Transport w 2D & 1D
strukturach półprzewodnikowych
Jacek A. MajewskiInstitut Fizyki Teoretycznej &
Interdyscyplinarne Centrum Modelowania Materiałów Universytet Warszawski ICM UW, Warszawa, 30 kwietnia, 2009 www.fuw.edu.pl
Rewolucja informatyczna
Wykładniczy wzrost
technologii informatycznych
Przetwarzana, zapamiętywana, i transmitowana gęstość informacji (liczba bitów na jednostkę pow.) rośniewykładniczo.
Ten szybki wzrost powoduje zmiany cywilizacyjne
wpływa silnie na światową gospodarkę
określa sposób w jaki pracujemy, bawimy się, etc.
Moore’s Law
Wykładniczy postęp w integracji
LiczbaLiczba Liczba
Liczba tranzystortranzystortranzystorótranzystoróóówwww w w w procesorzew procesorzeprocesorzeprocesorze podwaja
podwaja podwaja
podwaja sisisię co ~2 si co ~2 co ~2 co ~2 latalatalatalata PRAWO
MOORE’a
Rozwój technologii obwodów scalonych
Pierwsze obwody scalone Texas Instruments
Procesor Pentium
Wielkość chipu prawie bez zmian
kilka tranzystorów
Kilka milionów tranzystorów
MINIATURYZACJA!!!
Jack Jack Jack Jack KilbyKilbyKilbyKilby
½Nagrody Nobla z Fizyki w 2000 r.
Cena tranzystora 10-7$
Liczba sprzedanych tranzystorów w r. 2002 -- 1018
Przeciętna cena tranzystora
Technologia CMOS osiąga skalę nano
Nanotechnologia krzemowa
Obecnie produkowany procesor Intel’a w technologii 90 nm
Influenza virus
Dłudość bramki ~ 2000 x mniejsza od średnicy ludzkiego włosa Grubość tlenku po bramką ~1.2 nm(5 atomowych warstw Si)
Nanotechnologia
ma już obecnie
duże znaczenie w dziedzinie
elektroniki, gdzie utrzymuje się trend
w kierunku
miniaturyzacji
1965 - 1970, Integrated Circuits based on
bipolar transistors
Since 1980, Integrated Circuits based on
CMOS technology dominate,
Field Effect Transistor (FET)
Development of Integrated Circuits Technology
CMOS Technology
CMOS = Complementary Metal Oxide Semiconductor
Metal Oxide Semiconductor p-Silicon Depletion
V
G
= 0
x
Ec EvE
Fp-Si
M O
S
Q
Mx
SiO2 (Semiconductor = Silicon)CMOS Technology – Biased Gate
Metal Oxide Semiconductor p-Silicon Depletion
V
G> 0
x
Ec EvE
Fp-Si
Q
Mx
Semiconductor p-Silicon Depletion Metal Oxide InversionV
G>> 0
Ec EvE
Fp-Si
Q
M 2DEGQ
n S Gate DV
G≈
0
S Gate DV
G>> 0
Current does not flow Current flows
State “0” State “1”
CMOS Technology
Logical circuits (processors) Memories
cutoff mode active mode
Operation of n-MOS FET in enhancement mode
(V
G>
V
th)
GS Gate Source Drain VDS V ID ID VG1 VDS VG2 VG3 VG4 gm Current vs. voltage gm
δ
IDδ
VGS V DS =δ
VDSδ
ID gd VGS = Small-signal transconductance Channel conductanceTransit time (frequency)
Basic characteristics of a
Field Effect Transistor (FET)
Current characteristic
CMOS Technology
n-MOS
&
p-MOS
electrons in the channel Complementary holes in the channel n+ n+ p+ p+ n well p-type body
Complementary MOS technology employs MOS transistors of both polarities CMOS devices are more difficult to fabricate than NMOS, but many more powerful circuits are possible with CMOS configuration
n-MOS transistor p-MOS
transistor
I
DV
DSCurrent characteristics of
p-MOS
&
n-MOS
Enhancement mode n-MOS transistors have a positive threshold voltage
V
th> 0
The operational behavior of a p-MOS enhancement transistor
represents the complement of the operation of the n-MOS. Compared to the n-MOS transistor, basically all voltage quantities are reversed.
