Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

The problem is:

Let $x(t)$ a solution of $$x'=-x(t^2-x^2),$$ so that $|x(t_0)|\lt |t_0|$. Show that for all $|t|\gt |t_0|$, $|x(t)|\lt |t|$.

Suppose that $t_0\geq 0$. Consider


Then $(t_0,x(t_0))$ is over the red line. I don't see why: $x$ solution of the above ODE implies that the graph of $x$ stay between the blue lines to the right of the red line and to the left of the orange line.


The attempt becomes an answer...

share|improve this question
The way the problem is currently stated, it is not true. I numerically solved the equation using $x(0)=.75$, and if we let $t_0=-1$, $|x(-1)|<1$. However, $|x(1)|>1$. –  robjohn Apr 25 '12 at 23:19
@robjohn Thanks. I just copy the problem as is in the notes of the course. By now, I'll just skip this exercise. –  leo Apr 25 '12 at 23:34
If you take the problem as stated and restrict to either $t\ge0$ or $t\le0$, then it is true. It is only the correlation between $x(t_0)$ and $x(-t_0)$ that is bad. Are you studying flow? –  robjohn Apr 25 '12 at 23:41
@robjohn, I'm studying ordinary differential equations. If you refer to this kind of flow then I'm not. In my attempt to solve the exercise, I put the restriction $t_0\geq 0$. Is that a good approach to the problem? what do you think? –  leo Apr 25 '12 at 23:50
that will give you a workable problem. However, you should also be able to solve the problem for $t_0<0$ as well. The problems are not exactly isomorphic. –  robjohn Apr 26 '12 at 2:09

4 Answers 4

up vote 4 down vote accepted

For simplicity, multiply the equation by $2x$ and subtract $2t$ to get $$ (x^2-t^2)'=2x^2(x^2-t^2)-2t\tag{1} $$ First, let's deal with $t>0$.

$(1)$ says that if $x^2-t^2<0$, then we also have $(x^2-t^2)'<0$. Therefore, if $x(t_0)^2-t_0^2<0$, we have $x(t)^2-t^2<0$ for all $t\ge t_0$.

Next, let's deal with $t<0$.

$(1)$ says that if $x^2-t^2\ge0$, then we also have $(x^2-t^2)'\ge0$. Therefore, if $x(t)^2-t^2\ge0$, we have $x(t_0)^2-t_0^2\ge0$ for all $t_0\ge t$. The contrapositive says that for all $t\le t_0$, if $x(t_0)^2-t_0^2<0$, then we have $x(t)^2-t^2<0$.

A Graphical Explanation

Consider the direction field (aka slope field and flow diagram) for the solutions. That is, the vector field $\color{#0000c0}{\frac{\mathrm{d}}{\mathrm{d}t}(t,x)}$ which is tangent to all solutions $\color{#c00000}{(t,x)}$.


The important part to consider is the flow across the boundary $x^2=t^2$ between the regions where $\color{#00c000}{x^2< t^2}$ and $\color{#c000c0}{x^2>t^2}$. Notice that the on the boundary, the flow is horizontal since $x'=x(x^2-t^2)=0$.

$\hspace{4.2cm}$flow diagram

Thus, when $t<0$ (on the left), a solution can only move from the region where $\color{#00c000}{x^2< t^2}$ to the region where $\color{#c000c0}{x^2>t^2}$. When $t>0$ (on the right), a solution can only move from the region where $\color{#c000c0}{x^2>t^2}$ to the region where $\color{#00c000}{x^2< t^2}$.

share|improve this answer
Neat! :)${}{}{}{}{}$ –  t.b. Apr 27 '12 at 2:19

Consider first $t_0\geq 0$.

Suppose that the set $$C:=\{t\gt t_0:x(t)\geq t\}$$ is not empty, thus, since $C$ is bounded below by $t_0$, $$\tau=\inf C$$ is well defined. We can proof (by taking a sequence $C\ni t_n\to\tau$ and assuming continuity of $x$) that $\tau\in C$. In particular $\tau\gt t_0$ and then $$x(\tau)\geq\tau\gt 0,$$ then, there exist an $\alpha\gt 0$ such that $x$ is positive over $]\tau-\alpha,\tau+\alpha[\subset]t_0,\infty[$.

Then for all $t\in]\tau-\alpha,\tau[$, by the definition of $\tau$, $t\notin C$ (otherwise $t\in C$ with $t\lt\inf C$), so $0\lt x(t)\lt t$ and then $$x'(t)=x(t)(x(t)^2-t^2)\lt 0,$$ so $x$ is decreasing over $]\tau-\alpha,\tau[$.

Pick $t\in ]\tau-\alpha,\tau[$, then $$x(t)\geq x(\tau)\geq\tau\gt t,$$ thus $t\in C$, and that's contradictory.

Therefore $C=\emptyset$ and then for all $t\gt t_0$, $x(t)\lt t$.

share|improve this answer

Hint: If $x(t_0) < |t_0|$ then, by the differential equation, $x'(t_0) = -x(t_0)(\textrm{something positive})$.

share|improve this answer

If $x(t_0)=0$ then $x(t)=0$ for all $t\ge t_0$. Without loss of geneality, we assume $x(t_0)\ne0$.

If $0<x(t_0)<t_0$, you want to show that $0<x(t)\le t$ for all $t\ge t_0$. Now, $x'(t_0)=-x(t_0)(t_0^2-x(t_0)^2)<0$. By uniqueness of solutions $x(t)>0$ for all $t\>t_0$. And since the right hand side of the equation is negative on the region $\{(t,x):0<x<t,t>t_0\}$, it is easy to see that $x$ is decreasing.

If $x(t_0)=t_0$ then $x'(t_0)=0$. Taking derivatives in both sides of the equation we get $x''(t_0)=-2\,t_0\,x(t_0)<0$. Thus $x'(t)<0$ on some interval $(t_0,t_0+\delta)$, $\delta>0$, $x(t)<t$ on that same interval, and the argument in the previos paragraph carries through.

I leave to you the case $-t_0\le x(t_0)<0$.

share|improve this answer
I don't see why $x(t_0)=0$ implies $x(t)=0$. –  leo Apr 25 '12 at 18:52
@leo: note that $$ \frac{\mathrm{d}}{\mathrm{d}t}\log|x|=x^2-t^2 $$ When $|x|$ is close to $0$, $\log|x|$ is close to $-\infty$, yet the slope of $\log|x|$ is finite. Even if the slope were positive, which it isn't, it would take a long time to climb to something not close to $-\infty$. The closer $x$ gets to $0$, the longer it would take. Unfortunately, the slope of $\log|x|$ is not positive. –  robjohn Apr 25 '12 at 22:48
$x(t)=0$ is certainly a solution. Uniqueness of solution (the right hand side is $C^\infty$, so it is locally Lipschitz,) implies that if the solution vanishes at some point, it is identically equal to $0$. –  Julián Aguirre Apr 26 '12 at 8:22
I see. Thanks. ${}$ –  leo Apr 27 '12 at 3:02

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.