SOLUTION SET

to

Exercises in

FUNDAMENTALS OF MODERN VLSI DEVICES, 3rd ed.

published by

Cambridge University Press

Yuan Taur and Tak H. Ning

SOLUTION 1
 Solutions to Chapter 2 Exercises

2.1. From Eq. (2.4),

1 eE / kT
f ( E f  E )   ,
1  e E / kT eE / kT  1

and

1
f ( E f  E )  .
1  eE / kT

Adding the above two equations yields

eE / kT  1
f ( E f  E )  f ( E f  E )  E / kT  1.
e 1

2.2. Neglecting the hole (last) term in Eq. (2.19), one obtains
 ( E c  E f )/ kT  ( E c  E d  2 E f )/ kT
N ce  2N c e  Nd .

Treating exp(Ef/kT) as an unknown, the above equation is a quadratic equation with the solution

1  1  8( N d / N c )e( E c  E d )/ kT

E f / kT
e .
4e E d / kT

Here only the positive root has been kept. For shallow donors with low to moderate
concentration at room temperature, (Nd/Nc)exp(Ec  Ed)/kT  1, and the last equation can be
approximated by

4( N d / N c )e( E c  E d )/ kT N d E c / kT
 
E f / kT
e e ,
4e E d / kT Nc

which is the same as Eq. (2.20). If we compare the above relation with Eq. (2.19), it is clear
that in this case, exp[(Ed  Ef)/kT] << 1, and Nd+  Nd or complete ionization.

If the condition for low to moderate concentration of shallow donors is not met, then
exp[(Ed  Ef)/kT] is no longer negligible compared with unity. That means Nd+ < Nd (Eq.
(2.19)) or incomplete ionization (freeze-out). [Note that incomplete ionization never occurs for
shallow impurities: arsenic, boron, phosphorus, and antimony at room temperature, even for
doping concentrations higher than Nc or Nv. This is because in heavily doped silicon, the

SOLUTION 2
impurity level broadens and the ionization energy decreases to zero, as discussed in Subsection
[Link].]

2.3. (a) Substituting Eqs. (2.5) and (2.3) into the expression for average kinetic energy, one
obtains


 ( E  E f )/ kT
( E  Ec ) 3 / 2 e dE
K.E.  Ec
 .

 ( E  E f )/ kT
( E  Ec )1/ 2 e dE
Ec

Applying integration by parts to the numerator yields


(3 / 2)kT  ( E  Ec )1/ 2 e
 ( E  E f )/ kT
dE 3
K.E.  
Ec
 kT .

1/ 2  ( E  E f )/ kT
( E  Ec ) e dE 2
Ec

(b) For a degenerate semiconductor at 0 K, f(E) = 1 if E < Ef and f(E) = 0 if E > Ef. Here
Ef > Ec. Therefore,


Ef
( E  Ec )3/ 2 dE
3
K.E.  Ec
 ( E f  Ec ) .

Ef
( E  Ec ) dE 5 1/ 2
Ec

2.4. With the point charge Q at the center, construct a closed spherical surface S with radius r.
By symmetry, the electric field at every point on S has the same magnitude and points outward
perpendicular to the surface. Therefore,

 E  dS  4r E ,
2
S

where E is the magnitude of the electric field on S. 3-D Gauss’s law then gives

Q
E ,
4 si r 2

which is Coulomb’s law.

Since E =dV/dr, the electric potential at a point on the sphere is

Q
V ,
4 si r

SOLUTION 3
if one defines the potential to be zero at infinity.

2.5. (a) Construct a cylindrical Gaussian surface perpendicular to the charge sheet as shown:

A
Qs

The cross-sectional area is A. At the two ends of the cylinder, the electric field E is
perpendicular to the surface and pointing outward. Along the side surface of the cylinder, the
field is parallel to the surface, so EdS = 0. For an infinitely large sheet of charge, E is uniform
across A from symmetry. Therefore,

 E  dS  2AE .
S

The charge enclosed within the surface is QsA. From Gauss’s law, one obtains E = Qs/2.

(b) The field due to the positively charged sheet is Qs/2 pointing away from the sheet.
The field due to the negatively charged sheet is also Qs/2, but pointing toward the negatively
charged sheet. In the region between the two sheets, the two fields are in the same direction and
the total field adds up to Qs/, pointing from the positively charged sheet toward the negatively
charged sheet. In the regions outside the two parallel sheets, the fields are equal and opposite to
each other, resulting in zero net field.

2.6

SOLUTION 4
For x < 0,

d 2 (  i ) q 2 N d
  i  0
dx 2  si kT
General solution:

 i  Ae x / LD

where

 si kT
LD 
q2 Nd

For x > 0,

d 2 (  i ) q 2 N d q
  i   N d
dx 2
 si kT  si
General solution:
kT
 i  Be  x / LD  N d
qN d

Matching i and di/dx at x = 0,

kT
A  B  N d
2qN d

2.7
 ( E  E f ) / kT
Electron density per energy  N ( E ) f ( E )  E  EC e

Differentiate the above and set to zero gives

SOLUTION 5
 E  EC  kT / 2  13 mV

2.8
(a) The bands are flat in the region to the right  charge neutral.

Electron density: n = NC  exp[(EC – Ef)/kT] = 1.3  1016 cm-3.

Therefore, n-type with doping density Nd = Nd+ = 1.3  1016 cm-3.

(b) At point A: n = NC  exp[0.3/kT] = 2.8  1014 cm-3.

p = ni2/n = 3.5  105 cm-3.

At point B: n = NC  exp[0.4/kT] = 6  1012 cm-3.

p = ni2/n = 1.7  107 cm-3.

(c) At point A: Net charge = Nd+  n  1.3  1016 cm-3.

At point B: Net charge = Nd+  n  1.3  1016 cm-3.

2.9
(a)
N 
Eq. 2.20, Ec  E f  kT ln  c .
 Nd 
Table 2.1, Nc = 2.9 × 1019 cm-3
Therefore, Ec  E f  0.266 eV

(b)

 si kT
Eq. 2.67 LD   0.13  m
q2 N d

(c)
ρnεsi is the dielectric relaxation time.
-3
From Fig. 2.10, ρn = 4.3 -cm for n-type, 1015 cm density. So ρnεsi = 4.5 ps.

SOLUTION 6
(d)

 2 kT 
3/2
Eq. (2.10) Nc  2g mt2ml  2   T 3/ 2
 h 
Nc(100C) = (373/300)3/2 Nc(300 K) = 4.0 × 1019 cm-3

N 
Therefore, Ec  E f  kT ln  c   0.274 eV at 100C.
 Nd 

2.10
(a)
E f  Ei  0.2 eV
( E f  Ei ) / kT
Eq. (2.14), n  ni e  2.26  1013 cm3
( Ei  E f )/ kT
Eq. (2.15), p  ni e  4.43  106 cm3

Ionized donor density Nd+ = 1016 cm-3.

Therefore, not charge neutral, net positive charge.

(b)
See above.
(c)
The one in the middle so electrons fall toward positive charge in the center.

SOLUTION 7
 Solutions to Chapter 3 Exercises

3.1 Equation (3.138) gives

1
Mp
W

0  W

0  0
x

 exp   ( p   n )dx    n exp   ( p   n )dx  dx .  (1)

As Mp  , 1/Mp  0, the above equation gives

0  exp   ( p   n )dx     n exp   ( p   n )dx  dx

W W x
(2)
 0  0  0 


 exp ( p   n )W   n
( p   n )

exp ( p   n )W   1 . 
Therefore,  n   p exp ( p   n )W  , (4)

or   
W  ln  n /  p /  n   p . 
Similarly, Eq. (3.139) gives the same expression for W as Mn  .

x
3.2 Let u( x )    f ( x ' )dx ' . (1)
0

du( x )
Therefore   f (x) (2)
dx
deu( x )
and   f ( x )e u ( x ) . (3)
dx
Using (3), we have

f ( x )exp   f ( x ' )dx ' dx   f ( x )eu( x ) dx

W x W
0  0  0
(4)

W deu( x )
=  dx  1  eu(W ) .
0 dx

f ( x )exp   f ( x ' )dx ' dx  1  exp   f ( x ' )dx ' .

W x W
Therefore, 0  0   0 
(5)

A similar procedure can be used to show that

f ( x )exp   f ( x ' )dx ' dx  1  exp   f ( x ' )dx '

W W W
0  x   0 
(6)

W
by letting u( x )    f ( x ' )dx ' .
x

SOLUTION 8
3.3 From Eqs. (3.10), (3.11), and (3.13), we have
2 si ( Na  Nd ) m
Wd  ,
qNa Nd

 m   bi  Vapp ,
dQd (p - side)  si
and Cd   .
dVapp Wd

2( N a  N d )
Therefore,
1
  bi  Vapp  . (1)
Cd2
 si qN a N d

A plot of 1/ Cd2 versus Vapp gives a straight line, which intercept the horizontal Vapp axis at Vapp

= bi and has a slope of 2(Na + Nd)/siqNaNd.

Now, from Eq. (3.3), we have
n p  N N 
q bi  kT ln  n 0 2 p 0   kT ln  d 2 a  . (2)
 ni   ni 

Therefore, from the plot of 1/ Cd2 versus Vapp plot, we have

 2( N a  N d )
Slope  , (3)
 si qN a N d

1 kT  N a N d  2( N a  N d )
and  ln  . (4)
C (Vapp  0)
2
d q  ni2   si qN a N d
The two equations (3) and (4) can be solved for the two unknowns Na and Nd.

1/Cd2

Vapp
0  bi

SOLUTION 9
3.4

Na (x)

x
0 W(V)
n+ region p region

For the one-sided n+p diode, the incremental increase in the magnitude of the depletion-layer
charge dQd on the p-side as a result of an incremental increase in the applied voltage dV is
dQd  qN a (W )dW , (1)
where the depletion-layer width W is a function of V. The depletion-layer capacitance is
dQd dW
C  qN a (W ) , (2)
dV dV
which gives
C
N a (W )  . (3)
q dW / dV 

 si
From C ,
W
dW  dC  si C d (1 / C 2 )
we have   si2  . (4)
dV C dV 2 dV
Substituting (4) into (3), we have
2
N a (W )  .
q si [d (1 / C 2 ) / dV ]

3.5 Referring to Fig. 3.6 and Eq. (3.20), we have

Em (Wd  d )
 m   bi  V  , (1)
2
where V is the applied voltage. From Gauss’s law we have

SOLUTION 10
 Qd   siEm . (2)
Therefore,

1 dV dEm (Wd  d ) / 2 1  dWd 

   (Wd  d )  Em
dEm 
. (3)
Cd dQd d siEm  2 si 

Now, from Eq. (3.19), we have

dWd d ( xn  x p ) dx dx
Em  Em = Em n  Em p  ( xn  d )  x p . (4)
dEm dEm dEm dEm

Substituting (4) into (3), we have

1 Wd
 .
Cd  si

3.6 The band-to-band tunneling can be represented as tunneling through a triangular barrier
of height Eg, slope qE and tunneling distance Eg / qE. This is illustrated in the figure below.

Energy

Ec
slope = qE
Ev
Eg

0 W x
Ec

Ev
p-region n-region

The WKB tunneling coefficient is, from Eq. (3.127),

  4 W 
T ( E )  exp 
 h

0
2m*  q ( x)  E dx

(1)

with W = Eg / qE. For the coordinates indicated in the figure, we have

 q ( x)  Eg  qxE . (2)

Also, for the tunneling electron being considered, E = 0. Substituting into (1), we have
  4 W 
T ( E  0)  exp 
 h

0
2m* Eg  qxE dx

(3)

SOLUTION 11
   4 2m* E 3 / 2 
 exp  g
.
 3qE 

3.7
(a)
p-type
Ec

Ef
Ev

2V

Ec
Ef

Ev
n-type

kT  N a N d 
(b)  bi  ln   0.82 V
q  ni 2 


2 si ( N a  N d )( bi  VR )
Wd   0.27 m
qN a N d

qN a (Wd / 2)
(c) Emax   2.1105 V/cm
 si

3.8
(a)
Wd

Ec
Ei  bi
Ef Ec E f
Ev

Ev
p-type n+
x=0

SOLUTION 12
 N 
Ei  E f  kT ln a   0.38 eV
 ni 

Eg
(b)  bi   0.38 eV = 0.94 eV
2q

2 si bi
 0.20 m = 210 cm
5
(c) Wd 
qN a

N 
(d) E f  Ev  kT ln v  increases as temperature increases.
 Na 
So Ei  E f decreases as temperature increases.

And  bi decreases as temperature increases.

3.9 Using Eq. (3.45), Eq. (3.47) can be rearranged to give

 [n p (0)  n p 0 ]
  e q[ i (0) n (0)] kT  e q[ i (0) n ( Wd )] kT
ni
(1)
n p (0) N
  d e  q i kT
.
ni ni
Equation (3.43) gives
ni2 qVapp np0
e q i  
kT kT qVapp kT
e e . (2)
Nd Na Nd

Therefore, Eq. (1) can be reduced to


qVapp kT
n p (0) e
 . (3)
np0 1 

Using Eqs. (3.45), (2) and (3), we have

kT  N d 
n (Wd )  n (0)   i  ln  
q  n p (0) 
(4)
kT 1   e app 
 qV kT

 ln  .
q  1  

SOLUTION 13
3.10
xm
0 x

q (x) q

image(x)
q Bn q field

Ec
Ef
Total

Ev
Metal Silicon

Referring to the figure, the electrostatic potential associated with the image force is
q
 image ( x)  , (1)
16 0 x

and the electrostatic potential associated with the electric field in the silicon is given by
d 2 field ( x) qN d
  . (2)
dx 2
 si
Integrating (2) twice, subject to the boundary condition that the electric field is zero at x = Wd,
where Wd is the depletion layer width, and the boundary condition that field(0) = 0, we have

qN d  x 2 
 field ( x)     Wd x  . (3)
 si  2 
Therefore, the total electrostatic potential energy is

 q qN d  x 2 
PE ( x)  q[ image ( x)   field ( x)]  q   Wd x   . (4)
16 si x  si  2 

The peak of PE(x) is given by dPE(x)/dx = 0, that is

q qN dWd  x 
 1  m   0 . (5)
16 x 2
si m  si  Wd 
We will first make the assumption that xm/Wd << 1, and then come back to verify that our
assumption is correct. With this assumption, (5) gives

SOLUTION 14
 q
xm  , (6)
16 siEm

 d field ( x) qN dWd
where Em   (7)
dx x 0
 si

is the absolute value of the electric field in the silicon at the interface. The energy barrier
lowering is
 q qN d  x 2 
q | PE ( xm ) | q   Wd xm  m  (8)
16 si xm  si  2 

 q qN dWd xm  q 3Em
 q   .
16 si xm  si  4 si

To show that xm/Wd << 1, we note that Wd is related to the total band bending. For
simplicity we assume there is no external applied voltage, so that

2 si bi
Wd  , (9)
qN d

where bi is the built-in potential. For bi = 0.4 V and Nd = 11016 cm3, we have Wd =
2.3105 cm and xm = 3107 cm. That is, xm/Wd << 1 as assumed.

3.11 The multiplication factor for hole-initiated impact ionization is

1
Mp 
W

W x
 
 exp    p   n dx    n exp    p   n dx dx, (1)
0 0 0    
and that for electron-initiated impact ionization is
1
Mn W

W

W

 exp    n   p dx    p exp    n   p dx dx. (2)
0 0 x    
From Ex. 3.2, we have

0
W
 0
x
 
f  x  exp   f  x  dx dx  1  exp   f  x  dx
W

0  (3)

and

0
W
 W

x  
f  x  exp   f  x  dx dx  1  exp   f  x  dx .
0
W
 (4)

SOLUTION 15
Applying Eq. (3) to Eq. (1), we obtain

1
1
Mp 0 0
x

   p exp    p   n  dx dx.
W
 (5)

Similarly, applying Eq. (4) to Eq. (2), we have

1
1
Mn 0
W

x 
   n exp    n   p  dx dx.
W
 (6)

Avalanche breakdown occurs when carrier multiplication by impact ionization runs away, i.e.,
when the multiplication factor becomes infinite. Thus, Eq. (5) gives the hole-initiated breakdown
condition as

0
W

 p exp    p   n  dx dx  1.
x

0  (7)

Using Eq. (3) the left-hand side of Eq. (7) can be rearranged to give

0
W
 x

0  
 p exp    p   n  dx dx  1  exp    p   n  dx
W

0 
 
(8)
   n exp    p   n  dx dx.
W x

0 0

Substituting Eq. (7) into Eq. (8) gives

0
W

 n exp    p   n  dx dx
0
x

 exp       dx  .
(9)

W

0 p n

Dividing both sides of Eq. (9) by its RHS gives

0
W

 n exp  
x
W
 n 
  p  dx dx  1, (10)

which, according to Eq. (6), is simply the condition for electron-initiated breakdown. Thus, the
condition for avalanche breakdown is the same whether the breakdown process is initiated by
electrons or by holes.

SOLUTION 16
3.12

On p-side, Ei (xp) – Ef = kT ln(Na /ni) = 0.400 eV,

on n-side, Ef – Ei (xn) = kT ln(Nd /ni) = 0.417 eV. So bi = m = 0.817 V.

