V. VIJAY, CyberOptics Corporation
Fine particles (less than 5 micrometers in diameter) can have a disastrous impact on semiconductor manufacturing. From the earliest days, manufacturing facilities have deployed air filtering and recirculation to remove particles from the cleanroom, but particles may still be generated inside process tools, where they can cause defects and yield loss. Quickly identifying when and where airborne particles originate can be challenging, but it is critical to success. Conventional methods for monitoring and diagnosing contamination problems take considerable time to return results. Because of their intermittent nature, they are not good at detecting precisely when within the process cycle contamination appears, or at issuing immediate real-time alerts. As a result, contamination episodes are often not seen until defects are detected by downstream inspections. In-line particle sensing (IPS™) provides continuous, immediate, time-resolved monitoring that can shorten response times, protect work-in-progress, and facilitate root cause analysis.
During the manufacturing process, semiconductor devices are extremely vulnerable to defects caused by particulate contamination. Manufacturers go to great lengths to prevent contamination. The process takes place in a cleanroom, where air is constantly recirculated and filtered to remove particles. Even within the cleanroom, when wafers are transported from tool to tool, they are enclosed in specially designed air-tight containers and never exposed the ambient environment. However, contaminants generated inside process tools can still cause defects and yield loss, and engineers carefully monitor tools to detect contamination. Most monitoring relies on the inspection of monitor wafers, which are run through the process tool routinely, or in response to a contamination problem detected by downstream inspection of product wafers. This approach has shortcomings. It is not continuous and may miss an event that occurs between samples. It takes tool-time to run monitor wafers that could be used to process product wafers. It is slow, taking considerable time to return results and meanwhile putting additional wafers at risk. It is not time resolved, making it difficult to know when in the process cycle the contamination appeared. A new in-line particle sensor addresses these shortcomings. It can be installed in the process chamber exhaust line, where it can provide continuous, time-resolved monitoring of all gas flowing out of the chamber, detecting particles larger than 0.1μm.
EUV lithography is especially vulnerable to particulate contamination for several reasons. It is used to create the smallest structures in the circuit, where even the smallest particle can result in a killer defect. It is a bottleneck process – every layer begins with lithography, when it stops, the whole process stops. Moreover, EUV systems are capital intensive. Earning an adequate return on that substantial investment requires the system to operate constantly at maximum throughput with little or no downtime. For these reasons, maintaining system availability has the highest priority. Any time saved by early detection and rapid response to process excursions or contamination events is extremely valuable. Maximizing yield and tool uptime in photolithography processes requires best-in-class practices to achieve a contamination-free process environment. Unfortunately, identifying precisely when and where airborne particles originate is challenging with traditional methods.
There are three widely used methods for particle detection in photolithography systems: bench-top & hand-held airborne particle counters, monitor reticles (analogous to monitor wafers), and in-situ particle scanners. All three of these traditional methods have substantial drawbacks:
- Bench-top counters require long hoses to reach into the tool—and are often incapable of following the reticle path. Similarly, bench-top and handheld methods make it difficult (often impossible) to reach all locations of interest. In some cases, equipment engineers must crawl through the scanner to make accurate measurements, adding an additional contamination risk from the measurement itself.
- Monitor reticle scanning is time-consuming, creating long delays for test results. The process has many steps and typically involves considerable waiting for tool availability and lost tool time while the tool executes non-productive test cycles: 1) premeasure the test reticle, 2) load the test reticle into the tool, 3) put the test reticle through a normal operational cycle, 4) Remove the test reticle and wait for inspection tool availability, 5) perform the post measurement, 6) analyze results. If problems are found, repeat the measurement process. Monitor reticle scanning does not deliver time resolved results and cannot determine precisely when and where the reticle became contaminated.
- In-situ scanners are usually separate from the tool and look at particles deposited on a wafer not in the air. They are good at identifying particles but limited in their ability to pinpoint when and where in the wafer handling process the particles accumulated on the wafer. Like monitor reticles, they do not provide time-resolved data or timely alerts, which could prompt proactive intervention to avoid contamination of product wafers.
These issues have driven the development of a sensor that can detect airborne particles in vacuum and gas flow lines – the In-line Particle Sensor™ (IPS™, CyberOptics Corporation) shown in FIGURE 1. IPS mounts permanently on a supply or exhaust line and detects particles 24/7. It can detect particles of 0.1micron size and higher. Its 10-6 Torr operating range allows installation of the IPS on any vacuum line, where it can detect any particles that pass through its interaction region without risk of leakage. Although its design targets pressure ranges typical of chamber exhaust lines, where substantial gas flows move particles at relatively high speeds, it can also be used in high-vacuum lines where particles move more slowly.
Read the full article in the August/September issue of Semiconductor Digest.