Outline

6.12J / 3.155J Micr oelectr onic pr ocessing

Electro-chemical deposition
Why electro-chemical deposition?
How does it compare with sputtering, CVD, evaporation?
Excellent step coverage, esp. for small features
Reliability
Yield, Y ∝ e-AD
Failure Probability Density , Meant time to failure
“Infant mortality”, “steady state”, “wear out”

Electromigration
Manifestations of elecro-migration
Causes
Cures

Ap r i l 25 , 2005
 ELECTRO-CHEMICAL DEPOSITION
6.12J / 3.155J Micr oelectr onic pr ocessing

 We saw CVD Gas phase reactants: pg ≈ 1 mTorr to 1 atm.

Good step coverage, T > 350 K

 We saw sputtering Noble (+ reactive gas) p ≈ 10 mTorr; ionized particles

Industrial process, high rate, reasonable step coverage
Extensively used in electrical, optical, magnetic devices.

 We saw evaporation: Source material heated, peq.vap. = ~ 10-3 Torr, pg < 10-6 Torr
Generally no chemical reaction (except in “reactive deposition),
λ = 10’s of meters, Knudsen number NK >> 1
Poor step coverage, alloy fractionation: Δ pvapor
Historical (optical, electrical)

Ap r i l 25 , 2005
 ELECTRO-CHEMICAL DEPOSITION
See: Madou, Fundamentals of Microfabrication; Plummer, Sec. 9.3.10
6.12J / 3.155J Micr oelectr onic pr ocessing

Cathode Anode
Electro-chemical plating: -V +V
Reduction of metallic ions
from aqueous,
Cu+
inorganic salt electrolytes.
Mz+(solution) + ze- ⇒ M(lattice) e-
z electrons supplied
by external power supply

Electroless (autocatalytic) deposition process This is a

(no power supply) battery
reducing agent in the solution is the electron source
Ap r i l 25 , 2005
 Why ELECTRO DEPOSITION?
6.12J / 3.155J Micr oelectr onic pr ocessing

IC processing generally avoids wet processes

Increased use of Electro-depostion
driven by:
1) Cu replacing Al in metallization
Cu is difficult to remove by dry etch,

but is readily deposited electro-chemically

2) MEMS, need for high aspect ratio structures.

Ap r i l 25 , 2005
 How ELECTRO DEPOSITION?
6.12J / 3.155J Micr oelectr onic pr ocessing

Spin-coat
Expose
Photoresist Develop
Seed layer (15-30 nm) Au, Pt, Cu, Ni…
Adhesion layer (5-10 nm) Ti, Cr
(good oxides)
Dielectric substrate

Seed layer becomes cathode;

attracts M+ , coated with M layer

Ap r i l 25 , 2005
 ELECTRO DEPOSITION
6.12J / 3.155J Micr oelectr onic pr ocessing

Cathode Anode
-V +V CuSO4.5H2O

Cu+
At 0.377 V
e-

Film

Ap r i l 25 , 2005
 Aside: ELECTRO DEPOSITION vs. SPUTTERING
6.12J / 3.155J Micr oelectr onic pr ocessing

Cathode
is “target”,
Cathode Anode source material
-V +V P ≈ 10-100 m Torr
cathode anode
Ex
Cu+ - v Ar + v e!
⊕
V ≈ 1 kV
P ≈ 10-100 m Torr
e-

- ⊕
Film

Film

Ap r i l 25 , 2005
 Physical process
6.12J / 3.155J Micr oelectr onic pr ocessing

Cathode Anode dc c(!) " c(0)

-V +V =
dx #
c(") ! c(0)
Cu+ J = ! NqDF
#
e- Faraday Const
Diffusion const.
# electrons in reaction

1-D potential due to concentration diff Cu+ concentration

RT # c(0) & c(x)
! = ln % c∞ Cu+
NF $ c(") ('
At φc, all Cu+ at x = 0 are consumed
c(0) !
Jc is limited by
c"
J c = ! NqDF diffusion and x
# bulk concentration δ
Ap r i l 25 , 2005
 Physical process
6.12J / 3.155J Micr oelectr onic pr ocessing

c"
cCu+(x) J c = ! NqDF
c∞Cu+ #

c0 RT # c(0) &
!c = ln %
NF $ c(") ('
x
! = 2 Dt

) # NF"c & , J
J = J c +1 ! exp % ( . Activation
+* $ RT ' .-
!c current

!
Increased
stirring
Diffusion limited
Electro-deposition
Ap r i l 25 , 2005
 ELECTRO DEPOSITION vs. SPUTTERING
6.12J / 3.155J Micr oelectr onic pr ocessing

