跳到主要内容

5 Definite Integral

5.1 The Concept of the Definite Integrals

Darboux Integral

Consider a bounded function ff on a closed interval [a,b][a,b], and a partition of [a,b][a,b]:

P:a=x0<x1<<xN=b.\begin{equation*} P: \quad a = x_0 < x_1 < \cdots < x_N = b. \end{equation*}

For each subinterval [xk1,xk][x_{k-1}, x_k], define

Mk=supx[xk1,xk]f(x),mk=infx[xk1,xk]f(x).\begin{equation*} M_k = \sup_{x \in [x_{k-1},x_k]} f(x), \qquad m_k = \inf_{x \in [x_{k-1},x_k]} f(x). \end{equation*}

The upper Darboux sum of ff with respect to the partition PP is

S(f,P)=k=1NMk(xkxk1),\begin{equation*} \overline{S}(f,P) = \sum_{k=1}^{N} M_k (x_k - x_{k-1}), \end{equation*}

and the lower Darboux sum is

S(f,P)=k=1Nmk(xkxk1).\begin{equation*} \underline{S}(f,P) = \sum_{k=1}^{N} m_k (x_k - x_{k-1}). \end{equation*}

Define

S(f)=supPS(f,P),S(f)=infPS(f,P),\begin{equation*} \underline{S}(f) = \sup_P\, \underline{S}(f,P), \qquad \overline{S}(f) = \inf_P\, \overline{S}(f,P), \end{equation*}

where the supremum and infimum are taken over all partitions PP of [a,b][a,b].

  • S(f)\underline{S}(f) is called the Darboux lower integral of ff on [a,b][a,b];
  • S(f)\overline{S}(f) is called the Darboux upper integral of ff on [a,b][a,b].

If

S(f)=S(f),\begin{equation*} \underline{S}(f) = \overline{S}(f), \end{equation*}

then ff is said to be Darboux integrable on [a,b][a,b]. The common value is denoted by

abf(x)dx,\begin{equation*} \int_a^b f(x)\,dx, \end{equation*}

and is called the Darboux integral of ff on [a,b][a,b].

Theorem 5.1. For any two partitions P,QP, Q of the interval [a,b][a,b], let PQP \lor Q denote the partition consisting of all points in PQP \cup Q. We call PQP \lor Q the common refinement of PP and QQ. Then

S(f,Q)    S(f,PQ)    S(f,PQ)    S(f,P).\begin{equation*} \underline{S}(f, Q) \;\le\; \underline{S}(f, P \lor Q) \;\le\; \overline{S}(f, P \lor Q) \;\le\; \overline{S}(f, P). \end{equation*}

Theorem 5.2. A bounded function ff on the closed interval [a,b][a,b] is Darboux integrable if and only if for every ε>0\varepsilon > 0, there exists a partition PP of [a,b][a,b] such that

S(f,P)S(f,P)<ε.\begin{equation*} \overline{S}(f,P) - \underline{S}(f,P) < \varepsilon. \end{equation*}

Theorem 5.3. A bounded function ff on the closed interval [a,b][a,b] is Darboux integrable, and IRI \in \mathbb{R} is its Darboux integral, if and only if for every ε>0\varepsilon > 0, there exists a partition PP of [a,b][a,b] such that

Iε<S(f,P)IS(f,P)<I+ε.\begin{equation*} I - \varepsilon < \underline{S}(f,P) \le I \le \overline{S}(f,P) < I + \varepsilon. \end{equation*}

Theorem 5.4. Every monotonic function on a bounded closed interval [a,b][a,b] is Darboux integrable. More precisely, for every ε>0\varepsilon > 0, there exists δ>0\delta > 0 such that for any partition

a=x0<x1<<xN=b,\begin{equation*} a = x_0 < x_1 < \cdots < x_N = b, \end{equation*}

if

P:=max1kN(xkxk1)<δ,\begin{equation*} \|P\| := \max_{1 \le k \le N} (x_k - x_{k-1}) < \delta, \end{equation*}

then

S(f,P)S(f,P)<ε.\begin{equation*} \overline{S}(f,P) - \underline{S}(f,P) < \varepsilon . \end{equation*}

Theorem 5.5. If ff is a bounded function on the closed interval [a,b][a,b], and ff is continuous on the open interval (a,b)(a,b), then ff is Darboux integrable on [a,b][a,b]. More precisely, for every ε>0\varepsilon > 0, there exists δ>0\delta > 0 such that for any partition P:a=x0<x1<<xN=bP: a = x_0 < x_1 < \cdots < x_N = b,

if

P:=max1kN(xkxk1)<δ,\begin{equation*} \|P\| := \max_{1 \le k \le N} (x_k - x_{k-1}) < \delta, \end{equation*}

then

S(f,P)S(f,P)<ε.\begin{equation*} \overline{S}(f, P) - \underline{S}(f, P) < \varepsilon. \end{equation*}

Corollary 1. If a bounded function ff on the closed interval [a,b][a,b] has only finitely many discontinuities, then ff is Darboux integrable on [a,b][a,b].

