Gravity Simulator
General >> Discussion >> Weird jump after 50k years

Message started by Mal on 02/18/07 at 01:37:26

Title: Weird jump after 50k years
Post by Mal on 02/18/07 at 01:37:26

I'd be curious to know what people think is going on here - I'm running this at a 2048 timestep and I'm about 360,000 years into the simulation. (it's the Regina system from this thread (, without the white dwarf orbiting close to the primary star - note that in the attached file you'll have to manually delete Speck before running it). Olybrius (planet 4) is fairly wobbly, it has the thickest 'orbital envelope' of all the planets so far - but everything is still stable.

However, I notice something odd - around 50,000 years the eccentricity of Clement (planet 1) suddenly leaps up and starts oscillating around a new value. At the same time, Olybrius' eccentricity cycle changes as well. The ecc cycles of the other planets seem to shift upwards to some extent at the same time too. if you take a look at their Arg of Pericentres then you'll see the pattern there changes a bit as well. I didn't change any parameters halfway through or anything... can anyone reproduce this (or take a closer look, I saved every 100,000 years so I missed this).

It almost looks like something locked into a resonance, can GravSim do that?

Title: Re: Weird jump after 50k years
Post by Mal on 02/18/07 at 11:03:41

I'm doing another run at 2048 timestep with the same system, but recording output data every 100 years and autosaving every 5000 years,  up to 70,000 years. I'd take screenshots but unfortunately it seems you have to have the gravsim window open and on top all the time and plotting out the trails too in order to take the screenies - this isn't very useful because it slows down the simulation a lot and it's inconvenient because if you get something over the gravsim window then it won't take the screenshot anyway.

Hopefully this might reveal what's going on at 50k years. I think the ecc cycle of all the other planets except Assiniboia change/shifts up a notch around that time...

Title: Re: Weird jump after 50k years
Post by Mal on 02/18/07 at 14:04:03

Well this is really odd (see attachment. You don't need to worry about deleting Speck in this one, and I've included an autosave file starting at 20,000 years if you want to run it from there).

I ran the same sim again for 70,000 years with a 2048 timestep (recording output data every 100 years, autosaving every 5000 years), and nothing happened at 50k years this time - instead something odd happened at 28,000 years. If you plot the eccentricities, you'll see that at that time Olybrius's cycle doesn't go as far as it does around that year, and that there's a definite increase in Clement's eccentricity there (the curve suddenly becomes a straight line upwards)...

Anyone got any idea what is going on here?

Title: Re: Weird jump after 50k years
Post by Tony on 02/18/07 at 18:39:11

I haven't run the simulation, but it sounds similar to something I've encountered.  I started with the solar system, increased Venus' mass to 2 Jupiter masses.  After a certain amount of time Mars was ejected from the solar system.  I was surprised that Earth and Mercury were not effected.  Mars seemed unaffected at first, but when I let it run overnight, Mars was halfway to the stars by morning.  Reviewing the Autosaved simulations, Mars appeared stable until only a few centuries before its ejection.  Then its eccentricity started growing until it crossed the orbits of other planets.

If I ran this at a different time step, I got the same result, but after a different length of time.  The conclusion that I drew was that there is some planetary alignment that pushes Mars to the point of no return.  Maybe it happens on average once every 100,000 years.  And changing the timestep introduced enough chaos into the simulation that it shuffled the deck.

My guess is that although your simulation is very sensitive to initial conditions, it still demonstrates that Olybrius is not stable over periods of 10's of thousands of years.

Title: Re: Weird jump after 50k years
Post by Mal on 02/18/07 at 19:27:05

Yeah, but I didn't change the timestep :) - all I changed was how often the results were being recorded.

Title: Re: Weird jump after 50k years
Post by Tony on 02/19/07 at 02:19:49

Then that is weird.  

One possibility, but I'm not sure if it applies to your situation...

In a simulation where your final state is very sensitive to your starting conditions, you have to make sure that your initial conditions are exactly the same between two runs.  For example, if you create the simulation at timestep=1, run the simulation, and while the simulation is running, increase the timestep to 2048, then you are introducing something you can not duplicate between runs.  The next time you try this, you almost certaily won't increase the time step at the exact same rate.

I'm not sure if this is what is happening here, or if it is something else.  But if you are going to do 2 identical runs of a simulation, make sure you start paused, and adjust the timestep only when paused.

I'll download it tommorow and give it a try.

Title: Re: Weird jump after 50k years
Post by Mal on 02/19/07 at 08:03:48

That's possibly what might have happened.

Which raises the point, why do you set the program to start running time when it's turned on in the first place? Surely it's better to have time paused from the moment you start the program, that way you'll know exactly when the simulation started and don't have it running at the start to mess everything up.

Title: Re: Weird jump after 50k years
Post by Tony on 02/19/07 at 11:07:24

You can begin any simulation either paused or unpaused.   For example, open Gravity Simulator.  It opens the resume.gsim file.  Gravity Simulator always fires-up unpaused.  Pause resume.gsim, and open your simulation.  Your simulation will begin paused simply because Gravity Simulator was paused when you opened it.  Then you can adjust the time step while paused and ensure that multiple runs of the same sim will both begin on exactly the same "square 1".  But if you opened your simulation with Gravity Simulator unpaused, then your simulation will begin unpaused too.

If you accidently open your simulation unpaused, pause it and open it again.  It should be the first simulation on the "recently opened" list at the bottom of the File menu.

Title: Re: Weird jump after 50k years
Post by Mal on 02/19/07 at 14:31:38

Ok, but surely it'd be better to open every simulation in a paused state automatically? That saves the user from having to pause things while he sets up a new simulation for example. I would have thought it'd be better to for a user to consciously decide to start the simulation by unpausing it than to guess whether it was paused or unpaused last time it was opened.

Title: Re: Weird jump after 50k years
Post by Tony on 02/19/07 at 15:13:06

Mal wrote:
... than to guess whether it was paused or unpaused last time it was opened.

It doesn't matter whether it was paused or not the last time it was opened.  Gravity Simulator will always open unpaused.  Normally, when you open Gravity Simulator, you are not opening your simulation.  Rather, you are opening resume.gsim.  Just press pause, and open your simulation, and your simulation will open in the paused state.  The exception is if you have established a file association between Gravity Simulator and the .gsim file extension.  If you have, then you can immediately open your simulation by double clicking on yoursim.gsim.

There are two main reasons I like it opening unpaused.

1.  More often than not, I want my simulations to start right away.

2.  For a brand new user who just downloaded the program, the first thing they see when they run the program for the first time is a running simulation of the solar system.  After double clicking the Gravity Simulator executable icon, the planets immediately begin tracing paths.  This is supposed to generate a Wow effect.    If it started out paused, they would see a blank screen.  Wow effect would be gone. They'd have to read the directions to know to unpause the program.  And that would scare away a lot of people.

It would be possible to make the .gsim file remember the paused state of a simulation, but all it would be saving you is the trouble of pressing pause yourself, then choosing File > 1st file in the recent files list.

Gravity Simulator » Powered by YaBB 2.1!
YaBB © 2000-2005. All Rights Reserved.