Hacky way: you can just cast a float array. Where reals and imag values are going one after another.
It depends on if audio is interleaved or not. If it's interleaved (most of the cases) left and right channels are in the array with STRIDE 2
Lowest frequency in your case is frequency of a period of 1024 samples. In case of 44100kHz it's ~23ms, lowest frequency of the spectrum will be 1/(1024/44100) (~43Hz). Next frequency will be twice of this (~86Hz) and so on.