Chapter 1 Introduction
2.7 The theorem of MIS Capacitors measurement
As shown in Fig.2-10, the theorem which we use on capacitance measurement is: we add a constant DC bias on the metal gate and then add a small AC signal of different frequency. Due to AC signal is changing constantly so the attracted electric charges are changing constantly, too. As the result of the above-mentioned, we will get the capacitance using the changing amount of the electric charges divided by the changing amount of the voltage (2.1).
We will illustrate the three different conditions of ideal MOS diode capacitance under three different bias voltages. As shown in Fig. 2-11, the first condition, we add a minus bias voltage on the metal gate and then the energy band of the interface between oxide layer and semiconductor bend upward. The interface will attract some holes to accumulate around and we name this condition: accumulation. This is real oxide capacitance.
The second condition, we add a small positive bias voltage and then the energy band bends downward. The holes of around interface will be repelled and form a depletion layer. We name this condition: depletion. Because the measurement capacitance is the oxide capacitance series connected with the depletion capacitance, so the total capacitance is smaller. The third condition, we add a large positive bias voltage on the metal gate and then the energy band bends downward drastically. The interface around will not only form depletion layer but also attract some electrons.
When the numbers of minority electrons are more than majority holes and the surface will form inversion phenomenon. We name this condition is: inversion. If the measurement frequency is high, the capacitance is constant due to the width of depletion layer up to maximum. If the measurement frequency is low, the recombination and generation rate of the minority carrier will catch up with the changing of low frequency. The carrier changing will happen at interface around totally and the capacitance is larger and larger up to oxide capacitance [61, 62]. The capacitance-voltage curves of three conditions are shown in Fig. 2-12.
Chapter 3
Electrical Characteristics of Al/Ti/HfAlO(HfO 2 )/Si MIS Capacitors
3.1 Capacitance-Voltage Characteristics
In order to measure the C-V characteristics of our MIS capacitors we used HP2484C LCR meter in our experiments. We swept the gate bias from inversion region to accumulation region to obtain the curve at the frequency of 50 kHz. There are three kinds of plasma treatment with different source gas ( i.e. N2, N2O, and NH3 ) with CF4 plasma and they were treated with CF4 for different process time ( i.e. 0 sec, 30 sec, 60sec, 90 sec,120 sec). Hence, the relationship of difference process time with CF4 in one kinds of plasma treatment will be discussed.
Figure 3-1 reveals the capacitance-voltage (C-V) characteristics of HfAlOgate dielectrics treated with N2 plasma treatment for 30sec and CF4 plasma for different process time. The capacitor treated with CF4 for 30 sec and 60 sec shows almost the same maximum capacitance among these conditions of process time. In addition, the capacitor treated with CF4 for 60 second shows the good C values which are larger than the capacitor with the condition of no treatment. This phenomenon indicates that the N2 plasma treatment with CF4 was workable to improve the capacitance. Maybe it is caused by elimination of oxygen vacancies with optimal nitrogen and fluorine doping. The growing of interfacial oxide has also been restrained. On the other hand, the capacitance treated with CF4 for 90 sec is lower than others except for the no-treated sample. It seems that the plasma damage occur and then destroy the
structure of high-k capacitance when the duration of plasma treatment is too long. We can also find that the capacitance treated with CF4 for 60sec is larger than only treated with N2 plasma, so capacitance with optimal nitrogen and fluorine doping can have better performance than only with the nitrogen doping.
Figure 3-2 shows the capacitance-voltage (C-V) characteristics of HfAlOgate dielectrics treated with NH3 plasma 30sec and CF4 plasma treatment for different process time. Just like the group of N2 plasma treatment. The improvement of capacitance and the damage cause by excessive plasma treatment both can be seen. At this condition, the capacitance treated with CF4 plasma treatment for 60 second shows the largest value. After that, the capacitance becomes worse with the increase of the treatment time. By the way, all the other samples all have larger capacitance than the original sample. It is indicated that the optimal NH3 and CF4 plasma treatment is also a practicable method to improve the capacitance-voltage characteristics of HfAlOgate dielectrics.
Figure 3-3 shows the capacitance-voltage (C-V) characteristics of HfAlOgate dielectrics treated with N2Oplasma 30sec and CF4 plasma treatment for different process time. We can find that the capacitance with CF4 plasma treatment for 30sec have largest value than others, it is because the optimal condition between nitrogen and fluorine causes the elimination of oxygen vacancies in Hf-based layer and has no excess charge. Also, it is shown that the capacitors treated with CF4 for 0 sec, 60 sec, and 90 sec have larger capacitances than the origin sample. Take the view of 120 sec condition, its capacitance value is lower than the origin samples, it may be the interfacial layer growth. Besides 120 sec conditions, the other samples still remain good capacitances than origin sample. So it can be known that the long plasma time
will damage the sample and cause interfacial layer growth. It is suggested that plasma treatment with fluorine radical may cause additional etching for the dielectric layer.
