Limits...
Impact of device size and thickness of Al2O 3 film on the Cu pillar and resistive switching characteristics for 3D cross-point memory application.

Panja R, Roy S, Jana D, Maikap S - Nanoscale Res Lett (2014)

Bottom Line: The 8-μm devices show 100% yield of Cu pillars, whereas only 74% successful is observed for the 0.4-μm devices, because smaller size devices have higher Joule heating effect and larger size devices show long read endurance of 10(5) cycles at a high read voltage of -1.5 V.On the other hand, the resistive switching memory characteristics of the 0.4-μm devices with a 2-nm-thick Al2O3 film show superior as compared to those of both the larger device sizes and thicker (10 nm) Al2O3 film, owing to higher Cu diffusion rate for the larger size and thicker Al2O3 film.This conductive bridging resistive random access memory (CBRAM) device is forming free at a current compliance (CC) of 30 μA (even at a lowest CC of 0.1 μA) and operation voltage of ±3 V at a high resistance ratio of >10(4).

View Article: PubMed Central - PubMed

Affiliation: Thin Film Nano Tech. Lab., Department of Electronic Engineering, Chang Gung University, 259 Wen-Hwa 1st Rd., Kwei-Shan, Tao-Yuan, 333, Taiwan, panjarajeswar@gmail.com.

ABSTRACT
Impact of the device size and thickness of Al2O3 film on the Cu pillars and resistive switching memory characteristics of the Al/Cu/Al2O3/TiN structures have been investigated for the first time. The memory device size and thickness of Al2O3 of 18 nm are observed by transmission electron microscope image. The 20-nm-thick Al2O3 films have been used for the Cu pillar formation (i.e., stronger Cu filaments) in the Al/Cu/Al2O3/TiN structures, which can be used for three-dimensional (3D) cross-point architecture as reported previously Nanoscale Res. Lett.9:366, 2014. Fifty randomly picked devices with sizes ranging from 8 × 8 to 0.4 × 0.4 μm(2) have been measured. The 8-μm devices show 100% yield of Cu pillars, whereas only 74% successful is observed for the 0.4-μm devices, because smaller size devices have higher Joule heating effect and larger size devices show long read endurance of 10(5) cycles at a high read voltage of -1.5 V. On the other hand, the resistive switching memory characteristics of the 0.4-μm devices with a 2-nm-thick Al2O3 film show superior as compared to those of both the larger device sizes and thicker (10 nm) Al2O3 film, owing to higher Cu diffusion rate for the larger size and thicker Al2O3 film. In consequence, higher device-to-device uniformity of 88% and lower average RESET current of approximately 328 μA are observed for the 0.4-μm devices with a 2-nm-thick Al2O3 film. Data retention capability of our memory device of >48 h makes it a promising one for future nanoscale nonvolatile application. This conductive bridging resistive random access memory (CBRAM) device is forming free at a current compliance (CC) of 30 μA (even at a lowest CC of 0.1 μA) and operation voltage of ±3 V at a high resistance ratio of >10(4).

No MeSH data available.


Related in: MedlinePlus

Current-voltage characteristics of the Cu pillars.I-V characteristics of arbitrarily measured 50 devices with device size of (a) 8 × 8 μm2 under a high CC of 70 mA and of (b) 0.4 × 0.4 μm2 under a CC of 10 mA. The smallest size devices have largest failure of the Cu pillars, due to the Joule heating. The thickness of the Al2O3 layer is 20 nm.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4494016&req=5

Fig2: Current-voltage characteristics of the Cu pillars.I-V characteristics of arbitrarily measured 50 devices with device size of (a) 8 × 8 μm2 under a high CC of 70 mA and of (b) 0.4 × 0.4 μm2 under a CC of 10 mA. The smallest size devices have largest failure of the Cu pillars, due to the Joule heating. The thickness of the Al2O3 layer is 20 nm.

