Chapter 1 Introduction
1.5 Organization of the thesis
There are four chapters in this thesis. Chapter 1 describes the background and motivation for the high-κ dielectrics and application of the MIM structure.
In Chapter 2, we first describe the experimental procedure and then, show the basic characteristics of Al2O3 gate stacks on Si substrate, including basic C-V and I-V characteristics. The Al2O3 thin film in this thesis is deposited by metal-organic chemical vapor deposition (MOCVD), and the optimum conditions is found, including deposition temperature, oxygen purge time, oxygen purge flow.
Furthermore, we find optimum post-deposition annealing temperature and post-metal
deposition temperature.
In Chapter 3, we deposit Al2O3 on TiN metal with these optimum conditions which we find in Chapter 2. However, MIM structure performance is not as good as we expected. High leakage current density is fatalness of MIM structure. In order to suppress leakage current density, AlN buffer layer inserted between Al2O3 thin film and below TiN electrode is discussed. Besides, thinner the thickness of below TiN electrode is used as well.
Finally, in Chapter 4, the conclusions are made and the recommendation describes the topics which can be further researched.
Figure 1-1 The expected equivalent oxide thickness (EOT) trends from the published 2003- ITRS roadmap.
Figure 1-2 Power consumption and gate leakage current density comparing to the potential reduction in leakage current by an alternative dielectric exhibiting the same equivalent oxide thickness [1].
Material
HfO2 25 5.7 1.5 Mono.,Tetrag.,Cubic
ZrO2 25 7.8 1.4 Mono.,Tetrag.,Cubic
Table 1-1 Basic properties for many high-κ candidates [1].
Property TiN Ti Structure fcc hcp
Density 5.4 g/cm3 4.54 g/cm3
Melting point 2950°C 1940°C
Thermal conductivity 30 Watt/m-K 13 Watt/m-K Thermal expansion 9.36x10-6 /K 11x10-6 /K
Electrical resistivity 170µΩ-cm 39µΩ-cm
Table 1-2 Material properties of titanium nitride (TiN) and titanium (Ti) [14-16].
3.5
Figure 1-3 Band alignment of topical high-κ dielectrics.
Reaction ∆G° (Kcal / mol) Ti (s) + O2 (g) → TiO2 (s)
2Ti (s) + N2 (g) → 2TiN(s)
2TiN (s) + 2O2 (g) → 2TiO2 + N2 (g)
TiN (s) + 2H2S(g) → TiS2 + 1/2N2 (g) + 2H2 (g)
-175.180 -115.079 -108.960 +40.955
Table 1-3 Gibb’s free energy of titanium (Ti) and its combination[17].
CHAPTER 2
Characteristics of Al
2O
3Gate Dielectrics
2.1 Introduction
While the critical dimensions (CD) of complementary metal-oxide-semiconductor (CMOS) devices are scaling down, large direct-tunneling current is inevitable due to the ultra thin gate dielectric stack. On the next generation, we substitute high-dielectric constant (high-κ) materials for conventional silicon dioxide (SiO2). High-κ gate materials can maintain the same equivalent oxide thickness (EOT) with thicker physical thickness, and is therefore expected drastically reduced direct-tunneling current.
There are many methods to deposit high-κ gate dielectrics stack, such as physical vapor deposition (PVD) [17], atomic layer deposition chemical vapor deposition (ALCVD) [11,18-21], and metal-organic chemical vapor deposition (MOCVD) [22-24]. In the industrial production viewpoint, PVD is not an appropriate tool for high-κ film deposition due to both poor step coverage and bad uniformity. Nowadays, ALCVD and MOCVD have paid more efforts to be evaluated for high-κ dielectrics deposition in the industry. Table 2-1 is the comparison of deposition techniques which have been used. MOCVD has the advantage of superior step coverage, high deposition rate, good controllability of film composition, and excellent thickness uniformity on large dimension wafers. MOCVD system is therefore chosen to deposit
high-κ dielectrics in this thesis.
Recently, aluminum oxide (Al2O3) had been proved as promising candidates for the gate dielectrics of sub-0.1µm device due to their higher κ value, relatively high
φ
Band superior thermal stability [22]. Due to the high dielectric constant and high thermal stability, Al2O3 is suitable to be integrated into trench DRAM process and is therefore chosen in this thesis.
