Chapter 2 Fabrication and Experiment
2.3 Measurement and Parameter Extraction
3.1 Electrical Analysis 3.2 Material Analysis 3.3 Mechanism 3.4 Summary Chapter 4 Conclusion References
Vita
Fig. 1-1 Phase diagram for CO
2Table 1-1 Critical temperature and pressure for some common fluids
Table 1-2 Comparison of physical properties of CO
2
Fig. 1-2 Density – pressure – temperature surface for pure CO
2Chapter 2
Fabrication and Experiment
2.1 Devices Structure and Fabrication
In this experiment, we fabricate the a-Si TFTs and polycrystalline silicon thin films successfully. Moreover, we use supercritical fluid technology to treat the a-Si TFTs and poly silicon thin films. The cross sectional view of a-Si TFT is shown in Fig.
2-1 and the schematic cross sectional view of polycrystalline silicon thin film is shown in Fig. 2-2. For material analysis and measurements, polycrystalline silicon films were formed. The fabrication procedure is described as follows.
Fig. 2-1 shows a cross-sectional view of the a-Si TFT. This structure was fabricated based on the fabrication process described below. A corning glass was used as the substrate. The gate metal was Cr, deposited by DC sputtering, with the thickness of 150 nm and patterned by photolithography followed by wet etching processes. The conventional back channel etching (BCE) TFTs were used with 380 nm thick silicon nitride (SiNx), hydrogenated amorphous silicon layer of 150 nm thickness, and 50 nm thick n+ a-Si:H were deposited on the substrate by PECVD. 300 nm Al layer was deposited by thermal evaporation coater and then patterned for source and drain region by wet etching. Then, island region was patterned by dry etching processes.
Fig. 2-2 shows a cross-sectional view of polycrystalline silicon thin film. This structure was fabricated based on the fabrication process described below. First, silicon nitride (SiNx) films of 230 nm thicknesses were formed by LPCVD on a silicon substrate. Then, polycrystalline silicon films of 75 nm thicknesses were
formed on substrates using the low pressure chemical vapor deposition (LPCVD) method.
2.2 Experimental Procedures
Fig. 2-3 shows the schematic drawing of the supercritical fluids treatment system.
Pump A feeds a high-pressure syringe pump and supplies high-pressure ambient-temperature CO2. Pump B feeds a high-pressure syringe pump and supplies solvents. The sample is setting in the pressure chamber. And the flow rate is restricted by modulated tube.
In the experiment, three kinds of conditions have been used, respectively. They were described as follows. For the purpose of material and electrical analysis, a-Si TFTs and polycrystalline silicon thin films were treated at the same time.
1. Annealing-only
The samples were cut into pieces suitable for the size of pressure chamber (Fig. 2-3). The temperature of chamber was set for 150℃. Here the supercritical fluids system was the same as an airtight hot plate. When the temperature arrived to 150℃, the experiment began by placing the samples within the pressure chamber.
After 2 hours, we removed the samples from the supercritical fluids system.
2. H2O
The samples were cut into pieces suitable for the size of pressure chamber.
The temperature of chamber was set for 150℃. When the temperature arrived to 150℃, the experiment began by placing the samples within the pressure chamber. Then, the H2O from the pump B fell to the pressure chamber (Fig.
2-3).
After 2 hours, we removed the samples from the supercritical fluids system.
3. SCCO2 / Propyl alcohol / H2O
The samples were cut into pieces suitable for the size of pressure chamber.
The temperature of chamber was set for 150℃. Propyl alcohol and H2O were used for co-solvent. The percent of co-solvent was 10%. The experiment began by placing the samples within the pressure chamber in the first process step.
In the second process step, the syringe pumps of Pump A and Pump B (Fig.
2-3) were pressurized to 3000 psi, and then the supercritical CO2 carried the co-solvent into pressure chamber. The step provided for that CO2 and co-solvent in supercritical fluid state and ensured the pressure chamber was clean.
In the third process step, the syringe pumps of Pump A and Pump B were pressurized to 3000 psi, and more co-solvents were carried by supercritical CO2
into the chamber. During this soak step, the supercritical CO2 and co-solvent were maintained in contact with the sample.
After 2 hours, it was followed by dry step. The samples were soaked by supercritical CO2 at 3000 psi. Then they were flushed several times by supercritical CO2 at 3000psi. The dry step was performed to carry away the solvent on the samples. Lastly, the experiment ends by depressurizing the pressure chamber and removing the samples.
