The paper by Wiame et al presents a versatile approach to concurrently removing the fine structure of multiplets caused by homo- and/or heteronuclear coupling while achieving significant line-narrowing of NMR signals by eliminating the effects of inhomogeneous Bo. The paper is well-presented, with numerous examples, providing the NMR community with an excellent starting point for using this technique in a wide range of applications. As stated in the manuscript, "The data that support the findings of this study are available from the corresponding author upon request." I support the publication of this manuscript and like the authors to consider the points raised below.

1. The authors have provided typical parameters (pulse length, flip angle, power level, acquisition interval length, etc.), but they optimized these for different spectra (e.g., paragraphs 170 and 240). Could they provide some suggestions or rationale for choosing certain parameters? Additionally, what was the value of the after-pulse delay tau? The power deposition to the probe appears to be low; however, it would be desirable to put this into perspective using the technical specifications of current NMR probes.
2. Some spectra contain low-intensity peaks on each side of the collapsed singlet. Some comments on their origin would be useful. For example, in Figures 3 and 4, are these peaks present due to signal density caused by Bo inhomogeneity, but in others, such as Figures 5 and 8, perhaps not? The acquisition times used are short, hence the chunking artifacts are positioned much further away. Is it possible to link the presence of these peaks to some parameter of the pulse sequence?
3. The noise in the ³¹P SSL spectrum in Figure 11 is larger than in the reference spectrum. The stated peak height gains of approximately 3, 13, and 5 reflect the figure as presented, but perhaps not the SNR improvements, which should be considered when quantifying the signal enhancement.
4. Figure 9 demonstrates the use of polychromatic pulses, and it was interesting to see that the signals below 2.5 ppm preserved their multiplet structure, albeit with reduced intensities. The authors also demonstrated that applying SSL at frequencies 6.5 Hz apart is possible. It would be interesting to state how close ¹H multiplets can be for one of them to be collapsed by SSL, yielding a quantitative response without interference from the signal next door. Would the off-resonance signal still be a multiplet, preserving its structure?
5. In the latest iteration of SHARPER pulse sequences, the authors have moved away from using selective pulses during acquisition. Although functional (Uhrin and co-authors, https://doi.org/10.1039/d2cc01325h open_in_new), they unnecessarily shortened T_2eff, and a better outcome is achieved when the acquisition module consists of much shorter, hard pulses (Uhrin and co-authors, https://doi.org/10.1038/s41467-023-40130-2 open_in_new, https://doi.org/10.1021/jacs.5c11092 open_in_new). For the removal of J couplings only, typical chunk times (here AQ times) below 1 ms are used in combination with 90 or 180 deg pulses during the pulse/acquire period. It would be interesting to see how this approach compares to the SSL pulse in terms of the resulting signal linewidth, hence the signal hight and SNR.
6. Reflecting on the previous two points, applying SSL to a single frequency (provided that nearby signals are not affected) would present an advantage over SHARPER (even if some intensity drop is registered) as it would avoid the need for selective excitation.
7. In paragraph 250, the authors stated as a reason for narrowing ³¹P signals: "Surprisingly, the experimental line-widths were even narrower, since Δ1/2 = 0.7 Hz at all three sites, which may be due to the cancellation of broadening due to chemical exchange with Mg2+ ions." What is the mechanism of this narrowing? Chemical exchange usually broadens the signal. Can SSL remove exchange broadening?
8. In the concluding remarks, the authors list a number of areas with potential for the use of the SSL methodology, which is appreciated as it will accelerate the spread of this free-of-charge SNR enhancement methodology. As CPMG-type acquisition modules are widely used in applications where large Bo inhomogeneities need to be suppressed (e.g., the work of Prof Ville-Veikko Telkki and many others), it would be interesting to see if the SSL technique can also be applied in this field.
9. With regard to referencing prior work, since one of the applications of SSL presented in this paper is to LLS, work achieving similar outcomes in the LLS space could be mentioned. See Bodenhausen and co-authors, DOI: 10.1103/PhysRevLett.109.04760, and Levitt and co-authors, https://doi.org/10.1039/c2cp42553j open_in_new.
10. The approach taken by Andrew Simpson and co-workers (DREAMTIME, https://doi.org/10.1002/anie.202110044 open_in_new), a multi-focusing approach to increasing NMR sensitivity, should certainly be mentioned and discussed.

