RESEARCH ARTICLE www.advelectronicmat.de A Multilevel Magnetic Synapse Based on Voltage-Tuneable Magnetism by Nitrogen Ion Migration P. Monalisha,* Zheng Ma, Eva Pellicer, Enric Menéndez, and Jordi Sort* attempt to cope with the current tech- Advanced synaptic devices with simultaneous memory and processor nological needs, a new computing capabilities are envisaged as core elements of neuromorphic computing (NC) paradigm emulating how the human for low-power artificial intelligence. So far, most synaptic devices are based on brain works, with synchronized mem- resistive memories, where the device resistance is tuned with applied voltage ory and processing units, is emerging. However, much progress is still needed or current. However, the use of electric current in such resistive devices to optimize the key components of causes significant power dissipation due to Joule heating. Higher energy brain-inspired neuromorphic comput- efficiency has been reported in materials exhibiting voltage control of ing (NC): the artificial synapses, which magnetism (VCM). In particular, voltage-driven ion motion to modulate are central to learning and memory. magnetism (magneto-ionics) is an emerging VCM mechanism that can offer Tremendous progress has been made to fabricate synaptic devices exploiting new prospects for low-power implementation of NC. In the present work, various mechanisms including ferroelec- voltage-driven nitrogen ion motion is exploited in transition metal nitride tric memory (Fe-RAM),[2,3] phase change (CoFeN) thin films (i.e., nitrogen magneto-ionics) to emulate biological memory (PCM),[4,5] resistive random- synapses. In the proposed device, distinct multilevel non-volatile magnetic access memory (ReRAM),[6,7] electrolyte- states for analog computing and multi-state storage are realized. Moreover, gated transistors (EGT),[8,9,10] etc. essential synaptic functionalities of the human brain are successfully Along with charge-based devices, synaptic devices exploiting their spin simulated. The device exhibits an excellent synapse with a remarkable degree of freedom have been studied retention time (≈6 months), high switching ratio and large endurance (≈103 ), recently.[11,12,13,14] Devices based on spin- for hardware implementation of NC. This research provides new insight into transfer-torque magnetic random mem- exploiting magneto-ionic-based synaptic devices for spin-based neuromorphic ory (STT-MRAM),[15,16] skyrmion,[17–19] systems. and spin-orbit-torque magnetic ran- dom memory (SOT-MRAM),[20,21,22] etc have been implemented in synaptic electronics. The spintronics materials 1. Introduction are appealing for their fast switching, unlimited endurance and compatibility with the complementary metal oxide semiconduc- In the present era of artificial intelligence (AI) and the Inter- tors (CMOS) based technology.[20,23] However, in these magnetic net of Things (IoT), computers based on the conventional von synapses, a large fraction of energy is dissipated in the form of Neumann architecture, where the memory and processing units heat due to the Joule effect arising from the electric currents are placed separately, face important bottlenecks in terms of at- needed to switch the memory units and poses important limi- tainable speed and energy efficiency.[1] As an alternative, in an tations in terms of energy efficiency.[24] In this context, a new type of magnetic synapse operating with voltage, with minimized P. Monalisha, Z. Ma, E. Pellicer, E. Menéndez, J. Sort flowing current, might significantly boost energy efficiency. In Departament de Física this regard, materials whose magnetic properties can be directly Universitat Autònoma de Barcelona Cerdanyola del Vallès, Barcelona E-08193, Spain controlled with voltage (through the inverse magnetoelectric ef- E-mail: monalisha.peda@uab.cat; jordi.sort@uab.cat fect), could be a good alternative.[25] J. Sort Magneto-ionics, defined as voltage control of magnetism via Institució Catalana de Recerca i Estudis Avançats (ICREA) ion (O2– ,[26,27] Li,+ ,[28,29] H+ ,[30,31] ) migration, has acquired Pg. Lluís Companys 23, Barcelona E-08010, Spain a leading role in magnetoelectric systems since it can cause The ORCID identification number(s) for the author(s) of this article massive modulation of the magnetic properties in a perma- can be found under https://doi.org/10.1002/aelm.202300249 nent way. However, one of the drawbacks of this mechanism © 2023 The Authors. Advanced Electronic Materials published by is the limited switching speed attainable at room tempera- Wiley-VCH GmbH. This is an open access article under the terms of the ture, which is still rather slow (often > ms), because ion dif- Creative Commons Attribution License, which permits use, distribution fusion is a thermally-activated process.