Finally, all samples were placed on the hot plate, whose temperature was maintained at 200℃ for 1 hour. Afterward they were been measurement by material and electrical.
2.3 Measurement and Parameter Extraction
In this section, we will introduce the electrical measurement instrument and the methods of typical parameter extraction.
The electrical property analysis instrument mainly used the Agilent 4156A semiconductor analyzer. The Agilent 4156A semiconductor analyzer with probe stations was used to analyze the electrical properties of the devices. In our experiment, it was used for I-V measurement and bias-temperature-stress (BTS). The ground probe station was provided an electrical isolation. The water-cooled thermal plate within an optical shielding box and plate could be controlled by the TPO315A thermal controller between 25°C and 300°C. The source measurement units (SMUs) were used to control voltage sources where current flowing through could be measured.
The voltage or current sources were supplied by Agilent 4156A semiconductor analyzer.
The methods of parameter extraction such as threshold voltage, subthreshold swing, on/off current ratio, field-effect mobility μFE, and activation energy of the device were described as below.
1. Extraction of the threshold voltage
Threshold voltage is defined as the gate-source voltage at which conduction electrons begin to appear in the channel of TFT. A low threshold voltage is needed to assure the TFT operational region at a reasonable voltage ranges.
In the transfer characteristics, threshold voltage can be determined by the methods of linear line plot threshould voltage and constant current threshould voltage. In this thesis, we use the method of constant current threshould voltage in which the voltage at a specific drain current NId is taken as the threshold voltage. This technique is adopted in most studies of a-TFTs.
Constant current threshould voltage：
( )
2. Extraction of the subthreshold swing
Subthreshold swing S.S. (V/dec) is a typical parameter to describe the control ability of gate toward channel. It is defined as the amount of gate voltage required to increase/decrease one order of magnitude drain current. The subthreshold swing of the transfer characteristics should be low enough in order to achieve fast on/off state transition of the transistor. It strongly depends on the density of defect states in the channel material. An abrupt subthreshold swing ensures the transistor working at a high performance.
The subthreshold swing of the transfer characteristics is defined as (Figure 2-4):
( )d
In this thesis, we define that one-half of the gate voltage required to decrease the
threshold current by two orders of magnitude is subthreshold swing.
(2-4)
[ ( ) ( ) ]
(2-5)
−
The density of states (DOS) in TFT can be determined by the exponential transition between the on and off states. This parameter gives the information about the density of states at the interface between the amorphous silicon and the gate insulator (interface states) and in the bulk of the amorphous silicon (bulk states). The subthreshold swing could be affected also by states corresponding to defects at the channel surface (surface states). It is found that the surface states can have the same effect on the subthreshold swing as the interface states [30]. Here we assume that the densities of deep bulk states Nbs and interface states Nss are independent on the energy.
In the typical a-Si TFT, the defects in the amorphous silicon film and defect states near the insulator-silicon interface are mainly responsible for the performance of TFT in the subthreshold swing. For poly-Si TFT, defect states at the grain boundary are mostly responsible for the performance.
We will discuss the density of states in more details in the next chapter.
V
3. Extraction of the on/off Current Ratio
On/off ratio is another important factor of a-Si TFTs. High on/off ratio represents not only large turn-on current but also small off current (leakage current). High IDS,on/IDS,off (on/off) ratio is desirable in TFT, for good “transparent and non-transparent” states transition of the liquid crystal cell. To reach this goal, the transistor should have a high enough Ionand a low enough Ioff.
The limit of the on current for a-Si TFT is the electron mobility in the channel material. In general, the a-Si TFT owns an electron mobility lower than 1cm/Vs. This is because of the high defect concentration in a-Si:H (about 1015cm-3).
Lower value of the off currentis necessary for high on/off ratio of the a-Si TFT.
When a high negative voltage is applied to the gate, holes will accumulate near the gate. Then off current will increase. One solution for this is to introduce n+ a-Si layer into the drain and source contacts. The hole current will be eliminated by recombination in the n+ a-Si region. Also, this n+ a-Si doped layer can improve the on current and avoid Schottky barrier between the source / drain metal contacts and the channel material of amorphous silicon.
