THE RING OF MODULAR FORMS FOR THE EVEN UNIMODULAR LATTICE

OF SIGNATURE (2,18)

ATSUHIRA NAGANO AND KAZUSHI UEDA

Abstract. We show that the ring of modular forms with characters for the even unimodular lattice
of signature (2,18) is obtained from the invariant ring of Sym(Sym8 (V ) ⊕ Sym12 (V )) with respect to
the action of SL(V ) by adding a Borcherds product of weight 132 with one relation of weight 264,
where V is a 2-dimensional C-vector space. The proof is based on the study of the moduli space of
elliptic K3 surfaces with a section.
arXiv:2102.09224v1 [math.AG] 18 Feb 2021

1. Introduction
Let U be the even unimodular hyperbolic lattice of rank 2. A U-polarized K3 surface in the sense
of [Nik79] is a pair (U, j) of a K3 surface Y and a primitive lattice embedding j : U ֒→ Pic Y. As
explained, e.g., in [Huy], an elliptic K3 surface with a section corresponds naturally to a pseudo-ample
U-polarized K3 surface. Fix a primitive embedding of U to the K3 lattice Λ = U⊥U⊥U⊥E8 ⊥E8 ,
which is unique up to the left action of O(Λ), and let T = U⊥U⊥E8 ⊥E8 be the orthogonal lattice.
As explained in [Dol96, Section 3], the global Torelli theorem [PŠŠ71, BR75] and the surjectivity
of the period map [Tod80] show that the period map gives an isomorphism from the coarse moduli
scheme of pseudo-ample U-polarized K3 surfaces to the quotient M := Γ\D of the bounded Hermitian
domain
 
(1.1) D := [Ω] ∈ P(T ⊗ C)  (Ω, Ω) = 0, (Ω, Ω) > 0
of type IV by Γ := O(T).
The moduli space of elliptic K3 surfaces with a section attracts much attention recently, not only
from the point of view of modular compactification (see e.g. [AB, ABE] and references therein), but
also because of the relation with tropical geometry and mirror symmetry [HU19, OO].
A modular form on D with respect to Γ of weight k ∈ Z and character χ ∈ Char(Γ) := Hom(Γ, C× )
is a holomorphic function f : De → C on the total space
 
(1.2) De := [Ω] ∈ T ⊗ C  (Ω, Ω) = 0, (Ω, Ω) > 0
of a principal C× -bundle on D satisfying
(i) f (αz) = α−k f (z) for any α ∈ C× , and
(ii) f (γz) = χ(γ)f (z) for any γ ∈ Γ.
The vector spaces Ak (Γ, χ) of modular forms constitute the ring
∞
M M
(1.3) e
A(Γ) := Ak (Γ, χ)
k=0 χ∈Char(Γ)

of modular forms. We also write the subring of modular forms without characters as
M∞
(1.4) A(Γ) := Ak (Γ).
k=0

Let V := Spec C[x, w] be a 2-dimensional affine space over C. For k ∈ N, we write the k-th
symmetric product of V as Symk V . The special linear group SL2 acts naturally on S := Sym8 V ×
Sym12 V considered as an affine variety, whose coordinate ring will be denoted by
(1.5) C[S] = C[u8,0 , u7,1, . . . , u0,8 , u12,0, u11,1 , . . . , u0,12 ].
A. N. was partially supported by JSPS Kakenhi (18K13383) and MEXT LEADER.
K. U. was partially supported by JSPS Kakenhi (16H03930).
1
We let Gm act on S in such a way that ui,j has weight (i + j)/2. This Gm -action commutes with the
SL2 -action, so that the invariant subring C[S]SL2 has an induced Gm -action.
Building on [Mir81], it is shown in [OO, Theorem 7.9] that the period map induces an isomorphism
from Proj C[S]SL2 to the Satake–Baily–Borel compactification of Γ\D, so that one has an isomorphism
(1.6) A(Γ) ∼
= C[S]SL2
of graded rings.
The main result of this paper is the following:
Theorem 1.1. One has



