--- title: Additively Indecomposable Quadratic Forms url: https://www.emergentmind.com/topics/additively-indecomposable-quadratic-forms type: topic --- # Additively Indecomposable Quadratic Forms Additively indecomposable quadratic forms arise in several adjacent literatures, and the phrase is not entirely uniform across them. In the arithmetic of totally real number fields, a totally positive definite quadratic form \(Q\) over \(\mathcal O_K\) is called additively indecomposable if it cannot be written as \[ Q=Q_1+Q_2 \] with \(Q_1,Q_2\) nonzero totally positive semi-definite quadratic forms in the same number of variables [2509.23857]. This notion is tightly linked to additively indecomposable totally positive integers, namely elements of \(\mathcal O_K^+\) that cannot be expressed as sums of two totally positive integers, and to orthogonal indecomposability of quadratic or symplectic modules, where no nontrivial orthogonal direct-sum decomposition exists [2301.13222]. The subject therefore sits at the intersection of the additive geometry of the totally positive cone, the theory of universal quadratic forms, and decomposition theory for quadratic objects in characteristic \(2\) and over semirings. ## 1. Definitions and the range of meanings For a totally real number field \(K\) with ring of integers \(\mathcal O_K\), the order relation \[ \alpha \succ \beta \quad \Longleftrightarrow \quad \alpha-\beta\in \mathcal O_K^+ \] is standard in the literature on quadratic forms over \(\mathcal O_K\). A quadratic form in \(n\) variables is written \[ Q(x_1,\ldots,x_n)=\sum_{1\le i\le j\le n} a_{ij}x_ix_j,\qquad a_{ij}\in\mathcal O_K, \] and is called classical if \(2\mid a_{ij}\) for \(i\neq j\). It is totally positive semi-definite if it takes values in \(\mathcal O_K^+\cup\{0\}\) on \(\mathcal O_K^n\), and totally positive definite if the only zero is the trivial one. In this arithmetic setting, additively indecomposable means precisely the impossibility of decomposing \(Q\) as a sum of two nonzero totally positive semi-definite forms [2509.23857]. A second notion, central to the theory of universal forms, concerns totally positive integers. An element \(\alpha\in\mathcal O_K^+\) is indecomposable if it cannot be written as \[ \alpha=\beta+\gamma \qquad\text{with }\beta,\gamma\in\mathcal O_K^+. \] The survey literature treats these elements as additive atoms in the totally positive cone and uses them as the main arithmetic input in lower-bound arguments for universal forms [2301.13222]. A third notion appears in module-theoretic settings. For symplectic \(kG\)-modules in characteristic \(2\), indecomposable means the absence of a nontrivial orthogonal decomposition \[ (M,B)\cong (M_1,B_1)\perp (M_2,B_2) \] with both summands smaller symplectic \(kG\)-modules; this is explicitly stronger than indecomposability of \(M\) as a \(kG\)-module [1712.00313]. Over semirings with unique base, a quadratic module is indecomposable if it has no nontrivial orthogonal decomposition into basic summands [1509.01039]. A common source of confusion is therefore terminological rather than mathematical. In arithmetic papers, “additively indecomposable quadratic form” refers to decomposition as a sum of totally positive semi-definite forms; in characteristic-\(2\) representation theory and semiring theory, the operative notion is orthogonal indecomposability. The notions are related by the general theme of forbidding nontrivial additive splitting, but they are not identical. ## 2. Indecomposable totally positive integers as the arithmetic source The arithmetic theory is driven by the observation that universal forms are controlled not only by local representation theory, but also by the additive structure of \(\mathcal O_K^+\). A quadratic \(\mathcal O_K\)-lattice \((L,Q)\) is universal if \[ \forall \alpha\in \mathcal O_K^+ \ \exists v\in L \text{ such that } Q(v)=\alpha, \] and indecomposable totally positive integers are singled out because every universal form must represent them in a particularly rigid way [2301.13222]. For diagonal forms this rigidity is immediate. If \[ \alpha=a_1v_1^2+\cdots+a_rv_r^2 \] and \(\alpha\) is indecomposable, then only one term can contribute, so \(\alpha=a_iv_i^2\) for some \(i\). Consequently, each square class of indecomposables forces a coefficient, yielding direct lower bounds on the rank of universal diagonal