1. Introduction
1.1. Semiclassical electron transport
Semiclassical Boltzmann transport has been the principal theory in the field of modeling and simulation of semiconductor technology since its early development. To date, most commercial device simulations including the full-band Monte Carlo (FBMC) method are based on the solution of the Boltzmann transport equation (BTE) and its simplified derivatives such as the hydrodynamic (HD) equations and the drift-diffusion (DD) model. The Boltzmann transport equation expresses the global variation of the non-equilibrium distribution function,
where
Full analytical solutions of the BTE are possible only under very restrictive assumptions. The two classes of computational/numerical techniques that are used to solve the BTE are as follow [2][3][4][5]: (1) First, in
In the last decade, with the continued downsizing of device dimensions into the nanoscale regime, many new and challenging questions have emerged concerning the physics and characterization of these small devices. However, contrary to the technological advances, present state-of-the-art in device simulation is lacking in the ability to treat the new challenges pertaining to device scaling [12]. First, for devices where
2. The Monte Carlo method
Monte Carlo (MC) method was originally used and devised by Fermi, Von Neumann and Stanislaw Ulam to solve the BTE for transport of neutrons in the fissile material of the atomic bomb during the Manhattan Project of World War II [26]. Since these pioneering times in the mid 1940's, the popularity and use of the MC method has grown with the increasing availability of faster and cheaper digital computers. Its application to the specific problems of high-field electron transport in semiconductors is first due to Kurosawa [27] in 1966. Shortly afterwards the Malvern, UK, group [28] provided the first wide application of the method to the problem of the
The Monte Carlo method is the most popular method used to solve the Boltzmann transport equation without any approximation made to the distribution function. In the Monte Carlo method, particles are used to represent electrons (holes) within the device. For
Initialization.
Solution of the field (Poisson) equation: determine forces on electrons.
Simulate the electron dynamics (free-flight and scattering):
Accelerate the electrons.
Determine whether the electron experiences a scattering/collision.
If electron scatters, select the scattering mechanism.
Update the electron position, energy and wavevector.
Charge assignment to the numerical nodes/grids.
Compute measurable quantities such as average energy, average velocity, and mobility of the particle ensemble.
Repeat steps 2-5 for each iteration.
The
where
To simulate an actual
3. Description of the Monte Carlo device simulator
The major computational components of the 3-D Monte Carlo device simulator (MCDS 3-D) used in this work are shown in Figure 1. Regarding the Monte Carlo transport kernel, intravalley scattering is limited to acoustic phonons. For the intervalley scattering, both
Figure 2 shows the rates for various scattering mechanisms as a function of electron energy used in the MCDS 3-D simulator. At present, impact ionization and surface-roughness scattering are not included in the model. Impact ionization is neglected, as, for the drain biases used in the simulation, electron energy is typically insufficient to create electron-hole pairs. Also, since it is not quite clear how surface roughness scattering can be modeled when carriers are displaced from the interface due to the quantum confinement effects, it is believed that its inclusion is most likely to obscure the quantum confinement effects. Also, band-to-band tunneling and generation and recombination mechanisms have not been included in the simulations.
The Incomplete Lower-Upper (ILU) decomposition method has been employed for the solution of the 3D Poisson equation. To treat full Coulomb (electron-ion and electron-electron) interactions properly, the simulator implements two real-space molecular dynamics (MD) schemes: the particle-particle-particle-mesh (P3M) method and the corrected Coulomb approach. The effective force on an electron is computed as a combination of the short-range molecular dynamics force and the long-range Poisson force. The implementation details of these models and methodologies have been discussed in Ref. [35].
