Properties of HfAlO x MIS Capacitor with Dual Plasma
3.4.3 Breakdown voltage
Fig. 3-12 presents the breakdown characteristics of the HfAlOx thin films treated in CF4 plasma for different process durations and N2 (NH3) plasma for 90 sec.
Although the sample treated by single N2 (NH3) plasma shows lower gate leakage current than fresh sample, the leakage is suppressed mostly for the sample with dual plasma treatment (CF4 pre-treatment for 10 sec and N2 (NH3) post-treatment for 90 sec). The reduction of gate leakage current is the result of defect passivation. Nitrogen and fluorine atoms could reduce the oxygen vacancy related states and interface states within HfAlOx band gap. Nevertheless, while CF4 pre-treatment is longer than 10 sec, the non-ideal effect that gate leakage increased. The reason for this phenomenon might be attributed to plasma damage at the interface. These results consisted with the breakdown voltage behaviors and distribution, as shown in Fig. 3-13. It has been reported that the first breakdown happened in the low-k layer because the applied electric field would be largely distributed in the low-k region for high-k/low-k gate
98
stack [22]. In other words, interfacial layer (IL) represented the low-k layer and would cause the reliability problems. As shown in Fig. 3-13, the sample treated by dual plasma treatment (CF4 pre-treatment for 10 sec and N2 (NH3) post-treatment for 90 sec) shows the largest breakdown voltage because of the great improvement on interface quality. In short, the best condition of dual plasma treatment is CF4 pre-treatment in 10 sec with RF power 20W and N2 (NH3) post-treatment in 90 sec.
The breakdown voltage of all the samples is illustrated in Table 3-1 and Table 3-2.
3.4.4 Frequency dispersion characteristics
The C-V characteristics of the HfAlOx thin films with CF4 plasma for different process durations and N2 plasma for 90 sec measured under different frequencies were compared in Fig. 3-14. The range of measurement frequency was set form 1 kHz to 100 kHz. It could be noticed that the strong frequency dispersion in the accumulation region and the hump in the depletion region in the C-V curve for the fresh sample.
The effect of frequency dispersion is attributed to the response of charges to signal frequency [23]. Because some of the traps could follow the change of the applied gate voltage, the additional capacitance would be generated [23]. Compared to the fresh sample, the sample treated by N2 plasma showed relatively smaller frequency dispersion and smaller hump. Besides, it could be observed that the sample with dual plasma treatment (CF4 pre-treatment for 10 sec and N2 post-treatment for 90 sec) shows nearly no dispersion in the accumulation region and nearly no hump in the depletion region, which is independent of frequency. The reason of the improvement could be ascribed to that dual plasma treatment could effectively improve the interface quality [23-26]. However, while CF4 pre-treatment time is longer than 10 sec, the frequency dispersion and the hump became worst again because of plasma damage at the interface
99
Fig. 3-15 depicts the C-V characteristics of the HfAlOx thin films with CF4 plasma for different process durations and NH3 plasma for 90 sec measured under different frequencies. Similarly, the sample with dual plasma treatment (CF4 pre-treatment for 10 sec and NH3 post-treatment for 90 sec) shows nearly no dispersion in the accumulation region and nearly no hump in the depletion region, which is attributed to the improvement of the interface quality.
3.4.5 Constant voltage stress characteristics
The C-V curves before and after CVS testing of the HfAlOx thin films treated in CF4 plasma for different process durations and N2 plasma for 90 sec are shown in Fig.
3-16. The stress voltage was set as -3 V and the tress time was set in a range from 0 to 700 sec. It could be observed that all the C-V curves shift to left when stress time increase indicating that the generation of positive charges are trapped in the dielectric layer, which could be explained by Anode hole injection model [27]. The injected electrons traveled from the gate through the dielectric layer and arrived at the interface when constant negative bias stress, resulting in the de-passivation of Si3≡SiH centers and creating the Si3≡Si-dangling bonds at the interface (so called Pb0
centers) [28]. The released hydrogen transported through the dielectric field by electric field and trapped in the dielectric; as a consequence, it may create the positively charged centers [29]. On the other hand, the C-V curve of fresh sample exhibited hump after constant voltage stress, owing to the creation of interface defects.
Compared to other samples, it is worth mentioning that the C-V curves had smaller Vfb shift and less distortion for samples with CF4 pre-treatment for 10sec and N2
post-treatment for 90sec because less interface trap charges generated at the HfAlOx/Si interface. On the other hand, the similar phenomenon also could be observed for the sample treated in CF4 pre-treatment and NH3 post-treatment for
100 90sec, which is shown in Fig. 3-17.
3.5 Current conduction of Al/Ti/HfO
2/Si MIS capacitors
In order to understand the current conduction mechanism, three kinds of conduction mechanisms were be analyzed, including Schottky emission (S.E.), Frenkel-Poole (F-P) emission, and Fowler-Nordheim (F-N) tunneling. The leakage current was measured from 298 K to 398 K (25 K per step) because to analyze S.E.
and F-P emission needs the time dependence of the leakage current.
3.5.1 Schottky emission (S.E.)
The standard Schottky emission could be expressed as
where JSE is the current density, A* is the effective Richardson constant, E is the effective electric field, T is the absolute temperature, q is the electron charge, qB is the Schottky barrier height, k is Boltzmann’s constant, ε0 is the permittivity of free space, εr is the dynamic dielectric constant, m* is the electron effective mass in HfO2, m0 is the free electron mass. The electron effective mass is 0.1 m0.
