京都大学 情報学研究科 数理工学専攻 2020年8月実施 力学系数学
Author
Casablanca, 祭音Myyura
Description
日本語版
n > 1 n > 1 n > 1 を整数,f : R n → R n f : \mathbb{R}^n \rightarrow \mathbb{R}^n f : R n → R n を C 1 C^1 C 1 級関数として,R \mathbb{R} R 上の微分方程式系
d x d t = f ( x ) , x ∈ R n \begin{align}
\frac{dx}{dt} = f(x), \quad x \in \mathbb{R}^n \tag{1}
\end{align} d t d x = f ( x ) , x ∈ R n ( 1 )
を考える。x = ϕ ( t ) x = \phi(t) x = ϕ ( t ) を R \mathbb{R} R 上で有界な式 (1) の非定数解とする。
次式を式 (1) の解 x = ϕ ( t ) x = \phi(t) x = ϕ ( t ) のまわりの変分方程式という:
d y d t = D f ( ϕ ( t ) ) y , y ∈ R n (2) \frac{dy}{dt} = Df(\phi(t))y, \quad y \in \mathbb{R}^n \tag{2} d t d y = D f ( ϕ ( t )) y , y ∈ R n ( 2 )
ここで,D f ( x ) Df(x) D f ( x ) は f ( x ) f(x) f ( x ) のヤコビ行列で,各 j = 1 , 2 , … , n j = 1, 2, \ldots, n j = 1 , 2 , … , n に対して,f j ( x ) f_j(x) f j ( x ) と x j x_j x j をそれぞれ,f ( x ) f(x) f ( x ) と x x x の第 j j j 成分として,
D f ( x ) = ( ∂ f 1 ∂ x 1 ( x ) … ∂ f 1 ∂ x n ( x ) ⋮ ⋱ ⋮ ∂ f n ∂ x 1 ( x ) … ∂ f n ∂ x n ( x ) ) Df(x) = \begin{pmatrix}
\frac{\partial f_1}{\partial x_1}(x) & \dots & \frac{\partial f_1}{\partial x_n}(x) \\
\vdots & \ddots & \vdots \\
\frac{\partial f_n}{\partial x_1}(x) & \dots & \frac{\partial f_n}{\partial x_n}(x)
\end{pmatrix} D f ( x ) = ∂ x 1 ∂ f 1 ( x ) ⋮ ∂ x 1 ∂ f n ( x ) … ⋱ … ∂ x n ∂ f 1 ( x ) ⋮ ∂ x n ∂ f n ( x )
で与えられる n n n 次正方行列である。以下の問いに答えよ。
(i) 極限 a + = lim t → + ∞ ϕ ( t ) a_+ = \lim_{t \to +\infty} \phi(t) a + = lim t → + ∞ ϕ ( t ) と a − = lim t → − ∞ ϕ ( t ) a_- = \lim_{t \to -\infty} \phi(t) a − = lim t → − ∞ ϕ ( t ) が存在するとき,x = a + x = a_+ x = a + と a − a_- a − が式 (1) の定数解であることを示せ。また,変分方程式 (2) が lim t → ± ∞ ψ ( t ) = 0 \lim_{t \to \pm \infty} \psi(t) = 0 lim t → ± ∞ ψ ( t ) = 0 かつ R \mathbb{R} R 上で有界な解 y = ψ ( t ) y = \psi(t) y = ψ ( t ) をもつことを示せ。
(ii) 次式を満たす C 1 C^1 C 1 級関数 u : R n → R n u : \mathbb{R}^n \rightarrow \mathbb{R}^n u : R n → R n が存在するものとする。
D u ( x ) f ( x ) − D f ( x ) u ( x ) = 0 Du(x)f(x) - Df(x)u(x) = 0 D u ( x ) f ( x ) − D f ( x ) u ( x ) = 0
2 個のベクトル f ( ϕ ( 0 ) ) f(\phi(0)) f ( ϕ ( 0 )) と u ( ϕ ( 0 ) ) u(\phi(0)) u ( ϕ ( 0 )) が線形独立であるとき,変分方程式 (2) の線形独立な解を2個求めよ。
(iii) 次式を満たす n − 1 n - 1 n − 1 個の C 1 C^1 C 1 級関数 v j : R n → R n ( j = 1 , 2 , … , n − 1 ) v_j : \mathbb{R}^n \rightarrow \mathbb{R}^n \ (j = 1, 2, \ldots, n - 1) v j : R n → R n ( j = 1 , 2 , … , n − 1 ) が存在するものとする。
