Resolved Elliptic Genus (REG)

• Resolved elliptic genus (REG) is a context-dependent refinement that produces finite, modular invariants from the otherwise divergent or trivial elliptic genera in non-compact and singular theories.
• It unifies various constructions—from the modular lattice sum in the cigar SCFT to supersymmetry-index refinements in the D1-D5 CFT and discrepancy-corrected resolutions in algebraic geometry.
• REG techniques extract physically and mathematically meaningful data by addressing issues like continuum contributions, superselection sectors, and orbifold dependencies through structured regularization.

Resolved elliptic genus (REG) denotes several related but non-identical constructions in supersymmetric field theory, conformal field theory, and algebraic geometry. In the sources considered here, the term is used for a fully modular expression for the non-compact elliptic genus of the $SL(2,\mathbb R)/U(1)$ cigar SCFT, for a one-parameter refinement of the modified elliptic genus in the D1-D5 CFT on $T^4$, and for a resolution-based or orbifold-based elliptic genus attached to singular varieties and GIT phases. A closely related regularization program for ALE spaces defines regularized nonequivariant limits of equivariant elliptic genera and produces weak Jacobi forms proportional to the K3 elliptic genus (Troost, 2017, Hughes et al., 18 Mar 2026, Libgober, 2017, Zhou, 2015).

1. Terminology and range of meanings

The expression “resolved elliptic genus” is not uniform across the literature. In the non-compact SCFT of the cigar, it refers to the fully resolved cigar elliptic genus at level $k$, written as a modular lattice sum whose non-holomorphicity encodes the continuum contribution. In the D1-D5 CFT, it refers to a new supersymmetry index $\mathcal E_{N,\tilde\jmath_2}(q,y)$ that refines the modified elliptic genus by a superselection label $\tilde\jmath_2$. In algebraic geometry, it refers to the elliptic genus computed on a resolution $\widetilde X$ together with discrepancy data $\sum a_k E_k$, or equivalently, under suitable hypotheses, to the corresponding orbifold elliptic genus of a quotient presentation (Troost, 2017, Hughes et al., 18 Mar 2026, Libgober, 2017).

The shared feature is not a single universal definition, but the use of additional structure to remove an obstruction present in the naive genus. In one setting the obstruction is divergence or mock-modular behavior from a continuous spectrum; in another it is the near-triviality of the modified elliptic genus; in another it is the dependence of singular spaces on the choice of presentation or resolution. This suggests that “resolved” is best understood as a context-dependent term indicating that the elliptic genus has been made finite, modularly well behaved, sector-sensitive, or birationally well defined.

A further source of terminological overlap is the use of “regularized elliptic genus” for ALE spaces. That construction proceeds by restricting an equivariant genus to a subtorus, subtracting a pole, and taking a nonequivariant limit. The paper treats this as a regularization rather than a resolution, but it belongs to the same broader family of attempts to extract a finite two-variable genus from a non-compact or singular situation (Zhou, 2015).

2. The cigar SCFT: modular lattice sum and sigma-model derivation

For the $SL(2,\mathbb R)/U(1)$ “cigar” SCFT, the elliptic genus provides the simplest non-compact example of a non-holomorphic, or mock, Jacobi form. In standard conventions $q=e^{2\pi i\tau}$ and $y=e^{2\pi i z}$, the fully resolved cigar elliptic genus at level $T^4$0 is written as

$T^4$1

Because the sum runs over the two-dimensional lattice $T^4$2 in $T^4$3, the formula is manifestly modular covariant under $T^4$4 and elliptic in $T^4$5 (Troost, 2017).

The sigma-model derivation starts from the cigar target-space metric and dilaton

$T^4$6

with $T^4$7. One first T-duals to the $T^4$8-orbifolded trumpet and then lifts to its infinite cover, so that $T^4$9 becomes non-compact. In the worldsheet path integral, the oscillators factor out with

$k$0

while the zero modes $k$1, together with the right-moving fermions, are treated explicitly. In the Ramond-Ramond trace the left-moving fermion zero modes are lifted by the $k$2-twist, and one inserts once the Christoffel-symbol coupling

$k$3

to soak up the right-moving zero modes (Troost, 2017).

