35007 - Real Analysis


Sequences Successione 数列

Triangle inequality Disuguaglianza triangolare 三角不等式

\(\forall a,b \in \mathbb{R}, |a+b| \leq |a|+|b| \)

Proof

\( |a+b|^2 = (a+b)^2 = a^2 +2ab +b^2\)

\( (|a|+|b|)^2 = a^2 +2|a||b| + b^2\)

\( \implies |a+b| \leq |a|+|b| \)

Cauchy-Schwarz inequality

\( ( \sum^{k}_{n=1} a_n b_n )^{2} \leq ( \sum^{k}_{n=1} a_n^2 )( \sum^{k}_{n=1} b_n^2 ) \)

Cartesian product

Creating a new set from two sets by mapping each element in \(A\) to all elements in \(B\) by an ordered pair.

\( A \times B = \{ (a,b) : a \in A , b \in B \} \)

\( |A \times B| = |A||B| \)

Bounded set

\( S \text{ is bounded } \iff \exists M \gt 0 (\forall s \in S ( |s| \leq M))\)

Upper bound

\( M \text{ is an upper bound of } S \iff \forall s \in S( M \geq s) \)

Lower bound

\( M \text{ is a lower bound of } S \iff \forall s \in S( M \leq s) \)

Maximum function

\( \max (U) \) represents the largest element in \(U\)

Minimum function

\( \min (U) \) represents the smallest element in \(U\)

Supremum Supremo 上限

The smallest upper bound of a set in a partially ordered space

\( \sup (U) = \min \{ M : \forall u \in U, M \geq u \} \)

Infimum Infima 下限

The largest lower bound of a set in a partially ordered space

\( \inf (U) = \max \{ m : \forall u \in U, m \leq u \} \)

Archimedian property

Property of a space such that there does not exist a value that is infinitely large or small (there is always a number less than and greater than each number). This is an axiom for the real numbers.

\( \forall M \in \mathbb{R} ,\exists N \in \mathbb{N} : M \lt N\varepsilon \)

Least-upper-bound property

Property of a space such that all sets formed from that space bounded above have a supremum. This is an axiom for the real numbers.

\(\forall S \subset \mathbb{R} \setminus \emptyset : S \text{ is bounded above}, \exists r \in \mathbb{R} : r = \sup (S)\)

Sequence Successione 数列

Ordered, enumerated, countable collection of objects that allows for repetition

Formally it is defined as a function \( (x_n) : I \to X\)

Notation

Brackets enclose either the terms of the sequence or a symbol a subscript for indexing.

\( (1,1,2,3,5,8,13) \)

\( (F_n)^{6}_{n=0} \)

\( (F_n)^{\infty}_{n=0} \)

\( (F_n)_{n \in \mathbb{N}} \)

\( (1,1,2,3,5,8,13,\cdot) \)

Or for short hand, \( (F_n) \)

Monotone sequence

\( (x_n) \text{ is monotone increasing } \iff \forall n( x_{n+1} \geq x_{n}) \)

\( (x_n) \text{ is monotone decreasing } \iff \forall n( x_{n+1} \leq x_{n}) \)

Subsequence

Sequence formed from a countably infinite subset of terms from another sequence. Notationally, this is represented as the original sequence with the subscript being some 'monotone increasing indexing sequence' that jumps only to the terms that this subsequence includes.

Sequence that has some of its indexes skipped.

\( (x_{n_{k}}) \)

Epsilon variable

\(\varepsilon\) is a conventional symbol used to represent a variable with the following conditions:

Essentially, this is a variable that can be made arbitrarily close to 0, emulating infinitesimals. This will be necessary for defining limits

Convergence Convergenza 収束

\( x_n \text{ is convergent } \iff \exists L : \lim_{n \to \infty} x_{n} = L \)

Sequence property such that a sequence term can be chosen such that all subsequent terms are arbitrarily close to \(L\) (the sequence gets closer and closer to \(L\))

Epsilon-delta definition

\(\lim_{n \to \infty} x_n = L \iff (\forall \varepsilon ( \exists N : n\geq N \implies |x_n - L| \lt \varepsilon))\)

Topological definition

\(\lim_{n \to \infty} x_n = L \iff (\forall U : L \in U \subset \mathbb{R}, \exists N : n\geq N \implies x_{n} \in U))\)

Converges iff for any neigborhood of \(L\) there is a sequence term such that all subsequent terms are in that neighborhood

Theorems

\(\lim_{n \to \infty} x_n = x \land \lim_{n \to \infty} y_n = y \implies\)

\( a_n \text{ is convergent } \iff \exists! L : \lim_{n \to \infty} x_{n} = L \)

\(a_n \text{ is convergent } \implies a_{n} \text{ is bounded}\)

Proofs

By convergence, there is some \(N\) such that \( n \geq N \implies | x_n - L| \lt 1\)

One can construct \( M = \max \{x_1 , x_2, ..., x_{N}, |L|+1\} \implies |x_n| \lt M \implies x_n \text{ is bounded }\)

Bolzano-Weierstrass Theorem (BWT)

Theorem asserting that boundedness of a sequence ensures the existence of a convergent subsequence

\( x_{n} \text{ is bounded } \implies \exists x_{n_{k}} : x_{n_{k}} \text{ is convergent }\)

Proof

\(a_{n} \text{ is bounded } \implies \exists a_{b_{n}} : a_{b_{n}} \text{ is monotone } \)

\( a_{b_{n}} \text{ is monotone and bounded } \implies a_{b_{n}} \text{ is convergent}\)

Compactness

Property of a space such that all sequences have convergent subsequences in the space. In the case of subsets of Euclidean space \(\mathbb{R}^n\), this is characterized by BWT, however in the set of continuous functions \(C\), this is characterized by AAT (see Lebesgue Integration and Fourier Analysis)

\(S \text{ is compact } \iff \forall (x_n) \in S [ \exists (x_{n_{k}}) ( (x_{n_{k}}) \text{ is convergent })]\)

