APPLICABILITY OF LET TO SINGLE EVENTS IN MICROELECTRONIC STRUCTURES

LET is often used as a single parameter to determine the energy deposited in a microelectronic structure by a single event. The accuracy of this assumption is examined for ranges of ion energies and volumes of silicon appropriate for modem microelectronics. It is shown to be accurate only under very restricted conditions. Significant differences arise because (1) LET is related to energy lost by the ion, not energy deposited in the volume, and (2) LET is an average value and does not account for statistical variations in energy deposition. Criteria are suggested for determining when factors other than LET should be considered, and new analytical approaches are presented to account for them. One implication of these results is that improvements can be made in space upset rate predictions by incorporating the new methods into currently used codes such as CREME and CRUP. This necessitates implementing the calculations with a cosmic ray intensity spectrum that is a function of cosmic ray energy per unit mass rather than the currently used intensity vs. LET spectrum. Another implication is that an improved understanding of the information contained in measured heavy ion upset cross section curves can be gained through use of the new methods. This can be accomplished by, first of all, measuring and analyzing the curves as a function of energy deposited in the sensitive volume, not energy lost by the ion as is currently the practice. In some situations, improved values of SEU thresholds and critical charges will result. Selection of ions at test facilities can also be important when using different ions to test the same chip. A choice should be made such that the energy deposited by the two ions overlaps. This is not necessarily the same as choosing ions with overlapping effective LETs. Making the latter choice and plotting the experimental results as a function of effective LET can cause a discontinuity in the cross section curve at the point of the ion beam change. Finally, consideration of statistical fluctuations in energy deposition may lead to a better understanding of the origin of the width of the quickly rising portion of the cross section curve. All of the above considerations increase in significance for devices with smaller dimensions and for ions with higher incident energies. Therefore, they will continue to grow in importance for future generation devices.