Residually Nilpotent Hitchin Systems

• Residually nilpotent Hitchin systems are moduli problems where the Higgs field has a simple pole with its residue constrained to a nilpotent orbit closure via a Jacobson–Morozov model.
• The construction employs Jacobson–Morozov resolutions and spectral curve analysis to derive explicit local and global Hitchin bases, finite covers, and Prym-type abelianizations.
• Integrating Springer theory and parabolic constraints, these systems yield corrected affine Hitchin bases that ensure connected generic fibers and capture refined representation-theoretic properties.

Residually nilpotent Hitchin systems are Hitchin-type moduli problems in which the Higgs field is allowed a simple pole at a marked point, but its residue is constrained to a prescribed nilpotent orbit closure, or to the corresponding Jacobson–Morozov model, rather than merely to an arbitrary parabolic nilradical. In the strongest current formulation, this structure was introduced for $G=\mathrm{SO}_{2n}$ to study parabolic type-$D$ Hitchin systems, to make the local nilpotent behavior compatible with Springer theory, and to identify the correct “true” Hitchin base when the naive coefficient space is singular or non-normal (Wang et al., 21 Aug 2025). Closely related constructions appear in strongly parabolic and parahoric Hitchin systems, in orbit-closure Hitchin systems of types $B$ and $C$, and in compactified families of meromorphic Higgs bundles with prescribed nilpotent residues over the moduli of stable pointed curves (Su et al., 2019, Baraglia et al., 2016, Wang et al., 2024, Donagi et al., 2024).

1. Jacobson–Morozov formulation and the basic moduli problem

For the type-$D$ construction, one starts with a smooth pointed curve $(\Sigma,x)$, the group $G=\mathrm{SO}_{2n}$, and a nilpotent orbit $\mathbf O\subset \mathfrak{so}_{2n}$ represented by $e$. Choosing an $\mathfrak{sl}_2$-triple $D$0 yields the grading

$D$1

and the Jacobson–Morozov resolution

$D$2

The associated moduli space $D$3 parametrizes principal $D$4-bundles $D$5 on $D$6, Higgs fields $D$7, a $D$8-reduction of $D$9 at $B$0, and the condition that $B$1 land in the associated $B$2-bundle. In orthogonal-bundle language this is equivalently a tuple $B$3 with $B$4 preserving the filtration and landing in the image of $B$5 (Wang et al., 21 Aug 2025).

The adjective “residually nilpotent” refers exactly to this local condition: the Higgs field has a simple pole at $B$6, but its residue is constrained to lie in a nilpotent orbit closure, more precisely in the Jacobson–Morozov model $B$7. This is a refinement of ordinary parabolic Higgs bundles. It is designed so that the generic geometry of the Hitchin system reflects detailed representation-theoretic properties of the chosen nilpotent orbit, including specialness and generalized Springer data. In type $B$8, the moduli space $B$9 has two connected components $C$0, and for very even partitions one must further distinguish the two very even orbits $C$1 (Wang et al., 21 Aug 2025).

A related formal-local usage occurs in the type-$C$2 literature: over the formal disc $C$3 with fraction field $C$4, a local principal $C$5-Higgs bundle is called residually nilpotent associated to a nilpotent orbit $C$6 if $C$7. This local notion is used there as the marked-point model controlling global parabolic and orbit-closure Hitchin fibers (Wang et al., 2024).

2. Hitchin map, spectral curve, and local factorization

For $C$8 with partition $C$9, the type-$D$0 Hitchin map is defined by the characteristic-polynomial coefficients together with the Pfaffian. Writing

$D$1

the naive Hitchin base is

$D$2

and if

$D$3

then

$D$4

The Pfaffian is defined from the $D$5-structure by wedging $D$6 and using the framing $D$7 (Wang et al., 21 Aug 2025).

