Selected solutions to mathematical methods of classical mechanics

 
$\def\ae#1#2{\text{a.e.} {#1} \in {#2}} \def\brn{\mathbb{R}^N} \def\bR{\mathbb{R}} \def\bN{\mathbb{N}} \def\bC{\mathbb{C}} \def\bQ{\mathbb{Q}} \def\bZ{\mathbb{Z}} \def\bbf{\mathbb{F}} \def\ch{\mathcal{H}} \def\cb{\mathscr{B}} \def\cw{\mathcal{W}} \def\cd{\mathcal{D}} \def\cc{\mathcal{C}} \def\ck{\mathcal{K}} \def\cl{\mathcal{L}} \def\cs{\mathcal{S}} \def\cm{\mathcal{M}} \def\scl{\mathscr{L}} \def\lpr{L^2_{\br}[0,\pi]} \def\lpc{L^2_{\bc}[0,\pi]} \def\lppr{L^2_{\br}[-\pi,\pi]} \def\lppc{L^2_{\bc}[-\pi,\pi]} \def\lpf{L^2_{\bbf}[-\pi,\pi]} \def\cpr{C_{\br}[0,\pi]} \def\brk{\mathbb{R}^k} \def\brm{\mathbb{R}^m} \def\bpk{\mathbb{P}^{k-1}} \def\ixy{\langle x,y\rangle} \def\ixa{\langle x,a\rangle} \def\ixx{\langle x,x\rangle} \def\iyx{\langle y,x\rangle} \def\ixz{\langle x,z\rangle} \def\iyz{\langle y,z\rangle} \def\cat{\text{\rm cat}} \def\ve{\varepsilon} \def\bve{\overline{\ve}} \def\ox{\overline{X}} \def\tf{\tilde{f}} \def\tg{\tilde{g}} \def\th{\tilde{h}} \def\tx{\tilde{X}} \def\ta{\tilde{A}} \def\vp{\varphi} \def\tvp{\tilde{\vp}} \def\wc{\rightharpoonup} \def\hd{\hat{\delta}} \def\hve{\hat{\ve}} \def\unk{u_{n_k}} \def\bu{\boldsymbol{u}} \def\be{\boldsymbol{e}} \def\ba{\boldsymbol{a}} \def\bq{\boldsymbol{q}} \def\bv{\boldsymbol{v}} \def\bw{\boldsymbol{w}} \def\bx{\boldsymbol{x}} \def\bxd{\boldsymbol{\dot{x}}} \def\bxdd{\boldsymbol{\ddot{x}}} \def\ker{\text{Ker}\,} \def\tr{\text{tr}\,} \def\Re{\mathfrak{Re}} \def\Im{\mathfrak{Im}} \def\esssup{\text{ess}\,\sup} \def\sspan{\text{span}} \newcommand{\vat}{\bigg\arrowvert} \def\arcsinh{\text{arc}\sinh}$

Experimental facts

The principles of relativity and determinacy

The galilean group and Newton’s equations

证明空间$\bR\times\bR^3$的每个伽利略变换都可唯一得表示为一个旋转, 一个平移, 一个匀速运动的组合 (g = g1 ∘ g2 ∘ g3 )(所以, 伽利略变换群的维数是3 + 4 + 3 = 10)

Proof: 对任意伽利略变换g, 先考虑$g(t,\boldsymbol{0})$g是线性的, 可知, 有$\bv$, $\ba$,s, 使得$g(t,\boldsymbol{0}) = (t-s, \ba + t\bv) \in \bR\times\bR^3$. 下面考虑$\bR^3$中的标准正交基$\be_1,\be_2,\be_3$. 令$g(t,\be_i) - g(t,\boldsymbol{0})= (t-s,\bu_i)$. 由于同时事件空间的差是一个3维线性空间(定义).

所以$g(t,(a,b,c)) - g(t,\boldsymbol{0})= (t-s, G(a,b,c)^T)$, 其中$G = (\bu_1,\bu_2,\bu_3)$. 令$g_3(t,\bx) = (t,G \bx)$, $g_2(t,\bx) = (t,t \bv +\bx)$, $g_1(t,\bx) = (t-s,\ba+\bx)$. 则g = g1 ∘ g2 ∘ g3, 即为所求. ◻

证明一切伽利略空间均互相同构, 而特别是同构于坐标空间$\bR\times\bR^3$.

