MAT 5610, Analysis I
-
Upload
khangminh22 -
Category
Documents
-
view
0 -
download
0
Transcript of MAT 5610, Analysis I
MAT 5610: 0
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
MAT 5610, Analysis I
Wm C Bauldry
Spring, 2011
0 Today
MAT 5610: 1
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Analysis I
MAT 5610. Analysis I/(3).S. A rigorous treatment of topology of thereal numbers, continuity, differentiation, and sequences and series offunctions. Prerequisite: MAT 3220 (Introduction to Real Analysis I) orpermission of the instructor.
Our goal is a rigorous development of differential and integral calculusthrough sequences and series of functions. Chapters 5→ 8 of ourtext, Witold Kosmala’s A Friendly Introduction to Analysis, 2nd ed.Optional Supplement: David Bressoud’s A Radical Approach to RealAnalysis, 2nd ed.
Grading:
Projects / Presentations ≈ 100 pt.Homework & Proofs ≈ 100 pt.Midterm Exam ≈ 100 pt.Final Exam ≈ 100 pt.Total ≈ 400 pt.
MAT 5610: 2
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Semester Projects
• Individual ProjectGlossary: Build a glossary of the terms we use in analysis.
Start with basic items such as ‘open set’.
• Class ProjectsBibliography: Generate an annotated list of references for
• real analysis and advanced calculus,• calculus and teaching calculus. (sample)
Concept Map: Create a concept map of analysis.Look at the Derivative Map for a sample. There isfree software at the Institute for Human andMachine Cognition (IHMC) site.
MAT 5610: 3
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Reprise
DefinitionsOpen ball: B(a, ρ) = { |x− a| < ρ} (also called neighborhood)
Punctured ball: B′(a, ρ) = {0 < |x− a| < ρ} (deleted neighborhood)Open set: O is open⇐⇒ (∀x ∈ O) (∃δx > 0) [B(x, δx) ⊆ O]
Accum. point: a is an acc. pt. of S ⇐⇒ (∀ρ > 0) [B′(a, ρ) ∩ S 6= ∅ ]
Closed set: F is closed⇐⇒ F ⊇ {all accumulation pts of F}
Definition (Limit)Let φ :D → R and a be an accumulation point of D. Then the limit ofφ as x goes to a is L iff for every ε > 0 there is a δ > 0 such thatwhenever x ∈ B′(a, δ) ∩D, then f(x) ∈ B(L, ε). In symbols:
limx→a
φ(x) = L⇐⇒ (∀ε > 0) (∃δ > 0)[φ (B′(a, δ) ∩D) ⊆ B(L, ε)
]
MAT 5610: 4
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Re2prise
Definition (Continuity)Let f :D → R and a be a point of D. Then f is continuous at a iff forevery ε > 0 there is a δa > 0 such that whenever x ∈ B(a, δa) ∩D,then f(x) ∈ B(f(a), ε). In symbols:
f is cont at a⇐⇒ (∀ε > 0) (∃δa > 0)[f (B(a, δa) ∩D) ⊆ B(f(a), ε)
]
Definition (Uniform Continuity)Let f :D → R and E ⊆ D. Then f is uniformly cont. on E iff for everyε > 0 there is a δ > 0 such that whenever x1, x2 ∈ D with|x1 − x2| < δ, then |f(x1)− f(x2)| < ε. In symbols:
f is u.c. on E ⇔ (∀ε > 0) (∃δ > 0)[|x1 − x2|<δ ⇒ |f(x1)− f(x2)|<ε
]
The Four Types of Discontinuity:Simple: removable or jump. Essential: infinite or oscillating.
MAT 5610: 5
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Borzellino’s Example
f(x) =�x4 − x2
�3/2
What is limx→0
f(x)? Is f continuous at a = 0? 1
1Preview of a coming attraction: Is f differentiable at a = 0?
MAT 5610: 6
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Borzellino’s Second Example
f(x) =�
x sin(1/x)
What is limx→0
f(x)? Is f continuous at a = 0? 1
1Preview of a coming attraction: Is f differentiable at a = 0?
MAT 5610: 7
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Re3prise
PropositionsLet I be a closed, bounded interval. Let f be cont on I.• Then f is bounded on I.• Then f attains maximum and minimum values on I.• Then f has Bolzano’s Intermediate Value Property on I.• If f is nonconstant, then the range of f is also an interval.
Propositions• If f is cont on a closed, bounded set, then f is uniformly cont.• If f is unif cont on (a, b), then f(a+) and f(b−) are finite.• If f : [a,∞)→ R is cont and f(∞) is finite, then f is unif cont.
MAT 5610: 8
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Re4prise
Definition (Open Cover)An open cover of a set E is a family of open sets A = {Aδ | δ ∈ ∆} s.t.
• each Aδ is an open set, and
• E ⊆⋃δ∈∆ Aδ.
Definition (Compact)A set K is compact iff every open cover of K has a finite subcover; that is,K ⊆
⋃nj=1 Aδj whenever K ⊆
⋃δ∈∆ Aδ.
Theorem (Heine-Borel Theorem)A set K ⊆ R is compact iff K is closed and bounded.
PropositionIf f is continuous on a compact set K, then it is uniformly continuous on K.
MAT 5610: 9
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Re5prise
TheoremThe continuous image of a compact set is compact.
Proof.Suppose that f is continuous on the compact set K and that f(K)has an open cover O = {Oδ | δ ∈ ∆}. Whence f(K) ⊆ ⋃δ∈∆Oδ.
• Thus K ⊆ f−1 (f(K)) ⊆ f−1(⋃
δ∈∆Oδ)
=⋃δ∈∆ f−1(Oδ).
(i) Since f is continuous, each f−1(Oδ) is open, and(ii) since K is compact, there is a finite subcover s.t.K ⊆ ⋃nj=1 f
−1(Oδj ).
• Thence f(K) ⊆ f(⋃n
j=1 f−1(Oδj )
)⊆ ⋃nj=1Oδj ; i.e., f(K) is
contained in a finite subcover of O.
MAT 5610: 10
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Derivative
The “differential triangle” from Lectiones Opticæ et Geometricæ,Isaac Barrow, 1669. (Lect X, App, p 80(291). Fig. 115 to 121, p 386)
“. . . illustrates the way in which Barrow, Hudde and de Sluzewere working on the lines suggested by Fermat . . . ”
MAT 5610: 11
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Definitions
Definition (Elementary Calculus Version)The derivative of f at x = a is
f ′(a) = limx→a
f(x)− f(a)
x− aprovided the limit exists (and is finite).
Definition (Analysis Version1)The derivative of f at x = a, an acc pt of f ’s domain in the domain, isf ′(a) iff for any ε > 0 there is a δa > 0 s.t. whenever x ∈ B′(a, δa),then ∣∣∣∣
f(x)− f(a)
x− a − f ′(a)
∣∣∣∣ < ε.
Newton: y Leibniz:dy
dxArbogast / Euler: Dxy Lagrange: y′
1Cauchy, 1823. He used ε for ‘l’erreur’ and δ for ‘la distance’. [Bolzano (1817) but unknown andWeierstrass (c.1840)]. See Grabiner, “Who Gave You the Epsilon? Cauchy and the Origins ofRigorous Calculus,” MAA Monthly 90 (3), p 185–194. (Abel on Cauchy)
MAT 5610: 12
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
I. Differentiable?
Example (Good)Set f(x) = xn.Then
f ′(x) = limx→a
xn − an
x− a= limx→a
(xn−1 + xn−2a+ xn−3a2 + · · ·+ an−1) = nan−1.
Example (Bad)
Set f(x) =√|x| =
{√x x ≥ 0√−x x < 0
.
Then f ′(x) =
1/(2√x ) x > 0
? x = 0
−1/(2√−x ) x < 0
. What about f ′(0)?
limx→0−
f ′(x) = −∞ and limx→0+ f ′(x) =∞ Is there a tan line at 0?
MAT 5610: 13
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
II. Differentiable?
Example
Set h(x) =
{2− x x ≥ 019x
2 − x− 2 x < 0.
Then h′(x) =
−1 x > 0
? x = 029x− 1 x < 0
. What about h′(0)?
limx→0−
h′(x) = −1 = limx→0+
h′(x)
h(x) =
�2 − x x ≥ 019x2 − x − 2 x < 0
MAT 5610: 14
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Contrasting Definitions
Compare:Calculus Analysis
Limit ∀ε > 0 ∃δ > 0f(B′(a, δ) ∩D) ⊆ B(L, ε)
add ‘a is an accumulationpt. of D’
Continuity limx→a
f(x) = f(a) add ‘or a is an isolated pt.’
Derivative f ′(a) = limx→a
f(x)− f(a)
x− a essentially the same
MAT 5610: 15
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
What’s in a Definition
SubtletiesMany authors1 use:
The function y = f(x) has a tangent line at the point(x0, f(x0)) if and only if f is differentiable at x0.
• What about f(x) = 3√x at x = 0?
Others add:The curve has a vertical tangent at (x0, y0) if and only if theslope approaches ±∞ as x→ x0 from either side.
• What about g(x) =
{3√x+ 1 x ≥ 0
3√x x < 0
at x = 0?
1Eg., Calculus, Stewart, 6th ed, Brooks/Cole, 2008, p 113.
MAT 5610: 16
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Differentiability Implies . . .
TheoremIf f is differentiable at x = a, then f is continuous at a.
Proof. (outline).
limx→a
[f(x)− f(a)] = limx→a
[(f(x)− f(a)) · x− a
x− a
]
= limx→a
f(x)− f(a)
x− a · limx→a
x− a
= f ′(a) · 0 = 0
PropositionLet f :D → R and a ∈ D where a is an acc pt of D. Then f ′(a) existsiff qn = f(xn)−f(a)
xn−a always converges to the same real number z forany sequence xn → a with xn 6= a for all n. Then z = f ′(a).
MAT 5610: 17
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Algebra of Derivatives
Theorem (Algebra of Differentiability)Suppose f, g :D → R are differentiable at a and c ∈ R. Then
1. [c · f ]′(a) = c · [f ′(a)]
2. [f ± g]′(a) = [f ′(a)]± [g′(a)]
3. [f · g]′(a) = [f ′(a) · g(a)] + [f(a) · g′(a)]
4. [f/g]′(a) =[f ′(a) · g(a)]− [f(a) · g′(a)]
g2(a)provided g(a) 6= 0
CorollarySuppose fi :D → R is differentiable at a for i = 1..n. Then
1.
[n∑i=1
fi
]′(a) =
n∑i=1
[f ′i(a)]
2.
[n∏i=1
fi
]′(a) =
(n∏i=1
fi(a)
)n∑i=1
[f ′i(a)]
fi(a)(when fi(a) 6= 0, else . . . )
MAT 5610: 18
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Legis Catena
Theorem (The Chain Rule)Suppose f is continuous on [a, b], f ′(x) exists at some point x ∈ (a, b), g isdefined on an interval I containing the range of f, and g is differentiable atthe point f(x). Then [g ◦ f ]′(x) = g′(f(x)) f ′(x)
Proof.• Set y = f(x).
f(t)− f(x) = (t− x)[f ′(x) + u(t)]
g(s)− g(y) = (s− y)[g′(y) + v(s)]
where t ∈ [a, b], s ∈ I, and u→ 0 as t→ x and v → 0 as s→ y.
• Set s = f(t).g(f(t))− g(f(x)) = (t− x) · [f ′(x) + u(t)] · [g′(y) + v(s)]∗
g(f(t))− g(f(x))
t− x = [g′(y) + v(s)] · [f ′(x) + u(t)] when t 6= x
• Let t→ x. Then f cont⇒ s→ y. And the result obtains.
∗Needs f(B′(x, δ) ∩ dom(f)
)∩ dom(g) 6= ∅ !
MAT 5610: 19
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Malum Exemplum
ExampleLet f(x) = x4 − x2 and g(u) = u3/2. Let x = 0. What is (g ◦ f)′(0)?
Then f ′(x) = 4x3 − 2x and g′(u) = 3u1/2/2. So (g ◦ f)′(0) = 0?
No! Graph it! (Explain the problem.)
Exercise
Set Φ(x) =
{[x3 sin(1/x)
]3/2x 6= 0
0 x = 0.
1. Find Φ′(0), if it exists.2. Is Φ continuous at 0?3. Find limx→0 Φ(x), if it exists.
MAT 5610: 20
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Extremus Cuspis
TheoremSuppose f : (a, b)→ R is differentiable at x = c. If c is a relativeextremum of f , then f ′(c) = 0.
Proof.Wolog c is a rel max of f . There is a δ > 0 such that f(x) ≤ f(c) on(c− δ, c+ δ). Whence, for |h| < δ,
f(c+ h)− f(c)
h=
{negative h > 0
positive h < 0
Thus f ′(c) ≤ 0 and f ′(c) ≥ 0. So f ′(c) = 0.
Relation to calculus course’s First and Second Derivative Tests?
MAT 5610: 21
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Vado Domus
Theorem (Inverse Function Theorem)Suppose f is differentiable on the interval I and f ′ exists and isnonzero on I. Then
1. f is one-to-one on I2. f−1 is continuous on I3.(f−1
)′(y) = 1/f ′(x) or, with x = f−1(y), write x′ = 1/y′.
Proof.1. f ′ is nonzero on the interior of I, so f ′ is either always pos or
always neg. Hence f has no extrema on I. Thence f ismonotone. Whence f is one-to-one on I. (And f−1 exists on I.)
2. f is diffy on I ⇒ f is cont on I ⇒ f−1 is cont on I
3. f−1(yn)−f−1(y0)yn−y0 = xn−x0
f(xn)−f(x0) → 1f ′(x0) where xn → x0
MAT 5610: 22
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Horrendus Exempoator
Example (Weierstrass (1872) )
Define f(x) =∞∑n=0
an cos(bnπx) where a ∈ (0, 1), b > 0 is an odd
integer, and ab > 1 + 3π/2.
1. f is continuous everywhere2. f is differentiable nowhere
W (x) =∞�
n=0
cos(7nπx)
4n
FactoidThe set of discontinuities of f is an Fσ set (a countable union of closed sets).What about the set of f ’s nondifferentiable points?