Mentions: The I-V characteristics of randomly measured 50 pristine devices with two different sizes viz. 8 × 8 and 0.4 × 0.4 μm2 are shown in Figure 2. The thickness of Al2O3 film is 20 nm. The sweeping voltage direction is shown by the arrows 1 to 4, which also follows as 0 → +5 → 0 → -1.1 → 0 and 0 → +8 → 0 → -1 → 0 V for the devices with large and small sizes, respectively (Figure 2a,b). It is found that all 8-μm devices are operated at a high CC of 70 mA whereas many of the 0.4-μm devices show failure to reach even at a CC of 10 mA. By applying bias of -1 V on the TE, the 8-μm devices do not show RESET and few 0.4-μm devices show RESET. This suggests that the Joule heating burns the small size devices at a high current as well as device size-dependent filament diameter. Heat dissipation of larger size devices is higher than the smaller size devices. Thermal conductivities of Cu, Al, Al2O3, SiO2, TiN, and Si materials are 398, 244, 25.08, 1.38, 28.84, and 148 W/m/K, respectively [33]. This implies that heat will be dissipated through top electrode contact than the other sides. Therefore, the area of top electrode contact as well as device size will help to reduce heating effect, especially, when the device is operated at a high current of >10 mA. If the device does not show RESET, then stronger Cu filament (or pillar) is formed into the Al2O3 layer. The formation voltages (Vform) for the 8-, 4-, 2-, and 0.4-μm devices at 50% probability are 4.2, 4.5, 4.9, and 5.5 V, respectively (Figure 3a). Therefore, the value of Vform increases with decreasing the device sizes owing to lower leakage current as well as lower defects into the Al2O3 layer. On the other hand, the formation energy is lower for larger size devices than the smaller one owing to the higher diffusion rate of Cu ions with the area. The similar phenomena of Ag diffusion in SiO2 layer by in situ TEM observation have been reported by Yang et al. [34]. The Cu diffusion in ZrO2 layer by TEM observation was also reported by other group [35]. The number of successful devices with different device sizes ranging from 0.4 × 0.4 to 8 × 8 μm2 is shown in Table 1. The device size of less than 2 μm can carry current of 10 mA, while the larger size of 4-μm device can carry high current of 70 mA. Most important thing is that the larger size devices show 100% success, while the failure is increased with decreasing device size. It is expected that stronger Cu pillar is needed for 3D integration of cross-point nonvolatile memory. This will be easy way and low cost for application of 3D cross-point memory [32]. Therefore, we need those devices which can sustain at high current for long time, and we find that, the devices with large area are compatible for this purpose. Figure 3b shows the statistical distribution of currents at low resistance state (LRS) for the device-to-devices. The mean value and the standard deviation (σ) of currents for the 4-μm devices at a read voltage (Vread) of 1 V are 49.96 and 9.33 mA, while those values for the 8-μm devices are 46.14 and 6.61 mA, respectively. The read current of the 8-μm devices is slightly lower than that of the 4-μm devices owing to lower formation voltage. This implies that small amount of Cu diffusion into the Al2O3 films for the larger size devices than the smaller sizes. However, uniformity of the high current carrying Cu pillars is better for the 8-μm devices than those of the 4-μm devices. The mechanism of Cu pillar formation inside the pristine device is as follows. These are basically the CBRAM devices; however, 20-nm-thick Al2O3 film is studied for demonstration, and further study for real application of the Cu pillars into the 1-μm-thick Al2O3 films is necessary. When the positive bias is applied on the active Cu electrode, the Cuz+ (z = 1,2) ion is formed by oxidation, then those ions migrate through the switching medium in the presence of high electric field, and finally, they become reduced at the TiN BE. This formation process transforms the pristine device from its initial resistance state (IRS) to LRS as well as stronger Cu pillar is formed. By applying negative voltage on the TE, the Cu pillars of some smaller size devices are dissolved because of Joule heating. Robust Cu pillars have been investigated by measuring endurance properties below.Figure 2


Impact of device size and thickness of Al2O 3 film on the Cu pillar and resistive switching characteristics for 3D cross-point memory application.

