Sculpting with Ions 1.1.0 · Cryo-FIB

Cryo-FIB

Sculpting with Ions now prepares frozen cells for cryo-electron tomography.

Live: https://advanced-em.github.io/FIB-view/?mode=cryo

New

• Cryo stage (−196 to −100 °C) with a 45° pre-tilted shuttle, copper cooling braids and an anticontaminator, shown in the 3D instrument with their own instrument cards.
• Two frozen specimens: yeast cells plunge-frozen in vitreous ice on a holey-carbon gold grid, and high-pressure-frozen yeast for cryo slice & view (SE contrast of walls, nuclei, vacuoles and lipid in ice).
• Cryo lamella mode (key \): sputter coat, cold GIS organometallic Pt, milling angle (stage tilt − 7°), micro-expansion joints, milling passes at the current and target you choose, and a final coat. The true slab is tracked: thickness and t/λ, the damaged layer on each face (Ga⁺ ~42 nm, Ar⁺ ~31 nm, Xe⁺ ~18 nm at 30 kV), devitrification from beam heating or a warm stage, ice contamination growing in real time, bending without relief cuts, and a cryo-ET verdict.
• Physics: vitreous ice and frozen cells mill ~4× faster than silicon and charge above ~1.6 kV until coated; the cold GIS layer mills ~10× slower than ice; Ga⁺ 30 keV reaches ~52 nm in ice (binary-collision Monte Carlo).
• Training task Cryo lamella: 200 nm for cryo-ET with the true thickness hidden, and two experiments: A cryo lamella from a frozen cell and Gallium vs argon in ice.
• Voice and AI: "cryo lamella", "sputter coat", "milling angle 12", "mill to 800 nm at 300 pA", "anticontaminator on"; the analysis cryo_quality.
• Python: fib.cryo(...) and new settings (stage_temp_C, anticontaminator, milling_angle_deg, relief_cuts); example cryo_lamella.py (Ga vs Ar vs Xe).
• MCP: cryo_lamella (headless recipe → true result) and exhibit_cryo (drive a live exhibit).
• Guide: SOP-03 Cryo lamella for cryo-ET (12 pages, attached), with figures and numbers computed by the exhibit.

Tests

npm test (physics, 13 modes × 7 specimens, training tasks incl. cryo, cryo failure modes, grammar), the MCP tests and the Python tests pass.

Sculpting with Ions 1.0.0

Live exhibit: https://advanced-em.github.io/FIB-view/

The first release of Sculpting with Ions, the dual-beam sister exhibit of Seeing with Electrons: a focused ion beam / scanning electron microscope in the browser, where every image, trench and deposit is computed from the physics.

The instrument

• Ion beams: gallium from a liquid-metal source (2.5 nm probes) and xenon, argon, nitrogen and oxygen from a plasma source (up to 2.5 µA). Probe size, sputter yields, ranges and damage follow measured values and a binary-collision Monte Carlo (Ga 30 keV in Si: range ~27 nm, yield ~2).
• Twelve modes: SE, in-lens, BSE, EBSD, TKD and EDS with the electron beam; ion imaging, milling and gas-assisted deposition with the ion beam; slice & view; an eight-step TEM lamella lift-out; and Monte Carlo of electrons or ions.
• Both beams see the milled surface: the 52° SEM view looks into a trench and sees the cross-section face (Pt cap, oxide, copper wiring, tungsten vias, silicon), while ion imaging shows channelling contrast and erodes what it looks at.
• Experiments: cross-section a chip, gallium vs xenon vs oxygen on a battery cathode, channelling contrast, charging, nanograins (EBSD vs TKD), and a fly-through.

Training, AI and scripting

• Training tasks with a hidden answer: focus and stigmate the SEM; focus the ion beam without destroying the specimen; find the eucentric height; thin a lamella to 100 nm. With consent, sessions are recorded anonymously and can be shared with the research dataset on Hugging Face.
• AI mode: a command bar and push-to-talk voice ("switch to xenon", "mill a cross-section", "thin to 100 nm", "what is the sputter yield?"), and analyses checked against the simulator's truth.
• Agents: tools/fibview-mcp lets an AI agent run the twin headlessly or drive an open exhibit.
• Python: the fibview package and an in-browser IDE (Pyodide).
• Console: drive it with the TEM-view Console, any game controller, or the on-screen panel (V).

Guides

• SOP-01 · TEM lamella lift-out (11 pages)
• SOP-02 · Python scripting (8 pages)

Simulation only: nothing here controls a real instrument. All rights reserved while the licenses are decided (see LICENSE).