In Situ Real-time Visualization and Corrosion Testing of Stainless Steel 316LVM with Emphasis on Digital In-line Holographic Microscopy

Abstract

Digital in-line holographic microscopy (DIHM) has been incorporated as an additional simultaneous in situ optical technique with ellipsomicroscopy for surface imaging and microscopy to study metastable pitting corrosion on stainless steel 316LVM in simulated biological solutions. DIHM adds microscopic volume imaging, allows one to detect local changes of the index of refraction in the vicinity of a pitting event, and allows one to track tracer particles and/or material ejected from the pitting sites. To improve the pitting corrosion resistance of stainless steel 316LVM, a simple surface treatment was tested and the aforementioned imaging techniques were used to verify that pitting occurred only on the wire face. Treatments consisted of polishing the samples to remove the passive layer, then immersing the wires in 90 C nanopure water for several hours. Treated wires show a marked increase in pitting corrosion resistance over untreated wires: the pit initiation potential increases by a minimum of 200 mV. Additional testing with scanning electron microscopy and energy dispersive X-ray spectroscopy indicate that the removal of sulphide inclusions from the surface is the most probable cause of this enhancement. To increase holographic reconstruction performance, Graphics Processing Units (GPUs) have been used; 4 Mpixel holograms are reconstructed using the dot product approximation of the Kirchhoff-Fresnel integral in 60 ms on a Tesla c1060 GPU. Errors in sizes and positions can easily be as large as 5 to 10 % for regions where the dot product approximation is not valid, so algorithms with fewer or no approximations are also required. Reconstructions for arbitrary holographic geometries using the full Kirchhoff-Fresnel integral take approximately 1 hour (compared to 1 week on a quad-core CPU), and reconstructions using convolution methods, in which the results of 256 reconstructions at 4096 x 4096 pixels in one plane are combined, take 17 s. This method is almost exact, with approximations only in the obliquity factor.