The Effects of Ionizing Radiation on Space Electronics

What Could Go Wrong? The Effects of Ionizing Radiation on Space Electronics

John Scarpulla and Allyson Yarbrough

Space radiation comes in many forms and affects electronic components in diverse ways. Aerospace investigations of how energetic particles interact with integrated circuits and other electronics have been helping spacecraft designers and mission planners minimize the risk of component failure or performance degradation.

The harsh space environment can wreak havoc on unprotected electronics. Over time, exposure to energetic particles can degrade device performance, ultimately leading to component failure. Heavy ions, neutrons, and protons can scatter the atoms in a semiconductor lattice, introducing noise and error sources. Cosmic rays speeding through space can strike microcircuits at sensitive locations, causing immediate upsets known as single-event effects. Passive electronic components and even straightforward wiring and cabling can be seriously affected by radiation. Aerospace has been investigating the means by which heavy ions, protons, and electrons interact with microelectronics. This effort has helped spacecraft designers find ways to prevent serious anomalies on orbit (see sidebar, Space Radiation: Categories).

A typical integrated circuit contains various elements such as capacitors, resistors, and transistors embedded in a silicon substrate and connected by metallic vias (holes that allow electrical connections between front-side metal and the back-side ground plane or between planes in a multilayer structure). These elements are separated by dielectrics and covered by protective layers of passivating insulators and glass. Problems arise when space radiation subverts the normal function of these components or bridges the isolation between them.

Partial cross section of a typical silicon CMOS integrated circuit. Only one metallization layer is shown for simplicity (an actual circuit may have many more). In this figure, eight different p-n junctions (positive-negative, denoting the polarity) are visible; in a real integrated circuit there may be thousands or even millions. The cross section shows a bipolar transistor implanted in an n-well, a PMOS transistor in the same n-well, an NMOS transistor in a p-well, a polysilicon-oxide-polysilicon capacitor, and a polysilicon resistor. Radiation effects differ in each device even though they may be located in close proximity. Short circuit paths that cause latchup are also shown.

Various types of semiconductors are used in microelectronics. For example, the negative metal-oxide semiconductor (NMOS) transistor operation is based on the flow of negatively charged electrons. The positive metal-oxide semiconductor (PMOS) transistor operates based on the flow of positive charges, carried by so-called "holes" (a "hole" is the absence of an electron, or a missing bond that can hop from atom to atom like a positive charge carrier). The complementary metal-oxide semiconductor (CMOS) employs both of these on the same chip. CMOS technology is commonly found in digital circuits such as microprocessors and memories, analog circuits such as operational amplifiers and phase-locked loops, and mixed-signal devices such as analog-to-digital converters. All of these components are generally found aboard a spacecraft.

Total Dose Effects

Total dose refers to the integrated radiation dose that is accrued by satellite electronics over a certain period of time, say 1 year, or over a 15-year satellite mission (see sidebar, Radiation Dosimetry). The radiation has the capability to damage materials by virtue of its ability to ionize material. The energetic ions then can cause damage to materials by breaking and/or rearranging atomic bonds. In general, after exposure to sufficient total-dose radiation, most insulating materials such as capacitor dielectrics, circuit-board materials, and cabling insulators become less insulating or become more electrically leaky. Similarly, certain conductive materials, such as metal-film resistors, can change their characteristics under exposure to total-dose radiation. The metal conductors themselves and magnetic materials tend to be quite radiation hard or resistant to radiation effects. Semiconductor devices in particular exhibit a number of interesting effects. It is important to choose materials and components for satellite electronics that have the necessary radiation tolerance for the required mission. It is also necessary to design in margins or allowances for the expected component changes induced by the radiation environment.

Perhaps the most ubiquitous component in modern microelectronics is the MOS transistor. Coincidentally, it also can be particularly sensitive to radiation. The MOS transistor is an active component that controls the flow of current between its source and drain electrodes. Commonly used as a switch in digital circuits, it may be open or closed depending on whether a voltage is supplied to its control gate electrode. For example, when sufficient voltage is applied to the gate of an NMOS transistor, it allows current to flow; when the voltage remains below the critical threshold, the gate does not permit current to flow. The threshold voltage depends upon the device design and the materials used, but is usually 0.5 to 1.5 volts. The gate oxide, which isolates the gate from the source and drain, is an ideal insulator made of silicon dioxide.

