๆ˜ฏไธ€ไธช negative set, ้‚ฃไนˆ

๐นโŠ‚๐ธโŸน๐œˆ(๐น)โ‰ฅ๐œˆ(๐ธ)

ๅ› ่€Œ ๐ธ ๅฑ€้ƒจไนŸ็ญ‰ไปทไบŽๆ˜ฏไธ€ไธชๆ™ฎ้€š็š„ measure space, ๅชไธ่ฟ‡ๆ‰€ๆœ‰้›†ๅˆ็š„ measure ๅŠ ไธŠไบ†ไธ€ไธช่ดŸๅท.

Figureย 29:
Lemma 10.36

Countable union of positive / negative / null sets ไป็„ถๆ˜ฏ positive / negative / null sets.

Proof

Follows from Def. ไปปไฝ•ไธ€ไธช ๐ธ ็š„ๅญ้›†้ƒฝๅฏๅˆ†่งฃๆˆ่ฟ™ไธช ๐ธ1,๐ธ2,โ‹ฏ ไธญ็š„ๆŸไบ›้›†ๅˆ็š„ๅญ้›†็š„ at most ctbl disjoint union, whose measure add up to remain positive / negative / null measure.

โ–ก

Question: ็ป™ๅฎš signed measure ๐œˆ, ๅฎƒๆ˜ฏๅฆไธ€ๅฎš่ƒฝ่ขซ decompose into ไธคไธช positive measure ็š„ difference?

๐œˆ=๐œˆ+โˆ’๐œˆโˆ’?

Turns out that: there exists a canonical way to do this. ่ฟ™ไธชๅˆ†่งฃๅญ˜ๅœจไธ”ๆ˜ฏๅ”ฏไธ€็š„, ๅนถไธ”ๆญฃ็š„้ƒจๅˆ†ๅ’Œ่ดŸ็š„้ƒจๅˆ†ๆ˜ฏไธ็›ธไบค็š„ (ไธๅญ˜ๅœจไธ€ไธช้›†ๅˆๆ—ขๆœ‰้ž 0 ็š„ positive measure ๅˆๆœ‰้ž 0 ็š„ negative measure). ๆˆ‘ไปฌ็งฐ่ฟ™ไธช signed measure decomposition ไธบ Jordan decomposition.
ๆˆ‘ไปฌไธ‹่Š‚่ฏพไผš่ฏๆ˜Ž Jordan decomposition. ่ฟ™่Š‚่ฏพๆˆ‘ไปฌๅ…ˆ่ฏๆ˜Žไธ€ไธชๅพ—ๅˆฐ Jordan decomposition ็š„ๅ…ณ้”ฎๆญฅ้ชค: Hahn Decomposition.
Hahn Decomposition Theorem ่กจ็คบ: ไปปๆ„ไธ€ไธช signed measure ้ƒฝๆŠŠๆ•ดไธช็ฉบ้—ด ๐‘‹ ๅˆ’ๅˆ†ไธบไธคไธช a.e. ไธ็›ธไบค็š„ positive set ๐‘ƒ ๅ’Œ negative set ๐‘.
่ฟ™ไธช็ป“ๆžœๆ˜ฏ้žๅธธๆœ‰็”จ็š„. ๅ› ไธบๆˆ‘ไปฌ็Ÿฅ้“, ๅœจไธ€ไธช positive / negative / null set ๅ†…้ƒจ, ๆˆ‘ไปฌๅฏไปฅๆŠŠๅฎƒ็œ‹ไฝœๆˆไธ€ไธชๆ™ฎ้€š็š„ measure space. ๅ› ่€Œ, Hahn Decomposition Theorem ่ฏดๆ˜Žไบ†ไปปๆ„ไธ€ไธช signed measure ้ƒฝๆŠŠๆ•ดไธช็ฉบ้—ด ๐‘‹ ๅˆ’ๅˆ†ๆˆไธคไธชๆ™ฎ้€š็š„ measure space, ๅ…ถไธญไธ€ไธช็š„็ฌฆๅทๅ’Œ measure ่ฟ็ฎ—้ข ๅ€’ไธบ่ดŸ. ่ฟ™ๅฐฑๅŸบๆœฌ state ไบ† Jordan decomposition ็š„ๅ†…ๅฎน.

10.1.5 Hahn Decomposition

Theorem 10.56 : Hahn Decomposition Theorem

ๅฏนไบŽไปปๆ„ measurable space (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ไปปๆ„ signed measure ๐œˆ, ้ƒฝๅญ˜ๅœจไธ€ไธช positive set ๐‘ƒ ๅ’Œไธ€ไธช negative set ๐‘ s.t.

๐‘ƒโˆฉ๐‘=โŒ€

ๅนถไธ”

๐‘ƒโŠ”๐‘=๐‘‹

ๅณ๐‘‹ ่ขซ ๐œˆ ๅˆ’ๅˆ†ไธบไธ€ไธช positive measure space ๅ’Œไธ€ไธช negative measure space.
ๅนถไธ”, ่ฟ™ไธช decomposition ๆ˜ฏๅ”ฏไธ€็š„, in ๐œˆ-a.e. sense: ๅณ, ๅฆ‚ๆžœ ๐‘ƒโ€ฒ,๐‘โ€ฒ ๆ˜ฏ another pair of such decomposition, ๅฟ…็„ถๆœ‰:

๐‘ƒฮ”๐‘ƒโ€ฒ=๐‘ฮ”๐‘โ€ฒis null set
Figureย 30:
Proof

Uniqueness ๆ˜ฏ just be definition ็š„, ๅ› ไธบ้™คไบ† ๐‘ƒ,๐‘ ๅ†…้ƒจ็š„ null sets ๅฏไปฅ้šๆ„ไบค็ป™ๅฏนๆ–นไน‹ๅค–, ๅ…ถไป–ๅญ้›†้ƒฝๆ˜ฏไธฅๆ ผ็š„ positive set ๅ’Œ negative set, ไธๅฏ่ƒฝๆœ‰็ฌฌไบŒไธช decomposition. ๅ› ่€Œ STS existence.
WLOG ่€ƒ่™‘ ๐œˆ ไธ admit โˆž (่‡ณๅคš admit โˆ’โˆž). This makes sense ๅ› ไธบ otherwise we can consider โˆ’๐œˆ.
Set:

๐‘š:=sup{๐œˆ(๐ธ):๐ธโˆˆ๐’œ๏ธ€ positive set}

Pick seq of positive sets (๐’ซ๏ธ€๐‘—) in ๐’œ๏ธ€ s.t.

๐œˆ(๐‘ƒ๐‘—)โ†—๏ธŽ๐‘š

(่ฟ™ๆ˜ฏ doable ็š„ๅ› ไธบๅœจ positive sets ็š„้ƒจๅˆ†็ญ‰ไบŽไธ€ไธชๆญฃๅธธ็š„ measure space, ๅนถไธ”่ฟ™้‡Œ finite measure.) ๅนถ set

๐‘ƒโ‰”โ‹ƒ๐‘—=1โˆž๐‘ƒ๐‘—

ไปŽ่€Œ ๐‘ƒ ไนŸๆ˜ฏ positive ็š„ๅนถไธ”

๐œˆ(๐‘ƒ)=๐‘š<โˆž

Set:

๐‘:=๐‘ƒ๐‘

ๅช่ฆ show ๐‘ ๆ˜ฏไธ€ไธช negative set, ๅฐฑๅพ—่ฏไบ†.
ๆˆ‘ไปฌ argue by contradiction.
ๅ‡่ฎพ ๐‘ ไธๆ˜ฏ negative set, ้‚ฃไนˆๅญ˜ๅœจ ๐ดโŠ‚๐‘ s.t. ๐œˆ(๐ด)>0.
Pick ๐‘›1โˆˆโ„• the smallest number ไฝฟๅพ—ๅญ˜ๅœจ ๐ด1โŠ‚๐‘ s.t.

๐œˆ(๐ด1)โ‰ฅ1๐‘›1

Note ๐ด1 ไธๅฏ่ƒฝๆ˜ฏ positive set, ๅฆๅˆ™ ๐‘ƒโˆช๐ด1 ๅฐ†ๆ˜ฏไธ€ไธช positive set ๅนถไธ” ๐œˆ(๐‘ƒโˆช๐ด1)>๐‘š, contradicting with ๐‘š being the sup of measure among positive sets.
ๅ› ่€Œ ๐ด1 ไธญ, ๅฟ…้กปๅญ˜ๅœจ negative measure ็š„ set. ๆˆ‘ไปฌๅ† pick ๐‘›2โˆˆโ„•, the smallest number ไฝฟๅพ—ๅญ˜ๅœจ ๐ต2โŠ‚๐‘ s.t.

๐œˆ(๐ต2)โ‰คโˆ’1๐‘›2

ๅณ:

โˆ’1๐‘›2โˆ’1โ‰ค๐œˆ(๐ต2)โ‰คโˆ’1๐‘›2

ๅนถ Set

๐ด2:=๐ด1\๐ต2

ไปŽ่€Œ:

๐œˆ(๐ด2)โ‰ฅ๐œˆ(๐ด1)+1๐‘›2

ๆˆ‘ไปฌ recursively ๅš่ฟ™ไปถไบ‹, ๅพ—ๅˆฐ positive measure ็š„ seq (๐ด๐‘›) s.t.

๐‘โŠƒ๐ด1โŠƒ๐ด2โŠƒโ‹ฏ{๐œˆ(๐ด๐‘—)โ‰ฅ๐œˆ(๐ด๐‘—โˆ’1)+1๐‘›๐‘—for any ๐ธโŠ‚๐ด๐‘—,๐œˆ(๐ธ)<๐œˆ(๐ด๐‘—)+1๐‘›๐‘—+1โˆ’1

notice: ๐ด๐‘— ่ฟ™ไธช seq ็š„ measure ๆ˜ฏ้€’ๅขž็š„. ๆˆ‘ไปฌๅ–

๐ด:=โ‹‚๐‘—=1โˆž๐ด๐‘—

ไบŽๆ˜ฏ

๐œˆ(๐ด)=lim๐‘—โ†’โˆž๐œˆ(๐ด๐‘—)โ‰ฅโˆ‘๐‘—=1โˆž1๐‘›๐‘—

ๅ› ไธบ ๐ด ๆœ‰ positive measure, ่ฟ™ไธช measure ไธ€ๅฎšๆœ‰้™, ไปŽ่€Œ่ฟ™ไธช series ๆ”ถๆ•›, ๅ› ่€Œๆœ‰:

๐‘›๐‘—โ†’โˆžas ๐‘—โ†’โˆž

ๅ’Œไน‹ๅ‰ๅŒ็†, ๐ด ไธ่ƒฝๆ˜ฏ positive set, ๆ‰€ไปฅๅญ˜ๅœจ ๐ตโŠ‚๐ด ไฝฟๅพ— ๐œˆ(๐ต)<0.
Set

๐ดโ€ฒโ‰”๐ด\๐ต

ไบŽๆ˜ฏ

๐œˆ(๐ดโ€ฒ)>๐œˆ(๐ด)+1๐‘›,for some ๐‘›โ‰ฅ1

็”ฑไบŽ ๐‘›๐‘—โ†’โˆž, for some ๐‘—>1 ๆœ‰ ๐‘›<๐‘›๐‘—. ๆˆ‘ไปฌๅ–่ฟ™ไธช ๐‘— ๅนถ fix it. ็”ฑไบŽ ๐œˆ(๐ด) ๆฏ”ไปปไฝ• ๐œˆ(๐ด๐‘—) ้ƒฝๅคง, ๅฏไปฅๅพ—ๅˆฐ

๐œˆ(๐ดโ€ฒ)>๐œˆ(๐ด)+1๐‘›โ‰ฅ๐œˆ(๐ด๐‘—)+1๐‘›for all ๐‘—โ‰ฅ1

่ฟ™่ฏดๆ˜Ž, ๐ดโ€ฒ ๆ˜ฏไปŽ ๐ด๐‘— ไธญๅŽปๆމไบ†ไธ€ไธช่‡ณๅฐ‘ๆœ‰ 1๐‘› ็š„่ดŸๆต‹ๅบฆ็š„้›†ๅˆๅพ—ๅˆฐ็š„.
ไฝ†ๆ˜ฏ, recall how we picked ๐‘›๐‘—: ๐‘›๐‘— ๆ˜ฏ the smallest number ไฝฟๅพ—ๅญ˜ๅœจ ๐ตโŠ‚๐ด๐‘— s.t. ๐œˆ(๐ต)โ‰คโˆ’1๐‘›2, ๅ’Œ่ฟ™้‡Œ ๐‘›<๐‘›๐‘— ็Ÿ›็›พ. ไปŽ่€Œๅพ—่ฏ.

โ–ก

10.2 Jordan decomposition [Fol 3.1, finished]

ๅฏนไบŽไปปๆ„็š„ signed measure ๐œˆ, ๆˆ‘ไปฌๅทฒ็ป้€š่ฟ‡ Hahn-Decomposition ่ฏๆ˜Žไบ†ๅฎƒไธ€ๅฎšๆŠŠ้›†ๅˆๅˆ†ไธบไธ€ไธช positive set ๐‘ƒ ๅ’Œไธ€ไธช negative set ๐‘, ๅนถไธ” unique in ๐œˆ-a.e. sense.

Example 10.32

Consider mble space (โ„•,๐’ซ๏ธ€(โ„•)), ่€ƒ่™‘็”ฑ

๐œˆ({๐‘›})=๐‘›โˆ’3

ๅ’Œ countable subadditivity ็”Ÿๆˆ็š„ signed measure. ไปŽ่€Œ:

๐‘ƒ={1,2,3},๐‘=โ„•\๐‘ƒ

ไนŸๅฏไปฅๆŠŠ 3 ๅˆ’ๅˆ†่ฟ› ๐‘, ๅ› ไธบ {3},โŒ€ ๆ˜ฏ่ฟ™้‡Œๅ”ฏไธ€็š„ null set.

10.2.1 mutually singular s.m.

Definition 10.54 : mutually singular

ๆˆ‘ไปฌ็งฐไธคไธช signed measure ๐œˆ1,๐œˆ2 on (๐‘‹,๐’œ๏ธ€) ๆ˜ฏ mutually singular ็š„, ๅฆ‚ๆžœ ๐‘‹=๐ธ1โŠ”๐ธ2, ๅ…ถไธญ ๐ธ๐‘– ๆ˜ฏ ๐œˆ๐‘– ็š„ null set.
็ฎ€ๅ•่€Œ่จ€ๅฐฑๆ˜ฏ: ่ฟ™ไธคไธช measure ๅฏไปฅๆŠŠ

live on disjoint sets, ๅœจๅฏนๆ–น live on ็š„้ƒจๅˆ†ๆ€ปๆ˜ฏ null ็š„.

Figureย 32:
Example 10.33

1. ๆŠŠๆ‰€ๆœ‰ measurable set map to 0 ็š„ trivial measure ๅ’Œไปปๆ„ s.m. ้ƒฝ mutually singular.
2. ๅ†ๆฏ”ๅฆ‚:

(๐‘‹,๐’œ๏ธ€)=(โ„,โ„ฌ๏ธ€(โ„))

ๆˆ‘ไปฌ้€‰ๆ‹ฉ Lebesgue measure as ๐œˆ1, discrete measure as ๐œˆ2, Cantor measure as ๐œˆ3.

๐œˆ1:=๐‘š,๐œˆ2โ‰”โˆ‘๐‘—=1โˆž๐‘๐‘—๐›ฟ๐‘ฅ๐‘—,๐œˆ3:=๐œ‡๐ถ๐‘Ž๐‘›๐‘ก๐‘œ๐‘Ÿ

ๆˆ‘ไปฌๅ‘็Žฐ: ่ฟ™ไธ‰ไธช measure ไธญ็š„ไปปๆ„ไธคไธช้ƒฝๆ˜ฏ mutually singular ็š„.
ๅ› ไธบ discrete measure ๆ”ฏๆŒ็š„้›†ๅˆ {๐‘ฅ๐‘—}1โˆž ๆ˜ฏ countable ็š„, ๐‘š({๐‘ฅ๐‘—}1โˆž)=0; ่€ŒๅฏนไบŽ ({๐‘ฅ๐‘—}1โˆž)๐‘, ่ฟ™ไธช้›†ๅˆๆ˜ฏ discrete measure ็š„ null set, ๅ› ไธบๅฎƒๅนถไธๅŒ…ๅซๆŒ‡ๅฎš็š„ seq ไธญ็š„ไปปไฝ•ๅ…ƒ็ด , showing that

๐‘šโŠฅโˆ‘๐‘—=1โˆž๐‘๐‘—๐›ฟ๐‘ฅ๐‘—

ๅŒ็†, recall Cantor set ็š„ Lebesgue meausre ไธบ 0, ไปŽ่€Œๅฏไปฅ็”จ ๐ถ ๅ’Œ โ„\๐ถ ็š„ๅˆ†ๅ‰ฒๆฅ่ฏดๆ˜Ž

๐œ‡๐ถ๐‘Ž๐‘›๐‘ก๐‘œ๐‘ŸโŠฅ๐‘š

ๅนถไธ”ๅŒ็†, ็”ฑไบŽ Cantor measure ๆฒกๆœ‰ atom, ๅณๅ…ถไธญไปปไฝ•ไธ€ไธชๅ•็‚น้›†็š„ Cantor measure ้ƒฝๆ˜ฏ 0, ไปŽ่€Œไป็„ถๅฏไปฅ้‡‡็”จ {๐‘ฅ๐‘—}1โˆž ๅ’Œ ({๐‘ฅ๐‘—}1โˆž)๐‘ ็š„ๅˆ†ๅ‰ฒๆฅ่ฏดๆ˜Ž:

๐œ‡๐ถ๐‘Ž๐‘›๐‘ก๐‘œ๐‘ŸโŠฅโˆ‘๐‘—=1โˆž๐‘๐‘—๐›ฟ๐‘ฅ๐‘—

10.2.2 Jordan Decomposition Thm

็Žฐๅœจ, ๆˆ‘ไปฌๅฏนไบŽ ๐ธโˆˆ๐’œ๏ธ€ set

๐œˆ+(๐ธ):=๐œˆ(๐ธโˆฉ๐‘ƒ)โ‰ฅ0

ไปฅๅŠ

๐œˆโˆ’(๐ธ):=๐œˆ(๐ธโˆฉ๐‘)โ‰ฅ0
Lemma 10.37

ๅฏนไบŽ s.m. ๐œˆ, ๆˆ‘ไปฌ้€š่ฟ‡ Hahn Decomposition ๅพ—ๅˆฐ ๐‘ƒโŠ”๐‘=๐‘‹.
Now let

{๐œˆ+(๐ธ):=๐œˆ(๐ธโˆฉ๐‘ƒ)โ‰ฅ0๐œˆโˆ’(๐ธ):=๐œˆ(๐ธโˆฉ๐‘)โ‰ฅ0

Then:

  • ๐œˆ+,๐œˆโˆ’ ๆ˜ฏ (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ positive measure

  • ๐œˆ+,๐œˆโˆ’ ไธญ่‡ณๅฐ‘ๆœ‰ไธ€ไธชๆ˜ฏ finite measure (ๅฏนๅบ”ไบ† ๐œˆ admit ็š„ๆ˜ฏ โˆž ่ฟ˜ๆ˜ฏ โˆ’โˆž)

  • ๐œˆ=๐œˆ+โˆ’๐œˆโˆ’
  • ๐œˆ+โŠฅ๐œˆโˆ’
Proof

1. ๆ˜พ็„ถ, ๐œˆ+,๐œˆโˆ’ ้ƒฝๆ˜ฏ positive ๅ‡ฝๆ•ฐ, ๅนถไธ”็”ฑไบŽ

(โจ†๐‘—=1โˆž๐ธ๐‘—)โˆฉ๐‘ƒ=โจ†๐‘—=1โˆž(๐ธ๐‘—โˆฉ๐‘ƒ)

(ๅŒ็† for intersecting ๐‘), ๅฎƒไปฌๆปก่ถณ countable disjoint additivity, ๅ› ่€Œๆ˜ฏ (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ positive measure.
2. By signed measure ็š„ๅฎšไน‰, ๐œˆ+ ๅ’Œ ๐œˆโˆ’ ๅฟ…้กปๆœ‰ไธ€ไธช finite. ๅ› ่€Œ otherwise, ๅฆ‚ๆžœๅญ˜ๅœจๆŸไธช้›†ๅˆไธŠ่ฟ™ไธคไธช measure ้ƒฝ infinite measure ๅˆ™ not well-defined (contracting well-definedness of ๐œˆ); ๅฆ‚ๆžœไธๅญ˜ๅœจ่ฟ™ๆ ท็š„้›†ๅˆๅˆ™ ๐œˆ admit both โˆž and โˆ’โˆž (contradicting that ๐œˆ ๅช admit ่‡ณๅคšไธ€ไธชๆ— ็ฉท).
ย 

3.

๐œˆ=๐œˆ+โˆ’๐œˆโˆ’

ๆ˜ฏ็›ดๆŽฅ by Hahn Decomposition ็š„. ๅ› ไธบไปปไฝ•ไธ€ไธช measurable set ๐ธ ้ƒฝๅฏไปฅๆ‹†ๅˆ†ๆˆ

(๐ธโˆฉ๐‘ƒ)โŠ”(๐ธโˆฉ๐‘)

4. Directly follows from Hahn Decomposition.

โ–ก

ไธ‹้ขๆˆ‘ไปฌ่ฏๆ˜Ž Jordan decomposition:

Theorem 10.57 : Jordan decomposition theorem

ๅฏนไบŽไปปๆ„ s.m. ๐œˆ on (๐‘‹,๐’œ๏ธ€), ้ƒฝๅญ˜ๅœจๅ”ฏไธ€็š„ positive measure ๐œˆ+, ๐œˆโˆ’ s.t.

  • ๐œˆ+,๐œˆโˆ’ ๆ˜ฏ (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ positive measure

  • ๐œˆ+,๐œˆโˆ’ ไธญ่‡ณๅฐ‘ๆœ‰ไธ€ไธชๆ˜ฏ finite measure (ๅฏนๅบ”ไบ† ๐œˆ admit ็š„ๆ˜ฏ โˆž ่ฟ˜ๆ˜ฏ โˆ’โˆž)

  • ๐œˆ=๐œˆ+โˆ’๐œˆโˆ’
  • ๐œˆ+โŠฅ๐œˆโˆ’
Proof

Existence ๅฐฑๆ˜ฏๅ‰ไธ€ไธช lemma ไธ€ๆจกไธ€ๆ ท. ๆˆ‘ไปฌ็Ÿฅ้“, Jordan decomposition ็š„ๆต‹ๅบฆๅˆ†ๅ‰ฒๆฅ่‡ชไบŽ Hahn decomposition ็š„ๅ…จ้›†ๅˆ†ๅ‰ฒ.
STS Uniqueness:
ๆˆ‘ไปฌไปค ๐œˆ=๐œˆ+โˆ’๐œˆโˆ’ ไธบ้€š่ฟ‡ Hahn Decomposition ๅพ—ๅˆฐ็š„ Jordan decomposition, ๅ…ถไธญ ๐œˆ+โŠฅ๐œˆโˆ’ ๅˆ†ๅˆซ supported on ๐‘ƒ ๅ’Œ ๐‘.
Suppose ๐œˆ=๐œ‡+โˆ’๐œ‡โˆ’ ๆ˜ฏๅฆไธ€ไธช decomposition s.t. ๐œ‡+โŠฅ๐œ‡โˆ’. ไบŽๆ˜ฏๅญ˜ๅœจ ๐ธ,๐นโˆˆ๐’œ๏ธ€ s.t.

๐ธโŠ”๐น=๐‘‹,๐œ‡+(๐ธ)=๐œ‡โˆ’(๐น)=0

ๆˆ‘ไปฌๅ‘็Žฐ: ๐‘‹=๐ธโŠ”๐น ๆ˜ฏๅฆไธ€ไธช Hahn Decomposition of ๐œˆ. ๅ› ่€Œ

๐‘ƒฮ”๐ธ=๐‘ฮ”๐นis ๐œˆ-null

ไปŽ่€ŒๅฏนไบŽไปปๆ„ ๐ดโˆˆ๐’œ๏ธ€,

๐œ‡+(๐ด)=๐œ‡+(๐ดโˆฉ๐ธ)=๐œˆ(๐ดโˆฉ๐ธ)=๐œˆ(๐ดโˆฉ๐‘ƒ)=๐œˆ+(๐ด)

ๅ› ่€Œ

๐œ‡+=๐œˆ+

ไปฅๅŠๅŒ็†, ๐œˆโˆ’=๐œ‡โˆ’. ๅพ—่ฏ.

โ–ก

10.2.3 total variation measure

Definition 10.55 : total variation measure
|๐œˆ|:=๐œˆ++๐œˆโˆ’

Totcal variation measure ๅ’ŒๅŽŸ s.m. ็š„ๅ…ณ็ณป, ๅฏไปฅ็ฑปๆฏ”ไธ€ไธชๅ‡ฝๆ•ฐ็š„็ปๅฏนๅ€ผๅ‡ฝๆ•ฐๅ’Œๅฎƒ่‡ช่บซ็š„ๅ…ณ็ณป, ๅ› ไธบ

๐‘“=๐‘“+โˆ’๐‘“โˆ’,|๐‘“|=๐‘“++๐‘“โˆ’

ไฝ†ๆ˜ฏ่ฟ™้‡Œ, ่ฟ™ไธช |โ‹…| ็ฌฆๅทๅ’Œ็ปๅฏนๅ€ผ็š„ |โ‹…| ็ฌฆๅท็š„ๆ„ไน‰ๅนถไธไธ€่‡ด: ่ฟ™ไธช |๐œˆ| ๅนถไธๆ˜ฏ ๐œˆ ็š„็ปๅฏนๅ€ผๅ‡ฝๆ•ฐ. ๅœจ positive, negative, null sets ไธŠ, |๐œˆ| ็กฎๅฎžๆ˜ฏ ๐œˆ ็š„็ปๅฏนๅ€ผๅ‡ฝๆ•ฐ, ไฝ†ๆ˜ฏๅœจๅ†…้ƒจๆ—ขๆœ‰ positive measure ็š„้ƒจๅˆ†, ๅˆๆœ‰ negative measure ็š„้ƒจๅˆ†็š„้›†ๅˆ, ๅฎƒ็š„ total variation measure ๆ˜ฏ่ฆๆฏ”ๅฎƒ็š„ๅŽŸ s.m. ็š„็ปๅฏนๅ€ผๆ›ดๅคง็š„. ๅ› ่€Œๅฎƒๆ‰่ขซๅซๅšๅŽŸ s.m. ็š„ total variation measure, ่กจ็คบๆŸไธช้›†ๅˆๅ†…้ƒจ, ๅŽŸ s.m. ไปŽๆญฃๅˆฐ่ดŸ็š„ๆœ€ๅคงๅ˜ๅทฎ.

Lemma 10.38

|๐œˆ| ๆ˜ฏ (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ positive measure.
ๅนถไธ” |๐œˆ| finite iff ๐œˆ+ ๅ’Œ ๐œˆโˆ’ ้ƒฝ finite.
(Then we define: ๆˆ‘ไปฌ็งฐ ๐œˆ ๆ˜ฏ finite ็š„, if |๐œˆ| finite p.m.)

Proof

trivial.

โ–ก

10.2.4 integration w.r.t. s.m.

Definition 10.56 : integration w.r.t. signed measure

ๅฏนไบŽ signed measure ๐œˆ, ๆˆ‘ไปฌ set:

๐ฟ1(๐œˆ):=๐ฟ1(|๐œˆ|)=๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’)

ไธ”ๅฏนไบŽๆฏไธช ๐‘“โˆˆ๐ฟ1(๐œˆ), ๆˆ‘ไปฌ set:

โˆซ๐‘“๐‘‘๐œˆ:=โˆซ๐‘“๐‘‘๐œˆ+โˆ’โˆซ๐‘“๐‘‘๐œˆโˆ’
Proposition 10.27

ๆˆ‘ไปฌ็Ÿฅ้“, ๅฏนไบŽไปปๆ„ p.m. ๐œ‡ on (๐‘‹,๐’œ๏ธ€) ไปฅๅŠ ๐‘“โˆˆ๐ฟ1(๐œ‡),

๐œˆ(๐ธ):=โˆซ๐ธ๐‘“๐‘‘๐œ‡

ๅฎšไน‰ไบ† (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ไธ€ไธช s.m.
่€Œ้€š่ฟ‡็งฏๅˆ†ๅฎšไน‰ๅ‡บๆฅ็š„ s.m., ๅฏนไบŽไปปๆ„ไธ€ไธช ๐ธโˆˆ๐’œ๏ธ€, ๆœ‰:

๐œˆยฑ(๐ธ)=โˆซ๐ธ๐‘“ยฑ๐‘‘๐œ‡

ไปŽ่€Œ

|๐œˆ|(๐ธ)=โˆซ๐ธ|๐‘“|๐‘‘๐œ‡
Proof

่ฟ™ๆ˜ฏๅ› ไธบๆˆ‘ไปฌๅฎนๆ˜“้ชŒ่ฏ, by the procedure of Hahn decomp,

๐‘ฅโˆˆ๐‘ƒโ‡”๐‘“(๐‘ฅ)โ‰ฅ0

ๅ› ่€Œ

๐œˆ+(๐ธ)=โˆซ๐ธโˆฉ{๐‘“โ‰ฅ0}๐‘“๐‘‘๐œ‡=โˆซ๐ธ๐‘“+๐‘‘๐œ‡

โ–ก

We will learn that: ่ฟ™ไธช ๐‘“ ๆญฃๆ˜ฏ ๐œˆ w.r.t. ๐œ‡ ็š„ Radon-Nikodym derivative, ไปŽ่€Œ ๐‘“(๐‘ฅ) ่กจ็คบๅœจๆŸไธชๅ…ƒ็ด ๅค„, ๐œˆ ็›ธๅฏนไบŽ ๐œ‡ ็š„ๅ˜ๅŒ–่ถ‹ๅŠฟ. ่€Œ total variation measure of ๐œˆ ๆญฃๆ˜ฏๆŠŠๆ‰€ๆœ‰็š„ๅ…ƒ็ด ไธŠ็š„่ฟ™ไธชๅ˜ๅŒ–่ถ‹ๅŠฟ้ƒฝๅ–ๆญฃ (ๅณๅ–ๆ€ปๅ˜ๅŒ–้‡, ไธ็ฎกๆ–นๅ‘) ๅพ—ๅˆฐ็š„.

ไปค ๐œˆ be a s.m. on (๐‘‹,๐’œ๏ธ€), ๐ธโˆˆ๐’œ๏ธ€, ๅˆ™

  • |๐œˆ(๐ธ)|โ‰ค|๐œˆ|(๐ธ)
  • ๐ธ null w.r.t. ๐œˆโ‡”|๐œˆ|(๐ธ)=0โ‡”๐œˆ+(๐ธ)=๐œˆโˆ’(๐ธ)=0
  • ๅฆ‚ๆžœ ๐œ… ๆ˜ฏ (๐‘‹,๐’œ๏ธ€) ไธŠ็š„ๅฆไธ€ไธช s.m., ๅˆ™

    ๐œ…โŠฅ๐œˆโ‡”๐œ…โŠฅ|๐œˆ|โ‡”๐œ…โŠฅ๐œˆ+ and ๐œ…โŠฅ๐œˆโˆ’
Proof

By def ๆ˜“ๅพ—.

โ–ก

่ฟ™ไธค่Š‚่ฏพ็š„ๆ€ป็ป“
  • ๆˆ‘ไปฌๅฎšไน‰ไบ† signed measure;

  • ๆˆ‘ไปฌๅ‘็Žฐไธ€ไธช signed measure ๅฆ‚ๆžœไธ่ฎก่พƒ null sets, ไธ€ๅฎšๅฏไปฅๅ”ฏไธ€ๅœฐ่ขซๅˆ†่งฃๆˆไธ€ไธชๅ…จ positive set ๅ’Œไธ€ไธชๅ…จ negative set;

  • ๅนถไธ”้€š่ฟ‡่ฟ™ไธชๅฏน ๐‘‹ ็š„ไบŒๅˆ†, ๆˆ‘ไปฌไนŸๅพ—ๅˆฐไบ†ๅฏนๅŽŸ s.m. ๐œˆ ็š„ไบŒๅˆ† ๐œˆ=๐œˆ+โˆ’๐œˆโˆ’, ่ฟ™ไธชๅˆ†่งฃไนŸๆ˜ฏๅ”ฏไธ€็š„

  • ๆˆ‘ไปฌๅฎšไน‰ไบ† total varation measure of a s.m., |๐œˆ|:=๐œˆ++๐œˆโˆ’.

  • ๆˆ‘ไปฌๅฎšไน‰ไบ†ไป€ไนˆๆ ท็š„ๅ‡ฝๆ•ฐๅฏนไบŽไธ€ไธช s.m. ๐œˆ ๆ˜ฏๅฏ็งฏ็š„: ๅฏนไบŽ ๐œˆ+, ๐œˆโˆ’ ้ƒฝๅฏ็งฏๅณๅฏ. ไปŽ่€Œ general ็š„็งฏๅˆ†:

    โˆซ๐‘“๐‘‘๐œˆ:=โˆซ๐‘“๐‘‘๐œˆ+โˆ’โˆซ๐‘“๐‘‘๐œˆโˆ’=(โˆซโ„œ๐‘“๐‘‘๐œˆ++๐‘–โˆซโ„‘๐‘“๐‘‘๐œˆ+)โˆ’(โˆซโ„œ๐‘“๐‘‘๐œˆโˆ’+๐‘–โˆซโ„‘๐‘“๐‘‘๐œˆโˆ’)=((โˆซโ„œ๐‘“+๐‘‘๐œˆ+โˆ’โˆซโ„œ๐‘“โˆ’๐‘‘๐œˆ+)+๐‘–(โˆซโ„‘๐‘“+๐‘‘๐œˆ+โˆ’โˆซโ„‘๐‘“โˆ’๐‘‘๐œˆ+))โˆ’((โˆซโ„œ๐‘“+๐‘‘๐œˆโˆ’โˆ’โˆซโ„œ๐‘“โˆ’๐‘‘๐œˆโˆ’)+๐‘–(โˆซโ„‘๐‘“+๐‘‘๐œˆโˆ’โˆ’โˆซโ„‘๐‘“โˆ’๐‘‘๐œˆโˆ’))

    ่ฟ™ไธ€ไธชๅผๅญ้‡ŒๅŒ…ๅซไบ†ๅ…ซไธชๅฐ็งฏๅˆ†. ๆˆ‘ไปฌ็›ฎๅ‰ๅญฆๅˆฐ็š„ๅฐฑๆ˜ฏ่ฟ™ไนˆๅคš. ๅฆ‚ๆžœๅผ•ๅ…ฅ complex measure ็š„่ฏ,

Homework 9: on signed measure (50/50)

Three real Banach spaces and a fake one

  • Let

    โ„“0โˆžโ‰”{๐‘Ž=(๐‘Ž1,๐‘Ž2,โ‹ฏ)โˆฃ๐‘Ž๐‘–โˆˆโ„,lim๐‘›โ†’โˆž๐‘Ž๐‘›=0}.

    Prove that (โ„“0โˆž,โˆฅโ‹…โˆฅโˆž), where โˆฅ๐‘Žโˆฅโˆž=sup๐‘›|๐‘Ž๐‘›|, is a Banach space.

  • Let

    ๐ถ๐‘0(โ„)โ‰”{๐‘“:โ„โ†’โ„โˆฃ๐‘“ is continuous and bounded}.

    Prove that (๐ถ๐‘0(โ„),โˆฅโ‹…โˆฅโˆž), where โˆฅ๐‘“โˆฅโˆž=sup๐‘ฅโˆˆโ„|๐‘“(๐‘ฅ)|, is a Banach space.

  • Let

    ๐ถ00(โ„)โ‰”{๐‘“:โ„โ†’โ„โˆฃ๐‘“ is continuous, lim๐‘ฅโ†’ยฑโˆž๐‘“(๐‘ฅ)=0}.

    Prove that (๐ถ00(โ„),โˆฅโ‹…โˆฅโˆž), where โˆฅ๐‘“โˆฅโˆž=sup๐‘ฅโˆˆโ„|๐‘“(๐‘ฅ)|, is a Banach space.

  • Recall that

    ๐ถ๐‘0(โ„)={๐‘“:โ„โ†’โ„โˆฃ๐‘“ is continuous and ๐‘“=0 outside a bounded set}.

    Show that (๐ถ๐‘0(โ„),โˆฅโ‹…โˆฅโˆž), where โˆฅ๐‘“โˆฅโˆž=sup๐‘ฅโˆˆโ„|๐‘“(๐‘ฅ)|, is not a Banach space.

Proof

of (a): Since we showed in class that

โ„“โˆž=๐ฟโˆž(โ„•,๐’ซ๏ธ€(โ„•),๐œ‡๐‘๐‘œ๐‘ข๐‘›๐‘ก๐‘–๐‘›๐‘”)

and ๐ฟโˆž spaces are Banach, โ„“โˆž is Banach.
Thus it suffices to show that โ„“0โˆž is closed in โ„“โˆž, since a closed subset of a complete metric space is complete.
Let (๐‘Ž(๐‘˜))๐‘˜=1โˆž be a sequence in โ„“0โˆž converging in norm to ๐‘Žโˆˆโ„“โˆž, i.e.,

โˆฅ๐‘Ž(๐‘˜)โˆ’๐‘Žโˆฅโˆžโ†’0

Let ๐œ€>0.
Since โˆฅ๐‘Ž(๐‘˜)โˆ’๐‘Žโˆฅโˆžโ†’0, there exists ๐พ such that for all ๐‘˜โ‰ฅ๐พ,

โˆฅ๐‘Ž(๐‘˜)โˆ’๐‘Ž||=sup๐‘›|๐‘Ž๐‘›(๐‘˜)โˆ’๐‘Ž๐‘›|<๐œ€2

This implies that

โˆ€๐‘›,|๐‘Ž๐‘›(๐พ)โˆ’๐‘Ž๐‘›|<๐œ€

Since ๐‘Ž(๐พ)โˆˆโ„“0โˆž, ๐‘Ž๐‘›(๐พ)โ†’0 as ๐‘›โ†’โˆž. Thus there exists ๐‘โˆˆโ„• s.t. for all ๐‘›โ‰ฅ๐‘,

|๐‘Ž๐‘›(๐พ)|โ‰ค๐œ€2

Then for all ๐‘›โ‰ฅ๐‘, we have:

|๐‘Ž๐‘›|โ‰ค|๐‘Ž๐‘›โˆ’๐‘Ž๐‘›(๐พ)|+|๐‘Ž๐‘›(๐พ)|<๐œ€

This shows that

lim๐‘›โ†’โˆž|๐‘Ž๐‘›|<๐œ–

Since ๐œ€>0 is arbitrary, this implies

lim๐‘›โ†’โˆž๐‘Ž๐‘›=0

Hence ๐‘Žโˆˆโ„“0โˆž. So โ„“0โˆž is closed in โ„“โˆž, thus itself Banach.

โ–ก

Proof

of (b): Let (๐‘“๐‘›)๐‘›โˆˆโ„• be a Cauchy seq in (๐ถ๐‘0(โ„),โˆฅโ‹…โˆฅโˆž), then

โˆ€๐œ€>0,โˆƒ๐‘โˆˆโ„•๐‘ .๐‘ก.โˆฅ๐‘“๐‘›โˆ’๐‘“๐‘šโˆฅโˆž=sup๐‘ฅโˆˆโ„|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“๐‘š(๐‘ฅ)|<๐œ€

In particular, for each fixed ๐‘ฅโˆˆโ„, (๐‘“๐‘›(๐‘ฅ))๐‘›โˆˆโ„• is a Cauchy sequence in โ„, hence converges (since โ„ is complete). So we can define the pointwise limit:

๐‘“(๐‘ฅ)โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)

Claim 1: ๐‘“๐‘›โ†’๐‘“ in โˆฅโ‹…โˆฅโˆž.
Let ๐œ€>0.
Since (๐‘“๐‘›) is Cauchy in โˆฅโ‹…โˆฅโˆž, there exists ๐‘ such that:

โˆฅ๐‘“๐‘›โˆ’๐‘“๐‘šโˆฅโˆž<๐œ€,โˆ€๐‘›,๐‘šโ‰ฅ๐‘

Fix ๐‘šโ‰ฅ๐‘, and let ๐‘›โ†’โˆž. For each ๐‘ฅ, we get:

|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“๐‘š(๐‘ฅ)|<๐œ€โˆ€๐‘›โŸนlim๐‘›โ†’โˆž|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“๐‘š(๐‘ฅ)|=|๐‘“(๐‘ฅ)โˆ’๐‘“๐‘š(๐‘ฅ)|โ‰ค๐œ€

Since this is true for each ๐‘ฅโˆˆโ„, we obtain:

โˆฅ๐‘“โˆ’๐‘“๐‘šโˆฅโˆžโ‰ค๐œ€,for all ๐‘šโ‰ฅ๐‘

Since ๐œ€>0 is arbitrary, this shows that

lim๐‘›โ†’โˆžโˆฅ๐‘“โˆ’๐‘“๐‘›โˆฅโˆž=0

Claim 2: ๐‘“โˆˆ(๐ถ๐‘0(โ„),โˆฅโ‹…โˆฅโˆž).
Since lim๐‘›โ†’โˆžโˆฅ๐‘“โˆ’๐‘“๐‘›โˆฅโˆž=0, it also implies that the convergence is uniform.
We know the uniform limit of continuous functions is continuous, so ๐‘“ is continuous. It remains to show ๐‘“ is bounded, and this directly follows from the uniform convergence. We take ๐œ€=1. We have proved that there exists ๐‘ s.t. for all ๐‘šโ‰ฅ๐‘,

โˆฅ๐‘“โˆ’๐‘“๐‘šโˆฅโˆžโ‰ค1

Thus

sup๐‘ฅโˆˆโ„|๐‘“(๐‘ฅ)|โ‰คsup๐‘ฅโˆˆโ„|๐‘“๐‘(๐‘ฅ)|+1

Since ๐‘“๐‘›โˆˆ๐ถ๐‘0(โ„)), it is bounded, thus

sup๐‘ฅโˆˆโ„|๐‘“(๐‘ฅ)|<โˆž

showing that the limit function is bounded. This finishes the proof that ๐‘“โˆˆ๐ถ๐‘0(โ„). Thus, every Cauchy seq in (๐ถ๐‘0(โ„),โˆฅโ‹…โˆฅโˆž) converges in (๐ถ๐‘0(โ„),โˆฅโ‹…โˆฅโˆž), i.e. it is Banach.