Given $D$8, the spectral curve

$D$9

is defined by

$(\Sigma,x)$0

It carries the involution $(\Sigma,x)$1. Denoting by $(\Sigma,x)$2 its normalization and by $(\Sigma,x)$3 the quotient, one isolates an open subset $(\Sigma,x)$4 on which $(\Sigma,x)$5 has only simple zeros away from $(\Sigma,x)$6 and the local equation near $(\Sigma,x)$7 is generic in the Kazhdan–Lusztig stratum. This genericity is what allows local decomposition of lattices and Higgs bundles according to the irreducible factors of the local characteristic polynomial (Wang et al., 21 Aug 2025).

A major technical input is the decomposition theorem for local Higgs bundles. If the local characteristic polynomial is generic in the stratum corresponding to $(\Sigma,x)$8 and factors as $(\Sigma,x)$9, then, after suitable grouping into factors $G=\mathrm{SO}_{2n}$0, one obtains

$G=\mathrm{SO}_{2n}$1

In type $G=\mathrm{SO}_{2n}$2 the behavior depends on self-dual and dual pairs of local branches under $G=\mathrm{SO}_{2n}$3. The partition decomposes into blocks of types D1, D1$G=\mathrm{SO}_{2n}$4, and D2. D2 blocks are exactly the source of nontrivial abelian covers in generic Hitchin fibers, while D1$G=\mathrm{SO}_{2n}$5 blocks are exactly the source of singularities and finite normalization phenomena on the Hitchin base (Wang et al., 21 Aug 2025).

At the purely local level, type $G=\mathrm{SO}_{2n}$6 already exhibits additional coefficient constraints. For a tame type-$G=\mathrm{SO}_{2n}$7 Hitchin system with nilpotent residue in a special orbit $G=\mathrm{SO}_{2n}$8, even-type constraints occur iff the Hitchin partition has even parts, and odd-type constraints occur iff there is a marked pair $G=\mathrm{SO}_{2n}$9 of odd parts with $\mathbf O\subset \mathfrak{so}_{2n}$0. The odd-type choices are controlled by

$\mathbf O\subset \mathfrak{so}_{2n}$1

where $\mathbf O\subset \mathfrak{so}_{2n}$2 is the number of marked pairs, so a fixed nilpotent residue orbit can admit $\mathbf O\subset \mathfrak{so}_{2n}$3 local Hitchin bases (Balasubramanian et al., 2023).

3. Generic fibers, Prym-type abelianization, and self-duality

For $\mathbf O\subset \mathfrak{so}_{2n}$4, the generic fiber

$\mathbf O\subset \mathfrak{so}_{2n}$5

is analyzed by combining the local decomposition theorem with a parabolic BNR description. One obtains a natural morphism from $\mathbf O\subset \mathfrak{so}_{2n}$6 to the Prym geometry of the normalized spectral curve. The neutral Prym is

$\mathbf O\subset \mathfrak{so}_{2n}$7

and one also considers

$\mathbf O\subset \mathfrak{so}_{2n}$8

where $\mathbf O\subset \mathfrak{so}_{2n}$9 is the fixed-point divisor of $e$0. The reconstruction of a global Higgs bundle from a line bundle on $e$1 is not unique: at each D2 block one must choose an $e$2-isotropic subspace in a local orthogonal quotient. These choices produce finite covers of the Prym by abelian varieties with $e$3-torsion kernel, and the resulting fiber product over all D2 blocks defines an abelian variety $e$4 (Wang et al., 21 Aug 2025).

The generic-fiber theorem states that for generic $e$5, where $e$6 is the true affine base,

$e$7

is a torsor under $e$8. The abelian variety $e$9 is a finite cover of the Prym of the normalized spectral curve, hence it carries a natural polarization, and the generic fiber becomes connected after passing from the naive base to the true base. A common simplification is to identify the generic fiber with the Prym itself; in type $\mathfrak{sl}_2$0 that is generally false. The generic residually nilpotent Hitchin fiber is typically a finite $\mathfrak{sl}_2$1-isogeny cover of a Prym, with the covering determined by the local nilpotent-orbit data (Wang et al., 21 Aug 2025).

