I don't think such a sequence exists:
Sn = n/2 (2a + (n-1)d)
32 = n/2 (4 + (n-1)1.5)
64 = n (2.5 + 1.5n)
1.5n^2 + 2.5n - 64 = 0
3n^2 + 5n - 128 = 0
No integral roots to that equation, which is what we are looking for. N has to be a positive integer, as it denotes a term