A plot of ln(J/T2) versus E1/2 should be a straight line, as shown in Fig. 3-18. Eq.
(2) expressed the intercept of the Schottky emission plot with the vertical axis. The barrier height could be extracted from Eq. (2). The Schottky barrier height of all the samples is shown in Table 3-3. It is clear that the HfAlOx samples had larger barrier height than other samples after CF4 pre-treatment for 10 sec and nitrogen
101 post-treatment for 90 sec.
kT3.5.2 Frenkel-Poole (F-P) emission
When gate under negative bias, electrons will inject from gate into HfO2 dielectric layer and will be trapped into shallow trap levels. Thereafter, the electrons transported through the dielectric layer by hopping between these trap levels, leading to leakage current, called Frenkel-Poole (F-P) emission. The standard F-P emission could be expressed as [35]
trap energy in dielectric layer could be extracted form good F-P fitting, as shown in Fig. 3-19. The extracted trap levels are illustrated in Table 3-4.The extracted trap levels for the sample with and without N2 post-treatment were 1.27 eV and 0.92 eV, respectively. In contrast, the trap level for the sample treated by CF4 pre-treatment for 10 sec and N2 post-treatment for 90 sec was 1.49 eV. The deeper trap level means that most of shallow trap levels in HfO2 thin film can be eliminated by using dual plasma treatment. The elimination of shallow trap levels resulted in the reduction of F-P conduction current. Similarly, the sample treated in CF4 pre-treatment for 10 sec and NH3 post-treatment for 90 sec also exhibited larger trap levels than other samples. In short, dual plasma treatment could eliminate the
102
shallow trap levels and greatly reduce the gate leakage current.
3.5.2 Fowler-Nordheim (F-N) Tunneling
In higher electric field, the Fowler-Nordheim (F-N) tunneling dominated the conduction mechanism. The standard F-P emission could be expressed as
height, m* is the electron effective mass in HfO2, and the other notations were as same as mentioned before. The electron effective mass here is 0.1 m0. If the leakage current is dominated by the F-N mechanism, a plot of ln(J/E2) versus 1/E should be linear.Linear characteristic could be observed at high electric field, as shown in Fig. 3-20.
The slope of each curve in Fig. 3-20 and Eq. (5) could obtain the F-N barrier height, which were listed in Table 3-5.
23Table 3-5 listed the F-N barrier height of the samples with no treatment, single plasma treatment, and dual plasma treatment. It could be observed that samples treated in CF4 pre-treatment for 10 sec and nitrogen post-treatment for 90sec had larger value than other samples. The injection of electrons from the gate entered the conduction band of HfAlOx by tunneling through a triangular potential barrier. The injected electrons interacted with lattice or transferred its energy at anode, which resulted in the degradation of the dielectric layer. As a result, the sample with proper dual plasma treatment had bigger F-N barrier height and better reliability properties.
103
3.6 Conclusion
The interface quality and reliability properties of HfAlOx gate dielectric with dual plasma have been verified. Based on above results, the electrical characteristics including C-V, I-V, hysteresis, frequency dispersion, and CVS characteristics of HfAlOx gate dielectrics could be great improved by dual plasma treatment. According to our study, the best condition is CF4 pre-treatment for 10 sec and N2 (NH3) post-treatment for 90 sec time. According to the current conduction analysis, the dominant current conduction mechanism was Schottky emission type in the region of low to medium electric fields; Frenkel-Poole (F-P) emission operated in the region of medium to high fields; Fowler-Nordheim (F-N) tunneling was dominant at high fields.
In conclusion, dual plasma treatment could improve interface quality and enhance reliability properties of HfAlOx thin films.
104
Table 3-1 illustrates the basic electrical characteristics of the HfAlOx sample for dual plasma treatment (CF4 pre-treatment and N2 post-treatment).
Samples Fresh N2(120s) CF4(10s)+
Table 3-2 illustrates the basic electrical characteristics of the HfAlOx sample for dual plasma treatment (CF4 pre-treatment and NH3 post-treatment).
Samples Fresh NH3(120s) CF4(10s)+
105
Table 3-3 Schottky Barrier Height Extracted for The HfAlOx Samples with No Treatment, Single Plasma Treatment, and Dual Plasma Treatment.
Barrier height Fresh N2(90s) CF4(10s)+
qB 1.07±0.06 1.21±0.04 1.29±0.04 1.18±0.05
Barrier height Fresh NH3(90s) CF4(10s)+ NH3
qB 1.07±0.06 1.22±0.07 1.33±0.06 1.19±0.03
Table 3-4 F-P Trapping Level Extracted for The HfAlOx Samples with No Treatment, Single Plasma Treatment, and Dual Plasma Treatment.
Barrier height Fresh N2(90s) CF4(10s)+
Table 3-5 F-N Barrier Height Extracted for The HfAlOx Samples with No Treatment, Single Plasma Treatment, and Dual Plasma Treatment
Barrier
106
Fig. 3-1 Key process flows of two experiments in this work for (a) single nitrogen post-treatment and (b) dual plasma treatment (CF4 pre-treatment and nitrogen post-treatment)
107
Fig. 3-2 The C-V characteristics of the HfAlOx thin films with N2 post-treatment for different process durations.
Fig. 3-3 The I-V characteristics of the HfAlOx thin films with N2 post-treatment for different process durations.