betaFeedback

WorksPhysics and Astronomy

Analytical Modeling of I−VI-V characteristics using 2D Poisson Equations in AlN/ββ-Ga2_2O3_3 HEMT

Abstract

In this paper, physics-based analytical models using two-dimensional (2D) Poisson equations for surface potential, channel potential, electric field, and drain current in AlN/ββ-Ga2_2O3_3 high electron mobility transistor (HEMT) is presented. The analytical expression of different quantities is achieved based on full depletion approximation of the AlN barrier layer and polarization charge induced unified two-dimensional electron gas (2DEG) charge density model. For the validation of the developed model, results are compared with 2D numerical simulation results, and a good consistency is found between the two. The drain current model is also validated with experimental results of a similar dimension device. The developed model can be a good reference for different ββ-Ga2_2O3_3-based HEMTs.

Introduction

Nowadays, Gallium-oxide (Ga2O3\mathrm{Ga}_2\mathrm{O}_3) is being explored for its possible application in the decade-old high electron mobility transistors (HEMTs) technology, currently dominated by III-nitride material like gallium-nitride (GaN). GaN HEMTs have shown excellent performance for high power and high-frequency applications on the back of exciting features such as higher two-dimensional electron gas (2DEG) density ns∼1013 cm−2n_s \sim10^{13}\ \mathrm{cm}^{-2}, very high 2DEG mobility ∼1500 cm2/Vs\sim1500\ \mathrm{cm}^2/\mathrm{Vs}. However, rising emerging applications like electric vehicles (EVs) and robotics have expanded the investigation horizon from wide bandgap (WBG) to ultra-wide bandgap (UWB) semiconductors. Among the UWB semiconductors, β\beta-Ga2O3\mathrm{Ga}_2\mathrm{O}_3 is found to be more suitable for high voltage applications due to its ultra-large bandgap of ∼4.9 eV\sim4.9\ \mathrm{eV}, and breakdown field of ∼8 MV\sim8\ \mathrm{MV}. Moreover, β\beta-Ga2O3\mathrm{Ga}_2\mathrm{O}_3 single crystal substrate can be grown using melt-based techniques, which ensures large size and affordable wafers for different device technologies.

To date, various β\beta-Ga2O3\mathrm{Ga}_2\mathrm{O}_3 based experimental devices like Schottky diodes [1], [2], metal-oxide-semiconductor field-effect transistors MOSFETs[3]–[6], and β\beta-(Alx_xGa1−x_{1-x})2_2O3_3/Ga2_2O3_3 (AGO/GO) HEMTs [7], [8] with excellent DC and RF performances have been reported. However, since β\beta-Ga2O3\mathrm{Ga}_2\mathrm{O}_3 does not has any polarization property, and small band offset in AGO/GO HEMTs limit their performance as compared to their AlGaN/GaN counterpart. On the other hand, epitaxial layers of III-nitrides (AlN, GaN, InN) have been grown on β\beta-Ga2O3\mathrm{Ga}_2\mathrm{O}_3 and potential AlN/β\beta-Ga2O3\mathrm{Ga}_2\mathrm{O}_3 HEMT applications are anticipated [9]–[12]. Moreover, one of our previous works in AlN/β\beta-Ga2_2O3_3 HEMT estimated exceptional RF performance—cut-off frequency fTf_T of 167 GHz, and output power POUTP_{\mathrm{OUT}} of 2.91 W/mm [13]. Since the 2DEG density nsn_s in HEMTs has a critical role in device operation, recently we have reported validation of estimated nsn_s in AlN/β\beta-Ga2_2O3_3 HEMT through analytical modeling in [14]. Various physics-based models for different parameters in AlGaN/GaN HEMTs have been reported [15, 16, 17, 18, 19, 20], but almost none for β\beta-Ga2_2O3_3-based HEMTs. Looking at the accuracy of the mathematical models over empirical and approximation models, device physics-based mathematical models for β\beta-Ga2_2O3_3 based HEMT will be useful to analyze and validate the estimated results. Here, we present physics-based mathematical models for surface potential and electric fields and extended them to get the current-voltage model in AlN/β\beta-Ga2_2O3_3 HEMT with an etched AlN barrier layer. For the validation of results, developed model results are compared with simulation as well as experimental results earlier published for similar β\beta-Ga2_2O3_3 devices.

Device Structure

The cross-sectional schematic of the analysed device is shown in Figure 1. The epi-layer sequence is as follows: on a semi-insulating β\beta-Ga2_2O3_3 substrate an n-type doped 100 nm β\beta-Ga2_2O3_3 buffer layer with ND=1.5×1016N_D = 1.5 \times10^{16} /cm3^3 exists, followed by 10 nm AlN barrier layer which is undoped. Source and drain contacts are set as Ohmic with contact resistance of 0.4 Ω\Omega-mm. The gate contact is assumed as Schottky and a barrier potential of 0.8 eV is set. The device dimensions are kept the same as [5] with gate length LGL_G of 2.8 μ\mum, gate-drain spacing LGDL_{GD} of 1.3 μ\mum, and gate-source spacing LGSL_{GS} of 0.4 μ\mum. The gate width WW is equal to 100 μ\mum, and the output drain current is normalized with this quantity. In the simulation framework different physics-based models like Shockley-Read-Hall (SRH) recombination, Fermi-Dirac for carrier statistics, and negative differential conductivity (NDC) to capture electron velocity saturation effect are evoked. Default material parameters and spontaneous as well as piezoelectric polarization models are used from [21] for AlN barrier material. The β\beta-Ga2_2O3_3 material parameters used in the simulations and model development are taken from [22], [23], Table 1.

