6.

012 - Electronic Devices and Circuits

Lecture 3 - Solving The Five Equations - Outline
• Announcements
Handouts - 1. Lecture; 2. Photoconductivity; 3. Solving the 5 eqs.
See website for Items 2 and 3.
• Review
5 unknowns: n(x,t), p(x,t), Je(x,t), Jh(x,t), E(x,t)
5 equations: Gauss's law (1), Currents (2), Continuity (2)
What isn't covered: Thermoelectric effects; Peltier cooling
• Special cases we can solve (approximately) by hand
Carrier concentrations in uniformly doped material (Lect. 1)
Uniform electric field in uniform material (drift) (Lect. 1)
Low level uniform optical injection (LLI, τmin) (Lect. 2)
Photoconductivity (Lect. 2)
Doping profile problems (depletion approximation) (Lects. 3,4)
Non-uniform injection (QNR diffusion/flow) (Lect. 5)

• Doping profile problems

Electrostatic potential
Poisson's equation
Clif Fonstad, 9/17/09 Lecture 3 - Slide 1
 Non-uniform doping/excitation: Summary
What we have so far:
Five things we care about (i.e. want to know):
Hole and electron concentrations: p(x,t) and n(x,t)
Hole and electron currents: J hx (x,t) and J ex (x,t)
Electric field: E x (x,t)
And, amazingly, we already have five equations relating them:

= G # R $ Gext (x,t) # [ n(x,t) p(x,t) # n i2 ] r(t)

"p(x,t) 1 "J h (x,t)
Hole continuity: +
"t ! q "x

= G # R $ Gext (x,t) # [ n(x,t) p(x,t) # n i2 ] r(t)

Electron continuity:
"n(x,t) 1 "J e (x,t)
#
"t q "x
"p(x,t)
Hole current density: J h (x,t) = qµ h p(x,t)E(x,t) # qDh
"x
"n(x,t)
Electron current density: J e (x,t) = qµ e n(x,t)E(x,t) + qDe
"x
" [&(x)E x (x,t)]
Charge conservation: %(x,t) = $ q[ p(x,t) # n(x,t) + N d (x) # N a (x)]
"x

So...we're all set, right? No, and yes.....

Clif Fonstad, 9/17/09 We'll see today that it isn't easy to get a general solution, but we can prevail. Lecture 3 - Slide 2

!
 Thermoelectric effects* - the Seebeck and Peltier effects
(current fluxes caused by temperature gradients, and visa versa)

Hole current density, isothermal conditions:

Drift Diffusion
$ d [ q# ] ' $ dp '
J h = µ h p &" ) + qDh & " )
% dx ( % dx (
Hole potential Concentration
energy gradient gradient

Hole current density, non-isothermal conditions:

!
Drift Diffusion Seebeck Effect
$ d [q# ] ' $ dp ' $ dT '
J h = µ h p &" ) + q Dh & " ) + q Sh p&" )
% dx ( % dx ( % dx (
Temperature
gradient

Seebeck Effect: temperature gradient → current Generator

!Peltier Effect: current → temperature gradient Cooler/heater

* A cultural item; we will only consider isothermal

Clif Fonstad, 9/17/09 situations on 6.012 exams and problem sets. Lecture 3 - Slide 3
 Thermoelectric effects - the Seebeck and Peltier effects
(current fluxes caused by temperature gradients, and visa versa)

Two examples:
Right - The hot point probe, an
apparatus for determining the
carrier type of semiconductor
samples.
Below - A thermoelectric array
like those in thermoelectric
generators and solid-state
refrigerators.

Clif Fonstad, 9/17/09 Ref.: Appendix B in the course text. Lecture 3 - Slide 4
 Thermoelectric Generators and Coolers - modern examples
Electrical power for a trip to Pluto
Cooling/heating for the
necessities of life

Image of thermoelectric wine cooler

removed due to copyright restrictions.

Source: NASA.

