Homework for Part 4
Week 7 : Solve exercises 3, 4, 5.
Week 8 : Solve exercises 1, 2.
The exercises 3, 4, 5 are independent of the class. You just need to know that \(\displaystyle\dot{x} = x' = \frac{dx}{dt}.\)
Let \(f\) be a continuous function satisfying \[\vert f(t,x)-f(t,y) \vert \leq K \vert x-y \vert \quad \forall t, x, y \in \mathbb{R}.\]
To approximate the solution to \[\begin{cases} y'(t) = f(t,y(t)), \quad t \in [t_0, t_0 +T], \\ y(t_0) = y_0, \end{cases}\] one considers the subdivision of \([t_0,t_0+T]\) into \[t_0 < t_1 < t_2 < \ldots < t_N = t_0+T\] and the schema \[\begin{cases} y_0 = y_0 \\ y_{n+1} = y_n + h_n f(t_n,y_n), \quad h_n = t_{n+1} -t_n, \quad n=0, \ldots N-1. \end{cases}\]
i.e. the Euler schema where \(y_n\) is the approximation of \(y(t_n).\)
One sets \[\begin{aligned} \epsilon_n &= y(t_{n+}) - y(t_n) - h_n f(t_n,y(t_n)), &\quad n=0, \ldots, N-1, \\ e_n &= \vert y(t_n) - y_n \vert, &\quad n=0, \ldots, N. \end{aligned}\]
Show that \(e_{n+1} \leq (1+K h_n) e_n + \vert \epsilon_n \vert\) for \(n=0, \ldots, N-1.\)
Show that \(1 +x \leq e^x\) for all \(x \geq 0.\)
Show that \[e_n \leq \sum_{i=0}^{n-1} e^{(t_n - t_{i+1})K} \vert \epsilon_i \vert \quad \forall 0 < n \leq N.\]
Hint: Argue by induction.
One sets \(\varepsilon(h) = \max_{\vert \xi - \zeta \vert < h} \vert y'(\xi) - y'(\zeta) \vert,\) where \(h = \max_{n=0, \ldots, N-1} h_n.\)
Show that \(e_n \leq T e^{KT} \varepsilon(h).\)
Conclude that \(\max_n \vert y(t_n) - y_n\vert \to 0\) when \(h \to 0.\)
Write a code to compute the solution of the Price-Demand system
by the Euler Method,
by the Runge-Kutta Method.
One denote by \(\{e_1, e_2, e_3 \}\) the canonical basis of \(\mathbb{R}^3.\) For two vectors \(x, y\) of \(\mathbb{R}^3\) the exterior product \(x \wedge y\) is defined as \[x \wedge y = \begin{vmatrix} x_1 & y_1 & e_1 \\ x_2 & y_2 & e_2 \\ x_3 & y_3 & e_3 \end{vmatrix} = \begin{pmatrix} x_2 y_3 -y_2 x_3 \\ x_3 y_1 - y_3 x_1 \\ x_1 y_2 - y_1 x_2 \end{pmatrix}.\]
If \(x,y\) are differentiable show that \[\frac{d}{dt} x(t) \wedge y(t) = \dot{x} \wedge y + x \wedge \dot{y}.\] What generalisation of this formula can be provided ?
Hint: \(x \wedge y\) is bi-linear.
Let \(u=(u_1, u_2, 0 )^t\) and \(v=(v_1, v_2, 0)^t\) two vectors of \(\mathbb{R}^3.\)
Show that the area of the triangle defined by the origin \(O\) and the two vectors \(u\) and \(v\) (see Figure 1) is given by \[A = \frac{1}{2}\vert u_1 v_2 - v_1 u_2 \vert.\]
Hint: one can assume that the coordinates are chosen such that \(u\) lies on the \(x\)-axis of \(\mathbb{R}^2.\)
Consider \(x(t)\) a planar vector of \(\mathbb{R}^3\) depending of \(t.\)
From the previous point, show that the area swept between \(t\) and \(t+h\) by \(x(t)\) is given by \[A(t+h) - A(t) = \frac{1}{2}\left\vert x_1(t) \left(x_2(t+h) - x_2(t)\right) - \left( x_1(t+h) - x_1(t)\right)x_2(t)\right\vert.\] Derive that one has \[\frac{dA}{dt}(t) = \frac{1}{2} \left\vert x_1(t) x_2'(t)-x_1'(t)x_2(t)\right\vert.\]
Show that the area swept by \(x(t)\) between \(t_0\) and \(t\) is \[\displaystyle A(t) = \frac{1}{2} \int_{t_0}^t r^2(s)\theta'(s) ds\] where \(r\) and \(\theta\) are the polar coordinate of \(x(t)\) and \(s\) is the arc-length parameter.
For \(a \geq b > 0\) we set \(c^2 = a^2 -b^2.\)
In the plane \((x,y)\) one considers the curve defined by \(\displaystyle r(\theta) = \frac{b^2}{a-c \cos \theta},\) see Figure 5.
Show that \(\displaystyle\frac{x^2}{a^2}+ \frac{y^2}{b^2} = 1,\) i.e. the point \(P\) is on an ellipse.
Conversely, one considers the ellipse \(\displaystyle\frac{x^2}{a^2}+ \frac{y^2}{b^2} = 1.\) One sets \[x = r \cos \theta - c \quad , \quad y = r\sin \theta.\]
Show that if \((x,y)\) is located on the ellipse one has \(\displaystyle r = r(\theta) = \frac{b^2}{a-c \cos \theta}.\)
Show that the curve \[r(\theta) = \frac{h^2}{m} \frac{1}{1 - \varepsilon \cos \theta}, \qquad \text{with} \;\varepsilon<1\] is an ellipse.