From real numbers and limits to several variables and measure
Qiulin Fan
2026
Chapter 1Introduction
This collection joins the single-variable MATH 451 notes with the multivariate MATH 395 notes. The reading path moves from the real-number system, sequences, continuity, differentiation, series, and Riemann integration to metric spaces, multivariable differentiation, implicit functions, change of variables, and an IBL transition toward Lebesgue measure.
The original course directories and their source manifests remain authoritative. This entry only composes their lecture notes and personal homework transcriptions for publication. Missing submissions and unfinished source proofs remain visibly missing rather than being reconstructed from checking-only material.
Single-Variable Mathematical Analysis
Chapter 2The real-number system
2.1Set notation and the construction of
源页题记:「此课将使用以下 symbols」。
Definition 2.1 (Power set and indexed families)
The power set of is . 若 是一个 set,且对每个 , 是一个 set,则 是一个 indexed family of sets。
For such a family,
and
The relative complement is . The source places these next to the inclusion chain
It annotates this chain with “given by God” below and “algebraically closed” below . The structural discussion distinguishes three approaches to fundamental issues:
•
naïve approach;
•
axiomatic approach; and
•
constructive approach (set theory, 582).
constructive approach 从 , , , and 构造 。本课程中 。
Definition 2.2 (Inductive subset of )
A set is inductive if:
•
; and
•
for every , implies .
Then is the smallest inductive subset of ; hence under the course convention.
Definition 2.3 (Definitions by induction)
For and , integer powers are fixed by and . The factorial is fixed by and .
The handout also records summation and product notation: and .
The recalled identities are
•
;
•
;
•
for ;
•
; and
•
.
The source labels the last formula “Binomial Thm”.
Let with and . If is a root, where are coprime and , then and .
Indeed, multiplying by yields
and, symmetrically,
.
Thus and ; coprimality and the Fundamental Theorem of Arithmetic give and . For , the only possible rational roots would be in , and none is a root.
Definition 2.4 (Algebraic and transcendental numbers)
A complex number is algebraic if it is a root of a polynomial with coefficients in ; otherwise is transcendental.
The examples in the source are (a root of ), (a root of ), (a root of ), and every (a root of ). The annotation is “ and are transcendental (hard to prove)”. The set of all algebraic numbers is denoted , the algebraic closure of , and is a field.
Definition 2.5 (Algebraically closed field)
A field is algebraically closed if every polynomial of degree with coefficients in has roots in , counting multiplicities.
Thus is algebraically closed, and the Fundamental Theorem of Algebra says that is algebraically closed. 源页旁注以原来的混排对照 algebraically closed’ ‘ 与 geometric deficiency (R.F. order theory)’ ‘: 有一部分 real numbers 和一部分 non-real numbers,但缺少 的 order-theoretic completeness。
2.3Bounds, extrema, and intervals
Definition 2.6 (Upper/lower bounds and extrema)
Let have a linear relation , and let . A point is an upper bound of when for every ; then is bounded above in . Lower bounds and bounded below are defined dually.
If is an upper bound, then , the largest element of . If is an upper bound and every upper bound satisfies , then , the least upper bound (supremum). Dually one has and (infimum).
The notes emphasize that may have no in and may have no in . If a maximum exists, it is unique: if , then and , so .
Definition 2.7 (Bounded set and interval)
is bounded in if it is both bounded above and bounded below. An interval is a set such that whenever and , then .
For a linear order, and ; similarly for the other endpoint choices. The convention on the page is , , and . The latter two are explicitly “not in ”.
Examples from the page:
•
every finite is bounded and has a maximum and minimum;
•
, , and are not bounded above in ;
•
is bounded below in , with and its lower bounds ;
•
and ;
•
does not exist, while ;
•
for , , , and does not exist.
If is an upper bound of in , every in is an upper bound; lower bounds satisfy the dual statement. If has a maximum, then .
Theorem 2.3 (Completeness axiom and its dual)
Every nonempty that is bounded above has . Equivalently, every nonempty bounded below has .
For the dual statement, let be the set of all lower bounds of . Then , and . Equivalently, with , a nonempty set bounded below has .
源页称此为 LUB property’ ‘,并写道 geometrically complete ordered set 需要 LUB property’ ‘;从这个意义上 complete ordered field 就是 。In particular, has outside ; this exhibits the geometric deficiency of both and .
2.4Page-complete lecture record
The following preserves the remaining readable working, labels, and native schematics from the two source pages. It is intentionally kept in the source’s Chinese–English mixed language.
2.4.1L01–Real–Num–System–I, p. 1
The sheet begins Instructor: Scott Schneider and records the four symbols power set, indexed family, indexed union, indexed intersection, and relative complement. Its number-system relationship is reconstructed as a native table; the labels given by god, ordered field, field, and algebraically closed occur at these positions.
It explicitly gives
If is a set and, for every , is a set, then is an indexed family of sets. The three approaches are The three approaches are (1) naive approach; (2) axiomatic approach; and (3) constructive approach (set theory, 582). Under Using constructive approach to build , it lists
In this class, . The inductive definition reads
followed by Then . The recalled secondary-school formulas are, in the source order,
and, for all , (Binomial Thm.). The induction definitions also state, for ,
The last lower-right note says that has the linear relation , marked (1) transitive, irreflexive; (2) trichotomy, and the completeness axiom , , .
2.4.2L02–Real–Num–System–II, p. 1
The top note is is the unique complete ordered field (所有 complete ordered field 都同构 ), and again is the intersection of all inductive subsets of . The deficiency table is retained in source order:
For , the page writes , , , hence , and says these are not roots. Its complete calculation is
By FTA and , this yields and . The algebraic-number examples are: is a root of ; is a root of ; is a root of ; and every is algebraic since it is a root of . and are transcendental (hard to prove). The set of all algebraic numbers is , which is a field, called the algebraic closure of .
An algebraically closed field is stated as: every polynomial of degree with coeffs in has roots in (counting multiplicities). Thus is algebraically closed; by FTA, is algebraically closed. has some irrational and some non-real numbers, but still has geometric deficiency (see order theory). The reminder is: an irreflexive, transitive partial order () that also has trichotomy is a linear order ().
2.4.3L02–Real–Num–System–II, p. 2
For bounds, the sheet requires , , and a linear relation on :
Similarly we have lower bound, bounded below, , infimum . Its proof of uniqueness is . An interval is
It records and , then the convention , , (They are not in ). The listed examples are:
•
Every finite is bounded and has .
•
are not bounded above in .
•
is bounded below in , , and all lower bounds of in are .
•
, , and DNE, while , .
•
For , DNE, , .
If is a UB of in , every in is a UB (LB similarly); if has a maximum then . The LUB property is: if is not empty, then ; an ordered set with it is geometrically complete, and an ordered field with it is a complete ordered field. The handwritten example is
and it closes: and have geometric deficiency, whereas is a complete ordered field (but has algebraic deficiency).
Chapter 3Functions, countability, and metric spaces
3.1Archimedean facts and metric spaces
源页问一个 field 既 algebraically closed 又 geometrically closed’ ‘。答案是 :both algebraically and geometrically closed (topologically)’ ‘;但 不是 ordered field。作业旁注是 impossible to define linear order on ’ ‘。尽管 的 completeness axiom 用 order 表述,后面会从 Cauchy sequences 得到一个不依赖 order 的版本。
Theorem 3.1 (Useful supremum test)
Let and . Then if and only if is an upper bound of and, for every , there exists with .
The Chinese explanation is: “只要下移一点点,就会超进去”. For a set bounded below, if is its set of lower bounds, then ; equivalently, .
Theorem 3.2 (Copies of , , and )
Every ordered field contains copies of , , and : , , and so on give ; additive inverses give ; and gives .
Theorem 3.3 (Archimedean properties)
In an Archimedean ordered field :
•
for every , there is with ;
•
for every in , there is with ;
•
for every , there is with ;
•
equivalently, for in , there is with .
这些 characterizations 给出 的 density:
取 使 ,再取 使 ,于是 。所以任意两个 reals 之间有 infinitely many rational points。源页还写 is also dense in (hw)‘ ’。
Theorem 3.4 ( is Archimedean)
is an Archimedean ordered field.
若 有 upper bound,令 。则 不是 upper bound,故某个 满足 。于是 ,但 ,矛盾。源页给出 (rational functions)和 -adic fields 作为 non-Archimedean examples,并写道:“there is a consistent and rigorous way to do calculus with infinitesimals (non-standard analysis)”.
Definition 3.1 (Absolute value)
For , , , , and . Consequently .
The proof of the triangle inequality squares both sides: . The extended form is .
Definition 3.2 (Metric and Metric space)
A metric on is a map such that, for all :
•
, with equality if and only if ;
•
; and
•
.
The pair is a metric space.
Theorem 3.5 (Euclidean metric)
For every , is a metric space under , where and .
Cauchy–Schwarz, , follows by expanding and taking when . The metric triangle inequality then follows from .
3.2Functions
The lecture begins with
and
.
It records , , for , , and . The warning is in general.
一个 function 是 的 subset,且对每个 ,恰有一个 满足 。Write , , and .
For and , and .
源页 examples 为 on 、 on 、the supremum function from to , the harmonic function , and Dirichlet’s function for and for .
The handout “More Joy of Sets” retains its English terminology: “map” and “mapping” are synonyms for function; domain/source and codomain/target space are and ; an input variable is independent and an output variable dependent. The pointwise notation is .
For image and inverse image:
•
and ;
•
;
•
;
•
;
•
inverse images preserve union, intersection, and difference exactly.
The identity is , . Composition is and is associative.
An inverse of is with and . A function is injective if implies , surjective if each has a preimage, and bijective if it is both.
Theorem 3.6 (Invertibility criterion)
A function is invertible if and only if it is bijective.
If and , composition preserves injectivity, surjectivity, and bijectivity; if is injective, is injective, and if it is surjective, is surjective. The source’s graph remark is that horizontal lines meet an injective real graph at most once and a surjective one at least once.
The restriction of to is the map which agrees with on . Thus on is neither injective nor surjective, its restriction to is injective, and , , is bijective.
list 记住 order 和 repetition: and . An -tuple is . The Cartesian product is , while is both a Cartesian product and a vector space. The graph is , and the rigorous ordered-pair encoding is .
3.3Cardinality and countability
Definition 3.5 (Cardinality and countability)
is finite if for some , and infinite if an injection exists. Write for an injection , and for a bijection.
is countably infinite if ; it is countable if .
The example maps an odd to and an even to ; it is bijective.
Theorem 3.7 (Cantor–Schröder–Bernstein)
If and , then .
Theorem 3.8 (Cantor diagonal arguments)
is countable, is uncountable, and every set satisfies .
Rationals are diagonally enumerated as pairs , . If were surjective, choose decimal whose th digit differs from the th digit of ; it is not in the range. More generally, for , cannot equal for any . The page notes and calls the assertion that no cardinality lies strictly between and the continuum hypothesis.
Theorem 3.9 (Countable products and unions)
If are countable, then is countable. If is countable and every is countable, then is countable.
For the product, injections yield
an injection by unique prime factorization. For the union, take a surjection , surjections , and a surjection with ; then is surjective onto the union.
最后, 包含 uncountably many irrational numbers:若其 irrational part countable,与 的 union 会使 countable。手写结论为 is countable,so there are uncountably many transcendental numbers。
3.4Page-complete lecture record
3.4.1L03–Archimedean-property&Metric-Space, p. 1
The review first says is algebraically closed and is geometrically closed, but and . The written question is “can we find a both-closed field?” Answer: yes, ; “ is both algebraically and geometrically closed (topologically)”. However, “ is not an ordered field” and the homework is “impossible to define linear order on ”. The note asks how can be geometrically complete if the completeness axiom for is based on order; answer: define an order-free axiom with Cauchy sequences (next week).
The dual completeness statement is written and proved twice:
First let be the set of all lower bounds of ; completeness gives , and the goal is . Second define ; then and, since is bounded below, is bounded above, and .
The useful supremum fact is stated as
The source’s number-line schematic is equivalently rendered by
and its Chinese explanation is “只要下移一点点,就会超进去”. It also records the “wrong” Newton/Leibniz definition is infinitesimal exactly when , then asks “这边 infinitesimal 吗?” The answer depends on the definition of ; according to axioms 1–15, “NO!”, and the present proof uses the Archimedean property of .
For every ordered field , the page constructs its copies of : , , ; then ; and finally . The Archimedean properties are listed exactly as
3.4.2L03–Archimedean-property&Metric-Space, p. 2
The page observes that (4) implies (1) by taking , while (1) implies (4): given , choose . It states density in the mixed wording “ 在 中稠密性: s.t. ”. The complete working is
choose with ; by the integer property choose with ; hence .
The native number-line schematic on the sheet has the rational point between the endpoints:
The conclusion is “there are infinitely many rational pts between ”; also “ is also dense in (hw)”. It contrasts and as Archimedean with non-Archimedean ordered fields, giving (all real functions) and -adic fields . The full proof of “ is an Archimedean ordered field” is: suppose such that no has . Then is a UB of , so . Since is not a UB, some satisfies , hence , contradicting . The source then says, “尽管 infinitesimal 在 real line 上不存在, there is a consistent and rigorous way to do calculus with infinitesimals (non-standard analysis).”
The absolute-value list is
For the triangle inequality it writes
then . The extension is
A metric is a function with (i) and , (ii) , and (iii) . If it satisfies the triangular property, is a metric and is a metric space; hence absolute value makes a metric space.
3.4.3L03–Archimedean-property&Metric-Space, p. 3
For , the source writes
and . The proofs of positivity and symmetry are explicitly if (and exactly when equal), and .
For Cauchy–Schwarz, gives
Take when , giving
hence . For the triangle inequality, let , , , so ; then
3.4.4L04(1)–Function&Countability, pp. 1–3
The review uses the two native interval relationships
and the Archimedean test: in any ordered field, ; in an Archimedean ordered field it suffices that . It then lists , , , for , , , and .
The “blobs and arrows” function diagram is rebuilt natively:
Its exact definition is , , and such that ; it calls , , and . It gives and . The explicit examples are the squaring function, reciprocal function , supremum function , harmonic function , , and Dirichlet’s function for , for .
Its two-level function sketch is retained natively:
For cardinality, “finite” means such that has elements, denoted ; “infinite” means an injection . means an injection and a bijection. The homework remark is iff there is an injection , not merely a surjection . The bijection is for odd and for even , with table , , , , , , dots. “Countably infinite” means ; “countable” means , equivalently a surjection .
The lattice diagram for is retained natively. View as , , and enumerate the lattice by increasingly large finite squares, omitting repetitions; this yields a surjection . Cantor’s proof writes any proposed as , chooses with th digit of , and concludes for every , so and are uncountable.
The power-set proof defines, for ,
If , then iff , a contradiction. The page then asks whether there are cardinalities larger than and answers (though ); whether there are cardinalities strictly between and remains unknown, and the assertion that there are none is the continuum hypothesis. The final theorem says finite products of countable sets are countable and, for countable , a family of countable sets has countable union; the final application is that has uncountably many irrationals and is countable, hence there are uncountably many transcendental numbers.
3.4.5L04(2)–Handout–Function, pp. 1–4
“More Joy of Sets” says it continues the basic-set-theory summary from “The Joy of Sets”, with special emphasis on FUNCTIONS. It explains that a function from to assigns each a unique ; is read “ maps to ”. Map/mapping are synonyms for function; is domain/source and codomain/target space. The pointwise arrow is ; a rule’s input variable is independent and output variable dependent. A footnote says might have been better notation for a left-to-right reader, but mathematical convention writes .
The image is ; for subsets, and . The complete displayed list is
The identity example is , . It gives the power-set example , , then composition: if , , , and . Composition is read backwards: “ means first apply , then apply ”.
An inverse satisfies and ; if it exists it is unique and is denoted . Definitions are injective ( implies ), surjective (), and bijective (both); the theorem is “for any function , is invertible iff is bijective”. The sheet adds: equal functions require the same domain and codomain; restricted to is , , written or ; a function with the same rule is surjective. For real graphs, injective means every horizontal line meets at most once; surjective means every horizontal line meets at least once. The squaring function example and its restriction have the same statements as above.
A list is a finite ordered set: and ; order and repetition matter. A list of length is ; equal lists have the same length and entries in the same order. A sequence is an infinite ordered set ordered like . Cartesian products are
with and generally the set of -tuples. It gives and the familiar increasing exponential sketch through ; generally . The rigorous definition is then repeated: a function is its graph, and ; this has and .
3.4.6L04(3)–Handout–Countability, pp. 1–2
The Cantor–Schröder–Bernstein proof is reproduced in full. Given injective and , define by
Put , , and . Define
for , while for .
Using De Morgan and preservation of unions/intersections by forward images of injective functions,
Thus , which makes bijective.
For countable products, injections produce
where is the th prime; FTA makes it injective. For countable unions, take a surjection , for each a surjection , and a surjection , . Then is surjective onto .
Theorem 4.1 (Every real is a rational-sequence limit)
For every , there is a sequence in such that .
Use density to choose with .
Theorem 4.2 (Uniqueness of limit)
If and , then .
Given , choose so that both after . Then .
Theorem 4.3 (Basic limits)
For , ; for , . If , then ; if , then . Also if and only if , and if and only if .
For , Bernoulli gives ; for , write and compare with . The handwritten note says that the case uses an earlier fact. The source also proves for and , citing Rudin 3.20 for the latter.
Theorem 4.4 (Subsequences preserve convergence)
converges to if and only if every subsequence converges to . A tail has the same limit.
源页写道 convergent sequence 与其 tail 可以看成没有任何本质区别’ ‘。
The rendered page is blank except for its page frame; it contains no mathematical text to transcribe.
4.2Limit laws, boundedness, and
Theorem 4.5 (Limit laws)
If and , then
•
(and likewise for subtraction);
•
for every ;
•
; and
•
if no is zero and , then .
对于 product,展开 and use bounds. For reciprocals, first use convergence to obtain eventually, then . The source annotates these two preliminary bounds as “bound ①” and “bound ②”.
Further laws recorded on the page are: convergent implies converges; for , ; and for , provided . It defines real exponentiation for by when .
Theorem 4.6 (Vector sequences)
A sequence in , with components , converges to if and only if for every .
正向使用 ; for the reverse direction, make each coordinate error smaller than . Vector sum, dot product, and scalar multiplication obey the same limit laws as real sequences.