Cauchy criterion

\( (x_{n}) \text{ is Cauchy} \iff (\forall \varepsilon \gt 0( \exists N : n,m\geq N \implies |x_n - x_m| \lt \varepsilon))\)

Cauchy sequences are an alternative definition for convergent sequences that allow proof of convergence without knowing the actual limit

Theorems

\(x_{n} \text{ is Cauchy} \iff \lim_{n \to \infty} x_n = L\)

Completeness

Property of a space such that all Cauchy sequences have a limit within the space, that is, the space includes all limit points

\(S \text{ is complete} \iff forall \{x_n\} \subseteq S : \{x_n\} \text{ is Cauchy }, \lim_{n \to \infty} \{x_{n}\} \in S\)

Squeeze theorem

\( \lim_{n \to \infty} a_{n} = L \land \lim_{n \to \infty} c_{n} = L \land b_{n} \in [a_{n},c_{n}] \implies \lim_{n \to \infty} b_{n} = L\)

Floor function

\( \lfloor x \rfloor = \sup \{ m \in \mathbb{Z} : m \leq x \} \)

\( \lceil x \rceil = \inf \{ m \in \mathbb{Z} : m \geq x \} \)

Dedekind cut

Way of defining any real number by partitioning \(\mathbb{Q}\) by some inequality, for instance \(\sqrt{2}\) can be represented by the dedekind cut \((A,B) : A=\{a \in \mathbb{Q} : a^2 \lt 2 \lor a \lt 0\},B=\{ b \in \mathbb{Q} : b^2 \geq 2 \land b \geq 0 \} \)

Series Serie 級数

Series Serie 級数

Sequences that build by summation on the previous terms.

\( S_n \text{ is a series } \iff S_n = \sum^{n}_{k=1} a_k\)

Absolutely convergent series Serie convergente assoluta 絶対収束級数

A series is \(\sum_{n=1}^{\infty} |a_{n}| = S\)

\(\sum_{k=1}^{n} a_{k} \text{ is absolutely convergent } \iff \sum_{k=1}^{\infty} |a_{k}| \lt \infty \)

Conditionally convergent series

\(S_n \text{ is conditionally convergent } \iff S_n \text{ is convergent} \land \neg( S_n \text{ is conditionally convergent})\)

Harmonic series Serie armonica 調和級数

\( H_{n} = \sum_{k=1}^{n} \frac{1}{k}\)

Proposition

\( \sum^{n}_{k=1} H_{k} = (n+1)H_n -n\)

Geometric series Serie geometrica 幾何級数

\( S_{n} = \sum_{k=1}^{n} ar^{k-1} \)

Closed form

\( S_{n} = \frac{a(r^{n+1}-1)}{r-1} \)

\( |r| \lt 1 \implies \lim_{n \to \infty }S_{n} = \frac{a}{1-r}\)

\(\forall r \in \mathbb{C} [ |r| \lt 1 \implies \lim_{n \to \infty }S_{n} = \frac{a}{1-r} ] \)

Basel problem

\( \sum_{n=1}^{\infty} \frac{1}{n^2} = \frac{\pi^2}{6} \)

Euler's constant (\( \gamma \))

Constant representing the converging difference between the harmonic series and natural logarithm

\( \gamma = \lim_{n \to \infty} H_{n} -\ln (n) \)

Partial sum test

Given a series \(S = \sum_{k=1}^{\infty} a_{k}\), if you can form it into a sequence of partial sums \(x_{n} = \sum_{k=1}^{n} a_{k} \), and that sequence converges

Comparison test

If \(\sum_{n=1}^{\infty} a_{n}\) is convergent and there exists some term \(N\) such that any \(n \geq N\) means that \(a_n \geq b_n\), then \( \sum_{n=1}^{\infty} b_{n}\) is convergent too

\( \sum^{\infty}_{k=1} a_{k} \lt \infty \land \exists N : \forall n \geq N, a_n \geq b_n \implies \sum^{\infty}_{k=1} b_{k} \lt \infty\)

Limit comparison test

If the ratio of two series with strictly positive terms is convergent to some non zero constant, either both series converge or diverge

\( \lim_{n \to \infty} \frac{a_{n}}{b_{n}} = L \land L \neq 0 \implies \text{both convergent} \lor \text{both divergent}\)

Ratio test Prova del rapporto

Proving that the sequence converges by proving that the ratio of an iteration and the previous iteration is between 1 and -1 like such:

\(S = \sum_{n=0}^{\infty} a_n \text{ is convergent}\)

\(\lim_{n\to \infty} |\frac{a_{n+1}}{a_{n}}| = L\)

Proof

To prove for \(L \lt 1\), comparison to the geometric series will be made. \(L \lt 1 \implies \exists r: r \in (L,1) \). Then \(r - L\) can be used as an infinitesimal such that \(\exists N : \forall n \geq N, |\frac{a_{n+1}}{a_{n}} - L| \lt r-L\). Now by making double bounds to remove the absolute value, \( L-r \lt \frac{a_{n+1}}{a_{n}} - L \lt r-L \implies a_{n+1} \lt ra_{n}\). One can easily inductively show that \(a_{n} \lt r^{k}a_{n-k} \). Therefore there exists some \(N=n-k\) such that \(\forall n \geq N, a_{n} \leq \frac{a_{N}}{r^N} r^n \). This implies that each \(a_{n}\) is less than some geometric series term and since \(r \lt 1\), by the comparison test convergence is implied.

For the case \(L \gt 1\), consider instead \(L \gt 1 \implies \exists r: r \in (1,L) \), then similarly \( r-L \lt \frac{a_{n+1}}{a_{n}} - L \lt L-r \implies ra_{n} \lt a_{n+1} \) and generally \(a_{n} \gt r^{k}a_{n-k} \). The proof procedes similarly, however note that now \(r \gt 1\), which violates the radius of convergence for geometric series and therefore the series is divergent.

