Voltage-driven motion of nitrogen ions: a new paradigm for magneto-ionics

Magneto-ionics, understood as voltage-driven ion transport in magnetic materials, has largely relied on controlled migration of oxygen ions. Here, we demonstrate room-temperature voltage-driven nitrogen transport (i.e., nitrogen magneto-ionics) by electrolyte-gating of a CoN film. Nitrogen magneto-ionics in CoN is compared to oxygen magneto-ionics in Co3O4. Both materials are nanocrystalline (face-centered cubic structure) and show reversible voltage-driven ON-OFF ferromagnetism. In contrast to oxygen, nitrogen transport occurs uniformly creating a plane-wave-like migration front, without assistance of diffusion channels. Remarkably, nitrogen magneto-ionics requires lower threshold voltages and exhibits enhanced rates and cyclability. This is due to the lower activation energy for ion diffusion and the lower electronegativity of nitrogen compared to oxygen. These results may open new avenues in applications such as brain-inspired computing or iontronics in general.

(1) Redox potential measurements of the electrolytic cell should be conducted. In this work, the authors have applied a voltage as high as −50V to the cell. Such high voltage should induce a variety of electrochemical reaction not only in magneto-ionic target (CoN and Co<sub>3</sub>O<sub>4</sub> film) but also in the components of the used electrolytic cell; redox reaction of liquid electrolyte and Cu/Ti-metal film. At least, the redox potentials of CoN and Co<sub>3</sub>O<sub>4</sub> should be checked by measuring cyclic voltammetry. They should directly correlate with the responsiveness of magnetization to the applied voltages.
(2) Does the Cu-layer not react with migrated oxygen or nitrogen ions? The working electrode of electrolytic cell consists of Cu-metal. I am afraid that the Cu-layer might be oxidized under the high voltage and the oxidized Cu might react with the migrated oxygen or nitrogen ions. In such case, the formation of copper oxides and nitrides would affect the responsiveness of magnetization. Did the authors confirm the stability of the copper layer during voltage sweeping process? In case that oxidation of Cu occurs, the oxidation current would be detected during voltage sweeping process.
(3) How did the authors identify the gases produced under the gating voltage of −50 V as O<sub>2</sub> and N<sub>2</sub>? In Page 5, the occurrence of bubbling under the gating voltage of −50 V is reported for both Co<sub>3</sub>O<sub>4</sub> and CoN. The authors identify the generating gases as O<sub>2</sub> and N<sub>2</sub> from Co<sub>3</sub>O<sub>4</sub> and CoN, respectively. However, in such voltage as high as −50 V, solvent of electrolyte (propylene carbonate) might be electrochemically decomposed to generate gases. Was the electrochemical window of electrolyte confirmed? The author should state how they distinguish the evolution of O<sub>2</sub>/N<sub>2</sub> gas from that of other possible origins.
(4) Are the Na<sup>+</sup>-ions in liquid electrolyte inserted into the CoN/Co<sub>3</sub>O<sub>4</sub> film or not during voltage sweeping process? The liquid electrolyte used in this work contains Na<sup>+</sup>-ions. In the recent years, the Na+-ion is intensively investigated as charge carrier species in ion-batteries. In the present case, can the authors exclude the possibility Na<sup>+</sup>-ion insertion into CoN/Co<sub>3</sub>O<sub>4</sub>? In the case that the Na<sup>+</sup>-ion insertion occurs, Na<sup>+</sup>-ion insertion would contribute to redox reaction in CoN and Co<sub>3</sub>O<sub>4</sub>, and thus it would affect the responsiveness of magnetization to the applied voltages.
(5) In the Ab initio calculations, were the crystal structures of Co<sub>3</sub>O<sub>4</sub> and CoN reproduced in the lattice relaxation process after oxygen or nitrogen insertion? For comparison of the formation energy of Co-O and Co-N, the calculation for the oxygen and nitrogen insertion to the hexagonal closed-packed Co slab has been conducted (In Page 19 (Methods, Ab initio calculations). However, the sublattice-structures of Co in Co<sub>3</sub>O<sub>4</sub> and CoN are different from the hexagonal closed-packed structure. In the relaxation process of the atomic coordinates, were the crystal structures of Co<sub>3</sub>O<sub>4</sub> and CoN reproduced? If not, is it plausible to use the calculated values of energy barriers for explanation of the observed superior electrical magnetismcontrollability of CoN?
