Experimental evidence of gas-mediated heat transfer in porous solids measured by the flash method

Laser/light flash analysis (LFA) is becoming an increasingly popular assessment tool for the thermal properties of porous samples. Elevated temperature measurements are typically conducted under a protective atmosphere; specifically, under inert gas flow. Immersing a sample with interconnected porosity in a gas environment theoretically allows heat transfer via gas conduction or convection. To prove that LFA measurements are affected by the choice of the gas environment, measurements are conducted under inert gas flow (helium and nitrogen), static inert gas (nitrogen) and vacuum with two different instruments and two sets of samples: highly-conductive alloy foams and insulating lightweight concrete. A heat conduction model accounting for light penetration was used to extract the apparent thermal diffusivity of the solid–gas mixture. A correlation between the Biot number and the conductivity of the gas environment was registered. This is unusual for an LFA experiment where radiative heat losses, which are independent of the gas environment, are typically considered to be dominant. At the same time, variations in either the microgeometry of the porous sample or the type of carrier gas produce a systematic change in the shape of the time–temperature profiles and the duration of the temperature transient. In lightweight concrete samples, switching the instrument to helium flow enables a multifold increase in the apparent thermal diffusivity relative to the measurements done under vacuum. An attempt to explain this variation was made with a two-temperature model of coupled solid–gas conduction. It is argued that an additional factor is missing from the analysis. • Temperature transients in porous samples investigated under N 2 , He and vacuum. • Changing gas species may produce different concavity of time-temperature curves. • This non-linearity seems to increase in thermally insulating samples. • Dual-phase heat conduction likely not enough to explain data variations. • More complex models needed as traditional time axis re-scaling not applicable.