Multi-Bit Biomemory Based on Chitosan: Graphene Oxide Nanocomposite with Wrinkled Surface

Chitosan (CS) is one of the commonly affluent polysaccharides that are attractive biomaterials as they are easily found in different organisms and are biocompatible. An environment-friendly multi-bit biomemory was successfully achieved on the basis of CS as a favorable candidate for resistive-switching memory applications. By incorporating graphene oxide (GO) into CS, the multi-bit biomemory device (indium tin oxide (ITO)/CS:GO/Ni) was obtained through the solution-processable method, which had a high current ratio among a high, intermediate, and low resistance state as well as a low SET/RESET voltage. GO acting as trapping sites in the active layer might be responsible for the biomemory mechanism. This research opens up a new avenue towards renewable and environmentally benign CS-based materials for biodegradable electronic devices.


Introduction
Resistive-switching memory devices, originated from the beneficial properties like high density, large scalability, low power consumption, and high endurance and retention performance, have emerged as promising candidates for future high-performance nonvolatile data memory [1,2]. They mostly perform binary digital data storage by encoding "0" and "1" as a high and low resistance state (HRS and LRS) in response to an applied external electric voltage, substituted for flash memory and dynamic random access memory (DRAM). Nevertheless, the memory capacity on the basis of electrical bistability is limited at 2 n . A promising method to enhance the data storage capacity is by promoting the number of conductive states in memory devices to 3 or higher [3,4]. For next-generation ultrahigh density and high-speed biomemory, multi-bit resistive-switching biomemory devices can be seen as promising building blocks.
Si-based electronic devices have dramatically extended approaches in the domain of high-speed computers, high-efficiency photovoltaic cells, and portable consumer electronics [5][6][7][8][9]. As their operating lifetime gradually becomes shorter, not only a rapid depletion of natural and nonrenewable resources, but also a generation of large amounts of electronic waste with toxic or nondegradable materials has become ubiquitous nowadays. Hence, there is a great demand of sustainable biomaterial electronics. Biological polymers offer novel opportunities for nonvolatile memory by virtue of their biocompatibility, environmental benignity together with sustainability [10][11][12][13]. Biomaterial-based passive or active components in biocompatible and biodegradable devices revolutionize state-of-the-art electronic devices and systems that currently have counted on conventional semiconductor technologies [14,15].

Materials and Methods
Multilayers of GO (black powder) were purchased from Tanfeng Tech. Inc. (Suzhou, China) when carboxylated CS (M n = 48 kg/mol and a degree of deacetylation = 90 mol%) with molecular formula (C 6 H 11 NO 4 ) n was purchased from Aladdin (Tianjin, China). For the manufacture of biomemory devices, CS:GO nanocomposite solutions (5 mg/mL) with a concocting mass ratio between chemical components (2:1, 5:1, and 10:1, w/w) were homogeneously prepared in deionized water by ultrasonic vibration for 24 h. The substrates were precleaned by a sonication-aided washing process that included the following steps: acetone for 1 h → absolute alcohol for 1 h → deionized water for 40 min → drying in an oven at 40 • C for 30 min. At ambient temperature, biomemory devices were fabricated by subsequent steps. The nanocomposite solution was spun onto the precleaned glass slide coated with an ITO layer (the square resistance R sq ≤ 6 Ω/sq, 185 nm thick, transmittance ≥84%), during which the spin casting method was employed at a speed of 5000 rpm for 60 s. The spun films were dried on a hot plate at 60 • C overnight to remove the residuals of deionized water. The thickness of the obtained CS:GO nanocomposite films (~30 nm) was determined by NanoMap 500LS (3D Profilometer aep Technology, Santa Clara, CA, USA). With a shadow mask, Ni electrodes were eventually deposited onto nanocomposite films (coated on the substrates) by means of thermal evaporation at pressure approaching 10 −5 Torr. The top metal electrodes were determined to have a thickness of 200 nm and a size of 1.0 × 1.0 mm 2 .