(1.7) e
A(Γ) ∼
= C[S]SL2 [s132 ] s2132 − ∆264
where s132 is an element of weight 132 and ∆264 ∈ C[S]SL2 is an element of weight 264.
The proof is based on the construction of an algebraic stack which is isomorphic to the orbifold
quotient [D/O(T )] in codimension 1. The same strategy has been used in [HU] and [NU] to determine
the rings of modular forms with characters for the lattices U⊥U⊥E8 and U⊥U⊥A1 ⊥A1 respectively.
The modular form s132 is constructed in [FSM07, Lemma 5.1]. It can also be obtained either as the
quasi pull-back [GHS13, Theorem 8.2] of the Borcherds form Φ12 associated with the even unimodular
lattice of signature (2, 26) [Bor95, Section 10, Example 2], or by applying [Bor95, Theorem 10.1] to
the nearly holomorphic modular form
1728E4 1
(1.8) 3 2
= + 264 + 8244q + 139520q 2 + · · ·
E4 − E6 q
where
X ∞
n3 q n
(1.9) E4 = 1 + 240 n
= 1 + 240q + 2160q 2 + · · · ,
n=1
1−q
X∞
n5 q n
(1.10) E6 = 1 − 504 n
= 1 − 504q − 6632q 2 + · · · .
n=1
1 − q
In particular, it is a cusp form with character det admitting an infinite product expansion. See also
[DKW19, Section 5] and references therein for the case of the even unimodular lattice of signature
(2,10).
Since SL2 is reductive, the invariant ring C[S]SL2 is finitely generated, and there exists an algorithm
for computing a finite generating set (see e.g. [Stu08] and references therein). The element ∆264 can
also be computed algorithmically, and it is an interesting problem to describe them explicitly.

2. The coarse moduli space of U-polarized K3 surfaces

As is well known (cf. e.g. [SS10, Section 4]), a U-polarized K3 surface admits a Weierstrass model
of the form
(2.1) z 2 = y 3 + g2 (x, w; u)y + g3 (x, w; u)
in P(1, 4, 6, 1), where
8
X
(2.2) g2 (x, w; u) = u8−i,i x8−i w i
i=0
(2.3) = u8,0 x8 + u7,1 x7 u + · · · + u0,8 w 8 ,
12
X
(2.4) g3 (x, w; u) = u12−i,i x12−i w i
i=0
(2.5) = u12,0 x12 + u11,1 x11 u + · · · + u0,12 w 12
for
(2.6) u = ((u8,0 , . . . , u0,8 ), (u12,0 , . . . , u0,12)) ∈ S.
2
The hypersurface in P(1, 4, 6, 1) defined by (2.1) has a singularity worse than rational double points
on the fiber at a ∈ P1 if and only if orda (g2 ) ≥ 4 and orda (g3 ) ≥ 6 (see e.g. [Mir89, Proposition
III.3.2]). Let U ⊂ S be the open subscheme parametrizing hypersurfaces with at worst rational double
points.
The parameter u describing a given U-polarized K3 surface is unique up to the action of SL2 ×Gm ,
where Gm acts on P(1, 4, 6, 1) × Sym8 V × Sym12 V by
(2.7) Gm ∋ λ : ((x, y, z, w), (ui,j )i,j ) 7→ (x, λ2 y, λ3z, w), (λ(i+j)/2 ui,j )i,j )
rescaling the holomorphic volume form
wdx ∧ dy ∧ dz
(2.8) Ω = Res
z2 − y3 − g2 (x, w; u)y − g3 (x, w; u)
as
wdx ∧ d(λ2 y) ∧ d(λ3z)
(2.9) Ωλu = Res = λ−1 Ωu .
(λ3 z)2 − (λ2 y)3 − g2 (x, w; λ · u)(λ2y) − g3 (x, w; λ · u)
The categorical quotient T := U/ SL2 is the coarse moduli scheme of pairs (Y, Ω) consisting of a
U-polarized K3 surface Y and a holomorphic volume form Ω ∈ H 0 (ωY ) on Y . The fact that the
codimension of S \ U is greater than 2 implies an isomorphism
(2.10) C[S]SL2 ∼ = C[T ]
of graded rings. Since the character of C[S] as a SL2 ×Gm -module is given by
8
Y 12