Quantum mechanical size-quantization effects have been accounted for via a parameter-free effective potential scheme [38]. The approach is based on a perturbation theory around thermodynamic equilibrium and derived from the idea that the semiclassical Boltzmann equation with the quantum corrected potential and the Wigner equation should possess the same steady state. It leads to an effective potential/field, which takes into account the discontinuity at the Si/SiO2 barrier interface due to the difference in the semiconductor and the oxide affinities. It possesses no fitting parameters, as the size of the electron (wavepacket) is determined from its energy. The resultant quantum potential is, in general, two-degrees smoother than the original Coulomb and barrier potentials of the device, i.e. possesses two more classical derivatives, which essentially eliminates the problem of the statistical noise. The calculated quantum barrier field (QBF) for low-energy (left column) and high-energy (right column) electrons are shown in Figure 3 having the following salient features: (1) QBF decays almost exponentially with distance from the Si/SiO2 interface proper; (2) QBF increases with increasing the wavevector of the carriers along the normal direction; (3) The contour plots clearly reveal the fact that the electrons with lower momentum feel the quantum field far from the interface proper, and (4) A similar trend is also observed with the variation in electron energy—electrons with higher energy can reach the vicinity on the interface, thus, behaving as classical point-like particles.
The
where
where
where
or,
where
4. Statistical enhancement: The self-consistent event biasing scheme
Statistical enhancement in Monte Carlo simulations aims at reduction of the time necessary for computation of the desired device characteristics. Enhancement algorithms are especially useful when the device behavior is governed by rare events in the transport process. Such events are inherent for sub-threshold regime of device operation, simulations of effects due to discrete dopant distribution as well as tunneling phenomena. Virtually all Monte Carlo device simulators with statistical enhancement use
4.1. Theory of Event Biasing
The Ensemble Monte Carlo (EMC) technique is designed to evaluate averaged values
Where
where
and
If both, initial
While
A recursive replacement of equation (9) into itself gives rise to the von-Neumann expansion, where the solution
Consider the second term in (14) augmented with the help of two probabilities
It takes values determined by the second half with a probability given by the product in the first half in equation (15).
The iterative character of the multiple integral (15) has been used to introduce a consecutive procedure for construction of the trajectories. It can be shown that a single trajectory, obtained by successive applications of
The initial probability
Equation (18) accounts that only trajectories which belong to
The Boltzmann equation for Coulomb carriers becomes nonlinear via the interaction component
The charge density
Accordingly, the correct F is provided by the Poisson equation. As the evolution is Markovian,
4.2. Event biasing methods employed
Three different event-biasing techniques have been employed in this work. The chosen biasing techniques aim at increasing the number of
4.2.1. Biasing the initial/boundary temperature
Denoting the equilibrium distribution,
where,
which corresponds to higher temperature
The biasing scheme is illustrated in Figure 5. Increasing
4.2.2. Particle Split
Particle weight is controlled by choosing a desired weight
A choice of
4.2.3. Biasing Phonon Scattering (e-a)
Artificial carrier heating can be achieved by biasing the phonon scattering rates. For a given scattering mechanism, the probability for phonon absorption is increased at the expense of phonon emission, controlled by a parameter
If in the course of the simulation, a phonon absorption is selected, the particle weight is updated by a multiplication with
4.3. Results from biasing experiments
The MOSFET chosen for the simulation experiments has gate length of 15 nm, channel doping of
Figure 8 shows the standard deviation in electron number as a function of evolution time in a certain cell within the channel region. One can clearly see the improvement achieved from the use of event biasing techniques in the simulation. The standard deviation is least for method (a) where the initial/boundary temperature of the electrons is biased (at
Regarding the validation of the biasing techniques,
The corresponding
5. Conclusion
In conclusion, three event biasing techniques for particle-based Monte Carlo simulations of semiconductor devices have been derived in presence of both initial and boundary conditions and generalized for self-consistent simulations. All these approaches are confirmed and validated by simulations of a MOSFET with a channel length of 15 nm in the subthreshold regime of operation. A bias technique, particularly useful for small devices, is obtained by injection of hot carriers from the boundaries. The coupling with the Poisson equation requires a precise statistics in the source/drain regions. It is shown that a combination of event biasing and population control approaches is advantageous for this purpose.
