Additive Manufacturing (AM) offers unique design freedom in the fabrication of complex three-dimensional shapes. Consequently, in recent years, AM technologies have expanded into multiple sectors, such as aerospace, automotive industry 1,2 and medicine 3. One of the already commercialized AM techniques for metals and polymers is Laser-based Powder Bed Fusion (LPBF). In LPBF, parts are built up by sequentially adding precursor powder layers, followed by their selective consolidation by means of a laser, resulting in the densification of consecutive slices of a three-dimensional object4. The success of LPBF for producing metallic components with complex shapes has motivated scientists to investigate the possibility to use this technique for net shape ceramic parts, as their machining and shaping using conventional methods is challenging, time consuming and expensive due to their hardness and brittleness. However, laser-based AM of oxide ceramics implies additional challenges as compared to metals due to their significantly higher melting temperature, high viscosity, poor thermal shock resistance and poor absorption of laser light, as it is demonstrated in literature on LPBF of alumina 5–10 and zirconia9,11,12. The latter challenge can be well mitigated by introducing light absorption additives, such as carbon 6,13,14, titanium carbide 10,15 or other metal oxides 16–19. However, the issue of structural defect formation, such as pores and cracks, which is related to the extreme temperature gradients from high heating and cooling rates combined with low heat conductivity and high Youngs modulus, remains unsolved. The type and amount of defects as well as the final morphology of the samples depend strongly on the powder properties20 and the laser scanning parameters such as laser power, scanning speed, laser spot size and hatch distance 21. This dependence has been widely studied, mainly based on simulations22–25 and ex-situ characterization techniques, such as XRD, optical and electron microscopy. The effect of laser parameters on roughness, porosity and other structural defects is frequently investigated after manufacturing using X-ray19,26−28 and neutron 29 tomography.
The manufacturing of defect-free, high performance materials requires understanding of defect formation mechanisms and microstructural development as a function of process parameters. For that purpose, multiple in situ and operando techniques have been applied as for instance observation of sputter formation using a high speed camera30, IR imaging31, or two wavelength high speed-imaging thermography to measure operando thermal gradients and cooling rates during LPBF32.
In particular, operando techniques for LPBF process studies have been developed worldwide at synchrotrons, providing direct “live” insight into this complex manufacturing process33. This is reflected by numerous recent publications from such facilities, as the Swiss Light Source (SLS)34–37, Stanford Synchrotron Radiation Light source (SSRL) 38–40 Diamond Light Source (DLS) 41–44, European Synchrotron Radiation Facility (ESRF) 44,45 and the Argonne National Laboratory’s Advanced Photon Source (APS) 33,46.
Operando and in situ diffraction studies performed during LPBF allow for observation of phase transformations occurring prior and after melting and solidification as well as evaluation of temperature profiles and cooling rates35 34,37,47, while in situ X-ray radiography provides invaluable knowledge about melt pool dynamics 47–50 and formation of defects such as cracks 37,51 and porosity 49,52−54. Thanks to the rapid progress in detector and synchrotron development, X-ray radiography benefits from a high temporal resolution allowing frame rates up to 400 kHz 31,50 suitable for observing fluctuations of the keyhole, sputtering or powder particles ejection55. For metals, high speed X-ray radiography studies, have shown that porosity formation mechanisms are strongly dependent on the melt pool dynamics. Pores can be formed because of keyhole fluctuation and due to trapping pores at the end of a scanning track when the laser is turned off 22,24,40 or because of a lack of fusion. Keyhole pores typically have a high sphericity and are directly linked to the specific melt pool shape attributed to too high energy density. In contrast, lack of fusion results in elongated, irregularly shaped pores, which depend on the melt pool size, scanning pattern and wetting process between the substrate and the liquid phase.
However, not all phenomena occurring during the LPBF process can be observed with 2-dimensional imaging, because X-ray radiographic imaging provides a signal averaged over the penetration depth.
Operando radiography of LPBF is usually performed during one or a few laser tracks scanned perpendicular to the X-ray beam. In such measurements, it is not possible to spatially separate features along the beam direction, and thus, to link their location to the current laser position or to observe the interaction between scanning tracks. Many features, such as cracks or non-spherical pores can be detected only if they are well aligned with the X-ray beam. Moreover, any movement of liquid phase pores or powder particles, as well as crack propagation, can be observed only if it occurs parallel to the image plane.
In contrast, tomographic microscopy, a 3D imaging technique, allows to see defects that are oriented in any direction as well as to observe their motion in all directions. Single scan lines can be clearly distinguished and it is possible to distinguish between material being deposited during the imaging, and material that was solidified during deposition of the previous layer. Some advantages of 3D imaging in additive manufacturing were demonstrated by tomography studies performed before and after laser scanning single layers at ESRF 46. Thus, 3D imaging provides more detailed insight into the dynamic changes of the morphology allowing for observation of defect formation in the whole 3-dimensional manufactured object, which is more representative for the real process.
Performing fast tomographic imaging is, however, much more complex and challenging than radiography, since the observed object has to rotate during acquisition. Recent developments of synchrotron tomographic microscopy enabled ultrafast measurements allowing for 3D investigation of dynamic processes 56–58. Currently the highest reported time resolution of tomographic microscopy is 1000 tps (tomograms per second) 56, which required a rotation of a sample with a frequency of 500 Hz.
The typically used laser scanning speeds during LPBF of ceramics are in range from 1 to 100 mm/s 10,15,19,59. For this range, the achievable tomogram rate is already compatible with typical laser velocities for ceramics, as for instance for 10 mm/s laser scanning speed and acquisition with 1000 tps, the laser spot would move by 10 µm during acquisition of one tomogram. Thus, these parameters would allow to directly link the appearance of certain features with the melt pool dynamics. However, turning a sample including powder bed with such a high speed raises issues due to centrifugal forces acting on the powder.
In this paper, we present a method based on magnetism to assure stability of the powder and present the first operando tomographic microscopy studies performed during the LPBF process. The powder used in these studies is a magnetite-modified alumina.