You are on page 1of 7

Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

Analytical Modeling of Potential and Electric


Field Distribution and Simulation of Large
Signal Parameters for Dual Channel Dual
Material Gate AlGaN/GaN High Electron
Mobility Transistor
Rahis Kumar Yadav1, Pankaj Pathak2, R M Mehra3
Department of Electronics and Communication Engineering, School of Engineering and Technology,
Sharda University, Gr. Noida, India


Abstract—A two dimensional analytical model based on solution thermal stability, high breakdown benefits of electric field and
of 2-DEG charge expressions for potential and electric field development of polarization induced high sheet charge density
distribution in dual channel dual material gate AlGaN/GaN high of 2DEG at hetrointerface [2]. This kind of hetrostructures can
electron mobility transistor (DCDMG HEMT) is developed to provide much higher output power density compared with
demonstrate unique features of this device structure in conventional AlGaAs/GaAs-HEMT [3]-[6]. Long et al [6][7]
suppressing short channel effects (SCEs), reduction in overall
has fabricated and characterized AlGaAs/GaAs HFET
drain current collapse with further improvement in device
current drive capability. The model accurately predicts the structure, named as dual material gate HFET, in which two
channel potential and electric field distribution along the device different metals of different workfunction were selected and
channels. It is observed that work function difference of dual merged into a single gate by making them contact laterally.
metal gate leads to a screening effect of the drain potential The work function of metal region M1 was chosen to be
variation by second metal gate portion near the drain resulting in greater than that of metal region M2 leading to reduction in
suppressed drain induced barrier lowering (DIBL) and hot SCEs and supressed DIBL. This structure got improvement in
carrier effect. This further improves carrier transport efficiency the carrier transport efficiency due to step function in the
due to improved uniform distribution of electric field along both channel potential distribution.
the channels of device. Improvement in device current drive
In order to enhance the current drive capability of HEMTs,
capability is due to incorporation of lower channel in the device.
The resultant double channel dual material gate device get efforts were made by scientists to construct AlGaN/GaN
improved transconductance that suppresses current collapse double-channel HEMTs. However, due to the additional
that normally occurs in GaN based conventional HEMTs. Since AlGaN barrier between the two channels, accesses to the
current collapse mainly occurs in upper channel closer to the lower carrier channel w e r e usually inefficient thus double-
device surface while graded profile lower channel in the device is channel behaviors wa s not pronounced [1]. Fabrication of
at far distance resulting in minimal current collapse. The an AlGaN/GaN/AlGaN/GaN epilayer structure using a lower
proposed model also considers trap concentration, a responsible AlGaN barrier layer with smaller layer thickness and low and
factor, for current collapse for device analysis and graded Al composition was proved to be an effective
characterization. Finally theoretical prediction of our model
approach to implement the double-channel HEMT with the
results compared with device simulation results and experimental
data. Results are found to be in close agreement giving validity to second channel of high electron density and acceptable access
our proposed model. resistance Roanming Chu et al [1]. The resultant AlGaN/GaN
Keywords—2DEG-two dimensional electron gas, SCE- short double-channel HEMTs exhibit device performance better
channel effect, DCDMG-Dual Channel Dual Material Gate, than or comparable to baseline AlGaN/GaN HEMTs in all
HEMT-High Electron Mobility Transistor, AlGaN/GaN-Aluminum aspects. However, so for no analytical model is developed to
Gallium Nitride/Gallium Nitride. incorporate the effect of double channel with dual material
gate using AlGaN/GaN hetero-interface. In our DCDMG
I. INTRODUCTION AlGaN/GaN HEMT model we considered multilayer device
structure which provide better device performance in
B eing an important device for high frequency, high power
application AlGaN/GaN hetrostructure based High
Electron Mobility Transistors attracting a lot of attention of
comparison to base line GaN HEMTs structures.

device scintists word wide. Better performance of II. MODEL FORMULATION


AlGaN/GaN hetrostructures relies on high saturation velocity, 2.1. 2-D Analytical model for Channel Potential Analysis:
Following Fig.1 shows schematic cross sectional view of
DCDMG AlmGa1.mN / GaN HEMT with Al mole fraction
www.rsisinternational.org/IJRSI.html Page 49
Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

m=0.30, L1 and L2 are lengths of the laterally merged gate direction for upper channel can be approximated by following
metals M1 and M2 respectively having different workfunction parabolic function as
and forming single gate for the device. The source and drain 2
ф1 𝑥, 𝑦 = ф1𝐿 𝑥 + 𝐶11 𝑥 𝑦 − 𝑑1 + 𝐶12 𝑥 𝑦 − 𝑑1 (2)
regions are uniformally doped at Nd+ =1026 m-3. The work
function of metal gates M1 and M2 are фM1=5.3 V and Similarly, ф1𝐿 𝑥 is the upper channel potential, 𝑑1 = 𝑑𝑏1 +
фM2=4.1V respectively. 𝑑𝑠 + 𝑑𝑐 + 𝛥𝑑1 distance of upper channel from gate, 𝑑𝑏1 the
In normally on device the Al0.3 Ga0.7 N barrier layer are fully thickness of n-Al0.3 Ga0.7 N , 𝛥d1 depth of 2DEG in upper
depleted under normal operating conditions and electons are GaN layer from its top edge.
confined to the 2DEG in upper and lower GaN layers forming
hetrointerfaces for double channel effects. Similarly, ф2 𝑥, 𝑦 in lower channel can be obtained as
𝑑 2 ф2 (𝑥,𝑦) 𝑑 2 ф2 (𝑥,𝑦) −𝑞 𝑁𝑑
+ = 𝑓𝑜𝑟 0 ≤ 𝑥 ≤ 𝐿 (3)
𝑑𝑥 2 𝑑𝑦2 𝜀 𝑎 (𝑚 )
SOURCE GATE DRAIN