Riemann Integral

For a function f:[a,b]Rf:[a,b]\to \mathbb{R}, a partition of the interval [a,b][a,b] is

P: a=x0<x1<<xN=b.\begin{equation*} P:\ a = x_0 < x_1 < \cdots < x_N = b. \end{equation*}

Along with a set of sample points

ξ={ξi}i=1N,(ξi[xi1,xi]),\begin{equation*} \xi = \{\xi_i\}_{i=1}^{N}, \qquad (\xi_i \in [x_{i-1}, x_i]), \end{equation*}

the Riemann sum is defined as

S(f,P,ξ)=i=1Nf(ξi)(xixi1).\begin{equation*} S(f, P, \xi) = \sum_{i=1}^{N} f(\xi_i)(x_i - x_{i-1}). \end{equation*}

A function f:[a,b]Rf : [a,b] \to \mathbb{R} is said to be Riemann integrable on [a,b][a,b] if there exists a real number IRI \in \mathbb{R} such that for every ε>0\varepsilon > 0, there exists δ>0\delta > 0 with the property that for any tagged partition (P,ξ)(P,\xi) of [a,b][a,b], if the norm of the partition satisfies P<δ\|P\| < \delta, then

S(f,P,ξ)I<ε.\begin{equation*} \left| S(f, P, \xi) - I \right| < \varepsilon. \end{equation*}

In this case, we call II the Riemann integral of ff on [a,b][a,b], and write

I=abf(x)dx.\begin{equation*} I = \int_a^b f(x)\, dx. \end{equation*}

Theorem 5.6. For a function f:[a,b]Rf:[a,b]\to\mathbb{R}, the following statements are equivalent:

(1)f\text{(1)}\quad f is Riemann integrable on [a,b][a,b];

(2)f\text{(2)}\quad f is bounded on [a,b][a,b], and Darboux integrable;

(3)f\text{(3)}\quad f is bounded on [a,b][a,b], and the set of all discontinuities of ff in [a,b][a,b] forms a set of measure zero.

5.2 Properties and Computation of Definite Integrals

Theorem 5.7 (Newton-Leibniz Formula). If fR[a,b]f \in \mathcal{R}[a,b] and FF is an antiderivative of ff, then

abf(u)du=F(b)F(a).\begin{equation*} \int_a^b f(u)\,du = F(b) - F(a). \end{equation*}

Theorem 5.8. If ff is Riemann integrable on [a,b][a,b] and on [b,c][b,c], then ff is integrable on [a,c][a,c], and

abf(x)dx+bcf(x)dx=acf(x)dx.\begin{equation*} \int_a^b f(x)\,dx + \int_b^c f(x)\,dx = \int_a^c f(x)\,dx. \end{equation*}

Theorem 5.9. Assume fC[a,b]f \in \mathcal{C}[a,b], gR[a,b]g \in \mathcal{R}[a,b], and g(x)0g(x) \ge 0 for all x[a,b]x \in [a,b]. Then there exists ξ[a,b]\xi \in [a,b] such that

abf(x)g(x)dx=f(ξ)abg(x)dx.\begin{equation*} \int_a^b f(x)g(x)\,dx = f(\xi)\int_a^b g(x)\,dx. \end{equation*}

Theorem 5.10. Assume that fR[a,b]f \in \mathcal{R}[a,b]. Define

F(x)=axf(u)du.\begin{equation*} F(x) = \int_a^x f(u)\,du. \end{equation*}

Then FC[a,b]F \in \mathcal{C}[a,b]. Then

(1)\text{(1)}\quadThen the integral with variable upper limit F(x)=axf(t)dtF(x) = \int_a^x f(t) \mathrm{d}t is continuous on [a,b][a, b].

