Dopant Diffusion

(1) Predeposition dopant gas

dose control SiO2 SiO2

Doped Si region
Si

(2) Drive-in Turn off dopant gas

or seal surface with oxide

profile control SiO SiO2

SiO2
(junction depth; 2

concentration) Si

Note: Predeposition by diffusion can also be

replaced by a shallow implantation step.
1
 Dopant Diffusion Sources

(a) Gas Source: AsH3, PH3, B2H6

BN Si BN Si
(b) Solid Source

B oxide +SiO2

(c) Spin-on-glass SiO2+dopant oxide

2
(d) Liquid Source.

3
 Solid Solubility of Common Impurities in Si

oC

C0 (cm-3)

4
 Diffusion Coefficients of Impurities in Si

−EA
D = DO e kT

10-6
10-12 Cu

Au
B,P
10-13

As
10-14
5
 Temperature Dependence of D

−EA
D = D 0e kT Arrhenius Relationship

6
 Mathematics of Diffusion

C(x)

x
Fick’s First Law:
∂ C (x, t )
J (x , t ) = − D 
∂x
D : diffusion constant
2
[ D ] = cm
sec
7
 From the Continuity Equation

∂C (x,t)
+∇⋅J (x,t)=0
∂t
∂C(x,t) ∂J (x,t) ∂⎛ ∂C(x,t) ⎞
⇒ =− = ⎜D ⎟
∂t ∂x ∂x ⎝ ∂x ⎠
∂C(x,t) ∂⎛ ∂C(x,t) ⎞
⇒ = ⎜D ⎟
∂t ∂x ⎝ ∂x ⎠
“Diffusion Equation”
8
 Concentration independence of D

If D is independent of C
(i.e., D is independent of x).

∂C(x, t ) ∂ C(x, t ) 2
= D
∂t ∂x 2

Concentration Independent Diffusion Equation

State of the art devices use fairly high concentrations,

causing variable diffusivity and other significant side-
effects (transient-enhanced diffusion, for example.)
9
 A. Predeposition Diffusion Profile

•Boundary Conditions:
C(x = 0, t) = C0 = solid solubility of thedopant
C(x = , t ) = 0 Justification:
Si wafers are ~500um thick, doping
depths of interest are typically < several um

•Initial Condition:
C(x, t = 0) = 0 At time =0, there is no diffused dopant
in substrate

10
Diffusion under constant surface concentration
 2 2 xDt − y 2 
C (x, t ) = C 0  1 −  e dy 
 π 0 
 x 
= C  erfc
0  
 2 Dt 
2 Dt = Characteristic distance for diffusion.
C 0  Surface Concentration (solid solubility limit)
C0

t3>t2
t1 t2>t1
x
x=0
11
 Properties of Error Function erf(z)
and Complementary Error Function erfc(z)
z
2  -y2
erf (z) =  e dy erfc (z) ≡1 - erf (z)
π 0
erf (0) = 0 erf( ) = 1 erf(- ) = - 1
2 1 e-z2
erf (z)  z for z <<1 erfc (z)  for z >>1
  z
d erf(z) d erfc(z) 2 -z2
= - = e
dz dz 
d2 erf(z) 4 -z2
= - z e
dz2 
z 1 
 erfc(y)dy = z erfc(z) + (1-e-z2 )  erfc(z)dz =
1
0  0 

12
 Practical Approximations of erf and erfc
The value of erf(z) can be found in mathematical tables, as build-in functions in calculators and
spread sheets. If you have a programmable calculator, this approximation is accurate to 1 part in
2
107: erf(z) = 1 - (a1T + a2T2 +a 3T 3 +a 4T4 +a 5T 5) e-z
1
where T = 1+P z and P = 0.3275911
a1 = 0.254829592 a2 = -0.284496736 a3 = 1.421413741 a4 = -1.453152027 a5 = 1.061405429
1

10-1

10-2 exp(-z^2)
10-3 erfc(z)
10-4

10-5

10-6
10-7
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6
13
[1] Predeposition dose

Q (t ) =  C (x, t )dx


C 0  2 Dt
=  t
π

[2] Concentration gradient

∂C Co − x2
=− e 4 Dt
∂x π Dt
14
 B. Drive-in Profile

•Boundary Conditions :
C (x =  , t ) = 0
∂C Physical meaning of C/x =0: C(x)
=0 No diffusion flux in/out of the Si
∂x x=0
surface. Therefore, dopant dose is
conserved

•Initial Conditions : x
x=0
 x 
C (x , t = 0 ) = Co  erfc  
 2 (Dt ) 
Predep’s (Dt)

15
 C (x,t = 0 )  Q  δ(x )
Solution of Drive-in Profile
with Shallow Predeposition Approximation:
C(x,t=0)
C0  2 (Dt ) predep Approximate predep profile
Q= as a delta function at x=0
π
x

