Using lombscargle in orbital dynamics - question to adjust parameters

[gacarita]

Hi,

I am trying to use astropy.LombScargle in orbital dynamics to find resonances between giant planets and test particles, my orbits are not Keplerian and my integrator has adaptive time steps.

Running a test on an orbit that has a 1/1 resonance I found nothing close to 1/1 on the LS graphs.

Does anyone have any suggestions?

Here is my code:

ls = LombScargle(data_ptest_cartesian.t, data_ptest_cartesian.x,nterms=2)
ls1 = LombScargle(data_planet_cartesian.t, data_planet_cartesian.x,nterms=2)

freq,power = ls.autopower()

freq1,power1 = ls1.autopower()

plt.plot(freq, power,color="black")
plt.plot(freq1, power1,color="red")

I expected to see two central peaks very close together. I'm inserting the time and inertial cartesian coordinates for both bodys.

[pllim]

Thanks for reaching out! If you don't get replies here, you might want to try one of the mailing lists listed at https://www.astropy.org/help .

[adrn]

Hi @gabiruzo - Depending a bit on the context, LS (or at least the default, single-term LS) might not be the best approach here. Especially since you mention the orbits aren't Keplerian, there will be a few signals with different frequencies present in any cartesian coordinate you look at. So it might be helpful to consider doing something like orbital frequency analysis instead (e.g., http://github.com/adrn/superfreq). But I guess if you have non-uniform timesteps, you would either have to snap/interpolate your timesteps to a uniform grid, or would have to use something like a non-uniform Fourier transform instead (e.g., https://cims.nyu.edu/cmcl/nufft/nufft.html). If the frequency structure is simple, you might be able to get away with doing a multi-term LS...but I wouldn't count on it!

[gacarita]

Hello @adrn

For my case of 3 bodies and I am finding different frequency values ​​for the 1/1 type resonances, what I expected with the ls was to observe the ratio relationship between my test particle and the secondary body, I know that these relations are not they must be exact but for each 1/1 resonance orbit I am getting different values ​​between the second body and the test particle.
I am using a simplification of my integrator to obtain equispaced steps to use fft but the results seem strange just like in ls.
I would like to test SuperFreq but I need to better understand how it works.

would you have any suggestions?