Computational and Experimental Work

[absurdist]

Both computational and experimental physicists' jobs are to match the results proposed by the theory.
Can someone explain to me how the computational physicist studies nature? (ie does physics)?
Also in subjects such as computational particle physics, the only reason we need models is because its difficult to get experimental data? How then can we be sure that these models are representative of what's seen in nature?
Please correct me if I am wrong.

 

[nasu]

Both computational and experimental physicists' jobs are to match the results proposed by the theory.

I think that you've got it a little backwards. The experimental physicist test the theories. It is the job of theoretical physicist to match theories to the experimental data and not of the experimentalist to match data to theories (even though this may happen once in a while, intentionally or not).

 

[absurdist]

Sorry, my bad. But can you tell me if the computational physicist's job is effectively the same as the experimental physicist's? If yes, how do we make sure the results predicted by the computational physicist are right? And how are they useful in the study of nature?

 

[luben]

To me, I think computational physics is to try to compute the solution of a problem using a computer, which means, most of the time and to be more specific, solving a set of equations which is very very tedious to solve by hand or by theoretical tools. Once you get a solution, then it should correspond to both theories AND experiments if there are any (and given they are known to be correct ,of course). If there is nothing for you to compare with, then it may be harder for you to verify your solution.

 

[absurdist]

Let me get this straight: its not really much of a science then is it? It's just number crunching. To what end is computational physics used in terms of physics as a natural science?

 

[luben]

It is using much mathematics and on the other ends related to physics and engineering fields. It also requires knowledge of computer sciences, so sort of something a mixture of those. It has its own theories about what is a correct computation. From this viewpoint, it is a science of course.

About it uses, computaional physics is applied in may fields from product designs to aerospace projects. One of its strength is that, it gives you the whole solution (i.e. knowing everything in the field), which is usually costly/not possible from an experiment.

 

[nasu]

Let me get this straight: its not really much of a science then is it? It's just number crunching. To what end is computational physics used in terms of physics as a natural science?

Sometimes computational physics is somehow a replacement for experimental physics.
If you want to see what the results of the theory will be in a specific, complicated case, you can do some experiments or you can run a simulation.
The experiments can be sometimes too expensive or time consuming or impractical from other points of view.

 

[absurdist]

I agree completely about whatever you've both said but how is computational physics useful in terms of explaining the mysteries of the cosmos? I don't think it can at least directly explain anything about the nature of the universe.

 

[luben]

May I ask what sort of mysteries of cosmos in particular you are looking forward to be explained? Perhaps some specific examples can ring something up from my head

 

[absurdist]

Ok maybe I was exagerrating a little bit, but I meant stuff like this:

Applications of computational physics:
Computation now represents an essential component of modern research in accelerator physics, astrophysics, fluid mechanics, lattice field theory/lattice gauge theory (especially lattice quantum chromodynamics), plasma physics (see plasma modeling), solid state physics and soft condensed matter physics. Computational solid state physics, for example, uses density functional theory to calculate properties of solids, a method similar to that used by chemists to study molecules.

Courtesy Wikipedia

Also like modelling the behaviour of galaxies and similar stuff.


So where exactly does the aspect of physics as a natural/fundamental science come in?
Stuff like the theory of everything, behaviour of subatomic particles, etc. Where do you as a comp.phy get involved in this while you are modelling stuff?

 

[bcbwilla]

I agree completely about whatever you've both said but how is computational physics useful in terms of explaining the mysteries of the cosmos? I don't think it can at least directly explain anything about the nature of the universe.

Define "directly". Does theory directly explain anything about the nature of the universe? Does experiment? Or do you need both? Computational physics is often viewed as a method to bridge the two different aspects of physics when the theory is so complicated that you can't just solve a simple equation to get your answer.

You mentioned density functional theory in another post, which is a good example. Density functional theory is used to compute various chemical and material properties. It's pretty darn hard to solve the Schrodinger equation directly for anything more complicated than a Helium atom. This is where computers come in, which will happily iterate for days, minimizing the energy of a complicated system of atoms - which is basically impossible (impractical) to do by hand. A lot of theory needs to be developed first before these calculations can be made (hence the "theory" in density functional theory). Once they are made, the results can easily be compared to experiment.

Science is a team sport.

 

[cepheid]

Take numerical relativity as an example. I know some numerical relativists whose job it is to set up computer simulations of binary black hole systems in which the two black holes are spiraling in towards each other and will eventually merge. This problem is too complex to solve analytically (i.e. with pencil/paper). The binary black hole system produces gravitational waves throughout this process. The purpose of the simulation is to generate a gravitational waveform that then acts as a template for a gravitational wave detector like advanced LIGO. This template essentially tells them what they are looking for, so that they can attempt to pick out just such a signal from the noise. So I see numerical methods as an essential extension to the theorists' arsenal that allows them to do their job, which is to construct mathematical models of physical systems.

