We thank the reviewer for his detailed reading of our manuscript and for this comment, which is absolutely right. 

Indeed, in the case of dendritic lithium structures, the EPR spectrum is expected very sharp (around 0.05mT and less) as reported in a variety of EPR investigations. But also, the EPR spectrum should exhibit a pure Lorentzian line shape unlike the micrometric lithium aggregates exhibit a Dysonian shape. This effect on the line shape is due to the skin depth (δ_mw = 1.1 µm for Li⁰ at X-band) which limits the microwave penetration inside the conductor (i.e. through the metal thickness d). If d>δ_mw the EPR line exhibits an asymmetric peak (Dysonian with A/B > 1) whereas if d<δ_mw, the EPR line is a symmetric and intense (Lorentzian with A/B=1).

Initially, we performed EPR measurements using an amplitude modulation of 0.2mT to follow the EPR signature of the lithiated graphite which gives a linewidth around 0.4mT. At this step of investigations, we did not expect to observe a metallic lithium EPR signature. But after analysis using two Dysonian functions, we performed the second harmonic investigation to confirm the presence and the origin of the other hidden contribution which is assimilated to micrometric lithium structures.

In our case, we did not detect dendritic structures and the reason is twofold. On one hand, no sudden drop to 0V characterizing a short-circuit was detected during electrochemical cycling. On the other hand, no symmetric EPR spectrum was observed as expected for sub-micrometric structures like dendrites.

We detected micrometric lithium particles with a size much higher than 1.5 µm which is much larger than the skin depth. As a consequence, the linewidth of such particles is much larger than 0.05mT. Even if the EPR signal of the metal lithium aggregates, recorded using the first harmonic detection scheme is possibly over-modulated, the second harmonic mode does not show an excessive over-modulation for the EPR peak. In contrary, recorded a slightly over-modulated EPR line using the second harmonic mode increase the signal-to-noise ratio S/N.

As suggested by the reviewer, in the case of dendritic structures, an amplitude modulation less than 0.05mT will be necessary to get semi-quantitative information about such particles.

We thank the reviewer for his reading of the manuscript, which allows us to address some clarifications. 

Comment1: Thank you for this comment. Metallic Li⁰ species exhibit always a homogeneous EPR spectrum for which the EPR linewidth is mainly dominated by the spin-spin relaxation time T₂. It is possible to improve the spectral selectivity by playing with the relaxation (and then with the linewidth) of the detected signal. The second harmonic mode is sensitive to T₂, and usually a relatively broad signal (short T₂) gives an EPR signal using this detection mode. However, such signal appears flat in comparison to an EPR signal with a long T₂.

Two bibliographic references have been added (on page 6, line 121-122, in red) concerning the selectivity of the second harmonic detection mode.

1. Pali et al, J Magn Reson B. (1996), 113, 151-159
2. Marsh et al, J. Chem. Soc., Perkin Trans. 2, (1997), 2545-2548

Comment 2: Yes, it is possible to distinguish mossy and dendritic lithium structures using the second harmonic detection scheme. Indeed, the EPR spectrum of Li⁰ is indicative to the variety of morphologies. In the case of thick Li⁰, the EPR spectrum exhibits an asymmetric lineshape, also name Dysonian, with a linewidth around 0.15mT. In the case of mossy and dendritic lithium structures, the EPR signal is slightly Dysonian and pure Lorentizan respectively. The second harmonic EPR spectrum of mossy will be slightly asymmetric whereas the corresponding EPR signal of dendritic Li⁰ will be symmetric. Furthermore, mossy Li⁰ deposited during cycling exhibits a linewidth around 0.03mT whereas dendritic Li⁰ shows a linewidth around 0.005mT as reported by Niemöller et al (Sci. Rep. 8,1–7 (2018)). Consequently, dendritic Li⁰ will be much more amplify with the second harmonic mode than the mossy Li⁰ signal.

Furthermore, the new ELEXSYS II integrate today the new features for Signal Processing Unit (SPU) with Digital lock-in with up to 5 simultaneously detected harmonics with simultaneous detection of 0 and 90-degree modulation phases. In the case of multi-component with a similar g-factor, we believe that using multi-harmonic detection mode will be an invaluable method to clearly distinguish each contribution.

Comment 3: We only simulated the EPR signal from a theoretical model using the sum of two Lorentzian phase shifted functions defined in equation (1) in our manuscript. EasySpin is only used for loading data in MATLAB.

We thank the reviewer for this very positive appraisal of the work. As suggested by the reviewer and the reviewer #2, we added two bibliographic references about the second harmonic features (on page 6, line 121-122).

Concerning the A/B ratio, we thank the reviewer for his question. The variation of this ratio depends on the different Li⁰ morphologies (micro-aggregates, mossy or dendritic structures). Usually, thick Li⁰ exhibits an asymmetric lineshape (1^st harmonic - Dysonian), mossy Li⁰ a slightly asymmetric peak and dendritic Li⁰ a pure Lorentzian line. Consequently, the second harmonic EPR signal of Li⁰ micro-aggregates will exhibit an asymmetric lineshape while dendritic Li⁰ structures will show a symmetric EPR spectrum, with an apparent intensity higher than micro-aggregates (S/N ratio higher). An example of such second harmonic lineshapes is given in following Figure.