Anda di halaman 1dari 22

University of California,

Berkeley
EE230 - Solid State
Electronics Prof. J. Bokor
Transport Theory
(read Lundstrom 3.1 - 3.4)
Aim Develop a general approach for relating microscopic description of carrier motion to mac-
roscopic description.
Drift-Diffusion equation:
lassical approach to device design
material properties: ! D "
#oundar$
conditions
microph$sics is %uried device
in here structure
Device %ehavior descri%ed %$ independent specification of material and structure.
&odern devices
! D: can't %ur$ microph$sics (i.e. %allistic transport)
also depend e(plicitl$ on device structure (i.e. nonlocal transport)
arrier energ$ ) momentum distri%utions var$ strongl$ in devices
arrier scattering: *e$ to transport.
+arado(: no scattering --, net -./0
1
*
consider simple %and E = E
o
- E
o
cos(ka)
1 E
22 . --------
!
.
E
o
a
-------- sin( ka)

#olt3mann equation
4andout.fm
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor
periodic motion - #loch oscillation. 5his has %een e(perimentall$ o%served in superlattices
67oherent su%millimeter-8ave emission from #loch oscillations in a semiconductor super-
lattice!9 . :asch*e! et al.! +h$s. ;ev. Lett. </! 331= - 3322 (1==3). +D> cop$ is posted on
the class 8e% site.?
1lectrons oscillate through #@ 6#ragg diffraction? locali3ed in real space no net cur-
rent0
Acattering damps oscillation resulting in a net motion in an applied field. Bet the motion is
damped %$ scattering. 5he message is that transport depends on the balance between the
applied driving force and dissipation by scattering forces.
#olt3mann transport equation
Aemi-classical approach. Cn semiconductors! all the D& is %uried in mE. 1lectron motion
8ill %e descri%ed %$ classical mechanics! e(cept that scattering pro%a%ilities 8ill also %e
derived using quantum mechanics.
omplete description: could solve Fe8ton equations for each particle (electron):
d"
i
( , , )
------- . #E " F r " t
dt
random forces due to
phonons! impurities
for all i . 1, , $ . 5his is clearl$ not feasi%le.
Cnstead! 8e ta*e a statistical approach:
define distri%ution function (single particle):
f(r, ", t): pro%a%ilit$ of finding a particle at r ! 8ith momentum " at time t.
7phase space9: G dimensional space - %, y, &, "
%
, "
y
, "
&
5he particle densit$ is found as:
Aince momentum is discrete! 8e sho8 a summation over allo8ed values of p. 4o8ever! it is
- =3
-
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor
more convenient to convert to an appropriate integral.
1

( , )
-- f r " '
"
12
3

-- -----
2 (

1

'
vol----------
------of
-
-
-
one
state in
p-
spac
e
1
3

)3


f( r, "
4

(since f is
8ell locali3ed
inside the 1st
#@)
65his general method of converting
from a sum in p-space to an integral
is important! and 8ill come up again
and again.?
1(ample: &# distri%ution:
f
"
energ$
f(
.
e(p
E
E

k*
H
2
+
to get
the
carrier
densit$!
integrate
over
moment
um
space
----------
----
,
4

3 )
3

1st
#@
.
1
H

E
)
k*]

e
-------------- e(p[ (
E
,
4

)3

for reasona%le 5! onl$ populated


near %ottom of %and! so 8e can
e(tend the integral over all p-
space
- =4 -
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor

e
H"
2


2+

k*
"
2
d"d
p-space
5hus! the integral is:
Fo8 use:

%
2
1

% e
.
d
%
---
---
-
-
4

/
>inall$:
n
.
1
2+
k*
3

2
-
-
-------------
--- e(p[ ( E , H Ec) k*]
4

2
. $
c
e(p [ ( E
,
H E
c
)
k*]
8hich is our previous result.
Average *inetic energ$ densit$:
-( r ) .
1

"
2
H

E
H

"