(2)\text{(2)}\quadLet function ff be continuous at a point x0[a,b]x_0 \in [a, b]. Then FF is differentiable at x0x_0, and F(x0)=f(x0).F'(x_0) = f(x_0).

Theorem 5.11. Let function ff be continuous on [a,b][a, b]. Then

ddxaxf(t)dt=f(x)(axb).\begin{equation*} \frac{\mathrm{d}}{\mathrm{d}x} \int_a^x f(t) \mathrm{d}t = f(x) \quad (a \leqslant x \leqslant b). \end{equation*}

Theorem 5.12. Let f,gR[a,b]f, g \in \mathcal{R}[a,b]. Then

abf(x)g(x)dxabf(x)2dx  abg(x)2dx.\begin{equation*} \left| \int_a^b f(x)g(x)\,dx \right| \,\le\, \sqrt{\int_a^b |f(x)|^2\,dx}\; \sqrt{\int_a^b |g(x)|^2\,dx}. \end{equation*}

Theorem 5.13. If f(n+1)R[a,b]f^{(n+1)} \in \mathcal{R}[a,b], then for every x[a,b]x \in [a,b],

f(x)=f(a)+f(a)(xa)++f(n)(a)n!(xa)n+axf(n+1)(t)n!(xt)ndt.\begin{equation*} f(x) = f(a) + f'(a)(x-a) + \cdots + \frac{f^{(n)}(a)}{n!} (x-a)^n + \int_{a}^{x} \frac{f^{(n+1)}(t)}{n!} (x - t)^n \, dt. \end{equation*}

Theorem 5.14. Suppose that ff is continuous on an interval II, a,bIa, b \in I, and φ:[α,β]I\varphi : [\alpha, \beta] \to I satisfies φ(α)=a\varphi(\alpha) = a, φ(β)=b\varphi(\beta) = b, and φR[α,β]\varphi' \in \mathcal{R}[\alpha, \beta]. Then

abf(x)dx=αβf(φ(t))φ(t)dt.\begin{equation*} \int_a^b f(x)\,dx = \int_\alpha^\beta f(\varphi(t))\, \varphi'(t)\, dt. \end{equation*}

5.3 Improper Integrals

Improper Integral

Let f:[a,+)Rf:[a,+\infty)\to \mathbb{R} be Riemann integrable on every bounded closed interval [a,A][a,A].

If

limA+aAf(x)dx\begin{equation*} \lim_{A\to+\infty} \int_a^A f(x)\,dx \end{equation*}

exists, then the improper integral a+f(x)dx\displaystyle \int_a^{+\infty} f(x)\,dx is said to converge, and we define

a+f(x)dx=limA+aAf(x)dx.\begin{equation*} \int_a^{+\infty} f(x)\,dx = \lim_{A\to+\infty} \int_a^A f(x)\,dx. \end{equation*}

If the limit does not exist, the improper integral is said to diverge.

Similarly, we define

ag(x)dx=limAAag(x)dx.\begin{equation*} \int_{-\infty}^a g(x)\,dx = \lim_{A\to-\infty} \int_A^a g(x)\,dx. \end{equation*}

Let f:RRf:\mathbb{R}\to\mathbb{R} be Riemann integrable on every bounded closed interval [a,b][a,b].

If both improper integrals

a+f(x)dx,af(x)dx\begin{equation*} \int_a^{+\infty} f(x)\,dx, \qquad \int_{-\infty}^{a} f(x)\,dx \end{equation*}

converge, then the improper integral

+f(x)dx\begin{equation*} \int_{-\infty}^{+\infty} f(x)\,dx \end{equation*}

is said to converge, and we define

+f(x)dx=af(x)dx+a+f(x)dx.\begin{equation*} \int_{-\infty}^{+\infty} f(x)\,dx = \int_{-\infty}^{a} f(x)\,dx + \int_{a}^{+\infty} f(x)\,dx. \end{equation*}

Let f:[a,b)Rf:[a,b)\to\mathbb{R} be unbounded, but Riemann integrable on every bounded closed interval [a,bδ][a,b-\delta] for δ>0\delta>0. In this case, the integral abf(x)dx\displaystyle \int_a^b f(x)\,dx is called an improper integral, and bb is called an improper point.

