1
$\begingroup$

Firstly, take the family of differential equations $\dot x = > \frac{dx}{dt}=x^\alpha$, for any $\alpha \in \mathbb R$

The solution to these equations is $$(\text{for } \alpha=1):x(t)=x_0e^t$$ $$(\text{for } \alpha\neq 1): x(t)=\left(x_0^{1-\alpha}+(1-\alpha)t\right)^{\frac{1}{1-\alpha}}$$

In the second equation we can rewrite $\alpha$ as the deviation from $1$: $\alpha = 1+\delta$:

$$(\text{for } \alpha\neq 1): x(t)=\left(x_0^{-\delta}+-\delta > t\right)^{\frac{1}{-\delta}}$$

Note first that for any positive $\delta$ (no matter how miniscule), $x(t)$ approaches infinity at a finite (!!) time $t$. I will call such behaviour "infinitely explosive".

To see this, note that $\frac{1}{x(t)}$ approaches $0$ at finite time $t$ as follows: $\frac{1}{x(t)}=\left(x_0^{-\delta}+-\delta t\right)^{\frac{1}{\delta}}=0 \implies t= \frac {1}{x_0^{\delta}\delta }$, which is a finite number for any $\delta \in \mathbb R$.

On the other hand, for negative $\delta$, there is no $t$ such that $1/x(t)=0$ (unless, $x_0=0$). In fact, for $\alpha = 0$, it grows linearly, for $\alpha = 0.5$, quadratically, etc.

So this gives a very weird result: if we take the differential equation, and slowly "slide" $\alpha$ towards $1$, $x(t)$ slowly converges to the exponential function. This is intuitive once we realize that $e^x=\sum_{n=1}^\infty(\frac{x^n}{n!})$. However, when we then slide $\alpha$ only slightly above $1$, x(t) turns into an "infinitely explosive function" (reaches infinity in finite time).


Now Compare this with the following family of equations:

$$\dot x = \text{hypexp}(x,n), \text{for any } n\in \mathbb N$$ Whose solutions are $$x(t)= {e^{e^{...^{x}}}}$$ with $n+1$ amount of $e$'s

Where

$\text{hypexp}(x,0)= x$

$\text{hypexp}(x,1)= x\cdot ln(x)$

$\text{hypexp}(x,2) = x \cdot ln(x)\cdot ln(ln(x))$

$\text{hypexp}(x,n) = x \cdot ln(x)\cdot ln(ln(x))\cdot ln(ln(ln(x)))\cdot ...$


Now given $\dot x = f(x)$, define a function $f$ to be "explosively larger" than $g$ if $f$ approaches reaches infinity at smaller $t$ than $g$ for the same $x_0$. ( I don't know if this concept already has another name)

Now here comes the weird conclusion: no matter how large $n$ and how miniscule $\delta$, $f(x)=x\cdot x^{\delta}$ is always explosively larger than $g(x)=x\cdot ln(x) \cdot ln(ln(x))\cdot ...$, whereas if $\delta$ reaches $0$, $g$ is immediately monumentally larger (in the normal sense) than $f$ for any $t>1$.


Discussion

  1. Does anyone else find this as weird and interesting as I do?
  2. Is there some kind of "infintely explosive supremum"? A kind of "smallest" infinitely explosive (or "largest" non-infinitely explosive) function such that all non-infinitely explosive functions have to be smaller than it. Note that it would have to be larger than even bizarrely quickly increasing functions, such as Ackermann's function.
  3. Is there theory that makes sense of all this?
  • 1
    The phrase you want to look up is "finite time blow-up," as in "the solutions to this differential equation blow up in finite time."2017-02-25

0 Answers 0