Lab notebook on Sibasankar Panigrahy

Introducing WindowWise: plan, check and repair umbrella sampling

Fri, 09 Oct 2026 10:30:00 +0530

During my PhD I ran a very large number of umbrella-sampling simulations. The two mistakes I made most often were spacing the windows too far apart and using springs too soft for a steep part of the free-energy profile. Both fail quietly: WHAM still returns a smooth-looking curve, and the problem only shows up later, if at all.

Teaching attention to read a surface

Fri, 09 Oct 2026 07:00:00 +0530

Umbrella sampling is the workhorse for computing how strongly a polymer sticks to a surface. You pull the chain away in windows, sample each one, and stitch them together into a potential of mean force (PMF). It is reliable, and it is slow. If you want to screen polymer sequences against surface patterns, the combinatorics quickly outrun any compute budget.

Small crowders are not tiny PEG

Fri, 09 Oct 2026 06:30:00 +0530

Ask someone how a crowded environment affects a nanoparticle and you’ll usually hear about depletion. Big inert crowders such as PEG or dextran are excluded from a shell around the particle. That creates an osmotic push, and things compact or aggregate. It’s an elegant picture, the Asakura–Oosawa picture, and for large polymer crowders it works well.

Opening the lab notebook

Fri, 09 Oct 2026 06:00:00 +0530

For six years most of my thinking lived in places no one else could see: slide decks, half-finished Jupyter notebooks, and margins of printed papers. This notebook is an attempt to move some of it into the open.

What to expect

Short daily notes. A plot that surprised me, a parameter that mattered, a paper worth reading. Some are three paragraphs long.
Explainers. Longer pieces on methods I use all the time: Kirkwood–Buff theory, umbrella sampling, polarizable metal models, attention-based networks for materials.
Behind the papers. The reasoning and dead ends that don’t fit in a methods section.
Tool notes. Release notes and recipes for the software I build.

Why bother

Two reasons. Writing is the fastest way I know to find out whether I actually understand something. And most of what makes simulation work, like which force field to trust, how long to equilibrate, or why a free-energy profile looks wrong, is rarely written down anywhere useful.