See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/324182069

Permutation Orbifolds of the Heisenberg Vertex Algebra $\mathcal{H}(3)

Article in Journal of Mathematical Physics · April 2018

DOI: 10.1063/1.5045164

CITATIONS READS
7 271

3 authors, including:

Michael Penn
Randolph College
19 PUBLICATIONS 170 CITATIONS

SEE PROFILE

All content following this page was uploaded by Michael Penn on 28 April 2019.

The user has requested enhancement of the downloaded file.

 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX
ALGEBRA H(3)

ANTUN MILAS, MICHAEL PENN, HANBO SHAO

Abstract. We study the S3 -orbifold of a rank three Heisenberg vertex algebra in terms
of generators and relations. By using invariant theory, we prove that the orbifold algebra
has a minimal strong generating set of vectors whose conformal weights are 1, 2, 3, 4, 5, 62
(two generators of degree 6). The structure of the cyclic Z3 -oribifold is determined by
similar methods. We also study characters of modules for the orbifold algebra.

1. Introduction
n
For every vertex operator algebra V , the n-fold tensor product V ⊗ has a natural vertex
operator algebra structure. The symmetric group Sn acts on V (n) := V ⊗n by permuting
tensor factors and thus Sn ⊂ Aut(V (n)). Denote by V (n)Sn the fixed point vertex operator
subalgebra also called the Sn -orbifold of V (n). It is an open problem to classify irreducible
modules of V (n)Sn although it is widely believed that every such module should come
from a g-twisted V ⊗n -module for some g ∈ Sn . When it comes to the inner structure of
V (n)Sn (e.g., a strong system of generators) very little is known. Even for the Heisenberg
orbifold H(n)Sn , where H is the rank one Heisenberg vertex algebra, this problem seems
quite difficult.
Finite and permutation orbifolds have been extensively studied both in physics and math-
ematics literature. According to [5], the earliest study of permutation orbifolds seems to
be [17]. The first systematic construction of cyclic orbifolds, including their twisted sectors
appeared in [8]. Explicit formulas for characters and modular transformation properties of
permutation orbifolds of Rational Conformal Field Theories were given in [4].
In the literature on vertex algebras, the main focus has been on the classification and
construction of twisted modules starting with [15, 19]. Further developments include “Quan-
tum Galois” theory developed in [11], work on twisted sectors of permutation orbifolds [6]
(see also [7]), tensor category structure for general G-orbifolds of rational vertex algebras
[13, 18], etc. There are also numerous papers on orbifold vertex algebras for abelian groups
and groups of small order. Structure and representations of 2-permutation orbifolds were
subjects of [1, 2, 12]; see also a more recent work on 3-permutation orbifolds of lattice vertex
algebras [14]. The ADE orbifolds of rank one lattice vertex algebras (e.g. [10]) are impor-
tant for classification of c = 1 rational vertex algebras. Orbifolds of irrational C2 -cofinite
vertex algebras have been investigated in [3], [9] and [2]. A. Linshaw (see [9, 20, 21, 22],
etc.) extensively studied orbifolds of “free field” vertex operator (super)algebras using the
classical invariant theory [26] and its (super) extensions. As a consequence of the main
result in [22], every finite orbifold of an affine vertex algebra is finitely strongly generated.
In particular, this implies that H(n)Sn has a finite strong set of generators. Characters of
orbifolds of affine vertex algebras were investigated earlier in [16].
1
2 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

In this paper we are concerned with the structure of one of the simplest non-abelian
orbifolds, H(3)S3 where H is the rank one Heisenberg vertex algebra (case H(2)S2 is well-
understood [12, 1]). We prove three main results. The first result pertains to generators of
H(3)S3 . We show in Theorem 4.1 that this vertex algebra is isomorphic to a W -algebra
of type (2, 3, 4, 5, 62 ) tensored with a rank one Heisenberg vertex algebra. Here labels
2, 3, 4, 5, 62 indicate that our W -algebra is strongly generated by the Virasoro vector (of
degree two) and five primary vectors of degrees 3, 4, 5 and two of degree 6. These five gen-
erators are explicitly given in Section 4 where we denoted them by J1 , J2 , C1 , C2 and C3 .
Our second result is about the cyclic Z3 -orbifold of H(3) (see Theorem 5.1 for details).
In the last part (Sections 6 and 7) we discuss characters of certain H(3)S3 -modules and
their modular properties. It is expected that many irrational vertex algebras will enjoy mod-
ular invariance in a generalized sense involving iterated integrals instead of sums. For the
rank n permutation orbifold H(n)Sn , the character of a module M is expected to transform
as
  X n Z
1
(1.1) ch[M ] − = SM,Mλi ch[Mλi ](τ )dλi ,
τ Ri
i≥1

where SM,Mλi ∈ C and λi ∈ Ri , 1 ≤ i ≤ n parametrize certain H(n)Sn -modules. Our third

main result gives strong evidence for this conjecture for n = 3 (see Theorem 7.1).

2. The Sn -orbifold of the Heisenberg vertex operator algebra

Let H denote the rank one Heisenberg vertex operator algebra generated by α(−1)1, with
n
the usual conformal vector (and grading) given by ω = 21 α2 (−1)1. Let H(n) = H⊗ . For
convenience, we suppress the tensor product symbol and let α1 (−1)1 := α(−1)1 ⊗ · · · ⊗ 1 ∈
H(n), and similarly we define αi (−1)1, i ≥ 2, such that H(n) = hα1 (−1)1, · · · , αn (−1)1i.
We consider the natural action of Sn on H(n) given by
(2.1) σ · αi1 (m1 ) · · · αik (mk )1 = ασ(i1 ) (m1 ) · · · ασ(ik ) (mk )1,
for 1 ≤ ij ≤ n, mj < 0, and σ ∈ Sn .
We clearly have a natural linear isomorphism
(2.2) H(n) ∼
= C[xi (m)|1 ≤ i ≤ n, m ≥ 0]
induced by αi (−m − 1) 7→ xi (m) for m ≥ 0. Using the terminology of [21], we say that
C[xi (m)|1 ≤ i ≤ n, m ≥ 0] is the associated graded algebra of the vertex algebra H(n).
Further, we may endow the polynomial algebra C[xi (m)|1 ≤ i ≤ n, m ≥ 0] with the structure
of a ∂-ring by defining the map
∂ : C[xi (m)|1 ≤ i ≤ n, m ≥ 0] → C[xi (m)|1 ≤ i ≤ n, m ≥ 0]
(2.3)
xi (m) 7→ (m + 1)xi (m + 1),
where the action of ∂ is extended to the whole space via the Leibniz rule. This definition of
∂ is compatible with the translation operator in H(n) given by T (v) = v−2 1.
The following Lemma is from [20].

Lemma 2.1. Let A be a vertex algebra with a Z≥0 filtration, where à is the associated
∂-ring. If {ãi |i ∈ I} generates à then {ai |i ∈ I} strongly generates A, where ai and ãi are
related via the natural linear isomorphism described by the Z≥0 filtration.
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 3

Further, we recall that the invariant ring C[x1 , . . . , xn ]Sn has a variety of generating sets,
including the power sum polynomials
pi = xi1 + · · · + xin , 1 ≤ i ≤ n.
In addition to this, it is common to study the invariant theory of the ring of infinitely
many commuting copies of this polynomial algebra, where we denote by xi (m) the copy
of xi from the mth copy of the polynomial algebra. A Theorem of Weyl [26] shows that
C[xi (m)|1 ≤ i ≤ n, m ≥ 0]Sn is generated by the polarizations of these polynomials
n
X
(2.4) qk (m1 , . . . , mk ) = xi (m1 ) · · · xi (mk ),
i=1

for 1 ≤ k ≤ n. Now, applying Lemma 2.1, we have an initial strong generating set for the
orbifold H(n)Sn given by the vectors
n
X
(2.5) ωk (m1 , . . . , mk ) = αi (−1 − m1 ) · · · αi (−1 − mk )1,
i=1

for 1 ≤ k ≤ n and mj ≥ 0. It should be noted that the conformal vector of H(n) is 21 ω2 (0, 0),
making the orbifold a vertex operator algebra.