2 m 2 si m ( N a  N d )
Wd  xn  x p    1.782  105 cm
Em qN a N d

xn = [Na /(Na + Nd)]Wd = (1/3)Wd, xp = Wd  xn = (2/3)Wd,

Ei (0) – Ei (xn) = (1/2) xn  Em = (1/2)(Wd /3)(2m /Wd) = m /3 = 0.272 V.
At x = 0, Ei – Ef = [Ei (0) – Ei (xn)] – [Ef – Ei (xn)] = 0.272 – 0.417 = –0.145 eV.

Ei crosses Ef on the p-side when Ei (xp) – Ei (x) = Ei (xp) – Ef = 0.400 eV.

Since Ei (xp) – Ei (x)  (x + xp)2 and Ei (xp) – Ei (0) = (1/2) xp  Em = 2m /3 = 0.545 V,
solve x from (x + xp)2/xp2 = 0.400/0.545  x = –0.170105 cm.

SOLUTION 17
3.13

depletion
region

n-type (Nd p-type (Na

=1e18 cm-3) =1e17 cm-3)

x
0 10m

At x = 0, J n (0)   qDn ni2 / ( N a Ln )  exp qV / kT = 0.44 A/cm2.

Hole current equals the diffusion current on the n-side, i.e.,

J p (0)   qD p ni2 / ( N d L p )  exp qV / kT = 0.088 A/cm2.

At x = 10 m,

J n (10 m)  J n (0)e  x / Ln = 0.27 A/cm2.

The total current is continuous, so

J p (10 m)  J n (0)  J p (0)  J n (10 m) = 0.26 A/cm2.

3.14

(a) Built-in potential  bi  Eg / 2q  (kT / q) ln( N a / ni ) = 0.98 V.

Depletion layer width Wd   2 si ( bi  0.6) / qN a 

1/2
= 70 nm.

Depletion capacitance = si /Wd = 0.148 F/cm2.

(b) The total current is dominated by the electron diffusion current:

J n ( x  0)  [qDn ni2 / N a Ln ]exp( qVapp / kT ) = 9.2 mA/cm2.

(c) J n ( x)  J n ( x  0) exp( x / Ln ) = 5.6 mA/cm2.

SOLUTION 18
3.15

kT  N a N d 
(a)  bi  ln   0.804 V
q  ni 2 


2 si ( bi  0.5)( N a  N d )
Wd   1.31  10 5 cm
qN a N d

Nd
Qd ( p  side)  qN a x p  qN aWd  4.8  108 C/cm 2 , negative
Nd  Na

 si
(b) Cd   7.9  108 F/cm2
Wd

3.16

Eg kT  N a 
(a)  bi   ln   0.56  0.42  0.98 V
2q q  ni 

2 si ( bi  VF )
Wd   7  10 6 cm
qN a

qN aWd
Emax   1.1  105 V/cm
 si

2
ni qV F / kT
(b) np  e  1.15  1013 cm -3
Na

SOLUTION 19
3.17

(a)
p-type n-type

Ec Ec
Ef
Ef 0.5 V
Ev
Ev

kT  N a N d 
(b)  bi  ln   0.90 V
q  ni 2 


2 si ( N a  N d )( bi  0.5)
Wd   0.075 m
qN a N d

qN aWd N d /( N a  N d )
Emax   1.1105 V/cm
 si

2
ni qV / kT
(c) np  e  4.41  1011 /cm3
Na

3.18

(a)
kT  N a N d 
 bi  ln   0.70 V
q  ni 2 


(b)

2 si ( N a  N d ) bi
Wd   0.522 m
qN a N d

Na
xn  Wd  0.17 m
Na  Nd

Nd
xp  Wd  0.35 m
Na  Nd
SOLUTION 20
(c)
2 bi
Em   2.68  104 V/cm
Wd

SOLUTION 21
 Solutions to Chapter 4 Exercises

4.1. (a) Using the integral expressions of Qd and Qi, one has

dQd qN a (1  e q s / kT )
Cd    ,
d s Es

and

dQ q( ni / N a )( eq s / kT  1)
2
Ci   i  .
d s Es

Here Es = E ( = s).

(b) Take the square of Eq. (4.16) and differentiate with respect to s, we obtain

 
   
2
dQs n
2Qs  2 si qN a  1  e q s / kT  i 2 eq s / kT  1  .
d s  Na 

Using Qs =siE s (Gauss’s law) and the above two equations in (a), it is straightforward to show
that

dQs qN a  
   
2
 q s / kT ni q s / kT
Csi     1  e  2 e  1   Cd  Ci .
d s Es  Na 

(c) When s = 2B, exp(qs/kT) = (Na/ni)2 >> 1 and exp(qs/kT) << 1. From (a), one
has

qN a
Ci  Cd  .
Es

(d) It is clear from Fig. 4.7 that Qs and therefore Es takes off rapidly beyond strong
inversion. This means that Cd  E s1 decreases rapidly beyond strong inversion. We say that
the depletion layer (charge) is “screened” by the inversion layer. Note that Ci, on the other
hand, increases rapidly beyond strong inversion because of the exp(qs/kT) factor.

4.2
 ox
(a) Cox   Cmax  3.4  10 7 F/cm2  tox = 10 nm.
tox

SOLUTION 22
 CoxCd
(b) Cmin   7.3  108 F/cm2  Cd = 9.410-8 F/cm2
Cox  Cd

 si 2 si (2 B )
Cd   Wdm   1.1  10 5 cm
Wdm qN a

Take 2B  0.9 V  Na = 9.61016 cm-3, p-type.

1 1 L 1 kT
(c)   D    Cfb = 2.410-7 F/cm2
C fb Cox  si Cox  si q 2 N a

From C-V, Vfb = 0.4 V.

(d)

s    Eg / 2q   B  5.03 V  m  s  V fb  4.63 V

4.3. Eq. (4.33),

 si qN a  si qN a / 2
 s  Vg  V fb  2
 .
2Cox Cox

Solve for Vg,

2
 si qN a   qN / 2 
Vg  V fb     s  si a  .
2Cox 2  Cox 

Take a differential,

  qN / 2  1  qN / 2 s
Vg /  s  2   s  si a   1  si a .
 Cox  2  s Cox

This is simply s = Vg/(1 + Cd/Cox), where Cd is the depletion charge capacitance given by Eq.
(4.38).

4.4. By keeping only the dominant inversion charge term in Eq. (4.16) and substituting it in
Eq. (4.7), one obtains

SOLUTION 23
 2 si kTN a  ni  q s / 2 kT
Vg  V fb   s   e .
Cox  Na 

When Vg increases beyond the point where s = 2B, most of that increase appears as inversion
charge in the third term. There is very little change in the band bending beyond s = 2B. A
good approximate solution for s is then obtained by letting s = 2B for the second term and
solving s from the third term:

2kT  Cox (Vg  V fb  2 B ) 

 s  2 B  ln
  .
q  2 si kTN a 

Here exp(qB/kT) = Na/ni has been applied.

It is clear from the above equation that s  2B is a weak function of Vg. As an
example, for Na = 1017 cm3, tox = 100 Å, and Vg  Vfb = 4 V, one has 2B = 0.82 V, and s  2B
 7.2kT/q  0.19 V.

4.5. In Fig. 4.13(b), the total capacitance seen by the gate equals Cox and (Ci + Cd) connected
in series, i.e.,

d ( Qs ) Cox ( Ci  Cd )
C  .
dVg Cox  Ci  Cd

Since Qs = Qi + Qd and Qi/Qd = Ci/Cd, it follows that

d ( Qi ) Cox Ci
 .
dVg Cox  Ci  Cd

When Vg is below the threshold of strong inversion, Ci << Cox, therefore d(Qi)/dVg 
Ci/(1 + Cd/Cox) << Cox. Above strong inversion, Ci >> Cox (> Cd) (see Exercise 4.1), and one
has d(Qi)/dVg  Cox. Such a behavior is sketched schematically below.

d(-Qi ) Cox
dVg

Vg

SOLUTION 24
The sharp transition takes place at the threshold voltage. Qi(Vg) is simply the integrated area
under the above curve. Beyond threshold, Qi increases linearly with Vg with a slope equal to
Cox (see Fig. 5.4 in the text).

4.6.
 Eg   kT N d 
V fb    B    0.56  ln   (0.56  0.35)  0.21 V.
 2q   q ni 

Vfb Vg

Band diagram at Vg = 0:

Ef Ef

n+ n-type

At s = 2B = 0.7 V:

Qd 2 si qN d s 4.8 108
Vox     0.14 V.
Cox Cox Cox

4.7.

SOLUTION 25
 kT N a
(a) B  ln  0.42 V
q ni
Eg
V fb    B  0.98 V
2q
2 si qN a (2 B )
Vt  V fb  2 B   0.98  0.84  0.97  0.83 V.
Cox

(b)
C
Cmax

Cmin
Vfb Vt Vg

(c) Cmax  Cox  1.73 107 F/cm2

2 si (2 B )
Wdm   0.104 m
qN a

1 1 Wdm
 
Cmin Cox  si

Cmin  0.63 107 F/cm2

4.8.
Let the surface potential at zero gate voltage be s.

2 si qN a s
Then  V fb   s 
Cox

Solve the above quadratic equation in s :

SOLUTION 26
 2 si qN a 2 si qN a
s    2
 V fb
2Cox 4Cox

2 si qN a
2
 0.278 V,
4Cox

Therefore, s = 0.353 V

2
ni q s / kT
Electron density at the surface: n  e  8.3 108 cm-3
Na

hole density at the surface: p  N a e  q s / kT  1.2 1011 cm-3

4.9
(a) s = 0.5 V, Qd = (2si q Na s)1/2 = 4.110-8 C/cm2

Cox = ox/tox = 3.4510-7 F/cm2

Vg = s + Qd/Cox = 0.62 V.

(b) W d = (2si s/ q Na)1/2 = 0.25 m

si/W d = 4.110-8 F/cm2

C = Cox(si/W d)/[Cox + si/W d] = 3.6610-8 F/cm2.

(c) Vg = 2 V  inversion


s (max) = 2 B = 2 (kT/q) ln (Na/ni) = 0.72 V

Vox = Vg – s(max) = 1.28 V.

(d) Vg = -2 V  accumulation

In the bulk, Ef is (Eg/2q) – B = 0.2 V above Ev.

This is the most upward band bending at the surface,

i.e., s (min) = 0.2 V so Vox = Vg  s(min) = 1.8 V.

SOLUTION 27
4.10.
kT  N a 
B  ln   0.4083(V )
q  ni 
Qs  Cox (Vg  V fb  2 B )  1.09  106 (C / cm 2 )
 Qs / q
x poly   0.680(nm)
N poly

qN poly x 2poly
Since p   35.5(mV ) , the band-bending in the gate can be neglected to the first
2 si
order in the charge calculation.

4.11.
Electron energy level under triangular approximation, the electron energy levels are given by
2/3
 3hqEs  3 
Ej    j   where j  0,1,2, 
 4 2mx  4 

kT  N a 
B  ln   0.4679(V )
q  ni 
2 si
Wd  (2 B )  34.9(nm)
qN a
q Wd qN aWd
Es 
 si 0
N ( x)dx 
 si
 0.537( MV cm)

Substituting Es and j gives,

E0  0.1143(eV ) , E1  0.2010(eV ) , E2  0.2717(eV ) , …

4.12. We will do this by showing that

  
nT ( t )  N T 1  exp  N inj ( t ) , (1)

jG ( t  ) t
where N inj ( t )   dt  , (2)
q 0

is a solution to the differential equation. From (1), we have

 
dnT dN inj ( t )
 N T exp N inj ( t ) . (3)
dt dt

SOLUTION 28
From (2), we have
dN inj ( t ) jG ( t )
 . (4)
dt q

Substituting (4) into (3) and using (1), we have

dnT j
 G  ( N T  nT ) .
dt q

1E+13
NT = 5E12 cm-3
 = 1E-13 cm2
nT (cm-3)

1E+12

1E+11

2E+10
1E+11 1E+12 1E+13 5E+13
Ninj (cm-2)

SOLUTION 29
 Solutions to Chapter 5 Exercises

5.1.
kT N a
(a) B  ln  0.42 V
q ni
Eg
V fb    B  0.98 V
2q
2 si qN a (2 B )
Vt  V fb  2 B   0.98  0.84  0.97  0.83 V.
Cox

Ec

Ei
Ef
Ev
Ef

p-type silicon

Oxide

SOLUTION 30
(b)

2 si qN a (2 B  Vbs )
Vt  V fb  2 B   0.98  0.84  1.44  1V.30
Cox

Ec

Ei
Ef
Ev

Efn

Ef

p-type silicon

Oxide

5.2. Eq. (5.14) is an implicit equation relating s and V. Taking a differential yields

1/ 2
2 si kTN a  q s ni 2 q( s V )/ kT  q 
2
ni q( s V )/ kT 
0  d s    2e   s
d  2
e ( d s  dV ) .
2Cox  kT N a  kT  Na 

It is straightforward to re-group the terms and show that

d s ( ni / N a )eq s / kT
2 2

dV 1  ( ni 2 / N a 2 )eq s / kT  ( Cox 2 /  si qN a )( Qs / Cox )

in the limit of V  0, where

1/ 2
 q n
2

Qs  2 s kTN a  s  i 2 eq s / kT  .
 kT N a 

SOLUTION 31
5.3. Applying 1-D Gauss’s law to a box bounded by a plane at depth x from the surface and
another plane deep in the substrate where E = 0, one can write

1 
 Qd  qx n( x)dx  ,
xi
E( x ) 
 si  

where Qd is the depletion charge density and xi is the inversion layer depth. Therefore,

Qd q
   n( x ) n( x )dx dx .
xi xi xi xi
n( x )E ( x )dx  n( x )dx 
0  si 0  si 0 x

If we let n(x) = df(x)/dx, the second term on the RHS becomes

 1 xi d ( f 2 ) 
 f ( xi )  f ( x )dx 
q df q x i df

 si 0 i 0
2 0 dx
xi
 f ( x ) dx  dx  ,
dx  si  dx 

which can be further simplified to

q  
 
2
q  xi 
 f ( x ) f ( xi )  f ( 0)  f ( xi )  f ( 0)    f ( xi )  f (0)  0 n( x )dx  .
1 2 2 q 2

 si  i 2  2 si 2 si

Substituting it back, one has


xi
n( x )E ( x )dx Qd Qi
Eeff  0
  ,
 si 2 si

xi
n( x )dx
0

where

Qi  q  n( x )dx .
xi

5.4. This exercise is closely related to Exercise 4.5, in which Eq. (5.25) is derived:

d ( Qi ) Cox Ci
 .
dVg Cox  Ci  Cd

When Vg is below the threshold of strong inversion, d(Qi)/dVg  Ci/(1 + Cd/Cox) << Cox.
Above strong inversion, Ci >> Cox (> Cd), and one has d(Qi)/dVg  Cox. This behavior is
sketched below.

SOLUTION 32
 d(-Qi ) Cox
dVg

Vg

Qi(Vg) is simply the integrated area under the curve. Beyond threshold, Qi increases linearly
with Vg with a slope equal to Cox (see Fig. 5.4 in the text).

The sharp transition above can be used to define a kind of inversion charge threshold
voltage, Vtinv, where Ci = Cox. Since Cd  0 due to screening by inversion charge (Exercise
4.1), dQi/dVg  CoxCi/(Cox+Ci) = Cox/2 at Vg = Vtinv. In the region where Boltzmann
approximation is valid, Qs  Qi  exp(qs/2kT), hence Ci  (q/2kT)Qi and Qi  (2kT/q)Cox. Such
a threshold voltage is slightly higher than the “2B” threshold where Ci = Cd.

5.5

kT  N a  Cox 2 (Vgs  V fb   s ) 2 q s  
2
The implicit eq. for s,s and s,d is V  s  ln  2   
q  ni  2 si kTN a kT  
 
with V = 0 for s,s and V = Vds for s,d.

For s,s,

kT  N  C (V  V fb   ss ) q ss   kT  Cox (Vgs  V fb   ss ) q ss 
2 2

2 2 2

 ss  ln a2  ox gs    2  ln   
2 si kTN a 2 si kTN a
B
q  ni  kT   q  kT 
 
The s,s in the square bracket of ln can be approximated by 2B = 2 (kT/q) ln(Na/ni) = 0.87 V.

Cox = 4.910-7 F/cm2, therefore s,s = 0.87 + 0.0259ln[178.6  33.6] = 0.87 + 0.13 = 1.0 V.

For s,d in saturation, pinch-off or Qi  0, Vds  ,

Cox (Vgs  V fb   sd ) 2 q sd
2

 0
2 si kTN a kT

SOLUTION 33
The solution is
1 2 si qN a 2 si qN a 
 sd    2
 4(Vgs  V fb ) 
2  Cox Cox 

2siqNa/Cox2 = 0.28 V, so s,d = 1.38 V.

5.6
kT N a
B  ln  0.435 V
q ni

2 si (2 B )
Wdm   0.075 m
qN a

m = 1 + 3tox/Wdm = 1.4

Inverse SS = m60 mV/dec = 84 mV/decade

Qd  qN aWdm  2.4  107 C/cm2

Qi  Qs  Qd  7.6  107 C/cm2

The effective field is  Qd  Qi / 2 /  si  6.0  105 V/cm

From the universal mobility curve in Fig. 5.14, eff  400 cm2/V-s

5.7
Integrating from 0 to y after multiplying both sides of Eq. (5.8) by dy yields
V
I ds y  eff W  [Qi (V )]dV
0

 m 
 eff CoxW (Vgs  Vt )V  V 2  ,
 2 
where the simplified Qi(V), Eq. (5.32), has been used.
Substituting Ids from Eq. (5.27) into the equation above, one can solve for V(y):

Vgs  Vt  Vgs  Vt  y  Vgs  Vt 

2
y 2
V ( y)     2   Vds  Vds .
m  m  L m  L

SOLUTION 34
5.8
A more general form of the saturation voltage is obtained by letting Qi = 0 in Eq. (5.16) with ψs =
2ψB + V and solving for V = Vdsat [equivalent to solving dIds/dVds = 0 by differentiating Eq.
(5.22)].

