Mentored by Brian Schuster and Jonathan Ligda Automated femtosecond laser serial sectioning in a scanning electron microscope Wyatt Jordan Mentored by Brian Schuster and Jonathan Ligda Introduction Results Results (cont.) Left to right pos.’s: 1”, 14.5”, 28”, bottom of poster 29”, overall size 42x30, set-up coords: 12.33X6.91, 14.83X14.75 Analysis of a material’s microstructure can be applied to explain material properties and predict resistance to ballistic impact. The optimal method for acquiring microstructure data is through serial sectioning coupled with image segmentation to generate 3-dimensional maps of the microstructure. Electron BackScatter Diffraction (EBSD) data can also be segmented into 3-dimensional maps with the added advantage of recording the crystal orientation of a material. However, generating the images for segmentation is impractical for researchers unless automated and implemented by machinery (Spowart, 2006). This project contributed to the automation of serial sectioning at the Army Research Laboratory (ARL) by optimizing and testing a LabVIEW™ program for controlling the entire sectioning process inside a Scanning Electron Microscope (SEM). Serial sectioning simulations were performed on a sample of copper to assess the accuracy of the stages and overall program flow. These simulations were executed at several slice thicknesses to determine the sectioning capability of the microscope. Sectioning with a slice thickness of 100nm caused inconsistent fiducial locations and stage positions. Error in the Z-stage became consistent when sectioning slices of at least 500 nm (See Graph 1), and increased by almost exactly 115 nm per slice. The fiducial locations were recorded and the error calculated in order to gauge the effectiveness of the image alignment. An alternating pattern was produced by the fiducial error (See Graph 2) and most likely caused by a 250 nm tolerance parameter in the image matching code. Adjusting this parameter should reduce error in both the image alignment and stage positions. Reliability of the program was critical for ensuring the various probes’ safety throughout the sectioning process. The shield and FSM are calibrated by the user prior to sectioning to keep the beam contained and prevent it from reaching any sensitive probes. If any of these parameters are not set, the program will fail to begin sectioning, and prompt the user for the necessary information. Automated movements of the stage and shield require less than two minutes per slice. However, the required milling time and parameters are yet to be determined. Anticipated time for each slice with milling is no more than 10 minutes. 100 mm 5 µm Methods and Materials Picture 1: SEM vacuum chamber and 5-axis stage which is capable of tiliting the required 40° for milling samples Picture 2: Fiducial with a 15um diameter made by the FIB for image alignment The central piece of equipment was the FEI Nova Nanolab 600i SEM (Picture 1) which offers dual beam capability. The main beam is an electron column which is standard on all SEMs for imaging, and the secondary beam is a Focused Ion Beam (FIB) for milling. The ion beam has a minimal milling rate, so a Clark-MXR, Inc. CPA 2110 femtosecond laser was installed and aligned with the microscope (Figure 1). Since the laser pulses at a rapid rate and has a wavelength of 775 nm, just outside the visual spectrum, safety procedures were established to protect personnel and the EBSD probe. The laser’s internal shutter was networked to the microscope computer so that it could be controlled through the LabVIEW™ program. Precisely aligning the laser for each sample is critical for the safety of the microscope and accurate milling. A Fast Steering Mirror (FSM) model 300-01 from Newport© was installed on one of the SEM’s ports and a calibration program written so it could be aligned with each specific sample prior to sectioning. A piezoelectric linear actuator with a non-magnetic stainless steel shield was installed and programmed in order to protect the sensitive EBSD probe during milling. The shield had to be dynamic so that it did not block the electron column when images were captured. After each slice of material is removed the stage moves to the proper imaging position, captures an image, then moves back to the milling position with a slight increase in height for milling the next slice. The z-axis of the SEM is less accurate than the other axes so a fiducial (Picture 2) is initially made by the FIB for image alignment. The final stage adjustments prior to milling are executed through Runscript, the microscope’s standard programming language, which has built-in image matching functions. The accuracy of the program was assessed by observing the error in fiducial location and z-stage position for various slice thicknesses. Figure 1: 1. Vacuum Chamber 2. EBSD Probe 3. Focused Ion Beam 4. Main Electron Column 5. Laser Objective 6. Fast Steering Mirror 7. Femtosecond laser (Echlin, 2012) 4. 3. 6. 5. Conclusions The implementation of all components necessary for sectioning was completed in LabVIEW, proving that automated serial sectioning is practical with the equipment available at ARL. The simulations were successful, and indicated that the current set-up will allow for sectioning of materials in a single day, which is much faster than other automated processes. Image segmentation was not possible due to a hardware failure on the microscope, which delayed the project several months. Development of the program will continue over the summer, and the first 3-dimensional models of microstructure produced from microscope images. Once these models have been successfully generated, EBSD analysis will be incorporated to provide additional information about materials’ crystal structure. 7. 2. 1. References Echlin, M. P. [TriBeam System Diagram]. (2012). A New Tribeam System For Three-Dimensional Multimodal Materials Analysis. Review of Scentific Instruments 83.2 Spowart, J. (2006). Automated serial sectioning for 3-D analysis of microstructures. Scripta Materialia, 55(1), 5-10. Graph 1: Consistent error in Z-stage positions with r = -0.998 and error increasing 115 nm/slice Graph 2: Error alternates because of the tolerance set in Runscript, inital fiducial location of (-70 µm, -70 µm)