三、 Principle
3.5 The key physical parameters influence RRAM performance
3.5.5 Electric field
In RRAM, the resistance switching performance has much to do with the applied voltage. It also means that the electric field plays an important role in the resistance switching characteristics. Figure 3-44 shows of the influence of material surrounding environment on electric field[42]. The electric field enhancement structure is shown in the left region of figure 3-44. Also, the electric field distribution of 20 nm and 100 nm WOx RRAM device is shown in the right region of this figure with the same applied voltage. In this figure, it is clear to see that the electric field of 20 nm device is higher than 100nm device. The forming voltage also shows the decreasing relation with the device cell in this experiment. This result indicates the electric field in RRAM material directly influences the RRAM performance.
62
Figure 3-44. Electric enhancement structure (left) and the electric field distribution for 20nm (up-right) and 100nm (down-right) RRAM devices[42].
The dielectric constant of oxide layer also influences the electric field. Figure 3-45 shows the relationship between the forming voltage and dielectric constant[38]. In this figure, it is clearly shown that “forming” steps of RRAM can be regarded as soft-breakdown phenomenon of dielectric films controlled by applied electric field.
Because different dielectric constant materials show different electric field in oxidation layer, the forming voltage also shows different values. In the previous results, the electric field in the oxidation layer plays an important role in the RRAM performance.
Figure 3-45. Forming voltage verse dielectric constant of RRAM[38].
63
3.5.6 Density of Oxygen Vacancies
Oxygen vacancy is a key element in the resistance switching mechanism, and many studies report that it also influences the RRAM performance. In our research of downstream plasma (DP) experiments, the initial resistance depends on the DP time, and oxygen vacancies depend on the DP time. Figure 3-46 shows the WOx-based unipolar RRAM characteristic with different DP samples. Here, all the samples are forming-free and the initial resistance shows increasing relationship as DP increases.
However, this WOx-based RRAM with RTA oxidation process sample doesn’t show the forming-free characteristic. These results indicate that the oxygen vacancy is an important element in the resistance switching performance.
Figure 3-46. The unipolar operation character with different oxidation time.
64
Chapter 4 : Results and Discussion
In our experiments, we studied several materials such as nickel oxide (NiOx), titanium oxide (TiOx), and tungsten oxide (WOx) for the RRAM. NiOx, TiOx and WOx
are suitable RRAM materials because they show stable resistive switching character.
Moreover, due to the clear resistive switching character, simple structure, and CMOS fully compatible process of these three materials, presently many researchers still focus on these materials.
4.1 Outline of result
NiOx, TiOx and WOx all exhibited bistable resistive switching character by the bipolar operation. The resistance ratio (also called on/off ratio), cycle endurance, and data retention were clearly observed in the reliability test. Moreover, the WOx memory device was fully prepared by the semiconductor process, and it exhibits excellent performance in the reliability test. We believe that WOx is promising to replace the present non-volatile memory.
The polycrystalline NiOx RRAM showed bipolar resistive switching character, and electric character showed the relationship between thickness and oxygen content. The thicker thickness samples and higher oxygen content samples exhibit better performance in the reliability test. Also, the conduction mechanism can be well explained by Schottky emission and the resistive switching character shows the barrier high dependent relationship. Moreover, the dielectric constant of NiOx shows the relationship with oxygen ratio and it also influences the efficiency of applied bias.
The TiOx also showed bipolar resistive switching character and the conduction mechanism followed the Schottky emission. The electric character showed the barrier high dependence relationship. Moreover, the resistive switching behavior shows that thickness has little to do with trait. These results indicated the interface contribution in the resistive switching character. In addition, the TiOx/SiO2 hybrid system experiment indicated the interface contribution and this system exhibits better reliability performance.
For WOx RRAM, the resistive switching behavior was observed not only by bipolar operation but also by unipolar operation. The electrical character of WOx
followed VRH in HRS and the electrical behavior of LRS is close to minimum-metal-conductivity (MMC). It also showed the barrier high dependent relationship with the resistance state. Its excellent performance includes high on/off ratio (>1000), good endurance (>1000), high thermal stability (>2000 hrs at 250℃), good read disturb (>1000 sec, at 1 V), high speed operation (<10 ns), small size (~9 nm), low current consumption (<10 uA), and so on. What’s better, it can also be used in one-time programming (OTP) or multi-level cell (MLC) applications.
65
4.2 NiOx-based RRAM
For NiOx-based RRAM, there are three parts, including the material analysis, resistive switching character, and conduction mechanism analysis, which will be discussed in this section.