CMOS circuits - Inverter
The complementary electrical character characteristics of n-MOS and p-MOS transistors make them well suited for the design of logical gates used in many digital designs.
G
G
S
S
D
D
V
DDV
INV
OUTn-MOS p-MOS
VIN= 0 p-MOS = ON n- MOS = OFF VOUT= VDD
VIN= VDD p-MOS = OFF n- MOS = ON VOUT= 0
IN 1 OUT = IN
No current flow in steady state low power dissipation in CMOS circuits
CMOS Technology –
“Computational state” and material system
Abakus (~3000BC)
Processor based on CMOS
S Gate D
Current
The material system: CMOS transistors
The computational state: position of the beads
The material system: A set of stings in a frame
The computational state: electron current flow
S Gate D
CMOS Technology – Important Parameter
Important parameter – gate length
L
G
determines stage of integrationswitching speed
Nowadays it is necessary to remove ~1000 electrons from the channel
L
G
2D-Electron Gas
Electron mobility Bulk GaAs at T=300 K – 9 000 cm2V-1s-1 2DEG at T=300 K in HEMT – 10 000-12 000 cm2V-1s-1Electron mobility in 2DEG at low temperatures ( < 1 K) can reach 20 000 000 cm2V-1s-1 !!! EF z M AlGaAs GaAs 2DEG E c EF z AlGaAs GaAs 2DEG Ec Gated AlGaAs/GaAs heterostructure AlGaAS/GaAs heterostructure
Classical in-plane transport
Ballistic, Coherent Quantum Transport
Sour ce Gate Drai n GaAs - substrate AlGaAs 2DEG
Mesoscopic devices – macroscopic with quantum effects present
Semiclassical Transport
The motion of an electron can be described
semi-classically by its position (r) and momentum (η
η
η
ηk) as
accurately as necessary, without violating the uncertainty
principle
provided
:applied electric and/or magnetic fields vary slowly
over the dimension of the electron wave packet:
typical length >> ∆r =1/∆k >> a (lattice constant)
the phase information of the electron wavefunction
is not needed
Semiclassical Transport
1
( )
1
E
=
∇
=
+
×
d r
d t
d
q
d t
c
ε
kk
k
v
H
ℏ
ℏ
The semiclassical electron motion, in the absence of collisions, is described by the following equations:
Band structure of semiconductors
Silicon GaAs
Brillouin Zone of zinc-blende semiconductors
Band structure of semiconductors
simplified models
Parabolic bands Spherical * 2 ) ( 2 2 m k k Ep η = Ellipsoidal * 2 2 * 2 2 * 2 2 2 2 2 ) ( z z y y x x p m k m k m k E k =η +η +η Warped bands(
)
[
,θ,ϕ]
2[
1(
θ,ϕ)
]
g ak k Ek = µ Γ Γ Γ Γ L X1 X2The distribution function
3 ( , 2 (2 ) , ) dN f t d dπ
= k r Ω Ωk rrepresents the number of particles in the phase space volume dr dkaround point (r, k).
For a homogeneous solid in equilibrium, f(r,k,t)is equal to
1 0
( )
{exp[( ( )
) /
] 1}
−=
−
B+
f k
ε
k
µ
k T
Fermi-Dirac distribution function. nThe central quantity in the semiclassical transport is the
distribution function f (r,k,t),
which gives the occupation probability of a “state” characterized by band index n, k-vector k, and space vector r.
The distribution function
( ) Ω Ωr k d d t ( ) Ω Ωr + k d d t dt A group of electrons moves rigidly through the phase space
under the influence of external fields.
Individual electrons are scattered intoor outof the group.
The total change in the distribution function f due to the electron scattering (“lattice collisions”)
To calculate the distribution function
in a given external fields, we examine its behavior
with time
( , , )
f r k t
The Boltzmann Equation
( , , ) ∂ ( , , ) = ∂ coll df r k t f r k t dt t 1 ( ) k r= ∇εεεε k ɺ ℏ 1 q k E v H c = + × ɺ ℏ
The evolution of the distribution function is given by the Boltzmann equation
In the steady state the local differential quotient is zero and we are left with the usual form of the Boltzmann eq.