Charge Sheet model, Qi  Qs  Qd  Cox (Vgs  V fb  s )  2 si qN a s

Pinchoff happens when Qi = 0. Solve the quadratic eq. for s1/2:

 si qN a  si qN a
 s, max  Vgs  V fb  2
 2
2Cox 2Cox

 si qN a  si qN a 2(Vgs  V fb )Cox
2

 s, max  Vgs  V fb   1
Cox
2
Cox
2
 si qN a

 si qN a  si qN a 2(Vgs  V fb )Cox 2
Vdsat  Vgs  V fb  2 B   1 .
Cox2 Cox 2  si qN a
It turns out that Eq. (5.27) serves as a good approximation to the drain current over a much wider
range of voltages than expected. Even for a drain voltage several times greater than 2ψB, the
current Idsat is only slightly (≈5%) underestimated.

5.9. Let V = 0 in Eq. (5.12):

eq / kT
2
qni s
Qi 
Na  B E ( ) d .

In the subthreshold region, both the band bending  and the electric field E are mainly set by the
depletion charge. Because of the exponential factor in , the above integration of inversion
charge is dominated by that over a thin surface layer within which E() is essentially constant and
can be taken outside the integral. Thus one obtains
2
kTni q s / kT
Qi  e ,
Es N a

where Es is the surface electric field. Note that exp(qB/kT) << exp(qs/kT) in weak inversion.

SOLUTION 35
 This equation is generally valid for nonuniform (vertically) dopings with Na being the p-
type concentration at the edge of the depletion layer. (Note that the factor Na merely reflects the
fact in Fig. 4.6 that the band bending s is defined with respect to the bands of the neutral bulk
region of doping Na.)

5.10. From Eq. (5.8),

dV
I ds ( y )  eff WQi ( y ) .
dy

Current continuity requires that Ids be a constant, independent of y. For a short-channel or a

nonuniformly-doped (laterally) MOSFET,

dy  W Vds
 
L V ds
 eff dV  eff W .
0 Qi ( y ) I ds 0 I ds

Generalizing the result of the last exercise to laterally varying s and Es, one obtains

Vds 1 dy Na
 2 0 s
L L
  E ( y )e q s ( y )/ kT dy .
I ds eff W 0 Qi ( y ) eff WkTni

Since Es(y)  Vg  Vfb  s(y)/3tox is not a strong function of s, the exponential factor
dominates. This implies that the subthreshold current is controlled by the point of highest
barrier (lowest s) in the channel.

5.11 Charge Sheet model, Qi  Qs  Qd  Cox (Vgs  V fb  s )  2 si qN a s

Pinchoff happens when Qi = 0. Solve the quadratic eq. for s1/2:

 si qN a  si qN a
 s, max  Vgs  V fb  2
 2
2Cox 2Cox

 si qN a  si qN a 2(Vgs  V fb )Cox
2

So,  s, max  Vgs  V fb   1

Cox
2
Cox
2
 si qN a

SOLUTION 36
 Solutions to Chapter 6 Exercises

6.1 For the MOSFET scale length given by Eq. (6.23), show that
(a) λ ≥ Wdm and λ ≥ tox, i.e., λ is larger than the larger of Wdm, tox.
(b) In the special case of εox = εsi, λ = Wdm + tox.
(c) If tox ≪ Wdm (lower right corner of Fig. 6.7), λ ≈ Wdm + (εsi/εox)tox.
(a)
1 1
Let k = / f (k )  tan  ktox   tan  kWdm 
 ox  si
Assume Wdm > tox with no loss of generality.
As k goes up from 0 to [/(2Wdm)], f(k) stays positive. Once k = [/(2Wdm)]+, the 2nd tangent is
–∞ so f(k) becomes negative.
As k increases to [/Wdm], the 2nd tangent goes to zero.
If 2tox < Wdm: The 1st tangent is positive  f(k) changed sign between k = [/(2Wdm)]+ and
/Wdm  There is a  solution  (Wdm, 2Wdm).
If 2tox > Wdm: The 1st tangent goes to +∞ at k = [/(2tox)] < [/Wdm]  f(k) changed
sign between k = [/(2Wdm)]+ and /Wdm.
 In either case, a solution of  exists > Wdm, the larger of (Wdm, tox).
(b)
 πt   πWdm 
εox = εsi, tan  ox  + tan   = 0.
 λ   λ 
So tan[tox/ + Wdm/] = 0 which means [tox/ + Wdm/] = , 2, 3, etc.
The largest is λ = Wdm + tox.
(c)
If tox ≪ Wdm, the first angle  0, the second angle  
1 1 1  tox 1   Wdm 
tan  tox /    tan  Wdm /       0
 
So
 ox  si  ox   si 
which gives λ ≈ Wdm + (εsi/εox)tox.

6.2.

SOLUTION 37
 -9
x 10
1

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-1
0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
-8
x 10

f (n )   si tan(ti / n )   i tan(Wd / n )

From the figure above, 1 ~ 15.9(nm) , 2 ~ 7.3(nm) , 1 ~ 5.0(nm) .

exp L 21 
By substituting the numbers, ~ 40.14 , when L ~ 21 .
exp L 22 

6.3 Show that Eq. (6.38) for short-channel subthreshold current is reduced to Eq. (5.36) of long-
channel subthreshold current if 2(x, y) is taken to be the long channel potential v2(x) of Eq.
(6.10). [Hint: Convert dx to d/E and approximate E by Es, the field at the surface.]

ni2 1  e  qVds / kT
Eq. 6.38: I ds  eff kTW .
L dy

Na
e q 2  x , y / kT dx
Wdm

0
0

The integral over x can be approximated as

Wdm q 2  x , y  / kT s q 2  x , y  / kT d 2 1 kT q s / kT
 0
e dx   e
0 E

Es q
e

For long channel, there is no y-dependence. So the y-integral is just multiplication of L.

2
 x 
With Eq. 6.10 for long channel,  2 ( x)   s 1  
 Wd 

2 s 2qN a s
Es  
Wd  si
2 2
 si qN a  kT   ni  q
Therefore, I ds  eff
W
L
  
2 s  q   N a 
 e
s / kT
1  e  qVds / kT
.

SOLUTION 38
6.4
Under n = 1 velocity saturation, Eq. 6.36 becomes
eff dV / dy
I ds  W ( Qi ) .
1  ( eff / vsat )dV / dy
Solve for dV/dy:
dV I ds / eff
 ,
dy WQi  I ds / vsat

With -Qi given by Eq. 6.37,

ni2  qV  Wdm  q 2 ( x, y ) 
Qi ( y )  q exp    0 exp   dx,
Na  kT   kT 
We have
dV I ds / eff
 .
dy qW ( n 2 / N )  Wd

a 
q 2 ( x , y )/ kT  qV / kT
e dx e  I ds / vsat
i
 0 

6.5.
(a) I ds 

eff Cox (W / L ) (Vg  Vt )Vds  ( m / 2)Vds 2  (m = 1)
1  ( eff Vds / vsat L )
8
Vg -Vt =5.0 V
Drain current (mA)

6
4.0 V

4 3.0 V

2.0 V
2
1.0 V
0
0 1 2 3 4 5
Drain voltage (V)

(b) 1  2eff (Vg  Vt ) / ( mvsat L )  1

I dsat  CoxWvsat (Vg  Vt ) (m = 1)
1  2eff (Vg  Vt ) / ( mvsat L )  1

SOLUTION 39
 15
Vg -Vt =3.0 V

Saturation current (mA)

no velocity saturation

10

velocity saturation
0
0 1 2 3 4 5
Channel length (um)

6.6.
2 eff (Vgs  Vt )
Make use of z  , Eq. (6.49) becomes
mvsat L
mvsat L 1  z  1 mv L
 
2
I dsat  CinvWvsat z  CinvWvsat sat 1 z 1 .
2 eff 1 z 1 2 eff
Differentiating the above with respect to z, one obtains

dI dsat dI dsat 2 eff  

gmsat 
dVgs

dz mvsat L
 CoxWvsat 2  1 z 1 2 1
1 z
 CoxWvsat 1 

1

1  2 eff (Vg  Vt ) / mvsat L 

When L  0, gmsat  CoxWvsat (Eq. (6.79)).

When vsat  , the square root factor can be expanded in a Taylor series to yield

   (V  Vt )  W Vg  Vt
gmsat  CoxWvsat 1  1  eff g   eff Cox ,
  mvsat L  L m

which can also be derived from the long-channel current expression, Eq. (6.50).

6.7. Since Ids = WQiv and Qi = Cox(Vg  Vt) at the source, one has

1  2 eff (Vg  Vt ) / ( mvsat L )  1

v  vsat
1  2 eff (Vg  Vt ) / ( mvsat L )  1

SOLUTION 40
for the carrier velocity at the source end of the channel.

When L  0, v  vsat. When vsat  , the denominator approaches 2 and the square-
root expression in the numerator can be expaned in a Taylor series to yield

 eff (Vg  Vt )
v ,
2mL

which also follows from Ids = WQiv = WCox(Vg  Vt)v and the long-channel current, Eq. (6.50).

6.8. By definition, the total integrated inversion charge under the gate is

Qi ( total )   Qi ( y )Wdy .
L

Change variable from y to V:

Qi (V )WdV
Qi ( total )  
Vds
.
0 dV / dy

Substituting dV/dy from Eq. (3.74), one obtains

 eff W eff W 2 2 
Qi ( total )   
Vds
 Qi (V )  Qi (V ) dV .
0
 vsat I ds 

From Eq. (3.76), Qi(V) = Cox(Vg  Vt  mV), the above integral can be evaluated to yield

CoxWeff Cox W 2 eff

2

Qi ( total ) 
2mvsat
 (Vg  Vt )  (Vg  Vt  mVds ) 
2 2

3mI ds
 
(Vg  Vt )3  (Vg  Vt  mVds )3 . 
In the saturation region, Vds and Ids are given by Eqs. (6.48) and (6.49). It can be shown that the
above equation reduces to

1  2 eff (Vg  Vt ) / ( mvsat L )  (1 / 3)

Qi ( total )  WLCox (Vg  Vt ) .
1  2 eff (Vg  Vt ) / ( mvsat L )  1

qN a
6.9. (a) Vt  V fb  2 B  (Wdm  xs )
Cox

SOLUTION 41
 4 si B
Wdm   xs
2

qN a
Eliminate Na from the above eqs,
4 si B
xs   Wdm
Cox (Vt  V fb  2 B )
Vfb = -Eg/2q – B = -1.06 V, so xs = 17 nm.
Substituting in the 2nd eq., Na = 1.31017 cm-3.

(b) m = 1 + 3tox/Wdm = 1.21

Inv. Subth. Slope = m60mV/decade = 73 mV/decade

(c)  = Wdm + 3tox = 0.12 m

Lmin  2 = 0.24 m

6.10
The depletion charge density in silicon is
Qd = q Na (Wd – xs) = 6.410-8 C/cm2.
Surface field: Es = Qd/si = 61800 V/cm.
q qN a
Wd2  xs2   Es (Wd  xs )  0.37 V.
1
Wd
Surface potential: s 
 si  xs
xN a dx 
2 si 2

SOLUTION 42
 Solutions to Chapter 7 Exercises

7.1.

(a)
The expression for the drain current is given by
2 s
W 4 si  2kT    2  si ti 2 2 
I ds       tan     tan   . (1)
L tsi  q   2  itsi  d
The boundary conditions at the source and drain are:
q(Vgs  mi )  2 2 si kT  2 si ti
 ln    ln  s  ln  cos  s    tan  s (2)
2kT  tsi
2
q ni   i tsi s
q(Vgs  mi  Vds )  2 2 si kT  2 si ti
 ln    ln  d  ln  cos  d    tan  d . (3)
2kT  tsi
2
q ni   i tsi d
Here miV.
If the values of βs and βd are solved for given biases Vgs and Vds, the value of the current is
obtained by Eq. (1). The problem is then to solve the implicit equations (2) and (3) with given
boundary conditions.

1.0
Vg=0.0(V)
Vg=0.5(V)
0.8
Vg=1.0(V)
Vg=1.5(V)
0.6 Vg=2.0(V)
Ids (mA)

0.4

0.2

0.0
0.0 0.5 1.0 1.5 2.0
Vds (V)

Various numerical methods can be employed:

(i) Scan βs over 0   s   2 and use Eq. (2) to generate a look-up table mapping βs to Vgs. Pick
βs values that give Vgs = 0, 0.5, 1.0, 1.5, 2.0 V. Interpolation may be needed. For a given Vgs and
the corresponding βs, scan βd over (0, /2) and use Eq. (3) to calculate corresponding Vds. An Ids-
Vds curve can then be generated for each Vgs (βs).

(ii) The implicit equations can be solved by numerical methods such as Newton Raphson
method. Use of internal functions of mathematical tools is also acceptable. (e.g., fsolve or solve
of Matlab.)

SOLUTION 43
(b) tan(ti /  ) tan(tsi / 2 )   i /  si

-9
1.0x10

f=sitan(tox/))-oxcot(tsi/2)

-10
5.0x10

=19.80(nm)
0.0

-10
-5.0x10
5 10 15 20 25
 (nm)

By numerically plotting the function in the graph above, 1  19.80(nm) .

Therefore, the device can be scaled down to Lmin  21  39.6(nm) without severe SCEs.

7.2

2kT  tsi q 2 ni  2 x  
(a)  ( x)  V  ln  cos   ,
q  2 2 si kT  tsi  
Center potential at the source:

2kT  tsi q 2ni  2kT  2  2 si kT 

 (0)   ln   ln  
q  2  2 si kT  q  tsi q 2ni 

 cannot exceed /2, hence the maximum center potential is

2kT   2 si kT 
 (0)max  ln  
q  tsi q 2ni 

(b)

From the boundary condition,

q(Vgs  mi  V )  2 2 si kT  1  2 si   tox  1  
 ln    ln(1   )  ln     ln  1   .
  ox   R 
2
2kT  R q ni  2
At the source, V = 0. For Vgs >> threshold,  << 1.

SOLUTION 44
The potential solution is (V = 0)

kT  8 kT (1   ) 
 ( )  ln  2 2 si 2 2 
,
q  q ni R [1  (1   )(  / R ) ] 
At center, with  << 1,
kT  8 si kT 
 (   0)  ln  2 2  ,
q  q ni R 

7.3
(a) Let r  sitox/(oxtsi) and a  [2sikT/(q2ni)]1/2 to compact the expressions.

2kT  2  2 si kT  2kT
 ( x)  V  ln   lncos2 x / t si 
q  t si q 2 ni  q

2kT  2  2kT
At the source, V = 0. Surface potential:  (tsi / 2)  ln  a  lncos 
q  tsi  q
Condition for :
q(Vgs  mi  V ) 2 2 si kT  2 si tox
 ln    ln   ln cos     tan 
2kT  tsi
2
q ni   ox tsi

q(Vgs  mi )  2a 
 ln    ln   ln cos    2r  tan 
At the source,
2kT  tsi 
Vg >> threshold:   /2  tan term dominates.

q(Vgs  mi )  2a  r
 ln    2r  tan  
2kT  tsi  cos 
Then
2kT    2kT  q(Vgs  mi )  2a   2kT
 (tsi / 2)  ln  a  ln   ln     ln( r )
q  tsi  q  2 kT  t si   q

Simplification (optional):
2kT  a  q(Vgs  mi )  2a   
 (tsi / 2)  ln    ln   
q  rtsi  2kT  tsi   

(b)

From the boundary condition,

SOLUTION 45
  R q 2n 
q(Vgs  mi  V ) 1    si  tox 
 ln  i
  ln 1    ln   2 ln 1  
2kT  
 2 2 si kT    ox  R
At the source, V = 0. For Vgs >> threshold,  << 1, so the last term dominates and
q(Vgs  mi )  R  2  si  tox 
 ln    ln  1  
2kT  2a    ox  R
The potential solution is (V = 0)

2kT  R  (1   )  2  
 ( )   ln  1  
q  2a 1    R2 
At surface,
2kT  R  kT  4a 2  2kT  1 
 (   R)   ln     ln  2   ln  
q  2a  q  R  q  
Substituting the  solution from the above boundary condition:

2kT   q(Vgs  mi )  R   2a   2  t   

 (   R)  ln   ln     ln    ln  si ln  1  ox   
q   2kT  2a    R   ox  R   

7.4.

