Usually when performing calculations, we regard the energy levels of the system as being discrete. But in fact, excited states have a certain probability of decay via the emission of photons, and therefore have a finite lifetime. This implies that the levels become quasidiscrete, with a small but finite width; they can be written in the form $E-\tfrac{1}{2}i\Gamma$ where $\Gamma$ is the total probability of all possible decay channels (this fact is alternatively stated as the Optical Theorem). Often the width that develops is rather small compared to the gap between the discrete levels, so we can still measure sharp transitions. Someone else comments we shouldn't provide complete answers to homework questions, so instead I'm just going to elaborate on the concept and what it means since that's what you requested. For instance, another answer mentions the energy levels broaden into Lorentzians - it's not too hard to derive this fact.
Let's start from scratch, with the Schrodinger equation
\begin{equation}
\label{sequation}
i\frac{\partial \Psi}{\partial t}=\left(\hat{H}^{(0)}+\hat{V} \right)\Psi
\end{equation}
and expand the solution in terms of the wave functions of the unperturbed states of the system
\begin{equation}
\label{tdpt}
\Psi=\sum_\nu a_\nu (t) \Psi^{(0)}_\nu=\sum_\nu a_\nu (t) e^{-iE_\nu} \psi^{(0)}_\nu
\end{equation}
Inserting this expression either side of the Schrodinger equation and taking inner products with the state $\nu$ gives
\begin{equation}
\label{tdpt2}
i\frac{\partial a_\nu}{\partial t}=\sum_\nu\langle\nu|V|\nu '\rangle a_{\nu '} (t) \ e^{i(E_\nu-E_\nu')}
\end{equation}
The usual procedure for time dependent perturbation theory goes as follows. We assume the system is in some initial state with probability unity $a_1=1$ and $a_\nu=0$ for $\nu\neq 0$. Then to leading order we integrate both sides to find $a_\nu$ by only keeping terms on the left where $\nu'=1$ ie only keeping terms to leading order in $V$. The procedure for calculating $a_\nu$ more accurately proceeds by iteration. We are particularly interested in the long-time probability of decay
\begin{equation}
dw=|a_{\omega, 2}(\infty)|^2 \ d\omega
\end{equation}
where $\nu=2$ is some excited state and $\omega$ denotes the emission of a photon. In the usual case of time-dependent perturbation theory, we are effectively assuming our results apply to time longer than the inverse level spacing but short compared to the decay lifetime of the energy levels. Now let's relax that assumption - when we reach times comparable to the decay life of the state $1$ then $a_1=\exp{\left( -\tfrac{1}{2}\Gamma_1 t \right)}$ and the equation for $a_{\omega, 2}$ becomes
\begin{equation}
\label{modtdpt}
i\frac{d a_{\omega, 2}}{dt}=\langle \omega, 2| V |1\rangle e^{i(\omega-\omega_{12})t-\frac{1}{2}\Gamma_1 t}
\end{equation}
Integrating otherwise as usual, and substituting into the formula for the long-time decay probability gives
\begin{equation}
\label{bwdist}
dw=|\langle\omega, 2| V|1\rangle|^2 \frac{1}{(\omega-\omega_{12})^2+\frac{1}{4}\Gamma_1^2} \ d\omega
\end{equation}
If we assume that the width is small, then this expression is dominated by the frequency range $\omega\approx \omega_{12}$, and we recover the usual Fermi Golden Rule.
Now we define the expression
\begin{equation}
\Gamma_{1\rightarrow 2}=2\pi \sum_{pol,\boldsymbol{k}}|\langle \omega 2| V|1\rangle|^2
\end{equation}
which gives the total probability of emission, after summing over the polarisations and directions of motion of the photon. Then summing our expression for the long-time decay rate similarly over the polarisations and momenta we arrive at the total frequency distribution for emitted radiation
\begin{equation}
dw=\frac{\Gamma_{1\rightarrow 2}}{2\pi}\frac{1}{(\omega-\omega_{12})^2+\frac{1}{4}\Gamma_1^2} d\omega
\end{equation}
This broadening of the spectral line occurs for an isolated atom at rest, as distinct from broadening caused by the interaction of atoms with other atoms ($\textit{collision broadening}$) or by the presence of atoms in the source which move with various velocities ($\textit{Doppler broadening}$). It is called the $\textit{natural shape}$, and it's clear that it's a Lorentzian peak as claimed by the other answers. (Also: we could continue to add refinements to this calculation by taking into account the lifetime of the level $2$).