The flow distribution ill a silicon microchannel heat sink was studied using infrared micro-particle image velocimetry (IR mu PIV). The microchannel test piece consisted of seventy-six 110 pin wide x 371 pill deep channels etched into a silicon substrate. Inlet and outlet manifolds, also etched into the substrate, were fed by 1.4 mm inner-diameter tubing ports. An image-processing algorithm was developed that significantly improves the quality of IR mu PIV recordings in low signal-to-noise ratio environments. A general expression for the PIV measurement depth is presented, which is valid for PIV images that have undergone a threshold image-processing operation. Experiments were performed at two different flow rates: 10 ml/min (Re = 10.2) and 100 ml/min (Re = 102). Little flow maldistribution was observed at the lower flow rate. However, significant flow maldistribution was observed at Re = 102, with the channels near the center-line having all approximately 30% greater mass flux than the channels near the lateral edges of the heat sink. Numerical simulations Carried Out for flow in the microchannel heat sink agreed very well with the experimental measurements, validating the use of a computational approach for studying the effect of manifold design oil flow distribution in microchannel heat sinks. (C) 2007 Elsevier Ltd. All rights reserved.