Heterotic sigma models on $T^8$ and the Borcherds automorphic form $\Phi_{12}$

We consider the spectrum of BPS states of the heterotic sigma model with $(0,8)$ supersymmetry and $T^8$ target, as well as its second-quantized counterpart. We show that the counting function for such states is intimately related to Borcherds' automorphic form $\Phi_{12}$, a modular form which exhibits automorphy for $O(2,26;{\mathbb Z})$. We comment on possible implications for Umbral moonshine and theories of AdS$_3$ gravity.


Introduction
Studies of the entropy of supersymmetric black holes in string theory have led to the discovery of beautiful and unexpected relations between basic objects in the theory of automorphic forms, algebraic geometry, and indices of supersymmetric sigma models. A basic example is the formula of Dijkgraaf, Verlinde and Verlinde [1], capturing the degeneracy of 1/4-BPS dyons in the N = 4, d = 4 string theory obtained from compactifying type II string theory on K3 × T 2 or, equivalently, the heterotic string on T 6 . This theory has duality group SL(2, Z) × SO (6,22; Z), where the electric and magnetic charges are given by 28-dimensional vectors Q, P ∈ Γ 6,22 , and the SL(2, Z) factor is the electric-magnetic S-duality group of the theory.
The degeneracy of 1/4-BPS dyons, D(Q, P ), with charges (Q, P ) is then (−1) Q·P +1 D(Q, P ) = dΩ e πiTr (Ω·Λ) Φ 10 (Ω) where Ω = τ z z σ parametrizes the Siegel upper half-space of degree 2, and Λ = Q · Q Q · P Q · P P · P encapsulates the three T-duality invariants of the black hole charges. The beautiful function in the denominator of the integrand is the Igusa cusp form Φ 10 (Ω), a weight 10 Siegel automorphic form for the modular group SP (2, Z). For large charges this formula for the degeneracies has asymptotic growth reproducing the expected result for the Bekenstein-Hawking entropy of these black holes, The automorphic function Φ 10 also has a connection to elliptic genera of symmetric powers of the K3 surface, derived in [2] by considering the D1-D5 system on K3 × S 1 , which engineers 5D black holes. The end result is an elegant formula for the generating function, (Here, φ 10,1 is a named weak Jacobi form of weight 10 and index 1; explicit expressions can be found in [3]). Note the correction factor φ 10,1 spoils the automorphy of this function. In [4], a connection between 4D and 5D black holes [5] was used to explain the appearance of 1/φ 10,1 upon compactification and re-derive the result of [1].
The purpose of this note is to report analogous formulae governing BPS state counts in the heterotic sigma model with T 8 target. This is the model which would naturally arise on the worldsheet of heterotic string compactifications preserving half-maximal supersymmetry in two space-time dimensions. We will focus in this note on the physics of the 2d field theory and its supersymmetry-preserving excitations, and mostly limit any discussion of possible space-time interpretations to the concluding section. This work was largely motivated by trying to develop an understanding of BPS counts at Niemeier points in the moduli space of compactifications to 3 and 2 dimensions, and their possible interpretation in light of Mathieu and Umbral moonshine [6,7], in the picture advocated in [8].

The Borcherds modular form Φ 12
The hero of our story will be the Borcherds modular form Φ 12 [9]. A nice description of the relevant aspects of this form can be found in the work of Gritsenko [10], from which we borrow heavily.
Let Π 2,26 denote the (unique) even unimodular lattice of signature (2, 26). The Borcherds modular form Φ 12 is of weight 12 with respect to O + (Π 2,26 ). It is the unique cusp form with this property.
Of great interest for us will be the following explicit multiplicative lift formulae for Φ 12 . Consider the split with N ≡ N(R) chosen from among the 24 even unimodular positive-definite lattices in dimension 24, i.e., the 23 Niemeier lattices and the Leech lattice. 1 These lattices are uniquely classified by their root systems, R, which are unions of simply-lace ADE root systems of the same Coxeter number, and the Leech lattice is the unique 24-dimensional even positive-definite unimodular lattice with no roots. We will denote by h(R) the Coxeter number of the root system R associated with N, where we set h(R) = 1 when N is the Leech lattice.
To each possible choice of N, we can associate the refined lattice theta series Here, (v, w) denotes the lattice inner product of v, w ∈ N, and ξ ∈ N ⊗C can be thought of as a complex 24-vector of 'flavor' chemical potentials refining the lattice theta function.
Then the automorphic form Φ 12 (τ, ξ, σ), defined on can be constructed as a multiplicative lift of a conventional Jacobi form.
Define the Fourier coefficients f (n, λ) via This function is a weakly holomorphic Jacobi form of weight zero and index one for the lattice N. Then, Φ 12 is given by the formula p ≡ e 2πiσ , r λ = e 2πi(ξ,λ) .

