A D V A N C E D M A T E R I A L S & P R O C E S S E S | J A N U A R Y 2 0 2 0 1 7 (a) (b) (c) packages can be used to digitally recon- struct and interrogate acquired data, e.g., DREAM.3D and MIPAR [13,14] . A common form of mechanical serial sectioning using metallograph- ic grinding and polishing is fully auto- mated with Robo-Met.3D, which uses robotics to perform mechanical oper- ations and leverages an integrated in- verted optical microscope for imaging. Multiple polishing disks within the sys- tem allow polishing to be optimized for multiple material removal rates, surface finishes and reduction of preparation artifacts. The Robo-Met.3D at Sandia National Laboratories has been used to interrogate samples with dimensions ranging from hundreds of microme- ters (volume <1 mm 3 ) to comparatively large components with dimensions of 30-40 mm in length (volume <27,000 mm 3 ). The fidelity of the data produced is associated with the material remov- al rate (i.e., the amount of material re- moved between consecutive slices), which can be as fine as a single micron using Robo-Met.3D. Additionally, with mechanical serial sectioning, there is no sacrifice in resolution for increas- ing part size, as both a 1 mm 3 and a 27,000 mm 3 component can be imaged at a one-micron resolution. Acquisition times for a single slice with the authors’ Robo-Met.3D typically range from three to five minutes, and up to twenty to thirty minutes. Experiments vary from dozens to several thousand slices de- pending on sample size and the desired data fidelity. Due to the arduous and meticu- lous labor required for serial sectioning, M odern, advanced characteriza- tion techniques offer an unprec- edented understanding of ma- terials. Until recently, these techniques have primarily relied on 2D character- izations to probe materials. Over the past decade, however, 3D characteriza- tion has become increasingly practical within the materials community as the cost of computing resources decrease and software sophistication continues to increase [1-6] . These trends make auto- mation and data processing more feasi- ble than ever before. In many instances, 3D character- ization can provide valuable insight into microstructure, multimaterial sys- tems, and manufactured components beyond the reach of traditional 2D anal- ysis. Greater detail on features of inter- est is possible when compared to 2D techniques by enabling full volumetric analyses of spatial distributions, mor- phologies, interconnectivity of features, and feature volume fractions [4-6] . Mea- sures of this type can be further lever- aged to provide additional insight into material behavior, microstructure, and failure pathways. 3D characterization can be broad- ly grouped into two categories: nonde- structive testing (NDT) and destructive techniques. The resolutions and char- acterization volumes possible across most currently available and broad- ly deployed 3D characterization tech- niques are shown in Fig. 1(a). Atom probe, electron tomography, and fo- cused ion beam are all destructive techniques that encompass increas- ingly larger volumes from angstroms to microns, respectively. However, for the study of engineered components, the ability to resolve features within bodies ranging from millimeters to meters is needed. For this reason, ultrasound and microcomputed tomography (µCT) are two of the more widely used NDT tech- niques for larger parts. However, data acquired from these techniques can be challenging to interpret and resolu- tion remains dependent on sample size and geometry [6] . As shown in Fig. 1(a), mechanical serial sectioning intersects the observation volumes of both ultra- sound and µCT and provides for argu- ably the greatest trade-off in resolution and obtainable volume relative to typ- ical engineered component length scales, i.e., microns to millimeters. MECHANICAL SERIAL SECTIONING Mechanical serial sectioning is a destructive 3D characterization tech- nique in which: • Material is removed from a sample in a slice-by-slice fashion either by machining [4] , ablating using a fo- cused ion/laser beam [7] , or mechani- cal grinding and polishing [8] . • Characterization is then complet- ed on each successive slice, most commonly through optical [9] and/or electron microscopy [10-12] . This slice-by-slice collection meth- od produces a series of 2D images that when stacked together represent a 3D structure, see Figs. 1(b) and (c). Fol- lowing data collection, both custom and commercially available software Fig. 1 — (a) Resolutions and characterization volumes possible using a variety of 3D characterization techniques; illustration of the slice-by-slice collection method used to reconstruct a series of (b) individual 2D slices into (c) a reconstructed volume.