Image not extracted

Figure 1. 2D cross-sectional view of AlN/β\beta-Ga2_2O3_3 HEMT.

SymbolQuantityValuesSymbolQuantityValues
EgE_gEnergy bandgap4.9 eVNVN_VValence band density2.86×10182.86 \times10^{18}
χ\chiElectron affinity3.15 eVεr\varepsilon_rRelative permittivity10
me∗m_e^*Electron effective mass0.25m00.25m_0a0a_0Lattice constant3.04
DDDensity of states1.045×1018 m−2 V−11.045 \times10^{18}\ \mathrm{m}^{-2}\ \mathrm{V}^{-1}μn0\mu_{n0}Electron mobility140 cm2/Vs140\ \mathrm{cm}^{2}/\mathrm{Vs}
NCN_CConduction band density3.13×10183.13 \times10^{18}μp0\mu_{p0}Hole mobility50 cm2/Vs50\ \mathrm{cm}^{2}/\mathrm{Vs}

Table 1. Material parameters for β\beta-Ga2_2O3_3 [22], [23].

A list of symbols used in the model development is given in Table 2.

SymbolQuantitySymbolQuantity
ε1, ε2\varepsilon_1,\ \varepsilon_2Permittivity of AlN and β\beta-Ga2_2O3_3 material respectivelyqqUnit electron charge
d1, d2d_1,\ d_2Thickness of AlN and β\beta-Ga2_2O3_3 layer respectivelyVFBV_{\mathrm{FB}}Flat band voltage
NDN_DChannel doping concentrationVbiV_{\mathrm{bi}}Built-in potential
Φm\Phi_mGate metal workfunctionEFE_FFermi potential
nsn_s2DEG densitykkBoltzmann constant
VOFFV_{\mathrm{OFF}}Threshold voltageΔEC\Delta E_CConduction band offset
VthV_{\mathrm{th}}Thermal voltageχ\chiElectron affinity
CGC_GGate capacitanceTTTemperature

Table 2. Symbols used in analytical modeling

A physics-based analytical model for surface potential, electric field, and drain current are developed by solving two-dimensional (2-D) Poisson’s equation in the channel region marked as Region I, II, and III as shown in Figure 1. The 2-D Poisson’s equation for the potential function φ\varphi (x, y) in the channel (0≤x≥L30 \le x \ge L_3) is given by:

d2φi(x,y)dx2+d2φi(x,y)dy2=−qNDε2,i=1,2,3(1)\frac{d^2\varphi_i(x,y)}{dx^2}+\frac{d^2\varphi_i(x,y)}{dy^2}=-\frac{qN_D}{\varepsilon_2},\qquad i=1,2,3 \tag*{(1)}
d2φi(x,y)dx2+d2φi(x,y)dy2=0,i=4,5,6(2)\frac{d^2\varphi_i(x,y)}{dx^2}+\frac{d^2\varphi_i(x,y)}{dy^2}=0,\qquad i=4,5,6 \tag*{(2)}

where NDN_D, and ε2\varepsilon_2 are doping concentration and permittivity of β\beta-Ga2_2O3_3 buffer layer. Since potential at the AlN/β\beta-Ga2_2O3_3 interface is continuous, it is only solved for the region I, II, and III.

The potential function φ\varphi(x, y) along the y-direction in the channel can be approximated by a parabolic expression such as:

φi(x,y)=φsi(x)+Ai(x)y+Ai+1(x)y2,for i=1,2,3(3)\varphi_i(x,y)=\varphi_{si}(x)+A_i(x)y+A_{i+1}(x)y^2,\quad\text{for }i=1,2,3 \tag*{(3)}

where φsi(x)\varphi_{si}(x) are the surface potential, and Ai(x)A_i(x) are arbitrary constants for the region I, II, and III that are to be determined using boundary conditions. The Poisson’s equation (1) is solved simultaneously for all three regions in the channel using solution expressed in (3), and given as:

φ1(x,y)=φs1(x)+A1(x)y+A2(x)y2,for 0≤x≤L1(4)\varphi_1(x,y)=\varphi_{s1}(x)+A_1(x)y+A_2(x)y^2,\quad\text{for }0\le x\le L_1 \tag*{(4)}
φ2(x,y)=φs2(x)+A3(x)y+A4(x)y2,for L1≤x≤L2(5)\varphi_2(x,y)=\varphi_{s2}(x)+A_3(x)y+A_4(x)y^2,\quad\text{for }L_1\le x\le L_2 \tag*{(5)}
φ3(x,y)=φs3(x)+A5(x)y+A6(x)y2,for L2≤x≤L3(6)\varphi_3(x,y)=\varphi_{s3}(x)+A_5(x)y+A_6(x)y^2,\quad\text{for }L_2\le x\le L_3 \tag*{(6)}

where φ1(x,y)\varphi_1(x,y), φ2(x,y)\varphi_2(x,y), φ3(x,y)\varphi_3(x,y) and φs1(x)\varphi_{s1}(x), φs2(x)\varphi_{s2}(x), φs3(x)\varphi_{s3}(x) are the channel potential and surface potential in the channel in the region I, II, and III respectively. Furthermore, above equations (4), (5), and (6) are solved using the following boundary conditions for φi(x,y)\varphi_i(x,y) are given by:

dφ1(x,y)dy∣y=0=ε1ε2φs1(x)−VGS′d1(7)\left.\frac{d\varphi_1(x,y)}{dy}\right|_{y=0}=\frac{\varepsilon_1}{\varepsilon_2}\frac{\varphi_{s1}(x)-V'_{GS}}{d_1} \tag*{(7)}
dφ1(x,y)dy∣y=−d1=−qns1ε1=−E1(8)\left.\frac{d\varphi_1(x,y)}{dy}\right|_{y=-d_1}=-\frac{qn_{s1}}{\varepsilon_1}=-E_1 \tag*{(8)}
dφ2(x,y)dy∣y=0=ε1ε2φs2(x)−VGS′d1(9)\left.\frac{d\varphi_2(x,y)}{dy}\right|_{y=0}=\frac{\varepsilon_1}{\varepsilon_2}\frac{\varphi_{s2}(x)-V'_{GS}}{d_1} \tag*{(9)}
dφ2(x,y)dy∣y=−d1=−qns2ε1=−E2(10)\left.\frac{d\varphi_2(x,y)}{dy}\right|_{y=-d_1}=-\frac{qn_{s2}}{\varepsilon_1}=-E_2 \tag*{(10)}
dφ3(x,y)dy∣y=0=ε1ε2φs3(x)−VGS′d1(11)\left.\frac{d\varphi_3(x,y)}{dy}\right|_{y=0}=\frac{\varepsilon_1}{\varepsilon_2}\frac{\varphi_{s3}(x)-V'_{GS}}{d_1} \tag*{(11)}
dφ3(x,y)dy∣y=−d1=−qns1ε1=−E3(12)\left.\frac{d\varphi_3(x,y)}{dy}\right|_{y=-d_1}=-\frac{qn_{s1}}{\varepsilon_1}=-E_3 \tag*{(12)}
φs1(x)∣x=0=Vbi(13)\left.\varphi_{s1}(x)\right|_{x=0}=V_{bi} \tag*{(13)}
φs1(x)∣x=L3=VDS+Vbi(14)\left.\varphi_{s1}(x)\right|_{x=L_3}=V_{DS}+V_{bi} \tag*{(14)}
VGS′=VGS−VFB(15)V'_{GS}=V_{GS}-V_{FB} \tag*{(15)}

where VFB=Φm−Φs1V_{FB}=\Phi_m-\Phi_{s1}, where Φs1=χ1+EG1/2\Phi_{s1}=\chi_1+E_{G1}/2, since the AlN barrier layer is undoped.

Since β\beta-Ga2_2O3_3 does not possess any polarization property, the total polarization charge of the AlN/β\beta-Ga2_2O3_3 heterostructure is a sum of spontaneous as well as piezoelectric polarization of the AlN barrier layer. Total polarization of the AlN barrier PT=Psp+PpiP_T=P_{sp}+P_{pi}, where spontaneous polarization Psp=−0.09 C/m2P_{sp}=-0.09\ C/m^2 [21], and piezoelectric polarization PpiP_{pi} is calculated as given in [21]:

Ppi=2(as−a0a0)e31−C13C33e33(16)P_{pi}=2\left(\frac{a_s-a_0}{a_0}\right)e_{31}-\frac{C_{13}}{C_{33}}e_{33} \tag*{(16)}

where asa_s, a0a_0 are lattice constants of AlN and β\beta-Ga2_2O3_3; e31e_{31}, e33e_{33} and C13C_{13}, C33C_{33} are piezoelectric and elastic constants for AlN material respectively.

Based on full depletion of the AlN barrier layer, the polarization induced 2DEG charge density is given by [17]

ns=ε1qd1(VGO−EF−φi(x))(17)n_s=\frac{\varepsilon_1}{qd_1}\left(V_{GO}-E_F-\varphi_i(x)\right) \tag*{(17)}

where VGO=VGS−VOFFV_{GO}=V_{GS}-V_{OFF}, and φi(x)\varphi_i(x) is the channel potential in regions I, II, III for i=1i=1, 22, 33. Furthermore, cut-off voltage VOFFV_{OFF} is given as

VOFF=ΦB−ΔEC−PTCG(18)V_{OFF}=\Phi_B-\Delta E_C-\frac{P_T}{C_G} \tag*{(18)}

Where ΦB\Phi_B is the Schottky barrier height; ΔEC\Delta E_C is the conduction band offset between AlN and β\beta-Ga2_2O3_3 at the heterointerface; PTP_T is the total polarization charge density, and CGC_G is the gate capacitance.

Since surface potential and the lateral electric field between the two consecutive regions at the heterointerface are continuous. The associated boundary conditions at the interface of region I-II and region II-III are given as

φs1(x)∣x=L1=φs2(x)∣x=L2(19)\left.\varphi_{s1}(x)\right|_{x=L_1}=\left.\varphi_{s2}(x)\right|_{x=L_2} \tag*{(19)}
dφs1(x)dx∣x=L1=dφs2(x)dx∣x=L1(20)\left.\frac{d\varphi_{s1}(x)}{dx}\right|_{x=L_1}=\left.\frac{d\varphi_{s2}(x)}{dx}\right|_{x=L_1} \tag*{(20)}
φs2(x)∣x=L2=φs3(x)∣x=L2(21)\left.\varphi_{s2}(x)\right|_{x=L_2}=\left.\varphi_{s3}(x)\right|_{x=L_2} \tag*{(21)}
dφs2(x)dx∣x=L2=dφs2(x)dx∣x=L2(22)\left.\frac{d\varphi_{s2}(x)}{dx}\right|_{x=L_2}=\left.\frac{d\varphi_{s2}(x)}{dx}\right|_{x=L_2} \tag*{(22)}

