What is it that you actually do?
That’s a question that a lot people ask when they hear that I’m an astrophysicist. There’s a short answer and a long answer to this question. The short answer is that I research things. It’s the same as every other scientist. We have some question(s) and we work to find the answer to them. Currently I’m working on ways of simulating individual galaxies to answer questions related to dark matter halos and galaxy mergers. These are related to larger questions of how galaxies in the Universe evolve. It’s a lot of fun and a lot of work.
But that’s not a particularly informative answer. To make it informative I would need to write a full post about this work (which I may do at some point) to give the full context and more details about the work. Beyond that, the other issue with just telling you what I research is that it’s often not answering the question being asked. A lot of the time, the question ‘what do you actually do’ is often more about the day-to-day job of an astrophysicist (or scientist in general) than their topic of study. It’s less about what someone is researching and more about what is involved in research.
In this post I’m focusing on research because that’s the main job of most astrophysicists. Whether it’s at a university, observatory, or elsewhere, most of us are hired to conduct research. To be fair, some are hired for teaching or outreach or operations, but these jobs are significantly more rare than research positions. Of course, most everyone does end up doing some teaching and some outreach and various other interesting duties, but this isn’t the primary focus of their job. It may come as a surprise to many students, but, at most universities, teaching requirements are a service that someone does in order to continue being privileged enough to research. For a post-doc like myself, this is especially true as my contract is for something like 95% research and 5% services to the department. I have done some teaching (which I really enjoyed) and I have a number of roles within the department, but my main job is research. That’s what fills up my days at the office.
In general, there are perhaps four components to research; the question being investigated, the data needed for that question, the tools used to obtain and analyze the data, and the actual analysis of the data.
In astrophysics the questions usually are somewhat narrow, but they relate to larger, unanswered questions. I use the simulations that I mentioned earlier to answer very narrow questions about things like how bars form or how to quantify how asymmetric a galaxy is. But those questions come from larger questions about how galaxies evolve over time, what happens when different galaxies crash into each other, and what that tells us about the nature of the Universe as a whole!
Next, the data. It’s difficult to generalize about astronomical data because astronomical data is incredibly varied. In terms of observations, the data might take the form of images of some patch of sky, the spectrum of light from some galaxy, or the light curve of a particular star (light curves are measurements of the brightness of something as a function of time), or something else entirely. It might need to be in multiple types of light as well. For instance an observer studying supermassive black holes may need to look at each object in radio, optical, and X-ray.
Obtaining the data that an observer needs can be very complicated. They may need to apply for time to make new observations at different observatories. They may need to comb through old data (particularly from large surveys like WISE or SDSS). In these cases the observer might be able to apply to an existing observatory. In some cases, there might not be an instrument capable of producing the data that they need for their research. In that case, they may propose some new instrument and get involved with the design and construction of it.
Regardless of what they are studying, observers are rarely actually observing. They are figuring out what data they need, seeing whether that data exists, proposing new observations, and cleaning and analyzing the data that they have. In reality, data is often messy with all sorts of background and foreground things that must be accounted for. There can be contamination from nearby objects, instrumental effects, and the sky background that need to be removed. In radio, there is almost always some interference that must be accounted for. And after doing all that, it is usually necessary to produce ‘data products’. These include images, spectra (when the light is broken up into wavelength bins), SEDs (spectral energy distributions across different types of light), light curves, and more.
On the other hand there are astrophysicists who fit more on the theory side of the spectrum. For instance, I currently work mostly on tailored numerical simulations of individual galaxies or pairs of interacting galaxies. This involves generating a table of particles that represent the initial galaxy or galaxies and then loading those into a code to simulate their evolution under gravity. After the simulation is done, I need to analyze it what has happened to the system. While I do use a fairly standard tool, Gadget 2, for the simulation, I’ve had to develop new tools for generating the starting points for these simulations and to analyze the outputs. This is true for most people working on numerical simulations at all scales. Our data largely consists of the outputs from some numerical simulation package. In my case, it is tables of particle positions and velocities. I then usually need to change these into some sort of data product as well. Sometimes it’s mock images (or even movies), or velocity profiles, or something else entirely.
Both observers and theorists require tools to turn raw data into usable data products. And there are additional tools that are need to analyze those data products. Often times the tools that worked in the past don’t quite fit the current problem being researched. Observational data might be in some new format, the question might be slightly different, the sensitivity has changed, the simulation software is different, etc. As such, astrophysicists will need to modify those old tools, or develop new tools altogether. Astronomical toolkits tend to be much more ‘tailor-made’ rather than industrially produced. The tools are usually designed for the particular problem being investigated.
That being said, there are some relatively standard tools. For example Gadget 2 is a fairly standard tool for numerical simulations. I currently use it as is, but many other researchers have significantly modified this software for their research. They might need to include the effects of star formation or feedback from supermassive black holes so they modified the code to include recipes for those effects. The point is that, as I said in the previous paragraph, even standard tools often need to be modified for their research projects.
Both tool development and tool modification take a lot of time. As does data reduction and cleaning. As such, a typical day at the office for an astrophysicist is usually spent in front of a computer. They are finding the data that they need, devising plans on obtaining (or producing in the case of theorists) new data, writing proposals for access to instruments or computing facilities, developing new tools for analyzing their data, and actually doing that analysis.
All of this is hard work, but it’s quite rewarding. It requires a certain degree of creativity because of all the new things that need to be developed. It also requires a large knowledge base of what has been done so that the results from some project can be understood and placed into context. But more than anything else, it requires stubbornness. In my opinion, stubbornness (or better yet, persistence), is the number one characteristic of a scientist. It’s necessary because, as any scientist will tell you, research doesn’t usually go the way that it was initially planned. The data might be flawed, the tools might not work properly, or the method of analysis chosen just might not work. There are always roadblocks and complications that have to be solved. And where the research ends may be incredibly different than where it began!
So what is it I actually do? I do research. I find an interesting* question and then I try to answer it. I figure out what sort of data I need, be it a simulation or some observational data that I want to model. I sit in front of a computer and I develop the tools I need to analyze the data. I use these tools and then try to figure out what I’m learning from a particular analysis. I troubleshoot all the problems and bugs that show up, and try to figure out if what I’m getting is ‘real’** or from some mistake I made somewhere. And I keep repeating these steps over and over until I learn something new that relates to the initial question. It’s hard work, but it’s also very rewarding…and pretty fun!