(6) The definition of the sign of the applied voltage should be clearly stated.
In page 17 (Methods, magnetoelectric characterization), it is stated that the voltage was applied in the same fashion to that presented in references 7, 8 and 27. However, the direction of the battery connection of the external power supply is opposite between Fig.1 of this paper and the schematic figure in Ref.7. In Fig. 1, the positive electrode of the battery is connected to the thin film sample, whereas the negative electrode of the battery is connected in Ref.7. It is very confusing for readers. The author should clearly define the sign of the applied voltage in the experimental methods, rather than simply citing the literature.

1/17
In the following, we provide detailed answers to each of the reviewers' comments (written in blue and italic font). The added and modified parts appear in the revised version of the manuscript over a yellow background:

Reviewer #1
In the manuscript "Voltage-driven motion of nitrogen ions: a new paradigm for magneto-ionics" the authors set out to demonstrate a new type of a magneto-ionic (magnetoelectric) system driven by Nitrogen ions rather than conventionally used (most studied) Oxygen. The authors argue and ultimately prove that voltage-controlled nitrogen mediated redox reactions -bringing about the magnetelectric coupling -require lower threshold voltages and exhibit enhanced rates and cyclability.
To elucidate the phenomena behind the magnetization switching and phase transitions the authors have chosen a wide-ranging and well-thought set of characterization techniques. They inquired in detail into the structure, microstructure, elemental mapping, magnetic characteristics versus applied voltage etc. The conclusions reached are sound and sufficiently supported by the experimental data. The manuscript is clearly structured with the appropriate choice of references.
In particular, I find very appealing the combination of the semiconductor (or weakly metallic) material, undoubtedly resulting in the smooth and uniform redox process during the cycling, with the smart choice of the active ion.
In summary, this is a very interesting study; it should be published after addressing some remaining concerns pointed out below.
We do appreciate that Reviewer #1 considers our work very interesting and publishable after minor revision.
Some specific points recommended for considering: 1) As the authors concluded the full chemical reduction of the CoN film shown in the report (i.e. conversion from CoN to Co metal; for example at -50 V) is irreversible; on the other hand reversibility is demonstrated for the small voltages and shorter charging times (Fig. 2). I think it would be beneficial to discuss shortly -or in other words -an estimate of the actual thickness of the CoN participating actively in the reversible cycling. For example, let's make a very simple (may be too simple but worth trying anyway) approximation and take the experimental change of magnetization of 1emu/cm3 (as in Fig 2.) and compare it with the total Co(N) magnetization of 637 emu/cm 3 after full reduction (Table 1). Now taking 85 nm of the CoN film thickness one can estimate roughly the thickness of the magnetically active layer -(1/637) *85nm = 0.13nm. Does that mean that in the reversible cycling only a few top monolayers of the CoN are involved? How deep one can actually go with the CoN reduction to retain a full reversibility? Perhaps it would be instructive for the reader to shortly address those points since at the end the strength of the magnetic response to the electric stimulus (or strength of the magnetoelectric coupling) depends on those conditions. Reviewer #1 raises several interesting points regarding voltage cyclability in our system. His/her simple but representative approximation to estimate the magnetically active region indeed indicates that only a few monolayers are involved in the generation of a magnetization of 1 emu/cm 3 . This actually corresponds to the magnetic moment induced close to the threshold voltage (where magneto-ionic effects are triggered) if this voltage is applied for a relatively short time (a few minutes). As supported by the HAADF-STEM/EELS and PALS characterization, N migration begins at 2/17 the upper part of the CoN layer, i.e., at the boundary between the working electrode (the sample) and the electrolyte (which acts as N reservoir). With time, N ions from underneath regions become also activated. For example, Fig. 1g reveals that a magnetization of around 18 emu/cm 3 can be generated in CoN if the threshold voltage is applied for about 1 hour. This corresponds to approximately 2.5 nm and the behavior is still reversible. In any case, reversibility is constrained mainly due to the limited solubility of nitrogen in propylene carbonate (see, e.g., Chen, W. et al., Appl. Energy 240, 265-275 (2019)). An analogous reasoning can be made for oxygen from Co 3 O 4 due to its limited solubility in propylene carbonate (see, e.g., Read, J. et al., J. Electrochem. Soc. 150, A1351−A1356 (2003)). These low solubility limits cause N 2 (or O 2 ) bubbling to appear once propylene carbonate reaches nitrogen (or oxygen) supersaturation. The N 2 (or O 2 ) gases forming bubbles are released to the atmosphere and cannot be recovered. It might be that, with improved ion reservoirs (for instance, with solid state ionic conductors), reversibility could be enhanced in the future.
