Bioresorbable Resistive Switching Device Based on Organic/Inorganic Hybrid Structure for Transient Memory Applications

Biocompatible and biodegradable resistive random‐access memory devices using Mg/agarose/Al2O3/Mg‐based organic/inorganic structures are reported, showing nonvolatile bipolar resistive switching memory behavior. The organic/inorganic‐based hybrid active layer has large working windows (≈106) and is highly stable up to 200 continuous sweeps. The device can be also tuned into multilevel memory by varying the compliance current from a few microamperes to several milliamperes. The formation of metallic filaments inside the active layer during SET and RESET stages using an X‐ray photoelectron spectroscopy depth profile and magnesium (Mg) metallic ions that penetrate the agarose layer is confirmed. For the bioresorbable test in both deionized water and phosphate‐buffered saline solution, the erasing time of whole devices can be adjusted by using an ultrathin ALD‐grown Al2O3 film.


Introduction
The world is increasingly adopting artificial intelligence and a digital economy, which demands state-of-the-art data storage DOI: 10.1002/aelm.202300759 and memory processing.However, conventional computing systems have limitations in size, speed, density, and environmental friendliness. [1]Neuromorphic computing, also called brain-inspired computing, is believed to be one of the most promising strategies to overcome the existing von Neumann digital computing paradigms and eliminate a delay in computational processing between memory and computer by its parallel processing and self-learning.This technology could bring enormous benefits in terms of speed, latency, and low-power operation while managing a large amount of data and achieving previously unattainable performance.So far, memristor (memory + resistor) devices have received significant attention as new electronic memory devices for neuromorphic computing.Various appropriate structures with different mechanisms have been examined for this device, mainly including ferroelectric random-access memory (FeRAM), resistive random-access memory (RRAM), spin-transfer torque magnetic random-access memory (STT-RAM), and so on. [2]Among them, RRAM, with its metal/insulator (active layer)/metal structures, offers advantages like long endurance, high density, fast switching, and low power operation. [3,4]oth organic and inorganic materials can be used as the insulating layer in RRAM devices.Inorganic materials such as hafnium oxide (HfO 2 ), titanium dioxide (TiO 2 ), tantalum oxide (Ta 2 O 5 ), and aluminum oxide (Al 2 O 3 ) are commonly used due to their well-known resistive switching characteristic, simple manufacturing, low cost, high packing density, as well as the ability to compatible with the complementary metal oxide semiconductor technique (CMOS).However, they may exhibit variability, nonuniformity, limited endurance, and require complex deposition techniques.Besides, these inorganic materials can create electronic waste due to their nondegradable characteristics, and can also harm the environment if left exposed for a longer duration. [5]To address these challenges, organic materials-based transient RRAM has been proposed, utilizing biocompatible and biodegradable materials as the active layer. [6]13] Besides, this bioresorbable nature is advantageous for medical implants and temporary electronic devices, as it eliminates the need for additional removal procedures. [14,15]Moreover, selecting appropriate biocompatible and biodegradable metal electrodes and active layer materials is crucial.Metals like Mg, W, zinc (Zn), molybdenum (Mo), and iron (Fe) are used as top or bottom electrodes since they can dissolve easily in various environments at controllable rates. [16]Mg, in particular, stands out as a promising metal electrode due to its electrochemical activity, low melting point, abundance, and biocompatibility for various research fields including batteries, supercapacitors, and memory applications. [17,18]ere, we demonstrated the implementation of a transient RRAM consisting of a Mg/agarose/Al 2 O 3 /Mg hybrid structure.The natural biomaterial agarose is proposed here as an active layer for the RRAM devices.Agarose is a natural polymer derived from seaweed (red algae) and is composed of D-galactose and 3,6-anhydro-L-galactose (molecular formula: C 24 H 38 O 19 ). [19]t has high gel strength at low concentrations and transparency and is approved as a harmless and safe substance. [20]In particular, we explored two types of active layers with only inorganic and organic/inorganic structures.The organic/inorganic active layer device shows a higher on/off ratio of ≈10 6 between a low resistance state (LRS)/high resistance state (HRS) compared to a solely inorganic active structure and has a very low operating voltage of ≈2.29 V. We also confirm that the filament mechanism involves Mg metal oxidation as seen in XPS depth profile analysis.Hence, we successfully demonstrated the bioresorbable properties of these devices under various conditions.