2+

3
"
---------
-----
---------- e(p{ [
E
,
c
] k*}d
4

)
3
2+

1

H"
2

2+k*
.
4e(p[
( E
H
E
)
k*]

d
"
"
e
---------
-----
,
I

)
3
/
3
1

2
J

2
-
-
( 2 + k*)
I
3 1
2
+k
*
3

2
.
-
-
-----------
----- e(p[ ( E
H
E ) k*] -- k*
,
2 4
)
2

- =J
-
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor
- 3
Ao 8e can sa$ that the energ$ K electron . ---- . --k* . Again! a familiar result.
n 2
4o8 a%out current densit$L
J( r )
#

"
E
H
E
"
2

k
*
3
" . --------------
------
e(p
,
H
---------- d
)
3


4
"

+ 2+
. /
thermal equili%rium! f is even

J
.
/
Mer$ common and useful guess for f in nonequili%rium conditions is the 7drifted9 &a(-
8ellian:
" H
"
d 2
f( r, ", t) . e(p , ( r , t) H E
c
H
-------------
------

k*
2
+

quasi->ermi level
7drift9 velocit$
f
"
d&
"
&
- =G
-
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor
onsider heteroNunction %ipolar transistor (4#5) %and diagram
E
c
%allistic
pea*
scattered
electrons
emitter %ase
collector
electron distri%ution function at
c
ol
le
ct
o
r
f
2 components
t h er m al i 3 e d
component
7%allistic
9 pea*
"
&
O . 2+ E
c
4
o
8
c
o
u
l
d
8
e
calculate such a
distri%utionL
7onservation of
pro%a%ilit$9 --, continuit$
eqn for f.
#alance forces
and scattering

"
&
-
=
<
-
University of California, Berkeley
EE230 - Solid State Electronics
Prof. J.
Bokor
df( r , " , t) f
---------------------- . ---
dt t . H /
total derivative: driving terms:
all changes in f generation! recom%ination! scattering
5o see ho8 this is used! e(pand total derivative using the chain rule:
df
.
d
r

d" f
----
-
-
-
-
f "
----- "f " ----
dt
d
t r dt t
d"
Pse , .
---
--
.
dt
5his is *no8n as the #olt3mann transport equation (#51). ompare this to the carrier con-
tinuit$ eqn:
n 1

J .

n
----- " --
-
-
-
-
-
t #

t
.
H
/
:e can re8rite the #51 to highlight the analog$ to carrier continuit$:
f

f
--- " r ( v f ) " " ( ,f ) .
-
-
-
t

t
.
H
/
5his sho8s ho8 the #51 is a continuit$ equation for f. 5he vf term is analogous to an intu-
itivel$ appealing pro%a%ilit$ current in real space. 5he less intuitive element here is the ,f
term! 8hich represents the pro%a%ilit$ current in momentum space.
f
Fo8 let's e(amine the term: ---
t
. H /
2 processes contri%ute.
1. Actual generation-recom%ination processes such as photogeneration! defect
recom%ination! stimulated emission! ...
e(plicitl$ 8rite s(r! p! t) for these
2. arrier scattering . 5his ta*es carriers from one momentum state to another -
creates sources-sin*s in momentum space.
- =I -
University of California, Berkeley
EE230 - Solid State Electronics
Prof. J.
Bokor
f
for these processes.
:rite ---
t
coll
>or transport pro%lems! 8e 8ill %e mainl$ interested in collisions.
ollision integral
onsider the effect of collisions on the distri%ution at a particular point in phase
space: f0", r, t). 58o t$pes of collision processes affect f at this point:
(1) carriers at " scatter into "
(2) carriers at " scatter out to "
5he collision rate for process (1) is given %$:
f( " )[ 1 H f( " )]S( " , ")
pro%. of finding
pro%.
there
a carrier at "
is no
carrier
at "
Aimilarl$! the rate for process (2) is given %$:
transition rate for the
collision " "
(>ermi's golden rule)
5he total rate of change at "
due to collisions is then the
sum over all possi%le
processes scattering
scattering particles into "
minus the sum of all
processes scattering particles
out of ":
f( r , " , t)
----------------------
t
S( ", ")
could %e a
sum of
several
processes.
Qi
ven
S
"
"
the
#5
1
is
an
7integro-differential9 eqn
for f 0r, ", t). (i.e. a mess0)
5he name of the game is
finding suita%le
appro(imations and
computational techniques.
;ela(ation time appro(imation
(;5A)
Aimplest appro(imation for
the collision term:
- == -
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor
8here f
o
represents the equili%rium distri%ution.
5his simpl$ sa$s that equili%rium is restored in time . :e 8ill use this appro(imation
heavil$.
5he solution then is simpl$:
f( t ) . f
/
" [ f ( t=0) H f
/
]e
Ht


