Expectations

It’s important to always bear in mind that there is no readily applicable, exact methodology that can be applied to understand atomic structure from total scattering data. Reproducing the intricacies of even simple real-world systems is challenging, and the more complex the real-world system being measured is, the more prone to variance and error the actual samples are. Moreover, the samples you measure on a beamline are absolutely enormous by simulation standards, and are evolving on their own dynamic timescales (even if it is just standard vibrational entropy).

We must all accept that there are many sources of imperfection in what we’re trying to do:

Simulations are not perfect

They are just that - a simulation. They cannot possibly fully reproduce the size, timescales, and complexity of your actual samples, But they can do a pretty good job of approximating them.

Samples are not perfect

Purity, isotope substitution level, and mix ratio, for instance, can all be controlled by a careful scientist to a high degree of accuracy, but are never absolutely 100% accurate.

Experiments are not perfect

The instruments measuring your samples are well-designed, no doubt, but have all been built with some compromises in mind. In addition, the data reduction process use to generate your structure factors also make some approximations along the way, and so there is no “absolute” output from such processes either.

If one takes into account all of these factors it might seem almost impossible to make this kind of experimental technique and analysis work at all, but that is not the message here! The point is that due care must be taken at every stage in order to best quantify the limits of your knowledge about the system, to do the most appropriate and representative simulation that you can, and to accept that perfect agreement between simulation and experiment will likely never occur.