<p>The duty of both the computational and experimental physicist is to verify whats proposed by the theoretical physicsist, right? So do they play the same role?
But to create a model, the computational physicist needs experimental data to begin with or physical laws right, do these physical laws come from that theory? They must come from some fundamental physical laws?</p>
<p>I am not a computational scientist, but I work with many who are and what I have picked up includes …</p>
<br>
<br>
<p>a) Computation allows you to “observe” details that are difficult or impossible to observe in empirical experiments.</p>
<p>b) Computation allows you to run experiments that you are unable (or unwilling) to do empirically, perhaps because the experimental facility does not exist, because you do not have time on the experimental device, or because you want to run a series of experiments that are much too costly or time consuming to do empirically. One example is investigating theories of supernovae mechanisms. These are “difficult” to run experimental studies on. Another is the design of new materials, where the search space is very large and using only empirical experimentation would be very expensive.</p>
<p>c) Computation is also used to design experiments (eliminating ones that would be uninteresting, or perhaps even dangerous to run). For example, for the ITER facility, the device can handle only a certain number of experiments in which the plasma escapes the magnetic “walls” and hits the physical wall before having to be shut down or refurbished. The intent is to use computation to identify (and avoid) experimental settings that might cause this to happen, and so extend the lifespan of the device.</p>
<p>d) Computation also allows theories to be investigated without ever building an experimental facility, providing a form of triage by determining the logical consequences of the theory (identifying whether it matches existing experimental data).</p>
<br>
<br>
<p>e) Depends on the model and on the questions being asked. The physical laws come from the theorist (that’s what they are for 
). Some models are “first principal” and can be examined with little experimental data. Other models are incomplete, with ‘subprocesses’ represented solely by data (or data models). Computation is also not infallible, and people worry about validation and verification (V&V) and about uncertainty quantification (of both the underlying theoretical model and of the numerical approximations used in the computational model).</p>
<p>I’ll let you check out wikipedia for validation, verification, and uncertainty quantification. I would not do justice to the topics (and I always forget which is validation and which is verification anyway).</p>
<p>Sorry kind of zoned out but thanks that’s an excellent explanation of its uses!</p>
<p>BTW - sorry for the sloppy use of terminology. ‘empirical’ refers equally well to computational experiments (numerical simulations) as it does to noncomputational experiments (‘physical experiments’?). Similarly ‘experiments’ can be computational or physical. Hopefully my incorrect usage was still clear in context. As I said - I am not a computational scientist, and am not sure how things are normally labelled so as to avoid confusion. In biology they use the terminology ‘wet lab’ and ‘dry lab’, where dry lab refers to computational or analytic modeling . I haven’t heard similar terminology used by physicists. Probably just haven’t been paying attention.</p>