Figure 2-1 shows the detail schematic structure of the MOCVD system. The MOCVD chamber is equipped with a turbo-molecular pump and a liquid injection system which has four independent-controlled injectors. A liquid pump is consisted of the injector and pumps the precursors through a hot nickel frit with a proper rate. The vapors are carried with a 200 sccm flow of Argon to gas distribution ring which is located at a proper distance from the substrate. On the contrary of the conventional bubbler system, the liquid injection system is with sufficient temperature window to alleviate the thermal aging of the precursor. This is because the precursor remains in liquid state at room temperature until it is pumped into the vaporizer and injected into the deposition chamber. However, the precursor should keep at long-term chemical stability in solvent and non-reactive with other precursors in solvent [25].
The components of the vaporizer, the gas ring and the connecting tube are maintained at 190°C with heating tapes and blankets, while the substrate temperature is controlled at 500°C with quartz-halogen lamps and a thermocouple. A rotating susceptor is used for uniformly heating during processing. A flow of 100 sccm N2 is maintained throughout the deposition cycle. The base pressure of the MOCVD chamber is ~ 10-8 torr. The deposition pressure of the deposition is at the 5 mtorr where the gas-phase collisions are scarce.
2.2 Experimental details
In this chapter, LOCOS isolation was used to fabricate the metal-insulator-semiconductor (MIS) capacitors. The cross-sectional views and processing steps are shown in Figure 2-2. The MIS capacitors were fabricated on 6-inch (100)-oriented silicon wafer with 15~25 µΩ-cm resistivity. Prior to the growth of Al2O3 gate dielectrics, the native oxide was cleaned by the conventional RCA cleaning and diluted HF etching in sequence for the removal of particles and native oxides. Wafers were then processed to receive two different deposition temperatures (400°C and 450°C) by MOCVD. Aluminum precursor and oxygen gas were purged to deposited approximately 8 nm Al2O3 films, followed by three different RTA temperatures (800°C, 900°C, 1000°C) in N2 ambient for 30 sec.
Subsequently, a 2000Å titanium nitride (TiN) electrode was sputtered and patterned to form gate electrodes, followed by the 450°C post metal-deposition annealing (PMA) for in N2 ambient 30 sec. Finally, 5000Å aluminum was deposited on wafer backside with 400°C forming gas sintering for 30 min to create the backside contact.
Square or circular capacitors of different areas, ranging from 2.5╳10-5 to 1 ╳102 cm2, with LOCOS isolation are used to evaluate the gate oxide integrity. The physical gate oxide thickness was determined by spectroscopy ellipsometer and compared with transmission electron microscopy (TEM). The cross-section view of MIS capacitor was analyzed by transmission electron microscopy (TEM) as well. The equivalent oxide thickness (EOT) was extracted by fitting the measured high-frequency capacitance-voltage (C-V) data form Hewlett-Packard (HP) 4284LCR meter under an accumulation condition with quantum mechanical correction. The tunneling leakage
current density-voltage (J-V) was measured by semiconductor parameter analyzer HP4145A. Compositions of Al2O3 and TiN were analyzed by X-ray photoelectron spectroscopy (XPS).
2.3 Physical and Electrical Characteristics
2.3.1 Deposition Temperature and PDA Temperature
Figure 2-3(a) reveals the capacitance-voltage (C-V) characteristics of as-deposited Al2O3 gate dielectrics deposited at 400°C and 450°C. The capacitance deposited at 400°C was lower than 450°C-deposited sample with similarly optical thickness, which was evidenced by ellipsometer. Since high temperature deposition would transfer high surface mobility to the deposited atomics, Al2O3 dielectric deposited at 450°C would expect to have higher film density than 400°C-deposited dielectric. The corresponding current density-voltage (J-V) curves were presented in Figure 2-3(b). Al2O3 dielectric deposited at higher temperature was beneficial to suppress leakage current than lower temperature, which was consistent with the C-V characteristics shown in Fig. 2-3(a), i.e. high deposition temperature generated higher density Al2O3 thin films with larger capacitance and lower leakage current than low temperature deposited high-κ dielectrics.
Atomic layer does not combine very well in the as-deposited Al2O3 thin film.
After high temperature annealing, Al2O3 film is densified and shows more polarity and higher κ value [1、26]. Sequentially, PDA temperature effect is discussed. The PDA temperature effects on C-V characteristics of 400°C-deposited Al2O3 gate dielectrics is shown in Figure 2-4(a). Due to densification of Al O thin film [26、27], capacitance
increases with 800°C PDA. However, while PDA temperature is up to 900°C and 1000°C, capacitance descends. This is supposed that oxygen penetration through Al2O3 film over 900°C [28-31] and interact with Si-substrate resulting in thicker interfacial layer, i.e. EOT increases and capacitance under accumulation decreases.
