Colossal Tunneling Electroresistance in Co-Planar Polymer Ferroelectric Tunnel Junctions

neuromorphic applications. [1–5] Ferroelectric tunnel junctions (FTJs), an extension of polar switch proposed by Esaki et al., [3] is a metal/ferroelectric/metal junction with the ferroelectric layer being thin enough to allow for quantum mechanical tunneling of the charge carriers from one electrode through the ferroelectric tunnel barrier to the counter electrode. The resistance of the junction, or tunneling electroresistance (TER), switches between two nonvolatile states through the reorienta-tion of polarization of the insulating ferroelectric tunnel barrier. [6–10] field-effect using Simmons direct tunneling model. The present work is the first demonstration of functional FTJs manufactured via a scalable lithography-based nano-patterning technique and could pave the way to new and exciting memory device concepts. 75 ° C on a hotplate for 10 min. After deposition of the second electrode (Au, 80 nm), a thin coating of the commercial adhesive First Contact is applied from solution over the entire sample’s surface and left to dry at room temperature for 30 min. The adhesive layer is then manually peeled-off to remove regions of the top Au layer overlapping with the bottom SAM-treated Al electrodes. The resulting patterns consist of co-planar asymmetric electrodes with a nominal inter-electrode gap of ≈ 10 nm. ODPA is removed from the surface using 10 min UV/O 3 treatment. The adjacent Al/Au electrodes are shown in the top view scanning electron microscopy (SEM) image in Figure 1b, where the presence of a homogeneous interface with nm-scale gap between the two metal electrodes is clearly visible. The nanogap is continuous over large distances as demonstrated by atomic force microscope (AFM) image of the metal electrodes in Figure S1, Supporting Information. The extreme dimensionality of the nanogap can be better visualized in the cross-sectional transmission electron microscope (TEM) image shown in Figure 1c, where short inter-electrode distances < 10 nm can be discerned. the


Introduction
Two-terminal resistive switching devices that upon application of an external bias toggle between a low resistance ON-and a high resistance OFF-state, are ideal for memory and synaptic a reliable lithography-based technique that allows reproducible fabrication and enables upscaling and integration.
In-plane devices such as break junctions have been successfully used for the study of tunneling transport through single molecules. [30,34,35] The in-plane configuration has the advantage that the electrodes can be fabricated before deposition of the ferroelectric layer atop. However, techniques like break junctions have low device throughput and are not suited for upscaling. [7,8,36] It has been demonstrated recently that a lithography-based method, namely adhesion lithography (a-Lith), schematically shown in Figure 1a, can alleviate the poor scalability issues of the break junctions, and allow for rapid fabrication of nanogap electrodes, with typical gap spacing between 4-10 nm. [33,37] A-Lith has been successfully employed in for the fabrication of nanoscale memristors based on both inorganic and organic semiconductors. [33,37] Here, we employ a-Lith to fabricate nanogap with asymmetric electrodes and demonstrate FTJs based on the ferroelectric polymer P(VDF-TrFE) that show stable and reproducible giant TER approaching 10 6 %; to our knowledge this is the largest reported value to date for metal/polymer/metal FTJs. Asymmetric nanogaps with Al and Au electrodes with spacing of less than 10 nm, Figure 1b-c, are employed for the realization of the FTJs. The resulting FTJ is a non-volatile twoterminal resistive memory with rectifying behavior and shows time-invariance of the tunneling current in the programmed states. We show unambiguously that the current transport through a FTJ is dominated by tunneling that is modulated by ferroelectric polarization of P(VDF-TrFE) in the gap.

Demonstration of FTJ
To ensure the successful formation of electrically-isolated nanogaps, we perform current-voltage (I-V) measurements before deposition of the P(VDF-TrFE) layer. Bias is applied to the Au electrode, while the Al electrode is grounded. The bias is swept from 0 to +5 V, to −5 V and back to 0 V. The current for the as-prepared (empty) nanogaps, shown with black symbols in Figure 2a, is in the order of 10 pA, and close to the detection limit of our measurement setup. Hence, the Au and Al electrodes are electrically isolated and, thus, reliable nanogaps that are spatially separated have been formed.
Following the demonstration of the isolated electrodes with nanometric gaps, we measure current transport through Au/P(VDF-TrFE)/Al nanogap devices. A 300 nm-thick layer of P(VDF-TrFE) (63-35) (Solvay) is spin-coated at room temperature atop the fabricated nanogaps from 5 wt% cyclopentanone solution. [18] The devices are annealed at 140 °C in a partial vacuum (1 mbar) for 2 h to increase the crystallinity of the P(VDF-TrFE) layer.
