Introduction
Wide bandgap group III-nitride semiconductors are currently being pursued for possible high temperature and high power applications. Current gain as high as 105 is reported for GaN/SiC HBT [Reference Pankove, Chang, Lee and Molnar1]. In order to investigate high frequency performance the behavior of the base transit time, τb , needs to be investigated. Mohammad et al. [Reference Mohammad and Morkoc2] has reported the dependence of τb on base doping concentration in a graded GaN/InGaN HBT.
The double integral formulation of τb by Kroemer [Reference Kroemer3] for HBTs is based upon the assumption that excess minority carrier concentration at the edge of base-collector depletion layer is negligible. Roulston [Reference Roulston4] emphasized upon the use of a finite carrier velocity (saturation velocity) at the edge of the base-collector depletion region, thereby, introducing the component of base transit time due to velocity saturation. A more general formulation for carrier velocity at b-c heterojunction was used by Hafizi et al. [Reference Hafizi, Streit, Tran, Kobayashi, Umemoto, Oki and Wang5]. Jahan et al. [Reference Jahan and Anwar6], based upon a self-consistent calculation of thermionic and tunneling components of the total current, proposed a tunneling factor for the determination of carrier velocity at the b-c junction. In this paper, the method formulated by Jahan et al. [Reference Jahan and Anwar7] is used in the determination of the effective electron velocity at base-collector junction which affects the electrons transport across base-collector junction.
In this paper, τb in abrupt GaN/InGaN/GaN and AlGaN/GaN/AlGaN HBT is reported. The computation of τb includes the effects of bandgap narrowing, carrier saturation and partitioning of the total current into thermionic and tunneling components. Results obtained from an ensemble Monte Carlo simulation are used in the determination of low field mobility, μn .
Theory
Base transit time, τb , can be expressed as a sum of τb 1 + τb 2, where
and
where NAB (x) is the base doping concentration, nib 2 (x) = nie 2 exp(Δ g/kT) is the effective intrinsic carrier concentration in base region. The effective bandgap narrowing across the emitter-base heterojunction, Δ g ( InxGa1−xN)=x·Eg (InN)+(1−x)·Eg (GaN)−x· (1−x)Egb (Egb=1.0 eV)[Reference Koida, Itoh, Khan, Hiramatsu, Sawaki and Akasaki8] where x is the In-mole fraction in InxGa1−xN and nie is the effective intrinsic carrier concentration in emitter region.τb 1 is the component of the base transit time due to diffusion in the neutral base region and τb 2 accounts for the finite base-collector junction velocity, Sc. The effective minority carrier velocity, Sc, characterizing electron transport across the space-charge region at the base-collector junction is formulated as Sc =vth ·γ·exp[(qvjp −ΔEc)/kT] [Reference Parikh and Lindholm9], where ΔEc is the conduction band discontinuity, vjp is the applied voltage drop at the collector, γ is the tunneling transmission factor, and vth is the thermal velocity. The factor γ is determined by invoking the proper partitioning of the total current into tunneling and thermionic components [Reference Jahan and Anwar7]. WKB method is used to compute the transmission probability required in calculating the tunneling component of the total current.
Results and Discussions
5 × 4 μ m2 single finger npn GaN/InxGa1−xN and Al0.2Ga0.8N HBTs are investigated. The material parameters used in the simulation are as follows: me(GaN)=0.2m0, mh(GaN)=0.6m0 [Reference Khan, Skogman, Hove, Krishanakutty and Kolbas10], me(InN)=0.115m0 [Reference Morkoc, Strite, Gao, Lin, Sverdlov and Burns11], mh(InN)=1.6m0 [Reference Jenkins and Dow12], me(AlN)=0.314m0, mh(AlN)=0.71m0 [Reference Shur and Khan13], EgGaN(T)=3.503+5.08 × 10−4 × T2/(T-996) eV, EgInN(T)=2.01−1.8 × 10−4 ×T eV, EgAlN(T)=6.118−1.799 × 10−3T2/(T+1462) eV [Reference Dmitriev and Oruzheinikov14], ε GaN=9.5ε 0, ε InN=9.50 ε 0 and ε AlN==8.5ε 0 [Reference Shur and Khan13] where ε is static dielectric constant, m0 is electron rest mass, T is absolute temperature (K), and ε 0 is permittivity in vacuum. Emitter and collector doping concentrations equal 5 × 1017 cm−3. The base is 0.05um wide and is doped uniformly 1019 cm−3. The conduction band offset, ΔEc, is assumed to be 75% of the difference in bandgaps of the constituting semiconductor alloys [Reference Richard and Anwar15].
