Title: Semiconductor Device Noise Models Based on Semiclassical Transport 51153
1Semiconductor Device Noise Models Based on
Semiclassical Transport 5115-3
- Can E. Korman
- Department of ECE
- The George Washington University
- Washington, DC 20052
- Collaborators
- A. Piazza (Applied Wave Research, Inc.)
- I. D. Mayergoyz (University of Maryland)
Noise Information in Nanoelectronics, Sensors
Standards SPIE's First International Symposium on
Fluctuations and Noise - Santa Fe, NM - June 1-4,
2003
2Outline of the Talk
- Introduction
- Semiclassical Transport Theory
- Stochastic Differential Equations Theory
- Terminal Current Spectral Density Computation
- Numerical Results
- Numerical Techniques
- Conclusion
31- Introduction
- Motivation
- Noise characterization of devices
- Goal Connecting Electron Kinetics to Terminal
Noise Characteristics - Related noise calculation techniques
4Motivation
- Low power low voltages and currents
- Low voltage or current noise becomes critical
problem - Accurate noise modeling is an essential element
in modern device simulators
5Noise Characterization of DevicesTerminal Noise
Spectral Density
- Thermal noise
- GR noise
- Flicker or 1/f noise
- Shot noise
6Connecting Electron Kinetics to Terminal Noise
Characteristics
- Comprehensive approach for device noise
characterization Electron kinetics is described
by the Semiclassical Transport Theory
- Develop noise models in the framework of
Semiclassical Transport Theory suitable for
efficient numerical simulation and CAD
7Connecting Electron Kinetics to Terminal Noise
Characteristics
8Related Noise Calculation Techniques
- Probabilistic approach reasonable assumption on
the underlying microscopic process - Langevin Method add random source to
deterministic differential equation - Monte Carlo Method Direct simulation of random
electron kinetics - computationally intensive - SDE theory based model Connect electron
scattering characteristics to terminal noise
spectral density
92- Semiclassical Transport Theory
- The adiabatic principle allows to separate
crystal vibration from electron motion - The independent electron approximation treats
electron-electron interactions by one electron
effective potential - Each electron moves independently in a perfectly
periodic potential due to lattice and all other
electrons - Scattering mechanisms describe interaction of
electrons with the lattice
102- Semiclassical Transport Theory (cont.)
- Semiclassical Transport Theory predicts how, in
the absence of collisions, the position and wave
vector of each electron evolve in the presence of
macroscopic electric and magnetic fields - Maxwells Equations describe the macroscopic
fields - Electron position and wave vector correspond to a
wave packet of Bloch electrons - Scattering causes random sudden jumps in the
electron wave vector
112- Semiclassical Transport Theory (cont.)
- Electron scattering from one wave vector to
another is characterized by the transition
probability function - Paulis exclusion principle limits the scattering
rates - Such a process can be described in terms of
occupation probability that satisfies Boltzmann
transport equation (BTE)
12Motion of wave packet of Bloch electrons
- All distinct electron states are confined to a
unit cell in reciprocal space first Brillouin
zone - Velocity of wave packet
- Force due to External Fields - Wave packet
behaves classically
13Macroscopic External Fields Poissons Equation
- Electrostatic potential ?
