No abstract

No abstract

No abstract

ABSIIWCTIn this paper the hysteretic (historydependent) propagation gate delay of floating-body (FE) partially depleted (PD) SO1 CMOS circuits is investigated. The change in gate propagation delay with time is examined with no preconditioning of the floating-body. The simulationbased analysis includes the sensitivity of the hysteresis to supply voltage, WpNn (beta ratio), duty cycle, slew rate, output load, and initial state of the circuit. Basic physical mechanisms underlying the hysteretic circuit behavior are examined. The results identify the main contributors and general trends of hysteresis in FE PD/SOI circuits. The insight gained can ultimately be incorporated into conventional circuit timing tools. The results also reveal a circuit sizing methodology to minimize the hysteresis effects in circuits using PD/SOI technology. I. INTROOUC~ONHysteresis in floating-body partially depleted silicon-on-insulator (SOI) CMOS, a problematic effect that results in a history-dependent gate propagation delay, is due to the slow carrier recombination/generation processes in the body region of the FB SO1 MOSFET [l, 21. Hysteresis has been shown to cause pulses to sht?tch [3] and/or shrink [41 as they propagate down an inverter chain. Experimentally, hysteretic delays have been shown to range between 5-30%[5, 61, It is essential that circuit designers quantify and contain this delay variability without undermining the performance benefits that a PD/SOI technology offers.

Accurate proximity correction has proven essential for the patterning of submicron features using electron beam lithography. The use of a two-Gaussian model, which accounts for the finite beam size and forward scattering in the resist as well as backscattering, has demonstrated widespread success. It has been shown, however, that in certain instances, such as for features of order 100 nm or less or for exposure on high atomic number substrates, the two-Gaussian expression is unable to adequately fit the absorbed energy distribution in the resist. Suggested modifications, such as the addition of a third Gaussian term to account for large angle electron scattering, or the inclusion of an exponential term which may account for an increased absorption rate in high Z materials, have resulted in improved fits. This paper describes a study to determine the improvement gained in exposed features by including additional Gaussian terms in the expression for the absorbed energy distribution in the resist. A very high resolution probe (beam diameter ∼20 nm FWHM) is used so that forward scattering effects in the resist may be separated from the primary beam distribution. PMMA is exposed on Si and GaAs substrates at 25 keV using proximity correction parameters generated by curvefitting the three-Gaussian model and the two-Gaussian model to the absorbed energy distributions. The three-Gaussian model is seen to provide improved proximity correction particularly in the 100 nm size scale.