In Fig. 1(a) the low field mobility, μn , and the diffusion coefficient, Dn , are plotted as a function of temperature for In0.2Ga0.8N and GaN at a doping concentration of 1019 cm−3. The low field mobility data is obtained from an ensemble Monte Carlo simulation which accounts for piezoelectric, ionized impurity, alloy, intervalley, acoustic and polar optical phonon scattering [Reference Wu, Webster and Anwar16].The diffusion constant Dn =kT/q·μn ·[F 1/2 (ζn)/F −1/2 (ζn)] where ζn =(Efn −Ec)/kT is the reduced Fermi level for electrons and F ±1 2 (ζn) is the Fermi-Dirac integral. Dn initially increases followed by gradual decrease beyond T=200K. In Fig. 1(b) the base-collector electron junction velocity, Sc, is plotted as a function of base-collector bias, Vbc, for varying In-mole fraction with temperature as a parameter. Due to a smaller band offset at the b-c junction Sc approaches thermal velocity at lower Vbc for x=0.1 as compared to that at x=0.2. The behavior of Sc for AlGaN/GaN/AlGaN is similar to that of GaN/InGaN/GaN.
In Fig. 2(a) the base transit time τ b is plotted as a function of Vbe for Vbc = 3.0 V. τ b decreases with increasing In-mole fraction in GaN/InxGa1−xN/GaN HBTs, irrespective of temperature and this behavior can be explained with the aid of Fig. 1. A higher SC at x=0.1 results in a lower τ b2 as compared to that at x=0.2 for Vbc = 3V. The magnitude of τ b1 at x=0.1 is slightly greater than that at x=0.2. This may be attributed to a slightly higher low field mobility at x=0.1 that results in a lower diffusion constant as compared to that at x=0.2. The relative magnitudes of τ b1 and τ b2 reflects the role of the transport processes namely drift-diffusion versus thermionic field emission. As seen from Fig 2(b), the diffusion controlled component of the base transit time τ b1 is always the dominating time constant at elevated temperatures. However, at higher Dn at 300K results in a lower τ b1 and a higher τ b2. Or in other words, at 300K the contribution of thermionic field emission related component of the base transit time τ b2 becomes significant. τ b for Al.2Ga.8N/GaN HBTs at Vbc =3.0 V is controlled by τ b2 and is due to the small electron effective velocity.
The dependence of τb upon base width for a base doping concentration of 1019 cm−3 at room temperature is shown in Fig.3. Vbe=2V and Vbc=3V are assumed. As expected, τb increases with increasing base width. For extremely narrow base widths, the dominant component of τb is τ b2. As compared to InxGa1−xN-based HBTs, GaN-based HBTs have higherτb due to lower low field mobility and lower effective electron velocity at base-collector junction. The above calculation is performed using stationary transport. The validity of the above approximation is shown in Fig. 4, where the average velocity is plotted as a function of distance. As noted, the average velocity remains constant over a base width variation from 0.02μm to 0.8μm for both GaN and In.2Ga.8N in the presence of an applied field of 500KV/cm. For comparison a similar plot for GaAs is also shown where the applied electric field is 10KV/cm. For GaAs the effect of non-stationary transport is clear and for base widths less than 1τm assuming stationary transport data becomes questionable.
In Fig. 5, the unity gain cutoff frequency fT =1(2 · π · τec) is plotted as a function of collector current density, Jc, where τec = τe + τb + τc is the total transit time from emitter to the collector.τe is emitter charging time, expressed as τe = re ·Cje where re is the dynamic resistance and Cje is the base-emitter junction capacitance.τc = Rc · Cjc is the collector charging time where Rc is the collector resistance and Cjc is the collector junction capacitance. For collector current densities below 100 A/cm2 τ e dominant time constant, however for current densities above 100 A/cm2 τ b dominates. An fT of 19.1 GHz is obtained at 500K for GaN/In.2Ga.8N/GaN DHBT. A higher low field mobility and b-c junction velocity in Al0.2Ga0.8N/GaN DHBT produce a fT of 10.6 GHz at 300K .
Conclusion
Base transit time in abrupt GaN/InGaN/GaN and AlGaN/GaN/AlGaN HBTs are determined by accounting for bandgap narrowing, carrier degeneracy and the proper low field mobility. A narrow base width with a consequently large base resistance is recommended along with a lower In-mole fraction to realize superior unity gain current cut-off frequency, fT .fT =10.6 GHz at T=500K for Al0.2Ga0.8N/GaN HBT and fT = 19.1 GHz at T=300K for GaN/In0.2Ga0.8N have been demonstrated.