L1 L2
Where ф2 (𝑥, 𝑦) is and lower channel potential, Similar to as
stated above the potential profile in vertical direction for lower
channel can be apprximated by following parabolic function
M1 M2 2
ф2 𝑥, 𝑦 = ф2𝐿 𝑥 + 𝐶21 𝑥 𝑦 − 𝑑2 + 𝐶22 𝑥 𝑦 − (4)
dc = 3 nm (undoped Al0.3Ga0.7N) ф2𝐿 𝑥 is the lower channel potential, 𝑑2 = 𝑑1 + 𝑑𝑐𝑕1 +
𝑑𝑏2 + 𝛥𝑑2 is distance of lower channel from gate, 𝑑𝑐𝑕1 is
Nd+ db1 = 18 nm (Si-doped Al0.3Ga0.7N) Nd+ thickness of upper undoped-GaN layer , 𝑑𝑏2 the thickness of
Nd lower n-AlmGa1-m N and 𝛥𝑑2 is depth of 2DEG in lower GaN
ds = 3 nm (undoped Al0.3Ga0.7N) buffer layer.
In expressions (2) and (4), 𝐶11 𝑥 , 𝐶12 𝑥 , and 𝐶21 𝑥 , 𝐶22 𝑥
Upper 2-DEG
are arbitary coefficients.
Since in DCDMG Al0.3 Ga0.7 N /GaN HEMT gate is formed
dch1 = 14 nm (undoped GaN) by merging two different metal gates M1 and M2 of
workfunctions фM1 and фM2 respetively. Thus flatband
db2 = 21 nm AlmGa1-mN
voltages 𝑉𝐹𝐵1 and 𝑉𝐹𝐵2 of gates are given by expressions
(m graded from 3% to 6%)
𝑉𝐹𝐵1 = ф𝑀𝑆1 = ф𝑀1 − ф𝑆 (5)
Lower 2-DEG 𝑉𝐹𝐵2 = ф𝑀𝑆2 = ф𝑀2 − ф𝑆 (6)
-
dch2 = 2.5 µm (GaN) Where, ф𝑀𝑆1 and ф𝑀𝑆2 are metal to semiconductor work
functins for gate region M1 and gate regin M2 respectively.
SAPPHIRE Also,
𝐸𝑔
ф𝑆 = 𝜒𝑎 (𝑚) + − ф𝐹 (7)
2
𝑊𝑕𝑒𝑟𝑒, ф𝑆 is work function of n- Al0.3 Ga0.7 N, 𝜒𝑎 (𝑚) is
Fig.1. Cross sectional schematic of DCDMG AlGaN/GaN HEMT electron affinity , 𝐸𝑔 is band gap of n- Al0.3 Ga0.7 N at
based on double channel AlGaN/GaN HEMT epilayer on sapphire T=300K and ф𝐹 is fermi potential given by expression
substrate [1], Total gate length (L)=L1+L2, L1=L2=0.5µm, work
function of gate region M1 (фM1) =5.3V, work function of gate 𝑘𝑇 𝑁𝑑
ф𝐹 = 𝑙𝑛 (8)
region M2 (фM2) =4.1V [6][7], Gate width (W) =100𝝻m, 𝑞 𝑛𝑖
Nd=2x1024m-3, d1=db1+ds+dc+𝛥d1 and d2=d1+dch1+db2+Δd2 are
distance of upper channel and lower channel from gate , 𝛥d1 and Δd2 where 𝑛𝑖 is intrinsic career concentration and 𝑁𝑑 is donar
denotes 2DEG effective depth in upper and lowere GaN layers concentration in the material.
respectively. In the DCDMG Al0.3 Ga0.7 N HEMT structure gate has been
partitioned in to two parts, hence potential under gate region
In our analytical model the channel region it is devided into M1 and gate region M2 for upper channel will be given
two parts each under metal gate M1 and metal gate M2. The 2- separately as
D potential distribution ф1 (𝑥, 𝑦 and ф2 𝑥, 𝑦 in upper channel ф11 𝑥, 𝑦 = ф1𝐿1 (𝑥) + 𝐶111 𝑥 𝑦 − 𝑑1
can be obtained by solving the 2-D Poission,s equation as + 𝐶112 𝑥 𝑦 − 𝑑1 2 0 ≤ 𝑥 ≤ 𝐿1
(9)
𝑑 2 ф1 (𝑥,𝑦 ) 𝑑 2 ф1 (𝑥,𝑦) −𝑞𝑁𝑑
+ = 𝑓𝑜𝑟 0 ≤ 𝑥 ≤ 𝐿 (1) ф12 𝑥, 𝑦 = ф1𝐿2 𝑥 + 𝐶121 𝑥 (𝑦 − 𝑑1 )
𝑑𝑥 2 𝑑𝑦2 𝜀(𝑚 )
+ 𝐶122 𝑥 𝑦 − 𝑑1 2 𝐿1 ≤ 𝑥
where x and y denotes horizontal and vertical axis in the
≤ ( 𝐿1 + 𝐿2 )
device cross section respectively, Nd is doping density of the
(10)
Al0.3 Ga0.7 N layer, 𝜀(𝑚) is the dielectric constant, q is the
electronic charge of Al0.3 Ga0.7 N and L=L1+L2 is the effective Similarily potential under gate regions M1 and M2 for lower
channel length. Potential profile ф1 𝑥, 𝑦 in vertical y channel will be given separately as

www.rsisinternational.org/IJRSI.html Page 50
Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

