---
title: Nested Hilbert Schemes
url: https://www.emergentmind.com/topics/nested-hilbert-scheme
type: topic
---

# Nested Hilbert Schemes

Nested Hilbert schemes are moduli spaces of flags of finite subschemes with prescribed lengths. For a quasi-projective scheme \(X\), they generalize the ordinary Hilbert scheme of points by replacing a single zero-dimensional subscheme with a chain \(Z_{d_1}\subset \cdots \subset Z_{d_r}\subset X\). On smooth surfaces, the simplest case of lengths differing by \(1\) is smooth and admits explicit blow-up descriptions, while larger gaps and longer flags typically introduce singularities, nontrivial birational geometry, and, in some regimes, reducibility. The subject now sits at the intersection of deformation theory, singularity theory, representation theory, virtual intersection theory, and moduli-theoretic constructions on surfaces, curves, and affine spaces [1909.04321, 2109.09002, 1709.06105].

## 1. Definitions and foundational constructions

For a non-decreasing sequence of positive integers \(\underline d=(d_1\le \cdots \le d_r)\), the nested Hilbert functor assigns to a test scheme \(B\) the set of chains of \(B\)-flat finite closed subschemes
\[
Z^{(1)}\subset Z^{(2)}\subset \cdots \subset Z^{(r)}\subset X\times B
\]
with relative lengths \(d_i\). This functor is representable by a finite-type scheme \(\Hilb^{\underline d}X\); for \(r=1\) one recovers Grothendieck’s Hilbert scheme of points [2601.16765]. On a smooth surface \(S\), a standard two-step notation is
\[
\Hilb^{(n_1,n_2)}(S),
\]
the moduli space of pairs \(Z_1\supseteq Z_2\) of zero-dimensional closed subschemes of degrees \(n_1>n_2\). Equivalently, one works with chains of ideals \(I_2\subset I_1\subset \mathcal O_S\) of colengths \(n_2,n_1\) [2109.09002].

Several specializations recur throughout the literature. The length-one incidence space
\[
S^{[n,n+1]}=\{(I_n,I_{n+1},x)\mid I_{n+1}\subset I_n,\ I_n/I_{n+1}\cong k_x\}
\]
carries natural projections to \(S^{[n]}\), \(S^{[n+1]}\), and \(S\), together with a tautological line bundle whose fiber is the one-dimensional quotient \(I_n/I_{n+1}\). Higher nested spaces include triples such as
\[
S^{[n-1,n,n+1]}
\]
and quadruple spaces built from diagrams \(I_{n-1}\leftarrow I_n\to I_{n+1}\) with an additional intermediate ideal \(I_n'\), all successive colength \(1\) [2009.11267]. For families of curves \(\pi:\mathcal C\to B\), one similarly has relative nested Hilbert schemes \(\pi^{[m,m+1]}:\mathcal C^{[m,m+1]}\to B\) whose fibers are \(C_b^{[m,m+1]}\) [1802.07234]. Over a smooth curve and a Young diagram \(\lambda\), Monavari’s double-nested variant \(C^{[\mathbf n_\lambda]}\) parametrizes arrays of subschemes \(Z_{ij}\subset C\) ordered by the partial order on \(\lambda\) [2108.10827].

Natural morphisms organize these moduli. They include forgetful maps dropping one step of the flag, residual-point maps such as \(q_n\) and \(\mathrm{res}_{3,4}\), and incidence embeddings into products of ordinary Hilbert schemes. These morphisms often control both the local structure and the global birational geometry of the nested space [1909.04321, 2501.11055].

## 2. Smooth incidence geometry and local deformation theory

For a smooth surface \(S\), the ordinary Hilbert scheme \(S^{[n]}\) is smooth of dimension \(2n\). The first nontrivial nested case retains this smoothness: \(\Hilb^{[n-1,n]}(S)\) is smooth, projective, irreducible, and of dimension \(2n\), while the equivalent notation \(S^{[n,n+1]}\) gives a smooth variety of dimension \(2n+2\) [1909.04321, 2009.11267]. This case is exceptional among nested Hilbert schemes and serves as the basic incidence correspondence from which many later constructions are built.

