Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

An Improved Model For The I Mesfets by Evaluating The Potential Distribution Inside The Channel

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

J Comput Electron

DOI 10.1007/s10825-017-1010-y

An improved model for the I−V characteristics of submicron SiC


MESFETs by evaluating the potential distribution inside the
channel
M. M. Ahmed1 · M. Riaz1 · U. F. Ahmed2

© Springer Science+Business Media New York 2017

Abstract This paper presents a detailed mathematical 1 Introduction


model describing the I −V characteristics of submicron SiC
MESFETs. Poisson’s equation with appropriate boundary SiC offers low intrinsic carrier concentration, larger band gap
conditions has been solved to determine the potential distri- than Si, a high avalanche break down electric field, and high
bution inside the channel. The location (L 1 ) of the Schottky electron saturation velocity [1]. Thus, for high power appli-
barrier gate with a corresponding depletion layer width (u 1 ) cations, SiC MESFETs are preferred over GaAs MESFETs.
where the carrier’s velocity gets saturated has been evaluated. The current–Voltage (I −V ) characteristics of a SiC MES-
It has been demonstrated that, both L 1 and u 1 are bias depen- FET are the primary evaluation used to assess the quality
dent, and their values change by changing the drain biasing of a device. Because of the wide band gaps and high ther-
even after the onset of current saturation. This causes a modi- mal conductivity, SiC devices can handle relatively higher
fication in the depletion layer underneath the Schottky barrier power than GaAs/AlGaAs-based FETs, and are usually sub-
gate and, thus, changes the available channel cross-section jected to harsh environmental conditions. Moreover, in order
for the flow of current. It has been shown that finite out- to allow these devices to operate at microwave frequencies,
put conductance in the saturation region of operation, which it is mandatory that gate length, L g , of the device should be
is usually observed in submicron devices, can be explained in the submicron regime [2–4]. However, a submicron L g
with Schottky barrier depletion layer modification. The I −V device requires stringent fabrication constraints and, thus,
characteristics of submicron SiC MESFET are modeled and non-uniformities in the device characteristics are commonly
compared with conventional velocity saturation techniques, observed [5]. Like other MESFETs, submicron SiC MES-
where the depletion layer after the onset of current saturation FETs also exhibit short channel effects which include [6,7]:
has been treated as a constant. It is observed that the proposed (1) high output conductance in the saturation region of oper-
technique gave ∼15.9% improvement in the modeled char- ation; (2) compression in its transconductance and (3) shifts
acteristics of a submicron SiC MESFET. in the pinch-off voltage.
Short channel effects are classified as second order effects
Keywords SiC MESFET · Analytical model · Velocity by their nature, imposing additional challenges to engineers
saturation · I −V characteristics · Channel potential and scientists involved in the device modeling, making the
device’s analytical modeling a more demanding and cumber-
The authors would like to thank Prof. M. Sagheer and Prof. F. U.
some task.
Rehman for many helpful discussions. This paper discusses in detail the analytical modeling tech-
niques of submicron SiC MESFETs. In the first part of the
B M. M. Ahmed paper, the mobility of free carriers under a channel electric
mansoor@cust.edu.pk
field is discussed, whilst the second part of the paper deals
1 Department of Electrical Engineering, Capital University with the potential distribution inside the channel. In this sec-
of Science and Technology (CUST), Islamabad, Pakistan tion, regions have been identified underneath the Schottky
2 Department of Electrical Engineering, COMSATS Institute barrier gate responsible for the flow of the carriers initially
of Information Technology, Islamabad, Pakistan with constant mobility followed by the constant velocity

123
J Comput Electron

μ0
region. In the third part of the paper, the idea of depletion μ(E) =  1/β , (2)
layer modification has been introduced to achieve improved 1 + (μ0 E(x)/υs )β
accuracy in the device modeling, and finally, to establish the
validity of the proposed idea, the modeled and observed I −V where β is a constant having value 1 and υs is the saturation
characteristics have been presented. velocity. By using Eq. 2 one can write the field dependent
velocity, υ(E) of carriers moving in a SiC MESFET channel
as [12]
2 Field dependent mobility and velocity
μ0 E(x)
Consider an operating submicron SiC MESFET, as shown υ(E) =  1/β . (3)
in Fig. 1, the device is n-doped with channel doping 1.7 × 1 + (μ0 E(x)/υs )β
1017 cm−3 . The Schottky barrier gate is placed in recess,
which defines the desired target current, Idss and pinch-off
Figure 2a demonstrates the β dependent μ(E) for a SiC
voltage, VP . The depletion layer underneath the Schottky bar-
device; whereas, Fig. 2b illustrates the dependence of υ(E)
rier gate is divided into three regions [8]: Region-I is the one
on the chosen values of β as given in Eq. 3. The simulated
where carriers are moving below saturation velocity limit,
values of both υ(E) and μ(E) are identical to those which
and the mobility in this region can be taken as constant.
are reported in the literature for SiC devices [10,13–15].
Region-II is the region where carriers are moving with sat-
Figure 2 was plotted by using the data given in Table 1.
uration velocity and drain current of the device, under ideal
Plots of Fig. 2 show that by considering the material quality,
conditions, becomes independent of drain potential. Region-
device geometry and biasing conditions, α, β and Nr can be
III represents the extension of the depletion layer towards the
chosen appropriately to achieve a reasonable match between
drain side of the device. This region defines the maximum
modeled and experimental characteristics of SiC MESFETs.
electric field of the device and, thus, is responsible for its
breakdown.
It is an established fact that doping dependent mobility, 180
μ0 follows a Fermi–Dirac profile and can be expressed as (a)
[9,10] 150
= 0.9
μmax − μmin = 0.7
Mobility (cm /V-sec)

