---
title: Two-Centered Black Hole Index
url: https://www.emergentmind.com/topics/two-centered-black-hole-index
type: topic
---

# Two-Centered Black Hole Index

Searching arXiv for the cited papers to ground the article in the published literature.
arXiv search query: 2507.08551 Multi-centered Black Hole Quantum Mechanics and Generalized Error Functions
The two-centered black hole index is the protected contribution of a bound state of two BPS centers to the total BPS index in four-dimensional supersymmetric string compactifications. In the formulations relevant to type II Calabi–Yau compactifications, \( \mathcal N=2 \) supergravity, heterotic string theory on \(T^6\), and \(\mathbb Z_N\) CHL models, it is determined by the relative dynamics of two mutually non-local charges, by localization on the classical phase space of two-center solutions, or by extracting the polar part of meromorphic Siegel or Jacobi generating functions. Its central roles are to encode the jump of the index across walls of marginal stability, to supply the two-body factor in primitive wall-crossing formulae, and to provide the non-holomorphic completion needed for modularity in mock Jacobi and related automorphic structures [2507.08551, 1103.1887, 1011.1258].

## 1. Physical set-up and classical two-center geometry

For two mutually non-local BPS dyons of charges \(\gamma_1,\gamma_2\) in four dimensions, the center-of-mass decouples, and the relative motion is described by an \(\mathcal N=4\) supersymmetric quantum mechanics on \(\mathbb R^3\) with coordinate \(\vec r=\vec x_1-\vec x_2\). The effective Lagrangian can be written schematically as
\[
L=\frac m2(\dot{\vec x}^{\,2}+D^2+2\bar\lambda \dot\lambda)+(-U D+\vec A\cdot \dot{\vec x})+(\nabla U\cdot \bar\lambda \sigma \lambda),
\]
where \(m=m_1m_2/(m_1+m_2)\) is the reduced mass,
\[
U(r)=-\frac12\left(\frac{\kappa}{|r|}-c\right),\qquad \kappa\equiv \langle \gamma_1,\gamma_2\rangle\in \mathbb Z,
\]
\(c\propto \Im (e^{-i\phi} Z(\gamma_1,\gamma_2))\) is the real FI-parameter, \(\vec A\) is the Dirac monopole potential, and \(\lambda\) are two complex fermions. Classical BPS bound states solve \(U(r)=0\), so
\[
|r|=\frac{\kappa}{c}\qquad \text{for }\kappa c>0,
\]
and their moduli space is the two-sphere \(S^2\) of radius \(\kappa/c\) [2507.08551].

In four-dimensional \(N=2\) supergravity, the same structure appears through Denef’s equilibrium condition. For two centers of charges \(\Gamma_1,\Gamma_2\), the relative separation is fixed,
\[
\frac{\langle \Gamma_1,\Gamma_2\rangle}{|\vec r|}=c_1=-c_2,
\]
so after removing the overall center-of-mass motion the allowed relative orientations form a two-dimensional symplectic manifold \(M_2\cong S^2\). The pull-back of the canonical two-form gives
\[
\omega=\frac12\,\Gamma_{12}\sin\theta\, d\theta\wedge d\phi,\qquad \Gamma_{12}\equiv \langle \Gamma_1,\Gamma_2\rangle,
\]
and the moment map for rotations about the \(z\)-axis is
\[
J_3(\theta,\phi)=\frac12\,\Gamma_{12}\cos\theta.
\]
Equivalently, the classical angular momentum vector is
\[
\vec J=\frac12\,\langle \Gamma_1,\Gamma_2\rangle\,\hat r.
\]
These formulations identify the two-centered problem as the quantization of a compact \(S^2\) phase space in the absence of scaling solutions [1103.1887, 0807.4556].

