--- title: Lowest Discriminant Ideal url: https://www.emergentmind.com/topics/lowest-discriminant-ideal type: topic --- # Lowest Discriminant Ideal “Lowest discriminant ideal” is not a universally fixed term. In current noncommutative algebra, especially in the theory of affine Cayley–Hamilton Hopf algebras and algebras with trace, it denotes the first discriminant ideal in the filtration \(D_k(A/C,\operatorname{tr})\) whose zero set is nonempty [2307.15477]. In arithmetic geometry, the closely related standard object is the minimal discriminant ideal of an elliptic curve, formed from local minimal discriminant valuations [2402.19183]. In invariant theory and singularity theory, nearby constructions isolate the lowest-degree or lowest-Newton-part discriminant data rather than an identically named ideal [1206.2358, 2104.08567]. The shared principle is extraction of the earliest nontrivial discriminant information, but the algebraic objects, indexing conventions, and geometric meanings differ substantially. ## 1. Terminological scope and historical placement In the Hopf-algebraic literature, the term has a precise formal meaning. For an algebra with trace \((A,C,\operatorname{tr})\), the discriminant ideals and modified discriminant ideals form filtrations indexed by a positive integer \(k\), and the “lowest discriminant ideal” is the one at the smallest level where the corresponding vanishing locus becomes nonempty [2307.15477]. This formulation is used systematically for Cayley–Hamilton Hopf algebras, first under a basic-identity-fiber hypothesis and later under the weaker Chevalley-property hypothesis [2307.15477, 2506.21879, 2604.15986]. The phrase is not standard across all discriminant theories. In the prime PI algebra framework of Brown–Yakimov, there is no separately defined formal object called the “lowest discriminant ideal”; instead one studies the entire family \(D_\ell(R/Z(R);\operatorname{tr})\) and \(MD_\ell(R/Z(R);\operatorname{tr})\), with the top index \(\ell=n^2\) playing the decisive role for the Azumaya locus [1702.04305]. In invariant theory of matrices, the nearest analogue is the minimal-degree nonzero homogeneous component of the vanishing ideal of the locus of matrices with boundedly many distinct eigenvalues, together with the lowest-degree invariant equation, namely the subdiscriminant [1206.2358]. In the study of plane curve singularities, the relevant intrinsic datum is the initial Newton polynomial of a discriminant, determined by the ideals \((f)\) and \((g)\) up to rescaling [2104.08567]. The paper on the \(D\)-plus discriminant is explicitly not about a discriminant ideal under that name, although it develops Vieta ideals, quotient rings, and elimination procedures that are structurally adjacent [2105.03856]. This diversity suggests that the phrase names a family of “first nontrivial discriminant” constructions rather than a single standard invariant. ## 2. Discriminant ideals for algebras with trace Let \((R,C,\operatorname{tr})\) be an algebra with trace, where \(C\) is central in \(R\) and \(\operatorname{tr}:R\to C\) is \(C\)-linear and cyclic. The \(n\)-th discriminant ideal is the ideal \[ D_n(R/C,\operatorname{tr})\subseteq C \] generated by determinants \[ \det\!\big(\operatorname{tr}(y_i y_j)\big)_{1\le i,j\le n}, \qquad (y_1,\dots,y_n)\in R^n, \] and the \(n\)-th modified discriminant ideal is generated by \[ \det\!\big(\operatorname{tr}(y_i y'_j)\big)_{1\le i,j\le n}, \qquad (y_1,\dots,y_n),(y'_1,\dots,y'_n)\in R^n. \] If \(R\) is free of rank \(N\) over \(C\), then the discriminant itself is \[ D(R/C,\operatorname{tr})=\det\!