Retention

Retention is also a key feature about nonvolatile memory devices. It stands for the capability of retaining memory data for a long period without any data loss. For a commercially available nonvolatile memory product, the performance of retention is requested to keep ten years. The retention tests were carried out by DC biases which would switch devices to LRS or HRS at RT. Then, the data states were read out at RT once in a given period by applying a 0.2 V (or 0.5V) reading bias and the reading bias must be small enough in order not to vary the existed memory states.

•PtFe (50 nm)

•SiO ₂ (500 nm)

•P-Si (100)

•TiN •TiN •TiN •TiN

•TiN

•TiN

•SiO • SiO _{2 2} (50nm) (50nm)

•Ti

•Ti •Ti •Ti •Ti •Ti

•PtFe (50 nm)

•SiO ₂ (500 nm)

•P-Si (100)

•TiN •TiN •TiN •TiN

•TiN

•TiN

•SiO • SiO _{2 2} (50nm) (50nm)

•Ti

•Ti •Ti •Ti •Ti •Ti

Figure 2.1. Schematic diagram of Ti/TiN/SiO

2/PtFe/SiO2/Si structure, and its small size structure.

•P • P- -type Si type Si (100) substrate (100) substrate

•Wet oxidation (500nm) • Wet oxidation (500nm)

•FePt • FePt (50nm) (50nm)

• •50nm PECVD oxide 50nm PECVD oxide

•RCA clean • RCA clean

•30nm TiN and 80nm Ti deposition by • 30nm TiN and 80nm Ti deposition by sputtering, then TiN

sputtering, then TiN\ \Ti patterning & etching Ti patterning & etching

•P • P- -type Si type Si (100) substrate (100) substrate

•Wet oxidation (500nm) • Wet oxidation (500nm)

•

•FePt FePt (50nm) (50nm)

• •50nm PECVD oxide 50nm PECVD oxide

•RCA clean • RCA clean

•30nm TiN and 80nm Ti deposition by • 30nm TiN and 80nm Ti deposition by sputtering, then TiN

sputtering, then TiN\ \Ti patterning & etching Ti patterning & etching

•P • P- -type Si type Si (100) substrate (100) substrate

•Wet oxidation (500nm) • Wet oxidation (500nm)

•50nm Fe • 50nm Fe

_0.730.73

Pt Pt

_0.270.27

(composition by ICP- (composition by ICP -MS) MS)

•50nm PECVD oxide deposition and 30nm TiN • 50nm PECVD oxide deposition and 30nm TiN deposition by sputtering, then TiN patterning & etching deposition by sputtering, then TiN patterning & etching

•RCA clean • RCA clean

•200nm Via hole and 250nm contact hole, then • 200nm Via hole and 250nm contact hole, then 80nm Ti lithography patterning and etching 80nm Ti lithography patterning and etching

(100um

(100um × × 100um) 100um)

•P • P- -type Si type Si (100) substrate (100) substrate

•Wet oxidation (500nm) • Wet oxidation (500nm)

•50nm Fe • 50nm Fe

_0.730.73

Pt Pt

_0.270.27

(composition by ICP- (composition by ICP -MS) MS)

•50nm PECVD oxide deposition and 30nm TiN • 50nm PECVD oxide deposition and 30nm TiN deposition by sputtering, then TiN patterning & etching deposition by sputtering, then TiN patterning & etching

•RCA clean • RCA clean

•200nm Via hole and 250nm contact hole, then • 200nm Via hole and 250nm contact hole, then 80nm Ti lithography patterning and etching 80nm Ti lithography patterning and etching

(100um

(100um × × 100um) 100um)

Figure 2.2. Process flows of Ti/TiN/SiO

2/PtFe/SiO2/Si structure, and its small size structure.

Chapter 3

Results and discussions

In this chapter, four sections will be discussed. All of the devices are applied voltage in bottom electrodes, and the top electrodes are ground (Fig. 3-1).

(a) Device structures and characteristics.

(b) Bottom electrode metal effect.

(c) Thermal treatment effect.

(d) Small size effect.

3.1 Device Structures and Characteristics

In section 3.1, the resistive switching properties of different structures are examined and divided into three parts, including results of various bottom electrode metals, various top electrode metals, and various SiO2 thickness.