The same analysis yields a geometric characterization of special nilpotent orbits. A type-$\mathfrak{sl}_2$2 partition $\mathfrak{sl}_2$3 is special iff its transpose lies in $\mathfrak{sl}_2$4, equivalently iff it contains no D2 block of the form

$\mathfrak{sl}_2$5

Then

$\mathfrak{sl}_2$6

Equivalently, the generic Hitchin fiber is a torsor under a self-dual abelian variety exactly for special nilpotent orbits. Self-duality fails precisely when there are D2 blocks of length $\mathfrak{sl}_2$7; thus the generic abelian variety is self-dual iff every D2 block has length $\mathfrak{sl}_2$8 (Wang et al., 21 Aug 2025).

4. The true Hitchin base in type $\mathfrak{sl}_2$9

One of the central discoveries of the theory is that in type $D$00 the coefficient space $D$01 is often not the full affine base of the integrable system. The correct affine base is the affinization

$D$02

called the Coulomb Hitchin base. The naive coefficient space sits inside a larger affine space by quadratic square-root relations. Fixing a local coordinate $D$03 at $D$04, one defines

$D$05

by taking leading coefficients at the prescribed pole orders. For each maximal string of equal D1$D$06-blocks one constructs a polynomial $D$07, and the condition is that every $D$08 be a square. If $D$09 is the common zero-locus of these square relations and $D$10 its normalization, then

$D$11

is the normalization map (Wang et al., 21 Aug 2025).

Accordingly, the Hitchin map factors as

$D$12

with $D$13 generically one-to-one over $D$14. Algebraically, the normalization is expressed by degree-$D$15 Veronese pieces. If $D$16 is not very even, $D$17 is a product of an affine space and affine Veronese varieties of degree $D$18; if $D$19 is very even, it is disconnected, a disjoint union of two isomorphic affine spaces. Thus the true base is explicit and algebraic, but it is not, in general, the naive coefficient vector space (Wang et al., 21 Aug 2025).

This point is easy to blur in type $D$20, where local nilpotent constraints are often read off solely from pole and zero orders. Type $D$21 is different. In the local type-$D$22 analysis of special nilpotent residues, the coefficient space is cut out not only by valuation data but also by polynomial square relations among leading coefficients; the smooth local Hitchin image is obtained by retaining square-root coordinates rather than eliminating them. In that description, even-type constraints occur iff the partition has even parts, and the choice of local Hitchin base can depend on a subgroup $D$23 (Balasubramanian et al., 2023).

The type-$D$24 smoothness criterion is correspondingly combinatorial. The true base $D$25 is singular iff there exists a D1$D$26-block $D$27 with some $D$28 such that both $D$29 and $D$30 are not of type D1$D$31. If $D$32 is smooth, it is isomorphic to an affine space. The naive base $D$33 is singular iff there exists a D1$D$34-block in the partition; if $D$35 is smooth, it is also an affine space. A common misconception is therefore that the failure of the coefficient base is a purely local nuisance. In fact, the normalization is part of the global integrable-system structure, and it is precisely what restores connected generic fibers (Wang et al., 21 Aug 2025).

5. Relation to strongly parabolic and parahoric Hitchin systems

Residually nilpotent Hitchin systems sit inside a broader landscape of Higgs moduli with nilpotent local behavior, but they should not be conflated with every occurrence of nilpotence in Hitchin theory. In the strongly parabolic $D$36 system, one fixes flags $D$37 at marked points and requires

$D$38

This means the fiber endomorphism at each marked point strictly lowers a finite filtration, hence is nilpotent. The corresponding strongly parabolic Hitchin map lands in a smaller base

$D$39

its generic spectral curve is typically singular at the marked points, and for $D$40 the generic fiber is identified with a Picard of the normalization,

$D$41

The strongly parabolic global nilpotent cone $D$42 is equi-dimensional of half the dimension of the total space, and if the Higgs moduli is smooth then $D$43 is flat and surjective (Su et al., 2019).

In the parahoric setting, one replaces a parabolic by a parahoric Bruhat–Tits group scheme $D$44 over a curve. The cotangent stack $D$45 is the stack of parahoric Higgs bundles, and in the parabolic case the residue condition is explicitly

$D$46

with $D$47 the nilpotent radical of the parabolic Lie algebra. The parahoric global nilpotent cone

$D$48

is isotropic, the Hitchin Hamiltonians Poisson commute, the base $D$49 has dimension $D$50, and generic fibers are $D$51-gerbes over disjoint unions of abelian varieties. This yields complete integrability of the parahoric Hitchin map (Baraglia et al., 2016).

