LabForge recipe report · client & job ID redacted
Every field below is real and complete — the exact setpoint values are obscured, like an unlisted number. The sharp recipe, calculated for your chamber, ships with a Single Recipe — $599.
Develop an amorphous-carbon Zernike-type phase plate to correct aberration-induced phase-contrast loss in a TEM electron column.
The complete deliverable — nothing summarized
A LabForge recipe is a full, run-ready specification. This one contains all of the following, and so will yours:
route selection the candidate process + why it was chosen mechanism the governing equation + every derived value process spec every step, every setpoint — pressures, rates, powers, temperatures, times, tolerances equipment the exact tools the route requires expected result what the finished film does, quantified QC acceptance measurable pass/fail criteria failure modes where the process will drift, and the guard revision log the engine's own self-corrections, shown stated limits what the route does NOT do — plainly
Amorphous-carbon Zernike thin-film phase plate, back-focal-plane placement
The unscattered central beam passes through a FIB-milled center hole; scattered beams traverse the carbon film and acquire a π/2 phase shift, restoring low-spatial-frequency phase contrast.
mechanism Δφ = σ(E) · V0 · t target phase π/2 (1.5708 rad) σ (200 kV) 0.00729 rad/(V·nm) σ (300 kV) 0.00653 rad/(V·nm) V0 (a-C) 7.8 ± 0.6 V (mean inner potential) t (π/2) 30.9 nm @ 300 kV · 27.6 nm @ 200 kV working band 27–31 nm across 200–300 kV calibration V0·t confirmed per run by EELS log-ratio
The engine caught and corrected its own first draft
Before this recipe shipped, the engine ran its own numbers back through the physics — and flagged three corrections:
- Re-anchored the relativistic interaction constant σ — a one-voltage-step ladder error worth 10–15% drift in the phase calculation.
- Corrected the π/2 target thickness from 15 nm to 30.9 / 27.6 nm — which then reproduced the canonical published value (~31 nm carbon at 300 kV; Danev & Nagayama, Ultramicroscopy 88, 2001) exactly.
- Replaced a post-FIB O₂-plasma clean with a vacuum bake — the plasma step would have etched the ~30 nm carbon film it was meant to clean.
The engine self-corrected — and its corrected answer landed on a value published in the literature 25 years ago, without being told. That is the difference between derived physics and looked-up numbers.
Six steps — substrate prep to QC, every setpoint
Substrate preparation
Freshly cleaved mica carrier; the film is later floated onto a molybdenum grid. Plasma-clean the bare grid only — never the finished film (step 5).
carrier mica, ASTM V2 grade, freshly cleaved grid molybdenum, 300 mesh solvent clean acetone / IPA, ultrasonic, 5 min plasma (grid) Ar/O₂, 20 W, 2 min — bare grid only base pressure 8 × 10⁻⁷ Torr
Carbon film — e-beam evaporation
The load-bearing step. Recalibrate the QCM tooling factor against EELS log-ratio on a witness sample before the run — mean inner potential and film density co-vary, so the tooling factor is confirmed, not assumed.
target thickness 30.9 nm @ 300 kV · 27.6 nm @ 200 kV thickness band 27–31 nm (voltage-dependent) deposition rate 0.03 nm/s chamber pressure 5 × 10⁻⁷ Torr source pyrolytic graphite, 99.999% throw distance 300 mm substrate motion static (no rotation) substrate temp 293 K QCM density 2.0 g/cm³ (recalibrate tooling vs EELS)
Float-off & grid transfer
Release the carbon film from the mica onto a clean water surface and transfer it to the Mo grid. Handle wet-to-dry to avoid tearing the ~30 nm membrane.
FIB center-hole milling
Mill the central aperture that passes the unscattered beam. Milling time scales with the corrected (thicker) film.
ion source Ga⁺ @ 30 kV beam current 10 pA hole diameter 0.7 µm ± 0.05 milling time ~16 s (est., scales with thickness)
Contamination control — vacuum bake
No plasma on the finished film — O₂ would etch the amorphous carbon and drift the phase shift out of tolerance. Bake for contamination control, then operate the plate heated in-column.
vacuum bake 373–473 K · 30–60 min plasma on film none — O₂ etches a-C in-column op. heated holder, ~473–573 K storage high-vacuum desiccator, < 1 × 10⁻³ Torr
Characterization & QC
The measurements that confirm the plate meets spec before it goes into the column.
phase shift accept 1.5708 rad ± 10% thickness unif. ± 2 nm film thickness EELS log-ratio map (calibrates V0·t) phase verify Thon-ring diffractogram — graphene / catalase hole diameter SEM/TEM, 0.7 µm ± 0.05
The exact tool set this route needs
deposition e-beam PVD system with QCM rate monitor film transfer float-off station aperture FIB-SEM dual beam (Ga⁺) metrology / QC TEM (EELS + diffractogram capability) pre-clean plasma cleaner — pre-deposition, bare grid only
What the finished plate does
Installed at the back focal plane, the plate passes the unscattered central beam through the FIB hole untouched, while the scattered beams pick up a π/2 (1.5708 rad) phase shift traversing the carbon film. The result is restored low-spatial-frequency phase contrast — the weak-phase detail an aberrated, defocused column otherwise loses — at the ± 10% phase tolerance set in QC.
Where this process drifts — and the guard
O₂ plasma on the finished film
Etches the ~30 nm carbon, thins it by multiple nm, and drifts the phase shift out of the ± 10% window. Guard: plasma only on the bare grid (step 1); vacuum bake — not plasma — for the finished film (step 5).
Contamination & charging drift in-column
Beam-induced carbon build-up and charging shift the phase over a session. Guard: operate the plate heated (~473–573 K) and store under high vacuum between uses.
QCM tooling-factor error
Assumed density mis-reads deposited thickness, and V0·t misses π/2. Guard: recalibrate the tooling factor against an EELS log-ratio witness sample before the run (step 2).
Stated limitation of this route
This plate compensates the contrast-transfer consequence of aberrations and defocus. It does not null the Cs/Cc coefficients themselves — that requires a multipole corrector or an adaptive-field device. Every LabForge recipe states its limits this plainly.