Is it possible to do astrophysics research

Assuming that you know all the basic physics and have a good idea about the experimental methods in the field, you would have to know how to use Mathematica, MATLAB, C/C++, Fortran...things of that sort. But thats just part of the story. People who do it at the university would be using at least all this too

 

To do cutting edge reserach you would need computing power that you can't have at home (unless you are a multi-millionaire and are willing to throw in a LOT of money to get a cluster built). Then you would need the expertise to program it and to run it or hire people to do that (more money). Finally, you would need to have the physics expertise an dyou would probably end up having to spend time in a physics department anyway to discuss with people there and get some fresh ideas. It's not really feasible to do cutting edge computational physics at home. Only in pure theory could someone do research at home and even in that case it' srare that people can work in isolation, they would still want and need to spend time in a good physics department.

 

Ah! When I hear "doing research" it usuallly mean trying to discover something new. If you just want to do simulations and not research then for sure you can do a lot at home. But you might end up reinventing the wheel so a careful search on the web will probably reveal a lot of stuff already written in many different languages at at different levels of sophistication.

 

Yes, but after a while you also need to learn these languages/environments seriously rather than just playing around with code others have written. You don't want to be stuck with a coding (note that I did not use the word 'programming') problem when you're working on a scientific problem.

 

Start with something simple like solving heat/diffusion equations, then move into coupled systems of PDE's/ODE's. You could even write a code in a couple hundred line code for Bondi accretion.

 

What have you taken for math and physics courses?

 

I'll second what's been said about learning a language like C++ or fortran if you want to run large mathematical simulations. Pretty much everything in astronomy is written in one of those two. That being so, the amount of programing skills you actually need are pretty minimal. After all, most astro codes don't have user interfaces or display output; they just do lots of math over and over again. You probably could go far just knowing how to structure loops, write conditionals, and read and write to a file.

Knowing what math you want to computer to do, however, is a separate question.

Anyways, if you know the physics you'll be using, all you really need is a decent desktop running some linux or unix variant, a copy of gcc or gfortran (or some version of this that works on Windows), and some patience. Unless you want to do *really* detailed projects like supernova explosions or giant planet atmosphere models (which would require a supercomputer, or a cluster, at the least), you're looking at maybe 100 hours of CPU time, on the longer end. Most of the stuff I do takes shorter than 2 hours to complete on my desktop.

As an example of a CPU intensive code, check out John Chamber's Mercury integrator (http://www.arm.ac.uk/~jec/), which handles orbital dynamics, and has been used to model planet accretion.

Edit: Also, the book Numerical Recipes is great for looking up ways of doing standard routines (like integrating, parameter fitting, etc.)

 

Look up Numerical Recipes in C or even the 3rd edition is really good, but is written for C++ I believe. It seems to be the standard introduction to numerical methods for those in research. I find the book a bit sophisticated sometimes, and their coding style is a bit awkward, so their example codes may be a bit tough to decipher.

The only way your going to learn coding is to practice and write the codes on your own, so start small and learn how to solve heat equations etc.

 

You could always build a supercomputer out of PS3's running linux.

http://gravity.phy.umassd.edu/ps3.html

 

Frankly, I don't think this is realistic.

If you're playing around with simulations, the most you can possibly discover is something about the simulation, not anything about astrophysics.

You also have the twin problems that your resources are probably nowhere near as great as someone in a university, and you have many fewer opportunities for collaboration. Lest you think that is unimportant, I looked at the top cited papers in astrophysics (as of 2006) and of the top 50 papers, only three have single authors.

Finally, if it all it took for a major accomplishment were desire, and anyone with interest could do this from home, we'd be seeing a lot more results from people at home.

 

Most science comes in baby steps. Major turning points are rare. As long as you're aware of and content with this, where's the lack of realism?

Well, yes and no, but simulations as a research method do have a flaw pretty close to what you're saying. It can let you see how a theory plays out and demonstrate that it's plausible, and has been instrumental in helping choose between competing models on a few occasions. It is however highly interpretive and does more to support than prove theories.

Well yeah. That's just how it works these days in the academic arena. Have you looked at any of the papers out of CERN or Fermilab? Thousands of co-authors. How many do you want to bet actually wrote the paper?

There's plenty of data in observational astronomy that's contributed by amateurs. Not nearly as much as comes from major observatories, but a meaningful amount. Particularly in areas that are suited to smaller telescopes, for example data on variable stars. In the physics parts, the bar to entry is a bit higher, but it's hardly impossible to do meaningful research on some scale.

 

As it happens, I'm an author on a couple hundred of those papers. But aren't you making my point? "That's how things are done". Not by one person working from home.

You make a good point about amateur astronomers, and had the OP say that he wanted to do original research looking for, say, comets or minor planets, I would have been a lot more encouraging. I don't think one should consider it likely that one will make original discoveries in astrophysics running simulations from home.

 

I don't think simulations are one of the easier ways to go about it since in many cases it might require cluster time, but isn't the reason those papers have so many authors is because they contributed to gathering the relevant data? The actual analysis and writing for a particular paper is done by a small handful of people, run by a committee, and left for a period to see if anyone whose name is being put on it has objections or comments to raise.

The other issue is that scientists often consult other scientists when they need to deal with something outside their area of knowledge. In a national lab or a university, there are lots of other scientists about who could potentially help. Outside this environment, it's not as easy to make these connections, further limiting the range of problems you can try to work on. But in neither case is it prohibitive...if you have an interesting research proposal and are going about it professionally, I don't know that anyone *once you've done the additional work to develop access to these resources* would turn you away simply because you're working outside the traditional support structure.

For example, proposals for time on the Hubble telescope can come from pretty much anyone...researchers inside the traditional structures will have an easier time knowing what to do and getting help doing it when needed, but if you accept the limits of working outside that system and do the additional work, it won't keep you from doing useful science.

 

While I think it is an ambitious goal (to make discoveries in computational astrophysics at home that are genuine contributions to the existing body of knowledge), it seems that people working under significant restrictions or at large disadvantages often figure out an elegant method to solve a problem.

I'd say go cheap--probably build a basic, modern computer with lots of memory and a decent processor, and probably a decent graphics card and dual monitors (code in one, output/graphics in the other). Install linux, and learn the standard languages (C++, Fortran, whatever else is in vogue at the moment).

I might suggest that you get your feet wet by reading a book called "Black Holes: A Traveler's Guide" by Pickover. It's not a textbook (by ANY means) but some of my students have found the explanations and the code in the book to be useful as a way to get their feet wet in programming physics simulations.

 

While it would be an uphill battle, if a person has an interest in tackling astrophysics problems in his or her spare time by running simulations at home, I say go for it. I agree that university-type resources would not be available for large scale simulations, but are large scale simulation problems the only ones left to be solved in this field?

Even if this were the case, I think there is value in starting small to learn the basics of the problems one is trying to tackle.