• Nem Talált Eredményt

on a microscopic level, molecular dynamical simulations were also performed by Dr.

Dávid Zsolt Manrique at the Physics Department of Lancaster University [2]. The simulated system reproduced the experimental ndings and supported the role of the geometrical asymmetry.

As the last step to conrm this phenomenon, the switching behavior was tested on such a nanofabricated device, which mimics the geometry of the STM setup. The nanofabricated devices have further advantages compared to the STM arrangement, such as higher mechanical stability and they can be integrated into larger circuits.

This part of the research was my task.

Figure 4.3: Optical (I and II) and SEM (III) images about the asymmetric Ag-Ag2S-Ag on-chip samples after the electrical measurements. The samples are made of two large metal pads (I) and a constriction (III) at the middle with asymmetric geometry. On the SEM image the white scale bar in the lower right corner indicates 200nm [2].

During the lift-o signicantly portion of the devices aked o the substrate due to the poor adhesion to the Si3N4. To get better device fabrication yield titanium adhesion layer was deposited before the silver. In order to avoid the parallel conduc-tion at the constricconduc-tion, during the evaporaconduc-tion of the titanium the sample holder was tilted. Since at the constriction the height of the PMMA mask is similar to the width of the structure, the titanium is not present at the constriction under the silver, as illustrated in Figure 4.4.a. Due to the undercut prole of the double layer PMMA a narrow stripe appeared parallel to the edge of the structure, as it can be seen in the SEM image (Figure 4.4.b). The adhesion could be improved by deposit-ing a 1−2nm thick titanium layer under the silver. Such a thin Ti layer does not form continuous layer and thus do not provide parallel electron conduction. These deposition processes substantially increased the yield of the sample fabrication.

As it is discussed in Section 4.3 the sulfurization is performed by placing the pure silver devices into sulfur rich atmosphere. The aforementioned device are made of a single, uniformly thick silver layer. However, it is not feasible to expose only the constriction part to the sulfur. Therefore, if we performed the sulfurization process on the freshly made samples, all parts would transform to Ag2S and the devices

would not exhibit resistive switching behavior. We must localize the active region.

We attempted to achieve it in two ways. The one is to break or thin the silver wire at the narrowest cross-section using electromigration technique. This thinned region transforms to Ag2S sooner than the other parts of the device. By controlling the sulfurization time it is possible to form few tens of nanometer sized switching region.

The second option is to fabricate a shadows mask for the sulfurization process which is open at the constriction, but protects the rest parts of the device. During my work we tested three dierent mask structures.

As the simplest case we covered the whole chip by few 100nm thick PMMA layer and the polymer was removed at the constrictions using another lithography step. The geometry of the silver junctions was the same as presented above. In order to reduce the device to device deviation due to the sulfurization process, the devices were packed closer to each other in groups of ten. This design aids to expose constrictions to the same amount of sulfur. The device arrangement is shown in Figure 4.4.c. However, the sulfurization was not eective using this PMMA shadow mask, the samples showed hysteric I-V curves only after 30 minutes sulfurization time. The passivization against the sulfur may arise from the long-term mechanical instability of the PMMA slit, the surface of the constriction is covered by a thin polymer layer.

Figure 4.4: a) Schematics of deposition pattern in tilted sample holder (top panel) and in perpendicular position (bottom panel) using double-layered PMMA. b) SEM image of a device after the metal evaporation. The white scale bar at the lower left corner indicates 200nm. c) Optical image of a group of samples.

To overcome this problem the shadow mask was prepared by evaporating titanium layer to the top of the silver. It was carried out by tilting the sample holder at two

angles with the same magnitude but opposite direction. Figure 4.5 shows a SEM image about the surrounding of the constriction. Due to the two dierent tilted evaporation processes the parallel lines are present at both sides of the electrodes.

The dashed white lines around the constriction indicate the contour of the titanium, inside this region the silver is not covered for the same reason as discussed . Despite of the more sophisticated mask fabrication these devices did not show better sensibility to the sulfur. After exposing the devices to 30 minutes long sulfurization only a weak resistive switching eect could be observed. Despite of the tilted sample holder a thin titanium layer may have covered the surface of the silver at the constriction as well.

Figure 4.5: SEM image of a device using titanium shadows mask for sulfurization.

The dashed white lines around the constriction highlight the contour of the top tita-nium mask layer. The white scale bar at the lower left corner indicates 200nm

The usage of shadow mask can be avoided if the silver layer is not evenly thick.

This structure can be realized in the same way as the titanium mask. After the perpendicularly evaporated 40nm thick silver layer, additional silver was deposited in tilted sample holder at two dierent angles. Thereby the silver is thinner at the constriction in the same way as it is shown in Figure 4.5. The full deep sulfurization takes place sooner at this region than the other parts of the device.

Taking SEM images about the constriction before the sulfurization also puts the cleanness of the devices at risk. During scanning of the surface by the electron beam some organic contamination could burn onto the silver which forms a passivation layer. Thereby after the metal deposition it is not worth putting back the devices into the electron microscope to look at their structure.

The rst set of devices and the samples with masks were fabricated by Miklós Csontos. Later I also fabricated devices without masks. All samples were made at Microtechnology Department of MTA EK MFA.

4.3 Resistive switching in asymmetrically shaped