Proof: 由于$t:\bR^4\rightarrow \bR$, 是一个 Hilbert 空间上的线性范函, 有如下分解:$\bR^4 = \ker{t}\oplus\ker{t}^\perp$, 且ker t的维数是3. 设P, Q为对应的投影映射, 使$\bx = P\bx+Q\bx$, 其中$P\bx \in\ker{t}$, $Q\bx \in\ker{t}^\perp$. 又令Rker t$\bR^3$的一个等距同构.

任取A4中一点$\bx_0$. 可作$F: A^4 \rightarrow \bR\times\bR^3$, 其中$F(\bx) = (t(\bx-\bx_0),R\,P\bx)$. 即为所求同构. ◻

平面上一个可微运动的轨迹是否有图3形状? 其加速度矢量是否有图上所示的值?

可能, 如$\bx(t) = (t^3,t^2)$; 不可能 $\bxdd$ 指向曲线凹处.

若一力学系只由一点组成, 证明它在惯性参考系中的加速度为零.

Proof: 由例1,例2,加速度不依赖于$\bx,\bxd$t, 所以$F(\bx,\bxd,t)=\bv$, 又由旋转不变性对任意G, $\bv=F(G\bx,G\bxd,t)=GF(\bx,\bxd,t)=G\bv$ 所以只能有$\bv=0$. 即加速度为零. ◻

设一个力学系由两点组成, 在初始时刻其速度均为零, 证明他们将停留在初始时刻联结他们的直线上.

Proof: 只要证明$\bxdd_i$平行于$\bx_1-\bx_2$. 显然$\bx_1-\bx_2$非零, 可令$\bu=(\bx_1-\bx_2)/\left\|\bx_1-\bx_2\right\|$, 对任意与$\bu$垂直的单位向量$\bv$,令$\bw = \bu\times\bv$$G = (\bu, \bv,\bw)^T$, 为一个旋转变换. 由例3的相同方法,可知$F=\bxdd = (\bxdd_1,\bxdd_2)$ 只依赖于$\bx_1-\bx_2$$\bxd_1-\bxd_2$. 则$G\bxdd = G\,F(\bx_1-\bx_2,\bxd_1-\bxd_2,t) = F(G\bu,G\boldsymbol{0},t) \equiv F((\|\bu\|,0,0)^T,\boldsymbol{0},t)$. 由$\bv$的任意性, $\bxdd_i$垂直于$\bv$. 又考虑到$\bx_1 = \bx_2 = 0$ 所以$\bx_i(t)$始终在他们的连线上. ◻

设一个力学系由三点组成, 在初始时刻其速度均为零, 证明它们总位于初始时刻包含它们的平面上.

Proof: 类似前一题, 构造正交变换G. 令$\bu$是垂直于$\bx_1-\bx_2,\bx_1-\bx_3$的向量. $\bv$平行于$\bx_1-\bx_2$, $\bw=\bu\times\bv$, $G = (\bu,\bv,\bw)$,$G' = (-\bu,\bv,\bw)$,$\bx = (\bxdd_1,\bxdd_2,\bxdd_3)$. 由书中例, F只与$\bx_1-\bx_2,\bx_1-\bx_3$有关. 则(中部分不必考虑具体形式) $$G\bxdd=F(G(\bx_1-\bx_2,\bx_1-\bx_3),G(\bxd_1,\bxd_2,\bxd_3),t) = F((0,0,0;\star),\boldsymbol{0},t) = G' \bxdd$$, 得$\bu\bxdd = \boldsymbol{0}$, 同上题得证. ◻

设一个力学系由两点组成, 证明, 对任意初始条件都存在一个惯性参考系, 使得其中这两个点总位于一个固定平面上.

