A critical evaluation of thermal mass flow meters

Abstract

Many semiconductor processes require that stable and known flows of gas be delivered to the processing chamber. The thermal mass flow meter (TMFM) is used almost exclusively in the semiconductor industry for the admission of process gases. While TMFM’s have been used in the semiconductor industry for over twenty years, much still remains to be understood about their behavior. The abundance of TMFM manufacturers that make instruments which are supposedly interchangeable complicates the use of TMFM’s because the instruments generally have different designs and performance. While some attempt has been made via written standards to address the specifications of the instruments, these standards do not address all performance issues and cannot eliminate the systematic errors in the original manufacturers calibration of the TMFM’s. Further, the TMFM’s used to measure the process gases are generally calibrated with nitrogen and ‘‘corrected’’ for other gases, but the correction factors are not well understood and are of questionable reliability. It is also important to understand how the TMFM’s perform under conditions that differ from the laboratory conditions where they were calibrated and the measurement errors that are introduced as a result of these different operating conditions. This article presents data on the performance of five low-flow TMFM’s, from different manufacturers, with full scale ranges of 1.5×10−6–3.7×10−6 mol/s (2–5 sccm). The manufacturers’ calibration of the TMFM’s with nitrogen as compared to the National Institute of Standards and Technology (NIST) measured values differed by up to 17%. Three of the five tested TMFM’s were within the manufacturers’ stated tolerance of ±1% of full scale. While some of the instruments’ initial calibration was poor, all of the TMFM’s were stable to within ±1% of full scale over the test interval of nine months. The gas correction factors for five gases (argon, helium, hydrogen, sulfur hexafluoride, and hexafluoroethane) were measured and compared to manufacturers’ recommended values along with the temperature and flow dependence of the gas correction factors. Some of the gas correction factors agreed with the manufacturers’ recommended values to within ±1% while others differed by as much as 13%.