μ0 = μmin + , (1) 120


1 + (Nd /Nr )α = 0.5
2

90
where μmin and μmax are the minimum and maximum mobili-
ties, respectively. Nr and α represent empirical constants, and
60
Nd is the doping density. Field dependent mobility, μ(E) can
be expressed as [11]
30

Vgs 0
Vds (b)
2.0

1.6
Velocity ( 10 7 cm/sec )

Source Gate Drain

1.2
x'
I II III Depletion h(x)
Channel (a)
y a-h(x) 0.8
= 0.9
x
L1 L2 L3 = 0.7
0.4 = 0.5
Lg

Substrate
0.0
0 2 4 6 8 10 12 14 16 18
5
Field (10 V/cm)
Fig. 1 An operating submicron SiC MESFET with three distinct deple-
tion layer regions underneath a Schottky barrier gate Fig. 2 Field dependent a mobility and b velocity of a SiC MESFET

123
J Comput Electron

Table 1 Data used for simulation of Fig. 2   


Lg q Nd μ0 hd
Parameters Values Id dx + h x dh x
0 υs εs h0
μmin 40 cm2 /V-s (9)
μmax 950 cm2 /V-s 
(q Nd ) μ0 W
2 hd
Nr 1 × 1017 cm−3 = h x (a − h x ) dh x
εs
Nd 1.7 × 1017 cm−3 h0

α 2.6 or
 

q Nd μ0 a 2 h 2d h 20
Id Lg + −
3 Potential distribution inside the channel 2υs εs a2 a2

Consider Fig. 1, at Vds > 0 V, the drain current density, Jd   

inside SiC MESFET channel is given by Jd = q Nd υ, where (q Nd )2 μ0 W 3 h 2d h 20 h 3d h 30


= a 3 − −2 − .
υ is the low field velocity. The channel current can be defined 6εs a2 a2 a3 a3
as [16] (10)

Id = q Nd υW [a − h x ] , (4) The normalized depletion layer width, by using Eq. 7 at the


source, u 0 and drain, u d edges of the gate can be expressed
where W [a − h x ] is the available channel cross-section for respectively as:
the flow of current as shown in Fig. 1, q is the electronic  
charge, a represents epi-layer thickness and h x represents h0 1 2εs Vbi − Vg
depletion layer height at any point x. For low field regions, u0 = = (11)
a a q Nd
β can be substituted as unity in Eq. 3 and combining it with
Eq. 4 yields and
   
μ0 E (x) hd 1 2εs Vbi − Vg + Vd
Id = q N d W (a − h x ) (5) ud = = . (12)
1 + [μ0 E (x) /υs ] a a q Nd
or
Consider a submicron SiC MESFET with device variables
  that are defined in Table 2 [17]. For this device, the depletion
μ0 dV (x) dV (x)
Id 1 + = q Nd μ0 W (a − h x ) . (6) layer height, before the onset of current saturation, has been
υs dx dx
plotted by using Eq. 12 and shown in Fig. 3. It is obvious from
Since the depletion layer height underneath the Schottky bar- the figure that with increasing magnitude of Vg , the depletion
rier gate at point x is given by [16], layer terminates at lower Vd values because, it is linked with
saturation drain voltage, Vd(sat) , which itself is dependent on
2εs   Vg . Since the depletion layer is normalized with physical
h 2x = V (x) − Vg + Vbi , (7) channel thickness (a), its value below unity indicates that
q Nd
for given Vg biasing, a finite thickness of physical channel
where εs is the permittivity of SiC, Vg is the gate biasing, is still available for the conduction of current. Thus, current
V (x) is the potential at any point x because of the drain
biasing and Vbi is the built-in potential. From the above Table 2 Physical parameters of a submicron SiC MESFET [17]
expression, 2h x dh x = [2εs /q Nd ]dV (x) and substituting the
same in Eq. 6 gives Parameters Value Parameters Value

  ε0 8.85 × 10−12 Fm−1 Vbi 1.1 eV


μ0 q Nd q Nd εs 9.7ε0 Fm−1 γ 0.98
Id dx + h x dh x = q Nd μ0 W h x (a −h x ) dh x .
υs εs εs υs 2 × 107 cms−1 a 0.26 µm
(8) Ls 0.3 µm Lg 0.7 µm
β 0.87 Ld 0.8 µm
The total current can be calculated by integrating Eq. 8 over Nd 1.7 × 1017 cm−3 W 332 µm
its variable, x from 0 → L g , for which corresponding deple- q 1.6 × 10−19 C Rb 2 k
tion layer height changes from h 0 → h d as given below

123
J Comput Electron

1.0
where γ ∼ 0.98. Assuming that current saturation in a SiC
microwave MESFET is due to the velocity saturation, which
0.9
is attained by the carriers at a location underneath the Schot-
tky barrier gate where the depletion thickness is represented
Normalized Depletion Height

