In this section, the temperature of the process will be changed when the ZrO2 is sputtered on the Ti/Pt bottom electrode, and the most suitable temperature for the experiment will be determined. The performance of the RRAM will also be described, including high speed switching pulse, high reliability endurance, stress and the retention.
3.1.1 Temperature effect
The resistive switching behavior is found in all kinds of temperature process that sputter ZrO2, including room temperature (25oC) to the high temperature that RF sputter machine can stand (300oC). Although all of the temperature can be switched, there are still some different performances between the different temperatures. As the following description, the forming voltage、current of on-state and off-state、
endurance、dynamic pulse switch will be showed some different.
At first, the ZrO2-base RRAM is showed good characteristic [19][26], but the temperature effect is not described, so the purpose is to know what temperature is the most suitable for the experiment in the future. The ZrO2 is based on the 60mins sputtering time, so the thickness is the same, and the changing factor is the process temperature. It can be seen that all the temperature (25oC, 100oC, 150oC, 200oC,
250oC) has the forming voltage as shown in Fig. 3-1~ Fig. 3-6 , and the initial resistance in Fig. 3-7 that pick up the point at 0.3V can be known that (100oC, 150oC, 200oC, 250oC) is almost not changed, and the 25oC, 300oC are much lower, because the thickness is the same, so the difference of this characteristic is major at the point of the process temperature, although the XRD cannot show too much different of the sample for the lattice factor is all amorphous, but we still can find the most suitable temperature as the following description by checking the electrical property.
First, all the sample is given a high voltage as high as 5V and the 5mA compliance current to forming the sample, this purpose is to turn on the filament that included interface layer and bulk layer. After that, the sample is given a negative 2.5V to turn the state into the off-state, and then given a positive 1.5V to turn the state into the on-state, there is no need to limit the current for the interface layer can be a good compliance current resistance [26], it called self-compliance when switching the device, it’s one of a big advantage that RRAM using the structure of Ti top electrode on the ZrO2.
3.1.2 I-V curve
The I-V curve of the different temperature is shown in Fig. 3-8~ Fig. 3-12, and the on-state and off-state current is also shown in Fig. 3-13. As the plot shown, the on-state current is almost the same, the value is around 1mA. But at the condition of 200oC, the current is steadier for its variation is smaller compared with other condition, and the off-state current is a little bit smaller than other condition. It’s a advantage of this condition for the ratio will be bigger, and the smaller current, the better process for the RRAM when it is made on the smaller scale process as small as
the current 45nm process.
3.1.3 Statistic endurance test
The more cycles sample can operate, the more stability the sample can operate, for the DC sweep is the basic electrical property. Because all of the conditions can be successfully switch at bipolar mode ( positive turn on and negative turn off), and the next step is to know what kinds of condition can have the best stability, the sample is to be test on the statistic endurance test, which is continuous switching until the sample is almost breakdown. The test method is at the first cycle to the 100th cycle, the data will be record, and after the 100th , all the data will be record the last 10 cycles in every 100 cycles. As the endurance plots shown in Fig. 3-14~ Fig. 3-18, the results show that the 200oC also has the best endurance ability in the table 3-1, and it can be achieved over 10000 cycle times, the performance is much better than other conditions, in other words, its stability is the best.
3.1.4 Dynamic pulse test
After the statistic endurance test, it should be know that whether the dynamic pulse test is as good as the statistic test, because the real device is operated in the pulse mode. The dynamic pulse test is operated in the Aiglent 81110A, the measure method is that after the sample is given a pulse, the read voltage is at 0.3V, and continuous read 5 times in the Fig. 3-19. The sample is focus on the condition of 200oC for its DC sweep is better among all the conditions. The test is operated at 10ns, 20ns, 50ns, which is much faster than the current flash memory for its speed is around
us level [78][79].
All of the sample is on the situation of +6V and -3V for the turn on and turn off voltage, the only changing factor is the pulse width. At first, the sample is test at the 10ns in the Fig. 3-20, the soft error can be seen during the switching, for the shorter pulse width, the shorter reaction time the device can be operated, so the turn-on and turn-off mechanism in the filament is not operated completely, that is the reason the performance of the shorter pulse width is not as good as the longer pulse width as following description.
The sample is operated at 20ns, but its performance is also showed that the soft error also appears in the window in the Fig. 3-21 After that, the sample is operated at the 50ns in the Fig. 3-22, and it can shown that at this speed, the window is clean during the 1000 cycle times, there is no soft error during the window, it’s an improvement [74][75], on the other hand, the DC state will keep the same during the pulse test in Fig. 3-23. It showed that the Ti-ZrO2-Pt RRAM structure is a potential next generation non-volatile memory, which is much faster than the flash memory.
3.1.5 Stress test
The 200oC, 60mins–base ZrO2 is measured its stress ability, and the state is not change during the stress test that using the 0.3V to read the data. The on and off state are not changed during the 10000s stress test in Fig. 3-24. On the other hand, after the pulse test 1000 cycle times, the on-state and off-state will be lifted compared with the original stress test, for it stand the pulse voltage, but the state can still be keep the same.
3.1.6 Retention
Fig. 3-25 is the test for the retention, and it shows that the on-state and off-state can be kept until 106s at room temperature environment.