As-built Ti-6Al-4V thin parts were manufactured in three different orientations using EBM and characterized by laboratory X-ray tomography. Fatigue tests were performed. The comparison with results for machined samples from the literature showed a large reduction of fatigue strength. SEM observations of the fracture surfaces showed that surface defects which were identified as notch-like defects on tomographic images caused the failure. Their impact on fatigue results was rationalized by Kitagawa-Takahashi diagrams. A build orientation impact on the fatigue properties was observed and linked to its effect on defects distributions and crack growth. The limits of roughness measurements were also discussed.

Abstract. Understanding the slow densification process of polar firn into ice is essential in order to constrain the age difference between the ice matrix and entrapped gases. The progressive microstructure evolution of the firn column with depth leads to pore closure and gas entrapment. Air transport models in the firn usually include a closed porosity profile based on available data. Pycnometry or melting-refreezing techniques have been used to obtain the ratio of closed to total porosity and air content in closed pores, respectively. Xray-computed tomography is complementary to these methods, as it enables one to obtain the full pore network in 3-D. This study takes advantage of this nondestructive technique to discuss the morphological evolution of pores on four different Antarctic sites. The computation of refined geometrical parameters for the very cold polar sites Dome C and Lock In (the two Antarctic plateau sites studied here) provides new information that could be used in further studies. The comparison of these two sites shows a more tortuous pore network at Lock In than at Dome C, which should result in older gas ages in deep firn at Lock In. A comprehensive estimation of the different errors related to X-ray tomography and to the sample variability has been performed. The procedure described here may be used as a guideline for further experimental characterization of firn samples. We show that the closed-to-total porosity ratio, which is classically used for the detection of pore closure, is strongly affected by the sample size, the image reconstruction, and spatial heterogeneities. In this work, we introduce an alternative parameter, the connectivity index, which is practically independent of sample size and image acquisition conditions, and that accurately predicts the close-off depth and density. Its strength also lies in its simple computation, without any assumption of the pore status (open or close). The close-off prediction is obtained for Dome C and Lock In, without any further numerical simulations on images (e.g., by permeability or diffusivity calculations).