Nikulin Root Invariant in Lattice Geometry

• The Nikulin root invariant is a lattice-theoretic invariant that captures the configuration of ADE roots and their primitive embedding in numerical lattices.
• It underlies classification structures for Enriques, K3, Nikulin, and Kummer surfaces by encoding discriminant forms, parity data, and gluing information.
• This invariant refines 2-elementary lattice classifications and informs curve orbit enumeration and moduli problems across arithmetic and geometric contexts.

Searching arXiv for recent and foundational papers directly relevant to "Nikulin root invariant" and closely related lattice-theoretic uses. The Nikulin root invariant is a lattice-theoretic invariant attached to a surface or higher-dimensional holomorphic symplectic object through a configuration of roots, its primitive closure, and the associated discriminant-form and gluing data. In the most explicit recent formulation, for an Enriques surface $Y$ it is the pair

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$

where the sum runs over the connected components $\sigma$ of the graph $\overline{\Delta}(Y)$, each $R_\sigma$ is an irreducible ADE root lattice attached to $\sigma$, and the kernel records the mod-$2$ failure of primitiveness of the embedding into the numerical lattice $S_Y=\mathrm{Num}(Y)$ (Brandhorst et al., 10 Jul 2025). In adjacent literature on K3 surfaces, Nikulin surfaces, Kummer surfaces, Nikulin-type orbifolds, and Lorentzian Kac–Moody theory, the same idea appears through closely related packages of data: invariant and anti-invariant lattices, configurations of $(-2)$-classes or other root vectors, discriminant groups, parity invariants, and explicit gluing.

1. Explicit definition on Enriques surfaces

For an Enriques surface $Y$ over an algebraically closed field of characteristic $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$0, the numerical lattice is

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$1

which is even, unimodular of signature $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$2; in fact $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$3. A class $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$4 of a smooth rational curve has self-intersection $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$5. The splitting roots are

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$6

for $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$7 the K3-cover of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$8, and concretely

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$9

The effective splitting roots are

$\sigma$0

and the set of classes of smooth rational curves is

$\sigma$1

These constructions are the starting point for the paper’s working definition of the Nikulin root invariant (Brandhorst et al., 10 Jul 2025).

The image $\sigma$2 is viewed as an undirected graph whose vertices are the classes $\sigma$3, with an edge between $\sigma$4 if $\sigma$5. Each connected component $\sigma$6 of this graph is itself a simply-laced irreducible root system of ADE type. For such a component, the local data are the rank $\sigma$7, the dimension

$\sigma$8

the kernel dimension $\sigma$9, and the type

$\overline{\Delta}(Y)$0

A key restriction is that for each connected component $\overline{\Delta}(Y)$1, $\overline{\Delta}(Y)$2, and either $\overline{\Delta}(Y)$3, or $\overline{\Delta}(Y)$4 and then $\overline{\Delta}(Y)$5 or $\overline{\Delta}(Y)$6 (Brandhorst et al., 10 Jul 2025).

This explicit definition is already highly structured. It does not record only the ADE type. It records the direct sum $\overline{\Delta}(Y)$7 and the mod-$\overline{\Delta}(Y)$8 kernel measuring how the relevant root lattice sits inside $\overline{\Delta}(Y)$9. In the exceptional cases $R_\sigma$0 and $R_\sigma$1, the kernel distinguishes primitive from non-primitive embeddings, and that distinction is later reflected in orbit-counting formulas for smooth rational curves.

2. K3 surfaces, Nikulin surfaces, and Kummer root data

In the K3 setting, several papers work with the exact ingredients of a Nikulin root invariant without always naming the object. For a K3 surface with a Nikulin involution, the basic geometric configuration consists of eight disjoint $R_\sigma$2-curves on the quotient K3 surface $R_\sigma$3,

$R_\sigma$4

together with a class

$R_\sigma$5

The associated Nikulin lattice $R_\sigma$6 is the even lattice of rank $R_\sigma$7 generated by $R_\sigma$8 and

$R_\sigma$9

with $\sigma$0 and $\sigma$1 for $\sigma$2. For a genus-$\sigma$3 Nikulin surface, the lattice

$\sigma$4

is the polarization-plus-root datum fixed inside $\sigma$5; the paper describes this as the lattice underlying the moduli problem for Nikulin surfaces of genus $\sigma$6 (Farkas et al., 2011).