- Electric field E
- Conduction valance band electron distributions
f C and fV - Ionized impurity densities of donors acceptors
ND and NA
14Random Force Electron Scattering
- Almost instantaneous changes in the value of the
electron wave vector - Phonons deformation potential
- Thermal generation-recombination Shockley Read
Hall generation-recombination model - Ionized impurities screened Coulomb potential
- Phonon scattering elastic, inelastic, isotropic,
unisotropic, intervalley, intravalley, etc
15Scattering Transition in k-space
- Scattering probability per unit time
- Transition rate Born-Von Karman boundary
conditions
16Paulis Exclusion Principle Scattering
transition from k to k
- Transition into an occupied level is forbidden
- Transition rate into partially filled group of
level
- Steady-state transition rate
- f0 steady-state electron disribution function
17Electron motion in reciprocal-space
18Boltzmann Transport Equation (BTE) in terms of
electron occupation probability f0
LHS denotes flux in phase space
- RHS denotes flux due to scattering events
19Summary of Semiclassical Transport Model
- Electrons drift due to macroscopic electrostatic
field - Electrons scatter due to deviations from perfect
periodicity of potential - Paulis exclusion principle limits scattering
rates - Semiclassical transport embedded in Poissons and
Boltzmann transport equations
20Electron Kinetics Stochastic Differential
Equation (SDE) Electron motion can be
interpreted as a differential equation driven by
inhomogeneous randomly weighted Poisson process
SDE
- Scattering rate (Poisson)
- Transition rate
213 - Stochastic Differential Equations (SDE) Theory
- Differential equations describing processes
driven by random excitation - Wiener processes (continuous in time) - diffusion
- Poisson processes (discontinuous in time) -
scattering - These are continuous-state Markov processes
- Process is completely characterized by its
transition probability density function - Focker-Planck Equation
- Kolmogoroff-Feller equations
22Stochastic Differential Equation General Case
- Differential equation driven by stochastic
process j(y,t)
- Stochastic process j(y,t) is characterized by
- Intensity function (scattering rate)
- Transition rate
23Forward Kolmogoroff-Feller Equation (KFE)
General Case
- The stochastic process y is characterized by the
transition probability density function
24Kolmogoroff-Feller Equation for Electron Motion
Described by Semiclassical Transport Theory
The electron transition probability density
function ? satisfies
253- Summary of SDE
- Electron motion corresponds to Markov process
- Process completely characterized by transition
probability r - Transition probability function satisfies KFE
- KFE identical to linear BTE
- All stochastic information on electron kinetics
is embedded in KFE or equivalently, linear BTE
264 - Terminal Current Spectral Density Computation
Goal compute spectral density of the terminal
current
- Calculate the transition probability function r
- Relate electron motion in phase-space state to
the induced current at the device terminals
Ramo-Schockley theorem - Combine the results of Steps (1) and (2) to
calculate the autocovariance function or
spectral density of induced terminal noise.
27Ramo-Shockley Theorem
- The extension of Ramos theorem by Cavalleri et
al establishes a connection between induced
terminal currents and motion of carriers inside a
device - The model admits arbitrary mobile charge
distribution and conduction current density
28Ramo-Shockley Theorem (cont.)
- The model shows that for terminal potentials Vj
and terminal currrents Ij
E(0) is the electrostatic field due exclusively
to the applied potentials at the terminals j1,
, n
29Ramo-Shockley Theorem (cont.)
- At contact j the induced terminal current due to
each electron with position x and wave vector k
is
30Autocovariance function computation of terminal
current
By definition, the autocovariance function of the
observed terminal current is
In terms of the Ramo-Shockley function ij and an
auxiliary function g it can be shown that the
autocovariance function of terminal current j is
as follows
where ? is the transition probability density
function and f is the steady-state electron
distribution function
31Autocovariance function computation(cont.)