The paper by Wiame et al presents a versatile approach to concurrently removing the fine structure of multiplets caused by homo- and/or heteronuclear coupling while achieving significant line-narrowing of NMR signals by eliminating the effects of inhomogeneous Bo. The paper is well-presented, with numerous examples, providing the NMR community with an excellent starting point for using this technique in a wide range of applications. As stated in the manuscript, "The data that support the findings of this study are available from the corresponding author upon request." I support the publication of this manuscript and like the authors to consider the points raised below.

1. The authors have provided typical parameters (pulse length, flip angle, power level, acquisition interval length, etc.), but they optimized these for different spectra (e.g., paragraphs 170 and 240). Could they provide some suggestions or rationale for choosing certain parameters? Additionally, what was the value of the after-pulse delay tau? The power deposition to the probe appears to be low; however, it would be desirable to put this into perspective using the technical specifications of current NMR probes.
2. Some spectra contain low-intensity peaks on each side of the collapsed singlet. Some comments on their origin would be useful. For example, in Figures 3 and 4, are these peaks present due to signal density caused by Bo inhomogeneity, but in others, such as Figures 5 and 8, perhaps not? The acquisition times used are short, hence the chunking artifacts are positioned much further away. Is it possible to link the presence of these peaks to some parameter of the pulse sequence?
3. The noise in the ³¹P SSL spectrum in Figure 11 is larger than in the reference spectrum. The stated peak height gains of approximately 3, 13, and 5 reflect the figure as presented, but perhaps not the SNR improvements, which should be considered when quantifying the signal enhancement.
4. Figure 9 demonstrates the use of polychromatic pulses, and it was interesting to see that the signals below 2.5 ppm preserved their multiplet structure, albeit with reduced intensities. The authors also demonstrated that applying SSL at frequencies 6.5 Hz apart is possible. It would be interesting to state how close ¹H multiplets can be for one of them to be collapsed by SSL, yielding a quantitative response without interference from the signal next door. Would the off-resonance signal still be a multiplet, preserving its structure?
5. In the latest iteration of SHARPER pulse sequences, the authors have moved away from using selective pulses during acquisition. Although functional (Uhrin and co-authors, https://doi.org/10.1039/d2cc01325h open_in_new), they unnecessarily shortened T_2eff, and a better outcome is achieved when the acquisition module consists of much shorter, hard pulses (Uhrin and co-authors, https://doi.org/10.1038/s41467-023-40130-2 open_in_new, https://doi.org/10.1021/jacs.5c11092 open_in_new). For the removal of J couplings only, typical chunk times (here AQ times) below 1 ms are used in combination with 90 or 180 deg pulses during the pulse/acquire period. It would be interesting to see how this approach compares to the SSL pulse in terms of the resulting signal linewidth, hence the signal hight and SNR.
6. Reflecting on the previous two points, applying SSL to a single frequency (provided that nearby signals are not affected) would present an advantage over SHARPER (even if some intensity drop is registered) as it would avoid the need for selective excitation.
7. In paragraph 250, the authors stated as a reason for narrowing ³¹P signals: "Surprisingly, the experimental line-widths were even narrower, since Δ1/2 = 0.7 Hz at all three sites, which may be due to the cancellation of broadening due to chemical exchange with Mg2+ ions." What is the mechanism of this narrowing? Chemical exchange usually broadens the signal. Can SSL remove exchange broadening?
8. In the concluding remarks, the authors list a number of areas with potential for the use of the SSL methodology, which is appreciated as it will accelerate the spread of this free-of-charge SNR enhancement methodology. As CPMG-type acquisition modules are widely used in applications where large Bo inhomogeneities need to be suppressed (e.g., the work of Prof Ville-Veikko Telkki and many others), it would be interesting to see if the SSL technique can also be applied in this field.
9. With regard to referencing prior work, since one of the applications of SSL presented in this paper is to LLS, work achieving similar outcomes in the LLS space could be mentioned. See Bodenhausen and co-authors, DOI: 10.1103/PhysRevLett.109.04760, and Levitt and co-authors, https://doi.org/10.1039/c2cp42553j open_in_new.
10. The approach taken by Andrew Simpson and co-workers (DREAMTIME, https://doi.org/10.1002/anie.202110044 open_in_new), a multi-focusing approach to increasing NMR sensitivity, should certainly be mentioned and discussed.

This is an interesting ms with some nice results, well worth publishing, but it would benefit from some editing and from a more considered discussion and analysis.

Comments and corrections, trivial and otherwise, are listed below in order of page number.