3.1.1 Various bottom electrode metals

First of all, the resistive switching phenomenon is studied in Al (400nm) / SiO2

(50nm) / Pt (50nm) structure and its depiction of process flows in Fig. 3-2 and Fig.

3-3. The initial state of as deposition Al / SiO

2 / Pt structure processes high resistance value before forming process (Fig. 3-4). It should be noted that, in the previous research, the Au/SiO2/Al structure had resistance switching behavior [34]. However,

the Al/SiO2/Pt structure in my thesis does not possess resistance switching phenomena after the forming process (Fig. 3-5).

In order to study resistance switching behavior for SiO2 on alloy electrode, the Al (400nm) / SiO2 (50nm) / PtFe (50nm) is fabricated. Its structure and process flows are depicted in Fig. 3-6 and Fig. 3-7. Unlike Al / SiO2 / Pt structure, the Al / SiO2 / PtFe structure processes stably and repetitively resistance switching behaviors after the forming process (Fig. 3-8). In order to clarify the mechanisms of resistance switching behavior, seven parts will be discussed in this section.

3.1.1.1 Mechanism of Forming Process

The forming process means that the application of a large voltage a will soften the breakdown of a device. In order to avoid the device being permanent damage, the measurement parameter would set a compliance current which is controlled by a feedback system in apparatus, like Fig. 3-9. A large voltage would produce a high electrical field which induces impact ionization breakdown (Fig. 3-10) in SiO2 for Al / SiO2 / Pt structure [25]. Hence, there are many low resistance paths produced, like

Fig. 3-11.

For the Al / SiO2 / PtFe structure, it is at initial state and processes high resistance value before forming process (Fig. 3-12). During the forming process (Fig.

3-13), bottom PtFe electrode is applied in a large negative voltage which will produce

a high electrical field. Hence, the breakdown of impact ionization would happen in SiO2 and Fe2O3 which is formed on PtFe alloy electrode during the fabrication process (Fig. 3-14). Then, there are many low resistance paths produced in SiO2, and an

“electric faucet” would form in Fe2O3, depicted as Fig. 3-15 [37]. Further discussions

would be continued and its related switching mechanisms are similar with the set process.

3.1.1.2 Mechanism of Phase Change in the Reset Process

After the set and forming process, reset process is the second step to switch resistance. The reset process is that the bottom electrode applies positive voltage, and the top electrode is grounding, like Fig. 3-1. Due to the electrical field direction (Fig.

3-16) and high current flowed through the “electric faucet” [37], the oxygen ion and

localization Joule heating (Fig. 3-17) [39] cause the phase change of Fe3O4 (Fig. 3-18) which is formed during the forming or set process. The chemical reaction possesses relationship as following [18]:

2 Fe3O4 + O^2- => 3 Fe2O3 + 2 e

-The resistance between Fe2O3 and Fe3O4 is different because the band gap of Fe2O3 and Fe3O4 are 2.6eV and 1.6eV, respectively [40, 41]. The total resistance of the insulating layer is the low resistance paths in SiO2 series connection with Fe2O3 thin film which is less conductive. Hence, the phase change from Fe3O4 to Fe2O3 would switch the current or resistance from LRS to HRS in the process of reset.

3.1.1.3 Mechanism of Oxygen Vacancies in the Reset Process

To affect resistance switching behavior is another factor, the amounts of oxygen vacancies. Reset process follows the step of the set or following process. The voltage sweep mode is also the same with Fig. 3-1. Due to the electrical field direction (Fig.

3-16) and high current flowed through the “electric faucet” [37], the oxygen ion and

localization Joule heating (Fig. 3-17)

[39] would decrease the amounts of oxygen

vacancies which are increased at electric faucet region during the forming or set

process. The chemical reaction possesses relationship as following [23]:

V^’’ (donor) + O^2- => O

Due to oxygen vacancies could be seen as donor type defects [23], they would supply electron concentrations in the intrinsic n-type Fe2O3 semiconductor (further discussion about band diagram would be continued in the next section) [38]. The less amounts of oxygen vacancies exist at the electric faucet, the less conductive paths would form. Therefore, after the reset process, the total resistance of the insulating layer is the low resistance paths in SiO2 series connection with less oxygen vacancies of Fe2O3 thin film which is less conductive. The amounts of oxygen vacancies that are decreased in the intrinsic n-type Fe2O3 semiconductor would switch the current or resistance from LRS to HRS during the reset process.