I ds  s I ds  d
gm  
 s Vgs  d Vgs

W Vds W d dV
I ds  
L 
0
[Qi (V )]dV  
L  s
[ Qi (  )]
d
d,

q(Vgs  mi  V ) 2 2 si kT  2 si tox

 ln    ln   ln  cos     tan   f (  ).
2kT  tsi
2
q ni   oxtsi

I ds W dV W  2kT 
  Qi (  s )  Qi (  s )   f (  s )  ,
 s L d s
L  q 
I ds W dV W  2kT 
   Qi (  d )  Qi (  d )  f (  d )  ,
 d L d d
L  q 

q(Vgs  mi ) 2 2 si kT  q(Vgs  mi  Vds ) 2 2 si kT 

 ln    f (  s ),  ln    f (  d ).
2kT  tsi q 2ni  2kT  tsi q 2ni 

W  2kT  1 W  2kT  1 W
gm   Qi (  s )   f (  s )    Qi (  d )  f (  d )     Qi (  s )  Qi (  d ) 
L  q  (2kT / q) f (  s ) L  q  (2kT / q) f (  d ) L

d 2kT 2  tan 
Since Qi (  )  2 si  2 si ,
dx x tsi /2 q tsi

SOLUTION 46
 W kT  si
g m  8   s tan  s   d tan  d 
L q tsi

Likewise,
I ds  s I ds  d I ds  d W  2kT  1 W
gd      Qi (  d )  f (  d )     Qi (  d )
 s Vds  d Vds  d Vds L  q  ( 2 kT / q ) f (  d ) L

W kT  si
g d  8  d tan  d
L q tsi

7.5
At the source,
2kT 2  s tan  s
Qi  2 si ,
q tsi

where

q(Vgs  mi ) 2 2 si kT  2 si tox

 ln    ln  s  ln  cos  s    s tan  s .
2kT  tsi q 2ni   ox t si

Subthreshold,  s << 1, ln  s dominates.

q (Vgs mi )/2 kT tsi q 2ni

s  e
2 2 si kT

Therefore,
kT  s 2 q (V  )/ kT
Qi  8 si  qni tsi e gs mi
q tsi

Far above threshold,  s  /2,  s tan  s dominates.

2kT 2  s tan  s   2kT  2 2 si kT  

Qi  2 si  2 ox Vgs  mi  ln  
q tsi tox  q  tsi q 2ni  

7.6
 t  t  
tan  ox  tan  si   ox
    2   si

Let  =  tox/, then  tsi/(2) =  (4tox)/(2) = 2.

SOLUTION 47
So
2 tan  
tan  tan 2  tan   ox
1  tan   si
2

1
Solve tan  
1  2 si /  ox
Then
 tox

 1 
tan 1  
 1  2 / 
 si ox 

 tox  tox
If ox/si = 1,    6tox  15 nm.
 
Check:
tan 1 1 / 3  / 6

 tox t
Ifox/si = ,    ox  4tox  10 nm.
tan 1  / 4
1

7.7
(a)
q(Vgs  mi ) 2 2 si kT  2 si tox
 ln    ln   ln cos     tan 
2kT  tsi q 2ni   ox t si

mi = m – ( + Eg/2q) = 0.16 V. The LHS = 2.41  in subthreshold.

SOLUTION 48
 The RHS is dominated by ln , so  = e-2.41 = 0.09.

Electron charge density Qi = (8kT/q) (si/tsi)  tan  = 1.7410-9 C/cm2.

Acceptor charge density q Na tsi = 4.810-10 C/cm2.

Both charges are negative, so total net charge density = 2.2210-9 C/cm2.

(b) The threshold shift is q Na tsi / (2 Cox) = 0.139 V, more positive.

(c) Number of dopants is N = Na tsi W L = 27

The number fluctuation is sqrt (N) = 5.2

It is equivalent to an avg. doping of

sqrt (N)/(tsi W L) = Na = 5.81017 cm-3.

It causes a 1 threshold voltage uncertainty of

q Na tsi / (2 Cox) = 0.027 V.

7.8
q(Vgs  mi  V ) 2 2 si kT  2 si tox
 ln    ln   ln  cos     tan  .
2kT  tsi q 2ni   oxtsi

2kT  2 2 si kT 
Vt  mi  ln  
q  tsi q 2ni 
So,
2kT  2 si tox 
Vgs  Vt  V  ln   lncos     tan  
q   oxt si 

8kT  si
Qi   tan 
q t si
Therefore,
2 si tox
ln   lncos     tan 
1 2(Vgs  Vt  V )  oxt si tox ln   lncos  
   
Cinv Qi 2 si  ox 2 si
 tan   tan 
t si t si
and

SOLUTION 49
 2 si  tan 
Csi 
t si ln   lncos  

7.9
Integration of current continuity eq. from 0 to L yields
2 s
W 4 si  2kT    2  si tox 2 2 
I ds        tan   .
L tsi  q  
tan
2  ox tsi  d

Similar integration of current continuity eq. from 0 to y yields

2 s
W 4 si  2kT    2  si tox 2 2 
I ds        tan   .
y tsi  q  
tan
2  ox tsi 

For  << 1, the function in the square bracket  2/2.

Ratio of the two eqs. gives
L s  d
2 2

y s2   2

Also, for  << 1,

q(Vgs  mi  V )  2 2 si kT  2 si tox
 ln    ln   ln  cos     tan   ln  .
2kT 2
 tsi q ni   ox tsi
where V=0 for s and V=Vds for d.

Substituting , s, and d into the ratio-ed eq.,

L s  d 1  e qVds / kT
2 2
  .
y  s 2   2 1  e qV / kT
Solving V, we obtain
kT  1 
V ( y)  ln   qVds / kT 
.
q 1  ( y / L)(1  e )

7.10
Use the subthreshold model for SOI MOSFETs. Apply Eq. 7.10,
(tsi  3tox 2 )(Vgs1  mi1 )  3tox1 (Vgs 2  mi 2 )
 s1  .
tsi  3tox1  3tox 2

If Vgs2 = constant,

SOLUTION 50
 Vg1 t si  3tox1  3tox 2

 s1 t si  3tox 2
Therefore,
t si  3tox1  3tox 2
SS   60 mV/dec  75 mV/dec
t si  3tox 2

For Vt, solve Vgs1 with s1 = Eg/2q. Or use Eq. 7.13:

Eg
Vt  mi1  m   m  1 (Vgs 2  mi 2 )
2q
with
tsi  3tox1  3tox 2
m  1.25.
tsi  3tox 2

mi1 = 4.35 – (4.05 + 0.56) = –0.26 V, mi2 = 4.95 – (4.05 + 0.56) = 0.34 V, Vgs2 = 0.5 V 
Vt = 0.40 V.

7.11
tsi  3tox1  3tox 2 17
tsi = 5 nm, tox1 = tox2 = 2 nm, mi1 = Eg/2q, mi2 = 0, so m  .
tsi  3tox 2 11

1  1 11 6
 s1  (Vgs1  mi1 )   1   (Vgs 2  mi 2 )  (Vgs1  0.56)  Vgs 2
m  m 17 17

When both gates switch, Vgs1 = Vgs2.

s1 = 0.56 when Vgs1 = Vgs2 = Vt = 0.56(6/17) = 0.20 V.

Vgs2 = 0.

s1 = 0.56 when Vgs1 = Vt1 = 0.56(17/11 – 1) = 0.31 V.

From the above, the Vgs1/s1 ratio is 17/11.

Thus the inverse subthreshold slope is (17/11)60 mV/dec = 93 mV/dec.

From Eq. 7.8,

Vgs  mi1  (0) Vgs  mi 2  (tsi )
Qi   ox   ox .
tox tox

Above the threshold in (a), (0) is pinned to Eg/2q. Thus

Qi  Cox Vgs  Cox  Vgs   (tsi ) .

Assume negligible charge in silicon except near the front surface, we have

SOLUTION 51
  (tsi ) tsi 5
  .
Vgs tsi  3tox 11

Therefore,
Qi / Vgs  Cox  Cox 1   (tsi ) / Vgs   (17 / 11)Cox  2.67  F/cm2 .

SOLUTION 52
 Solutions to Chapter 8 Exercises

8.1. Under the scaling transformation, W  W/, L  L/, tox  tox/, Vds  Vds/, Vg 
Vg/, and Vt  Vt/, Eq. (5.23) becomes

W /   Vg Vt  Vds I ds
I ds   eff (Cox )  
L /      
;


and Eq. (5.31) becomes

2
W /  1  Vg Vt  I
I ds  eff (Cox )     ds .
L /  2m     

Note that both m = 1 + 3tox/Wdm and eff which is a function of Eeff given by Eq. (5.53) are nearly
invariant under constant field scaling.

8.2. Under the same scaling rules as above, Eq. (5.39) becomes

2
W /  kT  q(V V )/ mkT
I ds  eff (Cox ) ( m  1)  e g t .
L/  q 

The exp(qVds/kT) term has been neglected since typically Vds >> kT/q. In subthreshold, Vg < Vt
and exp[q(Vg  Vt)/mkT] > exp[q(Vg  Vt)/mkT] (note that  > 1), therefore, the subthreshold
current increases with scaling faster than Ids.

If the temperature also scales down by the same factor, i.e., T  T/, then Ids  Ids/,
same as the drift current in Exercise 8.1.

8.3. Since the factor z = 2eff(Vg  Vt)/(mvsatL) of Eq. (6.47) is invariant under the scaling
transformation, W  W/, L  L/, tox  tox/, Vds  Vds/, Vg  Vg/, and Vt  Vt/, the
square-root of (z + 1) and therefore the fraction in Eq. (6.49) is unchanged after scaling. The
saturation current Idsat of Eq. (6.49) then scales the same way as the fully saturation-velocity
limited current, Eq. (6.51), i.e.,

W V V  I
I dsat  (Cox ) vsat  g  t   dsat .
   

8.4. From Eqs. (8.8) and (8.9), the average CMOS inverter delay is

SOLUTION 53
  n  p CVdd  1 1 
     .
2 4  Wn I nsat Wp I psat 

If the inverter is driving another stage with the same n- to p-width ratio and if both the n- and p-
devices have the same capacitance per unit width, the load capacitance C is proportional to (Wn +
Wp), i.e.,

 1 1  1 1 W / Wn Wn / Wp
  (Wn  Wp )      p  .
 Wn I nsat Wp I psat  I nsat I psat I nsat I psat

The delay is a minimum when the n- to p-width ratio is such that the last two terms are equal,
i.e., when Wp/Wn = (Insat/Ipsat)1/2. Note that n < p for minimum delay.

8.5. For the following RC circuit,

Vdd

R V(t)

I(t) C

if the voltage source is switched to Vdd at t = 0, then

dV
Vdd  V (t )  RI ( t )  V ( t )  RC .
dt

With the initial condition V(t = 0) = 0, the solution for V(t) is


V ( t )  Vdd 1  e t / RC . 
The energy dissipated in R is
2
 Vdd  2 t / RC 1
E   RI 2 dt   dt  CVdd ,
2
e
0 R 0 2

SOLUTION 54
which is independent of R. The same amount of energy is stored in C.

If the voltage source is now switched to 0, one has

dV
V (t )  RC  0.
dt

With the initial condition V(t = 0) = Vdd, the solution for V(t) is

V (t )  Vdd e t / RC .

The energy stored in C is now all dissipated in R:

2
 V  2 t / RC 1
E RI dt  dd  dt  CVdd .
2 2
e
0 R 0 2

8.6. From Eq. (5.27) and the inversion charge expression above Eq. (5.70), the transit time is

WLCox (Vg  Vt  mVds / 2) L2

 tr  
 eff Cox (W / L)(Vg  Vt  mVds / 2)Vds  eff Vds

for a MOSFET device biased in the linear region.

From Eq. (5.31) and the inversion charge expression above Eq. (5.72), the transit time is

( 2 / 3)WLCox (Vg  Vt ) 4mL2

 tr  
 eff Cox (W / L)(Vg  Vt )2 / 2m 3 eff (Vg  Vt )

for a long-channel MOSFET biased in saturation.

8.7. From Eq. (6.49) and the inversion charge expression in Exercise 6.8, the transit time is

Qi L 1  2eff (Vg  Vt ) / ( mvsat L )  1 / 3

 tr  
I dsat vsat 1  2eff (Vg  Vt ) / ( mvsat L )  1

for a short-channel MOSFET biased in saturation. The limiting value of tr is L/vsat when the
device becomes fully velocity saturated as L  0.

8.8. When Ids is considered as a function of the external voltages: Ids(Vg,Vds), one has

SOLUTION 55
  I  I 
dI ds   ds  dVg   ds  dVds .
  Vg 
 Vds   Vds V g

From Eqs. (8.13) and (8.14),

dVds  dVds  ( Rs  Rd )dI ds ,

and

dVg  dVg  Rs dI ds .

Substituting them into the top equation and solving for dIds, one obtains

( I ds /  Vg )Vds
dI ds  dVg  (a similar term in dVds ) .
1  Rs ( I ds /  Vg )Vds  ( Rs  Rd )( I ds /  Vds )V g

Now Ids is expressed as a function of the internal voltages: Ids(Vg,Vds). One can identify the
first coefficient as the intrinsic transconductance, i.e.,

 I  gm
gm   ds   ,
  Vg 
 Vds 1  gm Rs  gds ( Rs  Rd )

where

 I 
gm   ds 
  Vg 
 Vds

is the extrinsic transconductance, and

I 
gds   ds 
  Vds V g

is the extrinsic output conductance.

8.9. The transmission line model of contact resistance in a planar geometry is represented by
the distributed network below. The current flows from a thin resistive film (diffusion with a
sheet resistivity sd) into a ground plane (metal) with an interfacial contact resistivity c between
them (Fig. 8.18).

SOLUTION 56
 I

Metal contact

Gdx
Rdx
I
x=0 V(x) x=lc
I(x)

Following a similar approach as in Eqs. (8.29)-(8.31), one can write

dV
V ( x  dx )  V ( x )  dx   I ( x ) Rdx ,
dx

and

dI
I ( x  dx )  I ( x )  dx  V ( x )Gdx .
dx

Here R = sd/W and G = W/c. From the above two equations, one obtains

d2 f 
 RGf  sd f ,
dx 2
c

where f(x) = V(x) or I(x).

8.10. The solution to the second-order differential equation in the above exercise is of the
sinh[(sd/c)1/2x] and cosh[(sd/c)1/2x] form. With the boundary condition I(x=lc) = 0 where x =
0 is the leading edge and x = lc is the far end of the contact window (see figure in Exercise 5.5),
the solution is

  
I ( x )  I0 sinh  sd (lc  x ) ,
  c 

where I0 is a constant multiplying factor. From the second equation in Exercise 8.9, the voltage
is

c dI c sd   
V (x)    I0 cosh  sd (lc  x ) .
W dx W  c 

The contact resistance is then

SOLUTION 57
 V ( x  0)  c  sd   sd 
Rco   coth lc .
I ( x  0) W   c 

8.11. From Eqs. (8.49) and (8.51), bmin <  if

Cout  CL  2Cout  2 CinCL .

This inequality is quadratic in CL1/2, which can be solved to yield

 
2
CL  Cin  Cin  Cout .

Under this condition, the insertion of one or a few properly designed buffer stage(s) will help
reduce the overall delay.

8.12. For an n-stage buffer with successive width ratios of k1, k2, k3,     , kn, the switching
resistances are Rsw, Rsw/k1, Rsw/k1k2, Rsw/k1k2k3,     , etc. The output capacitances are Cout,
k1Cout, k1k2Cout, k1k2k3Cout,     , etc.; and the input capacitances are k1Cin, k1k2Cin, k1k2k3Cin,
k1k2k3k4Cin,     , etc. The delay of the first stage is

 b,1  Rsw (Cout  k1Cin ) ;

the delay of the second stage is

Rsw
 b, 2  ( k1Cout  k1k2Cin )  Rsw ( Cout  k2Cin ) ;
k1

    , and so on, until the n-th stage:

 b, n  Rsw ( Cout  kn Cin ) .

The delay of the last stage is

Rsw  CL 
 b, n 1  ( k1k2    kn Cout  CL )  Rsw  Cout  .
k1k2   kn  k1k2    kn 

The total buffer delay is the sum of all the stage delays, i.e.,

 CL 
 b ( n)  Rsw  ( n  1)Cout  ( k1  k2       kn )Cin  .
 k1k2    kn 

SOLUTION 58
[Note that delay equations of the form Eq. (8.45) used above are defined in terms of propagation
delays through cascaded stages of identical width and loading (Fig. 8.32), and as such are not
strictly applicable to a chain of drivers with varying widths and rise times. When the input rise
time is shorter than that of the output as is the case with the above tapered drivers, the switching
resistance Rsw  d/dCL is somewhat less than that defined in Section [Link].]

8.13. Take any pair of k’s (positive numbers), e.g., k1 and k2, for a given product k1k2, the sum
k1 + k2 is a minimum when k1 = k2. This can be generalized to the result that for a given product
k1k2k3    kn, the sum k1 + k2 + k3 +    + kn is a minimum when k1 = k2 =      = kn. Let k1 = k2
=      = kn = k, the delay in Exercise 8.12 becomes

 CL 
 b ( n )  Rsw  ( n  1)Cout  nkCin  .
 kn 

Differentiating it with respect to k, one finds that the above expression has a minimum

  CL 
1/ ( n 1)

 b min ( n)  Rsw ( n  1)Cout  ( n  1)Cin   ,
  Cin  

when k = (CL/Cin)1/(n+1). [Note that the same result can be obtained alternatively by partially
differentiating the delay in Exercise 8.12 with respect to k1, k2, k3,     , kn respectively and
solving the n coupled equations.] bmin(n), as expected, is reduced to Eq. (8.51) if n = 1.

