Interlude: the Boltzmann Factor as a Character

For the reader from (either camp).

This chapter has no new computations. It names the single idea that the rest of the notes re-spell in five notations. If you read only one chapter slowly, read this one.

We now have both incarnations of the Boltzmann factor on the table: the number $\mathrm{e}^{-\beta E_{i}}$ of Lecture 1 and the operator $\mathrm{e}^{-\beta H}$ of Lecture 2. Before the object fragments into the notations of field theory, conformal field theory, and representation theory, let us abstract it. This is the piece that most often falls through the cracks: one learns each incarnation locally and never the idea behind all of them.

I.1 Three faces of one object

Face 1: the max-entropy weight.

Maximizing entropy subject to fixed mean values of conserved charges $Q_{a}$ forces the weight

w=\exp\Big{(}-\sum_{a}\beta_{a}Q_{a}\Big{)},

(I.1)

the $\beta_{a}$ being the Lagrange multipliers (Lecture 1, generalized). Nothing else extremizes entropy at fixed charges. This is why the exponential is not a modelling choice but a theorem: the Boltzmann factor is the max-entropy weight.

Face 2: the character (a homomorphism).

The map

Q\longmapsto\mathrm{e}^{-\beta Q}

(I.2)

is a continuous homomorphism from the additive monoid of conserved charges to the multiplicative monoid of weights-and, up to the choice of the constant $\beta$ , the only one:

\mathrm{e}^{-\beta(Q+Q^{\prime})}=\mathrm{e}^{-\beta Q}\,\mathrm{e}^{-\beta Q^% {\prime}}.

(I.3)

Additivity of charges becomes multiplicativity of weights. This single fact is why independent systems’ partition functions multiply (§4), why $Z$ factorizes over modes, and why product formulas- $\prod(1-q^{n})^{-1}$ , then $\eta$ , then $\theta$ -pervade the subject. A one-dimensional character is the embryo of every “character” in Lectures 4 and 5.

Face 3: Euclidean evolution.

Real-time evolution is the unitary $\mathrm{e}^{-\mathrm{i}Ht}$ . Under the Wick rotation $t\to-\mathrm{i}\beta$ its oscillating phase becomes the real, decaying weight $\mathrm{e}^{-\beta H}$ : through this analytic continuation inverse temperature is identified with an interval of imaginary (Euclidean) time, and $\beta$ measures its extent-a duration, not a temperature. Taking the trace glues the two ends of that interval into a circle of circumference $\beta$ (L3). In a two-dimensional CFT the space is itself a circle-the theory on a cylinder-so there are already two periodic directions, space and Euclidean time, and closing the time circle by the trace turns the cylinder into a torus (L4).

Remark.

These are not three analogies; they are three theorems about the same function. Face 1 says which function; Face 2 says how it composes; Face 3 says what geometry it secretly carries.

I.2 The general Boltzmann factor

Collect every commuting conserved charge-energy $H$ , particle number $N$ , momentum, an internal or flavour charge $J_{a}$ -and pair each with a conjugate potential:

\boxed{\;w=\exp\Big{(}-\beta H+\beta\mu N+\textstyle\sum_{a}\nu_{a}J_{a}+% \cdots\Big{)},\qquad Z=\operatorname{Tr}\,w.\;}

(I.4)

Two structural facts organize (I.4):

•

Extensive charge $\leftrightarrow$ intensive potential is a Legendre duality: the pairs $(E,\beta)$ , $(N,\mu)$ , $(J_{a},\nu_{a})$ . Differentiating $\log Z$ by a potential returns the mean of its charge; the second derivative returns the fluctuations.
•

The chemical potential is the prototype “extra knob.” It weights states by how much conserved charge they carry; equivalently a fugacity $z=\mathrm{e}^{\beta\mu}$ grades the trace. Refining the trace is turning a knob.