These comparisons clarify an important distinction. A residually nilpotent Hitchin system prescribes nilpotent local type at marked points, typically via a Jacobson–Morozov or orbit-closure model. By contrast, the global nilpotent cone refers to the fiber over $D$52 of a Hitchin map, that is, the locus where all characteristic coefficients vanish globally. The two notions interact, but they are not identical. Strongly parabolic and parahoric theories show that nilpotent local constraints are compatible with generic abelianization on the normalization of singular spectral curves; the residually nilpotent $D$53 theory sharpens this by tying the generic fiber and the affine base directly to Springer-theoretic data (Su et al., 2019, Baraglia et al., 2016).

6. Richardson orbits, finite covers, and degenerating families

The original motivation for introducing residually nilpotent Hitchin systems in type $D$54 was twofold: to analyze Hitchin systems with singular/nilpotent local behavior in a way compatible with Springer theory, and to determine the true base of parabolic $D$55-Hitchin systems. If $D$56 is a parabolic with nilradical $D$57, the ordinary parabolic moduli space $D$58 has the same naive Hitchin base as the residually nilpotent system associated with the Richardson orbit $D$59, but its generic fibers behave differently. For $D$60, there is a natural map

$D$61

of degree $D$62, where $D$63 is the generalized Springer map. Moreover, $D$64 has exactly $D$65 connected components, and each component maps isomorphically onto $D$66. If the Levi type is

$D$67

then

$D$68

Thus generic parabolic fibers over the naive base are generally disconnected, with disconnectedness measured exactly by generalized Springer theory (Wang et al., 21 Aug 2025).

The correction is a further finite cover

$D$69

constructed by adjoining new square roots, or geometrically, new Pfaffians associated with local direct summands $D$70. The parabolic Hitchin map factors through $D$71, this finite cover has degree $D$72, $D$73 is isomorphic to an affine space, and for generic $D$74 the fiber is a torsor under the self-dual abelian variety $D$75. This is the type-$D$76 realization of the “true Hitchin base” predicted by Tachikawa (Wang et al., 21 Aug 2025).

Orbit-closure and parabolic systems behave similarly in types $D$77 and $D$78, but with a different defect. There one constructs orbit-closure Hitchin systems and local residually nilpotent models associated to nilpotent orbits in $D$79 and $D$80. On the type-$D$81 side, generic fibers are Prym torsors on the normalization of the singular spectral curve; on the type-$D$82 side, generic fibers are torsors over finite covers of that Prym, with the defect measured by the partition invariant $D$83. If $D$84, the generic orbit-closure fibers are not dual abelian varieties, and duality is restored only after passing to the corresponding parabolic systems, where the finite covers are governed by Lusztig’s canonical quotient (Wang et al., 2024).

The same local-nilpotent viewpoint extends to degenerating families of pointed curves. For meromorphic Gieseker Higgs bundles of rank $D$85 on stable pointed curves, with residues constrained to prescribed nilpotent conjugacy classes $D$86, the naive family of coefficient spaces is replaced by a modified Hitchin base $D$87. The compactified Hitchin morphism

$D$88

is proper and flat, and $D$89 is a vector bundle of rank

$D$90

In the tame $D$91 family over $D$92, the corrected line bundles $D$93 similarly assemble the Hitchin bases into a vector bundle, and restricted nodes are encoded by a pair $D$94 consisting of a nilpotent orbit $D$95 and a simple subgroup $D$96 governing the surviving center parameters (Donagi et al., 2024, Balasubramanian et al., 2020).

This suggests a broad structural picture. Residually nilpotent Hitchin systems are not merely meromorphic analogues of ordinary Hitchin fibrations with an imposed residue condition. They are the mechanism through which nilpotent local singularities, singular spectral curves, normalization-based abelianization, generalized Springer theory, and corrected affine Hitchin bases become parts of a single geometric framework.