The distribution function can be calculated for known external fields and known
collision processes
( , , )
( , , )
( , , )
r k collf r k t
r
f r k t
f r k t
t
∂
⋅∇
+ ∇
=
∂
ɺ
The distribution function
Any quantity can be obtained from the knowledge of the distribution function:
( )
( )
carrier density
current density:
mean energy :
:
( , )
( , , )
( , )
( ) ( , , )
( , )
( , )
( ) ( , , )
n t
f
t d
t
f
t d
t
n t
E
f
t d
r
r k
k
J r
v k
r k
k
r
r
k
r k
k
=
=
=
∫
∫
∫
2
2
2
2
1
3 3π
π
E
The Boltzmann Equation - Scattering
Term
• Quantum mechanical probability (per unit time) of scattering between electronic states
k
andk’
( , ') P k k
• Number of particles (electrons) that has been scattered during time
δt
from volumed
3k
intod
3k’
1
3
[ ( , )
( , )]
3
(2 )
( , , )
′
Π
′
− Π
′
∫
∂
=
∂
colld k
k k
k k
f r k t
t
π
1
3
3
( )[1
( )] ( , )
3
(2 )
′
′
′
−
f k
f k
P k k dx kd k t
δ
π
( , ') :
k k
f k
( )[1
f k
( ')] ( , ')
P k k
Π
=
−
Scattering mechanisms
CARRIER-CARRIER Screened potential Plasmons DEFECTS Impurities Crystal defects Ionized or Neutral PHONONS Intravalley Intervalley DeformationElectrostatic Optical Acoustic Polar Piezoelectric Intervalley scattering Intravalley scatteringFermi Golden rule
Scattering rate = transition probability for unit time
Microscopic Theory of Scattering
(
)
(
)
[
E
c
E
c
]
c
H
c
c
c
P
(k,
;
k'
'
)
=
2
π
k,
Ik'
'
2δ
k'
,
'
−
k,
η
Interaction matrix element
HI= interaction Hamiltonian
k,k’ = carrier wave vector
c,c’ = crystal related coordinates (such as ion displacements)
Energy conservation
E(k,c) = energy of the system before scattering E(k’,c’) = energy of the system
after scattering
Intra-valley and Inter-valley scattering
Intra-valley Point Phonons for intra-valley scattering
, ,
X L
Γ
Γ
Γ
Γ
Γ
Γ
Γ
Γ
X
L
LA LA+TA LA+TA+LO+TO Inter-valleyTransition Phonons for inter-valleyscattering
L
X
X
Γ
Γ
Γ
Γ
L Γ Γ Γ Γ →→→→ L→→→→X ∆ ∆ ∆ ∆ ∆ ∆ ∆→→→→∆ X→→→→X LO + LA LO + LA (Opposite valley) (Non--opposite Valley) LO LA + TO LA LO (MV >>>>MIII) (MV <<<<MIII)Phonon
Phonon
scattering:
scattering:
Deformation
Deformation
potential
potential
Optical phonons
Intravalley, spherical parabolic bands ( ) q op op op n E m E P ω ρ ω πη µη + ± ∆ = 2 1 2 1 2 3 2 3 ( ) v q op if op f op g n E E m E P −∆ ± + ∆ = ω ρ ω πη 2 µη 1 2 1 2 3 2 3
Intervalley, spherical parabolic bands
gv= # equivalent valleys
Detailed knowledge of the band structure is very important !!
GaAs
GaAs
Linearization of the Boltzmann
Equation
• Splitting the distribution function into
• its equilibrium value f0 and a perturbation df (f1),
0 0 1 0 0 (1 0) 0 f f f f f φ f β f fφ ε ∂ = + = − = + − ∂ 0 0 0 (1 ) f f f f δ φ β φ ε ∂ ≡ − = − ∂ 3 0 3
[ ( )
( )]
( , )
(2 )
∂
′
′
′
= −
−
Π
∂
∫
f
d k
k
k
k k
t coll
β
φ
φ
π
• one can see that quadratic terms in df appear in the integrand of the collision term. These terms can be omitted for small perturbations, the collision term thereby being linearized.
