EECS 105 Fall 2003, Lecture 7

Lecture 7:
IC Resistors and Capacitors

Prof. Niknejad

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Lecture Outline

Review of Carrier Drift

Velocity Saturation

IC Process Flow

Resistor Layout

Diffusion

Review of Electrostatics

MIM Capacitors

Capacitor Layout

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Thermal Equilibrium

Rapid, random motion of holes and electrons at

“thermal velocity” vth = 107 cm/s with collisions
every τc = 10-13 s. 1 * 2 1
2 mn vth = 2 kT

Apply an electric field E and charge carriers

accelerate … for τc seconds λ = vthτ c
zero E field
λ = 107 cm / s × 10 −13 s = 10 −6 cm
v th

positive E (hole case)

aτ c

vth
x
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Drift Velocity and Mobility

For holes:  Fe   qE   qτ c 
vdr = a ⋅τ c =  τ c =  τ c =  E
m  m  m 
 p   p   p

vdr = µ p E

For electrons:
 Fe   − qE   qτ c 
vdr = a ⋅τ c =  τ c =  τ c = − E
m   m  m 
 p   p   p 

vdr = − µ n E

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Mobility vs. Doping in Silicon at 300 oK

“default” values: µ n = 1000 µ p = 400

Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Speed Limit: Velocity Saturation

c = 3 × 1010 m / s

Thermal Velocity

V 4 V cm V
10 4
= 10 =1
cm cm 10 µm
4
µm

The field strength to cause velocity saturation may seem very large
but it’s only a few volts in a modern transistor!
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Drift Current Density (Holes)

Hole case: drift velocity is in same direction as E

hole drift
current density
Jpdr

vdp
E

The hole drift current density is:

Jp dr
= q p µp E

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Drift Current Density (Electrons)

Electron case: drift velocity is in opposite direction as E

electron drift
current density
Jndr
vdn
J ndr = −(− q)nµ n E = qnµ n E
E

The electron drift current density is:

Jndr = (-q) n vdn units: Ccm-2 s-1 = Acm-2
J = J pdr + J ndr = (qpµ p + qnµ n )E
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Resistivity

Bulk silicon: uniform doping concentration, away from surfaces

n-type example: in equilibrium, no = Nd
When we apply an electric field, n = Nd

J n = qµ n nE = qµ n N d E

Conductivity σ n = qµ n N d ,eff = qµ n ( N d − N a )
1 1
Resistivity ρn = = Ω − cm
σ n qµ n N d ,eff
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

IC Fabrication: Si Substrate

Pure Si crystal is starting material (wafer)

The Si wafer is extremely pure (~1 part in a billion
impurities)

Why so pure?
– Si density is about 5 10^22 atoms/cm^3
– Desire intentional doping from 10^14 – 10^18
– Want unintentional dopants to be about 1-2 orders of
magnitude less dense ~ 10^12

Si wafers are polished to about 700 µm thick
(mirror finish)

The Si forms the substrate for the IC
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

IC Fabrication: Oxide

Si has a native oxide: SiO2

SiO2 (Quartz) is extremely stable and very
convenient for fabrication

It’s an insulators so it can be used for house
interconnection

It can also be used for selective doping

SiO2 windows are etched using photolithography

These openings allow ion implantation into selected
regions

SiO2 can block ion implantation in other areas

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

IC Fabrication: Ion Implantation

oxide

P-type Si Substrate

Si substrate (p-type) N-type diffusion region

Grow oxide (thermally)

Add photoresist

Expose (visible or UV source)

Etch (chemical such as HF)

Ion implantation (inject dopants)

Diffuse (increase temperature and allow dopants to diffuse)
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

“Diffusion” Resistor
Oxide
N-type Diffusion Region

P-type Si Substrate

Using ion implantation/diffusion, the thickness and

dopant concentration of resistor is set by process

Shape of the resistor is set by design (layout)

Metal contacts are connected to ends of the resistor

Resistor is capacitively isolation from substrate
– Reverse Bias PN Junction!
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Poly Film Resistor