- Terminal current autocovariance function
- Auxiliary function g satisfies the transient BTE
- Subject to the following special initial
conditions
32Current Spectral Density Computation
- The spectral density is defined as the Fourier
transform of the autocovariance function - Employing this definition it can be shown that
the spectral density can be calculated in terms
of an auxiliary function G
- G satisfies the Fourier transform of the
transient BTE
334 Summary of Terminal Current Spectral Density
Computation
- Key computations for the terminal current
spectral density function are reduced to solution
of Kolmogorov-Feller Equation or Boltzmann
Transport Equation - Ramo-Schockley theorem directly connects electron
motion to current induced at the device terminals - A computationally efficient approach in terms of
the effective distribution function G(x,k,w) was
presented
345 Sample Numerical Results
- Calculations for n-doped bulk silicon device
- Thermal noise
- Generation-Recombination noise
- Employ Spherical harmonics technique for
numerical calculations
35Spectral Density Thermal Noise
36Autocovariance Thermal Noise
37Autocovariance Thermal Noise
38Spectral Density GR Noise
39Spectral Density GR Thermal Noise
405 - Summary
- Simulation results are presented for bulk n-type
silicon as a function of - Electric field
- Temperature
- Electron capture cross section
- Results for thermal and Generation-Recombination
noise are presented - Numerical results are in agreement with measured
data - Results are obtained using computationally
efficient Spherical Harmonic technique
416 - Numerical Implementation
- Problem Steady state solution of Poisson and
non-linear Boltzmann transport equations - Problem Transient solution of linear BTE with
special initial condition in frequency domain - The BTE/KFE is a non-linear 7D equation (3D
space, 3D momentum and 1D time) - Spherical harmonics technique is employed to
reduce the number of dimensions in momentum space
42Mathematical techniques
- Poissons and non-linear Boltzmann transport
equations are consistently solved employing
Gummel block iteration method - BTE is iteratively solved in real-space
- At a given point in real space, 3D non-linear
equation in reciprocal-space is solved using
spherical harmonics - No time iteration is required equation is solved
in frequency domain
43Spherical Harmonics Polynomials
- Scattering involves transitions between spheres
- Spherical harmonics give a more efficient
representation of the distribution function
44Gummel Iteration
start
Poissons equation
BTE
stop
n
y
45Real-space iteration
start
BTE at xi
i0
i i1
n
stop
Last i
y
n
y
466 Summary of Numerical Techniques
- The approach requires the numerical solution of
- the steady-state Poisson equation
- the steady-state non-linear Boltzmann transport
equation - the transient linear Boltzmann transport equation
- Gummel iteration used to decouple Poisson and
Boltzmann transport equations - BTE solved in space iteratively
- BTE solved in reciprocal-space using spherical
harmonics - Transient solution performed in frequency domain
477 - Conclusion
- Unified approach to model noise in the framework
of semiclassical transport theory and stochastic
differential equations - Noise is exclusively due to electron scattering
- Model accounts for Paulis exclusion principle
- Random electron motion in phase-space directly
connected to induced terminal current Ramos
theorem - Adapted spherical harmonics technique to solve
transient BTE equation in frequency domain - Numerical solution approach based on well
established techniques practical for CAD
48Thank you for your attention
49Reference Publications and Presentations
- C. E. Korman and I. D. Mayergoyz, Semiconductor
Noise in the Framework of Semiclassical
Transport,'' Phys. Rev. B, Vol. 54, No. 24, pp.
17620-17627, 15 December, 1996. - A. J. Piazza, C. E. Korman and Amro M. Jaradeh,
A Physics Based Semiconductor Noise Model
Suitable for Efficient Numerical Implementation,
IEEE Tran. On CAD of Integrated Circuits and
Systems, Vol. 18, No. 12, pp. 1730-1740, December
1999. - A. J. Piazza and C. E. Korman, Semiconductor
Noise Computation Based on the Deterministic
Solution of Poisson and Boltzmann Transport
Equations, VLSI Design, 1998, vol. 8, Nos. (1-4),
pp. 381-385.
50Reference Publications and Presentations (cont.)
- C. E. Korman and A. J. Piazza, Computation of
Semiconductor Noise for Semiclassical Transport,
proceedings of the International Semiconductor
Device Research Symposium, Charlottesville,
Virginia, 1995. - A. J. Piazza and C. E. Korman, Computation of the
Spectral Density of Noise in Bulk Silicon Based
on the Solution of the Boltzmann Transport
Equation, VLSI Design, 1998, Vol. 6, Nos. (1-4),
pp. 185-189.