C(x,t) −x2
Q 4(Dt )drive−in
t1 C(x, t ) = e
t2
π(Dt )drive−in
x

16
 How good is the (x) approximation ?
For reference only
Dt
Let R  predep

Dt drive − in

C(x)/C0 Approximation over-estimates conc. here

Exact solution
R=1 Delta function
Approximation

Approximation
R=0.25
Good under-estimates
agreement concentration here.
x
17
 Summary of Predeposition + Drive-in

D1 = Diffusivity at Predeposition temperature

t1 = Predeposition time

D2 = Diffusivity at Drive-in temperature

t2 = Drive-in time
1

  D 1 t1 
− x2
 2C 0 2
C (x ) =    e 4 D 2t2

 π  D 2t 2 
*This will be the overall diffusion profile after a “shallow” predeposition
diffusion step, followed by a drive-in diffusion step.
18
Semilog Plots of normalized Concentration versus depth

Predeposition Drive-in
Drive-
in

19
Diffusion of Gaussian Implantation Profile

Note:  is the implantation dose

20
 Diffusion of Gaussian Implantation Profile (arbitrary Rp)
C
T h e e x a c t s o lu tio n s w ith = 0 a t x = 0 (.i.e . n o d o p a n t lo s s th ro u g h s u r fa c e )
x
c a n b e c o n s tru c te d b y a d d in g a n o th e r fu ll g a u s s ia n p la c e d a t -R p [M e th o d o f
Im a g e s ].
(x - R p)2 (x + R p)2
  [e
- 2 - 2
C (x , t) = 2 ( R p + 2 D t) + e 2 ( R p + 2 D t) ]
2
2 ( R p + 2 D t) 1 /2

W e c a n s e e th a t in th e lim it (D t) 1 /2 > > R p a n d  R p ,

 e - x /4 D t
2

C (x ,t) → (th e h a lf -g a u s s ia n d r iv e -in s o lu tio n )

(  D t) 1 /2

For reference only

21
 The Thermal Budget
Dopants will redistribute when subjected to various thermal cycles of
IC processing steps. If the diffusion constants at each step are
independent of dopant concentration, the diffusion equation can be
written as:
C 2C
t = D(t) x2
t
Let  (t)   D(t’)dt’
0

 D(t) =
t
C C  
Using =
t  • t
C      C or C = 2C
2
The diffusion equation becomes: • =
 t t • x2  x2

22
When we compare that to a standard diffusion equation with D being
C 2 C
time-independent:  (Dt) = 2, we can see that replacing the (Dt)
x
product in the standard solution by  will also satisfy the time-
dependent D diffusion equation.

Example
Consider a series of high-temperature processing cycles at
{temperature T1, time duration t1} ,{ temperature T2, time duration t2
}, etc. The corresponding diffusion constants will be D1, D2,... . Then,
 = D1t1+D2t2+..... = (Dt)effective

** The sum of Dt products is sometimes referred to as the “thermal

budget” of the process. For small dimension IC devices, dopant
redistribution has to be minimized and we need low thermal budget
processes.

23
 Thermal Budget
(Dt)effective = ∑(Dt) i
step i

Temp (t) Example

Dttotal of :

Well drive-in
well
drive-in S/D and
step Anneal
step
time S/D annealing

For a complete process flow, only those steps with high Dt

values are important

24
 Irvin’s Curves

p-type erfc
n-type erfc
p-type half-gaussian
n-type half-gaussian

Explicit relationship between:

N (surface concentration) ,
o
x (junction depth),
j
N (background concentration) ,
B
R (sheet resistance),
S

Once any three parameters are known, the fourth one can be determined.
25
 Motivation to generate the Irvin’s Curves
Both NB(4-point-probe), RS (4-point probe) and xj (junction staining) can be
conveniently measured experimentally but not No (requires secondary ion mass
spectrometry). However, these four parameters are related.

Approach
1)The dopant profile (erfc or half-gaussian ) can be uniquely determined if one
knows the concentration values at two depth positions.

2)We will use the concentration values No at x=0 and NB at x=xj to determine the
profile C(x). (i.e., we can determine the Dt value)

3)Once the profile C(x) is known, the sheet resistance RS can be integrated
numerically from:
1
Rs =
q  μ(x )C (x )− N B dx
xj
0
4) Irvin’s Curves are plots of No versus ( Rs• xj ) for various NB.
26
Illustrating the relationship of No, NB, xj, and RS

27
 2-Dimensional Diffusion with constant D
Example 1 Drive-in from line source with s atoms/cm
2C DC C
D 2 + r r = t Diffusion Equation in cylindrical coordinates
r
s 2
 C(r, t) = e-r /4Dt
2Dt