Results and Discussion
In order to fabricate a hybrid structure of an organic/inorganic film as an active layer for resistive switching (RS), the electrolyte solution viscosity is shown in Figure S4a (Supporting Information).The spin-coated agarose film shows a smooth surface as shown in Figure S4b,c (Supporting Information).The agarose film has a uniform thickness of ≈400 nm as shown in Figure S4d (Supporting Information).Figures 1a, S4e (Supporting Information) show a Mg/agarose/Al 2 O 3 /Mg RRAM crossbar array device, which consists of four layers including Al 2 O 3 , agarose, and Mg top and bottom electrodes.Current versus voltage (I-V) was measured at room temperature to investigate the performance of the organic/inorganic device in Figure 1b-h.In all cases, the positive voltage was applied to the top Mg electrode which was responsible for the SET process or LRS behavior, whereas negative voltage was applied to the bottom Mg electrode, which was responsible for a RESET process or HRS behavior in the device (Mg/agarose/Al 2 O 3 /Mg) as shown in Figure 1a.To set the forming process, the positive bias of (0 V → 7 V → 0 V) was swept across the top and bottom electrodes at the initial stage.The device was in HRS up to ≈2 V and converted from HRS to LRS at 3.94 V in Figure 1b.Then, this status came back to HRS under the negative sweeping.The forming process required a higher SET voltage than the next consecutive sweeps because of the agarose and Al 2 O 3 insulating properties. [11]In Figure 1c, both LRS and HRS are retained by the device for more than over 2 × 10 4 s with large memory windows (on/off ratio) of up to 10 6 by applying a continuous DC bias of 0.1 V.In order to confirm the cycle-tocycle variation in the LRS and HRS, the device was tested under 200 cycles, and it was under a compliance current of 10 −4 A to avoid a dielectric breakdown.We found that our device had highly stable behavior for the SET/RESET switching stages as shown in Figure 1d.The SET process was very sharp during the transition period whereas the RESET process frequently exhibited a stepwise current transition.This could be attributed to the rupture of thicker and multiple conducting filaments during filament formation. [21]The switching mechanism in our device involves the movement of Mg +2 within the polymer matrix, having different defect densities leading to the formation and dissolution of conductive filaments.The distribution of filaments and density of trap states can vary across the active area, resulting in variations in SET/RESET voltages.Also, the SET/RESET variation during the 200 sweeps could be attributed to stochastic dynamics of filament formation or rupturing depth.The endurance of devices had a high current ON/OFF ratio of ≈10 6 as shown in Figure 1e.[24][25] The endurance behavior of RRAM devices was usually altered based on device configurations and applied parameters such as an organic/inorganic structure, step voltage, stop voltage, and compliance current. [26,27]The cycle-to-cycle variation in the R LRS /R HRS was investigated in Figure 1f and found to be very stable up to 200 continuous sweeps with a ratio of ≈10 6 .A statistical distribution of the switching voltages (V SET /V RESET ) was directly extracted from the endurance sweeps (Figure 1d) and corresponding resistance values (R LRS /R HRS ) were taken at ±0.1 V as shown in Figure 1 g,h, respectively.The power consumption of the device was the major contribution due to the switching voltages and resistance during different states.The mean (μ) values of V SET and V RESET were 2.29 and −1.75 V with a standard deviation () of 0.34 and 0.25, respectively, as depicted in Figure 1g.Similarly, the mean (μ) values of R LRS and R HRS are 9.14 × 10 3 Ω and 1.56 × 10 9 Ω with standard deviation () 278.78 and 1.03 × 10 9 , respectively, as shown in Figure 1h.These extracted results meet low power consumption for our RRAM devices. [28]The correlation between the organic/inorganic hybrid structure and an only inorganic structure was investigated by fabricating a Mg/Al 2 O 3 /Mg device.The Mg/Al 2 O 3 /Mg device with 200 continuing I-V cycles had a memory window of ≈10 3 (Figure S6a,c, Supporting Information), which was lower than Mg/agarose/Al 2 O 3 /Mg (≈10 6 ).The lower memory windows in an inorganic structure mainly come from the ultralow thickness of Al 2 O 3 and the existence of oxygen vacancy-based filaments. [29]ith the existence of these vacancies, the conductive filament is difficult to remove completely after biasing a negative voltage and the distance between the broken filament is too