Professional Documents
Culture Documents
01 MOS Lecture
01 MOS Lecture
MOS Transistors
Dinesh Sharma
EE Department
IIT Bombay, Mumbai
August 4, 2023
MOS Transistors
The Parallel Plate Capacitor
+ + + + + + + + Q
ti
− − − − − − − − −Q
∇·D =0
plate capacitor.
As we add more and more +ve charge to the top plate, a
thicker and thicker depletion layer is required to balance
this charge.
So the capacitance keeps reducing as we make the top
plate more and more positive.
MOS Transistors
The MOS capacitor
Accumulation Inversion
Negative Bias: Depletion
Capacitance
For a negative DC bias on the top
plate, induced charge is provided
by holes which are mobile.
V
Capacitance
top plate, induced charge is
provided by acceptor ions in the
depletion layer, which are fixed in
their location. V
Accumulation Inversion
Positive Bias Beyond Inversion:
Depletion
Once inversion sets in, practically
Capacitance
all of the additional induced
negative charge is provided by
electrons which are mobile.
V
Inversion layer
Depletion Region
(Depletion width is constant after
inversion point, practically all the
Onset of inversion)
additional charge required in silicon is
(Only holes can respond to HF)
p type Si provided in the form of electrons. No more
Back Contact depletion charge is required.
Depletion width becomes nearly constant
beyond the inversion voltage.
MOS Transistors
The MOS capacitor
Inversion layer
Depletion Region form of electrons.
(Depletion width is constant after
Onset of inversion)
So the incremental charge in response to
(Only holes can respond to HF)
p type Si a high frequency AC signal has to be
Back Contact provided by holes.
This leads to a fluctuation in the width of
the depletion layer.
MOS Transistors
The MOS capacitor
To summarise:
In response to positive DC bias, the depletion width
increases till the inversion point and becomes constant at
this value for higher DC bias values.
This is because practically all the additional silicon charge
needed (corresponding to the bias voltage being higher
than the inversion voltage) is in the form of electrons.
If the superimposed AC signal is changing very fast, the
incremental charge in response to this signal cannot be
provided in the form of electrons.
This is because electrons cannot be generated and
transported fast enough to respond to the signal.
MOS Transistors
The MOS capacitor
Capacitance
Depletion
Capacitance
Depletion
V V
At high frequency, capacitance remains constant at its
minimum value as the bias voltage is increased beyond the
inversion point.
MOS Transistors
The MOS capacitor
Quantitative Analysis
Quantitative Analysis
o
X
Using a one dimensional analysis, we want to relate the
semiconductor charge to the applied gate voltage.
MOS Transistors
The MOS capacitor
Quantitative Analysis
Quantitative Analysis
Vg .
Let the electrostatic potential at any point x be
denoted by φ(x)
and let the potential at the silicon-oxide o
X
interface be φs .
MOS Transistors
The MOS capacitor
Quantitative Analysis
Quantitative Analysis
φs − Vg
So Area × ǫox = Total Charge in silicon
tox
MOS Transistors
The MOS capacitor
Quantitative Analysis
Quantitative Analysis
φs − Vg
= Total Charge in silicon
Area × ǫox
tox
If we define Qsi to be the semiconductor charge per unit area,
and Cox to be the parallel plate capacitance per unit area,
Qsi
Cox (φs − Vg ) = Qsi So Vg = φs −
Cox
Thus, the surface potential and the applied gate voltage can be
related to each other.
MOS Transistors
The MOS capacitor
Quantitative Analysis
∇·D=ρ
∂2φ
or = q(Nd+ − Na− + p − n)
− ǫsi
∂x 2
Since the electrostatic potential is dependent only on x, we can
change partial derivatives to total derivatives.
d2 φ
d dφ d
− 2 = − = (E)
dx dx dx dx
d2 φ
dE dφ d d 1 d 2
− 2 = = (E) = −E (E) = − E
dx dx dx dφ dφ 2 dφ
defining
q
u ≡ βφ where β ≡
KB T
We get
d2 φ 1 d 2 β d 2
− = − E = − E
dx 2 2 dφ 2 du
MOS Transistors
The MOS capacitor
Quantitative Analysis
∂2φ
−ǫsi = q(Nd+ − Na− + p − n)
∂x 2
represents the charge density. In the absence of an applied
voltage, this must be zero everywhere.
