Monte Carlo simulation of electronic characteristics in short channel δ-doped AlInAs/GaInAs HEMTs

doi:10.1016/S0026-2714(00)00211-0

Microelectronics Reliability

Volume 41, Issue 1, January 2001, Pages 73-77

https://doi.org/10.1016/S0026-2714(00)00211-0 Get rights and content

Abstract

We present a microscopic analysis of electronic noise in short channel δ-doped AlInAs/GaInAs HEMTs. A classical Monte Carlo device simulation, appropriately modified to locally introduce the effects of electron degeneracy and nonequilibrium screening, is used for the calculations. Even if the energy quantization in the channel is not taken into account in the Monte Carlo model, its validity has been checked by means of the comparison with experimental results of static characteristics, small signal behavior and noise performance in a recessed 0.1 μm T-gate δ-doped HEMT (InP based). The geometry and layer structure of the simulated HEMT is completely realistic, including recessed gate and δ-doping configuration and also the T-shape of the gate and the dielectric disposition has been included in the simulation

Introduction

With the aim of increasing the cut-off frequency of transistors, its gate length, L_g, has been reduced to the technological limit. However, this scaling process has the drawback that the parasitic gate resistance, r_g, increases proportionally to 1/L_g, thus deteriorating the transconductance and, consequently, other important figures of merit of the devices like current gain and noise figure. Accordingly, while decreasing L_g, the value of the parasitic gate resistance must be kept as low as possible. A good compromise between these two requirements is the use of the T-gate technology [1], [2], which allows to have a short L_g (corresponding to the base of the T) with a low value of r_g, similar to that associated to a longer gate (corresponding to the head of the T). The wider is the head of the T-gate the lower is r_g, but at the same time the gate capacitance increases [2]. Therefore the width of the T-gate head must be chosen as a trade-off between low resistance and low capacitance. The use of the recessed geometry to improve the device characteristics is also widespread. All these refinements lead to a rather complex configuration difficult to simulate. Indeed, usually Monte Carlo (MC) simulations of MESFETs and HEMTs only consider the transport and electric field inside the semiconductor. This approach is not completely correct, since it does not take into account the capacitive coupling between the semiconductor and the T-gate taking place through the dielectric (generally Si₃N₄) used to passivate the devices. In this work we will perform the simulation of the complete configuration and its validity will be checked through the comparison with experimental measurements of I–V characteristics, small signal equivalent circuit elements, and minimum noise figure of a real 0.1 μm recessed gate δ-doped HEMT.

Section snippets

Monte Carlo simulation

For the calculations we use a semiclassical ensemble MC simulator self-consistently coupled with a 2D Poisson solver (finite differences scheme, LU decomposition method) which allows the resolution of the potential in complicated geometries and nonuniform meshes. Three non-parabolic spherical valleys (Γ, L and X) with ionized impurity, alloy, polar and non polar optical phonon, acoustic phonon and intervalley scattering mechanisms are taken into account. Material parameters for the Al_0.48In_0.52

Device structure

The SEM photographs of the cross-section of a real 0.1 μm recessed-gate δ-doped HEMT fabricated at the IEMN shown in Fig. 1(a) outline the geometry of the usual recessed T-gate configuration [2], [3]. The layer structure of this HEMT, which has been reproduced in the simulation, was already presented in Ref. [3] and consists of an InP substrate, a 3000 Å Al_0.48In_0.52As buffer followed by a 250 Å thick Ga_0.47In_0.53As channel, three Al_0.48In_0.52As layers (a 50 Å spacer, a 5×10¹² cm^–2 δ-doped

Results

By adjusting separately the surface charge on the cap layer and on the bottom of the recess (whose values are, respectively, $6.2×10^{12} cm^{−2}$ and 4.3×10¹² cm^–2), the static I–V characteristics of the real HEMT have been reproduced quite closely by the simulation (Fig. 2).

In order to evidence the influence of the T-shape of the gate on the potential distribution, Fig. 3 shows the equipotential lines calculated with the MC simulation. The intrinsic bias is V_ds=0.75 V, V_gs=−1.0 V, corresponding,

Conclusions

Taking as a basis the geometry and parameters of a real structure, and using a MC simulation, we have performed a realistic analysis of a AlInAs/GaInAs, InP based HEMT, where the T-gate configuration has been included for first time in the MC simulations. The I–V characteristics, small signal equivalent circuit parameters and minimum noise figure measured in a real device are favorably compared with the simulation results. We have confirmed the importance of reducing the gate resistance to

Acknowledgements

This work was supported partially by the projects SA44/99 from the Consejerı́a de Educación y Cultura de la Junta de Castilla y León and PB97-1331 from the Dirección General de Enseñanza Superior e Investigación Cientı́fica.

References (7)

K Onodera et al.
Extremely low-noise performance of GaAs MESFET's with wide-head T-shaped gate
IEEE Trans Electron Dev
(1999)
L.D Nguyen et al.
50-nm self-aligned-gate pseudomorphic AlInAs/GaInAs high electron mobility transistors
IEEE Trans Electron Dev
(1992)
Mateos J, González T, Pardo D, Hoel V, Happy H, Cappy A. Improved Monte Carlo algorithm for the simulation of δ-doped...

There are more references available in the full text version of this article.

Cited by (0)

View full text