Its geometry is particularly explicit. The morphism
\[
S^{[n,n+1]}\to S^{[n]}\times S
\]
is the blow-up of the universal subscheme \(Z_n\subset S^{[n]}\times S\), and the map
\[
\gamma_n=(q_n,f_n):\Hilb^{[n-1,n]}(S)\to S\times S^{[n-1]}
\]
is the blow-up of \(S\times S^{[n-1]}\) along \(Z_{n-1}\) [1909.04321, 1708.00903]. In Zhao’s three-step geometry, the triple space \(S^{[n-1,n,n+1]}\) is smooth of dimension \(2n+1\) and is realized as the projectivization over \(S^{[n,n+1]}\) of a two-term complex whose cokernel is one of the tautological line bundles. The related quadruple space \(\mathcal F\) is smooth of dimension \(2n+2\) and maps birationally to two auxiliary triple spaces \( \mathfrak z_+\) and \(\mathfrak z_-\), with exceptional divisor given by a diagonal copy of \(S^{[n-1,n,n+1]}\) [2009.11267].

On \((\mathbb A^2)^{[n,n+1]}\), deformation theory has an explicit linear-algebraic form. For \(R=k[x,y]\), the tangent space at an ordinary point \([I]\in (\mathbb A^2)^{[n]}\) is
\[
T_{[I]}(\mathbb A^2)^{[n]}\cong \Hom_R(I,R/I),
\]
and for a nested pair \(I\subset J\) one has
\[
T_{(I,J)}(\mathbb A^2)^{[n,n+1]}
\cong
\ker\!\Bigl(
\Hom_R(I,R/I)\oplus \Hom_R(J,R/J)\to \Hom_R(I,R/J)
\Bigr).
\]
The torus-fixed points are indexed by a partition \(\lambda\vdash n+1\) together with a removable corner \(c\), and the fixed points in the fiber of the blow-up map over \((J,p)\) correspond to addable boxes of the smaller Young diagram. The tangent character is obtained from the usual arm-leg formula by a “shortening rule” along the distinguished row and column through \(c\) [2606.08120].

## 3. Singularities, irreducibility, and reducibility

Outside the length-one case, singular behavior is typical. For smooth connected surfaces in characteristic \(0\), the two-step nested schemes are “almost always singular” except for the trivial cases \(n_2=0\) or \(n_1-n_2=1\) [2109.09002]. Recent work has therefore focused on determining when these singularities are rational, canonical, Gorenstein, klt, or worse, and on identifying the threshold at which reducibility appears.

| Family | Geometric property | Source |
|---|---|---|
| \(\Hilb^{(n,1)}(S)\), \(\Hilb^{(n,2)}(S)\) | \(\Hilb^{(n,1)}(S)\) is normal, irreducible, and Cohen–Macaulay, and has rational singularities; \(\Hilb^{(n,2)}(S)\) is irreducible of dimension \(2n\), nonsingular in codimension \(3\), normal, Cohen–Macaulay, and has rational singularities | [2109.09002] |
| \(X^{[3,4]}\), \(X^{[2,3,4]}\), \(X^{[1,3,4]}\) | \(X^{[3,4]}\) has canonical Gorenstein singularities; \(X^{[2,3,4]}\) has canonical Gorenstein singularities; \(X^{[1,3,4]}\) has rational singularities | [2501.11055] |
| \(S^{[n,n+1,n+2]}\) | irreducible of dimension \(2n+4\), local complete intersection, and klt | [2106.06570] |

Zhao’s auxiliary spaces sharpen this singularity picture. The “minus” triple space \(\mathfrak z_-\) is an irreducible locally complete intersection of dimension \(2n+2\) with canonical singularities and hence rational singularities. The “plus” triple space \(\mathfrak z_+\) has two irreducible components meeting along a divisor, and the pair \((\mathfrak z_+,0)\) is semi-divisorial log terminal. These singularity classes are established by discrepancy computations and minimal-model-program techniques, and they are precisely the singularity conditions needed for the categorical pushforward statements used later in representation-theoretic applications [2009.11267].

