← Eyal Merhavi

Labelings, models, and gauge theory

A computational course for CS and math undergraduates.

The course starts from an abstract framework: random labelings of a graph by a group. It derives "quantum dynamics" as a consequence, rather than assuming it, and only then turns to physics, showing that lattice Yang–Mills is one model in the framework. Every idea is checked numerically on an example small enough to enumerate exactly.

Each chapter is a Jupyter notebook, shown here as a static page with its code, outputs and figures, and with the answers to the exercises folded away.

Run the notebooks in your browser ↗ · Code on GitHub ↗
The notebooks run on JupyterLite, starting with chapter 0. The first time, it downloads Python and its libraries (about 40 MB).

Four samples of a 10 by 10 grid with labels plus or minus one, at stiffness 0, 0.5, 1 and 2. The number of defect faces, shaded coral, drops from 49 to 23 to 12 to 3, while the gold edges look equally random in all four.
From chapter 2: samples of the coin model at four stiffnesses. The coral defect faces thin out as the stiffness grows; the gold edges look like the same noise every time.

Course map

All chapters are drafts.

I. The world of labelings

II. Dynamics from the distribution

Bridge

III. What can be proved

A dense, symmetric fractal pattern of dots shaped like a butterfly: the smoothness spectrum of a charged field on a 24 by 24 torus, plotted against the angle of the defect carried by every face.
From the interlude: give every face of a 24×24 torus the same defect and plot the possible smoothness costs of a charged field against the defect's angle. The picture is Hofstadter's butterfly.