If

limδ0+abδf(x)dx\begin{equation*} \lim_{\delta\to 0^+} \int_a^{\,b-\delta} f(x)\,dx \end{equation*}

exists, then the improper integral abf(x)dx\displaystyle \int_a^b f(x)\,dx is said to converge, and we define

abf(x)dx=limδ0+abδf(x)dx.\begin{equation*} \int_a^b f(x)\,dx = \lim_{\delta\to 0^+} \int_a^{\,b-\delta} f(x)\,dx. \end{equation*}

Similarly, one may define improper integrals where the left endpoint is an improper point.

Examples

  1. 1+1xpdx \begin{equation*} \int_{1}^{+\infty} \frac{1}{x^p} \mathrm{d}x \end{equation*}

    Solution:

    1A1xpdx={1A1pp1,p1,lnA,p=1.\begin{equation*} \int_{1}^{A} \frac{1}{x^p} \mathrm{d}x = \begin{cases} \frac{1 - A^{1-p}}{p-1}, & p \neq 1, \\ \ln A, & p = 1. \end{cases} \end{equation*}

    Therefore, the integral converges if and only if p>1p > 1. When it converges, the limit is:

    1+1xpdx=1p1\begin{equation*} \int_{1}^{+\infty} \frac{1}{x^p} \mathrm{d}x = \frac{1}{p-1} \end{equation*}
  2. 011xpdx \begin{equation*} \int_{0}^{1} \frac{1}{x^p} \mathrm{d}x \end{equation*}

    Solution:

    δ11xpdx={1δ1p1p,p1,lnδ,p=1.\begin{equation*} \int_{\delta}^{1} \frac{1}{x^p} \mathrm{d}x = \begin{cases} \frac{1 - \delta^{1-p}}{1-p}, & p \neq 1, \\ -\ln \delta, & p = 1. \end{cases} \end{equation*}

    Therefore, the integral converges if and only if p<1p < 1. When it converges, the limit is:

    011xpdx=11p\begin{equation*} \int_{0}^{1} \frac{1}{x^p} \mathrm{d}x = \frac{1}{1-p} \end{equation*}

Theorems 5.15. Suppose aωf(x)g(x)dx\int_{a}^{\omega} f'(x)g(x) \mathrm{d}x converges and limxωf(x)g(x)\lim\limits_{x \to \omega^-} f(x)g(x) converges. Then aωf(x)g(x)dx\int_{a}^{\omega} f(x)g'(x) \mathrm{d}x converges, and:

aωf(x)g(x)dx=f(x)g(x)aωaωf(x)g(x)dx.\begin{equation*} \int_{a}^{\omega} f(x)g'(x) \mathrm{d}x = f(x)g(x) \bigg|_{a}^{\omega} - \int_{a}^{\omega} f'(x)g(x) \mathrm{d}x. \end{equation*}

Theorem 5.16. The improper integral a+f(x)dx\int_{a}^{+\infty} f(x) \mathrm{d}x converges if and only if for any ε>0\varepsilon > 0, there exists Nε>0N_{\varepsilon} > 0 such that for any A2>A1>NεA_2 > A_1 > N_{\varepsilon}, the following holds:

A1A2f(x)dx<ε.\begin{equation*} \left| \int_{A_1}^{A_2} f(x) \mathrm{d}x \right| < \varepsilon. \end{equation*}

Absolutely Convergent

An improper integral a+f(x)dx\int_{a}^{+\infty} f(x) \mathrm{d}x is said to be absolutely convergent, or ff is absolutely integrable on the interval [a,+)[a, +\infty), if the improper integral a+f(x)dx\int_{a}^{+\infty} |f(x)| \mathrm{d}x converges.

From the inequality:

A1A2f(x)dxA1A2f(x)dx\begin{equation*} \left| \int_{A_1}^{A_2} f(x) \mathrm{d}x \right| \leq \int_{A_1}^{A_2} |f(x)| \mathrm{d}x \end{equation*}

It is known that any absolutely convergent improper integral is also convergent.

Definition. An improper integral is said to be conditionally convergent if it converges but is not absolutely convergent.

Theorem 5.17. Let ff be Riemann integrable on any bounded closed interval [a,A][a, A], and let gg be monotonic. If:

  • aAf(x)dx\int_{a}^{A} f(x) \mathrm{d}x is bounded for all A>aA > a,
  • limx+g(x)=0\lim\limits_{x \to +\infty} g(x) = 0,

then the improper integral a+f(x)g(x)dx\int_{a}^{+\infty} f(x)g(x) \mathrm{d}x converges.