8.14. To find the optimum n which gives the shortest overall delay, we let k = (CL/Cin)1/(n+1) and
treat it as a continuous variable. Then n + 1 = ln(CL/Cin)/ln(k) and the delay in Exercise 8.13
becomes

ln( CL / Cin )
 b min ( k )  Rsw ( Cout  kCin ) .
ln( k )

This has a minimum when dbmin/dk = 0, which yields an equation for k:

Cout
kln( k )  1  .
Cin

n is then the closest integer to

ln( CL / Cin )
n  1.
ln( k )

SOLUTION 59
 For typical Cout/Cin ratios not too different from unity, k is in the range of 3-5. Note that
k also gives the optimum width ratio between the successive buffer stages, i.e., k1 = k2 =      =
kn = k. Since Cout + kCin = kln(k)Cin, the minimum buffered delay can be written as

 b min  kRswCin ln( CL / Cin ),

which only increases logarithmically with load capacitance.

8.15.
Power = CVdd2f where f = 1/(2) and C = Cin+Cout+CL

So Power = Vdd2/2Rsw

If (Wn, Wp)  (kWn, kWp),

Rsw  Rsw/k

Cin+Cout  k(Cin+Cout)

Therefore,

Delay   Rsw(Cin+Cout+CL/k)

Power  k times original value.

SOLUTION 60
 Solutions to Chapter 9 Exercises

9.1 Consider hole current flow in an n-region. Eq. (2.78) gives

d p
J p   qpn  p . (1)
dx
Now, as shown in Fig. 3.10, dn(n-region)/dx  0 when there is negligible IR drop in the n-region.
Therefore, (1) can be rewritten as

J p ( n  region)   qpn  p
d
dx

 p  n .  (2)

Eq. (9.15) gives

kT  pn nn 
 p 
 n  ln 
q  nie2 
 (3)

for the n-region. Substituting (3) into (2), we obtain

 kT  1 dnn 1 dnie2   dp
J p ( n  region )  qpn  p    2    qDp n , (4)
 q  nn dx nie dx   dx

where we have used the Einstein relation Dp = kTp/q. If we compare (4) with Eq. (2.69), i.e.,
dpn
J p ( n  region)  qpn  pE ( n  region)  qDp , (5)
dx
 kT  1 dnn 1 dnie2  
we have E( n  region )     2  .
 q  nn dx nie dx  

9.2 Eq. (2.78) gives, for an n-region,

d p
J p   qpn  p . (1)
dx
As shown in Fig. 3.10, when there is negligible IR drop in the n-region, we have
dn (n  region)
 0. (2)
dx
Therefore, (1) and (2) give
d ( p   n )
J p ( n  region )   qpn  p . (3)
dx
From Eq. (9.15), for the n-region,

SOLUTION 61
 kT  pn nn 
( p   n )  ln  . (4)
q  nie2 

Substituting (4) into (3), and using Dp = kTp/q, we have

nie2 d  nn pn 
J p ( n  region )   qDp  .
nn dx  nie2 

9.3
pn (x)

WE1

polysilicon silicon

x
0
WE + WE1) W E
E-B junction

(a) Let pn(WE) be the excess hole density at the polysilicon-silicon interface.
Also let
x'  ( x  WE ) . That is, x’ is the distance between a point inside the polysilicon layer and the
polysilicon-silicon interface. The steady-state equation governing the excess hole density is
given by the hole-equivalent of Eq. (3.36), i.e.
d 2 pn pn
 2  0. (1)
dx 2 Lp

Applying (1) to hole distribution inside the polysilicon layer, we have

d 2 pn pn
 2  0. (2)
dx '2 LpE1

The general equation for (2) has the form

pn ( x' )  Ae  Be x ' /LpE1 .

x ' /L pE 1
(3)
At the ohmic contact at x’ = WE1, we have
WE 1 / L pE 1
0  Ae  Be
WE 1 / L pE 1
. (4)

SOLUTION 62
At the polysilicon-silicon interface (x’ = 0), we have
pn ( WE )  A  B . (5)

pn ( WE )
(4) and (5) give A (6)
1  e E1 pE1
2W / L

 pn ( WE )e E 1
2W / L pE 1

and B . (7)
1  e E 1 pE 1
2W / L

Substituting (6) and (7) into (3), we have

pn ( x ' )  pn ( WE )


sinh (WE1  x' ) / LpE1 . (8)
sinh(WE1 / LpE1 )

(b) The hole current density entering the polysilicon layer is given by Eq. (9.59), i.e.,
J p ( x  WE )  qpn ( WE )S p . (9)

Using (8), this current density is also

dpn ( x )
J p ( x  WE )   qDpE1 (10)
dx x  WE

dpn ( x ' )
 qDpE1
dx ' x '  0
pn ( WE )
  qDpE1 .
LpE1 tanh(WE1 / LpE1 )

DpE1
Equations (9) and (10) give Sp  .
LpE1 tanh(WE1 / LpE1 )

9.4 Neglecting parasitic resistances, Eqs. (9.182) and (9.183) give

I E   I EBO exp( qVBE
 / kT )  1   R IC (1)

and IC   ICBO exp( qVBC

 / kT )  1   F I E , (2)

which can be rearranged to give

kT  I E   R IC 
  ln 1 
q  I EBO 
VBE (3)

SOLUTION 63
 kT  IC   F I E 
  ln 1  .
q  ICBO 
and VBC (4)

Now,   VC  VE  (VC  VB )  (VB  VE )  VBE

VCE   VBC
 . (5)
Substituting (3) and (4) into (5) and rearranging, we obtain

kT  ICBO ( I EBO  I E   R IC ) 
  .
q  I EBO ( ICBO  IC   F I E ) 
VCE ln (6)

Now, Eq. (9.144) gives  R I R0   F IF 0 , (7)

and Eqs. (9.180) and (9.181) give
I EBO  I F 0 (1   R F ) , (8)
and ICBO  I R 0 (1   R F ) . (9)
ICBO I R 0  F
Therefore,   . (10)
I EBO I F 0  R

kT   F  I EBO  I E   R IC  
Substituting (10) into (6) gives  
VCE ln  .
q   R  ICBO  IC   F I E  

9.5 From Exercise (9.4), we have

kT   F  I EBO  I E   R IC  
 
VCE ln  . (1)
q   R  ICBO  IC   F I E  

Now, when emitter and collector resistances are included, we have

VCE  VC  VE  VC  rc IC  (VE  re I E )  VCE
  rc IC  re ( I B  IC ) , (2)
where we have used the fact that
I B  IC  I E  0 . (3)
Substituting (1) into (2), we obtain

kT   F  I EBO  I E   R IC  
VCE  ln    rc IC  re ( I B  IC ) . (4)
q   R  ICBO  IC   F I E  

If we neglect the saturation currents IEBO and ICBO, then using (3) we can rearrange (4) to give

kT   F  I B  IC   R IC  
VCE  ln    rC IC  re ( I B  IC ) (5)
q   R  F ( I B  I C )  IC  

SOLUTION 64
 kT  I B  IC (1   R ) 
 ln    re ( I B  IC )  rc IC .
q   R  I B  IC (1   F ) /  F  

For open-circuit collector, IC = 0, and (5) reduces to

kT  1 
VCE ( IC  0)  ln    re I B .
q R 

9.6 From Eq. (3.65), we have

sinh(WB  x ) / LnB 
n p  n p 0  n p 0 exp(qVBE / kT )  1 . (1)
sinh(WB / LnB )

If we assume diffusion current only, then

dnp cosh(WB  x ) / LnB   1 
J n ( x )  qDnB  np 0  exp( qVBE / kT )  1  . (2)
dx sinh(WB / LnB )  LnB 

Therefore, the base transport factor is

1
J ( x  WB )  W 
T  n   cosh B  . (3)
J n ( x  0)  LnB 

For a base region of WB = 100 nm and NB = 1018 cm3, we have, from Fig. 3.14, LnB  15 m, and
1
 .
01
 T   cosh   0.9998 .
 15 

9.7 Eq. (9.188) give the condition for collector-emitter breakdown as

 T M ( BVCEO )  1 . (1)
Now, from Eqs. (9.78) and (9.178), we have
0
  T , (2)
1 0

where 0 is the common-emitter current gain at negligible base-collector junction avalanche. (1)
and (2) give
1
M ( BVCEO )  1  .
0

SOLUTION 65
9.8 The equation

d 2 n p ( x) n p ( x)
  0, (1)
dx 2
2
LnB

where

 nB DnB LnB
 
LnB  (2)
1  j nB 1  j nB

has a general solution

 )  B exp( x LnB
n p ( x)  A exp( x LnB  ), (3)

where A and B can be determined from the boundary conditions for Δnp(x).
Expanding the Shockley diode equation, Eq. (3.61), and keeping only the first order term in vbe,
we have
 q 
n p (0, t )  n p 0 exp  qvBE (t ) kT   n p 0 exp(qVBE kT ) 1  vbee jt  . (4)
 kT 
Therefore, the boundary condition for Δnp(x) at x = 0 is
qvbe
n p (0)  n p 0 exp(qVBE / kT ) . (5)
kT
At the ohmic base contact, there are no excess electrons. Therefore, the boundary condition for
Δnp(x) at x = WB is
n p (WB )  0. (6)

Applying these boundary conditions to Eqs. (1) and (3), we obtain

n p 0 qvbe  qV  sinh  (WB  x) LnB
 
n p ( x)  exp  BE  .
kT  kT  sinh(WB LnB  )

9.9 From Eq. (2.97), the net recombination rate in the p-type base region of an npn
transistor is
n p ( x)  n p 0
U ( x)  . (1)
 nB
From Eq. (3.65), the electron distribution in the p-type base is
sinh  (WB  x) / LnB 
n p ( x)  n p 0  n p 0 [e qVBE / kT  1] . (2)
sinh(WB / LnB )

SOLUTION 66
For a typical transistor in operation, WB/LnB << 1, and qVBE/kT >> 1, and (2) reduces to
 x 
n p ( x)  n p 0  n p 0eqVBE / kT 1  .
 WB 
(3)
The base current component due to recombination in the base region is
WB
J B,rec   qU ( x)dx. (4)
0

Substituting (3) into (4), and using the fact that p p 0 n p 0  nieB
2
for the base region [see Eq.

(9.14)], and that pp0 = NB, we have

2
qWB nieB
J B ,rec  exp(qVBE / kT ). (5)
2 nB N B

Eq. (9.71) gives the base current due to hole injection from base into the emitter as
2
qD pE nieE exp  qVBE kT 
J B ,inj  . (6)

N E L pE tanh WE L pE 
J B,rec 2
WB L pE tanh(WE / L pE ) nieB NE
Therefore,  . (7)
J B,inj 2 nB D pE 2
nieE N B

For WE = 100 nm, NE = 1 × 1020 cm-3, NB = 1 × 1018 cm-3, and WB = 50 nm, we have, from Fig.
3.14, nB = 3×107 s, LpE = 0.38×104 cm, tanh(WE/LpE) = 0.26, p = 120 cm2/V-s. That is, from
Eq. (2.51), DpE = (kT/q)p = 0.026×120 cm2/s =3.12 cm2/s. Also, from Eq. (9.14), we have
2
nieB
2
 exp[(EgB  EgE ) / kT ]. (8)
nieE

From the Unified model in Fig. 9.4, EgB = 31 meV and EgE = 92 meV. Therefore
2
nieB 2
/ nieE  exp(61/ 26) = 0.096. We have

J B,rec 2
WB L pE tanh(WE / L pE ) nieB NE

J B,inj 2 nB D pE 2
nieE N B

= [50×107 × 0.38×104 × 0.26/ (2 × 3×107 × 3.12)] × 0.096 × 100 = 2.5×104.

9.10 Consider a wide-emitter but narrow-base n+-p diode, with emitter doping concentration
NE = 1020 cm3, base doping concentration NB = 1017 cm3, and base width WB = 100 nm.

SOLUTION 67
(a) Ignoring heavy-doping effect, the diffusion capacitance ratio is
C Dn 2 N E WB
 . (1)
C Dp 3 N B L pE

From Fig. 3.14, we have LpE = 0.38 m for holes in the n+ emitter. Substituting all these values
into (1) gives CDn/CDp = 175.

(b) When heavy-doping effect cannot be ignored, Eq. (9.133) gives

2 q 2 qA
C Dn  I nt B  J n tB ,
3 kT 3 kT
(2)
where A is the cross-sectional area of the diode. From Eq. (9.119), for a p-region with WB <<
LnB, we have
qDnB n p 0
Jn  exp(qVBE / kT )
WB
qDnB p p 0 n p 0
 exp(qVBE / kT ) (3)
WB p p 0
2
qDnB nieB
 exp( qVBE / kT ).
WB N B

From Eq. (9.106), we have

WB2
tB  . (4)
2 DnB

Substituting (3) and (4) into (2), we obtain

2
2 qA qnieBWB
CDn  exp(qVBE / kT ) . (5)
3 kT 2 N B

Similarly, for a n-type emitter region with WE << LpE, we have

1 q 1 qA
C Dp  I p pE  J p  pE (6)
2 kT 2 kT

SOLUTION 68
 1 qA qD p nieE pE
2
 exp(qVBE / kT )
2 kT N E L pE
2
1 qA qnieE L pE
 exp(qVBE / kT )
2 kT NE

where we have used the fact that L2 = D. (5) and (6) give

C Dn 2  nieB
2
 N E  WB
  2   . (7)
C Dp 3  nieE  N B  pE
L

Now, 2
nieB  ni2 exp( EgB / kT )

and 2
nieE  ni2 exp( EgE / kT ) .
2
nieB  E  EgE 
Therefore,  exp gB . (8)
2
nieE  kT 

From Fig. 9.4, we have EgB  2 meV for p-type base with doping concentration of 1017 cm3,
and EgE  92 meV for n-type emitter with doping concentration of 1020 cm3. Also, at room
temperature (300 K), kT = 26 meV. Therefore
2
nieB  E  EgE 
 exp gB  = 0.031. (9)
2
nieE  kT 

That is, the effect of heavy-doping reduces the ratio CDn/CDp from 175 to 5.4.

9.11
2
AE qDnB nieB exp(qVBE / kT )
I C (VBE )lowinjection  .
N BWB

1
A qD n 2 exp(qVBE / kT )  1  4n 2 exp(qVBE / kT )  
I C (VBE )all injection  E nB ieB 1   1  ieB  1  .
N BWB  4 N 2 
  B 

I C (VBE )lowinjection  1 4n 2 exp(qVBE / kT )  

Therefore, Ratio   1   1  ieB  1  . (1)
IC (VBE )all injection  4 N 2 
  B 

Eq. (9.14) gives 2

nieB  ni2 exp  EgB ( N B ) / kT  . Therefore, (1) becomes

SOLUTION 69
   4ni2 exp{[ qVBE  EgB ( N B )] / kT }  
 1
Ratio  1  1  1  . (2)
 4 N B2 
  

For Eg, we use the unified model given in Eq. (9.18), namely

 2 
  N    N  
Eg  N   6.92 ln  17 
  
ln 17  
 0.5  meV.
  1.3  10    1.3  10   
 

15
Ratio = IC(low inj) / IC(all inj)

10
NB=1E17
NB=1E18
NB=1E19
5

0
0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
VBE (V)

SOLUTION 70
 Solutions to Chapter 10 Exercises

10.1 For the polysilicon-emitter model described in Exercise 9.3,

 n2   W
GE (WE ,WE1 )  N E  2i   E 

LpE1 tanh WE1 / LpE1   . (1)
 nieE   DpE DpE1 

pn (x)

WE1

polysilicon silicon

x
0
WE + WE1) W E
E-B junction

Assume the lifetime in polysilicon layer is the same as in silicon, and assume the mobility in
polysilicon is smaller than that in silicon.

(a) For non-polysilicon-emitter: NE = 1020 cm3, WE = 300 nm, WE1 = 0. From Eq.
(9.14), we have
ni2
2
 exp(  E gE / kT ) . (2)
nieE

From Fig. 9.4, we have, for NE = 1020 cm3, EgE = 92 meV. Using kT = 26 meV, we have
ni2
2
 exp(  E gE / kT )  0.029 . (3)
nieE

From, Fig. 3.14, we have pE = 130 cm2/V-s. Therefore,

kT pE
DpE   338
. cm2/s. (4)
q

 n2  W
Therefore, GE (300 nm, 0)  N E  2i  E  2.6  1013 cm4-s.
 nieE  DpE

SOLUTION 71
 (b) For polysilicon-emitter: NE = 1020 cm3 and WE = 30 nm. With pE1/pE = 1/3, we
have
kT pE1
DpE1   113
. cm2/s. (5)
q

Now, from Fig. 3.14, we have LpE = 0.38 m. Therefore, we have

kT L
LpE1   pE1 pE  pE  0.22 m. (6)
q 3

Substituting all these values into (1) we have

WE1 (nm) GE(30 nm, WE1) (cm4-s)

50 1.521013

100 2.661013

200 4.361013

300 5.221013

(c)

2.2
GE(poly)/GE(non-poly)

2
1.8
1.6
1.4
1.2
1
0.8
0.6
0.4
0 50 100 150 200 250 300 350
Polysilicon-layer thickness (nm)

10.2 For a box-like profile, Eq. (10.5) gives

1
B
RSbi   q  p p ( x ) p ( x )dx
WB 1
  , (1)
0 qN B  pWB WB

SOLUTION 72
where B is the resistivity of the p-type base layer. Setting RSbi = 104 /, and using Fig. 2.10
for B, we obtain the values in the following table.