## 2. Quantization, localization, and the configurational factor

Quantizing the two-center phase space produces the universal configurational factor multiplying the single-center indices carried by the two constituents. In geometric quantization, one views \((S^2,\Omega)\) as a compact Kähler manifold, with stereographic coordinate
\[
z=\tan(\theta/2)e^{i\phi},
\]
Kähler potential
\[
K(z,\bar z)=2|J|\log(1+|z|^2),
\]
and prequantum line bundle \(\mathcal L\simeq O(2|J|)\). The holomorphic basis is
\[
\psi_m(z)=z^m,\qquad m=0,1,2,\ldots,2|J|.
\]
Without spin one gets \(2|J|+1\) states, while including the fermionic \(\mathrm{Spin}^c\) correction gives exactly \(2|J|\) states. Since \(|J|=\frac12|\langle \Gamma_1,\Gamma_2\rangle|\), the two-center configurational index is
\[
I_2^{\rm conf}(\Gamma_1,\Gamma_2)=|\langle \Gamma_1,\Gamma_2\rangle|.
\]
Including the sign from fermion zero modes and the horizon contributions \(\Omega_{\rm int}(\Gamma_p)\), the full two-center BPS index becomes
\[
I_2(\Gamma_1,\Gamma_2)=(-1)^{\langle \Gamma_1,\Gamma_2\rangle-1}\,|\langle \Gamma_1,\Gamma_2\rangle|\,\Omega_{\rm int}(\Gamma_1)\Omega_{\rm int}(\Gamma_2)
\]
[0807.4556].

The same factor follows from equivariant localization on \(M_2\). The only fixed points of the \(U(1)_{J_3}\)-action are the north and south poles, \(\theta=0,\pi\), corresponding to the two collinear configurations. Localization gives the refined two-center factor
\[
g_{\rm ref}(\gamma_1,\gamma_2;y)=(-1)^{\langle \gamma_1,\gamma_2\rangle+1}\,
\frac{y^{\langle \gamma_1,\gamma_2\rangle}-y^{-\langle \gamma_1,\gamma_2\rangle}}{y-y^{-1}},
\]
and in the unrefined limit
\[
g(\gamma_1,\gamma_2)=(-1)^{\langle \gamma_1,\gamma_2\rangle+1}\,|\langle \gamma_1,\gamma_2\rangle|.
\]
Accordingly, for two distinct centers,
\[
\Omega_2(\Gamma_1,\Gamma_2;y)=g_{\rm ref}(y)\,\Omega_1(\Gamma_1;y)\,\Omega_1(\Gamma_2;y).
\]
The factor
\[
\frac{y^{\Gamma_{12}}-y^{-\Gamma_{12}}}{y-y^{-1}}
\]
is the character of the spin-\(\frac{\Gamma_{12}-1}{2}\) representation carried by the relative motion [1103.1887, 1011.1258].

## 3. Supersymmetric quantum mechanics and the error-function completion

The two-body supersymmetric quantum mechanics admits a direct localization computation of the refined Witten index
\[
I_2(\beta,y)=\mathrm{Tr}\,\bigl[(-1)^F\,y^{2J_3}\,e^{-\beta H}\bigr].
\]
By the standard argument for supersymmetric \(\sigma\)-models with a potential, \(I_2\) localizes onto constant modes. One finds the finite-dimensional integral
\[
I_2(\beta,y)=(4\pi^2\beta)^{-1}\int d^3r\, dD\, d^2\lambda\, d^2\bar\lambda\,
y^{\kappa\cos\theta}\,
e^{-\beta[U D + (m/2)D^2+\nabla U\cdot \bar\lambda \sigma \lambda]}.
\]
The fermion integral yields
\[
\int d^2\lambda\, d^2\bar\lambda\,
e^{-\beta \nabla U\cdot \bar\lambda \sigma \lambda}
=-\beta^2(\nabla U)^2=-\beta^2\frac{\kappa^2}{4|r|^4},
\]
and after passing to spherical coordinates and performing the Gaussian \(D\)-integral one obtains
\[
I_2(\beta,y)=
-\frac12\,\frac{y^\kappa-y^{-\kappa}}{y-y^{-1}}
\left[\operatorname{sign}(\kappa)+E_1\!\left(c\sqrt{\frac{\beta}{8m}}\right)\right].
\]
In the unrefined limit,
\[
I_2(\beta)= -\frac{\kappa}{2}\left[\operatorname{sign}(\kappa)+E_1\!\left(c\sqrt{\frac{\beta}{8m}}\right)\right].
\]
Here
\[
E_1(x)=\operatorname{Erf}(\sqrt{\pi}\,x)\equiv \int_{-\infty}^{\infty} e^{-\pi (t-x)^2}\,\operatorname{sign}t\, dt
\]
[2507.08551].

