--- title: Almost-Uniform Steiner Bundles url: https://www.emergentmind.com/topics/almost-uniform-steiner-bundles type: topic --- # Almost-Uniform Steiner Bundles Almost-uniform Steiner bundles are Steiner bundles whose restriction data are generically constant but admit a controlled jumping locus. On projective space, the basic model is a bundle \(V\) on \(\mathbb{P}^n\) presented by a general linear resolution \[ 0 \to \mathcal{O}_{\mathbb{P}^n}(-1)^t \to \mathcal{O}_{\mathbb{P}^n}^{t+r} \to V \to 0, \] and the almost-uniform phenomenon appears in the splitting of \(V\) on lines: for general lines the splitting is balanced and constant, while deviations are confined to a determinantal locus of special lines or special matrices [2204.10247]. The term is not completely uniform across the literature: in the projective-space analysis of general Steiner bundles it is an inferred property rather than a formal definition, whereas in later work on \(\mathbb{P}^n\), following Ellia, almost-uniformity means that the set of jumping lines is finite and non-empty [2509.02014]. ## 1. Steiner resolutions and numerical invariants A Steiner bundle on \(\mathbb{P}^n\) in the sense of Dolgachev–Kapranov is a vector bundle \(V\) admitting an exact sequence \[ 0 \to \mathcal{O}_{\mathbb{P}^n}(-1)^t \to \mathcal{O}_{\mathbb{P}^n}^{t+r} \to V \to 0, \] with \(t>0\), \(r\ge n\), and general map. Here “general” means a member of a dense open subset of the affine parameter space \(\operatorname{Hom}(\mathcal{O}_{\mathbb{P}^n}(-1)^t,\mathcal{O}_{\mathbb{P}^n}^{t+r})\), so that the sequence is exact and \(V\) is locally free when \(r\ge n\). The dual presentation \[ 0 \to V^* \to \mathcal{O}_{\mathbb{P}^n}^{t+r} \to \mathcal{O}_{\mathbb{P}^n}(1)^t \to 0 \] is equally fundamental, because the maps on global sections govern the cohomological behavior of twists \(V(a)\) [2204.10247]. The basic invariants are determined directly by the resolution. Writing \(H\) for the hyperplane class, one has \[ c(V)=\frac{1}{(1-H)^t}. \] Hence \[ \operatorname{rk}(V)=r,\qquad c_1(V)=tH,\qquad \mu(V)=\frac{t}{r}, \] and, for \(k\ge 2\), \[ c_k(V)=\binom{t+k-1}{k}H^k, \] truncated at \(k\le n\). These formulas are the numerical substrate for the later cohomology, stability, ampleness, and restriction statements. They also show that, within a fixed pair \((t,r)\), the parameter space of general Steiner bundles is governed by the linear data of the defining matrix. A first structural point is that almost-uniformity is not an extra decoration imposed on an arbitrary bundle class. For Steiner bundles it is encoded already at the level of the two-step linear resolution: the same ratio \(t/r\) that fixes slope and border cohomological twists also fixes the generic splitting profile on lines. This suggests that almost-uniformity is an intrinsic asymptotic consequence of the Steiner presentation rather than an accidental feature of isolated examples. ## 2. Restriction to lines and balanced splitting Restricting the defining sequence to a line \(L\simeq \mathbb{P}^1\) gives \[ 0 \to \mathcal{O}_L(-1)^t \to \mathcal{O}_L^{r+t} \to V|_L \to 0. \] Therefore \(\deg(V|_L)=t\) and \(\operatorname{rk}(V|_L)=r\). For a general matrix and a general line, the restricted map is general, so \(V|_L\) is a general quotient of \(\mathcal{O}_L^{r+t}\) by \(\mathcal{O}_L(-1)^t\). Since every vector bundle on \(\mathbb{P}^1\) splits, the splitting on a general line is balanced and uniquely determined by \(t\) and \(r\): \[ V|_L \cong \mathcal{O}_L(\alpha)^u \oplus \mathcal{O}_L(\alpha+1)^{r-u}, \qquad \alpha=\Big\lfloor \frac{t}{r}\Big\rfloor,\quad u=r(\alpha+1)-t. \] This balanced splitting is constant on a dense open subset of the Grassmannian of lines, so the restriction type is uniform on general lines [2204.10247]. Within this projective-space framework, the paper does not formally define “almost-uniform,” but its restriction