ARTICLE https://doi.org/10.1038/s41467-020-19758-x OPEN Voltage-driven motion of nitrogen ions: a new paradigm for magneto-ionics Julius de Rojas 1, Alberto Quintana 2, Aitor Lopeandía1, Joaquín Salguero3, Beatriz Muñiz3, Fatima Ibrahim4, Mairbek Chshiev 4,5, Aliona Nicolenco1, Maciej O. Liedke 6, Maik Butterling6, Andreas Wagner 6, Veronica Sireus1, Llibertat Abad7, Christopher J. Jensen 2, Kai Liu 2, Josep Nogués8,9, José L. Costa-Krämer 3, Enric Menéndez 1 ✉ & Jordi Sort1,9 ✉ 1234567890():,; 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 Departament de Física, Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Spain. 2 Department of Physics, Georgetown University, Washington, DC 20057, USA. 3 IMN-Instituto de Micro y Nanotecnología (CNM-CSIC), Isaac Newton 8, PTM, 28760 Tres Cantos, Madrid, Spain. 4 Univ. Grenoble Alpes, CEA, CNRS, Spintec, 38000 Grenoble, France. 5 Institut Universitaire de France, 75231 Paris, France. 6 Institute of Radiation Physics, Helmholtz-Zentrum Dresden–Rossendorf, Dresden 01328, Germany. 7 Institut de Microelectrònica de Barcelona, IMB-CNM (CSIC), Campus UAB, 08193 Bellaterra, Spain. 8 Catalan Institute of Nanoscience and Nanotechnology (ICN2), CSIC and BIST, Campus UAB, Bellaterra, 08193 Barcelona, Spain. 9 Institució Catalana de Recerca i Estudis Avançats (ICREA), Pg. Lluís Companys 23, 08010 Barcelona, Spain. ✉email: enric.menendez@uab.cat; jordi.sort@uab.cat NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x M agneto-ionics1–17, i.e., the change in the magnetic polycrystalline (with cubic structure) and non-ferromagnetic properties of materials due to electric field-induced ion (Supplementary Information, Figs. S1–S4). The voltage actuation motion, is acquiring a leading role, among other mag- is carried out via electrolyte-gating using an electrochemical netoelectric mechanisms (intrinsic18 or extrinsic19 multiferroicity, capacitor configuration8 (Fig. 1a). In this way, the overall film area electric charge accumulation20–22), to control magnetism with exposed to the liquid electrolyte29,30 is activated, establishing a voltage23,24. This is triggered by its capability to largely modulate well-defined out-of-plane electric field8. To investigate magneto- magnetic properties in a permanent and energy-efficient way2,25. ionic motion, the films were electrolyte-gated at –50 V for 12 h Usually, magneto-ionic systems comprise layered hetero- and, during this time, consecutive magnetic hysteresis loops structures built around a ferromagnetic target material, such as (25 min duration each) were sequentially recorded. In Fig. 1b, c, Co or Fe, grown adjacent to solid-state electrolyte films (e.g., the red loops correspond to the first measurement under voltage GdOx2 or HfO226). Depending on the voltage polarity, these and the black arrows indicate the first sweeping leg of the cycle. In electrolytes accept or donate oxygen, acting as ion reservoirs. In this timeframe (during the descending branch of the loop), the this way, for instance, the effective magnetic easy axis of ferro- magnetization (M) of the CoN film significantly increases, whereas magnetic layers can be precisely controlled2. However, room- Co3O4 still remains paramagnetic, evidencing that nitrogen motion temperature ionic response is slow (∼102–103 s)2. Therefore, is significantly faster than oxygen transport. This is demonstrated since ion migration is a thermally activated process27, ion diffu- in Fig. 1d, which shows the saturation magnetization, MS, as a sion is commonly assisted with heat2, which is detrimental in function of time (Supplementary Fig. S5 shows details on MS terms of energy efficiency. Sometimes cyclability is limited due to quantification). MS evolves with time for both films, reaching, after cumulative irreversible changes that the target materials undergo the magneto-ionic motion has stabilized, maximum values of 588 from structural/compositional viewpoints2. Recently, through a and 637 emu cm–3 for Co3O4 and CoN, respectively (Table 1). By proton-based route, 10–1 s ionic motion has been demonstrated at linearly fitting MS vs. t during the first minutes of voltage appli- room-temperature, with good endurance, despite restricted cation (wherein MS in CoN fully saturates), magneto-ionic rates of hydrogen retention9. Hydrogen is mainly adsorbed rather than 467 and 2602 emu cm–3 h–1 are obtained for oxygen in Co3O4 and absorbed, which imposes stringent limitations on the thickness of nitrogen in CoN, respectively (Table 1), i.e., a five-fold enhance- the ferromagnet. Other approaches relying on the insertion/ ment for the latter. For both systems, when gating at –50 V, removal of ions, such as Li11–13,28 or F14, into a ferromagnet are bubbling occurs. Oxygen (in Co3O4) and nitrogen (in CoN) gas promising in terms of reversibility. However, due to incompat- evolution is most likely the major source of bubbling. When a large ibilities with complementary metal-oxide semiconductor (CMOS) negative voltage is applied, high concentrations of oxygen and architectures, applications in electronics in these cases are nitrogen anions accumulate near the counter-electrode. If their limited9. concentration is high enough, bubbles form when the oxygen (or An alternative approach is to use target materials, which are nitrogen) solubility limits36,37 are exceeded. When this happens, already oxidized. Magneto-ionics using structural oxygen (i.e., these elements are irreversibly lost from the system (sample + oxygen incorporated in the crystallographic structure of the electrolyte) and this is the main reason causing the lack of rever- actuated material) exhibits outstanding stability and sibility of magneto-ionic effects when positive