# Is there a mathematical way to calculate the bifurcation of a system?

1. Jan 25, 2007

### impendingChaos

Is there a mathematical way to calculate the bifurcation of a system?

2. Jan 25, 2007

### Crosson

Yes, take the case of a one-dimensional flow x' = f(x). If we have f(x) depend on a parameter r, then as r varies we observe bifurcations in the system.

More importantly, what do you mean by 'calculate the bifurcation'? If you mean to classify the type of bifurcation then the process is as I described above. If you mean to calculate the location of bifurcations on for example the period doubling route to chaos then this is another story.

3. Jan 26, 2007

### J77

Of course, the most basic way would be to calculate the eigenvalues from the jacobian of the linearised system.

Standard codimension one bifurcations (of steady states) can then be found as eigenvalues pass through the imaginary axis - upon variation of parameters.

Saddle-node bifurcation: real eigenvalue passes through.

Hopf: imaginary pair pass through...

(See Floquet theory for periodic solutions.)

4. Jan 26, 2007

### impendingChaos

Crosson, I believe it would be best for me to calculate the location of the bifurcations on the way to chaos as you say. Let me explain what I am working with a little better. I am basically analyzing the nonlinear data of a diode to look for chaos. Seeing that bifurcation was used to model chaotic behavior I was wondering if I could use this concept of bifurcation to analyze my nonlinear data and show its progression towards the hypothesized chaos. So how does one calculate these bifurcation positions?

5. Jan 26, 2007

### J77

You should read up on what a bifurcation is first.

The bit I've highlighted makes no sense.

If you have data only, you should plot the power spectra (via. fft) of it - periodic solutions will show as a single peak - if you have period-doubling, the peaks will double in quantity as you calculate the spectra at a different parameters (for the appropriate data); chaos will show as a broad spectra.

6. Jan 26, 2007

### Crosson

Everyone in this field uses TISEAN for nonlinear TIme SEries ANalysis. It is free and open source, and it contains a large number of tools to distinguish linear systems, from nonlinear systems, from stochastic systems.

http://www.mpipks-dresden.mpg.de/~tisean/

The reason I am suggesting this is because it sounds like you have data, but no mathematical model, which means that the analytical methods suggested by J77 don't apply.

7. Jan 26, 2007

### J77

The method in my second post is pretty standard for data analysis.

Know someone interested in this topic? Share this thread via Reddit, Google+, Twitter, or Facebook