0.8
by h 1 , then one can write
0.7  
h1
Id(sat) = q Nd W γ υs (a − h 1 ) = q Nd W aγ υs 1 −
0.6 a
(18)
0.5 Vg = 0V Id(sat) = q Nd W aγ υs (1 − u 1 ) , (19)
Vg = -2V
0.4 Vg = -4V
Vg = -6V
where u 1 is normalized depletion under which the carriers are
Vg = -8V moving with saturation velocity. The start of u 1 , according to
0.3
Fig. 1, is at the boundary of L 1 . Thus, potential drop by using
0 2 4 6 8 10 Eq. 12 for the region defined by L 1 , referred to as Region-I,
Drain Voltage (V) is given by (see “Appendix”)
Fig. 3 Variation in Schottky barrier depletion before the onset of cur-
rent saturation for a submicron SiC MESFET V (L 1 ) = u 21 Vp − Vbi + Vg = Vd(sat) , (20)

where
saturation in submicron SiC MESFET is caused by velocity
saturation of the carriers contrary to long channel devices q Nd 2
Vp = a (21)
where current saturation is associated with the channel pinch- 2εs
off [16]. Combining Eqs. 10, 11 and 12 (see “Appendix”)
or
   
Ip 3 u 2d − u 20 − 2 u 3d − u 30 V (L 1 ) = Vp u 21 − u 20 . (22)
Id(linear) = , (13)
1 + z u 2d − u 20
Assuming that the current in Region-I is represented by Ic
where then by changing the integration limits of Eq. 9 from 0 to L 1
and h 0 to h 1 , one can write [18] (see “Appendix”)
q Nd a 2 μ0  
z= (14)
2εs L g υs Ip L g /L 1 3(u 21 − u 20 ) − 2(u 31 − u 30 )
Ic = . (23)
1 + z(L g /L 1 )(u 21 − u 20 )
and
Since at the end of Region-I or at the start of Region-II,
q 2 Nd2 μ0 W a 3 Ic = Id(sat) , therefore, equating Eqs. 19 and 23
Ip = . (15)
6εs L g 
q 2 Nd2 μ0 W a 3
From Eq. 3 one can write q Nd W aγ υs (1 − u 1 ) =
6εs L g

υ(E) E(x) (24)


= . (16)   

μ0 (1 + [μ0 E(x)/υs ]) L g /L 1 3 u 21 − u 20 − 2 u 31 − u 30
× .
1 + z L g /L 1 u 21 − u 20
When E(x) reaches to the saturation field, E s , υ(E) reaches
to γ υs [17], therefore,
Using Eq. 14, we have the length L 1 where the carriers are
attaining saturation velocity
γ υs Es Es
= =
μ0 1 + γ υs /υs 1+γ    

u 21 − u 20 − 2/3 u 31 − u 30 
(17) L1 = Lgz − u1 − u0 .
2 2
γ (1 − u 1 )
(1 + γ )γ υs 2γ υs
Es = ≈ , (25)
μ0 μ0

123
J Comput Electron


To assess the depletion height at L 1 , let us assume a (A + B) × Dsinky + Gy − Vg + Vbi
two-dimensional potential distribution inside the channel q Nd yh 1 q Nd y 2 (33)
expressed by Poisson’s equation [8,19–21]: = − − Vg + Vbi ,
εs 2εs

∂ 2 V (x, y) ∂ 2 V (x, y) q Nd which gives


+ =− . (26)
∂x 2 ∂y 2 εs
q Nd h 1
 G= and A = −B. (34)
Using the co-ordinate transformation x = x − L g , as εs
explained in Fig.
 1, one can define Eq. 26 as a sum of two
potentials W x , y = ξ (y) + V (x, y) such that Thus,

d2 ξ q Nd W x  , y = 2 ADsinh kx  sinky + E x  y
= (27) q Nd h 1 y
dy 2 εs + − Vg + Vbi . (35)
εs
and
Differentiating Eq. 35
∂ 2 W (x  , y) ∂ 2 W (x  , y)
+ =0 (28) 
∂x2 ∂ y2 ∂  
W x , y 
∂y y=h 1
such that (36)
q Nd h 1
= 2 AD sin h kx  k cos kh 1 + E x  + .
q Nd 2 εs
W x  , y = V (x, y) + y . (29)
2εs
Now, consider Eq. 29
Equation 28 can be solved by assuming: (1) potential is a  
continuous function of (x, y) and it diminishes abruptly out- ∂  ∂  q Nd h 1
W x  , y  = V (x, y) + . (37)
side the depletion layer; (2) the potential at the gate edge is ∂y y=h 1 ∂ y y=h 1 εs
nothing more than the applied gate potential or built-in poten-
tial if gate biasing is zero and (3) the Schottky barrier gate At the edges of depletion, i.e., y = h 1
is ideal and the entire gate biasing is consumed in changing
the gate depletion layer. A general solution of Eq. 28 is by ∂ V (x, y)
= 0. (38)
separation of variables ∂y
 
W x  , y = (Aekx + Be−kx ) × (Ccosky + Dsinky ), (30) Comparing Eqs. 36 and 37

√ q Nd h 1 q Nd h 1
where k = λ. The following is also a solution 2 ADksinh kx  coskh 1 + E x  + = . (39)
εs εs
  

