Semiconductor Failure Analysis - Finding a One in a Billion Defect
Since the demonstration of the first integrated circuit in the late 1950s, semiconductor technology has developed explosively, growing at an exponential rate. The guidance computers that were used in the Apollo space program, performing the critical calculations necessary to land a manned spacecraft on the moon, have been completely dwarfed in complexity, memory capacity, and processing power by modern video game consoles and handheld .MP3 players. Where an early microchip might contain several hundred devices, today's IC is home to billions of transistors. Even though semiconductor technology has come so far from its inception, it is not yet infallible, and failures do occur as a result of improper processing, misuse, or simply due to the inexorable march of time. Finding a defect on such a complex device may bring to mind clichéd sayings about needles and haystacks; however, the process of semiconductor failure analysis brings together a comprehensive toolset, a breadth of industry experience, and a certain degree of intuition, all in order to find that one in a billion defect.
Almost all failure analysis work at the IC level starts with non-destructive testing. This may include an in-depth optical inspection of the device as received, a high-resolution x-ray inspection, or acoustic imaging of the failed device. While these steps may seem somewhat mundane, they often provide an analyst with crucial information that may dictate the entire course of the analysis. For example, an x-ray of a large microcontroller with hundreds of electrical connections for data input and output may reveal a bond wire that has been fused. Not only does this fused bond wire suggest a failure mechanism (excessive current), it also provides the analyst with a general area to inspect once the device has been decapsulated.
Though in some rare cases an analyst may move directly to cross-section or other destructive techniques following non-destructive testing, generally the next step in semiconductor failure analysis is to attempt to isolate the defect as precisely as possible by choosing an appropriate tool or technique based on the type of failure. A part with a short circuit, possibly caused by electrical overstress, may be a good candidate for thermal microscopy; a more subtle defect like a "functional failure" where the only noticeable problem is a very slight change in device characteristics may require the use of light emission microscopy to look for defects like gate oxide pinholes. With these techniques as well as analysis of a properly functioning unit, the failing device can be characterized and the most likely location of the defect can be identified.
Finally, armed with the results from their chosen isolation technique, the analyst can proceed to destructively analyze the failure. The failing device can be deprocessed and inspected layer-by-layer to expose a defect that may be hidden beneath the metal traces on the die, or cross-sectioned to examine construction parameters that may be responsible for the failure. With the problem thus characterized, the analyst can report their findings to the customer, who may choose to incorporate them into a corrective action plan.
Semiconductor failure analysis is an absolutely vital process for companies wanting to be competitive in the modern electronics industry. To get the full benefit, failure analysis should be performed by a lab with the experience, expertise, and skillset to deliver defect-free services, on time, every time.
Derek Snider has been an employee at Insight Analytical Labs since 2004, where he currently works as a Failure Analyst. He is an undergraduate student at the University of Colorado, Colorado Springs, where he is pursuing a Bachelors of Science degree in Electrical Engineering.