D v j ( x ) f ( x ) − D f ( x ) v j ( x ) = 0 ( j = 1 , 2 , … , n − 1 ) Dv_j(x)f(x) - Df(x)v_j(x) = 0 \quad (j = 1, 2, \ldots, n - 1) D v j ( x ) f ( x ) − D f ( x ) v j ( x ) = 0 ( j = 1 , 2 , … , n − 1 )
n n n 個のベクトル f ( ϕ ( 0 ) ) f(\phi(0)) f ( ϕ ( 0 )) と v j ( ϕ ( 0 ) ) ( j = 1 , 2 , … , n − 1 ) v_j(\phi(0)) \ (j = 1, 2, \ldots, n - 1) v j ( ϕ ( 0 )) ( j = 1 , 2 , … , n − 1 ) が線形独立であるとき,変分方程式 (2) の一般解を求めよ。
English Version
题目描述
设整数 n > 1 n>1 n > 1 ,f : R n → R n f:\mathbb R^n\to\mathbb R^n f : R n → R n 为 C 1 C^1 C 1 函数,考虑自治系统
d x d t = f ( x ) , x ∈ R n . (1) \frac{d\boldsymbol x}{dt}=f(\boldsymbol x),
\qquad \boldsymbol x\in\mathbb R^n. \tag{1} d t d x = f ( x ) , x ∈ R n . ( 1 )
令 x = ϕ ( t ) \boldsymbol x=\phi(t) x = ϕ ( t ) 为定义在 R \mathbb R R 上、有界且非常值的解。其沿轨道的变分方程为
d y d t = D f ( ϕ ( t ) ) y , y ∈ R n , (2) \frac{d\boldsymbol y}{dt}=Df(\phi(t))\boldsymbol y,
\qquad \boldsymbol y\in\mathbb R^n, \tag{2} d t d y = D f ( ϕ ( t )) y , y ∈ R n , ( 2 )
其中 D f Df D f 是 f f f 的 Jacobian 矩阵。回答:
若极限
a + = lim t → + ∞ ϕ ( t ) a_+=\lim_{t\to+\infty}\phi(t) a + = lim t → + ∞ ϕ ( t ) 与
a − = lim t → − ∞ ϕ ( t ) a_-=\lim_{t\to-\infty}\phi(t) a − = lim t → − ∞ ϕ ( t ) 存在,证明 x = a + \boldsymbol x=a_+ x = a + 和 x = a − \boldsymbol x=a_- x = a − 都是 (1) 的常值解;并证明 (2) 存在一个在 R \mathbb R R 上有界且满足
lim t → ± ∞ ψ ( t ) = 0 \lim_{t\to\pm\infty}\psi(t)=0 lim t → ± ∞ ψ ( t ) = 0 的解
y = ψ ( t ) \boldsymbol y=\psi(t) y = ψ ( t ) 。
假设存在 C 1 C^1 C 1 函数 u : R n → R n u:\mathbb R^n\to\mathbb R^n u : R n → R n 满足
D u ( x ) f ( x ) − D f ( x ) u ( x ) = 0. Du(\boldsymbol x)f(\boldsymbol x)
-Df(\boldsymbol x)u(\boldsymbol x)=0. D u ( x ) f ( x ) − D f ( x ) u ( x ) = 0.
若 f ( ϕ ( 0 ) ) f(\phi(0)) f ( ϕ ( 0 )) 与 u ( ϕ ( 0 ) ) u(\phi(0)) u ( ϕ ( 0 )) 线性无关,求变分方程 (2) 的两个线性无关解。
假设存在 n − 1 n-1 n − 1 个 C 1 C^1 C 1 函数
v j : R n → R n v_j:\mathbb R^n\to\mathbb R^n v j : R n → R n 满足
D v j ( x ) f ( x ) − D f ( x ) v j ( x ) = 0 ( j = 1 , … , n − 1 ) , Dv_j(\boldsymbol x)f(\boldsymbol x)
-Df(\boldsymbol x)v_j(\boldsymbol x)=0
\quad(j=1,\ldots,n-1), D v j ( x ) f ( x ) − D f ( x ) v j ( x ) = 0 ( j = 1 , … , n − 1 ) ,
且 f ( ϕ ( 0 ) ) , v 1 ( ϕ ( 0 ) ) , … , v n − 1 ( ϕ ( 0 ) ) f(\phi(0)),v_1(\phi(0)),\ldots,v_{n-1}(\phi(0)) f ( ϕ ( 0 )) , v 1 ( ϕ ( 0 )) , … , v n − 1 ( ϕ ( 0 )) 线性无关。求 (2) 的一般解。
Kai
(i)
Since ϕ ( t ) \phi(t) ϕ ( t ) is a solution, we have
d ϕ ( t ) d t = f ( ϕ ( t ) ) . \frac{d\phi(t)}{dt}=f(\phi(t)). d t d ϕ ( t ) = f ( ϕ ( t )) .