Adv. Electron. Mater. 2020, 6,1901091  Piezoelectric force microscopy (PFM) is performed to demonstrate ferroelectricity of the P(VDF-TrFE) layer, as shown in Figure S2, Supporting Information. A writing voltage of +7 V is applied to 7 × 7 µm 2 area, to put P(VDF-TrFE) in negative polarization state. Application of −7 V to a 4 × 4 µm 2 area reverses the polarization. The recorded phase map, Figure S2b, Supporting Information, confirms the complete reversibility of the ferroelectric domains. Moreover, the PFM phase image, Figure S3, Supporting Information, of the polarization map on the nanogap area, unambiguously, proves formation of a good contact between P(VDF-TrFE) and the electrodes.
To electrically characterize the FTJ, bias is applied on the Au electrode and varied from 0 to −5 V, to +5 V and back to 0 V. Figure 2a shows the measured I-V sweep. For the sweep direction from 0 to −5 V, the current is initially in the order of 10 pA. At nearly −2 V, the current begins an exponential rise with increasing |voltage|, and then shows a steep jump at −4 V from several nA to nearly 1 µA, eventually reaching 5 µA at −5 V. Upon back sweeping from −5 to 0 V, the current remains high and follows a different path, exhibiting a hysteretic behavior. During the sweep from 0 to +5 V, in the beginning, the current shows exponential rise with bias up to nearly +2 V. Above +2 V the current saturates, and between +4 and +5 V, the current increases slightly. For the scan direction +5 to 0 V, the current exponentially decreases with the bias. The full I-V loop is hysteretic and shows a bistable rectifying behavior. The junction switches to low-resistance on-and high-resistance offstate at −4 and +4 V, respectively. The spacing between the Au and Al electrodes for the FTJs is less than 10 nm (Figure 1c). Therefore, the electric field at which resistance switching occurs amounts to ≈400 MV/m. The switching electric field coincides with the coercive field that is reported for ultra thinfilms of P(VDF-TrFE). [38,39] The occurrence of the resistance switching at electric fields comparable to the coercive field of P(VDF-TrFE) is a strong indication that the modulation of the resistance is due to polarization switching of the P(VDF-TrFE) ferroelectric layer present in the nanogap. To explicitly attribute the measured I-V to polarization switching of P(VDF-TrFE), in the next step we wash away the polymer film from the junction. The current, as shown in Figure 2a, drops down to a value comparable to that of the as-prepared pristine junction, viz. 10 pA.
To demonstrate programmability of the Au/P(VDF-TrFE)/Al junction, pulses of ±5 V are applied for 100 ms to set the junction into high/low-resistance states. The programmed state is subsequently probed either using a voltage sweep between −3 and 3 V or a voltage pulse at ±3 V. As shown in Figure 2b, the Au/ (PVDF-TrFE)/Al junction can be programmed into two distinct on-and off-states. In the on-state, the junction shows rectifying behavior, which is highly desired for memory applications. [40][41][42] The device shows a colossal TER ratio of 8 × 10 5 %, which in combination with rectifying behavior, makes the coplanar FTJs stand out among other devices reported in literature as summarized in Table 1. The TER ratio has been obtained from the on-and off-state currents at −3 V which amount to 8 × 10 −7 A Adv. Electron. Mater. 2020, 6,1901091   CT-AFM 75 000 - [30] In-plane electrodes -- [31] In-plane electrodes by electromigration -- In-plane electrodes 3200 No [36] In-plane electrodes 250 No [43] Vertical electrodes 1000 No [44] In-plane electrodes 500 No Present work Co-planar nanogap by a-lith 800 000 Yes and 1 × 10 −10 A, respectively, and approaches 10 6 %. To probe the retention time of the programmed resistive states, a critical parameter for the applications of FTJs, we program the FTJs into the on-and off-states by application of ±5 V pulses, and read the states at −3 V in time. Both the on-and off-states do not show any deterioration for a period longer than 24 h, as shown in inset of Figure 2b. The work functions of the Au and Al electrodes amount to 4.9 and 4.2 eV, respectively. Therefore, there is a builtin potential, Δϕ, of 0.7 eV in the junction in the absence of the external bias. Considering that the nanogap is ≈10 nm, the built-in field amounts to 70 MV m −1 . The coercive field at which P(VDF-TrFE) switches is 400 MV m −1 , which is nearly six times larger than the built-in field. Therefore, no back switching or depolarization of P(VDF-TrFE) would be expected. Consequently, a long retention time is anticipated and indeed is experimentally measured. The programming and read voltages of, respectively, 5 and 3 V are compatible with today's electronic. The FTJs enable readout at any voltage below the coercive bias. The programming voltage of the FTJ depends on the gap size between the electrodes Hence, upon reducing the gap size, programming voltage drops. Optimization of the nanogap for low voltage operation however is beyond the scope of the present work. Moreover, the FTJs show good cycle endurance upon repeated cycling. As presented in Figure S4, Supporting Information, the TER remains unchanged upon cycling for more than 3000 cycles, which ensure endurance of the nanogap FTJs. The measured cycle endurance is acceptable for the targeted applications of the FTJs, namely as memory element in low-cost/disposable electronics and RFID tags.