Because the destruction of the oxide layer, the total capacitance was be affected and become lower.
Figure 3-4 shows the capacitance-voltage (C-V) characteristics of HfO2 gate dielectrics treated with N2 plasma 60 sec and CF4 plasma treatment for different process time. The capacitor treated with CF4 for 60 sec shows the maximum capacitance among these conditions of process time. Furthermore, the capacitor treated for 90 sec and 120 sec both show the bad capacitance values. Hence, we can suppose that there is an etched-thin film caused by Fluoridation of a long time, it would cause the distortion of the C-V curve.
Figure 3-5 shows the capacitance-voltage (C-V) characteristics of HfO2 gate dielectrics treated with NH3 plasma 90sec and CF4 plasma treatment for different process time. Just like the samples of N2 and CF4 plasma treatment, the improvement of capacitance could be seen at CF4 60sec plasma treatment. At this condition, the capacitance value shows 2.1*10 -6F/cm2 .By the way, the samples treated with CF4
plasma for 120 sec has a lower capacitance than the sample without treatment. It indicates that too much CF4 plasma treatment is also an etchant to etch thin films and degrade the capacitance-voltage characteristics of HfO2 gate dielectrics.
Figure 3-6 shows the capacitance-voltage (C-V) characteristics of HfO2 gate dielectrics treated with N2O plasma 90sec and CF4 plasma treatment for different process time. For the condition of CF4 plasma30 sec, we notice the capacitance with very large value than the origin sample, but it is still a little lower than the condition
of CF4 plasma 60 sec. This is due to no achievement of optimal condition between Nitridation and Fluorination. The best condition we thought is CF4 60sec plasma treatment, because the leakage current is reduced (see below) and the capacitance is the largest among these conditions. The improvement of capacitance could be seen.
At this condition, the capacitance treated with CF4 plasma for 90 sec and 120 sec show the worse value when Fluorination time is raised. By the way, all the samples which use CF4 plasma except for CF4 120sec have larger capacitance than the sample without treatment. it is indicated that N2O and CF4 plasma treatment, can also be a practicable method to improve the capacitance-voltage characteristics of HfO2 gate dielectrics. Finally, the sample treated with CF4 plasma for 120 sec is bad due to plasma damage.
Figure 3-7 shows the capacitance-voltage (C-V) characteristics of HfAlOgate dielectrics treated with N2, NH3, N2Oplasma treatment and CF4 plasma at optima condition. It is indicated that the capacitance treated with NH3 plasma treatment and CF4 plasma for 60 sec shows the most excellent value (i.e. 278% increasing about capacitance). Because NH3 compose of N atoms to repair defects and dangling bonds, and fluorine also can repair interface state and let the interface have no excess charge.
Among these samples, the capacitance treated with N2Oplasma treatment is worse because N2O has oxygen atoms and the growing of interfacial oxide is unavoidable while the oxygen atoms become radical and enter the interface. But there is a good effect on capacitance by Nitridation and Fluorination, it can maintain large enough capacitance value. Thus, the capacitance improvement by interface repair was easily observed by Nitridation and Fluorination.
Figure 3-8 shows the the capacitance-voltage (C-V) characteristics of HfO2 gate
dielectrics treated with N2 ,NH3,N2O plasma treatment and CF4 plasma at optimal condition. It is indicated that the capacitance treated with N2 plasma treatment and CF4 plasma for 60 sec shows the most excellent value (i.e. 258% increasing about capacitance). Among these samples, the reason why the sample treated with NH3 plasma has lower capacitance than N2O plasma treatment is complex. It may be that NH3 has the least N atoms to repair defects and dangling bonds. NH3 has hydrogen atoms to bond with dangling bonds but the bond is weak.
