In many reports, we know that top electrodes and resistive switching films are great influence on the RRAM [45]-[50]. Therefore, in order to reduce variables, we fixed resistive switching films for HfOX films, then to change the different electrode. In this section, we performed the resistive switching in various top electrodes/HfOX/TiN/Ti/SiO2/Si devices where various top electrodes where Pt, Pd, Cu, Ni, Ti, and Al.
3.2.1 Electrical properties of various top electrodes
At this section, we would show the basic memory device electrical properties of various top electrodes, including resistive switching characteristics, distribution of VSET and VRESET, resistance of HRS and LRS.
In Pt/HfOX/TiN device, first, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-3 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-4. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-5. Second, the compliance current changed 50 µA, and shows the typical I-V curve in Fig. 3-6. Distribution of VSET and VRESET are shown in Fig. 3-7. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-8. Form the results, we demonstrates when Pt as top electrode, which can use the dc voltage sweep with different compliance current. As mentioned in the previous section, smaller compliance current leads to the VSET and VRESET values increases and overlap each other, and resistance of HRS is larger various, but the endurance is more than 1200 cycles. On the contrary the lager compliance current leading to the VSET and VRESET values decreases and don＇t overlap each other, and various resistance of HRS decreases, but the endurance is less than 15 cycles.
In Pd/HfOX/TiN device, first, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-9 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-10. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-11. Second, the compliance current changed 50 µA, and shows the typical I-V curve in Fig. 3-12. Distribution of VSET and VRESET are shown in Fig. 3-13. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-14. Form the results, we demonstrates when Pd as top electrode, which can use the dc voltage sweep with different compliance current. As mentioned in the previous section, smaller compliance current leading to the VSET and VRESET values increases and overlap each other, and resistance of
HRS is larger various, but the endurance is more than 200 cycles. On the contrary the lager compliance current leading to the VSET and VRESET values decreases and don＇t overlap each other, and various resistance of HRS decreases, but the endurance is less than 30 cycles.
In Cu/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-15 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-16. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-17. Form the results, we demonstrates when Cu as top electrode, which only use the dc voltage sweep with mA compliance current. As mentioned in the previous section, lager compliance current leading to the VSET and VRESET values decreases and don＇t overlap each other, and various resistance of HRS decreases, but the endurance is less than 10 cycles.
In Ni/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 5 mA. In Fig. 3-18 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-19. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-20. Form the results, we demonstrates when Ni as top electrode, which only use the dc voltage sweep with mA compliance current. As mentioned in the previous section, lager compliance current leading to the VSET and VRESET values decreases and don＇t overlap each other, various resistance of HRS decreases, and the endurance is more than 250 cycles.
In Ti/HfOX/TiN device, is using the dc voltage sweep method with a current compliance of 3 mA. In Fig. 3-21 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-22. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-23. Form the results, we demonstrates when Ti as top electrode, which only use the dc voltage sweep with mA compliance current. As mentioned in the previous section, lager compliance current leading to the VSET and VRESET values decreases and don＇t overlap
each other, and various resistance of HRS decreases, but the endurance is less than 50 cycles.
In Al/HfOX/TiN device, forming process is formed as shown in Fig. 3-24. After forming process, the device couldn＇t cause resistive switching as shown in Fig. 3-25.
That is to say the resistive switching phenomenon couldn＇t be observed for Al/HfOX/TiN device.
3.2.2 Work function of various metals
According to textbooks, the values of work function φ of various metals are as follows; φPt=5.65 eV, φPd=5.12 eV, φNi=5.15 eV, φCu=4.65 eV, φTi=4.33 eV, and φAl=4.28 eV.
In the early reports, such as Sawa et al., which proposed that a Schottky barrier, the origin of the nonlinearity of the I-V curve, is formed by a low work function metal and p-type semiconductor PCMO, and the resistive switching mechanism on based Schottky barrier barrier modulation was suggested [51]. Afterward, Seo et al., demonstrated that the effective electric field across the NiO film was high enough to cause resistive switching phenomenon while Ohmic contact or a low Schottky barrier formed between the metal-insulator interface. There is no resistive switching phenomenon with Ti top electrode due to the considerable voltage drop across the well-established Schottky barrier at the Ti/NiO interface [52].
