Schubert calculus
Schubert calculus is a field of algebraic geometry focused on computing intersections of subvarieties in Grassmannians and flag varieties through the structure of their cohomology rings, where Schubert classes form a distinguished basis for enumerative invariants.[1][2][3] Originating in the late 19th century, Schubert calculus was pioneered by Hermann Schubert in his 1879 work Kalkül der abzählenden Geometrie, which addressed classical enumerative problems such as determining the number of lines in projective space intersecting four given lines or planes containing a specified line and point.[1] These methods gained rigorous foundation in the 20th century, partly in response to David Hilbert's 15th problem on axiomatizing enumerative geometry.[1] Central to the subject are flag manifolds, which parametrize chains of nested vector subspaces of a fixed dimension in \mathbb{C}^n, including the Grassmannian \mathrm{Gr}_k(\mathbb{C}^n) of k-dimensional subspaces and the full flag variety \mathrm{Fl}(\mathbb{C}^n).[2][4] Schubert varieties are defined as closures of orbits under the action of a Borel subgroup on these flag manifolds, indexed by partitions (for Grassmannians) or elements of the symmetric group S_n (for full flags), and they satisfy properties like being Cohen-Macaulay with torus actions featuring isolated fixed points.[1][4] The corresponding Schubert cycles, or classes [\Omega_\lambda] in the cohomology ring H^*(\mathrm{Gr}(n,k)), provide a basis over \mathbb{Z}, and the ring structure is determined by the product rule \sigma_u \cdot \sigma_v = \sum_w c_{u,v}^w \sigma_w, where the nonnegative integer coefficients c_{u,v}^w count geometric intersections and exhibit positivity properties.[2][3] These coefficients are computed combinatorially via rules such as the Pieri formula for products with special Schubert classes and the Littlewood-Richardson rule for general cases, linking the subject to symmetric functions and representation theory.[4] Beyond classical enumerative geometry, Schubert calculus extends to equivariant and quantum settings, influencing areas like combinatorics through the Bruhat order and Young's lattice, algebraic geometry via singularity studies, and number theory through connections to cohomology rings and invariants.[2][4] For instance, in the Grassmannian \mathrm{Gr}(2,4), the intersection of four codimension-1 Schubert varieties yields exactly two points, illustrating the precise counting enabled by these tools.[4]Preliminaries
Grassmannians
The Grassmannian variety, denoted \operatorname{[Gr](/page/GR)}(k,n), is defined as the set of all [k](/page/K)-dimensional subspaces of \mathbb{C}^n. It carries a natural structure as a smooth projective algebraic variety, parametrized by the choice of basis for each subspace modulo the action of \mathrm{GL}([k](/page/K),\mathbb{C}).[5] A key realization of \operatorname{[Gr](/page/GR)}(k,n) is via the Plücker embedding, which maps it into the projective space \mathbb{P}(\wedge^[k](/page/K) \mathbb{C}^n) by associating to each [k](/page/K)-plane the wedge product of a basis, yielding Plücker coordinates given by the determinants of [k](/page/K) \times [k](/page/K) minors of a matrix whose columns span the subspace. This embedding is closed, making \operatorname{[Gr](/page/GR)}(k,n) a projective subvariety, and it induces the ample line bundle \mathcal{O}(1) on the Grassmannian, which plays a central role in its ample cone and Picard group.[5][6] The dimension of \operatorname{Gr}(k,n) is k(n-k), reflecting the degrees of freedom in choosing k independent vectors in \mathbb{C}^n up to scaling. The general linear group \mathrm{GL}(n,\mathbb{C}) acts transitively on \operatorname{Gr}(k,n) by linear transformations, with the stabilizer of a fixed k-plane being a parabolic subgroup isomorphic to \mathrm{GL}(k,\mathbb{C}) \times \mathrm{GL}(n-k,\mathbb{C}), which admits a natural decomposition into affine cells.[5][7] Pioneered by Hermann Grassmann in his 1844 work Die lineale Ausdehnungslehre, ein neuer Zweig der Mathematik, the Grassmannian provides a foundational framework for higher-dimensional geometry, enabling coordinate representations of subspaces that extend classical Euclidean methods. In enumerative geometry, it serves as the ambient space for Schubert calculus, where special subvarieties known as Schubert varieties encode intersection-theoretic counts of geometric objects.[8][1]Partitions and Schubert varieties