Irreducibility results extend far beyond the consecutive triple \(S^{[n,n+1,n+2]}\). Gangopadhyay, Rasul, and Sebastian prove that the schemes
\[
S^{[n,m]},\quad
S^{[n,m,m+1]},\quad
S^{[n,n+1,m]},\quad
S^{[n,n+1,m,m+1]},\quad
S^{[n,n+2,m]},\quad
S^{[n,n+2,m,m+1]}
\]
are irreducible, with dimensions \(2m\) or \(2m+2\) as appropriate [2209.09003]. Ryan and Taylor further deduce irreducibility of \(S^{[n,n+2]}\), \(S^{[1,n,n+1,n+2]}\), \(S^{[1,n,n+1]}\), and \(S^{[1,n,n+2]}\) from the geometry of \(S^{[n,n+1,n+2]}\) [2106.06570].

Reducibility nevertheless occurs in several important regimes. For punctual nested Hilbert schemes on \(\mathbb A^2\), Bulois and Evain show that \(S^{[k,n]}\) is irreducible if and only if \(k\in\{0,1,n-1,n\}\), and the full nested punctual scheme \(S^{[n-k,n-k+1,\dots,n]}\) is irreducible if and only if \(k\in\{n-1,n\}\) or \(n\le 3\). In particular, \(S^{[2,n]}\) has exactly \(\lfloor n/2\rfloor\) irreducible components, each of dimension \(n-1\) [1306.4838]. Ryan and Taylor construct the first explicit reducible nested Hilbert scheme for flags of length \(k>22\) [2106.06570]. In higher ambient dimension, Graffeo and Lella produce non-smoothable elementary components on smooth \(n\)-folds for \(n\ge 4\), and a “sandwiching” construction yielding generically non-reduced elementary components [2601.16765]. A plausible implication is that irreducibility for surface nestings with small step gaps is exceptional rather than generic.

## 4. Relative, punctual, and curve-theoretic variants

For integral locally planar curves, nested Hilbert schemes behave differently from the surface case. Each fiber \(C_b^{[m,m+1]}\) is reduced and equidimensional of the expected dimension \(m+1\), and when \(C_b\) is planar it is a local complete intersection inside the ambient nested Hilbert scheme of \(\mathbb C^2\). In a locally versal family \(\pi:\mathcal C\to B\), one can shrink the base so that the entire relative nested Hilbert scheme \(\mathcal C^{[m,m+1]}\) is smooth. Under that smoothness hypothesis, the Beilinson–Bernstein–Deligne decomposition theorem applies to \(\pi^{[m,m+1]}\), and Felisetti proves a support theorem: no perverse summand in the decomposition of \(R\pi^{[m,m+1]}_*\mathbb Q\) is supported in positive codimension [1802.07234].

Monavari’s double-nested Hilbert scheme on a smooth quasi-projective curve \(C\) replaces a single chain by a Young-diagram-indexed collection \(\{Z_{ij}\}_{(i,j)\in\lambda}\) satisfying \(Z_{ij}\subset Z_{i'j'}\) whenever \((i,j)\le (i',j')\). The corresponding moduli space \(C^{[\mathbf n_\lambda]}\) is projective, and its dimension is
\[
\dim C^{[\mathbf n_\lambda]}
=
n_{00}
+
\sum_{(i,j)\in\lambda,\ i\ge1}(n_{ij}-n_{i-1,j})
+
\sum_{(i,j)\in\lambda,\ j\ge1}(n_{ij}-n_{i,j-1}).
\]
Its topological Euler characteristics satisfy the hook-length product formula
\[
\sum_{\mathbf n_\lambda} e\bigl(C^{[\mathbf n_\lambda]}\bigr)\,q^{|\mathbf n_\lambda|}
=
\prod_{\Box\in\lambda}(1-q^{h(\Box)})^{-e(C)}.
\]
Moreover, \(C^{[\mathbf n_\lambda]}\) is realized as the zero locus of a section of a vector bundle \(V\) on a smooth ambient space \(A\), which furnishes a perfect obstruction theory and virtual class
\[
\bigl[C^{[\mathbf n_\lambda]}\bigr]^{\vir}
=
c_{\mathrm{rank}(V)}(V)\cap [A].
\]
These constructions feed directly into the local stable-pairs theory of curves [2108.10827].