On the experimental side, data sets are only getting larger. A great deal of computing power is needed just to analyze them, often using parallel computing. A good experimentalist, in addition to being good with a wrench or a soldering iron, also needs to be a good computer scientist and statistician.

The bottom line is the computational methods are essential to 21st-century physics. Without them we would not be able to continue our investigation into the outstanding problems in our understanding of nature.

 

[absurdist]

Sorry to make you labour the point (this is not my forte as you can tell) but is this it ?

If you are checking a theory's predictions for complex systems where equations should be solved computationally, you are finding out more about the natural behaviour of the system if the theory's predictions then meet experimental results? But you know the theory has been used before and checked with experimental data. So if you mentioned schrodinger's equation, its only being applied to some other system but not to find out something completely new abt the fundamental behaviour of matter.

Calculating properties for complex systems is mainly what computational physicists do right?Theres not much fundamental science-related discoveries going on right?

From my research in engineering, you start off with theories and use them to compute models for more complex systems but that doesn't help develop a theory of any sort, only a model (approximation) used for some applications.

 

[Astronuc]

Let me get this straight: its not really much of a science then is it? It's just number crunching. To what end is computational physics used in terms of physics as a natural science?

Computational physics is very much a science. Think of how one would solve a system of coupled nonlinear PDEs that describe the physics of some physical system - e.g., a fission or fusion nuclear reactor, a propulsion system, a star or a class of stars, the formation of proto-stars, a supernova, a galaxy, a cell, a planetary weather system, a planetary tectonic system, . . . . One cannot do that by hand or a set of analytical equations.

One must use a complex numerical calculation to solve those for those systems. One must address multiple scales in dimension and time, and at each statepoint in the system, one must find a converged solution. There is a lot of applied mathematics and science in solution methods and numerical models.

If one wants to explore the technical limits or performance of an energetic system, one can numerically simulate a system that one could not perform experimentally without catastropic damage and potential fatalities and injuries to folks.

In my work, we simulate high energy input into nuclear fuel systems to explore under what conditions they fail. To perform experiments would be prohibitively expensive and problematic from the standpoint of radioactive contamination of the experiemental facilities. A limited number of experiments are performed, and from those experiments models are developed for predictive analysis. With predictive analysis, many more scenarios can be simulated/explored for a fraction of the cost of experiments.

Similarly, observational astronomers and meterologists can build models that enable them to predict future events based on an understanding of a limited number of observed events, or at least better understand a new observation.

 

[absurdist]

So I think then that computational physics has more importance in applied physics rather than in physics as a natural/fundamental science. Unless of course your simulation leads you to discover something surprising that no one saw coming and explains some great mystery in physics.
Has that ever happened?

 

[cepheid]

So I think then that computational physics has more importance in applied physics rather than in physics as a natural/fundamental science. Unless of course your simulation leads you to discover something surprising that no one saw coming and explains some great mystery in physics.
Has that ever happened?

Why do you think that? Because of the examples we just so *happened* to give? How the heck is astrophysics or general relativity "applied" physics? In the example I gave, you can vary the model parameters, including the mass ratio of the two black holes, their spin rates and their spin alignments, and see how the results change. Furthermore, when you combine the simulation efforts with the gravitational wave detection experiments, you DO realize that the overall goal is to test predictions made by general relativity about a fundamental aspect of nature: how gravity works.

Large cosmological N-body simulations that attempt to model the distribution of matter in the universe on the largest scales vs time are another example (and don't try to tell me that cosmology is applied physics). The simulation results (and their comparison with observations) have helped pin down some of the fundamental parameters in the cosmological model, including the densities of the universe's constituents: baryonic (atomic) matter, dark matter, and dark energy.

What is your definition of "fundamental?" Does your definition include only attempts to go beyond the standard model of particle physics? If so, it is too narrow. And I should point out that particle physicists undoubtedly make good use of numerical models in their work. I just can't go into detailed examples because it's not my field.

 

[AlephZero]

I would say at least 99% of real world working physicists don't spend any of their time even thinking about "great mysteries of physics", let alone explaining them. Mostly they leave "great mysteries" to the people who write pop-sci books.

But at a more mundane level, "discovering something surprising that no one saw coming through simulation" happens all the time. That's one good reason why people run simulations.

 

[absurdist]

Sorry, it was the most awful way possible to explain my question. What I meant by applied physics (and again, sorry) was that you are just testing the predictions of a theory on a more complex system than what it was already tested on. You already know how gravity works based on the theory, now you are just testing it in a more complex system. What's the purpose of that? To keep checking its universality?

And by your responses I think I am probably wrong but I was under the impression that computational physics is a more important (notice i wasnt categorical in the previous post) tool to engineering and applied physics than in pure physics (theoretical), where it is used to model only complex systems. Have all the feasible simple systems been tested experimentally for scientists to give so much importance to models now and forgive me but I am oblivious to modern science as I am not a scientist.

