1. IntroductionThe storage of ever-increasing amounts of information is an inevitable problem due to rapid advancements in science and technology [1]. When the amount of data to be stored expands to the zettabyte magnitude, higher requirements will be placed on memory devices [2]. So far, numerous novel devices based on new mechanisms, materials, and structures have been developed to respond to the limits of Moore’s law and solve the electric leakage problem in traditional silicon-based memory devices [3,4,5,6]. Among them, organic resistive memory devices have become highly anticipated due to their low costs, ease of processing, excellent scalability, and flexibility [7,8,9]. The active layers of organic memory devices range from simple small molecules [10], polymers [11], complexes with various ligands and coordination centers [12,13], and other biomolecules [14]. Metal complexes, especially transition metal complexes, show great potential in resistive memory devices due to the range of options of central metals and ligands with different functions [15]. For instance, Goswami and co-workers [16] found that rhodium coordinated with three azopyridine ligands that lacked memory properties and exhibited non-volatile memory behavior with a large ON/OFF current ratio of about 103. Similar results were obtained when they changed the metal to iridium [17]. Liu and co-workers [18] synthesized three kinds of conjugated polymeric materials containing cationic Ir(III) complexes in the main chain by changing the ligand structure of the iridium complex unit. Ir(III) complexes served as electron acceptors (A) in the polymer resistive memory device.PDA is a bionic inspired by the adhesive mussel byssus protein. Under alkaline conditions, dopamine self-polymerizes, even on smooth surfaces including metals, glass, ceramics, and plastics [19]. Compared with traditional spin-coated polymer-based memory devices, the in-situ growth of PDA on a substrate can achieve easier and closer bonding between materials and electrodes due to its simple preparation method and excellent adhesiveness [20,21,22]. Moreover, the surface of PDA is rich in catechols and amines, which can act as reaction sites with small molecules, polymers, or metal complexes [23]. By forming such a multifunctional reaction platform, the applications of PDA have been further expanded to chemical modification, and studies of PDA in resistive memory devices have also recently emerged [24,25]. It is worth mentioning that PDA also shows great potential applications in flexible and wearable memory devices [26].

In this contribution, PDA was self-assembled on an indium tin oxide (ITO)-coated polyethylene glycol terephthalate (PET) substrate as a reaction platform. By adding the bromine-containing initiator, BiBB, surface-initiated atom-transfer radical polymerization (SI-ATRP) was triggered to graft an iridium complex on PDA to form a polymeric active layer. The characterization of the structure and morphology confirmed that PDA and iridium polymers were in-situ and grown step-by-step on the substrate. The thin film formed by PDA and iridium polymers had a planar surface morphology with a thickness of about 80 nm. Then, a flexible device with the structure of Al/PDA-PPy3Ir/ITO was fabricated by evaporating ultra-pure aluminum as the top electrode. The device exhibited nonvolatile rewritable memory behaviors with a turn-on voltage of −1.0 V and an ON/OFF current ratio of 6.3 × 103. The memory behaviors of the flexible device were almost unchanged under the bending test. To compare SI-ATRP with the traditional spin-coating process, a heterojunction device Al/PPy3Ir/PDA/ITO was constructed by self-assembly of PDA on ITO/PET substrate and spin-coating of the polymerized iridium complex on the PDA layer. The results demonstrated that, though the contrast device also exhibited nonvolatile rewritable memory characteristics, the Al/PDA-PPy3Ir/ITO device prepared by SI-ATRP had a higher current switching ratio and reliability.