onductivit$ in the ;5A
Assume n-t$pe semiconductor! applied field E .
>urther assume s$stem is spatiall$ uniform so that
r
f . / .
f
Atead$ state implies --- . / .
t
5hen
;e8rite:
f . f
/
H #E "f
>or lo8 field! assume small change in the momentum gradient. 5hat is: "f "f
/
. 5hen:
f . f
o
H #E "f
o (1)
5o improve on the first order appro(imation! 8e could iterate. 5hat is! put this f in "f .
5his 8ould give a term 1( E
2
) in the solution for f. :e 8ill neglect this for no8.
5a*e E along &
R
.
5his loo*s li*e the first 2 terms of a 5a$lor series. 5hen! to 1st order: (assuming constant)
- 1//
-
University of California, Berkeley
EE230 - Solid State Electronics Prof. J. Bokor
8hich is the drifted &a(8ellian0
:e can derive another useful form. Let:

"
2
( k*)
EC H E, " ----------
2+

f
o
.
e
H
then

"
f

f
o
"
o
. ------

Hf
o
" v
------------- . -----
+k
* k*
1quation (1) %ecomes:
f .
f
#
E v
o
"
-----f
k*
f
o
"
f (2)
8ith f S f
o
.
Fo8 calculate drift velocit$

v
&
( f
o
" f)
v . ------------------------------
"
(3)
&

( f
o
" f)
"
Cf is constant (independent of " or 1)! then %$ inspection of drifted &a(8ellian:
urrent densit$ is given %$:
- 1/1
-
University of California, Berkeley
EE230 - Solid State Electronics
.
n#
2

---------
--E
&
+
. n#E
&
5his is the familiar result 8ith
Prof. J. Bokor
8
C
v
2
v
1T
f
o
ins
ert
(2)
#$ spherical s$mmetr$!
v
ing
thes
e
rela
tion
s!
8e
find
:
.
University of California, Berkeley
EE230 - Solid State Electronics
Cf 8e 8rite: v
&
. E
&
! then this defines more generall$:

#
. ---------------- .
+
5his special ensem%le average is defined
%$:
Uften ( E ) is given %$ a po8er
la8:
( E ) .
/
[ E ( k*)]
s
5hen:
1T ( "
2
2+)( "
2

2+k*)
s
e
H"
2


2+

k*

2 .

/
---------------------------------------------------------------
----------------------------
"

1T ( "
2

2+)e
H"
2


2+

k*

( "
2
2+k*)
s
e
H"
2


2+

k*
"
4
d"
. ------------------------------------------
--------------------------------/
/
e
H"
2

2+k*
"
4
d"
/
let y "
2
2+k*
dy .
----------------
2"d"
.
--------------------
2y
1
2
2+k
*
( + k*)
1


2
5hen

1
s H
-
-
.
-
1
/
3

-
University of California,
Berkeley
EE230 - Solid State Electronics

s

"
3
--

y
2
e
Hy
dy
.

/
----------------------------

-
-
/
y
2
e
Hy
dy
Psing the
function:
(
n )
( " "
1)
>inall$:
- 1/4
-

Anda mungkin juga menyukai