0x020 Foundation
Analysis is the art of taking limits.
1. Continuity
The basic relation between continuity concepts are:
1.1. Limits of Functions
Generally speaking, we want the value of \(\lim_{x \to c} f(x)\) to be independent of how we approach \(c\)
Definition (functional limit) Suppose \(x_0\) is a limit point of \(E\),
iff
It looks there are two version of functional limit definitions, the definition here is called the deleted limit version where the \(x = x_0\) is excluded from the requirements, the other version is called the non-deleted limit version where \(x = x_0\) can be included
These two definitions have slightly different conclusions sometimes (e.g: composition of function). This note will follow the deleted version.
example of deleted functional limit
Consider the function \(f(x)=0\) when \(x \neq 0\) and \(f(x)=1\) when \(x=0\) (there exists a jump discontinuity at \(0\)).
However, because the limit point is deleted, we can still argue
It is important to remember that limit points of \(E\) do not necessarily belong to \(E\) unless it is closed. In fact, \(x_0\) need not to be in the domain of \(f\).
limit point might not be in the domain
In the function \(f(x) = x \sin(1/x)\) with \(x>0\), \(x=0\) is not included in the domain but \(x=0\) is a limit point, so the limit on \(0\) can exist
1.1.1. Criterions
As mentioned above, the functional limit should be independent how we approach the limit point, therefore all sequence limit should converges to the same point.
Criterion (sequence criterion) the following statements are equivalent
- \(\lim_{x \to x_0; x \in E} f(x) = L\)
- \((a_n) \to x_0 \implies f(a_n) \to L\) where \(a_n \in E\)
This also gives us an approach to decide the divergence
Criterion (divergence) functional limit \(\lim_{x \to c} f(x)\) does not exist if
Criterion (Algebraic Limit Theorem) limit laws for functions
In the case of multivariable input and single output function, it is a bit tricky. Even limits along all lines exists and have the same value, the limit still might not exist
line limits does not guarantee limit
Define the function \(f\)
where \(\theta(x)\) is the angle measured from the positive part. All line limits is 0, but when taking curve limit \(||x||=\theta(x)\), then the curve limit is 1.
Again to make sure functional limit exist, ALL sequence limit should exist and equal not only line limits.
another counterexample of line limits
Consider the function \(f\)
It approaches 0 along the any line \(y=kx\), but approaches 0.5 along the line \(y=x^2\)
Proposition (limits are local) the following statements are equivalent
- \(\lim_{x \to x_0; x \in E} f(x) = L\)
- \((\forall \delta \in R) \lim_{x \to x_0; x \in E \cup (x - \delta, x+\delta)} f(x) = L\)
Proposition (multivariable convergence) Suppose \(U \subset \mathbf{R}^n\), \(f=(f_1, ..., f_m): U \to \mathbf{R}^m\) and \(x_0 \in \bar{U}\), then
is equivalent to
1.2. Continuity
The continuity class can be organized as follows:
Definition (continuity) A function \(f: A \to R\) is continuous at \(c \in A\) if for all \(\epsilon > 0\), there exists a \(\delta > 0\) such that
The most important point here is \(c\) must be included in the domain (unlike the functional limit where \(c\) is a limit point, which might not be in the domain)
Also notice because of this difference, here we require \(|x-c|<\delta\) rather than \(0 < |x-c| < \delta\) used in the functional limit.