Theorem 5.18. Let ff be Riemann integrable on any bounded closed interval [a,A][a, A], and let gg be monotonic. If:

  • The improper integral a+f(x)dx\int_{a}^{+\infty} f(x) \mathrm{d}x converges,
  • g(x)g(x) is bounded on the interval [a,+)[a, +\infty),

then the improper integral a+f(x)g(x)dx\int_{a}^{+\infty} f(x)g(x) \mathrm{d}x converges.

5.4 Applications of Definite Integrals

The Gamma Function

For any α>0\alpha > 0:

Γ(α)=0+xα1exdx\begin{equation*} \Gamma(\alpha) = \int_{0}^{+\infty} x^{\alpha-1}e^{-x} \mathrm{d}x \end{equation*}

The function satisfies Γ(α+1)=αΓ(α)\Gamma(\alpha+1) = \alpha\Gamma(\alpha) and Γ(1)=1\Gamma(1) = 1. Therefore, for any positive integer nn:

Γ(n+1)=n!\begin{equation*} \Gamma(n+1) = n! \end{equation*}

The Beta Function

For any α>0,β>0\alpha > 0, \beta > 0:

B(α,β)=01xα1(1x)β1dx\begin{equation*} B(\alpha, \beta) = \int_{0}^{1} x^{\alpha-1}(1-x)^{\beta-1} \mathrm{d}x \end{equation*}

It satisfies B(α+1,β)=αβB(α,β+1)B(\alpha+1, \beta) = \frac{\alpha}{\beta} B(\alpha, \beta+1).

  • The relationship between the Beta function and the Gamma function is given by the following identity:
B(α,β)=Γ(α)Γ(β)Γ(α+β)\begin{equation*} B(\alpha, \beta) = \frac{\Gamma(\alpha)\Gamma(\beta)}{\Gamma(\alpha+\beta)} \end{equation*}

The area of a Plane Region

Area formula for a simple closed parametric curve

For a curve given by γ:x=x(t),y=y(t),αtβ\gamma: x = x(t),\, y = y(t),\, \alpha \le t \le \beta, the area enclosed by (\gamma) is

Area=αβy(t)x(t)dt=αβx(t)y(t)dt.\begin{align*} \text{Area} &= - \int_{\alpha}^{\beta} y(t)\, x'(t)\, dt = \int_{\alpha}^{\beta} x(t)\, y'(t)\, dt. \end{align*}

Averaging these two expressions yields

Area=12αβ(x(t)y(t)x(t)y(t))dt.\begin{align*} \text{Area} &= \frac{1}{2} \int_{\alpha}^{\beta} \bigl( x(t) y'(t) - x'(t) y(t) \bigr)\, dt. \end{align*}

Equivalently, using the determinant notation,

Area=12αβx(t)x(t)y(t)y(t)dt.\begin{align*} \text{Area} &= \frac{1}{2} \int_{\alpha}^{\beta} \begin{vmatrix} x(t) & x'(t) \\ y(t) & y'(t) \end{vmatrix} dt. \end{align*}

Length of a Curve

Let x(t)=(x1(t),,xn(t))Rn\mathbf{x}(t) = (x_1(t), \dots, x_n(t)) \in \mathbb{R}^n be a parametrization of the curve γ\gamma, where αtβ\alpha \le t \le \beta. A partition of the interval [α,β][\alpha, \beta] is given by

P:α=t0<t1<<tn=β.\begin{equation*} \mathcal{P} : \alpha = t_0 < t_1 < \cdots < t_n = \beta. \end{equation*}

For points on the curve defined by

Pk=x(tk),\begin{equation*} P_k = \mathbf{x}(t_k), \end{equation*}

the length of the polygonal line P0P1PnP_0 P_1 \cdots P_n is

L(P)=k=1nPk1Pk=k=1nx(tk)x(tk1).\begin{equation*} L(\mathcal{P}) = \sum_{k=1}^{n} \| P_{k-1} P_k \| = \sum_{k=1}^{n} \| \mathbf{x}(t_k) - \mathbf{x}(t_{k-1}) \|. \end{equation*}

In coordinates, this becomes

L(P)=k=1n(x1(tk)x1(tk1))2++(xn(tk)xn(tk1))2.\begin{equation*} L(\mathcal{P}) = \sum_{k=1}^{n} \sqrt{ \bigl( x_1(t_k) - x_1(t_{k-1}) \bigr)^2 + \cdots + \bigl( x_n(t_k) - x_n(t_{k-1}) \bigr)^2 }. \end{equation*}

It is clear that as the partition is refined, the quantity L(P)L(\mathcal{P}) increases.