forms. The survey also records a basic but important fact: every totally positive unit is indecomposable [2301.13222]. In real quadratic fields there is a complete classical description due to Dress–Scharlau. If \(K=\mathbb Q(\sqrt D)\), the indecomposables are precisely the semiconvergents \[ \alpha=\alpha_{i,t}:=\alpha_i+t\alpha_{i+1}, \qquad i\ge -1\text{ odd},\quad 0\le tD-a^2=977\,393\,040 \] for the relevant \(a\) [1512.04691]. Higher-dimensional geometry provides another viewpoint. Under the Minkowski embedding \(\iota:K\hookrightarrow\mathbb R^n\), the Klein polyhedron \(K_K=\operatorname{Conv}(A^+)\) has boundary \(S_K\), called the sail. If \(a\in O_K\) lies on the sail \(S_K\), then \(a\) is indecomposable; in degree \(2\) the converse is true, but in higher degree there can be indecomposables strictly inside the Klein polyhedron [2403.18390]. This suggests that additively indecomposable quadratic forms are best viewed against a broader geometric background in which the relevant additive atoms are organized by continued fractions in degree \(2\) and by sails in higher degree. ## 3. Construction of additively indecomposable forms over totally real fields A basic construction principle is that indecomposable diagonal coefficients, together with a nonzero interaction term, force indecomposability of the form itself. If \[ Q(x,y)=\alpha x^2+\beta xy+\gamma y^2\in \mathcal O_K[x,y] \] is totally positive definite, and \(\alpha\) and \(\gamma\) are indecomposable integers in \(\mathcal O_K\) with \(\beta\neq 0\), then \(Q\) is additively indecomposable. As an immediate consequence, over every totally real field \(K\), there exists a non-classical additively indecomposable quadratic form in \(2\) variables; the example given is \[ x^2+xy+y^2, \] since \(1\) is indecomposable in every totally real field [2509.23857]. The same principle extends to longer “adjacency chains.” If \[ Q(x_1,\ldots,x_n)=\sum_{i=1}^n \alpha_i x_i^2+\sum_{j=1}^{n-1}\beta_{i,i+1}x_ix_{i+1} \] is totally positive definite, each \(\beta_{i,i+1}\neq 0\), and each \(\alpha_i\) is an indecomposable integer in \(\mathcal O_K\), then \(Q\) is additively indecomposable. The proof mechanism is combinatorial: indecomposability of the diagonal coefficients forces each diagonal term into only one summand in any putative decomposition \(Q=Q_1+Q_2\), and then a nonzero adjacent cross-term makes one summand fail total semidefiniteness because a \(2\times2\) principal submatrix has a zero diagonal entry but a nonzero off-diagonal entry [2509.23857]. This perspective yields a sharp contrast with the classical integral case. Over \(\mathbb Z\), there are no additively indecomposable classical quadratic forms in \(2\le n\le 5\) variables. Over totally real fields, by contrast, additively indecomposable classical binary forms do occur. In every real biquadratic field, there exists a classical, additively indecomposable quadratic form in \(2\) variables [2509.23857]. For simplest cubic fields \(K=(\rho)\), where \(\rho\) is the largest root of \[ x^3-ax^2-(a+3)x-1, \] the paper constructs explicit classical examples. One binary example is \[ 2x^2+2xy+(1+\rho+\rho^2)y^2, \] which is additively indecomposable because \(2(1+\rho+\rho^2)\succ 1\). A ternary example is \[ Q(x,y,z)=(1+\rho+\rho^2)x^2+(1+\rho'+\rho'^2)y^2+(1+\rho''+\rho''^2)z^2+2xy+2yz, \] whose total positive definiteness is checked by Sylvester’s criterion, with \[ \det(Q)=1+\rho'+\rho'^2\succ 0. \] Accordingly, there exist classical, additively indecomposable quadratic forms in \(2\) and \(3\) variables over every simplest cubic field \((\rho)\) with \(\mathcal O_K=\mathbb Z[\rho]\) [2509.23857]. ## 4. Universality, rank bounds, and arithmetic obstructions The strongest applications of indecomposability concern universal quadratic forms. The guiding principle is that a universal form must represent every indecomposable totally positive integer, and those representations are sufficiently rigid to force large rank. A particularly general criterion uses the codifferent \[ \mathcal O_K^\vee=\{\delta\in K:\Tr(\delta\alpha)\in \mathbb Z\ \forall \alpha\in\mathcal O_K\}. \] If there exists \(\delta\in \mathcal O_K^{\vee,+}\) such that \[ \Tr(\delta\alpha)=1, \] then \(\alpha\) must be indecomposable. More generally, if there are \(u\) elements \(\beta_1,\dots,\beta_u\in\mathcal O_K^+\) and some \(\delta\in\mathcal