---------------------------------------------------------------------- Report of Referee A -- LQ17413/Seidler ---------------------------------------------------------------------- The authors report on a spectral compression scheme for heralded single photons at 795nm using a time-lens based on an asymmetric cavity and a binary phase modulator. The heralded single photons are generated via four-wave mixing in a cloud of cold 87Rb atoms. They observe a spectral compression from 20.6(2)MHz to 8(2)MHz, which is close to the target 6MHz bandwidth to match the natural D transition linewidth in 87Rb. The spectral compression of the single photons is based on the temporal envelope manipulation of the wavepacket by first using an asymmetric cavity to disperse the photon wavepacket in time, and then applying a time-dependent phase shift with an EOM to change the spectral energy distribution. This work appears to be the first to report on the use of an optical cavity as the dispersive media, which has the advantage of (theoretically) being very low loss compared to other media like optical fibers. So this time-lens technique is discussed as a potentially lossless way to spectrally shape photons that will maintain high rate. The topic of spectral shaping photons to better interact with atomic transitions is current and novel, making it suitable for the PRL audience. This work is novel and deserves publication. However, it is not clear if these results are a significant improvement and wide-enough impact over previously reported techniques to warrant publication in PRL. If the main advantage is less loss, the only loss-related terms quoted from the experiment are quite high and probably don’t reflect the power of this technique. At this stage, I cannot recommend this paper for publication in PRL. Comments: - Could the authors elaborate more on the advantages of using an asymmetric cavity as the dispersive media as opposed to the other conventional methods? If the main difference/improvement in the author’s technique over currently published time-lens schemes is their choice of an asymmetric cavity as the dispersive media, then how much of dramatic of an improvement did this change make compared to previous results? Answer: The dispersion provided by a fiber is not enough to spread the time envelope of a norrowband photon significantly. The required fiber length is unfeasible and would lead to complete extinction of the photon due to losses. 132 ps / (nm km) (Addressed in paper edit; lines 98-116) - Are the authors aware of published work demonstrating a spectral compression down to a narrow linewidth like 8MHz? The references cited by the authors and other papers I could find seem to start with single photons with a much larger bandwidth (GHz) have compressed hundreds of GHz to tens of GHz, usually achieving a spectral compression factor of near 10. While this work does not achieve as high of a compression factor, it takes a much smaller starting bandwidth (20MHz) to 8(2)MHz, to make it closer to the desired 6MHz atomic transition. Is this the first report of compression near an atomic bandwidth? Response: PHYSICAL REVIEW X2, 021011 (2012) demonstrates a spectral compression of narrowband coherent pulses using a gradient echo memory. However, the spectral bandwidth they are operating on are below a few MHz and is not as adaptable to different ranges of bandwidths as our method. We added this reference in the introduction. (Addressed in paper edit; new citation added Sparkes2012 in line 73) - Does the asymmetric cavity have to stay phase locked with the EOM signal? If so, how is this done in the experiment? We do not require a phase lock. Instead, the cavity resonanace frequency is locked to the center frequency of the photon spectrum, such that it causes dispersion to the transitting photon. - The authors wrote they attribute the dip in the spectrum of the uncompressed photons to the reabsorption of the generated photons by the atomic cloud. However, later they write that the extracted linewidth of the dip does not match the 5S1/2 to 5P1/2 transition and further work needs to be done. How significant is this feature to the effectiveness of their time-lens method? If it just an artifact from the way they are generating the single photons, or could the source of this feature be related to the technique? Response: While even a high-reflective mirror will leak a small amount of light, it does not cause the observed dip in our experiment. The dip is also observed without the compression optics and is a charateristic of the photon pair source. (addressed in edit; line 340) - Concerning the phase flip being applied right after the first part of the dispersed photon exits the modulator and the second part starts to propagate through it. How did you measure/calculate/confirm this? Is it based on the length of the EOM and time for light to propagate from one end of the modulator to the other? How crucial is this timing of only having one photon in the modulator at a time? What is the single pair vs double pair rates? Response: The photon intensity envelope is roughly 80ns long, which corresponds to a distance of about 24m. The active length of the EOM is about 90mm long, which corresponds to a timing uncertainty of 0.3ns. At one time, only a small part of the photon resides inside the EOM and we can apply different phase shifts to different parts of the photon. General comments: - the uncompressed bandwidth is quoted as 20.6(2)MHz in some locations and 20(2)MHz in other places in the manuscript. Which is correct? Response: - The 20.6(2) MHz is the bandwidth inferred from the temporal profile, which was used as part of fitting parameters. 