isolated point
By this definition, we know functions are continuous at their isolated points of their domains
There are several characterizations of continuity:
Lemma (characterizations of continuity) \(f: X \to R\) is continuous at \(x_0 \in X\) iff \(x_0\) is a limit point and
Lemma (sequence version of continuity) Continuous function preserves the convergence
Definition (Discontinuity) There are three types of discontinuity
- If \(\lim_{x \to c} f(x)\) exists but has a value different from \(f(c)\), the discontinuity is called removable
- If \(\lim_{x \to c+} f(x) \neq \lim_{x \to c-} f(x)\), then \(f\) has a jump discontinuity at c
- If \(\lim_{x \to c}\) does not exist for other reasons, the discontinuity is called essential
nowhere continuous function
Dirichlet's function is an example of nowhere continuous function
discontinuous on \(Q\)
Thomae function is discontinuous on \(Q\) and continuous on \(I\)
1.2.1. Criterions
Criterion (Arithmetic operations and composition) if \(f, g\) are continuous at \(x_0\), then \(f+g, f-g, f/g, fg, \min(f,g), \max(f,g), g \circ f\) are also continuous at \(x_0\)
Criterion (limit composition) Let \(f,g\) be real functions, and \(\lim_{x \to p} f(x) = q, \lim_{x \to q} g(x) = r\), we have \(\lim_{x \to p} g(f(x)) = r\) when either of the following is satisfied:
- \(g(x)\) is continuous at \(q\)
- for some open interval \(I\) containing \(p\), it is true that \(f(x) \neq q\) for any \(x \in I\) except possibly \(x = p\). (this can happen, for example, \(f\) is monotone)
See the proofwiki
composition of non-continuous functions
Assume \(\lim_{x \to p} f(x) = q\), \(\lim_{x \to q} g(x)=r\), it may not be true that
For example, consider \(f(x)=0\) for all \(x\), \(g(x)=0\) if \(x \neq 0\) and \(g(x)=1\) if \(x = 0\), then \(\lim_{x \to 0} f(x) = 0\), \(\lim_{x \to 0} g(x) = 0\). however, \(\lim_{x \to 0} g(f(x)) = 1\)
To make the statement, we need to have the continuity of \(g\) on \(q\). (Note that in the non-deleted limit definition, this requirement can be dropped.)
In the case of single variable, we can define the left, right limits and use them to determine continuity.
Definition (left, right limits)
Criterion (left right limits and continuity) If \(f(x_0+), f(x_0-)\) both exist and equal to \(f(x_0)\), then \(f\) is continuous at \(x_0\)
Continuity has multiple good properties, for example, it preserves compactness and connectedness, which guarantees the continuous function has extreme value and intermediate value under certain conditions.
However, continuity are not good enough (not preserved) with respect to the pointwise convergence, measurable function improves in this point
continuity is not preserved over sup
See Abott Exercise 4.3.10 for this example
Consider the following function
Continuity is preserved over finite max operations: \(g(x)\) is continuous
However, it is not preserved over the infinite supermum
Note in contrast, measurable function is well-preserved wrt to the sup operations and pointwise convergence.
1.2.2. Continuity and Compactness
Theorem (preservation of compact sets) Let \(f: A \to R\) be continuous on \(A\). If \(K \subseteq A\) is compact, then \(f(K)\) is compact as well
Proposition (extreme value theorem) If \(f: A \to R\) is continuous on a compact set \(K\ subset R\), then \(f\) attains a maximum and mimum value.
1.2.3. Continuity and Connectedness
Theorem (preservation of connected sets) Let \(f: G \to R\) be continuous. If \(E \subseteq G\) is connected, then \(f(E)\) is connected as well
Theorem (Intermediate value theorem) Let \(f: [a,b] \to R\) be a continuous function on \([a,b]\). Let \(y\) be a real number between \(f(a), f(b)\). Then there exists \(c \in [a,b]\) such that \(f(c)=y\)
Conversely, a function having the intermediate value property does not necessariy need to be continuous.
1.3. Uniform Continuity
Continuity is a local property, where the choice of \(\delta\) can dependent the target point \(c\). Uniform continuity is a global property, where \(\delta\) should be same for all point in the domain.
Definition (uniform continuity) Let \(A \subset R\) and let \(f: A \to R\). We say that \(f\) is uniformly continuous if
Definition (equivalent sequences) Let \(m\) be an integer . Let \((a_n)_{n=m}^{\infty}, (b_n)_{n=m}^{\infty}\) be two sequences of real numbers, and let \(\epsilon > 0\) be given.