In the following we have programmed the device into the onstate and measured the I-V sweeps at different temperatures to study the charge transport mechanism in the Au/P(VDF-TrFE)/ Al nanogap junction. Figure 3a shows temperature dependence of the on-state current sampled at −3 V. We have employed Simmons model [45] as the simplest tunneling model to describe electron tunneling from one metallic contact to another through the P(VDF-TrFE) thin layer. The tunneling current shows a weak quadratic temperature described by [46]  (1) where J 0 K is the current passing through the junction at 0 K, α is a constant, k B is Boltzmann constant and T is temperature. Fitting Equation (1) to experimentally measured current, gives J 0 K and α that amount to 2.0 × 10 −7 A, and 4.5 × 10 21 N.m., respectively. The Au/P(VDF-TrFE)/Al is therefore a ferroelectric tunnel junction. The current in the on-state depends only on the tunnel barrier heights. [47][48][49] Following Simmons model [50,51] for tunneling through insulating films with a generalized potential barrier, the current flowing through the junction is exp e xp where J T , C, and ϕ are the current pre-factor, tunneling parameter, and the tunnel barrier height at the Fermi level of the electrodes. We note, C is proportional to the electron's effective mass.
Representative I-V characteristics of the FTJ at room-(293 K) and low-temperature of 113 K are shown in Figure 3b. Equation (2) describes very well the I-V characteristics of the FTJ at reverse bias, as shown in Figure 3b. We have determined the values for J 0 , C, and ϕ from the fits. Interestingly, all I-V curves could be fitted using similar C and ϕ of 1.18 ± 0.005 and 2.89 ± 0.02 eV, respectively. The temperature dependence comes from J T , which is well described by Equation (1).
For the positive bias, that is, when tunneling from the Au electrode, the current can be described with Equation (2) with the same C value but using a different J 0 and ϕ of ≈4 × 10 −9 A m −2 and 4.2 eV, respectively. Presence of a much larger tunnel barrier of 4.2 eV at the interface between the Au electrode and P(VDF-TrFE) gives much lower current at positive biases, and yield a rectifying behavior in the FTJ. We note that, in the off-state, both forward and reverse bias currents are low and comparable to the leakage (noise) current. Therefore, any attempt to analyze the off-state current, at this point, is prone to wishful interpretations.

Discussion
The choice of Simmons model is due to its simplicity, and obtaining reasonably good fits to experimental data shows its suitability. The Simmons model, considers only elastic electron tunneling, and is valid for low electric fields. The tunneling electrons, provided that they have enough energy, can lose energy through exciting vibrations of the CH or CF bonds, which leads to an inelastic tunneling path and an additional contribution to the tunneling current. However, the applied bias of ±3 V is low to activate inelastic processes. Application of more elaborate tunneling models that take for instance charge trapping into account, may produce better fits through introducing more fit parameters. Such models require a knowledge of electron trap density, distribution, and their energy levels, which are lacking for PVDF and P(VDF-TrFE), hence hindering application of such models. We note that the fits in Figure 3b have been obtained using a constant temperature-independent tunnel barrier, ϕ . Assuming ϕ as a floating fit parameter improves the fit quality but reduces the ϕ to a meaningless fit parameter.
