--- title: Geometric Degeneracy Variety url: https://www.emergentmind.com/topics/geometric-degeneracy-variety type: topic --- # Geometric Degeneracy Variety In current research usage, “geometric degeneracy variety” does not denote a single standardized object. The phrase is applied to loci defined by the failure of a generic geometric condition: orbit closure degeneration for invariant subspaces of nilpotent operators, hyperplane dependence on Segre varieties, drop of dual-variety dimension, enlargement of eigenspaces in matrix space, vanishing projected exchange fields in band theory, weakly special collapse in universal abelian schemes, or isotropy for a degenerate multilinear form. Across these settings, the common feature is that an ambient parameter or configuration space is stratified by a degeneracy condition expressible through rank, tangency, isotropy, closure, or vanishing equations [1202.2813] [2606.31965] [2507.17485] [2303.04936]. ## 1. Invariant-subspace varieties and orbit-closure degeneracy A particularly explicit algebraic realization occurs for invariant subspaces of nilpotent operators. For a partition \(\alpha=(\alpha_1\ge \alpha_2\ge \dots \ge \alpha_t)\), the associated nilpotent \(k[T]\)-module is \[ N_\alpha=\bigoplus_{i=1}^t k[T]/(T^{\alpha_i}), \] and an embedding of invariant subspaces is a monomorphism \(f:N_\alpha\hookrightarrow N_\beta\). Fixing \(\alpha,\beta,\gamma\), the locus \[ V_{\alpha,\gamma}^\beta(k)\subset V_{\alpha,\beta}(k) \] consists of monomorphisms with \(\mathrm{Coker}\,f\cong N_\gamma\); this is the main geometric degeneracy variety in that setting. It sits inside the affine space \(H_{\alpha,\beta}(k)=\mathrm{Hom}_k(N_\alpha,N_\beta)\cong k^{|\alpha||\beta|}\), and becomes a degeneration problem once \(k\) is assumed algebraically closed [1202.2813]. The relevant group is \[ G=\Aut_N(N_\alpha(k))\times \Aut_N(N_\beta(k)), \qquad (g,h)\cdot f=hfg^{-1}. \] Its orbits are precisely isomorphism classes of short exact sequences \[ 0\to N_\alpha\xrightarrow{f}N_\beta\to N_\gamma\to 0. \] For \(Y=(N_\alpha,N_\beta,f)\) and \(Z=(N_\alpha,N_\beta,g)\), geometric degeneration is defined by \[ Y\le_{\deg} Z \Longleftrightarrow \mathcal O_g\subseteq \overline{\mathcal O_f}. \] Inside the subcategory \(S_2(k)\), where \(\alpha_1\le 2\), this order is combinatorially controlled by Klein tableaux and arc diagrams. If \(\mathcal A\) is the arc diagram of \(Y\), with crossing number \(x(\mathcal A)\), then \[ \dim Gf=\deg h^\beta_{\alpha,\gamma}+\deg a_\beta-x(\mathcal A). \] Using \[ \deg h^\beta_{\alpha,\gamma}=n(\beta)-n(\alpha)-n(\gamma),\qquad \deg a_\alpha=|\alpha|+2n(\alpha), \] the orbit dimension is reduced by exactly one for each arc crossing. The degeneration order, the Ext-order, the Hom-order, and the arc order coincide on fixed partition type in \(S_2(k)\), and over an algebraically closed field they are also equivalent to the orbit-closure order \(\le_{\deg}\). Within each Littlewood–Richardson stratum there is a unique dominant Klein tableau with no crossings and maximal orbit dimension, and a unique maximal-crossing tableau with minimal orbit dimension [1202.2813]. ## 2. Determinantal and