Geometric Invariant Theory (GIT) Essentials

Updated 12 May 2026

Geometric Invariant Theory is a framework in algebraic geometry that constructs quotients of algebraic varieties through group actions, enforcing stability to build meaningful moduli spaces.
It employs tools like graded rings and the Hilbert–Mumford criterion to determine stability, leading to projective quotients with applications in vector bundles, hypersurfaces, and flag varieties.
Advanced computational methods and the study of VGIT enhance GIT's capacity to stratify varieties and drive birational transformations in modern moduli problems.

Geometric Invariant Theory (GIT) is a foundational framework in algebraic geometry for constructing quotients of algebraic varieties by group actions in a manner that retains rich geometric structure. Developed primarily by Mumford, GIT has proved essential in moduli theory, birational geometry, and the construction of moduli spaces with applications extending from vector bundles to moduli of higher-dimensional manifolds, and more recently to actions by non-reductive groups. The theory provides an algebraic and combinatorial apparatus for analyzing stability, constructing quotients, and understanding stratification in moduli problems.

1. Algebraic Structure and Quotients

Let $G$ be a reductive linear algebraic group acting linearly on a projective variety $X$ , with a chosen $G$ -linearized ample line bundle $L\rightarrow X$ . The key construction is the graded ring of invariants:

$R := \bigoplus_{m\geq 0} H^0(X, L^m)^G,$

which is finitely generated by reductivity. The projective GIT quotient is then defined as

$X^{ss}(L)//G := \mathrm{Proj}\, R,$

parametrizing $G$ -orbits of "semistable points" in $X$ . The quotient $X^{ss}(L)//G$ is itself a projective variety and, set-theoretically, parametrizes $S$ -equivalence classes of closed orbits of semistable points, with a geometric quotient on the subset of stable points (polystable orbits with finite stabilizer) (Lakhani, 2010, Dervan et al., 2023, Bérczi et al., 2015).

The construction extends to more general ambient spaces and line bundles, and underlies the formation of moduli spaces in algebraic geometry—for example, moduli of (semi)stable vector bundles, moduli of pointed curves, and Hilbert schemes (Zamora et al., 2019, Giansiracusa et al., 2011, Gulbrandsen et al., 2016).

2. Stability, Semistability, and the Hilbert–Mumford Criterion

GIT divides $X$ 0 into unstable, semistable, and stable loci, dictated by the absence or presence of invariant sections not vanishing at a given point. Explicitly:

$X$ 1 is semistable if there exists $X$ 2 and $X$ 3 with $X$ 4.
$X$ 5 is stable if the $X$ 6-orbit is closed in the semistable locus and stabilizer is finite.
$X$ 7 is unstable otherwise.

The Hilbert–Mumford criterion gives a numerical check for (semi)stability. For every 1-parameter subgroup $X$ 8:

$X$ 9

with

$G$ 0 semistable iff $G$ 1 for all $G$ 2,
$G$ 3 stable iff $G$ 4 for all $G$ 5 (Lakhani, 2010, Hattori et al., 2022, Zamora et al., 2019, Hanson et al., 24 Jun 2025, Bérczi et al., 2015).

Specialized computational approaches, such as partial orderings of monomials and explicit Weyl group reductions, streamline the process for explicit moduli problems and are formalized in packages like CompGIT (Hanson et al., 24 Jun 2025).

3. Stratification, Variation, and Birational Geometry

GIT stratifies projective varieties not only by stable/semistable loci but further by the type and severity of instability, with deep connections to canonical filtrations such as the Harder–Narasimhan filtration for vector bundles (Zamora et al., 2019).

The variation of GIT (VGIT) studies how the quotient depends on the chosen linearization (polarization). As shown by Dolgachev–Hu and Thaddeus, moving within the ample cone leads to a chamber structure with wall-crossings corresponding to birational flips between GIT quotients (Dervan et al., 2023, Seppänen et al., 2015, Giansiracusa et al., 2011). The “master space” constructed by Dervan–Reboulet parametrizes all possible GIT quotients across all birational models, capturing the full birational geometry of GIT quotients (Dervan et al., 2023).

In moduli problems (e.g., $G$ 6, Hilbert schemes), these structures produce a rich array of compactifications and flip-like birational morphisms between them, elucidating connections between disparate moduli spaces (Giansiracusa et al., 2011).

