Conclusions

• Implemented a suite of open-source phase retrieval algorithms
  • Particular emphasis on robustness against experimental artifacts
  • New preprint out now: 2309.05250
• Developed novel reconstruction algorithm → magnetic vector potential in 3D
  • Requires no prior-knowledge (e.g. sample periodicity)
  • Robust against low signal-to-noise datasets
  • Enables independent electrostatic/magnetic regularizations
• Demonstrated algorithm utility in simulated 3D
  • Enables atomic-scale imaging of antiferromagnetic order

Outlook

• Demonstrate technique experimentally → NiO and $\alpha \mathrm{Fe}_2\mathrm{O}_3$ crystals
• Investigate resolution limits set by turning objective lens off (Lorentz STEM)