[32] Voltage-driven H+ and reproduction in any medium, provided the original work is properly cited. ion transport has been reported to be faster (in some cases < 100 μs),[33] but H+ magneto-ionic effects can be highly sensitive to DOI: 10.1002/aelm.202300249 Adv. Electron. Mater. 2023, 9, 2300249 2300249 (1 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de environmental humidity leading to limited retention times.[30] reactive magnetron sputtering (see Experimental Details). The Recently, nitrogen magneto-ionics has proven to be also more CoFeN film thickness was fixed at 10 nm (i.e., relatively thin efficient than O2- magneto-ionics in terms of induced magne- to enhance the magneto-ionic effect).[36] All the magneto-ionic tization modulation, switching speed, and cyclability.[34] Nitro- measurements were carried out in a capacitor-like configuration gen magneto-ionics has been observed in binary and ternary at room temperature using a vibrating sample magnetometer transition metal (TM) nitrides.[31,35] Remarkably, compared to (VSM). The schematics of the device structure and the electri- the binary TM nitrides, the ternary TM nitrides often show cal connections are presented in Figure 1a. The Cu/Ti buffer a much larger magneto-ionic response due to synergetic mi- layer acted as the working electrode, while a Pt wire served as crostructural/ion diffusion and electric resistivity effects.[35] Re- the counter electrode in the proposed device structure. An an- cently, magneto-ionic-based synaptic devices have been reported hydrous propylene carbonate (PC) with dissolved Na+ and OH– using CoN, and only frequency-dependent behavior has been traces (10–25 ppm) was used as an electrolyte, which also acts shown. However, complete emulation of a wide range of synap- as an N3- ion reservoir. The working mechanism of the synap- tic functionalities is yet to be validated for the potential neuro- tic device is presented in Figure 1b. Upon voltage application, an morphic application of magnetic synapses. Additionally, a high electric double layer (EDL) is formed at the film/electrolyte inter- dynamic range, faster switching, good endurance and excellent face, which induces a large electric field perpendicular to the film retention are the important figure of merits of a synaptic de- surface. The Na+ ions dissolved in the liquid electrolyte facilitate vice for real-life technological applications. Besides, multilevel the formation of a strong electric field due to their small ionic ra- non-volatile states are crucial for analogue computing and high- dius when the working electrode is negatively biased. The strong density data storage. Due to its technological potential, a com- electric field drives N3- ions into/out of the sample. The working plete and systematic study on N3- magneto-ionic-based synaptic mechanism involves both electrostatic effect and electrochemical devices is highly desirable. reactions in the sample. With negative gating, the sample under- In this work, we propose a two-terminal (2T) magneto-ionic goes denitriding (removal of N3- ions), assisted by the negative synaptic device based on a ternary transition metal nitride electric field. However, positive gating results in sample nitriding thin film: (Co35 Fe65 )x N1–x (for simplicity, CoFeN). This mate- by re-insertion of N3- ions from the liquid, assisted by a positive rial is capable of inducing a large magneto-ionic effect com- electric field. The nitriding and denitriding processes in CoFeN pared to its binary nitride counterparts. Moreover, Co35 Fe65 is are reversible and can be accomplished multiple times by alter- well known for its very high saturation magnetization and high nating voltages of opposite polarities. magnetic anisotropy among all TM magnetic alloys at room To investigate the magnetic effects caused by N3- ion mo- temperature, thus promising for spin-based NC.