3.1.1.4 Band diagram in the Reset Process

During the reset process, the phase change of Fe2O3 and oxygen vacancies both play important roles in causing resistance switching. After reset process, the band diagram of intrinsic n-type Fe2O3 semiconductor shows in Fig. 3-19. Finally, the Frenkel-Poole emission is predicted by current fitting with different temperature (shows in section 3.3.4) (Fig. 3-20).

3.1.1.5 Mechanism of Phase Change in the Set Process

After the reset process, set process is the second step to switch resistance. The set process is that the bottom electrode applies negative voltage, and the top electrode is grounding, like Fig. 3-1. Due to the electrical field direction (Fig. 3-21) and the applied power by measurement system, the electric faucet would be opened [37]. That means the phase change of Fe O (Fig. 3-22) during the forming or set process. The

chemical reaction possesses relationship as following [18]:

2 Fe3O4 + O^2- <= 3 Fe2O3 + 2 e

-Due to band gap differences between Fe2O3 and Fe3O4, the total resistance of the insulating layer is the low resistance paths in SiO2 series connection with Fe3O4 thin film which is conductive. Hence, the phase change from Fe2O3 to Fe3O4 would switch the current or resistance from HRS to LRS during the set process.

3.1.1.6 Mechanism of Oxygen Vacancies in the Set Process

The role of Oxygen Vacancies is also an important factor to cause resistance switching. First, the set process is the following step to switch resistance after reset process. The voltage sweep mode is also the same with Fig. 3-1. Due to the electrical field direction (Fig. 3-21) and the applied power by measurement system, the electric faucet would be opened in Fe2O3 after forming or set process [37]. The reason is that the amounts of oxygen vacancies would increase in some regions, and these regions represent highly conductive paths in Fe2O3 thin film. The chemical reaction possesses relationship as following [23]:

V^’’ (donor) + O^2- <= O

From the previous sections, oxygen vacancies could be seen as donor type defects, which could supply electron (donor) concentrations in the intrinsic n-type Fe2O3 semiconductor (further discussion about band diagram would be continued in the following section) [38]. The more amounts of oxygen vacancies exist at the electric faucet, the more conductive paths would form. Hence, after the set process, the total resistance of the insulating layer is the low resistance paths in SiO2 series

connection with many oxygen vacancies of Fe2O3 thin film which is highly conductive. The amounts of oxygen vacancies increased in the intrinsic n-type Fe2O3

semiconductor would switch the current or resistance from HRS to LRS during reset process.

3.1.1.7 Band diagram in the Set Process

During the set process, the phase change of Fe2O3 and oxygen vacancies both play important roles to cause resistance switching. After set process, the band diagram of enhanced n-type Fe2O3 semiconductor shows in Fig. 3-23. Finally, the tunneling is predicted by current fitting with different temperature, where the current is less temperature sensitive at LRS (shows in section 3.3.4) (Fig. 3-24).

3.1.2 Effects of Top Electrode Metal Effects

In order to clarify where the key point to cause resistance switching is from, there were many studies had been done before. Therefore, many possible models have been proposed to explain where resistance switching occurred [28]. From the 3.1.1 section, the bottom PtFe electrode is the key point to cause resistance switching.

Hence, in order to prove that two different top electrode metals (Al (400nm) and Ti (80nm) / TiN (30nm) ) are fabricated, and using W probe becomes top electrode with the utmost care that tip does not scratch or perforate the oxide surface. The bistable resistance switching properties have been found in these three structures (Fig. 3-25

(a)-(c)). These evidences show that the top electrode region is not the dominant factor

to cause resistance switching. Furthermore, work function difference (TiN: 4.8~5.3 eV;

W: 4.5 eV; Al: 4.1 eV) in some cases would affect the resistance switching behavior, like contact between metal and p-type or n-type semiconductor [32, 33]. However, this effect does not observe in our case. To Sum up, the top electrode metals do not

affect the resistance switching behavior about the electrode/SiO2/PtFe structure.

