Failure of electronic components

Vias are a common source of unwanted serial resistance on chips; defective vias show unacceptably high resistance and therefore increase propagation delays. As their resistivity drops with increasing temperature, degradation of the maximum operating frequency of the chip the other way is an indicator of such a fault. Mousebites are regions where metallization has a decreased width; such defects usually do not show during electrical testing but present a major reliability risk. Increased current density in the mousebite can aggravate electromigration problems; a large degree of voiding is needed to create a temperature-sensitive propagation delay.[9]

Sometimes, circuit tolerances can make erratic behaviour difficult to trace; for example, a weak driver transistor, a higher series resistance and the capacitance of the gate of the subsequent transistor may be within tolerance but can significantly increase signal propagation delay. These can manifest only at specific environmental conditions, high clock speeds, low power supply voltages, and sometimes specific circuit signal states; significant variations can occur on a single die.[9] Overstress-induced damage like ohmic shunts or a reduced transistor output current can increase such delays, leading to erratic behavior. As propagation delays depend heavily on supply voltage, tolerance-bound fluctuations of the latter can trigger such behavior.

Gallium arsenide monolithic microwave integrated circuits can have these failures:[11]

• Degradation of I_DSS[12] by gate sinking and hydrogen poisoning. This failure is the most common and easiest to detect, and is affected by reduction of the active channel of the transistor in gate sinking and depletion of the donor density in the active channel for hydrogen poisoning.
• Degradation in gate leakage current. This occurs at accelerated life tests or high temperatures and is suspected to be caused by surface-state effects.
• Degradation in pinch-off voltage. This is a common failure mode for gallium arsenide devices operating at high temperature, and primarily stems from semiconductor-metal interactions and degradation of gate metal structures, with hydrogen being another reason. It can be hindered by a suitable barrier metal between the contacts and gallium arsenide.
• Increase in drain-to-source resistance. It is observed in high-temperature devices, and is caused by metal-semiconductor interactions, gate sinking and ohmic contact degradation.

Micro-photograph of a failed TO3 power transistor due to short circuit

Metallisation failures are more common and serious causes of FET transistor degradation than material processes; amorphous materials have no grain boundaries, hindering interdiffusion and corrosion.[13] Examples of such failures include:

• Electromigration moving atoms out of active regions, causing dislocations and point defects acting as nonradiative recombination centers producing heat. This may occur with aluminium gates in MESFETs with RF signals, causing erratic drain current; electromigration in this case is called gate sinking. This issue does not occur with gold gates.[13] With structures having aluminium over a refractory metal barrier, electromigration primarily affects aluminium but not the refractory metal, causing the structure's resistance to erratically increase. Displaced aluminium may cause shorts to neighbouring structures; 0.5-4% of copper in the aluminium increases electromigration resistance, the copper accumulating on the alloy grain boundaries and increasing the energy needed to dislodge atoms from them.[14] Other than that, indium tin oxide and silver are subject to electromigration, causing leakage current and (in LEDs) nonradiative recombination along chip edges. In all cases, electromigration can cause changes in dimensions and parameters of the transistor gates and semiconductor junctions.
• Mechanical stresses, high currents, and corrosive environments forming of whiskers and short circuits. These effects can occur both within packaging and on circuit boards.
• Formation of silicon nodules. Aluminium interconnects may be silicon-doped to saturation during deposition to prevent alloy spikes. During thermal cycling, the silicon atoms may migrate and clump together forming nodules that act as voids, increasing local resistance and lowering device lifetime.[2]
• Ohmic contact degradation between metallisation and semiconductor layers. With gallium arsenide, a layer of gold-germanium alloy (sometimes with nickel) is used to achieve low contact resistance; an ohmic contact is formed by diffusion of germanium, forming a thin, highly n-doped region under the metal facilitating the connection, leaving gold deposited over it. Gallium atoms may migrate through this layer and get scavenged by the gold above, creating a defect-rich gallium-depleted zone under the contact; gold and oxygen then migrate oppositely, resulting in increased resistance of the ohmic contact and depletion of effective doping level.[13] Formation of intermetallic compounds also plays a role in this failure mode.