Proof: 设两点初始时刻位置和速度分别为$\bx_1,\bx_2$$\bxd_1,\bxd_2$, 考虑一个以$\bxd_1$运动的参考系, 在其中$\bxd_1=\boldsymbol{0}$. 考虑单位向量$\bu$同时垂直于$\bx_1-\bx_2$$\bxd_2$. 另作单位向量$\bv,\bw$, 使它们成为两两正向量. 令$G = (\bu, \bv,\bw)$, $G' = (-\bu,\bv,\bw)$. $\bxdd = (\bxdd_1,\bxdd_2)$. 同时, F 只与$\bx_1-\bx_2$, $\bxd_1$,$\bxd_2$有关. 于是 $$G\bxdd = F(G(\bx_1-\bx_2),G\bxd_1,G\bxd_2) = F((0,\star)^T, \boldsymbol{0},(0,\star)) = F(G'(\bx_1-\bx_2),G'\bxd_1,G'\bxd_2) = G'\bxdd$$ 于是$\bxdd$$\bu$垂直, 结合$\bxd_1,\bxd_2$$\bu$垂直即得, 运动始终在和$\bu$垂直的平面内. ◻

证明镜子里的力学和我们的力学是一样的

Proof:$$G = \left( \begin{array}{ccc} -1 & 0 & 0 \\ 0 & 1 & 0 \\ 0 & 0 & 1 \\ \end{array} \right)$$ 即得. ◻

惯性参考系类是否唯一

Proof: 将惯性参考系类理解为, 若两个参考系为同一类, 则它们之间有一个伽利略变换. 那么改变时间间隔或者时间方向的两个惯性参考系之间是没有伽利略变换的. 若有, 则这个变换不保持时间间隔数t, 所以不是伽利略变换. ◻

Examples of mechanical systems

Investigation of the equations of motion

Systems with one degree of freedom

Draw the phase curves for the “equation of a pendulum one a rotating axis”: $\bxdd = -\sin \bx + M$

Find the tangent lines to the branches of the critical level corresponding to maximal potential energy E = U(ξ)

由于dx/dt = y, $dy/dt = -\sin \bx + M$. $dy/dx = (-\sin\bx+M)/y$. 解常微分方程得. $y^2/2 = E + \cos\bx + M\bx$. 所以相图如下(M = 0.15).

By U(x) + y2/2 = E, dy/dx =  − U′(x)/y (Since E = U(ξ) > U(x),y ≠ 0). By L’Hospital’s Rule limx → ξdy/dx =  − U″(x)/y. So $y'(\xi) = \pm \sqrt{-U''(\xi)}$. Then tangent lines are $y = \pm \sqrt{-U''(\xi)}(x-\xi)$. ps我的做法是对U(x) + y2/2 = E求两次导数。

Let S(E) be the area enclosed by the closed phase curve coresponding to the energy level E. Show that the period of mothion along this curve is equal to $$T = \frac{dS}{dE}$$.

Proof: 由唯一性定理, 首先知道这样的相曲线与x轴只能有两个交点设为x1 < x2 (否则如果有一个中间点, 那么相曲线必然在那边断开了,建议画草图看一下). 另外可知对每个半平面(上下), xy是一一对应的 (否则, 有一个转折点,于是就有$\bxd=0$,这于$y=\bxd\neq0$即在上, 下半平面内矛盾). 它们与能级E有关. 考虑相曲线围成的面积 S = ∫x1x2ydx + ∫x2x1ydx. (注意后面一项的积分方向). 结合$y = \pm \sqrt{2(E-U)}$. 所以 $$dS/dE = \int_{x_1}^{x_2} \frac{dx}{\sqrt{2(E-U)}} + \int_{x_1}^{x_2} \frac{dx}{\sqrt{2(E-U)}} + y dx/dE \vat_{x_1}^{x_2} + y dx/dE \vat_{x_2}^{x_1}$$ 注意到yx1, x2处取零, 上式第一项就是运动从x1x2的时间, 后一项是x2x1的时间, 和即为一个周期的运动时间. 所以T = dS/dE.

ps由公式U(x) + y2/2 = E可以看到相曲线是关于x轴对称的, 所以$T/2=\int_{x_1}^{x_2} \frac{dx}{\sqrt{2(E-U)}}$由Green公式, S/2 = ∫x1x2ydx求导即得结论。 ◻

Let E0 be the value of the potential function at a minimum point ξ. Find the period T0 = limE → E0T(E) of small oscillations in a neighborhood of the point ξ.

Consider a periodic motion along the closed phase curve corresponding to the energy level E. Is it stable in the sense of Liapunov?