28
 Two-Dimensional Drive-in Profile (cont.)
Example 2 : Semi-Infinite Plane Source
Diffusion mask yj (at x=0) < xj (at y → )
y

yj equal-conc. contours

xj
x
Q -x2/4Dt y
C(x,y,t) = e [ 1 + erf ( )]
2 Dt 2 Dt
where Q = predep dose in #/cm2

Rule of Thumb : yj ~0.7-0.8 xj

29
 Electric-field Enhancement

Example 1 : Diffusing specie generates build-in E-field

Na(x) Acceptor Boron p(x) =Na-(x)

Na(x) = Na-(x) hole gradient

Hole diffusion
tendency
E build-in
x
Complete acceptor ionization at diffusion At thermal equilibrium, hole current =0
temperature. Acceptor atoms are Hole gradient creates build-in
negatively charged. Mobile + holes are electric field to counteract the hole
near vincinity of the acceptor atoms. diffusion tendency

30
1
 Mobile holes tends to move
p+ away due to hole
concentration gradient

Ebuild-in due to the hole gradient

B-
B- acceptors
experience
an additional
drift force

Enhanced Diffusion for B- acceptor atoms

due to the Boron profile graduent

31
2
 Calculation of build-in Electric Field E

dp
J p = q  μ p  p  E − qDp  =0
dx
at thermalequilibrium.
kT 1 dp
E=  
q p dx
kT dln p
= 
q dx

32
3
 Calculation of total acceptor flux F
Na-(x)

Total acceptor atom flux = (diffusion + drift) components

F = - D A dNA/dx - A  E  NA

33
4
 Since A = DA/[kT/q] from Einstein’s Relationship

 dN A + d ln p  
F = −D A    NA
 dx dx 
−
From (1) p = n + N A charge Neutrality
(2) pn = ni 2 Law of Mass Action
NA +
N A + 4ni 2 2
(1) & (2) p =
2
dln p dN A 1
 = 
dx dx N A + 4ni 2
2

34
5
  
 
dN A  1 
 F = −DA   1 + 
dx 2
 1+
4 ni 
 2

 –– – – N A 
h − factor

If n i  A h→1 low dopant conc.

N ni A h→ 2 high dopant conc.
 N
Value of h depends on concentration NA

35
6
 Intrinsic Carrier Concentration ni

If n or p >
ni(Tdiffusion), we will
expect to observe high
concentration diffusion
effects.
Temp range for
dopant diffusion
36
7
 Example 2: Diffusion enhanced by build-in E-Field of
other diffusion species
Illustration: Uniform substrate concentration profile is
perturbed by E-field of another diffusion specie
As
diffusion
caused by As
conc gradient

Uniform B
conc in
substrate

B-

Professor N Cheung, U.C. Berk eley 37

8
 Dopant Diffusion Revisited

•Diffusion Mechanisms

•Dopant interactions with Point Defects of Si substrate

- Dopant Diffusivity affected by Dopant Concentration

- Dopant Diffusivity affected by Point Defect Concentration
- Point Defect Concentration affected by Dopant Concentration
- Point Defect Concentration affected by Processing Steps

9
 For reference only

Si Native
Point Si vacancy Si interstitial
Defects

1) Thermal-equilibrium values of Si
neutral interstitials and vacancies at
diffusion temperatures
<< doping concentration of interest At 1000oC, CIo* ~ 1012 /cm3
(1015 –1020 /cm3) CVo* ~ 1013 /cm3

2) Diffusivity of Si interstitials and

Si vacancies >>
diffusivity of dopants

10
 Diffusion Mechanisms in Si
(A) No Si Native Point Defect Required
Example: Cu, Fe, Li, H

Fast Diffusion
(a) Interstitial Diffusion

Cu

10-6 cm2/sec
Au

11
 Diffusion Mechanisms in Si
(B) Si Native Point Defects Required (Si vacancy and Si interstitials)

Example: Dopants in Si ( e.g. B, P,As,Sb)

(a) Substitutional Diffusion (b) Interstitialcy Diffusion

12
(B) Si Native Point Defects Required (Si vacancy and Si interstitials)
continued

(c) Kick-Out Diffusion

(d) Frank Turnbull Diffusion Slow Diffusion

B,P

10-12 cm2/sec
As
13
 Diffusivity Comparison:
Dopants, Si interstitial, and interstitial diffusers
For reference only

108 times higher

14
Diffusion Enhanced by Charged Point Defects
(Vacancies and Interstitials )

Vo = neutral
V+ = + vacancy
V- = - vacancy
Si vacancy V= = - - vacancy

…..