โ–ก

Proof

of (c): Let (๐‘“๐‘›)๐‘›โˆˆโ„• be a Cauchy seq in (๐ถ00(โ„),โˆฅโ‹…โˆฅโˆž), then for each fixed ๐‘ฅโˆˆโ„, (๐‘“๐‘›(๐‘ฅ))๐‘›โˆˆโ„• is a Cauchy sequence in โ„, so for the same reason as (b), we can define the pointwise limit:

๐‘“(๐‘ฅ)โ‰”lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)

And for the same reason as (b), we get

๐‘“๐‘›โ†’๐‘“ in โˆฅโ‹…โˆฅโˆž

which also implies that the pointwise convergence is uniform. Since each ๐‘“๐‘› is continuous, the uniform limit ๐‘“ is continuous.
Thus it suffices to show that lim๐‘ฅโ†’ยฑโˆž๐‘“(๐‘ฅ)=0.
Let ๐œ–>0. Since ๐‘“๐‘›โ†’๐‘“ uniformly, there exists ๐‘ such that for all ๐‘›โ‰ฅ๐‘, โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅโˆž<๐œ–/2. Also, since ๐‘“๐‘โˆˆ๐ถ00(โ„), there exists ๐‘€>0 such that |๐‘“๐‘(๐‘ฅ)|<๐œ–/2 for all |๐‘ฅ|>๐‘€.
Then for |๐‘ฅ|>๐‘€,

|๐‘“(๐‘ฅ)|โ‰ค|๐‘“(๐‘ฅ)โˆ’๐‘“๐‘(๐‘ฅ)|+|๐‘“๐‘(๐‘ฅ)|<๐œ–/2+๐œ–/2<๐œ–

So lim๐‘ฅโ†’ยฑโˆž๐‘“(๐‘ฅ)=0, i.e., ๐‘“โˆˆ๐ถ00(โ„). Thus, every Cauchy seq in (๐ถ00(โ„),โˆฅโ‹…โˆฅโˆž) converges in (๐ถ00(โ„),โˆฅโ‹…โˆฅโˆž), i.e. it is Banach.

โ–ก

Proof

of (d): We consider a continuous (smooth actually) function ๐œ™:โ„โ†’โ„ with supp(๐œ™)=[0,2] (here we take the closure):

๐œ™(๐‘ฅ)โ‰”{exp(โˆ’1๐‘ฅ(2โˆ’๐‘ฅ)),0<๐‘ฅ<2,0,otherwise
Figureย 33:

This function reaches its maximum at ๐‘ฅ=1,

โˆฅ๐œ™โˆฅโˆž=1๐‘’

For each integer ๐‘›โ‰ฅ1, define

๐œ™๐‘›(๐‘ฅ)=๐œ™(๐‘ฅโˆ’๐‘›)

Then each ๐œ™๐‘› is also continuous, and supp(๐œ™๐‘›)=[๐‘›,๐‘›+2].
Consider the sequence (๐‘†๐‘)1โˆž, defined as:

๐‘†๐‘(๐‘ฅ)โ‰”โˆ‘๐‘›=1๐‘2โˆ’๐‘›๐œ™๐‘›(๐‘ฅ)

Then each ๐‘†๐‘โˆˆ๐ถ๐‘0(โ„), since finite sum of continuous functions is also continuous, and supp(๐‘†๐‘)=[1,๐‘+2], thus each ๐‘†๐‘โˆˆ๐ถ๐‘0(โ„).
Claim: (๐‘†๐‘)1โˆž is Cauchy in the sup norm.
This is because for each (WLOG) ๐‘€>๐‘โˆˆโ„•,

โˆฅ๐‘†๐‘€โˆ’๐‘†๐‘โˆฅโˆž=โˆฅโˆ‘๐‘›=๐‘+1๐‘€12๐‘›๐œ™๐‘›โˆฅโˆžโ‰คโˆ‘๐‘›=๐‘+1๐‘€12๐‘›โˆฅ๐œ™โˆฅโˆžโ‰คโˆ‘๐‘›=๐‘+1โˆž12๐‘›โˆฅ๐œ™โˆฅโˆž=โˆ‘๐‘›=๐‘+1โˆž12๐‘›๐‘’=12๐‘๐‘’โ†’๐‘โ†’โˆž0

Thus for arbitrary ๐œ€>0, exists ๐พโˆˆโ„• s.t. for all ๐‘€,๐‘โ‰ฅ๐พ, โˆฅ๐‘†๐‘€โˆ’๐‘†๐‘โˆฅโˆž<๐œ€. And by same reason as (b), (c), (๐‘†๐‘)1โˆž converges by โˆฅโ‹…โˆฅโˆž into its pointwise limit:

๐‘†(๐‘ฅ)โ‰”โˆ‘๐‘›=1โˆž2โˆ’๐‘›๐œ™๐‘›(๐‘ฅ)

But ๐‘†(๐‘ฅ) does not have compact support, supp(๐‘†)=[0,โˆž). So ๐‘†โˆ‰๐ถ๐‘0(โ„). This serves as a counterexample showing that ๐ถ๐‘0(โ„) is not Banach.

โ–ก

๐œˆ+(๐ธ),๐œˆโˆ’(๐ธ),|๐œˆ|(๐ธ) ็š„formula from original ๐œˆ

Let ๐œˆ be a signed measure on (๐‘‹,๐’œ๏ธ€), and ๐ธโˆˆ๐’œ๏ธ€. Prove the following statements:

  • ๐œˆ+(๐ธ)=sup{๐œˆ(๐น)โˆฃ:๐นโˆˆ๐’œ๏ธ€,๐นโŠ‚๐ธ}, and ๐œˆโˆ’(๐ธ)=โˆ’inf{๐œˆ(๐น)โˆฃ๐นโˆˆ๐’œ๏ธ€,๐นโŠ‚๐ธ};

  • |๐œˆ|(๐ธ)=sup{โˆ‘๐‘–=1๐‘|๐œˆ(๐ธ๐‘–)|โˆฃ๐‘โˆˆโ„•,๐ธ=โ‹ƒ๐‘–=1๐‘๐ธ๐‘– disjoint union};

  • |๐œˆ|(๐ธ)โ‰ฅ|๐œˆ(๐ธ)|. In the case ๐œˆ finite, it achieves equality iff ๐ธ is positive or negative for ๐œˆ.

Proof

of (i): By the Hahn decomposition theorem, we can take a Hahn decomposition ๐‘‹=๐‘ƒโŠ”๐‘ where

๐œˆ(๐ด)โ‰ฅ0for all ๐ดโŠ‚๐‘ƒ,๐œˆ(๐ต)โ‰ค0for all ๐ตโŠ‚๐‘

Fix ๐ธโˆˆ๐’œ๏ธ€. By Jordan decomposition we have

๐œˆ+(๐ธ)=๐œˆ(๐ธโˆฉ๐‘ƒ)

Fix ๐นโŠ‚๐ธ, we have:

๐น=(๐นโˆฉ๐‘ƒ)โŠ”(๐นโˆฉ๐‘)

Since ๐œˆ(๐นโˆฉ๐‘)โ‰ค0, we have:

๐œˆ(๐น)โ‰ค๐œˆ(๐นโˆฉ๐‘ƒ)โ‰ค๐œˆ(๐ธโˆฉ๐‘ƒ)=๐œˆ+(๐ธ)

Since ๐น is arbitrary, this shows:

sup{๐œˆ(๐น)โˆฃ๐นโŠ‚๐ธ}โ‰ค๐œˆ+(๐ธ)

On the other hand, taking ๐น=๐ธโˆฉ๐‘ƒโŠ‚๐ธ, we get

๐œˆ(๐น)=๐œˆ(๐ธโˆฉ๐‘ƒ)=๐œˆ+(๐ธ)

Hence

sup{๐œˆ(๐น)โˆฃ๐นโŠ‚๐ธ}โ‰ฅ๐œˆ+(๐ธ)

Combining both inequalities gives

๐œˆ+(๐ธ)=sup{๐œˆ(๐น)โˆฃ๐นโŠ‚๐ธ}

Similarly, since ๐œˆ(๐นโˆฉ๐‘ƒ)โ‰ฅ0 and ๐œˆ(๐น)=๐œˆ(๐นโˆฉ๐‘ƒ)+๐œˆ(๐นโˆฉ๐‘), we have ๐œˆ(๐น)โ‰ฅ๐œˆ(๐นโˆฉ๐‘). And Since ๐œˆ(๐ธโˆฉ๐‘)=๐œˆ(๐นโˆฉ๐‘)+๐œˆ((๐ธ\๐น)โˆฉ๐‘) with ๐œˆ((๐ธ\๐น)โˆฉ๐‘)โ‰ค0, we get ๐œˆ(๐นโˆฉ๐‘)โ‰ฅ๐œˆ(๐ธโˆฉ๐‘).
Putting it together:

๐œˆ(๐น)โ‰ฅ๐œˆ(๐นโˆฉ๐‘)โ‰ฅ๐œˆ(๐ธโˆฉ๐‘)=โˆ’๐œˆโˆ’(๐ธ)

Since ๐น is arbitrary, this shows:

inf{๐œˆ(๐น)โˆฃ๐นโŠ‚๐ธ}โ‰ฅโˆ’๐œˆโˆ’(๐ธ)

On the other hand, taking ๐น=๐ธโˆฉ๐‘โŠ‚๐ธ, we get

๐œˆ(๐น)=๐œˆ(๐ธโˆฉ๐‘)=โˆ’๐œˆโˆ’(๐ธ)

Hence

inf{๐œˆ(๐น)โˆฃ๐นโŠ‚๐ธ}โ‰คโˆ’๐œˆโˆ’(๐ธ)

Combining both inequalities gives

๐œˆโˆ’(๐ธ)=โˆ’inf{๐œˆ(๐น)โˆฃ๐นโŠ‚๐ธ}

โ–ก

Proof

of (ii): Let ๐ธโˆˆ๐’œ๏ธ€. By def of total variation measure,

|๐œˆ|(๐ธ)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)

One direction of the equality is easy. Take a Hahn decomposition ๐‘‹=๐‘ƒโŠ”๐‘ where

๐œˆ(๐ด)โ‰ฅ0for all ๐ดโŠ‚๐‘ƒ,๐œˆ(๐ต)โ‰ค0for all ๐ตโŠ‚๐‘

Then by Jordan decomposition, we have:

๐œˆ+(๐ธ)=๐œˆ(๐ธโˆฉ๐‘ƒ),๐œˆโˆ’(๐ธ)=โˆ’๐œˆ(๐ธโˆฉ๐‘)

So by taking ๐ธ1:=๐ธโˆฉ๐‘ƒ, ๐ธ2:=๐ธโˆฉ๐‘, we have:

|๐œˆ|(๐ธ)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)=๐œˆ(๐ธ1)+๐œˆ(๐ธ2)

This shows that

|๐œˆ|(๐ธ)โ‰คsup{โˆ‘|๐œˆ(๐ธ๐‘–)|}

And for the other direction, for any disjoint measurable partition ๐ธ=โ‹ƒ๐‘–=1๐‘๐ธ๐‘–, we have

|๐œˆ(๐ธ๐‘–)|=|๐œˆ+(๐ธ๐‘–)โˆ’๐œˆโˆ’(๐ธ๐‘–)|โ‰ค๐œˆ+(๐ธ๐‘–)+๐œˆโˆ’(๐ธ๐‘–)=|๐œˆ|(๐ธ๐‘–)

Therefore

โˆ‘๐‘–=1๐‘|๐œˆ(๐ธ๐‘–)|โ‰คโˆ‘๐‘–=1๐‘|๐œˆ|(๐ธ๐‘–)=|๐œˆ|(โ‹ƒ๐‘–=1๐‘๐ธ๐‘–)=|๐œˆ|(๐ธ)

since |๐œˆ| is a p.m. and the ๐ธ๐‘–โ€™s are disjoint. Thus

sup{โˆ‘๐‘–=1๐‘|๐œˆ(๐ธ๐‘–)|}โ‰ค|๐œˆ|(๐ธ)

Combining the two inequalities gives

|๐œˆ|(๐ธ)=sup{โˆ‘๐‘–=1๐‘|๐œˆ(๐ธ๐‘–)||๐‘โˆˆโ„•,๐ธ=โ‹ƒ๐‘–=1๐‘๐ธ๐‘– disjoint}

proving the statement.

โ–ก

Proof

of (iii): Let ๐ธโˆˆ๐’œ๏ธ€. The ineq |๐œˆ|(๐ธ)โ‰ฅ|๐œˆ(๐ธ)| follows from triangular ineq on โ„:

|๐œˆ(๐ธ)|=|๐œˆ+(๐ธ)โˆ’๐œˆโˆ’(๐ธ)|โ‰ค๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)=|๐œˆ|(๐ธ)

Now we assume ๐œˆ is finite (i.e.ย |๐œˆ|(๐‘‹)<โˆž). The equality condition |๐œˆ(๐ธ)|=|๐œˆ|(๐ธ) is detailedly:

|๐œˆ+(๐ธ)โˆ’๐œˆโˆ’(๐ธ)|=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)

Since |๐œˆ|(๐‘‹)<โˆž, ๐œˆ+(๐ธ)<โˆž and ๐œˆโˆ’(๐ธ)<โˆž.
Case 1: ๐œˆ+(๐ธ)โ‰ฅ๐œˆโˆ’(๐ธ), then

|๐œˆ+(๐ธ)โˆ’๐œˆโˆ’(๐ธ)|=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)โ‡”๐œˆ+(๐ธ)โˆ’๐œˆโˆ’(๐ธ)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)โ‡”โˆ’๐œˆโˆ’(๐ธ)=๐œˆโˆ’(๐ธ)โ‡”๐œˆโˆ’(๐ธ)=0โ‡”๐ธโŠ‚๐‘ƒ

Case 2: ๐œˆ+(๐ธ)<๐œˆโˆ’(๐ธ), then

|๐œˆ+(๐ธ)โˆ’๐œˆโˆ’(๐ธ)|=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)โ‡”๐œˆโˆ’(๐ธ)โˆ’๐œˆ+(๐ธ)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)โ‡”โˆ’๐œˆ+(๐ธ)=๐œˆ+(๐ธ)โ‡”๐œˆ+(๐ธ)=0โ‡”๐ธโŠ‚๐‘

Therefore the equality condition implies that ๐ธ must be positive or negative for ๐œˆ; and in converse, if ๐ธ is neither positive nor negative set, in either case it implies |๐œˆ(๐ธ)|โ‰ |๐œˆ|(๐ธ), thus when ๐œˆ finite, |๐œˆ(๐ธ)|=|๐œˆ|(๐ธ) iff ๐ธ is positive or negative for ๐œˆ.

โ–ก

Signed integrals

Let ๐œˆ be a signed measure on (๐‘‹,๐’œ๏ธ€).

  • Prove that โˆซ๐‘”๐‘‘|๐œˆ|=โˆซ๐‘”๐‘‘๐œˆ++โˆซ๐‘”๐‘‘๐œˆโˆ’ for ๐‘”โˆˆ๐ฟ+(|๐œˆ|) or ๐‘”โˆˆ๐ฟ1(|๐œˆ|).

  • Define ๐ฟ1(๐œˆ)=๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’). Prove that ๐ฟ1(๐œˆ)=๐ฟ1(|๐œˆ|).

  • Define โˆซ๐‘“๐‘‘๐œˆ=โˆซ๐‘“๐‘‘๐œˆ+โˆ’โˆซ๐‘“๐‘‘๐œˆโˆ’ for ๐‘“โˆˆ๐ฟ1(๐œˆ). Prove that if ๐‘“โˆˆ๐ฟ1(๐œˆ), then

    |โˆซ๐‘“๐‘‘๐œˆ|โ‰คโˆซ|๐‘“|๐‘‘|๐œˆ|
  • Suppose that ๐œˆ is a finite measure (i.e. ๐œˆยฑ(๐‘‹)<โˆž.) Prove that if ๐ธโˆˆ๐’œ๏ธ€, then

    |๐œˆ|(๐ธ)=sup{|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โˆฃโˆฅ๐‘“โˆฅโˆžโ‰ค1}.
Proof

of (i): Take a Hahn decomposition ๐‘‹=๐‘ƒโŠ”๐‘.
Then by Jordan decomposition,

๐œˆ+(๐ธ)=๐œˆ(๐ธโˆฉ๐‘ƒ),๐œˆโˆ’(๐ธ)=โˆ’๐œˆ(๐ธโˆฉ๐‘),โˆ€๐ธโŠ‚๐‘‹

and therefore ๐‘ƒ is null set of ๐œˆโˆ’ and ๐‘ is null set of ๐œˆ+. So on ๐‘ƒ, |๐œˆ|=๐œˆ++๐œˆโˆ’=๐œˆ+; on ๐‘, |๐œˆ|=๐œˆ++๐œˆโˆ’=๐œˆโˆ’ Thus, suppose ๐‘”โˆˆ๐ฟ+(|๐œˆ|),

โˆซ๐‘”๐‘‘|๐œˆ|=โˆซ๐‘‹๐‘”๐‘‘|๐œˆ|=โˆซ๐‘ƒ๐‘”๐‘‘|๐œˆ|+โˆซ๐‘๐‘”๐‘‘|๐œˆ|since ๐‘‹=๐‘ƒโŠ”๐‘=โˆซ๐‘ƒ๐‘”๐‘‘๐œˆ++โˆซ๐‘๐‘”๐‘‘๐œˆโˆ’since |๐œˆ|=๐œˆ+,๐œˆโˆ’ on ๐‘ƒ,๐‘=โˆซ๐‘”๐‘‘๐œˆ++โˆซ๐‘”๐‘‘๐œˆโˆ’since ๐‘,๐‘ƒ is null for ๐œˆ+,๐œˆโˆ’

Suppose ๐‘”โˆˆ๐ฟ1(|๐œˆ|), then

โˆซ๐‘”๐‘‘|๐œˆ|=โˆซ๐‘‹๐‘”๐‘‘|๐œˆ|=โˆซ๐‘‹๐‘”+๐‘‘|๐œˆ|โˆ’โˆซ๐‘‹๐‘”โˆ’๐‘‘|๐œˆ|by def=(โˆซ๐‘ƒ๐‘”+๐‘‘๐œˆ++โˆซ๐‘๐‘”+๐‘‘๐œˆโˆ’)โˆ’(โˆซ๐‘ƒ๐‘”โˆ’๐‘‘๐œˆ++โˆซ๐‘๐‘”โˆ’๐‘‘๐œˆโˆ’)since ๐‘‹=๐‘ƒโŠ”๐‘=(โˆซ๐‘ƒ๐‘”+๐‘‘๐œˆ+โˆ’โˆซ๐‘ƒ๐‘”โˆ’๐‘‘๐œˆ+)+(โˆซ๐‘๐‘”+๐‘‘๐œˆโˆ’โˆ’โˆซ๐‘๐‘”โˆ’๐‘‘๐œˆโˆ’)=โˆซ๐‘ƒ๐‘”๐‘‘๐œˆ++โˆซ๐‘๐‘”๐‘‘๐œˆโˆ’since ๐‘”โˆˆ๐ฟ1(|๐œˆ|)=โˆซ๐‘”๐‘‘๐œˆ++โˆซ๐‘”๐‘‘๐œˆโˆ’since ๐‘,๐‘ƒ is null for ๐œˆ+,๐œˆโˆ’

This finishes the proof.

โ–ก

Proof

of (ii): WTS: ๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’)=๐ฟ1(|๐œˆ|).
(โ‡’): Suppose ๐‘“โˆˆ๐ฟ1(|๐œˆ|), i.e. โˆซ|๐‘“|๐‘‘|๐œˆ|<โˆž.
Let ๐œ™ be arbitrary positive-valued simple function:

๐œ™=โˆ‘๐‘—=1๐‘›๐‘Ž๐‘—๐œ’๐ธ๐‘—

then

โˆซ๐œ™๐‘‘|๐œˆ|=โˆ‘๐‘–=1๐‘›๐‘Ž๐‘—|๐œˆ|(๐ธ๐‘—)

Since ๐œˆโˆ’(๐ธ๐‘—),๐œˆ+(๐ธ๐‘—)โ‰ค๐œˆ+(๐ธ๐‘—)+๐œˆโˆ’(๐ธ๐‘—)=|๐œˆ|(๐ธ๐‘—) for each ๐‘—, we have

โˆซ๐œ™๐‘‘๐œˆ+,โˆซ๐œ™๐‘‘๐œˆโˆ’โ‰คโˆซ๐œ™๐‘‘|๐œˆ|

Since ๐œ™ is arbitrary, we have

โˆซ|๐‘“|๐‘‘๐œˆ+=sup{โˆซ๐œ™๐‘‘๐œˆ+:0โ‰ค๐œ™โ‰ค|๐‘“|,๐œ™ simple}โ‰คsup{โˆซ๐œ™๐‘‘|๐œˆ|:0โ‰ค๐œ™โ‰ค|๐‘“|,๐œ™ simple}=โˆซ|๐‘“|๐‘‘|๐œˆ|

Same for ๐œˆโˆ’. This shows that

โˆซ|๐‘“|๐‘‘๐œˆ+,โˆซ|๐‘“|๐‘‘๐œˆโˆ’โ‰คโˆซ|๐‘“|๐‘‘|๐œˆ|<โˆž

i.e. ๐‘“โˆˆ๐ฟ1(๐œˆ+) and ๐‘“โˆˆ๐ฟ1(๐œˆโˆ’), so ๐‘“โˆˆ๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’).
Thus

๐ฟ1(|๐œˆ|)โŠ‚๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’)

(โ‡): Suppose ๐‘“โˆˆ๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’), i.e.

โˆซ|๐‘“|๐‘‘๐œˆ+<โˆž,โˆซ|๐‘“|๐‘‘๐œˆโˆ’<โˆž

Since |๐‘“| is non-negative and measurable, we have |๐‘“|โˆˆ๐ฟ+(|๐œˆ|). Thus by (i) we have:

โˆซ|๐‘“|๐‘‘|๐œˆ|=โˆซ|๐‘“|๐‘‘๐œˆ++โˆซ|๐‘“|๐‘‘๐œˆโˆ’<โˆž

So ๐‘“โˆˆ๐ฟ1(|๐œˆ|).
This shows that:

๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’)โŠ‚๐ฟ1(|๐œˆ|)

Combining both direction, we finished the proof that:

๐ฟ1(๐œˆ+)โˆฉ๐ฟ1(๐œˆโˆ’)=๐ฟ1(|๐œˆ|)

โ–ก

Proof

of (iii): Suppose ๐‘“โˆˆ๐ฟ1(๐œˆ), then

|โˆซ๐‘“๐‘‘๐œˆ|=|โˆซ๐‘“๐‘‘๐œˆ+โˆ’โˆซ๐‘“๐‘‘๐œˆโˆ’|by defโ‰ค|โˆซ๐‘“๐‘‘๐œˆ+|+|โˆซ๐‘“๐‘‘๐œˆโˆ’|by tri ineqโ‰คโˆซ|๐‘“|๐‘‘๐œˆ++โˆซ|๐‘“|๐‘‘๐œˆโˆ’by property of ๐ฟ1 integration =โˆซ|๐‘“|๐‘‘|๐œˆ| from (i)

Therefore,

|โˆซ๐‘“๐‘‘๐œˆ|โ‰คโˆซ|๐‘“|๐‘‘|๐œˆ|

โ–ก

Proof

of (iv): Suppose that ๐œˆ is a finite measure (i.e. ๐œˆยฑ(๐‘‹)<โˆž), let ๐ธโˆˆ๐’œ๏ธ€.
We denote:

๐‘†โ‰”sup{|โˆซ๐ธ๐‘“๐‘‘๐œˆ||โˆฅ๐‘“โˆฅโˆžโ‰ค1}

First we show ๐‘†โ‰ค|๐œˆ|(๐ธ):
For any bounded measurable ๐‘“ with โˆฅ๐‘“โˆฅโˆžโ‰ค1,

|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โ‰คโˆซ๐ธ|๐‘“|๐‘‘|๐œˆ|by (iii)โ‰คโˆซ๐ธ1๐‘‘|๐œˆ|by linearity of integration=|๐œˆ|(๐ธ)

So by taking the supremum over such ๐‘“, we get:

๐‘†โ‰ค|๐œˆ|(๐ธ)

Next we will show |๐œˆ|(๐ธ)โ‰ค๐‘†:
We take a Hahn decomposition, getting ๐‘‹=๐‘ƒโŠ”๐‘ where

๐œˆ+(๐ต)=๐œˆ(๐‘ƒโˆช๐ต)โ‰ฅ0,๐œˆโˆ’(๐ต)=โˆ’๐œˆ(๐‘ƒโˆช๐ต)โ‰ค0,for all ๐ตโŠ‚๐‘‹

Then

|๐œˆ|(๐ธ)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)=๐œˆ(๐ธโˆฉ๐‘ƒ)โˆ’๐œˆ(๐ธโˆฉ๐‘)

Now define:

๐‘“โ‰”๐œ’๐‘ƒโˆ’๐œ’๐‘

Then ๐‘“ is measurable since ๐‘ƒ,๐‘ are measurable. And โˆฅ๐‘“โˆฅโˆžโ‰ค1 since ๐‘“(๐‘ฅ)โˆˆ{โˆ’1,1}โˆ€๐‘ฅโˆˆ๐‘‹ Compute:

โˆซ๐ธ๐‘“๐‘‘๐œˆ=โˆซ๐ธโˆฉ๐‘ƒ1๐‘‘๐œˆโˆ’โˆซ๐ธโˆฉ๐‘1๐‘‘๐œˆ=๐œˆ(๐ธโˆฉ๐‘ƒ)โˆ’๐œˆ(๐ธโˆฉ๐‘)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)=|๐œˆ|(๐ธ)

Thus

|๐œˆ|(๐ธ)=|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โ‰ค๐‘†

Combining both inequalities, we get:

|๐œˆ|(๐ธ)=๐‘†

โ–ก

finite signed measures on (๐‘‹,๐’œ๏ธ€) ๆ˜ฏไธ€ไธช NVM

Let (๐‘‹,๐’œ๏ธ€) be a measurable space.

  • Let ๐œ†, ๐œ‡ be finite positive measures on (๐‘‹,๐’œ๏ธ€). Let ๐œˆ=๐œ†โˆ’๐œ‡. Prove that

    ๐œˆ+(๐ธ)โ‰ค๐œ†(๐ธ),๐œˆโˆ’(๐ธ)โ‰ค๐œ‡(๐ธ),|๐œˆ|(๐ธ)โ‰ค๐œ†(๐ธ)+๐œ‡(๐ธ)

    for every ๐ธโˆˆ๐’œ๏ธ€.

  • Let ๐œˆ and ๐œ… be finite signed measures on (๐‘‹,๐’œ๏ธ€) (i.e. ๐œˆ(๐ธ),๐œ…(๐ธ)โˆˆโ„ for all ๐ธโˆˆ๐’œ๏ธ€). Show that

    |๐œˆ+๐œ…|(๐ธ)โ‰ค|๐œˆ|(๐ธ)+|๐œ…|(๐ธ)

    for every ๐ธโˆˆ๐’œ๏ธ€.

  • Let โ„ณ๏ธ€ be the collection of finite signed measure ๐œˆ on (๐‘‹,๐’œ๏ธ€). For ๐œˆโˆˆโ„ณ๏ธ€, define

    โˆฅ๐œˆโˆฅ=|๐œˆ|(๐‘‹)

    Prove that โˆฅโ‹…โˆฅ is a norm on โ„ณ๏ธ€ with an appropriate definition of the sum of two signed measures and the multiplication of a signed measure by a (real) scalar.

  • Suppose (๐‘‹,๐’œ๏ธ€)=(โ„,โ„ฌ๏ธ€(โ„)). Compute โˆฅ๐›ฟ๐‘ฅโˆ’๐›ฟ๐‘ฆโˆฅ for ๐‘ฅ,๐‘ฆโˆˆโ„.

Remark: the norm on โ„ณ๏ธ€ is called the the total variation norm.

Proof

of (a):
Recall in problem 2 we get:

๐œˆ+(๐ธ)=sup{๐œˆ(๐น):๐นโŠ‚๐ธ,๐นโˆˆ๐’œ๏ธ€},๐œˆโˆ’(๐ธ)=โˆ’inf{๐œˆ(๐น):๐นโŠ‚๐ธ,๐นโˆˆ๐’œ๏ธ€}

Claim 1: ๐œˆ+(๐ธ)โ‰ค๐œ†(๐ธ).
Let ๐นโŠ‚๐ธ, ๐นโˆˆ๐’œ๏ธ€. Then:

๐œˆ(๐น)=๐œ†(๐น)โˆ’๐œ‡(๐น)โ‰ค๐œ†(๐น)โ‰ค๐œ†(๐ธ)

since ๐นโŠ‚๐ธ and ๐œ† is positive. Taking the sup over all such ๐น, we get

๐œˆ+(๐ธ)=sup๐นโŠ‚๐ธ๐œˆ(๐น)โ‰ค๐œ†(๐ธ)

Claim 2: ๐œˆโˆ’(๐ธ)โ‰ค๐œ‡(๐ธ).
Similarly as Claim 1, for any ๐นโŠ‚๐ธ, since ๐œ† and ๐œ‡ are p.m., we have

๐œˆ(๐น)=๐œ†(๐น)โˆ’๐œ‡(๐น)โ‰ฅโˆ’๐œ‡(๐น)โ‰ฅโˆ’๐œ‡(๐ธ)โŸนโˆ’๐œˆ(๐น)โ‰ค๐œ‡(๐ธ)

Taking the inf over ๐นโŠ‚๐ธ, we get

๐œˆโˆ’(๐ธ)=โˆ’inf๐นโŠ‚๐ธ๐œˆ(๐น)โ‰ค๐œ‡(๐ธ)

Claim 3: |๐œˆ|(๐ธ)โ‰ค๐œ†(๐ธ)+๐œ‡(๐ธ).
This is just combining the two ineqs:

|๐œˆ|(๐ธ)=๐œˆ+(๐ธ)+๐œˆโˆ’(๐ธ)โ‰ค๐œ†(๐ธ)+๐œ‡(๐ธ)

โ–ก

Proof

of (b):
Let ๐ธโˆˆ๐’œ๏ธ€. WTS: |๐œˆ+๐œ…|(๐ธ)โ‰ค|๐œˆ|(๐ธ)+|๐œ…|(๐ธ).
Recall in problem 2 we showed that for a signed measure ๐œŽ and a measurable set ๐ธ , we have:

|๐œŽ|(๐ธ)=sup{โˆ‘๐‘–=1๐‘›|๐œŽ(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}

Let {๐ธ๐‘–}๐‘–=1๐‘› be any finite measurable partition of ๐ธ. Then for each ๐ธ๐‘–:

|(๐œˆ+๐œ…)(๐ธ๐‘–)|=|๐œˆ(๐ธ๐‘–)+๐œ…(๐ธ๐‘–)|โ‰ค|๐œˆ(๐ธ๐‘–)|+|๐œ…(๐ธ๐‘–)|(by tri ineq on โ„)

Summing over the partition, we have:

โˆ‘๐‘–=1๐‘›|(๐œˆ+๐œ…)(๐ธ๐‘–)|โ‰คโˆ‘๐‘–=1๐‘›|๐œˆ(๐ธ๐‘–)|+โˆ‘๐‘–=1๐‘›|๐œ…(๐ธ๐‘–)|

Now take the supremum over all such partitions of ๐ธ:

|๐œˆ+๐œ…|(๐ธ)=sup{โˆ‘๐‘–=1๐‘›|(๐œˆ+๐œ…)(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}โ‰คsup{โˆ‘๐‘–=1๐‘›|๐œˆ(๐ธ๐‘–)|+โˆ‘๐‘–=1๐‘›|๐œ…(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}โ‰คsup{โˆ‘๐‘–=1๐‘›|๐œˆ(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}+sup{โˆ‘๐‘–=1๐‘›|๐œ…(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}=|๐œˆ|(๐ธ)+|๐œ…|(๐ธ)

Since measurable ๐ธ is arbitrary, this finishes the proof.

โ–ก

Proof

of (c):

โ„ณ๏ธ€:={all finite signed measures on (๐‘‹,๐’œ๏ธ€)}

and for ๐œˆโˆˆโ„ณ๏ธ€, we define:

โˆฅ๐œˆโˆฅโ‰”|๐œˆ|(๐‘‹)

WTS: โˆฅโ‹…โˆฅ is a norm on โ„ณ๏ธ€.

  1. Positive Definiteness:
    Let ๐œˆโˆˆโ„ณ๏ธ€. Since |๐œˆ| is a positive measure, โˆฅ๐œˆโˆฅ=|๐œˆ|(๐‘‹)โ‰ฅ0.
    Since |๐œˆ| is a positive measure, โˆฅ๐œˆโˆฅ=|๐œˆ|(๐‘‹)โ‰ฅ0.
    Suppose |๐œˆ|(๐‘‹)=0, then ๐‘‹ is a |๐œˆ|-null set, so |๐œˆ|(๐ธ)=0 for all ๐ธโˆˆ๐’œ๏ธ€. Thus ๐œˆ=0.
    And suppose ๐œˆ=0, then |๐œˆ|=0 also, so |๐œˆ|(๐‘‹)=0.
    Thus, โˆฅ๐œˆโˆฅ=0 iff ๐œˆ=0. This finishes the proof of positive definiteness.

  2. Absolute Homogeneity:
    Since for any measurable set ๐ธ:

    |๐‘Ž๐œˆ|(๐ธ)=sup{โˆ‘๐‘–=1๐‘›|(๐‘Ž๐œˆ)(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}=sup{โˆ‘๐‘–=1๐‘›|๐‘Ž||๐œˆ(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}=|๐‘Ž|sup{โˆ‘๐‘–=1๐‘›|๐œˆ(๐ธ๐‘–)|:๐ธ=โจ†๐‘–=1๐‘๐ธ๐‘–}=|๐‘Ž|โ‹…|๐œˆ|(๐ธ)

    We have:

    โˆฅ๐‘Ž๐œˆโˆฅ=|๐‘Ž๐œˆ|(๐‘‹)=|๐‘Ž|โ‹…|๐œˆ|(๐‘‹)=|๐‘Ž|โ‹…โˆฅ๐œˆโˆฅ

    finishing the proof of absolute homogeneity.

  3. Triangle Inequality:
    Recall we just proved in (b) that for any measurable ๐ธ:

    |๐œˆ+๐œ…|(๐ธ)โ‰ค|๐œˆ|(๐ธ)+|๐œ…|(๐ธ)

    Thus

    โˆฅ๐œˆ+๐œ…โˆฅ=|๐œˆ+๐œ…|(๐‘‹)โ‰ค|๐œˆ|(๐‘‹)+|๐œ…|(๐‘‹)=โˆฅ๐œˆโˆฅ+โˆฅ๐œ…โˆฅ

    finishing the proof of triangle inequality.

So we can conclude that โˆฅ๐œˆโˆฅโ‰”|๐œˆ|(๐‘‹) defines a norm on โ„ณ๏ธ€, with the standard definitions of addition and scalar multiplication of signed measures.

โ–ก

Proof

of (d)
Suppose (๐‘‹,๐’œ๏ธ€)=(โ„,โ„ฌ๏ธ€(โ„)). Compute โˆฅ๐›ฟ๐‘ฅโˆ’๐›ฟ๐‘ฆโˆฅ for ๐‘ฅ,๐‘ฆโˆˆโ„.

Recall def: For any Borel set ๐ดโŠ‚โ„,

๐›ฟ๐‘ฅ(๐ด)={1if ๐‘ฅโˆˆ๐ด0otherwise

So we define the signed measure ๐œˆโ‰”๐›ฟ๐‘ฅโˆ’๐›ฟ๐‘ฆ as:

๐œˆ(๐ด)=๐›ฟ๐‘ฅ(๐ด)โˆ’๐›ฟ๐‘ฆ(๐ด)

If ๐‘ฅ=๐‘ฆ, then ๐›ฟ๐‘ฅ=๐›ฟ๐‘ฆ, then ๐œˆ=0, so โˆฅ๐œˆโˆฅ=0. This is the trivial case. if ๐‘ฅโ‰ ๐‘ฆ: We first compute the Jordan decomposition.
We know that ๐œˆ+(๐ธ)=sup{๐œˆ(๐น)โˆฃ:๐นโˆˆ๐’œ๏ธ€,๐นโŠ‚๐ธ}, and ๐œˆโˆ’(๐ธ)=โˆ’inf{๐œˆ(๐น)โˆฃ๐นโˆˆ๐’œ๏ธ€,๐นโŠ‚๐ธ}. For any ๐ธโˆ‹๐‘ฅ, we have

๐œˆ+(๐ธ)=๐œˆ({๐‘ฅ})=1

In other cases, we have:

๐œˆ+(๐ธ)=๐œˆ(๐ธ\{๐‘ฆ})=0

For any ๐ธโˆ‹๐‘ฆ, we have

๐œˆโˆ’(๐‘ฆ)=โˆ’๐œˆ({๐‘ฆ})=1

In other cases, we have:

๐œˆโˆ’(๐ธ)=โˆ’๐œˆ(๐ธ\{๐‘ฅ})=0

And we thus discover that:

๐œˆ+=๐›ฟ๐‘ฅ,๐œˆโˆ’=๐›ฟ๐‘ฆ

So

โˆฅ๐œˆโˆฅ=|๐œˆ|(โ„)=๐›ฟ๐‘ฅ(โ„)+๐›ฟ๐‘ฆ(โ„)=1+1=2

Thus we can conclude that

โˆฅ๐œˆโˆฅ={2if ๐‘ฅโ‰ ๐‘ฆ0otherwise

โ–ก

and more: finite signed measures on (๐‘‹,๐’œ๏ธ€) ็ป„ๆˆไธ€ไธช real Banach space

Prove that the normed vector space โ„ณ๏ธ€ in the previous problem is in fact a Banach space.