[37,38] In the tion, the CoFeN thin films were electrolyte-gated at room tem- as-grown state, CoFeN is paramagnetic. However, the applica- perature and magnetic hysteresis loops were recorded in situ. tion of external voltage triggers the emergence of a net mag- All the magnetic measurements were carried out by applying netic moment, which can be highly tuned through electrolyte the magnetic field parallel to the sample surface (in-plane). The gating (i.e., inducing a transition from paramagnetic to ferro- magnetization (M) versus applied magnetic field (H) hysteresis magnetic states). A very high ON/OFF ferromagnetism switch- measurements of the as-prepared and voltage-treated (VG = - ing ratio (with MS ≈ 880 emu cm-3 ) and a high magneto-ionic 12 V) sample are presented in Figure 1c.The as-grown CoFeN speed (12.3 emu cm-3 s-1 ) are achieved at room temperature. films exhibit a virtually paramagnetic behavior, with a very small Based on this magneto-ionic performance, several neuromor- MS (≈ 15 emu cm-3 ) likely arising from the presence of resid- phic functionalities have been demonstrated. First, we have re- ual metallic Fe or Co clusters, or from off-stoichiometric re- alized multilevel non-volatile magnetized states for analog com- gions in the as-grown CoFeN films.[35] Upon voltage treatment puting and multi-state storage, using a series of programmed (-12 V for 20 min), MS increases dramatically, showing a clear pulses. Then, we successfully emulated various characteristic paramagnetic-to-ferromagnetic switching. The M-H loop of the synaptic functions such as spike amplitude-dependent plastic- voltage-treated sample shows low coercivity (65 Oe) and large ity, spike duration-dependent plasticity, and long-term poten- squareness (MR /MS ≈ 0.84), suggesting in-plane anisotropy. The tiation/depression. Interestingly, the device exhibits excellent attained value of saturation magnetization by applying VG = – figures of merit such as large endurance (≈103 ) and remarkable 12 V is MS ≈ 880 emu cm-3 . With negative voltage, M increases retention time (≈6 months) for practical NC application. Since due to the extraction of N3- ions from the sample, thereby reduc- reading can be done externally (i.e., using a magnetometer), this ing CoFeN (paramagnetic) to metallic CoFe (ferromagnetic).[34–36] 2T device enables independent reading and writing operations, Thus, CoFeN exhibits clear OFF→ON switching of ferromag- unlike other conventional 2T synaptic devices, that allow easy in- netism. Remarkably, the obtained ON/OFF ratio in CoFeN is con- tegration in crossbar arrays.[39] Finally, the working mechanism siderably higher than previous values reported for binary FeN or of the device has been probed using X-ray photoelectron spec- CoN.[36] troscopy (XPS). Our results highlight the excellent platform of- The time evolution of M, while applying a gate voltage VG = fered by magneto-ionics to mimic biological synapses for build- -12 V, is presented in Figure 1d. An in-plane magnetic field of ing artificial NC systems. 10 kOe was applied (well above the saturation magnetic field) throughout the measurement. M quickly increased during the 2. Results and Discussion first few seconds of voltage application and then saturates. The speed of magnetization was calculated to be 12.3 emu cm-3 s-1 CoFeN thin films were deposited on Ti (20 nm)/Cu (60 nm) by linearly fitting the initial part (first 10 s) of the M versus t coated [100]-oriented Si substrate at room temperature using curve, shown as an inset in Figure 1d. The high dynamic range Adv. Electron. Mater. 2023, 9, 2300249 2300249 (2 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de Figure 1. a) Schematics of the CoFeN-based synaptic device and magneto-ionic measurement setup in a capacitor-like configuration. b) Schematics illustrating the working mechanism of the nitride system with negative voltage treatment involving both electrostatic and electrochemical effects. c) The M-H hysteresis loop of the CoFeN sample showing a paramagnetic to ferromagnetic switching with voltage treatment of -12 V for 20 min. d) The M-t curve of the sample with voltage (-12 V) treatment, and the inset show the linear fitting of the M-t curve to calculate the magneto-ionic speed. (magnetization generation ratio ∼ 880 emu cm-3 ) together with XPS spectra were recorded for the as-prepared and voltage- the high magneto-ionic speed (≈12.3 emu cm-3 s-1 ) makes CoFeN treated samples, after applying VG = –12 V for 15 min. particularly promising for NC applications. Figure 2a shows the Co 2p3/2 core-level spectrum of the The mechanism underlying the magnetic switching in the as-prepared CoFeN sample, evidencing the presence of only Co3+ investigated ternary nitride was probed using XPS, where (780.6 eV) peak.[40] On the contrary, the corresponding peak in the change in valence states of the elements was ana- the voltage-treated sample reveals the presence of metallic Co0 lyzed for samples before and after voltage treatment. The (777.9 eV) along with the Co3+ peak in Figure 2b.