I asked George Andrews about this problem and this was part of his reply: "Define g_1(n)=2n-1 and g_m(n)=g_m-1(n)(g_m-1+1) Thus g_1(n):1,3,5,7,9,11,13,15,... g_2(n):2,12,30,56,90,132,182,... g_3(n):6,156,930,3192,8190,17556,... g_4(n):42,24492,865830,... I claim that the sequence for a consists of all the values of g_m(n) for m>=1,n>=1"