MAT 5610: 23
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude
A theorem today relates . . .
MAT 5610: 24
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Most Valuable Theorems, 1691
Theorem (Rolle’s Theorem (1691))If f is (i) continuous on [a, b], (ii) differentiable on (a, b), and (iii)f(a) = f(b), then there is a point c ∈ (a, b) where f ′(c) = 0.
Proof.Wolog f is nonconstant. Assume f ′(x) 6= 0 on (a, b).
1. Since f is continuous on a compact interval, it has max and minvalues, say M 6= m.
2. If f ′(x) 6= 0 on (a, b), there are no internal extreme vals. So themax and min must be the endpoints: M = f(a) = f(b) = m BC
MAT 5610: 25
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Most Valuable Theorems, 1797
Theorem (Lagrange’s Mean Value Theorem (1797))If f is (i) continuous on [a, b] and (ii) differentiable on (a, b), then there
is a point c ∈ (a, b) where f ′(c) =f(b)− f(a)
b− a .
CorollaryIf f is (i) continuous on [a, b] and (ii) differentiable on (a, b), then thereis a point θ ∈ (0, 1) with f(x+ h)− f(x) = h f ′(x+ θh) wherex, x+ h ∈ [a, b].
Proof.
Apply Rolle’s thm to γ(x) = f(x)− f(b)− f(a)
b− a · x.
Simple Eg: | sin(x1)− sin(x2)| ≤ ?
MAT 5610: 26
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Most Valuable Theorems, 1821
Theorem (Cauchy’s Mean Value Theorem (1821))If f and g are (i) continuous on [a, b] and (ii) differentiable on (a, b),then there is a point c ∈ (a, b) where
f ′(c)g′(c)
=f(b)− f(a)
g(b)− g(a)(g(a) 6= g(b))
or, when g(b) = g(a), then g′(c) · [f(b)− f(a)] = 0.
Proof.Apply Rolle’s thm to
κ(x) = f(x) · [g(b)− g(a)]− g(x) · [f(b)− f(a)].
MAT 5610: 27
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Most Valuable Theorems
Let f(x) = cos(x2/5) and
g(x) = sin(x/3) on [0, 2π].
Set R(t) = [g(t), f(t)].
Example: Cauchy’s Mean Value Theorem
MAT 5610: 28
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Problem Children, I
Counterexamples1. The Heaviside function H(x) is not the derivative of any fcn.
2. f(x) =
{x2 sin(1/x) x 6= 0
0 x = 0has a discontinuous derivative
with f ′(0) = 0
3. g(x) =
{x4 (2 + sin(1/x)) x 6= 0
0 x = 0has abs min at 0, but g′(x)
oscillates in every neighborhood of 0.
4. h(x) =
{x+ 2x2 sin(1/x) x 6= 0
0 x = 0. f ′(0) = 1 but f is not
monotonic in any neighborhood of 0.
5. j(x) = eix fails LMVT
MAT 5610: 29
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Problem Children, II
Exercise (Prin of Math Analysis, Rudin. No. 13, Pg. 115.)13. Suppose a and c are real numbers, c > 0, and f is defined on
[0, 1] by
f(x) =
{xa sin(x−c) (if x 6= 0)
0 (if x = 0)
Prove the following statements:(a) f is continuous iff a > 0.
(b) f ′(0) exists iff a > 1.
(c) f ′ is bounded iff a ≥ 1 + c.
(d) f ′ is continuous iff a > 1 + c.
(e) f ′′(0) exists iff a > 2 + c.
(f ) f ′′ is bounded iff a ≥ 2 + 2c.
(g) f ′′ is continuous iff a > 2 + 2c.
⇒ Homework link (due Friday, Jan 28)
MAT 5610: 30
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Darboux’s Feat
Theorem (Darboux’s Intermediate Value Theorem (1875))Let I be an open interval, and let f :I → R be a differentiablefunction. Suppose [a, b] ⊆ I. For any m between f ′(a) and f ′(b),there exists a value c ∈ [a, b] such that f ′(c) = m.
Proof.Wolog f ′(a) ≤ m ≤ f ′(b). Define
fa(t) =
{f(a)−f(t)
a−t t 6= a
f ′(a) t = aand fb(t) =
{f(t)−f(b)
t−b t 6= a
f ′(b) t = b
Note: fa and fb are continuous and fa(b) = fb(a). So m is eitherbetween fa(a) and fa(b) or fb(a) and fb(b). Wolog, the former.Hence the IVT applied to fa gives an s ∈ (a, b] s.t.m = fa(s) = [f(s)− f(a)]/(s− a). Now, LMVT applied to f givesc ∈ [a, s] where [f(s)− f(a)]/(s− a) = f ′(c); i.e., m = f ′(c).
MAT 5610: 31
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Uniformity
Definition (Uniformly Differentiable)Let f be defined on an open interval I. Then f is uniformly differ-entiable iff f is diffy on I, and for every ε > 0 there is a δ > 0 s.t.
∣∣∣∣f(x1)− f(x2)
x1 − x2
∣∣∣∣ < ε
whenever 0 < |x1 − x2| < δ and xi ∈ I.
PropositionIf f is uniformly differentiable on I, then f ′ is continuous on I.
Example• L(t) = 1/t is uniformly diffy on every interval (a,∞) where a > 0.
• But L(t) = 1/t is not uniformly diffy on the interval (0,∞).
MAT 5610: 32
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Higher
Definition (Higher Derivatives)For n ≥ 2, define
dn
dxnf(x) =
d
dx
(dn−1
dxn−1f(x)
).
Exercises
1. Computedn
dxn(f ◦ g)(x) for n = 1..4.
2. Determine all derivatives of Vn(x) = |x|n.
3. Discuss derivatives of R(x) =
{x x ≥ 0
0 x < 0wrt Darboux’s thm.
MAT 5610: 33
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
In Search of Roots
DefinitionA root of multiplicity m ∈ N is a zero of f s.t. m is the least integer forwhich f can be written as
f(x) = (x− r)m q(x)
where limx→r q(x) 6= 0.
TheoremIf f is m times continuously differentiable, r is a root of multiplicity miff 0 = f(r) = f ′(r) = · · · = f (m−1)(r), but 0 6= f (m)(r).
Proof.⇒ Differentiate f(x) = (x− r)mq(x) and let x→ r.
⇐ Apply LMVT successively to [f(x)− f(r)]/(x− r).
MAT 5610: 34
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Taylor Made Theorems
Definition (Taylor Polynomial)If f has sufficiently many derivatives at x = a, the Taylor polynomial ofdegree n (or order n) is
pn(x) =
n∑
k=0
f (k)(a)
k!(x− a)k
where f (0)(a) = f(a).
Theorem (Taylor’s Theorem (1715∗))Suppose f has n+ 1 derivatives on a neighborhood of a. Thenf(x) = pn(x) +Rn(x) where
Rn(x) =f (n+1)(c)
(n+ 1)!(x− a)n+1
for some c between x and a.
∗Actually, first discovered by Gregory in 1671, ∼ 14 years before Taylor was born!
MAT 5610: 35
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Taylor Remainders
Forms of the Remainder∗
Lagrange (1797) Rn(x) =1
(n+ 1)!f (n+1)(c) (x− a)n+1
Cauchy (1821) Rn(x) =1
n!f (n+1)(c) (x− a)(x− c)n
Integral Form Rn(x) =1
n!
∫ x
a
f (n+1)(t) (x− t)n dt
Uniform Estimate Rn(x) = max∣∣f (n+1)(x)
∣∣ · rn+1
(n+ 1)!on B(a, r)
∗See, e.g., Whitaker and Watson, A Course of Modern Analysis, Cambridge, 1927.
MAT 5610: 36
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Proof Sewed Up
Proof.∗
1. The FToC⇒ f(x) = f(a) +
∫ x−a
0
f ′(x− t) dt.
2. Integrate by parts with u = f ′(x− t) and dv = dt:
f(x) = f(a) + f ′(a)(x− a) +
∫ x−a
0
f ′′(x− t) · t dt
3. Repeat the process choosing u = f (k)(x− t) and dv = tk−1/(k − 1)!arriving at
f(x) = f(a)+f ′(a)(x−a)+f ′′(a)
2!(x−a)2+· · ·+f (n)(a)
n!(x−a)n+Rn(a, x)
where
Rn(a, x) =1
n!
∫ x−a
0
f (n+1)(x− t) · tn dt
∗Why can’t this proof be used in our text?
MAT 5610: 37
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Second Proof
Proof.Let pn(x) be the Taylor polynomial for f centered at a.
1. Define M by f(b) = pn(b) +M(b− a)n+1.
2. Set φ(x) = f(x)− pn(x)−M(x− a)n+1.
Then φ(k)(a) = f (k)(a)− p(k)n (a) = 0 for k = 0..n by the defn of pn.
3. Now φ(a) = 0 = φ(b). So Rolle’s thm gives x1 ∈ (a, b) s.t. φ′(x1) = 0
4. Whence φ′(a) = 0 = φ′(x1). So Rolle’s thm gives x2 ∈ (a, x1) s.t.φ′′(x2) = 0.
5. Continue the process generating xi+1 ∈ (a, xi) s.t. φ(i+1)(xi+1) = 0.
6. After n+ 1 steps, we have
0 = φ(n+1)(xn+1) = f (n+1)(xn+1)−M (n+ 1)!;
i.e., M = f (n+1)(xn+1)/(n+ 1)! proving the thm.
MAT 5610: 38
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Taylor’s Proof
MAT 5610: 39
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Sample
ExampleFind the Maclaurin polynomial p3 with error bounds on [−1, 1] for the
sine integral Si(x) =
∫ x
0
sin(t)
tdt.
n f (n)(x) 1n!f
(n)(0)xn
0 Si(x) 0
1 sin(x)x x
2 cos(x)x − sin(x)
x2 0
3 − sin(x)x − 2 cos(x)
x2 + 2 sin(x)x3 − 1
18 x3
p3(x) = x− x3
18with |R3| ≤ max |Si(4)(x)| · 24
4!≤ 1
5
24
4!=
2
15
MAT 5610: 40
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Si and p3
-5 -2.5 0 2.5 5
-2.5
2.5Si(x) =
� x
0
sin(t)
tdt
p3(x) = x − x3
18
MAT 5610: 41
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Si and p3 Error
-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1
5!10-4
0.001
0.0015
Terr(x) = |Si(x) − p3(x)|
maxx∈[−1,1]
Terr(x) = Terr(1) ≈ 1.6386260× 10−3
MAT 5610: 42
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude
XKCD
MAT 5610: 43
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Try It!
ExerciseLet f(x) = tan(sin(x)) defined for all x ∈ R. Note f ∈ C∞.
1. Find an n ∈ N so that max |Rn(x)| ≤ 1 on [−π, π].
-π -0.5π 0 0.5π π
-2
-1
1
2f(x) = tan(sin(x))
To compute the nth Taylor polynomial in Maple, defineT:=n->convert(taylor(f(x),x,n+1),polynom)
MAT 5610: 44
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Try It! Computed
In Maple 15:
> f := x -> tan(sin(x)):> d := n -> (D@@n)(f);
n→ D(n)(f)
> c := n -> d(n)(0)/n!;
n→ d(n)(0)
n!
> N := 20:> <<$i=1..N>
|<d(i))(0)$i=1..N>
|evalf[5](<c(i)$i=1..N>)>
n f (n)(0) f (n)(0)/n!1 1 12 0 03 1 0.166674 0 05 -3 -0.0250006 0 07 -107 -0.0212308 0 09 -1095 -0.0030175
10 0 011 41897 0.001049612 0 013 3027637 0.0004862114 0 015 34528445 0.00002640416 0 017 -11832720271 -0.00003326718 0 019 -1190157296815 -0.000009783820 0 0
MAT 5610: 45
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Try It! Graphed
f(x) = tan(sin(x)) and T37(x).
MAT 5610: 46
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Differentiable Inequality
Theorem (Jensen’s Inequality, I (1906))Let f be twice continuously differentiable on (a, b) with f ′′(x) ≥ 0.Suppose xi ∈ (a, b) for i = 1..n. Then
f
(1
n
n∑
k=1
xk
)≤ 1
n
n∑
k=1
f(xk)
Proof.1. Let x∗ = 1
n
∑nk=1 xk. Define L(x) = f(x∗) + f ′(x∗) · (x− x∗).
2. L is linear gives f(x∗) = L(x∗) = 1n
∑nk=1 L(xk).
3. Since f is concave up (convex), then L(x) ≤ f(x) on (a, b).Whence
f
(1
n
n∑
k=1
xk
)=
1
n
n∑
k=1
L(xk) ≤ 1
n
n∑
k=1
f(xk).
MAT 5610: 47
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Mean Application
Theorem (Jensen’s Inequality (1906))If f is convex on (a, b), xk ∈ (a, b), and µk > 0 with
∑nk=1 µi = 1, then
f
(n∑k=1
µkxk
)≤
n∑k=1
µkf(xk)
Example (Arithmetic-Geometric Mean Inequality)1. Set f(x) = − ln(x) for x > 0. Then f ′′(x) > 0. Let xk > 0.
2. By Jensen’s inequality,
− ln
(1
n
n∑k=1
xk
)≤ − 1
n
n∑k=1
ln(xk) = − 1
nln
(n∏k=1
xk
)
3. Whence
ln
n
√√√√ n∏k=1
xk
≤ ln
(1
n
n∑k=1
xk
)⇒
A-GMI
n√x1x2 · · ·xn ≤
x1 + x2 + · · ·+ xnn
MAT 5610: 48
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Guillaume de l’Hopital
Actually:M. Guillaume-Francois-Antoine Marquis de l’Hopital, Marquis deSainte-Mesme, Comte d’Entremont et Seigneur d’Ouques-la-Chaise.
TheoremSuppose that f and g are
• continuous on [a, b],
• differentiable on (a, b),
• limx→a+ f(x) = 0 = limx→a+ g(x), and
• g′(x) 6= 0 on the interval (a, a+ ε).