The purely holomorphic or step-function index counts normalizable bound states and is obtained by replacing \(E_1(x)\to \operatorname{sign}(x)\),
\[
I_2^{\rm hol}=-\frac{\kappa}{2}\,[\operatorname{sign}(\kappa)+\operatorname{sign}(c)]
=-\kappa\,\theta(\kappa c).
\]
The difference \(I_2-I_2^{\rm hol}\) is the continuum or spectral-asymmetry piece,
\[
M_1(x)=E_1(x)-\operatorname{sign}(x),
\]
with standard integral representation
\[
M_1(x)=-\operatorname{sign}(x)\,\frac12\,\operatorname{erfc}(|x|\sqrt{\pi})
=-2\int_{|x|}^{\infty} e^{-\pi t^2}\,dt.
\]
Equivalently,
\[
R(u)=\int_u^\infty e^{-\pi t^2}\,dt=\frac12\,\operatorname{erfc}(u\sqrt{\pi}).
\]

Introducing \(\tau_2=\beta/(2\pi)\) and the stability ratio
\[
u\equiv \frac{\Im (\bar Z(\gamma_1)Z(\gamma_2))}
{\sqrt{2|Z(\gamma_1)Z(\gamma_2)Z(\gamma_1+\gamma_2)|}},
\]
the completed index can be written as
\[
I_2(\tau,z)=
-\frac{\kappa}{2}\,\frac{y^\kappa-y^{-\kappa}}{y-y^{-1}}\,
\bigl[\operatorname{sign}(\kappa)+\operatorname{Erf}(\sqrt{2\tau_2}\,u)\bigr],
\]
or equivalently
\[
\widehat h_2(\tau,z)=h_2(\tau,z)-\kappa\,R(\sqrt{2\tau_2}\,u).
\]
The non-holomorphic term exactly cancels the anomaly of the holomorphic term under \(S\)-duality, so the completion transforms as an ordinary non-holomorphic Jacobi form of weight \(k\), with \(k=-\tfrac12\) in the conventions of the cited treatment. Since only one error-function is needed, the result is a depth-one mock Jacobi form [2507.08551].

## 4. Wall-crossing, rational invariants, and continuum scattering states

Across the two-center wall \(\gamma\to \gamma_1+\gamma_2\), the total index jumps by the index of the two-center solution, with the centers carrying rational invariants
\[
\bar\Omega(\gamma)\equiv \sum_{m|\gamma}\frac{\Omega(\gamma/m)}{m^2}.
\]
The primitive wall-crossing formula is
\[
\Delta \bar\Omega(\gamma_1+\gamma_2)=
g(\gamma_1,\gamma_2)\,\bar\Omega(\gamma_1)\,\bar\Omega(\gamma_2)
=
(-1)^{\langle \gamma_1,\gamma_2\rangle+1}\,
|\langle \gamma_1,\gamma_2\rangle|\,
\bar\Omega(\gamma_1)\bar\Omega(\gamma_2).
\]
For the refined index,
\[
\Omega_{\rm ref}(\gamma,y)=\mathrm{Tr}_{H_\gamma}(-y)^{2J_3},
\]
one introduces
\[
\bar\Omega_{\rm ref}(\gamma,y)=
\sum_{m|\gamma}\bigl[m^{-1}\Omega_{\rm ref}(\gamma/m,y^m)\bigr],
\]
and the refined jump is
\[
\Delta \bar\Omega_{\rm ref}(\gamma_1+\gamma_2;y)=
g_{\rm ref}(\gamma_1,\gamma_2;y)\,
\bar\Omega_{\rm ref}(\gamma_1;y)\,
\bar\Omega_{\rm ref}(\gamma_2;y).
\]
The physical rationale is that Bose–Fermi combinatorics can be traded for Maxwell–Boltzmann statistics once \(\Omega\) is replaced by \(\bar\Omega\) [1011.1258].