Proof

In problem 4 we have shown that on (โ„ณ๏ธ€,โˆฅโ‹…โˆฅ) is a normed vector space, where

โ„ณ๏ธ€:={all finite signed measures on (๐‘‹,๐’œ๏ธ€)}

and

โˆฅ๐œˆโˆฅโ‰”|๐œˆ|(๐‘‹)

Now we prove that the NVM (โ„ณ๏ธ€,โˆฅโ‹…โˆฅ) is complete, i.e. it is a Banach space.
Let (๐œˆ๐‘›) be a Cauchy sequence in โ„ณ๏ธ€. We have

|๐œˆ๐‘›(๐ต)โˆ’๐œˆ๐‘š(๐ต)|=|(๐œˆ๐‘›โˆ’๐œˆ๐‘š)(๐ต)|โ‰คโˆฅ๐œˆ๐‘›โˆ’๐œˆ๐‘šโˆฅfor all ๐ตโˆˆ๐’œ๏ธ€

In particular, (๐œˆ๐‘›(๐ต))๐‘› is a Cauchy sequence for all ๐ตโˆˆ๐’œ๏ธ€. For each ๐ตโˆˆ๐’œ๏ธ€, this is a Cauchy seq in โ„, thus converges. So we can get:

๐œˆ(๐ต)โ‰”lim๐‘›๐œˆ๐‘›(๐ต)

as the pointwise limit (by a point we mean a set).
Claim 1: ๐œˆโˆˆโ„ณ๏ธ€.
Since for all ๐‘›, ๐œˆ๐‘›(โŒ€)=0, we have:

๐œˆ(โŒ€)โ‰”lim๐‘›๐œˆ๐‘›(โŒ€)=0

For a countable disjoint union of measurable sets ๐ธ=โจ†๐‘–=1โˆž๐ธ๐‘–,

lim๐‘›๐œˆ๐‘›(๐ธ)=lim๐‘›โˆ‘๐‘–๐œˆ๐‘›(๐ธ๐‘–)

is the limit of a finite sum of numerical sequences in โ„. So we can exchange the order of taking limit and sum. Then we get:

๐œˆ(๐ธ)=lim๐‘›๐œˆ๐‘›(๐ธ)=lim๐‘›โˆ‘๐‘–๐œˆ๐‘›(๐ธ๐‘–)=โˆ‘๐‘–lim๐‘›๐œˆ๐‘›(๐ธ๐‘–)=โˆ‘๐‘–๐œˆ(๐ธ๐‘–)

And notice, for each measurable set ๐ตโˆˆ๐’œ๏ธ€, since (๐œˆ๐‘›(๐ต))๐‘› is a Cauchy sequence in โ„, it is bounded, thus does not admit โˆž,โˆ’โˆž values. verifying that ๐œˆ is a valid signed measure.
Also, this means that taking Hahn Decomposition ๐‘‹=๐‘ƒโŠ”๐‘ by ๐œˆ, we have

๐œˆ+(๐‘‹)=๐œˆ(๐‘ƒ),๐œˆโˆ’(๐‘‹)=โˆ’๐œˆ(๐‘)

Since ๐œˆ(๐‘ƒ),๐œˆ(๐‘) are bounded, we have: Thus

|๐œˆ|(๐‘‹)=๐œˆ+(๐‘‹)+๐œˆโˆ’(๐‘‹)<โˆž

This verifies that ๐œˆ is a finite s.m.
Claim 2: ๐œˆ๐‘›โ†’๐œˆ in โˆฅโ‹…โˆฅ. Fix ๐œ€>0. There exists ๐‘ such that โˆฅ๐œˆ๐‘›โˆ’๐œˆ๐‘šโˆฅ<๐œ€/2 for all ๐‘š,๐‘›โ‰ฅ๐‘. Thus for all ๐‘›โ‰ฅ๐‘ we have:

|(๐œˆ๐‘›โˆ’๐œˆ)(๐ต)|=lim๐‘š|(๐œˆ๐‘›โˆ’๐œˆ๐‘š)(๐ต)|โ‰ค๐œ€/2,โˆ€๐ตโˆˆ๐’œ๏ธ€,โˆ€๐‘›โ‰ฅ๐‘

Notice that

๐œˆ+(๐ต)=sup{๐œˆ(๐ถ)โˆฃ๐ถโˆˆ๐’œ๏ธ€,๐ถโŠ‚๐ต}

and

๐œˆโˆ’(๐ต)=โˆ’inf{๐œˆ(๐ถ)โˆฃ๐ถโˆˆ๐’œ๏ธ€,๐ถโŠ‚๐ต}=sup{โˆ’๐œˆ(๐ถ)โˆฃ๐ถโˆˆ๐’œ๏ธ€,๐ถโŠ‚๐ต}

It follows that

(๐œˆ๐‘›โˆ’๐œˆ)+(๐‘‹)=sup{(๐œˆ๐‘›โˆ’๐œˆ)(๐ต)โˆฃ๐ตโˆˆ๐’œ๏ธ€}โ‰ค๐œ€/2,โˆ€๐‘›โ‰ฅ๐‘

Similarly,

(๐œˆ๐‘›โˆ’๐œˆ)โˆ’(๐‘‹)=sup{โˆ’(๐œˆ๐‘›โˆ’๐œˆ)(๐ต)โˆฃ๐ตโˆˆ๐’œ๏ธ€}โ‰ค๐œ€/2,โˆ€๐‘›โ‰ฅ๐‘

Thus

|๐œˆ๐‘›โˆ’๐œˆ|(๐‘‹)=(๐œˆ๐‘›โˆ’๐œˆ)+(๐‘‹)+(๐œˆ๐‘›โˆ’๐œˆ)โˆ’(๐‘‹)โ‰ค๐œ€

This holds for all ๐‘›โ‰ฅ๐‘. And since ๐œ€>0 is arbitrary, this proves that

lim๐‘›โ†’โˆžโˆฅ๐œˆ๐‘›โˆ’๐œˆโˆฅ=0

As a result, ๐œˆ๐‘›โ†’๐œˆ in โˆฅโ‹…โˆฅ, completeing the proof.

โ–ก

Nur fรผr Verrรผckte

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!) Does there exist a signed Borel measure ๐œˆ on โ„ with the property that for every ๐›ผโˆˆโ„ there exists a Borel set ๐ธโŠ‚โ„ with ๐œˆ(๐ธ)=๐›ผ.

11 Radon-Nikodym theorem

11.1 Radon-Nikodym Theorem [Fol 3.2]

ไปฅไธ‹ๆ˜ฏไธคไธช instructive ็š„ questions:
Question 1: Given ไธ‰ไธช s.m. on

๐œ‡=๐‘š+โˆ‘๐‘—=1โˆž๐‘๐‘—๐›ฟ๐‘ฅ๐‘—+๐œ‡๐ถ๐‘Ž๐‘›๐‘ก๐‘œ๐‘Ÿ

on (โ„,โ„ฌ๏ธ€(โ„)), ๆˆ‘ไปฌๅฏๅฆไปŽ ๐œ‡ ไธญ recover ๅ…ถไธญไธ€ไธช measure, without ๅฆๅค–ไธคไธช measure?

๐œ‡๐ถ๐‘Ž๐‘›๐‘ก๐‘œ๐‘Ÿ=(?)๐œ‡

Question 2: ็ป™ๅฎšไธ€ไธชไปปๆ„็š„ p.m. ๐œ‡, ไปฅๅŠไธ€ไธชไปปๆ„็š„ s.m. ๐œˆ on (๐‘‹,๐’œ๏ธ€),
ๅฆ‚ไฝ•ๅˆคๆ–ญๆ˜ฏๅฆๅญ˜ๅœจไธ€ไธช ๐‘“โˆˆ๐ฟ1(๐œ‡), ไฝฟๅพ—

๐œˆ(๐ธ)=โˆซ๐ธ๐‘“๐‘‘๐œ‡

ไปฅๅŠ, ๅฆ‚ๆžœๅญ˜ๅœจ, ๅฆ‚ไฝ•ๆ‰พๅˆฐ่ฟ™ๆ ท็š„ไธ€ไธช ๐‘“?

11.1.1 absolutely continuous: ๐œˆโ‰ช๐œ‡

Definition 11.57 : absolute continuity of signed measures

็ป™ๅฎš p.m. ๐œ‡ ๅ’Œ signed measure ๐œˆ on (๐‘‹,๐’œ๏ธ€), ๆˆ‘ไปฌ็งฐ ๐œˆ is absolutely continuous w.r.t. ๐œ‡, ๅฆ‚ๆžœ

โˆ€๐ธโˆˆ๐’œ๏ธ€,๐œ‡(๐ธ)=0โŸน๐œˆ(๐ธ)=0

ๅณ: ๐œˆ ็š„ null sets ๅŒ…ๅซไบ† ๐œ‡ ็š„ๆ‰€ๆœ‰ null sets. (๐œˆ ๆ‹ฅๆœ‰ๆฏ” ๐œ‡ ไธฅๆ ผๆ›ดๅคš็š„ null sets)
ๅ†™ไฝœ

๐œˆโ‰ช๐œ‡
Figureย 34: mutually singular and absolutely continuous

mutually singular ็š„่ฎฐๅท ๐œˆโŠฅ๐œ‡ ่กจ็คบ็š„ๆ˜ฏ ๐œˆ ๅ’Œ ๐œ‡ ๅ‡บ็Žฐๅ˜ๅŒ–็š„ๅŒบๅŸŸๅฎŒๅ…จไธๅŒ, ่€Œ ๐œˆโ‰ช๐œ‡ ่กจ็คบ็š„ๆ˜ฏ ๐œˆ ๅ‡บ็Žฐๅ˜ๅŒ–็š„ๅŒบๅŸŸๅฎŒๅ…จๅŒ…ๆ‹ฌๅœจ ๐œ‡ ๅ‡บ็Žฐๅ˜ๅŒ–็š„ๅŒบๅŸŸ้‡Œ (ๅ› ไธบ ๐œ‡ ไธๅ˜ๅŒ–็š„ๅŒบๅŸŸ่ขซๅŒ…ๆ‹ฌๅœจ ๐œˆ ไธๅ˜ๅŒ–็š„ๅŒบๅŸŸ้‡Œ).

Example 11.34

๐‘“โˆˆ๐ฟ1(๐œ‡), ๐œˆ(๐ธ):=โˆซ๐ธ๐‘“๐‘‘๐œ‡, ็”ฑ็งฏๅˆ†ๅฎšไน‰ๅ‡บ็š„ s.m., ๆ€ปๆ˜ฏๆปก่ถณ

๐œˆโ‰ช๐œ‡
Example 11.35
๐œˆ1:=๐‘š,๐œˆ2โ‰”โˆ‘๐‘—=1โˆž๐‘๐‘—๐›ฟ๐‘ฅ๐‘—,๐œˆ3:=๐œ‡๐ถ๐‘Ž๐‘›๐‘ก๐‘œ๐‘Ÿ

่ฟ™ไธ‰ไธช measure ๆœ‰

๐œˆ๐‘–โ‰ชฬธ๐œˆ๐‘—โˆ€๐‘–โ‰ ๐‘—

ๅฎƒไปฌๆ˜ฏ mutually singular ็š„. ๅฏนไบŽๅ…ถไธญไปปๆ„ไธคไธช ๐œˆ๐‘–,๐œˆ๐‘—, ๆœฌ่บซๅทฒ็ปๅญ˜ๅœจไธ€ไธชๅˆ’ๅˆ†ไฝฟๅพ— ๐œˆ๐‘– ๅœจ ๐ธ ไธŠๆ˜ฏ null ็š„่€Œ ๐œˆ๐‘— ๅœจ ๐ธ๐‘ ไธŠๆ˜ฏ null ็š„. ้‚ฃไนˆๅฆ‚ๆžœ ๐œˆ๐‘–โ‰ช๐œˆ๐‘—, ๅˆ™่ฏดๆ˜Ž ๐œˆ๐‘– ๅœจ ๐ธ๐‘ ไธŠไนŸๆ˜ฏ null ็š„, ้‚ฃไนˆ ๐œˆ๐‘– ๅœจๆ•ดไธช ๐‘‹ ไธŠ้ƒฝๆ˜ฏ null ็š„, ่ฏดๆ˜Ž ๐œˆ๐‘– ๆ˜ฏไธ€ไธช trivial measure.
ๆ˜พ็„ถ, ่ฟ™้‡Œไธ‰ไธช measure ้ƒฝไธๆ˜ฏ trivial measure, ๅ› ่€Œๅฎƒไปฌไน‹้—ดๆฒกๆœ‰ abs ctn ็š„ๅ…ณ็ณป.

Proposition 11.28 : absolutely continuous ็š„ๆ€ง่ดจ
  • ๅฏน total variation measure,

    |๐œˆ|โ‰ช๐œ‡โ‡”๐œˆ+โ‰ช๐œ‡ and ๐œˆโˆ’โ‰ช๐œ‡

    (ๅฎนๆ˜“่ฏๆ˜Ž)

  • ๐œˆโŸ‚๐œ‡ and ๐œˆโ‰ช๐œ‡โŸน๐œˆ=0

    (ๅˆšๆ‰ๅทฒ็ป่ฏๆ˜Ž)

ๆˆ‘ไปฌๅฏไปฅๆŠŠ absolutely ctn ็š„ๆฆ‚ๅฟตไปŽไธ€ไธช s.m. wrt ไธ€ไธช p.m. ๆ‰ฉๅฑ•ๅˆฐไธ€ไธช s.m. wrt ไธ€ไธช s.m., by taking ๅŽ้ข่ฟ™ไธช s.m. ็š„ total variation measure:

say ๐œˆโ‰ช๐œ‡, if ๐œˆโ‰ช|๐œ‡|

ไฝ†ๆ˜ฏ Folland ่กจ็คบๆˆ‘ไปฌไน‹ๅŽๅนถไธ้œ€่ฆ็”จๅˆฐ่ฟ™ไธชๆ›ด general ็š„ๅฎšไน‰. ๆ‰€ไปฅไธ็”จๅœจๆ„ๅฎƒ.

11.1.2 ๐œˆโ‰ช๐œ‡ ็š„็ญ‰ไปทๆกไปถ

question: ไธบไป€ไนˆ่ฟ™ไธชๅฎšไน‰่ฆๅซๅš absolutely continuous, ๅฎƒๅ’Œ continuous ่ฟ™ไธช่ฏๅˆฐๅบ•ๆœ‰ไป€ไนˆๅ…ณ็ณป. ไธ‹้ข่ฟ™ไธช theorem ่ฏดๆ˜Žไบ†่ฟ™ไธ€็‚น.

Theorem 11.58 : why it is called "absolutely continuous"

ไปค ๐œˆ ไธบไธ€ไธช finite s.m., ๐œ‡ ไธบไธ€ไธช p.m. on (๐‘‹,๐’œ๏ธ€).
Claim:

๐œˆโ‰ช๐œ‡โ‡”โˆ€๐œ–>0,โˆƒ๐›ฟ>0s.t.|๐œˆ(๐ธ)|<๐œ– whenever ๐œ‡(๐ธ)<๐›ฟ
Proof

(i) to (ii): ๆˆ‘ไปฌไฝฟ็”จๅ่ฏ, ๅˆฉ็”จ limsup.
Assume (i), ๅนถ suppose for contradiction that (ii) ไธๆˆ็ซ‹.
้‚ฃไนˆๅญ˜ๅœจ ๐œ–>0 s.t. ๅฏนไบŽไปปๆ„ ๐‘›โˆˆโ„•, ้ƒฝๅญ˜ๅœจไธ€ไธช seq ๐ธ๐‘›โˆˆ๐’œ๏ธ€ s.t. ๐œ‡(๐ธ๐‘›)โ‰ค12๐‘›, ๐œˆ(๐ธ๐‘›)โ‰ฅ๐œ– for each ๐‘›.
Set

๐ธ:=limโ€‰sup๐‘›๐ธ๐‘›=โ‹‚๐‘›=1โˆžโ‹ƒ๐‘˜=๐‘›โˆž๐ธ๐‘˜

ๆˆ‘ไปฌๆ ‡่ฎฐๅŽ้ข็š„ๆฏไธช้›†ๅˆไธบ:

๐น๐‘›:=โ‹ƒ๐‘˜=๐‘›โˆž๐ธ๐‘˜

ไบŽๆ˜ฏ

๐œ‡(๐น๐‘›)โ‰คโˆ‘๐‘˜=๐‘›โˆž12๐‘˜=12๐‘›

ไปŽ่€Œๅพ—ๅˆฐ,

๐œ‡(๐ธ)=0

่€Œ็”ฑไบŽ ๐œˆ(๐น๐‘›)โ‰ฅ๐œ– for each ๐‘›, we have

๐œˆ(๐ธ)โ‰ฅ๐œ–

่ฟ™ไธŽ ๐œˆโ‰ช๐œ‡ contradict. ไปŽ่€Œๅพ—่ฏ: ๐œˆโ‰ช๐œ‡โŸน๐›ฟ-๐œ– argument.
่€Œ ๐›ฟ-๐œ– argument โŸน ๐œˆโ‰ช๐œ‡ ๆ˜ฏ trivial ็š„.

โ–ก

11.1.3 RN derivative and RN Thm

11.1.4 RN derivative: (if exist) express how ๐œˆ can be induced from ๐œ‡

Definition 11.58 : Radon-Nikodym derivative

ๅฏนไบŽ {p.m. ๐œ‡s.m. ๐œˆ on (๐‘‹,๐’œ๏ธ€), ๅฆ‚ๆžœๅญ˜ๅœจไธ€ไธช ๐’œ๏ธ€-measurable ๐‘“, ไฝฟๅพ— ๐œˆ ไธบ the signed measure ๐œˆ induced by ๐œ‡ and ๐‘“:

๐œˆ(๐ธ)=โˆซ๐ธ๐‘“๐‘‘๐œ‡,โˆ€๐ธโˆˆ๐’œ๏ธ€

ๅˆ™็งฐ ๐‘“ is the Radon-Nikodym Derivative of ๐œˆ w.r.t. ๐œ‡. ๅ†™ไฝœ

๐‘“=๐‘‘๐œˆ๐‘‘๐œ‡

ๆˆ–่€…

๐‘‘๐œˆ=๐‘“๐‘‘๐œ‡

Radon-Nikodym derivative ๐‘“ ๅˆป็”ป็š„ๆ˜ฏๅœจๆฏไธ€็‚น ๐‘ฅโˆˆ๐‘‹ ไธŠ, signed ๆต‹ๅบฆ ๐œˆ ็›ธๅฏนไบŽๆต‹ๅบฆ ๐œ‡ ็š„ๅ˜ๅŒ–้€Ÿ็އ.
We sometimes call ๐œˆ the signed measure ๐‘“๐‘‘๐œ‡.

Example 11.36

ๅ– LS measure ๐œ‡๐น on (โ„,โ„ฌ๏ธ€(โ„)), with ๐น=๐‘’2๐‘ฅ.
้‚ฃไนˆ:

๐œ‡๐น((๐‘Ž,๐‘))=๐‘’2๐‘โˆ’๐‘’2๐‘Ž=โˆซ๐‘Ž๐‘2๐‘’2๐‘ฅ๐‘‘๐‘ฅ

ๆˆ‘ไปฌๅฏไปฅ check:

๐œ‡๐น(๐ธ)=โˆซ๐ธ2๐‘’2๐‘ฅ๐‘‘๐‘ฅ,โˆ€๐ธโˆˆโ„ฌ๏ธ€(โ„)

ๅ› ่€Œ

๐‘‘๐œ‡๐น๐‘‘๐‘š=2๐‘’2๐‘ฅ=๐นโ€ฒ(๐‘ฅ)
Proposition 11.29

ไปปๅ– measure ๐œ‡, ไปฅๅŠ extended ๐œ‡-integrable function ๐‘“, ้‚ฃไนˆthe signed measure ๐œˆ induced by ๐œ‡ and ๐‘“ ๅณ ๐œˆ(๐ธ):=โˆซ๐ธ๐‘“๐‘‘๐œ‡ ไธ€ๅฎšๆœ‰:

๐œˆโ‰ช๐œ‡
Proof

trivial.

โ–ก

Question: ๆˆ‘ไปฌๅฆ‚ไฝ•ๅˆคๆ–ญ่ฟ™ไธช RN derivative ๆ˜ฏๅฆๅญ˜ๅœจๅ‘ข? Radon Nikodym Theorem ๆญฃๆ˜ฏ่ฟ™ไธช้—ฎ้ข˜็š„็ญ”ๆกˆ.

11.1.5 RN Thm: ๐œŽ-finite ๐œˆโ‰ช๐œ‡โ‡” ๅญ˜ๅœจ RN derivative

Theorem 11.59 : Radon-Nikodym Theorem

ๅฏนไบŽ ๐œŽ-finite measure {p.m. ๐œ‡s.m. ๐œˆ on (๐‘‹,๐’œ๏ธ€),

๐œˆโ‰ช๐œ‡โ‡”โˆƒ ext. ๐œ‡-intble ๐‘“=๐‘‘๐œˆ๐‘‘๐œ‡

ๅนถไธ”่ฟ™ไธช RN derivative ๐‘“ ๆ˜ฏ unique ็š„, in ๐œ‡-a.e. sense. (ๅณๅœจ ๐œ‡ ็š„ไธ€ไธช null set ไน‹ๅค–ๅ”ฏไธ€).

Radon Nikodym Theorem ่กจ็คบ, ๅฏนไบŽ ๐œŽ-finite ็š„ ๐œˆ ๅ’Œ ๐œ‡, RN derivative ๅญ˜ๅœจ(ๅนถไธ”ไธ€ๅฎšๅ”ฏไธ€)ๅฝ“ไธ”ไป…ๅฝ“ ๐œˆโ‰ช๐œ‡. ๅณๅฏนไบŽไปปๆ„ไธคไธช abs ctn ็š„ measure, ๅช่ฆๅฎƒไปฌ ๐œŽ-finite, ๅฐฑๅฏไปฅ็”จไธ€ไธชๅ…ทไฝ“็š„ๅ‡ฝๆ•ฐ ๐‘“ ๆฅ่กจ่พพๅฎƒไปฌไน‹้—ด็š„ๅ…ณ็ณป.
่ฆ่ฏๆ˜Ž RN Theorem, ๆˆ‘ไปฌ่ฟ˜้œ€่ฆไธ€ไบ› Lemma.

Lemma 11.40

ๅฆ‚ๆžœ ๐œˆ,๐œ‡ ้ƒฝๆ˜ฏ finite positive measure on (๐‘‹,๐’œ๏ธ€) ๅนถไธ” ๐œ‡โŸ‚ฬธ๐œˆ, ้‚ฃไนˆไธ€ๅฎšๅญ˜ๅœจ ๐œ–>0 ไปฅๅŠ ๐ธโˆˆ๐’œ๏ธ€ with ๐œ‡(๐ธ)>0 s.t.

๐œˆโ‰ฅ๐œ–๐œ‡on ๐ธ
Proof

We look at ๐œˆโˆ’1๐‘›๐œ‡ for each ๐‘›โˆˆโ„•. ๅฎƒไปฌ้ƒฝๆ˜ฏ finite signed measure for sure.
่€ƒ่™‘ Hahn Decomposition Theorem ็ป™ๅ‡บ็š„ ๐‘ƒ๐‘›โŠ”๐‘๐‘› for each ๐‘›. ๅนถ set:

๐‘ƒ:=โ‹ƒ๐‘›๐‘ƒ๐‘›,๐‘:=โ‹‚๐‘›๐‘๐‘›=๐‘ƒ๐‘

ไบŽๆ˜ฏ: ๐‘ ๅฏนไบŽไปปๆ„ ๐‘›, ้ƒฝๆ˜ฏ ๐œˆโˆ’1๐‘›๐œ‡ ็š„ negative set.
่ฟ™่ฏดๆ˜Ž:

โˆ€๐‘›,0โ‰ค๐œˆ(๐‘)โ‰ค1๐‘›๐œ‡(๐‘)

ๅ› ่€Œไธ€ๅฎšๆœ‰:

๐œˆ(๐‘)=0

(่ฟ™ๆ˜ฏๆ˜พ็„ถ็š„, ๅ› ไธบ ๐‘ intersect ไบ†ๆ‰€ๆœ‰็š„ ๐œˆโˆ’1๐‘›๐œ‡ ็š„่ดŸ้›†, ๅœจ ๐‘› ๅคง็š„ๆ—ถๅ€™่ฟ™ไธช diff measure ๅŸบๆœฌ็ญ‰ไบŽ ๐œˆ, ่€Œ ๐œˆ ๆœฌ่บซๆ˜ฏ positive ็š„, ้‚ฃไนˆๆ˜พ็„ถ ๐œˆ(๐‘)=0.)
Case 1: ๅฆ‚ๆžœ ๐œ‡(๐‘ƒ)=0, ้‚ฃไนˆ ๐œ‡โŠฅ๐œˆ.
Case 2: Otherwise then ๅญ˜ๅœจๆŸไธช ๐œ‡(๐‘ƒ๐‘›)>0, ่ฏดๆ˜Ž ๐‘ƒ๐‘› ๆ˜ฏ ๐œˆโˆ’1๐‘›๐œ‡ ็š„ positive set, ๅ› ่€Œๅœจ ๐‘ƒ๐‘› ไธŠ, ๐œˆโ‰ฅ1๐‘›๐œ‡.

โ–ก

่ฟ™ไธช Lemma ่กจๆ˜Ž, ๅฏนไบŽไธคไธช positive measures, ๅฎƒไปฌ่ฆไนˆ mutually singular, ่ฆไนˆไธ€ๅฎšๅญ˜ๅœจๆŸไธช nontrivial ็š„้›†ๅˆไธŠ, ไธ€ไธช่ƒฝๅคŸไปฅไธ€ๅฎšๆฏ”ไพ‹ bound ๅฆๅค–ไธ€ไธช.
่ฟ™ๆ˜ฏๅ› ไธบ, ๅช่ฆ่ฟ™ไธคไธช positive measures ไธๆ˜ฏ mutually singular ็š„ (่ฏดๆ˜Žๅฎƒไปฌๆœ‰ๅ…ฑๅŒ็š„ๅญ˜ๅœจๅ˜ๅŒ–็š„ๅŒบๅŸŸ), ้‚ฃไนˆ note that positive measure ้š็€้›†ๅˆๅขžๅคงไธ€ๅฎšๆ˜ฏๅขžๅคง็š„, ๅ› ่€Œ็›ด่ง‰ไธŠ่‚ฏๅฎšๅญ˜ๅœจๆŸไธชๅญ้›†, ไฝฟๅพ—ๅ…ถไธŠ, ๅฎƒไปฌๅ…ถไธญไธ€ไธช่ƒฝๅคŸไปฅไธ€ๅฎšๆฏ”ไพ‹ bound ๅฆๅค–ไธ€ไธช.
็Žฐๅœจๆˆ‘ไปฌ่ฏๆ˜Ž RN Thm:

Proof

of RN Thm:
Step 1: ้ฆ–ๅ…ˆ็กฎ่ฎค uniqueness, if exist.
้ฆ–ๅ…ˆๆˆ‘ไปฌ assume ๐œˆ,๐œ‡ ้ƒฝๆ˜ฏ finite p.m.
ๆˆ‘ไปฌๅ…ˆ verity uniqueness: ๅ‡่ฎพ

๐‘‘๐œˆ=๐‘“1๐‘‘๐œ‡=๐‘“2๐‘‘๐œ‡,๐‘“๐‘– ext. ๐œ‡-intble

้‚ฃไนˆไปค ๐‘”:=๐‘“1โˆ’๐‘“2, ๆœ‰

โˆซ๐ธ๐‘”๐‘‘๐œ‡=0โˆ€๐ธโˆˆ๐’œ๏ธ€

ๆ‰€ไปฅ ๐‘”=0 a.e.
This shows the uniqueness.
็„ถๅŽๆˆ‘ไปฌ verity existence:
ๆˆ‘ไปฌ่€ƒ่™‘

โ„ฑ๏ธ€:={๐‘“โˆˆ๐ฟ+(๐œ‡):โˆซ๐ธ๐‘“๐‘‘๐œ‡โ‰ค๐œˆ(๐ธ),โˆ€๐ธโˆˆ๐’œ๏ธ€}

We can define partial order on โ„ฑ๏ธ€: ็งฐ ๐‘“1โ‰ค๐‘“2 if ๐‘“1(๐‘ฅ)โ‰ฅ๐‘“2(๐‘ฅ) for a.e. ๐‘ฅ.
ๆ˜พ็„ถ ๐‘“=0 ๆ˜ฏ โ„ฑ๏ธ€ ไธญๆœ€ๅฐ็š„ๅ…ƒ็ด . Idea: ๆˆ‘ไปฌๆƒณ่ฆๅพ—ๅˆฐ โ„ฑ๏ธ€ ไธญๆœ€ๅคง็š„ๅ…ƒ็ด  ๐‘“๐‘š๐‘Ž๐‘ฅ, ็œ‹็œ‹ๆ˜ฏๅฆ่ƒฝๅ–ๅˆฐๆ€ปๆ˜ฏๆœ‰

โˆซ๐ธ๐‘“๐‘š๐‘Ž๐‘ฅ๐‘‘๐œ‡=๐œˆ(๐ธ)

Step 2: Claim ๐‘“1,๐‘“2โˆˆโ„ฑ๏ธ€โŸน๐‘“โ‰”max{๐‘“1,๐‘“2}โˆˆโ„ฑ๏ธ€
Proof of Claim: for fixed ๐‘“1,๐‘“2, ่€ƒ่™‘ ๐ด:={๐‘“1>๐‘“2}. ไปปๅ– ๐ธโˆˆ๐’œ๏ธ€, ๆœ‰:

โˆซ๐ธ๐‘“๐‘‘๐œ‡=โˆซ๐ธโˆฉ๐ด๐‘“1๐‘‘๐œ‡+โˆซ๐ธโˆฉ๐ด๐‘๐‘“2๐‘‘๐œ‡โ‰ค๐œˆ(๐ธโˆฉ๐ด)+๐œˆ(๐ธโˆฉ๐ด๐‘)=๐œˆ(๐ธ)

Claim proved.
Step 3: ๆž„้€ ๅ‡บ potential RN derivative: ๆœ€ๅคง็š„ๅ…ƒ็ด  ๐‘“โˆˆโ„ฑ๏ธ€ ็Žฐๅœจๆˆ‘ไปฌ set

๐‘Ž:=sup{โˆซ๐‘“๐‘‘๐œ‡โˆฃ๐‘“โˆˆโ„ฑ๏ธ€}

ๆ˜พ็„ถๆœ‰:

1โ‰ค๐‘Žโ‰ค๐œˆ(๐‘‹)

pick ๐‘”๐‘›โˆˆโ„ฑ๏ธ€ s.t. โˆซ๐‘”๐‘›๐‘‘๐œ‡โ†—๏ธŽ๐‘Ž, ๅนถไธ” set

๐‘“๐‘›:=max{๐‘”1,โ‹ฏ,๐‘”๐‘›}

for each ๐‘›.
ๆ˜พ็„ถๆœ‰:

๐‘“๐‘›โ‰ค๐‘“๐‘›+1,โˆซ๐‘“๐‘›๐‘‘๐œ‡โ†—๏ธŽ๐‘Ž

ๅนถไธ”ๆ นๆฎๆˆ‘ไปฌ็š„ claim, ๆ‰€ๆœ‰ ๐‘“๐‘›โˆˆโ„ฑ๏ธ€.
ๆ นๆฎๅฏๆต‹ๅ‡ฝๆ•ฐ็š„ๆ€ง่ดจ,

โˆƒ๐‘“:=lim๐‘›๐‘“๐‘›โˆˆ๐ฟ+(๐œ‡), and โˆˆ๐ฟ1(๐œ‡) (since ๐œ‡ finite)

ๅนถไธ”ๆ นๆฎ monotone convergence theorem๏ผˆMCT๏ผ‰,

โˆซ๐‘“๐‘‘๐œ‡=lim๐‘›โ†’โˆžโˆซ๐‘“๐‘›๐‘‘๐œ‡=๐‘Ž

ๅนถไธ”, ๅฏนไบŽไปปๆ„ ๐ธ measurable, ๆ นๆฎ MCT ไนŸๆœ‰

โˆซ๐ธ๐‘“๐‘‘๐œ‡=lim๐‘›โ†’โˆžโˆซ๐ธ๐‘“๐‘›๐‘‘๐œ‡โ‰ค๐œˆ(๐ธ)

ๆˆ‘ไปฌ set:

๐œˆโ€ฒ(๐ธ):=โˆซ๐ธ๐‘“๐‘‘๐œ‡

Step 4: ่ฏๆ˜Ž ๐œˆโ€ฒ=๐œˆ.
Proof: ้ฆ–ๅ…ˆๆˆ‘ไปฌ็Ÿฅ้“ by def ๐œˆโ€ฒโ‰ค๐œˆ.
Set:

๐œˆฬƒ:=๐œˆโˆ’๐œˆโ€ฒโ‰ฅ0

By our assumption ๐œˆโ‰ช๐œ‡, ไปŽ่€ŒไนŸๆœ‰ ๐œˆฬƒโ‰ช๐œ‡.
ๅ› ่€Œๅช้œ€่ฆ่ฏๆ˜Ž ๐œˆฬƒโŠฅ๐œ‡, ๅฐฑๅฏไปฅๅพ—ๅˆฐ ๐œˆฬƒ=0, ไปŽ่€Œ่ฏๆ˜Žๅ‡บ ๐œˆโ€ฒ=๐œˆ.
่ฟ™ไธชๆ—ถๅ€™ Lemma ๅฐฑ่ตทไบ†ไฝœ็”จ:
Suppose for contradictin that ๐œˆฬƒโŸ‚ฬธ๐œ‡, ้‚ฃไนˆ by lemma, ็”ฑไบŽ ๐œˆฬƒ ๆ˜ฏไธ€ไธช finite positive measure, ๐œ‡ ไนŸๆ˜ฏไธ€ไธช finite positive measure, ๅˆ™ๅญ˜ๅœจ ๐œ–>0 ๅ’Œ nontrivial measurable ๐ธ, ไฝฟๅพ— ๐œˆฬƒโ‰ฅ๐œ–๐œ‡ on ๐ธ.
ไบŽๆ˜ฏ:

๐‘”:=๐‘“+๐œ–๐œ’๐ธโˆˆโ„ฑ๏ธ€

่€Œ โˆซ๐‘“๐‘‘๐œ‡=๐‘Ž, ๅ› ่€Œ

โˆซ๐‘”๐‘‘๐œ‡>๐‘Ž

่ฟ™ๅ’Œ ๐‘”โˆˆโ„ฑ๏ธ€ ๅ†ฒ็ช (ๅฆๅˆ™ๅฎƒ็š„็งฏๅˆ†ไธ€ๅฎšๅฐไบŽ็ญ‰ไบŽ ๐‘Ž).
ไปŽ่€Œ, ๐œ‡,๐œˆ ๆ˜ฏ finite p.m. ็š„ๆƒ…ๅ†ตๅพ—่ฏ.
Step 5: ๆŽจๅนฟ่‡ณ ๐œˆ finite s.m., ๐œ‡ finite p.m. ็š„ๆƒ…ๅ†ต.
็”ฑ Jordan decomposition theorem ๅ†™ๅ‡บ ๐œˆ+,๐œˆโˆ’, ๅ†็›ดๆŽฅ Apply Step 1 ๅณๅพ—่ฏ.
Step 6: ๆŽจๅนฟ่‡ณ ๐œˆ,๐œ‡ ๐œŽ-finite ็š„ๆƒ…ๅ†ต.
Proof: By ๐œŽ-finite ็š„ๅฎšไน‰, ๆˆ‘ไปฌๅฏไปฅ decompose

๐‘‹=โจ†๐‘›=1โˆž๐‘‹๐‘›

้‚ฃไนˆ by finite case, ๐œˆ|๐‘‹๐‘›, ๐œ‡|๐‘‹๐‘› is finite for each ๐‘›.
ๅ› ่€Œ

๐‘“๐‘›:=๐‘‘(๐œˆ|๐‘‹๐‘›)๐‘‘(๐œ‡|๐‘‹๐‘›)โˆƒ for each ๐‘›

ไบŽๆ˜ฏ, take

๐‘“:=โˆ‘๐‘›=1โˆž๐Ÿ๐‘‹๐‘›๐‘“๐‘›

ๅณๅฏๅพ—่ฏ.
Note: ่ฟ™้‡Œ็š„ ๐‘“ ๆ˜ฏ ext ๐œ‡-intble ็š„, ๅณ: ๐‘“+,๐‘“โˆ’ ่‡ณๅฐ‘ๆœ‰ไธ€ไธชๆ˜ฏ ext ๐œ‡-intble ็š„. ่ฟ™ follows from ๐œˆ ไฝœไธบไธ€ไธช signed measure ็š„ๅฎšไน‰: ๐œˆ ่‡ณๅคš admit +โˆž,โˆ’โˆž ไธญ็š„ไธ€ไธช.
Specially, ๅฆ‚ๆžœ ๐œˆ ๆ˜ฏไธ€ไธช positive measure, ้‚ฃไนˆ ๐‘“ ไธ€ๅฎšไนŸๆ˜ฏ้ž่ดŸ็š„, ไปŽ่€Œ ๐‘“โˆ’=0.

โ–ก

ไธ‹ไธ€ไธช lecture: ๆˆ‘ไปฌๅฐ† upgrade RN Thm to ไธ€ไธชๆ›ดๅŠ  general ็š„ version: Lebesgue Radon Nikodym Thm.

11.2 Lebesgue-Radon-Nikodym Theorem [Fol 3.2, finished; 3.3, finished]

recall Radon-Nikodym Theorem:

{๐œ‡๐œŽ-finite p.m.๐œˆ๐œŽ-finite s.m.๐œˆโ‰ช๐œ‡โŸน{โˆƒ!extended ๐œ‡-integrable๐‘“:๐‘‹โ†’โ„๐‘‘๐œˆ=๐‘“๐‘‘๐œ‡

ๆˆ‘ไปฌ็งฐ ๐‘“ ไธบ Radon-Nikodym Derivative:

๐œˆ(๐ธ)=โˆซ๐ธ๐‘“๐‘‘๐œ‡
Example 11.37

Application: conditional expectation.

(๐‘‹,๐’œ๏ธ€,๐œ‡)โ‰”([0,1),โ„ฌ๏ธ€([0,1)),๐‘š)

๐‘“:[0,1)โ†’โ„ Borel measurable.
Define:

๐ต:={โŒ€,[0,12),[12,1),๐‘‹}

๐‘“ ๅนถ้žไธ€ๅฎšๆ˜ฏ ๐ต-measurable ็š„.

11.2.1 LRNT: ไปปๆ„ ๐œŽ-finite ๐œˆ,๐œ‡, ๅฏๅฐ† ๐œˆ ๆ‹†่งฃๆˆ ๐œ†โŠฅ๐œ‡ ๅ’Œ ๐œŒโ‰ช๐œ‡

Theorem 11.60 : Lebesgue-Radon-Nikodym Theorem

ๅฆ‚ๆžœ {๐œ‡๐œŽ-finite p.m.๐œˆ๐œŽ-finite s.m. on (๐‘‹,๐’œ๏ธ€), ้‚ฃไนˆๅญ˜ๅœจๅ”ฏไธ€็š„ decomposition

๐œˆ=๐œ†+๐œŒ

where ๐œ†,๐œŒ ๆ˜ฏ ๐œŽ-finite ็š„ signed measure s.t. {๐œ†โŠฅ๐œ‡๐œŒโ‰ช๐œ‡.
(ไบŽๆ˜ฏ, by RNT, ๅญ˜ๅœจ ๐œ‡-unique ็š„ extended ๐œ‡-integrable ๐‘“:๐‘‹โ†’โ„ s.t. ๐‘‘๐œŒ=๐‘“๐‘‘๐œ‡ ).

Sktech of proof of LRN theorem: Assume for simplicity that ๐œ‡,๐œˆ ๆ˜ฏ finite p.m.
Like last time, look at

โ„ฑ๏ธ€:={๐‘“โˆˆ๐ฟ+:โˆซ๐ธ๐‘“๐‘‘๐œ‡โ‰ค๐œˆ(๐ธ)โˆ€๐ธโˆˆ๐’œ๏ธ€}/โˆผ

Saw: โ„ฑ๏ธ€ ๆœ‰ max element ๐‘“.
Define ๐œŒ by ๐‘‘๐œŒ=๐‘“๐‘‘๐œ‡.
Set:

๐œ†:=๐œˆโˆ’๐œŒ

Want: ๐œ†โŠฅ๐œ‡.
Prove by contradiction: ๅฆ‚ๆžœ ๐œ†โŸ‚ฬธ๐œ‡, ้‚ฃไนˆLemma 2 ๅ‘Š่ฏ‰ๆˆ‘ไปฌ: ๅญ˜ๅœจ ๐œ–>0 ๅ’Œ positive measure ็š„ ๐ธโˆˆ๐’œ๏ธ€ ไฝฟๅพ—:

๐œ†โ‰ฅ๐œ–๐œ‡

on ๐ธ.
Set

๐‘”:=๐‘“+๐œ–๐œ’๐ธ

ๅˆ™

โˆซ๐น๐‘”๐‘‘๐œ‡=โˆซ๐นโˆฉ๐ธ(๐‘“+๐œ–)๐‘‘๐œ‡+โˆซ๐นโˆฉ๐ธ๐‘๐‘“๐‘‘๐œ‡

ๅ› ่€Œ

๐œŒ(๐นโˆฉ๐ธ)+๐œ–๐œ‡(๐นโˆฉ๐ธ)+๐œŒ(๐นโˆฉ๐ธ๐‘)=๐œŒ(๐น)+๐œ–๐œ‡(๐นโˆฉ๐ธ)โ‰ค๐œˆ(๐น)โˆ’๐œ–๐œ‡(๐น)+๐œ–๐œ‡(๐นโˆฉ๐ธ)โ‰ค๐œˆ(๐น)

ๅ› ่€Œ ๐‘”โˆˆโ„ฑ๏ธ€ ไธ” ๐‘”>๐‘“.
ไปŽ่€Œๅพ—่ฏ ๐œ†โŠฅ๐œ‡. ไปŽ่€Œ existence proved.
Uniqueness part: Suppose we have

๐œˆ=๐œ†1+๐œŒ1=๐œ†2+๐œŒ2

where ๐œ†๐‘–โŠฅ๐œ‡, ๐œŒ๐‘–โ‰ช๐œ‡. ้‚ฃไนˆ

๐œ†1โˆ’๐œ†2=๐œŒ2โˆ’๐œŒ1

ๆˆ‘ไปฌ็Ÿฅ้“, ๐œ†1โˆ’๐œ†2 ๅ’Œ ๐œŒ2โˆ’๐œŒ1 ไนŸๆ˜ฏ signed measures. ๅนถไธ”,

(๐œ†1โˆ’๐œ†2)โŠฅ๐œ‡,(๐œŒ2โˆ’๐œŒ1)โ‰ช๐œ‡

By Lemma 1:

๐œ†1โˆ’๐œ†2=๐œŒ2โˆ’๐œŒ1=0

Properties of the RN derivative: (P91 in Folland)

๐‘‘(๐œˆ1+๐œˆ2)๐‘‘๐œ‡=๐‘‘๐œˆ1๐‘‘๐œ‡+๐‘‘๐œˆ2๐‘‘๐œ‡๐œˆโ‰ช๐œ‡,๐œ‡โ‰ช๐œ‡โŸน๐‘‘๐œˆ๐‘‘๐œ‡๐‘‘๐œ‡๐‘‘๐œˆ=1

๐œ‡-a.e. = ๐œˆ-a.e.