small, bringing about a low on/off ratio. [30]The increase in the memory window of a Mg/agarose/Al 2 O 3 /Mg device is mostly related to the contribution of the insulating nature and low electronic conductivity of the polymer matrix. [31]Therefore, the memory windows of Mg/Al 2 O 3 /Mg are much smaller than a Mg/agarose/Al 2 O 3 /Mg structure device.The LRS and HRS retention of a Mg/Al 2 O 3 /Mg device is ≈1 × 10 4 s under 0.1 V (Figure S6b, Supporting Information) which was lower than a Mg/agarose/Al 2 O 3 /Mg structure device.Also, the mean (μ) V SET and V RESET values of a Mg/Al 2 O 3 /Mg device are 3.05 and −1.86 V, respectively, as shown in Figure S6d (Supporting Information).The increase in V SET and V RESET may be due to the low ionic conducting nature of the Al 2 O 3 layer. [32] 2a, the lower voltage (region A) has a slope of ≈0.9 which follows the Ohm's law linear equation due to thermally generated minority carrier concentration. [33]A slope of ≈1.9 is obtained prior to filament formation which results in the presence of the space charge limited conduction (SCLC) mechanism in region B. [34] The SCLC region comes from trap-assisted Mg 2+ ions movements inside the thermally generated traps of agarose and Al 2 O 3 thin film. [35]he region C with slope ≈0.9 exhibits Ohmic conduction, which confirms the metallic filament during the LRS, whereas the negative slope demonstrates the broken filament process which can be caused by the movement of oxygen ions in this process.Figure 2b exhibits the double logarithmic I-V plot when the polarity of bias voltage was reversed during the RESET process.In the A region, the slope is ≈0.9 which is attributed to Ohmic conduction which was continued from the previous SET operation.When the bias voltage exceeds a certain threshold the electrochemical reaction and joule heating assist the filament to rupture at weak pinhole points (region B) leading to a dissolved filament. [36]A slope of ≈2.7 indicates an SCLC-driven trap filled with limited conduction as few charges still get trapped inside the thermally generated defects. [37,38]Further increase in voltage after dissolution of the filament, there is no rise in current during the reset process.When the biased voltage is decreased to a lower voltage, the provided power is inadequate to cause any change in the device state.Therefore, the device stays in an HRS with a slope of roughly 1.4 as shown in region C. [39] Figure 2c depicts the 3D layout diagram of the conduction mechanism during the SET and RESET process of the device.Probably, there are two interfaces, one at the top electrode with agarose and another at the bottom electrode with Al 2 O 3 which forms during the device fabrication process.The interface at the bottom Mg electrode and Al 2 O 3 layer is oxidized due to the initial oxidation of oxygen and Al precursors inside the ALD chamber as shown in Figure S7 (Supporting Information), while the top electrode is oxidized due to a reaction between the Mg and oxygen functional groups such as ─OH groups that exist in the agarose thin film. [35]In the SET process, a positive voltage sweep is employed to the top electrode which oxidizes the Mg electrode from Mg to Mg 2+ (Mg → Mg 2+ + 2e − ).These oxidized ions start diffusing toward the bottom electrode owing to the continuous electric field existence in that area.These diffused Mg 2+ ions are reduced to Mg by the electrochemical process (Mg 2+ + 2e − → Mg) at the bottom electrode and they initialize the filament formation process.This process continuously takes place until Mg makes a metallic filament connecting both the top and bottom electrodes.At the same time, the migration of oxygen ions also plays a vital role during the SET and RESET process because when an electric field is applied, the oxygen ions that exist in the Al 2 O 3 layer migrate in the opposite direction to the Mg ions because of an opposite charge.This phenomenon can disrupt or affect the Mg filament formation process, resulting in a slight current decrease during the SET process.The precise mechanism of filament generation and disintegration in hybrid RRAM is not well understood and is continuously being researched. [40]The instant Mg complete filaments formation between the top and bottom electrode causes a sudden rise in current at 3.94 V that was observed in the I-V curve which indicates the LRS and a slope of ≈0.9 was also obtained in a double logarithmic I-V curve to support metallic filament formation inside the device as shown in Figure 2a.During the RE-SET process, a reverse polarity was applied on the top electrode which increases the