Figure 3-9 shows the capacitance-voltage (C-V) characteristics of HfAlOgate dielectrics treated with N2, NH3
and
N2Oplasma treatment all for 30 sec. It is indicated that the capacitance treated with N2 plasma treatment for 30 sec shows the most excellent value. Fig.3-10 shows the capacitance-voltage (C-V) of HfO2 gate dielectrics treated with N2 plasma treatment for 60 sec, NH3 plasma treatment for 90 sec and N2O plasma treatment for 90 sec. It is indicated that the capacitance treated with NH3 plasma treatment for 90 sec shows the most excellent value3.2 Current-Voltage Characteristic
Figure 3-11 shows the J-V characteristics of p-type HfAlOcapacitors treated with N2 plasma 30 sec and CF4 treatment for different process time from 0 V to -2 V.
We observed that the gate leakage current density is suppressed while treatment conditions are 30sec, 60sec, 90 sec. It is indicated that N2 plasma and CF4 plasma treatment supply an effective barrier against the leakage current. The lower leakage shows that the weak structure of interface must be fixed by the plasma nitridation and fluorination, especially for 60 sec capacitor which both has the low leakage and largest capacitance value from Fig. 3-1. Gate leakage current density of origin insulator at VG = -1 V is about 9.025×10-2 A/cm2. From fig.3-11, however, gate
leakage current density of the capacitor treated for 60 sec CF4 plasma at VG = -1 V is only about 1.52×10-2A/cm2. It has less gate leakage than origin insulator about 1 orders. Furthermore, we notices that the 90 sec capacitor although has lower leakage than the origin, its capacitance has become degradation. This is an interesting phenomenon. Even though the plasma damage has begun to reduce C value, the leakage current is still kept very well. It means that the capacitance value is more easily affected by plasma damage than leakage current.
Figure 3-12 shows the J-V characteristics of p-type HfAlO capacitors treated by NH3 plasma 30sec and CF4 plasma with different process time from 0 V to -2 V. After NH3 plasma and CF4 plasma treatment, we can see the reduction of leakage current in contrast of the original sample. It is worthy to be noticed that the capacitors treated by 60 sec CF4 plasma which has the best C value also performs a low leakage current about 1.32×10-4A/cm2 at Vg=-1V. In addition, we find that the leakage current of 120 sec treatment are larger than the other sample, but they are all not larger than original sample. Relative to the case of N2 plasma, we can see that the level of leakage current increase obviously due to plasma damage.
Figure 3-13 shows the J-V characteristics of p-type HfAlOcapacitors treated by N2Oplasma 30sec and CF4 plasma with different process time from 0 V to -2 V.
Besides 120 sec sample, the other samples depict the presence of the reduction in leakage current. It is indicated that there are not only the effect of improving interface quality but also another effect to suppress the leakage current in the case. According to the discussion about Fig. 3-3, we know that the growth of interfacial oxide layer will decrease the C value. Now the interfacial layer introduces a hard barrier to suppress leakage current. Consequently, everythig except for that 120 sec condition,
all have revealed a lower leakage current value, due to the destruction caused by the plasma.
Figure 3-14 shows the J-V characteristics of p-type HfO2 capacitors treated by N2 plasma 30sec and CF4 plasma with different process time from 0 V to -2 V. We observed that the gate leakage current density is suppressed while treatment conditions are 60 sec and 90 sec. The results show that fluorine and nitrogen atoms were accumulate into the HfO2 dielectrics to form Hf–F and Hf-N bonding by CF4
and N2 plasma, resulting in the reduction of gate leakage current, charge trapping.
Especially for capacitor which treated with CF4 plasma 60 sec and it also has the lower leakage and largest capacitance value from Fig. 3-4. Gate leakage current density of no treatment insulator at VG = -2 V is about 1.35 A/cm2. From Fig.3-3, however, gate leakage current density of the capacitor treated for 60 sec CF4 plasma at VG = -2 V is about 0.25 A/cm2. It has less gate leakage than no treatment insulator about 1 order. Furthermore, we notices that the capacitor treated with N2 plasma for 130 sec has high leakage current, it is might be that the CF4 plasma is too little time to react with the film. Then, we observed that the leakage of the sample with 120 sec CF4 plasma is also larger than the origin due to plasma damage.
Figure 3-15 shows the J-V characteristics of p-type HfO2 capacitors treated by NH3 plasma 90sec and CF4 plasma with different process time from 0V to -2V. After NH3 plasma treatment, we could see the reduction of leakage current for 90 sec sample in contrast of no treatment sample. However, the sample of plasma treated for 60 sec got the little large gate leakage current but a sharp J-V curve and good C-V curve from Fig 3-5. It is indicated that the sample with CF4 60sec is good for interface characteristics. Relative to the case of N2 plus CF4 plasma, we could see that the
level of leakage current increasing obviously. It is possibly that the time with NH3 and CF4 plasma too long cause etching the thin film by fluorine radicals.