The following paragraphs and references have been added in the Results section (pages 5 and 9): "For both systems, when gating at -50 V, bubbling occurs. Oxygen (in Co 3 O 4 ) and nitrogen (in CoN) gas evolution is most likely the major source of bubbling. When a large negative voltage is applied, high concentrations of oxygen and nitrogen anions accumulate near the counter-electrode. If their concentration is high enough, bubbles form when the oxygen (or nitrogen) solubility limits 36,37 are exceeded. When this happens, these elements are irreversibly lost from the system (sample + electrolyte) and this is the main reason causing the lack of reversibility of magneto-ionic effects when positive voltage is applied." "As a first approach, a simple calculation considering the overall magnetization induced in CoN when applying -50 V (i.e., 637 emu/cm 3 , Table 1, when the whole CoN is affected, as will be shown in the forthcoming section) suggests that during this reversible cycling (1 emu/cm 3 , Fig. 2), only the uppermost 0.1-0.2 nm of the CoN layer are involved in the observed reversible effect. If the threshold negative voltage is applied for longer times (Fig. 1g), the induced magnetization results from nitrogen ion migration of an equivalent CoN thickness of 2-3 nm. Similar affected thicknesses during reversible voltage-driven magnetization cycling can be estimated for Co 3 O 4 ." 2) In Fig. 1f,g the reaction threshold voltages are estimated and applied for 1 hour (chemical reduction). Could you please comment on the fact that in one hour Co 3 O 4 reaches the magnetization of 200 emu/cm 3 while CoN shows only 15 emu/cm 3 . How this could be reconciled with the generally faster magnetic response of the CoN over Co 3 O 4 (Fig. 1d).
We would like to thank the reviewer for pointing out this issue. In fact, the determination of the exact threshold voltage is not so straightforward due to the interplay between voltage and time. Actually, after looking at Fig. 1f from the original version of the manuscript, it seemed that the magnetic signal was already increasing after applying -6 V. This observation and, particularly, the reviewer's comment made us repeat the same experiment for Co 3 O 4 but using -6 V instead of -8 V.

3/17
The result is shown in the new version of Fig. 1f and the corresponding hysteresis loops in the new version of Fig. S3. The new measurements reveal that after a better estimation of the threshold voltage for Co 3 O 4 (-6 V instead of -8 V), both layers attain similar saturation magnetization values after the voltage is applied for sufficiently long times. Remarkably, comparison of the new Fig. 1f with Fig. 1g further corroborates that magneto-ionic rates are faster for CoN than for Co 3 O 4 . We thank the reviewer for making us realize about this apparent contradiction.

3) A question out of curiosity: When all the CoN film is fully converted to Co metal, why is the measured Co magnetization after -50 V treatment (Table 1 Ms of 637) -less than half of the Co bulk value (bulk Ms of about 1400 emu/cm 3 )? Perhaps the authors could shortly comment on it in the manuscript.
We also noticed this reduction of M S . We believe that when nitrogen and oxygen diffuse towards the electrolyte, a large amount of vacancies and other structural defects are created in the films. In fact, the film thickness does not considerably reduce when negative voltage is applied, which means that the treated films (after applying negative voltage) have a lower density and possibly nanoporosity (due to coalescence of vacancies). Since it is not easy to quantify the nanoporosity degree or the amount of vacancies or regions with lower density, the estimated values of saturation magnetization are given after normalizing to the total volume of the film (i.e., no correction is applied to account for porosity or the aforementioned structural defects). By doing so, the net volume of ferromagnetic phase is probably overestimated, and this results in magnetization values which are lower than for bulk, fully dense, cobalt.
