Nonexistence of DSRGs with ten parameter sets: the rank of A−θIA - \theta I versus a submatrix prescribed at a vertex

Iurii Ivanov

Ivanov’s vertex-submatrix arguments and finite checks rule out ten listed parameter sets of directed strongly regular graphs and their complements. Eight pairs were marked open in the 6 November table cited by the paper. The other two pairs had earlier nonexistence proofs.

What was already known

Hobart and Brouwer’s table records known constructions, nonexistence results and open cases. The paper reworks two earlier rStation manuscripts and credits prior proofs for two parameter pairs.

AI involvement AI-led

Ivanov chose the problem and directed the research. He credits Claude agents with the arguments, proofs, checking programs and manuscript.

Importance Notable

The rank method compares the rank allowed by the graph spectrum with the rank forced by a vertex neighborhood. It resolves specific cases left open by the tabulated feasibility conditions.

Technical verification Partial

The project reports Lean proofs for nine of the ten parameter sets. For the remaining case, two separate programs ran the finite search and drat-trim checked the SAT solver’s certificates. The reduction to that search was reviewed by AI agents.

Formal checking

The project reports Lean proofs for nine of the ten parameter sets. For the remaining case, two separate programs ran the finite search and drat-trim checked the SAT solver’s certificates. The reduction to that search was reviewed by AI agents.

Checker
Lean 4 and drat-trim in the project’s checking workflow
Scope
Ivanov’s vertex-submatrix arguments and finite checks rule out ten listed parameter sets of directed strongly regular graphs and their complements. Eight pairs were marked open in the 6 November table cited by the paper. The other two pairs had earlier nonexistence proofs.
Basis
Reported by the linked checker

Human understanding Few

The manuscript gives the rank argument and its cases. Ivanov attributes that exposition to the model agents and reports no line-by-line human review.

Correction and publication history