\big(\operatorname{tr}(a_i a_j)\big)_{1\le i,j\le N} \] for a \(C\)-basis \(\{a_1,\dots,a_N\}\) of \(R\) [2404.10366]. The zero sets are denoted \[ V_k:=V(D_k(R/C,\operatorname{tr})). \] For finitely generated Cayley–Hamilton algebras under the standing hypotheses used in the Hopf-algebraic papers, Brown–Yakimov’s formula identifies these loci representation-theoretically: \[ V_k= \left\{ \mathfrak m\in \operatorname{MaxSpec} C \ \middle|\ \sum_{V\in \operatorname{Irr}(R/\mathfrak mR)}(\dim V)^20\mid D_k(H/C,\operatorname{tr})(\mathfrak m)=0\} = \frac{|G_0|\dim(V)^2}{|\operatorname{Stab}_{G_0}(V)|}+1. \] Consequently, the lowest discriminant ideal has level \(|G_0|+1\), and a maximal ideal lies in its zero set exactly when the fiber contains maximally stable irreducibles; equivalently, when all irreducibles in that fiber are maximally stable. The same theory identifies the orbit of the identity point under left and right winding automorphisms as a distinguished subset of the lowest discriminant locus, and in important cases that orbit is the whole lowest locus [2307.15477]. A later generalization replaces the basic-identity-fiber hypothesis by the Chevalley property for the identity fiber algebra. In that setting, the level of the lowest discriminant ideal is \[ \operatorname{FPdim}\big(\operatorname{Gr}(H/\mathfrak m_{\overline\varepsilon}H)\big)+1. \] Moreover, every nonempty zero locus of a discriminant ideal contains the orbit of the identity element of \(\operatorname{maxSpec}C\) under the left or right winding automorphism group action, and if \(H\) has the Chevalley property then all discriminant ideals are trivial [2506.21879]. The 2026 refinement strengthens the geometric picture. If the identity fiber algebra has the Chevalley property, then an irreducible \(H\)-module \(V\) has the property that \(V\otimes W\) is completely reducible for every irreducible \(H\)-module \(W\) if and only if \(V\) is annihilated by the lowest discriminant ideal. The corresponding lowest discriminant subvariety \[ \mathcal V_\ell=\mathcal V(D_\ell(H/C;\operatorname{tr})) \] is a closed subgroup of the affine algebraic group \(\operatorname{maxSpec}C\). Under the same hypothesis, \[ H\text{ has the Chevalley property} \iff H/\mathfrak m_{\overline\varepsilon}H\text{ has the Chevalley property and }D_\ell(H/C;\operatorname{tr})=0, \] equivalently, all discriminant ideals are trivial [2604.15986]. Taken together, these results place the lowest discriminant ideal at the opposite end of the representation-theoretic spectrum from the Azumaya locus: it detects the fibers with minimal square-dimension sum and, in the Hopf setting, the strongest tensor-stability or tensor-reducibility behavior. ## 4. Explicit computation for central extensions of Abelian groups A concrete computation is available for group algebras of central extensions of Abelian groups. Let \[ 1\longrightarrow Z\longrightarrow \Gamma\xrightarrow{f}G\longrightarrow 1, \qquad G=A\times B, \] where \(A,B\) are finitely generated Abelian groups, \(Z\) is finite Abelian, and the extension is represented by a cocycle \(\sigma\in Z^2(G,Z)\) lying in the subgroup \(N^2(A\times B,Z)\). Choose a central subgroup \(L\triangleleft\Gamma\) with \(\operatorname{Im}(\sigma)\subseteq L\) and finite index \[ m=[\Gamma:L], \qquad C=\Bbbk L, \qquad H=\Bbbk\Gamma, \] over an algebraically closed field with \(\operatorname{char}\Bbbk\notin[1,m]\). Then \(H\) is a finite free \(C\)-module of rank \(m\), \((H,C,\operatorname{tr}_{\rm reg})\) is a Cayley–Hamilton Hopf algebra of degree \(m\), the identity fiber is basic, and \(GKdim(H)=GKdim(C)\) [2404.10366]. For every maximal ideal \(\mathfrak m\in\operatorname{MaxSpec}C\), the fiber algebra decomposes as \[ H/\mathfrak mH\cong