q(Nd+ − Na− + p0 − n0 ) = 0
∂2φ
−ǫsi = q(Nd+ − Na− + p − n)
∂x 2
We have established that
d2 φ β d 2
− = − E and Nd+ − Na− = −(p0 − n0 )
dx 2 2 du
q qφ
where β≡ and u≡ = βφ
KB T KB T
βǫsi d 2
Therefore − E = q [p − p0 − (n − n0 )]
2 du
MOS Transistors
The MOS capacitor
Quantitative Analysis
Quantitative Analysis
βǫsi d 2
− E = q [p − p0 − (n − n0 )]
2 du
so
d 2 2qp0 p n0 n
E =− −1− −1
du βǫsi p0 p0 n0
n = n0 e u and p = p0 e−u
This gives,
d 2 2qp0 −u n0 u
E =− e − 1 − (e − 1)
du βǫsi p0
MOS Transistors
The MOS capacitor
Quantitative Analysis
Quantitative Analysis
d 2 2qp0 −u n0 u
E =− e − 1 − (e − 1)
du βǫsi p0
This can be integrated from x = ∞ (where E = 0 and u = 0) to x
to give
2qp0 u
Z
n0
E2 = −e−u + 1 + (eu − 1) du
βǫsi 0 p0
q ǫsi
r
where β ≡ , us ≡ βφs , and LD ≡
KB T qβp0
Here LD is the Extrinsic Debye Length.
MOS Transistors
The MOS capacitor
Quantitative Analysis
Maj. Carrier
1e−06 Charge
Q
total
1e−07
Q
Depl.
1e−08
1e−09
−0.4 −0.2 0 0.2 0.4 0.6 0.8 1
Gate Voltage (V)
MOS Transistors
The MOS capacitor
Quantitative Analysis
Qsi + Qox
Vg = Φms + φs −
Cox
Where Φms is the metal to semiconductor work function
difference.
Now given φs , we can compute the semiconductor charge, and
hence the gate voltage.
MOS Transistors
The MOS capacitor
Quantitative Analysis
0.8
Surface Potential (V)
0.6
0.4
0.2
0.0
−0.2
−4.0 −2.0 0.0 2.0 4.0
GATE VOLTAGE (V)
MOS Transistors
The MOS capacitor
Quantitative Analysis
total
2
Q
1e−07 inv
Q
depletion
1e−08
1e−09
−2 −1 0 1 2 3 4 5
Gate Voltage (V)
MOS Transistors
The MOS capacitor
Quantitative Analysis
2
positive gate voltages, the Q
1e−07 inv
entire semiconductor charge
is depletion charge. Q
depletion
As the voltage exceeds a 1e−08
1e−06
Qtotal
Metal Gate
n+ n+
P type Si
Metal Gate
n+ n+
P type Si
W
Y
S D
X dy
W
Y
S D
X dy
W
Y
S D
X dy
Z
Since there is no dependence on
z, the z integral just gives a
X dy multiplication by W.
Therefore,
∞
∂V (y)
Z
Id = µWq n(x, y) dx
x=0 ∂y
MOS Transistors
I-V characteristics of a MOS transistor
A simple MOS model
L
the value of n(x,y) is non zero in a
very narrow channel near the
W
Y
surface. We can assume that ∂V∂y(y )
S D is constant over this depth.
Z ∂V (y) ∞
Z
Id = µWq n(x, y)dx
X dy
∂y x=0
R∞
q x=0 n(x, y)dx = −Qn (y) where Qn (y) is the electron charge
per unit area in the semiconductor at point y in the channel.
(Qn (y) is negative, of course).