analysis yields a precise almost-uniform property: the splitting type is constant and balanced on a dense open set of lines, and the jumping locus where the splitting deviates is a proper closed subset. The explicitly analyzed jump is the appearance of a trivial summand \(\mathcal{O}_L\), equivalently \(H^0(V^*|_L)\ne 0\). For a fixed line this condition cuts codimension at least \(t-r+1\) in the fiber \(\operatorname{Hom}(\mathcal{O}_L^{r+t},\mathcal{O}_L(1)^t)\), while globally the locus of maps admitting some such line has codimension at least \(t-r-2n+3\) in the parameter space [2204.10247]. A different, stricter convention appears in later work on \(\mathbb{P}^n\): for a bundle \(F\), there is a dense open subset \(O_F\) of the Grassmannian of lines on which the splitting multiplicities are constant, and the complementary set \(J_F\) is the set of jumping lines; following Ellia, \(F\) is almost-uniform if \(J_F\) is finite and non-empty. In this language, uniformity means \(J_F=\varnothing\), while almost-uniformity means that the failure of uniformity is concentrated on finitely many lines [2509.02014]. A recurrent misconception is to identify almost-uniformity with homogeneity. The later representation-theoretic results show that these notions are distinct: uniform Steiner bundles can be non-homogeneous, and almost-uniformity concerns the geometry of the jumping locus rather than invariance under the action of \(\mathrm{GL}(A_r)\). ## 3. Cohomology, stability, and ampleness The cohomological behavior of general Steiner bundles is organized by a natural cohomology conjecture: every twist \(V(a)\) should have at most one nonvanishing cohomology group. For the dual kernel presentation \[ 0 \to V \to \mathcal{O}_{\mathbb{P}^n}^{t+r} \to \mathcal{O}_{\mathbb{P}^n}(1)^t \to 0, \] set \[ \alpha=\Big\lceil \frac{tn}{r}\Big\rceil,\qquad \beta=\Big\lfloor \frac{tn}{r}\Big\rfloor. \] Then \(V\) has natural cohomology if and only if \[ H^0(V(\beta-1))=0,\qquad H^1(V(\alpha-1))=0. \] This reduces all twists to two border cases via Castelnuovo–Mumford regularity and semicontinuity. The conjecture is proved on \(\mathbb{P}^2\), and asymptotically it holds under scaling: for \[ 0 \to V \to \mathcal{O}_{\mathbb{P}^n}^{m(t+r)} \to \mathcal{O}_{\mathbb{P}^n}(1)^{mt} \to 0, \] if \[ m \ge \max \left\{ \frac{1}{4(t+r)}\binom{\alpha+n-1}{n}^2\binom{\alpha+n}{n},\ \frac{1}{4t}\binom{\beta+n-1}{n}^3 \right\}, \] then \(V\) has natural cohomology. There is also an exact divisibility result: if \(\operatorname{rk}(V)=kn\), then the general Steiner bundle on \(\mathbb{P}^n\) has natural cohomology [2204.10247]. Stability is governed by a concrete slope window. If \[ \psi_n=\frac{n-1+\sqrt{n^2+2n-3}}{2}, \] then a general Steiner bundle \[ 0 \to \mathcal{O}_{\mathbb{P}^n}(-1)^t \to \mathcal{O}_{\mathbb{P}^n}^{r+t} \to V \to 0 \] is slope stable whenever \[ \frac{n}{t}\le \frac{r}{t}<\psi_n. \] For degree-\(d\) maps \[ 0 \to \mathcal{O}_{\mathbb{P}^n}(-d)^t \to \mathcal{O}_{\mathbb{P}^n}^{r+t} \to V \to 0, \] the same numerical range yields slope semistability for general \(V\). In \(\mathbb{P}^3\), the low-slope regime is handled by restriction to a quadric and the use of Abe’s and Rudakov’s classification [2204.10247]. Ampleness connects directly with the line-splitting picture. There is a closed subset \(S\) in the map space of codimension at least \(t-r-2n+3\) such that if the defining matrix lies outside \(S\), then \(V\) is ample. In particular, if \[ t-r>2n-3, \] the general Steiner bundle is ample. Since ampleness on lines is equivalent to positivity of all summands of \(V|_L\), this criterion turns the almost-uniform statement into a global positivity statement: in the ample range, every line carries only positive summands; below that range, failure of ampleness is still confined to a determinantal bad locus of controlled codimension [2204.10247]. The