In the context of the Grassmannian \mathrm{Gr}(k,n), which parametrizes k-dimensional subspaces of an n-dimensional vector space, Schubert varieties are indexed by partitions \lambda = (\lambda_1 \geq \lambda_2 \geq \cdots \geq \lambda_k \geq 0) that fit inside a k \times (n-k) rectangle, meaning \lambda_i \leq n-k for all i and the length l(\lambda) \leq k. The size of the partition is |\lambda| = \sum_i \lambda_i.[1] Fix a complete flag of subspaces F_\bullet in the n-dimensional space, such as the standard flag F_i = \langle e_1, \dots, e_i \rangle where \{e_1, \dots, e_n\} is the standard basis. The Schubert variety \Omega_\lambda is the closure of the Schubert cell \Omega_\lambda^\circ consisting of k-planes V \in \mathrm{Gr}(k,n) satisfying \dim(V \cap F_{n-k+i - \lambda_i}) = i for $1 \leq i \leq k, or equivalently, \Omega_\lambda = \{ V \in \mathrm{Gr}(k,n) \mid \dim(V \cap F_{n-k+i - \lambda_i}) \geq i \ \forall \, 1 \leq i \leq k \}. In terms of the conjugate partition \lambda', this can be expressed as \dim(V \cap F_{n-i}) \geq \lambda_i' for appropriate indices. The Schubert cell \Omega_\lambda^\circ is an affine space of dimension k(n-k) - |\lambda|, providing a cell decomposition of the Grassmannian.[1][9] The codimension of \Omega_\lambda in \mathrm{Gr}(k,n) is |\lambda|. Schubert varieties are normal and Cohen--Macaulay, possessing rational singularities, which implies that resolutions like the Bott--Samelson desingularization have fibers with cohomology vanishing above degree zero. They have smooth points except in degenerate cases, and \Omega_\lambda is smooth if and only if it is a linearly embedded sub-Grassmannian, corresponding to a rectangular partition.[9][10] The inclusions of Schubert varieties reflect the Bruhat order on the indexing set of partitions: \Omega_\mu \subset \Omega_\lambda if and only if \mu \geq \lambda componentwise (i.e., \mu_i \geq \lambda_i for all i, with the same length), which aligns with the reverse Bruhat order on the corresponding minimal-length coset representatives in the Weyl group. This partial order governs the stratification and geometry of the Grassmannian.[1][9]Notations and conventions
Standard index notation
In Schubert calculus, computations typically begin with the selection of a complete flag F_\bullet in the complex vector space \mathbb{C}^n, consisting of a nested sequence of subspaces \{0\} = F_0 \subset F_1 \subset \cdots \subset F_n = \mathbb{C}^n where \dim F_i = i for each i.[1] The standard choice for this flag is F_i = \operatorname{[span](/page/Span)}\{e_1, \dots, e_i\}, where \{e_1, \dots, e_n\} denotes the standard basis of \mathbb{C}^n.[9][1] Schubert varieties in the Grassmannian \operatorname{Gr}(k, n) are then indexed relative to this fixed flag, using partitions \lambda = (\lambda_1 \geq \lambda_2 \geq \cdots \geq \lambda_k \geq 0) that fit within the k \times (n-k) rectangle, meaning $0 \leq \lambda_i \leq n-k for all i.[9][1] The cohomology class of the Schubert variety \Omega_\lambda associated to the partition \lambda is denoted by \sigma_\lambda = [\Omega_\lambda] in H^{2|\lambda|}(\operatorname{Gr}(k,n); \mathbb{Z}), where |\lambda| = \sum \lambda_i is the codimension of \Omega_\lambda.[9][1] For the opposite Schubert varieties, defined relative to the opposite flag, the indexing employs the complementary (dual) partition \lambda^\vee, obtained by rotating \lambda 180 degrees in the rectangle to fit the dimensions (\lambda^\vee_i = (n-k) - \lambda_{k+1-i}).[9][1] Intersections between a Schubert variety \Omega_\lambda (with respect to a flag F) and an opposite Schubert variety \tilde{\Omega}_\mu (with respect to an opposite flag E) exhibit rigidity: when |\lambda| + |\mu| = \dim \operatorname{Gr}(k,n) = k(n-k) and \mu = \lambda^\vee (the dual partition obtained by rotating \lambda in the rectangle), their intersection is generically transverse, consisting of a single reduced point.