A quick review of MOS Capacitors

EE231 – Vivek Subramanian Slide 1-1

MOS Capacitors

What happens when the

work function is different?

EE231 – Vivek Subramanian Slide 1-2

1
MOS Band Diagram – Different Work function

EE231 – Vivek Subramanian Slide 1-3

Flat Band Condition E

0
χSiO2 =0.95 eV
Ec
qψM
3.1 eV 3.1 eV χSi qψs = χSi + (Ec –Ef )
=4.05eV

Ec, Ef Ec
Vfb
Ef
Ev
Ev
N+ -poly-Si 9 eV P-body

E0 : Vacuum level
Vfb = ψM −ψs
4.8 eV
E0 – Ef : Work function
E0 – Ec : Electron affinity
Si/SiO2 energy barrier Ev
SiO2
EE231 – Vivek Subramanian Slide 1-4

2
 Non-Flat-band conditions
What if
Vg ≠ V fb
voltage across the substrate,
i.e. the band bending in the
Vg = V fb + VOx + ϕ S substrate, also called surface potential

voltage
across
the oxide ϕS
VOx

Ef
substrate charge Ef Vg

− QS QS = Qdep + Qinv + Qacc

VOx =
COx
EE231 – Vivek Subramanian Slide 1-5

Surface Accumulation

3.1eV
Vg = V fb + φ s + Vox
Vox
Ec ,Ef − ( E f − Ev ) / kT
p = Nve
Ev E0
qV g
φs is negligible
qφs − QS
Ec VOx =
COx
Ef
Ev Vg = V fb + VOx + ϕ S
QS = −COxVOx
M O S = −COx (Vg − V fb )

EE231 – Vivek Subramanian Slide 1-6

3
 Depletion
qVox

qφs
Ec Vg = V fb + VOx + ϕ S
Ef
-- Ev
-- − QS − Qdep
qVg
Ec, Ef Wdep VOx = =
depletion COx COx
Ev
region
+ qN a X dep
=
COx

M O S 2qN aε Sϕ S
=
COx

Can solve for ϕS(Vg) and VOx(Vg)

EE231 – Vivek Subramanian Slide 1-7

Threshold (of Inversion)

Ec
nS = p0 = Na φst
Ei
( EC − E f ) S = ( E f − EV )bulk C=D A C = qφ Β
D
Ef
At Vt ϕ S = 2 ϕ B ≈ 0.8V A = B qVg = qVt
B
Ev
1
(
ϕ B = Ei − E f bulk
q
) Ec,Ef

kT N a
= ln ≈ 0.4V
q ni Ev
M O S
(Alternative definition: ϕS,th = ϕB + 0.45V)

VT = V g (ϕ S = 2ϕ B ) = V fb + ϕ S + VOx n = NCe
(
− E C − E f / kT )

VT = V fb + 2 ϕ B +
2 qε S N a 2ϕ B
(
− E f − E V / kT)
C Ox p = NV e
0.15V
≈ 10 19 cm − 3

EE231 – Vivek Subramanian Slide 1-8

4
 Inversion
ϕ S ≥ 2ϕ B ϕ S ~ 2ϕ B EC

Qdep + Qinv EF
Vg = Vfb + 2ϕB −
COx
2qεS Na 2ϕB Qinv Q ϕ S ≥ 2ϕ B
= Vfb + 2ϕB + − = VT − inv large Qinv
COx COx COx
large QS
Qdep
VT = Vfb + 2ϕB − large VOx
COx

Qinv = −COx(Vg −VT )

EE231 – Vivek Subramanian Slide 1-9

Review : Basic MOS Capacitor Theory

φs

2φB

Vg
Vf b Vt
accumulation depletion inversion

Wdep

Wdmax
Wdm ax = (2εs2φΒ /qΝ a )1/2

∝ (φs)1/2

Vg
Vfb Vt
accumulation depletion inversion

EE231 – Vivek Subramanian Slide 1-10

5
 Review : Basic MOS Capacitor Theory
Qdep = qNaWdep Qs

V
accumulation fb depletion inversion accumulation depletion inversion
Vg regime regime regime
(a) 0 Vt
–qNaWdep