3.1.3 Effects of SiO

2

Thicknesses

The final part wants to discuss different SiO2 thickness effects. Two different structures have been fabricated, and they only adjust SiO2 thickness (50nm and 30nm) in both structures (Fig. 3-26 (a)-(b)). In order to let them possess bistable resistance switching characteristics, forming process must do in both structures. From the electrical results, both structures need the same forming electric field (~5MV/cm) to let devices possess bistable resistance switching (Fig. 3-27 (a)-(b)). These phenomena indicated that SiO2 thickness would not affect resistance switching properties, and they only affect forming process (voltage amplitude) which lets SiO2 thin film produces impact ionization breakdown.

3.1.4 Summary I

From the above analysis, we could conclude three points to cause resistance switching characteristics from these different structures.

(a) Bottom electrode contains Fe elements play an important role to cause bistable resistance switching. Proposed models and mechanisms about electric faucet, Redox of Fe2O3, oxygen vacancies, and band diagram could explain the bistable resistance switching behavior.

(b) In the SiO2/FePt system, the resistive memory switching phenomena are independent of top electrode metal effects.

(c) SiO2 thickness effects do not affect resistance switching behaviors when it suffers from impact ionization breakdown.

3.2 Effects of Bottom Electrode Metals

In order to understand more details about resistance switching characteristics for Ti/TiN/SiO2/PtFe structure, the Ti/TiN/SiO2/Fe/Pt/Ti structure is also fabricated to demonstrate why the Ti/TiN/SiO2/FePt structure possesses resistance switching characteristics. Both structures and process flows show in Fig. 3-28 (a)-(c). It is interesting that the original structure (electrode/SiO2/Pt) does not possess resistance switching behavior. However, if the original structure is inserted by a thin Fe film between SiO2 and Pt, this modified structure comes to possess bistable resistance switching property.

3.2.1 Electrical properties

3.2.1.1 Bistable Resistance switching

Fig. 3-29 (a)-(b) are bistable resistance switching characteristics for

Ti/TiN/SiO2/FePt and Ti/TiN/SiO2/Fe/Pt/Ti structures. The forming process needs between both structures, and the forming electrical field is 5.13MV/cm and 5.46MV/cm respectively. According to 3.1.3 section analysis, SiO2 thickness will not affect resistance switching properties, and they only affect forming process (voltage amplitude) which lets SiO2 thin film produces impact ionization breakdown. Hence, forming electrical field is almost the same between both structures. There is another key point behind these phenomena that is the interfacial Fe element that can cause resistance switching behavior (The original structure of electrode/SiO2/Pt does not possess stable resistance switching behavior).

3.2.1.2 Current-Voltage Fitting

Fig. 3-30 shows the current-voltage fitting results of Ti/TiN/SiO

2/FePt and Ti/TiN/SiO2/Fe/Pt/Ti structures. There are three important points in this result.

First, at low resistance state, the current transports of both structures are similar and the fitting results are Ohmic law (Slope of ln (I) vs. ln (V) curve is 1).

Second, at high resistance state low voltage bias region, the current transports of both structures are also similar and the fitting results are Ohmic law (Slope of ln (I) vs.

ln (V) curve is 1).

Third, at high resistance state high voltage bias region, the current transports of both structures are similar again, and the fitting results are close to Frenkel-Poole emission (the relationship of ln (J/V) vs. V^1/2 curve is linear).

From the current-voltage fitting results, indicated that both of structures (Ti/TiN/SiO2/FePt and Ti/TiN/SiO2/Fe/Pt/Ti) possess the similar current transport characteristics.

3.2.1.3 I-V Characteristics Distribution

Fig. 3-31 (a)-(c) depicts that set (reset) voltage, set (reset) current and set (reset)

power dissipation for the both structures. All the statistics results are extracted from five samples and 300 times bistable resistance switching per sample.

From the set and reset voltage distribution, PtFe bottom electrode structure need less value to cause resistance switching than Fe bottom electrode structure (area is 100*100um² and compliance current is 5mA).

From the set current distribution, PtFe bottom electrode structure also needs less value than Fe bottom electrode structure (the reset current distributions for both structures are not obviously different, but still possess the same tendency: PtFe structure needs less value to cause reset process than Fe structure).