There are many methods to specify the on and off current. The easiest one is to define the maximum current as on current and the minimum leakage current as off current while drain voltage equal to 0.1V.
4. Extraction of the field-effect mobility
Another important parameter of the a-Si TFT is the filed-effect mobility. The field-effect mobility (μFE) is determined from the transconductance gm at low drain voltage. The transfer characteristics of a-Si TFTs are similar to those of conventional MOSFETs. Accordingly, the first order of the current-voltage relation in the bulk Si MOSFETs can be applied to the a-Si TFTs, which can be expressed as
2 ] W is channel width,
L is channel length,
VTH is the threshold voltage.
If VD is much smaller than VG - VTH (that is VD << VG - VTH ) and VG > VTH, the drain current can be approximated as:
D Thus, the transconductance is defined as
D
Therefore, the field-effect mobility can be obtained by gm(max)
5. Extraction of the activation energy
When gate voltage is applied in field effect structures, positive gate voltages cause accumulation of electrons near the gate/channel interface. The induced electrons fill the available states above the Fermi level. Then, the Fermi level is shifted towards the conduction band EC (towards higher energy). Negative gate
Consecutively, Fermi level shift towards the valence band EV (towards lower energy levels).
The rate at which EF moves towards the conduction band (in n-channel a-Si TFTs) depends on the density of states located in the band gap and on the distribution of tail states close to the conduction band. When gate voltage (Vg) is small, the Fermi level is located in deep states. Increasing Vg leads to a shift of the Fermi level towards the conduction band, and the tail band states become important.
Other way to shift the Fermi level is by thermal activation of the carriers. IDS in a-Si TFTs can be temperature activated [31-33]. From measurements of the temperature dependent current at constant gate voltage (Vg), we can deduce the activation energy Ea= EC-EF as a function of the gate voltage. To use the following dependence:
σ1 is channel conductivity at temperature T1, σ2 is channel conductivity at temperature T2,
1
IDS is the drain-source current at temperature T1,
2
IDS is the drain-source current at temperature T2, KB is the Boltzmann constant,
W is the length of the channel, L is the width of the channel, d is the thickness of the channel.
By measuring the drain-source voltage at different temperatures and keeping the same drain-source current, and then the activation energy (i.e. the Fermi level position)
can be determined from the slope of the Arrhenius plot (log(IDS) vs. 1000/T) by the following equation:
(2-14) − = × ( ) × −1000⎟⎟
If we apply different gate voltages (Vg), we will get different channel conductivity and corresponding drain-source current. Therefore, different Arrhenius plots can be obtained for each value of Vg. The activation energy can be found as a function of the gate voltage. For that reason, it is necessary to measure the transfer characteristics at different temperatures.
g
6. Extraction of the density of states
The information on DOS shape is important for understanding the physical mechanisms responsible for the device behaviors. The DOS shape is related to the threshold voltage, subthreshold swing, and field effect mobility of the TFTs.
Globus et al. [34] proposed a method for evaluation of DOS in a-Si TFTs, from the dependence of Ea vs. Vg. If it is assumed that the DOS does not suffer sharp changes for energy interval about kBT, the charge of acceptor-like states Qt, filled by the gate bias is given by
(2-15) Where q is the electronic charge, Vs is the surface potential, EF0 is the equilibrium Fermi level in the silicon layer, and g(E) is the density of states. The charge Qt can also be expressed as dielectric permittivity and gate dielectric thickness, and dt is the thickness of the space charge layer. From above equations, differentiating with respect to Vg, can be
( ) ( ) Fermi level. Hence, the density of localized states can be related to the derivative of the activation energy with respect to gate bias:
( )
If we assume that the band bending in the a-Si layer is small compared to the characteristic energy of the density of states variation, then dt ≈ t where t is the a-Si layer thickness, and above equation can reduce to
( )
This method for determination of the density of states is explained in details in [34]. According to ref. [34] this technique only accounts for the acceptor-like states in the band gap. Advantage of the method is its simplicity. It is necessary to perform only field-effect measurements at different temperatures. Using this method, the density of states can be evaluated in relatively large energy interval from the band gap.
It is suitable for evaluation of changes in the density of states due to bias stress.