[41] Similarly, Figure 2. XPS spectra of the CoFeN surface. The sample was treated with VG = -12 V for 15 min. a) Co 2p3/2 spectrum shows only the Co3+ peak in the as-prepared sample. b) Co 2p3/2 spectrum showing Co3+ and metallic Co0 peaks in the voltage-treated sample. c) Fe 2p3/2 shows only the Fe3+ peak in the as-prepared sample. d) Fe 2p3/2 spectrum showing Fe3+ and metallic Fe0 peaks in the voltage-treated sample. Adv. Electron. Mater. 2023, 9, 2300249 2300249 (3 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de ative gating (sweeping from 0 to -8 V, and from -8 to 0 V,) due to the aforementioned sample denitriding by the removal of N3- ions assisted by a negative electric field. Conversely, with posi- tive gating, M decreases as the N3- ions are inserted back into the sample, converting CoFe (ferromagnetic) back to CoFeN (param- agnetic). A small gap opening can be observed in Figure 3b, re- flecting partial irreversibility. However, with a faster VG sweeping fully closed loops can be obtained compromising the magnetiza- tion modulation, as shown in Figure S3 (Supporting Informa- tion). The M versus VG curve results in a loop, representing the non-volatile and virtually reversible modulation of sample mag- netization, making it suitable for the emulation of both short- term and long-term synaptic plasticity. Multilevel non-volatile states are crucial for analog computing and to increase the memory density for neuromorphic comput- ing applications.[45,46] Multilevel magnetic states have been made accessible in the CoFeN synaptic device by applying trains of neg- ative pulses (VG = -5 V, pulse duration tP = 90 s), spaced apart by 300 s, as shown in Figure 3c. We can see that during the negative gating periods, M increases due to sample denitriding. However, in the periods when voltage is stopped, M slightly decreases but retains a progressively higher value after each step, evidencing Figure 3. a) Schematics of a biological synapse demonstrating signal an intrinsic non-volatile nature. Such spontaneous partial deple- transmission from presynaptic to postsynaptic neurons. b) Variation of tion of the generated M has been observed in other systems.[34] sample magnetization as a function of VG with a sweeping speed of With successive pulses, CoFeN becomes progressively more den- 0.03 V s-1 , evidencing a non-volatile and reversible change. c) Multilevel itrided, resulting in states with higher MS . Hence, by applying non-volatile magnetic states were obtained by applying ten consecutive programming voltage pulses, we could realize multiple discrete gate pulses (VG = -5 V, tP = 90 s) spaced apart by 300 s. non-volatile states for NC applications. Synaptic plasticity can be tuned by varying the amplitude of ex- ternal stimuli. This is termed “spike amplitude-dependent plas- the Fe 2p3/2 core-level spectrum of the as-prepared sample, pre- ticity” (SADP). SADP was mimicked in the CoFeN-based synap- sented in Figure 2c, illustrates the existence of only Fe3+ peak tic device by applying a series of spikes of different amplitudes (710.6 eV).[42] Upon voltage treatment, there is the advent of (-4, -6, -8, -10, -12 V). The duration of each spike was main- metallic Fe0 peak (706.5 eV) along with Fe3+ peak in the corre- tained constant, i.e., tP = 90 s. Again, an external magnetic field sponding Fe 2p3/2 spectrum (Figure 2d).[43] Thus, Figure 2(b,d) of 10 kOe was applied throughout the measurements to mag- proves that, with negative voltage treatment, the sample under- netically saturate the generated ferromagnetic counterpart. As goes denitriding, hence reducing CoFeN to ferromagnetic CoFe, shown in Figure 4a, M increases with increasing spike ampli- in agreement with the magnetic results. Hence, the XPS study tude (more negative), due to a more pronounced paramagnetic- confirms the proposed working mechanism of the denitriding to-ferromagnetic phase transformation at a higher voltage. The process. increment of M with electrolyte-gating occurs due to the com- In the human brain, the synapse is a functional connection be- bined contribution of the electrostatic effect and electrochemi- tween neurons that allows signal transmission. The schematics cal reaction driven by the EDL. The latter involves denitriding of biological synapses, allowing signal transmission from presy- of the sample assisted by the electric field. Furthermore, the in- naptic neuron to postsynaptic neuron is depicted in Figure 3a. crement/decrement in sample M with negative/positive gating is Synapses exhibit a fascinating property called “synaptic plastic- analogous to the excitatory/inhibitory behavior of the magnetic ity” (change of synaptic weight depending on activity level), which synapse. While Figure 4a presents the excitatory behavior at dif- forms the basis for all learning and memory happening in the ferent spike amplitudes, the inhibitory behavior of the CoFeN human brain.