Techniques like scanning electron microscopy (SEM), fourier transform infrared (FTIR) spectroscopy, Raman spectroscopy, and thermogravimetric analyses (TGA) were implemented to characterize the morphological, structure, and thermal properties of CS:GO nanocomposites. Both morphological and cross-sectional profiles of the CS:GO nanocomposite films were implemented by Apreo Scanning Electron Microscope (Themoscientific; Waltham, MA, USA). FTIR spectroscopy (Foss DS 2500 Infrared Spectrometer; Munkedal, Sweden) using KBr pellets were carried out, sweeping from 400 cm −1 to 4000 cm −1 . Raman spectroscopy (Horiba Jobin Yvon; Villeneuve-d'Ascq, France) equipped with a 532 nm laser was employed to achieve the structure information of CS:GO nanocomposites, scanning from 100 cm −1 to 3200 cm −1 . Furthermore, TGA (TA Instruments; New Castle, DE, USA) was utilized to analyze thermal properties of CS:GO nanocomposites at a heating rate of 10 • C/min under N 2 of 100 mL/min. X-ray Diffraction (XRD) of GO powder was tested by X'Pert PRO XRD meter (Panalytical, Almelo, Finland) with Cu Kα radiation (wavelength = 1.540598 Å). The scan was taken in the 2θ range of 5 • -30 • with a scan step size of 0.0131303 • /mm. HR-TEM images for GO were characterized by a JEM-2100 TEM (JOEL; Tokyo, Japan) operated at 200 kV. The electrical measurements were fulfilled by a semiconductor parameter analyzer (Keithley 4200SCS; Solon, OH, USA). All electrical experiments were conducted without any device encapsulation. The device was initially scanned from 0 V to 6 V (1st sweep) and then scanned once more (2nd sweep), which was followed by a reverse scan from 0 V to −6 V (3rd sweep). Next, the succeeding step repeated it (4th sweep). The top Ni electrode was grounded as the bias voltage was applied to the bottom ITO electrode.

Characterization of CS:GO Nanocomposites
As exhibited in Figure 1a, a CS:GO nanocomposite film with ternary biomemory performance is sandwiched between an ITO bottom electrode and a Ni top electrode (i.e., ITO/CS:GO/Ni). The cross-sectional profiles of these three CS:GO nanocomposite films with a distinct weight ratio of 2:1, 5:1, and 10:1 (w/w), respectively, present a layered structure of GO (Figure 1b-d) with a thickness of roughly 30 nm. The reason for better compatibility between CS and GO is that amino and carboxyl groups in CS can be used to form hydrogen bonding with a variety of oxygen-containing groups in GO and simultaneously inhibit the aggregation of GO [23][24][25][26]. The morphological observation ( Figure 2) without agglomeration illustrates that more content of GO brings about more wrinkled nanocomposite film. The strong p-p stacking within GO is responsible for it [25,26]. It indicates that GO is well dispersed in CS, which is beneficial for the use of wrinkled surfaces for this application. For further insight into the wrinkled surface of CS:GO nanocomposite films, the GO aqueous solution (5 mg/mL) was spin-coated onto the glass substrate with ITO layer for which the SEM images of the GO film ( Figure 3) were employed to confirm that GO was responsible for the wrinkled surface of CS:GO films.     FTIR spectroscopy was applied to examine the chemical composition of GO, CS, and CS:GO nanocomposites. The spectra of CS and its blends were similar since they were polysaccharides ( Figure 4). The characteristic peaks of pure CS included a N-H bending peak at 1590-1660 cm −1 , a bridge−O stretching peak at 1154 cm −1 , and a C-O stretching peak at 1032-1094 cm −1 . The vibration of the OH group facilitated the peak at 3264 cm −1 when giving rise to the peak at 3270 cm -1 for all CS:GO nanocomposites. Moreover, the peak at 1152 cm -1 stemmed from the vibration of the saccharide structures that caused the peak at 1136 cm -1 , 1135 cm −1 , and 1152 cm -1 for CS:GO nanocomposites with 2:1, 5:1, and 10:1 (w/w), respectively. Observed from the FTIR spectrum of GO, GO had a variety of hydrophilic oxygen groups [27]. Among them, the peak at 3348 cm −1 was attributed to the asymmetric telescopic vibration peak of the hydroxyl (O-H) when the telescopic vibration of the carboxyl group (-COOH) brought about the peak at 1716 cm −1 . The stretching vibration of the undamaged carbon-carbon double bond (C=C) was attributed to the peak at 1621 cm −1 . These results suggested that both CS and GO could be spun together by this mixing system.  