The energy band diagram of the Al/P(VDF-TrFE)/Au FTJs in its un-polarized pristine state is shown in Figure 4a. To draw the band diagram we assumed that the lowest unoccupied molecular orbital (LUMO) of P(VDF-TrFE) lies at 3.5 eV. [52] The work function of Al and Au electrodes have been measured using Kelvin probe and amount to 4.2 and 4.9 eV, respectively. We have included a thin insulating barrier at the Al electrode to account for the presence of the native Al-oxide layer, which is typically ≈1 nm thick. Figure 4b shows the FTJ under negative polarization, at zero bias. Due to the presence of polarization charges, there is a slight distortion in the alignment of the energy levels, particularly of the LUMO of P(VDF-TrFE). We note the thickness of the skin depth for the metal electrode can be disregarded because of its small magnitude, compared to the thickness of the P(VDF-TrFE) layer. [48] Figure 4c, shows the FTJ in the on-state under the reverse bias of −3 V. The ferroelectric polarization is pointing toward the Au electrode. The effective tunnel barrier width is shown with the yellow arrow. Because both the external field and ferroelectric polarization are parallel, under the reverse bias condition, the tunneling barrier width is substantially lowered and the current is high. The forward bias situation is shown in Figure 4d. Ferroelectric polarization and applied external field are anti-parallel. Moreover, the majority of the applied potential is dropped over the P(VDF-TrFE) layer. Therefore, the Al-oxide barrier acts as an extra barrier which blocks charge transport. As a result, the effective tunnel barrier is large and the current under forward bias is substantially lower. Therefore, the presence of the Al-oxide interfacial layer would lead to a rectifying behavior of the junction. The tentative description based on the band diagram is in agreement with the calculated J 0 and ϕ for both reverse and forward bias situations.

Conclusion
In conclusion, we have demonstrated rectifying FTJs with P(VDF-TrFE) using co-planar nanogap asymmetric electrodes fabricated by adhesion lithography; a simple, scalable and highly cost-effective patterning technique. The FTJs exhibit ferroelectric polarization switching at room temperature and giant TER approaching 10 6 %, along with excellent data retention. Analysis of the current-voltage characteristics suggests direct tunneling as the dominant charge transport mechanism across the P(VDF-TrFE) ferroelectric tunnel junction. Barrier modulation at the Al/P(VDF-TrFE) electrode interface is responsible for the observation of giant TER effect in these planar FTJs that have been realized via wafer-scale adhesion lithography. The fabrication technique of the FTJs is a low-cost method that bears the potential for up-scaling and large-scale integration, and paves the way toward an industrially viable memory element for emerging large-area electronics. [53,54]

Experimental Section
Nanogap Fabrication by Adhesion Lithography: The nanogap electrodes with asymmetric Al/Au metal contacts are fabricated using a-Lith on silicon wafers (with thermally grown SiO 2 ). Figure 1a displays the process steps used for the formation of the device with co-planar nanogap electrodes. All Si/SiO 2 wafers are first cleaned by ultrasonication in deionized water, acetone and isopropanol for 10 min and then dried under a stream of dry nitrogen gas. Next, a 40 nm-thick layer of Al is deposited via thermal evaporation in high vacuum (10 −6 mbar) and subsequently patterned via standard lift-off photolithography and then immersed in an isopropanol solution containing 1 mm Octadecylphosphonic acid (ODPA) (Sigma-Aldrich) for 20 h to form a dense self-assembled monolayer (SAM) atop the native alumina (Al 2 O 3 ) layer. Subsequently the wafers are removed from the SAM solution, rinsed with isopropanol, and then dried/annealed at Adv. Electron. Mater. 2020, 6,1901091  75 °C on a hotplate for 10 min. After deposition of the second electrode (Au, 80 nm), a thin coating of the commercial adhesive First Contact is applied from solution over the entire sample's surface and left to dry at room temperature for 30 min. The adhesive layer is then manually peeled-off to remove regions of the top Au layer overlapping with the bottom SAM-treated Al electrodes. The resulting patterns consist of co-planar asymmetric electrodes with a nominal inter-electrode gap of ≈10 nm. ODPA is removed from the surface using 10 min UV/O 3 treatment. The adjacent Al/Au electrodes are shown in the top view scanning electron microscopy (SEM) image in Figure 1b, where the presence of a homogeneous interface with nm-scale gap between the two metal electrodes is clearly visible. The nanogap is continuous over large distances as demonstrated by atomic force microscope (AFM) image of the metal electrodes in Figure S1, Supporting Information. The extreme dimensionality of the nanogap can be better visualized in the cross-sectional transmission electron microscope (TEM) image shown in Figure 1c, where short inter-electrode distances <10 nm can be discerned.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.