dual-type degeneracy loci A second major usage concerns loci cut out by hyperplane dependence, Hessian rank bounds, or degeneration of Gauss images. For Segre varieties, if \[ \sigma_{m_\bullet}:\mathbb P^{m_1}\times\cdots\times \mathbb P^{m_r}\to \mathbb P^{M(m_\bullet)} \] is the Segre embedding and \(p_\bullet=(P_1,\dots,P_n)\), the Segre-determinantal locus \(\Seg_{m_\bullet,n}\) is the smallest closed subvariety containing all \(n\)-tuples whose Segre images lie on a common hyperplane. Writing \(\mathcal N_{m_\bullet,n}(p_\bullet)\) for the matrix whose rows are the Segre coordinates of the \(P_k\), one has \[ \Seg_{m_\bullet,n}=V(I_{m_\bullet,n}), \] where \(I_{m_\bullet,n}\) is generated by the maximal minors of \(\mathcal N_{m_\bullet,n}\). This ideal is prime, has height \(n-M(m_\bullet)\), the quotient ring is Cohen–Macaulay, and the maximal minors form a universal Gröbner basis. In the flatland case \(m_\bullet=(1,1,1)\), the image variety for three flatland cameras is exactly \(\Seg_{(1,1,1),n}\) [2606.31965]. For hypersurfaces, another degeneracy variety is \[ Dual_{k,d,N}\subset \mathbb P(S^d\mathbb C^N), \] the Zariski closure of degree-\(d\) hypersurfaces whose dual variety has dimension at most \(k\). If \(P\in S^dW^*\) is irreducible and \(w\) is a general point of the affine cone over \(Z(P)\), Katz’s formula gives \[ \dim Z(P)^*=\operatorname{rank}(H_{P,w})-2. \] Hence \(\dim Z(P)^*\le k\) is equivalent to rank bounds on Hessian restrictions, or equivalently to divisibility conditions \(P\mid \det(H_P|_F)\) for all \((k+3)\)-dimensional subspaces \(F\subset W\). The resulting set-theoretic equations form an \(SL_N\)-module of degree \((k+2)(d-1)\). The determinant orbit closure \(\overline{GL_{n^2}\cdot[\det_n]}\) is an irreducible component of \(Dual_{2n-2,n,n^2}\), and this yields the lower bound \[ \overline{dc}(\mathrm{perm}_m)\ge \frac{m^2}{2}. \] The same dual-geometric viewpoint controls degeneration in families: for \[ X^s=\{sF_d+F_{d_1}F_{d_2}=0\}\subset \mathbb P^{n+1}, \] the flat limit of the duals \((X^s)^*\) is reducible, with components \(X_{d_1}^*\) and \(X_{d_2}^*\) of multiplicity \(1\), \((X_{d_1}\cap X_{d_2})^*\) of multiplicity \(2\), and \((X_d\cap X_{d_1}\cap X_{d_2})^*\) of multiplicity \(1\) [1004.4802] [2312.17728]. ## 3. Tensor and matrix degeneracy varieties For tridimensional tensors, degeneracy is formulated through a kernel incidence condition. If \[ A=(a_{ijk})\in \mathbb C^p\otimes \mathbb C^q\otimes \mathbb C^r, \] the kernel \(K_A\subset \mathbb P^{p-1}\times \mathbb P^{q-1}\times \mathbb P^{r-1}\) consists of triples \((P,Q,T)\) satisfying the three systems \[ \sum_{i,j}a_{ijk}x_i y_j=0,\quad \sum_{i,k}a_{ijk}x_i z_k=0,\quad \sum_{j,k}a_{ijk}y_j z_k=0. \] The tensor is degenerate iff \(K_A\neq\emptyset\). Associated to \(A\) are matrices \(L,M,N\) of linear forms and determinantal schemes \(\mathcal L,\mathcal M,\mathcal N\). If \(A\) is degenerate and \((P,Q,T)\in K_A\), then \(P,Q,T\) are degenerate points of \(\mathcal L,\mathcal M,\mathcal N\), respectively. Conversely, a degenerate but non bi-degenerate point of one of these schemes implies degeneracy of \(A\). When \(r\le p+q-1\), the