Now using boundary conditions of (7) to (12), all six constants of (4) to (6) are determined to get the AlN/β\beta-Ga2_2O3_3 HEMT channel potential as

φ1(x,y)=φs1(x)+ε1ε2(φs1(x)−VGS′2d1)y+{−E12d1+ε1ε2(φs1(x)−VGS′2d1)}y2(23)\varphi_1(x,y)=\varphi_{s1}(x)+\frac{\varepsilon_1}{\varepsilon_2}\left(\frac{\varphi_{s1}(x)-V'_{GS}}{2d_1}\right)y+\left\{-\frac{E_1}{2d_1}+\frac{\varepsilon_1}{\varepsilon_2}\left(\frac{\varphi_{s1}(x)-V'_{GS}}{2d_1}\right)\right\}y^2 \tag*{(23)}
φ2(x,y)=φs2(x)+ε1ε2(φs2(x)−VGS′2d1)y+{−E22d1+ε1ε2(φs1(x)−VGS′2d1)}y2(24)\varphi_2(x,y)=\varphi_{s2}(x)+\frac{\varepsilon_1}{\varepsilon_2}\left(\frac{\varphi_{s2}(x)-V'_{GS}}{2d_1}\right)y+\left\{-\frac{E_2}{2d_1}+\frac{\varepsilon_1}{\varepsilon_2}\left(\frac{\varphi_{s1}(x)-V'_{GS}}{2d_1}\right)\right\}y^2 \tag*{(24)}
φ3(x,y)=φs3(x)+ε1ε2(φs3(x)−VGS′2d1)y+{−E32d1+ε1ε2(φs1(x)−VGS′2d1)}y2(25)\varphi_3(x,y)=\varphi_{s3}(x)+\frac{\varepsilon_1}{\varepsilon_2}\left(\frac{\varphi_{s3}(x)-V'_{GS}}{2d_1}\right)y+\left\{-\frac{E_3}{2d_1}+\frac{\varepsilon_1}{\varepsilon_2}\left(\frac{\varphi_{s1}(x)-V'_{GS}}{2d_1}\right)\right\}y^2 \tag*{(25)}

Now substituting values of the channel potential obtained above in (1), new equations for the region I, II, and III are obtained as

For region I (0≤x≤L10\leq x\leq L_1)

d2φs1(x)dx2−ε1ε2d12φs1(x)=ε1ε2d12{−qNDε2d12ε2+E1d1ε2d12ε1−VGS′}(26)\frac{d^2\varphi_{s1}(x)}{dx^2}-\frac{\varepsilon_1}{\varepsilon_2d_1^2}\varphi_{s1}(x)=\frac{\varepsilon_1}{\varepsilon_2d_1^2}\left\{\frac{-qN_D\varepsilon_2d_1^2}{\varepsilon_2}+\frac{E_1}{d_1}\frac{\varepsilon_2d_1^2}{\varepsilon_1}-V'_{GS}\right\} \tag*{(26)}

For region II (L1≤x≤L2L_1\leq x\leq L_2)

d2φs2(x)dx2−ε1ε2d12φs2(x)=ε1ε2d12{−qNDε2d12ε2+E2d1ε2d12ε1−VGS′}(27)\frac{d^2\varphi_{s2}(x)}{dx^2}-\frac{\varepsilon_1}{\varepsilon_2d_1^2}\varphi_{s2}(x)=\frac{\varepsilon_1}{\varepsilon_2d_1^2}\left\{\frac{-qN_D\varepsilon_2d_1^2}{\varepsilon_2}+\frac{E_2}{d_1}\frac{\varepsilon_2d_1^2}{\varepsilon_1}-V'_{GS}\right\} \tag*{(27)}

For region III (L2≤x≤L3L_2\leq x\leq L_3)

d2φs3(x)dx2−ε1ε2d12φs3(x)=ε1ε2d12{−qNDε2d12ε2+E3d1ε2d12ε1−VGS′}(28)\frac{d^2\varphi_{s3}(x)}{dx^2}-\frac{\varepsilon_1}{\varepsilon_2d_1^2}\varphi_{s3}(x)=\frac{\varepsilon_1}{\varepsilon_2d_1^2}\left\{\frac{-qN_D\varepsilon_2d_1^2}{\varepsilon_2}+\frac{E_3}{d_1}\frac{\varepsilon_2d_1^2}{\varepsilon_1}-V'_{GS}\right\} \tag*{(28)}

The solutions of equations (26) – (28) are obtained using the sum of complementary function (CF) and particular integral (PI) and given as follows

φS1(x)=Aek1x+Be−k1x−p1(29)\varphi_{S1}(x) = A e^{k_1 x} + B e^{-k_1 x} - p_1 \tag*{(29)}
φS2(x)=Cek2x+De−k2x−p2(30)\varphi_{S2}(x) = C e^{k_2 x} + D e^{-k_2 x} - p_2 \tag*{(30)}
φS3(x)=Eek3x+Fe−k3x−p3(31)\varphi_{S3}(x) = E e^{k_3 x} + F e^{-k_3 x} - p_3 \tag*{(31)}