current during the linear region of the fitted I-V curve (slope A ≈0.9) showing that the device still maintains its metallic filament behavior.After applying a negative voltage for a while, the electrochemical oxidation of Mg atoms starts and moves toward the top electrode.The continuous electrochemi-cal oxidation of these filament atoms (Mg) leads to the rupture of filaments at the weak point.Also, due to a rise in the current, it creates joule heating around the filaments, which supports the rupturing process.The rupture of filaments leads to a sudden decrease in current as can be seen in Figure 2b, and the device follows a slope of ≈2.7 and ≈1.4 which belongs to the trap-assisted SCLC conduction process in HRS.To further demonstrate evidence for oxidized Mg 2+ ions during filament formation, we used XPS depth analysis via Argon (Ar) ion etching.Figure S8a (Supporting Information) depicts the Mg 2p spectra depth profile of the agarose/Al 2 O 3 -based RRAM device as a function of etching thickness from the top surface of the Mg electrode to the interface area of the Mg/agarose.As shown in Figure 2d, with increasing etching time from 0 to 400 s, O 1s can be fitted using two components at binding energies 529.48 and 531.68 eV.These peaks in binding energies are referred to as MgO and MgO 2, respectively. [41]This can be confirmed from Mg 2p XPS spectra, which were fitted using two components with binding energies of ≈48.38 and 49.48 eV as shown in Figure 2e.With increasing etching time from 400 s as shown in Figure S8b (Supporting Information), the Mg atomic percentage decreased and C 1s atomic percentage increased.This indicates that the analyzed layer region is agarose, which leads to Mg─O peak enhancement between Mg and agarose.As a result, Mg is oxidized and forms a MgO x layer by reacting with the oxygen functional group inside agarose.These results clearly demonstrate the formation of MgO x which is shown in Figure 2c between Mg and agarose.From O 1s spectra after increasing the etching time to 600 s, an extra peak appears at a binding energy 530.98 eV, and this peak may be related to carbonate species (i.e., O─C and O = C bonds).Furthermore, the concentration of MgO 2 bonds increased, which is confirmed by the increasing Mg component in MgO 2 as presented in Figure 2e.However, the etching time increased to 1000, and for 1100.98 s, the atomic percentages of Mg and C 1s decreased, while O 1s and Al 2p percentages increased.Regarding the O 1s peak, we can see clearly that the MgO component disappeared and MgO x peak intensity increased.At 1100.98 s, from Figure S8b (Supporting Information), the Mg atomic percentage is highly decreased which may be related to the involved oxygen ions during the filament formation process and this can be conferred by decreasing MgO x peak intensity.Moreover, two new peaks with binding energies of 533.51 and 534.36 eV were formed, which may be related to Al─O and H─O bonds [42,43] as depicted in Figure 2d.This indicates the diffusion of Al atoms from the Al 2 O 3 layer into the polymer.This is also confirmed by the formation of a new peak at a binding energy of ≈46.64 eV in the Mg 2p , which may be associated with the formation of an Al─Mg bond.This result agrees with that observed by Liang et al. [44] for MgAl 2 O 4 .The XPS depth profile analysis and individual elemental spectra are also in accordance with the conduction mechanism explained in Figure 2c.We also prepared agarose-based RRAM with different configurations and electrodes to experimentally confirm the switching phenomenon.When the agarose was sandwiched between the high work function metal (inert) electrode (Au), it had no switching characteristics as shown in Figure S9a (Supporting Information).As Au is a noble metal with high electrode potential (1.40 eV), low reactivity, and cannot be oxidized easily, also there was an insufficient number of thermally generated charge carriers inside the active layer to start the conduction in the device.The I-V curve witnessed no switching behavior because it does not create any oxidized interface between the electrode and organic layer as shown in Figure S9a (Supporting Information).When an asymmetric structure was fabricated by changing the top Au electrode to Mg, we observed bipolar resistive switching behavior as shown in Figure S9b (Supporting Information).From these two structures, we confirmed that, in the case of Mg as a top electrode the filament is formed due to oxidation and there is a reduction of the top Mg electrode.Furthermore, to observe the difference in the I-V curve, agarose/Al 2 O 3 was inserted between Mg as a top and Au as a bottom electrode and found similar bipolar resistive switching behavior as shown in Figure S9c (Supporting Information).These results indicate that the Mg electrode plays an important role during the conductive formation process.