Panja R, Roy S, Jana D, Maikap S - Nanoscale Res Lett (2014)

Current-voltage characteristics of the Cu pillars.I-V characteristics of arbitrarily measured 50 devices with device size of (a) 8 × 8 μm2 under a high CC of 70 mA and of (b) 0.4 × 0.4 μm2 under a CC of 10 mA. The smallest size devices have largest failure of the Cu pillars, due to the Joule heating. The thickness of the Al2O3 layer is 20 nm.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4494016&req=5

Fig2: Current-voltage characteristics of the Cu pillars.I-V characteristics of arbitrarily measured 50 devices with device size of (a) 8 × 8 μm2 under a high CC of 70 mA and of (b) 0.4 × 0.4 μm2 under a CC of 10 mA. The smallest size devices have largest failure of the Cu pillars, due to the Joule heating. The thickness of the Al2O3 layer is 20 nm.
Mentions: The I-V characteristics of randomly measured 50 pristine devices with two different sizes viz. 8 × 8 and 0.4 × 0.4 μm2 are shown in Figure 2. The thickness of Al2O3 film is 20 nm. The sweeping voltage direction is shown by the arrows 1 to 4, which also follows as 0 → +5 → 0 → -1.1 → 0 and 0 → +8 → 0 → -1 → 0 V for the devices with large and small sizes, respectively (Figure 2a,b). It is found that all 8-μm devices are operated at a high CC of 70 mA whereas many of the 0.4-μm devices show failure to reach even at a CC of 10 mA. By applying bias of -1 V on the TE, the 8-μm devices do not show RESET and few 0.4-μm devices show RESET. This suggests that the Joule heating burns the small size devices at a high current as well as device size-dependent filament diameter. Heat dissipation of larger size devices is higher than the smaller size devices. Thermal conductivities of Cu, Al, Al2O3, SiO2, TiN, and Si materials are 398, 244, 25.08, 1.38, 28.84, and 148 W/m/K, respectively [33]. This implies that heat will be dissipated through top electrode contact than the other sides. Therefore, the area of top electrode contact as well as device size will help to reduce heating effect, especially, when the device is operated at a high current of >10 mA. If the device does not show RESET, then stronger Cu filament (or pillar) is formed into the Al2O3 layer. The formation voltages (Vform) for the 8-, 4-, 2-, and 0.4-μm devices at 50% probability are 4.2, 4.5, 4.9, and 5.5 V, respectively (Figure 3a). Therefore, the value of Vform increases with decreasing the device sizes owing to lower leakage current as well as lower defects into the Al2O3 layer. On the other hand, the formation energy is lower for larger size devices than the smaller one owing to the higher diffusion rate of Cu ions with the area. The similar phenomena of Ag diffusion in SiO2 layer by in situ TEM observation have been reported by Yang et al. [34]. The Cu diffusion in ZrO2 layer by TEM observation was also reported by other group [35]. The number of successful devices with different device sizes ranging from 0.4 × 0.4 to 8 × 8 μm2 is shown in Table 1. The device size of less than 2 μm can carry current of 10 mA, while the larger size of 4-μm device can carry high current of 70 mA. Most important thing is that the larger size devices show 100% success, while the failure is increased with decreasing device size. It is expected that stronger Cu pillar is needed for 3D integration of cross-point nonvolatile memory. This will be easy way and low cost for application of 3D cross-point memory [32]. Therefore, we need those devices which can sustain at high current for long time, and we find that, the devices with large area are compatible for this purpose. Figure 3b shows the statistical distribution of currents at low resistance state (LRS) for the device-to-devices. The mean value and the standard deviation (σ) of currents for the 4-μm devices at a read voltage (Vread) of 1 V are 49.96 and 9.33 mA, while those values for the 8-μm devices are 46.14 and 6.61 mA, respectively. The read current of the 8-μm devices is slightly lower than that of the 4-μm devices owing to lower formation voltage. This implies that small amount of Cu diffusion into the Al2O3 films for the larger size devices than the smaller sizes. However, uniformity of the high current carrying Cu pillars is better for the 8-μm devices than those of the 4-μm devices. The mechanism of Cu pillar formation inside the pristine device is as follows. These are basically the CBRAM devices; however, 20-nm-thick Al2O3 film is studied for demonstration, and further study for real application of the Cu pillars into the 1-μm-thick Al2O3 films is necessary. When the positive bias is applied on the active Cu electrode, the Cuz+ (z = 1,2) ion is formed by oxidation, then those ions migrate through the switching medium in the presence of high electric field, and finally, they become reduced at the TiN BE. This formation process transforms the pristine device from its initial resistance state (IRS) to LRS as well as stronger Cu pillar is formed. By applying negative voltage on the TE, the Cu pillars of some smaller size devices are dissolved because of Joule heating. Robust Cu pillars have been investigated by measuring endurance properties below.Figure 2

