Mixed‐Mode Electro‐Optical Operation of Ge2Sb2Te5 Nanoscale Crossbar Devices

The use of phase‐change materials for a range of exciting new optoelectronic applications from artificial retinas to ultrahigh‐resolution displays requires a thorough understanding of how these materials perform under a combination of optical and electrical stimuli. This study reports for the first time the complex link between the electronic and optical properties in real‐world crossbar nanoscale devices constructed by confining a thin layer of Ge2Sb2Te5 between transparent indium tin oxide electrodes, forming an optical nanocavity. A novel proof‐of‐concept device that can be operated by a combination of optical and electrical stimuli is presented, leading the way for the development of further applications based on mixed‐mode electro‐optical operation.


DOI: 10.1002/aelm.201700079
(PCMs). [6] Exploiting both optical and electrical properties of phase-change materials simultaneously, and the possibility of modulating optical effects using electrical excitations, has only very recently opened up new application fields, such as low-power, nonvolatile, phase-change nano displays. [7][8][9] A deeper understanding of the relationship between the optical and electrical properties of phase-change materials, such as GST, is now urgently required to spur the development of new optoelectronic applications.
Some previous studies have examined the basic relationship between reflectance and electrical resistance of phase-change materials. For instance, by comparing the crystallization times of Ge 15 Sb 85 in planar films excited optically and in bridge nanodevices excited electrically, it was shown that optical and electrical properties are interrelated. [10] Furthermore, simultaneous measurements of reflectivity and electrical resistivity of GST thin films during isothermal crystallization showed a decrease in resistivity preceding an increase in reflectivity, explained by the appearance of small crystalline nuclei that enable electrical conduction by percolation prior to the formation of a crystalline region large enough to be optically detectable. [11] Similar observations have been reported by measuring the transient crystallization response of as-deposited-amorphous GST exposed to singlenanosecond laser pulsed excitations. [12] In the above-mentioned works, films were driven from fully amorphous to fully crystalline by applying either a continuous heat source or laser pulse. However, intermediate partially crystalline states can of course also exist. In particular, GST allows for the existence of nanometer-sized crystalline nuclei in the amorphous phase by having a nucleation dominated crystallization mechanism. [13] The eventual increase in the size and number of crystalline nuclei usually accounts for the transition to the fully crystalline state. Therefore, by applying a series of pulses with a controlled amount of energy, it is possible to transition between fully amorphous and fully crystalline states in discrete incremental steps. This property has been referred to as accumulation and occurs by means of optical as well as electrical pulses. [14][15][16] In this work therefore we investigate the phase transition of GST by simultaneously examining the optical and electrical responses of crossbar-type nanoscale devices exposed to cumulative optical and electrical excitations.
In order to apply laser pulses and measure the reflectance of the samples, we used a custom-built reflective laser scanning The use of phase-change materials for a range of exciting new optoelectronic applications from artificial retinas to ultrahigh-resolution displays requires a thorough understanding of how these materials perform under a combination of optical and electrical stimuli. This study reports for the first time the complex link between the electronic and optical properties in real-world crossbar nanoscale devices constructed by confining a thin layer of Ge 2 Sb 2 Te 5 between transparent indium tin oxide electrodes, forming an optical nanocavity. A novel proof-of-concept device that can be operated by a combination of optical and electrical stimuli is presented, leading the way for the development of further applications based on mixed-mode electro-optical operation.