W x  , y = Aekx + Be−kx (Ccosky + Dsinky ) Since, the above expression holds for all x  , therefore, E = 0
 
+ E x y + F x + Gy + H, (31) and k = (π/2h 1 ). Now, one can write Eq. 35
   
where A to H are constants to be determined by the boundary 
π x πy
W x , y = 2 ADsinh sin
conditions. At y = 0, potential is −(Vg + Vbi ) and it does not 2h 1 2h 1
depend on x  . This implies that C = 0 and F = 0. Thus, q Nd h 1 y
+ − Vg + Vbi . (40)
H = −(Vg + Vbi ). At L 1 , one can write the potential as εs
H = −(Vg + Vbi ) and the potential caused by the depletion.
Hence, If we assume that E changes to E s at x = L g or x  = 0, then,

q Nd   y     
∂  πx π πy
V (L 1 , y) = y h1 − − Vg + Vbi . (32) W x , y = 2 ADcosh × ×sin
εs 2 ∂x 2h 1 2h 1 2h 1
(41)
Here it is assumed that the depletion is increasing linearly    
towards the drain side under gradual channel approximation. ∂W x , y  π πy
  = AD sin . (42)
Comparing Eqs. 31 and 32 at x  = 0, we have ∂x   h1 2h 1
x =0

123
J Comput Electron

Equation 42 represents a field, which is E s , thus us assume that in Region-III, the depletion height is repre-
  sented by h 1 , as that of Region-II, which approaches to an
π πy insignificant value after distance L 3 . Under these assump-
E s = −AD sin . (43)
h1 2h 1 tions, Region-III potential can be approximated as
   
At the boundaries of depletion, i.e., y = h 1 leads to 2E s au 1 π L3 2E s au 1 π Lg
V (L 3 ) ≈ sinh ≈ sinh ,
π 2au 1 π 8au 1
Es h1 (51)
AD = . (44)
π
where L 3 represents the width of Region-III, which for sub-
Now from Eq. 29
micron devices can be written as L 3 ∼ L g /4 [16,22]. Now
q Nd y 2 the potential distribution inside the channel of a SiC MES-
V (x, y) = W x  , y − . (45) FET can be summarized as
2εs

Combining Eqs. 40, 44, 45 and using x  = L 1 − L g we have V (L 1 ) + V (L 2 ) + V (L 3 ) = Vd (52)


  or
2E s h 1 π Lg − L1 πy 

V (x, y) = sinh sin   2E au π Lg − L1


s 1
π 2h 1 2h 1 VP u 1 − u 0 +
2 2
sinh
π 2au 1
q Nd h 1 y q Nd y 2
+ − − (Vg + Vbi ). (46) (53)
εs 2ε s  
2E s au 1 π Lg
+ sinh = Vd .
Towards the drain side at x = L g and y = h 1 , one can have π 8au 1


2E s h 1 π Lg − L1 To solve Eq. 53 for u 1 , after making substitution for L 1 from


V L g, h1 = sinh Eq. 25, we define a function f (u 1 ) as
π 2h 1
   2E au 
s 1
q Nd h 21 f (u 1 ) = VP u 21 − u 20 + sinh
+ − Vg + Vbi . (47) π
2ε s     2

π Lg π Lgz u 1 − u 20 − (2/3) u 31 − u 30
× −
Since h 1 = au 1 and maximum value of x = L g ; therefore, 2au 1 2au 1 γ (1 − u 1 )− u 21 − u 20
Eq. 47 can also be written as
   

2E s au 1 π Lg
2E s au 1 π Lg − L1 + sinh − Vd = 0. (54)
V L g, h1 = sinh π 8au 1
π 2h 1
Now the Newton’s iterative method [23,24], the defini-
+ VP u 21 − Vg + Vbi . (48)
tion u (n+1) = u n − f (u n )/D f (u n ) can be employed to find
u 1 . This evaluation would be for Vd ≥ Vd(sat) and for the
But Vg + Vbi = VP u 20 , therefore,
devices having short L g to ensure that the observed current

saturation is caused by velocity saturation of the carriers.
  2E au π Lg − L1
s 1 Thus, the proposed model may not be valid for long-channel
V L g , h 1 = VP u 1 −u 0 +
2 2
sinh .
π 2au 1 devices, wherein the current saturation is usually observed
(49) after attaining the pinch-off condition. Further, the validity
of the proposed model will be limited to the biasing condi-
In Eq. 49, VP (u 21 − u 20 ) represents the potential in Region- tions that maintain the device operation prior to the onset of
I, where carriers are moving below the saturation velocity; avalanche breakdown.
whereas, the second term of Eq. 49 represents the potential Figure 4 shows potential accumulation in all the three
drop in Region-II, as shown in Fig. 1, where carriers are regions of the device. Figure 4a represents potential drop in
moving with the saturation velocity thus, for L 2 = L g − L 1 Region-I for a SiC submicron MESFET. The figure is plotted
 
2E s au 1 π L2 using Eq. 22 and it shows a non-uniform variation in between
V (L 2 ) = sinh . (50) two consecutive curves while changing Vg from 0 to −8 V.
π 2au 1
This can be explained by the nonlinear dependence of h on
In Region-III, there is no gate biasing and the potential of Vg as evident from Eq. 7. Also in Fig. 4a, the magnitude
Region-II will decay exponentially as a function of x. Let of the potential in Region-I is decreasing by increasing Vd .

