---
title: Central Witten Index in Supersymmetric Theories
url: https://www.emergentmind.com/topics/central-witten-index
type: topic
---

# Central Witten Index in Supersymmetric Theories

The expression **central Witten index** is not used uniformly across the literature. Its common starting point is the ordinary Witten index
\[
I_W \equiv \mathrm{Tr}\!\left[(-1)^F e^{-\beta H}\right],
\]
which, in a supersymmetric theory with discrete spectrum, reduces to the difference between bosonic and fermionic zero-energy states. From this common core, different works develop distinct notions: a BPS index built from a central-charge-shifted Hamiltonian, an extension of the index to non-Fredholm operators through spectral regularization, a wall-crossing-safe sector of a quiver spectrum, or a torsion refinement encoding global anomalies. Several works also state explicitly that they do **not** introduce a central-charge-dependent refinement, even when the ordinary Witten index remains the central protected observable of the theory [1011.0940] [2404.18442] [2207.13858].

## 1. Ordinary Witten index and the baseline notion of protection

In the standard supersymmetric setting, the Witten index is
\[
I_W=\mathrm{Tr}\!\left[(-1)^F e^{-\beta H}\right].
\]
If the spectrum is discrete and supersymmetry is unbroken in the usual pairing sense, every positive-energy state is paired with a state of opposite fermion number, so only zero-energy states contribute:
\[
I_W = N_B(E=0)-N_F(E=0).
\]
A nonzero index implies the existence of at least one supersymmetric ground state and therefore excludes spontaneous supersymmetry breaking, whereas a vanishing index is only a necessary condition for such breaking, since it may also arise from equal numbers of bosonic and fermionic supersymmetric vacua [1011.0940].

In Euclidean path-integral language, this index is the partition function with **periodic** boundary conditions for both bosons and fermions,
\[
I_W = Z_{\rm P},
\]
in contrast with the thermal partition function, which uses antiperiodic fermions. The same literature emphasizes the topological character of the index: as long as supersymmetry is preserved and the zero-energy structure is not altered by singular spectral changes, the index is invariant under continuous deformations of couplings and masses [1006.2468].

This ordinary definition remains the reference point even in papers that discuss more elaborate structures. Kanamori’s lattice construction notes that “refined” or “central” versions can be imagined by inserting extra conserved charges, for example \(e^{-\beta Z}\), but does not introduce such refinements; it computes only the standard trace with \((-1)^F\) [1011.0940]. Likewise, the gauge-theory analysis of dynamical supersymmetry breaking defines the index by zero-mode counting and explicitly states that it does not introduce any central charge or central-charge-graded Witten index [1902.11078].

## 2. Central-charge and BPS-Hamiltonian refinements

A direct sense in which one obtains a **central** Witten index appears when the supersymmetry algebra is written in terms of a positive BPS Hamiltonian rather than the bare Hamiltonian. The general pattern described in the BMN analysis is
\[
\{Q,Q^\dagger\}=2(H-|Z|),
\]
so that one defines
\[
\Delta := H-|Z|,\qquad
I=\mathrm{Tr}\!\left[(-1)^F e^{-\beta \Delta}\right].
\]
In this form, the index receives contributions from states saturating the BPS bound \(H=|Z|\). With additional commuting charges \(Q_a\), one may further refine it to
\[
I(\{\mu_a\})=\mathrm{Tr}\!\left[(-1)^F e^{-\beta \Delta} e^{-\sum_a \mu_a Q_a}\right],
\]
provided the fugacity insertion is compatible with the chosen supercharge [2404.18442].

The BMN matrix quantum mechanics furnishes a concrete realization of this idea. Choosing a particular supercharge \(Q=Q_-^4\), the paper defines
\[
2\Delta := \{Q,Q^\dagger\}
= 2H - \frac{2\mu}{3}M^{12} - \frac{\mu}{3}(M^{45}+M^{67}+M^{89}),
\]
and the refined index becomes a trace over \((-1)^F\) weighted by \(\Delta\) and by chemical potentials for the Cartans \(M^{12}, M^{45}, M^{67}, M^{89}\). In this formulation, the index counts not only supersymmetric ground states but also the difference between bosonic and fermionic BPS states with nonzero spins. The same analysis shows that the resulting index sets a lower bound on the entropy and exhibits \(N^2\) growth, which is interpreted there as predicting BPS black holes in M-theory asymptotic to the plane-wave geometry [2404.18442].

This central-charge/BPS-Hamiltonian interpretation is structurally close to superconformal indices. In the same BMN work, the large-\(N\) irreducible-vacuum index is related to the superconformal index of ABJM theory after matching the \(\mathrm{SU}(2|4)\) and \(\mathrm{OSp}(8|4)\) generators and their fugacities [2404.18442]. In that sense, a central Witten index is a Witten index whose exponential weight is the positive operator singled out by the supersymmetry algebra itself.

