In this section, we will verify our conjecture and the condition is at 300ns still is used to in the experiment. At first, the normally on-state occurred because the time of
77
recombination is not enough and heating effect. We have to change the condition of pulse cycle and then make the device achieving the enhancement of endurance. Here, we define some parameters; one is the width between reset and set is called “a-region”
and the other one is the width between set and “next reset” is called “b-region”.
Fig-4.2-6 shows the condition. The changing condition a-region and b-region are extended or curtailed. Subsequently, we try to test the optimization condition and then apply the condition to analyze the effect, further. Therefore, we set three different a-region, a1=800ns, a2=15us and a3=20ns and the four different b-region, b1=800ns, b2=20ns, b3=20us, b4=15us. Next, five different compositions are applied to the device and then each pair of parameters is like Fig-4.2-7. The condition is inputted the device and is given 5000 cycles per time to the device. The device is biased ten times and then we gather statistics which is the distribution of on/off ratio. Fig-4.2-7 shows what we said the statistics. Each pair of parameters possesses some different effects from Fig-4.2-7. And Fig-4.2-7 indicates the optimization condition is the second pair of parameters (i.e. a2 and b2). This condition possesses lower leakage and lager on/off ratio so the condition is applied to the further experiment.
Subsequently, the long a-region and the short b-region are as the standard condition.
Fig-4.2-8 depicts the new changeable condition. And the a-region is modified another width is at 3.5us, 4.5us, 5.5us, 7.5us and 9.5us and the b-region is modified width is fixed at 20ns. Fig-4.2-9 indicates a special phenomenon which the on/off ratio is enlarged when the a-region lengthening gradually. On the other hand, the on-state independent with various a-regions and the off-state depend on the various a-regions.
(Conclusion)
From Fig-4.2-7 and Fig-4.2-9, the phenomena verify our conjectural the model at the section is 4-2.2. Here, the pulse cycle effect is explained by the lattice and non-lattice oxygen model. The lattice oxygen ions are the oxygen ions bond with the
78
oxygen vacancies and the non-lattice oxygen ions are the oxygen ions don’t bond with the oxygen vacancies. The amount of non-lattice oxygen ions is too many to rupture the conduction paths. On the contrary, conduction paths can be easy ruptured if the amount of lattice oxygen ions is more than the non-lattice because the non-lattice oxygen ions are simply captured at TiN. There are two reasons can cause the non-lattice oxygen increasing, first is the time of recombination and second is heating effect. The transition process can be described by the oxygen ions reaction. First, we can see the set process. In fact, the set process is a single reactive step which means the oxygen ions is pushed to the electrode (i.e. TiN). However, the reset process has to possess two reactive steps, first is the oxygen have to hop to the barrier from TiN to bulk and second is the oxygen ion recombine with the oxygen vacancy. Consequently, the time of recombine is not enough and then the non-lattice oxygen ions are produced. Here, we think the first step of reset process has to bias to the device but the second step is unnecessarily biasing the device. We have to increase a-region in order to achieve complete recombination because the time of recombination time is longer. The non-lattice oxygen effect is not serious at less pulse cycles but the effect is serious at many pulse cycles. On the other hand, the conduction paths can’t be ruptured because too many the non-lattice oxygen ions are captured. We can see Fig-4.2-5 shows we discuss the model. The heating effect existed in many pulse cycles but the effect make the oxygen ions possess too high mobility to recombine with vacancy. Therefore, the recombination and the heating effect determine the pulse cycle effect so we have to solve the two problems to avoid the normally on-state phenomenon. The a- and b-region have to be extended to achieve steady state.
However, we also find a batter pair of parameters which is the combination of a2 and b2. Fig-4.2-9 depicts we mentioned the model. The off-state increases with increasing the a-region because the amount of non-lattice oxygen ions are captured at TiN is less.
79
The oxygen ion can achieve complete bonding with vacancy if the enough recombination time and decreasing heating effect so the amount of non-lattice oxygen is decreased. Fig-4.2-9 indicates the non-lattice ions are relative to the width of a-region. On the contrary, the ions can’t achieve a situation of bonding with vacancy if the recombination time is not enough and the heating effect still exists.
Consequently, the state will remain on the on-state in the condition of many pulse cycles. The phenomenon is like Fig-4.2-3.