In essence, introduce a geometric term \(r \lt 1\), use this to form an infinitesimal and then make comparison to the geometric series

Leibniz test

\(S = \sum_{n=0}^{\infty} (-1)^{n}a_{n}\) converges when:

The intuition behind this is that if the sequence in the series is monotone decreasing and approaching zero, it is oscillating (due to the nature of being alternate) and the oscillation is shrinking to a certain point since it is monotone decreasing towards zero.

Limit supremum 上極限

Where the supremum denotes the smallest upper bound of a set, limit supremums denote the largest subsequential limit of a sequence (the largest value a subsequence of a given sequence can converge to), denoted as \(\limsup(x_n)\) or \(\lim_{n \to \infty}\sup \{x_n\}\)

\( \limsup \{x_n\} = \liminf\{x_n\} \implies \text{convergence}\)

\(n\)th Root test

\(S = \sum_{n=0}^{\infty} a_n\) converges when:

\( \limsup ( |a_{n}|^{\frac{1}{n}})=L\)

The proof for this relies on comparison test with a convergent geometric series for \(L \lt 1\), since a convergent geometric series may have terms of form \(r^{n}\) and we can delcare some \(r:L \lt r \lt 1\), which means \(|a_n|^{\frac{1}{n}} \lt r \implies |a_n| \lt r^n\)

P test

\(S = \sum_{n=0}^{\infty} \frac{1}{n^{p}}\) converges for \(p \gt 1\)

Cauchy condensation test

If a sequence is positive and non-increasing, then:

\( \sum_{n=1}^{\infty}a_n \text{ converges} \iff \sum_{n=0}^{\infty}2^{n}a_{2^{n}} \text{ converges} \)

The idea is because if it is non-increasing we can see that \(a_1 + a_2 \leq 2a_1 \implies \sum_{n=1}^{k} a_n \leq ka_1\), so we can extend this such that \(\sum_{n=1}^{\infty} a_n \leq a_1 + 2a_2 + 4a_4 + ...\) and that's how we obtain the inequality \(\sum_{n=1}^{\infty}x_n \leq \sum_{n=0}^{\infty}2^{n}x_{2n}\), and well if that series with the powers of 2 is bounded, and our series of interest is always less than that, then it converges too

\( \sum_{n=1}^{\infty}x_n \leq \sum_{n=0}^{\infty}2^{n}x_{2^{n}} \lt 2\sum_{n=1}^{\infty}x_n \)

Integral test

When \(f\) is positive, decreasing and continuous...

\(\sum_{n=1}^{\infty} f(n) \lt \infty \iff \int_{1}^{\infty} f(x)dx \lt \infty\)

Continuous functions

Limit Limite 制限

Concept that if \(x\) tends to \(x_0\), \(f\) converges to some value \(L\)

\( \lim_{x \to x_0}f(x) = L \iff |x - x_{0}| \lt \delta \implies |f(x)-L| \lt \varepsilon \)

Theorems

One-sided limits

Positive limit

\( \lim_{x \to a^{+}} =L \iff a \lt x \lt a+\delta \implies |f(x) - L| \lt \varepsilon \)

Negative limit

\( \lim_{x \to a^{-}} =L \iff a-\delta \lt x \lt a \implies |f(x) - L| \lt \varepsilon \)

Finding the limit using one-sided limits

\( \lim_{x \to a} =L \implies \lim_{x \to a^{+}} = \lim_{x \to a^{-}} = L\)

Multivariable limits

\( \lim_{(x,y) \to (x_0,y_0)}f(x,y) = L \iff \sqrt{(x - x_{0})^2 + (y-y_0)^2} \lt \delta \implies |f(x,y)-L| \lt \varepsilon \)

Polar coordinates

\( \lim_{(x,y) \to (0,0)}f(x,y) = \lim_{r \to 0^{+}} f(r\cos(\theta), r \sin (\theta) \)

Function composition

Parsing the output of one function to the input of another, nesting a function in another

\( (f \circ g)(x) = f(g(x)) \)

Continuous function funzione continua 連続関数

Function with the property that domain elements that are arbitrarily close map to image elements that are arbitrarily close

\( f \text{ is continuous on }U \iff \forall u \in U (f \text{ is continuous at }u) \)

\(U\) is a set

\( f \text{ is continuous at }x \iff [ \lim_{t \to x} f(t) = f(x)] \)

\( f \text{ is continuous at }x \iff [ \lim_{n \to \infty} x_{n} = x \implies \lim_{n \to \infty} f(x_n) = f(x)] \)

\( f \text{ is continuous at }x_0 \iff \forall \varepsilon \exists \delta_{x} \gt 0 ( |x_0 - x| \lt \delta_{x} \implies |f(x_0)-f(x)| \lt \varepsilon ) \)

Note that \(\delta_{x}\) is dependent on each domain value \(x\)

\( f \text{ is continuous at }x_0 \iff \forall N_1 (f(x_0)) [ \exists N_2(x_0) [ x \in N_2 (x_0) \implies f(x) \in N_1 (f(x_0)) ] ]\)

\( f \text{ is continuous on }\text{dom}(f) \iff \forall U \subset \text{codom}(f) ( U \text{ is open } \implies f^{-1}(U) \text{ is open }) \)

\( f \text{ is continuous on } U \iff \forall V \subset f(U) ( V \text{ is open } \implies f^{-1}(V) \text{ is open }) \)

Proof of definition equivalence

Assume the epsilon-delta definition satisfies for \(x\) and let \(y_n\) be a sequence approaching \(x\). This implies \( (\varepsilon , \delta_{x} ) \) can be chosen such that \(\exists N : n \geq N \implies |x-y_{n}| \lt \delta_{x} \implies |f(x) -f(y_n)| \lt \varepsilon\), however since these statements are equivalent to the definition of a limit. Hence \(\text{Epsilon-delta definition} \implies \text{Limit definition}\)

Assume the epsilon-delta definition fails for \(x\). Then even our sequence \(y_n\) may demonstrate \(\exists N : \forall n \geq N, |f(x) -f(y_n)| \geq \varepsilon\), which represents a sequence \(f(y_n)\) not approaching \(f(x)\), hence \(\neg \text{Epsilon-delta definition} \implies \neg\text{Limit definition}\) and therefore \(\text{Epsilon-delta definition} \iff \text{Limit definition}\)

