TCAD Simulation of a Polysilicon Thin Film Transistor For Active Matrix Liquid Crystal Displays

1 Introduction

Polysilicon thin film transistors are attractive as active devices in driver circuits of highly integrated active-matrix liquid crystal displays (AMLCDs) and as p-channel metal-oxide-semiconductor field effect transistors (MOSFETs) in static random access memory (SRAM) cells.

Mobility in polysilicon is higher than that in amorphous silicon thus enabling integration and improved performance (e.g. high transconductance) in multi layered MOSFET structures, and low cost AMLCDs which are becoming significantly important. As with amorphous films, polysilicon films are granular, and as such, the grain size of the polysilicon is critical in terms of dictating electron mobility and electrical device performance. For example several grains existing in the device channel will result in electron scattering and electron trapping by interface and sub level traps.

Much effort has been made to increase both grain size and mobility to promote single grain channels for the TFT [e.g. 1]. A popular technique for promoting single grain growth is excimer laser crystallization which is now a well established technique for producing polycrystalline silicon films on non refractory substrates [2]. This is a process in which the film is nearly completely melted such that the molten grains grow laterally from a few isolated solid portions of seed grains that are not included in the molten process. In such a process grains can have diameters far exceeding the film thickness due to super-lateral growth phenomenon resulting in near single grain channel films [2]. However, due to complicated growth physics and energy considerations, growth faults will occur delivering sites which can hinder electron motion and degrade device performance [e.g. 3].

Of critical importance when modelling TFT devices is to gain an accurate appreciation of these effects which are typically manifested within the density of states in a material’s bandgap in addition to introducing grain boundary regions within the device channel. Many density of states functions have been proposed in the past to accurately described the characteristics of the bandgap. However, as research continues, it is becoming important to be able to implement density of states functions specific to the device rater than general expressions that at times can prove inadequate.

For this study a real time polysilicon TFT device has been developed by industrial collaborators and an attempt at simulating its electrical performance has been undertaken using Silvaco TCAD simulation software. The device has undergone excimer laser crystallization and as no vertical grain boundaries are present in the device, these have been omitted from the simulation. However, in order to utilize in house developed physical expressions for the density of states, Silvaco provides a C-interpreter which permits the user to specify specific algorithms for device simulation which has been implemented here. In addition to simulating the effects of density of states within the bandgap, an attempt has also been made to include mobility factors variable with electric field.

 

2a Results and Discussion

A thin film transistor has been created using the ATHENA framework and is shown in Figure 1. The device has been saved as a structure file entitled tft_device.str with a device width of 8mm and channel length of 4mm. In order to refine the electrical characteristics a and assign density of states to the bandgap, ATLAS has been used. ATLAS is an intutative device simulator and an example of some of the input deck parameters used for this simulation is shown in Figure 2. After the initial go atlas command, the TFT structure file is invoked using the mesh infil=tft_device.str with the width set using width= 8mm.

Figure 1. Thin film device. Gate length=4mm. Device width=8mm.

 

Figure 2. Summary of input deck used.

 

Figure 3. (a) Density of states for acceptor.
(b) Density of states for donor.

 

Figure 3 illustrates the density of both donor like and acceptor like states. These profiles have been found at this time to give the best fit to experimental data. The C-Interpreter is invoked reference to the input deck shown in Figure 2 using the commands f.tftdon and f.tftacc. A simple C file has been written and saved as defect1.c. The commands dfile= and afile= provide a means of extracting the density profiles for the donor and acceptor states respectively. It has been found that for the double exponential expressions

(1)

(2)


the parameters to supply the best fit are shown in Table 1. Figures 4 and 5 show the simulated results for 0.1V and 5V on the drain contact respectively plotted on top of the experimental raw data in each case. Clearly there is an excellent agreement between simulated and experimental raw data. Particular attention is now given to the models used. To obtain such an agreement several models have been used fundamentally based on Fermi-Dirac statistics. Each model is stated on the models line as shown in Figure 2. Of particular interest is modeling the mobility of this device. It has been found that the Lombardi CVT model [4] invoked using cvt on the models statement line (see Figure 2) improves the fit as the electric field is increased. In particular the section of this model based on the surface roughness µsr has significant effects. It has been found that controlling the proportional constants of the surface roughness µsr,n and µsr,p for both the electrons and holes respectively help to improve the simulation where:


(3)

(4)

 

 
Ntail
(eV-1cm3)
Etail
(eV)
Ndeep
(eV-1cm3)
Edeep
(eV)
Donor
1.5e19
0.02
1e19
0.1
Acceptor
0.8e19
0.01
7e18
0.08

Table 1. Summarized parameter values for equations (1) and (2).

