Reducibility of spectral curves of finite Jacobi pencils

Boris Shapiro · 2026

Read Reducibility of spectral curves of finite Jacobi pencils by Boris Shapiro (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-14817.txt.

Introduction

Shapiro studies when the spectral curve of a finite pencil of Jacobi matrices is reducible. The pencil adds a diagonal matrix to a tridiagonal matrix with zero main entries, scaled by a parameter. Generic irreducibility is proved when entries are pairwise distinct. Several basic reducibility mechanisms are given, including disconnected Jacobi chains, constant eigenvalue branches, mirror-symmetric parts, and scalar blocks. A conjecture on reducibility is stated, with low-size evidence and counterexamples to optimistic claims.

Catalog record

CREATORBoris Shapiro
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27255
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.14817

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Reducibility of spectral curves of finite Jacobi pencils Boris Shapiro Abstract . We consider finite pencils of Jacobi matrices J n ​ ( w ) = A + w ​ B , J_{n}(w)=A+wB, where A A is diagonal and B B is tridiagonal with zero diagonal. The spectral curve is the affine plane curve χ n ​ ( λ , w ) = det ( λ ​ I + J n ​ ( w ) ) = 0 . \chi_{n}(\lambda,w)=\det(\lambda I+J_{n}(w))=0. The main question is to describe when this curve is reducible. We prove generic irreducibility for fixed pairwise distinct diagonal entries and discuss several elementary reducibility mechanisms. Besides disconnected Jacobi chains, constant eigenvalue branches, and reflection-symmetric components, one must also take into account reducibility caused by scalar diagonal blocks. We formulate a reducibility conjecture and record low-dimensional evidence and counterexamples to several overly optimistic classifications. A central point of the picture is a codimension-growth principle: apart from the cutting divisors b i = 0 b_{i}=0 , genuinely connected primitive reducibility should move to higher and higher codimension as the size of the chain grows. 2020 Mathematics Subject Classification 15A18, 15A29, 14H50, 47B36 1. Introduction Let J_{n}(w)=\begin{pmatrix}a_{1}&wb_{1}&0&\cdots&0\\ wb_{1}&a_{2}&wb_{2}&\ddots&\vdots\\ 0&wb_{2}&a_{3}&\ddots&0\\ \vdots&\ddots&\ddots&\ddots&wb_{n-1}\\ 0&\cdots&0&wb_{n-1}&a_{n}\end{pmatrix} be a finite Jacobi pencil. Our considerations of such pencils are motivated by similar infinite-dimensional objects studied in [ 1 , 3 ] , see discussions below. We use the sign convention \chi_{n}(\lambda,w)=\det(\lambda I+J_{n}(w)). The affine curve \chi_{n}(\lambda,w)=0 will be called the spectral curve of the pencil. Since w enters only through the off-diagonal entries, \chi_{n} is an even polynomial in w . We often write t=w^{2},\qquad P_{n}(\lambda,t)=\chi_{n}(\lambda,w). Then P_{n} is the continuant determined by P_{0}=1,\qquad P_{1}=\lambda+a_{1},\qquad P_{k}=(\lambda+a_{k})P_{k-1}-tb_{k-1}^{2}P_{k-2}. (1) The problem studied in this note is the following. For which parameters a_{i},b_{i} is \chi_{n}(\lambda,w) reducible in \mathbb{C}[\lambda,w] ? There are several elementary mechanisms. First, if some b_{i}=0 , the chain splits into two smaller Jacobi chains. Secondly, \chi_{n} can have a constant eigenvalue branch, i.e. a factor \lambda+a independent of w . Thirdly, a component of the chain may be invariant under reversal. Then the Jacobi matrix decomposes into symmetric and anti-symmetric subspaces for the reversal, giving a factorization of the characteristic polynomial. Finally, if a connected block has scalar diagonal part, say a_{r}=\cdots=a_{s} , then its characteristic polynomial is the characteristic polynomial of a constant tridiagonal matrix evaluated at \lambda+a_{r} and w ; over \mathbb{C} this usually splits into linear factors. The goal of the paper is to collect these elementary mechanisms, prove generic irreducibility, and formulate a conjectural picture. This finite-dimensional problem was motivated by the work of Djakov and Mityagin on spectral Riemann surfaces and trace formulae for infinite tridiagonal operator pencils. In that setting one studies operators of the form L+zB , where L is diagonal and B is tridiagonal, and the spectral Riemann surface records the analytic continuation and monodromy of the eigenvalue branches. The present paper studies a finite algebraic analogue: for a finite Jacobi pencil we ask when the polynomial spectral curve \chi_{n}(\lambda,w)=0 is reducible. Thus the reducibility problem considered below may be viewed as a finite-dimensional counterpart of the irreducibility questions for spectral Riemann surfaces in [ 1 ] ; the subsequent work on convergence radii and related tridiagonal spectral problems gives additional background for this point of view [ 2 , 3 ] . 