Is "compute one quantity two ways" a single theorem or several?

Run status: not attempted — no run in lab/runs/ names this case in its settles: field.

type: derivation (seat-only; no compute step expected) field: math

---

Object

Four results in unrelated areas share a form: a single quantity is evaluated by two different procedures, and setting the two evaluations equal is the content.

1. The trace of an operator. For a suitable linear operator on a function space, the sum of its eigenvalues equals the integral of its kernel along the diagonal:

> Σₙ λₙ = ∫ K(x,x) dx

The left side is spectral — it needs the eigenvalues. The right side is geometric — it needs only the kernel near the diagonal. Equating them turns a question about a spectrum into a question about a geometry, and this is the engine behind the Selberg trace formula, the Poisson summation formula, and index theorems.

2. Little's Law in queueing. For any queue in steady state, the average number of customers present equals the arrival rate times the average time each spends:

> L = λW

The proof counts the area under the occupancy curve twice: reading along the time axis gives L·T; reading customer by customer gives (λT)·W. It requires no assumption about arrival distribution, service distribution, service discipline, or number of servers — verified numerically to machine precision for Poisson, deterministic, and heavy-tailed arrivals alike.

3. Stokes' theorem. The integral of a differential form over a boundary equals the integral of its derivative over the interior:

> ∮_∂S ω = ∬_S dω

One quantity, computed on the edge or computed inside. Its instances include the relation between a holonomy around a closed loop and the enclosed area — the work done around a thermodynamic cycle, the phase acquired around a loop in phase space, the rotation accumulated by transporting a vector around a circuit.

4. The replica trick. An entropy is obtained by computing tr(ρⁿ) for integer n, where the calculation is tractable, and then analytically continuing in n to where it is wanted. The quantity is defined one way and evaluated another, with the continuation carrying the whole argument.

---

Friction

These are routinely grouped, and it is not clear what the grouping asserts.

The strong reading is that they are one theorem in four costumes — that there is a single underlying principle, "a quantity computed two ways must agree," of which each is an instance. On this reading, discovering the principle in one field should transfer technique to the others.

The weak reading is that "computed two ways" describes a proof strategy rather than a theorem, and the four are unrelated results that happen to be proved the same way — in which case grouping them explains nothing and transfers nothing.

The friction is that the four have visibly different requirements. Little's Law needs essentially no hypotheses. The trace identity needs the operator to be trace class, and fails — divergently — when it is not. Stokes' theorem needs a differential structure and an orientable domain. The replica trick needs an analytic continuation to exist and be unique, which is not automatic and has been the sticking point in its most prominent recent applications.

A single principle with four wildly different hypothesis sets is either not a single principle, or a single principle whose hypotheses are doing all the work.

---

Background (every fact needed, inline)

• Trace class means the sum of the singular values converges. An operator that merely preserves a measure — a composition operator induced by a measure-preserving map — is an isometry, has all singular values equal to 1, and is never trace class in infinite dimensions. The trace identity then has nothing finite on either side.
• Little's Law's proof is a double count of the same region of a (time × customer) plane. Swapping the order of summation over an indicator function is the whole argument.
• Stokes' theorem in its modern form subsumes the fundamental theorem of calculus, Green's theorem, the divergence theorem and the classical curl theorem. Its proof is local — verify on a cube, patch with a partition of unity.
• The replica trick's continuation is not determined by the integer values alone; infinitely many functions agree at the integers. Carlson's theorem supplies uniqueness under a growth condition, which must be checked and often cannot be.
• Two of the four have a differential structure (Stokes, and the trace identity in its heat-kernel form). Two do not (Little's Law, the replica trick).
• Two are exact identities with no limit taken (Little's Law, Stokes). Two require a limit or a continuation (the trace identity in infinite dimensions, the replica trick).

---

What a satisfactory answer must predict

An identity verdict, plus a discriminator. Either:

• one theorem — and then: what is it, stated so that all four are corollaries, and what single hypothesis specialises to trace-class, to orientability, and to Carlson's growth condition; or
• several — and then: what partition, and what feature separates the classes. A satisfactory partition must predict which side of the line a fifth instance falls on before it is examined.

A test case for any proposed partition: the Cauchy integral formula, which evaluates a function at an interior point by an integral over a surrounding contour. One quantity, two evaluations, a differential structure present, no limit required.

---

Provenance

None — this is a structural question about four stated results, not a claim about data.