Microscopic Indices and Protected Degeneracies

A microscopic black-hole calculation often produces a supersymmetric index, not the absolute number of states. The distinction is decisive: an index is a signed trace designed to cancel long multiplets, while an absolute degeneracy counts every state positively. We prove the deformation-invariance statement under precise spectral hypotheses, then show why continua and walls of marginal stability require a chamber label.

Required background. Gauge Theory and String-Theory Preparation supplies supersymmetry algebras and charge sectors. Gravitational Actions and Boundary Terms supplies the asymptotic boundary conditions that define the conserved black-hole charges.

Helpful background. Quantum Gravity frames the microscopic/macroscopic comparison, Quantum Information supplies traces and graded Hilbert spaces, and Black-Hole Geometry and Mechanics supplies the charge and entropy observables to be compared.

Index, helicity supertrace, and degeneracy

For a supersymmetric quantum system with fermion parity $(-1)^F$, the Witten index is

$$\mathcal I(\beta)=\operatorname{Tr}_{\mathcal H_\Gamma} \left[(-1)^F e^{-\beta H}\right].$$

The trace is taken in a fixed charge sector $\Gamma$ and with specified boundary conditions. Here $H$ is the nonnegative operator supplied by the chosen supercharge anticommutator. For charged BPS particles it measures energy above the relevant central-charge bound, rather than the unshifted ADM mass; a massive BPS state can therefore have $H=0$. Paired positive-$H$ bosonic and fermionic states cancel, while supersymmetric ground states contribute with sign. By contrast, the absolute degeneracy is

$$d(\Gamma)=\operatorname{Tr}_{\mathcal H_{\Gamma,\mathrm{BPS}}}\mathbf 1 =d_B+d_F,$$

whereas $\mathcal I=d_B-d_F$ in the simplest case. Therefore an index or helicity supertrace is not the absolute degeneracy unless an additional argument controls cancellations.

In four-dimensional theories, a BPS state breaks fermionic generators and carries their zero modes. An uninserted trace vanishes. One instead uses a helicity supertrace

$$B_{2n}(\Gamma)=\frac{(-1)^n}{(2n)!} \operatorname{Tr}_{\mathcal H_\Gamma} \left[(-1)^{2h}(2h)^{2n}\right],$$

with $2n$ chosen to soak up the broken-supersymmetry zero modes. Here $h$ is the third component of angular momentum in the rest frame. Other compactifications use elliptic genera or refined indices; their chemical potentials and spin structures are part of the definition, not optional notation.

Deformation invariance theorem

Theorem (graded-trace invariance). Let $H(\lambda)=\{Q(\lambda),Q(\lambda)^\dagger\}/2$ act on a $\mathbb Z_2$-graded Hilbert space in a fixed charge sector. Assume: the spectrum relevant to the trace is discrete; $e^{-\beta H}$ and its differentiated products are trace class; $Q$ is odd and preserves the domain; the deformation is supersymmetric and can be written $\partial_\lambda H=\{Q,V\}$ for an odd $V$; and no state enters from or escapes to a continuum boundary. Then $\partial_\lambda\mathcal I=0$.

The conclusion applies to the specified index. It neither proves that $d(\Gamma)$ is constant nor excludes jumps caused by changing the Hilbert space or boundary conditions.

Proof architecture

Differentiate under the trace, justified by the trace-class hypothesis:

$$\frac{\partial\mathcal I}{\partial\lambda} =-\beta\operatorname{Tr}\left[ (-1)^F\{Q,V\}e^{-\beta H} \right].$$

Because $[Q,H]=0$, move $Q$ through $e^{-\beta H}$. Graded cyclicity and $(-1)^FQ=-Q(-1)^F$ then give

$$\operatorname{Tr}\left[(-1)^FQV e^{-\beta H}\right] =-\operatorname{Tr}\left[(-1)^FV e^{-\beta H}Q\right],$$

which cancels the second term from the anticommutator. Hence $\partial_\lambda\mathcal I=0$. Equivalently, every nonzero-energy state belongs to a $Q$-doublet with opposite fermion parity, so its two trace contributions cancel. Witten develops this pairing and its use in supersymmetry breaking in Witten 1982, §§2–3, pp. 257–275.