Additionally, we explored the multilevel memory behavior of an organic/inorganic structure by allowing the device to carry different compliance currents within the same cell.The device can work for a compliance current ranging from a few μA to several mA as shown in Figure 3a.The compliance current applied during the SET process (LRS) can also be tracked to the same current in the RESET process before the device goes to HRS.The device maintained all the multilevel states for up to 10 3 s without any overwhelming degeneration as depicted in Figure 3b.The multilevel conduction state is an extremely crucial aspect for a neuromorphic device to realize high-density memory storage.Figure 3c demonstrates nonvolatile memory characteristics with repeated reading, programming, reading, and erasing sequences applying a drain voltage of 0.1, 4, 0.1, and −4 V, respectively.To check cell-to-cell variation in the V SET /V RESET , ten cells with a compliance current of 10 −4 A in the same device were measured.Figure S10a,b (Supporting Information) shows that all the cells exhibit similar I-V characteristics with ±0.9 V SET voltage variation and ±1 V RESET voltage variation with an on/off ratio of ≈10 5 .However, in our fabricated device, we observed that the I-V behavior of Mg/agarose/Al 2 O 3 /Mg changes after being stored under ambient conditions for over four months due to humidity and moisture absorption as shown in Figure S10c (Supporting Information).A biodegradability test of the device was conducted to evaluate the transience and implantable bioelectronics properties of the Mg/agarose/Al 2 O 3 /Mg device.We immersed the device in DI water at room temperature and monitored the time-dependent dissolution of transient electrodes in different time intervals as depicted in Figure 4a for the initial time.After 30 min of immersion, the top Mg electrode began reacting with the DI water and was consumed by a hydrolysis process (Equation S1, Supporting Information) as shown in Figure 4b.After 4 h (shown in Figure 4c), the bottom Mg electrodes were almost dissolved due to the Al 2 O 3 layer, which serves as an encapsulation layer to protect the bottom electrodes.Figure 4d shows that the bottom electrode (Mg) completely disappeared in DI water after 6 h as well.The same condition was adapted to further test the transience characteristics of this device in a PBS solution of pH 7.4 and Figure S11 (Supporting Information) shows the whole degradation process with respect to time after 3 h.To have better controllability over the degradation time, the Al 2 O 3 layer with different thicknesses was deposited on the top electrode and immersed in DI water and PBS solution as shown in Figures S12, S14, S16, and S18 (Supporting Information), respectively.In terms of 8 nm Al 2 O 3 , the degradation time between DI water and PBS solution is 24 h and 5 h 30 min, respectively.This device operates stable regarding I-V curve in DI water and PBS solution for 10 min and 5 min as shown in Figures S13 and S15 (Supporting Information), respectively.When we increase the thickness of the Al 2 O 3 layer from 8 to 17 nm, the degradation time between DI water and PBS solution also increases from 24 h and 5 h 30 min to 48 h and 14 h, respectively.This device works well in terms of I-V curve in DI water and PBS solution for 15 min and 10 min, as shown in Figure S17 and S19 (Supporting Information), respectively.A large variation in degradation and operation time was observed, which was expected for a controllable transient behavior.The performance comparisons of bipolar nonvolatile RRAM between the Mg/agarose/Al 2 O 3 /Mg device (as reported in this study) and another biomaterial RRAM device are summarized in Table S1 (Supporting Information).As per the literature survey and the table, the HRS/LRS ratio of the Mg/agarose/Al 2 O 3 /Mg device is the highest among the transient devices listed.In terms of endurance property, the Mg/agarose/Al 2 O 3 /Mg device exhibits very stable behavior compared with the Ag/pectin/FTO, Mg/gelatin/W, and other biomaterial-based devices.The retention time of the agarose device is comparably stable as compared with the other devices.Our results suggest that an organic/inorganic hybrid structured device could be used in future transient memory, hardware security, and implantable bioelectronics applications.

