TEM instrument time is valuable, and specimen loading is a careful, time-consuming step. After loading, the specimen often needs tens of minutes to several hours to stabilize mechanical and temperature drift for high-quality analysis. To make the most of each TEM session, users typically aim to prepare the largest possible electron-transparent area while keeping the specimen thin enough for reliable results. Larger usable areas are also important when analyzing large material structures, such as grains, when the region of interest cannot be localized precisely by non-TEM methods, or when beam sensitivity and contamination require frequent changes of the analyzed area, for example, after each frame. In practice, a TEM specimen must have a sufficiently large area and a very low thickness, typically around 100 nm or less, to support high-quality imaging and analysis.
During FIB preparation of a TEM specimen, the lamella may begin to polish non-uniformly once bending occurs. If the effect progresses, the thinnest area can be punched through. The images below show SEM views of TEM lamellae prepared from PET (left) and silicon (middle) with comparable width and thickness. The silicon lamella remains stable and uniformly polished, while the PET lamella clearly bends. Bending can also be observed in the FIB image during polishing (right).
Flag-post lamellae are usually more prone to bending because they are attached to the TEM grid on only one side. The risk increases when the lamella must be prepared without a supporting frame, for example to avoid induced stress, prevent shielding during final argon broad ion beam polishing, or enable TEM EDS analysis. In these cases, the free end tends to bend more strongly and can become over-polished. The SEM image below shows a GaAs superlattice lamella where the free end is visibly thinner than the main lamella area, and the platinum protective layer decreases toward the free end.
To improve mechanical stability, many FIB users position the lamella on top of a V-shaped grid finger so it can be attached on both sides, as shown below. This configuration can provide higher stability even without a bottom frame. The trade-off is a higher risk of grid-material redeposition on the lamella during polishing. In addition, two-sided attachments usually take more preparation time, and the quality or speed of attachment may differ from side to side depending on the GIS position and local gas flow.
As a general guideline, keeping a supporting frame around the lamella helps minimize bending and increases mechanical strength. A simple analogy is an A4 sheet of paper: unsupported, it bends easily; held on both sides, it becomes much more stable. The same principle applies to a canvas stretched over a frame. During lamella preparation, start polishing with an offset from the edge to create a window in the bulk material, and keep at least some frame at the bottom whenever possible. As the lamella becomes thinner, the polished window should become narrower. If the window remains too wide, the lamella is more likely to bend. In the example below, the left image shows a silicon lamella after 30 kV polishing, with a thickness of approximately 150 nm and a 2 µm window. The right image shows the same lamella after 5 keV polishing, with an approximately 80 nm thickness and a thinned window reduced to 1.5 µm.
For silicon, the following practical correlation between lamella thickness and maximum stable window width is recommended as a starting point: 150 nm – 5 µm; 50 nm – 3.5 µm; 30 nm – 2 µm; and 10 nm – 0.5 to 1 µm. Experienced users may achieve slightly wider widths, but these values provide a useful reference for FIB users building process confidence. Hard materials, including hard metals and alloys, sapphire, and diamond, can typically support larger widths at the same thickness as silicon; for example, a 150 nm steel lamella may reach a width of approximately 10 µm. Softer materials, such as polymers or composites, usually require narrower windows; for example, a 150 nm lamella may be limited to approximately 3 µm, with even smaller widths required at lower thicknesses.
Read more: Practical Tip | Lamella Preparation with Precise Thickness Estimation
Written by Maksym Klymov
Head of Applications Department, Tescan