MOS Transistors
I-V characteristics of a MOS transistor
A simple MOS model
∂V (y) ∞
Z
Id = µWq n(x, y)dx
∂y x=0
∂V (y)
= −µW Qn (y)
∂y
Integrating the drain current along the channel gives
Z L Z L
∂V (y)
Id dy = −µW Qn (y) dy
0 0 ∂y
Z Vd
Id × L = −µW Qn (y)dV (y)
0
Vd
W
Z
So, Id = −µ Qn (y)dV (y)
L 0
MOS Transistors
I-V characteristics of a MOS transistor
A simple MOS model
(Qox + Qdepl )
VT ≡ ΦMS + 2ΦF −
Cox
W Vd
Z
Therefore Id = µCox (Vg − VT − V (y))dV (y)
L 0
W 1
= µCox [(Vg − VT )Vd − Vd2 ]
L 2
MOS Transistors
I-V characteristics of a MOS transistor
A simple MOS model
W 1
ID = WCox (VG −VT −VD /2)µVD /L = µCox [(Vg −VT )Vd − Vd2 ]
L 2
MOS Transistors
I-V characteristics of a MOS transistor
A simple MOS model
p
−Qn = Cox Vg − ΦMS − V (y) − 2ΦF +Qox − 2ǫsi qNa (V (y) + 2ΦF )
This leads to
W Qox 1
Id = µCox Vg − ΦMS − 2ΦF + Vd − Vd2
L Cox 2
p #
2 2ǫsi qNa
− (Vd + 2ΦF )3/2 − (2ΦF )3/2
3 Cox
∂Id W
= µCox Vg − VT − Vd For Vd ≤ Vdss
∂Vd L
∂Id Idss
And = For Vd ≥ Vdss
∂Vd Vdss + VE
W 1 2
Where Idss ≡ µCox Vg − VT Vdss − Vdss
L 2
∂Id
On matching the value of ∂Vd on both sides of Vdss , we get
s
2 Vg − VT
Vdss = VE 1+ − 1
VE
MOS Transistors
I-V characteristics of a MOS transistor
Modeling the saturation region
s
2 Vg − VT
Vdss = VE 1+ − 1
VE
Since the value of Vdss does not change substantially from the
ideal saturation case, a simpler approach can be tried.
The drain current is calculated using the ideal saturation
model and its value is multiplied by a correction factor =
(1 + λVd ) in saturation as well as in linear regime.
This automatically assures continuity of Id and its
derivative.
λ is a fit parameter, whose value is ≈ 1/VE .
This approach is used in SPICE, a popular circuit simulation
program.
MOS Transistors
MOS Device Scaling
Scaling in general
So scale down r.
MOS Transistors
MOS Device Scaling
Grandma’s story
Remember the story that grandma used to tell about the giant
who was ten times taller, ten times broader and ten times fatter
than a human being?
And how terrified were we of this creature!
Consequences of Scaling
We assume classical or Constant Field scaling.
Device Area: Since W reduces by ↓ S and L reduces by ↓ S,
the area reduces by ↓ S 2 .
Packing Density: For a given chip area, the number of devices
which can be packed in this chip will go up by ↑ S 2 .
Cox : The gate capacitance per unit area is given by
ǫ/tox . Since tox scales down by ↓ S, Cox increases
by ↑ S. Cox determines the transconductance, so
this increase is good.
Load capacitance: All dimensions, including depletion widths
have been scaled down by ↓ S. Total capacitance
= ǫA/t. Now A reduces by ↓ S 2 , while the
dielectric thickness (be it oxide or depletion width)
reduces by ↓ S. The net effect is that total
capacitance = ǫArea ↓ S 2 /t ↓ S reduces by ↓ S.
MOS Transistors
Consequences of Scaling
Consequences of Scaling
Consequences of Scaling
Slew Rate: Slew rate is the rate of change of voltage at any
node. Since I = C dV dt , the slew rate goes as
I(↓ S)/C(↓ S). Thus slew rate remains unchanged
with scaling.
Delay: Delay is given by the total voltage change divided
by dV
dt . Since all voltages are scaled down by ↓ S,
while dVdt remains unchanged, the delay reduces
as ↓ S.
Static Power: It is given by V × I . So it scales as (↓ S)(↓ S),
that is ↓ S 2 .
Dynamic Power: Dynamic power is given by Ctotal V 2 f . This
scales as (↓ S)(↓ S 2 )(↑ S) Thus dynamic power
reduces as ↓ S 2 even when the frequency of
operation is increased by ↑ S to take advantage of
shorter delays, which scale down by (↓ S).
MOS Transistors
Consequences of Scaling
Consequences of Scaling
Classical Constant Field Scaling: All dimensions and voltages
divided by the factor S(> 1).