hyperdeterminant exists and \[ A\text{ is degenerate}\iff Det(A)=0. \] When \(r\ge p+q-1\), one has \[ A\text{ degenerate}\iff \mathcal L\text{ degenerate}\iff \mathcal M\text{ degenerate}. \] The same framework also relates degeneracy to conciseness, essential format, and tensor rank in small formats such as \((2,2,2)\), \((2,2,3)\), and \((2,2,r)\) [2605.10866]. For matrices, the geometric degeneracy variety is \[ \Sigma^{(n)}=\Sigma\subset M_n(\mathbb C), \] defined by the existence of an eigenvalue of geometric multiplicity at least \(2\). Equivalently, \[ \Sigma=\{A\in M_n(\mathbb C)\mid \exists \lambda\in \mathbb C,\ \operatorname{rank}(A-\lambda I)\le n-2\}. \] Its determinantal lift is \[ \widetilde\Sigma=\{(A,\lambda)\mid A-\lambda I\in \Sigma'\}, \qquad \Sigma'=\{A\in M_n(\mathbb C)\mid \operatorname{rank}(A)\le n-2\}. \] Here \(\Sigma'\) is determinantal, irreducible, and Cohen–Macaulay, while \(\Sigma\) itself is not Cohen–Macaulay for \(n\ge 3\). At a strictly \(k\)-fold eigenvalue \(\lambda_0\), the multiplicity of the corresponding local branch satisfies \[ \operatorname{mult}(\Sigma^{(n)},A_0;\lambda_0)=\frac{k^2(k^2-1)}{12}. \] For a holomorphic map germ \(f:(\mathbb C^3,0)\to (M_n(\mathbb C),A_0)\), isolated with respect to that branch, the number of complex Weyl points created by a generic perturbation is computed from the pullback of the \((n-1)\times (n-1)\) minors of \(f(x)-(\lambda+\lambda_0)I\). In the linear case this gives the upper bound \[ \sharp \mathbf{WP}\le \sharp \mathbf{cWP}=\frac{k^2(k^2-1)}{12}, \] for the Weyl points born from a strictly \(k\)-fold degeneracy [2507.17485]. ## 4. Degeneracy as directional, tangency, and vanishing-field geometry Some recent usages are geometric in a more structural sense. In projective geometric algebra, Euclidean PGA is built from a degenerate quadratic space \((V,B)\) with \[ \rad V=\mathbb F e_0,\qquad B(e_0,v)=0\ \text{for all }v\in V. \] The Clifford algebra decomposes as \[ \Cl(V)\cong \Cl(V/\mathbb F e_0)\ltimes_\alpha \Cl(V/\mathbb F e_0), \] where \(\alpha\) is the grade-involution. The radical line \(\mathbb F e_0\), the quotient \(V/\mathbb F e_0\) of parallel classes, and the square-zero ideal \(\Cl(V)e_0\) together supply a “degeneracy variety” of directions and ideal elements. The quotient \(V/\mathbb F e_0\) represents parallel classes of hyperplanes, and the Playfair projection \(\pi_W\) associated with a complement \(V=W\oplus \mathbb F e_0\) algebraizes the existence and uniqueness of parallels [2408.13441]. In spin-orbit-free compensated magnets, geometric spin degeneracy is encoded by the projected effective Zeeman field \[ H_{i,\mathrm{eff}}(\mathbf k)=\varepsilon_i(\mathbf k)\sigma_0+\mathbf f_i(\mathbf k)\cdot \boldsymbol{\sigma}, \qquad f_i^\alpha(\mathbf k)=\sum_{n=1}^N a_n^\alpha |u_{in}(\mathbf k)|^2. \] The degeneracy condition is \[ \mathbf f_i(\mathbf k)=0. \] Writing \(V_i(\mathbf k)=(|u_{i1}|^2,\dots,|u_{iN}|^2)\) in the probability simplex, each equation \[ \sum_{n=1}^N a_n^\alpha V_{in}=0 \] defines a “zero effective Zeeman field” hyperplane. Zero net magnetization, \[ \sum_n a_n^\alpha=0, \] forces these hyperplanes to pass through the