Picosecond Lasers for Single-Event Effects Testing

The interaction of a picosecond laser pulse with a semiconductor material can generate a high density of electron-hole pairs (also known as charge carriers), much like the passage of an ionizing particle through the device.

Picosecond Lasers for Single-Event Effects Testing

Steven Moss and Stephen LaLumondiere

In the past 10 years, Aerospace has developed a state-of-the-art facility that uses picosecond laser pulses to simulate the transient effects of energetic particles striking microelectronic devices. This system is used to diagnose radiation-hardened designs and to validate radiation-hardening techniques for mitigating the effects of space radiation on integrated circuits.

Microelectronic and optoelectronic devices used in satellite systems must operate in an extremely harsh environment. Energetic particles can strike sensitive nodes in devices, causing permanent damage or transient events. Phenomena associated with the trail of charge produced by the strike of a single energetic particle are commonly referred to, by members of the radiation-effects community, as single-event effects. These can cause temporary or permanent changes in the state or performance of a device.

Testing microelectronics for their susceptibility to single-event effects is typically done by exposing them to an ion beam from a particle accelerator. This method simulates the hostile space environment fairly well, but can be both costly and time consuming. To meet the need for a cheaper alternative, Aerospace began investigating the feasibility of using laser pulses to simulate the effects of energetic particles in 1992 (see sidebar, Origins of Laser Testing for Single-Event Effects).

Thanks to intensive efforts in the laser-test community, laser-based testing of microelectronic devices for single-event effects has gained widespread acceptance in the radiation-effects community as a useful complement to traditional testing methods. Today, the driving force behind the use of this technique is the ability to pinpoint sensitive nodes with submicron accuracy.

Laser Simulation of Cosmic Ray Effects

The interaction of a cosmic ray in an integrated circuit generates a dense electron-hole plasma inside the semiconductor material; so does the absorption of a picosecond laser pulse. Both the particle and laser interactions occur on a short time scale—much shorter than the response time of most microelectronic devices. Although the initial charge profile produced by absorption of a laser pulse is somewhat different from that produced by the interaction of a cosmic ray, both events produce a highly localized trail of charge capable of generating single-event effects in microelectronic devices.

In testing for susceptibility to single-event effects, a technique known as mode-locking is used to generate a train of laser pulses, each of which lasts only a few picoseconds. An electro-optic shutter grabs individual pulses, which are then focused through a microscope onto the device under test. A camera attached to the microscope shows the position of the laser beam. Devices are scanned beneath the laser beam to locate sensitive nodes. High-speed digital oscilloscopes, transient digitizers, and logic analyzers capture the response of devices to charges generated in the semiconductor material by the incident laser pulse.

Testing with heavy ions consists of irradiating the entire device in a particle-beam accelerator and determining the upset-sensitive cross section based upon the incident ion flux and the number of upsets observed. The technique is global in nature, generally indicating whether or not an upset occurred, but not where on the device it originated. Also, because the technique relies on random particle strikes over the entire area of the device, temporal information is lost.

Testing with a pulsed laser provides several capabilities not offered by particle-beam testing. For example, the small spot sizes achievable with a laser and the ability to precisely position the device relative to the laser beam allow sensitive device nodes to be pinpointed with submicron accuracy. The laser produces no permanent damage in the device, so repeated measurements can be made at a single sensitive location. The laser pulse can also be synchronized with the clock signal of the device to study temporal effects on sensitivity to single-event phenomena.

The laser system can also be used to verify operation of test equipment before embarking on the more costly journey to an accelerator facility. Unlike most particle-beam facilities, the laser facility does not require devices to be placed in a vacuum chamber for testing, and support electronics can be located close to the device under test. This is an extremely important feature when testing high-speed devices for their susceptibility to single-event transients.

The practicality of this technique is limited by the inability of the laser light to penetrate metal layers covering sensitive device nodes. Complex devices with many layers of metal limit the ability to determine the amount of incident light on a sensitive junction; however, other approaches such as thinning and testing devices from behind are viable alternates to the standard front test method.