- (\(\epsilon\)-close) We say that \((a_n)_{n=m}^{\infty}\) is \(\epsilon\)-close to \((b_n)_{n=m}^{\infty}\) iff \(a_n\) is \(\epsilon\)-close to \(b_n\) for each \(n \geq m\).
- (eventually \(\epsilon\)-close) We say that \((a_n)_{n=m}^{\infty}\) is eventually \(\epsilon\)-close to \((b_n)_{n=m}^{\infty}\) iff there exists an \(N \geq m\) such that the sequences \((a_n)_{n=N}^{\infty}\) and \((b_n)_{n=N}^{\infty}\) are \(\epsilon\)-close.
- (equivalent sequences) Two sequences \((a_n)_{n=m}^{\infty}, (b_n)_{n=m}^{\infty}\) are equivalenet iff for each \(\epsilon > 0\), the sequences are eventually \(\epsilon\)-close. Lemma Real sequences of \((a_n)_{n=1}^{\infty}, (b_n)_{n=1}^{\infty}\) are said to be equivalent iff \(\lim_{n \to \infty} (a_n - b_n) = 0\)
Proposition (uniformly continuity and sequences) following are logically equivalent
\(f: X \to R\) is uniformly continuous on \(X \subset R\) Whenever \((x_n)_{n=1}^{\infty}, (y_n)_{n=1}^{\infty}\) are two equivalent sequences consisting of elements of \(X\), the sequences \((f(x_n))_{n=1}^{\infty}, (f(y_n))_{n=1}^{\infty}\) are also equivalent
Proposition (uniform continuity preserves Cauchy sequences) if \(f\) is a uniformly continuous function. Let \((x_n)_{n=1}^{\infty}\) be a Cauchy sequence. Then \((f(x_n))_{n=1}^{\infty}\) is also Cauchy
Theorem (compact continuous function is uniformly continuous) If \(f: A \to R\) is continous on a compact set \(A\). Then \(f\) is also uniformly continuous
1.4. Lipschitz continuity
Definition (Lipschitz continuity) A function \(f: A \to R\) is called Lipschitz if there exists a bount \(M > 0\) such that for all \(x, y \in A, x \neq y\)
Lemma Lipschitz continuous implies uniform continuous
Example (contraction mapping) A contractive function is a function
where \(0 < c < 1\). It is a Lipschitz continuous function
1.5. Set of Discontinuity
Given a function \(f: R \to R\), define \(D_f \subset R\) to be the set of points where the function \(f\) fails to be continuous.
Defintion (classification of discontinuities) Generally speaking, discontinuities can be divided into three categories:
- removable discontinuity: \(\lim_{x \to c }f(x)\) exists but has a value different from $f©
- jump discontinuity: \(\lim_{x \to c^+} f(x) \neq \lim_{x \to c^-} f(x)\)
- essential discontinuity: \(\lim_{x \to c} f(x)\) does not exist for some other reason.
Theorem (set of discontinuity for an monotone function) A monotone function can only have jump discontinuities. The set of discontinuity is either finite or countable. This also indicates monontone function is Riemann integrable, because the set of discontinuities has a measure of 0.
Theorem (set of discontinuity for an arbitrary function) Let \(f: R \to R\) be an arbitrary function. Then \(D_f\) is an \(F_\sigma\) set.
