Modular Spectrum and Spectral Measures

The spectrum of a modular operator is spectral data of a positive self-adjoint operator attached to an algebra–state pair. In finite systems it can be listed as ratios of density-matrix eigenvalues. In continuum QFT it is usually continuous and must be handled through projection-valued or vector spectral measures. That difference is structural: a regulated entanglement spectrum is not automatically the spectrum of the limiting type-III modular operator.

Required background. Modular Hamiltonian definitions and domains supply spectral logarithms, supports, and unbounded generators.

Helpful background. Tomita–Takesaki flow supplies the relations among $\Delta$ , $J$ , and the invariant vector.

Spectral measures of Δ and K

Let $\Delta$ be the modular operator of a standard pair $(\mathcal A,\Omega)$ . The spectral theorem gives

\Delta=\int_{(0,\infty)}\lambda\,E_\Delta(d\lambda), \qquad K=-\log\Delta =\int_{\mathbb R}k\,E_K(dk),

with the change of variables $k=-\log\lambda$ . For a vector $\xi$ , the scalar measure

\mu_\xi^\Delta(B)=\langle\xi,E_\Delta(B)\xi\rangle

determines all bounded spectral observables and the domain tests for unbounded ones:

\lVert f(\Delta)\xi\rVert^2 =\int_{(0,\infty)}\lvert f(\lambda)\rvert^2\,d\mu_\xi^\Delta(\lambda).

Cross measures $\mu_{\xi,\eta}(B)=\langle\xi,E_\Delta(B)\eta\rangle$ determine modular correlators. They are generally complex measures and should not be called probability distributions unless $\xi=\eta$ is normalized.

The vector $\Omega$ is a $K=0$ eigenvector because $\Delta\Omega=\Omega$ . This fact does not imply that $0$ is an isolated eigenvalue or that other eigenvectors form a basis.

The structural map places Modular Spectrum and Spectral Measures between intrinsic modular data and the additional hypotheses that permit a geometric interpretation.

A standard algebra-state pair determines the Tomita operator, modular conjugation, modular flow, and relative modular data; only additional covariance or conformal hypotheses turn the flow into wedge or ball geometry. — Tomita polar decomposition intrinsically produces $J$ and $\Delta$ . Wedge boosts and CFT ball flow are special consequences of locality, covariance, the spectrum condition, the vacuum, and—only for the ball—the conformal map. The diagram is schematic and not to scale.

Reciprocal spectral symmetry

Modular conjugation obeys

J\Delta J=\Delta^{-1}, \qquad J KJ=-K.

Consequently the spectrum has reciprocal symmetry:

\lambda\in\operatorname{spec}(\Delta) \Longleftrightarrow \lambda^{-1}\in\operatorname{spec}(\Delta),

or equivalently $k\leftrightarrow-k$ for $K$ . The spectral projections transform as

J E_K(B)J=E_K(-B)

for Borel sets $B\subset\mathbb R$ . Vector measures need not themselves be even because $J\xi$ may differ from $\xi$ ; the relation is

\mu_{J\xi}^{K}(B)=\mu_\xi^{K}(-B).

This is a useful numerical check in a doubled finite model. Failure usually signals that the complementary inverse factor was omitted.

Point, continuous, and pair-dependent data

For a faithful $n$ -level density matrix with eigenvalues $p_i$ , standard-form matrix units are eigenvectors:

\Delta(\lvert i\rangle\langle j\rvert) =\frac{p_i}{p_j}\lvert i\rangle\langle j\rvert, \qquad K(\lvert i\rangle\langle j\rvert) =-\log\!\frac{p_i}{p_j}\lvert i\rangle\langle j\rvert.

The modular spectrum is therefore the ratio set $\{p_i/p_j\}$ , with multiplicities. A one-sided entanglement spectrum $\{-\log p_i\}$ is different data: only its pairwise differences appear in the standard modular spectrum.

In infinite dimensions, $\Delta$ may have absolutely continuous, singular continuous, and point parts. One should report which object has been computed:

eigenvalues of a regulated one-sided $K_A$ ;
the spectrum of a finite standard-form $\Delta$ ;
a vector spectral density for selected observables;
or an algebraic invariant such as the Connes spectrum obtained by intersecting modular spectra over a class of states.

These objects answer different questions. The Connes invariant classifies factor types and is not the ordinary spectrum of one chosen state; see Connes 1973, §§2–3 for the state-independent construction.

What survives a change of representation

Unitary equivalence of standard representations carries $E_\Delta(B)$ and all vector measures covariantly, so spectral type and multiplicity are invariant under that equivalence. Changing the state changes $\Delta$ and generally changes its measures. Multiplying a one-sided density matrix by a scalar shifts $-\log\rho$ but leaves ratio spectra and adjoint modular flow unchanged.

Changing the algebra is more consequential. Restricting to a subalgebra does not amount to deleting some eigenvalues; it constructs a new Tomita operator. Likewise, a UV cutoff can turn continuous spectral weight into a large discrete set without preserving multiplicities in a simple way.

For local type-III algebras, the absence of a trace and the strong accumulation of modular scales are not finite-size nuisances. A controlled continuum statement should therefore be phrased in terms of smeared spectral measures, convergence of correlators, or resolvents rather than convergence of an ordered eigenvalue list.

Spectral reconstruction from correlators

For bounded $A$ , set $\xi=A\Omega$ . Then

C_A(s)=\langle\xi,e^{-isK}\xi\rangle =\int_{\mathbb R}e^{-isk}\,d\mu_\xi^K(k).

Bochner positivity makes $C_A$ a positive-definite function, and Fourier inversion reconstructs the measure distributionally. In practice, a finite modular-time window convolves the true measure with the window transform; discrete peaks narrower than the resolution cannot be distinguished from continuous weight.

A numerical report should state the time window, sampling, window function, frequency resolution, positivity residual, and the KMS detailed-balance check. It should not label a smoothed finite-window spectrum “the modular spectrum” without this qualification.

Common pitfalls

Equating entanglement energies with modular eigenvalues. One-sided levels $-\log p_i$ and standard modular differences $-\log(p_i/p_j)$ are related but not identical.

Inferring pure point spectrum from a cutoff. Every finite matrix has discrete spectrum. Continuum spectral type requires convergence of measures or resolvents, not just denser eigenvalue plots.

Confusing a factor invariant with a state-dependent spectrum. The Connes spectrum involves an intersection over modular data. A single $\Delta_\Omega$ contains more state-specific information.

Before assigning geometric meaning to this modular statement, use the validity map to check standardness, spectral domains, normalization, geometric hypotheses, and approximation control independently.

A decision map separates exact modular conclusions from failures caused by missing faithfulness, uncontrolled operator domains, absent geometric theorems, or perturbative nonlocal kernels. — A modular-flow claim is only as strong as its algebra–state standardness, spectral domain, normalization, geometric hypotheses, and approximation control. Dashed branches mark conditional steps; the terminal warnings identify common out-of-domain uses. The diagram is schematic and not to scale.

References

Connes, Alain. “Une classification des facteurs de type III.” Annales Scientifiques de l’École Normale Supérieure 6 (1973): 133–252. Numdam.