If limx→a+
f ′(x)
g′(x)= L, then lim
x→a+
f(x)
g(x)= L
Proof.If xn → a, then [f(xn)− f(a)] / [g(xn)− g(a)] = f ′(cn) / g′(cn) & cn → a.
MAT 5610: 49
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Example
Example
Determine limn→∞
(√n2 + n+ 1− n
).
1.√n2 + n+ 1− n =
(√n2 + n+ 1
)2 − n2
√n2 + n+ 1 + n
=n+ 1√
n2 + n+ 1 + n
2.n+ 1√
n2 + n+ 1 + n=
1 + (1/n)√1 + (1/n) + (1/n2) + 1
→ 1√1 + 1
=1
2
3. Whence limn→∞
(√n2 + n+ 1− n
)=
1
2
MAT 5610: 50
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Diwrnod o Meddwl
Aha! If true, prove; if false, fix, then prove.1. Set y = (4x− 1)x, then y′ = x (4x− 1)x−1(4).
2. f(x) = x3 + x−√
2 has exactly one real root.
3. If f is differentiable at a, then f ′(a) is unique.
4. If f ′ is bounded, then f is bounded.
5. If f is bounded, then f ′ is bounded.
6. One point can be both an inflection pt. and a rel. maximum.
7. a is an inflection point iff f ′′(a) = 0.
8. If x ≥ 0, then√
1 + x < 1 + 12x.
9. If f is continuous and one-to-one, then f is differentiable.
10. If f and g are diffy and f(x)g(x) = 1, then f ′(x)f(x) + g′(x)
g(x) = 0.
MAT 5610: 51
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Diwrnod o Meddwl, II
Aha! If true, prove; if false, fix, then prove.
11. If limn→∞
n
[f
(a+
1
n
)− f(a)
]is finite, then f ′(a) exists.
12. If f is differentiable and uniformly continuous on (a, b), then f ′ isbounded on (a, b).
13. If f is continuously differentiable on an interval I, then f isuniformly continuous on I.
14. If a is a triple root of a three-times differentiable fcn f, then fmust have an inflection pt at a.
15. f(x) = e−x − 12x
2 + x− 1 has a triple root at 0.
16. f ′ may have more roots than f .
17. Let f ∈ C1(R) and f(0) = 0. If f ′(x) > 0 always, then f(x) > 0for x > 0.
MAT 5610: 52
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Bounded Variation
Definition (Partition)Let [a, b] be a compact interval. A partition of [a, b] is a set of n+ 1points P = {a = x0 < x1 < · · · < xn = b}. Set ‖P‖ = n.
Definition (Bounded Variation — Jordan(1881))Suppose f is defined on the compact interval [a, b]. The variation of fon [a, b] is
Vf (a, b) = supP∈P
‖P‖∑
k=1
|f(xk)− f(xk−1)|
where P is the collection of all partitions of [a, b]. Then f is ofbounded variation on [a, b] iff Vf (a, b) is finite.
¿How does variation relate to arclength (“rectification,” Neile (1657))?
MAT 5610: 53
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Sampling Variation
Examples• f(x) = sin(x) has bounded variation on any compact interval.
• g(x) =
{sin(1/x) x 6= 0
0 x = 0is not of bounded variation on any
compact interval containing 0.
• h(x) =
{x2 sin(1/x) x 6= 0
0 x = 0is of bounded variation on any
compact interval.
• j(x) =
{x2 sin(1/x2) x 6= 0
0 x = 0is not of bounded variation on any
compact interval containing 0 but j is differentiable at 0.
• A montone function?
MAT 5610: 54
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Variations on a Theorem
TheoremIf f and g are BV on [a, b], then f ± g and f · g are BV . If g isbounded away from 0, then f/g is of BV .
TheoremSuppose f is continuous on [a, b]. If f has a bounded derivative on(a, b), then f is BV on [a, b].
Proposition1. f is BV on [a, b] iff f can be written as the difference of two
increasing functions. (Jordan (1893))2. A BV fcn on [a, b] has only jump discontinuities.3. The discont of a BV fcn on [a, b] are countable. (A. Froda (1929))4. A BV fcn on [a, b] has a derivative a.e. (Lebesgue (1904)∗)
∗See Botsko
MAT 5610: 55
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude: Planimeter to ENIAC
Amler’s Planimeter, 1854
Thomson Differential Analyzer, 1876
ENIAC: Electronic NumericalIntegrator And Computer, 1946
MAT 5610: 56
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Integral
Classical Greek ProblemsTrisect an Angle: divide an angle into three equal parts. (Wantzel, 1837)
Square the Circle: create a square having the same area as a given circle.(Lindemann, 1882)
Duplicate the Cube: construct a second cube with double the volume of agiven cube. (Descartes, 1637)
Quadrature of the Lune: Hippocrates of Chios’ construction (c 440 BC) isone of the “Great Theorems.” (See Journey Through Genius)
Universal Etymological English Dictionary, Bailey (1731)QUADRATURE [quadratura, L] the making of a thing fquare, or the finding afquare equal to the area of any figure given.
QUADRATURE of Curves [in the higher Geometry] is the meafuring of theirarea, or the finding of a rectilinear fpace, equal to a curvilinear fpace.
MAT 5610: 57
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Integrators
Cavalieri, 1630’s Newton, 1670’s Leibniz, 1670’s Cauchy, 1820’s
Riemann, 1840’s Darboux, 1870’s Stieltjes, 1880’s Lebesgue, 1900’s
MAT 5610: 58
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Cauchy’s Integral
Definition (Cauchy’s Integral (1823))Suppose f is continuous on [a, b]. For the partition P = {x0 < x1 < · · · < xn}(x0 = a, xn = b) set ∆xk = (xk − xk−1). Define the Cauchy sum to be
C(P, f) =n∑k=1
f(xk−1)∆xk
If there is a number I s.t. for any ε > 0 there is a δ > 0 where |C(f, P )− I|<εwhenever max
P∆x < δ, then f is Cauchy integrable. Write I =
∫ baf(x) dx.
Theorem (Cauchy (1823))If f is continuous on [a, b], then f is Cauchy integrable.
Oops!The function f(x) = 1 for x 6= 1 & f(1) = 2 is not Cauchy integrable on [0, 2].
MAT 5610: 59
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Riemann’s Integral
Definition (Partition)Let [a, b] be a compact interval. A partition of [a, b] is a set of n+ 1 pointsP = {a = x0 < x1 < · · · < xn = b}. Set |P | = maxk=1..n ∆xk where∆xk = xk − xk−1. P is refinement of Q iff Q ⊆ P.
Definition (Riemann Sum)
The Riemann sum of a bounded function on [a, b] is R(P, f) =
n∑k=1
f(ck) ∆xk
where the “tags” ck ∈ [xk, xk−1].
Definition (Darboux Sums)The upper- & lower-Darboux sums of a bounded function on [a, b] are
U(P, f) =n∑k=1
Mk ∆xk and L(P, f) =
n∑k=1
mk ∆xk
respectively, with Mk= supx∈Ik
f(x) and mk= infx∈Ik
f(x) for Ik=[xk−1, xk].
MAT 5610: 60
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Order Lemmas
LemmaSuppose m ≤ f(x) ≤M on [a, b], and let P be a partition of [a, b].Then
m(b− a) ≤ L(P, f) ≤ R(P, f) ≤ U(P, f) ≤M(b− a)
for any set {ck} of tags with P .
LemmaSuppose f is bounded on [a, b]. Let P and Q be partitions of [a, b].Then
1. If Q is a refinement of P, then
L(P, f) ≤ L(Q, f) ≤ U(Q, f) ≤ U(P, f).
2. L(P, f) ≤ U(Q, f) for any partitions P and Q.
MAT 5610: 61
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Order Lemmas
Proof. (Second Order Lemma).1. Suppose Q = P ∪ {x∗} with x∗ ∈ (xi−1, xi). Since
infx∈[xi−1,xi]
f(x)
︸ ︷︷ ︸mi
≤ infx∈[xi−1,x∗]
f(x)
︸ ︷︷ ︸mi,1
or infx∈[x∗,xi]
f(x)
︸ ︷︷ ︸mi,2
,
we have
mi∆xi = mi(x∗−xi−1)+mi(xi−x∗) ≤ mi,1(x∗−xi−1)+mi,2(x∗−xi)
Whence L(P, f) ≤ L(Q, f).
If Q = P ∪ {x∗k | k = 1..m}, repeat the argument m times.
2. By 1, P ∪Q refines both P and Q, whence L(P ) ≤ L(P ∪Q) andU(P ∪Q) ≤ U(Q).
MAT 5610: 62
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Upper and Lower Riemann-Darboux Integrals
DefinitionFor a bounded function f on [a, b], define∗∫ b
a
f(x) dx = supPL(P, f) and
∫ b
a
f(x) dx = infPU(P, f).
PropositionIf f is bounded on [a, b] and P is any partition of [a, b], then
L(P, f) ≤∫ b
a
f(x) dx ≤∫ b
a
f(x) dx ≤ U(P, f).
Definition (Riemann/Darboux Integrable)A bounded function f is Riemann- or Riemann-Darboux integrable iff∫ b
a
f(x) dx =
∫ b
a
f(x) dx = A.
We write f ∈ R[a, b] and A =∫ baf(x) dx. Sample
∗Why don’t we use limits?
MAT 5610: 63
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Short Readings
Exercise (Read and Ponder)• Go to our “AsULearn class website.” Read the selections from:
1. A Garden of Integrals by Frank E. Burk
2. Lebesgue’s Theory of Integration: Its Origins and Development byThomas Hawkins
• Read §6.2 from David Bressoud’s A Radical Approach to RealAnalysis
• Investigate the theorem:Proposition: A bounded function f on [a, b] is Cauchy
integrable iff f is continuous.
How did Cauchy deal with simple discontinuities?
MAT 5610: 64
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
“Bad Integral! No Biscuit!”
Example (Bad Integral)
Determine∫ 1
−1
χQ(x) dx where χQ(x) =
{1 x ∈ Q0 x /∈ Q
.
For any P , we have L(P, χQ) = 0 and U(P, χQ) = 2. Thence∫ 1
−1
χQ(x) dx = 0 < 2 =
∫ 1
−1
χQ(x) dx,
whereupon the integral fails to exist.
Example (Good Integral)
Determine∫ 2
0
f(x) dx where f(x) =
{1 x 6= 1
2 x = 1.
The only “problem point” is at x = 1 where f is not continuous. Refine anypartition containing x = 1 as a tag by making ∆xk of that interval less thanε/8. The total contribution to any R-D sum from those terms is less than ε/2.Whence 2 ≤ L(P, f) ≤ U(P, f) ≤ 2 + ε, and so the integral equals 2.
MAT 5610: 65
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Function’s View
TheoremLet f be bounded on [a, b]. Then f ∈ R[a, b] iff for every ε > 0 there isa partition P s.t. U(P, f)− L(P, f) < ε.
Proof.(⇒) Let A =
∫ baf and ε > 0. Then
1. ∃PU s.t. U(PU ) < A+ ε/2 (Why?) and2. ∃PL s.t. L(PL) > A− ε/2. (Why?)
Set P = PL ∪ PU . Whence
A− ε/2 < L(PL) ≤ L(P ) ≤ U(P ) ≤ U(PU ) < A+ ε/2
(⇐) Let ε > 0. Suppose ∃P with U(P )− L(P ) < ε. ThenU(P ) < L(P ) + ε. Whence
L(P ) ≤∫ b
a
f ≤∫ b
a
f ≤ U(P ) < L(P ) + ε
MAT 5610: 66
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Who Lives in R?
Theorem{f | f is monotone on [a, b]} ⊂ R[a, b]
Proof.Wolog f is mono increasing. Let ε > 0 and set Pn = {xk | k = 0..n}where xk = a · (n− k)/n+ b · (k/n). Then ∆xk = (b− a)/n. We have
U(Pn)−L(Pn) =
n∑
k=1
[f(xk)−f(xk−1)] ·∆xk = [f(b)−f(a)] · b−an
=K
n
Hence, U(P )− L(P ) < ε for large enough n. Thus f ∈ R[a, b].
Example
Heaviside(x) =
{1 x ≥ 0
0 x < 0∈ R[a, b] for any a < b.
MAT 5610: 67
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Who Else Lives in R?
TheoremC[a, b] ⊂ R[a, b]: All continuous functions are integrable.
Proof.Let ε > 0.
1. Since f is continuous on a compact set, it is uniformly continuous.Whence ∃δ > 0 s.t. |f(t1)− f(t2)| < ε/(b− a) whenever |t1 − t2| < δ.
2. Choose P s.t. |P | < δ.
3. f cont on [a, b] says ∃tk, sk s.t. Mk = f(tk) and mk = f(sk) for each k.
4. Hence U(P )− L(P ) ≤n∑k=1
|f(tk)− f(sk)|∆xk
5. Since |P | < δ, thence U(P )− L(P ) < ε/(b− a) ·n∑k=1
∆xk = ε.
I.e., there is a partition P s.t. U(P )− L(P ) < ε
MAT 5610: 68
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Crazy Neighbors
Example
D Bernoulli and Goldbach soughtto extend n! to real values(1720’s). Euler’s solution (1729)is the Gamma function∗ definedby
Γ(x) =
∫ ∞
0
tx−1e−tdt
for x 6= 0,−1,−2, . . . .
Γ(x) =
� ∞
0
tx−1e−tdt
y = 1/Γ(x)
For x > 0, we see that Γ(x) is continuous. Whence Γ(x) is Riemannintegrable on any compact interval [a, b] for 0 < a < b <∞.
(Curious factoid:∫ x
1Γ(t) dt < Γ(x) for x ≥ 4.)
∗See the Handbook of Mathematical Functions, M Abramowitz & I Stegun, pg 255.(A result of Roosevelt’s WPA!) See also the DLMF Chap 5.