First, material analyses include the microstructure. Figure 4-1 shows the HRTEM image of our NiOx film in the metal-insulator-metal (MIM) structure. According to this image, we can observe clearly (especially in the enlarge image of figure 4-2) that the polycrystalline NiOx is everywhere in this film. Moreover, the SAED image (inset of figure 4-2) also shows the polycrystalline structure in NiOx film. Several studies[79-81]
also reported the polycrystalline structure in their NiOx RRAM research.
Figure 4-1. HRTEM image of polycrystalline NiOx film
66
Figure 4-2. The enlarged part (dash mark in fig.4-1) of polycrystalline NiOx film. Inset shows the SAED image.
Second, for resistive switching character, both unipolar and bipolar switching characters have been demonstrated in other studies[82-84]. These studies indicated the bistable resistive switching character in NiOx film. For the bipolar resistive switching NiOx RRAM, the resistance state is dependent on the direction of the applied voltage.
The HRS and LRS is formed by opposite bias. For the unipolar resistive switching NiOx
RRAM, the resistance state is dependent on the amount of applied voltage. The LRS and HRS is formed by large and small applied voltage, respectively.
In this study, we only found the bipolar resistive switching character in NiOx film.
Figure 4-3 shows the thickness-dependent bipolar switching phenomenon of NiOx films.
In our experiment, the thickness above 30 nm was necessary because the bistable resistive switching phenomenon disappeared when the thickness was below 30 nm.
With the increase of thickness, the more obvious resistive switching phenomenon was observed. Moreover, these thicker samples exhibited larger on/off ratio in the cycle endurance test. Accordingly, we defined the bistable states as high resistance state (HRS and low resistance state (LRS). In the programming operation, the resistance state is changed from HRS to LRS with an applied positive bias voltage. Similarly, the resistance state is changed from LRS to HRS by means of an applied negative bias voltage.
67
-7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 1E-12
1E-11 1E-10 1E-9 1E-8 1E-7 1E-6 1E-5 1E-4
Current (A)
Voltage (V)
30nm 75nm 150nm
Figure 4-3. The resistive switching character of NiOx films.
Figure 4-4 shows the relationship between sample thickness and the on/off ratio.
The ratio shows the thickness effect in this figure. The on/off ratio is about 0.5 when the thickness is about 30 nm. As the thickness increases, the on/off ratio also increased. The on/off ratio is about 100 when the thickness increases to 150 nm. This result indicates the resistance switching performance dependent on the thickness of NiOx.
Figure 4-4. The relationship between sample thickness and on/off ratio.
68
Figure 4-5 shows more than 200 times of resistive switching processes of cycle endurance test of thick (75 nm and 150 nm) NiOx films. In the endurance test, it was obvious that the resistive state is rising in this cycling process. It is attributed to the damage of the NiOx film and this damage induced the rising of resistance state. This phenomenon was also reported by Rossel et al.[52] and it is attributed to the localized
“burned” pits with create-like geometry. Therefore, the on/off ratio also showed a dropping in the endurance measurement. Figure 4-6 shows the on/off ratio dropping phenomenon and it indicates the on/off ratio dependence on the thickness of NiOx.
Figure 4-5. The endurance test of different thickness NiOx films.
Figure 4-6. The on/off ratio of different thickness NiOx films in the endurance measurement.
69
The oxygen content of NiOx film is known to play an important role in resistive switching character. Figure 4-7 shows the resistive switching character (I-V curve) of 150 nm NiOx film with various oxygen flow ratio. In this experiment, we defined O2/(Ar+O2) as oxygen flow ratio and this ratio shows a strong relationship with on/off ratio. The relationship between oxygen flow ratio and the on/off ratio is shown in figure 4-8. It is clear to see that the on/off ratio dramatically increases from X3 to X100 as we increase oxygen flow ratio from 33% to 43%. In this figure, it indicates that the high oxygen flow ratio sample exhibits better performance.
-20 -15 -10 -5 0 5 10 15 20
Figure 4-7. The resistive switching character of NiOx film with different oxygen flow ratio.
Figure 4-8. The relationship between on/off ratio and oxygen flow ratio.
70
Electrical analysis found the conduction mechanism was the Schottky emission in NiOx. Figure 4-9 shows the Schottky emission fitting curve in our sample. The linear dependency indicates that the transport mechanism is strongly dependent on the barrier height and the dielectric constant, as is shown in Eq. (3-2). The calculation of dielectric constants and the (B), which was defined by the barrier high difference between HRS and LRS by fitting the Schottky emission, were summarized in figure 4-10. The trend of
(B) is similar to the on/off ratio. It indicates the strong correlation between barrier high and on/off ratio. Several studies[85,86] also reported the relationship between barrier high and the resistive switching character. Moreover, a very high dielectric constant was obtained in the samples with oxygen flow ratio less than 33%, and the opposite case was observed in the sample with oxygen flow ratio higher than 43%. A sample with higher dielectric constant means it owns higher charge storage capability, which also influences the electric character in the resistive switching process.