In the above, we have defined
λ∈N f (0, λ)(λ, λ) 2 and used the notation (n, λ, m) > 0 to mean m > 0, or m = 0 and n > 0, or m = n = 0 and λ < 0. Furthermore, λ > 0 (or < 0) means that λ ∈ N has positive (respectively, negative) scalar product with a reference vector x ∈ N ⊗ R. The vector x must be chosen so that (x, λ) = 0 for all λ ∈ N, and different choices of x are related to one each other by automorphisms in O(Γ 2,26 ).
More precisely, the Niemeier and Leech points define cusps in the domain of definition of Φ 12 . These formulae should be thought of as expansions of the modular form around the cusps.
To make contact with the earlier work [8], it is useful to specialize the chemical potentials as follows. Choose a fixed lattice vector δ ∈ N. Then we can define with y = e 2πiz . One now obtains a Jacobi form of weight 0 and index (δ, δ)/2: These Jacobi forms, for suitable choices of δ, are the BPS counting functions discussed in [8]. That is, they control the coefficients in the expansion of a certain "F 4 " term in the low-energy effective action of heterotic string compactification to three dimensions, when the moduli are deformed a slight distance away from a point with Niemeier symmetry (the enhanced symmetry point itself having singular couplings).
A specialized form of Φ 12 can be obtained as a lift of these BPS counting functions as well: (1 − q n y l p m ) f δ (mn,l) with the prefactors A and C as above and B ≡ 1 2 λ>0 f (0, λ)(λ, δ). This object is an automorphic form on the Siegel upper half-space This structure also arises in the non-perturbative description of heterotic strings on T 7 [12].
One can think of M as parametrizing even unimodular lattices of signature (8, 24), Γ 8,24 (M). The worldsheet field theory at a given point in moduli space consists of 2d bosons propagating on the relevant lattice.
Let us consider this theory on a toroidal worldsheet. At any point in moduli space, there are 24 abelian currents of conformal dimension (1, 0). One can consider coupling these to background chemical potentials ("Wilson lines"). In this setup, the parameter τ of §2 can be considered as the modular parameter of the torus, while the 24 complex chemical potentials ξ should be thought of as these Wilson line degrees of freedom. The parameter p will emerge upon second-quantization. where N is again one of the 24 even unimodular positive-definite lattices, and Γ 8,0 the E 8 lattice. We will call this CFT V N .
One obtains BPS states in the heterotic sigma model by putting rightmovers in their ground state and performing a trace counting left-moving excitations. We restrict the counting to BPS states carrying no right-moving momentum, i.e. carrying zero charge with respect to the right-moving E 8 lattice: Tr V N (−1) F q L 0qL0 y ξ a ·Ja (with J a the 24 left-moving abelian currents). A subtlety arises because of the extra right-moving zero modes of this string background; without inserting extra factors of the right-moving fermion number operator F R in the trace, the computation will formally vanish due to the existence of fermion zero modes. One can circumvent this vanishing by inserting powers of F in the trace to soak up zero modes, as in [13]; however, this will alter the modularity properties and spoil the connection we intend to make to Φ 12 . Instead, we consider, following [14], a 'twisted' computation where we trace over leftmoving excitations but consider the ground states of the Z 2 orbifold of the right-moving sector -that is, we consider the right-movers to live in the Ramond sector of the Conway module constructed by Duncan [15]. 3 The orbifold action kills the extra zero modes, but leaves the left-moving sector untouched. The ground-state degeneracy of the right-movers gives an overall factor of 24, as in [14]. Including this degeneracy, the result (for a given N) is (whereṼ N is the twisted Hilbert space described above). Specializing chemical potentials as before by choosing a δ ∈ N, we see that the BPS counting function is F N,δ (τ, z).
We note here that the need to consider the twisted index computation to find a non-zero answer plays well with the structure of Φ 12 . While the heterotic compactification without the twist has a moduli space obtained by choosing left and right moving sub-lattices of Γ 8,24 , after twisting the E 8 lattice is rigidified. Therefore, compactifying on an additional T 2 to compute the index, one expects a moduli space based on Γ 2,26 . This makes it natural to expect a connection to an automorphic form for the Narain modular group O(2, 26; Z), such as Φ 12 .