voltage is applied. reversibility7,8. This has been demonstrated in electrolyte- Partial decomposition of electrolyte gate (i.e., propylene carbonate) gated20,29–33, thick (≥100 nm) paramagnetic Co3O4 films, in at high voltages cannot be completely ruled out38, although no which room-temperature voltage-controlled ON-OFF ferro- electron-transfer reactions can be observed in the cyclic voltam- magnetism has been achieved, benefiting from defect-assisted metry curve when using Pt as working electrode (Supplementary voltage-driven transport of structural oxygen. Nevertheless, there Fig. S6) and no bubbling was observed in this case. Bubbling is, in is an inherent voltage trade-off between induced magnetization, fact, more pronounced for CoN than for Co3O4, where the speed and cyclability. Specifically, the generated magnetization magneto-ionic response is stronger. The hysteresis cycles from increases with voltage, whereas cyclability degrades for exceed- electrolyte-gated CoN are more square-shaped than for Co3O4 ingly high voltages due to irreversible losses of oxygen (i.e., (Fig. 1b, c). They exhibit larger squareness [defined as the ratio bubbling)7,8. between the remnant magnetization (MR) and MS (MR/MS)], and Here, nitrogen magneto-ionics is demonstrated as an improved larger slopes at the coercivity (HC) normalized to MS [dM/dH(H = alternative to oxygen magneto-ionics. CoN and Co3O4 single-layer HC) MS–1] than the loops of Co3O4 throughout the time the voltage films are voltage-actuated to compare nitrogen vs. oxygen magneto- was applied (Fig. 1e). ionic performances. These materials were selected since they both Additionally, the HC corresponding to Co3O4 scales mono- exhibit voltage-induced ON-OFF ferromagnetic transitions. Both tonically with time, while in CoN HC shows a maximum at the films were grown by sputtering, have the same thickness and exhibit initial stages of gating and decreases afterwards (Supplementary a similar (nanocrystalline, face-centered cubic) microstructure. Fig. S7). The case of CoN bears a resemblance to the characteristic Remarkably, voltage-driven transport of structural nitrogen is dependence of HC with particle size in magnetic systems, energetically more favorable than oxygen, resulting in lower oper- consistent with a homogeneous generation of ferromagnetic ating voltages and enhanced cyclability. This together with the regions, uniformly evolving in size, starting from a super- lower electronegativity (i.e., weaker bonds with Co) of nitrogen with paramagnetic behavior (zero HC), followed by a single domain respect to oxygen leads to overall enhanced magneto-ionic effects. state (maximum HC) and, afterwards, reaching a multi-domain Controlled motion of nitrogen ions with voltage might enable the state with reduced HC39. The squareness of the loops and the rather use of magneto-ionics in new technological areas that require small HC for CoN indicate an in-plane anisotropy and likely a endurance and moderate operation speeds (e.g., neuromorphic reversal by domain wall motion, which hints at the uniformity of computing34 or micro-electro-mechanical systems35). the generated metallic Co phase (Table 1). Such low coercivities might be also associated with a highly nanostructured or even amorphous-like Co40. In contrast, the generated ferromagnetism in Results Co3O4 shows much larger HC, which results from the formation of Oxygen vs. nitrogen magneto-ionics: magnetoelectric char- isolated Co clusters immersed in a residual Co3O4 matrix7,8. acterization. Structural and magnetic characterization of the as- The onset voltage for magneto-ionic motion was determined in deposited films reveals that both Co3O4 and CoN films are as-prepared films by monotonically increasing the absolute value 2 NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x ARTICLE a Pt Liquid Electrolyte Magneto-io Metal Oxide nic Target Cu or Nitride Si Ti b c Co3O4 CoN 0.5 As-prepared 0.5 As-prepared M (103 emu cm–3) M (103 emu cm–3) 0 – 25 min 0 – 25 min 50 – 75 min 50 – 75 min 0.0 0.0 –0.5 –0.5 –10 –5 0 5 10 –10 –5 0 5 10 H (kOe) H (kOe) d e dM/dH [H =Hc ] MS–1 (kOe–1) 100 1.0 MS (103 emu cm–3) 75 0.5 MR /MS 0.5 50 CoN Co3O4 25 CoN Co3O4 0.0 0.0 0 0 2 4 6 8 10 12 0 2 4 6 8 10 12 time (h) time (h) f 50 g 50 –4 V –6 V +6 V –2 V –4 V +4 V 40 40 Co3O4 CoN MS (emu cm–3) MS (emu cm–3) 30 30 20 20 10 10 0 0 0 1 2 0 1 2 time (h) time (h) Fig. 1 Oxygen vs. nitrogen magneto-ionics: magnetoelectric characterization. a Schematic representation of the electrolyte-gating process in the Co3O4 and CoN films. b, c Hysteresis loops (each of 25 min of duration) of the as-prepared film (black) and under –50 V gating (the first (0–25 min) and the third (50–75 min) cycles are shown only) for the Co3O4 and CoN films, respectively, obtained by in-plane vibrating sample magnetometry. d Time evolution of the saturation magnetization (MS vs. t) and e squareness (MR/MS) and slope of hysteresis loop at HC normalized to MS (dM/dH[H = HC] MS–1) for each film. f, g Time evolution of MS when the gating is monotonically increased in steps of –2 V to determine the onset voltage required to display ferromagnetism for Co3O4 and CoN, respectively. Table 1 Oxygen vs. nitrogen magneto-ionics by magnetoelectric measurements. Film Onset voltage (V) Recovery voltage (V) dM/dt @ –50 V MS MR/MS (%) dM/dH MS–1 @ HC HC (Oe) (emu cm–3 h–1) (emu cm–3) (kOe–1) Co3O4 –6 +20 467 588 62 7.4 185 CoN –4 +4 2602 637 96 97.4 17 Onset and recovery voltages, magneto-ionic speed and magnetic properties of the generated ferromagnetic phases. Note that MS has been obtained by normalizing the magnetic moment to the nominal film thickness, without taking into consideration the formation of vacancies