Dissertation Organization

In this dissertation, we concentrated our effort to examine the effect of the plasma nitridation process and the plasma fluorination process to the electrical characteristics, the reliability and the thermal stability of pure HfO2 and HfAlOx thin films.

In the chapter 1 of this dissertation, we describe the background and the motivation of our research. In the chapter 2, we apply the inductance-coupled plasma (ICP) nitridation technology to pure HfO2 thin films in order to improve the electrical characteristics, the

reliabilities and the thermal stability of pure HfO2 thin films. In the chapter 3, we examine the similar ICP nitridation process to HfAlOx thin films to observe the process effect to the electrical characteristics, the reliabilities and the thermal stability of HfAlOx thin films.

Furthermore, in the chapter 4, we use the ICP fluorination process to improve the electrical characteristic and the reliabilities of the plasma-nitrided HfO2 thin films. In the chapter 5, we use the similar plasma fluorination process to improve the electrical characteristic and the reliabilities of the plasma-nitrided HfAlOx thin films.

Finally, in the chapter 6, a conclusion is given, and the future work about this dissertation is proposed.

Figure 1.1 The partial table from ITRS 2009, which indicates the predicted standard for high performance (HP) device technical requirements.

Ref: [1]. International Technology Roadmap for Semiconductors, presented at public.itrs.net (2009).

Figure 1.2 The partial table from ITRS 2009, which indicates the predicted standard for low operating power (LOP) device technical requirements.

Ref: [1]. International Technology Roadmap for Semiconductors, presented at public.itrs.net (2009).

Figure 1.3 The partial table from ITRS 2009, which indicates the predicted standard for low standby power (LSTP) device technical requirements.

Ref: [1]. International Technology Roadmap for Semiconductors, presented at public.itrs.net (2009).

Figure 1.4 The XRD spectrum of HfO2 dielectric after various PDA temperatures.

Ref: [8]. W. J. Zhu, T. Tamagawa, M. Gibson, T. Furukawa, and T. P. Ma, “Effect of Al inclusion in HfO2 on the physical and electrical properties of the dielectrics”, IEEE Electron Dev. Lett. 23, p. 649 (2002).

Figure 1.5 The EOT value of HfO2 dielectric after various PDA temperatures.

Ref: [8]. W. J. Zhu, T. Tamagawa, M. Gibson, T. Furukawa, and T. P. Ma, “Effect of Al inclusion in HfO2 on the physical and electrical properties of the dielectrics”, IEEE Electron Dev. Lett. 23, p. 649 (2002).

Figure 1.6 The plasma source of the plasma enhanced chemical vapor deposition (PECVD) system and inductively-coupled plasma (ICP) system.

Chapter 2

The Improvement Effect of the Plasma Nitridation Process to the Electrical Properties, the Reliability and the

Thermal Stability of HfO

thin films

2.1 Introduction

The rapid advancement of complementary metal oxide semiconductor (CMOS) field effect transistor process during the past few years has forced the microelectronics industry to face serious technological challenges. According to the predictions of the International Technology Roadmap for Semiconductor, the equivalent oxide thickness (EOT) and gate leakage currents of conventional gate dielectrics will reach their physical limits [1]. To solve the challenge about the excessive leakage current, many kinds of high-k dielectrics have emerged as the promising candidates to replace the ultrathin silicon oxynitride dielectrics for the advanced CMOS technologies [2-5]. The pure HfO2 is considered as a suitable gate dielectric material because of the acceptable band gap, which is wide enough to avoid the gate leakage forming the unacceptable power consumption, and the large dielectric constant, which is big enough to increase the physical thickness of the gate dielectric and maintain the relatively low EOT. However, there are several issues which have to be considered in order to integrate these Hf-based dielectrics into a conventional CMOS process flow such as the reliability and the thermal instability of these dielectrics [7, 30-31]. The incorporation of nitrogen into gate dielectrics by nitridation has been investigated with the aim of preventing dopant penetration [13-14]. Recently, the various nitridation processes have been shown to improve the thermal stability and the dielectric constant of Hf-based dielectrics [15-18, 21].

However, thermal nitridation is usually performed at high temperature and hydrogen-containing species that act as electron traps could be introduced into the thin film by thermal nitridation process. On the other hand, nitrogen could be incorporated by plasma nitridation process into the dielectric layer at lower temperature than by thermal nitridation process [22, 32]. The objective of this report is to examine the effect of different ICP plasma nitridation process to the electrical characteristics, the reliability and the thermal stability of HfO2 thin films [33]. According to this study, the plasma nitridation process could be an effective technology to improve the electrical characteristics, the reliability and the thermal stability of pure HfO2 thin films.

