What is simulated#

  • Column: behaves like DE-TEM-Channel’s JEOL Dummy.

    • JEOL function modes, TEM/STEM, and magnification and camera-length detent ladders.

    • Stage motion at realistic speeds, with a busy status while moving.

    • Spot size and probe mode set the beam current; intensity sets the illuminated area.

    • Defocus, astigmatism, beam/image shift, beam tilt and diffraction shift.

    • Axial aberrations in CEOS notation (C1 to C5), with an optional probe and/or image corrector (see Probe and image corrector).

    • Beam blank, screen and column valves.

  • Specimen: every holder and preparation of DE-Server’s former VirtualSpecimen, deterministic from a seed.

    • Holders: mesh and waffle grids, FIB lift-out lamellae, in-situ MEMS heating chip.

    • Preparations: nanoparticles, proteins in ice, thin films, bulk structures (strain, p-n junction, precipitates).

    • Every droplet or grain is a real crystal with a full 3D orientation (orix): random for particles, a fibre texture for thin films, near-single-crystal for lamellae.

    • In-situ crystallisation driven by the holder’s temperature history.

  • Image formation:

    • TEM: wave-optical, with a CTF: Thon rings, astigmatism, and the beam-tilt × defocus shift.

    • SAED, NBD and CBED: kinematic diffraction from diffsims structure factors. Stage tilt moves grains through Bragg conditions in every pattern and in bright-field contrast, and disks have the column’s convergence angle.

    • STEM: coherent 4D-STEM (probe wavefunction × transmission function, partial coherence) suitable for ptychography and tcBF, a fast kinematic model for large patterns, and virtual detectors.

    • Dose is physical: electrons per pixel per second from beam current and illuminated area.

    • Ground truth for scoring automation: noiseless flux, orientation maps (twin.ground_truth, an orix CrystalMap), and the ptychography object and probe.

  • Detector: every DE camera model (DE16/64, Celeritas, Apollo, …).

    • Poisson electrons, charge spread and gain non-uniformity.

    • Dark offset, fixed pattern, read noise and dark current.

    • Hot/dead pixels and bad columns, saturation, binning and counting.

    • Dark and gain references are therefore necessary, and they work.

  • Holder: a simulated DENS-style heating chip (optionally biasing), or a follower of a real DENS Impulse holder or DENS’s own simulator.