123
J Comput Electron

7
(a) Region-I Vg = -8V
mate the potential drop of Region-III by employing Eq. 51
Depletion Layer Potential (V) 6
Vg = -6V as shown in Fig. 4c. It is obvious from the figure that there
Vg = -4V
Vg = -2V is an enhancement of Region-III potential after the onset of
5
Vg = 0V current saturation such that its magnitude is relatively less
dependent on the device biasing. For the entire change in
4 Vd and Vg , the total observed variation in V (L 3 ) ∼ 0.5 V.
But, the important point to note is that the potential drop in
3 Region-III is ∼4 V which is a significant value and cannot
be neglected in the device modeling.
2
It is pertinent to mention here that the potential variation
20
(b) Region-II with respect to Vd , underneath the Schottky barrier gate, is a

Depletion Layer Potential (V)


continuous function for all the three regions discussed hith-
15 erto. In Region-I, prior to velocity saturation, some variation
Vg = -8V
Vg = -6V in the magnitude of potential is observed which presumably
Vg = -4V is due to the drifting of h 1 towards the source side with
Vg = -2V 10
Vg = 0V increasing values of Vd . Whereas, in Region-II, the poten-
tial increases linearly till the termination of the region. On
5 the other hand, in Region-III, as evident from Fig. 4c, the
potential variation as a function of Vd is nominal. However,
4.3 0
its magnitude is significant, thus, cannot be overlooked in the
(c) Region-III device modeling.
Depletion Layer Potential (V)

4.2 A two-region analytical model as proposed by Murray


Vg = -8V et al. [21] works reasonably well for low current (less than
Vg = -6V
4.1
Vg = -4V ∼10 mA) SiC MESFETs. However, it fails for large current
Vg = -2V
4.0 Vg = 0V
devices (∼50 mA or above) which is usually the case for
SiC power MESFETs. By incorporating the potential drop of
3.9 Region-III, Zhu et al. [17] have modified the Murray model
and demonstrated its improved performance over the two-
3.8
region model for relatively large current SiC MESFETs. In
3.7 this study, the performance of the Zhu et al. model is further
0 5 10 15 20 25 improved by incorporating the concept of depletion layer
Drain Voltage (V)
modification after the onset of current saturation, which was
Fig. 4 Depletion layer potential underneath the Schottky barrier gate taken as constant in Zhu model.
of a submicron SiC MESFET

This can only be possible if u 1 of Eq. 22 exhibits a decreasing 4 Modeled I−V characteristics
trend with increasing magnitude of Vd . This is because of the
fact that with increasing Vd the carriers will attain saturation A submicron SiC MESFET with device parameters as listed
velocity relatively early thus, reducing the overall length of in Table 2 has been selected [17]. A MATLAB code is devel-
L 1. oped using Eqs. 13 and 19 for the conditions shown below.
Figure 4b is attained using Eq. 50 and it represents the
potential drop for Region-II, where carriers are moving with
 
saturation velocity, υs . An examination of the figure shows Ip 3 u 2d −u 20 − 2 u 3d − u 30
that by increasing Vd , there is a linear increase in potential Id(linear) = for Vd < Vd(sat)
1 + z u 2d − u 20
accumulation in Region-II. The start of the drain voltages in
and
Fig. 4b is after meeting Vd = Vd(sat) condition, and the total
magnitude of the potential is dependent upon both Vd and Id(sat) = q Nd W aγ υs (1 − u 1 ) for Vd > Vd(sat) .
Vg . Thus, for a higher magnitude of Vg , Region-II will hold (55)
more potential after the onset of current saturation, which is
evident from Fig. 4b. In Eq. 55, u 1 is a constant as defined by the Zhu and Murray
Assuming that the extension of the depletion layer towards models, and the same is evaluated at Vd = Vd(sat) . On the
the drain side of a submicron SiC MESFET follows quarter contrary, our study showed that u 1 is a variable, which can be
circle approximation [22], i.e., L 3 = L g /4. One can esti- evaluated using Eq. 53 for Vd ≥ Vd(sat) and the evaluated data

123
J Comput Electron

can subsequently be employed to assess Id(sat) by involving The analytical models presented in [17,21] assumed that the
Eq. 55. depletion layer at location L 1 remains constant after the onset
Further, the effects of the parasitic resistances Rs , Rd and of current saturation, and any increase in the current after-
Rb have been incorporated by using wards, is associated with buffer layer conductance, which has
been incorporated in the model by Rb . Apart from the buffer
Vds = Vd + Id (Rs + Rd ), (56) layer conductance, our study showed that there is an increase
in the potential drop after the onset of current saturation in
where Region-II, as well as in Region-III as evident from Fig. 4,
which can potentially affect the modeled output characteris-
Ls
Rs = + Rc tics of a SiC MESFET.
Nd qμaW By involving Eqs. 25 and 53, one can evaluate the behavior
(57)
(L d − L 3 ) of L 1 and u 1 as a function of the applied potential. Figure 6
Rd = + Rc ,
Nd qμaW shows the variation in u 1 after the onset of the current satura-

where L s and L d represent separation between source-gate


and drain-gate, respectively, and Rc is the contact resistance. 1.00

For μ = 300 cm−2 (Vs)−1 , and Rc ∼ 1.5 , the calculated


values of Rs and Rd were 5.7 and 10.3  respectively. The 0.95
drain-to-source current, Ids was then evaluated by using

