arXiv++ Combinatorics

We study linear codes whose coordinates are the ordinary edges and self-loops of complete undirected graphs; a node erasure removes all coordinates incident with a failed vertex. The construction results are binary. For triple-node erasures, we extend the published cyclic construction by allowing a suitable cyclic check slope to depend on the prime graph length. An explicit determinant test proves that one of three fixed slope choices works at infinitely many prime lengths, unconditionally, and gives redundancy $3n-2$, one bit above the graph Singleton bound. We also give Singleton-optimal triple-node codes at $n=6,8,10,12$, together with a general ordinary-edge framework that isolates the remaining loop-completion problem. When $2$ is primitive modulo an odd prime $n$, a binary multi-slope construction corrects every $ρ$-node erasure for $2\leqρ<n$, with redundancy $ρn-(ρ-1)$ in the range $2\leqρ\leq(n+1)/2$. Returning to arbitrary prime powers, we derive exact generating transforms and inclusion--exclusion formulas for node-metric ball volumes, fixed-radius asymptotics, and packing, existence, and covering bounds. Finally, for the complementary clique-erasure metric, we obtain an exact weight enumerator and a Singleton-optimal node--clique duality.

Quadratic ball discrepancy defines an energy on codes in finite Hamming spaces. At perfect-code parameters, its exact minimizers are the perfect codes. We fix the alphabet size, length, and code cardinality and compare all codes with these parameters. We prove tiling-defect stability: excess discrepancy above the perfect-code benchmark controls the squared deviation of the distinguished ball-covering multiplicity from one. For one-error parameters satisfying sphere-packing and Lloyd integrality, the lower coefficient is $κ_{n,q}/q^2\geq1$. The uniform floor one is sharp, while the certified parameter-dependent coefficient can be much larger. An explicit parameter-dependent upper estimate is also available, and the two certified coefficients can be far apart. For any two-error parameter pair with $n\geq5$ satisfying sphere-packing divisibility and having two distinct integral Lloyd roots in the Hamming weight range, we obtain an explicit positive coefficient without assuming that a perfect code exists. For alphabets of size at least four, this conditional coefficient has a closed form and fixed-alphabet asymptotics. Direct certificates for the repetition and Golay families, combined with perfect-code classification, give tiling-defect stability for every nontrivial perfect code. Here stability concerns the ball-covering multiplicity profile, not symmetric-difference proximity to a particular perfect code. The defect is also a normalized chi-square smoothing error under uniform ball noise, so excess discrepancy controls holes, overlaps, defective ambient points, total variation, and Rényi divergence from uniformity of the ball-noise output. Competing codes need not be linear or satisfy a distance or error-correction constraint.

Stolarsky's invariance principle converts quadratic discrepancy into an energy-minimization problem. Barg developed its form for binary Hamming space and proved that binary perfect codes minimize total quadratic ball discrepancy among binary codes of the same length and cardinality. We prove an exact finite-alphabet counterpart: whenever a fixed Hamming space and cardinality admit a perfect code, the perfect codes are precisely the discrepancy minimizers. The competitors are arbitrary subsets of the prescribed cardinality, with no linearity or minimum-distance assumption. More sharply, we give explicit parameter-only lower benchmarks before existence is known. For every arithmetically admissible one-error parameter set, and for every nontrivial two-error parameter set satisfying the sphere-packing and integral Lloyd-root conditions, equality in the corresponding benchmark is equivalent to perfect tiling. The bounds are proved by explicit Fourier--Krawtchouk and polynomial certificates rather than by solving a numerical linear program. Thus, in every covered parameter regime, perfect-code existence is equivalent to attainment of an explicit variational target.

Let $G_q(d)$ be the unit-transfer graph on the nonnegative integer vectors whose $q$ coordinates sum to $d$, equivalently on the multiplicity profiles of size-$d$ multisets over $q$ symbols. The prime-checksum conjecture predicts that, for prime $q$ and all sufficiently large $d$, a largest independent set is a fiber of the natural cyclic checksum. We introduce weighted local covers of $G_q(d)$ by translated induced subgraphs. For fixed $q$, capped anchor profiles reduce the covering conditions for infinitely many degrees to a finite rational linear system. This method gives new proofs of the known cases $q=3$ and $q=4$ and determines $α(G_q(d))$ exactly for $q=5$ and $q=7$ in every degree, thereby proving the next two odd-prime cases of the conjecture. In the complementary regime where $d$ is fixed and $q$ grows, a partition-orbit reduction solves degree five for $q\ge7$, gives exact power-of-two families in degrees six, eight, and ten, and yields an asymptotically sharp upper bound through three terms for every fixed $d\ge7$. All computer-assisted assertions reduce to finite rational or integer systems and are supported by independently checkable certificates.

The Balister--Győri--Schelp (BGS) conjecture asks whether every zero-sum list of $2^{s-1}$ nonzero vectors in $\mathbb{F}_2^s$ is the prescribed-difference profile of a perfect matching. The conjecture remains open in general, whereas the classical Hall hyperplane case is solved when all prescribed differences cross between two affine copies of a hyperplane. We isolate the smallest mixed case beyond Hall: exactly two prescribed differences are internal. Although only two requests have changed type, the complete Hall permutation is replaced by a prescribed-difference bijection between two punctured copies of the hyperplane, with two unknown deleted vertices on each side. We call this the two-hole problem. We develop a new combinatorial method for prescribed-difference matchings, based on counting and the character structure of the binary vector space. Unlike the known Hall-type methods, which construct a matching through a sequence of local algorithmic choices, our approach proves existence through a global noncancellation phenomenon. This loss of algorithmic structure is compensated by a different advantage: the method can retain global boundary information that local exchanges do not control. As a first application, it gives a new proof of the binary Hall theorem, and it then yields a complete solution of the two-hole problem with no multiplicity assumption. We also give direct constructive proofs for symmetric even-multiplicity two-hole and four-hole families. More broadly, the new technique provides a framework for studying further subfamilies of the BGS problem by measuring how far their matching structure departs from the Hall case.