Where

k12=k22=ε1ε2d12(32)k_1^2 = k_2^2 = \frac{\varepsilon_1}{\varepsilon_2 d_1^2} \tag*{(32)}
p1=p3=−qNDε2k12+E1d1k12−VGS′(33)p_1 = p_3 = \frac{-qN_D}{\varepsilon_2 k_1^2} + \frac{E_1}{d_1 k_1^2} - V'_{GS} \tag*{(33)}
p2=−qNDε2k22+E2d1k22−VGS′(34)p_2 = \frac{-qN_D}{\varepsilon_2 k_2^2} + \frac{E_2}{d_1 k_2^2} - V'_{GS} \tag*{(34)}

Now the constants of (29) – (31) are calculated using boundary conditions given in (13), (14) and (19) – (22), and given as

A=e−k1L3e−k1L3−ek1L3(Vbi+p1+p2−p12e−k1L1+p3−p22e−k1L2)−ek1L3e−k1L3−ek1L3(p1−p22ek1L1+p2−p32ek1L2)−(Vbi+VDS+p3e−k1L3−ek1L3)(35)\begin{aligned} A ={}& \frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(V_{bi}+p_1+\frac{p_2-p_1}{2e^{-k_1L_1}}+\frac{p_3-p_2}{2e^{-k_1L_2}}\right)\\ &-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_1-p_2}{2e^{k_1L_1}}+\frac{p_2-p_3}{2e^{k_1L_2}}\right)-\left(\frac{V_{bi}+V_{DS}+p_3}{e^{-k_1L_3}-e^{k_1L_3}}\right) \tag*{(35)} \end{aligned}
B=ek1L3ek1L3−e−k1L3(Vbi+p1)−e−k1L3e−k1L3−ek1L3(p2−p12e−k1L1+p3−p22e−k1L2)−ek1L3e−k1L3−ek1L3(p1−p22ek1L1+p2−p32ek1L2)−(Vbi+VDS+p3e−k1L3−ek1L3)(36)\begin{aligned} B ={}& \frac{e^{k_1L_3}}{e^{k_1L_3}-e^{-k_1L_3}}(V_{bi}+p_1)-\frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_2-p_1}{2e^{-k_1L_1}}+\frac{p_3-p_2}{2e^{-k_1L_2}}\right)\\ &-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_1-p_2}{2e^{k_1L_1}}+\frac{p_2-p_3}{2e^{k_1L_2}}\right)-\left(\frac{V_{bi}+V_{DS}+p_3}{e^{-k_1L_3}-e^{k_1L_3}}\right) \tag*{(36)} \end{aligned}
C=e−k1L3e−k1L3−ek1L3(Vbi+p1+p2−p12e−k1L1+p3−p22e−k1L2)−e−k1L3e−k1L3−ek1L3(p1−p22ek1L1)−ek1L3e−k1L3−ek1L3(p2−p32ek1L2)−(Vbi+VDS+p3e−k1L3−ek1L3)(37)\begin{aligned} C ={}& \frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(V_{bi}+p_1+\frac{p_2-p_1}{2e^{-k_1L_1}}+\frac{p_3-p_2}{2e^{-k_1L_2}}\right)-\frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_1-p_2}{2e^{k_1L_1}}\right)\\ &-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_2-p_3}{2e^{k_1L_2}}\right)-\left(\frac{V_{bi}+V_{DS}+p_3}{e^{-k_1L_3}-e^{k_1L_3}}\right) \tag*{(37)} \end{aligned}
D=ek1L3ek1L3−e−k1L3(Vbi+p1)−ek1L3e−k1L3−ek1L3(p2−p12ek1L1)−e−k1L3e−k1L3−ek1L3(p3−p22e−k1L2)−ek1L3e−k1L3−ek1L3(p1−p22ek1L1+p2−p32ek1L2)−(Vbi+VDS+p3e−k1L3−ek1L3)(38)\begin{aligned} D ={}& \frac{e^{k_1L_3}}{e^{k_1L_3}-e^{-k_1L_3}}(V_{bi}+p_1)-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_2-p_1}{2e^{k_1L_1}}\right)\\ &-\frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_3-p_2}{2e^{-k_1L_2}}\right)-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_1-p_2}{2e^{k_1L_1}}+\frac{p_2-p_3}{2e^{k_1L_2}}\right)\\ &-\left(\frac{V_{bi}+V_{DS}+p_3}{e^{-k_1L_3}-e^{k_1L_3}}\right) \tag*{(38)} \end{aligned}
E=e−k1L3e−k1L3−ek1L3(Vbi+p1+p2−p12e−k1L1+p3−p22e−k1L2)−e−k1L3e−k1L3−ek1L3(p1−p22ek1L1)−e−k1L3e−k1L3−ek1L3(p2−p32ek1L2)−(Vbi+VDS+p3e−k1L3−ek1L3)(39)\begin{aligned} E ={}& \frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(V_{bi}+p_1+\frac{p_2-p_1}{2e^{-k_1L_1}}+\frac{p_3-p_2}{2e^{-k_1L_2}}\right)-\frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_1-p_2}{2e^{k_1L_1}}\right)\\ &-\frac{e^{-k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_2-p_3}{2e^{k_1L_2}}\right)-\left(\frac{V_{bi}+V_{DS}+p_3}{e^{-k_1L_3}-e^{k_1L_3}}\right) \tag*{(39)} \end{aligned}
F=ek1L3ek1L3−e−k1L3(Vbi+p1)−ek1L3e−k1L3−ek1L3(p2−p12e−k1L1)−ek1L3e−k1L3−ek1L3(p3−p22e−k1L2)−ek1L3e−k1L3−ek1L3(p1−p22ek1L1+p2−p32ek1L2)−(Vbi+VDS+p3e−k1L3−ek1L3)(40)\begin{aligned} F={}&\frac{e^{k_1L_3}}{e^{k_1L_3}-e^{-k_1L_3}}(V_{bi}+p_1)-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_2-p_1}{2e^{-k_1L_1}}\right)\\ &-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_3-p_2}{2e^{-k_1L_2}}\right)-\frac{e^{k_1L_3}}{e^{-k_1L_3}-e^{k_1L_3}}\left(\frac{p_1-p_2}{2e^{k_1L_1}}+\frac{p_2-p_3}{2e^{k_1L_2}}\right)\\ &-\left(\frac{V_{bi}+V_{DS}+p_3}{e^{-k_1L_3}-e^{k_1L_3}}\right) \tag*{(40)} \end{aligned}