Depeltion Layer ( u 1)
Ids = Id + Vds /Rb (58) 0.90

and
0.85
Vgs = 0V
Vgs = Vg + Ig (Rs ), (59) Vgs = -2V
0.80 Vgs = -4V
where for submicron devices Ig ∼ 10−9 A [6] and the buffer Vgs = -6V
Vgs = -8V
layer resistance, Rb , was evaluated using background doping
as 1015 cm−3 [16] and found as Rb ∼ 2 k.
0.75

The modeled and observed output characteristics of the 0 5 10 15 20 25


chosen device are shown in Fig. 5. It is pertinent to men- Drain-to-Source Voltage (V)
tion here that the modeled output characteristics showed
some discrepancies, especially near the pinch-off region. Fig. 6 Variation in depletion layer after the onset of current saturation
in a SiC MESFET

300 1.2

Vgs = 0V
Normalized Saturation Length ( L 1 /L g )

250
1.0 Vgs = 0V
Vgs = -2V
Drain-to-Source Current (mA)

Vgs = -4V
200
Vgs = -2V Vgs = -6V
0.8 Vgs = -8V
150
Vgs = -4V
100 0.6

Vgs = -6V
50
0.4
Vgs = -8V
0
0 5 10 15 20 25 5 10 15 20 25
Drain-to-Source Voltage (V) Drain-to-Source Voltage (V)

Fig. 5 Modeled (−) and measured (•) output characteristics of a sub- Fig. 7 Shows shifting of saturation length L 1 towards source side of
micron SiC MESFET the gate after the onset of current saturation in a SiC MESFET

123
J Comput Electron

Vgs attain the saturation velocity after traveling the entire Schot-
Vds tky barrier gate, i.e., up to Region-III of the device. By
increasing Vds , the x-directed field inside the channel also
increases, resulting in the shifting of L 1 toward Region-II,
which brings L 1 /L g < 1. Assuming that the current satu-
Source Drain
Gate ration in a short channel SiC MESFET is due to velocity
saturation [17,21], one can conclude that the point where
Depletion
these carriers are attaining the saturation velocity is bias
Channel (a) dependent, and it shifts towards the source side with increas-
y ing magnitude of Vds .
x The shifting of L 1 towards the source side of the deple-
tion as observed in Fig. 7, can be associated with the fact
Fig. 8 A cross-sectional view of an operating SiC MESFET where a that by increasing Vds the saturation point is gradually shift-
dotted line shows reduction in gate depletion layer with increased drain
biasing
ing towards the source and u 1 is thus decreasing as seen in
Fig. 6. As a result, there is an increase in the channel cross-
300 sectional area available for the flow of current as illustrated
Vgs = 0V
in Fig. 8. It has been shown in the figure that with the shift-
250
ing the saturation point towards the source side of Schottky
barrier gate, the channel is slightly widened, and the same
is represented in the figure by the dotted line. This plausible
Channel Current (mA)

200 explanation will lead to an obvious conclusion that finite out-


Vgs = -2V
put conductance in the saturation region of operation may be
150 associated with depletion layer modification with increasing
Vgs = -4V drain biasing [25].
By incorporating the depletion layer modification concept
100
as discussed hitherto, the device of Table 2 is again modeled,
Vgs = -6V and the result is shown in Fig. 9. A visual comparison of
50 Figs. 5 and 9 clearly shows a tangible improvement in the
Vgs = -8V modeled characteristics of Fig. 9, especially in the saturation
0 region of operation. In Fig. 5 at Vgs ∼ − 8 V, i.e., near to
0 5 10 15 20 25 the pinch-off, there is a relatively higher discrepancy in the
Drain-to-Source Voltage (V)
observed and modeled characteristics. By looking at Fig. 9
Fig. 9 Output characteristics of a submicron SiC MESFET: measured for the same curve (Vgs = −8 V), one can clearly gather that
(•) and modeled (−) with the proposed depletion layer modification this discrepancy has been adequately addressed.
technique Table 3 provides a comparison between the conventional
approach and the one which has been adopted in this research.
The table presents RMS errors for each Ids curve for the
tion. This figure clearly depicts that the depletion layer at the
device under discussion. The data of the table clearly demon-
point of saturation does not stay constant after the onset of
strate that the proposed technique offers an improvement in
current saturation; rather, it decreases with increasing values
SiC MESFET modeling. From average RMSE values one
of Vds .
can calculate that the proposed technique is 15.9%, better
Additionally, it is shown in Fig. 7 that the saturation length
than its counterpart.
L 1 is dependent on both Vds and Vgs of the device. A careful
examination of Fig. 7 reveals that for L 1 /L g > 1, carriers

Table 3 Comparison of RMS errors for conventional and proposed modeling technique for the device parameters listed in Table 2
Model Vgs = 0 V Vgs = −2 V Vgs = −4 V Vgs = −6 V Vgs = −8 V Avr. RMSE

