Semiconductor fabs split flow measurement into two distinct worlds: low-flow process gases metered by thermal mass flow controllers (MFCs) inside the tool, and cleanroom-grade liquids (DI water, slurries, photoresist solvents) measured downstream by corrosion-resistant or non-invasive meters [S1][S6].
Process gas uncertainty targets have tightened roughly an order of magnitude, from 1% of full scale around year 2000 to a 0.1% of reading target today, and NIST's SLowFlowS standard now measures gas flow down to 0.01 cm³/min with 0.06% of flow uncertainty in hours rather than days [S1].
Process gas side: thermal mass flow controller (MFC) is the default
An MFC regulates gas flow into the process chamber to produce the desired structures during chip fabrication, and any drift directly reduces device yield [S1]. Because many of these gases are hazardous and flow standards for hazardous gases do not exist, the industry currently relies on approximate "meter factors" to extrapolate a nitrogen calibration to other species, which is exactly the gap NIST's Semiconductor Low Flow Standard (SLowFlowS) and rate-of-rise (RoR) method aim to close [S1].
For fab-side selection, an MFC is a flow sensor plus a control valve, not a passive meter, and interchangeability matters: a replacement MFC must reproduce the same flow behaviour so the existing recipe transfers without requalification [S1]. A typical fab fleet runs tens of thousands of MFCs, all of which are now expected to be SI-traceable rather than vendor-traceable [S1].
Cleanroom air and equipment cooling: MEMS calorimetric flow sensors
For non-process airflow (fan control, tool cooling, filter loading detection), MEMS-based calorimetric flow sensors are the typical choice. MinebeaMitsumi's semiconductor-sensor flow product line is a calorimetric sensor built on MEMS technology, paired with a high-resolution ΔΣ ADC that outputs airflow and air velocity as digital values [S4].
DI water and process liquids: all-Teflon or clamp-on ultrasonic
On the wet side, the failure mode is contamination, not accuracy. ChemTec's all-Teflon liquid flow sensors are built specifically for semiconductor wet-etch and cleaning loops because Teflon tolerates aggressive chemistries (HF, H₂O₂, solvents) and sheds essentially no metallic ions into high-purity water [S6].
Where the line cannot be cut, clamp-on ultrasonic flow sensors measure from outside the pipe, with no moving parts and no wetted metal, which suits UPW (ultra-pure water) and chemical distribution loops where any intrusion is a particle or ion source [S3][S10]. One concrete 2026 product spec: the CPD Series clamp-on ultrasonic flow sensor covers 0.1–50 L/min at ±2% accuracy, with built-in LED display and air-bubble detection for DI water service [S10].
Side-by-side: gas MFC vs MEMS airflow vs liquid flow sensor for fab service
Use the matrix below to shortlist by service, not by brand. Semiconductor fabs typically stock all three categories because each handles a different part of the utility and process network [S1][S4][S6][S10].
Decision criteria: typical fluid, accuracy class, wetted materials, response time, line size. Gas MFCs win on low-flow precision (down to 0.01 cm³/min, 0.06% of flow with NIST SLowFlowS traceability), use stainless or specialised alloys, and target single-digit sccm to several slpm flows [S1]. MEMS calorimetric airflow sensors target m/s-range air at modest accuracy, are all-silicon with digital I²C-style output, and respond in the sub-second range [S4]. All-Teflon liquid flow sensors cover aggressive chemistries with full wetted-Teflon construction [S6], while clamp-on ultrasonic meters keep the wetted path untouched and read 0.1–50 L/min DI water at roughly ±2% [S10].
For a broader primer on principle, range, and fluid matchup across ultrasonic, magnetic, vortex, Coriolis, and thermal technologies, see How to Choose a Flow Sensor: Principle, Range, and Fluid Matchup.
Selection criteria that actually matter in a fab
Four criteria dominate every fab flow-sensor spec-out: SI-traceable calibration, wetted-material compatibility, cleanroom particle contribution, and response time. NIST's 34 L and 677 L pressure-volume-temperature-time (PVTt) standards now achieve flow uncertainty below 0.025% on non-hazardous gases, with the newer RoR extension reaching 0.1 cm³/min and SLowFlowS reaching 0.01 cm³/min at 0.06% of flow [S1].
Wetted material is binary: 316L stainless or equivalent for inert gases and most clean DI water, all-Teflon (PTFE/PFA) for HF, HCl, H₂O₂, and solvent streams where any metal ion is a yield killer [S6]. Particle addition matters most on UPW loops, which is why clamp-on ultrasonic meters are specified for those branches even though an inline magnetic or Coriolis meter would be more accurate on paper [S3][S10].
Limits, failure modes, and what not to use where
Thermal MFCs do not measure liquids and fail on two-phase or condensing flows because the calorimetric principle assumes a single-phase gas. Liquid flow sensors based on thermal dispersion inherit the same limitation and should not be specified for two-phase or boiling dielectric fluids [S1][S4].
Magnetic (magmeter) flow sensors require water-based conductive media above roughly 20 µS/cm and are not suitable for deionized water or for non-aqueous solvents like isopropyl alcohol or photoresist thinner, where ifm's own matrix shows magmeter accuracy degrading or the principle becoming inapplicable [S2]. Vortex and mechatronic (piston-type) meters assume water-like viscosity and roughly 95% water composition, so they are the wrong pick for concentrated chemistries and for gases [S2].
Clamp-on ultrasonic meters depend on acoustic coupling and pipe material; they read poorly on lined or heavily attenuating pipes, and air bubbles in the line can corrupt the time-of-flight signal, which is why the CPD Series pairs the measurement with an explicit air-bubble detect function [S10].
Standards, traceability, and 2026 buying signals
For gas MFCs, ask the vendor for NIST-traceable calibration via PVTt or RoR, the supported gases, and the demonstrated repeatability at your operating flow; the difference between a 1% of FS and a 0.1% of reading MFC is now the difference between legacy and leading-edge fab practice [S1].
For wet-side meters, require a written wetted-material list (316L, PTFE, PFA, FFKM), cleanroom particle contribution data, and CIP/SIP compatibility if the line sees periodic sanitisation [S6][S10]. Trackable signals to watch through the rest of 2026: NIST interlaboratory comparisons on additional hazardous process gases, more MEMS airflow sensors with wider dynamic range, and clamp-on ultrasonic offerings extending above 50 L/min for higher-flow UPW loops [S1][S4][S10].