In particular, our sample of various top electrodes did not show similar resistive switching characteristics as reported previously. Table 3-1 performs the comparison with the work function and resistive switching parameters of various metals. We demonstrates larger work function metal, which has better electrical properties, and can resistive switching for different compliance current, but there are different electrical properties on
different compliance current. The work function could not be clear to explain this phenomenon.
3.2.3 Free energy of various metals
Recently, electronegativity and free energy is often used to explain the mechanism RRAM. Hasan et al. proposed that under a positive bias, electromigration of oxygen ions forms thicker oxide, which dissociates under a negative bias, causes high and low resistance states, respectively [53]-[54]. Afterward, Lee et al. proposed that the importance of the reaction at the metal/NiO interface is demonstrated in relation to the free energy of oxide formation of electrode metals and Ni [47], [55]-[57].
Figure 3-26 shows Ellingham diagram of representative high-κ dielectrics showing the free energy of formation as a function of temperature, and Fig. 3-27 shows Ellingham diagram plotting the standard free energies of the formation of oxides of the metals. The free energy of oxidation of Al is close to the free energy of oxidation of Hf. Therefore, the Al/HfOX/TiN device showed no resistive switching characteristics. Because Al attracted and trapped more oxygen ions from HfOX films to form interface layer [58]-[68], which resulted the filaments were not ruptured.
The Ti top electrode was investigated, as Ti has slightly larger free energy of oxidation than free energy of oxidation of Hf. Therefore, the Ti/HfOX/TiN device showed resistive switching characteristics, but the endurance is bad. Because Ti still attracted and trapped some oxygen ions from HfOX films to form thin interface layer. When RESET process, the filaments were not ruptured clearly. Therefore, resistance of HRS and ratio decrease.
The Ni top electrode was also investigated, as Ni has higher free energy of oxidation than free energy of oxidation of Hf. Therefore, the Ni/HfOX/TiN device showed good
resistive switching characteristics. Because Ni did not attract and trap oxygen ions from HfOX films.
The Cu top electrode was also investigated, as Cu has higher free energy of oxidation than free energy of oxidation of Hf. However, the Cu/HfOX/TiN device showed bad resistive switching characteristics. Because Cu is a highly diffusive element in semiconductor materials due to its high diffusion coefficient, they could migrate into the NiO bulk from the top electrode and affect the resistive switching characteristics. The formula as follows.
D=D0 exp(-Q/KT) (3-1)
The Pt and Pd top electrodes were also investigated, as Pt and Pd have very high free energy of oxidation than free energy of oxidation of Hf. Therefore, the Pt/HfOX/TiN and Pd/HfOX/TiN device showed good resistive switching characteristics, and could sweep with different compliance current. However, sweep with larger compliance current resulted bad endurance. Because Pt and Pd have too higher free energy, when applied a dc voltage, the oxygen ions is direct from HfOX films to air. Table 3-2 performs the comparison with the free energy of oxide formation of electrode metals and resistive switching parameters of various metals.
In order to confirm our thoughts, we used Pd as top electrode, and injected N2 6 sccm 30 second. In Fig. 3-28 shows the typical I-V curve. Distribution of VSET and VRESET are shown in Fig. 3-29. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig.
3-30. Form the results, we demonstrates the Pd/HfOX/TiN with N2 showed better resistive switching characteristics than the Pd/HfOX/TiN without N2. The endurance is significant addition. Because N2 can block oxygen ions from HfOX films to air.
Afterward, we used TiN as top electrode under a mixed gas atmosphere with a partial pressure of P(Ar) : P(N2 ) = 24 : 9.6. In Fig. 3-31 shows the typical I-V curve. Distribution
of VSET and VRESET are shown in Fig. 3-32. Resistance of HRS and LRS by 0.08 V read bias are shown in Fig. 3-33. Form the results, we demonstrates the TiN/HfOX/TiN showed better resistive switching characteristics than the Ti/HfOX/TiN. The endurance is significant addition. Because TiN did not attract and trap oxygen ions from HfOX films. Table 3-3 performs the comparison with the Pd/HfOX/TiN with N2 versus without N2 and the Ti/HfOX/TiN versus TiN/HfOX/TiN.