Zhu et al. [17] 12.36 × 10−3 9.48 × 10−3 5.52 ×10−3 3.13 × 10−3 5.28 × 10−3 7.15 × 10−3
Proposed approach 12.36 × 10−3 8.57 × 10−3 4.60 ×10−3 2.87 × 10−3 1.67 × 10−3 6.01 × 10−3

123
J Comput Electron

5 Conclusion By re-arranging Eq. A.4, we get

In this paper an analytical model has been developed for sub-  2 2  


q Nd μ0 W a 3 /6εs L g 3 u 2d − u 20 − 2 u 3d − u 30
micron SiC MESFETs. The potential distribution inside the Id = .
1 + q Nd μ0 a 2 /2εs L g υs u 2d − u 20
channel has been evaluated by solving Poisson’s equation
with appropriate boundary conditions. It has been demon- (A.5)
strated that the location underneath the Schottky barrier gate,
where the carriers attain saturation velocity, is bias depen- Assuming
dent, and it moves towards the source side with increasing
drain voltage. Those results demonstrate that the depletion q 2 Nd2 μ0 W a 3
Ip = (A.6)
layer after attaining velocity saturation cannot be treated as 6εs L g
constant. It has been shown that, after the onset of current sat-
uration, there is a modification in the depletion layer, which and
in turn increases the available channel cross-section for the
flow of current. This contributes to a positive slope of the q Nd μ0 a 2
z= . (A.7)
output characteristics of the device, which has been conven- 2εs L g υs
tionally associated with the buffer layer conductance alone.
By incorporating depletion layer modification concepts, the Equation A.5, for the linear region, can be written as
I −V characteristics of submicron SiC MESFETs have been
modeled and an improvement of ∼15.9% is observed relative  
Ip 3 u 2d − u 20 − 2 u 3d − u 30
to the best reported technique. Id(linear) = . (A.8)
1 + z u 2d − u 20

To achieve Eq. 20, one can use Eq. 12 wherein u d has been
Appendix changed with u 1 and correspondingly h d with h 1
To achieve Eq. 13, consider Eq. 10  
h1 1 2εs Vbi − Vg + Vd
 
u1 = = . (A.9)
q Nd μ0 a 2 h 2d h 20 a a q Nd
Id Lg + −
2υs εs a2 a2
Squaring both sides
  
 
(q Nd )2 μ0 W 3 h 2d h 20 h 3d h 30 q Nd a 2
= a 3 − −2 − . u 21 = Vbi − Vg + Vd . (A.10)
6εs a2 a2 a3 a3 2εs
(A.1)
Using definition
As defined by Eqs. 11 and 12
q Nd a 2
  Vp = , (A.11)
h0 1 2εs Vbi − Vg 2εs
u0 = = (A.2)
a a q Nd
Equation A.10 can be written as
and
  u 21 Vp = Vbi − Vg + Vd . (A.12)
hd 1 2εs Vbi − Vg + Vd
ud = = . (A.3)
a a q Nd At u 1 carriers are attaining saturation velocity with corre-
sponding voltage, Vd(sat) , which at location L 1 can also be
Substituting u 0 and u d in Eq. A.1
referred to as V (L 1 ), therefore,
 
q Nd μ0 a 2  2 V (L 1 ) = u 21 Vp − Vbi + Vg = Vd(sat) . (A.13)
Id L g + ud − u0
2
2υs εs
(q Nd )2 μ0 W 3   2   
To achieve Eq. 23, consider Eq. 9, and by changing limits
= a 3 u d − u 20 − 2 u 3d − u 30 .
6εs for variable x, from 0 to L 1 and the corresponding depletion
(A.4) layer as h 0 to h 1 , one can write channel current, Ic as

123
J Comput Electron

  
L1 q Nd μ0 h1
3. Riaz, M., Ahmed, M.M., Munir, U.: An improved model for current
Ic dx + h x dh x
0 υs εs h0
voltage characteristics of submicron SiC MESFETs. Solid State
Electron. 121, 54–61 (2016)
4. Saremi, M., Afzali-Kusha, A., Mohammadi, S.: Ground plane fin-
 shaped field effect transistor GP-FinFET: a FinFET for low leakage
(q Nd )2 μ0 W h1
power circuits. Microelectron. Eng. 95, 74–82 (2012)
= h x (a − h x ) dh x (A.14) 5. Ahmed, M.M.: Effects of active-channel thickness on submicron
εs h0
GaAs metal semiconductor field-effect transistor characteristics. J.
Vac. Sci. Technol. B 16(3), 968–971 (1998)
or 6. Ahmed, M.M., Ahmed, H., Ladbrooke, P.H.: Effects of interface
 
states on submicron GaAs metal-semiconductor field-effect tran-
q Nd μ0 a 2 h 21 h 20 sistors assessed by gate leakage current. J. Vac. Sci. Technol. B
Ic L1 + − 13(4), 1519–1525 (1995)
2υs εs a2 a2 7. Ahmed, M.M.: Compression in transconductance at low gate
voltages in submicron GaAs metal semiconductor field-effect tran-
  