[44] In the present study, the magnetic switching based synapse is mimicked by applying spike (VG = 10 V, tP = (from paramagnetic to ferromagnetic) in the TM nitride-based 50 s), presented Figure S4 (in Supporting Information). device has been used to imitate various synaptic functionalities. Figure 4a reveals that M increases during 90 s of spike appli- Here, M switching is considered as the synaptic weight update cation, however, after spike removal, M relaxes tending to level and pulsed electric voltages as the external stimuli. off after 1500 s. The relaxation in M represents volatile magnetic To demonstrate the neuromorphic properties of the investi- changes, that can be used to mimic the short-term plasticity in gated magneto-ionic material, we first measured the variation of the human brain. Remarkably, M retains a constant and voltage- M as a function of gate voltage by sweeping VG in a clockwise tunable value for a longer time, presenting the non-volatile mag- fashion (from 0 to -8 V, then -8 to 0 V, then 0 to18 V, and fi- netic change analogous to long-term plasticity. The resultant MS nally back to 0 V), with a sweeping speed of 0.03 V s-1 , as shown is plotted as a function of spike amplitude in Figure 4b. The MS in Figure 3b. An in-plane magnetic field of 10 kOe was applied measured immediately after spike removal (at peak) is defined as throughout the measurement. The sample M increases with neg- volatile change, while the MS measured after 2000 s from spike Adv. Electron. Mater. 2023, 9, 2300249 2300249 (4 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de Figure 4. a) Spike amplitude-dependent plasticity emulated by firing spikes of different amplitudes (-4, -6, -8, -10, and -12 V) and of the same duration (tP = 90 s). b) Change in M as a function of spike amplitude. ΔM values were measured immediately (at peak) and after 2000 s of spike removal, presenting volatile and non-volatile changes, respectively. c) Time derivative of magnetization for different spike amplitudes (the inset shows the maximum magneto- ionic speed as a function of spike amplitude). d) Retention of the device treated at different gate voltages. After electrolyte gating in liquid, the sample was stored at ambient, and the M was measured at air periodically. The initial state (non-treated sample) is denoted as “Para”. removal is defined as non-volatile change.[47] Both the volatile and The CoFeN-based synaptic device shows remarkable retention non-volatile MS contributions increase monotonically as a func- in ambient of at least six months (till we measure). M is partic- tion of VG , representing the complete emulation of SADP. ularly stable for the high voltage treated sample due to the dom- Device speed is an important figure of merit of a synaptic de- inant denitriding that causes a non-volatile change. Moreover, vice. The speed of the magnetic switching determines the speed the energy consumption of the device is ≈120 μJ spike−1 . The of the device. The rate of change of M (magneto-ionic speed) larger energy consumption is attributed to the huge device size during the first 90 s of voltage application has been calculated (≈several mm2 ), which can be scaled down by lithography pat- by plotting the dM/dt for different spike amplitudes, as shown terning (to nanoscale regime) to have energy consumption com- in Figure 4c. The speed of the device increases systematically parable to biological synapses.[9,50,15] as a function of spike amplitude, as presented in the inset. The Similarly, extending the spike duration is another means to investigated CoFeN-based synaptic device exhibits higher speed tune the synaptic connection, termed “spike duration-dependent (12.3 emu cm-3 s-1 ) at room temperature, even at a much lower plasticity” (SDDP). To mimic SDDP in the nitride-based synaptic voltage (-12 V) compared to other magneto-ionic systems, which device, spikes of different duration (30, 60, 90, 120, 150 s) and the is very promising for practical NC applications. For instance, same amplitude (-8 V) were fired to the gate terminal. A magnetic 85 nm thick CoN and FeN exhibits magneto-ionic speed of 0.7 field of 10 kOe was applied to the sample throughout the mea- and 0.1 emu cm-3 s-1 respectively, with an applied voltage of - surement to saturate M. The results are shown in Figure 5a. With 50 V.[34,48] Even thinner (5 nm) films of CoN and Co3 O4 ex- increasing the spike duration, the attained M value becomes pro- hibits lower speed of 5.9 and 1.1 emu cm-3 s-1 with higher ap- gressively larger, strengthening the synaptic connection. In fact, plied voltages of −25 and -50 V, respectively.