SOLUTION 73
 NB (cm3) B (-cm) WB (nm)
61016 0.3 300
11017 0.2 200
2.91017 0.1 100
71017 0.05 50

1E+18
Base doping (cm-3)

5E+17

3E+17
2E+17

1E+17

5E+16
30 50 100 200 300 500
Base width (nm)

10.3 From Exercise 10.2, and Fig. 3.14, we have the following table.

WB (nm) NB (cm3) nB (cm2/V-s) DnB (cm2/s) tB (ps)

50 71017 37.5 9.75 1.3
100 2.91017 505 13.1 3.8
200 11017 760 19.8 10.1

where we have used DnB = kTnB/q and tB = WB2/2DnB. Also, since Fig. 3.14(a) does not have
values for NB < 11017 cm3, we have dropped the corresponding base width and base doping
concentration values in the table and the following plot.

SOLUTION 74
 12

Base transit time (ps)

10

0
0 50 100 150 200 250
Base width (nm)

10.4 From Eq. (9.94), we have

QpB qN BWB
VA   . (1)
CdBC CdBC

To avoid significant base widening, collector current density is kept at

JC
J C  0.3qvsat N C , or NC  , (2)
0.3qvsat

with vsat = 107 cm/s.

(a) For a one-sided BC junction, with VCB = 2 V, Eqs. (3.13) and (3.15) give

 si qN C si J C si
CdBC    . (3)
WdBC 2( bi  VCB ) 0.6vsat ( bi  VCB )

We can obtain bi from Fig. 3.4 (or set it to about 0.95 V as an approximation) and use si =
1.041012 F/cm to generate the following table.

JC (mA/m2) NC (cm3) bi (V) CdBC (fF/m2)

0.1 2.081016 0.925 0.243
0.2 4.16 0.945 0.343
0.3 6.24 0.95 0.420
0.4 8.32 0.96 0.484
0.5 1.041017 0.97 0.540

SOLUTION 75
 0.8 1.66 0.98 0.659
1.0 2.08 0.99 0.761
2.0 4.16 1.0 1.07
3.0 6.24 1.02 1.31
4.0 8.32 1.02 1.52
5.0 1.041018 1.02 1.69

3
2
CdBC (fF/um2)

0.5
0.3
0.2

0.1
0.1 0.2 0.3 0.5 1 2 3 5
JC (mA/um2)

(b) For a base design with qNBWB = 1.6106 C/cm2, we have the following table

JC (mA/m2) CdBC (fF/m2) VA (V)

0.1 0.243 65.8
0.2 0.343 46.6
0.3 0.420 38.1
0.4 0.484 33.1
0.5 0.540 29.6
0.8 0.659 24.3
1.0 0.761 21.0
2.0 1.07 15.0
3.0 1.31 12.2

SOLUTION 76
 4.0 1.52 10.5
5.0 1.69 9.5

100

50
VA (V)
30
20

10

5
0.1 0.2 0.3 0.5 1 2 3 5
JC (mA/um2)

10.5 An npn transistor with WB = 500 nm and negligible heavy doping effect in base layer,
and a graded base profile of
N B ( x )  A  x , (1)

A  N B (0)  2  1017 cm3, (2)

N B (WB )  2  1016 cm3 (3)

A  N B (WB )
and   3.6  1021 cm4. (4)
WB

From Eq. (10.11), the electric field due to the dopant distribution is
kT 1 dN B kT 
E( dopant )   . (5)
q N B ( x ) dx q NB(x)

We have the following table and plot of results.

x (nm) NB (cm3) E (V/cm)

0 2.0E17 -468
50 1.82E17 -514
100 1.64E17 -571

SOLUTION 77
 150 1.46E17 -641
200 1.28E17 -731
250 1.10E17 -851
300 9.2E16 -1017
350 7.4E16 -1265
400 5.6E16 -1671
450 3.8E16 -2463
500 2.0E16 -4680

0
Electric field (V/cm)

-1,000

-2,000

-3,000

-4,000

-5,000
0 100 200 300 400 500 600
x (nm)

10.6 An npn transistor with WB = 100 nm and non-negligible heavy doping effect in base
layer, and a graded base profile of
N B ( x )  A  x , (1)

A  N B (0)  5  1018 cm3, (2)

N B (WB )  5  1017 cm3 (3)

A  N B (WB )
and   4.5  1023 cm4. (4)
WB

From Eq. (10.11), the electric field due to the dopant distribution is
kT 1 dN B kT 
E( dopant )   , (5)
q N B ( x ) dx q NB(x)

and the field due to heavy-doping effect is

SOLUTION 78
 1 dE gB
E( HD )   . (6)
q dx
From Eq. (9.17), the bandgap narrowing in the base layer is


Eg ( N a )  9  F  F 2  0.5  meV (7)

where F  ln( N B / 1017 ) . (8)

 F  
Therefore, we have E( HD )  9  103 1   V/cm. (9)
 F 2  0.5  N B
We have the following table and plot of the results.

x (nm) NB (cm3) E (dopant)(V/cm) E (HD)(V/cm) E (Tot)(V/cm)

0 5.0E18 -2340 1607 -733
10 4.55 -2571 1765 -806
20 4.1 -2854 1958 -895
30 3.65 -3205 2198 -1007
40 3.2 -3656 2506 -1151
50 2.75 -4255 2913 -1342
60 2.3 -5087 3479 -1608
70 1.85 -6324 4317 -2008
80 1.4 -8357 5687 -2670
90 9.5E17 -12316 8330 -3985
100 5.0 -23400 15516 -7884

SOLUTION 79
 20,000

Electric field (V/cm)

10,000

-10,000
E(dopant) E(HD) E(Total)
-20,000

0 20 40 60 80 100 120
x (nm)

10.7 Polysilicon emitter with NE = 1020 cm3, AE = 1 m2, single-crystal region thickness WE =
30 nm, polysilicon layer thickness WE1 = 200 nm, metal-poly contact resistivity contact = 2107
-cm2. From Fig. 2.10, we have for the emitter doping concentration single-crystal = 8104 -
cm. Therefore, we have polysilicon = 24104 -cm. From the information, we can calculate
the emitter resistance components.
WE
rsin gle  crystal   sin gle  crystal  0.24
AE
WE1
rpolysilicon   polysilicon  4.8
AE
rcontact   contact / AE  20

q
10.8 JC 0  .
WB pp

0 2
DnB nieB
dx

(1)
For uniformly doped base with NB = 1018 cm3, WB = 100 nm, and assuming low current levels so
that pp = NB, we have
2
qDnB nieB
JC0  . (2)
N BWB

(a) Neglecting heavy doping effect, we have nieB = ni = 1.01010 cm3. Using the
mobility values in Fig. 3.14, we have

SOLUTION 80
 kT nB
DnB   8.58 cm2/s.
q
Putting these values into (1), we have JC0 (neglecting HD effect) = 1.371011 A/cm2.

(b) When heavy doping effect is included, we have from Eq. (9.14) and Fig. 9.4,
2
nieB  ni2 exp( E gB / kT )  ni2 exp( 43 / 26)  5.23ni2 . (3)

That is, JC0 is 5.23 times larger, or JC0 = 7.191011 A/cm2.

SOLUTION 81
10.9

xjE

Metal or metal silicide

n+

n+ emitter poly
p Ge
Eg0 = Egmax

x
0 WB0
WB

Wcap

Fig. 10.11. Schematic illustrating the emitter and base regions of a polysilicon-emitter SiGe-base bipolar
transistor having a constant Ge distribution in the base. The emitter depth xjE is assumed to be smaller
than the thickness Wcap of the starting Ge-free layer, resulting in a
small region of thickness Wcap xjE near the emitter end without Ge.

1
 W  x jE   WB 0  Wcap  
 exp Eg 0 / kT 
J C 0 (SiGe, x jE )
   cap  (1)
J C 0 (Si, x jE )
x jE Wcap
 WB 0  x jE   WB 0  x jE  
1
 Wcap  x jE   (WB 0  x jE )  (Wcap  x jE )  
     exp E g 0 / kT 
 WB 0  x jE  
  WB 0  x jE 
 
1
 Wcap  x jE   Wcap  x jE  
     1   exp E g 0 / kT  .
 WB 0  x jE   W x
  B0 jE

 

200
dE/kT=0.5 dE/kT=1 dE/kT=2.5 dE/kT=5
100
50
Current Ratio

20
10
5

2
1
0 0.2 0.4 0.6 0.8 1
(WcapxjE )/(WB0 xjE)

SOLUTION 82
 V A (SiGe, x jE )  Wcap  x jE   E   W  Wcap 
  exp g 0    B 0
 kT   W  x
 (2)
W x  
V A (Si, x jE )
x jE Wcap  B0 jE     B0 jE 
 Wcap  x jE   E   (W  x jE )  (Wcap  x jE ) 
  exp g 0    B 0
 kT  

W x   
 B0 jE     W B0 x jE 
 Wcap  x jE   E   W  x jE 
  exp g 0   1  cap
 kT   W  x 

W x 
 B0 jE     B0 jE 

200
dE/kT=0.5 dE/kT=1 dE/kT=2.5 dE/kT=5
100
Early Voltage Ratio

50

20
10
5

2
1
0 0.2 0.4 0.6 0.8 1
(WcapxjE )/(WB0 xjE)

10.10 The base bandgap narrowing parameter corresponding to Fig. 10.37 is given by

n+ Ge
n+ emitter poly

Egmax = Eg0

p
x
0 WB

WB1
x
E gB ,SiGe ( x)  E g 0 x < WB1 (1)
WB1

 E g 0 WB > x > WB1.

From Eq. (10.27), we have

SOLUTION 83
 J C 0 (SiGe)  WB
 WB
. (2)
J C 0 (Si)
 0
exp[E gB ,SiGe ( x) / kT ]dx

Using (1), we have

WB   E gB ,SiGe ( x)  WB1   xE g 0  WB   E g 0 
0
exp
 kT
dx   exp
 0
 WB1kT 
dx   exp
WB1
 kT 
dx (3)


kTWB1
E g 0
 
1  exp E g 0 / kT   WB  WB1 exp E g 0 / kT  .

Substituting (3) into (2), we have

1
 kT WB1  W  
J C 0 (SiGe)
    
1  exp E g 0 / kT   1  B1  exp E g 0 / kT  . (4)
J C 0 (Si)  E g 0 WB  WB  

Similarly, the VA ratio can be obtained from Eq. (10.32), namely

V A (SiGe) exp(E g max kT ) WB
V A (Si)

WB 0 exp[E gB,SiGe ( x) / kT ]dx . (5)

Substituting (3) into (5), we have

V A (SiGe) exp(E g 0 / kT )  kTWB1 

   
1  exp E g 0 / kT   WB  WB1 exp E g 0 / kT  (6)
V A (Si) WB  E g 0 

 W 

kT WB1
E g 0 WB

expE g 0 / kT   1  1  B1  . 
 WB 
The tB ratio can be obtained from Eq. (10.36), namely
t B (SiGe) 2 WB WB

t B (Si)

 WB2 0
exp[E gB , SiGe ( x) / kT ] exp[ E gB , SiGe ( x' ) / kT ]dx' dx .
x
(7)

The integration wrt x can be written as the sum of two parts, with one part for x < WB1 and
another part for x > WB1. That is, we have
WB WB

0
exp[E gB ,SiGe ( x) / kT ] exp[E gB ,SiGe ( x' ) / kT ]dx' dx
x
(8)

WB1  E gB ,SiGe ( x)  WB   E gB ,SiGe ( x' ) 

 exp   exp dx' dx
0
 kT  x
 kT 
WB  E gB ,SiGe ( x)  WB   E gB ,SiGe ( x' ) 
  exp   exp dx' dx .
WB1
 kT x  kT 
For the first integral on RHS of (8), we have

SOLUTION 84
 WB1  E ( x)  WB   EgB , SiGe ( x' ) 
0
exp gB , SiGe
 kT
  exp
 x
 kT
dx' dx

(9)

W B1  E ( x )  W B1   EgB , SiGe ( x' )  WB   EgB , SiGe ( x' ) 

 exp gB , SiGe  exp    exp dx' dx
0
 kT  x  kT  WB1  kT 

W B1  E ( x) W kT   Eg 0 x   W kT    Eg 0 

 exp gB , SiGe  B1 exp   (WB  WB1 )  B1  exp dx
0
 kT  Eg 0  kTWB1   Eg 0   kT 

WB21kT  W kT  W kT     Eg 0 
  (WB  WB1 )  B1   B1  1  exp  .
Eg 0  Eg 0   Eg 0    kT 
For the second integral on RHS of (8), we have
WB  EgB , SiGe ( x)  WB   EgB , SiGe ( x' ) 
WB1  kT x 
exp   exp
kT
dx' dx

(10)

 E ( x)    Eg 0 
WB  x dx
WB
 exp gB , SiGe  exp
W B1
 kT   kT 


1
WB  WB1 2 .
2
Therefore, (7) gives

2 WB21kT  WB1kT  WB1kT     Eg 0  1 

2
t B (SiGe)
    (W  W )     1  exp
 kT  2    W  W  
t B (Si)  WB2  Eg 0  B B1
Eg 0   Eg 0    
B b1


1 
2 2
 WB1  2kT  WB1  
 1      
  WB  E g 0  WB  
 

1  2kT WB1  WB1 kT WB1     E g 0  

  1  1  exp   .
  E g 0 WB  WB E g 0 WB    kT  

SOLUTION 85
10.11 The base bandgap narrowing parameter corresponding to Fig. 10.13 is given by

Egmax

Metal or metal silicide

xje Ge

n+ emitter poly
n+

Eg0
p
x
0 WB0
WB

Wcap

 x  Wcap 
EgB , SiGe ( x)  Eg 0   Eg max  Eg 0  x > Wcap (1)
W W 
 B0 cap 

 Eg 0 x < Wcap.

From Eq. (10.27) we have

J C 0 (SiGe, x jE )   (WB 0  x jE )
 WB 0
. (2)
J C 0 (Si, x jE )

x jE
exp[E gB ,SiGe ( x) / kT ]dx

The integral in (2) can be written as

WB 0   E gB ,SiGe ( x)  Wcap   E gB ,SiGe ( x)  WB 0   E gB ,SiGe ( x) 
 x jE
exp
 kT
dx   exp
 x jE
 kT
dx   exp
 Wcap
 kT
dx . (3)

The first integral on RHS of (3) is
Wcap   E gB ,SiGe ( x)    E g 0 
 x jE
exp
 kT
dx  (Wcap  x jE ) exp
  kT 
 . (4)

The second integral in (2) is

WB 0  E gB ,SiGe ( x)  WB 0   E g 0 (E g max  E g 0 )( x  Wcap ) 
Wcap 
exp
kT



dx  Wcap  kT 
exp
kT (WB 0  Wcap )
dx

(5)

  kT     (E g max  E g 0 )     E g 0 
 WB 0  Wcap   exp
 
  1 exp   .
 E 
 g max  E g 0    kT    kT 
Substituting (4) and (5) into (3), we have from (2)

SOLUTION 86
 J C 0 (SiGe, x jE )  exp(E g 0 / kT )
 .
J C 0 (Si, x jE )  Wcap  x jE   WB 0  Wcap  kT   E  E g max 
    1  exp g 0 
W x   W x  E  
 B0 jE   B0 jE  g max  E g 0    kT 

10.12 For evaluating the transit time ratio, it is convenient to write the base bandgap narrowing
parameter in the form

Egmax
Metal or metal silicide
xje Ge
n+ emitter poly
n+

Eg0
p
x
0 WB0
WB

Wcap

 ,SiGe ( x) ,
E gB ,SiGe ( x)  E g 0  E gB (1)

 x  Wcap 
where  , SiGe ( x)  
EgB Eg max  Eg 0  x > Wcap (2)
W W 
 B0 cap 

0 x < Wcap.
The transit time ratio is, from Eq. (10.54)
t B ( SiGe, x jE )
(3)
t B ( Si, x jE )
x jE Wcap

2 WB 0 WB 0

 (WB 0  x jE ) 2  x jE
exp[E gB , SiGe ( x) / kT ]
x
exp[E gB , SiGe ( x' ) / kT ]dx' dx

2 WB 0 WB 0

 (WB 0  x jE ) 2  x jE
exp[E ' gB , SiGe ( x) / kT ]
x
exp[E ' gB , SiGe ( x' ) / kT ]dx' dx

2 Wcap WB 0

 (WB 0  x jE ) 2  x jE
exp[E ' gB , SiGe ( x) / kT ]
x
exp[E ' gB , SiGe ( x' ) / kT ]dx' dx

SOLUTION 87
 2 WB 0 WB 0

 (WB 0  x jE ) 2 
Wcap
exp[E ' gB , SiGe ( x) / kT ]
x
exp[E ' gB , SiGe ( x' ) / kT ]dx' dx .