Conclusion
In summary, a hybrid structure device based on agarose/Al 2 O 3 was investigated for a biocompatible and biodegradable RRAM application.The Mg/agarose/Al 2 O 3 /Mg device can maintain up to 200 cycles and showed stable date retention over 2 × 10 4 s under 0.1 V. Furthermore, it operates at a low voltage and has a high on-off ratio of ≈10 6 at −0.1 V.Moreover, the hybrid structure device was able to operate in multilevel states with longterm retention of ≈1000 s, demonstrating the feasibility of highdensity memory applications.The interface between the top electrode and agarose film was analyzed and confirmed the switching mechanism of the devices.The transient behavior of our hybrid structure device dissolved completely in DI water and PBS solution due to the biodegradable nature of the Mg electrodes and agarose/Al 2 O 3 active layer.These results suggest that an organic/inorganic hybrid structure RRAM has potential prospects for applications in biomedical, transient electronics, and secure memory systems.

Experimental Section
Preparation of Agarose Solution: The agarose solution (0.1% w/v) was prepared by dissolving 0.1 g of agarose powder in 10 mL of Tris Acetate-EDTA buffer 10x (TAE buffer) (Sigma-Aldrich) and continuously agitating at 1000 rpm on a hot plate at 150 °C overnight./Mg RRAM devices were fabricated on a treated glass substrate that was cleaned with acetone and 2-propanol for 30 min during the ultrasonication and then dried using a nitrogen (N 2 ) gun.The surface of the glass was further treated with plasma cleaner for 5 min to achieve hydrophilicity.The bottom electrode with a 200 nm layer of Mg was defined by a shadow mask in an e-beam evaporator.A 5 nm thick layer of Al 2 O 3 which was confirmed by ellipsometry was then deposited over the Mg bottom electrode using atomic layer deposition (ALD) at a 267 °C temperature with trimethylaluminum (TMA) and oxygen gas as precursors.The 400 nm agarose layers were then spin-coated at 3000 rpm for 60 s and left on a hot plate at 80 °C for 10 min for drying and leftover solvent evaporation.Finally, the top electrode Mg of 200 nm was patterned by using a shadow mask using an e-beam evaporator as exhibited in Figure S1  devices utilizing ALD at 100 °C temperature with TMA and oxygen gas as precursors as demonstrated in Figure S2 (Supporting Information), respectively.Besides, Mg/Al 2 O 3 /Mg RRAM devices also were made on a treated glass substrate that was cleaned with acetone and 2-propanol for 30 min during ultrasonication and then dried using an N 2 gun.The surface of the glass was further treated with plasma cleaner for 5 min to achieve hydrophilicity.The bottom electrode with a 200 nm layer of Mg was defined by a shadow mask in an e-beam evaporator.A 5 nm thick layer of Al 2 O 3 which was checked by ellipsometry was then deposited over the Mg bottom electrode using ALD at a 267 °C temperature with TMA and oxygen gas as precursors.Lastly, the top electrode Mg of 200 nm was deposited by using a shadow mask using an e-beam evaporator as exhibited in Figure S3 (Supporting Information).The optical transmittance of the agarose film on a glass substrate displayed high transparency (≈78%) in the visible-light region from 390 to 700 nm as indicated in Figure S5 (Supporting Information).