[36,49] Furthermore, during spike application, MS increases, while after spike removal, sample retention is crucial for the hardware implementation of there is a relaxation of M, again leading to volatile and non- NC. Figure 4d shows the retention of the voltage-treated sam- volatile components of MS , similar to the SADP experiments. The ples at different amplitudes. For retention check, first the sam- MS variation with spike duration is summarized in Figure 5b. ples were voltage treated in the liquid electrolyte and immediately The MS values measured immediately after spike removal repre- removed to ambient for M measurement. Later on, the samples sents the volatile change. In turn, the MS measured after 2000 s were stored in the ambient and the M were measured periodi- of spike removal represents the non-volatile change. The volatile cally, showing the long-term stability and high retention of the change is mainly attributed to electrostatic effects and it repre- voltage-treated sample in air. sents short-term plasticity. The redistribution of N ions inside the Adv. Electron. Mater. 2023, 9, 2300249 2300249 (5 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de Figure 5. a) Spike duration-dependent plasticity mimicked by applying a series of spikes of different durations (30, 60, 90, 120, 150 s) and the same amplitude (VG = -8 V). b) The variation of ΔM as a function of spike duration for both volatile and non-volatile change. c) Several cycles of potentiation and depression were realized in the nitride-based synaptic device, illustrating excellent endurance. d) The enlarged view of a single cycle of potentiation (P) / depression (D). thin films after voltage treatment also contributes to the volatile netic states for analog computing and multi-state storage us- change of sample M. The non-volatile change, caused by the den- ing programmed voltage pulses. Furthermore, representative itriding process, represents long-term plasticity. Figure 5b shows synaptic functions such as spike amplitude-dependent plastic- that the MS increases monotonically with spike duration due to ity, spike duration-dependent plasticity, and long-term potentia- more sample denitriding with longer spikes, causing a more pro- tion/depression have been mimicked in the device. Interestingly, nounced paramagnetic to ferromagnetic phase transition. we have realized excellent endurance (≈103 ) and remarkable re- Long-term plasticity is a very important form of synaptic plas- tention time (≈6 months), turning it into a suitable device for ticity, that constitutes the basis for learning and memory in the real-life implementation in NC. The study suggests that the pro- human brain.[51] By applying a series of consecutive pulses, we posed magneto-ionic-based synaptic device with multilevel stor- could emulate the long-term potentiation (LTP) and long-term age may have great application prospects in spin-based NC. depression (LTD) in the nitride-based synaptic device. LTP was mimicked by applying a series of ten negative pulses (VG = -8 V, tP = 10 s, spaced apart by Δt = 10 s). This results in a massive 4. Experimental Section increment in M due to sample denitriding. M was then brought back to the initial state with a series of ten consecutive positive Sample Fabrication: The CoFeN thin films were deposited on Ti pulses (VG = 8 V, tP = 10 s, Δt = 10 s), emulating LTD. During (20 nm)/Cu (60 nm) coated [100]-oriented Si substrate using AJA Inter- national ATC 2400 sputtering system with a base pressure of ≈10–8 Torr. LTD/LTP the sample undergoes denitriding/nitriding with suc- High-purity Ar: N2 (50%:50%) environment with a pressure of 3×10–3 Torr cessive pulses. Figure 5c shows several cycles of potentiation (P) was maintained during the nitride film growth. The films were deposited at and depression (D) mimicked in the synaptic device. The device room temperature with a sputtering speed of 0.5 Å s-1 . One portion of the could be cycled > 103 times without undergoing any degradation, sample was masked during deposition to leave the Ti/Cu layer exposed to exhibiting excellent cyclability and endurance for practical NC ap- later make the contact and let it act as the working electrode. plications. Figure 5d shows an enlarged view of one cycle of po- Magnetoelectric Measurement: All the magnetoelectric measurements tentiation and depression. were carried out at room temperature, applying voltage in a capacitor- like configuration. A platinum wire was used as the counter electrode. An anhydrous propylene carbonate with dissolved Na+ (10–25 ppm) was 3. Conclusion used as the gate electrolyte, where