20(2)MHz is the bandwidth observed from the spectroscopy using the 2.6 MHz cavity. The distinction between the inferred and observed bandwidth has been made clearer in the edit. - p.g.1: spelling error “makeing” on pg 1, 2nd paragraph (corrected) - Fig 4: label black line too… difficult to see (corrected to new) - pg 4: Fig reference says “Fig. setup” instead of Fig 2 (found to be not necessary and was deleted) ---------------------------------------------------------------------- Report of Referee B -- LQ17413/Seidler ---------------------------------------------------------------------- This manuscript reports the compression of the bandwidth of wavepackets of single photons in the optical domain. This work is motivated by the fact that single-photon sources often have bandwidths that are larger than those of atomic quantum memories, so that an efficient storage of single photons is not possible. The present manuscript uses ‘”time lens” methods that were previously invented for ultrashort pulses, but adapts them to light that is already fairly narrowband. The bandwidth of a single-photon pulse is narrowed in two steps: The first step consists of an interaction with a cavity, which changes the temporal shape but leaves the spectral density unchanged, and the second step is the imprinting of a temporal phase (with an electro-optical modulator) that results in narrowing of the bandwidth. The first step is necessary to create a wider temporal shape that is consistent with the desired narrower bandwidth. The paper is well written with an introduction that is adequate for a general audience. The structure of the paper is clear and logical. The experimental results appear valid, and are in general agreement with the expectations from a theoretical prediction. While the bandwidth reduction is moderate (a factor 2.5) and accompanied by substantial loss (the overall transmission of the system is 21%), the authors argue that the loss could be improved upon. Response: - this was further elaborated in the edit. (lines 393-411) Regarding the PRL criteria of impact, innovation, and general interest, this paper could potentially qualify for PRL on the basis of impact and innovation. With respect to impact, however, one has to ask the question whether the current method, given its loss, is any better (in terms of spectral brightness) than a simple cavity filtering method when the loss of the new method is included. This is currently not the case (the brightness before loss increases by a factor 2-2.5, see Fig. 4, whereas the loss is a factor 5). To argue that this work can make an impact on the field, I would have at least liked to see a discussion in the manuscript whether this method can realistically provide higher brightness than spectral filtering, and by what realistic factor. If that factor in one or less, then the impact of this method will be quite limited, as other groups will simply continue to filter with a cavity. Response: - comparison to realistic improvement in losses and passive filtering were made in the edit. (lines 393-411) The manuscript might also qualify for PRL on the basis of its innovative aspect. As someone working in the general field, I found the paper interesting to read, and while the general method (spread wavepacket out in time, then imprint temporal phase) is not new, the way it is applied to single-photon wavepackets is new, and may inspire other ideas and experiments. For the general readership, it may be interesting to learn that it is possible to actively shape single-photon wavepackets, and in particular narrow their bandwidth. Overall, my recommendation (with moderate conviction) is publication of this article in PRL on the basis of its innovative aspect. My recommendation would be stronger if the authors could argue that a gain in spectral brightness over the much simpler method of passive spectral filtering with a cavity is realistic. Finally, I have a more technical remark: The authors mention that they don’t understand an absorption profile (Fig. 4) with a linewidth below the natural linewidth. I assume that the authors have considered a Raman absorption process between magnetic sublevels that could exhibit a smaller linewidth? Response: - We expect the splitting of magnetic sublevels in a field gradient to result in a inhomogenously broadening in the spectral profile, in which a Raman absorption process would exhibit a broadened absorption linewidth instead.