11.2.2 complex measure ไปฅๅŠ complex version of LRNT

Definition 11.59 : complex measure

ไธ€ไธช complex measure on a measurable space (๐‘‹,๐’œ๏ธ€) ๆ˜ฏไธ€ไธช map ๐œˆ:๐’œ๏ธ€โ†’โ„‚ satisfying ๐œˆ(โŒ€)=0 ไปฅๅŠ ctbl disjoint additivity.

Example 11.38

simple complex measures:

๐‘‹={1,2,โ‹ฏ,๐‘›} ๐œˆ p/s/c measure on ๐‘‹.
Since ๐‘‹={1,2,โ€ฆ,๐‘›}, a complex measure ๐œˆ is just a function

๐œˆ0:๐‘‹โ†’โ„‚,i.e., ๐œˆ0=(๐œˆ1,โ€ฆ,๐œˆ๐‘›)โˆˆโ„‚๐‘›.

่€Œ

๐œˆ(๐ธ)=โˆ‘๐‘ฅโˆˆ๐ธ๐œˆ0(๐‘ฅ)

๐œˆ positive: โˆˆโ„+๐‘› ๐œˆ signed: โˆˆโ„๐‘›

For discrete spaces, the total variation measure is defined pointwise:

|๐œˆ|(๐‘–)โ‰”|๐œˆ๐‘–|,for each ๐‘–=1,โ€ฆ,๐‘›.

So the total variation measure |๐œˆ| is just the vector of magnitudes:

|๐œˆ|=(|๐œˆ1|,|๐œˆ2|,โ€ฆ,|๐œˆ๐‘›|).

What is ๐‘‘๐œˆ๐‘‘|๐œˆ|?

Since this is a finite discrete setting, the Radon-Nikodym derivative is computed **pointwise**:

(๐‘‘๐œˆ๐‘‘|๐œˆ|)(๐‘–)={๐œˆ๐‘–|๐œˆ๐‘–|if ๐œˆ๐‘–โ‰ 0,0if ๐œˆ๐‘–=0.

So the result is a function ๐‘“:๐‘‹โ†’โ„‚, given by:

๐‘“(๐‘–)={๐œˆ๐‘–|๐œˆ๐‘–|if ๐œˆ๐‘–โ‰ 0,0if ๐œˆ๐‘–=0.๐‘“โ‰”๐‘‘๐œˆ๐‘‘|๐œˆ|=(๐œˆ1|๐œˆ1|,๐œˆ2|๐œˆ2|,โ€ฆ,๐œˆ๐‘›|๐œˆ๐‘›|),with the convention 00โ‰”0.

This derivative is a function that lives on the unit circle in โ„‚ (except at zero), and it satisfies:

|๐‘“(๐‘–)|=1whenever ๐œˆ๐‘–โ‰ 0.

Homework 10: on LRN Theorem and complex measure (40/40)

(Note: For this homework I applied for an one-day extension since I met with some emergent problem with my bank and rent payment.)

complex measure ็š„ total variation ็š„ formulas

Let ๐œˆ be a complex measure on a measurable space (๐‘‹,๐’œ๏ธ€). Prove that, for any ๐ธโˆˆ๐’œ๏ธ€:

|๐œˆ|(๐ธ)=sup{โˆ‘๐‘—=1๐‘›|๐œˆ(๐ธ๐‘—)|โˆฃ๐‘›โˆˆโ„•,๐ธ1โ€ฆ๐ธ๐‘›disjoint,๐ธ=โ‹ƒ๐‘—=1๐‘›๐ธ๐‘—}=sup{โˆ‘๐‘—=1โˆž|๐œˆ(๐ธ๐‘—)|โˆฃ๐ธ1,๐ธ2,โ€ฆdisjoint,๐ธ=โ‹ƒ๐‘—=1โˆž๐ธ๐‘—}=sup{|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โˆฃ๐‘“:๐‘‹โ†’โ„‚measurable,|๐‘“|โ‰ค1}.
Proof

Take some positive measure ๐œ‡ s.t. ๐œˆโ‰ช๐œ‡ (e.g. ๐œ‡:=|โ„œ๐œˆ|+|โ„‘๐œˆ|), then by RN Thm there exists ๐œ‡-unique RN derivative ๐‘“, and |๐œˆ| can be defined by

๐‘‘|๐œˆ|:=|๐‘“|๐‘‘๐œ‡

Now we denote:

๐œ‡1(๐ธ)โ‰”sup{โˆ‘๐‘—=1๐‘›|๐œˆ(๐ธ๐‘—)|โˆฃ๐‘›โˆˆโ„•,๐ธ1โ€ฆ๐ธ๐‘›disjoint,๐ธ=โ‹ƒ๐‘—=1๐‘›๐ธ๐‘—}๐œ‡2(๐ธ)โ‰”sup{โˆ‘๐‘—=1โˆž|๐œˆ(๐ธ๐‘—)|โˆฃ๐ธ1,๐ธ2,โ€ฆdisjoint,๐ธ=โ‹ƒ๐‘—=1โˆž๐ธ๐‘—}๐œ‡3(๐ธ)โ‰”sup{|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โˆฃ๐‘“:๐‘‹โ†’โ„‚measurable,|๐‘“|โ‰ค1}.

We will prove the equality by showing that ๐œ‡1โ‰ค๐œ‡2โ‰ค|๐œˆ|(๐ธ)โ‰ค๐œ‡3โ‰ค๐œ‡1.
Claim 1: ๐œ‡1โ‰ค๐œ‡2.
Proof: This is trivial since for each finite disjoint segmentation ๐ธ=โจ†๐‘—=1๐‘›๐ธ๐‘— of ๐ธ can be made into a countable segmentation of ๐ธ, by taking all ๐ธ๐‘=โŒ€ for ๐‘โ‰ฅ๐‘›+1. So every value included in {โˆ‘๐‘—=1๐‘›|๐œˆ(๐ธ๐‘—)|โˆฃ๐ธ=โจ†๐‘—=1๐‘›๐ธ๐‘—} is also in {โˆ‘๐‘—=1โˆž|๐œˆ(๐ธ๐‘—)|โˆฃ๐ธ=โจ†๐‘—=1โˆž๐ธ๐‘—}. Thus taking sup, we have the ineq.
Claim 2: ๐œ‡2โ‰ค|๐œˆ|โ‰ค๐œ‡3.
Since ๐œˆโ‰ช|๐œˆ| (Folland prop 3.13), by complex RN Thm we have have

๐‘“โ‰”๐‘‘๐œˆ๐‘‘|๐œˆ|โˆˆ๐ฟ1(|๐œˆ|)

Notice that ๐‘“ have absolute value 1, |๐œˆ|-a.e. (Folland prop 3.13)
Suppose ๐ธ=โŠ”1โˆž๐ธ๐‘—, we have:

โˆ‘๐‘—=1โˆž|๐œˆ(๐ธ๐‘—)|โ‰คโˆ‘๐‘—=1โˆž|๐œˆ|(๐ธ๐‘—)by property of total variation measure=|๐œˆ|(๐ธ)=โˆซ๐ธ1๐‘‘|๐œˆ|by ctbl disjoint additivity =โˆซ๐ธ|๐‘“|2๐‘‘|๐œˆ|=โˆซ๐ธ๐‘“ฬ„๐‘“๐‘‘|๐œˆ|since ๐‘“ have absolute value 1 ๐œˆ-a.e.=โˆซ๐ธ๐‘“ฬ„๐‘‘๐œˆ๐‘‘|๐œˆ|๐‘‘|๐œˆ|

To confirm this equal to โˆซ๐‘“ฬ„๐‘‘๐œˆ, we extend Folland prop 3.9 to the complex case.

Proposition 11.30

For complex measure ๐œˆ and ๐œŽ-finite positive measure ๐œ‡ s.t. ๐œˆโ‰ช๐œ‡, if ๐‘”โˆˆ๐ฟ1(๐œˆ), then

๐‘”(๐‘‘๐œˆ๐‘‘๐œ‡)โˆˆ๐ฟ1(๐œ‡),โˆซ๐‘”๐‘‘๐œˆ=โˆซ๐‘”(๐‘‘๐œˆ๐‘‘๐œ‡)๐‘‘๐œ‡

And the proof just follows from the finite signed-measure case, applied both to im part and re part.

โˆซ๐‘”๐‘‘๐œˆ=โˆซ๐‘”๐‘‘(โ„œ๐œˆ)+๐‘–โˆซ๐‘”๐‘‘(โ„‘๐œˆ)=โˆซ๐‘”(๐‘‘(โ„œ๐œˆ)๐‘‘๐œ‡)๐‘‘๐œ‡+๐‘–โˆซ๐‘”(๐‘‘(โ„‘๐œˆ)๐‘‘๐œ‡)๐‘‘๐œ‡=โˆซ๐‘”(โ„œ๐‘‘๐œˆ๐‘‘๐œ‡+๐‘–โ„‘๐‘‘๐œˆ๐‘‘๐œ‡)๐‘‘๐œ‡=โˆซ๐‘”(๐‘‘๐œˆ๐‘‘๐œ‡)๐‘‘๐œ‡

Now we back to Claim 2, since ๐‘“,๐‘“ฬ„โˆˆ๐ฟ1(๐œˆ), we have:

โˆ‘๐‘—=1โˆž|๐œˆ(๐ธ๐‘—)|โ‰ค|๐œˆ|(๐ธ)=โˆซ๐ธ๐‘“ฬ„๐‘‘๐œˆ๐‘‘|๐œˆ|๐‘‘|๐œˆ|=โˆซ๐ธ๐‘“ฬ„๐‘‘๐œˆโ‰ค|โˆซ๐ธ๐‘“ฬ„๐‘‘๐œˆ|

Since |๐‘“ฬ„|โ‰ค1 (in ๐œˆ-a.e. sense), this shows that every element in {โˆ‘๐‘—=1โˆž|๐œˆ(๐ธ๐‘—)|โˆฃ๐ธ=โจ†๐‘—=1โˆž๐ธ๐‘—} is less then or equal to |๐œˆ|(๐ธ)|, and |๐œˆ|(๐ธ)| is less then some element in {|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โˆฃmeasurable |๐‘“|โ‰ค1}, proves that ๐œ‡2โ‰ค|๐œˆ|โ‰ค๐œ‡3.
Claim 3: ๐œ‡3โ‰ค๐œ‡1.
For arbitrary simple function ๐œ™โ‰”โˆ‘1๐‘›๐‘๐‘˜๐œ’๐ธ๐‘˜ where |๐‘๐‘˜|โ‰ค1 for all ๐‘˜,๐ธ๐‘– s are disjoint and โ‹ƒ๐‘–=1๐‘›๐ธ๐‘–=๐ธ. We have

|โˆซ๐ธ๐œ™๐‘‘๐œˆ|โ‰คโˆ‘๐‘˜=1๐‘›|๐‘๐‘˜โˆซ๐ธ๐‘˜๐œ’๐ธ๐‘˜๐‘‘๐œˆ|=โˆ‘๐‘˜=1๐‘›|๐‘๐‘˜||๐œˆ(๐ธ๐‘˜)|โ‰คโˆ‘๐‘˜=1๐‘›|๐œˆ(๐ธ๐‘˜)|โ‰ค๐œ‡1(๐ธ)

Now we consider the general case: any measurable ๐‘“.
Fix arbitrary measurable ๐‘“ s.t. |๐‘“|โ‰ค1, since it is measurable, we can choose seq of simple functions (๐œ™๐‘›)1โˆž that approximate ๐‘“ pointwisely from below.

lim๐‘›โ†’โˆž๐œ™๐‘›=๐‘“

with

0โ‰ค|๐œ™1|โ‰ค|๐œ™2|โ‰คโ‹ฏโ‰ค|๐‘“|

Then |๐‘“| as a dominating function for (|๐œ™๐‘›|)๐‘›, by DCT we obtain:

โˆซ๐ธ๐‘“๐‘‘(โ„œ๐œˆ)=lim๐‘›โ†’โˆžโˆซ๐ธ๐œ™๐‘›๐‘‘(โ„œ๐œˆ)

and

โˆซ๐ธ๐‘“๐‘‘(โ„‘๐œˆ)=lim๐‘›โ†’โˆžโˆซ๐ธ๐œ™๐‘›๐‘‘(โ„‘๐œˆ)

Thus

โˆซ๐ธ๐‘“๐‘‘๐œˆ=โˆซ๐ธ๐‘“๐‘‘(โ„œ๐œˆ)+๐‘–โˆซ๐ธ๐‘“๐‘‘(โ„‘๐œˆ)=lim๐‘›โ†’โˆž(โˆซ๐ธ๐œ™๐‘›๐‘‘(โ„œ๐œˆ)+๐‘–โˆซ๐ธ๐œ™๐‘›๐‘‘(โ„‘๐œˆ))=lim๐‘›โ†’โˆžโˆซ๐ธ๐œ™๐‘›๐‘‘๐œˆ

Since for each ๐œ™๐‘›, we have 0โ‰ค|๐œ™๐‘›(๐‘ฅ)|โ‰ค|๐‘“(๐‘ฅ)|โ‰ค1 for a.e. ๐‘ฅโˆˆ๐ธ, we can apply the ineq we obtained that

|โˆซ๐ธ๐œ™๐‘›๐‘‘๐œˆ|โ‰ค๐œ‡1(๐ธ)

for each ๐‘›. Thus taking limit we get:

|โˆซ๐ธ๐‘“๐‘‘๐œˆ|โ‰ค๐œ‡1(๐ธ)

Taking supremum over ๐‘“, proves that ๐œ‡3(๐ธ)โ‰ค๐œ‡1(๐ธ).
Thus since we have shown ๐œ‡1โ‰ค๐œ‡2โ‰ค|๐œˆ|โ‰ค๐œ‡3โ‰ค๐œ‡1, every inequality above is an equality, i.e.

๐œ‡1=๐œ‡2=๐œ‡3=|๐œˆ|

finishing the proof.

โ–ก

complex measure ไธŽๅ…ถ total variation measure ไน‹้—ด็š„ๅ…ณ็ณป: ๆ•ดไฝ“ๅณๅฏๅ†ณๅฎšๅฑ€้ƒจ

Let ๐œˆ be a complex measure on a measurable space (๐‘‹,๐’œ๏ธ€).

11.2.3 ๐œˆ(๐‘‹)=|๐œˆ|(๐‘‹)โ‡”๐œˆ=|๐œˆ|โ‡”๐œˆ positive

  • ๐œˆ(๐‘‹)=|๐œˆ|(๐‘‹);

  • ๐œˆ is a (finite) positive measure;

  • ๐œˆ=|๐œˆ|.

Proof

(ii) โŸน (iii): If ๐œˆ is positive then ๐œˆโˆ’=0, so ๐œˆ=|๐œˆ|=๐œˆ+.
(iii) โŸน (i): Trivially true by taking ๐ธ=๐‘‹.
(i) โŸน (ii): Take some positive measure ๐œ‡ s.t. ๐œˆโ‰ช๐œ‡ (e.g. ๐œ‡:=|โ„œ๐œˆ|+|โ„‘๐œˆ|), then by RN Thm there exists ๐œ‡-unique RN derivative ๐‘“, and |๐œˆ| can be defined by

๐‘‘|๐œˆ|:=|๐‘“|๐‘‘๐œ‡

Then by def

โˆซ๐‘“๐‘‘๐œ‡=โˆซ|๐‘“|๐‘‘๐œ‡,๐‘–.๐‘’.โˆซโ„œ๐‘“๐‘‘๐œ‡+๐‘–โˆซโ„‘๐‘“๐‘‘๐œ‡=โˆซ|๐‘“|๐‘‘๐œ‡

Since the right hand side is real, we have:

โˆซ(|๐‘“|โˆ’โ„œ๐‘“)๐‘‘๐œ‡=0

Note that, |๐‘“|โˆ’โ„œ๐‘“ is always nonnegative, so this implies that โ„œ๐‘“=|๐‘“|๐œ‡-a.e.
Thus โ„‘๐‘“=0๐œ‡-a.e., so ๐‘“=|๐‘“| is real and positive ๐œ‡-a.e. Thus

๐œˆ(๐ธ)=โˆซ๐ธ๐‘“๐‘‘๐œ‡โˆˆโ„+,โˆ€๐ธโˆˆ๐’œ๏ธ€

finishing the proof that ๐œˆ is a positive measure.

โ–ก

11.2.4 |๐œˆ(๐‘‹)|=|๐œˆ|(๐‘‹)โ‡”๐œˆ=๐œ†|๐œˆ| for some |๐œ†|=1

Prove that the following two conditions are equivalent:

  • |๐œˆ(๐‘‹)|=|๐œˆ|(๐‘‹);

  • there exists a complex number ๐œ† with |๐œ†|=1 such that ๐œˆ=๐œ†|๐œˆ|.

Proof

(i) โŸน (ii): Since ๐œˆโ‰ช|๐œˆ|, by complex RN Thm we have RN derivative

โ„Žโ‰”๐‘‘๐œˆ๐‘‘|๐œˆ|โˆˆ๐ฟ1(|๐œˆ|)

Notice that โ„Ž have absolute value 1, |๐œˆ|-a.e.
Then by def of RN derivative we have

๐œˆ(๐‘‹)=โˆซ๐‘‹โ„Ž๐‘‘|๐œˆ|

Thus

|๐œˆ(๐‘‹)|=|โˆซ๐‘‹โ„Ž๐‘‘|๐œˆ||โ‰คโˆซ๐‘‹|โ„Ž|๐‘‘|๐œˆ|=โˆซ๐‘‹1๐‘‘|๐œˆ|=|๐œˆ|(๐‘‹)

Since we have |๐œˆ(๐‘‹)|=|๐œˆ|(๐‘‹), it implie that:

|โˆซ๐‘‹โ„Ž๐‘‘|๐œˆ||=โˆซ๐‘‹|โ„Ž|๐‘‘|๐œˆ|

Claim: โ„Ž is constant |๐œˆ|-a.e.
We first prove a lemma:

Lemma 11.41

Let ๐œ‡ be a finite positive measure.
For measurable function ๐‘“:๐‘‹โ†’โ„‚, if |๐‘“|=๐‘˜ a.e. for some nonzero constant ๐‘˜ and

|โˆซ๐‘“๐‘‘๐œ‡|=โˆซ|๐‘“|๐‘‘๐œ‡

then ๐‘“ must be a.e. constant.

Proof of Lemma: Set:

๐‘โ‰”โˆซ๐‘“๐‘‘๐œ‡|โˆซ๐‘“๐‘‘๐œ‡|

Then |๐‘|=1, and we consider:

โˆซ๐‘“๐‘‘๐œ‡=๐‘|โˆซ๐‘“๐‘‘๐œ‡|=๐‘โˆซ|๐‘“|๐‘‘๐œ‡

Define ๐‘”(๐‘ฅ)โ‰”๐‘ฬ„๐‘“(๐‘ฅ), so:

โˆซ๐‘”๐‘‘๐œ‡=๐‘ฬ„โˆซ๐‘“๐‘‘๐œ‡=๐‘ฬ„๐‘โˆซ|๐‘“|๐‘‘๐œ‡=โˆซ|๐‘“|๐‘‘๐œ‡

Notice โˆซ|๐‘“|๐‘‘๐œ‡โˆˆโ„+ and

โˆซ๐‘”๐‘‘๐œ‡=โˆซโ„œ๐‘”๐‘‘๐œ‡+๐‘–โˆซโ„‘๐‘”๐‘‘๐œ‡โˆˆโ„‚

Thus

โˆซโ„œ๐‘”๐‘‘๐œ‡=โˆซ|๐‘”|๐‘‘๐œ‡=โˆซ|๐‘“|๐‘‘๐œ‡โŸนโˆซ(โ„œ๐‘”โˆ’|๐‘”|)๐‘‘๐œ‡=0

Since by def:

0โ‰คโ„œ๐‘”โ‰ค|๐‘”|

We must have

โ„œ๐‘”=|๐‘”|๐‘Ž.๐‘’.

This proves that ๐‘” is a.e. real. And also since |๐‘”|=|๐‘“|=๐‘˜ a.e., ๐‘” is then constant ๐‘˜ a.e.
Therefore, ๐‘“ is constant ๐‘˜๐‘ฬ„ a.e.

Now we go back to the proof of the original statement. By our Lemma we get:

โ„Ž=|โˆซโ„Ž๐‘‘๐œ‡|โˆซโ„Ž๐‘‘๐œ‡ฬ„constant for |๐œˆ|-a.e. ๐‘ฅ

Therefore,

๐œˆ=|โˆซโ„Ž๐‘‘๐œ‡|โˆซโ„Ž๐‘‘๐œ‡ฬ„|๐œˆ|

This finishes the proof of (i) โŸน (ii).
(ii) โŸน(i): This direction is trivial. Since ๐œˆ=๐œ†|๐œˆ|, we have

|๐œˆ(๐‘‹)|=|๐œ†||๐œˆ|(๐‘‹)=1|๐œˆ|(๐‘‹)=|๐œˆ|(๐‘‹)

โ–ก

11.3 complex measures on (๐‘‹,๐’œ๏ธ€) ็ป„ๆˆไธ€ไธช complex Banach space

Let (๐‘‹,๐’œ๏ธ€) be a measurable space. Prove that the set โ„ณ๏ธ€ of complex measures on (๐‘‹,๐’œ๏ธ€) is a complex Banach space, with norm given by โˆฅ๐œˆโˆฅโ‰”|๐œˆ|(๐‘‹).

Proof

Claim 1: โ„ณ๏ธ€ is a complex vector space, with addition operation defined by the addition of two complex measures, and scalar multiplication defined by scaling a complex measure by a complex number.
Proof of Claim 1: For ๐œˆ,๐œ‡โˆˆโ„ณ๏ธ€, and ๐›ผโˆˆโ„‚, define:

  • (๐œˆ+๐œ‡)(๐ธ)โ‰”๐œˆ(๐ธ)+๐œ‡(๐ธ) for all ๐ธโˆˆ๐’œ๏ธ€

  • (๐›ผ๐œˆ)(๐ธ)โ‰”๐›ผโ‹…๐œˆ(๐ธ) for all ๐ธโˆˆ๐’œ๏ธ€.

Then: (๐œˆ+๐œ‡)(โŒ€)=0+0=0,(๐›ผ๐œˆ)(โŒ€)=๐›ผ0=0.
Also, ๐œˆ+๐œ‡ and ๐›ผ๐œˆ are both countably additive, since sum and scalar multiples preserve this property: for ๐ธ=โจ†๐‘—=1โˆž๐ธ๐‘— with each ๐ธ๐‘—โˆˆ๐’œ๏ธ€, we have:

(๐œˆ+๐œ‡)(โจ†๐‘—=1โˆž๐ธ๐‘—)=๐œˆ(โจ†๐‘—=1โˆž๐ธ๐‘—)+๐œ‡(โจ†๐‘—=1โˆž๐ธ๐‘—)=๐œˆ(๐ธ)+๐œˆ(๐ธ)=(๐œˆ+๐œ‡)(๐ธ)

and

(๐›ผ๐œˆ)(โจ†๐‘—=1โˆž๐ธ๐‘—)=๐›ผโ‹…๐œˆ(โจ†๐‘—=1โˆž๐ธ๐‘—)=๐›ผ๐œˆ(๐ธ)

So they are also complex measures, showing that โ„ณ๏ธ€ is closed under addition and scalar multiplication, thus a complex vector space.
Claim 2: total variation โˆฅ๐œˆโˆฅ:=|๐œˆ(๐‘‹)| defines a norm on โ„ณ๏ธ€.
Proof of Claim 2: To verify this is a norm, we check the norm requirements:

  • Nonnegative: โˆฅ๐œˆโˆฅโ‰ฅ0, and โˆฅ๐œˆโˆฅ=0โ‡”๐œˆ=0
    Proof: โˆฅ๐œˆโˆฅโ‰ฅ0 follows from that |๐œˆ| is a p.m.
    Since we know ๐œˆโ‰ช|๐œˆ|, if |๐œˆ|(๐‘‹)=0 then ๐‘‹ is a null set of |๐œˆ|, and thus is a null set for ๐œˆ, so ๐œˆ=0;
    Conversely, if ๐œˆ=0 then

    โˆฅ๐œˆโˆฅโ‰”|๐œˆ|(๐‘‹)=sup{โˆ‘๐‘—=1๐‘›|๐œˆ(๐ธ๐‘—)|:๐‘‹=โจ†๐‘—=1๐‘›๐ธ๐‘—}=sup{0}=0

    finishing the proof that โˆฅ๐œˆโˆฅ=0โ‡”๐œˆ=0

  • Homogeneity: โˆฅ๐›ผ๐œˆโˆฅ=|๐›ผ|โ‹…โˆฅ๐œˆโˆฅ
    Proof:

    โˆฅ๐›ผ๐œˆโˆฅโ‰”|๐›ผ๐œˆ|(๐‘‹)=sup{โˆ‘๐‘—=1๐‘›|๐›ผ๐œˆ(๐ธ๐‘—)|:๐‘‹=โจ†๐‘—=1๐‘›๐ธ๐‘—}=|๐›ผ|sup{โˆ‘๐‘—=1๐‘›|๐œˆ(๐ธ๐‘—)|:๐‘‹=โจ†๐‘—=1๐‘›๐ธ๐‘—}=|๐›ผ||๐œˆ|(๐‘‹)=|๐›ผ|โˆฅ๐œˆโˆฅ
  • Triangle inequality: โˆฅ๐œˆ+๐œ‡โˆฅโ‰คโˆฅ๐œˆโˆฅ+โˆฅ๐œ‡โˆฅ
    Proof:

    |๐œˆ+๐œ…|(๐‘‹)=sup{โˆ‘๐‘–=1๐‘›|(๐œˆ+๐œ…)(๐ธ๐‘–)|:๐‘‹=โจ†๐‘–=1๐‘๐ธ๐‘–}โ‰คsup{โˆ‘๐‘–=1๐‘›(|๐œˆ(๐ธ๐‘–)|+|๐œ…(๐ธ๐‘–)|):๐‘‹=โจ†๐‘–=1๐‘๐ธ๐‘–}by tri ineq in โ„=sup{โˆ‘๐‘–=1๐‘›|๐œˆ(๐ธ๐‘–)|+โˆ‘๐‘–=1๐‘›|๐œ…(๐ธ๐‘–)|:๐‘‹=โจ†๐‘–=1๐‘๐ธ๐‘–}โ‰คsup{โˆ‘๐‘–=1๐‘›|๐œˆ(๐ธ๐‘–)|:๐‘‹=โจ†๐‘–=1๐‘๐ธ๐‘–}+sup{โˆ‘๐‘–=1๐‘›|๐œ…(๐ธ๐‘–)|:๐‘‹=โจ†๐‘–=1๐‘๐ธ๐‘–}=|๐œˆ|(๐‘‹)+|๐œ…|(๐‘‹)

Here we have finished the proof of (โ„ณ๏ธ€,โˆฅโ‹…โˆฅ) being a normed โ„‚-vector space.
Claim 3: (โ„ณ๏ธ€,โˆฅโ‹…โˆฅ) is complete (thus Banach space)
Proof: Let (๐œˆ๐‘›) be a Cauchy sequence in โ„ณ๏ธ€. We have

|๐œˆ๐‘›(๐ต)โˆ’๐œˆ๐‘š(๐ต)|=|(๐œˆ๐‘›โˆ’๐œˆ๐‘š)(๐ต)|โ‰ค|(๐œˆ๐‘›โˆ’๐œˆ๐‘š)(๐‘‹)|=โˆฅ๐œˆ๐‘›โˆ’๐œˆ๐‘šโˆฅfor all ๐ตโˆˆ๐’œ๏ธ€

In particular, (๐œˆ๐‘›(๐ต))๐‘› is a Cauchy sequence for all ๐ตโˆˆ๐’œ๏ธ€. For each ๐ตโˆˆ๐’œ๏ธ€, this is a Cauchy seq in โ„‚, thus converges. So we can get:

๐œˆ(๐ต)โ‰”lim๐‘›๐œˆ๐‘›(๐ต)

as the pointwise limit (by a point we mean a set).
Claim 3.1: ๐œˆโˆˆโ„ณ๏ธ€.
Since for all ๐‘›, ๐œˆ๐‘›(โŒ€)=0, we have:

๐œˆ(โŒ€)โ‰”lim๐‘›๐œˆ๐‘›(โŒ€)=0

For a countable disjoint union of measurable sets ๐ธ=โจ†๐‘–=1โˆž๐ธ๐‘–,

๐œˆ(๐ธ)=lim๐‘›๐œˆ๐‘›(๐ธ)=lim๐‘›โˆ‘๐‘–๐œˆ๐‘›(๐ธ๐‘–)

We know by property of total variation measure that for each ๐‘› we have:

โˆ‘๐‘–|๐œˆ๐‘›(๐ธ๐‘–)|<|๐œˆ๐‘›|(๐‘‹)=โˆฅ๐œˆ๐‘›โˆฅ<๐‘€

for some uniform bound ๐‘€ for each ๐‘›, since โˆฅ๐œˆ๐‘›โˆฅ is a Cauchy seq in โ„‚. Thus we can exchange the order of taking limit and sum. Then we get:

๐œˆ(๐ธ)=lim๐‘›๐œˆ๐‘›(๐ธ)=lim๐‘›โˆ‘๐‘–๐œˆ๐‘›(๐ธ๐‘–)=โˆ‘๐‘–lim๐‘›๐œˆ๐‘›(๐ธ๐‘–)=โˆ‘๐‘–๐œˆ(๐ธ๐‘–)

verifying the countable disjoint additivity.
And notice, as we have mentioned, for each measurable set ๐ธโˆˆ๐’œ๏ธ€, since (๐œˆ๐‘›(๐ธ))๐‘› is a Cauchy sequence in โ„‚, it is bounded, verifying that ๐œˆ is a valid complex measure.
Claim 3.2: ๐œˆ๐‘›โ†’๐œˆ in โˆฅโ‹…โˆฅ.
Fix ๐œ–>0.
By Cauchy in โˆฅโ‹…โˆฅ, there exists ๐‘โˆˆโ„• s.t. for all ๐‘š,๐‘›โ‰ฅ๐‘ , we have

โˆฅ๐œˆ๐‘šโˆ’๐œˆ๐‘›โˆฅ=|๐œˆ๐‘šโˆ’๐œˆ๐‘›|(๐‘‹)<๐œ–

Fix ๐‘›โ‰ฅ๐‘, and consider the sequence ๐œˆ๐‘š. Then ๐œˆ๐‘šโ†’๐œˆ pointwise implies ๐œˆ๐‘›โˆ’๐œˆ๐‘šโ†’๐œˆ๐‘›โˆ’๐œˆ pointwise. Thus

โˆฅ๐œˆ๐‘›โˆ’๐œˆโˆฅ=|๐œˆ๐‘›โˆ’๐œˆ|(๐‘‹)โ‰คlimโ€‰inf๐‘šโ†’โˆž|๐œˆ๐‘›โˆ’๐œˆ๐‘š|(๐‘‹)<๐œ–

Since ๐œ–>0 is arbitrary, this shows that, โˆฅ๐œˆ๐‘›โˆ’๐œˆโˆฅโ†’0 as ๐‘›โ†’โˆž, proving the convergence is in norm.
Now we conclude that (โ„ณ๏ธ€,โˆฅโ‹…โˆฅ) is a Banach space.

โ–ก

Positivity

Let ๐œˆ1, ๐œˆ2 be complex measures on a measurable space (๐‘‹,๐’œ๏ธ€) such that โˆฅ๐œˆ1+๐œˆ2โˆฅ=โˆฅ๐œˆ1โˆฅ+โˆฅ๐œˆ2โˆฅ. Is it true that there exists a nonzero constant ๐‘Žโˆˆโ„‚ such that ๐‘Ž๐œˆ1 and ๐‘Ž๐œˆ2 are both positive measures?

Solution

No, not necessarily.

Proof

Consider ๐‘‹:={๐‘š,๐‘›}
Define ๐œˆ1,๐œˆ2 by atoms:

๐œˆ1({๐‘š})=๐œˆ2({๐‘š})=1,๐œˆ1({๐‘›})=๐œˆ2({๐‘›})=โˆ’1

Then

โˆฅ๐œˆ1+๐œˆ2โˆฅ=โˆฅ2๐œˆ1โˆฅ=|2๐œˆ1|(๐‘‹)=4=โˆฅ๐œˆ1โˆฅ+โˆฅ๐œˆ2โˆฅ

But there is no nonzero constant ๐‘Žโˆˆโ„‚ such that ๐‘Ž๐œˆ1 and ๐‘Ž๐œˆ2 are both positive measures.
This is because for any nonzero constant ๐‘Ž scaled on ๐œˆ1: if ๐‘Ž real, then it either flip, or preserve the sign of ๐œˆ1({๐‘š}) and ๐œˆ1({๐‘›}), where there is always one positive number and one negative number between them; if ๐‘Ž complex, then make the two numbers complex.
In both case, ๐œˆ1 cannot become a positive measure. And since ๐œˆ2 is defined the same as ๐œˆ1, same for it. Therefore it can never become positive measure by scaling a nonzero constant.

โ–ก

Averaging: Conditional Expectation

Let (๐‘‹,๐’œ๏ธ€,๐œ‡) be a finite measure space (i.e.ย a measure space such that ๐œ‡(๐‘‹)<โˆž). Let โ„ฌ๏ธ€โŠ‚๐’œ๏ธ€ be a sub-๐œŽ-algebra, and set ๐œˆโ‰”๐œ‡|โ„ฌ๏ธ€. Thus (๐‘‹,โ„ฌ๏ธ€,๐œˆ) is also a finite measure space.

  • Prove that if ๐‘“:๐‘‹โ†’โ„‚ is โ„ฌ๏ธ€-measurable, then ๐‘“ is ๐’œ๏ธ€-measurable. Is the converse true?

  • Suppose that ๐‘“โˆˆ๐ฟ1(๐œ‡). Prove that there exists a โ„ฌ๏ธ€-measurable function ๐‘”โˆˆ๐ฟ1(๐œˆ) such that โˆซ๐ธ๐‘“๐‘‘๐œ‡=โˆซ๐ธ๐‘”๐‘‘๐œˆ for all ๐ธโˆˆโ„ฌ๏ธ€. Also prove that any two such functions ๐‘” must agree outside a set of ๐œˆ-measure zero.

  • Construct ๐‘” explicitly in the case when ๐‘‹={1,2,3,4}, ๐’œ๏ธ€=๐’ซ๏ธ€(๐‘‹), ๐œ‡({๐‘–})=1/4 for ๐‘–โˆˆ๐‘‹, and โ„ฌ๏ธ€={โˆ…,{1,2},{3,4},๐‘‹}. Thus, given the four complex numbers ๐‘“(๐‘–), 1โ‰ค๐‘–โ‰ค4, you should find the four complex numbers ๐‘”(๐‘–), 1โ‰ค๐‘–โ‰ค4.

Hint: use the Radonโ€“Nikodym Theorem. Remark: if ๐œ‡ is a probability measure, then we can view ๐‘” as the conditional expectation of (the random variable) ๐‘“ with respect to the ๐œŽ-algebra โ„ฌ๏ธ€.

Proof

of (a): Suppose ๐‘“:๐‘‹โ†’โ„‚ is โ„ฌ๏ธ€-measurable, then for any Borel set ๐ตโŠ‚โ„‚, ๐‘“โˆ’1(๐ต)โˆˆโ„ฌ๏ธ€โŠ‚๐’œ๏ธ€, so ๐‘“ is ๐’œ๏ธ€ -measurable.
The converse is not true.
Consider ๐‘‹={0,1,2,3},๐ดโ‰”๐’ซ๏ธ€(๐‘‹),โ„ฌ๏ธ€โ‰”{โŒ€,๐‘‹}.
Consider ๐‘“:๐‘ฅโ†ฆ๐‘ฅ from ๐‘‹ to โ„.
๐‘“ is ๐’œ๏ธ€-measurable since ๐’œ๏ธ€ is the power set, containing all subsets of ๐‘‹.
But ๐‘“โˆ’1({0})={0}โˆ‰โ„ฌ๏ธ€. Thus ๐‘“ is not โ„ฌ๏ธ€-measurable.

โ–ก

Proof

of (b): Let ๐œˆโ‰”๐œ‡|โ„ฌ๏ธ€, and define a signed measure on โ„ฌ๏ธ€ by:

๐œ†(๐ธ)โ‰”โˆซ๐ธ๐‘“๐‘‘๐œ‡,๐ธโˆˆโ„ฌ๏ธ€

Then ๐œ†โ‰ช๐œˆ, since ๐œˆ(๐ธ)=๐œ‡(๐ธ)=0โŸน๐œ†(๐ธ)=0.
By Radon-Nikodym Thm, there exists a โ„ฌ๏ธ€-measurable function ๐‘”โˆˆ๐ฟ1(๐œˆ) such that

๐œ†(๐ธ)=โˆซ๐ธ๐‘”๐‘‘๐œˆ for all ๐ธโˆˆโ„ฌ๏ธ€

Then

โˆซ๐ธ๐‘“๐‘‘๐œ‡=โˆซ๐ธ๐‘”๐‘‘๐œˆ,โˆ€๐ธโˆˆโ„ฌ๏ธ€

Suppose ๐‘”1,๐‘”2 are both such functions, then

โˆซ๐ธ(๐‘”1โˆ’๐‘”2)๐‘‘๐œˆ=0โˆ€๐ธโˆˆโ„ฌ๏ธ€

Define

๐บ+:={๐‘”1โˆ’๐‘”2>0},๐บโˆ’:={๐‘”1โˆ’๐‘”2<0}

These two sets are in โ„ฌ๏ธ€ since ๐‘”1,๐‘”2 are โ„ฌ๏ธ€-measurable. Then we have:

โˆซ๐บ+(๐‘”1โˆ’๐‘”2)๐‘‘๐œˆ=โˆซ๐บโˆ’(๐‘”1โˆ’๐‘”2)๐‘‘๐œˆ=0

Since on ๐บ+ we have ๐‘”1โˆ’๐‘”2>0,

โˆซ๐บ+(๐‘”1โˆ’๐‘”2)๐‘‘๐œˆ=0โŸนโˆซ๐บ+|๐‘”1โˆ’๐‘”2|๐‘‘๐œˆ=0โŸน๐‘”1=๐‘”2๐œˆ-a.e. on ๐บ+โŸน๐œˆ(๐บ+)=0

Similarly, since on ๐บโˆ’ we have ๐‘”1โˆ’๐‘”2<0,

โˆซ๐บโˆ’(๐‘”1โˆ’๐‘”2)๐‘‘๐œˆ=0โŸนโˆ’โˆซ๐บโˆ’|๐‘”1โˆ’๐‘”2|๐‘‘๐œˆ=0โŸน๐‘”1=๐‘”2๐œˆ-a.e. on ๐บ+โŸน๐œˆ(๐บโˆ’)=0

Thus

๐œˆ{๐‘”1โ‰ ๐‘”2}=๐œˆ(๐บ+)+๐œˆ(๐บโˆ’)=0

This finishes the proof.

