Paul van Gerven
3 March

A new imaging technique being developed by TU Delft and ASML, augmenting an atomic force microscope with ultrahigh-frequency ultrasound functionality, could reveal sub-surface structures with single-digit nanometer resolution, making it an excellent candidate to expand the semiconductor industry’s metrology toolkit.

As chip structures keep shrinking, there’s less room for manufacturing errors with every generation. Only through a complex interplay of metrology, number crunching and real-time correction mechanisms are chipmakers able to keep yields up. Naturally, this is big business, one that ASML has embraced enthusiastically. It’s, therefore, no surprise that the Veldhoven-based equipment manufacturer has taken an interest in a new imaging technique that, unlike all other non-destructive imaging techniques, promises to be capable of peaking a few micron underneath surfaces.

The concept, which is like performing an ultrasound on the nanoscale, was conceived by Gerard Verbiest, currently an assistant professor at Delft University of Technology (TU Delft). “Working on sub-surface atomic force microscopy for my doctoral thesis, I realized that combining AFM with ultrasound should yield a much better way of mapping 3D nanostructures,” he says.

PU AFM 1
Gerard Verbiest next to his setup. Credit: TU Delft

Over the past year, Verbiest’s group has successfully added ultrasound functionality to an AFM and is currently exploring its potential. ASML is providing financial as well as technical support for the project, in return for ownership of the intellectual property. The company might, at some point, turn the technique into an industrial application, although a spokesperson wouldn’t elaborate on that.

Deflection

Ultrasound systems used for checking the health and development of unborn babies typically utilize frequencies of 10-20 MHz, corresponding to a wavelength of about a millimeter. “We need much higher frequencies to hone in on nanometer resolution – high up in the gigahertz range, in fact,” says Verbiest. That’s far from trivial, however, as piezoelectric transducers, currently the state-of-the-art method to generate ultrasonic waves, aren’t able to reach frequencies that high.

Verbiest found an alternative in picosecond ultrasonics (PU), a technique in which ultrahigh-frequency ultrasound is generated using ultrashort light pulses. After firing a ‘pump’ laser pulse at a substrate, it quickly heats up and expands. The strain caused by this process launches an ultrasonic pulse that propagates through the substrate until it’s (partially) reflected at an interface. The echo (or echoes) can be detected at the surface with a ‘probe’ pulse of light. Typically, it’s possible to image structures up to several micrometers underneath the surface.

“PU is already being used for imaging biological cells and semiconductor heterostructures, among other things. We expect that integrating it with AFM will increase the spatial resolution to the single-digit nanometer range. Ultimately, we hope to scan a surface with an AFM while simultaneously performing PU measurements, thus imaging both the surface as well as what’s beneath it. That’s still a long way off, however.”

From a technical perspective, an AFM instrument is relatively easily modified to add PU functionality. In its basic mode of operation, a sharp tip mounted on a cantilever is brought into close proximity with a surface. Forces between the tip and sample ‘pull’ on the cantilever, the deflection of which is measured using the reflection of a laser aimed at the tip-end of the cantilever. “Standard AFM tips are made of silicon coated with metal. We use this layer to generate the ultrasonic pulses as well as to detect the echoes, which travel back through the tip into the metal.”

PU AFM 2
Credit: TU Delft

Alignment

With the setup finished, Verbiest and colleagues can now demonstrate a proof-of-principle of their technique: turning the cacophony of echoes into useful information. The TU Delft researcher doesn’t expect that to be a major hurdle, however. After all, sound has been used ‘to see’ for decades already. Not just for checking up on babies, but also for sonar and finding oil wells, among other things. “Initially, we keep the tip in one place and measure simple samples of known structure to relate measurements to structural information. From there, we can work our way up to more complex samples and measurements.”

In parallel, Verbiest’s group will see how far they can take the technique. “Many questions remain unanswered. What resolution will we be able to attain? At what signal-to-noise ratio? How deep can we penetrate a material? This is a very interesting mix of fundamental research and engineering. For example, I expect that the cantilever and tip design will have a major impact on how much sound we can send into the material and how well we can detect it. The performance will also likely vary for different materials – we’d like to understand that.”

All in all, a lot of work still needs to be done before PU-AFM finds its way to commercial applications. ASML won’t say what it has in mind (and Verbiest isn’t allowed to), but it’s not hard to see how the ability to map the 3D structure of a non-transparent sample in a non-destructive way could be used for the alignment of chip layers, for example. Or for finding out the nature of a defect that has been spotted during IC manufacturing, so the appropriate knobs can be turned to correct the error.

Verbiest can divulge some applications outside chip manufacturing. “You could use it in cell biology to make a detailed 3D image of a single living cell, for example of the way mitochondria are folded in a cell. And in materials science, you could use it for research into heat transport in 2D materials such as graphene, or to characterize the interactions between stacked layers of 2D materials.” With the latter, we’re back to electronics, as stacked 2D materials are being investigated as a successor to silicon as chips’ base material.