Freeze-casting is a processing route for fabrication of materials with microchannels. In freeze-casting the anisotropic solidification of a suspension due to an applied thermal gradient facilitates the segregation of suspended particles in between the growing solidifying fluid phase. The solidified fluid phase is removed by sublimation and the structure is densified, now having pores where the crystalized fluid phase used to be.
In freeze-casted samples, domains, within which the orientation of the pores are similar, are always observed. The domains reflect the orientation of pores formed at nucleation [1]. Studies have investigated methods for domain ordering at the point of nucleation, primarily by a mold-altering approached; introduction of an insulating wedge between cooling source and suspension [2], patterning of the mold surface [3], or by using a mold with an additional cooling side [4].
However, although these approaches are reported successful, samples have been evaluated at only few cross-sections perpendicular to the freezing direction. For applications of mass transportation through the micro-channels on the scale of centimeters, such as heat exchangers, it is not evident that the domain ordering is maintained. It is therefore relevant to consider how domains evolve throughout freeze-casted structures.
Here we analyze the pore structure and associated flow properties of a full freeze-casted sample. We consider a sample of freeze-casted La0.66Ca0.33-xSrxMn1.05O3 (LCSM) following the procedure described by Christiansen [7,8]. Following freeze-casting and sintering, the sample is subjected to X-ray tomography to obtain a full 3D reconstruction of the centimeter high sample. X-ray tomography has previously been used to reconstruct morphology and microstructures for freeze-casted samples [5,6]. Using the tomography data we analyze the flux through the structure, quantifying the local flux at each channel as well as dead ends. We also quantify the structure tensor to determine the pore domains.
We find that only very few channels exhibit dead ends, but the flux can vary by a factor of 10 within a single channel. Quantifying the low flux points throughout the sample reveals no correlation with position. With regards to pore domains, the investigation reveals that the domains grow significantly in size throughout the length of the sample. However, the direction of the individual domains has a similar distribution at the bottom of the sample compared to the top, indicating that domains orientation is not correlated with size. The domain orientation, however, is not uniform but shows preferential directions, with the most preferred directions appearing 2.5 times as frequently as the least preferred directions. This must result from the initial freezing nucleation of the sample.