A detailed analysis of the types of vacancies generated during electrolyte gating of Co 3 O 4 films was reported in our previous publication (ACS Nano 12, 10291-10300 (2018)). Complex vacancies, involving groups of various atoms, are formed during magneto-ionic experiments. The resulting Co is highly nanocrystalline or even amorphous (particularly the uppermost portions of the films in direct contact with the electrolyte). Indeed, a decrease of density (decrease of Co signal) is obvious in the TEM image of the CoN sublayer 1 in Fig. 3g, possibly due to the occurrence of free volume or nanoporosity. Note that a reduction of M S has, in fact, been reported in amorphous Co nanoparticles compared to bulk fully dense Co (see for example Langmuir 25, 10209-10217 (2009)).

5/17
To make this point clearer, we have now stated in the manuscript (Table 1, caption, page 8) that the M S values are obtained by normalizing by the nominal film thickness, without taking into consideration eventual nanoporosity or vacancies generated during magneto-ionic treatments.
"Note that M S has been obtained by normalizing the magnetic moment to the nominal film thickness, without taking into consideration the formation of vacancies or eventual nanoporosity." 4) Why actually CoN is claimed to be semiconducting here (page 26) ? " Fig. S4) reveal their insulating and semiconductor character, respectively. This allows the layers to hold electric fields across them, which is a necessary condition for magneto-ionic phenomena." Fig. S4 actually suggests a metallic like electronic transport down to 100 K. The CoN resistivity at room temperature is low -4*10 -4 Ohm cm (for instance metallic alloys nichrome is of 10 -4 ohm cm). If the CoN is metallic-like at room temperature that would mean a creation of the Helmholtz double layer at the CoN interface (no electric field penetration into the bulk of the CoN). Perhaps this would explain why -contrary to Co 3 O 4 case -the redox reaction front moves into the material in a uniform way; this would mean a conversion reaction (CoN to Co) front moving with the Helmholtz double layer into the material interior; please address this issue and maybe consider discussing it in the manuscript.
The reviewer raises an interesting point concerning the electric properties of CoN. He/she is right that Fig. S4 suggests a metallic-like electronic transport down to 100 K (with positive dρ/dT). Below this temperature, the sign of the dρ/dT slope becomes negative, evidencing a complex electric transport behavior. We performed resistivity measurements on sputtered CoN layers grown at different N 2 pressures and the overall temperature dependence of resistivity is found to depend on the N content in the samples (see figure below). While only a negative slope is observed for the 75% N 2 film, a positive slope is measured in the whole temperature range for the film grown with 25% N 2 . In any case, the resistivity values at room temperature are higher than the values in metals such as Cu, Fe or Pt (which are of the order of 10 -6 -10 -5 Ω cm) and lower than values obtained in wide bandgap semiconductors. Nonetheless, Co-rich CoN films are often claimed to be metallic in the literature (e.g., Inorg. Chem. Front. 3, 236-242 (2016); Chem. Mater. 30, 5941-5950 (2018)).

6/17
To take this complex behavior into account, we have now replaced "semiconducting behavior" by "semiconducting/metallic behavior" throughout the manuscript when referring to the electric properties of CoN.
Importantly, as pointed out by the reviewer, the electric properties of the investigated materials are likely to play a significant role on the observed dissimilar magneto-ionic phenomena between the two investigated films. While electric field in metals is effectively screened at their outmost surface (electrostatic shielding), Co oxides are good insulators, and we believe that ionic movement in Co 3 O 4 proceeds through dielectric breakdown (local ionization), due to the strong covalent character of the bonding. In turn, CoN has orders of magnitude smaller resistivities compared to Co 3 O 4 , and the number of factors to consider (electric field screening, chemical bonding) are larger. Nonetheless, the metallic-like character of CoN at room temperature might indeed explain why, contrary to the Co 3 O 4 case, the redox reaction front in this case moves into the material in a uniform way (plane-wave-like migration front). We believe a clear description of these effects requires a more detailed investigation, which we will certainly undertake in the future. To address the reviewer's observation, the following sentence has been added in the Discussion of the new version of the manuscript (page 18): "The dissimilar electric properties of CoN and Co 3 O 4 are also likely to play a role in the way ions diffuse in the two layers."