or eventual nanoporosity. NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x of the negative gating in steps of –2 V to observe when the films thickness of 2–3 nm. Similar affected thicknesses during rever- started to display ferromagnetism (Fig. 1f, g and Supplementary sible voltage-driven magnetization cycling can be estimated for Fig. S3). Interestingly, the onset voltage for CoN (–4 V) is lower Co3O4. In any case, overall, Fig. 2 corroborates that magneto- than for Co3O4 (–6 V). To investigate reversibility of the ionic rates, onset voltages and endurance are better when using magneto-ionic process, both systems were kept at their respective nitrogen than oxygen migration. onset voltages for one hour and, then, the polarity was inverted (i.e., +4 and +6 V). Remarkably, while Co3O4 recovers only partially at +6 V, CoN fully recuperates the pristine paramagnetic Ion transport mechanisms. Cross-section lamellae of the as- state at +4 V. These results anticipate a higher activation energy prepared and electrolyte-gated samples treated at –50 V for 75 for oxygen transport than for nitrogen. In fact, to recover the min were studied by transmission electron microscopy (TEM) paramagnetic state in Co3O4 + 20 V are required (Table 1). The (Fig. 3). The as-prepared Co3O4 shows regular, columnar-shaped need to actuate at a higher voltage induces irreversible losses of grains (Fig. 3a and Supplementary Fig. S8a). Co (red) and O oxygen since the liquid electrolyte has a limited solubility of (blue) are homogeneously distributed in the as-grown film oxygen41. Note that if –50 V are applied, none of the treated films (Fig. 3b). Conversely, upon gating at –50 V, this morphology is recoverable in agreement with the irreversible loss of oxygen drastically changes redistributing the elements into O-rich and nitrogen through bubbling. channels in agreement with previously reported results7,8 (Fig. 3c, d and Supplementary Fig. S7c). This confirms that oxygen transport takes place via a two-fold mechanism: (i) uni- Magneto-ionic cyclability. To investigate the cyclability, the CoN form oxygen transport towards the electrolyte and (ii) localized and Co3O4 films were subjected to –4 V/ +4 V and –8 V/ +8 V oxygen migration along diffusion channels. High-resolution TEM pulses of relatively short duration (≈8.5 min/cycle). The duration (Fig. S9) shows remaining Co3O4 after applying –50 V. Electric- of the onset negative voltage pulse was selected in each material to field induced oxygen ions exchange with the liquid electrolyte is give a ΔM of approximately 1 emu cm–3 in the first cycle. This the main reason of the increased current densities observed in the resulted in times of approximately 4.1 min for CoN and 6.2 min cycling voltammetry curves (Supplementary Fig. S10). for Co3O4 (evidencing that the response of Co3O4 is delayed with The as-prepared CoN film shows an isotropic and highly respect to that of CoN). As seen in Fig. 2, cycling at –/+4 V nanostructured morphology with homogeneous composition results in a very stable periodic response for CoN. Conversely, at (Fig. 3e, f and Supplementary Fig. S7b). Remarkably, upon gating this voltage, no traces of magneto-ionic effects are observed in at –50 V for 75 min, a well-defined interface (resembling a Co3O4, corroborating the need of a higher onset voltage. Good diffusion front) parallel to the surface is distinguished, dividing cyclability is observed for Co3O4 at –8 V/ +8 V. Note that, in the film in two sublayers with different microstructures contrast to Fig. 1f, short pulses at –/+8 V allow recovering the (Supplementary Figs. S8d and S11). No traces of nitrogen are initial state in Co3O4 after each cycle. Both materials show no detected in either of the two sublayers, evidencing a full progressive irreversible gain/loss of magnetic signal upon suc- denitriding process, consistent with the magnetoelectric char- cessive cycling. However, while the ΔM amplitude in each cycle acterization. While sublayer 2 is nanocrystalline, sublayer 1 (in remains stable for CoN, a loss of about 25% is observed in Co3O4 contact with the electrolyte during magnetoelectric measure- after the first cycle. As a first approach, a simple calculation ments) is amorphous-like and exhibits lower density (decrease of considering the overall magnetization induced in CoN when Co signal in Fig. 3h), possibly due to free volume or nanoporosity. applying – 50 V (i.e., 637 emu cm−3, Table 1, when the whole This suggests a complex denitriding process in which, under large CoN is affected, as will be shown in the forthcoming section) negative voltages, not only does nitrogen migrate from the sample suggests that during this reversible cycling (1 emu cm−3, Fig. 2), to the electrolyte, but also Co is redistributed across the film only the uppermost 0.1–0.2 nm of the CoN layer are involved in thickness. Similar to Co3O4, these processes are essentially non- the observed reversible effect. If the threshold negative voltage is Faradaic and do not cause pronounced peaks in the cyclic applied for longer times (Fig. 1g), the induced magnetization voltammetry curves (Supplementary Fig. S10). from nitrogen ion migration leads to an equivalent CoN To further asses the microstructure of the films upon magneto- ionic actuation, variable energy positron annihilation lifetime 12 Co3O4 –4 / +4 V Pulses spectroscopy experiments42,43 were performed (Methods). In the Co3O4 –8 / +8 V Pulses as-prepared CoN sample, only τ1 and τ2 lifetime components are 10 CoN –4 / +4 V Pulses observed, indicating the absence of void-like structures (no τ3) in the pre-biasing state (Fig. 4). τ1 is around 0.28 ns in the top half of M (emu cm–3) 8 the film, which could