The Definition of Liapunove stability: (By Nonlinear Ordinary Differential Equations: nonlin Ordinary Diff Equat 3e P286)

Let $\bx^\ast(t)$ be a given real or complex solution vector of the n-dimentional system $\bxd = X(\bx,t)$. Then $\bx^\ast$ is Liapunov stable for t ≥ t0 if and only if, to each value of ε > 0, however small, there corresponds a value of δ > 0 (where δ may depend only on ε and t0) such that $$\|\bx(t_0) - \bx^\ast(t_0)\|<\delta \Rightarrow \|\bx(t) - \bx^\ast(t)\|<\varepsilon$$ for all t > t0, where $\bx(t)$ represents any other neighbouring solution.

Example:

Draw the image of the circle $x^2+(y-1)^2<\frac{1}{4}$ under the action of a transformation of the phase flow for the equations (a) of the “inverse pendulum”, $\ddot{x} = x$ and (b) of the “nonlinear pendulum”, $\ddot{x} = -\sin x$.

做首次积分, 由$\ddot{x} = x$dy/dx = x/y, 所以y2 − x2 = E. 考虑一支$\dot{x} = y = \sqrt{E+x^2}$. 解得$x = \sqrt{E} \sinh{(t+C)}$, 于是$y=\dot{x} = \sqrt{E} \cosh{(t+C)}$. 其中E = y2 − x2, $C = \arcsinh{(x/\sqrt{E})}$. 于是得到相流$g^t(x,y) = (\sqrt{E} \sinh{(t+C)},\sqrt{E}\cosh{(t+C)})$. 如图:

(b) 同样先首次积分, 用数值解法, 算出了图如下:

Systems with two degrees of freedom

Variational principles

Legendre transformations

Legendre transformation takes convex functions to convex functions.

Proof: 由 $d(p x-f(x)) / d x=0$ 得, $p=f^{\prime}(x)$, 所以 $d g / d p=x+p d x / d p-f^{\prime}(x) d x=x$. 于 是 $d^{2} g / d p^{2}=d x / d p$. 另一方面由 $d p / d x=f^{\prime \prime}(x), d x / d p=1 / f^{\prime \prime}(x)>0$ 所以 $g$ 也是凸的.

The Legendre transformation is involutive.

Proof: 考虑 $g(p)$ 的Legendre变换. 作函数 $G(x, p)=p x-g(p)$. 那么只要证明 $\max _{p} G(x, p)=$ $f(x)$. 首先, 若取 $x=x(p)$ (原Legendre变换,即 $f \rightarrow g$, 中的最大值点), 则利用原变换中的等式, 可 知 $\max _{p} G(x, p) \geq p x(p)-g(p)=f(x)$. 另一方面, 原Legendre变换中有对任意 $p, p x-f(x) \leq g(p)$, 于是有, $\max _{p} G(x, p)=\max _{p} p x-g(p) \leq f(x)$. 结合以上, 得结论.

Prove Liouville's formula $W=W_{0} e^{\int \operatorname{tr} A d t}$ for the Wronskian determinant of the linear system $\dot{\boldsymbol{x}}=A(t) \boldsymbol{x}$.

Proof:

Show that in a hamiltonian system it is impossible to have asymptotically stable equilibrum positions and asymptotically stable limit cycles in the phase space.

Proof: 若有, 对于某区域 $D$ 其体积将趋于 0 , 这与Liouville定理矛盾.

Consider the first digits of the numbers $2^{n}: 1,2,4,8,1,3,6,1,2,5,1,2,4, \ldots$ Does the digit 7 appears in this sequence? Which digit appears more often, 7 or $8 ?$ How many times more often?

Lagrangian mechanics on manifolds

Holonomic constraints

Differentiable manifolds

Show that SO(3) is homeomorphic to three-dimensional real projective space.

E. Noether’s theorem

Suppose that a particle moves in the field of the uniform helical line x = cos ϕ, y = sin ϕ, z = cϕ. Find the law of conservation corresponding to this helical symmetry.

任一质点若容许在螺线上运动, 且L在其上不变于是有容许映射h, 设$\bx = (\cos \phi,\sin\phi, c\phi)$. 则$h^s \bx = (\cos(\phi+s),\sin(\phi+s), c(\phi+s))$. 于是 $$dh^s\bx/ds \vat_{s = 0} = (-\sin\phi,\cos\phi,c) = (-y,x,c)$$. 又$M_3 = ([\bx,m\bxd],\be_3)= (m\bx,[\bx,\be_3]) = (m(\dot{x},\dot{y},\dot{z}),[(x,y,z),(0,0,1)]) = -m\dot{x}y + m\dot{y}x$ 于是I = mc + m(−y) + mx = cP3 + M3守恒.

