EBQM Research Log 3

Diary of a Marie Skłodowska-Curie Research Fellow

Been doing a lot of coding this month…!

Diary of a Marie Skłodowska-Curie Research Fellow: Entry #3

Hello, and welcome to the third post in this series of Research Logs for the Marie Skłodowska-Curie project ‘Ergodicity Breaking in Quantum Matter’ (ebqm.info). You can find the first post here.

The first month of 2022 has been a frantic one, full of incredibly exciting developments, and a lot of time this month was spent doing admin tasks and laying the groundwork for future projects while trying to wrap up some of the current ones. My main project has been somewhat on hold this month as I focused on improving the codebase I’ve been using so that it’s more scalable for the future planned projects, and will hopefully soon see the light of day as a publicly available Python package on GitHub or GitLab. As part of this, I’ve found several significant improvements in speed, memory usage and accuracy, and in fact the improvements are so great that I plan to update the results in my latest paper using the method (still currently under review!), as while the qualitative findings are unchanged, the improved method is much more accurate. (On a semi-related note, I also spent a bit of time this month finally preparing the Data Management Plan for the EBQM project, as this is a contractual requirement that I couldn’t do until I’d made a bit more progress with the codebase and knew roughly what type and quantity of data needed to be managed.)

On the non-research side, this also marks the first month of my “View from the arXiv” series of weekly posts highlighting newly released papers on the arXiv preprint repository, as well as the relaunch of regular (fortnightly) posts on my non-academic popular-science blog Broken Symmetry, alternating with newly-launched fortnightly blog posts here on my academic homepage. So, every week I put out a “View from the arXiv” post, as well as one post on either the pop-sci blog or a more technical post here on the new academic blog. I’m really enjoying having a writing schedule again, and so far the response has been very positive. The next step here is to try to grow the audience of both sites more by keeping up a regular schedule of quality content, and hopefully use this as a springboard to resume writing for other outlets as well, which I’ve not done now for a few years.

I’m also extremely excited to be giving a (virtual) theory discussion seminar at my alma mater, the University of St Andrews, in a few weeks. I’m planning to discuss many-body localization in quasiperiodic systems, but also want to take this opportunity to show off something new and a little bit less disordered than my usual fare, which I think might appeal to some people in St Andrews. Watch this space for more…!

With several new projects spinning up, the next main step for me is to try to wrap up the two current flow equation projects that I’m actively working on. I think I’ve fallen a little behind, as compared to the initially proposed schedule for the project, but the main reason for this is that opportunities have arisen for me to collaborate with colleagues on projects closely related to aspects of the project planned for later in the year, so essentially it made sense to move up the schedule for these projects while delaying slightly the completion of the early ones. Now that groundwork is in place for these projects, however, it’s time to tie up the first ones, so hopefully that can be done in the month ahead. I should have at least one paper coming out in the next month, as colleagues in Paris have sent me some exciting results on a project relating to Sachdev-Ye-Kitaev models, and in fact things are going so well that I think this should be out soon.

This month I made a grant application to the NVIDIA Academic Hardware Grant Program for two powerful GPUs on which to continue to develop the tensor flow equation method which I’ve been working with (which I’m toying with renaming ‘tensor bracket flow’ to distinguish it from the Google TensorFlow machine learning codebase). Though a very brief application in comparison to a lot of other schemes, this still took quite a bit of time this month, and in particular I spent a lot of time checking tech specs and discussing with our IT department whether our existing infrastructure was suitable for the GPUs I have applied for. Massive thanks to our IT staff at Freie Universität for their patience and help, and to François Rincon and Srikanth Sugavanam for feedback on the draft application. If successful, this application will represent a massive increase in computational power and enable us to do some really incredible things in the future, so keep your fingers crossed for me!

I also made another application to a Secret Exciting Thing and have another in the works for a Second Secret Exciting Thing, but I can’t say anything more about those at the moment - I’ll be sure to let you know if they’re successful. ;)

The biggest development of the month has undoubtedly been the arrival of my new M1 MacBook Pro, which I ordered in November but due to delays with the university’s supplier it only just arrived in the last week. Since starting my first postdoc in 2016, my daily driver has been a 2013 MacBook Air, which I’ve used to write every line of code in every project I’ve been involved with for the last 5+ years, as well as every paper relating to those projects, dozens of conference presentations and international trips (remember when travel was a thing…?), and even five successful NaNoWriMo attempts. It has served me well, but the lack of power has been apparent for quite some time, and as I moved more and more into high performance numerical work, it quickly became the bottleneck in my ability to make progress. Until now I’ve never had access to any funding to replace it, so I’m extremely excited to see what this much more powerful computer can do. The M1 chip is not yet directly compatible with everything I use, and in particular I had a little trouble with the (otherwise excellent) QuSpin exact diagonalization library, but I eventually got it to run by building it locally (with a few compiler tweaks) and so far I think it’s working just fine.

So - over the next month the plan is to finalise the Tensor (Bracket?) Flow Equation code, perhaps publicly release it on GitHub, give a talk on the method, hopefully update my previous paper using the method, prepare two new papers using the method, release a new paper on SYK physics, and keep the ball rolling on a bunch of exciting new projects currently getting underway. It’s going to be another intense month, but I’m looking forward to it. See you back here next month for another update on how it went!

Dr Steven J. Thomson
Dr Steven J. Thomson
Marie Skłodowska-Curie Research Fellow in Theoretical Condensed Matter Physics

Theoretical condensed matter physicist, currently a Marie Skłodowska-Curie Research Fellow at Freie Universität Berlin.