correspond to a cluster of more than 4 vacancies7. For the bottom half, τ1 is slightly lower, indicating 6 clusters with less than 4 vacancies. The second lifetime τ2 4 represents a mixture of signals from surface states and grain boundaries. In the first tens of nm, τ2 is larger than 0.5 ns, 2 indicating the presence of small voids. As seen in Fig. 4b, the relative intensity I1 decreases, while I2 increases reaching similar 0 intensities at the interface with the buffer layer, reflecting the 0 10 20 30 40 50 time (min) influence of the buffer polycrystallinity in the CoN growth, whose extent decreases with film thickness. Fig. 2 Cyclability. Magneto-ionic cyclability of the Co3O4 and CoN films Negative biasing of –20 V and –50 V increases τ1 and τ2 and subjected to –4 V/+4 V and –8 V/+8 V, and to –4 V/+4 V, respectively, this increase scales with voltage, indicating that the initial open with pulses of short duration (≈8.5 min/cycle). The pale yellow lines volumes become larger. Already for –20 V, τ3 emerges, showing represent the average curves. Horizontal lines (red, green) are guides to the the presence of large voids as it happens in Co3O47. However, for eye and span 1 emu cm−3. The data are shifted in ΔM-axis to make them Co3O4, a monotonic increase of relative intensity I2 across the distinguishable. Cyclability was carried out under the application of 5 kOe film thickness is observed7, whereas relative maxima (marked to ensure being above the anisotropy field and, thus, in saturation. with arrows in Fig. 4b) are found for CoN. The depth position of 4 NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x ARTICLE a b 50 nm Pt Co3O4 Cu c d 50 nm Pt "Co3O4" Cu e f 50 nm Pt CoN Cu g h 50 nm Pt sublayer 1 "CoN" sublayer 2 Cu Co O N Fig. 3 Compositional characterization by high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and electron energy loss spectroscopy (EELS). a–b, c–d, e–f, and g–h are the HAADF-STEM images and corresponding elemental EELS mappings of the areas marked in orange, respectively, of the as-prepared Co3O4 film, Co3O4 film subjected to a −50 V for 75 min, as-prepared CoN film and CoN film subjected to a −50 V for 75 min, respectively. The colors corresponding to each element for the EELS analyses are depicted at the bottom of the figure. these relative maxima increases with the applied voltage, re-introduced to the films with positive voltage), the insertion suggesting the occurrence of an interface migration front, in of an atom of either oxygen or nitrogen into a cobalt slab, with agreement with TEM observations. For –50 V, the migration either (0001) HCP or (111) FCC orientation, has been considered. front moves deeper into the film in agreement with a more Minimum energy paths were calculated by the nudged elastic intense denitriding/amorphization process. I3 tends to decrease band method (Methods). The obtained total energies per atom and vanishes with thickness, evidencing that larger voids are only normalized to the global minimum value are plotted in Fig. 5b as present at the top surface (as hinted by the TEM-EELS analysis; a function of the displacement z of the oxygen or nitrogen atom Fig. 3h). The ionic transport upon electrolyte-gating in CoN is from the cobalt reference layer. The reference z = 0 Å is assigned thus consistent with a uniform nitrogen migration through to the position of the outermost cobalt layer. The global energy vacancies and grain boundaries (τ1 and τ2 increase), leaving minimum is found at z = 1.14 Å for both oxygen and nitrogen behind larger grain boundaries and voids. cases. Another local minimum is located around z = –0.75 Å for oxygen and z = –1 Å for nitrogen. In turn, the calculated energy Co-O vs. Co-N formation energy. Neither Co3O4 nor CoN are barriers between the two minima are 1.54 (1.85) and 1.14 (1.37) ferromagnetic at room-temperature. The magnetic properties of eV/atom for oxygen and nitrogen displacement into HCP (FCC) these systems are strongly correlated to the amount of either oxygen Co surface, respectively. Thus, inserting nitrogen into cobalt is or nitrogen in the films. According to the virtual crystal approx- energetically more favorable (i.e., less energy required) compared imation, for CoN, beyond 50 at. % of nitrogen in the unit cell, the to inserting oxygen. Using the calculated energy barrier values, magnetic moment becomes negligible (Fig. 5a). This explains why one can infer the critical electric fields, EC, needed to overcome the as-prepared CoN film is not magnetic and it also sets a limit for the energy barrier by oxygen and nitrogen atoms. These electric irreversible losses of nitrogen beyond which successive voltage- fields can be estimated as EC ¼ ΔV Δz where ΔV and Δz represent, driven ON-OFF-ON ferromagnetism would be compromised. respectively, the electric potential per atom to overcome the To simulate the Co-O and Co-N formation energy (which energy barrier and the distance between minima that an occurs when the dissolved oxygen and nitrogen ions are atom must migrate25. EC is found to be 8.1 (8.5) V nm–1 and NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x As-prepared : 1, I1 2, I2 a –20 V @ 75 min: 1, I1 2, I2 2, I2 2.0 –50 V @ 75 min: 1, I1 2, I2 3, I3 1.5 c M (µB/Co) a 3 Positron Lifetimes, i (ns) 2 1.0 b a 0.8 0.6 0.5 CoN 0.4 0.0 0.30 0 20 40 60 0.25 Nitrogen Concentration (%) 0.20 1 2 3 4 b 2.5 EP (KeV) Co-O: HCP FCC Total Energy (eV atom–1) Co-N: HCP FCC 2.0 N or O b 100 Initial Position 1.5 z=0 Relative Intensities, Ii (%) Co 80 1.0 60 0.5 c 40 migration fronts 0.0 Cobalt Film Film Surface 20 –2 –1 0 1 2 3 a b z (Å) 0 1 2 3 4 Fig. 5 Ab initio calculations: magnetism in the Co-N system and Co-O vs. EP (KeV) Co-N formation energy. a