O_K^{\vee,+}\) with \[ \Tr(\beta_i\delta)=1\quad\text{for all }i, \] then \[ m(K)\ge \frac{u}{d},\qquad m_{\mathrm{class}}(K)\ge \frac{\sqrt u}{d}, \] where \(m(K)\) and \(m_{\mathrm{class}}(K)\) denote the minimal ranks of universal and classical universal lattices [2301.13222]. In the real quadratic case, the method is fully explicit. Over \(\mathbb Q(\sqrt6)\), the indecomposables up to multiplication by totally positive units fall into four square classes represented by \[ 1,\quad 3+\sqrt6,\quad 5+2\sqrt6,\quad 27+11\sqrt6, \] and this forces any diagonal universal form to have rank at least \(4\). At the same time, there is a uniform construction: if \(S\) is a set of representatives of indecomposables up to multiplication by totally positive units, then \[ \bigperp_{\sigma\in S}\langle \sigma,\sigma,\sigma,\sigma,\varepsilon\sigma,\varepsilon\sigma,\varepsilon\sigma,\varepsilon\sigma\rangle \] is universal and has \(8\#S\) variables, where \(\varepsilon\) is the totally positive fundamental unit [2301.13222]. The obstruction theory becomes asymptotic in families. For every positive integer \(r\), there are infinitely many quadratic fields \(\mathbb Q(\sqrt D)\) that do not have a universal lattice of rank \(\le r\). More precisely, for almost all squarefree \(D\), \[ m_{\mathrm{class}}(\mathbb Q(\sqrt D))\gg_\varepsilon D^{1/12-\varepsilon}, \qquad m(\mathbb Q(\sqrt D))\gg_\varepsilon D^{1/24-\varepsilon}. \] For simplest cubic fields \(K=\mathbb Q(\rho)\) with \[ f(x)=x^3-ax^2-(a+3)x-1, \] the indecomposables lead to concrete bounds \[ m_{\mathrm{class}}(K)\ge \frac{a^2}{6},\qquad m(K)\ge \frac{a}{3\sqrt2}, \] while there exists a diagonal universal form of rank \(\sim 3a^2\) [2301.13222]. Biquadratic fields furnish a particularly strong nonuniversality theory. There are sufficient conditions under which an indecomposable element of a quadratic subfield remains indecomposable in a biquadratic extension \(K=\mathbb Q(\sqrt p,\sqrt q)\), and these allow escalation arguments over \(K\). In particular, every classical universal totally positive quadratic form over \[ \mathbb Q(\sqrt2,\sqrt3) \] must have at least \(5\) variables, and every such form over \[ \mathbb Q(\sqrt6,\sqrt{19}) \] must have at least \(6\) variables [1802.07811]. The obstruction can be pushed further: no classical totally positive definite ternary quadratic form over the ring of integers of a totally real biquadratic field is universal [1909.05422]. The sail viewpoint packages these phenomena geometrically. For totally real biquadratic fields with unit signature rank at least \(3\), ranks of universal forms and numbers of indecomposables grow as a power of the discriminant, while the family \[ K=\mathbb Q(\sqrt5,\sqrt{p_n}) \] provides a contrasting example with \[ \frac{\log \Delta_K}{8w}\ \le\ \ell(K)\ \le\ \frac{\log \Delta_K}{8w}+1, \qquad w=\log(1+\sqrt5), \] so that the number of indecomposables modulo totally positive units grows only logarithmically [2403.18390]. A plausible implication is that “few” additively indecomposable integers and “few” additively indecomposable forms need not correlate uniformly across field families; the governing invariant is the ambient additive geometry rather than degree alone. ## 5. Orthogonal indecomposability in characteristic \(2\) In representation theory over a perfect field \(k\) of characteristic \(2\), quadratic forms are organized not by positivity but by their associated bilinear forms. For the Klein four group \[ G=C_2\times C_2=\{1,g_1,g_2,g_1g_2\}, \] a quadratic form on a \(kG\)-module \(M\) is a map \(q:M\to k\) such that \[ q(ax)=a^2 q(x),\qquad B_q(x,y)=q(x+y)-q(x)-q(y) \] is bilinear; in characteristic \(2\), \(B_q\) is alternating automatically. A symplectic \(kG\)-module \((M,B)\) is indecomposable if it cannot be written as a nontrivial orthogonal sum \[ (M,B)\cong (M_1,B_1)\perp (M_2,B_2) \] with both summands smaller symplectic \(kG\)-modules [1712.00313]. The classification in this setting is explicit. The indecomposable \(kG\)-modules that can carry indecomposable symplectic forms are \[ (KG)^2,\quad kG,\quad (kG)^2,\quad A_n\oplus B_n,\quad C_n(f),\quad C_n(f)^2,\quad C_n(\infty),\quad C_n(\infty)^2, \] and the paper determines all indecomposable symplectic forms on them up to isometry. The quadratic forms are then classified relative to the symplectic forms. A key structural fact