Device area ∝ W × L : (↓ S)(↓ S) ↓ S2
Cox ǫox /tox : const/(↓ S) ↑S
Ctotal 2
ǫA/t : (↓ S )/(↓ S) ↓S
VDS , VGS , VT Voltages : (↓ S) ↓S
Id µCox (W /L)(∝ V ) :2
(↑ S)(const)(↓ S 2 ) ↓S
dV
Slew Rate dt I/Ctotal : (↓ S)/(↓ S) const.
Delay V / dV
dt : (↓ S)/(const) ↓S
Static Power V × I : (↓ S)(↓ S) ↓ S2
dynamic power Ctotal V f : (↓ S)(↓ S )(↑ S) ↓ S 2
2 2
Impact of scaling
Moore’s “Law”
In 1965, Gordon Moore, the co-founder of Fairchild
Semiconductor as well as Intel, described a doubling every
year in the number of components per integrated circuit.
It is an observation of a trend and an empirical relationship
– not a physical or natural law! However, given the
prominence of Gordon Moore, it is widely referred to as
Moore’s Law.
In 1975, Moore modified his observation for the rate of
device scaling and predicted a doubling of device density
every two years.
It is remarkable that this trend has continued over several
decades. It is only in the last decade that the rate of
doubling has slowed down remarkably, as we hit several
physical limits.
MOS Transistors
Consequences of Scaling
The λ rules
It is hard to track and scale the optimum size of numerous
structures on an Integrated circuit.
It is common to describe feature sizes in units of a
parameter called λ.
Now sizes of various structures can be described in units
of λ.
As we scale technologies, we just scale the value of λ.
Feature sizes remain the same in units of λ. This is
convenient.
The smallest feature on a chip is the contact window. The
value of λ is so defined that the smallest feature size is 2λ.
The smallest registration rule – for example the extent to
which a contact window must be inside a diffused region –
is λ.
MOS Transistors
Consequences of Scaling
The scaling rate has slowed down after 2010. This is because
feature sizes had already reached about 20 nm – about 3% of
the wavelength of sodium light!
MOS Transistors
Consequences of Scaling
Resist Resist
Metal Metal
Resist Resist
undercut
Metal
MOS Transistors
Consequences of Scaling
Limits of scaling
Scaling is being limited now due to several reasons.
We are reaching limits of resolution possible with
photo-selective processes and etching etc.
Traditional Device Physics is not valid any more for such
small structures. Remember, the lattice constant of Silicon
is ≈ 0.5 nm. So there are as few as 20 atoms between
source and drain of a 10 nm channel MOSFET. Clearly,
conduction models based on statistics will not hold here.
Indefinite voltage scaling is not possible. If the voltage is
scaled down drastically, signal to noise ratio will become
poor and leakage currents will become dominant as KT /q
has not been scaled and current equations of junctions
involve qV /KT .
System considerations such as interconnect delay will limit
performance gain.
MOS Transistors
Limits of scaling
Unscaled Interconnect Delay
Interconnect Design
L LW
R=ρ , C=ǫ
L Wtm ti
tm W
ti L2
Charge Time ≈ RC = ρǫ
tm ti
To first order, delay is independent of W.
This is because increasing W reduces resistance but
increases capacitance in the same ratio.
Unfortunately W is about the only parameter that the circuit
designer can decide! (L is fixed by the distance between
the points to be connected, ρ, ǫ, tm and ti are decided by the
technology).
MOS Transistors
Limits of scaling
Concerns in Interconnect Design
ρǫ 2
Interconnect Delay = tm ti L ≡ AL2
For local interconnects, L scales the same way as tm , ti ,
so delay is invariant.
For Global Interconnects, L goes up with die size, while tm and
ti scale down. This leads to a sharp increase in delay.
MOS Transistors
Limits of scaling
Solutions for Interconnect Delay problem
Buffer Insertion
Avalanche Breakdown
GATE
S D Parasitic NPN
In saturation region, there is a high field
Bipolar Transistor region next to the drain.
Channel
Avalanche Breakdown
Punch Through
S
Due to the drain field, the source/substrate
D Parasitic NPN
junction will get forward biased and will
Bipolar Transistor
Channel
inject current into the channel, even if the
gate voltage is below VT .