Second quantization
Now, let us consider the second-quantized version of this counting function. To do this, consider the heterotic sigma model based on the conformal field theory Sym n (Ṽ N ). (Ṽ N appears because again, one can count BPS states without additional insertions to absorb fermion zero modes by quotienting by the natural Z 2 action on the E 8 lattice, before taking the symmetric product and putting right movers into their ground state). We still restrict ourselves to BPS states carrying zero charge with respect to the right-moving U(1) currents. Following the logic of Dijkgraaf, Moore, Verlinde and Verlinde [2], we see that .
Here, F Sym n (Ṽ N ) is the specialized BPS counting function for the CFT based on the nth symmetric product of the (twisted) heterotic sigma model, and where R + is the set of positive roots of the lattice N, and the sign depends on the choice of the set of positive roots. The factor of 24 again arises from the right-moving ground state degeneracy, as in [14].
We see that there is a precise analogy to the findings in [1,2]: just as an (automorphic correction of) 1 Φ 10 governs the BPS states of the sigma models on the Hilbert scheme of K3 surfaces, (an automorphic correction of) 1 Φ 12 governs the BPS states of the heterotic sigma models Sym n (Ṽ N ). The role of the factor ψ N,δ is superficially similar to that of the automorphic correction φ 10,1 in the former story. As mentioned earlier, this factor achieves an interpretation in the 4d/5d lift [4,5], and it would be interesting to give a precise similar interpretation to ψ N,δ .

Discussion
In this note, we've described how the (inverse of the) Borcherds modular form Φ 12 serves as a generating function for the BPS state degeneracies of heterotic sigma models on T 8 (after suitably twisting to kill right-moving zero modes), and their symmetric products. We conclude with several comments and possible avenues for further development.
• In light of AdS 3 /CFT 2 duality, it is natural to conjecture that these heterotic conformal field theories, Sym n (V N ), (at least at large central charge, i.e. large n) are dual to AdS 3 gravity theories with discrete symmetry groups corresponding to those of the associated Niemeier lattice -i.e. the Umbral groups and Conway's largest sporadic group. A criterion was developed in [16], using BPS degeneracies of 2d SCFTs to test for a possible large radius gravity dual. By modifying and checking this criterion for the case at hand, it has been found that these theories will have (at best) 'stringy' gravity duals -they will not achieve a (parametric) separation between the inverse AdS radius and M string as the central charge c → ∞ [17].
• The 2d heterotic compactifications governed by the CFTs V N are the dimensional reductions of the Niemeier points studied in [8] in relation to moonshine. It is tempting to try and connect Borcherds' modular form Φ 12 to Umbral moonshine. Making these thoughts precise is difficult because the notion of counting BPS states in gravity theories in d ≤ 3 flat dimensions is fraught with subtlety; charged or gravitating particles have strong infra-red effects in low dimensions. Possibly, finding an interpretation of the present formulae in the setting of the supersymmetry-protected amplitudes studied in [18] is a route forward.
• While the models V N depend on the choice of a Niemeier lattice N, the function counting the second quantized BPS states is the same in all cases. Indeed, these counting functions are just Fourier expansions at different cusps of the the same automorphic function Φ 12 . This is non-trivial: the points where the lattice Γ 8,24 splits into an E 8 lattice plus a Niemeier lattice are isolated points in the Narain moduli space of perturbative heterotic strings on T 8 . The basic reason behind this phenomenon is that these isolated points are the different decompactification limits in the Narain moduli space of heterotic strings on T 9 , with the condition that the Narain lattice Γ 9,25 splits as an orthogonal sum of E 8 and Γ 1,25 . This suggests that there might be an interpretation of Φ 12 in terms of compactification of the heterotic string to one dimension, possibly along the lines suggested in [14].
• It is known that 1/Φ 10 is related to the square of the denominator of a generalized Kac-Moody (GKM) algebra. There is a beautiful story relating this algebra to wall-crossing of 1/4-BPS dyons in N = 4, d = 4 string theory [19]. As we mentioned earlier, 1/Φ 12 is also the denominator of a GKM algebra-the fake monster Lie algebra. It would be interesting to explore a similar story relating this algebra to the BPS states in the theories we discuss.
• The Gromov-Witten theory of K3 × T 2 has recently been conjectured to be governed by the Igusa cusp form [20]. This follows from the role of Φ 10 in black hole entropy counts, and string duality. Given that the heterotic string on T 8 is dual to type IIA on K3 × T 4 , it is tempting to think that the appearance of Φ 12 in BPS counts in the present setting presages a similar role for Φ 12 in (flavored) enumerative geometry, perhaps of K3 × T 4 [21].