A complementary macroscopic description comes from the supersymmetric quantum mechanics on Taub–NUT for the relative dynamics of two \( \frac12\)-BPS centers. In that setting the localized index contains a temperature-independent contribution from discrete BPS bound states and a temperature-dependent contribution from a spectral asymmetry in the continuum. For each charge sector \(q\), in the limit \(y\to 1\), \(v\to 0\),
\[
I_q^+(\tau_2;\lambda)=
4|q| -4q\,\mathrm{erf}\!\Big(\sqrt{\frac{2\pi \tau_2}{R}}\,(q-R\lambda)\Big)
-\frac{2}{\pi}\sqrt{\frac{2R}{\tau_2}}\,
e^{-\frac{2\pi\tau_2}{R}(q-R\lambda)^2}.
\]
The first two terms arise from discrete BPS bound states, while the complementary-error-function and Gaussian term come from a spectral asymmetry in the continuum. Upon the identifications \(m=2R\), \(u_2=\Im z/\tau_2=-m\lambda\), and \(\ell=2q\), this agrees with the Fourier coefficients of the non-holomorphic completion \(A_m^*(\tau,\bar\tau;z)\). This provides a physical derivation of the completion term \(\Delta_m\) required for modularity [1808.05606].

## 5. Negative discriminant states and bound-state metamorphosis

In heterotic string theory on \(T^6\), and similarly in CHL orbifolds, one can define the T-duality invariants
\[
Q^2\equiv Q\cdot Q,\qquad P^2\equiv P\cdot P,\qquad Q\cdot P,
\]
and the discriminant
\[
\Delta(\Gamma)=Q^2P^2-(Q\cdot P)^2.
\]
Classical single-centered \( \frac14\)-BPS black holes exist only when \(\Delta>0\); in four-dimensional \(N=4\) supergravity their horizon area is proportional to \(\sqrt\Delta\), and no regular attractor flow can be sustained for \(\Delta\le 0\). Accordingly, any microscopic index for states with \(\Delta<0\) must arise from genuinely multi-center configurations [1104.1498].

The exact dyon index \(\Omega(Q,P)\) is extracted from Fourier coefficients of \(1/\Phi_{10}\), and one expects a schematic decomposition
\[
\Omega(\Gamma)=\Omega_{\rm single}(\Gamma)+\sum_{\Gamma_1+\Gamma_2=\Gamma}\Omega_{2\text{-center}}(\Gamma_1,\Gamma_2).
\]
For a bound state of two half-BPS centers,
\[
\Omega_{2\text{-center}}(\Gamma_1,\Gamma_2)=
(-1)^{\langle \Gamma_1,\Gamma_2\rangle+1}\,
|\langle \Gamma_1,\Gamma_2\rangle|\,
\Omega_{1/2}(\Gamma_1)\Omega_{1/2}(\Gamma_2),
\]
with
\[
\Omega_{1/2}(Q,P)=f(Q^2/2),
\]
where \(f(n)\) is the coefficient of \(q^n\) in
\[
q^{-1}\prod_{k=1}^{\infty}(1-q^k)^{-24}.
\]

A subtlety arises when one or both centers are “small” charges with square \(-2\). In that case one imposes the identification rule called bound-state metamorphosis. When one center \(\Gamma_1\) satisfies \(\Gamma_1^2=-2\) and \(u=\langle \Gamma_1,\Gamma_2\rangle\neq 0\), the split \((\Gamma_1,\Gamma_2)\) must be identified with the \(S\)-dual split obtained by shifting \(\Gamma_1\to \Gamma_1+u\,\Gamma_2\), \(\Gamma_2\to \Gamma_2\). Explicitly, if \(\Gamma_1=(Q,0)\), \(\Gamma_2=(0,P)\), and \(Q\cdot P=u\), then
\[
(\Gamma_1,\Gamma_2)\sim (\Gamma_1',\Gamma_2),
\qquad
\Gamma_1'=(Q+uP,0),\quad \Gamma_2=(-uP,P).
\]
When both centers have square \(-2\), one must continue to identify an infinite \(S\)-duality orbit of such splits. After subtracting all two-center contributions subject to these identifications, one finds
\[
\Omega_{\rm single}(\Gamma)=0\qquad \text{for }\Delta(\Gamma)<0.
\]
All negative-discriminant states are therefore accounted for by two-centered black holes [1104.1498].