cohomology, stability, and restriction numerics are aligned by the same ratio \(t/r\). The border twists \(\alpha,\beta\), the balanced degrees on general lines, and the slope window are all controlled by this single parameter. A plausible implication is that almost-uniformity for general Steiner bundles is not merely geometric but also cohomological: the same numerics constrain both line restrictions and the placement of nonzero cohomology. ## 4. Kronecker-quiver formulation A decisive reorganization of Steiner bundles on \(\mathbb{P}^n\) comes from the generalized Kronecker quiver \(K_r\). In the notation of the representation-theoretic literature, one sets \(r=n+1\) and identifies \(\mathbb{P}(A_r)\simeq \mathbb{P}^n\). A bundle \(F\) is Steiner if there are finite-dimensional \(k\)-vector spaces \(V_1,V_2\) and an exact sequence \[ 0 \to V_1\otimes_k \mathcal{O}_{\mathbb{P}(A_r)}(-1) \xrightarrow{\phi} V_2\otimes_k \mathcal{O}_{\mathbb{P}(A_r)} \to F \to 0. \] There is an equivalence \[ \widetilde{\Theta}\colon \operatorname{rep}_{\proj}(K_r,1)\xrightarrow{\sim}\operatorname{StVect}(\mathbb{P}(A_r)), \] from relative \(1\)-projective Kronecker representations to Steiner bundles. If the corresponding representation \(M\) has dimension vector \((a,b)\), then \[ \operatorname{rk}(\widetilde{\Theta}(M))=b-a,\qquad c_1(\widetilde{\Theta}(M))=a. \] Under this equivalence, the restriction of \(F=\widetilde{\Theta}(M)\) to a line is read off from the decomposition of the restricted representation \(\alpha^*(M)\) into preprojective indecomposables \(P_i(2)\), and the multiplicities of \(P_i(2)\) are exactly the splitting multiplicities of \(\mathcal{O}_{\mathbb{P}^1}(i)\) on that line [2509.02014]. This dictionary introduces the support \[ \operatorname{supp}(F)=\{\,i\in \mathbb{N}_0\mid b_i(F)\ne 0\,\} \] and the \(k\)-type \[ k=\max \operatorname{supp}(F), \] which records the highest twisting degree appearing in the splitting type on lines. It also separates homogeneity from uniformity. A bundle is homogeneous if \(g^*F\simeq F\) for every \(g\in \mathrm{GL}(A_r)\), and homogeneous bundles are uniform; however, the converse fails. The mechanism controlling splitting types is categorical. Restriction, inflation, and reflection functors produce adjoint pairs \[ \sigma_{K_r}^{-1}\circ \mathrm{inf}\ \dashv\ \sigma_{K_d}\circ \mathrm{res}, \qquad \sigma_{K_d}^{-1}\circ \mathrm{res}\ \dashv\ \sigma_{K_r}\circ \mathrm{inf}, \] and from preprojective \(K_2\)-modules one obtains test representations \(P_n^\pm(\ell)\) attached to lines \(\ell\subset \mathbb{P}(A_r)\). Their Hom-vanishing detects whether the restricted representation is supported in degrees \(n\) and \(n+1\): \[ \alpha^*(M)\in \operatorname{add}\{P_n(2),P_{n+1}(2)\} \Longleftrightarrow \operatorname{Hom}_{K_r}(P_{n+1}^-(\ell),M)=0= \operatorname{Hom}_{K_r}(M,P_n^+(\ell)). \] Thus the line-splitting problem becomes a problem of orthogonality to test modules [2509.02014]. ## 5. Existence results and prescribed jumping behavior The representation-theoretic approach produces explicit existence theorems beyond the classical homogeneous examples. For any \(n\ge 2\) and \(k\ge 1\), there exists a simple uniform Steiner bundle on \(\mathbb{P}^n\) of \(k\)-type with support \(\{k-1,k\}\) that is not homogeneous. There also exists an indecomposable uniform Steiner bundle on \(\mathbb{P}^n\) of \(k\)-type with support \(\{0,1,k-1,k\}\) that is not homogeneous. Since \(\{0,1,k-1,k\}\) is disconnected exactly for \(k\ge 4\), this yields uniform but non-homogeneous Steiner bundles with disconnected splitting type for every \(k\ge 4\) [2509.02014]. A more quantitative existence statement fixes rank and first Chern class. Given \(k\ge 2\), \(s\ge 2k^2(n-1)\), and \(\ell:=s-2k^2(n-1)\ge 0\), for each \[ (k-1)\big(k(n-1)+s\big)\le c\le (k-1)\big(k(n-1)+s\big)+\ell, \] the general Steiner bundle on \(\mathbb{P}^n\) of rank \(s\) and