ф21 𝑥, 𝑦 = ф2𝐿1 𝑥 + 𝐶211 𝑥 𝑦 − 𝑑2 sheet career densities for lower channel under M1 and M2
+ 𝐶212 𝑥 𝑦 − 𝑑2 2 𝑓𝑜𝑟 0 respectively [12]. These values can be obtained by [13] as
≤ 𝑥 ≤ 𝐿1 𝜀(𝑚 )
(11) 𝑛𝑠11 = 𝑉𝑔𝑠 − 𝑉𝑡𝑕1 − 𝑉𝑐1 𝑥 𝑓𝑜𝑟 0 ≤ 𝑥 ≤ 𝐿1 (29)
𝑞𝑑 1

ф22 𝑥, 𝑦 = ф2𝐿2 𝑥 + 𝐶221 𝑥 𝑦 − 𝑑2 𝑛𝑠12 =


𝜀(𝑚 )
𝑉𝑔𝑠 − 𝑉𝑡𝑕1 − 𝑉𝑐1 𝑥 𝑓𝑜𝑟 𝐿1 ≤ 𝑥 ≤ (𝐿1 + 𝐿2 )
+ 𝐶222 𝑥 𝑦 − 𝑑2 2 𝑓𝑜𝑟 𝐿1 ≤ 𝑥 𝑞𝑑 1

≤ ( 𝐿1 + 𝐿2 ) (30)
(12) 𝜀(𝑚 )
𝑛𝑠21 = 𝑉𝑔𝑠 − 𝑉𝑡𝑕2 − 𝑉𝑐2 𝑥 𝑓𝑜𝑟 0 ≤ 𝑥 ≤ 𝐿1
𝑞 𝑑2
The Poisson equations are solved separately for upper channel
and lower channel under each gate regions using the following (31)
boundary conditions 𝜀(𝑚 )
1. Channel potential at the interface of two dissimilar metals is 𝑛𝑠22 = 𝑉𝑔𝑠 − 𝑉𝑡𝑕2 − 𝑉𝑐2 𝑥 𝑓𝑜𝑟 𝐿1 ≤ 𝑥 ≤
𝑞𝑑 2
continuous [9-11] ( 𝐿1 + 𝐿2 ) (32)
ф11 𝑥, 𝑑1 = ф12 𝑥, 𝑑1 𝑎𝑡 𝑥 = 𝐿1 (13)
Where, 𝑉𝑐1 (𝑥) and 𝑉𝑐2 𝑥 are upper and lower channel
ф21 𝑥, 𝑑2 = ф22 𝑥, 𝑑2 𝑎𝑡 𝑥 = 𝐿1 (14) potentials at any point x along the channel due to 𝑉𝑑𝑠 as
applied drain to source voltage, 𝑉𝑔𝑠 is applied gate to source
2. Similarily electric field at the interface of two dissimilar
metal is also continuous [9] , thus voltage. The thresold voltages 𝑉𝑡𝑕1 of upper channel and 𝑉𝑡𝑕2
lower channels respectively are given as
ф11 𝑥,𝑦 ф12 𝑥,𝑦 𝑞(𝑁𝑑 −𝑁𝑡 ) 𝑑 𝑏2 1 𝜎𝑝𝑧 (𝑚 )𝑑 1
= 𝑎𝑡 𝑥 = 𝐿1 𝑓𝑜𝑟 𝑢𝑝𝑝𝑒𝑟 𝑐𝑕𝑎𝑛𝑛𝑒𝑙(15) 𝑉𝑡𝑕1 (𝑚) = 𝜒(𝑚) – 𝛥𝜑(𝑚) − − (33)
𝑑𝑥 𝑑𝑥 2 𝜀(𝑚 ) 𝜀(𝑚 )
ф21 𝑥,𝑦 ф22 𝑥,𝑦 𝑞(𝑁𝑑 −𝑁𝑡 ) 𝑑 𝑏2 2 𝜎𝑝𝑧 (𝑚 )𝑑 2
= 𝑎𝑡 𝑥 = 𝐿1 𝑓𝑜𝑟 𝑙𝑜𝑤𝑒𝑟 𝑐𝑕𝑎𝑛𝑛𝑒𝑙 (16) 𝑉𝑡𝑕2 (𝑚) = 𝜒(𝑚) – 𝛥𝜑(𝑚) − − (34)
𝑑𝑥 𝑑𝑥 2 𝜀(𝑚 ) 𝜀(𝑚 )

