fft – How interpolation of zeros in frequency domain data affects time domain version of the signal?

If I have four data d[0]-d[3] in time domain, say after taking FFT I find outputs D[0]-D[3]. If I interpolated zeros inbetween the frequency domain data’s i.e., D[0],0,D[1],0,…,D[3],0.
After that when I take IFFT of the frequency domain data, I see that it presents a scaled repeated version of the original time domain data’s d[0]-d[3]. Is there any mathematical explanation to this that why this is happening?
Also, when I insert those 4 zeros after all the FFT outputs such as D[0],D[1],D[2],D[3],0,0,0,0. I see taking IFFT of that zero-inserted frequency domain data shows different output than as explained for the first case above. What might cause the difference in the behaviors?