Proof idea: showing the following steps
- consider a concept of \(\alpha\)-continuity: for a fixed \(\alpha > 0\), the function is \(\alpha\)-continuity if there exists a \(\delta > 0\) such that for all \(y, z \in (x-\delta, x+\delta)\), it follows that \(|f(y) - f(z)| < \alpha\)
- show The set of \(f\) fails to be continuous \(D_f^{\alpha}\) is closed
- show \(D_f = \cup D^{\alpha_n}_f\) where \(\alpha_n = 1/n\)
discontinuity at \(Q\) and \(I\)
Thomae function is an example where the function's discontinuity set is \(Q\)
However, \(I\) is not a \(F_\sigma\) set, so there is no function \(f\) which is continuous at every rational and discontinuous at every irrational.
discontinuity at an arbitray closed set and open set
See exercise 4.3.14 of Abbott's book
For an arbitray closed set \(F\), we can construct a function whose value are \(2\) when \(x \notin F\) and Dirichlet function when \(x \in F\). Then \(F\) is discontinuous at precisely \(F\)
For an arbitrary open set \(O\), we can construct a function whose value are \(0\) when \(x \notin O\) and Dirichlet function multiplied by the distance function to \(O^c\): \(\inf \{ |x-a| | x \in O^c \}\). The multiplication makes sure \(x\) is continuous at the boundary of \(O\).
2. Differentiation
2.1. Real-value derivatives
Some definitions define differentiability only for interior points, others define it for limit points as well. We follow the latter definition.
Definition (differentiability at a point) Let \(X\) be a subset of \(\mathbb{R}\), \(x_0 \in X\) which is a limit point of \(X\), \(f: X \to \mathbb{R}\). If the limit
converges to some real number \(L\), we say that \(f\) is differentiable at \(x_0\) on \(X\) with derivative \(L\) and write \(f'(x_0) := L\). Otherwise \(f'(x_0)\) is undefined and \(f\) is not differentiable at \(x_0\).
Differentiability is a local property with respect to a point
Definition (differentiability on domain) Let \(X\) be a subset of \(R\) and let \(f: X \to R\) be a function. \(f\) is differentiable on \(X\) iff \(f\) is differentiable at every limit point \(x_0 \in X\)
Proposition (Newton's approximation) Let \(X\) be a subset of \(\mathbb{R}\), let \(x_0 \in X\) be a limit point of \(X\), let \(f: X \to \mathbb{R}\) a function and \(L\) be a real number. Following statements are logically equivalent:
\(f\) is differentiable at \(x_0\) on \(X\) with derivative \(L\)
for every \(\epsilon > 0\) there exists a \(\delta > 0\) such that \(f(x)\) is \(\epsilon |x-x_0|\)-close to \(f(x_0) + L(x-x_0)\) whenever \(x \in X\) is \(\delta\)-close to \(x_0\)
2.2. Multivariate Differentiability
Definition (directional derivative) Let \(E\) be a subset of \(\mathbb{R}^n\), \(f: E \to \mathbb{R}^m\) be a function, let \(x_0\) be an interior point of \(E\), and let \(v\) be a vector in \(R^n\). If the limit
exists, we say \(f\) is differentiable in the direction \(v\) at \(x_0\), and we denote the above limit by \(D_v f(x_0) \in \mathbb{R}^m\)
Lemma Suppose \(f(x)=(f_1(x), f_2(x), ..., f_n(x))\), then \(f\) is directional differentiable with \(e\) is equivalent to the statement that for all \(i\), \(f_i\) is directional differentiable with \(e\).
Directional Derivative is scalar multiplicative, the derivative of \(ku\) direction is \(k\) times the derivative of \(u\) direction.
Definition (directional linear) A function is called directional linear, iff there exists a linear transformation that can be used to compute all directional derivatives at that point
Definition (partial derivative) Let \(E\) be a subset of \(\mathbb{R}^n\), \(f: E \to \mathbb{R}^m\) and \(x_0\) is an interior point of \(E\), \(1\leq j \leq n\). Then the partial derivative of \(f\) with respect to \(x_j\) variable at \(x_0\), denoted \(\frac{\partial f}{\partial x_j}(x_0)\) is defined by
Next, we think about the differentiability for the multivariable function, the definition of differentiability for one variable cannot be directly to multivariable function \(f: R^n \to R^m\) because the denominator and nominator are vectors of different dimension.
