Could you please say the depth you talking about? We successfully use lift out procedure at different angles, you can try. But I can`t see the reason for avoidng curtain efect in this case.

You could grind the surface with abrasive paper under a shallow angle. After this procedure fib your lamella out from an area where your feature is couvered only by a thin layer of residue metal. This worked for me when I wanted to see the interface between two metals.

As an example, say lamella is 20microns wide and close to 20microns deep, the feature of interest is close to the bottom of the lamella (e.g. 16microns deep). When you say you lift out at different angles, are you talking about tilt angles? What kind of angles are you working with? 

Thanks Martin for your suggestion, unfortunately this would be tricky to do in this case. First of all, if our samples had lots of features of interest, this wouldn't have been an issue, typically we have only one or two sites in the entire specimen. We try to prepare the surface that the feature of interest is close, but that's very hard to do with very high precision, so we have to make sections to localize the feature of interest, then we know how far it is from the surface. At this point mechanical polishing would most likely destroy (or at least severely deform) the area we need to analyse.  

Ok, now I understand.

Here is another idea: lift out the lamella with a depth of 20µm and fix it to the copper grid. Then fib the upper 12µm away and put new deposit on the surface. After this you can thin the lamella and use it in TEM.

"when you say you lift out at different angles, are you talking about tilt angles? What kind of angles are you working with? " - we have lifted out lamellas at 0 (horizontal), 10 and 54 degrees (normal to ion beam incidence). In principle we can do it ant any any tilt angles. I am still wondering if this lifting can help you.

The last comment looks good. Two stage-preparation with the last reducing of the lamella size from 12 to let`s say 5 mkm at the copper grid can be useful.

I guess you could also thin it down to electron trasparency the usual way, but tilt the grid a few degrees off for each side of the lamella.

This way you should be able to get a parallel sides lamella, even for a depth of 20 um

To prepare FIB lamellae with a depth of 20µm and with electron transparency (!) is not a trivial issue. One solution would be to use a NanoMill system (if you have access to it). The system allows post-polishing of FIB lamellae with focused low-energy argon ion beam. So, after several preparation steps with NanoMill you will access to a ROI. If you don’t have the NanoMill in your lab, then another solution would be to re-polish as-prepared FIB lamella in FIB from back side (known as back side polishing, also helps to reduce curtaining in FIB) to electron transparency. For this preparation you might need a special FIB grid.

I know it's not a trivial issue... we have a lot of experience with FIB liftout and I'm still looking for better strategies for somewhat special specimen. I'm not a big fan of argon polishing. From what I've seen so far, if there's a small curtain in the beginning, it tends to make it worse. What do you mean by re-polish from back side? We always do polishing from both sides... Do you mean polishing from the bottom of lamella by flipping it? I thought about it, even done this before, but it's a very complicated procedure, especially if one doesn't have a flip stage... 

Alessandro, overtilt and undertilt is used during preparation of all lamellas. But it only goes that far in correcting artifacts (typically not far enough for such a case as ours). For very deep lamellas from polycrystalline materials curtain becomes very significant with use of large angles off normal, simply because polishing on the wide section at the bottom of the lamella is done outside the area with Pt protection. For us it's a problem as well.