If the supremum supPL(P)\displaystyle \sup_{\mathcal{P}} L(\mathcal{P}) exists, then the curve γ\gamma is said to be rectifiable, and

L(γ)=supPL(P)\begin{equation*} L(\gamma) = \sup_{\mathcal{P}} L(\mathcal{P}) \end{equation*}

is called the length of the curve γ\gamma.

L(γ)=αβx(t)dt.\begin{equation*} L(\gamma) = \int_{\alpha}^{\beta} \| \mathbf{x}'(t) \| \, dt. \end{equation*}

Curvature and Torsion of a Curve

l=l(t)=αtx(s)ds\begin{equation*} l = l(t) = \int_{\alpha}^{t} \|\mathbf{x}'(s)\|\, ds \end{equation*} l(t)=x(t)>0\begin{equation*} l'(t) = \|\mathbf{x}'(t)\| > 0 \end{equation*} x~(l)=x(t(l))(the reparametrization by arc length)\begin{equation*} \tilde{\mathbf{x}}(l) = \mathbf{x}(t(l)) \quad \text{(the reparametrization by arc length)} \end{equation*} x~(l)=x(t(l))t(l)=x(t)l(t)=x(t)x(t)is the unit tangent vector\begin{equation*} \tilde{\mathbf{x}}'(l) = \mathbf{x}'(t(l))\, t'(l) = \frac{\mathbf{x}'(t)}{l'(t)} = \frac{\mathbf{x}'(t)}{\|\mathbf{x}'(t)\|} \quad \text{is the unit tangent vector} \end{equation*} sinΔθ2=12x~(l+Δl)x~(l)\begin{equation*} \sin\frac{\Delta \theta}{2} = \frac{1}{2}\, \|\tilde{\mathbf{x}}'(l+\Delta l) - \tilde{\mathbf{x}}'(l)\| \end{equation*} =12x~(l)Δl+o(Δl)\begin{equation*} = \frac{1}{2}\, \|\tilde{\mathbf{x}}''(l)\,\Delta l + o(\Delta l)\| \end{equation*} =12x~(l)Δl+o(Δl)\begin{equation*} = \frac{1}{2}\, \|\tilde{\mathbf{x}}''(l)\|\,|\Delta l| + o(\Delta l) \end{equation*} Δθ=x~(l)Δl+o(Δl)\begin{equation*} \Delta \theta = \|\tilde{\mathbf{x}}''(l)\|\,\Delta l + o(\Delta l) \end{equation*} limΔl0ΔθΔl=x~(l)\begin{equation*} \lim_{\Delta l \to 0} \frac{\Delta \theta}{\Delta l} = \|\tilde{\mathbf{x}}''(l)\| \end{equation*} x~(l)=d2dl2x(t(l))=ddt(ddtx(t)dtdl)dtdl\begin{equation*} \tilde{\mathbf{x}}''(l) = \frac{d^2}{dl^2}\mathbf{x}(t(l)) = \frac{d}{dt}\left( \frac{d}{dt}\mathbf{x}(t)\, \frac{dt}{dl} \right)\frac{dt}{dl} \end{equation*} =(x(t)x(t))1x(t)\begin{equation*} = \left( \frac{\mathbf{x}'(t)}{\|\mathbf{x}'(t)\|} \right)' \frac{1}{\|\mathbf{x}'(t)\|} \end{equation*} =(x(t)x(t)12x(t)2x(t),x(t)x(t)3)1x(t)\begin{equation*} = \left( \frac{\mathbf{x}''(t)}{\|\mathbf{x}'(t)\|} - \frac{1}{2}\, \frac{\mathbf{x}'(t)\, 2\langle \mathbf{x}'(t),\mathbf{x}''(t)\rangle}{\|\mathbf{x}'(t)\|^3} \right) \frac{1}{\|\mathbf{x}'(t)\|} \end{equation*} =(x(t)x(t)x(t)x(t)x(t),x(t))1x(t)2\begin{equation*} = \left( \mathbf{x}''(t) - \frac{\mathbf{x}'(t)}{\|\mathbf{x}'(t)\|} \left\langle \frac{\mathbf{x}'(t)}{\|\mathbf{x}'(t)\|}, \mathbf{x}''(t) \right\rangle \right) \frac{1}{\|\mathbf{x}'(t)\|^2} \end{equation*} limΔl0ΔθΔl=x~(l)=κ(curvature)\begin{equation*} \lim_{\Delta l \to 0} \frac{\Delta \theta}{\Delta l} = \|\tilde{\mathbf{x}}''(l)\| = \kappa \quad\text{(curvature)} \end{equation*} x~(l): curvature vectorκ=x~(l): curvature\begin{equation*} \tilde{\mathbf{x}}''(l): \ \text{curvature vector} \qquad \kappa = \|\tilde{\mathbf{x}}''(l)\|: \ \text{curvature} \end{equation*} T(l)=x~(l): unit tangent vector of the curve\begin{equation*} \mathbf{T}(l) = \tilde{\mathbf{x}}'(l): \ \text{unit tangent vector of the curve} \end{equation*} N(l)=x~(l)x~(l): (unit) principal normal vector of the curve\begin{equation*} \mathbf{N}(l) = \frac{\tilde{\mathbf{x}}''(l)}{\|\tilde{\mathbf{x}}''(l)\|}: \ \text{(unit) principal normal vector of the curve} \end{equation*} T(l)=κN(l)\begin{equation*} \mathbf{T}'(l) = \kappa \mathbf{N}(l) \end{equation*} 0=N(l),N(l)=2N(l),N˙(l)N˙(l) is orthogonal to N(l)\begin{equation*} 0 = \langle \mathbf{N}(l), \mathbf{N}(l) \rangle' = 2\langle \mathbf{N}(l), \mathbf{\dot{N}}(l) \rangle \qquad \Longrightarrow \qquad \mathbf{\dot{N}}(l) \ \text{is orthogonal to } \mathbf{N}(l) \end{equation*} 0=T(l),N(l)=κN(l),N(l)+T(l),N˙(l)=κ+T(l),N˙(l)\begin{equation*} 0 = \langle \mathbf{T}(l), \mathbf{N}(l) \rangle' = \kappa \langle \mathbf{N}(l), \mathbf{N}(l)\rangle + \langle \mathbf{T}(l), \mathbf{\dot{N}}(l) \rangle = \kappa + \langle \mathbf{T}(l), \mathbf{\dot{N}}(l) \rangle \end{equation*}