Instead of this definition, we interpret that \(f\) is approximately linear near \(x_0\), which \(f(x) \approx f(x_0) + L(x-x_0)\), then the previous definition can be defined as
where the denominator and nominator are both scalar, this definition can be applied in the multivariable function as follows.
Definition ((total) differentiability) Let \(E\) be a subset of \(R^n\), \(f: E \to R^m\) be a function. \(x_0 \in E\) be a point, let \(L: R^n \to R^m\) be a linear transformation. We say \(f\) is differentiable at \(x_0\) with derivative \(L\) if we have
where \(L\) is called the total derivative or Jacobian matrix.
For a differentiable function,
Not differentiability implies directional linear, but directional linear does not imply differentiability.
Note the total differentiability is corresponding to the standard differentiability in one variable.
Definition (Jacobian matrix) The Jacobian matrix is defined as \(\(\mathbf J = \begin{bmatrix} \dfrac{\partial \mathbf{f}}{\partial x_1} & \cdots & \dfrac{\partial \mathbf{f}}{\partial x_n} \end{bmatrix} = \begin{bmatrix} \dfrac{\partial f_1}{\partial x_1} & \cdots & \dfrac{\partial f_1}{\partial x_n}\\ \vdots & \ddots & \vdots\\ \dfrac{\partial f_m}{\partial x_1} & \cdots & \dfrac{\partial f_m}{\partial x_n} \end{bmatrix}\)\)
Definition (Jacobian) Jacobian matrix is a square matrix when \(m=n\), Its determinant is called the Jacobian determinant, or simply the Jacobian.
The Jacobian at a given point gives important information about the behavior of \(f\) near the point. For example, the continuously differentiable function \(f\) is invertible near \(p\) when the Jacobian determinant at \(p\) is non-zero, which is known as the inverse function theorem
2.3. Smoothness
In general, the differentiability class can be organized as follows:
2.3.1. Differentiability and Continuity
Continuity does not guarantee the differentiability. Consider the following examples
continuous everywhere but not differentiable at 0
Consider the function \(g_1(x) = x \sin(1/x)\) when \(x \neq 0\), otherwise \(g_1(x)=0\). This function is continuous everywhere. But not differentiable at \(0\)
Weierstrass function: continuous everywhere but differentiable nowhere
The easy example is the absolute value function where 0 is not differentiable. Weierstrass has a even function where continuous but nowhere differentiable
where \(a,b\) are carefully chosen.
2.3.2. Differentiability and Continuously Differentiability
Being differentiable does not guarantee its derivative is continuous, showing
Here is the famous example
differentiable everywhere but has discontinuity
Consider the following function
It is differentiable everywhere, the derivative is given by
However, it has the discontinuity at 0 which is an essential discontinuity
Even it is differentiable, derivative functions might not be continuous, but it will satisfy the intermediate value properties as follows
Theorem (Darboux, intermediate value property of derivatives) If \(f\) is differentiable on an interval \([a,b]\), and if \(\alpha\) satisfies \(f'(a) < \alpha < f'(b)\), then there exists a point \(c \in (a,b)\) where \(f'(c) = \alpha\)
This theorem can be applied using the following interior extremum theorem, check the proof on Wikipedia
Theorem (Fermat, Interior Extreme Theorem) Let \(f\) be differentiable on an open interval \((a,b)\) and attains a maximum value at some point \(c \in (a,b)\), then \(f'(c) = 0\)
Note this does not include the boundary, it is possible that derivative is not zero when it attains extreme value at the boundary.