Reviewer #2
This paper reports electrical control of magnetism by voltage-driven ion transport (magneto-ionic control) for two-types of cobalt compounds; CoN and Co3O4. The authors control transport of nitrogen ion or oxygen ion by electrolyte-gating method in a CoN-and Co3O4-films, respectively. They have found reversible voltage-driven magnetic switching behavior between paramagnetic and ferromagnetic states at room temperature. By comparing the responsiveness of magnetization to gating voltage for CoN with that of Co3O4, they have also found lower operating voltage and the enhanced cyclability in CoN. The superior electrical magnetism-controllability of CoN is clearly demonstrated, and the results would be promising for the future application of magneto-ionics in devices. However, there are several comments and questionable points, which are shown below, on this paper in its present form. In particular, the electrochemical assessment of the electrolytic cell used for the measurements seems to be insufficient. In my opinion, they should be made clear before the manuscript is suitable for publication.
We thank the reviewer for his/her positive feedback, highlighting the superior responsiveness of CoN compared to Co 3 O 4 and considering that the results of our work are promising for future studies and applications of magneto-ionics. Following the reviewer's suggestion, the components of the electrochemical cell have been subjected to a series of cycling voltammetry experiments using propylene carbonate in order to test the existence of eventual electrochemical reactions. First, the potential window of the propylene carbonate electrolyte was confirmed using a Pt working electrode in the largest possible voltage window allowed by our instrumentation: -40/40 V (please note that to record the cyclic voltammetry curves a power supply different from the one used to run magneto-electric measurements was employed). As it can be seen from the figure below ( Figure S6 in the revised version of manuscript), the lack of oxidation/reduction peaks and the small currents involved (with currents densities of the order of µA/cm 2 ) indicate the absence of charge transfer reactions at the surface of the Pt working electrode. Importantly, no supporting electrolyte was used either here or during our magneto-ionic experiments. Therefore, the analyzed electrolyte is a priori resistive to charge transfer, contrary to aqueous media, Li-containing electrolytes or ionic liquids widely used in other magnetoelectric studies (e.g. Small 15, 1904523 (2019); Nano Lett. 16, 583−587 (2016); or Adv. Mater. 30, 1703908 (2018); etc.). It should be noted, nevertheless, that we have sometimes encountered polymerization effects in the working electrode when using propylene carbonate at even larger voltages (e.g., +100 V).

Fig. S6| Cyclic voltammetry curves of Pt working electrode in anhydrous propylene carbonate with traces of Na + and OHions.
Cyclic voltammetry curves using different potential windows are shown. The curves were recorded at 250 mV/s with a Pt wire pseudo-reference electrode. The arrows indicate the direction in which potential was scanned.
The reviewer is right in pointing out that any alteration of the Cu/Ti conducting layer underneath the targeted magneto-ionic materials, either Co 3 O 4 or CoN, could have some effects on the performance of the magneto-ionic systems. As demonstrated in a previous work from our group (see de Rojas, J. et al. Adv. Funct. Mater. 30, 2003704 (2020)), the electrical conductivity of the buffer layer can speed up/slow down the rate of induced magnetization as well as the total generated magnetization during the magneto-ionic processes. For this reason, the stability of the Si/Ti/Cu substrate immersed in the propylene carbonate electrolyte has been also tested. The results, shown in the Figure below (left), demonstrate the absence of distinct Cu oxidation/reduction peaks and very low current densities (background current) varying smoothly with potential, similarly to what was observed when cycling propylene carbonate using the Pt electrode. To further corroborate this point, the Si/Ti/Cu substate was then immersed in aqueous 1M NaOH solution and a cyclic voltammetry hysteresis loop was recorded under the same testing conditions. In this case, as it can be seen from the right panel of the figure below, several well-defined peaks corresponding to the formation of Cu oxides and hydroxides are observed. In addition, the current densities in the aqueous electrolyte are of the order of mA (much larger than when using propylene carbonate, where currents are of order of µA). This suggests that faradaic processes dominate over capacitive ones in an aqueous electrolyte, while there is no evidence of charge transfer reactions occurring in Si/Ti/Cu immersed in propylene carbonate (anhydrous) electrolyte. We should also point out that during the magneto-electric measurements there is no Cu exposed to the electrolyte. Thus, there is no means the Cu substrate can get oxidized or reduced with the propylene carbonate.