Introducing a holonomy $k$4 for $k$5 along the torus time cycle, the classical solution $k$6 induces a Gaussian action in $k$7 and $k$8. The resulting zero-mode integral is

$k$9

with $\mathcal E_{N,\tilde\jmath_2}(q,y)$0 the “order of the cover.” Undoing the infinite cover by summing over worldsheet windings $\mathcal E_{N,\tilde\jmath_2}(q,y)$1 in $\mathcal E_{N,\tilde\jmath_2}(q,y)$2 shifts $\mathcal E_{N,\tilde\jmath_2}(q,y)$3, trades the $\mathcal E_{N,\tilde\jmath_2}(q,y)$4-integral for a Gaussian in $\mathcal E_{N,\tilde\jmath_2}(q,y)$5, and reproduces the modular lattice-sum formula. The significance of this derivation is that the resolved genus is obtained directly from the non-linear sigma model rather than only from abstract modular or representation-theoretic arguments.

3. Infrared regulation, supersymmetry, and the mock-modular anomaly

A central issue in the non-compact problem is the role of infrared regulation. For each fixed right-moving momentum $\mathcal E_{N,\tilde\jmath_2}(q,y)$6 on the asymptotic circle of the cigar, one obtains a one-dimensional $\mathcal E_{N,\tilde\jmath_2}(q,y)$7 supersymmetric quantum mechanics on the half-line $\mathcal E_{N,\tilde\jmath_2}(q,y)$8, with superpotential $\mathcal E_{N,\tilde\jmath_2}(q,y)$9 as $\tilde\jmath_2$0 and boundary at $\tilde\jmath_2$1. The weighted trace

$\tilde\jmath_2$2

is ill-defined over a continuum without an IR regulator (Troost, 2017).

Two natural regulators on an interval $\tilde\jmath_2$3 are identified. One uses a delta-function potential $\tilde\jmath_2$4 chosen so that the boundary conditions at $\tilde\jmath_2$5 preserve supersymmetry. The other imposes Dirichlet boundary conditions at $\tilde\jmath_2$6 for both bosonic and fermionic components; this breaks supersymmetry but respects the density-of-states interpretation. With the supersymmetry-preserving regulator, boson and fermion levels remain exactly paired for all $\tilde\jmath_2$7, so $\tilde\jmath_2$8 is $\tilde\jmath_2$9-independent and the continuum gives no non-holomorphic contribution. With the Dirichlet regulator, the bosonic and fermionic spectral densities differ by the phase shift

$\widetilde X$0

and one gets

$\widetilde X$1

In the full cigar elliptic genus one must sum over all $\widetilde X$2 sectors and require an $\widetilde X$3-invariant IR prescription. The analysis shows that modular covariance forces the analogue of the Dirichlet prescription in each $\widetilde X$4-sector. This breaks right-moving supersymmetry at the IR cutoff and produces the characteristic non-holomorphic, $\widetilde X$5-dependent completion piece. Equivalently, right-moving supersymmetry and modular invariance cannot both be preserved in regulating the trace. In this formulation, the mock-modular anomaly of the non-compact elliptic genus is the manifestation of a regulator-dependent supersymmetry-modularity tension rather than a merely formal defect (Troost, 2017).

The same framework clarifies the flat-space limit. As $\widetilde X$6, the cigar flattens to $\widetilde X$7, and the Gaussian kernel becomes a broad regulator with $\widetilde X$8. For finite $\widetilde X$9, the resulting expression is the $\sum a_k E_k$0 partition sum with an explicit IR regulator; as $\sum a_k E_k$1, the divergence of the flat-space genus reappears. One may then impose either a modular-covariant minimal subtraction, recovering the non-holomorphic regulator, or a purely holomorphic regulator such as a Weierstraß-type subtraction. In this sense the finite-$\sum a_k E_k$2 cigar REG is a physical resolution of the flat-space singularity and yields the canonical modular completion of the naive flat-space result.