4. GIT for Moduli Spaces: Vector Bundles, Hypersurfaces, Flag Varieties

GIT underpins the construction of moduli spaces by providing well-behaved quotients that encode geometric features of objects up to isomorphism:

Vector bundles: Moduli spaces of (semi)stable bundles are GIT quotients of suitable Quot or parameter schemes by $G$ 7, with the slope stability condition matching the GIT numerical criterion under an appropriate embedding (Zamora et al., 2019). The Harder–Narasimhan filtration is mirrored by the choice of optimal destabilizing 1-PS.
Linear systems of hypersurfaces: For a Grassmannian of $G$ 8-planes in $G$ 9, stability criteria can be universally reduced to stability of unions of $L\rightarrow X$ 0 hypersurfaces, making analysis of GIT stability for linear systems tractable and generalizing classical cases (pencils, nets, etc.) (Hattori et al., 2022).
Flag varieties and chambers: GIT applied to actions on flag varieties yields explicit chamber decompositions and provides concrete models—e.g., families of Mori dream spaces associated to chambers, as for actions by principal $L\rightarrow X$ 1 (Seppänen et al., 2015). Walls in the chamber structure signal critical changes in the birational type of the quotient.
Singularity stratification: In problems such as quintic threefolds, non-stable locus classification is tied to explicit singularity types and their geometric combinatorics (Lakhani, 2010).

5. Non-Reductive GIT: Graded Unipotent Radicals and Enveloping Quotients

Classical GIT is intimately tied to reductive groups; the extension of GIT to non-reductive group actions involves additional structure:

When the unipotent radical $L\rightarrow X$ 2 is graded (i.e., there exists a central $L\rightarrow X$ 3 acting with strictly positive weights on $L\rightarrow X$ 4), one constructs an enlarged group $L\rightarrow X$ 5, facilitating finite generation of invariants after suitable twisting (Bérczi et al., 2016, Bérczi et al., 2017, Bérczi et al., 2016).
Stable loci are described in terms of $L\rightarrow X$ 6-attracting sets and require careful control of stabilizer dimensions.
Projective enveloping quotients: After imposing semistability equals stability and twisting, one forms the quotient

$L\rightarrow X$ 7

recovering a projective quotient with fibers corresponding exactly to orbits in the stable locus (Bérczi et al., 2016, Bérczi et al., 2016, Bérczi et al., 2015).

In practice, non-reductive GIT is crucial for constructing and stratifying moduli of unstable objects—e.g., sheaves or curves with fixed Harder–Narasimhan or singularity type (Bérczi et al., 2017, Bérczi et al., 2017, Bérczi et al., 2016). Explicit recursive procedures, including equivariant blow-ups, enforce stability conditions where classical GIT does not immediately apply.

6. Applications, Computational Aspects, and Recent Directions

GIT serves as the computational bedrock for a wide array of moduli constructions:

CompGIT and algorithmic GIT: Automated software tools, such as CompGIT, implement the stability and orbit stratification machinery for reductive GIT problems, streamlining the analysis of explicit examples and making large-scale computations feasible (Hanson et al., 24 Jun 2025).
Birational invariants and periods: GIT techniques underpin recent progress in representation theory, such as the calculation of the degree and period of stretched Kostka quasi-polynomials via GIT quotients of flag varieties (Besson et al., 2024).
Analytic and symplectic applications: Analytic GIT constructions, as in moduli of holomorphic vector bundles over Kähler manifolds, build smooth “classifying spaces” for polystable objects, with the Weil–Petersson form and Quillen metric becoming geometric data on the GIT quotient (Buchdahl et al., 2020).
Integrable systems and algebraic dynamics: GIT quotients parametrize moduli of dynamical data, such as conjugacy classes of endomorphisms with marked points, and provide natural ambient spaces for the algebraic study of integrable flows (Weinreich, 2021).

Limitations of current theory include the reliance on a $L\rightarrow X$ 8-equivariant embedding into projective space and the requirement of irreducible or simple group actions in most explicit calculations; ongoing developments target the extension to more general group actions and computational scalability (Hanson et al., 24 Jun 2025, Bérczi et al., 2016).