"…electrical power for the New Horizons spacecraft

and science instruments [is] provided by a single
radioisotope thermoelectric generator, or RTG."
Thermoelectric Wine Cooler
28 bottles
12˚C - 18˚C
Quiet, gas free, vibration free,
Launched environmentally friendly, LED
1/19/2006 display, interior light.
Zhongshan Candor Electric Appl. Co.

http://www.alibaba.com/
Clif Fonstad, 9/17/09 Lecture 3 - Slide 5
http://pluto.jhuapl.edu/
Source: NASA/Johns Hopkins University Applied Physics
Laboratory/Southwest Research Institute.
 Non-uniform doping/excitation: Back to work
(laying the groundwork to model diodes and transistors)
What we have:
Five things we care about (i.e. want to know):
Hole and electron concentrations: p(x,t) and n(x,t)
Hole and electron currents: J hx (x,t) and J ex (x,t)
Electric field: E x (x,t)
And, five equations relating them:

[ ]r(t)
"p(x,t) 1 "J h (x,t)
Hole continuity: + = G # R $ G (x,t) # n(x,t) p(x,t) # n 2

"t !q "x
ext i

= G # R $ Gext (x,t) # [ n(x,t) p(x,t) # n i2 ] r(t)

Electron continuity:
"n(x,t) 1 "J e (x,t)
#
"t q "x
"p(x,t)
Hole current density: J h (x,t) = qµ h p(x,t)E(x,t) # qDh
"x
"n(x,t)
Electron current density: J e (x,t) = qµ e n(x,t)E(x,t) + qDe
"x
" [&(x)E x (x,t)]
Charge conservation: %(x,t) = $ q[ p(x,t) # n(x,t) + N d (x) # N a (x)]
"x
We can get approximate analytical solutions in 5 important cases!
Clif Fonstad, 9/17/09 Lecture 3 - Slide 6

!
 Solving the five equations: special cases we can handle
1. Uniform doping, thermal equilibrium (nopo product, no, po):
" "
= 0, = 0, gL (x,t) = 0, J e = J h = 0 Lecture 1
"x "t
2. Uniform doping and E-field (drift conduction, Ohms law):
" "
= 0, = 0, gL (x,t) = 0, E x constant Lecture 1

! "x "t
3. Uniform doping and uniform low level optical injection (τmin):
"
= 0, gL (t), n' << po Lecture 2
! "x
3'. Uniform doping, optical injection, and E-field (photoconductivity):
"
= 0, E x constant, gL (t) Lecture 2
! "x
4. Non-uniform doping in thermal equilibrium (junctions, interfaces)
"
= 0, gL (x,t) = 0, J e = J h = 0 Lectures 3,4
! "t
5. Uniform doping, non-uniform LL injection (QNR diffusion)
"N d "N a "n' "
= = 0, n' # p', n' << p o , J e # qDe , # 0 Lecture 5
"x "x "x "t
Clif Fonstad, 9/17/09! Lecture 3 - Slide 7
 Non-uniform material with non-uniform excitations
(laying the groundwork to model diodes and transistors)
Where cases 2, 4, and 5 appear in important semiconductor devices

Junction diodes, LEDs:

p-type n-type

Flow problem
Case 5 - Flow
Flow
Case 5 problem
- Flow
Junctiion
Case 4 - problem
Junction

Bipolar transistors: B Junction

Case 4 problem
- Junctions

E n-type p n-type C

Flow problems
Case 5 - Flow
G
MOS transistors: S D

n+ n+
p-type
Depletion
Case approximation
4 - Interface
Diodes
Case 4 - Junctions Drift
Case 2 - Drift
(In subthrehsold: Case 5 - Flow)
Clif Fonstad, 9/17/09 Lecture 3 - Slide 8
 Case 4: Non-uniform doping in thermal equilibrium
Doping Profiles and p-n Junctions in TE: Na (x), Nd (x)
Any time the doping varies with position, we can no longer assume that
there is charge neutrality everywhere and that ρ(x) = 0. The dopants
are fixed, but the carriers are mobile and diffuse:
no(x),
Nd-Na Nd-Na Can't say: n o (x) = N D (x)
ρ(x) < 0
no(x)
ρ(x) > 0 !
E(x)