4. The D1-D5 CFT: a superselection-refined supersymmetry index

In the D1-D5 CFT on $\sum a_k E_k$3, the standard modified elliptic genus at fixed strand number $\sum a_k E_k$4 is

$\sum a_k E_k$5

and it receives only right-Ramond-ground-state contributions. For $\sum a_k E_k$6, one has $\sum a_k E_k$7 for all $\sum a_k E_k$8 except the vacuum. The resolved elliptic genus is introduced as a one-parameter refinement,

$\sum a_k E_k$9

with generating function

$SL(2,\mathbb R)/U(1)$0

Operationally, the construction inserts a Kronecker-$SL(2,\mathbb R)/U(1)$1 projector onto $SL(2,\mathbb R)/U(1)$2-algebra sectors of spin $SL(2,\mathbb R)/U(1)$3 in the refined grand-canonical partition function (Hughes et al., 18 Mar 2026).

The formalism is based on a Schur-Weyl decomposition of the BPS Hilbert space. Writing $SL(2,\mathbb R)/U(1)$4 for the graded single-strand left-moving space and $SL(2,\mathbb R)/U(1)$5 for the fixed right-ground-state space of dimension $SL(2,\mathbb R)/U(1)$6, one has

$SL(2,\mathbb R)/U(1)$7

Schur-Weyl duality then gives

$SL(2,\mathbb R)/U(1)$8

so that $SL(2,\mathbb R)/U(1)$9-invariance leaves

$q=e^{2\pi i\tau}$0

Equivalently, the grand partition function splits into $q=e^{2\pi i\tau}$1-sectors indexed by Young diagrams obeying the $q=e^{2\pi i\tau}$2-hook condition (Hughes et al., 18 Mar 2026).

The superselection structure comes from the supercharge deformed by the exactly marginal operator $q=e^{2\pi i\tau}$3. In the free basis, one identifies the effective Gava-Narain operator

$q=e^{2\pi i\tau}$4

which measures the second-order anomalous dimension through

$q=e^{2\pi i\tau}$5

The operator $q=e^{2\pi i\tau}$6 commutes with all left-moving total modes and preserves the subalgebra

$q=e^{2\pi i\tau}$7

Its irreducible representations are labeled by two spins $q=e^{2\pi i\tau}$8, have dimension $q=e^{2\pi i\tau}$9, and are depicted as diamond diagrams of size $y=e^{2\pi i z}$0 on the $y=e^{2\pi i z}$1 lattice. The crucial superselection rule is that the deformed supercharge connects only $y=e^{2\pi i z}$2-multiplets with the same $y=e^{2\pi i z}$3 spin $y=e^{2\pi i z}$4; no mixing occurs between diamond diagrams of different height. This permits a separate index in each fixed $y=e^{2\pi i z}$5 sector (Hughes et al., 18 Mar 2026).

The explicit REG formula is obtained by decomposing each $y=e^{2\pi i z}$6-sector character $y=e^{2\pi i z}$7 into $y=e^{2\pi i z}$8-characters $y=e^{2\pi i z}$9 and summing only over those $T^4$00-sectors that contain the chosen $T^4$01. Below the black-hole threshold, where the MEG is essentially trivial, each REG sector has nonzero coefficients and exhibits perfect agreement with supergravity Kaluza-Klein modes split into the same $T^4$02 sectors. Above the threshold, each sector shows Cardy growth,

$T^4$03

and the leading-growth coefficient is numerically independent of $T^4$04, indicating that black-hole microstates are distributed roughly equally among these superselection sectors. The REG is therefore the first index in this setting that is sensitive to which BPS multiplets remain short or recombine under marginal deformation.