3. Taking gate to source voltage in consideration, potential Where, 𝝌(m) as the Schottky barrier height, 𝛥𝜑(𝑚) as the
ф11 𝑥, 0 = 𝑉𝑔1 (17) conduction band discontinuity at the Al0.3 Ga0.7 N /GaN
hetrointerface. 𝑁𝑑 is Si doping concentration of n-AlGaN
ф12 𝑥, 0 = 𝑉𝑔2 (18) layer and 𝑁𝑡 is trap concentration [9], 𝜎𝑝𝑧 𝑚 net polarization
induced sheet carrier density at the hetrointerfaces and obtained as
ф21 𝑥, 0 = 𝑉𝑔1 (19) 𝜎𝑝𝑧 𝑚 = [ 𝑃𝑆𝑃 𝑚 − 𝑃𝑆𝑃 0 + 𝑃𝑝𝑧 ] (35)
ф22 𝑥, 0 = 𝑉𝑔2 (20) Where PSP (m) and PSP(0) is spontaneous polarization of AlmGa1-mN
and GaN material system [13-15].
where 𝑉𝑔1 = 𝑉𝑔𝑠 − 𝑉𝐹𝐵1 , 𝑉𝑔2 = 𝑉𝑔𝑠 − 𝑉𝐹𝐵2 , and 𝑉𝑔𝑠 is In our model piezoelectric polarization depends on the amount
of strain developed at hetrointerfaces in order to accommodate
gate to source voltage.
the difference in lattice constants of GaN and AlGaN. For
𝑑ф11 𝑥,𝑦 fully satrained device piezoelectric polarization dependent
4. = −𝐸𝑖𝑛𝑡 11 (21)
𝑑𝑦 𝑦=𝑑 1 charge density is given as
𝑎 0 −𝑎(𝑚 𝐶 (𝑚 )
𝑑ф12 𝑥,𝑦 𝑃𝑝𝑧 = 2 𝑒31 𝑚 − 𝑒33 𝑚 13 for 0≤m≤1
= −𝐸𝑖𝑛𝑡 12 (22) 𝑎 (𝑚 ) 𝐶 (𝑚 ) 33
𝑑𝑦 𝑦=𝑑 1 𝑎 0 − 𝑎(𝑚
𝜎𝑝𝑧 𝑚 = 2 𝑒31 𝑚
𝑑ф21 𝑥,𝑦 𝑎(𝑚)
= −𝐸𝑖𝑛𝑡 21 (23) 𝐶13 (𝑚)
𝑑𝑦 𝑦=𝑑 2 − 𝑒33 𝑚 + + 𝑃𝑆𝑃 𝑚 − 𝑃𝑆𝑃 0
𝐶33 (𝑚)
𝑑ф22 𝑥,𝑦
= −𝐸𝑖𝑛𝑡 22 (24) for 0≤m≤1 36)
𝑑𝑦 𝑦=𝑑 2
Where a(m) is lattice constant e31(m) and e32(m) are
where 𝐸𝑖𝑛𝑡 11 𝑎𝑛𝑑 𝐸𝑖𝑛𝑡 12 are fields at hetrointerface of upper piezoelecric constants and C13(m) and C33(m)are elastic
channel under M1 and M2. Similarily 𝐸𝑖𝑛𝑡 21 and 𝐸𝑖𝑛𝑡 22 are constants[14-16].
fields at hetrointerface of lower channel under M 1 and M2 [9- Increasing Al mole fraction in barrier layer increases
11]. Thus lattice mismatch that further results in stain relaxation.
𝑞𝑛 𝑛𝑠 11 For partially relaxed device with AlmGa1-mN layer thickness
𝐸𝑖𝑛𝑡 11 = (25) in the range of 18nm to 40 nm piezoelectric polarization
𝜀(𝑚 )
𝑞𝑛 𝑛𝑠 12 induced charge density depends on the Al mole fraction as
𝐸𝑖𝑛𝑡 12 = (26) [16].
𝜀(𝑚 )
𝑎 0 −𝑎(𝑚 𝐶13 𝑚
𝑞𝑛 𝑛𝑠 21 2 𝑒31 𝑚 − 𝑒33 𝑚
𝐸𝑖𝑛𝑡 21 = (27) 𝑎 𝑚 𝐶33 𝑚
𝜀(𝑚 )
𝑓𝑜𝑟 0 ≤ 𝑚 ≤ 0.38
𝑞𝑛 𝑛𝑠 22
𝐸𝑖𝑛𝑡 22 = (28) 𝑃𝑃𝑧 = 2 2.33−3.5 𝑚 𝐶13 𝑚 (37)
𝜀(𝑚 ) 𝑎 0 −𝑎 (𝑚 𝑒31 𝑚 − 𝑒33 𝑚
𝐶33 𝑚
𝑎 𝑚
here 𝑛𝑠11 and 𝑛𝑠12 are sheet career densities for upper 𝑓𝑜𝑟 0.38 ≤ 𝑚 ≤ 0.67
channel under M1 and M2 respectively, also 𝑛𝑠21 and 𝑛𝑠22 are 0 𝑓𝑜𝑟 0.67 ≤ 𝑚 ≤ 1

www.rsisinternational.org/IJRSI.html Page 51
Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

𝑑12
In above equation (37) device remains completely strained for where 𝜆1 =
2
Al molefraction up to 38% and partially relaxed for m from
38% to 67%. Device is fully relaxed for m greater than 67%. characteristic length of upper channel
The potential at source end for each channel is 𝑑2 2
and 𝜆2 =
ф11 0, 𝑑1 = 𝑉𝑏𝑖1 = ф1𝐿1 0 (38) 2

ф21 0, 𝑑2 = 𝑉𝑏𝑖2 = ф2𝐿1 0 (39) characteristic length of lower channel

Where𝑉𝑏𝑖1 and 𝑉𝑏𝑖2 is built in voltages. Introducing a new variable


−𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 11
5. 𝑇𝑕𝑒 𝑝𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 𝑎𝑡 𝑑𝑟𝑎𝑖𝑛 𝑒𝑛𝑑 𝑓𝑜𝑟 𝑒𝑎𝑐𝑕 𝑐𝑕𝑎𝑛𝑛𝑒𝑙 𝑖𝑠 𝜂11 (𝑥) = ф1𝐿1 𝑥 − 𝑉𝑔1 − + 𝜆1 2 (50)
𝜀(𝑚 ) 𝑑1
Ф12 𝐿1 + 𝐿2 , 𝑑2 = 𝑉𝑏𝑖1 + 𝑉𝑑𝑠 = Ф1𝐿2 𝐿1 + 𝐿2 (40) −𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 11
𝜂12 (𝑥) = ф1𝐿2 𝑥 − 𝑉𝑔2 − + 𝜆1 2 (51)
𝜀(𝑚 ) 𝑑1
Ф22 𝐿1 + 𝐿2 , 𝑑2 = 𝑉𝑏𝑖2 + 𝑉𝑑𝑠 = Ф2𝐿2 𝐿1 + 𝐿2 (41)
−𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 22
the constants in (9), (10), (11) and (12) can be found from 𝜂21 (𝑥) = ф2𝐿1 𝑥 − 𝑉𝑔1 − + 𝜆2 2 (52)
𝜀(𝑚 ) 𝑑2
boundry conditions and on substituting their values in (9), (10), −𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 22
(11) and (12), We get 𝜂22 (𝑥) = ф2𝐿2 𝑥 − 𝑉𝑔2 − + 𝜆2 2 (53)
𝜀(𝑚 ) 𝑑2
ф11 𝑥, 𝑦 = ф1𝐿1 𝑥 − 𝐸𝑖𝑛𝑡 11 𝑥 − (𝑦 − 𝑑1 )
𝑉𝑔1 − ф1𝐿1 𝑥 𝐸𝑖𝑛𝑡 11 on substituting (50) in (46) , (51) in (47) , (52) in (48) and (53)
+ 2 − 𝑦 − 𝑑1 2 in (49), we get
𝑑1 𝑑1
(42) 𝑑 2 𝜂 11 (𝑥) 𝜂 12 (𝑥)
− =0 (54)
𝑑𝑥2 𝜆12
ф12 𝑥, 𝑦 = ф1𝐿2 𝑥 − 𝐸𝑖𝑛𝑡 12 𝑥 − (𝑦 − 𝑑1 ) +
𝑉𝑔2−ф1𝐿2 𝑥 𝐸𝑖𝑛𝑡 12 2 𝑑 2 𝜂 21 (𝑥) 𝜂 22 (𝑥)
𝑑12
− 𝑦 − 𝑑1 (43) − =0 (55)
𝑑1 𝑑𝑥 2 𝜆22