–qNaWdmax Vf b
Vg
Vt
Qinv Qs = Qacc + Qdep + Qinv 0

accumulation Vfb depletion inversion

Vg Q inv
(b) Vt
slope = − C
ox slope = − Cox
small-signal capacitance:
C C = − dQS
dVg
Qa cc
COx

slope = − Cox
(c)

Vg VT
Vg
Vfb V Vfb
accumulation depletion t inversion

EE231 – Vivek Subramanian Slide 1-11

Advanced MOSCAP Physics and Technology

EE231 – Vivek Subramanian Slide 1-12

6
 So were we too simplistic?
• Problem with previous analysis
– Assumes that there are no free carriers in the depletion region
(depletion approximation)
– Obviously, this is not true (else, how could we have inversion
charge?)
N
a
ϕ(x)
p ( x ) = p0 e − qϕ ( x ) / kT
= p0 e − βϕ ( x ) ϕ (∞ ) = 0

q
β≡ ϕ(0) = ϕ
S EF Na, p-type
kT
EV

n( x ) = n0 e qϕ ( x )/ kT Eq. 3-9 SiO 2 Si

ni2
x
Na 0

EE231 – Vivek Subramanian Slide 1-13

A more general and accurate MOSCAP analysis

• If we don’t assume complete absence of carriers in the

depletion region, we have:
d 2ϕ ρ (x )
dx 2
=−
εS
=−
q
εS
p + N d+ − n − N a− ( )
=−
εS
q
(p e 0
− βϕ ( x )
− n0 e βϕ ( x ) − N a + N d )
• Now, we can proceed by noting that:

2
d 2ϕ 1 d  dϕ 
=  
dx 2 2 dϕ  dx 

EE231 – Vivek Subramanian Slide 1-14

7
 Calculation of E(x)
• Therefore, we have:
2
d 2ϕ 1 d  dϕ  −q
dx 2
=   =
2 dϕ  dx  εS
(
p0 e − βϕ − n0 e βϕ − N a + N d )
2
 dϕ  − 2q ϕ
⇒
dx
 =
ε ∫ (
p0 e − βϕ − n0 e βϕ − N a + N d dϕ )
  S
0

=
− 2q
[
− p0 (e − βϕ + βϕ − 1) − n0 (e βϕ − βϕ − 1)
εS β
]
1/ 2
dϕ  2 q 
=
dx  ε S β
(
p0 (e − βϕ + βϕ − 1) + n0 (e βϕ − βϕ − 1)  )


• Which is –E(x)

EE231 – Vivek Subramanian Slide 1-15

Calculation of QS
• Now, we can solve this equation at x=0 to find the peak
field. By Gauss’ law, we can therefore find the total
charge in the silicon
QS = −ε S ES by Gauss' s Law
dϕ dϕ
= εS = εS
dx S dx ϕ =ϕ S

=
2ε S q
β
[p (e
0
− βϕ S
) (
+ βϕ S − 1 + no e βϕ S − βϕ S − 1)] 1/ 2

• This equation is valid in all regions of operation, since we

haven’t made any region-specific assumptions so far
EE231 – Vivek Subramanian Slide 1-16

8
 Calculation of inversion and depletion charge
• From the previous analysis, we can also calculate the various charge
components, namely the inversion layer charge and the depletion layer
charge (and, of course, the accumulation charge in accumulation).
• For example, in inversion, we find:
xc ϕc
n(ϕ )dϕ
Q I = − q ∫ n( x)dx = −q ∫
0 ϕsurf
dϕ / dx

• Since we know E(x) and n(x), we can solve for the total inversion layer
charge.
• Similarly, we can also solve for QB, which consists of ionized
acceptors and holes.

EE231 – Vivek Subramanian Slide 1-17

Graphical analysis of charge distributions

• The results of the previous analysis are plotted below:

Accumulation Depletion Onset of Inversion

• Specific conclusions:
1. The inversion layer charge is extremely close to the surface
2. There are very few mobile carriers in the depletion region

i.e., our simplistic analysis in the previous section is actually reasonably

accurate.