Suppose that a rigid body is moving under its own inertia. Show that its center of mass moves linearly and uniformly. If the center of mass is at rest, then the angular momentum with repect to its is conserved

Proof: ◻

What quantity is conserved under the motion of a heavy rigid body if it is fixed at some point O? What if, in addition, the body is symmetric with respect to an axis passing through O?

Extend Noether’s theorem to non-autonomous lagrangian systems.

$M_1 = M\times \bR$为拓广的构形空间. 定义函数$L_1:TM_1\rightarrow \bR$$L\frac{dt}{d\tau}$. 局部坐标下: $$L_1\left(\bq,t,\frac{d\bq}{d\tau},\frac{dt}{d\tau}\right) = L\left(\bq,\frac{d\bq/d\tau}{dt/d\tau},t)dt\right)$$,

Differential forms

Exterior multiplication

Find the exterior k-th power of ω2.

$(\omega^2)^k =(\sum_{i=1}^{n} p_i\wedge q_i)^k =\sum_{i_1\cdots i_k} p_{i_1}\wedge q_{i_i}\wedge \cdots \wedge p_{i_k}\wedge q_{i_k}$.

考虑其中的项 pi1 ∧ qii ∧ ⋯ ∧ pik ∧ qik, 若i1ik 中有重复, 则为零. 若不充分, 将其变为形式: pi1 ∧ ⋯ ∧ pik ∧ qi1 ∧ ⋯ ∧ qik. 这需要经过1 + 2 + ⋯ + (k−1)次对换, 所以其符号为(−1)k(k−1)/2 再将{ik}按大小顺序排列, 此时由于p, q都做同样的置换, 所以这种置换必是偶的. 固定{ik}考虑共有多少这样的项?这是从kω2中取出ik组成排列的个数, 一共有k!个. 综合以上, 得 (ω2)k = (−1)k(k−1)/2i1 < ⋯ < ikpi1 ∧ ⋯ ∧ pik ∧ qi1 ∧ ⋯ ∧ qik .

Show that every differential 1-form on the line is the differential of some function.

Proof: $\bR$上的1-微分形式$\omega:T\bR \rightarrow \bR$, 在每一点x处, 可定义函数 f(x) = df(x,1). 显然f是连续的, 令F(x) = ∫0xf(t)dt. 则F1-微分形式为df = fdx, 与ω相同. ◻

Find differential 1-forms on the circle and the plane which are not the differential of any function.

Solution: 定义 1 微分形式 $\omega: T S \rightarrow \mathbb{R},$ 其中 $\omega(\boldsymbol{x}, t \boldsymbol{w})=t$ 其中 $\boldsymbol{w}$ 为切空间中与圆周相切, 成逆时针方向的单位切矢量.此不可能为某函数的微分, 因为圆周上连续函数必有最大值点, 在此点函数导数为零, 其1-形式也必须为零.

Show that every closed form on a vector space in an exterior derivative.