Punctual nested Hilbert schemes encode finer local structure. For a smooth pointed surface \((S,p)\), the punctual variety
\[
S_p^{[n_1,n_2]}
=
\{Z_1\subset Z_2\subset S\mid \chi(\mathcal O_{Z_i})=n_i,\ \Supp Z_i=\{p\}\}
\]
admits a Hilbert–Samuel stratification, and this stratification supports explicit motivic calculations. In particular,
\[
[S_p^{[2,n]}]
=
[S_p^{[n]}]\mathbb L-\mathbb L^{\,n-1}\mathbb L,
\]
and the corresponding Euler-characteristic generating series are
\[
G_2(q)=\frac{2-q}{(1-q)(1-q^2)}\,Z(q),\qquad
G_3(q)=\frac{3-q-q^2}{(1-q)(1-q^2)(1-q^3)}\,Z(q),
\]
where
\[
Z(q)=\sum_{n\ge0}\chi(S_p^{[n]})q^n=\prod_{k\ge1}(1-q^k)^{-1}.
\]
The number of irreducible components of maximal dimension satisfies
\[
\nu^{[2,n]}=\Bigl\lfloor \frac n2\Bigr\rfloor,\qquad
\nu^{[3,n]}=
\Bigl\lfloor\frac{n(n-6)}{12}\Bigr\rfloor
+
\Bigl\lfloor\frac{n-1}{2}\Bigr\rfloor
+1
\]
for all \(n\ge4\) [2503.14175].

## 5. Categorical, quiver, and birational frameworks

Nested Hilbert schemes enter categorical representation theory through geometric correspondences. Zhao’s construction of a categorical quantum toroidal action on Hilbert schemes uses the spaces \(S^{[n,n+1]}\), \(S^{[n-1,n,n+1]}\), \(\mathfrak z_\pm\), and \(\mathcal F\) to categorify the commutation of Nakajima’s Heisenberg operators and their quantum toroidal counterparts. The key geometric input is that the singularities of the relevant triple spaces are canonical or semi-dlt, so that pushforwards of structure sheaves behave as required for comparing correspondences on the smooth resolution \(\mathcal F\) [2009.11267].

Derived-category structures also reflect the incidence geometry. For a smooth projective surface \(S\), Scala proves a semiorthogonal decomposition
\[
D^b\!\bigl(\Hilb^{[n-1,n]}(S)\bigr)
=
\bigl\langle
D^b(S\times S^{[n-1]}),
D^b(S\times S^{[n-2]}),
\dots,
D^b(S\times S^{[1]}),
D^b(S)
\bigr\rangle.
\]
The same work shows that the universal ideal-sheaf functor is fully faithful if and only if \(H^1(S,\mathcal O_S)=H^2(S,\mathcal O_S)=0\), and that the \(\mathbb P^{n-1}\)-functor phenomenon occurs only for K3 surfaces [1909.04321].

Quiver and noncommutative models give alternative presentations of nested spaces. For
\[
X_n=\Tot\bigl(\mathcal O_{\mathbb P^1}(-n)\bigr),
\]
the length-\((k,k+1)\) nested Hilbert scheme is canonically isomorphic to a Nakajima-type quiver variety \(\mathcal M_\theta(v,w)\) for an enhanced Hirzebruch quiver with relations and stability parameter \(\theta\) [2403.18099]. In knot homology, Oblomkov and Rozansky define a noncommutative ambient space \(Hilb_{1,n}^{\mathrm{free}}\) and show that the cohomology sheaves of the associated \(2\)-periodic complexes \(S_\beta\) are set-theoretically supported on the ordinary nested Hilbert scheme \(Hilb_{1,n}\), the commutative locus inside the noncommutative model [1608.03227].

Birational geometry on nested Hilbert schemes is likewise explicit. For a smooth projective surface \(X\), the space \(X^{[n+1,n]}\) is the blow-up of \(X^{[n]}\times X\) along the universal subscheme \(Z^{[n]}\). When \(q(X)=0\),
\[
\Pic\bigl(X^{[n+1,n]}\bigr)\cong \Pic(X)^{\oplus 2}\oplus \mathbb Z^2,
\]
and Ryan–Yang compute nef cones for \(X=\mathbb P^2\), a Hirzebruch surface \(\mathbb F_e\), and a very general K3 surface. On Hirzebruch surfaces, this calculation recovers Butler’s projective-normality theorem via the identification \(X^{[2,1]}\cong \Bl_\Delta(X\times X)\) and a Kawamata–Viehweg vanishing argument [1708.00903].