2.0 2.5 3.0 3.5
10 12 14
ln ( J ) ( A/cm2 )
V1/2
Figure 4-9. Schottky emission approximation of NiOx film.
71
20 30 40 50 60
10 100 1000
Set Reset
()
O2 / Ar+O
2 (%)
i ()
Figure 4-10. Dielectric constant and the barrier high change with various oxygen flow ratio.
Figure 4-11 illustrates how the internal electric field affects the resistive switching process for samples with various oxygen flow ratio. If we apply the same voltage in two different permittivity samples, the real voltage in those samples will be different. For the high permittivity sample, it can storage more charges in the interface and those charges form an opposite voltage in the insulator at the same time. Therefore, the real voltage in the sample is lower than the applied voltage. Similarly, there are small opposite voltage formed in the insulator of the low permittivity sample. This might be attributed to the fact that the samples with few oxygen ratios needed higher voltage to switch the resistance state.
Figure 4-11. The influence of dielectric constant in the operation process.
72
Figure 4-12. The temperature dependent electric character of NiOx film.
Figure 4-12 shows the temperature-dependent electric character of NiOx film. It is clear to see the resistance decreases when the temperature decreases from room-temperature to 4K. The resistance of NiOx closes to linear decreasing trend when the temperature was above 60K. The minimum resistance appeared when the temperature was below 60K and it remained the same one. This result indicated a metal conduction character of our NiOx film.
In summary, NiOx shows the polycrystalline microstructure. The I-V curve shows clear bipolar resistive switching character. Both sample thickness and oxygen content influenced the bistable resistive switching phenomenon. The conduction mechanism follows the Schottky emission and the resistive switching character shows barrier high dependence relationship. Due to the temperature-dependent electric character, the NiOx
film showed a metal conduction character.
73
4.3 TiOx-based RRAM
For TiOx–based RRAM study in this section, we discuss three parts, including resistive switching character, conduction mechanism, and the interface contribution.
Figure 4-13 shows the schematic core-section of TiOx-based RRAM. The titanium oxide film is placed on the top of the 0.18 um diameter W-plug, and this W-plug is also called contact (CT) in the semiconductor fabrication. This CT is the connection between the memory cell and the source site, and the memory state is dependent on the resistance of titanium oxide.
Figure 4-13. Cross-sectional view of TiOx-based RRAM.
Figure 4-14. The bipolar resistive switching characterics of TiOx film.
First, the electric character exhibits the bipolar resistive switching character, which is shown in figure 4-14. The positive switching voltage is about +5 V and the negatvie switching voltage is about -2 V. This asymmetrical resistive switching behavior is different with NiOx film. Park et al.[87] also reported such an asymmetrical resistivite switching character in TiOx film.
74
Figure 4-15. The resistive switching phenomenon from high resistance state to low resistance state.
Figure 4-15 shows the resistive switching phenomenon from HRS to LRS with linear scale of current. We can find a sudden rise at +5.5V in the rising curve and the CF is formed at the same time.
Figure 4-16. Voltage dependent on/off ratio.
Figure 4-16 shows the relationship between the applied voltage and on/off ratio. In this figure, we found the maximum on/off ratio is 20 at the applied voltage about 4 V.
Also, we define this applied voltage as the transition voltage (VT) and we can use this
75
parameter to observe the electric character. The reading voltage here must be below the VT to avoid influencing the resistance stste.
Figure 4-17 shows resistive switching characters of TiOx at various thicknesses.
The electric characterstics indicate that these resistance switching behaviors are in thickness-indepentent relationship. All three I-V curves show similar resistive switching behavior and the resistance of these samples also shows thickness independent.
Figure 4-17. The resistive switching character of TiOx film at various thicknesses.
Moreover, both on/off ratio and VT are independent with sample thicknessas shown in figure 4-18. These results indicate the interface contribution of this TiOx material.
Figure 4-18. The thickness relationshop with transition voltage and the on/off ratio.
76
The curve fitting shown in figure 4-19 shows that all TiOx films obey the Schottky emission mechanism. According to this result, we got one more proof of interface contribution in our TiOx-based RRAM.
Figure 4-19. The curving fitting of I-V data for TiOx atvarious thicknesses.
Figure 4-20. The temperature dependent electric characteristics of TiOx film.
Figure 4-20 shows the temperature-dependent electric characteristics. In addition, according to the calculation of Schottky emission curve, we know the barrier high of HRS and LRS are about 0.6 eV and 0.73 eV, respectively.