5. Resolution, orbifolds, and GIT phases in algebraic geometry

In the algebro-geometric literature, the resolved elliptic genus is defined for singular varieties through discrepancy data on a resolution. For a normal projective variety $T^4$05 that is $T^4$06-Gorenstein with at worst Kawamata-log-terminal singularities, choose a log-resolution

$T^4$07

with $T^4$08. The resolved elliptic genus is then

$T^4$09

Blow-up invariance implies that the push-forward of this class is independent of the chosen resolution, and one recovers the singular elliptic genus of $T^4$10. When $T^4$11 is a global quotient with $T^4$12 smooth and $T^4$13 finite, an orbifold elliptic genus can be defined by a twisted-sector sum over commuting pairs $T^4$14; the orbifold-McKay correspondence identifies this with the resolution-based genus whenever a crepant resolution exists (Libgober, 2017).

This framework extends naturally to Witten’s phases of $T^4$15 theories. One fixes a smooth quasi-projective variety $T^4$16 with an action of a reductive group $T^4$17, a $T^4$18-equivariantly linearized line bundle $T^4$19, and a chosen linearization $T^4$20. The corresponding phase is the GIT quotient

$T^4$21

As $T^4$22 varies across chambers in the ample cone, one obtains Landau-Ginzburg, hybrid, gauged LG, and Calabi-Yau phases. To define the elliptic genus of a phase, one assumes at worst klt singularities, a global quotient presentation by a finite abelian group or equivalently the existence of a crepant resolution, and $T^4$23-normality with respect to the residual $T^4$24-action (Libgober, 2014).

Under these hypotheses one defines an equivariant orbifold elliptic class on a smooth cover $T^4$25, with commuting actions of a finite abelian group $T^4$26 and a torus $T^4$27, by summing over common fixed loci $T^4$28 and inserting the appropriate theta-function ratios determined by eigenbundles of the tangent bundle and by invariant divisors. The elliptic genus of the phase is then obtained by specializing the equivariant parameter $T^4$29 to $T^4$30. In the Landau-Ginzburg phase this reproduces the classic early-1990s formula for LG elliptic genera; under wall crossing, equivariant McKay correspondence implies invariance of the elliptic genus across $T^4$31-equivalent phases, including LG/CY and hybrid/CY correspondences (Libgober, 2014).

A recurrent misconception is to view the geometric REG as only a choice of resolution. The point of the blow-up and push-forward theorems is precisely that the resulting class is independent of the chosen log-resolution once discrepancy data are included. In this setting, “resolved” refers not merely to computing on a smooth model, but to a birationally stable prescription compatible with singular and orbifold presentations.

6. ALE regularization and broader structural themes

For ALE spaces of type $T^4$32, the relevant object is called a regularized elliptic genus, but its construction is closely aligned with the broader REG theme. Let $T^4$33 be the minimal resolution of $T^4$34, equipped with the standard $T^4$35-action. If the tangent weights at the fixed point $T^4$36 are

$T^4$37

then the equivariant elliptic genus is

$T^4$38

As a function of $T^4$39, this has first-order poles, and the naive nonequivariant limit does not exist (Zhou, 2015).

The regularization restricts to the one-parameter subtorus preserving the holomorphic volume form,

$T^4$40

for which

$T^4$41

As $T^4$42, one has a Laurent expansion

$T^4$43

with leading singular term

$T^4$44

Subtracting the pole part and taking the limit defines

$T^4$45

This is a weak Jacobi form of weight $T^4$46 and index $T^4$47, hence a multiple of the elliptic genus of a K3 surface, with proportionality constant $T^4$48 (Zhou, 2015).

Placed beside the other REG constructions, the ALE case makes the broader pattern especially clear. One removes poles by a controlled subtraction, just as the cigar construction resolves the flat-space singularity by a modularly covariant regulator, the D1-D5 construction resolves the triviality of the modified elliptic genus by sector decomposition, and the algebro-geometric construction resolves singular spaces by discrepancy-corrected resolution or orbifold formulas. The terminology is therefore context-sensitive, but the persistent mathematical theme is the extraction of a finite and structurally meaningful elliptic genus from a setting in which the naive object is singular, incomplete, or too coarse.