ф21 𝑥, 𝑦 = ф2𝐿1 𝑥 − 𝐸𝑖𝑛𝑡 21 𝑥 − (𝑦 − 𝑑2 ) + using (50) and (51) , η11 (x=0) = η111, η12 (x=L1+L2) = η122 and
𝑉𝑔1−ф2𝐿2 𝑥 𝐸 boundry conditions, similarly using (52) and (53) η21 (x=0) =
2 − 𝑖𝑛𝑡 21 𝑦 − 𝑑2 2 (44)
𝑑2 𝑑2 η211, η22 (x=L1+L2) = η222 and boundry conditions we get
ф22 𝑥, 𝑦 = ф2𝐿2 𝑥 − 𝐸𝑖𝑛𝑡 21 𝑥 − (𝑦 − 𝑑2 ) 𝑞𝑁 𝑑 2𝐸𝑖𝑛𝑡 11
𝑉𝑔2 − ф2𝐿1 𝑥 𝐸𝑖𝑛𝑡 22 𝜂111 = 𝑉𝑏𝑖 − 𝑉𝑔1 − − 𝜆1 2 (56)
2 𝜀(𝑚 ) 𝑑1
+ 2 − 𝑦 − 𝑑2
𝑑2 𝑑2
(45) 𝜂122 = 𝜂111 + 𝑉𝑑𝑠 – 𝑉𝐹𝐵1 – 𝑉𝐹𝐵2 − 𝑏12 − 𝑏11 𝜆1 2 (57)
𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 21
Channels potentials ф1L1 (x) and ф2L2(x) are obtained by 𝜂211 = 𝑉𝑏𝑖 − 𝑉𝑔1 − − 𝜆2 2 (58)
𝜀(𝑚 ) 𝑑2
substituting (42), (43) in (1) and (44), (45) in (2)..On
substitution we get 𝜂222 = 𝜂211 + 𝑉𝑑𝑠 – 𝑉𝐹𝐵1 – 𝑉𝐹𝐵2 − 𝑏22 − 𝑏21 𝜆2 2 (59)
𝑑 2 ф1𝐿1 𝑉𝑔1 − ф1𝐿1 (𝑥) −𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 11 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 11 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 12
+ = + (46) where 𝑏11 = − and 𝑏12 = − for upper
𝑑𝑥2 𝜆12 𝜀(𝑚 ) 𝑑1 𝜀(𝑚 ) 𝑑1 𝜀(𝑚 ) 𝑑1
𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 21 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 22
𝑑 2 ф1𝐿2 𝑉𝑔2 − ф1𝐿2 (𝑥) −𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 12 channel and 𝑏21 = − and 𝑏22 = − for
+ = + (47) 𝜀(𝑚 ) 𝑑2 𝜀(𝑚 ) 𝑑2
2
𝑑𝑥2 𝜆1 𝜀(𝑚 ) 𝑑1 lower channel respectively.
𝑑 2 ф2𝐿1 𝑉𝑔1 − ф2𝐿1 (𝑥) −𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 21
+ = + (48) Solving (54) for 𝜂11 (𝑥) and 𝜂12 (𝑥) and substituting in (46)
𝑑𝑥2 𝜆22 𝜀(𝑚 ) 𝑑2
and (47) we get channel potential for upper channel as:
𝑑 2 ф2𝐿2 𝑉𝑔2 − ф2𝐿2 (𝑥) −𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 22
+ = + (49)
𝑑𝑥2 𝜆22 𝜀(𝑚 ) 𝑑2

------------------------------------------------------------------------------------------- ----------------------------------------------------------------
𝐿 1 −𝑥 𝑥
𝜂 111 𝑠𝑖𝑛 𝑕 +𝜂 112 𝑠𝑖𝑛 𝑕 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 11
𝜆1 2
𝜆1 𝜆1
Ф1𝐿 1 𝑥 = 𝐿 +𝑉 +
𝑔1 − (60)
𝑠𝑖𝑛 𝑕 1 𝜀(𝑚 ) 𝑑1
𝜆1
𝑥 −𝐿 1 𝐿 1 +𝐿 2 −𝑥
𝜂 122 𝑠𝑖𝑛 𝑕 +𝜂 121 𝑠𝑖𝑛 𝑕 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 12
𝜆1 2
𝜆1 𝜆1
Ф1𝐿 2 𝑥 = 𝐿 +𝑉𝑔2 + − (61)
𝑠𝑖𝑛 𝑕 2 𝜀(𝑚 ) 𝑑1
𝜆1