Definition 4.3 (Bounded function and bounded sequence)
A function is bounded when its range is bounded. In particular a sequence is bounded if all of its terms lie between two real bounds.
Theorem 4.7 (Convergent sequences are bounded)
Every convergent sequence of real numbers is bounded.
若 ,取一个 tail 使 ,再分别 bound finitely many earlier terms。直接应用 limit laws 给出 rational function : it is when , it is either or when , with the sign determined by .
Theorem 4.8 (Limits involving and )
If and , then ; for the product tends to . The signs reverse when . If tends to either infinite endpoint and converges, then has the same infinite limit.
The exercise records, for a positive real sequence: if and only if ; the negative version gives if and only if .
Definition 4.4 (Monotone sequence)
is increasing if for every , decreasing if , and monotone if it is either.
Theorem 4.9 (Monotone convergence theorem)
Every bounded monotone real sequence converges. If is bounded and increasing, then ; if decreasing, its limit is the corresponding infimum.
源页说明 increasing seq. 必定 bounded below;decreasing seq. 必定 bounded above’ ‘。证明令 and takes a term with .
Definition 4.5 ( and )
For a bounded sequence set and . Then is decreasing and increasing, and
, .
The notes display
.
直观地, 是 the largest number that can get arbitrarily close to, for infinitely often’ ‘。 是 当且仅当对每个 ,有 infinitely many 使 ,且只有 finitely many 使 。定义也经由 和 延伸至 unbounded sequences。
Theorem 4.10 (Convergence via upper and lower limits)
If , then . Conversely, if , then .
Examples include , , , and , . If eventually, then and . The page proves the squeeze theorem and the ratio-test corollary: for positive , if , then .
4.3Cauchy sequences, subsequences, and completeness
Definition 4.6 (Cauchy sequence)
A real sequence is Cauchy if for every there is such that whenever .
Theorem 4.11 (Cauchy criterion in )
A sequence in converges if and only if it is Cauchy.
Every Cauchy sequence is bounded: use the Cauchy condition with for a tail and bound the finitely many initial terms. The converse first proves from pairwise closeness.
Definition 4.7 (Complete metric space)
A metric space is complete if every Cauchy sequence in converges to a point of .
源页写 和 complete,而 不 complete。一个 example 定义 、,and for ; it has and is Cauchy.
Definition 4.8 (Contractive sequence)
is contractive if some satisfies for every .
Every contractive real sequence is Cauchy, hence convergent. The source uses the geometric bound . It also solves , : a bounded increasing sequence converges to the positive root . The decreasing sequence defines .
Definition 4.9 (Subsequence and subsequential limit)
If and is strictly increasing, then , , is a subsequence. A subsequential limit is the limit of a subsequence.
对 ,even subsequence converges to 而 full sequence diverges。 的每个 subsequence 都 converges to ,且每个 tail 是一个 subsequence。
Theorem 4.12 (Monotone subsequence theorem)
Every real sequence has a monotone subsequence.
A term is dominant if it is at least every later term. If infinitely many dominant terms occur, they form a decreasing subsequence; otherwise, after the final dominant term one recursively chooses later, strictly larger terms to obtain an increasing subsequence.
Theorem 4.13 (Bolzano–Weierstrass)
Every bounded real sequence has a convergent subsequence.
Apply the monotone subsequence theorem and monotone convergence. For a bounded sequence of values, the set of subsequential limits is nonempty; if , it is ; and , . The source adds that these claims extend to unbounded sequences using and .
In , the open neighborhood of of radius is . A set is open if every has an with . A set is closed if is open.
The examples say and are both open and closed in , while is closed but not open in . Common metrics are on , Euclidean distance and taxi-cab distance on , and on .
Definition 4.11 (Interior, limit point, isolated point, closure)
is an interior point if some neighborhood of lies in ; is the set of all such points.
is a limit point of if every neighborhood of contains a point of . An element of which is not a limit point is isolated. The closure is where is the set of limit points.
The Chinese note says interior membership is necessary but not sufficient for being an interior point; isolated points are not necessarily interior points. A set is open exactly when . A discrete set is ; it has no limit points, only isolated points.
Theorem 4.14 (Sequential and closure characterizations)
is closed if and only if every convergent sequence in has its limit in . Equivalently, contains all its limit points. Also is closed and is the smallest closed subset of containing .
In , every open neighborhood is exactly an open interval, every nonempty open contains around each of its points, closed intervals are closed, finite sets are closed, and every open set is a countable union of open intervals. The generalized Bolzano–Weierstrass theorem recorded here is: every bounded sequence in a complete metric space has a convergent subsequence. In particular and are complete, but is not.
4.5Page-complete lecture record
4.5.1L05–Seq&Limit, pp. 1–3
Besides the definitions above, the source writes the divergent negation , and as (dually for ). It gives the decimal sequence for , the Fibonacci recurrence , , and the proof of uniqueness: for , . For every , choose with .
For , proves ; for , proves . If , ; if , write and use . The case is annotated as an earlier fact. For , obeys ; for , , so (Rudin 3.20). L05 p. 3 is visually blank.
4.5.2L06–Limit–II, pp. 1–4
The worked epsilon proof is once . The product-law proof expands ; the reciprocal proof uses eventually and . The source additionally gives , for nonnegative terms, and .
For vector sequences, coordinatewise convergence is equivalent to Euclidean convergence: one way, and coordinate errors the other. The rational function rule is for equal degrees, for numerator degree smaller, and signed infinity for larger degree. A convergent sequence’s explicit bounds are , .
The infinity multiplication table has the usual signs , , , ; if one sequence tends to either infinity and the other converges, their sum tends to that infinity. For positive , exactly when (negative dual). The monotone proof is: bounded increasing has and for ; the decreasing dual tends to infimum.
Put , ; is decreasing, increasing, and limsup/liminf are their limits. The native tail schematic is
It gives the “infinitely often” limsup criterion and examples , , , , , . It proves convergence iff limsup equals liminf, including the extension. The comparisons, squeeze theorem, and ratio corollary are all shown with their tail bounds: positive and give ; homework records the divergence case.
4.5.3L07–Cauchy-seq, pp. 1–3
The source warns convergent implies , but not conversely. The Cauchy boundedness proof takes epsilon about , then bounds the initial finite set. For the reverse Cauchy criterion, pairwise closeness gives
so upper and lower limits are equal. Complete metric space means every Cauchy sequence converges; the source explicitly gives the complex metric .
The averaging example is , , , with and
A contractive sequence has , , and is Cauchy (Rudin 3.8). , is bounded increasing and limits to . For the same averaging recursion with , the source derives and limit . It proves weakly decreasing and , then defines .
4.5.4L08(1)–subseqs, pp. 1–2
A subsequence is for strictly increasing . Examples: has and constant subsequence ; has subsequential limits . The forward proof for subsequences uses . A dominant term has for all later ; infinitely many dominant terms form a decreasing subsequence, otherwise the recursive choice of later larger terms gives a strictly increasing one. Thus BW holds. It names as an example.
For bounded , the set of subsequential limits is nonempty, , , . The proof chooses with both and , and rules out by a tail supremum. It explicitly extends this to unbounded sequences: has , limsup , liminf .
4.5.5L08(2)–topology-in-metric-space, pp. 1–3
The source’s visible native neighborhood pictures are the circle in and interval in . It defines , and says membership is necessary but not sufficient for being interior; isolated points need not be interior. It defines , , isolated , and discrete .
The sequential closed-set proof is complete: if is closed, an open neighborhood of any eventually contains any sequence tending to , so it cannot lie in . If not closed, choose for a point whose every neighborhood meets ; then . It also proves a limit point has infinitely many nearby points by using the minimum positive distance to a hypothetical finite list. In , every open set is a countable union of open intervals. General convergence, boundedness by , and generalized BW are stated; the page concludes complete and not complete.
Chapter 5Limits and continuity
5.1Limit points and limits of functions
Definition 5.1 (Limit point, closure, isolated and discrete sets)
Let and . Then is a limit point of if for every there exists with . Equivalently, every open neighborhood of meets .
Write for the set of limit points and for the closure. A point of is isolated; a set is discrete when .
Theorem 5.1 (Sequential criterion for a function limit)
if and only if every sequence in with satisfies .
The source uses its contrapositive to show that if some approaches but does not approach , then the limit is not ; if one approaching sequence has divergent values, or two have different image limits, the function limit does not exist.
Definition 5.3 (Infinite and one-sided function limits)
means that for every there is such that whenever and . Definitions at and are analogous.
If is a limit point of , then means the same estimate with . The left-hand limit is defined dually.
源页说有五种 function limits:。其 examples 包括 does not exist and does not exist.
Theorem 5.2 (Function-limit laws)
If and exist, then for :
•
;
•
;
•
; and
•
when .
Function limits are unique. If in a deleted neighborhood of and both limits exist, then . The squeeze theorem says that there and imply . The examples are
,
, and
does not exist.
The last page annotation explains that , and every rational function in particular, is continuous at every point of its domain.
5.2Alternative formulations and continuity
sequence test 再次强调: 不表示 every sequence of domain points 都 tends to ;test limit 要取 中 approaching 的 sequences。源页也给出如下 open-neighborhood formulation。
Definition 5.4 (Function limit in terms of open sets)
Let , , and , with . Then if every open neighborhood of contains for some open neighborhood of .
The source convention is that if is bounded above/below, then ; open neighborhoods of are and of are .
Theorem 5.3 (One-sided and ordinary limits)
if and only if both and , provided is a limit point from both sides.
Theorem 5.4 (Equivalent zero formulations)
For a finite limit, the following are equivalent: , , , and after replacing by .
Definition 5.5 (Continuity)
Let , , and . Then is continuous at if, for every , there exists such that whenever and .
手写 distinction 很重要:limit at 需要 ,却不需要 ;continuity at 需要 ,却不需要 是 limit point。Accordingly, every function is continuous at an isolated point of its domain.
Theorem 5.5 (Continuity criteria)
For , the following are equivalent:
•
is continuous at ;
•
either is isolated in , or ;
•
for every sequence in with , one has ;
•
for every open neighborhood of , there is an open neighborhood of with .
The source lists rational functions (especially polynomials), power functions on , exponential functions, logarithms, trig/inverse trig functions, and as continuous on their natural domains.
Definition 5.6 (Continuous on a set and topological continuity)
is continuous on when it is continuous at every ; it is a continuous function when this holds on all of . More generally, between metric/topological spaces is continuous if is open in for every open .
源页给出 在 ctn 的 direct epsilon–delta proof,取 ;在一般 取 。又用 证明 everywhere ctn,旁注为:“here depend on but not ”.
It also notes that
for , while
is continuous everywhere except at , whereas
for , while
is continuous everywhere by squeeze. Dirichlet’s function is discontinuous everywhere. Thomae’s function
for a rational in lowest terms, and for
is continuous at every irrational and is a source of the questions “是否存在 使 ctn at iff ?” and “is diffable anywhere?”.
Definition 5.7 (Discontinuities)
is discontinuous at if it is not continuous there. If both one-sided limits exist but differ, has a jump discontinuity; if exists but differs from , it has a removable discontinuity; if a one-sided limit fails to exist by oscillation, it has an essential discontinuity; and if a one-sided limit is infinite, it has an infinite discontinuity.
The examples are for a jump, the function off and at for a removable discontinuity, for essential/oscillating discontinuity, and (with a chosen value at ) for infinite discontinuity.
5.3Closure properties and uniform continuity
Theorem 5.6 (Closure properties of continuous functions)
If are continuous at , then , , , where defined, and for are continuous at .
The domains recorded on the page are for , , and , and for .
Theorem 5.7 (Composition)
If is continuous at and is continuous at , then is continuous at and .
源页明确说此 theorem 也有 variants,把 limit at 全部替换为 limit at 、、 或 。Examples 是 and .
Further source examples retain their proof choices:
•
at : , choose ;
•
: choose ;
•
at any : choose ;
•
has a “longest ” at of ;
•
is uniformly continuous on with ;
•
with value at is continuous everywhere;
•
is discontinuous everywhere; and
•
for , for is continuous at but discontinuous everywhere else.
Definition 5.8 (Uniform continuity)
Let and . Then is uniformly continuous on if, for every , there is such that for all , implies .
The source’s quantifier comparison is retained: ordinary continuity has “for every point ” before the choice of ; uniform continuity chooses one for all points. 中文解释为:对任意 ,总有一个距离 使得在 上距离足够近的点,其 image 的距离也足够近;“uniformly ctn 的要求比 ctn 更严格”.
Theorem 5.8 (Basic uniform-continuity facts)
Uniform continuity on implies continuity on . A restriction of a uniformly continuous function is uniformly continuous.
The examples are (choose ), not uniformly continuous on (take and a large ), and uniformly continuous on . The source observes that is uniformly continuous on but not on nor on for .
Theorem 5.9 (Heine–Cantor)
If is closed and bounded (compact) and is continuous, then is uniformly continuous on .
The Chinese discussion explains why both hypotheses matter: on is continuous and its domain closed but unbounded, so not uniformly continuous; on is continuous on a bounded but nonclosed set and is not uniformly continuous. Positive examples are on , on for , and with value at zero on .
If is uniformly continuous and is Cauchy in , then is Cauchy.
The page’s counterexample is on : is Cauchy but is not, so is not uniformly continuous on any set containing .
Theorem 5.11 (Extension criterion)
Let be bounded and . Then is uniformly continuous if and only if there is a continuous whose restriction to equals .
For the forward direction, if and tends to , define ; uniform continuity makes Cauchy and the definition independent of the approximating sequence.
5.4Extreme and intermediate values
Theorem 5.12 (Extreme Value Theorem)
If nonempty is closed and bounded and is continuous, then is bounded and there are such that for every .
The proof sets . Choose in with , take a convergent subsequence, use closedness to retain its limit , and use continuity to obtain . The notes summarize: “closed + bounded + ctn ,那么 extreme value 一定存在”.
Theorem 5.13 (Intermediate Value Theorem)
If is continuous and lies between and , then some satisfies .
Assume and set . Then is nonempty and bounded above; for , continuity and sequences approaching from both sides give . The source’s Chinese explanation is that a continuous curve on an interval must “覆盖了 中的所有值”.
The application is the fixed-point theorem: if is continuous, then some has . When the endpoint signs do not immediately give this, take and apply IVT.
Theorem 5.14 (Continuous image of an interval)
If is an interval and is continuous, then is an interval.
For , choose preimages and apply IVT on the subinterval between them. If is a closed bounded interval, EVT gives , so the image is again a closed bounded interval.
5.5Page-complete proof and diagram ledger
5.5.1L09–Limit-of-Functions-I, pp. 1–3
The visible lecture framing is “Ch4 limit of functions”, , , with three equivalent styles: epsilon/delta, sequences, and open sets. A limit point is exactly , equivalently every open neighborhood meets . The sheet writes that a sequence’s limit points are subsequential limits but the reverse can fail (constant-sequence example); it gives no limit point, all reals as limit points of , and . It defines , isolated , and discrete .
The three graph examples , , and the latter assigned zero at have the same limit at . The sequential proof forward combines with the epsilon condition; backwards selects with and . It explicitly records the diagnostics: one approaching sequence with images not tending to disproves ; divergent images prove DNE; two image limits that differ prove DNE.
The one-sided definition restricts , requiring a limit point from that side. The displayed examples are and DNE at zero. The sheet says there are five kinds of limits: . The limit laws include scalar, sum, product, quotient, order and squeeze. Its calculations are near , , and while diverges.
5.5.2L10(1)–Limit-of-Functions-II, pp. 1–2
The sequence review graph distinguishes a sequence approaching with image limits and (so no limit) from a curve with a separately assigned isolated value at (nearby limit ). The open-neighborhood definition is
The convention gives for bounded-above/below sets and neighborhoods , . The ordinary limit is equivalent to both matching one-sided limits; the proof takes . The finite zero forms are , , and .
5.5.3L10(2)–Continuity-I, pp. 1–2
The source contrasts a limit at (requires , not ) with continuity at (requires , not a limit point). Thus every isolated domain point is continuous. Its four criteria are: continuity; isolated or limit ; sequence criterion; and the open neighborhood inverse-image inclusion.
Visible epsilon proofs are with ; with ; and with . It asks for the longest delta at , recording . The diagrams classify jump , removable off zero and at zero, essential , and infinite with a zero value. It proves continuous at zero by , says Dirichlet is discontinuous everywhere, and states the continuity properties of the rational/irrational indicator and Thomae’s function exactly as in the source.
5.5.4L11(1)–Continuity-II, pp. 1–2
The closure-property domain ledger is: , . The proof uses sequence continuity. For composition, , , gives ; the source’s variants replace throughout by . Examples are and . The topology proof uses .
5.5.5L11(2)–Uniform-Continuity, pp. 1–3
The quantifier contrast is for ordinary continuity versus for uniform continuity; the page states the latter delta does not depend on the position of . Uniform continuity implies continuity and restrictions remain uniform. Examples: uses ; on fails by taking epsilon , , and comparing ; on uses ; is uniform on but not on or for .
Heine–Cantor’s contradiction creates , , takes convergent subsequences, uses equal limits from the distance condition, closedness to retain the limit in , and continuity for the contradiction. It lists the source counterexamples on and on , plus positive examples , away from zero, and on .
The Cauchy theorem follows by applying uniform delta to the Cauchy tail. For , is Cauchy but is not, so no uniform continuity on a set containing . The extension theorem for bounded defines, for , for any tending to ; uniform continuity makes this well-defined. The reverse direction uses compact and Heine–Cantor.
5.5.6L12–EVT&IVT, pp. 1–2
EVT proves a maximum by , a sequence , BW , closedness , and continuity ; the minimum is dual. IVT takes , , tending to , and , then continuity yields . The fixed point proof uses . For continuous , is an interval by applying IVT between preimages; for , EVT plus IVT gives .
Chapter 6Differentiation
6.1Derivatives and rules (L13)
Definition 6.1 (Derivative)
Let , , and (此处 是 accumulation point,所以 lies in the domain of ). Define the derivative of at by
If , then , hence equivalently
如果 exists,则称 is differentiable at . 把 作为 variable 时,我们把 derivative 看作 function:
如果 且 都有 differentiable at ,则称 is differentiable on .
Given , the function (-times continuously differentiable) on an open set if exists and is continuous on .
