Today, modern vehicles use more than 100,000 lines of digital code, a number that is expected to grow sixfold by 2025, and within 10 to 12 years, in-vehicle electronics are expected to account for half the value of electric and autonomous vehicles. These figures are thought-provoking considering that more than 30% of field failures today are caused by the electronics of the car.
Not only can car recalls damage a company’s reputation, it’s also costly, and manufacturers can’t bear the impact of an increased number of recalls related to electronics. A known risk of automotive electronics is latent defects, that is, failures that do not occur during testing in semiconductor fabs or during subsequent burn-in testing of component packaging. Over time, these defects can develop, causing failures that can lead to safety hazards and costly recalls.
Electronic device defects are already a costly problem, but a number of factors could make these defects a more significant problem in the future. The value of automotive electronics is on the rise, thanks in large part to digital control and comfort systems, advanced driver assistance systems (ADAS), and the continued development of fully autonomous vehicles. As the cost of vehicles increases, drivers and passengers increasingly rely on embedded systems to guide and control their travel, and buyers’ expectations for “excellent performance” also increase. As electronics are increasingly integrated into safety systems and self-driving cars emerge, failures in electronic systems can be potentially life-threatening. This will intensify the need for recalls when problems arise. Additionally, as vehicle usage patterns increase, costly potential defects will become more likely to manifest. According to Electroiq2, a typical car today contains between 5,000 and 8,000 microchips (chips). If a manufacturer produces 25,000 cars per day with a chip failure rate of one part per million (PPM), then 125 of the cars produced per day
The vehicle has potential chip quality issues. All things being equal, this number will grow exponentially as more electronics are installed in cars.
To meet these challenges, chipmakers will need to find ways to identify or prevent unreliable chips from being made at a lower cost, and to prevent unreliable chips from being installed in electronic assemblies and subsequently into automobiles. Take it out of the supply chain.
The cost of a faulty chip in a semiconductor fab is relatively small (1X). If the chip went through the packaging process and then was identified in the burn-in process, the cost of failure has risen to 10 times the cost of fab prevention. If this faulty chip made it into a fab and was installed in a car before the problem occurred, the cost would increase to an astronomical 1,000 times the cost of eliminating or preventing the problem in the fab, which Potential threats to safety or life have not been considered.
Problematic particles found
Clearly, the fab is the most logical and cost-effective place to prevent latent defects. The problem is that it is difficult to identify latent defects with techniques commonly used by most modern semiconductor fabs, which have traditionally focused on maximizing yield by allowing production processes to eliminate more easily identifiable “fatal” defects.These defects are usually caused by larger particles, which are
Particles can bridge adjacent lines or become embedded in layers such as gate oxide, causing vertical leakage. So-called “large particles” depend on the circuit spacing or pitch (the proximity of two adjacent circuits that carry electronic signals). Chip designs vary, so a particle that causes a “fatal” defect might be 90nm in one fab and 5nm in another. The metrology systems used in the fab were chosen with economics in mind, primarily to detect particles of a size that could pose a risk to the typical circuit spacing of the fab, and filters were specifically chosen to remove them from entering the manufacturing process deadly particles in fluids.
Because of their larger size, these lethal particles are often easier to spot using inline metrology and filtered to remove in the fab. Medium-sized particles are not easily found by traditional metering systems or removed by traditional filters, and can also cause problems. They can be found during the burn-in process of the device packaging step, but cannot be removed. As a result, equipment that is found to be faulty during aging can only be discarded, resulting in a loss of saleable merchandise. The equipment and methods for finding defects caused by these medium and large particles are well known and proven, and have become an integral part of the sustainable production of any electronic device.
Small power, big problem
Various tests have begun to map the relationship between potential defects and contaminants such as particles, gels, metal ions and organics. These are contaminants that remain after standard precautions and burn-in usability testing in the fab. Although contaminants that cause shorts, open circuits, or any electrolyte leakage are detected, small and medium-sized contaminants can still become embedded in the corresponding layers and cause problems over time.
The industry has studied the causes of reliability failures for decades. The reasons include electromigration, oxide breakdown, hot carrier injection (HCI), stress-induced cracking and negative bias temperature instability (NBTI) effects. In addition, more mechanisms related to diffusion, corrosion and plasticity were discovered. As reliability targets become more stringent, more mechanisms will be required to control potential defects related to particulate and metal contamination, thereby bringing reliability to new levels.
Effect of particle size on gate oxide
Figure 5 summarizes the potential impact of small, medium and large particles on gate oxide integrity at three problem levels. The largest particles can disrupt feature patterns on a chip or interfere with the layering of dissimilar materials and cause “deadly” defects. The removal of large particles can immediately improve yield and can be easily detected by metering systems and removed by standard semiconductor liquid filters and purifiers. The cost of stopping or controlling large particles is not high. However, once the wafers are produced in a semiconductor fab, these particles and any associated defects are permanently embedded in them beyond repair.