sistors. J. Vac. Sci. Technol. B 15(6), 2052–2056 (1997)
8. Khemissi, S., Azizi, C.: A two-dimensional analytical modeling of
(q Nd )2 μ0 W 3 h 21 h 20 h 31 h 30
= a 3 − −2 − . the current-voltage characteristics for submicron gate-length GaAs
6εs a2 a2 a3 a3 MESFETs. Int. J. Eng. Technol. 12, 27 (2012)
9. Caughey, D.M., Thomas, R.E.: Carrier mobilities in silicon empir-
(A.15)
ically related to doping and field. Proc. IEEE 55(12), 2192–2193
(1967)
Substitute u 0 and u 1 in Eq. A.15, as per the definition of 10. Mnatsakanov, T.T., Levinshtein, M.E., Pomortseva, L.I., Yurkov,
Eqs. A.2 and A.9 S.N.: Carrier mobility model for simulation of SiC-based electronic
devices. Semicond. Sci. Technol. 17(9), 974–977 (2002)
11. Lu, H., Zhang, Y., Zhang, Y., Zhang, T.: A comprehensive model
 
q Nd μ0 a 2  2 of frequency dispersion in 4H-SiC MESFET. Solid State Electron.
Ic L1 + u1 − u0
2
53(3), 285–291 (2009)
2υs εs 12. Khan, I.A., Cooper, J.A.: Measurement of high-field electron
transport in silicon carbide. IEEE Trans. Electron Devices 47(2),

(q Nd )2 μ0 W 3   2    269–273 (2000)
= a 3 u 1 − u 20 − 2 u 31 − u 30 . 13. Lv, H., Zhang, Y., Zhang, Y., Yang, L.A.: Analytic model of I-
6εs V characteristics of 4H–SiC MESFETs based on multiparameter
(A.16) mobility model. IEEE Trans. Electron Devices 51(7), 1065–1068
(2004)
14. Bertilsson, K., Nilsson, H.E., Hjelm, M., Petersson, C.S., Käckell,
Using Eqs. A.6 and A.7 P., Persson, C.: The effect of different transport models in simu-
lation of high frequency 4H–SiC and 6H–SiC vertical MESFETs.
   Solid State Electron. 45(5), 645–653 (2001)
Ic L 1 + z L g u 21 − u 20 15. Roschke, M., Schwierz, F.: Electron mobility models for 4H, 6H,
     and 3C SiC MESFETs. IEEE Trans. Electron Devices 48(7), 1442–
= Ip L g 3 u 21 − u 20 − 2 u 31 − u 30 (A.17) 1447 (2001)
16. Ladbrooke, P.H.: MMIC Design: GaAs FETs and HEMTs. Artech
House, Boston (1989)
or 17. Zhu, C.L., Rusli, Tin, C.C., Yoon, S.F., Ahn, J.: A three-region
  analytical model for short-channel SiC MESFETs. Microelectron.
I p L g 3 u 21 − u 20 − 2 u 31 − u 30 Eng. 83(1), 96–99 (2006)
Ic =   . (A.18) 18. Tsap, B.: Silicon carbide microwave field-effect transistor: effect
L 1 + z L g u 21 − u 20
of field dependent mobility. Solid State Electron. 38(6), 1215–1219
(1995)
Equation A.18 can also be written as 19. Lakhdar, N., Djeffal, F.: A two-dimensional analytical model of
subthreshold behavior to study the scaling capability of deep sub-
  micron double-gate GaN-MESFETs. J. Comput. Electron. 10(4),
I p (L g /L 1 ) 3 u 21 − u 20 − 2 u 31 − u 30
Ic =   . (A.19) 382–387 (2011)
1 + z(L g /L 1 ) u 21 − u 20 20. Chang, C.S., Day, D.Y.S.: Analytic theory for current–voltage char-
acteristics and field distribution of GaAs MESFET’s. IEEE Trans.
Electron Devices 36(2), 269–280 (1989)
21. Murray, S.P., Roenker, K.P.: An analytical model for SiC MES-
References FETs. Solid State Electron. 46(10), 1495–1505 (2002)
22. Enoki, T., Sugitani, S., Yamane, Y.: Characteristics includ-
1. Mahabadi, S.E.J., Moghadam, H.A.: Comprehensive study of a ing electron velocity overshoot for 0.1-μm-gate-length GaAs
4H–SiC MES–MOSFET. Physica E Low Dimens. Syst. Nanos- SAINT MESFET’s. IEEE Trans. Electron Devices 37(4), 935–941
truct. 74, 25–29 (2015) (1990)
2. Ahmed, M.M.: An improved method to estimate intrinsic small 23. Chun, C.: Iterative methods improving newton’s method by the
signal parameters of a GaAs MESFET from measured DC charac- decomposition method. Comput. Math. Appl. 50(10), 1559–1568
teristics. IEEE Trans. Electron Devices 50(11), 2196–2201 (2003) (2005)

123
J Comput Electron

24. Wang, X., Qin, Y., Qian, W., Zhang, S., Fan, X.: A family of newton 25. Ahmed, M.M.: Schottky barrier depletion modification-a source of
type iterative methods for solving nonlinear equations. Algorithms output conductance in submicron GaAs MESFETs. IEEE Trans.
8(3), 786–798 (2015) Electron Devices 48(5), 830–834 (2001)

123

You might also like