MAT 5610: 69
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Degenerate Function
Example (Burk’s Bad Badger)
Set f(x) =
{2x sin(1/x2)− 2x−1 cos(1/x2) x 6= 0
0 x = 0.∫ 1/
√π
0
f(t)dt =?
Cauchy: No! f is not continuous at 0!Riemann: No! f is unbounded at 0!
Lebesgue: No! |f | can’t be integrated, so neither can f !Newton/Leibniz: Dudes, whatev. Since d
dx
[x2 sin(1/x2)
]=f(x), it’s 0.
MAT 5610: 70
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Props to Riemann
Theorem (Riemann Integration is Linear)Suppose f , g ∈ R[a, b] and c ∈ R. Then
1. f ± g ∈ R[a, b], and∫ ba
(f ± g) =∫ baf ±
∫ bag.
2. cf ∈ R[a, b], and∫ ba
(cf) = c∫ baf.
Proof. (Key steps).
• ∃Pf , Pg s.t.∫ baf ≤ U(Pf ) < ε+
∫ baf and
∫ bag ≤ U(Pg) < ε+
∫ bag
• U(Pf ∪ Pg, f + g) ≤ U(Pf , f) + U(Pg, g)
∴∫ ba
(f + g) ≤ 2ε+∫ baf +
∫ bag
An analogous argument gives∫ ba(f + g) ≥
∫ baf +
∫ bag.
Whence∫ ba
(f + g) ≤∫ baf +
∫ bag ≤
∫ ba(f + g).
Hence∫ ba
(f + g) =∫ baf +
∫ bag
MAT 5610: 71
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
More Props to Riemann
TheoremSuppose f , g ∈ R[a, b]. If f(x) ≤ g(x) on [a, b], then
∫ baf ≤
∫ bag.
TheoremSuppose f is bounded on [a, b], and c ∈ [a, b]. Then (i) f ∈ R[a, b]⇐⇒(f ∈ R[a, c] and f ∈ R[c, b]) and (ii)
∫ baf =
∫ caf +
∫ bcf .
TheoremIf (i) f ∈ R[a, b], (ii) g ∈ R[c, d], and (iii) f([a, b]) ⊆ [c, d], then f ◦ g ∈ R[a, b].
CorollaryIf f, g ∈ R[a, b], and n ∈ N, then (i) fn ∈ R[a, b], and (ii) fg ∈ R[a, b].
Proof.X (i) g(x) = xn and (ii) fg = (f + g)2/4− (f − g)2/4
MAT 5610: 72
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude 2/17/11 9:10 PMPiled Higher and Deeper
Page 2 of 2http://www.phdcomics.com/comics/archive_print.php?comicid=1421
Piled Higher and Deeper by Jorge Cham www.phdcomics.com
title: "Guide to T.A. Office Hours" - originally published 2/16/2011
MAT 5610: 73
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
MVTs, I
Theorem (Bonnet’s Mean Value Theorem (1849))Suppose f ∈ C[a, b] and g ∈ R[a, b] with g(x) ≥ 0. There is a c ∈ (a, b) s.t.∫ b
a
f(x)g(x) dx = f(c) ·∫ b
a
g(x) dx
Proof.1. f ∈ C[a, b] =⇒ f is bounded: m ≤ f(x) ≤M2. g ≥ 0 =⇒ mg(x) ≤ f(x)g(x) ≤Mg(x)
=⇒ m∫ bag ≤
∫ bafg ≤M
∫ bag
3. Wolog∫ bag 6= 0. Then set A =
∫ bafg /
∫ bag.
4. Since m ≤ A ≤M and f ∈ C[a, b], there is a c ∈ (a, b) s.t. f(c) = A.
Corollary (Cauchy’s ‘Water in the Bucket’ Thm (1821))If f ∈ C[a, b], then there is a c ∈ (a, b) s.t.
∫ baf = f(c)(b− a).
MAT 5610: 74
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
MVTs, II
Theorem (The Second Mean Value Thm for Integrals)If f is monotonic on [a, b], then there is a c ∈ (a, b) s.t.∫ b
a
f(x) dx = f(a)(c− a) + f(b)(b− c)
Proof.• Define g(t) = f(a)(t− a) + f(b)(b− t) = µ t+ β.
• Since f is monotone, g(a) ≤∫ baf ≤ g(b)
• g has the IVP (Why?) which finishes the proof.
Theorem (Generalized Second MVT for Integrals)If f is monotonic on [a, b] and g ∈ C[a, b], then there is a c ∈ (a, b) s.t.∫ b
a
f(x)g(x) dx = f(a)
∫ c
a
g(x) dx+ f(b)
∫ b
c
g(x) dx
MAT 5610: 75
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Integral Inequalities
Theorem (Holder’s Inequality∗)If f, g ∈ R[a, b] and p, q ∈ [1,∞] are such that 1
p+ 1
q= 1, then∫ b
a
|fg| ≤[∫ b
a
|f |p]1/p
·[∫ b
a
|g|q]1/q
.
Corollary (Cauchy-Bunyakovsky-Schwarz Inequality)If f and g ∈ R[a, b], then[∫ b
a
|fg|]2
≤∫ b
a
|f |2 ·∫ b
a
|g|2.
Theorem (Minkowski’s Inequality)If f, g ∈ R[a, b] and p ≥ 1, then[∫ b
a
|f + g|p]1/p
≤[∫ b
a
|f |p]1/p
+
[∫ b
a
|g|p]1/p
∗See handout.
MAT 5610: 76
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
FToC
TheoremSuppose f is differentiable on [a, b] and f ′ ∈ R[a, b]. Then
∫ b
a
f ′(x) dx = f(x)∣∣∣b
a= f(b)− f(a)
Proof.1. Let P be a partition with tags ck chosen via the MVT s.t.
f ′(ck)(xk − xk−1) = f(xk)− f(xk−1).
2. With mk = inf f and Mk = sup f on [xk − xk−1], we seeL(P, f) ≤ R(P, f) ≤ U(P, f) =⇒ L(P, f) ≤ f(b)−f(a) ≤ U(P, f)
3. Since P is arbitrary and f ′ ∈ R[a, b], the result follows.
What about:∫ 1
0e−x
2
dx?∫ +1
−1U(x) dx?
∫ +1
−1
[x2 sin(1/x2)
]′dx?
MAT 5610: 77
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Results of FToC
Theorem (Integration by Parts)Suppose f and g are differentiable and f ′, g′ ∈ R[a, b]. Then
∫ b
a
f(x)g′(x) dx = f(b)g(b)− f(a)g(a)−∫ b
a
f ′(x)g(x) dx
Proof.1. f, g diffy⇒ f, g cont⇒ f, g ∈ R ⇒ f ′g, f g′ ∈ R ⇒ (fg)′ ∈ R2. Integrate the equation (fg)′ = f ′g + f g′ to finish.
Theorem (Indefinite Integral Theorem)Let f ∈ R[a, b]. Define F (x) =
∫ xaf(t) dt. Then
1. F is uniformly continuous on [a, b], and2. if f is cont at c, then F is diffy at c and F ′(c) = f(c).
MAT 5610: 78
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
IIT’s Proof
Proof. (Indefinite Integral Theorem).1. Let ε > 0. f ∈ R ⇒ f bounded, say by M . Choose δ < ε/M.
Let |x1 − x2| < δ. Then
|F (x1)− F (x2)| =∣∣∣∣∫ x1
a
f(t) dt−∫ x2
a
f(t) dt
∣∣∣∣ =
∣∣∣∣∫ x1
x2
f(t) dt
∣∣∣∣
Whence |F (x1)− F (x2)| ≤M · |x1 − x2| < ε.
2. Let f ∈ C[a, b] and ε > 0. Then ∃δ > 0 s.t. if |x− c| < δ, then|f(x)− f(c)| < ε. Then
∣∣∣∣F (x)− F (c)
x− c − f(c)
∣∣∣∣ =1
|x− c|
∣∣∣∣∫ x
c
f(t) dt− f(c)(x− c)∣∣∣∣
≤ 1
|x− c|
∫ x
c
|f(t)− f(c)| dt < 1
|x− c|
∫ x
c
ε dt < ε
MAT 5610: 79
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Fun Example
Example (O.M.G.)
Determine F (x) =
∫ x
0
sin(2t) dt.
• F1(x) = − 12 cos(2x) (check with differentiation)
• sin(2t) = 2 sin(t) cos(t). Let u = sin(t) and du = cos(t) dt. Then
F2(x) = sin2(x)
• sin(2t) = 2 cos(t) sin(t). Let u = cos(t) and du = − sin(t) dt. Then
F3(x) = − cos2(x)
Which is the correct F (x)? F (x) = F2(x), = F1(x) + 12 , = F3(x) + 1.
MAT 5610: 80
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Part II
Outerlude
MAT 5610: 81
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Riemann-Stieltjes Integration
Thomas Jan Stieltjes (1856–1894 ) modeled the mo-ments of thin rods with nonuniform mass distribu-tions. The nth moment of f about the point c is
µn =
∫ ∞
−∞(x− c)nf(x) dx
Stieltjes considered the distributions that came from nonuniform den-sities or from discrete masses placed along the x-axis, problems thatRiemann’s integral didn’t handle. He published his generalization ofRiemann’s integral in 1894, the year he died.
Stieltjes problem, now called the “Stieltjes Moment Problem,” is still thesubject for much research. (Also see the “Hamburger Moment Problem”.)
MAT 5610: 82
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Riemann-Stieltjes Integration
Definition (Upper and Lower Riemann-Stieltjes Integrals)Let f be bounded and α be monotonically increasing on [a, b]. For apartition P, define the upper- and lower Riemann-Stieltjes sums by
U(P, f, α) =
n∑
k=1
Mk∆αk and L(P, f, α) =
n∑
k=1
mk∆αk
where ∆αi = α(xi)− α(xi−1). Define the upper- and lowerRiemann-Stieltjes integrals as
∫ b
a
f dα = infPU(P, f, α) and
∫ b
a
f dα = supPL(P, f, α)
When α(x) = x, the Riemann-Stieltjes integral is just a Riemann integral.
MAT 5610: 83
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Riemann-Stieltjes Integrability
DefinitionIf∫fdα=
∫fdα, thenf is Riemann-Stieltjes integrable; write f ∈ R(α).
TheoremA function f is Riemann-Stieltjes integrable on [a, b] if and only if forevery ε > 0 there is a partition P of [a, b] such that
U(P, f, α)− L(P, f, α) < ε
Proof.Essentially the same as for Riemann-Darboux sums.
MAT 5610: 84
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Continuity & Riemann-Stieltjes Integrals
TheoremIf f is continuous on [a, b], then f ∈ R(α) on [a, b].
Proof.1. Let ε > 0. f is uniformly cont, so ∃δ > 0 s.t. if |x1 − x2| < δ, then|f(x1)− f(x2)| < ε/(α(b)− α(a)).
2. If |P | < δ, then on [xk−1, xk], we have Mk −mk<ε/(α(b)− α(a)).So
U(P, f, α)− L(P, f, α) =
n∑
k=1
(Mk −mk)∆αk
≤n∑
k=1
ε
α(b)− α(a)∆αk
=ε
α(b)− α(a)
n∑
k=1
∆αk = ε
MAT 5610: 85
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Monotonicity & Riemann-Stieltjes Integrals
TheoremIf f is monotonic on [a, b] and α is continuous, then f ∈ R(α) on [a, b].
Proof.1. Let ε > 0 and n ∈ N. Choose∗ P s.t. ∆αk = (α(b)− α(a))/n.
2. Wolog f mono incr. Then Mk = f(xk) and mk = f(xk−1).3. Whence
U(P, f, α)− L(P, f, α) =α(b)− α(a)
n
n∑
k=1
[f(xk)− f(xk−1)]
=α(b)− α(a)
n· [f(b)− f(a)] < ε.
∗Requires α to be continuous. Remember, α is still monotone increasing.
MAT 5610: 86
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Pointing to Continuity
TheoremIf f is bounded and has finitely many discontinuities on [a, b], and α iscontinuous at each discontinuity of f, then f ∈ R(α).
Proof.1. Let ε > 0, set M = sup |f(x)|, and let E = {discontinuities of f}.2. Since E is finite and α is cont on E, cover E with finitely many disjoint
intervals (uj , vj) s.t.∑
(α(vj)− α(uj)) < ε.
3. The set A = [a, b]−⋃
(uj , vj) is compact, so f is unif cont on A. Then∃δ > 0 s.t. if |x1 − x2| < δ, then |f(x1)− f(x2)| < ε.
4. Form P s.t.: • uj , vj ∈ P , • no point of (uj , vj) is in P , and • If xi−1 6=any uj , then ∆xi = xi − xi−1 < δ.
5. Then Mi −mi < 2M , and Mi −mi ≤ ε unless xi−1 = uj .
Whence U(P, f, α)− L(P, f, α) ≤ [α(b)− α(a)] ε︸ ︷︷ ︸xi 6=vj terms
+ 2Mε︸ ︷︷ ︸xi=vj terms
MAT 5610: 87
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Exercises
Exercises
1. Compute∫ 1
−1
5(x2 + 1) d(x3 + 1). ( HINT: 16 )
2. Define Uj(x) =
1 x > 0
uj x = 0
0 x < 0
for j = 1, 2, and 3
2.a Let u1 = 0.Prove: f ∈ R(U1) iff f(0+) = f(0) and then
∫f dU1 = f(0).
2.b Let u2 = 1.State and prove a similar result for U2.
2.c Let u3 = 1/2.Prove: f ∈ R(U3) iff f is continuous at 0.
2.d Prove:If f is continuous at 0, then
∫f dU1 =
∫f dU2 =
∫f dU3 = f(0).
MAT 5610: 88
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Exercise Hints
Exercises (Hints)1. Use equipartitions. Split U and L into decr and incr parts for f .
(A ‘coming attraction’ theorem makes it very easy.)