R_1\otimes\cdots\otimes R_k, \] where \[ R_i\cong \frac{\Bbbk\langle x_i,y_i\rangle} {(x_i^{l_i}-1,\ y_i^{l_i}-1,\ x_i y_i-\xi^{n/l_i}y_i x_i)}. \] The fiber is simple, it has an irreducible module \(V\) of dimension \[ \dim(V)=\prod_{i=1}^k l_i, \] and \[ \operatorname{Stab}(V)\cong (\Bbb Z_{l_1}\times\cdots\times \Bbb Z_{l_k})^2, \qquad |\operatorname{Stab}(V)|=(\dim V)^2. \] Hence every irreducible in every fiber is maximally stable, and \[ \sum_{W\in \operatorname{Irr}(H/\mathfrak mH)}(\dim W)^2=m \] for all \(\mathfrak m\) [2404.10366]. The discriminant-locus consequence is completely uniform: \[ V(D_k(H/C,\operatorname{tr}))= \begin{cases} \varnothing, & k\le m,\\[4pt] \operatorname{MaxSpec}C, & k>m. \end{cases} \] Therefore the lowest discriminant ideal occurs at level \[ m+1, \] and its zero set is the entire maximal spectrum. The result determines the vanishing locus rather than an explicit generator of the ideal [2404.10366]. The same paper gives an explicit description of the orbit of the identity under winding automorphisms: \[ \mathcal O\cong L/\operatorname{Im}(\sigma). \] This orbit is the locus of basic fibers, whereas the zero set of the lowest discriminant ideal is much larger in this example—indeed all of \(\operatorname{MaxSpec}C\) [2404.10366]. ## 5. Invariant-theoretic analogues: matrices, subdiscriminants, and lowest-degree equations For matrix varieties, the phrase “lowest discriminant ideal” is best interpreted analogically. Let \(\mathcal E_k\) be the variety of real symmetric \(n\times n\) matrices having at most \(n-k-1\) distinct eigenvalues. The vanishing ideal \(I(\mathcal E_k)\) is graded, and the paper on subdiscriminants determines both the first nonzero invariant part and the true minimal-degree nonzero part [1206.2358]. The \(k\)-subdiscriminant \(\operatorname{sDisc}_k\) is the unique lowest-degree invariant equation. More precisely, \[ I(\mathcal E_k)^{SO_n}_{(n-k)(n-k-1)} \] is spanned by \(\operatorname{sDisc}_k\), and there are no \(SO_n\)-invariants in lower degree. The same statement holds for general matrices with \(GL_n\)-invariants [1206.2358]. For the full ideal of real symmetric matrices, the actual first nonzero homogeneous piece occurs earlier: \[ I(\mathcal E_k)_{\frac{(n-k)(n-k-1)}{2}}\neq 0. \] This component contains the image of \[ \left(\bigwedge^{n-k-1}\mathcal N\right)^* \] under the comorphism of the equivariant map \[ \mathcal T_k(A)=\bigwedge_{i=1}^{n-k-1}\left(A^i-\frac1n\operatorname{Tr}(A^i)I\right). \] The variety \(\mathcal E_k\) is exactly the common zero locus of this module of lower-degree equations [1206.2358]. The invariant subdiscriminant is then recovered as a quadratic expression on such a minimal-degree module: if \(W\subset I(\mathcal E_k)_{\frac{(n-k)(n-k-1)}2}\) is a nonzero \(SO_n\)-submodule, there is a basis \(f_i\) of \(W\) such that \[ \operatorname{sDisc}_k=\sum_i f_i^2. \] Thus the matrix-theoretic analogue distinguishes sharply between the lowest nonzero graded piece of the full ideal and the lowest invariant discriminant equation [1206.2358]. This is structurally close to Hopf-theoretic lowest discriminant ideals in that both isolate a first nontrivial discriminant stratum, but the actual objects are different: graded components of vanishing ideals in one case, trace-determinantal ideals in the other. ## 6. Arithmetic geometry: minimal discriminant ideals of elliptic curves Over a number field \(K\), the relevant object is the minimal discriminant ideal of an elliptic curve. If \(\mathfrak p\) is a prime of \(\mathcal O_K\), a \(\mathfrak p\)-minimal Weierstrass model minimizes \(v_{\mathfrak p}(\Delta)\) among integral models. The global minimal