Bottom Line: The 8-μm devices show 100% yield of Cu pillars, whereas only 74% successful is observed for the 0.4-μm devices, because smaller size devices have higher Joule heating effect and larger size devices show long read endurance of 10(5) cycles at a high read voltage of -1.5 V.On the other hand, the resistive switching memory characteristics of the 0.4-μm devices with a 2-nm-thick Al2O3 film show superior as compared to those of both the larger device sizes and thicker (10 nm) Al2O3 film, owing to higher Cu diffusion rate for the larger size and thicker Al2O3 film.This conductive bridging resistive random access memory (CBRAM) device is forming free at a current compliance (CC) of 30 μA (even at a lowest CC of 0.1 μA) and operation voltage of ±3 V at a high resistance ratio of >10(4).

View Article: PubMed Central - PubMed

Affiliation: Thin Film Nano Tech. Lab., Department of Electronic Engineering, Chang Gung University, 259 Wen-Hwa 1st Rd., Kwei-Shan, Tao-Yuan, 333, Taiwan, panjarajeswar@gmail.com.

ABSTRACT
Impact of the device size and thickness of Al2O3 film on the Cu pillars and resistive switching memory characteristics of the Al/Cu/Al2O3/TiN structures have been investigated for the first time. The memory device size and thickness of Al2O3 of 18 nm are observed by transmission electron microscope image. The 20-nm-thick Al2O3 films have been used for the Cu pillar formation (i.e., stronger Cu filaments) in the Al/Cu/Al2O3/TiN structures, which can be used for three-dimensional (3D) cross-point architecture as reported previously Nanoscale Res. Lett.9:366, 2014. Fifty randomly picked devices with sizes ranging from 8 × 8 to 0.4 × 0.4 μm(2) have been measured. The 8-μm devices show 100% yield of Cu pillars, whereas only 74% successful is observed for the 0.4-μm devices, because smaller size devices have higher Joule heating effect and larger size devices show long read endurance of 10(5) cycles at a high read voltage of -1.5 V. On the other hand, the resistive switching memory characteristics of the 0.4-μm devices with a 2-nm-thick Al2O3 film show superior as compared to those of both the larger device sizes and thicker (10 nm) Al2O3 film, owing to higher Cu diffusion rate for the larger size and thicker Al2O3 film. In consequence, higher device-to-device uniformity of 88% and lower average RESET current of approximately 328 μA are observed for the 0.4-μm devices with a 2-nm-thick Al2O3 film. Data retention capability of our memory device of >48 h makes it a promising one for future nanoscale nonvolatile application. This conductive bridging resistive random access memory (CBRAM) device is forming free at a current compliance (CC) of 30 μA (even at a lowest CC of 0.1 μA) and operation voltage of ±3 V at a high resistance ratio of >10(4).

No MeSH data available.


Related in: MedlinePlus