On Kummer surfaces the same pattern is amplified. A Nikulin configuration is a set of $\sigma$7 disjoint smooth rational curves on a K3 surface. For $\sigma$8, the primitive closure $\sigma$9 of the lattice generated by these $2$0 $2$1-curves has discriminant group

$2$2

and its discriminant form is isometric to the discriminant form of $2$3. For a generic polarized abelian surface $2$4 with $2$5, the Néron–Severi group of $2$6 is a finite-index overlattice of $2$7, where

$2$8

The paper then constructs a second Nikulin configuration by replacing one root $2$9 with

$S_Y=\mathrm{Num}(Y)$0

which satisfies $S_Y=\mathrm{Num}(Y)$1, is orthogonal to the other $S_Y=\mathrm{Num}(Y)$2 curves, and yields a different embedding of a rank-$S_Y=\mathrm{Num}(Y)$3 orthogonal root configuration into $S_Y=\mathrm{Num}(Y)$4 (Roulleau et al., 2017).

These examples show the standard K3/Kummer content of the invariant: a primitive embedding of a root lattice, its orthogonal complement, and the discriminant-form data that control whether two root configurations are equivalent under automorphisms. In this sense, the Nikulin root invariant is not merely the list of roots; it is the root lattice together with the way it is glued into the ambient even lattice.

3. Discriminant forms, 2-elementary lattices, and parity data

A central theme in the literature is that Nikulin-style root invariants are refinements of 2-elementary lattice data. In the heterotic/K3 framework, one starts with the even self-dual lattice

$S_Y=\mathrm{Num}(Y)$5

and an involution $S_Y=\mathrm{Num}(Y)$6. The invariant lattice and orthogonal complement are

$S_Y=\mathrm{Num}(Y)$7

Both are even sublattices, and their discriminant groups are purely 2-elementary: $S_Y=\mathrm{Num}(Y)$8 The parity invariant is

$S_Y=\mathrm{Num}(Y)$9

and similarly $(-2)$0. The triple $(-2)$1 uniquely determines $(-2)$2 up to isomorphism, and $(-2)$3 is also uniquely determined up to isomorphism (Acharya et al., 2022).

The same paper refines this data by tracking the even/odd splitting of discriminant classes,

$(-2)$4

and by introducing a distinguished class $(-2)$5 with

$(-2)$6

The paper does not define a “Nikulin root invariant” explicitly, but it states that the relevant data are the pattern of root lattices inside $(-2)$7 and $(-2)$8 together with the discriminant-form and parity data $(-2)$9, and the distinguished class $Y$0 (Acharya et al., 2022).

This use clarifies a general point. In Nikulin-style classification problems, the invariant lattice, the anti-invariant lattice, and the discriminant form already provide a coarse classification. The root invariant is the refinement that remembers which root sublattices occur, how they embed, and how their classes behave modulo $Y$1. The Enriques-surface definition above is an explicit realization of exactly that principle.

4. Higher-dimensional analogues on Nikulin-type orbifolds

For irreducible symplectic orbifolds of Nikulin-type, recent work uses the same package of data in a higher-dimensional Beauville–Bogomolov setting. If $Y$2 is a Nikulin-type orbifold, then

$Y$3

and the relevant root-like subsets are

$Y$4

together with the wall-divisor sets $Y$5 and $Y$6. The paper states explicitly that it does not define or use a “Nikulin root invariant,” but the role played by such an invariant is distributed across the invariant lattice $Y$7, the coinvariant lattice $Y$8, the discriminant forms, and the subsets $Y$9, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$00, and $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$01. The deformation classification is then lattice-theoretic: two pairs $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$02 and $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$03 with finite-order symplectic automorphisms are deformation equivalent if and only if the invariant lattices are isometric, equivalently if and only if the coinvariant lattices are isometric (Brandhorst et al., 2024).

A related paper on standard involutions on Nikulin-type orbifolds makes the analogue even more explicit. There the global lattice is

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$04

and the induced involutions have coinvariant root lattices

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$05

The paper emphasizes that invariant lattice and coinvariant lattice do not determine the full structure, because the embedding of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$06 into $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$07 is not determined by its orthogonal complement. One must also specify the gluing. Lemma 2.3.8 gives the explicit half-sum generators for the correct gluing, and Theorem 2.3.10 states that if a Nikulin-type orbifold admits an embedding of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$08, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$09 or $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$10, satisfying exactly this gluing condition, then it admits a standard symplectic involution (Piroddi, 2024).

The higher-dimensional lesson is precise. In the K3 case one often speaks of the root invariant as a root lattice together with its discriminant-theoretic embedding data. In the Nikulin-type orbifold case, the same role is played by the coinvariant root lattice $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$11, its orthogonal complement, and the gluing that reconstructs the full Beauville–Bogomolov lattice. This suggests that the root invariant is best understood as an embedding invariant rather than a bare ADE label.

