“Soft errors have become a huge concern in advanced computer chips because, uncorrected, they produce a failure rate that is higher than all the other reliability mechanisms combined!”
• R. Baumann, IEEE-TDMR, 2005
While neutron radiation effects on Earth are extremely benign from the standpoint of a human being, the situation is markedly different for microelectronics. A microchip in an aircraft, for example, may be struck by a neutron every few minutes; and when such neutrons hit silicon, a nuclear reaction can occur which results in a shower of electrical charge that can significantly interfere with the normal operation of an integrated circuit.
Although such single event effects have been recognized as an issue for some time, the problem has become critical as a direct result of integrated circuit miniaturization. Decades of effort towards the increased miniaturization and memory density in computers has, for reasons that may be intuitive, increased the vulnerability of microelectronics to single event effects. Simply put: the smaller electronics get, the more susceptible they are to radiation.
High-energy particles contained in galactic cosmic rays (GCRs) enter the Earth’s atmosphere and interact with atmospheric nuclei—primarily nitrogen and oxygen. Collisions between GCRs and the atmosphere produce a wide variety of particles, most of which recombine quickly (charged particles are halted in relatively short range). However, significant quantities of high-energy neutrons are also produced by such collisions, due to cascading spallation reactions. While a large fraction of such neutrons are attenuated by the atmosphere, a still significant number penetrate the atmosphere and reach the Earth’s surface, where they can affect electronic devices.
Single Event Upsets
Neutron radiation effects can have a significant impact, particularly on memory, by causing it to change state (for example, by forcing a single bit in an SRAM FPGA to change state, from a 1 to 0 or vice-versa; this is referred to as a single event upset [SEU]). SEUs are also referred to as soft errors, and manifest themselves as corrupt program data. This can become more serious if the affected memory contains essential configuration information for a device; in these cases, the change in memory state can be much more critical, and may cause impaired operation or a complete loss of functionality. This more serious effect is often referred to as a firm error.
Single Event Latchups
Unlike memory errors, single event latchups (SELs) are potentially physically destructive conditions, in which a radiation-induced error causes a device to consume an excessive amount of current. In many SELs, this excessive induced current can be destructive if the current isn’t limited and removed “in time”. A “micro-latch” is a subset of SELs where a large but non-destructive amount of current is induced. Once latched-up, a device can only be recovered by removing all power. Such errors are often referred to as hard errors.
Single Event Burnout and Gate Rupture
Energetic neutrons can also induce catastrophic failures in power electronics, particularly by causing single event burnout and single event gate rupture in field effect transistors. Unlike other devices, power electronics do not follow down-scaling dimensions and voltage; instead, they exhibit larger sensitivity to neutron-induced effects, even in sea-level applications. Neutron-induced failure of high-voltage electronics can be particularly destructive and dramatic.