Improper filtration or metering methods can miss medium-sized particles that may or may not cause failure in burn-in testing.Since these particles do not completely destroy the feature pattern on the chip or interfere with the layering of different materials, they do not
Immediate failure of the device can result. Over time, this can cause the installed components to eventually fail. To prevent latent defects, these wafer defects should be discovered and handled in the semiconductor fab. These faults would be more difficult and more expensive to spot based on existing metrology techniques, but finding problems at this stage of manufacturing could remove substandard chips from the supply chain. These deficiencies can be prevented by strengthening filtration and purification operations.
The next challenge is small particles that may not be removed by fab filters or detected by metrology systems. Since they only partially disrupt the feature pattern on the chip, or partially interfere with the delamination of different materials, they do not cause immediate failure of the device, nor are they found in burn-in tests during chip and module manufacturing. The deterioration they can cause occurs more slowly, leading to potential failures that may not occur until months or years after the chip has passed all parametric validation, burn-in and functional testing and is in service.
Note from Figure 5 that the particle density increases as the particle gets smaller. Using nature as an analogy, the distribution of particles in chemicals is similar to geological conditions. There are far more grains of sand on Earth than boulders. In the same graph, defect density also increases as the particles get smaller. However, as the particles continue to get smaller, the defect density will decrease and freeze at a certain point. At this point, the particles are already small enough that potential defects are no longer possible and therefore do not need to be the focus of removal efforts. As with other particle sizes, the description of a “small” particle size will vary due to the tolerances of each circuit design.
Meeting the Challenges of the Billion Rates
The industry has long been aware of these issues. Manufacturers of high-end chips for AI (artificial intelligence), HPC (high-performance computing), cryptocurrency, 5G, and other storage- and processing-intensive applications are striving to achieve near-zero defects. However, automotive chips have traditionally been characterized by large circuit widths and stringent quality standards required in power applications, microcontrollers and low-complexity sensors to withstand harsh conditions over a life expectancy of 10-15 years. Tested for temperature, humidity and vibration conditions. As a result, contamination control typically focuses on removing larger particles so as not to create yield-threatening fatal defects. But as we give cars more automated control over our mobility needs and overall safety, automakers are increasingly aware that there can be a relationship between pollution and potential defects in individual failures and costly recalls. As the industry looks to reduce equipment failure rates from ppm to ppb levels, semiconductor manufacturers will have to further demonstrate the ability to meet these requirements. Demonstrable chip reliability will soon become a key competitive advantage as new device designs are introduced in automotive applications, creating more opportunities for organizations that can meet quality, cost, performance and reliability standards .
The contamination control solutions being evaluated will use detection methods using modern metrology tools and defect detection techniques, combined with preventive strategies using filtration and decontamination techniques. Each fab and process is unique, so there are different solutions to meet the different needs and constraints encountered by each fab and process to remove defect-causing contaminants. Provides the most comprehensive and predictable results by profiling contaminants in semiconductor manufacturing processes and implementing removal strategies. Depending on the metrology technique used, there is a limit to the particle size that can be detected. Particles smaller than the detection limit size can still pose a threat to circuit reliability without reducing wafer yield. As cars become more connected and electronics make up a larger share of a car’s value, automotive chipmakers need to explore ways to achieve higher reliability performance to address these small particles And the problems caused by impurities and the resulting potential defects are critical. These efforts can be verified through accelerated life cycle testing in the lab to understand the return on investment without years of field testing.
As cars become more integrated with advanced driver assistance systems4 (ADAS) and other digital systems, cars will contain a greater variety and number of chips. These digital systems will include traditional sensors, power devices, microcontrollers and storage,
And ADAS and other systems will bring our cell phones and other “high performance computing” processing and storage technologies to the car, creating one of the most complex digital systems. These cars are highly sensitive to potential defects that manifest after the chip is installed in the car. Whether it involves replacing faulty components or recalls and life safety, correcting these defects can be very costly. Foundry inspections and subsequent burn-in tests, designed to find chips that fail on the spot, have a hard time finding the small particles and metal contaminants that lead to potential defects. The problem for chipmakers is that increasing chip reliability through tougher testing could spark a philosophical debate about wafer yield trade-offs. One way to improve reliability without reducing yield is to remove these small particles and metal contaminants through more thorough filtration and purification before they enter the chip production process. As the relationship between potential defects and contaminants is further identified, this method of improving reliability could become a powerful competitive advantage and offer a very high return on investment.
Entegris is a leading supplier of specialty materials for the microelectronics industry and other high-tech industries. Entegris is ISO 9001 certified and has manufacturing, customer service and/or research facilities in the US, China, France, Germany, Israel, Japan, Malaysia, Singapore, South Korea and Taiwan.
1 A car is made of code: NXP website. https://blog.nxp.com/automotive/cars-are-made-of-code
2 Process Watch: Semiconductor Issues in Automotive. Electroiq website. http://electroiq.com/blog/2018/01/ process-watch-the-automotive-problem-with-semiconductors/
3 Price tags for automotive electronics: what really works. EDN website. https://www.edn.com/electronics-blogs/engineering-on-wheels/4458881/The-price-tag-of-automotive-electronics–What-s-really-at-play
4 How to Make Self-Driving Cars Reliable: Semiconductor Engineering Website. https://semiengineering.com/will-autonomous-vehicles-be-reliable/
The Links: A2475 7MBP100NA060-01