## 3. Spectral central values and the non-Fredholm Witten index

A different, highly analytic notion of centrality arises in the extension of the Witten index beyond the Fredholm case. For a densely defined closed operator \(T\), one defines the resolvent-regularized and semigroup-regularized Witten indices by
\[
W_r(T):=\lim_{\lambda\uparrow 0}
(-\lambda)\,\operatorname{tr}\!\left[(T^*T-\lambda I)^{-1}-(TT^*-\lambda I)^{-1}\right],
\]
and
\[
W_s(T):=\lim_{t\to\infty}
\operatorname{tr}\!\left[e^{-tT^*T}-e^{-tTT^*}\right],
\]
whenever the relevant trace-class hypotheses hold. In the Fredholm case, both reduce to the ordinary Fredholm index [1505.04895].

The central model in this theory is the operator
\[
\boldsymbol{D}_A = \frac{d}{dt}+A
\]
on \(L^2(\mathbb R;\mathcal H)\), associated with a self-adjoint path \(A(t)\) having asymptotes \(A_\pm\). When \(0\in\rho(A_+)\cap\rho(A_-)\), \(\boldsymbol{D}_A\) is Fredholm and
\[
W_r(\boldsymbol{D}_A)=W_s(\boldsymbol{D}_A)=\operatorname{ind}(\boldsymbol{D}_A)=\xi(0;A_+,A_-),
\]
where \(\xi(\lambda;A_+,A_-)\) is Krein’s spectral shift function [1505.04895].

The genuinely non-Fredholm case appears when \(0\in \sigma(A_+)\cup\sigma(A_-)\). Then the index is no longer an integer in general, but if \(0\) is a right and left Lebesgue point of the spectral shift function, the Witten index still exists and is given by the **symmetrized central value**
\[
W_r(\boldsymbol{D}_A)=W_s(\boldsymbol{D}_A)
=
\frac{\xi_L(0_+;A_+,A_-)+\xi_L(0_-;A_+,A_-)}{2}.
\]
This formula makes the “central” feature literal: the Witten index is extracted from the behavior of a spectral asymmetry function at the distinguished energy \(0\) [1505.04895].

The same analytic theme reappears in split-step quantum walks under non-Fredholm conditions. There the Witten index is defined through semigroup regularization and computed from the spectral shift function of a rank-one perturbation of a fourth-order difference operator. In the gapless case, the index can take half-integer values, a phenomenon attributed in that work to resonances at quasi-energies \(\pm1\) [2111.04108].

## 4. Topological centrality: bulk–edge indices and wall-crossing-safe sectors

In discrete-time supersymmetric quantum walks, the Witten index becomes a topological invariant central to phase classification. For one-dimensional split-step walks with chiral symmetry, the index is
\[
\operatorname{ind}(T,C)=\dim\ker Q_+ - \dim\ker Q_-,
\]
and in the Fredholm case it equals the Fredholm index of \(Q_+\). Matsuzawa’s index theorem expresses it as
\[
\operatorname{ind}(T,C)=\operatorname{wn}(F_+) - \operatorname{wn}(F_-),
\]
the difference of winding numbers of asymptotic symbol functions determined by the right- and left-limit coins. This is an explicit bulk–edge correspondence: the analytic Witten index equals a difference of bulk topological invariants, and a nonzero value guarantees protected \(\pm1\) eigenstates of the walk operator [1903.05061].

A closely related classification for anisotropic coins gives
\[
\operatorname{ind}(T,C)=
\begin{cases}
-\operatorname{sgn}p, & |a(L)|<|p|<|a(R)|,\\[4pt]
+\operatorname{sgn}p, & |a(R)|<|p|<|a(L)|,\\[4pt]
0, & \text{otherwise,}
\end{cases}
\]
showing that the index is completely determined by asymptotic bulk data. In this literature, “central” does not mean central charge; it means that the Witten index is the central bulk invariant controlling topologically protected bound states [1903.00559].

A parallel notion appears in \(\mathcal N=4\) quiver quantum mechanics. There the full refined Witten index depends on the FI chamber and is computed by Jeffrey–Kirwan residues, with wall crossing coming from the Coulomb branch at infinity [1407.2567]. However, the paper on mutation and quiver invariants isolates a chamber-independent sector,
\[
\Omega_{\rm Q}\big\rvert_{\rm Inv}
=
\lim_{\beta\to\infty}
\operatorname{Tr}_{L^2}\!\left[(-1)^{2J_3}\,{\bf y}^{2I+2J_3}\,e^{-\beta H_{\rm Q}(\zeta=0)}\right],
\]
called the **quiver invariant**. This is the wall-crossing-safe part of the spectrum, interpreted there as the intrinsic Higgs or single-centered contribution. The full index factorizes into Coulombic pieces built from induced quivers and products of subquiver invariants, and the modified mutation rule \(\tilde\mu\) preserves \(\Omega_{\rm Q}\rvert_{\rm Inv}\) for all nodes [1504.00068]. In this setting, the central Witten index is naturally the chamber-independent core of the BPS spectrum rather than the full FI-dependent index.