15
Possible Point Defect Configurations

For reference only

16
How processing steps affect point defect concentrations

•Neutral interstitial and vacancy point defects present at

thermal equilibrium
At 1000oC, CIo* ~ 1012 /cm3
CVo* ~ 1013 /cm3
• Charged Point Defects enhanced by heavy doping; total point
defect concentrations enhanced by ~10x
I+, Io, I-
V+, Vo, V-, V=
•Point defects Injected by interfaces during oxidation
(total point concentrations enhanced by ~10x)

•Implantation collisions (total point defect concentration

enhanced by ~ 1000X)

17
 Derivation of charged vacancy concentration

1) Negatively charged vacancies

Vo + r • electrons  V -r (r = 1,2,3…)
Law of Mass Action [Vo] n r [Vo] n i r
Kc = =
[V -r ] [V i -r ]
For the reaction :
a•A + b • B  c • C [V -r ] n
same Kc for undoped Si
r
=[ n ]
[V i -r ] i
where n = electron concentration ,
[A] a [B]b [ ] denotes the concentrations ,
= Kc subscript i denotes condition for intrinsic Si
[C] c
2) Positively charged vacancies
[ ] denotes Vo + r • holes  V +r (r =1,2,3…)
concentration [V +r] p ni
= [ n ] = [ n] r
r
Kc = equilibrium [V i +r ] i

Constant at Temp
p =hole concentration
T
See Diffusion Handout for Law of Mass Action Discussion
18
 Law of Mass Action Examples
1) Electron Hole recombination
e+h→0 [e]• [h] = K(T) or n•p =ni2

2) Generation of singly charged negative vacancies

e+Vo→V- [e]• [V o] / [V-] = K1(T) or [V-]  n

3) Generation of doubly charged negative vacancies

2e + V o → V = [e]2 • [V o] / [V=] = K2(T) or [V=]  n2

4) Generation of singly charged positive vacancies

h+Vo→V+ [h]• [V o] / [V+] = K3(T) or [V+]  p  1/n

19
 Charged Vacancy Enhanced Diffusion

 p −  n =  n
2

Dtotal = D + D   + D i    + D i    +...
o +
i
 ni   ni   ni 
Why each diffusivity written with ni as normalization factor ?
Consider the +charge vacancy first , D+  [V+].
From Law of Mass Action: [V+]  p
Therefore , D+  p and can be written as D+ = K• p
where K is a proportionality constant
Let us multiply K by ni and call it Di+, then D+ = Di+• (p / ni )
The physical meaning of Di+ is diffusivity D+ for undoped Si
because p =n= ni for undoped silicon
Similarly, Di- and Di= represent diffusivity values for undoped Si
due to –charge and = charge vacancies
20
 Diffusion Profile with enhanced D at high concentration

Log C(x)
High conc. profile:
D gets larger Log C(x)
when C(x)
is large

J large
Low conc profile:
Erfc or gaussian J small
x x

* C(x) goes deeper but the shape looks “flatter”at high conc. regions

21
Example : High Concentration Arsenic
diffusion profile becomes “box-like”

22
 Summary of High-Concentration Diffusion

First, Check if doping concentration is > ni(T) or < ni(T)

If doping conc < ni:

D= Do
Use constant diffusivity solutions
(diffusion profile is erfc or half-gaussian)

If doping conc > ni:

Use D = Dv = h [ Do + D+ + D- + D= + ...]
(Solution requires numerical techniques)

23
 Oxidation Enhanced Diffusion (OED)

Diffusivity is enhanced for B and P underneath a growing oxide

Example:
Channel-stop implant diffuses faster during field oxidation

Field Oxide

SiO2 Injected Si[I] interstitials

Si enhance dopant diffusion
Si[I]
DOED = D + D
D  [dXox / dt]n where 0.4 < n < 0.6
Note: The dopant Sb shows retarded diffusivity during oxidation
=> it diffuses primarily via the vacancy mechanism
24
 Transient Enhanced Diffusion (TED)
Dopant Implantation

C(x)
no annealing
Si Implantation
induced excess 900oC, several minutes
substrate point defects (After excess point defects
recombine. normal diffusion)
Implantation creates
large number of excess Si
interstitials and vacancies.
(1000X than thermal process). x
After seconds of annealing,
the excess point defects will 900oC, several sec (TED)
be reduced substantially by Extremely rapid diffusion due
recombination . to excess point defects
25
 Junction Depth versus post-implantation annealing time

Junction
Depth xj
Normal diffusion
(with high concentration effects)
TED

Annealing
time
seconds
••Difficult
Difficult to make ultra-shallow (< 100nm) junctions
with implantation and annealing
26
 Summary of Diffusion Module

•Diffusion Sources (gas, solid, liquid)

•Diffusion Constant D
•Concentration Profiles with D independent of Concentration
•The Thermal Budget
•Junction Depth and Sheet Resistance
•Irvin’s Curves
•Two-Dimensional diffusion profiles – qualitative
•Enhanced Diffusion due to electric field
•Point Defects in Si
•Diffusion due to point defects :
-Charged point defects,
-Oxidation induced point defects ( OED/ORD),
-Implantation induced point defects (TED)

27