Analysis of Surface Potential, Channel Potential, Electric Field, and Drain Current

Now the surface potential for the entire channel length (x=0x = 0 to L3L_3) using (29), (30), (31) is given as

φs(x)=φs1(x)+φs2(x)+φs3(x)(41)\varphi_s(x)=\varphi_{s1}(x)+\varphi_{s2}(x)+\varphi_{s3}(x) \tag*{(41)}

Furthermore, using (23), (24), and (25) and substituting y=d1y=d_1, the channel potential for entire channel length (x=0x = 0 to L3L_3) is given as

φ(x,d1)=φ1(x,d1)+φ2(x,d1)+φ3(x,d1)(42)\varphi(x,d_1)=\varphi_1(x,d_1)+\varphi_2(x,d_1)+\varphi_3(x,d_1) \tag*{(42)}

where channel potential for the region I, II, and III are given as

φ1(x,d1)=φs1(x)+ε1ε2(φs1(x)−VGS′)+−E12(43)\varphi_1(x,d_1)=\varphi_{s1}(x)+\frac{\varepsilon_1}{\varepsilon_2}\left(\varphi_{s1}(x)-V'_{GS}\right)+\frac{-E_1}{2} \tag*{(43)}
φ2(x,d1)=φs2(x)+ε1ε2(φs2(x)−VGS′)+−E22(44)\varphi_2(x,d_1)=\varphi_{s2}(x)+\frac{\varepsilon_1}{\varepsilon_2}\left(\varphi_{s2}(x)-V'_{GS}\right)+\frac{-E_2}{2} \tag*{(44)}
φ3(x,d1)=φs3(x)+ε1ε2(φs3(x)−VGS′)+−E32(45)\varphi_3(x,d_1)=\varphi_{s3}(x)+\frac{\varepsilon_1}{\varepsilon_2}\left(\varphi_{s3}(x)-V'_{GS}\right)+\frac{-E_3}{2} \tag*{(45)}

The lateral electric field along the 2DEG channel can be defined as

Ex=−dφ(x)dx(46)E_x=-\frac{d\varphi(x)}{dx} \tag*{(46)}

And from x=0x = 0 to L3L_3, the xx component of the electric field is given as

Ex=Ex1+Ex2+Ex3(47)E_x=E_{x1}+E_{x2}+E_{x3} \tag*{(47)}

Where lateral electric field components in the region I, II, III are obtained using (42), (43), (44) and (45) given as

Ex1=−dφs1(x)dx(1+ε1ε2)(48)E_{x1}=-\frac{d\varphi_{s1}(x)}{dx}\left(1+\frac{\varepsilon_1}{\varepsilon_2}\right) \tag*{(48)}
Ex2=−dφs2(x)dx(1+ε1ε2)(49)E_{x2}=-\frac{d\varphi_{s2}(x)}{dx}\left(1+\frac{\varepsilon_1}{\varepsilon_2}\right) \tag*{(49)}
Ex3=−dφs3(x)dx(1+ε1ε2)(50)E_{x3}=-\frac{d\varphi_{s3}(x)}{dx}\left(1+\frac{\varepsilon_1}{\varepsilon_2}\right) \tag*{(50)}

The drain current density is given as follows

IDS(x)W=qns(x)ve(x)(51)\frac{I_{\mathrm{DS}}(x)}{W}=q n_{\mathrm{s}}(x)v_{\mathrm{e}}(x) \tag*{(51)}

Where WW is the gate width, ve(x)v_{\mathrm{e}}(x) is the electron effective velocity, and is given as

ve(x)=μnExvsatμnE(x)+vsat(52)v_{\mathrm{e}}(x)=\frac{\mu_{\mathrm{n}}E_xv_{\mathrm{sat}}}{\mu_{\mathrm{n}}E(x)+v_{\mathrm{sat}} } \tag*{(52)}

Where electron mobility μn\mu_{\mathrm{n}} is given as per the negative differential mobility model from [21]

μn=μn0+vsatEx(ExEC)γ1+(ExEC)γ(53)\mu_{\mathrm{n}}=\frac{\mu_{\mathrm{n0}}+\frac{v_{\mathrm{sat}}}{E_x}\left(\frac{E_x}{E_{\mathrm{C}}}\right)^\gamma}{1+\left(\frac{E_x}{E_{\mathrm{C}}}\right)^\gamma} \tag*{(53)}

For β\beta-Ga2_2O3_3, the parameters μn0=140\mu_{\mathrm{n0}}=140 cm2^2/Vs, vsat=1.5×107v_{\mathrm{sat}}=1.5\times10^7 cm/s, critical field EC=2.25×105E_{\mathrm{C}}=2.25\times10^5 V/cm, and γ=2.84\gamma=2.84 are taken from [24]. After substituting nsn_{\mathrm{s}} from (17), the drain current given by (51) can be rewritten as