Chalcogenide-based phase-change materials have been the subject of much scientific interest because of their ability to reversibly switch between two solid states, amorphous and crystalline, at high speed [1] and with relatively low power consumption. [2] Transitions between states are achieved via localized heating with energy applied either optically or electrically. [3] One of the most common phase-change materials is Ge 2 Sb 2 Te 5 (GST). Each of the atomic configurations of GST have different physical (optical, electrical, and mechanical) properties. [4] The crystalline phase is more reflective and up to three orders of magnitude more electrically conductive than the amorphous phase. [1] These differences in the physical properties between phases, in addition to the ability to retain either phase state for a long time at ambient temperatures, [5] have been utilized in technologies such as rewritable optical data storage (DVD-RW, Blu-ray RW) and nonvolatile electronic phase-change memories The copyright line of this paper was changed 9 June 2017 after initial publication. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited. microscope (Figure 1a). A 658 nm (red) laser diode (ROHM RLD65PZB5) is coupled to a single mode optical fiber using an aspheric lens coupler/collimator (Thorlabs F280APC-B). Once coupled, the beam passes through a 90:10 fiber beam splitter (Thorlabs FC632-90B-APC) that allows the reflected signal to be sensed by a photodetector (New Focus 2001). The scanning objective was built in a confocal configuration by assembling a ferrule terminated single mode optical fiber (Thorlabs SMPF0206-APC) to a 0.29 pitch gradient index (grin) lens (Thorlabs GRIN2906). This configuration allows the fiber to be used as a pinhole [17] in the path of the reflected light. The optimum values for the trade-off between working distance and spot size were calculated following a transfer matrix model [18] and set to 328 and 2.24 µm, respectively. The scanning objective was mounted on top of the precision XY piezo stage of an Asylum MFP-3D atomic force microscope (AFM) using a custom-built base in contact to the XY piezo by a three groove kinematic clamp. By using the MFP-3D AFM it is possible to perform the XY movements to construct the image as well as to use one of the integrated AFM analog/digital converters to acquire the photodetector signal simultaneously with the raster scan. Such synchronization allows for precise positioning of the laser beam subsequently to the scan. The focus of the laser spot on the sample (Z axis) was controlled by a combination of a manual linear stage (Thorlabs DT12/M) for coarse adjustment and a custom-built multielement piezo linear actuator [19] for fine adjustment. Essentially, we adapted the extremely fine linear (x-y) piezo of our Asylum AFM to our home-built optical probe, in order to optically test our devices. Similar laser probing configurations have been reported previously in the literature. [20] In this case, however, we used fiber coupled optical components in order to reduce the footprint of the custom made laser accessory so it could be accommodated in the existing MPF-3D AFM system.
For our experiments we fabricated crossbar nanoscale devices following the process shown in Figure 1b-j, with indium tin oxide (ITO) transparent electrodes which confine a small GST area forming an optical nanocavity that is accessible optically and electrically. Starting with a thermally oxidized Si wafer (320 nm SiO 2 ) ( Figure 1b) and by employing a bilayer positive resist (polymethyl methacrylate (PMMA) 495 + PMMA950) (Figure 1c), the bottom electrode design was patterned by e-beam lithography ( Figure 1d). After development, reactive ion etching (RIE) was used to produce 20 nm deep trenches in the SiO 2 . The trenches were then filled with 20 nm ITO deposited by RF plasma sputtering ( Figure 1e). After lift-off, the sample was annealed at 300 °C for 30 min in air to improve the conductivity of the ITO bottom electrode. The trenches in the SiO 2 substrate serve as a simple planarization mechanism, which is used to increase the yield of the devices by avoiding short circuit between the top and bottom electrodes in the cross bar structure, in addition to improving thermal properties of the device. A second e-beam lithography process was used to pattern the top part of the cross bar device (Figure 1f-g). In this case a 15 nm thin-film of GST was deposited by RF sputtering, followed by 20 nm of ITO that serves as the transparent top electrode. Both materials were deposited within the same deposition run, avoiding the GST being exposed to oxygen from air ( Figure 1h). Then, a set of larger electrodes were added to access the device electrically. These electrodes were created by a third e-beam lithography process followed by thermal evaporation of 30 nm Cr as an adhesive layer followed by 60 nm  Au. The considerable thickness of the Cr adhesive layer (30 nm) was used to improve the electrical contact between the ITO and the Au electrodes. [21] A batch of 48 devices of each size 100, 200, 300, 500 nm, and 1 µm were fabricated simultaneously during the fabrication process, it was found during characterization that 500 nm and 1 µm gave the better response in terms of reliability and optical response and were therefore used for our experiments. Finally, electrical access to the devices was achieved by wire bonding each device to a chip carrier. The chip carrier was mounted on a custom-made magnetic holder designed to firmly fix the device in place during the raster scan of the laser microscope. Electrical measurements were performed with a source meter (Keithley 2614B).
It is known that thermal characteristics play an important role in the operation of phase-change memories. [22] Good heat containment accounts for more efficient devices, by reducing thermal losses. The use of ITO electrodes in the present design serves two purposes: it enhances thermal isolation because of its low thermal conductivity, ≈11 W mK −1 , compared to metallic electrodes while simultaneously allowing optical access to measure the changes in reflectance induced by the change of phase of the GST layer.
