Can the OSV conjecture be proven with the gravitational path integral?

What the OSV relation actually equates

The OSV conjecture links two objects that are computed in very different ways. On one side is a helicity supertrace: a trace over black hole microstates in 4d N=2 supergravity from a Calabi–Yau compactification of Type IIA string theory, with bosonic and fermionic states entering with opposite signs and each state weighted by (−1)^F (J_3)^2. The ensemble fixes the magnetic charges Γ_I and the electric chemical potentials µ_I, and holds the vector multiplet scalar moduli fixed at infinity. On the other side is the modulus squared of the A-model topological string partition function, evaluated at the attractor values of the scalar moduli, which are themselves determined by those charges and chemical potentials.

Because the topological string partition function has a genus expansion in the topological string coupling g_top, the conjecture is a statement about every order in g_top, with each order fixed exactly in terms of the inverse string tension α′. The paper's target is therefore not a leading approximation but an all-orders statement. Two features of the trace matter for reading the result. Only BPS states contribute, and the supercharge anticommutator {Q, Q†} annihilates them, so the index is expected to be independent of β; neither side of the relation carries a β label. And because the trace is an index, it computes a difference between bosonic and fermionic degeneracies, not their sum — the quantity that a Lorentzian extremal black hole entropy would approximate.

Finite-temperature saddles replace extremal geometries

The proof rests on a change of gravitational object. Historically, the black hole partition function entering OSV was discussed using Lorentzian extremal geometries with infinite β. The authors instead work with regular supersymmetric Euclidean backgrounds whose asymptotic boundary is S¹_β × S² at finite inverse temperature β. This matters because the finite-β path integral computes the difference of degeneracies required by the indexed partition function, whereas the extremal entropy approximates the logarithm of their sum.

Two structural facts make these saddles tractable. First, the two-derivative Euclidean Poincaré supergravity solutions are supersymmetric off-shell backgrounds of conformal supergravity to all orders in the derivative expansion, so higher-derivative corrections encoded by the prepotential can be carried along. Second, the Killing vector built from the Killing spinor on these backgrounds has only isolated fixed points, called nuts. For a single supersymmetric black hole there are exactly two: the north and south poles of the horizon. That isolated-fixed-point structure is what allows the on-shell action to be computed by equivariant localization rather than by brute force.

Why the action splits into two pole terms

Equivariant localization reduces the on-shell action of these saddles to data at the fixed points of the Killing vector. For each black hole, the action localizes entirely to the north and south poles, and the two contributions are determined respectively by the holomorphic and anti-holomorphic topological string free energies evaluated at the attractor values of the moduli. The 4d supergravity higher-derivative prepotential is encoded in the topological string free energy on the Calabi–Yau, so this localization formula is exactly the bridge between the gravitational action and the topological string partition function.

The paper also reports which terms do and do not contribute. In 4d N=2 conformal supergravity, D-terms do not contribute to the action, while the F-term contribution is computed explicitly. The resulting on-shell action is independent of the moduli of multi-centered non-extremal saddles associated with the positions and relative orientation of the centers, and — up to wall-crossing phenomena where a saddle could stop contributing to the path integral — also independent of the choice of scalar moduli at infinity. For the unshifted single-center saddle, the two pole contributions combine into the squared modulus of the topological string partition function, which is the OSV formula itself.

The mixed ensemble matches Type IIA boundary data

The calculation is naturally performed in the mixed ensemble, and the paper argues this is not an accident. The type IIA supergravity action is formulated in terms of the 10d metric, dilaton, B₂ field, and the RR forms C⁽¹⁾ and C⁽³⁾, so it is natural to fix the values of C⁽¹⁾ and C⁽³⁾ at asymptotic infinity. Fixing those values corresponds to fixing the number of D6- and D4-branes, which couple magnetically to those forms, together with chemical potentials for D0- and D2-branes, which couple electrically.