For $B_{2n}$ one must instead check the multiplet character: helicity insertions do not commute with every supercharge, so the uninserted cyclicity proof cannot simply be copied. Each fermionic zero-mode pair contributes a factor vanishing linearly at $y=0$ to the graded character $\operatorname{Tr}[(-1)^{2h}e^{2yh}]$. A multiplet with more than $2n$ such factors has zero $2n$th derivative there and hence zero $B_{2n}$. This explains both the cancellation of the relevant long multiplets and why the insertion must saturate the BPS multiplet’s universal zero modes. Extra zero modes or a changed supersymmetry fraction can require a different supertrace.

Walls, chambers, and contour dependence

A multicenter BPS bound state can become marginal when

$$\lvert Z(\Gamma_1+\Gamma_2;t_\infty)\rvert =\lvert Z(\Gamma_1;t_\infty)\rvert+\lvert Z(\Gamma_2;t_\infty)\rvert.$$

This codimension-one condition defines a wall of marginal stability. On the wall the relative-separation mode ceases to be bound; crossing it can add or remove states. Thus the index is locally constant only inside a chamber of asymptotic moduli. A complete microscopic answer is $\Omega(\Gamma;t_\infty\in\mathcal C)$, not merely $\Omega(\Gamma)$.

In $\mathcal N=4$ dyon counting, one meromorphic partition function can encode several chambers: the Fourier coefficient is selected by an integration contour, and moving between chambers crosses poles. Sen derives this contour prescription and the wall geometry in Sen 2007, §§2–4, pp. 6–30. The jump is the residue associated with the decay products. At large charge it can be exponentially smaller than the leading single-center entropy, but it is not literally absent.

This reconciles the chamber dependence of a microscopic index with the attractor entropy: the leading area of a single-center solution depends only on charges, while multicenter sectors and exponentially suppressed terms can retain asymptotic-moduli information.

Failure modes and counterexamples

The proof fails for a continuum reaching zero energy because the paired density of states can have a boundary contribution; formal cyclicity of a non-trace-class operator is not valid. It also fails when a wall changes normalizability: a wavefunction can spread to infinite intercenter separation without any finite-energy level crossing. These are not counterexamples to the theorem because they violate its hypotheses.

A simpler counterexample to “index equals degeneracy” is one bosonic and one fermionic BPS ground state. The absolute degeneracy is $d=2$, while $\mathcal I=1-1=0$. A large entropy inferred from an index therefore needs an argument that cancellations are subleading, a positivity theorem, or a protected sector in which all contributing states have the same sign. The comparison must also match charges, ensemble, angular-momentum insertions, hair modes, and chamber.

Practical comparison with black-hole entropy

Before taking a logarithm, state which object is counted. Remove universal center-of-mass or exterior “hair” factors consistently, specify whether the index is microcanonical or extracted from a grand-canonical generating function, and identify its contour. Then compare the asymptotic growth

$$\log |\Omega(\Gamma;\mathcal C)| =S_{\rm macro}(\Gamma)+o(S_{\rm macro})$$

only in a charge scaling where both sides are controlled. Agreement of the leading exponential does not establish equality term by term; precision entropy matching tests logarithmic and exponentially suppressed structure.

Exercises

Show explicitly why a long supersymmetry multiplet with paired states $|b\rangle$ and $|f\rangle=Q|b\rangle/\sqrt{2E}$ contributes zero to $\mathcal I(\beta)$, and identify the hypothesis that fails when the pair becomes a nonnormalizable scattering state at $E=0$.

Solution

The paired states have the same energy $E\gt0$ because $[Q,H]=0$, while their fermion parities differ. Their contribution is $e^{-\beta E}-e^{-\beta E}=0$. The normalization by $\sqrt{2E}$ uses a positive, discrete eigenvalue and a normalizable state. At a continuum threshold $E=0$, the partner may not lie in the Hilbert space and the heat-kernel trace can acquire a boundary term from the density of scattering states. The discreteness, trace-class, and no-escape hypotheses of the theorem then fail, so invariance cannot be concluded by graded cyclicity.

References

• Sen, Ashoke. “Walls of Marginal Stability and Dyon Spectrum in N=4 Supersymmetric String Theories.” Journal of High Energy Physics 05 (2007): 039. DOI. Open manuscript.
• Witten, Edward. “Constraints on Supersymmetry Breaking.” Nuclear Physics B 202 (1982): 253–316. DOI.