The Facility

Over the years, various organizations in the United States have used lasers to simulate single-event effects, but only Aerospace and the Naval Research Laboratory currently possess dedicated laser facilities for this work. Researchers in the radiation-effects community have come to rely upon these facilities because of the unique capabilities they provide.

A detailed one-dimensional sensitivity map displaying the threshold laser pulse energy required to induce latchup in the four-terminal test structure. The inset shows the location and direction of the scan. The most sensitive location (i.e., the location requiring the least pulse energy for latchup) is found near the edge of the negative-current well. The two double peaks represent where the laser was scanned over the metal contacts. The metal lines block any incident laser light. The laser spot has a Gaussian spatial profile with a spot size on the same order as the width of the metal lines. Consequently, there is always some light that propagates past the metal lines into the device, even when the laser spot is centered on one of the lines. Thus, the devices can still be latched up by increasing the laser pulse energy; however, as shown here, this requires a considerable increase in energy (view larger image).

For example, the Aerospace laser test system can produce a train of pulses at a variable repetition frequency or operate in a single-shot mode. The system uses dye lasers to generate picosecond optical pulses; the laser wavelength can be tuned over the visible spectrum and into the near infrared.

Two wavelengths are generally used at Aerospace to measure laser-induced single-event effects. The first, 600 nanometers, has a penetration depth of about 2 microns in silicon. The second, 815 nanometers, has a penetration depth of about 12 microns. The ability to vary the penetration depth allows for detailed studies of charge-collection mechanisms in a variety of devices. The ability to control the range—and the energy deposited over that range—is not easily achievable in accelerator-based testing. The penetration depth of an energetic particle depends on both the particle energy and its mass. In order to test at two different ranges with particles of the same linear energy transfer, particles with different mass are typically required (linear energy transfer is the amount of energy deposited per unit length by a particle along its path through a material). The Aerospace team that uses the Lawrence Berkeley cyclotron performs tests using a variety of particles with different energies and different masses, which allows characterization of most devices over a wide range of linear energy transfer (see Heavy-Ion Testing for Single-Event Effects).

In the Aerospace laser test facility, the device test fixture is mounted on a computer-controlled, two-dimensional positioning system and raster scanned beneath the laser beam. Positional accuracy is 0.1 micron. The laser beam is focused onto the device with a custom-built microscope. A camera attached to the microscope allows investigators to observe the exact location of the laser beam on the device. Various microscope objectives provide useful magnifications between 100X and 1000X, and the spot size of the incident laser beam can be varied between approximately 1 and 150 microns.

The testing process generally begins by scanning a device with the large-diameter laser spot at low magnification to identify sensitive regions. During this initial scan, both spatial coordinates and images of the sensitive regions are recorded.

Single-event transients from linear integrated circuits can be captured and registered as logical upsets in digital integrated circuits. The amplitude and width of this transient disturbance is proportional to the amount of charge collected by the sensitive device node. The use of picosecond lasers for this type of testing has aided in explaining the transient behavior of individual transistors in more complex integrated circuits. The three transients depicted here show how a single transistor in a high-speed operational amplifier responds to different amounts of energy deposited by the laser. The greater the amount of deposited energy, the larger the peak amplitude and width of the disturbance. For simplification, the energy deposited by the laser pulse has been normalized to the smallest amplitude transient. (view larger image).

Once the large-spot scan has been completed, a tightly focused laser spot at higher magnification is used to pinpoint sensitive nodes within the regions identified during the large-spot scan. The threshold for single-event effects can be determined by reducing the incident pulse energy until single-event effects are no longer observed. A fraction of the optical signal is sampled by a photodiode and monitored on an oscilloscope for calibrating the laser-pulse energy incident on the device. Thorough calibration of the system includes measurements of the reflectance from the semiconductor surface at sensitive locations.

Aerospace Activities

Early work at Aerospace focused on establishing a relationship between single-event effects induced by the pulsed laser and by energetic particles. For these measurements, basic four-terminal latchup test structures were chosen. These structures are routinely used for latchup research and are the simplest that can be used to study this phenomenon in complementary metal-oxide semiconductor (CMOS) devices.