The linearized collision term takes a particularly simple form
The Relaxation Time Approximation:
Drude’s Model of Carrier Transport
0 1 ( ) ( ) ( ) f k = f k + f k 1 ( , , ) coll f f r k t t ττττ ∂ = − ∂ 0 1 1 ( ) k f qE f f ττττ − ⋅∇ + = −
The distribution function:
Furthermore, assuming a constant electric field E and a spatially uniform charge electron distribution, the Boltzmann transport equation becomes
=
J
q nE
µ
The relaxation time approximation:
* q m τ µ= Mobility:
Ohm’s law from the Boltzmann transport equation
Another way of looking at this is to consider f as a Taylor series for f0
Displaced distribution 3 3 3 3 2 1 0 ( ) ( ) ( ) ( ) * * * ( ) ( ) k k d k f k d k f k q n m m J qn qn E m d kf k d kf k =
∫
=∫
=∫
∫
1( ) k 0( ) f k =q Eττττ ⋅∇f k 0 0 ( ) ( ) k ( ) ( ) f k = f k +q Eττττ ⋅∇f k + = f k+q Eττττ …Methods to solve the Boltzmann Equation
a) The Variational Methodc)
The Monte Carlo Method
The variational principle for the linearized form of the Boltzmann equation
b) Relaxation time approximation
Drift-diffusion equations
Monte Carlo procedure
The Boltzmann equation is solved by a stochasticevolution (r,k,n) (r’,k’,n) (r’,k’’,n’) Scattering time Initial
state Free flight State after
the scattering
∆t = 1 fs
F
Generation of random numbers with given
distribution: The direct technique
To sample x in (a,b) distributed according to the normalized probability distribution f(x) we do the following:
1) we generate a number
r
uniformly distributed in (0,1)2) then, xris chosen according the relation
( )
=
∫
( )
=
r x rf
x
dx
x
F
r
0 F (x ) 0 1 r xr dF dx a b but …1) we need to solve the integral 2) we need to invert the relation
r=F(x)
f(
x
)
a b x
Generation of random numbers with given
distribution: The rejection technique
To sample x in (a,b) distributed according to the normalized probability distribution f(x) we do the following:
1) we choose a constant Γ≥f(x)
2) we generate two uniformly distributed random numbers in (0,1), r, r’
3) we define x1=a+(b-a)r,y1=Γr’
4) if y1≤f(x1) then x1is retained as a choice of x, otherwise x1 is rejected and
steps 2)-4) are repeated
f( x ) a b x (x1,y1) (x2,y2) accepted reject
When f(x) is strongly peaked, many numbers of pair (x,y) might be generated before a successful trial, resulting in a large
expense of CPU time
Free-Flight (τ)
During the free flight carriers do not suffer any scattering.
r and k coordinates change according semiclassical theory
k
r
F
k
∂
∂
=
−
=
E
dt
d
e
dt
d
η
1
;
Determination of the flight duration
The probability that the electron will suffer its next scattering during the time dt around t is given by
( )
[
]
⋅
∫
[ ( )]=
− t dt t Pe
dt
t
P
dt
t
0 ' ')
(
kk
P
Probability of flight from 0 to t without scattering Probability to suffer a scattering in (t,t+dt)For simulation purposes, we need to generate t according to the P(t) distribution …too difficult !!