Similarily solving (55) for 𝜂21 (𝑥) and 𝜂22 (𝑥) and substituting in (49) and (50) we get potential for lower channel as:
𝐿 1 −𝑥 𝑥
𝜂 211 𝑠𝑖𝑛 𝑕 +𝜂 212 𝑠𝑖𝑛 𝑕 𝜆 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 21
𝜆2 2
𝜆2 2
Ф2𝐿1 𝑥 = 𝐿 +𝑉𝑔1 + − (62)
𝑠𝑖𝑛 𝑕 1 𝜀(𝑚 ) 𝑑2
𝜆2

𝑥 −𝐿 1 𝐿 +𝐿 −𝑥
𝜂 222 𝑠𝑖𝑛 𝑕 +𝜂 221 𝑠𝑖𝑛 𝑕 1 𝜆 2 𝑞𝑁𝑑 2𝐸𝑖𝑛𝑡 22
𝜆2 2
𝜆1 2
Ф2𝐿 2 𝑥 = 𝐿 +𝑉𝑔2 + − (63)
𝑠𝑖𝑛 𝑕 2 𝜀(𝑚 ) 𝑑2
𝜆2

www.rsisinternational.org/IJRSI.html Page 52
Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

Using boundry conditions (38), (40) and (56), (57) in (60) and (61) we get values of η112 and η121 as
𝐿2 𝐿1 𝐿1 𝐿2 2 𝐿1 𝐿2
𝜂 111 𝑠𝑖𝑛 𝑕 +𝜂 122 𝑠𝑖𝑛 𝑕 + 𝑉𝑔2 −𝑉𝑔1 𝑠𝑖𝑛 𝑕 𝑐𝑜𝑠 𝑕 +(𝑏12 −𝑏11 )𝜆 1 𝑠𝑖𝑛 𝑕 𝑐𝑜𝑠 𝑕
𝜆1 𝜆1 𝜆1 𝜆1 𝜆1 𝜆1
𝜂112 = 𝐿1 𝐿2 𝐿2 𝐿1 (64)
𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕 +𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕
𝜆1 𝜆1 𝜆1 𝜆1

𝐿1 𝐿2 𝐿1 𝐿2 2 𝐿2 𝐿1
𝜂 122 𝑠𝑖𝑛 𝑕 +𝜂 111 𝑠𝑖𝑛 𝑕 + 𝑉𝑔1 −𝑉𝑔2 𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕 + 𝑏11 −𝑏12 𝜆 1 𝑠𝑖𝑛 𝑕 𝑐𝑜𝑠 𝑕
𝜆1 𝜆1 𝜆1 𝜆1 𝜆1 𝜆1
𝜂121 = 𝐿1 𝐿2 𝐿2 𝐿1 (65)
𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕 +𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕
𝜆1 𝜆1 𝜆1 𝜆1

Now using boundry conditions (39), (41) and (58), (59) in (62) and (63), we get values of η212 and η221 as
𝐿2 𝐿1 𝐿1 𝐿2 𝐿1 𝐿2
𝜂 211 𝑠𝑖𝑛 𝑕 +𝜂 222 𝑠𝑖𝑛 𝑕 + 𝑉𝑔2 −𝑉𝑔1 𝑠𝑖𝑛 𝑕 𝑐𝑜𝑠 𝑕 +(𝑏22 −𝑏21 )𝜆 2 2 𝑠𝑖𝑛 𝑕 𝑐𝑜𝑠 𝑕
𝜆2 𝜆2 𝜆2 𝜆2 𝜆2 𝜆2
𝜂212 = 𝐿1 𝐿2 𝐿2 𝐿1
(66)
𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕 +𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕
𝜆 𝜆2 𝜆2 𝜆2

𝐿1 𝐿2 𝐿1 𝐿2 𝐿2 𝐿1
𝜂 222 𝑠𝑖𝑛 𝑕 +𝜂 211 𝑠𝑖𝑛 𝑕 + 𝑉𝑔1 −𝑉𝑔2 𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕 +(𝑏21 −𝑏22 )𝜆 2 2 𝑠𝑖𝑛 𝑕 𝑐𝑜𝑠 𝑕
𝜆2 𝜆2 𝜆2 𝜆2 𝜆2 𝜆2
𝜂221 = 𝐿1 𝐿2 𝐿2 𝐿1 (67)
𝑐𝑜𝑠 𝑕 𝑠𝑖𝑛 𝑕 +𝑐𝑜𝑠𝑕 𝑠𝑖𝑛 𝑕
𝜆2 𝜆2 𝜆2 𝜆2

----------------------------------------------------------------------------------------------------------------------------- ------------------------------
2.2 Electric Field Analysis: channels at the interface of M1 and M2 orinates due to the
The electron transport velocity through the channel is directly work function differences of two gate metal regions of
related with the electric field along the channel. The surface different work funtions. This further enhances drift velocity of
electric field component along the upper channel in x direction 2-DEG carriers, reduces DIBL and result into screening effect
under the Gate M1 is given as by M2 suppressing short channel effects (SCE) in upper and
𝑑ф11 𝑥 ,𝑦 𝑑ф1𝐿1 (𝑥,𝑦) lower channel effectvely.
𝐸11 𝑥 = = (68)
𝑑𝑥 𝑦=𝑑 1 𝑑𝑥
−1 𝐿 −𝑥 1 𝑥
𝜂 𝑐𝑜𝑠 𝑕 1 + 𝜂 𝑐𝑜𝑠 𝑕
𝜆 1 111 𝜆1 𝜆 1 112 𝜆1 3
𝐸11 𝑥 = 𝐿 (69)
𝑠𝑖𝑛 𝑕 1
𝜆1
and x component under M2 is
𝑑ф 𝑥 ,𝑦 𝑑ф (𝑥,𝑦) 2.5
𝐸12 𝑥 = 12 = 1𝐿2 (70) Upper channel
𝑑𝑥 𝑦=𝑑 1 𝑑𝑥
−1 𝐿 𝐿 −𝑥 1 𝑥 −𝐿 1
𝑐𝑜𝑠 𝑕 1+ 2
Channel Potential (V)