x
Electron diffusion
Electron drift In T.E., the
net fluxes
Hole drift
must be zero
Hole diffusion
Clif Fonstad, 9/17/09 Lecture 3 - Slide 9
 Non-uniform doping in thermal equilibrium, cont.
To treat non-uniformly doped materials we begin by looking at
them in thermal equilibrium, as we've said.
This is useful because in thermal equilibrium we must have:
gL (x,t) = 0
n(x,t) = n o (x)
p(x,t) = po (x)
J e (x,t) = 0
J h (x,t) = 0
Consequently, the 2 continuity equations in our 5 equations
reduce to 0 = 0, e.g.:
"n(x,t) 1 "J e!
(x,t)
# = gL (x,t) # [ n(x,t) $ p(x,t) # n o (x) $ po (x)] r(T)
"t q "x
0 0 0 0
Clif Fonstad, 9/17/09 Lecture 3 - Slide 10

!
 Non-uniform doping in thermal equilibrium, cont.
The third and fourth equations, the current equations,
give:
dn o (x) d# De 1 dn o (x)
0 = qµ e n o (x)E(x) + qDe " =
dx dx µ e n o (x) dx
dpo (x) d$ Dh 1 dpo (x)
0 = qµ h po (x)E(x) " qDh # ="
dx dx µ h po (x) dx
!
And Poisson’s equation becomes:

dE(x) d 2# (x) q
! = " 2
= [ po (x) " n o (x) + N d (x) " N a (x)]
dx dx $

In the end, we have three equations in our three remaining

unknowns, no(x), po(x), and φ(x), so all is right with the
! world.

Clif Fonstad, 9/17/09 Lecture 3 - Slide 11

 Non-uniform doping in thermal equilibrium, cont.
Looking initially at the first of our new set of equations,
we note that both sides can be easily integrated with
respect to position :
x d" De x 1 dn o (x)
# xo dx dx = # dx
µ e o n o (x) dx
x

De D n (x)
" (x) $ " (x o ) = [ln n o (x) $ ln n o (x o )] = e ln o
µe µ e n o (x o )
Next, raising both sides to the e power yields:
µe
[ " (x )#" (x o )]
! n o (x) = n o (x o )e De

We chose intrinsic material as our zero reference for the

electrostatic potential:
! " (x) = 0 where n o (x) = n i
µe
and arrive at : " (x )
n o (x) = n ie De

Clif Fonstad, 9/17/09 ! Lecture 3 - Slide 12

 Non-uniform doping in thermal equilibrium, cont.
From the corresponding equation for holes we also find :
µh
" # (x )
po (x) = n ie Dh

µh µe q Incredibly
Next use the Einstein relation: = = Multilingually
Dh De kT rhyming

!
Note: this relationship rhymes as written, as well as when inverted, and also
either way in Spanish. It is a very fundamental, and important, relationship!
Note : @ R.T. q kT " 40 V #1 and kT q " 25 mV
!
Using the Einstein relation we have:

! n o (x) = n ie q" (x ) kT and po (x) = n ie#q" (x ) kT

Finally, putting these in Poisson’s equation, a single
equation for φ(x) in a doped semiconductor in TE
materializes:
!
d 2" (x)
dx 2
= #
q
$
n i( [
e #q" (x )/ kT
# e q" (x )/ kT
) + N d (x) # N a (x) ]
Clif Fonstad, 9/17/09 Lecture 3 - Slide 13
 Non-uniform doping in thermal equilibrium, cont.
(an aside)
What do these equations say?
n o (x) = n ie q" (x ) kT and po (x) = n ie#q" (x ) kT

To see, consider what they tell us about the ratio of the hole
concentration at x2, where the electrostatic potential is φ2, and
that
! at x1, φ1: "q [ # (x 2 )"# (x1 )] / kT
po (x 2 ) = po (x1 )e
The thermal energy is kT, and the change in potential energy of a
hole moved from x1 to x2 is q(φ2 - φ1), so have:
! po (x 2 ) = po (x1 )e
"#PE x1 $x2 / kT

If the potential energy is higher at x2, than at x1, then the population
is lower at x2 by a factor e-ΔPE/kT.
That is, the population is lower at the top of a potential hill.
!
If the potential energy is lower, then the population is higher.
That is, the population is, conversely, higher at the bottom of a potential hill.