Cyclic voltammograms of Cu in anhydrous propylene carbonate with traces of of Na + and OHions (left) and in aqueous 1 M NaOH solution (right). The arrows indicate the direction in which potential was scanned.
Additionally, the electrochemical stability of the underneath Cu/Ti films was evaluated by TEM. As can be observed from electron energy loss spectroscopy images (EELS) in Fig. 3d and 3h of the manuscript, which belong to the samples treated during 75 min to -50 V, no oxygen or nitrogen signal is detected from the substrate Cu layer after the electrolyte gating process. This confirms the stability of the substrate layers and, therefore, corroborates that the induced effects are due to ion migration in the tested magneto-ionic materials.
Finally, cyclic voltammetry tests were performed on the CoN and Co 3 O 4 samples immersed in anhydrous propylene carbonate. The potential window was set from -20 to 20 V in order to detect any oxidation/reduction reactions taking place after overcoming the threshold magneto-ionic voltage (-4 V and -6 V, as evidenced from the magneto-electric measurements, for CoN and Co 3 O 4 , respectively). To acquire each voltammetry curve, the voltage was kept constant for 10 min in each point before sweeping voltage (due to a very slow electrode kinetics), mimicking the conditions of the magneto-ionic experiments. Unfortunately, as can be seen in the figure below ( Figure S10 of the revised version of manuscript), it is not possible to strictly determine the redox potentials of CoN and Co 3 O 4 , or to correlate the threshold voltage for magneto-ionics with electrochemical potentials for the oxidation/reduction of Co. The absence of clear maxima and the low values of current density indicate the occurrence of non-faradaic phenomena revealing that, despite the oxygen/nitrogen ion migration, these processes constitute a non-conventional redox reaction. Previous works from our group proved that the currents involved in the magneto-ionic events gated with anhydrous electrolytes are of the same order of magnitude (Navarro, C. et al., ACS Appl. Mater. Interfaces 10, 44897-44905 (2018)). In any case, the cyclic voltammetry results are consistent with the observed magneto-ionic effects since the non-linear increase in the negative current upon application of negative voltage could be attributed to the generation of metallic Co. However, in our system, the changes in the oxidation state of Co are not caused by the standard redox reactions but are 10/17 magneto-ionically driven with oxygen/nitrogen ion exchange between the cobalt oxide/nitride and the non-aqueous electrolyte (as evidenced by the detailed structural characterization provided).
To address these points the following sentences have been included in the revised version of the manuscript: "Partial decomposition of electrolyte gate (i.e., propylene carbonate) at high voltages cannot be completely ruled out 38 , although no electron-transfer reactions can be observed in the cyclic voltammetry curve when using Pt as working electrode (Fig. S6) and no bubbling was observed in this case." (page 5) "Electric-field induced oxygen ions exchange with the liquid electrolyte is the main reason of the increased current densities observed in the cycling voltammetry curves (Fig. S10)." "Similar to Co 3 O 4 , these processes are essentially non-Faradaic and do not cause pronounced peaks in the cyclic voltammetry curves (Fig. S10)." (page 11)  process? In case that oxidation of Cu occurs, the oxidation current would be detected during voltage sweeping process.
In the electrolytic cell configuration utilized in our experiments (i.e., Cu underlayer and Pt wire as working and counter electrodes, respectively), magneto-ionics is enabled by inducing oxygen and nitrogen ion migration towards the electrolyte by applying negative voltages. Specifically, ion migration starts at the top of the Co 3 O 4 and CoN films (i.e., at the boundary with the electrolyte, which acts as ion reservoir) and, with time, ions from underneath regions become activated. This way the Cu buffer layer cannot be oxidized or nitride, since the ions move towards the electrolyte, not towards Cu. As observed in the electron energy loss spectroscopy images of Figs. 3d and 3h, which correspond to the films Co 3 O 4 and CoN treated at -50 V for 75 min, respectively, neither oxygen nor nitrogen signals are detected within the Cu layer, confirming its chemical stability.