3. Results and DiscussionThe preparation of the Al/PDA-PPy3Ir/ITO device is briefly illustrated in Scheme 1. First, PDA was self-assembled as a connecting layer on the surface of an ITO-coated PET substrate. Then, the iridium complex (synthesis route shown in Figure S1) was polymerized on the surface of PDA via SI-ATRP with BiBB as the initiator [27,28,29,30]. Then, aluminum was evaporated on the surface of the iridium polymer under vacuum to obtain a device with a structure of Al/PDA-PPy3Ir/ITO. The detailed preparation procedure and conditions are shown in the Supporting Information.To confirm the growth of the active layer on the substrate surface, Fourier-transform infrared (FT-IR) spectroscopy, Raman spectroscopy, and X-ray photoelectron spectroscopy (XPS) were employed at different stages during the preparation of the active layer. By contrasting the position of new peaks in the XPS wide-scan spectra in Figure 1a, clear C 1s, N 1s, and O 1s peaks (green) were observed at 284.6 eV, 400 eV, and 529 eV, respectively, when PDA self-assembled on the surface to obtain the PDA/ITO substrate. In the C 1s core-level spectrum of this substrate in Figure 1b, the peaks at 284.6 eV, 285.5 eV, 286.2 eV, 287.5 eV, and 288.8 eV were attributed to C-H, C-N, C-O, C=O, and O=C-O, respectively [31], which suggested the successful growth of PDA. After anchoring the initiator, new peaks appeared in the wide-scan spectrum (blue). The peak in the core-level spectrum (Figure 1c) at 69.6 eV, which belonged to Br 3d derived from the terminal bromine atom of BiBB, proved the anchoring of the initiator [32]. Similarly, after the polymerization of the iridium complex (red), new peaks emerged at 60.3 eV and 63.3 eV in Figure 1d, which belonged to the Ir 4f7/2 and Ir 4f5/2 orbitals [33]. This implied the grafting of the iridium polymer during the preparation of the active layer. The characteristic peaks of methyl, amide bonds, and ester groups at 1385, 1612, and 1721 cm−1, corresponding to acyl bromide and iridium ligand, were observed in the FT-IR spectra in Figure 1e. These peaks indicated the introduction of BiBB and the iridium complex to the system [34]. As a supplement, Raman spectroscopy (Figure 1f) showed that the PDA-PPy3Ir/ITO substrate contained characteristic peaks at 1350 cm−1 and 1580 cm−1 for PDA and 1250 cm−1, 1440 cm−1 for the iridium polymer [35]. These spectroscopic characterizations provide strong evidence that active layer materials were in-situ modified on the substrate surface.PDA inevitably aggregates in an alkaline solution. Small particles will deposit at the bottom and continuously accumulate, which affects the surface morphology. To obtain a uniform surface morphology of the active layer after the reaction, the substrate was placed vertically in a stirred reactor. Atomic force microscopy (AFM) was employed to test the surface roughness. As shown in Figure 2a–c, the arithmetic mean surface roughness of PDA/ITO was Ra = 0.65 nm. The flatness was due to the self-assembly of PDA on the surface. After anchoring the initiator, the surface roughness increased marginally to 0.95 nm, but the overall surface roughness was still very low. After SI-ATRP, the overall roughness increased slightly, to 1.06 nm. Figure 2d shows the height profiles along three lines in Figure 2a–c. It can be clearly seen that the surface roughness gradually increased after the introduction of the initiator and iridium polymer. The thickness of the active layer was measured by field emission scanning electron microscopy (FESEM), as shown in Figure 2e,f. The thickness of the active layer after PDA self-assembly was approximately 53 nm, while the thickness increased to about 82 nm after SI-ATRP, indicating that the thickness of the grafted iridium polymer was 29 nm.The Al/PDA-PPy3Ir/ITO device (Figure 3a) was prepared after evaporating aluminum as the top electrode on the active layer. The device exhibited typical nonvolatile rewritable resistive memory behavior in subsequent electrical performance tests, with turn-on and turn-off voltages of −1.0 V and 3.2 V, respectively, and a current ratio of 6.3 × 103. When a negative scan voltage from 0 V to −3 V was applied, the current jumped from a low-conductive state (OFF) to a high-conductive state (ON) at −1.0 V. The corresponding current changed from 3.3 × 10−6 A to 2.1 × 10−2 A, which implied that the current switching ratio reached 6.3 × 103. This process corresponded to the “write” process of data storage (sweep i). The device remained in a highly conductive state and could be read by another negative scan as the “read” process (sweep ii). Subsequently, a positive voltage was applied, and the conductivity of the device in the “ON” state dropped precipitously at the voltage threshold of 3.2 V (sweep iii).The conductivity remained low after applying