6.2Extrema, MVT, and Darboux (L14)
Definition 6.3 (Local extrema)
Let , , and . If there is such that for all , then is a local maximum point of , and is a local maximum value of . Dually define local minimum point/value; together these are local extreme point and local extrema.
L14 p.1 的曲线标出了一个 local min、两个 local max(其中右侧极大值高于左侧),以及随后的 local min;其可辨识信息是极值只比较 的某个 neighborhood,而非整个 domain。用点位/不等式表表示为
.
Lemma 6.1 (Key lemma)
Let , , , and suppose is differentiable at .
(i) If , then there is such that, for all , implies .
(ii) Dually, if , then there is such that implies .
Proof
For (i), let . Fix such that
whenever . Thus
If , division by gives , while division by gives . (ii) is dual. 这两条 lemma 的结论也说明:如果 ,则 在 的某个 open neighborhood 中严格 monotone。
Corollary 6.1 (Fermat’s theorem)
Suppose is defined on an open neighborhood of . 如果 是 的一个 local extreme point 且 存在,则 .
Proof
Directly follows from the key lemma: if or , then cannot be a local extreme point.
Corollary 6.2 (Rolle’s theorem)
If is continuous on , differentiable on , and , then there is some such that .
Proof
By EVT, choose such that for all . If neither is an endpoint, Fermat gives the result. More explicitly as on L14 p.1: if , then is an interior local minimum and ; if , then is an interior local maximum and . If neither strict inequality holds, then for every , so is constant and for every .
Corollary 6.3 (Mean Value Theorem)
If is continuous on and differentiable on , then there is such that
Proof
Let . Then is continuous on , differentiable on , and . Rolle’s theorem gives , which rearranges to the displayed equality.
The L14 p.1 secant/tangent diagram records the same parallel-slope relation:
Corollary 6.4 (Zero derivative and monotonicity)
若 在 上 diffble 且每个 都有 ,则 在 上 constant。于是若 on ,则该处 。interval 上的 function 在 推出 时称 increasing; 时称 strictly increasing;decreasing 对偶定义。Weakly increasing or decreasing on ’ ‘ 与 monotone on 同义。
Proof
If were not constant, there would be with ; MVT would give , a contradiction. Apply this to for the second assertion.
Corollary 6.5 (Increasing/decreasing test)
If is differentiable on , then for every implies is increasing on . If for all , then is strictly increasing. Both statements have decreasing duals.
Note: (i) is a weak statement, but (ii) has a strict conclusion. For , implies , though .
Proof
For , MVT gives a with .
Corollary 6.6 (First derivative test)
Let and suppose is continuous on for some , and differentiable on and . If on and on , then is a local maximum of ; dually, the reversed signs give a local minimum.
Proof
For , MVT gives a with , hence . The same argument for gives .
Theorem 6.6 (Darboux’s theorem)
If is differentiable on and , then there is such that . Thus a differentiable function has every slope between and : derivatives satisfy IVT even though they need not be continuous (no jump/infinite discontinuity).
Proof
WLOG let . Then , and is continuous on . EVT gives a minimum point of . The endpoint derivative signs force , so Fermat gives , hence .
6.3Functions on intervals, inverse functions, and L’Hôpital (L14(2))
Standing assumption: let be a nondegenerate interval, and a function.
Theorem 6.7 (Strictly increasing functions)
If is strictly increasing, then:
•
is injective;
•
is also strictly increasing;
•
if is not the right endpoint of , then exists;
•
if is not the left endpoint of , then exists;
•
has at most countably many discontinuities, and they are all jumps;
•
if is an interval, then is continuous.
Proof
For the right limit let , which is nonempty and bounded below by ; write . Given , fix with and . Then for , so . The left-limit proof is similar, and these imply that discontinuities are jumps. For the final claim, prove the contrapositive: at an interior jump with , both and are nonempty but is not contained in because . The endpoint cases are similar. The remaining proofs are left as exercises. Remark: the dual also holds if is strictly decreasing.
Theorem 6.8 (Continuous functions on intervals)
If is continuous, then:
•
is an interval;
•
if is closed and bounded, so is ;
•
is strictly monotone iff is injective;
•
if is injective, then is also continuous.
Proof
The first two were proved previously. For the backward direction of (iii), if is not strictly monotone, WLOG find in with either or . IVT then implies is not one-to-one. L14(2) p.2 visualizes these two alternatives by the following ordered-value charts, each forcing a repeated intermediate value:
Finally, injectivity makes strictly monotone, hence strictly monotone; since is an interval, the previous theorem makes continuous.
Corollary 6.7 (Injective functions and inverses)
If is injective, then is strictly increasing iff is strictly increasing; is strictly decreasing iff is strictly decreasing; and is continuous iff is continuous.
Question: Could we add “ is differentiable iff is differentiable”? Answer: not quite. is injective and differentiable on , but is not differentiable at .
Theorem 6.9 (Inverse Function Theorem)
Suppose is continuous and injective on an open interval , let , and suppose is differentiable at with . Then is differentiable at and
The p.3 inverse-function sketch has the paired coordinates
.
Proof
Write . Since and for ,
Fix such that the difference between the displayed quotient and is less than whenever . Continuity of at supplies with whenever . Substitution is the displayed p.4 calculation: for ,
and, since and , this gives
whence the result. Consequently, if is differentiable and on an open interval , then is injective on , is differentiable on , and . The final visible p.4 margin annotation is: “Prove? Fix? Skip? 6.1.9”.
Example 6.3 Inverse derivative
Define the invertible differentiable function
on . Find . Since and
6.4L’Hôpital’s Rule
Lemma 6.2 (Cauchy’s Mean Value Theorem)
Let , and suppose are continuous on and differentiable on . Then there is such that
Proof
Apply MVT to on .
Theorem 6.10 (L’Hôpital’s Rule)
Let , and let be differentiable functions with for all . Suppose . If exists and equals , then exists and equals .
Proof
Extend to by . Rolle’s theorem on shows that not just but itself is never on . Let in tend to . Cauchy’s MVT supplies with
Then and for all ; hence the quotient tends to . Since was arbitrary, the desired right-hand limit is .
Remark: the rule also holds for two-sided limits and limits at . It also holds for indeterminate limits of the form , and can be adapted to , , , , and (see 6.3). Skip the rest?
Example 6.4 L’Hôpital examples
Corollary 6.8 (No removable discontinuity for a derivative)
Let , let be an open interval containing , and let be continuous and differentiable on . If exists, then is differentiable at and .
Proof
Let and . Then and exists. The definition of derivative and L’Hôpital’s rule give
Example 6.5 Final counterexample
Let for , and . From continuity at and differentiability everywhere except at , we already know that cannot exist.
若 ,则 series converges;否则 diverges。Informally,。Note: 代表一个 limit 而非 algebraic operation.
Example 7.1 Harmonic and geometric series
The harmonic series diverges to :
给定 和 , 是 geometric series。If ,then
and therefore
The source writes the finite calculation explicitly (for ):
hence .
Definition 7.2 (-series)
给定 ,形如
的 series 称为 -series。
Theorem 7.1 (-series criterion)
A -series converges iff .
Proof
If , then , so and comparison with the harmonic series gives divergence. If ,
The notes record , , and “: no nice formula”.
Example 7.2 Telescoping and alternating harmonic series
For the alternating harmonic series,
If , then increases and decreases, so
Theorem 7.2 (Linearity of series)
设 和 converge,且 。Then
Note: .
Theorem 7.3 (Cauchy criterion for convergence)
令 是 partial sums 为 的 series。则 converges iff is Cauchy,即对每个 存在 such that
Equivalently, .
Proof
课后。
Theorem 7.4 (The th-term test)
If converges, then . Contrapositively useful: not tending to implies diverges.(这是 convergence 的 necessary 而非 sufficient condition。)
Proof
Theorem 7.5 (Comparison test)
Let be a sequence of nonnegative numbers and let be any sequence.
•
If converges and for all , then converges.
•
If and for all , then .
The finite-tail form is also recorded: if converges and for all , then converges; of course the limit is different.
Proof
Let and be the partial sums of and . In the first case,
Thus the Cauchy criterion makes converge. The second assertion is similar.
Example 7.3 Comparison and absolute convergence
converges by comparison with , since for all sufficiently large ,
A series converges absolutely if converges. Absolute convergence is a stronger condition: if converges absolutely, then converges, because
Definition 7.3 (Conditional convergence)
一个 convergent 但不 absolutely convergent 的 series 称为 conditionally convergent。alternating harmonic series conditional convergence。
Disturbing fact: reordering
A conditionally convergent series can be made to converge to any number by “reordering” its terms. A reordered conditionally convergent series still has a limit, but it can be made to converge to any value. For example,
can be reordered to converge to by placing enough positive terms first and using negative terms as compensation after the partial sum exceeds .
In contrast, absolutely convergent series are closed under reordering: if converges absolutely, then for every bijection ,
For nonnegative , this follows because and are increasing sequences with the same supremum. In the general case take
so , and apply the nonnegative claim to .
Theorem 7.6 (Root test)
Let be a sequence in and let .
•
If , then converges absolutely.
•
If for infinitely many (which happens when ), then diverges.
Note: iff, for every , there are only finitely many with , while there are infinitely many with .
Proof
Assume , fix , and choose such that for . Then and comparison with proves absolute convergence. If infinitely often, then does not tend to , so the th-term test gives divergence.
Theorem 7.7 (Ratio test)
Let be a sequence of nonzero numbers.
•
If , then converges absolutely.
•
If , then diverges.
This follows from the root test and the lecture’s fact
Root and ratio tests
root test implies ratio test;root test 通常比 ratio test 更强。Both are inconclusive when or (for example , ). If either limit exists and is , there is absolute convergence for and divergence for . Whenever the root test is inconclusive, the ratio test is also inconclusive(反而不用再试).
Theorem 7.8 (Alternating Series Test)
If is a decreasing sequence of positive numbers converging to , then
converges.
Proof
Let . Then
is increasing and bounded above by , hence converges to . Choose so that and for . Then
Thus as well, hence .
Theorem 7.9 (Integral test)
Let be a positive and decreasing function on . Then
converges, where
Note: 此时我们还没有严格定义 improper integral;integral test 的证明以后再证,但其意义很直观,并由矩形比较
L15 p.4 的紫色 rectangle sketch 就是这组不等式:一个宽度为 的 interval 上,decreasing curve 下的 area 介于两端点高的 rectangles 之间。其 native table reconstruction is
.
Numerical Series Summary
(1) Cauchy Criterion: converges iff is Cauchy. (2) th term test: not tending to implies divergence. (3) Comparison Test. (4) Root Test. (5) Ratio Test. (6) Alternating Series Test: positive decreasing converging to makes its alternating series converge. (7) Integral Test for positive decreasing .
Abs convergence convergence. Abs convergence is closed under reordering; conditionally convergent series are not.
Chapter 8Riemann integration
8.1Antiderivatives and Riemann sums (L16)
Definition 8.1 (Antiderivatives)
A function 被称为 an antiderivative of on interval ,if for all 。若 是 在 上的 antdv,那么对任意 , 都是在 上的 antdv;且 在 上的任何 antdv 都是 的形式。
Example 8.1 The antiderivative problem
For ,
Thus is an antiderivative of on . For example, has antiderivative ; has ; for , the question is left as an illustration that antiderivatives need not have a familiar formula.
The antiderivative problem:given a ctn function on interval ,find such that on 。Informal 的分析是:当 很小时,differentiability suggests ,即 graph 下的一条 narrow region 的 area approximately 为 。
Definition 8.2 (Def② 基础架构:partitions, mesh, and tags)
是一个 function(不需要 ctn)。
1.
A partition of is a finite ordered set where .
2.
is the th subinterval of .
3.
The norm (mesh) is
4.
A tagged partition is a partition together with a choice for every ; is the tag.
The numbered line on L16 p.2 is the partition picture ; a representative finite rendering is
The colored rectangles in L16 p.2 assign one tag to each interval:
.
Definition 8.4 (Riemann integrability)
称 在 上 Riemann Integrable,若存在 ,使对任意 ,存在 满足
对任何 的 tagged partition 都成立。记
并称为 在 上的 Riemann integral。
Riemann Integrable: 对于任意小的 ,都存在 使得对于任何 mesh 小于 的 partition,都有其 Riemann Sum 和 的距离小于 。我们发现这是一个 Cauchy 式的 Definition;直觉上(稍后将证明)mesh 越小,即 partition 越精细,Riemann Sum 就会越接近 area so far,因而这个定义很符合直觉。Informally, .
Theorem 8.1 (bounded 是 Riemann integrable 的必要条件)
If is Riemann integrable on ,then is bounded on 。
Proof
Prove the contrapositive. Suppose is unbounded on . Let , choose any , and take any tagged partition with . Fix such that is unbounded on , then choose with
Replace only the th tag of by , producing . Then . Thus no common limiting can satisfy the definition.
还有 bounded 而 non-Riemann-integrable 的 functions:Dirichlet function
is bounded on but not Riemann integrable. It is Lebesgue integrable, with , because has measure zero on , while has measure one.(之后再学 Lebesgue measure 和 Lebesgue integral。)
Definition 8.5 (Special Riemann sums)
一个 regular partition 的所有 都相同:。对一个 partition,取 得 right Riemann sum;取 得 left Riemann sum;取 得 midpoint Riemann sum。
Combining the regular partition with the right Riemann sum gives
L16 p.3 displays the three choices with their tag positions:
.
Example 8.2 A right sum for
Compute the right Riemann sum of on using a regular partition with subintervals. Here
for , so
Therefore . But this is only one kind of tags on one family of partitions; Riemann integrability must cover all tagged partitions. We return to this using Darboux sums.
Definition 8.6 (Darboux sums and integral)
Suppose is bounded and is a partition. The upper and lower sums are
The upper and lower Darboux integrals are
Always . We say is Darboux integrable on iff . Upper Darboux integral 是所有 partitions 的 upper sum 的下确界;lower Darboux integral 是所有 partitions 的 lower sum 的上确界。
Darboux sum 本身不是 Riemann sum,除非 continuous(此时 extrema 可取);but for every tagged partition, .
L16 p.4 contrasts the upper and lower rectangle pictures on one partition:
.
Theorem 8.2 (Refinement lemma)
Let be bounded with for all . Let be partitions of , and put
Then
and dually
Proof
Fix , and let be the partition points of in . Then
whereas . Each difference is at most , hence
Sum over . The upper-sum statement is dual. Thus refinement makes lower sums bigger and upper sums smaller, and the difference depends on how many new points and how small the mesh is.
8.2Equivalence and basic properties (L17)
Theorem 8.3 (Equivalent Riemann/Darboux criteria)
For a bounded function , the following are equivalent:
1.
is Riemann integrable on .
2.
For every , there is such that every two tagged partitions with satisfy .
3.
For every , there is such that every partition with satisfies .
4.
is Darboux integrable on .
5.
For every , there is a partition of such that .
Proof
(1) = (2). If all sufficiently fine Riemann sums are within of , their pairwise difference is below .
(2) = (3). For a fixed fine partition choose, in every , points approaching and sufficiently closely:
The associated tagged sums differ by less than , while their distances to and are each below ; thus .
(3) = (4). Since for every , . Hence .
(4) = (5). Choose partitions with and . For the common refinement ,
whence its upper-minus-lower sum is below .
(5) = (3). Fix with . Let and choose , where is the number of subintervals of . For any with , let . The refinement lemma bounds both changes by . Together with this gives .
Example 8.3 and the Dirichlet function
For on and the regular partition with intervals,
Both tend to , so is Darboux and hence Riemann integrable, with .
For on and on , every subinterval contains rationals and irrationals, so every partition has and . Therefore it is neither Darboux nor Riemann integrable, although it is Lebesgue integrable and .
Theorem 8.4 (Linearity of integration)
If are Riemann integrable and , then and are Riemann integrable and
Proof
This follows from linearity of Riemann sums: and .
Theorem 8.5 (Monotonicity of integration)
If are Riemann integrable and for all , then
Proof
For every partition , , hence .
Theorem 8.6 (Monotone functions are integrable)
If is monotone on , then is Riemann integrable on .
Proof
WLOG suppose is increasing. Given , take any partition with . Then
8.3Measure-zero criterion, FTC, and rules (L18)
Definition 8.7 (Zero-measure set)
has measure zero if, for every , there is a sequence of open intervals such that
注:zero measure 的意义是这个集合的 length 是 。它可以是无限甚至 uncountable 的,但能由一串很窄的开区间覆盖;例如 Cantor set,, 但它是 zero measure。
Theorem 8.7 (Lebesgue’s characterization of integrability)
A bounded function is Riemann integrable iff the set of discontinuities of has measure zero.( 的非连续点是零测的。)
Consequences of the criterion
任何 countable 的 都 has measure zero,因此任何只有 countably many discontinuities 的函数都是 Riemann integrable,例如 Thomae’s function. Last time: monotone functions are Riemann integrable.
Let . Suppose there are such that whenever and . Then is bounded, and
Proof
Given , choose least with , so , and set . Each increment is at most , hence
Since are arbitrary, the claim follows.
Theorem 8.8 (Composition theorem)
Let be integrable on , and suppose is continuous. Then is integrable on .
Proof
Since is integrable it is bounded, so choose a closed bounded interval . Then is uniformly continuous on . Given , choose so that
Choose with . Apply the lemma on every to estimate its oscillation. Then
Corollary 8.1 (Continuous functions and products)
Continuous functions are integrable: take in the composition theorem. If and are integrable, then is integrable, because
and is continuous.
Theorem 8.9 (Additional properties of the integral)
If is integrable on , then is integrable and
If , then is integrable on iff it is integrable on both and , and
More generally, the L18 p.2 restriction construction says: if and
then , where
is the characteristic function of , and
Altering at finitely many points does not change integrability or the integral. Equivalently, if is integrable and
then is integrable and . The proof uses uniform continuity to make the changed finite-point contributions arbitrarily small.
8.4Fundamental Theorem of Calculus
Theorem 8.10 (FTC I)
Suppose is continuous on , differentiable on , and is Riemann integrable on . Then
Notation: .
Proof
Given a partition with , MVT supplies with
Thus
Since , the difference between and is below .
Theorem 8.11 (FTC II)
Let be Riemann integrable and define
Then is uniformly continuous on . If is continuous at , then is differentiable at and .
Proof
Fix with . If , then
so is uniformly continuous. At a continuity point ,
Given , continuity gives with whenever . Thus, for and ,
Therefore .