FTIR spectroscopy was applied to examine the chemical composition of GO, CS, and CS:GO nanocomposites. The spectra of CS and its blends were similar since they were polysaccharides ( Figure 4). The characteristic peaks of pure CS included a N-H bending peak at 1590-1660 cm −1 , a bridge−O stretching peak at 1154 cm −1 , and a C-O stretching peak at 1032-1094 cm −1 . The vibration of the OH group facilitated the peak at 3264 cm −1 when giving rise to the peak at 3270 cm −1 for all CS:GO nanocomposites. Moreover, the peak at 1152 cm −1 stemmed from the vibration of the saccharide structures that caused the peak at 1136 cm −1 , 1135 cm −1 , and 1152 cm −1 for CS:GO nanocomposites with 2:1, 5:1, and 10:1 (w/w), respectively. Observed from the FTIR spectrum of GO, GO had a variety of hydrophilic oxygen groups [27]. Among them, the peak at 3348 cm −1 was attributed to the asymmetric telescopic vibration peak of the hydroxyl (O-H) when the telescopic vibration of the carboxyl group (-COOH) brought about the peak at 1716 cm −1 . The stretching vibration of the undamaged carbon-carbon double bond (C=C) was attributed to the peak at 1621 cm −1 . These results suggested that both CS and GO could be spun together by this mixing system.   For a characterization of the layered structure, TEM, XRD together with Raman spectroscopy were performed. HR-TEM micrographs showed a two-dimensional (2D) multilayer structure of GO, demonstrated in Figure 5a,b. The selected area electron diffraction (SAED) pattern obtained from TEM is shown in Figure 5c. The formed diffuse ring pattern showed that GO was amorphous [23]. XRD pattern of GO (Figure 5d) showed the dominant peak at 10.8 • with interlayer spacing (0.82 nm) larger than pristine graphite (0.34 nm) owing to the addition of oxygen with functional groups [24]. The oxidation degree of GO could be calculated based on the XRD pattern [23,24].
For the Raman spectrum of CS (Figure 6a), the evaluation of CS predominately indicated peaks of symmetric stretches of the C-O-C bond in the 850-900 cm −1 region owing to glyosidic bonds while antisymmetric stretches of the C-O-C bond were identified in the 1060-1150 cm −1 region. The peak at 1265 cm −1 indicated the stretch of the C−OH bond as the one at 1378 cm −1 resulted from symmetrical CH 3 . Furthermore, the peak related to the symmetric bond of CH 2 was exhibited at 1459 cm −1 . The structural properties of GO and CS:GO nanocomposites were investigated through Raman spectra as well (Figure 6b). All of the samples displayed the distinctive D band (~1350 cm −1 ) and G band (~1590 cm −1 ), demonstrated in Table 1. The intensity ratio of D band to G band (I D /I G ) was in inverse proportion to the graphitization degree [28,29]. Compared with GO (I D /I G = 0.923), all the nanocomposites had higher I D /I G values, which suggested that more defects were created in CS:GO nanocomposites [30]. In the Raman spectra of GO and CS:GO nanocomposites, the 2D and D + G Raman bands from 2690 cm −1 to 2950 cm −1 were observed for all samples. It implied the presence of multilayers [31]. The 2D band was very sensitive to the stacking order of GO along the c-axis, and thus, functioned to evaluate the structural parameters of c-axis orientation [23]. The D + G band was only observed in samples with significant amounts of defects (trap sites) [32], which arose in all CS:GO nanocomposite samples. Table 1 represents the graph change in Raman intensity ratio (I 2D /I G ) against samples as well. CS:GO nanocomposites showed a higher graphene stacking as observed in the growth of the 2D band (I 2D /I G was proportional to the number of graphene layers) [30].