The corresponding current density-voltage (J-V) curves were presented in Figure 2-4(b). Al2O3 thin film processed after higher PDA temperature was beneficial to suppress leakage current [28-31], which was coincident with the C-V characteristics shown in Figure 2-4(a).
Figure 2-5(a) shows C-V curves of PDA temperature effect of 450°C-deposited Al2O3 gate dielectrics. Capacitance of 450°C-deposited Al2O3 gate dielectrics decreases as PDA temperature increased. Oxygen penetration through Al2O3 film contribute to interfacial layer between Al2O3 and Si-substrate [28-31], i.e. high PDA temperature generated thick interfacial layer between Al2O3 thin film and Si-substrate with lower capacitance. The corresponding current density-voltage (J-V) curves were presented in Figure 2-5(b). Leakage current density decreases with PDA temperature increased, which is supposed to be suppressed by thicker interfacial layer [28、29].
Al2O3 thin film processed after high PDA temperature was beneficial to suppress leakage current, which was consistent with the C-V characteristics shown in Figure 2-3(a).
Figure 2-6(a) presents the EOT variation for various PDA temperatures with 400°C and 450°C deposition. EOT reduction is ascribed to densification of 400°C-deposited Al2O3 film with PDA 800°C [27]. Then, EOT increases because of thicker interfacial layer [28-31] when PDA temperature is above 800°C. EOT of 450°C-deposited Al2O3 film increases after high temperature PDA, which is cause by thicker interfacial layer between Al2O3 thin films and Si-substrate. Figure 2-6(b)
displays the leakage current density variation for various PDA temperatures with 400°C and 450°C deposition. Lower leakage current density is observed with PDA temperature increased both on 400°C and 450°C-deposited Al2O3 thin film, i.e. proper PDA can get better performance of gate leakage current density.
Table 2-2 summarizes most of the electrical characteristics under all the conditions of deposition temperature (400°C and 450°C) with PDA temperature (800°C, 900°C, and 1000°C), including EOT, gate leakage current, hysteresis, and dispersion. In the viewpoint of lowest leakage current density and hysteresis, performance of Al2O3 thin film processed with 400°C deposition and 900°C PDA is best, i.e. the optimum condition of deposition and PDA temperature is determined.
2.3.2 J-V Curves Measurement under Various Temperatures
Figure 2-7(a) shows J-V characteristics of as-deposited MIS capacitor measured at various temperatures from 25°C to 150°C. The dependence of leakage current density and measured temperature is observed, i.e. leakage current density increases with measurement temperature increased. The dependence is suppressed after 900°C PDA, as shown in Figure 2-7(b). Conduction mechanism is found by fitting equation described as follows. Many conduction mechanisms are fitted, including Fowler-Nordheim Tunneling [32、33] , Frenkel-Poole Emission [32、33], Trap Assisted Tunneling [34、35], and Schottky Emission [32].
In the Fowler-Nordheim Tunneling model, leakage current occurs in the high field region. High electric field across on high-κ thin film inclines band diagram and electron can tunnel more easily. The equation of leakage current density is [36]:
The Fowler-Nordheim Tunneling plots were made for Jg (not shown in the thesis). In the Fowler-Nordheim Tunneling plots, Jg does not show a linearity relationship. The conduction mechanism is therefore not the Fowler-Nordheim Tunneling.
In the Frenkel-Poole Emission model, a lot of traps exist in high-κ thin film and electrons which get enough thermal energy can leap and stay in these traps temporarily and leak to substrate in the end. The equation of leakage current density is [36]:
The Frenkel-Poole Emission plots were made for Jg (not shown in the thesis). Jg
does not show a straight line in the Frenkel-Poole Emission plots, therefore the conduction mechanism is probability not the Frenkel-Poole Emission.
In the Trap Assisted Tunneling model, it is assumed that electrons first tunnel through the SiOx interfacial layer (direct-tunneling). Then, electrons tunnel through traps located below the conduction band of the high-κ thin film and leak to substrate finally [34]. The equation of leakage current density is [35]:
⎥⎦
The Trap Assisted Tunneling model plots were made for Jg (not shown in the thesis). Jg is not a straight line in the Trap Assisted Tunneling model plots, therefore the conduction mechanism is probability not the Trap Assisted Tunneling model.