Finally, from the set and reset power dissipation distribution, PtFe bottom electrode structure needs less value to cause resistance switching than Fe bottom electrode structure. From the above analysis, the conclusion is that the bottom PtFe electrode structure possesses property of saving more resistance switching power than bottom Fe electrode structure during set and reset process.

3.2.2 Summary II

To sum up, there are three conclusions between Ti/TiN/SiO2/FePt and Ti/TiN/SiO2/Fe/Pt/Ti structures.

(a) Interfacial Fe element is the most critical point to cause resistance switching.

(b) PtFe and Fe bottom electrode structures possess the similar electrical characteristics in HRS or LRS.

(c) During set and reset process, Ti/TiN/SiO2/FePt structure saves more resistance switching power than Ti/TiN/SiO2/Fe/Pt/Ti structure.

3.3 Thermal Treatment Effects

Three thermal treatment methods (as-deposition, furnace annealing 600^oC, 30min in vacuum and rapid thermal annealing 600^oC, 60s at atmospheric pressure in

N2 or air condition) are studied for their resistance switching characteristics for Ti/TiN/SiO2/FePt structure.

3.3.1 Electrical Characteristics

3.3.1.1 Bistable resistance switching and Endurance

Fig. 3-32 (a) and (b) show the bistable resistance switching behavior and its DC

endurance for as-deposition condition of the Ti/TiN/SiO2/PtFe structure. Although as-deposited sample possesses bistable resistance switching, its endurance is poor.

The reading current of the LRS and HRS is too close to determine memory ability (the LRS/HRS current ratio at least 1 order) and the endurance reliability is less than 300 times. One thing in the forming electrical field that deserves to be mentioned is 5.13MV/cm.

Fig. 3-33 (a) and (b) depict the bistable resistance switching and its DC

endurance for furnace annealing 600^oC 30min in vacuum of the Ti/TiN/SiO2/FePt structure. Furnace annealing sample possesses bistable resistance switching, and its endurance reliability is better than as-deposited sample. The 0.2V reading current ratio of the LRS/HRS is higher than 1 order over 1000 times that can determine “1”

or “0” by periphery circuit. Moreover, unlike the as-deposited sample, the forming electrical field of furnace annealing sample is 0.08 MV/cm.

Fig. 3-34 (a) and (b) indicate the bistable resistance switching and its DC

endurance for rapid thermal annealing 600^oC 60s in N2 (air the same) ambient of Ti/TiN/SiO2/FePt structure. Rapid thermal annealing sample possesses bistable resistance switching as well as Furnace annealing and as-deposited sample.

Furthermore, the endurance reliability is also good for rapid thermal annealing

600^oC 60s sample. The 0.2V reading current ratio of the LRS/HRS is higher than 1 order over 1000 times which can let the periphery circuit determine “1” or “0” bit.

Besides, the forming electrical field is 1.25 MV/cm for RTA sample.

Due to above analyses, the electrical characteristics and endurance reliability of different thermal treatment methods in Ti/TiN/SiO2/FePt structure can sum up by Fig.

3-35 and Table 3-1 which point out three key points:

1. Endurance reliability: FA ~ RTA > As-Deposition.

2. LRS current value: FA > RTA > As-Deposition.

3. HRS current value: FA ≧ RTA > As-Deposition.

3.3.2 Material analysis

3.3.2.1 SIMS

SIMS results are shown in Fig. 3-36, indicating the Fe concentration in the 50-nm SiO2 of different thermal treatment conditions. In RTA sample, like as-deposited sample, also possesses the similar Fe concentration profile which is near the interface region between SiO2 and PtFe bottom electrode. However, the Fe concentration profile for FA sample is different with the others. Due to FA sample which suffers long time annealing property, Fe element possesses more time to diffuse in SiO2 than as-deposited and RTA sample.

3.3.2.2 XPS

Fig 3-37 shows XPS results between SiO

2 and bottom PtFe electrode region in Ti/TiN/SiO2/FePt structure, and indicates that all of the thermal treatments possess the

same chemical state, Fe and Fe³⁺ (the 2p3/2 binding energy is 707eV and 710.6eV, respectively) [35].