Free flight: Self-scattering
Γ=1/τ0real scattering
self-scattering
In order to simplify the generation of the free flight, we define a new fictitious scattering (N+1 th) such that
( ) ( ) 1 const. 0 1 1 = = Γ = =
∑
+ = τ N i i P Pk kFor a constant scattering rate, the probability that the electron will suffer its next scattering during the time dt around t is given by
0 ) ( 0 τ
τ
t e dt dt t − ⋅ = PNow we can easily generate the free-flightτ distributed according P(t) via the direct technique
( )
r
ln
0τ
τ
=
−
with r uniformly distributed in (0,1) WhereΓ is the P(k) max for the energy interval considered in the simulationChoice of the scattering mechanism
impurity acustic def. polar optical intervalley self-scattering 0 Γ
After the free flight the carrier has a given energy E(k)
We look-up at the scattering table for this energy and by extracting a uniformly distributed numberrin (0,1) we chose the scattering mechanism:
Real scattering
we update the carrier state (E,k,band,valley) according to the scattering mechanism
Self-scattering
the carrier state remains unchanged (fictitious scattering)
rΓ
r
polar optical scattering
Synchronous Ensemble Monte Carlo
As soon as electron-electron interactions (also Poisson) or
explicitly time dependent phenomena occur,
synchronization of ensemble should be accounted
time 1 2 to+∆t to+2∆t to+3∆t to+4∆t to τ …
We define a simulation time step ∆∆∆∆t, after which
distribution function, potential etc. are updated
In the Ensemble Monte Carlo simulation many carriers
are simulated simultaneously
MC simulation of bulk GaAs
Γ L
X
At high electric field, electrons are scattered by phonons into satellite valleys
Hot Carriers
At high electric field, the kinetic energy of carriers is much larger than the thermal value
MC Simulation of Bulk GaAs
:
Velocity-Field Characteristics
For electric fields > 3-4 kV/cm, inter-valley transitions induce negative differential mobility.
High-field drift velocity: Theory
Electric field [kV/cm] 1 10 100 1000 6 10 7 10 D ri ft v e lo c it y [ c m /s ]
InN
Saturation velocity is determined by dominant energy loss mechanism -> polar optical phonon energy
⇒ ⇒ ⇒
⇒ High transit time frequencies
MC simulation: PM-HEMT (I
)
In0.25Ga0.75As/Al0.23Ga0.77As:GaAs HEMT2D simulation Electrons (ΓΓ,X,L) ΓΓ Holes (HH,LH,SO) Multigrid for Poisson Eq.
Electrons get hot at the Gate-Drain region
MC simulation: PM-HEMT (II)
In0.25Ga0.75As/Al0.23Ga0.77As:GaAs HEMT Potential profile - gate region
VD=2V High electric field at the
MC simulation: PM-HEMT (III)
Channel shrinks towards the end of the gate region Transfer of electrons to high energy valley
Transfer of electrons into surrounding barrier material
How far can we push Si CMOS?
Is it possible to keep exponential
growth?
S. E. Thompson & S. Parthasarathy, materialstoday, June 2006
Si technology industry time line:
possible time frame for new device types
What lies What lies What lies What lies beyond? beyond? beyond? beyond? Ge, III-V channel materials Carbon nanotubes Quantum Electronic Device Single Electron Transistor QED SET spintronics
Nanotechnology Eras
Ultimate electronics
Gate leakage current
S Gate D
Challenges – Gate Leakage
Gate OxideThinner gate oxides produce faster transistors The limit of Gate Oxide (SiO2) has been reached
30 nm transistor has 0.8 nm gate oxide layer
Thinner oxides leak more
Gate oxide can get so thin that it no longer acts as a good insulator
High-K Gate insulator
n+ n+
High-K Gate Dielectric
New material replaces SiO2
Thicker physical film but the same capacitance 10 000 lower gate leakage for the same capacitance Alternative gate dielectrics to reduce gate leakage TiO2, HfO2, ZrO2, Ta2O5
New Structures and Materials for
Nanoscale MOSFETs
Non-planar MOSFETs
Enhanced Hole mobility for Uniaxial Strained-Si
Strain in Si is introduced by Si1-xGexin the source/drain.
Higher hole mobility
Why Germanium MOS Transistors?
Higher (low field)
Higher (low field)
Higher (low field)
Higher (low field) mobilities
mobilities
mobilities
mobilities
Heteroepitaxial Growth of Ge on Si
With H With H With H
With H2222anneal, dislocations are confined to the anneal, dislocations are confined to the anneal, dislocations are confined to the anneal, dislocations are confined to the Si/
Si/ Si/
Si/GeGeGe interface leaving defect free top Geinterface leaving defect free top interface leaving defect free top interface leaving defect free top GeGeGeGe layers.layers.layers.layers.