2.2 Experimental

In this research, Al/Ti/HfO2/Si structures were fabricated to investigate the effect of the plasma nitridation to HfO2 dielectrics. A Si wafer with standard initial RCA cleaning was placed into the chamber and a 6-nm HfO2 layer was deposited on the wafer by the MOCVD system. Then the samples were annealed at 600 ℃ for 30 sec in pure N2 gas by rapid temperature annealing (RTA) process and nitrided by an ICP plasma treatment at the substrate temperature of 300 ℃, the process pressure of the plasma nitridation process was set as 1.33

× 10^-4 bar. Ar gas was added into the ICP chamber for activating the plasma while the nitridation process was performed. The flow rate of Ar was set as 10 sccm and the flow rate of the gas containing nitrogen, which is N2, NH3 or N2O, was set as 100 sccm. In an ICP system, a rf current in the coils generates a changing magnetic field, then the magnetic field induces a changing electric field through inductive coupling. Therefore, the inductively coupled electric field accelerates electrons. The electronic mean free path might be shorter than the distance between the anode and the cathode, so there are enough ionizing collisions to produce high density plasma [34]. After the plasma nitridation, there was an annealing process whose

condition was at 600 ℃ for 30 sec in pure N2 to eliminate the plasma damage [35-36]. A Ti film of 40 nm was deposited on the top side of the samples by sputtering. Then an Al film of 500 nm was thermally evaporated on the top side of the samples. The top electrodes were defined by a lithography process. Finally, the backside native oxide was stripped with diluted HF solution and the backside aluminum electrodes were evaporated by a thermal evaporation.

The top area of the Al/Ti/HfO2/Si MOS capacitors is 5000 μm². The capacitance-voltage (C-V) and the current density-voltage (J-V) characteristics of the MOS structures were measured by semiconductor parameter analyzer (HP4156C) and C-V measurement (HP4284) in order to evaluate the improvement effect of the plasma nitridation process and the best process condition. The experimental condition of the stress induced leakage current (SILC) measurement that was carried out in this study was set as constant voltage of 3 V for 180 sec.

The stress condition of constant voltage stress (CVS) measurement carried out in this study was set as constant voltage of 3 V to observe the change of the gate leakage current while the stress was applied.

2.3 Results and Discussion

In the beginning of this work, different process time was tested to decide the most suitable process condition for various plasma nitridation processes. After the process time had been determined, the improvement effect of the plasma nitridation process to the reliability of pure HfO2 thin films would be examined.

2.3.1 The Most Suitable Process Time

Figure 2.1 shows the C-V characteristics of the HfO2 gate dielectrics treated in ICP N2

plasma for different process times. The frequency used in the high frequency C-V

measurement was set as 50 kHz. The capacitors treated for 90 sec perform the maximum capacitance density among these samples with different process times. The EOT of HfO2 thin films decrease from 3.6 nm to 2.3 nm after N2 plasma nitridation. In addition, the capacitors treated for 30 sec and 60 sec also present the larger values than the capacitors without whole plasma nitridation process. The factor of improvement might be from that the PDA process [37-39] and the nitrogen incorporation in the HfO2 dielectrics, which could enhance the electronic polarization as well as the ionic polarization, so the dielectric constant of the HfO2

thin films increases just as Hf-silicate thin films [20, 40] and SiO2 thin films [41]. Besides, the capacitance density of the samples treated for 120 sec is degraded. The reason could be the damage caused by the N2 plasma.

The J-V characteristics of the HfO2 capacitors treated by ICP N2 plasma with different process times from 0 V to -2 V are described in Fig. 2.2. The gate leakage current density is suppressed while the treatment condition was 60 sec. The reduction of the leakage current could be attributed to the post-deposition annealing process [37-38]. The gate leakage current density of the samples not treated in ICP N2 plasma at Vg of -1 V is about 3.74 × 10^-4 A/cm² and the gate leakage current density of the capacitors treated in ICP N2 plasma for 60 sec at Vg of -1 V is about 2.22 × 10^-5 A/cm². Moreover, the leakage current densities of the samples treated in N2 plasma for shorter or longer time are larger than the one treated for 60 sec. The effect of the nitridation process with shorter process time might be not enough. On the other hand, while the nitridation process time is longer than 60 sec, the plasma damage from the plasma nitridation could cause the increase of the gate leakage density. In summary, the best process time of the plasma nitridation in N2 plasma is set as 60 sec. This appears that the samples treated in N2 plasma for 60 sec display the most excellent value (the EOT of the samples is about 2.3 nm).