Clif Fonstad, 9/17/09 Lecture 3 - Slide 14

 Non-uniform doping in thermal equilibrium, cont.
(continuing the aside)
The factor e-ΔPE/kT is called a Boltzman factor. It is a factor
relating the population densities of particles in many
situations, such as gas molecules in an ideal gas under the
influence of gravity (i.e, the air above the surface of the earth)
and conduction electrons and holes in a semiconductor.*
The potential energy difference for holes is qΔφ, while that for
electrons is -qΔφ. Thus when we look at the electron and hole
populations at a point where the electrostatic potential is φ,
relative to those where the potential is zero (and both
populations are ni) we have:
n o (x) = n ie q" (x ) kT and po (x) = n ie#q" (x ) kT
We will return to this picture of populations on either side of a
potential hill when we examine at the minority carrier
populations on either side of a biased p-n junction.
!
* Until the doping levels are very high, in which case the
Boltzman factor must be replaced by a Fermi factor.**
Clif Fonstad, 9/17/09 ** Don’t worry about it. Lecture 3 - Slide 15
 Doing the numbers:
I. D to µ conversions, and visa versa
To convert between D and µ it is convenient to say kT/q ≈ 25 mV,
in which case q/kT ≈ 40 V-1:
17˚C/62˚F
Example 1: µe = 1600 cm2/V-s, µh = 600 cm2/V-s
De = µe (q kT ) = 1600 /40 = 40 cm 2 /s
Dh = µh (q kT ) = 600 /40 = 15 cm 2 /s

II. Relating φ to n and p, and visa versa

To calculate φ knowing n or p it is better to say that kT/q ≈ 26 mV,
! because then (kT/q)ln10 ≈ 60 mV: 28˚C/83˚F
Example 1: n-type, ND = Nd - Na = 1016 cm-3
kT 1016 kT kT
"n = ln 10 = ln10 6 = ln10 # log10 6 $ 0.06 ln10 6 = 0.36 eV
q 10 q q
Example 2: p-type, NA = Na - Nd = 1017 cm-3
kT 1017 kT
" p = # ln 10 = # ln10 $ log10 7 % #0.06 $ 7 = #0.42 eV
! q 10 q
Example 3: 60 mV rule:
For every order of magnitude the doping is above (below) ni,
! the potential increases (decreases) by 60 meV.
Clif Fonstad, 9/17/09 Lecture 3 - Slide 16
 More numbers no[cm-3] po[cm-3] φ [V]

1019 − 101 − 0.54 −

1018 − 102 − 0.48 −
1017 − 103 − 0.42 −
Typical range
1016 − 104 − 0.36 −
n-type
1015 − 105 − 0.30 −
1014 − 106 − 0.24 −
1013 − 107 − 0.18 −
1012 − 108 − 0.12 −
1011 − 109 − 0.06 −
Intrinsic 1010 − 1010 − 0.00 −
109 − 1011 − -0.06 −
108 − 1012 − -0.12 −
107 − 1013 − -0.18 −
106 − 1014 − -0.24 −

p-type 105 − 1015 − -0.30 −

104 − 1016 − -0.36 −
Typical range
103 − 1017 − -0.42 −
102 − 1018 − -0.48 −
101 − 1019 − -0.54 −
Clif Fonstad, 9/17/09 Lecture 3 - Slide 17
 Non-uniform doping in thermal equilibrium,cont:
We have reduced our problem to solving one equation for
one unknown, in this case φ(x):

d 2" (x)
dx 2
= #
q
$
n i( e[#q" (x )/ kT
# e q" (x )/ kT
) + N d (x) # N a (x) ]
Once we find φ(x) we can find no and po from:

n o (x) = n ie q" (x ) kT and po (x) = n ie#q" (x ) kT

!
Solving Poisson’s equation for φ(x) is in general non-trivial,
and for precise answers a "Poisson Solver" program must
! be employed. However, in two special cases we can find
very useful, insightful approximate analytical solutions:
Case I: Abrupt changes from p- to n-type (i.e., junctions)
also: surfaces (Si to air or other insulator)
interfaces (Si to metal, Si to insulator, or Si to insulator to metal)
Case II: Slowly varying doping profiles.