Fundamental Equations of Plane Curve in R2\mathbb{R}^2

N˙(l)=κT(l)\begin{equation*} \dot{\mathbf{N}}(l) = -\kappa \mathbf{T}(l) \end{equation*}

In arc-length parametrization, the moving frame (T,N)(\mathbf{T}, \mathbf{N}) of a plane curve in R2\mathbb{R}^2 satisfies the fundamental equations:

(T N)=(T N)(0κκ0)\begin{equation*} (\mathbf{T} \ \mathbf{N})' = (\mathbf{T} \ \mathbf{N}) \begin{pmatrix} 0 & -\kappa \\ \kappa & 0 \end{pmatrix} \end{equation*}

Given a curvature function κ(l)\kappa(l) and any orthonormal initial vectors T0,N0\mathbf{T}_0, \mathbf{N}_0, the above system has a unique solution.

The curve is then obtained from

x(l)=x0+0lT(s)ds\begin{equation*} \mathbf{x}(l) = \mathbf{x}_0 + \int_0^l \mathbf{T}(s)\, ds \end{equation*}

A plane curve is uniquely determined by its curvature.

Fundamental Equations of Plane Curve in R3\mathbb{R}^3

B(l)=T(l)×N(l): the (unit) binormal vector of the curve\begin{equation*} \mathbf{B}(l) = \mathbf{T}(l) \times \mathbf{N}(l): \ \text{the (unit) binormal vector of the curve} \end{equation*}

N˙(l)\dot{\mathbf{N}}(l) is orthogonal to N(l)\mathbf{N}(l)

T(l),N˙(l)=κ\begin{equation*} \langle \mathbf{T}(l), \dot{\mathbf{N}}(l) \rangle = -\kappa \end{equation*} N˙(l)=κT(l)+τB(l)\begin{equation*} \dot{\mathbf{N}}(l) = -\kappa \mathbf{T}(l) + \tau \mathbf{B}(l) \end{equation*} τ=N˙(l),B(l)(torsion of the curve)\begin{equation*} \tau = \langle \dot{\mathbf{N}}(l), \mathbf{B}(l) \rangle \quad \text{(torsion of the curve)} \end{equation*} B˙(l)=T˙(l)×N(l)+T(l)×N˙(l)=τT(l)×B(l)=τN(l)\begin{equation*} \dot{\mathbf{B}}(l) = \dot{\mathbf{T}}(l) \times \mathbf{N}(l) + \mathbf{T}(l) \times \dot{\mathbf{N}}(l) = \tau\, \mathbf{T}(l) \times \mathbf{B}(l) = -\tau\, \mathbf{N}(l) \end{equation*}