Cross section of an NMOS transistor showing the gate oxide and conducting channel formed between the source and drain. The trapped charges shown in the inset are responsible for the threshold voltage shift, ultimately leading to failure.

Problems arise when this device is exposed to radiation. First, the gate oxide becomes ionized by the dose it absorbs. The free electrons and holes drift under the influence of the electric field that is induced in the oxide by the gate voltage. These holes and electrons would be fairly benign if they were to simply drift out of the oxide and disappear, but although the electrons are fairly mobile, the holes are not, and a small fraction of them become trapped in the gate oxide. After sufficient radiation dose, a large positive charge builds up, having the same effect as if a positive voltage were applied to the gate. With enough total dose, the device turns on even if no control voltage is applied. The transistor source-drain current can no longer be controlled by the gate, and remains on permanently.

The PMOS transistor exhibits a similar, but opposite, effect. When no voltage is supplied, the gate allows current to flow; when the voltage crosses a critical threshold, the gate prevents current from flowing. Therefore, when radiation traps enough positive charge in the gate oxide, the transistor remains off permanently. In a CMOS logic gate consisting of NMOS and PMOS transistors, the output will be frozen at either a "1" or a "0" after a sufficient dose is accumulated, and the device will cease to function.

Some integrated circuit manufacturers have tried to produce transistors with gate oxides that are "hard"—that is, they do not trap positive charges upon radiation exposure. These products can tolerate total-dose levels as high as 1 megarad without difficulties, making their use possible in satellite systems for many years. On the other hand, many commercial products lacking a hardened gate oxide (such as the processors used in desktop computers) might last a few days or weeks in a satellite orbit.

The CMOS integrated circuit market is extremely competitive, with succeeding generations of products offering greater processing power and speed. These gains are achieved by shrinking the transistors so that more can be packed on a single chip. As a consequence, the gate oxides in these shrinking transistors are growing thinner—just a few nanometers thick for the latest generation. Being thinner, the gate oxide traps less positive charge overall. Therefore, CMOS transistors are naturally becoming more radiation resistant. Still, gate oxides are not the only features affected by total ionizing dose.

The transistors in a CMOS device are isolated or separated by so-called field regions. Two different circuits that lie near each other will commonly be separated by a thick field oxide and sealed by an overlying metal conductor. Just like the gate oxide, the field oxide can trap positive charges through extended exposure to ionizing radiation. If enough charge is trapped, a channel of conducting electrons will form in the silicon under the field oxide. This effectively connects the two formerly isolated logic circuits, causing them both to malfunction.

A region of field oxide between two isolated NMOS transistors. When the field oxide traps radiation-induced charges, a conduction channel forms between the two transistors, destroying the isolation.

A similar effect can occur in a single transistor. Trapped charges in the field oxide form a leakage path along the edges parallel to normal conduction flow in an NMOS transistor. The silicon along these edges forms an unwanted conduction path. In modern CMOS devices, edge leakage is frequently the dominant mode limiting the total-dose hardness of the product. After a high total dose, the transistors cumulatively leak so much current that the power supply can no longer handle the load. The power dissipation rises to high levels, and the chip fails. A hardened field oxide is required to help prevent this occurrence.

Neutron or Proton Damage

When highly energetic neutrons or protons penetrate the crystal lattice of a semiconductor, such as silicon, atoms can get displaced through several mechanisms. For example, the incident particle can transfer some of its energy to the silicon nucleus, and if enough energy is transferred (approximately 25 electron volts), the nucleus gets knocked out of position. This is called elastic scattering, and the freed silicon atom can lose energy through ionization or by displacing other atoms. Inelastic scattering can also occur, whereby the struck nucleus absorbs the neutron or proton and then reemits it at a lower energy along with a gamma ray. This process also causes displacements. The displacements are essentially microscopic crystal imperfections that interfere with the orderly flow of charges from the source to the drain.

