Most flow meters covered so far measure volume — turbines count rotations, vortex meters count wakes, magnetic meters time a velocity. But for gases, volume is a lousy proxy for what you actually care about: mass. A cubic meter of natural gas at 10 bar contains ten times the molecules of the same cubic meter at 1 bar. Combustion, chemical reactions, and emissions reporting all care about molecules, not cubic meters. Thermal mass flow meters skip the pressure/temperature compensation dance entirely by measuring mass flow directly through heat transfer.

The physics: a gas flowing past a heated element carries heat away at a rate proportional to mass flow rate, specific heat, and temperature difference. Two common architectures:

• Constant power (Δт): Apply fixed power to a heater, measure how much cooler it stays as flow increases. Output is a temperature delta.
• Constant temperature (CTA): Use a feedback loop to keep the heated sensor at a fixed temperature above the gas stream. Power required to maintain that delta is the flow signal. This is what most modern instruments use — faster response, better linearity.

The governing relationship is King's Law: P = (A + B·√ṁ)·ΔT, where P is heater power, ṁ is mass flow, and A, B are calibration constants. The square-root dependence means thermal meters have excellent turndown at low flows — often 100:1 — where differential pressure meters get noisy.

Real-world example: Semiconductor fabs use thermal mass flow controllers (MFCs) to dose process gases like silane or ammonia into deposition chambers at flow rates of a few standard cubic centimeters per minute (sccm). A 10 sccm MFC controls flow to within ±1% of setpoint — accuracy that's impossible with volumetric meters because chamber pressure varies during the process. The "standard" in sccm refers to a reference condition (typically 0°C, 1 atm), and thermal meters output directly in those units without correction.

Rule of thumb / gotcha: Thermal meters are gas-specific. The signal depends on the gas's heat capacity (Cp). A meter calibrated for nitrogen reading 100 sccm will show roughly 72 sccm if you flow argon through it (Cp ratio ≈ 1.0/0.52 ≈ 1.4 the other way). Manufacturers publish "gas correction factors" (GCF) so one meter can serve multiple gases: actual_flow = indicated_flow × GCF. For mixtures, GCF must be calculated from mole-weighted Cp values — and that's why every fab MFC has a gas label.

Limits: condensable vapors fool the sensor (latent heat dwarfs sensible heat), and high-velocity streams can cool the element faster than the electronics can track. Stay subsonic and dry.