3. Warmup: H(2)S2
Here we describe the structure of H(2)S2 . This case is well-known, so we only provide
a few details. Denote by M (1)+ := H(1)Z2 the fixed point subalgebra under the action
α(−1)1 → −α(−1)1, where α is the Heisenberg generator (in the physics literature M (1)+ is
often denoted by W (2, 4)). This vertex algebra was thoroughly studied in [12] and elsewhere.
Let
h = α1 (−1)1 + α2 (−1)1, h⊥ = α1 (−1)1 − α2 (−1)1,
1 1 1
ω = α12 (−1)1 + α22 (−1)1 − α1 (−1)α2 (−1)1.
4 4 2
Observe that as vertex algebras
H(2) = H(1)h ⊗ H(1)h⊥
where H(1)h = hhi and H(1)h⊥ = hh⊥ i with conformal vector ω. The nontrivial element of
the group S2 fixes the first tensor factor and
h⊥ → −h⊥ .
Thus we immediately get
(3.1) H(2)S2 ∼
= H(1) ⊗ M (1)+ .
It is easy to see that

1
H= α1 (−1)4 +α2 (−1)4 −4α13 (−1)α2 (−1)−4α1 (−1)α23 (−1)+6α12 (−1)α22 (−1)+3α12 (−2)+3α22 (−2)
2

−6α1 (−2)α2 (−2)+2α1 (−1)α2 (−3)+2α1 (−3)α2 (−1)−2α1 (−3)α1 (−1)−2α2 (−3)α2 (−1) 1

is contained in hh⊥ i and is primary of conformal weight 4, and thus a generator of M (1)+ .
4 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

4. The Rank Three Case

4.1. The orbifold H(3)S3 . As described above we know that the orbifold H(3)S3 will be
strongly generated by the vectors
3
X
ω1 (a) = αi (−1 − a)1,
i=1
3
X
(4.1) ω2 (a, b) = αi (−1 − a)αi (−1 − b)1,
i=1
3
X
ω3 (a, b, c) = αi (−1 − a)αi (−1 − b)αi (−1 − c)1.
i=1

The following change of basis of the generating set will allow for an efficient reduction in
the generating set of the corresponding orbifold
1
β1 (−1)1 = √ (α1 (−1)1 + α2 (−1)1 + α3 (−1)1),
3
1
(4.2) β2 (−1)1 = √ (α1 (−1)1 + η 2 α2 (−1)1 + ηα3 (−1)1),
3
1
β3 (−1)1 = √ (α1 (−1)1 + ηα2 (−1)1 + η 2 α3 (−1)1),
3
where η is a primitive third root of unity. Using this generating set we have σ · β1 (−1)1 =
β1 (−1)1 for all σ ∈ S3 . Further, examining the action of the generators of S3 on the
generators of H(3) we have



(4.3) 2 3 · β2 (−1)1 = β3 (−1)1, 2 3 · β3 (−1)1 = β2 (−1)1,

and

3 · β3 (−1)1 = η 2 β3 (−1)1.



(4.4) 1 2 3 · β2 (−1)1 = ηβ2 (−1)1, 1 2

From this action, we see that an initial generating set for the orbifold H(3)S3 may be taken
to be
(4.5)
ω10 (a) = β1 (−1 − a)1,
ω20 (a, b) = β2 (−1 − a)β3 (−1 − b)1 + β3 (−1 − a)β2 (−1 − b)1,
ω30 (a, b, c) = β2 (−1 − a)β2 (−1 − b)β2 (−1 − c)1 + β3 (−1 − a)β3 (−1 − b)β3 (−1 − c)1,

for 0 ≤ a ≤ b ≤ c. In fact, we can explicitly write the relation between our original
generators (4.1) and our new generators (4.5) as follows
√
ω1 (a) = 3ω10 (a),
ω2 (a, b) = ω20 (a, b) + ω10 (a)−1 ω10 (b),
(4.6) 1 1
ω3 (a, b, c) = ω30 (a, b, c) + √ ω10 (a)−1 ω20 (b, c) + ω10 (b)−1 ω20 (a, c)
3 3
+ω1 (c)−1 ω20 (a, b) + ω10 (a)−1 ω10 (b)−1 ω10 (c) .
0


 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 5

We may make a similar change of variables for the generating set of C[xi (m)|1 ≤ i ≤
3, m ≥ 0] by setting
1
y1 (m1 ) = √ (x1 (m1 ) + x2 (m1 ) + x3 (m1 )),
3
1
(4.7) y2 (m2 ) = √ (x1 (m2 ) + η 2 x2 (m2 ) + ηx3 (m2 )),
3
1
y3 (m3 ) = √ (x1 (m3 ) + ηx2 (m3 ) + ηx3 (m3 )),
3
for mi ≥ 0, and
q10 (a) = y1 (a),
(4.8) q20 (a, b) = y2 (a)y3 (b) + y3 (a)y2 (b),
q30 (a, b, c) = y2 (a)y2 (b)y2 (c) + y3 (a)y3 (b)y3 (c),
for a, b, c ≥ 0.
Using the translation operator, T , restricted to the orbifold H(3)S3 , the initial strong
generating set (4.5) can be reduced per the following Lemma.
Lemma 4.1. The orbifold H(3)S3 is strongly generated by the vectors
ω10 (0),
(4.9) ω20 (0, 2a) for a ≥ 0,
ω30 (0, a, b) for 0 ≤ a ≤ b.
Proof. It is clear that since
1 0
(4.10) ω10 (a)−1 1 =
ω (0)−1−a 1
a! 1
the linear portion of the generating set (4.5) can be immediately minimized to contain the
single weight one element ω1 (0).
Moving on toward the quadratic terms in (4.5), we consider the vector spaces
(4.11) A2 (m) = span {ω20 (a, b)|a + b = m},
and a natural family of subspaces
(4.12) ∂A2 (m) = {v−2 1|v ∈ A2 (m)}.
Observe that
(4.13)
∂
ω20 (a, b)−2 1 = Res z −1 (Y (β2 (−1 − a)β3 (−1 − b)1, z) + Y (β3 (−1 − a)β2 (−1 − b)1, z))
∂z
 a
∂b ∂a ∂b

1 −1 ◦ ∂ ◦ ◦ ◦
= Res z ◦ β2 (z) b β3 (z)◦ + ◦ a β3 (z) b β2 (z)◦
a!b! ∂z a ∂z ∂z ∂z
= (a + 1)ω20 (a + 1, b) + (b + 1)ω20 (a, b + 1),
where βi (z) = Y (βi (−1)1, z). Also notice that for all a, b ≥ 0 we have ω20 (a, b) = ω20 (b, a).
We may take an initial basis for A2 (m) to be the set
n j m ko
(4.14) B2 (m) = ω20 (i, m − i)|0 ≤ i ≤
2
which implies that
(4.15) dim A2 (2n + 1) = dim A2 (2n) = n + 1.
6 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

Further the set

(4.16) {v−2 1|v ∈ B2 (m)}

is clearly a basis for ∂A2 (m). It follows that

A2 (2n) = ∂A2 (2n − 1) ⊕ Cω20 (0, 2n),

(4.17)
A2 (2n + 1) = ∂A2 (2n)

and thus by induction we may take a more convenient basis of A2 (m) to be

n j m ko
(4.18) ω20 (0, 2i)−1−m+2i 1|0 ≤ i ≤ .
2
Thus, a more efficient set of quadratic generators for the H(3)S3 may be taken to be ω20 (0, 2a)
for a ≥ 0.
Finally, we consider the cubic generators. Throughout, we use the fact that ω30 (a, b, c) is
invariant under any permutation of the entries a, b, c. Analogous to (4.13) we have

(4.19) ω30 (a, b, c)−2 1 = (a + 1)ω30 (a + 1, b, c) + (b + 1)ω30 (a, b + 1, c) + (c + 1)ω30 (a, b, c + 1),

which may be used inductively to write all cubic generators in terms of those of the form
ω30 (0, a, b) with 0 ≤ a ≤ b.


Remark 4.1. Using (4.6) we may take

ω1 (0),
(4.20) ω2 (0, 2a) for a ≥ 0,
ω3 (0, a, b) for 0 ≤ a ≤ b,

as our strong generating set. We take advantage of this translation tool as generators of
this form are somewhat more natural to the parent algebra, H(3).

Remark 4.2. This simplification of the generating set also holds if we consider the ∂-ring
associated to H(3)S3 , which has un-reduced generators given by (2.4). We may reduce these
to a minimal generating set for C[xi (m)|1 ≤ i ≤ 3, m ≥ 0]S3 given by
q1 (0),
(4.21) q2 (0, 2a) for a ≥ 0,
q3 (0, a, b) for 0 ≤ a ≤ b.