Three combined mixed-mode experiments were carried out to explore the simultaneous electrical and optical behavior of phase-change materials (in this case GST) in a nanoscale device context. The first experiment was to ascertain how optical pulses would modulate the resistance of the devices. We subjected our devices to a series of optical pulses to induce crystallization; the resistance of the device was then measured after each optical pulse. The optical pulses were 50 ns in duration and 13 mW in power (output power radiated from the grin lens), a combination that allows for a steady progression from the initial as-deposited amorphous state to the crystalline phase. Figure 2a shows the evolution of the resistance for two devices of dimensions 500 nm × 500 nm and 1 µm × 1 µm. As it can be seen, the resistance value in both devices gradually drops, by approximately three orders of magnitude in this case, as the cumulative effects of the optical pulses lead to crystallization. It is important to notice that despite the large change, the device is highly resistive at all times (≈1 MΩ in the low value) this is caused by the parasitic resistance of the ITO electrodes.
As the GST becomes more crystalline, the electrical resistance decreases in both devices of different sizes. The difference in the accumulative response of both devices is due to an increase of the thermal losses associated with wider electrodes. Therefore the optical pulses applied to the larger device were less effective in inducing crystallization, and more were required to reach the lowest resistive level.
In the second experiment, we examine how electrical excitation pulses affect the optical properties, specifically the reflectance, of a device. We thus applied a series of electrical pulses to a 500 nm × 500 nm crossbar cell, measuring its resistance and reflectance after each pulse. The applied voltage pulses were of 2.5 V and 100 ns duration, which again allows for a steady accumulative phase transition. Our optical measurements were performed by acquiring sequential images with our laser scanning microscope using a low intensity laser beam of 1 mW. The images were then post-processed following the methodology described in Section S1 of the Supporting Information (using such a methodology serves two purposes: first, after every scan it is possible to reposition the laser before applying the next pulse in case of unexpected stage drift; second, we can use the imaged substrate reflectance as a reference to compensate for laser intensity variations). Figure 2b shows the evolution of the reflectance and resistance of the crossbar device as a function of the number of electrical pulses applied. The electrical resistance drops as the number of pulses increases, as would be expected for a device undergoing crystallization via an accumulation process. Surprisingly, however, the reflectance of the devices decreases with increasing number of applied pulses. This behavior is counterintuitive, given that it is known that the reflectance of blanket GST films increases upon crystallization for wavelengths in the visible spectrum. [23] In this case, however, such response is a consequence of a shift, upon partial crystallization, in the resonance of the optical cavity formed by stacking the different layers (ITO/GST/ITO) that form the cell. This shift is produced by the change of the refractive index of the GST upon partial crystallization [7] (for more information see Section S2, Supporting Information).
It can also be seen in Figure 2b that the evolution of the reflectance occurs before the final resistance drop. The fact that b Figure 2. a) Evolution of electrical resistance in response to consecutive optical pulses. b) Evolution of resistance and optical reflectance in response to consecutive electrical pulses applied to a 500 nm device.
both measurements evolve differently suggests that the crystallization of the GST within the device may not be occurring homogeneously. These findings are similar to the previously observed by Liang et al., [12] who explained a delay in the evolution of the reflectance respect to the resistance by a 2D percolation model. Thus far, we have shown that subjecting the GST crossbar cell to optical pulses produces an observable change in resistance and that electrical pulses produce and observable change in reflectance. However, in order to further understand the relationship between the changes of these two physical properties we performed an additional experiment applying a series of 50 ns, 13 mW optical pulses to a 500 nm × 500 nm device and measured both the resistance and the reflectance after each pulse. These results are shown in Figure 3a. The graph has been divided into sections, and each section has been selected to highlight how the trends of the reflectance and resistance changes are either opposite or in tandem to each other as the GST freely changes phase.
In these results, the change in reflectance follows a similar behavior to the one observed in the previous experiment ( Figure 2b). As explained before, the reflectance initially decreases upon partial crystallization due to the resonance change of the optical nanocavity caused by the change of the refractive index of GST. Figure 3b shows the simulated nonmonotonic optical response at 658 nm of the optical nanocavity as a function of the crystalline fraction.