2.3.3. Partial Differentiability
Proposition (continuous partial implies continuous total differentiability) If all the partial derivatives \(\frac{\partial f}{\partial x_j}\) exists and continuous on \(x_0\), then \(f\) is differentiable at \(x_0\) and the linear transformation \(f'(x_0)\) is defined by
However, simple existence of partial derivatives does NOT imply total differentiability. Counter Example (existence of all partial does NOT imply even continuity) Consider the following case \(\(f(x, y) = \begin{cases} \frac{xy}{x^2+y^2} \text { if } (x,y) \neq (0,0) \\ 0 \text{ if } (x,y) = (0,0) \end{cases}\)\)
In this case, partial derivative exists for \(x,y\), but it is not even continuous. of course not differentiable nor continuously differentiable.
Theorem (mixed partial derivative are equal in \(C^2\) ) For a twice continuously differentiable function \(f\), the mixed second partial derivatives are equal:
2.3.4. Analytic Function
Definition (class \(C^k, C^{\infty}\)) The class \(C^k\) is a set of function \(f\) where its derivatives \(f', f'', ..., f^{(k)}\) exist and are continous. It is said to be in the class \(C^{\infty}\) or smooth, if The function has deriavtives of all orders.
A related concept is analytic function. For any \(C^{\infty}\) function, a Taylor seris is defined at arbitrary points, but its neighbor is not guaranteed to converge. To gaurantee convergence, it needs to be an analytic function.
Definition (analytic) A function is said to be analytic at \(x_0\) if there exists a neighborhood of \(x_0\) such that within the neighborhood, a series is convergent to \(f(x)\)
Defintion (analytic function, \(C^{\omega}\)) An analytic function is a smooth function that the Taylor series at any point \(x_0\) in its domain.
converges to \(f(x)\) for \(x\) in a neighborhood of \(x_0\) pointwise.
Note Analytic function is a strictly subset of smooth function, there are smooth functions that are not analytic function. Details will be given in the seris section.
smooth does not imply analytic
Consider the function
The Taylor series of this function is \(0\) everywhere, therefore it does not converges to the original \(f(x)\)
2.4. Mean Value Theorem
Mean Value Theorem is the crucial step in the proof of the Fundamental Theorem of Calculus
Theorem (Rolle) Let \(a<b\) be real numbers, and let \(g: [a,b] \to R\) be a continuous function which is differentiable on \((a,b)\). Suppose also that \(g(a)=g(b)\). Then there exists an \(x \in (a,b)\) such that \(g'(x) = 0\)
Mean Value Theorem can be reduced to Rolle's Theorem
Theorem (mean value theorem) If \(f: [a,b] \to R\) is continuous on \([a,b]\) and differentiable on \((a,b)\), then there exists a point \(c \in (a,b)\) where
Theorem (generalized mean value theorem) If \(f, g\) are continuously on the closed interval \([a,b]\) and differentiable on the open interval \((a,b)\), then there exists a point \(c \in (a,b)\) where
Theorem (L'Hospital) Let \(f,g\) be continous onan interval containing \(a\), \(f,g\) be differentiable on this interval with the possible exception of the point \(a\), it \(f(a)=g(a)=0\) or \(f(a)=g(a)=\infty\), then
3. Riemann Integral
Historically, the concept of integration was defined as the inverse process of differentiation. This approach, however, is unsatisfying because it results in a very limited number of functions that can be limited. For example, any function with a jump discontinuity cannot be integrated (because of Darboux's Theorem). Around 1850 in the work of Cauchy and Riemann, integration start to use the notion of area under the curve as a starting point for building rigorous definition of the integral.
The major function classes related to Riemann integral can be classified as follows. Note that integrable family is boarder than continous, but differentiable function is more narrow than continuous function.
counter example (integrable vs antidifferentiable)
Note that integrable function and function can be taken antiderivative are not equivalent. For example, the following example is integrable but not antidifferentiable.
The class of functions whose Antiderivative exists falls somewhere between the previous hierarchy, because every continuous function has antiderivative (guaranteed by the fundamental theorem), and function with antiderivative can be integrable, but not every integrable function has antiderivative. Note there are also some examples where functions are not continuous but has antiderivative.
