Dear Editor,

thank you for your positive assessment of our manuscript, and thanks to the reviewer for pointing out very valid points. Let me summarize our answers to them individually:

Point 1 about the quantum efficiency of the modulation scheme is a bit difficult to answer, because we start with a cw light beam with a constant power, and shape this into a pulse with a well-defined energy or average photon number. To make a direct comparison, one would need to specify what the starting pulse with a  well defined energy or photon number looks like out of which the excitation pulse is cut with the method presented in this paper. At the moment, we are considering a few options with very different pulse shapes, but none of that has reached an experimental state yet.
We operate the EOM in the low distortion regime, where the peak transfer efficiency into one of the side bands is around 5%, on top of the insertion loss of the modulator of about 50%. This bounds the our overall efficiency to maybe 2%, which indeed is very low. Obviously, for efficiently shaping a given single photon pulse emitted by some source into a single photon pulse suitable for efficient absorption one would need a different approach.

Point 2, a comparison between a square pulse and a rising exponential. The detailed analysis for various parameter regimes and field states is given in our reference 6 (Wang et al.). The details are a bit convoluted, but by en large, for optimized time constants, the exponentially rising pulses lead to an about 16-20% higher excitation probability than the rectangular pulses for otherwise similar conditions (coherent state vs. Fock state, and full versus partial spatial overlap between atomic emission mode and excitation field).

Both point 1 and 2 raise very valid questions, though. We tried to address both of them by adding a paragraph just before the summary putting the impact on the excitation probability in a context without going through all the details in reference 6:

"We should state that the technique presented here is able to generate pulses with an exponentially rising shape out of a continuous coherent light beam. To transfer an atom from the ground state to an excited state with a single photon field, such a pulse shape is a necessary condition, but it also requires a full spatial overlap between excitation mode and atomic dipole mode, and the field being in a single photon Fock state. For a limited spatial overlap, and for an average photon number of 1 either in a Fock state or a coherent state, the exponential pulse shape leads to an about 16-20%  higher excitation probability than an optimized rectangular pulse [6]. If our technique were used to reshape a field in a Fock state, one needs to consider also the losses in the EOM modulation scheme, leading to a relatively low transmission. This suggests the development of alternative pulse shaping techniques."

For point 3, we added a trace indicating the fit to exponential rising function for the photodiode response, changed the line connecting the measured points to dots, and explained the fit function in the main text reference by replacing the text    
  "... shows a rise with a time constant of about 10 ns."
by 
"...shows a rise with a time constant of $\tau=10.86\pm0.02$\,ns (from a fit with $f(t)=a e^{t/\tau}$ for $t<0$, solid line in the lower trace in the figure)."

We tried to add the fit for the RF signal as well to the figure, but the ratio of oscillation period versus rise time constant are so large that it was impossible to see any difference in the fit and the measured function, so we left this trace as it was.

With this, we hope to have addressed the points raised by the reviewer, and look forward to a favourable consideration.

With best Regards on behalf of all authors,

Christian Kurtsiefer