Propositions

\(f,g \text{ are continuous } \implies (f \circ g) \text{ is continuous}\)

\(f\text{ is monotone on } (a,b) \implies f \text{ is continuous with countable discontinuous points}\)

Uniformly continuous function Funzione uniformamente continua 一様連続関数

Notion of continuity on an open set \(U\) such for any epsilon, there is a single delta that can prove continuity for all elements of \(U\) simultaneously

\( f \text{ is uniformly continuous on }U \iff \forall \varepsilon \exists \delta [ \forall u,v \in U ( |u - v| \lt \delta \implies |f(u)-f(v)| \lt \varepsilon )] \)

Note that \(\delta\) is independent of domain values

Proposition

\(f \text{ is uniformly continuous } \implies f \text{ is continuous}\)

\(f \text{ is continuous on } [a,b] \iff f \text{ is uniformly continuous on } [a,b]\)

Lipschitz continuous function Funzione lipschitziana リプシッツ連続性

\(f\) is Lipschitz continouous if some \(M\) that satisfies the below inequality. \(M\) can be intuitively thought of as a 'maximum gradient'

\(f \text{ is Lipschitz continuous } \iff \exists M \gt 0 : |f(x) - f(y)| \leq M|x-y|\)

Extreme value theorem Teorema di Weierstrass 最大値最小値定理

All cotninuous functions on a closed interval have a maxima and minima

\( f \text{ is continuous on } [a,b] \implies \exists c,d : f(c) \leq f(x) \leq f(d), \forall x \in [a,b]\)

Intermediate Value Theorem (IVT)

Theorem asserting that all functions continuous on an interval with a positive range element an negative range element have some point \(c\) mapped to a range element \(0\) (note that the construction \(f(a)f(b)\) is negative iff \(f(a)\) and \(f(b)\) have different signs)

\(f \text{ is continuous on } [a,b] \land f(a)f(b) \lt 0 \implies \exists c \in [a,b] : f(c)=0\)

Differentiation Differenziazione 微分

Differentiable function Funzione differenziabile 微分可能関数

\( f \text{ is differentiable on } U \iff \forall x \in U [ f \text{ is differentiable at }x ] \)

\( f \text{ is differentiable at }x_0 \iff \exists f'(x_0) [ f'(x_0) = \lim_{x \to x_0} \frac{f(x) - f(x_0)}{x-x_0} ]\)

Derivative Derivativo 導関数

Scalar quantity evaluating the gradient of a differentiable function at a point.

represented by the Newtonian quotient; the ratio of the increase in range for an increase in domain for infinitesimally small values

\( f'(x) = \lim_{h \to 0} \frac{f(x+h) - f(x)}{h} = \lim_{t \to x} \frac{f(t) - f(x)}{x-t} \)

Proposition

\( f \text{ is differentiable on } X \implies f \text{ is continuous on } X\)

Product rule

\( (fg)'(x) = f(x)g'(x)+ f'(x)g(x) \)

Proof

\( (fg)'(x_0) = \lim_{x \to x_0} \frac{f(x)g(x) - f(x_0)g(x_0)}{x - x_0} \)

\( (fg)'(x_0) = \lim_{x \to x_0} \frac{f(x)g(x) - f(x)g(x_0) + f(x)g(x_0)- f(x_0)g(x_0)}{x - x_0} \) (add and subtract)

\( (fg)'(x_0) = \lim_{x \to x_0} f(x)\frac{g(x) - g(x_0)}{x-x_0} + g(x_0)\frac{f(x)- f(x_0)}{x - x_0} \) (factorize)

\( (fg)'(x_0) = f(x)g'(x)+ f'(x)g(x) \) (solve limit)

Chain rule

\( (f \circ g)'(x) = f'(g(x))g'(x)\)

Proof

\( (f \circ g)'(x_0) = \lim_{h \to 0} \frac{f(g(x+h)) - f(g(x))}{h}\)

\( (f \circ g)'(x_0) = \lim_{h \to 0} \frac{f(g(x+h)) - f(g(x))}{h} \frac{g(x+h) - g(x)}{g(x+h) - g(x)}\) (multiply by conjugate)

\( (f \circ g)'(x_0) = \lim_{h \to 0} \frac{f(g(x) + k) - f(g(x))}{k} \frac{g(x+h) - g(x)}{h}\)

\( (f \circ g)'(x) = f'(g(x))g'(x)\)

Quotient rule

\( (\frac{f}{g})'(x) = \frac{f(x)g'(x) - f'(x)g(x)}{g^2(x)} \)

Proof

\( (\frac{f}{g})'(x) = f(x)(\frac{1}{g(x)})' + \frac{f'(x)}{g(x)} \) (product rule)

\( (\frac{f}{g})'(x) = \frac{f(x)g'(x) - f'(x)g(x)}{g^2(x)} \) (chain rule)

Partial derivative

In mutlivariable calculus, with a function \(f : \mathbb{R}^2 \to \mathbb{R}\)

\( \frac{\partial f}{\partial x}(x,y) = \lim_{h \to 0} \frac{f(x+h,y) - f(x,y)}{h}\)

\( \frac{\partial f}{\partial y}(x,y) = \lim_{k \to 0} \frac{f(x,y+k) - f(x,y)}{k}\)

Maxima

A maxima is reached when \(f'(c) = 0\)

Rolle's theorem

On a differentiable interval \( I = (a,b) \) the following holds:

\( f(a)=f(b)=0 \implies \exists c \in (a,b) : f'(c)=0 \)

Between two equal points of a function, there is at least one point where the derivative is zero

Mean value theorem (MVT)

For all functions \(f\) on a differentiable interval \(I\), there exist some element \(c\) such that \(f'(c)\) represents a line from \(f(a)\) and \(f(b)\), i.e, the 'mean derivative'