We now present the following relations among the generators of C[xi (m)|1 ≤ i ≤ 3, m ≥
0]S3 .

Lemma 4.2. For a = (a1 , a2 , a3 , a4 , a5 , a6 ) with a1 , a2 , a3 , a4 , a5 , a6 ≥ 0 we have

D6C,1 (a) = q20 (a1 , a2 )q20 (a3 , a4 )q20 (a5 , a6 ) − q20 (a1 , a2 )q20 (a3 , a6 )q20 (a4 , a5 )
+ q20 (a1 , a4 )q20 (a2 , a6 )q20 (a3 , a5 ) − q20 (a1 , a4 )q20 (a2 , a3 )q20 (a5 , a6 )
(4.22)
+ q20 (a1 , a5 )q20 (a2 , a4 )q20 (a3 , a6 ) − q20 (a1 , a5 )q20 (a2 , a6 )q20 (a3 , a4 )
+ q20 (a1 , a6 )q20 (a2 , a3 )q20 (a4 , a5 ) − q20 (a1 , a6 )q20 (a2 , a4 )q20 (a3 , a5 ) = 0,
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 7

D6C,2 (a) = q30 (a1 , a2 , a3 )q30 (a4 , a5 , a6 ) − q30 (a1 , a2 , a4 )q30 (a3 , a5 , a6 )
1 1
(4.23) + q20 (a1 , a3 )q20 (a2 , a6 )q20 (a4 , a5 ) − q20 (a1 , a4 )q20 (a2 , a5 )q20 (a3 , a6 )
2 2
1 1
+ q20 (a1 , a5 )q20 (a2 , a3 )q20 (a4 , a6 ) − q20 (a1 , a5 )q20 (a2 , a4 )q20 (a3 , a6 ) = 0,
2 2
and
D5C (a) = q20 (a1 , a2 )q30 (a3 , a4 , a5 ) − q20 (a1 , a5 )q30 (a2 , a3 , a4 )
(4.24) − q20 (a2 , a5 )q30 (a1 , a3 , a4 ) − q20 (a3 , a4 )q30 (a1 , a2 , a5 )
+ q20 (a3 , a5 )q30 (a1 , a2 , a4 ) + q20 (a4 , a5 )q30 (a1 , a2 , a3 ) = 0,
for a = (a1 , a2 , a3 , a4 , a5 ) with a1 , a2 , a3 , a4 , a5 ≥ 0.
Remark 4.3. These relations play the role of the determinant (and similar) relations in
the invariant theory of classical Lie groups. We do not claim that these two families of
expressions generate all relations in this case. In fact, there are such relations at every
degree, and since the Frobenius number of 5 and 6 is 19, there are new relations at least up
to this degree.
We identify the relations from Lemma 4.2 via the isomorphism (2.2) with expressions in-
volving the generators of the orbifold H(3)S3 . For a = (a1 , a2 , a3 , a4 , a5 , a6 ) with a1 , a2 , a3 , a4 , a5 , a6 ≥
0 set
D61 (a) = ω20 (a1 , a2 )−1 ω20 (a3 , a4 )−1 ω20 (a5 , a6 ) − ω20 (a1 , a2 )−1 ω20 (a3 , a6 )−1 ω20 (a4 , a5 )
+ ω20 (a1 , a4 )−1 ω20 (a2 , a6 )−1 ω20 (a3 , a5 ) − ω20 (a1 , a4 )−1 ω20 (a2 , a3 )−1 ω20 (a5 , a6 )
(4.25)
+ ω20 (a1 , a5 )−1 ω20 (a2 , a4 )−1 ω20 (a3 , a6 ) − ω20 (a1 , a5 )−1 ω20 (a2 , a6 )−1 ω20 (a3 , a4 )
+ ω20 (a1 , a6 )−1 ω20 (a2 , a3 )−1 ω20 (a4 , a5 ) − ω20 (a1 , a6 )−1 ω20 (a2 , a4 )−1 ω20 (a3 , a5 ),

(4.26)
D62 (a) = ω30 (a1 , a2 , a3 )−1 ω30 (a4 , a5 , a6 ) − ω30 (a1 , a2 , a4 )−1 ω30 (a3 , a5 , a6 )
1 1
+ ω20 (a1 , a3 )−1 ω20 (a2 , a6 )−1 ω20 (a4 , a5 ) − ω20 (a1 , a4 )−1 ω20 (a2 , a5 )−1 ω20 (a3 , a6 )
2 2
1 0 1
+ ω2 (a1 , a5 )−1 ω2 (a2 , a3 )−1 ω2 (a4 , a6 ) − ω20 (a1 , a5 )−1 ω20 (a2 , a4 )−1 ω20 (a3 , a6 ),
0 0
2 2
and
D5 (a) = ω20 (a1 , a2 )−1 ω30 (a3 , a4 , a5 ) − ω20 (a1 , a5 )−1 ω30 (a2 , a3 , a4 )
(4.27) − ω20 (a2 , a5 )−1 ω30 (a1 , a3 , a4 ) − ω20 (a3 , a4 )−1 ω30 (a1 , a2 , a5 ) ,
+ ω20 (a3 , a5 )−1 ω30 (a1 , a2 , a4 ) + ω20 (a4 , a5 )−1 ω30 (a1 , a2 , a3 ),
for a = (a1 , a2 , a3 , a4 , a5 ) with a1 , a2 , a3 , a4 , a5 ≥ 0.
Observe that for i ∈ {1, 2}, the weight of the expression D6i (a) is |a| + 6 while the weight
of D5 (a) is |a| + 5, where we take |(u1 , . . . , un )| = u1 + · · · + un for any multi-index. By
Lemma 4.2 and repeated applications of the weak associativity properties of vertex algebras,
along with our linear isomorphism (2.2) we see that for i ∈ {1, 2}, we may rewrite
(4) (2)
(4.28) D6i (a) = D6 (a) + D6 (a),
8 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

where
(4) (4)
X
(4.29) D6 (a) = µa,b,c,d ω20 (a, b)−1 ω20 (c, d)
a,b,c,d≥0
a+b+c+d=|a|+2

and
(2) (2)
X
(4.30) D6 (a) = µa,b ω20 (a, b),
a,b≥0
a+b=|a|+4

(4) (2)
where µa,b,c,d and µa,b are appropriate constants. Importantly, the right hand side of (4.29)
has no terms that are “cubic” in the generators ω20 (a, b) or “quadratic” in the generators
ω30 (a, b, c), each of which would contain a combination of six of the original βi vectors with
i ∈ {2, 3}. We have a similar (and simpler) decomposition
(3)
(4.31) D5 (a) = D5 (a),
where
(3) (3)
X
(4.32) D5 (a) = µa,b,c ω30 (a, b, c),
a,b,c≥0
a+b+c=|a|+2

(3)
where µa,b,c are constants. Again, Lemma 4.2, weak associativity, and (2.2) implies that
the right hand side of (4.31) will not contain terms that are “products” of the generators
ω20 (a, b) and ω30 (a, b, c) or otherwise a combination of five of the original vectors β2 and β3 .
Now we are poised to use the expressions D6i (a) and D5 (a) to further reduce the gener-
ating set of H(3)S3 described in Lemma 4.1.
Lemma 4.3. The full list of quadratic generators described in Lemma 4.1 can be replaced
with the set {ω20 (0, 0), ω20 (0, 2), ω20 (0, 4)}.
Proof. Theorem 4.7 of [21] implies that, using only combination of quadratic generators, we
may immediately reduce our quadratic generators to the set
(4.33) {ω20 (0, 0), ω20 (0, 2), ω20 (0, 4), ω20 (0, 6), ω20 (0, 8)}
In order to remove the remaining two quadratic generators we use decoupling relations
that involve both quadratic and cubic generators. Using the expression D61 (0, 0, 0, 0, 1, 1),
the decomposition described in (4.28-4.30) may be used to construct the following equation
(4.34)
143 0 81 0 743 0 1
ω20 (0, 6) = ω (0, 4)−3 1 − ω (0, 2)−5 1 + ω (0, 0)−7 1 + ω 0 (0, 0)−1 ω20 (0, 4)
742 2 371 2 1113 2 1484 2
13 0 1 2
− ω (0, 0)−1 ω20 (1, 3) − ω 0 (0, 0)−1 ω20 (2, 2) + ω 0 (0, 1)−1 ω20 (0, 3)
2968 2 4452 2 1113 2
1 1 3
− ω 0 (0, 1)−1 ω20 (1, 2) + ω 0 (0, 2)−1 ω20 (0, 2) − ω 0 (0, 2)−1 ω20 (1, 1)
1113 2 2226 2 1484 2
1 1
− ω 0 (1, 1)−1 ω20 (1, 1) + ω 0 (0, 0)−1 ω20 (0, 0)−1 ω20 (0, 1)
2968 2 4452 2
1 1
− ω 0 (0, 0)−1 ω20 (0, 1)−1 ω20 (0, 1) + ω 0 (0, 0, 0)−1 ω30 (0, 1, 1)
8904 2 4452 3
1
− ω 0 (0, 0, 1)−1 ω30 (0, 0, 1),
4452 3
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 9