[9] This transversality holds for generic choices of the flags and underpins the well-definedness of intersection products in the cohomology ring, where the opposite class \tilde{\sigma}_{\lambda^\vee} is Poincaré dual to \sigma_\lambda.[9][1] A concrete illustration occurs in the Grassmannian \operatorname{Gr}(2,4), which has dimension 4 and whose Schubert classes are indexed by the six partitions fitting in the $2 \times 2 rectangle: (0,0), (1,0), (1,1), (2,0), (2,1), and (2,2).[1] These labels correspond to the Poincaré dual basis for the cohomology ring, with codimensions ranging from 0 to 4.[1]Alternative conventions
In the study of Schubert calculus, variations in the depiction of Young diagrams for partitions arise between the English and French conventions. The English convention, which is the default in most modern treatments, draws the diagram with rows of non-increasing lengths \lambda_1 \geq \lambda_2 \geq \dots \geq \lambda_\ell > 0, aligned left and top-justified, corresponding to matrix coordinates where the partition indices the row lengths.[11] In contrast, the French convention orients the diagram by flipping the English version upside down—placing the longest row at the bottom and rows decreasing upwards—aligning more closely with certain matrix representations in the southwest corner.[11] This orientation preserves the combinatorial content but alters the visual indexing of boxes, which can affect tableau fillings or software implementations; for instance, in some historical or quiver variety contexts, the French form directly arises from the positions of zeros in reduced row-echelon matrices defining Schubert cells.[11][12] Another alternative arises in the matrix formulation of Schubert conditions, where flags and subspaces are described via rank constraints on submatrices rather than dimension intersections with fixed flags. For a partition \lambda indexing a Schubert variety \Omega_\lambda in the Grassmannian \mathrm{Gr}(k,n), this notation specifies \Omega_\lambda = \{ V \in \mathrm{Gr}(k,n) \mid \dim(V \cap F_{k+1-j+\lambda_j}^\circ) = j \ \forall j \}, where F^\circ denotes the opposite flag and the conditions translate to prescribed ranks (e.g., \mathrm{rank}(V \cap F_i) \geq r_{i,j}) in the rank matrix r(\lambda) derived from the permutation or partition. These rank matrices compactly encode the Plücker ideal generators as minors of specific submatrices, providing an algebraic perspective equivalent to the geometric definition but useful for computational verification of ideals.[13] This approach contrasts with the standard dimension-based indexing by emphasizing determinantal conditions, which facilitate minimal equations for the variety.[13] In more general flag varieties G/B, Borel's combinatorial framework indexes Schubert varieties using elements of the Weyl group W rather than partitions, with the variety X_w = \overline{B w B / B} for w \in W.[14] For Grassmannians as partial flag varieties G/P with parabolic subgroup P corresponding to the block stabilizer, this reduces to cosets in the quotient W / W_P, where W = S_n and W_P = S_k \times S_{n-k} for \mathrm{Gr}(k,n), allowing permutations to label the cells via Bruhat order on the quotient.[14] This Weyl group notation unifies the treatment across Lie types and highlights equivariant structures, differing from the partition indexing in Grassmannians by incorporating full permutation data for inclusions and intersections.[14] For computational aspects, Lascoux's conventions in defining Schubert polynomials—originally via divided difference operators on permutations in one-line notation—have influenced software implementations, such as the Macaulay2 package MatrixSchubert, which computes \mathfrak{S}_w for w \in S_n using either divided difference algorithms or transition equations.[15][16] These tools adopt Lascoux-Schützenberger's permutation-based indexing, enabling verification of structure constants or equivariant classes, and extend to double Schubert polynomials \mathfrak{S}_w(x,y) via similar operators, providing a standardized interface for non-standard flag settings.[17][18]Ring structure