Variation of magnetic moment in Bohr magneton, μB, per Co atom as a function of increasing the N percentage in Fig. 4 Defect characterization by variable energy positron annihilation the CoN unit cell, shown in the inset, calculated within Virtual Crystal lifetime spectroscopy (VEPALS). a Positron lifetime components τi=1–3 and Approximation (VCA). b Calculated total energy per atom, normalized to b their relative intensities Ii=1–3 as a function of positron implantation the minimum energy value, as a function of the displacement between the energy Ep for as-prepared and –20 V and –50 V biased CoN films. The non- reference Co outermost surface atom and the inserted O or N atom. Both monotonic change of intensity I2 with depth is linked to the position of the HCP (0001) and FCC (111) surfaces are considered represented by interface between Co sublayers reminiscent of an ionic migration front as filled and open symbols, respectively. The black squares (red circles) highlighted in b. correspond to the O (N) energetic path, respectively. The five-monolayer- thick Co slab is shown in the right panel, where the dashed line indicates 5.3 (6.3) V nm–1 for oxygen and nitrogen migration into HCP the reference z position, which is the outermost Co surface monolayer. (FCC) Co, respectively, in good agreement with the onset voltages from magnetoelectric measurements (Fig. 1f, g) (thickness of the electronegativity of nitrogen is lower than that of oxygen, electric double layer <1 nm). It is important to point out that due resulting in weaker bonds with cobalt, allowing for an enhanced to the limitations of the applied density-functional theory and in ionic motion. particular the nudged elastic band method, the CoN and Co3O4 crystallographic structures were not reproduced, starting initially Discussion from the Co film while calculating the presented energy Our work demonstrates robust room-temperature nitrogen considerations. However, with the chosen method we were able magneto-ionics in CoN. Nitrogen magneto-ionics shows reduced to provide both a good explanation of the ionic energetics and activation energies for ionic transport, thus requiring lower vol- reasonable agreement with the experiment. More importantly, the tage actuation. Moreover, the magneto-ionic rates are faster than calculations allowed estimating the critical electric field needed to for oxygen magneto-ionics. This is linked to the conjugation of a overcome the barrier. It is worth noting that the same method lower critical electric field to overcome the energy barrier for ion was efficiently used to describe the voltage control of the magnetic diffusion and a lower electronegativity of nitrogen with respect to anisotropy by O migration at Fe/MgO25. Therefore, we believe oxygen. The dissimilar electric properties of CoN and Co3O4 are that the applied approach is plausible to compare O vs. N also likely to play a role in the way ions diffuse in the two layers. magneto-ionics. Thus, nitrogen magneto-ionics represents a robust alternative for Since the ionic radii of nitrogen ions are larger than the ionic efficient voltage-driven effects and may enable the use of radius of O2–, once the energy barrier for ion diffusion is magneto-ionics in devices that require endurance and moderate overcome ionic motion would be, in first approximation, speeds of operation, such as brain-inspired/stochastic computing expected to be larger for oxygen than for nitrogen. However, or magnetic micro-electro-mechanical systems. The reported the simulation results indicate the opposite, revealing that other effects are also appealing to extend the use of nitride semi- parameters, such as electronegativity, might play a more conductors in diverse applications such as electrochemical sen- dominant role than ion size in ionic motion. In fact, the Pauling sors, catalysis, batteries, spintronics, or iontronics. 6 NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x ARTICLE Methods group for CoN. The full structural relaxation was performed until the forces became Sample preparation. Eighty-five nanometer thick Co3O4 and CoN films were smaller than 1 meV Å−1, yielding a lattice constant of 4.41 Å. The virtual crystal grown by reactive sputtering on B-doped, highly conducting [100]-oriented Si approximation54 was used to model the variation of nitrogen per unit cell. To wafers (0.5 mm thick), previously coated with 20 nm of Ti and 60 nm of Cu. compare the Co-O and Co-N formation energy, the nudged elastic band method Depositions were carried always out while partly masking the Cu layer to serve as (NEB)55,56 was employed on the oxygen and nitrogen pathway into a five-monolayer working electrode. thick (0001) hexagonal close-packed Co slab. At each step, the atomic coordinates The Co3O4 films were grown at room-temperature in an AJA International were relaxed until the forces became smaller than 1 meV Å−1. A kinetic energy cut-off ATC 2400 Sputtering System with a base pressure in the 10–8 Torr range. High of 500 eV was used for the plane-wave basis set and 25 × 25 × 25 and 25 × 25 × 1 k- purity Ar and O2 gases were used. The target to substrate distance was around 8 cm point meshes were used to construct the first Brillouin zone in CoN unit cell and the and the sputtering rate of about 5 Å s−1. Co3O4 was grown in a 7% O2/93% Ar Co slab in the NEB calculations, respectively. atmosphere at a total pressure of 2.5 × 10–3 Torr. The CoN films were grown in a homemade triode sputtering system with a base Reporting summary. Further information on research design is available in the Nature pressure in the 10–8 Torr range. Ultra-high vacuum was ensured to minimize oxygen Research Reporting Summary linked to this