simplex center. The resulting degeneracy locus \[ \mathcal D_i=\{\mathbf k\in \mathrm{BZ}:\mathbf f_i(\mathbf k)=0\} \] can be a nodal surface, nodal line, or isolated nodal point, depending on dimension and the number of independent spin components [2604.13266]. A closely related tangency formulation appears in the Elekes–Szabó setting. For a smooth hypersurface \(V=Z(F)\subset \mathbb R^d\), local degeneracy at a coordinate-regular point is equivalent to the factorization \[ F_{x_i}=G(x_1,\dots,x_d)\,\alpha_{2,i}(x_i)\,\prod_{j\neq i}\alpha_{1,j}(x_j) \quad\text{on }V. \] The associated boundary varieties \[ \Sigma_i(V)=\{p\in V_{\mathrm{sm}}:F_{x_i}(p)=0\} \] record tangency to the coordinate hyperplane \(\{x_i=\mathrm{const}\}\). Under the factorization hypothesis, every local irreducible component of \(\Sigma_i(V)\) is contained in a coordinate slice \(V_j(c)=V\cap\{x_j=c\}\). For one-parameter families of hyperspheres, this forces a rigid classification: in dimensions \(d\ge 3\), degeneracy implies a concentric family; in dimension \(2\), the family is either concentric or consists of fixed-radius circles whose centers lie on a line parallel to a coordinate axis [2607.03366]. ## 5. Degeneracy loci in families, moduli, and cosmology In the universal family of principally polarized abelian varieties, \[ \pi:\mathfrak A_g\to \mathbb A_g, \] degeneracy is measured relative to weakly special or bi-algebraic geometry. For an irreducible closed \(X\subset \mathfrak A_g\), the quantity \[ \delta(X)=\dim X^{\mathrm{biZar}}-\dim \pi(X^{\mathrm{biZar}}) \] measures the vertical dimension of the minimal bi-algebraic closure. The \(t\)-th degeneracy locus is \[ X^{\deg(t)} = \bigcup_{\substack{Y\subset X\\ \dim Y>0\\ \delta(Y)<\dim Y+t}} Y. \] Although defined as a union, \(X^{\deg(t)}\) is Zariski closed. If \(X^{\deg(t)}\) contains a non-empty open subset of \(X^{\mathrm{an}}\) and the generic fiber is not contained in a proper algebraic subgroup, then there exists an endomorphism \(\varphi\) of the abelian scheme over the regular locus of \(\pi(X)\) such that \(\dim\ker \varphi_s=\delta(Y)\) on a large family of subvarieties \(Y\subset X\), the fibers \(Y_s^{\mathrm{biZar}}\) are finite unions of translates of \((\ker\varphi_s)^0\), and the quotient abelian varieties \(\varphi(\mathfrak A_g)_s\) are pairwise isomorphic for general \(s\). This degeneracy formalism is used in uniform Mordell–Lang arguments and is linked to a conjectural route toward relative Manin–Mumford through the density of \(X^{\deg(1)}\) [2303.04936]. In cosmology, the term is used differently. For the Hybrid \(f(Q)\) model \[ f(Q)=\alpha_1 Q+\alpha_2 Q_0+\alpha_3 \frac{Q_0^2}{Q}, \] “geometric degeneracy” means exact agreement with \(\Lambda\)CDM at the background level: the same \(H(z)\) and distance–redshift relations. Early-time viability forces \[ \alpha_1=1, \] after which the background reduces to \[ H^2(z)=H_0^2\left[\Omega_{m0}(1+z)^3+\Omega_{r0}(1+z)^4+\Omega_\Lambda\right]. \] The degeneracy is then broken only in the perturbation sector, because \[ \mu(z)=\frac{G_{\mathrm{eff}}(z)}{G_N}=\frac{1}{f_Q}, \] and the preferred parameter range implies \(G_{\mathrm{eff}}