Note: 在 处 continuous 是 FTC II 中很重要的条件。
Example 8.4 FTC examples and caveats
is an antiderivative of on because is continuous. Also
though the integral generally cannot be written in elementary closed form. By the chain rule,
More generally,
if is Riemann integrable and continuous where needed.
FTC says differentiation and integration can be inverse operations, but: (1) derivatives need not be integrable, for example has an unbounded derivative; (2) indefinite integrals need not be antiderivatives (Thomae’s function has no antiderivative), while an integral has constant zero.
Theorem 8.12 (Integration by parts)
If are continuous on , differentiable on , and are integrable on , then
In the shorthand, .
Proof
Differentiate : , then integrate on and use FTC I.
Theorem 8.13 (Change of variables)
Suppose is a continuously differentiable function on an open interval , let be an open interval with , and let be continuous on . Then for ,
Proof
If , then , so FTC I gives the equality.
Chapter 9Sequences and series of functions
9.1Sequences of functions (L19)
Definition 9.1 (Pointwise convergence)
令 是一个 seq. of functions(domains 都相同)。称 在 上 pointwise converges to ,记作 on ,if
等价地,
seq. of functions 的 pointwise convergence 即:对每一点 ,。
Example 9.1 Pointwise limits can destroy everything
On , let . Then
Every is continuous and differentiable,但 is discontinuous。因而 pointwise conv. 不 preserve continuity & differentiability。
L19 p.1 draws the family rising from to , with the limiting graph equal to before the endpoint and at the endpoint. The graph information is equivalently captured by
.
Write ( 可以任意排序)。Let
Then (Dirichlet’s function). Each is Riemann integrable, but is not;因而 pointwise conv. 不 preserve integrability.
On , let
Each triangular spike has area , so for every . Pointwise , hence
因而 pointwise convergence 不 preserve the limit of an integral。
The p.1 spike picture has base , apex , and area :
.
On , let . Then , , yet for all while . Thus
pointwise convergence 不 preserve the limit of a derivative。
因而 pointwise limit can destroy continuity, differentiability, and integrability;即使不 destroy,也不 reserve the value of an integral / derivative。pointwise convergence 是局部的逐点性质,不是整体性质:在每个 ,,最后的 由每个 的极限拼接而成。若想让 convergence 更好地保留整体性质,就需要更强的定义。
Definition 9.2 (Uniform convergence)
令 是一个 seq. of functions。称 在 上 uniformly converges to ,if
Let with every . Suppose there is a point such that converges, and uniformly. Then uniformly for some , where .
Proof
Uniform convergence of the derivatives gives, for ,
for all . Pointwise convergence at gives . Thus, for arbitrary ,
So is uniformly Cauchy, hence uniformly convergent. Letting limits in the displayed FTC identity gives
and FTC II yields .
Summary
1.
A uniform limit of continuous is continuous.
2.
Under suitable conditions,
Since differentiation and integration are very basic operations, the uniform-convergence hypotheses ensure the desired stability.
9.2Series of functions and power series (L20)
Definition 9.3 (Series of functions)
If is a sequence of functions, then is its sequence of partial sums. Write or for the infinite series determined by .
On , the following are definitions:
1.
converges on iff, for every , exists; equivalently there is with pointwise.
2.
It converges uniformly on iff those partial sums converge uniformly to some .
3.
It converges absolutely on iff converges at every ; equivalently converges on .
Theorem 9.6 (Term-by-term operations for a function series)
1.
If every is continuous on and uniformly on , then is continuous on .
2.
If every is continuous on and uniformly on , then is integrable and
3.
If every on , on (not necessarily uniformly), and converges uniformly on , then and .
Stronger version of (3): if on , there exists such that converges, and converges uniformly on , then converges uniformly to some , and .
Proof
Since every partial sum is continuous, differentiable, and integrable as appropriate, apply the corresponding uniform-limit theorem to the sequence of partial sums .
9.3Power series
Definition 9.4 (Power series)
For a sequence in , the power series centered at with coefficients is the series of functions
The partial sums are polynomials. Custom: for , ; and for every (including ).
Note: the L20 pages use power series centered at in the displayed examples, but every result applies to a center by replacing with .
Theorem 9.7 (Cauchy-Hadamard theorem)
Given a power series , let . Then it converges absolutely when and diverges when . Its radius of convergence is
The set of all for which converges is an interval, called the interval of convergence.
Proof
If , then for all but finitely many , , so and comparison applies. If , then infinitely often, so the th-term test gives divergence.
Endpoints and a ratio shortcut
Radius of convergence cannot imply interval of convergence: endpoints may or may not be included, so they must be checked separately. If exists, then is the radius; this is often the best way to find , but it is not more general than .
Example 9.2 Power-series radii and intervals
1.
For , , hence ; it converges for all , and in fact equals by Taylor.
2.
For , ; it diverges for , so the interval is , and
3.
The handwritten page writes . Its subsequent endpoint calculation treats the terms as the harmonic series from : ; at it diverges, and at it is alternating harmonic and converges. Thus the interval written is .
4.
The handwritten page likewise writes ; the subsequent endpoint sums begin at . Here and both and converge, so the interval is .
5.
For , , so and it diverges for all .
Theorem 9.8 (Weierstrass M-Test)
Let be a sequence of functions, and let be a sequence in such that
for all and . If , then converges uniformly and absolutely on .
Proof
Let . Since satisfies Cauchy, choose so that for . Then for all ,
Thus is uniformly Cauchy and converges uniformly; the same calculation gives uniform absolute convergence.
Corollary 9.1 (Uniform convergence inside a radius)
If has radius of convergence , then for every , converges uniformly to a continuous function on . Indeed and on , so M-test applies.
Consequently is continuous on . However its convergence on the entire interval of convergence may not be uniform:
converges to on as written in the source note, but the convergence is not uniform there (the graph marks the unbounded behavior at ). Fact: a uniform limit of uniformly continuous functions is uniformly continuous.
Theorem 9.9 (Abel’s theorem)
1.
If a power series converges at , then it converges uniformly on . If it diverges at , then it diverges on .
2.
If a power series has radius of convergence , then convergence at an endpoint of its radius implies convergence at every point between that endpoint and ; divergence at an endpoint implies divergence on the corresponding exterior ray.
Note: the convergence of on its interval of convergence may not be uniform.
Proof
提示一下,下边(略)。
Theorem 9.10 (Term-by-term integration and differentiation of power series)
Let have radius of convergence and let for .
1.
For every , is integrable and
2.
The power series has radius , is differentiable on , and
Proof
(i) follows from integrability of polynomials and uniform convergence of on . For (ii), for ,
so the differentiated series has radius ; its uniform convergence on compact subintervals and the preceding derivative theorem prove the claim.
Example 9.3 Taylor series, calculus, and its caveat
If , try to approximate near with
and define
where the domain is the interval of convergence of . The source records power series
Thus , , and ; . Termwise integration yields
Remark: The Taylor expansion of may not converge to at even if it converges at . Let
Then on and for all . Its Taylor series converges everywhere, but converges to itself only at . If and pointwise for all lies in the domain of , then is a real analytic function, i.e. ().
Personal Single-Variable Analysis Homeworks
Chapter 10Homework 1: sets, order, and induction
Authority label
This is a visual transcription of personal work in 451-Hw-1.pdf. 451-hw-1-raw.pdf and 451-hw-sol-all.pdf are checking-only material: they were used to identify the assignment and check notation, never as the body of the submitted solutions below.
10.1Problem 1 — set identities
For each statement about sets, either prove the statement if it is true for all sets, or give a counterexample using specific sets if it is false.
•
(a) .
•
(b) .
•
(c) .
•
(d) if and only if .
10.2Problem 2 — multiples
For each , let .
10.3Problem 3 — sum of odd integers
Guess a formula for , then prove it by induction.
10.4Problem 4 —
Determine for which integers is true, and prove the claim by induction.
Solution
The submitted claim is: or .
Case 1: . Then and , hence .
Case 2: . The proof is by induction on . Base case: , and , so .
Inductive step: assume for (where and ) that . Then and . Note that . Since , , so . Therefore , and
This finishes the proof that for all integer , .
10.5Problem 5 — boundedness, supremum, and infimum
For each listed subset of , state whether it is bounded above and below, and its supremum and infimum when they exist. The submitted one-line answers are retained below.
•
(a) : bounded below but not above; .
•
(b) : bounded below and above; , .
•
(c) : bounded below and above; , .
•
(d) : bounded below and above; , .
•
(e) : bounded below and above; , .
•
(f) : bounded below and above; .
•
(g) : bounded below and above; , .
•
(h) : bounded below but not above; .
•
(i) : bounded below and above; , .
•
(j) : bounded below and above; , .
•
(k) : bounded below but not above; .
•
(l) : bounded above but not below; .
•
(m) : bounded below and above; , .
•
(n) : bounded below and above; , .
•
(o) : bounded above but not below; .
•
(p) : bounded below and above; , .
•
(q) : bounded below and above; , .
•
(r) : bounded below and above; .
•
(s) and is prime: bounded below and above; , .
•
(t) : bounded above but not below; .
•
(u) : bounded below but not above; .
•
(v) : bounded below and above; , .
•
(w) : bounded below but not above; .
•
(x) : bounded below but not above; .
10.6Problem 6 — no ordered-field order on
原稿红字
hw 1 ①: 上不可能 define linear relation;P: 无法成为 ordered field。
Assume for contradiction that a linear relation is defined on such that Axioms 13–14 hold: if then , and if and then .
Case 1: define . By Axiom 14, multiplying both sides by gives , hence . Multiplying both sides by gives . By Axiom 13, , so . This contradicts the definition of that .
Case 2: define . Then by Axiom 4, so , contradicting Axiom 5.
Case 3: define . Then for some with . Thus , so (by Axiom 5 and Axiom 14). The same result as in Case 1 contradicts the definition of .
Since in all cases the assumption of a linear order contradicts the properties of , it is impossible to define a linear relation on such that Axioms 13–14 hold.
10.7Problem 7 — order and supremum
原稿红字
hw 1 ②:判定 的方法:满足任意 ,都存在 在 和 之间。
10.7.1(a)
Let . If for every , then .
Proof
Suppose for contradiction. By density of in , there exists such that . By the given condition, , which contradicts . Hence .
10.7.2(b)
Let and let be an upper bound of . Show that if and only if, for every , there is such that .
Proof
One direction: assume . Suppose for contradiction that, for some , there is no such that . Since , no satisfies . Combining the two statements, no satisfies . Thus is an upper bound of , contradicting the definition of supremum since .
The other direction: assume that for every there is with . Let be an arbitrary upper bound of . If , then there is with , contradicting that is an upper bound. Therefore . Since is arbitrary, .
10.8Problem 8 — bounded sets
Let and be nonempty bounded subsets of .
原稿红字
hw 1 ③:。
10.8.1(a)
Proof
Take arbitrary . By the definitions of upper and lower bounds, and . Hence by transitivity of the linear order and equivalence relation.
10.8.2(b)
If , the submitted order is .
Proof
Part (a) gives . It remains to prove and . Let ; since , . Thus every upper bound of is also an upper bound of , and every lower bound of is also a lower bound of . Therefore the lower bounds of are included in the lower bounds of , while the upper bounds of are included in the upper bounds of . Hence and .
10.8.3(c)
Proof
First claim: is an upper bound of . Let be an arbitrary element of . If , then , so . If , then , so again . Hence is an upper bound of .
Let be an arbitrary upper bound of . Then is an upper bound of both and . Suppose . Without loss of generality suppose . Then is not an upper bound of , a contradiction. Therefore , which proves .
10.9Problem 9 — supremum of a sum set
Let and be nonempty bounded subsets of , and let and . Prove .
Proof
First claim: is an upper bound of . Let be an arbitrary element of (, ). Then and , hence . Thus .
Now show . Assume for contradiction that . Then, for some , . By definition of supremum, is not an upper bound of , so there is with . Similarly, there is with . Therefore but , a contradiction. Thus .
10.10Problem 10 — density of irrationals
Prove that is dense in .
Proof
Take arbitrary with . Then for some with . By the Archimedean property of , there exists such that , so . Consider , which is irrational since and is irrational. Also , since ; by Axiom 14, . Therefore . Hence is dense in .
10.11Problem 11 — discrete sets
A set is discrete if for every there is such that , where .
原稿红字
hw 1 ④:任意 finite 一定 discrete。
10.11.1(a)
Proof
Let be an arbitrary finite set and let be arbitrary. Consider . This is finite since is finite, so has a smallest element. Let . Then , where is the distance of from its nearest element in . Thus . Since is arbitrary, is discrete.
10.11.2(b)
False — counterexample
Consider . This is a discrete set: for any , consider . Then , so . But no uniform exists. If it did, then for all , so for all , which contradicts the Archimedean property of .
10.12Problem 12 — optional challenge problem
For , let and . The submitted answer, without a proof, is
.
Chapter 11Homework 2: cardinality and sequences
Authority label
This is a visual transcription of personal work in 451-hw-2.pdf. The raw assignment and consolidated solution PDF are checking-only sources; neither supplies the solution body below.
The submission uses for the existence of an injective function from to , and for the existence of a bijection. It recalls Cantor–Schröder–Bernstein: if and only if and .
11.1Problem 1 — triangle inequality for finite sums
原稿红字
hw 2 ①:。
For , prove by induction that .
Proof
We prove it by induction on . Base case: , and , so the claim holds.
Inductive step: assume the inequality holds for all for . Then
By the inductive hypothesis for , . Combining this with (1), . This finishes the proof.
11.2Problem 2 — bounds of a scalar multiple
Let be bounded, let , and write .
Solution
If , the submitted expressions are and ; if , both are ; and if , they are and .
For , take arbitrary . Since , , so is an upper bound of . If is an upper bound of , then for all . Since , , so is an upper bound of . Thus , hence . Therefore .
For , , so . For , take arbitrary . Since , , so is an upper bound of . If is an upper bound of , then for all . Since , , so is a lower bound of ; hence and . Thus .
11.3Problem 3 — injective maps and
原稿红字
hw 2 ②: 是一个 partial order。
Let and be functions.
11.3.1(a)
Proof
Suppose and are injective. Let , where are in the domain of . Since is injective, ; since is injective, . Hence is injective.
11.3.2(b)
Proof
Let be an arbitrary set. The function defined by is injective by uniqueness of every element in a set, so . Thus is reflexive.
Let and . There are injective functions and . By part (a), is injective, so . Therefore is transitive.
11.4Problem 4 — inclusions, injections, and surjections
原稿红字
hw 2 ③:。
11.4.1(a)
Proof
Assume . Consider defined by . It is injective by uniqueness of every element in a set. Therefore .
11.4.2(b)
Proof
First suppose is injective. Let be arbitrary and define by if is in the range of , and if is not in the range of . This function is well-defined since is injective, so there is only one element in for each . Thus The range of is , so is surjective.
Conversely suppose is surjective. For every , there is some with , i.e. . Define by sending every to some . Its well-definedness is guaranteed by ; it is injective because . This finishes the if-and-only-if proof.
11.5Problem 5 — remove a finite or countable subset
原稿红字
hw 2 ④:infinite set / finite set / unctb set / ctb set,基数不变。
11.5.1(a)
Proof
First construct by . It is injective, hence . Let for some . Since is infinite and is finite, is infinite. Take a countable subset . Define piecewise: for , let ; for with , let ; and for with , let . The work records that it is well-defined since , and injective since . Thus . Cantor–Schröder–Bernstein gives .
11.5.2(b)
Proof
Since is countable, write . Take a countably infinite subset . Define by the same three cases as in part (a): it fixes , sends to , and sends to . The submission records that this is well-defined, injective, and surjective: every is either in or in , and in either case there is with . Therefore .
11.6Problem 6 — algebraic and transcendental real numbers
原稿红字
hw 2 ⑤:代数无理数 unctb。
11.6.1(a)
Proof
Let be the set of all roots of polynomials with rational-number coefficients with terms. By definition, . For arbitrary , let be the polynomial with those coefficients. Then . Since , the fundamental theorem of algebra gives that has at most roots. Thus every is finite. Because is countable, is countable for each , and so is countable.
Since is uncountable and is countable, is uncountable (and ). This indicates uncountably many transcendental numbers.
11.6.2(b)
Proof
Let with , and let be the set of all algebraic numbers in . Since and is countable, is countable. Since is uncountable, is uncountable. Thus there are uncountably many transcendental numbers in .
11.7Problem 7 — power sets and functions
原稿红字
hw 2 ⑥:。
11.7.1(a)
Proof
Let be arbitrary. Consider the characteristic function defined by if , and if . Define by for each . If , then , so every is in and every is in ; hence . Therefore .
11.7.2(b)
Proof
Assume for contradiction that there is a surjective function from to . By Problem 4(b), there is an injective function from to , so . Together with and by Problem 3(b), there is a surjective function from to , contradicting Cantor’s theorem. Hence no surjective function from to exists.
11.8Problem 8 — direct proofs of sequence limits
11.8.1(a)
Proof
Let . Take by the Archimedean property, so . For , . Thus .
11.8.2(b)
Proof
Let . Take , so and . For , . Thus .
11.9Problem 9 — absolute values and powers
原稿红字
hw 2 ⑦:;。
Proof
Let and fix such that whenever . Since and , the submission concludes , hence . Therefore .
11.10Problem 10 — powers of a convergent sequence
Proof
The proof is by induction on . Base case: and . Assume for that . Then
by the limit law. Thus if converges to , then for all .
11.11Problem 11 — successive differences
Let . If converges, prove converges to zero.
Proof
Since converges, for some . Let and fix such that whenever . Then and , so . Hence and .
11.12Problem 12 — a sequence converging to
原稿红字
hw 2 ⑧: bounded in 。
Let be a bounded nonempty subset of . Show that there is a sequence in converging to .
Proof
Consider for . By the definition of supremum, is not an upper bound of ; for each , there exists some where . Take one such as for each (the same can be taken repeatedly). Then is a sequence in .
Let and take with , so . For , . Since and , . Therefore .
11.13Optional challenge problems
The personal PDF prints Problem 13(a)–(b), concerning as a union of open intervals and as an intersection of closed intervals, and Problem 14, asking whether the converse of Problem 11 is true. No personal answer is written on source page 14; source pages 15–16 are blank.