article. contamination and, thus, to rule out traces of oxygen magneto-ionics. The target to substrate distance was around 10 cm and the sputtering rate about 1 Å s−1. CoN was grown in a 50% N2/50% Ar atmosphere at a total pressure of 8 × 10–3 Torr. Data availability The data used in this article are available from the corresponding authors upon request. Magnetoelectric characterization. Magnetoelectric measurements were carried out by performing vibrating sample magnetometry while electrolyte gating the Received: 17 June 2020; Accepted: 26 October 2020; films in a capacitor configuration at room-temperature. A magnetometer from Micro Sense (LOT-Quantum Design), with a maximum applied magnetic field of 2 T, was used. The samples are mounted in a homemade electrolytic cell filled with anhydrous propylene carbonate with Na+ solvated species (5–25 ppm). The magnetic properties were measured along the film plane upon applying different voltages, using an external Agilent B2902A power supply, between the Cu working electrode and the counter-electrode (Pt wire). The sign of voltage was such that References negative charges accumulate at the working electrode when negative voltage was 1. Molinari, A., Hahn, H. & Kruk, R. Voltage-control of magnetism in all- applied (and vice versa for positive voltages). The Na+ solvated species in the solid-state and solid/liquid magnetoelectric composites. Adv. Mater. 31, electrolyte are aimed at reacting with any traces of water30. The magnetic signal is 1806662 (2019). normalized to the volume of the sample exposed to the electrolyte. Note that the 2. Bauer, U. et al. Magneto-ionic control of interfacial magnetism. Nat. Mater. hysteresis loops were background-corrected using the signal at high fields (i.e., 14, 174–181 (2015). fields always far above saturation fields) to eliminate linear contributions. 3. Gilbert, D. A. et al. Structural and magnetic depth profiles of magneto-ionic heterostructures beyond the interface limit. Nat. Commun. 7, 12264 (2016). Structural and compositional measurements. θ/2θ X-ray diffraction (XRD) pat- 4. Gilbert, D. A. et al. Controllable positive exchange bias via redox-driven terns were recorded on a Materials Research Diffractometer (MRD) from Malvern oxygen migration. Nat. Commun. 7, 11050 (2016). PANalytical company, equipped with a PIXcel1D detector, using Cu Kα radiation. The 5. Duschek, K., Uhlemann, M., Schlörb, H., Nielsch, K. & Leistner, K. patterns were analyzed using a full-pattern Rietveld refinement method. Electrochemical and in situ magnetic study of iron/iron oxide films oxidized High-resolution transmission electron microscopy (HRTEM), high-angle annular and reduced in KOH solution for magneto-ionic switching. Electrochem. dark-field scanning transmission electron microscopy (HAADF-STEM), and electron Commun. 72, 153–156 (2016). energy loss spectroscopy (EELS) were performed on a TECNAI F20 HRTEM/STEM 6. Duschek, K., Petr, A., Zehner, J., Nielsch, K. & Leistner, K. All-electrochemical microscope operated at 200 kV. Cross sectional lamellae were prepared by focused ion voltage-control of magnetization in metal oxide/metal nanoislands. J. Mater. beam and placed onto a Cu transmission electron microscopy grid. Chem. C. 6, 8411–8417 (2018). 7. Quintana, A. et al. Voltage-controlled ON-OFF ferromagnetism at room Transport measurements. To determine electric properties, both films (Co3O4 temperature in a single metal oxide film. ACS Nano 12, 10291–10300 (2018). and CoN) were deposited onto high resistivity Si substrates. To assess the semi- 8. de Rojas, J. et al. Boosting room-temperature magneto-ionics in a non- conducting/metallic behavior of CoN, resistivity values were acquired from 30 to magnetic oxide semiconductor. Adv. Funct. Mater. 30, 2003704 (2020). 300 K. In all cases, the van der Pauw configuration was used. 9. Tan, A. J. et al. Magneto-ionic control of magnetism using a solid-state proton pump. Nat. Mater. 18, 35–41 (2019). 10. Gößler, M. et al. Magneto-ionic switching of superparamagnetism. Small 15, Variable energy positron annihilation lifetime spectroscopy. Variable energy 1904523 (2019). positron annihilation lifetime spectroscopy (VEPALS) measurements were con- ducted at the mono-energetic positron source (MePS) beamline, which is an end 11. Dasgupta, S. et al. Intercalation‐driven reversible control of magnetism in bulk station of the radiation source ELBE (Electron Linac for beams with high Brilliance ferromagnets. Adv. Mater. 26, 4639–4644 (2014). and low Emittance) at Helmholtz-Zentrum Dresden-Rossendorf (Germany)42 12. Dasgupta, S. et al. Toward on-and-off magnetism: reversible electrochemistry using a digital lifetime CeBr3 scintillator detector with a homemade software to control magnetic phase transitions in spinel ferrites. Adv. Funct. Mater. 26, employing a SPDevices ADQ14DC-2X with 14 bit vertical resolution and 2GS s–1 7507–7515 (2016). (GigaSamples per second) horizontal resolution and with a time resolution func- 13. Zhu, X. et al. In situ nanoscale electric field control of magnetism by tion down to about 0.205 ns. The resolution function required for spectrum ana- nanoionics. Adv. Mater. 28, 7658–7665 (2016). lysis uses two Gaussian functions with distinct intensities depending on the 14. Vasala, S. et al. Reversible tuning of magnetization in a ferromagnetic positron implantation energy, Ep, and appropriate relative shifts. All spectra con- Ruddlesden–Popper‐type manganite by electrochemical fluoride‐ion tained at least 1 × 107 counts. The spectra were deconvoluted using the non-linearly intercalation. Adv. Electron. Mater. 