3.1. Partitions
Definition (length of intervals) If \(I\) is a bounded interval, we define the length of \(I\), denoted \(|I|\). If \(I\) is one of tthe intervals \([a,b], (a,b), [a,b), (a,b]\) for some real numbers \(a<b\), then we define \(|I| := b-a\). Otherwise if \(I\) is a point or the empty set, we define \(|I|=0\)
Definition (Partitions) Let \(I\) be a bounded interval. A partition of \(I\) is a finite set \(P\) of bounded intervals contained in \(I\), such that every \(x \in I\) lies in exactly one of the bounded intervals \(J \in P\)
Theorem (length is finitely additive) Let \(I\) be a bounded interval, n be a natural number, and let \(P\) be a partition of \(I\) of cardinality \(n\) Then
Definition (finer and coarser partitions) Let \(I\) be a bounded interval, \(P, P'\) be two partitions of \(I\). We say that \(P'\) is finer than \(P\) if for every \(J \in P'\), there exists a \(K \in P\) such that \(J \subseteq K\)
Definition (common refinement) Let \(I\) be a bounded interval, \(P, P'\) be two partitions of \(I\). We define the common refinement \(P \# P'\) to be the set
Definition (lower sum, upper sum) The lower sum, upper sum of a bounded function \(f\) with respect to partition \(P\) is given by
where \(m_k := \inf\{ f(x): x\in [x_{k-1}, x_k] \}\), \(M_k := \sup\{ f(x): x\in [x_{k-1}, x_k] \}\)
lower sum, upper sum
Define \(f: [0, 1] \to R\) by \(f(x) = x^2\), let \(P_n\) denote the partition \(0, 1/n, 2/n, ..., 1\), then
3.2. Integrability
Definition (upper integral, lower integral) Let \(\mathcal{P}\) be the collection of all possible partitions of the interval \([a,b]\). The upper integral, lower integral of \(f\) is defined to be
Definition (Riemann Integrability) A bounded function \(f\) defined on the interval \([a,b]\) is Riemann-integrable if \(U(f) = L(f)\). In this case
Criterion (Integrability) A bounded function \(f\) is integrable on \([a,b]\) iff for every \(\epsilon > 0\) there exists a partition \(P_{\epsilon}\) of \([a,b]\) such that
By applying this criterion, we know several classes of functions are integrable
Criterion (continuous function is integrable) If \(f\) is continuous on \([a,b]\), then it is integrable
Proof: This can be shown by controlling the difference between max and min within each range by exploiting uniform continuity, (which is guaranteed by the domain compactness)
Criterion (monotone function is integrable) If \(f: [a,b] \to \mathbb{R}\) is monotone on the set \([a,b]\), then it is integrable.
Criterion (function with a single discontinuity at an endpoint is integrable) If \(f: [a,b] \to R\) is bounded and \(f\) is integrable on \([c,b]\) for all \(c \in (a,b)\), then \(f\) is integrable on \([a,b]\)
A function with single discontinuity is still integrable
Consider the function with the domain \([0,1]\)
The upper sum \(U(f,P)\) is always 2, the lower sum is less than 2, but can be controlled by embedding the discontinuity into a small subinterval.
Any function with a finite number of discontinuity is integrable, some function with infinite discontinuities still might be integrable, depending on its measure of discontinuity.
Function with \(N\) discontinuity is integrable
Consider the function
$$f(x) = \begin{cases} 1 & \text{ if } x=1/n \text{ for some } n \in N \ 0 & \text{ otherwise} \end{cases}
This function has discontinuity of \(N\) and is integrable (by considering partition sequence with length of \(1/n^2\))
Function with \(Q\) discontinuity is integrable (Thomae's function)
Thomae's function has discontiuities at every rational number
Its discontinuity has measure 0 and is integrable.