The activation energy method is fast and simple and is suitable to observe the changes in the density of states due to bias stress, illumination, etc. The main disadvantage of this method and also of the method of the subthreshold swing is that they do not permit the separation of the bulk states from the interface sates at the channel / gate insulator. Both methods are based on field-effect measurements that are strongly dependent on the quality of the interface between the channel material and the gate insulator. This does not permit the exact evaluation of the intrinsic DOS of the channel material, independently from the properties of this interface.
Glass Cr Al
n+ a-Si a-Si SiNx
Gate~1500Å, SiNx~3800Å, a-Si~1500Å, n+ a-Si~500Å
Source/Drain~3000Å
Fig. 2-1 Cross-sectional view of the a-Si TFT
（Conventional back channel etching structure was used）
Si Wafer
LPCVD SiNx 2300Å LPCVD Poly-Si 750Å
Fig. 2-2 Cross-sectional view of polycrystalline silicon thin film
CO2
chamber
Fig. 2-3 Schematic drawing of the supercritical fluids treatment system
Fig. 2-4 Definition of subthreshold swing (S.S.)
Chapter 3
Result and Discussion
3.1 Electrical Analysis
Figure 3-1 shows transfer characteristics of a-Si TFTs (log (NId)-Vg) before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 0.1V and temperature was 30℃. After annealing, subthreshold swing was the same as before annealing. This means that deep states in the channel region (amorphous silicon) were alike while it was been heated 2 hours at 150℃. Especially, off current was increase after annealing. Figure 3-2 shows transfer characteristics of a-Si TFTs (NId-Vg) before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. From slope of the curve, we observed that before and after annealing the mobility were similar. However, threshold voltage showed slight improvement after annealing.
Table 3-1 shows parameters (field-effect mobility, subthreshold swing and threshold voltage) of a-Si TFTs before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 0.1V and temperature was 30℃. Here we used random sampling and obtained 3 series of measured results. In these result, it could observed that mobility was slight decrease after annealing. However, threshold voltage and subthreshold swing showed slight improvement after annealing.
Figure 3-3 shows transfer characteristics of a-Si TFTs (log (NId)-Vg) before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were
measured at a drain voltage 10V and temperature was 30℃. Subthreshold swing was the same whatever before or after annealing. But, the off current increased after annealing in comparison with before annealing.
Figure 3-4 shows output characteristics of a-Si TFTs before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The output characteristics were measured in temperature of 30 ℃ . It was been observed that drain current increased after annealing.
Figure 3-5 shows transfer characteristics of a-Si TFTs before annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 0.1V and temperature 30℃~60℃. Figure 3-6 shows transfer characteristics of a-Si TFTs after annealing in supercritical fluids treatment system for 2 hours at 150℃.
TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 0.1V and temperature 30℃~60℃. Figure 3-7 shows transfer characteristics of a-Si TFTs before annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 10V and temperature 30℃~60℃. Figure 3-8 shows transfer characteristics of a-Si TFTs after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 10V and temperature 30℃~60℃.
Figure 3-9 shows temperature activation of the drain-source current of the only annealing 2 hours at 150℃ sample (Figure 3-6) for different gate voltages. From the slope of Figure 3-9, we could extract the activation energy.
Figure 3-10 shows activation energy vs. gate voltage for the samples which before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. For gate voltages from -6V to 6V, the activation energy (Ea) before and after annealing was the same.
Therefore, defect states before and after annealing were similarity.
Figure 3-11 shows density of states (DOS) vs. trap state energy (Et) for the samples which before and after annealing in supercritical fluids treatment system for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. Where trap state energy (Et) was the needful energy which electron in the trap of the amorphous silicon band gap jumped to conduction band (EC). It was observed that density of states (DOS) before and after annealing was almost the same. That is to say defect states before and after annealing were similarity.
Figure 3-12 shows transfer characteristics of a-Si TFTs (log (NId)-Vg) before and after H2O passivation for 2 hours at 150℃. TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics were measured at a drain voltage 0.1V (and 5V) and temperature 30℃. Subthreshold swing, threshold voltage and mobility were serious decrease after H2O passivation in comparison with before passivation. Figure 3-13 shows transfer characteristics of a-Si TFTs (NId-Vg) before and after H2O passivation for 2 hours at 150℃.
Table 3-2 shows parameters (field-effect mobility, subthreshold swing and threshold voltage) of a-Si TFTs before and after H2O passivation for 2 hours at 150℃.
TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics
TFTs had a gate width of 20μm and a gate length of 10μm. The transfer characteristics