Ok finally, you scientists may consider me stupid for asking this but: Why do we keep testing theories if most models or experimental data have shown that they are right?
Why can't we move on to finding another new theory for someother property/behaviour in nature?

And again sorry, most of what I know abt science is the romanticized version of what's shown on TV.

 

[absurdist]

I would say at least 99% of real world working physicists don't spend any of their time even thinking about "great mysteries of physics", let alone explaining them. Mostly they leave "great mysteries" to the people who write pop-sci books.

But at a more mundane level, "discovering something surprising that no one saw coming through simulation" happens all the time. That's one good reason why people run simulations.

But I don't understand, as an intellectual race, why isn't our primary goal to solve these mysteries. WHY aren't we focussing all our faculties and resources to this one goal of using physics, math and biology to explain our existence. I mean I am not smart enough but I would die happy thinking I made some effort. Is it because of our capitalistic society?

Or do you just think I have been watching Brian Greene, Through the wormhole w/ Freeman, etc too often..

 

[cepheid]

You can't investigate most interesting problems in theoretical physics today without doing numerical simulations.

We still don't understand fully how a core-collapse supernova explodes. If you think that a simulation of this is just "simulating a complex system whose underlying physics we already understand" then you're wrong. I would argue that we don't understand something about the underlying physics. Furthermore, I don't think anyone fully understands the equation of state of matter in the core of a neutron star. Simulations are essential for seeing whatt new theories predict about such systems.

As AlephZero has alluded to, most physicists aren't in the business of trying to discover new "laws" of nature, but that doesn't mean that simulations (along with experiments and more traditional theory) haven't done so in the past. EDIT: and it doesn't mean they won't continue to do so. Much of science is serendipitous. You are focused on investigating one problem, and you discover something else entirely that is new and interesting.

 

[cepheid]

Regarding gravity: without simulations to predict what systems that generate gravitational waves will do, we can't conduct experiments to look for them in real life. And without such a detection, we're kind of missing a key confirmation of the best theory of gravity we've come up with to date. If we don't see any gravitational waves at all, even at the sensitivities that theory predicts we ought to be able to, then it will be clear, I think, that general relativity is incomplete in some way. Granted, general relativity has held up spectacularly well against experimental testing so far. But stating that we fully understand everything about gravity and are simply applying that knowledge to more complex systems is quite an overstatement in my opinion. GR should continue to be subjected to experimental test. The fact that GR is inconsistent with quantum mechanics and that we have no quantum theory of gravity yet is argument alone that something is missing.

 

[cepheid]

But I don't understand, as an intellectual race, why isn't our primary goal to solve these mysteries. WHY aren't we focussing all our faculties and resources to this one goal of using physics, math and biology to explain our existence. I mean I am not smart enough but I would die happy thinking I made some effort. Is it because of our capitalistic society?

Or do you just think I have been watching Brian Greene, Through the wormhole w/ Freeman, etc too often..

I don't understand what you are saying here. As many of my posts have pointed out, we ARE trying to understand the nature of the universe, galaxies, stars, fundamental particles and everything in between. Simulations play an essential role in that quest. That is what we have been trying to explain numerous times in this thread.

 

[absurdist]

Got it cepheid . Thanks for dumbing it down!

I was digressing regarding AlephZero's point about not solving 'great mysteries'
by trying to imply that if there were more of us who would become physicists, it would be likely that someone would crack one of those discrepancies or unsolved mysteries.
But I understand now how computational physics fits into the big picture of pure physics.

The kind of physics I have learned is mostly classical and I suppose most of the theories and laws there are carved in stone. THIS is why it was difficult to see how simulations could have possibly helped elucidate something new.

 

[luben]

Absurdist, having the laws and theories carved in stone doesn't mean you cannot discover new physical phenomena from these laws. if you ever try some simulations, for example for some nonlinear problems, you can easily get very rich phenomena under different dimensionless parameters regimes. Those phenomena can easily become new discovery, since they are very sensitive to the conditions you imposed to the problem. Discovering the true nature of these new phenomena has great values in understanding the world, too. So you get what I mean? This is one thing. Computation helps by giving the whole picture of the solution which experiments usually incapable of measuring/recording, and the results are in absolute details which experiments cannot rival sometimes. It is more convenient to look into the solution and to dig up the nature of any new phenomena. This is why computational physics is indispensable for modern researchers in many fields.

 

[absurdist]

Point taken Luben!

Does this mean if you are an engineer who is great at computations and simulations, you could easily apply the same principles to work being conducted in science or is there a large difference between the kind of simulations and modelling done in science versus engineering?

Also stuff like Monte Carlo Simulations and CFD, there's a very unclear line between physics and engineering, isn't it?