and thus we may remove ω20 (0, 6) from (4.33). Using (4.34) and D61 (0, 0, 0, 0, 2, 2) we can
similarly write ω20 (0, 8) as a vertex algebraic polynomial using the quadratic generators
ω20 (0, 0), ω20 (0, 2), ω20 (0, 4), along with cubic generators ω30 (0, 0, 0), ω30 (0, 0, 1), ω30 (0, 1, 1),
ω30 (0, 0, 2), and ω30 (0, 2, 2).


Lemma 4.4. The full list of cubic generators described in Lemma 4.1 can be be replaced
with the set
(4.35) {ω30 (0, 0, 0), ω30 (0, 0, 2), ω30 (0, 1, 2)}.
Moreover, these generators together with ω10 (0), ω20 (0, 0), ω20 (0, 2), ω20 (0, 4) form a strong gen-
erating set of H(3)S3 .

Proof. To get started we notice that

1
ω30 (0, 0, 1) = ω30 (0, 0, 0)−2 1
3
2
(4.36) ω30 (0, 1, 1) = −ω30 (0, 0, 2) + ω30 (0, 0, 0)−3 1
3
2 1
ω30 (0, 0, 3) = − ω30 (0, 1, 2) + ω30 (0, 0, 2)−2 1,
3 3
in other words, all cubic generators of conformal weight 6 or less may be written in terms of
the generators{ω30 (0, 0, 0), ω30 (0, 0, 2), ω30 (0, 1, 2)} using only the translation operator. Fur-
thermore, we have

(4.37)
1 0 1 1
ω30 (0, 1, 3) = − ω (0, 0, 4) − ω30 (0, 1, 2)−2 1 + ω30 (0, 0, 2)−3 1,
12 3 72 72
4 1 1
ω30 (0, 2, 2) = 3ω30 (0, 0, 4) + ω30 (0, 1, 2)−2 1 − ω30 (0, 0, 2)−3 1 − ω30 (0, 0, 0)−5 1,
3 3 3
1 1 1 1
ω30 (0, 2, 3) = − ω30 (0, 1, 4) + ω30 (0, 0, 4)−2 1 + ω30 (0, 1, 2)−3 1 − ω30 (0, 0, 0)−6 1,
30 60 35 36
5 1 1 1
ω30 (0, 3, 3) = ω30 (0, 0, 6) − ω30 (0, 2, 4) + ω30 (0, 1, 4)−2 1 + ω30 (0, 0, 4)−3 1
4 2 5 20
1 1 1
+ ω30 (0, 1, 2)−4 1 − ω30 (0, 0, 2)−5 1 − ω30 (0, 0, 0)−7 1,
6 12 4
5 15 1 2
ω30 (0, 3, 4) = ω30 (0, 1, 6) + ω30 (0, 0, 6)−2 1 − ω30 (0, 2, 4)−2 1 + ω30 (0, 1, 4)−3 1
7 7 3 3
5 5 10
+ ω30 (0, 1, 2)−5 1 − ω30 (0, 0, 2)−6 − ω30 (0, 0, 0)−8 1,
9 9 9
8 29
ω30 (0, 4, 4) = −21ω30 (0, 0, 8) + 4ω30 (0, 2, 6) − ω30 (0, 1, 6)−2 1 + ω30 (0, 0, 6)−3 1
7 7
2 0 8 0 1 0
− ω3 (0, 2, 4)−3 1 + ω3 (0, 1, 4)−4 1 − ω3 (0, 0, 4)−5 1
3 5 5
16 0 7 0 47
+ ω3 (0, 1, 2)−6 1 − ω3 (0, 0, 2)−7 1 − ω30 (0, 0, 0)−9 1,
9 3 9
0
that is, all vectors of the form ω3 (0, a, b) with 0 ≤ a ≤ b ≤ 4 can be written, using only the
translation operator, in terms of our proposed generating set, (4.35), with the addition of the
vectors ω30 (0, 0, 4), ω30 (0, 1, 4), ω30 (0, 2, 4), and ω30 (0, c, d) with 0 ≤ c ≤ d where d ≥ 5. The
10 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

remainder of our argument will be concerned with eliminating the need for these additional
vectors.
Using D5 (0, 0, 0, 1, 1) and the decomposition described in (4.31-4.32) we have
(4.38)
16 4 24 2
ω30 (0, 0, 4) = − ω30 (0, 1, 2)−2 1 + ω30 (0, 0, 2)−3 1 + ω30 (0, 0, 0)−5 1 − ω20 (0, 0)−1 ω30 (0, 1, 1)
15 15 45 5
4 0 0 2 0 0
+ ω2 (0, 1)−1 ω3 (0, 0, 1) − ω2 (1, 1)−1 ω3 (0, 0, 0),
5 5
where again this calculation was performed using [25]. Using D5 (0, 0, 1, 1, 1) a similar
decoupling equation for ω30 (0, 1, 4) can be found. Furthermore a linear combination of
D5 (0, 0, 0, 1, 3) and D5 (0, 0, 1, 1, 2) leads to the decoupling equation
(4.39)
1 0 1 1
ω30 (0, 2, 4) = ω (0, 0)−1 ω30 (1, 2, 2) + ω20 (0, 0)−1 ω30 (0, 1, 3) − ω20 (0, 1)−1 ω30 (0, 1, 2)
18 2 64 18
1 1 1
− ω20 (0, 1)−1 ω30 (0, 0, 3) − ω20 (0, 2)−1 ω30 (0, 1, 1) + ω20 (1, 2)−1 ω30 (0, 0, 1)
64 18 18
1 1 7
− ω20 (0, 3)−1 ω30 (0, 1, 1) + ω20 (1, 3)−1 ω30 (0, 0, 1) − ω30 (0, 0, 0)−7 1
64 64 18
59 0 187 0 115 0 37
− ω (0, 0, 2)−5 1 + ω (0, 0, 4)−3 1 + ω (0, 1, 2)−4 1 + ω30 (0, 1, 4)−2 1.
144 3 240 3 288 3 60
Next, for a ∈ N, each of the six expressions D5 (0, 0, 0, 1, a − 3), D5 (0, 0, 0, 2, a − 4),
D5 (0, 0, 1, 1, a − 4), D5 (0, 0, 0, 3, a − 5), D5 (0, 0, 1, 2, a − 5), and D5 (0, 1, 1, 1, a − 5) can be
expanded as linear combinations of the six vectors ω30 (0, 0, a), ω30 (0, 1, a − 1), ω30 (0, 2, a − 2),
ω30 (0, 3, a−3), ω30 (0, 4, a−4), and ω30 (0, 5, a−5) along with terms of the form ω3 (0, m1 , m2 )−1−n 1
where m1 + m2 + n = a with n ≥ 1. The linear independence of these can be determined
by considering the determinant of the matrix, A, whose (i, j) entry is the coefficient of the
ith vector in the j th expression as described above. A lengthy elementary calculation shows
that, in the case that a is even
1
det A = (a − 4)2 (a − 3)(a − 1)(a + 2)(5331a6 − 70325a5 + 314669a4 − 613567a3
(4.40) 6
+ 97384a2 + 1614156a − 1835568),
and a similar expression exists for odd a. In particular, for a ≥ 5, we can write each of
the vectors ω30 (0, 0, a), ω30 (0, 1, a − 1), ω30 (0, 2, a − 2), ω30 (0, 3, a − 3), ω30 (0, 4, a − 4), and
ω30 (0, 5, a − 5) as a vertex algebraic polynomial in terms of lower weight terms.
Next, for 5 ≤ a < b, D5 (b + 2, a − 1, 0, 0, 0) can be used to construct
(4.41)
2
ω30 (0, a + 1, b) = ((−1)b − (−1)a )ω30 (0, a − 1, b + 2)
a(a + 1)(b + 1)
1 + 3(−1)a 1 − 3(−1)b (−1)b − (−1)a
  