## 6. Generating functions, polar parts, and modular structure

For fixed magnetic invariant \(m>0\), the Fourier–Jacobi coefficient of the inverse Igusa cusp form is
\[
\psi_m(\tau,z)=\sum_{n,\ell\in \mathbb Z} d^{\rm dyon}(n,\ell,m)\,q^n\zeta^\ell.
\]
Physically this generating function decomposes as
\[
\psi_m(\tau,z)=\psi_m^F(\tau,z)+\psi_m^P(\tau,z),
\]
where \(\psi_m^F\) captures the single-centered black holes and is holomorphic in \(z\), while \(\psi_m^P\) captures two-centered bound states and is meromorphic in \(z\). The theorem of Dabholkar–Murthy–Zagier gives a unique non-holomorphic completion \(\Delta_m(\tau,\bar\tau,z)\) such that
\[
\widehat\psi_m^P=\psi_m^P+\Delta_m,\qquad
\widehat\psi_m^F=\psi_m^F-\Delta_m,
\]
and both completed pieces transform as true Jacobi forms of weight \(-10\) and index \(m\). In particular,
\[
\psi_m^P(\tau,z)=\frac{d(m)}{\eta(\tau)^{24}}\,A_{2,m}(\tau,z),
\qquad
\widehat A_{2,m}=A_{2,m}+A_m^*,
\]
so the completion of the two-center part is controlled by \(A_m^*(\tau,\bar\tau;z)\) [1808.05606].

In the heterotic \(T^6\) setting, the two-centered generating function \(Z_2(\Omega)\) is defined as the sum of all terms subtracted from \(1/\Phi_{10}\) in order to remove the two-center contributions. Near \(z=0\),
\[
\frac1{\Phi_{10}(\Omega)}\sim
(e^{\pi i z}-e^{-\pi i z})^{-2}\,\eta(\tau)^{-24}\eta(\sigma)^{-24},
\]
which is the generating function of two-center bound states formed from two half-BPS cores. Subtracting the leading double pole and its \(PSL(2,\mathbb Z)\)-images yields \(Z_2(\Omega)\), and one defines
\[
F(\Omega)=\frac1{\Phi_{10}(\Omega)}-Z_2(\Omega).
\]
The Fourier coefficients of \(F\) in the attractor chamber are moduli-independent, vanish whenever \(4mn-\ell^2<0\) or \(m<0\) or \(n<0\), and are free of wall-crossing [2510.05219].

In \(\mathbb Z_N\) CHL models, the two-centered generating function is
\[
S_k(\Omega)=
N\sum_{m,n,\ell\in \mathbb Z} (-1)^{\ell+1}\,d^{\rm two}(m,n,\ell)\,
e^{2\pi i \mathrm{Tr}(T\Omega)},
\qquad
T=\begin{pmatrix}m&\ell/2\\ \ell/2&n/N\end{pmatrix},
\]
and the single-centered generating function is
\[
F_k(\Omega)=\frac1{\Phi_k(\Omega)}-S_k(\Omega).
\]
The explicit decomposition \(S_k=\mathcal{CF}_1+\mathcal{CF}_2+\mathcal{CF}_3+\mathcal{CF}_4\) subtracts the \(n_2=0\) poles of \(1/\Phi_k\), while the last term implements the quotient by the metamorphosis subgroup \(G_r^N\). For \(N=2,3\), each of the four sums \(\mathcal{CF}_i\) is absolutely convergent and uniformly so on compact subsets of the attractor chamber, and \(F_k\) is holomorphic in the domain \(\det(\Im \Omega)>1/4\) with only \(n_2\ge 1\) poles [2606.19479].