\( I = (a,b), f \in C^{1}(I) \implies \exists c \in I : f'(c)=\frac{f(b)-f(a)}{b-a} \)

Generalized mean value theorem

\( f(a)=f(b) \implies \frac{f'(c)}{g'(c)}=\frac{f(b)-f(a)}{g(b)-g(a)} : c \in (a,b) \)

L'Hôpital's rule

L'Hôpital's rule asserts that when \( \lim_{x \to a^{+}} f(x) = \lim_{x \to a^{+}} g(x) = 0, \pm \infty\), the ratio of these two functions tends to the same value as the ratio of their derivative functions

\(\lim_{x \to a}f(x) = \lim_{x \to a} g(x) = 0, \pm \infty \land f,g \in C^{1} \implies \lim_{x \to a} \frac{f(x)}{g(x)} = \lim_{x \to a} \frac{f'(x)}{g'(x)} \)

Inverse function theorem

\(f \in C^{1}(I) \text{ is injective } \land f'(a) \neq 0, a \in I \implies (f^{-1})'(f(a)) = \frac{1}{f'(a)}\)

Proof

Injectivity of \(f\) implies existence of a continuous \(f^{-1}\), hence

\( (f^{-1})'(f(a)) = \lim_{x \to a} \frac{f^{-1}(f(x)) - f^{-1}(f(a))}{f(x)-f(a)} \)

\( (f^{-1})'(f(a)) = \lim_{x \to a} \frac{x - a}{f(x)-f(a)} \)

\( (f^{-1})'(f(a)) = \frac{1}{f'(a)}\)

Convex function

Class of functions such that between any two range elements \(x,y\) all range elements are less than the line connecting \(x\) and \(y\)

\(f \text{ is convex } \iff ( a+b=1 \implies f(ax+by) \leq af(x)+bf(y) )\)

Theorems

\(f \text{ is convex on open interval } \implies \exists f'(x^{+}), f'(x^{-}) : f'(x^{+}) = \lim_{h \to 0^{+}} \frac{f(x+h)-f(x)}{h}, f'(x^{-}) = \lim_{h \to 0^{-}} \frac{f(x+h)-f(x)}{h}\)

\(f \text{ is convex on open interval } I \implies f \text{ is Lipschitz continuous on } I\)

\(f \text{ is convex on on open interval } I \implies f \in C^{1}(I) \text{ with countable indifferentiable points}\)

\(f \in C^{1}(I) \implies (f \text{ is convex on } I \iff f' \text{ is monotone increasing on } I) \)

Concave function

\(f \text{ is convex } \iff ( a+b=1 \implies f(ax+by) \geq af(x)+bf(y) )\)

Power series

Infinite series of the following form

\( \sum_{n=0}^{\infty} a_n (x-c)^{n} \)

Radius of convergence

When a power series converges iff \(|x - c| \lt R\), \(R\) is said to be the radius of convergence, series of these forms can always have the ratio test applied, hence:

\(R = ( \lim_{n \to \infty } |\frac{a_{n+1}}{a_{n}}| )^{-1}\)

Theorems

\( ( f(x) = \sum_{n=0}^{\infty} a_n (x-c)^{n} \lt \infty \iff |x-c| \lt R) \implies (f'(x) = \sum_{n=1}^{\infty} n a_n (x-c)^{n-1} \lt \infty \iff |x-c| \lt R) \)

\( ( f(x) = \sum_{n=0}^{\infty} a_n (x-c)^{n} \lt \infty \iff |x-c| \lt R) \implies (f'(x) = \sum_{n=1}^{\infty} n a_n (x-c)^{n-1} \lt \infty \iff |x-c| \lt R) \)

Taylor series

Power series relating to a function

analytic functions equal their own Taylor series

\( f \sim \sum_{n=0}^{\infty} \frac{f^{(n)}(c)(x-c)^{n}}{n!}\)

Its construction can be though of as a polynomial such that at the expansion point, all orders of derivatives equal the function it models

Taylor's theorem

Theorem stating a property that the Taylor series error (denoted \(F(x_0)\)) has. When this property is established, it can be noted that \(\lim_{n \to \infty} F(x_0) = 0\) (when approaching infinite terms, the Taylor series equals the function)

\( P_n(x) = \sum_{k=0}^{n} \frac{f^{k}(x_0)(x-x_0)^{k}}{k!} \)

\( f \in C^{n+1}(I) \implies \exists \xi \in [x,x_0] : f(x) = P_n(x)+ \frac{f^{(n+1)}(\xi)(x-x_0)^{n+1}}{n!}\)

Proof

Consider the error of a \(C^{n}(I)\) Taylor approximation when using \(t\) as the point of expansion \( F(t) = f(x) - P_{n}(t) \). Note the following:

One desires some constructed function \(G\) in terms of the error function such that \(G(x)=G(x_0)=0\) (in order to apply Rolle's theorem to bound the error)

One such construction is \(G(t) = F(t) - F(x_0)(\frac{x-t}{x-x_0})^{n+1}\) and teh one collates some properties to verify that we can apply Rolle's theorem with \(G\), and finds the \(G'\) to employ in Rolle's theorem

Rolles' theorem says \(\xi \in [x,x_0] : -\frac{f^{(n+1)}(\xi)(x-\xi)^{n}}{n!} + F(x_0)\frac{(n+1)(x-\xi)^{n}}{(x-x_0)^{n+1}} = 0\)

Rearranging proves that \(F(x_0) = \frac{f^{(n+1)}(\xi)(x-x_0)^{n+1}}{(n+1)!}\)

Sine product series

\( \sin (z) = z \prod_{n=1}^{\infty} (1-(\frac{z}{n\pi})^2)\)

Analytic function

Function that equals its Taylor series on \(U\)

\(f \text{ is analytic on }U \iff \forall u \in U [f(u) = \sum_{n=0}^{\infty} \frac{f^{(n)}(c)(x-c)^{n}}{n!} ]\)

\(C^{k}\) spaces

Property of functions measured by differentiability classes; sets of functions that hold the same type of differentiabiliy

