Leak Testing Meets New Semiconductor Industry Needs.
Helium leak detectors have evolved quite dramatically in recent years. Thanks to the electronics revolution, expanded features and interface capabilities have become more user friendly and more suitable for advanced applications, such as those found in the semiconductor industry.
Manufacturing integrated circuits requires many complex steps that use high-purity gases inside process chambers under vacuum conditions. To maintain the purity of these gases and the integrity of wafer surfaces during each manufacturing step, the semiconductor industry began switching to so-called "dry" vacuum pumping packages in the early 1990s.
Dry pumps contain no vacuum-side hydrocarbon lubricants to possibly contaminate process chambers by vaporizing and then backstreaming lubricant materials from the pump to the processing chamber. Because hydrocarbons in even small amounts (ppm levels) can dramatically affect product yields, oil-based diffusion pumps were replaced by dry turbo pumps, molecular drag pumps, and hybrid pumps, and rotary vane roughing/backing pumps were replaced by membrane, scroll, or frictionless multistage Roots technologies.
This trend was quickly extended to leak detection in 1991 when Alcatel Vacuum Technology, Hingham, Mass., introduced the first compact high-performance dry leak detector at that year's Semicon West show.
One reason that dry leak detection is important is that oil in a standard leak detector backstreams most during the roughing phase when the chamber being tested is exposed directly to the rotary vane pump that standard leak detectors use. For a given type of pump lubricant, the pressure, temperature, and time of operation are the most critical parameters in regulating backstreaming amounts. Lower pressures, higher temperatures, and longer times of operation are associated with increased backstreaming levels.
Oil backstreaming is even more critical for "wet" leak detectors that operate in a counter-flow mode where the rotary vane pump is always connected to the chamber or part being tested, thereby exposing it to hydrocarbons on a continuous basis. The problem is somewhat less severe for units that provide fine leak testing, since these contain a high-vacuum pump that functions as a barrier, preventing most heavy molecules, such as hydrocarbons, from migrating toward the chamber.
Another important application of dry leak detectors involves the high-purity gases used in semiconductor processes. These gases are supplied by means of gas tanks located anywhere from several meters to several hundred meters away from the process chambers they serve. Connecting the two are gas lines that are stainless steel tubes 0.6-5.0 cm in dia. These gas lines must be leak tested to tight specifications, with the standard being 1 x [10.sup.-9] atm [cm.sup.3]/sec for helium. This leak rate corresponds to an air leak of 3.7 x [10.sup.-10] atm [cm.sup.3]/sec, which represents a theoretical leak diameter of approximately 0.25 [micro]m under one atmosphere of pressure differential.
The reason for such fine leak-rate requirements is to guarantee that no external gas species will ever penetrate the gas lines or that none of the gases that are used in a process will leak outside the gas lines. Due to the nature of process gases such as arsenic, silane, and chlorine, the latter is a safety issue for everyone working in the fabs.
Leak testing gas lines down to [10.sup.-9] atm [cm.sup.3]/sec is a challenge and used to be time consuming. The traditional method consisted of connecting the leak detector to one end of the gas line while the other end was sealed. Helium would then be sprayed around each weld and connection susceptible to leaking. This technique involved a very long process and did not guarantee the proper leak rate specification.
Extensive leak tests run by Alcatel have shown that for a gas line that's 0.6 cm in dia and 91 m or longer, such a leak cannot be detected. The main reason is that gas in the line is maintained in the molecular-flow regime, so helium molecules moving randomly take a very long time (more than 45 min) to go from one end of the gas line to the other. During that time, the helium gets diluted inside the gas line and cannot be measured. These tests have been run with leak detectors capable of testing far beyond the [10.sup.-9] range (5 x [10.sup.-12] atm [cm.sup.3]/sec).
