E19 · Lift-out and thinning

FIB Lamella: From Bulk to TEM Grid

How do you lift out a thin specimen without losing its support or useful material? Attach the needle before cutting the last bridge, and attach the lamella to its grid before removing the needle. Leave enough material for a low-energy finishing pass: two specimens of equal thickness can retain different amounts of unaffected interior.

Prepare two trenches, attach a needle before cutting the last bridge, and drag the supported lamella toward a TEM grid. Fix it to the grid before separating the needle. Then thin both faces and compare a gentle finishing pass with a damaging one.

Current support $\text{—}$

Modeled lamella thickness $\text{—}$

Modeled unaffected core $\text{—}$

Affected layer on each face $\text{—}$

More measurements and target

Retained Ga from face milling $\text{—}$

Retained Xe from face milling $\text{—}$

Mean implanted atomic fraction $\text{—}$

Last committed milling energy $\text{—}$

Trench and face-milling time $\text{—}$

Trench and face-milling ion number $\text{—}$

Mean trench depth $\text{—}$

Trench depth RMS $\text{—}$

Mean central substrate loss $\text{—}$

Mean assigned damage-length proxy $\text{—}$

Expected incident ion number $\text{—}$

Gross removed atom number $\text{—}$

Returned atom number $\text{—}$

Escaped atom number $\text{—}$

Expected writing duration $\text{—}$

Selected-cell net depth $\text{—}$

Experiment target $\text{—}$

Simulated examples; each plot identifies its data source.

Incident energy

E

30\,\mathrm{keV}

Needle distance from the attachment point

x_n

-600\,\mathrm{nm}

Lifted lamella travel toward the grid

x_g

2400\,\mathrm{nm}

Removal from each face in the next pass

d

30\,\mathrm{nm}

Advanced settings

Writing current

I

100\,\mathrm{pA}

Incident dose per specimen area

F

4000\,\mathrm{\frac{ions}{nm^2}}

Assigned yield stripe amplitude

a

30\,\mathrm{\%}

Assumed rocking averaging

R

0\,\mathrm{\%}

Ion source: 0 Ga, 1 Xe

0

Incidence from initial surface normal

\theta

0\,\mathrm{{}^\circ}

Assigned returned-material fraction

f_r

20\,\mathrm{\%}

Central silicon-equivalent sacrificial cap

t_c

0\,\mathrm{nm}

Selected cell index

k

1164

Dose/depth display gain only

G

1

Learning goal and model limits

Predict · write · inspect · explain

Remove material without ruining the specimen

Make clean trenches, keep the lamella supported while lifting it onto the grid, then detach the needle. Thin both faces and finish with a low-energy pass. Aim for 60–150 nm total thickness, at most 5 nm of affected surface on either face, and at least 85 percent unaffected core. These are targets for the assigned model.

Model assumptions

The face-milling history also tracks separate Ga and Xe implantation profiles in an assigned silicon-equivalent host. Exponential stopping ranges and retained fractions are teaching parameters; sputtering removes previously implanted atoms and each pass adds new ions. The profile starts at the first face pass and excludes implantation from trenching or the support steps. Its composition is a calculated expectation, separate from the affected-layer proxy, not a measured concentration. This teaching workflow follows a supported lamella from trench preparation through undercutting, needle attachment, release, transfer, grid attachment, needle separation and two-face thinning. Each operation keeps its settings. The lamella starts 600 nm thick and 1500 nm long; its height comes from the committed trench depth. Both faces lose the chosen thickness in a pass. The removed volume and an assigned silicon-like atomic density determine the modeled ion number and milling time. Ga/Xe sputter yields, the trench stripes, redeposition and the affected surface layer are assigned teaching laws. The surface layer is a damage proxy, not a measured amorphous thickness or implanted-ion profile. Contact tolerances, ideal welds and a fixed raised transfer path replace forces and collision dynamics. Undercutting and weld deposition are state changes, not dose-calibrated processes; reported ion totals cover trench and face milling only. There is no calibrated material transport, evolving shadowing, thermal damage, fracture mechanics, real SEM acquisition or instrument recipe. The inspection rays are illustrative. Height, lateral distance, thickness in the data card and hardware use separate schematic scales.