6, 1900974 (2020). least-squared-based package PALSfit fitting software43 into discrete lifetime com- 15. Lu, N. et al. Electric-field control of tri-state phase transformation with a ponents, which directly confirm different defect types—i.e., sizes—(Fig. 4). selective dual-ion switch. Nature 546, 124–128 (2017). The corresponding relative intensities (Ii) reflect to a large extent the 16. Taniguchi, K. et al. In situ reversible ionic control for nonvolatile magnetic concentration of each defect type. In general, positron lifetime (τi) is directly phases in a donor/acceptor metal‐organic framework. Adv. Funct. Mater. 27, proportional to defect size, i.e., the larger the open volume, the lower the probability 1604990 (2017). and the longer it takes for positrons to be annihilated with electrons42–47. The 17. Molinari, A., Hahn, H. & Kruk, R. Voltage-controlled on/off switching of positron lifetime and its intensity are probed as a function of positron implantation ferromagnetism in manganite supercapacitors. Adv. Mater. 30, 1703908 energy Ep or, in other words, implantation depth (thickness). The shortest lifetime (2018). component (τ1 < 0.32 ns) represents positron annihilation inside vacancy clusters 18. Eerenstein, W., Mathur, N. D. & Scott, J. F. Multiferroic and magnetoelectric (likely within grains). The intermediate lifetime (0.35 < τ2 < 0.90 ns) accounts for materials. Nature 442, 759–765 (2006). annihilation at larger vacancy clusters (linked to grain boundaries), surface states, 19. Wang, Y., Hu, J., Lin, Y. & Nan, C.-W. Multiferroic magnetoelectric composite and small voids/pores (0.28−0.37 nm in diameter, calculated based on the shape- nanostructures. NPG Asia Mater. 2, 61–68 (2010). free model for pore-size estimation of Wada et al.48). The longest lifetime 20. Weisheit, M. et al. Electric field-induced modification of magnetism in thin- component (2.3 < τ3 < 3.3 ns) indicates contributions of larger voids (0.58–0.74 nm film ferromagnets. Science 315, 349–351 (2007). in diameter). 21. Ovchinnikov, I. V. & Wang, K. L. Theory of electric-field-controlled surface ferromagnetic transition in metals. Phys. Rev. B 79, 020402(R) (2009). Ab initio calculations. The first-principles calculations were based on the projector- 22. Quintana, A. et al. Voltage-induced coercivity reduction in nanoporous alloy augmented wave (PAW) method49 as implemented in the VASP package50–52 using films: a boost toward energy-efficient magnetic actuation. Adv. Funct. Mater. the generalized gradient approximation53. We used cubic unit cell with a F43 m space 27, 1701904 (2017). NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-020-19758-x 23. Song, C., Cui, B., Li, F., Zhou, X. & Pan, F. Recent progress in voltage control 53. Perdew, J. P., Burke, K. & Ernzerhof, M. Generalized gradient approximation of magnetism: materials, mechanisms, and performance. Prog. Mater. Sci. 87, made simple. Phys. Rev. Lett. 77, 3865–3868 (1996). 33–82 (2017). 54. Bellaiche, L. & Vanderbilt, D. Virtual crystal approximation revisited: 24. Hu, J.-M. & Nan, C.-W. Opportunities and challenges for magnetoelectric application to dielectric and piezoelectric properties of perovskites. Phys. Rev. devices. APL Mater. 7, 080905 (2019). B 61, 7877–7882 (2000). 25. Ibrahim, F., Hallal, A., Dieny, B. & Chshiev, M. Establishing characteristic 55. Henkelman, G., Uberuaga, B. P. & Jónsson, H. A climbing image nudged behavior of voltage control of magnetic anisotropy by ionic migration. Phys. elastic band method for finding saddle points and minimum energy paths. J. Rev. B 98, 214441 (2018). Chem. Phys. 113, 9901–9904 (2000). 26. Yan, Y. N. et al. Electrical control of Co/Ni magnetism adjacent to gate oxides 56. Henkelman, G. & Jónsson, H. Improved tangent estimate in the nudged elastic with low oxygen ion mobility. Appl. Phys. Lett. 107, 122407 (2015). band method for finding minimum energy paths and saddle points. J. Chem. 27. Strukov, D. B. & Williams, R. S. Exponential ionic drift: fast switching and low Phys. 113, 9978–9985 (2000). volatility of thin-film memristors. Appl. Phys. A 94, 515–519 (2009). 28. Zhang, Q. et al. Lithium-ion battery cycling for magnetism control. Nano. Lett. 16, 583–587 (2016). Acknowledgements 29. Leighton, C. Electrolyte-based ionic control of functional oxides. Nat. Mater. Financial support by the European Research Council (2014-Consolidator Grant Agree- 18, 13–18 (2019). ment No. 648454 and 2019-Proof of Concept Grant Agreement N° 875018), the Spanish 30. Navarro-Senent, C., Quintana, A., Menéndez, E., Pellicer, E. & Sort, J. Government (MAT2017-86357-C3-1-R), the Generalitat de Catalunya (2017-SGR-292 Electrolyte-gated magnetoelectric actuation: phenomenology, materials, and 2018-LLAV-00032), the French ANR (ANR-18-CE24-0017 Project “FEOrgSpin”) mechanisms and prospective applications. APL Mater. 7, 030701 (2019). and FEDER (MAT2017-86357-C3-1-R and 2018-LLAV-00032) is acknowledged. ICN2 is 31. Navarro-Senent, C. et al. Large magnetoelectric effects in electrodeposited funded by the CERCA program/Generalitat de Catalunya and supported by the Severo nanoporous microdisks driven by effective surface charging and magneto- Ochoa Centres of Excellence program (Spanish Research Agency, grant no. SEV-2017- ionics. ACS Appl. Mater. Interfaces 10, 44897–44905 (2018). 0706). Work at GU has been supported by SMART (2018-NE-2861), one of seven centers 32. Robbennolt, S. et al. Electric field control of magnetism in iron oxide of nCORE, a Semiconductor Research Corporation program sponsored by NIST, and nanoporous thin films. ACS Appl. Mater. Interfaces 11, 37338–37346 (2019). NSF (DMR−1905468, DMR−1828420). The PALS measurements were carried out at 33. Robbennolt, S. et al. Reversible, electric-field induced magneto-ionic control of ELBE at the Helmholtz-Zentrum Dresden-Rossendorf e. V., a member of the Helmholtz magnetism in mesoporous cobalt ferrite thin films. Sci. Rep. 9, 10804 (2019). Association. We would like to thank A.G. Attallah and E. Hirschmann for assistance 34. Mishra, R., Kumar, D. & Yang, H. Oxygen-migration-based spintronic device during PALS. emulating a biological synapse. Phys. Rev. Appl. 11, 054065 (2019). 