If $s(t)$ is your signal in time domain, you want to make the operation $s(t+\Delta t)$, which according to the Fourier's transform properties, is equivalent to the operation :
$$
S(f)e^{j2\pi f\Delta t}
$$
$S(f)$ being $s(t)$ Fourier's transform.
In case of a discrete Fourier's transform, which is defined as
$$
S[k] = \sum^{N-1}_{n=0}{s[n]e^{\frac{-i2\pi kn}{N}}}
$$
with $s$ the signal and N the number of points of both signal and DFT.
The operation to be applied in order to induce a shift by a number c of samples to $s[n]$ and obtain $s[n+c]$ is :
$$
e^{\frac{-i2\pi kc}{N}}S[k]
$$
Which in matlab code would be done with :
exp(-j*2*pi*(1:(N-1))*c/N).*S_fdt;
flag
link below your question. $\endgroup$