discriminant ideal is \[ D_E=\prod_{\mathfrak p}\mathfrak p^{\,v_{\mathfrak p}(\Delta_{\mathfrak p}^{\min})}. \] This is the correct global invariant because a global minimal model need not exist over a general number field [2402.19183]. Two elliptic curves \(E\) and \(E'\) over \(K\) are called discriminant ideal twins if they are not \(K\)-isomorphic and have the same conductor and the same minimal discriminant ideal. They are discriminant twins if, in addition, for each prime \(\mathfrak p\), there exist \(\mathfrak p\)-minimal models with equal local discriminants. The distinction is essential: equality of ideals is weaker than equality of local discriminant elements [2402.19183]. For prime isogenies with \(p\in\{3,5,7,13\}\), where \(X_0(p)\) has genus \(0\), the paper gives a complete parameterized classification. If \(E_1\) and \(E_2\) are \(p\)-isogenous and parameterized by \(\mathcal C_{p,1}(t_0,d_0)\) and \(\mathcal C_{p,2}(t_0,d_0)\), then \[ \frac{\Delta_{p,1}(t,d)}{\Delta_{p,2}(t,d)}=t^{p-1}p^{-12}. \] For \(p\in\{3,5,7,13\}\), \(E_1\) and \(E_2\) are discriminant ideal twins if and only if there exist \(t_0,d_0\in\mathcal O_K\) such that for every prime \(\mathfrak p\), \[ v_{\mathfrak p}(t_0)=\frac{12}{p-1}k, \qquad 0\le k\le v_{\mathfrak p}(p), \] and the two curves have the same Kodaira–Néron type at \(\mathfrak p\). They are discriminant twins if and only if, additionally, \[ t_0\in \mathcal O_K^{12/(p-1)}. \] For \(p=2\), the analogous valuation condition is necessary but not sufficient [2402.19183]. A parallel classification is proved for \(p^2\)-isogenies with \(p=3\) or \(5\), the cases where \(X_0(p^2)\) has genus \(0\). In the \(9\)-isogeny case, discriminant ideal twins are characterized by \[ \nu_{\mathfrak p}(t-3)=3k_{\mathfrak p}, \qquad 0\le k_{\mathfrak p}\le \nu_{\mathfrak p}(3), \] and discriminant twins require the extra condition \[ (t-3)^8\in \mathcal O_K^{12}. \] In the \(25\)-isogeny case, discriminant ideal twins are characterized by \[ \nu_{\mathfrak p}(t-1)=k_{\mathfrak p}, \qquad 0\le k_{\mathfrak p}\le \nu_{\mathfrak p}(5), \] and this already implies the stronger discriminant-twin property [2403.01287]. Here “lowest discriminant ideal” is genuinely arithmetic: it is the ideal of minimal local discriminant exponents, finer than the conductor and not preserved by isogeny in general. ## 7. Nearby constructions: lowest Newton data, minimal plane-curve discriminants, and discriminant substitutes Several adjacent literatures isolate “lowest” discriminant information without defining a lowest discriminant ideal in the Hopf-theoretic or arithmetic sense. For a holomorphic map germ \[ (f,g):(\mathbb C^2,0)\to(\mathbb C^2,0) \] with isolated zero, the intrinsic datum is the initial Newton polynomial of the discriminant. It is determined, up to rescaling variables, by the ideals \((f)\) and \((g)\) in \(\mathbb C\{x,y\}\). The preserved object is therefore the compact-edge part of the discriminant, rather than an ideal in coefficient space [2104.08567]. In the geometry of plane curves, the discriminant of a separable polynomial \(f\in K[x,y]\) with respect to \(y\) is measured by \(\deg_x\Delta_y(f)\). For irreducible \(f\), \[ \deg_x\Delta_y(f)\ge 2g+d_y-1, \] and the minimal case is \[ \deg_x\Delta_y(f)=d_y-1. \] Irreducible monic polynomials with minimal discriminant are exactly coordinate polynomials. This is a lowest-discriminant problem in the sense of extremal degree, not an ideal-theoretic construction [1507.01091]. The \(D\)-plus discriminant furnishes a different kind of substitute for a vanishing discriminant. For \[ p(x)=a_n\prod_{i=1}^m(x-\rho_i)^{\mu_i}, \] the paper defines \[ D^+(p)=\prod_{1\le i