The resulting crystal lattice contains voids where the silicon atoms were knocked out of position and clusters where they came to rest. These sites, known as traps or recombination centers, respectively, can be a source of problems in some semiconductor devices.

For example, a bipolar-junction transistor functions as a current amplifier. A p-n junction is the place where a p-type material meets an n-type material. There are two types of bipolar-junction transistors—n-p-n and p-n-p—which are created by sandwiching semiconductor of one doping type between two other layers of the opposite type. The principle of operation of bipolar transistors is by charge-carrier diffusion, which is different from the MOS transistor, whose principle of operation is by drift. In an n-p-n bipolar transistor, electrons are emitted by the emitter n-type layer into a middle material known as the base, where they diffuse to the collector n-type layer at the opposite side. If the transistor were perfect, all the electrons that traverse the middle material would be collected. In actuality, some are lost through recombination with holes. The transistor gain is therefore defined as the amount of current that reaches the collector compared with the amount that recombines with the base.

Operation of a bipolar junction transistor. Electrons are emitted at one end, diffuse through the middle material (the base), and are eventually collected. If the transistor were perfect, all the emitted electrons would be collected; however, some are lost through recombination with holes in the base.

When the transistor is exposed to neutrons or protons, displacement damage and new recombination centers are created. This increases the likelihood that electrons will recombine with holes in the base material. Higher neutron or proton fluxes give rise to higher rates of recombination and lower transistor gain. Eventually, the transistor fails because its gain drops too low to provide amplification. This is the dominant failure mode in bipolar integrated circuits.

Bipolar-junction transistors are also sensitive to total ionizing dose. The phenomenon is similar to that observed in MOS transistors, where an unwanted conducting channel is formed adjacent to the surfaces of the field oxide. These channels cause unwanted current that can eventually cause device failure. Similarly, MOS transistors are somewhat sensitive to displacement damage. Some of the charges are scattered by the damage sites, and the transistors exhibit a loss of conductance and an increase in noise. These degradations are themselves capable of causing circuit failures.

Path of a cosmic ray through an NMOS transistor

Path of a cosmic ray through the drain of an NMOS transistor. The charge liberated in the ion's wake is collected by the funneling mechanism and through diffusion away from the junction. A short is momentarily created between the substrate (normally grounded) and the drain terminal (normally connected to a positive power supply voltage).

Single-Event Effects

This category of radiation effects is the only one in which a single particle is the source of the trouble (see Picosecond Lasers for Single-Event Effects Testing and Heavy-Ion Testing for Single-Event Effects). Highly energetic ions such as cosmic rays can easily penetrate the structure of a spacecraft, pass through internal components, and exit the structure in a straight line. Shielding against them is simply not practical. Because the heavy particles are omnidirectional, they impinge on an integrated circuit at random times and locations, with random angles of incidence.

The concept of total ionizing dose is not useful to describe a single particle; instead, a quantity called the linear energy transfer is used. As the particle traverses the material of interest, it deposits energy along its path. Linear energy transfer is the amount of energy deposited per unit of distance traveled, normalized to the material's density. It is usually expressed in MeV-cm²/mg. A typical satellite environment will include a wide variety of particles with various amounts of kinetic energy corresponding to a wide spectrum of linear energy transfer.

An energetic ion passes through a semiconductor device in a few picoseconds. As it does so, it leaves behind a "track" or column of ionized material typically ranging from a few tenths of a micron to a few microns in diameter. The ionized track contains equal numbers of electrons and holes and is therefore electrically neutral. The total number of charges is proportional to the linear energy transfer of the incoming particle. It is as if a conducting wire were suddenly inserted into the semiconductor device, disturbing the electric fields and normal current paths (see table, Calculating Electron-Hole Pairs).

If a cosmic ray passes through the drain region of an NMOS transistor, a short is momentarily created between the substrate (normally grounded) and the drain terminal (normally connected to a positive power supply voltage). When this happens, a spike of current flows for an instant. The amount of charge that is "collected" from the ion track before it dissipates or disappears by recombination is significant: Every device has a certain critical charge, which, if exceeded, results in a single-event upset, burnout, or other undesirable phenomenon.