2. Let P be any partition containing the point x = 0.
2.1 The only interval with ∆αk 6= 0 is [0, xk]. All but the intervalcontaining 0 have ∆αk = 0. Then consider lim
xk→0
(Mk −mk)
2.2 Repeat with [xk, 0].Let P be any partition not containing the point x = 0.
2.3 Then 0 ∈ (xk∗−1, xk∗), so U(P )− L(P ) = (Mk∗ −mk∗)∆αk∗ .Consider what happens as both xk∗−1, xk∗ → 0.
MAT 5610: 89
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude, II
MAT 5610: 90
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Alpha’s Bucket
PropositionIf f is bounded on [a, b] and f ∈ R(α), then there exists m and M ∈ Rsuch that
m[α(b)− α(a)] ≤∫ b
a
f(x) dα(x) ≤M [α(b)− α(a)]
This result matches Cauchy’s Water in the bucket theorem.
Example
Consider∫ 1
−1
5(x2 + 1) d(x3 + 1).
Since ∆α = α(1)− α(−1) = 2 and 5 ≤ f(x) ≤ 10, we see
5 · 2 = 10 ≤∫ 1
−1
5(x2 + 1) d(x3 + 1) ≤ 10 · 2 = 20.
MAT 5610: 91
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Composition
TheoremIf f is bounded, f ∈ R(α), and φ is cont on range(f), then φ◦f ∈ R(α).
Proof.1. Let ε > 0. Suppose m ≤ f(x) ≤M. Then ∃δ > 0 s.t. δ < ε and if|s− t| < δ, then |φ(s)− φ(t)| < ε. Let M∗i ,m∗i be the “M & m”s for φ◦f
2. ∃P s.t. U(P, f, α)− L(P, f, α) < δ2
3. Divide P by i ∈ A if Mi −mi < δ and i ∈ B otherwise
4. For i ∈ A, we have M∗i −m∗i <ε⇒∑i∈A
(M∗i −m∗i )∆αi<ε[α(b)−α(a)]
5. For i ∈ B, M∗i −m∗i ≤ 2 sup |φ|.
6. Now∑i∈B
δ∆αi ≤(3.)
∑i∈B
(Mi−mi)∆αi <(2.)
δ2 ⇒∑i∈B
∆αi < δ
7. Whence U(P, φ◦f, α)− L(P, φ◦f, α) =
[∑i∈A
+∑i∈B
](M∗i −m∗i )∆αi
≤ ε[α(b)−α(a)] + [2 sup |φ|] δ < const · ε
MAT 5610: 92
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Exercises, II
Exercises1. Evaluate
1.1∫ π/2
0
cos(x) d sin(x)
1.2∫ 2
0
x d[x]
1.3∫ π
0
x d cos(x) (Surely, cos is decreasing! Yes, but don’t call me ...)
2. Prove: If g is monotone increasing on [a, b], then∫ b
a
dg exists
and is equal to g(b)− g(a).
3. Suppose that f and α are both discontinuous at x = 0. What can
you say about∫ 1
−1
f dα?
MAT 5610: 93
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude, III
MAT 5610: 94
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
RSI: Algebra
Theorem (RS Integration is Bilinear)
Let f and g ∈ R(α) on [a, b] and c ∈ R. Then
1.∫ b
a
(cf) dα = c
∫ b
a
f dα and∫ b
a
f d(cα) = c
∫ b
a
f dα
2.∫ b
a
(f + g) dα=
∫ b
a
f dα+
∫ b
a
g dα and∫ b
a
f d(α+ β)=
∫ b
a
f dα+
∫ b
a
f dβ
3. f · g ∈ R(α) (NB: f and f · g ∈ R(α) 6⇒ g ∈ R. Eg? )
Theorem (RS Integration is Ordered)If f and g ∈ R(α) on [a, b] and f(x) ≤ g(x), then
1.∫ b
a
f dα ≤∫ b
a
g dα
2.∣∣∣∣∫ b
a
f dα
∣∣∣∣ ≤ ∫ b
a
|f | dα
MAT 5610: 95
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Making a Massive Point
DefinitionDefine the shifted Heaviside function to be
Ux0(x) =
{0 x ≤ x0
1 x > x0
(Note: Ux0(x) = U(x− x0).)
TheoremIf f is bounded on [a, b] and continuous at x0 ∈ (a, b), then
∫ b
a
f dUx0 = f(x0)
MAT 5610: 96
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Wow
TheoremLet
∑n
cn be a convergent sequence of positive terms and {xn} be a
sequence of distinct points in (a, b). Set α(x) =
∞∑k=1
ck Uxk (x). If f is
continuous on [a, b]. Then ∫ b
a
f dα =∞∑k=1
ckf(xk)
ExampleLet ck = 1/k2 and xk = (k − 1)/k. Set α =
∑n cn Uxn . Then∫ 1
0
1
1 + xdα =
∞∑k=1
1
k2
1
1 + (k − 1)/k=
∞∑k=1
1
k(2k − 1)= ln(4)
(Note:∫ 1
0
1/(1 + x) dx = ln(2).)
MAT 5610: 97
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Wow2
TheoremIf f is bounded and α′ ∈ R(α), then f ∈ R(α) iff fα′ ∈ R and∫ b
a
f dα =
∫ b
a
f · α′ dx
Example
Compute∫ 1
−1
5(x2 + 1) d(x3 + 1).
∫ 1
−1
5(x2 + 1) d(x3 + 1)=
∫ 1
−1
5(x2 + 1)(3x2) dx=15
∫ 1
−1
x4 + x2 dx = 16
For a ‘reasonable’ α, separate∫fdα into an integral and a sum as∫ b
a
fdα =
∫ b
a
fα′dx+∑n
Jα(xn)f(xn)
where Jα(xn) = α(xn+)− α(xn−) for the ‘mass points’ xn of α.
MAT 5610: 98
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
RSI: Last Word
Theorem (Holder’s Inequality)Let f and g be in R(α) and let p, q > 0 be such that 1/p+ 1/q = 1.Then ∣∣∣∣∣
∫ b
a
fg dα
∣∣∣∣∣ ≤[∫ b
a
|f |pdα]1/p [∫ b
a
|g|qdα]1/q
For p = 2, this is the Cauchy-Bunyakovski-Schwarz inequality.
Theorem (Minkowski’s Inequality)Let p > 1 and let fp and gp be in R(α). Then
[∫ b
a
|f + g|p dα]1/p
≤[∫ b
a
|f |pdα]1/p
+
[∫ b
a
|g|pdα]1/p
MAT 5610: 99
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Exercises, III
Exercises1. Compute
1.1∫ 7/2
0
x d[x]
1.2∫ 2
0
x d[x]
2. Set α(x) = x+ bxc. Find the value of∫ 3
0
exdα.
3. Prove the “RS Integration is Bilinear theorem” (#86)
4. State and prove analogs of the First and Second Mean ValueTheorems for Integration for Riemann-Stieltjes Integrals.
5. “Write a homework problem suitable for this section and solve it.”
MAT 5610: 100
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude
MAT 5610: 101
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Gamma Function
Γ(x) =
� ∞
0
tx−1e−tdt
y = 1/Γ(x)
MAT 5610: 102
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Who? What? Where? When?
Leonhard Euler at 22 (in letters to Goldbach in 1729−30) briefly described the gamma function. Then Eu-ler published the details in De progressionibus tran-scendentibus seu quarum termini generales alge-braice dari nequeunt.† Euler merged an interpola-tion problem (a generating function for the triangularnumbers) with integration. Euler discovered that[(
2
1
)n1
n+ 1
] [(3
2
)n2
n+ 2
] [(4
3
)n3
n+ 3
]· · · = n!
and used interpolation via integration to extend thefunction to noninteger values. He called it C(x); laterLegendre named it Γ(x).
†On transcendental progressions whose general term cannot be expressedalgebraically
MAT 5610: 103
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Defining Gamma
Definition (Euler’s Gamma Integral (1730))
If x is not a negative integer, define Γ(x) =
∫ ∞0
tx−1e−tdt.
Examples
1. Γ(1) =
∫ ∞0
e−tdt. So Γ(1) = 1.
2. Γ(2) =
∫ ∞0
t e−tdt. So Γ(2) = − (t+ 1)e−t∣∣∣∞0
= 1.
3. Γ(x+ 1) =
∫ ∞0
tx e−tdt. So
Γ(x+ 1) = txe−t∣∣∣∣∞0
+ x
∫ ∞0
tx−1e−tdt = xΓ(x)
4. Γ(n+ 1) = nΓ(n) = n(n− 1)Γ(n− 1) = n(n− 1)(n− 2)Γ(n− 2)= n(n− 1)(n− 2)(n− 3)Γ(n− 3) = · · · = n!
MAT 5610: 104
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Gamma’s Rays
Theorem (Weierstrass’s Product Formula)
Γ(x) =e−γx
x
∞∏
n=1
(1 +
x
n
)−1
ex/n
where γ ≈ 0.5772156649 is the Euler-Mascheroni constant.
Theorem (Euler’s Reflection Formula)
Γ(x) · Γ(1− x) =π
sin(πx)
Corollary
Γ
(1
2
)= (−0.5)! =
√π and Γ
(3
2
)= 0.5 ! =
√π
2
MAT 5610: 105
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Gamma’s Mirror
Proof (Euler’s Reflection Formula).
1.1
Γ(x)= x eγx
∞∏
n=1
(1 +
x
n
)e−x/n
2.1
Γ(x)· 1
Γ(−x)= −x2eγxe−γx
∞∏
n=1
(1− x2
n2
)e−x/nex/n
= −x2∞∏
n=1
(1− x2
n2
)
3. Since Γ(1− x) = (−x)Γ(−x), then
1
Γ(x)· 1
Γ(−x)= x
∞∏
n=1
(1− x2
n2
)
4. Euler had previously shown the product on the right equal tosin(πx)/π (in his solution of the Basel Problem.)
MAT 5610: 106
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Almost Gamma
Theorem (Stirling’s Approximation)
Γ(x+ 1) ≈(xe
)x√2πx
Corollary
n! ≈ (n/e)n√
2πn
Theorem (Ramanujan’s Double Inequality)
[8x3 + 4x2 + x+
1
100
]1/6
<Γ(x+ 1)
(x/e)x√π<
[8x3 + 4x2 + x+
1
30
]1/6
MAT 5610: 107
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Project
Exercises
1. Show√
2
2Γ
(1
4
)Γ
(3
4
)= π.
2. Verify the identities√
3
2
3∏k=1
Γ
(k
3
)= π and
27
8
6∏k=1
Γ
(k
3
)= π2
Project (Spring Break Project for the Intrepid Student)1. Problem #11426, Am. Math. Monthly, Vol.116, April, 2009:
Γ(
114
)Γ(
914
)Γ(
1114
)Γ(
1314
)Γ(
514
)Γ(
314
) = 2
2. For some rational r(n) = p/q,n∏
k=1gcd(k,n)=1
Γ
(k
n
)=p
q· π
MAT 5610: 108
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude: An Application of Integration
MAT 5610: 109
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Fourier’s Problem
J B J Fourier had an interesting life. He was arrestedat times as a revolutionary, others as a counter-revolutionary. He taught at the Ecole Polytechnique,then served as Napolean’s scientific adviser in theinvasion/occupation of Egypt. He retuned to Franceand the Ecole in 1801. Then Napolean appointedhim as Prefect of the Department of Isere. (Fourierwas not happy, but couldn’t refuse.) He presentedOn the Propagation of Heat in Solid Bodies in 1807.His theory was controversial. However, in 1822, Fourier, was awardeda prize for the second version of this work. (In it, he introduced thenotation
∫ ba
.) Fourier, who wasn’t overly modest, wrote in his preface
The theory of heat will hereafter form one of the mostimportant branches of general physics.
MAT 5610: 110
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Defining Fourier’s SeriesThe main tool Fourier introduced was trigonometric series.
Theorem (Fourier Series of f )
f(x) =a0
2+
∞∑k=1
ak cos(kx) + bk sin(kx) −→Dirichlet
f(x+) + f(x−)
2
where
ak =1
π
∫ π
−πf(x) cos(kx) dx and bk =
1
π
∫ π
−πf(x) sin(kx) dx
The definition depends on the following identities. Suppose m,n ∈ N, then
1.∫ π
−πcos(mx) cos(nx) dx =
0 m 6= n
π m = n 6= 0
2π m = n = 0
2.∫ π
−πcos(mx) sin(nx) dx = 0
3.∫ π
−πsin(mx) sin(nx) dx =
0 m 6= n
π m = n 6= 0
0 m = n = 0
MAT 5610: 111
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Square Example
-4 -3 -2 -1 0 1 2 3 4
-1.2
-0.8
-0.4
0.4
0.8
1.2
-4 -3 -2 -1 0 1 2 3 4
-1.2
-0.8
-0.4
0.4
0.8
1.2
n = 10 n = 20
-4 -3 -2 -1 0 1 2 3 4
-1.2
-0.8
-0.4
0.4
0.8
1.2
-4 -3 -2 -1 0 1 2 3 4
-1.2
-0.8
-0.4
0.4
0.8
1.2
n = 30 n = 40
f(x) = sign(sin(2x)) =⇒ SN (x) =4
π
b(N+2)/4c∑k=1
sin(2(2k − 1)x)
2k − 1
MAT 5610: 112
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Selected Fourier Series
Function on [−π, π] Fourier Series
f(x) = x 2
∞∑k=1
(−1)k+1 sin(kx)
k
f(x) = |x| π
2− 4
π
∞∑k=1
cos((2k − 1)x)
(2k − 1)2
f(x) = x2 π2
3+ 4
∞∑k=1
(−1)kcos(kx)
k2
f(x) = sin2(x)1
2− 1
2cos(2x)
f(x) = | sin(x)| π
2− 4
π
∞∑k=1
cos(2kx)
4k2 − 1
f(x) =
2π − 4|x|
π2|x| < π/2
0 otherwise
1
2π+
8
π3
∞∑k=1
1− cos(kπ/2)
k2cos(kx)
MAT 5610: 113
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Selected Graphs
S10(f) for ...