Cathode
is “target”,
Cathode Anode source material
-V +V P ≈ 10-100 m Torr
cathode anode
Ex
Cu+ - v Ar + v e!
⊕
V ≈ 1 kV
e-

+ VD.C. + VD.C.
Cu+ Ar+

e- e-

Plasma
High conductivity
Ap r i l 25 , 2005
 Feature size dependence
6.12J / 3.155J Micr oelectr onic pr ocessing

J
Activation This applies to
!c current large substrates: δ/L << 1
! L
δ
Diffusion limited
Electro-deposition
δ
But for small features: δ/L >> 1 L

Ap r i l 25 , 2005
 Small features
6.12J / 3.155J Micr oelectr onic pr ocessing

+V
++++
δ
But for small features: δ/L >> 1 - - - -
+ + + +
-V

L
Electro-deposition is faster,
+V
the smaller the feature
great for vias, step coverage… δ

-V

L
Ap r i l 25 , 2005
 Leveling agents
6.12J / 3.155J Micr oelectr onic pr ocessing

Small organic and inorganic molecules

control Cu interaction with substrate.
These molecules are known generically as
‘accelerators’, ‘levelers’ or ‘inhibitors’.

“Leveling agents” preferentially adhere

on high, convex surfaces,
(not in trenches),
inhibit high-point deposition,
level the film surface.

Ap r i l 25 , 2005
 Leveling agents, accelerators, inhibitors
6.12J / 3.155J Micr oelectr onic pr ocessing

Little is understood concerning how these molecules

affect growing Cu surface, nor with Cu ions in solution.

Polyethylene glycol (PEG)

works via synergy with Cu+ and Cl-.

Most common accelerator, bis(3-sulfopropyol) disulfide (SPS),

interacts with the surface,
increases electron transfer rate across the interface.

Ap r i l 25 , 2005
 Applications of electrochemical deposition
6.12J / 3.155J Micr oelectr onic pr ocessing

printed circuit boards,

magnetic alloys for computer memories,
coatings for hard disk drives,
wear resistant coatings,
corrosion resistant alloys,
metal matrix composites ,
electroreformed laser mirrors,
electrochromic materials,
decorative coatings,
oxides,
organic polymers.

Ap r i l 25 , 2005
 Outlook
6.12J / 3.155J Micr oelectr onic pr ocessing

As the feature size decreases,

present methodology of depositing Cu on Si will change.

Use of a seed layer, deposited by PVD or CVD

will likely be eliminated.
Cu will be directly deposited onto the diffusion barrier,
need development of new bath chemistries.

Possible: direct deposition of Cu onto oxide and nitrides

in various environments.

Ap r i l 25 , 2005
 Reliability of semiconductor I Cs
plus Electrodeposition
6.12J / 3.155J Micr oelectr onic pr ocessing

Read Campbell, p. 425 -428 and Ch. 20. Sec. 20.1, 20.2; Plummer, Sec. 11.5.6

1. IC reliability:
Yield = (# operating parts) / (total # produced)

Failure of devices occurs by various mechanisms:

 Particles on surface interrupt depositions, flaw devices
 Oxides, dielectrics fail by charging or dielectric breakdown,
 Metals fail by corrosion and
Electro-migration:

2. Electromigration:
 Manifestation Hillocks and voids
 Modeling Currents, Thermal gradients, electric fields

Ap r i l 25 , 2005
 Reliability of semiconductor I Cs
Why is this an issue?
6.12J / 3.155J Micr oelectr onic pr ocessing

Net yield is product: Y1 x Y2 x Y3… (e.g., a 10-step process each 95% =>60% yield)

Defect density, D, has decreased

“Learning curve”: yield vs. lot number with succeeding higher-density
dynamic random access memories …
and average over last 7 lots.

Ap r i l 25 , 2005
 Killer defects
Defect area density
6.12J / 3.155J Micr oelectr onic pr ocessing

Simplest yield model assumes independent, randomly-distributed defects,

(Poisson distribution):

A = chip area AD = probability

Yield of defect
Y ! e"AD D = defects/area overlapping
chip

AD Y = (1! G)e!AD(" )
Fraction of disk area
in which all circuits fail

Particle control: Class (Max #/ft3) > 0.5 µm

1 1
10 10
100 100
1000 1000

Ap r i l 25 , 2005
 Killer defects
Defect size
6.12J / 3.155J Micr oelectr onic pr ocessing

Defects are not randomly distributed spatially

(e.g. stress concentrations generate dislocations, stacking faults),
nor randomly distributed by size, δ, i.e. D = D(δ):
Empirical distribution of defect sizes: 1
All fail
!q
1! G
D(! ) = c , 0 < ! < !0
!0q +1

!0p"1
D(! ) = c p , !0 < ! < !max
!