another positive voltage sweep (sweep iv), which corresponded to the “erase” and “read” process of data storage, respectively. Similar cycles occurred in the subsequent repeated sweeps v–viii which demonstrated that the device possessed nonvolatile rewritable memory characteristics.The effect between current and time was investigated by applying a constant voltage of −0.5 V. The current remained at about 10−2 A and 10−6 A, respectively, even on the time scale of 104 s (Figure 3c). Large currents were very rare, indicating high stability of the device during continuous operation. Pulse voltage tests were also used to evaluate the stability of the device. More than 108 times reading pulse tests were performed with a pulse period of 20 μs and width of 10 μs under an applied voltage of −0.5 V. As a result, the device was insensitive to pulse voltage stimulation since the current did not change significantly (Figure 3d). During 200 switching cycles, as shown in Figure 3e, the device exhibited strong cyclic stability in both the ON and OFF states. Figure 3f gave the cumulative probability plots of ON and OFF state at −0.5 V, which exhibited a small variation.Flexibility is the crucial property for PDA-based devices in application of flexible and wearable resistive switching memory. I–V characteristic measurement was applied under both tensile and compressive strain (Figure 4a). As shown in Figure 4b, the device exhibits a bistable switching behavior under tensile strain, with turn-on voltage of −1.5 V and turn-off voltage of 4.2 V. Similarly, the device still exhibits typical nonvolatile rewritable memory performance under compressive strain, with turn-on and turn-off voltage of −1.5 V and 4.5 V, respectively. The cumulative probability diagrams of the turn-on and turn-off voltages of the device under different strains show stable and concentrated distribution (Figure 4d). In addition, when tensile strain and compressive strain are applied, the current values of the device in the ON state and OFF state almost remain unchanged, and the ON/OFF current ratio is greater than 104. To compare the differences between SI-ATRP and traditional spin-coating processes, an additional contrast device was designed. First, AIBN was used as an initiator, and the iridium complex was polymerized into an iridium polymer by free-radical polymerization [36] (Figure S2). After the self-assembly of PDA on the ITO substrate, the iridium polymer was spin-coated on the surface of the PDA film. The top aluminum electrode was evaporated to obtain an Al/PPy3Ir/PDA/ITO device. XPS spectra are shown in Figure S3. Subsequently, a morphology analysis was carried out (Figure S4). AFM images showed that the roughness of the active layer obtained by spin-coating was 2.02 nm, which was much greater than that of the Al/PDA-PPy3Ir/ITO device. Due to the large molecular weight of the monomer, the iridium polymer was prone to agglomeration instead of being evenly dispersed on the surface. A bright bulge emerged in the 3D AFM and FESEM images. The surface of PDA-PPy3Ir/ITO was more uniform than that of PPy3Ir/PDA/ITO, indicating that the SI-ATRP method was more suitable for forming a smooth active layer film.In subsequent electrical performance tests, the Al/PPy3Ir/PDA/ITO device obtained by spin-coating also showed nonvolatile memory performance with turn-on and turn-off voltages of −1.8 V and 4.4 V, respectively, and a current ratio of 13.3 (Figure S5). Specifically, when the negative voltage was scanned from 0 to −2.5 V, the current jumped from 4.5 × 10−3 A to 6.0 × 10−2 A when the voltage reached −1.8 V. The device changed from a low-conductive state to a high-conductive state. The device remained in the high-conductive state when the negative voltage was reapplied. After applying a positive scanning voltage, the device returned to a low-conductive state once the voltage reached 4.4 V. It retained its low conductivity under a subsequent positive voltage, showing nonvolatile rewritable resistive memory behavior. Then, the reliability of the memory device was studied by measuring its electrical performance under a constant voltage and pulse voltage. When the constant voltage of −0.5 V lasted for more than 104 s, the OFF state and ON state currents were about 10−3 A and 10−2 A, respectively, but there were sometimes large fluctuations. As the current switching ratio was only 13.3, there was a greater risk of misreading. A similar problem occurred during the pulse voltage test. Different from the stable distribution of current values in the Al/PDA-PPy3Ir/ITO device in the two conductive states, the device obtained by spin-coating showed large fluctuations during the switching cycle test. The device obtained by SI-ATRP had a higher current switching ratio, which reduced the misreading risk, and better time retention, pulse