-4 -3 -2 -1 0 1 2 3 4
-2.5
2.5
-4 -3 -2 -1 0 1 2 3 4
0.25
0.5
0.75
f(x) = |x| f(x) = triangle pulse
-4 -3 -2 -1 0 1 2 3 4
2.5
5
7.5
10
-4 -3 -2 -1 0 1 2 3 4
-2.5
2.5
f(x) = x2 f(x) = x
MAT 5610: 114
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Fourier Spectrum
Definition (Spectrum of a Fourier Series)The spectrum φ of a series is the amplitude of the harmonics. Thus
φ(k) =√a2k + b2k.
(Also called line spectrum, power spectrum, etc.)
The spectrum indicates the amount of “energy” or “power” of a givenharmonic in the Fourier series. (Think of an audio signal.)
Example
Suppose that f(x) = 2
∞∑
k=1
(−1)k+1 sin(kx)
k. Then φ(k) = 2/k.
The line spectrum is a plot of φ v k.
MAT 5610: 115
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Line Spectrum
ExampleConsider the square wave f(x) = sign(sin(2x)). Then
ak = 0 and bk =4
π·{
12k−1 k = 2 mod 4
0 otherwise
Whereupon f ’s line spectrum is
MAT 5610: 116
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Application
A line spectrum lets us compare instruments or assists with creatingelectronically generated music. Other applications include measuringflow and pressure in arteries, measuring biological morphology, etc.
Brown’s‡ spectrum analysis of the opening chord in the Beatle’s “AHard Day’s Night,” solved the ‘what instruments were played’ mystery.‡“Mathematics, Physics and A Hard Day’s Night,” Jason Brown, 2004.
MAT 5610: 117
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude
MAT 5610: 118
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Seriesly Fun
Definition• A series is a sequence of partial sums with the nth term given bySn =
∑nk=p ak for some fixed p ∈ Z
• A sequence {bn}n≥p can be written as a series Sn =∑nk=p ak by
setting an = bn − bn−1 and bp−1 = 0.
Examples
1.∞∑
k=1
1
k2=π2
6
2.∞∑
k=0
1
k!= e
3.∞∑k=1
1
k(k + 1)=1=
∞∑k=2
1
k(k − 1)
4.∞∑
k=1
(−1)k+1
k= ln(2)
5.∞∑
k=1
(−1)k+1 sin(k)
k=
arctan
(sin(1)
1 + cos(1)
)
MAT 5610: 119
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Harmonious Series
RECALL:
DefinitionA series
∑∞k=p ak converges to A ∈ R iff for every ε > 0 there is an
n∗ ∈ N s.t. whenever n ≥ n∗, then |Sn −A| < ε.
Proposition• If
∑ak converges, then ak → 0.
• If ak 6→ 0, then∑ak diverges.
Example (Nicole Oresme (c.1360))The harmonic series
∑1/k diverges.
Take segments that are greater than 1/2 to show S2n ≥ 1 + n/2.
MAT 5610: 120
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Critical Series
PropositionThe geometric series
∑a rk converges iff the constant |r| < 1, then∞∑
k=0
a rk =a
1− r .
Proof.1. If |r| ≥ 1, then a rk 6→ 0. Whence
∑a rk diverges.
2. Let |r| < 1 and n > 1. Then Sn − rSn = a− a rn+1. Hence
Sn =a(1− rn+1)
1− r
3. Since |r| < 1, then rn → 0. Thence Sn converges.
What about∑
sin(k)k?
MAT 5610: 121
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Geometric Sample
ExampleWrite 0.1011 as a rational number.
1. 0.1011 = 1011 · 10−4 + 1011 ·(10−4
)2+ 1011 ·
(10−4
)3+ · · ·
2. =
∞∑
k=1
1011 ·(10−4
)k
3. = 1011
(−1 +
∞∑
k=0
(10−4
)k)
4. = 1011
(−1 +
1
1− 10−4
)
5. = 1011
(−1 +
10000
9999
)= 1011
(1
9999
)=
337
3333
MAT 5610: 122
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Maple Sample
Maple Summing
How does Maple find that∞∑
k=1
(−1)k+1 sin(k)
k= arctan
(sin(1)
1 + cos(1)
)?
1. First, enter> infolevel[sum]:=5:
to raise the level of “user information” printed.
2. Now enter the summation
>∞∑
k=1
(−1)k+1
ksin(k)
and wait for the results.∗
∗Plan B.
MAT 5610: 123
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Cauchy and Friends
Theorem (Cauchy Criterion for Series)∑ak converges iff for each ε > 0 there is an n∗ ∈ N s.t. whenever
n ≥ m ≥ n∗, then
|Sn − Sm−1| = |am + am+1 + am+2 + · · ·+ an| < ε
Proof.X (Series are sequences; cf. Cauchy’s Criterion for Sequences.)
Corollary1. If ak = bk for k ≥ k∗, and
∑bk converges, then
∑ak converges.
2. If∑ |ak| converges, then
∑ak converges.
3. If∑ak diverges, then
∑ |ak| diverges
MAT 5610: 124
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Tests, I
Theorem (The Integral Test)If f : [1,∞)→ R is continuous, decreasing, and nonnegative, then
∞∑
k=1
f(k) converges iff∫ ∞
1
f(x) dx converges
Proof.Let ak = f(k). Since f is decr and cont, ak+1 ≤ f(x) ≤ ak for k ≥ 1.Whence
ak+1 ≤∫ k+1
k
f(x) dx ≤ ak
ThereforeSk+1 − a1 ≤
∫ k+1
1
f(x) dx ≤ Sk
The series bounds the integral, the integral bounds the series. Sinceall terms are nonneg, Bolzano-Weierstrass Thm gives the result.
MAT 5610: 125
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Integral Test Figure
∞∑k=2
ak ≤∫ ∞
1
f(x) dx
∫ ∞1
f(x) dx ≤∞∑k=1
ak
So∫ ∞
2
f(x) dx ≤∞∑
k=2
ak ≤∫ ∞
1
f(x) dx ≤∞∑
k=1
ak
MAT 5610: 126
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Tests, II
Theorem (The Comparison Test)Suppose
∑ak and
∑bk are two series with bk ≥ 0 for all k.
1. If |ak| ≤Mbk eventually (for some M ∈ R+) and∑bk converges,
then∑ |ak| and
∑ak both converge.
2. If mbk ≤ |ak| eventually (for some m ∈ R+) and∑bk diverges,
then∑ |ak| diverges.∗
Proof.1. Wolog |ak| ≤Mbk for all k. Since
∑bk converges, so does∑
Mbk, say to S. Then
0 ≤N∑
k=1
|ak| ≤N∑
k=1
Mbk ≤∞∑
k=1
Mbk = S
Whence∑ |ak| is a bounded, increasing series, thus convergent.
∗ ∑ ak may or may not diverge; e.g., ak = (−1)k+1/k and bk = 1/k.
MAT 5610: 127
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Useful Series Set
Proposition (p-Series∗)
The real series ζ(p) =
∞∑k=1
1
kpconverges iff p > 1.
Proof.X (Apply the integral test.)
Examples
1.∞∑k=1
1/k =∞
2.∞∑k=1
1/k2 = π2/6
3.∞∑k=1
1/k3 = ζ(3)
4.∞∑k=1
1/k4 = π4/90
5.∞∑k=1
1/k5 = ζ(5)
6.∞∑k=1
1/k6 = π6/945
∗See the Riemann Zeta function. Also $
MAT 5610: 128
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Tests, III
Theorem (The Limit Comparison Test)Suppose
∑ak and
∑bk are two positive-term series.
1. If limn→∞
anbn
is finite and∑bn converges, then
∑an converges.
2. If limn→∞
bnan
is finite and∑bn diverges, then
∑an diverges.
Proof.Set L = lim an/bn.
1. If L = 0, then ∃M ∈ R+ s.t. an ≤Mbn. Now compare.
2. If L > 0, then 12L bn ≤ an ≤ 3
2L bn eventually. Again compare.
Corollary
Suppose∑ak &
∑bk are positive-term series, and lim
n→∞
anbn
= L with
0 < L <∞. Then∑ak &
∑bk converge or diverge together.
MAT 5610: 129
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Fun Time
Exercises
1.∞∑k=1
(−1)kk
(k + 1) ek
2.∞∑n=1
102n
(2n− 1)!
3.∞∑j=1
3j
j3
4.∞∑i=1
(−1)i 23i
32i
5.∞∑m=1
(√
5− 1)m
m2 + 1
6.∞∑k=1
k!
kk
7.∞∑n=1
(−1)n−1
n2 + 1
8.∞∑j=2
(−1)j
j ln(j)
9.∞∑i=1
(−1)nn
n2 + 1
10.∞∑m=1
(−1)mm3
(m2 + 1)4/3
11. 13
+ 1·43·6 + 1·4·7
3·6·9 + · · ·
12. 29
+ 2·59·12
+ 2·5·89·12·15
+
13.∞∑k=1
k + 2
(k+1)√k+3
Challenge:∞∑k=1
(a+ kb)rk =a− (a− b)r
(1− r)2; |r| < 1
(cacdcadcac2c2cddcd)
MAT 5610: 130
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Tests, IV
Theorem (Ratio Comparison Test)Let
∑an &
∑bn be positive term series. If an+1/an ≤ bn+1/bn eventually,
and∑bn converges, then
∑an converges.
Theorem (D’Alembert’s Ratio Test)Let
∑an be a series with positive terms. If there is a constant α ∈ [0, 1) s.t.
an+1/an ≤ α eventually, then∑an converges. If an+1/an ≥ 1 eventually,
then∑an diverges.
Corollary (Cauchy’s Ratio Test)Let
∑an be a series with positive terms. Set r = limn→∞
an+1
an. Then:
• If r < 1, the series converges.
• If r > 1, the series diverges.
• If r = 1, the test fails.
MAT 5610: 131
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Proofs
Proof (Ratio Comparison Test).1. an/bn is decreasing eventually, thus bounded. Thus an ≤Mbn.2. Since
∑bn converges, then
∑an converges by comparison.
Proof (D’Alembert’s Ratio Test).1. α < 1 gives
∑αn converges.
2. Then an+1/an ≤ αn+1/αn = α implies that∑an converges.
Proof (Cauchy’s Ratio Test).X
MAT 5610: 132
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Tests, V
Theorem (Root Test)Let
∑an be a series with positive terms. If ∃α ∈ [0, 1) s.t. n
√an ≤ α
eventually, then∑an converges. If n
√an > 1 eventually, then
∑an diverges.
Theorem (Cauchy’s Root Test)Let
∑an be a series with positive terms. Set ρ = lim
n→∞n√an. Then:
• If ρ < 1, the series converges.
• If ρ > 1, the series diverges.
• If ρ = 1, the test fails.
Proof.1. n√an ≤ ρ < 1 implies an ≤ ρn < 1.
2. Since ρ < 1, then∑ρn converges. By comparison,
∑an converges.
MAT 5610: 133
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Tests, VI
Theorem (Raabe’s Test∗)
Let∑an be a series with lim
n→∞
∣∣∣an+1
an
∣∣∣ = 1. Set ρ = limn→∞
n[1−∣∣∣an+1
an
∣∣∣].
• If ρ < 1, the series diverges.• If ρ > 1, the series converges absolutely.• If ρ = 1, the test fails.
Example
Consider the series S = 1− 1
2−∞∑
k=2
1 · 3 · · · (2n− 3)
2n n!
1.∣∣∣∣an+1
an
∣∣∣∣ =2n− 1
2n+ 2
2. n[1−∣∣∣∣an+1
an
∣∣∣∣]
=3n
2n+ 2→ 3
2
∗Raabe’s Test (1832) is a special case of Kummer’s Test (1835)
MAT 5610: 134
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Challenge Prize Problem
Ready. Set. Go. . .What conditions are needed on α, β, and γ (in addition to γ 6= 0, −1,−2, . . . ) so that the hypergeometric function
2F1(α, β; γ; z) =
∞∑
k=0
(α)k(β)k(γ)k
· zk
k!
=
∞∑
k=0
α(α+ 1) · · · (α+ k − 1) · β(β + 1) · · · (β + k − 1)
γ(γ + 1) · · · (γ + k − 1)· z
k
k!
converges at z = 1.
(See Gauss’s Hypergeometric Theorem.)
MAT 5610: 135
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Converge? Absolutely!
Definition• A series of the form
∑(−1)kak or
∑(−1)k+1ak with ak ≥ 0 is an
alternating series.• If
∑ |ak| converges, then∑ak converges absolutely.
• If∑ |ak| diverges but
∑ak converges, then
∑ak converges
conditionally.
Theorem (Leibniz’s Alternating Series Test)Let an ≥ 0 for all n. If 1. an is eventually decreasing, and 2. an → 0,then
∑(−1)k+1ak converges. (N.B. |S − Sn| < an+1.)
Examples
1.∑
1k =∞;
∑ (−1)k+1
k = ln(2) 2.∑
1k2 = π2
6 ;∑ (−1)k+1
k2 = π2
12
Fun Time
MAT 5610: 136
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Rearranging the Terms
Theorem• If
∑ak converges to S absolutely and
∑bk is a rearrangement of
∑ak,
then∑bk converges to S absolutely.
• If∑ak converges to S conditionally and α ∈ [−∞,+∞], then there is a
rearrangement∑bk that converges to α.
• If∑ak converges to S conditionally and α, β ∈ R, then there is a
rearrangement∑bk that oscillates between α and β.
Example∞∑k=1
(−1)k+1
k= ln(2), conditionally convergent. Choose α ∈ R.
1. Add positive terms until 1 + 1/3 + · · ·+ 1/(2n1 − 1) first exceeds α
2. Add negative terms until 1 + · · ·+ 1/(2n1 − 1)− 1/2− · · · − 1/(2n2) firstis less than α.
3. Rinse and repeat.
MAT 5610: 137
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude
MAT 5610: 138
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Functioning Sequences & Series
Definition (Pointwise Convergence)A sequence of fcns {fn} converges pointwise to f on D ⊆ R iff foreach xo ∈ D, the sequence fn(x0)→ f(x0). In symbols:
fn→f ⇐⇒ (∀ε>0)(∀x0∈D)(∃n∗∈N)(∀n>n∗)[|fn(x0)− f(x0)| < ε
]
Examples
1. fn(x) = xn → f(x) =
{0 |x| < 1
1 x = 1on D = (−1, 1]. ((−1)n div.)