In order to check this interface contribution, we prepared a hybrid sample with a 40 Å SiO2 layer between TiOx and bottom electrode. This hybrid sample (TiOx-SiO2) shows great improvement in the data retention test, and it also improves the on/off ratio.
These improvements are shown in figure 4-21. In this figure, the LRS and HRS almost overlap, and the on/off ratio almost disappears after 100-sec retention test for TiOx film.
However, the on/off ratio always keeps at about 20 times for the hybrid sample.
77
Figure 4-21. Data retention of TiOx-based RRAM with/without SiO2 hybrid system.
Figure 4-22 shows the cycle endurance test of TiOx-based RRAM. In this figure, the resistance increases in the cycling test. It is attributed to the damage of the TiOx film and this damage induces the rising of resistance. This result is similar to the NiOx film.
After 50 times of cycling operation, the LRS is close to the initial of HRS.
Figure 4-22. Cycle endurance test of TiOx-based RRAM.
In hybrid system, the cycle endurance performance is also improved. Figure 4-23 shows the cycle endurance test of TiOx-based RRAM with SiO2 hybrid system. It is clear to see both LRS and HRS keep their resistance in the cycling operation test. This figure shows above 1000 times of cycle endurance performance of our hybrid system, and the on/off ration keeps at about 100X in this test. This hybrid system experiment also indicates the interface contribution of TiOx RRAM.
78
Figure 4-23. Cycle endurance test of TiOx-based RRAM with SiO2 hybrid system.
Figure 4-24 shows the read disturb test of the hybrid system. The reading voltage is only 50 mV to avoid damaging the memory cell. This figure indicates over 500 times read disturb performance of our hybrid system. Both LRS and HRS keep their resistance state, and the on/off ratio almost keeps at about 200X in this test. In this figure, the HRS decreases slowly in the read disturb test. It indicates that small reading voltage influences the HRS.
Figure 4-24. Read disturb test of TiOx-based RRAM with SiO2 hybrid system.
In summary, the TiOx exhibits asymmetrical bipolar resistive switching character, and it shows interface contribution. The conduction mechanism follows the Schottky emission and the resistive switching character shows barrier high dependence relationship. The barrier high of LRA and HRS are about 0.6 and 0.73 eV, respectively.
Moreover, the TiOx-SiO2 hybrid system also indicates the interface contribution, which improves the data retention cycle endurance and read disturb performance.
79
4.4 WOx-based RRAM
WOx-based RRAM has been studies by several research groups[24,28,88-91]
. In the section, we discuss the process flow, microstructure, composition analysis, conduction mechanism, and the application for WOx. Due to its better performance and widely applications, this material is promising for the next general nonvolatile memory applications.
Figure 4-25 shows the process flow of WOx-based RRAM. First, we use a dummy wafer as the substrate and then deposit an inter metal dielectric (IMD1) for the isolation with Si wafer. Secondly, the TiN/AlSiCu/TiN was deposited for bottom electrode (ML1) and the SiO2 was deposited for the second inter metal dielectric (IMD2). Thirdly, via pattern is defined by a photo lithography process and the via plug structure was created by via etching process. After the creation of via plug structure, an TiN adhesion layer was deposited and the W deposition was achieved by CVD process. After the W deposition, we used CMP process to make the surface smooth and form a W-plug structure at the same time. Finally, the oxidation process is formed by the down-stream plasma process and the top electrode (ML2) is defined by the second metal line process.
Figure 4-25. The WOx base RRAM process flow
Figure. 4-26 shows the TEM image of WOx film and the inset image shows the W-plug structure. The memory cell is placed on the top of the W-plug and it shows a uniform thickness about 120 Å of WOx film. This image also indicates that our plasma oxidation is a uniform process and it can avoid the leakage of the short channel in the electrical measurement. Figure 4-27 shows the enlarged part of the WOx layer. This image shows that the amorphous WOx is everywhere in the memory cell.
80
Figure 4-26. The TEM image of our WOx film.
Figure 4-27. The enlarge image of TEM
Figure 4-28 shows the XPS data of WOx film for O and W analysis. Figure 4-28 (a)~(d) shows the peak profile of the surface at the depth of 15Å , 70Å , and 140Å , respectively. In this figure, the WO3 peak on the wafer surface (a) and the pure W peak
Figure 4-28 shows the XPS data of WOx film for O and W analysis. Figure 4-28 (a)~(d) shows the peak profile of the surface at the depth of 15Å , 70Å , and 140Å , respectively. In this figure, the WO3 peak on the wafer surface (a) and the pure W peak