Acknowledgements
Shaikh Ahmed would like to thank Sharnali Islam for her assistance in editing the text. This work was partially supported by the National Science Foundation (Grant No. ECCS-1102192). Computational resource/time supported by the ORAU/ORNL High-Performance Computing Grant 2009 is acknowledged. Mihail Nedjalkov would like to acknowledge the Austrian Science Fund (FWF) P21685-N22 and the Bulgarian Science Fund DTK 02/44.
References
- 1.
Iafrate G. Introduction. In: Ferry DK., Jacoboni C. (ed.) Quantum Transport in Semiconductors. Plenum Press: New York and London; 1992. pxv-xxi. - 2.
Ferry DK. Semiconductor Transport. Taylor & Francis: London, UK; 2000. - 3.
Lewis EE., Miller, WF. Computational Methods of Neutron Transport, American Nuclear Society, Inc.: La Grange Park, Illinois; 1993. - 4.
Majumdar A. Development of a Multiple Perturbation Monte Carlo Method for Eigenvalue Problems and Implementation on Parallel Processors. Ph.D. dissertation: The University of Michigan; 1996. - 5.
Duderstadt JJ., Martin WR. Transport Theory. John Wiley & Sons: New York; 1979. - 6.
Hammersley JM. et al. Monte Carlo Methods. John Wiley & Sons: New York; 1964. - 7.
Lux I., Koblinger L. Monte Carlo Particle Transport Methods: Neutron and Photon Calculations. CRC Press: Florida; 1991. - 8.
Rubinstein RY. Simulation and the Monte Carlo Method. John Wiley & Sons: New York; 1981. - 9.
Spanier J., Gelbard EM. Monte Carlo Principles and Neutron Transport Problems. Addison-Wesley: Reading, Massachusetts; 1969. - 10.
Kosina H., Nedjalkov M., Selberherr S. Theory of the Monte Carlo method for semiconductor device simulation. IEEE Transactions on Electron Devices 2000; 47 1898–1908. - 11.
Fjeldly TA., Shur MS. Simulation and Modeling of Compound Semiconductor Devices. In: Shur MS. (ed.) Compound Semiconductor Electronics: The Age of Maturity. World Scientific; 1996. 317–364. - 12.
Vasileska D., Mamaluy D., Khan H. R., Raleva K., Goodnick SM. Semiconductor Device Modeling. Journal of Computational and Theoretical Nanoscience 2008; 5 1–32. - 13.
Ferry DK., Goodnick SM. Transport in Nanostructures. Cambridge University Press; 1997. - 14.
Wigner E. On the quantum correction for thermodynamic equilibrium. Phys. Rev 1932; 40 749–759. - 15.
Feynman R., Kleinert H. Effective classical partition functions. Phys. Rev. A 1986; 34 5080–5084. - 16.
Datta S. Quantum Transport: Atom to Transistor. Cambridge University Press; 2005. - 17.
Lake R., Klimeck G., Bowen RC., Jovanovic D. Single and multiband modeling of quantum electron transport through layered semiconductor devices J. Appl. Phys. 1997; 81 7845. - 18.
Iafrate GJ., Grubin HL., Ferry DK. Utilization of quantum distribution functions for ultra-submicron device transport. Journal de Physique 1981; 42 (Colloq. 7) 307–312. - 19.
Ferry DK., Zhou JR. Form of the quantum potential for use in hydrodynamic equations for semiconductor device modeling. Phys. Rev. B 1993; 48 7944–7950. - 20.
Ferry DK. Effective potential and the onset of quantization in ultrasmall MOSFETs. Superlattices and Microstructures 2000; 28 419–423. - 21.
Ringhofer C., Gardner CL. Smooth QHD model simulation of the resonant tunneling diodes. VLSI Design 1998; 8 143–146. - 22.
Vasileska D., Khan H., Ahmed SS. Modeling Coulomb effects in nanoscale devices. Journal of Computational and Theoretical Nanoscience 2008; 5(9) 1793–1827. - 23.