โ–ก

Solution

of (c): Given:

  • ๐‘‹={1,2,3,4}

  • ๐’œ๏ธ€=๐’ซ๏ธ€(๐‘‹)

  • ๐œ‡({๐‘–})=1/4 for each ๐‘–

  • โ„ฌ๏ธ€={โˆ…,{1,2},{3,4},๐‘‹}

Suppose we have: ๐‘“:๐‘‹โ†’โ„‚, so ๐‘“(๐‘–)โˆˆโ„‚ for ๐‘–=1,2,3,4. We want to find: ๐‘”(๐‘–)โˆˆโ„‚,๐‘–=1,2,3,4, such that ๐‘” is โ„ฌ๏ธ€-measurable and

โˆซ๐ธ๐‘“๐‘‘๐œ‡=โˆซ๐ธ๐‘”๐‘‘๐œˆ for all ๐ธโˆˆโ„ฌ๏ธ€

Notice that โ„ฌ๏ธ€={โˆ…,{1,2},{3,4},๐‘‹}, we must set ๐‘”(1)=๐‘”(2) and ๐‘”(3)=๐‘”(4), this is because, suppose if we set ๐‘”(1)โ‰ ๐‘”(2), then it will happen that

1โˆˆ๐‘”โˆ’1(๐‘”(1))โˆŒ2

No set in โ„ฌ๏ธ€ satisfy this condition, thus ๐‘”โˆ’1(๐‘”(1))โˆ‰โ„ฌ๏ธ€, contradicts that ๐‘” is โ„ฌ๏ธ€-measurable.
Thus we set

๐‘”(1)=๐‘”(2)=๐‘Ž,๐‘”(3)=๐‘”(4)=๐‘

We have:

โˆซ{1,2}๐‘”๐‘‘๐œˆ=โˆซ{1,2}๐‘“๐‘‘๐œˆ=๐‘“(1)๐œ‡({1})+๐‘“(2)๐œ‡({2})=๐‘“(1)+๐‘“(2)4

and

โˆซ{3,4}๐‘”๐‘‘๐œˆ=โˆซ{1,2}๐‘“๐‘‘๐œˆ=๐‘“(3)๐œ‡({3})+๐‘“(4)๐œ‡({4})=๐‘“(3)+๐‘“(4)4

while on the other hand

โˆซ{1,2}๐‘”๐‘‘๐œˆ=๐‘”(1)+๐‘”(2)4=๐‘Ž2,โˆซ{3,4}๐‘”๐‘‘๐œˆ=๐‘”(3)+๐‘”(4)4=๐‘2

Thus ๐‘” is defined by:

๐‘”(1)=๐‘”(2)=๐‘“(1)+๐‘“(2)2,๐‘”(3)=๐‘”(4)=๐‘“(3)+๐‘“(4)2

Thus what ๐‘” expressses: is the conditonal expectation of ๐‘“ on {1,2},{3,4}.
(Therefore it can be generalized: given any sub ๐œŽ-algebra โ„ฌ๏ธ€โŠ‚๐’œ๏ธ€, there exists a ๐œ‡|โ„ฌ๏ธ€-unique โ„ฌ๏ธ€ measurable function ๐‘”โˆˆ๐ฟ1(๐œ‡|โ„ฌ๏ธ€), that is the conditional expectation

๐‘”=๐”ผ[๐‘“โˆฃโ„ฌ๏ธ€]

s.t. for ๐ตโˆˆโ„ฌ๏ธ€,

โˆซ๐ต๐‘“๐‘‘๐œ‡=โˆซ๐ต๐”ผ[๐‘“โˆฃโ„ฌ๏ธ€]๐‘‘๐œ‡

it gives the average of ๐‘“ on sets in โ„ฌ๏ธ€.)

Nur fรผr Verrรผckte

(Itโ€™s really not necessary to attempt these problems. Do not, under any circumstances, hand them in!) To any measure space (๐‘‹,๐’œ๏ธ€) we can associate a new measure space (๐‘Œ,โ„ฌ๏ธ€), where ๐‘Œ is the Banach space of complex measures on (๐‘‹,๐’œ๏ธ€), and โ„ฌ๏ธ€ is the Borel ๐œŽ-algebra on ๐‘Œ.

  • Does this operation define a functor from the category of measurable spaces to itself. Is this functor (if well defined) full? Is it faithful? Is it essentially surjective?

  • Does the operation above admit any nontrivial fixed points (up to isomorphism)?

12 differentiation on real spaces

12.1 differentiation of regular Borel measures on โ„๐‘› [Fol 3.4, finished]

Definition 12.60 : regular Borel measure

ไธ€ไธช Borel measure ๐œˆ on โ„๐‘› ่ขซ็งฐไธบ regular ็š„, if ๐œˆ is locally finite (finite on every compact set).

Theorem 12.61 : regular Borel measure: ่•ดๅซไบ† regularity

ไธ€ไธช regular Borel measure ๐œˆ on โ„๐‘› ไธ€ๅฎšๆปก่ถณ:

  1. outer regularity:

    ๐œˆ(๐ธ)=inf{๐œˆ(๐‘ˆ)โˆฃ๐ธโŠ‚๐‘ˆ open}
  2. inner regularity:

    ๐œˆ(๐ธ)=inf{๐œˆ(๐‘ˆ)โˆฃ๐ธโŠ‚๐‘ˆ open}
Example 12.39

ไปปไฝ• LS measure on โ„ (restrict to Borel sets) ้ƒฝๆ˜ฏ regular measure. Lebesgue measure ๐‘š on โ„๐‘› (restrict to Borel sets) ๆ˜ฏ regular measure.

ๅฝ“็„ถ, ่ฟ™ไธชๆฆ‚ๅฟตไนŸๅฏไปฅๆŽจๅนฟ่‡ณ signed/coplex measure ไธŠ.

Definition 12.61 : regular signed/complex measure

ไธ€ไธช signed/complex measure ๐œˆ on โ„๐‘› ่ขซ็งฐไธบ regular measure, if |๐œˆ| is regular ็š„.

Lemma 12.42

ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(โ„๐‘›), ๅˆ™ ๐‘“๐‘‘๐‘š ๆ˜ฏไธ€ไธช regular measure. ๅฆ‚ๆžœ ๐‘“:โ„๐‘›โ†’โ„ ๆ˜ฏ extended-integrable ็š„, ๅˆ™

๐‘“โˆˆ๐ฟ๐‘™๐‘œ๐‘1(๐‘š)โ‡”๐‘“๐‘‘๐‘š is a regular measure
Proof

Folland p99. ่ฟ™ๆ˜พ็„ถ, ๅ› ไธบ ๐‘“ locally integrable ๅฐฑ่ฏดๆ˜Ž ๐‘“๐‘‘๐‘š ๆ˜ฏ locally finite ็š„.

โ–ก

Lemma 12.43

ๅฆ‚ๆžœ ๐œŒ,๐œ† ๆ˜ฏ signed/complex measure ๅนถไธ” ๐œŒโŠฅ๐œ†, ้‚ฃไนˆ

๐œŒ,๐œ† are regular measuresโ‡”๐œŒ+๐œ† is a regular measure
Proof

ๅœจ hw 10 ไธญ.
Note: STS it for positive measure, ่ฟ™ๆ˜ฏๅ› ไธบๅฏนไบŽ regular signed / complex measure ่€Œ่จ€,

๐œ†โŸ‚๐œŒโ‡”โˆƒ๐ดโˆˆ๐’œ๏ธ€ s.t. |๐œ†|(๐ด๐‘)=0 and |๐œŒ|(๐ด)=0โ‡”|๐œ†|โŸ‚|๐œŒ|

ๅนถไธ”ไปŽ่€Œ

๐œˆ=๐œ†+๐œŒ,๐œ†โŸ‚๐œŒโŸน|๐œˆ|=|๐œ†+๐œŒ|=|๐œ†|+|๐œŒ|

่ฟ™ไธ€ๅ‘ฝ้ข˜็š„่ฏๆ˜ŽไนŸๅœจ hw 10 ไธญ,

โ–ก

12.1.1 LDT meets LRNT: ไปปไฝ• regular Borel measure ๐œˆ on โ„๐‘› ๅฏนไบŽ ๐‘š ็š„ RN-derivative = relative density

Theorem 12.62 : LDT meets LRNT: computing RN derivative on โ„๐‘›

Let ๐œˆ be a regular Borel measure on โ„๐‘›, with LRN decomposition

๐œˆ=๐œ†+๐œŒ,๐‘‘๐œŒ=๐‘“๐‘‘๐‘š,๐œ†โŠฅ๐‘š

ๅณ

๐‘‘๐œˆ=๐‘‘๐œ†+๐‘“๐‘‘๐‘š

้‚ฃไนˆ: ๅฏนไบŽ ๐‘š-a.e. ๐‘ฅโˆˆโ„๐‘›, ้ƒฝๆœ‰:

lim๐‘Ÿโ†’0๐œˆ(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=๐‘“(๐‘ฅ)
Proof

็”ฑ ๐œˆ=๐œ†+๐œŒ ๅพ—ๅˆฐ:

๐œˆ(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=๐œ†(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))+๐œŒ(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

By LDT, ๆˆ‘ไปฌๆœ‰:

lim๐‘Ÿโ†’0๐œŒ(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=๐‘“(๐‘ฅ), for ๐‘š-a.e. ๐‘ฅ.

ไบŽๆ˜ฏ, ๅŽŸๅ‘ฝ้ข˜ๅณ่ฝฌๅŒ–ไธบ WTS:

lim๐‘Ÿโ†’0๐œ†(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=0, for ๐‘š-a.e. ๐‘ฅ.

(Notice: ่ฟ™้‡Œ 0 ไนŸๅฐฑๆ˜ฏ ๐‘‘๐œ†/๐‘‘๐‘š, ๅ› ไธบไธคไธช mutually singular ็š„ measure๏ผŒๅ…ถ RN derivative = 0 a.e.).
By lemma: ๅ› ไธบ ๐œˆ regular, ไธ” ๐œ†โŸ‚๐œŒ ,ๅฏไปฅๆŽจๅ‡บ: ๐œ†,๐œŒ ไนŸๆ˜ฏ regular ็š„.
WLOG ๆˆ‘ไปฌๅฏไปฅ suppose ๐œ† ๆ˜ฏ positive measure, ๅ› ไธบ ๐œ†โŸ‚๐‘šโ‡”|๐œ†|โŸ‚๐‘š, ๅนถไธ”|๐œ†(๐ธ)|โ‰ค|๐œ†|(๐ธ) for any ๐ธ. ๅ› ่€Œ ๐œ† ๆ˜ฏ positive measure ็š„ๆƒ…ๅ†ตไธญ่ฟ™ไธชๆž้™ไธบ 0 ไนŸ่‡ช็„ถๆŽจๅนฟๅˆฐ complex measure ไธŠ.
ๆณจๆ„: ็”ฑไบŽ ๐œ†โŸ‚๐‘š, ๆˆ‘ไปฌๅฏไปฅ้€‰ๅ– Borel set ๐ด such that

๐œ†(๐ด)=0,๐‘š(๐ด๐‘)=0

ไปŽ่€Œ: ๅช้œ€่ฆ่ฏๆ˜Ž lim๐‘Ÿโ†’0๐œ†(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=0 for a.e. ๐‘ฅโˆˆ๐ด ๅฐฑๅฏไปฅไบ†, ๅ› ไธบ ๐ด๐‘ ๆœฌ่บซไนŸๆ˜ฏ ๐‘š ็š„ null set.
ๆˆ‘ไปฌ set:

๐น๐‘˜:={๐‘ฅโˆˆ๐ด:lim๐‘Ÿโ†’0๐œ†(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))โ‰ฅ1๐‘˜}

ไปŽ่€Œ STS: ๅฏนไบŽไปปๆ„ ๐‘˜, ๐‘š(๐น๐‘˜)=0.
ๆˆ‘ไปฌ Fix ไธ€ไธช ๐‘˜, by ๐œ† ็š„ inner regularity, STS: ๅฏนไบŽไปปๆ„็š„ cpt ๐พโŠ‚๐น๐‘˜ compact, ้ƒฝๆœ‰ ๐‘š(๐พ)=0.
ไบŽๆ˜ฏๆˆ‘ไปฌ fix ไธ€ไธช compact set ๐พโŠ‚๐น๐‘˜, ๅนถ fix ๐œ–>0, STS: ๐‘š(๐พ)<๐œ–.
By ๐œ† ็š„ outer regularity, ๅญ˜ๅœจ ๐‘ˆ๐œ–โŠƒ๐ด open ไฝฟๅพ—

๐œ†(๐‘ˆ๐œ–)<๐œ–3๐‘›๐‘˜

By ๐น๐‘˜ ็š„ๅฎšไน‰, ๅฏนไบŽไปปๆ„็š„ ๐‘ฅโˆˆ๐น๐‘˜, ้ƒฝๅญ˜ๅœจๆŸไธช ๐‘Ÿ๐‘ฅ>0 ไฝฟๅพ—

๐œˆ(๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ))>1๐‘˜๐‘š(๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ))

Since

๐พโŠ‚โ‹ƒ๐‘ฅโˆˆ๐พ๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ)

ไปŽ่€Œ by finite open covering thm, ไธ€ๅฎšๅญ˜ๅœจๆŸไธช finite set ๐พโ€ฒ ไฝฟๅพ—

๐พโŠ‚โ‹ƒ๐‘ฅโˆˆ๐พโ€ฒ๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ)

ๆˆ‘ไปฌ recall VItali covering lemma: For given collection of balls {๐ต๐‘—โŠ‚โ„๐‘›}๐‘—=1๐‘˜, ๅญ˜ๅœจ disjoint subcollection {๐ต๐‘—1,โ‹ฏ,๐ต๐‘—๐‘š} ไฝฟๅพ—

โ‹ƒ๐‘—=1๐‘˜๐ต๐‘—โŠ‚โ‹ƒ๐‘–=1๐‘š(3๐ต๐‘—๐‘–)

ไปฃๅ…ฅ่ฟ™้‡Œ, ๅพ—ๅˆฐ: ๅญ˜ๅœจ ๐พโ€ณโŠ‚๐พโ€ฒ s.t. for all ๐‘ฅโˆˆ๐พโ€ณ, ๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ) ้ƒฝๆ˜ฏ disjoint ็š„, with

๐พโŠ‚โ‹ƒ๐‘ฅโˆˆ๐พโ€ณ3๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ)

ไบŽๆ˜ฏ

๐‘š(๐พ)โ‰คโˆ‘๐‘ฅโˆˆ๐พโ€ณ๐‘š(3๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ))=3๐‘›โˆ‘๐‘ฅโˆˆ๐พโ€ณ๐‘š(๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ))โ‰ค3๐‘›๐‘˜โˆ‘๐‘ฅโˆˆ๐พโ€ณ๐œ†(๐ต(๐‘ฅ,๐‘Ÿ๐‘ฅ))โ‰ค3๐‘›๐‘˜๐œ†(๐‘ˆ๐œ–)โ‰ค๐œ–

Since ๐œ– ไปปๆ„, ๐‘š(๐พ)=0.
Since ๐พ ไปปๆ„, ๐‘š(๐น๐‘˜)=0.
Since ๐‘˜ ไปปๆ„,

๐‘š({๐‘ฅโˆˆ๐ด:lim๐‘Ÿโ†’0๐œ†(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))>0})=0

ไปŽ่€Œๅพ—่ฏ.

โ–ก

12.1.2 Differentiation on โ„

12.1.3 {positive regular Borel measures on โ„}โ‰ƒ{distribution functions ๐น:โ„โ†’โ„}

Recall that: ๆฏไธช distribution function (้žไธฅๆ ผ increasing, right ctn function) ้ƒฝๅฏนๅบ”ไบ†ๅ”ฏไธ€็š„ไธ€ไธช regular Borel measures ๐œ‡๐น on โ„, ๅไน‹ไบฆ็„ถ.
็ป™ๅฎšไธ€ไธช regular Borel measures ๐œ‡๐น,

๐น๐œ‡(๐‘ฅ)โ‰”{๐œ‡((0,๐‘ฅ]),๐‘ฅโ‰ฅ0โˆ’๐œ‡((๐‘ฅ,0]),๐‘ฅ<0

ไธบๅฎƒ็š„ unique distribution function. ๅณ ๐œ‡((๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Ž), for all h-intervals.
่€Œ็ป™ๅฎšๅฏนไบŽ distribution function ๐น, ๆˆ‘ไปฌ define ๐œ‡0 by:

๐œ‡0(โ‹ƒ๐‘–=1๐‘›(๐‘Ž๐‘–,๐‘๐‘–])=โˆ‘๐‘–=1๐‘›(๐น(๐‘๐‘–)โˆ’๐น(๐‘Ž๐‘–))

็„ถๅŽ by Hahn-Kolmogrov, extend to a regular Borel measure ๐œ‡๐น, ไฝฟๅพ— ๐œ‡๐น((๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Ž) for any h-interval, i.e. ๐น ๆ˜ฏ ๐œ‡๐น ็š„ distribution function, ๅนถไธ” unique in the sense that ไปปๆ„ๅ…ถไป–็š„ such function ๐บ ๅฆ‚ๆžœไนŸๆ˜ฏ๐œ‡๐น ็š„ distribition function, ๅˆ™ๅฟ…็„ถๆœ‰ ๐นโˆ’๐บ ไธบ const. ไปŽ่€Œ

๐œ‡๐นโ†”๏ธŽ๐น

ไน‹้—ดๆž„ๆˆไบ†ไธ€ไธช measures ๅ’Œ functions ็š„็ฉบ้—ด็š„ bijection.

distribution function ๅ’Œ regular measure ไน‹้—ด็š„ๅฏนๅบ”ๅ…ณ็ณป, ๅ…ณ้”ฎ็”จๅค„ๅœจไบŽไป€ไนˆๅ‘ข? ๆˆ‘ไปฌ recall ๅˆšๅˆšๆ‰่ฏๆ˜Ž็š„ๅฎš็†, ไธ่ฟ‡ไฝฟ็”จไธ€ไธชๆ›ด general ็š„ version (can easily be extended from what we proved):

Theorem 12.63 : slightly more general version of LRNT meets LDT

Let ๐œˆ be a regular Borel measure on โ„๐‘›, with LRN decomposition

๐œˆ=๐œ†+๐œŒ,๐‘‘๐œŒ=๐‘“๐‘‘๐‘š,๐œ†โŠฅ๐‘š

้‚ฃไนˆ: ๅฏนไบŽ ๐‘š-a.e. ๐‘ฅโˆˆโ„๐‘›, ๅ–ไปปๆ„ nicely shrinking {๐ธ๐‘Ÿ} to ๐‘ฅ, ้ƒฝๆœ‰:

lim๐‘Ÿโ†’0๐œˆ(๐ธ๐‘Ÿ)๐‘š(๐ธ๐‘Ÿ)=๐‘“(๐‘ฅ)

ๅ› ่€Œๅฆ‚ๆžœๆˆ‘ไปฌๆœ‰ไธ€ไธช โ„ ไธŠ็š„ regualr measure ๐œ‡๐บ, ้‚ฃไนˆ่€ƒ่™‘ ๐ธ๐‘Ÿ:=(๐‘ฅ,๐‘ฅ+๐‘Ÿ], ๆˆ‘ไปฌๆœ‰:

๐œ‡๐บ(๐ธ๐‘Ÿ)๐‘š(๐ธ๐‘Ÿ)=๐บ(๐‘ฅ+๐‘Ÿ)โˆ’๐บ(๐‘ฅ)๐‘Ÿ

ไปŽ่€Œๆˆ‘ไปฌๅ‘็Žฐ for a.e. ๐‘ฅ, ้ƒฝๆœ‰:

lim๐‘Ÿโ†’0๐œ‡๐บ(๐ธ๐‘Ÿ)๐‘š(๐ธ๐‘Ÿ)=๐บโ€ฒ(๐‘ฅ)

ๆˆ‘ไปฌๅฏไปฅๅพ—ๅˆฐ: ่ฟ™ไธช regualr measure ๐œ‡๐บ ็›ธๅฏนไบŽ ๐‘š ็š„ LRN derivative, ๅฐฑ็ญ‰ไบŽๅฎƒ็š„ distribution function ็š„ derivative! ็”š่‡ณ, ๆˆ‘ไปฌๅฏไปฅ็”ฑๆญคๅˆคๆ–ญ: ๅฆ‚ๆžœ ๐บ:โ„โ†’โ„ ๆ˜ฏไธ€ไธช distribution function (increasing, right ctn), ้‚ฃไนˆๅฎƒ็š„ derivative ไธ€ๅฎšๆ˜ฏ a.e. ๅญ˜ๅœจ็š„! Since ๐œ‡๐บ regular โŸน ๐บ locally intble โŸน by LDT, ่ฟ™ไธช density limit ๆ˜ฏ a.e. ๅญ˜ๅœจ็š„.
่‡ณๆญค, ๆˆ‘ไปฌๅ‘็Žฐไบ† Monotone Differentiation Theorem.

12.1.4 Monotone Differentiation Theorem

Theorem 12.64 : Monotone Differentiation Theorem

ไปค ๐น:โ„โ†’โ„ ไธบไธ€ไธช increasing (nondecreasing) function, set:

๐บ(๐‘ฅ):=๐น(๐‘ฅ+)

ๅณ ๐น ็š„ๅณๆž้™ๅ‡ฝๆ•ฐ. (note: ๐บ ไธ€ๅฎšๆ˜ฏ increasing ไธ” right ctn ็š„, ๅ› ่€Œๆ˜ฏไธ€ไธช distribution function)
ๅˆ™ๆœ‰:

  • ๐ท๐นโ‰”{๐‘ฅ:๐น disctn at ๐‘ฅ} ๆ˜ฏ่‡ณๅคš ctbl ็š„ (ไปŽ่€Œไธ€ๅฎš zero measure)

  • ๐น,๐บ ้ƒฝ differentaitble ๐‘š-a.e., ๅนถไธ”

    ๐นโ€ฒ=๐บโ€ฒ a.e.
Proof

of (a): STS that, ๅฏนไบŽไปปๆ„ ๐‘š,๐‘›โˆˆโ„•,

๐‘๐‘š,๐‘›:={๐‘ฅโˆˆ[โˆ’๐‘š,๐‘š]:๐น(๐‘ฅ+)โˆ’๐น(๐‘ฅโˆ’)โ‰ฅ1๐‘›}

ๆ˜ฏไธ€ไธช finite set. ไปŽ่€Œ ๐ท๐น=โ‹ƒ๐‘š,๐‘›๐‘๐‘š,๐‘› ๆ˜ฏ at most ctbl ็š„.
่€Œ ๐‘๐‘š,๐‘› ็กฎๅฎžๆ˜ฏ finite ็š„, ๅ› ไธบ ๐น(โˆ’๐‘š)โˆ’๐น(๐‘š) ๆ˜ฏ bounded ็š„, ๆ‰€ไปฅ ๐‘๐‘š,๐‘› ไธ€ๅฎšๆ˜ฏ finite ็š„. (่‡ณๅคš็ปๅކ ๐น(โˆ’๐‘š)โˆ’๐น(๐‘š)/(1/๐‘›) ไธช่ฟ™ๆ ท็š„็‚น).

โ–ก

Proof

of (b): ้ฆ–ๅ…ˆๆˆ‘ไปฌ็Ÿฅ้“, ๐บ right ctn + increasing โŸน๐œ‡๐บ ๆ˜ฏไธ€ไธช LS (thus regular when restricted to Borel sets) measure on โ„.
Apply LDT to ๐œ‡๐บ, take ๐ธ๐‘Ÿ:=(๐‘ฅ,๐‘ฅ+๐‘Ÿ] as the shrinking family to ๐‘ฅ.
ไบŽๆ˜ฏ

๐œ‡๐บ(๐ธ๐‘Ÿ)๐‘š(๐ธ๐‘Ÿ)=๐บ(๐‘ฅ+๐‘Ÿ)โˆ’๐บ(๐‘ฅ)๐‘Ÿ

็”ฑ LDT โŸน ไธŠๅผ็š„ limit exist for a.e. ๐‘ฅ ( ๐œ‡๐บ w.r.t. ๐‘š ็š„ RN derivative), ๅฎƒๅฐฑๆ˜ฏ ๐บโ€ฒ, ไธ” ๐บโ€ฒ a.e. ๅญ˜ๅœจ, ็ญ‰ไบŽๅ…ถ induce ็š„ LS measure ็š„ RN derivaive.
็Žฐๅœจ remains to show: ๐น ็š„ derivative ไนŸ a.e. ๅญ˜ๅœจ, ๅนถไธ”ๅ’Œ ๐บ ็š„็›ธ็ญ‰. ๆˆ‘ไปฌ set:

๐ป:=๐บโˆ’๐น

ไปŽ่€Œ STS: ๐ปโ€ฒ a.e. ๅญ˜ๅœจไธ”ไธบ 0.
้ฆ–ๅ…ˆ, ๐ป>0 ๅนถไธ” ๐ปโ‰ 0 ๅชๆœ‰ๅฏ่ƒฝๅœจ discontinuous points (which is at most ctbl)ไธŠ. We set:

๐œ‡:=โˆ‘๐‘ฅโˆˆ๐ท๐น๐ป(๐‘ฅ)๐›ฟ๐‘ฅ

ไปŽ่€ŒๅฏนไบŽไปปๆ„ๅŒบ้—ด ๐ผ,

๐œ‡(๐ผ)=โˆ‘๐‘ฅโˆˆ๐ท๐นโˆฉ๐ผ๐ป(๐‘ฅ)

็”ฑไบŽ ๐น,๐บ locally intble, ่ฟ™ไธช ๐œ‡ ๆ˜ฏไธ€ไธช regular Borel measure.
ๅนถไธ”, ่ฟ™ไธช ๐œ‡ ็š„ null set ไธบ ๐ท๐‘“๐‘, ่€Œ ๐ท๐‘“, as we have proved, is at most ctbl, ๅ› ่€Œๆ˜ฏ ๐‘š ็š„ null set. ไปŽ่€Œๅพ—ๅˆฐ:

๐œ‡โŸ‚๐‘š

ไบŽๆ˜ฏ

๐‘‘๐œ‡๐‘‘๐‘š=0๐‘Ž.๐‘’

ไปŽ่€Œ็”ฑ LDT ๅพ—:

๐œ‡((๐‘ฅโˆ’๐‘Ÿ,๐‘ฅ+๐‘Ÿ))2๐‘Ÿโ†’๐‘Ÿโ†’00๐‘Ž.๐‘’.

ๅ› ่€ŒๅฏนไบŽไปปๆ„็š„ โ„Ž>0,

|๐ป(๐‘ฅ+โ„Ž)โˆ’๐ป(๐‘ฅ)โ„Ž|โ‰ค๐ป(๐‘ฅ+โ„Ž)+๐ป(๐‘ฅ)|โ„Ž|โ‰ค๐œ‡((๐‘ฅโˆ’2|โ„Ž|,๐‘ฅ+2|โ„Ž|))4|โ„Ž|โ†’โ„Žโ†’00

finishing the proof.

โ–ก

12.2 functions of bounded variation: ๐นโˆˆ๐ต๐‘‰ [Fol 3.5]

ไธŠไธ€่Š‚่ฏพๆˆ‘ไปฌ่ฏๆ˜Žไบ† Monotone Differentiation Theorem: ๅฎƒ่กจๆ˜Ž็š„ๆ˜ฏ, ไปปไฝ• โ„โ†’โ„ ็š„ non-decreasing function ้ƒฝๆ˜ฏ differentiable a.e. ็š„.
ๆˆ‘ไปฌ็Ÿฅ้“ไธ€ไธชๅ‡ฝๆ•ฐๅœจไธ€ๆ•ดไธชๅŒบ้—ดไธŠ differentiable ๅ…ถๅฎžๆ˜ฏไธ€ไธชๆฏ”ไปทไธฅๆ ผ็š„ๆกไปถ, ไฝ†ๆ˜ฏ differentiable a.e. ็š„ๆกไปถๅฐฑ็•ฅๅฅฝ่พพๅˆฐไธ€ไบ›.
Question: ๅฆ‚ๆžœไธ€ไธชๅ‡ฝๆ•ฐๅœจ [๐‘Ž,๐‘] ไธŠ differentiable a.e., ้‚ฃไนˆ in a.e. sense, ๅฏไปฅๅœจ [๐‘Ž,๐‘] ไธŠๅฎšไน‰ๅฎƒ็š„ derivative ๐นโ€ฒ. ้‚ฃไนˆ, ๆ˜ฏๅฆไธ€ๅฎšๆœ‰

๐น(๐‘)โˆ’๐น(๐‘Ž)=โˆซ๐‘Ž๐‘๐นโ€ฒ(๐‘ฅ)๐‘‘๐‘ฅ

ๅ‘ข? ็ญ”ๆกˆ่‚ฏๅฎšๆ˜ฏไธไธ€ๅฎš็š„. ไปฅไธ‹ๆ˜ฏไธ‰ไธชๅไพ‹: 1. Heaviside function; 2. Cantor function; 3.๐นโ€ฒ(๐‘ฅ)=0 a.e., but not 0 on a null set.
่ฟ™ไนŸๅพˆๆ˜พ็„ถ: ๅ› ไธบๅ•็‚น็š„ๅ€ผๆ˜ฏๆ— ๆณ•ๆŽงๅˆถ็š„. ๆˆ‘ไปฌๅช่ƒฝๆŽงๅˆถ in sense of a.e. , ๅ› ่€Œๆœ‰ outlier ็š„ ๐‘Ž,๐‘ ๆ˜ฏๅพˆๆญฃๅธธ็š„.
ๆˆ‘ไปฌไน‹ๅŽๅฐ† revisit ่ฟ™ไธ€้—ฎ้ข˜, ็ป™ๅ‡บ่ฟ™ไธช็ญ‰ๅผๆˆ็ซ‹็š„ condition.

ๆŽฅไธ‹ๆฅๆˆ‘ไปฌๅฐ†

12.2.1 total variation function ๐‘‡๐น of a function ๐น

Definition 12.62 : total variation function

็ป™ๅฎšไธ€ไธช function ๐น:โ„โ†’โ„‚, ๆˆ‘ไปฌๅฎšไน‰ๅฎƒ็š„ total variation function ๐‘‡๐น ไธบ:

๐‘‡๐น:โ„โ†’[0,โˆž]๐‘ฅโ†ฆsup{โˆ‘๐‘—=1๐‘›|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—+1)|:โˆ’โˆž<๐‘ฅ0<โ‹ฏ<๐‘ฅ๐‘›=๐‘ฅ}
Lemma 12.44

ๅฏนไบŽไปปๆ„็š„ ๐น:โ„โ†’โ„‚, ๐‘‡๐น ้ƒฝๆ˜ฏ increasing ็š„; ๅนถไธ”ๅฏนไบŽไปปๆ„ ๐‘Ž<๐‘, ๆœ‰:

๐‘‡๐น(๐‘)=๐‘‡๐น(๐‘Ž)+๐‘‡๐น(๐‘Ž;๐‘)

where

๐‘‡๐น(๐‘Ž;๐‘)=sup{โˆ‘๐‘—=1๐‘›|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—+1)|:๐‘=๐‘ฅ0<โ‹ฏ<๐‘ฅ๐‘›=๐‘Ž}

่กจ็คบ ๐น ็š„ๅฎšไน‰ๅŸŸ้™ๅˆถๅœจ [๐‘Ž,๐‘] ไธŠ็š„ total variation.

Proof

ๆ˜พ็„ถ, ็”ฑไบŽ total variation ๆ˜ฏ increasing ็š„, ๆˆ‘ไปฌๆ€ปๆ˜ฏๅฏไปฅ greedyly ้€‰ๆ‹ฉ partition. ๅฏนไบŽไธ€ไธช partition, ๆ€ปๆ˜ฏๅฏไปฅๆ’ๅ…ฅไธ€ไธชไธญ้—ด็‚นๆŠŠๅฎƒๅˆ†ๆˆไธคๅŠ, ่€Œ่ฟ™ไธคๅŠ็š„ sub partition ็š„ total variation ็š„ๅ’Œ โ‰ฅ ๅŽŸๅ…ˆ็š„ partition ็š„ total variation.

โ–ก

12.2.2 space of functions of bounded variation: ๐ต๐‘‰ ็š„ๅŸบๆœฌๆ€ง่ดจ

Definition 12.63 : function of bounded variation

ๅฆ‚ๆžœ ๐‘‡๐น(โˆž)<โˆž, ๆˆ‘ไปฌ็งฐ ๐น:โ„โ†’โ„‚ is of bounded variation ็š„, ๅ†™ไฝœ ๐นโˆˆ๐ต๐‘‰.

Definition 12.64 : function of bounded variation on an interval

ๅฆ‚ๆžœ ๐‘‡๐น(๐‘Ž;๐‘)<โˆž, ๆˆ‘ไปฌ็งฐ ๐น:โ„โ†’โ„‚ is of bounded variation on [๐‘Ž,๐‘], ๅ†™ๆˆ ๐นโˆˆ๐ต๐‘‰([๐‘Ž,๐‘]).

้ฆ–ๅ…ˆๆ˜พ็„ถ, ๐นโˆˆ๐ต๐‘‰ ๅฏไปฅ reduce to real-valued ็š„ๆƒ…ๅ†ตๆฅ่ฎจ่ฎบ.

Proposition 12.31
๐นโˆˆ๐ต๐‘‰โ‡”โ„œ๐‘“โˆˆ๐ต๐‘‰ and โ„‘๐‘“โˆˆ๐ต๐‘‰

12.2.3 ๐ต๐‘‰ as a vector space

Lemma 12.45 : ๐ต๐‘‰ ๆ˜ฏไธ€ไธช complex vector space

ๅฆ‚ๆžœ ๐น,๐บโˆˆ๐ต๐‘‰, ้‚ฃไนˆๅฏนไบŽไปปๆ„็š„ ๐‘Ž,๐‘โˆˆโ„‚, we have

๐‘‡๐‘Ž๐น+๐‘๐บโ‰ค|๐‘Ž|๐‘‡๐น+|๐‘|๐‘‡๐บ

ไปŽ่€Œ

๐‘Ž๐น+๐‘๐บโˆˆ๐ต๐‘‰
Proof

ๆ˜“ๅพ—. ๆ˜พ็„ถ, ๅ‡ฝๆ•ฐ็š„ total variation ๆ˜ฏ็บฟๆ€งๅฏๅŠ ็š„.

โ–ก

12.2.4 ๐นโˆˆ๐ต๐‘‰ ็š„ ๐‘‡๐น ็š„ limit behavior

ๆˆ‘ไปฌ็Ÿฅ้“๏ผŒ๐นโˆˆ๐ต๐‘‰ if ๐‘‡๐น(โˆž)<โˆž. ่€Œๅ…ณไบŽ ๐‘‡๐น(โˆ’โˆž), ๅŒๆ ทๆœ‰ๅผบ็ป“่ฎบ:

Proposition 12.32
๐นโˆˆ๐ต๐‘‰โŸน๐‘‡๐น(โˆ’โˆž)=0
Proof

Let ๐œ–>0.
ไปŽ่€ŒๅฏนไบŽไปปๆ„็š„ ๐‘ฅโˆˆโ„, since ๐นโˆˆ๐ต๐‘‰ ้‚ฃไนˆ ๐‘‡๐น bounded, ๐‘‡๐น(๐‘ฅ) ๆ˜ฏไธ€ไธช real number.
ๅ› ่€Œๆˆ‘ไปฌๅฏไปฅๆ‰พๅˆฐไธ€็ป„ partition points ๐‘ฅ0<โ‹ฏ<๐‘ฅ๐‘› ไฝฟๅพ—

โˆ‘1๐‘›|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|โ‰ฅ๐‘‡๐น(๐‘ฅ)โˆ’๐œ–

ไปŽ่€Œ

๐‘‡๐น(๐‘ฅ)โˆ’๐‘‡๐น(๐‘ฅ0)โ‰ฅ๐‘‡๐น(๐‘ฅ)โˆ’๐œ–

ไปŽ่€Œ

๐‘‡๐น(๐‘ฆ)โ‰ค๐œ–,โˆ€๐‘ฆโ‰ค๐‘ฅ0

Since ๐œ–>0 arbitrary, ่ฟ™่ฏๆ˜Žไบ† ๐‘‡๐น(โˆ’โˆž)=0

โ–ก

Lemma 12.46 : ๐นโˆˆ๐ต๐‘‰ right ctn โŸน๐‘‡๐น right ctn

๐นโˆˆ๐ต๐‘‰ right ctn โŸน๐‘‡๐น ไนŸ right ctn

Proof

Let ๐‘ฅโˆˆโ„,๐œ–>0.
Let

๐›ผ:=๐‘‡๐น(๐‘ฅ+)โˆ’๐‘‡๐น(๐‘ฅ)

WTS: ๐›ผ=0.
By right ctnity of ๐น ๅ’Œ ๐‘‡๐น increasing, ๆˆ‘ไปฌๅฏไปฅ้€‰ๆ‹ฉ ๐›ฟ>0, ๅŒๆ—ถๆปก่ถณ: |๐น(๐‘ฅ+โ„Ž)โˆ’๐น(๐‘ฅ)|<๐œ–, ๐‘‡๐น(๐‘ฅ+โ„Ž)โˆ’๐‘‡๐น(๐‘ฅ+)<๐œ– whenever 0<โ„Ž<๐›ฟ.
Fix ไธ€ไธชๆปก่ถณ 0<โ„Ž<๐›ฟ ็š„ โ„Ž. ๅ…ถๅŽ็š„่ฏๆ˜Ž่ง Folland 104.

โ–ก

12.2.5 ๅฑžไบŽ ๐ต๐‘‰,๐ต๐‘‰(๐ผ) ็š„ๅ‡ฝๆ•ฐ

Lemma 12.47 : ๅ“ชไบ›ๅ‡ฝๆ•ฐไธ€ๅฎš ๐ต๐‘‰ or ๐ต๐‘‰(๐ผ)
  1. ๅฆ‚ๆžœ ๐น:โ„โ†’โ„ bounded ไธ” increasing, ้‚ฃไนˆ ๐นโˆˆ๐ต๐‘‰ ไธ” ๐‘‡๐น(๐‘ฅ)=๐น(๐‘ฅ)โˆ’๐น(โˆ’โˆž).

  2. ๅฆ‚ๆžœ ๐น:โ„โ†’โ„ ๆ˜ฏ Lipschitz countinuous ็š„, ้‚ฃไนˆ๐นโˆˆ๐ต๐‘‰(๐ผ) for ไปปๆ„็š„ cpt interval ๐ผ

  3. ๅฆ‚ๆžœ ๐น:โ„โ†’โ„ ๆ˜ฏ differentiable ไธ” ๐นโ€ฒ bounded ็š„, ้‚ฃไนˆ ๐นโˆˆ๐ต๐‘‰(๐ผ) for ไปปๆ„็š„ cpt interval ๐ผ

Proof

(1) trivial.
(2) by def: ่€ƒ่™‘ Lipschitz const ๐‘€, ๅˆ™ ๐‘‡๐น(๐‘Ž;๐‘)โ‰ค๐‘€(๐‘โˆ’๐‘Ž).
(3): ่ฟ™ๆ˜ฏ (2) ็š„ๆŽจ่ฎบ, ๅ› ไธบ recall: by MCT ๅฏๅพ—: ๐น:โ„โ†’โ„ ๆ˜ฏ differentiable ไธ” ๐นโ€ฒ bounded โŸน๐น Lipstchiz ctn.

โ–ก

Proposition 12.33

ไปฅไธ‹ๆ˜ฏไธ€ไบ›็ปๅ…ธ็š„ๅ‡ฝๆ•ฐ็š„ variational behavior:

  1. ๐‘“(๐‘ฅ)=sin(๐‘ฅ): ๅฑžไบŽ ๐ต๐‘‰(๐ผ) for ไปปๆ„ cpt ๐ผ, ไฝ†ไธๅฑžไบŽ ๐ต๐‘‰.

  2. ๐‘“(๐‘ฅ)=๐‘ฅsin1๐‘ฅ,๐‘“(0)=0: ๅฑžไบŽ๐ต๐‘‰(๐ผ) iff 0โˆ‰๐ผ.

  3. ๐‘“(๐‘ฅ)=๐‘ฅ2sin1๐‘ฅ2,๐‘“(0)=0: ๅฑžไบŽ๐ต๐‘‰(๐ผ) iff 0โˆ‰๐ผ.