Results from these measurements showed that it was possible to correlate the thresholds for heavy-ion-induced latchup and laser-induced latchup in CMOS devices from a number of different vendors. Additional studies were performed to validate the effectiveness of various techniques to produce devices that were "hardened by design" (see Designing Integrated Circuits to Withstand Space Radiation).

More recently, Aerospace has been investigating single-event upsets, single-event latchup, and single-event transients in various analog, digital, and mixed-signal devices.

Transient Testing

Single-event transients appear as brief current spikes that can lead to anomalies in other components, such as logic circuits, downstream from the affected component. They can also propagate through logic gates in digital integrated circuits and be captured as upsets by clocked logic.

The commercial demand for high-speed, low-power devices is driving down the minimum feature sizes in microelectronics. As a result, single-event transients are causing greater concern for space-systems engineers. Reduced feature sizes and operating voltages mean that less charge is required to generate upsets, and also mean that modern devices will be fast enough to respond to single-event transients that were too short to propagate through older, slower logic.

By synchronizing the clock frequency of a device with the laser pulse, the temporal dependence of single-event upset on clock cycle can be investigated using a pulsed laser. The upper traces in this figure show the laser pulse arriving just prior to the falling edge of the clock signal and no upset observed at the device output; however, whenever the laser pulse arrives slightly after the falling edge of the clock waveform, the next clock cycle is phase shifted, or delayed in time by one-half of the clock period (view larger image).

Aerospace first reported the use of a picosecond laser as a diagnostic tool for understanding the origins of single-event transients in analog devices in 1993. Operational amplifiers, known to experience single-event transients on orbit, were first tested with heavy ions at a cyclotron and then subjected to laser testing to identify the approximate areas of sensitive transistors. The results showed that the laser could be used to identify sensitive transistors and to reproduce the transient behavior observed during energetic particle tests.

Since then, pulsed lasers have been used on numerous occasions to complement a limited set of particle-beam data and to expand the knowledge of how device sensitivity varies under different operating conditions. To date, laser-based testing has been used to examine single-event transients in a variety of analog devices commonly found in space systems, including operational amplifiers, comparators, and mixed-signal components.

Latchup Testing of Commercial Parts

Recently, Aerospace collaborated with researchers from NASA's Jet Propulsion Laboratory (JPL) to identify the mechanisms responsible for destructive failures observed in an analog-to-digital converter induced by heavy ions during latchup testing. A substantial number of these devices suffered catastrophic failures during these tests, but the complexity of the devices made it difficult to identify the failure mode.

By using the pulsed laser, Aerospace was able to pinpoint the sensitive nodes and view, in real time, the destructive failure mode. This allowed the researchers from JPL to determine that the current density from latchup was so great in these converters that the aluminum metal lines were actually melting and ejecting molten aluminum from beneath the metal encapsulant layer.

Once the location of this failure mechanism had been identified with the pulsed laser, the devices were reexamined using heavy-ion irradiation, and the same failure mode was obvious. The laser tests also provided direct evidence for nondestructive, latent damage to metal lines and vias subject to such high-current densities as a result of latchup. These were the first experiments in which destructive failures and latent damage were observed and recorded in real time.

Before and after images showing destructive failure of metal lines in an analog-to-digital converter as a result of latchup. The highlighted regions (in the photo on the right) are areas where molten metal was ejected from the metal line.

Aerospace has also tested a number of complex microprocessors and digital signal processors. In the case of the Motorola 68302 microprocessor, for example, heavy-ion testing revealed a number of different single-event upset signatures and indicated that the device was fairly sensitive to energetic-particle-induced latchup; however, observations had not shown this microprocessor to be prone to latchup on orbit. The laser was used to probe the different parts of the microprocessor responsible for these types of effects and pinpoint the nodes that were sensitive to latchup.

