Because in many cases you want to have quantiative explanations or predictions for whatever you are doing. The moment that you have a model that provides a good agreement with experiments, you know that you have a firm handle on the problem at hand. The moment that a model stops agreeing with your experimental results, you have a new problem to solve.

Not all problems require models, though. If your goal is to synthesize a material and you want to check its composition, a good experimental design and understanding of the driving forces controlling your growth is most likely what you need. In those cases, you need a good mental model rather than a mathematical model. Still, data analytics and machine learning could be the right thing for you. When I talk about models and simulations here, I am refering to models mainly based on physics/chemistry and math.

What do I need to get going?

  1. Going back to the fundamentals. You don’t need a supercomputer or a fancy simulation to get going. In many cases, going back to the basics, to essentially textbook-level equations can get you a long way. A lot of phenomena in physics/engineering/materials science are formulated using simple models or things like partial differential equations. Going back to the basics can get you a long way. I still do it, pretty much all the time.

  2. Using simulation tools. Some problems cannot be solved using simple equations that have a nice closed or analytic solution. In that case, you probably need to use computational methods to set up your problem and get your results. You are in luck though: we live in a golden era where there are plenty of tools available that you can learn how to use to solve your questions without having to code them from scratch. Many of them are open source, which means that they are free to use. Good computing literacy is the most useful skill.

  3. Coding your own simulation tools. You don’t have a ready-made simulation to solve your problem. No biggie: there are tons of programming tools and libraries that can get you a long way. Some of them are proprietary tools that you can probably access at no cost through your institution or company. Think Matlab or Mathematica. Many of them, though, are open source, which means that you can reuse them and adapt them to your own problems. This is probably the hardest path out there, but also the most powerful. Learning a useful programming language is a must.

If you are really serious about simulations you should learn a programming language

Regardless of whether you can get away with Excel or pen and pencil, I cannot emphasize enough how important it is for a scientist/engineer to learn how to code. And you are really primed for it: one of the hardest aspect of learning how to code is to find a motivation to do it. Otherwise it often becomes an academic exercise that is likely to fail unless you have a strong driving force (as we say in Spanish, the graveyard is full of good intentions).

In your case, you already have hundreds of interesting problems that you can use as motivation to learn: reading files and automatically plotting the data, building a simple simulation that you can validate with another tool or an analytical solution, learning how to process data so that you don’t have to open a file, copy two columns, remove the headers and save it in a different file manually. There are infinite possibilities.

So the obvious next question is: which programming language? The answer is: whatever your community is using. And my default answer would be: Python. My own trajectory was (starting in middle school way back then) Basic -> C -> C++ -> Fortran -> Python -> Julia -> Rust, with few exotic detours such as Scheme, Java/Jython, Clojure, and Haskell. And I basically followed the needs/opportunities at the time: I had a computer that did Basic, then someone lent me a book of C, my college offered a few credits intro to C++, my group and collaborators worked with Fortran, another postdoc pointed me to Python, and the rest were the result of my own explorations, interests, and needs at work.

Models and simulations are a great scientific equalizer

One final point: not everybody has access to state of the art experimental equipment or funds to run an experimental lab with the tools you need to make a research impact. However, in many fields (not all of them, though) you can make an impact in your research community by answering the right question with the right tool. And while it is true that few of us have access to the computational resources required to run large simulations, a model that is simple enough to run on a budget laptop and that gets you 80% there has its value.

Models and simple simulations can therefore be an equalizer in scientific research, if you find a relevant problem that can be solved or at least partially solved with models you can put together. It doesn’t come for free though: in order to get there, you will have to invest your time and brain power to get the right skills, the background you need, and to identify the right challenge. But at least there is a path.