CS:GO nanocomposites [30]. In the Raman spectra of GO and CS:GO nanocomposites, the 2D and D + G Raman bands from 2690 cm −1 to 2950 cm −1 were observed for all samples. It implied the presence of multilayers [31]. The 2D band was very sensitive to the stacking order of GO along the c-axis, and thus, functioned to evaluate the structural parameters of c-axis orientation [23]. The D + G band was only observed in samples with significant amounts of defects (trap sites) [32], which arose in all CS:GO nanocomposite samples. Table 1 represents the graph change in Raman intensity ratio (I2D/IG) against samples as well. CS:GO nanocomposites showed a higher graphene stacking as observed in the growth of the 2D band (I2D/IG was proportional to the number of graphene layers) [30]. TGA and derivative thermogravimetry (DTG) of GO and CS:GO nanocomposites in N2 atmosphere were carried out to study their thermal stability (Figure 7). The weight loss of GO started below 100 °C upon heating, and the major weight loss of GO occurred (60%) at ~200 °C, since oxygen-containing functional groups were pyrolyzed and released gases which resulted in rapid thermal expansion of GO [33]. The TGA plot for GO revealed that there was a small residual weight of 6% after heating to 600 °C under N2 atmosphere. TGA and DTG curves for CS:GO nanocomposites were similar to that of CS with three main decomposition stages. The first stage of thermal decomposition started from 44 °C to 287 °C. This stage was originated from physically adsorbed water surface of the polymer. With a maximum peak at 318 °C in the DTG curve, the second stage occurred between 287 °C and 400 °C, and bore the final mass loss of 59.5%. It was bound up with the thermal degradation of the ether bond. Furthermore, the third stage occurred between 400 °C and 600 °C, which might be correlated with the degradation of volatile products of the material. It evidenced the good dispersion of GO in the CS films, which corresponded with the characterization of SEM, FTIR, and Raman spectroscopy.     TGA and derivative thermogravimetry (DTG) of GO and CS:GO nanocomposites in N 2 atmosphere were carried out to study their thermal stability (Figure 7). The weight loss of GO started below 100 • C upon heating, and the major weight loss of GO occurred (60%) at~200 • C, since oxygen-containing functional groups were pyrolyzed and released gases which resulted in rapid thermal expansion of GO [33]. The TGA plot for GO revealed that there was a small residual weight of 6% after heating to 600 • C under N 2 atmosphere. TGA and DTG curves for CS:GO nanocomposites were similar to that of CS with three main decomposition stages. The first stage of thermal decomposition started from 44 • C to 287 • C. This stage was originated from physically adsorbed water surface of the polymer. With a maximum peak at 318 • C in the DTG curve, the second stage occurred between 287 • C and 400 • C, and bore the final mass loss of 59.5%. It was bound up with the thermal degradation of the ether bond. Furthermore, the third stage occurred between 400 • C and 600 • C, which might be correlated with the degradation of volatile products of the material. It evidenced the good dispersion of GO in the CS films, which corresponded with the characterization of SEM, FTIR, and Raman spectroscopy.
bound up with the thermal degradation of the ether bond. Furthermore, the third stage occurred between 400 °C and 600 °C, which might be correlated with the degradation of volatile products of the material. It evidenced the good dispersion of GO in the CS films, which corresponded with the characterization of SEM, FTIR, and Raman spectroscopy.    (Figure 8a). That the current abruptly increased at a SET voltage of 1 V (V SET1 ) leaded to the device turning from HRS to an intermediate resistance state (IRS), whose process was denoted as SET1. It suddenly gave rise to a second current growth for the device switching from IRS to LRS when the applied bias rose up to 2.75 V (V SET2 ). That could be regarded as a SET2 process. Furthermore, the performance of the biomemory devices was preserved in LRS during the voltage sweeping from 0 V to 6 V again. The current, when the negative bias was swept from 0 V to −6 V, dramatically decreased at a bias voltage of −3.8 V (V RESET ). A transition from LRS to HRS for the device was displayed as a RESET process. In addition, the current proportion of HRS, IRS, and LRS (I HRS /I IRS /I LRS ) was estimated to be about 1:10 2 :10 3 during the biomemory process. whose process was denoted as SET1. It suddenly gave rise to a second current growth for the device switching from IRS to LRS when the applied bias rose up to 2.75 V (VSET2). That could be regarded as a SET2 process. Furthermore, the performance of the biomemory devices was preserved in LRS during the voltage sweeping from 0 V to 6 V again. The current, when the negative bias was swept from 0 V to -6 V, dramatically decreased at a bias voltage of -3.8 V (VRESET). A transition from LRS to HRS for the device was displayed as a RESET process. In addition, the current proportion of HRS, IRS, and LRS (IHRS/IIRS/ILRS) was estimated to be about 1:10 2 :10 3 during the biomemory process.  