The paper reaches two main algebraic conclusions. For every n at least 4, it constructs a subring R_n of n-by-n matrices over F2, the field with two elements, and proves that its null ideal is not a two-sided ideal of the polynomial ring R_n[x]. At the same time, it proves that the null ideal is two-sided whenever the ring’s Jacobson radical has nilpotency at most 3, meaning that repeated products of that radical become zero within the stated bound.
The question concerns the null ideal N(R), a set of polynomials tied to a finite associative ring with unity. The conjecture under examination says that N(R) is always a two-sided ideal of the surrounding polynomial ring. In practical terms, two-sidedness asks whether multiplying a null polynomial on either side by a polynomial from R[x] keeps the result in N(R). The paper works with a central indeterminate, so the polynomial variable commutes with the ring’s elements.
The proof narrows the test
The analysis reduces the two-sidedness test to right multiplication by the idempotents in E(R). Idempotents are elements that remain unchanged when multiplied by themselves. This reduction turns the broader question into a focused check on a distinguished set of ring elements.
For the low-nilpotency result, the proof works inside the Jacobson radical J. It uses Peirce decompositions—ways of breaking the ring into pieces determined by orthogonal idempotents—to build two-sided ideals inside J. The decomposition supplies the structure used to prove the nilpotency-at-most-3 guarantee.
The exception has a precise shape
The universal conjecture does not survive the known upper-triangular example. That ring is a subring of 4 × 4 upper-triangular matrices over F2; its null ideal is not two-sided, and its Jacobson radical has nilpotency 4. The example has order 128, but the paper does not settle whether some smaller finite ring can produce the same failure.
The paper then gives a formal definition for the kind of counterexample it seeks. Such a ring has two F2 residue summands, has radical nilpotency n at least 4, and contains an element meeting a specified nonzero-power condition. In the paper’s notation, the relevant residue object is R/J. The two-summand requirement is necessary for a non-two-sided null ideal, but it is not enough by itself to guarantee one.
When those construction conditions are met, the paper produces an explicit polynomial witness. Its right product with the designated element, written as fe in the algebraic argument, falls outside N(R). That single failure of closure is enough to show that the null ideal is not two-sided.
One construction, all matrix sizes
The family R_n extends the exception across every matrix size n at least 4. For each such n, the paper constructs a subring of M_n(F2) whose Jacobson radical has nilpotency n and whose null ideal is not two-sided. This is a family of explicit algebraic examples, not a claim that every ring of the same size or nilpotency behaves this way.
The proof has a parity wrinkle. For n at least 4, even-indexed R_n satisfy every condition in the paper’s formal counterexample definition. Odd-indexed R_n satisfy all of those conditions except one, labeled Definition 5.1(4b), so the odd cases need a separate argument.
For odd n at least 5, that extra step supplies an element beta in eJf but outside eJ²f, with the relation beta alpha^(n−2) = alpha^(n−1) for every alpha in J. In less compressed terms, the paper finds a bridge element that reproduces the power needed by the counterexample argument even though the formal definition is not met in full.
Another lemma gives a concrete witness inside every member of the family: for each n at least 4, there is an element r_n in J(R_n) for which (r_n² + r_n)^(n−1) is nonzero. The calculation records the nonvanishing ingredient behind the family’s counterexamples.
The boundary is still incomplete
The results leave the smallest possible counterexample unresolved. The known example has order 128, and the paper leaves open whether a finite ring of order below 128 can have a non-two-sided null ideal. The family theorem does not answer that order question because it establishes examples through a defined matrix-ring construction.
A second question concerns the residue structure. Every constructed R_n has R_4 as a residue ring, and the paper asks whether a non-two-sided example can exist without that residue ring. That possibility remains unresolved.
The supplied document is marked arXiv:2608.25853v1, so readers are seeing the result in preprint form. Its stated setting is a finite associative ring with unity and a polynomial ring in a central indeterminate.
Paper data and sources
Original title: Some results on null ideals of finite rings
Authors: Nicholas J. Werner
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-26
DOI: Not available
Original paper · Full text