EE231 – Vivek Subramanian Slide 1-18

9
 Extraction of C-V characteristics
• From our equation for QS, we can also determine more accurate C-V
characteristics that predicted by the previous analysis.
2ε S q
QS =
β
[p (e
0
− βϕ S
) (
+ βϕ S − 1 + no e βϕ S − βϕ S − 1 )] 1/ 2

--
Es - - -- -
--- -
-

dQS qε S β ( )
p 0 − e − βϕ S + 1 + no e βϕ S − 1 ( )
CS = − =
dϕ S 2 [ ( ) (
p 0 e − βϕ S + βϕ S − 1 + no e βϕ S − βϕ S − 1 )]
1/ 2
QS = Qinv + Qdep + Qacc
qε S β p S − nS − N a + N d
= C
2 p S + n S + (βϕ S − 1)( N a − N d ) − 2n0

Vg
V

EE231 – Vivek Subramanian Slide 1-19

Extraction of C at VT
• We can use this equation to determine specific capacitance values

qε S β p S − nS − N a + N d
CS =
2 p S + n S + (βϕ S − 1)( N a − N d ) − 2n0

• At VG = VT, we have:
qβε S 2Na qε S N a εS εS
CS = = = = COx
2 2 βϕ B N a ϕB ϕ Bε S / qN a X d max / 2
εS
Cdep =
X d max
2ε S 2ϕ B
• Where X d max =
qN a dQinv
Cinv = −
dϕ S
• This value is different from the value predicted by the simple model
by a factor if 2, since the latter does not include the inversion charge
present at VT

EE231 – Vivek Subramanian Slide 1-20

10
 Extraction of C at VFB
• Similarly, at VG = VFB, we find that:
COx
εS
CS (V fb ) =
ϕ
S

εS Debye length
Cinv Cdep Cacc
qN a β

X dep for band

bending of kT/2q
∆ϕs
1 kT
=
β q
x=0
EV

• Again, we find that the simplistic model is somewhat inaccurate,

since it doesn’t include the “wiggle” around flat-band, which affects
the charge in the silicon. This effect is small, of course.

EE231 – Vivek Subramanian Slide 1-21

MOSCAP LFCV Characteristics

• We can plot the variation in CS in the various regions to find the
MOSCAP LFCV characteristics:
CS(ϕS)

εS εS
Cdep ≈ C Ox
LD 2ε S X d max
X d max εS Cs
Cinv =
ϕS X d max
2ϕ B

acc
QS (ϕ S )
Cdep
Vg = V fb + ϕ S − 0
COx C C

LF : Q inv , Q acc , Q dep , i.e., n, p respond to

changes in ϕ S , eg., p = p 0 e − βϕ ,
ϕS Vg
i.e. n and p are at equilibrium. 0

EE231 – Vivek Subramanian Slide 1-22

11
 MOSCAP LFCV Characteristics
• Graphically:
QS
CS

dQS
CS = − C Ox

acc
2ϕB
ϕS dϕ S
0

ϕ
S

∝ ϕs
Cg
Vg

M
C Ox
ϕs

S Cs

Substrate cap.
ϕ => Vg
S

EE231 – Vivek Subramanian Slide 1-23

MOSCAP HFCV Characteristics

• We can perform a similar analysis for HFCV:
COx
ϕS COx

Cs
Cinv
Cdep Cacc

C S (ϕ S ) for HF?

CS =
qε S β ( )
p 0 − e − βϕ S + 1 + n0 e βϕ S − 1 ( )
2 [ ( ) (
p0 e − βϕ S + βϕ S − 1 + n0 e βϕ S − βϕ S − 1 )]
1/ 2

Let n 0 → 0 (since the inversion layer does not respond quickly enough)
C

C
CS(HF)

εS
HF or DD ~
X d max
ϕ
ϕS S
V
g

EE231 – Vivek Subramanian Slide 1-24

12
 Summary of MOSCAP CV Characteristics
C
LF
Vg
Linear ramp
+ ac signal HF
t DD
Vg
n
I. LF: n in equilibrium with ϕs(Vg)
LF
n : nbias (Vgbias ) + nac (Vgac )
t