Hard to measure,
Therefore Y ≈ (1-G) exp(-AD)
G is fractional area where all fail

Meander-line
process control
module

Ap r i l 25 , 2005
 Wafer size and reliability challenges
6.12J / 3.155J Micr oelectr onic pr ocessing

Wafers now 300 mm,

soon 450 mm.

Now 5 billion square in. of ckts/yr,

at $25/sq in. => $125 billion/yr

That’s 800 acres, 1 sq. mile

Ckts/wafer

Moore’s law
Ckts/area doubles every 18 mos.

More chips/wafer => more chips

per process cycle
Now 3 yrs
greater demands on deposition uniformity
Ap r i l 25 , 2005
 “Perfect storm” in microfabrication
6.12J / 3.155J Micr oelectr onic pr ocessing

Convergence of 3 simultaneous disruptive events (all on larger wafers):

1) Smaller line-width, conductors, capacitors…
Now λ = 248 nm, KrF excimer => 130 nm lines (Pentium 4)
λ = 193 nm, ArF excimer => 110 nm lines

Soon λ = 157 nm, Ar excimer => 70 nm lines

T < 100 C
2) New materials
Al conductors => Cu (TaN buffer) T < 1000 C
SiO2 dielectrics (3 GHz)
=> lower κ, C-doped SiO2 (5 GHz)
T < 500 C

3) New deposition processes

CVD, PVD now being joined/replaced by
“atomic layer deposition” (ALD)
Ap r i l 25 , 2005
 Reliability definitions
6.12J / 3.155J Micr oelectr onic pr ocessing

Cumulative failure distribution function, F (t): R (t )

1
F (t) = fraction of failures up to time, t. F ( t)
Survival or reliability distribution function, R (t):
R (t )= 1 - F (t ) 0
0 t

Failure probability density function, f (t): f ( t)

f (t) = dF/dt 0
(This is key to predicting failure rates) 0 t

"

Mean time to failure, MTTF: MTTF = # t ! f (t)dt

0

Median time to failure, t50: time after which half of devices have failed.

Ap r i l 25 , 2005
 Reliability definitions
6.12J / 3.155J Micr oelectr onic pr ocessing

1
λ(t)
Failure probability density/number remaining:
λ(t) = f(t)/R(t)
0
0 t
Failure rate during time δt, λ(t):
R(t) " R(t + #t) 1 dR(t) 1 dF(t)
! (t) = =" =
#tR(t) R(t) dt R(t) dt

Failure rate ! (t) = "

1 dR(t)
= const. = ! 0 (fractional failure frequency)
in steady state: R(t) dt
Steady-state survival
Hence: or reliability drops off
R(t) ! e" #0 t exponentially with time
steady state: "
1
f ss (t) =
dF
dt
=!
dR
dt
" #0e! #0 t MTTFss = # t ! f (t)dt = $0
0

Ap r i l 25 , 2005
 Different failure processes
6.12J / 3.155J Micr oelectr onic pr ocessing

Failure rate:

λ (t )

!(t) = !0

0
t
“Infant Steady
mortality” state Wearout

Ea
!
Different failure processes have
different thermally activated rates: r = r0 e kBT

Ap r i l 25 , 2005
 Mean time to failure
6.12J / 3.155J Micr oelectr onic pr ocessing

The mean time to failure (MTTF)

(related to inverse of rate): "n + Ea / kB T
MTTF ! J e
n ≈ 2 to 3.

(Most activated mechanisms of f ailur e have a f or m like this)

Expressed in log form as:
En
ln < t fail >= ln(A) ! nln(J) +
kBT

Plotted vs Log(J), right Log(MTTF)

in which case the slope gives the power, -n.

For increased operating current,

MTTF drops off as the -nth power of J.
For higher operating temperatures,
lifetime curve is shifted down, quicker failure.

log(J)
Or, ln(MTTF -1), mean failure rate, could be plotted vs. 1/T (Arrhenius plot)…

Ap r i l 25 , 2005
 Mitigating thermally activated failure
6.12J / 3.155J Micr oelectr onic pr ocessing

Thermally activated failure rates: Caution on accelerated aging:

Ea
!
r = r0 e kBT Accel. aging
test
Ln(r)
RT
Operate at lower temp.,
Operating
lower current density temperature

Use “burn-out” to elim. early fails

1/kBT

…you may get

wrong activation energy.

Example: electromigration. ..
Ap r i l 25 , 2005
 Electromigration: Manifestation
6.12J / 3.155J Micr oelectr onic pr ocessing

Hillock formation Whisker bridging

two interconnects

4 microns
! = "y E # 10 9 Pa # 2 $10%4 lbs /micron 2 Hillocks and voids
(Current flow )
(Pics: Aluminum films, Bell Labs)

Ap r i l 25 , 2005
 Electromigration: electron wind moves atoms
6.12J / 3.155J Micr oelectr onic pr ocessing

Electromigration: mode of failure in high-current-density heterostructures.