Broglie L. De. Sur la possibilité de relier les phénomènes d'interference et de diffraction á la théorie des quanta de luminère. C. R. Acad. Sci. Paris 1926; 183 447–448. - 24.
Madelung E. Quantatheorie in hydrodynamischer form. Z. Phys. 1926; 40 322–326. - 25.
Bohm D. A suggested interpretation of the quantum theory in terms of hidden variables I and II. Phys. Rev. 1952; 85 166–193. - 26.
Fischetti M., Laux S. Monte Carlo study of electron transport in silicon inversion layers. Physical Review B 1993; 48 2244–22743. - 27.
Kurosawa. J. Phys. Soc. Jpn. 1966; 21 424. - 28.
Fawcett W., Boardman DA., Swain S. J. Phys. Chem. Solids 1970; 31 1963. - 29.
Jacoboni C., Reggiani L. The Monte Carlo Method for the Solution of Charge Transport in Semiconductors with Applications to Covalent Materials. Rev. Modern Phys 1983; 55 645–705. - 30.
Price PJ. Semiconductors and Semimetals 1979; 14 249–308. - 31.
Gross WJ. Three-dimensional particle-based simulations of deep sub-micron MOSFET devices. Ph. D. dissertation: Arizona State University; 1999. - 32.
Lundstrom M. Fundamentals of Carrier Transport. Cambridge University Press; 2000. - 33.
Tomizawa K. Numerical Simulation of Submicron Semiconductor Devices. Artech House: Boston; 1993. - 34.
Kunikuyo T., Takenaka M., Kamakura Y., Yamaji M., Mizuno H., Morifuji M. A Monte Carlo simulation of anisotropic electron transport in silicon using full band structure and anisotropic impact-ionization model. Journal of Applied Physics 1994; 75 297–312. - 35.
Ahmed SS. Quantum and Coulomb Effects in Nanoscale Devices. Ph.D. dissertation: Arizona State University; 2005. - 36.
Ferry DK. First-Order Optical and Intervalley Scattering in Semiconductors. Phys. Rev. B 1976; 14 1605–1609. - 37.
Gross WJ., Vasileska D., Ferry DK. 3D Simulations of Ultra-Small MOSFETs with Real-Space Treatment of the Electron-Electron and Electron-Ion Interactions. VLSI Design 2000; 10 437–452. - 38.
Ahmed SS., Ringhofer C., Vasileska D. Parameter-Free Effective Potential Method for Use in Particle-Based Device Simulations. IEEE Transactions on Nanotechnology 2005; 4 465–471. - 39.
Wordelman C., Kwan T., Snell C. Comparison of statistical enhancement methods for Monte Carlo semiconductor simulations. IEEE Transactions on Computer-Aided Design 1998; 17(12) 1230–1235. - 40.
Rota L., Jacoboni C., Poli P. Weighted Ensemble Monte-Carlo. Solid-State Electronics 1989; 32(12) 1417–1421. - 41.
Kosina H., Nedjalkov M., Selberherr S. The stationary Monte-Carlo method for device simulaton–Part I and Part II. Journal of Applied Physics 2003; 93(6) 3553–3571. - 42.
Jacoboni C. Principles of Quantum Transport. In: Ferry D., Jacoboni C. (ed.) Quantum Transport in Semiconductors. Plenum Press: New York and London; 1992. p1–15. - 43.
Jacoboni C. A New Approach to Monte Carlo Simulation. PRIEDM, IEEE Electron Devices Society 1989; 469–472. - 44.
Chau R., Boyanov B., Doyle B., Doczy M., Datta S., Hareland S., Jin B., Kavalieros J., Metz M. Silicon nano-transistors for logic applications. Proceedings of the 4th Int. Symp. on Nanostructures and Mesoscopic Systems, 17-21 February 2003, Tempe Mission Palms Hotel, Temepe, Arizona, USA.