Proof: 证明按提示, 需要证明 $(k-1)$-形式 $p \omega^{k}$ 存在唯一, 其值为 $$ (p \omega)_{x}\left(\xi_{1}, \cdots, \xi_{k-1}\right)=\int_{0}^{1} \omega_{t_{x}}\left(x, t \xi_{1}, \cdots, t \xi_{k-1}\right) d t $$ 只对单形证明即可, 设单形 $c_{k-1}=\left\{\sum_{i=1}^{k-1} s_{i} \xi_{i}: 0 \leq s_{i} \leq 1, \forall i\right\}$, 则 $$ p c_{k-1, \boldsymbol{x}}=\left\{t\left(\boldsymbol{x}+\sum_{i=1}^{k-1} s_{i} \xi_{i}\right): 0 \leq s_{i} \leq 1, \forall i, 0 \leq t \leq 1\right\} $$ 由于\begin{array}{c} \int_{c_{k-1}} p \omega^{k}=\int_{c_{k-1}} p \omega^{k}\left(\xi_{1}, \cdots, \xi_{k-1}\right) d s_{1} \cdots, d s_{k-1} \\ \int_{p c_{k-1}} \omega^{k}=\int_{p c_{k-1}} \omega^{k}\left(\boldsymbol{x}+\sum_{i=1}^{k-1} s_{i} \xi_{i}, t \xi_{1}, \cdots, t \xi_{k-1}\right) d t d s_{1} \cdots d s_{k-1} \end{array}在 $s_{i}=0$ 处求导, 于是由定义即得上式. 事实上可直接定义上式为所求 $k-1$-形式. 还需证明 $d \circ p+p \circ d=1$. 在 $\mathbb{R}^{m}$ 上考虑 $k$ 形式. 只需考虑 $\omega^{k}=\phi(x) d x_{1} \wedge \cdots \wedge d x_{n}$. 计算 $p \omega^{k}$ 中项 $d x_{i_{1}} \wedge d x_{i_{k-1}}$ 的系数. 下面 $i$ 表示去掉指标 $i$. 将 $\boldsymbol{e}_{i}$ 的各组合代入得, $$ p \omega^{k}=\sum_{i}\left((-1)^{i+1} \int_{0}^{1} t^{k-1} \phi(t \boldsymbol{x}) x_{i} d t\right) d x_{1} \wedge \cdots \wedge \hat{d x}_{i} \wedge d x_{k} $$ 下面计算 $d p \omega$. 按运算规则. 项 $d x_{s} \wedge d x_{1} \wedge \cdots \wedge d \hat{x}_{i} \cdots \wedge d_{k}$ 前的系数为. 若 $s \neq 1, \cdots, k$. $$ a_{(s, 1, \cdot \hat{,}, \cdots, k)}=(-1)^{i+1} \int_{0}^{1} t^{k} \frac{\partial \phi}{\partial x_{s}}(t \boldsymbol{x}) x_{i} d t $$ 若 $s$ 在 $1, \cdots, k$ 中, 设为 $i$, 则会产生项 $$ \left((-1)^{i+1} \int_{0}^{1} t^{k} \frac{\partial \phi}{\partial x_{i}}(t \boldsymbol{x}) x_{i} d t+(-1)^{i+1} \int_{0}^{1} t^{k-1} \phi(t \boldsymbol{x}) d t\right) d x_{i} \wedge d x_{1} \wedge \cdots \wedge d \hat{x}_{i} \wedge \cdots \wedge d x_{k} $$再计算 $d \omega^{k}$. 显然 $$ d \omega^{k}=\sum_{s \neq 1, \cdots, k} \frac{\partial \phi}{\partial x_{s}}(\boldsymbol{x}) d x_{s} \wedge d x_{1} \wedge \cdots \wedge d_{k} $$ 考虑 $p d \omega$ 中项 $\left.d x_{s} \wedge d x_{1} \wedge \cdots \wedge d \hat{x}_{i} \wedge \cdots \wedge d_{k}\right)$ 的系数若 $s \neq 1, \cdots, k$. 则系数为 $$ (-1)^{i+2} \int_{0}^{1} t^{k} \frac{\partial \phi}{\partial x_{s}}(t \boldsymbol{x}) x_{i} d t $$ 而 $d x_{1} \wedge \cdots \wedge d_{k}$ 的系数为 $$ \sum_{s \neq 1, \cdots, k} \int_{0}^{1} t^{k} \frac{\partial \phi}{\partial x_{s}}(t \boldsymbol{x}) x_{s} d t $$ 合之得 $$ \begin{aligned} d p \omega^{k}+p d \omega^{k} &=\left(\int_{0}^{1} k t^{k-1} \phi(t \boldsymbol{x}) d t+\sum_{s=1, \cdots, m} \int_{0}^{1} t^{k} \phi(t \boldsymbol{x}) x_{s} d t\right) d x_{1} \wedge \cdots \wedge d x_{k} \\ &=\left(\int_{0}^{1} k t^{k-1} \phi(t \boldsymbol{x}) d t+\int_{0}^{1} t^{k} d \phi(t \boldsymbol{x})\right) d x_{1} \wedge \cdots \wedge d x_{k} \\ &=\left(\int_{0}^{1} k t^{k-1} \phi(t \boldsymbol{x}) d t+\left.t^{k} \phi(t \boldsymbol{x})\right|_{0} ^{1}-\int_{0}^{1} k t^{k-1} \phi(t \boldsymbol{x}) d t\right) d x_{1} \wedge \cdots \wedge d x_{k} \\ &=\phi(\boldsymbol{x}) d x_{1} \wedge \cdots \wedge d x_{k} \end{aligned} $$得证.