a+b+1
+ + + ω30 (0, 0, a + b + 1)
b a+1 b+2 a+b+1
+ Ψ,
where Ψ is a vertex algebraic polynomial with terms of the form ω30 (0, m1 , m2 )−1−n 1 and
ω20 (r1 , r2 )−1 ω30 (0, s1 , s2 ) where m1 + m2 + n = a + b + 1 and r2 + r2 + s1 + s2 = a + b − 1
with n ≥ 1.
Finally, (4.38) – (4.41), provide us with a clear inductive path to eliminating all cubic
generators other than ω30 (0, 0, 0), ω30 (0, 0, 2), and ω30 (0, 1, 2). This proves the first assertion.
To prove the second claim we use Lemma 4.3, where we reduced all quadratic generators
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 11

down to ω20 (0, 0), ω20 (0, 2), ω20 (0, 4). Although in this reduction we additionally used cubic
generators ω30 (0, 0, 1), ω30 (0, 1, 1), and ω30 (0, 2, 2), we can remove them using the first two
relations in (4.36), second relation in (4.37) and relation (4.38). 

Theorem 4.1. (i) The vertex operator algebra H(3)S3 is simple of type (1, 2, 3, 4, 5, 62 ), i.e.
it is strongly generated by seven vectors whose conformal weights are: 1, 2, 3, 4, 5, 6, 6. This
generating set is minimal.
(ii) H(3)S3 is isomorphic to H(1) ⊗ W , where W is of type (2, 3, 4, 5, 62 ).
(iii) H(3)S3 is not freely generated (by any set of generators).

Proof. Clearly, by a result from [11], this vertex algebra is simple. By Lemma 4.4 and
earlier discussion we see that H(3)S3 is strongly generated by vectors of conformal weights:
1 ( linear generator), 2, 4, 6 (quadratic generators) and 3, 5, 6 (cubic generators). From the
character formula (see Proposition 6.1) we see that

1
q 1/8 ch[H(3)S3 ](τ ) − = O(q 9 ),
(q; q)∞ (q 2 ; q)∞ (q 3 ; q)∞ (q 4 ; q)∞ (q 5 ; q)∞ (q 6 ; q)2∞

where (a; q)∞ := i≥0 (1 − aq i ). An easy analysis shows that dropping one (or more)
Q

generators from this generating set would imply that certain graded dimensions of H(3)S3
are strictly bigger than the corresponding graded dimension for the smaller subalgebra.
Thus, the proposed set of generators must be a minimal generating set. Clearly, this vertex
algebra is not freely generated by these generators due to

1 − q9
q 1/8 ch[H(3)S3 ](τ ) − = O(q 10 ),
(q; q)∞ (q 2 ; q)∞ (q 3 ; q)∞ (q 4 ; q)∞ (q 5 ; q)∞ (q 6 ; q)2∞

which implies that there must be a nontrivial relation at degree 9. In fact, one can quickly
argue this is impossible simply from the fact that ch[H(3)S3 ] is modular, which is impossible
to achieve with a free generating set.
For (ii), we first observe that hω1 (0)i is isomorphic to H(1). By taking the commutant
W := Comm(H(1), H(3)S3 ), we get H(1) ⊗ W ∼ = H(3)S3 . Each generator of H(3)S3 , except
ω1 (0), can be written as a linear combination of elements in H(1) ⊗ W with a nonzero
component in W (otherwise it would imply that some of the generators are contained in
hω1 (0)i, contradicting the minimality in (i)). These nonzero components form a generating
set of W .


Using (4.6) we may take the original invariants ω1 (0), ω2 (0, 0), ω2 (0, 2), ω2 (0, 4), ω3 (0, 0, 0),
ω3 (0, 0, 2), and ω3 (0, 1, 2) as our minimal strong generating set. Furthermore, the following
12 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

change of variables allows us to express the orbifold in terms of primary generators

1 24 30
h = ω1 (0), ω = ω2 (0, 0), J1 = 2ω2 (0, 2) − ω−1 ω − ω−3 1,
2 37 37
460 80 256
J2 = 24ω2 (0, 4) − ω2 (0, 2)−3 1 − ω−1 ω2 (0, 2) + ω−1 ω−1 ω
81 27 801
364 3472 40744
− ω−2 ω−2 1 + ω−3 ω + ω−5 1,
2403 2403 2403
1
C1 = ω3 (0, 0, 0) − h−1 h−1 h,
3
2 8 2
(4.42) C2 = 2ω3 (0, 0, 2) − ω3 (0, 0, 0)−3 1 − h−3 h−1 h + h−2 h−2 h
15 45 15
16 16
− ω−1 ω3 (0, 0, 0) + h−1 h−1 h−1 ω,
45 135
1 6 2
C3 = 2ω3 (0, 1, 2) + ω3 (0, 0, 2)−2 1 − ω3 (0, 0, 0)−4 − ω−1 ω3 (0, 0, 0)−2 1
10 25 25
3 1 2 2
+ ω−2 ω3 (0, 0, 0) − h−1 h−1 h−1 ω−2 1 + h−2 h−1 h−1 ω + ω−2 ω−2 ω−2 1
25 25 25 25
4 6
− h−3 h−2 h + h−4 h−1 h.
25 75
5. The orbifold H(3)Z3
We now
consider the orbifold of H(3) under the action of the cyclic subgroup Z3 ∼
=
1 2 3 ⊂ S3 . Using a similar strategy to the previous subsection we may take
ω10 = β1 (−1)1, ω2,3
0
(a, b) = β2 (−1 − a)β3 (−1 − b)1,
0
(5.1) ω2,2,2 (a, b, c) = β2 (−1 − a)β2 (−1 − b)β2 (−1 − c)1,
0
ω3,3,3 (a, b, c) = β3 (−1 − a)β3 (−1 − b)β3 (−1 − c)1
as our initial generating set. Analogous to Lemma 4.1 we have the following initial reduction
of the generating set
Lemma 5.1. The orbifold H(3)Z3 is strongly generated by vectors
ω10 (0), ω2,3
0
(0, a) for a ≥ 0,
(5.2) 0 0
ω2,2,2 (0, a, b) for 0 ≤ a ≤ b, ω3,3,3 (0, a, b) for 0 ≤ a ≤ b.
Proof. The linear generator is the same vector found in Lemma 4.1 which may be used to
produce all vectors of the form ω10 (a) for a ≥ 0 as before.
The quadratic generators will be reduced using an argument similar to the proof of Lemma
0 0
4.1, where the main difference is due to the fact that for a 6= b, ω2,3 (a, b) 6= ω2,3 (b, a). We
begin by setting
(5.3) A(m) = span{ω2,3 (a, b)|a + b = m} and ∂A(m) = {v−2 1|v ∈ A(m)}
and notice that for all m ≥ 0, ∂A(m) is a subspace of A(m + 1) of co-dimension 1. In fact,
we have
0
(5.4) A(m + 1) = ∂A(m) ⊕ Cω2,3 (0, m + 1),
from which it follows that the set
0
(5.5) {ω2,3 (0, i)−1−m+i 1}
is a basis of A(m). Our result follows.
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 13

Again, the reduction of the cubic generators follows from an analogue of (4.13). In
particular
0 0 0 0
(5.6) ωi,i,i (a, b, c)−2 1 = (a+1)ωi,i,i (a+1, b, c)+(b+1)ωi,i,i (a, b+1, c)+(c+1)ωi,i,i (a, b, c+1),
for i ∈ {1, 2} may be used inductively to achieve the desired set, where we use the fact that
0
ωi,i,i (a, b, c) is fixed under any permutation of the entries a, b, c.

