Gas plasmas have become a fundamental building block in many semiconductor manufacturing processes. Plasma torches used to create these gas plasmas have three components: an induction coil, a plasma confinement tube, and a gas distributor or torch head that introduces multiple gases into the torch. RF generators supply the high-frequency electrical energy needed to transform the plasma-forming gases flowing through the torch, typically oxygen or a fluorine-bearing gas, into a plasma. The RF generators used for semiconductor manufacturing typically operate in the low megahertz or tens of megahertz frequency range and are expected to output high RF power at those frequencies for long periods. For example, ALD and CVD processes use RF generators with output powers on the order of a few kilowatts.
About three years ago, a major semiconductor device maker experienced a recurring problem with its RF generators. The company found that more than half of the RF generators it deployed in its manufacturing lines were failing within the first two years of service. Further, the same model RF generators obtained from the same RF generator vendor simply were not behaving similarly when used for exactly the same processes under exactly the same conditions. Nor were these supposedly identical generators operating for consistent lengths of time before failing. Clearly there was variation from one generator to the next, even within the same model.
A further complication occurred during procurement of these RF generators. Procurement people were acquiring generators using general specification requirements and these requirements were, at times, opaque to the intended process application. In some cases, equipment was being purchased in bulk quantities and then assigned to different processes on the semiconductor manufacturing lines. When these generators were deployed, they had not been designed or optimized for the specific task to which they were assigned, exacerbating the reliability problem.
The RF generator suppliers felt that they would be able to supply more reliable generators if they could collaborate with their customers so that they could purpose-build their generators for the intended uses. However, the semiconductor makers preferred to keep the specifics of the manufacturing process applications for these generators proprietary, for obvious reasons. To make matters worse, customers did not always return failed units to RF generator vendors for analysis. Instead, the RF generators were sometimes sent out to be refurbished by third parties or repair depots, and then redeployed. As a result, failure analysis proved challenging to obtain.
This is exactly the type of situation that SEMI’s Semiconductor Component, Instrument and Subsystem (SCIS) technical community exists to address. SCIS develops test methods aimed at measuring component defects for the greater semiconductor manufacturing community. SCIS tackled this RF generator problem and developed a standard test method for measuring specific RF generator characteristics. Using this test method, RF generator manufacturers can publish results for their generators in a standardized way that allows their customers to make fair, application-specific comparisons among models and vendors.
Many aspects of an RF generator needed to be considered. A key aspect that interested integrated device makers (IDMs) and capital equipment OEMs was a transient-response test for RF generators.
A transient-response test standard established by the SEMI-E135 standard did exist, but its tests were run only with 50-ohm RF output loads. SCIS decided to expand this transient-response test by adding high- and low-impedance load tests to the existing 50-ohm load test.
The initial response to this plan was not enthusiastic. The semiconductor makers feared that this simple expansion of an existing test standard would not produce a test regimen that would help solve what they considered to be the real problem: RF generator reliability. However, a major semiconductor equipment OEM differed, and felt that the two additional load conditions would provide a much better understanding of an RF generator’s capabilities. A second major semiconductor equipment OEM also got involved by providing additional, valuable feedback on the developing RF generator testing standard.
In the end, the general feeling in the community is that this newly revised standard levels the playing field and makes it easier for customers to compare RF generators from different generator vendors. Now that this revised SEMI-E135 standard with the additional output load resistances has been published, the SCIS technical community has gained broader support and is now digging into the creation of a reliability test standard for RF generators to meet the greater semiconductor manufacturing community’s strong need for such a standard.
How SEMI Standards are Made
This sequence of events illustrates how standards are developed at SEMI. The SCIS technical community (or some other technical community within SEMI) develops and incubates test methods until a document is ready for standardization. At that point, a SEMI Standards task force is created. Companies within SCIS work with the task force (or become the task force) to ready the document for standardization. For the SEMI-E135 revision, the list of participating companies encompassed the entire semiconductor manufacturing community including RF generator suppliers, semiconductor capital equipment OEMs, and IDMs. All stakeholders participate.
Figure 1 illustrates the sequence of events that occurred during the revision of the SEMI-E135 standard, after the test methods had been developed by SCIS as discussed above.
Figure 1: Timeline for SEMI-E135 RF generator test standard revision after SCIS had developed the new load tests.
Balloting, as illustrated in Figure 1, is the main way that SEMI obtains global consensus in the standards-making process. To achieve this, SEMI sends out the standard ballot proposal, or in this case a major revision of an existing standard. The changes to SEMI-E135 were sufficiently extensive that it was treated as a complete rewrite to this standard.
On first ballot, the revised SEMI-E135 standard received several rejection votes, which also included suggested modifications that would remove the objections. These ballot rejections caused the proposed standard to be further revised, with both technical as well as editorial changes, triggering a SEMI Standards process called a Ratification Ballot. This approach takes less time than starting the balloting process over again. The final revised standard was published in September 2018.
Having all stakeholders participate in the early development of the revised standard helped move the standard through the balloting process immensely, but customer participation was especially important. In the end, the semiconductor device makers and equipment OEMs are the ultimate beneficiaries of a standard like SEMI-E135. When end customers help to drive a standard’s development, there’s added pressure to move the standard along in the standardization process and the standard is far more likely to be useful for their purposes.
And that’s a very good thing.
For those looking to learn more about SCIS or engage in ongoing efforts, please contact Paul Trio, senior manager of Strategic Initiatives at SEMI, at ptrio@semi.org.