Year-end appeal: Please make a donation to the OEIS Foundation to support ongoing development and maintenance of the OEIS. We are now in our 61st year, we have over 378,000 sequences, and we’ve reached 11,000 citations (which often say “discovered thanks to the OEIS”).
%I #24 Aug 28 2022 08:28:44
%S 0,1,4,125,84277,1435150,9061191,249858189,2799936925,146234239784,
%T 1139643680683264,7471434609455791,21274660147684109,
%U 2911209509190673141,15845190736671957299,995980378496501932493,213688560255016550712685,28372206851301867342910959
%N Numbers k such that floor(k^2*phi) is a square, where phi = A001622 is the golden ratio.
%C Numbers k such that there exists m such that m^2 <= k^2*phi < m^2+1. For k > 0, m/k is a lower convergent to sqrt(phi) since m/k < sqrt(phi) < sqrt(m^2+1)/k, so |sqrt(phi) - m/k| < (sqrt(m^2+1)-m)/k < 1/(2*k^2) (see the Mathematics Stack Exchange link). As a result, this is a subsequence of {A225205(2*n): n>=0}. The terms > 0 are b(0), b(2), b(4), b(10), b(12), b(14), ... for b = A225205. Note that sqrt(phi) - A225204(2*r)/A225205(2*r) < 1/(A225205(2*r)*A225205(2*r+1)) (by Theorem 5 of the Wikipedia link), so A225205(2*r) is a term if sqrt(A225204(2*r)^2+1) + A225204(2*r) < A225205(2*r+1).
%C Consider the numbers s such that A035513(s,0) = A000201(s) = floor(s*phi) and A035513(s,1) = A003622(s) = floor(floor(s*phi)*phi) are both squares. If s has this property, then clearly floor(s*phi) is the square of a term. However, for k > 0 being a term, k^2 is not always in A000201. This happens if and only if {k^2*phi} < phi^(-2), where {} denotes the fractional part (see A000201); for example k = 2799936925, for which floor(k^2*phi) = 3561574786^2. In this case, floor(k^2*phi) is not in A003622 since n -> floor(n*phi) is an injection. Suppose that A035513(p,q) = floor(k^2*phi) for p >= 1, q >= 2, then A035513(p,q-1) = floor(k^2*phi) since floor((k^2+1)*phi) = floor(k^2*phi)+1 in this case.
%H Jianing Song, <a href="/A354549/b354549.txt">Table of n, a(n) for n = 1..140</a>
%H Mathematics Stack Exchange, <a href="https://math.stackexchange.com/questions/4446187/if-bigx-fracpq-big-frac12q2-then-p-q-is-necessarily-one-of-the">If |x - p/q| < 1/(2*q^2) then p/q is necessarily one of the convergents</a>
%H Wikipedia, <a href="https://en.wikipedia.org/wiki/Continued_fraction#Some_useful_theorems">Continued fraction</a>
%e 4 is a term since floor(4^2*phi) = 5^2.
%e 125 is a term since floor(125^2*phi) = 159^2.
%o (PARI) A000201(n) = (n+sqrtint(5*n^2))\2; isok(n) = issquare(A000201(n^2))
%o (PARI) print1("0, 1, "); my(cofr=A331692_vector_bits(1000), conv=matrix(2,#cofr)); conv[,1]=[1,1]~; conv[,2]=[4,3]~; for(n=3, #cofr, conv[,n]=cofr[n]*conv[,n-1]+conv[,n-2]; if(n%2 == 1 && (conv[1,n]^2+1)^2 - (conv[1,n]^2+1)*(conv[2,n]^2) - (conv[2,n]^2)^2 > 0, print1(conv[2,n], ", ")))
%o \\ Here conv[1,n] = A225204(n-1), conv[2,n] = A225205(n-1), n odd implies conv[1,n]/conv[2,n] < sqrt((1+sqrt(5))/2); let A = conv[1,n]^2+1, B = conv[2,n]^2, then A^2 - A*B - B^2 > 0 implies A/B > (1+sqrt(5))/2
%o \\ Modified by _Jianing Song_, Aug 28 2022 according to _Kevin Ryde_'s program for A331692
%Y Cf. A001622, A000201, A003622, A035513, A225204, A225205.
%K nonn
%O 1,3
%A _Jianing Song_, Aug 18 2022