This is an extreme case of drain induced barrier lowering.
It results in heavy current flow even though the transistor is
supposed to be ‘OFF’.
This is known as “Punch Through”
MOS Transistors
Parasitic Devices
Diffusion Line
Oxide
enough potential.
Voltages much higher than VDD can occur in a MOS circuit
due to charge pumping effects.
Doping levels and field oxide thickness have to be chosen
so that this parasitic device never turns ON.
MOS Transistors
Parasitic Devices
Output
(Metal)
The figure on the
VDD Input
left shows a badly
Poly Poly
Input Output
VDD (Metal)
laid out inverter.
(Metal) cross section
through this line While the lay out
VDD
Gnd
Substrate does form an
Contact nMOS source pMOS source Well Contact
R well
inverter as
Vertical PNP
p+ n+ p+ n+
p n
desired, it also
Vertical PNP
p substrate n Well
Horizontal
NPN
Horizontal
forms a parasitic
p substrate
NPN latch-up structure.
Output
1 the p+ source of a pMOS
VDD Input (Metal)
transistor connected to VDD
Poly Poly
Input Output
VDD (Metal)
(which becomes the emitter),
(Metal) cross section
Gnd
through this line
2 the n well (which becomes the
VDD
Substrate
R well Contact nMOS source pMOS source Well Contact base), and
Vertical PNP
p+ n+ p+ n+
p n
3 the p substrate (which
Vertical PNP
p substrate n Well
Horizontal
NPN
Horizontal
NPN
p substrate becomes the collector of this
transistor).
The n well is connected to VDD through a resistive path, which
represents the resistance of the n well to the well contact.
MOS Transistors
Parasitic Devices
VDD Input
Output
(Metal)
1 the n+ source of an nMOS
Input Output
Poly Poly transistor connected to ground
(Metal)
VDD (Metal)
cross section (which becomes the emitter),
through this line
VDD Gnd
Substrate
Contact nMOS source pMOS source Well Contact
2 the p substrate, (which
R well
Vertical PNP
p substrate n Well
Horizontal
NPN Horizontal
3 the n well, (which becomes
p substrate
NPN
the collector).
Since the collector of the npn and the base of the pnp are
both formed by the n well, these two are connected.
Similarly, the collector of the pnp and the base of the npn
are formed by the p substrate, so these are also
connected.
MOS Transistors
Parasitic Devices
Preventing Latch up
Preventing Latch up
Latchup prevention requires care in process design, as well as
in layout rules.
nMOS pMOS n Well
VDD Input
Output
(Metal)
The doping gradient of the n well should
Input Output
Poly Poly
VDD (Metal)
be made retrograde. (Doping should
(Metal) cross section
VDD Gnd
Substrate
through this line
increase as we go deeper). This kills the
Contact nMOS source pMOS source Well Contact
p+ n+ p n p+ n+
Vertical PNP
p substrate n Well
Horizontal
NPN Horizontal
NPN
p substrate
The n well should be deep to kill the gain
Substrate contact n source
Gnd Gnd
p source
VDD
Well contact
VDD of the pnp transistor.
p+ n+ n+ p+ p+ n+
n well
n channel transistors should be placed far
p substrate
from the edge of the n well to increases
Rwell the base width of npn transistors and
Vertical pnp
reduce their current gain.
horizontal npn Rsub p channel transistors should also be
placed far from the well edge.
MOS Transistors
Parasitic Devices
Preventing Latch up
Substrate contact n source p source Well contact
Gnd Gnd VDD VDD
The n well should have a guard ring
p+ n+ n+ p+ p+ n+
n well connected to VDD , which will collect any
p substrate
current which could form the base current
Rwell of the pnp.
Vertical pnp
In layout, substrate and well contacts
horizontal npn Rsub should be placed frequently to reduce the
value of Rwell and Rsubstrate .
Normally, the n well is connected to VDD and the p well is
connected to ground. Thus there is no forward bias for emitter
base junctions and bipolar action is prevented.
However, if layout rules are not observed, latch up may be
triggered by stray currents from junctions near avalanche, or
due to exposure to light etc.