We now present analogues of Lemmas 4.3 and 4.4 in this setting.
Lemma 5.2. The full list of quadratic generators described in (5.2) can be replaced with
0 0 0 0
(5.7) ω2,3 (0, 0), ω2,3 (0, 1), ω2,3 (0, 2), and ω2,3 (0, 3).
Proof. Our main tool for this argument will be the expression
0 0 0 0
(5.8) D2 (a) = ω2,3 (0, a)−1 ω2,3 (1, 1) − ω2,3 (1, a)−1 ω2,3 (0, 1),
which may be expanded to
(5.9)     
a+2 a+4 0 a+3 0
D2 (a) = ω2,3 (0, a + 4) + 2 ω2,3 (1, a + 3) − (−1)a (a + 1)ω2,3
0
(a + 3, 1).
2 2 3
An elementary calculation allows us to write
0 0 0
(5.10) ω2,3 (1, a + 3) = −(a + 4)ω2,3 (0, a + 4) + ω2,3 (0, a + 3)−2 1,
and more generally
a+3
X
0
ω2,3 (a + 3, 1) = (−1)a+j+1 (a − j + 4)ω2,3
0
(0, a − j + 4)−1−j 1
j=0

(5.11) = (−1)a+1 (a + 4)ω2,3

0
(0, a + 4)
a+3
X
+ (−1)a+j+1 (a − j + 4)ω2,3
0
(0, a − j + 4)−1−j 1.
j=1

Now combining (5.9)-(5.11) we have

1 0
D2 (a) = − (a + 6)(a + 4)(a + 1)(a − 1)ω2,3 (0, a + 4)
 12 
a+3 0
+2 ω2,3 (0, a + 3)−2 1
(5.12) 3
a+3
X
+ (−1)j (a + 1)(a − j + 4)ω2,3
0
(0, a − j + 4)−1−j 1,
j=1
0
which, for a = 0 and a ≥ 2, may be used to solve for ω2,3 (0, a + 4) in terms of vectors of
lower conformal weight.
Inductively, together with the special case
0 1 0 0 0 0
ω2,3 (0, 5) = (ω2,3 (0, 0)−1 ω2,3 (1, 2) − ω2,3 (1, 0)−1 ω2,3 (0, 2))
(5.13) 7
3 0 3 0 1 0
+ ω2,3 (0, 4)−2 1 − ω2,3 (0, 3)−3 1 + ω2,3 (0, 2)−4 1,
7 7 7
this allows us to remove all but the necessary generators. 
14 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

Lemma 5.3. The full list of cubic generators described in (5.2) can be replaced with
0 0 0 0
(5.14) ω2,2,2 (0, 0, 0), ω2,2,2 (0, 0, 2), ω3,3,3 (0, 0, 0), ω3,3,3 (0, 0, 2).

Proof. We follow a similar strategy to the proof of Lemma 4.4. In this case, our argument
is greatly simplified due to the fact that our decoupling relations are simpler and occur at
0 0
a lower initial conformal weight. We focus on the ω2,2,2 (0, a, b) terms, as the ω3,3,3 (0, a, b)
are similar, starting with the observation that

0 1 0
ω2,2,2 (0, 0, 1) = ω (0, 0, 0)−2 1
(5.15) 3 2,2,2
0 0 2 0
ω2,2,2 (0, 1, 1) = −ω2,2,2 (0, 0, 2) + ω2,2,2 (0, 0, 0)−3 1,
3
meaning that, by using the translation operator, all cubic generators of weight five or less
may be written in terms of (5.14).
The expression
0 0 0 0
(5.16) D3 (a, b) = ω2,3 (0, a)−1 ω2,2,2 (0, 0, b) − ω2,3 (b, a)−1 ω2,2,2 (0, 0, 0)

will be our main tool moving forward. A special case of (5.16) can be used to construct the
equation
(−1)a
 
0 1 0
(5.17) ω2,2,2 (0, 0, a) = ω2,2,2 (0, 0, a − 1)−2 1 + D3 (a − 3, 1) .
3a + 1 a−2
3
which can be used inductively to remove all generators of the form ω2,2,2 (0, 0, a) for a ≥ 3
from the generating set. Furthermore, a more general version yields
(−1)a
   
0 1 2a + 3b a + b 0
(5.18) ω2,2,2 (0, a, b) = ω2,2,2 (0, 0, a + b) + D(a − 2, b)
2 a+b a a−1
3
which can be used to remove all generators of the form ω2,2,2 (0, a, b) where 2 ≤ a ≤ b. This
0
leaves only the generators of the form ω2,2,2 (0, 1, a) which can be removed using

0 1 0 0
(5.19) ω2,2,2 (0, 1, a) = (ω (0, 0, a)−2 1 − (a + 1)ω2,2,2 (0, 0, a + 1)).
2 2,2,2


Theorem 5.1. (i) The vertex operator algebra H(3)Z3 is simple of type (1, 2, 33 , 4, 53 ), i.e.
it is strongly generated by seven vectors whose conformal weights are: 1, 2, 3, 3, 3, 4, 5, 5, 5.
This generating set is minimal.
(ii) H(3)Z3 is isomorphic to H(1) ⊗ W , where W is of type (2, 33 , 4, 53 ).
(iii) H(3)Z3 is not freely generated (by any set of generators).

Proof. This proof is similar to the proof of Theorem 4.1. In particular, we use the fact that
the free W-algebra on a generating set containing the same weight elements with any one
removed has certain graded dimensions strictly less than those described by the character
of H(3)Z3 , (6.1). 

Finally, the orbifold H(3)Z3 is described in terms of primary generators with the following
vectors.
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 15

(5.20)
1 1 1
h = ω10 (0), ω = ω20 (0, 0) + ω10 (0)−1 ω10 (0), J1 = ω20 (0, 1) − ω−2 1 + h−2 h,
2 2 2
0 0 2 22 77
J2 = 2ω2 (0, 2) − ω2 (0, 1)−2 1 + ω−1 ω − h−1 h−1 ω + h−3 h
3 3 9
83
− h−2 h−2 1,
12
8
J3 = 6ω20 (0, 3) − ω20 (0, 2)−2 1 + 4h−2 h−1 ω + 8h−1 h−1 ω−2 1
3
34 40
− 16h−3 ω + 8h−1 ω−3 1 + 4h4−1 h − h−2 h−2 h − h−3 h−1 h + 4h4−1 h
3 3
58 88
+ h−3 h−2 1 − 23h−4 h + h−5 1,
3 5
(2) 0 (2) 0 1 0 8 0
C1 = ω2,2,2 (0, 0, 0), C2 = ω2,2,2 (0, 0, 2) − ω2,2,2 (0, 0, 0)−3 1 − ω−1 ω2,2,2 (0, 0, 0),
15 15
(3) 0 (3) 0 1 0 8 0
C1 = ω3,3,3 (0, 0, 0), C2 = ω3,3,3 (0, 0, 2) − ω3,3,3 (0, 0, 0)−3 1 − ω−1 ω3,3,3 (0, 0, 0).
15 15

6. The character of H(n)Sn

We first derive a general formula for the character of the Sn -orbifold of a vertex algebra.
Although formulas of this form have appeared in the literature (e.g., [4] or [16]), we include
the proof for completeness.

Theorem 6.1. Let V be a VOA and f Sn (q) the character of the Sn -fixed point subalgebra
n
V (n)Sn ⊂ V ⊗ (under the usual action). Then
1 X
f Sn (q) = trV ⊗n gq L(0)−c/24
n!
g∈Sn

where g acts by permutation of tensor factors.

Proof. Consider the group algebra C[Sn ] and the idempotent

1 X
e= g ∈ C[Sn ], e2 = e
|Sn |
g∈Sn

where |Sn | = n!. The space Im(e) is precisely V (n)Sn . Since the eigenvaues of e are 1 and
0 (and 1 for the Sn -fixed subalgebra), the character can be computed simply by taking the
trace of e:
trV ⊗n eq L(0)−c/24 = trV (n)Sn q L(0)−c/24 .


Now we specialize to n = 3 and V = H(1).