II HF: n
ϕs≤2 ϕB QS
HF
n: nbias (Vgbias) Vg = Vfb + 2ϕB +
n ac =0 Xd ≤ Xdmax COx
t

III Deep Depletion: n QS

ϕS ϕs>2 ϕB C Xd > Xdmax
Vg = V fb + 2ϕ B +
nbias=0 COx HF
n: 2ϕ B
n ac=0 DD
t t Vg

EE231 – Vivek Subramanian Slide 1-25

The Charge Sheet Model

• Problems with the simple model
– Inaccurate in depletion
– Inaccurate in accumulation
– Inaccurate in weak inversion ( 2ϕB > ϕS > ϕB)
• Problems with the general model
– Requires numerical solution for QI

The charge sheet model provides a reasonable tradeoff

between the two. It isn’t as accurate in depletion or
accumulation, but these regions aren’t as important for
MOSFET operation

EE231 – Vivek Subramanian Slide 1-26

13
 Main Assumptions
• Mobile charge exists beyond the onset of weak inversion
(i.e., QI > 0 for ϕS > ϕB
– unlike the simple model, which assumes that the inversion charge
is zero for ϕS < 2ϕB

• Mobile charge is present in a negligibly thin layer

– similar to the simple model

• Depletion region has a sharp boundary

– similar to the simple model

• The surface potential is not clamped past threshold

– unlike the simple model, which assumes that ϕS is clamped at 2ϕB
for all values of VG past threshold

EE231 – Vivek Subramanian Slide 1-27

Derivation of QS
• As in the general model, we have:

d 2ϕ − q
= (
p e − βϕ − n0 e βϕ − N a + N d
εS 0
)
dx 2
1/ 2
dϕ  2q 
=
dx  ε S β
(
p 0 (e − βϕ + βϕ − 1) + n0 (e βϕ − βϕ − 1)  )

2ε S q
QS =
β
[p (e
0
− βϕ S
) ( )]
+ βϕ S − 1 + no e βϕ S − βϕ S − 1
1/ 2

• To simplify, assume we only care about QS in weak and strong

inversion (i.e., ϕS > ϕB). Then:
2qN aε S  n2 
| QS |=  βϕ S + i 2 e βϕ S  =| Qinv + Qdep |
β  Na 

EE231 – Vivek Subramanian Slide 1-28

14
 Derivation of Qdep and Qinv
• Then, we can determine Qdep as in the simple model, and subtract to
find Qinv
Qdep = 2qN aε S ϕ S

2qN aε S  n2 
Qinv =  βϕ S + i 2 e βϕ S  − 2qN aε S ϕ S
β  Na 
QS (ϕ S )
Vg = V fb + ϕ S −
COx

• Note that Qinv(VG) still requires an iterative solution

• In strong inversion, since the exponent dominates, we can simplify to:
2qε S ni e βϕ S
2
Qinv =
βN a
• This exponential dependence implies that large changes in Qinv result
from small changes in ϕS, which means that ϕS is essentially clamped,
as was assumed in the simple model.

EE231 – Vivek Subramanian Slide 1-29

Simplification under weak inversion

• In weak inversion, Qinv < QB. This allows us to simplify Qinv:
2ϕ B > ϕ S > ϕ B
 ni
2  qε S N a ni
2
Qinv = 2ε S qN aϕ S  1 + e βϕ s − 1 ≈ − e βϕ s
 ϕ β N
2
 2ϕ S βN a 2
 S a 
log Q inv
log Q inv From first 2
terms of taylor
series
∝ e qϕ S / kT
1 decade of 1 decade of
increase increase
for 60 mV in ϕ s for 60/ η mV in V g
ϕS
Vg

QS (ϕ S ) typically 80 mV
Vg = V fb + ϕ S −
COx

• This equation clearly incorporates the effect of subthreshold current in

MOSFETs, unlike the simple VT equations studied previously (for
example, in EE130)

EE231 – Vivek Subramanian Slide 1-30

15
 Oxide Charges
• In general, these charges all
modify the threshold voltage
based on their charge centroid
x0
1
∆VT = −
ε SiO ε 0 ∫ xρ
0
ox ( x)dx
2