Figure 3-16 shows the J-V characteristics of p-type HfO2 capacitors treated by N2O plasma 90sec and CF4 plasma with different process time from 0V to -2V. After N2O plasma treatment, we could see the reduction of leakage current in contrast of no treatment samples except the 30sec sample. It is indicated that the plasma treatment could be less for fluorine radicals to fixed interface trapped charge. However, the sample of plasma treated for 60 sec got the small gate leakage current and a good C-V curve from Fig 3-6. Relative to the case of N2 plasma, we could see that the level of leakage current decreasing obviously. It is possibly due to the additional oxidation layer formed by oxygen radical. The interfacial oxidation layer will let the dielectric thicker to prevent from gate leakage.
3.3 Summary
By the compare of the samples which has the best capacitance in their own gas, we can realize the most suitable treatment condition which both has the best capacitance and lower leakage current. Hence, we significantly find a relative optimum condition among above discussion. It is proved that without thick oxidation layer, it can also reach the smallest leakage current when there is suitable time treatment.
If we take a look at all the samples, we find that most of the N2, N2O, and NH3
the N element and F element can fix the interface and promote the electrical properties include of C-V curve and J-V curve. As for the Hf02 type of material, oxygen radical can cause growth in the interfacial layer, therefore by comparing with the N20 plasma treatment's sample, it shows a lower C value. Just because the oxidation phenomenon, the films will become thicker so that the plasma damage will not easily affect the leakage current profile.
Chapter 4
Reliability of Al/HfAlO(HfO 2 )/Si MIS Capacitors
4.1 Hysteresis
The name of Hysteresis was borrowed from electromagnetics. It is means that when a ferromagnetic material is magnetized in one direction, it will not relax back to zero magnetization when the applied magnetizing field is removed. It must be driven back to zero by the additional opposite direction magnetic field. If an alternating magnetic field is applied to the material, its magnetization will trace out a loop called a hysteresis loop [34].
The hysteresis phenomenon is similar to the C-V curve in the MIS capacitor device. When we apply a voltage in reverse, it will not fit the original C-V curve measured previously. It is due to the interface traps which can trap charges to have impact on the flat band voltage and C-V curve. [23] Fig. 4-1 shows the hysteresis of p-type HfAlOgate dielectrics treated without PDA, plasma treatment and PNA and the hysteresis is 98 mV. Fig. 4-2 shows the hysteresis of p-type HfAlOgate dielectrics treated with N2 plasma 30 sec plus CF4 plasma treatment for 60 sec process time.
Hysteresis of p-type HfAlOcapacitors is changed with the plasma treatment and its value is 7mV. The hysteresis is suppressed by means of the fixing ability at the interface. This means that fluorine incorporation into the HfO2 gate dielectrics to strengthens the HfO2 thin film.
Figure 4-3 shows the hysteresis of p-type HfAlOgate dielectrics treated with NH3 plasma 30sec plus CF4 plasma treatment for 60 sec process time. The tendency of hysteresis is similar with the case of N2 plasma treatment and it’s value is 19 mV.
Fig. 4-4 shows the hysteresis of p-type HfAlO gate dielectrics treated with N2O plasma 30 sec and CF4 plasma treatment for 60 sec process time and its value is 41 mV. It also shows a likely tendency. As a consequence, the plasma treatment can improve the reliability of hysteresis for all the different optimal plasma gas treatment.
Among these samples, we can find that the hysteresis of N2 plasma treatment for 30 sec is the smallest but the other values are larger.
Figure 4-5 shows the hysteresis of p-type HfO2 gate dielectric without any treatment and the hysteresis is 32 mV. Fig. 4-6 shows the hysteresis of p-type HfO2
gate dielectrics with N2 plasma 60 sec plus CF4 plasma treatment for 60 sec process time. The hysteresis voltage is 8mV and also small than the origin.
Figure 4-7 shows the hysteresis of p-type HfO2 gate dielectrics (MOCVD) with NH3 plasma 30sec plus CF4 plasma treatment for 60 sec process time. The hysteresis voltage is 11mV.Fig. 4-8 shows the hysteresis of p-type HfO2 gate dielectrics (MOCVD) with N2Oplasma 30 sec and CF4 plasma treatment for 60 sec process time.
The hysteresis voltage is 9mV, so nitridation and fluorination could decrease the trap density and let the thin film sustain high thermal stress.
Therefore, we can speculate that the sample without plasma treatment which is
Therefore, we can speculate that the sample without plasma treatment which is