is that once the associated bilinear form \(B_q\) is fixed, any other quadratic form with the same \(B_q\) differs by a diagonal correction: \[ Q' = Q + D \] for some diagonal matrix \(D\). The existence of a \(G\)-invariant quadratic form above a given symplectic form is therefore a system of explicit diagonal constraints, solved module by module [1712.00313]. This literature uses “indecomposable quadratic form” in the orthogonal-sum sense. It therefore matches the additive language only after translation: the relevant additive operation is block-diagonal orthogonal addition, not pointwise addition of totally positive semi-definite forms. The distinction is substantive. In arithmetic papers, the obstruction comes from the additive semigroup \(\mathcal O_K^+\); in characteristic \(2\), the obstruction comes from invariant-theoretic structure and isometry. A related characteristic-\(2\) result concerns principal indecomposable modules of finite groups. A principal indecomposable module \(P\) has quadratic type if it affords a non-degenerate \(G\)-invariant quadratic form, and this occurs if and only if there exist involutions \(s,t\in G\) such that \(st\) has odd order and \[ \frac{\varphi(st)}{2} \] is not an algebraic integer, where \(\varphi\) is the Brauer character of the corresponding simple module. Moreover, the number of isomorphism classes of quadratic principal indecomposable \(G\)-modules is equal to the number of strongly real conjugacy classes of odd order elements of \(G\) [1803.03182]. This is another instance in which indecomposability is module-theoretic and orthogonal rather than arithmetic. ## 6. Semiring analogues, graph-theoretic structure, and conceptual synthesis Over semirings, especially in the setting of free modules with unique base, quadratic-form decomposition acquires a combinatorial form. A quadratic form \[ q:V\to R \] satisfies \[ q(ax)=a^2q(x), \qquad q(x+y)=q(x)+q(y)+b(x,y) \] for some symmetric bilinear companion \(b\), but because semirings do not have subtraction, a quadratic form can have many companions. Orthogonality is phrased through quasilinearity on pairs of submodules, and a basic submodule \(W\subseteq V\) is indecomposable if it admits no nontrivial orthogonal decomposition \[ W = X \perp Y,\qquad X,Y\neq 0 \] [1509.01039]. The decisive structural theorem is graph-theoretic. Given a base \(B\) and a quasiminimal companion \(b\), one defines an equivalence relation on \(B\) by connectivity through chains \[ e=e_0,e_1,\dots,e_r=f \] with \[ b(e_i,e_{i+1})\neq 0. \] If \(W_k\) is the basic submodule spanned by an equivalence class \(B_k\), then each \(W_k\) is indecomposable and \[ V=\bigperp_{k\in K} W_k. \] Moreover, these \(W_k\) are exactly the indecomposable basic orthogonal summands. Thus a quadratic module is indecomposable if and only if the associated base graph is connected [1509.01039]. This setting also admits a cancellation theorem analogous to Witt cancellation and a tensor-product indecomposability theorem. If \(W_1\) is isotypically finite and \[ W_1\cong W_1', \qquad W_1\perp W_2 \cong W_1'\perp W_2', \] then \[ W_2\cong W_2'. \] For tensor products, if \(U=(U,y)\) is an indecomposable bilinear module and \(V=(V,q)\) an indecomposable quadratic module, then \(U\otimes_b V\) is indecomposable except in the special case where \(y\) is alternate, \(q\) is diagonally zero, and both \(U\) and \(V\) contain only even cycles, in which case exactly two indecomposable components arise [1509.01039]. Taken together, the arithmetic, representation-theoretic, and semiring literatures point to a unified conceptual picture. In the arithmetic of totally real fields, universal quadratic forms are governed by the fine additive geometry of \(\mathcal O_K^+\), and indecomposable totally positive algebraic integers behave as additive atoms that force rank and sometimes nonuniversality [2301.13222]. In characteristic \(2\) and over semirings, indecomposability is encoded by the impossibility of orthogonal decomposition, often reducible to explicit matrix normal forms or to connectivity in an interaction graph [1712.00313]. The common principle is that quadratic objects become rigid when their admissible additive splittings are sharply constrained; the ambient category determines what “addition” means. Source: https://www.emergentmind.com/topics/additively-indecomposable-quadratic-forms