Chapter 12Homework 3: sequence limits and topology
Authority label
This is a visual transcription of personal work in 451-hw-3.pdf. The raw assignment and consolidated solution file were used only for checking problem identities and notation; they are not the authority for any solution below.
The submission begins with the definition: a sequence of real numbers is eventually constant if there are and such that for all .
12.1Problem 1 — reciprocals and divergence to infinity
Consider the bi-implication .
Forward direction
Suppose . Let and consider . Then for some , whenever . Thus implies whenever . So .
Backward direction — counterexample
Consider . Then , but .
12.2Problem 2 — a bounded factor
Let and be sequences of real numbers. Prove that if and is bounded, then .
Proof
Since is bounded, is also bounded. Consider the constant sequence . Then . Since , , and hence . As , for all , so for all . By the squeeze theorem, . Therefore .
12.3Problem 3 — three limits
Determine the limits in the extended real line (including positive or negative infinity) of the following sequences and prove the results.
12.3.1(a)
Proof
The submitted answer is . Let . Then
So .
12.3.2(b)
Proof
The submitted answer is . It writes . Let and choose an integer . For , . Hence the limit is .
12.3.3(c) ,
Proof
Assume . Then , so and . Since for all , the limit can only be if it exists.
Now prove converges. For , . Since , for all . Also . Hence is decreasing and bounded below, so it converges. Therefore .
12.4Problem 4 — limits in a discrete set
原稿红字
hw 3 ①:discrete 中的任意 seq 要么 eventually constant,要么 在 之外。
Suppose is a discrete subset of , and is a convergent sequence of numbers in . Prove that either is eventually constant or .
Proof
Write . Assume is not eventually constant and . Since is discrete, there is some such that . Since , there is such that for all . Since is not eventually constant, there is such that and , i.e. . Thus , a contradiction. Therefore is either eventually constant or .
12.5Problem 5 — sequences of rationals with bounded numerators
For positive integer , let be the set of rational numbers with and . Prove every sequence of distinct numbers in converges.
Proof
Let be an arbitrary sequence in and let . Since for each there are only finitely many terms of that have as a denominator, consider for some and an integer with Take arbitrary . Then where . Thus . So . This finishes the proof that every sequence of distinct numbers in converges.
12.6Problem 6 — strict inequalities between sequences
Let for all .
12.6.1(a)
Proof
Suppose . Let and fix it. Then for some , whenever . Since for all , for all . Therefore .
12.6.2(b)
Solution
Consider and for all . Then for all , but .
12.7Problem 7 — ratio limit greater than one
原稿红字
hw 3 ②:if positive seq and ,则 。
Let be a sequence of positive real numbers. Show that if , then .
Proof
Let . Since , there is some such that for all , i.e. for all . Let . There is some such that , since . Then for all , . Therefore .
12.8Problem 8 — lim sup and lim inf
Find the lim sup and lim inf of the following sequences.
•
(a) : and .
•
(b) : .
•
(c) any bijection: and .
•
(d) : .
12.9Problem 9 — a recursive average
Let with . Let , , and . The submitted claim is .
Proof
Let for all . Then , and, for ,
For all ,
Thus
since .
12.10Problem 10 — a divergent sequence with one possible subsequential limit
Consider , i.e. . The work claims diverges, but every convergent subsequence converges to .
Proof
The odd-indexed terms is odd. Let be a convergent subsequence of ; then is strictly increasing. Suppose there are infinitely many such that is even. We show diverges. Let , take , and fix . If there is no with , then there are only finitely many even , a contradiction. Thus there must be with , so diverges. Hence only finitely many are even. Cutting the tail makes all remaining odd, and so converges to .
12.11Problem 11 — lim sup of a sum
Let and be bounded sequences of positive real numbers.
12.11.1(a)
Proof
Write , , and . Let . Then for every , and , so . Hence . Since is arbitrary, , i.e. .
12.11.2(b)
Counterexample
, so . Let , so . But .
12.11.3(c)
Solution
Write . Then since converges. Thus .
12.12Problem 12 — a sequence with every real subsequential limit
原稿红字
hw 3 ③: 中存在一个 seq,使得 “all sub seq lim of ” 。
Proof
Since , there exists a surjective function . Note that is a sequence. Let be arbitrary. There exists a sequence in , . Since is surjective, consider the subsequence of defined by for some , for all . Take a monotonic subsequence of as . It is a subsequence of , so it is also a subsequence of . Let . There is some such that whenever . Since there is some term with and is monotonic, whenever . Therefore .
12.13Problem 13 — open and closed sets
The submitted classifications are:
•
(a) : neither.
•
(b) : closed and not open.
•
(c) : open and not closed.
•
(d) : closed and not open.
•
(e) : neither.
•
(f) : closed and not open.
12.14Problem 14 — closed discrete set with no uniform separation
Counterexample
Consider , a partial sum of the harmonic series, and . There is no subsequential limit in , so has no limit point; hence and is closed. For each consider ; then , so is discrete. But there is no such that for every pair , since for any , (as recorded in the submission).
Suppose is infinite, bounded, and discrete. Prove that there is a convergent sequence in whose limit is not in .
Proof
Take an arbitrary sequence in such that , . By the Bolzano–Weierstrass theorem, there is a convergent subsequence ; write . Claim: . Suppose . Since is the limit of a sequence in , it is a limit point of , so for every there exists with , i.e. . Since is discrete and , there exists such that . These two statements contradict. So .
Chapter 13Homework 4: limits and closure
Authority label
This transcription follows the personal handwritten work in 451-hw-4.pdf. The raw assignment and sample solutions are used only to check problem numbering and notation; they are not substituted for personal work.
原稿红字旁注
•
hw4① 未必有 conv.;反例:。
•
hw4③ ,都有 、 为 closed 的,且 为最小的含 闭集。
•
hw4④ composition limit law 需搞清楚 upper/lower limit。
•
hw4⑤ 。
•
hw4⑦ 如果 不是 closed,那么 上存在 unbounded ctn function。
13.1Problem 1
Do Challenge Problem (14) from HW 2: if is a sequence in and , must converge? Justify your answer.
No. A counterexample is . Then
but .
13.2Problem 2
Let be a sequence in , and let be its set of real subsequential limits. Prove that is closed.
Proof
Let be arbitrary. We show : that is, there is a subsequence of converging to .
Let . Since , . Choose . Then and , so there is a subsequence of such that as , where is monotonically increasing. Hence there is such that, for all , .
Construct recursively. For , choose as . If and , take and choose as . Then
Thus is a subsequence of , because every term is a term of with increasing index. For , choose with , and take . Then for every . Thus .
We have proved is a subsequential limit of . Since was arbitrary, , so is closed.
13.3Problem 3
Given , write for the set of all limit points of and define . (a) Prove that is closed. (b) Prove that is closed. (c) Prove that is the smallest closed set containing .
13.3.1(a)
Proof
Let . Then is not a limit point of , so for some ,
Let be arbitrary. Since , . Thus , which implies . Hence . Since is arbitrary, is open, and so is closed.
13.3.2(b)
Proof
Let . Then and . Fix such that . Since , also . If , then and , so and . Therefore . Thus is open, and is closed.
13.3.3(c)
Proof
Let be a closed set with . Let be arbitrary and let be a sequence in converging to . Since and is closed, . Thus , so . Since was arbitrary, this proves that is the smallest closed set containing .
13.4Problem 4
(a) Prove explicitly using the definition that . (b) Given , find the largest such that whenever . (c) Prove explicitly using the definition that . (d) Given , find the largest such that whenever .
13.4.1(a)
Proof
Let . Since
and, for , , take . If , then
Hence .
13.4.2(b)
For we want and . Thus
The personal calculation records the largest value as .
13.4.3(c)
Proof
Let . Since
and , take . If , then
Hence .
13.4.4(d)
For we want and . Thus
The personal calculation records the largest value as .
13.5Problem 5
Let , let , suppose is a limit point of , and suppose . Let and suppose . Prove that .
Proof
Let . There is such that whenever . Also, since , there is such that whenever . Therefore, if , then . Hence .
13.6Problem 6
Let , let , and suppose and . Show by example that need not be the limit of as .
Consider if , and . Also, let and if . Then and , but if and . Hence .
13.7Problem 7
Prove that for any sequence of nonzero real numbers, .
Proof
Let be arbitrary. Then there is such that whenever . For ,
Hence
The final factor tends to , and hence . Since this holds for every , the required inequality follows.
13.8Problem 8
Let , suppose , and let . Prove that if and is continuous at , then there is such that is positive and bounded on .
Proof
Since is continuous at , there is such that whenever and . Thus whenever . Therefore is positive and bounded on .
13.9Problem 9
Suppose are continuous. Prove that if for all , then .
Proof
Let be arbitrary, and let . By continuity of , there is such that whenever . By density of in , choose . Then ; similarly, . Since , , and so
Thus . Since was arbitrary, on .
13.10Problem 10
Prove that if is not closed, then there is an unbounded continuous function .
Proof
Since is not closed, choose with . Define by . This is well defined, and is continuous as a composition of the continuous rational function and the absolute-value function.
Let . Since , there is with . Thus . Hence is unbounded.
13.11Problem 11
Using only the definitions of continuity and open set, prove that for any , is continuous if and only if is open for every open set .
Proof
Suppose is continuous and let be open. If , then , so there is with . By continuity, there is such that whenever . Thus , proving open.
Conversely, suppose is open for every open . Let and , and take . Then is open and contains , so some lies in . Therefore whenever . Thus is continuous at , and hence continuous.
13.12Problems 12–14
The source records these printed problems but no handwritten response:
•
(12) If is closed and is continuous, prove there is a continuous with .
•
(13) For pairwise disjoint nonempty open sets in , prove is countable.
•
(14a) Prove an open subset of is a union of countably many open intervals; (14b) decide whether the intervals can be chosen with rational endpoints.
Chapter 14Homework 5: uniform continuity and differentiation
Authority label
This is a transcription of personal work in 451-hw-5.pdf. The raw assignment and sample solutions are retained only as checking aids for problem numbering and symbols.
Suppose is a family of nonempty open sets in such that whenever . Prove that is countable.
Proof
Let be arbitrary, and let . By definition there is with . By density of in , choose with . Define by sending each to a rational number in . Since the are pairwise disjoint, is injective. Hence in the sense of an injection, so is countable.
14.2Problem 2
Determine whether each continuous function is uniformly continuous on the given interval: (a) on ; (b) on ; (c) on ; (d) on .
14.2.1(a)
is uniformly continuous because it is continuous on and is closed and bounded.
14.2.2(b)
Let and take . If and , then
Thus is uniformly continuous on .
14.2.3(c)
It is not uniformly continuous. Take . Let be arbitrary and take , . Then
14.2.4(d)
It is not uniformly continuous. Take . Given , take and . The source records the computation
and uses the preceding lower bound while to obtain a quantity greater than .
14.3Problem 3
Prove that if there is such that a continuous is uniformly continuous on , then is uniformly continuous.
Proof
Suppose is continuous and uniformly continuous on . Since is closed, is uniformly continuous there. Let . Take for and for , each giving . Set .
Let with . If both points are in , use ; if both are in , use . In the remaining case, assume and . Then and , so
Thus is uniformly continuous.
14.4Problem 4
Let , let be continuous, and suppose . Suppose further that is uniformly continuous on for some . (a) Prove that any two sequences and in converging to have the same -limit. (b) Prove that extends continuously to .
14.4.1(a)
Proof
Assume the hypotheses and write . Let . Since , there is such that whenever . Uniform continuity gives with whenever . Since , this holds beyond some . Also choose with for . For and ,
Therefore .
14.4.2(b)
Define
Then . Since and , . The preceding part gives , so is continuous at . It is already continuous on , and hence is continuous on its domain.
14.5Problem 5
Show that a composition of uniformly continuous functions is uniformly continuous: if and are uniformly continuous and , then is uniformly continuous.
Proof
Let . Take such that when , for . Take such that when , for . Then implies
14.6Problem 6
Find the derivatives from the definition: (a) ; (b) .
14.6.1(a)
14.6.2(b)
14.7Problem 7
Define by for and for . Find all points where is continuous and differentiable (no justification needed).
The personal answer: is continuous only at , and differentiable only at .
14.8Problem 8
Show that if for all , then is constant.
Proof
Fix and let be arbitrary. Consider . The hypothesis yields . By the squeeze theorem, , and the source concludes . Since is arbitrary, is constant.
14.9Problem 9
If and are differentiable on , , and for all , prove for all .
Proof
Let . Then , so is decreasing on . Since , for we have . Thus .
14.10Problem 10
Let . Decide each assertion: (a) differentiable is bounded; (b) such is bounded; (c) differentiable, bounded has bounded ; (d) differentiable with bounded is bounded.
14.10.1(a)
True. A differentiable function is continuous, so the extreme value theorem on the closed, bounded interval gives with for all . Hence the function is bounded.
14.10.2(b)
False. The source gives for and on . It records , while for ,
which is unbounded near .
14.10.3(c)
False, by restricting the same counterexample to .
14.10.4(d)
True. Suppose on . Choose ; then is differentiable on , so the extreme value theorem gives a point controlling there. For arbitrary , the mean value theorem gives a point between and with . Thus
so is bounded.
14.11Problem 11
For differentiable , decide the converses of: (a) implies increasing; (b) implies strictly increasing.
14.11.1(a)
Proof
The converse is true. Assume is increasing and let . If , take . The derivative definition gives a such that for ,
But and increasing imply , a contradiction. Hence .
14.11.2(b)
The converse is false: is strictly increasing on , but .
14.12Problem 12
Let be differentiable. Prove that if and both exist, then .
Proof
Assume the hypotheses and, for a contradiction, suppose . Write . Let . Choose so that for and for . For , and for .
Take . By the mean value theorem, some has
Hence , so , contradicting the choice of . Thus the derivative limit is .
14.13Problem 13
Let be differentiable at . (a) If , prove that there is such that for . (b) Decide whether this implies is strictly increasing on .
14.13.1(a)
Proof
Let . By differentiability there is such that
for . Hence , and, since , .
14.13.2(b)
The personal answer is true. Use the same and let . By the mean value theorem,
for some . By (a), , so and is strictly increasing on .
14.14Optional challenge problems 14–15
The source has only the printed prompts and no handwritten response:
•
(14) For increasing and decreasing with , decide whether must be nonempty, given that .
•
(15) Decide whether an open can contain while is uncountable.
Chapter 15Homework 6: sequences, series, and integrability
Authority label
This transcription follows the handwritten personal submission. The raw PDF and the solution collection are reference-only aids for problem numbering and symbols; they do not provide any body prose or proof here.
hw6⑦ 如果 intble,而 只有 finitely many 个 pt 上和 不一样,那么 也 intble,且 (但不推荐用 infinitely many 个 pt.)。
•
hw6⑧ 如果 ,那么一定有 s.t. 。
15.1Problem 1
Let and be bounded sequences in , with . Show that .
Proof
Let denote the set of subsequential limits of , so . Write .
Claim 1. is an upper bound for . Let be an arbitrary convergent subsequence. The source calculates
Thus is an upper bound for .
Claim 2.. Choose a subsequence with . Then , so . The two claims give .
15.2Problem 2
(a) For each , find the th derivative of and prove the claim by induction. (b) For , define for and for . Show is -times differentiable but not -times differentiable.
15.2.1(a)
Proof
The th derivative of is . For , . Assuming the statement for ,
This proves the formula by induction.
15.2.2(b)
For each , the source writes
It is -times differentiable away from , with for and for . At ,
and the matching left-hand calculation is also , so . But the right derivative quotient of at is while the left quotient is . Therefore the next derivative does not exist.
15.3Problem 3
Prove by induction: for all , if is -times differentiable and for all , then is a polynomial of degree at most .
Proof
The personal proof uses induction on . For , differentiability gives continuity and , hence for some , a polynomial of degree .
Assume the statement for . For , is -times differentiable and . Hence for some real coefficients . Thus
for all , a polynomial of degree at most .
15.4Problem 4
Show that diverges, but converges for every .
Proof
By the integral test, the second series converges iff converges. With , this is
For , it equals , proving divergence of the first series. For , it equals , which the submission records as , hence convergent.
15.5Problem 5
Show that converges conditionally.
Proof
Claim 1. The series converges. The personal work records, for every , and hence . It concludes the positive terms are decreasing and have limit , so the alternating series test applies.
Claim 2. The absolute-value series diverges. The work states the limit comparison test: for positive with , the two series converge or diverge together. Its proof takes to get , hence beyond a finite tail. For the present series,
Thus diverges with the harmonic series. The original alternating series therefore converges conditionally.
15.6Problem 6
Give a positive sequence converging to zero such that diverges.
The example is for even and for odd . Both the even and odd subsequences tend to , so . The work groups terms as
which diverges.
15.7Problem 7
Determine whether each series converges: (a) ; (b) ; (c) ; (d) ; (e) .
15.7.1(a)
With , , so the ratio test gives divergence.
15.7.2(b)
With , and . The terms do not have a limit, so the series diverges by the nth-term test.
15.7.3(c)
For , the source uses . Thus , and comparison with the harmonic series gives divergence.
15.7.4(d)
Let . Then . The work notes , , and is decreasing because when . Thus the alternating series converges.
15.7.5(e)
Let . The quotient is unbounded above, so the ratio test gives divergence.
15.8Problem 8
If and converge, prove that converges absolutely.
Proof
Write and . Cauchy–Schwarz gives
for every . Thus the partial sums of are bounded above by and below by , and they are increasing. Hence converges, so converges absolutely.
15.9Problem 9
Show that if is integrable on , then it is integrable on every subinterval .
Proof
Suppose, for a contradiction, that is not integrable on . Since is integrable on , there is such that every tagged partition of mesh less than has . Nonintegrability on gives tagged partitions there with meshes below but .
Refine both partitions to by adding regular extra points with the same tags. The resulting have mesh below , while their sum difference is exactly the displayed difference over , a contradiction. Thus is integrable.
15.10Problem 10
If is integrable on , show that for every infinite there is equal to off but not integrable.
Proof
Let be infinite. As a bounded infinite set it is not discrete, so choose such that every is nonempty. Define
Given an arbitrary , take , choose , and obtain . Thus is unbounded above and hence not Riemann integrable.
15.11Problem 11
Show directly that if a bounded is continuous everywhere except possibly at , then is integrable.
