2
$\begingroup$

What is known about fine stability properties of ODEs of the following kind? $$ \dot{x} = Ax + b + \phi(x),\quad x\in \mathbb{R}^d ,$$ where $d\geq 1$; $A$ is a constant matrix with all e.v. having real part $<0$; $b$ is a constant vector and $\phi:\mathbb{R}^d\to \mathbb{R}^d_{\geq 0}$ is a function, growing faster than linear on $+\infty$. A particular $\phi$ to think about is $\phi(x) = c\exp(x)$, for a fixed $c>0$ and $\exp(x)$ means coordinate-wise application of the exponential.

For example, for $d=1$ it is easy to see that there is a stable fixed point $x_s$, which is followed by an unstable one $x_u>x_s$, after which each solution goes to infinity. I would like to find out whether there are some qualitative differences in higher dimensions. A dream result would be a detailed description of the attraction basin for the stable fixed point.

UPD: It is not true even in 1D that there is always a stable point. But one can always derive conditions for its existence by looking at how many solutions does equation $Ax+b+\phi(x) = 0$ have and calculating derivatives at the roots.

Are there some classical (or not so) references covering this kind of equations?

$\endgroup$
5
  • $\begingroup$ Let $d=1$. $\dot{x}=-x-1+e^x$ has only one equilibrium, asymptotically stable from the left and unstable from the right. $\dot{x}=-x+e^x$ has no equilibria. $\endgroup$
    – user539887
    Commented Apr 23, 2018 at 6:42
  • $\begingroup$ Let again $d=1$. Even if we assume superlinear growth plus "the right direction" we need not obtain the result you mention. A trivial example is $\dot{x}=-x+(x^3+x)$: $\phi(x)=x^3+x$ has superlinear growth, is positive close to $+\infty$, negative close to $-\infty$, but $0$ is the only unstable equilibrium. You should make additional assumptions on $\phi$. $\endgroup$
    – user539887
    Commented Apr 23, 2018 at 11:25
  • $\begingroup$ @user539887 you are right, indeed one has to have some additional assumptions for what I claimed to be true. Some inequalities relating $A, b$ and $\phi$ should be satisfied. In 1D for $A=-1/10$, $b=-1$ and $\phi(x) =\frac{ e^x}{100} $. The conditions in 1D are not difficult to write, but it becomes less obvious in higher dimensions when $A$ is not diagonal. $\endgroup$
    – demitau
    Commented Apr 23, 2018 at 18:52
  • $\begingroup$ I think that you are interested in solutions staying, for positive time, in the nonnegative orthant $\mathbb{R}^d_{\ge0}$. Am I right? $\endgroup$
    – user539887
    Commented Apr 23, 2018 at 19:04
  • $\begingroup$ @user539887 not necessarily. In my mental picture all the solutions from minus infinity approach the stable point, For the parameters above, the stable point is close to $x=-10$ and the unstable one is close to $5$. The stable point has a non-compact basin of attraction. If we have diagonal $A$ and coordinate-wise repetition of parameters, the picture persists but now there are some saddle points appearing too. $\endgroup$
    – demitau
    Commented Apr 23, 2018 at 19:14

0

You must log in to answer this question.

Browse other questions tagged .