In Fig. 2.3 and Fig. 2.4, the C-V and the J-V characteristics of the HfO2 gate dielectrics treated in ICP NH3 plasma for different process times are presented. As mentioned before, the reason of the improvement effect in the NH3 plasma nitridation process could be the same as the one in the N2 plasma nitridation process [20, 37]. From the similar analysis, the best process time of the plasma nitridation in NH3 plasma is set as 90 sec. The EOT of HfO2 thin films decrease from 3.6 nm to 2.3 nm after NH3 plasma nitridation. The gate leakage current density of the capacitors treated in ICP NH3 plasma for 90 sec at Vg of -1 V is about 1.62 × 10^-5 A/cm². This indicates that the samples treated in NH3 plasma for 90 sec displays the most excellent value (the EOT of the samples is about 2.3 nm).

In Fig. 2.5 and Fig. 2.6, the C-V and the J-V characteristics of the HfO2 gate dielectrics treated in ICP N2O plasma for different process times are demonstrated. There could be other reasons for the improvement effect of the N2O plasma nitridation process. Since a high concentration of oxygen vacancies causes electrons to be generated and a large leakage current to flow, treatment with plasma that contains oxygen could yield active oxygen atoms and reduce oxygen vacancies to improve the quality of dielectric films [34]. The EOT of HfO2

thin films decrease from 3.6 nm to 2.7 nm after N2O plasma nitridation. The gate leakage current density of the samples treated in ICP N2O plasma for 60 sec at Vg of -1 V is about 2.37 × 10^-5 A/cm² and the gate leakage current density of the capacitors treated in ICP N2O plasma for 90 sec at Vg of -1 V is about 2.22 × 10^-5 A/cm². So when the process time is set as 90 sec, the reduction of the gate leakage density of the nitrided sample would be more obvious. The decrease in the leakage current from nitridation process might be believed to be caused by electron trapping [42-43]. As mentioned above, the best process time of the plasma nitridation in N2O plasma is set as 90 sec. The samples treated in N2O plasma for 90 sec perform that the EOT of the samples is about 2.7 nm. The reason that the EOT of the sample

treated by N2O plasma was bigger than the ones treated by N2 or NH3 plasma could be the excess oxygen in N2O plasma to make the HfO2 thin films become thicker.

2.3.2 Reliability

Figure 2.7 shows the hysteresis characteristics of the HfO2 gate dielectrics treated in different ICP plasma. Hysteresis measurement was started from positive to negative bias (negative sweeping, 1 to −2 V), and then swept back from negative to positive bias (positive sweeping, −2 to 1 V) at a frequency of 50 kHz. The hysteresis phenomenon of the C-V curves can be observed for all samples, which is caused by the existence of negative charges trapped in the dielectric defect states when the capacitors are stressed [44]. These defect states are called slow trapping sites [45]. The hysteresis characteristic could be improved by various ICP plasma nitridation process as presented in Fig. 2.7. The C-V curve shift of the sample nitrided by N2O plasma is about 6 mV and the voltage shift in the C-V curve of the sample nitrided by N2 plasma is about 9 mV. That is, the hysteresis phenomenon of pure HfO2

dielectrics could be restrained to be less than 10 mV by both N2 and N2O nitridation processes.

Figure 2.8 displays the SILC curve of p-type HfO2 gate dielectrics treated with N2 plasma process. After the samples stressed by constant voltage of 3 V for 180 sec, the degree of leakage current degradation could reflect the reliability of the samples. First, Fig. 2.8 presents that the degradation could be improved by N2 plasma nitridation process. The increase of gate leakage current at -1 V of HfO2 thim film changes from 159.06 % to 79.19 % after N2 plasma nitridation. Secondly, since HfO2 dielectric would recrystallize when suffering a high temperature process, the leakage current would increase after the extra high temperature RTA

process (800 ℃ for 30 sec in N2). Even so, the degradation of the gate leakage would be slighter with N2 plasma nitridation process.

As shown in Fig. 2.9, the SILC characteristics of p-type HfO2 gate dielectrics nitrided by NH3 plasma process present that the shift of leakage current because of the constant voltage stress could also be restrained by NH3 plasma nitridation. The increase of gate leakage current at -1 V of HfO2 thim film changes from 159.06 % to 50.00 % after NH3 plasma nitridation.

Furthermore, even when nitride sample suffered an additional high temperature RTA process (800 ℃ for 30 sec in N2), the increase of the gate leakage current would be still suppressed.

Figure 2.10 describes the SILC curves of p-type HfO2 gate dielectrics treated with N2O plasma process. As Fig. 2.10 presents, the shift of leakage current due to the constant voltage stress could also be restrained by N2O plasma nitridation. The increase of gate leakage current at -1 V of HfO2 thim film changes from 159.06 % to 87.80 % after N2O plasma nitridation.

But the improvement effect of the N2O plasma nitridation process to the SILC phenomenon of the HfO2 thin films seems unapparent in comparison with other plasma nitridation process.