Clif Fonstad, 9/17/09 Lecture 3 - Slide 18

 Non-uniform doping in thermal equilibrium, cont.:
Case I: Abrupt p-n junctions
Consider the profile below:
Nd-Na
NDn

x
- NAp
p-type n-type

po = N Ap , n o = n i2 N Ap n o = N Dn , po = n i2 N Dn
? kT
ln( N Dn /n i ) # " n
ln( N Ap /n i ) $ " p
kT "=
"= # n o (x) = ?
q q
po (x) = ?
Clif Fonstad, 9/17/09
" (x) = ? Lecture 3 - Slide 19

! !
 Abrupt p-n junctions, cont:
First look why there is a dipole layer in the vicinity of the
junction, and a "built-in" electric field.
no , po
NDn
NDn
NAp
NAp
ni2/NAp ni2/NDn
x
Hole diffusion Electron diffusion

ρ(x) Drift balances

qNDn diffusion in the
steady state.
+Q 0
0
x
-Q
-qNAp
Hole drift Electron drift
Clif Fonstad, 9/17/09 Lecture 3 - Slide 20
 Abrupt p-n junctions, cont:
If the charge density is no longer zero there must be an
electric field: εEx(x) = ∫ρ(x)dx
Ex
0 0
x

Epk

and an electrostatic potential step: φ(x) = -∫Ex(x)dx

φ(x)
φn

x
φp

Clif Fonstad, 9/17/09

Ok, but how do we find φ(x)? Lecture 3 - Slide 21
 Abrupt p-n Junctions: the general strategy
We have to solve an non-linear, second order differential
equation for φ:
d 2" (x)
dx 2
= #
$ (x)
%
= #
q
%
[
n i (e#q" (x )/ kT # e q" (x )/ kT ) + N d (x) # N a (x)]
$(x)
Or, alternatively " (x) = # && dx + Ax 2 + Bx
%
In the case of an abrupt p-n junction we have a pretty good
idea of what ρ(x) must look like, and we know the details
will be lost anyway after integrating twice, so we can try
!
the following iteration strategy:
Guess a starting ρ(x).
Integrated once to get E(x), and again to get φ(x).
Use φ(x) to find po(x), no(x), and, ultimately, a new ρ(x).
Compare the new ρ(x) to the starting ρ(x).
- If it is not close enough, use the new ρ(x) to iterate again.
- If it is close enough, quit.
Clif Fonstad, 9/17/09 Lecture 3 - Slide 22
 To figure out a good first guess for ρ(x), look at how quickly
no and po must change by looking first at how φ changes:
φ(x)
60 mV φn
φn

700 to 900 mV
x
φp 60 mV
A 60 mV change in φ
φp decreases no and po
10x and ρ increases to
…and what it ρ(x) 90% of its final value.
means for ρ(x):
qNDn The change in ρ must
90% be much more abrupt!
0 +Q 0
x
90% -Q
-qNAp

The observation that ρ changes a lot, when φ changes

Clif Fonstad, 9/17/09 a little, is the key to the depletion approximation. Lecture 3 - Slide 23
 6.012 - Electronic Devices and Circuits
Lecture 3 - Solving The Five Equations - Summary
• Non-uniform excitation in non-uniform samples
The 5 unkowns: n(x,t), p(x,t), Je(x,t), Jh(x,t), E(x,t)
The 5 equations: coupled, non-linear differential equations
• Special cases we can solve (approximately)
Carrier concentrations: (Lect. 1)
Drift: Jdrift = Je,drift + Jh,drift = q (µe no + µh po) E = σo E (Lect. 2)
Low level optical injection: dn'/dt – n'/tmin ≈ gL(t) (Lect. 2)
Doping profile problems: junctions and interfaces
Non-uniform injection: QNR flow problems
• Using the hand solutions to model devices
pn Diodes: two flow problems and a depl. approx.
BJTs: three flow problems and two depl. approx.’s
MOSFETs: three depl. approx.’s and one drift
• Non-uniform doping in T.E.
Relating no, po, and electrostatic potential, φ
Poisson's equation: two situations important in devices
Clif Fonstad, 9/17/09 Lecture 3 - Slide 24
MIT OpenCourseWare
http://ocw.mit.edu

6.012 Microelectronic Devices and Circuits

Fall 2009

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.