The fabrication of three-dimensional (3D) microfluidic networks entirely made of SU-8 with integrated electrodes is reported. The described technology allows the fabrication of uncrosslinked SU-8 dry film on a polyester (PET) sheet and its subsequent lamination to form closed microstructures. Unlike other reported methods, transferred layers are patterned following the bonding step allowing a more accurate and simple alignment between levels than techniques using already patterned layers. Dry release of the complete polymer microstructure was demonstrated. Flexible microfluidic chips were obtained. This technique uses simple tools and no wafer bonder is used but lamination techniques which are more collective processes. Limitations in the method for layers thicker than 50 µm have been observed and are discussed. Hydraulic flow experiments have been performed to study the deformation of the cover layer which could influence adjacent flow in a three-dimensional configuration. Important deformations have been observed for layers 10 µm thick and an average pressure greater than 100 kPa. No deformations have been noted for layers with thicknesses greater than 35 µm and for average pressures up to 200 kPa. No failures occurred within the range of the experimental set-up, i.e. up to 300 kPa.

Micro Particle Image Velocimetry (μ-PIV) is a non-intrusive technique widely used nowadays to experimentally obtain the velocity field of a micro flow. The main goal of this research was to examine the influence of particle concentration and the number of images acquired, on the accuracy of the μ-PIV velocity measurement. For this reason, a comparison between experimental and analytical values was made. It has been demonstrated that the influence of the seeding concentration on the accuracy of the velocity measurements, into the investigated range, can be considered insignificant. On the other hand, the number of images selected for the cross-correlation is more important for the accuracy of the measurements. By increasing the quantity of images processed it is possible to artificially increase the seeding concentration and reduce the scatter. However, this considerably increases the processing time for the experiment. A trade-off is required between obtaining a highly accurate result without losing precious experimental down time. When the range of the concentration is fixed, it is possible to set the maximum inaccuracy allowance tolerated for the experiment. There is a compromise between a better precision and adequate time to process the data.

No abstract