Ulrich Wildness in Algebraic Geometry

• Ulrich wildness is a property of polarized varieties where one finds families of pairwise non-isomorphic, indecomposable H-Ulrich bundles with arbitrarily large moduli dimensions.
• The framework integrates methods from ACM theory, linear resolutions, deformation theory, and generalized Clifford algebras to characterize the growth of Ulrich bundle families.
• Concrete instances, including relative hypersurfaces, threefold scrolls over Hirzebruch surfaces, and Del Pezzo threefolds, illustrate how extension constructions and moduli space dimensions signal wild representation type.

Searching arXiv for papers on Ulrich wildness and related Ulrich bundle representation type. Ulrich wildness is the condition that a polarized variety $(X,H)$ support families of pairwise non-isomorphic indecomposable $H$-Ulrich bundles of arbitrarily large dimension; in relative form, the same phenomenon is formulated for relatively Ulrich bundles over a base. The notion measures the complexity of the Ulrich category rather than merely the existence of Ulrich bundles, and recent work places it at the intersection of ACM theory, linear resolutions, deformation theory, extension constructions, and representation theory via generalized Clifford algebras (Fania et al., 14 Jul 2025, Mondal et al., 2 Apr 2026).

1. Definitions, cohomological criteria, and complexity invariants

Let $(X,H)$ be a polarized smooth projective variety of dimension $n$. A vector bundle $E$ on $X$ is $\mathcal O_X(H)$-Ulrich if

$$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$

For threefolds this becomes the vanishing of $H^i(X,E(-H))$, $H^i(X,E(-2H))$, and $H$0 for all $H$1. Equivalent formulations used in the literature are that $H$2 is ACM and has a linear minimal free resolution over the ambient projective space, or that its pushforward under a general linear projection has a trivial resolution. The Hilbert polynomial is extremal: $H$3 and in particular $H$4 for rank $H$5 (Fania et al., 14 Jul 2025, Fania et al., 4 Jun 2026).

Ulrich bundles are Gieseker semistable and slope-semistable; in the formulations cited here, stability and slope-stability coincide for Ulrich bundles, and instability is governed by extensions of lower-rank Ulrich bundles. The Ulrich dual

$H$6

is again Ulrich and has the same rank. These basic properties make Ulrich bundles suitable for deformation-theoretic and moduli-theoretic analysis (Fania et al., 14 Jul 2025).

Two numerical invariants organize the subject. The set of Ulrich ranks is

$H$7

and the Ulrich complexity is

$H$8

Thus $H$9 means that $(X,H)$0 carries an Ulrich line bundle. In the examples discussed below, this minimal complexity occurs for decomposable threefold scrolls over Hirzebruch surfaces and for relative hyperplanes, whereas higher-degree relative hypersurfaces exhibit explicit rank-one obstructions (Fania et al., 14 Jul 2025, Mondal et al., 2 Apr 2026).

2. Wildness as a representation-type condition

Ulrich wildness is modeled on the finite/tame/wild trichotomy for representations of algebras and ACM categories. In the geometric formulation adopted in papers, a smooth projective variety $(X,H)$1 is geometrically $(X,H)$2-Ulrich wild if it supports $(X,H)$3-dimensional families of pairwise non-isomorphic indecomposable $(X,H)$4-Ulrich bundles for arbitrarily large $(X,H)$5. Equivalently, for every $(X,H)$6 there exists a family of dimension $(X,H)$7 of such bundles. This is stronger than the statement that infinitely many Ulrich bundles exist: it asserts unbounded moduli-theoretic complexity (Fania et al., 4 Jun 2026, Fania et al., 14 Jul 2025).

Two recurrent diagnostics certify wildness. The first is the existence of moduli components whose dimensions grow without bound with the rank. The second is the presence of indecomposable families with self-extension spaces growing without bound, for example bundles $(X,H)$8 with

$(X,H)$9

Stability is especially useful because stable bundles are indecomposable, so smooth or generically smooth moduli components of stable Ulrich bundles immediately produce wildness when their dimensions are unbounded (Ciliberto et al., 2022, Mondal et al., 2 Apr 2026).

The main settings treated in the cited works can be summarized as follows.

Setting	Mechanism	Wildness output
Relative hypersurfaces $n$0	Equivalence with linear representations of generalized Clifford algebras	Families $n$1 with $n$2 (Mondal et al., 2 Apr 2026)
Decomposable threefold scrolls over $n$3	Generically smooth moduli components or alternating extensions	Geometrically $n$4-Ulrich wild, often with $n$5 (Fania et al., 4 Jun 2026, Fania et al., 14 Jul 2025)
Del Pezzo threefolds and the $n$6-Veronese $n$7	Smooth moduli of stable Ulrich bundles in arbitrarily large rank	Ulrich wildness from unbounded families of stable bundles (Ciliberto et al., 2022)

This framework shows that Ulrich wildness is not a single construction but a class of mechanisms: deformation growth, extension growth, and representation-theoretic parametrization all lead to the same representation-type conclusion.

