next up previous contents
Next: 6. Summary and Conclusions Up: 5. Results Previous: 5.2 High-Field Electron Transport

Subsections



5.3 Device Simulation

With the ever increasing costs of developing state-of-the-art microelectronic technologies and devices, the optimization of semiconductor manufacturing processes on a mere experimental basis is quite debatable. This entails the employment of numerical tools which can simulate the behavior of these upcoming technologies at different levels, such as fabrication (process simulation), electrical characterization (device simulation) and integrated circuit (circuit level). While a process simulator enables the simulation of implantation, etching, diffusion, oxidation etc., the device simulator utilizes the information from the process simulator for predicting the performance of the devices in terms of current/voltage and capacitance/voltage characteristics. Additionally, the device simulator can be used to extract parameters as well as calibrate models for circuit simulation.

In this work, MINIMOS-NT [IuE04] has been used, which is a general-purpose semiconductor device simulator that provides steady-state, transient, and small-signal analysis of arbitrary two- and three-dimensional device structures. The simulator is capable of dealing with different material systems such as Si/Ge and group III/Vs compound semiconductors [Palankovski04] as well as advanced device structures including SOI. The tool is equipped with a comprehensive set of physically based models. As a part of this work, the simulator has been extended to include the models that describe the effect of strain on the electron mobility. The models are suitable for drift-diffusion based device simulations.


5.3.1 Effective Mobility


Parameter Unit Value Parameter Unit Value
$ B$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 485.6 $ E_\ensuremath{{\mathrm{ref}}}$ $ \ensuremath{{\mathrm{Vm^{-1}}}}$ 1.0e4
$ \gamma_1$ 1 0.79 $ S_\ensuremath{{\mathrm{ref}}}$ $ \ensuremath{{\mathrm{Vm^{-1}}}}$ 123.7
$ \tau_1$ 1 $ -2.8$ $ A$ 1 2.8
$ C_1$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 1.73 $ \times 10^{4}$ $ S_{exp}$ 1 0.05
$ C_3$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 1.23 $ \times 10^{3}$ $ D_1$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 3.376 $ \times 10^{23}$
$ C_2$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 155 $ D_3$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 2.645 $ \times 10^{20}$
$ N_\ensuremath{{\mathrm{ref}}}$ $ \ensuremath{{\mathrm{cm^{-3}}}} $ 1.0 $ \times 10^{17}$ $ D_2$ $ \ensuremath{{\mathrm{cm^2V^{-1}s^{-1}}}}$ 2.645 $ \times 10^{20}$
$ f_1$ 1 1.7 $ E_1$ $ \ensuremath{{\mathrm{MVcm^{-1}}}}$ 0.2
$ f_2$ 1 1.3 $ E_2$ $ \ensuremath{{\mathrm{MVcm^{-1}}}}$ 1.5


Table 5.8: Parameter values for the inversion layer mobility model.


For the evaluation of the effective mobility MOS structure were created with uniform substrate doping concentrations ranging from $ 10^{16}$ cm$ ^{-3}$ to $ 10^{18}$ cm$ ^{-3}$. The low-field mobility was calculated as described in Section 4.5. Due to the quantum confinement of electrons, the peak of the inversion layer concentration is not at the interface but is slightly displaced. This effect leads to a decrease in the net inversion layer capacitance, which reduces the drain current. To capture this phenomenon for device simulation purposes efficiently two approaches [van Dort94] [Jungemann01] have been suggested. In this work, the approach based on modifying the conduction band edge through a quantum corrected potential has been used [Jungemann01].

Fig 5.17a shows a comparison of the simulated effective mobilities with the universal mobility curves obtained by Takagi [Takagi94]. The parameter values used for the simulations are summarized in Table 5.8. The results show a very good agreement between the simulated and measured values. Fig. 5.17b shows the mobility enhancement as obtained from (4.98). The model shows good agreement to the strained mobility data obtained by Currie [Currie01] at MIT for the case of biaxially strained Si grown on SiGe with 20% Ge content.

\includegraphics[width=2.8in]{figures/rot_UnivMob-Uns.ps}         \includegraphics[width=2.8in]{figures/rot_EEeff2025-new3.ps}
$\textstyle \parbox{2.8in}{(a) \hfill (b)}$

Figure 5.17: Comparison of the effective mobility versus field for (a) unstrained Si and (b) strained Si from the model and measurements reported from various groups. MIT [Currie01], IBM [Rim02], Toshiba [Maeda03], Hitachi [Sugii01], Stanford [Welser94a]

.


5.3.2 The dotFET Structure

The dotFET is a novel device concept proposed by Schmidt and Eberl [Schmidt01] that utilizes the advantages of strain and SOI technology and has the potential for enhancing device performance. Its structure relies on the self-assembly of coherent, defect-free Ge dots on Si. The Ge dots are grown in the Stranski-Krastanow [Stranski37] mode based on interface thermodynamics, wherein a strained SiGe film grows pseudomorphically up to a few monolayers followed by the formation of three-dimensional islands. Depending on the growth conditions, the shape of these islands can vary from pyramids, to truncated pyramids to huts and dome shaped clusters. The SiGe dots are deposited on a pre-patterned (001) Si substrate and have dimensions ranging from 100 to 300 nm. Next, a Si capping layer is grown on top of the SiGe dot at low temperature, after which the gate oxide is grown and the polysilicon gate formed. Due to the three-dimensional nature of the dot, the strain in the capping layer is not uniform but distributed with both uniaxial and shear components present. Then, the Ge dot can be removed using selective wet chemical etching to leave a Si free standing bridge, which forms the channel of the actual device structure. To provide additional mechanical stability to the free standing bridge, a nitride layer could be deposited before the SiGe dot removal. The dotFET structure is shown in Fig. 5.18. This structure combines potentially four advantages.