Figure 2.11 demonstrates the gate current shift of p-type HfO2 gate dielectrics treated with N2 plasma for different annealing process during CVS of 3 V. Figure 2.11 indicates that the gate current shift of the thin film treated with N2 plasma for 60 sec is smaller than the one without nitridation. After N2 plasma nitridation 60 sec, the gate leakage shift shrinks as 3.76 %. Besides, the sample with N2 plasma treatment for 60 sec that suffered an additional thermal process of 800 ℃ also has smaller current shift. In other word, for substrate injection, the gate current shift could be suppressed by N2 plasma nitridation.

The CVS characteristics of samples nitride by miscellaneous kinds of ICP plasma are described in Fig. 2.12. All the gate current shifts of the samples with nitridation could be decreased. The gate leakage shift shrinks as 11.64 % after 90 sec NH3 plasma nitridation and the gate leakage shift shrinks as 14.36 % after 60 sec NH3 plasma nitridation. Among all nitrided process, the ICP N2 plasma nitridation process presents the best result. In summary, the improvement effect of the ICP N2 plasma nitridation would be most obvious to the reliability of HfO2 thin films.

2.3.3 Thermal Stability

The nitrogen was incorporated into the dielectric could maintain an amorphous homogeneous film without phase separation at high temperature [21]. The nitridation effect to the thinner HfO2 thin films (the EOT of the samples was about 1.5 nm) was also corroborated in other experimental result.

In Figure 2.13 and Figure 2.14, the C-V and the J-V characteristics of the HfO2 gate dielectrics treated by different plasma nitridation processes and thermal treatments have been shown. As demonstrated in Figure 2.14, for the samples which were just deposited in the ALD system and not nitrided, the C-V characteristic of the samples without the high temperature annealing (in N2 at 850 ℃ for 30 sec) was very different from the samples with the annealing. So from the electrical characteristic, the original samples could not sustain the high temperature annealing. In the meantime, for the samples which were nitrided in N2

plasma for 90 sec, the C-V characteristic of the samples without the high temperature annealing was very similar to the samples with the annealing. So it seemed to prove that the nitridation process could improve the thermal stability of the HfO2 thin films. If the nitridation time had not been enough, the thermal stability of the high-k thin films would be

not enough either. Just like the samples treated in N2 plasma for 30 sec could not sustain the high temperature annealing. In Figure 2.14, we observed that the J-V curve of the sample with nitridation which suffered high temperature annealing could maintain a lower value than the one without nitridation. The above electrical characteristics could also confirm the improved effect of the plasma nitridation to the thermally stability of the HfO2 thin films.

2.3.4 Physical Analysis

Fig. 2.15 is XPS analysis of the Hf 4f electronic spectra of the samples treated in ICP N2

plasma for 60 sec. For the nitrided HfO2 thin films, the Hf 4f peaks of the XPS spectra would shift to lower binding energy because the binding energy of Hf-N bonds is lower than Hf-O bonds [60]. So it indicates that the presence of Hf-O-N bonds in HfO2 thin films after ICP N2

plasma nitridation.

2.4 Summary

According to above research, after the best process time was decided from the C-V and J-V characteristics, the improvement effect of the ICP plasma process to the reliability of pure HfO2 thin films was verified from the hysteresis, SILC and CVS characteristics. For different ICP plasma nitridation process, the influence would be diverse. In conclusion, the ICP N2

plasma nitridation process could achieve the best performance to the reliability of HfO2 thin films. The EOT of HfO2 changes from 3.6 nm to 2.3 nm and the gate leakage current density changes from 3.74 × 10^-4 A/cm² to 2.22 × 10^-5 A/cm² after N2 plasma nitridation.

Furthermore, the plasma nitridation could be also used to improve the thermal stability of the HfO2 thin films to bear the high temperature process at 850 ℃ for at least 30 sec.

Figure 2.1 The C-V characteristics of the HfO2 thin films treated in N2 plasma for different process times.

Figure 2.2 The J-V characteristics of the HfO2 thin films treated in N2 plasma for different process times.

Figure 2.3 The C-V characteristics of the HfO2 thin films treated in NH3 plasma for different process times.

Figure 2.4 The J-V characteristics of the HfO2 thin films treated in NH3 plasma for different process times.

Figure 2.5 The C-V characteristics of the HfO2 thin films treated in N2O plasma for different process times.

Figure 2.6 The J-V characteristics of the HfO2 thin films treated in N2O plasma for different process times.

Figure 2.7 The hysteresis characteristics of the HfO2 thin films nitrided by different ICP plasma process.

Figure 2.8 The SILC characteristics of the HfO2 thin films nitrided by ICP N2 plasma.

Figure 2.9 The SILC characteristics of the HfO2 thin films nitrided by ICP NH3 plasma.

Figure 2.10 The SILC characteristics of the HfO2 thin films nitrided by ICP N2O plasma.