Film and Device Characterizations: The electrical performance of the devices was determined by biasing top electrodes at room temperature and pressure using a probe station and Keithley 4200A.DC characteristics were measured using the voltage -sweep mode.Surface morphological images were recorded using an atomic force microscope (AFM; SPA-400, SEIKO) in tapping mode.The thickness of the polymer was measured by using Field-emission scanning electron microscopy (FE-SEM) HITACHI (HITACHI SU-8010).The thickness of the Al 2 O 3 layer was characterized by utilizing a spectroscopic ellipsometer (Alpha-SE).The optical transparency of the agarose film was investigated by using an Agilent 8453 UV-Vis spectrophotometer.X-ray photoelectron spectroscopy (Thermofisher Nexsa) was used to analyze the chemical elements of the device.

Figure 1 .
Figure 1.a) 3D schematics of the Mg/agarose/Al 2 O 3 /Mg crossbar array device structure and memory function, b) I-V characteristics of Mg/agarose/Al 2 O 3 /Mg device under forming process and typical sweep, c) LRS and HRS retention at I CC = 10 −4 A, d) 200 continue I-V sweeps, e) on-off ratio of Mg/agarose/Al 2 O 3 /Mg at −0.1 V, f) resistance-cycle number curve of Mg/agarose/Al 2 O 3 /Mg RRAM device under positive and negative bias voltage of ±0.1 V, g) cumulative probability of SET and RESET voltage of 200 sweeps, and h) cumulative probability of LRS and HRS of 200 sweeps.
Furthermore, the mean (μ) values for R LRS /R HRS of a Mg/Al 2 O 3 /Mg device are 1.49 × 10 4 Ω and 1.13 × 10 8 Ω (Figure S6e, Supporting Information), which are lower than the values of a Mg/agarose/Al 2 O 3 /Mg structure device.The performance of the Mg/Al 2 O 3 /Mg RRAM was further enhanced by forming a hybrid structure as depicted in previous sections and allowed us to investigate its mechanism for insight into performance by combining one more layer to generate a hybrid structure.The switching mechanism of an Mg/agarose/Al 2 O 3 /Mg conductive bridge RAM (CBRAM) can be understood by referring to the double logarithmic plot (log I-log V) as shown in Figure 2a,b.The hybrid structure has electro-migration of Mg 2+ ions in the agarose and Al 2 O 3 thin film under a continuous electric field.During the positive voltage as in Figure

Figure 2 .
Figure 2. a) Double logarithmic I-V curve of Mg/agarose/Al 2 O 3 /Mg device, illustrating the slope values of linear fitted curve at positive bias, b) double logarithmic I-V curve of Mg/agarose/Al 2 O 3 /Mg device, illustrating the slope values of linear fitted curve at negative bias, c) schematic presentation of resistive switching mechanism in LRS and HRS, Fitted XPS raw data for d) O 1s spectra, and e) Mg 2p spectra.

Figure 3 .
Figure 3. a) Multilevel storage of Mg/agarose/Al 2 O 3 /Mg device under varying SET compliance currents from 0.008 to 10 mA, b) retention characteristics (10 3 s) with the corresponding increasing SET I CC , and c) electrical response of Mg/agarose/Al 2 O 3 /Mg memory that repeats reading, programming, reading, and erasing sequence by applying a drain voltage of 0.1, 4, 0.1, and −4 V, respectively.

Figure 4 .
Figure 4. Evolution images of the Mg/agarose/Al 2 O 3 /Mg device in DI water: a) initial time, b) 30 min, c) 4 h, and d) 6 h at room temperature.