Polysilicon Film (N+ or P+ type) Oxide

P-type Si Substrate

To lower the capacitive parasitics, we should build

the resistor further away from substrate

We can deposit a thin film of “poly” Si (heavily
doped) material on top of the oxide

The poly will have a certain resistance (say 10
Ohms/sq)
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Ohm’s Law

Current I in terms of Jn V = IR

Voltage V in terms of electric field I = JA = JtW

I = JA = JtW = σ t W E
E =V / L
σ tW
– Result for R I = JA = JtW = V
L 1 L ρ L
R= R=
W σt W t
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Sheet Resistance (Rs)

IC resistors have a specified thickness – not

under the control of the circuit designer

Eliminate t by absorbing it into a new
parameter: the sheet resistance (Rs)

ρL  ρ  L  L
R= =    = Rsq  
Wt  t  W  W 

“Number of Squares”

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Using Sheet Resistance (Rs)

Ion-implanted (or “diffused”) IC resistor

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Idealizations

Why does current density Jn “turn”?

What is the thickness of the resistor?

What is the effect of the contact regions?

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Diffusion

Diffusion occurs when there exists a concentration
gradient

In the figure below, imagine that we fill the left
chamber with a gas at temperate T

If we suddenly remove the divider, what happens?

The gas will fill the entire volume of the new
chamber. How does this occur?

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Diffusion (cont)

The net motion of gas molecules to the right
chamber was due to the concentration gradient

If each particle moves on average left or right then
eventually half will be in the right chamber

If the molecules were charged (or electrons), then
there would be a net current flow

The diffusion current flows from high
concentration to low concentration:

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Diffusion Equations

Assume that the mean free path is λ

Find flux of carriers crossing x=0 plane
n (λ )
F = vth (n(−λ ) − n(λ ) )
n (0 )
1
2
n ( −λ )
1  dn   dn  
F= vth   n(0) − λ  − n(0) + λ  
2  dx   dx  
1 1
n(−λ )vth n(λ )vth dn
2 2 F = −vth λ
dx
−λ 0 λ dn
J = − qF = qvth λ
dx

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Einstein Relation

The thermal velocity is given by kT
1
2 mn*vth2 = 12 kT
Mean Free Time
λ = vthτ c
τc kT qτ c
vth λ = vth2 τ c = kT =
mn* q mn*

dn  kT  dn
J = qvth λ = q µ n 
dx  q  dx
 kT 
Dn =   µ n
 q 

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Total Current and Boundary Conditions

When both drift and diffusion are present, the total
current is given by the sum:
dn
J = J drift + J diff = qµ n nE + qDn
dx

In resistors, the carrier is approximately uniform
and the second term is nearly zero

For currents flowing uniformly through an interface
(no charge accumulation), the field is discontinous
J1 (σ 1 ) J1 = J 2
σ 1 E1 = σ 2 E2
E1 σ 2
J 2 (σ 2 ) =
Department of EECS
E2 σ 1 University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Electrostatics Review (1)

Electric field go from positive charge to negative
charge (by convention)
+++++++++++++++++++++

−−−−−−−−−−−−−−−

Electric field lines diverge on charge
ρ
∇⋅E =
ε

In words, if the electric field changes magnitude,
there has to be charge involved!

Result: In a charge free region, the electric field
must be constant!

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Electrostatics Review (2)

Gauss’ Law equivalently says that if there is a net
electric field leaving a region, there has to be
positive charge in that region:
+++++++++++++++++++++

−−−−−−−−−−−−−−−

Electric Fields are Leaving This Box!

Q
∫ E ⋅ dS = ε
Recall:
ρ Q
∫ ∇ ⋅ E dV =∫ dV = Q / ε ∫ ∇ ⋅ E dV =∫ E ⋅ dS =
V V
ε V S
ε
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Electrostatics in 1D

Everything simplifies in 1-D
dE ρ ρ
∇⋅E = = dE = dx
dx ε ε

ρ ( x' )
x
E ( x) = E ( x0 ) + ∫ dx'
x 0
ε

Consider a uniform charge distribution E (x) ρ0

x1
Zero field ρ (x) ε
boundary ρ0
condition
x x x1
x1 ρ ( x' ) ρ0
E ( x) = ∫ dx' = x
0
ε ε
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Electrostatic Potential