2. gn(x) = 1n cos(nx)→ 0 on D = R
3. hn(x) = sin(nx)→ 0 on D = {kπ | k ∈ Z} and div otherwise.
4. jn(x) = 2π arctan(nx)→ signum(x) on D = R
MAT 5610: 139
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Problems
ExamplesContinuity: Both xn and 2
π arctan(nx) are continuous fcns that havediscontinuous limits
limn→∞
limx→a
fn(x) 6= limx→a
limn→∞
fn(x)
Differentiability: fn(x) = 1n sin(nx)→ 0, but f ′n(x) = cos(nx) diverges
for all x 6= 0
d
dxlimn→∞
fn(x) 6= limn→∞
d
dxfn(x)
Integrability: nxn → 0 on [0, 1), but∫ 1
0nxn dx→ 1
∫ b
a
limn→∞
fn(x) dx 6= limn→∞
∫ b
a
fn(x) dx
MAT 5610: 140
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude: L2
Definition (L2-norm)For f , define the mean-square or L2-norm on [a, b] to be
‖f‖2 =
[∫ b
a
|f(x)|2 dx]1/2
Definition (Mean-Square Convergence)A seq {fn} converges to f in the mean on [a, b] iff lim
n→∞‖fn − f‖2 = 0.
Examples
1. Let fn(x)= xn on D=[0, 1]. Since ‖xn‖2 =√∫ 1
0x2ndx =1/
√2n+1→0,
thence xn → 0 in the mean on [0, 1].
2. Theorem. Riesz-Fischer (1907).The Fourier series of an L2 fcn f converges to f in the mean.
‡ Fourier Squares
MAT 5610: 141
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Uniform Convergence
Definition (Uniform Convergence)A sequence of fcns {fn} converges uniformly to f on D ⊆ R iff forevery ε > 0 there is an n∗ = n∗(ε) ∈ N such that for any x ∈ D, ifn > n∗, then |fn(x)− f(x)| < ε.
Examples
1. fn(x) =xn
1 + nxn→ 0 uniformly on [0, 1]
2. gn(x) = 1√n
cos(nx)→ 0 uniformly on R
3. hn(x) =1
n(x2 + 1)→ 0 uniformly on R (but
1
nx2 + 1doesn’t)
4. jn(x) = xn → f(x) =
{0 |x| < 1
1 x = 1pw, not uniformly, on [0, 1].
MAT 5610: 142
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Two Graphs
Uniform Convergence Nonuniform Convergence
MAT 5610: 143
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
How To...
Proposition
fn(x) =x
nx+ 1converges uniformly to f(x) = 0 on [0,∞).
Proof.1. If x = 0, then fn(0) ≡ 0. Hence |fn(0)− 0| < ε for any ε > 0.
2. If x > 0, then |fn(x)− 0| =∣∣∣∣ x
nx+ 1
∣∣∣∣ < ∣∣∣ xnx ∣∣∣ =1
n< ε for n suffic large.
Propositiongn(x)→0 pw, not unif on [0, 1], where gn(x) = 4n2x for 0 ≤ x < (2n)−1, andgn(x) = 4n(1−nx) for (2n)−1≤x≤n−1, and 0 otherwise.
Proof.Consider gn(1/(2n)) = 2n 6< ε for n sufficiently large.
MAT 5610: 144
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Outerlude
MAT 5610: 145
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Props
QueryWe know that a function continuous on a compact set is uniformlycontinuous. Does a sequence converging on a compact set convergeuniformly?
Theorem (Cauchy Criterion for Sequences)A sequence of fcns {fn} converges uniformly on D ⊆ R iff for eachε > 0 there is an n∗ s.t. for all x ∈ D, if n,m ≥ n∗, then|fn(x)− fm(x)| < ε.
Proof.1. Build an f by considering fn(x) for each x ∈ D2. 4 shows fn → f uniformly
MAT 5610: 146
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude: L∞
Definition (L∞-norm)For f , define the sup- or L∞-norm on [a, b] to be
‖f‖∞ = supx∈[a,b]
|f(x)|
��x�1 = 1��x�2 = 1 ��x�∞ = 1
TheoremA sequence fn converges to f in the sup norm on [a, b]; i.e., ‖fn − f‖∞ → 0iff fn converges to f uniformly on [a, b].
Proof.Observe that |fn(x)− f(x)| < ε for all x ∈ [a, b] iff ‖fn − f‖∞ < ε.
MAT 5610: 147
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude: L∞ Examples
Example
Consider fn(x) =x
1 + n2x2. (fn → 0 pw on [−1, 1])
L∞: fn → 0 in L∞[−1, 1].
• f ′n(x) =1− n2x2
(1 + n2x2)2 =⇒ f ′n(x) = 0 at x = ±1/n. Thus
−1
2n≤ fn(x) ≤ 1
2n
• So ‖fn − 0‖∞ = 12n→ 0 on [−1, 1] (fn → 0 unif on [−1, 1])
L2: fn → 0 in L2[−1, 1].
•√∫ 1
−1|fn(x)|2dx =
√arctan(n)
n3− 1
n2(n2 + 1)→ 0.
• Thus ‖fn − 0‖2 → 0 on [−1, 1]
MAT 5610: 148
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude II: Extra for Experts
Theorem (Dini’s Theorem (1878))Suppose {fn(x)} is a sequence of functions converging to f onD ⊆ R. If
1. each fn is continuous on D,2. f is continuous on D,3. fn → f monotonically, and4. D is compact (on R: compact = closed and bounded),
then fn converges to f uniformly on D.
1. fn(x) = 1 on 0 < x < 1/n and 0 otherwise converges to 0 onD = [0, 1] (fn is not continuous)
2. xn → U(x− 1) on [0, 1] (f = U is not continuous)3. “Marching triangles”→ 0 on [0, 1] (convergence is not monotonic)4. xn → 0 on [0, 1) ([0, 1) is not compact)
MAT 5610: 149
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Props to UniCon
TheoremIf a sequence of continuous functions {fn} converges uniformly on D,the limit function f is continuous on D.
Proof.The proof hinges on the inequalities
|f(x)− f(x0)| ≤ |f(x)−fn1(x)|︸ ︷︷ ︸n1>n∗
+ |fn1(x)−fn1(x0)|︸ ︷︷ ︸|x−x0|<δ
+ |fn1(x0)−f(x0)|︸ ︷︷ ︸n1>n∗
<ε
3+ε
3+ε
3= ε
for some n1 > n∗ and all x ∈ B(x0, δ).
Converse? No — the “marching triangles”.
MAT 5610: 150
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
More Props to UniCon
TheoremIf a sequence of continuous functions {fn} converges uniformly to fon D = [a, b], then
limn→∞
[∫ b
a
fn(x) dx
]=
∫ b
a
[limn→∞
fn(x)]dx.
I.e., if Fn =∫ bafn and F =
∫ baf , then Fn → F.
Proof.The proof hinges on the inequalities∣∣∣∣∣
∫ b
a
fn(x) dx−∫ b
a
f(x) dx
∣∣∣∣∣ =
∣∣∣∣∣
∫ b
a
[fn(x)−f(x)] dx
∣∣∣∣∣
≤∫ b
a
|fn(x)−f(x)| dx <∫ b
a
ε dx = O(ε)
MAT 5610: 151
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Yet Even More Props to UniCon
TheoremSuppose {fn} is a sequence of continuously∗ differentiable functionsconverging pointwise to f on [a, b]. If the sequence of derivatives {f ′n}converges uniformly on [a, b], then f is continuously differentiable and• fn → f uniformly on [a, b], and• f ′n → f ′ uniformly on [a, b].
Proof.The proof is based on three ideas:
1.∫ xaf ′n(t) dt = fn(x)− fn(a)
2. f(x)−f(a)= limn→∞
[fn(x)−fn(a)] = limn→∞
∫ xaf ′n(t) =
∫ xa
[limn→∞
f ′n(t)]
3. f ′(x) = ddx
∫ xaf ′(t) dt
∗‘Continuously’ can be relaxed. ‘Pointwise’ can be relaxed to a single point.
MAT 5610: 152
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
A Sample Case
Exercises
1.{
n√
sin(x)}
converges uniformly on [ε, 12π] for ε > 0, but pw on (0, 1
2π].
2. Determine limn→∞
∫ a
−ae−nx
2
dx where a > 0.
3. Set Sn(x) = nxe−nx2
. What does Sn converge to? Show
limn→∞
∫ a
0
Sn(x) dx 6=∫ a
0
limn→∞
Sn(x) dx
4. Let fn(x) =x
1− nx2for |x| ≤ 1.
4.1 What is f(x) = limn→∞
fn(x)?
4.2 Is the convergence uniform?
4.3 Is f ′(x) = limn→∞
f ′n(x)? Explain.
MAT 5610: 153
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Series
Definition
Function Series: The sequence of partial sums Sn(x) =n∑k=1
fk(x)
Series Conv PW: The series converges pointwise on D if {Sn} convergespointwise on D
Series Conv Unif: The series converges uniformly on D if {Sn} convergesuniformly on D
Series Conv Abs: The series converges absolutely on D if {Sn} convergesabsolutely on D
Series Representation: A function f has a series representation on Diff ∃Sn → f on D
TheoremIf a series
∑k fk(x) of continuous functions converges uniformly to S(x),
then S is continuous.
MAT 5610: 154
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Samples
Exercise
1. S(x) =
∞∑
k=1
x2
(1 + x2)k
• S(0) = 0
• 0 <1
(1 + x2)< 1 for x 6= 0
• Set a = x2 and r = 1/(1 + x2), then use∞∑k=0
ark = a/(1− r).
Whence for x 6= 0,
S(x) =∞∑k=1
x2
(1 + x2)k=
x2
1− 11+x2
− x2 = 1
• Thus S(x) =
{1 x 6= 0
0 x = 0.
• Since S is not continuous, but each fn is, the convergence ispointwise and absolutely.
MAT 5610: 155
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Convergence Modes
Pointwise convergence
Absoluteconvergence
Uniformconvergence
Cesàroconvergence
Cesaro Convergence: Let Sn(x) =n∑k=0
fn(x) be a series. Then
SnCesaro−→ S(x) iff
1
n+ 1(S0 + S1 + · · ·+ Sn)→ S
E.g.,∞∑n=0
(−1)nCesaro−→ 1
2
MAT 5610: 156
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Two Big Theorems
Theorem (Cauchy Criterion)The series S(x) =
∑fn(x) converges uniformly on D iff for any ε > 0
there is an n∗ ∈ N s.t. whenever m,n ≥ n∗, then for all x ∈ D,
|Sn(x)− Sm(x)| =∣∣∣∣∣
n∑
k=m+1
fn(x)
∣∣∣∣∣ < ε
Theorem (Weierstrass M -Test)Suppose that |fn(x)| ≤Mn for all x ∈ D and for each n. If
∑Mn
converges, then∑fn(x) converges uniformly and absolutely on D.
Example
What can be said about∞∑
k=1
(−1)k+1 sin(k4x)
k2?
MAT 5610: 157
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
The Example
-5 -4 -3 -2 -1 0 1 2 3 4 5
-1.2
-0.8
-0.4
0.4
0.8
1.2
-5 -4 -3 -2 -1 0 1 2 3 4 5
-1.2
-0.8
-0.4
0.4
0.8
1.2
n = 25
-5 -4 -3 -2 -1 0 1 2 3 4 5
-16
-8
8
16
-5 -4 -3 -2 -1 0 1 2 3 4 5
-16
-8
8
16
d
dx
∞∑
k=1
(−1)k+1 sin(k4x)
k26=∞∑
k=1
(−1)k+1 d
dx
sin(k4x)
k2
MAT 5610: 158
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Integrating Series
Theorem
If Sn(x) =n∑fn(x) is integrable on [a, b] for each n ∈ N and Sn → S
uniformly, then ∫ b
a
[n∑
k=0
fn(x)
]dx =
n∑
k=0
[∫ b
a
fn(x) dx
]
Example
1
1 + t2=
∞∑
k=0
(−1)kt2k on |t| < 1 and uniformly on |t| ≤ ρ < 1
Whence∫ x
0
1
1 + t2dt =
∞∑
k=0
(−1)k∫ x
0
t2kdt
Wherefore arctan(x) =
∞∑
k=0
(−1)kx2k+1
2k + 1for |x| ≤ ρ < 1
MAT 5610: 159
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Arctan? < 1?
arctan(x) =
∞∑
k=0
(−1)kx2k+1
2k + 1for |x| ≤ ρ < 1︸ ︷︷ ︸
Why?
x
y
zf(z) = |arctan(z)|
f(z) = arctan(z) has poles at z = ±i
MAT 5610: 160
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Differentiating Series
Theorem
Suppose Sn(x) =n∑fn(x) is a series of continuously differentiable
functions converging pointwise to S(x) on [a, b]. If the series of
derivativesn∑f ′n(x) converges uniformly on [a, b], then
• S(x) is continuously differentiable,•∑fn(x)→ S(x) uniformly on [a, b], and
•∑f ′n(x)→ S′(x) uniformly on [a, b].
Example ∑xn =
1
1− x =⇒∑
nxn−1 =1
(1− x)2=⇒∑
n(n− 1)xn−2 =2
(1− x)3=⇒
∑n(n− 1)(n− 2)xn−3 =
3!
(1− x)4
MAT 5610: 161
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Fun Day
ProblemI. Where does
∞∑
n=1
n(x− 1)n
2n(3n− 1)
converge?
II. Let a > 0. Investigate
d
dx
∞∑
n=1
xn
n2an?=
∞∑
n=1
d
dx
xn
n2an
III. Set fn(x) = nxe−nx2
andf(x) =
∑n fn(x) on [0, 1].
Investigate∫ 1
0
f(x) dx?=
∞∑
n=0
∫ 1
0
fn(x) dx
IV. Bessel’s function is
Jp(x) =
∞∑
n=0
(−1)n(x/2)2n+p
n! (n+ p)!.