Smooth function

\(f \text{ is smooth on }U \iff f \in C^{\infty}(U)\)

Continuously differentiable function

\(f \text{ is continuously differentiable on }U \iff f \in C^{1}(U)\)

Riemann integral Integrale di Riemann リーマン積分

Interval partition

Strictly increasing finite sequence starting at the initial point of the interval and arriving at the final point of the interval

\( (x_i)^{n}_{i=0} \text{ is a partition of }[a,b] \iff \)

A partition may conveniently also refer to the set of the terms formed by this sequence

Uniform partition

Partition where each term is equally spaced from one another

\(x_{i} = a + \frac{(b-a)i}{n} \text{ is a uniform partition of }[a,b]\)

Mesh

Largest distance between terms in a partition

\( \| \mathcal{P} \| = \max_{i \in \mathbb{N} \cap [1,|\mathcal{P}|]} ( x_i -x_{i-1} ) \)

Riemann integrable function Funzione Riemann-integrabile リーマン可積分関数

Function that on some interval of its image can have its integral evaluated by partitioning the interval to arbitrary precision.

Definition

\( f \text{ is Riemann integrable on } [a,b] \iff \overline{\int_{b}^{a}} f = \underline{\int_{b}^{a}} f\)

Riemann's criterion

Alternative definition of a Riemann integrable function

\( f \text{ is Riemann integrable on } [a,b] \iff \forall \varepsilon \exists \mathcal{P} \subset [a,b] ( |U(f,\mathcal{P}) - L(f, \mathcal{P})| \lt \varepsilon\)

Riemann integral Integrale di Riemann リーマン積分

Scalar quantity relating to area bound by a Riemann integrable function on an interval.

\( \int^{b}_{a} f(x)dx = \overline{\int_{b}^{a}} f = \underline{\int_{b}^{a}} f \)

\( \int^{b}_{a} f(x)dx = \lim_{n \to \infty} \sum_{i=1}^{n} f(x_i) \Delta x_i\)

Propositions

\( \int_{a}^{b} cdx = c(b-a)\)

\( |\int_{a}^{b} f(x)dx| \leq \int_{a}^{b} |f(x)| dx\)

\(f \text{ is continuous on } [a,b] \implies f \text{ is Riemann integrable on } [a,b]\)

\( \int_{a}^{b} (f(x)+g(x))dx = \int_{a}^{b} f(x)dx + \int_{a}^{b} g(x))dx\)

Fundamental Theorem of Calculus (FTC)

Part 1

\(f \text{ is continuous on } [a,b] \implies \frac{d}{dx} \int_{a}^{x} f(t)dt = f(x)\)

Part 2

\(f \text{ is Riemann integrable on } [a,b] \implies \int_{a}^{b} f(x)dx = F(b)-F(a) : F'=f\)

Proof

Part 1

Consider \(F(x) = \int_{a}^{x} f(t)dt\) and note the following:

\(|F(x) -F(y)| \leq \int_{y}^{x} |f(t)|dt \leq \int_{y}^{x} Mdt = M|x-y| \implies F \text{ is Lipschitz continuous on } [a,b] \implies F \text{ is continuous on } [a,b]\)

Due to continuity of \(f\) and \(F\), one can choose \((\varepsilon , \delta_{x})\) to show that the \(F'=f\) as such, letting \(|x-y| \lt \delta_{x}\)

\( |\frac{F(x)-F(y)}{x-y} - f(y)| \leq \frac{1}{|x-y|} \int^{x}_{y} |f(t) - f(y)|dt \leq \frac{1}{|x-y|} \int^{x}_{y} \varepsilon dt = \varepsilon \implies F'=f\)

Part 2

Consider \(G(x) = \int^{x}_{a} f(t)dt\) and note the following:

\(G(b) -F(b) = G(a) - F(a) \implies G(b) -G(a) = F(b) -F(a) \implies \int_{a}^{b} f(x)dx = F(b)-F(a)\)

Integration by parts

\( (fg)'(x) = \int f(x)g'(x)+ f'(x)g(x) \implies \)

\( \int f(x)g'(x) dx = (fg)(x) - \int f'(x)g(x) dx \)

Integration by substitution

\( \int (f \circ u)(x) u'(x) dx = \int f(u) du \)

Mean value theorem of integrals

For all functions \(f,g\) on a Riemann integrable interval \(I\), there exist some element \(c\) such that the integral on \(f(x)g(x)\) equals the integral on \(f(c)g(x)\), i.e the 'mean integral' caused by \(f\)

\( I = (a,b), f,g \text{ are Riemann integrable on }I \land \forall x, g(x) \geq 0 \implies \exists c \in I : \int^{b}_{a} f(x)g(x)dx=f(c)\int^{b}_{a} g(x)dx \)

Improper Riemann integrals

If a function is discontinuous at \(a\), then it is an improper Riemann integral and providing that there is an existing limit, it can be found as

\( \int_{a}^{b} f(x)dx = \lim_{X \to a} \int_{X}^{b} f(x)dx \)

Cauchy Principle Value (CPV)

When an improper Riemann integral fails on some interval \((-\infty,\infty)\), a Cauchy principle value may return a valid answer

\( \text{pv}\int_{-\infty}^{\infty} f(x)dx = \lim_{X \to \infty} \int_{-X}^{X} f(x)dx \)

Again, prividing this exists

Double integrals

\(\int_{b}^{a} (\int_{d}^{c} f(x,y)dy) dx\)

Fubini's theorem

\(\int_{b}^{a} (\int_{d}^{c} f(x,y)dy) dx =\int_{d}^{c} (\int_{b}^{a} f(x,y)dx) dy\)

Cauchy-Schwarz inequality (integration)

\( ( \int_{b}^{a} f(x)g(x)dx )^{2} \leq ( \int_{b}^{a} f^2(x)dx )( \int_{b}^{a} g^2(x)dx ) \)

Arc length of a function

To find the length of an arc of a function, we can derive it from the priciple of the distance between two infinitesimal points.

