Why aren't there computer simulations for these things?

The cosmos at our fingertips.
Post Reply
longtry
Ensign
Posts: 55
Joined: Tue Dec 11, 2018 6:59 am

Why aren't there computer simulations for these things?

Post by longtry » Tue Nov 29, 2022 4:42 am

The first one I think of is bipolar jets. If their cause and mechanism are debated so hotly, why haven't someone build a model and run it on a supercomputer? Then by removing or adjusting some component - say, magnetism - the answer must be found soon enough?

The second one is the solar system's history. If we can predict objects' orbits and calculate spacecraft's trajectories so accurately, why not build a simulation that runs backward into the past? At some points when things seem missing, that must be the moment an event happened, like a collision, or an ejection of a planet, etc.

Since people don't seem to be doing those things, I guess there must be reasons. What are they? Do you have any predictions on when this type of simulation will be possible in the future (or never)?

User avatar
Chris Peterson
Abominable Snowman
Posts: 18102
Joined: Wed Jan 31, 2007 11:13 pm
Location: Guffey, Colorado, USA
Contact:

Re: Why aren't there computer simulations for these things?

Post by Chris Peterson » Tue Nov 29, 2022 3:40 pm

longtry wrote: Tue Nov 29, 2022 4:42 am The first one I think of is bipolar jets. If their cause and mechanism are debated so hotly, why haven't someone build a model and run it on a supercomputer? Then by removing or adjusting some component - say, magnetism - the answer must be found soon enough?
There are. Indeed, these are modeled all the time. But they are only as good as the understanding of the underlying physics. Like the nature of the magnetic fields. Most of our understanding of jets comes from numerical modeling.
The second one is the solar system's history. If we can predict objects' orbits and calculate spacecraft's trajectories so accurately, why not build a simulation that runs backward into the past? At some points when things seem missing, that must be the moment an event happened, like a collision, or an ejection of a planet, etc.
This is also done. But N-body systems are chaotic, so there is no stopping the accumulation of errors over time. The state of the Solar System is not deterministic, past or future.
Chris

*****************************************
Chris L Peterson
Cloudbait Observatory
https://www.cloudbait.com

longtry
Ensign
Posts: 55
Joined: Tue Dec 11, 2018 6:59 am

Re: Why aren't there computer simulations for these things?

Post by longtry » Wed Nov 30, 2022 7:02 am

these are modeled all the time. But they are only as good as the understanding of the underlying physics. Like the nature of the magnetic fields. Most of our understanding of jets comes from numerical modeling.
Hmm. I understand that as, "relativistic experiments couldn't exist/be done effectively without Einstein". If that's correct, then what kind of theoretical physics we're lacking in order to make breakthroughs in decoding polar jets?
But N-body systems are chaotic, so there is no stopping the accumulation of errors over time. The state of the Solar System is not deterministic, past or future.
Thanks. What is the farthest to the past (and the future) of the Solar system that simulations have done (with high certainty)?

User avatar
Chris Peterson
Abominable Snowman
Posts: 18102
Joined: Wed Jan 31, 2007 11:13 pm
Location: Guffey, Colorado, USA
Contact:

Re: Why aren't there computer simulations for these things?

Post by Chris Peterson » Wed Nov 30, 2022 3:27 pm

longtry wrote: Wed Nov 30, 2022 7:02 am
these are modeled all the time. But they are only as good as the understanding of the underlying physics. Like the nature of the magnetic fields. Most of our understanding of jets comes from numerical modeling.
Hmm. I understand that as, "relativistic experiments couldn't exist/be done effectively without Einstein". If that's correct, then what kind of theoretical physics we're lacking in order to make breakthroughs in decoding polar jets?
There are just a lot of unknowns. We don't really know what the interior of these stars looks like, for instance. So there are lots of free variables in the simulations (aka "fudge factors") that are manipulated until the output looks like reality, and then hopefully those values can inform us about the underlying physics. It's an iterative process.
longtry wrote: Wed Nov 30, 2022 7:02 am
But N-body systems are chaotic, so there is no stopping the accumulation of errors over time. The state of the Solar System is not deterministic, past or future.
Thanks. What is the farthest to the past (and the future) of the Solar system that simulations have done (with high certainty)?
It depends on what we're looking at, and what sort of precision we need. The numerical simulations I do involve the motion of dust and small bodies, and that's pretty good over a few thousand years. The best ephemeris generators for the larger bodies look about 10,000 years past and future before the errors get significant. One problem is that they are only as good as our knowledge of the actual position of the planets, and that's not as good as you might think, especially for the outer planets that we've only observed for an orbit or two (or less than an orbit).
Chris

*****************************************
Chris L Peterson
Cloudbait Observatory
https://www.cloudbait.com

longtry
Ensign
Posts: 55
Joined: Tue Dec 11, 2018 6:59 am

Re: Why aren't there computer simulations for these things?

Post by longtry » Thu Dec 01, 2022 5:06 am

they are only as good as our knowledge of the actual position of the planets, and that's not as good as you might think, especially for the outer planets that we've only observed for an orbit or two (or less than an orbit).
I kinda see the problem with them. This dependence on outer factors might mean we'll only get really high quality simulations in, like, 100 years.

Can we develop a machine-learning AI which can tinker with the inputs of simulations (especially cases like the bipolar) and speed up the guessing process of iterations?

User avatar
Chris Peterson
Abominable Snowman
Posts: 18102
Joined: Wed Jan 31, 2007 11:13 pm
Location: Guffey, Colorado, USA
Contact:

Re: Why aren't there computer simulations for these things?

Post by Chris Peterson » Thu Dec 01, 2022 5:09 am

longtry wrote: Thu Dec 01, 2022 5:06 am
they are only as good as our knowledge of the actual position of the planets, and that's not as good as you might think, especially for the outer planets that we've only observed for an orbit or two (or less than an orbit).
I kinda see the problem with them. This dependence on outer factors might mean we'll only get really high quality simulations in, like, 100 years.

Can we develop a machine-learning AI which can tinker with the inputs of simulations (especially cases like the bipolar) and speed up the guessing process of iterations?
Can't get around the chaotic nature of the system, though. It's likely impossible to accurately predict the positions of the planets beyond a few tens of thousands of years, even given perfect knowledge of their orbital parameters.
Chris

*****************************************
Chris L Peterson
Cloudbait Observatory
https://www.cloudbait.com

Post Reply