From lim t → + ∞ ϕ ( t ) = a + \lim_{t\to+\infty}\phi(t)=a_+ lim t → + ∞ ϕ ( t ) = a + and continuity, we have
lim t → + ∞ d ϕ ( t ) d t = f ( a + ) . \lim_{t\to+\infty}\frac{d\phi(t)}{dt}=f(a_+). t → + ∞ lim d t d ϕ ( t ) = f ( a + ) .
If L = f ( a + ) ≠ 0 L=f(a_+)\ne0 L = f ( a + ) = 0 , put q = L / ∥ L ∥ q=L/\lVert L\rVert q = L / ∥ L ∥ . For all sufficiently large t t t ,
q ⊤ ϕ ′ ( t ) > ∥ L ∥ 2 . q^\top\phi'(t)>\frac{\lVert L\rVert}{2}. q ⊤ ϕ ′ ( t ) > 2 ∥ L ∥ .
Integrating this inequality contradicts the convergence of q ⊤ ϕ ( t ) q^\top\phi(t) q ⊤ ϕ ( t ) . Hence f ( a + ) = 0 f(a_+)=0 f ( a + ) = 0 , so x = a + x=a_+ x = a + is a constant solution. The same argument at − ∞ -\infty − ∞ gives f ( a − ) = 0 f(a_-)=0 f ( a − ) = 0 , so x = a − x=a_- x = a − is also a constant solution.
Notice that
d d t f ( ϕ ( t ) ) = D f ( ϕ ( t ) ) d ϕ ( t ) d t . \frac{d}{dt}f(\phi(t))
=Df(\phi(t))\frac{d\phi(t)}{dt}. d t d f ( ϕ ( t )) = D f ( ϕ ( t )) d t d ϕ ( t ) .
Set
ψ ( t ) = ϕ ′ ( t ) = f ( ϕ ( t ) ) . \psi(t)=\phi'(t)=f(\phi(t)). ψ ( t ) = ϕ ′ ( t ) = f ( ϕ ( t )) .
Then
ψ ′ ( t ) = D f ( ϕ ( t ) ) ψ ( t ) , lim t → ± ∞ ψ ( t ) = f ( a ± ) = 0. \psi'(t)=Df(\phi(t))\psi(t),\qquad
\lim_{t\to\pm\infty}\psi(t)=f(a_\pm)=0. ψ ′ ( t ) = D f ( ϕ ( t )) ψ ( t ) , t → ± ∞ lim ψ ( t ) = f ( a ± ) = 0.
Since ϕ ( R ) \phi(\mathbb R) ϕ ( R ) is bounded, its closure is compact; continuity of f f f therefore implies that ψ ( t ) = f ( ϕ ( t ) ) \psi(t)=f(\phi(t)) ψ ( t ) = f ( ϕ ( t )) is bounded on R \mathbb R R . Since ϕ \phi ϕ is nonconstant, ψ \psi ψ is not identically zero.
(ii)
The two solutions are
y 1 ( t ) = f ( ϕ ( t ) ) , y 2 ( t ) = u ( ϕ ( t ) ) . y_1(t)=f(\phi(t)),\qquad y_2(t)=u(\phi(t)). y 1 ( t ) = f ( ϕ ( t )) , y 2 ( t ) = u ( ϕ ( t )) .
Indeed,
y 1 ′ = D f ( ϕ ) y 1 , y 2 ′ = D u ( ϕ ) f ( ϕ ) = D f ( ϕ ) u ( ϕ ) = D f ( ϕ ) y 2 . y_1'=Df(\phi)y_1,\qquad
y_2'=Du(\phi)f(\phi)=Df(\phi)u(\phi)=Df(\phi)y_2. y 1 ′ = D f ( ϕ ) y 1 , y 2 ′ = D u ( ϕ ) f ( ϕ ) = D f ( ϕ ) u ( ϕ ) = D f ( ϕ ) y 2 .
They are linearly independent because their values at t = 0 t=0 t = 0 are linearly independent.
(iii)
By the same calculation, f ( ϕ ( t ) ) f(\phi(t)) f ( ϕ ( t )) and v j ( ϕ ( t ) ) v_j(\phi(t)) v j ( ϕ ( t )) ( j = 1 , … , n − 1 ) (j=1,\ldots,n-1) ( j = 1 , … , n − 1 ) are solutions. Their values at t = 0 t=0 t = 0 form a basis of R n \mathbb R^n R n , hence the general solution is
y ( t ) = c 0 f ( ϕ ( t ) ) + ∑ j = 1 n − 1 c j v j ( ϕ ( t ) ) , c 0 , c 1 , … , c n − 1 ∈ R . y(t)=c_0f(\phi(t))+\sum_{j=1}^{n-1}c_jv_j(\phi(t)),
\qquad c_0,c_1,\ldots,c_{n-1}\in\mathbb R. y ( t ) = c 0 f ( ϕ ( t )) + j = 1 ∑ n − 1 c j v j ( ϕ ( t )) , c 0 , c 1 , … , c n − 1 ∈ R .