3.3.2.3 XRD

Fig. 3-38 shows XRD results of different thermal treatment conditions. In the

XRD result of the as-deposited sample, there is no apparent Fe2O3 crystallinity at 33.19^o

[36]. It means that the as-deposited sample possesses Fe

³⁺ chemical state at SiO2 and PtFe bottom electrode interface region from the XPS result, but the crystallinity of Fe2O3 is amorphous and small amount from the XRD result. However, FA and RTA conditions possess obvious crystallinity because both of them must be suffered high temperature annealing which would let Fe2O3 possess enough energy to crystallize. Furthermore, RTA sample has more crystalline than FA sample because of the annealing ambient difference. The annealing ambient of the FA sample suffered high temperature annealing (600^oC) in vacuum can not provide enough O element into SiO2 and PtFe bottom electrode interface region. Hence, from the XRD result, the crystallinity peak of the Fe2O3 is broad. On the other hand, the annealing ambient of the RTA sample suffered high temperature (600^oC) in atmosphere can provide O element into SiO2 and PtFe bottom electrode interface region. Therefore, from the XRD result, the crystallinity peak of the Fe2O3 is sharp.

3.3.3 Comprehensive Comparison

3.3.3.1 Endurance Reliability

From the above discussions, Fe2O3 thin film which is between SiO2 and bottom PtFe electrode region in Ti/TiN/SiO2/FePt structure is the most important component to cause bistable resistance switching, and the thermal treatment conditions would affect the endurance reliability and current value. Hence, XPS and XRD analyses are

the powerful tools which could provide the information about Fe2O3 chemical state and crystallinity. From the above electrical characteristics, as-deposited sample possesses the poorest reliability (about 300 times bistable resistance switching) in comparison with FA and RTA samples (over 1000 times bistable resistance switching).

This property is similar with the XPS and XRD results because both of the analysis tools show small amounts of Fe2O3 chemical state (integral area of XPS fitting) and crystallinity for as-deposited sample. On the other hand, RTA and FA samples possess more amounts of Fe2O3 chemical state (integral area of XPS fitting) and better crystallinity than as-deposited sample. Therefore, the relationship between Fe2O3 thin film quality and endurance reliability is very close. FA and RTA samples possess good endurance reliability because they have large amounts of Fe2O3 thin film between SiO2 and PtFe bottom electrode interface region and good crystallinity of Fe2O3. On the contrary, as-deposited sample does not possess good endurance reliability due to its less amounts of Fe2O3 thin film and bad crystallinity of Fe2O3.

3.3.3.2 The Low Resistance State

Fig 3-39 shows the possible mechanisms for FA and RTA condition at LRS.

From the SIMS results, Fe element of FA sample diffuses deeper and higher concentrations in SiO2 than RTA and as-deposited sample. These Fe ions produce many oxygen vacancies in SiO2, and these oxygen vacancies would affect the forming electrical field and set voltage, like Fig. 3-40. The forming electrical field and set voltage for FA sample is smaller than RTA sample (as-deposited sample is similar).

Furthermore, the more oxygen vacancies exist in SiO2 and Fe2O3 thin film, the more conductive paths form during the set or forming process. FA sample makes more current flow than RTA sample under the same electric measurement condition.

3.3.3.3 The High Resistance State

From the SIMS results, Fe element of FA sample diffuses more depths and concentrations in SiO2 than RTA and as-deposited sample. After the reset process, the electric faucets in Fe2O3 are close and the device is at high resistance state (the total resistance is dominated by Fe2O3 thin film resistance). These results indicate that the quality of the Fe2O3 layer (the amounts of oxygen vacancies) is the key point to

From the SIMS results, Fe element of FA sample diffuses more depths and concentrations in SiO2 than RTA and as-deposited sample. After the reset process, the electric faucets in Fe2O3 are close and the device is at high resistance state (the total resistance is dominated by Fe2O3 thin film resistance). These results indicate that the quality of the Fe2O3 layer (the amounts of oxygen vacancies) is the key point to

在文檔中 氧空缺與相變化在氮化鈦/二氧化矽/鐵白金結構電阻式非揮發性隨機存取記憶體的影響 (頁 39-0)