Proof

(1) ๆ˜พ็„ถ; (2),(3) ่ง HW 11. ๅ…ถๅฎžๅฎƒไปฌๅŸบๆœฌ็›ธๅŒ. (โŸน): if 0โˆ‰๐ผ then ๐นโˆˆ๐ต๐‘‰(๐ผ). ๆ˜ฏ็ฎ€ๅ•็š„, we differentiate ๐น(๐‘ฅ)=๐‘ฅsin(1/๐‘ฅ) for ๐‘ฅโ‰ 0:

๐นโ€ฒ(๐‘ฅ)=๐‘‘๐‘‘๐‘ฅ(๐‘ฅโ‹…sin(1๐‘ฅ))=sin(1๐‘ฅ)+๐‘ฅโ‹…cos(1๐‘ฅ)โ‹…(โˆ’1๐‘ฅ2)=sin(1๐‘ฅ)โˆ’1๐‘ฅcos(1๐‘ฅ)

ๅœจไธๅซ 0 ็š„ๅŒบ้—ดไธŠ, ๅฎƒๆ˜ฏ bounded ็š„. ไบŽๆ˜ฏ by lemma ๅพ—่ฏ.
(โŸธ): if ๐นโˆˆ๐ต๐‘‰(๐ผ) then 0โˆ‰๐ผ. This is equiv to: if 0โˆˆ๐ผ then ๐นโˆ‰๐ต๐‘‰(๐ผ).
Suppose 0โˆˆ๐ผ=[๐‘Ž,๐‘] then ๐‘Žโ‰ค0 and ๐‘โ‰ฅ0, one of which is strict. WLOG we suppose ๐‘>0.
ๆˆ‘ไปฌ็š„ idea ๆ˜ฏ harmonic series. ่€ƒ่™‘

๐‘ฆ๐‘›โ‰”1๐‘›๐œ‹+๐œ‹/2โ†’0+

we have:

๐น(๐‘ฆ๐‘›)=๐‘ฆ๐‘›sin(1๐‘ฆ๐‘›)=1๐‘›๐œ‹+๐œ‹/2โ‹…sin(๐‘›๐œ‹+๐œ‹/2)

For odd ๐‘›, ๐น(๐‘ฆ๐‘›)=โˆ’1๐‘›๐œ‹+๐œ‹/2, for even ๐‘›, ๐น(๐‘ฆ๐‘›)=1๐‘›๐œ‹+๐œ‹/2. Since ๐‘>0, for some ๐‘0 we have ๐‘ฆ๐‘0<๐‘. Then we consider the partition: pick ๐‘โˆˆโ„•, and use ๐‘ฅ0=0,๐‘ฅ1=๐‘ฆ๐‘0+๐‘โˆ’1,๐‘ฅ2=๐‘ฆ๐‘0+๐‘โˆ’2,โ‹ฏ,๐‘ฅ๐‘=๐‘ฆ๐‘0,๐‘ฅ๐‘+1=๐‘ as the partition points of [0,๐‘].
Then we have

โˆ‘๐‘›=1๐‘+1|๐น(๐‘ฅ๐‘›)โˆ’๐น(๐‘ฅ๐‘›โˆ’1)|โ‰ฅโˆ‘๐‘›=๐‘0๐‘0โˆ’2+๐‘1๐œ‹๐‘›+๐œ‹/2+1๐œ‹(๐‘›+1)+๐œ‹/2โ‰ฅ2โˆ‘๐‘›=๐‘0๐‘0โˆ’2+๐‘1๐œ‹๐‘›+๐œ‹/2

As ๐‘โ†’โˆž, this sum โˆ‘๐‘›=1๐‘+2|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|โ†’โˆž, by the harmonic series.

โ–ก

12.2.6 Jordan decomposition for ๐‘“โˆˆ๐ต๐‘‰: ๐‘“=12(๐‘‡๐น+๐น)โˆ’12(๐‘‡๐นโˆ’๐น)

Lemma 12.48

ๅฆ‚ๆžœ real-valued ๐นโˆˆ๐ต๐‘‰, ้‚ฃไนˆ ๐‘‡๐น+๐น,๐‘‡๐นโˆ’๐น ้ƒฝๆ˜ฏ increasing ็š„.

Proof

ไปปๅ– ๐‘ฅ<๐‘ฆ.
Let ๐œ–>0.
Can find ๐‘ฅ0<๐‘ฅ1<โ‹ฏ<๐‘ฅ๐‘=๐‘ฅ, s.t.

โˆ‘1๐‘|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|โ‰ฅ๐‘‡๐น(๐‘ฅ)โˆ’๐œ–

ไปŽ่€Œ

๐‘‡๐น(๐‘ฆ)โ‰ฅโˆ‘1๐‘|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|+|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|โ‰ฅ๐‘‡๐น(๐‘ฅ)โˆ’๐œ–+|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|

็”ฑไบŽ ๐œ–>0 ไปปๆ„, ๅฏไปฅๅพ—ๅˆฐ:

๐‘‡๐น(๐‘ฆ)โˆ’๐‘‡๐น(๐‘ฅ)โ‰ฅ|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|

ๅ› ่€Œ:

(๐‘‡๐น(๐‘ฆ)โˆ’๐น(๐‘ฆ))โˆ’(๐‘‡๐น(๐‘ฅ)โˆ’๐น(๐‘ฅ))โ‰ฅ|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|โˆ’(๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ))โ‰ฅ0

โ–ก

Theorem 12.65 : Jordan decomposition for ๐นโˆˆ๐ต๐‘‰

ๅฏนไบŽ ๐น:โ„โ†’โ„ (ๆณจๆ„ๆ˜ฏ real-valued):

๐นโˆˆ๐ต๐‘‰โ‡”๐น ็ญ‰ไบŽไธคไธช bounded increasing functions ็š„ๅทฎ

Specially,

๐นโˆˆ๐ต๐‘‰โ‡”๐‘‡๐นยฑ๐น bounded

ๅ› ่€Œ for ๐นโˆˆ๐ต๐‘‰, ๆˆ‘ไปฌๆ€ปๆ˜ฏๅฏไปฅๆŠŠๅฎƒๅ†™ไฝœ

๐น=12(๐‘‡๐น+๐น)โˆ’12(๐‘‡๐นโˆ’๐น)

where we call it as the Jordan decomposition of ๐นโˆˆ๐ต๐‘‰. ๅ…ถไธญ, 12(๐‘‡๐น+๐น) ่ขซ็งฐไธบ ๐น ็š„ positive variation; 12(๐‘‡๐นโˆ’๐น) ่ขซ็งฐไธบ ๐น ็š„ negative variation.

Proof

ๆ˜พ็„ถ, ๐นโˆˆ๐ต๐‘‰โŸน๐น bdd, ๅ› ไธบ

|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|โ‰ค๐‘‡๐น(โˆž)โˆ’๐‘‡๐น(โˆ’โˆž)

For ๐นโˆˆ๐ต๐‘‰, we have ๐‘‡๐น(โˆž)<โˆž,๐‘‡๐น(โˆ’โˆž)=0.
ๅˆ ๐นโˆˆ๐ต๐‘‰โŸน๐‘‡๐น bdd by def, ๆˆ‘ไปฌๅพ—ๅˆฐ:

๐นโˆˆ๐ต๐‘‰โŸน๐‘‡๐นยฑ๐น bounded

่€Œๅๅ‘ trivial (bounded function ็š„ๅทฎไป็„ถ bounded).

โ–ก

12.2.7 corollaries of Jordan decomposition

Corollary 12.28

Let ๐นโˆˆ๐ต๐‘‰. By Jordan decomposition, ๐น ็ญ‰ไบŽไธคไธช bounded increasing functions ็š„ๅทฎ.ไปŽ่€Œๆˆ‘ไปฌ by MDT ๅพ—:

  • ๐น(๐‘ฅ+),๐น(๐‘ฅโˆ’) ๅญ˜ๅœจ for all ๐‘ฅ; ๐น(ยฑโˆž) ไนŸๅญ˜ๅœจ.

  • ๐ท๐น={๐‘ฅ:๐น disctn at ๐‘ฅ} ๆ˜ฏ at most ctbl ็š„.

  • ๅฎšไน‰ ๐บ(๐‘ฅ):=๐น(๐‘ฅ+), ๅˆ™ ๐น,๐บ ้ƒฝ a.e. differentiable ไธ” ๐นโ€ฒ=๐บโ€ฒ ๐‘š-a.e.

่ฟ™้‡Œๆœ€้‡่ฆ็š„ๆ˜ฏ:

๐นโˆˆ๐ต๐‘‰โŸน๐น a.e. differentiable

12.3 ๐‘๐ต๐‘‰๐ด๐ถ ็ฉบ้—ด, ไปฅๅŠๅ…ถไธŠ็š„ FTC for Lebesgue integral [Fol 3.5, finished]

12.3.1 ๐‘๐ต๐‘‰ ๅŠๅ…ถๆ€ง่ดจ

Definition 12.65 : NBV

For ๐น:โ„โ†’โ„‚, ๆˆ‘ไปฌๅฎšไน‰: ๐นโˆˆ๐‘๐ต๐‘‰, if ๐นโˆˆ๐ต๐‘‰ ไธ” ๐น right ctn, ๐น(โˆ’โˆž)=0.

Proposition 12.34

๐‘๐ต๐‘‰โŠ‚๐ต๐‘‰ ๆ˜ฏไธ€ไธช linear subspace.

ๆˆ‘ไปฌๅทฒ็ป็Ÿฅ้“:

{positive regular Borel measures on โ„}โ‰ƒ{distribution functions ๐น:โ„โ†’โ„}

Notice: ่ฟ™ไธ€็‚น is achieved by

๐น๐œ‡(๐‘ฅ)โ‰”{๐œ‡((0,๐‘ฅ]),๐‘ฅโ‰ฅ0โˆ’๐œ‡((๐‘ฅ,0]),๐‘ฅ<0

็ญ‰ๅŒไบŽ:

๐œ‡๐น((๐‘Ž,๐‘])=๐น(๐‘)โˆ’๐น(๐‘Ž)

ๆณจๆ„: positive regular Borel measure ๅ’Œ distribution functions ้ƒฝๆœ‰ไธ€ไธชๅ…ฑๅŒ็‚น: ๅฎƒไปฌๅœจ bounded set ไธŠๆ˜ฏ bounded ็š„, ไฝ†ๆ˜ฏๆ•ดไฝ“ๅฏไปฅ unbounded.
่€Œ็Žฐๅœจๆˆ‘ไปฌ่ฏๆ˜Ž:

12.3.2 {complex Borel measures on โ„}โ‰ƒ๐‘๐ต๐‘‰

ๆณจๆ„: ไธ€ไธช complex Borel measure ๅ’Œไธ€ไธช ๐นโˆˆ๐‘๐ต๐‘‰ ้ƒฝๆ˜ฏ finite ็š„.

Theorem 12.66 : {complex Borel measures on โ„}โ‰ƒ๐‘๐ต๐‘‰
  1. ๅฏนไบŽ โ„ ไธŠ็š„ complex measure ๐œ‡, defining

    ๐น(๐‘ฅ):=๐œ‡((โˆ’โˆž,๐‘ฅ])

    ๅˆ™ๆœ‰:

    ๐นโˆˆ๐‘๐ต๐‘‰
  2. ๅฏนไบŽ ๐นโˆˆ๐‘๐ต๐‘‰, ไธ€ๅฎšๅญ˜ๅœจๆŸไธช unique complex measure ๐œ‡๐น on โ„, ไฝฟๅพ—

    ๐œ‡((โˆ’โˆž,๐‘ฅ])=๐น(๐‘ฅ)โˆ€๐‘ฅ
Proof

(1): ๅฝ“ ๐œ‡ ๆ˜ฏ positive ็š„ๆƒ…ๅ†ตไธ‹, ๆ˜ฏๆ˜พ็„ถ็š„. complex ็š„ๆƒ…ๅ†ตๅฐฑๆ˜ฏ re/im ้ƒจๅˆ†ๅˆ†ๅˆซๅ ๅŠ ๅณๅฏ.
(2): ๅŒๆ ท, WLOG ๆˆ‘ไปฌๅฏไปฅๅ‡่ฎพ ๐น ๆ˜ฏ real-valued ็š„.
๐นโˆˆ๐‘๐ต๐‘‰โŸน๐น=๐น+โˆ’๐นโˆ’, ่ฟ™ไธคไธช้ƒฝๆ˜ฏ bounded increasing functions ไธ” NBV, ไปŽ่€Œๅญ˜ๅœจไธคไธช finite signed measure ๆปก่ถณ:

๐œ‡ยฑ((โˆ’โˆž,๐‘ฅ])=๐นยฑ(๐‘ฅ)โˆ’๐นยฑ(โˆ’โˆž)

ๅ†ๅฎšไน‰

๐œ‡:=๐œ‡+โˆ’๐œ‡โˆ’

ๅณๅฏ

โ–ก

Theorem 12.67 : ๐œ‡๐น ็š„ total variation measure = ๐œ‡๐‘‡๐น

ๅฏนไบŽไปปๆ„็š„ ๐นโˆˆ๐‘๐ต๐‘‰, ๆˆ‘ไปฌๆœ‰:

|๐œ‡๐น|=๐œ‡๐‘‡๐น

Specially when ๐น ๆ˜ฏ real-valued ๆƒ…ๅ†ตไธ‹, ้‚ฃไนˆ ๐œ‡๐น ๆ˜ฏไธ€ไธช finite positive measure, ไธ”ๆœ‰

๐œ‡ยฑ=๐œ‡๐นยฑ

Now: Given ๐นโˆˆ๐‘๐ต๐‘‰ with associated c.m. ๐œ‡๐น, ไป€ไนˆๆ—ถๅ€™ ๐œ‡๐นโŸ‚๐‘š, ไป€ไนˆๆ—ถๅ€™ ๐œ‡๐นโ‰ช๐‘š?

Theorem 12.68 : characterization of ๐œ‡๐นโŸ‚๐‘š ๅ’Œ ๐œ‡๐นโ‰ช๐‘š, for ๐นโˆˆ๐‘๐ต๐‘‰

ๅฏนไบŽ ๐นโˆˆ๐‘๐ต๐‘‰, ๆˆ‘ไปฌๅทฒ็ป็Ÿฅ้“: ๐นโ€ฒ ๐‘š-a.e. ๅญ˜ๅœจ, ไธ” ๐นโ€ฒโˆˆ๐ฟ1(๐‘š).
Now we claim, ๆœ‰:

๐œ‡๐นโŸ‚๐‘šโ‡”๐นโ€ฒ=0๐‘š-a.e.

ไปฅๅŠ

๐œ‡๐นโ‰ช๐‘šโ‡”๐น(๐‘ฅ)=โˆซโˆ’โˆž๐‘ฅ๐นโ€ฒ(๐‘ก)๐‘‘๐‘กโˆ€๐‘ฅ
Proof

Let ๐‘ฅโˆˆโ„. Applying LDT and LRNT, with ๐ธ๐‘Ÿ:=(๐‘ฅ,๐‘ฅ+๐‘Ÿ]:

lim๐‘Ÿโ†’0๐œ‡๐น(๐ธ๐‘Ÿ)๐‘š(๐ธ๐‘Ÿ)=lim๐‘Ÿโ†’0๐น(๐‘ฅ+๐‘Ÿ)โˆ’๐น(๐‘ฅ)๐‘Ÿ=๐นโ€ฒ

ๅ› ่€Œ ๐นโ€ฒ ๅฐฑๆ˜ฏ่ฟ™ไธช RN derivative. ๅฏนไบŽ

๐œ‡๐น=๐œ†+๐œŒ

where ๐œ†โŸ‚๐‘š,๐œŒโ‰ช๐‘š, ๆˆ‘ไปฌๆœ‰:

๐น(๐‘ฅ):=๐œ‡๐น((โˆ’โˆž,๐‘ฅ])=๐œ†((โˆ’โˆž,๐‘ฅ])+๐œŒ((โˆ’โˆž,๐‘ฅ])

ๆˆ‘ไปฌ็Ÿฅ้“, ๐œ‡๐นโŸ‚๐‘šโ‡”๐œŒ=0, ไปŽ่€Œ by LDT meets LRNT, we know that

๐นโ€ฒ=0๐‘Ž.๐‘’.

่€Œ ๐œ‡๐นโ‰ช๐‘šโ‡”๐œ†=0, ไปŽ่€Œ็›ดๆŽฅ :

๐น(๐‘ฅ):=๐œŒ((โˆ’โˆž,๐‘ฅ])=๐นโ€ฒ๐‘‘๐‘š((โˆ’โˆž,๐‘ฅ])=โˆซโˆ’โˆž๐‘ฅ๐นโ€ฒ(๐‘ก)๐‘‘๐‘ก

โ–ก

12.3.3 ๐ด๐ถ ๅŠๅ…ถๆ€ง่ดจ

Definition 12.66 : absolutely continuous function

ๆˆ‘ไปฌๅฎšไน‰ ๐น:โ„โ†’โ„‚ ๆ˜ฏ absolutely ctn ็š„, if ๅฏนไบŽไปปๆ„ ๐œ–>0 ้ƒฝๅญ˜ๅœจ ๐›ฟ>0 ไฝฟๅพ—ๅฏนไบŽไปปๆ„็š„ disjoint intervals (๐‘Ž1,๐‘1),โ‹ฏ,(๐‘Ž๐‘,๐‘๐‘), ้ƒฝๆœ‰:

โˆ‘1๐‘|๐น(๐‘๐‘—)โˆ’๐น(๐‘Ž๐‘—)|<๐œ–wheneverโˆ‘1๐‘|๐‘๐‘—โˆ’๐‘Ž๐‘—|<๐›ฟ
Definition 12.67 : absolutely continuous function on a cpt interval

ๆˆ‘ไปฌๅฎšไน‰ ๐น:๐ผโ†’โ„‚ ๆ˜ฏ absolutely ctn ็š„, if ๅฏนไบŽไปปๆ„ ๐œ–>0 ้ƒฝๅญ˜ๅœจ ๐›ฟ>0 ไฝฟๅพ—ๅฏนไบŽไปปๆ„็š„ disjoint intervals (๐‘Ž1,๐‘1),โ‹ฏ,(๐‘Ž๐‘,๐‘๐‘)โŠ‚๐ผ, ้ƒฝๆœ‰:

โˆ‘1๐‘|๐น(๐‘๐‘—)โˆ’๐น(๐‘Ž๐‘—)|<๐œ–wheneverโˆ‘1๐‘|๐‘๐‘—โˆ’๐‘Ž๐‘—|<๐›ฟ
Lemma 12.49 : ๐นโˆˆ๐‘๐ต๐‘‰ abs ctn โ‡” ๐œ‡๐นโ‰ช๐‘š

ๅฏนไบŽ ๐นโˆˆ๐‘๐ต๐‘‰,

๐นโˆˆ๐ด๐ถโ‡”๐œ‡๐นโ‰ช๐‘š
Proof

ๆˆ‘ไปฌ recall, abs ctn ้™คไบ† "๐‘š ็š„ nullsets ไนŸไธ€ๅฎšๆ˜ฏ ๐œ‡๐น ็š„ null sets" ไน‹ๅค–, ่ฟ˜ๆœ‰ๅฆไธ€ไธช characterization: ๐œ‡๐นโ‰ช๐‘š ๅฝ“ไธ”ไป…ๅฝ“ๅฏนไบŽไปปๆ„ ๐œ–>0 ้ƒฝๅญ˜ๅœจ ๐›ฟ>0 ไฝฟๅพ— ๐‘š(๐ธ)<๐›ฟโŸน|๐œ‡๐น(๐ธ)|<๐œ–.
ๆ˜พ็„ถ, ่ฟ™ไธช characterization ๅ’Œ่ฟ™้‡Œ็š„ๅ‘ฝ้ข˜ๆœ‰ๅ…ณ. ๆˆ‘ไปฌๅ‘็Žฐ, ๐œ‡๐นโ‰ช๐‘šโŸน๐นโˆˆ๐ด๐ถ ็›ดๆŽฅ naturally follows from ่ฟ™ไธช form. Let ๐œ–>0, ๅญ˜ๅœจ ๐›ฟ ไฝฟๅพ— ๐‘š(๐ธ)<๐›ฟโŸน|๐œ‡๐น(๐ธ)|<๐œ–. ้‚ฃไนˆ่€ƒ่™‘ ๐ธ=โจ†1๐‘(๐‘Ž๐‘—,๐‘๐‘—) with ๐‘š(๐ธ)<๐›ฟ, ็›ดๆŽฅๆœ‰

|๐œ‡๐น(๐ธ)|=โˆ‘1๐‘|๐น(๐‘๐‘—)โˆ’๐น(๐‘Ž๐‘—)|<๐œ–

ไปŽ่€Œๅพ—่ฏ.
่€Œๅๅ‘, ๆˆ‘ไปฌ่€ƒ่™‘ ๐‘š(๐ธ)=0, ๅนถๅˆฉ็”จ outer regularity ๅ–ไธ€ไธช้€ผ่ฟ‘ๅฎƒ็š„ open set (ๆฏไธชๆ˜ฏ union of finite disjoint open intervals) seq, ้€ผ่ฟ‘ ๐ธ, with ๐‘š(๐‘ˆ1)<๐›ฟ. ็”ฑ ๐นโˆˆ๐ด๐ถ ๅฏไปฅๅพ—ๅˆฐ

๐œ‡๐น(๐‘ˆ๐‘—)โ‰ค๐œ‡๐น(๐‘ˆ1)<๐œ–

for all ๐‘—, ไปŽ่€Œ ๐œ‡๐น(๐ธ)โ‰ค๐œ–. ไปŽ่€Œๅพ—่ฏ, since ๐œ– arbitrary.

โ–ก

12.3.4 FTC for Lebesgue integral on โ„: requires ๐‘๐ต๐‘‰+๐ด๐ถ

Corollary 12.29

ๅฆ‚ๆžœ ๐‘“โˆˆ๐ฟ1(๐‘š), ้‚ฃไนˆ

๐น(๐‘ฅ):=โˆซโˆ’โˆž๐‘ฅ๐‘“(๐‘ก)๐‘‘๐‘กโˆˆ๐‘๐ต๐‘‰โˆฉ๐ด๐ถ,๐‘“=๐นโ€ฒ๐‘Ž.๐‘’.

Conversely, ๅฆ‚ๆžœ ๐นโˆˆ๐‘๐ต๐‘‰โˆฉ๐ด๐ถ, ้‚ฃไนˆ

๐นโ€ฒโˆˆ๐ฟ1(๐‘š),๐น(๐‘ฅ)=โˆซโˆ’โˆž๐‘ฅ๐นโ€ฒ(๐‘ก)๐‘‘๐‘ก

12.3.5 FTC for Lebesgue integral on a cpt interval: ๅช้œ€่ฆ ๐ด๐ถ

ๆฏ”่ตทๅˆšๆ‰็š„ FTC-I, FTC-II ็š„ๆกไปถ่ฆๅฎฝๆพๅพˆๅคš, ๅช้œ€่ฆ ๐น ๅœจๅฎƒ้œ€่ฆ่ขซ็”จๅˆฐ็š„ compact interval ไธŠ AC ๅณๅฏไปฅ. ่ฟ™ๆ˜ฏๅ› ไธบ, ๆˆ‘ไปฌไธ้œ€่ฆ็”จๅˆฐ NBV ๅช้œ€่ฆ BV, ๅนถไธ”ๅœจ cpt interval ไธŠ, AC ๆœฌ่บซๅฐฑๅฏไปฅๆŽจๅ‡บ BV.

Lemma 12.50

ๅฆ‚ๆžœ ๐นโˆˆ๐ด๐ถ([๐‘Ž,๐‘]), ้‚ฃไนˆ ๐นโˆˆ๐ต๐‘‰([๐‘Ž,๐‘]).
ๅณ

๐ด๐ถ([๐‘Ž,๐‘])โŠ‚๐ต๐‘‰([๐‘Ž,๐‘])
Proof

่ฟ™ๆ˜ฏๆ˜พ็„ถ็š„. ๆˆ‘ไปฌ็œ‹ๅˆฐ ๐นโˆˆ๐ด๐ถ([๐‘Ž,๐‘]) ็š„ๅฎšไน‰: on [๐‘Ž,๐‘] we have:

โˆ‘1๐‘|๐น(๐‘๐‘—)โˆ’๐น(๐‘Ž๐‘—)|<๐œ–wheneverโˆ‘1๐‘|๐‘๐‘—โˆ’๐‘Ž๐‘—|<๐›ฟ

ๆˆ‘ไปฌไธๅฆจ่€ƒ่™‘ ๐œ–=1, ็„ถๅŽๅฏไปฅๅˆ’ๅˆ† [๐‘Ž,๐‘] into ไธ€ไธชไธชๆ€ป้•ฟๅบฆไธบ ๐›ฟ2 ็š„็”ฑ disjoint intervals ๆž„ๆˆ็š„ๅ— (่กฅ็ฉบ็ผบๆฒกไบ‹), ไปŽ่€Œ Bound ไฝ่ฟ™ไธชๅˆ’ๅˆ†ไธŠ็š„ variation by parition โˆ‘1๐‘0|๐น(๐‘๐‘—)โˆ’๐น(๐‘Ž๐‘—)|. ่€Œๆˆ‘ไปฌๅ‘็Žฐ: ่ฟ™ไธชๆ—ถๅ€™ๆˆ‘ไปฌไธ่ฎบๆ€Žไนˆ fine ่ฟ™ไธชๅˆ’ๅˆ†, ๆฏไธชๅ—็š„ๆ€ป้•ฟๅบฆๆ€ปๅฝ’ๆ˜ฏไธๅ˜็š„, ไปŽ่€Œไป็„ถๅฏไปฅไฝฟ็”จๅŽŸๅ…ˆ็š„ bound.
ๆˆ‘ไปฌ recall: for total variation, partition ็š„้€‰ๅ–ๆ˜ฏ greddy ็š„. ไปŽ่€Œ่ฟ™ๅฐฑ่ถณไปฅๅพ—่ฏ.

โ–ก

Theorem 12.69 : FTC-II for Lebesgue integral on a cpt interval

TFAE:

  • ๐นโˆˆ๐ด๐ถ[๐‘Ž,๐‘]

  • ๐น ๆ˜ฏ diffble a.e. on [๐‘Ž,๐‘] ็š„, ไธ” ๐นโ€ฒโˆˆ๐ฟ1([๐‘Ž,๐‘],๐‘š), ไธ”

    ๐น(๐‘ฅ)โˆ’๐น(๐‘Ž)=โˆซ๐‘Ž๐‘ฅ๐นโ€ฒ(๐‘ก)๐‘‘๐‘ก

    for all ๐‘ฅโˆˆ[๐‘Ž,๐‘].

Proof

้ฆ–ๅ…ˆ, ๐นโˆˆ๐ด๐ถ[๐‘Ž,๐‘]โŸน๐นโˆˆ๐ต๐‘‰[๐‘Ž,๐‘]โŸน๐น diffble a.e. on [๐‘Ž,๐‘].
Notice that: ่ฟ™้‡Œๆˆ‘ไปฌๅช่€ƒ่™‘ ๐น|[๐‘Ž,๐‘], ไบŽๆ˜ฏๆˆ‘ไปฌๅฏไปฅๅฐ†ๅ…ถไป–้ƒจๅˆ†็š„ๅ€ผ้ƒฝ่ฎพไธบ smooth ็š„, ๅนถ normalize it: ๆŠŠ ๐น(๐‘ฅ)โ‰”0 for ๐‘ฅ<๐‘Ž, ๐น(๐‘ฅ):=๐‘ for ๐‘ฅ>๐‘.
ๅˆ ๐นโˆˆ๐ด๐ถโŸน๐น right ctn for sure, ๆˆ‘ไปฌ then have: ๐นโˆˆ๐‘๐ต๐‘‰, ไปŽ่€Œ

โ–ก

12.3.6 characterization for Lipschitz ctn

Theorem 12.70 : characterization for Lipschitz ctn

ๅฏนไบŽ ๐น:โ„โ†’โ„‚, we have:

๐น Lipschitz ctn with const ๐‘€โ‡”๐นโˆˆ๐ด๐ถ and |๐นโ€ฒ(๐‘ฅ)|โ‰ค๐‘€ a.e.
Proof

่ง HW 12.

โ–ก

ไปŽ่€Œๆˆ‘ไปฌๅพ—ๅˆฐ: ctnity ๆกไปถ็š„้€’ๆŽจๅ…ณ็ณป:

Lipschitz ctn โŸน abs ctn โŸน uniformly ctn โŸน ctn

ๅ…ณไบŽ่ฟž็ปญๅ‡ฝๆ•ฐ็š„ๅฏๅฏผๆ€ง: Lipschitz ctn ๅฏไปฅๆŽจๅพ— a.e. diffble + bounded derivative; abs ctn ๅฏไปฅๆŽจๅพ— a.e. diffble ไธ”ๅœจ cpt interval ไธŠ derivative ๐ฟ1; ่€Œๅพ€ๅŽ็š„ uniform ctn ๅˆ™ไธ่•ดๅซๅฏๅฏผๆ€งๆกไปถ.

่€Œๅ…ณไบŽๅ’Œๅฏๅฏผๆ€ง็ดงๅฏ†็›ธๅ…ณ็š„ๅ˜ๅทฎๆ€ง่ดจ: abs ctn ๅœจ bounded interval ไธŠๆ˜ฏๆฏ” BV, NBV ๆ›ดๅผบ็š„ๆกไปถ, ่€Œๅœจ โ„ ไธŠๅˆ™ๅนถไธๆ˜ฏ.

ๅœจ โ„ ไธŠ, NBV + AC ็š„ๅ‡ฝๆ•ฐๅฏ่ฟ็”จ FTC. ่€Œๅœจ bounded area ไธŠ AC ็š„ๅ‡ฝๆ•ฐๅฐฑๅฏไปฅ่ฟ็”จ FTC.

Homework 11: on regular Borel measure and functions of bounded variation (36/40)

Measurability of densities of measures

Suppose ๐œ‡ is a regular (positive) Borel measure on โ„๐‘›.

  • Prove that the functions ๐‘“ฬ„:โ„๐‘›โ†’[0,+โˆž] and ๐‘“ยฏ:โ„๐‘›โ†’[0,+โˆž] defined by

    ๐‘“ฬ„(๐‘ฅ)โ‰”limโ€‰sup๐‘Ÿโ†’0+๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ)),and๐‘“ยฏ(๐‘ฅ)โ‰”limโ€‰inf๐‘Ÿโ†’0+๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

    where ๐‘š denotes Lebesgue measure, are Borel measurable.

  • Prove that the set

    ๐ด={๐‘ฅโˆˆโ„๐‘›โˆฃthe limit lim๐‘Ÿโ†’0+๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ)) exists in [0,+โˆž])}

    is Borel measurable.

  • Give an example where ๐ดโ‰ โ„๐‘›.

Hint: we are taking the limsup over an uncountable set, so you probably need to use some properties of the functions ๐‘Ÿโ†ฆ๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ)) and ๐‘Ÿโ†ฆ๐‘š(๐ต(๐‘ฅ,๐‘Ÿ)), in addition to properties of ๐‘ฅโ†ฆ๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ)) and ๐‘ฅโ†ฆ๐‘š(๐ต(๐‘ฅ,๐‘Ÿ)).

Proof

of (a): We prove a lemma:

Lemma 12.51

For regular positive Borel measure ๐œ‡ on โ„๐‘›, fixing ๐‘Ÿ>0, ๐‘ฅโ†ฆ๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ)) is Borel measurable.

Proof of Lemma: We recall

๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ))=โˆซ๐œ’๐ต(๐‘ฅ,๐‘Ÿ)๐‘‘๐œ‡=โˆซ๐œ’๐ต(๐‘ฅ,๐‘Ÿ)(๐‘ฆ)๐‘‘๐œ‡(๐‘ฆ)

We define

๐‘“(๐‘ฅ,๐‘ฆ)=๐œ’๐ต(๐‘ฅ,๐‘Ÿ)(๐‘ฆ)

which is a function from โ„๐‘›ร—โ„๐‘›โ†’โ„, and takes value between 0 and 1.
Thus for ๐‘Žโ‰ฅ1,

๐‘“โˆ’1((๐‘Ž,โˆž))=โŒ€โˆˆโ„ฌ๏ธ€(โ„๐‘›)โŠ—โ„ฌ๏ธ€(โ„๐‘›)

for ๐‘Ž<0,

๐‘“โˆ’1((๐‘Ž,โˆž))=๐‘“โˆ’1({0,1})=โ„๐‘›ร—โ„๐‘›โˆˆโ„ฌ๏ธ€(โ„๐‘›)โŠ—โ„ฌ๏ธ€(โ„๐‘›)

For 0โ‰ค๐‘Ž<1, ๐‘“โˆ’1((๐‘Ž,โˆž))=๐‘“โˆ’1({1}). Note this set is:

๐‘“โˆ’1((๐‘Ž,โˆž))={(๐‘ฅ,๐‘ฆ)โˆˆโ„๐‘›ร—โ„๐‘›:๐‘ฆโˆˆ๐ต(๐‘ฅ,๐‘Ÿ)}={(๐‘ฅ,๐‘ฆ)โˆˆโ„๐‘›ร—โ„๐‘›:โˆฅ๐‘ฅโˆ’๐‘ฆโˆฅ<๐‘Ÿ}

Since ๐‘”:(๐‘ฅ,๐‘ฆ)โ†ฆโˆฅ๐‘ฅโˆ’๐‘ฆโˆฅ2 is continuous function, and

๐‘“โˆ’1((๐‘Ž,โˆž))={(๐‘ฅ,๐‘ฆ)โˆˆโ„๐‘›ร—โ„๐‘›:โˆฅ๐‘ฅโˆ’๐‘ฆโˆฅ<๐‘Ÿ}=๐‘”โˆ’1(๐‘Ÿ)

is open, since it is preimage of an open set, under a continuous function.
Thus

๐‘“โˆ’1((๐‘Ž,โˆž))โˆˆโ„ฌ๏ธ€(โ„2๐‘›)=โ„ฌ๏ธ€(โ„๐‘›)โŠ—โ„ฌ๏ธ€(โ„๐‘›)

Thus ๐‘“ is Borel measurable function, and since it is nonnegative, ๐‘“โˆˆ๐ฟ+(โ„2๐‘›), thus by Tonelliโ€™s Theorem,

๐‘ฅโ†ฆโˆซ๐‘“๐‘ฅ(๐‘ฆ)๐‘‘๐œ‡(๐‘ฆ)=๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ)) is Borel measurable

finishing the proof of Lemma.
Define for ๐‘Ÿ>0

๐‘“๐‘Ÿ(๐‘ฅ)โ‰”๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

Notice that for each ๐‘Ÿ, ๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=๐‘๐‘›๐‘Ÿ๐‘›>0 is constant regardless of ๐‘ฅ, so ๐‘“๐‘˜ is Borel measurable as a product of a Boreal measurable function and a constant.
ย So

๐‘“ฬ„(๐‘ฅ)=limโ€‰sup๐‘Ÿโ†’0+๐‘“๐‘Ÿ(๐‘ฅ)=lim๐œ–>0sup0<๐‘Ÿ<๐œ–๐‘“๐‘Ÿ(๐‘ฅ)

For fixed ๐œ–>0, we define โ„Ž๐œ–(๐‘ฅ):=sup0<๐‘Ÿ<๐œ–๐‘“๐‘Ÿ(๐‘ฅ), then for ๐‘Žโˆˆโ„, we have

โ„Ž๐œ–((๐‘Ž,โˆž))=โ‹ƒ0<๐‘Ÿ<๐œ–๐‘“๐‘Ÿ((๐‘Ž,โˆž))=โ‹ƒ0<๐‘Ÿ<๐œ–,๐‘Ÿโˆˆโ„š๐‘“๐‘Ÿ((๐‘Ž,โˆž))

is Bore measurable, Thus โ„Ž๐œ– is a Borel measurable function, then

๐‘“ฬ„=lim๐œ–>0โ„Ž๐œ–=lim๐‘›โ†’โˆžโ„Ž1๐‘›

is a Borel measurable function as limit of a seq of Borel measurable functions. ่ฟ™้‡Œๆณจๆ„: Reducing limup (or liminf) over an uncountable sets to a countable one requires upper/lower semicontinuity. ๅ› ่€Œๆˆ‘ไปฌ้œ€่ฆ่ฏดๆ˜Žไธ€ไธ‹ ๐‘“๐‘Ÿ ๆ˜ฏ right ctn in r ็š„. Same trick is applied to ๐‘“ยฏ. We set ๐‘”๐œ–(๐‘ฅ):=inf0<๐‘Ÿ<๐œ–๐‘“๐‘Ÿ(๐‘ฅ) and have ๐‘“ยฏ(๐‘ฅ)=lim๐‘›โ†’โˆž๐‘”1๐‘› is Borel measurable, finishing the proof.

โ–ก

Proof

of (b):

๐ดโ‰”{๐‘ฅโˆˆโ„๐‘›:the limit lim๐‘Ÿโ†’0+๐œ‡(๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ)) exists in [0,+โˆž])}={๐‘ฅโˆˆโ„๐‘›:๐‘“ฬ„(๐‘ฅ)=๐‘“ยฏ(๐‘ฅ)}

Notice:

Lemma 12.52

if (๐‘‹,๐’œ๏ธ€) is a measurable space; ๐‘“,๐‘”:๐‘‹โ†’โ„ are (๐’œ๏ธ€,โ„ฌ๏ธ€(โ„)) -measurable functions, then

๐น(๐‘ฅ)โ‰”(๐‘“(๐‘ฅ),๐‘”(๐‘ฅ)):๐‘‹โ†’โ„2

is a (๐’œ๏ธ€,โ„ฌ๏ธ€(โ„2))-measurable function.

Proof of Lemma: We have shown in hw8 that, ๐‘“ is an product measurable function if ๐‘“โˆ’1(๐ต1ร—๐ต2) is measurable for each measurable rectangle ๐ต1ร—๐ต2.
And for measurable rectangle ๐‘ˆร—๐‘‰โŠ‚โ„2, we have:

๐นโˆ’1(๐‘ˆร—๐‘‰)=๐‘“โˆ’1(๐‘ˆ)โˆฉ๐‘”โˆ’1(๐‘‰)โˆˆ๐’œ๏ธ€

proving the lemma.
And back to the original statement, we define:

๐น(๐‘ฅ)=(๐‘“ฬ„(๐‘ฅ),๐‘“ยฏ(๐‘ฅ))

Then we notice that

๐ด={๐‘ฅโˆˆโ„๐‘›:๐‘“ฬ„(๐‘ฅ)=๐‘“ยฏ(๐‘ฅ)}=๐นโˆ’1({(๐‘ฅ,๐‘ฅ)|๐‘ฅโˆˆโ„})

Since the diagonal {(๐‘ฅ,๐‘ฅ)|๐‘ฅโˆˆโ„} is a closed set, it is a Borel set. And by lemma, ๐น is a Borel measurable function, implying that ๐ด is Borel measurable.

โ–ก

Example 12.40

of (c): Consider

๐ผ:={0}โˆชโ‹ƒ๐‘—=0โˆž[23โ‹…12๐‘—,12๐‘—]

Set

๐‘”:=๐œ’๐ผ,๐œ‡(๐ธ):=โˆซ๐ธ๐‘”๐‘‘๐‘š

Then we look at ๐‘ฅ=0, we have:

๐œ‡(๐ต(0,๐‘Ÿ))๐‘š(๐ต(0,๐‘Ÿ))=๐‘š(๐ต(0,๐‘Ÿ)โˆฉ๐ผ)๐‘š(๐ต(0,๐‘Ÿ))

So

lim๐‘Ÿโ†’0+๐œ‡(๐ต(0,๐‘Ÿ))๐‘š(๐ต(0,๐‘Ÿ))=lim๐‘Ÿโ†’0+๐‘š(๐ผโˆฉ๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

is exactly the density of ๐ผ at 0, and we have shown in class that this limit does not exist, in the sense that its limsup is not equal to its liminf, i.e.

limโ€‰sup๐‘Ÿโ†’0+๐‘š(๐ผโˆฉ๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))=:๐‘“ฬ„(๐‘ฅ)โ‰ ๐‘“ยฏ(๐‘ฅ)โ‰”limโ€‰inf๐‘Ÿโ†’0+๐‘š(๐ผโˆฉ๐ต(๐‘ฅ,๐‘Ÿ))๐‘š(๐ต(๐‘ฅ,๐‘Ÿ))

Here we explain it in detailed:

Figureย 38:

If we take ๐‘Ÿ๐‘˜=12๐‘˜ for ๐‘˜โˆˆโ„•, we have:

๐ต(0,๐‘Ÿ๐‘˜)=(โˆ’๐‘Ÿ๐‘˜,๐‘Ÿ๐‘˜)=(โˆ’12๐‘˜,12๐‘˜)

Then for each ๐‘˜,

๐‘š(๐ผโˆฉ๐ต(0,๐‘Ÿ๐‘˜)=โˆ‘๐‘—=๐‘˜โˆž13โ‹…12๐‘—=13โ‹…โˆ‘๐‘—=๐‘˜โˆž12๐‘—=13โ‹…12๐‘˜โˆ’1,๐‘š(๐ต(0,๐‘Ÿ๐‘˜))=2๐‘Ÿ๐‘˜=22๐‘˜

So for each ๐‘˜,

๐œ‡(๐ต(0,๐‘Ÿ๐‘˜))๐‘š(๐ต(0,๐‘Ÿ๐‘˜))=13

so we have:

๐‘“ฬ„(0)โ‰ฅ13

But if we take ๐‘Ÿ๐‘˜=23โ‹…12๐‘˜, then for each ๐‘˜,

๐‘š(๐ผโˆฉ๐ต(0,๐‘Ÿ๐‘˜))=โˆ‘๐‘—=๐‘˜+1โˆž13โ‹…12๐‘—=13โ‹…โˆ‘๐‘—=๐‘˜+1โˆž12๐‘—=13โ‹…12๐‘˜,๐‘š(๐ต(0,๐‘Ÿ๐‘˜))=2๐‘Ÿ๐‘˜=22๐‘˜

So for each ๐‘˜,

๐œ‡(๐ต(0,๐‘Ÿ๐‘˜))๐‘š(๐ต(0,๐‘Ÿ๐‘˜))=16

so we have:

๐‘“ยฏ(0)โ‰ค16

Proving that

๐‘“ฬ„(0)โ‰ ๐‘“ยฏ(0)

This serves as an counterexample of ๐ดโ‰ โ„๐‘› (๐‘›=1 here)

Lebesgue decomposition ๐œˆ=๐œ†+๐œŒโŸน|๐œˆ|=|๐œ†|+|๐œŒ|

  • Let ๐œˆ be a regular complex or finite signed Borel measure on โ„๐‘›, and let ๐œˆ=๐œ†+๐œŒ be its Lebesgue decomposition with respect to Lebesgue measure ๐‘š, so that ๐œ†โŸ‚๐‘š and ๐œŒโ‰ช๐‘š. Prove that the Lebesgue decomposition of the total variation measure |๐œˆ| with respect to ๐‘š is given by |๐œˆ|=|๐œ†|+|๐œŒ|. In other words, prove that |๐œˆ|=|๐œ†|+|๐œŒ|, |๐œ†|โŸ‚๐‘š, and |๐œŒ|โ‰ช๐‘š.

