# Parametric plot of the critical points of an ODE

I have a single ODE of the form

x'[t] = -(y*n*n + z*n) x[t] + w (1-x[t]) x[t]


where the second part of the ODE is the logistic growth with the maximum allowed population set to 1 and growth rate equals to w.

The first part is some sort of polynomial exploitation/growth function whose effect is defined according to y, z and n.

I am only interested in the values of x[t] within the range [0, 1].

I am able to plot and manipulate the stream plot for it:

x'[t] == f[t, x[t]];
fo1[t_, x_] := -x*(y*n^2 + z*n) + w (1 - x) x /. {w -> 1};
Manipulate[
Show[
StreamPlot[{1, fo1[t, x] /. {y -> my, z -> mz, n -> mn}}, {t, 0,25}, {x, -0.25, 1.25},
VectorPoints -> Automatic,
VectorStyle -> Black, StreamStyle -> Orange,
VectorStyle -> Arrowheads[0.01], ImageSize -> Medium,
PlotRange -> {{0, 25}, {-0.25, 1.15}}]],
{{mn, 1, "[0,100]"}, 0, 1},
{{mz, 1.05, "[-1,1]"}, -2, 2},
{{my, -0.345, "[-1,1]"}, -2, 2}]


from which it can be seen that according to the parameter y, z and n (assuming w ia fixed) the number and type of the critical points change.

With the given values, for instance, there is an unstable point at 0 and a stable point at ~0.3

What I would like to achieve is a parametric plot with two of the above parameters on the Cartesian axes (e.g., y and z) and which divides the plot into regions according to the number of critical points present in it. If possible, I would also like to each region with its type.

For instance something like what is shown in the figure below: So far I am not able to go any further than making the stream plot. I have no idea on how to proceed.

Any help is appreciated.

• You could use ParametricNDSolve. What is the initial value x? The condition for a critical point is x'[t]=0  ? Jan 7, 2019 at 21:15
• Is u[t] supposed to be x[t]? Jan 7, 2019 at 21:21
• The ODE seems to have a 1D phase space -- there are no saddles. Generically, you have either 1 source/1 sink. Jan 7, 2019 at 21:34
• @MichaelE2 yes, I apologize for the mistake. And yes I know that there is no saddle in this specific case. Thanks. Jan 8, 2019 at 9:17
• @UlrichNeumann the initial value x might have different values. But for now, it is okay to fix it (e.g. 0.5). Jan 8, 2019 at 9:29

Analysis is easier if we define a new compound parameter, c=y*n*n + z*n. The resulting equation is still a logistic equation, with two equilibria:

Solve[0 == -c x + (1 - x) x, x]
(* {{x -> 0}, {x -> 1 - c}} *)


The trivial equilibrium is unstable when $$c<1$$ and stable when $$c>1$$. The latter situation corresponds to the second equilibrium being negative, so that's the condition for one vs two meaningful (non-negative) equilibria.

Now we can substitute the definition of c in terms of y and z to solve for the critical boundary between these cases:

Solve[1 == (y*n^2 + z*n), y]
(* {{y -> (1 - n z)/n^2}} *)


This is a line, which looks like:

n=1;
Plot[(1 - n z)/n^2, {z, -1, 1}, AxesLabel -> {z, y}] Above the line, there is only x=0 which is stable. Below the line, x=0 is unstable and there is a positive, stable equilibrium.