In arc-length parametrization, the moving frame (T,N,B)(\mathbf{T}, \mathbf{N}, \mathbf{B}) of a space curve in R3\mathbb{R}^3 satisfies:

(T N B)=(T N B)(0κ0κ0τ0τ0)\begin{equation*} (\mathbf{T}\ \mathbf{N}\ \mathbf{B})' = (\mathbf{T}\ \mathbf{N}\ \mathbf{B}) \begin{pmatrix} 0 & -\kappa & 0 \\ \kappa & 0 & -\tau \\ 0 & \tau & 0 \end{pmatrix} \end{equation*}

A space curve in R3\mathbb{R}^3 is uniquely determined by its curvature and torsion.

Frenet–Serret Formulas

T=xx\begin{equation*} \mathbf{T} = \frac{\mathbf{x}'}{\|\mathbf{x}'\|} \end{equation*} dtdl=1x\begin{equation*} \frac{dt}{dl} = \frac{1}{\|\mathbf{x}'\|} \end{equation*} T=xxxx3x,x=1x(xTT,x)\begin{equation*} \mathbf{T}' = \frac{\mathbf{x}''}{\|\mathbf{x}'\|} - \frac{\mathbf{x}'}{\|\mathbf{x}'\|^{3}} \langle \mathbf{x}', \mathbf{x}'' \rangle = \frac{1}{\|\mathbf{x}'\|} \left( \mathbf{x}'' - \mathbf{T}\langle \mathbf{T}, \mathbf{x}'' \rangle \right) \end{equation*} T˙=Tdtdl(curvature vector)\begin{equation*} \dot{\mathbf{T}} = \mathbf{T}'\frac{dt}{dl} \quad\text{(curvature vector)} \end{equation*} κ=1x2xTT,x\begin{equation*} \kappa = \frac{1}{\|\mathbf{x}'\|^{2}} \left\|\mathbf{x}'' - \mathbf{T}\langle\mathbf{T}, \mathbf{x}''\rangle\right\| \end{equation*} N=T˙T˙=xTT,xxTT,x\begin{equation*} \mathbf{N} = \frac{\dot{\mathbf{T}}}{\|\dot{\mathbf{T}}\|} = \frac{\mathbf{x}'' - \mathbf{T}\langle \mathbf{T},\mathbf{x}''\rangle} {\left\|\mathbf{x}'' - \mathbf{T}\langle \mathbf{T},\mathbf{x}''\rangle\right\|} \end{equation*} B=x×xx×x\begin{equation*} \mathbf{B} = \frac{\mathbf{x}' \times \mathbf{x}''}{\|\mathbf{x}' \times \mathbf{x}''\|} \end{equation*} B=x×xBB,x×xx×x\begin{equation*} \mathbf{B}' = \frac{\mathbf{x}' \times \mathbf{x}''' - \mathbf{B} \langle \mathbf{B}, \mathbf{x}' \times \mathbf{x}''' \rangle} {\|\mathbf{x}' \times \mathbf{x}''\|} \end{equation*} B˙=Bdtdl=τN\begin{equation*} \dot{\mathbf{B}} = \mathbf{B}'\frac{dt}{dl} = - \tau \mathbf{N} \end{equation*} τ=B˙,N\begin{equation*} \tau = -\langle \dot{\mathbf{B}}, \mathbf{N} \rangle \end{equation*} τ=x,T×Nx×x=x,Bx×x=x,x×xx×x2\begin{equation*} \tau = \frac{\langle \mathbf{x}''',\mathbf{T}\times\mathbf{N}\rangle} {\|\mathbf{x}'\times\mathbf{x}''\|} = \frac{\langle\mathbf{x}''',\mathbf{B}\rangle} {\|\mathbf{x}'\times\mathbf{x}''\|} = \frac{\langle \mathbf{x}''', \mathbf{x}' \times \mathbf{x}'' \rangle} {\|\mathbf{x}' \times \mathbf{x}''\|^{2}} \end{equation*}