voltage endurance, and switching cycle stability than the device prepared by the spin-coating method.Iridium complexes have been previously used as electron acceptors in resistive memory devices [20,37], so the non-volatile rewritable characteristics of the Al/PDA-PPy3Ir/ITO device may be assigned to the charge transfer between iridium polymer and PDA. The existence of charge transfer was tested by spectroscopy. The UV-vis spectra of the active layer material PDA-PPy3Ir in different solvents are shown in Figure 5a. A wide absorption band appeared in the range of 325–425 nm. The characteristic peak appeared at 382 nm in toluene and shifted to 378 nm and 375 nm when the solvent was changed to the more polar tetrahydrofuran (THF) and N,N-dimethylformamide (DMF). This blue-shift indicated that the absorption may be attributed to the n-π* transition of the iridium complex [37]. A similar phenomenon also appeared in the steady-state fluorescence spectra in Figure 5b. In toluene, only one emission band at 512 nm appeared when the excitation wavelength was 390 nm. When THF and DMF were used, the peak position gradually shifted to 515 nm and 521 nm, respectively. The emission peak position exhibited a red-shift of 9 nm and its intensity decreased significantly, which indicated that charge transfer may exist in PDA-PPy3Ir.To provide evidence of intramolecular charge transfer, electron paramagnetic resonance (EPR) tests of the active layer material before and after UV illumination were carried out. The result in Figure 5c showed that the active layer had a strong EPR signal with a g value of 2.0023 and a peak-to-peak width (ΔHpp) of 7 G before illumination. After UV irradiation for 5 min (wavelength 365 nm), the g value and ΔHpp remained almost unchanged (g = 2.0024, ΔHpp = 6 G). However, the signal intensity was obviously weakened, which may be attributed to charge transfer between PDA (electron donor) and iridium polymer (electron acceptor) after irradiation, while unpaired electrons or radicals recombined. The existence of light-induced intramolecular charge transfer was proved [38].

EHOMO/ELUMO = −(Eox/red − Eox.(ferrocene)) − 4.8 eV

(1)

The cyclic voltammetry of the active layer showed multiple oxidation and reduction peaks in Figure 5d. According to the equations above, where EHOMO, ELUMO, Eox/red, Eox.(ferrocene), and Eg represent the energy level of the highest occupied molecular orbital (HOMO), lowest unoccupied molecular orbital (LUMO), oxidation/reduction potential of the material, the oxidation potential of ferrocene (0.38 eV) in this electrochemical system, and the bandgap of the material, respectively [39]. Since the oxidation potential was 0.86 V, and the first reduction potential was −0.77 V, the potentials of the HOMO and LUMO were calculated to be −5.28 eV and −3.65 eV, relative to the saturated calomel electrode (SCE) after correction. For the electrode material, the potential of ultra-pure Al, as the top electrode, was −4.28 eV relative to the SCE electrode, while the potential of ITO, as the bottom electrode, was −4.8 eV. The energy barrier between ITO and the HOMO energy level was smaller than that between Al and the LUMO energy level (Figure 5e). This indicates that charge transfer in the active layer was dominated by hole transfer [39]. Under the applied electric field, efficient hole transfer from donor (PDA) to acceptor (iridium polymer) occurs in Al/PDA-PPy3Ir/ITO device. When the charge transfer state is reached, the conductivity of the D-A system rises sharply, which corresponds to the device switching to the high conduction state (ON) after the SET process (Figure S6). The AC impedance spectra (Figure 5f) of the active layer showed that the total resistance of the active layer was lower than that of pure PDA. The equivalent circuit of the entire system was R(CR)(CR)W, and the resistance was composed of the total resistance of the solution (Rt), contact resistance (Rs), and charge transfer resistance (Rct). Since the total resistance of the solution was the same as that of the electrolyte solution, and the contact resistance between the platinum electrode and the solution was identical, the difference of resistance was attributed to the lower charge transfer resistance of PDA-PPy3Ir [40].

For the Al/PPy3Ir/PDA/ITO device obtained by the spin-coating method, a donor-acceptor heterojunction memory device with a structure of Al/acceptor polymer/donor polymer/ITO was constructed. The iridium polymer was the electron acceptor layer, and PDA was the electron donor layer. This spin-coating method relied on the physical connection between layers and was weaker than covalent bonds between iridium polymer and PDA. This may be the reason why the device obtained by the SI-ATRP method was more reliable than the one obtained by the spin-coating method.