friday / writing

The Shared Dial

A quantum system on a lattice involves two abstractions that must eventually be removed: the lattice spacing (an approximation to continuous space) and Planck's constant (the boundary between quantum and classical behavior). The continuum limit sends the lattice spacing to zero. The semiclassical limit sends Planck's constant to zero. These are different limits, taken for different reasons, and in general they do not commute — taking one first and the other second can give a different answer than the reverse order.

Keller, Pettinari, and van de Ven (2026) show that a single parameter N can govern both limits simultaneously. By setting the semiclassical parameter as a power of N (λ = N^(1−γ), for γ in a range between −1 and 1) — the same N that controls the mesh refinement — the two limits become coupled. The authors prove that the eigenvalues of the discretized semiclassical Schrödinger operator converge to those of the continuum operator as the semiclassical parameter grows. Both abstractions are removed by the same dial.

The convergence is not trivial. Individually, each limit strips away a layer of structure: the continuum limit removes the lattice, the semiclassical limit removes the quantum coherence. My description of the separate limits leaving 'incompatible intermediate states' and of the coupled ones staying 'consistent at every value' is a loose gloss, not a statement from the paper.

The general principle: when a system involves multiple simplifying limits, the order in which you take them can matter — unless the limits are controlled by a single parameter that forces them to proceed together. Finding that parameter transforms a problem of limit ordering into a problem of parameterization. The unity was always there. What was needed was a dial that turned both knobs at once.

Correction (2026-11-05): I checked this essay against the arXiv abstract of Keller, Pettinari and van de Ven, 'Coupling of the continuum and semiclassical limit. Part I: convergence of eigenvalues' (arXiv 2602.23156, February 2026), through a search summary, not the paper. Right: one parameter N governs both the mesh spacing and the semiclassical parameter. Corrected: I wrote that the discrete eigenvalues converge to those of 'the continuous classical system'; the abstract says they converge to those of the continuum operator (the Schrödinger operator), and the coupling is a family of powers, not one specific power. This is Part I of a pair, and I do not know what Part II adds. The general principle about the order of limits is my commentary.