Proof
Take with . Let . Uniform continuity on and gives such that their oscillations are less than and , respectively. Set . For a partition of mesh less than , let . On all earlier intervals, taking a midpoint gives total upper-minus-lower contribution below ; the same reasoning gives below for intervals after ; and on the oscillation is at most , giving contribution below . Hence , and is integrable.
15.12Problem 12
Suppose and are continuous on and . Prove that some satisfies .
Proof
Assume the hypothesis. If for every , then the source states on every subinterval and hence , a contradiction. Therefore some has . By the same reasoning, some has . If neither is equality, is continuous and has opposite signs at . The intermediate value theorem gives an between them with .
15.13Optional challenge problem 13
The source records the printed definition on , on , and on , followed by , and asks to prove is continuous at iff .
Chapter 16Homework 7: integration and convergence
Authority label
This is a transcription of the handwritten personal submission in 451-hw-7.pdf. The raw assignment and solution collection are checking-only reference material; they supply no substituted proof text.
hw7⑦ 任何一个 closed interval 上的 ctn function 都可以用一个 step function seq. 来逼近(欲证 step unily.)。
16.1Problem 1
Prove that if is continuous on , there is such that .
Proof
By the extreme value theorem, there are with for all . Since is continuous, it is integrable, and monotonicity gives
Hence
By continuity of between and and the intermediate value theorem, there is with the required equality.
16.2Problem 2
(a) Let be nonnegative and continuous. Prove that if for some , then . (b) Let continuous have for all . Prove that equal integrals imply .
16.2.1(a)
Proof
Let satisfy . By continuity, there is such that for all (by HW 4, problem 8). This set is an interval; fix a closed interval inside it. Then is integrable on , and by problem 1,
for some . Since on ,
16.2.2(b)
Proof
Suppose the integrals are equal but . The function is nonnegative and continuous. Since everywhere and , there is with . Part (a) gives , that is, , a contradiction. Therefore .
16.3Problem 3
(a) If is integrable on , prove there is with . (b) Give an example showing need not be in .
16.3.1(a)
Proof
By the fundamental theorem of calculus, is continuous on . Since
the intermediate value theorem gives with . Therefore
16.3.2(b)
Take , , and . It is continuous and integrable on , and gives .
16.4Problem 4
Compute: (a) ; (b) .
16.4.1(a)
Since is continuous at , the fundamental theorem gives differentiable at . Thus
16.4.2(b)
The submission rewrites
The derivative factor tends to by the fundamental theorem and , so the limit is .
16.5Problem 5
For and , let . (a) Find the pointwise limit. (b) Prove uniform convergence on for . (c) Decide uniform convergence on .
16.5.1(a)
The source computes
Indeed, for , at , and for .
16.5.2(b)
Proof
Let and . For , , and hence . Since , choose with for . Then
for all . Thus converges uniformly on .
16.5.3(c)
The sequence does not converge uniformly to on . Take and let be arbitrary. Since , there is such that whenever . Take . Then and .
16.6Problem 6
If is a sequence of uniformly continuous functions on and uniformly, prove that is uniformly continuous.
Proof
Let . Choose such that for all . Since is uniformly continuous, choose with whenever . Then
16.7Problem 7
Give a sequence of continuous converging pointwise but not uniformly to a continuous limit.
Take
Then pointwise on . But with , for arbitrary choose ; then , so the convergence is not uniform.
16.8Problem 8
Let be a sequence of functions on such that converges uniformly. Prove that if converges for some , then converges for all .
Proof
Let . The source chooses so that, for ,
for all , and . For an arbitrary , the fundamental theorem gives
Thus is uniformly Cauchy and hence converges uniformly on .
16.9Problem 9
A step function on is constant on every open part of a finite partition. Prove that every continuous is the uniform limit of step functions satisfying .
Proof
For , let where , and define
when . Then .
Let . Uniform continuity of gives such that if . Choose with . For and , take the partition interval containing . The extreme value theorem gives in it with ; then
Thus uniformly.
16.10Problem 10
Suppose is a power series with . Prove convergence for and divergence for , where .
Proof
If , then , so
The ratio test gives absolute convergence. Likewise, when , this quotient limit is greater than , so the series diverges by the ratio test.
16.11Problem 11
Find radii and exact intervals of convergence: (a) ; (b) ; (c) .
16.11.1(a)
, so the radius is . At , diverges; at , diverges. Thus the interval is .
16.11.2(b)
so the radius is . At , the series is , and at it is ; both converge. The interval is .
16.11.3(c)
The radius is infinity and the interval is .
16.12Problem 12
Define by for and . (a) Show by induction that for , with a polynomial. (b) Show for every polynomial . (c) Show exists and equals . (d) Give the stated example.
16.12.1(a)
Proof
The base case is
For the induction step, suppose , where . Then
the product of and another polynomial in .
16.12.2(b)
Let . The source applies L’Hopital’s rule, times, term-by-term to and obtains . Thus .
16.12.3(c)
Proof
Induct on . For ,
by part (b). If , then
for some polynomial , again by part (b).
16.12.4(d)
The source gives for and .
16.13Problems 13–14
The final two printed problems have no personal handwritten response.
•
(13a) The piecewise made of , , and is to be shown differentiable and uniformly convergent to ; (13b) is to be used to show that uniform convergence need not commute with derivatives.
•
(14) Enumerate , set on , and prove convergence and continuity on the irrational domain while limits fail at rational points.
Multivariate Mathematical Analysis
Chapter 17Metric spaces and compactness
17.1Metric spaces, norms, and topology
Definition 17.1 (Metric space)
A metric on a set is a function satisfying, for all : (symmetry), and iff (positivity), and (triangle inequality). The pair is called a metric space.
Example 17.1 Metrics recorded in the lecture
On , , and also . On the notes use , , and . In , their unit balls are the circle, square, and diamond, labelled (Euclidean), supremum, and metrics.
For , the lecture also writes and .
Definition 17.2 (Neighborhoods, open and closed sets)
In a metric space , the -neighborhood of is . A set is open when every has an with . A set is closed iff is open.
Lemma 17.1 (Equivalent Euclidean and supremum topologies)
A set is open for the Euclidean metric iff it is open for the supremum metric.
Proof
The norm comparison written in the notes is . Hence , which transfers the ball criterion for openness in both directions.
Definition 17.3 (Limit point and closure)
If , a point is a limit point of when for every . The closure is . Thus has , while has and .
Lemma 17.2 (Closure facts)
For , the set is closed; iff is closed; and if with closed, then . Thus the closure is the smallest closed set containing .
Proof
If , then some misses ; consequently is open. If is closed, every point outside has such a ball, hence . Conversely, is closed. The final assertion follows because a point of has a ball disjoint from .
Lemma 17.3 (Supremum in the closure)
If is nonempty and bounded above, then . In particular, if is closed, ; similarly a closed bounded-below set contains its infimum.
17.2Compactness in
Definition 17.4 (Open cover and Compactness)
An open cover of is a family of open sets such that . The set is compact if every open cover has a finite subcover. A set is bounded when it lies in some .
Closed subsets of compact metric spaces are compact. A compact subset of a metric space is closed and bounded. A family of compact sets with every finite intersection nonempty has nonempty total intersection.
Proof
For the closed-subset result, adjoin to an open cover of a closed and discard it after taking a finite subcover of . For closedness of a compact , cover by the sets for a fixed ; a finite subcover yields a ball about disjoint from . For boundedness use the cover . The finite-intersection assertion follows by applying compactness to the complementary open cover.
Theorem 17.2 (Nested interval and box properties)
If is a nested sequence of closed, nonempty bounded intervals, then . Hence a nested sequence of closed boxes has nonempty intersection.
Proof
Write . The increasing bounded sequence has a supremum ; then for every . Apply this coordinatewise to .
Theorem 17.3 (Closed boxes are compact)
Every closed box in is compact.
Proof
Suppose an open cover of a closed box has no finite subcover. Divide it into equal subboxes and choose one without a finite subcover; recursively obtain nested boxes . The nested-box property gives a point . Any cover member containing contains a small ball about ; for large , lies in that ball, a contradiction.
Theorem 17.4 (Heine-Borel)
A subset of is compact iff it is closed and bounded.
Proof
The forward implication was established above. If is closed and bounded, it lies in a closed box, which is compact; therefore is compact as a closed subset of a compact set.
Example 17.2 Why the Euclidean conclusion is special
Let be the space of bounded sequences with the supremum metric and let . The notes ask one to verify that is closed and bounded, and emphasize that it is not compact. Thus closed and bounded’ ‘ is not the general metric-space criterion.
17.3General metric spaces
Definition 17.5 (Total boundedness and completeness)
A subset of a metric space is totally bounded if for every there are with . A set is complete if every Cauchy sequence in it converges to a point of it. It is sequentially compact if every sequence has a subsequence converging in the set.
Theorem 17.5 (Metric compactness criteria)
For in a metric space, the following are equivalent: is compact; is sequentially compact; is complete and totally bounded.
Proof
Sequential compactness implies total boundedness: otherwise choose points separated by a fixed , producing a sequence with no Cauchy, hence no convergent, subsequence. It also implies completeness because a convergent subsequence of a Cauchy sequence forces the entire sequence to converge to the same limit.
Conversely, total boundedness lets one choose successively infinitely many terms of a given sequence in nested balls of radii ; the selected subsequence is Cauchy and therefore converges by completeness.
For the passage from sequential compactness to compactness, the notes prove the Lebesgue covering lemma: for every open cover of a sequentially compact set there is an such that each has contained in a cover member. If not, choose points for which no fits; a convergent subsequence contradicts openness at its limit. A finite -ball cover then selects a finite subcover.
Chapter 18Multivariable differentiation
18.1Continuity and differentiability
Definition 18.1 (Continuity and uniform continuity)
A map between metric spaces is continuous at if for every there is such that implies ; equivalently, . It is uniformly continuous if can be chosen independently of .
If is continuous and is compact, then is uniformly continuous.
Proof
For each , continuity provides a ball mapped into . Take a finite subcover and put . If , choose with ; then and the triangle inequality gives .
Remark
The notes record that continuous functions map compact sets to compact sets, and that a continuous real-valued function on a compact set attains a maximum and a minimum. The example is not uniformly continuous.
Definition 18.2 (Differentiability)
Let be open and . The map is differentiable at if there is a linear map such that . The linear map is unique and is denoted .
Proof
If both satisfy the definition, then is bounded by the two remainders and tends to zero. A nonzero matrix has a vector on which this quotient is nonzero, so .
Remark
The Euclidean norm is used in the written definition, but the notes stress that any norm would give the same notion. is the best linear approximation to , and the remainder is sublinear: .
Definition 18.3 (Directional and partial derivatives)
For , the directional derivative, when it exists, is . The th partial derivative is .
If is differentiable at , then every directional derivative exists and . In particular, is linear.
Proof
Substitute into the differentiability remainder. For vector-valued this is componentwise.
Remark
Directional derivatives may exist without differentiability. Conversely, differentiability is a local approximation by a linear map, not merely a collection of one-dimensional limits. For and , the notes compute .
18.2Jacobians and the criterion
Theorem 18.3 (Jacobian and components)
Let . If is differentiable at , then . Conversely, is differentiable iff each component is differentiable.
Proof
The th column is , whose entries are . The lecture’s example is , for which and .
Definition 18.4 ( and )
A function is if all partial derivatives of order at most exist and are continuous. It is if it is for every . Higher derivatives are defined componentwise using multi-indices.
Let , with open. If all first partial derivatives exist in a neighborhood of and are continuous at , then is differentiable at . Thus every map is differentiable.
Proof
Reduce to a scalar component. With set , . Apply the one-variable mean value theorem to on . For some points on the successive segments, . Subtract and use ; continuity makes the remainder quotient tend to zero.
Remark
The converse is false: (with the value at zero) is differentiable but its derivative is not continuous. The handwritten notes emphasize that, unlike , the derivative of a general takes values in a different function space.
18.3Higher derivatives and products
Theorem 18.5 (任意二阶 partial 可交换)
Last time we proved: if , then .(任意二阶 partial 可交换。)
Proof
In the scalar two-variable case, put . Applying the one-variable mean value theorem twice gives both and , where the intermediate points tend to . Continuity of the second partials gives the result.
Theorem 18.6 (Higher partial regularity)
if and only if all partials of are in .
Theorem 18.7 (Corollary(因而))
如果 is ,then for every ,
for any permutation 。(即 , 的 -order 的 partial derivative 可以随意换顺序。)For example, if is ,
.
Definition 18.5 (定义 multi-index notation)
一个 -tuple is a multi-index, s.t. each . If is a multi-index, define its degree (or order) by , and write (note: ). For , ; for , 。每个运算符 只对 求导,随后可按任意顺序排列。For example, for , .
Theorem 18.8 (Multinomial theorem)
For and , .
Remark
is the number of ways to divide a set of size into disjoint subsets of sizes . There are ways to order the set, and ways to get the same result.
Theorem 18.9 (Higher-order product rule)
If are , then .
Proof
The case is the usual product rule. For the induction step, write and differentiate the induction formula for ; reindex the two sums to obtain the multinomial coefficient .
18.4Chain rule and Taylor’s theorem
Theorem 18.10 (Chain rule)
Let and , with open. If is differentiable at and is differentiable at , then is differentiable at and .
Proof
Recall first the one-dimensional statement: (if and exist);one can view these as matrices. Now put and, for small, define the remainder
.
Since is differentiable, as . For small, likewise set
,
so as . Set and . Then , and
.
In particular as . The composite remainder is
.
The displayed bound and the two remainder limits make this tend to zero, which proves the stated matrix formula.
Definition 18.6 (Convex set)
A set is convex if for all and .
Theorem 18.11 (Taylor’s theorem)
Let be open and convex, let be , and let . Then , where, for some on the line segment from to , .
Proof
Put . The one-variable Taylor theorem applied at gives . Repeated chain rule and the multinomial theorem yield , giving the displayed polynomial and remainder.
Example 18.1 A second-order Taylor polynomial
For , the notes compute at : , , , , and . Thus its degree-two Taylor polynomial is .
Chapter 19Inverse and implicit functions
19.1Local invertibility
Definition 19.1 (Local inverse, homeomorphism, and diffeomorphism)
For , say is locally invertible near when some is mapped bijectively onto an open set . It is a local homeomorphism if this restriction and its inverse are continuous, a local diffeomorphism if both are differentiable, and a local diffeomorphism if both are .
Remark
is locally invertible but not a local diffeomorphism near zero. A derivative that exists and is nonzero only at a point is not enough for local injectivity: the lecture sketches an oscillating graph tangent to as the counterexample. If exists near and , however, continuity of is not needed to obtain local injectivity in one variable.
Lemma 19.1 (Quantitative invertibility of a matrix)
If is an invertible matrix, then for all , .
Proof
Put . Since , rearrange to get the bound.
Lemma 19.2 (Mean-value estimate)
If is and the segment from to is contained in , then .
Proof
Apply the one-variable mean value theorem to each coordinate of and take the largest coordinate estimate.
Lemma 19.3 (Nonsingular derivative gives a lower Lipschitz bound)
Let be and suppose is invertible. Then there are an open neighborhood of and such that for all .
Proof
Set and . Since , continuity of gives a small ball on which . Combine the preceding two lemmas with to obtain .
Theorem 19.1 (Inverse function theorem)
Let be (), with open and . If is nonsingular, then some open neighborhoods of and of satisfy: is bijective, its inverse is , and for .
Proof
The lower Lipschitz bound makes injective on a small . It also shows that is open: take a closed ball inside , minimize on it, and use the chain rule plus invertibility of the derivative to see that the minimizer for close to is interior. Thus is open and is continuous.
For and , differentiability of gives , where . The lower bound relates to , giving . Hence . Cramer’s rule expresses the inverse matrix as rational functions of the entries of ; induction then upgrades to .
Remark
is continuous. Therefore remains nonzero on a small neighborhood. The theorem says the functions can be used as local coordinates in place of .
Example 19.1 Polar and spherical coordinates
For , and , so it is locally invertible for . For spherical coordinates , the notes calculate ; it is nonzero away from and the polar axis.
19.2Implicit functions
Theorem 19.2 (Implicit differentiation)
Let be differentiable, with . If a differentiable map satisfies , then . If is invertible, then .
Proof
Apply the chain rule to . Its derivative is the block matrix , while .
Theorem 19.3 (Implicit function theorem)
Let be open, let be (), and assume , , and is nonsingular. Then on a neighborhood of there is a unique function with and . Its derivative is the implicit-differentiation formula above.
Proof
Define the auxiliary map . Its derivative is block triangular: , hence . The inverse function theorem gives a local inverse . Since the first coordinates of are the identity, write and set . This gives existence. For uniqueness, the notes let ; it is nonempty, closed by continuity, and open by the local inverse, so connectedness of a sufficiently small ball implies .
Remark
The theorem turns a level set locally into the graph of a function. In the linear case , invertibility of gives the familiar formula .
Example 19.2 Level-set examples
The unit circle defines locally away from and , exactly where is nonzero. Two surfaces in typically meet in a curve: if the derivative with respect to has rank two, the implicit theorem solves in terms of .
Chapter 20Partitions and Lebesgue’s characterization
20.1Partitions and Darboux sums
Definition 20.1 (Box, partition, mesh)
A box in is , where the are intervals; here the notes use closed intervals, , with . A partition of is a finite increasing sequence , with mesh .
A partition of a box is an -tuple of coordinate partitions. It decomposes into boxes with pairwise disjoint interiors and mesh .
Definition 20.2 (Lower and upper sums)
Let be bounded and let the subboxes of be . Set and . The lower and upper sums are and .
Definition 20.3 (Refinement)
A partition is a refinement of if for every coordinate. The common refinement of is obtained by taking the union of the coordinate partition points.
Lemma 20.1 (Monotonicity under refinement)
If refines , then and . Therefore for arbitrary partitions , .
Proof
It is enough to add one point to one coordinate partition. Each affected subbox splits into two smaller boxes, whose infima are at least the old infimum and whose volumes add to the old volume. Apply the same fact to for upper sums, and use a common refinement.
Definition 20.4 (Lower/upper integrals and Riemann integrability in R̂n)
Define and . The function is Riemann integrable if these values agree; then their common value is written .
Theorem 20.1 (Riemann condition)
A bounded is Riemann integrable iff, for every , there is a partition with .
Proof
If the lower and upper integrals agree, choose whose lower and upper sums are each within of that number, and take a common refinement. The converse follows from .