The process of energetic-ion-induced charge collection is complex and rapid, and is not completely understood. It consists in part of charge "funneling," where distortion of the normal electric field patterns of a device allows more charge to be collected than could normally be transported into the sensitive region. Charges are also pulled away from the drain-substrate junction by diffusion. The ability of a device to collect charge from the ion track determines its sensitivity to cosmic rays.

Latchup occurs when the source of one MOS transistor forges a pathway to the drain of another. A transient radiation pulse can generate the current needed to bridge this gap. Current will then continue to flow unregulated between the two components. The entire circuit must be powered down to break the connection. In some cases, the circuit may be permanently damaged.

If a device is large, it presents a greater target for cosmic rays. It is therefore more likely to receive a "hit" than a smaller device. This relationship is described by an attribute known as the "cross section" of the device, which is calculated as the ratio of the number of single-event upsets to the particle flux over a given surface area. In determining the sensitivity of a device to single-event effects, two important parameters to consider are the threshold linear energy transfer, above which upsets or single events are seen, and the saturation cross section, i.e., the cross section at high values of linear energy transfer.

Researchers have identified various types of single-event effects, varying in their degree of seriousness.

A single-event transient, for example, is a temporary spike or signal caused by a heavy ion. In some cases, this spike can excite analog circuits into temporary or permanent oscillation. In digital circuits, the spike may propagate through many logic gates, causing system malfunction. In mixed-signal devices, a transient generated in the analog part of the device can propagate into the digital part, causing logic-level shifts.

A single-event upset usually manifests itself as a "bit-flip" or change of state in a logic circuit. If enough of these upsets occur, or if a single critical node is affected, a computer can freeze up and must be rebooted. Single-event upsets occur in computer memories, microprocessors, controllers, and almost any digital circuit containing latches or memory elements.

Single-event latchup is triggered when a heavy ion causes current to flow unregulated between components on an integrated circuit. When PMOS and NMOS transistors are integrated into the same area of a silicon substrate, they can form a parasitic or undesired circuit element (called a thyristor) if struck by an energetic ion. A thyristor is an interconnected n-p-n and p-n-p bipolar transistor; the current amplified by the n-p-n transistor supplies the p-n-p transistor, which in turn supplies it back to the n-p-n transistor, creating a feedback loop. Thyristors are perfectly legitimate devices in their own right, and are used for regeneratively switching large currents. But they are, by nature, feedback devices, and can be turned on or "latched" when the initiating current exceeds a threshold value that allows the feedback process to begin. Thus, when an energetic particle traverses the region of a CMOS integrated circuit containing the parasitic n-p-n and p-n-p transistors, it can generate enough current to trigger the thyristor, provided the particle has sufficient linear energy transfer. If this happens, the effected portion of the CMOS integrated circuit will be driven into latchup.

Current is generated in a p-n junction exposed to transient radiation. It can be modeled as a transient current source in parallel with a diode.

As long as the power supply maintains the voltage equal to or greater than the thyristor "holding" voltage, the latchup condition remains. The entire integrated circuit must be powered down to correct the condition. In many cases, the current is sufficient to burn out the transistors or metallization in the latchup path, permanently damaging the circuit (a phenomenon known as single-event burnout). In other cases, latchup does not cause damage, and the device is universally recoverable. The outcome depends on the circuit design, the geometry, and the presence of any current-limiting resistances. This serious problem makes it very difficult to use most commercial integrated circuits in an environment where heavy-particle radiation may be encountered. Bipolar integrated circuits are particularly sensitive to latchup.

Other single-event phenomena are even more complex. For example, in certain MOS transistors, the gate oxide can be ruptured by the passage of a cosmic ray. While not completely understood, this so-called single-event gate rupture may be caused by a combination of charge-multiplicative breakdown and injection of charges into the gate oxide.

Conclusion

Understanding how space radiation interacts with microelectronics is the first step in establishing ways to mitigate adverse effects. Research at Aerospace has been instrumental in revealing the underlying mechanisms that lead to radiation-induced effects. Semiconductor manufacturing processes continue to evolve, and new technologies present new opportunities for complex interactions. Continued research at Aerospace will help spacecraft designers and mission planners account for all possible failure modes.