  • Let ๐œ‡1 and ๐œ‡2 be positive, mutually singular Borel measures on โ„๐‘›. Prove that ๐œ‡1+๐œ‡2 is regular iff ๐œ‡1 and ๐œ‡2 are both regular.

Remark: these results were used the the proof of Theorem 3.22 in Folland. Please donโ€™t use any results fromย ยง7.

Proof

of (a): Recall that for two complex measures ๐œ†,๐œŒ, we define they are mutually singular if:

๐œ†โŸ‚๐œŒโ‡”๐œ†๐‘ŸโŸ‚๐œŒ๐‘Ÿ,๐œ†๐‘ŸโŸ‚๐œŒ๐‘–,๐œ†๐‘–โŸ‚๐œŒ๐‘Ÿ,๐œ†๐‘–โŸ‚๐œŒ๐‘–

We first show an equivalent form of it, for further use.

Lemma 12.53

For two complex measures ๐œ†,๐œŒ

๐œ†โŸ‚๐œŒโ‡”โˆƒ๐ดโˆˆ๐’œ๏ธ€ s.t. |๐œ†|(๐ด๐‘)=0 and |๐œŒ|(๐ด)=0โ‡”|๐œ†|โŸ‚|๐œŒ|

Proof of the lemma: The second equivalence follows from definition (since total variation measure is positive), and the backward direction of the first equivalence follows from that the null set of the total variation measure is also the null set for original complex measure (thus null set for the positive and imaginary part).
For the forward direction of the first equivalence,

๐œ†๐‘ŽโŸ‚๐œŒ๐‘โŸนโˆƒ๐ด๐‘Ž๐‘โˆˆ๐’œ๏ธ€:๐ด๐‘Ž๐‘ is null set for ๐œŒ๐‘ and ๐ด๐‘Ž๐‘๐‘ is null set for ๐œ†๐‘ŽโŸนโˆƒ๐ด๐‘Ž๐‘โˆˆ๐’œ๏ธ€:|๐œ†๐‘Ž|(๐ด๐‘Ž๐‘๐‘)=0,|๐œŒ๐‘|(๐ด๐‘Ž๐‘)=0

Define:

๐ดโ‰”(๐ด๐‘Ÿ๐‘Ÿโˆฉ๐ด๐‘Ÿ๐‘–)โ‹ƒ(๐ด๐‘–๐‘Ÿโˆฉ๐ด๐‘–๐‘–)โˆˆ๐’œ๏ธ€

Since ๐ด๐‘Ÿ๐‘Ÿโˆฉ๐ด๐‘Ÿ๐‘– is a null set for ๐œŒ๐‘Ÿ,๐œŒ๐‘–, thus a null set for |๐œŒ|. And (๐ด๐‘Ÿ๐‘Ÿโˆฉ๐ด๐‘Ÿ๐‘–)๐‘=๐ด๐‘Ÿ๐‘Ÿ๐‘โˆช๐ด๐‘Ÿ๐‘–๐‘. Since these two are both null set for ๐œ†๐‘Ÿ and union of null sets is null set, (๐ด๐‘Ÿ๐‘Ÿโˆฉ๐ด๐‘Ÿ๐‘–)๐‘ is also a null set for ๐œ†๐‘Ÿ.
Similarly, ๐ด๐‘–๐‘Ÿโˆฉ๐ด๐‘–๐‘– is a null set for |๐œŒ| and (๐ด๐‘–๐‘Ÿโˆฉ๐ด๐‘–๐‘–)๐‘ is a null set for ๐œ†๐‘–.
Thus, ๐ด is a null set for |๐œŒ|, and ๐ด๐‘=(๐ด๐‘Ÿ๐‘Ÿโˆฉ๐ด๐‘Ÿ๐‘–)๐‘โ‹‚(๐ด๐‘–๐‘Ÿโˆฉ๐ด๐‘–๐‘–)๐‘ is a null set for both ๐œ†๐‘Ÿ and ๐œ†๐‘–, thus a null set for ๐œ†.
This finishes the construction of ๐ด, proving our lemma. Now we can apply the equivalent conditions of ๐œ†โŸ‚๐œŒ for positive, signed and complex measures.
Now we prove this statement which immediately implies what we want:

Proposition 12.35

If complex measure ๐œ† and ๐œŒ on the same measurable space are mutually singular, then

|๐œ†+๐œŒ|=|๐œ†|+|๐œŒ|

Proof of Proposition: Since ๐œ†โŸ‚๐œŒ, there exists a measurable set ๐ดโІ๐‘‹ such that:

|๐œ†|(๐ด๐‘)=0 and |๐œŒ|(๐ด)=0

Let ๐œˆโ‰”๐œ†+๐œŒ. Let ๐ธโˆˆ๐’œ๏ธ€.

Then

|๐œˆ|(๐ธ)=|๐œˆ|((๐ธโˆฉ๐ด)โŠ”(๐ธโˆฉ๐ด๐‘))=|๐œˆ|(๐ธโˆฉ๐ด)+|๐œˆ|(๐ธโˆฉ๐ด๐‘)=|๐œ†+๐œŒ|(๐ธโˆฉ๐ด)+|๐œ†+๐œŒ|(๐ธโˆฉ๐ด๐‘)=|๐œ†|(๐ธโˆฉ๐ด)+|๐œŒ|(๐ธโˆฉ๐ด๐‘)since ๐œ†=0 on ๐ด๐‘ and ๐œŒ=0 on ๐ด=|๐œ†|(๐ธ)+|๐œŒ|(๐ธ)since |๐œ†| is 0 on ๐ธโˆฉ๐ด๐‘, |๐œŒ| is 0 on ๐ธโˆฉ๐ด

finishing the proof the the proposition.
Now we look back at the original statement: For Lebesgue decomposition ๐œˆ=๐œ†+๐œŒ, we have ๐œ†โŸ‚๐‘š and ๐œŒโ‰ช๐‘š. ๐œ†โŸ‚๐‘š implies that there exists a measurable set ๐ดโІ๐‘‹ such that:

|๐œ†|(๐ด๐‘)=0 and ๐‘š(๐ด)=0

Since ๐œŒโ‰ช๐‘š, null sets of ๐‘š are also null sets of ๐œŒ, thus |๐œŒ|(๐ด)=0. Thus we have

๐œ†โŸ‚๐œŒ

By our just proved proposition we have:

|๐œˆ|=|๐œ†+๐œŒ|=|๐œ†|+|๐œŒ|

And it also follows from our lemma that

๐œ†โŸ‚๐‘šโŸน|๐œ†|โŸ‚๐‘š

and |๐œŒ|โ‰ช๐‘š is trivial, since |๐œŒ| and ๐œŒ have the same null sets.
This finishes the proof that: if Lebesgue decomposition of ๐œˆ is ๐œˆ=๐œ†+๐œŒ, then Lebesgue decomposition of the total variation measure |๐œˆ| with respect to ๐‘š is given by |๐œˆ|=|๐œ†|+|๐œŒ|.

โ–ก

Proof

of (b): First we show (โŸน:) if ๐œ‡1 and ๐œ‡2 are both regular then ๐œ‡1+๐œ‡2 is regular.
Let ๐ด be a Borel set. Since ๐œ‡1 and ๐œ‡2 are regular, we have:

๐œ‡1(๐ด)=inf๐ดโŠ‚๐‘ˆ๐œ‡1(๐‘ˆ)=sup๐พโŠ‚๐ด๐œ‡1(๐พ),๐œ‡2(๐ด)=inf๐ดโŠ‚๐‘ˆ๐œ‡2(๐‘ˆ)=sup๐พโŠ‚๐ด๐œ‡2(๐พ)

Set ๐œ‡=๐œ‡1+๐œ‡2, then

๐œ‡(๐ด)=inf๐ดโŠ‚๐‘ˆ(๐œ‡(๐‘ˆ))=inf๐ดโŠ‚๐‘ˆ(๐œ‡1(๐‘ˆ)+๐œ‡2(๐‘ˆ))โ‰ฅinf๐ดโŠ‚๐‘ˆ๐œ‡1(๐‘ˆ)+inf๐ดโŠ‚๐‘ˆ๐œ‡2(๐‘ˆ)=๐œ‡1(๐ด)+๐œ‡2(๐ด)

Also on the other direction,

๐œ‡(๐ด)=sup๐พโŠ‚๐ด(๐œ‡(๐พ))=sup๐พโŠ‚๐ด๐œ‡1(๐พ)+๐œ‡2(๐พ))โ‰คsup๐พโŠ‚๐ด๐œ‡1(๐พ)+sup๐พโŠ‚๐ด๐œ‡2(๐พ)=๐œ‡1(๐ด)+๐œ‡2(๐ด)

Combining these two ineq chains, all inequalities is indeed equality. Thus we have

๐œ‡(๐ด)=inf๐ดโŠ‚๐‘ˆ(๐œ‡(๐‘ˆ))=sup๐พโŠ‚๐ด(๐œ‡(๐พ))=๐œ‡1(๐ด)+๐œ‡2(๐ด)

The first two equalities shows regularities, and the last equality shows finiteness. This finishes the proof of forward direction.
Next we show: (โŸธ:) if ๐œ‡1+๐œ‡2 is regular then ๐œ‡1 and ๐œ‡2 are both regular.
Let ๐ด be a Borel set.
First, suppose ๐ด is compact. Then (๐œ‡1+๐œ‡2)(๐พ)<โˆž. Notice, since ๐œ‡1,๐œ‡2 are positive measures, ๐œ‡1+๐œ‡2โ‰ฅ๐œ‡1,๐œ‡2, thus we sure have

๐œ‡1(๐ด),๐œ‡2(๐ด)<โˆž

This shows the local finiteness of ๐œ‡1,๐œ‡2. It remains to show the outer regularity of ๐œ‡1,๐œ‡2. (Note: local finiteness โŸน outer regularity is reached using tools in Ch7, so we still need to show outer regularity here; for local finiteness + outer regularity โŸน inner regularity, it have similar steps as Thm 1.18, so it is done.)
Since ๐œ‡1โŸ‚๐œ‡2, there exists measurable ๐ธโŠ‚โ„๐‘› s.t.

๐ธ null for ๐œ‡1,๐ธ๐‘ null for ๐œ‡2

By outer regularity of ๐œ‡1+๐œ‡2, we can construct a seq of open sets ๐‘ˆ๐‘˜โŠƒ๐ด s.t.

(๐œ‡1+๐œ‡2)(๐‘ˆ๐‘˜)<(๐œ‡1+๐œ‡2)(๐ด)+12๐‘˜

Thus we have

lim๐‘˜โ†’โˆž(๐œ‡1+๐œ‡2)(๐‘ˆ๐‘˜)=(๐œ‡1+๐œ‡2)(๐ด)

And notice that, for each ๐‘˜,

(๐œ‡1+๐œ‡2)(๐‘ˆ๐‘˜)=(๐œ‡1+๐œ‡2)(๐‘ˆ๐‘˜โˆฉ๐ธ)+(๐œ‡1+๐œ‡2)(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)=๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)+๐œ‡2(๐‘ˆ๐‘˜โˆฉ๐ธ)since ๐ธ null for ๐œ‡1,๐ธ๐‘ null for ๐œ‡2

And for ๐ด, similarly we have:

(๐œ‡1+๐œ‡2)(๐ด)=๐œ‡1(๐ดโˆฉ๐ธ๐‘)+๐œ‡2(๐ดโˆฉ๐ธ)

Since ๐‘ˆ๐‘˜โŠƒ๐ด, we have ๐‘ˆ๐‘˜โˆฉ๐ธโŠƒ๐ดโˆฉ๐ธ, thus ๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)โ‰ฅ๐œ‡1(๐ดโˆฉ๐ธ๐‘), and similarly ๐œ‡2(๐‘ˆ๐‘˜โˆฉ๐ธ)โ‰ฅ๐œ‡2(๐ดโˆฉ๐ธ).
Thus

(๐œ‡2+๐œ‡2)(๐‘ˆ๐‘˜)โˆ’(๐œ‡2+๐œ‡2)(๐ด)=๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)+๐œ‡2(๐‘ˆ๐‘˜โˆฉ๐ธ)โˆ’(๐œ‡1(๐ดโˆฉ๐ธ๐‘)+๐œ‡2(๐ดโˆฉ๐ธ))=๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)โˆ’๐œ‡1(๐ดโˆฉ๐ธ๐‘)+(๐œ‡2(๐‘ˆ๐‘˜โˆฉ๐ธ)โˆ’๐œ‡2(๐ดโˆฉ๐ธ))โ‰ฅ๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)โˆ’๐œ‡1(๐ดโˆฉ๐ธ๐‘)(since ๐œ‡2(๐‘ˆ๐‘˜โˆฉ๐ธ)โˆ’๐œ‡2(๐ดโˆฉ๐ธ)โ‰ฅ0)=๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ๐‘)+๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ)โˆ’๐œ‡1(๐ดโˆฉ๐ธ๐‘)โˆ’๐œ‡1(๐ดโˆฉ๐ธ)(since ๐œ‡1(๐‘ˆ๐‘˜โˆฉ๐ธ),๐œ‡2(๐ดโˆฉ๐ธ)=0)=๐œ‡1(๐‘ˆ๐‘˜)โˆ’๐œ‡1(๐ด)โ‰ฅ0

Therefore

(๐œ‡1+๐œ‡2)(๐‘ˆ๐‘˜)โ†˜๏ธŽ๐‘˜โ†’โˆž(๐œ‡1+๐œ‡2)(๐ด)โŸน๐œ‡1(๐‘ˆ๐‘˜)โ†˜๏ธŽ๐‘˜โ†’โˆž๐œ‡1(๐ด)

Since ๐‘ˆ๐‘˜โŠƒ๐ด for each ๐‘˜, this shows the outer regularity:

๐œ‡1(๐ด)=inf๐‘ˆ open โŠƒ๐ด๐œ‡1(๐‘ˆ)

And dually, through exact same steps we can get:

๐œ‡2(๐‘ˆ๐‘˜)โ†˜๏ธŽ๐‘˜โ†’โˆž๐œ‡2(๐ด),๐œ‡2(๐ด)=inf๐‘ˆ open โŠƒ๐ด๐œ‡2(๐‘ˆ)

finishing the proof.

โ–ก

A convergence problem

Let ๐‘“โˆˆ๐ฟ1(โ„). For ๐‘›โˆˆโ„•, define ๐‘“๐‘›:โ„โ†’โ„ as follows. For ๐‘˜โˆˆโ„ค and ๐‘ฅโˆˆ[๐‘˜๐‘›,๐‘˜+1๐‘›), set

๐‘“๐‘›(๐‘ฅ)โ‰”๐‘›โˆซ๐‘˜๐‘›๐‘˜+1๐‘›๐‘“(๐‘ก)๐‘‘๐‘ก
  • Prove that ๐‘“๐‘›โ†’๐‘“ a.e.

  • Prove that ๐‘“๐‘›โ†’๐‘“ in ๐ฟ1.

Hint: for (a), use the Lebesgue differentiability theorem; for (b) you may want to approximate ๐‘“ by a nice function.

Proof

of (a):

Figureย 39:
๐‘“๐‘›(๐‘ฅ)=๐‘›โˆซ๐‘˜๐‘›๐‘˜+1๐‘›๐‘“(๐‘ก)๐‘‘๐‘ก=11/๐‘›โˆซ๐ผ๐‘›,๐‘˜๐‘“(๐‘ก)๐‘‘๐‘ก=1๐‘š(๐ผ๐‘›,๐‘˜)โˆซ๐ผ๐‘›,๐‘˜๐‘“(๐‘ก)๐‘‘๐‘ก

Thus ๐‘“๐‘›(๐‘ฅ) is the average of ๐‘“ over the interval ๐ผ๐‘›,๐‘˜โ‰”[๐‘˜๐‘›,๐‘˜+1๐‘›), where ๐‘ฅโˆˆ๐ผ๐‘›,๐‘˜.
Fixing ๐‘ฅโˆˆโ„, for each ๐‘› we set ๐ธ๐‘›(๐‘ฅ)โ‰”๐ผ๐‘›,๐‘˜ for ๐ผ๐‘›,๐‘˜ s.t. ๐‘ฅโˆˆ๐ผ๐‘›,๐‘˜. Notice that for each ๐‘›,

โจ†๐‘˜๐ผ๐‘›,๐‘˜=โ„

so this ๐ธ๐‘› is well-defined.
And for each ๐ธ๐‘›, we have

๐ธ๐‘›(๐‘ฅ)=[๐‘˜๐‘›,๐‘˜+1๐‘›)โŠ‚(๐‘ฅโˆ’2๐‘›,๐‘ฅ+2๐‘›)=๐ต(๐‘ฅ,2๐‘›)

And

๐‘š(๐ธ๐‘›(๐‘ฅ))=1๐‘›=14๐‘š(๐ต(๐‘ฅ,2๐‘›))

This shows that ๐ธ๐‘›(๐‘ฅ) nicely shrinks to ๐‘ฅ as ๐‘›โ†’โˆž. Then by LDT, we have

lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)=lim๐‘›โ†’โˆž1๐‘š(๐ธ๐‘›(๐‘ฅ))โˆซ๐ธ๐‘›(๐‘ฅ)๐‘“(๐‘ก)๐‘‘๐‘ก=๐‘“(๐‘ฅ)

for ๐‘š-a.e. ๐‘ฅ.
This finishes the proof.

โ–ก

Proof

of (b): WTS:

lim๐‘›โ†’โˆžโˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ1=โˆซ|๐‘“๐‘›(๐‘ฅ)โˆ’๐‘“(๐‘ฅ)|๐‘‘๐‘ฅ=0

Since ๐‘“โˆˆ๐ฟ1(โ„), we can select ๐œ™โˆˆ๐ถ๐‘0(โ„) a ctn compactly supported function (e.g., can take bump function) such that

โˆฅ๐‘“โˆ’๐œ™โˆฅ1<๐œ€/3

Now define ๐œ™๐‘› by averaging ๐œ™ over the same intervals:

๐œ™๐‘›(๐‘ฅ)โ‰”๐‘›โˆซ๐‘˜/๐‘›(๐‘˜+1)/๐‘›๐œ™(๐‘ก)๐‘‘๐‘ก=1๐‘š(๐ผ๐‘›,๐‘˜)โˆซ๐ผ๐‘›,๐‘˜๐œ™(๐‘ก)๐‘‘๐‘ก , for ๐‘ฅโˆˆ[๐‘˜๐‘›,๐‘˜+1๐‘›)

Then by tri eq on ๐ฟ1(๐‘š),

โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ1โ‰คโˆฅ๐‘“๐‘›โˆ’๐œ™๐‘›โˆฅ1+โˆฅ๐œ™๐‘›โˆ’๐œ™โˆฅ1+โˆฅ๐œ™โˆ’๐‘“โˆฅ1

First, โˆฅ๐œ™โˆ’๐‘“โˆฅ1<๐œ€/3 by construction. Next, fixing ๐‘›,๐‘˜, we write the value of ๐‘“๐‘›(๐‘ฅ) over the interval ๐ผ๐‘›,๐‘˜โ‰”[๐‘˜๐‘›,๐‘˜+1๐‘›) as ๐‘“๐‘›,๐‘˜, and value of ๐œ™๐‘›(๐‘ฅ) over the interval ๐ผ๐‘›,๐‘˜โ‰”[๐‘˜๐‘›,๐‘˜+1๐‘›) as ๐œ™๐‘›,๐‘˜. Then for each ๐‘›,๐‘˜

โˆฅ๐‘“๐‘›|๐ผ๐‘›,๐‘˜โˆ’๐œ™๐‘›|๐ผ๐‘›,๐‘˜โˆฅ1=โˆซ๐ผ๐‘›,๐‘˜|๐‘“๐‘›,๐‘˜โˆ’๐œ™๐‘›,๐‘˜|๐‘‘๐‘ฅ=1๐‘›|๐‘“๐‘›,๐‘˜โˆ’๐œ™๐‘›,๐‘˜|=1๐‘›โ‹…๐‘›|โˆซ๐ผ๐‘›,๐‘˜(๐‘“(๐‘ก)โˆ’๐œ™(๐‘ก))๐‘‘๐‘ก|=|โˆซ๐ผ๐‘›,๐‘˜(๐‘“(๐‘ก)โˆ’๐œ™(๐‘ก))๐‘‘๐‘ก|

Since ๐‘“๐‘›โˆ’๐œ™๐‘›=โˆ‘๐‘˜โˆˆโ„ค๐‘“๐‘›|๐ผ๐‘›,๐‘˜โˆ’๐œ™๐‘›|๐ผ๐‘›,๐‘˜, by Minkowskiโ€™s ineq we then have:

โˆฅ๐‘“๐‘›โˆ’๐œ™๐‘›โˆฅโ‰คโˆ‘๐‘˜โˆˆโ„คโˆฅ๐‘“๐‘›|๐ผ๐‘›,๐‘˜โˆ’๐œ™๐‘›|๐ผ๐‘›,๐‘˜โˆฅ1=โˆ‘๐‘˜โˆˆโ„ค|โˆซ๐ผ๐‘›,๐‘˜(๐‘“(๐‘ก)โˆ’๐œ™(๐‘ก))๐‘‘๐‘ก|โ‰คโˆ‘๐‘˜โˆˆโ„คโˆซ๐ผ๐‘›,๐‘˜|๐‘“(๐‘ก)โˆ’๐œ™(๐‘ก)|๐‘‘๐‘ก=โˆซ|๐‘“(๐‘ก)โˆ’๐œ™(๐‘ก)|๐‘‘๐‘ก=โˆฅ๐‘“โˆ’๐œ™โˆฅ1<๐œ–3

This shows that, for every ๐‘›โˆˆโ„•, we all have โˆฅ๐‘“๐‘›โˆ’๐œ™๐‘›โˆฅ<๐œ–3.
And finally for ๐œ™๐‘›โˆ’๐œ™, since ๐œ™โˆˆ๐ถ๐‘0(โ„)โŠ‚๐ฟ1(โ„), by (a) we already have ๐œ™๐‘›โ†’๐œ™ a.e.; and, since ๐œ™ have compact support, say ๐พ with ๐‘š(๐พ)<โˆž and it is continuous on the compact support, it is uniformly continuous and bounded. Say |๐œ™|<๐‘€ for some ๐‘€>0.
Then the function ๐‘”=๐‘€ on ๐พ and ๐‘”=0 on ๐พ๐‘ can serve as a dominating function for ๐œ™๐‘›, with โˆซ๐‘”=๐‘€โ‹…๐‘š(๐พ)<โˆž. Then by DCT, we have: ๐œ‘๐‘›โ†’๐œ‘ in ๐ฟ1.
So for some ๐‘โˆˆโ„•, โˆฅ๐œ™๐‘›โˆ’๐œ™โˆฅ1<๐œ–/3 for all ๐‘›โ‰ฅ๐‘.
Therefore for all ๐‘›โ‰ฅ๐‘, we have:

โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ1โ‰คโˆฅ๐‘“๐‘›โˆ’๐œ™๐‘›โˆฅ1+โˆฅ๐œ™๐‘›โˆ’๐œ™โˆฅ1+โˆฅ๐œ™โˆ’๐‘“โˆฅ1<๐œ–

This finishes the proof that

lim๐‘›โ†’โˆžโˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ1=0

โ–ก

Oscillations: ๐น(๐‘ฅ)=๐‘ฅsin1๐‘ฅ,๐‘ฅ2sin1๐‘ฅ2โˆˆ๐ต๐‘‰(๐ผ)โ‡”0โˆ‰๐ผ

  • Define ๐น:โ„โ†’โ„ by ๐น(๐‘ฅ)=๐‘ฅsin1๐‘ฅ for ๐‘ฅโ‰ 0 and ๐น(0)=1. Prove that if ๐ผ=[๐‘Ž,๐‘]โŠ‚โ„ is a compact interval, so that โˆ’โˆž<๐‘Ž<๐‘<โˆž, then ๐นโˆˆBV(๐ผ) iff 0โˆ‰๐ผ.

  • Define ๐น:โ„โ†’โ„ by ๐น(๐‘ฅ)=๐‘ฅ2sin1๐‘ฅ2 for ๐‘ฅโ‰ 0 and ๐น(0)=0. Prove that ๐น is differentiable everywhere (including at ๐‘ฅ=0) but that ๐นโˆ‰BV([โˆ’1,1]).

Proof

of (a):
We first verify (โŸน): if 0โˆ‰๐ผ then ๐นโˆˆ๐ต๐‘‰(๐ผ).
We differentiate ๐น(๐‘ฅ)=๐‘ฅsin(1/๐‘ฅ) for ๐‘ฅโ‰ 0:

๐นโ€ฒ(๐‘ฅ)=๐‘‘๐‘‘๐‘ฅ(๐‘ฅโ‹…sin(1๐‘ฅ))=sin(1๐‘ฅ)+๐‘ฅโ‹…cos(1๐‘ฅ)โ‹…(โˆ’1๐‘ฅ2)=sin(1๐‘ฅ)โˆ’1๐‘ฅcos(1๐‘ฅ)

WLOG suppose ๐‘Ž>0, then on [๐‘Ž,๐‘] we have:

0โ‰ค|๐นโ€ฒ|โ‰ค1+1๐‘Ž

Then for arbitrary division of [๐‘Ž,๐‘], say ๐‘Ž=๐‘ฅ0โ‰คโ‹ฏโ‰ค๐‘ฅ๐‘›=๐‘, for all ๐‘— we have:

|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|โ‰ค(1+1๐‘Ž)(๐‘ฅ๐‘—โˆ’๐‘ฅ๐‘—โˆ’1)

Thus

โˆ‘๐‘—=1๐‘›|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|โ‰ค(1+1๐‘Ž)(๐‘โˆ’๐‘Ž)=๐‘โˆ’๐‘Ž+๐‘๐‘Žโˆ’1

Taking sup over all partition of [๐‘Ž,๐‘], proving that ๐‘‡๐น(๐‘Ž;๐‘)โ‰ค๐‘โˆ’๐‘Ž+๐‘๐‘Žโˆ’1, proving that ๐นโˆˆ๐ต๐‘‰([๐‘Ž,๐‘]); If ๐‘Ž<0 then ๐‘<0 also, then 0โ‰ค|๐นโ€ฒ|โ‰ค1โˆ’1๐‘, by same reasoning showing that ๐นโˆˆ๐ต๐‘‰([๐‘Ž,๐‘]).
Then we verify: (โŸธ): if ๐นโˆˆ๐ต๐‘‰(๐ผ) then 0โˆ‰๐ผ. This is equiv to: if 0โˆˆ๐ผ then ๐นโˆ‰๐ต๐‘‰(๐ผ).
Suppose 0โˆˆ๐ผ=[๐‘Ž,๐‘] then ๐‘Žโ‰ค0 and ๐‘โ‰ฅ0, one of which is strict. WLOG we suppose ๐‘>0.
Consider this seq:

๐‘ฆ๐‘›โ‰”1๐‘›๐œ‹+๐œ‹/2โ†’0+

we have:

๐น(๐‘ฆ๐‘›)=๐‘ฆ๐‘›sin(1๐‘ฆ๐‘›)=1๐‘›๐œ‹+๐œ‹/2โ‹…sin(๐‘›๐œ‹+๐œ‹/2)

For odd ๐‘›, ๐น(๐‘ฆ๐‘›)=โˆ’1๐‘›๐œ‹+๐œ‹/2, for even ๐‘›, ๐น(๐‘ฆ๐‘›)=1๐‘›๐œ‹+๐œ‹/2.
Since ๐‘>0, for some ๐‘0 we have ๐‘ฆ๐‘0<๐‘. Then we consider the partition: pick ๐‘โˆˆโ„•, and use ๐‘ฅ0=0,๐‘ฅ1=๐‘ฆ๐‘0+๐‘โˆ’1,๐‘ฅ2=๐‘ฆ๐‘0+๐‘โˆ’2,โ‹ฏ,๐‘ฅ๐‘=๐‘ฆ๐‘0,๐‘ฅ๐‘+1=๐‘ as the partition points of [0,๐‘].
Then we have

โˆ‘๐‘›=1๐‘+1|๐น(๐‘ฅ๐‘›)โˆ’๐น(๐‘ฅ๐‘›โˆ’1)|โ‰ฅโˆ‘๐‘›=๐‘0๐‘0โˆ’2+๐‘1๐œ‹๐‘›+๐œ‹/2+1๐œ‹(๐‘›+1)+๐œ‹/2โ‰ฅ2โˆ‘๐‘›=๐‘0๐‘0โˆ’2+๐‘1๐œ‹๐‘›+๐œ‹/2

As ๐‘โ†’โˆž, this sum โˆ‘๐‘›=1๐‘+2|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|โ†’โˆž, by the harmonic series. Then taking sup over all partitions, the sup is unbounded, showing that ๐นโˆ‰๐ต๐‘‰([0,๐‘]), thus ๐นโˆ‰๐ต๐‘‰(๐ผ). Same reasoning when we suppose ๐‘Ž<0 is strict.

โ–ก

Proof

of (b): For ๐‘ฅโ‰ 0: sin(1/๐‘ฅ2) is differentiable as the composition of two differentiable functions, thus differentiable; and ๐น(๐‘ฅ)=๐‘ฅ2sin(1/๐‘ฅ2) is the product of differentiable functions, so ๐น is differentiable.
For ๐‘ฅ=0:

lim๐‘ฅโ†’0๐น(๐‘ฅ)โˆ’๐น(0)๐‘ฅโˆ’0=lim๐‘ฅโ†’0๐‘ฅ2sin(1๐‘ฅ2)๐‘ฅ=lim๐‘ฅโ†’0๐‘ฅsin(1๐‘ฅ2)

Since |sin(1/๐‘ฅ2)|โ‰ค1, we get |๐‘ฅsin(1/๐‘ฅ2)|โ‰ค|๐‘ฅ|โ†’0 as ๐‘ฅโ†’0, thus ๐น is differentiable at ๐‘ฅ=0, and ๐นโ€ฒ(0)=0.
This proves that, ๐น is differentiable everywhere on โ„.
Now we show that ๐นโˆ‰BV([โˆ’1,1]):
Consider this seq:

๐‘ฆ๐‘›โ‰”1๐‘›๐œ‹+๐œ‹/2โ†’0+

we have:

๐น(๐‘ฆ๐‘›)=๐‘ฆ๐‘›2sin(1๐‘ฆ๐‘›2)=1๐‘›๐œ‹+๐œ‹/2โ‹…sin(๐‘›๐œ‹+๐œ‹/2)

For odd ๐‘›, ๐น(๐‘ฆ๐‘›)=โˆ’1๐‘›๐œ‹+๐œ‹/2, for even ๐‘›, ๐น(๐‘ฆ๐‘›)=1๐‘›๐œ‹+๐œ‹/2.
Notice that ๐‘ฆ1<1, so we then consider the partition: pick ๐‘โˆˆโ„•, and use ๐‘ฅ0=0,๐‘ฅ1=๐‘ฆ๐‘,๐‘ฅ2=๐‘ฆ๐‘โˆ’1,โ‹ฏ,๐‘ฅ๐‘=๐‘ฆ1,๐‘ฅ๐‘+1=1 as the partition points of [0,1].
Then we have

๐‘‡๐น(1)โˆ’๐‘‡๐น(โˆ’1)โ‰ฅโˆ‘๐‘›=1๐‘+1|๐น(๐‘ฅ๐‘›)โˆ’๐น(๐‘ฅ๐‘›โˆ’1)|โ‰ฅโˆ‘๐‘›=2๐‘|๐น(๐‘ฆ๐‘›)โˆ’๐น(๐‘ฆ๐‘›โˆ’1)|โ‰ฅโˆ‘๐‘›=2๐‘1๐œ‹๐‘›+๐œ‹/2+1๐œ‹(๐‘›โˆ’1)+๐œ‹/2โ‰ฅ2โˆ‘๐‘›=2๐‘1๐œ‹๐‘›+๐œ‹/2

This sum is unbounded as ๐‘โ†’โˆž by the harmonic series. Then taking sup over all partitions, the sup is unbounded, showing that ๐นโˆ‰๐ต๐‘‰([โˆ’1,1]).

โ–ก

Everywhere unbounded variation

Construct a function ๐นโˆˆ๐ถ00(โ„) (see HW9) such that ๐น does not have bounded variation on any interval [๐‘Ž,๐‘] with ๐‘Ž<๐‘. Hint: construct ๐น based on functions like the ones in the previous problem.

Solution

We consider this function as the building block:

๐บ(๐‘ฅ)={๐‘ฅsin1๐‘ฅ,๐‘ฅโˆˆ(โˆ’1๐œ‹,0)โˆช(0,1๐œ‹)0, elsewhere

We know that, this function is continuous (we know in elementary real analysis course that it is true for ๐‘ฅโˆˆ(โˆ’1๐œ‹,1๐œ‹), and ๐บโ†’0 as ๐‘ฅโ†’ยฑโˆ’1๐œ‹, so it is true all over the domain.) and similar reasoning as question 4(a), we can verift that, ๐บโˆ‰BV(๐ผ) for any ๐ผโˆ‹0.
Also, it is clear that

lim๐‘ฅโ†’ยฑโˆž๐บ(๐‘ฅ)=๐บ(1)=0

Thus we have:

๐บโˆˆ๐ถ00(โ„)

And notice this function has uniform bound 1: setting ๐‘ก=1๐‘ฅ, so ๐‘ฅ=1๐‘ก, and

|๐บ(๐‘ฅ)|=|1๐‘กsin(๐‘ก)|=|sin(๐‘ก)๐‘ก|โ‰ค1โˆ€๐‘กโ‰ 0

So by translating, stretching and scaling it, we can define for each ๐‘›:

๐บ๐‘›(๐‘ฅ)=12๐‘›๐บ(๐‘ฅโˆ’๐‘ฅ๐‘›๐œŽ๐‘›)

where we will delicately choose ๐‘ฅ๐‘›,๐œŽ๐‘›.
By defining the partial sum seq:

๐น๐‘(๐‘ฅ)=โˆ‘๐‘›=1๐‘๐บ๐‘›(๐‘ฅ)

Then by geometric seq, such function is also uniformly bounded by 1, and it is continuous since it is finite sum of continuous functions, and also have ๐น๐‘(๐‘ฅ)โ†’0 as ๐‘ฅโ†’โˆž, so for each ๐‘ we have ๐น๐‘โˆˆ๐ถ00(โ„). And ๐น๐‘ is an increasing seq (not really), so define:

๐นโ‰”lim๐‘โ†’โˆž๐น๐‘=โˆ‘๐‘›=1โˆž๐บ๐‘›

Then ๐น๐‘โ†’๐น uniformly as ๐‘โ†’โˆž. This is since ๐บ๐‘› is uniformly bounded by 12๐‘›: For ๐œ–>0, there exists ๐‘0 s.t. 12๐‘โˆ’1<๐œ–, and then for all ๐‘€โ‰ฅ๐‘0, we have

|๐น๐‘€(๐‘ฅ)โˆ’๐น(๐‘ฅ)|โ‰คโˆ‘๐‘=๐‘0โˆž12๐‘=12๐‘โˆ’1<๐œ–

Thus, we also have

๐นโˆˆ๐ถ00(โ„)

since it is uniform limit of continuous functions, and the limit to ยฑโˆž remains 0. This is regardless of the choice of ๐‘ฅ๐‘›,๐œŽ๐‘› for each ๐‘›.
Then, to finish the construction, it remains for us to choose ๐‘ฅ๐‘›,๐œŽ๐‘› for each ๐‘›, to let ๐น have the property that ๐น does not have bounded variation on any compact interval.
Let {๐‘ฅ๐‘›} be the enumeration of a dense subset of โ„. e.g. Let it be the enumeration of โ„š.
We inductively pick ๐œŽ๐‘›: for each ๐‘›, we pick ๐œŽ๐‘›โˆˆ(0,1) s.t. for all 1โ‰ค๐‘—โ‰ค๐‘›โˆ’1, we have |๐‘ฅ๐‘›โˆ’๐‘ฅ๐‘—|>2๐œŽ๐‘› and |๐‘ฅ๐‘›+1โˆ’๐‘ฅ๐‘›|>2๐œŽ๐‘›.
Now let ๐ผ=[๐‘Ž,๐‘] be an arbitrary compact interval. WTS: ๐นโˆ‰๐ต๐‘‰(๐ผ).
By density of the seq, there exists ๐‘ฅ๐‘› such that ๐‘ฅ๐‘›โˆˆ๐ผ.
We consider the subinterval:

๐ผโ€ฒโ‰”(๐‘ฅ๐‘›โˆ’๐œŽ๐‘›,๐‘ฅ๐‘›+๐œŽ๐‘›)โŠ‚๐ผ

This construction ensures that the ๐บ1,โ‹ฏ,๐บ๐‘›โˆ’1,๐บ๐‘›+1 will not have some offsetting variation such to make the variation of ๐บ๐‘› interfered (suspectively finite): for each 1โ‰ค๐‘—โ‰ค๐‘› and ๐‘—=๐‘›+1, we have:

๐บ๐‘—โˆˆ๐ต๐‘‰(๐ผโ€ฒ)

since ๐‘ฅ๐‘—โˆ‰๐ผโ€ฒ. This is by question 4(a). This means that we can ignore these terms when showing ๐นโˆ‰๐ต๐‘‰(๐ผโ€ฒ).
And for we know that

๐บ๐‘›โˆ‰๐ต๐‘‰(๐ผโ€ฒ)

since ๐‘ฅ๐‘›โˆˆ๐ผ, as verified by question 4(a).
And for the rest ๐บ๐‘›+2,โ‹ฏ, their total variation contributed to this the total variation of ๐น on ๐ผ is at most a half of ๐บ๐‘› (by geometric seq).
Thus the only term matters is ๐บ๐‘›. Since ๐บ๐‘›โˆ‰๐ต๐‘‰(๐ผโ€ฒ), we have ๐น=โˆ‘๐‘›=1โˆž๐บ๐‘›โˆ‰๐ต๐‘‰(๐ผโ€ฒ), thus ๐นโˆ‰๐ต๐‘‰(๐ผ) since ๐ผโŠƒ๐ผโ€ฒ.
This finishes the proof.
(Rigorous reasoning is as question 4, we construct partitions to apply harmonic seq to the variation by the partition, and ๐บ๐‘›+2,โ‹ฏ can at most halve it, which does not matter.)

Homework 12: on absolutely continuous functions (40/40)

Some of the following questions will be graded. Do them, and do hand them in.

Terminologies ็š„ communication: |๐œ‡๐น|=๐œ‡๐‘‡๐น

Let ๐น:โ„โ†’โ„ be a function in NBV. Prove that total variation of the complex measure associated to ๐น is the complex measure associated to the total variation of ๐น. In other words, prove that |๐œ‡๐น|=๐œ‡๐‘‡๐น. Hint: see Exercise 28 in Chapter 3 of [Folland]; proofs by terminology alone are not valid.