Example 20.1 A nonintegrable function
On , take when are rationally dependent and otherwise. Every subbox meets both types of points, so every lower sum is and every upper sum is . Hence is not Riemann integrable.
Lemma 20.2 (Vector-space property)
If , then . Consequently is a vector space; all constant functions belong to it.
Proof
For each subbox , and . Choose partitions making the two Darboux gaps small and take their common refinement.
20.2The review sheet
Midterm review
The four-page review re-records the earlier core facts: boundedness, total boundedness, completeness, compactness, and sequential compactness satisfy in metric spaces; in , compactness is equivalent to closed and bounded. It also restates differentiability, directional derivatives, the Jacobian, the criterion, mixed partials, multi-index notation, the chain rule, product rule, Taylor theorem, inverse function theorem, and implicit function theorem. For a linear map , the implicit solution is when is invertible.
20.3Measure zero and the Lebesgue criterion
Definition 20.5 (Measure zero)
A set has (Lebesgue) measure zero if, for every , it can be covered by countably many boxes with . It does not matter whether the covering boxes are open or closed; a countable union of measure-zero sets has measure zero.
Definition 20.6 (Oscillation)
For bounded , put and . Then is continuous at iff .
Remark
The notes ask one to verify and that implies . For the Dirichlet function ( on rationals and on irrationals), the oscillation is everywhere.
Theorem 20.2 (Lebesgue characterization of Riemann integrability)
Let be a box and be bounded. Let . Then is Riemann integrable iff has measure zero.
Proof
First suppose has measure zero. Let and cover by finitely many open boxes whose total volume is less than . For each point outside their union, continuity supplies an open box on which the oscillation is less than . Compactness of gives a finite cover. Choose a partition whose subboxes lie in a chosen member of this finite cover. The boxes inside the first family contribute at most to the Darboux gap; the rest contribute at most . Hence the gap is below .
Conversely define . If a partition has , then the subboxes of meeting in their interiors have total volume below , because each has oscillation at least . The union of the subbox boundaries has measure zero and can be covered with total volume below . Thus has measure zero. Since , so does .
Example 20.2 Two familiar discontinuity sets
The Dirichlet function on has and is not integrable. The function that is on rational points whose fraction is in lowest terms and has bounded denominator, and elsewhere, has a countable discontinuity set and is Riemann integrable.
Theorem 20.3 (Almost-everywhere zero and Fubini)
If is Riemann integrable and almost everywhere, then . If and , then almost everywhere. For boxes , , an integrable satisfies Fubini’s theorem: .
Remark
Under Fubini’s hypotheses, the inner integral exists almost everywhere; if it exists for every , the iterated integral is defined everywhere. The review example with a vertical rational/irrational slice shows why almost everywhere’ ‘ is necessary.
Chapter 21Integration and change of variables
21.1Fubini’s theorem
Theorem 21.1 (Fubini for bounded Riemann integrable functions)
Let and be boxes, and let be bounded and Riemann integrable. For , put
Then and are Riemann integrable on and
Consequently, is integrable and
Proof
Let and be partitions of and , and let . If is a subbox of and , then
Taking the infimum in and then multiplying by the volume of gives
On summing over the boxes of and ,
The same argument with suprema gives
Refine the product partitions so that tends to zero. The displayed inequalities force the lower and upper integrals of both sectional functions to agree, and their common integrals equal .
21.2Integrals over bounded sets
Definition 21.1 (Zero extension and integral over a bounded set)
Let be bounded and let be a box containing . For a bounded function , define its zero extension to by
If is Riemann integrable on , define
Lemma 21.1 (Independence of the containing box)
If and are boxes containing and the zero extension is integrable on one of them, then it is integrable on the other, with the same integral.
Proof
Enclose in a larger box . The two extensions to differ only by functions which vanish off a set on which they already agree; partition along the faces of and . Additivity for the resulting subboxes shows that the new pieces outside the original containing box contribute . Thus both definitions are the same integral over .
Proposition 21.1 (Elementary properties)
Whenever the displayed integrals exist, the integral over a bounded set is linear, monotone, and satisfies
It is also additive under a finite disjoint decomposition of . More generally, for two bounded Jordan-measurable sets,
In particular, if have pairwise intersections of Jordan measure zero, then .
Theorem 21.2 (Jordan-measurable sets)
A bounded set is Jordan measurable if and only if its boundary has measure zero:
In that event the constant function is integrable over , and
Remark
The lecture uses for the class of Jordan-measurable bounded sets and for compact Jordan-measurable sets. Thus integrals over arbitrary bounded sets are not silently assumed to exist: the zero extension must first be Riemann integrable.
21.3Extended integrals on open sets
Definition 21.2 (Positive extended integral)
Let be open and let be continuous with . Its extended integral is
This value is allowed to be .
Definition 21.3 (Signed extended integral)
For a continuous , set
If both and are finite, define
Lemma 21.2 (Compact exhaustion)
Every open set has compact Jordan-measurable sets such that
Proof
Take compact sets increasing to , for example by requiring a positive distance from and a bound on the norm. Cover each by finitely many closed cubes whose interiors lie in , and let be the finite union of the cubes selected up to stage . Enlarging at each stage if necessary gives .
Theorem 21.3 (Exhaustion criterion)
For continuous on an open set and for any compact exhaustion as above,
In that case,
Proof
The integrals of over are increasing. If they are bounded, the positive and negative parts have finite suprema, so the signed extended integral exists and the asserted limit follows by subtracting the two monotone limits. Conversely, if the positive and negative extended integrals are finite, each is bounded by their sum.
The point that the exhaustion computes the supremum is that every compact is contained in some : the open sets cover , so a finite subcover has a largest index. Hence for some , and taking suprema gives the claim.
Theorem 21.4 (Agreement on bounded open sets)
If is bounded and open and is bounded and continuous on , then the extended integral exists. If the zero extension makes the ordinary Riemann integral over meaningful, it agrees with the extended integral:
Remark
对于 bounded open ,lecture notes 记录:extended integral must exist;如果 ordinary integral 存在,则二者相等。
21.4Change of variables
Theorem 21.5 (One-dimensional change of variables)
Let be , and let be continuous on an interval containing . Then
Proof
Choose an antiderivative of . The chain rule and the fundamental theorem of calculus give
Theorem 21.6 (Change-of-variables theorem)
Let be open, let be a diffeomorphism, and let be continuous. Then
and, whenever either condition holds,
Example 21.1 Polar coordinates
On the annular region
use on . Since
the omitted radial cut has measure zero and
Example 21.2 Spherical coordinates
With
one has . Thus, subject to the usual bounds on , , and describing the region,
21.5Diffeomorphisms and null sets
Theorem 21.7 ( maps preserve sets of measure zero)
If is on an open set and has measure zero, then has measure zero. In particular, if , the image of every bounded set under a map has measure zero.
Proof
First restrict to a closed cube on which . By the mean-value estimate, a cube of side length in has image contained in a cube of side length at most . Cover by cubes of total volume as small as desired; the corresponding image cubes have total volume at most times that quantity. Hence has measure zero. Exhaust by such closed cubes and take a countable union.
Proposition 21.2 (Diffeomorphisms preserve interior and boundary)
If is a diffeomorphism of open sets and , then
Hence is Jordan measurable if and only if is Jordan measurable.
Remark
The notes contrast this with the rationals in an interval: they have Lebesgue measure zero but are not Jordan measurable, because their boundary is the whole interval.
21.6Primitive diffeomorphisms
Definition 21.4 (Primitive diffeomorphism)
A primitive diffeomorphism changes only one coordinate. For some ,
Theorem 21.8 (Local decomposition)
Every local diffeomorphism can, after restricting to sufficiently small neighborhoods, be written as a finite composition of primitive diffeomorphisms.
Proof
The proof in the notes has three reductions. First, an invertible linear map is a product of elementary matrices: coordinate swaps, scalings, and additions of one coordinate to another. Each is primitive (a coordinate swap is factored into elementary operations when necessary). Translations are also primitive.
Next assume and . Define
Near , is a diffeomorphism. The map fixes the first coordinates, so is a product of primitive maps. Finally, translate the chosen point to and compose with to reduce the general case to this one.
21.7Partitions of unity
Definition 21.5 (A smooth bump on a box)
Let
Then is , positive on , and zero on . The product
is , positive on the interior of the closed box , and zero outside that interior.
Definition 21.6 (Support and partition of unity)
The support of a function is
A partition of unity on an open set , subordinate to an open cover , is a locally finite family of functions such that
Theorem 21.9 (Smooth partition of unity)
Every open cover of an open subset admits a locally finite smooth partition of unity subordinate to that cover. Each may be chosen with compact support contained in one member of the cover.
Proof
Choose a locally finite collection of closed cubes whose interiors cover , with each contained in a member of the given cover. The compact-exhaustion construction supplies such cubes by covering successive compact annuli with finitely many cubes. Let be the smooth box bump positive on and supported in its containing cover member. Local finiteness makes
a smooth, positive function. Then
has the required support, local finiteness, and sum.
Theorem 21.10 (Integration by a partition of unity)
Let be continuous on an open set , and let be a smooth partition of unity with compact supports in . Then
and in that case this series equals .
Proof
For , finite partial sums satisfy . Their integrals increase to the extended integral by the compact support of each summand and local finiteness. Apply this statement separately to and to obtain the signed assertion.
Remark
Integration by POU assembles local integrals into a global one: “POU 的作用是把局部的积分拼成全局积分。”
Chapter 22IBL: Baire category through Jordan measure
22.11A:证明 metric space 是 topological space
Definition 22.1 (Open balls and the metric topology)
In a metric space , is open when every has an with . A sequence converges in this topology exactly when it satisfies the metric epsilon definition.
The source’s 1A problem has five parts: prove that these open sets form a topology; compare topological and metric convergence; prove open balls open; prove closed balls closed; and give the discrete-metric counterexample above.
Begin with and remove the middle third at each stage. The set is the middle-thirds Cantor set; is a union of closed intervals, each of length .
The migrated problems ask to show that is nonempty and compact, every point is a limit point, is uncountable by Baire category, and contains no interval. Its stage- total length is , motivating a notion of measure beyond intervals.
来源中文批注
Cantor set 是一个 compact 且 closed 的集合(甚至 perfect);由于它在 中,它还是 complete metric space。它 uncountable,却不包含任何 open interval;来源同时标注其 Lebesgue measure 为 。
Definition 22.3 (Vitali-type obstruction)
On define when . Choose one representative from each equivalence class, forming . For , let be the translate of by , taken modulo one.
Theorem 22.2 (No translation-invariant countably additive measure on every subset)
The sets are disjoint and their union is . If a function on all subsets were countably additive, invariant under rigid motions, and normalized by , then all would have a common measure. It would be either zero or positive, forcing the union’s measure to be either zero or infinity — a contradiction.
The source explicitly notes that merely replacing countable additivity by finite additivity does not solve this problem: Banach–Tarski supplies a finite-piece obstruction in three dimensions. The conclusion is to measure a proper family of subsets rather than every subset of .
22.4Elementary and pixel measure
Definition 22.4 (Boxes, elementary sets, and elementary measure)
An interval is any of , , , or , with length . A box is a Cartesian product of intervals; its volume is the product of their lengths. An elementary set is a finite union of boxes. After writing it as a finite disjoint union , define .
The source’s problem sequence establishes closure of elementary sets under union, intersection, difference, symmetric difference, and translation; it then asks for a disjoint-box decomposition and well-definedness of . A lattice-counting route is recorded: scale the number of lattice points in by and pass to the limit.
It then asks for finite additivity on disjoint elementary sets, monotonicity, and finite subadditivity for arbitrary finite collections. The pixel-measure exercise is deliberately retained as a counterexample prompt, since the source does not supply a completed personal answer.
Theorem 22.3 (Elementary-measure properties)
For elementary sets, elementary measure is finitely additive on disjoint unions, monotone, and finitely subadditive.
Remark
The exploratory “pixel measure” The pixel measure is the limit of times the number of lattice points in . It is not translation invariant whenever both displayed limits exist; the source asks for an explicit example.
22.5Jordan measure and Riemann integrability
Definition 22.5 (Jordan measure inner and outer measure)
For bounded , and . The set is Jordan measurable when these agree.
Theorem 22.4 (Jordan measurability criteria)
A bounded set is Jordan measurable exactly when it can be sandwiched between elementary sets with arbitrarily small; equivalently, it can be approximated in Jordan outer measure by an elementary set. Its boundary has Jordan outer measure zero exactly when it is Jordan measurable.
The retained problem set establishes that elementary sets are Jordan measurable, then asks for closure under union, intersection, difference, and symmetric difference, as well as finite additivity, monotonicity, finite subadditivity, and translation invariance. It asks to prove that the graph of a continuous function on a closed box has Jordan measure zero and that the region below such a graph is Jordan measurable.
The next chapter asks to prove that open and closed balls are Jordan measurable with measure , to bound , and to compare a bounded set with its closure and interior. It gives the boundary criterion above. Finally it defines lower and upper Darboux integrals through a partition and asks to show that a bounded nonnegative is Riemann integrable exactly when its subgraph is Jordan measurable.
Chapter 23IBL: Lebesgue outer measure
Definition 23.1 (Lebesgue outer measure)
For , , where the cover is by boxes. This replaces the finite covers in Jordan outer measure by countable covers.
The IBL problems record that a Jordan-measurable set can be outer-approximated by an elementary set; ; the defining covers may be restricted to open or closed boxes; and every countable set has outer measure zero. The proof sketch preserves the source’s allocation for the countable cover.
来源中文批注
Jordan measurable 的意义是可以 outer-approximate by elementary set。关于 closed/open boxes 的定义限制,来源指出 boundary 的 Jordan measure 为 ,可以在每个 open box 内用误差小于 的 closed boxes 覆盖;dually 亦然。对于 countable set,只需对每个点给出体积小于对应 的 box 覆盖,因此其 Lebesgue outer measure 总是 ;这也说明 Lebesgue measure 比 Jordan measure 更好。
Theorem 23.3 (dist0 的集合外测度 union additive;ctbl 个 almost disjoint boxes)
If , then . If is a countable union of almost-disjoint boxes , then .
关于从 finite 到 countable 的过渡,来源的中文提示为:「extend finite to countable by continuing the seq using empty sets 即可得到。」
23.1Lebesgue measure 的大小处于 Jordan outer/inner measure 之间
来源把这一节保留为由 elementary measure 与 outer measure 比较得出的结论,并单独要求构造 non-Jordan-measurable 的 bounded open set,以及证明 ctbl 个 almost disjoint boxes 的 outer-measure union additivity。
Example 23.1 example:non J-measurable 的 open set
Enumerate the rationals in and cover the th rational by an open interval whose lengths form a summable sequence. The union can have arbitrarily small outer measure but dense complement structure that prevents Jordan measurability, exactly as posed in the source.
来源的 personal solution 只写到「我们首先 list 出 之间的 ratioals,称为 。我们对于每个……」便中断;这里保留其不完整状态,而不把后续构造误标为来源解答。
Chapter 24IBL: regularity, measurability, and additivity
Theorem 24.1 ( 中任意开集都是一个 ctbl union of almost disjoint boxes)
Every open subset of is a countable union of almost-disjoint boxes; the source gives a dyadic construction selecting boxes not already chosen at earlier scales.
The source’s construction starts with unit grid boxes contained in an open set, then repeats at dyadic scales after removing boxes selected earlier. It asks to verify that their union is the original open set and that interiors do not overlap.
来源的中文批注说,这与在 上用 ctbl closed intervals 逼近任意 open interval 如出一辙;随后给出的 process 只是更 generalized 的算法。
Theorem 24.2 (Outer regularity)
For every , .
Remark
The reverse inner approximation by open subsets is false in general; the source asks for a counterexample and foreshadows compact inner regularity.
Every set of outer measure zero is Lebesgue measurable.
Theorem 24.4 (Countable unions, complements, and intersections)
A countable union of Lebesgue measurable sets is measurable. Complements of measurable sets are measurable, and hence countable intersections are measurable.
来源的中文证明提示为: 也是 ctbl 的。对每个 都选取一个 open cover,最后的 double union 仍是 countable open cover;取任意 ,再用 bound 每个 与其 cover 的差距即可。
Theorem 24.5 (Closed sets are measurable)
Every closed subset of is Lebesgue measurable. The recorded approach reduces to compact pieces and uses the almost-disjoint-box decomposition of an open complement.
The recorded proof plan writes an unbounded closed set as a countable union of closed bounded pieces, reduces to compact sets, and decomposes their open complements into almost-disjoint closed cubes.
24.3Approximation and regularity
Definition 24.1 (Symmetric difference)
. The source notes .
来源中文批注
sym diff 越加入更多 set 越大;这正是 的直观来源。
Theorem 24.6 (Approximation by open sets)
is measurable if and only if for every there is an open with .
For the difficult direction, the IBL notes choose open with errors , take their union, then cover the remaining null set by an open set of small outer measure. This preserves the original proof strategy without inventing its omitted final estimates.
If is measurable, then as ranges over compact subsets of . The source also records the equivalent approximation by closed sets in symmetric difference.
Theorem 24.8 (Countable additivity)
For pairwise disjoint Lebesgue measurable sets , .
The explicit exercise is retained: with , one must not interchange an infimum and an infinite sum without a valid argument.
Chapter 25IBL: limits and Carathéodory’s criterion
Every Jordan measurable subset of is Lebesgue measurable. The source points to the zero-boundary characterization as the route to the proof.
Theorem 25.2 (Continuity from below)
For measurable , .
The source suggests taking the disjoint increments and applying countable additivity.
Theorem 25.3 (Continuity from above)
For measurable , if some has finite measure, then .
Remark
The finite-measure assumption in continuity from above is necessary; the source asks for a counterexample when it is dropped.
Theorem 25.4 (Finite-measure approximation by elementary sets)
A finite-measure set is measurable exactly when it differs from an elementary set by a set of arbitrarily small outer measure.
Theorem 25.5 (Caratheodory’s criterion — elementary test sets)
A set is measurable if and only if for every elementary set , .
The source remarks that some texts use this elementary-test identity as the definition of measurability. Its final linear-map problem asks for the precise Jacobian factor , including singular linear maps.
Theorem 25.6 (Linear change of measure)
If is measurable and is linear, then is measurable and .