In the Schottky Emission model, the Schottky emission is generated by the
thermionic effect and is caused by the electron transport across the potential energy barrier at a metal-insulator interface. The equation of leakage current density is [36]:
( )
The Schottky Emission model plots were made for Jg in Figure 2-8. Jg shows a clear linearity in the Schottky Emission model plots, therefore the conduction mechanism is probability the Schottky Emission model.
From the conduction mechanism fitting, we speculate that the conduction mechanism of MIS structure is Schottky Emission.
2.3.3 Transmission Electron Microscopy (TEM) Analysis
It is found that 900°C PDA can improve performance of MIS, including EOT reduction, leakage current density suppression and little dependence of leakage current density and measured temperature. The mechanism is analyzed from TEM images. As shown in Figure 2-10, thickness of Al2O3 thin film is recognized about 7.44 nm- 9 nm and interfacial layer is about 1.06 nm- 1.13 nm. After 900°C PDA, thickness of Al2O3 thin film is a little thinner about 7.06 nm- 8.21 nm and interfacial layer is about 1.81 nm-2.14 nm, as shown in Figure 2-11(b). On the other hand, the κ value of as-deposited Al2O3 thin film is about 7.12 and that is about 9.53 after 900°C PDA. The κ value is increased and interfacial layer is thicker after high temperature annealing. It can be understood reasonably that during high temperature annealing Al2O3 film is densified and thickness of interfacial layer is increasing due to oxygen penetration [28-31].
2.3.4 Oxygen Purge Conditions Optimization and Post Metal-deposition Anneal (PMA) effect
Al2O3 thin film is deposited by reaction of aluminum precursor and oxygen gas.
Oxygen purge time and oxygen purge flow is relative to characteristics of Al2O3 thin film. To optimize the performance of MIS, we purge oxygen before/ after aluminum precursor purges. The detail process flow is below:
O2 Purge－＞ Al2O3 deposition －＞ O2 Purge
We are wondering how long we should purge exactly. In the first part of experiment, oxygen is purged before Al2O3 deposition and is skipped after Al2O3
deposition. The second part of experiment is contrary. List in the Table 2-3.
On the other hand, in the trench DRAM technology, trench capacitor will undergo high thermal budget while source/ drain activation annealing. To simulate industrial process, high temperature annealing after metal electrode deposition is necessarily. PMA temperature over 900°C is expected in our research. At the beginning, PMA temperature is determined at 450°C in order to get a smoother interface between upper electrode TiN and Al2O3 thin film.
Figure 2-11(a) shows EOT performance both with 450°C PMA and without 450°C PMA under oxygen purge 15、30、60、120 sec before Al2O3 deposition. Higher EOT is observed before 450°C PMA. However, EOT thickness reduces after PMA 450°C. From Figure 2-11(b), leakage current is lowest with oxygen purge 30 sec after PMA 450°C. It is found that performance of Al2O3 deposition is best due to lowest leakage current at the similar EOT, i.e. optimum condition of oxygen purge time before Al2O3 deposition is 30 sec.
Subsequently, we focus on oxygen purge time after Al2O3 deposition. Figure 2-12(a) displays EOT performance both with 450°C PMA and without 450°C PMA
under oxygen purge 30、70、450 sec after Al2O3 deposition. Before 450°C PMA, increased EOT is observed after PDA 900°C, ascribing to oxygen penetration through Al2O3 thin film and inducing thicker interfacial layer [28]. Similar to Figure 2-11(a), EOT decreases after PMA 450°C. On the other hand, leakage current density is lowest when oxygen purges 30 sec after PMA 450°C, as shown in Figure 2-12(b). It is found that performance of Al2O3 deposition is best due to lowest leakage current at the similar EOT, i.e. optimum condition of oxygen purge time after Al2O3 deposition is determined at 30 sec.
Finally, we focus on oxygen purge flow effect with Al2O3 deposition. Figure 2-13(a) shows the oxygen purge flow effect on EOT. We notice that without PMA, EOT increases slightly with oxygen flow rate increasing, causing by more oxygen penetration and thicker interfacial layer [28]. Similar to Figure 2-11(a) and Figure 2-12(a), EOT decreases after PMA 900°C. Consideration of entirely interaction of aluminum precursor and oxygen gas, oxygen purge flow must be large enough to eliminate residual aluminum atoms. However, too large oxygen purge flow induces thick interfacial layer. We therefore experiment further only on oxygen purge flow is 500 sccm and 1000 sccm. Figure 2-13(b) shows the leakage current density with Al2O3 thin film deposited under oxygen flow is 500 and 1000 sccm. It is observed that the leakage current density is lower when Al2O3 thin film deposited under 1000 sccm oxygen flow, i.e. optimum condition of oxygen purge flow during Al2O3 deposition is determined at 1000 sccm.