3. Relative hypersurfaces, generalized Clifford algebras, and relative wildness

For a smooth connected projective scheme $n$8 over an algebraically closed field, a locally free sheaf $n$9 of rank $E$0, and the Grothendieck projective bundle $E$1, a relative hypersurface $E$2 of degree $E$3 is the zero locus of a section of $E$4. When $E$5, the defining datum is a section $E$6. A vector bundle $E$7 on $E$8 is relatively Ulrich with respect to $E$9 if it is globally generated and

$X$0

By base change, this is equivalent to requiring that every fiber $X$1 be an Ulrich bundle on the fiber hypersurface $X$2. Such bundles are characterized by a linear presentation on the ambient projective bundle: $X$3 Conversely, any rank-$X$4 bundle on $X$5 admitting such a linear presentation is relatively Ulrich (Mondal et al., 2 Apr 2026).

The associated generalized Clifford algebra is

$X$6

A representation of $X$7 is an $X$8-algebra morphism $X$9 for a locally free sheaf $\mathcal O_X(H)$0, and one has the divisibility constraint $\mathcal O_X(H)$1. The crucial bridge is the linearization map

$\mathcal O_X(H)$2

constructed from the degree-$\mathcal O_X(H)$3 part of the representation. It satisfies

$\mathcal O_X(H)$4

so it is a globalized matrix-factorization-type object encoding the hypersurface equation. From $\mathcal O_X(H)$5, one obtains the linear resolution above; conversely, a relatively Ulrich bundle with such a resolution canonically determines a representation of $\mathcal O_X(H)$6. The result is a functorial equivalence of categories between linear Clifford representations of $\mathcal O_X(H)$7 and relatively Ulrich bundles on $\mathcal O_X(H)$8, generalizing the absolute Ulrich–Clifford correspondence of Coskun–Kulkarni–Mustopa (Mondal et al., 2 Apr 2026).

This algebraic description yields a relative wildness theorem. Under the hypotheses that $\mathcal O_X(H)$9 is flat and projective, $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$0 for $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$1, there exists a relatively simple Ulrich bundle $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$2 with $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$3, and $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$4 has wild representation type, the category of relatively Ulrich bundles on $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$5 has wild representation type. The key construction is the exact fully faithful tensor functor

$$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$6

which preserves indecomposability and injects

$$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$7

Hence one obtains families $$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$8 of indecomposable relatively Ulrich bundles with

$$H^i\!\left(X,E(-jH)\right)=0 \quad \text{for all } 0\le i\le n,\; 1\le j\le n.$$9

At the rank-one end, the same paper isolates a sharp dichotomy: $H^i(X,E(-H))$0 is relatively Ulrich if and only if $H^i(X,E(-H))$1. Thus relative hyperplanes have minimal Ulrich complexity one, whereas degree $H^i(X,E(-H))$2 hypersurfaces exhibit a homological obstruction that forces the use of matrix factorizations or generalized Clifford algebras rather than trivial geometric line bundles (Mondal et al., 2 Apr 2026).

4. Decomposable threefold scrolls over Hirzebruch surfaces

A second major class of wild examples is furnished by decomposable threefold scrolls over Hirzebruch surfaces. Let $H^i(X,E(-H))$3 be the Hirzebruch surface, let

$H^i(X,E(-H))$4

and consider

$H^i(X,E(-H))$5

The polarization $H^i(X,E(-H))$6 is very ample if and only if $H^i(X,E(-H))$7, the degree is

$H^i(X,E(-H))$8

and the canonical class is

$H^i(X,E(-H))$9

For any $H^i(X,E(-2H))$0-Ulrich bundle $H^i(X,E(-2H))$1 of rank $H^i(X,E(-2H))$2, the slope is

$H^i(X,E(-2H))$3

These explicit intersection-theoretic formulas make the scroll case especially tractable for moduli calculations and extension constructions (Fania et al., 4 Jun 2026, Fania et al., 14 Jul 2025).

One outcome is that the Ulrich complexity can be minimal. On the scrolls considered in (Fania et al., 14 Jul 2025), one has $H^i(X,E(-2H))$4, with complete classifications of $H^i(X,E(-2H))$5-Ulrich line bundles in several regimes. For example, when $H^i(X,E(-2H))$6, the only $H^i(X,E(-2H))$7-Ulrich line bundles are

$H^i(X,E(-2H))$8

and additional line bundles $H^i(X,E(-2H))$9 appear when $H$00 or $H$01. These line bundles serve as building blocks for higher-rank constructions via nontrivial extensions (Fania et al., 14 Jul 2025).

