TEM specimen preparation by Au + , Ga + , Si + and Si ++ focused ion beams for off-axis electron holography G. BenAssayag a, * , P. Salles a , F. Bertin b , D. Cooper b a CEMES-CNRS/Toulouse University, 29 Rue J. Marvig, 31055-Toulouse, France b CEA, LETI, Minatec, 17 Rue des Martyrs, F38054 Grenoble, France article info Article history: Received 14 September 2009 Received in revised form 30 October 2009 Accepted 4 November 2009 Available online 14 November 2009 Keywords: Focused ion beam Holography TEM preparation abstract Silicon specimens containing p–n junctions have been prepared by focused ion beam milling using gal- lium, single or double charged silicon, and gold focused ion beams. These specimens have been examined before and after annealing for Si prepared samples using off-axis electron holography. The total thickness of the electrically ‘inactive’ layer and of the amorphous region have been measured for the Au, Ga, Si + and Si ++ ions, respectively. We also show that the thickness of this damaged layer can be predicted using col- lision simulations. Ó 2009 Elsevier B.V. All rights reserved. 1. Introduction There is a need of the semiconductor industry for mapping the dopant concentration with nm scale resolution [1]. Focused ion beam (FIB) milling using Ga + ions has been used extensively to fab- ricate membranes for examination by TEM. It is well-known that FIB milling introduces artifacts in semiconductor specimens such as amorphous surface layers, an electrically ‘inactive’ thickness, high level Ga + impurities, and thickness variations caused by dif- ferential milling for heterogeneous samples. Usually the presence of the amorphous layers is not critical and can be managed. For off-axis electron holography, it is essential to have parallel-sided membranes prepared by FIB with nanometer scale site specificity to ease the interpretation. But the presence of the electrically ‘inac- tive’ thickness in the specimens is more problematic and arises from the ion beam induced defects in the crystalline regions which can deactivate the dopants resulting in a measured potential that is much less than predicted by theory [2,6]. Although the electrically ‘inactive’ thickness can be almost eliminated by annealing the specimens at low temperature, it has been shown that the Ga ions which are dopants in Si are implanted in quantities that are similar to typical values of the dopant concentration. To be able to mea- sure the doping level of real devices after FIB thinning it is neces- sary to switch from gallium to other ion species which do not act as dopants of silicon. In this work we used different species such as Si and Au to prepare the TEM samples. Si + , Si ++ , Au + and Ga + species at an accelerating voltage of 30 kV are tested [7]. The amorphous and the ‘‘electrically inactive” zones generated by the ion milling are estimated for the different ion species either by standard transmis- sion electron microscopy or electron holography of the p–n junc- tions acquired on a FEI Titan microscope operated at 200 kV. 2. Experimental 2.1. Mass filtered focused ion beam column The focused ion beam system used during this work for TEM thinning is derived from a standard XB1540 Zeiss CrossBeam. In this equipment the standard focused ion column is replaced by an OrsayPhysics column including an E Â B (Wien) filter located between the condenser and objective lenses. In such a filter the tra- jectory of an ion crossing the filter having a velocity equal to E di- vided by B (where E and B, are respectively the electric and magnetic fields) is straight. The other species are deviated from the optical axis according to their different mass and a mass sepa- rating aperture located below the filter can stop the undesirable ions. The best filtered probe resolution is obtained when the cross- over is located in the middle of the mass filter as the best mass res- olution is achieved for a crossover being focused in the mass aperture plane. This mass filter has a resolution of roughly 100 which allows a sufficient mass separation for most of suitable ion species. In particular it is possible to separate the different silicon isotopes (28, 29 and 30). The final filtered FIB resolution is slightly degraded (15 nm FWHM for Si + ion probe) compared to a standard FIB column on the Zeiss XB1540. Fig. 1 is a schematic diagram of 0167-9317/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.mee.2009.11.013 * Corresponding author. Tel.: +33 5 62 25 79 24. E-mail address: benassay@cemes.fr (G. BenAssayag). Microelectronic Engineering 87 (2010) 1579–1582 Contents lists available at ScienceDirect Microelectronic Engineering journal homepage: www.elsevier.com/locate/mee