Refinement of low-resistance Ni-Ge-Au ohmic contacts to n + GaAs using screening and response surface experiments
- PDF / 147,503 Bytes
- 6 Pages / 612 x 792 pts (letter) Page_size
- 5 Downloads / 244 Views
MATERIALS RESEARCH
Welcome
Comments
Help
Refinement of low-resistance Ni–Ge–Au ohmic contacts to n1 GaAs using screening and response surface experiments Nancy E. Lumpkin and Warren King Division of Radiophysics, CSIRO, P.O. Box 77, Epping, New South Wales 2121, Australia
T. L. Tansley Physics Department, Semiconductor Science & Technology Laboratories, Macquarie University, New South Wales 2109, Australia (Received 10 October 1995; 15 January 1996)
Multivariable screening and response surface experiments have been performed to model ohmic contact resistance sRc d of a Ni –Ge– Au ohmic metal process for n1 GaAs-based high electron mobility transistors (HEMTs). Seven variables were examined via a fractional factorial screening experiment to rank the effects of each process variable. The results of the screening experiment indicated that the most significant variables were total Ge and Au evaporated thickness, Ge-to-Au ratio, and the post-alloy cooling time. Response surface experiments were designed around these three variables to examine the first- and second-order effects. The results enabled the development of an empirical model of ohmic contact resistance from which a new low value of 0.03 6 0.03 V ? mm (one-sigma) was predicted. Twenty confirmation runs on the new process indicated an average Rc of 0.06 6 0.02 V ? mm (one-sigma), with a range of 0.02 V ? mm to 0.11 V ? mm, a reduction from the previous average process value of 0.14 6 0.07 V ? mm (one-sigma).
I. INTRODUCTION
Multivariable experimental design approaches have traditionally been used by chemists to examine and optimize yields in the production of chemicals.1,2 In recent years this approach has been adopted by the microelectronics industry to refine production processes with impressive results.3 Multivariable design approaches have expedited the development of advanced processes such as fine-line lithography and new plasma etching steps.4–6 Multivariable design experiments systematically examine the effect of each variable and its interactions to maximize the information gained and minimize the number of experimental runs. This leads to a more efficient use of time and therefore reduces costs.7,8 There are also the advantages of quantifying interaction effects not observed in single variable experiments; if interaction effects are not considered, optimum results may not be achieved. First-order (linear) and secondorder (nonlinear) estimates of the influence of each variable can be computed and empirical models constructed from these data.9,10 We describe here the refinement of a low-resistance Ni–Ge–Au ohmic metal process for contacting n1 GaAs HEMT structures. The high electron mobility of GaAs makes it an ideal material for microwave and millimeter-wave applications such as monolithic microwave integrated circuit (MMIC) amplifier technology. However, the desirable 1238
http://journals.cambridge.org
J. Mater. Res., Vol. 11, No. 5, May 1996
Downloaded: 14 Mar 2015
device characteristics of high gain and low noise can be rapidly degraded
Data Loading...