first Chern class \(c\) is simple, uniform of \(k\)-type with support \(\{k-1,k\}\), and is not homogeneous. These results show that uniformity does not force homogeneous symmetry, even inside the rigid class of Steiner bundles. Almost-uniform bundles in the finite-jumping-line sense are also constructed systematically. For any finite non-empty set of lines \(L\subset G(\mathbb{P}^n)\), there is a full subcategory of Steiner bundles consisting of almost-uniform bundles whose set of jumping lines is exactly \(L\), and this subcategory corresponds via \(\widetilde{\Theta}\) to a wild subcategory of \(\operatorname{rep}(K_{n+1})\). A concrete criterion is \[ J_{\widetilde{\Theta}(M)} = \{\,u\in \operatorname{Gr}_2(A_r)\mid \operatorname{Hom}_{K_r}(P_1^-(u),M)\ne 0\,\}. \] If this set is finite and non-empty, then \(\widetilde{\Theta}(M)\) is almost-uniform, and outside the jumping lines the generic splitting type is \[ \widetilde{\Theta}(M)|_\ell \cong \mathcal{O}_\ell^{\oplus \Delta_M(2)} \oplus \mathcal{O}_\ell(1)^{\oplus \dim_k M_1}. \] In particular, for any line \(\ell\subset \mathbb{P}^n\), the bundle \[ E_\ell:=\widetilde{\Theta}(P_1^-(\ell)) \] has rank \(2n-1\) and \(J_{E_\ell}=\{\ell\}\) [2509.02014]. The wildness statement is conceptually important. It shows that once the jumping set is allowed to be finite and prescribed, almost-uniform Steiner bundles form a class with rich internal complexity rather than a near-classifiable boundary case between uniform and non-uniform behavior. ## 6. Grassmannians, jumping pairs, and broader generalizations On Grassmannians, Steiner bundles are defined by the universal subbundle rather than by \(\mathcal{O}(-1)\). If \(G(k,n)=G(k+1,V^*)\), an \((s,t)\)-Steiner bundle \(F\) is given by \[ 0 \to S\otimes U \to T\otimes \mathcal{O}_{G(k,n)} \to F \to 0, \] with \[ \operatorname{rk}(F)=t-s(k+1). \] There is a lower bound \[ \operatorname{rk}(F)\ge \min\big((k+1)(n-k),(n-k)\dim S\big). \] The central replacement for line-splitting is the geometry of jumping pairs. If \(\varphi:T^*\to S^*\otimes H^0(U^\vee)\) is the linear map encoding the Steiner resolution, then a pair \((a,\Gamma)\in G(1,S^*)\times G(k+1,H^0(U^\vee))\) is a jumping pair when \(a\otimes \Gamma\subset \operatorname{Im}\varphi\), and the jumping-pair variety satisfies \[ \widetilde{J}(F)=\operatorname{Im}\nu\cap G(k+1,T_0^*), \] where \(\nu\) is the generalized Segre embedding and \(T_0^*=\operatorname{Im}\varphi\). The expected dimension has a lower bound, tangent-space calculations give an upper bound, and Steiner bundles whose jumping locus has maximal dimension are exactly Schwarzenberger bundles [1208.0571]. A parallel theory exists on arbitrary smooth projective varieties. For a strongly exceptional pair \((F_0,F_1)\), an \((F_0,F_1)\)-Steiner bundle has resolution \[ 0 \to S\otimes F_0 \to T\otimes F_1 \to E \to 0. \] In the special case \((F_0,\mathcal{O}_X)\), the paper introduces jumping pairs \[ \widetilde{J}(E)\subset G(1,S^*)\times G(f_0,H^0(F_0^\vee)), \] proves an expected lower bound and a tangent-space upper bound for its dimension, and classifies the extremal maximal-dimension case as Schwarzenberger. The paper does not explicitly define almost-uniformity, but its jumping-pair formalism makes “small jumping locus” the natural measure of departure from uniformity: empty jumping locus corresponds to uniformity, while a small proper jumping locus corresponds to generic constancy with controlled exceptional behavior [1306.0746]. The representation-theoretic Grassmannian formulation sharpens this viewpoint further. For \(\mathrm{Gra}(k^r)\), there is an equivalence \[ \mathcal{O}_d:\operatorname{rep}_{\proj}(K_r,d)\xrightarrow{\sim}\operatorname{StVect}(\mathrm{Gra}(k^r)), \] and for \(M\in \operatorname{rep}(K_r)\) the closed \(d\)-th rank variety \[ V(K_r,d)_M=\{\,b\in \operatorname{Gr}_d(A_r)\mid \operatorname{rk}(\mathcal{U}_{M,b})