To resolve this issue, Alcatel has developed a technique that meets these critical leak requirements. The technique consists of using a high-purity carrier gas such as nitrogen to speed up the response time by a factor of at least 10 compared to the conventional method. The presence of the carrier gas takes the gas inside the line into the viscous regime, thereby resolving the problem of poor conductance. Using this technique, a 1 x [10.sup.-9] atm [cm.sup.3]/sec leak rate can be detected in less than 4 min for a gas line 0.6 cm in dia and 122 m long. The recovery time can be made even faster by increasing the flow of the carrier gas to help remove the helium in the gas line.
To help implement this technique, Alcatel supplies charts that indicate the estimated response time based on the internal diameter, the flow rate of the carrier gas, and the length of the gas line. This tool is very convenient and ideal for the engineer or technician performing the leak test, since it helps guarantee that no leaks will be missed.
Alcatel has also designed special calibrated leaks that are mounted in-line with the 0.6-cm gas line using metal seals of VCR type. These dedicated calibrated leaks range from 1 x [10.sup.-9] to 1 x [10.sup.-5] atm [cm.sup.3]/sec. Their internal design allows the now of carrier-gas molecules to sweep the surface of the calibrated leak membrane, transporting the helium molecules toward the other end of the gas line where the leak detector is located. This is in contrast to conventional calibrated leaks mounted on a `T' setup, where the response would be neither as fast nor as precise.
Speed is often crucial in semiconductor manufacturing, making it essential that tests for leaks be performed as quickly as possible. This means having a unit that can respond rapidly to a leak and then recover rapidly after a large leak has been detected.
The response time of a leak tester is measured by the time needed to detect 63% of the total helium signal. The response time is the ratio of the total volume evacuated by the leak detector divided by the helium pumping speed available, in the case of an application where no auxiliary pumps are used during the leak test. What this means is that, for a given volume, the more helium pumping speed available at the inlet of the leak detector, the faster the response time.
Once the unit has measured a gross leak, it must be able to recover from it and then go back to the same helium background level it had prior to the leak test. This characteristic is defined as the recovery time. It is directly dependent on the helium pumping speed.
In a perfect world, for a given volume, pumping speed is the only contributing factor to the response and recovery times. In reality, another parameter comes into play: the conductance of the vacuum line used to connect the leak detector to the chamber or part being tested. Conductance has a greater impact in the molecular-flow regime than in the viscous regime. Generally speaking, conductance is maximized by keeping the vacuum lines as short as possible and their diameters as large as possible.
As leak specifications become tighter and tighter, it is essential that basic fundamentals not be overlooked. A leak of 1 x [10.sup.-7] atm [cm.sup.3]/sec under a pressure differential of 1 MPa corresponds to a theoretical hole diameter of less than 0.5 [micro]m (for a wall thickness of 0.6 cm). Such a hole can be easily clogged by grease, fingerprints, contamination, and water. In short, cleanliness can be a limiting factor in the size of leaks that can be found and should be given close consideration.
Besides having the potential of clogging leaks, contamination can also be a contributing factor in the helium background. Helium will get trapped in the contamination and will then be released when the part is exposed to the leak detector, thus increasing the background level.
It is recommended that parts be as clean and dry as possible prior to leak testing. Obviously the handling of the parts is also crucial and should be addressed. Another important parameter to be considered is the test pressure. It is strongly recommended that parts be tested under the same pressure as the operating pressure. If the pressures are different, the part will undergo a different mechanical strain, and leaks might be missed. Indeed, leaks might not develop under a lesser pressure, but will develop once the pressure is raised to the operating level.
Did you find this article useful? If so ... Circle 205
For more information on helium leak detection, contact:
Alcatel High Vacuum Technology 617-331-4200 Ext. 240 firstname.lastname@example.org
Jean-Pierre Deluca is product development manager for leak detection at Alcatel High Vacuum Technology Inc., Hingham, Mass.
|Printer friendly Cite/link Email Feedback|
|Comment:||Leak Testing Meets New Semiconductor Industry Needs.|
|Publication:||R & D|
|Date:||Oct 1, 1998|
|Previous Article:||E-Beam Systems Combine Durability and Precision.|
|Next Article:||Protecting Cluster Tools From Photoresist Residues.|