Cohomology and Chow rings
The rational cohomology ring H^*( \mathrm{Gr}(k,n); \mathbb{Z} ) of the Grassmannian \mathrm{Gr}(k,n) is a free \mathbb{Z}-module concentrated in even degrees, with a basis given by the Schubert classes \sigma_\lambda, which are the Poincaré duals of the fundamental classes of the Schubert varieties.[19] These classes \sigma_\lambda, indexed by partitions \lambda fitting inside a k \times (n-k) rectangle, form a \mathbb{Z}-basis for the ring, establishing it as the central object where Schubert calculus computations take place.[20] The Chow ring A^*(\mathrm{Gr}(k,n)) is the ring of algebraic cycles on \mathrm{Gr}(k,n) modulo rational equivalence, graded by codimension. For the Grassmannian, a smooth projective variety over \mathbb{C}, the cycle class map induces an isomorphism A^*(\mathrm{Gr}(k,n)) \otimes [\mathbb{Q}](/page/Q) \cong H^{2*}(\mathrm{Gr}(k,n); \mathbb{Q} ), and in fact the integral version A^*(\mathrm{Gr}(k,n)) \cong H^{2*}(\mathrm{Gr}(k,n); \mathbb{Z} ) holds due to the torsion-freeness of the cohomology.[19] This isomorphism identifies the Schubert classes in the Chow ring with their counterparts in cohomology, preserving the ring structure under the intersection product in Chow and the cup product in cohomology.[21] A presentation of the ring is given by H^*(\mathrm{Gr}(k,n); \mathbb{Z} ) \cong \mathbb{Z}[ c_1(T), \dots, c_k(T), c_1(Q), \dots, c_{n-k}(Q) ] / ( c(T) c(Q) = 1 ), where T is the tautological subbundle of rank k, Q is the tautological quotient bundle of rank n-k, and c(\cdot) denotes the total Chern class; this arises from the splitting principle and the Whitney sum formula for the trivial bundle T \oplus Q \cong \mathbb{C}^n.[22] Equivalently, using the Chern classes h_i = c_i(Q) of the quotient bundle as generators (each of degree $2i), the ring admits a presentation \mathbb{Z}[h_1, \dots, h_{n-k}] / (e_1, \dots, e_k), where the e_i are the relations obtained from the higher-degree terms in the formal inverse Chern class c(T) = 1 / c(Q), corresponding to the vanishing of complete homogeneous symmetric polynomials h_{k+j} in the Chern roots of Q for j = 1, \dots, k.[21] Symmetrically, an analogous presentation holds using the Chern classes of the tautological subbundle, reflecting the duality between sub- and quotient bundles via complete flags.[22] The Betti numbers of \mathrm{Gr}(k,n) satisfy \dim H^{2d}( \mathrm{Gr}(k,n); \mathbb{Z} ) equal to the number of integer partitions \lambda = (\lambda_1 \geq \dots \geq \lambda_k \geq 0) with \lambda_1 \leq n-k and |\lambda| = \sum \lambda_i = d, which counts the basis elements \sigma_\lambda in degree $2d.[20] The total rank is the binomial coefficient \binom{n}{k}, matching the number of such partitions overall.[19]Schubert basis and dimension
In the cohomology ring of the Grassmannian \mathrm{Gr}(k,n), the Schubert classes \sigma_\lambda, where \lambda ranges over all partitions with at most k parts each at most n-k, form an integral \mathbb{Z}-basis for H^*(\mathrm{Gr}(k,n); \mathbb{Z}). This basis arises from the fundamental classes of the Schubert varieties \Omega_\lambda, and its elements provide a complete additive description of the cohomology groups.[1] The multiplicative structure of the ring is encoded in the intersections of these varieties, reflecting the geometric intersections in the ambient space.[23] The degree of each Schubert class \sigma_\lambda is determined by the codimension of the corresponding variety: \deg(\sigma_\lambda) = 2|\lambda|, where |\lambda| = \sum_{i=1}^k \lambda_i is the codimension of \Omega_\lambda in \mathrm{Gr}(k,n).[23] This follows directly from the dimension of the Schubert variety, which is \dim(\mathrm{Gr}(k,n)) - |\lambda|, with \dim(\mathrm{Gr}(k,n)) = k(n-k). Consequently, the cohomology is concentrated in even degrees, and the graded structure takes the form H^*(\mathrm{Gr}(k,n); \mathbb{Z}) = \bigoplus_{d=0}^{k(n-k)} H^{2d}(\mathrm{Gr}(k,n); \mathbb{Z}), where the components H^{2d} are free \mathbb{Z}-modules of rank equal to the number of partitions \lambda with |\lambda| = d.[23] This Schubert basis is unique up to sign conventions, which depend on choices of orientations for the Schubert varieties and the ambient Grassmannian.[23] Such choices affect the signs in the basis elements but preserve the ring structure and the integral basis property.[1]Inclusion relations