## 5. Lattice and disorder-averaged realizations

The lattice literature emphasizes the path-integral realization
\[
I_W = Z_{\rm P},
\]
the periodic partition function, and addresses the problem that Monte Carlo methods usually measure normalized expectation values rather than the absolute normalization of \(Z_{\rm P}\). Kanamori’s construction fixes the lattice path-integral measure and uses a Gaussian regularization functional to extract the partition function. After integrating out fermions and reweighting the determinant phase, the Witten index is obtained as
\[
w = Z_{\rm P}
=
C\,
\frac{\langle \sigma[D_{\rm P}] \rangle_{0,{\rm P}}}
{\left\langle \exp\!\left[+S'_{\rm P}-\tfrac{1}{2}\sum_i \mu^2 \phi_i^2\right]\right\rangle_{0,{\rm P}}},
\]
with \(C=\mu^{-N}\). Applied to \(\mathcal N=2\) supersymmetric quantum mechanics, this reproduces the expected values \(w=1\) for quartic superpotentials and \(w=0\) for cubic superpotentials. The construction is explicitly for the ordinary Witten index; the paper states that no refined or central-charge-dependent index is introduced [1011.0940].

A different protected setting is the disorder-averaged \(\mathcal N=4\) SYK-type quantum mechanics. There the exact Witten index is
\[
\mathcal I=(p-1)^N,
\]
independent of the disordered couplings \(C_{i_1\cdots i_p}\). The disorder-averaged collective-field formalism still reproduces this exact answer, and when several copies are disorder-averaged together, nontrivial couplings between the copies are generated but the index remains factorized. In that sense, the Witten index functions as a central protected quantity that survives both disorder averaging and multi-copy couplings [2504.05379].

These constructions make clear that “central” can refer not only to central charges but also to robustness under discretization, reweighting, and disorder averaging. At the same time, the lattice papers are explicit that any genuinely central-charge-sensitive refinement would require additional operator insertions or twisted boundary conditions beyond the standard periodic partition function [1011.0940].

## 6. Noncompact dynamics, anomaly refinements, and the range of meanings of “central”

In noncompact supersymmetric dynamics, Lee and Yi distinguish sharply between the physical \(L^2\) Witten index and twisted partition functions computed with infrared regulators. The physical index is
\[
\mathcal I_W^{L^2}
=
\sum_n (-1)^n \dim H^n_{L^2}(\mathcal M),
\]
and is the regulator-independent quantity that counts normalizable supersymmetric ground states. By contrast, the localization quantity \(\Omega(y;x)\) generally computes a bulk term contaminated by continuum contributions and can be fractional. In many one-dimensional gauge systems, however, the integral \(L^2\) Witten index can still be extracted from the twisted partition function through rational invariants and Möbius inversion, exactly as in wall-crossing formulae [1602.03530]. Here the “central” index is the integral, physical \(L^2\) index rather than the more readily computed twisted partition function.

A far-reaching refinement appears in heterotic anomaly theory. For compact \(2\)-dimensional \(\mathcal N=(0,1)\) SQFTs, the ordinary index is
\[
I_{\mathcal T}(q)=\operatorname{Tr}_{\mathcal T}(-1)^F q^{H_L}\bar q^{H_R},
\]
and its coefficients encode perturbative target-space anomalies. For mildly noncompact theories, the paper introduces a **torsion Witten index**
\[
J_{\mathcal B}(q),
\]
defined through APS-type gluing and only well-defined modulo weakly holomorphic modular forms and integral Laurent series. This torsion refinement captures global anomalies and is presented as one of the invariants suggested by topological modular forms. For actual heterotic theories, the torsion index is expected to vanish [2207.13858].

Accordingly, the phrase **central Witten index** can denote several non-equivalent constructions. In one usage it is the BPS index built from a central-charge-shifted Hamiltonian \(\Delta\). In another it is the central spectral value of a spectral shift function at zero. In quantum walks and quiver theories it is the central topological or wall-crossing-safe part of a spectrum. In anomaly theory it becomes a torsion refinement. Several works remain with the ordinary Witten index and explicitly note that no central-charge-dependent version is being introduced [1902.11078].

Source: https://www.emergentmind.com/topics/central-witten-index