Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Consider the 2nd order ODE $$ \ddot{x}+\frac32x^2=0. $$ Denote by $u$ the maximal solution of the associated Cauchy problem with initial condition $(x(0),\dot{x}(0))=(0,1)$.

The problem is to prove the existence of some $T>0$ such that $\dot{u}(T)=0$.

Any help will be appreciated.

Remark: So far I know that if $I$ denotes the domain of existence of $u$, then $$ \dot{u}^2(t)+u^3(t)=1\quad \forall t \in I. $$ In particular, if such a $T$ exists then $u(T)=1$.

share|cite|improve this question
What kind of proof do you need? In hamiltonian systems of 1 degree of freedom it is very easy to plot the trajectories, it's just the contour plot of the hamiltonian function. – tst Feb 3 '13 at 5:47
An other method is to write the system as $$\dot{x}=y$$ $$\dot{y}=-\frac{3}{2}x^2$$ and look at the vector field for the point $(0,1)$ onwards. You should be able to use continuity and get your result. – tst Feb 3 '13 at 5:52
@tst: the vector field shows the path that $(x,\dot{x})$ would take to get from $(0,1)$ to $(1,0)$, but does it show that this path would take finite time? – robjohn Feb 3 '13 at 15:26
up vote 1 down vote accepted

Let $(T_u^-,T_u^+) \ni 0$ be the maximal domain of existence for the solution $u$, and suppose $\dot{u}(t)\ne 0$ for every $t \in [0,T_u^+)$. Since $\dot{u}(0)>0$ and $\dot{u}$ is continuous, it follows that $\dot{u}>0$ on $[0,T_u^+)$.

Using the Hamiltonian nature of the original DE we have $$\tag{1} \dot{u}^2(t)+u^3(t)=1 \quad \forall t \in (T_u^-,T_u^+). $$ It follows that the restriction $u_0$ of $u$ to $[0,T_u^+)$ solves the problem $$\tag{2} \dot{x}=\sqrt{1-x^3},\ x(0)=0. $$ If $u_1:(t_1,t_2) \to \mathbb{R}$ denotes the maximal solution for (2), then its restriction to $[0,T_u^+)$ is precisely $u_0$. Hence $$ T_u^+\le t_2=\int_0^1\frac{dx}{\sqrt{1-x^3}} \in (0,\infty). $$ If $T_u^+<t_2$, then $$ \tilde{u}:(T_u^-,t_2) \to \mathbb{R},\ \tilde{u}(t)=\begin{cases} u(t)&\text{ for }\ t \in (T_u^-,0]\\ u_1(t)&\text{ for }\ t \in (0,t_2) \end{cases} $$ is a solution of the original problem which is defined on a larger interval than $(T_u^-,T_u^+)$, contradicting the maximality of $(T_u^-,T_u^+)$. Hence we necessarily have $t_2=T_u^+$, and on $[0,T_u^+)$ the solution is given by $u(t)=G^{-1}(t)$, where $$ G(x)=\int_0^x\frac{ds}{\sqrt{1-s^3}} $$ NB. $G$ is continuous on $(-\infty,1]$ and differentiable on $(-\infty,1)$.

Using the definition of $u$ on $[0,T_u^+)$, we have $$ \lim_{t\uparrow T_u^+}u(t)=1, $$ and thanks to (1) we get $$ \lim_{t\uparrow T_u^+}\dot{u}(t)=0. $$ Let us now consider the IVP $$\tag{3} \ddot{x}+\frac32x^2=0,\ x(T_u^+)=1,\ \dot{x}(T_u^+)=0, $$ or equivalenty $$\tag{4} (\dot{x},\dot{y})=f(x,y),\ (x(T_u^+),y(T_u^+))=(1,0), $$ where $$ f: \mathbb{R}^2 \to \mathbb{R}^2,\ f(x,y)=(y,-\frac32x^2) $$ is $C^1$ ($C^\infty$ actually), and therefore locally Lipschitz. Thus we can solve (4) (at least locally), i.e. we have a solution $$ u_2:(T_u^+-\delta, T_u^++\delta) \to \mathbb{R} $$ for (4), where $\delta>0$. Gluing $u$ with $u_2$ at $t=T_u^+$ we obtain a solution of the original problem which is defined on $(T_u^-,T_u^++\delta)$. Again this contradicts the maximality of $(T_u^-,T_u^+)$. Thus, there must exists some $T>0$ such that $\dot{u}(T)=0$.

share|cite|improve this answer

As $\dot x(0)=1$ there is a neighborhood $U$ of $t=0$ such that $\dot x(t)\ne0$ for $t\in U$. Within $U$ the given DE is equivalent with $$0=2\dot x\ddot x + 3x^2\dot x=(\dot x^2+ x^3)^\cdot \ ,$$ therefore $\dot x^2+x^3$ is constant in $U$, and as $x(0)=0$, $\ \dot x(0)=1$ this constant has to be $1$. It follows that the solution of the given initial problem is the solution of the first order initial value problem $$\dot x=\sqrt{1-x^3}\ ,\qquad x(0)=0\ ,\tag{1}$$ at least as long as $\dot x(t)>0$.