metallic sodium was used to elimi- nate any traces of residual water in the electrolyte. The in situ magnetic A magneto-ionic-based synaptic device based on CoFeN, where measurements were carried out using a VSM (Micro Sense LOT-Quantum we have used voltage-induced magnetic switching to emulate bi- design) with a maximum in-plane magnetic field of 20 kOe. The M-H ological synapses, is proposed. The strong (880 emu cm-3 ) and loops were corrected for the linear slope at high fields to remove the dia- magnetic/paramagnetic contributions from the Si substrate and the VSM fast (12.3 emu cm-3 s-1 ) magneto-ionic response make the de- holder. The voltage pulses were applied using an Agilent B2902A power vice promising for NC. The insertion/removal of N3- ions, re- supply to emulate neuromorphic functionalities. sponsible for such magnetic switching, have been experimen- Structural Characterization: XPS measurements were carried out us- tally demonstrated. We have realized multilevel non-volatile mag- ing a PHI 5500 Multitechnique System (Physical Electronics), with a Adv. Electron. Mater. 2023, 9, 2300249 2300249 (6 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de monochromatic Al K𝛼 (E = 1486.6 eV) source. The XPS measurements [10] C. Ge, C. xiang Liu, Q. li Zhou, Q. hua Zhang, J. yu Du, J. kun Li, C. were carried out ex situ, immediately after voltage treatment of the sam- Wang, L. Gu, G. zhen Yang, K. juan Jin, Adv. Mater. 2019, 31, 1900379. ples, and subsequently washing them using deionized water. CasaXPS [11] Q. Zhang, Y. Zhao, C. He, Y. Huo, B. Cui, Z. Zhu, G. Zhang, G. Yu, B. software was used for XPS peak deconvolution and a Shirley background He, Y. Zhang, H. Lyu, Y. Guo, J. Qi, S. Shen, H. Wei, B. Shen, S. Wang, was used for all elements. Adv. Electron. Mater. 2022, 8, 2200845. [12] J. Liu, T. Xu, H. Feng, L. Zhao, J. Tang, L. Fang, W. Jiang, Adv. Funct. Mater. 2022, 32, 2107870. Supporting Information [13] G. W. Burr, R. M. Shelby, A. Sebastian, S. Kim, S. Kim, S. Sidler, K. Virwani, M. Ishii, P. Narayanan, A. Fumarola, L. L. Sanches, I. Boybat, Supporting Information is available from the Wiley Online Library or from M. Le Gallo, K. Moon, J. Woo, H. Hwang, Y. Leblebici, Adv Phys X 2017, the author. 2, 89. [14] J. Grollier, D. Querlioz, K. Y. Camsari, S. Fukami, M. D. Stiles, Nat. Electron. 2020, 3, 360. [15] D. Kumar, H. J. Chung, J. Chan, T. Jin, S. Ter Lim, S. S. P. Parkin, R. Acknowledgements Sbiaa, S. N. Piramanayagam, ACS Nano 2023, 17, 6261. P. M. and Z.M. equally contributed to this work. Financial support [16] X. Zhang, W. Cai, M. Wang, B. Pan, K. Cao, M. Guo, T. Zhang, H. by the European Research Council (2021-ERC-Advanced “REMINDS” Cheng, S. Li, D. Zhu, L. Wang, F. Shi, J. Du, W. Zhao, Adv. Sci. 2021, grant no. 101054687), the Spanish Government (PID2020-116844RB- 8, 2004645. C21 and PDC2021-121276-C31), and the Generalitat de Catalunya (2021- [17] R. Chen, C. Li, Y. Li, J. J. Miles, G. Indiveri, S. Furber, V. F. Pavlidis, C. SGR-00651) and MCIN/AEI/10.13039/501100011033 & European Union Moutafis, Phys. Rev. Appl. 2020, 14, 014096. NextGenerationEU/PRTR (grant CNS2022–135230) are acknowledged. [18] K. M. Song, J. S. Jeong, B. Pan, X. Zhang, J. Xia, S. Cha, T. E. Park, K. E.M. is a Serra Húnter Fellow. Kim, S. Finizio, J. Raabe, J. Chang, Y. Zhou, W. Zhao, W. Kang, H. Ju, S. Woo, Nat. Electron. 2020, 3, 148. [19] Y. Huang, W. Kang, X. Zhang, Y. Zhou, W. Zhao, Nanotechnology 2017, 28, 08LT02. Conflict of Interest [20] T. Leonard, S. Liu, M. Alamdar, H. Jin, C. Cui, O. G. Akinola, L. Xue, T. The authors declare no conflict of interest. P. Xiao, J. S. Friedman, M. J. Marinella, C. H. Bennett, J. A. C. Incorvia, Adv. Electron. Mater. 2022, 8, 2200563. [21] R. S. Yadav, P. Gupta, A. Holla, K. I. Ali Khan, P. K. Muduli, D. Bhowmik, ACS Appl. Electron. Mater. 2023, 5, 484. Data Availability Statement [22] Y. Cao, A. W. Rushforth, Y. Sheng, H. Zheng, K. Wang, Adv. Funct. Mater. 2019, 29, 1808104. The data that support the findings of this study are available from the cor- responding author upon reasonable request. [23] J. Zhou, J. Chen, Adv. Electron. Mater. 2021, 7, 2100465. [24] V. Milo, G. Malavena, C. M. Compagnoni, D. Ielmini, Materials 2020, 13, 166. [25] B. Sun, G. Zhou, L. Sun, H. Zhao, Y. Chen, F. Yang, Y. Zhao, Q. Song, Keywords Nanoscale Horiz. 2021, 6, 939. [26] D. A. Gilbert, A. J. Grutter, E. Arenholz, K. Liu, B. J. Kirby, J. A. magneto-ionics, neuromorphic computing, synaptic devices, transition- Borchers, B. B. Maranville, Nat. Commun. 2016, 7, 12264. metal nitride, voltage control of magnetism [27] A. Molinari, H. Hahn, R. Kruk, Adv. Mater. 