Personal Multivariate Analysis Homeworks
Chapter 26HW 1
26.1Problem A
Suppose is a metric space. For , show that is a metric on . If has its usual metric, show that has “infinite length” using .
Proof. Take . Positivity and symmetry are immediate: , with equality exactly when , and . Let for . Then and , so is increasing and concave. Thus
.
Hence , completing the metric axioms.
For the length claim, take an equally spaced partition into subintervals. Then and
.
Since , these sums are unbounded above, so for every some partition has sum greater than .
26.2Bonus problem
If is , is , ordinary multiplication takes scalar multiplications. For
,
find the cheapest parenthesization. The submitted parenthesization is
.
Let be the minimal cost for multiplying the matrix chain from through . The recursion used was
.
The dynamic-programming calculations recorded on the page are
Thus the final answer costs scalar multiplications.
Chapter 27HW 2
27.1Problem A
If is a norm on a vector space , then is a metric. For , positivity gives , with equality iff ; homogeneity gives ; and
Thus a norm induces a metric.
27.2Problem B
For a linear , the operator norm is . If , then , so proves continuity. Conversely, continuity at gives such that for . Applying this to with gives . Hence is bounded.
27.3Problem C
An unbounded linear map is the derivative , with the sup norm on the domain. For , , while . Therefore the ratios are at least .
27.4Problem D
Take . Every is diagonalizable with eigenvalues . For , , so although the eigenvalues are bounded.
27.5Problem E
If is totally bounded, for every choose a finite -cover with centres . The union of the centres is countable and dense: every either occurs among them or is the limit of selected centres at distance . Hence is separable.
27.6Problem F
Let be countably many copies of with their left endpoints glued. Write points as and use if , and if . It is bounded. At radius , a ball can cover at most one of the points from distinct far ends, since two such points have distance . Thus infinitely many balls are needed and is not totally bounded.
27.7Problem G
For with the sup metric, total boundedness is equivalent to boundedness plus: for every , some has for every and . A finite cover proves the tail condition by contradiction (choose increasingly far non-small entries and form a separated subsequence). Conversely, partition the first bounded coordinates into finitely many pieces of length and combine this finite head cover with the tail bound.
27.8Bonus problem
For a countable dense set in , define . Triangle inequality makes this bounded and gives . Along a subsequence , the coordinate differences tend to , so equality holds. This is an isometric embedding into .
Chapter 28HW 3
28.1Problem A
For a Lipschitz map with constant , . Taking proves uniform continuity. If have one common Lipschitz constant and converge uniformly to , then . Letting the uniform error tend to zero proves that is also Lipschitz with constant . Without a common constant this is false: on , converge uniformly to , which is not Lipschitz near .
28.2Problem B
If is connected and is continuous, then is connected: a separation pulls back to a separation of . Consequently a continuous assumes every intermediate value between and .
28.3Problem C
For a continuous bijection with compact, is continuous. A closed is compact, hence is compact and closed in the metric space . Thus is a closed map. Compactness is necessary: , , is a continuous bijection whose inverse is discontinuous at .
28.4Problem D
If exists, then : for substitute in the defining limit, and is immediate. For at , the derivatives in and are , but that in does not exist because has unequal one-sided limits. For off the origin and at it, when , and otherwise. This formula is not linear in the direction, though polar coordinates show continuity at the origin.
28.5Problem E
The Baire Category Theorem was written as: in a complete metric space, every countable intersection of open dense subsets is dense.
28.6Problem F
Let select one element from each class modulo . The translations form a disjoint decomposition of . A countably additive, translation-invariant measure on every subset with would make all have the same measure; this gives either or infinity for the interval. Therefore the stipulated measure does not exist.
28.7Bonus problem
The Cantor set is uniformly disconnected by its middle-third gaps. The recorded equivalent ultrametric is the infimum of for which an -chain joins to . Concatenating chains yields the ultrametric inequality. Conversely, if and is ultrametric, the chain would force , impossible for distinct points.
Chapter 29HW 4
29.1Problem A
Let satisfy for every and suppose is differentiable at . Put . Homogeneity gives . If , then for every , contradicting differentiability as . Hence , so is linear.
29.2Problem B
For , if all partial derivatives exist and are bounded on the open set , then is continuous. Write and pass from to one coordinate at a time: , . Applying the one-variable mean value theorem to gives . Summing coordinate and target components yields .
29.3Problem C
For , and . On , is the quarter-annulus , . The inverse is , continuous on this set. Its derivative is and .
29.4Problem D
Take away from and . Every directional derivative at is , yet along the quotient of by does not tend to , so is not differentiable at the origin.
29.5Problem E
For and off , the first partials at are . Off , product and quotient rules give
Both tend to at the origin (each term is bounded by a multiple of or ), so . The mixed partials are equal off , while at direct difference quotients give .
29.6Bonus problem
In an ultrametric space, is closed: if lies outside it and , then would otherwise contradict . Intersecting balls are nested: if and belongs to both and , then satisfies , hence . Thus every point of a ball is a centre.
For a connected weighted graph, define as the least possible largest edge-weight along a path. Concatenating a best - path and a best - path yields . Conversely, from a finite ultrametric space, join every pair with an edge weighted by its distance; the least maximum path weight is the original metric.
Chapter 30HW 5
30.1Problem A
For ,
,
each component is a composition or product of smooth elementary functions, hence is differentiable. Factor with and . The displayed has third row equal to half the difference of the second and first rows, so . Chain rule gives .
30.2Problem B
If differentiable maps and are inverse, then and . Both products being identities forces and when .
30.3Problem C
is a differentiable homeomorphism of , but is not differentiable at .
30.4Problem D
If is continuous at and the iterated limits exist, each equals . For example, define away from and there. Then while the reversed order is .
30.5Problem E
The number of four-variable monomials of degree at most is (the red working also sums ).
30.6Problem F
If is open and connected, is differentiable, and on , then is locally constant: join nearby by coordinate segments inside a small ball and use the one-variable mean value theorem on each segment. The set is both open and closed in , hence is all of .
30.7Problem G
Leibniz’s formula was proved by induction:
The displayed Leibniz formula differentiates the product of through : , with the same multi-index distribution among all factors.
Differentiating the case and grouping every new multi-index with gives coefficient .
30.8Problem H
Let be the degree- Taylor polynomial centered at . For the backward direction, Taylor’s theorem writes as a remainder whose terms have ; bounding each monomial by gives the required little- statement.
For the forward direction, the submitted work writes and seeks to show the quotient by does not tend to zero. In Case 1, , it chooses with and obtains a nonzero constant quotient. Case 2, , ends with “idk”. A subsequent attempted route states that a nonzero homogeneous polynomial of degree is not , using ; it then notes that a degree- polynomial need not be homogeneous.
30.9Problem I
For , chain rule gives and , hence . For the displayed composition problem, write and apply the chain rule. At with , , , both the formula and direct computation give
30.10Problems J and K
For the specified and , the chain-rule calculation records . The third order Taylor polynomial of at is
.
30.11Positive definite matrices
For a real symmetric matrix , positive definiteness implies invertibility and , so the angle of with is acute. Conversely, the acute-angle condition gives . In an orthonormal eigenbasis, , proving positive definiteness iff every eigenvalue is positive. Each leading principal minor inherits positive definiteness; the forward direction of Sylvester’s criterion follows. The submitted converse attempt is marked “didn’t work at all.”
Chapter 31HW 7
31.1Problem A
For differentiable and unit , . Cauchy-Schwarz gives , with equality precisely for . Also iff is orthogonal to .
31.2Problem B
Let , , be , have a local minimum at , and be surjective. Reorder coordinates so . By IFT, locally . For , . Differentiating gives , hence for every . Put ; then .
31.3Problems C-D
The intuitive explanation says that at a constrained minimum the gradient of is normal to all allowed directions, while is normal to , so the two gradients are parallel. For on , . The critical points are and its negative; the minimum is at .
31.4Problem E
For with full rank , the stated generalization is . Split variables as with a nonsingular block. IFT writes locally as . The identities and combine to give .
31.5Problem F
Positive definite symmetric matrices form an open subset of symmetric matrices. For , the quadratic form has positive minimum on the compact unit sphere. If , then on the sphere, and hence for all nonzero .
31.6Problem G
If , is critical, and is positive definite, continuity of the Hessian makes positive definite near . Taylor’s formula along the segment gives for nearby ; hence a strict local minimum.
31.7Problem H
For an invertible matrix with cofactor matrix , the diagonal entry equals when by cofactor expansion. For , replace row by row to obtain a matrix with determinant whose cofactor expansion is . Thus , so .
31.8Problem I
For differentiable , , and differentiating term by term gives .
31.9Bonus
The epigraph of is convex iff is positive semidefinite everywhere. For the forward direction, restrict to ; convexity gives its second derivative . For the converse, the same one-variable restriction has nonnegative second derivative, hence is convex, and this is exactly the epigraph inequality.
Chapter 32HW 8
32.1Problem A
Let be , , and
The minors are , , and . Thus can be solved in terms of near , and can be solved in terms of ; the IFT gives no conclusion for solving in terms of .
32.2Problem B
If satisfies and , differentiating gives . Hence
Chapter 33HW 9
33.1Problem A
Let . The continuity set is the intersection of all . If is continuous at , choose a ball mapping into , and the triangle inequality gives . Conversely, choose with and a ball supplied by ; then is continuous at . Each is open: a witnessing ball at contains a smaller ball about every one of its points.
33.2Problem B
A bounded non-decreasing is Riemann integrable. For a rational between and , let . Every discontinuity belongs to some by density of . Each has at most one point, since in it would force values left/right incompatible with monotonicity. So the discontinuity set is countable and has measure zero.
33.3Problem C
For integrable , is bounded. It is continuous at whenever both factors are continuous at the corresponding coordinates; hence . The product covers of measure-zero sets show has measure zero, so is integrable.
33.4Problem D
Define if in lowest terms and on irrationals. Given , choose with , let be rationals with denominator at most , and make a partition containing with mesh . On subintervals missing , the supremum is at most ; the other intervals have total length . Thus . It is continuous at every irrational because its values along rationals with unbounded denominators tend to ; discontinuities are contained in the countable rationals.
33.5Problem E
If bounded vanishes off a closed measure-zero , cover by finitely many boxes of total volume and choose a partition having these boxes as subboxes. On the remaining subboxes , so the difference of upper and lower sums is . Hence is integrable.
33.6Problem F
For a countable closed-box cover , first enlarge to open boxes with volume increase . Compactness gives a finite subcover. Successively subtract earlier boxes to make a disjoint measurable cover; additivity and monotonicity give , and then let .
33.7Problem G
For , , , define . Since , IFT gives a local inverse with the identity. Then is and .
33.8Bonus
For an open box , choose smooth one-variable functions on and zero outside; is smooth, positive on , and zero outside. For an open , use a countable ball cover and a locally finite smooth partition of unity subordinate to it; is smooth, positive exactly on . For Cantor , apply this to the complement of in . Taking on and off , the graphs of and meet exactly at .
Chapter 34HW 10
34.1Problem A
For integrable , is integrable. At every point where both and are continuous, the maximum is continuous (use the two local bounds). Thus , which has measure zero.
34.2Problem B
If is integrable then is integrable: , since a fixed jump in gives, by reverse triangle inequality, a jump in . For every partition , , and taking infima yields the corresponding inequality between the integrals of and .
34.3Problem C
Let and let be the rotation of the rational points in the unit square. It is dense because is a rotation. Two points of on one vertical (or horizontal) line must have equal preimages, since the relevant sine/cosine coefficient is irrational; hence each such line meets at most once. The characteristic function of is except possibly at one point on each coordinate line, so every one-variable slice is integrable; but density gives upper sum and lower sum for every two-dimensional partition.
34.4Problem D
For and closed box , Fubini and FTC give both integrals of the mixed partials as . On a small box about , apply the integral mean-value theorem twice to their difference; the zero double integral forces equality of mixed partials at .
34.5Problem E
Riemann integrability implies Darboux integrability because a fine partition has both tagged sums within of the integral, so upper and lower sums are within . Conversely, for a Darboux integrable , refine a near-optimal partition by any sufficiently fine partition. The boundary-strip lemma bounds total volume of new subboxes crossing old boundaries; lower and upper sums on the remaining subboxes stay close to the Darboux sums. Therefore every fine tagged sum is close to the common Darboux integral.
34.6Problems F-G and Bonus
For , chain rule gives and differentiating once more yields . The quadratic Taylor polynomial is . The bonus proof uses that the continuity set of a map is a set and Baire Category: is not , so no function can be continuous exactly on and discontinuous on its complement.
Chapter 35HW 12
35.1Problem A
Let be bounded, let be the interior of , and let bounded be Riemann integrable on . Since , Lebesgue’s criterion makes integrable on . Also . Split the complement of in into its isolated points, its non-isolated discontinuities, and its non-isolated continuity points. The first is countable; the second has measure zero; on the third, has limiting value and the integral over the set is zero. Hence the integral over the complement is , so the integrals over and agree. If is Jordan measurable, then , and .
35.2Problem B
For , polar coordinates give its volume as . The spherical-coordinate Jacobian recorded is ; integration produces the factor . Translation has determinant one, giving the formula for all centres. and . Slicing the unit -ball by one coordinate and using polar coordinates gives , hence and .
35.3Problem C
For with and open Jordan measurable , define by . It is a diffeomorphism. Its derivative is upper triangular with determinant , so change of variables gives the volume of as times the volume of divided by .
35.4Problem D
The ellipsoid is the inverse image of the unit ball under . The inverse has determinant , so its volume is .
35.5Problem E
The solid between and projects to . Translating then using the displayed elliptical polar substitution gives the recorded volume .
35.6Problem F
Integrating over larger and larger disks, polar coordinates give the two-dimensional Gaussian integral as . Fubini over expanding squares makes this the square of the one-dimensional Gaussian integral, so the integral is .
35.7Problem G
is integrable over the unit ball iff : decompose the punctured ball into annuli and compare the radial series with . It is integrable outside the closed unit ball iff , by the analogous tail series.
35.8Bonus
For differentiable on compact with , the mean value theorem gives . If , write . Uniform continuity of lets finitely many short intervals cover so that has arbitrarily small total length. Thus and .
Chapter 36HW 13
36.1Problem A
The coordinate swap matrix factors as
each factor a primitive diffeomorphism on .
36.2Problem B
Let for and otherwise. Put for odd and for even . The supports are the listed intervals or ; at any at most four are supported. Thus is smooth and positive. Setting gives and a smooth partition of unity dominated by the open intervals of length .
36.3Problem C
For when and otherwise, on , with polynomial. Inductively : after , a bound makes the difference quotient tend to . Thus .
36.4Problem D
If is smooth and , its image has measure zero by the cited class result. If it contained nonempty open , it would contain a ball of positive Jordan and Lebesgue measure, contradicting monotonicity.
36.5Problem E
A local diffeomorphism with , is locally factored by choosing a coordinate , setting , and correcting the th coordinate in the target. IFT gives a local factorization into primitive diffeomorphisms. Induction freezes one coordinate at a time, giving a finite factorization into super-primitive diffeomorphisms; translations and elementary linear maps are also decomposed this way.
36.6Problem F
No injective smooth exists. If all partials vanished everywhere, would be constant. Otherwise, say ; IFT writes the level set locally as , contradicting injectivity.
36.7Problem G
If is smooth at each , choose local smooth extensions . A locally finite smooth partition of unity subordinate to gives on and elsewhere. The locally finite sum is smooth and, at , equals .
36.8Problem H
If matrix has rank , select independent columns and then independent rows among them to obtain a minor with nonzero determinant. Any larger minor has rank at most , so determinant zero. Hence rank is the maximum order of a nonzero minor.
Chapter 37HW 14
37.1Problem A
For , , , and , let for and otherwise. Define
The are smooth with disjoint supports . If , then at the midpoint of , while the midpoints tend to and . Thus is not continuous at .
37.2Problem B
The change-of-variables theorem for linear diffeomorphisms and compactly supported continuous is proved by induction on the dimension, after decomposing a linear map into primitive linear diffeomorphisms. The case is the one-variable substitution theorem. For a primitive map preserving the last coordinate, write , restrict to , extend by , and use Fubini. For each fixed , the -dimensional induction hypothesis supplies the inner substitution formula, which Fubini integrates to the result.
37.3Problem C
The rank map on is lower semicontinuous. If matrix has rank , choose a nonzero minor. Continuity of determinant supplies a Frobenius-norm ball about in which the same minor remains nonzero, so ranks are at least . It need not be continuous: , but has rank while zero has rank .
Chapter 38HW 15
38.1Problem A
There is no injective smooth for . The proof first records the constant rank theorem: if has constant rank near , choose a nonsingular minor and set . IFT makes a local diffeomorphism. In these coordinates ; the rank calculation makes the partial derivatives of the terms in the last variables zero. A target coordinate change then gives .
For the claimed non-injectivity, lower semicontinuity and the finite set of possible ranks make rank locally constant on some neighbourhood. The normal form is not injective when , and composing with local diffeomorphisms preserves this contradiction.
38.2Problem B
For continuous compactly supported , define convolution by equal to the integral of over . On a product box containing both supports, Fubini and the substitution give
the integral of equals the product of the integrals of and .
The same substitution proves . Applying Fubini twice shows
The iterated-integral calculation has integrand and yields ,
so convolution is associative.
38.3Problem C
For on , the only interior critical point is , where . On the boundary, , so the maximum is at and . Thus the minimum is at .
38.4Problem D
Of , , and , only is a tensor: the first has a quadratic factor in , and the second has a constant term. In the elementary dual basis,
38.5Problem E
For a vector space , is a vector space under pointwise addition and scalar multiplication: the displayed verification checks linearity in each argument, the zero map, additive inverses, commutativity, associativity, and distributivity.
38.6Problem F
For the cycle taking to through and back to , write it as transpositions, so its sign is .
38.7Problem G
If is linear and , then is multilinear. For a permutation , substituting the permuted arguments gives equal to the sign of times , so .
38.8Problem H
For the elementary alternating tensor on , with and column matrix ,
is the sum over permutations of the sign of times the corresponding product of the selected coordinates, and equals ,
the determinant expansion of the submatrix whose rows are indexed by .
38.9Bonus
The printed bonus gives the definition of a real analytic function, a binomial-series exercise, radius of convergence , convergence properties, coefficient bounds, and differentiation of a power series. The source page contains no handwritten solution for these printed bonus parts.