This curve is divided into two regions: in region (I) the reflectance decreases as the GST partially crystallizes; in region (II) a further increase of crystallization induces an increase in reflectance, reaching the peak value when the GST becomes fully crystalline. Given that an increasing crystallization fraction will produce a monotonic decrease of resistance, as was shown previously in Figure 2a, within region (I) the trends in the change of resistance and reflectance will be opposite to each other, whereas in region (II) they will be the same. Thus, different degrees of crystallization are responsible for the changes shown in the various regions (A-M) of Figure 3a, as the GST in the device traverses between regions (I) and (II) of the optical response shown in Figure 3b. This remarkable optoelectronic response is a consequence of the optical resonance characteristics of the optical nanocavity formed by the crossbar cell, along with the fact that, for the particular optical excitation pulses used here, the cell "oscillates" between regions (I) and (II) due to a process of partial crystallization and partial reamorphization induced by successive excitations in tandem with the change in absorbance of GST between amorphous and crystalline states [24] (this can be inferred from the observation that the partial reamorphization takes place only after sufficient crystallization has occurred-see Figure 3a, region A).
Up to this point, we have shown that a correspondence exists between resistance and reflectance when either optical or electrical excitation pulses are applied and that such correspondence depends on the degree of partial crystallization. In our third experiment we demonstrate true combined electrooptical operation of our devices (mixed-mode) at the nanoscale. A 500 nm × 500 nm device was conditioned optically until it reached the lowest resistance state. We then iteratively applied a series of 50 optical pulses of 80 ns at 65 mW of optical power (measured at the GRIN lens output) to reamorphize the GST, followed by an incremental staircase voltage signal ( Figure S4, Supporting Information) from 0 to 4 V to electrically crystallize the material again. Figure 4 shows the result of ten iterations of our mixedmode experiment. It is seen that after the application of a series  of optical pulses, an increment in the resistance is observed. Similarly, after an electrical triangular staircase signal is applied, the resistance decreases again as the GST recrystallizes. The change of the resistance is approximately one order of magnitude, which is significantly less than the change observed in the first transition from as-deposited amorphous state reported previously in Figures 1 and 2. Also the operation occurs in the lower range of the resistance, near the fully crystalline region. Both of these indicate that the optical pulses do not fully reamorphize the material (see Section S4, Supporting Information). However, partial reamorphization does occur and modulates the resistance, demonstrating mixed-mode operation for the very first time in nanoscale devices. We repeated the behavior seen in Figure 4 for over 100 iterations.
In conclusion, we have experimentally shown for the very first time in nanoscale GST phase-change devices that there exists a complex relationship between the resistance, reflectance, and crystallinity of the GST. Such a relationship exists irrespective of the means used to induce the crystallization (i.e., electrical or optical pulses). An optically induced phasechange can readily produce a change in device resistance while at the same time an electrically induced phase-change can be observed as a change in reflectance. We have also shown that partial crystallization due to accumulation of pulses occurs, and we suggest that partial crystallization of the GST within our mixed-mode device is not homogeneous. Further investigation is required to understand whether crystalline and amorphous phase segregation occur within the device, in order to optimize the optical contrast and energy required for the operation. Moreover, we have provided a proof-of-concept GST nanodevice in mixed-mode electro-optical operation by crystallizing it electrically and reamorphizing it optically while the resistance values show two different repeatable states. Our results are a first step toward the realization of a memflector [25] (the optoelectronic analogue of the memristor), as well as other opto electronic devices for future applications that require a memory element capable of operating electrically and optically. Such applications include optoelectronic interfaces for integrated photonic circuits, [26,27] in addition to potential new technologies like accumulative optical pulse detectors or synthetic retinas.

Experimental Section
E-Beam Lithography: A Jeol-5500 Series E-beam lithography tool and a PMMA bilayer resist (PMMA495 + PMMA950) were used. The final lithographic step to create access electrodes used a single PMMA495 resist layer.
Bottom Electrode Deposition: This was carried out in a Nordiko system; ITO was sputtered from a 2 in. target (Testbourne, UK) with a base pressure of 1 × 10 −7 mbar, operating pressure of 7 × 10 −3 mbar, 100 sccm Ar @ 30 W for 9 min 30 s.
Radio Frequency Sputtering GST/ITO Top Electrode: Ge 2 Sb 2 Te 5 was sputtered from a 2 in target (Super Conductor Materials, USA) with a base pressure of 1 × 10 −7 mbar, a chamber pressure of 7 × 10 −3 mbar, flow of 100 sccm Ar, at a power of 30 W for 4 min 10 s. The ITO top electrode was then sputtered in situ, with the same parameters as bottom ITO electrode.
Thermal Evaporation: 30 nm of Cr + 60 nm Au was evaporated from a Cr rod and Au wire on W crucible, respectively, at rate of 0.1 nm s −1 monitored by a crystal deposition monitor. Chamber base pressure was 2 × 10 −6 mbar.

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