Show that y = Jp(x) solves
x2y′′ + xy′ + (x2 − p2)y = 0.
MAT 5610: 162
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Fun Day, II
ProblemV. Is∞∑
n=1
(−1)(n+1)/2 · 1− (−1)n
2
Cesaro-convergent?
VI. Is∞∑
n=1
(−1)n · n− 1
n+ 1
Cesaro-convergent?
VII. Calculate1
f(x)when
f(x) =
∞∑
n=0
xn.
VIII. Calculate1
g(x)when
g(x) =
∞∑
n=0
xn
n!.
MAT 5610: 163
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Very Powerful Series
Definition (Power Series)
• A power series is a series of the form∞∑
k=0
ak(x− a)k.
• The radius of convergence is the largest value R ∈ [0,∞] s.t. thepower series converges for all |x− a| < R.
TheoremSuppose
∑k ak(x− a)k converges for x∗ s.t. |x∗ − a| = r0. Then∑
k ak(x− a)k converges absolutely for all x s.t. |x− a| < r0.
Proof.Based on the inequalities |(x− a)/r0| < 1 and
∣∣ak(x− a)k∣∣ ≤
∣∣akrk0∣∣ ·∣∣∣∣x− ar0
∣∣∣∣k
≤M ·∣∣∣∣x− ar0
∣∣∣∣k
MAT 5610: 164
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
More Power
CorollaryIf∑k ak(x− a)k diverges for x∗, then
∑k ak(x− a)k diverges for all x
s.t. |x− a| ≥ |x∗ − a|.
TheoremLet R be the radius of conv of
∑k ak(x− a)k. Then
∑k ak(x− a)k
1. converges absolutely on |x− a| < R,2. converges uniformly on |x− a| ≤ ρ < R, and3. diverges on |x− a| > R.
• If∑k ak(R)k conv, then
∑k ak(x− a)k conv uniformly on
−R < −ρ ≤ |x− a| ≤ R.
• If∑k ak(−R)k conv, then
∑k ak(x− a)k conv uniformly on
−R ≤ |x− a| ≤ ρ < R.
MAT 5610: 165
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
More Power
Examples (“Try not. Do, or do not. There is no try.”)
1. sin(x) =
∞∑
k=0
(−1)kx2k+1
(2k + 1)!
2. f(x) =
∞∑
k=0
k!xk
3. ln(1 + x) =
∞∑
k=1
(−1)k+1xk
k
4. ln
[√1 + x
1− x
]=
∞∑
k=1
x2k−1
2k − 1
5. g(x) =
∞∑
k=1
ln(k + 1)√k + 1
(x− 5)k
6.xex
(ex − x)2=
∞∑
k=1
k xke−kx
7.1√
1 + x= 1− 1
2x+
1 · 32 · 4 x
2 − 1 · 3 · 52 · 4 · 6 x
3 +1 · 3 · 5 · 72 · 4 · 6 · 8 x
4 − . . .
8. ln
[x− 3
x− 4
]=
1
x− 3+
1
2(x− 3)2+
1
3(x− 3)3+
1
4(x− 3)4+ . . .
MAT 5610: 166
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
More Power Solutions, I
Examples
7.1√
1 + x= 1− 1
2x+
1 · 32 · 4 x
2 − 1 · 3 · 52 · 4 · 6 x
3 +1 · 3 · 5 · 72 · 4 · 6 · 8 x
4 − . . .
a.1 · 3 · 52 · 4 · 6 =
1 · 2 · 3 · 4 · 5 · 6(2 · 4 · 6)2
=6!
26 · (3!)2=
6!
43 · (3!)2(n = 3)
b.1 · 3 · 5 · 72 · 4 · 6 · 8 =
8!
28 · (4!)2=
8!
44 · (4!)2(n = 4)
c. an =(−1)n(2n)!
4n(n!)2=
(−1)n
4n·
(2n
n
)
d. rn(x) =
∣∣∣∣ (2(n+ 1))!xn+1
4n+1((n+ 1)!)2· 4n(n!)2
(2n)!xn
∣∣∣∣ =2n+ 1
2n+ 2|x| −→
n→∞|x|
Whence the RoC is 1; i.e., the series converges on |x| < 1.
Testing the endpoints shows the series converges at x = 1 anddiverges at x = −1. (The recursive form of an is easier.)
MAT 5610: 167
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
More Power Solutions, II
Examples
8. ln
[x− 3
x− 4
]=
1
x− 3+
1
2(x− 3)2+
1
3(x− 3)3+
1
4(x− 3)4+ . . .
a. Let the series be S. Substitute y = 1/(x− 3). Then
S = y +1
2y2 +
1
3y3 +
1
4y4 + . . .
b. Thus an = 1/n.
c. So rn =
∣∣∣∣yn+1/(n+ 1)
yn/n
∣∣∣∣ =n
n+ 1|y| −→
n→∞|y|
d. S converges for |y| < 1. That is, for1
|x− 3| < 1. Or |x− 3| > 1.
Testing the endpoints shows S(4) diverges and S(2) converges.
MAT 5610: 168
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Cauchy–Hadamard Theorem
Theorem (Cauchy (1821)–Hadamard (1888) Theorem)
The radius of convergence R of the power series∞∑
k=0
ak(x− a)k is
given by 1
R= lim sup
n→∞n√|an|
Proof.Wolog a = 0. Since
lim sup n√|anxn| = |x| lim sup n
√|an| =
|x|R
the result follows from Cauchy’s root test.
MAT 5610: 169
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Differentiate and Integrate
Theorem
If the power series∞∑
k=0
ak(x− a)k has RoC R ∈ (0,∞], the power
series∞∑
k=1
k ak(x− a)k−1 and∞∑
k=0
akk + 1
(x− a)k+1
have the same radius of convergence R.
Proof.L TTRAA E!
NB: Differentiating may lose convergence at the endpoints; e.g.,∑
1n x
n
MAT 5610: 170
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Unicity
Theorem (Unicity of Power Series Representation)Suppose f(x) =
∑n an(x− a)n and f(x) =
∑n bn(x− a)n on some
interval |x− a| < R where R > 0. Then an = bn for all n ∈ N.
Proof.Setting x = a shows that a0 = b0. Differentiate both series, then setx = a again. Now we have a1 = b1. (Do it again!) Induction showsan = bn for all n. Actually, we can show an = f (n)(a)/n!.
Examplesin(x) = x− 1
6x3 + 1
120x5 − 1
5040x7 + 1
362880x9 + . . .
sin(x) = −(x− π) + 16 (x− π)
3 − 1120 (x− π)
5+ 1
5040 (x− π)7 − . . .
a0 = b0 =∑∞k=1
(−1)k+1
(2k−1)! π2k−1, a1 = b1 =
∑∞k=0
(−1)k
(2k)! π2k, . . .
MAT 5610: 171
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Unicity: Caveat Emptor
-5 -4 -3 -2 -1 0 1 2 3 4 5
-3
-2
-1
1
2
3
T11(x − 1)
T11(x)
f(x) = tan(sin(x))
MAT 5610: 172
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Unicity: Caveat Emptor, II
f(x) = tan(sin(x))
T11,0(x) = x+ 0.1667x3 − 0.0250x5 − 0.0212x7 − 0.0030x9 + 0.0011x11
converging on (−1.87, 1.87)
T11,1(x) = 1.119 + 1.217(x− 1)− 0.2122(x− 1)2 − 0.7861(x− 1)3
− 0.6239(x− 1)4 + 0.0625(x− 1)5 + 0.4961(x− 1)6
+ 0.3716(x− 1)7 − 0.0054(x− 1)8 − 0.3161(x− 1)9
− 0.2243(x− 1)10 + 0.0042(x− 1)11
= −0.0910 + 2.045x− 5.126x2 + 15.50x3 − 31.53x4
+ 44.06x5 − 44.75x6 + 30.28x7 − 13.97x8 + 4.241x9
− 0.6887x10 + 0.0435x11
converging on (−0.17, 2.17)
MAT 5610: 173
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Taylor Series
Theorem (Taylor Series∗)Suppose f(x) =
∑∞k=0 ak(x− a)k converges with RoC R > 0. Then
• f ∈ C∞(|x− a| < R)
• f (n)(a) = ann! for n = 1, 2, . . .
Example (Bad example! No Biscuit! — Cauchy (1823))
Consider f(x) = e−1/x2 with f(0) = 0.
Since f (n)(0) = 0 for all n, we have the Maclaurin series for f is 0.
Example (Good example! Biscuit!)Consider f(x) = sin(x2).
Since sin(x)=
∞∑k=1
(−1)k+1 x2k−1
(2k − 1)!, then f(x)=
∞∑k=1
(−1)k+1 x2(2k−1)
(2k − 1)!.
‡Taylor (1712). Earlier: Gregory (1671), Leibniz (1670s), Newton (1691), J Bernoulli(1694), de Moivre (1708). Remainder formulas: Cauchy (1822), Lagrange (1772)
MAT 5610: 174
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Taylor Series Convergence
Taylor ConvergenceIF
1. f (n)(a) has a “reasonable” form, and2. one of the remainders
Lagrange’s form Rn(x) = f(n+1)(c)(n+1)! (x− a)n+1
Cauchy’s form Rn(x) = f(n+1)(c)n! (x− c)n(x− a)
Integral form Rn(x) = 1n!
∫ xaf (n+1)(t)(x− t)n dt
etc., is “reasonably” easy to find the limit of as n→∞,THEN we can determine a Taylor series and its radius ofconvergence.
MAT 5610: 175
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Weierstrass’s Two Stone Caps
Theorem (Weierstrass (1872))Let
f(x) =
∞∑
n=1
bn cos(anx).
Then1. f converges uniformly for b < 1, and2. f is nowhere differentiable when ab > 1 + 3
2 π.
Theorem (Weierstrass (1885))Let f ∈ C([a, b]). For every ε > 0 there is a polynomial pε(x) s.t.
|f(x)− pε(x)| < ε
for all x ∈ [a, b]. This is also written as ‖f − pε‖∞ < ε.
A continuous function f may not have a derivative anywhere, butis still within ε of an infinitely differentiable function everywhere.
MAT 5610: 176
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Interlude: Algebra for Analysis
Definition (The Field of Rational Functions on R)The field of rational functions on R is
R(x) =
{p(x)
q(x)
∣∣∣∣ p, q ∈ R[x] and q 6≡ 0
}.
with the usual addition and multiplication.
Definition (Derivation)An operator D on R(x) is a derivation if and only if for any c ∈ R andany α and β ∈ R(x), we have
1. D(x) = 1,2. D(c α) = cD(α) for any constant c,3. D(α+ β) = D(α) +D(β),4. D(α · β) = D(α) · β + α ·D(β) (the Leibniz rule).
MAT 5610: 177
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Properties of D
TheoremLet c ∈ R and n ∈ N. Then
1. D(1) = 0
2. D(c) = 0
3. D(xn) = nxn−1
4. D(x−n) = −nx−n−1
Proof.1. Apply the Leibniz rule to 1 = 1 · 12. Use linearity on c = c · 13. Apply the Leibniz rule to xk+1 = xk · x and induct4. Apply the Leibniz rule to xn · x−n = 1
MAT 5610: 178
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Chains
Definition (The Chain Rule)For α and β ∈ R(x), define
D(α ◦ β) = (D(α) ◦ β) ·D(β)
TheoremLet p, q ∈ N and α, β ∈ R(x) with β 6≡ 0. Then
1. D(x1/q) = (1/q)x1/q−1
2. D(xp/q) = (p/q)xp/q−1
3. D (α/β) = [D(α) · β − α ·D(β)]/β2
Proof.1. Apply the chain rule to (x1/q)q = x
2. Modified ditto
3. Look at (α/β) · β = α
MAT 5610: 179
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Two Extensions to R(x)
Definition• Define the function L by the relation: for any nonzero α ∈ R(x),
D(L ◦ α) =D(α)
α
• Define the function E by the relation: for any nonzero α ∈ R(x),
D(E ◦ α) = E(α) ·D(α)
Exercise1. Determine D(L ◦ α) when α(x) = x.2. Determine D(E ◦ β) when β(x) = x.
DefinitionFor any γ ∈ R(x)[L,E], define
∫γ = Γ iff D(Γ) = γ.
MAT 5610: 180
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
J-Lio’s Result
Theorem (Liouville’s Principle)If a function f has an elementary integral, then the integral has theform ∫
f = F0 +
n∑
k=1
ckL(Fk)
where the ck ∈ R and the Fk ∈ R(x). The function L is the logarithmdefined above.
Problem
• Does∫xx + xx ln(x)
?=
∫xx +
∫xx ln(x)? No!
Read Risch’s two papers:• “The Problem of Integration in Finite Terms,” Robert H. Risch, Trans
AMS, Vol. 139 (May, 1969), pp. 167–189.• “Algebraic Properties of the Elementary Functions of Analysis,” Robert
H. Risch, Am J Math, Vol. 101, No. 4 (Aug., 1979), pp. 743–759.
MAT 5610: 181
Introduction Review Derivative Integral Riemann-Stieltjes Integral Sequences & Series
Bad Function! No Biscuit!
TheoremThere is no elementary antiderivative for f(x) = ex
2
.
Proof.
Suppose F =
∫ex
2
. We know D(ex2
) = ex2
2x. Applying Liouville’s principle
implies F = α(x) ex2
with α ∈ R(x). Thus F ′ = α′ex2
+ α 2x ex2
= ex2
. Soα′ + 2xα = 1. Comparing degrees shows α can’t be a poly, and so must be arational function. Let α = p(x)/q(x). Then q must have a root (possiblycomplex), so write α = h(x)/(x− z0)m where h(z0) 6= 0 is a rational function.Substitute into the DE α′ + 2xα = 1:[
h′(x)
(x− z0)m−m h(x)
(x− z0)m+1
]+ 2x
h(x)
(x− z0)m= 1
Taking the limit of both sides as x→ z0 yields the contradiction∞ = 1.Whence F cannot exist in R(x, L,E).