(Most literature on electromigration deals with metallic conductors in semiconductor devices)

J = nqv
Large current density, J => not only charge transport
but also mass transport of
charged particles, e’s or h’s.

When charge carriers collide with atoms (“electron wind”),

they impart a small momentum to atoms,
sweeping them in the direction of the carrier drift.

Ap r i l 25 , 2005
 Electromigration: electron wind moves atoms
6.12J / 3.155J Micr oelectr onic pr ocessing

Electromigration flux of species A, jA = cAvdrift,

requires the force on an ion A
J = nqv
due to the electric current:

F = qZA E = qZA J !
* *

(Z*q . nq v ) ρ

Ion - carrier
interaction

Here q is the electronic charge, ZA* is the effective ion valence

and E is the electric field (force per unit charge)
producing the electric current density, J = E/ρ
Yield stress
of Al is about
! = F / area " 100 GPa
70 GPa
Ap r i l 25 , 2005
 Electromigration: grain-boundary diffusion
6.12J / 3.155J Micr oelectr onic pr ocessing

Most electromigration takes place along grain boundaries.

DA = DA exp[!Ea /(k BT)]

o

DA is the grain boundary diffusion coefficient of A .

(EA typically ≈ 0.5 - 0.8 eV vs. bulk about 1.4 eV)

Flux of species A, JA, is proportional to the product

F = qZA J! ):
*
(volume concentration of A) x (velocity of A resulting from

DA F DA qZA* J!
J A = cA v A = cA = cA
RT RT

Here use Nernst-Einstein equation for drift velocity of a particle

at temperature T under influence of force F: v = DAF/RT.

Ap r i l 25 , 2005
 Electromigration: grain-boundary diffusion
6.12J / 3.155J Micr oelectr onic pr ocessing

DA F DA qZA* J!
J A = cA v A = cA = cA
RT RT

Electromigration is problematic
• at high current density, J
• high resistivity, ρ (many electron-atom collisions),
• for large grain-boundary diffusion, D
• at high T (which is in exponent of DA),
• for light metals (DA0 is inverse function of mass of A)

Ap r i l 25 , 2005
 Electromigration damage: due to flux divergence or temperature gradients
6.12J / 3.155J Micr oelectr onic pr ocessing

Fick's second law of diffusion states that change in concentration of species A

occurs as a result of a divergence in JA, i.e. a variable concentration gradient:

!cA !J A ! 2 cA
=" = DA 2
!t !x !x
Add temperature-dependent term to time rate of change of concentration as follows:

dc A "J A "J A dT
=! !
dt "x "T dx

Isothermal mass transport temperature gradients

due to flux divergence associated with
jA such as at local hot or cold spots
grain boundary junctions. couple with temperature
dependence of JA.

Ap r i l 25 , 2005
 Electromigration vs. linewidth/grain size
(Thompson-Frost model)
6.12J / 3.155J Micr oelectr onic pr ocessing

Yield
w/d50 ≈ 3.0 d50

w/d50

DAc A d#
JA = (qZ*A J ! + " ) w/d50 ≈ 1.3
k BT dx Voids hillocks
Equilibrium: Mass flow

d! qZA* J# , ! = "ax + b
="
dx $ Lp
$& Z* qJ" ' Lp
! max = ±
% # ( 2 w/d50 ≈ 0.3
! crit 2"
JLp <
Z * q#
Ap r i l 25 , 2005
 Electromigration summary
6.12J / 3.155J Micr oelectr onic pr ocessing

electron wind,
mass transport

Voids, depletion
Accumulation, hillocks

Most electromigration takes place 4 microns

along grain boundaries.
Grain boundaries that run parallel to current direction
are most problematic.

A factor cos (θ) is often attached to the atomic flux expression to reflect this fact;
a is the angle between the current and the grain boundary.

Ap r i l 25 , 2005
 Summary
6.12J / 3.155J Micr oelectr onic pr ocessing

Electro-chemical deposition
Need driven by Cu and MEMS
Potential initiates M+ diffusion; diffusion limited
Excellent step coverage, esp. for small features
Reliability
Yield, Y ∝ e-AD
F (t) = fraction of failures up to time, t. Prob. Den: f (t) = dF/dt
Infant mortality, steady state, wear out; different processes

Electromigration
Momentum transfer from charge carriers to atoms
Aggrevated by large J, high T, small mass (Al)
Role of grain boundaries and T gradients

Ap r i l 25 , 2005