A Critical Component of Semiconductor Processing
Semiconductor process tools used in chip fabs are highly complex systems costing millions of dollars. Every part of the system must work flawlessly time-after-time, 24/7. An incredible amount of intellectual capital is invested in the design, manufacturing and maintenance of these systems. What is interesting is that there are components in these systems based on technologies more than 50 years old. One of these components is the humble mass flow controller or MFC. In comparison with the cost of a semiconductor tool the cost of an individual MFC is miniscule. Yet, if it fails, the result can be the loss of a wafer-lot. Depending on the product and where in the process the failure occurs, the cost could be a million dollars, or more. Variability in performance of one MFC can affect wafer yield. Yet, there can be upwards of 40 MFCs installed on a single process tool.
What amazes me is how under appreciated the MFC is. It is generally viewed as a commodity. As a result, there is a continual downward price pressure that reduces the MFC manufacturer’s margins. Lower margins mean that less investment goes into product improvements. As a result, the MFC industry is stagnating. The earlier market leaders have gone through wave after wave of consolidation. New entrants are not focused on the semiconductor industry, generally, and so develop products that are not optimal for semiconductor processing.
My early experience with MFCs was as a user. I had about a dozen installed in each of three R&D semiconductor tools. Due to limited funds and natural curiosity, I tried to repair my own MFCs. I succeeded about half the time. I later joined one of my MFC suppliers. Part of my training was to meet with one of the pioneers who developed the early thermal MFC. My first task was to update the “conversion factors” for the company’s products. Thermal MFCs or tMFCs measure flow by measuring the heat transfer from a small heater to a temperature sensor. Conversion factors are supposed to allow one to measure flow of a gas different from the one on which the MFC was calibrated. My trainer explained that the conversion factors were based on the heat capacity of the gas. However, it was still not understood why some gases were better matched than others. When I explained that the basis for heat capacity in molecules is their internal degrees of freedom, the company launched an investigation into a spectroscopic approach to mass flow.
Working with a number of MFC users in the semiconductor industry, I discovered that there was a general frustration with MFCs. They require one to know a body of knowledge that often was outside of their experience or training. For example, I had a customer with a PhD in electrical engineering call me one day asking what the pressure drop was for our MFCs. I asked him what flow rate he was interested in. He replied “Any flow rate.” I explained that the valve in a MFC acts like a variable resistor. Flow is analogous to electrical current and pressure is analogous to voltage. So, for a given pressure (voltage), the flow (current) varied with the valve position (resistance). Without a specific flow rate, I could not provide a definite pressure drop. He got it and so we went on to a productive discussion.
When I joined the next MFC company, one of my goals was to eliminate the conversion factors. Leveraging the digitally-based products we developed, our team introduced a calibration technique that stored multiple calibration curves to accurately flow a number of gases. Prior to our product launch, I wrote an article in a trade magazine reporting that conversion factors only work for a very narrow range of process conditions. This helped to increase early market acceptance. However, we wanted to move beyond multiple calibrations. We then had the opportunity to develop a completely new product line. This led to the development of Coriolis MFCs. These required a single factory calibration and could be used for any fluid with equal accuracy. However, the company was sold and then sold again; and the Coriolis MFC did not move beyond MVP.
This has led to a vision for a successor to the MFC that is very user friendly, reliable and robust; a product that would add significant value to the user. The technologies are available from various industries to employ in the development of an entirely new product that would be optimal for semiconductor manufacturing and applicable to many other industries, as well. I am currently, looking for funding sources for its development.