IDS(x)=WCG(VGO−EF)μnExvsatμnEx+vsat(54)I_{\mathrm{DS}}(x)=WC_{\mathrm{G}}(V_{\mathrm{GO}}-E_{\mathrm{F}})\frac{\mu_{\mathrm{n}}E_xv_{\mathrm{sat}}}{\mu_{\mathrm{n}}E_x+v_{\mathrm{sat}}} \tag*{(54)}

Results and Discussion

The simulation results for the device structure (Figure 1) are obtained using the 2D device simulator—ATLAS [21]. The physics-based models and material properties are discussed in previous sections. The device simulator is first calibrated with experimental results measured in [5]. The mathematical model data are obtained using the computational tool—MATLAB [25]. The surface potential profile, channel potential, and electric field profile are extracted from structure files saved at different bias voltages. The surface potential plots are shown in Figure 2. These data are extracted from at the heterointerface of different structure files for three different VGSV_{\mathrm{GS}} of 1 V, −1-1 V, and −3-3 V and constant VDSV_{\mathrm{DS}} of 10 V. The computed data from derived model (41) are also shown.

Image not extracted

Figure 2. The surface potential at the AlN/β\beta-Ga2_2O3_3 interface for different VGS=1V_{\mathrm{GS}}=1, −1-1, −3-3 V at fixed VDS=10V_{\mathrm{DS}}=10 V. Full channel length subdivided into LGSL_{\mathrm{GS}}, LGL_{\mathrm{G}}, and LGDL_{\mathrm{GD}} is also drawn in the lower part of the figure.

The channel potential and electric field simulated data are extracted in the channel 5 nm below the AlN/β-Ga2O3\mathrm{AlN}/\beta\text{-}\mathrm{Ga}_2\mathrm{O}_3 interface. The model data for channel potential and electric field are computed from (42), and (47) respectively. Simulated and model data for channel potential and electric field are shown in Figure 3, and Figure 4 respectively.

Image not extracted

Figure 3. Simulated and modeled data of channel potential along the channel below 5 nm from heterointerface for different VGS=1V_{\mathrm{GS}} = 1, −1-1, −3-3 V at fixed VDS=10V_{\mathrm{DS}} = 10 V. The plot is almost similar to surface potential plot except under LGSL_{\mathrm{GS}} in the latter.

Image not extracted

Figure 4. Lateral electric field along the channel below 5 nm from heterointerface for different VGS=1V_{\mathrm{GS}} = 1, −1-1, −3-3 V at fixed VDS=10V_{\mathrm{DS}} = 10 V. Peak electric field of 1.8×1061.8 \times10^{6} V/cm is evident under the drain side gate edge.

The transfer characteristic of the proposed device is shown in Figure 5. Drain current model data and experimental result relating to similar dimension device is also indicated. Except for the sub-threshold region, there is a good consistency between model data and simulation results. Additionally, a fair degree of agreement is evident with experimental data reported in [5].

Image not extracted

Figure 5. Simulated, Experimental, and modeled data of IDI_\mathrm{D} – VGV_\mathrm{G} characteristics of the analysed device.

Conclusion

The analytical model of surface potential, channel potential, and electric field for AlN/β-Ga2O3\mathrm{AlN}/\beta\text{-}\mathrm{Ga}_2\mathrm{O}_3 is presented. The 2D Poisson equation is solved to develop an expression of potential and lateral field along the channel and consequently extended to find expression of drain current. For the validation of the simulation and model data, the simulation deck is calibrated with device dimensions similar to the experimental one as reported elsewhere. Analytical model data of surface potential, channel potential, and electric field are rigorously validated with different values of gate voltages. Furthermore, drain current model results are compared with simulation results as well as validated by experimental data. A good level of agreement among all three is obtained. Further work in the physics-based analytical models for the β-Ga2O3\beta\text{-}\mathrm{Ga}_2\mathrm{O}_3-based devices can be helpful to understand device mechanism and useful to predict its potential for futuristic high power applications.