Start with the essentials

Focus question: How do you lift out a thin specimen without losing its support or useful material?
One-sentence intuition: Keep one support in place until the next connection is made.

Core mathematical model

Remove equal thicknesses from two faces

t_{j+1}=\max(0,t_j-2d_j)

The setting is removal from each face. If the two cuts meet, the lamella is lost. The initial modeled thickness is 600 nm.

Count the removed volume

\begin{aligned}V_j&=L H(t_j-t_{j+1})\\N_j&=nV_j/Y_j\\T_j&=eN_j/I_j\end{aligned}

Length and height stay fixed during face milling. The density and yield are assigned. This counts the ion charge for normal-incidence face passes; the trench-angle setting does not change that separate geometry.

Interior left between affected surfaces

t_{\rm core}=\max(0,t-2\delta)

The two surface layers use one assigned depth. Removing a face can clear old damage and create a new layer. The model does not infer a measured implantation profile.

Common difficulties

Teaching parameters

Typical misconceptionThe displayed settings are a preparation recipe for my material.

Better mental modelThe face-milling history also tracks separate Ga and Xe implantation profiles in an assigned silicon-equivalent host. Exponential stopping ranges and retained fractions are teaching parameters; sputtering removes previously implanted atoms and each pass adds new ions. The profile starts at the first face pass and excludes implantation from trenching or the support steps. Its composition is a calculated expectation, separate from the affected-layer proxy, not a measured concentration. This teaching workflow follows a supported lamella from trench preparation through undercutting, needle attachment, release, transfer, grid attachment, needle separation and two-face thinning. Each operation keeps its settings. The lamella starts 600 nm thick and 1500 nm long; its height comes from the committed trench depth. Both faces lose the chosen thickness in a pass. The removed volume and an assigned silicon-like atomic density determine the modeled ion number and milling time. Ga/Xe sputter yields, the trench stripes, redeposition and the affected surface layer are assigned teaching laws. The surface layer is a damage proxy, not a measured amorphous thickness or implanted-ion profile. Contact tolerances, ideal welds and a fixed raised transfer path replace forces and collision dynamics. Undercutting and weld deposition are state changes, not dose-calibrated processes; reported ion totals cover trench and face milling only. There is no calibrated material transport, evolving shadowing, thermal damage, fracture mechanics, real SEM acquisition or instrument recipe. The inspection rays are illustrative. Height, lateral distance, thickness in the data card and hardware use separate schematic scales.

Run the experiment

01
Compare damage and implanted ions
After mounting and thinning, inspect the implantation profile. Change the ion source in advanced controls and commit a small final pass. Compare retained Ga and Xe with the affected-layer readout.
What to observe: A new source adds its own species while removing some atoms from the earlier pass. A smaller affected layer does not by itself specify the remaining composition.
02
See why a thin slice is not enough
Compare Thin and mounted with Thin but surface-damaged. Check the affected layer and the unaffected core.
What to observe: Equal total thickness can leave quite different amounts of useful interior.
03
Prepare a new specimen
Choose New specimen. Set the trench dose and protection in Advanced settings, mill the trenches, then undercut while retaining one bridge.
What to observe: A shallow cut cannot release this lamella. The retained bridge still carries it.
04
Let the needle take over
Drag the gold needle handle to the attachment point and attach it. Cut the last bridge, then lift clear.
What to observe: The model refuses release before attachment and motion before release.
05
Transfer the support to the grid
Drag the supported lamella to the grid stop, lower it and weld it to the post. Separate the needle only after this connection is made.
What to observe: Touching the grid is different from being attached to it.
06
Choose the last face passes
Thin both faces, leaving material for a low-energy finish. Check the 60–150 nm thickness target and the 5 nm surface-layer target.
What to observe: A later slider change cannot alter an earlier pass. Milling through either face is irreversible for that specimen.

Core mathematical model

Remove equal thicknesses from two faces

Count the removed volume

Interior left between affected surfaces

Common difficulties

Teaching parameters

Run the experiment

Compare damage and implanted ions

See why a thin slice is not enough

Prepare a new specimen

Let the needle take over

Transfer the support to the grid

Choose the last face passes