2019, 31, 1806662. [28] M. Ameziane, R. Mansell, V. Havu, P. Rinke, S. van Dijken, Adv. Funct. Received: April 14, 2023 Mater. 2022, 32, 2113118. Revised: June 7, 2023 [29] G. Wei, L. Wei, D. Wang, Y. Chen, Y. Tian, S. Yan, L. Mei, J. Jiao, Appl. Published online: June 21, 2023 Phys. Lett. 2017, 110, 062404. [30] A. J. Tan, M. Huang, C. O. Avci, F. Büttner, M. Mann, W. Hu, C. Mazzoli, S. Wilkins, H. L. Tuller, G. S. D. Beach, Nat. Mater. 2019, 18, 35. [1] J. Backus, Commun. ACM 1978, 21, 613. [31] G. Chen, C. Ophus, A. Quintana, H. Kwon, C. Won, H. Ding, Y. Wu, [2] S.-T. Yang, X.-Y. Li, T.-L. Yu, J. Wang, H. Fang, F. Nie, B. He, L. Zhao, W.- A. K. Schmid, K. Liu, Nat. Commun. 2022, 13, 1350. M. Lü, S.-S. Yan, A. Nogaret, G. Liu, L.-M. Zheng, Adv. Funct. Mater. [32] D. B. Strukov, R. S. Williams, Appl. Phys. A 2009, 94, 515. 2022, 32, 2202366. [33] M. Huang, M. U. Hasan, K. Klyukin, D. Zhang, D. Lyu, P. Gargiani, M. [3] S. Boyn, J. Grollier, G. Lecerf, B. Xu, N. Locatelli, S. Fusil, S. Girod, Valvidares, S. Sheffels, A. Churikova, F. Büttner, J. Zehner, L. Caretta, C. Carrétéro, K. Garcia, S. Xavier, J. Tomas, L. Bellaiche, M. Bibes, A. K. Y. Lee, J. Chang, J. P. Wang, K. Leistner, B. Yildiz, G. S. D. Beach, Barthélémy, S. Saïghi, V. Garcia, Nat. Commun. 2017, 8, 14736. Nat. Nanotechnol. 2021, 16, 981. [4] S. G. Sarwat, B. Kersting, T. Moraitis, V. P. Jonnalagadda, A. Sebastian, [34] J. de Rojas, A. Quintana, A. Lopeandía, J. Salguero, B. Muñiz, F. Nat. Nanotechnol. 2022, 17, 507. Ibrahim, M. Chshiev, A. Nicolenco, M. O. Liedke, M. Butterling, [5] D. Kuzum, R. G. D. Jeyasingh, B. Lee, H.-S. P. Wong, Nano Lett. 2012, A. Wagner, V. Sireus, L. Abad, C. J. Jensen, K. Liu, J. Nogués, J. 12, 2179. L. Costa-Krämer, E. Menéndez, J. Sort, Nat. Commun. 2020, 11, [6] S. Paramanik, A. Maiti, S. Chatterjee, A. J. Pal, Adv. Electron. Mater. 5871. 2022, 8, 2100237. [35] Z. Tan, S. Martins, M. Escobar, J. De Rojas, F. Ibrahim, M. Chshiev, A. [7] F. Zhou, Z. Zhou, J. Chen, T. H. Choy, J. Wang, N. Zhang, Z. Lin, S. Quintana, A. Lopeandia, L. Costa-kr, E. Men, J. Sort, ACS Appl. Mater. Yu, J. Kang, H. S. P. Wong, Y. Chai, Nat. Nanotechnol. 2019, 14, 776. Interfaces 2022, 14, 44581. [8] Q. Wang, T. Zhao, C. Zhao, W. Liu, L. Yang, Y. Liu, D. Sheng, R. Xu, Y. [36] Z. Tan, J. de Rojas, S. Martins, A. Lopeandia, A. Quintana, M. Cialone, Ge, X. Tu, H. Gao, C. Zhao, Adv. Electron. Mater. 2022, 8, 2101260. J. Herrero-Martín, J. Meersschaut, A. Vantomme, J. L. Costa-Krämer, [9] W. Xu, S. Y. Min, H. Hwang, T. W. Lee, Sci. Adv. 2016, 2, 1501326. J. Sort, E. Menéndez, Mater. Horiz. 2023,10, 88. Adv. Electron. Mater. 2023, 9, 2300249 2300249 (7 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH 2199160x, 2023, 8, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aelm.202300249 by Readcube (Labtiva Inc.), Wiley Online Library on [27/09/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License www.advancedsciencenews.com www.advelectronicmat.de [37] G. Scheunert, O. Heinonen, R. Hardeman, A. Lapicki, M. Gubbins, [45] P. Monalisha, S. Li, T. Jin, P. S. A. Kumar, S. N. Piramanayagam, Nan- R. M. Bowman, Appl. Phys. Rev. 2016, 3, 011301. otechnology 2023, 34, 165201. [38] A. Ghemes, O. Dragos-Pinzaru, H. Chiriac, N. Lupu, M. Grigoras, D. [46] B. Zhang, W. Lv, Y. Guo, B. Wang, K. Luo, W. Li, J. Cao, Adv. Electron. Shore, B. Stadler, I. Tabakovic, J. Electrochem. Soc. 2016, 164, D13. Mater. 2022,9, 2200939. [39] Y. Van De Burgt, A. Melianas, S. T. Keene, G. Malliaras, A. Salleo, Nat. [47] P. Monalisha, S. Li, T. Jin, P. S. A. Kumar, S. N. Piramanayagam, J. Electron. 2018, 1, 386. Phys. D Appl. Phys. 2022, 56, 015302. [40] J. Li, Z. Li, F. Ning, L. Zhou, R. Zhang, M. Shao, M. Wei, ACS Omega [48] J. De Rojas, J. Salguero, F. Ibrahim, M. Chshiev, A. Quintana, A. 2018, 3, 1675. Lopeandia, M. O. Liedke, M. Butterling, E. Hirschmann, A. Wagner, [41] J. Yu, Y. Zhong, X. Wu, J. Sunarso, M. Ni, W. Zhou, Z. Shao, Adv. Sci. L. Abad, J. L. Costa-Krämer, E. Menéndez, J. Sort, ACS Appl. Mater. 2018, 5, 1800514. Interfaces 2021, 13, 30826. [42] A. Rajan, M. Sharma, N. K. Sahu, Sci. Rep. 2020, 10, 15045. [49] S. Martins, J. De Rojas, Z. Tan, M. Cialone, A. Lopeandia, J. Herrero- [43] K. Idczak, R. Idczak, R. Konieczny, Phys. B Condens. Matter 2016, 491, Martin, J. L. Costa-Kramer, E. Menendez, J. Sort, Nanoscale 2022, 14, 37. 842. [44] M. B. Kennedy, Cold Spring Harb. Perspect. Biol. 2013, 8, [50] Y. Lee, H. L. Park, Y. Kim, T. W. Lee, Joule 2021, 5, 794. a016824. [51] D. V. Buonomano, W. Maass, Nat. Rev. Neurosci. 2009, 10, 113. Adv. Electron. Mater. 2023, 9, 2300249 2300249 (8 of 8) © 2023 The Authors. Advanced Electronic Materials published by Wiley-VCH GmbH