From looking at the DE it is obvious that the solution $t\mapsto x(t)$ of $(1)$ will in a primary phase $[0,T[\ $ increase from $0$ to $1$. During this phase we have $\dot x(t)>0$. If $T=\infty$ we are done; otherwise we have to investigate what happens as $t\to T\!-$. In order to determine $T$ we separate the variables in $(1)$, i.e., we write the DE in the form $$dt={dx\over\sqrt{1-x^3}}\ .$$ From this we derive $$T=\lim_{\xi\to 1-}\int_0^\xi{dx\over\sqrt{1-x^3}}<\int_0^1{dx\over\sqrt{1-x^2}}={\pi\over2}<\infty\ .$$ (In his answer Did has given the exact value of $T$.)

So we are left with the question what happens at time $T$. Note that $\lim_{t\to T-} x(t)=1$ and $\lim_{t\to T-}\dot x(t)=0$; furthermore it is easy to verify that $\lim_{t\to T-}\ddot x(t)=-{3\over2}$. Our solution is a solution of the original DE $\ddot x +{3\over2}x^2=0$; it therefore should be possible (see below) to identify it with the solution to the initial value problem $$\left\{\eqalign{\dot x&=y\cr \dot y&=-{3\over2} x^2\cr}\right.\quad ,\quad x(T)=1, \quad y(T)=0\ .\tag{2}$$ The solution of $(2)$ is by general principles an analytical function of $\tau:=t-T$ in a neighborhood of $\tau=0$. Comparing coefficients one gets the expansion $$x(T+\tau)=1-{3\over4}\tau^2 +{3\over16}\tau^4 -{3\over64}\tau^6 +{9\over896}\tau^8 + \ ? \tau^{10}\ .$$ If the indicated identification can be done this would mean that our solution $x(\cdot)$ can be continued past $t=T$ (albeit with another constant when we reduce to a first order problem).

As a final note I'd like to remark that even for the linear DE $\ddot x+x=0$ we run into the same difficulties if we don't know about trigonometric functions.

share|cite|improve this answer
...Except that the function $x$ is defined as a solution of the 2nd order DE in the post, not as a solution of (1). Hence the continuation you suggest is illegal because $x''$ would be discontinuous at $T$. It seems that in fact, $x(T+h)=1-\frac34h^2+o(h^2)$. – Did Feb 3 '13 at 11:35

Multiply by $2\dot{x}$ $$ 2\dot{x}\ddot{x}+3x^2\dot{x}=0\tag{1} $$ and integrate (using the initial conditions) to get $$ \dot{x}^2+x^3=1\tag{2} $$ Since $\ddot{x}\le0$, $\dot{x}$ starts at $1$ and decreases. While $\dot{x}>0$, $x$ increases from $0$. In fact, $\dot{x}=\sqrt{1-x^3}$. Should $x$ increase to $1$ at finite time $T$, we then have $\dot{x}=0$ and $\ddot{x}=-\frac32$. This precludes the solution to $(2)$ where $x=1$ for $t\ge T$. Thus, for $t\gt T$, $\dot{x}<0$ and $x$ decreases from $1$. In fact, $\dot{x}=-\sqrt{1-x^3}$ and we get that $x(2T)=0$.

$x$ does indeed increase to $1$ in finite time. In fact, the time from $x=0$ to $x=1$ is $$ \begin{align} T &=\int_0^1\frac{\mathrm{d}x}{\sqrt{1-x^3}}\\ &=\frac13\int_0^1u^{-2/3}(1-u)^{-1/2}\,\mathrm{d}u\\ &=\frac13\mathrm{B}\left(\frac13,\frac12\right)\\ &=\frac13\frac{\Gamma\left(\frac13\right)\Gamma\left(\frac12\right)}{\Gamma\left(\frac56\right)}\\ &\stackrel.=1.4021821053254542612\tag{3} \end{align} $$ As mentioned above, $\dot{x}(T)=0$ and $x(2T)=0$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.