Prove the homotopy formula $i_{X} d+d i_{X}=L_{x}$.

Proof:


第十章51节某习题知乎 可能和问的问题不一样,无视掉物理背景,我来计算 $\lim_{t\to\infty}\frac{1}{t}\mathrm{arg}\left(\sum_ia_ie^{i\omega_it}\right)$ 这是可以靠arnold这一章的空间平均=时间平均定理解决的,过程如下: 首先 $\frac{1}{t}\mathrm{arg}\left(\sum_ia_ie^{i\omega_it}\right)=\frac{1}{t}\arctan\left(\frac{\sum_ia_i\sin(\omega_it)}{\sum_ia_i\cos(\omega_it)}\right)$ 求个导数写成积分的样子 $\frac{1}{t}\arctan\left(\frac{\sum_ia_i\sin(\omega_it)}{\sum_ia_i\cos(\omega_it)}\right)=\frac{1}{t}\left(C+\int_{0}^t\mathrm{d}\tau\,\frac{\sum_{i,j}a_ia_j\omega_i\cos((\omega_i-\omega_j)\tau)}{\sum_{i,j}a_ia_j\cos((\omega_i-\omega_j)\tau)}\right)$ 常数C在极限过程中并没有什么用,下面不妨看做0。 如果我们直接对这个式子应用化时间平均的极限为空间平均,我们需要找一条轨道$\phi(t)=\phi(0)+\omega t$,一条显然的轨道就是$\phi(t)=0+\omega t$,这个时候我们被积函数的形式为$f(\omega)$,然后积分对n个$\omega$积分之后就变成了一个无关于$\omega$的一个数。显然,这样的转化对问题没什么帮助,所以,我们需要去寻找另一组有理无关的频率。为了可解性等一大堆理由,我们相信,这样的频率应当取做$\omega$的某种线性组合,且从$\omega$的有理无关,可以推出他们相互有理无关。 记$\omega_{ij}=\omega_{i}-\omega_j$,可以注意到恒等式: $\omega_{ij}-\omega_{ik}+\omega_{jk}=0$ 所以如果我们选$\Omega_{i}=\omega_{i}-\omega_1$这n-1个变量(当i=1的时候为零)作为我们的频率(显然是有理无关的),此时对于其他的i和j不等于1,我们有 $\omega_{ij}=\Omega_{i}-\Omega_{j}$ 那么上面积分中的被积式就是如下形式: $\sum_ia_i\omega_i f(\Omega_2 \tau,\cdots,\Omega_i \tau,\cdots,\Omega_n \tau)$ 直接化作对空间的积分就得到了 $\lim_{t\to\infty}\frac{1}{t}\mathrm{arg}\left(\sum_ia_ie^{i\omega_it}\right)=\sum_i\frac{a_i\omega_i}{(2\pi)^{n-1}} \int_{T^{n-1}}\mathrm{d}\Omega\,f(\Omega_2,\cdots,\Omega_i,\cdots,\Omega_n)$ 后面就是爆算积分一波带走,但是,不做计算也可以直接看到,极限是$\omega_i$的线性组合。 特殊函数没怎么学过,不知道有没有公式简化这个积分。 ------- 1. 三角的情况: $\sum_i\frac{a_i\omega_i}{(2\pi)^{2}}\int_0^{2\pi}\mathrm{d}\Omega_2\int_0^{2\pi}\mathrm{d}\Omega_3\,\frac{\sum_{j}a_j\cos(\Omega_i-\Omega_j)}{\sum_{i,j}a_ia_j\cos(\Omega_i-\Omega_j)}$ 即使是二重积分,计算量也特别感人。数值计算了一下,大概是对的(验证了直角三角形和等边三角形)。 2. 不知在哪个年份的夏日的下午的教室里,我写过4根杆的Lagrange函数,但年代久远,物理早已忘光了。 3. 有理无关是为了空间平均=时间平均。我觉得三角的情况可能和三角形稳定性没什么关系。