The inclusions among Schubert varieties in the Grassmannian \mathrm{Gr}(k,n) are determined by the Bruhat order on the set of partitions \lambda with at most k parts each at most n-k. Specifically, the Schubert variety \Omega_\mu is contained in \Omega_\lambda if and only if \mu \geq \lambda in the componentwise partial order on partitions, meaning \mu_i \geq \lambda_i for all i, or equivalently, the Young diagram of \mu contains the Young diagram of \lambda. This relation arises from the Bruhat order on the minimal-length right coset representatives in S_n / (S_k \times S_{n-k}), where the order on cosets corresponds to the containment of the associated Schubert varieties.[9][24] This geometric inclusion induces a graded poset structure on the Schubert basis \{\sigma_\lambda\} of the cohomology ring, graded by the codimension |\lambda| of the varieties, where the order is reversed relative to the Bruhat order: \sigma_\mu \leq \sigma_\lambda if \Omega_\mu \supset \Omega_\lambda. The poset structure facilitates algebraic operations in Schubert calculus, such as pullbacks of Schubert classes under morphisms between Grassmannians that respect the order, and specializations where classes supported on larger varieties degenerate to those on subvarieties.[9] Combinatorially, the inclusion relations mirror the containment order on Young diagrams: \Omega_\mu \subset \Omega_\lambda precisely when every box in the diagram of \lambda is also in the diagram of \mu. This provides a direct bijection between geometric subvariety inclusions and the lattice of partitions under diagram containment, underscoring the combinatorial foundation of the geometry.[5] In the Grassmannian \mathrm{Gr}(2,4), which parametrizes lines in \mathbb{P}^3, an example of such an inclusion chain is \Omega_{(2,2)} \subset \Omega_{(2,1)} \subset \Omega_{(2,0)} \subset \Omega_{(1,0)} \subset \Omega_{(0,0)}. Here, \Omega_{(2,2)} is a point (a fixed line), \Omega_{(2,1)} consists of lines lying in a fixed plane and passing through a fixed point, \Omega_{(2,0)} of lines incident to a fixed line, \Omega_{(1,0)} of lines meeting a fixed plane, and \Omega_{(0,0)} is the entire Grassmannian, reflecting the classical incidence geometry of points, lines, and planes in \mathbb{P}^3 via intersection conditions with a fixed flag.[9]Multiplication formulas
Pieri formula
The Pieri formula provides an explicit rule for multiplying a general Schubert class by a special Schubert class (corresponding to a partition of the form (r, 0, \dots, 0) with $0 \leq r \leq n-k) in the cohomology ring of the Grassmannian \mathrm{Gr}(k, n). Named after the Italian mathematician Mario Pieri (1860–1913), who contributed to enumerative geometry in the early 20th century, the formula states: \sigma_r \cdot \sigma_\lambda = \sum_{\nu} \sigma_\nu, where the sum is over all partitions \nu obtained from \lambda by adding a total of r boxes to its Young diagram such that no two boxes are added to the same column (i.e., \nu / \lambda is a horizontal strip of size r), and \nu fits inside the k \times (n-k) rectangle. All coefficients in the expansion are 1, reflecting the geometric positivity of intersections. This rule, originally developed for classical enumerative problems, simplifies computations and serves as a building block for the general Littlewood-Richardson rule.[22]Giambelli formula