New Materials for Si- Technology
Moore’s Law increasingly relies on material innovations
Epitaxial Growth of Vertical Nanowires
Controlled Growth and Structures of
Molecular-Scale Silicon Nanowires
(a) TEM images of 3.8-nm SiNWs
grown along the <110> direction (c) cross-sectional image (b) & (d) models based on Wulff construction
High Performance Silicon Nanowire
Field Effect Transistors
Yi Cui, et al. NANO LETTERS 3, 149 (2003)
Comparison of SiNW FET transport parameters with those for
state-of-the-art planar MOSFETs show that
“SiNWs have the potential to exceed substantially conventional devices, and thus could be ideal building blocks for future nanoelectronics.”
Nanocable device concepts
Gargini-type nanotransistorhigh k gate dielectric, e.g. ZrO2
nanorod,
e.g Si/Ge or an alloy
metal gate electrode possible MOS transistor structure
P. Gargini, (Director - Technology Strategy Intel Fellow) ‘Enlightenment beyond Classical CMOS’ ISS US 2002.
Si NW FETs: Experimental Status: ‘Top Down’
Ge Nanowires Synthesized
by Low Temperature CVD
Ge Nanowire FET with High K Gate Dielectric
Key Challenge: Controlled growth Key Challenge: Controlled growthKey Challenge: Controlled growth Key Challenge: Controlled growth
Technologia półp rzewodników Supramol ekularna c hemia CZAS SKALA
10 nm?
2015 ?
Elektronika oparta o nano-struktury
półprzewodnikowe i duże molekuły
“Top down”
“Bottom up”
?
?
Czy uda się utrzymać wykładniczy
charakter wzrostu?
Nanorurki węglowe podstawą przyszłych
technologii informacyjnych ?
Delft, Tans, et al.,
Nature393 (1998)
Harvard, Javel et al.,
Nano Letters393 (1998)
Parametry tranzystorów lepsze od 2D FETs stosowanych w obecnych procesorach opartych o technologię CMOS
Quantum Coherent
Transport Theory
Source Drain Gate
V
V
Gµ
1µ
2 ChannelQuantum transport – the study of current flow
on an atomistic scale
Physics of open systemsConductance from Transmission
( , ) ( , ) F L F F f E µ =f E E +V fF( ,E µR)=fF( ,E EF) Device Lead Lead Contact Contact
µ
µ
µ
µ
L= EF+ VL
R
µ
µ
µ
µ
R= EFµL- electrochemical potential = chemical potential + voltage V – external bias
EF– Fermi energy (chemical potential in the absence of bias)
The Landauer approach – very useful in describing mesoscopic transport The current through a conductor (device) is expressed in terms of probability that an electron can transmit through it
Leads are reservoirs of electrons in which energy- and momentum relaxation processes are so effective that the electron system remains in equilibrium even under a given applied voltage bias
1 ( , ) exp[ ( )] 1 F f E E µ β µ ≡ − +
The electron concentration in the leads is so high that the electrostatic potential in each lead is taken to be constant (as for the case of metal)
Electric current through the device
Conductance from Transmission
||, , k n m || || || || , 0 2 ( ) ( ( , , ) ) L L n m k c e I v T E f E k n m L > µ − =
∑∑
− || || || || , 0 2 ( ) ( ( , , ) ) R R n m k c e I v T E f E k n m L < µ − =∑∑
− || || || || || , 0 2 ( )[ ( ( , , ) ) ( ( , , ) )] L R R L n m k c e I I I v T E f E k n m f E k n m L > µ µ = + =∑ ∑
− − −The number of electrons in state
||
2 ( ( , , )f E k n m −µL) /Lc
per unit length The total contribution to the electric current from the electrons entering from the left
Similarly, for the electrons entering from the right
The total electric current through the device
L
E V
Fµ
= +
R
E
Fµ
=
Quantized Resistance of a ballistic
waveguide - Experiment
Measured conductance vs gate voltage