−1 Y
−1
(2.11) 1 − q 2i−8 t4 1 − q 2i−12 t6 ,
i=0 i=0
the Hilbert series of the invariant ring is given by
∞ 8 12
!
X
Y
Y
−1
SL2 2i−8 4 −1
(2.12) dim C[S] i
ti = Resq=0 (q −1 − q) 1−q t 1 − q 2i−12 t6
i=0 i=0 i=0

as explained, e.g., in [Muk03, Section 4.4]. It follows from the global Torelli theorem and the
surjectivity of the period map that the period map induces a ring isomorphism
∼
(2.13) A(Γ) −
→ C[T ],
which preserves the grading by (2.9). The isomorphism (1.6) follows from (2.10) and (2.13).

3. Modular forms with characters

The coarse moduli space M := Γ\D of U-polarized K3 surfaces is an open subvariety of its Satake–
Baily–Borel compactification Proj A(Γ) ∼ = P(49 , 613 )//SL2 . Although M and the orbifold quotient
M := [Γ\D] are closely related, the canonical morphism M → M is not an isomorphism even in
codimension 1. In order to obtain an orbifold which is isomorphic to M in codimension 1 (so that
the total coordinate rings are isomorphic), consider the stack


(3.1) P := P 49 , 613 / SL2 ,
defined as the quotient of C24 \ 0 by the action of SL2 ×Gm . The morphism M → M lifts to a
morphism M → P, which is an isomorphism in codimension 0, since the generic stabilizers are {± id}
on both sides.
Stabilizers of M along divisors come from reflections. One divisor with a generic stabilizer comes
from the reflection with respect to a (−2)-vector whose reflection hyperplane corresponds to the
locus where the Picard lattice contains U⊥A1 . In order to describe this locus, first consider the
discriminant
(3.2) h(x, w; u) := 4g2 (x, w; u)3 + 27g3 (x, w; u)2
of y 3 + g2 (x, w; u)y + g3 (x, w; u) as a polynomial in y, which is P
homogeneous of degree 24 in (x, w)
and degree 12 in u. Note that the discriminant of a polynomial ni=0 ai xi w n−i with respect to (x, w)
is homogeneous of degree 2(n − 1) in Z[a0 , . . . , an ] if deg a0 = · · · = deg an = 1. It follows that
3
the discriminant k552 (u) of h(x, w; u) with respect to (x, w) is a homogeneous polynomial of degree
2 · 23 · 12 = 552 in u. A general point on the divisor D552 of P defined by k552 (u) corresponds to the
locus where two fibers of Kodaira type I1 collapse into one fiber. This divisor has two components; a
general point on one component corresponds to the case when there exists a point p = [x : w] on P1
such that neither g2 nor g3 vanishes at p, and a general point on the other component corresponds
to the case when both g2 and g3 vanishes at p. In the former case, the resulting singular fiber is of
Kodaira type I2 , and the surface acquires an A1 -singularity. In the latter case, the resulting singular
fiber is of Kodaira type II, and the surface does not acquire any new singularity. The defining equation
of the latter component is the resultant of g2 and g3 . It is given as the determinant
 
 u8,0 u7,1 · · · u0,8 
 
 u8,0 · · · u1,7 u0,8 
 
 .. .. .. 
 . . . 
 .. .. .. 
 . . . 
(3.3) 
r96 (t) =  
u · · · u u 
 8,0 1,7 0,8 
u 
 12,0 u11,1 · · · · · · u1,11 u0,12 
 . .. . .. . .. 
 
 
 u12,0 u11,1 · · · · · · u1,11 u0,12 
of the Sylvester matrix, which is homogeneous of degree
(3.4) 12 × 4 + 8 × 6 = 96.
As shown in [HU, Lemma 6.1], the polynomial k552 (t) is divisible by r96 (t)3 , and the quotient
(3.5) ∆264 (t) := k552 (t)/r96 (t)3
defines the reflection hyperplane along a (−2)-vector.
Recall from [AGV08, Cad07] that the root construction is an operation which adds a stabilizer
along a divisor. Let T be the stack obtained from P by the root construction of order 2 along the
divisor on P defined by ∆264 (t), which is the quotient of the double cover of P branched along ∆264 (t)
by the group G of deck transformations. The Picard group of T (or the G-equivariant Picard group
of P) is generated by the pull-back OT (1) := p∗ OP (1) of the generator OP (1) of the Picard group of P
by the structure morphism p : T → P and the line bundle OT (D132 ) such that the space H 0 (OT (D132 ))
is generated by an element s132 satisfying s2132 = ∆264 ∈ H 0 (OT (264)) ∼
= H 0 (OP (264)). Note also that
ωP ∼= OP (a) where
8
X 12
X
(3.6) a=− deg u8−i,i − deg u12−i,i = −9 × 4 − 13 × 6 = −114.
i=0 i=0

The ramification formula for the canonical bundle gives

(3.7) ωT ∼= p∗ ωP ⊗ OT (D132 )
(3.8) ∼
= OT (−114) ⊗ OT (132 + (−132 + D132 ))
(3.9) ∼
= OT (18) ⊗ OT (−132 + D132 ).
Note that OT (−132 + D132 ) is an element of order two in Pic T. By comparing (3.9) with
(3.10) ωM ∼
= OM (dim M) ⊗ det = OM (18) ⊗ det
which follows from (the proof of) [HU, Proposition 5.1], one concludes that M has no further stabilizer
along a divisor, so that the lift M → T of M → P is an isomorphism in codimension 1. It follows that
the injective map Z × Char(Γ) → Pic M, (i, χ) 7→ OM (i) ⊗ χ is surjective, and the total coordinate
ring (also known as the Cox ring) of M is given by
M ∞
M
(3.11) H 0 (L) ∼
= H 0 (OM (i))/(s2132 − ∆264 (t)).
L∈Pic M i=0
4
 References
[AB] Kenneth Ascher and Dori Bejleri, Compact moduli of elliptic K3 surfaces, arXiv:1902.10686. 1
[ABE] Valery Alexeev, Adrian Brunyate, and Philip Engel, Compactifications of moduli of elliptic K3 surfaces:
stable pair and toroidal, arXiv:2002.07127. 1
[AGV08] Dan Abramovich, Tom Graber, and Angelo Vistoli, Gromov-Witten theory of Deligne-Mumford stacks,
Amer. J. Math. 130 (2008), no. 5, 1337–1398. MR 2450211 (2009k:14108) 4
[Bor95] Richard E. Borcherds, Automorphic forms on Os+2,2 (R) and infinite products, Invent. Math. 120 (1995),
no. 1, 161–213. MR 1323986 2
[BR75] Dan Burns, Jr. and Michael Rapoport, On the Torelli problem for kählerian K − 3 surfaces, Ann. Sci. École
Norm. Sup. (4) 8 (1975), no. 2, 235–273. MR 0447635 1
[Cad07] Charles Cadman, Using stacks to impose tangency conditions on curves, Amer. J. Math. 129 (2007), no. 2,
405–427. MR 2306040 (2008g:14016) 4
[DKW19] C. Dieckmann, A. Krieg, and M. Woitalla, The graded ring of modular forms on the Cayley half-space of
degree two, Ramanujan J. 48 (2019), no. 2, 385–398. MR 3911795 2
[Dol96] I. V. Dolgachev, Mirror symmetry for lattice polarized K3 surfaces, J. Math. Sci. 81 (1996), no. 3, 2599–
2630, Algebraic geometry, 4. MR 1420220 (97i:14024) 1
[FSM07] Eberhard Freitag and Riccardo Salvati Manni, Modular forms for the even modular lattice of signature
(2, 10), J. Algebraic Geom. 16 (2007), no. 4, 753–791. MR 2357689 2
[GHS13] V. Gritsenko, K. Hulek, and G. K. Sankaran, Moduli of K3 surfaces and irreducible symplectic manifolds,
Handbook of moduli. Vol. I, Adv. Lect. Math. (ALM), vol. 24, Int. Press, Somerville, MA, 2013, pp. 459–526.
MR 3184170 2
[HU] Kenji Hashimoto and Kazushi Ueda, The ring of modular forms for the even unimodular lattice of signature
(2,10), arXiv:1406.0332. 2, 4
[HU19] , Reconstruction of general elliptic K3 surfaces from their Gromov–Hausdorff limits, Proc. Amer.
Math. Soc. 147 (2019), no. 5, 1963–1969. MR 3937674 1
[Huy] Daniel Huybrechts, Lectures on K3 surfaces, available at http://www.math.uni-bonn.de/people/huybrech/K3Globa
1
[Mir81] Rick Miranda, The moduli of Weierstrass fibrations over P1 , Math. Ann. 255 (1981), no. 3, 379–394.
MR 615858 2
[Mir89] , The basic theory of elliptic surfaces, Dottorato di Ricerca in Matematica. [Doctorate in Mathemat-
ical Research], ETS Editrice, Pisa, 1989. MR 1078016 (92e:14032) 3
[Muk03] Shigeru Mukai, An introduction to invariants and moduli, Cambridge Studies in Advanced Mathematics,
vol. 81, Cambridge University Press, Cambridge, 2003, Translated from the 1998 and 2000 Japanese editions
by W. M. Oxbury. MR 2004218 3
[Nik79] V. V. Nikulin, Finite groups of automorphisms of Kählerian K3 surfaces, Trudy Moskov. Mat. Obshch. 38
(1979), 75–137. MR 544937 (81e:32033) 1
[NU] Atsuhira Nagano and Kazushi Ueda, The ring of modular forms of O(2, 4; Z) with characters, to appear in
Hokkaido Mathematical Journal, arXiv:2005.00231. 2
[OO] Yuji Odaka and Yoshiki Oshima, Collapsing K3 surfaces, tropical geometry and moduli compactifications of
Satake, Morgan-Shalen type, arXiv:1810.07685. 1, 2
[PŠŠ71] I. I. Pjateckiı̆-Šapiro and I. R. Šafarevič, Torelli’s theorem for algebraic surfaces of type K3, Izv. Akad. Nauk
SSSR Ser. Mat. 35 (1971), 530–572. MR 0284440 (44 #1666) 1
[SS10] Matthias Schütt and Tetsuji Shioda, Elliptic surfaces, Algebraic geometry in East Asia—Seoul 2008, Adv.
Stud. Pure Math., vol. 60, Math. Soc. Japan, Tokyo, 2010, pp. 51–160. MR 2732092 (2012b:14069) 2
[Stu08] Bernd Sturmfels, Algorithms in invariant theory, second ed., Texts and Monographs in Symbolic Compu-
tation, SpringerWienNewYork, Vienna, 2008. MR 2667486 (2011j:13009) 2
[Tod80] Andrei N. Todorov, Applications of the Kähler-Einstein-Calabi-Yau metric to moduli of K3 surfaces, Invent.
Math. 61 (1980), no. 3, 251–265. MR 592693 1

Faculty of Mathematics and Physics, Institute of Science and Engineering, Kanazawa University,
Kakuma, Kanazawa, Ishikawa, 920-1192, Japan
Email address: [email protected]

Graduate School of Mathematical Sciences, The University of Tokyo, 3-8-1 Komaba, Meguro-ku,
Tokyo, 153-8914, Japan.
Email address: [email protected]