The 9d boundary of the type IIA uplift of these saddles has the form S¹_β × S² × Y. For a general Calabi–Yau manifold Y there are no 1-cycles, so one must specify the holonomy of C⁽¹⁾ on S¹_β, and similarly the period of C⁽³⁾ on the h^{1,1}(Y) = n_v three-cycles S¹_β × C_A. The periods of C⁽³⁾ along the 2(h^{2,1}(Y)+1) three-cycles of Y set boundary conditions for D2-brane instantons, which do not contribute to the perturbative series. The authors conclude that the relevant type IIA path integral is specified by electric chemical potentials for D0- and D2-branes and by the number, or charge, of D4- and D6-branes — precisely the 4d supergravity mixed ensemble used in the analysis.

A saddle-by-saddle refinement beyond single centers

The same localization formula applies to every saddle that contributes to the helicity supertrace, not only the single-center one. Two classes matter. Shifted saddles arise because 4d Abelian gauge field configurations may require additional patches to quantize electric charges properly in the index. Multi-center black hole configurations contribute as well, and the paper states that such contributions can be non-perturbatively small, suppressed by e^{−#S} relative to the leading saddle, where S is the entropy of that saddle and # is a positive constant.

Together these give a non-perturbative refinement of OSV in Newton's constant that holds saddle by saddle. The refinement accounts for the quantization of electric charges and remains relevant when the single-center saddle is absent or subdominant. The paper's appendix makes the point concrete with an explicit two-derivative example: for a prepotential F⁽⁰⁾(X) = (X¹)³/X⁰ and magnetic charges Γ₁ = (0, 10L²), Γ₂ = (6L, 10L²), the two-black-hole on-shell action is 550πL⁵ while the competing single-black-hole action is 400πL⁵, so log(Z_2BH/Z_1BH) = 150πL⁵ > 0. The unshifted two-center saddle therefore dominates exponentially at large L, in a regime where g_top = O(L⁻¹) and the large-volume approximation is controlled.

What the proof does not yet cover

The derivation is at leading order in Newton's constant G_N, while it is all orders in g_top. Loop corrections around the Euclidean saddles are not included. The authors note that if the factorization underlying OSV extends to these quantum corrections, the one-loop determinant should likewise factorize into north- and south-pole contributions, and they point to the Atiyah–Bott fixed-point theorem and the equivariant index of the fluctuation complex as a route, with careful treatment of boundary conditions, boundary contributions, and zero modes still required.

A second scope condition concerns which saddles the path integral actually picks up. The paper computes the on-shell action of each saddle but explicitly does not determine which of the saddles are selected by the gravitational path integral. Wall-crossing phenomena, where a saddle could stop contributing, are likewise left as a qualification on the moduli-independence of the action rather than resolved. The proof should therefore be read as a statement about the action of each contributing saddle, not as a complete evaluation of the path integral measure.

Where the all-orders action pays off

The most distinctive payoff is for small black holes, where higher-derivative corrections are essential even at leading order. The all-orders expression for the on-shell action in terms of the topological string coupling lets the authors determine the small-black-hole contribution directly. Chen, Murthy, and Turiaci had conjectured a formula for the higher-derivative on-shell action and, assuming its temperature independence, evaluated it in the zero-temperature limit to obtain the leading microscopic index. The conformal supergravity and equivariant localization analysis supplies a direct finite-temperature derivation including both F-terms and D-terms, confirming that conjectured index result in the canonical ensemble and explaining how to include the small-black-hole contribution within the OSV relation in the mixed ensemble.

The paper also flags where the construction could extend: uplifting the 4d saddles on the M-theory circle and decompactifying it to reach 5d black holes, black rings, and black lenses; taking Γ⁰ = 0 and a low-temperature limit that scales the S¹_M size with the thermal circle to obtain asymptotically AdS₃ × S² saddles computing contributions to the elliptic genus of the dual MSW CFT; and asking whether an OSV-type relation exists for asymptotically AdS₄ black holes, where the link to topological strings is described as more tenuous. The priority for a reader is the single-center result: the north/south-pole factorization is what turns the OSV relation from a conjecture into a derivation to all orders in g_top, and the multi-center and shifted saddles are what turn it into a saddle-by-saddle statement rather than a single-saddle coincidence.

Sources used in this answer