\( L = \int_{a}^{b} \sqrt{1 + f'(x)^2} dx\)

Proof

Partition the function as \(\mathcal{P} = \{ x_0,x_1,...,x_n \}\)

Denote \( \Delta x = x_{i} - x_{i-1} , \Delta y = f(x_{i}) - f(x_{i-1})\)

\( L = \lim_{n \to \infty} \sum^{n}_{i=1} \sqrt{\Delta x^2 + \Delta y^2}\)

\( \exists \xi \in [ x_{i-1},x_i ] : f'(\xi)=\frac{\Delta y}{\Delta x} \)

\( L = \lim_{n \to \infty} \sum^{n}_{i=1} \sqrt{\Delta x^2 + (f'(\xi)\Delta x)^2}\)

\( L = \lim_{n \to \infty} \sum^{n}_{i=1} \sqrt{1 + f'(\xi)^2} \Delta x\)

\( L = \int_{a}^{b} \sqrt{1 + f'(x)^2} dx\)

Sequences of functions

Sequence of functions

A sequence \(f_n : \mathbb{N} \times X \to Y\) such that each term is a real function

Pointwise convergence

Convergence such that considering a variable setting for the sequence of functions variable, a term in a sequence of functions can be chosen to be arbitrarily close to a function \(f\).

\( \lim_{n \to \infty} f_n = f \text{ pointwise } \iff \forall x (\forall \varepsilon (\exists N \in \mathbb{N} : n\geq N \implies |f_n(x) - f(x)| \lt \varepsilon ))\)

Uniform convergence

Convergence such that a term in a sequence of functions can be chosen to be arbitrarily close to function \(f\) without dependence on the sequence of functions variable

\( \lim_{n \to \infty} f_n = f \text{ uniformly } \iff \forall \varepsilon (\exists N \in \mathbb{N} : n\geq N \implies |f_n(x) - f(x)| \lt \varepsilon) \)

\( \lim_{n \to \infty} f_n = f \text{ uniformly } \iff \forall \varepsilon (\exists N \in \mathbb{N} : n\geq N \implies \sup_{x \in I} |f_n(x) - f(x)| \lt \varepsilon ))\)

Theorems

Proofs

Due to the assumed conditions, sufficient \(k,\delta\) can be chosen such that the following hold:

\( |x-y| \lt \delta \implies |f(x) - f(y)|\)

\(=|f(x) -f_{k}(x) -f_{k}(y) + f_{k}(x) + f_{k}(y) -f(y)| \) (term injection)

\( \lt |f(x) - f_{k}(x)| + |f_{k}(x) - f_{k}(y)| + |f_{k}(y) - f(y)|\) (triangle inequality)

\( \lt 3\frac{\varepsilon}{3} = \varepsilon \implies f \text{ is continuous} \)

Due to the assumed conditions, sufficient \(k,\delta\) can be chosen such that the following hold:

\( |x-y| \lt \delta \implies |f(x) - f(y)|\)

\(=|f(x) -f_{k}(x) -f_{k}(y) + f_{k}(x) + f_{k}(y) -f(y)| \)

\( \lt |f(x) - f_{k}(x)| + |f_{k}(x) - f_{k}(y)| + |f_{k}(y) - f(y)|\)

\( \lt 3\frac{\varepsilon}{3} = \varepsilon \implies f \text{ is continuous} \)

Weierstrass M test

Uniform convergence test that permits comparison with a convergent series to prove uniform convergence (since convergent series lack parameters)

\( \sum^{\infty}_{n=1} M_n \lt \infty \implies \sum_{n=1}^{\infty} f_n(x) \text{ is uniformly convergent on X}\)

Proof

Consider \(S_n(x) = \sum_{k=1}^{n} f_k(x) \) and note the following:

\(| S_n(x) - S_m(x) | \lt \sum_{k=m+1}^{n} M_n \lt \varepsilon \implies S_n(x) \text{ is uniformly Cauchy} \implies S_n(x) \text{ is uniformly convergent}\)

Note that the property of uniformly Cauchy was implied due to absense of reliance on the parameter \(x\)

Uniform cauchy

\( f_n \text{ is uniformly cauchy } \iff \forall \varepsilon ( \exists N \in \mathbb{N} : n,m\geq N \implies |f_n(x) - f_m(x)| \lt \varepsilon\)

Theorem

\( f_n \text{ is uniformly cauchy } \iff \lim_{n \to \infty} f_n = f \text{ uniformly }\)

Arzela's Bounded Convergence Theorem (ABCT)

If a sequence of functions is pointwise convergent, uniformly bounded, and Riemann integrable, then limits and integrals can be swapped

\( \lim_{n \to \infty} f_n = f \text{ pointwise on } [a,b] \land (f_n)^{\infty}_{n=1} \text{ is uniformly bounded } \land f_n \text{ is Riemann integrable on }[a,b] \land f \text{ is integrable on }[a,b] \implies\)

\( \lim_{n \to \infty} \int^{b}_{a} f_n (x)dx = \int^{b}_{a} \lim_{n \to \infty} f_n (x)dx = \int^{b}_{a} f(x)dx \)

Weierstrass Approximation Theorem

Theorem asserting that any continuous function on a closed and bounded interval can be uniformly approximated by some polynomial with arbitrarily small error

\( f \text{ is continuous on } [a,b] \implies \exists P : \sup_{x \in [a,b] }|f(x) - P(x)| \lt \varepsilon \)

Stone-Weierstrass Theorem

\( X \text{ is a compact Hausdorff space} \land A \text{ is a subalgebra of } C(X,\mathbb{R}) \implies A \text{is dense in } C(X) \)

Characteristic function

A class of function with a range of \( \{ 0,1\} \) that returns \(1\) iff the input belongs to some set \(S\)

\( \chi_{S}(x)= \begin{cases} 1 & x \in S \\ 0 & x \notin S \end{cases} \)

Dirichlet function

A characteristic function that returns 1 on a rational input and 0 on an irrational input. it is an example of a completely discontinuous function