5. Variants in arithmetic and physical literature

The same Nikulin-style philosophy appears in several adjacent settings, although the phrase itself is used with different degrees of explicitness.

Context	Core data	Role
Enriques surface $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$12	$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$13	Controls ADE-components and curve orbits
Nikulin/Kummer K3 surface	Root configuration, primitive closure, orthogonal class, discriminant form	Distinguishes Kummer structures and moduli components
Nikulin-type orbifold	Invariant lattice, coinvariant root lattice, gluing	Classifies induced involutions and deformation types
Heterotic/K3 involution	$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$14, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$15, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$16, class $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$17	Controls allowed shifts, phases, and root content
Lorentzian Kac–Moody setting	$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$18	Encodes simple roots, Weyl chamber, Weyl vector

In Lorentzian Kac–Moody theory, the phrase is not used verbatim in the cited paper, but the reconstruction is explicit. For a hyperbolic lattice $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$19 of signature $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$20, a finite simple root system $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$21, and a Weyl vector $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$22 satisfying

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$23

the essential invariant is the isomorphism class of the triple $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$24, together with the arithmetic-type conditions ensuring automorphic correction. The paper classifies exactly $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$25 elliptic-type rank-$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$26 systems of this form. In that Gritsenko–Nikulin sense, the root invariant is the Lorentzian lattice, the simple root configuration, the Weyl chamber, and the Weyl vector (Allcock, 2012).

This usage is not identical to the Enriques-surface definition, but the family resemblance is strong. In each case the invariant packages root data with the ambient lattice structure and with a finite quadratic-form or Weyl-theoretic constraint. The common feature is that roots alone are insufficient; the invariant is the rooted embedding.

6. Classification power, applications, and scope

The most concrete application currently recorded is on Enriques surfaces. Let $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$27 be the number of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$28-orbits of connected components of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$29 of type $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$30, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$31 the number of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$32-orbits of type $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$33, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$34 the number of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$35-orbits of type $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$36, $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$37 the number of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$38-orbits of type $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$39, and $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$40 the number of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$41-orbits of type $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$42. Then

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$43

Thus the Nikulin root invariant, together with the Vinberg group, determines the number of $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$44-orbits of smooth rational curves on $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$45 (Brandhorst et al., 10 Jul 2025).

In Kummer-surface geometry, the same type of invariant distinguishes non-equivalent Nikulin configurations on the same K3 surface. For $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$46, the paper proves that the natural Nikulin configuration $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$47 and the modified configuration $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$48 are not $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$49-equivalent, so the corresponding abelian surfaces $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$50 and $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$51 are not isomorphic. The distinction is detected by the different embeddings of the $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$52-root configuration and its orthogonal polarization class in $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$53 (Roulleau et al., 2017).

In the theory of Prym and spin moduli, the same lattice-plus-root package underlies the moduli space $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$54 of genus-$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$55 Nikulin surfaces, the Prym–Nikulin locus in $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$56, and the Grassmannian model for genus $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$57. There the fixed lattice

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$58

with $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$59 generated by eight disjoint $$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$60-classes and their half-sum is the invariant that organizes the moduli problem (Farkas et al., 2011).

Two misconceptions are corrected by the recent literature. First, the phrase “Nikulin root invariant” is not used uniformly: several papers are built on the same lattice-theoretic data while explicitly not packaging them under that name (Brandhorst et al., 2024, Piroddi, 2024, Acharya et al., 2022). Second, the invariant is not only an ADE type. In the explicit Enriques-surface definition the kernel

$$\left(R := \bigoplus_{\sigma} R_\sigma,\ \ker\big(R\otimes \mathbb{F}_2 \to S_Y\otimes \mathbb{F}_2\big)\right),$$61

is essential, and in the Nikulin-type orbifold setting the gluing of invariant and coinvariant lattices is essential. A plausible implication is that the modern mathematical content of the term is best captured not by a single root system, but by a root system together with its discriminant-theoretic realization inside a fixed ambient lattice.

In this sense, the Nikulin root invariant is a unifying object across several branches of geometry and arithmetic: it records root configurations, orthogonal complements, parity and discriminant-form data, and the gluing necessary to recover the ambient lattice. Where an explicit definition is available, it serves as a complete combinatorial-lattice encoding of the relevant geometric configuration; where the term is absent, the same invariant structure still governs deformation, automorphism, and moduli classification.