Upon application of positive voltages, used to invert the magneto-ionic process (i.e., always applied after negative voltages), the ions oxygen or nitrogen ions previously dissolved in the electrolyte are reincorporated in the films, oxidizing/nitriding the generated Co-rich regions back to the nonmagnetic oxide or nitride phases. However, for the conditions used in our experiments, we believe the ions from the solution cannot reach the Cu underlayer. Actually, if that would be the case, metallic Co may form again once Cu would integrate the oxygen or nitrogen from the Co 3 O 4 or CoN layers. This would give rise to an increase of magnetization, which is not observed for positive voltages. In addition, as it was previously demonstrated with cycling voltammetry tests discussed in response to comment # 1, we do not observe currents high enough to be unambiguously attributed to the oxidation/reduction of the Cu substrate in anhydrous propylene carbonate.
Cu oxidation or nitriding may not be ruled out if a sufficiently high positive voltage was applied first. In that case oxygen or nitrogen ions would indeed tend to diffuse towards the Cu seed layer instead of the electrolyte. The reviewer is right in pointing out that any alteration of the Cu working electrode would definitely affect the induced magneto-ionic effects. A decrease in ionic motion and generated magnetic moment would be expectable if conductivity of the Cu layer was reduced. The stability of the electrolyte was assessed by cycling voltammetry measurements, which indicated the occurrence of capacitive processes in the potential window -40/40 V (see response to comment #1). Thus, in this potential range, the formed bubbles are highly likely to be linked to the formation of oxygen and nitrogen molecule gases, since there is no evidence of electron charge transfer reactions. We agree with the reviewer that a partial cathodic decomposition of propylene carbonate may occur at high voltages, being propene (see Eichinger, G. J. Electroanal. Chem. Interf. Electrochem. 74, 183-193 (1976)) a source of gas (bubbling). However, no bubbling was observed when performing analogous magneto-ionic experiments on metallic alloy systems or during cyclic voltammetry using Pt as working electrode. Moreover, bubbling was, in fact, more prominent for CoN (compared to 12/17 Co 3 O 4 ), where magneto-ionic effects are in fact more pronounced. Note that for negative voltages, oxygen (or nitrogen) ions are released to the electrolyte, get dissolved, and eventually form bubbles at the counter-electrode once the solubility limits for O 2 and N 2 are exceeded (Refs. 35 and 36 of the revised manuscript). Due to the rather limited solubility of both gases in propylene carbonate, it is indeed expected that for very long gating voltages and time, the ions are released in the form of gas from the electrolyte into the atmosphere.
The following sentences have been added in the revised version of the manuscript (page 5): "For both systems, when gating at -50 V, bubbling occurs. Oxygen (in Co 3 O 4 ) and nitrogen (in CoN) gas evolution is most likely the major source of bubbling. When a large negative voltage is applied, high concentrations of oxygen and nitrogen anions accumulate near the counter-electrode. If their concentration is high enough, bubbles form when the oxygen (or nitrogen) solubility limits 36,37 are exceeded. When this happens, these elements are irreversibly lost from the system (sample + electrolyte) and this is the main reason causing the lack of reversibility of magneto-ionic effects when positive voltage is applied. Partial decomposition of electrolyte gate (i.e., propylene carbonate) at high voltages cannot be completely ruled out 38 , although no electron-transfer reactions can be observed in the cyclic voltammetry curve when using Pt as working electrode (Fig. S6) and no bubbling was observed in this case. Bubbling is, in fact, more pronounced for CoN than for Co 3 O 4 , where the magneto-ionic response is stronger." (4) Are the Na + ions in liquid electrolyte inserted into the CoN/Co 3 O 4 film or not during voltage sweeping process?
The liquid electrolyte used in this work contains Na + -ions. In the recent years, the Na + -ion is intensively investigated as charge carrier species in ion-batteries. In the present case, can the authors exclude the possibility Na + -ion insertion into CoN/Co 3 O 4 ? In the case that the Na + -ion insertion occurs, Na + -ion insertion would contribute to redox reaction in CoN and Co 3 O 4 , and thus it would affect the responsiveness of magnetization to the applied voltages.