Proof

Set:

๐บ(๐‘ฅ):=|๐œ‡๐น|((โˆ’โˆž,๐‘ฅ])

Claim 1: It suffices to show that ๐บ=๐‘‡๐น.
Proof of Claim 1: Since ๐นโˆˆ๐‘๐ต๐‘‰, ๐œ‡๐น is then a complex (regular) Borel measure, as we have shown in class; And by def, |๐œ‡๐น|(๐ธ)=โˆซ๐ธ|๐‘“|๐‘‘๐‘š where ๐‘“=๐‘‘๐œ‡๐น๐‘‘๐‘šโˆˆ๐ฟ1(๐‘š), thus |๐œ‡๐น| is also a complex (regular) Borel measure since it is finite.
Thus ๐บโˆˆ๐‘๐ต๐‘‰ and its association with |๐œ‡๐น| is unique. if ๐บ=๐‘‡๐น, it is then also uniquely associated with ๐œ‡๐‘‡๐น, which implies that ๐œ‡๐‘‡๐น=|๐œ‡๐น|.
Claim 2: ๐บ=๐‘‡๐น Indeed.
Proof of Claim 2:
First we verity that ๐‘‡๐นโ‰ค๐บ:
By def:

๐‘‡๐น(๐‘ฅ)=sup{โˆ‘๐‘—=1๐‘›|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|:๐‘›โˆˆโ„•,โˆ’โˆž<๐‘ฅ0<โ€ฆ<๐‘ฅ๐‘›=๐‘ฅ}=sup{โˆ‘๐‘—=1๐‘›|๐œ‡๐น(โˆ’โˆž,๐‘ฅ๐‘—]โˆ’๐œ‡๐น(โˆ’โˆž,๐‘ฅ๐‘—โˆ’1]|:๐‘›โˆˆโ„•,โˆ’โˆž<๐‘ฅ0<โ€ฆ<๐‘ฅ๐‘›=๐‘ฅ}=sup{โˆ‘๐‘—=1๐‘›|๐œ‡๐น(๐‘ฅ๐‘—,๐‘ฅ๐‘—โˆ’1]|:๐‘›โˆˆโ„•,โˆ’โˆž<๐‘ฅ0<โ€ฆ<๐‘ฅ๐‘›=๐‘ฅ}โ‰คsup{|๐œ‡๐น(โˆ’โˆž,๐‘ฅ0]|+โˆ‘๐‘—=1๐‘›|๐œ‡๐น(๐‘ฅ๐‘—,๐‘ฅ๐‘—โˆ’1]|:๐‘›โˆˆโ„•,โˆ’โˆž<๐‘ฅ0<โ€ฆ<๐‘ฅ๐‘›=๐‘ฅ}โ‰คsup{โˆ‘๐‘—=1๐‘›|๐œ‡๐น(๐ธ๐‘—)|:(โˆ’โˆž,๐‘ฅ]=โจ†๐‘—=1๐‘›๐ธ๐‘—}=|๐œ‡๐น|((โˆ’โˆž,๐‘ฅ])=๐บ(๐‘ฅ)

This proves this direction.
Then we verity that ๐บโ‰ค๐‘‡๐น:
Claim 2.1: |๐œ‡๐น(๐ธ)|=๐œ‡๐‘‡๐น(๐ธ) for all borel set ๐ธ.
First, for h-interval ๐ธ=(๐‘Ž,๐‘], we have:

|๐œ‡๐น(๐ธ)|=|๐œ‡๐น(๐‘Ž,๐‘]|=|๐œ‡๐น(โˆ’โˆž,๐‘]โˆ’๐œ‡๐น(โˆ’โˆž,๐‘Ž]|=|๐น(๐‘)โˆ’๐น(๐‘Ž)|โ‰คsup{โˆ‘๐‘—=1๐‘›|๐น(๐‘ฅ๐‘—)โˆ’๐น(๐‘ฅ๐‘—โˆ’1)|:๐‘›โˆˆโ„•,๐‘Ž=๐‘ฅ0<โ€ฆ<๐‘ฅ๐‘›=๐‘},by tri ineq=๐‘‡๐น(๐‘)โˆ’๐‘‡๐น(๐‘Ž)=๐œ‡๐‘‡๐น(โˆ’โˆž,๐‘]โˆ’๐œ‡๐‘‡๐น(โˆ’โˆž,๐‘Ž]=๐œ‡๐‘‡๐น(๐‘Ž,๐‘]=๐œ‡๐‘‡๐น(๐ธ)

Also for intervals like (โˆ’โˆž,๐‘], we have

|๐œ‡๐น((โˆ’โˆž,๐‘])|=|โˆ‘๐‘˜=1โˆž๐œ‡๐น((๐‘โˆ’๐‘˜,๐‘+1โˆ’๐‘˜])|โ‰คโˆ‘๐‘˜=1โˆž|๐œ‡๐น((๐‘โˆ’๐‘˜,๐‘+1โˆ’๐‘˜])|โ‰คโˆ‘๐‘˜=1โˆž๐œ‡๐‘‡๐น((๐‘โˆ’๐‘˜,๐‘+1โˆ’๐‘˜])=๐œ‡๐‘‡๐น((โˆ’โˆž,๐‘])

Thus |๐œ‡๐น(๐ธ)|=๐œ‡๐‘‡๐น(๐ธ) is true for all left-open, right-closed intervals ๐ธ, and thus also true for all finite disjoint unions of left-open, right-closed intervals. Notice that, the set of all finite disjoint unions of left-open, right-closed intervals is an algebra, we denote it by ๐’œ๏ธ€. So

|๐œ‡๐น(๐ธ)|โ‰ค๐œ‡๐‘‡๐น(๐ธ),โˆ€๐ธโˆˆ๐’œ๏ธ€

Now we define:

๐’ž๏ธ€โ‰”{๐ธโˆˆโ„ฌ๏ธ€(โ„):|๐œ‡๐น(๐ธ)|โ‰ค๐œ‡๐‘‡๐น(๐ธ)}

Then we have:

๐’œ๏ธ€โŠ‚๐’ž๏ธ€

Notice that increasing sequence (๐ธ๐‘˜)๐‘˜=1โˆž in ๐’ž๏ธ€, we have:

|๐œ‡๐น(โ‹ƒ๐‘˜=1โˆž๐ธ๐‘˜)|=|๐œ‡๐น(โจ†๐‘˜=1โˆž(๐ธ๐‘˜\โ‹ƒ๐‘—=1๐‘˜โˆ’1๐ธ๐‘—))|โ‰คโˆ‘๐‘˜=1โˆž|๐œ‡๐น(๐ธ๐‘˜\โ‹ƒ๐‘—=1๐‘˜โˆ’1๐ธ๐‘—)|โ‰คโˆ‘๐‘˜=1โˆž๐œ‡๐‘‡๐น(๐ธ๐‘˜\โ‹ƒ๐‘—=1๐‘˜โˆ’1๐ธ๐‘—)=๐œ‡๐‘‡๐น(โ‹ƒ๐‘˜=1โˆž๐ธ๐‘˜)

Showing that ๐’ž๏ธ€ is closed under countable increasing unions. Similarly, ๐’ž๏ธ€ is closed under countable decreasing intersections. This shows that ๐’ž๏ธ€ is a monotone class. Since ๐’ž๏ธ€โŠƒ๐’œ๏ธ€ which is an algebra that generates the ๐œŽ-algebra โ„ฌ๏ธ€(โ„), we have by the monotone class lemma:

โ„ฌ๏ธ€(โ„)โŠ‚๐’ž๏ธ€

This finishes the proof that |๐œ‡๐น(๐ธ)|=๐œ‡๐‘‡๐น(๐ธ) for all borel set ๐ธ.

Then we have:

|๐œ‡๐น|(๐ธ)=sup{โˆ‘๐‘˜=1โˆž|๐œ‡๐น(๐ธ๐‘˜)|:๐ธ=โจ†๐‘˜=1โˆž๐ธ๐‘˜}โ‰คsup{โˆ‘๐‘˜=1โˆž๐œ‡๐‘‡๐น(๐ธ๐‘˜):๐ธ=โจ†๐‘˜=1โˆž๐ธ๐‘˜}=sup{๐œ‡๐‘‡๐น(๐ธ)}=๐œ‡๐‘‡๐น(๐ธ)

Therefore we have

๐บโ‰ค๐‘‡๐น

Combining both directions we have

๐บ=๐‘‡๐น

which shows by Claim 1 that

|๐œ‡๐น|=๐œ‡๐‘‡๐น

โ–ก

Characterization of Lipschitz continuity: ๐ด๐ถ + bounded derivativeโ‡”Lipschitz continuity:

Consider a function ๐น:โ„โ†’โ„. Show that |๐น(๐‘ฅ)โˆ’๐น(๐‘ฆ)|โ‰ค๐‘€|๐‘ฅโˆ’๐‘ฆ| for all ๐‘ฅ,๐‘ฆ (i.e. ๐น is Lipschitz continuous with Lipschitz constant at most ๐‘€) iff ๐น is absolutely continuous, and |๐นโ€ฒ(๐‘ฅ)|โ‰ค๐‘€ for Lebesgue a.e.ย ๐‘ฅ.

Proof

Forward Direction (โŸน): Suppose ๐น is Lipschitz continuous, and take Lipschitz constant ๐‘€>0 such that |๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|โ‰ค๐‘€|๐‘ฆโˆ’๐‘ฅ| for all ๐‘ฅ,๐‘ฆโˆˆโ„.
Let ๐œ–>0.
Let (๐‘Ž1,๐‘1),(๐‘Ž2,๐‘2),โ€ฆ,(๐‘Ž๐‘›,๐‘๐‘›) be a finite collection of disjoint intervals with โˆ‘๐‘˜=1๐‘›(๐‘๐‘˜โˆ’๐‘Ž๐‘˜)<๐œ–๐‘€ then we have:

โˆ‘๐‘˜=1๐‘›|๐น(๐‘๐‘˜)โˆ’๐น(๐‘Ž๐‘˜)|โ‰คโˆ‘๐‘˜=1๐‘›๐‘€|๐‘๐‘˜โˆ’๐‘Ž๐‘˜|=๐‘€โˆ‘๐‘˜=1๐‘š(๐‘๐‘˜โˆ’๐‘Ž๐‘˜)<๐‘€๐œ€๐‘€=๐œ€

This shows that ๐น is absolutely continuous. And since ๐น is absolutely continuous, its restriction on any compact interval is of bounded variation, thus differentiable a.e.; thus ๐น is differentiable ๐‘š-a.e.
Then for ๐‘š-a.e. ๐‘ฅโˆˆโ„, we have:

|๐นโ€ฒ(๐‘ฅ)|=|lim๐‘ฆโ†’๐‘ฅ๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)๐‘ฆโˆ’๐‘ฅ|=lim๐‘ฆโ†’๐‘ฅ|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)||๐‘ฆโˆ’๐‘ฅ|โ‰คlim๐‘ฆโ†’๐‘ฅ๐‘€|๐‘ฆโˆ’๐‘ฅ||๐‘ฆโˆ’๐‘ฅ|=lim๐‘ฆโ†’๐‘ฅ๐‘€=๐‘€

This finishes the proof of the forward direction.
Backward Direction (โŸน): Suppose ๐น is absolutely continuous, and |๐นโ€ฒ(๐‘ฅ)|โ‰ค๐‘€ for ๐‘š-a.e.ย ๐‘ฅ.
Let ๐‘ฅ,๐‘ฆโˆˆโ„ and ๐‘ฅโ‰ค๐‘ฆ then on [๐‘ฅ,๐‘ฆ] we have:

|๐น(๐‘ฆ)โˆ’๐น(๐‘ฅ)|=|โˆซ๐‘ฅ๐‘ฆ๐นโ€ฒ๐‘‘๐‘š|โ‰คโˆซ๐‘ฅ๐‘ฆ|๐นโ€ฒ|๐‘‘๐‘šโ‰คโˆซ๐‘ฅ๐‘ฆ๐‘€๐‘‘๐‘š=๐‘€(๐‘ฆโˆ’๐‘ฅ)=๐‘€|๐‘ฆโˆ’๐‘ฅ|

Therefore ๐น is Lipschitz continuous with Lipschitz constant ๐‘€.

โ–ก

๐ด๐ถ function ไฟ็•™ null sets

Let ๐น:โ„โ†’โ„ be an absolutely continuous function. Prove that ๐น maps null sets to null sets. In other words, if ๐ธโŠ‚โ„ is a set of Lebesgue measure zero, then ๐น(๐ธ)={๐น(๐‘ฅ)โˆฃ๐‘ฅโˆˆ๐ธ} is also of Lebesgue measure zero. (In particular, ๐น(๐ธ) is Lebesgue measurable, cf.ย HW4#6.)

Proof

Fix ๐น:โ„โ†’โ„ abs ctn, and ๐ธโŠ‚โ„ s.t. ๐‘š(๐ธ)=0.
Let ๐œ–>0.
Since ๐นโˆˆ๐ด๐ถ, there exists some ๐›ฟ>0 s.t. for any disjoint intervals (๐‘Ž1,๐‘1),โ‹ฏ,(๐‘Ž๐‘›,๐‘๐‘›) s.t. โˆ‘1๐‘›(๐‘๐‘—โˆ’๐‘Ž๐‘—)<๐›ฟ, we have: โˆ‘1๐‘|๐น(๐‘๐‘—)โˆ’๐น(๐‘Ž๐‘—)|<๐œ–.
Fix this ๐›ฟ. Since ๐‘š(๐ธ)=0, there exists finite collection of bounded open intervals (๐‘1,๐‘‘1),โ‹ฏ.(๐‘๐‘›,๐‘‘๐‘›) such that

๐ธโŠ‚โ‹ƒ1๐‘›(๐‘๐‘—,๐‘‘๐‘—)

with

โˆ‘1๐‘›๐‘š(๐‘๐‘—,๐‘‘๐‘—)=โˆ‘1๐‘›(๐‘‘๐‘—โˆ’๐‘๐‘—)<๐›ฟ

Notice that, though these open intervals are not necessarily disjoint, but finite union of bounded open intervals can be expressed as finite union of disjoint open intervals. We just need to connect those open intervals that has intersection.
By doing this, we get some disjoint intervals (๐‘Ž1,๐‘1),โ‹ฏ,(๐‘Ž๐‘,๐‘๐‘) from (๐‘1,๐‘‘1),โ‹ฏ.(๐‘๐‘›,๐‘‘๐‘›), with

๐ธโŠ‚โ‹ƒ1๐‘(๐‘Ž๐‘—,๐‘๐‘—)=โ‹ƒ1๐‘›(๐‘๐‘—,๐‘‘๐‘—)

and (since new intervals remove the intersection part and keep the union:)

โˆ‘1๐‘๐‘š(๐‘Ž๐‘—,๐‘๐‘—)โ‰คโˆ‘1๐‘›๐‘š(๐‘๐‘—,๐‘‘๐‘—)<๐›ฟ

Now we can apply the absolute continuity. Since ๐นโˆˆ๐ด๐ถ, it is continuous for sure. Thus on [๐‘Ž๐‘—,๐‘๐‘—], it takes max and min value respectively on some ๐‘ฅ๐‘—,๐‘ฆ๐‘—โˆˆ[๐‘Ž๐‘—,๐‘๐‘—]. Then

๐น([๐‘Ž๐‘—,๐‘๐‘—])=[๐น(๐‘ฆ๐‘—),๐น(๐‘ฅ๐‘—)]

So

๐น((๐‘Ž๐‘—,๐‘๐‘—))โŠ‚[๐น(๐‘ฆ๐‘—),๐น(๐‘ฅ๐‘—)]

This is by the intermediate value theorem. We denote the open interval using ๐‘ฅ๐‘—,๐‘ฆ๐‘— as endpoints as ๐ผ๐‘—. We then have ๐ผ๐‘—โŠ‚[๐‘Ž๐‘—,๐‘๐‘—].
Thus

โˆ‘1๐‘|๐ผ๐‘—|<๐›ฟ

and by abs ctnity, we have :

โˆ‘1๐‘|๐น(๐‘ฆ๐‘—)โˆ’๐น(๐‘ฅ๐‘—)|<๐œ–

Since ๐ธโŠ‚โ‹ƒ1๐‘(๐‘Ž๐‘–,๐‘๐‘–), we have

๐น(๐ธ)โŠ‚๐น(โ‹ƒ1๐‘(๐‘Ž๐‘–,๐‘๐‘–))=โ‹ƒ1๐‘๐น((๐‘Ž๐‘–,๐‘๐‘–))โŠ‚โ‹ƒ1๐‘[๐น(๐‘ฆ๐‘—),๐น(๐‘ฅ๐‘—)]

so we then have

๐‘š(๐น(๐ธ))โ‰ค๐‘š(โ‹ƒ1๐‘[๐น(๐‘ฆ๐‘—),๐น(๐‘ฅ๐‘—)])โ‰คโˆ‘1๐‘๐‘š([๐น(๐‘ฆ๐‘—),๐น(๐‘ฅ๐‘—)])=โˆ‘1๐‘|๐น(๐‘ฆ๐‘—)โˆ’๐น(๐‘ฅ๐‘—)|<๐œ–

Since ๐œ–>0 is arbitrary, this finishes the proof that

๐‘š(๐น(๐ธ))=0

โ–ก

๐ต๐‘‰ function ๆฏ็‚น็š„ left right limit ไธ€ๅฎšๅญ˜ๅœจ

Prove directly from the definition that if ๐น:โ„โ†’โ„ is a function of bounded variation, then ๐น admits a left and a right limit at every point. In other words, for any ๐‘Žโˆˆโ„, the limits

lim๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ)andlim๐‘ฅโ†’๐‘Žโˆ’๐น(๐‘ฅ)

both exist. Do not use the Jordan decomposition. Hint: as is often the case, limits can be studied through limsup and liminf.

Proof

Let ๐น:โ„โ†’โ„ be a function of bounded variation, fix ๐‘Žโˆˆโ„.
Define

๐ฟโ‰”limโ€‰sup๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ),๐‘™โ‰”limโ€‰inf๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ)

Then we have ๐ฟโ‰ฅ๐‘™. We will show ๐ฟ=๐‘™.
Let ๐œ–>0.
Suppose for contradiction that ๐ฟ>๐‘™+๐œ–.
Let ๐‘Ž๐‘›โ†’๐‘Ž be a seq. By the def of limsup and lininf, there must exists a subseq ๐‘Ž๐‘›๐‘— such that for some ๐‘1, we have:

|๐ฟโˆ’๐น(๐‘Ž๐‘›๐‘—)|<๐œ–4,โˆ€๐‘—โ‰ฅ๐‘1

And there must exists a subseq ๐‘Ž๐‘š๐‘˜ such that for some ๐‘2, we have:

|๐น(๐‘Ž๐‘š๐‘˜)โˆ’๐‘™|<๐œ–4,โˆ€๐‘˜โ‰ฅ๐‘2

Then for all ๐‘—,๐‘˜โ‰ฅmax(๐‘1,๐‘2) we have:

|๐น(๐‘Ž๐‘›๐‘—)โˆ’๐น(๐‘Ž๐‘š๐‘˜)|โ‰ฅ|๐ฟโˆ’๐‘™|โˆ’|๐ฟโˆ’๐น(๐‘Ž๐‘›๐‘—)|โˆ’|๐น(๐‘Ž๐‘š๐‘˜)โˆ’๐‘™|>๐œ–2

Notice: for any ๐‘—โ‰ฅmax(๐‘1,๐‘2) and given start ๐พ0โˆˆโ„•, there exists some ๐‘˜โ‰ฅmax(๐พ0,๐‘1,๐‘2) s.t.

๐‘Ž๐‘š๐‘˜<๐‘Ž๐‘›๐‘—

This is because ๐‘Ž๐‘š๐‘˜โ†’๐‘Ž as ๐‘˜โ†’โˆž.
And this is same on the ๐‘˜ side.

Figureย 40: unbounded total variation by alternating limsup/inf seq

Thus, by picking ๐‘—0=max(๐‘1,๐‘2), we can pick ๐‘˜0 s.t. ๐‘Ž๐‘š๐‘˜0<๐‘Ž๐‘›๐‘—0, and then pick ๐‘—1 s.t. ๐‘Ž๐‘š๐‘—1<๐‘Ž๐‘›๐‘˜0 ; and inductively, for the pick of ๐‘—๐‘, we can always pick ๐‘˜๐‘ s.t. ๐‘Ž๐‘š๐‘˜๐‘<๐‘Ž๐‘›๐‘—๐‘ an then pick ๐‘Ž๐‘š๐‘—๐‘+1<๐‘Ž๐‘›๐‘˜๐‘.
We do this process to get the finite seq ๐‘—0,๐‘˜0,๐‘—1,๐‘˜1,โ‹ฏ,๐‘—๐‘,๐‘˜๐‘ for some int ๐‘. Then we have:

๐‘‡๐น([๐‘Ž,๐‘Ž๐‘›๐‘—0])โ‰ฅ|๐น(๐‘Ž๐‘›๐‘—0)โˆ’๐น(๐‘Ž๐‘š๐‘˜0)|+|๐น(๐‘Ž๐‘›๐‘˜0)โˆ’๐น(๐‘Ž๐‘š๐‘—1)|+โ‹ฏ+|๐น(๐‘Ž๐‘›๐‘—๐‘)โˆ’๐น(๐‘Ž๐‘š๐‘˜๐‘)|+|๐น(๐‘Ž๐‘›๐‘˜๐‘)โˆ’๐น(๐‘Ž)|โ‰ฅ๐‘๐œ–2

As ๐‘โ†’โˆž, we have ๐‘‡๐น([๐‘Ž,๐‘Ž๐‘—0])โ‰ฅ๐‘๐œ–2โ†’โˆž. Thus by def, ๐‘‡๐น([๐‘Ž,๐‘Ž๐‘—0])=โˆž, contradicting the assumption that ๐น is a function of bounded variation.
Thus by contradiction, it shows that

๐ฟโ‰ค๐‘™+๐œ–

Since ๐ฟโ‰ฅ๐‘™ and ๐œ–>0 is arbitrary, this finishes the proof that

๐ฟ=๐‘™

Since we have limโ€‰sup๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ)=limโ€‰inf๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ), we then have:

lim๐‘ฅโ†’๐‘Ž+๐น(๐‘ฅ)โˆƒ

By same reasoning, we can get that

lim๐‘ฅโ†’๐‘Žโˆ’๐น(๐‘ฅ)โˆƒ

โ–ก

๐ด๐ถ๐ฟ1 ๅ‡ฝๆ•ฐ็š„ๅฏผๆ•ฐ็ปๅฏนๅ€ผ็š„ๆ€ป็งฏๅˆ†ไธบ 0โŸน๐‘“=0

Let ๐‘“:โ„โ†’โ„ be an absolutely continuous function. Assume that ๐‘“โˆˆ๐ฟ1(โ„), and that

lim๐‘กโ†’0+โˆซโˆ’โˆžโˆž|๐‘“(๐‘ฅ+๐‘ก)โˆ’๐‘“(๐‘ฅ)๐‘ก|๐‘‘๐‘ฅ=0.

Prove that ๐‘“=0. Hint: consult Fatou Samba but ignore any dance moves.

Proof

We define:

๐ท๐‘ก(๐‘ฅ):=๐‘“(๐‘ฅ+๐‘ก)โˆ’๐‘“(๐‘ฅ)๐‘ก

Since ๐‘“โˆˆ๐ด๐ถ, we have that ๐‘“โ€ฒโˆˆ๐ฟ1(๐‘š) exists a.e., thus by def of derivative we have: So we take a seq of functions ๐‘”๐‘›:=|๐ท1/๐‘›|, we then have:

lim๐‘›โ†’โˆž๐‘”๐‘›=lim๐‘กโ†’0+|๐ท๐‘ก|=|๐‘“โ€ฒ| a.e.

Notice we are given the condition that:

lim๐‘กโ†’0+โˆฅ๐ท๐‘กโˆฅ1=lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›=0

Since fixing ๐‘ก, ๐‘“(๐‘ฅ+๐‘ก) and ๐‘“(๐‘ฅ) are measurable functions, ๐ท๐‘ก is also measurable, and thus ๐‘”๐‘›โˆˆ๐ฟ+(๐‘š) for each ๐‘›. (we can ignore the points where the limit does not exist, since the set of these points has Lebesgue measure 0.)
Applying Fatouโ€™s Lemma we have:

โˆซlimโ€‰inf๐‘›โ†’โˆž๐‘”๐‘›๐‘‘๐‘ฅโ‰คlimโ€‰inf๐‘›โ†’โˆžโˆซ๐‘”๐‘›๐‘‘๐‘ฅ=lim๐‘›โ†’โˆžโˆซ๐‘”๐‘›=0

Since ๐‘”๐‘› and limโ€‰inf๐‘›โ†’โˆž๐‘”๐‘›=lim๐‘›โ†’โˆž๐‘”๐‘› are nonnegative, we have:

|๐‘“โ€ฒ|=lim๐‘›โ†’โˆž๐‘”๐‘›=0a.e.

Thus

๐‘“โ€ฒ=0a.e.

Since by AC, we can apply FTC: Let [๐‘Ž,๐‘] be an arbitrary interval, then by FTC we have:

๐‘“(๐‘ฅ)โˆ’๐‘“(๐‘Ž)=โˆซ๐‘Ž๐‘ฅ0๐‘‘๐‘ฆ=0,โˆ€๐‘ฅโˆˆ[๐‘Ž,๐‘]

Thus

๐‘“(๐‘ฅ)=๐‘“(๐‘Ž),โˆ€๐‘ฅโˆˆ[๐‘Ž,๐‘]

Since the interval [๐‘Ž,๐‘] is arbitrary, this proves: ๐‘“ is a constant function. (By taking ๐ผ๐‘›:=[โˆ’๐‘›,๐‘›] over ๐‘›โˆˆโ„•, we can get ๐‘“(๐‘ฅ)=0 for all ๐‘ฅโˆˆโ„.)
Suppose for contradiction that ๐‘“=๐‘โ‰ 0, then

โˆซ|๐‘“|=โˆซโ„|๐‘|=โˆž

contradicting ๐‘“โˆˆ๐ฟ1(๐‘š), thus we have

๐‘“=0

This finishes the proof.

โ–ก

13 the dual of ๐ฟ๐‘ spaces

13.1 the dual of ๐ฟ๐‘-I [Fol 6.2]

ๅฏนๅบ”: Folland 5.1, 6.2.
(ๅŽŸๆœฌ่ฟ™ๆ˜ฏๅœจ lec 25 ็š„ไฝ็ฝฎ่ฎฒ็š„, ไฝ†ๆ˜ฏๅฝ“ๆ—ถ็”ฑไบŽๆฒกๆœ‰ Radon-Nikodym Thm, ๆฒกๆœ‰่ถณๅคŸ็š„ๅทฅๅ…ทๅŽปๅฎŒๆˆ

(๐ฟ๐‘)โˆ—=๐ฟ๐‘ž

็š„่ฏๆ˜Ž (ๅทฎไบ†ไธ€ไธช proof surjectivity of the isometry ๐‘”โ†’โ„“๐‘”). ๅ› ่€Œๆˆ‘ๆŠŠๅฎƒๆ”พๅœจ่ฟ™้‡Œ, ่ก”ๆŽฅไธ‹้ขๅ‡ ไธช lectures, ๅฎŒๆˆ 6.2 ่ฟ™ไธ€่Š‚.
ๆˆ‘ไปฌ้ฆ–ๅ…ˆ็ปƒไน ไธ€ไธช example of Hรถlderโ€™s ineq ๆฅๅ›žๅฟ†ไธ€ไธ‹:
recall Hรถlderโ€™s ineq: for 1โ‰ค๐‘,๐‘žโ‰คโˆž,1๐‘+1๐‘ž=1โŸน

||๐‘“๐‘”||1โ‰ค||๐‘“||๐‘||๐‘”||๐‘ž
Example 13.41

Prove:

๐‘“โˆˆ๐ฟ3([โˆ’1,1],๐‘š)โŸนโˆซโˆ’11|๐‘“(๐‘ฅ)||๐‘ฅ|๐‘‘๐‘ฅ<โˆž
Proof

Apply Hรถlderโ€™s: ๆ—ข็„ถ ๐‘“โˆˆ๐ฟ3, ้‚ฃไนˆๆˆ‘ไปฌๅฐฑๆ‹‰ๆปก, take ๐‘=3, correspondingly ๐‘ž=3/2:

โˆซโˆ’11|๐‘“(๐‘ฅ)||๐‘ฅ|๐‘‘๐‘ฅโ‰ค(โˆซโˆ’11|๐‘“(๐‘ฅ)|3๐‘‘๐‘ฅ)13(โˆซโˆ’111|๐‘ฅ|34๐‘‘๐‘ฅ)23

both integrals evaluate <โˆž

โ–ก

13.1.1 intro to dual space

่ฟ™้‡Œๅช่ฎจ่ฎบ ๐•‚โ‰”โ„ or โ„‚.
recall, ๅฏนไบŽไธ€ไธช ๐•‚-vector space ๐‘‰, ไธ€ไธช linear functional of ๐‘‰ ๅฐฑๆ˜ฏไธ€ไธช linear function

๐‘“:๐‘‰โ†’๐•‚

ๅฏนไบŽไฝœไธบ NVS ็š„ ๐‘‰, ๆˆ‘ไปฌ่ฟ˜ๅฏไปฅๅฎšไน‰ไธ€ไธช linear functional ็š„ boundedness.

Definition 13.68 : bounded linear functional

Let ๐‘‰ be a ๐•‚-NVS, ๐‘“:๐‘‰โ†’๐•‚ be a linear functional.
ๆˆ‘ไปฌ็งฐ ๐‘“ bounded, if exist ๐ถ>0 s.t.

|๐‘“(๐‘ฃ)|โ‰ค๐ถ||๐‘ฃ||,โˆ€๐‘ฃโˆˆ๐‘‰
Proposition 13.36 : linear functional bounded โ‡” ctn at 0

if ๐‘“:๐‘‰โ†’๐•‚ is a linear functional, TFAE:

  • ๐‘“ bounded

  • ๐‘“ continuous

  • ๐‘“ continuous at 0โˆˆ๐‘‰

Proof

(ii) to (iii): trivial.
(i) to (ii): ๅ‡่ฎพ ๐‘“ bounded, ้‚ฃไนˆๅฏไปฅ pick ๐ถ s.t. |๐‘“(๐‘ฃ)|โ‰ค๐ถ||๐‘ฃ||.
Pick ๐‘ฃ0โˆˆ๐‘‰,๐œ–>0. Set ๐›ฟโ‰”๐œ–๐ถ. Then

||๐‘ฃโˆ’๐‘ฃ0||<๐›ฟโŸน|๐‘“(๐‘ฃ)โˆ’๐‘“(๐‘ฃ0)|=|๐‘“(๐‘ฃโˆ’๐‘ฃ0)|โ‰ค๐ถ||๐‘ฃโˆ’๐‘ฃ0||<๐œ–

ไปŽ่€Œ ctn.
(iii) to (i): โˆƒ๐›ฟ>0 s.t. ||๐‘ฃ||โ‰ค๐›ฟโŸน|๐‘“(๐‘ฃ)|โ‰ค1.
ไบŽๆ˜ฏ โˆ€๐‘ฃโˆˆ๐‘‰\{0}, ้ƒฝๆœ‰

|๐‘“(๐‘ฃ)|=|๐‘“(๐‘ฃโ‹…๐›ฟ||๐‘ฃ||)|๐›ฟ||๐‘ฃ||โ‰ค๐›ฟ||๐‘ฃ||

taking ๐ถ=1๐›ฟ, ๅพ—ๅˆฐ boundedness.

โ–ก

Definition 13.69 : dual space

If ๐‘‰ is a NVS, ๆˆ‘ไปฌๅฎšไน‰ๅฎƒ็š„ dual space as:

๐‘‰โˆ—โ‰”{bounded linear functionals ๐‘“:๐‘‰โ†’๐•‚}
Definition 13.70 : norm of dual space: ๅณ dual norm

Given ๐‘“โˆˆ๐‘‰โˆ—, set

||๐‘“||โˆ—:=sup๐‘ฃโˆˆ๐‘‰\{0}|๐‘“(๐‘ฃ)|||๐‘ฃ||=sup||๐‘ฃ||=1|๐‘“(๐‘ฃ)|

where โˆฅ๐‘ฃโˆฅ ่กจ็คบ็š„ๆ˜ฏ ๐‘‰ ไธŠไฝฟ็”จ็š„ norm. ่ฟ™ไธช norm ่ขซ็งฐไธบ dual norm.

่ฟ™ไธชๅฝขๅผๆ˜ฏๆˆ‘ไปฌๅœจๅ„็งๅœฐๆ–น่ง่ฟ‡้žๅธธๅคšๆฌก็š„ operator norm, ๅชไธ่ฟ‡่ฟ™้‡Œ, ๆŒ‡ๅฎšไธ€ไธช NVS, ๅฏนไบŽๅ…ถ dual space ไธŠ็š„ linear functional, ๅฎƒๆ˜ฏๅ›บๅฎš็š„, ไธ้œ€่ฆๆŒ‡ๅฎš ๐‘ฃ ๅ’Œ ๐‘“(๐‘ฃ)ไฝฟ็”จๅ“ชไธช norm, ๅ› ไธบ ๐‘“(๐‘ฃ) ๅฐฑๆ˜ฏๆ ‡้‡, ่€Œ ๐‘ฃ from ๅŽŸ NVS, ๅทฒ็ปๆŒ‡ๅฎšๅฅฝ norm.

13.1.2 ๐‘‰โˆ— being a Banach space

Theorem 13.71 : dual space is always Banach

ๅฏนไบŽไปปๆ„็š„ NVS ๐‘‰: ๐‘‰โˆ— ้ƒฝๆ˜ฏไธ€ไธช Banach space. (not assuming ๐‘‰ Banach).

Proof

First we can confirm ๐‘‰โˆ— is a VS, ๅ› ไธบๅฎƒ็”ฑ linear functions of the same size ็ป„ๆˆ.
Claim 1: ๐‘‰โˆ— ๆ˜ฏไธ€ไธช NVS.
ๅ› ไธบไปปๅ– ๐‘ฃโˆˆ๐‘‰,๐œ†โˆˆ๐•‚ ้ƒฝๆœ‰ |๐‘“(๐œ†๐‘ฃ)|=|๐œ†|โ‹…|๐‘“(๐‘ฃ)|, ไปŽ่€Œ

๐‘“โˆˆ๐‘‰โˆ—,๐œ†โˆˆ๐•‚โŸน||๐œ†๐‘“||โˆ—=|๐œ†|โ‹…||๐‘“||โˆ—

ไปฅๅŠ

๐‘“,๐‘”โˆˆ๐‘‰โˆ—,๐‘ฃโˆˆ๐‘‰โŸน|(๐‘“+๐‘”)(๐‘ฃ)|=|๐‘“(๐‘ฃ)+๐‘”(๐‘ฃ)|โ‰ค|๐‘“(๐‘ฃ)|+|๐‘”(๐‘ฃ)|

ๅ› ่€Œ

๐‘“,๐‘”โˆˆ๐‘‰โˆ—โŸนโˆฅ๐‘“+๐‘”โˆฅโˆ—โ‰คโˆฅ๐‘“โˆฅโˆ—+โˆฅ๐‘”โˆฅโˆ—

ไธ‹้ขๆˆ‘ไปฌ verify ๐‘‰โˆ— Banach.
Claim 2: ไธ€ไธช Cauchy seq in ๐‘‰โˆ— ไธ€ๅฎš pointwise converge to some ๐‘“.
Pick (๐‘“๐‘›)1โˆž, ไธ€ไธช Cauchy seq in ๐‘‰โˆ—. Let ๐œ–<0, ๅญ˜ๅœจ ๐‘ ไฝฟๅพ—ๅฏนไบŽไปปๆ„ ๐‘š,๐‘›โ‰ฅ๐‘ ้ƒฝๆœ‰ โˆฅ๐‘“๐‘›โˆ’๐‘“๐‘šโˆฅโˆ—<๐œ–, ๆˆ‘ไปฌ็ฎ€ๅ†™ไธบ:

โˆฅ๐‘“๐‘›โˆ’๐‘“๐‘šโˆฅโˆ—โ†’0

ๅ› ่€ŒๅฏนไบŽไปปๆ„ ๐‘ฃโˆˆ๐‘‰, we have

|๐‘“๐‘›(๐‘ฃ)โˆ’๐‘“๐‘š(๐‘ฃ)|โ‰คโˆฅ๐‘“๐‘›โˆ’๐‘“๐‘šโˆฅโˆ—||๐‘ฃ||โ†’0

ๅนถไธ”ๆˆ‘ไปฌ็Ÿฅ้“ ๐•‚ ๆ˜ฏ complete ็š„, ๅ› ่€Œ ๐‘“๐‘›(๐‘ฃ) converges in ๐•‚ to some element, declared to be ๐‘“(๐‘ฃ).
ๅณ ๐‘“๐‘›โ†’๐‘“ pointwisely:

lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฃ)=๐‘“(๐‘ฃ)

(่ฟ™ๆ˜ฏ่‡ช็„ถ็š„, ๅ› ไธบๅฆ‚ๆžœ linear function ๐‘“โˆ’๐‘” ็š„ operator norm ๆ˜ฏ 0, ้‚ฃไนˆ่ฏดๆ˜Žๅฎƒไปฌๆฏซๆ— ๅทฎๅˆซ, ๅฆๅˆ™ไธ€ๅฎšๆœ‰ๆŸไธชๅœฐๆ–น ๐‘“,๐‘” ็š„ image ไธไธ€ๆ ท, ไฝฟๅพ—่ฟ™ไธช norm ไธๆ˜ฏ 0.)
Claim 3: ๐‘“ ๆ˜ฏ linear ็š„, ๅนถไธ” bounded (ไปŽ่€Œ ctn), ๅณ ๐‘“โˆˆ๐‘‰โˆ—.
linearity: ็”ฑไบŽๆฏไธช ๐‘“๐‘› ้ƒฝๆ˜ฏ linear ็š„,

๐‘“๐‘›(๐‘ฅ+๐›ผ๐‘ฆ)=๐‘“๐‘›(๐‘ฅ)+๐›ผ๐‘“๐‘›(๐‘ฆ)

ๅ› ่€Œ

๐‘“(๐‘ฅ+๐›ผ๐‘ฆ)=lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ+๐›ผ๐‘ฆ)=lim๐‘›โ†’โˆž(๐‘“๐‘›(๐‘ฅ)+๐›ผ๐‘“๐‘›(๐‘ฆ))=lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)+๐›ผlim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฆ)=๐‘“(๐‘ฅ)+๐›ผ๐‘“(๐‘ฆ)

ๅ› ๆญค ๐‘“ ๆ˜ฏ็บฟๆ€ง็š„.
(Note: ่ฟ™้‡Œ่ฏๆ˜Žไบ† linear map ็š„ pointwise ๆž้™ไธ€ๅฎšไนŸๆ˜ฏ linear map.)
Boundedness: Note a standard fact from metric spaces: every Cauchy sequence is bounded.
ๅ› ่€Œ ๐‘“๐‘› ๆ˜ฏไธ€ไธช bounded seq, ๅณๅญ˜ๅœจ ๐‘€>0 such that โˆฅ๐‘“๐‘›โˆฅโ‰ค๐‘€ for all ๐‘›. Then

|๐‘“(๐‘ฅ)|=|lim๐‘›โ†’โˆž๐‘“๐‘›(๐‘ฅ)|โ‰คlim๐‘›โ†’โˆž|๐‘“๐‘›(๐‘ฅ)|โ‰คlim๐‘›โ†’โˆžโˆฅ๐‘“๐‘›โˆฅโˆ—โˆฅ๐‘ฅโˆฅโ‰ค๐‘€โˆฅ๐‘ฅโˆฅ

Hence ๐‘“ is bounded (continuous), and โˆฅ๐‘“โˆฅโˆ—โ‰ค๐‘€. Claim 4: โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅโˆ—โ†’0, proving ๐‘‰โˆ— ๆ˜ฏ Banach ็š„. WTS:

โˆฅ๐‘“๐‘›โˆ’๐‘“โˆฅ=supโˆฅ๐‘ฅโˆฅ=1|(๐‘“๐‘›โˆ’๐‘“)(๐‘ฅ)|โ†’0

//TO BE DONE.

โ–ก

Actually ่ฟ™ไธช Theorem ๆœ‰ๆ›ด general ็š„ๅฝขๅผ:

Theorem 13.72

ๅฏนไบŽไปปๆ„ nvm ๐‘‰ ๅ’Œ Banach ๐‘Š, โ„’๏ธ€(๐‘‰,๐‘Š) ไธ€ๅฎšๆ˜ฏ Banach ็š„.

Proof ่ง Folland 5.4.

13.1.3 (๐ฟ๐‘)โˆ—=๐ฟ๐‘ž, 1๐‘+1๐‘ž=1

Theorem 13.73 : ๅฏนไบŽไบ’ไธบ conjugate exponent ็š„ ๐‘,๐‘ž, ๐ฟ๐‘ ๆ˜ฏ ๐ฟ๐‘ž ็š„ dual space

For 1<๐‘,๐‘ž<โˆž with 1๐‘+1๐‘ž=1, we have:

(๐ฟ๐‘)โˆ—=๐ฟ๐‘ž

In particular the Hilbert space:

(๐ฟ2)โˆ—=๐ฟ2
Proof

Define map

๐ฟ๐‘žโ†’(๐ฟ๐‘)โˆ—๐‘”โ†ฆโŒ€๐‘”

where

โŒ€๐‘”(๐‘“)โ‰”โˆซ๐‘“๐‘”,๐‘“โˆˆ๐ฟ๐‘

It is well-defined by Hรถlder:

๐‘“โˆˆ๐ฟ๐‘,๐‘”โˆˆ๐ฟ๐‘žโŸน๐‘“๐‘”โˆˆ๐ฟ1

and

||๐‘“๐‘”||1=โˆซ|๐‘“๐‘”|โ‰ค||๐‘“||๐‘||๐‘”||๐‘ž

Easy:

โŒ€๐‘”(๐‘“1+๐‘“2)=โŒ€๐‘”(๐‘“1)+โŒ€๐‘”(๐‘“2)

Also

|โŒ€๐‘”(๐‘“)|=|โˆซ๐‘“๐‘”|โ‰คโˆซ|๐‘“๐‘”|โ‰ค||๐‘“||๐‘โ‹…||๐‘”||๐‘ž

Thus

โŒ€๐‘”โˆˆ(๐ฟ๐‘)โˆ—

โ–ก

13.2 the dual of ๐ฟ๐‘-II [Fol 6.2]

13.3 the dual of ๐ฟ๐‘-III [Fol 6.2, finished]