## 7. Finite-temperature supergravity saddles and the disappearance of the bound state

A direct gravitational derivation of the two-centered contribution to the index uses the Euclidean path integral with periodic fermions,
\[
Z_{\rm grav}(\beta,\Gamma)=
\mathrm{Tr}_\Gamma\bigl[(-1)^F e^{-\beta H}\bigr].
\]
Because of supersymmetry, the index localizes semiclassically onto finite-temperature BPS saddles in which the geometry caps off smoothly in the bulk at each center. For a two-center split \(\Gamma=\Gamma_1+\Gamma_2\), one introduces doubled poles
\[
\Gamma_i=\gamma_i+\widetilde\gamma_i,\qquad
\gamma_i=\frac12\Gamma_i+\delta_i,\qquad
\widetilde\gamma_i=\frac12\Gamma_i-\delta_i,
\]
with \(\delta_i\) fixed by the new attractor equations, and the harmonic function
\[
H(x)=h+\sum_{i=1}^2 \frac{\gamma_i}{|x-x_i|}+\sum_{i=1}^2\frac{\widetilde\gamma_i}{|x-\tilde x_i|}.
\]
The metric and one-form satisfy
\[
ds^2=\frac1{\Sigma(H)}(d\tau+\omega_E)^2+\Sigma(H)\,dx^i dx^i,
\qquad
d\omega_E=\langle dH,H\rangle,
\]
where
\[
\Sigma(H)=\langle H,\Omega_*(H)\rangle \,\langle H,\overline{\Omega}_*(H)\rangle.
\]
Smoothness imposes the regularity conditions
\[
\langle \gamma_i,H(x_i)\rangle=\frac{\beta}{4\pi},
\qquad
\langle \widetilde\gamma_i,H(\tilde x_i)\rangle=-\frac{\beta}{4\pi}.
\]
These are the finite-temperature analogues of the Denef integrability conditions [2507.07166].

For \(N=2\), the regularity conditions can be solved in closed form, leaving a single real modulus \(r=|x_1-x_2|\) in a finite interval \(r\in [r_-,r_+]\). In the extremal limit,
\[
r_*=-\frac{\langle \Gamma_1,\Gamma_2\rangle}{\langle \Gamma_1,h\rangle},
\]
which reproduces the Denef formula, and the full four-dimensional moduli space after quotienting translations has topology
\[
\mathcal M_2\cong S^1\times S^3.
\]
The on-shell action is
\[
I_{\rm saddle}(\beta)=\beta\,|Z(\Gamma;\Omega_\infty)|-\sum_{i=1}^2 S_{\rm ext}(\Gamma_i),
\]
so
\[
Z_{\rm grav}^{(2)}(\Gamma_1,\Gamma_2;t_\infty)\simeq
e^{-I_{\rm saddle}}
=
e^{-\beta |Z(\Gamma;\Omega_\infty)|}\,
e^{\sum_i \pi \langle \gamma_i,\widetilde\gamma_i\rangle }.
\]

As the asymptotic moduli vary, the interval \([r_-,r_+]\) shrinks and collapses when
\[
\langle \Gamma_1,h\rangle\to 0
\qquad \Longleftrightarrow \qquad
\Im\bigl(Z(\Gamma_1)\overline{Z}(\Gamma_2)\bigr)=0.
\]
At that point the two-center index saddle ceases to exist. The net jump in the index is then
\[
\Delta Z=
(-1)^{\langle \Gamma_1,\Gamma_2\rangle-1}
|\langle \Gamma_1,\Gamma_2\rangle|\,
Z_{\rm ext}(\Gamma_1)\,Z_{\rm ext}(\Gamma_2),
\]
which matches the standard primitive wall-crossing formula. A plausible implication is that the two-centered index admits a unified interpretation across microscopic partition functions, supersymmetric quantum mechanics, and finite-temperature supergravity saddles: in each description, the disappearance of the bound state is controlled by the same wall of marginal stability [2507.07166].

Source: https://www.emergentmind.com/topics/two-centered-black-hole-index