The first double integral in (3) is for x < Wcap, and its value is
2 Wcap WB 0

 (WB 0  x jE ) 2  x jE
exp[E ' gB , SiGe ( x) / kT ]
x
exp[E ' gB , SiGe ( x' ) / kT ]dx' dx (4)

2 Wcap WB 0

 (WB 0  x jE ) 2  x jE x
exp[E ' gB , SiGe ( x' ) / kT ]dx' dx

2 Wcap Wcap

 (WB 0  x jE ) 2  x jE x
exp[E ' gB , SiGe ( x' ) / kT ]dx' dx

2 Wcap WB 0

 (WB 0  x jE ) 2  
x jE Wcap
exp[E 'gB , SiGe ( x' ) / kT ]dx' dx

2 Wcap

 (WB 0  x jE ) 2 x jE
(Wcap  x)dx

2 Wcap WB 0  ( x'Wcap )(E g max  Eg 0 ) 


 (WB 0  x jE ) 2  Wcap 
exp
(WB 0  Wcap )kT
 dx' dx

x jE

2(Wcap  x jE ) 2

 (WB 0  x jE ) 2

2(WB 0  Wcap )(Wcap  x jE )  kT   Eg 0  Eg max 

   1  exp  .
 (WB 0  x jE ) 2  E 
 g max  Eg 0    kT 
The second double integral in (3) is for x > Wcap, and its value is
2 WB 0 WB 0

 (WB 0  x jE ) 2 Wcap
exp[E 'gB , SiGe ( x) / kT ]
x
exp[E 'gB , SiGe ( x' ) / kT ]dx' dx (5)

2 WB 0  ( x  Wcap )(E g max  E g 0 )  WB 0  ( x'Wcap )(E g 0  E g max ) 


 (WB 0  x jE ) 2  Wcap
exp 
 (WB 0  Wcap )kT
 
 x
exp 
 (WB 0  Wcap )kT
 dx' dx


2(WB 0  Wcap ) 2  kT   kT   E  E g max  

  1    1  exp g 0   .
2 
 (WB 0  x jE )  E g max  E g 0   E  
  g max  E g 0   kT  
Substituting (4) and (5) into (3), we have
t B ( SiGe, x jE ) 2(Wcap  x jE ) 2

t B ( Si, x jE )
x jE Wcap
 (WB 0  x jE ) 2

SOLUTION 88
 2(WB 0  Wcap )(Wcap  x jE )   E  Eg max 
  kT  1  exp g 0 
 (WB 0  x jE )  E  
 g max  Eg 0  
2
 kT 

2(WB 0  Wcap ) 2  kT   kT   E  E g max  

  1    1  exp g 0   .
2 
 (WB 0  x jE )  E g max  E g 0   E  
  g max  E g 0   kT  

10.13

n+ Ge
n+ emitter poly

Egmax = Eg0

p
x
0 WB

WB1

x
E gB ,SiGe ( x)  E g 0 x < WB1 (1)
WB1

 E g 0 WB > x > WB1.

1
 kT WB1  W  
J C 0 (SiGe)
   
1  exp E g 0 / kT   1  B1  exp E g 0 / kT   (2)
J C 0 (Si)  E g 0 WB  WB  

100000
dE/kT=2.5 dE/kT=5 dE/kT=10
10000
Current Ratio

1000

100

10

1
0 0.2 0.4 0.6 0.8 1
WB1/WB

 W 
V A (SiGe)

kT WB1
E g 0 WB

expE g 0 / kT   1  1  B1  .  (3)
V A (Si)  WB 

SOLUTION 89
 10000
3000

Early Voltage Ratio

1000
dE/kT=2.5
300
dE/kT=5
100
30 dE/kT=10

10
3
1
0 0.2 0.4 0.6 0.8 1
WB1/WB

t B (SiGe) 1  2kT  WB1  

2 2
 WB1  
 1       (4)
t B (Si)   WB  E g 0  WB  
 

1  2kT WB1  WB1 kT WB1     E g 0  

  1  1  exp   .
  E g 0 WB  WB E g 0 WB    kT  

1
0.9 dE/kT=2.5 dE/kT=5 dE/kT=10
0.8
Transit Time Ratio

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
WB1/WB

SOLUTION 90
10.14 The Ge profile of interest, from Fig. 10.19, is

Electron flow
(normal) Electron flow
(reverse)
Egmax
Ge
B

n+ p n+

x
0 WB

The current ratio, from Eq. (10.27) is

J C 0 (SiGe)  WB
 WB
. (1)
J C 0 (Si)
0
exp[E gB ,SiGe ( x) / kT ]dx

For the forward current ratio, we use in (1)

xE g max
E gB , SiBe , FORWARD ( x)  . (2)
WB

For the reverse current ratio, we should use in (1)

(WB  x)E g max
E gB , SiBe , REVERSE ( x)  (3)
WB

Substituting (3) into (1), we have

J C 0 (SiGe, reverse)   WB
 (4)
J C 0 (Si, reverse) WB  ( x  WB )E g max 
0
exp
 WB kT
dx

  WB

  E g max  WB  xE g max 
exp   exp dx
 kT   WB kT 
0

SOLUTION 91
   WB

  E g max  WB kT    E g max  
exp   exp   1
    kT
 kT  E g max    
  E g max / kT
 .
1  exp( E g max / kT )

10.15 From Fig. 10.21, the two transistors of interest are:

Egmax Egmax
Metal or metal silicide

Metal or metal silicide

Ge
Ge
B B
n+ emitter poly

n+ emitter poly
n+ n+
p p

x x
0 WB 0 WB
(a) (b)

From Eq. (10.32), we have for transistor b,

 E gB ,SiGe (WB ) 
exp 
V A (SiGe, b) DnB (SiGe, WB )  (WB )  kT  WB exp[E
V A (Si, b)

DnB (SiGe)  WB 0 gB , SiGe ( x ) / kT ]dx (1)

DnB (SiGe,WB )  (WB ) 1 WB  ( x  WB )E g max 


DnB (SiGe)  WB 0
exp
 WB kT
dx


DnB (SiGe,WB )  (WB ) exp(E g max / kT ) WB  xE g max 


DnB (SiGe)  WB 0
exp
 WB kT
dx


DnB (SiGe,WB )  (WB ) exp(E g max / kT ) WB  xE g max 


DnB (SiGe)  WB 0  WB kT
exp dx

DnB (SiGe, WB )  (WB ) exp(E g max / kT ) WB kT

DnB (SiGe)  WB E g max
exp(E g max / kT )  1  

SOLUTION 92
 
kT
E g max

exp(E g max / kT ) exp(E g max / kT )  1 . 
The VA ratio for transistor a is given by Eq. (10.33), namely
V A (SiGe, a)
V A (Si, a)

kT
E g max

exp(E g max / kT )  1 .  (2)

Since VA(Si, a) = VA(Si, b), we have

V A (SiGe, b)   E g max 
 exp  . (3)
V A (SiGe, a)  kT 
From Eq. (10.36), we have for transistor b
t B (SiGe, b) 2 WB WB

t B (Si, b)

 WB2 0
exp[E gB , SiGe ( x) / kT ] exp[E gB , SiGe ( x' ) / kT ]dx' dx
x
(4)

2 WB WB  ( x'WB )E g max 


 WB2 0
exp[E gB , SiGe ( x) / kT ] exp 
x
 WB kT
 dx' dx


2 WB kTWB    E g max   x E g max 


 WB2  E g max
1  exp  exp  exp[E gB , SiGe ( x) / kT ]dx
  kT   WB kT 
0

2 kTWB  (WB  x)E g max  2 WB kTWB


WB

 WB2 0 E g max
exp 
 WB kT
 dx 
  WB2 0 E g max
dx

2 kTWB  kTWB  E g max    E g max  

  exp  1  exp   WB 
 WB E g max  E g max
2
 kT    kT  

2 kT  kT  E g max    E g max  
  exp  1  exp   1 .
 E g max  E g max  kT    kT  

From Eq. (10.37), we have for transistor a

 
t B (SiGe, a )

2kT
 E g max
1 
kT

1  exp E g max / kT   .  (5)
t B (Si, a )  E g max 

Therefore, we have

SOLUTION 93
 kT  E g max     E g max 
exp  1  exp   1
t B (SiGe, b) E g max  kT    kT 
 . (6)
t B (SiGe, a ) kT    E g max 
1 1  exp 
E g max   kT 

Solutions to Chapter 11 Exercises

11.1 Ignoring the collector-substrate capacitance term, Fig. 9.26 gives the short-circuit-load
equivalent circuit

rb C rc
B C
CdCS,tot
C r' (1+g'm re) r'0

g'm vbe /(1+g'm re)

re

E E

From Eq. (9.161), we have for the small-signal baseemitter voltage

  vbe  ic re  ib (rb  re )  vbe  ic re ,
vbe (1)
where we have assumed the base current to be small compared to the collector current.
Similarly, we have
  vcb  ic rc  ibrb  vcb  ic rc   vbc  ic rc .
vcb (2)
Now, for the short-circuit load, vbc = vbe, so (2) gives
   vbe  ic rc .
vcb (3)
From the equivalent circuit, we have

g m  g m 
ic   
vbe  jC  vcb  jC   vbe  jC  rc ic , (4)
1  g m re 1  g m re 
which can be rearranged to give

SOLUTION 94
  g m 
1  jC r i
 c c   jC  vbe . (5)
1  g m re 
For a transistor having collector doping concentration of 2.51016 cm3, Fig. 3.5 shows that CdBC
is about 0.45 fF/m2. If the basecollector junction area is 2 m2 and the frequency is 1 GHz,
then CCdBC,tot103 mhos. For a bipolar transistor biased to have IC = 2 mA, gm =
qIC/kT = 7.7102 mhos. That is, we can assume C to be small compared to gm. We can
also show that Crc is small compared to unity. Therefore, (5) gives
g m
ic  vbe . (6)
1  g m re

Similarly, we have from the equivalent circuit

vbe
ib    jC  vbc
 jC vbe  . (7)
rb  re  r (1  g m re )

Since 1/gm is on the order of 13 , r = 0/gm is on the order of 1300 , which is large
compared to rb and re. Therefore, using (1) and (3), we can rewrite (7) as
vbe
ib   jC (vbe  ic re )  jC  (vbe  ic rc )
r (1  g m re )
j C  rc  C re g m 
(8)

 j C  C   
1
  vbe
 r (1  g m re ) 1  g m re 
ic g m
and  ( )   . (9)
 (1  g m re ) j C  C    j C  rc  C re g m
ib 1
r
(9) can be rearranged to give
g m r
 ( ) 

1  jr (C  C )  g m (re  rc )C
.
 (10)

In the high-frequency limit, (10) becomes

g m
 ( ) 

j (C  C )  g m (re  rc )C
,
 (11)

which drops to unity at a frequency fT given by

g m
1

2f T (C  C )  g m (re  rc )C  (12)

SOLUTION 95
 1 kT
or F  (C dBE ,tot  C dBC ,tot )  C dBC ,tot (re  rc ) , (13)
2 f T qI C

where we have used Eqs. (9.153), (9.156), (9.157), and (9.170), i.e.
 IC qI
g m   C,

 VBE kT

C  CdBC ,tot ,

C  CdBE ,tot  CDE

qI C
and C DE   F   F g m .
kT

11.2 NPN transistor with AE = 2 m2, ABC = 10 m2, deep emitter with NE = 1020 cm3, NB =
1018 cm3, WB = 100 nm, 0 = 100, and NC = 51016 cm3. Assume one-sided junction
approximation, with VBE = 0.8 V and VCB = 2 V.

(a) From Eq. (9.134), the diffusion capacitance due to holes in the deep emitter is
CDp  qI p pE / 2kT . For an npn transistor with deep emitter, Ip is the base current IB. From Eq.

(9.170), E is given by the relation CDp   E qIC / kT . That is E given by

 pE I B  pE
 E (deep emitter)   . (1)
2 IC 20

From Fig. 3.14(b), we have pE = 51010 s. Therefore

 pE
 E (deep  emitter)   2.5 10 12 s. (2)
2 0

From Eq. (9.133), the diffusion capacitance due to electrons in the narrow base is
CDn  2qI ntB / 3kT . For an npn transistor, In is the collector current IC. Therefore, from Eqs.

(9.166) to (9.171), we have

2
 B  tB . (3)
3
From Fig. 3.14(a), we have nB = 330 cm2/V-s. Therefore DnB = kTnB/q = 8.58 cm2/s.
Therefore, Eq. (10.21) gives

SOLUTION 96
 WB2
tB   5.83  1012 s, (4)
2 DnB

or  B  3.89 1012 s.

From Eq. (3.15), we have

2 si ( bi  VCB )
WdBC  . (5)
qN C

From Fig.3.4, we have bi  0.95 V. Therefore

2 si ( bi  VCB )
WdBC   2.77  105 cm. (6)
qN C

As discussed towards the end of Section [Link], the space-charge-region delay time is Wd/2vsat,
where Wd is the space-charge-region width and vsat is the saturated electron velocity. Therefore,
we have
WdBC
t BC  .  1012 s,
 138 (7)
2vsat

where we have used vsat = 107 cm/s. Similarly, we have

2 si ( bi  VBE )
WdBE  .  106 cm,
 171 (8)
qN B

WdBE
and t BE   0.86  1012 s, (9)
2 v sat

where we have used bi = 1.025 V from Fig. 3.4. Therefore, we have
 F   E   B   BE   BC   E   B  t BE  t BC  8.63 10 12 s. (10)

(b) From Eq. (3.13), we have

AE  si
C dBE ,tot   1.22 10 14 F, (11)
WdBE

ABC  si
and C dBC ,tot   3.75  10 15 F. (12)
WdBC

(c) For JC = 0.3qvsatNC = 2.4104 A/cm2, we have

SOLUTION 97
 IC  AE J C  4.8  104 A (13)
1 kT
and  F  (C dBE ,tot  C dBC ,tot )  (8.63 10 12  0.86  10 12 ) s (14)
2 f T qI C

or fT = 16.9 GHz.

(d) For (re + rc) = 50 , we have

(re  rc )C dBC ,tot  0.19 10 12 s, (15)

which is very small compared to F in this example. From (14) and (15), we have
1 kT
F  (C dBE  C dBC )  C dBC (re  rc )  (9.59 10 12  0.19  10 12 ) s,
2 f T qI C

(16)
or fT = 16.3 GHz.

11.3 Transistor in Exercise 11.2 is replaced by a pedestal-collector design, with NC(int) =

51016 cm3, and NC(ext) = 51015 cm3, i.e., NC(ext) is reduced by 10 times. The
corresponding areas are ABC(int) = AE = 2 m2 and ABC(ext) = (ABC  AE) = 8 m2. From (5) of
Ex. 11.2, we have

2 si ( bi  VCB ) 2 si ( 0.95  2)
WdBC (int)    2.77  105 cm. (1)
qN C (int) qN C (int)

Similarly, we have

2 si ( bi  VCB ) 2 si (0.89  2)
WdBC (ext )    8.67  10 5 cm. (2)
qN C (ext ) qN C (ext )

AE  si
Therefore, C dBC ,tot (int)   0.75110 15 F, (3)
WdBC (int)

ABC (ext ) si
CdBC ,tot (ext )   0.960  10 15 F (4)
WdBC (ext )

and CdBC ,tot  CdBC ,tot (int)  C dBC ,tot (ext )  1.7110 15 F. (5)

To repeat Ex. 11.2, we simply use this value of CdBC,tot in Ex. 11.2.

SOLUTION 98
 (a) Compared to Ex. 11.2, E, B, tBE and tBC all remain unchanged by the pedestal
collector design, since NC(int) is unchanged. Therefore, F is also unchanged.

(b) CdBE,tot is also unchanged, and CdBC,tot is given by (5) above.

(c) JC is determined by NC(int) and hence is unchanged. Therefore, IC is also

unchanged, and
1 kT
 F  (C dBE ,tot  C dBC ,tot )  (8.63 10 12  0.75  10 12 ) s (6)
2 f T qI C

or fT = 17.0 GHz.

(d) For (re +rc) = 50 ,

(re  rc )C dBC ,tot  0.09 10 12 s, (7)

1 kT
and F  (C dBE ,tot  C dBC ,tot )  C dBC ,tot (re  rc )  (9.38  10 12  0.09  10 12 ) s (8)
2 f T qI C

or fT = 16.8 GHz.

11.4 Same as Ex. 11.3, but with NC(int) = 11017 cm3. In this case, WdBC(int) is reduced and
JC is increased.

2 si ( bi  VCB ) 2 si ( 0.97  2)
WdBC (int)   .  105 cm
 196 (1)
qN C (int) qN C (int)

AE  si
C dBC ,tot (int)   1.06 10 15 F (2)
WdBC (int)

and CdBC ,tot  CdBC ,tot (int)  CdBC ,tot (ext )  (1.06  0.960)  10 15 F  2.02  10 15 F. (3)

(a) Compared to Ex. 11.3, E, B, and tBE all remain unchanged by the pedestal collector
design. But tBC is reduced because of the larger NC(int). We have from (1)
WdBC (int)
t BC   0.98  10 12 s. (4)
2 v sat

Therefore,

SOLUTION 99
  F   E   B  t BE  t BC  (2.5  3.89  0.86  0.98)  10 12 s  8.23  1012 s. (5)

(b) CdBE,tot is unchanged, and CdBC,tot is given by (3) above.

(c) JC is increased to JC = 0.3qvsatNC(int) = 4.8104 A/cm2, and IC is increased to

IC  AE J C  9.6  104 A. (6)
1 kT
Therefore,  F  (C dBE ,tot  C dBC ,tot )  (8.23 10 12  0.38  10 12 ) s (7)
2 f T qI C

or fT = 18.5 GHz.

(d) For (re + rc) = 50 ,

(re  rc )C dBC ,tot  0.10 10 12 s, (8)

1 kT
and F  (C dBE ,tot  C dBC ,tot )  C dBC ,tot (re  rc )  (8.6110 12  0.10 10 12 ) s
2 f T qI C

(9)
or fT = 18.3 GHz.

