Energy Maximisation for Integral Circulant Graphs with Opposite-Parity Exponents

@misc{jiang2026energy,
  title = {{Energy Maximisation for Integral Circulant Graphs with Opposite-Parity Exponents}},
  author = {Jianwei Jiang and Chunhua Yang},
  howpublished = {arXiv},
  year = {2026},
  month = aug,
  doi = {10.48550/arxiv.2608.29523},
  eprint = {2608.29523},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2608.29523},
}

TY  - UNPB
TI  - Energy Maximisation for Integral Circulant Graphs with Opposite-Parity Exponents
AU  - Jianwei Jiang
AU  - Chunhua Yang
PY  - 2026
DA  - 2026/08/30
PB  - arXiv
DO  - 10.48550/arxiv.2608.29523
UR  - https://arxiv.org/abs/2608.29523
AB  - For a finite graph, its energy is the sum of the absolute values of its adjacency eigenvalues. Let n be the product of powers of distinct odd primes, where all exponents are positive integers, the sum of the exponents is odd, and every prime whose exponent is odd is at least five. Among all nonempty sets of proper divisors of n, the energy of the corresponding integral circulant graph is uniquely maximised by the checkerboard divisor set consisting of those divisors for which the sum of the prime-power exponents is even. The maximum energy is one half of the sum of n and the product, over all prime factors, of explicit polynomial quantities. For exponent k and parameter x, the corresponding polynomial is given by two k plus one times x to the k-th power, together with four times the sum from zero to k minus one of j plus one times x to the j-th power with alternating sign determined by the parity of k minus j. For the proof, we use weighted prime-power Ramanujan transforms, which give an exact energy model in tensor form and extend to a polynomial matrix family for real parameters greater than one. A semidefinite parity theorem shows that the matrix obtained from this polynomial family by an explicit positive diagonal normalisation is a real symmetric contraction whose unique eigenvalue of modulus one is the simple eigenvalue determined by the parity of the exponent; this holds for even exponents at least two and parameters at least three, and for odd exponents at least one and parameters at least five. Tensorising this spectral structure and centring the binary divisor tensor yield a sharp one-norm bound for the corresponding tensor operator on sign vectors. In particular, for n equal to p squared times q cubed, with distinct odd primes p and q and with q at least five, this proves Roldan's conjectured maximality in this range.
ER  -