References

  1. [1]Z. Hu et al., “Field-Plated Lateral β-Ga2O3 Schottky Barrier Diode with High Reverse Blocking Voltage of More Than 3 kV and High DC Power Figure-of-Merit of 500 MW/cm2,” IEEE Electron Device Lett., pp. 1–1, 2018, doi: 10.1109/LED.2018.2868444.DOI
  2. [2]J. Yang, F. Ren, M. Tadjer, S. J. Pearton, and A. Kuramata, “Ga2O3 Schottky rectifiers with 1 ampere forward current, 650 V reverse breakdown and 26.5 MW.cm-2 figure-of-merit,” AIP Adv., vol. 8, no. 5, 2018, doi: 10.1063/1.5034444.DOI
  3. [3]M. Higashiwaki et al., “Depletion-mode Ga2O3 metal-oxide-semiconductor field-effect transistors on β-Ga2O3 (010) substrates and temperature dependence of their device characteristics,” Appl. Phys. Lett., vol. 103, no. 12, pp. 1–5, 2013, doi: 10.1063/1.4821858.DOI
  4. [5]Z. Xia et al., “Delta Doped β-Ga2O3 Field Effect Transistors With Regrown Ohmic Contacts,” IEEE Electron Device Lett., vol. 39, no. 4, pp. 568–571, Apr. 2018, doi: 10.1109/LED.2018.2805785.DOI
  5. [6]Z. Feng et al., “Normally-Off-β-Ga2O3 Power MOSFET with Ferroelectric Charge Storage Gate Stack Structure,” IEEE Electron Device Lett., vol. 41, no. 3, pp. 333–336, 2020, doi: 10.1109/LED.2020.2970066.DOI
  6. [7]S. Krishnamoorthy et al., “Modulation-doped (β-Al 2 O 3) 2 O 3 / Ga 2 O 3 field-effect transistor,” Appl. Phys. Lett., vol. 111, no. 2, pp. 023502, Jul. 2017, doi: 10.1063/1.4993569.
  7. [8]Y. Zhang et al., “Demonstration of high mobility and quantum transport in modulation-doped β-(AlxGa1-x)2O3/Ga2O3 heterostructures,” Appl. Phys. Lett., vol. 112, no. 17, pp. 1–6, 2018, doi: 10.1063/1.5025704.DOI
  8. [9]W. Wei et al., “Valence band offset of β-Ga2O3/wurtzite GaN heterostructure measured by X-ray photoelectron spectroscopy,” Nanoscale Res. Lett., vol. 7, no. 1, p. 562, 2012, doi: 10.1186/1556-276X-7-562.DOI
  9. [12]S. Lyu and A. Pasquarello, “Band alignment at β -Ga 2 O 3 /III-N (III = Al, Ga) interfaces through hybrid functional calculations,” Appl. Phys. Lett., vol. 117, no. 10, p. 102103, Sep. 2020, doi: 10.1063/5.0020442.DOI
  10. [13]R. Singh, T. R. Lenka, R. T. Velpula, B. Jain, H. Q. T. Bui, and H. P. T. Nguyen, “A novel β-Ga2O3 HEMT with fT of 166 GHz and X-band POUT of 2.91 W/mm,” Int. J. Numer. Model. Electron. Networks, Devices Fields, vol. 34, no. 1, pp. 1–11, 2021, doi: 10.1002/jnm.2794.DOI
  11. [14]R. Singh, T. R. Lenka, and H. P. T. Nguyen, “Analytical Study of Conduction Band Discontinuity Supported 2DEG Density in AlN/β-Ga2O3 HEMT,” Facta Universitatis, Series: Electronics and Energetics, 06 May 2021. (Accepted, in press)
  12. [15]S. P. Kumar, A. Agrawal, R. Chaujar, S. Kabra, M. Gupta, and R. S. Gupta, “Threshold voltage model for small geometry AlGaN/GaN HEMTs based on analytical solution of 3-D Poisson’s equation,” Microelectronics J., vol. 38, no. 10–11, pp. 1013–1020, 2007, doi: 10.1016/j.mejo.2007.09.001.DOI
  13. [16]S. Khandelwal, N. Goyal, and T. A. Fjeldly, “A physics-based analytical model for 2DEG charge density in AlGaN/GaN HEMT devices,” IEEE Trans. Electron Devices, vol. 58, no. 10, pp. 3622–3625, 2011, doi: 10.1109/TED.2011.2161314.DOI
  14. [17]S. Khandelwal and T. A. Fjeldly, “A physics based compact model of I-V and C-V characteristics in AlGaN/GaN HEMT devices,” Solid State. Electron., vol. 76, pp. 60–66, 2012, doi: 10.1016/j.sse.2012.05.054.DOI
  15. [18]F. M. Yigletu, S. Khandelwal, T. A. Fjeldly, and B. Iniguez, “Compact charge-based physical models for current and capacitances in AlGaN/GaN hemts,” IEEE Trans. Electron Devices, vol. 60, no. 11, pp. 3746–3752, 2013, doi: 10.1109/TED.2013.2283525.DOI
  16. [19]M. Kaddeche, A. Telia, and A. Soltani, “Analytical modeling and analysis of Al m Ga1-mN/GaN HEMTs employing both field-plate and high-k dielectric stack for high-voltage operation,” J. Comput. Electron., vol. 12, no. 3, pp. 501–510, 2013, doi: 10.1007/s10825-013-0468-5.DOI
  17. [20]S. Khandelwal, S. Ghosh, Y. S. Chauhan, B. Iniguez, and T. A. Fjeldly, “Surface-Potential-Based RF Large Signal Model for Gallium Nitride HEMTs,” 2015 IEEE Compd. Semicond. Integr. Circuit Symp. CSICS 2015, pp. 1–4, 2015, doi: 10.1109/CSICS.2015.7314527.DOI
  18. [21]Device Simulation Software, ATLAS User’s Manual, Silvaco, Santa Clara, CA, USA, 2019.
  19. [22]A. Mock et al., “Band-to-band transitions, selection rules, effective mass, and excitonic contributions in monoclinic β-Ga 2 O 3,” Phys. Rev. B, vol. 96, no. 24, pp. 1–14, 2017, doi: 10.1103/PhysRevB.96.245205.DOI
  20. [23]S. Poncé and F. Giustino, “Structural, electronic, elastic, power, and transport properties of β-Ga2O3 from first principles,” Physical Review Research, vol. 2, no. 3. p. 033102, 2020, [Online]. Available: https://link.aps.org/doi/10.1103/PhysRevResearch.2.033102.DOI
  21. [24]K. Ghosh and U. Singisetti, “Ab initio velocity-field curves in monoclinic β-Ga 2 O 3,” J. Appl. Phys., vol. 122, no. 3, p. 035702, Jul. 2017, doi: 10.1063/1.4986174.DOI
  22. [25]MATLAB Version 7.10.0 (R2018a). Natick, Massachusetts: The MathWorks Inc.; 2018.

Paper details

vlnc:W_vfvb9hbp
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents