Automorphisms of K3 Surfaces

• Automorphisms of K3 surfaces are isometries of the second cohomology that preserve both the lattice structure and Hodge decomposition, linking the Néron–Severi and transcendental lattices.
• The classification into symplectic, non-symplectic, and mixed automorphisms reveals distinct fixed point phenomena and group structures, providing insights into surface geometry and moduli.
• Lattice-theoretic techniques, such as chamber decomposition and Borcherds' method, enable explicit computation of automorphism groups and analysis of dynamical properties like Salem numbers and entropy.

Automorphisms of K3 Surfaces

A K3 surface is a compact, simply connected, smooth complex surface with trivial canonical bundle, or, equivalently, with trivial first Chern class and $h^{2,0}=1$. The automorphism group of a K3 surface—meaning the group of biregular automorphisms—has deep connections to lattice theory, Hodge theory, moduli, deformation theory, and arithmetic, and is tightly controlled by the geometry of both the Néron–Severi group and the transcendental lattice. The interplay between symplectic, non-symplectic, and infinite-order automorphisms gives rise to a multitude of phenomenologically distinct classes, each with a rich structure and extensive classification.

1. Fundamental Definitions and Lattice Structure

Automorphisms of a K3 surface $X$ are isometries of the integral second cohomology $H^2(X,\mathbb{Z})$ preserving the intersection form and the Hodge decomposition $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$, together with the Kähler cone. The group $\operatorname{Aut}(X)$ is understood via the induced map: $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$ and the Global Torelli theorem relates automorphisms to Hodge isometries fixing an ample class (Taki, 2018).

Key lattices:

• The Néron–Severi lattice:

$$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$

of signature $(1,\rho(X)-1)$ with $1 \le \rho(X) \le 20$ for complex K3 surfaces.

• The transcendental lattice:

$$T(X) := \mathrm{NS}(X)^\perp \subset H^2(X,\mathbb{Z})$$

Any $X$0 acts as an isometry of both $X$1 and $X$2, preserving the cup-product form.

2. Classification: Symplectic, Non-Symplectic, and Mixed Automorphisms

Symplectic Automorphisms

An automorphism $X$3 is symplectic if $X$4 on the unique (up to scale) nowhere vanishing holomorphic 2-form $X$5; equivalently, $X$6 acts trivially on $X$7.

Nikulin–Mukai–Kondo established that any finite symplectic automorphism group embeds into the Mathieu group $X$8 and that the order of a finite symplectic automorphism is at most $X$9, with the fixed locus comprising only isolated points (e.g., $H^2(X,\mathbb{Z})$0 for order $H^2(X,\mathbb{Z})$1, $H^2(X,\mathbb{Z})$2 for order $H^2(X,\mathbb{Z})$3, $H^2(X,\mathbb{Z})$4 for order $H^2(X,\mathbb{Z})$5, etc.) (Taki, 2018).

Non-Symplectic Automorphisms

A finite-order automorphism is non-symplectic if it acts on $H^2(X,\mathbb{Z})$6 by multiplication with a primitive root of unity of order $H^2(X,\mathbb{Z})$7. If the induced action is by a primitive $H^2(X,\mathbb{Z})$8th root (i.e., no proper divisor), $H^2(X,\mathbb{Z})$9 is purely non-symplectic. Such automorphisms often admit fixed loci consisting of smooth curves and isolated points, and induce cyclic group structures on the transcendental lattice, realized via modules over cyclotomic rings $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$0 (Keum, 2012, Taki, 2018).

The fixed lattice $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$1 is hyperbolic and classifiable via root systems and elementary lattices (e.g., $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$2-elementary if $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$3 has order $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$4 prime).

The possible orders $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$5 of finite order automorphisms are completely classified: $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$6 where $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$7 is Euler's totient function. The maximal possible order is $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$8 in characteristic $$H^2(X,\mathbb{C}) = H^{2,0} \oplus H^{1,1} \oplus H^{0,2}$$9 (Keum, 2012).

Mixed Cases and Infinite Automorphism Groups

Automorphisms may have symplectic and non-symplectic components and can generate, in certain situations, infinite automorphism groups, especially when the Picard number is low and the Weyl group $\operatorname{Aut}(X)$0 has infinite index in $\operatorname{Aut}(X)$1. For example, with $\operatorname{Aut}(X)$2 or $\operatorname{Aut}(X)$3, one can have infinite cyclic, infinite dihedral, or even non-abelian free groups of automorphisms (Lee, 2022, Hashimoto et al., 2023, Lee, 2024).

3. Explicit Models: High-Order Non-Symplectic Automorphisms

Purely non-symplectic automorphisms of high order are realized in explicit models, with uniqueness results in many cases:

Order 50 Example and Uniqueness

• For $\operatorname{Aut}(X)$4, any K3 surface with a cyclic automorphism of order 50 is isomorphic to

$\operatorname{Aut}(X)$5

with automorphism

$\operatorname{Aut}(X)$6

where $\operatorname{Aut}(X)$7 is a primitive 50th root of unity. The action is purely non-symplectic; no non-trivial power fixes the global 2-form (Keum, 2015).

The eigenvalue structure on the second cohomology is

$\operatorname{Aut}(X)$8

and the fixed locus of powers of $\operatorname{Aut}(X)$9 is determined by analysis of the Lefschetz formula (Keum, 2015).

Maximal-Order Examples

• The maximal possible order for complex K3 automorphisms is 66, realized, e.g., on

$\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$0

which is isolated in the moduli space ($\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$1 dimension $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$2) (Keum, 2012).

4. Lattice-Theoretic Techniques and Their Applications

The automorphism group is tightly controlled by lattice theory, enabled through several methodologies:

Chamber-Walking and Borcherds' Method

For high Picard number (e.g., singular or special K3s), automorphism groups are computed using chamber-decomposition of the nef (or ample) cone, enabled by embedding $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$3 in a high-rank even unimodular lattice (e.g., $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$4). Fundamental domains (chambers) tessellate the positive cone, with walls given by $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$5-vectors. Borcherds' method allows explicit computation of generators via wall-crossing, and the construction of relations by tracking face adjacencies (Shimada, 2013, Shimada, 16 Feb 2025).

Example: For the Apéry–Fermi K3 surface (Picard number 19),

$\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$6

Borcherds' method yields a finite generating set, including projective automorphisms, fiberwise translations, and Mordell–Weil group elements. The automorphism group acts transitively on rational curves and partitions pairs of disjoint $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$7-curves into finitely many orbits (Shimada, 16 Feb 2025).

Cyclic, Dihedral, and Free Automorphism Groups

• For generic Picard number 2, automorphism groups are typically infinite cyclic or infinite dihedral, classified by the ambiguity of the underlying rank-2 lattice and the presence of involutory isometries (Lee, 2022).
• In Picard number 3, Hashimoto–Lee demonstrate that $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$8 can be realized as a free group of arbitrarily large rank, isomorphic to a congruence subgroup of $\operatorname{Aut}(X) \to O(H^2(X,\mathbb{Z}))$9 for suitable $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$0 (Hashimoto et al., 2023).

5. Salem Numbers, Entropy, and Dynamics

For automorphisms of infinite order, the dynamical complexity is captured by the spectral radius (the dynamical degree), which is often a Salem number.

• McMullen's theorem: The characteristic polynomial of $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$1 for any $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$2 factors as the product of at most one Salem polynomial and cyclotomic polynomials (Shimada, 2015, Bayer-Fluckiger, 2021).
• For supersingular K3 surfaces in odd characteristic, explicit automorphisms with irreducible Salem polynomials of degree 22 are constructed, ensuring maximal possible entropy; these are generated by products of involutions corresponding to double-plane morphisms (Shimada, 2015).

Automorphisms with Salem numbers of small degree are also realized on non-projective K3 surfaces, with explicit conditions for the characteristic polynomial (e.g., the vanishing of global obstructions in the Bayer–Fluckiger lattice theory framework) (Bayer-Fluckiger, 2021).

6. Automorphism Groups, Picard Rank, and Arithmetic Aspects

The structure and size of $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$3 depend sensitively on the Picard rank:

• For $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$4, automorphism groups can be infinite; explicit criteria in terms of the absence of $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$5-classes and isotropic vectors are known.
• For singular K3 surfaces ($$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$6), infinite automorphism groups are realized, with explicit presentations via lattice isometries; in contrast, when the Weyl group is finite-index in $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$7, the automorphism group is finite (Shimada, 2013, Taki, 2018).
• Supersingular K3 surfaces in positive characteristic have automorphism groups reflecting Frobenius invariants—Artin invariants for supersingulars, height for finite-height surfaces—with delicate congruence conditions controlling when a given automorphism structure appears (Jang, 2013).

Finiteness properties: The automorphism group of any projective K3 surface is finitely generated (Hashimoto et al., 2023). For some explicit classes (e.g., Picard number 2 or 3), specific group-theoretic realizations (free, cyclic, dihedral) can be exhibited.

7. Notable Examples and Moduli Implications

Automorphisms are realized in moduli-theoretically distinguished loci; for many (particularly maximal-order) automorphisms, the moduli correspond to isolated points.

Order $$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$8	Example Realization	Moduli Dim
66	$$\mathrm{NS}(X) := H^{1,1}(X) \cap H^2(X,\mathbb{Z})$$9	$(1,\rho(X)-1)$0
50	$(1,\rho(X)-1)$1 in $(1,\rho(X)-1)$2	$(1,\rho(X)-1)$3
44,54...	Unique up to isomorphism	$(1,\rho(X)-1)$4

Symplectic automorphisms have their own isolated cases with fixed-point data determined entirely by the order—for order $(1,\rho(X)-1)$5, the number and type of fixed points are uniquely classified (Taki, 2018).

Infinite automorphism groups reflect large moduli (e.g., lattice polarization of generic type for free groups) and manifest in moduli of lattice-polarized or Jacobian-fibered K3 surfaces.

---

References

• (Keum, 2015) "K3 surfaces with an order 50 automorphism"
• (Hashimoto et al., 2023) "Free automorphism groups of K3 surfaces with Picard number 3"
• (Lee, 2024) "Generalized Fibonacci numbers and automorphisms of K3 surfaces with Picard number 2"
• (Lee, 2022) "K3 surfaces with Picard number two"
• (Shimada, 2013) "An algorithm to compute automorphism groups of K3 surfaces and an application to singular K3 surfaces"
• (Shimada, 2015) "Automorphisms of supersingular K3 surfaces and Salem polynomials"
• (Keum, 2012) "Orders of automorphisms of K3 surfaces"
• (Jang, 2013) "The representations of the automorphism groups and the Frobenius invariants of K3 surfaces"
• (Taki, 2018) "Automorphisms of K3 surfaces and their applications"
• (Shimada, 16 Feb 2025) "The automorphism group of an Apéry-Fermi K3 surface"
• (Bayer-Fluckiger, 2021) "Isometries of lattices and automorphisms of K3 surfaces"
• (Garbagnati et al., 2021) "Order 3 symplectic automorphisms on K3 surfaces"
• (Artebani et al., 2019) "Order 9 automorphisms of K3 surfaces"
• (Comparin et al., 2019) "On some K3 surfaces with order sixteen automorphism"
• (Artebani et al., 2011) "Symmetries of order four on K3 surfaces"
• (Bayer-Fluckiger, 2023) "Automorphisms of K3 surfaces, cyclotomic polynomials and orthogonal groups"

For additional technical details, explicit examples, and algorithmic aspects, see the referenced works.