Most stress-related semiconductor failures are electrothermal in nature microscopically; locally increased temperatures can lead to immediate failure by melting or vaporising metallisation layers, melting the semiconductor or by changing structures. Diffusion and electromigration tend to be accelerated by high temperatures, shortening the lifetime of the device; damage to junctions not leading to immediate failure may manifest as altered current–voltage characteristics of the junctions. Electrical overstress failures can be classified as thermally-induced, electromigration-related and electric field-related failures; examples of such failures include:

• Thermal runaway, where clusters in the substrate cause localised loss of thermal conductivity, leading to damage producing more heat; the most common causes are voids caused by incomplete soldering, electromigration effects and Kirkendall voiding. Clustered distribution of current density over the junction or current filaments lead to current crowding localised hot spots, which may evolve to a thermal runaway.
• Reverse bias. Some semiconductor devices are diode junction-based and are nominally rectifiers; however, the reverse-breakdown mode may be at a very low voltage, with a moderate reverse bias voltage causing immediate degradation and vastly accelerated failure. 5 V is a maximum reverse-bias voltage for typical LEDs, with some types having lower figures.
• Severely overloaded Zener diodes in reverse bias shorting. A sufficiently high voltage causes avalanche breakdown of the Zener junction; that and a large current being passed through the diode causes extreme localised heating, melting the junction and metallisation and forming a silicon-aluminium alloy that shorts the terminals. This is sometimes intentionally used as a method of hardwiring connections via fuses.[14]
• Latchups (when the device is subjected to an over- or undervoltage pulse); a parasitic structure acting as a triggered SCR then may cause an overcurrent-based failure. In ICs, latchups are classified as internal (like transmission line reflections and ground bounces) or external (like signals introduced via I/O pins and cosmic rays); external latchups can be triggered by an electrostatic discharge while internal latchups cannot. Latchups can be triggered by charge carriers injected into chip substrate or another latchup; the JEDEC78 standard tests susceptibility to latchups.[9]

Electrostatic discharge (ESD) is a subclass of electrical overstress and may cause immediate device failure, permanent parameter shifts and latent damage causing increased degradation rate. It has at least one of three components, localized heat generation, high current density and high electric field gradient; prolonged presence of currents of several amperes transfer energy to the device structure to cause damage. ESD in real circuits causes a damped wave with rapidly alternating polarity, the junctions stressed in the same manner; it has four basic mechanisms:[15]

• Oxide breakdown occurring at field strengths above 6–10 MV/cm.
• Junction damage manifesting as reverse-bias leakage increases to the point of shorting.
• Metallisation and polysilicon burnout, where damage is limited to metal and polysilicon interconnects, thin film resistors and diffused resistors.
• Charge injection, where hot carriers generated by avalanche breakdown are injected into the oxide layer.

Catastrophic ESD failure modes include:

• Junction burnout, where a conductive path forms through the junction and shorts it
• Metallisation burnout, where melting or vaporizing of a part of the metal interconnect interrupts it
• Oxide punch-through, formation of a conductive path through the insulating layer between two conductors or semiconductors; the gate oxides are thinnest and therefore most sensitive. The damaged transistor shows a low-ohmic junction between gate and drain terminals.

A parametric failure only shifts the device parameters and may manifest in stress testing; sometimes, the degree of damage can lower over time. Latent ESD failure modes occur in a delayed fashion and include:

• Insulator damage by weakening of the insulator structures.
• Junction damage by lowering minority carrier lifetimes, increasing forward-bias resistance and increasing reverse-bias leakage.
• Metallisation damage by conductor weakening.

Catastrophic failures require the highest discharge voltages, are the easiest to test for and are rarest to occur. Parametric failures occur at intermediate discharge voltages and occur more often, with latent failures the most common. For each parametric failure, there are 4–10 latent ones.[16] Modern VLSI circuits are more ESD-sensitive, with smaller features, lower capacitance and higher voltage-to-charge ratio. Silicon deposition of the conductive layers makes them more conductive, reducing the ballast resistance that has a protective role.

The gate oxide of some MOSFETs can be damaged by 50 volts of potential, the gate isolated from the junction and potential accumulating on it causing extreme stress on the thin dielectric layer; stressed oxide can shatter and fail immediately. The gate oxide itself does not fail immediately but can be accelerated by stress induced leakage current, the oxide damage leading to a delayed failure after prolonged operation hours; on-chip capacitors using oxide or nitride dielectrics are also vulnerable. Smaller structures are more vulnerable because of their lower capacitance, meaning the same amount of charge carriers charges the capacitor to a higher voltage. All thin layers of dielectrics are vulnerable; hence, chips made by processes employing thicker oxide layers are less vulnerable.[14]

Current-induced failures are more common in bipolar junction devices, where Schottky and PN junctions are predominant. The high power of the discharge, above 5 kilowatts for less than a microsecond, can melt and vaporise materials. Thin-film resistors may have their value altered by a discharge path forming across them, or having part of the thin film vaporized; this can be problematic in precision applications where such values are critical.[17]