We thank the reviewer for highlighting this important point. It is well known that Na + , like Li + , can mediate in redox reactions and therefore, modify the ferromagnetic properties of target materials (see Dasgupta (2017)). Further evidence of Na + not taking place as a mechanism in Co 3 O 4 reduction into Co, was demonstrated in a previous publication from our group (see Fig. S12 in Quintana, A. et al. ACS Nano 12, 10291-10300 (2018)). In that work, magneto-ionic experiments were performed in similar Co 3 O 4 films. In order to discard Na + mediated phenomena, we performed EELS analysis on the Na K edge, which revealed no measurable signal (i.e., no Na + inclusion or its presence was well below the detection limit), thereby essentially discarding any Na + reduction mediation process.  We thank the reviewer for this insightful comment and questions. The primary purpose of our ab initio calculations was to provide insights supporting the experimental observation and compare the nitrogen versus oxygen magneto-ionics. For that, we employed a simplified approach by calculating straightforwardly the energy barrier that an O or N atom needs to overcome to get inserted inside a Co film. Due to the limitations of the applied density-functional theory and, in particular, the nudged elastic band method, the CoN and Co 3 O 4 crystallographic structures were not reproduced, thus starting initially from the Co film while calculating the presented energetics. We agree with the reviewer that reproducing those crystals would be the ideal approach; however, with the chosen method we were able to provide both a good explanation of the ionic energetics and reasonable agreement with the experiment. More importantly, the calculation allowed estimating the critical electric field needed to overcome the barrier. It is worth to note that the same method was efficiently used to describe the voltage control of the magnetic anisotropy by O migration at Fe/MgO [Ibrahim, F. et al., Phys. Rev. B 98, 214441 (2018)]. Therefore, we believe that the applied approach is plausible to compare the O and N magneto-ionics.
We also agree with the reviewer that the Co sublattices in CoN and Co 3 O 4 are different from Co HCP structure. Due to the observation of both HCP (0001) and FCC (111) Co phases in the experimental samples [supplementary Table S3] and following the reviewer's comment, we performed additional calculations using a Co FCC (111) film. The obtained results are presented in the figure below and show larger energy barriers in the case of FCC-Co compared to its HCP-Co counterpart, but our main conclusions on a lower energy barrier in case of N (compared to O) still holds, independently of the Co phase. Indeed, taking into account an energy barrier ΔV= 1.37 eV/atom (1.85 eV/atom) for FCC Co-N (Co-O), respectively, the corresponding critical electric field needed to overcome this barrier is estimated as E c = 6.3 V/nm (8.5 V/nm) that is still in reasonable agreement with experiment.
Following the reviewer's comments, we have updated Fig. 5(b) and the corresponding discussion in the main text (pages 15, 16, 17): 14/17 "In turn, the calculated energy barriers between the two minima are 1.54 (1.85) and 1.14 (1.37) eV/atom for oxygen and nitrogen displacement into HCP (FCC) Co surface, respectively." "E C is found to be 8.1 (8.5) V nm -1 and 5.3 (6.3) V nm -1 for oxygen and nitrogen migration into HCP (FCC) Co, respectively, in good agreement with the onset voltages from magnetoelectric measurements (Figs. 1f and 1g) (thickness of the electric double layer < 1 nm). It is important to point out that due to the limitations of the applied density-functional theory and in particular the nudged elastic band method, the CoN and Co 3 O 4 crystallographic structures were not reproduced, starting initially from the Co film while calculating the presented energy considerations. However, with the chosen method we were able to provide both a good explanation of the ionic energetics and reasonable agreement with the experiment. More importantly, the calculations allowed estimating the critical electric field needed to overcome the barrier. It is worth noting that the same method was efficiently used to describe the voltage control of the magnetic anisotropy by O migration at Fe/MgO 22 . Therefore, we believe that the applied approach is plausible to compare O vs. N magnetoionics.".
15/17   Fig. 1, the positive electrode of the battery is connected to the thin film sample, whereas the negative electrode of the battery is connected in Ref.7. It is very confusing for readers. The author should clearly define the sign of the applied voltage in the experimental methods, rather than simply citing the literature.
We do agree with the reviewer. To avoid misunderstandings, citation to Refs. 7, 8 and 27 has been eliminated and the following sentence has been added in the Methods section (page 19): "The sign of voltage was such that negative charges accumulate at the working electrode when negative voltage was applied (and vice versa for positive voltages). in a similar fashion to that presented in references 7, 8 and 27."