In previous discussions, we demonstrates reactive metal (lower free energy of oxidation of the metal) is easier combined with oxygen from resistive switching films and forms metal oxide, which resulted the filaments were not ruptured and no resistive switching characteristics. However, the inert metal (higher free energy of oxidation of the metal) is hard easier combined with oxygen from resistive switching films. Therefore, the inert metal as top electrode has good resistive switching characteristics, but can＇t resistive switching in larger compliance current. Because inert metal have too higher free energy, when applied a dc voltage, the oxygen ions is direct from HfOX films to air. In this paper, we have studied to use two-layer metal as top metal, the first layer is inert metal and the second layer is thin reactive metal, such as Pd/Al, Pt/Ti, etc. Because thin reactive metal can trap oxygen ions to form thin interface layer and avoid oxygen ions from HfOX films to air, and Inert metal is protection reactive metal which avoid to direct contact with air.
3.3.1 Electrical properties
First, we fixed the thickness of Pd for 30nm, and to change the thickness of Al. In Pd/Al(20 Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-34 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-35. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-36. The Pd/Al(20 Å)/HfOX/TiN device shown good resistive switching characteristics. The VSET and VRESET values don＇t overlap each other, and the endurance is more than 300 cycles.
In Pd/Al(50Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-37 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-38. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-39. The Pd/Al(50 Å)/HfOX/TiN device shown good resistive switching characteristics. The VSET and VRESET values don＇t overlap each other, and the endurance is more than 550 cycles.
In Pd/Al(75Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-40 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-41. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-42. The Pd/Al(75 Å)/HfOX/TiN device shown good resistive switching characteristics. The VSET and VRESET values don＇t overlap each other, the endurance is more than 200 cycles, but the yield decreases.
In Pd/Al(100 Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-43 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-44. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-45. The Pd/Al(100 Å)/HfOX/TiN device shown good resistive switching characteristics. The VSET and VRESET values don＇t overlap each other, the endurance is
more than 450 cycles, but the yield decreases.
In Pd/Al(150 Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-46 shows the typical I-V curve for SET process. In Fig. 3-47 shows the typical I-V curve for RESET process. The Pd/Al(150 Å)/HfOX/TiN device shown bad resistive switching characteristics, the endurance is less than 3 cycles.
In Table 3-4 performs the comparison with the different thickness of Al of Pd/Al/HfOX/TiN structure and resistive switching parameters of the devices. Form the results, we demonstrates Pd/Al/HfOX/TiN structure which has good resistive switching characteristics, but thickness of Al is thicker to result that the yield and resistance decrease [69]-[70]. Because thickness of Al is thicker, which forms thicker AlOX at Al/HfOX interface to result the filaments were not ruptured.
In order to confirm our thoughts, we etched aboriginal Al/HfOX/TiN device for etching Al solution. The proportion of chemical solution is H3PO4 : HNO3 : CH3COOH : H2O = 50 : 2 : 10 : 9. Afterward, we etched thickness of Al of the device to 30 Å. In Al(30 Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-48 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-49. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-50. The Al(30 Å)/HfOX/TiN device has resistive switching phenomenon. Afterward, we used direct sputtering Al 30 Å as top electrode. In Al(30 Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-51 shows the typical I-V curve.
Distribution of VSET and VRESET are shown in Fig. 3-52. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-53. The Al(30 Å)/HfOX/TiN device has resistive switching phenomenon.
Form the results, we demonstrates thickness of Al is thicker to result that the yield decreases, and thickness of Al is thin, which has resistive switching phenomenon.
However, thickness of Al is too thin to result oxygen ions through AlOX from HfOX films to air. Therefore, we chosen Pd/Al(35 Å)/HfOX/TiN structure. In Pd/Al(35 Å)/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA and 10mA. In Fig. 3-54 and Fig. 3-55 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-56 and Fig. 3-57. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-58. In Pd/Al(35 Å)/HfOX/TiN device shown very good resistive switching characteristics. The VSET and VRESET values don＇t overlap each other, and the endurance is more than 4000 cycles.