3               delete extra brackets following INFERNO

4               The introduction to SSL lacks clarity and depth. It should be made clear in Fig. 1 that delta t is 1/sw, and that the period labelled “Acq” contributes just one complex point to the final FID; innocent readers will expect “Acq” to denote acquisition of an entire FID. Throughout the ms the description of experimental methods is missing small but critical pieces of information, e.g. the duration of the delay tau, whether “points” means complex points or real points, etc. I strongly encourage the authors to place their experimental data in a public repository so that interested readers can replicate their work, or at the least fill in some of the gaps left by the ms.

                  Explanations are needed for the orders of magnitude chosen (e.g. pulse durations taup are limited by spectrometer RF bandwidths, flip angles need to be low enough to elicit linear responses, RF amplitudes are determined by beta and taup). It is unhelpful to describe the sequence of M short pulses as ‘a “comb” in the manner of DANTE’; it is the M sets of N pulses that are such combs. It would be helpful to the reader to explain that sequence (a) is a DANTE selective irradiation on resonance; adding further phase-rolled pulses in (b) amounts to running a series of different DANTE sequences in parallel. It is not correct to describe the M pulses as “a superposition”: that would place them one on top of another, they are a sequence or a succession not a superposition.

                  A significant problem with both the exposition and the analysis of polychromatic SSL is that the Bloch-Siegert effect appears to have been totally ignored. This is a major complication when RF irradiation is, as here, carried out at multiple frequencies simultaneously. The practical consequence here is that the effective DANTE irradiation frequencies produced will differ materially from those calculated using eq. (1). Packages such as Wavemaker routinely correct for such effects.

                  A second significant complication that needs addressing is that modern NMR instruments do not acquire a single complex data point per dwell time; rather, digital signal processing is used to distil such single points from a much more rapid stream of data. Where that stream is, as here and in any “windowed” acquisition, interrupted, complications and distortions result that are an inevitable consequence of the loss of receiver data continuity while RF pulses are applied.

                  Throughout the ms a range of different prescriptions are given for the choice of spin lock field amplitude. A clear explanation is needed of how to make rational decisions about this: too low an amplitude will degrade spin locking, too high will lead to spectra distorted by the B-S effect (v supra). Advice to use an RF amplitude comparable to J is unhelpful: many multiplets have widths much greater than any individual coupling, and using only a 10 Hz field with a 50 Hz wide multiplet will degrade the results obtainable.

5               an N repetition -> N repetitions

5               It is a little misleading to say that “the novelty of SSL lies in the acquisition of data points in the intervals between the pulses”. Windowed acquisition has been used for 60 years or so (e.g. in WAHUHA, and in multiple papers from the Freeman group), and is still available in some routine liquid state spectrometers for selective homonuclear decoupling.

6               1 Hz FWHH is about twice the typical manufacturer’s specification, so “careful shimming” is a bit of an exaggeration.

6               If the natural abundance of 29Si is < 5% then the decoupling referred to cannot explain a 20% increase in peak integral.

6               Here and throughout, the discussion of the linewidths achievable is a bit confused. Under the experimental conditions reported, the experimental linewidth contains a substantial contribution from the limited FID duration. The spins neither know nor care if a receiver is switched on during a DANTE sequence, so the T1rho measured during a windowed sequence should be exactly the same as that in the corresponding unwindowed sequence (see e.g. the comment at the foot of p. 9).

7               miss-  -> mis-

                  typically -> often [not all methylenes are in alkyl chains]

8               both experiments -> experiments both

9,11       Figs. 5 and 7 are missing basic information – “8 nonlinear steps” is not helpful!

11            Applying Lorentzian broadening does not “improve the digital resolution” in the accepted sense of the term. Why were the data not zero-filled, rather than degraded in this way?

12            The squiggly arrow in Fig. 8 and subsequent figures needs explanation (O1?)

12            The interesting observation that the proposed method can fail for the simplest possible coupled spin system, AX, needs both comment and explanation. Under what conditions does the breakdown occur? How can this be avoided?

13            Lines narrowed -> Line-narrowed

13            The disappointingly messy results in Fig.9, where artefacts exceed some of the real signals in amplitude, need comment. Why are there so many spurious satellites (B-S effects? DSP artefacts?)? Why are they predominantly in negative absorption mode?

15            The suggestion that SSL cancels broadening due to Mg2+ exchange is implausible: it would imply a shift change on binding that is small compared to the spin lock field amplitude.

17            The comparison with pure shift methods is more favourable than stated: in pure shift NMR the resolution limit is 1/(pi T2 star) not 1/(pi T2), and in the presence of a spin-lock field scalar contributions to T2 should be reduced, boosting the SSL signal still further.

17            Authors contributions -> Author contributions or Authors’ contributions