 

Figure 4. Simulation data plotted over experimental
raw data. Drain voltage = 0.1V.

 

Figure 5. Simulation data plotted over experimental raw
data. Drain voltage=5.0V. Two simulations are shown
with band-to-band tunneling and with band-to-band
tunneling with Poole-Frenkel effect.

 

The best fit has been obtained by setting each parameter equal to 1.65e13vs-1. An interface fixed trap density of 1.83e12 has also been used. Recombination mechanisms used during this study are standard and include Shockley-Read-Hall (SRH) and Auger recombination. These are implemented on the model statement as srh and auger respectively. The carrier lifetimes have also been specified using taun0 and taup0 for electrons and holes respectively.

 

2b Effects of Band-to-Band and Poole-Frenkel Tunneling

When polysilicon TFTs are used as switching elements in active matrices, the off current has to be low, since it limits the time the video information can remain on a pixel before refreshing. The off current depends on the generation-recombination mechanisms occurring in the depletion region at the drain junction. At low Vds the leakage current is dominated by thermal generation occurring in the depletion layer close to the drain. More interestingly however is what happens at high Vds. At high Vds, high electric fields will be present at the drain junction and field enhanced generation mechanisms dominate the leakage current. Several mechanisms have been proposed to account for this and include field enhanced thermal emission i.e. Poole-Frenkel, trap assisted tunneling, phonon assisted tunneling and band-to-band tunneling. Within the TFT module it is possible to account for these effects. For this simulation two models have been used. These include band-to-band tunneling and band-to-band tunneling in addition to Poole-Frenkel effect which will be briefly be summarized. This is shown in Figure 5.

The band-to-band tunneling effect is initiated by the bbt.std on the models statement and it has three parameters that the user can alter in order to better characterize a device. If a sufficiently high electric field exists within a device local band bending may be sufficient to permit electrons to tunnel by field emission from the valence band into the conduction band. To implement this effect the current continuity equations within ATLAS are altered accordingly. The right hand side of the continuity equations are altered by the tunneling generation rate GBBT where:

(5)

Here E is the magnitude of the electric field and BB.A, BB.GAMMA and BB.B are user definable parameters that can be set within the input deck. For this simulation satisfactory values have be found and are summarized in table 2.

Tunneling
Parameters
BB.A
(V/cm)
BB.GAMMA
(arb.)
BB.B
(arb.)
Initial Guess
5e20
2.0
1.6e7

Table 2. Summarized parameter values for band-to-band tunneling.

 

The Poole Frenkel barrier lowering effect enhances the emission rate for trap-to-band phonon assisted tunneling and pure thermal emissions at low electric fields. The Poole-Frenkel effect occurs when the Coulombic potential barrier is lowered sufficiently due to the electric field. The Poole-Frenkel effect is modeled by including field effect enhancement terms for Coulombic wells and thermal emissions in the capture cross sections [5]. This model also includes the trap assisted tunneling effects in the Dirac well. This model is invoked by specifying the commands trap. tunnel and trap.coulombic on the models statements. It can be seen from Figure 5 that by including Poole-Frenkel effects the leakage current has been increased slightly and has semblance to the experimental data.

 

3 Summary

A thin film transistor for active matrix liquid crystal displays has successfully been simulated using Silvaco International’s extensive technology computer aided design (TCAD) toolset. It is clear that the simulation data is in excellent agreement with experimental data from the TFT device. Accurate density of states expressions have been presented having the form of a double exponential expression. Poole-Frenkel in addition to band-to-band tunneling have been demonstrated to increase leakage current improves the fit between simulation data and experimental data. This report was intended to demonstrate the capability of Silvaco software in modeling thin film transistors together with demonstrating the ease of use and user friendly approach of a complete modular system with powerful C-interpreter interface.

 

References

  1. Liu G., Fonash S.J., Appl. Phys. Lett. vol. 62, p2554 (1993).
  2. Im J.S., Kim H.J., Thompson M.O., Appl. Phys. Lett., vol. 63, p1969 (1993).
  3. Armstrong G.A., Uppal S., IEEE Trans. Electron Device letters, EDL 18, p315 (1997).
  4. ATLAS User’s Manual Vol. I.
  5. Lui O.K.B., Miglioratop., Solid state electronics, vol. 14, 4, p575 (1997).


Download pdf version of this article