35. Marauska, S. et al. MEMS magnetic field sensor based on magnetoelectric Author contributions composites. J. Micromech. Microeng. 22, 065024 (2012). E.M. and J.S. had the original idea and led the investigation. J.d.R., E.M. and J.S. designed 36. Read, J. et al. Oxygen transport properties of organic electrolytes and the experiments. A.Q., C.J.J., and K.L. synthesized the Co3O4 films and J.S., B.M., J.N., performance of lithium oxygen battery. J. Electrochem. Soc. 150, and J.L.C. prepared the CoN films. E.M., A.Q., and K.L. designed the sample holder to A1351–A1356 (2003). carry out the magnetoelectric measurements. J.d.R. and E.M. carried out the magneto- 37. Chen, W. et al. A new approach to the upgrading of the traditional propylene electric measurements and analyzed the data. J.d.R., A.Q., C.J.J., and E.M. performed the carbonate washing process with significantly higher CO2 absorption capacity XRD characterization and analyzed the data. A.L., L.A., and E.M. carried out the TEM and selectivity. Appl. Energy 240, 265–275 (2019). and STEM characterization and analyzed the corresponding data. A.N. performed the 38. Eichinger, G. Cathodic decomposition reactions of propylene carbonate. J. electrochemical studies. F.I. and M.C. performed the ab initio calculations. J.S., B.M., J.N., Electroanal. Chem. Interf. Electrochem. 74, 183–193 (1976). and J.L.C. carried out the transport measurements. M.O.L., M.B., and A.W. characterized 39. Leslie-Pelecky, D. L. & Rieke, R. D. Magnetic properties of nanostructured the samples by PALS and analyzed the data. All authors discussed the results and materials. Chem. Mater. 8, 1770–1783 (1996). commented on the article. The article was written by J.d.R., E.M., and J.S. 40. McHenry, M. E., Willard, M. A. & Laughlin, D. E. Amorphous and nanocrystalline materials for applications as soft magnets. Prog. Mater. Sci. 44, 291–433 (1999). Competing interests 41. Baird, W. R. & Foley, R. T. Solubility of oxygen in selected organic solvents. J. The authors declare no competing interests. Chem. Eng. Data 17, 355–357 (1972). 42. Wagner, A., Butterling, M., Liedke, M. O., Potzger, K. & Krause-Rehberg, R. Positron annihilation lifetime and Doppler broadening spectroscopy at the Additional information Supplementary information is available for this paper at https://doi.org/10.1038/s41467- ELBE facility. AIP Conf. Proc. 1970, 040003 (2018). 020-19758-x. 43. Olsen, J. V., Kirkegaard, P., Pedersen, N. J. & Eldrup, M. PALSfit: a new program for the evaluation of positron lifetime spectra. Phys. Stat. Sol. (c.) 4, Correspondence and requests for materials should be addressed to E.M. or J.S. 4004–4006 (2007). 44. Krause-Rehberg, R. & Leipner, H. in Positron Annihilation in Semiconductors Peer review information Nature Communications thanks the anonymous reviewer(s) for (Springer, Berlin, 1999). their contribution to the peer review of this work. 45. Tuomisto, F. & Makkonen, I. Defect identification in semiconductors with positron annihilation: experiment and theory. Rev. Mod. Phys. 85, 1583–1631 Reprints and permission information is available at http://www.nature.com/reprints (2013). 46. Khan, E. H., Weber, M. H. & McCluskey, M. D. Formation of isolated Zn Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in vacancies in ZnO single crystals by absorption of ultraviolet radiation: a published maps and institutional affiliations. combined study using positron annihilation, photoluminescence, and mass spectroscopy. Phys. Rev. Lett. 111, 017401 (2013). 47. Liedke, M. O. et al. Open volume defects and magnetic phase transition in Fe60Al40 transition metal aluminide. J. Appl. Phys. 117, 163908 (2015). Open Access This article is licensed under a Creative Commons 48. Wada, K. & Hyodo, T. A simple shape-free model for pore-size estimation Attribution 4.0 International License, which permits use, sharing, with positron annihilation lifetime spectroscopy. J. Phys. Conf. Ser. 443, adaptation, distribution and reproduction in any medium or format, as long as you give 012003 (2013). appropriate credit to the original author(s) and the source, provide a link to the Creative 49. Blöchl, P. E. Projector augmented-wave method. Phys. Rev. B 50, 17953–17979 Commons license, and indicate if changes were made. The images or other third party (1994). material in this article are included in the article’s Creative Commons license, unless 50. Kresse, G. & Hafner, J. Ab initio molecular dynamics for liquid metals. Phys. indicated otherwise in a credit line to the material. If material is not included in the Rev. B 47, 558–561 (1993). article’s Creative Commons license and your intended use is not permitted by statutory 51. Kresse, G. & Furthmüller, J. Efficient iterative schemes for ab initio total- regulation or exceeds the permitted use, you will need to obtain permission directly from energy calculations using a plane-wave basis set. Phys. Rev. B 54, 11169–11186 the copyright holder. To view a copy of this license, visit http://creativecommons.org/ (1996). licenses/by/4.0/. 52. Kresse, G. & Furthmüller, J. Efficiency of ab-initio total energy calculations for metals and semiconductors using a plane-wave basis set. Comput. Mater. Sci. 6, 15–50 (1996). © The Author(s) 2020 8 NATURE COMMUNICATIONS | (2020)11:5871 | https://doi.org/10.1038/s41467-020-19758-x | www.nature.com/naturecommunications