Newer CMOS output buffers using lightly doped silicide drains are more ESD sensitive; the N-channel driver usually suffers damage in the oxide layer or n+/p well junction. This is caused by current crowding during the snapback of the parasitic NPN transistor.[18] In P/NMOS totem-pole structures, the NMOS transistor is almost always the one damaged.[19] The structure of the junction influences its ESD sensitivity; corners and defects can lead to current crowding, reducing the damage threshold. Forward-biased junctions are less sensitive than reverse-biased ones because the Joule heat of forward-biased junctions is dissipated through a thicker layer of the material, as compared to the narrow depletion region in reverse-biased junction.[20]

A resistor removed from a high voltage tube circuit shows damage from voltaic arcing on the resistive metal oxide layer.

Resistors can fail open or short, alongside their value changing under environmental conditions and outside performance limits. Examples of resistor failures include:

• Manufacturing defects causing intermittent problems. For example, improperly crimped caps on carbon or metal resistors can loosen and lose contact, and the resistor-to-cap resistance can change the values of the resistor[2]
• Surface-mount resistors delaminating where dissimilar materials join, like between the ceramic substrate and the resistive layer.[21]
• Nichrome thin-film resistors in integrated circuits attacked by phosphorus from the passivation glass, corroding them and increasing their resistance.[22]
• SMD resistors with silver metallization of contacts suffering open-circuit failure in a sulfur-rich environment, due to buildup of silver sulfide.[6]
• Copper dendrites growing from Copper(II) oxide present in some materials (like the layer facilitating adhesion of metallization to a ceramic substrate) and bridging the trimming kerf slot.[3]

Potentiometers and trimmers are three-terminal electromechanical parts, containing a resistive path with an adjustable wiper contact. Along with the failure modes for normal resistors, mechanical wear on the wiper and the resistive layer, corrosion, surface contamination, and mechanical deformations may lead to intermittent path-wiper resistance changes, which are a problem with audio amplifiers. Many types are not perfectly sealed, with contaminants and moisture entering the part; an especially common contaminant is the solder flux. Mechanical deformations (like an impaired wiper-path contact) can occur by housing warpage during soldering or mechanical stress during mounting. Excess stress on leads can cause substrate cracking and open failure when the crack penetrates the resistive path.[2]

Capacitors are characterized by their capacitance, parasitic resistance in series and parallel, breakdown voltage and dissipation factor; both parasitic parameters are often frequency- and voltage-dependent. Structurally, capacitors consist of electrodes separated by a dielectric, connecting leads, and housing; deterioration of any of these may cause parameter shifts or failure. Shorted failures and leakage due to increase of parallel parasitic resistance are the most common failure modes of capacitors, followed by open failures. Some examples of capacitor failures include:

• Dielectric breakdown due to overvoltage or aging of the dielectric, occurring when breakdown voltage falls below operating voltage. Some types of capacitors "self-heal", as internal arcing vaporizes parts of the electrodes around the failed spot. Others form a conductive pathway through the dielectric, leading to shorting or partial loss of dielectric resistance.[2]
• Electrode materials migrating across the dielectric, forming conductive paths.[2]
• Leads separated from the capacitor by rough handling during storage, assembly or operation, leading to an open failure. The failure can occur invisibly inside the packaging and is measurable.[2]
• Increase of dissipation factor due to contamination of capacitor materials, particularly from flux and solvent residues.[2]

In addition to the problems listed above, electrolytic capacitors suffer from these failures:

• Aluminium versions having their electrolyte dry out for a gradual leakage, equivalent series resistance and loss of capacitance. Power dissipation by high ripple currents and internal resistances cause an increase of the capacitor's internal temperature beyond specifications, accelerating the deterioration rate; such capacitors usually fail short.[2]
• Electrolyte contamination (like from moisture) corroding the electrodes, leading to capacitance loss and shorts.[2]
• Electrolytes evolving a gas, increasing pressure inside the capacitor housing and sometimes causing an explosion; an example is the capacitor plague.
• Tantalum versions being electrically overstressed, permanently degrading the dielectric and sometimes causing open or short failure.[2] Sites that have failed this way are usually visible as a discolored dielectric or as a locally melted anode.[6]

Metal oxide varistors typically have lower resistance as they heat up; if connected directly across a power bus, for protection against electrical transients, a varistor with a lowered trigger voltage can slide into catastrophic thermal runaway and sometimes a small explosion or fire.[23] To prevent this, the fault current is typically limited by a thermal fuse, circuit breaker, or other current limiting device.