• In addition, they may alter

mobility due to coulombic
scattering

Q + Qit ∫0 xρ ( x )dx
TOx
Q
Vg = φ MS + ϕS − S − = V fb + ϕ S − S
COx ε Ox COx

EE231 – Vivek Subramanian Slide 1-31

Mobile Ions
• People observed odd shifts in C-Vs

• Reason: Mobile charge was moving towards /

away from interface, changing charge centroid
EE231 – Vivek Subramanian Slide 1-32

16
 Interface traps

Traps cause “sloppy” C-V and also

greatly degrade mobility in channel

EE231 – Vivek Subramanian Slide 1-33

Noise due to Interface

Traps
Telegraphic noise in Id of a small
MOSFET is the signature of a
signal interface trap.
ninv×W×L
Id ∝ µNinv
When a single trap changes from
“empty” to “filled” ∆Ninv=-1.

∆ Id ∝ µ ∆ Ninv + Ninv ∆ µ
= -µ + Ninv ∆ µ
∆I d −1 ∆µ
= +
Id Ninv µ

∆ µ<0 : acceptor type

∆ µ>0 : donor type

EE231 – Vivek Subramanian

17
 “Electrical Tox”Raised by Inversion
Change Centroid and Gate Depletion

Physical Tox
Electrical Toxe

EE231 – Vivek Subramanian

Polysilicon gate depletion Npoly

Impact on IV Cpoly Xdep

increase the effective Tox by ~5Å -20Å COx

Qinv = ∫ CdVg Cdep

−1 −1
 1 1   W 
decrease of Vg by ~0.2V, C = +  =  Tox + dpoly 
C C  ε εs
also has impact on CV and S  ox poly   ox 
ε ox
Cdep =
S = 60mv(1 + ) Tox + Wdpoly / 3
COx

EE231 – Vivek Subramanian

18
 How to Reduce Gate-Depletion Effect
• Metal gate - process integration issues
Or increase active doping concentration in the gate:
• In-situ or POCl3 doped poly-Si gate
– Not suitable for dual-gate CMOS technology
• Higher dosage for ion-implanted poly-Si gate
– Cost, damage and boron penetration issues
or higher activation temperature
– S/D diffusion and boron penetration
• Poly-Si1-xGex-gate technology

EE231 – Vivek Subramanian

Quantum effect (in the inversion layer)

Sec 3.3 presents the “classical” analysis based on Poisson’s equation
and Fermi-Dirac function (or Boltzmann’s relations)
Quantum confinement in the potential well at the Si/SiO2 interface
creates discrete subbands of energy levels.
Ref. Stern, “self-consistent …”, Phy, Rev. B. vol. 5, p.4891, 1972
2/3
 2hqΕ S  3 
Ec Ej ≈   j +  j = 0,1,2
 4 2mx  4 
mx ≈ 0.9m0 for (100) Si

SiO2
Assume only ground subband is populated
E0~60mV, depending
9ε Si h 2
on Nsub and Qinv X inv ≈
 1 
40Å 16 π 2 m x q  Q dep + Q inv 
 3 
x

EE231 – Vivek Subramanian

19
 Quantum effect (in the inversion layer)
-Effect on VT
ϕS has to be larger than “2ϕB”by, say 60mV depending on Nsub.
 Cdep 
∆Vt , ie ∆Vg = ∆ϕ S 1 + , ~ 100mV
 COx 
Empirical model: Rios, “ A Physical Compact MOSFET Model ….”, IEDM
P.937, 1995

-Effect on CV

ε Si
Cinv ≡
X inv
-Effect on IV
Similar to CV, but there is a subtle difference between AC charge
centroid and DC charge centroid.

EE231 – Vivek Subramanian

“Real” MOSCAP CV Characteristics

1x10-6
Capacitance (F/cm2)

8x10-7

6x10-7

4x10-7 Tox=30A
Measured Data
Cox, Tox=30A
Nsub=5.2x1017cm-3 Classical
-7 Npoly=4.5x1019cm-3 QM+PD
2x10

-4 -3 -2 -1 0 1 2 3 4
Gate Voltage(V)
EE231 – Vivek Subramanian

20