The electric field (force) is related to the potential
(energy):
dφ
E=−
dx

Negative sign says that field lines go from high

potential points to lower potential points (negative
slope)

Note: An electron should “float” to a high potential
point: φ dφ
dφ 1
Fe = − e
Fe = qE = −e dx
dx e
φ2
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

More Potential

Integrating this basic relation, we have that the
potential is the integral of the field: φ (x)
r
φ ( x) − φ ( x0 ) = − ∫ E ⋅ dl r
C
dl

In 1D, this is a simple integral:
E
x
φ ( x0 )
φ ( x) − φ ( x0 ) = − ∫ E ( x' )dx'
x0

Going the other way, we have Poisson’s equation in

1D:
d 2φ ( x) ρ ( x)
=−
dx 2
ε
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Boundary Conditions

Potential must be a continuous function. If not, the fields
(forces) would be infinite

Electric fields need not be continuous. We have already
seen that the electric fields diverge on charges. In fact,
across an interface we have:

∆x
∫ ε E ⋅ dS = −ε E S + ε
1 1 2 E2 S = Qinside
E1 (ε 1 ) Qinside ∆x
→ 0 → 0
− ε 1 E1S + ε 2 E2 S = 0
E2 (ε 2 ) S E1 ε 2
=
E2 ε 1

Field discontiuity implies charge density at surface!

Department of EECS University of California, Berkeley

EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

IC MIM Capacitor
Bottom Plate Top Plate Bottom Plate

Contacts

Thin Oxide

Q = CV

By forming a thin oxide and metal (or polysilicon) plates, a

capacitor is formed

Contacts are made to top and bottom plate

Parasitic capacitance exists between bottom plate and
substrate
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Review of Capacitors
Q
∫ E ⋅ dS = ε
+++++++++++++++++++++
+
− Vs
Q
−−−−−−−−−−−−−−−
∫ E ⋅ dS = − ε
Vs Q = CVs
∫ E ⋅ dl = E0tox = Vs E0 =
tox
Vs Q Aε
Q A= C=
∫ E ⋅ dS = E A = ε
0
tox ε tox

For an ideal metal, all charge must be at surface

Gauss’ law: Surface integral of electric field over
closed surface equals charge inside volume
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Capacitor Q-V Relation

+++++++++++++++++++++
Q y
−−−−−−−−−−−−−−−

Vs Q( y )

Q = CVs

Total charge is linearly related to voltage

Charge density is a delta function at surface (for
perfect metals)
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

A Non-Linear Capacitor

+++++++++++++++++++++
Q y
−−−−−−−−−−−−−−−

Vs Q( y )

Q = f (Vs )

We’ll soon meet capacitors that have a non-linear Q-V

relationship

If plates are not ideal metal, the charge density can penetrate
into surface
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

What’s the Capacitance?

For a non-linear capacitor, we have
Q = f (Vs ) ≠ CVs

We can’t identify a capacitance

Imagine we apply a small signal on top of a bias
voltage:
df (V )
Q = f (Vs + vs ) ≈ f (Vs ) + vs
dV V =Vs
Constant charge

The incremental charge is therefore:

df (V )
Q = Q0 + q ≈ f (Vs ) + vs
dV V =Vs
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Small Signal Capacitance

Break the equation for total charge into two terms:
Incremental
Charge

df (V )
Q = Q0 + q ≈ f (Vs ) + vs
dV V =Vs

Constant
Charge

df (V )
q= vs = C vs
dV V =Vs

df (V )
C≡
dV V =Vs
Department of EECS University of California, Berkeley
EECS 105 Fall 2003, Lecture 7 Prof. A. Niknejad

Example of Non-Linear Capacitor

Next lecture we’ll see that for a PN junction, the
charge is a function of the reverse bias:
V Voltage Across NP
Q j (V ) = −qN a x p 1 −
φb Junction

Charge At N Side of Junction

Constants

Small signal capacitance:

dQ j qN a x p 1 C j0
C j (V ) = = =
dV 2φb V V
1− 1−
φb φb
Department of EECS University of California, Berkeley