bias for ITO/CS:GO(10:1)/Ni was 0.8 V, 1.5 V, and -4.4 V, respectively. Control experiments were elaborately conducted when an active CS layer was sandwiched between ITO and Ni electrodes (ITO/CS/Ni). No hysteresis effects were observed in I-V characteristics of ITO/CS/Ni. It was evident that the ternary resistive-switching behaviors of CS:GO nanocomposites were originated from GO. Under ambient condition, retention tests were performed on ITO/CS:GO/Ni in LRS, IRS, and HRS at a reading voltage of -0.1 V (Figure 11). During the retention time of 10 4 s, the resistance of CS:GO nanocomposite films in LRS, IRS, and HRS (RLRS, RIRS, and RHRS) continuously kept stable without any degradation. These biomemory devices demonstrated the capacity of multi-bit nonvolatile random access memory. In an effort to further the conduction mechanism of the ternary resistive-switching behavior achieved from ITO/CS:GO/Ni devices, I-V profiles were plotted in a log-log scale ( Figure 12). In the low voltage region of HRS, a linear fitting (slope of ~1) for the double logarithmic scale indicated an In an effort to further the conduction mechanism of the ternary resistive-switching behavior achieved from ITO/CS:GO/Ni devices, I-V profiles were plotted in a log-log scale ( Figure 12). In the low voltage region of HRS, a linear fitting (slope of~1) for the double logarithmic scale indicated an Ohmic conduction behavior. Nevertheless, the log-log curves following a quadratic nature were deviated to realize a higher slope region (~2) after increasing the applied voltage. Since CS is an insulator in the native state, the injected charge carriers were believed to be gradually captured in the trap centers of GO up to saturation. The logI-logV curve for LRS (a slope of 1) transparently showed an Ohmic conduction behavior. It was consistent with the formation of a conducting path in the device after the SET2 process. In general, the features in the trap controlled space charge limited conduction (SCLC) model followed the correlation I ∝ V n , where V was denoted as an applied bias between the two electrodes and n was denoted as a positive number. The aforementioned conclusions were further corroborated upon application of both Poole-Frenkel and Schottky emission mechanisms to fit the resistive-switching behavior. The plots of lnI ∝ V and lnI ∝ V 1/2 before transition to Ohmic conduction were not linear and showed parabolic behavior. It supported that the ternary resistive-switching mechanism was based on the SCLC mechanism with GO acting as trap centers in the CS:GO matrix. applied bias between the two electrodes and n was denoted as a positive number. The aforementioned conclusions were further corroborated upon application of both Poole-Frenkel and Schottky emission mechanisms to fit the resistive-switching behavior. The plots of lnI ∝ V and lnI ∝ V 1/2 before transition to Ohmic conduction were not linear and showed parabolic behavior. It supported that the ternary resistive-switching mechanism was based on the SCLC mechanism with GO acting as trap centers in the CS:GO matrix.

Conclusions
Unlike state-of-the-art Si-based devices, CS-based nanocomposites blended with GO by means of the solution-processable method were seen as the biodegradable layer for multi-bit biomemory. It had a remarkable potential for nonvolatile biomemory, with a large current proportionality IHRS/IIRS/ILRS (1:10 2 :10 3 ), excellent cycling performance (>100 cycles), and long data retention characteristics (>10 4 s) at ambient temperature. By combining the nature of CS and the carrier

Conclusions
Unlike state-of-the-art Si-based devices, CS-based nanocomposites blended with GO by means of the solution-processable method were seen as the biodegradable layer for multi-bit biomemory. It had a remarkable potential for nonvolatile biomemory, with a large current proportionality I HRS /I IRS /I LRS (1:10 2 :10 3 ), excellent cycling performance (>100 cycles), and long data retention characteristics (>10 4 s) at ambient temperature. By combining the nature of CS and the carrier transport information extracted from double-logarithmic I-V characteristics, GO acting as trapping sites in the CS:GO nanocomposite layer might be responsible for the ternary biomemory mechanism. This paper shows the feasibility to fabricate multi-bit biomemory devices based on CS:GO nanocomposite, which paves the way for next-generation green data-storage devices and biointegrated electronics.