Does SEM Have a Future?

CD-SEM's importance to metrology cannot be overrated. However, it seems that current tools are restricted to running at energies between 1 keV and a few hundred electron volts, because that is the energy range found to give the best beam interactions, while minimizing charging and limiting resist shrinkage.

With structures hundreds of nanometers in size, this choice did not matter. However, now they are in the tens of nanometers and getting smaller, which is causing problems. For example, low-energy electron wavelengths are large at a few hundred volts; their wavelength is a significant fraction of an angstrom. Consequently, diffraction effects must be controlled or beam spot size is affected. Chromatic aberration resulting from the natural energy spread of the electron source is even more important. Here, electrons of different energies are focused to different planes. In the <3 keV range, chromatic aberration's effect on spot size is critical.

To minimize diffraction, the beam's convergence angle must be kept large; however, to reduce chromatic aberration effects, it must be small. The best accommodation between these competing requirements limits resolution to a few nanometers at ~1 keV. If the electron source's energy spread is reduced, resolution is enhanced. Some new sources are laser-driven and have a spread about one-tenth of that of a conventional field-emission gun. But even that benefit is probably insufficient to justify the engineering effort.

The void between layers in this semiconductor device cross-section is clearly imaged by a helium ion microscope. (Source: Carl Zeiss SMT)

Aberration correction is another obvious option. Aberration correctors for SEMs that can set the aberration to be either zero, positive or negative are available. This enables the same options as with a glass optical system to achieve the theoretical limiting performance. On paper, this looks promising, because it should be possible to simultaneously reduce the probe size to <1 nm at 1 keV and get more beam current. However, aberration correctors are complex devices with 48 or 64 active optical elements, all of which must be simultaneously optimized, requiring considerable computing power to maintain performance. An even more significant problem is that in an aberration-corrected microscope, the depth of field is exceedingly small (1 or 2 nm). Because small-scale structures have high aspect ratios, it is impossible to get an in-focus image of anything beyond a very thin slice.

An energy increase is another possibility. At higher energies, SEM performance improves without significant limits. The downside is a high-energy electron beam (e-beam) may increase damage.

Prof. David C. Joy of the University of Tennessee (Knoxville, Tenn.) and the Center for NanoPhase Materials Science at Oak Ridge National Laboratories (Oak Ridge, Tenn.), believes this is misleading. "On the basis of measurements we performed during a project for Sematech, we determined that under proper conditions, radiation damage is manageable, particularly since the standard technique is not to look at a working structure, but at something in the kerf area."

This is a promising scenario because it does not require any new technology, making it possible to immediately build a microscope that meets required parameters and works at 50 or 100 keV instead of 1 keV or 100 eV. Work done at IBM (Yorktown Heights, N.Y.) with semiconductor devices using SEM energies up to 400 keV has demonstrated depth sectioning through the device, impossible at lower energies, while maintaining a resolution level that is otherwise conventionally unapproachable.

However, Joy thinks that the most exciting option may be abandoning electrons for helium ions. "Helium ions have a very short wavelength, so there is no diffraction limit; their range is much smaller than that of an electron beam, and so can be employed at significantly higher energies which makes the source brighter," he said. "The secondary electron yield generated by ions is at least a factor of 10 higher than a corresponding secondary electrons signal generated. The signal-to-noise ratio is better and, on the basis of some modeling we have been doing and some initial data we produced, secondary electron escape depth — the path secondary electrons travel — seems lower in the case of ion-induced secondaries than for electron-induced ones. The capability to determine the position of any edge is improved by three to four times."

This means it should be possible to push ion-induced secondary electron imaging below the nanometer scale mark. Better yet, ion technology has a long way to go before it reaches the situation that the CD-SEM is now in, where it is limited by fundamental parameters.

Find more information on inspection, measurement and test.