The wildness mechanisms split into two regimes. In the inherited case $H$02, the paper “Ulrich wildness of some decomposable threefold scrolls over $H$03” proves that for any $H$04 the moduli space of rank-$H$05 $H$06-Ulrich bundles with explicit Chern classes contains a generically smooth component $H$07, rational for $H$08 and unirational for $H$09, whose general point is slope-stable. Its dimension is

$H$10

Since these dimensions are unbounded, $H$11 is geometrically $H$12-Ulrich wild, with $H$13 and no slope-stable Ulrich rank gaps. In the obstructed case $H$14, generically smooth moduli components may fail because of $H$15-obstructions, yet wildness survives: alternating extensions produce arbitrarily large families of indecomposable, pairwise non-isomorphic Ulrich bundles in every rank (Fania et al., 4 Jun 2026).

A closely related construction in (Fania et al., 14 Jul 2025) treats the range $H$16. Starting from $H$17 and $H$18, one defines a sequence $H$19 by alternating nonsplit extensions

$H$20

Ulrichness persists at every step, the Chern classes are computed explicitly, and there exists a generically smooth moduli component $H$21 whose general member $H$22 is slope-stable and Ulrich. The dimensions are

$H$23

This again yields geometric $H$24-Ulrich wildness and establishes $H$25. In the special case $H$26, additional rank-$H$27 families arise from instanton monads on $H$28, providing Ulrich bundles that are neither extension-type nor pull-backs (Fania et al., 14 Jul 2025).

5. Del Pezzo threefolds, the $H$29-Veronese $H$30, and smooth stable moduli

Ulrich wildness also appears on smooth Fano threefolds through stable moduli spaces. For any smooth Fano threefold $H$31 of index two, the paper “Ulrich bundles on Del Pezzo threefolds” proves that for every integer $H$32, the moduli space of stable Ulrich bundles of rank $H$33 and determinant $H$34 is smooth of dimension $H$35. For the index-four case, namely $H$36 embedded by the $H$37-Veronese, the same statement holds for every even $H$38, while no odd-rank Ulrich bundles exist. In both cases the unbounded growth of these smooth stable moduli spaces implies Ulrich wildness (Ciliberto et al., 2022).

The proof strategy is deformation-theoretic. For a stable Ulrich bundle $H$39, smoothness of the moduli at $H$40 follows from the vanishing of $H$41, while the dimension is computed from $H$42. The same paper also develops a preliminary existence criterion for Ulrich bundles on smooth projective threefolds in terms of curves with specified numerical and cohomological properties. In Serre-type form, an Ulrich bundle $H$43 of rank $H$44 and determinant $H$45 is related to an exact sequence

$H$46

where $H$47 is a smooth curve and $H$48 has dimension $H$49. For Fano threefolds this criterion specializes to explicit numerical conditions involving the index and degree (Ciliberto et al., 2022).

The Fano examples clarify a central point in the theory. Wildness need not depend on a visible supply of Ulrich line bundles or on a direct extension tower from line bundles; it can instead be detected through smooth, high-dimensional moduli of stable higher-rank bundles. This complements the scroll and relative-hypersurface constructions, where extension theory and representation theory are more explicit.

6. Obstructions, minimality phenomena, and current directions

A persistent theme in the subject is that wildness coexists with sharp low-rank obstructions. For relative hypersurfaces $H$50, $H$51 is relatively Ulrich if and only if $H$52. For $H$53, fiberwise cohomology and Serre duality produce nonvanishing in degree $H$54, so trivial line bundles cannot satisfy the Ulrich vanishing window. This is precisely the point at which matrix factorizations and generalized Clifford algebras enter: they replace “purely geometric” rank-one candidates by homologically nontrivial constructions (Mondal et al., 2 Apr 2026).

On decomposable threefold scrolls, the analogous obstruction is modular rather than rank-one. In the obstructed regime $H$55, the pairs $H$56 and $H$57 develop higher $H$58-obstructions, preventing the kind of generically smooth components available when $H$59. Nonetheless, alternating extension arguments still yield indecomposable families in all ranks, so wildness persists even when smooth moduli geometry breaks down. By contrast, (Fania et al., 14 Jul 2025) does not claim Ulrich wildness for the general regime $H$60; there the high-rank structure and positivity of moduli dimensions remain delicate (Fania et al., 4 Jun 2026, Fania et al., 14 Jul 2025).

These results suggest a stratified picture of Ulrich wildness. At one extreme lie varieties of minimal Ulrich complexity, where line bundles generate extension towers and moduli components. At another lie higher-degree or obstructed situations, where wildness is visible only after passing to matrix-factorization, Clifford-algebra, or deformation-theoretic machinery. The explicit open directions recorded in the recent literature include sharpening existence and classification results for relatively Ulrich bundles in higher degree and rank, analyzing stability conditions and moduli through irreducible generalized Clifford representations, relaxing hypotheses such as $H$61 in relative wildness theorems, and understanding the larger-$H$62 scroll regimes where wildness is not yet asserted (Mondal et al., 2 Apr 2026, Fania et al., 14 Jul 2025).

In this sense, Ulrich wildness functions as a unifying invariant across several domains: it detects unbounded complexity in Ulrich categories, organizes the geometry of moduli spaces, and links cohomological linearity to representation-theoretic growth.