Proposition 6.1.
∞ ∞ ∞ ∞
!
S3 q −1/8 Y 1 Y 1 Y 1 Y 1
ch[H(3) ](τ ) = n 3
+ 3 2n n
+ 2 .
6 n=1
(1 − q ) n=1
(1 − q ) n=1 (1 − q ) n=1
(1 − q 3n )
16 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

Proof. It suffices to consider elements: g = 1, g = (12) and g = (123). By the previous

theorem we have
1 
ch[H(3)S3 ](τ ) = trH(3) q L(0)−3/24 + 3 · trH(3) (12)q L(0)−3/24 + 2 · trH(3) (123)q L(0)−3/24
6
Clearly,
∞
Y 1
trH(3) q L(0)−3/24 = q −1/8 .
n=1
(1 − q n )3
For tr(12) we compute the trace on the following basis of H(3)
B = {pλ1 ⊗ pλ2 ⊗ pλ3 , λi ∈ P}
where P is the set of partitions, and for pλi we take obvious monomials of Heisenberg
elements. The matrix representation of (12) ∈ End(H(3)) in this basis has a non-zero entry
on the diagonal (=1) if and only if the corresponding basis element is
pλ ⊗ pλ ⊗ pν .
Since all pλ are generated by α(−i), vectors contributing to the trace are generated by
α(−i) ⊗ α(−i) ⊗ α(−j). Thus
∞ ∞
L(0)−3/24 −1/8
Y 1 Y 1
trH(3) (12)q =q 2i )
.
i=1
(1 − q j=1
(1 − qj )

One similarly argues for 3-cycles, so we get

∞
Y 1
trH(3) (123)q L(0)−3/24 = q −1/8 .
i=1
(1 − q 3i )

Corollary 6.1.
∞ ∞
!
Z3 q −1/8 Y 1 Y 1
ch[H(3) ](τ ) = n 3
+ 2 .
3 n=1
(1 − q ) n=1
(1 − q 3n )

7. Modular invariance of H(3)S3 -characters and quantum dimensions

By a result from [11] (especially Section 7 devoted to dihedral groups), we have a decom-
position of H(3) in terms of H(3)S3 -modules:
(7.1) H(3) = H(3)S3 ⊕ H(3)S3 ,sgn ⊕ V (2) ⊗ H(3)S3 ,st
where V (2) is the 2-dimensional standard representation of S3 , and H(3)S3 ,sgn and H(3)S3 ,st
are irreducible H(3)S3 -modules.
Lemma 7.1. We have
∞ ∞ ∞ ∞
!
S3 ,sgn q −1/8 Y 1 Y 1 Y 1 Y 1
ch[H(3) ](τ ) = n )3
− 3 2n ) n)
+ 2
6 n=1
(1 − q n=1
(1 − q n=1
(1 − q n=1
(1 − q 3n )
∞ ∞
!
q −1/8 Y 1 Y 1
ch[H(3)S3 ,st ](τ ) = 2 −2 .
6 n=1
(1 − q n )3 n=1
(1 − q 3n )
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 17

Proof. The character of the sign module [11]

H(3)S3 ,sgn = {a ∈ H(3) : σ(a) = sgn(σ)a, σ ∈ S3 }
is computed along the lines of Proposition 6.1 with the difference that in the sign represen-
tation 2-cycles act as −1. This sign contributes with a negative sign to the middle terms of
the formula.
The second relation follows directly from the first formula, Proposition 6.1, and decom-
position (7.1). 
Next we construct several H(3)S3 -modules coming from ordinary and g-twisted H(3)-
modules, g ∈ S3 .
We first look at H(3)S3 -modules coming from irreducible H(3)-modules (Fock represen-
tations). Let w = (w1 , w2 , w3 ) ∈ C3 . We denote by Fw1 ,w2 ,w3 the irreducible H(3)-module
with highest weight w such that αi (0) acts as multiplication by wi . Similarly, Fw denotes
the Fock representation for H(1) with highest weight w.
Interestingly, Fw1 ,w2 ,w3 is generically irreducible as an H(3)S3 -module.
Lemma 7.2. Let w1 , w2 , w3 ∈ C3 such that wi 6= wj , i 6= j. Then the H(3)-module
Fw1 ,w2 ,w3 is irreducible as an H(3)S3 -module.
Proof. Let V be a simple VOA, g be an automorphism of V of prime order p, and M be a
simple V -module such that g ◦ M is not isomorphic to M as V -modules. Then according
to Theorem 6.1 [11], M is a simple V <g> -module. To apply this result we first observe
embeddings
H(3)S3 ,→ H(3)<(12)> ,→ H(3).
Because of w1 6= w2 , (12) ◦ Fw1 ,w2 ,w3  Fw1 ,w2 ,w3 as H(3)-modules. Indeed, for (12) ◦
Fw1 ,w2 ,w3 actions of α1 (0) and α2 (0) are switched, so (12)◦Fw1 ,w2 ,w3 ∼
= Fw2 ,w1 ,w3 . Therefore
Fw1 ,w2 ,w3 is irreducible as a G(3) := H(3)(<12>) -module. Similarly, we consider (123) ◦
Fw1 ,w2 ,w3 and Fw1 ,w2 ,w3 as G(3)-modules. Because of w1 6= w3 and w3 6= w2 , they are not
isomorphic as G(3)-modules and thus Fw1 ,w2 ,w3 is irreducible as a G(3)<(123)> -module. The
proof follows. 
In the study of characters we require modular variables so we let q = e2πiτ , where τ ∈ H
(the upper half-plane). Let
η(τ ) = q 1/24 (q; q)∞ ,
the Dedekind η-function. Clearly, for every w,
2 2 2 2 2 2
q −3/24+w1 /2+w2 /2+w3 /2 q w1 /2+w2 /2+w3 /2
(7.2) ch[Fw1 ,w2 ,w3 ](τ ) = = .
(q; q)∞ η(τ )3
Next we consider g-twisted H(3)-modules, so g is either a 2− or 3-cycle. We may assume,
without loss of generality, that g is either (12) or (123).
We first construct a family of θ-twisted H(3)-modules, where θ := (12) switches the first
two ”coordinates”. By using the isomorphism in Section 3, H(2)<(12)> ∼ = H(1) ⊗ M (1)+ ,
for every w1 , w3 ∈ C, we get a θ-twisted H(3)-module
Fw1 ,w3 (θ) := Fw1 ⊗ M (1)(θ) ⊗ Fw3 ,
where M (1)(θ) ∼
= C[α(−1/2), α(−3/2), ....] is a Z2 -twisted H(1)-module studied in [12]. This
1
module is of conformal weight 16 and its character is easily computed. We quickly get
2 2 2 2
q 1/16−3/24+w1 /2+w3 /2 q −1/16+w1 /2+w3 /2
ch[Fw1 ,w3 (θ)](τ ) = 1/2 = 1/2 1/2
(q ; q)∞ (q; q)∞ (q; q)∞ (q ; q )∞ (q; q)∞
18 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

2 2
q w1 /2+w3 /2
(7.3) = .
η(τ /2)η(τ )

Further, we construct a family of (123)-twisted H(3)-modules. Let σ = (123). We

first construct a σ-twisted Heisenberg algebra ĥσ as in [19]. It is isomorphic to C[t, t−1 ] ⊕
t1/3 C[t, t−1 ] ⊕ t2/3 C[t, t−1 ] ⊕ Ck as a graded Lie algebra. Then for every w ∈ C we have an
associated σ-twisted H(3)-representation Fw (σ) ∼ = U (hσ<0 ). Moreover,
2 2
q w /2+h−1/8 q w /2+h−3/24
ch[Fw (σ)](τ ) = 1/3 2/3
= 1/3 1/3 ,
(q; q)∞ (q ; q)∞ (q ; q)∞ (q ; q )∞
Pp−1
where h = 4p12 i=1 i(p − i)ri , and p is the order of σ and ri (1 ≤ i ≤ 3) are dimensions of
the eigenspaces of σ. Plugging in p = 3 and ri = 1 into the last formula we obtain h = 19
and thus
2 2
q w /2−1/72 q w /2
(7.4) ch[Fw (σ)](τ ) = 1/3 1/3 = .
(q ; q )∞ η(τ /3)

Here we do not pursue decomposition and irreducibility of Fw1 ,w3 (σ) and Fw (σ) as
H(3)S3 -modules. This will be a subject of [24].

7.1. Modular invariance. In this part we obtain a modularity result for the character of
H(3)S3 under the S-transformation, τ → − τ1 . This was briefly discussed in the introduction.
We make use of a well-known modular relation for the Dedekind η-function:
√
(7.5) η(−1/τ )n = ( −iτ )n η(τ )n , n ∈ N.

Let us also recall a higher dimensional Gauss’ integral formula

Z
2 2 1
q w1 /2+···+wn /2 dw = √ .
R n ( −iτ )n

From the Gauss formula and (7.5) we immediately get the following relations.