\( \chi_{\mathbb{Q}}(x) = \begin{cases} 1 & x \in \mathbb{Q} \\ 0 & x \notin \mathbb{Q} \end{cases} = \lim_{n,k \to \infty} \cos(n!\pi x)^{2k} \)

Dini's Theorem

\( \forall n, f_n \text{ is continuous and monotone on } [a,b] \land \lim_{n \to \infty} f_n = f \text{ pointwise} \land f\text{ is continuous } \implies \lim_{n \to \infty} f_n = f \text{ uniformly}\)

Proof

WLOG assume \(f_n\) is monotone decreasing, and \(\lim_{n \to \infty} f_n = 0\)

Define \(M_n = \sup \{f_n (x) :x \in [a,b] \}\) and note the proposition \(\lim_{n \to \infty} M_n = 0\) implies uniform convergence since if the supremum of some \(f_n\) can be chosen to be arbitrarily close to 0, the whole domain of \(f_n\) is even closer to 0, hence uniformly convergent.

By WOC assume \(\forall n, M_n \gt \varepsilon\), implying that \( \exists x_n : f_n (x_n) \gt \varepsilon\). \(\lim_{n \to \infty} f_n (x_{n_{k}}) = 0\) therefore Note the following properties:

Since \(f_n (L) \lt \frac{\varepsilon}{2}\), injecting this into the statement of continuity gives \( |f_n (x) - f_n (L) + f_n (L)| = |f_n (x)| \lt \varepsilon \), however this contradicts \( f_n (x_n) \gt \varepsilon \), therefore \(\forall n, M_n \lt \varepsilon\), and

Methods of integration

Gamma function Funzione gamma ガンマ関数

\( \displaystyle \Gamma (z) = \int_{0}^{\infty} t^{z-1}e^{-t} dt \)

Propositions

Proofs

The fact that \(\Gamma (z+1) = z \Gamma (z) \) can be proved using integration by parts

\( \Gamma(z) = \int_{0}^{\infty} t^{z-1}e^{-t} dt = [-t^{z-1}e^{-t}]^{\infty}_{0} - \int_{0}^{\infty} -(z-1)t^{(z-1)-1}e^{-t} dt\)

\( = \int_{0}^{\infty} (z-1)t^{(z-1)-1}e^{-t} dt = (z-1) \Gamma (z-1)\)

\(\Gamma (n) = (n-1)! \) is a simple corrolary of this fundamental property since \( \Gamma(1) = 1 = 0!\), and so \(\Gamma(k)=(k-1)!\) and finally \(\Gamma(k+1)=k \Gamma (k)= k(k-1)!=k!\)

Euler's product form of the Gamma function can be derived by abusing the fact that \(\forall z, \lim_{n \to \infty} \frac{n!(n+1)^z}{(n+z)!} = 1\). We can then assume that \(\lim_{n \to \infty} \frac{n!(n+1)^z \Gamma(z)}{(n+z)!} = \Gamma (z)\)

\( \implies \lim_{n \to \infty} \frac{n!(n+1)^z (z-1)!}{(n+z)!} = \Gamma (z)\)

\( \implies \frac{1}{z}\lim_{n \to \infty} \frac{n!(n+1)^z z!}{(n+z)!} = \Gamma (z)\)

\( \implies \frac{1}{z}\lim_{n \to \infty} \frac{ (n (n-1) ... 1)( \frac{2}{1} \frac{3}{2} \frac{4}{3}\frac{5}{4} ... \frac{n+1}{n} )^z }{ ((z+1)(z+2)...(n+z))} = \Gamma (z)\)

\( \implies \Gamma (z) = \frac{1}{z} \prod^{\infty}_{n=1} [ \frac{1}{1+ \frac{z}{n}} (1+ \frac{1}{n})^z ] \)

Duplication formula

Multiplication formula

Beta function Funzione betaベータ関数

\( \displaystyle B (a,b) = \int_{0}^{1} t^{a-1} (1-t)^{b-1} dt : a,b \gt 0\)

Propositions

Leibniz integral rule (Feynman's technique)

Integration technique by passing a differential operator under the integral for a non-integrated variable

\( \displaystyle \frac{d}{dt} (\int_{b}^{a} f(x,t) dx ) = \int_{a}^{b} \frac{\partial}{\partial t} f(x,t) dx \)

\( \displaystyle \frac{d}{dt} (\int_{u_2(t)}^{u_1(t)} f(x,t) dx) = f(u_2(t),t)u'_2(t) -f(u_1(t),t)u'_1(t) + \int_{u_2(t)}^{u_1(t)} \frac{\partial}{\partial t} f(x,t) dx\)

Fubini's theorem フビニの定理

Theorem that permits swapping order of multivariable integrals, useful for integrating under integral sign

\( \displaystyle \int_{a}^{b} \int_{c}^{d} f(x,y) dydx =\int_{c}^{d} \int_{a}^{b}f(x,y)dxdy \)

Series method

Integration technique that replaces factors within an integral with a series

Gaussian integral Integrale di Gauss ガウスの積分

\( \displaystyle \int_{-\infty}^{\infty} e^{-x^2}dx = \sqrt{\pi} \)

Proof

\( I^2 = (\int_{-\infty}^{\infty} e^{-x^2}dx)^2 = (\int_{-\infty}^{\infty} e^{-x^2}dx)(\int_{-\infty}^{\infty} e^{-y^2}dy)\)

\( \int^{2\pi}_{0} \int_{0}^{\infty} e^{-r^2}r dr d\theta\)

\( 2\pi \int_{0}^{\infty} e^{-r^2}r dr \)

\( 2\pi \int_{0}^{\infty} e^{-r^2}r dr \)

\( 2\pi \int^{0}_{-\infty} \frac{1}{2} e^{u} du \)

\( \pi \int^{0}_{-\infty} e^{u} du \)

\( I^2 = \pi \)

\( I = \sqrt{\pi} \)

Dirichlet integral

\( \int_{-\infty}^{\infty} \frac{\sin(x)}{x} dx = \pi \)