  1. The short channel behavior that hampers the off-state leakage current of small devices is improved.
  2. The presence of strain in the Si layer increases the carrier mobility and thereby reduces the delay time of the transistor.
  3. The Ge dots also offer the possibility of aligning themselves in stacks on top of each other to form a 3D stack. Therefore the area consumption is less.

\includegraphics[width=2.9in,angle=0]{figures/dotfet_freg2.eps}

Figure 5.18: Schematic cross section of the dotFET structure.

The performance of the dotFET is strongly dependent on the amount of strain and its distribution within the Si capping layer. The strain state of the Si depends on (i) growth conditions for the capping layer, (ii) size, shape and composition of the Ge dots, (iii) material choice of the gate dielectric, (iv) fabrication of source/drain junctions, (v) impact on the Si bridge during and after the removal of the dot. The strain distribution can be determined either by using experimental techniques such as X-ray diffraction or theoretically from atomistic simulations using Tersoff potentials [Marchetti05] or finite element calculation methods.

5.3.2.1 The Strain Profile

In this work, the strain distribution used was provided by University Milano [Vastola] who have used an atomistic molecular dynamics code based on the Tersoff [Tersoff89] interatomic potential for the calculations. The code delivers the stress and strain tensor components at each atomic site. From the complete data set a (010) plane passing through the island apex is chosen. Fig. 5.19 shows the different components of the strain tensor. The strain distribution is highly non-uniform with large amount of strain present close to the Si/SiO$ _2$ interface.

5.3.2.2 Interfacing with MINIMOS-NT

The geometry of the dotFET device structure was created using the process simulator Athena [ATHENA02]. A layer of Ge was grown on a uniformly doped Si substrate and the Ge layer was etched to create a dome shape. A Si layer was grown on top of this dome followed by an oxide layer. Next, a polysilicon gate was formed and then source and drain implants were performed to obtain the device structure. The diameter of the dome was kept to be 160 nm while the thickness of the Si capping layer was 30nm. The structure was next modified to keep/remove the Ge dot using the devEdit [Sil02] tool as well as to make the device structure compatible for conversion to MINIMOS-NT format. After the structure was converted from the .str (Athena) format to the .pif (MINIMOS-NT) format, it was remeshed using the cgg tool [Cervenka99]. As a final step, a cut (x-z plane) of the strain distribution was read into the device simulator taking into account the proper scaling of the x and z coordinates. The z-direction denotes the growth direction ([001]) whereas the x-direction is along the channel. All the steps after the creation of the device structure using Athena were automated through scripting. Fig. 5.20 shows the final device structure with a substrate doping of $ 10^{15}\ensuremath{{\mathrm{cm^{-3}}}}$ and a 3 nm thick oxide layer. An additional threshold voltage adjust implant of $ 8\times10^{11} \ensuremath{{\mathrm{cm^{-2}}}}$ at 2.3 keV was performed.

\includegraphics[width=2.5in]{figures/rot_Exx-pif.eps} \includegraphics[width=2.5in]{figures/rot_Eyy-pif.eps}

\includegraphics[width=2.5in]{figures/rot_Ezz-pif.eps} \includegraphics[width=2.5in]{figures/rot_Exy-pif.eps}

\includegraphics[width=2.5in]{figures/rot_Eyz-pif.eps} \includegraphics[width=2.5in]{figures/rot_Exz-pif.eps}

Figure 5.19: Distribution of the strain tensor components as obtained from atomistic growth simulations [Vastola].

5.3.2.3 Simulation Results

Fig. 5.21 shows the distribution of the low-field mobility in the unstrained and strained Si bridge. It can be clearly seen from Fig. 5.21b that strain does increase the value of the low-field mobility in the channel area. This enhancement of the mobility comes from the significantly large ($ \ge 2\%$) values of the diagonal components of the strain tensor existing in and around the channel region.

\includegraphics[width=3.0in]{figures/rot_DeviceDoping-pif3.eps}
Figure 5.20: The dotFET device structure used for simulation.



\includegraphics[width=2.5in]{figures/rot_Mobility_uns2.eps}         \includegraphics[width=2.5in]{figures/rot_Mobility_str2.eps}
$\textstyle \parbox{2.5in}{(a)\hfill (b)}$

Figure 5.21: Distribution of the mobility for an (a) unstrained and (b) a strained Si bridge..

This mobility enhancement directly affects the transfer characteristics, as shown in Fig. 5.22. A 37% and 43% enhancement in the drain current is observed for the cases when the Ge dot is present and removed, respectively. The threshold voltage of the device can be estimated to be 0.1 V. Fig. 5.23 shows the output characteristics for two different values of $ V_\mathrm{GS}$. There is about 40% enhancement in the linear regime and more than 10% enhancement in the saturation region, for both the cases where the Ge dot is present and removed. It should be noted that the simulation results present an optimistic scenario where in the strain values are large. However, depending on the processing conditions, the final fabricated device structure may experience some strain relaxation, thereby resulting in a reduced enhancement in the drain currents.

\includegraphics[width=2.7in]{figures/rot_Vts_all.eps}

Figure 5.22: Simulated transfer characteristics of the dotFET.

\includegraphics[width=2.7in]{figures/rot_IdVds_all.eps}

Figure 5.23: Simulated output characteristics of the dotFET.


next up previous contents
Next: 6. Summary and Conclusions Up: 5. Results Previous: 5.2 High-Field Electron Transport

S. Dhar: Analytical Mobility Modeling for Strained Silicon-Based Devices