Sums of triangular numbers and sums of squares
Abstract
For non-negative integers a,b, and n, let N(a, b; n) be the number of representations of n as a sum of squares with coefficients 1 or 3 (a of ones and b of threes). Let N*(a,b; n) be the number of representations of n as a sum of odd squares with coefficients 1 or 3 (a of ones and b of threes). We have that N*(a,b;8n+a+3b) is the number of representations of n as a sum of triangular numbers with coefficients 1 or 3 (a of ones and b of threes). It is known that for a and b satisfying 1≤ a+3b ≤ 7, we have N*(a,b;8n+a+3b)= 22+a4+ab N(a,b;8n+a+3b) and for a and b satisfying a+3b=8, we have N*(a,b;8n+a+3b) = 22+a4+ab ( N(a,b;8n+a+3b) - N(a,b; (8n+a+3b)/4) ). %& t(8,0;n) = 136 ( N(8,0;8n+8) - N(8,0;2n+2) ). eq315 Such identities are not known for a+3b>8. In this paper, for general a and b with a+b even, we prove asymptotic equivalence of formulas similar to the above, as n→∞. One of our main results extends a theorem of Bateman, Datskovsky, and Knopp where the case b=0 and general a was considered. Our approach is different from Bateman-Datskovsky-Knopp's proof where the circle method and singular series were used. We achieve our results by explicitly computing the Eisenstein components of the generating functions of N*(a,b;8n+a+3b) and N(a,b;8n+a+3b). The method we use is robust and can be adapted in studying the asymptotics of other representation numbers with general coefficients.
Turn this paper into a lesson
ArcXiv compiles a structured reading guide from this paper's metadata: plain-English importance, contributions, prerequisite concepts, which sections to read first, flashcards, and a quiz. Grounded in the abstract, never invented.