The Giambelli formula expresses a general Schubert class \sigma_\lambda in the cohomology ring of the Grassmannian \mathrm{Gr}(k, n) as a determinant involving special Schubert classes \sigma_m, which are the Chern classes of the tautological quotient bundle. For a partition \lambda = (\lambda_1 \geq \lambda_2 \geq \cdots \geq \lambda_k \geq 0) with \lambda_1 \leq n-k, let r = l(\lambda) be the length of \lambda (the number of nonzero parts, padded with zeros to reach k if necessary). Then, \sigma_\lambda = \det\left( \sigma_{\lambda_i + j - i} \right)_{1 \leq i,j \leq r}, where \sigma_m = 0 if m < 0 and \sigma_0 = 1. This determinantal representation, originally due to Giovanni Zeno Giambelli in 1903, facilitates explicit computations of Schubert classes and their relations in the ring.[25] Geometrically, the formula arises from viewing the Schubert variety \Omega_\lambda as an iterated degeneracy locus in a flag bundle over the Grassmannian, where the class is computed via the Thom-Porteous formula applied to the universal maps between tautological bundles. The special Schubert classes \sigma_m correspond to the Chern classes c_m(\mathcal{Q}) of the quotient bundle \mathcal{Q}, and the determinant captures the expected codimension of the locus defined by rank conditions on these maps. This perspective, modernized by Kempf and Laksov in 1974, unifies the Giambelli formula with degeneracy locus theory and extends it to more general settings.[26] In the Grassmannian \mathrm{Gr}(2, n), for a partition \lambda = (a, b) with a \geq b \geq 0 and a + b \leq n-2, assuming b \geq 1 so l(\lambda) = 2, the formula simplifies to \det\begin{pmatrix} \sigma_a & \sigma_{a+1} \\ \sigma_{b-1} & \sigma_b \end{pmatrix} = \sigma_a \sigma_b - \sigma_{a+1} \sigma_{b-1}. This expression equals the class \sigma_{(a,b)}, and when b = 0, it reduces to \sigma_a since \sigma_{-1} = 0. For instance, in \mathrm{Gr}(2,4), \sigma_{(1,1)} = \sigma_1^2 - \sigma_2. The Giambelli formula has a close analogue in the theory of symmetric functions, where the Schur function s_\lambda is given by s_\lambda = \det(h_{\lambda_i + j - i})_{1 \leq i,j \leq l(\lambda)}, with h_m the complete homogeneous symmetric polynomials. This parallelism reflects the isomorphism between the cohomology ring of \mathrm{Gr}(k, n) (for large n) and the ring of symmetric polynomials, modulo relations, where the special Schubert classes \sigma_m map to the h_m. The Pieri formula may be applied to evaluate the products of special classes appearing in the determinant if needed.General intersection product
The product of two Schubert classes in the cohomology ring (or Chow ring) of the Grassmannian \mathrm{Gr}(k,n) is given by the formula \sigma_\lambda \cdot \sigma_\mu = \sum_\nu c^\nu_{\lambda \mu} \sigma_\nu, where \lambda, \mu, \nu are partitions fitting inside a k \times (n-k) rectangle, and the c^\nu_{\lambda \mu} are the Littlewood-Richardson coefficients.[27] These coefficients are nonnegative integers that count the number of standard Young tableaux of skew shape \nu / \lambda with content \mu. The sum is over all \nu such that the coefficient is positive, and the formula holds in both the cohomology and Chow rings due to the isomorphism between them for Grassmannians. The Schubert basis provides a \mathbb{Z}-basis for these rings, ensuring that the Littlewood-Richardson coefficients are uniquely determined as the structure constants of the ring multiplication, independent of any particular geometric or combinatorial realization. This uniqueness follows from the Poincaré duality and the fact that Schubert classes form a basis indexed by partitions. The positivity of the coefficients arises geometrically from the fact that intersections of Schubert varieties can be deformed into unions of other Schubert varieties with nonnegative multiplicities, as established by classical intersection theory. These coefficients can be computed combinatorially via iterated applications of the Pieri formula or through models such as hives, which label triangular arrays satisfying certain inequalities. The hive model, in particular, provides an efficient way to enumerate the contributing terms by counting integer labelings (hives) with boundary conditions given by \lambda and \mu, yielding \nu at the output. For example, in \mathrm{Gr}(2,4), the product \sigma_{(1,0)} \cdot \sigma_{(1,0)} = \sigma_{(2,0)} + \sigma_{(1,1)}, corresponding to the intersection of two special Schubert divisors degenerating into one codimension-2 class and one codimension-1 class, each with multiplicity one.Connections to other theories
Relation to Chern classes
In the Grassmannian \mathrm{Gr}(k,n), the tautological subbundle S of rank k has fiber over a point [V] equal to the k-plane V \subset \mathbb{C}^n, while the quotient bundle Q = \mathbb{C}^n / S has rank n-k. The total Chern classes satisfy c(S) c(Q) = 1 since S \oplus Q is the trivial bundle of rank n. In terms of the special Schubert classes \sigma_{(i)}, which generate the cohomology ring, these are given by c_i(Q) = \sigma_{(i)} for $1 \leq i \leq n-k, and thus c(S) = 1 - h_1 + h_2 - \cdots + (-1)^k h_k, where h_i = \sigma_{(i)}.[22] The general Schubert classes \sigma_\lambda in H^*(\mathrm{Gr}(k,n)) relate to these Chern classes via the Giambelli formula, which expresses \sigma_\lambda as a determinantal polynomial in the Chern classes of the dual tautological subbundle: \sigma_\lambda = \det\left( c_{\lambda_i - i + j}(S^*) \right)_{1 \leq i,j \leq \ell(\lambda)}, where \ell(\lambda) is the length of the partition \lambda. This determinantal form arises from the geometry of degeneracy loci and aligns with the dual Jacobi-Trudi identity in symmetric function theory. Equivalently, the Schubert classes correspond to Schur polynomials in the generators h_1, \dots, h_k: \sigma_\lambda = s_\lambda(h_1, \dots, h_k).[9] A presentation of the cohomology ring is H^*(\mathrm{Gr}(k,n), \mathbb{Z}) \cong \mathbb{Z}[c_1(S), \dots, c_k(S), c_1(Q), \dots, c_{n-k}(Q)] / \langle c(S) c(Q) - 1 \rangle, where c_i(Q) = h_i. In this ring, the Schubert classes \sigma_\lambda form a \mathbb{Z}-basis, and the relations encode the Whitney sum formula for the trivial ambient bundle. This structure provides an alternative to the combinatorial basis, facilitating computations via symmetric function identities.[22] A key application arises in evaluating integrals of Schubert classes over the Grassmannian, which correspond to degrees of cycles. Specifically, the integral \int_{\mathrm{Gr}(k,n)} \sigma_\lambda = \delta_{\lambda, (n-k)^k}, where (n-k)^k is the rectangular partition representing the class of a point (with integral 1), and 0 otherwise. This follows from the orthogonality of the Schubert basis with respect to the pushforward to the point and can be verified using the degree of the Plücker embedding, where the hyperplane class is \sigma_1 = h_1, and higher powers decompose via the ring structure into multiples of the point class.[9]Generalizations beyond Grassmannians
The full flag variety \mathrm{Fl}(n), which parametrizes complete flags of subspaces in \mathbb{C}^n, is the homogeneous space \mathrm{GL}(n,\mathbb{C})/B, where B is the Borel subgroup of upper triangular matrices.[14] Schubert varieties in \mathrm{Fl}(n) are the closures of B-orbits on this space and are indexed by permutations w \in S_n, partially ordered by the Bruhat order on S_n.[14] These varieties form a basis for the cohomology ring H^*(\mathrm{Fl}(n);\mathbb{Z}), extending the Schubert basis from Grassmannians, which arise as partial flag varieties.[14] A combinatorial model for this cohomology ring is provided by double Schubert polynomials \mathfrak{S}_w(x;y), introduced by Lascoux and Schützenberger, which represent the equivariant Schubert classes under the torus action on \mathrm{Fl}(n).[28] These polynomials are defined recursively using divided difference operators, and their multiplication in the ring corresponds to the intersection product of Schubert classes via these operators, yielding a positive expansion in the Schubert basis.[29][28] Analogues of the Pieri formula for flag varieties compute the product of a Schubert class by a special Schubert class (corresponding to a Grassmannian divisor) as a nonnegative sum over Schubert classes covering the initial permutation in a compatible way, expressed combinatorially via Schubert polynomials.[30] Similarly, a Giambelli-type formula expresses Schubert polynomials for vexillary permutations as determinants of Schur polynomials, while general cases rely on more intricate determinantal structures from Rothe diagrams.[31] The full structure constants of the intersection product in the Schubert basis, however, are governed by Kazhdan-Lusztig polynomials, which encode the coefficients in the change of basis to the Kazhdan-Lusztig basis and ensure positivity properties for intersections.[32][33] Quantum Schubert calculus extends this framework by deforming the cohomology ring with a quantum parameter q, where the structure constants incorporate Gromov-Witten invariants counting rational curves intersecting Schubert varieties. In particular, for homogeneous spaces beyond Grassmannians, such as other partial flag varieties, Buch, Chaput, Manivel, and Perrin developed explicit descriptions of the quantum cohomology ring, showing that three-point Gromov-Witten invariants determine the full structure and recover classical limits. This quantum deformation highlights connections to enumerative geometry, with positivity results mirroring classical cases.Concrete examples
Chow ring of Gr(2,4)
The Grassmannian \mathrm{Gr}(2,4) parametrizes 2-dimensional subspaces of \mathbb{C}^4 and has complex dimension 4. It embeds into \mathbb{P}^5 via the Plücker embedding, where it realizes as the smooth Klein quadric hypersurface defined by the equation p_{12}p_{34} - p_{13}p_{24} + p_{14}p_{23} = 0 in Plücker coordinates [p_{ij}]. The Chow ring A^*(\mathrm{Gr}(2,4)) is generated by the Schubert classes and has rank 6 as a free \mathbb{Z}-module, reflecting the even-degree Betti numbers $1,1,2,1,1$ of its cohomology ring (shifted by dimension). A standard presentation of the Chow ring is A^*(\mathrm{Gr}(2,4)) \cong \mathbb{Z}[\sigma_1, \sigma_2]/I, where \sigma_1 = \sigma_{(1,0)} and \sigma_2 = \sigma_{(1,1)} are the codimension-1 and codimension-2 special Schubert classes (corresponding to the first Chern class c_1(\mathcal{Q}) and second Chern class c_2(\mathcal{Q}) of the tautological quotient bundle \mathcal{Q} of rank 2), and I is the ideal generated by the degree-6 relation \sigma_1^3 - 2\sigma_1 \sigma_2 = 0 and the degree-8 relation \sigma_1^4 - 3\sigma_1^2 \sigma_2 + 2\sigma_2^2 = 0. These relations arise from the vanishing of the higher elementary symmetric functions e_3 = e_4 = 0 in the Chern roots of \mathcal{Q} \oplus \mathcal{O}^{\oplus 2} on the trivial rank-4 bundle. The \mathbb{Z}-basis for A^*(\mathrm{Gr}(2,4)) consists of the Schubert classes \{1, \sigma_{(1,0)}, \sigma_{(2,0)}, \sigma_{(1,1)}, \sigma_{(2,1)}, \sigma_{(2,2)}\}, spanning codimensions $0,1,2,2,3,4respectively. Using the [Pieri formula](/page/Pieri_formula), the relation\sigma_1^2 = \sigma_{(2,0)} + \sigma_{(1,1)}expresses the nonspecial class\sigma_{(2,0)}in terms of the generators, while the [Giambelli formula](/page/Giambelli_formula) gives\sigma_{(2,1)} = \sigma_1 \sigma_{(1,1)} - \sigma_{(2,0)}and\sigma_{(2,2)} = \sigma_{(1,1)}^2. Substituting the degree-6 relation yields \sigma_1^3 = 2 \sigma_{(2,1)}and\sigma_{(1,1)}^2 = \sigma_{(2,2)}$, with the degree-8 relation ensuring consistency in the top degree. The full ring structure is determined by these relations and the Pieri rule for products with \sigma_1, yielding the multiplication table in the Schubert basis:| \cdot | $1$ | \sigma_{(1,0)} | \sigma_{(2,0)} | \sigma_{(1,1)} | \sigma_{(2,1)} | \sigma_{(2,2)} |
|---|---|---|---|---|---|---|
| $1$ | $1$ | \sigma_{(1,0)} | \sigma_{(2,0)} | \sigma_{(1,1)} | \sigma_{(2,1)} | \sigma_{(2,2)} |
| \sigma_{(1,0)} | \sigma_{(1,0)} | \sigma_{(2,0)} + \sigma_{(1,1)} | \sigma_{(2,1)} | \sigma_{(2,1)} | \sigma_{(2,2)} | $0$ |
| \sigma_{(2,0)} | \sigma_{(2,0)} | \sigma_{(2,1)} | \sigma_{(2,2)} | \sigma_{(2,2)} | $0$ | $0$ |
| \sigma_{(1,1)} | \sigma_{(1,1)} | \sigma_{(2,1)} | \sigma_{(2,2)} | \sigma_{(2,2)} | $0$ | $0$ |
| \sigma_{(2,1)} | \sigma_{(2,1)} | \sigma_{(2,2)} | $0$ | $0$ | $0$ | $0$ |
| \sigma_{(2,2)} | \sigma_{(2,2)} | $0$ | $0$ | $0$ | $0$ | $0$ |