Criterion (Lebesgue) Let \(f: [a,b] \to R\) be a bounded function. \(f\) is Riemann-integrable iff the set of noncontinous point has measure zero
Function with uncountable discontinuity might be integrable (Cantor)
Consider the indicator function over the Cantor set
where \(C\) is the cantor set. This function has uncountable discontinuity but with measure 0, therefore it is Riemann Integrable
Dirichlet's function is not Riemann Integrable
The Dirichlet's function is not Riemann Integrable because the noncontinuous point has measure 1.
3.3. Properties of the Integral
Theorem Assume \(f: [a,b] \to R\) is bounded and \(c \in (a,b)\). Then \(f\) is integrable on \([a,b]\) iff \(f\) is integrable on \([a,c]\) and \([c,b]\), in which case
Theorem Assume \(f\) and \(g\) are integrable functions on the interval \([a,b]\)
The function \(f+g\) is integrable on \([a,b]\)
For \(k \in R\), the function \(kf\) is integrable
If \(f(x) < g(x)\) on \([a,b]\)
The function \(|f|\) is integrable and
Theorem (Integral Limit Theorem) Assume that \(f_n \to f\) uniformly on \([a,b]\) and each \(f_n\) is integrable. Then \(f\) is integrable and
3.4. Fundamental Theorem of Calculus
Note that Stoke theorem can be thought as generalization of those fundamental theorems.
Theorem (fundamental theorem of calculus 1, Integration of Differentiation) if \(f: [a,b] \to R\) is integrable and \(f\) is antidifferentiable such that \(F: [a,b] \to R\) is antiderivative of \(f\) then
Theorem (fundamental theorem of calculus 2, Differentiation of Integration) Let \(g: [a,b] \to R\) be integrable and for \(x \in [a,b]\) define
then \(G\) is continuous on \([a,b]\) If \(g\) is continuous at some point \(c \in [a,b]\), then \(G\) is differentiable at \(c\) and \(G'(c) = g(c)\)
continuity is required in the 2nd statement
Consider the function
It has 1 discontinuity at 0 and its integration is
\(F(x)\) is not differentiable at this point!
Note the other side is not true: \(G\) being able to differentiate at \(c\) does not imply that \(g(c)\) is continuous.
Lemma (Integration by parts) Assume \(f(x), g(x)\) are continuously differentiable, then
Note the requirement is a bit stronger than necessary.
Lemma (Integration by substitution) Assume \(f(x)\) is continuous, and \(g(x)\) is continuously differentiable, then
proof of integration by substitution
As \(f\) is continuous and \(g\) is continuously differentiable. We know \(f\) has an antiderivative \(F\) such that \(F' = f\) and \(F(g(x))\) is well-defined. By chain rule, we know
which is also continuous. Therefore by the fundamental theorem
Also we know by the fundamental theorem
Therefore the integration by substitution is proved.
Leibnitz's rule is an application of the Fundamental Theorem and chain rules. It characterizes the interchanging the differnetiation and integral.
Theorem (Leibnitz's Rule) If \(f(x, \theta), a(\theta), b(\theta)\) are differentiable with respect to \(\theta\), then
If both \(a(\theta), b(\theta)\) are constant, we have a special case of Leibnitz's Rule \(\(\frac{d}{d\theta} \int_{a}^{b} f(x, \theta) dx = \int_{a}^{b} \frac{\partial}{\partial \theta} f(x, \theta) dx\)\)
4. Reference
[1] Tao, Terence. Analysis. Vol. 1. Hindustan Book Agency, 2006.
[2] Tao, Terence. Analysis. Vol. 2. Hindustan Book Agency, 2006.
[3] Abbott, Stephen. Understanding analysis. Vol. 2. New York: Springer, 2001.
[4] Lax, Peter D., and Maria Shea Terrell. Multivariable Calculus with Applications. Springer, 2017.
[5] 杉浦光夫. "解析入門 I." 東京大学出版会
[6] 杉浦光夫. "解析入門 II." 東京大学出版会