Autonomous Multimodal Spectrum Imaging Using Hybrid DMScript and Python Code in DigitalMicrograph

Session

Poster Session Three

Authors

Dr Liam Spillane (2), Dr Bernhard Schaffer (2), Dr Paul Thomas (2), Dr Michael Zachman (1)

Affiliations

1. Centre for Nanophase Materials Sciences, Oak Ridge National Laboratory
2. Gatan

Keywords

- Spectroscopy and spectrum imaging

- Automated Control, Advanced Data Processing and Modelling Techniques

- 4D-STEM, phase-sensitive, ptychography and Lorentz techniques

- Correlative microscopy

Abstract text

Programmable mask-based scanning capabilities of a flexible next generation scan control system (Digiscan3) are used in combination with a hybrid DMScript and Python code framework, to explore and demonstrate novel spectrum image (SI) functionality for autonomous experimental setup and data acquisition.

Fast, dose-efficient SI data acquisition functionality is inbuilt to Gatan’s DigitalMicrograph. This allows spatially correlated multimodal data to be collected routinely at sub-nm or atomic spatial resolution. Datasets typically comprise electron energy-loss spectroscopy (EELS), energy dispersive X-Ray spectroscopy (EDS) and/or (energy-filtered) 4DSTEM data taken from fixed points, line-profiles, or rectangular area scans. While the native features have a high degree of flexibility, the ability to add modifications in bespoke, user-defined ways is desirable. In the case of single shot in-situ experiments where all multimodal data must be acquired at fixed in-situ conditions, script-based automation was shown to be an effective means of acquiring serial multimodal data elements to maximise in-situ stimulus resolution whilst also minimizing time spent at the TEM [1]. Specimens in which features of interest represent a minimal fraction of the total scan area are non-suitable for rectangular area-based scanning. Scan efficiency and data statistics in this case can be greatly enhanced by adding automated feature recognition [2].

Hybrid scripting has recently been demonstrated as a robust and scalable approach to customizing SI data acquisition in DigitalMicrograph. The hybrid framework presented utilizes interoperable DMScript and Python codes to leverage strengths of both languages, providing novel features such as autonomous feature recognition, scan point and/or scan area assignment, and programmable TEM stage control [2]. Programmable mask-based scanning is now possible using a next generation scan control system (Digiscan3). Such a scan modality provides high selectivity and scan efficiency, combined with dose averaging provided by the array-based scanning approach. 

Here we present automated multimodal SI data acquisition from a number of material systems. These include sparsely distributed samples such as metal nanoparticles, oxide nanoparticles, and PGM-free catalysts, in addition to samples containing irregularly shaped target regions, such as 3DNAND or finFET semiconductor devices. This sample set is used to highlight benefits of programmable mask scanning when used in an autonomous framework.

Figure 1. DigitalMicrograph screenshot showing output from a custom automated SI acquisition launched using a hybrid script. Raw ADF images and particle analysis mask generated by ParticleSpy Python package [3] are shown. EELS elemental maps for silver M4,5, nickel L2,3 and C K are generated from autonomously acquired EELS SI data are also shown in addition to an extracted spectrum from particle cluster 4.

References

[1] Spillane, L et al., Microsc. Microanal. 27 (Suppl 1), (2021), p.630

[2] Spillane L & Schaffer B, Microsc. Microanal. 28 (Suppl 1), (2022), p.2922

[3] Slater, T. (2021). ePSIC-DLS/ParticleSpy:0.5.2