To summarize briefly, we find out optimum conditions of oxygen purge time and oxygen purge flow. The optimum process is as follows:
O2 Purge 30 sec －＞ Al2O3 deposition －＞ O2 Purge 30 sec
For the time being, there is still one question in our mind, which is the mechanism of PMA effect resulting in improving MIS performance, including EOT reduction and leakage current density suppression. Subsequently, we discuss this phenomenon by TEM and XPS analyses.
2.3.5 Transmission Electron Microscopy (TEM) and X-ray Photoelectron Spectroscopy (XPS) Analyses
Figure 2-14 shows the TEM images of the sample which was processed by PMA 450°C. There is an obvious interfacial layer between upper electrode TiN and Al2O3
thin film about 1.25 nm, which is not seen in Figure 2-9 and Figure 2-10. From TEM images, we estimate the thickness of Al2O3 thin film and bottom interfacial layer is 7.94nm- 8.44nm, and 1.44nm- 2.04nm, respectively, and is not obviously changed comparing to Figure 2-10, i.e. 450°C PMA does not change the whole dielectric gate stack except the top interfacial layer between upper electrode TiN and Al2O3 thin film. Total dielectric gate stack is a little thicker because of top interfacial layer.
However this is not coincident with electrical characteristics (EOT of total dielectric gate stack reduces after 450°C PMA). Therefore, it is necessarily to understand the composition of top interfacial layer by XPS analysis.
The XPS analyses are performed using a ULVAC-PHI PHI Quantera SXM spectrometer: the instrument employs a 180° spherical capacitor analyzer and 32 channel detectors, and a scanning monochromated (A1 anode) X-ray source. For the present study Mg Kα radiation is used, at energy 1253.6 eV, and the source is operated at 400W (15 kV and 27 mA) applied voltage and emission current, respectively. There is a 5 kV ion gun which can clean sample surface and analyze depth profile. Etching rate of ion gun is 70Å SiO2/ min. Surface element component
analysis can reach 10 Å below the surface.
The test sample structure is Al2O3 80Å /CVD TiN 2000Å, without upper electrode TiN deposition. In order to expose the interface of Al2O3 thin film and below electrode CVD TiN, the surface of Al2O3 thin film is removed by Argon gas sputtering. As sputtering time increases, the composition closer to the interface of Al2O3 thin film and below electrode CVD TiN will be detected.
Figure 2-15 displays the survey of XPS analysis. Ti peak is detected with sputter time over 1 min, and is stronger while sputter time increased. It is ascribed to approach the bottom TiN electrode. Element of Ti、Al、O and N is detected at the same time. Figure 2-16 shows the XPS binding energy spectrum of Ti. As sputter time increased, the peak of TiN is getting strong because of closer to below TiN electrode, which is consistent with Figure 2-15. Furthermore, the peak of TiO2 is detected as well. While sputter time increased, the peak of TiO2 is getting obvious because of closer to interfacial layer. We therefore speculate that the interface of Al2O3 thin film and TiN electrode is TiO2. From TEM analysis, there is an interfacial layer between Al2O3 thin film and upper electrode TiN. The samples are under 450°C PMA in N2 ambient for 30 sec. A little bit oxygen in N2 ambient penetrate through the upper TiN electrode and form the interfacial layer. According to XPS analyses, we recognize the interfacial layer between Al2O3 and TiN is TiO2, whose κ value is
Figure 2-15 displays the survey of XPS analysis. Ti peak is detected with sputter time over 1 min, and is stronger while sputter time increased. It is ascribed to approach the bottom TiN electrode. Element of Ti、Al、O and N is detected at the same time. Figure 2-16 shows the XPS binding energy spectrum of Ti. As sputter time increased, the peak of TiN is getting strong because of closer to below TiN electrode, which is consistent with Figure 2-15. Furthermore, the peak of TiO2 is detected as well. While sputter time increased, the peak of TiO2 is getting obvious because of closer to interfacial layer. We therefore speculate that the interface of Al2O3 thin film and TiN electrode is TiO2. From TEM analysis, there is an interfacial layer between Al2O3 thin film and upper electrode TiN. The samples are under 450°C PMA in N2 ambient for 30 sec. A little bit oxygen in N2 ambient penetrate through the upper TiN electrode and form the interfacial layer. According to XPS analyses, we recognize the interfacial layer between Al2O3 and TiN is TiO2, whose κ value is