Agreement between the laser-based test and the heavy-ion test led investigators to look for an alternative explanation for the apparent absence of latchup events on orbit. They noted that the telemetry data from the satellite allows checking of the device current only 0.0002 percent of the time. They therefore concluded that the part probably is experiencing latchup on-orbit, but the high-current state is not detectable because the limited duty factor of the sampling telemetry makes it highly unlikely that a high-current event will be detected before the system is reset (effectively correcting the latchup condition).

In another instance, Aerospace assessed the single-event latchup vulnerability of a 24-bit digital signal processor for the Milsatcom program office. A highly detailed map of latchup-sensitive locations on this device was generated, and more than 3700 individual nodes were identified as being susceptible to laser-induced latchup. Some of these sites were susceptible at low linear energy transfer values, which indicated that the part would probably experience latchup on orbit. As a result, researchers concluded that this part would not be an acceptable candidate for the mission under consideration.

Validating Hardened Designs

Heavy-ion tests on memory cells designed to be resistant to single-event effects generated both single-event upset and single-event latchup. In accelerator-based testing, little information could be extracted from these tests because the latchup threshold was only slightly higher than the upset threshold. Irradiation of the entire device produced latchup and upsets randomly; however, in many instances, the device experienced upset but was then driven into latchup before information about the upset could be retrieved. Consequently, for devices such as these, it was not possible to cleanly extract information about the upset threshold and cross section using the standard, heavy-ion test procedures.

Because the laser beam can be focused onto a single node, Aerospace used the pulsed laser to identify the locations responsible for latchup and then conducted a detailed analysis of the device layout to identify the root cause. Laser testing was also used to identify the nodes responsible for the upset, without any interference from the latchup problem. Electrical simulations of the circuit then helped reveal an unexpected dual-node upset mechanism. The upset was a result of simultaneous charge collection at two sensitive locations. Understanding the mechanisms responsible for the high sensitivity to single-event effects allowed for circuit design changes that improved the memory cells' resistance to single-event effects.

Similarly, heavy-ion single-event testing of an application-specific integrated circuit identified a susceptibility both to single-event latchup and single-event upset. The linear energy transfer threshold for inducing latchup was low enough to prompt the use of the pulsed laser to identify areas on the chip that were responsible for these events. The results from these measurements were provided to the contractor, and the circuit was redesigned with appropriate modifications. Subsequent testing showed no evidence of latchup.

A radiation-hardened version of a 32-bit digital signal processor was also tested for laser-induced latchup and compared with the corresponding commercial version. During heavy-ion testing, the hardened devices exhibited no latchup for effective linear energy transfer values as high as 120 MeV-cm²/mg.

The commercial version, on the other hand, exhibited latchup during heavy-ion testing at an effective linear energy transfer value of only 12 MeV-cm²/mg. In fact, laser testing allowed the identification of more than 60 single-event latchup locations on this device. The same locations on the hardened version were then interrogated with the laser, but no latchup was observed. This result provided confidence in the radiation-hardened design and further confirmed the effectiveness of the laser for latchup screening of hardened devices.

Continuing Investigations

Aerospace is involved in collaborative research efforts to study novel approaches for hardening commercially available integrated circuits against single-event latchup. Additional efforts seek to gauge the space suitability of commercially available devices that take advantage of advanced manufacturing processes. The picosecond-laser facility is also being used to study the effectiveness of various design strategies for mitigating the effects of single-event transients in digital integrated circuits.

The use of a large-diameter laser beam allows for rapid identification of regions in complex integrated circuits that are susceptible to single-event effects. A tightly focused laser beam, on the order of a micron in diameter, can then be used to precisely pinpoint sensitive junctions at the transistor level. This technique provides valuable information to the designer who is interested in understanding and mitigating single-event effects in spaceborne microelectronics.

High-speed integrated circuits are transitioning from silicon-based semiconductors to compound semiconductors, such as gallium arsenide, indium phosphide, and silicon germanium. Aerospace investigation of these devices will include the picosecond laser system to help characterize their sensitivity to single-event effects.

While laser-induced single-event effect testing will not replace conventional particle-beam testing, it has become a well-established technique for providing a better understanding of the nature of single-event effects in complex modern microelectronic devices and for validating design-hardening methods to mitigate single-event effects in these devices.