Lemma 7.3.
2 2 2
q w1 /2+w2 /2+w3 /2
Z
1
= dw, w = (w1 , w2 , w3 )
η(−1/τ )3 R3 η(τ )3
1 √ Z q w12 /2+w32 /2
= 2 dw, w = (w1 , w3 )
η(−1/τ )η(−2/τ ) R2 η(τ )η(τ /2)

1 √ Z q w2 /2
= 3 dw.
η(−3/τ ) R η(τ /3)

Notice that only the numerator of the integrand depends on w. The denominator is
placed inside the integral so that it matches the shape of (1.1).
Proposition 6.1, the character formulas for Fw1 ,w2 ,w3 , Fw1 ,w3 (θ) and Fw (σ), that is, for-
mulas (7.2), (7.3) and (7.4), together with Lemma 7.3 imply

Theorem 7.1. The character of H(3)S3 has a modular invariance property, in the sense of
(1.1).
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 19

7.2. Quantum dimensions. Let V be a vertex operator algebra. For a V -module M we

define its quantum dimension:
ch[M ](it)
qdim[M ] := lim .
t→0+ ch[V ](it)
In a situation where V is rational (in the strongest sense) this limit always exists and is
nonzero (see [13]). But for general V this limit can be zero or even +∞.
As our characters can be expressed in terms of η(τ ), we will make use of an asymptotic
formula
 
1 πn
(7.6) η(it)n ∼ n e− 12t , (t → 0+ ).
t2
Proposition 7.1. As H(3)S3 -modules, we have:
(a) For any Fock representation Fw1 ,w2 ,w3 ,
qdim[Fw1 ,w2 ,w3 ] = 6.
(b) For representations appearing inside H(3):
qdim[H(3)S3 ] = 1
qdim[H(3)S3 ,sgn ] = 1
qdim[H(3)S3 ,st ] = 2.
(c) Fw1 ,w3 (θ) and Fw (σ) have quantum dimension +∞.
Proof. We first observe that in the formulas for ch[H(3)S3 ](τ ), ch[H(3)S3 ,sgn ](τ ) and ch[H(3)S3 ,st ](τ )
q −1/8
the infinite product (q;q) 3 = η(τ1 )3 , multiplied with a constant, dominates the asymptotics
∞
t → 0+ . Therefore in order to compute quantum dimensions for these modules, we only
have to compute the ratio of these constants. This implies assertions (a) and (b).
For (c), we need more precise asymptotic behaviors. We have
1 t π
∼ √ e 4t ,
η(τ )η(τ /2) 2
1/2
1 t π
∼ √ e 4t ,
η(τ /3) 3
1 π

3
∼ t3/2 e 4t .
η(τ )
From these formulas we get
ch[Fw1 ,w3 (θ)(it) 1
S
∼√ ,
ch[H(3) ](it)
3
2t1/2
ch[Fw (σ)(it) 1
∼√ ,
ch[H(3)S3 ](it) 3t
so both ratios have growing terms at t = 0. The proof follows.

We finish with a conjecture which will be addressed in [24].
Conjecture 1. Every irreducible (ordinary) H(3)S3 -module M appears in the decomposition
of a g-twisted H(3)-module, for g ∈ S3 . Moreover,
qdim(M ) ∈ {1, 2, 6, +∞}.
20 ANTUN MILAS, MICHAEL PENN, HANBO SHAO

8. Future work
Remark 8.1. (1) In order to confirm Conjecture 1, we first have to describe the Zhu
algebra of H(3)S3 . This will be addressed in [24].
(2) In our very recent work [23], we determine strong minimal generating sets for the
permutation orbifold F(3)S3 , where F is the free fermion vertex algebra and for
SF(3)S3 , where SF is the rank one symplectic fermion vertex superalgebra [9].
We prove that these vertex algebras are of type 12 , 2, 4, 92 and 12 , 2, 33 , 43 , 55 , 64 ,
respectively.
Acknowledgements: We thank Drazen Adamovic, Thomas Creutzig and especially
Andrew Linshaw for useful discussions. We also thank the referee for constructive criticism
and other valuable comments. Several computations in this paper are performed by the
OPE package, [25], for Mathematica.
The first author was partially supported by the NSF grant DMS-1601070.

References
[1] T. Abe, Fusion Rules for the Free Bosonic Orbifold Vertex Operator Algebra, Journal of Algebra
229 (2000): 333-374.
[2] T. Abe, C2 -cofiniteness of 2-cyclic permutation orbifold models, Communications in Mathematical
Physics 317 (2013): 425-445.
[3] D. Adamovic, X. Lin, and A. Milas, ADE subalgebras of the triplet vertex algebra: A-type,
Communications in Contemporary Mathematics 15 (2013): 1350028.
[4] P. Bantay, Characters and modular properties of permutation orbifolds, Physics Letters B 419
(1998): 175-178.
[5] P. Bantay, Permutation orbifolds, Nuclear Physics B 633 (2002): 365-378.
[6] K. Barron, C. Dong, and G. Mason. Twisted sectors for tensor product vertex operator algebras
associated to permutation groups, Communications in mathematical physics 227 (2002): 349-384.
[7] K. Barron, Y.-Z. Huang, and J. Lepowsky, An equivalence of two constructions of permutation-
twisted modules for lattice vertex operator algebras, Journal of Pure and Applied Algebra 210
(2007): 797-826.
[8] L. Borisov, M. B. Halpern, and C. Schweigert, Systematic approach to cyclic orbifolds, Interna-
tional Journal of Modern Physics A 13 (1998): 125-168.
[9] T. Creutzig and A. Linshaw, Orbifolds of symplectic fermion algebras, Transactions of the
American Mathematical Society 369 (2017): 467-494
[10] C. Dong and C. Jiang, Representations of the vertex operator algebra VL2 A4 , Journal of Algebra

377 (2013): 76-96.

[11] C. Dong and G. Mason, On quantum Galois theory, Duke Mathematical Journal 86 (1997):
305-321.
[12] C. Dong and K. Nagatomo, Classification of irreducible modules for the vertex operator algebra
M (1)+ , Journal of Algebra 216 (1999): 384-404.
[13] C. Dong, L. Ren, and F. Xu, On orbifold theory, Advances in Mathematics 321 (2017): 1-30.
[14] C. Dong, F. Xu, and N. Yu, The 3-permutation orbifold of a lattice vertex operator algebra,
Journal of Pure and Applied Algebra 222 (2018): 1316-1336.
[15] I. Frenkel, J. Lepowsky and A. Meurman, Vertex Operator Algebras and the Monster, Pure and
Applied Math., Vol. 134, Academic Press, 1988.
[16] V. Kac and I. Todorov, Affine orbifolds and rational conformal field theory extensions of W1+∞ ,
Communications in mathematical physics 190 (1997): 57-111.
[17] A. Klemm and M. Schmidt, Orbifolds by cyclic permutations of tensor product conformal field
theories, Physics Letters B 245 (1990): 53-58.
[18] A. Kirillov Jr, On G–equivariant modular categories, arXiv.0401119 (2004).
[19] J. Lepowsky, Calculus of twisted vertex operators, Proceedings of the National Academy of
Sciences 82 (1985): 8295-8299.
[20] A. Linshaw, A Hilbert theorem for vertex algebras, Transformation Groups 15 (2010): 427- 448.
[21] A. Linshaw, Invariant Theory and the Heisenberg Vertex Algebra, International Mathematics
Research Notices 17 (2012): 4014-4050
 PERMUTATION ORBIFOLDS OF THE HEISENBERG VERTEX ALGEBRA H(3) 21

[22] A. Linshaw, Invariant subalgebras of affine vertex algebras, Advances in Mathematics 234 (2013):
61-84.
[23] A. Milas, M. Penn and J. Wauchope, S3 -orbifolds of F (3) and SF (3), submitted.
[24] A. Milas, M. Penn and J. Wauchope, in preparation.
[25] K. Theilmans, A Mathematica package for computing operator product expansions, International
Journal of Modern Physics, (1991) C2:787-798.
[26] H. Weyl, The Classical Groups: Their Invariants and Representations, Princeton University
Press, 1946.

Department of Mathematics and Statistics, SUNY-Albany

E-mail address: [email protected]

Mathematics Department, Randolph College

E-mail address: [email protected]

Department of Mathematics and Computer Science, Colorado College

E-mail address: h [email protected]

View publication stats