𝜂 + 𝜂 𝑐𝑜𝑠 𝑕
𝜆 1 121 𝜆1 𝜆 1 122 𝜆1
𝐸12 𝑥 = 𝐿 (71)
𝑠𝑖𝑛 𝑕 2 2
𝜆1
The surface electric field component along the lower channel
in x direction under the Gate region M1 is given as
𝑑ф21 𝑥,𝑦 𝑑ф2𝐿1 (𝑥,𝑦 ) 1.5
𝐸21 𝑥 = = (72) Lower channel
𝑑𝑥 𝑦=𝑑 2 𝑑𝑥
−1 𝐿 −𝑥 1 𝑥
𝜂 𝑐𝑜𝑠 𝑕 1𝜆 + 𝜆 𝜂 212 𝑐𝑜𝑠 𝑕 𝜆
𝜆 2 211 2 2 2
𝐸21 𝑥 = 𝐿 (73) 1
𝑠𝑖𝑛 𝑕 1
𝜆2
and x component under gate region M2 is given as
𝑑ф 𝑥,𝑦 𝑑ф (𝑥,𝑦 )
𝐸22 𝑥 = 22 = 2𝐿2 (74) 0.5
𝑑𝑥 𝑦=𝑑 2 𝑑𝑥
−1 𝐿 𝐿 −𝑥 1 𝑥 −𝐿 1
𝜂 221 𝑐𝑜𝑠 𝑕 1+ 2 + 𝜂 𝑐𝑜𝑠 𝑕 0 0.2 0.4 0.6 0.8 1
𝜆2 𝜆2 𝜆 2 222 𝜆2
𝐸22 𝑥 = 𝐿2 (75)
𝑠𝑖𝑛 𝑕
𝜆2
Normalised Channel Length (x/L)
Fig. 2 Variation of channel potential with normalized channel length in
DCDMG Al0.3 Ga0.7 N/GaN HEMT solid lines shows model data and
III. RESULTS AND DISCUSSIONS dash lines shows ATLAS simulation data.

To verify the present DCDMG Al0.3 Ga0.7 N/GaN -HEMT


model, ATLAS TCAD [8] device simulator is used to analyse
The potential distribution plot shows that gate M1 functions as
channel electric field, potential distribution, current voltage
control gate and gate M2 works as screen gate in device. Since
characteristics as well as transconductance of device. Also
potential begins to drop more in the channels under screen
model and simulation results are compared with experimental
gate M2 under drain junction reions leading to more reduced
data [1][6][7]. Fig.2 shows the potentials distribution along
DIBL. Plot also depicts that potential distribution in upper
both the channels as extracted from the analytical model as
channel is higher than that of the lower channel.
well as simulated data. These results are in good agreement
Fig.3 depicts longitudinal electric field variation along the
with pulished DMG AlGaN/GaN HEMT results [9]. In this
channels for DCDMG Al0.3 Ga0.7 N/GaN HEMT model and
figure, the solid lines shows model results and dash lines
simulation data for channel lenth L=1µm. Due to the peaks in
represent the simulated results which further depends on
electric field profile at the interface of two gate metals in
device biasing. The upper channel potential starts from 1.25
DCDMG AlGaN/GaN HEMT structure, the electric field
volts and increases with hump up to 2.0 volts. This figure
under M2 in both the channels get enhanced. Since the
clearly shows the step in the potential distribution along the
electron transport velocity is related to the electric field
www.rsisinternational.org/IJRSI.html Page 53
Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

distribution along the channels. The velocity with which collapse due to presence of two channels in the device
charge carriers enter and drift in the channels also increases structure. Output current voltage characteristics also shows
significantly higher darin current to that of the conventional
4 10 GaN based HEMTs. The significant enhancement in
8 transconductance of DCDMG Al0.3 Ga0.7 N/GaN HEMT has
2 also been verified from Fig. 5. The simulated input
6 characteristics plot for the device reveals a much higher drive

Lower channel Electric Field x105 (V/cm)


Upper Channel Electric Field x105 (V/cm)

0 current in comparison of conventional GaN based HEMTs


4 [10] due to the increased carier transport velocity in the device
0 0.2 0.4 0.6 0.8 1
-2 and better gate controllability.Thus our proposed DCDMG
2
Upper channel Al0.3 Ga0.7 N/GaN HEMT structure can improve current drive
0 capabilty and overall device transconductance by reducing
-4
current collapse, SCEs and hot electron effects much
-2 effectively than conventional GaN HEMT structures [11-14].
-6 A close matching between model results, simulation results
Lower channel -4 and published experimental data as described above prove the
-8 accuracy and validity of our proposed device model.
-6
-10 -8 IV. CONCLUSIONS
Normalized Channel Position (x/L)
Analytical modeling concept of dual channel dual material
Fig.3 Longitudinal Electric Field variation with gate Al0.3 Ga0.7 N/GaN HEMT has been examined first time
normalized channel position for DCDMG AlGaN/GaN by developing solid state physics baed 2-D analytical model
HEMT . Solid lines shows model data and dash lines and simulating device structure using industry validated
shows ATLAS data. ATLAS TCAD [8]. The analytical expressions for upper and
lower channels potential distribution and electric field
distribution have been obtained and found in good agreement.
leading to more efficient carrier transport along both the The results are further caliberated with published experimental
channels. Also presence of lower workfunction of gate metal data.
M2 reduces peak electric field at the drain end leading to
reduction in hot carrier effects in the device.