To see a knob actually move the grading, watch spin. A conserved angular-momentum projection $J$ has quantized eigenvalues: whole units $\dots,-1,0,1,\dots$ on integer-spin (bosonic) states, half-units $\dots,-\tfrac{3}{2},-\tfrac{1}{2},\tfrac{1}{2},\dots$ on half-integer-spin (fermionic) ones. Its conjugate potential is a rotation angle $\theta$ , which enters the trace as a holonomy-a twist applied once around the Euclidean time circle of Face 3-so that a state of spin $j$ picks up the phase $\mathrm{e}^{\mathrm{i}\theta j}$ on the way round:

\operatorname{Tr}\,\mathrm{e}^{\mathrm{i}\theta J}\,\mathrm{e}^{-\beta H}\;=\;% \sum_{j}\mathrm{e}^{\mathrm{i}\theta j}\,\operatorname{Tr}_{\mathcal{H}_{j}}% \mathrm{e}^{-\beta H},

one term for each spin sector $\mathcal{H}_{j}$ . Although $\theta$ runs continuously, nothing depends on it except through that holonomy-and what the holonomy detects is the quantization of $j$ . At the trivial twist $\theta=0$ every sector is counted equally; at a full turn $\theta=2\pi$ the phase becomes $\mathrm{e}^{2\pi\mathrm{i}J}=(-1)^{F}$ , the familiar sign a fermion picks up under a $2\pi$ rotation-the coarse $(-1)^{F}$ grading of L3–L4 in embryo.

Particle number and energy grade the trace the same way; only the spectrum being sorted changes. Particle number is likewise quantized, so its fugacity $z=\mathrm{e}^{\beta\mu}$ organizes the trace into the grand-canonical series $\sum_{N}z^{N}Z_{N}$ , and raising $\mu$ makes each added particle cheaper, tilting the sum toward larger $N$ . Energy is not quantized in integer units: $\beta$ weights each state by its actual energy through the smooth envelope $\mathrm{e}^{-\beta E}$ , sliding the weight rather than sorting it-raising $\beta$ (cooling) suppresses high-energy states toward the ground state, lowering it spreads the weight across the whole range. In every case the exponent is a charge and the knob its conjugate potential; whether turning it sorts the trace into discrete sectors or slides a continuous weight is fixed by the charge, not the knob.

An aside on scale. Because $\beta$ is the circumference of the Euclidean time circle (Face 3), it is also a length; dialing it from small to large carries the theory from short-distance toward long-distance physics. Turning the $\beta$ knob is, in this sense, moving along a renormalization-group scale-a connection we flag here and leave to a course on critical phenomena.

I.3 The same knobs, renamed, downstream

Here is the cross-disciplinary payoff to keep visible for the rest of the notes. Each conjugate potential we meet now reappears later wearing a geometric or formal name.

Conjugate potential	in L1–L2 it is	downstream it becomes
$\beta$ (to energy $H$ )	inverse temperature	$\operatorname{Im}\tau$ on the torus (L4); the modular parameter of a character (L5)
$\mu$ (to charge $N$ )	chemical potential / fugacity $z$	the elliptic variable $y$ conjugate to $J_{0}$ (L4); Cartan variables of Weyl–Kac (L5); the $y$ of the elliptic genus
$\theta$ (to fermion number $F$ )	sign grading: $(-1)^{F}$ at $\theta=2\pi$	a spin structure (L3–L4); the Witten index / elliptic genus (L5)
source (to a current)	external field	a background gauge field / Wilson line (L3–L4)

For the curious (a categorical wink).

Face 2 says a Boltzmann factor is a monoidal functor $(\text{charges},+,0)\to(\mathbb{C},\times,1)$ . That is the one-object baby version of the bordism functor $Z\colon\operatorname{Bord}_{2}\to\operatorname{Vect}$ of the finale: assign weights compatibly with a composition law. From this height, “partition function $=$ trace $=$ character” is the single statement that the assignment is functorial.

A checklist for everything that follows

Whenever a later chapter writes a new $\mathrm{e}^{-(\cdots)}$ or $q^{(\cdots)}$ , ask three questions:

1.

Which charges sit in the exponent?
2.

What are their conjugate potentials?
3.

What geometric object closes the trace?

Those three answers are the entire dictionary we are about to assemble.