## 6. Virtual geometry, localization, and asymptotic invariants

A central modern viewpoint treats nested Hilbert schemes as degeneracy loci. Gholampour–Thomas realize the two-step surface nesting \(\Hilb^{n_1,n_2}(S)\) as the deepest degeneracy locus of a two-term complex
\[
E^\bullet\simeq R\pi_*R\mathcal Hom(\mathcal I_1,\mathcal I_2)
\]
on \(A=\Hilb^{n_1}(S)\times \Hilb^{n_2}(S)\). This gives a canonical perfect obstruction theory and a virtual class computable by Thom–Porteous. In the broader points-and-curves setting,
\[
S^{[n_1,n_2]}_\beta
=
\{I_1(-D)\subset I_2\subset \mathcal O_S\mid [D]=\beta,\ \length(\mathcal O_S/I_i)=n_i\}
\]
is identified with a virtual resolution \(\widetilde D_1(o)\) of a first degeneracy locus, and its virtual cycle is pushed forward by a Chern-class formula. These constructions are then modified to recover the virtual cycles used in Vafa–Witten, Seiberg–Witten, local PT, and local DT theories [1709.06105, 1902.04128].

In equivariant \(K\)-theory on \(\mathbb C^2\), Minddal introduces a nested noncommutative Hilbert scheme \(ncNHilb^{(1,n)}(\mathbb C^2)\) and realizes the classical nested Hilbert scheme \(NestHilb^{n+1,n}(\mathbb C^2)\) as the commutativity locus cut out by sections of equivariant bundles. The resulting perfect obstruction theory agrees with that of Gholampour–Sheshmani–Yau. For the forgetful morphism to \(\Hilb^n(\mathbb C^2)\), the virtual structure sheaf satisfies a factorization with twist
\[
\alpha_1
=
\chi_T\bigl(\Hilb^1(\mathbb C^2),\Lambda_{-1}(\omega_{\mathbb C^2}^{[1]})\bigr)
=
1-t_1t_2,
\]
and the multivariate generating series becomes
\[
G(q,u;t_1,t_2)
=
PE\!\left(
\frac{q+u(1-t_1t_2)}{(1-t_1)(1-t_2)}
\right).
\]
A key combinatorial step is the independence of a localization correction factor from the underlying Young diagram \(\lambda\) [2606.30172].

A different ambient-space formalism works in arbitrary dimension. In the non-associative model \(naNHilb^{\underline d}(\mathbb A^n)\), one replaces the commutative associative algebra structure by a commutative non-associative one on a flag of vector bundles. The ambient space is smooth, the classical nested Hilbert scheme \(NHilb^{\underline d}(\mathbb A^n)\) is the zero locus of the associator section
\[
s_{\mathrm{ass}}=\psi_2\circ(\psi_2\otimes id)-\psi_2\circ(id\otimes\psi_2)
\]
of an associativity bundle \(E_{\mathrm{ass}}\), and the induced virtual class is computed by equivariant localization as a sum over admissible nested partitions. On the punctual locus, all virtual integrals are governed by a multivariable iterated residue formula [2512.11538].

Nested Hilbert schemes also carry asymptotic convex geometry. On \(\Hilb^{n,n+1}(\mathbb C^2)\), global sections of
\[
O(m,k)=\pi_n^*O(m)\otimes \pi_{n+1}^*O(k)
\]
are identified with the \(\mathrm{sgn}(m+k)\)-component of \(J^{m+k}\cap I^k\), where \(J\) and \(I\) are Haiman-type diagonal ideals. A lexicographic trailing-term analysis yields explicit inequalities on exponent vectors, and the resulting Newton–Okounkov body is
\[
\Delta\bigl(O(m,k)\bigr)
=
\mathbb R_{\ge0}^2\times \Delta(m+k,k),
\]
with \(\Delta(m+k,k)\) a rational-polyhedral region defined by those inequalities [2510.07420]. This suggests that, even for smooth length-one nests, the asymptotic linear series already remembers the full combinatorics of the extra point and its interaction with diagonal vanishing conditions.

Source: https://www.emergentmind.com/topics/nested-hilbert-scheme