It can be concluded with the above interpretations that


DCDMG Al0.3Ga0.7N/GaN HEMT architecture provides
Fig.4 shows current voltage characteristics of DCDMG Al0.3 screening effects in small channel devices thus suppressing
Ga0.7 N/GaN HEMT for different gate to source biases when short channel effects (SCE), reducing hot electron effects due
Fig.1 device structure simulated using ATLAS TCAD device to introduction of dual material gate and minimising overall
simulaor. Plot clearly shows variation of drain current with current collapse due to introduction of double channels. In
variation on drain to source voltage at various gate to source addition, due to enhanced channel potential distribution and
biases in both linear and saturation regions. Simulated results step function in channel electric field distribution, the more
shows that DCDMG AlGaN/GaN HEMT device has effective uniform carrier drift velocity in both the channels is obtained
control on drain current with reduction in overall current that results in higher switching speed of the device.

www.rsisinternational.org/IJRSI.html Page 54
Volume II, Issue V, May 2015 IJRSI ISSN 2321 - 2705

REFERENCES
[1] Roanming Chu,Yugang Zhou, Zie Liu, Deliang
Wang, Kevin J. Chen, “AlGaN/GaN Double Channel
HEMT,” IEEE Transaction on Electron Devices,
vol.52, no.4, Apr. 2005.
[2] O. Ambacher, J. Smart, J. R. Shealy, N. G. Weimann, K. Chu,
M.Murphy, W. J. Schaff, L. F. Eastman, R. Dimitrov, L. Wittmer,
M. Stutzmann, W. Rieger, and J. Hilsenbeck, “Two-dimensional
electron gases induced by spontaneous and piezoelectric
polarization charges in N- and Ga-face AlGaN/GaN
heterostructures”, J. Appl. Phys, vol. 85, no. 6, pp. 3222–3233, Mar.
1999.
[3] Y. F. Wu, A. Saxler, M. Moore, R. P. Smith, S. Sheppard, P.
M.Chavarkar, T. Wisleder, U. K. Mishra, and P. Parikh, “30-W/mm
GaN HEMTs by field plate optimization”, IEEE Electron Device
Lett., vol.25, no. 3, pp. 117–119, Mar. 2004.
[4] V. Tilak, B. Green, V. Kaper, H. Kim, T. Prunty, J. Smart, J. Shealy,
and L. Eastman, “Influence of barrier thickness on the high power
performance of AlGaN/GaN HEMTs”, IEEE Electron Device Lett.,
vol. 22, no. 11, pp. 504–506, Nov. 2001.
[5] K.Kasahara, N. Miyamoto, Y. Ando, Y. Okamoto, T. Nakayama, and
M.Kuzuhara, “Ka-band 2.3 W power AlGaN/GaN hetrojunction FET,”
in IEDM Tech. Dig., pp. 667–680, Dec. 2002.
[6] W. Long, H. Ou, J. M. Kuo and K.K Chin., "Dual-Material G a t e
(DMG) Field Effect T r a n s i s t o r ” , IEEE Trans. Electron
Devices, vol. 46, pp. 865-70, May 1999.
[7] W. Long et.al. “Methods for fabricating a dual material gate of a
short channel field effect transistor”, US Patent, US006153534,
Nov 2000.
[8] Silvaco TCAD, ATLAS Device Simulator, www.silvaco.com.
[9] Sona P.Kumar et all, “Analytical modeling and simulation of
subthreshold behavior in nanoscale dual material gate AlGaN/GaN
HEMT” Elsevier, Superlattices and Microstructures, vol. 44, pp.
37-53, Mar. 2008.
[10] Jia-Chuhan Lin, Yu-Chieh Chen, Wei-Chih Tsai ,Po-Yu Yang
“Structure design criteria of dual channel high mobility electron
transistor”, Solid-State Electronics, vol.51, pp. 64-68, 2007.
[11] Sona P.Kumar, A. Agarwal, R Chaujar, S. Kabra , M. Gupta, R.S.
Gupta, “Threshold Voltage model for small geometry AlGaN/GaN
HEMT based on analytical solution of 3-D Poisson’s equations”
Microelectronics Journal, vol.38,pp. 1013-1020, Oct.2007.
[12] Sona P.Kumar, A. Agarwal, R Chaujar, M. Gupta, R.S. Gupta,
“Performance Assessment and sub-threshold analysis of gate
material engineered AlGaN/GaN HEMT for enhanced carrier
transport efficiency”, Microelectronics Journal, vol.39,pp.1416-
1424, Aug.2008.
[13] Rashmi “Modelin and characterization of spontaneous and
piezoelectric dependent lattice mismatched AlGaN/GaN HEMT
for microwave and millimeter wave applications” PhD Thesis,
UDSC, India, 2001.
[14] E.S. Hellman, “The polarity of GaN: A critical review”, MRS
Internet J. Nitride Semiconduct. Res. 3, vol. 11, pp. 1-11, 1998.
[15] Y.Zhang,I.P.Smorchkova, C.R.Elsass, S.Kellar, J.P.EIbbetson,
S.Denbaars, U.K.Mishra ans J.Singh “ Charge control and mobility
in AlGaN/ GaN Transistor:Experimental and theoretical studies”
J.Appl, vol.87, no11, pp.7981-7987, 2000.
[16] T.Li,R.P.Joshi and C.Fazi “Montecarlo evaluation of degeneracy
and interface roughness effects on electron transport in AlGaN/
GaN hetrostructures” J.Appl,Phys vol.88, no.2, pp.829-837,2000.

www.rsisinternational.org/IJRSI.html Page 55

You might also like