2. Generic irreducibility Theorem 2.1 . Assume that a_{1},\ldots,a_{n} are pairwise distinct. Then, for fixed a_{1}, \ldots, a_{n} , the polynomial \chi_{n}(\lambda,w) is irreducible in \mathbb{C}[\lambda,w] for all (b_{1},\ldots,b_{n-1}) outside a proper algebraic subset of \mathbb{C}^{n-1} . Proof. For fixed a_{1},\ldots,a_{n} , reducibility of the monic polynomial \chi_{n} is a Zariski closed condition in the variables b_{1},\ldots,b_{n-1} . Indeed, for each possible degree of a non-trivial factor one writes an undetermined factorization and eliminates the coefficients of the two factors. It remains to exhibit one point in the parameter space at which \chi_{n} is irreducible. We use a standard monodromy argument. Choose non-zero b_{i} outside a finite union of proper algebraic hypersurfaces specified below. At w=0 the n branches of the covering \chi_{n}(\lambda,w)=0\longrightarrow\mathbb{C}_{w} are the distinct points \lambda=-a_{i} . Fix an adjacent pair i,i+1 . If all couplings except b_{i} are set to zero, the two corresponding sheets are described by the elementary 2\times 2 block \begin{pmatrix}a_{i}&b_{i}w\\ b_{i}w&a_{i+1}\end{pmatrix}, and its two eigenvalues have square-root ramification at the two points w=\pm{a_{i}-a_{i+1}\over 2b_{i}}. For sufficiently small but non-zero neighbouring couplings this simple ramification persists by the implicit function theorem. Avoiding the finitely many resultants which would make this branch point collide with another ramification point or with a third sheet, the local monodromy around it is the transposition of the two sheets analytically continued from -a_{i} and -a_{i+1} . Doing this for each adjacent pair and then avoiding the union of the corresponding exceptional hypersurfaces, the monodromy group contains all adjacent transpositions (1\ 2),(2\ 3),\ldots,(n-1\ n). These transpositions generate S_{n} , so the monodromy is transitive. Hence \chi_{n} is irreducible over \mathbb{C}(w) . Since \chi_{n} is monic in \lambda , Gauss’ lemma gives irreducibility in \mathbb{C}[\lambda,w] . Therefore the reducible locus is a proper algebraic subset. ∎ Lemma 2.2 . Assume that b_{i}\neq 0 for all i . Then consecutive continuants are coprime: \gcd(\chi_{n},\chi_{n-1})=1 in \mathbb{C}[\lambda,w] . Proof. The continuant recurrence gives \chi_{n}=(\lambda+a_{n})\chi_{n-1}-b_{n-1}^{2}w^{2}\chi_{n-2}. If a nonconstant polynomial d divides both \chi_{n} and \chi_{n-1} , then d divides w^{2}\chi_{n-2} . Since \chi_{n-1}(\lambda,0)=\prod_{j=1}^{n-1}(\lambda+a_{j}), the polynomial \chi_{n-1} is not divisible by w . Hence d divides \chi_{n-2} . Repeating the argument inductively yields that d divides \chi_{1}=\lambda+a_{1} and \chi_{0}=1 , impossible. ∎ Corollary 2.3 . For connected Jacobi chains the spectral curves \chi_{n}(\lambda,w)=0 and \chi_{n-1}(\lambda,w)=0 have no common irreducible component. Corollary 2.4 . Under the assumptions of Theorem 2.1 , the Galois group of \chi_{n}(\lambda,w) over \mathbb{C}(w) is generically the full symmetric group S_{n} . Proof. The proof of Theorem 2.1 shows that the monodromy group contains all adjacent transpositions. Hence it equals S_{n} . ∎ 3. Elementary reducibility mechanisms Put c_{i}=b_{i}^{2} . Proposition 3.1 . The following conditions imply reducibility of \chi_{n} . (1) If b_{i}=0 , then \chi_{n}=\chi_{i}^{(1,\ldots,i)}\,\chi_{n-i}^{(i+1,\ldots,n)}, where the two factors are the characteristic polynomials of the two resulting smaller Jacobi chains. (2) If \chi_{n}(-a,w)\equiv 0 , then \lambda+a divides \chi_{n} . (3) Suppose that a connected block indexed by r,\ldots,s is reflection-symmetric up to diagonal sign changes, i.e. a_{r+k}=a_{s-k},\qquad b_{r+k}^{2}=b_{s-k-1}^{2} for all admissible k . Then the characteristic polynomial of this block factors as the product of its restrictions to the symmetric and anti-symmetric subspaces of the reversal. (4) If a connected block I=[r,s] has scalar diagonal part, a_{r}=a_{r+1}=\cdots=a_{s}=a, then J_{I}(w)=aI+wB_{I},\qquad\chi_{I}(\lambda,w)=\det((\lambda+a)I+wB_{I}). Consequently, over \mathbb{C} , \chi_{I}(\lambda,w)=\prod_{\mu\in\operatorname{Spec}(B_{I})}(\lambda+a+\mu w), with multiplicities. In particular, every scalar diagonal block of length at least two gives a reducible spectral curve. Proof. The first statement is immediate from block diagonal decomposition. The second statement is the factor theorem. For the third statement, after changing signs of basis vectors one may assume b_{r+k}=b_{s-k-1} . The corresponding Jacobi block commutes with the reversal operator. Hence it preserves the eigenspaces of the reversal with eigenvalues +1 and -1 , and its characteristic polynomial is the product of the two restricted characteristic polynomials. The fourth statement follows by diagonalizing the constant matrix B_{I} over \mathbb{C} . ∎

Read the complete text