11.5 Fig. 10.6 is shown here for use in this exercise.

1E+4
SiGe/Si Improvement Factor

Current gain Early voltage Base transit time

1E+3

1E+2

1E+1

1E+0

1E-1
0 2 4 6 8 10
Egmax/kT

Fig. 10.6. Relative improvement factors for current gain, Early voltage, and base transit time of a SiGe-
base bipolar transistor over a Si-base bipolar transistor, as a function of the maximum base bandgap
narrowing. A linearly graded Ge profile is assumed. Also,  and  are set to unity.

SOLUTION 100
For Egmax = 100 meV, we have Egmax/kT  4. The incorporate of Ge does not change the
capacitances or the depletion layer thickness. It increases the current gain and reduces the base
and emitter transit times.

(a) Compared to Ex. 11.4, tBE and tBC are unchanged, but
 E (SiGe)  0 (Si) 1
  (1)
 E (Si)  0 (SiGe) 4

2.5 10 12

or  E (SiGe)  s  0.625 10 12 s, (2)
4
where we have used E(Si) = 2.51012 s from (2) of Ex. 11.2. Also, from the plot above, we
have
 B (SiGe) t B (SiGe)
  0 .4 , (3)
 B (Si) t B (Si)

or  B (SiGe)  0.4  3.89 1012 s  1.56 1012 s

(4)
where we have used B(Si) = 3.891012 s from (4) of Ex. 11.2. Therefore,
 F   E   B  t BE  t BC  (0.625  1.56  0.98  0.86)  1012 s  4.03  1012 s. (5)

(b) Compared to Ex. 11.4, CdBE,tot and CdBC,tot are unchanged.

(c) Compared to Ex. 11.4, JC and IC are unchanged, and

1 kT
 F  (C dBE ,tot  C dBC ,tot )  (4.03  10 12  0.38  10 12 ) s,
2 f T qI C

(6)
or fT = 36.1 GHz.

(d) Compared to Ex. 11.4, (rc + re)CdBC,tot is unchanged, and

SOLUTION 101
 1 kT
F  (C dBE ,tot  C dBC ,tot )  C dBC ,tot (re  rc )  (4.41 10 12  0.10 10 12 ) s (7)
2 f T qI C

or fT = 35.3 GHz.

11.6 Two-port network for an intrinsic transistor (parasitic resistances ignored) biased in
forward-active mode, adapted from Fig. 9.25 (small-signal hybrid-π model of a bipolar transistor
when parasitic resistances are neglected).
C

B C

v'1 i1 C r' r'0 v'2 i2

g'm v'be
E E

(a) The matrix representation for the two-port network is

i1     
  ˆ  v1   Y11 Y12  v1  .
i    v  Y  Y   v 
Y (1)
 2  2   21 22   2 
Referring to the network, we can write
1
i1  v1  j (C  C )v1  jC v2 ,
r

1
and i2  v2  g m v1  jC v2  jC v1.
r0

Compare with the matrix representation in (1), we can write

1 g
Y 11   j (C  C )  m  j (C  C ), (2)
r  0
Y 12   jC , (3)

Y 21  g m  jC , (4)

and
1
Y 22   jC . (5)
r 0

SOLUTION 102
In Eq. (2), we have used the fact that 1/ r  gm / 0 where β0 is the static current gain [see Eq.

(9.154)]. Note that the real part of Y 12 is zero because the base current is not a function of

collector voltage for a transistor biased in the forward-active mode.

(b) To derive the admittance matrix for an extrinsic transistor, we need to include the
IR drops in the resistors rb (= rbi + rbx), re and rc. The small-signal forms of Eqs. (9.161) and
(9.162) are
vbe  vbe  ib (rb  re )  ic re (6)

and
vce  vce  ib re  ic (rc  re ), (7)

which, in matrix notation, imply

 v1   v1   rb  re re   i1 
 v    v    r rc  re  i2 
. (8)
 2  2  e
Inverting the matrix in (1), and substituting into (8), we have

 v1  
 ˆ 1  rb  re re   i1 
v    Y    r    . (9)
 2    e rc re  
 i2 
The admittance matrix for the extrinsic transistor is
i1  ˆ v1  Y11 Y12  v1 
i   Y  v   Y Y  v  . (10)
 2  2   21 22   2 
Comparing (9) and (10) with Eqs. (8.71) and (8.72) for the case of a MOSFET, we see the
equivalencies between rb and Rg, between re and Rs, and between rc and Rd. Therefore, from Eq.
(8.73) for an extrinsic MOSFET, we can write, to first order in parasitic resistance, the
corresponding matrix elements for an extrinsic bipolar transistor as
Y11  Y 11  Y 112 rb  Y 12 Y 21 rc  (Y 11  Y 12 )(Y 11  Y 21 )re , (11)

Y12  Y 12  Y 11Y 12 rb  Y 12 Y 22 rc  (Y 11  Y 12 )(Y 12  Y 22 )re , (12)

Y21  Y 21  Y 11Y 21 rb  Y 21Y 22 rc  (Y 11  Y 21 )(Y 21  Y 22 )re , (13)

Y22  Y 22  Y 12 Y 21 rb  Y222 rc  (Y 12  Y 22 )(Y 21  Y 22 )re . (14)

SOLUTION 103
 (c) To obtained the cutoff frequency, we use the short-circuit-load configuration. For
the intrinsic transistor, that means v2  0. The intrinsic device current gain is

i2 Y 21
   . (15)
i1 v 2  0
Y 11

Therefore, the cutoff frequency of the intrinsic device is given by

Y g 2m  T2 C2

1  21  , (16)
Y 11 g 2m
  (C  C )
2 2

 02 T

which can be rearranged to give

g 2m
g 2m 
 02 g m
T  2 f T   , (17)
(C  C )  C
2 2
(C  C ) 2  C2

where we have use the approximation 1/β0 << 1. For a typical modern bipolar transistor which
has reduced Cμ (= CdBC,tot), we have Cπ (= CdBE,tot + CDE) >> Cμ. Therefore, (17) can be
simplified to give the commonly used approximation, i. e., Eq. (11.8), of
g m
T  2 f T  . (18)
C  C

For the extrinsic device, the short-circuit-load extrinsic current gain, to the first order of rb, rc,
and re, is

i2 Y21 Y 21  1  rbY 11  rcY 22  re (Y 11  Y 21 )(Y 21  Y 22 ) Y 21 
    
i1 v2  0
Y11 Y 11  1  rbY 11  rcY 12 Y 21 Y 11  re (Y 11  Y 12 )(Y 11  Y 21 ) Y 11 
(19)
Y
 21 1  (rc  re )(Y 22  Y 12 Y 21 Y 11 )  re (Y 12  Y 11Y 22 Y 21 ) .
Y 11

Notice that β is independent of rb. Noting that Re(Y 12 )  0 and that (rc  re )Re(Y 22 ) 1 , we
have, to the first order of rb, rc, and re,
Y 21
  1  (rc  re )Re(Y 12 Y 21 Y 11 )  re Re(Y 11Y 22 Y 21 )
Y 11
(20)
Y
 21 1  (rc  re )Re(Y 12 Y 21 Y 11 ) .
Y 11

SOLUTION 104
where we have used the fact that re Re(Y 11Y 22 / Y 21 ) 1 for a typical transistor in writing the
last equation. The cutoff frequency is obtained by setting |β| = 1, i.e.,

g 2m  T2C2  (rc  re )T2 g m (C2  0  C C  C2 ) 

1 1  . (21)
( g m  0 )2  T2 (C  C )2  ( g m  0 ) 2  T2 (C  C ) 2 
Noting that 1/β0 ≪ 1 and keeping only the terms to the first order of rc and re, we can rearrange
(21) to give
g 2m  T2C2  T2 (C  C )2  2(rc  re ) g m T2 (C  C )C , (22)

which in turn can be rearranged to give

1 (C  C ) 2  C2  g m (C  C )C 

 1  (re  rc ) . (23)
T g m  (C  C ) 2  C2 

Using (17) and (18) and the approximations involved in them, we can simplify (23) to the
commonly used approximation, i. e., Eq. (11.9), of
1 1 1
  (re  rc )C   (re  rc )CdBC ,tot . (24)
T  T  T

11.7 From Eq. (8.69) and the discussion in Section 8.5.4, fmax is given by the equation
2
4 Re(Y11 ) Re(Y22 )  Y12  Y21* . (1)

From Ex. 11.6, we have

Y12  Y 12  Y 11Y 12 rb  Y 12 Y 22 rc  (Y 11  Y 12 )(Y 12  Y 22 )re , (2)

and Y21  Y 21  Y 11Y 21 rb  Y 21Y 22 rc  (Y 11  Y 21 )(Y 21  Y 22 )re . (3)

Keeping only the terms to the first order of rb, rc, and re, we have
Re(Y11 )Re(Y22 )  g m  0 r 0  rb  g m2  02 r 0  g m  2C2  0   2 (C  C ) 2 r 0 
 rc  g m  0 r02  g m  2C2  0   2C2 r0  (4)
 re  2 g m2  0 r 0   2C2 r 0  g m  0 r 02  .
2
A similar expansion and approximation can be made for Y12  Y21* , and (1) can then be used to

determine fmax. However, if we make the large Early voltage approximation early on, the
procedure and the final expression for fmax can be greatly simplified. For a typical modern bipolar

SOLUTION 105
transistor operating at IC = 1 mA, Cμ (= CdBC,tot) > 1 fF, β0 > 100, r0  105  , g m  4 102 1

and fmax > 30 GHz. Thus, for a typical modern bipolar transistor, the terms of the second order in
1 r0 in (4) can be ignored compared to those of the first order in 1 r0 . That is, (4) can be

approximated by

Re(Y11 )Re(Y22 )  g m  0 r 0  rb  g m2  02 r 0  g m  2C2  0   2 (C  C ) 2 r0 

(5)
 rc  g m  2C2  0   2C2 r0   re  2 g 2m  0 r0   2C2 r0  .

Again, keeping only the terms to the first order of rb, rc, and re, we have
2
Y12  Y21*  g 2m  rb  4 g m  2 (C  C )C  2 g 3m  0 
 rc  4 g m  2C2  2 g m2 r0   re  2 g 3m  2 g m2  0 r 0  2 g m2 r0  (6)
 g 2m  rb  4 g m  2 (C  C )C  2 g 3m  0   4rc g m  2C2  2re g 3m ,

where in writing the last equation, we have used the fact that g m 1 r0 , re r0 1 and

rc r0 1 . Substituting (5) and (6) into (1), and again noting that g m 1 r0 , rc r0 , rb r0 ,

re r0 and β0 ≫ 1, we have the equation for fmax in the form

g m2 (1  2 g m rb  0  2 g m re )
 
(7)
 max
2
4 g m rb (C  C )C  (C  C ) 2 g m r0   rcC2  reC2 g m r0 .

To the first order of rb, rc, and re, (7) can be rearranged to give
g m (1  g m rb  0  g m re )
max  . (8)

4 g m rb (C  C )C  (C  C ) 2 g m r0   rcC2  reC2 g m r0 
For a typical modern bipolar transistor, g m r0  qVA kT  4 103 (see Section [Link]). While it

is true that Cπ = (CdBE,tot + CDE) ≫ Cμ, Cπ is not larger than Cμ by a factor of g m r0 without

severe base widening occurring. In other words, for a transistor operated at typical current
densities of interest, the terms proportional to 1 gm r0 in the denominator can be ignored. That

is, we have
g m (1  g m rb  0  g m r e )
max  . (9)
4 g m  rb (C  C )C  rcC2 

SOLUTION 106
Equation (9) is equivalent to that obtained by making the large Early voltage approximation
( 1/ r0  0 ) from the very beginning. The familiar approximation for ωmax is obtained if we

approximate the numerator by g m and keep only the term containing rb in the denominator, i.e.,

g m f T
max   , (10)
4 g m (C  C )C rb 8 C rb

where we have used Eq. (11.7) for fT .

11.8 A symmetric lateral npn transistor on SOI having the following physical parameters:
WE = 45 nm, WEC = 40 nm, NE = 4×1020 cm3, NB = 5×1018 cm3, LE = 2 m and tsi = 100 nm
(i.e., AE = 0.1 × 2 m2).
 F   E   B   BE   BC ,

2 I B WE2 W W (V  ,V  ) N n 2
 E (shallow emitter)   E B BE BC 2 B ieE ,
3I C 2 D pE 3DnB N E nieB

 , VBC
2t B WB2 (VBE  )
B   ,
3 3DnB
 )
WdBE (VBE
 BE  t BE 
2vsat
 )
WdBC (VBC
 BC  t BC  .
2vsat

(a) To calculate WB, WdBE, WdBC, E, B, BE, BC, and F for a given WEC, we need to
use the following equations (the equation numbers are those in the book) :
 ,VBC
WE C  WB (VBE  )  WdBE (VBE
 )  WdBC (VBC
 ). (10.72)

 )  2 si ( bi  VBE
WdBE (VBE  ) / q( N B  n), (10.78)

 )  2 si ( bi  VBC
WdBC (VBC  ) / q ( N B  n), (10.79)

where, n is given by
J C  qvsat n. (10.80)

[However, in the low-injection approximation, we ignore mobile space charge and assume
n = 0 in this exercise.]

SOLUTION 107
  si
Cd  . (3.26)
Wd

 )  LE CBC , fringe ,
C dBC ,tot  AE CdBC (VBC (11.19)

 )  LE CBE , fringe ,
CdBE ,tot  AE CdBE (VBE (11.20)

   
p0 Eg n0 Eg  nie2  ni2 exp Eg / kT .   (9.14)

The Klaassen unified model for the apparent bandgap-narrowing parameter

 2 
  N    N  
Eg  N   6.92 ln  17 
  
ln 17  
 0.5  meV. (9.18)
  1.3  10    1.3  10   
 
DnB for the p-type base can be obtained from Fig. 3.14.
The calculated results are shown in the table in (c).

(b) The collector current can be calculated from Eqs. (10.74) and (10.75), namely
 ,VBC
IC (VBE  )  IC 0 (VBE
 ,VBC
 ) exp(qVBE
 / kT ),

with
1
2
qAE DnB nieB  1  / kT )  
4n 2 exp(qVBE
 ,VBC
I C 0 (VBE  ) 1   1  ieB  1  .
 ,VBC
N BWB (VBE  )  4 N 2 
  B 
The calculated results are shown in table in (c).

(c) The calculated values of fT and fT , assuming (re  rc )  200  , are shown in

table.

 (V) WB
VBE WdBE WdBC E B BE BC F IC fT fT
(nm) (nm) (nm) (ps) (ps) (ps) (ps) (ps) (mA) (GHz) (GHz)
0.88 9.56 7.19 23.2 0.0274 0.0453 0.0359 0.116 0.225 0.18 197 156

0.90 9.94 6.82 23.2 0.0285 0.0490 0.0341 0.116 0.228 0.38 305 217

0.92 10.4 6.43 23.2 0.0297 0.0531 0.0321 0.116 0.231 0.77 418 269

SOLUTION 108
0.94 10.8 6.01 23.2 0.0310 0.0578 0.0300 0.116 0.235 1.5 508 303

0.96 11.3 5.56 23.1 0.0325 0.0635 0.0278 0.116 0.239 3.0 564 322

0.98 12.0 5.07 23.0 0.0343 0.0707 0.0253 0.115 0.245 5.6 589 329

Fig. 11.5 (vertical SiGe-base transistor) and comparison with calculated fT of symmetric lateral
transistor. The solid black line is fmax of SiGe device. The dash black line is fT of SiGe device.
The red dots are calculated fT of lateral device.

(d) A symmetric lateral transistor on SOI has significantly smaller capacitance than a
vertical transistor. At low currents, fT is determined by capacitance, and not by transit times. At
high currents, fT is determined by the transit times, which are not too different for a vertical
transistor and a lateral transistor. This explains the significantly higher fT for a lateral transistor at
lower currents.

SOLUTION 109
 Solutions to Chapter 12 Exercises

12.1 Consider a bipolar transistor biased to operate in the saturation region. Ignoring parasitic
resistances, the collector current is given by Eq. (12.5) and the base current is given by Eq.
(12.6), namely,
 I 
I C  I B 0 F e qVBE / kT   0 (1  e  qVCE / kT )  B 0 R e qVCE / kT  ,
 I B0F 

and

 I 
I B  I B 0 F e qVBE / kT 1  B 0 R e  qVCE / kT  .
 I B0F 

The current gain is

I B 0 R  qVCE / kT
 0 (1  e  qV CE / kT
) e
IC I BoF
  .
IB I
1  B 0 R e  qVCE / kT
I B0F

Plot the current gain  as a function of qVCE/kT for qVCE/kT = 0 to qVCE/kT = 4, using 0 = 100
and IB0R/IB0F = 1. Repeat using 0 = 100 and IB0R/IB0F = 10. This exercise shows that for
practical transistors the current gain is less than 1 only for very small VCE values.

100
90
80
70
60
IB0R/IB0F = 1
IC / IB

50
40 IB0R/IB0F = 10

30
20
10
0
-10
0 0.5 1 1.5 2 2.5 3 3.5 4
qVCE / kT

SOLUTION 110
 10
9
8
7
6
IB0R/IB0F = 1

IC / IB
5
4 IB0R/IB0F = 10

3
2
1
0
-1
0 0.05 0.1 0.15 0.2
qVCE / kT

SOLUTION 111