3.3.2 Retention property
For a nonvolatile memory, the data storage time, called retention time, is an important indicator [71]-[77]. It means that how long the stored resistance values can be remained. In order to accelerate the degradation of Pd/HfOX/TiN structure, the retention test is also tested under the 85℃. As showed in Fig. 3-59, the retention time of the two states is at 1000 seconds and the resistance ratio only over 1 by 0.08 V read bias. Afterward, in order to accelerate the degradation of Pd/Al(35 Å)/HfOX/TiN structure, the retention test is also tested under the 175℃, as show in Fig. 3-60. Retention time at the high temperature is at least over 5×104 seconds, and the resistance ratio still remains over 50.
3.3.3 Thermal stability
Current transport mechanisms for the LRS and the HRS were analyzed by measuring the I-V curve at the temperature ranging from 25 to 175 ℃. For the LRS, the I-V curve follows Ohm＇s law and the resistivity of the HfOX films shows slightly negative temperature dependence, as shown in Fig. 3-61. Metallic filaments in HfOX films induced
by the set process may be responsible for this observation. On the other hand, the I-V curve of the HRS, as shown in Fig. 3-62, has positive and weak temperature dependence at low bias. This positive temperature dependence for the HRS current suggests that the dominant current transport mechanism for the Pd/Al/HfOX/TiN structure during set process may be tunneling, Schottky emission, or Pool-Frenkel emission. However, we ruled out Schottky emission and Pool-Frenkel emission as possible candidates due to the very weak temperature dependence of the HRS current [36], [78]-[84].
3.3.4 Curve fitting of current-voltage plots
Figure 3-63 and Fig. 3-64 show the plot of ln |J| versus ln |V| of our Pd/Al(35 Å)/HfOX/TiN structure. In the LRS current and HRS current at small bias region, the curve present Ohmic behavior whose slopes are close to unity. The ln |J/V2| versus 1/V curve is presented in Fig. 3-65. The linear line with a negative slope at the large bias region of the plot shows that the HRS current at large bias is due to the Fowler-Nordheim (FN) tunneling [36], [85]-[93].
In the fitting curve, the filament model for the resistive switching in the Pd/Al(35 Å)/HfOX/TiN structure can be described as follows: When HfOX films is destroyed during the forming process, the conductive path is created. An Ohmic current-voltage behavior is observed for devices. As the device was switched back to the HRS, the HfOX films near the anode is recovered by the partial rupture of filaments near anode/ HfOX films interface.
3.3.5 Material analyses
In order to analyze Al/HfOX interface and provide extra information for the elucidation of the switching mechanism, XPS, EDS, and TEM are taken to inspect the
atomic percentage, binding energy in the samples and the cross section view. The further details are described as below.
The depth profile of the chemical composition of Pd/Al/HfOX structure measured by XPS is shown in Fig. 3-66 and Fig. 3-67. We could see that O2 and Al at the same depth, which is formed AlOX at interface. The binding energy of Pd/Al/HfOX structure measured by XPS is shown Fig. 3-68. The primitive binding energy of Al2O3 is 74.4 eV. The peak of plot is close to 74.4 eV, which was demonstrated the AlOX is formed at interface.
Figure 3-69 is the TEM micrograth of Pd/Al/HfOX/TiN/Ti/SiO2/Si structure. Fig. 3-70 is high resolution TEM image of Al/HfOX interface, which shows Al thickness from 3.5 nm to 9 nm. Therefore, we could demonstrate AlOX which was formed at Al/HfOX interface.
Figure 3-71 is EDS analysis of the Al layer, which shows the Al layer has oxygen content. Therefore, we could demonstrate AlOX which was formed at Al/HfOX interface.
3.3.6 Electrical properties of Pd/Ti/HfOX/TiN structure
In order to validate our argument, we decided to change the reactive metal. The structure is Pd/Ti(30 Å)/HfOX/TiN. In the device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-72 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-73. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-74. The Pd/Ti(30Å)/HfOX/TiN device shown good resistive switching characteristics. The VSET and VRESET values don＇t overlap each other, and the endurance is more than 350 cycles.